Gene Regulatory Networks [1 ed.] 9780128131800, 9780128131817

Table of contents :
Contributors
Preface
A gene regulatory network for cell fate specification in Ciona embryos
Introduction
Basic properties for the gene regulatory network in early ascidian embryos
Regulatory genes
Initial conditions for the gene regulatory network
Outputs of the gene regulatory network
Potentially unique properties of the gene regulatory network in early ascidian embryos
The gene regulatory network accounts for changes of gene expression in space and time
The initial set up in the 16-cell embryo by maternal factors
Specific gene expression in the 32-cell embryo
Specific gene expression at the 64-cell stage and thereafter
Specification of endodermal fate
Fate specification of mesenchymal cells
Fate specification of muscle cells and adult heart precursors
Fate specification of the notochord and the posterior neural plate
Epidermal cells
Cells for the lateral neural plate and its border
Cells for the anterior neural plate and its border
Perspectives
References
Early Xenopus gene regulatory programs, chromatin states, and the role of maternal transcription factors
Introduction
Roles of maternal TFs during germ layer specification
Localization of maternal gene products
Endodermally enriched TFs
Vegt
Otx1
Sox7
Ectodermally enriched TFs
Foxi2
Grhl1
Ubiquitously expressed TFs
Foxh1
Sox3
Pou5f3
Intracellular mediators of signaling pathways
Gdf1/Nodal/Smad2,3
Wnt/TCF/Ctnnb1 (β-catenin)
Other TFs
Enhancers, promoters and chromatin states
Genomic approaches to identifying enhancers
Chromatin states and ZGA
Repressive chromatin marking
Network structure analysis of the early Xenopus GRN
Summary and prospects
Acknowledgments
References
Dynamic and self-regulatory interactions among gene regulatory networks control vertebrate limb bud morpho ...
Introduction
Setting the stage: Molecular control of the limb field positions along the primary body axis
``Pre-patterning´´: Polarizing the limb bud axes upstream of morphogenetic SHH signaling
Establishment of AER-FGF signaling is linked dorso-ventral and proximo-distal axis patterning
The self-regulatory limb bud signaling system coordinately controls antero-posterior and proximo-distal axis patternin ...
Propagation and termination of the self-regulatory limb bud signaling system
A Turing-type mechanism controls the definitive periodic pattern of digit rays
Brief outlook
Acknowledgments
References
Gene regulatory networks during the development of the Drosophila visual system
Introduction
Design principles of the gene regulatory networks (GRNs) in the Drosophila visual system
The eye
The eye disc GRN
The GRN of the morphogenetic furrow
The GRN specifying photoreceptor fate
The GRN that determines photoreceptor terminal features: Ommatidial subtypes
The GRN that determines photoreceptor terminal features: Axon targeting
GRNs in the eye: A brief summary
The GRN that specifies lamina neurons in response to photoreceptor signals
Temporal and spatial GRNs specify medulla neurons
The GRNs leading to the formation of the lobula and lobula plate
Conclusion and perspectives
Acknowledgments
References
Cell fate decisions during the development of the peripheral nervous system in the vertebrate head
Introduction
The preplacodal region: A territory of sense organ and cranial ganglia progenitors
Subdivision of the embryonic ectoderm prior to gastrulation
The neural plate border
Signaling events at the neural plate border
Transcription factor interactions: From neural plate border to neural crest and placode precursors
Segregation of neural crest and sensory placode precursors
Subdividing the placode territory along the anterior-posterior axis
Otic and epibranchial placode formation: Signals and regulatory circuits
FGF signaling induces otic-epibranchial progenitors
Segregation of otic-epibranchial progenitors: Signals and regulatory circuits
Regulatory circuits at otic and epibranchial placode stages
Refining the OEP network: A cis-regulatory perspective
Conclusion
References
A Hox gene regulatory network for hindbrain segmentation
Introduction
A GRN controls segmental Hox expression in the hindbrain
Signaling determines early axial allocation
Wnt
RA
FGF
Spatial subdivision of the hindbrain territory by segmentation genes
Rhombomeric deployment of cell segregation factors
Establishment and maintenance of segmental Hox expression
Hox targets and cofactors
Hox regulatory networks and the neural crest GRN
Conclusions and key questions
Acknowledgments
References
Logical modeling of cell fate specification-Application to T cell commitment
Introduction
Overview of T cell development
Logical modeling of T cell specification
Delineation of the regulatory graph controlling T cell specification
Multilevel nodes
Logical rules
Computation of model stable states in the wild-type situation
Computation of model stable states upon environmental or genetic perturbations
Assessing the reachability of specific expression patterns from specific initial states
Stochastic simulations
Conclusions
Acknowledgments
References
Repressive interactions in gene regulatory networks: When you have no other choice
Introduction
The developing neural tube: A GRN in time and space
Concentration interpretation: More than Gli affinities
Gradient interpretation in time
Cis regulatory elements and the role of repression
From subcircuits to complex dynamics
What can the GRN do for you?
Molecular mechanisms for implementing the GRN cross-repressive interactions
CRE repression
Repression by protein-protein interactions
Repression at the RNA level
Conclusions
Acknowledgments
References
The function of architecture and logic in developmental gene regulatory networks
Introduction
The structure of developmental GRNs in theory and experiment
Logic processing in developmental GRNs
Architecture of circuit modules that are prevalent in developmental GRNs
The connection between circuit architecture and developmental function
Organization of circuit modules within developmental GRNs
Evolution of network architecture
Conclusion
Acknowledgments
References
The evolution of the gene regulatory networks patterning the Drosophila Blastoderm
Introduction
The Drosophila syncytial blastoderm
Axial patterning
Terminal patterning
Gap genes
Pair-rule genes
Segment polarity genes
Conclusions
References
The notochord gene regulatory network in chordate evolution: Conservation and divergence from Ciona to ve ...
The notochord, a synapomorphy of the chordate phylum and a sine qua non of chordate development
Evolutionary history of the notochord
Notochord development in Ciona
Morphogenetic milestones
The molecular blueprint for notochord formation in Ciona
Notochord induction and specification
Invagination and convergent extension
Formation of the notochordal sheath and elongation
Tubulogenesis and disappearance
Time is of the essence: Temporal cis-regulatory control of gene expression in a fast-developing chordate
The GRN underlying notochord formation in Ciona
The main regulatory nodes of the notochord GRN: Brachyury and Foxa2
Drawing edges, connecting nodes: Reconstructing the gene batteries controlled by notochord transcription factors
Reverse-engineering the notochord GRN: Lessons from notochord cis-regulatory modules
Ciona notochord CRMs can be synergistically activated by Foxa.a in cooperation with either Bra or other transcriptio ...
Additional notochord cis-regulatory mechanisms identified in Ciona
Comparative studies of notochord cis-regulatory regions across chordates
A comparative view of the notochord GRN in Ciona and other chordates
Concluding remarks and future perspectives
Acknowledgments
References
The notochord gene regulatory network in chordate evolution: Conservation and divergence from Ciona to ve ...
The notochord, a synapomorphy of the chordate phylum and a sine qua non of chordate development
Evolutionary history of the notochord
Notochord development in Ciona
Morphogenetic milestones
The molecular blueprint for notochord formation in Ciona
Notochord induction and specification
Invagination and convergent extension
Formation of the notochordal sheath and elongation
Tubulogenesis and disappearance
Time is of the essence: Temporal cis-regulatory control of gene expression in a fast-developing chordate
The GRN underlying notochord formation in Ciona
The main regulatory nodes of the notochord GRN: Brachyury and Foxa2
Drawing edges, connecting nodes: Reconstructing the gene batteries controlled by notochord transcription factors
Reverse-engineering the notochord GRN: Lessons from notochord cis-regulatory modules
Ciona notochord CRMs can be synergistically activated by Foxa.a in cooperation with either Bra or other transcriptio ...
Additional notochord cis-regulatory mechanisms identified in Ciona
Comparative studies of notochord cis-regulatory regions across chordates
A comparative view of the notochord GRN in Ciona and other chordates
Concluding remarks and future perspectives
Acknowledgments
References
On the specificity of gene regulatory networks: How does network co-option affect subsequent evolution?
How does one part become different from other parts?
Immediate outcomes of network co-option
Wholesale co-option
Partial network co-option and functionally-divergent network co-option
Co-option resulting in downstream regulatory expression only (aphenotypic co-option)
How do GRNs maintain or recover specificity after network co-option?
Changes in trans
Changes in cis
Regulatory input diversification
Specificity conferred by redundant CREs
The action of selection on co-opted networks over time
Initiating trans change with positive fitness effects
Initiating trans change with detrimental fitness effects
Initiating trans change with selectively neutral effects
Network co-option and the origin and diversification of traits
Concluding remarks
Acknowledgments
Glossary
References
Evolutionary dynamics of gene regulation
Conceptual approaches to understanding the evolution of gene regulation
The origin of metazoan regulatory novelties
The expansion and exploitation of combinatoric complexity
Evolutionary repatterning of gene regulatory networks
Micro- and macroevolutionary changes in gene regulatory networks
Do novel morphologies involve mechanistically distinct pathways?
Towards a general theory of GRN evolution
Concluding remarks
Acknowledgments
References
Further reading

Citation preview

CURRENT TOPICS IN DEVELOPMENTAL BIOLOGY “A meeting-ground for critical review and discussion of developmental processes” A.A. Moscona and Alberto Monroy (Volume 1, 1966)

SERIES EDITOR Paul M. Wassarman Department of Cell, Developmental and Regenerative Biology Icahn School of Medicine at Mount Sinai New York, NY, USA

CURRENT ADVISORY BOARD Blanche Capel Wolfgang Driever Denis Duboule Anne Ephrussi

Susan Mango Philippe Soriano Cliff Tabin Magdalena Zernicka-Goetz

FOUNDING EDITORS A.A. Moscona and Alberto Monroy

FOUNDING ADVISORY BOARD Vincent G. Allfrey Jean Brachet Seymour S. Cohen Bernard D. Davis James D. Ebert Mac V. Edds, Jr.

Dame Honor B. Fell John C. Kendrew S. Spiegelman Hewson W. Swift E.N. Willmer Etienne Wolff

Academic Press is an imprint of Elsevier 50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States 525 B Street, Suite 1650, San Diego, CA 92101, United States The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom 125 London Wall, London, EC2Y 5AS, United Kingdom First edition 2020 Copyright © 2020 Elsevier Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. ISBN: 978-0-12-813180-0 ISSN: 0070-2153 For information on all Academic Press publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Zoe Kruze Editorial Project Manager: Shellie Bryant Production Project Manager: Denny Mansingh Cover Designer: Greg Harris Typeset by SPi Global, India

Contributors James Briscoe The Francis Crick Institute, London, United Kingdom Ailin Leticia Buzzi Centre for Craniofacial and Regenerative Biology, Faculty of Dentistry, Craniofacial and Oral Sciences, King’s College London, London, United Kingdom Elisabetta Cacace European Molecular Biology Laboratory, Genome Biology Unit, Heidelberg, Germany Yen-Chung Chen Department of Biology, New York University, New York, NY, United States Ariel D. Chipman The Department of Ecology, Evolution & Behavior, The Hebrew University of Jerusalem, Edmond J. Safra Campus, Givat Ram, Jerusalem, Israel Jin S. Cho Department of Developmental and Cell Biology, University of California, Irvine, CA, United States Ken W.Y. Cho Department of Developmental and Cell Biology; Center for Complex Biological Systems, University of California, Irvine, CA, United States Samuel Collombet European Molecular Biology Laboratory, Genome Biology Unit, Heidelberg, Germany;  cole Normale Superieure Computational Systems Biology Team, Institut de Biologie de l’E  cole Normale Superieure, Universite PSL, Paris, France (IBENS), CNRS, INSERM, E M. Joaquina Dela´s The Francis Crick Institute, London, United Kingdom Claude Desplan Department of Biology, New York University, New York, NY, United States Anna Di Gregorio Department of Basic Science and Craniofacial Biology, New York University College of Dentistry, New York, NY, United States Douglas H. Erwin Department of Paleobiology, National Museum of Natural History, Washington, DC, United States Robb Krumlauf Stowers Institute for Medical Research, Kansas City, MO; Department of Anatomy and Cell Biology, Kansas University Medical Center, Kansas City, KS, United States

xi

xii

Contributors

Eden McQueen Department of Biological Sciences, University of Pittsburgh, Pittsburgh, PA, United States Kitt D. Paraiso Department of Developmental and Cell Biology; Center for Complex Biological Systems, University of California, Irvine, CA, United States Hugo J. Parker Stowers Institute for Medical Research, Kansas City, MO, United States Isabelle S. Peter Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, United States Mark Rebeiz Department of Biological Sciences, University of Pittsburgh, Pittsburgh, PA, United States Yutaka Satou Department of Zoology, Graduate School of Science, Kyoto University, Kyoto, Japan Andrea Streit Centre for Craniofacial and Regenerative Biology, Faculty of Dentistry, Craniofacial and Oral Sciences, King’s College London, London, United Kingdom Denis Thieffry  cole Normale Superieure Computational Systems Biology Team, Institut de Biologie de l’E  cole Normale Superieure, Universite PSL, Paris, France; (IBENS), CNRS, INSERM, E Cancer Science Institute of Singapore, National University of Singapore, Singapore, Singapore Alexandre Thiery Centre for Craniofacial and Regenerative Biology, Faculty of Dentistry, Craniofacial and Oral Sciences, King’s College London, London, United Kingdom Junseok Yong Department of Developmental and Cell Biology, University of California, Irvine, CA, United States Rolf Zeller Developmental Genetics, Department Biomedicine, University of Basel, Basel, Switzerland Aimee Zuniga Developmental Genetics, Department Biomedicine, University of Basel, Basel, Switzerland

Preface Gene regulatory networks (GRNs) offer an unprecedented view on the genomic control of developmental and evolutionary processes. By regulating gene expression, GRNs control the molecular changes in genome activity that drive development and enable evolution. A lot remains to be learned about how GRNs control developmental mechanisms, and this volume provides just a snapshot of a growing field. Many bits and pieces of regulatory circuits have been discovered, or are currently being investigated, in different animals and particularly also in plants. In many developmental contexts, functional analyses of GRNs are starting to illuminate the control systems that generate causality in the developmental process. This volume focuses in particular on developmental systems where decades of research have provided sufficient tools and insights to generate a causal understanding of the underlying GRNs and the evolution thereof. The articles in this collection discuss current insights into how GRNs operate in various contexts in development and evolution, and also address some of the challenges that lie ahead and that will be solved in coming years as research in this scientific area advances. The first two chapters of this volume deal with the question of how maternal inputs are interpreted by zygotic GRNs to drive the initial specification of cell fates and to control the spatial organization of early embryos. In Chapter 1, Satou presents a comprehensive account of the GRNs that control the earliest cell fate specification in Ciona embryos. The GRN models compiled in this review explain the specification of cell fates throughout the embryo, from the earliest activation of the zygotic genome by maternal transcription factors up to 112 cell stage. Similarly, in Chapter 2, Paraiso et al. discuss the initial activation of the zygotic genome by maternally localized transcription factors in Xenopus embryos, with particular focus on the endoderm GRN. Later in development, the formation of body parts also depends on the interpretation of localized transcription factors and signaling molecules which define embryonic positions along the major body axes. These regulatory inputs then activate body part-specific GRNs that control the specification and organization of cell fates. In Chapter 3, Zuniga and Zeller review how GRNs control the positioning and specification of the major axes and cell fates in the developing limb of the mouse embryo. xiii

xiv

Preface

These GRNs provide an insightful account for how networks of interconnected transcription factors and signaling gradients establish the coordinate system that organizes a newly developing body part. The next three chapters deal with the particular challenges of organizing the developing nervous system. Chapter 4 presents a review of the GRNs controlling the specification and coordination of different layers of the visual system in Drosophila, from the specification of progenitor cells to the terminal differentiation of different types of neurons (Chen and Desplan). A further rapidly growing field focuses on the development of vertebrate sensory neurons and the formation of placodes. Chapter 5 presents a comprehensive review of the GRNs controlling different stages of the specification of placodes, from those defining the progenitors of placodes and neural crest cells to those specifying the diverse types of placodes according to their position along the anterior–posterior axis (Thiery et al.). Organization along the anterior–posterior axis is also crucial in the development of the vertebrate hindbrain. Hox transcription factors play a particularly important role in the control of positional identity of hindbrain rhombomeres, and the regulatory interactions in this GRN have been particularly well explored at the cis-regulatory level, as reviewed in Chapter 6 (Parker and Krumlauf ). Developmental GRNs controlling the specification of body axes and cell fates consist of the interactions between many transcriptional regulators, which complicates not only the experimental analysis of these networks, but also the assessment of their system level functions once experimental insights into the players and interactions have been obtained. In order to assess the behavior of these networks and to determine the contribution of individual parts to the overall function of a GRN, more formal analytic approaches become necessary. Chapters 7–9 address different approaches to the computational modeling of network function and to deciphering the function of circuit structure. Chapter 7 presents a detailed account of an approach to Boolean logic modeling of developmental GRNs (Cacace et al.). On the other hand, continuous modeling has been successfully used to capture the function of a regulatory circuit that forms discrete cell fate domains in response to a signaling gradient. This is particularly well resolved in the vertebrate neural tube, as reviewed by Dela´s and Briscoe in Chapter 8. Furthermore, a comparison of the topology of various GRNs indicates that network structure and regulatory logic have an important function in the system level control of developmental processes, as discussed in Chapter 9 (Peter).

Preface

xv

The evolution of the animal body plan is a challenging topic to address experimentally as well as conceptually. The growing understanding of developmental GRNs provides an important opportunity in this field. By viewing animal evolution as the outcome of change of developmental processes, and therefore of change in developmental GRNs, several evolutionary mechanisms have been discovered that explain evolutionary change as well as conservation. The last four chapters in this volume address this topic. From an experimental perspective, Chapter 10 summarizes recent insights into the GRNs controlling segmentation in different arthropods and shows how these findings can illuminate the possible evolutionary history of this network (Chipman). Similarly, Chapter 11 discusses the evolution of the notochord GRN, starting from the perspective of the Ciona GRN and its relevance to understanding the evolution of the notochord GRN in vertebrates (Di Gregorio). And finally, the last two chapters of this volume address the mechanisms of evolutionary change in GRNs. Chapter 12 discusses the possible consequences of the evolutionary co-option of transcription factors or network subcircuits to novel developmental contexts (McQueen and Rebeiz). Chapter 13 provides a more general overview of evolutionary changes that enabled adaptations as well as macroevolutionary change of the body plan and considers how evolutionary modifications of the regulatory genome have contributed to the generation of morphological novelty (Erwin). Just about 50 years ago, in 1969, Roy Britten and Eric Davidson envisioned a first model for how gene regulation occurs during animal development. Although it would take several decades before this problem became accessible to experimental analysis, this model was nevertheless fundamental for the discovery of GRNs that control genome activity in development. This volume is a celebration of the intellectual, technical, and experimental achievements that have contributed to the current insights into the control systems in so many developmental and evolutionary processes. ISABELLE S. PETER Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, United States

CHAPTER ONE

A gene regulatory network for cell fate specification in Ciona embryos Yutaka Satou* Department of Zoology, Graduate School of Science, Kyoto University, Kyoto, Japan *Corresponding author: e-mail address: [email protected]

Contents 1. Introduction 2. Basic properties for the gene regulatory network in early ascidian embryos 2.1 Regulatory genes 2.2 Initial conditions for the gene regulatory network 2.3 Outputs of the gene regulatory network 2.4 Potentially unique properties of the gene regulatory network in early ascidian embryos 3. The gene regulatory network accounts for changes of gene expression in space and time 3.1 The initial set up in the 16-cell embryo by maternal factors 3.2 Specific gene expression in the 32-cell embryo 3.3 Specific gene expression at the 64-cell stage and thereafter 4. Perspectives References

2 4 4 5 5 7 10 10 12 16 25 26

Abstract Ascidian embryos are used as a model system in developmental biology due to their unique properties, including their invariant cell division patterns, being comprised of a small number of cells and tissues, the feasibility of their experimental manipulation, and their simple and compact genome. These properties have provided an opportunity for examining the gene regulatory network at the single cell resolution and at a genome-wide scale. This article summarizes when and where each regulatory gene is expressed in early ascidian embryos, and the extent to which the gene regulatory network explains each gene expression.

Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.01.001

#

2020 Elsevier Inc. All rights reserved.

1

2

Yutaka Satou

1. Introduction Ascidians are invertebrate chordates that have been used in research for over 100 years, since the studies in 1905 by Conklin (1905a, 1905b, 1905c). Cell division patterns in ascidian embryos are invariant among individuals, and cells are easily identifiable under a binocular stereo microscope. Taking advantage of these features, each cell is uniquely named, and cell lineages have been traced (Conklin, 1905c; Nishida, 1987; Nishida & Satoh, 1983, 1985). These experiments have revealed that most cells obtain one unique developmental fate by the 112-cell stage (Fig. 1A). That is, individual cells in the 112-cell embryos are precursors for cells that appear in neurula to larval stages (Fig. 1B and C). In other words, cell types that are found in A

vegetal hemisphere

animal hemisphere

posterior neural plate (PNP) (Zic-r.b, Foxb) notochord (T)

anterior neural plate (ANP) (Zic-r.b, Dmrt.a) anterior neural plate border (ANB) (Foxc, Dmrt.a)

lateral neural plate and its border (LNPB) (Nodal, Zic-r.b) B8.15

mesenchyme (Twist-r) muscle (Mrf) muscle/heart precursor (Mrf/Mesp) endoderm germ cells (Lhx3/4, Nkx2-1/4) (Pem-1)

anterior

ANB

ANP

C

PNP

LNPB

posterior

B

epidermis (Tfap2-r.b, Dlx.b)

central nervous system epidermis

notochord mesenchyme palp endoderm

muscle

Fig. 1 Development of Ciona embryos. (A) Illustrations for the vegetal (left) and animal views (right) of the 112-cell embryo (approximately 4.5 h after fertilization at 18 °C). Cells with different developmental fates are marked with different patterns and colors. Representative genes expressed in individual lineages are shown in parentheses. (B) Illustration of the initial neurula (approximately 6 h after fertilization at 18 °C) showing the neural plate and its border region. (C) Illustration of larva with simple architecture. The tadpole larva hatches in 18 h after fertilization at 18 °C.

A gene regulatory network in early ascidian embryos

3

larvae are largely specified by the 112-cell stage. For example, a blastomere called B8.15 gives rise to larval tail muscle cells, but does not contribute to other tissues. This article reviews functions of the gene regulatory network (GRN) in the specification of cell types until the 112-cell stage. Since early embryos contain only up to 112 cells, the gene expression patterns in these embryos can be examined at single-cell resolution by in situ hybridization. The effects of gene knock-down/out and overexpression can therefore be evaluated on a cell-by-cell basis. Accordingly, the GRN for cell fate specification has been determined at this resolution. Ascidian embryos have long been regarded to exhibit a “mosaic” mode of development, in which localized maternal determinants play essential roles (Conklin, 1905a). The basis for this idea has played an important role in the study of developmental mechanisms, as it highlights the importance of localized maternal factors. Using ascidian embryos, many studies have attempted to identify such localized maternal factors, resulting in the identification of several important maternal factors. The activities of these factors define the initial conditions for the GRN (see Section 2.2). In addition to the simplicity of embryos, their genome is also simple and compact. The size of the genomes of most ascidians have been estimated to be less than 200 Mb (Lemaire, 2011). The genome size of Ciona intestinalis had been estimated to be 160 Mb (Dehal et al., 2002; Satou, Mineta, et al., 2008), and the most recent assembly has indicated that it is approximately 125 Mb (Satou et al., 2019). This simplicity is explained partly by the observation that the ascidian genome contains fewer paralogs than vertebrate genomes (Chiba et al., 2003; Hino, Satou, Yagi, & Satoh, 2003; Kawashima, Tokuoka, Awazu, Satoh, & Satou, 2003; Sasakura et al., 2003; Sasakura, Yamada, Takatori, Satou, & Satoh, 2003; Satou, Imai, et al., 2003; Satou, Sasakura, et al., 2003; Satou & Satoh, 2005; Satou, Wada, Sasakura, Satoh et al., 2008; Wada et al., 2003; Yagi et al., 2003; Yamada, Kobayashi, Degnan, Satoh, & Satou, 2003). Accordingly, the number of genes encoded in the ascidian genome is less than 16,000, which is much smaller than the number of genes in vertebrate genomes. In addition, genes are encoded much more compactly in the ascidian genome. As 16,000 genes are encoded in a genome of 125 Mb, one gene is encoded by every 7.8-kb region on average. Indeed, the upstream
3-kb regions can recapitulate the endogenous expression patterns for many genes. The most extreme case may be that of a muscle actin gene (HrMA4a) of Halocynthia roretzi, as its 103-bp upstream region is sufficient for the reporter to be expressed specifically in larval muscle cells (Satou & Satoh, 1996). The simplicity and compactness

4

Yutaka Satou

of the genomes provides two advantages for the study of the GRN (Kubo, Imai, & Satou, 2009; Lemaire, 2009; Satou & Imai, 2015; Satou, Satoh, & Imai, 2009). First, genome-wide scale functional analyses are feasible, since the number of genes is small and functional redundancy among paralogs is rare. Second, the identification of cis-regulatory sequences is relatively easy. These features make the ascidian embryo a unique experimental system. It has been revealed that the animal we used to call Ciona intestinalis contains two cryptic species, the first (and most widely used species) is Ciona robusta (C. intestinalis type A), and the second, Ciona intestinalis (Brunetti et al., 2015). However, because these two species have not been discriminated in many previous studies, it is not possible to discuss the GRNs of these two species separately. Accordingly, these two animals are not discriminated in this review. Embryos of Ciona savignyi and Halocynthia roretzi are also commonly used for GRN studies. Because most findings in these species are applicable to C. intestinalis, previous studies using these two species will be cited throughout this review.

2. Basic properties for the gene regulatory network in early ascidian embryos 2.1 Regulatory genes Due to the simplicity of the genome, the number of genes for transcription factors (TFs) and signaling molecules, which constitute the GRN nodes, in ascidian embryos is small. For example, a previous study identified 46 genes encoding bHLH transcription factors (Satou, Imai, et al., 2003), in contrast to the
120 genes identified in the human genome. Similarly, the most recent genome assembly includes genes for 26 bZip TFs, 15 Ets TFs, 28 Fox TFs, 20 Sox/Hmg TFs, 86 homeodomain TFs, 18 nuclear receptors, 10 T-box TFs, 5 Smad TFs, 7 type-A ephrins, 7 fibroblast growth factors (Fgfs), 12 transforming growth factor β (Tgfβ) family molecules (including bone morphogenetic proteins (Bmps)), 11 Wnts, 2 Hedgehogs, and 3 Notch receptor ligands (Satou et al., 2019). Genes encoding TFs similar to other well-known TFs have also been extensively surveyed. These include genes for 2 Tfap2, 2 Gata, 3 Prdm1, 7 Zic, 2 Dmrt, Snai, and Gli. In addition, the genome of ascidian embryos contains nearly 300 genes encoding proteins with C2H2-type zinc fingers (Miwata et al., 2006). Therefore, approximately 600 genes constitute the nodes with outgoing edges within the GRN encoded in the ascidian genome, although some zinc finger proteins may not be transcription factors.

A gene regulatory network in early ascidian embryos

5

This review focuses on the GRN for fate specification in the ascidian embryo up to the 112-cell stage. This network is a sub-network of the entire GRN, which contains
600 regulatory gene nodes. According to previous comprehensive in situ hybridization assays, only about 90 regulatory genes are expressed zygotically up to the 112-cell stage (Imai, Hino, Yagi, Satoh, & Satou, 2004; Miwata et al., 2006). Hence, the sub-network that is discussed here consists of these regulatory nodes.

2.2 Initial conditions for the gene regulatory network In the ascidian embryo, the first zygotic gene expression begins at the 16-cell stage, although very weak expression can also be found at the 8-cell stage in the animal hemisphere (Oda-Ishii, Abe, & Satou, 2018; Oda-Ishii & Satou, 2018). The zygotic gene expression patterns of the regulatory genes categorize the 16 cells into five groups, a5.3/a5.4, b5.3/b5.4, A5.1/A5.2, B5.1, and B5.2 groups (Fig. 2A). The zygotic genetic program begins on the basis of these initial gene expression in each cell. The initial zygotic gene expression is activated by the activity of maternal regulatory factors. In other words, maternal factors create the five different territories in 16-cell embryos. Therefore, the activity of maternal factors should be different among these five categories, and provide the initial conditions for the GRN in the individual cells within the 16-cell embryo. As will be discussed below, Ctnnb (β-catenin), Tcf7, Gata.a, Macho-1 (Zic-r.a), and Pem-1 are known to be involved in this initial process (Bertrand, Hudson, Caillol, Popovici, & Lemaire, 2003; Imai, Takada, Satoh, & Satou, 2000; Nishida & Sawada, 2001; Rothb€acher, Bertrand, Lamy, & Lemaire, 2007; Satou et al., 2002; Yoshida, Marikawa, & Satoh, 1996).

2.3 Outputs of the gene regulatory network As the GRN directly governs gene expression, the dynamics of the GRN for fate specification should result in specific gene expression patterns in individual cells of the ascidian embryo. Indeed, at the 112-cell stage when most cells obtain one unique developmental fate, cells with different cell fates express different sets of genes from one another. In other words, the dynamics of the GRN for fate specification produce different gene expression patterns in the 112-cell embryo, and thereby produce precursors for cells that appear in neurula to larval stages (Fig. 1B and C). Specifically, endoderm precursors express Lhx3/4 and Nkx2-1/4 (Ristoratore et al., 1999; Satou, Imai, & Satoh, 2001b), notochord

A

a5.3/a5.4 b5.3/b5.4 A5.1/A5.2 B5.1 B5.2

Cells

Zygotically expressed genes

Gene Expression pattern

Pdrm1-r.a* Efna.d, Tfap2-r.b, Gdf1/3-r Sox1/2/3* Foxa.a* Hes.a* Fgf9/16/20, Foxd, Lefty, Foxtun2 Tbx6-r.b, Admp, Wnttun5 *Regulatory mechanisms for expression of these genes are not completely understood

B

a5.4

B5.2

A5.2 B5.1

A5.1

posterior

b5.3

.4

a5.3

b5

anterior

animal

vegetal

Fig. 2 Gene regulatory mechanisms in the 16-cell embryo. (A) Genes expressed from the zygotic genome at the 16-cell stage (approximately 2 h after fertilization at 18 °C). Fourteen genes show seven different expression patterns. (B) Lateral view of the 16-cell embryo in the center of the panel. Each cell is labeled with a unique name. Sister cell relationships are denoted by short bars. Cells with different gene expression profiles are marked with different colors or patterns. Differences in the GRN dynamics of the five territories are shown in boxes. Although the anterior and posterior cells in the animal hemisphere express slightly different sets of genes, the mechanisms that produce this difference has not been revealed and the GRN activity in these two territories are shown in one box. The networks were illustrated using Biotapestry (Longabaugh, 2012). Note that not all the genes are included in the illustrations, for simplicity.

A gene regulatory network in early ascidian embryos

7

precursors express T (Brachyury) (Corbo, Levine, & Zeller, 1997; Yasuo & Satoh, 1993), mesenchyme precursors express Twist-r (Imai, Satoh, & Satou, 2003; Tokuoka, Imai, Satou, & Satoh, 2004), muscle precursors express Mrf (Myod) (Araki, Saiga, Makabe, & Satoh, 1994; Meedel, Chang, & Yasuo, 2007; Meedel, Farmer, & Lee, 1997; Satoh, Araki, & Satou, 1996), a bipotential cell pair with the muscle fate and the adult heart precursor fate expresses Mrf and Mesp (Satou, Imai, & Satoh, 2004), precursors for the posterior neural plate region (PNP) express Foxb (Hashimoto, Enomoto, Kumano, & Nishida, 2011; Imai et al., 2004), epidermal cells express Dlx. b and Tfap2-r.b (Caracciolo, Di Gregorio, Aniello, Di Lauro, & Branno, 2000; Imai, Hikawa, Kobayashi, & Satou, 2017; Imai et al., 2004; Irvine, Cangiano, Millette, & Gutter, 2007), the anterior neural plate region (ANP), which contributes mainly to the central nervous system, expresses Zic-r.b and Dmrt.a (Imai, Levine, Satoh, & Satou, 2006; Imai, Satoh, & Satou, 2002; Tresser et al., 2010; Wada & Saiga, 2002), the anterior neural plate border (ANB), which contributes to the adhesive organs (palps) and the primordium for the adult oral siphon, expresses Foxc and Dmrt.a (Imai et al., 2004, 2006; Lamy, Rothbacher, Caillol, & Lemaire, 2006; Wagner & Levine, 2012), and the lateral neural plate region and the lateral neural plate border (LNPB), which contribute to dorsal epidermal cells, the dorsal part of the nerve cord, tail epidermal sensory neurons, bipolar tail neurons, and muscle cells near the tail tip (Nishida, 1987; Pasini et al., 2006; Stolfi, Ryan, Meinertzhagen, & Christiaen, 2015), express Msx (Aniello et al., 1999; Roure & Darras, 2016; Waki, Imai, & Satou, 2015). Most of these genes are expressed specifically in each of these lineages, and play important roles in specification of these lineages. For example, T is specifically expressed in notochord precursors, and the differentiation of notochord is severely disturbed by the knockdown or knockout of T (Chiba, Jiang, Satoh, & Smith, 2009; Satou, Imai, & Satoh, 2001a). In Section 3, how the GRN dynamics establish these specific gene expression patterns in the 112-cell embryo is discussed.

2.4 Potentially unique properties of the gene regulatory network in early ascidian embryos It is empirically known that regulatory genes activated in one cell cycle begin to regulate their downstream genes after the subsequent cell division. For instance, as will be discussed in the following sections, Foxd expression begins at the 16-cell stage, whereby the expression of its target genes begins at the 32-cell stage or thereafter (Imai et al., 2002; Hudson, Sirour, & Yasuo, 2016). This property is very helpful for analyzing GRN dynamics, because

8

Yutaka Satou

changes in GRN dynamics always occur once during one cell cycle. This may be due to the delay between gene expression and translation and the rapid division of cell in early embryos. However, alternatively, there may be an unknown mechanism that disturbs immediate translation. Embryos up to the 112-cell stage mostly consist of two layers of cells: animal hemisphere cells and vegetal hemisphere cells. Because signaling molecules are acting in such small and simple space of embryos, it is not difficult to infer which cells signaling molecules are acting on. Fgf9/16/20 is expressed in three pairs of cells (A5.1, A5.2, and B5.1) in the vegetal hemisphere of the 16-cell embryo (Fig. 2). Its protein product is expected to begin acting at the 32-cell stage. All cells in the 32-cell embryo directly contact the descendants of cells that express Fgf9/16/20 at the 16-cell stage (Tassy, Daian, Hudson, Bertrand, & Lemaire, 2006) (see Fig. 3). This implies that the downstream MAPK pathway is activated in all cells of the 32-cell embryo by this molecule. However, as will be discussed later, a membraneanchored signaling molecule, Efna.d, is expressed in the animal hemisphere. This molecule downregulates the MAPK pathway in six pairs of the animal cells that mainly contribute to epidermal cells (Haupaix et al., 2013; Ohta & Satou, 2013; Picco, Hudson, & Yasuo, 2007). As a result, the MAPK pathway is activated in the remaining two pairs of the animal cells (ANP/ANB and LNPB lineages), and in all vegetal cells. If Efna.d is knocked down, Fgf9/16/20 activates the MAPK pathway in all cells, which can be detected using an antibody for doubly-phosphorylated Erk (dpErk) (Ohta & Satou, 2013). In addition, taking an advantage of the simplicity of embryos, the shape and position of the individual cells within early embryos has been analyzed using confocal microscopy, and embryos have been reconstructed threedimensionally as virtual embryos (Tassy et al., 2006, 2010), which provide information regarding the distances between any pair of cells, as well as the area of the contact surfaces (when cells come into contact). Under the assumptions that cells with more contact to signal-emitting-cells receive more signaling molecules and that cells closer to signal-emitting-cells receive more signaling molecules, the order in which the cells receive more signaling molecules is easily predictable for each signaling ligand. These predictions are useful when examining which cells receive a given signaling molecule sufficiently. For example, the bilaterally symmetrical 32-cell ascidian embryo contains 16 pairs of cells. If we can determine the rank order for a given signaling molecule, there are 17 (¼16 + 1) possibilities, that is, no cells receive the signal sufficiently; only the first rank of cells

ANP/ANB lineage

LNPB lineage

Efna.d animal

b6.5

A6.4

Fgf9/16/20

B6.2

A6.3 A6 .1

B6.1

B6.4 B6.3 posterior

6.2 a6.5 A

anterior

a6.7

b6.8 b6.6

b6.7

a6 .

6

a6.8

vegetal

Fig. 3 Gene regulatory mechanisms in the 32-cell embryo. A lateral view of the 32-cell embryo (approximately 3 h after fertilization at 18 °C) is shown in the center. Each cell is labeled with a unique name. Cells are largely specified to six territories, each of which express different sets of genes and is represented by different colors and/or patterns. Sister cell relationships are denoted by short bars. Differences in the GRN dynamics of the five territories are shown in boxes. The most posterior cells are transcriptionally quiescent, and their GRN activity is not shown. Signaling sources of Fgf9/16/20 and Efna.d are shown in small boxes under the assumption that cells descended from cells with expression of a ligand gene at the preceding stage express the encoded protein. Thick lines in embryos indicate the boundary between the animal and vegetal hemispheres.

10

Yutaka Satou

receive the signal sufficiently; only the first and second ranks receive the signal sufficiently, and so on. By assessing whether each of these possibilities is compatible with observations in experiments, the ranges of action of the signaling molecules can be predicted. As was mentioned above, the MAPK pathway is activated in the two pairs of cells in the animal hemisphere of the 32-cell embryo. Among the animal cells, these two pairs contact cells with Fgf9/16/20 expression more extensively and cells with Efna.d expression less extensively than the remaining six pairs (Ohta, Waki, Mochizuki, & Satou, 2015; Tassy et al., 2006). These types of calculation can be used to predict the extent of cell-cell interactions in ascidian embryos.

3. The gene regulatory network accounts for changes of gene expression in space and time 3.1 The initial set up in the 16-cell embryo by maternal factors As mentioned above, five important maternal factors have been identified: Gata.a, Tcf7, Ctnnb, Macho-1, and Pem-1. The first, Gata.a (Bertrand et al., 2003; Rothb€acher et al., 2007), is detectable in all nuclei from the 2-cell to 16-cell stages (Oda-Ishii et al., 2018, 2016). Tcf7 is also detected in all nuclei in early embryos (Oda-Ishii et al., 2016). Ctnnb, which acts as a cofactor of Tcf7, is weakly detected in the nuclei of vegetal-hemisphere cells at the 8-cell stage and strongly at the 16-cell stage (Hudson, Kawai, Negishi, & Yasuo, 2013; Kawai, Iida, Kumano, & Nishida, 2007; Oda-Ishii et al., 2018, 2016), although the mechanism for controlling the nuclear localization of Ctnnb has not been revealed. The mRNA for Macho-1 (Zic-r.a) and Pem-1 are localized in the posterior pole (the posterior end of B5.2; Fig. 2) (Kumano, Takatori, Negishi, Takada, & Nishida, 2011; Nishida & Sawada, 2001; Yoshida et al., 1996). As discussed below, Macho-1 protein is detected in the most posterior cells and its sister cells (Oda-Ishii et al., 2016), while Pem-1 protein is detectable only in the most posterior cells (Shirae-Kurabayashi, Matsuda, & Nakamura, 2011). Because Ctnnb is localized only in the nuclei of cells in the vegetal hemisphere, a complex of Tcf7 and Ctnnb activates its targets in the vegetal cells (A5.1, A5.2, and B5.1 cell pairs), except the most posterior cells (B5.2) (Fig. 2). Fgf9/16/20, Foxd, and Lefty are indeed targets of Tcf7/Ctnnb, and expressed in the vegetal hemisphere (Oda-Ishii et al., 2016).

A gene regulatory network in early ascidian embryos

11

Gata.a is responsible for the zygotic activation of genes in the animal hemisphere (Bertrand et al., 2003; Rothb€acher et al., 2007). Although Gata.a is observed in all nuclei of the 16-cell embryo (Oda-Ishii et al., 2016), Gata.a activity is suppressed by nuclear Ctnnb via protein-protein interactions in the vegetal hemisphere (Oda-Ishii et al., 2016; Rothb€acher et al., 2007). Therefore, Gata.a acts only in the animal hemisphere. Efna.d and Tfap2-r.b, which are target genes of Gata.a, are expressed across the entire animal hemisphere at the 16-cell stage (Fig. 2). Tbx6-r.b, Admp, and Wnttun5 are expressed in the posterior vegetal cells (B5.1), except for the most posterior cells. While Tcf7/Ctnnb is required for the expression of these genes, as in the case of the genes expressed in the entire vegetal hemisphere, knockdown experiments have shown that Macho-1 is additionally required (Yagi, Satoh, & Satou, 2004) (Fig. 2B). Although the simplest hypothesis may be the combinatorial regulation of Tcf7/Ctnnb and Macho-1, the finding of cis-regulatory elements responsible for repressing the expression in the entire vegetal hemisphere in the upstream regions of Tbx6-r.b and Wnttun5 indicates that Macho-1 somehow antagonizes this repression in the posterior vegetal quadrant of the 16-cell embryo (Oda-Ishii et al., 2016). Macho-1 mRNA and Pem-1 mRNA are associated with the cortical network of rough ER (cER) in mature oocytes. After fertilization, these mRNAs and cER concentrate into a subcellular structure called the centrosome attracting body, located in the posterior pole of the embryo (Nishida, 2005; Nishikata, Hibino, & Nishida, 1999; Prodon, Chenevert, & Sardet, 2006). Due to its localization, their protein levels are highest in the most posterior cells. Macho-1 protein is detectable strongly in the most posterior cells (B5.2) and relatively weakly in their sister cells (B5.1) at the 16-cell stage (Oda-Ishii et al., 2016), while Pem-1 protein is detectable only in the most posterior cells (B5.2) (Shirae-Kurabayashi et al., 2011) (Fig. 2). This may be due to the difference in the sensitivity of the antibodies used. However, it is more likely that the distribution patterns are indeed different, since functional experiments have shown that Macho-1 acts in B5.1 while Pem-1 acts only in the most posterior cells (B5.2). Germ line cells are derived from the most posterior cells (B5.2), where Macho-1 and Pem-1 mRNAs are localized. Pem-1 suppresses the activity of RNA polymerase II, and, therefore, the most posterior cells are transcriptionally inactive, which is a common feature for germ line cells (Kumano et al., 2011; Shirae-Kurabayashi et al., 2011). That is, the most posterior

12

Yutaka Satou

cells with Pem-1 do not express genes zygotically at the 16-cell stage. Upon each cell division of this most posterior lineage with the germ cell fate, one somatic daughter cell lineage is produced and becomes transcriptionally active, while the other daughter cell retains the germ cell fate and is transcriptionally inactive. In this way, these five maternal factors set up the first zygotic gene expression patterns in the 16-cell embryo. However, the expression of a small number of genes has not yet been explained. Hes.a is expressed in all cells, except the most posterior cells (Fig. 2A). This gene may have two regulatory elements, each of which is regulated by Gata.a and Tcf7/ Ctnnb, respectively, although this possibility has not been tested. Foxa.a is expressed in the vegetal cells and in the anterior animal cells (Fig. 2A); while the expression in the vegetal cells is under the control of Ctnnb (Hudson et al., 2016), the expression in the anterior animal cells cannot be explained by Ctnnb activity alone. Sox1/2/3 is expressed in the animal cells and in the anterior vegetal cells, and Prdm1-r.a is expressed only in the anterior animal cells. These expression patterns cannot be explained by combinations of the abovementioned mechanisms. Additional maternal factors are likely to be responsible for the expression of these genes. Despite such unsolved issues, the initial gene expression profile indicates that the localized activities of the maternal factors provide five different initial conditions for GRN dynamics in the 16-cell embryo.

3.2 Specific gene expression in the 32-cell embryo After the 16-cell stage, cell divisions become asynchronous, as nuclear Ctnnb controls the progression of the cell cycle (Dumollard, Hebras, Besnardeau, & McDougall, 2013). As a result, the blastomeres in the vegetal hemisphere divide earlier than the blastomeres in the animal hemisphere, which results in a 24-cell stage for a short period of time before the 32-cell stage. The cells around the vegetal pole mainly give rise to endodermal tissues (A6.1, A6.3, and B6.1; Fig. 3); the most posterior vegetal cells retains their germ cell fate (B6.3; Fig. 3); the other marginal vegetal cells give rise to mesodermal tissues and the PNP (A6.2, A6.4, B6.2, and B6.4; Fig. 3); two pairs of the animal hemisphere cells (a6.5 and b6.5) contribute to the neural plate and its border cells (ANP, ANB, and LNPB); the remaining animal hemisphere cells mainly give rise to epidermal cells (Nishida & Satoh, 1985). Thus, at this stage, three germ layers are largely separated, wherein each cell lineage expresses a specific set of genes.

A gene regulatory network in early ascidian embryos

13

The most anterior pair of the animal cells (a6.5), which contributes to the ANP and ANB, expresses Otx and Dmrt.a, and the cell pair that contributes to the LNPB (b6.5) expresses Otx and Nodal (upper boxes in Fig. 3). These genes are under the control of the MAPK pathway (Bertrand et al., 2003; Hudson, Darras, Caillol, Yasuo, & Lemaire, 2003; Hudson & Lemaire, 2001; Imai et al., 2006; Imai, Satoh, & Satou, 2002a; Khoueiry et al., 2010). This pathway is controlled positively by Fgf9/16/20 and negatively by Efna.d (Haupaix et al., 2013; Picco et al., 2007; Shi & Levine, 2008). Because Fgf9/16/20 is expressed in the vegetal hemisphere of the 16-cell embryos, Fgf9/16/20 is expected to be secreted from their descendants (see the left small box in Fig. 3). Efna.d is expressed in all animal cells of the 16-cell embryo, and therefore its protein product is expected to be present in all animal cells of the 32-cell embryo (see the right small box in Fig. 3). As Efna.d is a signaling molecule anchored to the cell membrane, this signal is transmitted only to neighboring cells. According to a 3D-reconstructed virtual embryo (Tassy et al., 2006, 2010), animal hemisphere cells without Otx expression have larger areas of contact with cells expressing Efna.d, and animal cells with Otx expression (a6.5 and b6.5) have larger areas of contact with cells expressing Fgf9/16/20 (Ohta & Satou, 2013; Tassy et al., 2006). More intuitively, cells around the animal pole are extensively surrounded by cells with Efna.d expression, and marginal cells in the animal hemisphere (a6.5 and b6.5; the ANP/ANB and LNPB lineages) are not so extensively surrounded by cells with Efna.d expression (see Fig. 3); for example, b6.8 is surrounded thoroughly by cells with Efna.d expression, while b6.5 is surrounded by two cells with Efna.d expression and three cells with Fgf9/16/20 expression (see Fig. 3). Indeed, the MAPK pathway is activated in a6.5 and b6.5, but not in the other animal hemisphere cells (Hudson et al., 2003; Ohta & Satou, 2013; Ohta et al., 2015). In addition, Foxa.a regulates Dmrt.a positively, and Nodal negatively. Because Foxa.a is expressed in the anterior cells (its mRNA is expressed in their parental cells of the 16-cell embryo; see Fig. 2), Dmrt.a is expressed only in the ANP/ANB lineage, and Nodal is expressed only in the LNPB lineage (Lamy et al., 2006; Ohta & Satou, 2013) (Fig. 3). Gata.a is present almost at the same level during early embryogenesis (Oda-Ishii et al., 2018), and indeed it keeps activating Efna.d and Tfap2r.b in cells with epidermal fate of the 32-cell embryo, as it does in the 16-cell embryo. In the neural lineage, Gata.a acts as an effector of the MAPK signaling pathway, in addition to a canonical effector Ets1/2 (Bertrand et al., 2003).

14

Yutaka Satou

Admp and Gdf1/3-r, which both belong to the TGFβ-superfamily, also regulate Otx and Nodal expression, as the knockdown of Admp and Gdf1/3-r was found to result in the ectopic expression of Otx and Nodal (Ohta & Satou, 2013). However, these factors do not give positional information and therefore are not included in Fig. 3. This negative regulation is required to avoid Otx and Nodal responding to weakly activated MAPK pathway in the animal cells with epidermal fate, since Efna.d cannot completely downregulate the MAPK pathway. Admp is a ligand to activate the BMP pathway. Thus, the combination of the positive action of the Fgf signaling and the negative action of the BMP signaling induces Otx and Nodal expression in the neural (ANP/ANB and LNBP) lineages, which is reminiscent of neural induction in vertebrate embryos (Delaune, Lemaire, & Kodjabachian, 2005; Marchal, Luxardi, Thome, & Kodjabachian, 2009; Munoz-Sanjuan & Brivanlou, 2002; Streit, Berliner, Papanayotou, Sirulnik, & Stern, 2000; Wilson, Graziano, Harland, Jessell, & Edlund, 2000). Meanwhile, although these two signaling pathways simultaneously establish the dorso-ventral body axis in vertebrate embryos, the dorso-ventral axis is not evident in this stage of the ascidian embryo. Because BMP signaling is used for the patterning of epidermal cells along the dorso-ventral axis of embryos at the tailbud stage (which is much later than the 112-cell stage) (Imai, Daido, Kusakabe, & Satou, 2012; Pasini et al., 2006), these two events, neural induction and dorso-ventral axis establishment, may have had been independent in ancestral chordates. In the vegetal cells of the 32-cell embryo, except B6.3 and B6.4 (the most posterior cells with the germ cell fate and their siblings), nuclear Ctnnb again plays a key role in establishing the expression patterns in the vegetal hemisphere. Although its mechanism is not yet understood, nuclear Ctnnb is observed only in the nuclei of cells around the vegetal pole (Hudson et al., 2013). That is, among the descendants of cells with nuclear Ctnnb at the 16-cell stage, only the daughters closer to the vegetal pole (A6.1, A6.3, and B6.1) have nuclear Ctnnb (see Fig. 3). Because Gata.a is distributed throughout the embryos and Ctnnb is not present in the nuclei of the marginal vegetal cells (A6.2, A6.4, and B6.2; Fig. 3), Gata.a begins to act in these marginal cells (Imai, Hudson, Oda-Ishii, Yasuo, & Satou, 2016). In their parental cells in the 16-cell embryo, Foxa.a, Foxd, and Fgf9/16/20 are expressed (Fig. 2). In combination with the factors encoded by these mRNAs, nuclear Ctnnb activates Lhx3/4 in the cells close to the vegetal pole, and Gata.a activates Zic.r-b in the marginal cells (compare the bottom two boxes in Fig. 3) (Hudson et al., 2016).

A gene regulatory network in early ascidian embryos

15

Gata.a activates different genes between the animal cells and the vegetal marginal cells of the 32-cell embryo. As mentioned above, Foxa.a, Foxd, and the MAPK pathway activity are necessary for Gata.a to activate Zic-r. b expression. However, because the animal cells lack one or more of these factors (see Fig. 3), Gata.a does not activate Zic-r.b in animal cells. On the other hand, because Foxd, which is expressed in the vegetal cells, represses Otx and Dmrt.a (Tokuhiro et al., 2017), these genes are not activated in the vegetal marginal cells by Gata.a (the bottom box on the left-hand side in Fig. 3; Note that Otx is activated in the vegetal cells through a mechanism independently of Gata.a, as described in the next paragraph). In this way, the GRN shows how these different types of cell express different genes. However, it does not completely account for the gene expression in the 32-cell embryos. In addition to the neural lineage cells in the animal hemisphere, vegetal cells also express Otx and Nodal via independent regulatory mechanisms. Specifically, Otx is expressed in the posterior vegetal cells except B6.3, and Nodal is expressed in the vegetal cells around the vegetal pole (A6.1, A6.3, and B6.1). Macho-1 regulates the expression of Otx and Nodal indirectly or directly in these cells (Yagi et al., 2004). In addition, the expression of Otx is regulated by the MAPK pathway (Hudson et al., 2003). Thus, the expression of Otx and Nodal in the vegetal hemisphere is regulated differently from their expression in the animal hemisphere; however, the mechanisms are not completely understood. Among the blastomeres in the 32-cell embryo, B6.4 cells, which are siblings of the cells with the germ cell fate, are unique, since the zygotic transcription of this cell pair begins at the 32-cell stage. As discussed above, in the vegetal marginal cells, except B6.4, Zic-r.b is activated through the combinatorial action of Foxa.a, Foxd, Gata.a, and the MAPK signaling pathway. However, B6.4 also expresses Zic-r.b at the 32-cell stage, although Foxa.a or Foxd is not expressed in the parental cells. Indeed, the knockdown of Gata.a abolishes the expression of Zic-r.b in A6.1, A6.3, and B6.1, but not in B6.4 (Imai et al., 2016). This observation indicates that Zic-r.b expression is regulated differently in B6.4, although this mechanisms is not yet completely understood. Similarly, Snai, which is not included in Fig. 3, is expressed in the posterior vegetal cells (B6.1, B6.2, and B6.4), except the most posterior cells. In B6.4, a constitutively active form of Raf (MAP3K), which is encoded by an alternatively spliced mRNA, activates the MAPK pathway, and then it activates Snai cooperatively with Macho-1 in B6.4. On the other hand, in B6.1 and B6.2, Tbx6-r.b activates Snai (Tokuoka, Kobayashi, & Satou, 2018).

16

Yutaka Satou

3.3 Specific gene expression at the 64-cell stage and thereafter 3.3.1 Specification of endodermal fate The endoderm is derived from three pairs of cells around the vegetal pole in the 32-cell embryo (see Fig. 3). Lhx3/4 plays an important role in the specification of the endoderm, as the knockdown of this gene results in severe loss of expression of endodermal markers (Satou et al., 2001b). However, Lhx3/4 is also expressed in a cell pair, which gives rise to muscle and adult heart precursor cells, and regulates Mesp expression there (Christiaen, Stolfi, Davidson, & Levine, 2009). This indicates that specification of endoderm fate is not the only function of Lhx3/4, because Mesp plays an essential role in specification of the heart precursors (Satou et al., 2004). Therefore, it is unlikely that Lhx3/4 alone specifies endodermal fate. Nkx2-1/4 (Ttf1 or Titf1) is expressed only in the endodermal lineage under the control of Fgf9/16/20, Foxa.a, and Foxd at the 64-cell stage and thereafter (Fig. 4) (Imai et al., 2006; Ristoratore et al., 1999; Spagnuolo & Di Lauro, 2002; Tokuhiro et al., 2017). Otx is also expressed under the control of Foxa.a in the endodermal cells at the 64-cell stage, and activates zygotic Gata.a expression in the endoderm at the tailbud stage (Imai et al., 2006). In this way, the endoderm precursor cells specifically express the set of Lhx3/4, Nkx2-1/4, and Otx at the 64-cell stage (Fig. 4). It is possible that these transcription factors regulate different sets of target genes. However, the GRN does not completely account for the gene expression in this lineage. As shown in Fig. 2, the expression of Fgf9/16/20, Foxa. a, and Foxd begins at the 16-cell stage, and their protein products activate their targets at the 32-cell stage (Fig. 3). Although these three factors are the only known upstream factors of Nkx2-1/4 expression, Nkx2-1/4 is not activated at the 32-cell stage. It is likely that there are unidentified transcription factors that repress Nkx2-1/4 at the 32-cell stage, or begin to activate Nkx2-1/4 at the 64-cell stage. 3.3.2 Fate specification of mesenchymal cells Among the three pairs of cells with Lhx3/4 expression at the 32-cell stage, two pairs close to the vegetal pole give rise to endodermal cells only. The remaining one pair (A6.3) produces precursor cells for the mesenchyme and for the endoderm at the 64-cell stage. Specifically, the daughter cells that are located marginally (A7.6) are mesenchyme precursor cells, and the others are endoderm precursor cells (see Figs. 3 and 4). The difference between the siblings depends on two signaling pathways: the Nodal and

17

112-cell embryo

animal

vegetal

posterior

64-cell embryo

anterior

A gene regulatory network in early ascidian embryos

A7.6 B7.3

Mesenchyme (B; posterior)

Endoderm

Mesenchyme (A7.6-derived)

Mesenchyme (A7.6)

Fgf9/16/20

Efna.d

Nodal

Fig. 4 Gene regulatory mechanisms in endodermal and mesenchymal cells of the 64-cell and 112-cell embryo. Lateral views of the 64-cell and 112-cell embryo (approximately 4–4.5 h after fertilization at 18 °C) are shown at the top. Endodermal and mesenchymal cells are marked with different colors and/or patterns. Sister cell relationships are denoted by short bars. Differences in the GRN dynamics of the representative endodermal and mesenchymal cells are shown in boxes. Arrowheads indicate the cells derived from B7.3, which give rise to the posterior eight notochord cells (black) and mesenchymal cells (white). Signaling sources of Fgf9/16/20, Efna.d, and Nodal are shown in small boxes under the assumption that cells descended from cells with expression of a ligand gene at the preceding stage express the encoded protein. Thick lines in embryos indicate the boundary between the animal and vegetal hemispheres.

18

Yutaka Satou

MAPK signaling pathways. Nodal is first expressed in the LNPB lineage, and in three pairs of the vegetal cells near the vegetal pole at the 32-cell stage (see Fig. 3). The Nodal signaling pathway positively regulates genes specific for the mesenchyme precursors (Hudson & Yasuo, 2006; Imai et al., 2006; Shi & Levine, 2008). The MAPK signaling pathway is activated in the endodermal precursors but not in the mesenchyme precursors. As in the 32-cell embryos, Fgf9/16/20 and Efna.d positively and negatively regulate this pathway, respectively. The mesenchyme precursor, but not the endoderm precursor, is located next to animal cells with Efna.d expression. Thus, the MAPK pathway is activated in the endoderm precursors and suppressed in the mesenchyme precursors (Shi & Levine, 2008). The Nodal pathway positively regulates a specific gene set, including Fgf8/17/18, Dlk (Delta-like or Delta2), Hand-r (Notrlc or Hand-like), and Myt1 (Mytf), while the MAPK pathway negatively regulates the same set (compare the two boxes on the left-hand side in Fig. 4). As a result, these genes are expressed specifically in the mesenchyme precursor pair. In addition to cells derived from A7.6, mesenchymal cells are also differentiated from the posterior vegetal quadrant. These two populations of mesenchymal cells have slightly different properties; the A7.6-derived population contributes to oral siphon muscle and blood cells in adults, while the other contributes to blood cells but not to the oral siphon muscle (Tokuoka, Satoh, & Satou, 2005). Nevertheless, all mesenchyme precursors express Twist-r, and this gene plays an essential role in specification of these cells. In the former population, Twist-r begins to be expressed under the control of Foxa.a, Foxd, Hand-r, and Otx after the 112-cell stage (Imai et al., 2006, 2003) (the lower right box in Fig. 4). In the latter population, Twist-r is activated by the combinatorial action of Zic-r.b, Otx, and MAPK signaling (the lower right box in Fig. 4). The latter mesenchymal precursors are derived from the marginal vegetal cells of the 32-cell embryo. From these marginal cells, muscle precursor cells are also derived. However, because the MAPK signaling pathway is activated in the mesenchyme precursor cells but not in their sibling muscle precursors, Twist-r is specifically expressed in the mesenchyme precursors but not in the muscle precursors (Imai et al., 2003; Kim, Kumano, & Nishida, 2007; Nishida, 2003). However, this does not completely explain how Twist-r is expressed specifically in the mesenchyme precursors, since Zic-r.b, Otx, and MAPK signaling act in the ANP and LNPB cells, as will be discussed in Sections 3.3.6 and 3.3.7 (see also Fig. 6). The current version of the GRN does not provide an explanation for why these cells do not express Twist-r.

A gene regulatory network in early ascidian embryos

19

Another unsolved issue is the initiation of Twist-r expression. The latter population consists of two distinct cell lineages. Among them, the posterior lineage initiates Twist-r expression at the 64-cell stage, while the anterior lineage initiates expression at the 112-cell stage. This delay is important, as the anterior lineage (B7.3) retains the developmental potential for the formation of the mesenchyme and notochord at the 64-cell stage. At the 112-cell stage, the developmental fate of its anterior daughter is restricted to the notochord, while that of the posterior daughter is restricted to the mesenchyme (arrowheads in Fig. 4). However, it has not been elucidated why Twist-r is not activated in the anterior lineage at the 64-cell stage. 3.3.3 Fate specification of muscle cells and adult heart precursors In the muscle lineage, Mrf is activated by the combinatorial action of Tbx6-r.b and Zic-r.b. Subsequently, the combined action of Zic-r.b and Mrf re-activates Tbx6-r.b. This positive feedback circuit maintains Tbx6-r.b and Mrf expression in early embryos (Yu, Oda-Ishii, Kubo, & Satou, 2019). At the 64-cell and 112-cell stages, two pairs of cells, which are siblings of the cells of mesenchymal fate, and their descendants express Mrf under the control of Tbx6-r.b and Zic-r.b (the lower box in Fig. 5). In these cells, some muscle structural genes, including genes for muscle actin and myosin regulatory light chain, are activated initially only by Tbx6-r.b and later by the combinatorial action of Tbx6-r.b and Mrf (Kubo et al., 2010; Kugler et al., 2010; Yu et al., 2019). Muscle cells are also derived from a cell pair, called B7.5, of the 64-cell embryo. The parental cell pair in the 32-cell embryo has the germ cell fate, and is therefore transcriptionally silent. In other words, the B7.5 lineage initiates zygotic transcription at the 64-cell stage. In this lineage, Tbx6-r.b expression begins under the control of Macho-1, and Lhx3/4 expression begins under the control of Ctnnb (Christiaen et al., 2009). Zic-r.b expression also begins at the 64-cell stage (Imai et al., 2004, 2002; Yu et al., 2019). Until the early gastrula stage, Mrf is activated under the control of Tbx6-r.b and Zic-r.b, and Mesp is activated under the control of Tbx6-r.b, Lhx3/4, and MAPK signaling (the upper boxes in Fig. 5) (Christiaen et al., 2009; Satou et al., 2004; Yu et al., 2019). Mrf and Mesp are essential transcription factors for the specification of muscle and adult heart precursors, respectively. Thus, this B7.5 cell pair is a bipotential cell pair that gives rise to muscle and adult heart precursors. Later, Fgf signaling acts on a subset of their descendants to specify the heart precursor fate, and the remaining descendants give rise to muscle (Davidson, Shi, Beh, Christiaen, & Levine, 2006).

20

anterior

animal

64-cell embryo

112-cell embryo

vegetal

posterior

Yutaka Satou

B7.5 B7.5

Muscle / Adult heart precursor

Fgf9/16/20

Muscle / Adult heart precursor

Fgf9/16/20 Muscle

late early muscle structural genes

Fig. 5 Gene regulatory mechanisms in muscle and adult heart lineage cells of the 64-cell and 112-cell embryo. Lateral views of the 64-cell and 112-cell embryo are shown at the top. Muscle and adult heart lineage cells are marked with different colors or patterns. Sister cell relationships are denoted by short bars. Differences in the GRN dynamics of the representative muscle and adult heart lineage cells are shown in boxes. Signaling sources of Fgf9/16/20 are shown in small boxes under the assumption that cells descended from cells with Fgf9/16/20 expression at the preceding stage express Fgf9/16/20 protein. Thick lines in embryos indicate the boundary between the animal and vegetal hemispheres. The panel showing Fgf9/16/20 expression in the 64-cell embryo is the same as that shown in Fig. 4.

3.3.4 Fate specification of the notochord and the posterior neural plate The ascidian larval tail contains 40 notochord cells, and the anterior 32 cells are derived from the anterior two pairs of the vegetal marginal cells in the32-cell embryo. However, these pairs in the 32-cell embryo have the notochordal fate and the PNP fate. After the division between the 32- and 64-cell stages (Fig. 6), the anterior daughters give rise to

21

A gene regulatory network in early ascidian embryos

Epidermis

LNPB

posterior

anterior

animal

Fgf9/16/20

B7.3

vegetal

Efna.d

Nodal PNP Notochord (anterior)

Fig. 6 Gene regulatory mechanisms in the lineages of the notochord, epidermis, PNP, and LNPB of the 64-cell embryo. A lateral view of the 64-cell embryo is shown in the center. Cells for the notochord (anterior), epidermis, PNP, and LNPB lineages are marked with different colors and/or patterns. Sister cell relationships are denoted by short bars. Differences in the GRN dynamics of the representative cells are shown in boxes. Signaling sources of Fgf9/16/20, Efna.d, and Nodal are shown in small boxes under the assumption that cells descended from cells with expression of a ligand gene at the preceding stage express the encoded protein. These panels are the same as those shown in Fig. 4. Thick lines in embryos indicate the boundary between the animal and vegetal hemispheres.

the PNP precursors, and the posterior daughters give rise to the notochord precursors. This difference between the siblings is primarily established by the activity of the MAPK signaling pathway. The anterior daughters, but not the posterior daughters, are located next to the animal cells, which

22

Yutaka Satou

express Efna.d. As in the case of the A6.3 daughters, which give rise to the mesenchyme precursors and the endoderm precursors, Efna.d signal downregulates the MAPK pathway in the anterior daughters, so that the MAPK pathway is activated only in the posterior daughters (Haupaix et al., 2013). In the posterior daughters, MAPK signaling and Zic-r.b combinatorially activate T (Brachyury) (Yagi, Satoh, & Satou, 2004). MAPK signaling also represses Foxb expression in the posterior daughters, while Zic-r.b activates Foxb in the anterior daughters, in which the MAPK pathway is downregulated (Hashimoto et al., 2011; Imai et al., 2006). Foxa.a, whose mRNA begins to be expressed at the 16-cell stage, is also required for the expression of T and Foxb (the lower box in Fig. 6). Thus, the anterior and posterior rows of cells express Foxb and T, respectively. The posterior 8 notochordal cells in the larval tail are derived from the posterior vegetal quadrant. As described above, at the 64-cell stage, this lineage of cells retains both the notochordal and mesenchymal fate (B7.3 in Fig. 6). After the next division, the fate of the anterior daughter is restricted to the notochord and the fate of the posterior daughter is restricted to the mesenchyme. This notochord precursor is located between the mesenchyme precursors derived from the anterior vegetal quadrant and the posterior vegetal quadrant (black arrowheads in Fig. 4). The former mesenchyme precursors (A7.6 lineage) express Dlk (Delta-like or Delta2; Fig. 4), which encodes a ligand for the Notch signaling pathway, and the expression of T is induced through this signaling pathway (Corbo, Fujiwara, Levine, & Di Gregorio, 1998; Hudson & Yasuo, 2006; Imai, Satoh, & Satou, 2002b). The overexpression of a constitutively active form of Notch suggests that only the posterior vegetal cells are competent for Dlk/Notch signaling in C. savignyi embryos (Imai et al., 2002b), although the current version of the GRN does not explain which factors make these cells competent. In this way, T is expressed in the anterior and posterior lineages of cells via the different regulatory mechanisms. T subsequently activates the genes necessary for the formation of the notochord both directly and indirectly (see chapter “The notochord gene regulatory network in chordate evolution: Conservation and divergence from Ciona to vertebrates” by Di Gregorio in this volume; Hotta, Takahashi, Erives, Levine, & Satoh, 1999; Katikala et al., 2013; Takahashi et al., 1999; Yasuo & Satoh, 1998). 3.3.5 Epidermal cells In the animal hemisphere, Dlx.b is activated by Sox1/2/3 (Imai et al., 2017; Irvine, Vierra, Millette, Blanchette, & Holbert, 2011) (Fig. 6). Although Sox1/2/3 is expressed in the animal and vegetal hemispheres (see Fig. 2),

A gene regulatory network in early ascidian embryos

23

Foxd represses Dlx.b in the vegetal hemisphere (not indicated in Figures, for simplicity), and therefore Dlx.b is specifically expressed in the animal hemisphere (Imai et al., 2006; Tokuhiro et al., 2017). Tfap2-r.b is activated directly by maternal Gata.a at the 16-cell stage, as described in Section 3.1. Subsequently, the activity of the MAPK pathway downregulates Tfap2-r.b expression in the ANB/ANP and LNPB lineages (Imai et al., 2017) (Figs. 6 and 7). In other words, Tfap2-r.b is initially activated in the animal hemisphere, and the expression is subsequently restricted to the epidermal lineage of cells. Tfap2-r.b and Dlx.b then cooperatively specify the epidermal fate (Imai et al., 2017). 3.3.6 Cells for the lateral neural plate and its border LNPB cells are derived from two cells (b6.5) that express Otx and Nodal at the 32-cell stage (Fig. 3) (Hudson & Yasuo, 2005, 2006; Imai et al., 2006). Msx is expressed specifically in this lineage under the control of this Nodal signaling and Otx at the 64-cell stage (Fig. 6) (Imai et al., 2006). Then, Msx plays a central role in the specification of epidermal sensory neurons (Roure & Darras, 2016; Roure, Lemaire, & Darras, 2014; Waki et al., 2015). 3.3.7 Cells for the anterior neural plate and its border The ANP and ANB cells are mainly derived from cells (a6.5) that express Dmrt.a and Otx at the 32-cell stage (see Fig. 3). At the 64-cell stage, the daughter cells of a6.5 retain the ANP fate and the ANB fate, and are located next to the vegetal hemisphere cells, which express Fgf9/16/20. Indeed, the MAPK pathway is subsequently activated and downregulates Tfap2-r.b in a6.5-derived cells, as described above. Although MAPK signaling potentially works as an activator for Zic-r.b, Hes.a, Prdm1-r.a, and its paralog Prdm1-r.b works as repressors for Zic-r.b (Ikeda, Matsuoka, & Satou, 2013) (Fig. 7; Hes.a is not indicated for simplicity). The next division yields the ANP and ANB cells (Fig. 7), where only the ANP cells retain extensive contact with the vegetal cells with Fgf9/16/20 expression. Prdm1-r.a represses the transcription of its own gene and Prdm1-r.b, resulting in their expression being turned off before the 112-cell stage. For this reason, at the 112-cell stage, among the descendants of a6.5, cells in contact with Fgf9/16/20-expressing cells initiate Zic-r.b expression, and are specified as ANP cells (Ikeda et al., 2013). In these ANP cells, the same Fgf signal represses Foxc (Wagner & Levine, 2012). The remaining descendants (ANB) do not sufficiently receive Fgf9/16/20 molecules, so that Zic-r.b is not activated and Foxc is not repressed. Instead, Otx and Dmrt.a act as

24

Yutaka Satou

ANB

ANB / ANP

64-cell embryo

112-cell embryo

vegetal

a7.13

Fgf9/16/20

anterior

animal

posterior

ANP

Fgf9/16/20

Fig. 7 Gene regulatory mechanisms in the ANB and ANP lineages of the 64-cell and 112-cell embryo. Lateral views of the 64-cell and 112-cell embryo are shown at the bottom. The ANB and ANP lineage cells are marked with different colors and/or patterns. These two lineages are not separated in the 64-cell embryo. Sister cell relationships are denoted by short bars. Differences in the GRN dynamics of the representative ANB and ANP cells are shown in boxes. Signaling sources of Fgf9/16/20 are shown in small boxes under the assumption that cells descended from cells with Fgf9/16/20 expression at the preceding stage express Fgf9/16/20 protein. These panels are also shown in Figs. 4–6. Thick lines in embryos indicate the boundary between the animal and vegetal hemispheres.

A gene regulatory network in early ascidian embryos

25

positive regulators for Foxc, and, as a result, these ANB cells express Foxc (Ikeda et al., 2013; Imai et al., 2006; Wagner & Levine, 2012), and eventually give rise to the oral siphon primordium and the palps (Abitua et al., 2015; Liu & Satou, 2019). The cells (a7.13) next to the daughter cells of a6.5 also contribute to the ANP and ANB (see Fig. 7). However, the regulatory mechanism for this lineage is largely unknown.

4. Perspectives The ascidian GRN characterized thus far has largely succeeded in explaining how each regulatory gene is activated in specific cells at specific stages. In many cases, the MAPK signaling pathway has been found to play a key role in establishing differential gene expression patterns between sibling cells. Particularly in early embryos, Fgf9/16/20 is expressed in many cells, and this secreted molecule potentially activates the MAPK pathway in all cells. However, the MAPK pathway is differentially activated between siblings through the antagonizing activity of Efna.d, as this signaling molecule is bound to the cell membrane and therefore acts only on cells in vicinity (Haupaix et al., 2013). In ascidian embryos, specification proceeds in a limited space with a small number of cells, which may have constrained the GRN to utilize the Fgf/Ephrin system extensively during ascidian evolution (Lemaire, 2009; Satou & Imai, 2015). Many studies have focused on how specific gene expression is evoked. However, at least in some cases, it is also important to understand when and how genes are inactivated. Prdm1-r.a is turned off by negative autoregulation after two rounds of cell division. As discussed, this precise termination is important for Zic-r.b expression in the ANP (Ikeda et al., 2013). Prdm1-r.a also terminates Foxa.a expression in the ANP lineage before the activation of Zic-r.b by the MAPK signaling pathway. If Foxa.a expression is not properly terminated, MAPK signaling, Zic-r.b, and Foxa.a act together in this lineage. Because this combination of the regulatory factors is sufficient to activate T, T is ectopically expressed in such experimental embryos, which consequently develop ectopic notochord cells instead of the brain (Ikeda & Satou, 2017). The mechanism by which expression of regulatory genes is terminated will be required to be analyzed more extensively in the future. In this review, we described gene expression in two values: expressed or not expressed. In other words, the functions of the GRN have been

26

Yutaka Satou

explained in a binary space. This is likely to indicate that the ascidian early embryo does not regulate gene expression rigorously in a quantitative manner in most cases. This property may not be specific to ascidian embryos, as the endomesodermal GRN in early sea urchin embryos is represented by Boolean functions (Peter, Faure, & Davidson, 2012). Meanwhile, quantitative regulation may be important for non-regulatory genes. In the notochord precursors, where T is specifically expressed, many non-regulatory genes are regulated directly or indirectly by T (Kubo et al., 2010), and genes with more T-binding sites tend to be expressed earlier (Katikala et al., 2013). In the muscle lineage cells of early embryos, some muscle structural genes begin to be expressed under the control of Tbx6-r.b. It has been suggested that the expression of genes that bind to Tbx6-r.b more extensively begins earlier (Yu et al., 2019). Thus, quantitative regulation may be more common for non-regulatory genes in late stage embryos. Cell properties emerge from GRN dynamics, which are constrained by GRN structure. To understand how these dynamics emerge remains a challenge. One approach would be to determine Boolean or quantitative regulatory functions for individual genes and to analyze behavior of the network (Peter et al., 2012; Piran, Halperin, Guttmann-Raviv, Keinan, & Reshef, 2009). Another approach would involve predicting dynamics directly from the network structures (Kobayashi, Maeda, Tokuoka, Mochizuki, & Satou, 2018; Liu, Slotine, & Barabasi, 2011; Mochizuki, 2016; Zanudo, Yang, & Albert, 2017). The first approach is straightforward and may provide more precise predictions. However, the experimental determination of regulatory functions is often difficult, and may therefore be an unrealistic approach for large networks. On the other hand, the second approach could be feasible even in large networks, as it does not require individual regulatory functions. However, it often requires making assumptions and will therefore provide more limited predictions. Nevertheless, both approaches would require the precise determination of the GRN structure, and the ascidian early embryo provides a good experimental platform to this end.

References Abitua, P. B., Gainous, T. B., Kaczmarczyk, A. N., Winchell, C. J., Hudson, C., Kamata, K., et al. (2015). The pre-vertebrate origins of neurogenic placodes. Nature, 524, 462–465. Aniello, F., Locascio, A., Villani, M. G., Di Gregorio, A., Fucci, L., & Branno, M. (1999). Identification and developmental expression of Ci-msxb: A novel homologue of Drosophila msh gene in Ciona intestinalis. Mechanisms of Development, 88, 123–126.

A gene regulatory network in early ascidian embryos

27

Araki, I., Saiga, H., Makabe, K. W., & Satoh, N. (1994). Expression of Amd1, a gene for a Myod1-related factor in the ascidian Halocynthia roretzi. Roux’s Archives of Developmental Biology, 203, 320–327. Bertrand, V., Hudson, C., Caillol, D., Popovici, C., & Lemaire, P. (2003). Neural tissue in ascidian embryos is induced by FGF9/16/20, acting via a combination of maternal GATA and Ets transcription factors. Cell, 115, 615–627. Brunetti, R., Gissi, C., Pennati, R., Caicci, F., Gasparini, F., & Manni, L. (2015). Morphological evidence that the molecularly determined Ciona intestinalis type A and type B are different species: Ciona robusta and Ciona intestinalis. Journal of Zoological Systematics and Evolutionary Research, 53, 186–193. Caracciolo, A., Di Gregorio, A., Aniello, F., Di Lauro, R., & Branno, M. (2000). Identification and developmental expression of three Distal-less homeobox containing genes in the ascidian Ciona intestinalis. Mechanisms of Development, 99, 173–176. Chiba, S., Awazu, S., Itoh, M., Chin-Bow, S. T., Satoh, N., Satou, Y., et al. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. IX. Genes for muscle structural proteins. Development Genes and Evolution, 213, 291–302. Chiba, S., Jiang, D., Satoh, N., & Smith, W. C. (2009). Brachyury null mutant-induced defects in juvenile ascidian endodermal organs. Development, 136, 35–39. Christiaen, L., Stolfi, A., Davidson, B., & Levine, M. (2009). Spatio-temporal intersection of Lhx3 and Tbx6 defines the cardiac field through synergistic activation of Mesp. Developmental Biology, 328, 552–560. Conklin, E. G. (1905a). Mosaic development in ascidian eggs. The Journal of Experimental Zoology, 2, 145–223. Conklin, E. G. (1905b). Organ-forming substances in the eggs of ascidians. The Biological Bulletin, 8, 205–230. Conklin, E. G. (1905c). The organization and cell-lineage of the ascidian egg. Journal of the Academy of Natural Sciences of Philadelphia, 13, 1–119. Corbo, J. C., Fujiwara, S., Levine, M., & Di Gregorio, A. (1998). Suppressor of hairless activates brachyury expression in the Ciona embryo. Developmental Biology, 203, 358–368. Corbo, J. C., Levine, M., & Zeller, R. W. (1997). Characterization of a notochord-specific enhancer from the Brachyury promoter region of the ascidian, Ciona intestinalis. Development, 124, 589–602. Davidson, B., Shi, W. Y., Beh, J., Christiaen, L., & Levine, M. (2006). FGF signaling delineates the cardiac progenitor field in the simple chordate, Ciona intestinalis. Genes & Development, 20, 2728–2738. Dehal, P., Satou, Y., Campbell, R. K., Chapman, J., Degnan, B., De Tomaso, A., et al. (2002). The draft genome of Ciona intestinalis: Insights into chordate and vertebrate origins. Science, 298, 2157–2167. Delaune, E., Lemaire, P., & Kodjabachian, L. (2005). Neural induction in Xenopus requires early FGF signalling in addition to BMP inhibition. Development, 132, 299–310. Dumollard, R., Hebras, C., Besnardeau, L., & McDougall, A. (2013). Beta-catenin patterns the cell cycle during maternal-to-zygotic transition in urochordate embryos. Developmental Biology, 384, 331–342. Hashimoto, H., Enomoto, T., Kumano, G., & Nishida, H. (2011). The transcription factor FoxB mediates temporal loss of cellular competence for notochord induction in ascidian embryos. Development, 138, 2591–2600. Haupaix, N., Stolfi, A., Sirour, C., Picco, V., Levine, M., Christiaen, L., et al. (2013). p120RasGAP mediates ephrin/Eph-dependent attenuation of FGF/ERK signals during cell fate specification in ascidian embryos. Development, 140, 4347–4352. Hino, K., Satou, Y., Yagi, K., & Satoh, N. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. VI. Genes for Wnt, TGFbeta, Hedgehog and JAK/STAT signaling pathways. Development Genes and Evolution, 213, 264–272.

28

Yutaka Satou

Hotta, K., Takahashi, H., Erives, A., Levine, M., & Satoh, N. (1999). Temporal expression patterns of 39 Brachyury-downstream genes associated with notochord formation in the Ciona intestinalis embryo. Development, Growth & Differentiation, 41, 657–664. Hudson, C., Darras, S., Caillol, D., Yasuo, H., & Lemaire, P. (2003). A conserved role for the MEK signalling pathway in neural tissue specification and posteriorisation in the invertebrate chordate, the ascidian Ciona intestinalis. Development, 130, 147–159. Hudson, C., Kawai, N., Negishi, T., & Yasuo, H. (2013). β-catenin-driven binary fate specification segregates germ layers in ascidian embryos. Current Biology, 23, 491–495. Hudson, C., & Lemaire, P. (2001). Induction of anterior neural fates in the ascidian Ciona intestinalis. Mechanisms of Development, 100, 189–203. Hudson, C., Sirour, C., & Yasuo, H. (2016). Co-expression of Foxa.a, Foxd and Fgf9/16/20 defines a transient mesendoderm regulatory state in ascidian embryos. eLife, 5, e14692. Hudson, C., & Yasuo, H. (2005). Patterning across the ascidian neural plate by lateral Nodal signalling sources. Development, 132, 1199–1210. Hudson, C., & Yasuo, H. (2006). A signalling relay involving Nodal and Delta ligands acts during secondary notochord induction in Ciona embryos. Development, 133, 2855–2864. Ikeda, T., Matsuoka, T., & Satou, Y. (2013). A time delay gene circuit is required for palp formation in the ascidian embryo. Development, 140, 4703–4708. Ikeda, T., & Satou, Y. (2017). Differential temporal control of Foxa.a and Zic-r.b specifies brain versus notochord fate in the ascidian embryo. Development, 144, 38–43. Imai, K. S., Daido, Y., Kusakabe, T. G., & Satou, Y. (2012). Cis-acting transcriptional repression establishes a sharp boundary in chordate embryos. Science, 337, 964–967. Imai, K. S., Hikawa, H., Kobayashi, K., & Satou, Y. (2017). Tfap2 and Sox1/2/3 cooperatively specify ectodermal fates in ascidian embryos. Development, 144, 33–37. Imai, K. S., Hino, K., Yagi, K., Satoh, N., & Satou, Y. (2004). Gene expression profiles of transcription factors and signaling molecules in the ascidian embryo: Towards a comprehensive understanding of gene networks. Development, 131, 4047–4058. Imai, K. S., Hudson, C., Oda-Ishii, I., Yasuo, H., & Satou, Y. (2016). Antagonism between β-catenin and Gata.a sequentially segregates the germ layers of ascidian embryos. Development, 143, 4167–4172. Imai, K. S., Levine, M., Satoh, N., & Satou, Y. (2006). Regulatory blueprint for a chordate embryo. Science, 312, 1183–1187. Imai, K., Satoh, N., & Satou, Y. (2002a). Early embryonic expression of FGF4/6/9 gene and its role in the induction of mesenchyme and notochord in Ciona savignyi embryos. Development, 129, 1729–1738. Imai, K. S., Satoh, N., & Satou, Y. (2002b). An essential role of a FoxD gene in notochord induction in Ciona embryos. Development, 129, 3441–3453. Imai, K., Satoh, N., & Satou, Y. (2003). A Twist-like bHLH gene is a downstream factor of an endogenous FGF and determines mesenchymal fate in the ascidian embryos. Development, 130, 4461–4472. Imai, K. S., Satou, Y., & Satoh, N. (2002). Multiple functions of a Zic-like gene in the differentiation of notochord, central nervous system and muscle in Ciona savignyi embryos. Development, 129, 2723–2732. Imai, K., Takada, N., Satoh, N., & Satou, Y. (2000). β-catenin mediates the specification of endoderm cells in ascidian embryos. Development, 127, 3009–3020. Irvine, S. Q., Cangiano, M. C., Millette, B. J., & Gutter, E. S. (2007). Non-overlapping expression patterns of the clustered Dll-A/B genes in the ascidian Ciona intestinalis. Journal of Experimental Zoology. Part B, Molecular and Developmental Evolution, 308, 428–441. Irvine, S. Q., Vierra, D. A., Millette, B. J., Blanchette, M. D., & Holbert, R. E. (2011). Expression of the Distalless-B gene in Ciona is regulated by a pan-ectodermal enhancer module. Developmental Biology, 353, 432–439.

A gene regulatory network in early ascidian embryos

29

Katikala, L., Aihara, H., Passamaneck, Y. J., Gazdoiu, S., Jose-Edwards, D. S., Kugler, J. E., et al. (2013). Functional Brachyury binding sites establish a temporal read-out of gene expression in the Ciona notochord. PLoS Biology, 11, e1001697. Kawai, N., Iida, Y., Kumano, G., & Nishida, H. (2007). Nuclear accumulation of betacatenin and transcription of downstream genes are regulated by zygotic Wnt5alpha and maternal Dsh in ascidian embryos. Developmental Dynamics, 236, 1570–1582. Kawashima, T., Tokuoka, M., Awazu, S., Satoh, N., & Satou, Y. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. VIII. Genes for PI3K signaling and cell cycle. Development Genes and Evolution, 213, 284–290. Khoueiry, P., Rothbacher, U., Ohtsuka, Y., Daian, F., Frangulian, E., Roure, A., et al. (2010). A cis-regulatory signature in ascidians and flies, independent of transcription factor binding sites. Current Biology, 20, 792–802. Kim, G. J., Kumano, G., & Nishida, H. (2007). Cell fate polarization in ascidian mesenchyme/muscle precursors by directed FGF signaling and role for an additional ectodermal FGF antagonizing signal in notochord/nerve cord precursors. Development, 134, 1509–1518. Kobayashi, K., Maeda, K., Tokuoka, M., Mochizuki, A., & Satou, Y. (2018). Controlling cell fate specification system by key genes determined from network structure. iScience, 4, 281–293. Kubo, A., Imai, K. S., & Satou, Y. (2009). Gene-regulatory networks in the Ciona embryos. Briefings in Functional Genomics & Proteomics, 8, 250–255. Kubo, A., Suzuki, N., Yuan, X., Nakai, K., Satoh, N., Imai, K. S., et al. (2010). Genomic cisregulatory networks in the early Ciona intestinalis embryo. Development, 137, 1613–1623. Kugler, J. E., Gazdoiu, S., Oda-Ishii, I., Passamaneck, Y. J., Erives, A. J., & Di Gregorio, A. (2010). Temporal regulation of the muscle gene cascade by Macho1 and Tbx6 transcription factors in Ciona intestinalis. Journal of Cell Science, 123, 2453–2463. Kumano, G., Takatori, N., Negishi, T., Takada, T., & Nishida, H. (2011). A maternal factor unique to ascidians silences the germline via binding to P-TEFb and RNAP II regulation. Current Biology, 21, 1308–1313. Lamy, C., Rothbacher, U., Caillol, D., & Lemaire, P. (2006). Ci-FoxA-a is the earliest zygotic determinant of the ascidian anterior ectoderm and directly activates Ci-sFRP1/5. Development, 133, 2835–2844. Lemaire, P. (2009). Unfolding a chordate developmental program, one cell at a time: Invariant cell lineages, short-range inductions and evolutionary plasticity in ascidians. Developmental Biology, 332, 48–60. Lemaire, P. (2011). Evolutionary crossroads in developmental biology: The tunicates. Development, 138, 2143–2152. Liu, B., & Satou, Y. (2019). Foxg specifies sensory neurons in the anterior neural plate border of the ascidian embryo. Nature Communications, 10, 4911. Liu, Y. Y., Slotine, J. J., & Barabasi, A. L. (2011). Controllability of complex networks. Nature, 473, 167–173. Longabaugh, W. J. (2012). BioTapestry: A tool to visualize the dynamic properties of gene regulatory networks. Methods in Molecular Biology, 786, 359–394. Marchal, L., Luxardi, G., Thome, V., & Kodjabachian, L. (2009). BMP inhibition initiates neural induction via FGF signaling and Zic genes. Proceedings of the National Academy of Sciences of the United States of America, 106, 17437–17442. Meedel, T. H., Chang, P., & Yasuo, H. (2007). Muscle development in Ciona intestinalis requires the b-HLH myogenic regulatory factor gene Ci-MRF. Developmental Biology, 302, 333–344. Meedel, T. H., Farmer, S. C., & Lee, J. J. (1997). The single MyoD family gene of Ciona intestinalis encodes two differentially expressed proteins: Implications for the evolution of chordate muscle gene regulation. Development, 124, 1711–1721.

30

Yutaka Satou

Miwata, K., Chiba, T., Horii, R., Yamada, L., Kubo, A., Miyamura, D., et al. (2006). Systematic analysis of embryonic expression profiles of zinc finger genes in Ciona intestinalis. Developmental Biology, 292, 546–554. Mochizuki, A. (2016). Theoretical approaches for the dynamics of complex biological systems from information of networks. Proceedings of the Japan Academy. Series B, Physical and Biological Sciences, 92, 255–264. Munoz-Sanjuan, I., & Brivanlou, A. H. (2002). Neural induction, the default model and embryonic stem cells. Nature Reviews. Neuroscience, 3, 271–280. Nishida, H. (1987). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. III. Up to the tissue restricted stage. Developmental Biology, 121, 526–541. Nishida, H. (2003). Spatio-temporal pattern of MAP kinase activation in embryos of the ascidian Halocynthia roretzi. Development, Growth & Differentiation, 45, 27–37. Nishida, H. (2005). Specification of embryonic axis and mosaic development in ascidians. Developmental Dynamics, 233, 1177–1193. Nishida, H., & Satoh, N. (1983). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. I. Up to the eight-cell stage. Developmental Biology, 99, 382–394. Nishida, H., & Satoh, N. (1985). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. II. The 16- and 32-cell stages. Developmental Biology, 110, 440–454. Nishida, H., & Sawada, K. (2001). macho-1 encodes a localized mRNA in ascidian eggs that specifies muscle fate during embryogenesis. Nature, 409, 724–729. Nishikata, T., Hibino, T., & Nishida, H. (1999). The centrosome-attracting body, microtubule system, and posterior egg cytoplasm are involved in positioning of cleavage planes in the ascidian embryo. Developmental Biology, 209, 72–85. Oda-Ishii, I., Abe, T., & Satou, Y. (2018). Dynamics of two key maternal factors that initiate zygotic regulatory programs in ascidian embryos. Developmental Biology, 437, 50–59. Oda-Ishii, I., Kubo, A., Kari, W., Suzuki, N., Rothbacher, U., & Satou, Y. (2016). A maternal system initiating the zygotic developmental program through combinatorial repression in the ascidian embryo. PLoS Genetics, 12, e1006045. Oda-Ishii, I., & Satou, Y. (2018). Initiation of the zygotic genetic program in the ascidian embryo. Seminars in Cell & Developmental Biology, 84, 111–117. Ohta, N., & Satou, Y. (2013). Multiple signaling pathways coordinate to induce a threshold response in a chordate embryo. PLoS Genetics, 9, e1003818. Ohta, N., Waki, K., Mochizuki, A., & Satou, Y. (2015). A Boolean function for neural induction reveals a critical role of direct intercellular interactions in patterning the ectoderm of the ascidian embryo. PLoS Computational Biology, 11, e1004687. Pasini, A., Amiel, A., Rothbacher, U., Roure, A., Lemaire, P., & Darras, S. (2006). Formation of the ascidian epidermal sensory neurons: Insights into the origin of the chordate peripheral nervous system. PLoS Biology, 4, e225. Peter, I. S., Faure, E., & Davidson, E. H. (2012). Predictive computation of genomic logic processing functions in embryonic development. Proceedings of the National Academy of Sciences of the United States of America, 109, 16434–16442. Picco, V., Hudson, C., & Yasuo, H. (2007). Ephrin-Eph signalling drives the asymmetric division of notochord/neural precursors in Ciona embryos. Development, 134, 1491–1497. Piran, R., Halperin, E., Guttmann-Raviv, N., Keinan, E., & Reshef, R. (2009). Algorithm of myogenic differentiation in higher-order organisms. Development, 136, 3831–3840. Prodon, F., Chenevert, J., & Sardet, C. (2006). Establishment of animal-vegetal polarity during maturation in ascidian oocytes. Developmental Biology, 290, 297–311.

A gene regulatory network in early ascidian embryos

31

Ristoratore, F., Spagnuolo, A., Aniello, F., Branno, M., Fabbrini, F., & Di Lauro, R. (1999). Expression and functional analysis of Citiff1, an ascidian NK-2 class gene, suggest its role in endoderm development. Development, 126, 5149–5159. Rothb€acher, U., Bertrand, V., Lamy, C., & Lemaire, P. (2007). A combinatorial code of maternal GATA, Ets and β-catenin-TCF transcription factors specifies and patterns the early ascidian ectoderm. Development, 134, 4023–4032. Roure, A., & Darras, S. (2016). Msxb is a core component of the genetic circuitry specifying the dorsal and ventral neurogenic midlines in the ascidian embryo. Developmental Biology, 409, 277–287. Roure, A., Lemaire, P., & Darras, S. (2014). An otx/nodal regulatory signature for posterior neural development in ascidians. PLoS Genetics, 10, e1004548. Sasakura, Y., Shoguchi, E., Takatori, N., Wada, S., Meinertzhagen, I. A., Satou, Y., et al. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. X. Genes for cell junctions and extracellular matrix. Development Genes and Evolution, 213, 303–313. Sasakura, Y., Yamada, L., Takatori, N., Satou, Y., & Satoh, N. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. VII. Molecules involved in the regulation of cell polarity and actin dynamics. Development Genes and Evolution, 213, 273–283. Satoh, N., Araki, I., & Satou, Y. (1996). An intrinsic genetic program for autonomous differentiation of muscle cells in the ascidian embryo. Proceedings of the National Academy of Sciences of the United States of America, 93, 9315–9321. Satou, Y., & Imai, K. S. (2015). Gene regulatory systems that control gene expression in the Ciona embryo. Proceedings of the Japan Academy. Series B, Physical and Biological Sciences, 91, 33–51. Satou, Y., Imai, K. S., Levine, M., Kohara, Y., Rokhsar, D., & Satoh, N. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. I. Genes for bHLH transcription factors. Development Genes and Evolution, 213, 213–221. Satou, Y., Imai, K., & Satoh, N. (2001a). Action of morpholinos in Ciona embryos. Genesis, 30, 103–106. Satou, Y., Imai, K. S., & Satoh, N. (2001b). Early embryonic expression of a LIM-homeobox gene Cs-lhx3 is downstream of beta-catenin and responsible for the endoderm differentiation in Ciona savignyi embryos. Development, 128, 3559–3570. Satou, Y., Imai, K., & Satoh, N. (2004). The ascidian Mesp gene specifies heart precursor cells. Development, 131, 2533–2541. Satou, Y., Mineta, K., Ogasawara, M., Sasakura, Y., Shoguchi, E., Ueno, K., et al. (2008). Improved genome assembly and evidence-based global gene model set for the chordate Ciona intestinalis: New insight into intron and operon populations. Genome Biology, 9, R152. Satou, Y., Nakamura, R., Yu, D., Yoshida, R., Hamada, M., Fujie, M., et al. (2019). A nearly complete genome of Ciona intestinalis type A (C. robusta) reveals the contribution of inversion to chromosomal evolution in the genus Ciona. Genome Biology and Evolution, 11, 3144–3157. Satou, Y., Sasakura, Y., Yamada, L., Imai, K. S., Satoh, N., & Degnan, B. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. V. Genes for receptor tyrosine kinase pathway and Notch signaling pathway. Development Genes and Evolution, 213, 254–263. Satou, Y., & Satoh, N. (1996). Two cis-regulatory elements are essential for the musclespecific expression of an actin gene in the ascidian embryo. Development, Growth & Differentiation, 38, 565–573.

32

Yutaka Satou

Satou, Y., & Satoh, N. (2005). Cataloging transcription factor and major signaling molecule genes for functional genomic studies in Ciona intestinalis. Development Genes and Evolution, 215, 580–596. Satou, Y., Satoh, N., & Imai, K. S. (2009). Gene regulatory networks in the early ascidian embryo. Biochimica et Biophysica Acta, 1789, 268–273. Satou, Y., Wada, S., Sasakura, Y., & Satoh, N. (2008). Regulatory genes in the ancestral chordate genomes. Development Genes and Evolution, 218, 715–721. Satou, Y., Yagi, K., Imai, K. S., Yamada, L., Nishida, H., & Satoh, N. (2002). macho-1Related genes in Ciona embryos. Development Genes and Evolution, 212, 87–92. Shi, W., & Levine, M. (2008). Ephrin signaling establishes asymmetric cell fates in an endomesoderm lineage of the Ciona embryo. Development, 135, 931–940. Shirae-Kurabayashi, M., Matsuda, K., & Nakamura, A. (2011). Ci-Pem-1 localizes to the nucleus and represses somatic gene transcription in the germline of Ciona intestinalis embryos. Development, 138, 2871–2881. Spagnuolo, A., & Di Lauro, R. (2002). Cititf1 and endoderm differentiation in Ciona intestinalis. Gene, 287, 115–119. Stolfi, A., Ryan, K., Meinertzhagen, I. A., & Christiaen, L. (2015). Migratory neuronal progenitors arise from the neural plate borders in tunicates. Nature, 527, 371–374. Streit, A., Berliner, A. J., Papanayotou, C., Sirulnik, A., & Stern, C. D. (2000). Initiation of neural induction by FGF signalling before gastrulation. Nature, 406, 74–78. Takahashi, H., Hotta, K., Erives, A., Di Gregorio, A., Zeller, R. W., Levine, M., et al. (1999). Brachyury downstream notochord differentiation in the ascidian embryo. Genes & Development, 13, 1519–1523. Tassy, O., Daian, F., Hudson, C., Bertrand, V., & Lemaire, P. (2006). A quantitative approach to the study of cell shapes and interactions during early chordate embryogenesis. Current Biology, 16, 345–358. Tassy, O., Dauga, D., Daian, F., Sobral, D., Robin, F., Khoueiry, P., et al. (2010). The ANISEED database: Digital representation, formalization, and elucidation of a chordate developmental program. Genome Research, 20, 1459–1468. Tokuhiro, S., Tokuoka, M., Kobayashi, K., Kubo, A., Oda-Ishii, I., & Satou, Y. (2017). Differential gene expression along the animal-vegetal axis in the ascidian embryo is maintained by a dual functional protein Foxd. PLoS Genetics, 13, e1006741. Tokuoka, M., Imai, K. S., Satou, Y., & Satoh, N. (2004). Three distinct lineages of mesenchymal cells in Ciona intestinalis embryos demonstrated by specific gene expression. Developmental Biology, 274, 211–224. Tokuoka, M., Kobayashi, K., & Satou, Y. (2018). Distinct regulation of Snail in two muscle lineages of the ascidian embryo achieves temporal coordination of muscle development. Development, 145, dev163915. Tokuoka, M., Satoh, N., & Satou, Y. (2005). A bHLH transcription factor gene, Twist-like1, is essential for the formation of mesodermal tissues of Ciona juveniles. Developmental Biology, 288, 387–396. Tresser, J., Chiba, S., Veeman, M., El-Nachef, D., Newman-Smith, E., Horie, T., et al. (2010). doublesex/mab3 related-1 (dmrt1) is essential for development of anterior neural plate derivatives in Ciona. Development, 137, 2197–2203. Wada, S., & Saiga, H. (2002). HrzicN, a new Zic family gene of ascidians, plays essential roles in the neural tube and notochord development. Development, 129, 5597–5608. Wada, S., Tokuoka, M., Shoguchi, E., Kobayashi, K., Di Gregorio, A., Spagnuolo, A., et al. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. II. Genes for homeobox transcription factors. Development Genes and Evolution, 213, 222–234. Wagner, E., & Levine, M. (2012). FGF signaling establishes the anterior border of the Ciona neural tube. Development, 139, 2351–2359.

A gene regulatory network in early ascidian embryos

33

Waki, K., Imai, K. S., & Satou, Y. (2015). Genetic pathways for differentiation of the peripheral nervous system in ascidians. Nature Communications, 6, 8719. Wilson, S. I., Graziano, E., Harland, R., Jessell, T. M., & Edlund, T. (2000). An early requirement for FGF signalling in the acquisition of neural cell fate in the chick embryo. Current Biology, 10, 421–429. Yagi, K., Satoh, N., & Satou, Y. (2004). Identification of downstream genes of the ascidian muscle determinant gene Ci-macho1. Developmental Biology, 274, 478–489. Yagi, K., Satou, Y., Mazet, F., Shimeld, S. M., Degnan, B., Rokhsar, D., et al. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. III. Genes for Fox, ETS, nuclear receptors and NFkappaB. Development Genes and Evolution, 213, 235–244. Yagi, K., Satou, Y., & Satoh, N. (2004). A zinc finger transcription factor, ZicL, is a direct activator of Brachyury in the notochord specification of Ciona intestinalis. Development, 131, 1279–1288. Yamada, L., Kobayashi, K., Degnan, B., Satoh, N., & Satou, Y. (2003). A genomewide survey of developmentally relevant genes in Ciona intestinalis. IV. Genes for HMG transcriptional regulators, bZip and GATA/Gli/Zic/Snail. Development Genes and Evolution, 213, 245–253. Yasuo, H., & Satoh, N. (1993). Function of vertebrate T gene. Nature, 364, 582–583. Yasuo, H., & Satoh, N. (1998). Conservation of the developmental role of Brachyury in notochord formation in a urochordate, the ascidian Halocynthia roretzi. Developmental Biology, 200, 158–170. Yoshida, S., Marikawa, Y., & Satoh, N. (1996). Posterior end mark, a novel maternal gene encoding a localized factor in the ascidian embryo. Development, 122, 2005–2012. Yu, D., Oda-Ishii, I., Kubo, A., & Satou, Y. (2019). The regulatory pathway from genes directly activated by maternal factors to muscle structural genes in ascidian embryos. Development, 146, dev173104. Zanudo, J. G. T., Yang, G., & Albert, R. (2017). Structure-based control of complex networks with nonlinear dynamics. Proceedings of the National Academy of Sciences of the United States of America, 114, 7234–7239.

CHAPTER TWO

Early Xenopus gene regulatory programs, chromatin states, and the role of maternal transcription factors Kitt D. Paraisoa,b, Jin S. Choa, Junseok Yonga, Ken W.Y. Choa,b,∗ a

Department of Developmental and Cell Biology, University of California, Irvine, CA, United States Center for Complex Biological Systems, University of California, Irvine, CA, United States Corresponding author: e-mail address: [email protected]

b ∗

Contents 1. Introduction 2. Roles of maternal TFs during germ layer specification 2.1 Localization of maternal gene products 2.2 Endodermally enriched TFs 2.3 Ectodermally enriched TFs 2.4 Ubiquitously expressed TFs 2.5 Intracellular mediators of signaling pathways 2.6 Other TFs 3. Enhancers, promoters and chromatin states 3.1 Genomic approaches to identifying enhancers 3.2 Chromatin states and ZGA 3.3 Repressive chromatin marking 4. Network structure analysis of the early Xenopus GRN 5. Summary and prospects Acknowledgments References

36 36 36 41 42 43 45 46 47 47 49 50 52 54 55 55

Abstract For decades, the early development of the Xenopus embryo has been an essential model system to study the gene regulatory mechanisms that govern cellular specification. At the top of the hierarchy of gene regulatory networks, maternally deposited transcription factors initiate this process and regulate the expression of zygotic genes that give rise to three distinctive germ layer cell types (ectoderm, mesoderm, and endoderm), and subsequent generation of organ precursors. The onset of germ layer specification is also closely coupled with changes associated with chromatin modifications. This review will examine the timing of maternal transcription factors initiating the zygotic genome activation, the epigenetic landscape of embryonic chromatin, and the network structure that governs the process. Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.02.009

#

2020 Elsevier Inc. All rights reserved.

35

36

Kitt D. Paraiso et al.

1. Introduction After fertilization, the embryonic genome is inactive until transcription is initiated during the maternal-to-zygotic transition, whereby the onset of embryonic genome transcription is called zygotic genome activation (ZGA). At present, the functional relationship among transcription factors (TFs), co-regulators and the epigenetic landscape around ZGA is still poorly understood and several major questions remain. For instance, what combination of maternal TFs contributes to the initiation of genome activation? How do the maternal TFs bind to the chromatin and coordinate the opening or closing of chromatin for their accessibility? How are the chromatin states in the form of histone and DNA modifications established during ZGA? What kinds of network structures are operational at the early stage of development to ensure dynamic changes in gene expression among different cell types? The Xenopus species are model organisms well-suited to address these critical questions. They are highly tolerant toward RNA and DNA microinjection to obtain knockdown or overexpression phenotypes to uncover TF functions. The genomes of the traditionally used pseudotetraploid Xenopus laevis (Session et al., 2016) and the diploid cousin Xenopus tropicalis (Hellsten et al., 2010; Mitros et al., 2019) have been sequenced, which makes both amenable to a variety of high-throughput genomic approaches. In addition, the embryos of both species are amenable to CRISPR-Cas9 mutagenesis-based approaches (Aslan, Tadjuidje, Zorn, & Cha, 2017; Blitz, Biesinger, Xie, & Cho, 2013; Blitz, Fish, & Cho, 2016; Nakayama et al., 2013; Shi et al., 2015; Wang et al., 2015). In this review, we will cover the essential players in the first network connections in the vertebrate gene regulatory network (GRN) occurring at the onset of ZGA. First, we will discuss the maternal TFs and signaling molecules that confer germ layerspecific GRNs. Second, we will highlight the cis-regulatory regions in the form of enhancers and promoters with particular focus on findings from the genomic data. Finally, we will approach the GRN from a systems biological perspective and discuss the network architecture of the early Xenopus GRN.

2. Roles of maternal TFs during germ layer specification 2.1 Localization of maternal gene products After fertilization, the Xenopus zygote undergoes multiple rounds of division to give rise to smaller cells (blastomeres) without increasing the overall size of

Early Xenopus gene regulatory programs

37

the embryo. During the early stages of this process, individual blastomeres are pluripotent and remain uncommitted to specific lineages (Heasman, Wylie, Hausen, & Smith, 1984; Snape, Wylie, Smith, & Heasman, 1987). Gradually, cells along the animal-vegetal axis acquire germ layer identities. Specifically, ectoderm is formed in the animal cap (top side of the embryo), endoderm is formed in the vegetal mass, and mesoderm is induced at the equatorial region between the animal and the vegetal poles of the Xenopus embryo. During this process, maternally deposited mRNAs and proteins are inherited into individual blastomeres. Some maternal products show localized expression, while the others are uniformly distributed along the animal-vegetal axis (Fig. 1). The localization of these maternal products controls the germ layer cell identities by initiating their respective GRN programs. Multiple large-scale (Cuykendall & Houston, 2010; Flachsova, Sindelka, & Kubista, 2013) and genome-wide (De Domenico, Owens, Grant, Gomes-Faria, & Gilchrist, 2015; Paraiso et al., 2019) screens identified hundreds of genes with localized expression, which includes a little over a dozen TFs and signaling molecules in both Xenopus laevis and Xenopus tropicalis embryos. Among these TFs are otx1, vegt and sox7 gene products, which are enriched in the vegetal tissue, whereas the foxi2 and grhl1 gene products are enriched animally. Additionally, ubiquitously expressed foxh1, sox3, pou5f3 (oct60) TFs, and mediators of TGF-β (Smad2/3) signaling and Wnt (ctnnb1/β-catenin and tcf7l1) signaling play central roles in specifying the identities of germ layer-specific gene regulatory programs. We will discuss the roles of these TFs in germ layer specification (summarized in Table 1).

Fig. 1 Maternal transcription factors are differentially inherited by blastomeres during cleavage stages. Gene products in the animal blastomeres are inherited by ectodermal cells, while those in the vegetal blastomeres are inherited by mesendodermal cells. Gene products such as Foxh1 are ubiquitously expressed in the oocyte and are inherited by blastomeres across the animal-vegetal axis.

Table 1 Maternal TFs and their gene regulatory function during ZGA and germ layer formation. Maternal expression in Maternal TF the animal-vegetal axis Gene regulatory function References

Vegt

Vegetal pole

• Direct activator of endodermal genes

Lustig, Kroll, Sun, and Kirschner (1996), Zhang and King (1996), Stennard, Carnac, and Gurdon (1996), Horb and Thomsen (1997), Zhang et al. (1998), Kofron et al. (1999), Taverner et al. (2005) and Paraiso et al. (2019)

• Direct repressor of ectodermal and mesodermal genes

• Regulates RNA Polymerase II binding and transcription in enhancers

• Co-binds with Foxh1 and Otx1 to site-select endodermal enhancers and super-enhancers Otx1

Vegetal pole

• Direct activator of endodermal genes

Pannese, Cagliani, Pardini, and Boncinelli (2000) and Paraiso et al. (2019)

• Direct repressor of ectodermal and mesodermal genes

• Synergizes/antagonizes with Vegt to regulate genes

• Regulates RNA Polymerase II binding and transcription in enhancers

• Co-binds with Foxh1 and Vegt to site-select endodermal enhancers and super-enhancers Sox7

Vegetal pole

• Activates endodermal genes • Co-binds with Foxh1 and Smad2,3 near dorsal mesendodermal genes

Zhang, Basta, Fawcett, and Klymkowsky (2005) and Charney, Forouzmand, et al. (2017)

Foxi2

Animal pole

• Activates ectodermal genes

Cha, McAdams, Kormish, Wylie, and Kofron (2012)

Grhl1

Animal pole

• Unknown (maternally)

Tao, Kuliyev, et al. (2005)

• Zygotic Grhl1 regulates epidermal ectoderm formation Foxh1

Ubiquitous

• TF for the Nodal signaling pathway

Chen, Rubock, and Whitman (1996), Chen et al. (1997), Kofron et al. (2004), Chiu et al. (2014), Reid et al. (2016), Charney, Forouzmand, et al. (2017) and Paraiso et al. (2019)

• Directly activates mesendodermal genes • Acts as dual TF to regulate gene expression • Regulates RNA Polymerase II binding and transcription in enhancers

• Co-regulates with a variety of TFs in diverse spatial coordinates in the early embryo including Vegt, Otx1, Sox7 and Smad2,3 to site-select enhancers and super-enhancers Sox3

Ubiquitous

• Suppresses Nodal signaling in the ectoderm

Zhang, Basta, Jensen, and Klymkowsky (2003), Zhang and Klymkowsky (2007) and Gentsch, Spruce, Owens, and Smith (2019)

• Co-regulates gene expression with PouV TFs • Regulates zygotic TF binding • Mediates chromosome conformation and accessibility changes during ZGA Continued

Table 1 Maternal TFs and their gene regulatory function during ZGA and germ layer formation.—cont’d Maternal expression in Maternal TF the animal-vegetal axis Gene regulatory function References

Pou5f3

Ubiquitous

• Suppresses Nodal signaling

Cao, Siegel, and Kn€ ochel (2006), Cao et al. (2007), Chiu et al. (2014) and Gentsch et al. (2019)

• Suppresses the function of tissue-specific factors such as Vegt and Ctnnb1

• Co-regulates gene expression with Sox3 • Regulates zygotic TF binding • Mediates chromosome conformation and accessibility changes during ZGA Smad2,3

Ubiquitous

• Mediates Nodal signaling pathway with Foxh1 Birsoy, Kofron, Schaible, Wylie, and Heasman (2006) and Chiu et al. (2014)

• Activates endoderm gene expression • Induces mesoderm formation Tcf/Ctnnb1 Ubiquitous

• Specifies the dorsal identities in the embryo

• Co-regulates with Nodal signaling to specify the dorsal mesendodermal fates

Ku and Melton (1993), Tao, Yokota, et al. (2005), Brannon, Gomperts, Sumoy, Moon, and Kimelman (1997), Laurent, Blitz, Hashimoto, Rothb€acher, and Cho (1997) and Nakamura, de Paiva Alves, Veenstra, and Hoppler (2016)

Early Xenopus gene regulatory programs

41

2.2 Endodermally enriched TFs 2.2.1 Vegt One of the most well-studied maternal determinants of germ layer specification in the Xenopus embryo is the vegetally localized T-box TF, Vegt (previously also called Xombi, Antipodean, or Brat) (Horb & Thomsen, 1997; Lustig et al., 1996; Stennard et al., 1996; Zhang & King, 1996). Gain-of-function of Vegt in the putative ectoderm exhibits strong endodermal-inductive properties, while loss-of-function of endogenous Vegt results in conversion of the putative endodermal cells into mesoand ectodermal cell fates (Kofron et al., 1999; Zhang et al., 1998). This is consistent with the finding that Vegt regulates the expression of Nodal ligands (Kofron et al., 1999), which are one of the earliest transcribed zygotic genes necessary for the initiation of both the endoderm and mesoderm (reviewed by Schier, 2003). In addition to its role as an activator of transcription, Vegt is involved in repression of genes. Gain-of-function in ectodermal cells caused significant down-regulation of ectodermal genes (Taverner et al., 2005), while Vegt loss-of-function in endodermal cells caused up-regulation of meso- and ectodermal genes (Paraiso et al., 2019). The genome-wide binding of Vegt also supports the dual function of Vegt, as Vegt occupies both cis-regulatory modules (CRMs) of mesendoderm and ectodermal genes (Gentsch et al., 2013; Paraiso et al., 2019). Following the discovery of Vegt, the role of T-box TFs during early endoderm formation has become well appreciated in a variety of vertebrates including axolotl (Nath & Elinson, 2007; Perez et al., 2007), zebrafish (Bjornson et al., 2005; Xu et al., 2014), and human (Teo et al., 2011). 2.2.2 Otx1 Otx1 was previously shown to be maternally deposited and vegetally localized (De Domenico et al., 2015; Pannese et al., 2000), and the role of Otx1 in endoderm formation has been recently described (Paraiso et al., 2019). Otx1 is a critical collaborator of Vegt during endoderm formation. ChIP-seq analysis shows that Otx1 can bind to selected mesendoderm CRMs during cleavage stages in a manner whereby the majority of Otx1 binding sites overlap with Vegt binding (Paraiso et al., 2019). Interestingly, whereas Vegt can activate both mesoderm and endoderm gene expression, Otx1 only activates endoderm genes while concurrently repressing mesodermal genes. How these additive/synergistic and antagonistic

42

Kitt D. Paraiso et al.

interactions are differentiated at the level of CRMs is currently unknown. Importantly, the maternal expression and vegetal localization of vertebrate Otx1 and invertebrate Otx1 homologs appear to be conserved across metazoans (Paraiso et al., 2019). The oocyte expression pattern, along with functional evidence in echinoderms (Hinman, Nguyen, & Davidson, 2003; Peter & Davidson, 2010) and tunicates (Wada & Saiga, 1999) on the role of otx genes, suggests a deep conservation of otx function at least in bilaterian endodermal GRNs. 2.2.3 Sox7 The HMG-domain containing TF Sox7 is maternally expressed locally in the vegetal cells (Owens et al., 2016; Zhang et al., 2005). Sox7 belongs to the F-type sub-family of Sox TFs, which includes a well known zygotically expressed endodermal differentiation factor Sox17. Sox7 regulates the expression of Nodal ligands, endodermal markers such as a2m (formerly endodermin), and endodermal TFs mixer and sox17b (Zhang et al., 2005). Notably, the depletion of Sox7 does not cause major phenotypic disruption on endoderm formation, suggesting that a combinatorial function of maternal TFs is needed during this process. Indeed, Sox7 appears to co-bind with other maternal TFs during early gastrula stages (Charney, Forouzmand, et al., 2017), and is a predicted co-bound factor with Otx1 and Vegt (Paraiso et al., 2019). How Sox7 functionally co-regulates with Otx1, Vegt and other maternal TFs is unknown.

2.3 Ectodermally enriched TFs 2.3.1 Foxi2 Maternally expressed forkhead domain TF Foxi2 is highly enriched in the animal region of the Xenopus embryo (Cha et al., 2012). Foxi2 is required for the expression of ectodermal genes such as lhx5 and cdh1 (e-cadherin). Additionally, maternal Foxi2 has been shown to directly activate the zygotic expression of another Foxi TF, foxi1, by binding to the foxi1 promoter. Like foxi2, foxi1 is an important regulator of ectoderm formation (Mir et al., 2007). How the function of these related Fox TFs overlap is unknown and it will be interesting to dissect the genome-wide roles of foxi2 and foxi1. Particularly, since Foxi2 acts high in the ectodermal gene regulatory hierarchy, Foxi2 ChIP-seq data will uncover important CRMs regulating the ectodermal GRN.

Early Xenopus gene regulatory programs

43

2.3.2 Grhl1 Grainyhead TFs are highly conserved across diverse animal species, and are responsible for epidermal barrier formation. Grh1 is maternally expressed and enriched animally (Paraiso et al., 2019). While not much is known about the role of maternal grhl1, zygotically expressed Grhl1 is essential for ectoderm formation (Tao, Kuliyev, et al., 2005). As one of the few ectodermally localized maternal TFs, analysis of the grhl1 downstream targets in the GRN will be critical in understanding the early ectoderm differentiation process.

2.4 Ubiquitously expressed TFs 2.4.1 Foxh1 Foxh1, a member of the forkhead family, is maternally supplied and expressed ubiquitously in the cleavage stage Xenopus embryo (Kofron et al., 2004). Foxh1 was initially identified as a Nodal signaling co-factor (Chen et al., 1996, 1997), which recruits co-effector Smad2/3 to activate downstream target gene expression. Early studies of Foxh1 function complemented by recent genomics approaches suggest transcriptional roles beyond collaboration with Nodal signaling during mesoderm induction. In the dorsal mesendoderm, Foxh1 collaborates with intracellular mediators of Wnt signaling, Tcf3 and Ctnnb1 (β-catenin), to regulate the expression of Nodal genes (Charney, Forouzmand, et al., 2017; Kofron et al., 2004; Reid et al., 2016). Genome-wide approaches identified further roles of Foxh1 in regulating genes involved in formation of the endoderm such as sox17a (Chiu et al., 2014). Additionally, Foxh1 has been implicated as a repressor of gene expression (Chiu et al., 2014; Reid et al., 2016) to possibly inhibit precocious activation of ZGA. The mechanism by which Foxh1 toggles between these roles is unclear, although hypotheses can be inferred from genome-wide chromatin data (Fig. 2). For tissue-specific functions, Foxh1 appears to co-bind to different sets of TFs. Persistent Foxh1 binding through the blastula and gastrula stages co-localizes with Smad2/3 and Sox7, and this binding is enriched near dorsal mesendoderm genes (Charney, Forouzmand, et al., 2017). In the endoderm, Foxh1 co-binds with endodermal TFs Vegt and Otx1 in the enhancers of target genes (Paraiso et al., 2019). Interestingly, the collaborative nature of Foxh1 and T-box TF interactions in mesendoderm formation has been documented in early zebrafish embryos whereby zebrafish Foxh1 collaborates with the Vegt paralog Eomes (Nelson et al., 2014; Slagle, Aoki, & Burdine, 2011). For negative regulatory functions, Foxh1 appears to switch partners

44

Kitt D. Paraiso et al.

Fig. 2 Modes of Foxh1 co-binding in the chromatin. In combination with different partner proteins, Foxh1 co-binds to putative CRMs and activate target genes in a spatially regulated manner. (A) Foxh1 recruits Tle/Groucho and represses nodal target genes in ectoderm. (B) Foxh1, Otx1, and Vegt interact and activate endodermal genes. (C) Foxh1, Sox7, and Smad2,3 interact and activate mesodermal genes. (D) Foxh1 and Smad2,3 mediate the Nodal signaling pathway and activate mesendodermal genes.

whereby Foxh1 recruits the Groucho family of co-repressors Tle1/2/4 (Charney, Forouzmand, et al., 2017) to mediate the switch to Foxh1’s repressive roles (Reid et al., 2016). 2.4.2 Sox3 Sox3 is a member of the B group of the large Sox proteins and binds to variants of a common core consensus sequence AACAAT (Mertin, McDowall, & Harley, 1999; van Beest et al., 2000). Sox3 is expressed both maternally and zygotically in ectoderm, and is involved in neural ectoderm specification (Rogers, Archer, Cunningham, Grammer, & Casey, 2008). Sox3’s maternal role was inhibited using an affinity-purified antibody, which blocks Sox3 binding to DNA (Zhang et al., 2003). This indicated its primary function as a mesodermal suppressor in the ectoderm by negatively regulating Nodal signaling (Zhang & Klymkowsky, 2007). While maternal sox3 mRNA is predominantly expressed in the ectoderm, it is also detected in the vegetal mass of early cleavage and blastula embryos (Blitz et al., 2017; Paraiso et al., 2019), suggesting its possible roles in the mesoderm and the endoderm as well. Indeed, the role of Sox3 in Xenopus embryos has been expanded to pioneering roles in chromatin opening and mediating chromosome conformation to regulate early gene expression. Sox3 synergistically acts with Pou5f3 (an ortholog of mammalian Oct4/ Pou5f1) to regulate ZGA (Gentsch et al., 2019) similar to what has been proposed in early zebrafish embryos (Lee et al., 2013). This functional collaboration by Sox3 and Pou5f3 in Xenopus is also reminiscent of the

Early Xenopus gene regulatory programs

45

establishment of pluripotency in mammalian embryonic stem cells by these factors (Takahashi & Yamanaka, 2006). 2.4.3 Pou5f3 Xenopus and zebrafish PouV family TFs are evolutionarily closely related to mammalian Pou5f1/Oct4 (Frankenberg, Pask, & Renfree, 2010; Hellsten et al., 2010), which plays crucial roles during early mammalian embryogenesis and embryonic stem cell pluripotency. In Xenopus, the PouV genes pou5f3.1, pou5f3.2 and pou5f3.3 (previously oct91, oct25, and oct60, respectively) are expressed ubiquitously in the early embryo (Chiu et al., 2014). Temporally, pou5f3.2 and pou5f3.3 are both expressed maternally, but pou5f3.3 RNA is more abundant than pou5f3.2 before the blastula stage (Hinkley, Martin, Leibham, & Perry, 1992). Injection of a cocktail of pou5f3.1, pou5f3.2 and pou5f3.3 MOs into Xenopus embryos caused axial defects including gross head abnormalities and shortening of the trunk and tail (Chiu et al., 2014; Morrison & Brickman, 2006). Interestingly, Pou motifs are enriched in regions of Foxh1 ChIP-seq peaks, and PouV knockdown showed up-regulation of Foxh1 target genes such as (cer1, foxa4, gata4, gsc, nodal2, snai1, and vegt) (Chiu et al., 2014). These results suggest that PouV proteins negatively regulate the expression of a subset of Nodal target genes. In Xenopus, Pou5f3 was shown to interact with Sox3 and initiates local chromatin remodeling to facilitate poised or active transcription during ZGA (Gentsch et al., 2019). Sox3 and PouV interaction may be additive or redundant as loss of function of Sox3 and PouV individually is ineffective, whereas simultaneous depletion led to much stronger phenotypes. Interestingly, the maternal PouV TFs was also shown to inhibit the function of the TFs Vegt and Ctnnb1 (Cao et al., 2007) and Nodal signaling (Cao et al., 2006) in ectoderm, implying their diverse functions during early embryogenesis.

2.5 Intracellular mediators of signaling pathways 2.5.1 Gdf1/Nodal/Smad2,3 Maternally expressed gdf1 (formerly vg1) transcripts encoding for a Nodal ligand is expressed in the vegetal mass along with vegt transcripts. Gdf1 loss-of-function results in reduction in Smad2/3 phosphorylation (Birsoy et al., 2006) and results in shortened axial elongation, indicative of convergent extension defects. The activated Smad2/3-Smad4 complex regulates target genes, which include foxh1, eomes, foxh1.2, gtf2i, gtf2ird1, mixer, tcf3 (also known as e2a) and tp53. The role of Gdf1 appears to be specific to

46

Kitt D. Paraiso et al.

the mesoderm-inducing function of endodermal cells as Gdf1 loss-offunction specifically affects the organizer (dorsal mesoderm) gene expression, including BMP antagonists chrd and nog, and the zygotic Nodal gene xnr1, but not the endodermal gene sox17a (Birsoy et al., 2006). During cleavage stages, nodal5 and nodal6 are activated by Vegt, and these ligands contribute to the earliest zygotic activation of the Nodal signaling pathway. This role of Nodal signaling during endoderm formation has been well documented in a variety of vertebrates (reviewed by Schier, 2003). 2.5.2 Wnt/TCF/Ctnnb1 (β-catenin) Maternal wnt11b, is localized to the vegetal pole in the egg, relocated to the dorsal vegetal cells following cortical rotation, and activates a canonical Wnt signaling pathway to specify the dorsal fate (Ku & Melton, 1993; Tao, Yokota, et al., 2005). Molecularly, nuclear β-catenin protein accumulates dorsally, forms a complex with Lef/Tcf TFs, and directly regulates sia1 and sia2 homeobox gene expression (Brannon et al., 1997; Laurent et al., 1997). At the same time, dorsally enriched maternal Wnt/β-catenin signaling activates the nodal5 and nodal6 genes (Xanthos et al., 2002; Yang, Tan, Darken, Wilson, & Klein, 2002). Co-occurrence between Foxh1 (a major Smad2/3 co-factor) and β-catenin ChIP-seq peaks supports the model that Nodal and Wnt signaling pathways crosstalk and co-regulate the expression of dorsal mesoderm genes, including hhex, lhx1, otx2, cer1 and gsc (Chiu et al., 2014; Nakamura et al., 2016).

2.6 Other TFs Hundreds of TFs are detectably expressed at the RNA level maternally and dozens are expressed in animally or vegetally localized manner (Fig. 3). Among these TFs are those previously discussed and many more have unknown functions. Examples include snai1, pbx1 and gbpb1 which are vegetally localized. Particularly interesting TFs among the ubiquitously expressed may be members of the Zic family as Zic motifs are often enriched within ChIP-seq peaks of the above mentioned maternal TFs (Paraiso et al., 2019). Currently, the role of Zic family members in ZGA and germ layer specification is not well understood. It will be interesting to see how these maternal TFs are integrated to initiate GRNs of the three germ layers.

Early Xenopus gene regulatory programs

47

Fig. 3 Expression of transcription factors at the 8-cell stage in the animal and vegetal blastomeres from RNA sequencing. Hundreds of transcription factors are expressed, while only a little over a dozen show localized expression. Adapted from Paraiso, K. D., Blitz, I. L., Coley, M., Cheung, J., Sudou, N., Taira, M., & Cho, K. W. Y. (2019). Endodermal maternal transcription factors establish super-enhancers during zygotic genome activation. Cell Reports, 27, 2962–2977.e5.

3. Enhancers, promoters and chromatin states 3.1 Genomic approaches to identifying enhancers Much work has been performed in order to understand the function of CRMs during early Xenopus embryogenesis. The CRMs upstream of genes such as gsc (reviewed by Koide, Hayata, & Cho, 2005) and hhex (Rankin, Kormish, Kofron, Jegga, & Zorn, 2011) have been functionally dissected. However, the depth of understanding of the regulatory regions of these two genes are the exception in our understanding of the CRMs in the genome. In addition, recent genomic datasets have uncovered putative enhancers downstream the hhex gene body (Fig. 4), outside the 6 kb upstream region of hhex that has been analyzed (Rankin et al., 2011). This downstream region is bound by TFs such as Foxh1, Otx1 and Vegt; and is contained within the hhex endodermal super-enhancer (Paraiso et al., 2019). Due to these reasons, the use of ChIP-seq, ATAC-seq,

48

Kitt D. Paraiso et al.

Fig. 4 Genome browser showing the hhex loci with Foxh1, Otx1 and Vegt ChIP-seq signal, and the endodermal super-enhancer associated with this loci. Boxed are putative enhancers bound by maternal TFs that are previously untested for enhancer activity.

Fig. 5 Marks of an active enhancer. Genome-wide approaches have used chromatin marks such as H3K4me1, H3K27ac, and Ep300 binding; extragenic RNA Polymerase II binding; transcription factor binding; chromatin accessibility and enhancer RNA transcription to identify putative active enhancers.

DNAse-seq, etc., to identify the genomic coordinates of epigenetically marked histones and open chromatin have been attractive approaches in identifying putative CRMs in the genome. Promoter regions have been found to be associated with H3K4me3 marking (Heintzman et al., 2007). Enhancers on the other hand have been associated with a variety of features, in addition to binding of multiple TFs (Fig. 5). Chromatin marks such as H3K4me1 (Heintzman et al., 2007), H3K27ac (Creyghton et al., 2010), Ep300 (Heintzman et al., 2007), and DNA accessibility (Boyle et al., 2008) have been used to identify active enhancers. Additionally, extragenic binding

Early Xenopus gene regulatory programs

49

of RNA Polymerase II (De Santa et al., 2010) and transcripts (enhancer RNAs) (Kim et al., 2010) have been associated with marks of active enhancers, genome-wide. For Section 3.2, we highlight the chromatin states of the early embryo as from the view of promoter and enhancer epigenetic marks.

3.2 Chromatin states and ZGA In most animals, the early embryonic genome is transcriptionally silent and is programmed into a pluripotent state after the union of the egg and the sperm genome. This development is initially under the control of maternal products, including TFs that are loaded into the female gamete during oogenesis. These maternal TFs play central roles in coordinating the initiation of zygotic GRNs by binding to the CRMs of the genome to regulate the transcriptional responses of target genes. In addition, the presence of other transcriptional regulators such as co-activators/repressors and the chromatin state surrounding the CRMs influence gene expression. Significant efforts have been placed to uncover the chromatin state of these TF-bound CRMs during ZGA to comprehend the relationship between epigenetic regulation and gene expression during the earliest cell fate decision process. The onset of zygotic genome activation (ZGA) is one of the first major milestones in embryonic development, and the timing varies significantly among different animals (reviewed by Jukam, Shariati, & Skotheim, 2017). For example, in mice, this process begins right after the first cleavage cycle (24 h post-fertilization) while in Drosophila melanogaster, ZGA occurs at the 14th nuclear cycle (2.5h post-fertilization) at 6000 nuclei stage. In Xenopus laevis, ZGA generally occurs after the first 12 cleavage divisions during the early blastula stage, which is also known as the mid-blastula transition (MBT). However, recent high-resolution transcriptome profiling (Collart et al., 2014; Owens et al., 2016) has revealed that zygotic transcripts of pre-mir427 are detected as early as the 8-cell stage (at the third cleavage), which is significantly earlier than the classically defined MBT (Newport & Kirschner, 1982). Additionally, from the same transcriptomic data, dozens of zygotic transcripts are first detected at the 128- and 256-cell stages, including nodal5 and nodal6 (Yang et al., 2002). These data indicate that Xenopus ZGA is not a single switch-like temporal event, but instead occurs broadly during a time window where new transcription gradually begins. Evidence from Xenopus suggests that gene promoter marking in the form of histone H3 lysine 4 trimethylation (H3K4me3) are largely established

50

Kitt D. Paraiso et al.

around the period of zygotic genome activation (Akkers et al., 2009; Hontelez et al., 2015). Loss of transcription through α-amanitin treatment does not affect H3K4me3 marking, suggesting that the establishment of this mark is controlled by maternal factors (Hontelez et al., 2015). However, the mechanism of regulating the timing of the methyltransferase activity is unknown. Interestingly, the appearance of the H3K4me3 mark during early development varies across species. Zebrafish and Drosophila H3K4me3 marking occurs largely during ZGA (Li, Harrison, Villalta, Kaplan, & Eisen, 2014; Vastenhouw et al., 2010), similar to Xenopus. While in mice, unusually broad, non-canonical H3K4me3 domains (wider than 5 kb) were observed in matured oocytes (Dahl et al., 2016; Liu et al., 2016; Zhang et al., 2016). During ZGA, these broad domains disappear and H3K4me3 marking become restricted to the conventional transcription start site (TSS) of transcriptionally active genes. Early embryonic enhancers are labeled with a variety of histone marks. These include the general enhancer histone mark H3K4me1, active enhancer histone mark H3K27ac (histone H3 lysine 27 trimethylation) and the co-activator Ep300 (Gupta, Wills, Ucar, & Baker, 2014; Hontelez et al., 2015). Just like H3K4me3, these enhancer marks appear to be largely established during ZGA, consistent with findings in Drosophila (Li et al., 2014). At present, how H3K4me1 and H3K27ac accumulation on CRMs in Xenopus embryos is regulated is unknown, although numerous histone methyl- and acetyl-transferases that regulate their deposition are maternally expressed in the early Xenopus embryo (Collart et al., 2014; Owens et al., 2016). Importantly, the presence of these enhancer marks is highly correlated with maternal TF binding to CRMs (Charney, Forouzmand, et al., 2017; Gentsch et al., 2019; Paraiso et al., 2019). It is therefore tempting to speculate that maternal TFs are somehow involved in the deposition of these histone marks, perhaps by recruiting specific histone modifying complexes to CRMs. While a Wnt signaling co-activator, Ctnnb1 (β-catenin), was shown to be required for the deposition of H3R8me2 mark through recruitment of the methyltransferase Prmt2 (Blythe, Cha, Tadjuidje, Heasman, & Klein, 2010) in the promoter region of organizer genes, it is currently unclear whether the H3R8 mark is a critical regulator of ZGA.

3.3 Repressive chromatin marking The question of how a given gene is dynamically modified, sometimes with active histone marks in one cell type, but with repressive states in another cell type, is a central question to understanding the germ layer gene regulatory

Early Xenopus gene regulatory programs

51

program. The PRC2 complex deposits H3K27me3 and represses target gene expression (Bannister & Kouzarides, 2011). In most species, the increase of H3K27me3 begins to emerge after ZGA (Li et al., 2014; Vastenhouw et al., 2010). This implies that transcriptional quiescence before ZGA is not imposed by H3K27me3-marked repression. In mammalian embryonic stem cells (ESCs), the co-occurrence of active H3K4me3 and repressive H3K27me3 chromatin modifications has been described as a bivalent mode on promoters of poised developmental genes (Bernstein et al., 2006; Mikkelsen et al., 2007). However, this bivalent mode has not been detected during early embryogenesis of mouse, fly and frog embryos (Akkers et al., 2009; Liu et al., 2016; Zhang et al., 2016), suggesting that the bivalency marking is not a common state of the vertebrate embryonic genome. For instance, sequential ChIP-seq experiments carried out using Xenopus embryos reveal that the bivalent marking (H3K4me3 and H3K27me3) of genes is not a prevalent configuration in Xenopus embryos (Akkers et al., 2009). Another interesting finding is the spatially regulated activity of H3K27me3 marking. When histone marking of endodermally expressed genes was examined, the genes were specifically marked by H3K27me3 repressive mark in ectodermal (animally located) cells. This suggests that H3K27me3 participates in repressing unwanted endodermal gene expression in ectoderm, but not in endoderm, thus contributing to the spatially distinct chromatin states in different cell types. How can the CRM of a given gene be marked by an active histone mark, while in other tissues the same CRM is marked by a repressive mark? Through the examination of various maternal TF ChIP-seq data, it was noted that maternal TFs co-binding to CRMs are associated with H3K4me1 or H3K27ac marks in later developmental stages (Paraiso et al., 2019). Co-binding of maternal TFs is particularly enriched in clusters of endodermal enhancers with high-levels of H3K4me1 marking or superenhancers, which are associated with key cell identity genes (Loven et al., 2013; Whyte et al., 2013). Current tissue-specific perturbation data show that maternal TFs regulate the RNA Polymerase II occupancy and enhancer RNA transcription in these co-bound CRMs (Paraiso et al., 2019). Possibly, these maternal TFs also regulate deposition of H3K4me1, similar to what has been shown in whole embryo datasets where Sox3 and Pou5f3 perform this function (Gentsch et al., 2019). Further study examining the biochemistry of the interactions between these maternal TFs and histone modifiers should provide useful insights to the dynamic epigenetic regulation occurring during ZGA and germ layer specification.

52

Kitt D. Paraiso et al.

4. Network structure analysis of the early Xenopus GRN We previously curated the endodermal gene regulatory network during early Xenopus development from fertilization through early gastrulation (Charney, Paraiso, Blitz, & Cho, 2017; Koide et al., 2005). We used a bipartite criteria in identifying regulatory targets. First, a candidate target gene has to be perturbed by loss-of-function and/or gain-of-function experiments performed on the TFs. Second, there has to be evidence of directness in the regulatory interaction. We called an interaction “putatively direct” if the activator is capable of inducing target gene transcription even in the presence of protein synthesis inhibitors. We called an interaction “direct” if there is an identifiable cis-regulatory region, which can be implicated to the activating TF. This can be in the form of reporter gene assays, chromatin immunoprecipitation experiments, DNAse footprinting, and/or electrophoretic mobility shift assays. The resultant network describes the first few hours of Xenopus development starting from the control of maternal TFs in establishing the germ layers and ends with the role of early activated zygotic TFs during early gastrulation. Presented in Fig. 6 is a simplified GRN showing how combinatorial interactions of maternal TFs regulate spatially distinct expression patterns of zygotic targets, accounting for recently published evidences of regulatory interactions. The Xenopus GRN provides the opportunity to dissect and understand the underlying network of interactions between TFs and their target genes by examining network motifs, or smaller network structures that appear frequently within the GRN. Network motifs can be classified based on the number of involved nodes: single-, two- and three-node motifs. The singe node motifs (a single gene regulated by its own protein product) represent a positive or a negative auto-regulation. Two node network motifs (two genes, X and Y, mutually regulate each other) can involve with a positive, negative or double negative feedback loops. Lastly, the feedforward loops of network motifs are made of three nodes (two genes regulate a third downstream gene) are made of three nodes. Although the various types of network motifs present in a GRN were previously identified, the frequency of each type of network motifs present in GRN architecture have not been fully explored. In addition, how their presence is relevant to the function of individual TFs is uncertain. We analyzed the entire literature of published Xenopus mesendoderm GRN structures (Charney, Paraiso, et al., 2017) and reported on the

Fig. 6 A maternal TF-centric GRN. Shown are a subset of known maternal TFs and their target zygotic genes highlighting the area of activity in the early embryo. Neither shown are target genes with more complex expression patterns, nor regulatory connections with other types of combinations of maternal TF input. For a more comprehensive GRN, see Charney, Paraiso, et al. (2017).

54

Kitt D. Paraiso et al.

frequency of network motifs present among 23 TFs. We found 4 singlenode motifs (2 positive and 2 negative regulation), 5 two-node motifs (3 positive feedbacks, 1 negative feedbacks, 1 double negative feedbacks), and 88 three-node motifs. Of these, 63 are feed forward loop (FFL) type I, representing 70% of all three-node motifs. More specifically, in the Xenopus endodermal GRN, what appears to be common is the formation of feedforward loops whereby the product of gene A activates gene B, and both factors A and B activate the expression of gene C. This network motif appears to form whereby A is a maternal TF, B is an early expressed zygotic TF and C is a later expressed zygotic gene. In the majority of cases, the initial activator appears to be either Ctnnb1 (β-catenin), Foxh1, Smad2/3, or Vegt (gene A). These maternal factors activate the expression of early and midblastula zygotic genes such as wnt8a, sia1, sia2, mix1, gsc, and the nodal genes (gene B), which, in turn, activate the expression a larger number of later expressed mesendodermal genes (gene C). A notable feature of coherent feedforward loops is that they may be useful in refining temporally the regulation of a cascade of gene expression (Mangan, Zaslaver, & Alon, 2003). This network structure could have implications in the differences in timing of gene induction from the embryonic genome, as seen in the broad window by which the timing of ZGA occurs. Interestingly, the relative abundance of this motif has been noted in the GRNs of E. coli and S. cerevisiae (Mangan et al., 2003), as well as the developmental GRN of sea urchin (Peter & Davidson, 2017). Further meta-analysis of other established developmental GRNs in vertebrates suggests that the abundance of this feedforward loop is a staple of GRNs, as seen in the network structure of the C. elegans endoderm (Maduro, 2017), the mammalian T-cell (Kueh & Rothenberg, 2012), the mammalian pancreatic (Servitja & Ferrer, 2004) and the vertebrate neural crest (Simo˜es-Costa & Bronner, 2015) GRNs. How this network motif and other motifs are functionally relevant is yet to be tested in the Xenopus system.

5. Summary and prospects The current Xenopus GRN is based on hand curated data accumulated over decades of work (Charney, Paraiso, et al., 2017). In the future, with the accumulation of more genomic data such as RNA-seq, ChIP-seq and ATAC-seq, we expect this GRNs to be built through integration of these datasets. A major challenge of such an approach is to determine which predicted interactions between TF and CRMs are functional as hundreds

Early Xenopus gene regulatory programs

55

and thousands of such sites are predicted by bioinformatic approaches. Thus, what is needed in GRN science in the future is high-throughput approaches to validate the predicted TF-CRM interactions in vivo. Advances in the use of CRISPR/Cas9 mediated deletions (Blitz et al., 2013, 2016; Nakayama et al., 2013; Wang et al., 2015) and knock-ins (Aslan et al., 2017; Shi et al., 2015), along with high-throughput reporter genes assays such as STARR-seq (self-transcribing active regulatory region sequencing) (Arnold et al., 2013) are likely to provide the opportunity to fill in this gap. We therefore expect the Xenopus and other animal model systems to facilitate significant advances in gene regulatory biology.

Acknowledgments Work in Ken Cho’s lab is funded by the National Institute of Health R01GM126395, and the National Science Foundation Division of Organismal Systems #1755214. We would like to thank Dr. Ira L. Blitz for useful discussions and comments on the manuscript.

References Akkers, R. C., van Heeringen, S. J., Jacobi, U. G., Janssen-Megens, E. M., Franc¸oijs, K. J., Stunnenberg, H. G., et al. (2009). A hierarchy of H3K4me3 and H3K27me3 acquisition in spatial gene regulation in Xenopus embryos. Developmental Cell, 17, 425–434. Arnold, C. D., Gerlach, D., Stelzer, C., Bory n, Ł. M., Rath, M., & Stark, A. (2013). Genome-wide quantitative enhancer activity maps identified by STARR-seq. Science, 339, 1074–1077. Aslan, Y., Tadjuidje, E., Zorn, A. M., & Cha, S. W. (2017). High-efficiency non-mosaic CRISPR-mediated knock-in and indel mutation in F0 Xenopus. Development, 144, 2852–2858. Bannister, A. J., & Kouzarides, T. (2011). Regulation of chromatin by histone modifications. Cell Research, 21, 381–395. Bernstein, B. E., Mikkelsen, T. S., Xie, X., Kamal, M., Huebert, D. J., Cuff, J., et al. (2006). A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell, 125, 315–326. Birsoy, B., Kofron, M., Schaible, K., Wylie, C., & Heasman, J. (2006). Vg1 is an essential signaling molecule in Xenopus development. Development, 133, 15–20. Bjornson, C. R., Griffin, K. J., Farr, G. H., 3rd, Terashima, A., Himeda, C., Kikuchi, Y., et al. (2005). Eomesodermin is a localized maternal determinant required for endoderm induction in zebrafish. Developmental Cell, 9, 523–533. Blitz, I. L., Biesinger, J., Xie, X., & Cho, K. W. Y. (2013). Biallelic genome modification in F (0) Xenopus tropicalis embryos using the CRISPR/Cas system. Genesis, 51, 827–834. Blitz, I. L., Fish, M. B., & Cho, K. W. Y. (2016). Leapfrogging: Primordial germ cell transplantation permits recovery of CRISPR/Cas9-induced mutations in essential genes. Development, 143, 2868–2875. Blitz, I. L., Paraiso, K. D., Patrushev, I., Chiu, W. T. Y., Cho, K. W. Y., & Gilchrist, M. J. (2017). A catalog of Xenopus tropicalis transcription factors and their regional expression in the early gastrula stage embryo. Developmental Biology, 426, 409–417. Blythe, S. A., Cha, S. W., Tadjuidje, E., Heasman, J., & Klein, P. S. (2010). β-Catenin primes organizer gene expression by recruiting a histone H3 arginine 8 methyltransferase, Prmt2. Developmental Cell, 19, 220–231.

56

Kitt D. Paraiso et al.

Boyle, A. P., Davis, S., Shulha, H. P., Meltzer, P., Margulies, E. H., Weng, Z., et al. (2008). High-resolution mapping and characterization of open chromatin across the genome. Cell, 132, 311–322. Brannon, M., Gomperts, M., Sumoy, L., Moon, R. T., & Kimelman, D. (1997). A β-catenin/XTcf-3 complex binds to the siamois promoter to regulate dorsal axis specification in Xenopus. Genes & Development, 11, 2359–2370. Cao, Y., Siegel, D., Donow, C., Kn€ ochel, S., Yuan, L., & Kn€ ochel, W. (2007). POU-V factors antagonize maternal VegT activity and beta-Catenin signaling in Xenopus embryos. EMBO Journal, J26, 2942–2954. Cao, Y., Siegel, D., & Kn€ ochel, W. (2006). Xenopus POU factors of subclass V inhibit activin/nodal signaling during gastrulation. Mechanisms of Development, 123, 614–625. Cha, S. W., McAdams, M., Kormish, J., Wylie, C., & Kofron, M. (2012). Foxi2 is an animally localized maternal mRNA in Xenopus, and an activator of the zygotic ectoderm activator Foxi1e. PLoS One, 7, e41782. Charney, R. M., Forouzmand, E., Cho, J. S., Cheung, J., Paraiso, K. D., Yasuoka, Y., et al. (2017). Foxh1 occupies cis-regulatory modules prior to dynamic transcription factor interactions controlling the mesendoderm gene program. Developmental Cell, 40, 595–607.e4. Charney, R. M., Paraiso, K. D., Blitz, I. L., & Cho, K. W. Y. (2017). A gene regulatory program controlling early Xenopus mesendoderm formation: Network conservation and motifs. Seminars in Cell & Developmental Biology, 66, 12–24. Chen, X., Rubock, M. J., & Whitman, M. (1996). A transcriptional partner for MAD proteins in TGF-beta signalling. Nature, 383, 691–696. Chen, X., Weisberg, E., Fridmacher, V., Watanabe, M., Naco, G., & Whitman, M. (1997). Smad4 and FAST-1 in the assembly of activin-responsive factor. Nature, 389, 85–89. Chiu, W. T., Charney, L. R., Blitz, I. L., Fish, M. B., Li, Y., Biesinger, J., et al. (2014). Genome-wide view of TGFβ/Foxh1 regulation of the early mesendoderm program. Development, 141, 4537–4547. Collart, C., Owens, N. D., Bhaw-Rosun, L., Cooper, B., De Domenico, E., Patrushev, I., et al. (2014). High-resolution analysis of gene activity during the Xenopus mid-blastula transition. Development, 141, 1927–1939. Creyghton, M. P., Cheng, A. W., Welstead, G. G., Kooistra, T., Carey, B. W., Steine, E. J., et al. (2010). Histone H3K27ac separates active from poised enhancers and predicts developmental state. Proceedings of the National Academy of Sciences of the United States of America, 107, 21931–21936. Cuykendall, T. N., & Houston, D. W. (2010). Identification of germ plasm-associated transcripts by microarray analysis of Xenopus vegetal cortex RNA. Developmental Dynamics, 239, 1838–1848. Dahl, J. A., Jung, I., Aanes, H., Greggains, G. D., Manaf, A., Lerdrup, M., et al. (2016). Broad histone H3K4me3 domains in mouse oocytes modulate maternal-to-zygotic transition. Nature, 537, 548–552. De Domenico, E., Owens, N. D., Grant, I. M., Gomes-Faria, R., & Gilchrist, M. J. (2015). Molecular asymmetry in the 8-cell stage Xenopus tropicalis embryo described by single blastomere transcript sequencing. Developmental Biology, 408, 252–268. De Santa, F., Barozzi, I., Mietton, F., Ghisletti, S., Polletti, S., Tusi, B. K., et al. (2010). A large fraction of extragenic RNA pol II transcription sites overlap enhancers. PLoS Biology, 8 e1000384. Flachsova, M., Sindelka, R., & Kubista, M. (2013). Single blastomere expression profiling of Xenopus laevis embryos of 8 to 32-cells reveals developmental asymmetry. Scientific Reports, 3, 2278. Frankenberg, S., Pask, A., & Renfree, M. B. (2010). The evolution of class V POU domain transcription factors in vertebrates and their characterisation in a marsupial. Developmental Biology, 337, 162–170.

Early Xenopus gene regulatory programs

57

Gentsch, G. E., Owens, N. D., Martin, S. R., Piccinelli, P., Faial, T., Trotter, M. W., et al. (2013). In vivo T-box transcription factor profiling reveals joint regulation of embryonic neuromesodermal bipotency. Cell Reports, 4, 1185–1196. Gentsch, G. E., Spruce, T., Owens, N. D. L., & Smith, J. C. (2019). Maternal pluripotency factors initiate extensive chromatin remodelling to predefine first response to inductive signals. Nature Communications, 10, 4269. Gupta, R., Wills, A., Ucar, D., & Baker, J. (2014). Developmental enhancers are marked independently of zygotic Nodal signals in Xenopus. Developmental Biology, 395, 38–49. Heasman, J., Wylie, C. C., Hausen, P., & Smith, J. C. (1984). Fates and states of determination of single vegetal pole blastomeres of X. laevis. Cell, 37, 185–194. Heintzman, N. D., Stuart, R. K., Hon, G., Fu, Y., Ching, C. W., Hawkins, R. D., et al. (2007). Distinct and predictive chromatin signatures of transcriptional promoters and enhancers in the human genome. Nature Genetics, 39, 311–318. Hellsten, U., Harland, R. M., Gilchrist, M. J., Hendrix, D., Jurka, J., Kapitonov, V., et al. (2010). The genome of the Western clawed frog Xenopus tropicalis. Science, 328, 633–636. Hinkley, C. S., Martin, J. F., Leibham, D., & Perry, M. (1992). Sequential expression of multiple POU proteins during amphibian early development. Molecular and Cellular Biology, 12, 638–649. Hinman, V. F., Nguyen, A. T., & Davidson, E. H. (2003). Expression and function of a starfish Otx ortholog, AmOtx: A conserved role for Otx proteins in endoderm development that predates divergence of the eleutherozoa. Mechanisms of Development, 120, 1165–1176. Hontelez, S., van Kruijsbergen, I., Georgiou, G., van Heeringen, S. J., Bogdanovic, O., Lister, R., et al. (2015). Embryonic transcription is controlled by maternally defined chromatin state. Nature Communications, 6, 10148. Horb, M. E., & Thomsen, G. H. (1997). A vegetally localized T-box transcription factor in Xenopus eggs specifies mesoderm and endoderm and is essential for embryonic mesoderm formation. Development, 124, 1689–1698. Jukam, D., Shariati, S. A. M., & Skotheim, J. M. (2017). Zygotic genome activation in vertebrates. Developmental Cell, 42, 316–332. Kim, T. K., Hemberg, M., Gray, J. M., Costa, A. M., Bear, D. M., Wu, J., et al. (2010). Widespread transcription at neuronal activity-regulated enhancers. Nature, 465, 182–187. Kofron, M., Demel, T., Xanthos, J., Lohr, J., Sun, B., Sive, H., et al. (1999). Mesoderm induction in Xenopus is a zygotic event regulated by maternal VegT via TGFbeta growth factors. Development, 126, 5759–5770. Kofron, M., Puck, H., Standley, H., Wylie, C., Old, R., Whitman, M., et al. (2004). New roles for FoxH1 in patterning the early embryo. Development, 131, 5065–5078. Koide, T., Hayata, T., & Cho, K. W. Y. (2005). Xenopus as a model system to study transcriptional regulatory networks. Proceedings of the National Academy of Sciences of the United States of America, 102, 4943–4948. Ku, M., & Melton, D. A. (1993). Xwnt-11: A maternally expressed Xenopus wnt gene. Development, 119, 1161–1173. Kueh, H. Y., & Rothenberg, E. V. (2012). Regulatory gene network circuits underlying T cell development from multipotent progenitors. Wiley Interdisciplinary Reviews Systems Biology and Medicine, 4, 79–102. Laurent, M. N., Blitz, I. L., Hashimoto, C., Rothb€acher, U., & Cho, K. W. Y. (1997). The Xenopus homeobox gene twin mediates Wnt induction of goosecoid in establishment of Spemann’s organizer. Development, 124, 4905–4916. Lee, M. T., Bonneau, A. R., Takacs, C. M., Bazzini, A. A., DiVito, K. R., Fleming, E. S., et al. (2013). Nanog, Pou5f1 and SoxB1 activate zygotic gene expression during the maternal-to-zygotic transition. Nature, 503, 360–364.

58

Kitt D. Paraiso et al.

Li, X. Y., Harrison, M. M., Villalta, J. E., Kaplan, T., & Eisen, M. B. (2014). Establishment of regions of genomic activity during the Drosophila maternal to zygotic transition. eLife, 3, e03737. Liu, X., Wang, C., Liu, W., Li, J., Li, C., Kou, X., et al. (2016). Distinct features of H3K4me3 and H3K27me3 chromatin domains in pre-implantation embryos. Nature, 537, 558–562. Loven, J., Hoke, H. A., Lin, C. Y., Lau, A., Orlando, D. A., Vakoc, C. R., et al. (2013). Selective inhibition of tumor oncogenes by disruption of super-enhancers. Cell, 153, 320–334. Lustig, K. D., Kroll, K. L., Sun, E. E., & Kirschner, M. W. (1996). Expression cloning of a Xenopus T-related gene (Xombi) involved in mesodermal patterning and blastopore lip formation. Development, 122, 4001–4012. Maduro, M. F. (2017). Gut development in C. elegans. Seminars in Cell & Developmental Biology, 66, 3–11. Mangan, S., Zaslaver, A., & Alon, U. (2003). The coherent feedforward loop serves as a signsensitive delay element in transcription networks. Journal of Molecular Biology, 334, 197–204. Mertin, S., McDowall, S. G., & Harley, V. R. (1999). The DNA-binding specificity of SOX9 and other SOX proteins. Nucleic Acids Research, 27, 1359–1364. Mikkelsen, T. S., Ku, M., Jaffe, D. B., Issac, B., Lieberman, E., Giannoukos, G., et al. (2007). Genome-wide maps of chromatin state in pluripotent and lineage-committed cells. Nature, 448, 553–560. Mir, A., Kofron, M., Zorn, A. M., Bajzer, M., Haque, M., Heasman, J., et al. (2007). FoxI1e activates ectoderm formation and controls cell position in the Xenopus blastula. Development, 134, 779–788. Mitros, T., Lyons, J. B., Session, A. M., Jenkins, J., Shu, S., Kwon, T., et al. (2019). A chromosome-scale genome assembly and dense genetic map for Xenopus tropicalis. Developmental Biology, 452, 8–20. Morrison, G. M., & Brickman, J. M. (2006). Conserved roles for Oct4 homologues in maintaining multipotency during early vertebrate development. Development, 133, 2011–2022. Nakamura, Y., de Paiva Alves, E., Veenstra, G. J., & Hoppler, S. (2016). Tissue- and stagespecific Wnt target gene expression is controlled subsequent to β-catenin recruitment to cis-regulatory modules. Development, 143, 1914–1925. Nakayama, T., Fish, M. B., Fisher, M., Oomen-Hajagos, J., Thomsen, G. H., & Grainger, R. M. (2013). Simple and efficient CRISPR/Cas9-mediated targeted mutagenesis in Xenopus tropicalis. Genesis, 51, 835–843. Nath, K., & Elinson, R. P. (2007). RNA of AmVegT, the axolotl orthologue of the Xenopus meso-endodermal determinant, is not localized in the oocyte. Gene Expression Patterns, 7, 197–201. Nelson, A. C., Cutty, S. J., Niini, M., Stemple, D. L., Flicek, P., Houart, C., et al. (2014). Global identification of Smad2 and Eomesodermin targets in zebrafish identifies a conserved transcriptional network in mesendoderm and a novel role for Eomesodermin in repression of ectodermal gene expression. BMC Biology, 12, 81. Newport, J., & Kirschner, M. (1982). A major developmental transition in early Xenopus embryos: I. characterization and timing of cellular changes at the midblastula stage. Cell, 30(3), 675–686. Owens, N. D. L., Blitz, I. L., Lane, M. A., Patrushev, I., Overton, J. D., Gilchrist, M. J., et al. (2016). Measuring absolute RNA copy numbers at high temporal resolution reveals transcriptome kinetics in development. Cell Reports, 14, 632–647. Pannese, M., Cagliani, R., Pardini, C. L., & Boncinelli, E. (2000). Xotx1 maternal transcripts are vegetally localized in Xenopus laevis oocytes. Mechanisms of Development, 90, 111–114.

Early Xenopus gene regulatory programs

59

Paraiso, K. D., Blitz, I. L., Coley, M., Cheung, J., Sudou, N., Taira, M., et al. (2019). Endodermal maternal transcription factors establish super-enhancers during zygotic genome activation. Cell Reports, 27, 2962–2977.e5. Perez, O., Benı´tez, M. S., Nath, K., Heasman, J., Del, P. E. M., & Elinson, R. P. (2007). Comparative analysis of Xenopus VegT, the meso-endodermal determinant, identifies an unusual conserved sequence. Differentiation, 75, 559–565. Peter, I. S., & Davidson, E. H. (2010). The endoderm gene regulatory network in sea urchin embryos up to mid-blastula stage. Developmental Biology, 340, 188–199. Peter, I. S., & Davidson, E. H. (2017). Assessing regulatory information in developmental gene regulatory networks. Proceedings of the National Academy of Sciences of the United States of America, 114, 5862–5869. Rankin, S. A., Kormish, J., Kofron, M., Jegga, A., & Zorn, A. M. (2011). A gene regulatory network controlling hhex transcription in the anterior endoderm of the organizer. Developmental Biology, 351, 297–310. Reid, C. D., Steiner, A. B., Yaklichkin, S., Lu, Q., Wang, S., Hennessy, M., et al. (2016). FoxH1 mediates a Grg4 and Smad2 dependent transcriptional switch in Nodal signaling during Xenopus mesoderm development. Developmental Biology, 414, 34–44. Rogers, C. D., Archer, T. C., Cunningham, D. D., Grammer, T. C., & Casey, E. M. (2008). Sox3 expression is maintained by FGF signaling and restricted to the neural plate by Vent proteins in the Xenopus embryo. Developmental Biology, 313, 307–319. Schier, A. F. (2003). Nodal signaling in vertebrate development. Annual Review of Cell and Developmental Biology, 19, 589–621. Servitja, J. M., & Ferrer, J. (2004). Transcriptional networks controlling pancreatic development and beta cell function. Diabetologia, 47, 597–613. Session, A. M., Uno, Y., Kwon, T., Chapman, J. A., Toyoda, A., Takahashi, S., et al. (2016). Genome evolution in the allotetraploid frog Xenopus laevis. Nature, 538, 336–343. Shi, Z., Wang, F., Cui, Y., Liu, Z., Guo, X., Zhang, Y., et al. (2015). Heritable CRISPR/ Cas9-mediated targeted integration in Xenopus tropicalis. The FASEB Journal, 29, 4914–4923. Simo˜es-Costa, M., & Bronner, M. E. (2015). Establishing neural crest identity: A gene regulatory recipe. Development, 142, 242–257. Slagle, C. E., Aoki, T., & Burdine, R. D. (2011). Nodal-dependent mesendoderm specification requires the combinatorial activities of FoxH1 and Eomesodermin. PLoS Genetics, 7, e1002072. Snape, A., Wylie, C. C., Smith, J. C., & Heasman, J. (1987). Changes in states of commitment of single animal pole blastomeres of Xenopus laevis. Developmental Biology, 119, 503–510. Stennard, F., Carnac, G., & Gurdon, J. B. (1996). The Xenopus T-box gene, Antipodean, encodes a vegetally localised maternal mRNA and can trigger mesoderm formation. Development, 122, 4179–4188. Takahashi, K., & Yamanaka, S. (2006). Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell, 126, 663–676. Tao, J., Kuliyev, E., Wang, X., Li, X., Wilanowski, T., Jane, S. M., et al. (2005). BMP4dependent expression of Xenopus Grainyhead-like 1 is essential for epidermal differentiation. Development, 132, 1021–1034. Tao, Q., Yokota, C., Puck, H., Kofron, M., Birsoy, B., Yan, D., et al. (2005). Maternal wnt11 activates the canonical wnt signaling pathway required for axis formation in Xenopus embryos. Cell, 120, 857–871. Taverner, N. V., Kofron, M., Shin, Y., Kabitschke, C., Gilchrist, M. J., Wylie, C., et al. (2005). Microarray-based identification of VegT targets in Xenopus. Mechanisms of Development, 122, 333–354.

60

Kitt D. Paraiso et al.

Teo, A. K., Arnold, S. J., Trotter, M. W., Brown, S., Ang, L. T., Chng, Z., et al. (2011). Pluripotency factors regulate definitive endoderm specification through eomesodermin. Genes & Development, 25, 238–250. van Beest, M., Dooijes, D., van De Wetering, M., Kjaerulff, S., Bonvin, A., Nielsen, O., et al. (2000). Sequence-specific high mobility group box factors recognize 10–12-base pair minor groove motifs. The Journal of Biological Chemistry, 275, 27266–27273. Vastenhouw, N. L., Zhang, Y., Woods, I. G., Imam, F., Regev, A., Liu, X. S., et al. (2010). Chromatin signature of embryonic pluripotency is established during genome activation. Nature, 464, 922–926. Yang, J., Tan, C., Darken, R. S., Wilson, P. A., & Klein, P. S. (2002). Beta-catenin/Tcfregulated transcription prior to the midblastula transition. Development, 129, 5743–5752. Wada, S., & Saiga, H. (1999). Vegetal cell fate specification and anterior neuroectoderm formation by Hroth, the ascidian homologue of orthodenticle/otx. Mechanisms of Development, 82, 67–77. Wang, F., Shi, Z., Cui, Y., Guo, X., Shi, Y. B., & Chen, Y. (2015). Targeted gene disruption in Xenopus laevis using CRISPR/Cas9. Cell & Bioscience, 5, 15. Whyte, W. A., Orlando, D. A., Hnisz, D., Abraham, B. J., Lin, C. Y., Kagey, M. H., et al. (2013). Master transcription factors and mediator establish super-enhancers at key cell identity genes. Cell, 153, 307–319. Xanthos, J. B., Kofron, M., Tao, Q., Schaible, K., Wylie, C., & Heasman, J. (2002). The roles of three signaling pathways in the formation and function of the Spemann Organizer. Development, 129, 4027–4043. Xu, P., Zhu, G., Wang, Y., Sun, J., Liu, X., Chen, Y. G., et al. (2014). Maternal Eomesodermin regulates zygotic nodal gene expression for mesendoderm induction in zebrafish embryos. Journal of Molecular Cell Biology, 6, 272–285. Zhang, C., Basta, T., Fawcett, S. R., & Klymkowsky, M. W. (2005). SOX7 is an immediateearly target of VegT and regulates Nodal-related gene expression in Xenopus. Developmental Biology, 278, 526–541. Zhang, C., Basta, T., Jensen, E. D., & Klymkowsky, M. W. (2003). The β-catenin/VegTregulated early zygotic gene Xnr5 is a direct target of SOX3 regulation. Development, 130, 5609–5624. Zhang, J., Houston, D. W., King, M. L., Payne, C., Wylie, C., & Heasman, J. (1998). The role of maternal VegT in establishing the primary germ layers in Xenopus embryos. Cell, 94, 515–524. Zhang, J., & King, M. L. (1996). Xenopus VegT RNA is localized to the vegetal cortex during oogenesis and encodes a novel T-box transcription factor involved in mesodermal patterning. Development, 122, 4119–4129. Zhang, C., & Klymkowsky, M. W. (2007). The Sox axis, Nodal signaling, and germ layer specification. Differentiation, 75, 536–545. Zhang, B., Zheng, H., Huang, B., Li, W., Xiang, Y., Peng, X., et al. (2016). Allelic reprogramming of the histone modification H3K4me3 in early mammalian development. Nature, 537, 553–557.

CHAPTER THREE

Dynamic and self-regulatory interactions among gene regulatory networks control vertebrate limb bud morphogenesis e Zuniga∗, Rolf Zeller∗ Aime Developmental Genetics, Department Biomedicine, University of Basel, Basel, Switzerland *Corresponding authors: e-mail address: [email protected]; [email protected]

Contents 1. Introduction 2. Setting the stage: Molecular control of the limb field positions along the primary body axis 3. “Pre-patterning”: Polarizing the limb bud axes upstream of morphogenetic SHH signaling 4. Establishment of AER-FGF signaling is linked dorso-ventral and proximo-distal axis patterning 5. The self-regulatory limb bud signaling system coordinately controls anteroposterior and proximo-distal axis patterning and outgrowth 6. Propagation and termination of the self-regulatory limb bud signaling system 7. A Turing-type mechanism controls the definitive periodic pattern of digit rays 8. Brief outlook Acknowledgments References

62 64 68 71 74 77 78 81 82 82

Abstract Vertebrate limb bud outgrowth and patterning is controlled by two instructive signaling centers, the apical ectodermal ridge (AER) and the polarizing region in the posterior limb bud mesenchyme. Molecular analysis of limb bud development has identified a self-regulatory signaling system that operates between the AER and mesenchyme and orchestrates the dynamic progression of limb bud outgrowth and patterning. The first focus of this review are the gene regulatory networks (GRNs) and interactions that control the positioning of the fore- and hindlimb fields along the primary body axis, establish the initial axis polarity and control the precise positioning of the signaling centers. These early processes are largely controlled by activating and inhibiting interactions among types of transcriptional regulators expressed in specific territories. Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.02.005

#

2020 Elsevier Inc. All rights reserved.

61

62

Aimee Zuniga and Rolf Zeller

The second focus deals with the dynamic interactions among the GRNs that control limb bud patterning and outgrowth by responding to inputs from the self-regulatory limb bud signaling system. The final part describes the GRN interactions regulating digit morphogenesis and the Turing-type system that controls the periodicity of the digit ray pattern. This review highlights the significant progress made toward an integrative analysis and understanding of the morpho-regulatory systems that orchestrate patterning and outgrowth of vertebrate limb buds in time and space.

1. Introduction The analysis of limb bud morphogenesis is an experimental paradigm to study the molecular networks and cellular interactions that govern vertebrate organogenesis. In tetrapods (vertebrates with four limbs), limb bud development is initiated by specification of the fore- and hindlimb fields at specific positions along the primary embryonic axis. The pioneering studies by Saunders et al. showed that limb bud outgrowth and patterning is controlled by two instructive signaling centers (Fig. 1A; in depth review by Tickle, 2017): during the onset of limb bud development, the morphologically distinct apical ectoderm ridge (AER) forms at the dorso-ventral (DV) interface of the limb bud ectoderm and its microsurgical removal from chicken wing buds showed that the AER is required for proximo-distal (PD) outgrowth and patterning (Saunders, 1948). The second signaling center is the polarizing region (or ZPA, Fig. 1A) that is located in the posterior mesenchyme. The polarizing region was identified by its ability to induce mirror image duplications of digits following its ectopic transplantation into the anterior chicken wing bud (Maccabe, Gasseling, & Saunders, 1973). These studies revealed its instructive role in antero-posterior (AP) patterning and prompted Wolpert to propose that mesenchymal cells acquire their positional information by responding to thresholds of a diffusible morphogen signal produced by the rather small number of polarizing region cells (Tickle, 1981; Wolpert, 1969). The AP limb axis is defined as the axis running from digit 1 (thumb) to digit 5 (little finger), PD from the scapula to the autopod and DV from the back of the hand to the palm (Fig. 1B). Molecular analysis identified fibroblast growth factors (FGFs) as the instructive signals produced by the AER (Mariani, Ahn, & Martin, 2008; Niswander, Tickle, Vogel, Booth, & Martin, 1993) and Sonic hedgehog (SHH) as the polarizing region signal required for AP limb axis development (Chiang et al., 2001; Riddle, Johnson, Laufer, & Tabin, 1993). It was shown AER-FGFs and SHH function as part of a positive epithelial-mesenchymal feedback loop

Fig. 1 See legend on next page.

64

Aimee Zuniga and Rolf Zeller

that controls patterning and proliferative expansion of the responding limb bud mesenchymal progenitors (LMPs; Fig. 1C; Laufer, Nelson, Johnson, Morgan, & Tabin, 1994; Niswander, Jeffrey, Martin, & Tickle, 1994). SHH signaling propagates AER-FGF expression via transcriptional upregulation of Gremlin1 (GREM1)-mediated antagonism of BMPs in the posterior-distal limb bud mesenchyme (Zuniga, Haramis, McMahon, & Zeller, 1999). These interactions result in establishment of a system of interlinked signaling feedback loops with self-regulatory propagation and termination properties (Fig. 1C; Benazet et al., 2009; Scherz, Harfe, McMahon, & Tabin, 2004; Verheyden & Sun, 2008; Zuniga et al., 1999). This review focuses on the gene regulatory networks (GRNs) and signaling systems that control the dynamic progression of limb bud development from limb field positioning to establishing the definitive periodic digit pattern during handplate (autopod) formation. An important aspect is the understanding of the activating and repressive interactions that govern the temporal and spatial dynamics of gene expressions and interactions. Throughout the review, progress made toward systems-level understanding of limb bud organogenesis will be highlighted.

2. Setting the stage: Molecular control of the limb field positions along the primary body axis The first distinct molecular markers delineating the fore- and hindlimb fields in the lateral plate mesenchyme (LPM) are the T-box transcription Fig. 1 Tetrapod limb bud development. (A) Scanning EM picture of a mouse embryo with enlarged forelimb bud at embryonic day E11.5. The two signaling centers, apical ectodermal ridge (AER) at the distal tip of the limb bud and polarizing region (ZPA) in the posterior limb bud mesenchyme are schematically indicated. In the dorsal view of the forelimb bud the antero-posterior (Ant-Post) and proximo-distal (Prox-Dist) limb bud axes are indicated. (B) Mouse limb skeleton at E16.5. Alizarin red (dark) and alcian blue (gray) reveal ossified bone and cartilage respectively. All major skeletal elements are labeled together with the stylopod, zeugopod and autopod. The anterior-most digit 1 correspond to the thumb, and the most-posterior digit 5 to the little finger. The three limb axes are indicated (Ant, anterior; Post, posterior; Prox, proximal; Dist, distal; Dors, dorsal; Vent, ventral). (C) Schemes show the regulatory interactions among the different signaling pathways that constitute the self-regulatory feedback signaling system that controls limb bud outgrowth and patterning. The system progresses in a self-regulatory manner from initiation to progression and self-termination (see text). Positive inputs on gene expression (activation, upregulation) are indicated by the following symbol: ⊸. Repressive inputs (downregulation, termination) are indicated by the following symbol: a. Terminating interactions are indicated in light gray.

Self-regulatory control of limb development

65

factors Tbx5 and Tbx4, which are activated during limb field specification and required for normal forelimb and hindlimb development respectively (for a detailed review of Tbx gene functions see Sheeba & Logan, 2017). In addition, the Pitx1 paired homeodomain transcription factor is required to up-regulate Tbx4 expression and interacts with TBX4 to promote hindlimb-specific characteristics (Duboc & Logan, 2011). The exact positions of the fore- and hindlimb fields within the LPM are specified during posterior growth and elongation of the primary body axis and in particular by the sequential 30 –50 activation of the Hox genes encoded by the four vertebrate gene clusters (Fig. 2A; Deschamps & Duboule, 2017). The proposal that the axial Hox expression code provides positional information for the fore- and hindlimb fields is corroborated to some extent by loss-of-function genetics in the mouse as forelimb buds are shifted more anteriorly in mouse

Fig. 2 Limb field positioning and pre-patterning. (A) Schemes showing the key players of the GRNs involved in positioning and establishing the forelimb (upper scheme) and hindlimb fields (lower scheme). Established genetic interactions are shown in italics, while the function of Raldh2 in retinoic acid (RA) synthesis is indicated in normal writing. (B) Limb bud pre-patterning is controlled by an anterior (upper scheme) and posterior GRN (lower scheme) that are globally antagonistic. Note that Isl1 activates Hand2 in posterior hindlimb buds instead of posterior Hox genes (in forelimb buds). Capitals indicate direct interactions of transcription factors (GLI3R, IRX3/5, HAND2, TBX3, ETS) with their target genes. Positive inputs on gene expression (activation, upregulation) are indicated by the following symbol: ⊸. Repressive inputs (downregulation, termination) are indicated by the following symbol: a.

66

Aimee Zuniga and Rolf Zeller

embryos lacking the Hoxb5 gene (Rancourt, Tsuzuki, & Capecchi, 1995). In addition, hindlimbs are displaced posteriorly in mouse embryos lacking the Hoxc8 gene (van den Akker et al., 2001). These studies provided the first genetic and functional evidence that axial Hox genes are essential for correct limb field positioning, but extensive further genetic analysis failed to provide additional evidence. Instead, it revealed that axial Hox9 activity is required to regulate the expression of genes in the posterior part of the forelimb field (Xu & Wellik, 2011) and uncovered the essential functions of the 50 HoxA and 50 HoxD genes (distal HoxA/D genes) during limb bud outgrowth and patterning (reviewed by Kherdjemil & Kmita, 2018; Zakany & Duboule, 2007). Only recently Moreau et al. (2019) have identified molecular interactions among Hox genes that impact positioning of the forelimb field in chicken embryos. In particular, they linked these findings to natural variation in forelimb positioning in different bird species as a consequence of changes affecting the timing of Hox gene activation during gastrulation. Their study establishes that fore-, inter- and hindlimb progenitor cells are generated sequentially and express specific combinations of Hox genes. In particular, the collinear expression of the Hox4/5 and Hox9 transcription factors in the LPM regulates the position and size of the Tbx5 expression domain, which defines the forelimb field position (Fig. 2A, upper panel; Minguillon et al., 2012; Moreau et al., 2019; Nishimoto, Minguillon, Wood, & Logan, 2014). Experimentally-induced posterior expansion/ displacement of either the Hoxb4 or Hoxb9 domains results in posterior expansion of Tbx5 expression and a corresponding posterior shift of the forelimb field (Moreau et al., 2019). The authors detect strikingly similar molecular changes in ostrich embryos whose forelimb buds develop at more posterior somite positions in comparison to chicken embryos. Conversely, the Hox expression domains are shifted anteriorly in zebra finch embryos in agreement with the anterior shift of their forelimb buds in comparison to chicken embryos (Moreau et al., 2019). In mouse embryos lacking the Gdf11 signaling molecule the hindlimb bud is displaced posteriorly, revealing the importance of GDF11 signaling for hindlimb field positioning (Fig. 2A, lower panel). This posterior shift correlates well with the posterior expansion of the corresponding Hox domains in the mutant LPM (McPherron, Lawler, & Lee, 1999). An experimental delay in the onset of GDF11 signaling in early chicken embryos also delays activation of the Hox9 to Hox13 paralogues in the LMP, which not only shifts their expression domains but also the hindlimb field posteriorly (Matsubara et al., 2017). In contrast, precocious

Self-regulatory control of limb development

67

activation of GDF11 signaling expands the Hox domains anteriorly and shifts the hindlimb field anteriorly. Furthermore, the heterochronic differences in GDF activation observed in different species correlate well with the observed natural variation in hindlimb positioning. For example, delayed and thereby more posterior GDF11 activation is observed in tetrapod species with hindlimb buds located at comparatively more posterior somite positions than mouse embryos (Matsubara et al., 2017). Collectively these studies indicate that temporal heterochrony in the collinear activation of Hox genes (Deschamps & Duboule, 2017) and Gdf11 in the LMP and posterior axial mesenchyme respectively, underlies the species-specific variations in fore- and hindlimb bud positions (Fig. 2A). Almost four decades ago it was shown that retinoic acid (RA) is able to induce an ectopic polarizing region in chicken wing buds and RA was proposed to be the long-thought morphogen that specifies AP digit identities (Tickle, Alberts, Wolpert, & Lee, 1982). However, subsequent genetic and experimental analysis provided evidence that endogenous RA appears to control limb field positioning and regulates gene expression in the proximal limb bud mesenchyme. In responsive cells RA interacts with its nuclear receptor heterodimers, and this complex activates the RA-response elements (RARE) which are cis-regulatory modules (CRMs) controlling the expression of RA target genes (Cunningham & Duester, 2015). Established RA target genes include the Hox4/5 genes that function in forelimb field positioning (Fig. 2A, upper panel). In agreement, genetic analysis of mouse embryos lacking key enzymes for RA synthesis such Rdh10 or Raldh2/3 shows that the resulting RA deficiency disrupts establishment of the Tbx5 expression domain in the LPM (Cunningham, Chatzi, Sandell, Trainor, & Duester, 2011; Cunningham et al., 2013; Zhao et al., 2009). Further molecular analysis showed that RA signaling prevents expansion of Fgf8 expression into the developing forelimb field and this inhibition enables the activation of Tbx5 in the forelimb field (Fig. 2A, upper panel; Cunningham et al., 2013). A recent study using early chicken embryos implicates the transcription factor Cux2 in establishment of the forelimb field by directly regulating Raldh2 and Hoxb in the LPM (Fig. 2A, upper panel; Ueda et al., 2019). Together, these studies reveal several of the key players in the transcription factor networks that regulate the species-specific positioning of fore- and hindlimb fields in the LMP of tetrapod embryos (Fig. 2A).

68

Aimee Zuniga and Rolf Zeller

3. “Pre-patterning”: Polarizing the limb bud axes upstream of morphogenetic SHH signaling While classical studies indicated that morphogenetic SHH signaling by the posterior polarizing region establishes the AP limb bud axis (see Section 1), it is now clear that the nascent limb bud mesenchyme is polarized by interactions among transcription factors upstream of morphogenetic signaling (Fig. 2B). Genetic analysis first evidenced the pre-patterning mechanism in which the Gli3 and Hand2 transcriptional regulators function in a mutually antagonistic manner, which polarizes their expression along the AP axis prior to activation of SHH signaling (Galli et al., 2010; Osterwalder et al., 2014; te Welscher, Fernandez-Teran, Ros, & Zeller, 2002). This interaction is functionally relevant as genetic inactivation of Gli3 causes anterior expansion of Hand2 expression and preaxial polydactyly. Conversely, Gli3 expands posteriorly in Hand2-deficient mouse limb buds concurrent with the disruption of Shh activation (Galli et al., 2010; te Welscher, Fernandez-Teran, et al., 2002). While only one rudimentary digit forms in Hand2-deficient limbs, limb buds lacking both Gli3 and Hand2 are apolar (i.e., completely lack AP polarity) and the resulting limbs are severely polydactylous (Galli et al., 2010). The GLI3 repressor isoform (GLI3R) and the HAND2 bHLH transcription factor are initially co-expressed, but the antagonistic interactions trigger rapid establishment of distinct anterior GLI3R- and posterior HAND2-positive mesenchymal domains (Osterwalder et al., 2014). These studies reveal the essential functions of GLI3R and HAND2 in the AP polarization of the limb bud mesenchyme before SHH signaling. Several other transcriptional regulators are part of the GRNs that establish the initial axis polarity in the limb field mesenchyme (Fig. 2B). The transcriptional upregulation of Gli3 in the anterior mesenchyme depends on the Irx3/5 and the Sall4 transcriptional regulators by directly interacting with a Gli3 enhancer active in the anterior limb bud mesenchyme (Akiyama et al., 2015; Li, Sakuma, et al., 2014) and in turn, GLI3 up-regulates their expression as part of a positive transcriptional feedback (Yokoyama et al., 2017). In Irx3/5-deficient mouse hindlimb buds, Hand2 expression and SHH pathway activity are anteriorly expanded in agreement with the reduced Gli3 expression. This causes agenesis of the anterior skeletal elements in Irx3/5-deficient hindlimbs, which is largely restored by additional genetic reduction of Shh (Li, Sakuma, et al., 2014).

Self-regulatory control of limb development

69

Very similar molecular changes and skeletal defects are observed in Sall4deficient limb buds, and molecular analysis showed that Sall4 acts epistatic to Irx3/5 (Akiyama et al., 2015). Sall4 also acts genetically upstream of the interactions of Plzf with Gli3 that specify proximal identities (Barna, Pandolfi, & Niswander, 2005) and of the Hox5 genes that restrict 50 HoxD genes and Shh activation to the posterior mesenchyme (Akiyama et al., 2015; Xu et al., 2013). These interactions provide molecular insights into how AP and PD patterning are co-regulated from early limb field/bud stages onward (Fig. 2B). Recent analysis has identified a large number of additional transcription factors expressed in the anterior limb bud mesenchyme at early stages (Yokoyama et al., 2017), which indicates that the GRNs specifying the initial AP polarity are likely more complex. Expression of Hand2 in the posterior limb bud mesenchyme is regulated by a transcriptional network of different genes (Fig. 2B). Hox9 paralogues are required to activate Hand2 expression in the forelimb bud mesenchyme, while it is activated by the LIM-homeodomain transcription factor Isl1 in the hindlimb field (Itou et al., 2012; Xu & Wellik, 2011). Furthermore, the PBX homeodomain proteins participate in the early transcriptional upregulation of Hand2 and other posterior genes as their expression is reduced or absent in mutant hindlimb buds from early stages onward (Capellini et al., 2006). The activation of Hand2 (and Shh) expression are also disrupted in forelimb buds lacking both HoxA and HoxD gene functions (Sheth et al., 2013). These studies establish that in addition to GLI3Rmediated repression, Hand2 expression is positively regulated by different homeobox transcription factors. In summary, these studies corroborate the essential functions of the GLI3R-HAND2 pre-patterning system in setting up the initial mesenchymal pattern upstream morphogenetic signaling. Recently, it has been proposed that a better understanding of this pre-patterning system (Fig. 2B) may inform the etiology of human congenital limb malformations as mutations affecting the human orthologues of these pre-patterning genes alter human limb skeletal development in a very similar manner (Tao, Kawakami, Hui, & Hopyan, 2017). The HAND2-controlled GRN relevant to pre-patterning was identified by Osterwalder et al. (2014) using an epitope-tagged Hand2 mouse allele to detect the direct transcriptional targets of the endogenous HAND2 protein complexes. This analysis established that GLI3 and HAND2 directly regulate each other’s expression to establish the anterior GLI3R and posterior HAND2-positive limb bud mesenchymal compartments

70

Aimee Zuniga and Rolf Zeller

(Fig. 2B; see also Lewandowski et al., 2015; Vokes, Ji, Wong, & McMahon, 2008). In addition, HAND2 directly regulates the activation and/or upregulation of target genes in the posterior mesenchyme. In contrast, anterior and proximal target genes are repressed as they are up-regulated in Hand2-deficient limb buds (Osterwalder et al., 2014). This can be explained by the fact that HAND2 can function either as a transcriptional activator or repressor by forming heterodimers with other transcriptional regulators (Zhang et al., 2010). Tbx3 is a HAND2 target whose expression is positively regulated and genetic analysis showed that Tbx3 is required in parallel to Hand2 to restrict Gli3 expression form the posterior mesenchyme. In fact, the HAND2 GRN architecture reveals several regulatory interactions indicative of a transcriptional “reinforcement” mechanism: (1) HAND2 is required to activate Shh and up-regulates Ets1/2 expression, which reinforces the upregulation of Shh expression. (2) HAND2 and its transcriptional target TBX3 function in parallel to establish a sharp posterior Gli3 expression boundary (Fig. 2B; Lettice et al., 2012; Osterwalder et al., 2014). Tbx3 is expressed in both the anterior and posterior mesenchyme at early stages and its genetic inactivation in mouse limb buds causes loss of the posterior-most digit 5 and preaxial polydactyly in forelimbs, while only one rudimentary digit forms in Tbx3-deficient hindlimbs (Frank, Emechebe, Thomas, & Moon, 2013). In the posterior forelimb bud, Tbx3 is required to maintain Tbx5 expression and to up-regulate both Hand2 and Shh pathway genes (Emechebe et al., 2016), which reveals the existence of a positive feedback between Tbx3 and Hand2 (Fig. 2B). In the anterior mesenchyme, the TBX3 protein associates with the cilia of LMPs to regulate stability of the GLI3 full-length activator (GLI3A) and its processing to the GLI3R isoform in a complex with KIF7 and SUFU (Emechebe et al., 2016). In Tbx3-deficient forelimb buds, the levels of both GLI3 isoforms are very much reduced, which is the molecular cause of the preaxial polydactyly similar to the one caused by decreased GLI3R activity (Wang, Ruther, & Wang, 2007). The studies by Osterwalder et al. (2014) and Emechebe et al. (2016) reveal the multiple functions of TBX3 in the molecular circuitry that controls establishment of the posterior and anterior limb bud mesenchymal domains (Fig. 2B). Furthermore, it has been shown that TBX3 interacts with components of the RNA splicing machinery in mouse embryos (Kumar et al., 2014), which shows that TBX3 impacts the molecular control of limb bud development at multiple levels.

Self-regulatory control of limb development

71

4. Establishment of AER-FGF signaling is linked dorso-ventral and proximo-distal axis patterning The formation of a functional AER expressing several FGF and BMP ligands at the DV interface of the limb bud ectoderm is induced by mesenchymal signals (Fig. 3A and schemes in Fig. 3B). During formation of limb buds, the activation of Fgf8 expression depends on mesenchymal FGF10

Fig. 3 Establishment of AER-FGF signaling is directly linked to DV axis patterning and promotes coordinated PD and AP axis patterning and outgrowth. (A) Scheme of the relevant molecular interactions. Establishment of AER-Fgf8 expression is initiated by mesenchymal FGF10 signaling in concert with ectodermal WNT signaling (GRNs indicated by no. 1), which also induces concurrent establishment of DV axis polarity (GRN no. 2). AER-Fgfs are key regulators of both PD limb axis outgrowth and patterning (GRN no. 3) and establishment of the SHH/GREM1/AER/FGF signaling system in the posterior limb bud mesenchyme (GRN no. 4). Note that CYB26b1 degrades retinoic acid (RA) and the GREM1 protein directly antagonizes the BMP4 ligand. Positive inputs on gene expression (activation, upregulation) are indicated by the following symbol: ⊸. Repressive inputs (downregulation, termination) are indicated by the following symbol: a. (B) Limb bud diagrams illustrating the spatial territories involved in establishment of AER and ZPA and axes polarity. Upper panel: dorsal-ventral polarity (posterior view); lower panel: proximo-distal and anterior-posterior polarities (dorsal view).

72

Aimee Zuniga and Rolf Zeller

which signals to its cognate FGF receptor 2b in the overlying ectoderm (De Moerlooze et al., 2000). In addition to FGF10, AER formation depends critically on mesenchymal BMP4 signaling, which is transduced by BMP receptor 1 and SMAD4 in the ectoderm (Fig. 3A—GRN no. 1; Ahn, Mishina, Hanks, Behringer, & Crenshaw, 2001; Benazet & Zeller, 2013; Pajni-Underwood, Wilson, Elder, Mishina, & Lewandoski, 2007). These two signaling pathways act together with WNT/β-Catenin-mediated signal transduction in the ectoderm to establish DV axis polarity concurrent with the Fgf8-expressing AER (Soshnikova et al., 2003). In addition, the p63 and DLX5/6 transcriptional regulators are required as their genetic inactivation disrupts AER morphology, causing strikingly similar limb congenital malformations in both mice and humans (Gurrieri & Everman, 2013). A recent study showed that the SP6 and SP8 transcriptional activators activate AER-Fgf8 expression downstream of WNT/β-Catenin-mediated signal transduction (Fig. 3A—GRN no. 1; Haro et al., 2014). In addition, SP6 and SP8 are required in concert with BMP signaling to activate the En1 homeodomain transcription factor in the ventral ectoderm, which in turn initiates establishment of DV limb axis polarity (Fig. 3A—GRN no. 2; Haro et al., 2014). EN1 represses Wnt7a from the ventral ectoderm and WNT7a signaling from the dorsal ectoderm induces Lmx1b expression in the underlying mesenchyme (reviewed by Delgado & Torres, 2017). The differential expression of En1 in the ventral and Wnt7a/Lmx1b in the dorsal limb bud hallmark establishment of the DV axis during the onset of limb bud development (Fig. 3A—GRN no. 2). LMX1b is part of a LIM transcription factor network together with LHX2 and LHX9 that coordinately controls Shh, Grem1 and Fgf10 expression during distal limb bud outgrowth and patterning (Tzchori et al., 2009). Several additional Fgfs (Fgf4, Fgf9 and Fgf17) are sequentially activated in the Fgf8-expressing AER, which increases FGF signaling to the underlying mesenchyme. Mouse genetics showed that AER-FGFs regulate PD outgrowth and patterning in an instructive manner (Fig. 3A—GRN no. 3; Mariani et al., 2008). It has been proposed that the proximal and distal identities of LMPs are specified early, but that their PD identities are determined just prior to initiating their chondrogenic differentiation (Tabin & Wolpert, 2007). The transcription factors implicated in specification and determination of mesenchymal PD identities include the Meis1 and Meis2 transcription factors that are initially expressed throughout the nascent limb bud mesenchyme, but become proximally restricted during outgrowth (Mercader et al., 1999). Implantation of RA-loaded beads into the limb

Self-regulatory control of limb development

73

bud mesenchyme up-regulates Meis expression, which indicated that their expression could be directly controlled by RA signaling from the lateral plate mesenchyme (Mercader et al., 2000). However, genetic disruption of RA synthesis in the mouse limb field showed that the expression of Meis2 and other putative RA targets such as Hand2 and Shh is not affected (Cunningham et al., 2011, 2013; Zhao et al., 2009). While genetic analysis of Meis1/2 functions during mouse limb bud development is still lacking, ectopic expression of Meis1 causes distal to proximal transformations (Mercader et al., 2009). A recent study shows that the polycomb repressive complex PCR1 which functions in histone H2A mono-ubiquitylation, represses Meis2 during distal limb bud outgrowth by competing with inductive RA signals (Yakushiji-Kaminatsui et al., 2018). This is reinforced by AER-FGF signaling which activates the expression of the RA-degrading enzyme Cyp26b1 in the distal mesenchyme as soon as the AER is formed (Fig. 3A—GRN no. 3; Yashiro et al., 2004). This results in establishment of distinct proximal Meis2/Rarb and distal Cyb26b1 expression domains already during the earliest limb bud stages. During limb bud outgrowth, Cyb26b1 expression is up-regulated as part of the SHH/GREM1/ AER-FGF feedback loop that coordinately controls AP und PD limb axes development (Fig. 3A—GRNs no. 3, 4; Probst et al., 2011). In Cyp26b1deficient mouse embryos, limb development is disrupted due to mesenchymal cell death and the limb skeleton is severely truncated similar to what is observed by the teratogenic effects of excessive RA (Yashiro et al., 2004; Zhou & Kochhar, 2004). Collectively, these studies imply that limb bud outgrowth and formation of skeletal elements are incompatible with active RA signaling. While the early expression of 50 Hox genes (paralogous groups 9–13) is nested along the AP limb bud axes, their late expression domains mark specific PD domains during limb bud outgrowth (Fig. 3A—GRN no. 3; reviewed by Kherdjemil & Kmita, 2018; Zakany & Duboule, 2007). In agreement with their late PD expression domains, mouse genetics showed that the combined inactivation of either the Hox9 or Hox10 paralogues disrupts normal stylopod morphogenesis, i.e., alters proximal but not distal skeletal development (Fromental-Ramain, Warot, Lakkaraju, et al., 1996; Wellik & Capecchi, 2003). The combined inactivation of the more distally expressed Hoxa11 and Hoxd11 genes disrupts zeugopod morphogenesis, while the Hoxa13 and Hoxd13 paralogues are essential for normal autopod and digit ray development (Davis, Witte, Hsieh-Li, Potter, & Capecchi, 1995; Fromental-Ramain, Warot, Messadecq, et al., 1996). The spatio-temporal

74

Aimee Zuniga and Rolf Zeller

expression of 50 HoxD genes is controlled by enhancers embedded in the bi-modal chromatin topology of the genomic landscape that regulates their expression. These chromatin topology domains control the early and late 50 HoxD expression domains by distinct and temporally controlled interactions of the relevant enhancers with their respective target promoters (Fabre, Benke, Manley, & Duboule, 2015). With respect to 50 HoxA genes, it has recently been shown that the polycomb repressive complex PRC2 has a dual role in regulating their spatio-temporal expression in mouse limb buds (Gentile et al., 2019). Direct interactions of PRC2 with HoxA promoters repress target gene expression by generating inactive chromatin. In addition, PRC2 promotes long-range interactions among the different PRC2 bound regions, which results in formation of different chromatin topologies among the genes of the HoxA cluster in the proximal and distal limb bud. Interestingly, these distant PRC2 interactions not only repress target gene expression but also promote interactions of the neighboring, transcriptionally active HoxA genes with their distant enhancers (Gentile et al., 2019).

5. The self-regulatory limb bud signaling system coordinately controls antero-posterior and proximo-distal axis patterning and outgrowth Posteriorly restricted Shh activation is key to normal limb bud development and establishment of an ectopic anterior SHH signaling domain is one of the most common causes of preaxial digit polydactylies in humans and other species (Zuniga, Probst, & Zeller, 2012). Shh expression in the posterior limb bud mesenchyme is controlled by a far upstream and evolutionary conserved CRM called ZRS and mutations altering this CRM cause a large number of different preaxial polydactylies (Lettice, Hill, Devenney, & Hill, 2008). Genetic inactivation or functional degeneration of the ZRS disrupts ougrowth, results in loss of posterior limb skeletal elements and is one of the key alterations underlying evolutionary limb loss in, e.g., snakes (Kvon et al., 2016; Lettice et al., 2008). The transcription regulators controlling limb pre-patterning (Fig. 2B) also directly impact the posterior restriction and activation of Shh expression (Fig. 3A—GRN no. 4). Mouse genetics showed that the transcription factors encoded by the HoxA and HoxD gene clusters (Kherdjemil & Kmita, 2018; Kmita et al., 2005; Sheth et al., 2013) and AER-Fgf expression up-regulate Shh expression in the posterior limb bud mesenchyme (Mariani et al., 2008).

Self-regulatory control of limb development

75

In fact, HAND2, ETS and HOX transcription factor complexes interact with distinct regions in the ZRS to transactivate and up-regulate Shh expression (Galli et al., 2010; Lettice et al., 2012; Osterwalder et al., 2014). Interestingly, GLI3R competes with HAND2 to form transcription complexes with HOX13, which modulates activating and repressive inputs on Shh expression (Galli et al., 2010). One major aspect of Shh regulation in limb buds is the repression of Shh expression in the anterior mesenchyme. The genetic interaction of Hox5 genes with Plzf not only regulates AP prepatterning, but is also required to suppress establishment of an anterior ectopic Shh expression domain (Xu et al., 2013). In addition, the AERFGF dependent expression of the Etv4 and Etv5 transcription factors in the distal limb bud mesenchyme also represses Shh transcription in the anterior limb bud mesenchyme (Mao, McGlinn, Huang, Tabin, & McMahon, 2009; Zhang, Verheyden, Hassell, & Sun, 2009). Interestingly, HAND2 competes with EVT proteins to form heterodimers with the bHLH transcription factor TWIST1 which is also contributes to anterior repression of Shh (Fig. 2B; Firulli et al., 2005). Genetic analysis supports the functional relevance of these interactions as Etv4/5 and Hand2 act in an antagonistic manner to regulate Shh expression (Zhang et al., 2010). In addition, Gata6 is expressed in the anterior-distal mesenchyme and has been shown to directly bind to specific sites in the ZRS (Kozhemyakina, Ionescu, & Lassar, 2014). This interaction is of repressive nature as genetic inactivation of Gata6 results in anterior ectopic Shh expression and hindlimb preaxial polydactyly, while its overexpression inhibits both Shh and Grem1 expression (Kozhemyakina et al., 2014). Together, these studies reveal the multi-level control of Shh by the transcriptional complexes interacting with the ZRS to precisely position and restrict the Shh expression domain to the posterior limb bud mesenchyme (Figs. 2B and 3A—GRN no. 4). The BMP antagonist Grem1 is a key node in the self-regulatory limb bud signaling system and its expression is activated by mesenchymal cells responding to either BMP or SHH signaling (Benazet et al., 2009; Zuniga et al., 1999). It is important to note that in the absence of early SMAD4-mediated BMP signal transduction (Fig. 1C), Grem1 expression is activated by SHH, which reveals the functional complementarity of these two limb bud signaling pathways with respect to Grem1 activation (Benazet et al., 2009, 2012; Zuniga et al., 1999). Normally, the BMP-mediated activation of Grem1 expression results in rapid downregulation of BMP activity, such that the further increase and distal-anterior expansion of Grem1 transcription depends on GLI-mediated SHH signal transduction

76

Aimee Zuniga and Rolf Zeller

during outgrowth. Indeed, the Grem1 cis-regulatory landscape harbors several CRMs enriched in GLI chromatin complexes that function in regulation of the Grem1 expression dynamics (Li, Lewandowski, et al., 2014; Vokes et al., 2008; Zuniga, Laurent, et al., 2012; Zuniga et al., 2004). Not unexpected, 50 HOXA, 50 HOXD and LHX transcription factors also participate in the activation and subsequent distal-anterior expansion of Grem1 expression (Kmita et al., 2005; Sheth et al., 2013, 2016; Tzchori et al., 2009). Yet again these studies provide an excellent example of how gene expression dynamics are orchestrated by competing and complementary interactions of transcriptional activators and repressors that impact the Grem1 cis-regulatory landscape (Zuniga, Laurent, et al., 2012; Zuniga et al., 2004). These transcriptional inputs not only activate Shh and Grem1 expression but also trigger establishment of the SHH/GREM1/ AER-FGF signaling system that coordinately controls outgrowth and patterning in concert with WNT signaling (Figs. 1C and 3A—GRN no. 4; Benazet et al., 2009; ten Berge, Brugmann, Helms, & Nusse, 2008; reviewed by Zuniga, 2015). Genetic inactivation of Shh from specific mouse limb bud stages onward revealed two temporally distinct functions for SHH signaling: during the onset, SHH is first required to specify the AP identities of LMPs in agreement with its proposed morphogen functions and subsequently to promote proliferative expansion (Zhu et al., 2008). SHH signal transduction inhibits the constitutive processing of the full-length GLI3 activator (GLI3A) to the GLI3R isoform in responding mesenchymal cells, which results in a graded distribution with GLI3A being predominant in the posterior and GLI3R in the anterior mesenchyme (Wang, Fallon, & Beachy, 2000; Wang et al., 2007). Genetic evidence in the mouse shows that the transcriptional response to SHH is mostly mediated by GLI3A and GLI2 activator complexes in the posterior limb bud mesenchyme (Bowers et al., 2012). In contrast, Gli1 is not required for normal AP patterning but its transcriptional upregulation provides a direct transcriptional readout for SHH signaling in the limb bud mesenchyme. This analysis showed that the progenitors of digits 5 to 2, but not the anterior-most digit 1, are exposed to SHH signal transduction (Ahn & Joyner, 2004). Furthermore, mapping the descendants of Shh-expressing cells showed that the posterior digits 5, 4 and the posterior half of digit 3 consist largely of Shh descendants (Harfe et al., 2004). Therefore, only the progenitors of the anterior part of digit 3 and digit 2 depend on long-range SHH signaling. The study by Harfe et al. (2004) indicated that LMPs are exposed to SHH signaling as part of an expansion-based

Self-regulatory control of limb development

77

temporal gradient, while long-range morphogenetic signaling impacts digit 2 progenitors at early stages (Zhu et al., 2008). Prolonged exposure of posterior cells to SHH signaling results in desensitization as is apparent from the reduction in Gli1 expression (Harfe et al., 2004). Consistent with these spatio-temporal kinetics, GLIA target genes are expressed in distinct spatio-temporal patterns and many are up-regulated when GLI3-mediated repression is lost (Lewandowski et al., 2015; Probst et al., 2011; te Welscher, Zuniga, et al., 2002; Vokes et al., 2008). In particular, Lewandowski et al. (2015) established that GLIA target genes whose expression in posteriordistal mesenchyme is dynamic depend on sustained SHH signaling. In contrast, GLIA target genes whose expression is restricted to the posterior or core mesenchyme only require SHH signaling during early limb bud development. The results of this genome-wide analysis agree with previous analysis that revealed the differential temporal dependence of Grem1, Jag1 and 50 HoxD transcription on SHH signaling (Panman et al., 2006). Taken together these studies unravel the complex regulatory circuits that govern spatio-temporal target gene expression as part of the self-regulatory signaling system that orchestrates gene expression during limb bud outgrowth (Fig. 1C).

6. Propagation and termination of the self-regulatory limb bud signaling system The self-regulatory properties of the SHH/GREM1/AER-FGF signaling system are apparent from its network architecture which consists of several interlinked feedback loops (Benazet et al., 2009). The upregulation and distal-anterior progression of Grem1 expression lowers mesenchymal BMP activity by direct antagonism, which enables further upregulation of AER-Fgfs and mesenchymal Shh expression (Benazet et al., 2009). These signal/antagonist interactions and WNT signaling promote survival and proliferative expansion of LMPs (Benazet et al., 2009; ten Berge et al., 2008), including the expanding population of Shh descendants in the posterior limb bud mesenchyme (Harfe et al., 2004). Their expansion increases the spatial separation between the Shh and Grem1 expression domains as Shh descendants are refractory to Grem1 expression (Scherz et al., 2004). Experimental analysis indicated that the widening gap between the Shh and Grem1 domains results in self-termination of the feedback signaling system. Furthermore, genetic analysis and experimental manipulation of AER-FGF signal transduction identified several additional regulatory

78

Aimee Zuniga and Rolf Zeller

interactions that contribute to self-termination (Fig. 1C). Initially it was shown that the progressive increase in AER-FGF signal transduction triggers an inhibitory feedback in the distal mesenchyme that terminates Grem1 expression (Verheyden & Sun, 2008). This study did however not identify the transcriptional repressors that function in termination of Grem1 expression. The Tbx2 repressor is up-regulated in the posterior-distal hindlimb bud mesenchyme and represses Grem1 transcription by binding to CRMs in the Grem1 cis-regulatory landscape (Farin et al., 2013). In the anterior handplate of mouse forelimb buds, GLI3R restricts and terminates Grem1 expression, which in turn enables the BMP/SMAD4-dependent exit of proliferating LMPs toward chondrogenic differentiation (Benazet et al., 2012; Lopez-Rios et al., 2012). In particular, GLI3R constrains LMP proliferation by downregulating positive regulators of the cell-cycle such as Cdk6. This analysis shows that GLI3R constrains digit numbers to pentadactyly by its dual function as a cell-cycle regulator and and as a gatekeeper toward chondrogenic differentiation (Lopez-Rios et al., 2012). A recent study using chicken wing buds uncovered an intrinsic cell-cycle timer that operates in distal LMPs and functions in terminating their proliferation during initiation of digit ray formation (Pickering et al., 2018). This cell-cycle timer is regulated and terminated by a progressive increase in Bmp expression and by a switch from BMP antagonism to BMP signaling in the wing bud mesenchyme.

7. A Turing-type mechanism controls the definitive periodic pattern of digit rays As SHH signaling specifies the AP identities of LMPs prior to their massive proliferative expansion (Zhu et al., 2008), another mechanism must determine the definitive digit pattern during hand- and footplate development. Loss-of-function analysis in mice has established that the nested and distal-anteriorly expanding expression domains of the 50 HoxA/D genes (11–13 paralogues) in the autopod mesenchyme control digit numbers and identities (reviewed by Kherdjemil & Kmita, 2018; Zakany & Duboule, 2007). Furthermore, it was proposed that the ratio between 50 HOXD and GLI3R proteins regulates digit numbers as 50 HOXD proteins forming heterodimers with GLI3R convert the repressor into a transcription activator complex (Chen et al., 2004). In contrast, the signals that regulate the periodic pattern and numbers of digits remained largely unknown. Analysis of digit ray formation during chicken

Self-regulatory control of limb development

79

Fig. 4 Control of digit formation by a Turing-type patterning mechanism. The scheme is an overlay over an enhanced and transposed Sox9 expression pattern in a mouse limb bud at E11.5. Indicated are the five digits in anterior to posterior sequence (D1–D5) and their respective interdigit domains (ID). The scheme shows the regulation of the digit/ interdigit pattern and phalanx-joint formation by the Wnt and Bmp signaling pathways, which also impact Sox9 expression in the digit mesenchyme. The scheme to the left depicts the Turning-type network controlling the establishment of the periodic digit pattern. This network and phalanx-joint formation are modulated by 50 Hox genes and AER-Fgfs.

autopod development showed that identities are “fixed” (determined) late by differential BMP signaling from the interdigit (ID) mesenchyme to the forming digits (Dahn & Fallon, 2000). This differential BMP signaling from the ID mesenchyme generates unique phospho(p)-SMAD1/5/8 activity signatures in the distal-most mesenchyme of the forming digit rays, which are the so-called phalanx-forming regions (PFR; Fig. 4; Suzuki, Hasso, & Fallon, 2008). The mesenchymal cells of the PFR give rise to chondrocytes, and their identity is determined as they condense to form the cartilage models of the future digit phalanges. This study provided experimental evidence that the specific pSMAD1/5/8 signatures are part of the mechanism that defines the definitive digit identities (Suzuki et al., 2008). Genetic analysis in mice confirmed these findings and showed in addition that pSMAD1/5/8 activation in the ID mesenchyme is inhibited by β-Catenin-mediated WNT signal transduction

80

Aimee Zuniga and Rolf Zeller

(Witte, Chan, Economides, Mundlos, & Stricker, 2010). This study also provided evidence that the variable degrees of digit shortening observed in mouse and human brachydactyly phenotypes are a consequence of impairing the pSMAD1/5/8-positive PFR. It was then shown that during mouse digit ray development, GLI3R functions in the ID mesenchyme to up-regulate BMP signaling while Bmp expression is inhibited by 50 HoxA/D gene functions (Huang et al., 2016). In mouse embryos, these 50 HoxA/D-Gli3 interactions also control the periodic pattern and numbers of digit primordia during handplate development (Fig. 4; Sheth et al., 2012). Progressive genetic reduction of the 50 HoxA/D gene dosage in Gli3-deficient limb buds results in increasingly more severe polydactyly. As the digit numbers increase, they become progressively thinner and more densely packed. Data-based simulations of the interactions underlying these polydactylous phenotypes allowed the authors to hypothesise that 50 HOXA/D genes determine the digit periodicity in concert with FGF signaling by impacting the wavelength of a self-organizing Turing-type patterning system (Sheth et al., 2012). By combining molecular analysis and experimental manipulation with data-based model simulations, Raspopovic, Marcon, Russo, and Sharpe (2014) identified the BMP and WNT signaling pathways as potential Turing morphogens (Fig. 4). Indeed, BMP and WNT signal transduction control the periodic expression of the Sox9 transcription factor in the osteo-chondrogenic progenitors (Akiyama et al., 2005) that give rise to the digit ray primordia. In this three node Turing-type patterning system, BMP2 signaling from the ID mesenchyme positively regulates the periodic Sox9 expression, while WNT signal transduction is out of phase with Sox9 and highest in the ID mesenchyme where it inhibits Sox9 expression (Raspopovic et al., 2014). Sox9 is an integral part of this Turing system as its inactivation disrupts the periodicity of gene expression in both digit rays and ID mesenchyme. Simulations of this Turing network including its modulation by 50 HOXA/D transcription factors and FGF signaling (Fig. 4) recapitulate the formation of wild-type and experimentally or genetically altered digit patterns in silico (Raspopovic et al., 2014; Sheth et al., 2012). Last but not least, comparative analysis and simulation of shark fin and mouse digit ray development indicates that their diverse morphologies can be explained by adaptive re-organization of an evolutionary ancient BMP-Sox9-WNT Turing system (Onimaru et al., 2015).

Self-regulatory control of limb development

81

8. Brief outlook As exemplified by the research described in this review, significant progress has been made in uncovering the GRNs that orchestrate tetrapod limb bud morphogenesis. This has significantly increased our understanding of the complexity of interactions and provided insights into the interplay among the different signaling pathways and downstream transcriptional mediators. In particular, the systematic analysis of limb bud development is providing deep functional insights into the signaling systems, GRNs and cis-trans-regulatory interactions that control the morpho-regulatory dynamics during limb bud development. It is also eminent that data-based simulations are powerful tools to gain insights into the regulatory logics defined by the molecular architecture of the GRNs that orchestrate limb bud organogenesis (Iber & Zeller, 2012; Uzkudun, Marcon, & Sharpe, 2015). Significant redundancy is observed at all levels within the signaling systems and GRNs controlling limb bud development. However, it is clear that novel approaches able to capture temporal and spatial differences at high resolution are required to gain further insights into these dynamics. For example, single cells analysis of limb buds at different stages in combination with data-based simulations of the underlying network interactions may start to provide this spatio-temporal resolution. In particular, the genetic analysis of the cis-regulatory control of gene expression is revealing the functional complementation and redundancy among the CRMs that regulate particular genes during limb bud development (Osterwalder et al., 2018). The majority of genes or gene clusters functioning in limb buds are embedded in large genomic landscapes (or TADs: topologically associating domains) and their dynamic expression is regulated by CRM clusters embedded in chromatin domains that can switch between active and inactive states (reviewed by Darbellay & Duboule, 2016). Loss-of-function analysis of individual CRMs is often hampered by extensive functional complementation and/or redundancy, which points to cis-regulatory robustness of gene expression. Last but not least, the analysis of CRMs and GRNs from different species starts to provide fascinating molecular insights into the molecular changes underlying evolutionary diversification of tetrapod limbs allowing them to serve different specialized functions (see, e.g., Gehrke & Shubin, 2016; Lopez-Rios et al., 2014).

82

Aimee Zuniga and Rolf Zeller

Acknowledgments The authors wish to thank the Swiss National Science Foundation (SNF 310030_184734), the European Research Council (ERC-AdG 695032 INTEGRAL) and the University of Basel for generously supporting their research on the molecular mechanisms governing vertebrate limb bud development and evolutionary diversification for many years. The authors apologize to all researchers whose contributions could not be discussed in this review.

References Ahn, S., & Joyner, A. L. (2004). Dynamic changes in the response of cells to positive hedgehog signaling during mouse limb patterning. Cell, 118, 505–516. Ahn, K., Mishina, Y., Hanks, M. C., Behringer, R. R., & Crenshaw, E. B., 3rd. (2001). BMPR-IA signaling is required for the formation of the apical ectodermal ridge and dorsal-ventral patterning of the limb. Development, 128, 4449–4461. Akiyama, R., Kawakami, H., Wong, J., Oishi, I., Nishinakamura, R., & Kawakami, Y. (2015). Sall4-Gli3 system in early limb progenitors is essential for the development of limb skeletal elements. Proceedings of the National Academy of Sciences of the United States of America, 112, 5075–5080. Akiyama, H., Kim, J. E., Nakashima, K., Balmes, G., Iwai, N., Deng, J. M., et al. (2005). Osteo-chondroprogenitor cells are derived from Sox9 expressing precursors. Proceedings of the National Academy of Sciences of the United States of America, 102, 14665–14670. Barna, M., Pandolfi, P. P., & Niswander, L. (2005). Gli3 and Plzf cooperate in proximal limb patterning at early stages of limb development. Nature, 436, 277–281. Benazet, J. D., Bischofberger, M., Tiecke, E., Goncalves, A., Martin, J. F., Zuniga, A., et al. (2009). A self-regulatory system of interlinked signaling feedback loops controls mouse limb patterning. Science, 323, 1050–1053. Benazet, J. D., Pignatti, E., Nugent, A., Unal, E., Laurent, F., & Zeller, R. (2012). Smad4 is required to induce digit ray primordia and to initiate the aggregation and differentiation of chondrogenic progenitors in mouse limb buds. Development, 139, 4250–4260. Benazet, J. D., & Zeller, R. (2013). Dual requirement of ectodermal Smad4 during AER formation and termination of feedback signaling in mouse limb buds. Genesis, 51, 660–666. Bowers, M., Eng, L., Lao, Z., Turnbull, R. K., Bao, X., Riedel, E., et al. (2012). Limb anterior-posterior polarity integrates activator and repressor functions of GLI2 as well as GLI3. Developmental Biology, 370, 110–124. Capellini, T. D., Di Giacomo, G., Salsi, V., Brendolan, A., Ferretti, E., Srivastava, D., et al. (2006). Pbx1/Pbx2 requirement for distal limb patterning is mediated by the hierarchical control of Hox gene spatial distribution and Shh expression. Development, 133, 2263–2273. Chen, Y., Knezevic, V., Ervin, V., Hutson, R., Ward, Y., & Mackem, S. (2004). Direct interaction with Hoxd proteins reverses Gli3-repressor function to promote digit formation downstream of Shh. Development, 131, 2339–2347. Chiang, C., Litingtung, Y., Harris, M. P., Simandl, B. K., Li, Y., Beachy, P. A., et al. (2001). Manifestation of the limb prepattern: Limb development in the absence of sonic hedgehog function. Developmental Biology, 236, 421–435. Cunningham, T. J., Chatzi, C., Sandell, L. L., Trainor, P. A., & Duester, G. (2011). Rdh10 mutants deficient in limb field retinoic acid signaling exhibit normal limb patterning but display interdigital webbing. Developmental Dynamics, 240, 1142–1150. Cunningham, T. J., & Duester, G. (2015). Mechanisms of retinoic acid signalling and its roles in organ and limb development. Nature Reviews. Molecular Cell Biology, 16, 110–123.

Self-regulatory control of limb development

83

Cunningham, T. J., Zhao, X., Sandell, L. L., Evans, S. M., Trainor, P. A., & Duester, G. (2013). Antagonism between retinoic acid and fibroblast growth factor signaling during limb development. Cell Reports, 3, 1503–1511. Dahn, R. D., & Fallon, J. F. (2000). Interdigital regulation of digit identity and homeotic transformation by modulated BMP signaling. Science, 289, 438–441. Darbellay, F., & Duboule, D. (2016). Topological domains, metagenes, and the emergence of pleiotropic regulations at Hox loci. Current Topics in Developmental Biology, 116, 299–314. Davis, A. P., Witte, D. P., Hsieh-Li, H. M., Potter, S. S., & Capecchi, M. R. (1995). Absence of radius and ulna in mice lacking hoxa-11 and hoxd-11. Nature, 375, 791–795. De Moerlooze, L., Spencer-Dene, B., Revest, J. M., Hajihosseini, M., Rosewell, I., & Dickson, C. (2000). An important role for the IIIb isoform of fibroblast growth factor receptor 2 (FGFR2) in mesenchymal-epithelial signalling during mouse organogenesis. Development, 127, 483–492. Delgado, I., & Torres, M. (2017). Coordination of limb development by crosstalk among axial patterning pathways. Developmental Biology, 429, 382–386. Deschamps, J., & Duboule, D. (2017). Embryonic timing, axial stem cells, chromatin dynamics, and the Hox clock. Genes & Development, 31, 1406–1416. Duboc, V., & Logan, M. P. (2011). Pitx1 is necessary for normal initiation of hindlimb outgrowth through regulation of Tbx4 expression and shapes hindlimb morphologies via targeted growth control. Development, 138, 5301–5309. Emechebe, U., Kumar, P. P., Rozenberg, J. M., Moore, B., Firment, A., Mirshahi, T., et al. (2016). T-box3 is a ciliary protein and regulates stability of the Gli3 transcription factor to control digit number. eLife, 5, e07897. Fabre, P. J., Benke, A., Manley, S., & Duboule, D. (2015). Visualizing the HoxD gene cluster at the nanoscale level. Cold Spring Harbor Symposia on Quantitative Biology, 80, 9–16. Farin, H. F., Ludtke, T. H., Schmidt, M. K., Placzko, S., Schuster-Gossler, K., Petry, M., et al. (2013). Tbx2 terminates shh/fgf signaling in the developing mouse limb bud by direct repression of gremlin1. PLoS Genetics, 9, e1003467. Firulli, B. A., Krawchuk, D., Centonze, V. E., Vargesson, N., Virshup, D. M., Conway, S. J., et al. (2005). Altered Twist1 and Hand2 dimerization is associated with Saethre-Chotzen syndrome and limb abnormalities. Nature Genetics, 37, 373–381. Frank, D. U., Emechebe, U., Thomas, K. R., & Moon, A. M. (2013). Mouse TBX3 mutants suggest novel molecular mechanisms for Ulnar-mammary syndrome. PLoS One, 8, e67841. Fromental-Ramain, C., Warot, X., Lakkaraju, S., Favier, B., Haack, H., Birling, C., et al. (1996). Specific and redundant functions of the paralogous Hoxa-9 and Hoxd-9 genes in forelimb and axial skeleton patterning. Development, 122, 461–472. Fromental-Ramain, C., Warot, X., Messadecq, N., LeMeur, M., Dolle, P., & Chambon, P. (1996). Hoxa-13 and Hoxd-13 play a crucial role in the patterning of the limb autopod. Development, 122, 2997–3011. Galli, A., Robay, D., Osterwalder, M., Bao, X., Benazet, J. D., Tariq, M., et al. (2010). Distinct roles of Hand2 in initiating polarity and posterior Shh expression during the onset of mouse limb bud development. PLoS Genetics, 6, e1000901. Gehrke, A. R., & Shubin, N. H. (2016). Cis-regulatory programs in the development and evolution of vertebrate paired appendages. Seminars in Cell & Developmental Biology, 57, 31–39. Gentile, C., Berlivet, S., Mayran, A., Paquette, D., Guerard-Millet, F., Bajon, E., et al. (2019). PRC2-associated chromatin contacts in the developing limb reveal a possible mechanism for the atypical role of PRC2 in HoxA gene expression. Developmental Cell, 50, 184–196.

84

Aimee Zuniga and Rolf Zeller

Gurrieri, F., & Everman, D. B. (2013). Clinical, genetic, and molecular aspects of splithand/foot malformation: An update. American Journal of Medical Genetics. Part A, 161A, 2860–2872. Harfe, B. D., Scherz, P. J., Nissim, S., Tian, H., McMahon, A. P., & Tabin, C. J. (2004). Evidence for an expansion-based temporal Shh gradient in specifying vertebrate digit identities. Cell, 118, 517–528. Haro, E., Delgado, I., Junco, M., Yamada, Y., Mansouri, A., Oberg, K. C., et al. (2014). Sp6 and Sp8 transcription factors control AER formation and dorsal-ventral patterning in limb development. PLoS Genetics, 10, e1004468. Huang, B. L., Trofka, A., Furusawa, A., Norrie, J. L., Rabinowitz, A. H., Vokes, S. A., et al. (2016). An interdigit signalling centre instructs coordinate phalanx-joint formation governed by 50 Hoxd-Gli3 antagonism. Nature Communications, 7, 12903. Iber, D., & Zeller, R. (2012). Making sense-data-based simulations of vertebrate limb development. Current Opinion in Genetics & Development, 22, 570–577. Itou, J., Kawakami, H., Quach, T., Osterwalder, M., Evans, S. M., Zeller, R., et al. (2012). Islet1 regulates establishment of the posterior hindlimb field upstream of the Hand2-Shh morphoregulatory gene network in mouse embryos. Development, 139, 1620–1629. Kherdjemil, Y., & Kmita, M. (2018). Insights on the role of hox genes in the emergence of the pentadactyl ground state. Genesis, 56, e23046. Kmita, M., Tarchini, B., Zakany, J., Logan, M., Tabin, C. J., & Duboule, D. (2005). Early developmental arrest of mammalian limbs lacking HoxA/HoxD gene function. Nature, 435, 1113–1116. Kozhemyakina, E., Ionescu, A., & Lassar, A. B. (2014). GATA6 is a crucial regulator of Shh in the limb bud. PLoS Genetics, 10, e1004072. Kumar, P. P., Franklin, S., Emechebe, U., Hu, H., Moore, B., Lehman, C., et al. (2014). TBX3 regulates splicing in vivo: A novel molecular mechanism for Ulnar-mammary syndrome. PLoS Genetics, 10, e1004247. Kvon, E. Z., Kamneva, O. K., Melo, U. S., Barozzi, I., Osterwalder, M., Mannion, B. J., et al. (2016). Progressive loss of function in a limb enhancer during snake evolution. Cell, 167, 633–642 e611. Laufer, E., Nelson, C. E., Johnson, R. L., Morgan, B. A., & Tabin, C. (1994). Sonic hedgehog and Fgf-4 act through a signaling cascade and feedback loop to integrate growth and patterning of the developing limb bud. Cell, 79, 993–1003. Lettice, L. A., Hill, A. E., Devenney, P. S., & Hill, R. E. (2008). Point mutations in a distant sonic hedgehog cis-regulator generate a variable regulatory output responsible for preaxial polydactyly. Human Molecular Genetics, 17, 978–985. Lettice, L. A., Williamson, I., Wiltshire, J. H., Peluso, S., Devenney, P. S., Hill, A. E., et al. (2012). Opposing functions of the ETS factor family define Shh spatial expression in limb buds and underlie polydactyly. Developmental Cell, 22, 459–467. Lewandowski, J. P., Du, F., Zhang, S., Powell, M. B., Falkenstein, K. N., Ji, H., et al. (2015). Spatiotemporal regulation of GLI target genes in the mammalian limb bud. Developmental Biology, 406, 92–103. Li, Q., Lewandowski, J. P., Powell, M. B., Norrie, J. L., Cho, S. H., & Vokes, S. A. (2014). A Gli silencer is required for robust repression of gremlin in the vertebrate limb bud. Development, 141, 1906–1914. Li, D., Sakuma, R., Vakili, N. A., Mo, R., Puviindran, V., Deimling, S., et al. (2014). Formation of proximal and anterior limb skeleton requires early function of Irx3 and Irx5 and is negatively regulated by Shh signaling. Developmental Cell, 29, 233–240. Lopez-Rios, J., Duchesne, A., Speziale, D., Andrey, G., Peterson, K. A., Germann, P., et al. (2014). Attenuated sensing of SHH by Ptch1 underlies evolution of bovine limbs. Nature, 511, 46–51.

Self-regulatory control of limb development

85

Lopez-Rios, J., Speziale, D., Robay, D., Scotti, M., Osterwalder, M., Nusspaumer, G., et al. (2012). GLI3 constrains digit number by controlling both progenitor proliferation and BMP-dependent exit to chondrogenesis. Developmental Cell, 22, 837–848. Maccabe, A. B., Gasseling, M. T., & Saunders, J. W., Jr. (1973). Spatiotemporal distribution of mechanisms that control outgrowth and anteroposterior polarization of the limb bud in the chick embryo. Mechanisms of Ageing and Development, 2, 1–12. Mao, J., McGlinn, E., Huang, P., Tabin, C. J., & McMahon, A. P. (2009). Fgf-dependent Etv4/5 activity is required for posterior restriction of Sonic Hedgehog and promoting outgrowth of the vertebrate limb. Developmental Cell, 16, 600–606. Mariani, F. V., Ahn, C. P., & Martin, G. R. (2008). Genetic evidence that FGFs have an instructive role in limb proximal-distal patterning. Nature, 453, 401–405. Matsubara, Y., Hirasawa, T., Egawa, S., Hattori, A., Suganuma, T., Kohara, Y., et al. (2017). Anatomical integration of the sacral-hindlimb unit coordinated by GDF11 underlies variation in hindlimb positioning in tetrapods. Nature Ecology & Evolution, 1, 1392–1399. McPherron, A. C., Lawler, A. M., & Lee, S. J. (1999). Regulation of anterior/posterior patterning of the axial skeleton by growth/differentiation factor 11. Nature Genetics, 22, 260–264. Mercader, N., Leonardo, E., Azpiazu, N., Serrano, A., Morata, G., Martinez, C., et al. (1999). Conserved regulation of proximodistal limb axis development by Meis1/Hth. Nature, 402, 425–429. Mercader, N., Leonardo, E., Piedra, M. E., Martinez, A. C., Ros, M. A., & Torres, M. (2000). Opposing RA and FGF signals control proximodistal vertebrate limb development through regulation of Meis genes. Development, 127, 3961–3970. Mercader, N., Selleri, L., Criado, L. M., Pallares, P., Parras, C., Cleary, M. L., et al. (2009). Ectopic Meis1 expression in the mouse limb bud alters P-D patterning in a Pbx1-independent manner. The International Journal of Developmental Biology, 53, 1483–1494. Minguillon, C., Nishimoto, S., Wood, S., Vendrell, E., Gibson-Brown, J. J., & Logan, M. P. (2012). Hox genes regulate the onset of Tbx5 expression in the forelimb. Development, 139, 3180–3188. Moreau, C., Caldarelli, P., Rocancourt, D., Roussel, J., Denans, N., Pourquie, O., et al. (2019). Timed collinear activation of Hox genes during gastrulation controls the avian forelimb position. Current Biology, 29, 35–50. Nishimoto, S., Minguillon, C., Wood, S., & Logan, M. P. (2014). A combination of activation and repression by a collinear Hox code controls forelimb-restricted expression of Tbx5 and reveals Hox protein specificity. PLoS Genetics, 10, e1004245. Niswander, L., Jeffrey, S., Martin, G. R., & Tickle, C. (1994). A positive feedback loop coordinates growth and patterning in the vertebrate limb. Nature, 371, 609–612. Niswander, L., Tickle, C., Vogel, A., Booth, I., & Martin, G. R. (1993). FGF-4 replaces the apical ectodermal ridge and directs outgrowth and patterning of the limb. Cell, 75, 579–587. Onimaru, K., Kuraku, S., Takagi, W., Hyodo, S., Sharpe, J., & Tanaka, M. (2015). A shift in anterior-posterior positional information underlies the fin-to-limb evolution. eLife, 4, e07048. Osterwalder, M., Barozzi, I., Tissieres, V., Fukuda-Yuzawa, Y., Mannion, B. J., Afzal, S. Y., et al. (2018). Enhancer redundancy provides phenotypic robustness in mammalian development. Nature, 554, 239–243. Osterwalder, M., Speziale, D., Shoukry, M., Mohan, R., Ivanek, R., Kohler, M., et al. (2014). HAND2 targets define a network of transcriptional regulators that compartmentalize the early limb bud mesenchyme. Developmental Cell, 31, 345–357.

86

Aimee Zuniga and Rolf Zeller

Pajni-Underwood, S., Wilson, C. P., Elder, C., Mishina, Y., & Lewandoski, M. (2007). BMP signals control limb bud interdigital programmed cell death by regulating FGF signaling. Development, 134, 2359–2368. Panman, L., Galli, A., Lagarde, N., Michos, O., Soete, G., Zuniga, A., et al. (2006). Differential regulation of gene expression in the digit forming area of the mouse limb bud by SHH and gremlin 1/FGF-mediated epithelial-mesenchymal signalling. Development, 133, 3419–3428. Pickering, J., Rich, C. A., Stainton, H., Aceituno, C., Chinnaiya, K., Saiz-Lopez, P., et al. (2018). An intrinsic cell cycle timer terminates limb bud outgrowth. eLife, 7, e37429. Probst, S., Kraemer, C., Demougin, P., Sheth, R., Martin, G. R., Shiratori, H., et al. (2011). SHH propagates distal limb bud development by enhancing CYP26B1-mediated retinoic acid clearance via AER-FGF signalling. Development, 138, 1913–1923. Rancourt, D. E., Tsuzuki, T., & Capecchi, M. R. (1995). Genetic interaction between hoxb-5 and hoxb-6 is revealed by nonallelic noncomplementation. Genes & Development, 9, 108–122. Raspopovic, J., Marcon, L., Russo, L., & Sharpe, J. (2014). Modeling digits. Digit patterning is controlled by a Bmp-Sox9-Wnt Turing network modulated by morphogen gradients. Science, 345, 566–570. Riddle, R. D., Johnson, R. L., Laufer, E., & Tabin, C. (1993). Sonic hedgehog mediates the polarizing activity of the ZPA. Cell, 75, 1401–1416. Saunders, J. W., Jr. (1948). The proximo-distal sequence of origin of the parts of the chick wing and the role of the ectoderm. The Journal of Experimental Zoology, 108, 363–403. Scherz, P. J., Harfe, B. D., McMahon, A. P., & Tabin, C. J. (2004). The limb bud Shh-Fgf feedback loop is terminated by expansion of former ZPA cells. Science, 305, 396–399. Sheeba, C. J., & Logan, M. P. (2017). The roles of T-box genes in vertebrate limb development. Current Topics in Developmental Biology, 122, 355–381. Sheth, R., Barozzi, I., Langlais, D., Osterwalder, M., Nemec, S., Carlson, H. L., et al. (2016). Distal limb patterning requires modulation of cis-regulatory activities by HOX13. Cell Reports, 17, 2913–2926. Sheth, R., Gregoire, D., Dumouchel, A., Scotti, M., Pham, J. M., Nemec, S., et al. (2013). Decoupling the function of Hox and Shh in developing limb reveals multiple inputs of Hox genes on limb growth. Development, 140, 2130–2138. Sheth, R., Marcon, L., Bastida, M. F., Junco, M., Quintana, L., Dahn, R., et al. (2012). Hox genes regulate digit patterning by controlling the wavelength of a Turing-type mechanism. Science, 338, 1476–1480. Soshnikova, N., Zechner, D., Huelsken, J., Mishina, Y., Behringer, R. R., Taketo, M. M., et al. (2003). Genetic interaction between Wnt/beta-catenin and BMP receptor signaling during formation of the AER and the dorsal-ventral axis in the limb. Genes & Development, 17, 1963–1968. Suzuki, T., Hasso, S. M., & Fallon, J. F. (2008). Unique SMAD1/5/8 activity at the phalanxforming region determines digit identity. Proceedings of the National Academy of Sciences of the United States of America, 105, 4185–4190. Tabin, C., & Wolpert, L. (2007). Rethinking the proximodistal axis of the vertebrate limb in the molecular era. Genes & Development, 21, 1433–1442. Tao, H., Kawakami, Y., Hui, C. C., & Hopyan, S. (2017). The two domain hypothesis of limb prepattern and its relevance to congenital limb anomalies. Wiley Interdisciplinary Reviews: Developmental Biology, 6, e270. te Welscher, P., Fernandez-Teran, M., Ros, M. A., & Zeller, R. (2002). Mutual genetic antagonism involving GLI3 and dHAND prepatterns the vertebrate limb bud mesenchyme prior to SHH signaling. Genes & Development, 16, 421–426.

Self-regulatory control of limb development

87

te Welscher, P., Zuniga, A., Kuijper, S., Drenth, T., Goedemans, H. J., Meijlink, F., et al. (2002). Progression of vertebrate limb development through SHH-mediated counteraction of GLI3. Science, 298, 827–830. ten Berge, D., Brugmann, S. A., Helms, J. A., & Nusse, R. (2008). Wnt and FGF signals interact to coordinate growth with cell fate specification during limb development. Development, 135, 3247–3257. Tickle, C. (1981). The number of polarizing region cells required to specify additional digits in the developing chick wing. Nature, 289, 295–298. Tickle, C. (2017). An historical perspective on the pioneering experiments of John Saunders. Developmental Biology, 429, 374–381. Tickle, C., Alberts, B., Wolpert, L., & Lee, J. (1982). Local application of retinoic acid to the limb bond mimics the action of the polarizing region. Nature, 296, 564–566. Tzchori, I., Day, T. F., Carolan, P. J., Zhao, Y., Wassif, C. A., Li, L., et al. (2009). LIM homeobox transcription factors integrate signaling events that control three-dimensional limb patterning and growth. Development, 136, 1375–1385. Ueda, S., Cordeiro, I. R., Moriyama, Y., Nishimori, C., Kai, K. I., Yu, R., et al. (2019). Cux2 refines the forelimb field by controlling expression of Raldh2 and Hox genes. Biology Open, 8, bio040584. Uzkudun, M., Marcon, L., & Sharpe, J. (2015). Data-driven modelling of a gene regulatory network for cell fate decisions in the growing limb bud. Molecular Systems Biology, 11, 815. van den Akker, E., Fromental-Ramain, C., de Graaff, W., Le Mouellic, H., Brulet, P., Chambon, P., et al. (2001). Axial skeletal patterning in mice lacking all paralogous group 8 Hox genes. Development, 128, 1911–1921. Verheyden, J. M., & Sun, X. (2008). An Fgf/Gremlin inhibitory feedback loop triggers termination of limb bud outgrowth. Nature, 454, 638–641. Vokes, S. A., Ji, H., Wong, W. H., & McMahon, A. P. (2008). A genome-scale analysis of the cis-regulatory circuitry underlying sonic hedgehog-mediated patterning of the mammalian limb. Genes & Development, 22, 2651–2663. Wang, B., Fallon, J. F., & Beachy, P. A. (2000). Hedgehog-regulated processing of Gli3 produces an anterior/posterior repressor gradient in the developing vertebrate limb. Cell, 100, 423–434. Wang, C., Ruther, U., & Wang, B. (2007). The Shh-independent activator function of the full-length Gli3 protein and its role in vertebrate limb digit patterning. Developmental Biology, 305, 460–469. Wellik, D. M., & Capecchi, M. R. (2003). Hox10 and Hox11 genes are required to globally pattern the mammalian skeleton. Science, 301, 363–367. Witte, F., Chan, D., Economides, A. N., Mundlos, S., & Stricker, S. (2010). Receptor tyrosine kinase-like orphan receptor 2 (ROR2) and Indian hedgehog regulate digit outgrowth mediated by the phalanx-forming region. Proceedings of the National Academy of Sciences of the United States of America, 107, 14211–14216. Wolpert, L. (1969). Positional information and the spatial pattern of cellular differentiation. Journal of Theoretical Biology, 25, 1–47. Xu, B., Hrycaj, S. M., McIntyre, D. C., Baker, N. C., Takeuchi, J. K., Jeannotte, L., et al. (2013). Hox5 interacts with Plzf to restrict Shh expression in the developing forelimb. Proceedings of the National Academy of Sciences of the United States of America, 110, 19438–19443. Xu, B., & Wellik, D. M. (2011). Axial Hox9 activity establishes the posterior field in the developing forelimb. Proceedings of the National Academy of Sciences of the United States of America, 108, 4888–4891. Yakushiji-Kaminatsui, N., Kondo, T., Hironaka, K. I., Sharif, J., Endo, T. A., Nakayama, M., et al. (2018). Variant PRC1 competes with retinoic acid-related signals to repress Meis2 in the mouse distal forelimb bud. Development, 145, dev16634.

88

Aimee Zuniga and Rolf Zeller

Yashiro, K., Zhao, X., Uehara, M., Yamashita, K., Nishijima, M., Nishino, J., et al. (2004). Regulation of retinoic acid distribution is required for proximodistal patterning and outgrowth of the developing mouse limb. Developmental Cell, 6, 411–422. Yokoyama, S., Furukawa, S., Kitada, S., Mori, M., Saito, T., Kawakami, K., et al. (2017). Analysis of transcription factors expressed at the anterior mouse limb bud. PLoS One, 12, e0175673. Zakany, J., & Duboule, D. (2007). The role of Hox genes during vertebrate limb development. Current Opinion in Genetics & Development, 17, 359–366. Zhang, Z., Sui, P., Dong, A., Hassell, J., Cserjesi, P., Chen, Y. T., et al. (2010). Preaxial polydactyly: Interactions among ETV, TWIST1 and HAND2 control anterior-posterior patterning of the limb. Development, 137, 3417–3426. Zhang, Z., Verheyden, J. M., Hassell, J. A., & Sun, X. (2009). FGF-regulated Etv genes are essential for repressing Shh expression in mouse limb buds. Developmental Cell, 16, 607–613. Zhao, X., Sirbu, I. O., Mic, F. A., Molotkova, N., Molotkov, A., Kumar, S., et al. (2009). Retinoic acid promotes limb induction through effects on body axis extension but is unnecessary for limb patterning. Current Biology, 19, 1050–1057. Zhou, J., & Kochhar, D. M. (2004). Cellular anomalies underlying retinoid-induced phocomelia. Reproductive Toxicology, 19, 103–110. Zhu, J., Nakamura, E., Nguyen, M. T., Bao, X., Akiyama, H., & Mackem, S. (2008). Uncoupling Sonic hedgehog control of pattern and expansion of the developing limb bud. Developmental Cell, 14, 624–632. Zuniga, A. (2015). Next generation limb development and evolution: Old questions, new perspectives. Development, 142, 3810–3820. Zuniga, A., Haramis, A. P., McMahon, A. P., & Zeller, R. (1999). Signal relay by BMP antagonism controls the SHH/FGF4 feedback loop in vertebrate limb buds. Nature, 401, 598–602. Zuniga, A., Laurent, F., Lopez-Rios, J., Klasen, C., Matt, N., & Zeller, R. (2012). Conserved cis-regulatory regions in a large genomic landscape control SHH and BMP-regulated Gremlin1 expression in mouse limb buds. BMC Developmental Biology, 12, 23. Zuniga, A., Michos, O., Spitz, F., Haramis, A. P., Panman, L., Galli, A., et al. (2004). Mouse limb deformity mutations disrupt a global control region within the large regulatory landscape required for Gremlin expression. Genes & Development, 18, 1553–1564. Zuniga, A., Probst, S., & Zeller, R. (2012). The molecular basis of human congenital limb malformations. Wiley Interdisciplinary Reviews: Developmental Biology, 6, 803–822.

CHAPTER FOUR

Gene regulatory networks during the development of the Drosophila visual system Yen-Chung Chen, Claude Desplan∗ Department of Biology, New York University, New York, NY, United States ∗ Corresponding author: e-mail address: [email protected]

Contents 1. Introduction 2. Design principles of the gene regulatory networks (GRNs) in the Drosophila visual system 3. The eye 3.1 The eye disc GRN 3.2 The GRN of the morphogenetic furrow 3.3 The GRN specifying photoreceptor fate 3.4 The GRN that determines photoreceptor terminal features: Ommatidial subtypes 3.5 The GRN that determines photoreceptor terminal features: Axon targeting 3.6 GRNs in the eye: A brief summary 4. The GRN that specifies lamina neurons in response to photoreceptor signals 5. Temporal and spatial GRNs specify medulla neurons 6. The GRNs leading to the formation of the lobula and lobula plate 7. Conclusion and perspectives Acknowledgments References

90 92 93 93 94 96 98 99 102 102 105 108 111 113 114

Abstract The Drosophila visual system integrates input from 800 ommatidia and extracts different features in stereotypically connected optic ganglia. The development of the Drosophila visual system is controlled by gene regulatory networks that control the number of precursor cells, generate neuronal diversity by integrating spatial and temporal information, coordinate the timing of retinal and optic lobe cell differentiation, and determine distinct synaptic targets of each cell type. In this chapter, we describe the known gene regulatory networks involved in the development of the different parts of the visual system and explore general components in these gene networks. Finally, we discuss the advantages of the fly visual system as a model for gene regulatory network discovery in the era of single-cell transcriptomics.

Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.02.010

#

2020 Elsevier Inc. All rights reserved.

89

90

Yen-Chung Chen and Claude Desplan

1. Introduction Cell diversity in the nervous system has long fascinated generations of scientists. Back in the nineteenth century, the pioneering works of Camillo Golgi and Santiago Ramo´n y Cajal spearheaded a series of studies on how neurons differ in shape, position, and targeting properties (Golgi, 1885; Ramo´n y Cajal, 1911). The nervous system is so sophisticated that even after more than 100 years of vigorous interrogation, there are still many unknown types of neurons to be discovered (Boldog et al., 2018), which makes one wonder what underlies this enormous diversity. In vertebrates and invertebrates, the nervous system originates from a small population of stem cells that seem homogenous in the embryo. The stem cells then generate a wide array of neuron types based on their spatial location and the time when the neurons are born (Holguera & Desplan, 2018). The specification process requires distinct transcription factors expressed in specific spatial and temporal domains and results in a wide diversity of neurons generated in precisely regulated quantity and location (Doe, 2017; Jessell, 2000; Li, Chen, & Desplan, 2013; Shirasaki & Pfaff, 2002). Each type of neurons has a stereotypical pattern of innervation, sends and receives signals from a unique set of synaptic partners, and often has a distinctive mode of firing. Researchers have used reverse genetics to identify many of the genes that are involved in the establishment of features like guidance or selection of targets (Kolodkin, Matthes, & Goodman, 1993; Kolodziej et al., 1996; Messersmith et al., 1995; Mitchell et al., 1996; Seeger, Tear, Ferres-Marco, & Goodman, 1993). Furthermore, recent advances in technology have allowed electrophysiological features and gene expression to be systematically compared and have demonstrated that similarity in gene expression correlates with similarity in firing patterns (Cadwell et al., 2016; Fuzik et al., 2016). Starting from either the developmental process or the terminal features of neurons, previous studies have broadened our knowledge on how neurons obtain their identities and morphological and physiological features. Nonetheless, how taking a developmental identity leads to the establishment and maintenance of specific terminal features remains largely unknown. Understanding regulatory genes and the regulatory networks in which they are embedded would make it possible to explore the combinatorial regulation of such terminal features. A complete gene regulatory network at different times during development would encompass not only the sum

Gene regulatory networks in Drosophila visual system

91

of the relationships between each regulator and its target genes but also the interactions among all the regulators and effectors that are involved in a process. The construction of a gene regulatory network would provide insight into how the genes important for the morphological and functional features that we use to define a cell type are regulated such that they are expressed specifically and in a timely manner. With more information about regulatory interactions, we could examine the properties that emerge at the level of control loops, for example, how robustness to disruption is achieved beyond simple redundancy. The visual system of Drosophila is an ideal platform to learn about gene regulatory networks and to bridge the gap between development and terminal features that establish diversity in the nervous system. First, the neurons of the visual system have been studied extensively for decades, and as a result, the morphology and the choice of neurotransmitter of a large proportion of the neurons are characterized (Fischbach, 1989; Raghu & Borst, 2011; Raghu, Claussen, & Borst, 2013; Varija Raghu, Reiff, & Borst, 2011). Second, the high-resolution connectivity of most neuropils within the optic lobe has been actively determined by electron microscopy (Rivera-Alba et al., 2011; Shinomiya et al., 2019; Takemura et al., 2013, 2015, 2017). These known features that distinguish different types of neurons are invaluable for the discovery of gene regulatory networks, as changes in many of these features can be assessed following perturbations of potential regulators, making it easier to build a predictive model that links regulators to function. Finally, powerful genetic tools in Drosophila allow manipulation of the potential regulators in a timely and versatile fashion. For example, previous investigations in the developing optic lobe have identified transcription factors that are limited to specific spatial domains or temporal windows, and precisely timed perturbation of these transcription factors have demonstrated that they are required for the proper development of distinct types of neurons in the optic lobe (Erclik et al., 2017; Erclik, Hartenstein, Lipshitz, & McInnes, 2008; Gold & Brand, 2014; Li et al., 2013; Suzuki, Trush, Yasugi, Takayama, & Sato, 2016). Traditionally, gene regulation is discovered gene-by-gene, uncovering individual regulators for a target gene, or alternatively, the targets of a regulator are discovered by direct binding to the target sequences or by correlated changes upon perturbation of the regulator. This strategy enables a detailed but highly focused understanding of parts of larger gene regulatory networks. Recently, with insights from previous research and advances in high-throughput profiling of gene expression, systematic inference of gene

92

Yen-Chung Chen and Claude Desplan

regulatory networks has provided another strategy to investigate broad regulations (Reviewed by Huynh-Thu & Sanguinetti, 2019). This strategy holds the promise to discover novel regulatory logic that might ultimately allow us to integrate what we have learned previously from single genes into a complete network. This review focuses on the regulators in the gene regulatory network during the development of the visual system of Drosophila and discusses the promises that the fly visual system holds as a platform for systematic gene regulatory network discovery and inference.

2. Design principles of the gene regulatory networks (GRNs) in the Drosophila visual system Development is a process of gradually specifying various cell types that serve different functions from a small population of stem cells and organizing the specified cells to form a functional entity. The process is precisely regulated in number and in time. For example, neurons in the Drosophila optic lobe are generated in a fixed ratio to the number of unit-eyes, and the specification of the neurons happens in a time window from the third larval instar to early pupation. Precision and robustness of the developmental processes are critical for organisms to function properly and are made possible by sophisticated interactions among genes that regulate the timing and amount of the expression of other genes. A gene regulatory network is the sum of the interactions between genes in time and space and can be further partitioned into interconnected subnetworks that serve specific functions. In the development of the visual system of Drosophila, five subnetworks are shared across the visual system but wired differently to fit the specific needs of each region: 1. Tissue commitment subnetwork: An early subnetwork is activated during embryonic development and determines which tissue or organ is formed. Activation of the subnetwork is often sufficient for inducing ectopic formation of the corresponding tissue. 2. Progenitor expansion subnetwork: After establishing tissue identity, a progenitor subnetwork makes sure the tissue is formed in the proper size and delays further differentiation from happening until receiving proper extrinsic cues. 3. Patterning subnetwork: Along with the expansion of progenitors, the tissue is compartmentalized by a subnetwork that contains

Gene regulatory networks in Drosophila visual system

93

spatial information and modulates the progenitor subnetwork. Compartmentalization allows different cell types to be generated and organize the developing tissue according to the body plan. 4. Differentiation subnetwork: The differentiation subnetwork integrates multiple extrinsic cues to initiate specification in an accurately timed fashion and results in detailed functional subdivision of cell types. 5. Terminal specification subnetwork: A subnetwork that involves a temporal series of regulators underlies the specification of different neuron types. Different cell types are specified according to the integration of the temporal information and the spatial information from the compartmentation subnetwork. The specification process results in the initiation of the subnetworks that establish the terminal features that make each neuron type distinct.

3. The eye The Drosophila eye has a highly repetitive structure, consisting of approximately 800 unit-eyes called ommatidia. Each ommatidium contains eight photoreceptor cells, four cone cells, and pigment cells. The Drosophila eye originates from the eye imaginal disc, and previous studies of its development have uncovered a versatile gene regulatory network that is modulated by various extrinsic cues (Reviewed in Roignant & Treisman, 2009; Treisman, 2013). Many of the design principles of GRNs in the visual system of Drosophila were discovered in the eye and were later shown to apply to other parts of the visual system.

3.1 The eye disc GRN At the root of the eye development network are Twin of Eyeless (Toy) and Eyeless (Ey), the Drosophila Pax-6 homologs. Toy and Ey regulate the tissue commitment subnetwork of the eye and are expressed in the eye disc during embryonic stages that define eye disc progenitors, and their ectopic expression in heterologous imaginal discs induces ectopic eye formation (Czerny et al., 1999; Halder, Callaerts, & Gehring, 1995). At the first larval instar, Homothorax (Hth) expression begins, and Hth, Ey, Teashirt (Tsh), and Toy form the progenitor expansion subnetwork that maintains proliferation of the eye disc progenitor cells and suppresses the differentiation subnetwork (Bessa, Gebelein, Pichaud, Casares, & Mann, 2002; Lopes & Casares, 2010; Zhu, Palliyil, Ran, & Kumar, 2017).

94

Yen-Chung Chen and Claude Desplan

The developing eye is compartmentalized into dorsal and ventral compartments. The dorsal compartment is defined by the expression of Pannier (Pnr) and its target Wingless (Wg) at the dorsal margin of the eye disc. Wg signaling initiates the patterning subnetwork to delineate the dorsal and ventral compartments by activating the transcription factors Araucan and Mirror encoded by the Iroquois complex (IrodC) in the dorsal compartment (Cavodeassi, Diez Del Corral, Campuzano, & Domı´nguez, 1999; Maurel-Zaffran & Treisman, 2000). Dorsoventral compartmentation also prepares the extrinsic cues that are required for differentiation. Fringe (Fng), a glycosyl transferase, is expressed in the ventral compartment and modifies Notch: Glycosylated Notch does not respond to its ventrally expressed ligand Serrate but responds to dorsally expressed Delta (Dl). The selective sensitivity to Dl ensures that Notch signaling is only active at the boundary between dorsal and ventral compartment (Br€ uckner, Perez, Clausen, & Cohen, 2000; Cho & Choi, 1998; Domı´nguez & De Celis, 1998; Panin, Papayannopoulos, Wilson, & Irvine, 1997; Papayannopoulos, Tomlinson, Panin, Rauskolb, & Irvine, 1998) (Fig. 1A). Notch signaling at the dorsoventral midline is required to initiate the morphogenetic furrow at the posterior margin of the eye and terminal specification by promoting proliferation that brings posterior cells out of reach of Wg (see next section) (Kenyon, Ranade, Curtiss, Mlodzik, & Pignoni, 2003; Kumar & Moses, 2001).

3.2 The GRN of the morphogenetic furrow At early third larval instar, Hedgehog (Hh) and Decapentaplegic (Dpp), two secreted signaling molecules, are expressed at the posterior margin of the eye disc before differentiation starts (Baker et al., 2018; Borod & Heberlein, 1998; Pignoni et al., 1997). Dpp signaling is induced by Hh signaling and suppresses the expression of Hth. Retinal differentiation is a progressive process and is initiated when Hh signaling induces a dorsoventral stripe of Dpp expressing cells at the posterior margin, termed the morphogenetic furrow (MF) (Fig. 1A, lower panel). Dpp signaling from the MF acts at long distance on cells anterior to it and initiates the differentiation subnetwork, forming the pre-proneural zone (Bessa et al., 2002; Greenwood & Struhl, 1999). Dpp signaling allows Ey to activate Sine oculis (So) and Eyes absent (Eya) in the posterior eye disc at the second larval instar (Curtiss & Mlodzik, 2000; Halder et al., 1998; Heberlein, Heberlein, Luk, & Donohoe, 1995; Lopes & Casares, 2010;

Fig. 1 Gene regulatory networks during retinal development. (A) The eye disc progenitor gene regulatory network responds to extrinsic cues to initiate differentiation. (B) Cross-regulation of fate-specifying transcription factors induced by short-range extrinsic cues from photoreceptors specified earlier in the cluster allows sequential generation of photoreceptors in an ommatidium. (C) Spalt governs the spectral differences of outer and inner photoreceptors. (D) A network of transcription factors integrates spatial extrinsic cues and stochastic intrinsic choice to define the ommatidial subtypes that form the retinal mosaic.

96

Yen-Chung Chen and Claude Desplan

Niimi, Seimiya, Kloter, Flister, & Gehring, 1999; Ostrin et al., 2006) (Fig. 1A, lower panel). Eya and So are required to maintain the expression of Dpp, forming a positive feedback loop, and to directly activate the expression of Daschund (Dac) (Pappu et al., 2005; Pignoni et al., 1997). Eya, So, and Dac form the subnetwork that initiates differentiation and suppresses the progenitor subnetwork (Chen, Amoui, Zhang, & Mardon, 1997) (Fig. 1A, upper panel). Within the MF, short-acting Hh signaling from the posterior margin of the eye disc cooperates with Eya, So, and Dac and activates the expression of a proneural gene, Atonal (Ato) (Borod & Heberlein, 1998; Tanaka-Matakatsu & Du, 2008) (Fig. 1A, lower panel). The expression of Ato marks the beginning of specification of R8 photoreceptors and the recruitment of photoreceptor preclusters that later develop into individual ommatidia ( Jarman, Grell, Ackerman, Jan, & Jan, 1994). Differentiating photoreceptors begin to express Hh. Hh signaling from these differentiating photoreceptors reaches the pre-proneural zone and induces Dpp expression, moving the MF anteriorly. Dpp signaling from the moving MF induces a new pre-proneural zone (Borod & Heberlein, 1998). The speed of MF propagation is modulated by the interaction of extrinsic cues to prevent ectopic or precocious differentiation of photoreceptors and to allow the development of a properly sized and patterned eye. For example, Wg signaling from the anterior eye disc suppresses the expression of Ato and slows down MF propagation, while the expression of Wg is suppressed by Dpp signaling (Cadigan, Jou, & Nusse, 2002; Hazelett, Bourouis, Walldorf, & Treisman, 1998). In the pre-proneural zone, Dpp signaling modulates the speed of MF propagation by inducing Hairy, a repressor of Ato, and expression of Delta in the MF suppresses Hairy to allow differentiation to proceed (Baonza & Freeman, 2001; Brown, Sattler, Paddock, & Carroll, 1995; Greenwood & Struhl, 1999) (Fig. 1A, upper panel).

3.3 The GRN specifying photoreceptor fate Within the morphogenetic furrow, the regulation of Ato governs the initiation of photoreceptor preclusters by specifying the R8 photoreceptors that provide the signals required for the sequential specification of all other photoreceptors. Ato is first expressed at a basal level in a dorsoventral stripe in the MF and then resolves into cell clusters containing approximately 15 cells (Dokucu, Zipursky, & Cagan, 1996) (Fig. 1A and B). Within each cluster, only one cell keeps expressing Ato and is specified as an R8 photoreceptor while the other cells lose Ato. The R8 cell expresses Dl and Scabrous (Sca) to

Gene regulatory networks in Drosophila visual system

97

inhibit Ato expression in the other cella via Notch-mediated lateral inhibition (Baker, Mlodzik, & Rubin, 1990; Baker, Yu, & Han, 1996; Chen & Chien, 1999; Dokucu et al., 1996; Lim, Jafar-Nejad, Hsu, & Choi, 2008). R8 is responsible for sequentially triggering the differentiation of other cells in an ommatidium via Notch and EGFR signaling (Baker & Yu, 1998; Jarman et al., 1994; Tio, Ma, & Moses, 1994; Tio & Moses, 1997). Ato in R8 cells activates Rhomboid (Rho), which cleaves Spitz (Spi), a secreted EGF ligand that activates EGFR signaling in surrounding non-R8 cells (Baonza, Casci, & Freeman, 2001) (Fig. 1B). EGFR signaling triggers two non-specified cells to express Rough (Ro), a suppressor of R8 fate, and to differentiate into R2 and R5. Ro activates the expression of Rhomboid in R2 and R5, thus increasing EGFR signaling in neighboring cells (Freeman, 1996; Freeman, Kimmel, & Rubin, 1992; Kimmel, Heberlein, & Rubin, 1990; Pepple et al., 2008; Tomlinson, Kimmel, & Rubin, 1988; Yogev, Schejter, & Shilo, 2008). This higher EGFR activity specifies two more cells as R3 and R4 cells that express Seven-up (Svp) (Mlodzik, Hiromi, Weber, Goodman, & Rubin, 1990). Svp and EGFR signaling suppress R7 fate in both R3/4 and R1/6 cells (see below) (Begemann, Michon, Voorn, Wepf, & Mlodzik, 1995). Within the R3 and R4 precursors, the cell receiving a polarizing signal from the poles of the eye via Frizzled (Fz) becomes R3 and expresses Dl that specifies its neighboring precursor as an R4 cell via Notch signaling (Fanto & Mlodzik, 1999) (Fig. 1B). R8, R2, R5, R3, and R4 form a photoreceptor precluster. EGFR signaling is necessary and sufficient for maintaining G1 arrest in this precluster, while promoting mitosis in the surrounding unspecified cells, initiating the “second mitotic wave” (Baonza & Freeman, 2005; Baonza, Murawsky, Travers, & Freeman, 2002; Domı´nguez, Wasserman, & Freeman, 1998; Firth & Baker, 2005; Yang & Baker, 2003). The unspecified cells divide once more to generate enough cells to form the remaining ommatidial cells. Cells closest to the precluster receive EGF signaling and start to differentiate. The first two specified cells express Svp and Delta and differentiate to R1/R6, while the third cell is specified by Notch signaling from R1/R6 and the Bride of sevenless (Boss) local signal from R8. This cell becomes an R7 inner photoreceptor that expresses Prospero (Pros) to suppress the R8 fate and consolidate the R7 fate (Cooper & Bray, 2000; Karpilow, Pimentel, Shamloula, & Venkatesh, 1996; Miller, Lyons, & Herman, 2009; Miller, Seymour, King, & Herman, 2008; Tomlinson, Mavromatakis, & Struhl, 2011; Tomlinson & Struhl,

98

Yen-Chung Chen and Claude Desplan

2001) (Fig. 1B). Finally, cone cells and pigment cells are specified by EGFR and Notch signaling without Boss/Sev signaling that cannot reach the cells outside of the photoreceptor cluster.

3.4 The GRN that determines photoreceptor terminal features: Ommatidial subtypes Ommatidia contain two types of photoreceptors. The outer photoreceptors R1–R6 are involved in motion detection and dim light vision. They express the broad-spectrum Rhodopsin Rh1. The inner photoreceptors R7 and R8 are involved in color discrimination (Yamaguchi, Desplan, & Heisenberg, 2010; Yamaguchi, Wolf, Desplan, & Heisenberg, 2008). Ommatidia can be separated into subtypes based on the Rhodopsins expressed in their inner photoreceptors: pale and yellow ommatidia are distributed stochastically in the retina. In yellow ommatidia (65%), R7 expresses UV-sensitive Rh4, and R8 expresses green-sensitive Rh6; in pale ommatidia, R7 expresses UV-sensitive Rh3, and R8 expresses blue-sensitive Rh5. In the dorsal third of the retina, R7s in yellow ommatidia co-express the UV Rhodopsins Rh3 and Rh4, perhaps for improved UV detection for solar orientation, while R7s in pale ommatidia express normally Rh3 alone (Hardie, 1985; Mazzoni et al., 2008). In a single row of ommatidia at the dorsal rim area, both R7 and R8 express Rh3 and are specialized in detecting the vector of polarized UV light for navigation (Weir & Dickinson, 2012; Wernet et al., 2012, 2003). In outer photoreceptors, Sine oculis activates the expression of the transcription factor Glass (Gl) that links retina patterning to the morphogenesis of the rhabdomere and the expression of Rhodopsins (Bernardo-Garcia, Fritsch, & Sprecher, 2016; Moses, Ellis, & Rubin, 1989). Gl directly activates the expression of Rh1, along with Orthodenticle (Otd) and Pph13 (Ellis, O’Neill, & Rubin, 1993; Mishra et al., 2010). Otd activates Defective proventriculus (Dve) at a high level that suppresses inner photoreceptor Rhodopsins. In contrast, Spalt (Sal) is expressed in inner photoreceptors and suppresses the default expression of Rh1 and Dve, thus allowing different combinations of Rhodopsins to be expressed in yellow and pale ommatidia (Cook, Pichaud, Sonneville, Papatsenko, & Desplan, 2003; Mollereau et al., 2001) (Fig. 1C). The ratio of yellow and pale ommatidia is controlled by the stochastic activation of the transcription factor Spineless (Ss) that specifies yellow R7s ( Johnston et al., 2011; Thanawala et al., 2013; Yan et al., 2017). In yellow R7s, Ss forms a heterodimer with Tango (Tgo) to directly activate Rh4 and to activate Dve at a low level that suppresses Rh3 ( Johnston et al., 2011;

Gene regulatory networks in Drosophila visual system

99

Thanawala et al., 2013; Yan et al., 2017). In pale R7s lacking Ss, Otd and Sal activate the expression of Rh3. In the yellow R7s of the dorsal third, the expression of Iro-C genes overcomes the suppression of Rh3 by Dve while still maintaining expression of Ss at a lower level (Mazzoni et al., 2008; Thanawala et al., 2013), and as a result, dorsal third yellow R7s express both Rh3 and Rh4. Finally, in the dorsal rim area (DRA), Wingless signaling and Iro-C activate Homothorax (Hth) to suppress Ss, and consequently Rh4. As a result, DRA R7s only express Rh3 (Wernet et al., 2006, 2003). Interestingly, R8s in the DRA also express Hth and Rh3 (see below and Fig. 1D). The expression of Rhodopsins in R8 is coordinated with that of its partner R7. In a sevenless mutant that lacks R7s, most R8s express Rh6 and resemble yellow R8s, suggesting that pale R7 provides a signal that allows for the expression of Rh5 in R8s (Chou et al., 1999; Papatsenko, Sheng, & Desplan, 1997). Pale R7s activate a bistable loop in R8s consisting of two genes cross-regulating each other, Warts (Wts), a tumor suppressor kinase, and Melted (Melt), a growth regulator. Sens, a critical transcription factor for specifying R8, promotes by default the expression of Wts, which represses Melted and activates Rh6. In pR8, Activin and BMP signals from pale R7 activate Melt, which suppresses Wts and allows Otd and Yki to activate Rh5 ( Jukam et al., 2013; Jukam & Desplan, 2011; Wells, Pistillo, Barnhart, & Desplan, 2017). In the dorsal third yellow ommatidia that express both Rh3 and Rh4 in R7, R8s still express Rh6, suggesting that the Activin signal from R7 is suppressed by Ss, independently from the presence of Rh3. In the dorsal rim area, Wingless signaling and the dorsal Iro-C genes cooperate to induce the expression of Hth not only in R7 but also in R8. In these R8s, Hth suppresses Sens and consequently Rh6 while Hth, Extradenticle, Otd, and Sal act cooperatively to activate Rh3 that is therefore expressed in both R7 and R8. By expressing Rh3 in both R7 and R8, DRA ommatidia are not involved in color vision but instead compare the angle of light polarization (Wernet et al., 2003; Wernet & Desplan, 2014) (Fig. 1D).

3.5 The GRN that determines photoreceptor terminal features: Axon targeting Immediately after specification, the photoreceptors extend their axons to different layers of the optic lobe. Then, much later, they form the rhabdome, the light-gathering structure that contains the Rhodopsins that detect different spectra of light (Mollereau et al., 2001). Outer photoreceptors (R1–R6) detect motion; they express the broad-spectrum Rh1 and terminate their

100

Yen-Chung Chen and Claude Desplan

Fig. 2 Anatomy of the optic lobe and neurogenesis in the lamina. (A) Anatomy of the developing optic lobe at third larval instar (left panel) and adult optic lobe (right panel). The outer proliferation center generates lamina and medulla neurons, while the inner proliferation center generates distal cells (T2, T2a, T3, C2, C3) and lobula plate neurons (T4/T5). OPC: outer proliferation center; IPC: inner proliferation center; NE: neuroepithelium; NB: neuroblast; LN: Lamina neuron; LPC: Lamina precursor cell; LPN: lobula plate neuron. Yellow arrowheads: The direction of the wave of neurogenesis. (B) Photoreceptor axon bundles secrete signal molecules to recruit laminal precursors and trigger lamina neuron differentiation. This is mediated by signals secreted by wrapping glia. (C) Signals from photoreceptor axon bundles trigger the differentiation of lamina precursors and recruit the precursors to form lamina cartridges.

axons in the lamina, the first neuropil of the optic lobe (see Fig. 2A for the anatomy of the optic lobe). Inner photoreceptors (R7 and R8) express distinct Rhodopsins that are sensitive to different wavelengths of light and are involved in color vision ( Jukam, Lidder, & Desplan, 2008; O’Tousa et al., 1985). R7 and R8 terminate their axons in the medulla part of the optic lobe (Reviewed in Morante & Desplan, 2004). Glia-derived signals in the lamina are critical for axons from outer photoreceptors to terminate in the lamina while the axons from the inner photoreceptors extend to the medulla (Poeck, Fischer, Gunning, Zipursky, & Salecker, 2001). Several genes are

Gene regulatory networks in Drosophila visual system

101

required for proper targeting of outer photoreceptors. For example, the loss of Off-track, a receptor tyrosine kinase, in outer photoreceptors results in axons wrongly terminating in the medulla (Cafferty, Yu, & Rao, 2004). Similarly, Misshapen, a serine/threonine kinase, is expressed in photoreceptor cells under the control of Glass, interacts with the SH2/SH3 adaptor protein Dreadlocks, and is required for proper axon termination in the lamina (Ruan, Long, Vuong, & Rao, 2002; Ruan, Pang, & Rao, 1999; Treisman, Ito, & Rubin, 1997). The inner photoreceptor R8 exhibits stepwise targeting: during midpupation, two genes, Gogo and Flamingo (Fmi), are required to guide the R8 axons into the medulla. Gogo is mainly expressed in R8s and the optic lobe, while Flamingo is expressed unevenly in the growth cones of R1–R6 but not the optic lobe (Lee et al., 2003; Tomasi, Hakeda-Suzuki, Ohler, Schleiffer, & Suzuki, 2008). The chemotropic guidance molecule Netrin is expressed in the R8 target layer and is recognized by its R8-expressed receptor, Frazzled, that facilitates targeting specificity (Timofeev, Joly, Hadjieconomou, & Salecker, 2012). The fate-specifying factors could also be involved in the regulation of axon targeting. For example, in R8s, Sens cooperates with Otd to regulate the targeting subnetwork in R8s. Otd is required to activate the expression of Gogo and Fmi (Mencarelli & Pichaud, 2015), and along with Otd, Sens activates Capricious, a R8-specific cell adhesion molecule required for its targeting. In R7s, NF-YC suppresses the R8 axon targeting subnetwork (Morey et al., 2008). Although in most cases, it is not yet known how regulators of axon targeting are linked to the regulatory networks that specify the fate of the photoreceptors, a few regulators that coordinate photoreceptor axon targeting have been identified (Hoi, Xiong, & Rebay, 2016; Kniss, Holbrook, & Herman, 2013; Kulkarni, Ertekin, Lee, & Hummel, 2016; Oliva & Sierralta, 2010). For example, Brakeless, a transcriptional repressor expressed in photoreceptors, is required for proper targeting of the outer photoreceptors to the lamina by repressing Runt, which is only expressed in inner photoreceptors. Mis-expression of Brakeless or Runt in photoreceptors does not alter the expression of the fate-specifying factors like Pros or Rough, suggesting this cross-regulation happens downstream of or parallel to the fate-specifying factors. However, mis-expression of Runt in R1–R6 leads to their ectopic projection to the medulla (Kaminker, Canon, Banerjee, & Salecker, 2002; Rao, Pang, Ruan, Gunning, & Zipursky, 2000; Senti, Keleman, Eisenhaber, & Dickson, 2000).

102

Yen-Chung Chen and Claude Desplan

3.6 GRNs in the eye: A brief summary Gene regulation during fly retina development has been extensively studied and provides the likely framework that times and coordinates the specification of terminal features. The first step during embryonic retina development is Ey and Toy promoting the potency of the progenitor cells to form an eye. During the proliferation of retinal progenitors, extrinsic cues modulate the progenitor gene regulatory network, first to compartmentalize the eye disc into dorsal and ventral compartments, and later to trigger differentiation. Dpp, Notch, and Hh signaling pathways activate the retinal determination genes, Eya, So, and Dac, which initiate cell-type specification by promoting a sequential specification subnetwork regulated by Ato in the photoreceptor preclusters. During terminal differentiation, the Rhodopsin subnetwork regulated by Glass and Spalt is activated to define motionvs. color-sensing photoreceptors. Ommatidial subtypes are defined by a stochastic subnetwork regulated by Ss and by signaling by Iro-C and Wg. Less is known about the subnetwork that determines axon targeting. Some fatespecifying factors, like Sens, is shown to also play a central role in axon targeting, while it is not yet known if the regulators that distinguish innervation of lamina vs. medulla is also under the factors that secure the identity of each photoreceptor cell, like Ro, Pros, and Svp, or alternatively, if the innervation and targeting regulators are directly under the control of retinal determination genes and form their own subnetwork.

4. The GRN that specifies lamina neurons in response to photoreceptor signals Outer photoreceptors R1–R6 innervate the lamina part of the optic lobe. The neuroepithelium that gives rise to the optic lobe proliferates and segregates into two domains at the first larval instar: The Fas3-expressing inner proliferation center (IPC) that gives rise to lobula plate neurons, and the Outer Proliferation Center (OPC) that does not express Fas3 (Gold & Brand, 2014; Tayler, Robichaux, & Garrity, 2004) (Fig. 2A). The inner part of the OPC crescent will become the lamina, the part of the optic lobe first involved in motion vision, while the outer part of the crescent will give rise to the medulla (Huang & Kunes, 1996, 1998; Huang, Shilo, & Kunes, 1998). Lamina neurons develop synchronously with the MF that moves through the eye and critically depend on innervation by photoreceptors R1–R6. Axon bundles of outer photoreceptors

Gene regulatory networks in Drosophila visual system

103

extend to the optic lobe neuroepithelium and induce proliferation of their target field and differentiation into five types of lamina neurons, L1–L5. The progenitor expansion subnetwork for lamina neurogenesis includes Hth, Eya, and So, which promotes cell proliferation and suppresses a differentiation subnetwork (see below). Photoreceptor axons secrete several signaling molecules when they arrive in the developing OPC. The neuroepithelial cells in the inner OPC crescent respond to Hedgehog produced from photoreceptor axons (Huang & Kunes, 1996, 1998; Selleck & Steller, 1991) (Fig. 2C). Hedgehog signaling mediates the transition from neuroepithelium to lamina precursor cells (LPCs) by rewiring the progenitor expansion subnetwork: Eya and So activate Dac in the presence of Hh signaling, which in turn suppresses the expression of Hth. Therefore, the precursor subnetwork still promotes proliferation but also allows the differentiation subnetwork to be activated (Pin˜eiro, Lopes, & Casares, 2014). Down-regulation of Hth allows the expression of the transcription factor Single-minded (Sim), which activates the expression in LPCs of the cell adhesion molecule Hibris that interacts with Roughest in photoreceptors, thus allowing the lamina precursor cells to be recruited to photoreceptor axon bundles to assemble into lamina cartridges (Pin˜eiro et al., 2014; Sugie, Umetsu, Yasugi, Fischbach, & Tabata, 2010; Umetsu, Murakami, Sato, & Tabata, 2006). Glial precursors are recruited earlier to extending photoreceptor axons and differentiate in response to signals secreted by the photoreceptor axon bundles (Perez & Steller, 1996; Winberg, Perez, & Steller, 1992). Glia precursor cells originate from the optic stalk and migrate to the developing lamina along photoreceptor axons that produce FGF (Choi & Benzer, 1994; Chotard & Salecker, 2007). Photoreceptor axon-derived FGF is required for the differentiation of the wrapping glia that wrap the bundle of photoreceptor axons as they progress towards the lamina (see below) (Franzdo´ttir et al., 2009). Photoreceptor axons also secrete an EGF ligand (Spitz) that is specifically processed and secreted in the axons of photoreceptors and induces the differentiation of LPCs into the five types of lamina neurons, L1–L5 (Huang et al., 1998; Huang & Kunes, 1996, 1998; Yogev et al., 2008). However, this effect is indirect and is in fact mediated by wrapping glia that are the cells that receive the EGF signal from photoreceptor axons and secrete in response Insulinlike peptides as they extend down the photoreceptor axon bundle into the developing lamina. Insulin-like peptides are required for the sequential specification of lamina neuron L1–L5 (Fernandes, Chen, Rossi, Zipfel, & Desplan, 2017; Rossi & Fernandes, 2018) (Fig. 2B). Therefore, lamina

104

Yen-Chung Chen and Claude Desplan

precursor cells are produced and differentiate into distinct cell types in response to photoreceptor innervation. Supernumerary cells recruited to the cartridges are eliminated by apoptosis, and the lamina does not form at all in the absence of photoreceptors (Huang et al., 1998; Huang & Kunes, 1996, 1998). Lamina neurons then innervate different layers in the medulla. A number of transcriptional regulators and other factors are required for precise layer selection in a cell type-specific manner. N-Cadherin, Acj6, and Earmuff are examples, but how the regulation of these genes is coupled to cell type specification remains unclear (Certel, Clyne, Carlson, & Johnson, 2000; Nern, Zhu, & Zipursky, 2008; Peng et al., 2018). In short, the gene regulatory network of lamina neurogenesis responds to extrinsic cues provided by photoreceptor axon bundles. The tissue commitment gene regulatory subnetwork that determines whether the OPC neuroepithelium would give rise to LPCs or to medulla neuroblasts is still unclear. One candidate regulator is Tailless (Tll), which is broadly expressed in OPC neuroepithelium at early first larval instar and later becomes restricted to the neuroepithelium lateral to the lamina furrow to generate lamina neurons. Tll is required for lamina precursor cell specification (Guillermin, Perruchoud, Sprecher, & Egger, 2015). The progenitor expansion network responsible for promoting proliferation and suppression of premature lamina neuron differentiation includes Hth, Eya, and So, while Hh signaling from photoreceptor axons rewires the progenitor subnetwork by inducing Dac, which suppresses Hth and allows the activation of the differentiation subnetwork in response to glia-derived insulin-like peptide. The division and recruitment of LPCs to photoreceptor axons that sequentially arrive in a posterior-to-anterior order ensure retinotopy along the anterior-posterior axis. The terminal specification subnetwork that specifies the five lamina neurons L1-L5 remains elusive. There could be an intrinsic transcription factor cascade triggered by the glial derived Hh signal, and this cascade would allow different cell types to form when the LPCs receive glial-derived signal at different time windows in the cascade; alternatively, the sequential specification of lamina precursor cells could be achieved by their interaction with the extending wrapping glia or by short range signals that are provided by the neurons specified early and tune the fate potential of cells around them, resembling the sequential specification of different photoreceptor subtypes in the eye.

Gene regulatory networks in Drosophila visual system

105

5. Temporal and spatial GRNs specify medulla neurons The medulla is the largest of the optic ganglia, containing more than 40,000 neurons belonging to around 100 different neuron types with distinct morphology and innervation pattern. The medulla neuropil is subdivided into 10 layers (Fischbach, 1989; Morante & Desplan, 2008) and is made of 800 columns corresponding to the 800 ommatidia in the eye. The outer part of the OPC neuroepithelium crescent gives rise to the medulla. It proliferates under the control of a progenitor expansion subnetwork regulated by Eya, So, and Hth until the beginning of the third larval instar (Apitz & Salecker, 2016). Eya, So, and Hth are common regulators in the progenitor expansion subnetwork for neurogenesis of the lamina and medulla. A comparison of the progenitor expansion subnetworks in the retina, lamina, and medulla provides an example of how GRNs are built with a shared design but wired differently for each part of the visual system: Hth is involved in the progenitor subnetwork of the retina, lamina, and medulla and maintains progenitor proliferation, while Eya and So are only required for proliferation in the lamina and medulla but not in the retina, where they are regulators of the specification subnetwork and activate the expression of the proneural factor Ato (Apitz & Salecker, 2016; Bonini, Leiserson, & Benzer, 1993). The expansion of the OPC epithelium is also dependent on the expression of Vsx1 (Erclik et al., 2008). During its very early proliferation, the OPC neuroepithelium is compartmentalized along the dorsoventral axis by a patterning subnetwork containing Vsx, Optix, and Rx, which each define specific compartments and negatively cross-regulate each other, and by signaling molecules that subdivide all compartments (Hh) or only the Rx compartments (Wg and Dpp). Each spatial domain has the potential to generate different neuron types (Chen et al., 2016; Erclik et al., 2017; Evans et al., 2009; Gold & Brand, 2014; Kaphingst & Kunes, 1994). At the beginning of the third larval instar, a proneural wave begins at the outer edge of the OPC neuroepithelium crescent, in which neuroepithelial cells respond to the combinatorial action of EGFR, Fat/Hippo, JAK/STAT, and Notch signaling to activate the differentiation subnetwork. In the neuroepithelium, EGFR signaling induces the expression of Rhomboid, which processes the EGFR ligand Spi to reinforce EGFR signaling that activates the expression of the Notch ligand Delta (Yasugi, Sugie, Umetsu, & Tabata, 2010). Increasing Delta activates Notch signaling that works with

106

Yen-Chung Chen and Claude Desplan

the JAK/STAT and Fat/Hippo signaling pathways to slow down the proneural wave and to maintain the integrity of the optic neuroepithelium (Kawamori, Tai, Sato, Yasugi, & Tabata, 2011; Ngo et al., 2010; Reddy, Rauskolb, & Irvine, 2010; Wang, Li, Zhou, Yue, & Luo, 2011). During the differentiation into neuroblasts, ever increasing Delta suppresses Notch signaling, presumably by cis-inhibition (Egger, Gold, & Brand, 2010), allowing the expression of lethal of scute (L(1)Sc), a proneural transcription factor. As a result, the combinatorial action of extrinsic cues leads to a wave of L(1)sc that specifies the medulla neural stem cells (neuroblasts) that progressively emerge from the neuroepithelium (Egger et al., 2010; Orihara-Ono, Toriya, Nakao, & Okano, 2011; Wallace, Liu, & Vaessin, 2000; Weng, Haenfler, & Lee, 2012). Neuroblasts divide asymmetrically, re-generating a neuroblast and producing a ganglion mother cell (GMC) (Egger, Boone, Stevens, Brand, & Doe, 2007). Deadpan (Dpn) is expressed in the neuroblasts to promote self-renewal, while in the ganglion mother cell, L(1)sc, as part of the Achaete-Scute proneural complex, activates Prospero (Pros), which promotes cell cycle exit and thus only allows a single division of the GMC to produce two neurons (Choksi et al., 2006; San-Jua´n & Baonza, 2011; Yasugi, Fischer, Jiang, Reichert, & Knoblich, 2014) (Fig. 3A). A series of transcription factors are sequentially expressed in the medulla neuroblasts undergoing asymmetric division and are part of the terminal specification subnetwork. Homothorax (Hth), then Eyeless (Ey), Sloppy paired 1 (Slp1), Dichaete (D), and finally Tail-less (Tll). These temporal transcription factors (tTFs) form a temporal sequence as the dividing neuroblasts age. tTFs in this temporal cascade cross-regulate each other. Specifically, Ey, Slp1, and D are required to turn on the next tTF, while Slp1, D, and Tll are required to repress the previous tTF. Different GMCs are produced by each neuroblast at each temporal window. These GMCs undergo Notch-dependent asymmetric division, giving rise to two daughter cells with different fates. For example, the GMC emerging from neuroblasts expressing Hth give rise to a Notch-on neuron that expresses Bsh and Apterous (Ap) and becomes a Mi1 neuron, and a Notch-off cell that expresses Lim3 and Svp and becomes a Pm1, Pm2 or Pm3 neurons, depending on the location of the neuroblasts (see below) (Erclik et al., 2017; Li, Chen, & Desplan, 2013; Suzuki, Kaido, Takayama, & Sato, 2013) (Fig. 3B). In the current model, medulla neurons can be categorized into two types: Uni-columnar neurons are present as one per medulla column

Fig. 3 Neurogenesis in the medulla and lobula complex. (A) EGF signaling initiates the proneural wave in the OPC neuroepithelium. (B) Main OPC neuroblasts from the Vsx, Optix and Dpp domains undergo temporal transition of transcription factors (tTFs) and generate distinct neuron types based on the combination of the tTFs, the domain of neuroepithelium from which neuroblasts originate, and Notch signaling. (C) The domain of the neuroepithelium that expresses Wingless at the tip of OPC gives rise to tOPC neuroblasts. tOPC neuroblasts generate neural diversity with a cascade of tTFs that is similar but distinct from that of OPC neuroblasts. (D) Neuroepithelial cells delaminate from the proximal IPC (p-IPC) and migrate to the distal IPC (d-IPC) where they differentiate into neuroblasts. d-IPC neuroblasts undergo a temporal transcription factor transition and first give rise to distal cells and later to T4/T5 neurons. T4/T5 neurons are then specified into four subtypes by two consecutive Notch-dependent divisions.

108

Yen-Chung Chen and Claude Desplan

(i.e., 800 neurons of each type) and are likely produced by each neuroblast independently of its spatial position in the neuroepithelium. In contrast, multi-columnar neurons innervate multiple medulla columns and are thus less abundant. The specification of multi-columnar neurons requires the integration of the spatial compartment in which the neuroblast was born and the temporal cascade of tTFs. For example, Pm1/Pm2/Pm3 neurons are multi-columnar and originate from Hth-expressing/Notch-off GMCs. Pm3 neurons come exclusively from the Vsx1-expressing neuroblasts, while Pm1 neurons are generated from ventral Rx-expressing neuroblasts and Pm2 from dorsal Rx neuroblasts (Erclik et al., 2017). The intersection of temporal, spatial, and Notch signaling therefore generates a wide array of neuron types in the medulla. The neuroblasts at the tip of the OPC neuroepithelium (tOPC), which expresses Rx and Wingless, follow a modified tTF cascade as they sequentially express Distal-less (Dll), Ey, Slp, and D. tOPC neuroblasts at the Dll stage are type 0 and directly differentiate into a single neuron. On the other hand, GMCs from the Ey, Slp, and D window give rise to two cells which differ in Notch signaling. Notch signaling then regulates an apoptotic switch: Notch-on neurons die in the Ey window while it is the Notchoff neurons that die in the Slp and D windows (Bertet et al., 2014) (Fig. 3C). Little is known about how the terminal features like the choice of neurotransmitters or the synaptic partners are determined downstream of the specification subnetwork in the medulla, a highly heterogeneous ganglion. However, recent advances in single-cell mRNA-seq will allow gene expression to be studied even in very heterogeneous brain tissues. Indeed, transcription factors that are required for the expression of neurotransmitter-related genes were recently identified from a single-cell transcriptome atlas of the Drosophila optic lobe (Konstantinides et al., 2018). Further investigation of gene expression in the developing optic lobe are warranted to reveal the gene regulatory networks governing the terminal features like target selection or synapse specificity, which is likely to happen earlier in development (Li et al., 2017).

6. The GRNs leading to the formation of the lobula and lobula plate The inner proliferation center (IPC) neuroepithelium generates distal cells (C2, C3, T2, T2a and T3) and lobula plate neurons (T4 and T5) (Reviewed in Contreras, Sierralta, & Oliva, 2019). Anatomically, the

Gene regulatory networks in Drosophila visual system

109

IPC neuroepithelium can be subdivided into proximal, surface, and distal domains (p-IPC, s-IPC, d-IPC) (Apitz & Salecker, 2015). The tissue commitment subnetwork that distinguish IPC from OPC and the progenitor expansion subnetwork that maintains the expansion of IPC neuroepithelium have not yet been discovered, but we know that the progenitor subnetwork is distinct from the other regions because Eya and So are not expressed in p-IPC neuroepithelium (Apitz & Salecker, 2016). The p-IPC neuroepithelium is spatially compartmentalized into dorsal and ventral subdomains that express Dpp and flank a central subdomain that expresses Brinker (Brk). Another domain at the ventral tip of the IPC expresses, and is specified by Wg (Apitz & Salecker, 2018), and has a different type of neurogenesis (Filipe Pinto-Teixeira, personal communication). Dpp signaling from the dorsal and ventral subdomains of the p-IPC triggers the differentiation subnetwork in p-IPC neuroepithelial cells and stimulates their migration from the p-IPC to the d-IPC. The migration requires Escargot, a Snail-type transcription factor that mediates epithelialmesenchymal transition. p-IPC neuroepithelial cells express L(1)sc when they differentiate into migratory progenitors. Migratory progenitors than differentiate into Dpn- or Brk-expressing neuroblasts when they arrive at the lower d-IPC where they intermingle with each other to become d-IPC neuroblasts. The terminal specification subnetwork promotes d-IPC neuroblasts to progress through a series of tTFs as they divide and are pushed from the lower to the upper d-IPC by newly arriving progenitors (Apitz & Salecker, 2015). Young neuroblasts in the lower d-IPC express Ase and Dichaete and generate distal cells (C2, C3, T2, T2a and T3) that express Acj6 or Toy. As neuroblasts age, Ase and Dichaete are down-regulated, while Ato, Tll and Dac are up-regulated. Ato promotes the proliferation of the neuroblasts and later cell cycle exit by inducing the expression of Brat. The Ato-expressing neuroblasts in the upper d-IPC give rise to lobula plate neurons T4 and T5, which express Dac and Acj6 (Apitz & Salecker, 2015; Mora et al., 2018; Pinto-Teixeira et al., 2018). Upper d-IPC neuroblasts undergo two asymmetric divisions: The neuroblasts accumulate Pros asymmetrically, and when they divide, one of the daughter cell inherits cortical Pros while the other expresses Pros right after cell division, producing two Pros-expressing GMCs and then four neurons (Pinto-Teixeira et al., 2018). T4 and T5 neurons detect local motion and can each be categorized into four subtypes according to the four cardinal directions that they are tuned

110

Yen-Chung Chen and Claude Desplan

to detect (front-to-back, a subtype; back-to-front, b subtype; up, c subtype, or down, d subtype) (Maisak et al., 2013). Their diversity arises from the intersection of the spatial compartment they originated from (Dpp or Brk) and from Notch signaling: Horizontal motion detectors T4 and T5 subtypes a and b are generated exclusively from the Brk-expressing domain in the p-IPC neuroepithelium, while vertical motion detector subtypes c and d originate from the Dpp-expressing domains (Apitz & Salecker, 2018; Pinto-Teixeira et al., 2018). The first Notch-dependent asymmetric division of the upper d-IPC neuroblast decides the directionality of the local motion that the neurons generated detect: For the Brk+ p-IPC-derived neuroblast, the asymmetric division delineates subtype a (front-to-back motion-sensing neurons) vs. b (back-to-front motion-sensing neurons), while, for the Dpp+ p-IPC-derived neuroblast, the division delineates subtype c (upward motion-sensing neurons) vs. d (downward motion-sensing neurons). For the second asymmetric division of the GMC, the Notch-off daughter cell becomes a T4 neuron while the Notch-on daughter cell becomes a T5 neuron (Pinto-Teixeira et al., 2018) (Fig. 3D). Many questions about the neurogenesis in the IPC remain unanswered to date. For example, how different types of distal cells are specified is still unknown. It is likely that a similar network is modulated by extrinsic cues and contains subnetworks for neuroepithelium expansion, spatial compartmentation, and tTF transitions to generate different neuron types. T4/T5 neurons present a unique model for studying the subnetworks that define axon and dendrite targeting and synaptic specificity: Although T4/T5 neurons are transcriptionally indistinguishable in the adult (Konstantinides et al., 2018), T4 neurons terminate their dendrites in the medulla, while T5 neurons terminate theirs in the lobula; on the other hand, T4/T5 a, b, c or d subtypes differ in the lobula plate layer in which the neurons project their axons to. Recent single-cell transcriptomic studies have shown that distinct genes are differentially expressed in each T4/T5 subtype during development. These subtype-specific genes are likely to be the effectors of the subnetworks that governs layer selection during neurite targeting (Kurmangaliyev, Yoo, LoCascio, & Zipursky, 2019). Further studies are required to uncover the regulators that integrate the inputs from the two Notch-dependent division and Dpp signaling and activate the downstream subnetworks that control dendrite and axon targeting and assemble the motion detection circuit.

Gene regulatory networks in Drosophila visual system

111

7. Conclusion and perspectives Gene regulatory networks of the development of the Drosophila visual system can be subdivided into interconnected subnetworks, and the interplay of subnetworks underlies the specification of the progenitors to generate the eye or different parts of the optic ganglia, the expansion of the progenitor pool to meet the number of neurons that is required for proper function, the compartmentalization of the progenitor pool along body axes to generate different neurons, and the temporal sequence of regulators that specify different types of neurons cooperatively with spatial information from the compartmentalization. General design principles of the gene regulatory networks of the development of the Drosophila visual system could apply to neural development in different species. In some cases, not only the gene but also its function can be evolutionarily conserved across multiple species, so the regulatory network in which a Drosophila transcription factor is involved will help learn what its vertebrate orthologue controls. For example, Homothorax (Hth), the key factor of the subnetwork that expands the progenitor pool in the Drosophila eye, lamina, and medulla, has several vertebrate orthologues, the MEIS family proteins, which serve similar roles in vertebrates to promote cell proliferation in vertebrate body plan patterning (Bessa et al., 2008). Alternatively, a gene that is conserved across species could function differently, even if it is involved in the same development process. If this is the case, gene regulatory networks in Drosophila could still suggest how general developmental processes are structured mechanistically. For example, sequential specification of different cell types from neural stem cells is observed in several mammalian systems such as the retina and cortex (Chugh et al., 2014). Although it remains unknown whether the temporal transcription factors are conserved in the brain of fruit flies and mammals, the detailed network structure of temporal transcription factor cascade we learn from Drosophila could give hints on the regulators and the topology of the networks that drive the temporal sequence of specification in mammals. Additionally, a comparison of networks that serve similar purposes in Drosophila and other species could shed light on how innovations are achieved in evolution. For instance, the difference in progenitor expansion accounts for one of the major differences between the nervous system of human, non-human primates, and other mammals (Boyd et al., 2015;

112

Yen-Chung Chen and Claude Desplan

reviewed in Dehay, Kennedy, & Kosik, 2015), and the molecular identity of neural stem cells with increased proliferation potential in primates has recently been characterized (Pollen et al., 2015). By comparing the progenitor expansion subnetwork that drives proliferation in different species, we could understand the molecular basis of neocortex expansion. There are still many missing pieces in the gene regulatory network. In many cases, although the regulators for certain features have been identified, how these regulators are regulated by the fate specifying transcription factors remains elusive. For example, although we know both the regulators important for the specification of inner photoreceptors and the regulators allowing inner photoreceptors to target the medulla, how these subnetworks interact is still mostly unknown. Similarly, even if the regulators required for fate specification are identified, the targets of those regulators that define terminal features are often unknown. The spatiotemporal patterning in medulla and lobula plate is a great example in which we know regulators required for fate specification but still do not have a precise idea of what these regulators regulate to allow the neurons to acquire their terminal features. With the advent of high-throughput single-cell mRNA profiling techniques, computational approaches aiming to uncover regulatory networks systematically from profiling results are being actively developed and give promising results in filling the gaps in the networks. For example, Sens is predicted to be a direct target of Ato in one such study. If Sens were to be a bona fide target of Ato, this would explain how R8 fate is maintained after transiently expressed Ato is turned off. It also favors a cascade that is initiated by Ato and later results in the default expression of Rh6, a Rhodopsin specific to R8 photoreceptors and medulla targeting of R8 cells (Aerts et al., 2010; Davie et al., 2018; Mencarelli & Pichaud, 2015). Existing techniques that infer gene regulation rely on bulk mRNA profiling data from different conditions of samples that are assumed to be homogeneous. The number of samples is often limited, and diversity within the sample is averaged and thus undetectable. The ability of single-cell RNA-seq to distinguish individual cells allows unsupervised clustering of cells according to their states, and the scalability to capture hundreds of thousands of cells makes it easier to model complicated and non-linear relationships between genes. Scalability also enables the capture of continuous processes and the detection of regulators that are transiently

Gene regulatory networks in Drosophila visual system

113

expressed. A recent study from Konstantinides et al. used single-cell RNAseq to predict the regulators of neurotransmitter choices in neurons of the developing optic lobe and demonstrated that each of the predicted regulators is only required for the expression of neurotransmitter-related genes in a subset of neuron types. The phenotypic convergence of distinct regulators explains how core regulatory complexes specific for cell types could regulate features that are shared between multiple cell types (Konstantinides et al., 2018). The advances of experimental techniques bring not only potentials but also challenges for the study of gene regulatory networks: the tremendous number of observations from single-cell RNA-seq experiments is computationally demanding to analyze and requires more efficient algorithms. The sparseness and noisiness warrant refinements of existing analytical tools before being effectively applied. Additionally, the resulting gene regulatory networks from inference could be too complicated to interpret and test with experiments. The Drosophila visual system is an excellent model of modern gene regulatory network study. First, our understanding of regulators at different stages of development and the characterization of neural diversity and various terminal features provide a tremendous system to test newly developed analytic tools and to compare the performance of existing tools. Second, the number of subnetworks described in previous research limits the number of regulatory relationships to infer and examine in order to bridge the gaps within validated subnetworks. Finally, the powerful genetic tools available in flies allow precisely timed manipulations in specific neuron types and consequently provide the means to examine experimentally complex predicted subnetworks. Leveraging the rapidly evolving techniques and the advantages of the Drosophila visual system, we could soon be equipped to answer the long-held outstanding questions: What regulators does it take to define a fate? How is fate linked to connectivity and morphology? Are there other features that differ between neuron types?

Acknowledgments € We would like to thank Nikolaos Konstantinides, Nes¸et Ozel, Filipe Pinto-Teixeira, Anthony Rossi, Felix Simon, and Jessica Treisman for discussions and input. We also want to apologize to the researchers whose work contributed to our current understanding of the development of the Drosophila visual system but could not be included in this review. This work was supported by a grant from NIH R01 EY13010.

114

Yen-Chung Chen and Claude Desplan

References Aerts, S., Quan, X. J., Claeys, A., Sanchez, M. N., Tate, P., Yan, J., et al. (2010). Robust target gene discovery through transcriptome perturbations and genome-wide enhancer predictions in drosophila uncovers a regulatory basis for sensory specification. PLoS Biology, 8, e1000435. https://doi.org/10.1371/journal.pbio.1000435.g001. Apitz, H., & Salecker, I. (2015). A region-specific neurogenesis mode requires migratory progenitors in the Drosophila visual system. Nature Neuroscience, 18, 46–55. https:// doi.org/10.1038/nn.3896. Apitz, H., & Salecker, I. (2016). Retinal determination genes coordinate neuroepithelial specification and neurogenesis modes in the Drosophila optic lobe. Development (Cambridge, England), 143, 2431–2442. https://doi.org/10.1242/dev.135004. Apitz, H., & Salecker, I. (2018). Spatio-temporal relays control layer identity of directionselective neuron subtypes in Drosophila. Nature Communications, 9, 2295. https://doi. org/10.1038/s41467-018-04592-z. Baker, N. E., Mlodzik, M., & Rubin, G. M. (1990). Spacing differentiation in the developing Drosophila eye: A fibrinogen-related lateral inhibitor encoded by scabrous. Science, 250, 1370–1377. https://doi.org/10.1126/science.2175046. Baker, L. R., Weasner, B. M., Nagel, A., Neuman, S. D., Bashirullah, A., & Kumar, J. P. (2018). Eyeless/Pax6 initiates eye formation non-autonomously from the peripodial epithelium. Development (Cambridge, England), 145, dev163329. https://doi.org/10.1242/ dev.163329. Baker, N. E., & Yu, S. Y. (1998). The R8-photoreceptor equivalence group in Drosophila: Fate choice precedes regulated Delta transcription and is independent of Notch gene dose. Mechanisms of Development, 74, 3–14. https://doi.org/10.1016/S0925-4773(98) 00054-9. Baker, N. E., Yu, S., & Han, D. (1996). Evolution of proneural atonal expression during distinct regulatory phases in the developing Drosophila eye. Current Biology, 6, 1290–1302. https://doi.org/10.1016/s0960-9822(02)70715-x. Baonza, A., Casci, T., & Freeman, M. (2001). A primary role for the epidermal growth factor receptor in ommatidial spacing in the Drosophila eye. Current Biology, 11, 396–404. https://doi.org/10.1016/S0960-9822(01)00125-7. Baonza, A., & Freeman, M. (2001). Notch signalling and the initiation of neural development in the Drosophila eye. Development, 128, 3889–3898. Baonza, A., & Freeman, M. (2005). Control of cell proliferation in the Drosophila eye by notch signaling. Developmental Cell, 8, 529–539. https://doi.org/10.1016/j.devcel. 2005.01.019. Baonza, A., Murawsky, C. M., Travers, A. A., & Freeman, M. (2002). Pointed and Tramtrack69 establish an EGFR-dependent transcriptional switch to regulate mitosis. Nature Cell Biology, 4, 976–980. https://doi.org/10.1038/ncb887. Begemann, G., Michon, A. M., Voorn, L. v., Wepf, R., & Mlodzik, M. (1995). The Drosophila orphan nuclear receptor seven-up requires the Ras pathway for its function in photoreceptor determination. Development, 121, 225–235. Bernardo-Garcia, F. J., Fritsch, C., & Sprecher, S. G. (2016). The transcription factor glass links eye field specification with photoreceptor differentiation in Drosophila. Development (Cambridge, England), 143, 1413–1423. https://doi.org/10.1242/ dev.128801. Bertet, C., Li, X., Erclik, T., Cavey, M., Wells, B., & Desplan, C. (2014). Temporal patterning of neuroblasts controls notch-mediated cell survival through regulation of hid or reaper. Cell, 158, 1173–1186. https://doi.org/10.1016/j.cell.2014.07.045. Bessa, J., Gebelein, B., Pichaud, F., Casares, F., & Mann, R. S. (2002). Combinatorial control of Drosophila eye development by eyeless, homothorax, and teashirt. Genes & Development, 16, 2415–2427. https://doi.org/10.1101/gad.1009002.

Gene regulatory networks in Drosophila visual system

115

Bessa, J., Tavares, M. J., Santos, J., Kikuta, H., Laplante, M., Becker, T. S., et al. (2008). meis1 regulates cyclin D1 and c-myc expression, and controls the proliferation of the multipotent cells in the early developing zebrafish eye. Development, 135, 799–803. https://doi.org/10.1242/dev.011932. Boldog, E., Bakken, T. E., Hodge, R. D., Novotny, M., Aevermann, B. D., Baka, J., et al. (2018). Transcriptomic and morphophysiological evidence for a specialized human cortical GABAergic cell type. Nature Neuroscience, 21, 1185–1195. https://doi.org/10.1038/ s41593-018-0205-2. Bonini, N. M., Leiserson, W. M., & Benzer, S. (1993). The eyes absent gene: Genetic control of cell survival and differentiation in the developing Drosophila eye. Cell, 72, 379–395. https://doi.org/10.1016/0092-8674(93)90115-7. Borod, E. R., & Heberlein, U. (1998). Mutual regulation of decapentaplegic and hedgehog during the initiation of differentiation in the Drosophila retina. Developmental Biology, 197, 187–197. https://doi.org/10.1006/dbio.1998.8888. Boyd, J. L., Skove, S. L., Rouanet, J. P., Pilaz, L. J., Bepler, T., Gord^an, R., et al. (2015). Human-chimpanzee differences in a FZD8 enhancer alter cell-cycle dynamics in the developing neocortex. Current Biology, 25, 772–779. https://doi.org/10.1016/j. cub.2015.01.041. Brown, N. L., Sattler, C. A., Paddock, S. W., & Carroll, S. B. (1995). Hairy and Emc negatively regulate morphogenetic furrow progression in the drosophila eye. Cell, 80, 879–887. https://doi.org/10.1016/0092-8674(95)90291-0. Br€ uckner, K., Perez, L., Clausen, H., & Cohen, S. (2000). Glycosyltransferase activity of Fringe modulates Notch-Delta interactions. Nature, 406, 411–415. https://doi.org/ 10.1038/35019075. Cadigan, K. M., Jou, A. D., & Nusse, R. (2002). Wingless blocks bristle formation and morphogenetic furrow progression in the eye through repression of daughterless. Development, 129, 3393–3402. Cadwell, C. R., Palasantza, A., Jiang, X., Berens, P., Deng, Q., Yilmaz, M., et al. (2016). Electrophysiological, transcriptomic and morphologic profiling of single neurons using Patch-seq. Nature Biotechnology, 34, 199–203. https://doi.org/10.1038/nbt.3445. Cafferty, P., Yu, L., & Rao, Y. (2004). The receptor tyrosine kinase off-track is required for layer-specific neuronal connectivity in Drosophila. Development, 131, 5287–5295. https://doi.org/10.1242/dev.01406. Cavodeassi, F., Diez Del Corral, R., Campuzano, S., & Domı´nguez, M. (1999). Compartments and organising boundaries in the Drosophila eye: The role of the homeodomain Iroquois proteins. Development, 126, 4933–4942. Certel, S. J., Clyne, P. J., Carlson, J. R., & Johnson, W. A. (2000). Regulation of central neuron synaptic targeting by the Drosophila POU protein, Acj6. Development, 127, 2395–2405. Chen, R., Amoui, M., Zhang, Z., & Mardon, G. (1997). Dachshund and eyes absent proteins form a complex and function synergistically to induce ectopic eye development in Drosophila. Cell, 91, 893–903. https://doi.org/10.1016/S0092-8674(00)80481-X. Chen, C. K., & Chien, C. T. (1999). Negative regulation of atonal in proneural cluster formation of Drosophila R8 photoreceptors. Proceedings of the National Academy of Sciences of the United States of America, 96, 5055–5060. https://doi.org/10.1073/ pnas.96.9.5055. Chen, Z., Del Valle Rodriguez, A., Li, X., Erclik, T., Fernandes, V. M., & Desplan, C. (2016). A unique class of neural progenitors in the Drosophila optic lobe generates both migrating neurons and glia. Cell Reports, 15, 774–786. https://doi.org/10.1016/j. celrep.2016.03.061. Cho, K. O., & Choi, K. W. (1998). Fringe is essential for mirror symmetry and morphogenesis in the Drosophila eye. Nature, 396, 272–276. https://doi.org/10.1038/24394.

116

Yen-Chung Chen and Claude Desplan

Choi, K.-W., & Benzer, S. (1994). Migration of glia along photoreceptor axons in the developing drosophila eye. Neuron, 12, 423–431. https://doi.org/10.1016/0896-6273(94) 90282-8. Choksi, S. P., Southall, T. D., Bossing, T., Edoff, K., de Wit, E., Fischer, B. E. E., et al. (2006). Prospero acts as a binary switch between self-renewal and differentiation in Drosophila neural stem cells. Developmental Cell, 11, 775–789. https://doi.org/10.1016/ j.devcel.2006.09.015. Chotard, C., & Salecker, I. (2007). Glial cell development and function in the Drosophila visual system. Neuron Glia Biology, 3, 17–25. https://doi.org/10.1017/S1740925X07000592. Chou, W. H., Huber, A., Bentrop, J., Schulz, S., Schwab, K., Chadwell, L. V., et al. (1999). Patterning of the R7 and R8 photoreceptor cells of Drosophila: Evidence for induced and default cell-fate specification. Development, 126, 607–616. Chugh, K., Simons, B. D., Krieger, T. G., Kodish, O., Streicher, C., Papusheva, E., et al. (2014). Deterministic progenitor behavior and unitary production of neurons in the neocortex. Cell, 159, 775–788. https://doi.org/10.1016/j.cell.2014.10.027. Contreras, E. G., Sierralta, J., & Oliva, C. (2019). Novel strategies for the generation of neuronal diversity: Lessons from the fly visual system. Frontiers in Molecular Neuroscience, 12, 140. Cook, T., Pichaud, F., Sonneville, R., Papatsenko, D., & Desplan, C. (2003). Distinction between color photoreceptor cell fates is controlled by prospero in Drosophila. Developmental Cell, 4, 853–864. https://doi.org/10.1016/S1534-5807(03)00156-4. Cooper, M. T. D., & Bray, S. J. (2000). R7 photoreceptor specification requires Notch activity. Current Biology, 10, 1507–1510. https://doi.org/10.1016/S0960-9822(00)00826-5. Curtiss, J., & Mlodzik, M. (2000). Morphogenetic furrow initiation and progression during eye development in Drosophila: The roles of decapentaplegic, hedgehog and eyes absent. Development, 127, 1325–1336. Czerny, T., Halder, G., Kloter, U., Souabni, A., Gehring, W. J., & Busslinger, M. (1999). Twin of eyeless, a second Pax-6 gene of Drosophila, acts upstream of eyeless in the control of eye development. Molecular Cell, 3, 297–307. https://doi.org/10.1016/S10972765(00)80457-8. Davie, K., Janssens, J., Koldere, D., De Waegeneer, M., Pech, U., Kreft, Ł., et al. (2018). A single-cell transcriptome atlas of the aging Drosophila brain. Cell174, https://doi. org/10.1016/j.cell.2018.05.057. 982-998.e20. Dehay, C., Kennedy, H., & Kosik, K. S. (2015). The outer subventricular zone and primatespecific cortical complexification. Neuron, 85, 683–694. https://doi.org/10.1016/j. neuron.2014.12.060. Doe, C. Q. (2017). Temporal patterning in the Drosophila CNS. Annual Review of Cell and Developmental Biology, 33, 219–240. https://doi.org/10.1146/annurev-cellbio-111315125210. Dokucu, M. E., Zipursky, S. L., & Cagan, R. L. (1996). Atonal, rough and the resolution of proneural clusters in the developing Drosophila retina. Development, 122, 4139–4147. Domı´nguez, M., & De Celis, J. F. (1998). A dorsal/ventral boundary established by Notch controls growth and polarity in the Drosophila eye. Nature, 396, 276–278. https://doi. org/10.1038/24402. Domı´nguez, M., Wasserman, J. D., & Freeman, M. (1998). Multiple functions of the EGF receptor in Drosophila eye development. Current Biology, 8, 1039–1048. https://doi.org/ 10.1016/S0960-9822(98)70441-5. Egger, B., Boone, J. Q., Stevens, N. R., Brand, A. H., & Doe, C. Q. (2007). Regulation of spindle orientation and neural stem cell fate in the Drosophila optic lobe. Neural Development, 2, 1. https://doi.org/10.1186/1749-8104-2-1.

Gene regulatory networks in Drosophila visual system

117

Egger, B., Gold, K. S., & Brand, A. H. (2010). Notch regulates the switch from symmetric to asymmetric neural stem cell division in the Drosophila optic lobe. Development, 137, 2981–2987. https://doi.org/10.1242/dev.051250. Ellis, M. C., O’Neill, E. M., & Rubin, G. M. (1993). Expression of Drosophila glass protein and evidence for negative regulation of its activity in non-neuronal cells by another DNA-binding protein. Development, 119, 855–865. Erclik, T., Hartenstein, V., Lipshitz, H. D., & McInnes, R. R. (2008). Conserved role of the Vsx genes supports a monophyletic origin for bilaterian visual systems. Current Biology, 18, 1278–1287. https://doi.org/10.1016/j.cub.2008.07.076. Erclik, T., Li, X., Courgeon, M., Bertet, C., Chen, Z., Baumert, R., et al. (2017). Integration of temporal and spatial patterning generates neural diversity. Nature, 541, 365–370. https://doi.org/10.1038/nature20794. Evans, C. J., Olson, J. M., Ngo, K. T., Kim, E., Lee, N. E., Kuoy, E., et al. (2009). G-TRACE: Rapid Gal4-based cell lineage analysis in Drosophila. Nature Methods, 6, 603–605. https://doi.org/10.1038/nmeth.1356. Fanto, M., & Mlodzik, M. (1999). Asymmetric Notch activation specifies photoreceptors R3 and R4 and planar polarity in the Drosophila eye. Nature, 397, 523–526. https://doi.org/ 10.1038/17389. Fernandes, V. M., Chen, Z., Rossi, A. M., Zipfel, J., & Desplan, C. (2017). Glia relay differentiation cues to coordinate neuronal development in Drosophila. Science, 357, 886–891. https://doi.org/10.1126/science.aan3174. Firth, L. C., & Baker, N. E. (2005). Extracellular signals responsible for spatially regulated proliferation in the differentiating Drosophila eye. Developmental Cell, 8, 541–551. https://doi.org/10.1016/j.devcel.2005.01.017. Fischbach, D. (1989). The optic lobe of Drosophila melanogaster. I. A Golgi analysis of wild-type structure. Cell and Tissue Research, 258, 441–475. https://doi.org/10.1007/ BF00218858. Franzdo´ttir, S. R., Engelen, D., Yuva-Aydemir, Y., Schmidt, I., Aho, A., & Kl€ambt, C. (2009). Switch in FGF signalling initiates glial differentiation in the Drosophila eye. Nature, 460, 758–761. https://doi.org/10.1038/nature08167. Freeman, M. (1996). Reiterative use of the EGF receptor triggers differentiation of all cell types in the Drosophila eye. Cell, 87, 651–660. https://doi.org/10.1016/S0092-8674 (00)81385-9. Freeman, M., Kimmel, B. E., & Rubin, G. M. (1992). Identifying targets of the rough homeobox gene of Drosophila: Evidence that rhomboid functions in eye development. Development, 116, 335–346. Fuzik, J., Zeisel, A., Mate, Z., Calvigioni, D., Yanagawa, Y., Szabo, G., et al. (2016). Integration of electrophysiological recordings with single-cell RNA-seq data identifies neuronal subtypes. Nature Biotechnology, 34, 175–183. https://doi.org/10.1038/ nbt.3443. Gold, K. S., & Brand, A. H. (2014). Optix defines a neuroepithelial compartment in the optic lobe of the Drosophila brain. Neural Development, 9, https://doi.org/10.1186/17498104-9-18. Golgi, C. (1885). Sulla fina anatomia degli organi centrali del sistema nervoso. S. Calderini. Greenwood, S., & Struhl, G. (1999). Progression of the morphogenetic furrow in the Drosophila eye: The roles of hedgehog, decapentaplegic and the Raf pathway. Development, 126, 5795–5808. Guillermin, O., Perruchoud, B., Sprecher, S. G., & Egger, B. (2015). Characterization of tailless functions during Drosophila optic lobe formation. Developmental Biology, 405, 202–213. https://doi.org/10.1016/j.ydbio.2015.06.011.

118

Yen-Chung Chen and Claude Desplan

Halder, G., Callaerts, P., Flister, S., Walldorf, U., Kloter, U., & Gehring, W. J. (1998). Eyeless initiates the expression of both sine oculis and eyes absent during Drosophila compound eye development. Development, 125, 2181–2191. Halder, G., Callaerts, P., & Gehring, W. (1995). Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science, 267, 1788–1792. https://doi. org/10.1126/science.7892602. Hardie, R. C. (1985). In H. Autrum, D. Ottoson, E. R. Perl, R. F. Schmidt, H. Shimazu, & W. D. Willis (Eds.), Functional organization of the fly retina (pp. 1–79). Berlin, Heidelberg: Springer Berlin Heidelberg. https://doi.org/10.1007/978-3-642-70408-6_1. Hazelett, D. J., Bourouis, M., Walldorf, U., & Treisman, J. E. (1998). Decapentaplegic and wingless are regulated by eyes absent and eyegone and interact to direct the pattern of retinal differentiation in the eye disc. Development, 125, 3741–3751. Heberlein, U., Heberlein, U., Luk, A. Y., & Donohoe, T. J. (1995). Growth and differentiation in the Drosophila eye coordinated by hedgehog. Nature, 373, 709–711. https:// doi.org/10.1038/373709a0. Hoi, C. S. L., Xiong, W., & Rebay, I. (2016). Retinal axon guidance requires integration of eya and the Jak/Stat pathway into phosphotyrosine-based signaling circuitries in drosophila. Genetics, 203, 1283–1295. https://doi.org/10.1534/genetics.115.185918. Holguera, I., & Desplan, C. (2018). Neuronal specification in space and time. Science, 362, 176–180. https://doi.org/10.1126/science.aas9435. Huang, Z., & Kunes, S. (1996). Hedgehog, transmitted along retinal axons, triggers neurogenesis in the developing visual centers of the Drosophila brain. Cell, 86, 411–422. https://doi.org/10.1016/S0092-8674(00)80114-2. Huang, Z., & Kunes, S. (1998). Signals transmitted along retinal axons in Drosophila: Hedgehog signal reception and the cell circuitry of lamina cartridge assembly. Development, 125, 3753–3764. Huang, Z., Shilo, B. Z., & Kunes, S. (1998). A retinal axon fascicle uses Spitz, an EGF receptor ligand, to construct a synaptic cartridge in the brain of Drosophila. Cell, 95, 693–703. https://doi.org/10.1016/S0092-8674(00)81639-6. Huynh-Thu, V. A., & Sanguinetti, G. (2019). Gene regulatory network inference: An introductory survey. Methods in Molecular Biology, 1883, 1–23. https://doi.org/10.1007/9781-4939-8882-2_1. Jarman, A. P., Grell, E. H., Ackerman, L., Jan, L. Y., & Jan, Y. N. (1994). Atonal is the proneural gene for Drosophila photoreceptors. Nature, 369, 398–400. https://doi.org/ 10.1038/369398a0. Jessell, T. M. (2000). Neuronal specification in the spinal cord: Inductive signals and transcriptional codes. Nature Reviews. Genetics, 1, 20–29. https://doi.org/10.1038/35049541. Johnston, R. J., Otake, Y., Sood, P., Vogt, N., Behnia, R., Vasiliauskas, D., et al. (2011). Interlocked feedforward loops control cell-type-specific rhodopsin expression in the drosophila eye. Cell, 145, 956–968. https://doi.org/10.1016/j.cell.2011.05.003. Jukam, D., & Desplan, C. (2011). Binary regulation of hippo pathway by Merlin/NF2, Kibra, Lgl, and melted specifies and maintains postmitotic neuronal fate. Developmental Cell, 21, 874–887. https://doi.org/10.1016/j.devcel.2011.10.004. Jukam, D., Lidder, P., & Desplan, C. (2008). Rhodopsins in Drosophila color vision. In J. Tombran-Tink & C. J. Barnstable (Eds.), Visual transduction and non-visual light perception (pp. 251–266). Totowa, NJ: Humana Press. https://doi.org/10.1007/9781-59745-374-5_11. Jukam, D., Xie, B., Rister, J., Terrell, D., Charlton-Perkins, M., Pistillo, D., et al. (2013). Opposite feedbacks in the hippo pathway for growth control and neural fate. Science, 342, 1238016. https://doi.org/10.1126/science.1238016. Kaminker, J. S., Canon, J., Banerjee, U., & Salecker, I. (2002). Control of photoreceptor axon target choice by transcriptional repression of Runt. Nature Neuroscience, 5, 746–750. https://doi.org/10.1038/nn889.

Gene regulatory networks in Drosophila visual system

119

Kaphingst, K., & Kunes, S. (1994). Pattern formation in the visual centers of the Drosophila brain: Wingless acts via decapentaplegic to specify the dorsoventral axis. Cell, 78, 437–448. https://doi.org/10.1016/0092-8674(94)90422-7. Karpilow, J. M., Pimentel, A. C., Shamloula, H. K., & Venkatesh, T. R. (1996). Neuronal development in the Drosophila compound eye: Photoreceptor cells R1, R6, and R7 fail to differentiate in the retina aberrant in pattern (rap) mutant. Journal of Neurobiology, 31, 149–165. https://doi.org/10.1002/(SICI)1097-4695(199610)31:23.0.CO;2-B. Kawamori, H., Tai, M., Sato, M., Yasugi, T., & Tabata, T. (2011). Fat/hippo pathway regulates the progress of neural differentiation signaling in the Drosophila optic lobe. Development, Growth & Differentiation, 53, 653–667. https://doi.org/10.1111/j.1440169X.2011.01279.x. Kenyon, K. L., Ranade, S. S., Curtiss, J., Mlodzik, M., & Pignoni, F. (2003). Coordinating proliferation and tissue specification to promote regional identity in the Drosophila head. Developmental Cell, 5, 403–414. https://doi.org/10.1016/S15345807(03)00243-0. Kimmel, B. E., Heberlein, U., & Rubin, G. M. (1990). The homeo domain protein rough is expressed in a subset of cells in the developing Drosophila eye where it can specify photoreceptor cell subtype. Genes & Development, 4, 712–727. https://doi.org/10.1101/ gad.4.5.712. Kniss, J. S., Holbrook, S., & Herman, T. G. (2013). R7 photoreceptor axon growth is temporally controlled by the transcription factor Ttk69, which inhibits growth in part by promoting transforming growth factor-β/activin signaling. The Journal of Neuroscience, 33, 1509–1520. https://doi.org/10.1523/JNEUROSCI.2023-12.2013. Kolodkin, A. L., Matthes, D. J., & Goodman, C. S. (1993). The semaphorin genes encode a family of transmembrane and secreted growth cone guidance molecules. Cell, 75, 1389–1399. https://doi.org/10.1016/0092-8674(93)90625-Z. Kolodziej, P. A., Timpe, L. C., Mitchell, K. J., Fried, S. R., Goodman, C. S., Jan, L. Y., et al. (1996). Frazzled encodes a Drosophila member of the DCC immunoglobulin subfamily and is required for CNS and motor axon guidance. Cell, 87, 197–204. https://doi.org/ 10.1016/S0092-8674(00)81338-0. Konstantinides, N., Kapuralin, K., Fadil, C., Barboza, L., Satija, R., & Desplan, C. (2018). Phenotypic convergence: Distinct transcription factors regulate common terminal features. Cell, 174, 622–635.e13. https://doi.org/10.1016/j.cell.2018.05.021. Kulkarni, A., Ertekin, D., Lee, C. H., & Hummel, T. (2016). Birth order dependent growth cone segregation determines synaptic layer identity in the Drosophila visual system. eLife, 5, e13715. https://doi.org/10.7554/eLife.13715. Kumar, J. P., & Moses, K. (2001). The EGF receptor and notch signaling pathways control the initiation of the morphogenetic furrow during Drosophila eye development. Development, 128, 2689–2697. Kurmangaliyev, Y. Z., Yoo, J., LoCascio, S. A., & Zipursky, S. L. (2019). Modular transcriptional programs separately define axon and dendrite connectivity. eLife, 8, e50822. https://doi.org/10.7554/eLife.50822. Lee, R. C., Clandinin, T. R., Lee, C. H., Chen, P. L., Meinertzhagen, I. A., & Zipursky, S. L. (2003). The protocadherin flamingo is required for axon target selection in the Drosophila visual system. Nature Neuroscience, 6, 557–563. https://doi.org/ 10.1038/nn1063. Li, X., Chen, Z., & Desplan, C. (2013). Temporal patterning of neural progenitors in drosophila. Current Topics in Developmental Biology, 105, 69–96. https://doi.org/10.1016/ B978-0-12-396968-2.00003-8. Li, X., Erclik, T., Bertet, C., Chen, Z., Voutev, R., Venkatesh, S., et al. (2013). Temporal patterning of Drosophila medulla neuroblasts controls neural fates. Nature, 498, 456–462. https://doi.org/10.1038/nature12319.

120

Yen-Chung Chen and Claude Desplan

Li, H., Horns, F., Xie, Q., Xie, Q., Li, T., Luginbuhl, D. J., et al. (2017). Classifying Drosophila olfactory projection neuron subtypes by single-cell RNA sequencing. Cell, 171, 1206.e22-1207. https://doi.org/10.1016/j.cell.2017.10.019. Lim, J., Jafar-Nejad, H., Hsu, Y. C., & Choi, K. W. (2008). Novel function of the class I bHLH protein daughterless in the negative regulation of proneural gene expression in the Drosophila eye. EMBO Reports, 9, 1128–1133. https://doi.org/10.1038/ embor.2008.166. Lopes, C. S., & Casares, F. (2010). Hth maintains the pool of eye progenitors and its downregulation by Dpp and Hh couples retinal fate acquisition with cell cycle exit. Developmental Biology, 339, 78–88. https://doi.org/10.1016/j.ydbio.2009.12.020. Maisak, M. S., Haag, J., Ammer, G., Serbe, E., Meier, M., Leonhardt, A., et al. (2013). A directional tuning map of Drosophila elementary motion detectors. Nature, 500, 212–216. https://doi.org/10.1038/nature12320. Maurel-Zaffran, C., & Treisman, J. E. (2000). Pannier acts upstream of wingless to direct dorsal eye disc development in Drosophila. Development, 127, 1007–1016. Mazzoni, E. O., Celik, A., Wernet, M. F., Vasiliauskas, D., Johnston, R. J., Cook, T. A., et al. (2008). Iroquois complex genes induce co-expression of rhodopsins in Drosophila. PLoS Biology, 6, 825–835. https://doi.org/10.1371/journal.pbio.0060097. Mencarelli, C., & Pichaud, F. (2015). Orthodenticle is required for the expression of principal recognition molecules that control axon targeting in the Drosophila retina. PLoS Genetics, 11, e1005303. https://doi.org/10.1371/journal.pgen.1005303. Messersmith, E. K., Leonardo, E. D., Shatz, C. J., Tessier-Lavigne, M., Goodman, C. S., & Kolodkin, A. L. (1995). Sernaphorin III can function as a selective chemorepellent to pattern sensory projections in the spinal cord. Neuron, 14, 949–959. https://doi.org/ 10.1016/0896-6273(95)90333-X. Miller, A. C., Lyons, E. L., & Herman, T. G. (2009). Cis-inhibition of Notch by endogenous Delta biases the outcome of lateral inhibition. Current Biology, 19, 1378–1383. https:// doi.org/10.1016/j.cub.2009.06.042. Miller, A. C., Seymour, H., King, C., & Herman, T. G. (2008). Loss of seven-up from Drosophila R1/R6 photoreceptors reveals a stochastic fate choice that is normally biased by Notch. Development, 135, 707–715. https://doi.org/10.1242/dev.016386. Mishra, M., Oke, A., Lebel, C., McDonald, E. C., Plummer, Z., Cook, T. A., et al. (2010). Pph13 and orthodenticle define a dual regulatory pathway for photoreceptor cell morphogenesis and function. Development, 137, 2895–2904. https://doi.org/10.1242/ dev.051722. Mitchell, K. J., Doyle, J. L., Serafini, T., Kennedy, T. E., Tessier-Lavigne, M., Goodman, C. S., et al. (1996). Genetic analysis of Netrin genes in Drosophila: Netrins guide CNS commissural axons and peripheral motor axons. Neuron, 17, 203–215. https://doi.org/10.1016/S0896-6273(00)80153-1. Mlodzik, M., Hiromi, Y., Weber, U., Goodman, C. S., & Rubin, G. M. (1990). The drosophila seven-up gene, a member of the steroid receptor gene superfamily, controls photoreceptor cell fates. Cell, 60, 211–224. https://doi.org/10.1016/0092-8674(90) 90737-Y. Mollereau, B., Dominguez, M., Webel, R., Colley, N. J., Keung, B., De Celis, J. F., et al. (2001). Two-step process for photoreceptor formation in drosophila. Nature, 412, 911–913. https://doi.org/10.1038/35091076. Mora, N., Oliva, C., Fiers, M., Ejsmont, R., Soldano, A., Zhang, T. T., et al. (2018). A temporal transcriptional switch governs stem cell division, neuronal numbers, and maintenance of differentiation. Developmental Cell, 45, https://doi.org/10.1016/j. devcel.2018.02.023. 53-66.e5. Morante, J., & Desplan, C. (2004). Building a projection map for photoreceptor neurons in the Drosophila optic lobes. Seminars in Cell & Developmental Biology, 15, 137–143. https://doi.org/10.1016/j.semcdb.2003.09.007.

Gene regulatory networks in Drosophila visual system

121

Morante, J., & Desplan, C. (2008). The color-vision circuit in the medulla of Drosophila. Current Biology, 18, 553–565. https://doi.org/10.1016/j.cub.2008.02.075. Morey, M., Yee, S. K., Herman, T., Nern, A., Blanco, E., & Zipursky, S. L. (2008). Coordinate control of synaptic-layer specificity and rhodopsins in photoreceptor neurons. Nature, 456, 795–799. https://doi.org/10.1038/nature07419. Moses, K., Ellis, M. C., & Rubin, G. M. (1989). The glass gene encodes a zinc-finger protein required by Drosophila photoreceptor cells. Nature, 340, 531–536. https://doi.org/ 10.1038/340531a0. Nern, A., Zhu, Y., & Zipursky, S. L. (2008). Local N-cadherin interactions mediate distinct steps in the targeting of lamina neurons. Neuron, 58, 34–41. https://doi.org/10.1016/j. neuron.2008.03.022. Ngo, K. T., Wang, J., Junker, M., Kriz, S., Vo, G., Asem, B., et al. (2010). Concomitant requirement for Notch and Jak/Stat signaling during neuro-epithelial differentiation in the Drosophila optic lobe. Developmental Biology, 346, 284–295. https://doi.org/ 10.1016/j.ydbio.2010.07.036. Niimi, T., Seimiya, M., Kloter, U., Flister, S., & Gehring, W. J. (1999). Direct regulatory interaction of the eyeless protein with an eye-specific enhancer in the sine oculis gene during eye induction in Drosophila. Development, 126, 2253–2260. Oliva, C., & Sierralta, J. (2010). Regulation of axonal development by the nuclear protein hindsight (pebbled) in the Drosophila visual system. Developmental Biology, 344, 911–921. https://doi.org/10.1016/j.ydbio.2010.06.007. Orihara-Ono, M., Toriya, M., Nakao, K., & Okano, H. (2011). Downregulation of Notch mediates the seamless transition of individual Drosophila neuroepithelial progenitors into optic medullar neuroblasts during prolonged G1. Developmental Biology, 351, 163–175. https://doi.org/10.1016/j.ydbio.2010.12.044. Ostrin, E. J., Li, Y., Hoffman, K., Liu, J., Wang, K., Zhang, L., et al. (2006). Genome-wide identification of direct targets of the Drosophila retinal determination protein eyeless. Genome Research, 16, 466–476. https://doi.org/10.1101/gr.4673006. O’Tousa, J. E., Baehr, W., Martin, R. L., Hirsh, J., Pak, W. L., & Applebury, M. L. (1985). The Drosophila ninaE gene encodes an opsin. Cell, 40, 839–850. https://doi.org/ 10.1016/0092-8674(85)90343-5. Panin, V. M., Papayannopoulos, V., Wilson, R., & Irvine, K. D. (1997). Fringe modulates Notch-ligand interactions. Nature, 387, 908–912. https://doi.org/10.1038/43191. Papatsenko, D., Sheng, G., & Desplan, C. (1997). A new rhodopsin in R8 photoreceptors of Drosophila: Evidence for coordinate expression with Rh3 in R7 cells. Development, 124, 1665–1673. Papayannopoulos, V., Tomlinson, A., Panin, V. M., Rauskolb, C., & Irvine, K. D. (1998). Dorsal-ventral signaling in the Drosophila eye. Science, 281, 2031–2034. https://doi.org/ 10.1126/science.281.5385.2031. Pappu, K. S., Ostrin, E. J., Middlebrooks, B. W., Sili, B. T., Chen, R., Atkins, M. R., et al. (2005). Dual regulation and redundant function of two eye-specific enhancers of the Drosophila retinal determination gene dachshund. Development, 132, 2895–2905. https://doi.org/10.1242/dev.01869. Peng, J., Santiago, I. J., Ahn, C., Gur, B., Tsui, C. K., Su, Z., et al. (2018). Drosophila fezf coordinates laminar-specific connectivity through cell-intrinsic and cell-extrinsic mechanisms. eLife, 7, e33962. https://doi.org/10.7554/eLife.33962. Pepple, K. L., Atkins, M., Venken, K., Wellnitz, K., Harding, M., Frankfort, B., et al. (2008). Two-step selection of a single R8 photoreceptor: A bistable loop between senseless and rough locks in R8 fate. Development, 135, 4071–4079. https://doi.org/10.1242/ dev.028951. Perez, S. E., & Steller, H. (1996). Migration of glial cells into retinal axon target field in Drosophila melanogaster. Journal of Neurobiology, 30, 359–373. https://doi.org/ 10.1002/(SICI)1097-4695(199607)30:33.0.CO;2-3.

122

Yen-Chung Chen and Claude Desplan

Pignoni, F., Hu, B., Zavitz, K. H., Xiao, J., Garrity, P. A., & Zipursky, S. L. (1997). The eye-specification proteins So and Eya form a complex and regulate multiple steps in Drosophila eye development. Cell, 91, 881–891. https://doi.org/10.1016/S00928674(00)80480-8. Pin˜eiro, C., Lopes, C. S., & Casares, F. (2014). A conserved transcriptional network regulates lamina development in the Drosophila visual system. Development (Cambridge, England), 141, 2838–2847. https://doi.org/10.1242/dev.108670. Pinto-Teixeira, F., Koo, C., Rossi, A. M., Neriec, N., Bertet, C., Li, X., et al. (2018). Development of concurrent retinotopic maps in the fly motion detection circuit. Cell, 173, 485-498.e11. https://doi.org/10.1016/j.cell.2018.02.053. Poeck, B., Fischer, S., Gunning, D., Zipursky, S. L., & Salecker, I. (2001). Glial cells mediate target layer selection of retinal axons in the developing visual system of Drosophila. Neuron, 29, 99–113. https://doi.org/10.1016/s0896-6273(01)00183-0. Pollen, A. A., Nowakowski, T. J., Chen, J., Retallack, H., Sandoval-Espinosa, C., Nicholas, C. R., et al. (2015). Molecular identity of human outer radial glia during cortical development. Cell, 163, 55–67. https://doi.org/10.1016/j.cell.2015.09.004. Raghu, S. V., & Borst, A. (2011). Candidate glutamatergic neurons in the visual system of drosophila. PLoS One, 6, e19472. https://doi.org/10.1371/journal.pone.0019472. Raghu, S. V., Claussen, J., & Borst, A. (2013). Neurons with GABAergic phenotype in the visual system of Drosophila. The Journal of Comparative Neurology, 521, 252–265. https:// doi.org/10.1002/cne.23208. Ramo´n y Cajal, S. (1911). Histologie du syste`me nerveux de l’homme & des vertebres. In E. Franc¸a (Ed.), Histologie du syste`me nerveux de l’homme & des vertebres. Paris: Maloine. https://doi.org/10.5962/bhl.title.48637. Rao, Y., Pang, P., Ruan, W., Gunning, D., & Zipursky, S. L. (2000). Brakeless is required for photoreceptor growth-cone targeting in Drosophila. Proceedings of the National Academy of Sciences of the United States of America, 97, 5966–5971. https://doi.org/ 10.1073/pnas.110135297. Reddy, B. V. V. G., Rauskolb, C., & Irvine, K. D. (2010). Influence of Fat-Hippo and Notch signaling on the proliferation and differentiation of Drosophila optic neuroepithelia. Development, 137, 2397–2408. https://doi.org/10.1242/dev.050013. Rivera-Alba, M., Vitaladevuni, S. N., Mischenko, Y., Lu, Z., Takemura, S. Y., Scheffer, L., et al. (2011). Wiring economy and volume exclusion determine neuronal placement in the Drosophila brain. Current Biology, 21, 2000–2005. https://doi.org/10.1016/j. cub.2011.10.022. Roignant, J.-Y., & Treisman, J. E. (2009). Pattern formation in the Drosophila eye disc. The International Journal of Developmental Biology, 53, 795–804. https://doi.org/10.1387/ ijdb.072483jr. Rossi, A. M., & Fernandes, V. M. (2018). Wrapping glial morphogenesis and signaling control the timing and pattern of neuronal differentiation in the Drosophila lamina. Journal of Experimental Neuroscience, 12, 1179069518759294. https://doi.org/10.1177/ 1179069518759294. Ruan, W., Long, H., Vuong, D. H., & Rao, Y. (2002). Bifocal is a downstream target of the Ste20-like serine/threonine kinase misshapen in regulating photoreceptor growth cone targeting in Drosophila. Neuron, 36, 831–842. https://doi.org/10.1016/S08966273(02)01027-9. Ruan, W., Pang, P., & Rao, Y. (1999). The SH2/SH3 adaptor protein dock interacts with the Ste20-like kinase misshapen in controlling growth cone motility. Neuron, 24, 595–605. https://doi.org/10.1016/S0896-6273(00)81115-0. San-Jua´n, B. P., & Baonza, A. (2011). The bHLH factor deadpan is a direct target of Notch signaling and regulates neuroblast self-renewal in Drosophila. Developmental Biology, 352, 70–82. https://doi.org/10.1016/j.ydbio.2011.01.019.

Gene regulatory networks in Drosophila visual system

123

Seeger, M., Tear, G., Ferres-Marco, D., & Goodman, C. S. (1993). Mutations affecting growth cone guidance in drosophila: Genes necessary for guidance toward or away from the midline. Neuron, 10, 409–426. https://doi.org/10.1016/0896-6273(93)90330-T. Selleck, S. B., & Steller, H. (1991). The influence of retinal innervation on neurogenesis in the first optic ganglion of drosophila. Neuron, 6, 83–99. https://doi.org/10.1016/08966273(91)90124-I. Senti, K. A., Keleman, K., Eisenhaber, F., & Dickson, B. J. (2000). Brakeless is required for lamina targeting of R1-R6 axons in the Drosophila visual system. Development, 127, 2291–2301. Shinomiya, K., Huang, G., Lu, Z., Parag, T., Xu, C. S., Aniceto, R., et al. (2019). Comparisons between the ON- and OFF-edge motion pathways in the Drosophila brain. eLife, 8, e40025. https://doi.org/10.7554/eLife.40025. Shirasaki, R., & Pfaff, S. L. (2002). Transcriptional codes and the control of neuronal identity. Annual Review of Neuroscience, 25, 251–281. https://doi.org/10.1146/annurev.neuro. 25.112701.142916. Sugie, A., Umetsu, D., Yasugi, T., Fischbach, K. F., & Tabata, T. (2010). Recognition of pre- and postsynaptic neurons via nephrin/NEPH1 homologs is a basis for the formation of the Drosophila retinotopic map. Development, 137, 3303–3313. https://doi.org/ 10.1242/dev.047332. Suzuki, T., Kaido, M., Takayama, R., & Sato, M. (2013). A temporal mechanism that produces neuronal diversity in the Drosophila visual center. Developmental Biology, 380, 12–24. https://doi.org/10.1016/j.ydbio.2013.05.002. Suzuki, T., Trush, O., Yasugi, T., Takayama, R., & Sato, M. (2016). Wnt signaling specifies anteroposterior progenitor zone identity in the Drosophila visual center. The Journal of Neuroscience, 36, 6503–6513. https://doi.org/10.1523/JNEUROSCI.0864-16.2016. Takemura, S. Y., Bharioke, A., Lu, Z., Nern, A., Vitaladevuni, S., Rivlin, P. K., et al. (2013). A visual motion detection circuit suggested by Drosophila connectomics. Nature, 500, 175–181. https://doi.org/10.1038/nature12450. Takemura, S. y., Nern, A., Chklovskii, D. B., Scheffer, L. K., Rubin, G. M., & Meinertzhagen, I. A. (2017). The comprehensive connectome of a neural substrate for ‘ON’ motion detection in Drosophila. eLife, 6, e24394. https://doi.org/10.7554/ eLife.24394. Takemura, S. Y., Xu, C. S., Lu, Z., Rivlin, P. K., Parag, T., Olbris, D. J., et al. (2015). Synaptic circuits and their variations within different columns in the visual system of Drosophila. Proceedings of the National Academy of Sciences of the United States of America, 112, 13711–13716. https://doi.org/10.1073/pnas.1509820112. Tanaka-Matakatsu, M., & Du, W. (2008). Direct control of the proneural gene atonal by retinal determination factors during Drosophila eye development. Developmental Biology, 313, 787–801. https://doi.org/10.1016/j.ydbio.2007.11.017. Tayler, T. D., Robichaux, M. B., & Garrity, P. A. (2004). Compartmentalization of visual centers in the Drosophila brain requires Slit and Robo proteins. Development, 131, 5935–5945. https://doi.org/10.1242/dev.01465. Thanawala, S. U., Rister, J., Goldberg, G. W., Zuskov, A., Olesnicky, E. C., Flowers, J. M., et al. (2013). Regional modulation of a stochastically expressed factor determines photoreceptor subtypes in the drosophila retina. Developmental Cell, 25, 93–105. https://doi. org/10.1016/j.devcel.2013.02.016. Timofeev, K., Joly, W., Hadjieconomou, D., & Salecker, I. (2012). Localized netrins act as positional cues to control layer-specific targeting of photoreceptor axons in drosophila. Neuron, 75, 80–93. https://doi.org/10.1016/j.neuron.2012.04.037. Tio, M., Ma, C., & Moses, K. (1994). Spitz, a Drosophila homolog of transforming growth factor-α, is required in the founding photoreceptor cells of the compound eye facets. Mechanisms of Development, 48, 13–23. https://doi.org/10.1016/0925-4773(94)90002-7.

124

Yen-Chung Chen and Claude Desplan

Tio, M., & Moses, K. (1997). The Drosophila TGFα homolog Spitz acts in photoreceptor recruitment in the developing retina. Development, 124, 343–351. Tomasi, T., Hakeda-Suzuki, S., Ohler, S., Schleiffer, A., & Suzuki, T. (2008). The transmembrane protein Golden goal regulates R8 photoreceptor axon-axon and axon-target interactions. Neuron, 57, 691–704. https://doi.org/10.1016/j.neuron.2008.01.012. Tomlinson, A., Kimmel, B. E., & Rubin, G. M. (1988). Rough, a Drosophila homeobox gene required in photoreceptors R2 and R5 for inductive interactions in the developing eye. Cell, 55, 771–784. https://doi.org/10.1016/0092-8674(88)90133-X. Tomlinson, A., Mavromatakis, Y. E., & Struhl, G. (2011). Three distinct roles for Notch in Drosophila R7 photoreceptor specification. PLoS Biology, 9, e1001132. https://doi.org/ 10.1371/journal.pbio.1001132. Tomlinson, A., & Struhl, G. (2001). Delta/Notch and Boss/Sevenless signals act combinatorially to specify the Drosophila R7 photoreceptor. Molecular Cell, 7, 487–495. https:// doi.org/10.1016/S1097-2765(01)00196-4. Treisman, J. E. (2013). Retinal differentiation in Drosophila. Wiley Interdisciplinary Reviews: Developmental Biology, 2, 545–557. https://doi.org/10.1002/wdev.100. Treisman, J. E., Ito, N., & Rubin, G. M. (1997). Misshapen encodes a protein kinase involved in cell shape control in Drosophila. Gene, 186, 119–125. https://doi.org/ 10.1016/S0378-1119(96)00694-4. Umetsu, D., Murakami, S., Sato, M., & Tabata, T. (2006). The highly ordered assembly of retinal axons and their synaptic partners is regulated by hedgehog/single-minded in the Drosophila visual system. Development, 133, 791–800. https://doi.org/10.1242/ dev.02253. Varija Raghu, S., Reiff, D. F., & Borst, A. (2011). Neurons with cholinergic phenotype in the visual system of Drosophila. The Journal of Comparative Neurology, 519, 162–176. https://doi.org/10.1002/cne.22512. Wallace, K., Liu, T.-H., & Vaessin, H. (2000). The pan-neural bHLH proteins DEADPAN and ASENSE regulate mitotic activity and cdk inhibitor dacapo expression in the Drosophila larval optic lobes. Genesis, 26, 77–85. https://doi.org/10.1002/ (SICI)1526-968X(200001)26:13.0.CO;2-R. Wang, W., Li, Y., Zhou, L., Yue, H., & Luo, H. (2011). Role of JAK/STAT signaling in neuroepithelial stem cell maintenance and proliferation in the Drosophila optic lobe. Biochemical and Biophysical Research Communications, 410, 714–720. https://doi.org/ 10.1016/j.bbrc.2011.05.119. Weir, P. T., & Dickinson, M. H. (2012). Flying Drosophila orient to sky polarization. Current Biology, 22, 21–27. https://doi.org/10.1016/j.cub.2011.11.026. Wells, B. S., Pistillo, D., Barnhart, E., & Desplan, C. (2017). Parallel Activin and BMP signaling coordinates R7/R8 photoreceptor subtype pairing in the stochastic Drosophila retina. eLife, 6, e25301. https://doi.org/10.7554/eLife.25301. Weng, M., Haenfler, J. M., & Lee, C. Y. (2012). Changes in Notch signaling coordinates maintenance and differentiation of the Drosophila larval optic lobe neuroepithelia. Developmental Neurobiology, 72, 1376–1390. https://doi.org/10.1002/dneu.20995. Wernet, M. F., & Desplan, C. (2014). Homothorax and Extradenticle alter the transcription factor network in Drosophila ommatidia at the dorsal rim of the retina. Development (Cambridge, England), 141, 918–928. https://doi.org/10.1242/dev.103127. Wernet, M. F., Labhart, T., Baumann, F., Mazzoni, E. O., Pichaud, F., & Desplan, C. (2003). Homothorax switches function of Drosophila photoreceptors from color to polarized light sensors. Cell, 115, 267–279. https://doi.org/10.1016/S0092-8674(03) 00848-1. ¸ elik, A., Duncan, D. M., Duncan, I., & Desplan, C. (2006). Wernet, M. F., Mazzoni, E. O., C Stochastic spineless expression creates the retinal mosaic for colour vision. Nature, 440, 174–180. https://doi.org/10.1038/nature04615.

Gene regulatory networks in Drosophila visual system

125

Wernet, M. F., Velez, M. M., Clark, D. A., Baumann-Klausener, F., Brown, J. R., Klovstad, M., et al. (2012). Genetic dissection reveals two separate retinal substrates for polarization vision in Drosophila. Current Biology, 22, 12–20. https://doi.org/ 10.1016/j.cub.2011.11.028. Winberg, M. L., Perez, S. E., & Steller, H. (1992). Generation and early differentiation of glial cells in the first optic ganglion of Drosophila melanogaster. Development, 115, 903–911. Yamaguchi, S., Desplan, C., & Heisenberg, M. (2010). Contribution of photoreceptor subtypes to spectral wavelength preference in Drosophila. Proceedings of the National Academy of Sciences of the United States of America, 107, 5634–5639. https://doi.org/10.1073/ pnas.0809398107. Yamaguchi, S., Wolf, R., Desplan, C., & Heisenberg, M. (2008). Motion vision is independent of color in Drosophila. Proceedings of the National Academy of Sciences of the United States of America, 105, 4910–4915. https://doi.org/10.1073/pnas.0711484105. Yan, J., Anderson, C., Viets, K., Tran, S., Goldberg, G., Small, S., et al. (2017). Regulatory logic driving stable levels of defective proventriculus expression during terminal photoreceptor specification in flies. Development, 144, 844–855. https://doi.org/10.1242/ dev.144030. Yang, L., & Baker, N. E. (2003). Cell cycle withdrawal, progression, and cell survival regulation by EGFR and its effectors in the differentiating Drosophila eye. Developmental Cell, 4, 359–369. https://doi.org/10.1016/S1534-5807(03)00059-5. Yasugi, T., Fischer, A., Jiang, Y., Reichert, H., & Knoblich, J. A. (2014). A regulatory transcriptional loop controls proliferation and differentiation in Drosophila neural stem cells. PLoS One, 9, e97034. https://doi.org/10.1371/journal.pone.0097034. Yasugi, T., Sugie, A., Umetsu, D., & Tabata, T. (2010). Coordinated sequential action of EGFR and Notch signaling pathways regulates proneural wave progression in the Drosophila optic lobe. Development, 137, 3193–3203. https://doi.org/10.1242/ dev.048058. Yogev, S., Schejter, E. D., & Shilo, B.-Z. (2008). Drosophila EGFR signalling is modulated by differential compartmentalization of rhomboid intramembrane proteases. The EMBO Journal, 27, 1219–1230. https://doi.org/10.1038/emboj.2008.58. Zhu, J., Palliyil, S., Ran, C., & Kumar, J. P. (2017). Drosophila Pax6 promotes development of the entire eye-antennal disc, thereby ensuring proper adult head formation. Proceedings of the National Academy of Sciences of the United States of America, 114, 5846–5853. https:// doi.org/10.1073/pnas.1610614114.

CHAPTER FIVE

Cell fate decisions during the development of the peripheral nervous system in the vertebrate head Alexandre Thiery, Ailin Leticia Buzzi, Andrea Streit∗ Centre for Craniofacial and Regenerative Biology, Faculty of Dentistry, Craniofacial and Oral Sciences, King’s College London, London, United Kingdom ∗ Corresponding author: e-mail address: [email protected]

Contents 1. Introduction 2. The preplacodal region: A territory of sense organ and cranial ganglia progenitors 3. Subdivision of the embryonic ectoderm prior to gastrulation 4. The neural plate border 5. Signaling events at the neural plate border 6. Transcription factor interactions: From neural plate border to neural crest and placode precursors 7. Segregation of neural crest and sensory placode precursors 8. Subdividing the placode territory along the anterior-posterior axis 9. Otic and epibranchial placode formation: Signals and regulatory circuits 9.1 FGF signaling induces otic-epibranchial progenitors 9.2 Segregation of otic-epibranchial progenitors: Signals and regulatory circuits 9.3 Regulatory circuits at otic and epibranchial placode stages 10. Refining the OEP network: A cis-regulatory perspective 11. Conclusion References

128 130 132 134 139 141 144 145 147 147 151 152 153 155 156

Abstract Sensory placodes and neural crest cells are among the key cell populations that facilitated the emergence and diversification of vertebrates throughout evolution. Together, they generate the sensory nervous system in the head: both form the cranial sensory ganglia, while placodal cells make major contributions to the sense organs—the eye, ear and olfactory epithelium. Both are instrumental for integrating craniofacial organs and have been key to drive the concentration of sensory structures in the vertebrate head allowing the emergence of active and predatory life forms. Whereas the gene Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.04.002

#

2020 Elsevier Inc. All rights reserved.

127

128

Alexandre Thiery et al.

regulatory networks that control neural crest cell development have been studied extensively, the signals and downstream transcriptional events that regulate placode formation and diversity are only beginning to be uncovered. Both cell populations are derived from the embryonic ectoderm, which also generates the central nervous system and the epidermis, and recent evidence suggests that their initial specification involves a common molecular mechanism before definitive neural, neural crest and placodal lineages are established. In this review, we will first discuss the transcriptional networks that pattern the embryonic ectoderm and establish these three cell fates with emphasis on sensory placodes. Second, we will focus on how sensory placode precursors diversify using the specification of otic-epibranchial progenitors and their segregation as an example.

1. Introduction Sensory placodes are epithelial patches that form in the surface ectoderm next to the neural tube. In amniotes, they become visible as patches of columnar epithelium from the 10-somite stage onwards, and subsequently undergo complex morphogenetic events to form the olfactory epithelium, the adenohypophysis, the lens or the inner ear, or remain as neurogenic patches that form the cranial sensory ganglia (Fig. 1A and B) (for review: Baker & Bronner-Fraser, 2001; Grocott, Tambalo, & Streit, 2012; Patthey, Schlosser, & Shimeld, 2014; Schlosser, 2010). Two placodes are nonneurogenic: the adenohypophysis and the lens. While the former

Fig. 1 Placodes and their derivatives. (A) Placode derivatives. Diagram showing a side view of a 20–26-somite stage chick embryo with placode derivatives color coded; for details see text. (B) Sensory placodes. Diagram showing a 10-somite stage embryo. Placodes (color-coded as in A) have formed thickened, columnar epithelia located in the surface ectoderm along the neural tube. (C) At head process stages, sensory progenitors occupy the nonneural ectoderm surrounding the neural plate, the preplacodal region. In the head, neural crest cell precursors are sandwiched between the neural plate and posterior preplacodal region; no neural crest cells form in the most anterior ectoderm. Mesodermal cells provide inducing signals for sensory progenitors.

Cell fate decisions in the cranial sensory nervous system

129

develops at the midline and gives rise to the anterior pituitary producing neuroendocrine cells, the latter generates lens fiber and lens epithelial cells and becomes incorporated into the eye. The ophthalmic and maxillomandibular trigeminal placodes (profundal and trigeminal in anamniotes) and the epibranchial placodes (geniculate, petrosal and nodose) are neurogenic centers from which neuroblasts delaminate to form the Vth, VIIth, IXth and Xth cranial ganglia, respectively. The trigeminal ganglion transmits somatosensory information like temperature, touch and pain from the face to the central nervous system, while the epibranchial ganglia provide gustatory information from the oral cavity and viscerosensory information from the heart and other visceral organs. In addition, in aquatic anamniotes specialized lateral line placodes, rostral and caudal to the otic placode, give rise to the lateral line system that detects electric fields and water movement along the entire body axis. Lateral line placodes form both sensory cells and the neurons that innervate them. The olfactory placode is located next to its target, the olfactory bulb, and produces a number of different cell types including olfactory sensory neurons responsible for odorant and pheromone perception, a variety of migratory neurons that enter the brain and stem cells that regenerate sensory neurons throughout life. Finally, the otic placode develops next to the hindbrain; the placode then invaginates to form the otic vesicle, which subsequently undergoes complex morphogenetic changes. Over time it is transformed into the complex architecture of the inner ear with the semicircular canals and associated sensory patches responsible for the perception of balance, acceleration and body position, and the cochlea and its sensory hair cells for sound perception. Like the other placodes, the otic territory also generates neurons of the cochlear-vestibular (VIII) ganglion that project to their target nuclei in the hindbrain. As evident from this brief summary, the sensory placodes contribute to diverse adult organs and tissues. Yet during early development they arise from a pool of sensory progenitors that initially have the potential to give rise to any placode derivative. As development proceeds placode progenitors become different from each other, acquire their unique identity before generating neurons, sensory cells and many specialized cell types. Just before somite formation, these progenitors reside in the ectoderm surrounding the anterior neural plate (Fig. 1C), which gives rise to the brain, and have a unique transcriptional signature expressing nuclear factors of the Six and Eya families as well as neural and non-neural markers. Sandwiched between and partially overlapping with the neural plate and sensory progenitors are future neural crest cells (Fig. 1C), which will ultimately come to reside

130

Alexandre Thiery et al.

in the dorsal neural tube, from where they delaminate to form many different cell types including sensory ganglia, melanocytes and the craniofacial skeleton in the head (for review, see Baker & Bronner-Fraser, 1997; Pla & Monsoro-Burq, 2018; Prasad, Charney, & Garcia-Castro, 2019; Simoes-Costa & Bronner, 2015). Thus, all three cell populations are specified in close association. The general view holds that these cell fates are allocated around the time of gastrulation and early neurulation in a stepwise process under the influence of different signals and their downstream effectors. However, recent evidence indicates that even around the time of neural tube closure ectodermal cells continue to contribute to different lineages with the balance of ‘lineage determinants’ defining their ultimate fate, suggesting that we may have to revise our view on how and when cell fate decisions occur in vivo (Roellig, Tan-Cabugao, Esaian, & Bronner, 2017). Here, we will discuss the transcriptional networks that control the segregation of neural, neural crest and placodal cells and the specification of the otic and epibranchial lineage as an example for how placode identity is established. Comparison across different species can be challenging due to the difference in experimental design and timing, and because different members of the same transcription factor family are used in different species. Throughout this review we will therefore use gene nomenclature and embryo anatomy of an amniote embryo, the chick, but incorporate evidence from other species as appropriate. Like human embryos, avian embryos are initially a flat disc making it easy to follow dynamic changes in gene expression; their early development is relatively slow compared to Xenopus and fish allowing the dissection of fate decisions over time. Importantly, recent studies in chick have systematically probed signaling and transcriptional interactions using medium or high throughput analysis of transcriptional changes thus providing a good basis to reconstruct gene regulatory networks in vivo.

2. The preplacodal region: A territory of sense organ and cranial ganglia progenitors While precursors for different placodes are distributed widely in the ectoderm at the time of gastrulation, by head fold stages they have converged to the ectoderm surrounding the anterior neural plate in a territory termed the preplacodal region (PPR; Fig. 1C) (Bhat & Riley, 2011; Bhattacharyya, Bailey, Bronner-Fraser, & Streit, 2004; Dutta et al., 2005; Ezin, Fraser, & Bronner-Fraser, 2009; Fernandez-Garre, Rodriguez-Gallardo, Gallego-Diaz, Alvarez, & Puelles, 2002; Garcia-Martinez, Alvarez, & Schoenwolf, 1993;

Cell fate decisions in the cranial sensory nervous system

131

Hatada & Stern, 1994; Kozlowski, Murakami, Ho, & Weinberg, 1997; Pieper, Eagleson, Wosniok, & Schlosser, 2011; Streit, 2002; Xu, Dude, & Baker, 2008). Here, placode precursors are initially intermingled with neural, neural crest, and epidermal cells, but the degree of mixing differs in chick and Xenopus and might be overestimated due to labeling techniques (Pieper et al., 2011). As neurulation progresses, progenitors for different tissues and for different placodes are increasingly restricted to distinct domains. Classic transplantation experiments in amphibians showed that at neurula stages cells within the PPR are competent to give rise to any placode, while more recent experiments in chick and frog highlighted that the placode territory is the only region of the ectoderm competent to respond to placode inducing signals (Baker, Stark, Marcelle, & Bronner-Fraser, 1999; Bhattacharyya & Bronner-Fraser, 2008; Gallagher, Henry, & Grainger, 1996; Groves & Bronner-Fraser, 2000; Henry & Grainger, 1990; Jacobson, 1963; Ladher, Anakwe, Gurney, Schoenwolf, & FrancisWest, 2000; Martin & Groves, 2006). In addition, the PPR is uniquely specified: irrespective of their later fate all placode precursors are specified as lens at the 0–1 somite stage, while no other parts of the ectoderm show this property (Bailey, Bhattacharyya, Bronner-Fraser, & Streit, 2006). When chick preplacodal explants from any part along the rostro-caudal axis are cultured in isolation they rapidly activate the lens marker Pax6 and after prolonged culture lens-specific α-crystallin and δ-crystallin (Bailey et al., 2006). Thus, the PPR is endowed with special properties that distinguish it from other ectodermal territories. However, it remains to be elucidated whether this territory contains multipotent precursors or a mixture of progenitors prespecified as individual placodes, which segregate through cell movements over time. Single cell lineage tracing combined with single cell transcriptomics will be required to shed light on this question. Finally, preplacodal cells express a unique set of genes, namely members of the Six and Eya families of nuclear factors. These gene families were first identified in Drosophila as sine oculis and eyes absent, respectively, where they form a regulatory network with eyeless (Pax6) and dachshund to control compound eye development (Pignoni et al., 1997). In vertebrates, the Six gene family contains three subgroups, Six1/2, Six3/6 and Six4/5, with each gene comprising a Six domain and a Six-type homeodomain (Kawakami, Sato, Ozaki, & Ikeda, 2000). Six proteins regulate transcription through direct interaction with regulatory regions and the recruitment of other co-factors. At the neural plate border, Six1 can act as a transcriptional

132

Alexandre Thiery et al.

repressor or activator, depending upon the presence of the cofactors Eya1 and Groucho, respectively (Brugmann, Pandur, Kenyon, Pignoni, & Moody, 2004). There are four vertebrate Eya genes (Eya1-4), of which Eya1 is expressed in mouse, fish and frog preplacodal cells, while in chick, Eya2 is expressed (Ahrens & Schlosser, 2005; Hanson, 2001; Ishihara, Ikeda, Sato, Yajima, & Kawakami, 2008; Kozlowski, Whitfield, Hukriede, Lam, & Weinberg, 2005; Litsiou, Hanson, & Streit, 2005; McLarren, Litsiou, & Streit, 2003; Sahly, Andermann, & Petit, 1999; Wawersik & Maas, 2000). Together, Six1/4 and Eya1/2 expression domains encompass all placode progenitors and later continue to be expressed in all placodes, except for the lens. Importantly, they are crucial for the development of all placode derivatives including the olfactory epithelium, the inner ear and the sensory ganglia as evidenced by loss-of-function experiments in mouse, chick, fish, and frog (Chen, Kim, & Xu, 2009; Friedman, Makmura, Biesiada, Wang, & Keithley, 2005; Konishi, Ikeda, Iwakura, & Kawakami, 2006; Kozlowski et al., 2005; Laclef et al., 2003; Li et al., 2003; Ozaki et al., 2004; Xu et al., 1999; Zheng et al., 2003; Zou, Silvius, Rodrigo-Blomqvist, Enerback, & Xu, 2006). In humans, mutations in Six/Eya result in complex syndromes such as Branchio-Oto-Renal syndrome (mutations in Eya1 or Six1) that affect sense organ development (Abdelhak et al., 1997; Johnson et al., 1999; Ruf et al., 2004; Schonberger et al., 2005; Zhang, Knosp, Maconochie, Friedman, & Smith, 2004). In summary, sensory placode progenitors reside in a continuous band of ectoderm next to the neural plate, are characterized by a unique set of molecular markers and have special properties distinct form other ectodermal territories. Over time, these progenitors become different from each other and differentiate into mature placodes and their specialized cell types.

3. Subdivision of the embryonic ectoderm prior to gastrulation The embryonic ectoderm gives rise to the central and peripheral nervous systems as well as the epidermis. Already prior to gastrulation, the ectoderm begins to be molecularly subdivided into different domains that seem to foreshadow the future neural and epidermal territories. In amniotes, this is evident by patterning of the epiblast along the medial-lateral axis, and similar molecular subdivision is observed in anamniotes along the dorso-ventral axis. For example, in chick preneural markers like ERNI, Sox3, Otx2 and Geminin are strongly expressed within the medial epiblast, whereas

Cell fate decisions in the cranial sensory nervous system

133

nonneural Bmp4, Bmp7, Gata2/3 and Dlx5 are largely restricted to the periphery within the area pellucida and/or the extraembryonic area opaca (Fig. 2A) (Bally-Cuif, Gulisano, Broccoli, & Boncinelli, 1995; Papanayotou et al., 2008; Pera, Stein, & Kessel, 1999; Rex et al., 1997; Sheng & Stern, 1999; Streit et al., 1998; Streit, Berliner, Papanayotou, Sirulnik, & Stern, 2000). Likewise, in Xenopus, preneural and nonneural transcripts are highly expressed in the dorsal and ventral ectoderm, respectively (Pieper, Ahrens, Rink, Peter, & Schlosser, 2012). Recent experiments in the chick have characterized the transcriptional state of early epiblast cells in response to neural and placode inducing signals. This has revealed new transcriptional regulators with preneural expression profiles including BMI1, N-Myc, Mafa, Sox11, Trim24, Prdm1 and Znf462 (Hintze et al., 2017; Trevers et al., 2018). Like the preneural markers originally identified, many of these transcripts are confined to neural plate and/or placode and neural crest territories by gastrulation stages. However, it is important to note that prior and throughout gastrulation no sharp gene expression boundaries are apparent and that preneural and nonneural markers overlap in an intermediate region. Experiments in the chick aimed at exploring the signals that activate differential gene expression prior to gastrulation have demonstrated a role for FGF, BMP and Wnt signaling. When medial epiblast explants are cultured in the presence of Wnts or FGF inhibitors, preneural gene expression is inhibited (Wilson et al., 2001; Wilson, Graziano, Harland, Jessell, & Edlund, 2000). In contrast, exposure of the area opaca epiblast—which is competent to respond to neural, neural crest and placode inducing signals—to FGF signaling induces the expression of ERNI, Sox3 and Geminin as well as the FGF mediator Etv4 and newly identified preneural genes like N-Myc, Trim24 and Znf462 (Albazerchi & Stern, 2007; Hintze et al., 2017; Linker & Stern, 2004; Papanayotou et al., 2008; Streit et al., 2000; Trevers et al., 2018). Other transcripts like Otx2 require inhibition of both Wnt and BMP signaling in addition to FGF exposure, while genes expressed in the lateral epiblast like Msx1 and Gata2/3 require BMP and canonical Wnt signaling (Albazerchi & Stern, 2007). Like in chick, Msx1, Gata3, Foxi1 and members of the Tfap2 family are under the control of BMPs in Xenopus and zebrafish and experiments in the latter clearly establish a role for BMP signaling prior to, but not during gastrulation (Beanan, Feledy, & Sargent, 2000; Hoffman, Javier, Campeau, Knight, & Schilling, 2007; Hong & Saint-Jeannet, 2007; Kwon, Bhat, Sweet, Cornell, & Riley, 2010; Luo, Lee, Saint-Jeannet, & Sargent, 2003; Matsuo-Takasaki,

134

Alexandre Thiery et al.

Matsumura, & Sasai, 2005; Suzuki, Ueno, & Hemmati-Brivanlou, 1997). Canonical Wnt signaling, however, differentially regulates non-neural transcripts; although it activates Msx1 and Gata3, it represses Foxi1 and Dlx3. The integration of different signals is key to control early patterning events. Inhibition of FGF activates Bmp4 and 7, which in turn prevent preneural gene expression, while activation of Wnt signaling inhibits FGF activity (Wilson et al., 2001). In amniotes, the extraembryonic hypoblast (anterior visceral endoderm in the mouse) underlying the epiblast emerges as the source of FGFs, Wnt and BMP antagonists, and exposure of competent epiblast to the hypoblast induces a large number of preneural genes (Albazerchi & Stern, 2007; Chapman, Brown, Lees, Schoenwolf, & Lumsden, 2004; Streit, 2008; for review: Stern & Downs, 2012; Trevers et al., 2018). In contrast, BMP and Wnt family members are expressed in the extraembryonic region adjacent to the nonneural domain (Skromne & Stern, 2001; Streit et al., 1998; Wilson et al., 2001). In summary, prior to gastrulation antagonism between FGF signaling on the one hand and BMP/Wnt signaling on the other, begins to partition the epiblast into preneural and nonneural territories. These domains are characterized by distinct transcription factors, although their expression overlaps in an intermediate region. At these stages, the transcription factor interactions have not been explored. However, it is likely that positive feedback loops and cross-repressive interactions as observed at later stages (see below) are already in place to implement the subdivision of the ectoderm downstream of these signals.

4. The neural plate border As development proceeds preneural and non-neural territories become increasingly distinct with the onset of new transcriptional regulators. As the neural plate is induced and definitive neural markers like Sox2 become expressed at gastrulation stages, preneural and nonneural genes continue to overlap in the adjacent ectoderm, while some future neural crest factors like FoxD3 become widely activated in the neural plate and the adjacent ectoderm (Khudyakov & Bronner-Fraser, 2009; Papanayotou et al., 2008; Rex et al., 1997). By early neural plate stages gene expression domains are more complex. Nonneural genes are expressed in nested domains with Tfap2 family members and Dlx5/6 abutting Sox2, while Gata2/3 and Foxi1/3 are confined more laterally. Msx1 and FoxD3 are now restricted to a stripe of cells at the edge of the neural plate, coexpressed with Pax3 and 7

Cell fate decisions in the cranial sensory nervous system

135

(Fig. 2A) (Khudyakov & Bronner-Fraser, 2009; McLarren et al., 2003; Pieper et al., 2012; Streit, 2002; Woda, Pastagia, Mercola, & Artinger, 2003). Together they are part of a regulatory network to specify neural crest cells (see below). Partially overlapping with these genes, the placode progenitor specifier genes Six1/4 and Eya1/2 are activated; by late neural plate stages, expression domains become further restricted and appear to mark distinct territories. How are definitive neural crest and placode progenitors specified in the ectoderm next to the neural plate? Two apparently opposing models have been discussed (for review: Schlosser, 2014). The binary competence model proposes that the ectoderm is initially subdivided into neural and non-neural territories and that neural crest cells arise from the former, while placodes arise from the latter. The neural plate border model suggests that both populations emerge gradually over time as gene expression domains become refined from pregastrula to late neural plate stages; this implies that neural plate border cells are in a distinct transitory or unstable state and that the border represents a territory of mixed or multipotent precursors. The term neural plate border (NPB) was coined to describe the ectoderm where preneural and non-neural transcripts overlap, because of its “mixed” transcriptional identity, because fate maps in chick, frog and fish reveal that it contains progenitors for all four ectodermal derivatives—the neural plate, neural crest, placodes and epidermis - and because experimental apposition of neural plate and epidermis generates a new territory—the border to which both tissues contribute (Bhattacharyya et al., 2004; Dickinson, Selleck, McMahon, & Bronner-Fraser, 1995; Dutta et al., 2005; Ezin et al., 2009; Fernandez-Garre et al., 2002; Garcia-Martinez et al., 1993; Kozlowski et al., 1997; Moury & Jacobson, 1989; Pieper et al., 2011; Selleck & Bronner-Fraser, 1995; Streit, 2002; Streit & Stern, 1999; Xu et al., 2008). Transcriptional profiling in Xenopus and chick shows that, in addition to preneural and nonneural genes, this region also expresses some genes that characterize pluripotent pregastrula epiblast (Buitrago-Delgado, Nordin, Rao, Geary, & LaBonne, 2015; Trevers et al., 2018). It has been suggested that this may reflect the continued ability of NPB cells to generate multiple fates. This idea is supported by fate mapping studies in chick showing that future neural, neural crest and placodal cells continue to be intermingled at the NPB until early somite stages, while the border becomes molecularly partitioned into subdomains (Fig. 2) (Bhattacharyya et al., 2004; Streit, 2002; Xu et al., 2008; for review: Grocott et al., 2012; McCabe & Bronner-Fraser, 2009; Meulemans & Bronner-Fraser, 2004; Moody &

136

Fig. 2 See legend on opposite page.

Alexandre Thiery et al.

Cell fate decisions in the cranial sensory nervous system

137

LaMantia, 2015; Park & Saint-Jeannet, 2010; Pla & Monsoro-Burq, 2018; Schlosser, 2006; Simoes-Costa & Bronner, 2015; Streit, 2008). More recently, analysis of only four proteins in chick—Sox2 (neural), Tfap2a (nonneural), Pax7 (neural crest) and Six1 (placodal)—revealed surprising heterogeneity in gene expression at the NPB even as late as neural tube stages, with cells coexpressing combinations of up to four markers (Roellig et al., 2017). Lineage tracing using a “neural-specific” Sox2 enhancer reveals that cells that are initially Sox2 positive can contribute not only to the neural plate, but also to the neural crest and epidermis (Roellig et al., 2017). Thus, while classifying cells based on broad transcriptional domains is useful, it is clear from the study of only a few select markers that cells in the border territory are heterogeneous and exist in different transcriptional states. In contrast, the degree of cell intermingling in the future crest and placode territory is less pronounced in Xenopus (Pieper et al., 2011). Together with a series of transplantation experiments this observation forms the basis for the binary competence model (Pieper et al., 2012). When

Fig. 2 Gene regulatory networks at the neural plate border. (A) Regulatory states at the neural plate border. Changes in gene expression is represented diagrammatically. At early gastrula stages, preneural (green) and nonneural (brown, light brown) overlap in the neural plate border (grey in diagram on the left), where future neural crest factors (red) are also expressed. By head process stages, the neural plate has formed expressing definitive neural markers (blue), while nonneural genes subdivide the neural plate border into different domains with some factors (light brown) abutting the neural plate. Neural crest and sensory progenitor markers become expressed within these two subdomains, but partially overlap. (B) Regulatory interactions that establish the preplacodal region and the expression of the core Six-Eya network. Neural plate border genes are activated by signals from within the ectoderm and from the underlying mesoderm, and provide critical input for Six1, Six4 and Eya2. While the Six-Eya cassette promotes its own expression, it represses neural crest factors, thus stabilizing the sensory progenitor regulatory state. Solid lines represent direct interactions, dashed lines represent interactions extracted from functional experiments that may be direct or indirect and blue diamonds indicate evidence from enhancer analysis. For details see text and Supplementary Table 1 in the online version at https://doi.org/10.1016/bs.ctdb.2020. 04.002. (C) Regulatory interactions that establish the core neural crest cells factors FoxD3 and Snail2. Induced signals from adjacent tissues, neural plate border genes promote their own expression and activate the neural crest specifiers FoxD3 and Snail2, while inhibiting the expression of the Six/Eya network. Solid lines represent direct interactions, dashed lines represent interactions extracted from functional experiments that may be direct or indirect and blue diamonds indicate evidence from enhancer analysis. For details see text and Supplementary Table 1 in the online version at https://doi. org/10.1016/bs.ctdb.2020.04.002.

138

Alexandre Thiery et al.

definitive neural plate is grafted into the edge of the neural plate, placode precursor markers Six1 and Eya1 are only induced in the nonneural ectoderm, while neural crest genes are activated in the grafted neural plate itself. Thus, competence to form placodes is lost from neural tissue, while neural crest competence is not. These findings led to the idea that by neurula stages neural and neural crest cells are segregated from placodes and future epidermis. However, if the same experiment is performed earlier, the neural plate can form both cell populations in agreement with earlier experiments in amphibians and in chick (Dickinson et al., 1995; Moury & Jacobson, 1989; Selleck & Bronner-Fraser, 1995; Streit & Stern, 1999). Like in chick, changes in non-neural and neural gene expression accompany these changes in competence. It is therefore likely that rather than using different modes of neural crest and placode precursor specification both models can be reconciled when considering developmental time. At early gastrula stages cells of different fates are largely intermingled expressing both preneural and nonneural factors, but as development proceeds their potential is gradually restricted as the ectoderm is molecularly subdivided and “fate determinants” become expressed. Support for the NPB model also comes from specification assays performed in the chick. As discussed above before gastrulation the epiblast is roughly subdivided into preneural and nonneural domains with an intermediate region where markers for both overlap. When medial and lateral epiblast is cultured in isolation, explants first express preneural (ERNI, Sox3) and nonneural markers (Dlx5, Gata3, Msx1) together with neural crest (Snail2, Pax7), neural plate (Sox2) and placode progenitor genes (Six4, Eya2, Pax6) before differentiating into the placode derivative lens, definitive neural and neural crest cells (Linker et al., 2009; Trevers et al., 2018; Wilson et al., 2000). No difference is observed between lateral and medial epiblast explants suggesting that despite a difference in gene expression epiblast cells have the same potential to generate all three neural lineages normally derived from the NPB. Therefore, before gastrulation epiblast cells are in a transcriptional state that resembles the NPB. Although “fate-specific” programs are gradually activated in the NPB, cells in this territory are not yet committed to their final identity and fates remain fluid. This is clearly demonstrated by the fact that misexpression of different transcription factors shifts gene expression domains and cell fate. For example, overexpression of the placode precursor gene Six1 causes future neural crest and neural cells to adopt a placodal identity, whereas its downregulation leads to a narrowing of the preplacodal domain

Cell fate decisions in the cranial sensory nervous system

139

(Brugmann et al., 2004; Christophorou, Bailey, Hanson, & Streit, 2009). Likewise, misexpression of Sox2 within the putative neural crest domain downregulates the definitive neural crest maker Snail2, while misexpression of the NPB factor Pax3 can expand or abolish the future neural crest domain depending on its concentration and cofactors (Hong & Saint-Jeannet, 2007; Wakamatsu, Endo, Osumi, & Weston, 2004). These experiments suggest that mutually repressive interactions between neural, neural crest and placodal factors refine cell fate choices and we will discuss the regulatory circuits below. Furthermore, they indicate that their interactions may maintain gene expression boundaries while cell fates remain plastic. Together with the transcriptional heterogeneity discussed above, a scenario emerges where NPB cells appear to be in a transcriptionally unstable state and retain their ability to generate multiple cell types for much longer than previously thought. Advancement in single cell transcriptomics together with lineage tracing of single cells with defined gene expression profiles will provide novel insight into how individual cells acquire their identity at the NPB.

5. Signaling events at the neural plate border The molecular interactions regulating cell fate choices at the NPB have been reviewed extensively (Grocott et al., 2012; McCabe & Bronner-Fraser, 2009; Meulemans & Bronner-Fraser, 2004; Moody & LaMantia, 2015; Park & Saint-Jeannet, 2010; Pla & Monsoro-Burq, 2018; Schlosser, 2006; Simoes-Costa & Bronner, 2015; Streit, 2008). Following the activation of preneural and nonneural genes and the induction of the definitive neural plate, FGF, Wnt and BMP pathways continue to control play a role in regulating gene expression and cell fates at the NPB (Fig. 2; Supplementary Table 1 in the online version at https://doi.org/10.1016/ bs.ctdb.2020.04.002). Early models for ectodermal patterning proposed that cells read out a discrete level of BMP activity with the secretion of BMP antagonists from the organizer and midline establishing low to high gradient along the mediallateral axis, in turn imparting neural, neural crest, placodal and epidermal fates respectively (Aybar & Mayor, 2002; Vonica & Brivanlou, 2006). Accordingly, Xenopus animal caps from embryos injected with BMPs or BMP antagonists develop into neural, NPB and epidermal fates depending on the level of BMP activity and intermediate levels of BMP signaling have been shown to specify neural crest cells in fish (Brugmann et al., 2004; Glavic et al., 2004; LaBonne & Bronner-Fraser, 1998; Marchant,

140

Alexandre Thiery et al.

Linker, Ruiz, Guerrero, & Mayor, 1998; Reichert, Randall, & Hill, 2013; Schumacher, Hashiguchi, Nguyen, & Mullins, 2011; Tribulo, Aybar, Nguyen, Mullins, & Mayor, 2003; Wilson, Lagna, Suzuki, & HemmatiBrivanlou, 1997). However, animal caps contain NPB cells, which in chick are the only cells responsive to BMP modulation at late gastrula stages, suggesting that at the role of BMP signaling is more complex (Linker et al., 2009; Streit et al., 1998; Streit & Stern, 1999). BMP signaling via SMAD1/5/8 activates the NBP genes Foxi1/3, Gata2/3, Dlx5/6, Tfap2, Msx1 and Zic1 (for review: Moody & LaMantia, 2015; Pla & MonsoroBurq, 2018; Schlosser, 2014; Simoes-Costa & Bronner, 2015). Recent experiments in fish clearly revealed two phases of BMP activity (Kwon et al., 2010). Using pharmacological inhibition of BMPs at different time points and inducible BMP antagonists this study shows that active BMP is required for placode progenitor formation before gastrulation for the induction of Tfap2a/c, Gata3 and Foxi1 at the NPB, but must be blocked later to allow sensory progenitor gene expression. In addition, Wnt and FGF pathways are repeatedly used during the specification of NPB fates (Chang & Hemmati-Brivanlou, 1998; LaBonne & Bronner-Fraser, 1998; Litsiou et al., 2005; Monsoro-Burq, Fletcher, & Harland, 2003; Monsoro-Burq, Wang, & Harland, 2005; Patthey, Gunhaga, & Edlund, 2008; Villanueva, Glavic, Ruiz, & Mayor, 2002; for review: Streit, 2008). As discussed above, FGF signaling is critical for the induction of preneural genes prior to gastrulation and initiates the neural, neural crest and placodal programs (Ahrens & Schlosser, 2005; de Croze, Maczkowiak, & Monsoro-Burq, 2011; Delaune, Lemaire, & Kodjabachian, 2005; Hintze et al., 2017; Litsiou et al., 2005; Marchal, Luxardi, Thome, & Kodjabachian, 2009; Sasai, Lu, Piccolo, & De Robertis, 1996; Streit et al., 2000; Stuhlmiller & Garcia-Castro, 2012; Trevers et al., 2018; Wilson et al., 2000). At the time of NBP specification, various FGF family members are expressed in the underlying paraxial and lateral head mesoderm and the neural plate (Ahrens & Schlosser, 2005; Monsoro-Burq et al., 2003; Ohuchi, Kimura, Watamoto, & Itoh, 2000; Shamim & Mason, 1999). FGF activity is required for the expression of early NPB markers such as Msx1, Tfap2a, Zic1 and Pax3, which in turn are necessary for placode and neural crest specification (Barrallo-Gimeno, Holzschuh, Driever, & Knapik, 2004; de Croze et al., 2011; Garnett, Square, & Medeiros, 2012; Hong & Saint-Jeannet, 2007; Kwon et al., 2010; Maharana & Schlosser, 2018; Monsoro-Burq et al., 2003; Nichane et al., 2008; Sato, Sasai, & Sasai, 2005; Stuhlmiller & Garcia-Castro, 2012;

Cell fate decisions in the cranial sensory nervous system

141

Suzuki et al., 1997; Woda et al., 2003). Likewise, canonical Wnt signaling is needed for the activation of Tfap2, Msx1, and Pax3 with Tfap2 and Msx1 being direct targets (de Croze et al., 2011; Monsoro-Burq et al., 2005). Wnt signaling also directly controls Gbx2 in the posterior ectoderm (Li, Kuriyama, Moreno, & Mayor, 2009). Finally, in zebrafish zic3 and pax3a are regulated by the combination of FGF, Wnt and BMP signaling; this information is integrated by multiple enhancers that respond differentially to these signals (Garnett et al., 2012). In summary, a complex signaling network acts to control the expression of NBP markers at late gastrula and early neurula stages. However, how this is integrated by regulatory elements that control the expression of downstream transcription factors is largely unexplored. Below we will discuss how NBP cells maintain their identity through mutual repression and positive feedback loops among NPB transcription factors.

6. Transcription factor interactions: From neural plate border to neural crest and placode precursors By head process stage, the NPB is characterized by the expression of NPB specifiers including Zic, Tfap2, Pax, Msx and Dlx family members. NPB specifiers can be defined as transcription factors which are both expressed at the NBP and required for the establishment NPB gene expression and in turn for placodal and neural crest specification (Meulemans & Bronner-Fraser, 2004; Pla & Monsoro-Burq, 2018). However, the expression of NPB specifiers is not uniform throughout the NPB, with certain genes preferentially upregulated in domains which roughly mark future neural crest and placodal populations (Fig. 2; Supplementary Table 1 in the online version at https://doi.org/10.1016/bs.ctdb.2020.04.002). Tfap2a has been identified as a key regulator of fate choice at the NPB and regulates both placodal and neural crest targets. Which fate is induced depends upon the presence of additional factors that cooperate with Tfap2a, which in turn is regulated by the modulation of BMP, FGF and Wnt signaling. Pax3 and Msx1 are directly activated by Tfap2a in the presence of Wnt signaling, while Zic1 is also activated indirectly (de Croze et al., 2011). Interestingly, Zic1 is required for placodal specification and positively regulates Six/Eya expression (Hong & Saint-Jeannet, 2007). However, when expressed together, Zic1, Pax3 and Tfap2a are sufficient to induce the neural crest specifiers Snail2 and FoxD3 in the nonneural ectoderm (de Croze et al., 2011). Msx1, Pax3/7, Axud1 and Zic1 interact and form

142

Alexandre Thiery et al.

a positive feedback loop with Tfap2a, resulting in the upregulation of their own expression (Bhat, Kwon, & Riley, 2013; de Croze et al., 2011; Monsoro-Burq et al., 2005; Plouhinec et al., 2014; Sato et al., 2005; Simoes-Costa, McKeown, Tan-Cabugao, Sauka-Spengler, & Bronner, 2012; Simoes-Costa, Stone, & Bronner, 2015). Downstream, they activate the expression of neural crest specifiers Snail2 and FoxD3, with Msx1, Axud1 and Pax7 cooperating directly to upregulate FoxD3 via the FoxD3-NC1 and FoxD3-NC2 enhancers (de Croze et al., 2011; SimoesCosta et al., 2012, 2015). In addition to positive feedback loops, the neural crest program reinforces itself through repressing preplacodal specifiers. Axud1, Pax3/7, Msx1 and FoxD3 all repress Six1 and Eya1/2, with Pax7 and Msx1 acting directly by binding to the Six1-14 enhancer (Maharana & Schlosser, 2018; Sato et al., 2010; Simoes-Costa et al., 2015). These genes form a core network which initiates neural crest specification, with the downregulation of Tfap2a, Pax3/7, Zic1, Axud1 or Msx1 leading to the loss of neural crest specifiers Snail2 and FoxD3 (Basch, Bronner-Fraser, & Garcia-Castro, 2006; de Croze et al., 2011; Gutkovich et al., 2010; Hong & Saint-Jeannet, 2007; Li et al., 2017; Monsoro-Burq et al., 2005; Sato et al., 2005; Simoes-Costa et al., 2015). Transcriptional analysis following Axud1 downregulation reveals that not only are neural crest specifiers lost, but also core elements of the preplacodal program including Six1, Eya1 and Irx1 are upregulated (Simoes-Costa et al., 2015). This highlights the unstable state of cells at the NPB and the importance of crossrepression and positive feedback loops in establishing a stable early neural crest transcriptional state distinct from the preplacodal state (Fig. 2C). Specification of the sensory placodes centers around the induction of Six1/4 and Eya1/2 (Fig. 2B). They are expressed throughout the PPR where they label all placodal progenitors, and act as a hub to integrate a complex upstream signaling cascade and activate placode specific transcription downstream (Bhattacharyya et al., 2004; Christophorou et al., 2009; Kobayashi, Osanai, Kawakami, & Yamamoto, 2000; Litsiou et al., 2005; McLarren et al., 2003; Pandur & Moody, 2000; Schlosser & Ahrens, 2004; Streit, 2002; Xu et al., 2008). Upstream of Six/Eya, a number of factors including Dlx, Gata, Foxi, Zic and Tfap2 family members interact to activate Six/Eya and repress alternative fates (Grocott et al., 2012; Moody & LaMantia, 2015; Schlosser, 2006; Streit, 2018). A recent study showed that the underlying prechordal mesendoderm and lateral head mesoderm provide the necessary signals required for PPR induction, including signals which impart regional character of the PPR along the

Cell fate decisions in the cranial sensory nervous system

143

anterior-posterior axis (Hintze et al., 2017; Litsiou et al., 2005). In response to these tissues a hierarchy of transcription factors is induced and this has provided new factors, including Znf462 and Pdlim4, which regulate the activation of Six1 and Eya2 (Hintze et al., 2017). Knock-down of Znf462 results in downregulation of Foxi3, Gata3 and Dlx6, which are in turn important regulators of Six1 and Eya2. Transiently elevated BMP signaling is required for the co-induction of PPR competence factors Foxi1 and Gata3 alongside Tfap2a/c, although downstream BMP attenuation is required for placodal specification (discussed below) (Ahrens & Schlosser, 2005; Kwon et al., 2010; Litsiou et al., 2005). Interestingly, when BMP is partially inhibited, expression of Gata3 and FoxI1 is lost at the NPB, while Tfap2a/c are induced together with the neural crest specifier FoxD3 (Bhat et al., 2013; Brugmann et al., 2004; Kwon et al., 2010). Tfap2a, Foxi1 and Gata3 positively interact with each other: downregulation of either transcript results in a reduction of the others. This positive feedback loop serves to activate the Six/Eya cassette downstream (Bhat et al., 2013; Matsuo-Takasaki et al., 2005; Pieper et al., 2012). While neural crest specifiers are dependent upon positive feedback interactions with their upstream inducers, Six/Eya regulate PPR induction in a different manner. Once induced, Six1 and Eya1/2 autoregulate their own expression, while downregulating the expression of their upstream activators, Dlx5/6, Gata2/3 and Prdm1 (Brugmann et al., 2004; Christophorou et al., 2009; Maharana & Schlosser, 2018; Prajapati, Hintze, & Streit, 2019). They also act to prevent neural crest induction through repression of Msx1, Pax3/7 and FoxD3 (Brugmann et al., 2004; Christophorou et al., 2009; McLarren et al., 2003). These interactions serve to establish robust expression of Six1 and Eya1/2 and define the preplacodal state (Fig. 2B). In summary, NPB specifiers and their downstream targets form a complex regulatory network containing a number of cross-repressive interactions and feedback loops, which specify cell fate according to the combination of transcription factors expressed. Given the cell heterogeneity and large number of cross-repressive interactions, NPB cells appear to be in an unstable state. Our understanding of the regulatory cascade which determines fate allocation in this territory is far from complete. However, the recent surge in transcriptomic assays has paved the way for novel insight into the dynamic gene interactions at the NPB but will require the combination of different approaches to extract biologically meaningful information.

144

Alexandre Thiery et al.

7. Segregation of neural crest and sensory placode precursors While the regulatory circuits described above are crucial to maintain the NPB, continued signaling events via BMP and Wnt modulation cooperate with cell specific transcriptional programs controlling the specification of neural crest and placodal lineages (Fig. 2). By neural plate stages BMP2, 4, and 7 are strongly expressed in the forming neural folds next to the sensory progenitor domain and colocalizing with future neural crest cells, as is the BMP target Msx1 (Ahrens & Schlosser, 2005; Chapman, Schubert, Schoenwolf, & Lumsden, 2002; Fainsod, Steinbeisser, & De Robertis, 1994; Litsiou et al., 2005; McLarren et al., 2003; Streit et al., 1998; Streit & Stern, 1999; Suzuki et al., 1997; Tribulo et al., 2003). Indeed, active BMP signaling has been implicated in neural crest cell formation (Liem Jr., Tremml, Roelink, & Jessell, 1995). However, the strong expression of BMPs is at odds with the intermediate level of phospho-Smads 1/5/8 in the NPB and the idea that intermediate levels of activity specify neural crest cells (LaBonne & Bronner-Fraser, 1998; Marchant et al., 1998; Stuhlmiller & GarciaCastro, 2012; Tribulo et al., 2003; Villanueva et al., 2002; Wilson et al., 1997). A recent study in chick revealed that the intercellular factors CKIP-1/Smurf modulate BMP signaling in the NPB by regulating Smad degradation and consequently keep BMP activity at the intermediate level required for neural crest formation (Piacentino & Bronner, 2018). In contrast, placode progenitors require low or no BMP signaling. Experiments in fish have shown that BMP inhibition no longer prevents the expression of preplacodal markers, while in chick and frog, inhibition of BMP signaling expands the placodal domain and overexpression of BMP4 inhibits Six1 expression (Ahrens & Schlosser, 2005; Bhat et al., 2013; Kwon et al., 2010; Litsiou et al., 2005). Indeed, placode progenitors are protected from BMP activity by antagonists emanating from the underlying mesoderm and the placodal domain itself (Chapman et al., 2004; Esterberg & Fritz, 2009; Litsiou et al., 2005; Ogita et al., 2001; Rodriguez Esteban et al., 1999). Thus, BMP activity must be carefully controlled in time and space to allow the specification of neural crest and placode progenitors form the NPB. As discussed above Wnt signaling plays an important role in NBP gene activation and consequently is critical for the generation of neural crest cells

Cell fate decisions in the cranial sensory nervous system

145

and sensory progenitors. At early somite stages this pathway continues to play an important role and several Wnts are expressed in or surrounding the NPB. While Wnt6 is expressed in the non-neural ectoderm, Wnt1 is confined to the neural folds and Wnt8c to the lateral and posterior mesoderm (Baranski, Berdougo, Sandler, Darnell, & Burrus, 2000; Garcia-Castro, Marcelle, & Bronner-Fraser, 2002; Litsiou et al., 2005). When NBP explants are cultured in the presence of Wnt agonists, neural crest markers are upregulated, while Wnt inhibition leads to the loss of neural crest specifiers like FoxD3 and Snail2, with the latter likely to be a direct Wnt target (Garcia-Castro et al., 2002; Patthey et al., 2008; Vallin et al., 2001). Recent evidence from chick has identified the transcription factor Axud as critical link between Wnt signaling and its downstream targets in neural crest cells (Simoes-Costa & Bronner, 2015). In contrast to the neural crest, sensory progenitors form in the absence or at low levels of Wnt signaling. Activation of the placodal program in chick NPB explants requires Wnt inhibition (Patthey et al., 2008; Patthey, Edlund, & Gunhaga, 2009). When Wnt antagonists are misexpressed in vivo, either in the NPB in chick or when injected into the 32-cell stage in Xenopus, the preplacodal domain expands at the expense of neural crest cells, while Wnt activation has the opposite effect (Brugmann et al., 2004; Litsiou et al., 2005). Therefore, the level of Wnt activity at least in part controls whether NPB cells differentiate into crest or placodes. In addition, Wnt signaling from the trunk region ensures restriction of sensory progenitor formation to the head ectoderm (Brugmann et al., 2004; Litsiou et al., 2005). At early somite stages, Wnt antagonists are expressed in the mesoderm underlying placode progenitors and thus protect them from inhibitory Wnts emanating from surrounding tissues. Finally, retinoic acid has also been implicated in confining the placode territory to the anterior ectoderm, although the molecular mechanisms have not been explored (Villanueva et al., 2002). Thus, Wnt and BMP signaling are used repeatedly during the activation of the neural crest and placode program and their activity is tightly controlled to as different cell fates emerge from the NPB.

8. Subdividing the placode territory along the anterior-posterior axis While sensory progenitors are not yet committed to a specific placode fate and are competent to give rise to any placode, different transcription factors already subdivide the PPR along the rostro-caudal axis.

146

Alexandre Thiery et al.

For example, Otx2 is expressed in the anterior PPR, while Gbx2 is restricted to its posterior portion (Acampora et al., 1995; Acampora, Gulisano, Broccoli, & Simeone, 2001; Bally-Cuif et al., 1995; Li et al., 2009; Simeone, Acampora, Gulisano, Stornaiuolo, & Boncinelli, 1992; Simeone et al., 1993; Tour, Pillemer, Gruenbaum, & Fainsod, 2001; von Bubnoff, Schmidt, & Kimelman, 1996). Acting on the prepatterned ectoderm, two different tissues induce sensory progenitors, the lateral head mesoderm and the anterior mesendoderm (Hintze et al., 2017; Litsiou et al., 2005). Both induce the expression of generic sensory progenitor markers, but over time the lateral mesoderm promotes caudal character (Gbx2++, Foxi3+), whereas the anterior mesendoderm promotes anterior genes including Otx2, Six3, the neuropeptide pNoc and its receptor SSTR5. Thus, regionalization of the placode territory occurs concomitant with its induction. As a result, different pairs of transcription factors form expression boundaries at distinct rostro-caudal positions (Fig. 3) (for review: Grocott et al., 2012; Moody & LaMantia, 2015; Saint-Jeannet & Moody, 2014; Schlosser, 2006, 2014). While Otx2 and Gbx2 abut at the boundary between trigeminal and otic-epibranchial precursors, Six3 and Irx1/2/3 do so just posterior to lens progenitors. Like in the neural tube, cross-repressive interactions between them sharpen the boundaries endowing cells with distinct

Fig. 3 Rostro-caudal subdivision of the preplacodal territory. At neural plate stages, pairs of transcription factors form partially overlapping gene expression boundaries. By early somite stages boundaries have been sharpened and members of the Pax gene family divide the placode territory into anterior, intermediate and posterior placodal domains. Diagram on the left: gene expression domains along the rostro-caudal axis; diagram on the right: chick embryo at 5–7-somite stages and the expression domains of Pax6, Pax3 and Pax2.

Cell fate decisions in the cranial sensory nervous system

147

rostro-caudal identity (Broccoli, Boncinelli, & Wurst, 1999; Glavic, Gomez-Skarmeta, & Mayor, 2002; Hidalgo-Sanchez, Millet, BlochGallego, & Alvarado-Mallart, 2005; Joyner, Liu, & Millet, 2000; Katahira et al., 2000; Li & Joyner, 2001; Millet et al., 1999; Steventon, Mayor, & Streit, 2012; Wassarman et al., 1997). At early somite stages, additional transcription factors become expressed, notably Pax family members with Pax6 encompassing adenohypophyseal, olfactory and lens precursors, Pax3 labeling the ophthalmic trigeminal territory and Pax2 otic-epibranchial progenitors (OEPs). These genes require transcriptional input from the Six/Eya cassette, although it remains to be determined whether this occurs through direct interaction with relevant regulatory elements (Christophorou et al., 2009). The Pax proteins themselves cross repress each other: misexpression of Pax3 represses Pax6 and Pax2 and vice versa (Christophorou, Mende, Lleras-Forero, Grocott, & Streit, 2010; Dude et al., 2009; Wakamatsu, 2011). As development proceeds other factors become activated and the PPR is divided into successively smaller molecular domains as individual placodes emerge. Eventually, each placode expresses its unique transcription factor code endowing cells with their unique identity (Grocott et al., 2012; Moody & LaMantia, 2015; Saint-Jeannet & Moody, 2014; Schlosser, 2006). The signals that control placode induction from sensory progenitor cells have been explored extensively, however the downstream transcription factor networks are less well understood largely because only few regulatory elements have been characterized that integrate both signaling and transcriptional inputs. In the following section, we will focus on the derivatives of the posterior PPR, the otic and epibranchial placodes, as an example for how sensory progenitors acquire different fates.

9. Otic and epibranchial placode formation: Signals and regulatory circuits 9.1 FGF signaling induces otic-epibranchial progenitors At early somite stages otic-epibranchial progenitors (OEPs) are located in the surface ectoderm next to the neural tube just rostral to the first somite (Bhat & Riley, 2011; Pieper et al., 2011; Streit, 2002). They are characterized by the expression of Pax2 and are intermingled with neural crest and epidermal precursors (Groves & Bronner-Fraser, 2000; Streit, 2002; Torres, Go´mez-Pardo, & Gruss, 1996; Xu et al., 2008). As development proceeds, cell fates segregate and the otic and epibranchial placodes emerge as distinct domains: the otic epithelium forms close to the neural tube, while

148

Alexandre Thiery et al.

epibranchial placodes—geniculate, petrosal and nodose—form dorsal of the pharyngeal pouches. Signals from the surrounding tissues are key to induce OEPs and definitive placodes from the posterior PPR. While gain and loss-of-function approaches in mouse, chick, and fish clearly demonstrate that FGFs from the underlying mesoderm initiate the OEP program (Fgf19 in chick, Fgf8 in zebrafish, Fgf10 in mouse), the downstream transcriptional networks have only recently been identified (Chen & Streit, 2013; Ladher et al., 2000; Phillips, Bolding, & Riley, 2001; Sai & Ladher, 2015; Wright & Mansour, 2003). As a first response, FGF signaling stabilizes posterior PPR identity (Fig. 4A). Working through the MAP-kinase pathway and AP1 phosphorylation, FGF signaling leads to the rapid deposition of active histone marks near ear specific enhancers, which in turn promotes the activation of genes associated to them (Tambalo, Anwar, Ahmed, & Streit, 2020; Yang et al., 2013). Among its direct targets are the SoxD group factor Sox13 and the FGF effectors Etv4 and 5, while Gbx2 and Foxi3, which are already expressed, are further upregulated indirectly (Anwar et al., 2017; Hidalgo-Sanchez, Alvarado-Mallart, & Alvarez, 2000; Khatri & Groves, 2013; Niss & Leutz, 1998; Ohyama & Groves, 2004; Solomon, Logsdon Jr., & Fritz, 2003; Tambalo et al., 2020). The regulatory region that controls Foxi3 expression contains both AP1 and SoxD motifs suggesting that these factors contribute to its activation. Foxi3 forms a positive feedback loop with the PPR specifiers Six1 and Eya2, while network inference predicts activation of Foxi3 by Gbx2 and vice versa (Fig. 4A) (Anwar et al., 2017; Khatri, Edlund, & Groves, 2014; Maharana, Pollet, & Schlosser, 2017; Solomon et al., 2003). Like posterior PPR genes, Pax2 is also rapidly induced by FGF signaling and its expression requires dual input from the FGF mediators Etv4/5 on one hand and the posterior PPR factors Foxi3, Gbx2 and Six1/Eya2 on the other (Brugmann & Moody, 2005; Chen et al., 2017; Christophorou et al., 2010; Hans, Liu, & Westerfield, 2004; Khatri et al., 2014; Nissen, Yan, Amsterdam, Hopkins, & Burgess, 2003). While promoting OEP identity the FGF pathway also represses late otic placode transcripts (Lmx1b, Myb) and alternative fates: anterior PPR markers like Pax6 and Otx2 and neural crest markers like Id2 and 4 are rapidly downregulated (Fig. 4A) (Tambalo et al., 2020). Among these Pax6, Id2 and 4 appear to be direct targets. Together, these observations suggest that FGF signaling initiates a series of feed forwards loops, in which posterior PPR genes maintain their own expression and activate the earliest OEP specific factor Pax2, while simultaneously repressing alternative fates. Once activated, OEPs may

Cell fate decisions in the cranial sensory nervous system

149

Fig. 4 Otic-epibranchial progenitor specification and segregation. (A–C) Left: embryo diagrams at appropriate developmental stages; right: gene networks showing the signaling and transcription factor interactions that induce and maintain cell fates (see Supplementary Table 1 in the online version at https://doi.org/10.1016/bs.ctdb. 2020.04.002). Dashed lines represent gene interactions deduced largely from knock down experiments in different species. Solid lines represent direct interactions verified by experiments involving inhibition of protein synthesis by cycloheximide (orange diamonds) or by enhancer analysis (blue diamonds: presence of transcription factor (Continued)

150

Alexandre Thiery et al.

then be independent of sustained signaling. Thus, FGF signaling initially reinforces a regulatory state characteristic for posterior sensory progenitors, which in turn leads to initiation of the OEP program. Downstream of the posterior PPR-OEP module, other transcription factors are activated in a temporal hierarchy, and loss-of-function experiments reveal some of their regulatory interactions (Fig. 4B) (Chen et al., 2017; Yang et al., 2013). While Sox8 seems to be regulated by FGF signaling alone presumably through Etv4/5 or AP1, Pax2 and Etv4/5 work in concert to activate many of the downstream events (Chen et al., 2017; Yang et al., 2013). Pax2 controls the expression Lmx1a, Zbtb16, Prdm1 and Sall4 and maintains the sensory progenitor gene Six1, while Lmx1a regulates Zbtb16, Sox13 and Pax2. Thus, Pax2 and Lmx1a mutually activate their expression, while at the same time repressing factors that characterize other placodes like Pax3 (Chen et al., 2017; Dude et al., 2009). Indeed, Pax2 is critical for ear formation; in mouse loss of Pax2 function leads to failure of the cochlea to develop from the medial otic vesicle, while functional studies in chick point to an important role at early placode development and in maintaining OEP proliferation (Burton, Cole, Mulheisen, Chang, & Wu, 2004; Christophorou et al., 2010; Favor et al., 1996; Freter et al., 2012; Torres et al., 1996). In summary, FGF signaling activates a small regulatory circuit, in which transcription factors maintain their own expression and provide input for the next tier of factors, which in turn promote OEP character, while repressing alternative fates. Fig. 4—Cont’d binding site and mutational analysis; pink diamonds: presence of transcription factor binding sites). (A) At neural plate stages OEPs lie in the posterior PPR (pink). FGF signaling from the mesoderm enhances the expression of posterior PPR genes (Etv4, Sox13, Foxi3, Gbx2), while repressing transcripts expressed in anterior placode progenitors and neural crest cells. Together with Six1/Eya2 posterior PPR factors form positive feedback loops to enhancer their own expression and activate the OEP marker Pax2. Note: interaction of Foxi3 and Gbx2 is predicted by network inference (Anwar, Tambalo, Ranganathan, Grocott, & Streit, 2017). (B) At early somite stages, new OEP factors are activated downstream of Pax2 and posterior PPR genes. Pax2 and Lmx1a maintain their own expression in a positive feedback loop and that of their targets, while repressing alternative fates (Pax6, Otx2, Pax3). (C) Under the influence of canonical Wnt signaling and Notch activity, otic specific transcription factors are induced in OEPs close to the neural tube, while epibranchial placodes require continued FGF signaling and BMP activity promotes neurogenesis.

Cell fate decisions in the cranial sensory nervous system

151

9.2 Segregation of otic-epibranchial progenitors: Signals and regulatory circuits After their induction at early somite stages, OEPs remain widely scattered and segregate over time as morphological placodes appear (Steventon, Mayor, & Streit, 2016; Streit, 2002). Mesoderm-derived FGF not only induces OEPs, but also activates the expression of Wnt family members in the hindbrain (Ladher et al., 2000; Wright & Mansour, 2003). In turn, canonical Wnt signaling activates the next tier of otic transcription factors Nkx5.1, SOHo-1, Dlx5 and Sox2 in cells close to the neural tube promoting otic character, as well as promoting otic expression of Six1 and Eya1 by activating their enhancers (see below) (Fig. 4C) (Freter, Muta, Mak, Rinkwitz, & Ladher, 2008; Uchikawa, Ishida, Takemoto, Kamachi, & Kondoh, 2003). In contrast, cells far away from the neural tube receive low levels or no Wnt and develop into epibranchial placodes. Consequently, activation of canonical Wnt signaling in mouse leads to the expansion of otic markers (Dlx5, Pax2, Pax8) at the expense of Foxi2+ epidermis (Ohyama, Mohamed, Taketo, Dufort, & Groves, 2006). Interestingly, unlike FGF which induces Pax2, canonical Wnt signaling regulates its levels, which in turn has been proposed to control otic vs epibranchial fate (McCarroll et al., 2012; McCarroll & Nechiporuk, 2013). In zebrafish, cells with high levels of pax2a contribute to the otic placode, whereas cells with low pax2a levels turn into epibranchial cells. While otic placode formation clearly requires Wnts, activation of the pathway represses epibranchial fate (Freter et al., 2008). In contrast, continued FGF signaling is necessary for the specification of epibranchial placodes (Fig. 4C) (Freter et al., 2008; Sun et al., 2007). While it has been suggested that attenuation of FGF signaling is required for continued otic placode development, loss of the FGF antagonists Spry1 and -2 increases the otic domain at the expense of non-otic ectoderm (Freter et al., 2008; Mahoney Rogers, Zhang, & Shim, 2011). Therefore, the role of continued FGF signaling in the otic placode is controversial, but current observations suggests that fine tuning of this pathway is critical. Finally, Notch signaling has been implicated in otic placode development by enhancing Wnt activity (Fig. 4C) ( Jayasena, Ohyama, Segil, & Groves, 2008). In mouse, overactivation of Notch results in the expansion of the otic domain, while nonotic ectoderm and epibranchial territories are smaller. Conversely, inhibition of Notch signaling has the opposite effect.

152

Alexandre Thiery et al.

9.3 Regulatory circuits at otic and epibranchial placode stages As both placodes mature new transcription factors become expressed over time and are likely to work in a temporal hierarchy. Some OEP transcripts continue to be expressed in both otic and epibranchial placodes (e.g. Pax2, Zbtb16, Sox3 and the late factor Bach2), while others become confined to the epibranchial (e.g. Tfap2e, Klf7, Prdm1, Eya4) or otic territories (e.g. Lmx1a, Sox8, Sall4) (Fig. 4C) (Chen et al., 2017). The latter may be important during the segregation of both fates, however, their function is largely unexplored and the gene networks controlling epibranchial placode formation have not been investigated systematically. Interestingly, SoxE group factors like Sox8 and 10 are prominently expressed in the otic territory but absent in epibranchial placodes. Conversely, SoxB (Sox2, Sox3) group members are expressed in both as is Pax2. Sox proteins often work together with cofactors including Pax proteins to activate downstream target genes. It is therefore possible that different combinations of these family members play a role in otic vs epibranchial fate decisions (Kondoh & Kamachi, 2010). The OEP factors Etv4/5, Pax2 and Lmx1a seem to control some late otic transcripts (Chen et al., 2017). However, while these factors continue to be coexpressed in the otic domain it remains to be determined whether these interactions are direct or indirect. Few otic specific enhancers have been identified, whose analysis begins to shed light on the regulatory circuits (see below). Network interference approaches have predicted some small regulatory modules involving the transcription factors Hesx1 and Foxg1, which become expressed in the otic placode downstream of OEP module. Foxg1 is predicted to repress anterior placode marker and indeed when misexpressed downregulates the lens marker Six3 (Fig. 4C) (Anwar et al., 2017). Likewise, Hesx1 seems to act as a transcriptional repressor: overexpression leads to downregulation of Foxi3, Etv5 and Eya2 and, in agreement with these findings, the expression of these factors gradually decreases in the otic placode over time (Fig. 4C) (Anwar et al., 2017). At placode stages continued signaling events pattern the otic territory into subdomains that foreshadow different parts of the adult ear, while signaling to the epibranchial placodes mainly activates neurogenesis. While inhibition of BMP signaling results in loss of neurogenic markers like Phox2a, NeuroM and Hu, BMP7 is sufficient to upregulate Phox2a and neurofilament in non-neural cranial ectoderm (Begbie, Brunet, Rubenstein, & Graham, 1999; Kriebitz et al., 2009). Thus, in epibranchial placodes BMP signaling positively regulates neurogenesis. In contrast, in the otic placode

Cell fate decisions in the cranial sensory nervous system

153

Lmx1b appears to be downstream of BMP signaling and represses the neurogenic genes (Abello et al., 2010; Abello, Khatri, Giraldez, & Alsina, 2007). Thus, BMP signaling may differentially regulate neurogenesis in both placodes. In summary, as otic and epibranchial placodes become distinct new transcription factor are activated in a temporal hierarchy downstream of the OEP module. A reoccurring theme is that signaling events initiate feed forward loops to stabilize a regulatory state: once activated they promote their own expression and that of “like” genes, while repressing alternative states.

10. Refining the OEP network: A cis-regulatory perspective Despite recent efforts to unravel the gene regulatory network that controls OEP specification and segregation, mechanistic information on gene regulation and transcription factor interactions downstream of the signaling pathways is still poor. cis-Regulatory elements that control gene expression in time and space are key to provide mechanistic understanding of developmental as well as evolutionary processes (Davidson & Erwin, 2006; Levine, Cattoglio, & Tjian, 2014; Levine & Davidson, 2005). Acting as transcription factor hubs they integrate both signaling and transcriptional inputs. Below we discuss recent advances in enhancer identification in the context of OEP specification and diversification. Foxi3 is crucial for ear formation across vertebrates and is already expressed in posterior sensory progenitors (Birol et al., 2016; Khatri et al., 2014). ChIPseq for H3K27ac and H3K27me3 from FGF2-treated posterior PPR has recently identified two Foxi3 enhancer elements that are active at different time: Foxi3-E1 recapitulates early Foxi3 expression in ear progenitors, while Foxi3-E2 is active later, when Foxi3 becomes confined to the nonotic ectoderm (Tambalo et al., 2020). Foxi3-E1 contains AP1, Sall1, SoxD and -E, and Tead1 motifs and, as discussed above, is activated by FGF signaling, presumably through AP1 (Fig. 4A). In addition, SoxD motifs are required for its activity; with Sox13 being the only SoxD factor expressed it is likely that it regulates Foxi3 expression in OEP progenitors. Foxi3-E2 harbors Six1, Otx2 and Tcf4 motifs, however their requirement has not been investigated. It is possible that Tcf4 contributes to Foxi3 repression in the otic placode (Fig. 4C), while Six1 and Otx2 might activate it in nonotic placodal ectoderm.

154

Alexandre Thiery et al.

While Six1 and Eya1 are essential for sensory progenitor formation, they continue to be important as placodes mature and generate specialized cell types (Ishihara, Ikeda, et al., 2008). Several enhancers have been identified using evolutionary conservation across different species, which faithfully recapitulate the expression domains of Six1 and Eya1 in mouse and chick (Ishihara, Sato, Ikeda, Yajima, & Kawakami, 2008; Sato et al., 2012; Sato, Yajima, Furuta, Ikeda, & Kawakami, 2015). Eya1 enhancers CS1-3, CS1-5, CS1-23, CS1-26 are active in the otic placode and/or cranial ganglia from placode stages onwards. CS1-5 harbors several motifs including Tfap2a, Lef/Tcf, Sox, Foxi1 and Dlx3 and most of the corresponding transcription factors are expressed in the placode. Mutational analysis reveals that CS1-5 contains both positive and negative elements. Deletion of the Tcf/Lef binding site reduces enhancer activity, while mutation of the Sox motif increases activity in the head mesenchyme and neural crest cells suggesting that Sox factors are important to restrict enhancer activity to the otic placode. Interestingly, in the chick otic placode Lef1 and Eya1 begin to be expressed simultaneously indicating that Wnt signaling may activate Eya1 through Lef1 (Fig. 4C) (Chen et al., 2017). Likewise, eight Six1 enhancers were identified to be conserved across vertebrate species, of which four are active in both otic and epibranchial placodes (Sato et al., 2015). Sox, Fox and Six motifs are essential for mSix1-21 activity in the otic placode, while mutation of Pax and Lef/Tcf motifs results in a significant decrease of enhancer activity (Fig. 4C). Likewise, the Six1-8 enhancer harbors a number of transcription factor binding sites with Tcf/Lef, Smad, bHLH, POU, YY and nuclear hormone receptor (NR) motifs necessary for normal levels of its activity (Sato et al., 2015). Many of the corresponding transcription factors are indeed expressed in the otic placode at the time when the enhancer is activated; the requirement of Tcf/Lef and Smad motifs suggests that Six1 expression in the otic placode is controlled by both Wnt and BMP signaling (Chen et al., 2017). A number of Sox transcription factors are expressed in the otic and/or epibranchial placodes; while the SoxB group members Sox2 and -3 are found in both, the SoxE group members Sox8, -9 and -10 are restricted to the otic territory. Sox3 expression is differentially regulated: evolutionary conserved enhancers for Sox3-D1 and -D3 are active in the otic placode, while Sox3-U3 and -U8 control Sox3 expression in epibranchial placodes (Nishimura et al., 2012). Sox3-D3 contains a 68 bp core element sufficient to drive reporter activity in the otic placode and is controlled by Sall4 and SoxE-group factors (Okamoto et al., 2018). Likewise, Sox2 is temporally

Cell fate decisions in the cranial sensory nervous system

155

and spatially regulated by a cohort of enhancers. NOP1 is active in the otic placode from the 12-somite stage onwards (Uchikawa et al., 2003). Mutational analysis demonstrates that FGF-responsive, SoxB1/E, Lef/Tcf and Sall4 motifs are crucial for its activity with SoxB1/E and Sall4 cooperating. In contrast, Homeo-box, Fox, and Zeb/Snail/E2 box seem to mediate its repression (Sugahara, Fujimoto, Kondoh, & Uchikawa, 2018). The SoxE group member Sox10 begins to be expressed in the otic placode around the 10-somite stage, and two highly conserved regions recapitulate its expression pattern in chick (Betancur, Bronner-Fraser, & Sauka-Spengler, 2010; Cheng, Cheung, Abu-Elmagd, Orme, & Scotting, 2000). In the otic placode, Sox10-E2 requires input from Sox8, cMyb and Etv4 (Fig. 4C), while the same enhancer is active in neural crest cells where it is controlled by Sox9, Ets1 and cMyb (Betancur, SaukaSpengler, & Bronner, 2011; Murko & Bronner, 2017). Sall4 expression already begins in sensory progenitors, but it is then restricted to the otic placode (Barembaum & Bronner-Fraser, 2007, 2010). An enhancer element CR-F replicates its expression from placode stages onwards and contains Pax2, Etv4, and Ets motifs. Among these, Etv4 is required for enhancer activity, while mutation of the Pax2 binding site results in decreased activity, and both corresponding transcription factors can activate it ectopically (Barembaum & Bronner-Fraser, 2010). Taken together, otic-specific enhancer analysis highlights a crucial role for transcription factors of the Sox family and for Sall4, as well as for FGF, Wnt and Notch signaling pathways, while less is known for the gene networks that regulate epibranchial fates. Expanding this information will be crucial to provide mechanistic understanding on the gene regulatory level, which in turn will shed new light on how cells are committed to specific fates and how this may go awry in disease.

11. Conclusion In summary, placode progenitor specification is intimately linked with the induction of central nervous system and neural crest precursors. Initial events are shared by all three cell populations, but fates diverge under the influence of local signaling events which in turn induce different combinations of transcription factors. Neural plate border specifiers provide critical inputs for both placodal and neural crest cells, however, their combination determines the ultimate fate of NPB cells. Recent evidence not only points to unexpected transcriptional heterogeneity of neural plate border cells, but

156

Alexandre Thiery et al.

also suggests that cell fates remain flexible until at least the time of neural tube closure. To provide mechanistic understanding the identification of cisregulatory elements is crucial, however this approach is currently in its infancy in the context of placode development as illustrated here for oticepibranchial progenitors. Nevertheless, an emerging theme is that local signals activate transcriptional feedforward loops, where ‘like’ genes promote their own expression and that of their targets, while repressing determinants for alternative fates. Advances in single cell approaches for transcriptome and epigenome profiling from developing embryos will allow us to explore the entire repertoire of neural plate border and placodal cells. Functional experiments are essential for testing predicted interactions, although it becomes increasingly challenging to prioritize interactions for further testing with the availability of more sequencing data. Integrating this information into meaningful gene regulatory networks requires a combination of experimental and computational approaches to refine the number of predictions to be experimentally tested and to generate biologically meaningful insights.

References Abdelhak, S., Kalatzis, V., Heilig, R., Compain, S., Samson, D., Vincent, C., et al. (1997). A human homologue of the Drosophila eyes absent gene underlies branchio-oto-renal (BOR) syndrome and identifies a novel gene family. Nature Genetics, 15, 157–164. Abello, G., Khatri, S., Giraldez, F., & Alsina, B. (2007). Early regionalization of the otic placode and its regulation by the Notch signaling pathway. Mechanisms of Development, 124, 631–645. Abello, G., Khatri, S., Radosevic, M., Scotting, P. J., Giraldez, F., & Alsina, B. (2010). Independent regulation of Sox3 and Lmx1b by FGF and BMP signaling influences the neurogenic and non-neurogenic domains in the chick otic placode. Developmental Biology, 339, 166–178. Acampora, D., Gulisano, M., Broccoli, V., & Simeone, A. (2001). Otx genes in brain morphogenesis. Progress in Neurobiology, 64, 69–95. Acampora, D., Mazan, S., Lallemand, Y., Avantaggiato, V., Maury, M., Simeone, A., et al. (1995). Forebrain and midbrain regions are deleted in Otx2-/- mutants due to a defective anterior neuroectoderm specification during gastrulation. Development, 121, 3279–3290. Ahrens, K., & Schlosser, G. (2005). Tissues and signals involved in the induction of placodal Six1 expression in Xenopus laevis. Developmental Biology, 288, 40–59. Albazerchi, A., & Stern, C. D. (2007). A role for the hypoblast (AVE) in the initiation of neural induction, independent of its ability to position the primitive streak. Developmental Biology, 301, 489–503. Anwar, M., Tambalo, M., Ranganathan, R., Grocott, T., & Streit, A. (2017). A gene network regulated by FGF signalling during ear development. Scientific Reports, 7, 6162. Aybar, M. J., & Mayor, R. (2002). Early induction of neural crest cells: Lessons learned from frog, fish and chick. Current Opinion in Genetics & Development, 12, 452–458. Bailey, A. P., Bhattacharyya, S., Bronner-Fraser, M., & Streit, A. (2006). Lens specification is the ground state of all sensory placodes, from which FGF promotes olfactory identity. Developmental Cell, 11, 505–517.

Cell fate decisions in the cranial sensory nervous system

157

Baker, C. V., & Bronner-Fraser, M. (1997). The origins of the neural crest. Part I: Embryonic induction. Mechanisms of Development, 69, 3–11. Baker, C. V., & Bronner-Fraser, M. (2001). Vertebrate cranial placodes I. Embryonic induction. Developmental Biology, 232, 1–61. Baker, C. V., Stark, M. R., Marcelle, C., & Bronner-Fraser, M. (1999). Competence, specification and induction of Pax-3 in the trigeminal placode. Development, 126, 147–156. Bally-Cuif, L., Gulisano, M., Broccoli, V., & Boncinelli, E. (1995). c-otx2 is expressed in two different phases of gastrulation and is sensitive to retinoic acid treatment in chick embryo. Mechanisms of Development, 49, 49–63. Baranski, M., Berdougo, E., Sandler, J. S., Darnell, D. K., & Burrus, L. W. (2000). The dynamic expression pattern of frzb-1 suggests multiple roles in chick development. Developmental Biology, 217, 25–41. Barembaum, M., & Bronner-Fraser, M. (2007). Spalt4 mediates invagination and otic placode gene expression in cranial ectoderm. Development, 134, 3805–3814. Barembaum, M., & Bronner-Fraser, M. (2010). Pax2 and Pea3 synergize to activate a novel regulatory enhancer for spalt4 in the developing ear. Developmental Biology, 340, 222–231. Barrallo-Gimeno, A., Holzschuh, J., Driever, W., & Knapik, E. W. (2004). Neural crest survival and differentiation in zebrafish depends on mont blanc/tfap2a gene function. Development, 131, 1463–1477. Basch, M. L., Bronner-Fraser, M., & Garcia-Castro, M. I. (2006). Specification of the neural crest occurs during gastrulation and requires Pax7. Nature, 441, 218–222. Beanan, M. J., Feledy, J. A., & Sargent, T. D. (2000). Regulation of early expression of Dlx3, a Xenopus anti-neural factor, by beta-catenin signaling. Mechanisms of Development, 91, 227–235. Begbie, J., Brunet, J. F., Rubenstein, J. L., & Graham, A. (1999). Induction of the epibranchial placodes. Development, 126, 895–902. Betancur, P., Bronner-Fraser, M., & Sauka-Spengler, T. (2010). Genomic code for Sox10 activation reveals a key regulatory enhancer for cranial neural crest. Proceedings of the National Academy of Sciences of the United States of America, 107, 3570–3575. Betancur, P., Sauka-Spengler, T., & Bronner, M. (2011). A Sox10 enhancer element common to the otic placode and neural crest is activated by tissue-specific paralogs. Development, 138, 3689–3698. Bhat, N., Kwon, H. J., & Riley, B. B. (2013). A gene network that coordinates preplacodal competence and neural crest specification in zebrafish. Developmental Biology, 373, 107–117. Bhat, N., & Riley, B. B. (2011). Integrin-alpha5 coordinates assembly of posterior cranial placodes in zebrafish and enhances Fgf-dependent regulation of otic/epibranchial cells. PLoS One, 6, e27778. Bhattacharyya, S., Bailey, A. P., Bronner-Fraser, M., & Streit, A. (2004). Segregation of lens and olfactory precursors from a common territory: Cell sorting and reciprocity of Dlx5 and Pax6 expression. Developmental Biology, 271, 403–414. Bhattacharyya, S., & Bronner-Fraser, M. (2008). Competence, specification and commitment to an olfactory placode fate. Development, 135, 4165–4177. Birol, O., Ohyama, T., Edlund, R. K., Drakou, K., Georgiades, P., & Groves, A. K. (2016). The mouse Foxi3 transcription factor is necessary for the development of posterior placodes. Developmental Biology, 409, 139–151. Broccoli, V., Boncinelli, E., & Wurst, W. (1999). The caudal limit of Otx2 expression positions the isthmic organizer. Nature, 401, 164–168. Brugmann, S. A., & Moody, S. A. (2005). Induction and specification of the vertebrate ectodermal placodes: Precursors of the cranial sensory organs. Biology of the Cell, 97, 303–319.

158

Alexandre Thiery et al.

Brugmann, S. A., Pandur, P. D., Kenyon, K. L., Pignoni, F., & Moody, S. A. (2004). Six1 promotes a placodal fate within the lateral neurogenic ectoderm by functioning as both a transcriptional activator and repressor. Development, 131, 5871–5881. Buitrago-Delgado, E., Nordin, K., Rao, A., Geary, L., & LaBonne, C. (2015). Shared regulatory programs suggest retention of blastula-stage potential in neural crest cells. Science, 348, 1332–1335. Burton, Q., Cole, L. K., Mulheisen, M., Chang, W., & Wu, D. K. (2004). The role of Pax2 in mouse inner ear development. Developmental Biology, 272, 161–175. Chang, C., & Hemmati-Brivanlou, A. (1998). Neural crest induction by Xwnt7B in Xenopus. Developmental Biology, 194, 129–134. Chapman, S. C., Brown, R., Lees, L., Schoenwolf, G. C., & Lumsden, A. (2004). Expression analysis of chick Wnt and frizzled genes and selected inhibitors in early chick patterning. Developmental Dynamics, 229, 668–676. Chapman, S. C., Schubert, F. R., Schoenwolf, G. C., & Lumsden, A. (2002). Analysis of spatial and temporal gene expression patterns in blastula and gastrula stage chick embryos. Developmental Biology, 245, 187–199. Chen, B., Kim, E. H., & Xu, P. X. (2009). Initiation of olfactory placode development and neurogenesis is blocked in mice lacking both Six1 and Six4. Developmental Biology, 326, 75–85. Chen, J., & Streit, A. (2013). Induction of the inner ear: stepwise specification of otic fate from multipotent progenitors. Hearing Research, 297, 3–12. Chen, J., Tambalo, M., Barembaum, M., Ranganathan, R., Simoes-Costa, M., Bronner, M. E., et al. (2017). A systems-level approach reveals new gene regulatory modules in the developing ear. Development, 144, 1531–1543. Cheng, Y., Cheung, M., Abu-Elmagd, M. M., Orme, A., & Scotting, P. J. (2000). Chick sox10, a transcription factor expressed in both early neural crest cells and central nervous system. Brain Research. Developmental Brain Research, 121, 233–241. Christophorou, N. A., Bailey, A. P., Hanson, S., & Streit, A. (2009). Activation of Six1 target genes is required for sensory placode formation. Developmental Biology, 336, 327–336. Christophorou, N. A., Mende, M., Lleras-Forero, L., Grocott, T., & Streit, A. (2010). Pax2 coordinates epithelial morphogenesis and cell fate in the inner ear. Developmental Biology, 345, 180–190. Davidson, E. H., & Erwin, D. H. (2006). Gene regulatory networks and the evolution of animal body plans. Science, 311, 796–800. de Croze, N., Maczkowiak, F., & Monsoro-Burq, A. H. (2011). Reiterative AP2a activity controls sequential steps in the neural crest gene regulatory network. Proceedings of the National Academy of Sciences of the United States of America, 108, 155–160. Delaune, E., Lemaire, P., & Kodjabachian, L. (2005). Neural induction in Xenopus requires early FGF signalling in addition to BMP inhibition. Development, 132, 299–310. Dickinson, M. E., Selleck, M. A., McMahon, A. P., & Bronner-Fraser, M. (1995). Dorsalization of the neural tube by the non-neural ectoderm. Development, 121, 2099–2106. Dude, C. M., Kuan, C. Y., Bradshaw, J. R., Greene, N. D., Relaix, F., Stark, M. R., et al. (2009). Activation of Pax3 target genes is necessary but not sufficient for neurogenesis in the ophthalmic trigeminal placode. Developmental Biology, 326, 314–326. Dutta, S., Dietrich, J. E., Aspock, G., Burdine, R. D., Schier, A., Westerfield, M., et al. (2005). pitx3 defines an equivalence domain for lens and anterior pituitary placode. Development, 132, 1579–1590. Esterberg, R., & Fritz, A. (2009). dlx3b/4b are required for the formation of the preplacodal region and otic placode through local modulation of BMP activity. Developmental Biology, 325, 189–199.

Cell fate decisions in the cranial sensory nervous system

159

Ezin, A. M., Fraser, S. E., & Bronner-Fraser, M. (2009). Fate map and morphogenesis of presumptive neural crest and dorsal neural tube. Developmental Biology, 330, 221–236. Fainsod, A., Steinbeisser, H., & De Robertis, E. M. (1994). On the function of BMP-4 in patterning the marginal zone of the Xenopus embryo. The EMBO Journal, 13, 5015–5025. Favor, J., Sandulache, R., Neuhauser-Klaus, A., Pretsch, W., Chatterjee, B., Senft, E., et al. (1996). The mouse Pax2(1Neu) mutation is identical to a human PAX2 mutation in a family with renal-coloboma syndrome and results in developmental defects of the brain, ear, eye, and kidney. Proceedings of the National Academy of Sciences of the United States of America, 93, 13870–13875. Fernandez-Garre, P., Rodriguez-Gallardo, L., Gallego-Diaz, V., Alvarez, I. S., & Puelles, L. (2002). Fate map of the chicken neural plate at stage 4. Development, 129, 2807–2822. Freter, S., Muta, Y., Mak, S. S., Rinkwitz, S., & Ladher, R. K. (2008). Progressive restriction of otic fate: The role of FGF and Wnt in resolving inner ear potential. Development, 135, 3415–3424. Freter, S., Muta, Y., O’Neill, P., Vassilev, V. S., Kuraku, S., & Ladher, R. K. (2012). Pax2 modulates proliferation during specification of the otic and epibranchial placodes. Developmental Dynamics, 241, 1716–1728. Friedman, R. A., Makmura, L., Biesiada, E., Wang, X., & Keithley, E. M. (2005). Eya1 acts upstream of Tbx1, Neurogenin 1, NeuroD and the neurotrophins BDNF and NT-3 during inner ear development. Mechanisms of Development, 122, 625–634. Gallagher, B. C., Henry, J. J., & Grainger, R. M. (1996). Inductive processes leading to inner ear formation during Xenopus development. Developmental Biology, 175, 95–107. Garcia-Castro, M. I., Marcelle, C., & Bronner-Fraser, M. (2002). Ectodermal Wnt function as a neural crest inducer. Science, 297, 848–851. Garcia-Martinez, V., Alvarez, I. S., & Schoenwolf, G. C. (1993). Locations of the ectodermal and nonectodermal subdivisions of the epiblast at stages 3 and 4 of avian gastrulation and neurulation. The Journal of Experimental Zoology, 267, 431–446. Garnett, A. T., Square, T. A., & Medeiros, D. M. (2012). BMP, Wnt and FGF signals are integrated through evolutionarily conserved enhancers to achieve robust expression of Pax3 and Zic genes at the zebrafish neural plate border. Development, 139, 4220–4231. Glavic, A., Gomez-Skarmeta, J. L., & Mayor, R. (2002). The homeoprotein Xiro1 is required for midbrain-hindbrain boundary formation. Development, 129, 1609–1621. Glavic, A., Maris Honore, S., Gloria Feijoo, C., Bastidas, F., Allende, M. L., & Mayor, R. (2004). Role of BMP signaling and the homeoprotein Iroquois in the specification of the cranial placodal field. Developmental Biology, 272, 89–103. Grocott, T., Tambalo, M., & Streit, A. (2012). The peripheral sensory nervous system in the vertebrate head: A gene regulatory perspective. Developmental Biology, 370, 3–23. Groves, A. K., & Bronner-Fraser, M. (2000). Competence, specification and commitment in otic placode induction. Development, 127, 3489–3499. Gutkovich, Y. E., Ofir, R., Elkouby, Y. M., Dibner, C., Gefen, A., Elias, S., et al. (2010). Xenopus Meis3 protein lies at a nexus downstream to Zic1 and Pax3 proteins, regulating multiple cell-fates during early nervous system development. Developmental Biology, 338, 50–62. Hans, S., Liu, D., & Westerfield, M. (2004). Pax8 and Pax2a function synergistically in otic specification, downstream of the Foxi1 and Dlx3b transcription factors. Development, 131, 5091–5102. Hanson, I. M. (2001). Mammalian homologues of the Drosophila eye specification genes. Seminars in Cell & Developmental Biology, 12, 475–484. Hatada, Y., & Stern, C. D. (1994). A fate map of the epiblast of the early chick embryo. Development, 120, 2879–2889.

160

Alexandre Thiery et al.

Henry, J. J., & Grainger, R. M. (1990). Early tissue interactions leading to embryonic lens formation in Xenopus laevis. Developmental Biology, 141, 149–163. Hidalgo-Sanchez, M., Alvarado-Mallart, R., & Alvarez, I. S. (2000). Pax2, Otx2, Gbx2 and Fgf8 expression in early otic vesicle development. Mechanisms of Development, 95, 225–229. Hidalgo-Sanchez, M., Millet, S., Bloch-Gallego, E., & Alvarado-Mallart, R. M. (2005). Specification of the meso-isthmo-cerebellar region: The Otx2/Gbx2 boundary. Brain Research. Brain Research Reviews, 49, 134–149. Hintze, M., Prajapati, R. S., Tambalo, M., Christophorou, N. A. D., Anwar, M., Grocott, T., et al. (2017). Cell interactions, signals and transcriptional hierarchy governing placode progenitor induction. Development, 144, 2810–2823. Hoffman, T. L., Javier, A. L., Campeau, S. A., Knight, R. D., & Schilling, T. F. (2007). Tfap2 transcription factors in zebrafish neural crest development and ectodermal evolution. Journal of Experimental Zoology, Part B, Molecular and Developmental Evolution, 308, 679–691. Hong, C. S., & Saint-Jeannet, J. P. (2007). The activity of Pax3 and Zic1 regulates three distinct cell fates at the neural plate border. Molecular Biology of the Cell, 18, 2192–2202. Ishihara, T., Ikeda, K., Sato, S., Yajima, H., & Kawakami, K. (2008). Differential expression of Eya1 and Eya2 during chick early embryonic development. Gene Expression Patterns, 8, 357–367. Ishihara, T., Sato, S., Ikeda, K., Yajima, H., & Kawakami, K. (2008). Multiple evolutionarily conserved enhancers control expression of Eya1. Developmental Dynamics, 237, 3142–3156. Jacobson, A. G. (1963). The determination and positioning of the nose, lens and ear. III. Effects of reversing the antero-posterior axis of epidermis, neural plate and neural fold. The Journal of Experimental Zoology, 154, 293–303. Jayasena, C. S., Ohyama, T., Segil, N., & Groves, A. K. (2008). Notch signaling augments the canonical Wnt pathway to specify the size of the otic placode. Development, 135, 2251–2261. Johnson, K. R., Cook, S. A., Erway, L. C., Matthews, A. N., Sanford, L. P., Paradies, N. E., et al. (1999). Inner ear and kidney anomalies caused by IAP insertion in an intron of the Eya1 gene in a mouse model of BOR syndrome. Human Molecular Genetics, 8, 645–653. Joyner, A. L., Liu, A., & Millet, S. (2000). Otx2, Gbx2 and Fgf8 interact to position and maintain a mid-hindbrain organizer. Current Opinion in Cell Biology, 12, 736–741. Katahira, T., Sato, T., Sugiyama, S., Okafuji, T., Araki, I., Funahashi, J., et al. (2000). Interaction between Otx2 and Gbx2 defines the organizing center for the optic tectum. Mechanisms of Development, 91, 43–52. Kawakami, K., Sato, S., Ozaki, H., & Ikeda, K. (2000). Six family genes—Structure and function as transcription factors and their roles in development. BioEssays, 22, 616–626. Khatri, S. B., Edlund, R. K., & Groves, A. K. (2014). Foxi3 is necessary for the induction of the chick otic placode in response to FGF signaling. Developmental Biology, 391, 158–169. Khatri, S. B., & Groves, A. K. (2013). Expression of the Foxi2 and Foxi3 transcription factors during development of chicken sensory placodes and pharyngeal arches. Gene Expression Patterns, 13, 38–42. Khudyakov, J., & Bronner-Fraser, M. (2009). Comprehensive spatiotemporal analysis of early chick neural crest network genes. Developmental Dynamics, 238, 716–723. Kobayashi, M., Osanai, H., Kawakami, K., & Yamamoto, M. (2000). Expression of three zebrafish Six4 genes in the cranial sensory placodes and the developing somites. Mechanisms of Development, 98, 151–155.

Cell fate decisions in the cranial sensory nervous system

161

Kondoh, H., & Kamachi, Y. (2010). SOX-partner code for cell specification: Regulatory target selection and underlying molecular mechanisms. The International Journal of Biochemistry & Cell Biology, 42, 391–399. Konishi, Y., Ikeda, K., Iwakura, Y., & Kawakami, K. (2006). Six1 and Six4 promote survival of sensory neurons during early trigeminal gangliogenesis. Brain Research, 1116, 93–102. Kozlowski, D. J., Murakami, T., Ho, R. K., & Weinberg, E. S. (1997). Regional cell movement and tissue patterning in the zebrafish embryo revealed by fate mapping with caged fluorescein. Biochemistry and Cell Biology, 75, 551–562. Kozlowski, D. J., Whitfield, T. T., Hukriede, N. A., Lam, W. K., & Weinberg, E. S. (2005). The zebrafish dog-eared mutation disrupts eya1, a gene required for cell survival and differentiation in the inner ear and lateral line. Developmental Biology, 277, 27–41. Kriebitz, N. N., Kiecker, C., McCormick, L., Lumsden, A., Graham, A., & Bell, E. (2009). PRDC regulates placode neurogenesis in chick by modulating BMP signalling. Developmental Biology, 336, 280–292. Kwon, H. J., Bhat, N., Sweet, E. M., Cornell, R. A., & Riley, B. B. (2010). Identification of early requirements for preplacodal ectoderm and sensory organ development. PLoS Genetics, 6, e1001133. LaBonne, C., & Bronner-Fraser, M. (1998). Neural crest induction in Xenopus: Evidence for a two-signal model. Development, 125, 2403–2414. Laclef, C., Hamard, G., Demignon, J., Souil, E., Houbron, C., & Maire, P. (2003). Altered myogenesis in Six1-deficient mice. Development, 130, 2239–2252. Ladher, R. K., Anakwe, K. U., Gurney, A. L., Schoenwolf, G. C., & Francis-West, P. H. (2000). Identification of synergistic signals initiating inner ear development. Science, 290, 1965–1968. Levine, M., Cattoglio, C., & Tjian, R. (2014). Looping back to leap forward: Transcription enters a new era. Cell, 157, 13–25. Levine, M., & Davidson, E. H. (2005). Gene regulatory networks for development. Proceedings of the National Academy of Sciences of the United States of America, 102, 4936–4942. Li, J. Y., & Joyner, A. L. (2001). Otx2 and Gbx2 are required for refinement and not induction of mid-hindbrain gene expression. Development, 128, 4979–4991. Li, B., Kuriyama, S., Moreno, M., & Mayor, R. (2009). The posteriorizing gene Gbx2 is a direct target of Wnt signalling and the earliest factor in neural crest induction. Development, 136, 3267–3278. Li, X., Oghi, K. A., Zhang, J., Krones, A., Bush, K. T., Glass, C. K., et al. (2003). Eya protein phosphatase activity regulates Six1-Dach-Eya transcriptional effects in mammalian organogenesis. Nature, 426, 247–254. Li, Y., Zhao, D., Horie, T., Chen, G., Bao, H., Chen, S., et al. (2017). Conserved gene regulatory module specifies lateral neural borders across bilaterians. Proceedings of the National Academy of Sciences of the United States of America, 114, E6352–E6360. Liem, K. F., Jr., Tremml, G., Roelink, H., & Jessell, T. M. (1995). Dorsal differentiation of neural plate cells induced by BMP-mediated signals from epidermal ectoderm. Cell, 82, 969–979. Linker, C., De Almeida, I., Papanayotou, C., Stower, M., Sabado, V., Ghorani, E., et al. (2009). Cell communication with the neural plate is required for induction of neural markers by BMP inhibition: evidence for homeogenetic induction and implications for Xenopus animal cap and chick explant assays. Developmental Biology, 327, 478–486. Linker, C., & Stern, C. D. (2004). Neural induction requires BMP inhibition only as a late step, and involves signals other than FGF and Wnt antagonists. Development, 131, 5671–5681.

162

Alexandre Thiery et al.

Litsiou, A., Hanson, S., & Streit, A. (2005). A balance of FGF, BMP and WNT signalling positions the future placode territory in the head. Development, 132, 4051–4062. Luo, T., Lee, Y. H., Saint-Jeannet, J. P., & Sargent, T. D. (2003). Induction of neural crest in Xenopus by transcription factor AP2alpha. Proceedings of the National Academy of Sciences of the United States of America, 100, 532–537. Maharana, S. K., Pollet, N., & Schlosser, G. (2017). Identification of novel cis-regulatory elements of Eya1 in Xenopus laevis using BAC recombineering. Scientific Reports, 7, 15033. Maharana, S. K., & Schlosser, G. (2018). A gene regulatory network underlying the formation of pre-placodal ectoderm in Xenopus laevis. BMC Biology, 16, 79. Mahoney Rogers, A. A., Zhang, J., & Shim, K. (2011). Sprouty1 and Sprouty2 limit both the size of the otic placode and hindbrain Wnt8a by antagonizing FGF signaling. Developmental Biology, 353, 94–104. Marchal, L., Luxardi, G., Thome, V., & Kodjabachian, L. (2009). BMP inhibition initiates neural induction via FGF signaling and Zic genes. Proceedings of the National Academy of Sciences of the United States of America, 106, 17437–17442. Marchant, L., Linker, C., Ruiz, P., Guerrero, N., & Mayor, R. (1998). The inductive properties of mesoderm suggest that the neural crest cells are specified by a BMP gradient. Developmental Biology, 198, 319–329. Martin, K., & Groves, A. K. (2006). Competence of cranial ectoderm to respond to Fgf signaling suggests a two-step model of otic placode induction. Development, 133, 877–887. Matsuo-Takasaki, M., Matsumura, M., & Sasai, Y. (2005). An essential role of Xenopus Foxi1a for ventral specification of the cephalic ectoderm during gastrulation. Development, 132, 3885–3894. McCabe, K. L., & Bronner-Fraser, M. (2009). Molecular and tissue interactions governing induction of cranial ectodermal placodes. Developmental Biology, 332, 189–195. McCarroll, M. N., Lewis, Z. R., Culbertson, M. D., Martin, B. L., Kimelman, D., & Nechiporuk, A. V. (2012). Graded levels of Pax2a and Pax8 regulate cell differentiation during sensory placode formation. Development, 139, 2740–2750. McCarroll, M. N., & Nechiporuk, A. V. (2013). Fgf3 and Fgf10a work in concert to promote maturation of the epibranchial placodes in zebrafish. PLoS One, 8, e85087. McLarren, K. W., Litsiou, A., & Streit, A. (2003). DLX5 positions the neural crest and preplacode region at the border of the neural plate. Developmental Biology, 259, 34–47. Meulemans, D., & Bronner-Fraser, M. (2004). Gene-regulatory interactions in neural crest evolution and development. Developmental Cell, 7, 291–299. Millet, S., Campbell, K., Epstein, D. J., Losos, K., Harris, E., & Joyner, A. L. (1999). A role for Gbx2 in repression of Otx2 and positioning the mid/hindbrain organizer. Nature, 401, 161–164. Monsoro-Burq, A. H., Fletcher, R. B., & Harland, R. M. (2003). Neural crest induction by paraxial mesoderm in Xenopus embryos requires FGF signals. Development, 130, 3111–3124. Monsoro-Burq, A. H., Wang, E., & Harland, R. (2005). Msx1 and Pax3 cooperate to mediate FGF8 and WNT signals during Xenopus neural crest induction. Developmental Cell, 8, 167–178. Moody, S. A., & LaMantia, A. S. (2015). Transcriptional regulation of cranial sensory placode development. Current Topics in Developmental Biology, 111, 301–350. Moury, J. D., & Jacobson, A. G. (1989). Neural fold formation at newly created boundaries between neural plate and epidermis in the axolotl. Developmental Biology, 133, 44–57. Murko, C., & Bronner, M. E. (2017). Tissue specific regulation of the chick Sox10E1 enhancer by different Sox family members. Developmental Biology, 422, 47–57.

Cell fate decisions in the cranial sensory nervous system

163

Nichane, M., de Croze, N., Ren, X., Souopgui, J., Monsoro-Burq, A. H., & Bellefroid, E. J. (2008). Hairy2-Id3 interactions play an essential role in Xenopus neural crest progenitor specification. Developmental Biology, 322, 355–367. Nishimura, N., Kamimura, Y., Ishida, Y., Takemoto, T., Kondoh, H., & Uchikawa, M. (2012). A systematic survey and characterization of enhancers that regulate Sox3 in neuro-sensory development in comparison with Sox2 enhancers. Biology (Basel), 1, 714–735. Niss, K., & Leutz, A. (1998). Expression of the homeobox gene GBX2 during chicken development. Mechanisms of Development, 76, 151–155. Nissen, R. M., Yan, J., Amsterdam, A., Hopkins, N., & Burgess, S. M. (2003). Zebrafish foxi one modulates cellular responses to Fgf signaling required for the integrity of ear and jaw patterning. Development, 130, 2543–2554. Ogita, J., Isogai, E., Sudo, H., Sakiyama, S., Nakagawara, A., & Koseki, H. (2001). Expression of the Dan gene during chicken embryonic development. Mechanisms of Development, 109, 363–365. Ohuchi, H., Kimura, S., Watamoto, M., & Itoh, N. (2000). Involvement of fibroblast growth factor (FGF)18-FGF8 signaling in specification of left-right asymmetry and brain and limb development of the chick embryo. Mechanisms of Development, 95, 55–66. Ohyama, T., & Groves, A. K. (2004). Expression of mouse Foxi class genes in early craniofacial development. Developmental Dynamics, 231, 640–646. Ohyama, T., Mohamed, O. A., Taketo, M. M., Dufort, D., & Groves, A. K. (2006). Wnt signals mediate a fate decision between otic placode and epidermis. Development, 133, 865–875. Okamoto, Y., Nishimura, N., Matsuda, K., Ranawakage, D. C., Kamachi, Y., Kondoh, H., et al. (2018). Cooperation of Sall4 and Sox8 transcription factors in the regulation of the chicken Sox3 gene during otic placode development. Development, Growth & Differentiation, 60, 133–145. Ozaki, H., Nakamura, K., Funahashi, J., Ikeda, K., Yamada, G., Tokano, H., et al. (2004). Six1 controls patterning of the mouse otic vesicle. Development, 131, 551–562. Pandur, P. D., & Moody, S. A. (2000). Xenopus Six1 gene is expressed in neurogenic cranial placodes and maintained in the differentiating lateral lines. Mechanisms of Development, 96, 253–257. Papanayotou, C., Mey, A., Birot, A. M., Saka, Y., Boast, S., Smith, J. C., et al. (2008). A mechanism regulating the onset of Sox2 expression in the embryonic neural plate. PLoS Biology, 6, e2. Park, B. Y., & Saint-Jeannet, J. P. (2010). Induction and segregation of the vertebrate cranial placodes. San Rafael (CA): Morgan & Claypool Life Sciences. Patthey, C., Edlund, T., & Gunhaga, L. (2009). Wnt-regulated temporal control of BMP exposure directs the choice between neural plate border and epidermal fate. Development, 136, 73–83. Patthey, C., Gunhaga, L., & Edlund, T. (2008). Early development of the central and peripheral nervous systems is coordinated by Wnt and BMP signals. PLoS One, 3, e1625. Patthey, C., Schlosser, G., & Shimeld, S. M. (2014). The evolutionary history of vertebrate cranial placodes—I: Cell type evolution. Developmental Biology, 389, 82–97. Pera, E., Stein, S., & Kessel, M. (1999). Ectodermal patterning in the avian embryo: epidermis versus neural plate. Development, 126, 63–73. Phillips, B. T., Bolding, K., & Riley, B. B. (2001). Zebrafish fgf3 and fgf8 encode redundant functions required for otic placode induction. Developmental Biology, 235, 351–365. Piacentino, M. L., & Bronner, M. E. (2018). Intracellular attenuation of BMP signaling via CKIP-1/Smurf1 is essential during neural crest induction. PLoS Biology, 16, e2004425.

164

Alexandre Thiery et al.

Pieper, M., Ahrens, K., Rink, E., Peter, A., & Schlosser, G. (2012). Differential distribution of competence for panplacodal and neural crest induction to non-neural and neural ectoderm. Development, 139, 1175–1187. Pieper, M., Eagleson, G. W., Wosniok, W., & Schlosser, G. (2011). Origin and segregation of cranial placodes in Xenopus laevis. Developmental Biology, 360, 257–275. Pignoni, F., Hu, B., Zavitz, K. H., Xiao, J., Garrity, P. A., & Zipursky, S. L. (1997). The eye-specification proteins So and Eya form a complex and regulate multiple steps in Drosophila eye development. Cell, 91, 881–891. Pla, P., & Monsoro-Burq, A. H. (2018). The neural border: Induction, specification and maturation of the territory that generates neural crest cells. Developmental Biology, 444(Suppl. 1), S36–S46. Plouhinec, J. L., Roche, D. D., Pegoraro, C., Figueiredo, A. L., Maczkowiak, F., Brunet, L. J., et al. (2014). Pax3 and Zic1 trigger the early neural crest gene regulatory network by the direct activation of multiple key neural crest specifiers. Developmental Biology, 386, 461–472. Prajapati, R. S., Hintze, M., & Streit, A. (2019). PRDM1 controls the sequential activation of neural, neural crest and sensory progenitor determinants. Development, 146, 1–11. Prasad, M. S., Charney, R. M., & Garcia-Castro, M. I. (2019). Specification and formation of the neural crest: Perspectives on lineage segregation. Genesis, 57, e23276. Reichert, S., Randall, R. A., & Hill, C. S. (2013). A BMP regulatory network controls ectodermal cell fate decisions at the neural plate border. Development, 140, 4435–4444. Rex, M., Orme, A., Uwanogho, D., Tointon, K., Wigmore, P. M., Sharpe, P. T., et al. (1997). Dynamic expression of chicken Sox2 and Sox3 genes in ectoderm induced to form neural tissue. Developmental Dynamics, 209, 323–332. Rodriguez Esteban, C., Capdevila, J., Economides, A. N., Pascual, J., Ortiz, A., & Izpisua Belmonte, J. C. (1999). The novel Cer-like protein Caronte mediates the establishment of embryonic left-right asymmetry. Nature, 401, 243–251. Roellig, D., Tan-Cabugao, J., Esaian, S., & Bronner, M. E. (2017). Dynamic transcriptional signature and cell fate analysis reveals plasticity of individual neural plate border cells. eLife, 6, e21620. Ruf, R. G., Xu, P. X., Silvius, D., Otto, E. A., Beekmann, F., Muerb, U. T., et al. (2004). SIX1 mutations cause branchio-oto-renal syndrome by disruption of EYA1-SIX1-DNA complexes. Proceedings of the National Academy of Sciences of the United States of America, 101, 8090–8095. Sahly, I., Andermann, P., & Petit, C. (1999). The zebrafish eya1 gene and its expression pattern during embryogenesis. Development Genes and Evolution, 209, 399–410. Sai, X., & Ladher, R. K. (2015). Early steps in inner ear development: Induction and morphogenesis of the otic placode. Frontiers in Pharmacology, 6, 19. Saint-Jeannet, J. P., & Moody, S. A. (2014). Establishing the pre-placodal region and breaking it into placodes with distinct identities. Developmental Biology, 389, 13–27. Sasai, Y., Lu, B., Piccolo, S., & De Robertis, E. M. (1996). Endoderm induction by the organizer-secreted factors chordin and noggin in Xenopus animal caps. The EMBO Journal, 15, 4547–4555. Sato, S., Ikeda, K., Shioi, G., Nakao, K., Yajima, H., & Kawakami, K. (2012). Regulation of Six1 expression by evolutionarily conserved enhancers in tetrapods. Developmental Biology, 368, 95–108. Sato, S., Ikeda, K., Shioi, G., Ochi, H., Ogino, H., Yajima, H., et al. (2010). Conserved expression of mouse Six1 in the pre-placodal region (PPR) and identification of an enhancer for the rostral PPR. Developmental Biology, 344, 158–171. Sato, T., Sasai, N., & Sasai, Y. (2005). Neural crest determination by co-activation of Pax3 and Zic1 genes in Xenopus ectoderm. Development, 132, 2355–2363.

Cell fate decisions in the cranial sensory nervous system

165

Sato, S., Yajima, H., Furuta, Y., Ikeda, K., & Kawakami, K. (2015). Activation of Six1 expression in vertebrate sensory neurons. PLoS One, 10, e0136666. Schlosser, G. (2006). Induction and specification of cranial placodes. Developmental Biology, 294, 303–351. Schlosser, G. (2010). Making senses development of vertebrate cranial placodes. International Review of Cell and Molecular Biology, 283, 129–234. Schlosser, G. (2014). Early embryonic specification of vertebrate cranial placodes. Wiley Interdisciplinary Reviews: Developmental Biology, 3, 349–363. Schlosser, G., & Ahrens, K. (2004). Molecular anatomy of placode development in Xenopus laevis. Developmental Biology, 271, 439–466. Schonberger, J., Wang, L., Shin, J. T., Kim, S. D., Depreux, F. F., Zhu, H., et al. (2005). Mutation in the transcriptional coactivator EYA4 causes dilated cardiomyopathy and sensorineural hearing loss. Nature Genetics, 37, 418–422. Schumacher, J. A., Hashiguchi, M., Nguyen, V. H., & Mullins, M. C. (2011). An intermediate level of BMP signaling directly specifies cranial neural crest progenitor cells in zebrafish. PLoS One, 6, e27403. Selleck, M. A., & Bronner-Fraser, M. (1995). Origins of the avian neural crest: The role of neural plate-epidermal interactions. Development, 121, 525–538. Shamim, H., & Mason, I. (1999). Expression of Fgf4 during early development of the chick embryo. Mechanisms of Development, 85, 189–192. Sheng, G., & Stern, C. D. (1999). Gata2 and Gata3: Novel markers for early embryonic polarity and for non-neural ectoderm in the chick embryo. Mechanisms of Development, 87, 213–216. Simeone, A., Acampora, D., Gulisano, M., Stornaiuolo, A., & Boncinelli, E. (1992). Nested expression domains of four homeobox genes in developing rostral brain. Nature, 358, 687–690. Simeone, A., Acampora, D., Mallamaci, A., Stornaiuolo, A., D’Apice, M. R., Nigro, V., et al. (1993). A vertebrate gene related to orthodenticle contains a homeodomain of the bicoid class and demarcates anterior neuroectoderm in the gastrulating mouse embryo. The EMBO Journal, 12, 2735–2747. Simoes-Costa, M., & Bronner, M. E. (2015). Establishing neural crest identity: A gene regulatory recipe. Development, 142, 242–257. Simoes-Costa, M. S., McKeown, S. J., Tan-Cabugao, J., Sauka-Spengler, T., & Bronner, M. E. (2012). Dynamic and differential regulation of stem cell factor FoxD3 in the neural crest is encrypted in the genome. PLoS Genetics, 8, e1003142. Simoes-Costa, M., Stone, M., & Bronner, M. E. (2015). Axud1 integrates Wnt signaling and transcriptional inputs to drive neural crest formation. Developmental Cell, 34, 544–554. Skromne, I., & Stern, C. D. (2001). Interactions between Wnt and Vg1 signalling pathways initiate primitive streak formation in the chick embryo. Development, 128, 2915–2927. Solomon, K. S., Logsdon, J. M., Jr., & Fritz, A. (2003). Expression and phylogenetic analyses of three zebrafish FoxI class genes. Developmental Dynamics, 228, 301–307. Stern, C. D., & Downs, K. M. (2012). The hypoblast (visceral endoderm): An evo-devo perspective. Development, 139, 1059–1069. Steventon, B., Mayor, R., & Streit, A. (2012). Mutual repression between Gbx2 and Otx2 in sensory placodes reveals a general mechanism for ectodermal patterning. Developmental Biology, 367, 55–65. Steventon, B., Mayor, R., & Streit, A. (2016). Directional cell movements downstream of Gbx2 and Otx2 control the assembly of sensory placodes. Biology Open, 5, 1620–1624. Streit, A. (2002). Extensive cell movements accompany formation of the otic placode. Developmental Biology, 249, 237–254.

166

Alexandre Thiery et al.

Streit, A. (2008). The cranial sensory nervous system: specification of sensory progenitors and placodes. In F. Watt & F. Gage (Eds.), StemBook. The Stem Cell Research Community. Streit, A. (2018). Specification of sensory placode progenitors: signals and transcription factor networks. The International Journal of Developmental Biology, 62, 195–205. Streit, A., Berliner, A. J., Papanayotou, C., Sirulnik, A., & Stern, C. D. (2000). Initiation of neural induction by FGF signalling before gastrulation. Nature, 406, 74–78. Streit, A., Lee, K. J., Woo, I., Roberts, C., Jessell, T. M., & Stern, C. D. (1998). Chordin regulates primitive streak development and the stability of induced neural cells, but is not sufficient for neural induction in the chick embryo. Development, 125, 507–519. Streit, A., & Stern, C. D. (1999). Establishment and maintenance of the border of the neural plate in the chick: Involvement of FGF and BMP activity. Mechanisms of Development, 82, 51–66. Stuhlmiller, T. J., & Garcia-Castro, M. I. (2012). FGF/MAPK signaling is required in the gastrula epiblast for avian neural crest induction. Development, 139, 289–300. Sugahara, S., Fujimoto, T., Kondoh, H., & Uchikawa, M. (2018). Nasal and otic placode specific regulation of Sox2 involves both activation by Sox-Sall4 synergism and multiple repression mechanisms. Developmental Biology, 433, 61–74. Sun, S. K., Dee, C. T., Tripathi, V. B., Rengifo, A., Hirst, C. S., & Scotting, P. J. (2007). Epibranchial and otic placodes are induced by a common Fgf signal, but their subsequent development is independent. Developmental Biology, 303, 675–686. Suzuki, A., Ueno, N., & Hemmati-Brivanlou, A. (1997). Xenopus msx1 mediates epidermal induction and neural inhibition by BMP4. Development, 124, 3037–3044. Tambalo, M., Anwar, M., Ahmed, M., & Streit, A. (2020). Enhancer activation by FGF signalling during otic induction. Developmental Biology, 457, 69–82. Torres, M., Go´mez-Pardo, E., & Gruss, P. (1996). Pax2 contributes to inner ear patterning and optic nerve trajectory. Development (Cambridge, England), 122, 3381–3391. Tour, E., Pillemer, G., Gruenbaum, Y., & Fainsod, A. (2001). The two Xenopus Gbx2 genes exhibit similar, but not identical expression patterns and can affect head formation. FEBS Letters, 507, 205–209. Trevers, K. E., Prajapati, R. S., Hintze, M., Stower, M. J., Strobl, A. C., Tambalo, M., et al. (2018). Neural induction by the node and placode induction by head mesoderm share an initial state resembling neural plate border and ES cells. Proceedings of the National Academy of Sciences of the United States of America, 115, 355–360. Tribulo, C., Aybar, M. J., Nguyen, V. H., Mullins, M. C., & Mayor, R. (2003). Regulation of Msx genes by a Bmp gradient is essential for neural crest specification. Development, 130, 6441–6452. Uchikawa, M., Ishida, Y., Takemoto, T., Kamachi, Y., & Kondoh, H. (2003). Functional analysis of chicken Sox2 enhancers highlights an array of diverse regulatory elements that are conserved in mammals. Developmental Cell, 4, 509–519. Vallin, J., Thuret, R., Giacomello, E., Faraldo, M. M., Thiery, J. P., & Broders, F. (2001). Cloning and characterization of three Xenopus slug promoters reveal direct regulation by Lef/beta-catenin signaling. The Journal of Biological Chemistry, 276, 30350–30358. Villanueva, S., Glavic, A., Ruiz, P., & Mayor, R. (2002). Posteriorization by FGF, Wnt, and retinoic acid is required for neural crest induction. Developmental Biology, 241, 289–301. von Bubnoff, A., Schmidt, J. E., & Kimelman, D. (1996). The Xenopus laevis homeobox gene Xgbx-2 is an early marker of anteroposterior patterning in the ectoderm. Mechanisms of Development, 54, 149–160. Vonica, A., & Brivanlou, A. H. (2006). An obligatory caravanserai stop on the silk road to neural induction: inhibition of BMP/GDF signaling. Seminars in Cell & Developmental Biology, 17, 117–132.

Cell fate decisions in the cranial sensory nervous system

167

Wakamatsu, Y. (2011). Mutual repression between Pax3 and Pax6 is involved in the positioning of ophthalmic trigeminal placode in avian embryo. Development, Growth & Differentiation, 53, 994–1003. Wakamatsu, Y., Endo, Y., Osumi, N., & Weston, J. A. (2004). Multiple roles of Sox2, an HMG-box transcription factor in avian neural crest development. Developmental Dynamics, 229, 74–86. Wassarman, K. M., Lewandoski, M., Campbell, K., Joyner, A. L., Rubenstein, J. L., Martinez, S., et al. (1997). Specification of the anterior hindbrain and establishment of a normal mid/hindbrain organizer is dependent on Gbx2 gene function. Development, 124, 2923–2934. Wawersik, S., & Maas, R. L. (2000). Vertebrate eye development as modeled in Drosophila. Human Molecular Genetics, 9, 917–925. Wilson, S. I., Graziano, E., Harland, R., Jessell, T. M., & Edlund, T. (2000). An early requirement for FGF signalling in the acquisition of neural cell fate in the chick embryo. Current Biology, 10, 421–429. Wilson, P. A., Lagna, G., Suzuki, A., & Hemmati-Brivanlou, A. (1997). Concentrationdependent patterning of the Xenopus ectoderm by BMP4 and its signal transducer Smad1. Development, 124, 3177–3184. Wilson, S. I., Rydstrom, A., Trimborn, T., Willert, K., Nusse, R., Jessell, T. M., et al. (2001). The status of Wnt signalling regulates neural and epidermal fates in the chick embryo. Nature, 411, 325–330. Woda, J. M., Pastagia, J., Mercola, M., & Artinger, K. B. (2003). Dlx proteins position the neural plate border and determine adjacent cell fates. Development, 130, 331–342. Wright, T. J., & Mansour, S. L. (2003). Fgf3 and Fgf10 are required for mouse otic placode induction. Development, 130, 3379–3390. Xu, P. X., Adams, J., Peters, H., Brown, M. C., Heaney, S., & Maas, R. (1999). Eya1-deficient mice lack ears and kidneys and show abnormal apoptosis of organ primordia. Nature Genetics, 23, 113–117. Xu, H., Dude, C. M., & Baker, C. V. (2008). Fine-grained fate maps for the ophthalmic and maxillomandibular trigeminal placodes in the chick embryo. Developmental Biology, 317, 174–186. Yang, L., O’Neill, P., Martin, K., Maass, J. C., Vassilev, V., Ladher, R., et al. (2013). Analysis of FGF-dependent and FGF-independent pathways in otic placode induction. PLoS One, 8, e55011. Zhang, Y., Knosp, B. M., Maconochie, M., Friedman, R. A., & Smith, R. J. (2004). A comparative study of Eya1 and Eya4 protein function and its implication in branchio-oto-renal syndrome and DFNA10. Journal of the Association for Research in Otolaryngology, 5, 295–304. Zheng, W., Huang, L., Wei, Z. B., Silvius, D., Tang, B., & Xu, P. X. (2003). The role of Six1 in mammalian auditory system development. Development, 130, 3989–4000. Zou, D., Silvius, D., Rodrigo-Blomqvist, S., Enerback, S., & Xu, P. X. (2006). Eya1 regulates the growth of otic epithelium and interacts with Pax2 during the development of all sensory areas in the inner ear. Developmental Biology, 298, 430–441.

CHAPTER SIX

A Hox gene regulatory network for hindbrain segmentation Hugo J. Parkera, Robb Krumlaufa,b,* a

Stowers Institute for Medical Research, Kansas City, MO, United States Department of Anatomy and Cell Biology, Kansas University Medical Center, Kansas City, KS, United States *Corresponding author: e-mail address: [email protected] b

Contents 1. Introduction 2. A GRN controls segmental Hox expression in the hindbrain 2.1 Signaling determines early axial allocation 2.2 Spatial subdivision of the hindbrain territory by segmentation genes 2.3 Rhombomeric deployment of cell segregation factors 2.4 Establishment and maintenance of segmental Hox expression 2.5 Hox targets and cofactors 3. Hox regulatory networks and the neural crest GRN 4. Conclusions and key questions Acknowledgments References

170 173 175 180 183 184 186 189 192 193 193

Abstract In vertebrates, the hindbrain serves as a highly conserved complex coordination center for regulating many fundamental activities of the central nervous system, such as respiratory rhythms, sleep patterns and equilibrium, and it also plays an important role in craniofacial development. The basic ground plan that underlies the diverse functions of the hindbrain and its neural crest derivatives is established and patterned by a process of segmentation. Through a dynamic series of signaling and regulatory interactions the developing hindbrain is transiently compartmentalized into a set of seven segmental units, termed rhombomeres. The nested expression of the Hox family of transcription factors is tightly coupled to the process of segmentation and provides a molecular code for specifying the unique regional properties of the hindbrain and its neural crest derived craniofacial structures. The high degree of similarity in hindbrain architecture between diverse vertebrates has enabled cross-species regulatory analysis. This has facilitated the experimental assembly of the signaling and regulatory interactions, which underlie the process of segmentation, into a Hox-dependent gene regulatory network (GRN) model. This hindbrain GRN is a key regulatory feature of head patterning, conserved to the base of vertebrate evolution. This regulatory framework also serves as a basis for comparing and contrasting GRNs that govern cranial neural crest formation and axial patterning and provide insight into regulatory mechanisms associated with Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.03.001

#

2020 Elsevier Inc. All rights reserved.

169

170

Hugo J. Parker and Robb Krumlauf

the evolution of novel vertebrate traits. The purpose of this review is to discuss the majorfeatures of the GRN for hindbrain segmentation and its relationship to the broader functional role of the hindbrain in patterning head development.

1. Introduction During vertebrate development, the hindbrain territory undergoes a process of segmentation, forming a transient series of seven swellings along the anterior-posterior (A-P) axis, called rhombomeres (r) (Alexander, Nolte, & Krumlauf, 2009; Krumlauf, 2016; Lumsden, 2004), as shown in Fig. 1A. Rhombomeres represent lineage-restricted compartments of cells, with restricted cell mixing between the adjacent segments (Birgbauer & Fraser, 1994; Fraser, Keynes, & Lumsden, 1990; Guthrie & Lumsden, 1991; Guthrie, Prince, & Lumsden, 1993). This iterated series of functional territories can respond independently to axial patterning signals, creating regional diversity in the developing hindbrain. This is reflected in the patterns of neurogenesis and neuroanatomy of the early hindbrain, as seen in the segmental organization of neurons and cranial nerve roots, establishing a fundamental neuronal connectivity between the hindbrain, other brain centers, and the periphery (Chandrasekhar, 2004; Clarke & Lumsden, 1993; Gilland & Baker, 2005; Lumsden & Keynes, 1989). The hindbrain also contributes more broadly to head development and craniofacial patterning through the generation of the cranial neural crest (cNCC), which delaminates from the neural tube and migrates in a series of discrete streams to populate the pharyngeal arches (Le Douarin & Kalcheim, 1999; Trainor, Bronner-Fraser, & Krumlauf, 2004). An early segmental plan can be seen at the molecular level, as revealed by the restricted expression domains of many developmental genes which encode transcription factors functionally linked with regulating steps in the segmentation process (Alexander et al., 2009; Lumsden & Krumlauf, 1996; Parker & Krumlauf, 2017). This includes the highly conserved Hox genes, which are coupled to hindbrain segmentation, exhibiting nested and striped expression domains that correspond to the boundaries between rhombomeres (Alexander & Krumlauf, 2009; Lumsden & Krumlauf, 1996; Wilkinson, Bhatt, Cook, Boncinelli, & Krumlauf, 1989), as depicted in Fig. 1A. Hox genes play key roles in specifying segmental identity to rhombomeres and cranial neural crest cells (Alexander et al., 2009;

Hindbrain segmentation GRN

171

Fig. 1 (A) A depiction of the mouse embryonic hindbrain, viewed dorsally, with anterior to the top. Rhombomeres (r1–r8) and pharyngeal arches (pa1–pa4) annotated. On the left-hand side, rhombomeric expression domains of segmentation genes (Krox20, Kreisler) and Hox genes are shown. Darker shading of Hox expression domains (e.g., Hoxb1 in r4) denotes higher expression levels. Hoxa1 is transiently expressed during early hindbrain segmentation. On the right-hand side, the migratory streams of neural crest cells from the rhombomeres to the pharyngeal arches are indicated (arrows) and Hox expression domains in the neural crest are shown. (B) The four Hox gene clusters in mouse are shown. Hox genes are assigned to 13 paralogue groups based on their unique sequence features. The spatial and temporal collinearity of Hox gene expression across the clusters are indicated, as are their differing responsiveness to retinoic acid and FGF signaling. Hox genes of paralogue groups 1–4 are expressed in the segmental hindbrain. HB, hindbrain; MB, midbrain; pa, pharyngeal arch; r, rhombomere; s, somite.

172

Hugo J. Parker and Robb Krumlauf

Tumpel, Wiedemann, & Krumlauf, 2009). In mammals, Hox genes reside in four genomic clusters, which arose via ancient duplication events in early vertebrates (Hoegg & Meyer, 2005; Krumlauf, 1994; McGinnis & Krumlauf, 1992). They can be classified into 13 paralogue groups (PG) based on their sequence features, and display temporal and spatial collinearity along their clusters, whereby their timing and spatial domains of expression along the A-P axis correlate with their relative gene order along each chromosomal cluster (Deschamps & Duboule, 2017; Duboule & Dolle, 1989; Graham, Papalopulu, & Krumlauf, 1989; Kmita & Duboule, 2003; Lewis, 1978) (Fig. 1B). The Hox genes also display opposing gradients of responsiveness to retinoic acid (RA) and fibroblast growth factor (FGF) axial signaling pathways along the clusters (Fig. 1B). This gives rise to nested domains of expression, providing combinatorial Hox codes that specify regional properties along the A-P axis in multiple tissues, as shown for the hindbrain and neural crest in Fig. 1A. Hox expression domains are sometimes offset between rhombomeres and pharyngeal arches, in part due to differences between adjacent rhombomeres in their neural crest contributions. Even-numbered rhombomeres provide the major contributions of neural crest, while very few crest cells emerge from odd-numbered rhombomeres (Birgbauer, Sechrist, Bronner-Fraser, & Fraser, 1995; Golding, Trainor, Krumlauf, & Gassmann, 2000). There are also tissue-specific responses to environmental signals that modulate Hox expression in neural crest versus hindbrain domains, such as the repression of Hoxa2 expression in PA1 but not r2 (Trainor, Ariza-McNaughton, & Krumlauf, 2002), yet the regulatory mechanisms underlying this tissue-specificity are not fully understood (Parker, Pushel, & Krumlauf, 2018). Patterns of segmental Hox gene expression in the developing head are remarkably well conserved across vertebrates (Godsave et al., 1994; Parker, Bronner, & Krumlauf, 2014, 2019; Prince, Moens, Kimmel, & Ho, 1998), and Hox gene perturbation experiments have revealed roles in rhombomere formation, specification of rhombomere segmental identity and craniofacial patterning (Alexander et al., 2009; Lumsden, 2004; Minoux & Rijli, 2010; Moens & Prince, 2002; Parker & Krumlauf, 2017; Tumpel et al., 2009). Detailed analysis of cis-regulatory elements of Hox genes and other segmental regulators in the developing hindbrain have been reviewed in detail elsewhere (Parker, Bronner, & Krumlauf, 2016; Parker & Krumlauf, 2017; Thierion et al., 2017; Torbey et al., 2018; Tumpel et al., 2009). In this chapter, we provide an outline of the gene regulatory network governing segmental expression of Hox genes in the hindbrain.

Hindbrain segmentation GRN

173

We discuss recent findings at multiple levels of the network, including how retinoid signaling coordinates maintenance of homogeneous segments, and how segmental border sharpness is subsequently coupled to induction of specialized rhombomere boundary cells during hindbrain tissue patterning (Addison, Xu, Cayuso, & Wilkinson, 2018; Cayuso, Xu, Addison, & Wilkinson, 2019). We then outline key gaps in knowledge that merit further research, including how Hox regulation is integrated between the hindbrain and neural crest.

2. A GRN controls segmental Hox expression in the hindbrain Data from gene expression and perturbation analyses have identified many of the genes and signals that control aspects of the segmentation process (Fig. 1), and in combination with the characterization of their associated cis-regulatory elements, are contributing to the characterization of the gene regulatory network for early hindbrain development in vertebrates. The data derive predominantly from experiments in mouse, chick and zebrafish, but in light of the high degree of similarity in architecture and function of the hindbrain many aspects of the network appear to be widely conserved across vertebrate species. This evolutionary conservation was advantageous for identifying key conserved cis-regulatory elements in the GRN by crossspecies sequence comparison. While regulatory interactions via these cisregulatory elements have not been independently validated in each species, the experimental data from multiple vertebrate species has been drawn upon to formulate a GRN model that depicts our current understanding of the dynamic processes of hindbrain segmentation and patterning. This model describes a hypothetical vertebrate GRN that lays down the basic ground plan for the elaboration and coordination of programs for neural differentiation, circuit formation and head development. This is important as it provides a regulatory framework for investigating hindbrain and head development, for understanding how perturbations in this process lead to disease, for directed tissue engineering of specific hindbrain and cranial neural crest cell types, and for inferring the molecular basis of hindbrain and head evolution in early vertebrates. The network mediating segmental gene expression in the hindbrain is organized hierarchically and can be conceptualized as comprising multiple layers or modules of circuitry that govern successive stages of the developmental patterning process (Fig. 2). The GRN model describes regulatory

174

Hugo J. Parker and Robb Krumlauf

Fig. 2 The structure of the GRN for hindbrain segmentation and Hox patterning. The GRN can be broadly conceptualized as layers of regulatory circuitry mediating different aspects of the developmental process: early A-P signaling, segmental subdivision, secondary signaling, cell segregation and regulation of boundary cell identity, segmental patterning, and culminating in rhombomere differentiation programs. Key genes and signaling molecules are highlighted for each layer. Arrows represent the flow of regulatory information between the layers. The approximate developmental timing of the corresponding events in the mouse embryo are given in embryonic days post fertilization (E).

events that begin during gastrulation, around embryonic day (E)6.5 in mouse, through to rhombomeric patterning programs at approximately E11. In the primary phase of A-P patterning of the hindbrain, cues from a series of local signaling centers set up opposing signaling gradients that provide initial A-P inputs into the GRN (Deschamps & Duboule, 2017; Deschamps & van Nes, 2005; Diez del Corral et al., 2003; Diez del Corral & Storey, 2004; Duester, 2008; Rhinn & Dolle, 2012; Sirbu, Gresh, Barra, & Duester, 2005). Subsequent levels of the network include regulatory interactions mediating the activation of segmentation genes such as Kreisler, Krox20 and Hox genes, as well as the initiation of an FGF secondary signaling center and the deployment of cell segregation factors. This circuitry sets up a phase of segmental patterning leading to the generation of a rhombomeric Hox code, which is maintained and refined through auto- and cross-regulation between these segmentally expressed genes. Further regulatory programs are then deployed in each rhombomere to drive their independent differentiation and developmental trajectories.

Hindbrain segmentation GRN

175

2.1 Signaling determines early axial allocation During gastrulation and early somitogenesis, through cues emanating from embryonic signaling centers the hindbrain territory becomes delineated and subdivided along the A-P axis of the neuroepithelium by the deployment of transcription factors in discrete expression domains. The crosstalk and regulatory interactions between these transcription factors delimit abutting territories of their expression, positioning future segmental boundaries and imparting midbrain, hindbrain and spinal cord regional identities to the neuroepithelium. The interface of Otx2 and Gbx1/2 expression determines the anterior limit of the hindbrain at the midbrain-hindbrain boundary (Millet et al., 1999), while Cdx genes promote posterior embryonic development and specify spinal cord fate (Deschamps & van Nes, 2005; Metzis et al., 2018; Young et al., 2009). Numerous studies have identified key roles for major signaling pathways—Wnt, FGF and retinoic acid (RA)— in generating these early axially restricted patterns of gene expression in neural progenitors and the developing neuroepithelium, and they have continuing roles in refining interactions in the hindbrain GRN. These signals are spatially and temporally dynamic, and have an early posteriorizing effect, with Wnt and FGF signaling preceding the initiation of RA synthesis during gastrulation. In a cascade of regulatory events, this early signaling triggers the formation of secondary signaling centers, such as the somites (Begemann, Schilling, Rauch, Geisler, & Ingham, 2001) and isthmic organizer (IO) at the midbrain-hindbrain boundary ( Joyner, 1996; Lee, Danielian, Fritzsch, & McMahon, 1997; Reifers et al., 1998), which further refine A-P patterning. In turn, regulatory feedback from the A-P patterning genes also modifies properties and cues from the secondary signaling centers. Collectively, these temporally and spatially dynamic sources of cues underlie the diverse roles of the major signaling pathways in the progressive regulatory steps that govern hindbrain segmentation and patterning. 2.1.1 Wnt Early graded Wnt signaling has been shown to influence expression of members of the Otx, Gbx, Hox and Cdx families in neural progenitors during gastrulation (Nordstrom, Jessell, & Edlund, 2002; Nordstrom, Maier, Jessell, & Edlund, 2006; Rhinn, Lun, Luz, Werner, & Brand, 2005). Lrp5 and Lrp6 transmembrane proteins are Wnt co-receptors that are important components of this canonical Wnt signaling that patterns the embryo during gastrulation (Kelly, Pinson, & Skarnes, 2004). Wnt3 appears to be a key Wnt

176

Hugo J. Parker and Robb Krumlauf

ligand in this early phase of embryonic Wnt signaling, with expression restricted to the posterior epiblast and the associated visceral endoderm during early gastrulation in mouse, becoming localized to the primitive streak as gastrulation proceeds (Liu et al., 1999). A collection of Wnt-dependent enhancers has been identified 30 of the HoxA cluster that initiate Hoxa1 transcription in progenitor cells of the posterior primitive streak region in response to Wnt3 (Neijts et al., 2016). As the Wnt3 domain expands rostrally during gastrulation, Hoxa1 becomes activated more rostrally in neuromesodermal progenitors. Cdx genes are also activated at more caudal levels by Wnt signaling, where they potentiate transcription of central and 50 HoxA genes by binding to the central part of the HoxA cluster (50 to Hoxa4) (Neijts, Amin, van Rooijen, & Deschamps, 2017). This has led to a model in which two regulatory phases—(1) early initiation of Hoxa1 by Wnt, (2) later regulation of HoxA genes by Cdx—initiate collinear Hox gene transcription in the primitive streak, which confers a spatial patterning program upon these axial progenitors (Neijts & Deschamps, 2017). At early somite stages, these progenitors transmit their Hox identities to their neuroectodermal and mesodermal progenies upon differentiation, with further germ layer-specific regulatory processes then influencing subsequent Hox expression patterns (Deschamps & Duboule, 2017). In the neuroepithelium, these regulatory processes include the differential responses of Hox genes to RA and FGF signaling gradients (Bel-Vialar, Itasaki, & Krumlauf, 2002; Pownall, Tucker, Slack, & Isaacs, 1996).

2.1.2 RA From late gastrulation RA is synthesized by RALDH2 in the presomitic mesoderm and nascent somites adjacent to the caudal hindbrain territory (Begemann et al., 2001; Deschamps & van Nes, 2005; Molotkova, Molotkov, Sirbu, & Duester, 2005) (Fig. 3A). It diffuses into the hindbrain neuroepithelium and is degraded anteriorly by CYP26 enzymes, which are themselves upregulated anteriorly in response to RA (Gould, Itasaki, & Krumlauf, 1998; Hernandez, Putzke, Myers, Margaretha, & Moens, 2007; Sirbu et al., 2005; White, Nie, Lander, & Schilling, 2007; White & Schilling, 2008). As development proceeds, expression of Raldh2 is downregulated in anterior somites as they begin to differentiate and progressively activated in newly forming posterior somites during elongation of the embryonic axis. This interplay between synthesis and degradation gives rise to a temporally dynamic concentration gradient of RA in the hindbrain,

Hindbrain segmentation GRN

177

Fig. 3 Signaling during hindbrain segmental patterning. Depictions of the hindbrain are shown in dorsal view with anterior to the left. (A) Shifting activity domains of retinoic acid (RA) signaling in the developing hindbrain are shown. RA from the somitic mesoderm diffuses anteriorly in the neuroepithelium, being degraded by CYP26 enzymes. Cyp26 genes are initially upregulated in the isthmic organizer in response to FGF signaling and subsequently display dynamic expression in the hindbrain. RA activates the expression of key genes in the hindbrain GRN, such as HoxPG1 genes, vHnf1 and Hoxb4, as depicted. The shifting boundaries of Cyp26 expression and RA responsiveness are illustrated for three sequential phases, as characterized in the zebrafish hindbrain at pre-rhombomeric and rhombomeric stages. (B) FGF signaling centers (shading) characterized in the zebrafish or chick hindbrain. In zebrafish, hoxb1a activates expression of fgf3/8 to create an r4 signaling center, which influences expression of val/Kreisler and krox20 in conjunction with vHnf1. The isthmic organizer is a source of FGF signals that pattern r1 and restrict anterior Hox expression. In chick, FGF signaling from r2 and r4 influences expression of EphA4 in r3 and r5, either directly (dashed arrows) or via Krox20. r, rhombomere; s, somite.

178

Hugo J. Parker and Robb Krumlauf

with low levels anteriorly and high levels posteriorly (Deschamps & van Nes, 2005; Schilling, Sosnik, & Nie, 2016; Shimozono, Iimura, Kitaguchi, Higashijima, & Miyawaki, 2013). The RA gradient is directly interpreted by RA response elements (RAREs) in the enhancers of numerous target genes, including key players in the hindbrain GRN such as Hoxa1, Hoxb1, vHnf1, Cdx1, Hoxb4 and Hoxd4 (Nolte, De Kumar, & Krumlauf, 2019; Tumpel et al., 2009). RAREs are bound by heterodimeric complexes of retinoic acid receptors (RARs and RXRs) which are able to modulate transcription through their ability to recruit co-activators and co-repressors in a ligand-dependent manner (Chen & Evans, 1995; Glass & Rosenfeld, 2000; Horlein et al., 1995; Lavinsky et al., 1998). These RAREs respond to the RA gradient at different times and positions along the A-P axis (Dupe et al., 1997; Gould et al., 1998; Marshall et al., 1994; Studer, Popperl, Marshall, Kuroiwa, & Krumlauf, 1994), generating ordered gene expression domains (Fig. 3A). There is also direct cross-regulation between the Hox genes and retinoic acid receptors that sets up a feedback circuit which refines domains of RA signaling and brings them into alignment with rhombomere boundaries (Serpente et al., 2005). The mechanisms underlying this diversity in RA responsiveness are not fully understood, but in some cases the spacing between direct repeats of RAREs, or variation in sequences flanking the RAREs, can impact their response to RA (Evans & Mangelsdorf, 2014; Mangelsdorf et al., 1995; Nolte et al., 2006; Umesono, Murakami, Thompson, & Evans, 1991). The molecular mechanisms used by RAREs to activate the Hox genes also appear to be different. For example, the Hoxa1 and Hoxb1 loci both contain 30 RAREs which are required for their activation in neural tissues, but Hoxa1 is activated earlier than Hoxb1 (Dupe et al., 1997; Marshall et al., 1994; Studer et al., 1998). This temporal difference in activation is associated with the presence of paused polymerase on the Hoxa1 promoter in combination with the removal of a co-repressor mediated by RA, while for Hoxb1 RA signaling stimulates de novo initiation of transcription of the gene (De Kumar et al., 2015; Lin et al., 2011). This suggests that cooperativity between retinoid receptors and other transcription factors influences the sensitivity and timing for how RAREs read the RA gradient in the hindbrain. Multiple RAREs have been identified in the Hox clusters and contribute to the general colinear RA response of Hox genes in neural cells (Nolte et al., 2019; Simeone et al., 1990, 1991). However, each Hox gene does not contain an independent RARE which governs its response to RA (Nolte et al., 2019). Reporter assays and mutational analysis of endogenous

Hindbrain segmentation GRN

179

RAREs have revealed that this coordinate response to RA of individual Hox clusters can be mediated through the action of shared RAREdependent enhancers. These enhancers embedded in the clusters, not only regulate the adjacent Hox genes, but they act more globally through long-range interactions to coregulate multiple Hox genes in the cluster (Ahn, Mullan, & Krumlauf, 2014; Gould, Morrison, Sproat, White, & Krumlauf, 1997; Nolte, Jinks, Wang, Martinez Pastor, & Krumlauf, 2013; Oosterveen et al., 2003; Qian et al., 2018; Sharpe, Nonchev, Gould, Whiting, & Krumlauf, 1998). Hence, a series of interactions involving multiple RAREs in individual Hox clusters underlie the coordinate response of these genes to the temporally dynamic gradient of RA which lays down the segmental Hox code in hindbrain segments. There is experimental evidence from regulatory comparisons of amphioxus, a cephalochordate, and mouse Hox clusters indicating that the position of some of these RAREs are highly conserved (Manzanares et al., 2000; Wada, Escriva, Zhang, & Laudet, 2006). In combination with recent evidence that Wnt signaling is important for regulating Hox genes in hemichordates (Darras et al., 2018), this suggests that RA and Wnt signaling are part of an ancient mechanism for generating nested domains of Hox expression along the A-P axis in patterning the nervous system. The temporally dynamic nature of the RA gradient is due in part to the changing expression domains of Raldh2 and Cyp26 genes during development. Hoxa1 and its cofactor Pbx1 have been shown to directly regulate Raldh2 expression in the presomitic mesoderm in mouse and frog, but it is unclear to what extent this Hox-dependent regulation mediates initiation versus maintenance of Raldh2 expression (Vitobello et al., 2011). Since Hox gene expression in the developing mesoderm is initially RA-independent, it has been proposed that Hox genes themselves, in this case Hoxa1, influence their expression in the hindbrain neuroepithelium through a feed-forward transcriptional mechanism mediated by direct control of mesodermal Raldh2 expression and the resulting production of RA that diffuses to the neuroectoderm (Vitobello et al., 2011). The Cyp26 genes play combinatorial roles in modulating RA-dependent hindbrain patterning events. In zebrafish, cyp26a1 is expressed in the anterior neuroectoderm under complex transcriptional control by both RA and FGF signaling, where it plays a major RA-degrading role during gastrulation (Hernandez et al., 2007; White et al., 2007). cyp26b1 and cyp26c1 have slightly later onset and exhibit dynamic rhombomere-specific expression in r2–r6 (Hernandez et al., 2007). Loss-of-function studies are revealing that these genes play multiple

180

Hugo J. Parker and Robb Krumlauf

roles during hindbrain development. They influence the establishment of initial anterior borders of RA-regulated gene expression, such as Hoxa1 in pre-r3, and refine or tighten the boundaries between mutually exclusive gene expression territories, such as between Krox20 and Hoxb1 in r3/r4 (Addison et al., 2018; Hernandez et al., 2007; Sirbu et al., 2005; White et al., 2007). Furthermore, genomic studies are revealing that Cyp26 genes can be direct targets of Hox proteins, indicating that Hox genes themselves may influence the local levels of RA (De Kumar, Parker, Paulson, Parrish, Zeitlinger, et al., 2017; Qian et al., 2018). 2.1.3 FGF FGF signaling plays multiple roles in hindbrain segmental patterning. Increasing FGF concentrations lead to expression of progressively more posterior Hox genes, suggesting that FGF signaling from the caudal mesoderm during gastrulation has an early posteriorizing effect on the neuroectoderm (Diez del Corral & Storey, 2004; Pownall, Isaacs, & Slack, 1998; Pownall et al., 1996). Additionally, at mid gastrula to early segmentation stages, two secondary signaling centers have been identified in zebrafish that release planar FGF signals to pattern the developing hindbrain—the isthmic organizer and presumptive r4 (Mason, 2007) (Fig. 3B). FGF signaling from the IO inhibits Hox gene expression in r1, sets the anterior limit of Hoxa2 expression in r2 and is required for development of the cerebellum (Irving & Mason, 2000; Liu et al., 2003; Mason, Chambers, Shamim, Walshe, & Irving, 2000). In presumptive r4 of zebrafish, HoxPG1-dependent expression of fgf3/8 generates FGF signals that influence caudal hindbrain patterning (Maves, Jackman, & Kimmel, 2002). These FGF signals act synergistically with vHnf1, itself upregulated caudal to r4 in response to RA, to activate val (Kreisler) expression in r5–r6 (Hernandez, Rikhof, Bachmann, & Moens, 2004). This convergence of RA and FGF signaling in Kreisler regulation highlights how these signaling pathways can be integrated by target genes during caudal hindbrain patterning. There is also evidence from analysis in chicken embryos, that r2 and r4 function as local signaling centers in the hindbrain that maintain the expression of EphA4 in r3 and r5 through FGF signaling independent of initiation by Krox20 (Cambronero, ArizaMcNaughton, Wiedemann, & Krumlauf, 2020; Theil et al., 1998) (Fig. 3B).

2.2 Spatial subdivision of the hindbrain territory by segmentation genes The early phase of A-P signaling inputs triggers a regulatory cascade among the responding segmentation genes and their targets in the GRN which then

Hindbrain segmentation GRN

181

leads to a regulatory phase that specifies the segmental subdivision of the hindbrain. In this step, the tight A-P registration of expression domains of key segmentation genes (Hoxa1, vHnf1, Krox20, Kreisler and Cdx1) prefigures rhombomere boundaries. This series of steps is illustrated in Fig. 4. An important cis-regulatory feature in the wiring of this level of the GRN is the pivotal role for direct cross-regulatory interactions between many of the segmentation genes. This circuitry ensures that the spatially-restricted domains of segmental gene expression, triggered by transient and dynamic A-P signaling inputs, can be sharpened and maintained by positive-feedback loops and refined by mutual repression between abutting domains. This enables the hindbrain region to be progressively subdivided into seven discrete segments each with unique identities by the integration of multiple regulatory inputs on the cis-regulatory elements of downstream Hox genes and other targets (Fig. 4). The regulatory interactions between HoxPG1 and Krox20 in delineating r3–r5 exemplify how their dynamic patterns of gene expression are sculpted by the GRN architecture. RA signaling drives the early expression of HoxPG1 factors up to the presumptive r3 territory (Fig. 3), where they contribute to the initiation of Krox20 expression (Helmbacher et al., 1998; Wassef et al., 2008) (Fig. 4). In turn, Krox20 represses the HoxPG1 genes and maintains its own expression through positive autoregulation (Bouchoucha et al., 2013; Chomette, Frain, Cereghini, Charnay, & Ghislain, 2006). This regulatory interaction establishes a Krox20 expression domain that defines the future r3 segment. With the posterior regression of HoxPG1 expression due to temporal shifts in RA levels, the Krox20 expression in r3 is restricted from expanding posteriorly due to the maintenance of Hoxb1 in r4 by a positive auto-regulatory circuit (Popperl et al., 1995; Studer et al., 1998). Hoxb1 has the ability to repress Krox20 via Nlz factors (Labalette et al., 2015; Mechta-Grigoriou, Garel, & Charnay, 2000), leading to the establishment of a stable interface between anterior Krox20 expression and posterior Hoxb1 expression due to mutual repression, which defines the r3/r4 border. Posteriorly, Krox20 is initiated in r5 by vHnf1 and Kreisler, where positive autoregulation and mutual repression with Hoxb1 define the r4/r5 boundary (Chomette et al., 2006). The elucidation of these regulatory events has been driven by detailed identification and characterization of Krox20 cis-regulatory elements in mouse, chick and zebrafish. These elegant studies have uncovered a complex regulatory landscape, comprising multiple regulatory elements that exhibit cooperation, functional versatility and plasticity of their roles across evolution (Thierion et al., 2017; Torbey et al., 2018).

182

Fig. 4 See figure legend on opposite page.

Hugo J. Parker and Robb Krumlauf

Hindbrain segmentation GRN

183

2.3 Rhombomeric deployment of cell segregation factors The borders between the opposing gene expression territories that prefigure the prospective rhombomeres are initially ragged, with individual cells at first expressing combinations of transcription factors that confer opposing rhombomere identities (such as Krox20 and Hoxb1). Sharpening these borders is important for the generation of rhombomeric segments with homogenous identity. Recent mechanistic studies have shed light on how this is achieved through a combination of cell sorting and cell identity switching (Addison et al., 2018). Cells from adjacent rhombomeres exhibit restricted intermingling (Guthrie & Lumsden, 1991; Guthrie et al., 1993), mediated by segmentally expressed cell surface molecules including Eph receptors and ephrins, which influence selective repulsion, tension and adhesion between cells (Cayuso, Xu, & Wilkinson, 2015; Mellitzer, Xu, & Wilkinson, 1999; Nieto, Gilardi-Hebenstreit, Charnay, & Wilkinson, 1992; Sela-Donenfeld & Wilkinson, 2005; Xu, Mellitzer, Robinson, & Wilkinson, 1999). Krox20 directly regulates EphA4 expression in r3 and r5, indicating a link between segmental subdivision and cell segregation modules of the hindbrain GRN (Fig. 2) (Theil et al., 1998). Despite this mechanism for cell sorting, it was found that some cells can intermingle between r3/r4 territories during early segmentation (Addison et al., 2018; Birgbauer & Fraser, 1994; Fraser et al., 1990). Heterotopic grafting experiments in zebrafish and mice have shown that hindbrain cells, when placed in an adjacent rhombomeric environment, are able to switch their identities through local signaling (Schilling, Prince, & Ingham, 2001; Trainor & Krumlauf, 2000a, 2000b). This shows that rhombomeric cells at this stage display plasticity and are able to alter their fates based on signaling cues. This mechanism has been shown to be important in sharpening rhombomere boundaries through cell identity switching in normal hindbrain development. Retinoid signaling and cross-regulatory interactions Fig. 4 Regulatory interactions between Hox genes and other segmentally expressed genes in the developing hindbrain at three successive stages. Expression domains of genes are shown in relation to the positions of the forming rhombomere boundaries in the hindbrain, depicted in dorsal view with anterior to the left. Green and red arrows indicate positive and negative regulatory interactions, respectively. In the pre-rhombomeric hindbrain, A-P signaling inputs, including RA signaling, contribute to early nested gene expression domains. This triggers a cascade of regulatory interactions resulting in segmental subdivision of the hindbrain territory, positioning of rhombomere boundaries (r1–r7), and the initiation and refinement of rhombomeric Hox gene expression.

184

Hugo J. Parker and Robb Krumlauf

used in establishing segmental territories underlie this process. An RARE-dependent enhancer of Hoxb1 is required along with Krox20 for repressing its expression in r3 and r5 (Studer et al., 1994). The segmental differences in RA levels between r3 and r4 are mediated by CYP26dependent degradation (Addison et al., 2018). Since cyp26 gene expression was found to be regulated by krox20 in r3/r5, different levels of RA signaling are directly coupled to segment identity, enabling maintenance of homogenous segmental identity through cell fate switching, despite the intermingling of cells during rhombomere formation (Wilkinson, 2018). Upon rhombomeric segmentation, specialized boundary cells form at rhombomere borders and play important roles during subsequent hindbrain development (Amoyel, Cheng, Jiang, & Wilkinson, 2005; Cheng et al., 2004; Heyman, Faissner, & Lumsden, 1995; Sela-Donenfeld, Kayam, & Wilkinson, 2009; Weisinger, Kohl, Kayam, Monsonego-Ornan, & SelaDonenfeld, 2012). These include acting as a physical barrier to prevent cell intermingling, influencing patterns of differentiation and neural organization of adjacent cells via cell signaling pathways, and providing a source of progenitor cells (Peretz et al., 2016; Sela-Donenfeld et al., 2009). These hindbrain boundaries are under high mechanical stress during segmentation, and recent studies in zebrafish have begun to uncover how the physical properties of the boundary cell environment are coupled to their cell fate decisions, such as proliferation or differentiation. Segmental EphA4 signaling increases cortical tension by regulating myosin phosphorylation, which contributes to border sharpening (Calzolari, Terriente, & Pujades, 2014; Cayuso et al., 2019). Myosin phosphorylation also leads to nuclear translocation of Taz, a component of Hippo signaling, which then induces boundary cell formation by activating boundary gene expression (Voltes et al., 2019). Thus, Hippo signaling couples border sharpness to boundary cell induction and ensures the correct organization of signaling centers at rhombomere borders. Hippo signaling is subsequently required for maintaining progenitor behavior in boundary cells. Hence, the segmental expression of EphA4 in r3 and r5, downstream of Krox20, coordinates both cell sorting and identity at rhombomere boundaries in hindbrain tissue patterning (Fig. 2).

2.4 Establishment and maintenance of segmental Hox expression Hox expression in the developing hindbrain can be classified into three separate and sequential phases—(1) initiation prior to segmentation,

Hindbrain segmentation GRN

185

(2) segmental restriction and (3) post-segmental maintenance. Early presegmental nested expression is generated in response to A-P signaling inputs, with segment-restricted expression then being progressively established and refined, as shown in Fig. 4. Detailed regulatory studies in mouse, chick and zebrafish have identified a diverse array of cis-regulatory elements that contribute to the majority of the segmental Hox expression domains in the hindbrain (Alexander et al., 2009; Parker & Krumlauf, 2017; Tumpel et al., 2009). Each Hox gene is controlled by multiple regulatory elements, which contribute to specific spatio-temporal subsets of the segmental pattern (Tumpel et al., 2009). Characterization of the functional sites and upstream regulators in these cis-regulatory modules has provided insights into how they are wired into the GRN. There are important direct inputs from the segmental regulatory genes Krox20 and Kreisler (Manzanares et al., 1999, 1997, 2002; Nonchev, Maconochie, et al., 1996; Nonchev, Vesque, et al., 1996; Sham et al., 1993), and extensive auto- and cross-regulatory interactions between Hox genes (Gould et al., 1998, 1997; Maconochie et al., 1997; Manzanares et al., 2001; T€ umpel et al., 2007). For instance, expression of Hoxb1 in r4 is generated initially by a combination of the broad activation of HoxPG1 genes in response to RA through 30 RAREs (Dupe et al., 1997; Marshall et al., 1994; Studer et al., 1998). Subsequently, it is restricted to r4 through repression in r3/ r5 by a Krox20- and RA-responsive 50 element (Studer et al., 1994) and repression in r6 by Hoxa3 and Hoxb3 (Gaufo, Thomas, & Capecchi, 2003; Wong et al., 2011). Expression is maintained in r4 by a 50 Hoxresponse element that integrates auto- and cross-regulatory activation by Hoxb1, Hoxb2 and Hoxa1 (Davenne et al., 1999; Pattyn et al., 2003; Popperl et al., 1995; Studer et al., 1998). The regulatory elements mediating expression of Hoxa2 in r2–r5 each have separate, rhombomere-specific activities that combine additively to generate the major aspects of its segmental expression in the hindbrain. Krox20 directly up-regulates Hoxa2 and Hoxb2 in r3 and r5 via paralogous enhancers 50 to each gene (Nonchev, Vesque, et al., 1996; Sham et al., 1993), while their expression in r4 is regulated by Hoxb1 via an intronic enhancer of Hoxa2 and a 50 enhancer of Hoxb2 (Maconochie et al., 1997; T€ umpel et al., 2007). Hoxa2 is regulated in r2 by an enhancer located in its homeodomain-encoding exon (T€ umpel, Cambronero, Sims, Krumlauf, & Wiedemann, 2008). HoxPG3 genes are activated in r5 by Krox20 and Kreisler (Manzanares et al., 1999, 1997, 2002), while Hoxb4 and Hoxd4 are activated posteriorly to the r6/r7 boundary by RA

186

Hugo J. Parker and Robb Krumlauf

(Gould et al., 1998, 1997; Morrison, Moroni, Ariza-McNaughton, Krumlauf, & Mavilio, 1996; Nolte et al., 2006). Hoxa3 and Hoxb4 subsequently maintain their expression through autoregulation (Gould et al., 1997; Manzanares et al., 2001). In addition to auto- and cross-regulation, Hox genes also reinforce their expression via feedback loops with their segmental regulators, such as Hoxa2 enhancing Krox20 expression in r3 (Wassef et al., 2008), and HoxPG4 genes regulating the retinoic acid receptor β gene (Serpente et al., 2005). Downstream of segmental Hox expression, Hox factors impart unique identities to each segment by regulating rhombomere-specific suites of differentiation genes (Fig. 2).

2.5 Hox targets and cofactors The nature of Hox downstream target genes and the mechanisms mediating their regulation by specific Hox factors are key questions for generating a more complete understanding of how rhombomere-specific neuroanatomies are acquired. Genome wide binding analyses are beginning to provide insight into Hox targets (Amin et al., 2015; De Kumar, Parker, Parrish, et al., 2017; De Kumar, Parker, Paulson, Parrish, Zeitlinger, et al., 2017; Donaldson et al., 2012), and this is currently an active area of investigation important for understanding final outputs of the hindbrain GRN that specifies the distinct properties of each segment. RNA sequencing approaches are beginning to enhance our understanding of segmentally expressed genes in the hindbrain, which will help in expanding the scope of the GRN in relation to rhombomeric differentiation programs, in particular neurogenesis (Tambalo, Mitter, & Wilkinson, 2020). Early segmental Hox expression is generally homogenous throughout the dorsoventral and mediolateral extent of the neural tube, but subsequently becomes restricted to specific pools of differentiating neurons at later stages of development. It is interesting to note that, in mouse, members of the HoxA and HoxC clusters become ventrally restricted while HoxB gene expression becomes dorsally restricted in correlation with the birth of different classes of neurons (Gaunt, 1991; Graham, Maden, & Krumlauf, 1991; Krumlauf, 2016). Functional studies have shown roles for Hox genes in regulating the identity of motor neurons in the spinal cord (Dasen & Jessell, 2009). However, nothing is known about the cis-regulatory basis for D-V restriction of Hox gene expression in hindbrain neurogenesis, so an important question for future research will be to understand the signals and regulatory mechanisms that underlie this aspect of the later stages of hindbrain segmental patterning.

Hindbrain segmentation GRN

187

A modest number of studies have employed genome-wide approaches to identify transcripts that are differentially expressed upon manipulation of the rhombomeric expression of specific Hox genes, such as HoxPG1 genes in zebrafish and mouse (Choe, Zhang, Hirsch, Straubhaar, & Sagerstrom, 2011; Gouti & Gavalas, 2008; Makki & Capecchi, 2011; Qian et al., 2018; Rohrschneider, Elsen, & Prince, 2007; Tvrdik & Capecchi, 2006). These studies revealed putative Hox targets from multiple functional categories of genes, including transcription factors, signaling molecules, cell sorting factors, structural proteins and factors that regulate pluripotency. More recently, direct targets of Hoxa1 have been identified in neural cell culture, by using chromatin immunoprecipitation followed by next generation sequencing (ChIP-seq) (De Kumar, Parker, Parrish, et al., 2017; De Kumar, Parker, Paulson, Parrish, Pushel, et al., 2017; De Kumar, Parker, Paulson, Parrish, Zeitlinger, et al., 2017). This approach provided further evidence that Hoxa1 may regulate pluripotency factors to influence differentiation in the neuroectoderm. Hoxa1 shows dynamic patterns of occupancy over pluripotency genes during embryonic stem cell differentiation into neuroectoderm, and significant co-occupancy with Nanog, a pluripotency factor, on many common target sites. Additionally, Hoxa1 and Nanog bind to each other’s regulatory regions, implicating a mechanism of direct mutually repressive cross-regulatory feedback between these factors. This leads to a model in which direct inputs from Hoxa1 and Nanog on shared targets, as well as mutual repression between Hoxa1 and the pluripotency network, modulates the balance between pluripotency and differentiation during early hindbrain development. This may be relevant to the neural crest cells that emanate from the hindbrain at these axial levels, as Hoxa1 is not expressed in neural crest cells but is required for the early formation and migration of r4-derived NCCs into pharyngeal arch 2 (Gavalas, Trainor, Ariza-McNaughton, & Krumlauf, 2001). The studies described above provide insight into the diverse manner in which regulatory inputs from Hox genes can be integrated with other patterning cues to generate restricted expression patterns of their downstream targets. For example, while hoxb1a is expressed broadly in r4 of the zebrafish hindbrain, many of its targets are expressed in restricted rhombomeric subdomains (Rohrschneider et al., 2007). This suggests that, in addition to the early pan-r4 input of hoxb1a, additional patterning cues that mediate dorsoventral, mediolateral, and temporal regulation are also being integrated to drive restricted expression of these target genes. This integration is likely to be carried out by the cis-regulatory elements of these hoxb1a target genes;

188

Hugo J. Parker and Robb Krumlauf

however, it is unknown which of these differentially expressed genes are direct or indirect targets. While many of the factors that influence tissueand domain-specificity of cis-regulation at Hox target enhancers remain to be characterized, numerous lines of evidence implicate the TALE family of transcription factors (including PBX and MEIS) as conserved Hox cofactors (Mann, Lelli, & Joshi, 2009; Merabet & Mann, 2016). They have been shown to form complexes with Hox factors and modulate their binding specificities (Rezsohazy, Saurin, Maurel-Zaffran, & Graba, 2015; Slattery et al., 2011), as well as interacting as cofactors with a variety of other homeodomain and non-homeodomain classes of transcription factors. Indeed, PBX and MEIS play important roles in patterning diverse tissues during development, many of which are independent of Hox factors (Schulte & Frank, 2014; Selleri, Zappavigna, & Ferretti, 2019). In vertebrates, there are multiple Pbx and Meis genes, which are expressed in the hindbrain and are required for its patterning (Moens & Selleri, 2006; Selleri et al., 2019; Waskiewicz, Rikhof, Hernandez, & Moens, 2001). In zebrafish, their perturbation leads to a lack of segmentation of r2–r6 and an r1-like neural patterning extending along the A-P axis of the hindbrain (Choe, Vlachakis, & Sagerstrom, 2002; Waskiewicz, Rikhof, & Moens, 2002). Binding sites for PBX and Meis are present and important for the regulatory activity of the many auto- and cross-regulatory enhancers of Hox genes described above (Ferretti et al., 2005, 2000; Manzanares et al., 2001). This has led to a model in which Pbx and Meis proteins act at multiple levels in the hindbrain segmentation GRN, including roles as Hox cofactors. Recent genome-wide studies have identified direct cross-regulatory interactions between Hox and TALE genes in mouse embryos and ES cells, pointing toward complex regulatory feedback mechanisms between these genes during development (De Kumar, Parker, Paulson, Parrish, Pushel, et al., 2017; Penkov et al., 2013). This provides a regulatory circuitry that ensures that the appropriate TALE cofactors are co-expressed with Hox genes for acting on downstream target genes. Hoxa1-bound regions in differentiating mouse ES cells are enriched for binding motifs for Pbx and Meis, and genome-wide binding analysis of multiple TALE members (Pbx, Meis, Tgif, Prep) revealed that almost all Hoxa1-bound loci also display occupancy of one or more TALE members (De Kumar, Parker, Paulson, Parrish, Pushel, et al., 2017). This suggests that, beyond Pbx and Meis, the family of TALE proteins may represent a wide repertoire of Hox cofactors and coregulate Hox-responsive enhancers through distinct mechanisms. Thus, a complex regulatory network comprising extensive

Hindbrain segmentation GRN

189

auto- and cross-regulatory interactions among Hox and TALE genes appears to be initiated in the neuroectoderm during early hindbrain development.

3. Hox regulatory networks and the neural crest GRN Hindbrain segmentation plays a broader role in craniofacial development through the generation of cranial neural crest which forms much of the bone and connective tissues in the developing head. The neural crest borders the neural plate during early vertebrate embryogenesis and gives rise to a transient population of migratory, multipotent cells that generate a variety of derivatives throughout the body (Le Douarin & Kalcheim, 1999; Trainor et al., 2004). In cranial regions, cranial neural crest cells are specified in a domain adjacent to the midbrain and hindbrain, from which they migrate into the pharyngeal arches and differentiate into the majority of the bones, connective tissues and certain nerves of the head. The cellular programs and regulatory mechanisms governing the biology of NCCs are rapidly being elucidated, leading to the formulation of a core NCC GRN model (Martik & Bronner, 2017; Meulemans & Bronner-Fraser, 2004; Simoes-Costa & Bronner, 2015). This GRN comprises a hierarchical series of circuits that contribute to stages of cNCC development including specification, delamination, migration and differentiation (Fig. 5A). Much like the GRN for hindbrain segmentation, this core NCC GRN appears to be highly conserved across vertebrates (Green, Simoes-Costa, & Bronner, 2015; Sauka-Spengler, Meulemans, Jones, & Bronner-Fraser, 2007). NCCs that originate at different axial levels have distinct differentiation potentials; for instance, cranial NCCs can give rise to cartilage while trunk NCCs cannot. The molecular programs that underlie the divergent properties of NCCs at different axial levels are starting to be elucidated (Rothstein, Bhattacharya, & Simoes-Costa, 2018; Simoes-Costa & Bronner, 2016). It has recently been shown that axial-specific NCC gene expression profiles differ between species, suggesting that the NCC GRN circuitry has been progressively elaborated at different axial levels during vertebrate evolution, possibly contributing to the evolution of vertebrate head complexity (Green, Uy, & Bronner, 2017; Martik et al., 2019). Hox genes display nested domains of expression in cNCCs, and perturbation studies have demonstrated important roles for these genes at multiple stages of NCC development, including specification, migration and differentiation (Hunt et al., 1991; Minoux & Rijli, 2010; Parker et al., 2018; Trainor & Krumlauf, 2001). However, the mechanisms by which Hox genes

190

Hugo J. Parker and Robb Krumlauf

Fig. 5 (A) Outstanding questions relating to the regulatory interactions between GRNs mediating hindbrain segmentation and Hox-dependent A-P patterning, Hox activity in cranial NCC and NCC formation and diversification. The broad structures of the hindbrain segmentation GRN, Hox cNCC GRN and cNCC GRN are depicted alongside each other, with interactions between these regulatory networks represented by arrows. Dashed arrows and question marks indicate putative or inferred interactions that remain to be clarified. Early Hox expression in the neuroepithelium in response to signaling pathways is maintained by auto- and cross-regulation, but it is unclear to what extent this Hox patterning is carried over during development of cNCCs. HoxPG1 genes appear to influence early cNCC development by interacting with factors in the neural plate border or neural crest specification circuitry, although direct interactions remain to be identified. Other key questions relate to the roles of cNCC GRN factors in mediating or maintaining Hox expression in cNCCs, the role of Hox auto- and cross-regulation in cNCCs, and the interactions between Hox genes and NCC migration and diversification circuits. (B) The tissue-specific activities (shading) of characterized mouse HoxPG2 NCC enhancers in hindbrain and pharyngeal arch domains. Hoxa2 is regulated in r4 and r4-derived NCC by two independent enhancers, implying distinct regulatory mechanisms underly neural and NCC expression for this gene. In contrast, Hoxb2 is regulated in r4 and r4-derived NCC by a single, shared enhancer, suggesting shared regulation in these two tissues. Rhombomeres (r) and pharyngeal arches are numbered.

Hindbrain segmentation GRN

191

mediate A-P patterning of the craniofacial skeleton, and their roles within NCC gene regulatory circuits, remain to be determined. Hox genes are not yet integrated within the current core NCC GRN. Hence, despite the established functional roles of Hox genes in cranial NCC (Minoux & Rijli, 2010; Parker et al., 2018), whether the Hox code is coupled to this GRN and if so at what level, remains to be determined (Fig. 5A). Current knowledge indicates that RA-dependent Hoxa1 expression in the neuroepithelium precedes NCC delamination, with the Hoxa1 lineage giving rise to all r4-derived NCCs (Makki & Capecchi, 2010). Perturbation of HoxPG1 genes in the neuroepithelium in both mouse and frog leads to a loss of r4-derived NCCs, possibly through impaired NCC specification or migration (Gavalas et al., 2001; McNulty, Peres, Bardine, van den Akker, & Durston, 2005). Microarray studies in mouse and zebrafish, and DNA binding analyses in mouse ESCs have identified putative downstream targets of HoxPG1 genes that are consistent with regulation of NCC specification factors (including Foxd3, tfap2 and Zic1) and pluripotency genes (Choe et al., 2011; De Kumar, Parker, Parrish, et al., 2017; Makki & Capecchi, 2011). Thus, multiple lines of evidence implicate roles for HoxPG1 genes in early NCC development, but the underlying mechanisms involved remain unclear (Fig. 5A). Identification of direct targets of these factors in the NCC lineage will be important for understanding how HoxPG1 genes interact with components of the NCC GRN. Explaining how Hox A-P patterning is integrated between neural tissue and the NCC-derived pharyngeal apparatus raises a variety of conceptual questions regarding vertebrate head development and evolution. For instance, to what extent does Hox patterning in cNCCs involve a passive carry-over of the hindbrain Hox code versus generation of an independent NCC Hox code? Do auto-/cross-regulation play a role in maintaining Hox regulatory states in NCCs as they delaminate and migrate from the neuroepithelium? Do shared or separate cis-regulatory elements mediate Hox expression in NCCs versus neural domains? This last question is significant from an evolutionary perspective, since bona-fide NCCs appear to be a vertebrate novelty. Understanding how Hox expression in cNCCs is regulated will help to address whether this evolved through co-option of neural enhancers, or via de novo evolution of NCC enhancers. However, in contrast to our knowledge of the regulatory interactions mediating rhombomeric Hox expression, relatively little is known about the cis-regulatory basis for Hox expression in cNCCs.

192

Hugo J. Parker and Robb Krumlauf

The HoxPG2 genes are best studied in this regard, since Hoxa2 has been shown to be an important selector gene for regulating the identity of the craniofacial skeleton that derives from pharyngeal arch 2 (GendronMaguire, Mallo, Zhang, & Gridley, 1993; Kitazawa et al., 2015; Parker et al., 2018; Rijli et al., 1993; Santagati, Minoux, Ren, & Rijli, 2005). However, regulatory analysis of the HoxPG2 genes finds that each gene supports a different model with respect to shared versus independent regulation in hindbrain and NCC domains. In mouse, Hoxa2 has independent enhancers regulating its expression in r4 and r4-derived NCCs (Maconochie et al., 1999; T€ umpel et al., 2007), while Hoxb2 expression in these domains is mediated by a shared enhancer (Ferretti et al., 2000; Maconochie et al., 1997) (Fig. 5B). Detailed cross-species comparisons, including with the jawless sea lamprey, recently revealed the Hoxa2 and Hoxb2 NCC enhancers to be ancient paralogues, regulated by a HoxTALE regulatory circuit that was present in ancestral vertebrates (Parker, De Kumar, et al., 2019). An equivalent lamprey enhancer was identified that appears to have retained ancestral activity in both NCC and hindbrain, while the paralogous Hoxa2 and Hoxb2 NCC enhancers differentially partitioned NCC and hindbrain activities in the gnathostome lineage. This suggests that NCC expression of HoxPG2 may have evolved through co-option of an ancestral neural element. It will be interesting to investigate the basis for NCC expression of other Hox genes, and to search for these regulatory interactions deeper in the deuterostome phylogeny, to understand how Hox expression in cNCCs arose during vertebrate evolution.

4. Conclusions and key questions Vertebrate hindbrain development is orchestrated by a cascade of gene regulatory interactions that governs formation and patterning of rhombomeres. Major progress has been made in elucidating many of the components of this network, as well as their regulatory circuitry. The resulting hindbrain GRN model provides a framework for interpreting data from multiple experimental sources and enables comparison of regulatory mechanisms between species, revealing deep conservation of the network across all vertebrates. Recent studies have advanced our knowledge at multiple levels of the GRN particularly with respect to early and later phases. This is illustrated by findings on the early activation of Hox expression in neuromesodermal progenitors by Wnt signaling, the regulation of cell sorting and identity switching during early rhombomere formation, the induction of boundary cells between rhombomeres, and the identification

Hindbrain segmentation GRN

193

of Hox downstream targets and Hox cross-regulatory interactions with cofactors. Future progress in expanding the GRN to include rhombomeric and neural crest differentiation programs should be informative for linking this regulatory circuitry to the control of morphological diversity in head development and evolution. Besides establishing the formation of neuronal circuitry, the hindbrain also plays a key role in patterning head development via the cranial neural crest, where Hox factors have well established roles. However, there has been relatively little progress in understanding how Hox genes are integrated within the gene regulatory networks governing neural crest development. Outstanding questions relate to the degree of direct regulatory crosstalk between Hox genes and the core NCC GRN, and whether/how Hox factors regulate different targets in neural versus neural crest tissue. Current and future efforts toward characterization of the downstream targets and their cognate cis-regulatory elements of Hox and NCC GRN transcription factors will be important for clarifying these regulatory interactions. While many of the upstream regulatory inputs into rhombomeric Hox expression have been identified, relatively little is known about how Hox genes are regulated in NCCs, and the degree to which this regulation is shared or independent from their regulation in the neuroepithelium. From existing work on HoxPG2 enhancers, there is evidence for both shared and independent regulation between Hox gene expression in the hindbrain rhombomeres and NCCs. This has interesting implications for the evolution of vertebrate head patterning, suggesting that Hox regulation in NCCs may have evolved in part via transfer of the hindbrain Hox code, through elaboration or co-option of pre-existing hindbrain enhancers. Identification and characterization of Hox NCC cis-regulatory elements will be important for investigating these regulatory and evolutionary models in more detail.

Acknowledgments We are grateful to Mark Miller for assistance with figure design and to members of the Krumlauf lab for valuable discussions on the topic of this review. Work in the author’s lab is funded by the Stowers Institute for Medical Research.

References Addison, M., Xu, Q., Cayuso, J., & Wilkinson, D. G. (2018). Cell identity switching regulated by retinoic acid signaling maintains homogeneous segments in the hindbrain. Developmental Cell, 45(606–620), e603. Ahn, Y., Mullan, H. E., & Krumlauf, R. (2014). Long-range regulation by shared retinoic acid response elements modulates dynamic expression of posterior Hoxb genes in CNS development. Developmental Biology, 388, 134–144.

194

Hugo J. Parker and Robb Krumlauf

Alexander, T., & Krumlauf, R. (2009). Mammalian embryo: Hox genes, eLS citable reviews in life sciences. Chichester: John Wiley & Sons. Ltd. Alexander, T., Nolte, C., & Krumlauf, R. (2009). Hox genes and segmentation of the hindbrain and axial skeleton. Annual Review of Cell and Developmental Biology, 25, 431–456. Amin, S., Donaldson, I. J., Zannino, D. A., Hensman, J., Rattray, M., Losa, M., et al. (2015). Hoxa2 selectively enhances Meis binding to change a branchial arch ground state. Developmental Cell, 32, 265–277. Amoyel, M., Cheng, Y. C., Jiang, Y. J., & Wilkinson, D. G. (2005). Wnt1 regulates neurogenesis and mediates lateral inhibition of boundary cell specification in the zebrafish hindbrain. Development, 132, 775–785. Begemann, G., Schilling, T. F., Rauch, G. J., Geisler, R., & Ingham, P. W. (2001). The zebrafish neckless mutation reveals a requirement for raldh2 in mesodermal signals that pattern the hindbrain. Development, 128, 3081–3094. Bel-Vialar, S., Itasaki, N., & Krumlauf, R. (2002). Initiating Hox gene expression: In the early chick neural tube differential sensitivity to FGF and RA signaling subdivides the HoxB genes in two distinct groups. Development, 129, 5103–5115. Birgbauer, E., & Fraser, S. E. (1994). Violation of cell lineage restriction compartments in the chick hindbrain. Development, 120, 1347–1356. Birgbauer, E., Sechrist, J., Bronner-Fraser, M., & Fraser, S. (1995). Rhombomeric origin and rostrocaudal reassortment of neural crest cells revealed by intravital microscopy. Development, 121, 935–945. Bouchoucha, Y. X., Reingruber, J., Labalette, C., Wassef, M. A., Thierion, E., DesmarquetTrin Dinh, C., et al. (2013). Dissection of a Krox20 positive feedback loop driving cell fate choices in hindbrain patterning. Molecular Systems Biology, 9, 690. Calzolari, S., Terriente, J., & Pujades, C. (2014). Cell segregation in the vertebrate hindbrain relies on actomyosin cables located at the interhombomeric boundaries. The EMBO Journal, 33, 686–701. Cambronero, F., Ariza-McNaughton, L., Wiedemann, L. M., & Krumlauf, R. (2020). Interrhombomeric interactions reveal roles for fibroblast growth factors signaling in segmental regulation of EphA4 expression. Developmental Dynamics, 249, 354–368. Cayuso, J., Xu, Q., Addison, M., & Wilkinson, D. G. (2019). Actomyosin regulation by Eph receptor signaling couples boundary cell formation to border sharpness. eLife, 8, e49696. Cayuso, J., Xu, Q., & Wilkinson, D. G. (2015). Mechanisms of boundary formation by Eph receptor and ephrin signaling. Developmental Biology, 401, 122–131. Chandrasekhar, A. (2004). Turning heads: Development of vertebrate branchiomotor neurons. Developmental Dynamics, 229, 143–161. Chen, J. D., & Evans, R. M. (1995). A transcriptional co-repressor that interacts with nuclear hormone receptors. Nature, 377, 454–457. Cheng, Y. C., Amoyel, M., Qiu, X., Jiang, Y. J., Xu, Q., & Wilkinson, D. G. (2004). Notch activation regulates the segregation and differentiation of rhombomere boundary cells in the zebrafish hindbrain. Developmental Cell, 6, 539–550. Choe, S. K., Vlachakis, N., & Sagerstrom, C. G. (2002). Meis family proteins are required for hindbrain development in the zebrafish. Development, 129, 585–595. Choe, S. K., Zhang, X., Hirsch, N., Straubhaar, J., & Sagerstrom, C. G. (2011). A screen for hoxb1-regulated genes identifies ppp1r14al as a regulator of the rhombomere 4 Fgf-signaling center. Developmental Biology, 358, 356–367. Chomette, D., Frain, M., Cereghini, S., Charnay, P., & Ghislain, J. (2006). Krox20 hindbrain cis-regulatory landscape: Interplay between multiple long-range initiation and autoregulatory elements. Development, 133, 1253–1262. Clarke, J. D., & Lumsden, A. (1993). Segmental repetition of neuronal phenotype sets in the chick embryo hindbrain. Development, 118, 151–162.

Hindbrain segmentation GRN

195

Darras, S., Fritzenwanker, J. H., Uhlinger, K. R., Farrelly, E., Pani, A. M., Hurley, I. A., et al. (2018). Anteroposterior axis patterning by early canonical Wnt signaling during hemichordate development. PLoS Biology, 16, e2003698. Dasen, J. S., & Jessell, T. M. (2009). Hox networks and the origins of motor neuron diversity. Current Topics in Developmental Biology, 88, 169–200. Davenne, M., Maconochie, M. K., Neun, R., Pattyn, A., Chambon, P., Krumlauf, R., et al. (1999). Hoxa2 and Hoxb2 control dorsoventral patterns of neuronal development in the rostral hindbrain. Neuron, 22, 677–691. De Kumar, B., Parker, H. J., Parrish, M. E., Lange, J. J., Slaughter, B. D., Unruh, J. R., et al. (2017). Dynamic regulation of Nanog and stem cell-signaling pathways by Hoxa1 during early neuro-ectodermal differentiation of ES cells. Proceedings of the National Academy of Sciences of the United States of America, 114, 5838–5845. De Kumar, B., Parker, H. J., Paulson, A., Parrish, M. E., Pushel, I., Singh, N. P., et al. (2017). HOXA1 and TALE proteins display cross-regulatory interactions and form a combinatorial binding code on HOXA1 targets. Genome Research, 27, 1501–1512. De Kumar, B., Parker, H. J., Paulson, A., Parrish, M. E., Zeitlinger, J., & Krumlauf, R. (2017). Hoxa1 targets signaling pathways during neural differentiation of ES cells and mouse embryogenesis. Developmental Biology, 432, 151–164. De Kumar, B., Parrish, M. E., Slaughter, B. D., Unruh, J. R., Gogol, M., Seidel, C., et al. (2015). Analysis of dynamic changes in retinoid-induced transcription and epigenetic profiles of murine Hox clusters in ES cells. Genome Research, 25, 1229–1243. Deschamps, J., & Duboule, D. (2017). Embryonic timing, axial stem cells, chromatin dynamics, and the Hox clock. Genes & Development, 31, 1406–1416. Deschamps, J., & van Nes, J. (2005). Developmental regulation of the Hox genes during axial morphogenesis in the mouse. Development, 132, 2931–2942. Diez del Corral, R., Olivera-Martinez, I., Goriely, A., Gale, E., Maden, M., & Storey, K. (2003). Opposing FGF and retinoid pathways control ventral neural pattern, neuronal differentiation, and segmentation during body axis extension. Neuron, 40, 65–79. Diez del Corral, R., & Storey, K. G. (2004). Opposing FGF and retinoid pathways: A signalling switch that controls differentiation and patterning onset in the extending vertebrate body axis. BioEssays, 26, 857–869. Donaldson, I. J., Amin, S., Hensman, J. J., Kutejova, E., Rattray, M., Lawrence, N., et al. (2012). Genome-wide occupancy links Hoxa2 to Wnt-beta-catenin signaling in mouse embryonic development. Nucleic Acids Research, 40, 3990–4001. Duboule, D., & Dolle, P. (1989). The structural and functional organization of the murine HOX gene family resembles that of Drosophila homeotic genes. The EMBO Journal, 8, 1497–1505. Duester, G. (2008). Retinoic acid synthesis and signaling during early organogenesis. Cell, 134, 921–931. Dupe, V., Davenne, M., Brocard, J., Dolle, P., Mark, M., Dierich, A., et al. (1997). In vivo functional analysis of the Hoxa-1 30 retinoic acid response element (3’RARE). Development, 124, 399–410. Evans, R. M., & Mangelsdorf, D. J. (2014). Nuclear receptors, RXR, and the big bang. Cell, 157, 255–266. Ferretti, E., Cambronero, F., T€ umpel, S., Longobardi, E., Wiedemann, L. M., Blasi, F., et al. (2005). Hoxb1 enhancer and control of rhombomere 4 expression: Complex interplay between PREP1-PBX1-HOXB1 binding sites. Molecular and Cellular Biology, 25, 8541–8552. Ferretti, E., Marshall, H., P€ opperl, H., Maconochie, M., Krumlauf, R., & Blasi, F. (2000). Segmental expression of Hoxb2 in r4 requires two separate sites that integrate cooperative interactions between Prep1, Pbx and Hox proteins. Development, 127, 155–166.

196

Hugo J. Parker and Robb Krumlauf

Fraser, S., Keynes, R., & Lumsden, A. (1990). Segmentation in the chick embryo hindbrain is defined by cell lineage restrictions. Nature, 344, 431–435. Gaufo, G. O., Thomas, K. R., & Capecchi, M. R. (2003). Hox3 genes coordinate mechanisms of genetic suppression and activation in the generation of branchial and somatic motoneurons. Development, 130, 5191–5201. Gaunt, S. J. (1991). Expression patterns of mouse Hox genes: Clues to an understanding of developmental and evolutionary strategies. BioEssays, 13, 505–513. Gavalas, A., Trainor, P., Ariza-McNaughton, L., & Krumlauf, R. (2001). Synergy between Hoxa1 and Hoxb1: The relationship between arch patterning and the generation of cranial neural crest. Development, 128, 3017–3027. Gendron-Maguire, M., Mallo, M., Zhang, M., & Gridley, T. (1993). Hoxa-2 mutant mice exhibit homeotic transformation of skeletal elements derived from cranial neural crest. Cell, 75, 1317–1331. Gilland, E., & Baker, R. (2005). Evolutionary patterns of cranial nerve efferent nuclei in vertebrates. Brain, Behavior and Evolution, 66, 234–254. Glass, C. K., & Rosenfeld, M. G. (2000). The coregulator exchange in transcriptional functions of nuclear receptors. Genes & Development, 14, 121–141. Godsave, S., Dekker, E. J., Holling, T., Pannese, M., Boncinelli, E., & Durston, A. (1994). Expression patterns of Hoxb genes in the Xenopus embryo suggest roles in anteroposterior specification of the hindbrain and in dorsoventral patterning of the mesoderm. Developmental Biology, 166, 465–476. Golding, J. P., Trainor, P., Krumlauf, R., & Gassmann, M. (2000). Defects in pathfinding by cranial neural crest cells in mice lacking the Neuregulin receptor ErbB4. Nature Cell Biology, 2, 103–109. Gould, A., Itasaki, N., & Krumlauf, R. (1998). Initiation of rhombomeric Hoxb4 expression requires induction by somites and a retinoid pathway. Neuron, 21, 39–51. Gould, A., Morrison, A., Sproat, G., White, R. A., & Krumlauf, R. (1997). Positive crossregulation and enhancer sharing: Two mechanisms for specifying overlapping Hox expression patterns. Genes & Development, 11, 900–913. Gouti, M., & Gavalas, A. (2008). Hoxb1 controls cell fate specification and proliferative capacity of neural stem and progenitor cells. Stem Cells, 26, 1985–1997. Graham, A., Maden, M., & Krumlauf, R. (1991). The murine Hox-2 genes display dynamic dorsoventral patterns of expression during central nervous system development. Development, 112, 255–264. Graham, A., Papalopulu, N., & Krumlauf, R. (1989). The murine and Drosophila homeobox gene complexes have common features of organization and expression. Cell, 57, 367–378. Green, S. A., Simoes-Costa, M., & Bronner, M. E. (2015). Evolution of vertebrates as viewed from the crest. Nature, 520, 474–482. Green, S. A., Uy, B. R., & Bronner, M. E. (2017). Ancient evolutionary origin of vertebrate enteric neurons from trunk-derived neural crest. Nature, 544, 88–91. Guthrie, S., & Lumsden, A. (1991). Formation and regeneration of rhombomere boundaries in the developing chick hindbrain. Development, 112, 221–229. Guthrie, S., Prince, V., & Lumsden, A. (1993). Selective dispersal of avian rhombomere cells in orthotopic and heterotopic grafts. Development, 118, 527–538. Helmbacher, F., Pujades, C., Desmarquet, C., Frain, M., Rijli, F. M., Chambon, P., et al. (1998). Hoxa1 and Krox20 synergize to control the development of rhombomere 3. Development, 125, 4739–4748. Hernandez, R. E., Putzke, A. P., Myers, J. P., Margaretha, L., & Moens, C. B. (2007). Cyp26 enzymes generate the retinoic acid response pattern necessary for hindbrain development. Development, 134, 177–187.

Hindbrain segmentation GRN

197

Hernandez, R. E., Rikhof, H. A., Bachmann, R., & Moens, C. B. (2004). vhnf1 integrates global RA patterning and local FGF signals to direct posterior hindbrain development in zebrafish. Development, 131, 4511–4520. Heyman, I., Faissner, A., & Lumsden, A. (1995). Cell matrix specialisations of rhombomere boundaries. Developmental Dynamics, 204, 301–315. Hoegg, S., & Meyer, A. (2005). Hox clusters as models for vertebrate genome evolution. Trends in Genetics, 21, 421–424. Horlein, A. J., Naar, A. M., Heinzel, T., Torchia, J., Gloss, B., Kurokawa, R., et al. (1995). Ligand-independent repression by the thyroid hormone receptor mediated by a nuclear receptor co-repressor. Nature, 377, 397–404. Hunt, P., Gulisano, M., Cook, M., Sham, M. H., Faiella, A., Wilkinson, D., et al. (1991). A distinct Hox code for the branchial region of the vertebrate head. Nature, 353, 861–864. Irving, C., & Mason, I. (2000). Signalling by FGF8 from the isthmus patterns anterior hindbrain and establishes the anterior limit of Hox gene expression. Development, 127, 177–186. Joyner, A. L. (1996). Engrailed, Wnt and Pax genes regulate midbrain—Hindbrain development. Trends in Genetics, 12, 15–20. Kelly, O. G., Pinson, K. I., & Skarnes, W. C. (2004). The Wnt co-receptors Lrp5 and Lrp6 are essential for gastrulation in mice. Development, 131, 2803–2815. Kitazawa, T., Fujisawa, K., Narboux-Neme, N., Arima, Y., Kawamura, Y., Inoue, T., et al. (2015). Distinct effects of Hoxa2 overexpression in cranial neural crest populations reveal that the mammalian hyomandibular-ceratohyal boundary maps within the styloid process. Developmental Biology, 402, 162–174. Kmita, M., & Duboule, D. (2003). Organizing axes in time and space; 25 years of colinear tinkering. Science, 301, 331–333. Krumlauf, R. (1994). Hox genes in vertebrate development. Cell, 78, 191–201. Krumlauf, R. (2016). Hox genes and the hindbrain: A study in segments. Current Topics in Developmental Biology, 116, 581–596. Labalette, C., Wassef, M. A., Desmarquet-Trin Dinh, C., Bouchoucha, Y. X., Le Men, J., Charnay, P., et al. (2015). Molecular dissection of segment formation in the developing hindbrain. Development, 142, 185–195. Lavinsky, R. M., Jepsen, K., Heinzel, T., Torchia, J., Mullen, T. M., Schiff, R., et al. (1998). Diverse signaling pathways modulate nuclear receptor recruitment of N-CoR and SMRT complexes. Proceedings of the National Academy of Sciences of the United States of America, 95, 2920–2925. Le Douarin, N., & Kalcheim, C. (1999). The neural crest (2nd ed.). Cambridge, UK; New York, NY, USA: Cambridge University Press. Lee, S. M., Danielian, P. S., Fritzsch, B., & McMahon, A. P. (1997). Evidence that FGF8 signalling from the midbrain-hindbrain junction regulates growth and polarity in the developing midbrain. Development, 124, 959–969. Lewis, E. B. (1978). A gene complex controlling segmentation in Drosophila. Nature, 276, 565–570. Lin, C., Garrett, A. S., De Kumar, B., Smith, E. R., Gogol, M., Seidel, C., et al. (2011). Dynamic transcriptional events in embryonic stem cells mediated by the super elongation complex (SEC). Genes & Development, 25, 1486–1498. Liu, A., Li, J. Y., Bromleigh, C., Lao, Z., Niswander, L. A., & Joyner, A. L. (2003). FGF17b and FGF18 have different midbrain regulatory properties from FGF8b or activated FGF receptors. Development, 130, 6175–6185. Liu, P., Wakamiya, M., Shea, M. J., Albrecht, U., Behringer, R. R., & Bradley, A. (1999). Requirement for Wnt3 in vertebrate axis formation. Nature Genetics, 22, 361–365.

198

Hugo J. Parker and Robb Krumlauf

Lumsden, A. (2004). Segmentation and compartition in the early avian hindbrain. Mechanisms of Development, 121, 1081–1088. Lumsden, A., & Keynes, R. (1989). Segmental patterns of neuronal development in the chick hindbrain. Nature, 337, 424–428. Lumsden, A., & Krumlauf, R. (1996). Patterning the vertebrate neuraxis. Science, 274, 1109–1115. Maconochie, M., Krishnamurthy, R., Nonchev, S., Meier, P., Manzanares, M., Mitchell, P. J., et al. (1999). Regulation of Hoxa2 in cranial neural crest cells involves members of the AP-2 family. Development, 126, 1483–1494. Maconochie, M. K., Nonchev, S., Studer, M., Chan, S. K., Popperl, H., Sham, M. H., et al. (1997). Cross-regulation in the mouse HoxB complex: The expression of Hoxb2 in rhombomere 4 is regulated by Hoxb1. Genes & Development, 11, 1885–1896. Makki, N., & Capecchi, M. R. (2010). Hoxa1 lineage tracing indicates a direct role for Hoxa1 in the development of the inner ear, the heart, and the third rhombomere. Developmental Biology, 341, 499–509. Makki, N., & Capecchi, M. R. (2011). Identification of novel Hoxa1 downstream targets regulating hindbrain, neural crest and inner ear development. Developmental Biology, 357, 295–304. Mangelsdorf, D. J., Thummel, C., Beato, M., Herrlich, P., Schutz, G., Umesono, K., et al. (1995). The nuclear receptor superfamily: The second decade. Cell, 83, 835–839. Mann, R. S., Lelli, K. M., & Joshi, R. (2009). Hox specificity unique roles for cofactors and collaborators. Current Topics in Developmental Biology, 88, 63–101. Manzanares, M., Bel-Vialer, S., Ariza-McNaughton, L., Ferretti, E., Marshall, H., Maconochie, M. K., et al. (2001). Independent regulation of initiation and maintenance phases of Hoxa3 expression in the vertebrate hindbrain involves auto and crossregulatory mechanisms. Development, 128, 3595–3607. Manzanares, M., Cordes, S., Ariza-McNaughton, L., Sadl, V., Maruthainar, K., Barsh, G., et al. (1999). Conserved and distinct roles of kreisler in regulation of the paralogous Hoxa3 and Hoxb3 genes. Development, 126, 759–769. Manzanares, M., Cordes, S., Kwan, C.-T., Sham, M.-H., Barsh, G., & Krumlauf, R. (1997). Segmental regulation of Hoxb3 by kreisler. Nature, 387, 191–195. Manzanares, M., Nardelli, J., Gilardi-Hebenstreit, P., Marshall, H., Giudicelli, F., MartinezPastor, M. T., et al. (2002). Krox20 and kreisler co-operate in the transcriptional control of segmental expression of Hoxb3 in the developing hindbrain. The EMBO Journal, 21, 365–376. Manzanares, M., Wada, H., Itasaki, N., Trainor, P. A., Krumlauf, R., & Holland, P. W. (2000). Conservation and elaboration of Hox gene regulation during evolution of the vertebrate head. Nature, 408, 854–857. Marshall, H., Studer, M., Popperl, H., Aparicio, S., Kuroiwa, A., Brenner, S., et al. (1994). A conserved retinoic acid response element required for early expression of the homeobox gene Hoxb-1. Nature, 370, 567–571. Martik, M. L., & Bronner, M. E. (2017). Regulatory logic underlying diversification of the neural crest. Trends in Genetics, 33, 715–727. Martik, M. L., Gandhi, S., Uy, B. R., Gillis, J. A., Green, S. A., Simoes-Costa, M., et al. (2019). Evolution of the new head by gradual acquisition of neural crest regulatory circuits. Nature, 574, 675–678. Mason, I. (2007). Initiation to end point: The multiple roles of fibroblast growth factors in neural development. Nature Reviews. Neuroscience, 8, 583–596. Mason, I., Chambers, D., Shamim, H., Walshe, J., & Irving, C. (2000). Regulation and function of FGF8 in patterning of midbrain and anterior hindbrain. Biochemistry and Cell Biology, 78, 577–584.

Hindbrain segmentation GRN

199

Maves, L., Jackman, W., & Kimmel, C. B. (2002). FGF3 and FGF8 mediate a rhombomere 4 signaling activity in the zebrafish hindbrain. Development, 129, 3825–3837. McGinnis, W., & Krumlauf, R. (1992). Homeobox genes and axial patterning. Cell, 68, 283–302. McNulty, C. L., Peres, J. N., Bardine, N., van den Akker, W. M., & Durston, A. J. (2005). Knockdown of the complete Hox paralogous group 1 leads to dramatic hindbrain and neural crest defects. Development, 132, 2861–2871. Mechta-Grigoriou, F., Garel, S., & Charnay, P. (2000). Nab proteins mediate a negative feedback loop controlling Krox-20 activity in the developing hindbrain. Development, 127, 119–128. Mellitzer, G., Xu, Q., & Wilkinson, D. (1999). Eph receptors and ephrins restrict cell intermingling and communication. Nature, 400, 77–81. Merabet, S., & Mann, R. S. (2016). To be specific or not: The critical relationship between Hox and TALE proteins. Trends in Genetics, 32, 334–347. Metzis, V., Steinhauser, S., Pakanavicius, E., Gouti, M., Stamataki, D., Ivanovitch, K., et al. (2018). Nervous system regionalization entails axial allocation before neural differentiation. Cell, 175, 1105-1118.e1117. Meulemans, D., & Bronner-Fraser, M. (2004). Gene-regulatory interactions in neural crest evolution and development. Developmental Cell, 7, 291–299. Millet, S., Campbell, K., Epstein, D. J., Losos, K., Harris, E., & Joyner, A. L. (1999). A role for Gbx2 in repression of Otx2 and positioning the mid/hindbrain organizer. Nature, 401, 161–164. Minoux, M., & Rijli, F. M. (2010). Molecular mechanisms of cranial neural crest cell migration and patterning in craniofacial development. Development, 137, 2605–2621. Moens, C. B., & Prince, V. E. (2002). Constructing the hindbrain: Insights from the zebrafish. Developmental Dynamics, 224, 1–17. Moens, C. B., & Selleri, L. (2006). Hox cofactors in vertebrate development. Developmental Biology, 291, 193–206. Molotkova, N., Molotkov, A., Sirbu, I. O., & Duester, G. (2005). Requirement of mesodermal retinoic acid generated by Raldh2 for posterior neural transformation. Mechanisms of Development, 122, 145–155. Morrison, A., Moroni, M. C., Ariza-McNaughton, L., Krumlauf, R., & Mavilio, F. (1996). In vitro and transgenic analysis of a human HOXD4 retinoid-responsive enhancer. Development, 122, 1895–1907. Neijts, R., Amin, S., van Rooijen, C., & Deschamps, J. (2017). Cdx is crucial for the timing mechanism driving colinear Hox activation and defines a trunk segment in the Hox cluster topology. Developmental Biology, 422, 146–154. Neijts, R., Amin, S., van Rooijen, C., Tan, S., Creyghton, M. P., de Laat, W., et al. (2016). Polarized regulatory landscape and Wnt responsiveness underlie Hox activation in embryos. Genes & Development, 30, 1937–1942. Neijts, R., & Deschamps, J. (2017). At the base of colinear Hox gene expression: Cis-features and trans-factors orchestrating the initial phase of Hox cluster activation. Developmental Biology, 428, 293–299. Nieto, M. A., Gilardi-Hebenstreit, P., Charnay, P., & Wilkinson, D. G. (1992). A receptor protein tyrosine kinase implicated in the segmental patterning of the hindbrain and mesoderm. Development, 116, 1137–1150. Nolte, C., De Kumar, B., & Krumlauf, R. (2019). Hox genes: Downstream “effectors” of retinoic acid signaling in vertebrate embryogenesis. Genesis, 57, e23306. Nolte, C., Jinks, T., Wang, X., Martinez Pastor, M. T., & Krumlauf, R. (2013). Shadow enhancers flanking the HoxB cluster direct dynamic Hox expression in early heart and endoderm development. Developmental Biology, 383, 158–173.

200

Hugo J. Parker and Robb Krumlauf

Nolte, C., Rastegar, M., Amores, A., Bouchard, M., Grote, D., Maas, R., et al. (2006). Stereospecificity and PAX6 function direct Hoxd4 neural enhancer activity along the antero-posterior axis. Developmental Biology, 299, 582–593. Nonchev, S., Maconochie, M., Vesque, C., Aparicio, S., Ariza-McNaughton, L., Manzanares, M., et al. (1996). The conserved role of Krox-20 in directing Hox gene expression during vertebrate hindbrain segmentation. Proceedings of the National Academy of Sciences of the United States of America, 93, 9339–9345. Nonchev, S., Vesque, C., Maconochie, M., Seitanidou, T., Ariza-McNaughton, L., Frain, M., et al. (1996). Segmental expression of Hoxa-2 in the hindbrain is directly regulated by Krox-20. Development, 122, 543–554. Nordstrom, U., Jessell, T. M., & Edlund, T. (2002). Progressive induction of caudal neural character by graded Wnt signaling. Nature Neuroscience, 5, 525–532. Nordstrom, U., Maier, E., Jessell, T. M., & Edlund, T. (2006). An early role for WNT signaling in specifying neural patterns of Cdx and Hox gene expression and motor neuron subtype identity. PLoS Biology, 4, e252. Oosterveen, T., Niederreither, K., Dolle, P., Chambon, P., Meijlink, F., & Deschamps, J. (2003). Retinoids regulate the anterior expression boundaries of 5’ Hoxb genes in posterior hindbrain. The EMBO Journal, 22, 262–269. Parker, H. J., Bronner, M. E., & Krumlauf, R. (2014). A Hox regulatory network of hindbrain segmentation is conserved to the base of vertebrates. Nature, 514, 490–493. Parker, H. J., Bronner, M. E., & Krumlauf, R. (2016). The vertebrate Hox gene regulatory network for hindbrain segmentation: Evolution and diversification: Coupling of a Hox gene regulatory network to hindbrain segmentation is an ancient trait originating at the base of vertebrates. BioEssays, 38, 526–538. Parker, H. J., Bronner, M. E., & Krumlauf, R. (2019). An atlas of anterior hox gene expression in the embryonic sea lamprey head: Hox-code evolution in vertebrates. Developmental Biology, 453, 19–33. Parker, H. J., De Kumar, B., Green, S. A., Prummel, K. D., Hess, C., Kaufman, C. K., et al. (2019). A Hox-TALE regulatory circuit for neural crest patterning is conserved across vertebrates. Nature Communications, 10, 1189. Parker, H. J., & Krumlauf, R. (2017). Segmental arithmetic: Summing up the Hox gene regulatory network for hindbrain development in chordates. Wiley Interdisciplinary Reviews: Developmental Biology, 6, e286. Parker, H. J., Pushel, I., & Krumlauf, R. (2018). Coupling the roles of Hox genes to regulatory networks patterning cranial neural crest. Developmental Biology, 444(Suppl. 1), S67–S78. Pattyn, A., Vallstedt, A., Dias, J. M., Samad, O. A., Krumlauf, R., Rijli, F. M., et al. (2003). Coordinated temporal and spatial control of motor neuron and serotonergic neuron generation from a common pool of CNS progenitors. Genes & Development, 17, 729–737. Penkov, D., Mateos San Martin, D., Fernandez-Diaz, L. C., Rossello, C. A., Torroja, C., Sanchez-Cabo, F., et al. (2013). Analysis of the DNA-binding profile and function of TALE homeoproteins reveals their specialization and specific interactions with Hox genes/proteins. Cell Reports, 3, 1321–1333. Peretz, Y., Eren, N., Kohl, A., Hen, G., Yaniv, K., Weisinger, K., et al. (2016). A new role of hindbrain boundaries as pools of neural stem/progenitor cells regulated by Sox2. BMC Biology, 14, 57. Popperl, H., Bienz, M., Studer, M., Chan, S. K., Aparicio, S., Brenner, S., et al. (1995). Segmental expression of Hoxb-1 is controlled by a highly conserved autoregulatory loop dependent upon exd/pbx. Cell, 81, 1031–1042. Pownall, M. E., Isaacs, H. V., & Slack, J. M. (1998). Two phases of Hox gene regulation during early Xenopus development. Current Biology, 8, 673–676.

Hindbrain segmentation GRN

201

Pownall, M., Tucker, A., Slack, J., & Isaacs, H. (1996). eFGF, Xcad3 and Hox genes form a molecular pathway that establishes the anteroposterior axis in Xenopus. Development, 122, 3881–3892. Prince, V. E., Moens, C. B., Kimmel, C. B., & Ho, R. K. (1998). Zebrafish hox genes: Expression in the hindbrain region of wild-type and mutants of the segmentation gene, valentino. Development, 125, 393–406. Qian, P., De Kumar, B., He, X. C., Nolte, C., Gogol, M., Ahn, Y., et al. (2018). Retinoidsensitive epigenetic regulation of the Hoxb cluster maintains normal hematopoiesis and inhibits leukemogenesis. Cell Stem Cell, 22(740–754), e747. Reifers, F., Bohli, H., Walsh, E. C., Crossley, P. H., Stainier, D. Y., & Brand, M. (1998). Fgf8 is mutated in zebrafish acerebellar (ace) mutants and is required for maintenance of midbrain-hindbrain boundary development and somitogenesis. Development, 125, 2381–2395. Rezsohazy, R., Saurin, A. J., Maurel-Zaffran, C., & Graba, Y. (2015). Cellular and molecular insights into Hox protein action. Development, 142, 1212–1227. Rhinn, M., & Dolle, P. (2012). Retinoic acid signalling during development. Development, 139, 843–858. Rhinn, M., Lun, K., Luz, M., Werner, M., & Brand, M. (2005). Positioning of the midbrainhindbrain boundary organizer through global posteriorization of the neuroectoderm mediated by Wnt8 signaling. Development, 132, 1261–1272. Rijli, F. M., Mark, M., Lakkaraju, S., Dierich, A., Dolle, P., & Chambon, P. (1993). A homeotic transformation is generated in the rostral branchial region of the head by disruption of Hoxa-2, which acts as a selector gene. Cell, 75, 1333–1349. Rohrschneider, M. R., Elsen, G. E., & Prince, V. E. (2007). Zebrafish Hoxb1a regulates multiple downstream genes including prickle1b. Developmental Biology, 309, 358–372. Rothstein, M., Bhattacharya, D., & Simoes-Costa, M. (2018). The molecular basis of neural crest axial identity. Developmental Biology, 444(Suppl. 1), S170–S180. Santagati, F., Minoux, M., Ren, S. Y., & Rijli, F. M. (2005). Temporal requirement of Hoxa2 in cranial neural crest skeletal morphogenesis. Development, 132, 4927–4936. Sauka-Spengler, T., Meulemans, D., Jones, M., & Bronner-Fraser, M. (2007). Ancient evolutionary origin of the neural crest gene regulatory network. Developmental Cell, 13, 405–420. Schilling, T. F., Prince, V., & Ingham, P. W. (2001). Plasticity in zebrafish hox expression in the hindbrain and cranial neural crest. Developmental Biology, 231, 201–216. Schilling, T. F., Sosnik, J., & Nie, Q. (2016). Visualizing retinoic acid morphogen gradients. Methods in Cell Biology, 133, 139–163. Schulte, D., & Frank, D. (2014). TALE transcription factors during early development of the vertebrate brain and eye. Developmental Dynamics, 243, 99–116. Sela-Donenfeld, D., Kayam, G., & Wilkinson, D. G. (2009). Boundary cells regulate a switch in the expression of FGF3 in hindbrain rhombomeres. BMC Developmental Biology, 9, 16. Sela-Donenfeld, D., & Wilkinson, D. G. (2005). Eph receptors: Two ways to sharpen boundaries. Current Biology, 15, R210–R212. Selleri, L., Zappavigna, V., & Ferretti, E. (2019). ‘Building a perfect body’: Control of vertebrate organogenesis by PBX-dependent regulatory networks. Genes & Development, 33, 258–275. Serpente, P., Tumpel, S., Ghyselinck, N. B., Niederreither, K., Wiedemann, L. M., Dolle, P., et al. (2005). Direct crossregulation between retinoic acid receptor {beta} and Hox genes during hindbrain segmentation. Development, 132, 503–513. Sham, M. H., Vesque, C., Nonchev, S., Marshall, H., Frain, M., Das Gupta, R. D., et al. (1993). The zinc finger gene Krox-20 regulates Hoxb-2 (Hox2.8) during hindbrain segmentation. Cell, 72, 183–196.

202

Hugo J. Parker and Robb Krumlauf

Sharpe, J., Nonchev, S., Gould, A., Whiting, J., & Krumlauf, R. (1998). Selectivity, sharing and competitive interactions in the regulation of Hoxb genes. The EMBO Journal, 17, 1788–1798. Shimozono, S., Iimura, T., Kitaguchi, T., Higashijima, S., & Miyawaki, A. (2013). Visualization of an endogenous retinoic acid gradient across embryonic development. Nature, 496, 363–366. Simeone, A., Acampora, D., Arcioni, L., Andrews, P. W., Boncinelli, E., & Mavilio, F. (1990). Sequential activation of HOX2 homeobox genes by retinoic acid in human embryonal carcinoma cells. Nature, 346, 763–766. Simeone, A., Acampora, D., Nigro, V., Faiella, A., D’Esposito, M., Stornaiuolo, A., et al. (1991). Differential regulation by retinoic acid of the homeobox genes of the four HOX loci in human embryonal carcinoma cells. Mechanisms of Development, 33, 215–227. Simoes-Costa, M., & Bronner, M. E. (2015). Establishing neural crest identity: A gene regulatory recipe. Development, 142, 242–257. Simoes-Costa, M., & Bronner, M. E. (2016). Reprogramming of avian neural crest axial identity and cell fate. Science, 352, 1570–1573. Sirbu, I. O., Gresh, L., Barra, J., & Duester, G. (2005). Shifting boundaries of retinoic acid activity control hindbrain segmental gene expression. Development, 132, 2611–2622. Slattery, M., Riley, T., Liu, P., Abe, N., Gomez-Alcala, P., Dror, I., et al. (2011). Cofactor binding evokes latent differences in DNA binding specificity between Hox proteins. Cell, 147, 1270–1282. Studer, M., Gavalas, A., Marshall, H., Ariza-McNaughton, L., Rijli, F. M., Chambon, P., et al. (1998). Genetic interactions between Hoxa1 and Hoxb1 reveal new roles in regulation of early hindbrain patterning. Development, 125, 1025–1036. Studer, M., Popperl, H., Marshall, H., Kuroiwa, A., & Krumlauf, R. (1994). Role of a conserved retinoic acid response element in rhombomere restriction of Hoxb-1. Science, 265, 1728–1732. Tambalo, M., Mitter, R., & Wilkinson, D. G. (2020). A single cell transcriptome atlas of the developing zebrafish hindbrain. Development, 147, dev184143. https://doi.org/10.1242/ dev.184143. Theil, T., Frain, M., Gilardi-Hebenstreit, P., Flenniken, A., Charnay, P., & Wilkinson, D. G. (1998). Segmental expression of the EphA4 (Sek-1) receptor tyrosine kinase in the hindbrain is under direct transcriptional control of Krox-20. Development, 125, 443–452. Thierion, E., Le Men, J., Collombet, S., Hernandez, C., Coulpier, F., Torbey, P., et al. (2017). Krox20 hindbrain regulation incorporates multiple modes of cooperation between cis-acting elements. PLoS Genetics, 13, e1006903. Torbey, P., Thierion, E., Collombet, S., de Cian, A., Desmarquet-Trin-Dinh, C., Dura, M., et al. (2018). Cooperation, cis-interactions, versatility and evolutionary plasticity of multiple cis-acting elements underlie krox20 hindbrain regulation. PLoS Genetics, 14, e1007581. Trainor, P. A., Ariza-McNaughton, L., & Krumlauf, R. (2002). Role of the isthmus and FGFs in resolving the paradox of neural crest plasticity and prepatterning. Science, 295, 1288–1291. Trainor, P. A., Bronner-Fraser, M., & Krumlauf, R. (2004). Neural crest cells. In R. P. Lanza (Ed.), Handbook of stem cells (pp. 219–232). Boston, MA: Elsevier Academic. Trainor, P., & Krumlauf, R. (2000a). Plasticity in mouse neural crest cells reveals a new patterning role for cranial mesoderm. Nature Cell Biology, 2, 96–102. Trainor, P. A., & Krumlauf, R. (2000b). Patterning the cranial neural crest: Hindbrain segmentation and Hox gene plasticity. Nature Reviews. Neuroscience, 1, 116–124. Trainor, P. A., & Krumlauf, R. (2001). Hox genes, neural crest cells and branchial arch patterning. Current Opinion in Cell Biology, 13, 698–705.

Hindbrain segmentation GRN

203

T€ umpel, S., Cambronero, F., Ferretti, E., Blasi, F., Wiedemann, L. M., & Krumlauf, R. (2007). Expression of Hoxa2 in rhombomere 4 is regulated by a conserved crossregulatory mechanism dependent upon Hoxb1. Developmental Biology, 302, 646–660. T€ umpel, S., Cambronero, F., Sims, C., Krumlauf, R., & Wiedemann, L. M. (2008). A regulatory module embedded in the coding region of Hoxa2 controls expression in rhombomere 2. Proceedings of the National Academy of Sciences of the United States of America, 105, 20077–20082. Tumpel, S., Wiedemann, L. M., & Krumlauf, R. (2009). Hox genes and segmentation of the vertebrate hindbrain. Current Topics in Developmental Biology, 88, 103–137. Tvrdik, P., & Capecchi, M. R. (2006). Reversal of hox1 gene subfunctionalization in the mouse. Developmental Cell, 11, 239–250. Umesono, K., Murakami, K. K., Thompson, C. C., & Evans, R. M. (1991). Direct repeats as selective response elements for the thyroid hormone, retinoic acid, and vitamin D3 receptors. Cell, 65, 1255–1266. Vitobello, A., Ferretti, E., Lampe, X., Vilain, N., Ducret, S., Ori, M., et al. (2011). Hox and Pbx factors control retinoic acid synthesis during hindbrain segmentation. Developmental Cell, 20, 469–482. Voltes, A., Hevia, C. F., Engel-Pizcueta, C., Dingare, C., Calzolari, S., Terriente, J., et al. (2019). Yap/Taz-TEAD activity links mechanical cues to progenitor cell behavior during zebrafish hindbrain segmentation. Development, 146, dev176735. Wada, H., Escriva, H., Zhang, S., & Laudet, V. (2006). Conserved RARE localization in amphioxus Hox clusters and implications for Hox code evolution in the vertebrate neural crest. Developmental Dynamics, 235, 1522–1531. Waskiewicz, A. J., Rikhof, H. A., Hernandez, R. E., & Moens, C. B. (2001). Zebrafish Meis functions to stabilize Pbx proteins and regulate hindbrain patterning. Development, 128, 4139–4151. Waskiewicz, A. J., Rikhof, H. A., & Moens, C. B. (2002). Eliminating zebrafish pbx proteins reveals a hindbrain ground state. Developmental Cell, 3, 723–733. Wassef, M. A., Chomette, D., Pouilhe, M., Stedman, A., Havis, E., Desmarquet-Trin Dinh, C., et al. (2008). Rostral hindbrain patterning involves the direct activation of a Krox20 transcriptional enhancer by Hox/Pbx and Meis factors. Development, 135, 3369–3378. Weisinger, K., Kohl, A., Kayam, G., Monsonego-Ornan, E., & Sela-Donenfeld, D. (2012). Expression of hindbrain boundary markers is regulated by FGF3. Biology Open, 1, 67–74. White, R. J., Nie, Q., Lander, A. D., & Schilling, T. F. (2007). Complex regulation of cyp26a1 creates a robust retinoic acid gradient in the zebrafish embryo. PLoS Biology, 5, e304. White, R. J., & Schilling, T. F. (2008). How degrading: Cyp26s in hindbrain development. Developmental Dynamics, 237, 2775–2790. Wilkinson, D. G. (2018). Establishing sharp and homogeneous segments in the hindbrain. F1000Res, 7, pii: F1000 Faculty Rev-1268. Wilkinson, D. G., Bhatt, S., Cook, M., Boncinelli, E., & Krumlauf, R. (1989). Segmental expression of Hox-2 homoeobox-containing genes in the developing mouse hindbrain. Nature, 341, 405–409. Wong, E. Y., Wang, X. A., Mak, S. S., Sae-Pang, J. J., Ling, K. W., Fritzsch, B., et al. (2011). Hoxb3 negatively regulates Hoxb1 expression in mouse hindbrain patterning. Developmental Biology, 352, 382–392. Xu, Q., Mellitzer, G., Robinson, V., & Wilkinson, D. G. (1999). In vivo cell sorting in complementary segmental domains mediated by Eph receptors and ephrins. Nature, 399, 267–271. Young, T., Rowland, J. E., van de Ven, C., Bialecka, M., Novoa, A., Carapuco, M., et al. (2009). Cdx and Hox genes differentially regulate posterior axial growth in mammalian embryos. Developmental Cell, 17, 516–526.

CHAPTER SEVEN

Logical modeling of cell fate specification—Application to T cell commitment Elisabetta Cacacea, Samuel Collombeta,b, Denis Thieffryb,c,* a

European Molecular Biology Laboratory, Genome Biology Unit, Heidelberg, Germany.
cole Normale Sup
erieure (IBENS), CNRS, Computational Systems Biology Team, Institut de Biologie de l’E
cole Normale Sup
erieure, Universit
e PSL, Paris, France. INSERM, E c Cancer Science Institute of Singapore, National University of Singapore, Singapore, Singapore. *Corresponding author: e-mail address: [email protected] b

Contents 1. Introduction 2. Overview of T cell development 3. Logical modeling of T cell specification 3.1 Delineation of the regulatory graph controlling T cell specification 3.2 Multilevel nodes 3.3 Logical rules 3.4 Computation of model stable states in the wild-type situation 3.5 Computation of model stable states upon environmental or genetic perturbations 3.6 Assessing the reachability of specific expression patterns from specific initial states 3.7 Stochastic simulations 4. Conclusions Acknowledgments References

206 207 211 211 213 215 218 219 223 224 229 232 232

Abstract Boolean approaches and extensions thereof are becoming increasingly popular to model signaling and regulatory networks, including those controlling cell differentiation, pattern formation and embryonic development. Here, we describe a logical modeling framework relying on three steps: the delineation of a regulatory graph, the specification of multilevel components, and the encoding of Boolean rules specifying the behavior of model components depending on the levels or activities of their regulators. Referring to a non-deterministic, asynchronous updating scheme, we present several complementary methods and tools enabling the computation of stable activity patterns, the verification of the reachability of such patterns, as well as the generation of mean temporal evolution curves and the computation of the probabilities to reach distinct activity patterns. Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.02.008

#

2020 Elsevier Inc. All rights reserved.

205

206

Elisabetta Cacace et al.

We apply this logical framework to the regulatory network controlling T lymphocyte specification. This process involves cross-regulations between specific T cell regulatory factors and factors driving alternative differentiation pathways, which remain accessible during the early steps of thymocyte development. Many transcription factors needed for T cell specification are required in other hematopoietic differentiation pathways and are combined in a fine-tuned, time-dependent fashion to achieve T cell commitment. Using the software GINsim, we integrated current knowledge into a dynamical model, which recapitulates the main developmental steps from early progenitors entering the thymus up to T cell commitment, as well as the impact of various documented environmental and genetic perturbations. Our model analysis further enabled the identification of several knowledge gaps. The model, software and whole analysis workflow are provided in computerreadable and executable form to ensure reproducibility and ease extensions.

1. Introduction During metazoan development, cell fate specification and differentiation events are driven by complex interaction networks, which involve cross-talking signaling pathways, transcriptional regulatory circuits and epigenetic remodeling. The delineation of these networks has been recently boosted by the rapid development of powerful pan-genomic methods to profile the epigenome, the transcriptome, the proteome, and the interactome, at the tissue level and even at the single-cell level. A proper understanding of the dynamical behavior of cellular networks requires the integration of massive and diverse datasets into predictive models, which in turn entails the development of rigorous and scalable computational methods and tools. Various modeling frameworks, ranging from differential equations and stochastic models to Boolean networks, have been already applied to model cellular networks and embryonic development (Le Nove`re, 2015). Each of these frameworks has its own assets and limitations. At one side of this spectrum, quantitative models (e.g., based on differential or stochastic equations) require precise information about the mechanisms and kinetic parameters involved. In the absence of such information, predictions made with such models remain rather qualitative. At the other side of the spectrum, Boolean models rely on qualitative information on key regulatory factors and interactions, enabling the reproduction of the most salient dynamical properties of the corresponding network, e.g., alternative stable states, typically representing cell differentiation states, or periodic behavior such as cell cycle or circadian oscillations (Abou-Jaoud
e et al., 2016).

Logical modeling of cell fate specification

207

The choice of a modeling framework depends on the biological processes under study, on the data currently available, as well as on the questions to be addressed. Contrarily to the widespread belief that one needs a deep and comprehensive understanding of a biological system before engaging into its dynamical modeling, we are convinced that an early recourse to formal modeling can be of great help to the biologist to (i) clarify their current understanding of the process under study, (ii) rigorously infer all the consequences of the underlying hypotheses, and (iii) rationally design novel experiments. Whatever the formalism considered, the first step in the modeling of the regulatory network underlying cell fate decisions most often implies the delineation of a graph, where nodes correspond to genes or gene products (e.g., proteins, mRNAs, miRNAs) and arcs represent regulatory interactions (e.g., transcriptional activations or inhibitions, protein modifications) (see, e.g., Longabaugh, Davidson, & Bolouri, 2005). A sophisticated graphical format, called SBGN, has been recently proposed by the systems biology community to standardize the drawing of complex molecular regulatory maps, thereby easing their interpretation and fostering their exchange (Le Nove`re et al., 2009). However, the translation of such maps into dynamical models is far from straightforward and still a field of active research. In this chapter, we present a stepwise approach to model developmental processes, which starts with the delineation of a regulatory graph encompassing all key regulatory factors and interactions driving the process under study. This graph is then supplemented with Boolean rules to generate a predictive dynamical model. The delineation of such qualitative, yet rigorous, dynamical models is greatly facilitated by the availability of user-friendly software, such as GINsim (Naldi, Hernandez, Aboujaoud
e, et al., 2018), including a wide array of analysis functionalities. The resulting models can then be refined with probabilistic transition rates, to enable more quantitative temporal simulations, taking advantage of software tools such as MaBoSS (Stoll et al., 2017). To illustrate our stepwise approach, we present hereafter a comprehensive model of the regulatory network controlling the commitment of pluripotent progenitors into T cells in the mouse thymus.

2. Overview of T cell development T cell development stands out from other hematopoietic lineages differentiations in that (i) it is initiated and controlled by a precise environmental cue, i.e., the Notch ligand DLL-4 supplied by the thymic stroma,

208

Elisabetta Cacace et al.

(ii) it results from the sequential exclusion of alternative fates, whose programs partially overlap with the T-specific program, (iii) it is not directed by a master regulator, but rather by a cocktail of transcription factors that in themselves are not T-specific (Rothenberg, 2019a, 2019b). How these overlapping gene-regulatory modules are rearranged and controlled has been partly elucidated. Analysis of gene expression and transcription factor binding during in vitro and in vivo differentiation of T cells has shed light on key checkpoints of T cell specification (De Obaldia & Bhandoola, 2015; Mingueneau et al., 2013; Yui, Feng, & Rothenberg, 2010; Yui & Rothenberg, 2014; Zhang, Mortazavi, Williams, Wold, & Rothenberg, 2012). Conditional genetic perturbations have contributed to disentangle the relative roles of specific transcription factors and gene regulatory circuits at different stages of development up to commitment (Kueh et al., 2016; Oravecz et al., 2015; Pai et al., 2003; Scripture-Adams et al., 2014). During T cell development, three main phases can be identified, orchestrated by distinct regulatory modules: (i) the exclusion of alternative fates and repression of stem-cell associated genes, leading to a definite commitment; (ii) the generation and selection of clonal T cell receptors (TCRs), which require sequential steps of gene rearrangement and strict quality control checkpoints, and (iii) the selection of effector functions, specified by similar regulatory mechanisms across T cells and innate lymphoid cells (ILCs) (Rothenberg, 2019a). In this modeling study, we focus on the early phase of T cell differentiation (Fig. 1), from the homing of bone marrow precursors into the thymic cortex to the generation of committed pre-T cells, ready to undergo VD(J) recombination, which is governed by recombinases Rag1 and Rag2 and followed by TCR selection. These correspond to the stages ETP (Kit++, CD25
, CD44+), DN2a (Kit++, CD25+, CD44+) and DN2b (Kit+, CD25+, CD44+), where DN stands for double-negative, referring to the surface expression of the CD4 and CD8 antigens. Hence, we do not model the stages of development occurring after commitment, leading to the generation of a functional TCR and to the selection of all T-subset effector functions, corresponding to the DN3 and double-positive (DP) and single-positive (SP) stages (for a review devoted to the modeling of this process, see Outters, Jaeger, Zaarour, & Ferrier, 2015). Hematopoietic precursors with lymphoid potential home in the thymus thanks to chemokine signaling (CCL21 and CCL25) and P-selectin expression (PSGL-1) (Love & Bhandoola, 2011). Composed of cortical and medullary epithelial cells, the thymic microenvironment provides environmental cues that support the development and survival T precursors, such as

Logical modeling of cell fate specification

209

Fig. 1 Early phases of T cell development, from bone marrow precursors to commitment (DN2b stage). HSC, hematopoietic stem cells; LMPP, lymphoid-primed multipotent progenitors; thymus-settling: thymus-settling progenitors; ETP, early T cell precursors; DN2a, double-negative 2a; DN2b, double-negative 2a; DP, double-positive. Dashed line: commitment. Stages occurring between commitment and the double-positive (DP) phase are not depicted. Main regulations underpinning exclusion of alternative fates and commitment are synthetically depicted. Network components are colored (see legend) according to the developmental program for which they are most important in this context. This color code is further used in the regulatory graph (Fig. 2), as well as in the graphical representations of the results of stochastic simulations (Figs. 4 and 5). Green arrow: positive regulation; Red arrow: negative regulation.

interleukin 7 (IL7), Kit ligand and delta-like ligand 4 (DLL4), the ligand of the receptor Notch1 expressed on the thymic precursor surface (Buono et al., 2016; Rothenberg, 2019a). Notch signaling is the main driver of T-fate specification, which first involves the downregulation of stem-cell related factors, such as Gfi1b, Flt3, Scl, Lmo2, Kit, with Kit remaining expressed at lower levels on the surface of these cells until commitment (Zhang et al., 2012). In this process, the formation of different heterodimers of E proteins plays a crucial role: E2A switches its heterodimeric partners from Scl

210

Elisabetta Cacace et al.

(with additional binding of Lmo2) with the E-protein HEB (encoded by Tcf12), typical of the pre- and post-commitment phases. The first complexes are necessary to ensure Notch1 expression (Miyazaki et al., 2017) and sustain pro-T cell survival until the DN2a stage, while the increase in HEB expression and the shift to E2A-HEB heterodimers is fundamental to activate Rag recombinases and CD3 genes (Del Real & Rothenberg, 2013; Miyazaki et al., 2011; Yashiro-Ohtani et al., 2009). While the cocktail of transcription factors necessary for commitment is refined throughout the progression from ETP to DN2a stage, access to alternative fates is restricted. E proteins restrict the access to the ILC program, antagonizing Id proteins (Braunstein & Anderson, 2011; Miyazaki et al., 2017; Wang et al., 2017), and buffer the expression of Gata3 (Del Real & Rothenberg, 2013; Miyazaki et al., 2017; Xu et al., 2013). TCF-1 (encoded by Tcf7) and Lef1 (encoded by a paralog of Tcf7) are two other key factors for the specification of T cell fate, whose expression is supported by Notch signaling and Gata3. While Lef1 is expressed later around commitment, together with Bcl11b and Ets1, Tcf1 is important for precursor viability from the ETP stage, as shown by loss-of-function experiments, exerting a positive regulation on Gata3, Bcl11b, Il2ra (CD25), as well as on components of the TCR and its signaling machinery (Germar et al., 2011; Hattori, Kawamoto, Fujimoto, Kuno, & Katsura, 1996; Hosoya et al., 2009; Scripture-Adams et al., 2014; Ting, Olson, Barton, & Leiden, 1996; Weber et al., 2011). Runx1 is another important factor controlling Bcl11b expression (Kueh et al., 2016) and limiting the expression of PU.1 (Hoogenkamp et al., 2007; Hosokawa, Ungerb€ack, et al., 2018; Huang et al., 2008; Zarnegar, Chen, & Rothenberg, 2010). Notch signaling, TCF-1, Gata3 and Runx1 activities together enable the expression of Bcl11b, whose upregulation marks the commitment step and the transition to DN2b thymocytes (Ha et al., 2017; Ikawa et al., 2010; Kueh et al., 2016; Li, Leid, & Rothenberg, 2010). Before Bcl11b activation, other lymphoid and myeloid programs remain accessible (Yui & Rothenberg, 2014). Noteworthy, as soon as precursors are subjected to Notch signaling (Garcı´a-Ojeda et al., 2013; Heinzel, Benz, Martins, Haidl, & Bleul, 2007; Scripture-Adams et al., 2014), the B fate is excluded and is de facto inaccessible in the thymus, despite sharing common regulatory modules with the T program, such as the network of E protein activity and the somatic hypermutation apparatus needed to generate clonal receptor genes.

Logical modeling of cell fate specification

211

In T cells, this process is launched by Bcl11b, which, together with the other T-determining factors, upregulate the recombinase enzymes Rag1 and Rag2, responsible for VD(J) recombination. Concurrently, the surrogate alpha chain of the TCR is expressed on the cell surface, together with the accessory components CD3d, CD3e and CD3g. Furthermore, other T cell signaling molecules start to be expressed at this stage, including the kinases Lat and Lck, crucial for the signaling cascades initiated by TCR stimulation, during development and later on by antigen binding. Although T cell development has been extensively studied, our understanding of this process is still incomplete, and quantitative data on the interplay of different regulatory modules are lacking. Hence, formal qualitative modeling approaches are particularly suited to integrate available data, check their coherence, and identify gaps. Here, we show here how logical modeling can (i) recapitulate key state transition events, including commitment, (ii) recapitulate or predict the outcome of specific genetic and environmental perturbations, and (iii) identify gaps in our current understanding of the underlying regulatory network.

3. Logical modeling of T cell specification 3.1 Delineation of the regulatory graph controlling T cell specification The first step of the model construction consisted in mapping the main functional regulations among the different regulatory factors involved in the specification of T cell identity and in the exclusion of alternative cell fate programs. In this respect, we aimed to integrate different kinds of experimental evidence, from gene expression data to more specific information pointing to direct regulatory connections, such as transcription factor binding assays, including the assessment of the effects of binding site mutations, the use of promoter or enhancer reporter assays, as well as pan-genomic approaches such as ChIP-seq. This integration was facilitated by the comprehensive work of Ellen Rothenberg’s group, which resulted in the publication of several up-to-date reviews of T cell development, including detailed regulatory schemes (see, e.g., Georgescu et al., 2008; Kueh & Rothenberg, 2012; Rothenberg, 2019b). Here, we relied on a simple graph-based representation, where each node represents a gene activity (potentially via a modified protein) and where each arc represents a positive or negative influence from the source gene product onto the expression or activity of the target gene product.

212

Elisabetta Cacace et al.

We mapped the main signaling inputs, IL7 as survival cue and Delta as main commitment inducer, distinguishing the gene encoding for a receptor, the presence of the receptor, and its activation by the corresponding ligand. In the case of Notch, we further defined a specific node accounting for the Notch transcription complex (NTC), which directly regulates the expression of target genes in the nucleus. Nrarp and Deltex-1 are further considered for their role in the control of Notch expression, readjusting NTC levels and limiting excessive leukemogenic Notch activity (Dumortier et al., 2006; Krebs, Deftos, Bevan, & Gridley, 2001; Yun & Bevan, 2003). Although several factors are shared among different regulatory programs, we represent each of them by a single component accounting for their different roles. The specification of logical rules then ensures that the different programs are encoded by distinct combinations of such factors. For example, this is the case for PU.1, a key regulator of the myeloid program, but also of the B and T fate. Indeed, it has been shown that myeloid fate remains available before commitment when Notch signaling is removed (Balciunaite, Ceredig, & Rolink, 2005; Bell & Bhandoola, 2008; Kueh et al., 2016; Wada et al., 2008; Yui et al., 2010) and cells are exposed to a myeloid-permissive environment (in presence of SCF, Flt3L, IL3, G-CSF, GM-CSF and M-CSF (De Obaldia et al., 2013), or even after commitment if PU.1 is overexpressed (Franco et al., 2006; Laiosa, Stadtfeld, & Graf, 2006; Lefebvre et al., 2005). We modeled the exclusion of the myeloid fate by encoding the well-established inhibition of C/EBPa by Hes1, a target of Notch (De Obaldia et al., 2013), leaving PU.1 expressed at a level lower than that required for the myeloid program, which can only be reached upon exposure to M-CSF (Del Real & Rothenberg, 2013; Franco et al., 2006). In the case of E proteins, which form several heterodimers with distinct partners, we specified a component for progenitor-associated E protein activity, which represents dimers of E2A (Tcf3) and Scl (with Lmo2 also binding to the complex), belonging to the stem-cell genes that are silenced by Notch signaling and definitely shut down with commitment. We defined another model component representing the heterodimer E2AHEB, which supports Notch1 expression (Miyazaki et al., 2017) and later fosters the expression of Rag-recombinases and CD3 genes, preventing an heterodimerization of E2A with Id proteins, which would correspond to a shift toward the innate lymphoid program (NK and ILC fates) (Braunstein & Anderson, 2011; Miyazaki et al., 2017; Wang et al., 2017).

Logical modeling of cell fate specification

213

In our model, the exclusion of the B-cell fate is enabled by the inhibition of EBF1 by Notch and Gata3, which restricts the B-cell program as soon as thymocytes are subjected to Notch signaling (Garcı´a-Ojeda et al., 2013; Scripture-Adams et al., 2014). We do not explicitly model the shift toward the mast-cell program observed upon Gata3 overexpression (Taghon, Yui, & Rothenberg, 2007), but we assigned a third level to Gata3 to represent this overexpression, which is restrained by E2A-HEB in the pre-commitment phase, and by Bcl11b after commitment. By and large, the resulting regulatory graph shown in Fig. 2 covers most regulatory links reported in the literature, avoiding the inclusion of indirect links redundant with established sequences of direct links. In particular, this graph encompasses most regulatory components and links reported in the regulatory map included in the most recent review on the subject (Rothenberg, 2019b). Encompassing 56 components, this regulatory graph has been encoded using GINsim (Naldi, Hernandez, Abou-Jaoud
e, et al., 2018). The components and regulatory links of this model are documented by textual comments and links to external databases. The model file is provided as Supplementary File S1 in the online version at https://doi.org/10.1016/ bs.ctdb.2020.02.008, while the current release of GINsim (v3.0) is available online (http://ginsim.org).

3.2 Multilevel nodes A first approximation consists in assuming that each gene can be OFF or ON, or that its product can be “inactive” or “active.” Under this assumption, we obtain a Boolean model. However, whenever biologically justified, a Boolean model can be refined by using multilevel variables to account for the fact that a given gene product can regulate different targets, depending on qualitatively distinct concentration or activity ranges (as in the case of morphogens). In our current T cell specification model, most regulatory components are associated with Boolean variables (taking only two values, 0 or 1, denoting inactivity or activity, respectively), with the exception of the components Notch_gene, Lef1, Gata3, Pu1 and Runx1, represented by rectangular shaped nodes in Fig. 2. These five components are associated with ternary variables (taking the logical values 0, 1 and 2, corresponding to negligible, medium and high levels, respectively), based on considerations

214

Elisabetta Cacace et al.

Fig. 2 Regulatory graph integrating the documented cross regulations between the components controlling the specification of T cell precursors entering in the thymus up to commitment (DN2a stage) This regulatory graph encompasses 56 components, denoting transcription factors, environmental cues, TCR components and enzymes, such as RAG recombinases. Nodes are depicted in different colors depending on their belonging to different regulatory modules: red nodes represent progenitor-associated components, orange nodes myeloid-associated factors, light-blue nodes represent the IL7 signaling components, dark-blue nodes correspond to B-cell distinctive factors. Key components of T cell specification are depicted using different shades of green: from dark green nodes, representing Notch signaling components; to intermediate green corresponding to components needed for commitment, to light green ones, representing genes expressed in the post-commitment phase and involved in the TCR assembly and signaling. Ellipses and rectangles denote Boolean and ternary components, respectively. Inhibitory interactions are represented by blunt red arcs, while positive interactions are depicted as green arrows; bimodal interactions (i.e., when a component has distinct regulatory effects according to its level or to the presence of cofactors) are depicted as blue hybrid arrows.

derived from experimental data (these are detailed in the corresponding node annotations in the model file provided as Supplementary File S1 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008). In addition, the interactions requiring the highest regulator level (value 2)

Logical modeling of cell fate specification

215

must be explicitly declared. For example, we considered data demonstrating that a strong overexpression of Gata3 impedes T program completion and redirects the precursors toward a mast-cell fate (Taghon et al., 2007; Xu et al., 2013). However, at lower levels Gata3 is essential for commitment, as shown by knockout experiments (Hattori et al., 1996; Hosoya et al., 2009; Scripture-Adams et al., 2014; Ting et al., 1996), and for restriction of the B-cell fate as soon as bone marrow precursors settle in the thymus (Garcı´a-Ojeda et al., 2013; Scripture-Adams et al., 2014). Hence, we defined a ternary node for Gata3, with the level 1 enabling the interactions required for normal T cell development, and the level 2 associated with the specific interactions enabled by Gata3 overexpression. We also introduced a level 2 for the node Pu1 to account for its inhibitory impact on all genes that induce and follow commitment (represented in our model by the nodes Bcl11b, CD3e, CD3g, E2A_gene, Ets1, Gfi1, Gata3, HEB, Hes1, Hex, Lat, Myb, Ikaros, Rag1, Runx1, TCF1, Zap70): indeed, overexpression of PU.1 can experimentally prevent T cell commitment and shift cells toward the myeloid and dendritic cell program (Champhekar et al., 2015; Del Real & Rothenberg, 2013; Franco et al., 2006; Laiosa, Stadtfeld, & Graf, 2006). Concordantly, we defined a positive regulation from Pu1 onto CEBPb, requiring Pu1 at level 2 and maintaining the access to the myeloid program. Similarly, we defined two positive interactions from Pu1, although requiring its level 2, onto the progenitorassociated genes Bcl11a and Lmo2. The multilevel node for Runx1 accounts for its dynamic interactions with other factors, revealed by loss- and gain-of function experiments (Hoogenkamp et al., 2007; Hosokawa, Ungerb€ack, et al., 2018; Huang et al., 2008; Zarnegar et al., 2010), particularly with PU.1, whose repression right before commitment is strongly dependent on Runx1 upregulation. In the cases of Notch_gene and Lef1, the third level (value 2) is not used for any regulatory interaction threshold, but rather to denote specific upregulation events.

3.3 Logical rules A regulatory graph must be supplemented with rules to obtain a predictive dynamical model. In the case of Boolean nodes, we associate a Boolean rule to each component, which specifies how its regulatory inputs influence its activity. These rules are usually written by combining literals (denoting the presence of regulators) with the Boolean operators NOT (inhibition),

216

Elisabetta Cacace et al.

AND (synergistic effect) and OR (redundant effects). In the case of ternary components, we have to define distinct rules to reach the level 1 (medium) vs the level 2 (high). To define these rules, we relied on experimental evidence whenever possible. For example, the well-characterized “AND” logic governing the activation of Bcl11b by Gata3, Tcf1, Runx1 and Notch signaling (Kueh et al., 2016) is directly encoded in the logical rule associated with Bcl11b. Whenever specific experimental evidence was lacking, we used restrictive generic rules, specifying that a target component is activated in the presence of all its activator and in the absence of all its repressors. These rules were then used to compute the model stable states and to perform wild-type and mutant simulations. Based on the outcome of these analyses, the rules were edited to improve the qualitative match between stable states and simulation results with available transcriptome and other data (e.g., mutant phenotypes). The current set of rules are listed in the Supplementary Table S1 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008 and encoded in the model file provided as Supplementary File S1 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008. Using the resulting model, we were able to recapitulate key features of physiological T cell commitment, as well as the effects of documented genetic and environmental perturbations. In particular, we could explore the impact of loss- and gain-of-function alterations of key components of the T developmental program, and probe the access to alternative developmental programs, which are still available to T precursors before commitment. To investigate the impact of each model perturbation, we first observed the stable states that the model can attain, and we analyzed the state transition graphs generated upon each perturbation when computationally feasible. We subsequently used a powerful model checking approach to verify the reachability of specific gene expression patterns from given initial conditions, and further performed stochastic simulations to investigate the relative probabilities of reaching different outcomes, comparing the wild-type results with those obtained for different perturbations of interest. All the analyses reported hereafter have been encoded into a Jupyter Python notebook which is available online together with our GINsim model file and a Docker image containing all the software tools (see Table 1) necessary to reproduce our results (https://github.com/colomoto/colomotodocker). Furthermore, to ease access to detailed results, we provide an HTML export of this notebook as Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008.

Table 1 Software used in this modeling study. Software name Description Use

URL

References

GINsim

Logical modeling software, Model building and written in Java, Jar distribution visualization

https://ginsim.org

Naldi, Hernandez, Abou-Jaoud
e, et al. (2018)

bioLQM

Java library

https://github.com/colomoto/ bioLQM

Naldi (2018)

Pint

Static analysis of automata Reachability analyses networks, written in OCamel

https://loicpauleve.name/pint/

Paulev
e (2017)

MaBoSS

Stochastic simulations of Boolean networks, C ++ and perl/python scripts

Docker

Computational platform using multi-platform packaging of all https://docs.docker.com/ OS-level virtualization to the software mentioned into https://github.com/colomoto the CoLoMoTo image deliver software packages

Jupyter Open-source web application Notebook enabling the creation and sharing of documents containing live code, equations, visualizations and text

Model conversion Stable state computation

https://maboss.curie.fr/ Computation of mean time plots, estimation of the relative probabilities to reach specific patterns

Encoding of the analysis workflow the tools listed above, with textual explanations

https://jupyter.org/

Stoll et al. (2017)

Naldi, Hernandez, Levy, et al. (2018) Levy et al. (2018); Naldi, Hernandez, Levy, et al. (2018)

218

Elisabetta Cacace et al.

3.4 Computation of model stable states in the wild-type situation

T cells null NK / ILCs myeloid B cells CE B CE Pb BP a Id2 Id1 Sc Id3 l_E 2A S Lm cl o Gf 2 i1b Kit Flt 3 Pu 1 Ly l1 Hh e Ika x ros EB F Pa 1 Fo x5 xo 1 E2 Bcl11 E2 A_g a A_ en pro e tein Ets pro 1 g_ as My s_E b pr CD ot Ru 25 n Ga x3 ta HE TC 3 B_ F1 ge ne HE HE B_ B E Bc 2A l11 Ru b nx 1 Le f1 Gf i1 pT TC a R Ra b g CD 1 3 CD e 3g L Za at p7 0 No tch Hes1 _g en No NT e IL7 tch1_ C Ra rec IL7 _gen Ra e _a c Sta t CD t5 4 CD 5 4 De 4 lte Nr x a MC rp SF IL De 7 lta

States

Once a logical model is defined, we need to confront its behavior to existing data. In this respect, the first step usually consists in computing the asymptotic behavior of the model, i.e., the existence of stable states or of cyclic attractors. When dealing with cellular networks, stable states often denote specific cell types or activation states, while cyclic attractors denote periodic behaviors such as cell cycle or circadian rhythms. Interestingly, computer scientists have developed powerful methods to compute or approximate the attractors of discrete event systems, which have been recently applied to large models of cellular networks (Abou-Jaoud
e et al., 2016; Albert & Thakar, 2014). In particular, it is possible to compute directly all model stable states for logical models with hundreds of components. It is also possible to compute stationary logical patterns that can capture more complex attractors (Gan & Albert, 2018; Klarner, Streck, & Siebert, 2017). Using an algorithm (implemented in GINsim, as well as in the Java library bioLQM) enabling the direct computation of stable states, we obtain 28 stable states (cf. notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008). Some of these states differ only by the values of some input variables MCSF, Delta and IL7. These stable states hence correspond to similar fates and are grouped together in Fig. 3. A stable state corresponding to T cell commitment (DN2b) is obtained only in the presence of Delta (in the presence or not of MCSF and IL7). This T-committed state has received appropriate stimulus from Notch signaling, denoted by the presence of NTC. It further features the panel of typical T cell genes needed for commitment (Gata3, TCF1, Bcl11b, Runx1), and has appropriately activated TCRb, which represents here the TCR genes and is thus a reporter of commitment completion.

Model components

Fig. 3 Stable states attained by the model for the wild-type genetic background. Yellow: level 1; white: level 0; orange: level 2; gray: the component value can take the value 0 or 1 without altering the values of the other components of the stable state.

Logical modeling of cell fate specification

219

Interestingly, Notch signaling fully restricts the myeloid fate, represented in our model by a higher level of Pu1 and by the activation of C/EBPα and C/EBPβ. In this state we still observe the expression of Scl, a progenitor-associated gene, that has been hypothesized to regulate the expression of C/EBPα (Cooper, Guo, & Friedman, 2015), indicating that this state corresponds to a myeloid progenitor rather than to a more differentiated stage. Indeed, it has been shown that myeloid fate can only be reached upon withdrawal of progenitors from the thymic microenvironment and in presence of myeloid-supporting cytokines (Balciunaite et al., 2005; Bell & Bhandoola, 2008; Kueh et al., 2016; Wada et al., 2008; Yui et al., 2010), which are represented by the macrophage-colony stimulating factor (MCSF) in our model. Regardless of the presence of Notch, our model gives rise to other stable states, which display combinations of activated factors corresponding to the B-cell fate (with activation of EBF1 and Pax5), or to the innate lymphoid fate (NK or ILC, with the activation of Id2). Of note, Id protein expression is one of the main hallmarks of an ILC2 state, a kind of ILCs characterized by the expression of genes common to the T program (e.g., TCF1, GATA3 and Bcl11b), which can be reached from pro-T cells following ectopic expression of Id proteins (Braunstein & Anderson, 2011; Miyazaki et al., 2017).

3.5 Computation of model stable states upon environmental or genetic perturbations In the context of the logical framework, it is relatively easy to assess the impact of microenvironmental or genetic perturbations. Here, we focus on the lack of Notch signaling (input Delta set to zero), as well as on loss-of-function or gain-of-function of the key T cell factors Bcl11b, Gata3, Runx1 and TCF1. A loss-of-function (LoF) is simply defined by setting the corresponding variable permanently to zero, whereas a gainof-function (GoF) is defined by setting the corresponding variable permanently to its maximal value (i.e., value 1 in the case of Boolean variables, value 2 in the case of ternary variables). We further assessed the effect of the joint ectopic expression of C/EBPa and Pu1 (at level 2), as this situation corresponds to published reprogramming experiments (Laiosa, Stadtfeld, Xie, de Andres-Aguayo, & Graf, 2006). The notebook (Supplementary File S2 in the online version at https:// doi.org/10.1016/bs.ctdb.2020.02.008) lists the stable patterns obtained for

220

Elisabetta Cacace et al.

each of the aforementioned perturbations. To ease comparison, the stable patterns have been projected onto a smaller set encompassing 13 components, which are sufficient to distinguish the main cell fates: C/EBPα (as representative of the myeloid fate), Id2 (as representative of Id proteins and of the innate lymphoid program), Scl (as indicator of progenitorassociated genes), Pu1 (as key regulator of both T-fate and alternative fates), EBF1 and Pax5 (as markers of the B-cell fate), NTC (to monitor Notch signaling activity), Gata3 and TCF1 (as earlier T cell genes), Bcl11b and Runx1 (as later, commitment-related T cell genes) and pTa and TCRb (as post-commitment indicators). For the sake of space, we only summarize here the most salient results. We first sought to simulate perturbations of the thymic microenvironment that recapitulate documented experimental perturbations. The removal or blockade of Notch signaling has been experimentally achieved by various means: (i) comparing the development of thymocytes in OP9DLL4 co-culture versus OP9 control (Del Real & Rothenberg, 2013; Taghon, David, Zu´n˜iga-Pfl€ ucker, & Rothenberg, 2005), (ii) preventing the cleavage of the intracellular domain of the Notch1 receptor by γ-secretase inhibitors (Del Real & Rothenberg, 2013; Franco et al., 2006; Van de Walle et al., 2009), (iii) using bone marrow precursors lacking Notch1 (Radtke et al., 2000; Wilson, MacDonald, & Radtke, 2001) or RBP-Jκ (Han et al., 2002), (iv) overexpressing Notch inhibitors such as Deltex (Izon et al., 2002), or (v) using dominant-negative constructs of Notch co-activators of the mastermind family (MAML) (Taghon et al., 2009). In the absence of the Delta stimulus, our model leads to a stable state corresponding to a complete failure of commitment, without neither the key T-factors activated before commitment (TCF1, GATA3 and Runx1), nor downstream T-specific genes such as TCRb, resulting in a redirection toward the B-cell program, as experimentally observed (Feyerabend et al., 2009; Han et al., 2002; Koch et al., 2008; Wilson et al., 2001). We can also identify cells that are shifted toward a high-Id protein state that can be interpreted as an NK-like state. This is consistent with previous findings using γ-secretase inhibitors (De Smedt, Magda, Reynvoet, Leclercq, & Plum, 2005) and dominant-negative MAML constructs (Taghon et al., 2009). Concordantly with published data (Li et al., 2010), a LoF of Bcl11b prevents ETP precursors from reaching the T-committed state, despite the presence of thymic microenvironment stimuli, represented by IL7 and Delta. The corresponding stable pattern recapitulates some key effects

Logical modeling of cell fate specification

221

of Bcl11b: (i) the repression of Id proteins and of NK/ILC-like fate (Li et al., 2010; Longabaugh et al., 2017), particularly relevant in the mouse model (Hosokawa, Romero-Wolf, et al., 2018), (ii) the latest stage reached by T precursors without Bcl11b, i.e., a viable population corresponding to the DN2a stage, equipped with key T cell genes already expressed in the pre-commitment phase (e.g., Gata3 and TCF1), but unable to proceed to commitment and to express later T genes (e.g., TCRb). We further observe a stable state featuring B-cell genes (e.g., EBF1), which is not actually found in experimental conditions (Li et al., 2010). It remains to be established whether the expression of these genes is too low to be experimentally detected or if some additional putative regulation from Bcl11b onto B-cell genes should be assumed. We further investigate the expression levels of EBF-1 and Pax5 and the proportion of cells that can actually access this fate by performing stochastic simulations with the software MaboSS (see Section 3.7). GoF perturbation of Bcl11b leads to a wide range of stable states, including states that are normally excluded by the time Bcl11b is expressed upon commitment (e.g., myeloid-like and B-cell-like). Among the stable states attained, we find a B-cell like fate where EBF1 and Pax5 are activated with Bcl11b, pointing to some missing regulation for the exclusion of the B-cell program, which is known to be restrained by Notch signaling and Gata3 (Garcı´a-Ojeda et al., 2013; Scripture-Adams et al., 2014). We investigate the actual reachability of these states and the relative proportions of cells that can attain them below in the following section. We then explored the impact of perturbations of the other key factors ensuring commitment, i.e., Gata3, TCF1 and Runx1. Gata3 is essential for precursor survival and contributes to the exclusion of the B-cell fate shortly after the progenitors settle in the thymus (Garcı´aOjeda et al., 2013; Hattori et al., 1996; Hosoya et al., 2009; Scripture-Adams et al., 2014; Ting et al., 1996). This is confirmed by the stable patterns reached upon LoF simulation, where the B-cell fate is available, while generation of committed cells is abolished, with the resulting stable state lacking the activation of the key commitment factor Bcl11b and the later TCR-genes TCRb and pTa, although retaining expression of TCF1 and Runx1. This is in part concordant with experimental data based on slow GATA-3 downregulation using RNA interference and acute conditional deletion (Scripture-Adams et al., 2014): TCF1 inhibition becomes evident only when Gata3 levels are markedly reduced, probably because of its strong binding affinity (Scripture-Adams et al., 2014; Zhang et al.,

222

Elisabetta Cacace et al.

2012). This calls for further refinement of the Gata3-TCF1 regulation, possibly introducing an additional level for Gata3. NK-like and myeloid-like fates remain available, to an extent further assessed below in the stochastic simulation of this perturbation. Gata3 GoF (set in our model at level 2) results in the loss of the stable pattern associated with T cell commitment, as experimentally observed in the mouse model (Taghon et al., 2007; Xu et al., 2013). In addition, the B and myeloid stable patterns are also lost, while the stable pattern marked by the expression of Id2 is conserved. Although we cannot recapitulate the experimentally observed shift toward a mast-cell fate, whose main components are not included in the model, we can reproduce some of the effects that were observed on Scl (upregulation), PU.1 and Tcf1 (downregulation) (Taghon et al., 2007), which all appear in one of the three resulting stable patterns. Overall, this perturbation requires more refined modeling of existing interactions, together with the inclusion of more components relative to the mast-cell fate, but these preliminary results are encouraging. LoF of TCF1 is best known to induce a substantial loss of viability in the ETP pool (Germar et al., 2011; Scripture-Adams et al., 2014; Weber et al., 2011). TCF1 LoF leads to a stable pattern lacking Bcl11b but still expressing Gata3, thus failing to generate a population of committed cells, together with the alternative stable states above mentioned. TCF1 overexpression does not lead to major alterations of the stable patterns and has been shown to accelerate T-specification, rather than excluding alternative fates (Weber et al., 2011). More insights on this perturbation can be provided by reachability analysis and stochastic simulations (see below). Additionally, we simulated perturbations of the factor Runx1. Its LoF leads to a pervasive alteration of the stable pattern landscape, which corresponds to the abolishment of hematopoietic cells as a whole, in agreement with experimental data (Speck, 2001; Talebian et al., 2007). Runx1 GoF does not impede neither T cell commitment nor B-cell development, but it does block myeloid fate. Experimental Runx1 overexpression (Wong et al., 2010) has been performed only in post-commitment stages, showing that downregulation is necessary to proceed to DP stages, but no evidence is available for Runx1 GoF in pre-commitment stages, where its expression already increases from the ETP stage. Finally, the joint ectopic activation of CEBPa and Pu1 (at level 2) gives rise to a single stable pattern corresponding to a myeloid fate, which matches the results of previously reported reprogramming experiments (Laiosa, Stadtfeld, Xie, et al., 2006).

Logical modeling of cell fate specification

223

3.6 Assessing the reachability of specific expression patterns from specific initial states The computation of stable states provides important information regarding the activity patterns compatible with each input combinations. However, the mere existence of a stable pattern corresponding to a given cellular state, e.g., DN2a thymocytes, does not necessarily imply that this pattern can be reached from biologically plausible initial conditions, i.e., from a logical state corresponding to the multipotent progenitors entering into the thymus, hereafter called early progenitors (EP). The dynamical behavior of a logical model is usually represented in terms of an oriented graph, where each node denotes a state of the system, i.e., a specific combination of levels of the gene products (0 or 1 in the simplest Boolean case), connected by transitions (changes of values) enabled by the Boolean rules. In this respect, two main approaches are usually alternatively considered. The first approach considers that, at a given state, one can simultaneously apply all the changes dictated by the rules, and returns a unique state. Under this synchronous updating assumption, the dynamics is deterministic, as any state can lead to only one other state. Hence, this assumption cannot directly account for alternative cell fate decisions from a given cellular state. In contrast, the asynchronous updating approach considers that, whenever several changes are dictated by the rules at a given state, only one of them is performed at a time, giving rise to as many independent transitions. This approach is thus intrinsically non-deterministic, computationally more complex, but also more realistic from a biological point of view. As the size of state transition graphs increases exponentially with the number of regulatory components considered (in particular under the asynchronous approach), it becomes rapidly impossible to analyze state transition graphs exhaustively. To verify the reachability (i.e., the existence of a state transition path) of a given pattern (e.g., stable state) from specific initial conditions, we can rely on powerful model checking approaches developed by computer scientists, which are based on static analysis and thereby avoid the computation of very large state transition graphs (Abou-Jaoud
e et al., 2015, 2016; Naldi, Hernandez, Levy, et al., 2018). Here, we use a recent discrete automata modeling framework called Pint (Paulev
e, 2017), which enables very efficient reachability verification. To launch a pattern reachability verification with Pint, we need (i) to convert our GINsim model into the format expected by this software, which can be performed automatically using the bioLQM library

224

Elisabetta Cacace et al.

(Naldi, 2018); (ii) define an initial state or pattern; (iii) define a target state or pattern. As detailed in the notebook (Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008), using Pint, we first confirmed the reachability of the DN2b pattern from the EP state. Turning to model perturbations, we could verify that the model correctly recapitulates the loss of the reachability of the DN2b pattern in the absence of Delta (Feyerabend et al., 2009; Han et al., 2002; Koch et al., 2008; Wilson et al., 2001), as well as for the loss-of function of Bcl11b (Kueh et al., 2016; Li et al., 2010; Yui et al., 2010), Gata3 (Garcı´aOjeda et al., 2013; Hosoya et al., 2009; Scripture-Adams et al., 2014), TCF1 (Germar et al., 2011; Scripture-Adams et al., 2014; Weber et al., 2011) or Runx1 (Speck, 2001; Talebian et al., 2007). Reachability of the DN2b pattern is also impaired by a gain-of-function of Gata3 (Taghon et al., 2007; Xu et al., 2013). In contrast, a gain-of function of Bcl11b, Runx1 or TCF1 does not impede the reachability of the DN2b pattern. This is concordant with existing experimental data in the case of TCF1 (Weber et al., 2011) and Bcl11b (Ha et al., 2017), while no experimental evidence is available for Runx1 GoF after the ETP stage (in HSPCs, GoF perturbation results in an acute myeloid leukemia phenotype (Behrens et al., 2017)).

3.7 Stochastic simulations The reachability analysis described in the preceding section is helpful to verify the coherence of the model with data already available on the potential of thymocytes at different stages, in the wild-type situation or in the presence of genetic perturbations. Indeed, we have seen that Pint can provide information about the reachability of a given pattern from chosen initial conditions. However, the probability to reach such pattern from given initial conditions must also be assessed. This can be addressed using other model checking tools, but we opted here to rather use a probabilistic extension of the logical formalism, which has been implemented in the software suite MaBoSS (https://maboss.curie.fr/) (Stoll et al., 2017). This software relies on the use of the Gillespie algorithm to compute temporal activity profiles based on predefined Boolean rules. To launch a probabilistic simulation with MaBoSS, we need (i) to convert our GINsim model into the format expected by this software, which can be performed automatically using the bioLQM library (Naldi, 2018);

Logical modeling of cell fate specification

225

(ii) define a set of initial states; (iii) specify up- and down-rates for each Boolean variable, (iv) specify a series of simulation parameters, such as the number of trajectories to generate, the maximal duration and the time step to be used; and (v) specify the components to be reported. As MaBoSS can handle only purely Boolean models, our multilevel model is first translated into a Boolean one, which amounts to encode each ternary variable into two Boolean ones. For example, Gata1 (taking the values 0, 1 and 2) is then encode into two Boolean variables Gata1_b1 and Gata1_b1, which simply indicate that Gata1 is over its first threshold and over its second threshold, respectively (note that having simultaneously Gata1_b1 ¼ 0 and Gata1_b1 ¼ 1 is not allowed). Regarding initial states, we performed wild-type simulations starting with an ETP-like state, with IL7, Delta, CEBPa, Id1, Id2,Id3, Scl, Scl_E2A, Lmo2, Gfi1b, Kit, Flt3, Pu1_b1, Lyl1, Hhex, Bcl11a, E2A_ gene, E2A_protein, Myb, Gata3_b1, TCF1, HEB_gene, HEB, HEB_ E2A, Runx1_b1, Gfi1, Notch_gene_b1, NTC, Notch1_rec, IL7Ra_gene, IL7Ra_act, Stat5, and Nrarp all taking the value 1, while the remaining variables initially take the value 0. In the cases of genetic perturbations, this initial state was modified only when required to properly simulate the corresponding loss-of-function or gain-of-function background. As kinetic information necessary to define specific up- and downtransition rates is mostly missing, we simply considered equiprobable transition rates whenever multiple transitions are enabled by the logical rules. For all MaBoSS simulations, we used default parameters, excepting for the maximal duration of simulations, which was chosen such as to lead to mean equilibrium (see notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008). Finally, we selected a limited number of markers to ease the analysis of our stochastic simulations: Bcl11b, TCF1, Gata3_b1, Gata3_b2, EBF1, Pax5, CEBPa, Scl, Id2 and TCRb. Together, these markers provide enough information to evaluate the progression toward the specification of the main blood cell progenitors. The results of MaBoSS simulations can be processed and displayed in the form of pie charts, which represent the probabilities to have each combination of markers activated at the states reached at the end of individual simulations. They can also be displayed as temporal profiles showing the evolution of marker activity probabilities. Furthermore, MaBoSS can generate time plots for the state and transition entropies. The state entropy provides a measure of the diversity of the states

226

Elisabetta Cacace et al.

(higher the state entropy, higher the diversity of states), while the transition entropy quantifies the number of transitions. Hence, whenever a simulation ends up with only stable states, the transition entropy drops to zero, while the state entropy sets to a stable value. The notebook (Supplementary File S2 in the online version at https:// doi.org/10.1016/bs.ctdb.2020.02.008) reports probabilistic simulations for the wild-type and the different perturbations mentioned above, with the results displayed in terms of both pie chart and time plots. Hereafter, we focus on selected results, starting with the simulation of wild-type ETP evolution in the presence of IL7 and Delta (Fig. 4). In the presence of Delta and IL7 inputs, a substantial proportion of ETP cells (24%) progresses toward commitment, with Gata1, TCF1 and Bcl11b all set to the value 1. We also see that a smaller percentage of ETP cells progress toward a B-cell fate (4%), while a majority (72%) of cells only display Id2 markers at the end of the simulations. As the transition entropy (TH) drops to zero and as the state entropy (H) reaches a low value (1, cf. notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008), the simulations clearly end in a limited number of stable states. However, the relatively low proportion of committed T cells is somewhat disappointing. This said, as we did not tune transition rates, this is not surprising. We should not look to closely at quantitative values, but rather compare the proportions obtained for the wild-type situation with those obtained for different perturbations. The bottom part of Fig. 4 displays the results obtained starting from the same initial conditions but in absence of Delta. Consistently with published results (Feyerabend et al., 2009; Han et al., 2002; Koch et al., 2008; Wilson et al., 2001), T cell commitment is completely lost, while the proportion of B progenitors is substantially increased (7%). Fig. 5 displays the pie charts obtained for six selected perturbations (the corresponding time plots are provided in the notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020. 02.008). As experimentally demonstrated and previously discussed in Section 3.5, the LoF of Bcl11b (Fig. 5, panel A) leads to a complete failure in commitment (Li et al., 2010). Among the stable states obtained, one corresponds to B fate (epitomized by the expression of EBF1 and Pax5), which is not experimentally documented. Within the stochastic simulation framework, we can now observe that the proportion of cells attaining it is indeed very low, suggesting that, even if present, this state might be very difficult to

Logical modeling of cell fate specification

227

Fig. 4 Stochastic simulations of the wild-type upon physiological environmental stimuli (co-existence of IL7 and Notch signaling) or in the presence of IL7 but in the absence of Notch signaling, performed with MaBoSS software. 1000 simulations were performed, starting from initial conditions corresponding to ETPs. The pie charts display the proportions of the different combinations of selected differentiation markers activated at the end of the simulations. The temporal evolutions of the mean values of these 10 markers are displayed on the right (with two Boolean variables for Gata3, denoting medium and high levels, respectively). In these time plots, the x- and y-axes represent time (in arbitrary units) and activation frequency, respectively. According to the model, in the presence of the two inputs Delta and IL7, about 25% of the early progenitors reach full T commitment (with Bcl11b, GATA3, TCF1 and TCRb ON), whereas a small percentage reaches a B-cell state (3%, with both EBF1 and Pax5 ON). In contrast, in the presence of IL7 input only, T cell commitment is fully lost, while the proportion of progenitors taking a B-cell like fate increases to about 7%.

observe experimentally. The vast majority (74%) of the population is shifted toward the innate lymphoid program, while about 23% of the population is represented by uncommitted cells expressing Gata3 and TCF1. GoF perturbation of Bcl11b yields to a marginal increase in committed T cells (Fig. 5, panel B), consistently with the effects of the transduction of Bcl11b in CD34+ progenitors cultured on OP9-DLL1 stroma cells (Ha et al., 2017). We further observe a B-like fate with atypical coexpression of Bcl11b (8%).

228

Elisabetta Cacace et al.

Fig. 5 Stochastic simulations of the effects of genetic perturbations using MaBoSS. Pie charts reporting the proportions of the different combinations of selected differentiation markers activated at the end of the simulations are displayed, for different perturbations: (A) Bcl11b loss-of-function, (B) Bcl11b gain-of-function, (C) Gata3 loss-of-function, (D) Gata3 gain-of-function, (E) the ectopic activation of CEBPa and Pu1, and (F) the response of early progenitors to MCSF, in the absence of Delta and IL7. All these perturbations completely impair T cell specification, except for Bcl11b gain-of-function, which results in a significant increase of the progenitors reaching T cell commitment. Interestingly, the two last cases (E, F) predominantly give rise to a myeloid-like fate. Further details on these simulations (including time plots) are included in the notebook provided as Supplementary File S2 in the online version at https://doi.org/10.1016/ bs.ctdb.2020.02.008.

Logical modeling of cell fate specification

229

Following a loss-of-function of Gata3 (Fig. 5, panel C), a proportion of simulations (24%) still lead to a state with TCF1 ON, but in the absence of Bcl11b, thus denoting a failure of T cell commitment. In addition, we observe a slight increase of the B-cell fate (up to 5%). Following a gain-of-function of Gata3 (Fig. 5, panel D), both B and T cell markers are completely lost, denoting a complete blocking of lymphoid development. By imposing a gain-of-function of both CEBPa and Pu1 (Fig. 5, panel E), we could simulate reprogramming experiments performed by the group of Thomas Graf, which consisted in forcing the ectopic expression of these two factors in thymocyte progenitors. Consistently with the published results (Laiosa, Stadtfeld, & Graf, 2006), our simulations give rise to a large proportion of states characterized by high levels of both factor activities, pointing to a myeloid conversion. We also simulated the evolution of ETP in absence of Delta and IL7, but in presence of MCSF (Fig. 5, panel F), which gives rise to a widespread activation of both CEBPA and Pu1 at their maximal levels, hence denoting progression toward myeloid fate. The loss-of-function of Runx1 (cf. notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008) results in the loss of all markers, which is consistent with the essential role of this factor for the development of all immune cell lineages. In contrast, according to our model, the gain-of-function of Runx1 (cf. notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/ bs.ctdb.2020.02.008) does not substantially affect progression to B and T cell fates. Finally, TCF1 gain-of-function did not substantially affect the proportion of states corresponding to T cell commitment (Weber et al., 2011), while the loss-of-function of Bcl11b (Kueh et al., 2016; Li et al., 2010; Yui et al., 2010) or of TCF1 (Germar et al., 2011; Scripture-Adams et al., 2014; Weber et al., 2011) lead to the blockade of T cell development, as expected (cf. notebook, Supplementary File S2 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008).

4. Conclusions The logical framework is particularly well suited to build comprehensive dynamical models of developmental processes. Relatively easy to define, Boolean models rely on qualitative data to generate qualitative

230

Elisabetta Cacace et al.

but clear-cut results, e.g., concerning the possibility for a given regulatory network to give rise to various stable states representing different cellular fates. Interestingly, when justified, it is possible to refine logical models through the parsimonious introduction of multilevel components, the use of sophisticated updating schemes, or yet by considering transition rates to generate stochastic simulations and estimate the relative probabilities of concurrent fates. Interestingly, a relatively wide spectrum of complementary software tools are currently available to define and analyze Boolean or multilevel logical models, involving up to hundreds of regulatory components (see, e.g., Abou-Jaoud
e et al., 2016). We have illustrated the assets of the logical framework through the delineation and dynamical analysis of a comprehensive network model encompassing the main regulatory components and interactions controlling the specification of T cells in the thymus, from early progenitors to commitment (DN2a stage). Over the last decade, a number of detailed maps of the gene-regulatory network (GRN) driving T cell commitment have been proposed by the group of Ellen Rothenberg (Georgescu et al., 2008; Kueh & Rothenberg, 2012; Rothenberg, 2019b). The same group further designed and analyzed a four-component-dynamical model recapitulating several key features of this process, in particular commitment irreversibility and delay upon Notch signaling exposure (Manesso, Chickarmane, Kueh, Rothenberg, & Peterson, 2013; Manesso, Kueh, Freedman, Rothenberg, & Peterson, 2016). More recently, starting from a systematic analysis of the capacity of three-components regulatory motifs combining several interlocked positive circuits to generate four attractors, Ye et al. proposed that such a small core network could account for a stepwise commitment of T cell progenitors in the thymus (Ye, Kang, Bailey, Li, & Hong, 2019). Relying on differential equations, these models capture the crucial role of Notch signaling, considered as an input, in triggering key T cell transcriptional factors (Gata3, TCF1 and Bcl11b), which collectively set up T cell commitment and simultaneously repress Pu1 and thereby the alternative myeloid fate. In contrast, the logical model of early steps of T cell development described here encompasses a much wider range of relevant components (56 signaling and transcriptional factors), including components and interactions for which quantitative interaction data are not available and cannot be easily inferred, in particular key factors involved in the specification of alternative fates (i.e., myeloid, B cell and innate lymphoid fates).

Logical modeling of cell fate specification

231

This model was obtained after various refinement rounds involving systematic verifications of documented properties, starting with the generation of relevant stable states matching B, T, innate lymphoid and myeloid expression patterns, in the wild-type situation and for various documented genetic perturbations. Using a model checking framework, we further systematically assessed the reachability of the T cell commitment state (DN2b) from the early progenitor state. Finally, we performed stochastic Boolean simulations to further estimate the probabilities to reach alternative gene expression patterns depending on the genetic background. Overall, our model qualitatively recapitulates the physiological commitment endpoints for most of the documented environmental and genetic perturbations. However, at this point, we have to acknowledge that our current model still shows several important limitations. First of all, our model does not explicitly account for the effect of the signaling inputs and transcriptional status on cell survival and proliferation. Secondly, our model does not implement the post-commitment downregulation of PU.1 by Runx1 (Rothenberg, 2019a), which would require a further refinement of the modeling of PU.1 regulation, presumably through the consideration of at least one additional threshold (i.e., using a quaternary variable). Thirdly, our current model does not properly restrict the access to the B-cell program, which is repressed early during intrathymic development, before the loss of the access to NK, ILC and myeloid fates (Bell & Bhandoola, 2008; Rothenberg, 2011; Wada et al., 2008). Nevertheless, our model analysis points to gaps in our current knowledge of the molecular mechanisms underlying B-cell fate restriction in the thymus. Fourthly, cross-interactions between T cell and other blood cell specification circuits (e.g., myeloid and innate lymphoid), which still need to be fully deciphered, are only partially implemented in the model at the moment. Finally, our stochastic simulations currently subestimate the proportion of early progenitors undergoing commitment in the wild-type conditions, compared to the reported yield of T cell reaching commitment starting from the Gata3 + -TCF1 + ETP subset (Zhou et al., 2019). However, this does not impede our model to qualitatively recapitulate committed T cell yield variations following specific genetic perturbations (e.g., increase in Bcl11b GoF, or full collapse in Gata3, TCF1 or Bcl11b LoF). Overall, our current comprehensive dynamical model for T cell commitment constitutes a solid basis for extensions and refinements to address these

232

Elisabetta Cacace et al.

limitations. In this respect, we provide our model in a standard computerreadable format (SBML-qual file, provided as Supplementary File S3 in the online version at https://doi.org/10.1016/bs.ctdb.2020.02.008, and further deposited into the BioModels database with the ID MODEL2002170001), together with a complete analysis workflow as supplementary material. We have shown how environmental and genetic perturbations can be tested on the model, recapitulating known biology. The model can also be used to predict the outcome of novel perturbations that have not been yet tested experimentally, helping to decipher specific regulatory interactions and informing the design of novel experiments. Such approach has proved effective in elucidating the functional relevance of physical interactions between transcription factors and their targets (Collombet et al., 2017). Once the limitations mentioned above will be overcome, the model will be used to probe the role of specific regulations for T cell development and to disentangle overlapping circuits controlling alternative developmental programs.

Acknowledgments We shall warmly thank Ellen Rothenberg for her kind expert advice over the course of this project. We would like to further acknowledge the technical support of Aur
elien Naldi during the preparation of the Jupyter notebook integrating our final model analyses. The D.T. laboratory was supported by grants from the French Plan Cancer, in the context of the projects CoMET (2014–2017) and SYSTAIM (2015–2019), as well as by a grant from the French Agence Nationale pour la Recherche, in the context of the project TMod (2016–2019).

References Abou-Jaoud
e, W., Monteiro, P. T., Naldi, A., Grandclaudon, M., Soumelis, V., Chaouiya, C., et al. (2015). Model checking to assess T-helper cell plasticity. Frontiers in Bioengineering and Biotechnology, 2, 86. https://doi.org/10.3389/fbioe.2014.00086. Abou-Jaoud
e, W., Traynard, P., Monteiro, P. T., Saez-Rodriguez, J., Helikar, T., Thieffry, D., et al. (2016). Logical modeling and dynamical analysis of cellular networks. Frontiers in Genetics, 7, 94. https://doi.org/10.3389/fgene.2016.00094. Albert, R., & Thakar, J. (2014). Boolean modeling: A logic-based dynamic approach for understanding signaling and regulatory networks and for making useful predictions. Wiley Interdisciplinary Reviews. Systems Biology and Medicine, 6(5), 353–369. https://doi. org/10.1002/wsbm.1273. Balciunaite, G., Ceredig, R., & Rolink, A. G. (2005). The earliest subpopulation of mouse thymocytes contains potent T, significant macrophage, and natural killer cell but no B-lymphocyte potential. Blood, 105(5), 1930–1936. https://doi.org/10.1182/ blood-2004-08-3087.

Logical modeling of cell fate specification

233

Behrens, K., Maul, K., Tekin, N., Kriebitzsch, N., Indenbirken, D., Prassolov, V., et al. (2017). RUNX1 cooperates with FLT3-ITD to induce leukemia. The Journal of Experimental Medicine, 214(3), 737–752. https://doi.org/10.1084/jem.20160927. Bell, J. J., & Bhandoola, A. (2008). The earliest thymic progenitors for T cells possess myeloid lineage potential. Nature, 452(7188), 764–767. https://doi.org/10.1038/ nature06840. Braunstein, M., & Anderson, M. K. (2011). HEB-deficient T-cell precursors lose T-cell potential and adopt an alternative pathway of differentiation. Molecular and Cellular Biology, 31(5), 971–982. https://doi.org/10.1128/MCB.01034-10. Buono, M., Facchini, R., Matsuoka, S., Thongjuea, S., Waithe, D., Luis, T. C., et al. (2016). A dynamic niche provides kit ligand in a stage-specific manner to the earliest thymocyte progenitors. Nature Cell Biology, 18(2), 157–167. https://doi.org/10.1038/ncb3299. Champhekar, A., Damle, S. S., Freedman, G., Carotta, S., Nutt, S. L., & Rothenberg, E. V. (2015). Regulation of early T-lineage gene expression and developmental progression by the progenitor cell transcription factor PU.1. Genes & Development, 29(8), 832–848. https://doi.org/10.1101/gad.259879.115. Collombet, S., van Oevelen, C., Sardina Ortega, J. L., Abou-Jaoud
e, W., Di Stefano, B., Thomas-Chollier, M., et al. (2017). Logical modeling of lymphoid and myeloid cell specification and transdifferentiation. Proceedings of the National Academy of Sciences of the United States of America, 114(23), 5792–5799. https://doi.org/10.1073/pnas. 1610622114. Cooper, S., Guo, H., & Friedman, A. D. (2015). The + 37 kb Cebpa enhancer is critical for Cebpa myeloid gene expression and contains functional sites that bind SCL, GATA2, C/EBPα, PU.1, and additional Ets factors. PLoS ONE, 10(5), e0126385, https://doi.org/10.1371/journal.pone.0126385. Del Real, M. M., & Rothenberg, E. V. (2013). Architecture of a lymphomyeloid developmental switch controlled by PU.1, notch and Gata3. Development, 140(6), 1207–1219. https://doi.org/10.1242/dev.088559. De Obaldia, M. E., Jeremiah Bell, J., Wang, X., Harly, C., Yashiro-Ohtani, Y., DeLong, J. H., et al. (2013). T cell development requires constraint of the myeloid regulator C/EBP-α by the notch target and transcriptional repressor Hes1. Nature Immunology, 14(12), 1277–1284. https://doi.org/10.1038/ni.2760. De Obaldia, M. E., & Bhandoola, A. (2015). Transcriptional regulation of innate and adaptive lymphocyte lineages. Annual Review of Immunology, 33, 607–642. https://doi.org/ 10.1146/annurev-immunol-032414-112032. De Smedt, M., Hoebeke, I., Reynvoet, K., Leclercq, G., & Plum, J. (2005). Different thresholds of notch signaling bias human precursor cells toward B-, NK-, Monocytic/ dendritic-, or T-cell lineage in thymus microenvironment. Blood, 106(10), 3498–3506. https://doi.org/10.1182/blood-2005-02-0496. Dumortier, A., Jeannet, R., Kirstetter, P., Kleinmann, E., Sellars, M., dos Santos, N. R., et al. (2006). Notch activation is an early and critical event during T-cell leukemogenesis in Ikaros-deficient mice. Molecular and Cellular Biology, 26(1), 209–220. https://doi.org/ 10.1128/MCB.26.1.209-220.2006. Feyerabend, T. B., Terszowski, G., Tietz, A., Blum, C., Luche, H., Gossler, A., et al. (2009). Deletion of Notch1 converts pro-T cells to dendritic cells and promotes thymic B cells by cell-extrinsic and cell-intrinsic mechanisms. Immunity, 30(1), 67–79. https://doi.org/ 10.1016/j.immuni.2008.10.016. Franco, C. B., Scripture-Adams, D. D., Proekt, I., Taghon, T., Weiss, A. H., Yui, M. A., et al. (2006). Notch/delta signaling constrains reengineering of pro-T cells by PU.1. Proceedings of the National Academy of Sciences of the United States of America, 103(32), 11993–11998. https://doi.org/10.1073/pnas.0601188103.

234

Elisabetta Cacace et al.

Gan, X., & Albert, R. (2018). General method to find the attractors of discrete dynamic models of biological systems. Physical Review E, 97(4-1), 042308. https://doi.org/ 10.1103/PhysRevE.97.042308. Garcı´a-Ojeda, M. E., Klein Wolterink, R. G. J., Lemaıˆtre, F., Goff, O. R.-L., Hasan, M., Hendriks, R. W., et al. (2013). GATA-3 promotes T-cell specification by repressing B-cell potential in pro-T cells in mice. Blood, 121(10), 1749–1759. https://doi.org/ 10.1182/blood-2012-06-440065. Georgescu, C., Longabaugh, W. J. R., Scripture-Adams, D. D., David-Fung, E.-S., Yui, M. A., Zarnegar, M. A., et al. (2008). A gene regulatory network armature for T lymphocyte specification. Proceedings of the National Academy of Sciences of the United States of America, 105(51), 20100–20105. https://doi.org/10.1073/pnas. 0806501105. Germar, K., Dose, M., Konstantinou, T., Zhang, J., Wang, H., Lobry, C., et al. (2011). T-cell factor 1 is a gatekeeper for T-cell specification in response to notch signaling. Proceedings of the National Academy of Sciences of the United States of America, 108(50), 20060–20065. https://doi.org/10.1073/pnas.1110230108. Han, H., Tanigaki, K., Yamamoto, N., Kuroda, K., Yoshimoto, M., Nakahata, T., et al. (2002). Inducible gene knockout of transcription factor recombination signal binding protein-J reveals its essential role in T versus B lineage decision. International Immunology, 14(6), 637–645. https://doi.org/10.1093/intimm/dxf030. Hattori, N., Kawamoto, H., Fujimoto, S., Kuno, K., & Katsura, Y. (1996). Involvement of transcription factors TCF-1 and GATA-3 in the initiation of the earliest step of T cell development in the thymus. The Journal of Experimental Medicine, 184(3), 1137–1147. https://doi.org/10.1084/jem.184.3.1137. Ha, V. L., Luong, A., Li, F., Casero, D., Malvar, J., Kim, Y. M., et al. (2017). The T-ALL related gene BCL11B regulates the initial stages of human T-cell differentiation. Leukemia, 31(11), 2503–2514. https://doi.org/10.1038/leu.2017.70. Heinzel, K., Benz, C., Martins, V. C., Haidl, I. D., & Bleul, C. C. (2007). Bone marrowderived hemopoietic precursors commit to the T cell lineage only after arrival in the Thymic microenvironment. Journal of Immunology, 178(2), 858–868. https://doi.org/ 10.4049/jimmunol.178.2.858. Hoogenkamp, M., Krysinska, H., Ingram, R., Huang, G., Barlow, R., Clarke, D., et al. (2007). The Pu.1 locus is differentially regulated at the level of chromatin structure and noncoding transcription by alternate mechanisms at distinct developmental stages of hematopoiesis. Molecular and Cellular Biology, 27(21), 7425–7438. https://doi.org/ 10.1128/MCB.00905-07. Hosokawa, H., Romero-Wolf, M., Yui, M. A., Ungerb€ack, J., Quiloan, M. L. G., Matsumoto, M., et al. (2018). Bcl11b sets pro-T cell fate by site-specific cofactor recruitment and by repressing Id2 and Zbtb16. Nature Immunology, 19(12), 1427–1440. https:// doi.org/10.1038/s41590-018-0238-4. Hosokawa, H., Ungerb€ack, J., Wang, X., Matsumoto, M., Nakayama, K. I., Cohen, S. M., et al. (2018). Transcription factor PU.1 represses and activates gene expression in early T cells by redirecting partner transcription factor binding. Immunity, 48(6), 1119–1134.e7. https://doi.org/10.1016/j.immuni.2018.04.024. Hosoya, T., Kuroha, T., Moriguchi, T., Cummings, D., Maillard, I., Lim, K.-C., et al. (2009). GATA-3 is required for early T lineage progenitor development. The Journal of Experimental Medicine, 206(13), 2987–3000. https://doi.org/10.1084/jem.20090934. Huang, G., Zhang, P., Hirai, H., Elf, S., Yan, X., Chen, Z., et al. (2008). PU.1 is a major downstream target of AML1 (RUNX1) in adult mouse hematopoiesis. Nature Genetics, 40(1), 51–60. https://doi.org/10.1038/ng.2007.7. Ikawa, T., Hirose, S., Masuda, K., Kakugawa, K., Satoh, R., Shibano-Satoh, A., et al. (2010). An essential developmental checkpoint for production of the T cell lineage. Science, 329(5987), 93–96. https://doi.org/10.1126/science.1188995.

Logical modeling of cell fate specification

235

Izon, D. J., Aster, J. C., He, Y., Weng, A., Karnell, F. G., Patriub, V., et al. (2002). Deltex1 redirects lymphoid progenitors to the B cell lineage by antagonizing Notch1. Immunity, 16(2), 231–243. https://doi.org/10.1016/s1074-7613(02)00271-6. Klarner, H., Streck, A., & Siebert, H. (2017). PyBoolNet: A python package for the generation, analysis and visualization of Boolean networks. Bioinformatics, 33(5), 770–772. https://doi.org/10.1093/bioinformatics/btw682. Koch, U., Fiorini, E., Benedito, R., Besseyrias, V., Schuster-Gossler, K., Pierres, M., et al. (2008). Delta-like 4 is the essential, nonredundant ligand for Notch1 during Thymic T cell lineage commitment. The Journal of Experimental Medicine, 205(11), 2515–2523. https://doi.org/10.1084/jem.20080829. Krebs, L. T., Deftos, M. L., Bevan, M. J., & Gridley, T. (2001). The Nrarp gene encodes an ankyrin-repeat protein that is transcriptionally regulated by the notch signaling pathway. Developmental Biology, 238(1), 110–119. https://doi.org/10.1006/dbio. 2001.0408. Kueh, H. Y., & Rothenberg, E. V. (2012). Regulatory gene network circuits underlying T cell development from multipotent progenitors. Wiley Interdisciplinary Reviews. Systems Biology and Medicine, 4(1), 79–102. https://doi.org/10.1002/wsbm.162. Kueh, H. Y., Yui, M. A., Ng, K. K. H., Pease, S. S., Zhang, J. A., Damle, S. S., et al. (2016). Asynchronous combinatorial action of four regulatory factors activates Bcl11b for T cell commitment. Nature Immunology, 17(8), 956–965. https://doi.org/10.1038/ni.3514. Laiosa, C. V., Stadtfeld, M., & Graf, T. (2006). Determinants of lymphoid-myeloid lineage diversification. Annual Review of Immunology, 24, 705–738. https://doi.org/10.1146/ annurev.immunol.24.021605.090742. Laiosa, C. V., Stadtfeld, M., Xie, H., de Andres-Aguayo, L., & Graf, T. (2006). Reprogramming of committed T cell progenitors to macrophages and dendritic cells by C/EBP alpha and PU.1 transcription factors. Immunity, 25(5), 731–744. https:// doi.org/10.1016/j.immuni.2006.09.011. Lefebvre, J. M., Haks, M. C., Carleton, M. O., Rhodes, M., Gomathinayagam, S., Celeste Simon, M., et al. (2005). Enforced expression of Spi-B reverses T lineage commitment and blocks beta-selection. Journal of Immunology, 174(10), 6184–6194. https://doi.org/ 10.4049/jimmunol.174.10.6184. Le Nove`re, N. (2015). Quantitative and logic modelling of molecular and gene networks. Nature Reviews. Genetics, 16(3), 146–158. https://doi.org/10.1038/nrg3885. Le Nove`re, N., Hucka, M., Mi, H., Moodie, S., Schreiber, F., Sorokin, A., et al. (2009). The systems biology graphical notation. Nature Biotechnology, 27(8), 735–741. https://doi. org/10.1038/nbt.1558. Levy, N., Naldi, A., Hernandez, C., Stoll, G., Thieffry, D., Zinovyev, A., et al. (2018). Prediction of mutations to control pathways enabling tumor cell invasion with the CoLoMoTo interactive notebook (Tutorial). Frontiers in Physiology, 9( July), 787. https://doi.org/10.3389/fphys.2018.00787. Li, L., Leid, M., & Rothenberg, E. V. (2010). An early T cell lineage commitment checkpoint dependent on the transcription factor Bcl11b. Science, 329(5987), 89–93. https:// doi.org/10.1126/science.1188989. Longabaugh, W. J. R., Davidson, E. H., & Bolouri, H. (2005). Computational representation of developmental genetic regulatory networks. Developmental Biology, 283(1), 1–16. https://doi.org/10.1016/j.ydbio.2005.04.023. Longabaugh, W. J. R., Zeng, W., Zhang, J. A., Hosokawa, H., Jansen, C. S., Li, L., et al. (2017). Bcl11b and combinatorial resolution of cell fate in the T-cell gene regulatory network. Proceedings of the National Academy of Sciences of the United States of America, 114(23), 5800–5807. https://doi.org/10.1073/pnas.1610617114. Love, P. E., & Bhandoola, A. (2011). Signal integration and crosstalk during thymocyte migration and emigration. Nature Reviews. Immunology, 11(7), 469–477. https://doi. org/10.1038/nri2989.

236

Elisabetta Cacace et al.

Manesso, E., Chickarmane, V., Kueh, H. Y., Rothenberg, E. V., & Peterson, C. (2013). Computational modelling of T-cell formation kinetics: Output regulated by initial proliferation-linked deferral of developmental competence. Journal of the Royal Society, Interface/the Royal Society, 10(78), 20120774. https://doi.org/10.1098/rsif.2012.0774. Manesso, E., Kueh, H. Y., Freedman, G., Rothenberg, E. V., & Peterson, C. (2016). Irreversibility of T-cell specification: Insights from computational modelling of a minimal network architecture. PLoS ONE, 11(8), e0161260. https://doi.org/10.1371/ journal.pone.0161260. Mingueneau, M., Kreslavsky, T., Gray, D., Heng, T., Cruse, R., Ericson, J., et al. (2013). The transcriptional landscape of αβ T cell differentiation. Nature Immunology, 14(6), 619–632. https://doi.org/10.1038/ni.2590. Miyazaki, M., Miyazaki, K., Chen, K., Jin, Y., Turner, J., Moore, A. J., et al. (2017). The E-id protein axis specifies adaptive lymphoid cell identity and suppresses thymic innate lymphoid cell development. Immunity, 46(5), 818–834.e4. https://doi.org/10.1016/ j.immuni.2017.04.022. Miyazaki, M., Rivera, R. R., Miyazaki, K., Lin, Y. C., Agata, Y., & Murre, C. (2011). The opposing roles of the transcription factor E2A and its antagonist Id3 that orchestrate and enforce the naive fate of T cells. Nature Immunology, 12(10), 992–1001. https://doi.org/ 10.1038/ni.2086. Naldi, A. (2018). BioLQM: A Java toolkit for the manipulation and conversion of logical qualitative models of biological networks. Frontiers in Physiology, 9(November), 1605. https://doi.org/10.3389/fphys.2018.01605. Naldi, A., Hernandez, C., Abou-Jaoud
e, W., Monteiro, P. T., Chaouiya, C., & Thieffry, D. (2018). Logical modeling and analysis of cellular regulatory networks with GINsim 3.0. Frontiers in Physiology, 9(June), 646. https://doi.org/10.3389/fphys. 2018.00646. Naldi, A., Hernandez, C., Levy, N., Stoll, G., Monteiro, P. T., Chaouiya, C., et al. (2018). The CoLoMoTo interactive notebook: Accessible and reproducible computational analyses for qualitative biological networks. Frontiers in Physiology, 9, 680. https://doi.org/ 10.3389/fphys.2018.00680. Oravecz, A., Apostolov, A., Polak, K., Jost, B., Le Gras, S., Chan, S., et al. (2015). Ikaros mediates gene silencing in T cells through polycomb repressive complex 2. Nature Communications, 6, 8823. https://doi.org/10.1038/ncomms9823. Outters, P., Jaeger, S., Zaarour, N., & Ferrier, P. (2015). Long-range control of V(D)J recombination & allelic exclusion: Modeling views. Advances in Immunology, 128, 363–413. https://doi.org/10.1016/bs.ai.2015.08.002. Pai, S.-Y., Truitt, M. L., Ting, C.-N., Leiden, J. M., Glimcher, L. H., & Ho, I.-C. (2003). Critical roles for transcription factor GATA-3 in thymocyte development. Immunity, 19(6), 863–875. https://doi.org/10.1016/s1074-7613(03)00328-5. Paulev
e, L. (2017). Pint: A static analyzer for transient dynamics of qualitative networks with IPython interface. In Vol. 10545. CMSB 2017—15th conference on computational methods for systems biology (pp. 306–316). Springer International Publishing. https://doi.org/ 10.1007/978-3-319-67471-1_20. Radtke, F., Ferrero, I., Wilson, A., Lees, R., Aguet, M., & MacDonald, H. R. (2000). Notch1 deficiency dissociates the Intrathymic development of dendritic cells and T cells. The Journal of Experimental Medicine, 191(7), 1085–1094. https://doi.org/10.1084/jem.191. 7.1085. Rothenberg, E. V. (2011). T cell lineage commitment: Identity and renunciation. Journal of Immunology, 186(12), 6649–6655. https://doi.org/10.4049/jimmunol.1003703. Rothenberg, E. V. (2019a). Programming for T-lymphocyte fates: Modularity and mechanisms. Genes & Development, 33(17–18), 1117–1135. https://doi.org/10.1101/gad. 327163.119.

Logical modeling of cell fate specification

237

Rothenberg, E. V. (2019b). Dynamic control of the T-cell specification gene regulatory network. Current Opinion in Systems Biology, 18, 62–76. https://doi.org/10.1016/j.coisb. 2019.10.012. Scripture-Adams, D. D., Damle, S. S., Li, L., Elihu, K. J., Qin, S., Arias, A. M., et al. (2014). GATA-3 dose-dependent checkpoints in early T cell commitment. Journal of Immunology, 193(7), 3470–3491. https://doi.org/10.4049/jimmunol.1301663. Speck, N. A. (2001). Core binding factor and its role in normal hematopoietic development. Current Opinion in Hematology, 8(4), 192–196. https://doi.org/10.1097/00062752200107000-00002. Stoll, G., Caron, B., Viara, E., Dugourd, A., Zinovyev, A., Naldi, A., et al. (2017). MaBoSS 2.0: An environment for stochastic Boolean modeling. Bioinformatics, 33(14), 2226–2228. https://doi.org/10.1093/bioinformatics/btx123. Taghon, T. N., David, E.-S., Zu´n˜iga-Pfl€ ucker, J. C., & Rothenberg, E. V. (2005). Delayed, asynchronous, and reversible T-lineage specification induced by notch/delta signaling. Genes & Development, 19(8), 965–978. https://doi.org/10.1101/gad.1298305. Taghon, T., Van de Walle, I., De Smet, G., De Smedt, M., Leclercq, G., Vandekerckhove, B., et al. (2009). Notch signaling is required for proliferation but not for differentiation at a well-defined beta-selection checkpoint during human T-cell development. Blood, 113(14), 3254–3263. https://doi.org/10.1182/blood-2008-07-168906. Taghon, T., Yui, M. A., & Rothenberg, E. V. (2007). Mast cell lineage diversion of T lineage precursors by the essential T cell transcription factor GATA-3. Nature Immunology, 8(8), 845–855. https://doi.org/10.1038/ni1486. Talebian, L., Li, Z., Guo, Y., Gaudet, J., Speck, M. E., Sugiyama, D., et al. (2007). T-lymphoid, megakaryocyte, and granulocyte development are sensitive to decreases in CBFbeta dosage. Blood, 109(1), 11–21. https://doi.org/10.1182/blood-200605-021188. Ting, C. N., Olson, M. C., Barton, K. P., & Leiden, J. M. (1996). Transcription factor GATA-3 is required for development of the T-cell lineage. Nature, 384(6608), 474–478. https://doi. org/10.1038/384474a0. Van de Walle, I., De Smet, G., De Smedt, M., Vandekerckhove, B., Leclercq, G., Plum, J., et al. (2009). An early decrease in notch activation is required for human TCR-alphabeta lineage differentiation at the expense of TCR-Gammadelta T cells. Blood, 113(13), 2988–2998. https://doi.org/10.1182/blood-2008-06-164871. Wada, H., Masuda, K., Satoh, R., Kakugawa, K., Ikawa, T., Katsura, Y., et al. (2008). Adult T-cell progenitors retain myeloid potential. Nature, 452(7188), 768–772. https://doi. org/10.1038/nature06839. Wang, H.-C., Qian, L., Zhao, Y., Mengarelli, J., Adrianto, I., Montgomery, C. G., et al. (2017). Downregulation of E protein activity augments an ILC2 differentiation program in the thymus. Journal of Immunology, 198(8), 3149–3156. https://doi.org/10.4049/ jimmunol.1602009. Weber, B. N., Chi, A. W.-S., Chavez, A., Yashiro-Ohtani, Y., Yang, Q., Shestova, O., et al. (2011). A critical role for TCF-1 in T-lineage specification and differentiation. Nature, 476(7358), 63–68. https://doi.org/10.1038/nature10279. Wilson, A., MacDonald, H. R., & Radtke, F. (2001). Notch 1-deficient common lymphoid precursors adopt a B cell fate in the thymus. The Journal of Experimental Medicine, 194(7), 1003–1012. https://doi.org/10.1084/jem.194.7.1003. Wong, W. F., Nakazato, M., Watanabe, T., Kohu, K., Ogata, T., Yoshida, N., et al. (2010). Over-expression of Runx1 transcription factor impairs the development of thymocytes from the double-negative to double-positive stages. Immunology, 130(2), 243–253. https://doi.org/10.1111/j.1365-2567.2009.03230.x.

238

Elisabetta Cacace et al.

Xu, W., Carr, T., Ramirez, K., McGregor, S., Sigvardsson, M., & Kee, B. L. (2013). E2A transcription factors limit expression of Gata3 to facilitate T lymphocyte lineage commitment. Blood, 121(9), 1534–1542. https://doi.org/10.1182/blood-2012-08-449447. Yashiro-Ohtani, Y., He, Y., Ohtani, T., Jones, M. E., Shestova, O., Xu, L., et al. (2009). Pre-TCR signaling inactivates Notch1 transcription by antagonizing E2A. Genes & Development, 23(14), 1665–1676. https://doi.org/10.1101/gad.1793709. Ye, Y., Kang, X., Bailey, J., Li, C., & Hong, T. (2019). An enriched network motif family regulates multistep cell fate transitions with restricted reversibility. PLoS Computational Biology, 15(3,) e1006855. https://doi.org/10.1371/journal.pcbi.1006855. Yui, M. A., Feng, N., & Rothenberg, E. V. (2010). Fine-scale staging of T cell lineage commitment in adult mouse thymus. Journal of Immunology, 185(1), 284–293. https://doi.org/ 10.4049/jimmunol.1000679. Yui, M. A., & Rothenberg, E. V. (2014). Developmental gene networks: A triathlon on the course to T cell identity. Nature Reviews. Immunology, 14(8), 529–545. https://doi.org/ 10.1038/nri3702. Yun, T. J., & Bevan, M. J. (2003). Notch-regulated ankyrin-repeat protein inhibits Notch1 signaling: Multiple Notch1 signaling pathways involved in T cell development. Journal of Immunology, 170(12), 5834–5841. https://doi.org/10.4049/jimmunol.170.12.5834. Zarnegar, M. A., Chen, J., & Rothenberg, E. V. (2010). Cell-type-specific activation and repression of PU.1 by a complex of discrete, functionally specialized cis-regulatory elements. Molecular and Cellular Biology, 30(20), 4922–4939. https://doi.org/10.1128/MCB. 00354-10. Zhang, J. A., Mortazavi, A., Williams, B. A., Wold, B. J., & Rothenberg, E. V. (2012). Dynamic transformations of genome-wide epigenetic marking and transcriptional control establish T cell identity. Cell, 149(2), 467–482. https://doi.org/10.1016/j.cell.2012.01.056. Zhou, W., Yui, M. A., Williams, B. A., Yun, J., Wold, B. J., Cai, L., et al. (2019). Single-cell analysis reveals regulatory gene expression dynamics leading to lineage commitment in early T cell development. Cell Systems, 9(4), 321–337.e9. https://doi.org/10.1016/j.cels. 2019.09.008.

CHAPTER EIGHT

Repressive interactions in gene regulatory networks: When you have no other choice M. Joaquina Delás, James Briscoe∗ The Francis Crick Institute, London, United Kingdom ∗ Corresponding author: e-mail address: [email protected]

Contents 1. Introduction 2. The developing neural tube: A GRN in time and space 2.1 Concentration interpretation: More than Gli affinities 2.2 Gradient interpretation in time 3. Cis regulatory elements and the role of repression 4. From subcircuits to complex dynamics 5. What can the GRN do for you? 6. Molecular mechanisms for implementing the GRN cross-repressive interactions 6.1 CRE repression 6.2 Repression by protein-protein interactions 6.3 Repression at the RNA level 7. Conclusions 8. Acknowledgments References

240 241 242 247 248 252 254 257 258 259 260 261 262 262

Abstract Tightly regulated gene expression programs, orchestrated by complex interactions between transcription factors, control cell type specification during development. Repressive interactions play a critical role in these networks, facilitating decision-making between two or more alternative cell fates. Here, we use the ventral neural tube as an example to illustrate how cross repressive interactions within a network drive pattern formation and specify cell types in response to a graded patterning signal. This and other systems serve to highlight how external signals are integrated through the cis regulatory elements controlling key genes and provide insight into the molecular underpinning of the process. Even the simplest networks can lead to counterintuitive results and we argue that a combination of experimental dissection and modeling approaches will be necessary to fully understand network behavior and the underlying design principles. Studying these gene regulatory networks as a whole ultimately allows us to extract fundamental properties applicable across systems that can expand our mechanistic understanding of how organisms develop. Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.03.003

#

2020 Elsevier Inc. All rights reserved.

239

240

M. Joaquina Delás and James Briscoe

1. Introduction The large diversity of cell types that form complex organs and tissues in multicellular organisms originate from small groups of pluripotent progenitors that establish the tissue. In most cases, each of the progenitors that seed a tissue bears the competency to generate any of the final cell types. Over time and usually in response to external signaling molecules, progenitors transition from multipotency into specialized and functionally distinct cells. This complexity does not arise all at once but from a progressive restriction in a cell’s differentiation potential to particular fates that eventually leads to the terminal differentiation event. The identity of cells in a tissue can be defined by expression of a characteristic sets of genes. These genes include those encoding the effector proteins responsible for a cell’s function (e.g., the T cell receptor subunits in a T cell, or specific receptors and neurotransmitters in a neuron), as well as the genes that encode for transcriptional regulators that control the genes being transcribed or repressed in the cell. This has focused attention on gene regulation and the function of transcription factors in establishing cell function and controlling cell type diversity. A common observation in developing tissues is that the combinations of transcription factors (TFs) that control cell identity form a network of interactions. This network is determined by the sequence-specific binding specificities of the TFs that direct them to target genes. This creates complex sets of regulatory interactions that shape the gene expression patterns in different cell types. The gene regulatory network (GRN) theory (Davidson, 2010) provides a framework to understand these relationships and explore their outputs. In this view, the nodes of the network comprise the TFs that direct the gene expression programs of the system. The network edges are the interactions between these nodes, mainly specified by cis-regulatory elements (CREs), non-coding DNA stretches where TFs bind to exert their regulatory functions. These CREs interact with multiple TFs and act as the information processing devices of the GRN (Davidson, 2010). Conversely, each TF interacts with multiple CREs controlling expression of many genes including their own expression and that of other TFs. These recursive links within a GRN, together with external signals regulating transcriptional effectors, create the dynamics of gene expression. Understanding how these dynamics arise and the effect they have on the spatial and temporal pattern of gene expression within a tissue provides a casual explanation for cell fate specification during development and for the homeostatic maintenance of adult tissues.

Repressive interactions in gene regulatory networks

241

GRNs operating in several tissues have been particularly well studied. Davidson and colleagues assembled the most detailed GRN to date, that of sea urchin endomesoderm development, comprising spatial expression of TFs and their interactions (Peter & Davidson, 2011). Beyond being a diagram of interactions, this tool can also be used to model the dynamics of the system. In this way one can follow the sequence of gene expression and the resulting state transitions in the system and compare them to experimental observations. In order to do this, a temporal component must be added to the interactions, either by assuming each transaction in the system—the activation or repression of a gene—takes a fixed amount of time or by inferring the production and degradation dynamics of each factor. While the sea urchin endomesoderm represents a particularly large and comprehensive description of a GRN, smaller but functionally relevant GRNs can allow the investigation of dynamical behaviors of TFs during cell fate decisions and reveal mechanistic insights that would not be obvious by studying the different TF relationships in isolation. Indeed, dissecting GRNs in this way can identify principles that characterize mechanisms commonly used to perform specific tissue patterning functions. Based on paradigms such as cellular reprograming and transdifferentiation, where TFs can instruct the fate of a cell by overwriting the current gene expression state, the transactivating function of TFs is often emphasized (Niwa, 2018). Positively acting TFs can result in self-reinforcing activation to produce stable cellular states. However, negative regulation between TFs and the action of TFs repressing effector genes have an equally important role in determining a cell’s fate. Using the vertebrate neural tube development as our central theme, we will explore the role of repressive interactions in cell fate decisions and explore how the GRN concept can help us understand and extend principles of gene regulation.

2. The developing neural tube: A GRN in time and space It is conceptually easy to imagine how equivalent cells could acquire different fates if they are exposed to different signals, with each signal inducing a different cell type. However, in many tissues a single signal secreted from a localized source generates a spatial gradient and instructs multiple cell fates in a position dependent manner (Rogers & Schier, 2011). The interactions between the TFs already present in the cell and those produced in response to signaling interpret this graded input and generate gene expression patterns in a tissue. Work in the vertebrate neural tube serves as a good example of how these regulatory interactions can be brought together into a

242

M. Joaquina Delás and James Briscoe

GRN that explains pattern formation. Because the interactions can be rather complex, mathematical modeling helps understand the network behavior and extract the working principles. In the ventral neural tube distinct neural progenitors form in response to a gradient of Sonic Hedgehog (Shh) initially emanating from the notochord and later from the floor plate. Each of these progenitors will generate specific neuronal subtypes, and the location, numbers and connections these neurons make are essential for the proper relay of sensory information to motor outputs (Lai, Seal, & Johnson, 2016). Along this dorsoventral (DV) axis, the expression of different TFs can be detected at different distances from the source and span varying widths of the DV axis. This results in 11 different progenitor domains that can be easily identified by their expression of a particular combination of TFs (Fig. 1A). Closest to the source of Shh are the p3 progenitors, characterized by expression of homeobox TF Nkx2.2. Adjacent to the p3 domain are the motor neuron progenitors (pMN), which express the bHLH TF Olig2 and low levels of Pax6, followed by the p2 progenitors, which express high levels of Pax6 and Irx3. These three domains all also express Nkx6.1, which will not be expressed dorsally beyond p2 (Briscoe et al., 1999; Briscoe, Pierani, Jessell, & Ericson, 2000; Ericson et al., 1997; Novitch, Chen, & Jessell, 2001). Each of these progenitor domains generate postmitotic neuronal subtypes of defined identity, motor neurons from the pMN domain and V3 and V2 interneurons from the p3 and p2 domain, respectively (Sagner & Briscoe, 2019). The use of graded signals offers a new layer of information, position— determined by the distance from the source—that cells must interpret to adopt their appropriate identity. This poses a challenge. Cells have to interpret not just the identity of the signal but additional properties, such as signal concentration, and this information has to be integrated to direct cell fate decisions.

2.1 Concentration interpretation: More than Gli affinities A simple and elegant way in which a graded signal could control differential gene expression is if the different concentration of signal received by the cell would result in different levels of effector TF. By having CREs with different affinities for the effector TF or different numbers of binding sites, the corresponding genes would respond to different concentrations of the effector and thus respond in a position dependent manner. These types of affinity-based models were initial proposed to explain, for example, how

Repressive interactions in gene regulatory networks

243

Fig. 1 The vertebrate neural tube GRN. (A) Cross-section of an embryo and diagram of the dorsoventral domains that arise in response to signaling gradients, with a focus in the ventral part. (B) Resulting expression pattern diagrams for mutant embryos with altered signaling dynamics or lacking TFs characteristic of specific domains. (C) Interactions that make up the ventral neural tube GRN: cross-repressive interactions between domain-specific TFs Nkx2.2, Olig2, Pax6 and Irx3; widespread positive inputs into all domain represented by, but not limited to, Sox2; and positive and negative inputs from the Shh pathway effector, GliA/GliR. Arrows indicate activating interactions; T bars indicate repressive ones.

the Bicoid (bcd) gradient could activate different targets along the Drosophila embryo anterior-posterior axis. Binding sites for bcd with different affinities drove expression of a reporter gene at different distances from the source, with high affinity sites leading to expression in up to half the embryo, and weakest sites in a small domain next to the source (Driever, Thoma, & N€ usslein-Volhard, 1989; Struhl, Struhl, & Macdonald, 1989). However, a broader survey of bcd binding sites associated with target genes found no strong correlation between domain expression and predicted bcd binding affinity (Ochoa-Espinosa et al., 2005) suggesting this simplest interpretation might not be enough. The vertebrate ventral neural tube is a good system to exemplify some of the shortcomings of this model and illustrate how GRNs can help explain differential target gene expression in a way that does not depend solely on binding affinity of an activator.

244

M. Joaquina Delás and James Briscoe

The intracellular signaling pathway for Shh results in the regulation of transcriptional effectors, members of the Gli protein family (Hui & Angers, 2011). There are three Gli proteins in mammals, Gli1, Gli2 and Gli3. In the absence of Shh signaling full length Gli2 and Gli3 are processed into repressive forms (GliR) (Pan, Bai, Joyner, & Wang, 2006; Wang, Fallon, & Beachy, 2000), with Gli3 being the principal repressor (Litingtung & Chiang, 2000; Persson et al., 2002). In response to Shh, however, activating isoforms of Gli are produced (GliA) at the expense of GliR isoforms. All three Gli proteins contribute to the activating function, but Gli1 is itself a downstream gene of the Gli pathway, thus augmenting the response to the signaling inputs (Lee, Platt, Censullo, & Ruiz i Altaba, 1997). Parallel negative feedback loops are also in place to keep the pathway in check (Hui & Angers, 2011). Because both the GliA and GliR isoforms bind to the same site it is hard to predict the effects that changing the affinity would have. An increase in affinity would increase the propensity for both GliA and GliR binding and vice versa. Indeed, although alterations to the predicted Gli-binding affinity in the regulatory elements of neural tube TF genes has an effect in reporter assays, especially for elements expressed in the most ventral domains (Peterson et al., 2012), the effects are not always straightforward. For example, when the CRE of ventral gene Nkx2.2 is mutated to contain the Gli binding site of a much more dorsal gene, Dbx1, the ventral expression domain of the reporter in vivo does not change. Furthermore, ectopic expression of the reporter is observed dorsally (Oosterveen et al., 2012). Together with cellular reporter data showing that low concentrations of GliR are enough to suppress the activity of GliA, this argues for an instructive role of Gli repressors (Oosterveen et al., 2012). This logic would dictate that cells are primarily interpreting the gradient in GliR, and de-repression of elements controls expression patterns. Moreover, mice lacking both Shh and Gli3, the main source of GliR in mammals (Hui & Angers, 2011), are able to generate many of the ventral neural progenitor cell types, including pMNs, although not those closest to the source (floor plate and p3) (Litingtung & Chiang, 2000; Persson et al., 2002) (Fig. 1B). Therefore, the Shh-induced transcriptional de-repression of these genes, and not solely the actions of GliA, plays a central role in generating most of the progenitors along the neural tube DV axis. Genetic mutant analysis indicates that information integration is even more complex. The different domain-defining genes mentioned, Nkx2.2, Olig2, Pax6 and Irx3 are in fact repressive TFs that are essential for appropriate domain formation. For instance, in mouse embryos mutant for Olig2

Repressive interactions in gene regulatory networks

245

and/or Pax6, the Nkx2.2-expressing domain expands in size, away from the Shh source, to occupy the domain of cells where the deleted genes used to be expressed (Balaskas et al., 2012; Ericson et al., 1997; Zhou & Anderson, 2002) (Fig. 1B). Given that the intrinsic affinity of the Nkx2.2 regulatory element for Gli-binding proteins has not changed, this indicates that the presence of Pax6 and/or Olig2 negatively affects the ability of Gli activator to induce expression of Nkx2.2 (Balaskas et al., 2012). Alternatively, these factors could affect how the Shh signal is transduced or feedback to the signaling pathway (Lek et al., 2010), however, reporter assays suggested this was not the case (Balaskas et al., 2012). Indeed, genetic evidence has shown repressive interactions between most of these TFs (Briscoe et al., 1999; Novitch et al., 2001) that, together with genomic analysis of binding sites (Oosterveen et al., 2012; Peterson et al., 2012), supports a model where the cell fate decisions in the ventral neural tube are controlled by a gene regulatory network of repressive TFs. The boundary between p3, the most ventral domain, and pMN, is defined by the cross-repression between Nkx2.2 and, Pax6 and Olig2. Nkx2.2 and Olig2 also form two cross repressive interactions with Irx3, expressed in the p2 domain. Pax6 is repressed unidirectionally by Olig2 and is present in the pMN domain albeit at low levels. The ventral boundary of Pax6 will be established by Nkx2.2 expression in the p3 domain, which represses this gene (Fig. 1C). The abundant knowledge of how repressors control neural tube patterning raises the question of what performs the activator function. As mentioned, the Shh pathway, via the resulting GliA/R balance generates the spatial asymmetry that initiates pattern, yet the specific activator function of Gli is only required for the generation of the p3 domain, the ventral-most cell type. SoxB1 members Sox1–3 are uniformly expressed in neural progenitors, required to maintain the cells in this state and perform partially overlapping functions (Bylund, Andersson, Novitch, & Muhr, 2003; Graham, Khudyakov, Ellis, & Pevny, 2003). They are thus thought to represent a positive input throughout the different domains of the neural tube. Supporting this, Sox2 binding to cis-regulatory elements near many of the key neural progenitor TFs has been shown to be required for their expression and occur in the same cis-regulatory elements as Gli binding (Oosterveen et al., 2012; Peterson et al., 2012). Sox family TFs are known to act in complex with other TFs in numerous developmental contexts, a mechanism that is believed to be required for its transcriptional effector activity (Kamachi & Kondoh, 2013) and it is possible that Sox and GliA interact to support activation of target genes. While GliA could act as the

246

M. Joaquina Delás and James Briscoe

activator partner for some the most ventral cell types, this would not explain what the activator input is elsewhere. Retinoic acid (RA) and the activating function of its receptors are known to be required for the specification of intermediate and ventral neural progenitors (Novitch, Wichterle, Jessell, & Sockanathan, 2003; Pierani, Brenner-Morton, Chiang, & Jessell, 1999). RA is secreted by the adjacent somites and it could represent another uniform positive input for neural progenitor gene expression. This means ventral neural progenitors are simultaneously receiving a graded signaling input encoding positional information in the form of a bi-functional repressor-activator TF (Gli), positive inputs from RA and pan-neurally expressed factors such as Sox2, and repressive inputs from domain-specific TFs. Only by integrating these three types of information can cells establish the appropriate gene expression pattern that will determine neural progenitor cell fate. While the information integration must ultimately happen at the level of the CREs that drive gene expression, how it occurs mechanistically is not well understood. The complex interplay between inputs and regulatory interactions among TFs makes the system difficult to understand. Mathematical models are essential to assimilate observation and test assumptions. Analyzing how the network responds over time after receiving the input reveals a temporal dimension to the gene expression response that is hard to tackle experimentally. The outputs obtained from theoretical approaches are compared to the experimental knowledge from in vivo perturbations and adjusted to reflect the expected behavior. Following these studies, a mathematical model that describes the regulation of gene expression (TF binding, gene transcription, etc.) (Sherman & Cohen, 2012) was used to model the ventral neural tube with inputs from Gli, and with Olig2, Nkx2.2, Pax6 and Irx3 as nodes (Cohen, Page, Perez-Carrasco, Barnes, & Briscoe, 2014). While the spatial outputs were constrained to represent the experimental knowledge, the parameters recovered that fit these requirements also recapitulated the known temporal dynamics of these genes in vivo across the different positions. This is especially interesting for Olig2 and Nkx2.2-expressing cells, as it is known that Olig2 expression is initiated before Nkx2.2 ( Jeong, 2004; Stamataki, Ulloa, Tsoni, Mynett, & Briscoe, 2005) and cell transitions from Olig2 + to Nkx2.2+ have been shown by lineage tracing (Dessaud et al., 2007). These results emphasize that the response to a graded input occurs not only in space but also in time, and that this dynamic response is important for determining cell fate decisions.

Repressive interactions in gene regulatory networks

247

2.2 Gradient interpretation in time If the response occurs over time and it is not an immediate result of the signal concentration, the amount of time that a cell is exposed to the signal can affect its interpretation. Indeed, this is the case in the ventral neural tube, where a temporal adaptation to Shh links a cell’s response to both concentration and duration of exposure (Dessaud et al., 2007). Initially shown for the transition of p3 through an Olig2+ state, this mechanism was shown to be generally applicable to all the ventral neural cell types. While different concentrations of Shh alone could not differentially generate all the expected neural sub-types, different exposure times of chick explants to the same concentration could produce peaks in specific neuronal subtypes (Dessaud et al., 2010). The dynamics of the network, whereby both the level and time of the signaling input are integrated to dictate the transcriptional output can explain how these two seemingly distinct parameters are linked. This combined integration of concentration and time by the GRN can also help explain previous observations of a temporal component to pattern formation in other systems. In early Xenopus laevis embryos, cells from the animal pole that are exposed to increasing concentrations of purified activin generate different fates (mesenchyme at low concentration versus notochord and muscle at higher). Yet increasing exposure times at a constant concentration also had a similar effect (Green, Howes, Symes, Cooke, & Smith, 1990). In Drosophila wing imaginal discs, Wnt is expressed in the dorsoventral boundary, from where it is secreted to form a gradient (Zecca, Basler, & Struhl, 1996). Despite its well-established role as a secreted morphogen, a study showed that membrane tethered Wg (the main Drosophila Wnt protein) could replace endogenous Wnt function with little if any adverse consequence for pattern formation and only mild growth defects (Alexandre, Baena-Lopez, & Vincent, 2014). Wnt cannot form a gradient in these flies and only cells adjacent to the source receive signaling. It was thus puzzling how the long-range target genes of this morphogen were being activated. Wnt expression turned out to be initially expressed throughout the wing disc and its expression gradually refined to the dorsoventral boundary, thereby generating a temporal gradient along the dorsoventral axis (Alexandre et al., 2014). The authors hypothesize this temporal gradient might act through epigenetic mechanisms to maintain expression of target genes. The strategy to interpret a graded signal and convert it to distinct cellular fates is not limited to the ventral neural tube. BMPs secreted from the roof

248

M. Joaquina Delás and James Briscoe

plate have been shown to control the identity of the dorsal neuronal subtypes that are produced (Liem, Tremml, & Jessell, 1997). The neuronal progenitors that generate these neurons are also defined by a code of TFs with crossrepressive interactions (Lai et al., 2016). In vitro assays indicate that BMPs can induce dorsal cell types (Andrews et al., 2017; Gupta et al., 2018; Liem et al., 1997; Tozer, Le Dreau, Martı´, & Briscoe, 2013). Recent work has proposed that BMPs may not act solely in graded manner to specify cell type identity and instead the identity of BMP ligands is important (Andrews et al., 2017). However, the combined activity of BMP ligands appears graded at early developmental stages, as assays for phosphorylated SMADs (the intracellular effectors of BMP signaling) reveal a dorsal to ventral gradient (Tozer et al., 2013; Zagorski et al., 2017). These seemingly contradictory findings might also be reconciled if the timing of exposure plays a role in the interpretation of BMP signaling (Duval et al., 2019). Moreover, BMP inhibitors have been shown to be present in the ventral neural tube (Liem, Jessell, & Briscoe, 2000), bringing together the idea that these the gene regulatory network integrates both signaling pathways across the DV axis. This combination of signals has been proposed to confer accurate positional information along the entire axis (Zagorski et al., 2017). By analyzing the positional error, the authors showed that either signal could only confer positional identity close to the source, whereas the combination of both gradients can be used to accurately provide positional information for cells in the center of the neural tube, distant from both sources. This was proposed to depend on the GRN and suggested that the regulatory interactions between TFs in the network are able to compute a number of inputs to successfully and robustly determine cell fate.

3. Cis regulatory elements and the role of repression The genetic evidence from mutants and forced expression experiments, together with the dynamical models suggests that cell type-specific gene expression programs are established by integrating multiple inputs. In the ventral neural tube these include the pan-neural Sox2 TFs and related factors, which provide a uniform positive input, the bifunctional Gli TF, acting as a repressor or activator, in a Shh-dependent manner, and a number of cross-repressive interactions between TFs. Several different domains are formed in response to Shh signaling, with some of the same activators, such as Sox2, promoting all of them. As a mechanism to ensure reliable selection of a domain-specific gene expression program, each domain-specific TF

Repressive interactions in gene regulatory networks

249

repress all other possible fates achieving specificity by ‘selectivity by exclusion.’ Integrating the positive and negative inputs into a gene are the CREs, which provide the platform through which the different signals are assimilated to direct appropriate gene repression or activation. Analysis of conserved noncoding DNA sequences containing Gli binding sites (Oosterveen et al., 2012) and genome-wide DNA-binding of Gli1 and Sox2 (Peterson et al., 2012) identified potential CREs controlling expression of many of the ventral neural tube TFs. Extensive in vivo reporter assays indeed showed that these elements drive gene expression in the appropriate neural tube domain and that they contained predicted binding site for Sox2 and Gli that are required for their function (Oosterveen et al., 2012; Peterson et al., 2012). This work emphasized how either SoxB1 or Gli inputs in isolation are not sufficient to drive expression, therefore supporting that the combination of inputs is necessary. Direct binding of domain-specific repressive TFs Olig2, Nkx2.2 and Nkx6.1 to DNA was further assessed genome-wide, revealing that these factors bind some of the same regions as Sox2 and Gli (Kutejova, Sasai, Shah, Gouti, & Briscoe, 2016; Nishi et al., 2015) thus repressing each other’s expression (Fig. 2A). Strikingly, these unbiased global assays revealed that the TFs bind not only to the regulatory regions of the neighboring domain TFs, but also repress expression of non-adjacent domains TFs, as well as that of effector genes that characterize those alternative fates (Kutejova et al., 2016; Nishi et al., 2015). This way, while the positive acting TFs, including signaling inputs from Shh, are directly and concurrently activating gene expression of several ventral domains, the domain-specific TFs expressed in response to these inputs directly inhibit both the TFs and effector genes of discordant domains. This results in a densely interconnected repressive network (Fig. 2B). The binding regions for three of these domain-specific TFs in the ventral cell types (Nkx6.1, Nkx2.2 and Olig2) showed extensive overlap, with the potential target genes including genes from alternative domains (Nishi et al., 2015). This overlap suggests the network operates through shared CREs, which integrate negative inputs to ensure that only genes for the appropriate domain are active. This ‘selectivity by exclusion’ model controls the specification of progenitor identity in the vertebrate neural tube. While positively acting factors, including Shh-induced Gli activity and uniformly expressed TFs such as SoxB1 promote multiple identities, the network of transcriptional

250

M. Joaquina Delás and James Briscoe

Fig. 2 Cis regulatory elements and the role of repression. (A) Activating and repressive inputs are integrated to direct gene expression outcomes. (B) Cross-repressive interactions between TFs, and additional repression of target genes by all alternative TFs leads to a ‘selectivity by exclusion’ fate decision. (C) Three types of inputs are integrated: broadly activators that promote multiple fates, signaling inputs that introduce a bias, and finally, cell type specific repressors, responsible for the acquisition and commitment to a gene expression program.

repressors induced by this signal determines the resulting specific cell identity. This strategy enables progenitor cells to select a signal definitive transcriptional program from several initially permitted options (Fig. 2C). The design principles extracted from this system seem to apply in other systems, such as the pair rule stripe formation in Drosophila. In this case, uniformly distributed positive inputs, including the TF Zelda, promote several alternative fates along the anterior-posterior axis of the fly embryo. Also positively acting, bcd is expressed in a gradient and involved in activating possible outcomes. It is the network of cross-repressing TFs, the gap genes, that will enforce a selectivity by an exclusion mechanism resulting in the characteristic stripe formation. An extensive body of work has built the cross repressive interactions between gap genes ( Jaeger, 2011). While their direct regulation of each other’s target genes is less explored, it would provide a mechanism for the pattern that forms during development.

Repressive interactions in gene regulatory networks

251

It is tempting to speculate that similar ‘selectivity by exclusion’ principles apply broadly across different developmental systems. The best documented examples are cross-repressive fates where a binary lineage choice needs to be made. In the sea urchin skeletogenic micromere lineage, for example, the mesoderm lineage is repressed by skelogenic determinants, thus ensuring confined fate decisions (Oliveri, Tu, & Davidson, 2008). And perhaps one of the classical paradigms is mammalian hematopoietic stem cell differentiation, where the PU1(Spi1)-Gata1 antagonistic interaction will dictate the myeloid-erythroid cell fate decision (Orkin, 2000). Enforced expression of either TF will direct or reprogram cells to their respective fates (Kulessa, Frampton, & Graf, 1995; Nerlov & Graf, 1998) through a combination of antagonistic interaction (Rekhtman, Radparvar, Evans, & Skoultchi, 1999) and by promoting the expression of their own gene expression program. Additional mechanisms are required to proceed past the binary choice into lineage commitment and differentiation. In the PU1-Gata1 example, each of these TFs are activators that promote their own expression by binding to CREs near their transcription start site, which results in commitment to the lineage choice (Reviewed in (Graf & Enver, 2009)). Commitment in the neural tube is essential to communicate progenitor identity to the neuronal progeny. Here, each neuronal subtype will differentiate from their respective progenitor and the number and position where they are born are essential for tissue function. This added layer of complexity is harder to study in bulk and it has been addressed at the single cell level in spinal cord motor neuron development (Sagner et al., 2018). Interestingly, in the predicted transition of Olig2 + pMN to differentiated motor neurons, there was a further increase in Olig2, which was validated in vivo (Sagner et al., 2018). This would suggest a model where the decision between alternative progenitor domains is driven by intermediate levels of these TFs, whereas once the TF levels increase above a threshold the cells enters an irreversible commitment phase, which will conclude with neuronal differentiation. An equivalent pattern has been observed for other TFs in predicted pseudotime transitions between progenitors from different domains and their respective neuronal subtypes (Delile et al., 2019). An interesting mechanistic challenge is how these repressive interactions occur at the level of the CREs. The binding of specific activating and repressing inputs will reliably result in a gene expression outcome. This integration of inputs can occur at least at two different levels. In some circumstances all the TFs regulating a gene will bind in the same element, whereas in other

252

M. Joaquina Delás and James Briscoe

cell states, there can be multiple enhancers actively controlling a gene. How these interactions would then occur in 3D and whereas repressive inputs override activating ones are still open questions.

4. From subcircuits to complex dynamics Specific relationships between TFs lead to a predictable functional outcome. Recurrent basic topologies, known as motifs or subcircuits, have been proposed to carry out stereotypical information-processing functions and have been extensively studied in the developmental biology context (Alon, 2007). Even in relatively simple networks, motifs such as feedback loops can result in dynamical behaviors that can be hard to understand. For this reason, the ability to mathematically model the specific interactions and the more complex circuits they form has made it possible to understand the functionality of some circuits and predict the biologically use of others. The initial building blocks are activation and repression, which can then be combined resulting in more sophisticated behaviors (Fig. 3A). For example, if a TF is activated and then it self-represses, this negative autoregulation facilitates reaching steady state levels faster and can help reduce the cell-tocell variations in TF levels (Alon, 2007). Combining two genes with mutually cross repressing interactions, such as the Nkx2.2-Olig2 or Pu1-Gata1 examples, produces a bistable switch. Each TF promotes its own gene expression program while repressing the alternative one. This offers a mechanism for establishing alternative fates (Fig. 3B). Other motifs have been described that can ensure that a factor is only expressed for a short period of time, if the activator of that TF also directs induction of its repressor. In development, this transient expression is useful to generate intermediary cell types (Fig. 3C). At the molecular level, depending on the parameters of the system, this is a type of incoherent feedforward that can also provide a mechanism for detecting fold changes in the input signal, at least according to theoretical studies (Goentoro, Shoval, Kirschner, & Alon, 2009). The dynamics of this circuit results in changes in the input signal effecting a response and is less sensitive to the absolute levels of signal. Thus, different cells can generate the same gene response as a result of similar change in input signal, even if they are not exposed to similar levels of signal. This could have several advantages such as producing the same response even in the presence of noise. Many other motifs or sub-circuits have been defined and shown to function in development, for purposes such as boundary maintenance or terminal binary cell fate choice (Davidson, 2010).

Repressive interactions in gene regulatory networks

253

Fig. 3 Example motifs and resulting gene expression outcomes. (A) A simple activating signal leads to gene activation. (B) Antagonistic interactions between two TFs can switch the gene expression pattern upon signal induction. (C) Incoherent feedforward loop induces a gene and its repressor. (D) The AC-DC multi-functional circuit can lead to different behaviors under different signaling regimes or starting conditions.

Analysis of subcircuits sometimes assumes the topology of the regulatory interactions is sufficient to define its activity. However, this is not always the case ( Jimenez, Cotterell, Munteanu, & Sharpe, 2017). This is particularly true as topologies become more complex. Structure alone is not sufficient to determine behavior. Depending on the starting conditions and the strength of the regulatory interactions, networks can yield different outcomes. This leads to the concept of multifunctional motifs and is where a dynamical systems approach to GRNs becomes instrumental in understanding function. For the neural tube and the gap genes in the Drosophila blastoderm an interesting example is a circuit that has come to be known as the AC-DC motif (Panovska-Griffiths, Page, & Briscoe, 2013). Composed of two well-characterized topologies, a cross repressive interaction, which would usually lead to bistability, and a repressilator (Elowitz & Leibier, 2000), which displays oscillatory behavior, this circuit is capable of

254

M. Joaquina Delás and James Briscoe

displaying both bistability and oscillatory behavior, depending on the relative strength of the regulatory interactions. Theoretical work has shown these two behaviors are possible within the same parameter space, for example with the same regulatory interaction strength but at different signaling levels (Perez-Carrasco et al., 2018) (Fig. 2D). The AC-DC motif was first noted in the ventral neural tube, where the cross repressive interaction is present between Pax6 and Nkx2.2, combined with the additional repression of Olig2 by Nkx2.2, and Pax6 by Olig2 (Balaskas et al., 2012; Panovska-Griffiths et al., 2013). In this system, the interactions lead to a multistable behavior. In the gap gene network, however, it has also been proposed that three overlapping AC-DC circuits control anterior-posterior patterning, two of them displaying multistability while the third would follow an oscillatory behavior (Verd, Monk, & Jaeger, 2019). A more unbiased theoretical survey of circuits that can perform multiple behaviors further revealed that multi-functional circuits do not need to arise from a composite of two known topologies. These “emergent class” multifunctional circuits cannot be decomposed into separate functional modules and their behavior is very hard to intuitively predict or understand by simply looking at the topology ( Jimenez et al., 2017). This underscores how the same set of genetic interactions can lead to complex behavior depending on the starting conditions and the strength of the interactions. In developmental biology, this can result in different behaviors of the same network in different environmental contexts (different timepoints or location in the embryo), and is especially relevant in an evolutionary context, as changes in the interaction strength (i.e., TF binding site mutations) or the initial conditions could potentially lead to markedly different behavior of the circuit, rather than a subtle change in the levels of a molecule, thus driving evolution of new cell types.

5. What can the GRN do for you? Understanding the structure and dynamics of a GRN can help explain counterintuitive behaviors and formulate testable predictions. Formulating the interactions as a network allows us to focus on the system as a whole, and the application of theoretical approaches derived from other disciplines, such as dynamical systems helps in this process. Modeling the behavior of a network is a powerful approach to extract principles, explain unforeseen phenotypes, and make predictions that can be compared to experimental

Repressive interactions in gene regulatory networks

255

outcomes. Combined experimental and theoretical studies in the neural tube have provided several examples that illustrate some of these aspects. These include how the GRN confers robustness to signaling changes and precision in the choice of cell fate leading to sharp boundaries, and memory of the signaling input that ensures maintenance of the patterned tissue after the signal ceases. One such scenario was the observation that embryos lacking Gli3, which carries out most of the repressive Gli function (Hui & Angers, 2011) displays little if any defects in the spatial patterning of the ventral neural tube, whereby pMN and p3 cells still acquire their identities at the same distance from the source as in wild type (Persson et al., 2002). This robustness to signaling changes was unexpected as, in the absence of repressor, the Shh pathway results in much larger transcriptional activation effects. This was confirmed experimentally with a reporter for Shh signaling activity (Balaskas et al., 2012). However, the increased signaling did not result in a change in the pattern of cell fates (Fig. 4A). The simplest network that explains the formation of these ventral cell types, including Gli inputs, Nkx2.2, Olig2 and Pax6 (Balaskas et al., 2012), was modified to include an increase in the signaling input. This also predicted that, if the signal increase was only temporary, there would be no effect in the cell type specification (Fig. 4A). Consistent with this, reporter activity assays indicated that signaling levels returned to wild-type levels in Gli3 mutants after their initial marked increase (Balaskas et al., 2012). Thus, the structure and dynamics of the neural tube GRN appears to confer robustness to transient alternations in the input signals. This explained why loss of Gli3 had relatively mild effects on pattern formation despite the dramatic effects on signaling, but also could be important in normal development to ensure that stochastic fluctuations in signal do not introduce noise into the neural tube pattern. This robustness is a feature of the GRNs that has been documented in other systems (Prill, Iglesias, & Levchenko, 2005). In contrast to the robust response to fluctuations in signaling, the other puzzling behavior of this system was how cells adopt their fates at a reproducible distance from the signaling source resulting in sharp boundaries. This boundary precision is surprising because the difference in signaling between either side of the boundary is expected to be small, which might be anticipated to result in errors in boundary positioning. Recent quantitative experimental measurements of wild type and mutants, combined with theoretical approaches, show that the network itself is, yet again, facilitating precision in boundary formation (Exelby et al., 2019). This was shown

256

M. Joaquina Delás and James Briscoe

Fig. 4 Properties of GRNs extracted from the ventral neural tube studies. (A) A temporary increase in the levels of signaling does not alter the final patterning outcome, as predicted in silico and observed experimentally. (B) Boundary precision encoded in the dynamics of the GRN. Alterations in the nodes or edges have unexpected phenotypes that can be modeled in silico. (C) The pattern is maintained even after the signaling inputs are lost, a property of the network known as hysteresis.

for p3-pMN boundary. In embryos mutant for Pax6, in addition to an increase in the size of the p3 domain, the distinct separation between p3 and pMN cell types is lost (Balaskas et al., 2012; Exelby et al., 2019) (Fig. 4B). In addition, deletion of a cis regulatory element that controls Olig2 expression resulted in fewer Olig2-expressing cells and to more intermixing of Olig2-expressing pMN and Nkx2.2-expressing p3 cells—a lossof-sharpness phenotype similar to that observed in Pax6 mutants (Exelby et al., 2019) (Fig. 4B). The precise molecular mechanisms driving these phenotypes are different but studying the network as a whole and modeling its

Repressive interactions in gene regulatory networks

257

dynamics pointed to a similar underlying principle for how these interactions control developmental cell fate decisions. This suggested that while a 2-node bistable motif is enough to generate a boundary, the dynamics of a 3-node network—the strength and directionality of the interactions and the resulting speed in the transitions—ensure that the boundary is sharp. These predictions can now be tested in other experimental systems that also produce sharp boundaries such as Drosophila embryo A-P patterning. A final feature of the neural tube GRN that was revealed by quantitative modeling was how gene expression domains are maintained after their initial establishment. In the neural tube, Shh signaling decreases after reaching a peak early in development, this is because cells progressively receive less Shh as the tissue grows, moving cells away from the source of the signal, and at the same time signal-adaptation mechanisms lead to a decreased in the effective intracellular response (Dessaud et al., 2007). Despite this, the spatial domains of gene expression patterns are maintained as the tissue grows. Analysis of the model of the GRN revealed that it produces hysteresis. This refers to a property of a dynamical system in which the history of the system influences its future behavior, an effect that effectively provides memory, and explains why gene expression is maintained even if the input decreases. Experimentally, this means as long as the cells have been exposed to sufficient signal concentration for enough time, they will express the appropriate neural progenitor fate even if the pathway reduces its strength (Balaskas et al., 2012; Zagorski et al., 2017). It is proposed that the network interactions, in this case through a series of cross repressive TFs, stabilize the cell fate decisions that were installed earlier by the GRN responding to gradients of patterning signals (Fig. 4C). Together, these analyses provided insight into how cross-repressive interactions in GRNs can guide cell fate decisions.

6. Molecular mechanisms for implementing the GRN cross-repressive interactions Most of the interactions in developmental GRN diagrams are initially derived from genetic gain and loss-of-function studies. At the molecular level, how these genetic interactions are implemented is not completely understood. For TFs, the initial assumption is that they act via their DNA-binding domain regulating gene expression although this is not always necessarily the case. An equivalent gene expression outcome could be observed under the same genetic perturbation conditions if the repression

258

M. Joaquina Delás and James Briscoe

is mediated at the protein or RNA level. Experimental evidence for all of these scenarios have emerged and the interplay of these mechanisms cannot be excluded.

6.1 CRE repression The repression of an otherwise activating CRE by the binding of a transcriptional repressor provides the most common mechanism. For example, Olig2 has been shown to bind to the regulatory region of Nkx2.2, and vice-versa (Kutejova et al., 2016). These cis-regulatory regions were previously shown to also be bound by Shh downstream effector Gli and panneural TF Sox2 (Oosterveen et al., 2012; Peterson et al., 2012). This raises the possibility that the binding of a repressor overrides the positive input and allows the integration of the different inputs to direct target gene expression (Fig. 5A). In the simplest case, a single CRE integrates all the inputs into a gene to dictate gene expression outcome. In reality, many genes have multiple CREs, this number being especially high for developmental genes (Osterwalder et al., 2018) and specific well-documented examples are available in Drosophila (Hong, Hendrix, & Levine, 2008; Perry, Boettiger,

Fig. 5 Molecular mechanisms for implementing repressive interactions. Examples of TF B repressing TF A by (A) directly binding a CRE, (B) interacting with A and preventing it from carrying out its activatory function or (C) inhibiting mRNA translation or promoting its degradation.

Repressive interactions in gene regulatory networks

259

Bothma, & Levine, 2010). This redundancy contributes to robust gene expression, as mutants of one of the enhancers are sensitive to other genetic or environmental perturbations. Mechanistically, however, if one of these CREs is bound by a repressor and the others are not, it raises the question of whether the transcriptional start site is ultimately integrating the inputs of the diverse CREs. Alternatively, a repressor directly binding to the TSS could exert a dominant role. Repressive TFs were traditionally divided in to short and long-range acting, based on the type of cofactor that the TF recruited (Turki-Judeh & Courey, 2012). This was a helpful way of conceptually distinguishing a mechanism where a bound TF could locally repress transcription versus the more complex scenario, whereby long-range mechanisms affecting the chromatin environment were involved. However, this turned out not to be as straightforward. For example, Nkx repressive neural TFs were shown to bind to Groucho/TLE (Muhr, Andersson, Persson, Jessell, & Ericson, 2001). While these are thought to act as a long-range repressors, it has also been shown to be able to act at short range (TurkiJudeh & Courey, 2012). Intriguingly, genome-wide binding data shows widespread interaction of multiple repressive TFs to TSS, as well as distal CREs (Kutejova et al., 2016; Nishi et al., 2015). Beyond the short vs long-range of action, once the repressor recruits further chromatin modifying cofactors to the loci, these could potentially use the 3D contacts to spread as it has been shown to occur in X chromosome inactivation (Engreitz et al., 2013), or a multitude of chromosomal interactions could dictate the ultimate fate of a loci, as is the case for olfactory receptor choice (Monahan, Horta, & Lomvardas, 2019). With increasing evidence for enhancer activation acting via phase separated hubs (Mir, Bickmore, Furlong, & Narlikar, 2019), the questions of whether a repressor’s function would be to ensure the CRE gets excluded from these locations and how this would ensure appropriate exclusion of the TSS remain opened.

6.2 Repression by protein-protein interactions Although DNA-based regulatory mechanisms are widespread, they are by no means the only way to implement cross-repressive interactions. Both Gata1 and PU1 are transcriptional activators, despite their well-documented antagonistic roles in cell fate decision in blood progenitors. Rather than acting via each other’s regulatory elements, each of these factors inhibits the ability of the other to activate its target genes via direct interactions between

260

M. Joaquina Delás and James Briscoe

the TFs (Rekhtman et al., 1999) or by preventing the ability of the competing TF to bind DNA targets (Nerlov, Querfurth, Kulessa, & Graf, 2000) (Fig. 5B). Because this would not affect the rate at which a competing factor is produced or degraded, the system additionally relies on positive feedback loops for each TF (Graf & Enver, 2009) and activation of one of the two downstream programs results in a two-step commitment process (Huang, Guo, May, & Enver, 2007). A prediction of this mechanism is that both PU1 and Gata1 would be expressed in the same cell prior to the cell fate commitment. Intriguingly, live cell imaging of endogenously tagged TFs did not detect coexpression of these two factors. Instead, the TF first detected would ultimately increase expression and lead to the corresponding cell fate commitment. These observations are consistent with a model where the myeloid-erythroid decision has already been taken and the cells are primed to differentiate into either lineage, with Pu1 and Gata1 reinforcing and stabilizing the decision (Hoppe et al., 2016). Complementary studies challenged the sensitivity of the methods and supported the initial findings that showed cells will indeed coexpress these two TFs at the progenitor stage (Palii et al., 2019). Mechanistically, these protein-protein interactions could occur on or off the DNA. A TF could prevent the antagonistic TF from binding its target, as reported for Gata1-PU1. Alternatively, a scenario can be envisioned in which either TF could recruit activating cofactors only when bound to the CRE in isolation, whereas binding of both TFs could prevent the gene activating function. While ultimately having the same effect, this configuration provides a mechanistic explanation for why a single TF could apparently activate some of the genes where it is bound while seemingly repress others.

6.3 Repression at the RNA level Repression of a TF can also be mediated at the RNA level, as has been shown to be the case for Olig2, which is targeted by mir-17 (Chen et al., 2011). Expressed in the neural tube in the same domain as Irx3, another TF that represses and is repressed by Olig2, mir-17-3p targets Olig2 30 UTR and its deletion of the mir-17–92 cluster that contains this microRNA leads to an expansion of the Olig2-expressing domain (Chen et al., 2011). The upstream regulation of this cluster has not been addressed in the context of neural tube development and it is not known whether its transcription is regulated in the same way as Irx3, including repression by Olig2.

Repressive interactions in gene regulatory networks

261

Another well-described RNA-mediated mechanism is the direct binding of bcd to the 30 UTR of cad, which blocks translation initiation. While the repressor is a TF, the resulting effect occurs at the RNA level (Dubnau & Struhl, 1996) (Fig. 5C). RNA-based regulation is likely to be more pervasive that we are currently aware, both affecting translation and transcription (Xiao et al., 2019), however dissection these regulatory interactions is still a challenging task. Although any of the abovementioned implementations of a crossrepressive interaction will achieve the same result, their dynamics can be very different. For example, while cis-regulatory element repression will not have any effect on the already produced mRNAs, a microRNAmediated mechanism would result in mRNA degradation and reduced translation. A combination of both, as suggested for the Olig2-Irx3 TF pair, could increase the speed of the transition and therefore lock the cell fate decisions earlier than either mechanism alone. This is again where modeling the dynamics of the system could provide insight, by suggesting whether either CRE-mediated, RNA-mediated or both repressions are at play. In each of these scenarios, predictions of the outcome upon perturbation can be made and compared to experimental results, which could help identify the contribution of each mechanism to the regulatory interaction. This type of interplay between theoretical and experimental approaches can maximize our understanding of the system.

7. Conclusions The gene expression patterns that dictate cell function and lineage are established via networks of regulatory interactions. Many of these regulatory factors are repressive TFs, which can effectively execute either binary choices between two cell states or, by combining multiple interactions, underlie the interpretation of graded signals. The ventral neural tube serves as an excellent example of how Shh promotes multiple alternative fates that are formed at different distances from the signal source. Information processing at the level of the GRN largely relies on cross-repressive TFs that ensure only the position appropriate gene expression program is implemented. As shown in this system, beyond simply repressing other TFs, this network can incorporate a second layer of regulation, whereby each TF also represses the downstream effectors of all alternative programs. These networks can be complex, and they are difficult to intuitively understand to predict their behaviors. Knowing the structure of a network is often insufficient. Some circuits, such as the AC-DC motif and other

262

M. Joaquina Delás and James Briscoe

multistable motifs, display different behaviors depending on initial conditions or strength of the interaction. This is where mathematical modeling can greatly aid in understanding the functionality of a network. Additionally, it can reveal testable predictions that would be harder to conceptualize based solely on experimental data and intuitive reasoning. By taking a step back from the individual components, network modeling helps extract principles that are applicable across organisms and cell types. The interaction of different activating and repressive TFs with CREs will ultimately result in gene regulation outcomes. How this is regulated at the mechanistic level in the context of multiple TFs remains poorly understood. Activating TFs seem to promote multiple cell fates while cross-repressive interactions ensure only the appropriate gene expression program is established. This logically implies an overriding role for the repressors. How this is implemented molecularly at the level of the CREs, especially when multiple CREs are at play, is still work in progress. Dissecting the specific components and interactions of a GRN is time-consuming and has limited the number of models to which we have access. New technologies that allow for rapid gene perturbations at single cell resolution now allow us to tackle cell types with low cell numbers or embryonic systems that are historically less well characterized. Simultaneously, chromatin accessibility techniques can also greatly facilitate identification of CREs, which can also be now perturbed without the need of genetic knockouts. Extending the GRN framework to more developmental cell fate decisions and exploiting the synergy with modeling approaches will enable us to identify general principles of the design rules that dictate embryonic patterning and differentiation.

8. Acknowledgments We are grateful to Andreas Sagner and Teresa Rayon for comments on this manuscript. Work in the JB lab is supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK, the UK Medical Research Council and Wellcome Trust (all under FC001051) and the European Research Council under European Union (EU) Horizon 2020 research and innovation program grant 742138.

References Alexandre, C., Baena-Lopez, A., & Vincent, J.-P. (2014). Patterning and growth control by membrane-tethered wingless. Nature, 505, 180–185. Alon, U. (2007). Network motifs: Theory and experimental approaches. Nature Reviews. Genetics, 8, 450–461.

Repressive interactions in gene regulatory networks

263

Andrews, M. G., Del Castillo, L. M., Ochoa-Bolton, E., Yamauchi, K., Smogorzewski, J., & Butler, S. J. (2017). BMPs direct sensory interneuron identity in the developing spinal cord using signal-specific not morphogenic activities. eLife, 6, 383. Balaskas, N., Ribeiro, A., Panovska, J., Dessaud, E., Sasai, N., Page, K. M., et al. (2012). Gene regulatory logic for Readingthe sonic hedgehog Signaling gradient in the vertebrate neural tube. Cell, 148, 273–284. Briscoe, J., Pierani, A., Jessell, T. M., & Ericson, J. (2000). A homeodomain protein code specifies progenitor cell identity and neuronal fate in the ventral neural tube. Cell, 101, 435–445. Briscoe, J., Sussel, L., Serup, P., Hartigan-O’Connor, D., Jessell, T. M., Rubenstein, J. L., et al. (1999). Homeobox gene Nkx2.2 and specification of neuronal identity by graded sonic hedgehog signalling. Nature, 398, 622–627. Bylund, M., Andersson, E., Novitch, B. G., & Muhr, J. (2003). Vertebrate neurogenesis is counteracted by Sox1-3 activity. Nature Neuroscience, 6, 1162–1168. Chen, J.-A., Huang, Y.-P., Mazzoni, E. O., Tan, G. C., Zavadil, J., & Wichterle, H. (2011). Mir-17-3p controls spinal neural progenitor patterning by regulating Olig2/Irx3 crossrepressive loop. Neuron, 69, 721–735. Cohen, M., Page, K. M., Perez-Carrasco, R., Barnes, C. P., & Briscoe, J. (2014). A theoretical framework for the regulation of Shh morphogen-controlled gene expression. Development, 141, 3868–3878. Davidson, E. H. (2010). Emerging properties of animal gene regulatory networks. Nature, 468, 911–920. Delile, J., Rayon, T., Melchionda, M., Edwards, A., Briscoe, J., & Sagner, A. (2019). Single cell transcriptomics reveals spatial and temporal dynamics of gene expression in the developing mouse spinal cord. Development, 146(12), dev.173807. Dessaud, E., Ribes, V., Balaskas, N., Yang, L. L., Pierani, A., Kicheva, A., et al. (2010). Dynamic assignment and maintenance of positional identity in the ventral neural tube by the Morphogen sonic hedgehog. PLoS Biology, 8, e1000382–14. Dessaud, E., Yang, L. L., Hill, K., Cox, B., Ulloa, F., Ribeiro, A., et al. (2007). Interpretation of the sonic hedgehog morphogen gradient by a temporal adaptation mechanism. Nature, 450, 717–720. Driever, W., Thoma, G., & N€ usslein-Volhard, C. (1989). Determination of spatial domains of zygotic gene expression in the Drosophila embryo by the affinity of binding sites for the bicoid morphogen. Nature, 340, 363–367. Dubnau, J., & Struhl, G. (1996). RNA recognition and translational regulation by a homeodomain protein. Nature, 379, 694–699. Duval, N., Vaslin, C., Barata, T. C., Frarma, Y., Contremoulins, V., Baudin, X., et al. (2019). BMP4 patterns Smad activity and generates stereotyped cell fate organization in spinal organoids. Development, 146, dev175430. Elowitz, M. B., & Leibier, S. (2000). A synthetic oscillatory network of transcriptional regulators. Nature, 403, 335–338. Engreitz, J. M., Pandya-Jones, A., McDonel, P., Shishkin, A., Sirokman, K., Surka, C., et al. (2013). The Xist lncRNA exploits three-dimensional genome architecture to spread across the X chromosome. Science, 341, 1237973. Ericson, J., Rashbass, P., Schedl, A., Brenner-Morton, S., Kawakami, A., van Heyningen, V., et al. (1997). Pax6 controls progenitor cell identity and neuronal fate in response to graded Shh signaling. Cell, 90, 169–180. Exelby, K., Herrera-Delgado, E., Perez, L. G., Perez-Carrasco, R., Sagner, A., Metzis, V., et al. (2019). Precision of tissue patterning is controlled by dynamical properties of gene regulatory networks. bioRxiv. https://doi.org/10.1101/721043. Goentoro, L., Shoval, O., Kirschner, M. W., & Alon, U. (2009). The incoherent feedforward loop can provide fold-change detection in gene regulation. Molecular Cell, 36, 894–899.

264

M. Joaquina Delás and James Briscoe

Graf, T., & Enver, T. (2009). Forcing cells to change lineages. Nature, 462, 587–594. Graham, V., Khudyakov, J., Ellis, P., & Pevny, L. (2003). SOX2 functions to maintain neural progenitor identity. Neuron, 39, 749–765. Green, J. B., Howes, G., Symes, K., Cooke, J., & Smith, J. C. (1990). The biological effects of XTC-MIF: Quantitative comparison with Xenopus bFGF. Development, 108, 173–183. Gupta, S., Sivalingam, D., Hain, S., Makkar, C., Sosa, E., Clark, A., et al. (2018). Deriving dorsal spinal sensory interneurons from human pluripotent stem cells. Stem Cell Reports, 10, 390–405. Hong, J.-W., Hendrix, D. A., & Levine, M. S. (2008). Shadow enhancers as a source of evolutionary novelty. Science, 321, 1314. Hoppe, P. S., Schwarzfischer, M., Loeffler, D., Kokkaliaris, K. D., Hilsenbeck, O., Moritz, N., et al. (2016). Early myeloid lineage choice is not initiated by random PU.1 to GATA1 protein ratios. Nature, 535, 299–302. Huang, S., Guo, Y.-P., May, G., & Enver, T. (2007). Bifurcation dynamics in lineagecommitment in bipotent progenitor cells. Developmental Biology, 305, 695–713. Hui, C.-C., & Angers, S. (2011). Gli proteins in development and disease. Annual Review of Cell and Developmental Biology, 27, 513–537. Jaeger, J. (2011). The gap gene network. Cellular and Molecular Life Sciences, 68, 243–274. Jeong, J. (2004). Growth and pattern of the mammalian neural tube are governed by partially overlapping feedback activities of the hedgehog antagonists patched 1 and Hhip1. Development, 132, 143–154. Jimenez, A., Cotterell, J., Munteanu, A., & Sharpe, J. (2017). A spectrum of modularity in multi-functional gene circuits. Molecular Systems Biology, 13, 925. Kamachi, Y., & Kondoh, H. (2013). Sox proteins: Regulators of cell fate specification and differentiation. Development, 140, 4129–4144. Kulessa, H., Frampton, J., & Graf, T. (1995). GATA-1 reprograms avian myelomonocytic cell lines into eosinophils, thromboblasts, and erythroblasts. Genes & Development, 9, 1250–1262. Kutejova, E., Sasai, N., Shah, A., Gouti, M., & Briscoe, J. (2016). Neural progenitors adopt specific identities by directly repressing all alternative progenitor transcriptional programs. Developmental Cell, 36, 639–653. Lai, H. C., Seal, R. P., & Johnson, J. E. (2016). Making sense out of spinal cord somatosensory development. Development, 143, 3434–3448. Lee, J., Platt, K. A., Censullo, P., & Ruiz i Altaba, A. (1997). Gli1 is a target of sonic hedgehog that induces ventral neural tube development. Development, 124, 2537–2552. Lek, M., Dias, J. M., Marklund, U., Uhde, C. W., Kurdija, S., Lei, Q., et al. (2010). A homeodomain feedback circuit underlies step-function interpretation of a Shh morphogen gradient during ventral neural patterning. Development, 137, 4051–4060. Liem, K. F., Jessell, T. M., & Briscoe, J. (2000). Regulation of the neural patterning activity of sonic hedgehog by secreted BMP inhibitors expressed by notochord and somites. Development, 127, 4855–4866. Liem, K. F., Tremml, G., & Jessell, T. M. (1997). A role for the roof plate and its resident TGFβ-related proteins in neuronal patterning in the dorsal spinal cord. Cell, 91, 127–138. Litingtung, Y., & Chiang, C. (2000). Specification of ventral neuron types is mediated by an antagonistic interaction between Shh and Gli3. Nature Neuroscience, 3, 979–985. Mir, M., Bickmore, W., Furlong, E. E. M., & Narlikar, G. (2019). Chromatin topology, condensates and gene regulation: Shifting paradigms or just a phase? Development, 146(19), dev.182766. Monahan, K., Horta, A., & Lomvardas, S. (2019). LHX2- and LDB1-mediated trans interactions regulate olfactory receptor choice. Nature, 565, 448–453.

Repressive interactions in gene regulatory networks

265

Muhr, J., Andersson, E., Persson, M., Jessell, T. M., & Ericson, J. (2001). Groucho-mediated transcriptional repression establishes progenitor cell pattern and neuronal fate in the ventral neural tube. Cell, 104, 861–873. Nerlov, C., & Graf, T. (1998). PU.1 induces myeloid lineage commitment in multipotent hematopoietic progenitors. Genes & Development, 12, 2403–2412. Nerlov, C., Querfurth, E., Kulessa, H., & Graf, T. (2000). GATA-1 interacts with the myeloid PU.1 transcription factor and represses PU.1-dependent transcription. Blood, 95, 2543–2551. Nishi, Y., Zhang, X., Jeong, J., Peterson, K. A., Vedenko, A., Bulyk, M. L., et al. (2015). A direct fate exclusion mechanism by sonic hedgehog-regulated transcriptional repressors. Development, 142, 3286–3293. Niwa, H. (2018). The principles that govern transcription factor network functions in stem cells. Development, 145. Novitch, B. G., Chen, A. I., & Jessell, T. M. (2001). Coordinate regulation of motor neuron subtype identity and pan-neuronal properties by the bHLH repressor Olig2. Neuron, 31, 773–789. Novitch, B. G., Wichterle, H., Jessell, T. M., & Sockanathan, S. (2003). A requirement for retinoic acid-mediated transcriptional activation in ventral neural patterning and motor neuron specification. Neuron, 40, 81–95. Ochoa-Espinosa, A., Yucel, G., Kaplan, L., Pare, A., Pura, N., Oberstein, A., et al. (2005). The role of binding site cluster strength in Bicoid-dependent patterning in Drosophila. Proceedings of the National Academy of Sciences of the United States of America, 102, 4960–4965. Oliveri, P., Tu, Q., & Davidson, E. H. (2008). Global regulatory logic for specification of an embryonic cell lineage. Proceedings of the National Academy of Sciences, 105, 5955–5962. Oosterveen, T., Kurdija, S., Alekseenko, Z., Uhde, C. W., Bergsland, M., Sandberg, M., et al. (2012). Mechanistic differences in the transcriptional interpretation of local and long-range Shh Morphogen Signaling. Developmental Cell, 23, 1006–1019. Orkin, S. H. (2000). Diversification of haematopoietic stem cells to specific lineages. Nature Reviews Genetics, 1, 57–64. Osterwalder, M., Barozzi, I., Tissie`res, V., Fukuda-Yuzawa, Y., Mannion, B. J., Afzal, S. Y., et al. (2018). Enhancer redundancy provides phenotypic robustness in mammalian development. Nature, 554, 239–243. Palii, C. G., Cheng, Q., Gillespie, M. A., Shannon, P., Mazurczyk, M., Napolitani, G., et al. (2019). Single-cell proteomics reveal that quantitative changes in co-expressed lineagespecific transcription factors determine cell fate. Cell Stem Cell, 24(5), 812–820. Pan, Y., Bai, C. B., Joyner, A. L., & Wang, B. (2006). Sonic hedgehog signaling regulates Gli2 transcriptional activity by suppressing its processing and degradation. Molecular and Cellular Biology, 26, 3365–3377. Panovska-Griffiths, J., Page, K. M., & Briscoe, J. (2013). A gene regulatory motif that generates oscillatory or multiway switch outputs. Journal of the Royal Society, Interface, 10, 20120826. Perez-Carrasco, R., Barnes, C. P., Schaerli, Y., Isalan, M., Briscoe, J., & Page, K. M. (2018). Combining a toggle switch and a Repressilator within the AC-DC circuit generates distinct dynamical Behaviors. Cell Systems, 6, 521–530.e523. Perry, M. W., Boettiger, A. N., Bothma, J. P., & Levine, M. (2010). Shadow enhancers foster robustness of Drosophila gastrulation. Curbio, 20, 1562–1567. Persson, M., Stamataki, D., Welscher, P. T., Andersson, E., B€ ose, J., R€ uther, U., et al. (2002). Dorsal-ventral patterning of the spinal cord requires Gli3 transcriptional repressor activity. Genes & Development, 16, 2865–2878. Peter, I. S., & Davidson, E. H. (2011). A gene regulatory network controlling the embryonic specification of endoderm. Nature, 474, 635–639.

266

M. Joaquina Delás and James Briscoe

Peterson, K. A., Nishi, Y., Ma, W., Vedenko, A., Shokri, L., Zhang, X., et al. (2012). Neural-specific Sox2 input and differential Gli-binding affinity provide context and positional information in Shh-directed neural patterning. Genes & Development, 26, 2802–2816. Pierani, A., Brenner-Morton, S., Chiang, C., & Jessell, T. M. (1999). A sonic hedgehogindependent, retinoid-activated pathway of neurogenesis in the ventral spinal cord. Cell, 97, 903–915. Prill, R. J., Iglesias, P. A., & Levchenko, A. (2005). Dynamic properties of network motifs contribute to biological network organization. PLoS Biology, 3, e343–e1892. Rekhtman, N., Radparvar, F., Evans, T., & Skoultchi, A. I. (1999). Direct interaction of hematopoietic transcription factors PU.1 and GATA- 1: Functional antagonism in erythroid cells. Genes & Development, 13, 1398–1411. Rogers, K. W., & Schier, A. F. (2011). Morphogen gradients: From generation to interpretation. Annual Review of Cell and Developmental Biology, 27, 377–407. Sagner, A., & Briscoe, J. (2019). Establishing neuronal diversity in the spinal cord: A time and a place. Development, 146(22), dev182154. Sagner, A., Gaber, Z. B., Delile, J., Kong, J. H., Rousso, D. L., Pearson, C. A., et al. (2018). Olig2 and Hes regulatory dynamics during motor neuron differentiation revealed by single cell transcriptomics. PLoS Biology, 16, e2003127–e2003137. Sherman, M. S., & Cohen, B. A. (2012). Thermodynamic state ensemble models of cisregulation. PLoS Computational Biology, 8, e1002407. Stamataki, D., Ulloa, F., Tsoni, S. V., Mynett, A., & Briscoe, J. (2005). A gradient of Gli activity mediates graded sonic hedgehog signaling in the neural tube. Genes & Development, 19, 626–641. Struhl, G., Struhl, K., & Macdonald, P. M. (1989). The gradient morphogen bicoid is a concentration-dependent transcriptional activator. Cell, 57, 1259–1273. Tozer, S., Le Dreau, G., Martı´, E., & Briscoe, J. (2013). Temporal control of BMP signalling determines neuronal subtype identity in the dorsal neural tube. Development, 140, 1467–1474. Turki-Judeh, W., & Courey, A. J. (2012). Groucho. A corepressor with instructive roles in development. Current Topics in Developmental Biology, 98, 65–96. Verd, B., Monk, N. A., & Jaeger, J. (2019). Modularity, criticality, and evolvability of a developmental gene regulatory network. eLife, 8, 450. Wang, B., Fallon, J. F., & Beachy, P. A. (2000). Hedgehog-regulated processing of Gli3 produces an anterior/posterior repressor gradient in the developing vertebrate limb. Cell, 100, 423–434. Xiao, R., Chen, J.-Y., Liang, Z., Luo, D., Chen, G., Lu, Z. J., et al. (2019). Pervasive chromatin-RNA binding protein interactions enable RNA-based regulation of transcription. Cell, 178, 107–121.e118. Zagorski, M., Tabata, Y., Brandenberg, N., Lutolf, M. P., Tkacik, G., Bollenbach, T., et al. (2017). Decoding of position in the developing neural tube from antiparallel morphogen gradients. Science, 356, 1379–1383. Zecca, M., Basler, K., & Struhl, G. (1996). Direct and long-range action of a wingless morphogen gradient. Cell, 87, 833–844. Zhou, Q., & Anderson, D. J. (2002). The bHLH transcription factors OLIG2 and OLIG1 couple neuronal and glial subtype specification. Cell, 109, 61–73.

CHAPTER NINE

The function of architecture and logic in developmental gene regulatory networks Isabelle S. Peter* Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA, United States *Corresponding author: e-mail address: [email protected]

Contents 1. Introduction 2. The structure of developmental GRNs in theory and experiment 3. Logic processing in developmental GRNs 4. Architecture of circuit modules that are prevalent in developmental GRNs 5. The connection between circuit architecture and developmental function 6. Organization of circuit modules within developmental GRNs 7. Evolution of network architecture 8. Conclusion Acknowledgments References

268 270 274 276 280 287 289 290 290 290

Abstract An important contribution of systems biology is the insight that biological systems depend on the function of molecular interactions and not just on individual molecules. System level mechanisms are particularly important in the development of animals and plants which depends not just on transcription factors and signaling molecules, but also on regulatory circuits and gene regulatory networks (GRNs). However, since GRNs consist of transcription factors, it can be challenging to assess the function of regulatory circuits independently of the function of regulatory factors. The comparison of different GRNs offers a way to do so and leads to several observations. First, similar regulatory circuits operate in various developmental contexts and in different species, and frequently, these circuits are associated with similar developmental functions. Second, given regulatory circuits are often used at particular positions within the GRN hierarchy. Third, in some GRNs, regulatory circuits are organized in a particular order in respect to each other. And fourth, the evolution of GRNs occurs not just by co-option of regulatory genes but also by rewiring of regulatory linkages between conserved regulatory genes, indicating that the organization of interactions is important. Thus, even though in most

Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.04.001

#

2020 Elsevier Inc. All rights reserved.

267

268

Isabelle S. Peter

instances the function of regulatory circuits remains to be discovered, it becomes evident that the architecture and logic of GRNs are functionally important for the control of genome activity and for the specification of the body plan.

1. Introduction Gene regulatory networks (GRNs) control gene expression throughout development and adult life in animals and plants (Peter & Davidson, 2015). By regulating gene expression, GRNs control a variety of developmental and cellular functions. One of the most important functions of GRNs is to control the organization of the body plan, the arrangement of body parts and cell fates within the overall context of an organism. Decades of genetic mutation experiments demonstrate that dramatic phenotypic changes of the body plan, such as gain or loss of segments or body parts, or phenotypic transformations of body parts, are frequently caused by mutations in regulatory genes encoding transcription factors (Halder, Callaerts, & Gehring, 1995; Lewis, 1978; Nusslein-Volhard & Wieschaus, 1980; Quiring, Walldorf, Kloter, & Gehring, 1994). These observations indicated early on that transcriptional gene regulation is particularly important for the specification of the animal body plan (Levine, 2010; Spitz & Furlong, 2012). But despite the importance of transcription factors and intercellular signaling molecules in regulating gene expression and defining the animal body plan, these molecules are often deployed in many different developmental contexts, and individual regulatory factors are therefore not sufficient to explain the specificity of developmental mechanisms. Transcription factors and signaling molecules function as part of GRNs in which they control the expression of regulatory genes and all other genes by means of regulatory interactions (Davidson, 2006; Davidson et al., 2002; Peter & Davidson, 2015). GRNs are organized as networks because each gene is regulated by many transcription factors and each transcription factor regulates many genes. Transcription factors therefore represent the nodes of these networks and are essential for regulating gene expression in development. However, the importance of higher level circuit architecture in controlling development is also becoming increasingly clear, although its functional contribution is less well understood. Ultimately, development is controlled at all levels of network organization, from nodes to subcircuits to large GRNs, and one of the goals will be to understand how each of these levels contribute to the overall developmental process (Fig. 1).

Architecture and logic in developmental GRNs

269

Fig. 1 Levels of organization of the genomic control system. Transcription factors (TFs) represent the individual nodes of the networks. Modular subcircuits consist of small groups of transcription factors with a particular design of regulatory interactions. GRNs include all functional regulatory interactions between TFs and subcircuits that control developmental processes. Genomic programs for development integrate regulatory information at the level of transcription factors, subcircuit modules, and large GRNs and all levels should be considered in computational models reproducing developmental gene expression. Evolution of the body plan occurs by changes in the developmental program at all levels of network organization.

Only few GRNs have been characterized experimentally to reveal both the regulatory genes at their nodes as well as the regulatory interactions between them. Where GRNs have been analyzed, they provide an opportunity to consider the system level information that is encoded in GRNs. What can we learn from analyzing GRNs? The direct comparison of mechanisms across different developmental contexts is often challenging because each developmental process is controlled by unique combinations of transcription factors and signaling molecules. On the other hand, development includes many processes that occur in similar ways in different parts of an embryo and even in different species. These processes include for example cell fate specification, patterning, cell migration, cell fate exclusion, proliferation, cell type differentiation, and many more. By comparing GRNs that control similar processes in different developmental contexts, it becomes possible to abstract from individual factors and signals, which are mostly context specific, and to reveal instead the function of network circuitry that connects these molecules. The functional analysis of GRNs that consist of different sets of transcription factors and signaling molecules thus allows us to distinguish between the molecular function of regulatory molecules and the system level function of regulatory circuits in development.

270

Isabelle S. Peter

2. The structure of developmental GRNs in theory and experiment Before discussing how GRNs are organized and how they control developmental mechanisms, it is worth addressing how the structure of genomic instructions was envisioned before the molecular details of developmental gene regulation were discovered. The idea that genomes must contain instructions that define the animal body plan is not new and was formulated in the following way by E.B. Wilson in 1924: “How are the operations of development so coordinated as to give rise to a definitely ordered system? It is our scientific habit of thought to regard the operation of any specific system as determined primarily by its specific physical-chemical composition …. This mechanistic assumption implies some specific structure or material configuration in the system, and since the organization of the egg is hereditary, the structure or configuration must be preserved by cell division without loss of its specific character …” (Wilson, 1924). A major step toward solving this configuration came from a theoretical model of gene regulation that was formulated about 50 years ago. Roy Britten and Eric Davidson proposed that a key mechanism controlling development must rely on the regulation of gene expression (Britten & Davidson, 1969). “Cell differentiation is based almost certainly on the regulation of gene activity, so that for each state of differentiation a certain set of genes is active in transcription and other genes are inactive.” The model assumed that transcriptional regulation is most important for regulating cell differentiation, since most of the genome is inactive and since different types of cells express different RNAs. A further prediction was that gene regulation has to be sequence specific. At the time, little was known about the molecular biology of gene regulation and not surprisingly, the molecular details in the gene regulatory model would be different from the structure of regulatory systems that were experimentally discovered decades later (Davidson et al., 2002). According to this model, developmental gene expression was thought to be regulated by dedicated signals (Britten & Davidson, 1969). Each signal could regulate multiple target genes, but the network was thought to be a one-step instruction where the expression of a differentiation gene battery would occur wherever the instructive signal is present. One can imagine that these signals might instruct cell type-specific gene expression or the expression of genes associated with cell proliferation, epithelial to mesenchymal

Architecture and logic in developmental GRNs

271

transition, or any other cellular behavior that is not exclusive to any given cell type. Curiously, intercellular signals turn out to be important for regulating developmental gene expression, but they are only one part of the overall regulatory system. In the 1969 model, sequence-specific gene regulation is mediated by activator RNAs, although the authors state that gene regulation could also be mediated by proteins without changing the principle predictions. The model for gene regulation has the following structure. A sensor DNA sequence containing binding sites for pattern forming agents, such as for example hormones or other signals, drive the expression of activator RNAs. An activator RNA which has regulatory function in turn binds to recognition sites in responder DNA (now referred to as cis-regulatory modules) that controls the expression of producer genes. These producer genes encode non-regulatory molecules that contribute to the structure and function of the cell. In modern language, these producer genes correspond to differentiation genes. In this early model of gene regulation, each activator RNA can regulate multiple producer genes and each producer gene is regulated by multiple activator RNAs. Thus each patterning agent can regulate a multitude of target genes, a gene battery that executes tissue-specific molecular functions. And each producer gene can be expressed in multiple patterns, as it can be regulated by specific activator RNAs in each context. The conceptual predictions of this model were fundamental for driving the efforts to obtain experimental proof. Once it became possible to obtain experimental evidence, several of the predictions of the Britten-Davidson model were confirmed. First, transcriptional gene regulation is indeed crucial for developmental cell fate specification (Betancur, Bronner-Fraser, & Sauka-Spengler, 2010b; Chen, Xu, Mei, Yu, & Small, 2012; Christiaen et al., 2008; Davidson et al., 2002; Levine & Tjian, 2003; Oliveri, Tu, & Davidson, 2008; Peter & Davidson, 2011b; Stathopoulos, Van Drenth, Erives, Markstein, & Levine, 2002). A multitude of gene regulatory mechanisms have been discovered in the decades since the publication of the Britten-Davidson model, all of them affecting in some ways either the level or the activity of the ultimate gene product. But the mechanisms of gene regulation that control the distinction of cell types rely primarily on transcriptional regulation. Second, transcription factors regulate gene expression by sequence-specific interactions with DNA binding site motifs (Badis et al., 2009; Weirauch et al., 2014). Third, in multicellular organisms, gene expression is regulated by multiple transcription factors binding to their

272

Isabelle S. Peter

recognition sites within cis-regulatory modules (CRMs), similar to the predicted regulation by multiple activator RNAs (Levine, 2010). Fourth, intercellular signaling interactions that are mediated by signaling molecules are often crucial for activating gene expression in specific tissues, although instead of directly binding to DNA, signaling ligands activate signal transduction pathways that ultimately regulate the activity of transcription factors (Barolo & Posakony, 2002). In this way, signaling interactions can simultaneously activate hundreds of target genes, and many of these target genes are also expressed in other developmental contexts under the control of a different signaling ligand, just as predicted by the model. Although the Britten-Davidson model offers explanations for how specific sets of differentiation genes are co-expressed in given developmental contexts, it does not address how the initial signals are controlled. In biological reality, this problem is solved in an elegant way, which is that developmental signals, transcriptional regulators, and differentiation genes are all regulated by the same transcriptional control machinery. This means that transcriptional regulators are both cause and effect in the control of developmental gene expression. Due to the combinatorial control of gene expression, the transcriptional control system is organized as a network where multiple transcription factors regulate the expression of each gene and each transcription factor controls the expression of multiple target genes. When modeled as a network where regulatory genes are represented as nodes, each node will receive multiple inputs representing the regulatory interactions controlling its expression, and produce one output with linkages to multiple target genes. As accurate as some of these predictions are that were made 50 years ago, it is particularly interesting to explore what is missing from the original model of a GRN for development. Interestingly, several missing features turn out to be crucial for the establishment of patterns and the differential specification of cell fates. First, gene regulation includes repression and not just activation of gene expression (Barolo & Posakony, 2002; Kraut & Levine, 1991; Ozdemir, Ma, White, & Stathopoulos, 2014). Transcriptional repression plays an important role in pattern formation and the exclusion of alternative cell fates. Second, since gene regulation occurs by combinations of transcription factors, the regulatory logic by which these transcription factors function together in the regulation of particular genes becomes important. In the model, the regulatory inputs were assumed to operate independently, with each input being sufficient for gene activation, while in reality, transcription factors often operate in AND logic

Architecture and logic in developmental GRNs

273

together with other transcription factors (Istrail & Davidson, 2005; Peter, Faure, & Davidson, 2012). This means that a given transcription factor A may only activate the expression of target gene X if transcription factor B is also present. The consequence of regulatory AND logic is that gene X will only be expressed in cells where both transcription factors are present (Fig. 2). This regulatory feature is particularly important since transcription factor A can also be expressed in a different cell type and activate gene Y together with transcription factor C, without affecting the expression of gene X. In GRNs, the same transcription factors and signaling ligands can therefore be re-used repeatedly in entirely different developmental contexts and yet still control context-specific gene expression. Third, the probably most important and interesting feature of transcriptional gene regulation is that it operates as a gene regulatory network system and not just

Fig. 2 Effects of regulatory logic on spatial gene expression. Transcription factors regulate gene expression by activation or repression, in AND logic or OR logic. Gene A is regulated by three transcription factor inputs, Ι1, Ι2, and Ι3, that are expressed in three spatial patterns as shown in schematic embryos on top. Truth tables show the state of expression of gene A as expressed (1) or not expressed (0) depending on the presence or absence of regulatory inputs and the regulatory logic by which they function. The various spatial expression patterns of gene A in response to the three inputs is shown in schematic embryos.

274

Isabelle S. Peter

a linear hierarchy. Thus the genes encoding transcription factors and signaling molecules are themselves controlled by the same regulatory mechanisms. Instead of passively responding to a patterning agent, gene expression patterns are established through cross-regulation and feedback regulation between regulatory genes encoding transcription factors and signaling molecules. These last two properties of gene regulation, regulatory logic and network architecture, are the reason why transcription factors and signaling molecules can execute context-specific regulatory functions in many different developmental processes. Transcription factors and signaling molecules are very different from enzymes that operate specific reactions with specific substrates and products. The molecular function of transcription factors is to bind small DNA motifs in the genome and, together with other transcription factors, control gene expression. The developmental function of a transcription factor is not an intrinsic function of this molecule but a function of the cis-regulatory modules (CRMs) to which this transcription factor binds, and a function of the regulatory state, the set of transcription factors that are co-expressed in the same nucleus (Peter, 2017). Thus the ectopic expression of mouse Pax6 in Drosophila contributes to the development of Drosophila compound eyes despite the fact that in the endogenous developmental context, this transcription factor contributes to the formation of mouse eyes (Halder et al., 1995). In order to characterize the distinct context-specific functions of transcription factors and signaling molecules, it is therefore important to consider regulatory functions in the context of regulatory logic and circuit architecture.

3. Logic processing in developmental GRNs In the theoretical model discussed above, each individual regulatory RNA is capable of activating gene expression, and the same gene can be regulated by different regulators that operate independently of one another. Where single transcription factors are sufficient to regulate gene transcription, transcriptional regulation occurs mostly by additive OR logic such that different transcription factors can alternatively regulate gene expression. This is indeed true for many bacterial transcription factors that are sufficient to regulate gene expression when bound to a promoter sequence. However, transcriptional regulation in multicellular animals depends on the simultaneous function of multiple transcription factors and co-factors (Levine, 2010).

Architecture and logic in developmental GRNs

275

During animal development, gene expression is regulated by CRMs that each includes binding sites for several transcription factors, and often multiple binding sites for each factor (Levine, 2010; Spitz & Furlong, 2012). Some transcription factors that bind together to a CRM indeed function redundantly, by OR logic, such that removing one of them will not have a drastic effect on CRM function and gene expression. However, the rules by which CRMs decipher the combinatorial inputs are often more complex than that. Where detailed experiments have been conducted to dissect individual developmental CRMs, the evidence shows that different transcription factor binding sites within the same CRM are often not redundant but contribute in a non-linear way to the activity of a CRM. For example, the sparkling enhancer of the Drosophila pax2 gene that drives expression in eyes was systematically analyzed by mutation of short sequences throughout the enhancer (Swanson, Evans, & Barolo, 2010). When mutated, several short sequences of GFP plasmid (Corbo, Levine, & Zeller, 1997), incubated until the mid-gastrula ( 5.6 hpf ) (B), initial tailbud (C, C0 ), mid-tailbud I (D), late-tailbud II (E, E0 ), late tailbud III (F–G) stages (Hotta et al., 2007). (C0 ) High-magnification view of the region boxed in yellow in (C), showing intercalating notochord cells. Noticeably, intercalation at the anterior and posterior ends of the notochord is nearly complete (discussed in Section 3.2.2). (E’) High-magnification view of the notochord from an embryo at approximately the same stage as the embryo in (E). The notochord cells have a characteristic stack-of-coins arrangement. Nuclei of neighboring cells are stained with DAPI (blue). (F) Notochord of a late tailbud II embryo, displaying several intercellular lumen pockets. (F0 ) High-magnification view of the region boxed in (F), showing intercellular lumen pockets. (G) High-magnification view of the notochord of an embryo co-electroporated with the Bra>GFP and Bra>Slc26αa:: mCherry plasmids (the latter plasmid was kindly provided by Dr. Wei Deng, SARS International Centre for Marine Molecular Biology, Bergen, Norway), to highlight the apical domains of adjacent notochord cells. Scale bars: in (B, C), 40 μm; in (D, G), 10 μm. Timeline of ascidian embryonic development at 18 °C; developmental stages and timing are according to Hotta, Mitsuhara, et al. (2007). “Late gastrula” is also referred to as “neural plate.” Hpf, hours post-fertilization. (H) The approximate expression time frames of Bra and those of its target genes are symbolized by horizontal bars, and their respective lengths are matched to the developmental timeline shown above. Expression of Bra (red bar) is detected in notochord precursors beginning at the 64-cell stage and persists through notochord development (Corbo, Levine, et al., 1997). Expression time frames of early-onset Bra target genes are shown in pink; middle-onset targets

The notochord gene regulatory network in chordate evolution

335

while the increasingly large extracellular lumen pockets tilt and begin to fuse with each other (Fig. 2A). Eventually, these cavities coalesce into a single lumen that occupies the central-most region of the tail (Fig. 3C). Tubulogenesis transforms the notochord from a rod of cells into a fluid-filled tube. The notochordal sheath is now fully developed (Fig. 3D and D0 ) and ready to oppose both the expansion of the notochordal lumen on its inner surface and the pull exerted by the contractile tail muscles on its outer surface (Razy-Krajka & Stolfi, 2019). Tubulogenesis enables the notochord to function as an efficient hydrostatic skeleton, which together with the rhythmic contractions of the muscle cells located on both its sides and the yawing of the trunk allows the swimming movements of the hatched larva (Bone, 1992; Kier, 2012). The strategy employed to complete tubulogenesis and notochord stiffening in Ciona is rather unusual, and it is noteworthy that this process is divergent even among ascidian embryos, as the size and pervasiveness of the lumen vary among different species (reviewed in Smith, 2018). In the vertebrate notochord, tubulogenesis does not take place and stiffening of this structure is achieved through the extensive vacuolization of notochord cells. In Xenopus, the swelling of the individual vacuoles increases the stiffness of the notochord and generates enough pressure to drive its elongation (Adams, Keller, & Koehl, 1990; Fig. 3B and B0 ). Nevertheless, tubulogenesis is a generalized mechanism employed by different organisms to form hollow structures, and remarkably, formation of the dorsal longitudinal anastomotic vessels in zebrafish closely resembles the process of lumen formation that is observed in the Ciona notochord (Herwig et al., 2011).

are colored in light blue, and late-onset Bra target genes are colored in lilac. Schematics on the left side represent minimal notochord CRMs (pink bars) associated with either early-onset targets, which contain two or three presumably cooperative Bra binding sites (red ovals), or middle-onset CRMs (light blue bar) containing individual Bra binding sites (single red oval). Schematics at the bottom left depict how Bra might activate transcription (right angle arrow) of middle-onset transcription factor genes through individual binding sites (single red oval), whose products (yellow hexagon and green oblique bar), in turn, activate expression of late-onset target genes by binding their CRMs (Katikala et al., 2013). Gene names are abbreviated as follows, with respective gene models: thbsp, thrombospondin 3A (KH.C6.164); laminin, laminin c1 (KH.C7.167); collagen, fibrillar collagen 1 (CiFCol1, Wada, Okuyama, Satoh, & Zhang, 2006; KH.C7.633); Noto5 (KH.L153.32); ERM, ezrin-radixin-moesin (KH.C12.129); Tbx2/3 (KH.L96.87); Noto1 (KH.L20.18); Noto8, Calmodulin-like3 (KH.C11.665); Noto4/PID1 (KH.L18.30); Noto9/KTN1/RRBP1 (KH.L13.3); NFAT5 (KH.C3.133); ACL, ATP citrate lyase (KH.C3.228); b4GalT, β-1,4-galactosyltransferase (KH.C11.44).

336

Anna Di Gregorio

Fig. 3 Late stages of notochord formation in Ciona and Xenopus embryos. (A) Brightfield microphotograph of a Ciona robusta embryo, approximate stage late tailbud III (Hotta, Mitsuhara, et al., 2007) (dorsal view). Both sensory organs contain melanin (brown spots in the center of the “head”) and intercellular lumen pockets are regularly spaced between notochord cells throughout the length of the tail. Anterior is on the left side. Scale bar: 20 μm. (A0 ) High-magnification view of the region boxed in violet in (A), displaying 7 notochord cells (N) intercalated by intercellular lumen pockets (LP), in the characteristic shape of concavities. The notochord is flanked bilaterally by non-syncytial muscle (M), and the entire body is covered by epidermis (E). (B) Bright-field microphotograph of a coronal section of a Xenopus laevis tadpole (NF stage 45; Nieuwkoop & Faber, 1967), embedded in paraffin and stained by hematoxylin and eosin. Anterior is on the left side. The notochord is visible in the middle of the section as a multicellular structure with a foamy appearance. At this stage the notochord is completely vacuolated, extends cranially to the infundibulum, and is flanked bilaterally by skeletal muscle. Scale bar: 100 μm. (B0 ) High-magnification view of the region boxed in (B), displaying several notochord cells, each mostly occupied by a large vacuole (V), flanked by skeletal muscle (M). Colors have been removed and contrast has been increased, to better

The notochord gene regulatory network in chordate evolution

337

3.2 The molecular blueprint for notochord formation in Ciona 3.2.1 Notochord induction and specification At the molecular level, notochord formation is initiated by a cascade of inductive events that begin in the 16-cell embryo with the translocation of β-catenin to the nuclei of the mesendodermal precursors. In particular, as the primary notochord precursors, the A5.1 and the A5.2 blastomere pairs, concertedly divide, the nuclear localization of β-catenin becomes progressively restricted to their descendants fated to form endoderm (Imai, Takada, Satoh, & Satou, 2000), and is no longer observed in the precursors of notochord and neural cells (A6.2 and A6.4 blastomere pairs) (Hudson, Kawai, Negishi, & Yasuo, 2013; Satou, 2020). Nuclear localization of β-catenin is an evolutionarily conserved process that induces transcription of several target genes, primarily through the mediation of transcription factors of the TCF/LEF family (e.g., Valenta, Hausmann, & Basler, 2012). One of the target genes directly activated by β-catenin through TCF/LEF sites found in its promoter region encodes for the transcription factor FoxD, which is required for notochord induction (Imai, Satoh, & Satou, 2002). In fact, FoxD is responsible for activating transcription of ZicL (recently renamed Zic-r.b.), a zinc-finger transcription factor, in A6.2 and A6.4 blastomere pairs, as well as in the B6.2 and B6.4 blastomere pairs of the 32-cell embryo (Imai, Satou, & Satoh, 2002), by acting in combination with maternally stored p53/73-a (Noda, 2011) and with the Ciona counterpart of vertebrate Foxa2, currently called Foxa.a (formerly forkhead/HNF-3β or FoxA-a). While activating expression of endomesodermal genes, at the same time FoxD represses expression of key specifiers of neural fate, Otx, Prdm1r.a, Prdm1-r.b, and Dmrt1 (Tokuhiro et al., 2017). Interestingly, two of these proteins, Prdm1-r.a and Prdm1-r.b, act as transcriptional repressors,

visualize cell and tissue boundaries. (C) High-magnification view of a longitudinal optical section the tail of a hatching Ciona larva electroporated at the one-cell stage with the Bra>GFP plasmid, displaying a fully formed lumen and residual fluorescent notochord cells (N). (D) Transmission electron microphotograph of a small region of the tail of a Ciona embryo, approximately late tailbud III, seen in longitudinal section. The center of the tail is occupied by a large lumen pocket (LP) and residues of notochord cells (N) can be seen around it. Muscle cells (M), rich in mitochondria, can be distinguished on both sides of the notochord. Part of the epidermal cells (E), organized in a monolayer, are visible at the periphery of the section. Magnification 2900, scale bar: 10 μm. (D0 ) High-magnification view of the region boxed in (D), showing in the center a close-up of the notochordal sheath (NS), which is characterized by electron-dense material rich in collagen and extracellular matrix proteins (Cloney, 1964). Magnification 19,000, scale bar: 1 μm.

338

Anna Di Gregorio

and together with a third repressor, Hes.a, are responsible for preventing the simultaneous expression of Foxa.a and ZicL in the brain precursors, which would trigger ectopic expression of Bra in this “non-notochord” lineage (Ikeda & Satou, 2017). These elegant experiments revealed that the limited repertoire of transcription factors present in the Ciona genome can be used in different combinations to generate complex developmental outcomes, and shed light on the role of temporally and spatially localized repression events in cell-fate restriction. One cell division after its activation, ZicL binds the Bra promoter region through two binding sites located 170 bp upstream of its transcription start site, and induces Bra expression in the A-line notochord cells (Yagi, Satou, & Satoh, 2004). The expression of Bra marks the definitive commitment of these cells to the notochord fate. In addition to this gene regulatory cascade, another mechanism controls the expression of Bra in notochord cells. The nuclear localization of β-catenin induces also the expression of FGF9/16/20, which in turn activates expression of Bra via the MAPK/MEK/ERK signaling pathway. FGF9/16/20 is required for the induction of notochord fate, and subsequently, together with FGF8/17/18, maintains the A-lineage committed to form notochord (Yasuo & Hudson, 2007). At the same time, the activity of the MAPK/MEK/ERK signaling pathway, and thus the expression of Bra, are restricted to these cells by the expression of Ci-Ephrin-Ad; embryos lacking the appropriate activity of Ci-Ephrin-Ad ectopically express Bra in neural precursors, as a result of ectopic ERK activation via its dephosphorylation (Picco, Hudson, & Yasuo, 2007; Shi & Levine, 2008). The transcriptional effector of the FGF signaling pathway in the notochord is an ETS family transcription factor(s), and ETS binding sites are found in the Bra “shadow” enhancer (Farley, Olson, Zhang, Rokhsar, & Levine, 2016). Notochord fate induction in the blastomeres of the B-line that will form the secondary notochord involves two different molecular pathways: Nodal and Delta/Notch. The B8.6 secondary notochord precursors receive inductive Nodal signals from their adjacent b-line blastomeres, the b6.5 pair (Hudson & Yasuo, 2006). Primary notochord precursors, A7.6, express the Nodal-downstream gene Delta2, which encodes a divergent, membranebound Notch ligand (Hudson & Yasuo, 2005). Delta2 signaling from the A7.6 blastomeres activates the canonical Delta/Notch pathway in the B-line notochord precursors, and the interaction of the Notch intracellular domain with the transcription factor Suppressor of Hairless, Su(H), induces activation of Bra expression through Su(H) binding sites that have been identified in the Bra promoter region and are required for its function

The notochord gene regulatory network in chordate evolution

339

(Corbo, Fujiwara, Levine, & Di Gregorio, 1998). β-catenin/TCF/LEF, FGF/MAPK/MEK/ERK, Nodal and Delta/Notch/Su(H) are all evolutionarily conserved pathways, and have been adapted in a unique way in ascidians to induce the notochord fate in two separate cell lineages. Although this developmental strategy has been reported thus far only in ascidians, it is intriguing that in Xenopus embryos one of the Su(H) transcription factors, XSu(H)2, is required for expression of Xenopus brachyury (Xbra) (Ito, Katada, Miyatani, & Kinoshita, 2007).

3.2.2 Invagination and convergent extension Notochord precursors invaginate during gastrulation; at this time, they acquire a wedge-shaped aspect and form a monolayer epithelium (Munro & Odell, 2002b). As shown by studies in another ascidian, Boltenia villosa, at the beginning of neurulation notochord precursors divide one last time, and 1 h later, the resulting cells begin to extend F-actin-based protrusion and to acquire motility (Munro & Odell, 2002a). Remarkably, the formation of these protrusions is part of a developmental cell behavior intrinsic to the notochord cells, and is exhibited autonomously by notochord cells in isolation; however, within the embryo, the appropriate planar polarization of the protrusions, which is specific to the interior basolateral edges of the notochord cells, requires interactions with adjacent tissues (Munro & Odell, 2002a). More generally, within the embryo, notochord cells exchange mechanical patterning cues with the muscle cells flanking them (Di Gregorio, Harland, Levine, & Casey, 2002). Shortly after neurulation, two arcs, each containing 20 notochord cells, form on both sides of the embryonic midline, and need to intercalate mediolaterally in order to form a single cylindrical rudiment in the center of the developing tail. This mediolateral intercalation, or convergence, is part of the overall process of convergent extension that will drive tail elongation in the absence of further cell divisions. In vertebrate embryos, actomyosin-based polarized protrusions and/or lamellipodia reportedly generate the forces that drive mediolateral convergence of intercalating cells (e.g., Pfister, Shook, Chang, Keller, & Skoglund, 2016). Together with the extension of protrusions, as discussed in Munro and Odell (2002a), formation of the notochord by invagination of the epithelial sheet that constitutes the roof of the archenteron is a conserved morphogenetic mechanism observed throughout the chordate spectrum, from amphioxus to mouse embryos (Conklin, 1928; Sulik et al., 1994).

340

Anna Di Gregorio

Studies carried out in Ciona savignyi through the analysis of a spontaneous notochord mutation, called aimless, indicate that convergent extension is mediated by the non-canonical planar cell polarity (PCP) pathway ( Jiang, Munro, & Smith, 2005). The aimless mutation has been traced back to the loss of function of the gene prickle, which in Ciona is required for the polarization of notochord cell protrusions and for the asymmetric localization of another PCP component, disheveled (dsh), exclusively to the side of the notochord cells that directly borders the muscle ( Jiang et al., 2005). Shortly after the completion of intercalation, Prickle becomes specifically localized to the anterior pole of the notochord cells, followed by another PCP component, Strabismus/van Gogh, and by one of the Ciona myosin proteins (Kourakis et al., 2014). On the other hand, the nuclei, at least in primary notochord cells, remain located posteriorly, although in ascidians other than Ciona their localization can vary (Kourakis et al., 2014; Newman-Smith, Kourakis, Reeves, Veeman, & Smith, 2015). An additional instructive role for the process of mediolateral intercalation and the formation of protrusions is provided by yet another tissue neighboring the notochord, the developing nervous system. Before and after intercalation, the notochord cells express an FGF receptor, Ci-FGFR, which is activated by Ci-FGF3; the latter is expressed in the ventral-most of the four cells that compose the nerve cord, as observed in cross-section (Fig. 1A). Ci-FGF3 morphant embryos display defects in convergent extension, which have been attributed to an impaired formation of appropriately polarized protrusions (Shi, Peyrot, Munro, & Levine, 2009). Differently from the FGF-mediated notochord induction, this process is independent of the MAPK pathway (Shi et al., 2009). Interestingly, not all notochord cells intercalate simultaneously. Notochord cells located at the anterior-most and posterior-most extremities of this structure intercalate faster, and begin narrowing sooner, than the cells located in its midsection (Fig. 2C0 ). Additionally, unequal cell divisions occur in both A-line and B-line precursors and reduce the volume of the cells that will end up forming the tips of the notochord. Together, these cellular behaviors cause the characteristic tapering of the notochord (Veeman & Smith, 2013; Winkley, Ward, Reeves, & Veeman, 2019). 3.2.3 Formation of the notochordal sheath and elongation As the notochord cells interdigitate during intercalation, their protrusions diminish and they actively secrete ECM components, which will participate in the formation of a basal lamina and of a structure that was originally

The notochord gene regulatory network in chordate evolution

341

described as a collagen-based notochordal sheath (Cloney, 1964). In support of these first findings, recent studies have shown that several collagen genes, representative of at least four collagen families, including fibrillar collagen, are expressed in the notochord cells at these stages (Katikala et al., 2013; Wada et al., 2006; reviewed in Satoh, 2014). Alongside collagen genes, notochord cells express numerous other ECM components, including alpha and beta laminins, fibronectin, entactin, cadherins, thrombospondin and others ( Jos
e-Edwards, Oda-Ishii, Nibu, & Di Gregorio, 2013; Katikala et al., 2013; Kugler et al., 2008). Along with ECM components sensu stricto, also numerous enzymes involved in their secretion and post-translational modification are synthesized by the notochord cells, including leprecan/ P3H1, which encodes the collagen modifier prolyl 3-hydroxylase 1 (P3H1) (Dunn & Di Gregorio, 2009), lysyl oxidases, which are involved in collagen and elastin cross-linking, furin, coatomer proteins, and others (Kugler et al., 2008). Studies of Ciona notochord genes have guided the identification of related genes in mouse, including three orthologs of the single-copy Ciona leprecan/P3H1 gene, named Leprecan/P3H1, Leprecanlike1/P3H2 and Leprecan-like2/P3H3, which are all expressed in notochord cells, as well as in various additional vertebrate-specific territories that are not present in ascidian embryos, including the developing vertebral cartilages (Capellini, Dunn, et al., 2008). The characterization of the Ciona notochord transcriptome through RNA-Seq has confirmed the abundant expression of ECM components and their modifiers by the notochord cells (Reeves, Wu, Harder, & Veeman, 2017). The ECM is also responsible for forming a boundary that under normal circumstances is able to reduce the motility of intercalating notochord cells, thus organizing this critical morphogenetic movement. This mechanism, called “boundary capture,” has been described in amphibian embryos as well (Keller et al., 2000; Smith, 2018). The role of different ECM components in notochord development has been investigated using spontaneous mutants in Ciona savignyi and various gene knockdown techniques in Ciona robusta. Work on another spontaneous Ciona savignyi mutant, chongmague, which carries a mutation in the laminin-α3/4/5 gene, shows that when the ECM is lacking this laminin the notochord cells begin intercalation but maintain their protrusions and their mobility, and migrate randomly to ectopic locations in the tail (Veeman et al., 2008). In Ciona robusta, the notochord cis-regulatory module (CRM, or enhancer) associated with the laminin c1 gene is directly controlled by Bra through cooperative binding sites (Katikala et al., 2013; Fig. 2H).

342

Anna Di Gregorio

The knockdown via CRISPR/Cas9-mediated editing of another ECM component, fibronectin (Ci-FN1-containing in Jos
e-Edwards et al., 2013; renamed Ci-Fn in Segade, Cota, Famiglietti, Cha, & Davidson, 2016), also disrupts intercalation, although it does not seem to affect the structure of the notochordal sheath (Segade et al., 2016). Instead, formation of the notochordal sheath appears impaired by the shRNA-mediated knockdown of leprecan/P3H1 (Dunn & Di Gregorio, 2009). Analogously, CRISPR/ Cas9-mediated editing of this gene causes the formation of a slightly shorter, kinked tail. These results suggest that a reduction in leprecan/P3H1 function might reduce the post-translational modification of collagen molecules in the notochordal sheath, thus lowering its structural integrity and rigidity (Maguire et al., 2018). Notochord elongation in Ciona is characterized by a striking change in cell shape, which takes place through a divergent molecular mechanism. The notochord cells transition from the stack-of-coins shape that they display in middle and late-tailbud to an elongated “drum-like” shape, while maintaining their volumes constant. This impressive process relies upon the formation of an equatorially positioned actomyosin ring that constricts the notochord cells and promotes their elongation (Sehring et al., 2014). Interestingly, the actomyosin ring is highly similar, in both structure and molecular composition, to the cytokinetic ring that is formed during normal cell divisions; however, this ring assembles at the anterior aspect of the notochord cells at the time when the notochord has the appearance of a stack of coins, and interacts with the anteriorly localized PCP components, Dsh, prickle and Strabismus/van Gogh (described above). At this point, the ring repositions itself to the middle of each cell, begins contracting and produces a constriction along its circumference, thus inducing the elongation of each cell and its transition to the characteristic “drum-like” shape (Lu, Bhattachan, & Dong, 2019; Sehring et al., 2015). The actomyosin ring eventually disassembles, through a still unknown mechanism, and tail elongation proceeds through tubulogenesis (Lu et al., 2019).

3.2.4 Tubulogenesis and disappearance The process of tubulogenesis in Ciona has been reconstructed in exquisite detail through a combination of molecular and morphometric studies (Deng et al., 2013; Denker et al., 2013, 2015; Denker & Jiang, 2012; Dong, Deng, & Jiang, 2011; Dong et al., 2009) and is briefly summarized here.

The notochord gene regulatory network in chordate evolution

343

At the onset of tubulogenesis, extracellular vacuoles, or more precisely, intercellular lumen pockets, which were originally mistaken for large intracellular vacuoles by analogy to those observed in vertebrate notochord cells (Fig. 3B and B0 ), form between notochord cells and gradually induce another change in their shape, from “drum-like” to biconcave. All notochord cells form two apical domains, each facing another notochord cell. This peculiar change in cell polarity marks a mesenchymal-to-epithelial transition of the notochord cells ( Jiang & Smith, 2007). Both apical domains in each notochord cell contain Par3/Par6/aPKC protein complexes, organized in patches, as well as two sets of tight junctions (Denker et al., 2013; Oda-Ishii, Ishii, & Mikawa, 2010). The molecular event that triggers the formation of the lumen pockets is the localization, on each apical domain, of the anion transporter SLC26αa, which is necessary for the formation of the notochord lumen and its expansion through the transport of fluid from the notochord cells toward the lumen (Deng et al., 2013). More generally, SLC26 is an evolutionarily conserved family of transporters that participate in numerous events in development and disease, and in humans are expressed in different regions of the renal tubules that are involved in ion transport (e.g., Markovich & Aronson, 2007). As the lumen pockets increase in size by virtue of newly synthesized apical membrane and SLC26mediated ion flow, the notochord cells remain sealed together by tight junctions in the regions where their adjacent apical domains are juxtaposed (Denker et al., 2015). Claudin-rich tight junctions gradually replace the adherens junctions between notochord cells, another event that is required for the proper formation of the lumen pockets (Denker et al., 2013; Dong et al., 2009). Accordingly, transcripts for one of the Ciona claudins, Claudin16/17/19, whose expression is regulated by the transcription factor Bhlh-tun1, become detectable shortly before the onset of this process (Kugler et al., 2019). In vivo confocal studies and morphometric analyses have shown that cortical actin and ezrin-radixin-moesin (ERM) are also required for lumen formation, along with a microtubule network that forms at the apical cortex of the notochord cells through an actin-dependent subcellular movement (Dong et al., 2011). The microtubule network allows the notochord cells to form lamellipodia that enable them to crawl on the notochordal sheath and eventually flatten to an endothelial-like appearance to facilitate the coalescence of the lumen pockets and consequent formation of the definitive fluid-filled lumen (Dong et al., 2011; Fig. 3C). The interaction of ERM with the multifunctional protein 14-3-3εa at the basal cortex of the notochord cells is necessary for the transport of components required

344

Anna Di Gregorio

for lumen formation from the basal region of these cells to their lumenfacing domain (Mizotani et al., 2018). A notochord CRM directly controlled by Bra through cooperative binding sites has been identified in the ERM genomic locus (Katikala et al., 2013; Fig. 2H). Ciona larvae with a fully differentiated notochord are able to swim for a few hours and to select an appropriate substrate to which they will adhere in preparation for metamorphosis. In laboratory cultures, at the temperature of 18 °C, larval hatching occurs 18 h after fertilization (Fig. 2), and swimming continues for about 5 h after hatching; at that point, the larvae attach to a submerged substrate, or to the bottom of Petri dishes filled with seawater, and begin retracting their tail after a minimum of 28 min after adhesion (Matsunobu & Sasakura, 2015). Most of the cells of the tail undergo caspase-dependent apoptosis, including cells of the epidermis and the posterior-most notochord cells (Krasovec, Robine, Qu
einnec, Karaiskou, & Chambon, 2019). Apoptosis is determined and organized by the expression of Ci-Hox12, through an evolutionarily conserved mechanism; on the other hand, the propagation of the apoptotic wave that starts from the posterior end of the tail and proceeds anteriorly is considered a rather unique adaptation among chordates (Karaiskou, Swalla, Sasakura, & Chambon, 2015). A notochord residue remains visible in the metamorphosing larva and in the resulting juvenile, but differently from the notochord remnants that in vertebrates are incorporated in the nuclei pulposi, these residual notochord cells are entirely eliminated as post-metamorphic development proceeds. The expression of Bra is reportedly undetectable after metamorphosis (Chiba, Jiang, Satoh, & Smith, 2009), while expression of transcription factors that are exclusively active in adult structures is activated. Of note, differently from Bra, several notochord genes are rerouted to different cell-types and newly formed structures after the disappearance of the notochord. One representative example is Bhlh-tun1, a transcription factor that functions in the developing notochord (Kugler et al., 2019), and after metamorphosis is expressed first in atrial siphon muscle precursors, and later on in the oral siphon muscle (Razy-Krajka et al., 2014).

4. Time is of the essence: Temporal cis-regulatory control of gene expression in a fast-developing chordate Comparably to other marine organisms, ascidian embryos develop through a larval stage; however, differently from other marine larvae,

The notochord gene regulatory network in chordate evolution

345

ascidian embryos are unable to feed themselves and must rely upon yolk granules and other maternally stored nutrients for their survival. This condition imposes strict limits on the duration of the pre-metamorphic embryogenesis, which at 18 °C is completed within roughly 1 day and takes place even more expeditiously at higher temperatures (Satoh, 2014). At 18 °C, the notochord completes its development, from specification to lumen formation, in 14 h (Fig. 2). Based on timescales determined for mammalian cells, transcription of an average Ciona gene (10.5 kb) is expected to require 10 min, and maturation, translation and folding might take approximately 1–3 additional minutes each (Phillips, Kondev, Theriot, & Garcia, 2012); this suggests that within the 12–14 h required for notochord development, the notochord GRN could, theoretically, utilize a multi-tiered cascade of transcription factors. However, a large fraction of the notochord CRMs analyzed thus far rely upon Bra/T-box binding sites (generic consensus: TNNCAC; Di Gregorio & Levine, 1999; Dunn & Di Gregorio, 2009; Katikala et al., 2013; Jos
e-Edwards et al., 2013, 2015; Thompson & Di Gregorio, 2015). Interestingly, bona fide Bra-downstream notochord genes (described in detail in Section 5.2) are expressed in the developing notochord at different developmental stages (Hotta et al., 2000; Hotta, Takahashi, Erives, Levine, & Satoh, 1999; Takahashi et al., 1999), even though Bra transcripts are present at all stages of notochord development (Corbo, Levine, et al., 1997) and are, arguably, translated into active Bra protein that is detectable in the nuclei of late tailbud embryos (Katikala et al., 2013). In particular, transcripts for early-onset Bradownstream genes are first detected in notochord precursors around early gastrulation; middle-onset target genes become detectable around the late gastrula/neural plate stage, and late-onset target genes are first detected around the beginning of neurulation (Hotta et al., 1999; Katikala et al., 2013). The molecular mechanisms responsible for these differences in the temporal read-out of Bra-downstream gene expression have been investigated through the characterization of the notochord CRMs associated with genes representative of early-, middle- and late-onset Bra targets, and the minimal sequences and binding sites required for their function have been identified (Katikala et al., 2013). The results of these experiments suggest that notochord CRMs that are associated with early-onset notochord genes require multiple functional Bra binding sites (Katikala et al., 2013; Fig. 2H). This category includes thrombospondin 3A (thbsp), and laminin c1, which encode evolutionarily conserved ECM components (Urry, Whittaker, Duquette, Lawler, & DeSimone, 1998), fibrillar collagen 2A1 (CiFCol1), a

346

Anna Di Gregorio

presumptive component of the notochordal sheath, which in vertebrates is reportedly co-opted to cartilaginous structures (Cloney, 1964; Katikala et al., 2013; Wada et al., 2006), and ERM, which is required for cell-shape changes and lumen formation (detailed in Section 3.2.4) (Dong et al., 2011; Hotta, Yamada, Ueno, Satoh, & Takahashi, 2007; Katikala et al., 2013). Middle-onset notochord genes, such as Noto9/KTN1/RRBP1, which is first detected in notochord cells by the neural plate/early neurula stage, and encodes a ribosome-binding protein that is also detected in the notochord of Xenopus embryos (Liu et al., 2016), are controlled through notochord CRMs that rely upon individual functional Bra binding sites (Katikala et al., 2013; Fig. 2H). This model predicts that removal of a functional Bra binding site from the CRM of an early-onset notochord gene that relies upon two Bra binding sites would delay its onset of activity, essentially turning it into a middle-onset CRM. Accordingly, this is the result that was obtained when either one of the two Bra binding sites of the laminin c1 notochord CRM was mutated; either mutation had the effect of turning this CRM into a middle-onset one (Katikala et al., 2013). Lastly, the notochord CRMs of the two late-onset Bra target genes analyzed, ATP-citrate lyase (ACL), which is required for the establishment of cell polarity and for intercalation (Hotta, Yamada, et al., 2007), and β1,4-Galactosyltransferase (β4GalT), which is presumed to be involved in the formation of the notochordal sheath, are devoid of functional Bra binding sites and are likely controlled by Bra indirectly, through a relay mechanism that involves the activity of Bra-downstream intermediary transcription factors (Fig. 2H) (Katikala et al., 2013). Recent ATAC-Seq studies have provided accurate genome-wide profiles of the chromatin state in developing Ciona embryos (Madgwick et al., 2019; Racioppi, Wiechecki, & Christiaen, 2019), which will be instrumental for testing and refining this mechanistic model.

5. The GRN underlying notochord formation in Ciona A GRN can be envisioned as a blueprint of the regulatory interactions existing among genes expressed in a certain tissue, organ, or structure. These interactions are responsible for generating a specific spatial and temporal regulatory state that will, in turn, originate a specific phenotypic output (Peter, 2017). These gene regulatory circuitries are encoded in each genome in form of transcription factor genes and cis-regulatory regions, and as such, are subjected to mutations and evolutionary changes that can lead to

The notochord gene regulatory network in chordate evolution

347

variation in the processes that they control, and in their phenotypic manifestations (e.g., Levine, 2010). Genes whose products control the expression of other genes, such as genes encoding for transcription factors, RNAbinding proteins, and other regulatory molecules, are positioned internally within the GRN, while genes whose products do not control other genes, but rather act as direct effectors of cellular differentiation and morphogenesis, such as genes encoding for enzymes, ECM components and structural proteins, are located at the periphery (Davidson, 2006; Peter & Davidson, 2015). The reconstruction of the Ciona notochord GRN began with the identification of its first regulatory nodes, Brachyury and Foxa.a (Corbo, Erives, Di Gregorio, Chang, & Levine, 1997; Corbo, Levine, et al., 1997; Di Gregorio, Corbo, & Levine, 2001). Toward the end of 2002, the public release of the first assembly of the Ciona intestinalis (now robusta) genome (Dehal et al., 2002) enabled the identification of several notochord genes through sequence homology (Kugler et al., 2008). Hundreds of notochord genes were identified through the coordinated efforts of a group of laboratories that have carried out the systematic identification of the expression patterns of several genes spanning a wide range of developmental stages, from unfertilized eggs and zygotes, to larvae and juveniles (Fujiwara et al., 2002; Kusakabe et al., 2002; Miwata et al., 2006; Nishikata et al., 2001; Ogasawara et al., 2002; Satou et al., 2002, 2001; Satou, Kawashima, Shoguchi, Nakayama, & Satoh, 2005). A few years later, another remarkably comprehensive study, specifically focused on transcription factors and signaling molecules, identified a group of potential regulatory nodes of the notochord GRN (Imai, Hino, Yagi, Satoh, & Satou, 2004; Satou & Satoh, 2005; Satou, Wada, Sasakura, & Satoh, 2008). Together, these outstanding foundational projects provided an enviable arsenal of biomolecular resources for Ciona, and propelled the field of ascidian biology toward the use of a system biology approach for studies of developmental processes and functional genomics. This research culminated with the morpholinomediated knockdown of 53 zygotically expressed transcription factor genes and 23 signal transduction molecules (Imai, Levine, Satoh, & Satou, 2006). This work initiated the reconstruction of the genetic blueprints underlying the formation of most of the tissues found in the Ciona embryo (Imai et al., 2006; Satou et al., 2005 http://ghost.zool.kyoto-u.ac.jp/otherfr_kh.html). The translucency of Ciona embryos, the ease of transgenesis and their availability in large quantities allow the use of fluorescence-activated cell sorting (FACS) for the purification of specific cell populations (Christiaen et al., 2008; Christiaen, Wagner, Shi, & Levine, 2009). FACS-mediated

348

Anna Di Gregorio

isolation of notochord cells was used in combination with microarray screens, and has added to the Ciona notochord GRN a group of transcription factors, described below, which had escaped previous searches, as well as 456 previously published effector genes and >100 novel notochord genes ( Jos
e-Edwards et al., 2011; our unpublished results). Taken together, all these studies have yielded 700 bona fide notochord genes, whose expression has been validated by WMISH. More recently, RNA-Seq on FACSpurified cell populations and single-cell RNA-Seq (scRNA-Seq) have been used to elucidate the full complement of tissue-specific transcriptomes present in Ciona before metamorphosis (Cao et al., 2019; Horie et al., 2018; Reeves et al., 2017). Preliminary analyses of the results of these experiments suggest that the number of genes expressed in the Ciona notochord might rise sharply to a few thousands. The Ciona genome spans 123 Mb and contains a little over 14,000 genes (Dehal et al., 2002; Satou et al., 2019). The total number of transcription factor genes is 385 (Dehal et al., 2002; Satou et al., 2019; Stolfi & Christiaen, 2012); of these, about 50–60 are expressed in the notochord during different stages of its development, with widely variable intensity and duration. This number does not include most of the zinc-finger proteins expressed in the notochord, some of which might function as transcription factors (Miwata et al., 2006), and might further increase after the validation of new candidate notochord genes identified by scRNA-Seq (Cao et al., 2019; Horie et al., 2018).

5.1 The main regulatory nodes of the notochord GRN: Brachyury and Foxa2 Two evolutionarily conserved transcriptional regulators, Brachyury and Foxa2, appear reiteratively in chordate evolution as components of notochord GRNs, and are both present, in single copy, in the Ciona genome (Corbo, Erives, et al., 1997, Corbo, Levine, et al., 1997, Di Gregorio et al., 2001). Ciona Brachyury (Bra; originally published as Ci-Bra; Corbo, Levine, et al., 1997) is a member of the T-box family of transcription factors (Di Gregorio, 2017; Takatori et al., 2004), while Foxa.a is a member of the forkhead/winged-helix family (Di Gregorio et al., 2001; Imai et al., 2004). These two genes possess orthologs in numerous organisms throughout, and outside, the phylum Chordata, and their functions in development and evolution of multicellular organisms decidedly precede the role in notochord formation that they acquired in the chordate lineage. Mice heterozygous for mutations in the Bra locus are characterized by a short tail (Brachyury, in Greek), while mice homozygous for mutant alleles of Bra

The notochord gene regulatory network in chordate evolution

349

die in utero because they lack notochord and allantois, and display abnormal somites (Stott, Kispert, & Herrmann, 1993; Wilkinson, Bhatt, & Herrmann, 1990). Likewise, ENU-generated Bra mutant Ciona embryos display a very short tail and lack an organized notochord (Chiba et al., 2009). However, differently from most other chordates examined thus far, Bra is notochord-specific in Ciona (Corbo, Levine, et al., 1997), and this renders ascidian embryos ideally suited for studies of the notochord-specific function of this transcription factor. In mouse embryos, Foxa2 is expressed in notochord, nervous system, endoderm and additional territories (Sasaki & Hogan, 1993). Mice carrying a homozygous mutation in the Foxa2 locus lack an organized node and fail to develop a notochord (Ang & Rossant, 1994). Resembling its vertebrate counterparts, Ciona Foxa.a is expressed in notochord, endoderm, and in the ventral-most cells of the nerve cord, which are considered a rudimentary floor plate (Corbo, Erives, et al., 1997; Di Gregorio et al., 2001). The crossregulatory relationship between these two transcription factors has been investigated in the ascidians Ciona and Halocynthia and in vertebrates. In Ciona, the results of the morpholino-mediated knockdown of Foxa.a indicate that this transcription factor controls the expression of numerous genes, among which Bra, which is down-regulated in Foxa.a morphants (Imai et al., 2006). These results are confirmed by parallel studies in Halocynthia roretzi, which show that Hr-FoxA is sufficient to elicit ectopic expression of Hr-Bra (Kumano, Yamaguchi, & Nishida, 2006), and by experiments carried out in mouse embryos (Tamplin et al., 2008). On the other hand, Bra controls Foxa2 expression in mouse embryoid bodies (Lolas, Valenzuela, Tjian, & Liu, 2014), and the Foxa.a locus in Ciona is bound by Bra in chromatin immunoprecipitation on DNA microarray (ChIP-chip) and in electrophoretic mobility shift assay (EMSA) (Kubo et al., 2010; our unpublished results). The presumptive positive feedback between these two transcription factors suggests that they constitute a subcircuit of the Ciona notochord GRN. Moreover, experimental evidence suggests that both transcription factors feedback on their own transcription, with different effects. In the case of Bra, ChIP-chip data suggest the presence of autoregulatory sites in the Bra locus (Kubo et al., 2010), and interestingly, morpholino-mediated knockdown experiments indicate that Bra negatively regulates its own transcription (Imai et al., 2006). In the case of Foxa.a, the occupancy of the Foxa.a locus by its own protein product is confirmed by EMSA experiments that show that binding sites for winged-helix proteins within the Foxa.a cis-regulatory region are bound by an in vitro synthesized

350

Anna Di Gregorio

Foxa.a protein (Di Gregorio et al., 2001). Mutation analysis of Foxa.a binding sites within the Foxa.a cis-regulatory region in vivo indicated the positive autoregulatory role of these sequences (Di Gregorio et al., 2001; Fig. 4). A positive autoregulatory loop is predicted for mouse Foxa2 as well (Tamplin et al., 2008). Bra and Foxa2 orthologs are reiteratively incorporated into different GRNs that operate in diverse structures/tissues of non-chordate embryos (Davidson, 2006, 2010; Davidson & Erwin, 2006). In Ciona and mouse embryos, these two transcription factors appear to have established positive cross-regulatory interactions that have locked them into a subcircuit that is presumably specific to the notochord and required for its development. In support of this possibility, work on Ciona notochord CRMs has provided the first evidence that these two transcription factors synergistically control notochord gene expression ( Jos
e-Edwards et al., 2015; Passamaneck et al., 2009; Section 6.1). Future studies on the structure of the notochord GRNs in other chordates will clarify whether the positive feedback circuitry between Bra- and Foxa2-related transcription factors is evolutionarily conserved, i.e., whether the Bra/Foxa2 subcircuit constitutes a notochord “kernel” (Davidson & Erwin, 2006; Peter & Davidson, 2015).

5.2 Drawing edges, connecting nodes: Reconstructing the gene batteries controlled by notochord transcription factors The lineage of the notochord cells in Ciona is invariant, as is the final number of 40 post-mitotic cells (Nishida & Satoh, 1983, 1985). The overwhelming majority of notochord genes identified thus far are expressed fairly homogeneously in all 40 notochord cells. However, a few genes, such as multidom, exhibit a peculiar mosaic expression that is apparently random and lineageindependent (Oda-Ishii & Di Gregorio, 2007), while others display a more regionalized expression along the anterior-posterior axis of the notochord (Reeves, Thayer, & Veeman, 2014; Utsumi, Shimojima, & Saiga, 2004), and a few examples of genes specifically expressed in secondary notochord have been recently reported (Harder, Reeves, Byers, Santiago, & Veeman, 2019). Interestingly, Multidom is a protein characterized by multiple domains (Oda-Ishii & Di Gregorio, 2007) including a region of sequence similarity to Delta-like proteins. The full complement of genes that exhibit these heterogeneous expression patterns, and their impact on the global notochord developmental program, are yet to be determined. Nevertheless, the Ciona notochord could be envisioned as a

The notochord gene regulatory network in chordate evolution

351

Fig. 4 A close-up of the Bra/Foxa.a subcircuit and its relationship with notochord transcription factors, differentiation genes, and CRMs. An extremely detailed and comprehensive view of the Ciona notochord specification GRN has been garnered through expression studies and morpholino-mediated knockdown experiments (Imai et al., 2004, 2006), and is available online on the Ghost database website (http://ghost.zool. kyoto-u.ac.jp/otherfr_kh.html Satou et al., 2005). This is a summary of the Bra/Foxa.a subcircuit, mainly based on our findings on notochord transcription factors that were not characterized in those previous studies, and on their target genes and notochord CRMs. For simplicity, information about the temporal onset of individual genes has been omitted from this circuit diagram. After the identification of Bra and Foxa.a, and of their cis-regulatory regions (Corbo, Erives, et al., 1997; Corbo, Levine, et al., 1997; Di Gregorio et al., 2001), subsequent studies have uncovered the notochord expression of Bhlh-tun1 and Tbx2/3 (Imai et al., 2004), MIER1 and Xbp1 (Kugler et al., 2008), and of the seven additional notochord transcription factors shown in colored font, most of which are conserved in vertebrate notochord and/or nuclei pulposi (Jos
e-Edwards et al., 2011). The notochord CRM of Bhlh-tun1 is controlled directly by Bra, Foxa.a and by an early-onset homeodomain transcription factor (early HD, gray), which remains to be ascertained (Kugler et al., 2019). Transcription of NFAT5 is directly regulated by Bra (our unpublished results), as is that of Tbx2/3, which is activated through cooperative Bra binding sites (Jos
e-Edwards et al., 2013). Peripheral components (differentiation effector genes) of this part of the GRN are shown in black font; among these, effector genes whose notochord CRMs have been fully characterized are annotated in bright blue font. The role of a Myb-like transcription factor in the activation of some of the notochord CRMs (C6-sulfotransferase and carboxypeptidase) has been inferred on the basis of the minimal sequences required for notochord activity of these cis-regulatory regions (Jos
e-Edwards et al., 2015). Twenty-one notochord genes have been identified as targets of Bhlh-tun1, including Claudin16/17/19, whose notochord CRM has been identified (Kugler et al., 2019). Transcription of several notochord genes is influenced by Xbp1 (our unpublished results), and 20 notochord genes are controlled by Tbx2/3, including Fos-a and Duox-c, which are down-regulated by it (Jos
e-Edwards et al., 2013). The fibronectin notochord CRM was characterized by Segade et al. (2016) and relies upon a T-box binding site, which could be bound by Bra and/or Tbx2/3, and a divergent Fox-like binding site. The Ephrin 3 notochord CRM is controlled directly by Bra, but requires for its function also an (AC)6 microsatellite repeat (Jos
e-Edwards et al., 2015), an arrangement (Continued)

352

Anna Di Gregorio

self-contained biological system that is relatively impermeable to variability and perturbation, and as such, can be systematically dissected through the analysis of its regulatory nodes and of the gene batteries that they (co-) regulate. According to the recent estimates obtained from scRNA-Seq studies, the notochord GRN might involve the activity of a few thousand genes to complete all the steps of its morphogenesis. If all these genes were equally controlled by the 60 transcription factors expressed in the notochord, in a series of nearly parallel independent cascades, this would assign a few hundred target genes to each transcription factor. A more realistic scenario depicts the Bra/Foxa.a subcircuit presiding a regulatory differentiation hierarchy where either transcription factor, or both, control the majority of the peripheral nodes of the GRN, i.e., the differentiation effector genes, either directly, or indirectly through the use of transcriptional intermediaries. Extremely detailed GRN maps have been reconstructed for the pre-metamorphic tissues of the Ciona embryo through the systematic morpholino-mediated knockdown of numerous genes encoding for transcription factors (Imai et al., 2006; Satoh, 2014; Satou et al., 2005 http://ghost.zool.kyoto-u.ac.jp/otherfr_kh.html). In Fig. 4, we provide a high-magnification view of the Bra/Foxa.a subcircuit that is activated after notochord specification is completed (Section 3.2.1; Satou, 2020). This summary is mainly focused on transcription factors identified more recently, which therefore were not included in the global GRN maps, and are currently being characterized ( Jos
e-Edwards et al., 2011; Kugler et al., 2008, 2019; our unpublished results). The first two of these transcription factors, Ciona Xbp1 and Mi-er1, were identified through a survey for orthologs of vertebrate notochord genes in Ciona (Kugler et al., 2008), while Sall-a (related to both human genes SALL1 and SALL3), NFAT5, Stat5/6b, Klf15, Fos-a, AFF2/3/4, and Lmx-like were identified, along with Fig. 4—Cont’d that is also found in Bra-bound genomic regions in mouse embryonic stem cells (Evans et al., 2012). Regulatory interactions that could be either direct or indirect are interrupted by diagonal black lines. Positive interactions are shown as arrows, inhibitory interactions are indicated by flat-headed arrows. Dashed lines indicate interactions that are hypothesized and need to be confirmed. Negative autoregulatory feedback for Bra, revealed by morpholino-mediated knockdown (Imai et al., 2006), is indicated by a flat-headed circular arrow. Positive autoregulatory feedback for Foxa.a (circular arrow) was uncovered by in vivo mutation analyses coupled with in vitro DNA-binding assays (Di Gregorio et al., 2001). All additional data for notochord CRMs are from Dunn and Di Gregorio (2009), Passamaneck et al. (2009), Jos
e-Edwards et al. (2015), Katikala et al. (2013), Thompson and Di Gregorio (2015). Abbreviations: Olfactom.1, olfactomedin 1; Sulfotr., sulfotransferase.

The notochord gene regulatory network in chordate evolution

353

previously published transcription factors and numerous effector genes, through the microarray screen detailed above ( Jos
e-Edwards et al., 2011; our unpublished results). As discussed in the respective publications, with the exception of Klf15, all these transcription factors are related to vertebrate counterparts that are expressed in notochord and/or nuclei pulposi; it is noteworthy that a recent study reported that SALL3 is a unique marker of spine chordoma, while LMX1A is predominant in skull base chordoma (Bell et al., 2018). In situ hybridization experiments on embryos that were either carrying a mutation in the Bra locus or that were expressing a repressor form of this transcription factor showed that embryos in which the levels or the functions of Bra are altered, the expression of NFAT5, AFF2/3/4, Fos-a and Klf15 is no longer detectable in the misshapen cells that replace the notochord, while expression of Lmx-like remains relatively unaltered ( Jos
e-Edwards et al., 2011). These results suggest that Lmx-like, which is robustly expressed in notochord cells starting from gastrulation and throughout the main steps of notochord formation, might be part of a branch of the notochord GRN that is completely or partially independent of Bra ( Jos
e-Edwards et al., 2011). Considerable effort has been directed toward drawing edges between these regulatory nodes of the GRN and their respective effectors. Fig. 4 includes partial lists of target genes that have been identified for a few of the notochord transcription factors, as well as genes whose notochord CRMs have been completely characterized and have guided the elucidation of the transcription factors controlling them (detailed in Section 6). The notochord-specific expression of Bra has enabled the successful identification of notochord genes controlled by this transcription factor through its ectopic expression in neural and endodermal precursors. Transgenic embryos carrying a fusion of the Foxa.a promoter region to the Bra cDNA ectopically express Bra in endoderm and neural precursors and display a characteristic phenotype. These embryos were isolated for RNA extraction, followed by subtractive hybridization with RNAs extracted from stage-matched wild-type controls (Takahashi et al., 1999). This study identified 50 bona fide notochord transcriptional targets of Bra, including tropomyosin-like, prickle, ERM, leprecan/P3H1, Noto4/PID1, CiFCol1, multidom, ACL, b4GalT, and several others, whose function and/or transcriptional regulation have been studied in depth (Di Gregorio & Levine, 1999; Dunn & Di Gregorio, 2009; Hotta et al., 1999, 2000; Hotta, Takahashi, Satoh, & Gojobori, 2008; Hotta, Yamada, et al., 2007; Jiang et al., 2005; Katikala et al., 2013; Maguire et al., 2018; Oda-Ishii &

354

Anna Di Gregorio

Di Gregorio, 2007; Takahashi et al., 1999, 2010). This experiment has been more recently repeated using RNA-Seq, and has identified 925 putative Bra targets; however, many of these genes are still uncharacterized, and only a fraction of those genes for which expression patterns are available are expressed in the notochord (Reeves et al., 2017). Furthermore, ChIP-chip experiments have provided a genome-wide map of Bra-bound genomic loci, indicating that approximately 2100 individual genes are occupied by a transgenic Bra protein in embryos at the 112-cell stage; 194 of these genes encode for transcription factors, including Foxa.a (Kubo et al., 2010). Surprisingly, the overlap between the putative Bra targets identified through these surveys is limited to 459 genes (Reeves et al., 2017). We suggest that this discrepancy might be attributable to the technical differences in the methods employed for these experiments (subtractive hybridization, ChIP-chip and FACS/RNA-Seq) and also to the differences in the developmental stages that were assayed. These observations could also indicate that the Bra-downstream transcriptional targets constitute a large contingent of genes that changes dynamically as notochord morphogenesis proceeds, possibly in response to changes in chromatin configuration and accessibility, or to fluctuations in the levels of functional Bra protein. The very limited overlap between Bra target genes datasets identified through ChIP experiments in mouse (1942 Bra targets), zebrafish (218 Bra targets) and Xenopus tropicalis (1040 Bra targets) supports this hypothesis (Lolas et al., 2014). Furthermore, the analysis of human BRA target genes identified via ChIP-Seq in human ES cells has identified a group of roughly 800 BRA targets that appear specific to this population (Faial et al., 2015). In the case of Foxa.a, ChIP-chip experiments indicate that 3653 regions of the Ciona genome are bound by this transcription factor in 112-cell embryos, including the loci of 245 transcription factor genes, among which are Bra and Foxa.a itself (Kubo et al., 2010). Subtractive microarray screens were used to investigate the gene battery downstream of Tbx2/3, the only other T-box transcription factor reportedly expressed in notochord cells, which, differently from Bra, is not notochord-specific and is expressed also in various regions of CNS and epidermis (Imai et al., 2004; Jos
e-Edwards et al., 2013; Takatori et al., 2004). This survey led to the identification of 20 notochord genes whose expression is influenced by this transcription factor, and of 61 genes that are controlled by Tbx2/3 in other territories encompassed by its expression ( Jos
e-Edwards et al., 2013). The Tbx2/3 locus is bound by both

The notochord gene regulatory network in chordate evolution

355

Bra and Foxa.a (Kubo et al., 2010) and expression of Tbx2/3 in the notochord is lost in Bra mutant embryos, while it remains unaltered in CNS and epidermis ( Jos
e-Edwards et al., 2013). The characterization of a notochord CRM isolated from the Tbx2/3 locus revealed that this transcription factor is indeed regulated by Bra through cooperative binding sites ( Jos
e-Edwards et al., 2013; Fig. 2H). Part of the Tbx2/3 targets are shared with Bra, which suggests that Tbx2/3 controls part of the Bra-downstream notochord GRN; these genes include Noto4/PID1, which is required for notochord intercalation (Yamada, Ueno, Satoh, & Takahashi, 2011) and whose notochord CRM is controlled through an individual functional Bra binding site (Fig. 2H; Katikala et al., 2013), and fibronectin, whose notochord CRM contains binding sites for transcription factors of the T-box and Fox families, and is required for intercalation as well (Segade et al., 2016). These findings are in agreement with the effects of the over-expression of mutant forms of Tbx2/3 on notochord intercalation ( Jos
e-Edwards et al., 2013). It is noteworthy that the consensus binding sites determined through SELEX assays for Bra and Tbx2/3 are very similar (Nitta et al., 2019), and this suggests the hypothesis that related T-box binding sites might be used interchangeably by Bra in early embryos and by Tbx2/3 at later stages. It is also possible that slight differences in sequence between these consensus binding sites might determine more selective binding by these transcription factors in vivo. Microarray screens were employed also to determine the identities of notochord genes controlled by Bhlh-tun1; the locus of this transcription factor is bound by both Bra and Foxa.a in early embryos (Kubo et al., 2010) and notochord expression of Bhlh-tun1 is lost in Bra mutants (Kugler et al., 2019). Its minimal notochord CRM requires both Bra and Foxa.a binding sites for its function, along with a binding site for a homeodomain-containing factor (Kugler et al., 2019). Among the 21 notochord genes whose transcription is influenced by Bhlh-tun1, 16 are shared with Bra and encode for ECM components, transmembrane transporters, effectors of cell-shape changes, and a few enzymes; one of the Bhlh-tun1-downstream genes is Claudin16/17/19, whose product is likely involved in joining together notochord cells at the onset of tubulogenesis (Section 3.2.4) (Kugler et al., 2019). Lastly, microarray screens and RNA-Seq experiments have revealed numerous notochord target genes for Xbp1 (Kugler et al., 2008; Fig. 4; our unpublished results) and NFAT5 ( Jos
e-Edwards et al., 2011; our unpublished results).

356

Anna Di Gregorio

Aside from the interactions between Bra and Foxa.a, little is known about possible cross-regulatory interactions between other transcription factors within the notochord GRN. As discussed above, the cross-regulatory interaction between Bra and Foxa.a (Fig. 4) is presumed on the basis of chromatin occupancy and morpholino knockdown results; however, it has been suggested that the regulation of Bra by Foxa.a could be achieved indirectly through FoxD and ZicL, since the promoters of both these genes are occupied by Foxa.a in early embryos (Kubo et al., 2010). The positive regulation/feedback of Foxa.a by Bra remains to be confirmed in Ciona, although it could be predicted on the basis of the occupancy of the Foxa. a locus by Bra, observed through ChIP-chip assays (Kubo et al., 2010), and of the results obtained in mouse embryoid bodies, which indicate that Foxa2 is a direct Bra target and its transcription is activated by it (Lolas et al., 2014). Microarray screens revealed that Tbx2/3 represses expression of Fos-a, and that Bhlh-tun1 activates Lhx3/4/5, which is, however, characterized by a very short window of expression in notochord precursors ( Jos
eEdwards et al., 2013; Kugler et al., 2019). Additional cross-regulatory interactions will likely surface as more cis-regulatory regions and gene batteries are characterized for more notochord transcription factors.

6. Reverse-engineering the notochord GRN: Lessons from notochord cis-regulatory modules Over the past two decades, a straightforward electroporation protocol for transgenesis, the availability of Ciona embryos and their rapid development have facilitated the discovery of a large number of cis-regulatory regions active in notochord cells, along with numerous enhancers active in other tissues (reviewed in Di Gregorio & Levine, 2002; Irvine, 2013; Kusakabe, 2005; Wang & Christiaen, 2012). Nearly 40 notochord cisregulatory regions have been identified and systematically dissected, and their minimal functional sequences have been used to reverse-engineer the notochord GRN through the identification of the transcription factors responsible for their activity (Anno, Satou, & Fujiwara, 2006; Christiaen et al., 2008; Corbo, Levine, et al., 1997; Di Gregorio & Levine, 1999; Dunn & Di Gregorio, 2009; Farley et al., 2016; Harder et al., 2019; Jos
e-Edwards et al., 2015, 2013; Katikala et al., 2013; Kugler et al., 2019; Passamaneck et al., 2009; Segade et al., 2016; Thompson & Di Gregorio, 2015; our unpublished results). Many of these CRMs rely upon

The notochord gene regulatory network in chordate evolution

357

Bra/T-box binding sites for their function (Fig. 2H, Fig. 4), and the relationship between their temporal onsets and the presence of multiple/ individual functional Bra binding sites within their minimal sequences have been discussed in Section 4 (Fig. 2H).

6.1 Ciona notochord CRMs can be synergistically activated by Foxa.a in cooperation with either Bra or other transcription factors The characterization of the Ci-tune CRM provided the first evidence of a notochord CRM controlled synergistically by orthologs of Brachyury and Foxa2 (Passamaneck et al., 2009), followed by the identification of two additional CRMs controlled in a similar way in a subsequent study ( Jos
eEdwards et al., 2015). The results of genome-wide ChIP-chip assays indicate that approximately 1020 individual Ciona loci are occupied by both Bra and Foxa.a in early embryos (Kubo et al., 2010). Apart from synergizing with Bra, Foxa.a is able to trigger notochord gene expression through multiple binding sites (Anno et al., 2006), or by acting in concert with other transcription factors. The analysis of a Ciona notochord CRM, Ci-CRM112, which is associated with an olfactomedin-like gene, indicates that Foxa.a can work in concert with transcription factors of the homeodomain (e.g., Lmx-like) and AP1 family (e.g., Fos-a) ( Jos
eEdwards et al., 2015; dashed violet and orange lines in Fig. 4). These results are consistent with the chromatin-opening and remodeling properties that characterize transcription factors of the Fox family (e.g., Lalmansingh, Karmakar, Jin, & Nagaich, 2012).

6.2 Additional notochord cis-regulatory mechanisms identified in Ciona The notochord CRM that is associated with the early homeodomain transcription factor gene Mnx relies upon a combination of binding sites for Ets and ZicL, as does another notochord CRM located upstream of the main Bra enhancer/promoter region (Farley et al., 2016; Matsumoto, Kumano, & Nishida, 2007; Yagi et al., 2004). Furthermore, at least three notochord CRMs rely for their activity upon minimal sequences that resemble binding sites for transcription factors of the Myb family, while one CRM, Ci-CRM26, requires a binding site for a basic helix-loop-helix (bHLH) transcription factor ( Jos
e-Edwards et al., 2015). Regulation of notochord gene expression in the secondary notochord cells requires multiple inputs, both negative and positive. The notochord

358

Anna Di Gregorio

cis-regulatory region associated with the gene KH.C11.331, which encodes for a putative ECM component of the fibulin/hemicentin family, contains a silencer that prevents expression in primary notochord cells, as well as sequences that mediate its response to FGF and Wnt signaling pathways, both of which control patterning of the posterior regions of the tail (Harder et al., 2019). Even though the molecular players responsible for the transcriptional repression of these genes in the primary lineage are yet to be identified, this finding lays the foundation for future studies of the molecular differences between primary and secondary notochord in ascidians.

6.3 Comparative studies of notochord cis-regulatory regions across chordates The lack of self-evident linear conservation between the cis-regulatory regions identified in Ciona and those identified in other chordates can be overcome by focusing on functional transcription factor binding sites as the main building blocks and descriptors of notochord CRMs. This reductionist approach has allowed us to begin a comparative study of notochord cis-regulatory regions across chordates, and to tentatively group Bra- and/or Foxa2-dependent notochord CRMs from Ciona and from vertebrates within a chordate-wide, basal strategy of cis-regulatory control of notochord gene expression ( Jos
e-Edwards et al., 2015). This group of CRMs is currently the largest, and includes CRMs containing Bra/T-box binding sites, Fox binding sites, or combinations of both. In vertebrates, notochord cis-regulatory regions that rely upon Bra binding sites for their function are the Sonic hedgehog (Shh) ar-C intronic notochord and floor plate enhancer identified in zebrafish (M€ uller et al., 1999) and the eFGF promoter/ enhancer region identified in Xenopus (Casey, O’Reilly, Conlon, & Smith, 1998; Tada, Casey, Fairclough, & Smith, 1998). An intriguing variation on this theme was identified through the analysis of the Ciona Ephrin3 notochord CRM (Fig. 4), which necessitates both a Bra binding site and an (AC)6 microsatellite repeat ( Jos
e-Edwards et al., 2015). This unexpected association of a Bra binding site(s) with a (AC)6 repetitive sequence has been demonstrated, through ChIP-chip assays, to be bound by mouse Bra in embryonic stem cells (Evans et al., 2012). The results obtained in Ciona predict that at least a fraction of these regions within the mouse genome might possess cis-regulatory activity. To date, notochord CRMs equally dependent upon both Brachyury and Foxa2 binding sites have been identified only in Ciona, although it seems

The notochord gene regulatory network in chordate evolution

359

predictable that this configuration might be present in vertebrate cisregulatory regions as well, given that the functions of these transcription factors, their co-expression in the notochord, and their binding sites are evolutionarily conserved across chordates. Numerous notochord CRMs regulated exclusively through multiple Foxa2 binding sites have been identified in mouse embryos, through a survey centered on Foxa2-downstream genes (Tamplin, Cox, & Rossant, 2011). In most of the Fox-containing CRMs that have been identified, Foxa2 binding sites work cooperatively, as is the case for mouse Pkd1/1-1, Shh, Bicc1-1, which contain five, three and two Foxa2 binding sites, respectively ( Jeong & Epstein, 2003; Tamplin et al., 2011); however, the Sox9 E1 enhancer contains an individual Foxa2 binding site (Bagheri-Fam et al., 2006). In zebrafish, a few notochord CRMs contain individual Foxa2 binding sites functionally associated with a recurrent sequence motif (Rastegar et al., 2008). In addition to the chordate-wide Bra- and/or Foxa2-dependent notochord CRMs, the Ciona genome contains notochord CRMs that depend upon binding sites for transcription factors of the bHLH, Myb, Zic, Ets and homeodomain families (Farley et al., 2016; Jos
e-Edwards et al., 2015; Katikala et al., 2013); notochord CRMs of these categories are yet to be identified in vertebrates. Conversely, a few vertebrate notochord CRMs contain sequences that are yet to be identified in Ciona, such as the orphan binding site (OBS) contained in the node and nascent notochord enhancer (NOCE) of mouse Noto (Alten et al., 2012), and the binding site for a Tead transcription factor in the Foxa2 notochord cis-regulatory region (Sawada et al., 2005). These findings hint at the existence of vertebrate-specific mechanisms of activation of notochord gene expression, which could have either evolved after the divergence of tunicates from the chordate lineage leading to vertebrates, or could have been present in a common chordate ancestor and might have been selectively lost in either ascidians or in all tunicates.

7. A comparative view of the notochord GRN in Ciona and other chordates Ascidian embryos lack an organizer, a structure analogous to the dorsal lip of the blastopore in Xenopus, or an equivalent signaling source that is physically distinguishable, can be transplanted from embryo to embryo and is sufficient to induce formation of ectopic structures in its new embryonic context (e.g., Anderson & Stern, 2016). Nevertheless, several of the

360

Anna Di Gregorio

signaling molecules involved in vertebrate notochord induction, such as chordin, bone morphogenetic protein (BMP) and FGF are involved in the early events of notochord specification in Ciona (discussed in Kourakis & Smith, 2005; Passamaneck & Di Gregorio, 2005; Satou, 2020) and in Halocynthia (Darras & Nishida, 2001). After its specification, the Ciona notochord GRN shares with its vertebrate counterparts the central role of Bra and Foxa2. However, some of the key components of vertebrate GRNs appear to be missing from the Ciona notochord GRN. Among them is the homeodomain transcription factor Noto, which is required for notochord formation across chordates, from zebrafish to mouse (e.g., Morley et al., 2009; Talbot et al., 1995; Zizic Mitrecic, Mitrecic, Pochet, Kostovic-Knezevic, & Gajovic, 2010). Although a putative Ciona Noto gene has been identified, its expression in the notochord appears very weak, transient, and limited to the posteriormost region of the tail in the early tailbud stage. On the other hand, in Halocynthia roretzi, expression of Hr-Noto appears stronger, although it is still limited to the posterior-most primary notochord and all secondary notochord cells (Utsumi et al., 2004). Besides Noto, other homeodomain transcription factors, namely, representatives of the Hox and Pbx subfamilies, which are found in vertebrate notochords (e.g., Capellini et al., 2008; Wang et al., 2014), seem remarkably absent from the Ciona notochord (e.g., Keys et al., 2005; Satou et al., 2001). Ciona possesses an incomplete, fragmented Hox cluster, whose components are expressed in a variety of pre- and post-metamorphic tissues and structures, ranging from the larval epidermis and nervous system to the juvenile digestive tract and spermiduct sensory organ, with the notable exception of the notochord and its precursors (Di Gregorio et al., 1995; Ikuta, Yoshida, Satoh, & Saiga, 2004; Keys et al., 2005; Passamaneck & Di Gregorio, 2005; Spagnuolo et al., 2003; Tajima et al., 2020). However, the absence of Hox genes expression from the Ciona notochord contrasts with findings in the amphioxus Branchiostoma lanceolatum, where expression of Hox14 is observed in the posterior notochord and is influenced by retinoic acid (Pascual-Anaya et al., 2012); even among other tunicates, the larvacean Oikopleura dioica displays colinear Hox gene expression in the notochord (Seo et al., 2004). Together, these data favor the hypothesis of a specific loss of Hox gene expression in the Ciona notochord and possibly in the notochord of other ascidians. Overall, the entire homeobox gene family appears underrepresented in the Ciona notochord, except for the early expression of Mnx and Lhx3/4/5 and the more sustained expression of Ciona Islet

The notochord gene regulatory network in chordate evolution

361

(Giuliano, Marino, Pinto, & De Santis, 1998; Imai et al., 2004). It is plausible that transient, weak expression of additional homeobox genes will be revealed through the validation of recent scRNA-Seq data (Cao et al., 2019; Horie et al., 2018). Another remarkably divergent characteristic of the Ciona notochord is the absence of members of the Sonic hedgehog (Shh) pathway (Takatori, Satou, & Satoh, 2002), whose orthologs are present in the genome but are expressed in the visceral ganglion and pharyngeal endoderm of the larva (Islam, Moly, Miyamoto, & Kusakabe, 2010). In vertebrates, Shh is abundantly expressed in both notochord and floor plate (Fig. 1C), and is required for notochordal sheath formation and patterning of the nuclei pulposi (Choi & Harfe, 2011); its down-regulation has been associated with degeneration of the intervertebral discs (Rajesh & Dahia, 2018). In amphioxus, AmphiHh and other members of the Shh pathway, Patched, Smoothened and Suppressor of Fused, are expressed in the notochord and additional structures, thus more closely resembling the situation of vertebrates (Lin et al., 2009; Shimeld, 1999). Hence, also the loss of the Shh pathway from the notochord seems to have occurred selectively in the ascidian lineage.

8. Concluding remarks and future perspectives More tunicate genomes are currently being sequenced and compared, and together with the elucidation of novel gene expression patterns, the results of these investigations are expected to pinpoint the evolutionary timing of the loss of Noto, Hox and Shh from the ascidian notochord. The much-needed characterization of additional notochord CRMs in vertebrates and cephalochordates will expand the comparative studies of notochord cis-regulatory regions that we have initiated using Ciona, and will increase the knowledge of the repertoire of cis-regulatory strategies employed by divergent chordates to achieve notochord gene expression. These comparative studies will also highlight clade-specific molecular mechanisms that are responsible for the morphological and functional differences in the notochords of different chordates, and will sharpen the distinction between basal chordate-wide cis-regulatory mechanisms, cladespecific divergent strategies, and vertebrate innovations. Clustered regularly interspaced short palindromic repeats and Cas9 endonuclease (CRISPR/Cas9)-mediated genome editing (Sasaki, Yoshida, Hozumi, & Sasakura, 2014; Stolfi, Gandhi, Salek, & Christiaen, 2014) is being successfully employed to delete cis-regulatory regions active

362

Anna Di Gregorio

in heart and cardiopharyngeal lineages (Racioppi et al., 2019) and to ablate notochord CRMs (our unpublished results), and will clarify the functional requirement and pervasiveness of redundant or additive notochord cisregulatory regions. Ongoing studies on the role of non-coding RNAs (Chen, Pedro, & Zeller, 2011; reviewed in Velandia-Huerto, Brown, Gittenberger, Stadler, & Bermu´dez-Santana, 2018) and on the function of transient, post-specification expression of transcriptional repressors in the notochord (Erives, Corbo, & Levine, 1998), together with further studies of Braindependent gene regulation ( Jos
e-Edwards et al., 2011; Kugler et al., 2008) will increase the resolution and depth of the sophisticated GRN that powers the simple notochord of Ciona.

Acknowledgments Thanks to all present and past lab members and collaborators. I am particularly grateful to Emerita Prof. Kathleen Sulik (University of North Carolina), to Prof. Jean-Pierre SaintJeannet (NYU College of Dentistry), Ms. Yushi Wu, Drs. Jordan M. Thompson, Yale Passamaneck, and Julie Maguire, for generously sharing their original microphotographs. I remain indebted to Drs. Eric Davidson and Michael Levine for invaluable discussions and for their contagious passion for regulatory sequences and gene regulatory networks. Special thanks to Cettina and Franco Di Gregorio and to Ms. Marianna Nibu for encouragement and inspiration. Research in our lab is currently supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health, under award number R03HD098395.

References Adams, D. S., Keller, R., & Koehl, M. A. (1990). The mechanics of notochord elongation, straightening and stiffening in the embryo of Xenopus laevis. Development, 110(1), 115–130. Almaza´n, A., Ferra´ndez-Rolda´n, A., Albalat, R., & Can˜estro, C. (2019). Developmental atlas of appendicularian Oikopleura dioica actins provides new insights into the evolution of the notochord and the cardio-paraxial muscle in chordates. Developmental Biology, 448(2), 260–270. Alten, L., Schuster-Gossler, K., Eichenlaub, M. P., Wittbrodt, B., Wittbrodt, J., & Gossler, A. (2012). A novel mammal-specific three partite enhancer element regulates node and notochord-specific Noto expression. PLoS One, 7(10), e47785. Anderson, C., & Stern, C. D. (2016). Organizers in development. Current Topics in Developmental Biology, 117, 435–454. Ang, S. L., & Rossant, J. (1994). HNF-3 beta is essential for node and notochord formation in mouse development. Cell, 78, 561–574. Anno, C., Satou, A., & Fujiwara, S. (2006). Transcriptional regulation of ZicL in the Ciona intestinalis embryo. Development Genes and Evolution, 216(10), 597–605. Annona, G., Holland, N. D., & D’Aniello, S. (2015). Evolution of the notochord. EvoDevo, 6, 30.

The notochord gene regulatory network in chordate evolution

363

Bagheri-Fam, S., Barrionuevo, F., Dohrmann, U., G€ unther, T., Sch€ ule, R., Kemler, R., et al. (2006). Long-range upstream and downstream enhancers control distinct subsets of the complex spatiotemporal Sox9 expression pattern. Developmental Biology, 291(2), 382–397. Bell, A. H., DeMonte, F., Raza, S. M., Rhines, L. D., Tatsui, C. E., Prieto, V. G., et al. (2018). Transcriptome comparison identifies potential biomarkers of spine and skull base chordomas. Virchows Archiv, 472(3), 489–497. Bone, Q. (1992). On the locomotion of ascidian tadpole larvae. Journal of the Marine Biological Association of the United Kingdom, 72(1), 161–186. Brown, F. D., Prendergast, A., & Swalla, B. J. (2008). Man is but a worm: Chordate origins. Genesis, 46(11), 605–613. Brunet, T., Lauri, A., & Arendt, D. (2015). Did the notochord evolve from an ancient axial muscle? The axochord hypothesis. BioEssays, 37(8), 836–850. Cao, C., Lemaire, L. A., Wang, W., Yoon, P. H., Choi, Y. A., Parsons, L. R., et al. (2019). Comprehensive single-cell transcriptome lineages of a proto-vertebrate. Nature, 571(7765), 349–354. Capellini, T. D., Dunn, M. P., Passamaneck, Y. J., Selleri, L., & Di Gregorio, A. (2008). Conservation of notochord gene expression across chordates: Insights from the Leprecan gene family. Genesis, 46(11), 683–696. Capellini, T. D., Zewdu, R., Di Giacomo, G., Asciutti, S., Kugler, J. E., Di Gregorio, A., et al. (2008). Pbx1/Pbx2 govern axial skeletal development by controlling Polycomb and Hox in mesoderm and Pax1/Pax9 in sclerotome. Developmental Biology, 321(2), 500–514. Casey, E. S., O’Reilly, M. A., Conlon, F. L., & Smith, J. C. (1998). The T-box transcription factor Brachyury regulates expression of eFGF through binding to a non-palindromic response element. Development, 125(19), 3887–3894. Chen, J. S., Pedro, M. S., & Zeller, R. W. (2011). miR-124 function during Ciona intestinalis neuronal development includes extensive interaction with the Notch signaling pathway. Development, 138(22), 4943–4953. Chiba, S., Jiang, D., Satoh, N., & Smith, W. C. (2009). Brachyury null mutant-induced defects in juvenile ascidian endodermal organs. Development, 136(1), 35–39. Choi, K. S., Cohn, M. J., & Harfe, B. D. (2008). Identification of nucleus pulposus precursor cells and notochordal remnants in the mouse: Implications for disk degeneration and chordoma formation. Developmental Dynamics, 237, 3953–3958. Choi, K. S., & Harfe, B. D. (2011). Hedgehog signaling is required for formation of the notochord sheath and patterning of nuclei pulposi within the intervertebral discs. Proceedings of the National Academy of Sciences of the United States of America, 108(23), 9484–9489. Christiaen, L., Davidson, B., Kawashima, T., Powell, W., Nolla, H., Vranizan, K., et al. (2008). The transcription/migration interface in heart precursors of Ciona intestinalis. Science, 320(5881), 1349–1352. Christiaen, L., Wagner, E., Shi, W., & Levine, M. (2009). Isolation of individual cells and tissues from electroporated sea squirt (Ciona) embryos by fluorescence-activated cell sorting (FACS). Cold Spring Harbor Protocols, (12), pdb.prot5349. Cleaver, O., & Krieg, P. A. (2001). Notochord patterning of the endoderm. Developmental Biology, 234(1), 1–12. Cloney, R. A. (1964). Development of the ascidian notochord. Acta Embryologiae et Morphologiae Experimentalis, 7, 111–130. Conklin, E. G. (1905). The organization and cell-lineage of the ascidian egg. Journal of the Academy of Natural Sciences of Philadelphia, 13, 1–119. Conklin, E. G. (1928). The embryology of amphioxus. Journal of Morphology, 54, 69–151. Corallo, D., Trapani, V., & Bonaldo, P. (2015). The notochord: Structure and functions. Cellular and Molecular Life Sciences, 72(16), 2989–3008.

364

Anna Di Gregorio

Corbo, J. C., Erives, A., Di Gregorio, A., Chang, A., & Levine, M. (1997). Dorsoventral patterning of the vertebrate neural tube is conserved in a protochordate. Development, 124(12), 2335–2344. Corbo, J. C., Fujiwara, S., Levine, M., & Di Gregorio, A. (1998). Suppressor of hairless activates brachyury expression in the Ciona embryo. Developmental Biology, 203(2), 358–368. Corbo, J. C., Levine, M., & Zeller, R. W. (1997). Characterization of a notochord-specific enhancer from the Brachyury promoter region of the ascidian, Ciona intestinalis. Development, 124(3), 589–602. Darras, S., & Nishida, H. (2001). The BMP signaling pathway is required together with the FGF pathway for notochord induction in the ascidian embryo. Development, 128(14), 2629–2638. Davidson, E. H. (2006). The regulatory genome: Gene regulatory networks in development and evolution. San Diego: Academic Press. Davidson, E. H. (2010). Emerging properties of animal gene regulatory networks. Nature, 468(7326), 911–920. Davidson, E. H., & Erwin, D. H. (2006). Gene regulatory networks and the evolution of animal body plans. Science, 311(5762), 796–800. de Bree, K., de Bakker, B. S., & Oostra, R. J. (2018). The development of the human notochord. PLoS One, 13(10), e0205752. de Sousa, C. S. M., de Castro, B. B., de Miranda, C. L. V. M., Bastos, B. B., & Avelino, M. C. (2018). Split notochord syndrome: A case in point. Radiologia Brasileira, 51(1), 59–60. Dehal, P., Satou, Y., Campbell, R. K., Chapman, J., Degnan, B., De Tomaso, A., et al. (2002). The draft genome of Ciona intestinalis: Insights into chordate and vertebrate origins. Science, 298(5601), 2157–2167. Delsuc, F., Brinkmann, H., Chourrout, D., & Philippe, H. (2006). Tunicates and not cephalochordates are the closest living relatives of vertebrates. Nature, 439(7079), 965–968. Deng, W., Nies, F., Feuer, A., Bocina, I., Oliver, D., & Jiang, D. (2013). Anion translocation through an Slc26 transporter mediates lumen expansion during tubulogenesis. Proceedings of the National Academy of Sciences of the United States of America, 110(37), 14972–14977. Denker, E., Bocina, I., & Jiang, D. (2013). Tubulogenesis in a simple cell cord requires the formation of bi-apical cells through two discrete Par domains. Development, 140(14), 2985–2996. Denker, E., & Jiang, D. (2012). Ciona intestinalis notochord as a new model to investigate the cellular and molecular mechanisms of tubulogenesis. Seminars in Cell & Developmental Biology, 23(3), 308–319. Denker, E., Sehring, I. M., Dong, B., Audisso, J., Mathiesen, B., & Jiang, D. (2015). Regulation by a TGFβ-ROCK-actomyosin axis secures a non-linear lumen expansion that is essential for tubulogenesis. Development, 142(9), 1639–1650. Di Gregorio, A. (2017). T-box genes and developmental gene regulatory networks in ascidians. Current Topics in Developmental Biology, 122, 55–91. Di Gregorio, A., Corbo, J. C., & Levine, M. (2001). The regulation of forkhead/HNF-3beta expression in the Ciona embryo. Developmental Biology, 229(1), 31–43. Di Gregorio, A., Harland, R. M., Levine, M., & Casey, E. S. (2002). Tail morphogenesis in the ascidian, Ciona intestinalis, requires cooperation between notochord and muscle. Developmental Biology, 244(2), 385–395. Di Gregorio, A., & Levine, M. (1999). Regulation of Ci-tropomyosin-like, a Brachyury target gene in the ascidian, Ciona intestinalis. Development, 126(24), 5599–5609. Di Gregorio, A., & Levine, M. (2002). Analyzing gene regulation in ascidian embryos: New tools for new perspectives. Differentiation, 70(4–5), 132–139. Di Gregorio, A., Spagnuolo, A., Ristoratore, F., Pischetola, M., Aniello, F., Branno, M., et al. (1995). Cloning of ascidian homeobox genes provides evidence for a primordial chordate cluster. Gene, 156(2), 253–257.

The notochord gene regulatory network in chordate evolution

365

Dong, B., Deng, W., & Jiang, D. (2011). Distinct cytoskeleton populations and extensive crosstalk control Ciona notochord tubulogenesis. Development, 138(8), 1631–1641. Dong, B., Horie, T., Denker, E., Kusakabe, T., Tsuda, M., Smith, W. C., et al. (2009). Tube formation by complex cellular processes in Ciona intestinalis notochord. Developmental Biology, 330(2), 237–249. Dunn, M. P., & Di Gregorio, A. (2009). The evolutionarily conserved leprecan gene: Its regulation by Brachyury and its role in the developing Ciona notochord. Developmental Biology, 328(2), 561–574. Erives, A., Corbo, J. C., & Levine, M. (1998). Lineage-specific regulation of the Ciona snail gene in the embryonic mesoderm and neuroectoderm. Developmental Biology, 194(2), 213–225. Evans, A. L., Faial, T., Gilchrist, M. J., Down, T., Vallier, L., Pedersen, R. A., et al. (2012). Genomic targets of Brachyury (T) in differentiating mouse embryonic stem cells. PLoS One, 7(3), e33346. Faial, T., Bernardo, A. S., Mendjan, S., Diamanti, E., Ortmann, D., Gentsch, G. E., et al. (2015). Brachyury and SMAD signalling collaboratively orchestrate distinct mesoderm and endoderm gene regulatory networks in differentiating human embryonic stem cells. Development, 142(12), 2121–2135. Farley, E. K., Olson, K. M., Zhang, W., Rokhsar, D. S., & Levine, M. S. (2016). Syntax compensates for poor binding sites to encode tissue specificity of developmental enhancers. Proceedings of the National Academy of Sciences of the United States of America, 113(23), 6508–6513. Fujiwara, S., Maeda, Y., Shin-I, T., Kohara, Y., Takatori, N., Satou, Y., et al. (2002). Gene expression profiles in Ciona intestinalis cleavage-stage embryos. Mechanisms of Development, 112(1–2), 115–127. Giuliano, P., Marino, R., Pinto, M. R., & De Santis, R. (1998). Identification and developmental expression of Ci-isl, a homologue of vertebrate islet genes, in the ascidian Ciona intestinalis. Mechanisms of Development, 78(1–2), 199–202. Harder, M., Reeves, W., Byers, C., Santiago, M., & Veeman, M. (2019). Multiple inputs into a posterior-specific regulatory network in the Ciona notochord. Developmental Biology, 448(2), 136–146. Hartvigsen, J., Hancock, M. J., Kongsted, A., Louw, Q., Ferreira, M. L., Genevay, S., et al. (2018). What low back pain is and why we need to pay attention. Lancet, 391(10137), 2356–2367. Herwig, L., Blum, Y., Krudewig, A., Ellertsdottir, E., Lenard, A., Belting, H. G., et al. (2011). Distinct cellular mechanisms of blood vessel fusion in the zebrafish embryo. Current Biology, 21(22), 1942–1948. Horie, R., Hazbun, A., Chen, K., Cao, C., Levine, M., & Horie, T. (2018). Shared evolutionary origin of vertebrate neural crest and cranial placodes. Nature, 560(7717), 228–232. Hotta, K., Mitsuhara, K., Takahashi, H., Inaba, K., Oka, K., Gojobori, T., et al. (2007). A web-based interactive developmental table for the ascidian Ciona intestinalis, including 3D real-image embryo reconstructions: I. From fertilized egg to hatching larva. Developmental Dynamics, 236(7), 1790–1805. Hotta, K., Takahashi, H., Asakura, T., Saitoh, B., Takatori, N., Satou, Y., et al. (2000). Characterization of Brachyury-downstream notochord genes in the Ciona intestinalis embryo. Developmental Biology, 224(1), 69–80. Hotta, K., Takahashi, H., Erives, A., Levine, M., & Satoh, N. (1999). Temporal expression patterns of 39 Brachyury-downstream genes associated with notochord formation in the Ciona intestinalis embryo. Development, Growth & Differentiation, 41(6), 657–664.

366

Anna Di Gregorio

Hotta, K., Takahashi, H., Satoh, N., & Gojobori, T. (2008). Brachyury-downstream gene sets in a chordate, Ciona intestinalis: Integrating notochord specification, morphogenesis and chordate evolution. Evolution & Development, 10(1), 37–51. Hotta, K., Yamada, S., Ueno, N., Satoh, N., & Takahashi, H. (2007). Brachyurydownstream notochord genes and convergent extension in Ciona intestinalis embryos. Development, Growth & Differentiation, 49(5), 373–382. Hudson, C., Kawai, N., Negishi, T., & Yasuo, H. (2013). β-Catenin-driven binary fate specification segregates germ layers in ascidian embryos. Current Biology, 23(6), 491–495. Hudson, C., & Yasuo, H. (2005). Patterning across the ascidian neural plate by lateral Nodal signalling sources. Development, 132(6), 1199–1210. Hudson, C., & Yasuo, H. (2006). A signalling relay involving Nodal and Delta ligands acts during secondary notochord induction in Ciona embryos. Development, 133(15), 2855–2864. Ikeda, T., & Satou, Y. (2017). Differential temporal control of Foxa.a and Zic-r.b specifies brain versus notochord fate in the ascidian embryo. Development, 144(1), 38–43. Ikuta, T., Yoshida, N., Satoh, N., & Saiga, H. (2004). Ciona intestinalis Hox gene cluster: Its dispersed structure and residual colinear expression in development. Proceedings of the National Academy of Sciences of the United States of America, 101(42), 15118–15123. Imai, K. S., Hino, K., Yagi, K., Satoh, N., & Satou, Y. (2004). Gene expression profiles of transcription factors and signaling molecules in the ascidian embryo: Towards a comprehensive understanding of gene networks. Development, 131(16), 4047–4058. Imai, K. S., Levine, M., Satoh, N., & Satou, Y. (2006). Regulatory blueprint for a chordate embryo. Science, 312(5777), 1183–1187. Imai, K. S., Satoh, N., & Satou, Y. (2002). An essential role of a FoxD gene in notochord induction in Ciona embryos. Development, 129(14), 3441–3453. Imai, K. S., Satou, Y., & Satoh, N. (2002). Multiple functions of a Zic-like gene in the differentiation of notochord, central nervous system and muscle in Ciona savignyi embryos. Development, 129(11), 2723–2732. Imai, K., Takada, N., Satoh, N., & Satou, Y. (2000). (Beta)-catenin mediates the specification of endoderm cells in ascidian embryos. Development, 127(14), 3009–3020. Irvine, S. Q. (2013). Study of Cis-regulatory elements in the ascidian Ciona intestinalis. Current Genomics, 14(1), 56–67. Islam, A. F., Moly, P. K., Miyamoto, Y., & Kusakabe, T. G. (2010). Distinctive expression patterns of Hedgehog pathway genes in the Ciona intestinalis larva: Implications for a role of Hedgehog signaling in postembryonic development and chordate evolution. Zoological Science, 27(2), 84–90. Ito, M., Katada, T., Miyatani, S., & Kinoshita, T. (2007). XSu(H)2 is an essential factor for gene expression and morphogenesis of the Xenopus gastrula embryo. The International Journal of Developmental Biology, 51(1), 27–36. Jeong, Y., & Epstein, D. J. (2003). Distinct regulators of Shh transcription in the floor plate and notochord indicate separate origins for these tissues in the mouse node. Development, 130(16), 3891–3902. Jiang, D., Munro, E. M., & Smith, W. C. (2005). Ascidian prickle regulates both mediolateral and anterior-posterior cell polarity of notochord cells. Current Biology, 15(1), 79–85. Jiang, D., & Smith, W. C. (2007). Ascidian notochord morphogenesis. Developmental Dynamics, 236(7), 1748–1757. Johnson, Z. I., Shapiro, I. M., & Risbud, M. V. (2014). Extracellular osmolarity regulates matrix homeostasis in the intervertebral disc and articular cartilage: Evolving role of TonEBP. Matrix Biology, 40, 10–16. Jos
e-Edwards, D. S., Kerner, P., Kugler, J. E., Deng, W., Jiang, D., & Di Gregorio, A. (2011). The identification of transcription factors expressed in the notochord of Ciona intestinalis adds new potential players to the brachyury gene regulatory network. Developmental Dynamics, 240(7), 1793–1805.

The notochord gene regulatory network in chordate evolution

367

Jos
e-Edwards, D. S., Oda-Ishii, I., Kugler, J. E., Passamaneck, Y. J., Katikala, L., Nibu, Y., et al. (2015). Brachyury, Foxa2 and the cis-regulatory origins of the notochord. PLoS Genetics, 11(12), e1005730. Jos
e-Edwards, D. S., Oda-Ishii, I., Nibu, Y., & Di Gregorio, A. (2013). Tbx2/3 is an essential mediator within the Brachyury gene network during Ciona notochord development. Development, 140(11), 2422–2433. Karaiskou, A., Swalla, B. J., Sasakura, Y., & Chambon, J. P. (2015). Metamorphosis in solitary ascidians. Genesis, 53(1), 34–47. Katikala, L., Aihara, H., Passamaneck, Y. J., Gazdoiu, S., Jos
e-Edwards, D. S., Kugler, J. E., et al. (2013). Functional Brachyury binding sites establish a temporal read-out of gene expression in the Ciona notochord. PLoS Biology, 11(10), e1001697. Keller, R., Davidson, L., Edlund, A., Elul, T., Ezin, M., Shook, D., et al. (2000). Mechanisms of convergence and extension by cell intercalation. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 355(1399), 897–922. Keys, D. N., Lee, B. I., Di Gregorio, A., Harafuji, N., Detter, J. C., Wang, M., et al. (2005). A saturation screen for cis-acting regulatory DNA in the Hox genes of Ciona intestinalis. Proceedings of the National Academy of Sciences of the United States of America, 102(3), 679–683. Kier, W. M. (2012). The diversity of hydrostatic skeletons. The Journal of Experimental Biology, 215(Pt. 8), 1247–1257. Kispert, A., Koschorz, B., & Herrmann, B. G. (1995). The T protein encoded by Brachyury is a tissue-specific transcription factor. The EMBO Journal, 14(19), 4763–4772. Kjaer, I., Becktor, K. B., Nolting, D., & Fischer Hansen, B. (1997). The association between prenatal sella turcica morphology and notochordal remnants in the dorsum sellae. Journal of Craniofacial Genetics and Developmental Biology, 17(3), 105–111. Kourakis, M. J., Reeves, W., Newman-Smith, E., Maury, B., Abdul-Wajid, S., & Smith, W. C. (2014). A one-dimensional model of PCP signaling: Polarized cell behavior in the notochord of the ascidian Ciona. Developmental Biology, 395(1), 120–130. Kourakis, M. J., & Smith, W. C. (2005). Did the first chordates organize without the organizer? Trends in Genetics, 21(9), 506–510. Krasovec, G., Robine, K., Qu
einnec, E., Karaiskou, A., & Chambon, J. P. (2019). Ci-hox12 tail gradient precedes and participates in the control of the apoptotic-dependent tail regression during Ciona larva metamorphosis. Developmental Biology, 448(2), 237–246. Kubo, A., Suzuki, N., Yuan, X., Nakai, K., Satoh, N., Imai, K. S., et al. (2010). Genomic cis-regulatory networks in the early Ciona intestinalis embryo. Development, 137(10), 1613–1623. Kugler, J. E., Passamaneck, Y. J., Feldman, T. G., Beh, J., Regnier, T. W., & Di Gregorio, A. (2008). Evolutionary conservation of vertebrate notochord genes in the ascidian Ciona intestinalis. Genesis, 46(11), 697–710. Kugler, J. E., Wu, Y., Katikala, L., Passamaneck, Y. J., Addy, J., Caballero, N., et al. (2019). Positioning a multifunctional basic helix-loop-helix transcription factor within the Ciona notochord gene regulatory network. Developmental Biology, 448(2), 119–135. Kumano, G., Yamaguchi, S., & Nishida, H. (2006). Overlapping expression of FoxA and Zic confers responsiveness to FGF signaling to specify notochord in ascidian embryos. Developmental Biology, 300(2), 770–784. Kusakabe, T. (2005). Decoding cis-regulatory systems in ascidians. Zoological Science, 22(2), 129–146. Kusakabe, T., Yoshida, R., Kawakami, I., Kusakabe, R., Mochizuki, Y., Yamada, L., et al. (2002). Gene expression profiles in tadpole larvae of Ciona intestinalis. Developmental Biology, 242(2), 188–203. Lacalli, T. (2012). The Middle Cambrian fossil Pikaia and the evolution of chordate swimming. EvoDevo, 3(1), 12.

368

Anna Di Gregorio

Lalmansingh, A. S., Karmakar, S., Jin, Y., & Nagaich, A. K. (2012). Multiple modes of chromatin remodeling by Forkhead box proteins. Biochimica et Biophysica Acta, 1819(7), 707–715. Lauri, A., Brunet, T., Handberg-Thorsager, M., Fischer, A. H., Simakov, O., Steinmetz, P. R., et al. (2014). Development of the annelid axochord: Insights into notochord evolution. Science, 345(6202), 1365–1368. Lawson, L., & Harfe, B. D. (2015). Notochord to nucleus pulposus transition. Current Osteoporosis Reports, 13(5), 336–341. Lemaire, P. (2009). Unfolding a chordate developmental program, one cell at a time: Invariant cell lineages, short-range inductions and evolutionary plasticity in ascidians. Developmental Biology, 332(1), 48–60. Levine, M. (2010). Transcriptional enhancers in animal development and evolution. Current Biology, 20(17), R754–R763. Lin, Y., Cai, Z., Huang, S., Yang, L., Wang, C., Liu, Z., et al. (2009). Ptc, Smo, Sufu, and the Hedgehog signaling pathway in amphioxus. Evolution & Development, 11(6), 710–718. Liu, G. H., Mao, C. Z., Wu, H. Y., Zhou, D. C., Xia, J. B., Kim, S. K., et al. (2016). Expression profile of rrbp1 genes during embryonic development and in adult tissues of Xenopus laevis. Gene Expression Patterns, 23–24, 1–6. Lolas, M., Valenzuela, P. D., Tjian, R., & Liu, Z. (2014). Charting Brachyury-mediated developmental pathways during early mouse embryogenesis. Proceedings of the National Academy of Sciences of the United States of America, 111(12), 4478–4483. Lu, Q., Bhattachan, P., & Dong, B. (2019). Ascidian notochord elongation. Developmental Biology, 448(2), 147–153. Madgwick, A., Magri, M. S., Dantec, C., Gailly, D., Fiuza, U. M., Guignard, L., et al. (2019). Evolution of embryonic cis-regulatory landscapes between divergent Phallusia and Ciona ascidians. Developmental Biology, 448(2), 71–87. Maguire, J. E., Pandey, A., Wu, Y., & Di Gregorio, A. (2018). Investigating evolutionarily conserved molecular mechanisms controlling gene expression in the notochord. Advances in Experimental Medicine and Biology, 1029, 81–99. Mallatt, J., & Holland, N. (2013). Pikaia gracilens Walcott: Stem chordate, or already specialized in the Cambrian? Journal of Experimental Zoology. Part B, Molecular and Developmental Evolution, 320, 247–271. Mansfield, J. H., Haller, E., Holland, N. D., & Brent, A. E. (2015). Development of somites and their derivatives in amphioxus, and implications for the evolution of vertebrate somites. EvoDevo, 6, 21. Markovich, D., & Aronson, P. S. (2007). Specificity and regulation of renal sulfate transporters. Annual Review of Physiology, 69, 361–375. Matsumoto, J., Kumano, G., & Nishida, H. (2007). Direct activation by Ets and Zic is required for initial expression of the Brachyury gene in the ascidian notochord. Developmental Biology, 306(2), 870–882. Matsunobu, S., & Sasakura, Y. (2015). Time course for tail regression during metamorphosis of the ascidian Ciona intestinalis. Developmental Biology, 405(1), 71–81. McCann, M. R., & S
eguin, C. A. (2016). Notochord cells in intervertebral disc development and degeneration. Journal of Developmental Biology, 4(1), 3. McCann, M. R., Tamplin, O. J., Rossant, J., & S
eguin, C. A. (2012). Tracing notochordderived cells using a Noto-cre mouse: Implications for intervertebral disc development. Disease Models & Mechanisms, 5, 73–82. Meinertzhagen, I. A., & Okamura, Y. (2001). The larval ascidian nervous system: The chordate brain from its small beginnings. Trends in Neurosciences, 24(7), 401–410. Miwata, K., Chiba, T., Horii, R., Yamada, L., Kubo, A., Miyamura, D., et al. (2006). Systematic analysis of embryonic expression profiles of zinc finger genes in Ciona intestinalis. Developmental Biology, 292(2), 546–554.

The notochord gene regulatory network in chordate evolution

369

Miyamoto, D. M., & Crowther, R. J. (1985). Formation of the notochord in living ascidian embryos. Journal of Embryology and Experimental Morphology, 86, 1–17. Mizotani, Y., Suzuki, M., Hotta, K., Watanabe, H., Shiba, K., Inaba, K., et al. (2018). 14-33εa directs the pulsatile transport of basal factors toward the apical domain for lumen growth in tubulogenesis. Proceedings of the National Academy of Sciences of the United States of America, 115(38), E8873–E8881. Morley, R. H., Lachani, K., Keefe, D., Gilchrist, M. J., Flicek, P., Smith, J. C., et al. (2009). A gene regulatory network directed by zebrafish No tail accounts for its roles in mesoderm formation. Proceedings of the National Academy of Sciences of the United States of America, 106(10), 3829–3834. Morris, S. C., & Caron, J. B. (2012). Pikaia gracilens Walcott, a stem-group chordate from the Middle Cambrian of British Columbia. Biological Reviews of the Cambridge Philosophical Society, 87, 480–512. M€ uller, F., Chang, B., Albert, S., Fischer, N., Tora, L., & Str€ahle, U. (1999). Intronic enhancers control expression of zebrafish sonic hedgehog in floor plate and notochord. Development, 126(10), 2103–2116. Munro, E. M., & Odell, G. M. (2002a). Polarized basolateral cell motility underlies invagination and convergent extension of the ascidian notochord. Development, 129(1), 13–24. Munro, E. M., & Odell, G. (2002b). Morphogenetic pattern formation during ascidian notochord formation is regulative and highly robust. Development, 129(1), 1–12. Newman-Smith, E., Kourakis, M. J., Reeves, W., Veeman, M., & Smith, W. C. (2015). Reciprocal and dynamic polarization of planar cell polarity core components and myosin. eLife, 13(4), e05361. Nibu, Y., Jos
e-Edwards, D. S., & Di Gregorio, A. (2013). From notochord formation to hereditary chordoma: The many roles of Brachyury. BioMed Research International, 2013, 826435. Nieuwkoop, P. D., & Faber, J. (1967). Normal table of Xenopus laevis (Daudin). Amsterdam: North Holland. Nishida, H., & Satoh, N. (1983). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. I. Up to the eight-cell stage. Developmental Biology, 99, 382–394. Nishida, H., & Satoh, N. (1985). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. II. The 16- and 32-cell stages. Developmental Biology, 110, 440–454. Nishikata, T., Yamada, L., Mochizuki, Y., Satou, Y., Shin-i, T., Kohara, Y., et al. (2001). Profiles of maternally expressed genes in fertilized eggs of Ciona intestinalis. Developmental Biology, 238(2), 315–331. Nitta, K. R., Vincentelli, R., Jacox, E., Cimino, A., Ohtsuka, Y., Sobral, D., et al. (2019). High-throughput protein production combined with high- throughput SELEX identifies an extensive atlas of Ciona robusta transcription factor DNA-binding specificities. Methods in Molecular Biology, 2025, 487–517. Noda, T. (2011). The maternal genes Ci-p53/p73-a and Ci-p53/p73-b regulate zygotic ZicL expression and notochord differentiation in Ciona intestinalis embryos. Developmental Biology, 360(1), 216–229. Oda-Ishii, I., & Di Gregorio, A. (2007). Lineage-independent mosaic expression and regulation of the Ciona multidom gene in the ancestral notochord. Developmental Dynamics, 236(7), 1806–1819. Oda-Ishii, I., Ishii, Y., & Mikawa, T. (2010). Eph regulates dorsoventral asymmetry of the notochord plate and convergent extension-mediated notochord formation. PLoS One, 5(10), e13689. Ogasawara, M., Sasaki, A., Metoki, H., Shin-i, T., Kohara, Y., Satoh, N., et al. (2002). Gene expression profiles in young adult Ciona intestinalis. Development Genes and Evolution, 212(4), 173–185.

370

Anna Di Gregorio

Ortolani, G. (1954). Risultati definitivi sulla distribuzione dei territory presuntivi degli organi nel germe di Ascidie allo stadio VIII, determinati con le marche al carbone. Pubblicazioni della Stazione Zoologica di Napoli. Section II: History and Philosophy of the Life Sciences, 25, 161–187. Pascual-Anaya, J., Adachi, N., Alvarez, S., Kuratani, S., D’Aniello, S., & Garcia-Ferna`ndez, J. (2012). Broken colinearity of the amphioxus Hox cluster. EvoDevo, 3(1), 28. Passamaneck, Y. J., & Di Gregorio, A. (2005). Ciona intestinalis: Chordate development made simple. Developmental Dynamics, 233(1), 1–19. Passamaneck, Y. J., Katikala, L., Perrone, L., Dunn, M. P., Oda-Ishii, I., & Di Gregorio, A. (2009). Direct activation of a notochord cis-regulatory module by Brachyury and FoxA in the ascidian Ciona intestinalis. Development, 136(21), 3679–3689. Pennati, R., Ficetola, G. F., Brunetti, R., Caicci, F., Gasparini, F., Griggio, F., et al. (2015). Morphological differences between larvae of the Ciona intestinalis species complex: Hints for a valid taxonomic definition of distinct species. PLoS One, 10(5), e0122879. Peter, I. S. (2017). Regulatory states in the developmental control of gene expression. Briefings in Functional Genomics, 16(5), 281–287. Peter, I. S., & Davidson, E. H. (2015). Genomic control process, development and evolution. Amsterdam: Elsevier. Peterson, K. J., Cameron, R. A., Tagawa, K., Satoh, N., & Davidson, E. H. (1999). A comparative molecular approach to mesodermal patterning in basal deuterostomes: The expression pattern of Brachyury in the enteropneust hemichordate Ptychodera flava. Development, 126(1), 85–95. Pfister, K., Shook, D. R., Chang, C., Keller, R., & Skoglund, P. (2016). Molecular model for force production and transmission during vertebrate gastrulation. Development, 143(4), 715–727. Phillips, R., Kondev, J., Theriot, J., & Garcia, H. (2012). Physical biology of the cell (2nd ed.). New York: Garland Science, Taylor and Francis Group LLC. Picco, V., Hudson, C., & Yasuo, H. (2007). Ephrin-Eph signalling drives the asymmetric division of notochord/neural precursors in Ciona embryos. Development, 134(8), 1491–1497. Racioppi, C., Wiechecki, K. A., & Christiaen, L. (2019). Combinatorial chromatin dynamics foster accurate cardiopharyngeal fate choices. Elife, 8, e49921. Rajesh, D., & Dahia, C. L. (2018). Role of sonic hedgehog signaling pathway in intervertebral disc formation and maintenance. Current Molecular Biology Reports, 4(4), 173–179. Rastegar, S., Hess, I., Dickmeis, T., Nicod, J. C., Ertzer, R., Hadzhiev, Y., et al. (2008). The words of the regulatory code are arranged in a variable manner in highly conserved enhancers. Developmental Biology, 318(2), 366–377. Razy-Krajka, F., Lam, K., Wang, W., Stolfi, A., Joly, M., Bonneau, R., et al. (2014). Collier/ OLF/EBF-dependent transcriptional dynamics control pharyngeal muscle specification from primed cardiopharyngeal progenitors. Developmental Cell, 29(3), 263–276. Razy-Krajka, F., & Stolfi, A. (2019). Regulation and evolution of muscle development in tunicates. EvoDevo, 10, 13. Reese, D. E., Hall, C. E., & Mikawa, T. (2004). Negative regulation of midline vascular development by the notochord. Developmental Cell, 6(5), 699–708. Reeves, W., Thayer, R., & Veeman, M. (2014). Anterior-posterior regionalized gene expression in the Ciona notochord. Developmental Dynamics, 243(4), 612–620. Reeves, W. M., Wu, Y., Harder, M. J., & Veeman, M. T. (2017). Functional and evolutionary insights from the Ciona notochord transcriptome. Development, 144(18), 3375–3387. Rhee, J. M., Oda-Ishii, I., Passamaneck, Y. J., Hadjantonakis, A. K., & Di Gregorio, A. (2005). Live imaging and morphometric analysis of embryonic development in the ascidian Ciona intestinalis. Genesis, 43(3), 136–147.

The notochord gene regulatory network in chordate evolution

371

Sasaki, H., & Hogan, B. L. (1993). Differential expression of multiple fork head related genes during gastrulation and axial pattern formation in the mouse embryo. Development, 118(1), 47–59. Sasaki, H., Yoshida, K., Hozumi, A., & Sasakura, Y. (2014). CRISPR/Cas9-mediated gene knockout in the ascidian Ciona intestinalis. Development, Growth & Differentiation, 56(7), 499–510. Satoh, N. (2001). Ascidian embryos as a model system to analyze expression and function of developmental genes. Differentiation, 68(1), 1–12. Satoh, N. (2014). Developmental genomics of ascidians. Hoboken: Wiley-Blackwell. Satoh, N., Tagawa, K., Lowe, C. J., Yu, J. K., Kawashima, T., Takahashi, H., et al. (2014). On a possible evolutionary link of the stomochord of hemichordates to pharyngeal organs of chordates. Genesis, 52(12), 925–934. Satoh, N., Tagawa, K., & Takahashi, H. (2012). How was the notochord born? Evolution & Development, 14(1), 56–75. Satou, Y. (2020). A gene regulatory network for cell fate specification in Ciona embryos. Current Topics in Developmental Biology, 139, 1–33. Satou, Y., Kawashima, T., Shoguchi, E., Nakayama, A., & Satoh, N. (2005). An integrated database of the ascidian, Ciona intestinalis: Towards functional genomics. Zoological Science, 22(8), 837–843. Satou, Y., Nakamura, R., Yu, D., Yoshida, R., Hamada, M., Fujie, M., et al. (2019). A nearly complete genome of Ciona intestinalis type a (C. robusta) reveals the contribution of inversion to chromosomal evolution in the genus Ciona. Genome Biology and Evolution, 11(11), 3144–3157. Satou, Y., & Satoh, N. (2005). Cataloging transcription factor and major signaling molecule genes for functional genomic studies in Ciona intestinalis. Development Genes and Evolution, 215(11), 580–596. Satou, Y., Takatori, N., Fujiwara, S., Nishikata, T., Saiga, H., Kusakabe, T., et al. (2002). Ciona intestinalis cDNA projects: Expressed sequence tag analyses and gene expression profiles during embryogenesis. Gene, 287(1–2), 83–96. Satou, Y., Takatori, N., Yamada, L., Mochizuki, Y., Hamaguchi, M., Ishikawa, H., et al. (2001). Gene expression profiles in Ciona intestinalis tailbud embryos. Development, 128(15), 2893–2904. Satou, Y., Wada, S., Sasakura, Y., & Satoh, N. (2008). Regulatory genes in the ancestral chordate genomes. Development Genes and Evolution, 218(11 12), 715–721. Sawada, A., Nishizaki, Y., Sato, H., Yada, Y., Nakayama, R., et al. (2005). Tead proteins activate the Foxa2 enhancer in the node in cooperation with a second factor. Development, 132, 4719–4729. Segade, F., Cota, C., Famiglietti, A., Cha, A., & Davidson, B. (2016). Fibronectin contributes to notochord intercalation in the invertebrate chordate, Ciona intestinalis. EvoDevo, 7(1) 21. Sehring, I. M., Dong, B., Denker, E., Bhattachan, P., Deng, W., Mathiesen, B. T., et al. (2014). An equatorial contractile mechanism drives cell elongation but not cell division. PLoS Biology, 12(2), e1001781. Sehring, I. M., Recho, P., Denker, E., Kourakis, M., Mathiesen, B., Hannezo, E., et al. (2015). Assembly and positioning of actomyosin rings by contractility and planar cell polarity. eLife, 21(4). e09206. Seo, H. C., Edvardsen, R. B., Maeland, A. D., Bjordal, M., Jensen, M. F., Hansen, A., et al. (2004). Hox cluster disintegration with persistent anteroposterior order of expression in Oikopleura dioica. Nature, 431(7004), 67–71. Sheyn, D., Ben-David, S., Tawackoli, W., Zhou, Z., Salehi, K., Bez, M., et al. (2019). Human iPSCs can be differentiated into notochordal cells that reduce intervertebral disc degeneration in a porcine model. Theranostics, 9(25), 7506–7524.

372

Anna Di Gregorio

Shi, W., & Levine, M. (2008). Ephrin signaling establishes asymmetric cell fates in an endomesoderm lineage of the Ciona embryo. Development, 135(5), 931–940. Shi, W., Peyrot, S. M., Munro, E., & Levine, M. (2009). FGF3 in the floor plate directs notochord convergent extension in the Ciona tadpole. Development, 136(1), 23–28. Shimeld, S. M. (1999). The evolution of the hedgehog gene family in chordates: Insights from amphioxus hedgehog. Development Genes and Evolution, 209(1), 40–47. Sivan, S. S., Hayes, A. J., Wachtel, E., Caterson, B., Merkher, Y., Maroudas, A., et al. (2014). Biochemical composition and turnover of the extracellular matrix of the normal and degenerate intervertebral disc. European Spine Journal, 23(Suppl. 3), S344–S353. Smith, W. C. (2018). Cellular processes of notochord formation. Advances in Experimental Medicine and Biology, 1029, 165–177. Søviknes, A. M., & Glover, J. C. (2008). Continued growth and cell proliferation into adulthood in the notochord of the appendicularian Oikopleura dioica. The Biological Bulletin, 214(1), 17–28. Spagnuolo, A., Ristoratore, F., Di Gregorio, A., Aniello, F., Branno, M., & Di Lauro, R. (2003). Unusual number and genomic organization of Hox genes in the tunicate Ciona intestinalis. Gene, 309(2), 71–79. Stemple, D. L. (2005). Structure and function of the notochord: An essential organ for chordate development. Development, 132, 2503–2512. Stolfi, A., & Christiaen, L. (2012). Genetic and genomic toolbox of the chordate Ciona intestinalis. Genetics, 192(1), 55–66. Stolfi, A., Gandhi, S., Salek, F., & Christiaen, L. (2014). Tissue-specific genome editing in Ciona embryos by CRISPR/Cas9. Development, 141(21), 4115–4120. Stott, D., Kispert, A., & Herrmann, B. G. (1993). Rescue of the tail defect of Brachyury mice. Genes & Development, 7(2), 197–203. Sulik, K., Dehart, D. B., Iangaki, T., Carson, J. L., Vrablic, T., Gesteland, K., et al. (1994). Morphogenesis of the murine node and notochordal plate. Developmental Dynamics, 201(3), 260–278. Suzuki, M. M., & Satoh, N. (2000). Genes expressed in the amphioxus notochord revealed by EST analysis. Developmental Biology, 224(2), 168–177. Tada, M., Casey, E. S., Fairclough, L., & Smith, J. C. (1998). Bix1, a direct target of Xenopus T-box genes, causes formation of ventral mesoderm and endoderm. Development, 125(20), 3997–4006. Tajima, Y., Hozumi, A., Yoshida, K., Treen, N., Sakuma, T., Yamamoto, T., et al. (2020). Hox13 is essential for formation of a sensory organ at the terminal end of the sperm duct in Ciona. Developmental Biology, 458(1), 120–131. Takahashi, H., Hotta, K., Erives, A., Di Gregorio, A., Zeller, R. W., Levine, M., et al. (1999). Brachyury downstream notochord differentiation in the ascidian embryo. Genes & Development, 13(12), 1519–1523. Takahashi, H., Hotta, K., Takagi, C., Ueno, N., Satoh, N., & Shoguchi, E. (2010). Regulation of notochord-specific expression of Bra downstream genes in Ciona intestinalis embryos. Zoological Science, 27(2), 110–118. Takatori, N., Hotta, K., Mochizuki, Y., Satoh, G., Mitani, Y., Satoh, N., et al. (2004). T-box genes in the ascidian Ciona intestinalis: Characterization of cDNAs and spatial expression. Developmental Dynamics, 230(4), 743–753. Takatori, N., Satou, Y., & Satoh, N. (2002). Expression of hedgehog genes in Ciona intestinalis embryos. Mechanisms of Development, 116(1–2), 235–238. Talbot, W. S., Trevarrow, B., Halpern, M. E., Melby, A. E., Farr, G., Postlethwait, J. H., et al. (1995). A homeobox gene essential for zebrafish notochord development. Nature, 378(6553), 150–157.

The notochord gene regulatory network in chordate evolution

373

Tamplin, O. J., Cox, B. J., & Rossant, J. (2011). Integrated microarray and ChIP analysis identifies multiple Foxa2 dependent target genes in the notochord. Developmental Biology, 360(2), 415–425. Tamplin, O. J., Kinzel, D., Cox, B. J., Bell, C. E., Rossant, J., & Lickert, H. (2008). Microarray analysis of Foxa2 mutant mouse embryos reveals novel gene expression and inductive roles for the gastrula organizer and its derivatives. BMC Genomics, 9, 511. Tang, R., Jing, L., Willard, V. P., Wu, C. L., Guilak, F., Chen, J., et al. (2018). Differentiation of human induced pluripotent stem cells into nucleus pulposus-like cells. Stem Cell Research & Therapy, 9(1) 61. Thompson, J. M., & Di Gregorio, A. (2015). Insulin-like genes in ascidians: Findings in Ciona and hypotheses on the evolutionary origins of the pancreas. Genesis, 53(1), 82–104. Tokuhiro, S. I., Tokuoka, M., Kobayashi, K., Kubo, A., Oda-Ishii, I., & Satou, Y. (2017). Differential gene expression along the animal-vegetal axis in the ascidian embryo is maintained by a dual functional protein Foxd. PLoS Genetics, 13(5), e1006741. Urano, A., Suzuki, M. M., Zhang, P., Satoh, N., & Satoh, G. (2003). Expression of musclerelated genes and two MyoD genes during amphioxus notochord development. Evolution & Development, 5(5), 447–458. Urry, L. A., Whittaker, C. A., Duquette, M., Lawler, J., & DeSimone, D. W. (1998). Thrombospondins in early Xenopus embryos: Dynamic patterns of expression suggest diverse roles in nervous system, notochord, and muscle development. Developmental Dynamics, 211, 390–407. Utsumi, N., Shimojima, Y., & Saiga, H. (2004). Analysis of ascidian Not genes highlights their evolutionarily conserved and derived features of structure and expression in development. Development Genes and Evolution, 214(9), 460–465. Valenta, T., Hausmann, G., & Basler, K. (2012). The many faces and functions of β-catenin. The EMBO Journal, 31(12), 2714–2736. Veeman, M. T., Nakatani, Y., Hendrickson, C., Ericson, V., Lin, C., & Smith, W. C. (2008). Chongmague reveals an essential role for laminin-mediated boundary formation in chordate convergence and extension movements. Development, 135(1), 33–41. Veeman, M. T., & Smith, W. C. (2013). Whole-organ cell shape analysis reveals the developmental basis of ascidian notochord taper. Developmental Biology, 373(2), 281–289. Velandia-Huerto, C. A., Brown, F. D., Gittenberger, A., Stadler, P. F., & Bermu´dezSantana, C. I. (2018). Nonprotein-coding RNAs as regulators of development in tunicates. Results and Problems in Cell Differentiation, 65, 197–225. Wada, H., Okuyama, M., Satoh, N., & Zhang, S. (2006). Molecular evolution of fibrillar collagen in chordates, with implications for the evolution of vertebrate skeletons and chordate phylogeny. Evolution & Development, 8, 370–377. Wang, W., & Christiaen, L. (2012). Transcriptional enhancers in ascidian development. Current Topics in Developmental Biology, 98, 147–172. Wang, S., Furmanek, T., Kryvi, H., Krossøy, C., Totland, G. K., Grotmol, S., et al. (2014). Transcriptome sequencing of Atlantic salmon (Salmo salar L.) notochord prior to development of the vertebrae provides clues to regulation of positional fate, chordoblast lineage and mineralisation. BMC Genomics, 15, 141. Ward, L., Pang, A. S. W., Evans, S. E., & Stern, C. D. (2018). The role of the notochord in amniote vertebral column segmentation. Developmental Biology, 439(1), 3–18. Wilkinson, D. G., Bhatt, S., & Herrmann, B. G. (1990). Expression pattern of the mouse T gene and its role in mesoderm formation. Nature, 343(6259), 657–659. Winkley, K., Ward, S., Reeves, W., & Veeman, M. (2019). Iterative and complex asymmetric divisions control cell volume differences in Ciona notochord tapering. Current Biology, 29(20), 3466–3477.

374

Anna Di Gregorio

Yagi, K., Satou, Y., & Satoh, N. (2004). A zinc finger transcription factor, ZicL, is a direct activator of Brachyury in the notochord specification of Ciona intestinalis. Development, 131(6), 1279–1288. Yamada, S., Ueno, N., Satoh, N., & Takahashi, H. (2011). Ciona intestinalis n4 contains a phosphotyrosine interaction domain and is involved in the midline intercalation of notochord cells. The International Journal of Developmental Biology, 55(1), 11–18. Yasuo, H., & Hudson, C. (2007). FGF8/17/18 functions together with FGF9/16/20 during formation of the notochord in Ciona embryos. Developmental Biology, 302(1), 92–103. Zizic Mitrecic, M., Mitrecic, D., Pochet, R., Kostovic-Knezevic, L., & Gajovic, S. (2010). The mouse gene Noto is expressed in the tail bud and essential for its morphogenesis. Cells, Tissues, Organs, 192(2), 85–92.

CHAPTER ELEVEN

The notochord gene regulatory network in chordate evolution: Conservation and divergence from Ciona to vertebrates Anna Di Gregorio∗ Department of Basic Science and Craniofacial Biology, New York University College of Dentistry, New York, NY, United States *Corresponding author: e-mail address: [email protected]

Contents 1. The notochord, a synapomorphy of the chordate phylum and a sine qua non of chordate development 2. Evolutionary history of the notochord 3. Notochord development in Ciona 3.1 Morphogenetic milestones 3.2 The molecular blueprint for notochord formation in Ciona 4. Time is of the essence: Temporal cis-regulatory control of gene expression in a fast-developing chordate 5. The GRN underlying notochord formation in Ciona 5.1 The main regulatory nodes of the notochord GRN: Brachyury and Foxa2 5.2 Drawing edges, connecting nodes: Reconstructing the gene batteries controlled by notochord transcription factors 6. Reverse-engineering the notochord GRN: Lessons from notochord cis-regulatory modules 6.1 Ciona notochord CRMs can be synergistically activated by Foxa.a in cooperation with either Bra or other transcription factors 6.2 Additional notochord cis-regulatory mechanisms identified in Ciona 6.3 Comparative studies of notochord cis-regulatory regions across chordates 7. A comparative view of the notochord GRN in Ciona and other chordates 8. Concluding remarks and future perspectives Acknowledgments References

326 331 332 332 337 344 346 348 350 356 357 357 358 359 361 362 362

Abstract The notochord is a structure required for support and patterning of all chordate embryos, from sea squirts to humans. An increasing amount of information on notochord development and on the molecular strategies that ensure its proper morphogenesis has been Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.01.002

#

2020 Elsevier Inc. All rights reserved.

325

326

Anna Di Gregorio

gleaned through studies in the sea squirt Ciona. This invertebrate chordate offers a fortunate combination of experimental advantages, ranging from translucent, fast-developing embryos to a compact genome and impressive biomolecular resources. These assets have enabled the rapid identification of numerous notochord genes and cis-regulatory regions, and provide a rather unique opportunity to reconstruct the gene regulatory network that controls the formation of this developmental and evolutionary chordate landmark. This chapter summarizes the morphogenetic milestones that punctuate notochord formation in Ciona, their molecular effectors, and the current knowledge of the gene regulatory network that ensures the accurate spatial and temporal orchestration of these processes.

Abbreviations ATAC-Seq bp bHLH cDNA ChIP ChIP-chip CRM DAPI EMSA ENU FACS GFP GRN GSK3 hpf hr(s) iPSCs kb Mb mya NFAT5 RNA-Seq scRNA-Seq WMISH

assay for transposase-accessible chromatin with high-throughput sequencing base pair (s) basic helix-loop-helix complementary DNA chromatin immunoprecipitation chromatin immunoprecipitation on DNA microarray cis-regulatory module 40 ,6-diamidino-2-phenylindole electrophoretic mobility shift assay N-ethyl-N-nitrosourea fluorescence-activated cell sorting green fluorescent protein gene regulatory network glycogen synthase kinase 3 hours post-fertilization hour(s) induced pluripotent stem cells kilobase(s), or 1000 base pairs megabase(s), one million base pairs million years ago nuclear factor of activated T cells 5 RNA-Sequencing single-cell RNA-Sequencing whole-mount in situ hybridization(s)

1. The notochord, a synapomorphy of the chordate phylum and a sine qua non of chordate development The chordate phylum is a large division of the animal kingdom that includes three subphyla of widely different organisms, ranging from marine

The notochord gene regulatory network in chordate evolution

327

invertebrate chordates, such as lancelets and sea squirts, to humans and all other vertebrates. All chordate embryos share, among a few other hallmarks, a defining structural feature, the notochord (chorda dorsalis). The notochord consists of an axial rod of cells of mesodermal origin that support the developing embryo, which in vertebrates is almost completely replaced, as embryogenesis proceeds, by the developing vertebral column. The simplest arrangement of the notochord described to date in an extant chordate is found in ascidians, where the notochord is composed of 40 cells aligned in single file in the center of the tail (circled in red in Fig. 1A). In a cross-section of the ascidian tail, the central notochord cell is flanked on each side by large muscle cells (orange ovals in Fig. 1A), and it adjoins the smaller, dorsally located neural tube, which in cross-section is made of only four cells (circled in blue in Fig. 1A), and the ventral endodermal strand (Fig. 1A, yellow circle). The relative sizes of notochord and neural tube vary considerably among different species of vertebrates. In cross-sectioned Xenopus embryos, the notochord is multicellular and highly vacuolated (circled in red in Fig. 1B), and, as in ascidian embryos, it is larger than its overlying neural tube (circled in blue in Fig. 1B). Instead, in mouse embryos, the notochord is much smaller than the neural tube (circled in red and blue in Fig. 1C, respectively). In human embryos as well, the notochord is considerably smaller (Fig. 1D, red circle) and contains less cells than the neural tube. In all chordates, the notochord is found in very close connection with the ventral-most region of the neural tube, the floor plate, and in most chordates the notochord plays an inductive role on the formation of this structure and on the dorsal-ventral regionalization of the neural tube. In humans and most other vertebrates, the notochord is also in contact with the bilaterally located somites and with the roof of the gut, which is located ventrally (Fig. 1D). The sclerotomes of the somites give rise to the vertebral bodies, and although the establishment of the segmental pattern of the vertebrae follows different modalities between amniotes and teleosts, these patterning events are strictly reliant upon the notochord (e.g., Ward, Pang, Evans, & Stern, 2018). In human embryos, the notochord starts to form around day 17 of gestation (Carnegie Stage 7) as a notochordal process, and the definitive notochord is completed between day 23 and day 30 (Carnegie Stages 10–12) (de Bree, de Bakker, & Oostra, 2018). After gastrulation, the notochordal process is incorporated into the endoderm of the roof of the developing gut, and remains attached ventrally and laterally to the endoderm as an epithelial-like formation, the notochordal plate. The dorsal aspects of these structures, as well as the definitive notochord in the initial steps of its formation, are closely adjoined to the floor of the developing

328

Anna Di Gregorio

Fig. 1 The notochord and its surrounding structures in divergent chordates. Cross-sectional views of the notochord and its neighboring structures in representative chordate embryos. (A) Microphotograph of a sectional view of the tail of a Ciona embryo. The notochord spans a single cell-diameter and, in this posterior region of the tail, is flanked by two muscle cells on each side. (B) Low-magnification microphotograph of a cross-section of the tail of a Xenopus laevis tadpole (NF stage 45; Nieuwkoop & Faber, 1967), embedded in paraffin and stained by hematoxylin and eosin. The notochord (circled in red) is multicellular and the cells show large vacuoles. Kindly provided by Dr. Jean-Pierre Saint-Jeannet, New York University College of Dentistry. (C) Microphotograph of a cross-section of an E10.5 mouse embryo hybridized in situ with an antisense RNA probe for Sonic hedgehog (partial view) (Capellini, Dunn, Passamaneck, Selleri, & Di Gregorio, 2008). (D) Scanning electron microphotograph of a human embryo, Carnegie stage 11. The notochord (circled in red) is in direct contact with the neural tube dorsal to it, with somites on each side, and with the endoderm, located ventrally. Kindly provided by Prof. Kathleen Sulik, University of North Carolina. Red circles, notochord; orange ovals, muscle cells; blue ovals, neural tube; yellow circle, endodermal cells. In (D), the neural tube is not circled to preserve the image’s details. Scale bars: 10 μm in (A); 100 μm in (B–D).

The notochord gene regulatory network in chordate evolution

329

neural tube (de Bree et al., 2018). As the notochord becomes part of the mesodermal mesenchyme, it gradually separates from the endoderm, though maintaining contact with the oropharynx cranially and with the hindgut caudally; eventually, it detaches from the neural tube, at least in its cranial-most region (de Bree et al., 2018). The close relationship between the human notochord and its surrounding structures and tissues (Fig. 1D) explains the severity of notochord-derived birth defects. Spina bifida, the most frequent permanently disabling birth defect, and other congenital malformations, including encephalocele and anencephaly, are attributed to impaired signaling between notochord, sclerotome and ectoderm (Kjaer, Becktor, Nolting, & Fischer Hansen, 1997), along with the much rarer group of birth defects collectively known as “split notochord syndrome,” which are characterized by either a partial duplication or by a bifurcation of the lumbosacral region of the spine (de Sousa, de Castro, de Miranda, Bastos, & Avelino, 2018). Similarly, mouse embryos in which notochord formation is impaired as a result of a mutation in the HNF-3β locus (hereinafter referred to as Foxa2) fail to develop a floor plate and therefore lack essential signals required for axon guidance (Ang & Rossant, 1994). In addition to patterning the neural tube, in vertebrates the notochord exchanges developmental cues with the paraxial mesoderm and with the somites that derive from it, and is required for the proper morphogenesis of endodermal derivatives (Cleaver & Krieg, 2001; Corallo, Trapani, & Bonaldo, 2015; Stemple, 2005). Experiments carried out in quail embryos have also uncovered the existence of notochord-derived instructive signals that prevent the formation of blood vessels along the embryonic midline (Reese, Hall, & Mikawa, 2004). In invertebrate chordates, the aquatic Tunicates, or Urochordates, (ascidians, larvaceans and salps) and Cephalochordates (amphioxus), the notochord remains the main structural support for the developing embryo. In tunicates with a dual life-cycle, such as the ascidian Ciona, the notochord disappears at the time of metamorphosis during the process of tail retraction. In other tunicates, such as the larvacean Oikopleura, the notochord persists throughout the life of the animal, and is composed of cells that continue to divide until they add up to 120–160 in total (Almaza´n, Ferra´ndezRolda´n, Albalat, & Can˜estro, 2019; Søviknes & Glover, 2008). In vertebrates, as the notochord is gradually replaced by the vertebrae, its remnants constitute the nuclei pulposi, the innermost compartment of the intervertebral discs (Lawson & Harfe, 2015). The long-debated notochordal origin of the nuclei pulposi was confirmed in mouse embryos through

330

Anna Di Gregorio

lineage-tracing experiments that followed notochord cells expressing the LacZ reporter gene to their final location in the central-most region of the intervertebral discs (Choi, Cohn, & Harfe, 2008; McCann, Tamplin, Rossant, & S
eguin, 2012). Within the nuclei pulposi, notochord cells secrete extracellular matrix (ECM) molecules, which form a gelatinous, highly hydrated cushioning structure that contains a loose network of proteoglycans held together by type II collagen and elastin fibers (Sivan et al., 2014). Hence, the secretory activity of the notochord cells is mainly responsible for the shock-absorbing properties of the intervertebral discs, which are necessary for the movements and flexibility of the backbone. A transcription factor expressed in nuclei pulposi, NFAT5, ensures that the correct osmotic equilibrium is maintained within these gelatinous structures by activating the expression of ECM components and matrix homeostasis genes, such as aquaporins and solute transporters, in response to hyperosmotic stress ( Johnson, Shapiro, & Risbud, 2014). The nuclei pulposi are encased by the rigid annuli fibrosi, which are composed of bundles of type I collagen fibers surrounded by elastic fibers; this structural arrangement enables them to counteract the pressure exerted by the turgor of the nuclei pulposi. The nuclei pulposi are avascular, and receive nutrients from the vertebral bodies, to which they are anchored by means of the cartilage end plates (McCann & S
eguin, 2016). Degeneration of the notochord cells in the nuclei pulposi causes the onset of intervertebral disc degeneration and consequent back pain, one of the most common afflictions and the leading cause of disability in the adult population worldwide (Hartvigsen et al., 2018). For these reasons, therapeutic strategies are aimed at restoring the supply of healthy notochord cells in the nuclei pulposi, for example, by treating human induced pluripotent stem cells (iPSCs) with GSK3 inhibitor to program them toward a mesodermal fate, and by subsequently transfecting them with plasmids overexpressing Brachyury (Bra), the main marker and inducer of notochord cell fate (Kispert, Koschorz, & Herrmann, 1995; Sheyn et al., 2019; Tang et al., 2018). Occasionally, notochord remnants can give rise to chordomas, extremely rare tumors that affect approximately one individual in a million, and nevertheless account for 20% of the primary tumors of the spine (reviewed in Nibu, Jos
e-Edwards, & Di Gregorio, 2013). The paramount importance of the notochord for embryonic development and the crucial role of its deriving cells in adult posture, locomotion and disease, motivate the significant interest in studies of the gene regulatory network (GRN) responsible for its formation. Here we review

The notochord gene regulatory network in chordate evolution

331

the information that has been gathered on notochord formation in an invertebrate chordate, the tunicate Ciona, and the progress made on the reconstruction of its notochord GRN.

2. Evolutionary history of the notochord Fossil records have traced the origins of the notochord to the Middle Cambrian (510–495 mya), and the extinct eel-shaped Pikaia gracilens is still argued to be either a basal chordate or a more specialized, divergent one (Lacalli, 2012; Mallatt & Holland, 2013; Morris & Caron, 2012). Structures that could represent evolutionary precursors of the notochord have been sought in non-chordate phyla; the stomochord of different species of hemichordates (acorn worms) has been repeatedly probed for notochord marker genes, and is currently considered more closely related to chordate organs of pharyngeal origin than to an ancestral notochord (Peterson, Cameron, Tagawa, Satoh, & Davidson, 1999; Satoh et al., 2014). More recently, a new candidate ancestor of the notochord has been tentatively identified through studies of the axochord, a muscular structure found in the ventral midline of the annelid worms Platynereis dumerilii and Capitella teleta, which, unlike the hemichordate stomochord, expresses a Bra ortholog and other molecular identifiers of notochord cells (Brunet, Lauri, & Arendt, 2015; Lauri et al., 2014). Different theories on chordate ancestry and on the origins of the notochord, as either a chordate-specific structure that appeared exclusively in this phylum, or as a chordate-specific modification of an ancestral non-chordate organ, are extensively discussed in detailed reviews (e.g., Annona, Holland, & D’Aniello, 2015; Brown, Prendergast, & Swalla, 2008; Satoh, Tagawa, & Takahashi, 2012). Among extant chordates, Cephalochordates, which include a few genera of amphioxus (lancelet), are considered the most basal subphylum of chordates (Delsuc, Brinkmann, Chourrout, & Philippe, 2006). The amphioxus notochord is a permanent structure and is flanked by somites, similarly to the vertebrate notochord. However, unlike the notochord of ascidians and vertebrates, the notochord of amphioxus is composed, in cross-section, of multiple cells that belong to two different types. Contractile notochord cells contain myofibrils, express muscle-related genes and occupy the center of the notochord, while non-contractile M€ uller cells are located peripherally and express a fibrillar collagen gene, ColA, along with a subset of muscle genes, such as calponin (Mansfield, Haller, Holland, & Brent, 2015; Urano, Suzuki, Zhang, Satoh, & Satoh, 2003). Quantitative studies of the

332

Anna Di Gregorio

complement of genes expressed in the amphioxus notochord indicate that 11% of the transcripts found in its cells are muscle-related genes, while 6% are ECM genes that participate in the formation of the notochordal sheath that envelopes the notochord (Suzuki & Satoh, 2000). On the other hand, the ascidian notochord is composed of cells that resemble a primitive type of cartilage (Stemple, 2005).

3. Notochord development in Ciona In this section we summarize the main morphogenetic milestones that define notochord formation in the ascidian Ciona. We mostly refer to studies carried out in Ciona robusta (formerly Ciona intestinalis type A; Pennati et al., 2015), and its distant relative Ciona savignyi. We review the cellular events that characterize each of these developmental landmarks, the pivotal molecular events underlying them, and the current knowledge of their evolutionary conservation across chordates. Even though Ciona intestinalis type A has been recently renamed Ciona robusta and several ascidian gene names have been updated, to prevent confusion we will refer, in some instances, to the gene names as they were reported in their original publications.

3.1 Morphogenetic milestones Notochord development has been studied and visualized in ample detail in ascidian embryos, by reason of their year-round availability, abundance, fast development, invariant cell-lineage and translucency (e.g., Denker & Jiang, 2012; Lemaire, 2009; Maguire, Pandey, Wu, & Di Gregorio, 2018; Passamaneck & Di Gregorio, 2005; Satoh, 2001). At this point in time, most imaging studies of notochord development have been carried out in Ciona robusta and Ciona savignyi. The process of notochord formation has been largely investigated in these species, first through the use of bright-field time-lapse imaging and microphotography (Miyamoto & Crowther, 1985), and in recent times through the use of confocal time-lapse imaging, 3-D reconstructions, and morphometric analyses (e.g., Denker et al., 2015; Denker, Bocina, & Jiang, 2013; Denker & Jiang, 2012; Jiang & Smith, 2007; Rhee, Oda-Ishii, Passamaneck, Hadjantonakis, & Di Gregorio, 2005; Smith, 2018). In Ciona and other ascidians, the fully developed notochord consists of 40 post-mitotic cells, 32 of which are grouped in the anterior-most “primary,” or “anterior,” or “A-line” notochord, while the remaining eight are located more posteriorly and are known as “secondary,” or “posterior” or “B-line” notochord (Satou, 2020).

The notochord gene regulatory network in chordate evolution

333

These terms refer to the different lineage and location of these cells (for a review of the conventional nomenclature of blastomeres in the ascidian embryo, see Meinertzhagen & Okamura, 2001; Satoh, 2014). In the invariantly cleaving ascidian embryo, the precursors of the notochord cells can be distinguished as early as the eight-cell stage, where the four blastomeres that will give rise to the notochord precursors are located ventrally and are indicated as the A4.1 and B4.1 blastomere pairs (Conklin, 1905; Nishida & Satoh, 1985; Ortolani, 1954). The anteriorly located A4.1 blastomeres divide to form, in what has become a 16-cell embryo, the A5.1 and the A5.2 blastomere pairs, which are precursors of the A-line notochord, and of neural and endodermal cells. Through subsequent cell divisions the B4.1 blastomere pair descendants, B5.1, B6.2, B7.3 and eventually B8.6, give rise to the secondary notochord (e.g., Satoh, 2014). In the 64-cell embryo, the four precursors of the primary notochord, blastomere pairs A7.3 and A7.7, become clonally restricted to form notochord cells, while the B7.3 blastomere pair is still fated to form both mesenchyme and secondary notochord (green cells in the 64-cell drawing, Fig. 2A). One cell division later, in the 112-cell embryo, eight primary notochord precursors (blastomere pairs A8.5, A8.6, A8.13, A8.14) and two secondary notochord precursors, the B8.6 blastomere pair, can be distinguished (green cells in the 112-cell drawing, Fig. 2A). At gastrulation, 16 primary notochord and 4 secondary notochord precursors are formed; these cells coordinately invaginate as a monolayer epithelium over the archenteron (Fig. 2A and B). By the end of neurulation, the notochord precursors divide for the last time, the definitive 40 post-mitotic notochord cells are formed, and are distributed in two rows of 20 cells each, on both sides of the embryonic midline. At this point, the definitive notochord cells begin to intercalate by converging toward the midline ( Jiang & Smith, 2007) (Fig. 2A, C and C0 ). As a result of intercalation, the notochord cells form a single row, with the shape of a stack of coins, in the center of the tail (Fig. 2A, D and E0 ). Since the 40 definitive notochord cells no longer divide, tail elongation is a direct consequence of their ability to extensively change their shape and stretch along their anterior-posterior axis. During this process, extracellular lumen pockets begin to form between notochord cells (Fig. 2A, F, F0 and G0 ; Fig. 3A and A0 ), and continue to increase in size as the tail completes its elongation (Denker & Jiang, 2012). At the same time, the secretory activity of the notochord cells gradually leads to the formation of the notochordal sheath, a layer of extracellular matrix that continues to thicken throughout tail formation (drawn in purple in Fig. 2A; shown in detail in Fig. 3D and D0 ),

334

Anna Di Gregorio

Fig. 2 Notochord formation in Ciona. (A) Summary of the main stages of notochord formation in Ciona. The first four drawings from the left depict embryos at the stages indicated below each image, with notochord precursors colored in green. Subsequent drawings display part of the 40 definitive notochord cells (green), to illustrate the main changes that they undergo. Cell nuclei, yellow; lumen pockets, red circles/ovals; notochordal sheath, purple lines of increasing thickness. (B–G) Microphotographs of Ciona embryos electroporated at the one-cell stage with the Bra >GFP plasmid (Corbo, Levine, & Zeller, 1997), incubated until the mid-gastrula ( 5.6 hpf ) (B), initial tailbud (C, C0 ), mid-tailbud I (D), late-tailbud II (E, E0 ), late tailbud III (F–G) stages (Hotta et al., 2007). (C0 ) High-magnification view of the region boxed in yellow in (C), showing intercalating notochord cells. Noticeably, intercalation at the anterior and posterior ends of the notochord is nearly complete (discussed in Section 3.2.2). (E’) High-magnification view of the notochord from an embryo at approximately the same stage as the embryo in (E). The notochord cells have a characteristic stack-of-coins arrangement. Nuclei of neighboring cells are stained with DAPI (blue). (F) Notochord of a late tailbud II embryo, displaying several intercellular lumen pockets. (F0 ) High-magnification view of the region boxed in (F), showing intercellular lumen pockets. (G) High-magnification view of the notochord of an embryo co-electroporated with the Bra>GFP and Bra>Slc26αa:: mCherry plasmids (the latter plasmid was kindly provided by Dr. Wei Deng, SARS International Centre for Marine Molecular Biology, Bergen, Norway), to highlight the apical domains of adjacent notochord cells. Scale bars: in (B, C), 40 μm; in (D, G), 10 μm. Timeline of ascidian embryonic development at 18 °C; developmental stages and timing are according to Hotta, Mitsuhara, et al. (2007). “Late gastrula” is also referred to as “neural plate.” Hpf, hours post-fertilization. (H) The approximate expression time frames of Bra and those of its target genes are symbolized by horizontal bars, and their respective lengths are matched to the developmental timeline shown above. Expression of Bra (red bar) is detected in notochord precursors beginning at the 64-cell stage and persists through notochord development (Corbo, Levine, et al., 1997). Expression time frames of early-onset Bra target genes are shown in pink; middle-onset targets

The notochord gene regulatory network in chordate evolution

335

while the increasingly large extracellular lumen pockets tilt and begin to fuse with each other (Fig. 2A). Eventually, these cavities coalesce into a single lumen that occupies the central-most region of the tail (Fig. 3C). Tubulogenesis transforms the notochord from a rod of cells into a fluid-filled tube. The notochordal sheath is now fully developed (Fig. 3D and D0 ) and ready to oppose both the expansion of the notochordal lumen on its inner surface and the pull exerted by the contractile tail muscles on its outer surface (Razy-Krajka & Stolfi, 2019). Tubulogenesis enables the notochord to function as an efficient hydrostatic skeleton, which together with the rhythmic contractions of the muscle cells located on both its sides and the yawing of the trunk allows the swimming movements of the hatched larva (Bone, 1992; Kier, 2012). The strategy employed to complete tubulogenesis and notochord stiffening in Ciona is rather unusual, and it is noteworthy that this process is divergent even among ascidian embryos, as the size and pervasiveness of the lumen vary among different species (reviewed in Smith, 2018). In the vertebrate notochord, tubulogenesis does not take place and stiffening of this structure is achieved through the extensive vacuolization of notochord cells. In Xenopus, the swelling of the individual vacuoles increases the stiffness of the notochord and generates enough pressure to drive its elongation (Adams, Keller, & Koehl, 1990; Fig. 3B and B0 ). Nevertheless, tubulogenesis is a generalized mechanism employed by different organisms to form hollow structures, and remarkably, formation of the dorsal longitudinal anastomotic vessels in zebrafish closely resembles the process of lumen formation that is observed in the Ciona notochord (Herwig et al., 2011).

are colored in light blue, and late-onset Bra target genes are colored in lilac. Schematics on the left side represent minimal notochord CRMs (pink bars) associated with either early-onset targets, which contain two or three presumably cooperative Bra binding sites (red ovals), or middle-onset CRMs (light blue bar) containing individual Bra binding sites (single red oval). Schematics at the bottom left depict how Bra might activate transcription (right angle arrow) of middle-onset transcription factor genes through individual binding sites (single red oval), whose products (yellow hexagon and green oblique bar), in turn, activate expression of late-onset target genes by binding their CRMs (Katikala et al., 2013). Gene names are abbreviated as follows, with respective gene models: thbsp, thrombospondin 3A (KH.C6.164); laminin, laminin c1 (KH.C7.167); collagen, fibrillar collagen 1 (CiFCol1, Wada, Okuyama, Satoh, & Zhang, 2006; KH.C7.633); Noto5 (KH.L153.32); ERM, ezrin-radixin-moesin (KH.C12.129); Tbx2/3 (KH.L96.87); Noto1 (KH.L20.18); Noto8, Calmodulin-like3 (KH.C11.665); Noto4/PID1 (KH.L18.30); Noto9/KTN1/RRBP1 (KH.L13.3); NFAT5 (KH.C3.133); ACL, ATP citrate lyase (KH.C3.228); b4GalT, β-1,4-galactosyltransferase (KH.C11.44).

336

Anna Di Gregorio

Fig. 3 Late stages of notochord formation in Ciona and Xenopus embryos. (A) Brightfield microphotograph of a Ciona robusta embryo, approximate stage late tailbud III (Hotta, Mitsuhara, et al., 2007) (dorsal view). Both sensory organs contain melanin (brown spots in the center of the “head”) and intercellular lumen pockets are regularly spaced between notochord cells throughout the length of the tail. Anterior is on the left side. Scale bar: 20 μm. (A0 ) High-magnification view of the region boxed in violet in (A), displaying 7 notochord cells (N) intercalated by intercellular lumen pockets (LP), in the characteristic shape of concavities. The notochord is flanked bilaterally by non-syncytial muscle (M), and the entire body is covered by epidermis (E). (B) Bright-field microphotograph of a coronal section of a Xenopus laevis tadpole (NF stage 45; Nieuwkoop & Faber, 1967), embedded in paraffin and stained by hematoxylin and eosin. Anterior is on the left side. The notochord is visible in the middle of the section as a multicellular structure with a foamy appearance. At this stage the notochord is completely vacuolated, extends cranially to the infundibulum, and is flanked bilaterally by skeletal muscle. Scale bar: 100 μm. (B0 ) High-magnification view of the region boxed in (B), displaying several notochord cells, each mostly occupied by a large vacuole (V), flanked by skeletal muscle (M). Colors have been removed and contrast has been increased, to better

The notochord gene regulatory network in chordate evolution

337

3.2 The molecular blueprint for notochord formation in Ciona 3.2.1 Notochord induction and specification At the molecular level, notochord formation is initiated by a cascade of inductive events that begin in the 16-cell embryo with the translocation of β-catenin to the nuclei of the mesendodermal precursors. In particular, as the primary notochord precursors, the A5.1 and the A5.2 blastomere pairs, concertedly divide, the nuclear localization of β-catenin becomes progressively restricted to their descendants fated to form endoderm (Imai, Takada, Satoh, & Satou, 2000), and is no longer observed in the precursors of notochord and neural cells (A6.2 and A6.4 blastomere pairs) (Hudson, Kawai, Negishi, & Yasuo, 2013; Satou, 2020). Nuclear localization of β-catenin is an evolutionarily conserved process that induces transcription of several target genes, primarily through the mediation of transcription factors of the TCF/LEF family (e.g., Valenta, Hausmann, & Basler, 2012). One of the target genes directly activated by β-catenin through TCF/LEF sites found in its promoter region encodes for the transcription factor FoxD, which is required for notochord induction (Imai, Satoh, & Satou, 2002). In fact, FoxD is responsible for activating transcription of ZicL (recently renamed Zic-r.b.), a zinc-finger transcription factor, in A6.2 and A6.4 blastomere pairs, as well as in the B6.2 and B6.4 blastomere pairs of the 32-cell embryo (Imai, Satou, & Satoh, 2002), by acting in combination with maternally stored p53/73-a (Noda, 2011) and with the Ciona counterpart of vertebrate Foxa2, currently called Foxa.a (formerly forkhead/HNF-3β or FoxA-a). While activating expression of endomesodermal genes, at the same time FoxD represses expression of key specifiers of neural fate, Otx, Prdm1r.a, Prdm1-r.b, and Dmrt1 (Tokuhiro et al., 2017). Interestingly, two of these proteins, Prdm1-r.a and Prdm1-r.b, act as transcriptional repressors,

visualize cell and tissue boundaries. (C) High-magnification view of a longitudinal optical section the tail of a hatching Ciona larva electroporated at the one-cell stage with the Bra>GFP plasmid, displaying a fully formed lumen and residual fluorescent notochord cells (N). (D) Transmission electron microphotograph of a small region of the tail of a Ciona embryo, approximately late tailbud III, seen in longitudinal section. The center of the tail is occupied by a large lumen pocket (LP) and residues of notochord cells (N) can be seen around it. Muscle cells (M), rich in mitochondria, can be distinguished on both sides of the notochord. Part of the epidermal cells (E), organized in a monolayer, are visible at the periphery of the section. Magnification 2900, scale bar: 10 μm. (D0 ) High-magnification view of the region boxed in (D), showing in the center a close-up of the notochordal sheath (NS), which is characterized by electron-dense material rich in collagen and extracellular matrix proteins (Cloney, 1964). Magnification 19,000, scale bar: 1 μm.

338

Anna Di Gregorio

and together with a third repressor, Hes.a, are responsible for preventing the simultaneous expression of Foxa.a and ZicL in the brain precursors, which would trigger ectopic expression of Bra in this “non-notochord” lineage (Ikeda & Satou, 2017). These elegant experiments revealed that the limited repertoire of transcription factors present in the Ciona genome can be used in different combinations to generate complex developmental outcomes, and shed light on the role of temporally and spatially localized repression events in cell-fate restriction. One cell division after its activation, ZicL binds the Bra promoter region through two binding sites located 170 bp upstream of its transcription start site, and induces Bra expression in the A-line notochord cells (Yagi, Satou, & Satoh, 2004). The expression of Bra marks the definitive commitment of these cells to the notochord fate. In addition to this gene regulatory cascade, another mechanism controls the expression of Bra in notochord cells. The nuclear localization of β-catenin induces also the expression of FGF9/16/20, which in turn activates expression of Bra via the MAPK/MEK/ERK signaling pathway. FGF9/16/20 is required for the induction of notochord fate, and subsequently, together with FGF8/17/18, maintains the A-lineage committed to form notochord (Yasuo & Hudson, 2007). At the same time, the activity of the MAPK/MEK/ERK signaling pathway, and thus the expression of Bra, are restricted to these cells by the expression of Ci-Ephrin-Ad; embryos lacking the appropriate activity of Ci-Ephrin-Ad ectopically express Bra in neural precursors, as a result of ectopic ERK activation via its dephosphorylation (Picco, Hudson, & Yasuo, 2007; Shi & Levine, 2008). The transcriptional effector of the FGF signaling pathway in the notochord is an ETS family transcription factor(s), and ETS binding sites are found in the Bra “shadow” enhancer (Farley, Olson, Zhang, Rokhsar, & Levine, 2016). Notochord fate induction in the blastomeres of the B-line that will form the secondary notochord involves two different molecular pathways: Nodal and Delta/Notch. The B8.6 secondary notochord precursors receive inductive Nodal signals from their adjacent b-line blastomeres, the b6.5 pair (Hudson & Yasuo, 2006). Primary notochord precursors, A7.6, express the Nodal-downstream gene Delta2, which encodes a divergent, membranebound Notch ligand (Hudson & Yasuo, 2005). Delta2 signaling from the A7.6 blastomeres activates the canonical Delta/Notch pathway in the B-line notochord precursors, and the interaction of the Notch intracellular domain with the transcription factor Suppressor of Hairless, Su(H), induces activation of Bra expression through Su(H) binding sites that have been identified in the Bra promoter region and are required for its function

The notochord gene regulatory network in chordate evolution

339

(Corbo, Fujiwara, Levine, & Di Gregorio, 1998). β-catenin/TCF/LEF, FGF/MAPK/MEK/ERK, Nodal and Delta/Notch/Su(H) are all evolutionarily conserved pathways, and have been adapted in a unique way in ascidians to induce the notochord fate in two separate cell lineages. Although this developmental strategy has been reported thus far only in ascidians, it is intriguing that in Xenopus embryos one of the Su(H) transcription factors, XSu(H)2, is required for expression of Xenopus brachyury (Xbra) (Ito, Katada, Miyatani, & Kinoshita, 2007).

3.2.2 Invagination and convergent extension Notochord precursors invaginate during gastrulation; at this time, they acquire a wedge-shaped aspect and form a monolayer epithelium (Munro & Odell, 2002b). As shown by studies in another ascidian, Boltenia villosa, at the beginning of neurulation notochord precursors divide one last time, and 1 h later, the resulting cells begin to extend F-actin-based protrusion and to acquire motility (Munro & Odell, 2002a). Remarkably, the formation of these protrusions is part of a developmental cell behavior intrinsic to the notochord cells, and is exhibited autonomously by notochord cells in isolation; however, within the embryo, the appropriate planar polarization of the protrusions, which is specific to the interior basolateral edges of the notochord cells, requires interactions with adjacent tissues (Munro & Odell, 2002a). More generally, within the embryo, notochord cells exchange mechanical patterning cues with the muscle cells flanking them (Di Gregorio, Harland, Levine, & Casey, 2002). Shortly after neurulation, two arcs, each containing 20 notochord cells, form on both sides of the embryonic midline, and need to intercalate mediolaterally in order to form a single cylindrical rudiment in the center of the developing tail. This mediolateral intercalation, or convergence, is part of the overall process of convergent extension that will drive tail elongation in the absence of further cell divisions. In vertebrate embryos, actomyosin-based polarized protrusions and/or lamellipodia reportedly generate the forces that drive mediolateral convergence of intercalating cells (e.g., Pfister, Shook, Chang, Keller, & Skoglund, 2016). Together with the extension of protrusions, as discussed in Munro and Odell (2002a), formation of the notochord by invagination of the epithelial sheet that constitutes the roof of the archenteron is a conserved morphogenetic mechanism observed throughout the chordate spectrum, from amphioxus to mouse embryos (Conklin, 1928; Sulik et al., 1994).

340

Anna Di Gregorio

Studies carried out in Ciona savignyi through the analysis of a spontaneous notochord mutation, called aimless, indicate that convergent extension is mediated by the non-canonical planar cell polarity (PCP) pathway ( Jiang, Munro, & Smith, 2005). The aimless mutation has been traced back to the loss of function of the gene prickle, which in Ciona is required for the polarization of notochord cell protrusions and for the asymmetric localization of another PCP component, disheveled (dsh), exclusively to the side of the notochord cells that directly borders the muscle ( Jiang et al., 2005). Shortly after the completion of intercalation, Prickle becomes specifically localized to the anterior pole of the notochord cells, followed by another PCP component, Strabismus/van Gogh, and by one of the Ciona myosin proteins (Kourakis et al., 2014). On the other hand, the nuclei, at least in primary notochord cells, remain located posteriorly, although in ascidians other than Ciona their localization can vary (Kourakis et al., 2014; Newman-Smith, Kourakis, Reeves, Veeman, & Smith, 2015). An additional instructive role for the process of mediolateral intercalation and the formation of protrusions is provided by yet another tissue neighboring the notochord, the developing nervous system. Before and after intercalation, the notochord cells express an FGF receptor, Ci-FGFR, which is activated by Ci-FGF3; the latter is expressed in the ventral-most of the four cells that compose the nerve cord, as observed in cross-section (Fig. 1A). Ci-FGF3 morphant embryos display defects in convergent extension, which have been attributed to an impaired formation of appropriately polarized protrusions (Shi, Peyrot, Munro, & Levine, 2009). Differently from the FGF-mediated notochord induction, this process is independent of the MAPK pathway (Shi et al., 2009). Interestingly, not all notochord cells intercalate simultaneously. Notochord cells located at the anterior-most and posterior-most extremities of this structure intercalate faster, and begin narrowing sooner, than the cells located in its midsection (Fig. 2C0 ). Additionally, unequal cell divisions occur in both A-line and B-line precursors and reduce the volume of the cells that will end up forming the tips of the notochord. Together, these cellular behaviors cause the characteristic tapering of the notochord (Veeman & Smith, 2013; Winkley, Ward, Reeves, & Veeman, 2019). 3.2.3 Formation of the notochordal sheath and elongation As the notochord cells interdigitate during intercalation, their protrusions diminish and they actively secrete ECM components, which will participate in the formation of a basal lamina and of a structure that was originally

The notochord gene regulatory network in chordate evolution

341

described as a collagen-based notochordal sheath (Cloney, 1964). In support of these first findings, recent studies have shown that several collagen genes, representative of at least four collagen families, including fibrillar collagen, are expressed in the notochord cells at these stages (Katikala et al., 2013; Wada et al., 2006; reviewed in Satoh, 2014). Alongside collagen genes, notochord cells express numerous other ECM components, including alpha and beta laminins, fibronectin, entactin, cadherins, thrombospondin and others ( Jos
e-Edwards, Oda-Ishii, Nibu, & Di Gregorio, 2013; Katikala et al., 2013; Kugler et al., 2008). Along with ECM components sensu stricto, also numerous enzymes involved in their secretion and post-translational modification are synthesized by the notochord cells, including leprecan/ P3H1, which encodes the collagen modifier prolyl 3-hydroxylase 1 (P3H1) (Dunn & Di Gregorio, 2009), lysyl oxidases, which are involved in collagen and elastin cross-linking, furin, coatomer proteins, and others (Kugler et al., 2008). Studies of Ciona notochord genes have guided the identification of related genes in mouse, including three orthologs of the single-copy Ciona leprecan/P3H1 gene, named Leprecan/P3H1, Leprecanlike1/P3H2 and Leprecan-like2/P3H3, which are all expressed in notochord cells, as well as in various additional vertebrate-specific territories that are not present in ascidian embryos, including the developing vertebral cartilages (Capellini, Dunn, et al., 2008). The characterization of the Ciona notochord transcriptome through RNA-Seq has confirmed the abundant expression of ECM components and their modifiers by the notochord cells (Reeves, Wu, Harder, & Veeman, 2017). The ECM is also responsible for forming a boundary that under normal circumstances is able to reduce the motility of intercalating notochord cells, thus organizing this critical morphogenetic movement. This mechanism, called “boundary capture,” has been described in amphibian embryos as well (Keller et al., 2000; Smith, 2018). The role of different ECM components in notochord development has been investigated using spontaneous mutants in Ciona savignyi and various gene knockdown techniques in Ciona robusta. Work on another spontaneous Ciona savignyi mutant, chongmague, which carries a mutation in the laminin-α3/4/5 gene, shows that when the ECM is lacking this laminin the notochord cells begin intercalation but maintain their protrusions and their mobility, and migrate randomly to ectopic locations in the tail (Veeman et al., 2008). In Ciona robusta, the notochord cis-regulatory module (CRM, or enhancer) associated with the laminin c1 gene is directly controlled by Bra through cooperative binding sites (Katikala et al., 2013; Fig. 2H).

342

Anna Di Gregorio

The knockdown via CRISPR/Cas9-mediated editing of another ECM component, fibronectin (Ci-FN1-containing in Jos
e-Edwards et al., 2013; renamed Ci-Fn in Segade, Cota, Famiglietti, Cha, & Davidson, 2016), also disrupts intercalation, although it does not seem to affect the structure of the notochordal sheath (Segade et al., 2016). Instead, formation of the notochordal sheath appears impaired by the shRNA-mediated knockdown of leprecan/P3H1 (Dunn & Di Gregorio, 2009). Analogously, CRISPR/ Cas9-mediated editing of this gene causes the formation of a slightly shorter, kinked tail. These results suggest that a reduction in leprecan/P3H1 function might reduce the post-translational modification of collagen molecules in the notochordal sheath, thus lowering its structural integrity and rigidity (Maguire et al., 2018). Notochord elongation in Ciona is characterized by a striking change in cell shape, which takes place through a divergent molecular mechanism. The notochord cells transition from the stack-of-coins shape that they display in middle and late-tailbud to an elongated “drum-like” shape, while maintaining their volumes constant. This impressive process relies upon the formation of an equatorially positioned actomyosin ring that constricts the notochord cells and promotes their elongation (Sehring et al., 2014). Interestingly, the actomyosin ring is highly similar, in both structure and molecular composition, to the cytokinetic ring that is formed during normal cell divisions; however, this ring assembles at the anterior aspect of the notochord cells at the time when the notochord has the appearance of a stack of coins, and interacts with the anteriorly localized PCP components, Dsh, prickle and Strabismus/van Gogh (described above). At this point, the ring repositions itself to the middle of each cell, begins contracting and produces a constriction along its circumference, thus inducing the elongation of each cell and its transition to the characteristic “drum-like” shape (Lu, Bhattachan, & Dong, 2019; Sehring et al., 2015). The actomyosin ring eventually disassembles, through a still unknown mechanism, and tail elongation proceeds through tubulogenesis (Lu et al., 2019).

3.2.4 Tubulogenesis and disappearance The process of tubulogenesis in Ciona has been reconstructed in exquisite detail through a combination of molecular and morphometric studies (Deng et al., 2013; Denker et al., 2013, 2015; Denker & Jiang, 2012; Dong, Deng, & Jiang, 2011; Dong et al., 2009) and is briefly summarized here.

The notochord gene regulatory network in chordate evolution

343

At the onset of tubulogenesis, extracellular vacuoles, or more precisely, intercellular lumen pockets, which were originally mistaken for large intracellular vacuoles by analogy to those observed in vertebrate notochord cells (Fig. 3B and B0 ), form between notochord cells and gradually induce another change in their shape, from “drum-like” to biconcave. All notochord cells form two apical domains, each facing another notochord cell. This peculiar change in cell polarity marks a mesenchymal-to-epithelial transition of the notochord cells ( Jiang & Smith, 2007). Both apical domains in each notochord cell contain Par3/Par6/aPKC protein complexes, organized in patches, as well as two sets of tight junctions (Denker et al., 2013; Oda-Ishii, Ishii, & Mikawa, 2010). The molecular event that triggers the formation of the lumen pockets is the localization, on each apical domain, of the anion transporter SLC26αa, which is necessary for the formation of the notochord lumen and its expansion through the transport of fluid from the notochord cells toward the lumen (Deng et al., 2013). More generally, SLC26 is an evolutionarily conserved family of transporters that participate in numerous events in development and disease, and in humans are expressed in different regions of the renal tubules that are involved in ion transport (e.g., Markovich & Aronson, 2007). As the lumen pockets increase in size by virtue of newly synthesized apical membrane and SLC26mediated ion flow, the notochord cells remain sealed together by tight junctions in the regions where their adjacent apical domains are juxtaposed (Denker et al., 2015). Claudin-rich tight junctions gradually replace the adherens junctions between notochord cells, another event that is required for the proper formation of the lumen pockets (Denker et al., 2013; Dong et al., 2009). Accordingly, transcripts for one of the Ciona claudins, Claudin16/17/19, whose expression is regulated by the transcription factor Bhlh-tun1, become detectable shortly before the onset of this process (Kugler et al., 2019). In vivo confocal studies and morphometric analyses have shown that cortical actin and ezrin-radixin-moesin (ERM) are also required for lumen formation, along with a microtubule network that forms at the apical cortex of the notochord cells through an actin-dependent subcellular movement (Dong et al., 2011). The microtubule network allows the notochord cells to form lamellipodia that enable them to crawl on the notochordal sheath and eventually flatten to an endothelial-like appearance to facilitate the coalescence of the lumen pockets and consequent formation of the definitive fluid-filled lumen (Dong et al., 2011; Fig. 3C). The interaction of ERM with the multifunctional protein 14-3-3εa at the basal cortex of the notochord cells is necessary for the transport of components required

344

Anna Di Gregorio

for lumen formation from the basal region of these cells to their lumenfacing domain (Mizotani et al., 2018). A notochord CRM directly controlled by Bra through cooperative binding sites has been identified in the ERM genomic locus (Katikala et al., 2013; Fig. 2H). Ciona larvae with a fully differentiated notochord are able to swim for a few hours and to select an appropriate substrate to which they will adhere in preparation for metamorphosis. In laboratory cultures, at the temperature of 18 °C, larval hatching occurs 18 h after fertilization (Fig. 2), and swimming continues for about 5 h after hatching; at that point, the larvae attach to a submerged substrate, or to the bottom of Petri dishes filled with seawater, and begin retracting their tail after a minimum of 28 min after adhesion (Matsunobu & Sasakura, 2015). Most of the cells of the tail undergo caspase-dependent apoptosis, including cells of the epidermis and the posterior-most notochord cells (Krasovec, Robine, Qu
einnec, Karaiskou, & Chambon, 2019). Apoptosis is determined and organized by the expression of Ci-Hox12, through an evolutionarily conserved mechanism; on the other hand, the propagation of the apoptotic wave that starts from the posterior end of the tail and proceeds anteriorly is considered a rather unique adaptation among chordates (Karaiskou, Swalla, Sasakura, & Chambon, 2015). A notochord residue remains visible in the metamorphosing larva and in the resulting juvenile, but differently from the notochord remnants that in vertebrates are incorporated in the nuclei pulposi, these residual notochord cells are entirely eliminated as post-metamorphic development proceeds. The expression of Bra is reportedly undetectable after metamorphosis (Chiba, Jiang, Satoh, & Smith, 2009), while expression of transcription factors that are exclusively active in adult structures is activated. Of note, differently from Bra, several notochord genes are rerouted to different cell-types and newly formed structures after the disappearance of the notochord. One representative example is Bhlh-tun1, a transcription factor that functions in the developing notochord (Kugler et al., 2019), and after metamorphosis is expressed first in atrial siphon muscle precursors, and later on in the oral siphon muscle (Razy-Krajka et al., 2014).

4. Time is of the essence: Temporal cis-regulatory control of gene expression in a fast-developing chordate Comparably to other marine organisms, ascidian embryos develop through a larval stage; however, differently from other marine larvae,

The notochord gene regulatory network in chordate evolution

345

ascidian embryos are unable to feed themselves and must rely upon yolk granules and other maternally stored nutrients for their survival. This condition imposes strict limits on the duration of the pre-metamorphic embryogenesis, which at 18 °C is completed within roughly 1 day and takes place even more expeditiously at higher temperatures (Satoh, 2014). At 18 °C, the notochord completes its development, from specification to lumen formation, in 14 h (Fig. 2). Based on timescales determined for mammalian cells, transcription of an average Ciona gene (10.5 kb) is expected to require 10 min, and maturation, translation and folding might take approximately 1–3 additional minutes each (Phillips, Kondev, Theriot, & Garcia, 2012); this suggests that within the 12–14 h required for notochord development, the notochord GRN could, theoretically, utilize a multi-tiered cascade of transcription factors. However, a large fraction of the notochord CRMs analyzed thus far rely upon Bra/T-box binding sites (generic consensus: TNNCAC; Di Gregorio & Levine, 1999; Dunn & Di Gregorio, 2009; Katikala et al., 2013; Jos
e-Edwards et al., 2013, 2015; Thompson & Di Gregorio, 2015). Interestingly, bona fide Bra-downstream notochord genes (described in detail in Section 5.2) are expressed in the developing notochord at different developmental stages (Hotta et al., 2000; Hotta, Takahashi, Erives, Levine, & Satoh, 1999; Takahashi et al., 1999), even though Bra transcripts are present at all stages of notochord development (Corbo, Levine, et al., 1997) and are, arguably, translated into active Bra protein that is detectable in the nuclei of late tailbud embryos (Katikala et al., 2013). In particular, transcripts for early-onset Bradownstream genes are first detected in notochord precursors around early gastrulation; middle-onset target genes become detectable around the late gastrula/neural plate stage, and late-onset target genes are first detected around the beginning of neurulation (Hotta et al., 1999; Katikala et al., 2013). The molecular mechanisms responsible for these differences in the temporal read-out of Bra-downstream gene expression have been investigated through the characterization of the notochord CRMs associated with genes representative of early-, middle- and late-onset Bra targets, and the minimal sequences and binding sites required for their function have been identified (Katikala et al., 2013). The results of these experiments suggest that notochord CRMs that are associated with early-onset notochord genes require multiple functional Bra binding sites (Katikala et al., 2013; Fig. 2H). This category includes thrombospondin 3A (thbsp), and laminin c1, which encode evolutionarily conserved ECM components (Urry, Whittaker, Duquette, Lawler, & DeSimone, 1998), fibrillar collagen 2A1 (CiFCol1), a

346

Anna Di Gregorio

presumptive component of the notochordal sheath, which in vertebrates is reportedly co-opted to cartilaginous structures (Cloney, 1964; Katikala et al., 2013; Wada et al., 2006), and ERM, which is required for cell-shape changes and lumen formation (detailed in Section 3.2.4) (Dong et al., 2011; Hotta, Yamada, Ueno, Satoh, & Takahashi, 2007; Katikala et al., 2013). Middle-onset notochord genes, such as Noto9/KTN1/RRBP1, which is first detected in notochord cells by the neural plate/early neurula stage, and encodes a ribosome-binding protein that is also detected in the notochord of Xenopus embryos (Liu et al., 2016), are controlled through notochord CRMs that rely upon individual functional Bra binding sites (Katikala et al., 2013; Fig. 2H). This model predicts that removal of a functional Bra binding site from the CRM of an early-onset notochord gene that relies upon two Bra binding sites would delay its onset of activity, essentially turning it into a middle-onset CRM. Accordingly, this is the result that was obtained when either one of the two Bra binding sites of the laminin c1 notochord CRM was mutated; either mutation had the effect of turning this CRM into a middle-onset one (Katikala et al., 2013). Lastly, the notochord CRMs of the two late-onset Bra target genes analyzed, ATP-citrate lyase (ACL), which is required for the establishment of cell polarity and for intercalation (Hotta, Yamada, et al., 2007), and β1,4-Galactosyltransferase (β4GalT), which is presumed to be involved in the formation of the notochordal sheath, are devoid of functional Bra binding sites and are likely controlled by Bra indirectly, through a relay mechanism that involves the activity of Bra-downstream intermediary transcription factors (Fig. 2H) (Katikala et al., 2013). Recent ATAC-Seq studies have provided accurate genome-wide profiles of the chromatin state in developing Ciona embryos (Madgwick et al., 2019; Racioppi, Wiechecki, & Christiaen, 2019), which will be instrumental for testing and refining this mechanistic model.

5. The GRN underlying notochord formation in Ciona A GRN can be envisioned as a blueprint of the regulatory interactions existing among genes expressed in a certain tissue, organ, or structure. These interactions are responsible for generating a specific spatial and temporal regulatory state that will, in turn, originate a specific phenotypic output (Peter, 2017). These gene regulatory circuitries are encoded in each genome in form of transcription factor genes and cis-regulatory regions, and as such, are subjected to mutations and evolutionary changes that can lead to

The notochord gene regulatory network in chordate evolution

347

variation in the processes that they control, and in their phenotypic manifestations (e.g., Levine, 2010). Genes whose products control the expression of other genes, such as genes encoding for transcription factors, RNAbinding proteins, and other regulatory molecules, are positioned internally within the GRN, while genes whose products do not control other genes, but rather act as direct effectors of cellular differentiation and morphogenesis, such as genes encoding for enzymes, ECM components and structural proteins, are located at the periphery (Davidson, 2006; Peter & Davidson, 2015). The reconstruction of the Ciona notochord GRN began with the identification of its first regulatory nodes, Brachyury and Foxa.a (Corbo, Erives, Di Gregorio, Chang, & Levine, 1997; Corbo, Levine, et al., 1997; Di Gregorio, Corbo, & Levine, 2001). Toward the end of 2002, the public release of the first assembly of the Ciona intestinalis (now robusta) genome (Dehal et al., 2002) enabled the identification of several notochord genes through sequence homology (Kugler et al., 2008). Hundreds of notochord genes were identified through the coordinated efforts of a group of laboratories that have carried out the systematic identification of the expression patterns of several genes spanning a wide range of developmental stages, from unfertilized eggs and zygotes, to larvae and juveniles (Fujiwara et al., 2002; Kusakabe et al., 2002; Miwata et al., 2006; Nishikata et al., 2001; Ogasawara et al., 2002; Satou et al., 2002, 2001; Satou, Kawashima, Shoguchi, Nakayama, & Satoh, 2005). A few years later, another remarkably comprehensive study, specifically focused on transcription factors and signaling molecules, identified a group of potential regulatory nodes of the notochord GRN (Imai, Hino, Yagi, Satoh, & Satou, 2004; Satou & Satoh, 2005; Satou, Wada, Sasakura, & Satoh, 2008). Together, these outstanding foundational projects provided an enviable arsenal of biomolecular resources for Ciona, and propelled the field of ascidian biology toward the use of a system biology approach for studies of developmental processes and functional genomics. This research culminated with the morpholinomediated knockdown of 53 zygotically expressed transcription factor genes and 23 signal transduction molecules (Imai, Levine, Satoh, & Satou, 2006). This work initiated the reconstruction of the genetic blueprints underlying the formation of most of the tissues found in the Ciona embryo (Imai et al., 2006; Satou et al., 2005 http://ghost.zool.kyoto-u.ac.jp/otherfr_kh.html). The translucency of Ciona embryos, the ease of transgenesis and their availability in large quantities allow the use of fluorescence-activated cell sorting (FACS) for the purification of specific cell populations (Christiaen et al., 2008; Christiaen, Wagner, Shi, & Levine, 2009). FACS-mediated

348

Anna Di Gregorio

isolation of notochord cells was used in combination with microarray screens, and has added to the Ciona notochord GRN a group of transcription factors, described below, which had escaped previous searches, as well as 456 previously published effector genes and >100 novel notochord genes ( Jos
e-Edwards et al., 2011; our unpublished results). Taken together, all these studies have yielded 700 bona fide notochord genes, whose expression has been validated by WMISH. More recently, RNA-Seq on FACSpurified cell populations and single-cell RNA-Seq (scRNA-Seq) have been used to elucidate the full complement of tissue-specific transcriptomes present in Ciona before metamorphosis (Cao et al., 2019; Horie et al., 2018; Reeves et al., 2017). Preliminary analyses of the results of these experiments suggest that the number of genes expressed in the Ciona notochord might rise sharply to a few thousands. The Ciona genome spans 123 Mb and contains a little over 14,000 genes (Dehal et al., 2002; Satou et al., 2019). The total number of transcription factor genes is 385 (Dehal et al., 2002; Satou et al., 2019; Stolfi & Christiaen, 2012); of these, about 50–60 are expressed in the notochord during different stages of its development, with widely variable intensity and duration. This number does not include most of the zinc-finger proteins expressed in the notochord, some of which might function as transcription factors (Miwata et al., 2006), and might further increase after the validation of new candidate notochord genes identified by scRNA-Seq (Cao et al., 2019; Horie et al., 2018).

5.1 The main regulatory nodes of the notochord GRN: Brachyury and Foxa2 Two evolutionarily conserved transcriptional regulators, Brachyury and Foxa2, appear reiteratively in chordate evolution as components of notochord GRNs, and are both present, in single copy, in the Ciona genome (Corbo, Erives, et al., 1997, Corbo, Levine, et al., 1997, Di Gregorio et al., 2001). Ciona Brachyury (Bra; originally published as Ci-Bra; Corbo, Levine, et al., 1997) is a member of the T-box family of transcription factors (Di Gregorio, 2017; Takatori et al., 2004), while Foxa.a is a member of the forkhead/winged-helix family (Di Gregorio et al., 2001; Imai et al., 2004). These two genes possess orthologs in numerous organisms throughout, and outside, the phylum Chordata, and their functions in development and evolution of multicellular organisms decidedly precede the role in notochord formation that they acquired in the chordate lineage. Mice heterozygous for mutations in the Bra locus are characterized by a short tail (Brachyury, in Greek), while mice homozygous for mutant alleles of Bra

The notochord gene regulatory network in chordate evolution

349

die in utero because they lack notochord and allantois, and display abnormal somites (Stott, Kispert, & Herrmann, 1993; Wilkinson, Bhatt, & Herrmann, 1990). Likewise, ENU-generated Bra mutant Ciona embryos display a very short tail and lack an organized notochord (Chiba et al., 2009). However, differently from most other chordates examined thus far, Bra is notochord-specific in Ciona (Corbo, Levine, et al., 1997), and this renders ascidian embryos ideally suited for studies of the notochord-specific function of this transcription factor. In mouse embryos, Foxa2 is expressed in notochord, nervous system, endoderm and additional territories (Sasaki & Hogan, 1993). Mice carrying a homozygous mutation in the Foxa2 locus lack an organized node and fail to develop a notochord (Ang & Rossant, 1994). Resembling its vertebrate counterparts, Ciona Foxa.a is expressed in notochord, endoderm, and in the ventral-most cells of the nerve cord, which are considered a rudimentary floor plate (Corbo, Erives, et al., 1997; Di Gregorio et al., 2001). The crossregulatory relationship between these two transcription factors has been investigated in the ascidians Ciona and Halocynthia and in vertebrates. In Ciona, the results of the morpholino-mediated knockdown of Foxa.a indicate that this transcription factor controls the expression of numerous genes, among which Bra, which is down-regulated in Foxa.a morphants (Imai et al., 2006). These results are confirmed by parallel studies in Halocynthia roretzi, which show that Hr-FoxA is sufficient to elicit ectopic expression of Hr-Bra (Kumano, Yamaguchi, & Nishida, 2006), and by experiments carried out in mouse embryos (Tamplin et al., 2008). On the other hand, Bra controls Foxa2 expression in mouse embryoid bodies (Lolas, Valenzuela, Tjian, & Liu, 2014), and the Foxa.a locus in Ciona is bound by Bra in chromatin immunoprecipitation on DNA microarray (ChIP-chip) and in electrophoretic mobility shift assay (EMSA) (Kubo et al., 2010; our unpublished results). The presumptive positive feedback between these two transcription factors suggests that they constitute a subcircuit of the Ciona notochord GRN. Moreover, experimental evidence suggests that both transcription factors feedback on their own transcription, with different effects. In the case of Bra, ChIP-chip data suggest the presence of autoregulatory sites in the Bra locus (Kubo et al., 2010), and interestingly, morpholino-mediated knockdown experiments indicate that Bra negatively regulates its own transcription (Imai et al., 2006). In the case of Foxa.a, the occupancy of the Foxa.a locus by its own protein product is confirmed by EMSA experiments that show that binding sites for winged-helix proteins within the Foxa.a cis-regulatory region are bound by an in vitro synthesized

350

Anna Di Gregorio

Foxa.a protein (Di Gregorio et al., 2001). Mutation analysis of Foxa.a binding sites within the Foxa.a cis-regulatory region in vivo indicated the positive autoregulatory role of these sequences (Di Gregorio et al., 2001; Fig. 4). A positive autoregulatory loop is predicted for mouse Foxa2 as well (Tamplin et al., 2008). Bra and Foxa2 orthologs are reiteratively incorporated into different GRNs that operate in diverse structures/tissues of non-chordate embryos (Davidson, 2006, 2010; Davidson & Erwin, 2006). In Ciona and mouse embryos, these two transcription factors appear to have established positive cross-regulatory interactions that have locked them into a subcircuit that is presumably specific to the notochord and required for its development. In support of this possibility, work on Ciona notochord CRMs has provided the first evidence that these two transcription factors synergistically control notochord gene expression ( Jos
e-Edwards et al., 2015; Passamaneck et al., 2009; Section 6.1). Future studies on the structure of the notochord GRNs in other chordates will clarify whether the positive feedback circuitry between Bra- and Foxa2-related transcription factors is evolutionarily conserved, i.e., whether the Bra/Foxa2 subcircuit constitutes a notochord “kernel” (Davidson & Erwin, 2006; Peter & Davidson, 2015).

5.2 Drawing edges, connecting nodes: Reconstructing the gene batteries controlled by notochord transcription factors The lineage of the notochord cells in Ciona is invariant, as is the final number of 40 post-mitotic cells (Nishida & Satoh, 1983, 1985). The overwhelming majority of notochord genes identified thus far are expressed fairly homogeneously in all 40 notochord cells. However, a few genes, such as multidom, exhibit a peculiar mosaic expression that is apparently random and lineageindependent (Oda-Ishii & Di Gregorio, 2007), while others display a more regionalized expression along the anterior-posterior axis of the notochord (Reeves, Thayer, & Veeman, 2014; Utsumi, Shimojima, & Saiga, 2004), and a few examples of genes specifically expressed in secondary notochord have been recently reported (Harder, Reeves, Byers, Santiago, & Veeman, 2019). Interestingly, Multidom is a protein characterized by multiple domains (Oda-Ishii & Di Gregorio, 2007) including a region of sequence similarity to Delta-like proteins. The full complement of genes that exhibit these heterogeneous expression patterns, and their impact on the global notochord developmental program, are yet to be determined. Nevertheless, the Ciona notochord could be envisioned as a

The notochord gene regulatory network in chordate evolution

351

Fig. 4 A close-up of the Bra/Foxa.a subcircuit and its relationship with notochord transcription factors, differentiation genes, and CRMs. An extremely detailed and comprehensive view of the Ciona notochord specification GRN has been garnered through expression studies and morpholino-mediated knockdown experiments (Imai et al., 2004, 2006), and is available online on the Ghost database website (http://ghost.zool. kyoto-u.ac.jp/otherfr_kh.html Satou et al., 2005). This is a summary of the Bra/Foxa.a subcircuit, mainly based on our findings on notochord transcription factors that were not characterized in those previous studies, and on their target genes and notochord CRMs. For simplicity, information about the temporal onset of individual genes has been omitted from this circuit diagram. After the identification of Bra and Foxa.a, and of their cis-regulatory regions (Corbo, Erives, et al., 1997; Corbo, Levine, et al., 1997; Di Gregorio et al., 2001), subsequent studies have uncovered the notochord expression of Bhlh-tun1 and Tbx2/3 (Imai et al., 2004), MIER1 and Xbp1 (Kugler et al., 2008), and of the seven additional notochord transcription factors shown in colored font, most of which are conserved in vertebrate notochord and/or nuclei pulposi (Jos
e-Edwards et al., 2011). The notochord CRM of Bhlh-tun1 is controlled directly by Bra, Foxa.a and by an early-onset homeodomain transcription factor (early HD, gray), which remains to be ascertained (Kugler et al., 2019). Transcription of NFAT5 is directly regulated by Bra (our unpublished results), as is that of Tbx2/3, which is activated through cooperative Bra binding sites (Jos
e-Edwards et al., 2013). Peripheral components (differentiation effector genes) of this part of the GRN are shown in black font; among these, effector genes whose notochord CRMs have been fully characterized are annotated in bright blue font. The role of a Myb-like transcription factor in the activation of some of the notochord CRMs (C6-sulfotransferase and carboxypeptidase) has been inferred on the basis of the minimal sequences required for notochord activity of these cis-regulatory regions (Jos
e-Edwards et al., 2015). Twenty-one notochord genes have been identified as targets of Bhlh-tun1, including Claudin16/17/19, whose notochord CRM has been identified (Kugler et al., 2019). Transcription of several notochord genes is influenced by Xbp1 (our unpublished results), and 20 notochord genes are controlled by Tbx2/3, including Fos-a and Duox-c, which are down-regulated by it (Jos
e-Edwards et al., 2013). The fibronectin notochord CRM was characterized by Segade et al. (2016) and relies upon a T-box binding site, which could be bound by Bra and/or Tbx2/3, and a divergent Fox-like binding site. The Ephrin 3 notochord CRM is controlled directly by Bra, but requires for its function also an (AC)6 microsatellite repeat (Jos
e-Edwards et al., 2015), an arrangement (Continued)

352

Anna Di Gregorio

self-contained biological system that is relatively impermeable to variability and perturbation, and as such, can be systematically dissected through the analysis of its regulatory nodes and of the gene batteries that they (co-) regulate. According to the recent estimates obtained from scRNA-Seq studies, the notochord GRN might involve the activity of a few thousand genes to complete all the steps of its morphogenesis. If all these genes were equally controlled by the 60 transcription factors expressed in the notochord, in a series of nearly parallel independent cascades, this would assign a few hundred target genes to each transcription factor. A more realistic scenario depicts the Bra/Foxa.a subcircuit presiding a regulatory differentiation hierarchy where either transcription factor, or both, control the majority of the peripheral nodes of the GRN, i.e., the differentiation effector genes, either directly, or indirectly through the use of transcriptional intermediaries. Extremely detailed GRN maps have been reconstructed for the pre-metamorphic tissues of the Ciona embryo through the systematic morpholino-mediated knockdown of numerous genes encoding for transcription factors (Imai et al., 2006; Satoh, 2014; Satou et al., 2005 http://ghost.zool.kyoto-u.ac.jp/otherfr_kh.html). In Fig. 4, we provide a high-magnification view of the Bra/Foxa.a subcircuit that is activated after notochord specification is completed (Section 3.2.1; Satou, 2020). This summary is mainly focused on transcription factors identified more recently, which therefore were not included in the global GRN maps, and are currently being characterized ( Jos
e-Edwards et al., 2011; Kugler et al., 2008, 2019; our unpublished results). The first two of these transcription factors, Ciona Xbp1 and Mi-er1, were identified through a survey for orthologs of vertebrate notochord genes in Ciona (Kugler et al., 2008), while Sall-a (related to both human genes SALL1 and SALL3), NFAT5, Stat5/6b, Klf15, Fos-a, AFF2/3/4, and Lmx-like were identified, along with Fig. 4—Cont’d that is also found in Bra-bound genomic regions in mouse embryonic stem cells (Evans et al., 2012). Regulatory interactions that could be either direct or indirect are interrupted by diagonal black lines. Positive interactions are shown as arrows, inhibitory interactions are indicated by flat-headed arrows. Dashed lines indicate interactions that are hypothesized and need to be confirmed. Negative autoregulatory feedback for Bra, revealed by morpholino-mediated knockdown (Imai et al., 2006), is indicated by a flat-headed circular arrow. Positive autoregulatory feedback for Foxa.a (circular arrow) was uncovered by in vivo mutation analyses coupled with in vitro DNA-binding assays (Di Gregorio et al., 2001). All additional data for notochord CRMs are from Dunn and Di Gregorio (2009), Passamaneck et al. (2009), Jos
e-Edwards et al. (2015), Katikala et al. (2013), Thompson and Di Gregorio (2015). Abbreviations: Olfactom.1, olfactomedin 1; Sulfotr., sulfotransferase.

The notochord gene regulatory network in chordate evolution

353

previously published transcription factors and numerous effector genes, through the microarray screen detailed above ( Jos
e-Edwards et al., 2011; our unpublished results). As discussed in the respective publications, with the exception of Klf15, all these transcription factors are related to vertebrate counterparts that are expressed in notochord and/or nuclei pulposi; it is noteworthy that a recent study reported that SALL3 is a unique marker of spine chordoma, while LMX1A is predominant in skull base chordoma (Bell et al., 2018). In situ hybridization experiments on embryos that were either carrying a mutation in the Bra locus or that were expressing a repressor form of this transcription factor showed that embryos in which the levels or the functions of Bra are altered, the expression of NFAT5, AFF2/3/4, Fos-a and Klf15 is no longer detectable in the misshapen cells that replace the notochord, while expression of Lmx-like remains relatively unaltered ( Jos
e-Edwards et al., 2011). These results suggest that Lmx-like, which is robustly expressed in notochord cells starting from gastrulation and throughout the main steps of notochord formation, might be part of a branch of the notochord GRN that is completely or partially independent of Bra ( Jos
e-Edwards et al., 2011). Considerable effort has been directed toward drawing edges between these regulatory nodes of the GRN and their respective effectors. Fig. 4 includes partial lists of target genes that have been identified for a few of the notochord transcription factors, as well as genes whose notochord CRMs have been completely characterized and have guided the elucidation of the transcription factors controlling them (detailed in Section 6). The notochord-specific expression of Bra has enabled the successful identification of notochord genes controlled by this transcription factor through its ectopic expression in neural and endodermal precursors. Transgenic embryos carrying a fusion of the Foxa.a promoter region to the Bra cDNA ectopically express Bra in endoderm and neural precursors and display a characteristic phenotype. These embryos were isolated for RNA extraction, followed by subtractive hybridization with RNAs extracted from stage-matched wild-type controls (Takahashi et al., 1999). This study identified 50 bona fide notochord transcriptional targets of Bra, including tropomyosin-like, prickle, ERM, leprecan/P3H1, Noto4/PID1, CiFCol1, multidom, ACL, b4GalT, and several others, whose function and/or transcriptional regulation have been studied in depth (Di Gregorio & Levine, 1999; Dunn & Di Gregorio, 2009; Hotta et al., 1999, 2000; Hotta, Takahashi, Satoh, & Gojobori, 2008; Hotta, Yamada, et al., 2007; Jiang et al., 2005; Katikala et al., 2013; Maguire et al., 2018; Oda-Ishii &

354

Anna Di Gregorio

Di Gregorio, 2007; Takahashi et al., 1999, 2010). This experiment has been more recently repeated using RNA-Seq, and has identified 925 putative Bra targets; however, many of these genes are still uncharacterized, and only a fraction of those genes for which expression patterns are available are expressed in the notochord (Reeves et al., 2017). Furthermore, ChIP-chip experiments have provided a genome-wide map of Bra-bound genomic loci, indicating that approximately 2100 individual genes are occupied by a transgenic Bra protein in embryos at the 112-cell stage; 194 of these genes encode for transcription factors, including Foxa.a (Kubo et al., 2010). Surprisingly, the overlap between the putative Bra targets identified through these surveys is limited to 459 genes (Reeves et al., 2017). We suggest that this discrepancy might be attributable to the technical differences in the methods employed for these experiments (subtractive hybridization, ChIP-chip and FACS/RNA-Seq) and also to the differences in the developmental stages that were assayed. These observations could also indicate that the Bra-downstream transcriptional targets constitute a large contingent of genes that changes dynamically as notochord morphogenesis proceeds, possibly in response to changes in chromatin configuration and accessibility, or to fluctuations in the levels of functional Bra protein. The very limited overlap between Bra target genes datasets identified through ChIP experiments in mouse (1942 Bra targets), zebrafish (218 Bra targets) and Xenopus tropicalis (1040 Bra targets) supports this hypothesis (Lolas et al., 2014). Furthermore, the analysis of human BRA target genes identified via ChIP-Seq in human ES cells has identified a group of roughly 800 BRA targets that appear specific to this population (Faial et al., 2015). In the case of Foxa.a, ChIP-chip experiments indicate that 3653 regions of the Ciona genome are bound by this transcription factor in 112-cell embryos, including the loci of 245 transcription factor genes, among which are Bra and Foxa.a itself (Kubo et al., 2010). Subtractive microarray screens were used to investigate the gene battery downstream of Tbx2/3, the only other T-box transcription factor reportedly expressed in notochord cells, which, differently from Bra, is not notochord-specific and is expressed also in various regions of CNS and epidermis (Imai et al., 2004; Jos
e-Edwards et al., 2013; Takatori et al., 2004). This survey led to the identification of 20 notochord genes whose expression is influenced by this transcription factor, and of 61 genes that are controlled by Tbx2/3 in other territories encompassed by its expression ( Jos
e-Edwards et al., 2013). The Tbx2/3 locus is bound by both

The notochord gene regulatory network in chordate evolution

355

Bra and Foxa.a (Kubo et al., 2010) and expression of Tbx2/3 in the notochord is lost in Bra mutant embryos, while it remains unaltered in CNS and epidermis ( Jos
e-Edwards et al., 2013). The characterization of a notochord CRM isolated from the Tbx2/3 locus revealed that this transcription factor is indeed regulated by Bra through cooperative binding sites ( Jos
e-Edwards et al., 2013; Fig. 2H). Part of the Tbx2/3 targets are shared with Bra, which suggests that Tbx2/3 controls part of the Bra-downstream notochord GRN; these genes include Noto4/PID1, which is required for notochord intercalation (Yamada, Ueno, Satoh, & Takahashi, 2011) and whose notochord CRM is controlled through an individual functional Bra binding site (Fig. 2H; Katikala et al., 2013), and fibronectin, whose notochord CRM contains binding sites for transcription factors of the T-box and Fox families, and is required for intercalation as well (Segade et al., 2016). These findings are in agreement with the effects of the over-expression of mutant forms of Tbx2/3 on notochord intercalation ( Jos
e-Edwards et al., 2013). It is noteworthy that the consensus binding sites determined through SELEX assays for Bra and Tbx2/3 are very similar (Nitta et al., 2019), and this suggests the hypothesis that related T-box binding sites might be used interchangeably by Bra in early embryos and by Tbx2/3 at later stages. It is also possible that slight differences in sequence between these consensus binding sites might determine more selective binding by these transcription factors in vivo. Microarray screens were employed also to determine the identities of notochord genes controlled by Bhlh-tun1; the locus of this transcription factor is bound by both Bra and Foxa.a in early embryos (Kubo et al., 2010) and notochord expression of Bhlh-tun1 is lost in Bra mutants (Kugler et al., 2019). Its minimal notochord CRM requires both Bra and Foxa.a binding sites for its function, along with a binding site for a homeodomain-containing factor (Kugler et al., 2019). Among the 21 notochord genes whose transcription is influenced by Bhlh-tun1, 16 are shared with Bra and encode for ECM components, transmembrane transporters, effectors of cell-shape changes, and a few enzymes; one of the Bhlh-tun1-downstream genes is Claudin16/17/19, whose product is likely involved in joining together notochord cells at the onset of tubulogenesis (Section 3.2.4) (Kugler et al., 2019). Lastly, microarray screens and RNA-Seq experiments have revealed numerous notochord target genes for Xbp1 (Kugler et al., 2008; Fig. 4; our unpublished results) and NFAT5 ( Jos
e-Edwards et al., 2011; our unpublished results).

356

Anna Di Gregorio

Aside from the interactions between Bra and Foxa.a, little is known about possible cross-regulatory interactions between other transcription factors within the notochord GRN. As discussed above, the cross-regulatory interaction between Bra and Foxa.a (Fig. 4) is presumed on the basis of chromatin occupancy and morpholino knockdown results; however, it has been suggested that the regulation of Bra by Foxa.a could be achieved indirectly through FoxD and ZicL, since the promoters of both these genes are occupied by Foxa.a in early embryos (Kubo et al., 2010). The positive regulation/feedback of Foxa.a by Bra remains to be confirmed in Ciona, although it could be predicted on the basis of the occupancy of the Foxa. a locus by Bra, observed through ChIP-chip assays (Kubo et al., 2010), and of the results obtained in mouse embryoid bodies, which indicate that Foxa2 is a direct Bra target and its transcription is activated by it (Lolas et al., 2014). Microarray screens revealed that Tbx2/3 represses expression of Fos-a, and that Bhlh-tun1 activates Lhx3/4/5, which is, however, characterized by a very short window of expression in notochord precursors ( Jos
eEdwards et al., 2013; Kugler et al., 2019). Additional cross-regulatory interactions will likely surface as more cis-regulatory regions and gene batteries are characterized for more notochord transcription factors.

6. Reverse-engineering the notochord GRN: Lessons from notochord cis-regulatory modules Over the past two decades, a straightforward electroporation protocol for transgenesis, the availability of Ciona embryos and their rapid development have facilitated the discovery of a large number of cis-regulatory regions active in notochord cells, along with numerous enhancers active in other tissues (reviewed in Di Gregorio & Levine, 2002; Irvine, 2013; Kusakabe, 2005; Wang & Christiaen, 2012). Nearly 40 notochord cisregulatory regions have been identified and systematically dissected, and their minimal functional sequences have been used to reverse-engineer the notochord GRN through the identification of the transcription factors responsible for their activity (Anno, Satou, & Fujiwara, 2006; Christiaen et al., 2008; Corbo, Levine, et al., 1997; Di Gregorio & Levine, 1999; Dunn & Di Gregorio, 2009; Farley et al., 2016; Harder et al., 2019; Jos
e-Edwards et al., 2015, 2013; Katikala et al., 2013; Kugler et al., 2019; Passamaneck et al., 2009; Segade et al., 2016; Thompson & Di Gregorio, 2015; our unpublished results). Many of these CRMs rely upon

The notochord gene regulatory network in chordate evolution

357

Bra/T-box binding sites for their function (Fig. 2H, Fig. 4), and the relationship between their temporal onsets and the presence of multiple/ individual functional Bra binding sites within their minimal sequences have been discussed in Section 4 (Fig. 2H).

6.1 Ciona notochord CRMs can be synergistically activated by Foxa.a in cooperation with either Bra or other transcription factors The characterization of the Ci-tune CRM provided the first evidence of a notochord CRM controlled synergistically by orthologs of Brachyury and Foxa2 (Passamaneck et al., 2009), followed by the identification of two additional CRMs controlled in a similar way in a subsequent study ( Jos
eEdwards et al., 2015). The results of genome-wide ChIP-chip assays indicate that approximately 1020 individual Ciona loci are occupied by both Bra and Foxa.a in early embryos (Kubo et al., 2010). Apart from synergizing with Bra, Foxa.a is able to trigger notochord gene expression through multiple binding sites (Anno et al., 2006), or by acting in concert with other transcription factors. The analysis of a Ciona notochord CRM, Ci-CRM112, which is associated with an olfactomedin-like gene, indicates that Foxa.a can work in concert with transcription factors of the homeodomain (e.g., Lmx-like) and AP1 family (e.g., Fos-a) ( Jos
eEdwards et al., 2015; dashed violet and orange lines in Fig. 4). These results are consistent with the chromatin-opening and remodeling properties that characterize transcription factors of the Fox family (e.g., Lalmansingh, Karmakar, Jin, & Nagaich, 2012).

6.2 Additional notochord cis-regulatory mechanisms identified in Ciona The notochord CRM that is associated with the early homeodomain transcription factor gene Mnx relies upon a combination of binding sites for Ets and ZicL, as does another notochord CRM located upstream of the main Bra enhancer/promoter region (Farley et al., 2016; Matsumoto, Kumano, & Nishida, 2007; Yagi et al., 2004). Furthermore, at least three notochord CRMs rely for their activity upon minimal sequences that resemble binding sites for transcription factors of the Myb family, while one CRM, Ci-CRM26, requires a binding site for a basic helix-loop-helix (bHLH) transcription factor ( Jos
e-Edwards et al., 2015). Regulation of notochord gene expression in the secondary notochord cells requires multiple inputs, both negative and positive. The notochord

358

Anna Di Gregorio

cis-regulatory region associated with the gene KH.C11.331, which encodes for a putative ECM component of the fibulin/hemicentin family, contains a silencer that prevents expression in primary notochord cells, as well as sequences that mediate its response to FGF and Wnt signaling pathways, both of which control patterning of the posterior regions of the tail (Harder et al., 2019). Even though the molecular players responsible for the transcriptional repression of these genes in the primary lineage are yet to be identified, this finding lays the foundation for future studies of the molecular differences between primary and secondary notochord in ascidians.

6.3 Comparative studies of notochord cis-regulatory regions across chordates The lack of self-evident linear conservation between the cis-regulatory regions identified in Ciona and those identified in other chordates can be overcome by focusing on functional transcription factor binding sites as the main building blocks and descriptors of notochord CRMs. This reductionist approach has allowed us to begin a comparative study of notochord cis-regulatory regions across chordates, and to tentatively group Bra- and/or Foxa2-dependent notochord CRMs from Ciona and from vertebrates within a chordate-wide, basal strategy of cis-regulatory control of notochord gene expression ( Jos
e-Edwards et al., 2015). This group of CRMs is currently the largest, and includes CRMs containing Bra/T-box binding sites, Fox binding sites, or combinations of both. In vertebrates, notochord cis-regulatory regions that rely upon Bra binding sites for their function are the Sonic hedgehog (Shh) ar-C intronic notochord and floor plate enhancer identified in zebrafish (M€ uller et al., 1999) and the eFGF promoter/ enhancer region identified in Xenopus (Casey, O’Reilly, Conlon, & Smith, 1998; Tada, Casey, Fairclough, & Smith, 1998). An intriguing variation on this theme was identified through the analysis of the Ciona Ephrin3 notochord CRM (Fig. 4), which necessitates both a Bra binding site and an (AC)6 microsatellite repeat ( Jos
e-Edwards et al., 2015). This unexpected association of a Bra binding site(s) with a (AC)6 repetitive sequence has been demonstrated, through ChIP-chip assays, to be bound by mouse Bra in embryonic stem cells (Evans et al., 2012). The results obtained in Ciona predict that at least a fraction of these regions within the mouse genome might possess cis-regulatory activity. To date, notochord CRMs equally dependent upon both Brachyury and Foxa2 binding sites have been identified only in Ciona, although it seems

The notochord gene regulatory network in chordate evolution

359

predictable that this configuration might be present in vertebrate cisregulatory regions as well, given that the functions of these transcription factors, their co-expression in the notochord, and their binding sites are evolutionarily conserved across chordates. Numerous notochord CRMs regulated exclusively through multiple Foxa2 binding sites have been identified in mouse embryos, through a survey centered on Foxa2-downstream genes (Tamplin, Cox, & Rossant, 2011). In most of the Fox-containing CRMs that have been identified, Foxa2 binding sites work cooperatively, as is the case for mouse Pkd1/1-1, Shh, Bicc1-1, which contain five, three and two Foxa2 binding sites, respectively ( Jeong & Epstein, 2003; Tamplin et al., 2011); however, the Sox9 E1 enhancer contains an individual Foxa2 binding site (Bagheri-Fam et al., 2006). In zebrafish, a few notochord CRMs contain individual Foxa2 binding sites functionally associated with a recurrent sequence motif (Rastegar et al., 2008). In addition to the chordate-wide Bra- and/or Foxa2-dependent notochord CRMs, the Ciona genome contains notochord CRMs that depend upon binding sites for transcription factors of the bHLH, Myb, Zic, Ets and homeodomain families (Farley et al., 2016; Jos
e-Edwards et al., 2015; Katikala et al., 2013); notochord CRMs of these categories are yet to be identified in vertebrates. Conversely, a few vertebrate notochord CRMs contain sequences that are yet to be identified in Ciona, such as the orphan binding site (OBS) contained in the node and nascent notochord enhancer (NOCE) of mouse Noto (Alten et al., 2012), and the binding site for a Tead transcription factor in the Foxa2 notochord cis-regulatory region (Sawada et al., 2005). These findings hint at the existence of vertebrate-specific mechanisms of activation of notochord gene expression, which could have either evolved after the divergence of tunicates from the chordate lineage leading to vertebrates, or could have been present in a common chordate ancestor and might have been selectively lost in either ascidians or in all tunicates.

7. A comparative view of the notochord GRN in Ciona and other chordates Ascidian embryos lack an organizer, a structure analogous to the dorsal lip of the blastopore in Xenopus, or an equivalent signaling source that is physically distinguishable, can be transplanted from embryo to embryo and is sufficient to induce formation of ectopic structures in its new embryonic context (e.g., Anderson & Stern, 2016). Nevertheless, several of the

360

Anna Di Gregorio

signaling molecules involved in vertebrate notochord induction, such as chordin, bone morphogenetic protein (BMP) and FGF are involved in the early events of notochord specification in Ciona (discussed in Kourakis & Smith, 2005; Passamaneck & Di Gregorio, 2005; Satou, 2020) and in Halocynthia (Darras & Nishida, 2001). After its specification, the Ciona notochord GRN shares with its vertebrate counterparts the central role of Bra and Foxa2. However, some of the key components of vertebrate GRNs appear to be missing from the Ciona notochord GRN. Among them is the homeodomain transcription factor Noto, which is required for notochord formation across chordates, from zebrafish to mouse (e.g., Morley et al., 2009; Talbot et al., 1995; Zizic Mitrecic, Mitrecic, Pochet, Kostovic-Knezevic, & Gajovic, 2010). Although a putative Ciona Noto gene has been identified, its expression in the notochord appears very weak, transient, and limited to the posteriormost region of the tail in the early tailbud stage. On the other hand, in Halocynthia roretzi, expression of Hr-Noto appears stronger, although it is still limited to the posterior-most primary notochord and all secondary notochord cells (Utsumi et al., 2004). Besides Noto, other homeodomain transcription factors, namely, representatives of the Hox and Pbx subfamilies, which are found in vertebrate notochords (e.g., Capellini et al., 2008; Wang et al., 2014), seem remarkably absent from the Ciona notochord (e.g., Keys et al., 2005; Satou et al., 2001). Ciona possesses an incomplete, fragmented Hox cluster, whose components are expressed in a variety of pre- and post-metamorphic tissues and structures, ranging from the larval epidermis and nervous system to the juvenile digestive tract and spermiduct sensory organ, with the notable exception of the notochord and its precursors (Di Gregorio et al., 1995; Ikuta, Yoshida, Satoh, & Saiga, 2004; Keys et al., 2005; Passamaneck & Di Gregorio, 2005; Spagnuolo et al., 2003; Tajima et al., 2020). However, the absence of Hox genes expression from the Ciona notochord contrasts with findings in the amphioxus Branchiostoma lanceolatum, where expression of Hox14 is observed in the posterior notochord and is influenced by retinoic acid (Pascual-Anaya et al., 2012); even among other tunicates, the larvacean Oikopleura dioica displays colinear Hox gene expression in the notochord (Seo et al., 2004). Together, these data favor the hypothesis of a specific loss of Hox gene expression in the Ciona notochord and possibly in the notochord of other ascidians. Overall, the entire homeobox gene family appears underrepresented in the Ciona notochord, except for the early expression of Mnx and Lhx3/4/5 and the more sustained expression of Ciona Islet

The notochord gene regulatory network in chordate evolution

361

(Giuliano, Marino, Pinto, & De Santis, 1998; Imai et al., 2004). It is plausible that transient, weak expression of additional homeobox genes will be revealed through the validation of recent scRNA-Seq data (Cao et al., 2019; Horie et al., 2018). Another remarkably divergent characteristic of the Ciona notochord is the absence of members of the Sonic hedgehog (Shh) pathway (Takatori, Satou, & Satoh, 2002), whose orthologs are present in the genome but are expressed in the visceral ganglion and pharyngeal endoderm of the larva (Islam, Moly, Miyamoto, & Kusakabe, 2010). In vertebrates, Shh is abundantly expressed in both notochord and floor plate (Fig. 1C), and is required for notochordal sheath formation and patterning of the nuclei pulposi (Choi & Harfe, 2011); its down-regulation has been associated with degeneration of the intervertebral discs (Rajesh & Dahia, 2018). In amphioxus, AmphiHh and other members of the Shh pathway, Patched, Smoothened and Suppressor of Fused, are expressed in the notochord and additional structures, thus more closely resembling the situation of vertebrates (Lin et al., 2009; Shimeld, 1999). Hence, also the loss of the Shh pathway from the notochord seems to have occurred selectively in the ascidian lineage.

8. Concluding remarks and future perspectives More tunicate genomes are currently being sequenced and compared, and together with the elucidation of novel gene expression patterns, the results of these investigations are expected to pinpoint the evolutionary timing of the loss of Noto, Hox and Shh from the ascidian notochord. The much-needed characterization of additional notochord CRMs in vertebrates and cephalochordates will expand the comparative studies of notochord cis-regulatory regions that we have initiated using Ciona, and will increase the knowledge of the repertoire of cis-regulatory strategies employed by divergent chordates to achieve notochord gene expression. These comparative studies will also highlight clade-specific molecular mechanisms that are responsible for the morphological and functional differences in the notochords of different chordates, and will sharpen the distinction between basal chordate-wide cis-regulatory mechanisms, cladespecific divergent strategies, and vertebrate innovations. Clustered regularly interspaced short palindromic repeats and Cas9 endonuclease (CRISPR/Cas9)-mediated genome editing (Sasaki, Yoshida, Hozumi, & Sasakura, 2014; Stolfi, Gandhi, Salek, & Christiaen, 2014) is being successfully employed to delete cis-regulatory regions active

362

Anna Di Gregorio

in heart and cardiopharyngeal lineages (Racioppi et al., 2019) and to ablate notochord CRMs (our unpublished results), and will clarify the functional requirement and pervasiveness of redundant or additive notochord cisregulatory regions. Ongoing studies on the role of non-coding RNAs (Chen, Pedro, & Zeller, 2011; reviewed in Velandia-Huerto, Brown, Gittenberger, Stadler, & Bermu´dez-Santana, 2018) and on the function of transient, post-specification expression of transcriptional repressors in the notochord (Erives, Corbo, & Levine, 1998), together with further studies of Braindependent gene regulation ( Jos
e-Edwards et al., 2011; Kugler et al., 2008) will increase the resolution and depth of the sophisticated GRN that powers the simple notochord of Ciona.

Acknowledgments Thanks to all present and past lab members and collaborators. I am particularly grateful to Emerita Prof. Kathleen Sulik (University of North Carolina), to Prof. Jean-Pierre SaintJeannet (NYU College of Dentistry), Ms. Yushi Wu, Drs. Jordan M. Thompson, Yale Passamaneck, and Julie Maguire, for generously sharing their original microphotographs. I remain indebted to Drs. Eric Davidson and Michael Levine for invaluable discussions and for their contagious passion for regulatory sequences and gene regulatory networks. Special thanks to Cettina and Franco Di Gregorio and to Ms. Marianna Nibu for encouragement and inspiration. Research in our lab is currently supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health, under award number R03HD098395.

References Adams, D. S., Keller, R., & Koehl, M. A. (1990). The mechanics of notochord elongation, straightening and stiffening in the embryo of Xenopus laevis. Development, 110(1), 115–130. Almaza´n, A., Ferra´ndez-Rolda´n, A., Albalat, R., & Can˜estro, C. (2019). Developmental atlas of appendicularian Oikopleura dioica actins provides new insights into the evolution of the notochord and the cardio-paraxial muscle in chordates. Developmental Biology, 448(2), 260–270. Alten, L., Schuster-Gossler, K., Eichenlaub, M. P., Wittbrodt, B., Wittbrodt, J., & Gossler, A. (2012). A novel mammal-specific three partite enhancer element regulates node and notochord-specific Noto expression. PLoS One, 7(10), e47785. Anderson, C., & Stern, C. D. (2016). Organizers in development. Current Topics in Developmental Biology, 117, 435–454. Ang, S. L., & Rossant, J. (1994). HNF-3 beta is essential for node and notochord formation in mouse development. Cell, 78, 561–574. Anno, C., Satou, A., & Fujiwara, S. (2006). Transcriptional regulation of ZicL in the Ciona intestinalis embryo. Development Genes and Evolution, 216(10), 597–605. Annona, G., Holland, N. D., & D’Aniello, S. (2015). Evolution of the notochord. EvoDevo, 6, 30.

The notochord gene regulatory network in chordate evolution

363

Bagheri-Fam, S., Barrionuevo, F., Dohrmann, U., G€ unther, T., Sch€ ule, R., Kemler, R., et al. (2006). Long-range upstream and downstream enhancers control distinct subsets of the complex spatiotemporal Sox9 expression pattern. Developmental Biology, 291(2), 382–397. Bell, A. H., DeMonte, F., Raza, S. M., Rhines, L. D., Tatsui, C. E., Prieto, V. G., et al. (2018). Transcriptome comparison identifies potential biomarkers of spine and skull base chordomas. Virchows Archiv, 472(3), 489–497. Bone, Q. (1992). On the locomotion of ascidian tadpole larvae. Journal of the Marine Biological Association of the United Kingdom, 72(1), 161–186. Brown, F. D., Prendergast, A., & Swalla, B. J. (2008). Man is but a worm: Chordate origins. Genesis, 46(11), 605–613. Brunet, T., Lauri, A., & Arendt, D. (2015). Did the notochord evolve from an ancient axial muscle? The axochord hypothesis. BioEssays, 37(8), 836–850. Cao, C., Lemaire, L. A., Wang, W., Yoon, P. H., Choi, Y. A., Parsons, L. R., et al. (2019). Comprehensive single-cell transcriptome lineages of a proto-vertebrate. Nature, 571(7765), 349–354. Capellini, T. D., Dunn, M. P., Passamaneck, Y. J., Selleri, L., & Di Gregorio, A. (2008). Conservation of notochord gene expression across chordates: Insights from the Leprecan gene family. Genesis, 46(11), 683–696. Capellini, T. D., Zewdu, R., Di Giacomo, G., Asciutti, S., Kugler, J. E., Di Gregorio, A., et al. (2008). Pbx1/Pbx2 govern axial skeletal development by controlling Polycomb and Hox in mesoderm and Pax1/Pax9 in sclerotome. Developmental Biology, 321(2), 500–514. Casey, E. S., O’Reilly, M. A., Conlon, F. L., & Smith, J. C. (1998). The T-box transcription factor Brachyury regulates expression of eFGF through binding to a non-palindromic response element. Development, 125(19), 3887–3894. Chen, J. S., Pedro, M. S., & Zeller, R. W. (2011). miR-124 function during Ciona intestinalis neuronal development includes extensive interaction with the Notch signaling pathway. Development, 138(22), 4943–4953. Chiba, S., Jiang, D., Satoh, N., & Smith, W. C. (2009). Brachyury null mutant-induced defects in juvenile ascidian endodermal organs. Development, 136(1), 35–39. Choi, K. S., Cohn, M. J., & Harfe, B. D. (2008). Identification of nucleus pulposus precursor cells and notochordal remnants in the mouse: Implications for disk degeneration and chordoma formation. Developmental Dynamics, 237, 3953–3958. Choi, K. S., & Harfe, B. D. (2011). Hedgehog signaling is required for formation of the notochord sheath and patterning of nuclei pulposi within the intervertebral discs. Proceedings of the National Academy of Sciences of the United States of America, 108(23), 9484–9489. Christiaen, L., Davidson, B., Kawashima, T., Powell, W., Nolla, H., Vranizan, K., et al. (2008). The transcription/migration interface in heart precursors of Ciona intestinalis. Science, 320(5881), 1349–1352. Christiaen, L., Wagner, E., Shi, W., & Levine, M. (2009). Isolation of individual cells and tissues from electroporated sea squirt (Ciona) embryos by fluorescence-activated cell sorting (FACS). Cold Spring Harbor Protocols, (12), pdb.prot5349. Cleaver, O., & Krieg, P. A. (2001). Notochord patterning of the endoderm. Developmental Biology, 234(1), 1–12. Cloney, R. A. (1964). Development of the ascidian notochord. Acta Embryologiae et Morphologiae Experimentalis, 7, 111–130. Conklin, E. G. (1905). The organization and cell-lineage of the ascidian egg. Journal of the Academy of Natural Sciences of Philadelphia, 13, 1–119. Conklin, E. G. (1928). The embryology of amphioxus. Journal of Morphology, 54, 69–151. Corallo, D., Trapani, V., & Bonaldo, P. (2015). The notochord: Structure and functions. Cellular and Molecular Life Sciences, 72(16), 2989–3008.

364

Anna Di Gregorio

Corbo, J. C., Erives, A., Di Gregorio, A., Chang, A., & Levine, M. (1997). Dorsoventral patterning of the vertebrate neural tube is conserved in a protochordate. Development, 124(12), 2335–2344. Corbo, J. C., Fujiwara, S., Levine, M., & Di Gregorio, A. (1998). Suppressor of hairless activates brachyury expression in the Ciona embryo. Developmental Biology, 203(2), 358–368. Corbo, J. C., Levine, M., & Zeller, R. W. (1997). Characterization of a notochord-specific enhancer from the Brachyury promoter region of the ascidian, Ciona intestinalis. Development, 124(3), 589–602. Darras, S., & Nishida, H. (2001). The BMP signaling pathway is required together with the FGF pathway for notochord induction in the ascidian embryo. Development, 128(14), 2629–2638. Davidson, E. H. (2006). The regulatory genome: Gene regulatory networks in development and evolution. San Diego: Academic Press. Davidson, E. H. (2010). Emerging properties of animal gene regulatory networks. Nature, 468(7326), 911–920. Davidson, E. H., & Erwin, D. H. (2006). Gene regulatory networks and the evolution of animal body plans. Science, 311(5762), 796–800. de Bree, K., de Bakker, B. S., & Oostra, R. J. (2018). The development of the human notochord. PLoS One, 13(10), e0205752. de Sousa, C. S. M., de Castro, B. B., de Miranda, C. L. V. M., Bastos, B. B., & Avelino, M. C. (2018). Split notochord syndrome: A case in point. Radiologia Brasileira, 51(1), 59–60. Dehal, P., Satou, Y., Campbell, R. K., Chapman, J., Degnan, B., De Tomaso, A., et al. (2002). The draft genome of Ciona intestinalis: Insights into chordate and vertebrate origins. Science, 298(5601), 2157–2167. Delsuc, F., Brinkmann, H., Chourrout, D., & Philippe, H. (2006). Tunicates and not cephalochordates are the closest living relatives of vertebrates. Nature, 439(7079), 965–968. Deng, W., Nies, F., Feuer, A., Bocina, I., Oliver, D., & Jiang, D. (2013). Anion translocation through an Slc26 transporter mediates lumen expansion during tubulogenesis. Proceedings of the National Academy of Sciences of the United States of America, 110(37), 14972–14977. Denker, E., Bocina, I., & Jiang, D. (2013). Tubulogenesis in a simple cell cord requires the formation of bi-apical cells through two discrete Par domains. Development, 140(14), 2985–2996. Denker, E., & Jiang, D. (2012). Ciona intestinalis notochord as a new model to investigate the cellular and molecular mechanisms of tubulogenesis. Seminars in Cell & Developmental Biology, 23(3), 308–319. Denker, E., Sehring, I. M., Dong, B., Audisso, J., Mathiesen, B., & Jiang, D. (2015). Regulation by a TGFβ-ROCK-actomyosin axis secures a non-linear lumen expansion that is essential for tubulogenesis. Development, 142(9), 1639–1650. Di Gregorio, A. (2017). T-box genes and developmental gene regulatory networks in ascidians. Current Topics in Developmental Biology, 122, 55–91. Di Gregorio, A., Corbo, J. C., & Levine, M. (2001). The regulation of forkhead/HNF-3beta expression in the Ciona embryo. Developmental Biology, 229(1), 31–43. Di Gregorio, A., Harland, R. M., Levine, M., & Casey, E. S. (2002). Tail morphogenesis in the ascidian, Ciona intestinalis, requires cooperation between notochord and muscle. Developmental Biology, 244(2), 385–395. Di Gregorio, A., & Levine, M. (1999). Regulation of Ci-tropomyosin-like, a Brachyury target gene in the ascidian, Ciona intestinalis. Development, 126(24), 5599–5609. Di Gregorio, A., & Levine, M. (2002). Analyzing gene regulation in ascidian embryos: New tools for new perspectives. Differentiation, 70(4–5), 132–139. Di Gregorio, A., Spagnuolo, A., Ristoratore, F., Pischetola, M., Aniello, F., Branno, M., et al. (1995). Cloning of ascidian homeobox genes provides evidence for a primordial chordate cluster. Gene, 156(2), 253–257.

The notochord gene regulatory network in chordate evolution

365

Dong, B., Deng, W., & Jiang, D. (2011). Distinct cytoskeleton populations and extensive crosstalk control Ciona notochord tubulogenesis. Development, 138(8), 1631–1641. Dong, B., Horie, T., Denker, E., Kusakabe, T., Tsuda, M., Smith, W. C., et al. (2009). Tube formation by complex cellular processes in Ciona intestinalis notochord. Developmental Biology, 330(2), 237–249. Dunn, M. P., & Di Gregorio, A. (2009). The evolutionarily conserved leprecan gene: Its regulation by Brachyury and its role in the developing Ciona notochord. Developmental Biology, 328(2), 561–574. Erives, A., Corbo, J. C., & Levine, M. (1998). Lineage-specific regulation of the Ciona snail gene in the embryonic mesoderm and neuroectoderm. Developmental Biology, 194(2), 213–225. Evans, A. L., Faial, T., Gilchrist, M. J., Down, T., Vallier, L., Pedersen, R. A., et al. (2012). Genomic targets of Brachyury (T) in differentiating mouse embryonic stem cells. PLoS One, 7(3), e33346. Faial, T., Bernardo, A. S., Mendjan, S., Diamanti, E., Ortmann, D., Gentsch, G. E., et al. (2015). Brachyury and SMAD signalling collaboratively orchestrate distinct mesoderm and endoderm gene regulatory networks in differentiating human embryonic stem cells. Development, 142(12), 2121–2135. Farley, E. K., Olson, K. M., Zhang, W., Rokhsar, D. S., & Levine, M. S. (2016). Syntax compensates for poor binding sites to encode tissue specificity of developmental enhancers. Proceedings of the National Academy of Sciences of the United States of America, 113(23), 6508–6513. Fujiwara, S., Maeda, Y., Shin-I, T., Kohara, Y., Takatori, N., Satou, Y., et al. (2002). Gene expression profiles in Ciona intestinalis cleavage-stage embryos. Mechanisms of Development, 112(1–2), 115–127. Giuliano, P., Marino, R., Pinto, M. R., & De Santis, R. (1998). Identification and developmental expression of Ci-isl, a homologue of vertebrate islet genes, in the ascidian Ciona intestinalis. Mechanisms of Development, 78(1–2), 199–202. Harder, M., Reeves, W., Byers, C., Santiago, M., & Veeman, M. (2019). Multiple inputs into a posterior-specific regulatory network in the Ciona notochord. Developmental Biology, 448(2), 136–146. Hartvigsen, J., Hancock, M. J., Kongsted, A., Louw, Q., Ferreira, M. L., Genevay, S., et al. (2018). What low back pain is and why we need to pay attention. Lancet, 391(10137), 2356–2367. Herwig, L., Blum, Y., Krudewig, A., Ellertsdottir, E., Lenard, A., Belting, H. G., et al. (2011). Distinct cellular mechanisms of blood vessel fusion in the zebrafish embryo. Current Biology, 21(22), 1942–1948. Horie, R., Hazbun, A., Chen, K., Cao, C., Levine, M., & Horie, T. (2018). Shared evolutionary origin of vertebrate neural crest and cranial placodes. Nature, 560(7717), 228–232. Hotta, K., Mitsuhara, K., Takahashi, H., Inaba, K., Oka, K., Gojobori, T., et al. (2007). A web-based interactive developmental table for the ascidian Ciona intestinalis, including 3D real-image embryo reconstructions: I. From fertilized egg to hatching larva. Developmental Dynamics, 236(7), 1790–1805. Hotta, K., Takahashi, H., Asakura, T., Saitoh, B., Takatori, N., Satou, Y., et al. (2000). Characterization of Brachyury-downstream notochord genes in the Ciona intestinalis embryo. Developmental Biology, 224(1), 69–80. Hotta, K., Takahashi, H., Erives, A., Levine, M., & Satoh, N. (1999). Temporal expression patterns of 39 Brachyury-downstream genes associated with notochord formation in the Ciona intestinalis embryo. Development, Growth & Differentiation, 41(6), 657–664.

366

Anna Di Gregorio

Hotta, K., Takahashi, H., Satoh, N., & Gojobori, T. (2008). Brachyury-downstream gene sets in a chordate, Ciona intestinalis: Integrating notochord specification, morphogenesis and chordate evolution. Evolution & Development, 10(1), 37–51. Hotta, K., Yamada, S., Ueno, N., Satoh, N., & Takahashi, H. (2007). Brachyurydownstream notochord genes and convergent extension in Ciona intestinalis embryos. Development, Growth & Differentiation, 49(5), 373–382. Hudson, C., Kawai, N., Negishi, T., & Yasuo, H. (2013). β-Catenin-driven binary fate specification segregates germ layers in ascidian embryos. Current Biology, 23(6), 491–495. Hudson, C., & Yasuo, H. (2005). Patterning across the ascidian neural plate by lateral Nodal signalling sources. Development, 132(6), 1199–1210. Hudson, C., & Yasuo, H. (2006). A signalling relay involving Nodal and Delta ligands acts during secondary notochord induction in Ciona embryos. Development, 133(15), 2855–2864. Ikeda, T., & Satou, Y. (2017). Differential temporal control of Foxa.a and Zic-r.b specifies brain versus notochord fate in the ascidian embryo. Development, 144(1), 38–43. Ikuta, T., Yoshida, N., Satoh, N., & Saiga, H. (2004). Ciona intestinalis Hox gene cluster: Its dispersed structure and residual colinear expression in development. Proceedings of the National Academy of Sciences of the United States of America, 101(42), 15118–15123. Imai, K. S., Hino, K., Yagi, K., Satoh, N., & Satou, Y. (2004). Gene expression profiles of transcription factors and signaling molecules in the ascidian embryo: Towards a comprehensive understanding of gene networks. Development, 131(16), 4047–4058. Imai, K. S., Levine, M., Satoh, N., & Satou, Y. (2006). Regulatory blueprint for a chordate embryo. Science, 312(5777), 1183–1187. Imai, K. S., Satoh, N., & Satou, Y. (2002). An essential role of a FoxD gene in notochord induction in Ciona embryos. Development, 129(14), 3441–3453. Imai, K. S., Satou, Y., & Satoh, N. (2002). Multiple functions of a Zic-like gene in the differentiation of notochord, central nervous system and muscle in Ciona savignyi embryos. Development, 129(11), 2723–2732. Imai, K., Takada, N., Satoh, N., & Satou, Y. (2000). (Beta)-catenin mediates the specification of endoderm cells in ascidian embryos. Development, 127(14), 3009–3020. Irvine, S. Q. (2013). Study of Cis-regulatory elements in the ascidian Ciona intestinalis. Current Genomics, 14(1), 56–67. Islam, A. F., Moly, P. K., Miyamoto, Y., & Kusakabe, T. G. (2010). Distinctive expression patterns of Hedgehog pathway genes in the Ciona intestinalis larva: Implications for a role of Hedgehog signaling in postembryonic development and chordate evolution. Zoological Science, 27(2), 84–90. Ito, M., Katada, T., Miyatani, S., & Kinoshita, T. (2007). XSu(H)2 is an essential factor for gene expression and morphogenesis of the Xenopus gastrula embryo. The International Journal of Developmental Biology, 51(1), 27–36. Jeong, Y., & Epstein, D. J. (2003). Distinct regulators of Shh transcription in the floor plate and notochord indicate separate origins for these tissues in the mouse node. Development, 130(16), 3891–3902. Jiang, D., Munro, E. M., & Smith, W. C. (2005). Ascidian prickle regulates both mediolateral and anterior-posterior cell polarity of notochord cells. Current Biology, 15(1), 79–85. Jiang, D., & Smith, W. C. (2007). Ascidian notochord morphogenesis. Developmental Dynamics, 236(7), 1748–1757. Johnson, Z. I., Shapiro, I. M., & Risbud, M. V. (2014). Extracellular osmolarity regulates matrix homeostasis in the intervertebral disc and articular cartilage: Evolving role of TonEBP. Matrix Biology, 40, 10–16. Jos
e-Edwards, D. S., Kerner, P., Kugler, J. E., Deng, W., Jiang, D., & Di Gregorio, A. (2011). The identification of transcription factors expressed in the notochord of Ciona intestinalis adds new potential players to the brachyury gene regulatory network. Developmental Dynamics, 240(7), 1793–1805.

The notochord gene regulatory network in chordate evolution

367

Jos
e-Edwards, D. S., Oda-Ishii, I., Kugler, J. E., Passamaneck, Y. J., Katikala, L., Nibu, Y., et al. (2015). Brachyury, Foxa2 and the cis-regulatory origins of the notochord. PLoS Genetics, 11(12), e1005730. Jos
e-Edwards, D. S., Oda-Ishii, I., Nibu, Y., & Di Gregorio, A. (2013). Tbx2/3 is an essential mediator within the Brachyury gene network during Ciona notochord development. Development, 140(11), 2422–2433. Karaiskou, A., Swalla, B. J., Sasakura, Y., & Chambon, J. P. (2015). Metamorphosis in solitary ascidians. Genesis, 53(1), 34–47. Katikala, L., Aihara, H., Passamaneck, Y. J., Gazdoiu, S., Jos
e-Edwards, D. S., Kugler, J. E., et al. (2013). Functional Brachyury binding sites establish a temporal read-out of gene expression in the Ciona notochord. PLoS Biology, 11(10), e1001697. Keller, R., Davidson, L., Edlund, A., Elul, T., Ezin, M., Shook, D., et al. (2000). Mechanisms of convergence and extension by cell intercalation. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 355(1399), 897–922. Keys, D. N., Lee, B. I., Di Gregorio, A., Harafuji, N., Detter, J. C., Wang, M., et al. (2005). A saturation screen for cis-acting regulatory DNA in the Hox genes of Ciona intestinalis. Proceedings of the National Academy of Sciences of the United States of America, 102(3), 679–683. Kier, W. M. (2012). The diversity of hydrostatic skeletons. The Journal of Experimental Biology, 215(Pt. 8), 1247–1257. Kispert, A., Koschorz, B., & Herrmann, B. G. (1995). The T protein encoded by Brachyury is a tissue-specific transcription factor. The EMBO Journal, 14(19), 4763–4772. Kjaer, I., Becktor, K. B., Nolting, D., & Fischer Hansen, B. (1997). The association between prenatal sella turcica morphology and notochordal remnants in the dorsum sellae. Journal of Craniofacial Genetics and Developmental Biology, 17(3), 105–111. Kourakis, M. J., Reeves, W., Newman-Smith, E., Maury, B., Abdul-Wajid, S., & Smith, W. C. (2014). A one-dimensional model of PCP signaling: Polarized cell behavior in the notochord of the ascidian Ciona. Developmental Biology, 395(1), 120–130. Kourakis, M. J., & Smith, W. C. (2005). Did the first chordates organize without the organizer? Trends in Genetics, 21(9), 506–510. Krasovec, G., Robine, K., Qu
einnec, E., Karaiskou, A., & Chambon, J. P. (2019). Ci-hox12 tail gradient precedes and participates in the control of the apoptotic-dependent tail regression during Ciona larva metamorphosis. Developmental Biology, 448(2), 237–246. Kubo, A., Suzuki, N., Yuan, X., Nakai, K., Satoh, N., Imai, K. S., et al. (2010). Genomic cis-regulatory networks in the early Ciona intestinalis embryo. Development, 137(10), 1613–1623. Kugler, J. E., Passamaneck, Y. J., Feldman, T. G., Beh, J., Regnier, T. W., & Di Gregorio, A. (2008). Evolutionary conservation of vertebrate notochord genes in the ascidian Ciona intestinalis. Genesis, 46(11), 697–710. Kugler, J. E., Wu, Y., Katikala, L., Passamaneck, Y. J., Addy, J., Caballero, N., et al. (2019). Positioning a multifunctional basic helix-loop-helix transcription factor within the Ciona notochord gene regulatory network. Developmental Biology, 448(2), 119–135. Kumano, G., Yamaguchi, S., & Nishida, H. (2006). Overlapping expression of FoxA and Zic confers responsiveness to FGF signaling to specify notochord in ascidian embryos. Developmental Biology, 300(2), 770–784. Kusakabe, T. (2005). Decoding cis-regulatory systems in ascidians. Zoological Science, 22(2), 129–146. Kusakabe, T., Yoshida, R., Kawakami, I., Kusakabe, R., Mochizuki, Y., Yamada, L., et al. (2002). Gene expression profiles in tadpole larvae of Ciona intestinalis. Developmental Biology, 242(2), 188–203. Lacalli, T. (2012). The Middle Cambrian fossil Pikaia and the evolution of chordate swimming. EvoDevo, 3(1), 12.

368

Anna Di Gregorio

Lalmansingh, A. S., Karmakar, S., Jin, Y., & Nagaich, A. K. (2012). Multiple modes of chromatin remodeling by Forkhead box proteins. Biochimica et Biophysica Acta, 1819(7), 707–715. Lauri, A., Brunet, T., Handberg-Thorsager, M., Fischer, A. H., Simakov, O., Steinmetz, P. R., et al. (2014). Development of the annelid axochord: Insights into notochord evolution. Science, 345(6202), 1365–1368. Lawson, L., & Harfe, B. D. (2015). Notochord to nucleus pulposus transition. Current Osteoporosis Reports, 13(5), 336–341. Lemaire, P. (2009). Unfolding a chordate developmental program, one cell at a time: Invariant cell lineages, short-range inductions and evolutionary plasticity in ascidians. Developmental Biology, 332(1), 48–60. Levine, M. (2010). Transcriptional enhancers in animal development and evolution. Current Biology, 20(17), R754–R763. Lin, Y., Cai, Z., Huang, S., Yang, L., Wang, C., Liu, Z., et al. (2009). Ptc, Smo, Sufu, and the Hedgehog signaling pathway in amphioxus. Evolution & Development, 11(6), 710–718. Liu, G. H., Mao, C. Z., Wu, H. Y., Zhou, D. C., Xia, J. B., Kim, S. K., et al. (2016). Expression profile of rrbp1 genes during embryonic development and in adult tissues of Xenopus laevis. Gene Expression Patterns, 23–24, 1–6. Lolas, M., Valenzuela, P. D., Tjian, R., & Liu, Z. (2014). Charting Brachyury-mediated developmental pathways during early mouse embryogenesis. Proceedings of the National Academy of Sciences of the United States of America, 111(12), 4478–4483. Lu, Q., Bhattachan, P., & Dong, B. (2019). Ascidian notochord elongation. Developmental Biology, 448(2), 147–153. Madgwick, A., Magri, M. S., Dantec, C., Gailly, D., Fiuza, U. M., Guignard, L., et al. (2019). Evolution of embryonic cis-regulatory landscapes between divergent Phallusia and Ciona ascidians. Developmental Biology, 448(2), 71–87. Maguire, J. E., Pandey, A., Wu, Y., & Di Gregorio, A. (2018). Investigating evolutionarily conserved molecular mechanisms controlling gene expression in the notochord. Advances in Experimental Medicine and Biology, 1029, 81–99. Mallatt, J., & Holland, N. (2013). Pikaia gracilens Walcott: Stem chordate, or already specialized in the Cambrian? Journal of Experimental Zoology. Part B, Molecular and Developmental Evolution, 320, 247–271. Mansfield, J. H., Haller, E., Holland, N. D., & Brent, A. E. (2015). Development of somites and their derivatives in amphioxus, and implications for the evolution of vertebrate somites. EvoDevo, 6, 21. Markovich, D., & Aronson, P. S. (2007). Specificity and regulation of renal sulfate transporters. Annual Review of Physiology, 69, 361–375. Matsumoto, J., Kumano, G., & Nishida, H. (2007). Direct activation by Ets and Zic is required for initial expression of the Brachyury gene in the ascidian notochord. Developmental Biology, 306(2), 870–882. Matsunobu, S., & Sasakura, Y. (2015). Time course for tail regression during metamorphosis of the ascidian Ciona intestinalis. Developmental Biology, 405(1), 71–81. McCann, M. R., & S
eguin, C. A. (2016). Notochord cells in intervertebral disc development and degeneration. Journal of Developmental Biology, 4(1), 3. McCann, M. R., Tamplin, O. J., Rossant, J., & S
eguin, C. A. (2012). Tracing notochordderived cells using a Noto-cre mouse: Implications for intervertebral disc development. Disease Models & Mechanisms, 5, 73–82. Meinertzhagen, I. A., & Okamura, Y. (2001). The larval ascidian nervous system: The chordate brain from its small beginnings. Trends in Neurosciences, 24(7), 401–410. Miwata, K., Chiba, T., Horii, R., Yamada, L., Kubo, A., Miyamura, D., et al. (2006). Systematic analysis of embryonic expression profiles of zinc finger genes in Ciona intestinalis. Developmental Biology, 292(2), 546–554.

The notochord gene regulatory network in chordate evolution

369

Miyamoto, D. M., & Crowther, R. J. (1985). Formation of the notochord in living ascidian embryos. Journal of Embryology and Experimental Morphology, 86, 1–17. Mizotani, Y., Suzuki, M., Hotta, K., Watanabe, H., Shiba, K., Inaba, K., et al. (2018). 14-33εa directs the pulsatile transport of basal factors toward the apical domain for lumen growth in tubulogenesis. Proceedings of the National Academy of Sciences of the United States of America, 115(38), E8873–E8881. Morley, R. H., Lachani, K., Keefe, D., Gilchrist, M. J., Flicek, P., Smith, J. C., et al. (2009). A gene regulatory network directed by zebrafish No tail accounts for its roles in mesoderm formation. Proceedings of the National Academy of Sciences of the United States of America, 106(10), 3829–3834. Morris, S. C., & Caron, J. B. (2012). Pikaia gracilens Walcott, a stem-group chordate from the Middle Cambrian of British Columbia. Biological Reviews of the Cambridge Philosophical Society, 87, 480–512. M€ uller, F., Chang, B., Albert, S., Fischer, N., Tora, L., & Str€ahle, U. (1999). Intronic enhancers control expression of zebrafish sonic hedgehog in floor plate and notochord. Development, 126(10), 2103–2116. Munro, E. M., & Odell, G. M. (2002a). Polarized basolateral cell motility underlies invagination and convergent extension of the ascidian notochord. Development, 129(1), 13–24. Munro, E. M., & Odell, G. (2002b). Morphogenetic pattern formation during ascidian notochord formation is regulative and highly robust. Development, 129(1), 1–12. Newman-Smith, E., Kourakis, M. J., Reeves, W., Veeman, M., & Smith, W. C. (2015). Reciprocal and dynamic polarization of planar cell polarity core components and myosin. eLife, 13(4), e05361. Nibu, Y., Jos
e-Edwards, D. S., & Di Gregorio, A. (2013). From notochord formation to hereditary chordoma: The many roles of Brachyury. BioMed Research International, 2013, 826435. Nieuwkoop, P. D., & Faber, J. (1967). Normal table of Xenopus laevis (Daudin). Amsterdam: North Holland. Nishida, H., & Satoh, N. (1983). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. I. Up to the eight-cell stage. Developmental Biology, 99, 382–394. Nishida, H., & Satoh, N. (1985). Cell lineage analysis in ascidian embryos by intracellular injection of a tracer enzyme. II. The 16- and 32-cell stages. Developmental Biology, 110, 440–454. Nishikata, T., Yamada, L., Mochizuki, Y., Satou, Y., Shin-i, T., Kohara, Y., et al. (2001). Profiles of maternally expressed genes in fertilized eggs of Ciona intestinalis. Developmental Biology, 238(2), 315–331. Nitta, K. R., Vincentelli, R., Jacox, E., Cimino, A., Ohtsuka, Y., Sobral, D., et al. (2019). High-throughput protein production combined with high- throughput SELEX identifies an extensive atlas of Ciona robusta transcription factor DNA-binding specificities. Methods in Molecular Biology, 2025, 487–517. Noda, T. (2011). The maternal genes Ci-p53/p73-a and Ci-p53/p73-b regulate zygotic ZicL expression and notochord differentiation in Ciona intestinalis embryos. Developmental Biology, 360(1), 216–229. Oda-Ishii, I., & Di Gregorio, A. (2007). Lineage-independent mosaic expression and regulation of the Ciona multidom gene in the ancestral notochord. Developmental Dynamics, 236(7), 1806–1819. Oda-Ishii, I., Ishii, Y., & Mikawa, T. (2010). Eph regulates dorsoventral asymmetry of the notochord plate and convergent extension-mediated notochord formation. PLoS One, 5(10), e13689. Ogasawara, M., Sasaki, A., Metoki, H., Shin-i, T., Kohara, Y., Satoh, N., et al. (2002). Gene expression profiles in young adult Ciona intestinalis. Development Genes and Evolution, 212(4), 173–185.

370

Anna Di Gregorio

Ortolani, G. (1954). Risultati definitivi sulla distribuzione dei territory presuntivi degli organi nel germe di Ascidie allo stadio VIII, determinati con le marche al carbone. Pubblicazioni della Stazione Zoologica di Napoli. Section II: History and Philosophy of the Life Sciences, 25, 161–187. Pascual-Anaya, J., Adachi, N., Alvarez, S., Kuratani, S., D’Aniello, S., & Garcia-Ferna`ndez, J. (2012). Broken colinearity of the amphioxus Hox cluster. EvoDevo, 3(1), 28. Passamaneck, Y. J., & Di Gregorio, A. (2005). Ciona intestinalis: Chordate development made simple. Developmental Dynamics, 233(1), 1–19. Passamaneck, Y. J., Katikala, L., Perrone, L., Dunn, M. P., Oda-Ishii, I., & Di Gregorio, A. (2009). Direct activation of a notochord cis-regulatory module by Brachyury and FoxA in the ascidian Ciona intestinalis. Development, 136(21), 3679–3689. Pennati, R., Ficetola, G. F., Brunetti, R., Caicci, F., Gasparini, F., Griggio, F., et al. (2015). Morphological differences between larvae of the Ciona intestinalis species complex: Hints for a valid taxonomic definition of distinct species. PLoS One, 10(5), e0122879. Peter, I. S. (2017). Regulatory states in the developmental control of gene expression. Briefings in Functional Genomics, 16(5), 281–287. Peter, I. S., & Davidson, E. H. (2015). Genomic control process, development and evolution. Amsterdam: Elsevier. Peterson, K. J., Cameron, R. A., Tagawa, K., Satoh, N., & Davidson, E. H. (1999). A comparative molecular approach to mesodermal patterning in basal deuterostomes: The expression pattern of Brachyury in the enteropneust hemichordate Ptychodera flava. Development, 126(1), 85–95. Pfister, K., Shook, D. R., Chang, C., Keller, R., & Skoglund, P. (2016). Molecular model for force production and transmission during vertebrate gastrulation. Development, 143(4), 715–727. Phillips, R., Kondev, J., Theriot, J., & Garcia, H. (2012). Physical biology of the cell (2nd ed.). New York: Garland Science, Taylor and Francis Group LLC. Picco, V., Hudson, C., & Yasuo, H. (2007). Ephrin-Eph signalling drives the asymmetric division of notochord/neural precursors in Ciona embryos. Development, 134(8), 1491–1497. Racioppi, C., Wiechecki, K. A., & Christiaen, L. (2019). Combinatorial chromatin dynamics foster accurate cardiopharyngeal fate choices. Elife, 8, e49921. Rajesh, D., & Dahia, C. L. (2018). Role of sonic hedgehog signaling pathway in intervertebral disc formation and maintenance. Current Molecular Biology Reports, 4(4), 173–179. Rastegar, S., Hess, I., Dickmeis, T., Nicod, J. C., Ertzer, R., Hadzhiev, Y., et al. (2008). The words of the regulatory code are arranged in a variable manner in highly conserved enhancers. Developmental Biology, 318(2), 366–377. Razy-Krajka, F., Lam, K., Wang, W., Stolfi, A., Joly, M., Bonneau, R., et al. (2014). Collier/ OLF/EBF-dependent transcriptional dynamics control pharyngeal muscle specification from primed cardiopharyngeal progenitors. Developmental Cell, 29(3), 263–276. Razy-Krajka, F., & Stolfi, A. (2019). Regulation and evolution of muscle development in tunicates. EvoDevo, 10, 13. Reese, D. E., Hall, C. E., & Mikawa, T. (2004). Negative regulation of midline vascular development by the notochord. Developmental Cell, 6(5), 699–708. Reeves, W., Thayer, R., & Veeman, M. (2014). Anterior-posterior regionalized gene expression in the Ciona notochord. Developmental Dynamics, 243(4), 612–620. Reeves, W. M., Wu, Y., Harder, M. J., & Veeman, M. T. (2017). Functional and evolutionary insights from the Ciona notochord transcriptome. Development, 144(18), 3375–3387. Rhee, J. M., Oda-Ishii, I., Passamaneck, Y. J., Hadjantonakis, A. K., & Di Gregorio, A. (2005). Live imaging and morphometric analysis of embryonic development in the ascidian Ciona intestinalis. Genesis, 43(3), 136–147.

The notochord gene regulatory network in chordate evolution

371

Sasaki, H., & Hogan, B. L. (1993). Differential expression of multiple fork head related genes during gastrulation and axial pattern formation in the mouse embryo. Development, 118(1), 47–59. Sasaki, H., Yoshida, K., Hozumi, A., & Sasakura, Y. (2014). CRISPR/Cas9-mediated gene knockout in the ascidian Ciona intestinalis. Development, Growth & Differentiation, 56(7), 499–510. Satoh, N. (2001). Ascidian embryos as a model system to analyze expression and function of developmental genes. Differentiation, 68(1), 1–12. Satoh, N. (2014). Developmental genomics of ascidians. Hoboken: Wiley-Blackwell. Satoh, N., Tagawa, K., Lowe, C. J., Yu, J. K., Kawashima, T., Takahashi, H., et al. (2014). On a possible evolutionary link of the stomochord of hemichordates to pharyngeal organs of chordates. Genesis, 52(12), 925–934. Satoh, N., Tagawa, K., & Takahashi, H. (2012). How was the notochord born? Evolution & Development, 14(1), 56–75. Satou, Y. (2020). A gene regulatory network for cell fate specification in Ciona embryos. Current Topics in Developmental Biology, 139, 1–33. Satou, Y., Kawashima, T., Shoguchi, E., Nakayama, A., & Satoh, N. (2005). An integrated database of the ascidian, Ciona intestinalis: Towards functional genomics. Zoological Science, 22(8), 837–843. Satou, Y., Nakamura, R., Yu, D., Yoshida, R., Hamada, M., Fujie, M., et al. (2019). A nearly complete genome of Ciona intestinalis type a (C. robusta) reveals the contribution of inversion to chromosomal evolution in the genus Ciona. Genome Biology and Evolution, 11(11), 3144–3157. Satou, Y., & Satoh, N. (2005). Cataloging transcription factor and major signaling molecule genes for functional genomic studies in Ciona intestinalis. Development Genes and Evolution, 215(11), 580–596. Satou, Y., Takatori, N., Fujiwara, S., Nishikata, T., Saiga, H., Kusakabe, T., et al. (2002). Ciona intestinalis cDNA projects: Expressed sequence tag analyses and gene expression profiles during embryogenesis. Gene, 287(1–2), 83–96. Satou, Y., Takatori, N., Yamada, L., Mochizuki, Y., Hamaguchi, M., Ishikawa, H., et al. (2001). Gene expression profiles in Ciona intestinalis tailbud embryos. Development, 128(15), 2893–2904. Satou, Y., Wada, S., Sasakura, Y., & Satoh, N. (2008). Regulatory genes in the ancestral chordate genomes. Development Genes and Evolution, 218(11 12), 715–721. Sawada, A., Nishizaki, Y., Sato, H., Yada, Y., Nakayama, R., et al. (2005). Tead proteins activate the Foxa2 enhancer in the node in cooperation with a second factor. Development, 132, 4719–4729. Segade, F., Cota, C., Famiglietti, A., Cha, A., & Davidson, B. (2016). Fibronectin contributes to notochord intercalation in the invertebrate chordate, Ciona intestinalis. EvoDevo, 7(1) 21. Sehring, I. M., Dong, B., Denker, E., Bhattachan, P., Deng, W., Mathiesen, B. T., et al. (2014). An equatorial contractile mechanism drives cell elongation but not cell division. PLoS Biology, 12(2), e1001781. Sehring, I. M., Recho, P., Denker, E., Kourakis, M., Mathiesen, B., Hannezo, E., et al. (2015). Assembly and positioning of actomyosin rings by contractility and planar cell polarity. eLife, 21(4). e09206. Seo, H. C., Edvardsen, R. B., Maeland, A. D., Bjordal, M., Jensen, M. F., Hansen, A., et al. (2004). Hox cluster disintegration with persistent anteroposterior order of expression in Oikopleura dioica. Nature, 431(7004), 67–71. Sheyn, D., Ben-David, S., Tawackoli, W., Zhou, Z., Salehi, K., Bez, M., et al. (2019). Human iPSCs can be differentiated into notochordal cells that reduce intervertebral disc degeneration in a porcine model. Theranostics, 9(25), 7506–7524.

372

Anna Di Gregorio

Shi, W., & Levine, M. (2008). Ephrin signaling establishes asymmetric cell fates in an endomesoderm lineage of the Ciona embryo. Development, 135(5), 931–940. Shi, W., Peyrot, S. M., Munro, E., & Levine, M. (2009). FGF3 in the floor plate directs notochord convergent extension in the Ciona tadpole. Development, 136(1), 23–28. Shimeld, S. M. (1999). The evolution of the hedgehog gene family in chordates: Insights from amphioxus hedgehog. Development Genes and Evolution, 209(1), 40–47. Sivan, S. S., Hayes, A. J., Wachtel, E., Caterson, B., Merkher, Y., Maroudas, A., et al. (2014). Biochemical composition and turnover of the extracellular matrix of the normal and degenerate intervertebral disc. European Spine Journal, 23(Suppl. 3), S344–S353. Smith, W. C. (2018). Cellular processes of notochord formation. Advances in Experimental Medicine and Biology, 1029, 165–177. Søviknes, A. M., & Glover, J. C. (2008). Continued growth and cell proliferation into adulthood in the notochord of the appendicularian Oikopleura dioica. The Biological Bulletin, 214(1), 17–28. Spagnuolo, A., Ristoratore, F., Di Gregorio, A., Aniello, F., Branno, M., & Di Lauro, R. (2003). Unusual number and genomic organization of Hox genes in the tunicate Ciona intestinalis. Gene, 309(2), 71–79. Stemple, D. L. (2005). Structure and function of the notochord: An essential organ for chordate development. Development, 132, 2503–2512. Stolfi, A., & Christiaen, L. (2012). Genetic and genomic toolbox of the chordate Ciona intestinalis. Genetics, 192(1), 55–66. Stolfi, A., Gandhi, S., Salek, F., & Christiaen, L. (2014). Tissue-specific genome editing in Ciona embryos by CRISPR/Cas9. Development, 141(21), 4115–4120. Stott, D., Kispert, A., & Herrmann, B. G. (1993). Rescue of the tail defect of Brachyury mice. Genes & Development, 7(2), 197–203. Sulik, K., Dehart, D. B., Iangaki, T., Carson, J. L., Vrablic, T., Gesteland, K., et al. (1994). Morphogenesis of the murine node and notochordal plate. Developmental Dynamics, 201(3), 260–278. Suzuki, M. M., & Satoh, N. (2000). Genes expressed in the amphioxus notochord revealed by EST analysis. Developmental Biology, 224(2), 168–177. Tada, M., Casey, E. S., Fairclough, L., & Smith, J. C. (1998). Bix1, a direct target of Xenopus T-box genes, causes formation of ventral mesoderm and endoderm. Development, 125(20), 3997–4006. Tajima, Y., Hozumi, A., Yoshida, K., Treen, N., Sakuma, T., Yamamoto, T., et al. (2020). Hox13 is essential for formation of a sensory organ at the terminal end of the sperm duct in Ciona. Developmental Biology, 458(1), 120–131. Takahashi, H., Hotta, K., Erives, A., Di Gregorio, A., Zeller, R. W., Levine, M., et al. (1999). Brachyury downstream notochord differentiation in the ascidian embryo. Genes & Development, 13(12), 1519–1523. Takahashi, H., Hotta, K., Takagi, C., Ueno, N., Satoh, N., & Shoguchi, E. (2010). Regulation of notochord-specific expression of Bra downstream genes in Ciona intestinalis embryos. Zoological Science, 27(2), 110–118. Takatori, N., Hotta, K., Mochizuki, Y., Satoh, G., Mitani, Y., Satoh, N., et al. (2004). T-box genes in the ascidian Ciona intestinalis: Characterization of cDNAs and spatial expression. Developmental Dynamics, 230(4), 743–753. Takatori, N., Satou, Y., & Satoh, N. (2002). Expression of hedgehog genes in Ciona intestinalis embryos. Mechanisms of Development, 116(1–2), 235–238. Talbot, W. S., Trevarrow, B., Halpern, M. E., Melby, A. E., Farr, G., Postlethwait, J. H., et al. (1995). A homeobox gene essential for zebrafish notochord development. Nature, 378(6553), 150–157.

The notochord gene regulatory network in chordate evolution

373

Tamplin, O. J., Cox, B. J., & Rossant, J. (2011). Integrated microarray and ChIP analysis identifies multiple Foxa2 dependent target genes in the notochord. Developmental Biology, 360(2), 415–425. Tamplin, O. J., Kinzel, D., Cox, B. J., Bell, C. E., Rossant, J., & Lickert, H. (2008). Microarray analysis of Foxa2 mutant mouse embryos reveals novel gene expression and inductive roles for the gastrula organizer and its derivatives. BMC Genomics, 9, 511. Tang, R., Jing, L., Willard, V. P., Wu, C. L., Guilak, F., Chen, J., et al. (2018). Differentiation of human induced pluripotent stem cells into nucleus pulposus-like cells. Stem Cell Research & Therapy, 9(1) 61. Thompson, J. M., & Di Gregorio, A. (2015). Insulin-like genes in ascidians: Findings in Ciona and hypotheses on the evolutionary origins of the pancreas. Genesis, 53(1), 82–104. Tokuhiro, S. I., Tokuoka, M., Kobayashi, K., Kubo, A., Oda-Ishii, I., & Satou, Y. (2017). Differential gene expression along the animal-vegetal axis in the ascidian embryo is maintained by a dual functional protein Foxd. PLoS Genetics, 13(5), e1006741. Urano, A., Suzuki, M. M., Zhang, P., Satoh, N., & Satoh, G. (2003). Expression of musclerelated genes and two MyoD genes during amphioxus notochord development. Evolution & Development, 5(5), 447–458. Urry, L. A., Whittaker, C. A., Duquette, M., Lawler, J., & DeSimone, D. W. (1998). Thrombospondins in early Xenopus embryos: Dynamic patterns of expression suggest diverse roles in nervous system, notochord, and muscle development. Developmental Dynamics, 211, 390–407. Utsumi, N., Shimojima, Y., & Saiga, H. (2004). Analysis of ascidian Not genes highlights their evolutionarily conserved and derived features of structure and expression in development. Development Genes and Evolution, 214(9), 460–465. Valenta, T., Hausmann, G., & Basler, K. (2012). The many faces and functions of β-catenin. The EMBO Journal, 31(12), 2714–2736. Veeman, M. T., Nakatani, Y., Hendrickson, C., Ericson, V., Lin, C., & Smith, W. C. (2008). Chongmague reveals an essential role for laminin-mediated boundary formation in chordate convergence and extension movements. Development, 135(1), 33–41. Veeman, M. T., & Smith, W. C. (2013). Whole-organ cell shape analysis reveals the developmental basis of ascidian notochord taper. Developmental Biology, 373(2), 281–289. Velandia-Huerto, C. A., Brown, F. D., Gittenberger, A., Stadler, P. F., & Bermu´dezSantana, C. I. (2018). Nonprotein-coding RNAs as regulators of development in tunicates. Results and Problems in Cell Differentiation, 65, 197–225. Wada, H., Okuyama, M., Satoh, N., & Zhang, S. (2006). Molecular evolution of fibrillar collagen in chordates, with implications for the evolution of vertebrate skeletons and chordate phylogeny. Evolution & Development, 8, 370–377. Wang, W., & Christiaen, L. (2012). Transcriptional enhancers in ascidian development. Current Topics in Developmental Biology, 98, 147–172. Wang, S., Furmanek, T., Kryvi, H., Krossøy, C., Totland, G. K., Grotmol, S., et al. (2014). Transcriptome sequencing of Atlantic salmon (Salmo salar L.) notochord prior to development of the vertebrae provides clues to regulation of positional fate, chordoblast lineage and mineralisation. BMC Genomics, 15, 141. Ward, L., Pang, A. S. W., Evans, S. E., & Stern, C. D. (2018). The role of the notochord in amniote vertebral column segmentation. Developmental Biology, 439(1), 3–18. Wilkinson, D. G., Bhatt, S., & Herrmann, B. G. (1990). Expression pattern of the mouse T gene and its role in mesoderm formation. Nature, 343(6259), 657–659. Winkley, K., Ward, S., Reeves, W., & Veeman, M. (2019). Iterative and complex asymmetric divisions control cell volume differences in Ciona notochord tapering. Current Biology, 29(20), 3466–3477.

374

Anna Di Gregorio

Yagi, K., Satou, Y., & Satoh, N. (2004). A zinc finger transcription factor, ZicL, is a direct activator of Brachyury in the notochord specification of Ciona intestinalis. Development, 131(6), 1279–1288. Yamada, S., Ueno, N., Satoh, N., & Takahashi, H. (2011). Ciona intestinalis n4 contains a phosphotyrosine interaction domain and is involved in the midline intercalation of notochord cells. The International Journal of Developmental Biology, 55(1), 11–18. Yasuo, H., & Hudson, C. (2007). FGF8/17/18 functions together with FGF9/16/20 during formation of the notochord in Ciona embryos. Developmental Biology, 302(1), 92–103. Zizic Mitrecic, M., Mitrecic, D., Pochet, R., Kostovic-Knezevic, L., & Gajovic, S. (2010). The mouse gene Noto is expressed in the tail bud and essential for its morphogenesis. Cells, Tissues, Organs, 192(2), 85–92.

CHAPTER TWELVE

On the specificity of gene regulatory networks: How does network co-option affect subsequent evolution? Eden McQueen, Mark Rebeiz* Department of Biological Sciences, University of Pittsburgh, Pittsburgh, PA, United States *Corresponding author: e-mail address: [email protected]

Contents 1. How does one part become different from other parts? 2. Immediate outcomes of network co-option 2.1 Wholesale co-option 2.2 Partial network co-option and functionally-divergent network co-option 2.3 Co-option resulting in downstream regulatory expression only (aphenotypic co-option) 3. How do GRNs maintain or recover specificity after network co-option? 3.1 Changes in trans 3.2 Changes in cis 4. The action of selection on co-opted networks over time 4.1 Initiating trans change with positive fitness effects 4.2 Initiating trans change with detrimental fitness effects 4.3 Initiating trans change with selectively neutral effects 5. Network co-option and the origin and diversification of traits 6. Concluding remarks Acknowledgments Glossary References

376 379 379 382 384 385 386 387 390 390 393 394 395 397 398 398 398

Abstract The process of multicellular organismal development hinges upon the specificity of developmental programs: for different parts of the organism to form unique features, processes must exist to specify each part. This specificity is thought to be hardwired into gene regulatory networks, which activate cohorts of genes in particular tissues at particular times during development. However, the evolution of gene regulatory networks sometimes occurs by mechanisms that sacrifice specificity. One such mechanism is network co-option, in which existing gene networks are redeployed in new developmental

Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.03.002

#

2020 Elsevier Inc. All rights reserved.

375

376

Eden McQueen and Mark Rebeiz

contexts. While network co-option may offer an efficient mechanism for generating novel phenotypes, losses of tissue specificity at redeployed network genes could restrict the ability of the affected traits to evolve independently. At present, there has not been a detailed discussion regarding how tissue specificity of network genes might be altered due to gene network co-option at its initiation, as well as how trait independence can be retained or restored after network co-option. A lack of clarity about network co-option makes it more difficult to speculate on the long-term evolutionary implications of this mechanism. In this review, we will discuss the possible initial outcomes of network co-option, outline the mechanisms by which networks may retain or subsequently regain specificity after network co-option, and comment on some of the possible evolutionary consequences of network co-option. We place special emphasis on the need to consider selectively-neutral outcomes of network co-option to improve our understanding of the role of this mechanism in trait evolution.

1. How does one part become different from other parts? The biology of gene regulatory networks (GRNs) has played a key role in our understanding of how parts are differentiated during development (Davidson, 2010). Each gene (or network “node”) within a GRN is deployed through the action of transcription factors which bind specifically to cis-regulatory elements (CREs) to activate its tissue-specific expression (Farley, Olson, & Levine, 2015; Levine, 2010). Phenotypic changes can often be traced to changes in GRN structure that have tissue-specific effects (Carroll, 2008; Stern & Orgogozo, 2008; Wray, 2007), and thus understanding the mechanisms by which GRNs can be modified gives us insight into evolution. One mechanism that has emerged as a potential player in the evolution of GRNs is the phenomenon of “gene network co-option” (For definition, see Box 1), particularly in the origins of novel phenotypes (Monteiro, 2012; Olson, 2006; Peter & Davidson, 2015; Shubin, Tabin, & Carroll, 2009; True & Carroll, 2002). Changes to a single regulator (an “initiating trans change”) in an existing GRN could recruit many terminal effectors in just one or a few steps to produce a novel phenotype, rather than a slow accumulation of the necessary mutations in the CRE of each effector (Fig. B1). While co-option is a mechanism for rapidly establishing a complex network in a tissue, because the CREs of a co-opted GRN have their function expanded, network co-option is predicted to cause an immediate loss of the

Network co-option and subsequent evolution

377

tissue specificity for the reused CREs (Duboule & Wilkins, 1998; Rebeiz, Patel, & Hinman, 2015; Rice & Rebeiz, 2019). If a large number of co-opted CREs are causally linked to an increased number of phenotypes (i.e., increased pleiotropy for every co-opted CRE), this lack of specificity could be detrimental over evolutionary time, as it may preclude the independent movement of affected traits toward fitness maxima, at least via

BOX 1 The term “co-option” has been applied to a range of phenomena. For instance, co-option is sometimes used to describe a case in which a gene network with a known ancestral function may evolve along a certain lineage to be employed instead for a novel derived function (e.g., (Hinman, Nguyen, & Davidson, 2007; Suryamohan et al., 2016). Our usage here is somewhat more restricted. For our purposes, “gene network co-option” (or “network co-option”) is a specific way of modifying the developmental “program”. In network co-option, a regulatory factor is deployed in a new location or at a new time during development such that this factor interacts with already existing cis-regulatory elements (CREs) in the next developmental time step. These extant CREs were previously functional in the process of specifying some other trait, i.e., regulated nodes in an existing part (gene regulatory network or “GRN”) of the developmental program. Thus, the activation of these CREs may initiate a second instantiation of some or all of the subsequent time steps of that preexisting program (Fig. B1). The regulatory machinery that defines the GRN of this other trait is therefore being reused, recruited, or “co-opted” to a new location or at a novel point in time (Shubin et al., 2009; True & Carroll, 2002). Our usage therefore defines co-option as a mechanism, not as an outcome per se. This distinction is important, as the deployment of an existing GRN in a novel location could occur by other mechanisms, such as de-novo construction of network connections or some combination of de-novo building and co-option. “Co-option” of a terminal effector gene via changes to that gene’s locus is not conceptually distinct from what we describe here (e.g., Gompel, Prud’homme, Wittkopp, Kassner, & Carroll, 2005), but our focus is specifically co-option of multiple interconnected elements in networks simultaneously. Other closely related and interesting phenomena that we do not discuss here are co-option of host gene expression by pathogens (e.g., Faust, Binning, Gross, & Frankel, 2017; Saeij et al., 2007) and the alternate developmental trajectories induced in cancer cells via co-option of extant network architecture (e.g., Minafra et al., 2014; Shah et al., 2013). It would be interesting to connect these areas of research to the concepts discussed here in the future. Continued

378

Eden McQueen and Mark Rebeiz

BOX 1—CONT’D

Fig. B1 Redeployment of a gene regulatory network via gene network co-option. (Left) The ancestral condition reflects that gene X directly regulates downstream targets only in cellular context 1 and not in context 2 because it is not expressed there. (Right) Co-option occurs when there is an expansion of the expression domain of gene X, such that it is now also expressed in cellular context 2 in the derived condition. The novel expression of gene X in cellular context 2 results in the redeployment of the downstream targets of gene X in cellular context 2, employing its existing cis-regulatory elements (CREs).

changes to those CREs (Fisher, 1930; Hansen, 2003). That is, an excess of pleiotropic linkage between traits as a result of co-option could ultimately act as a relative constraint, impinging on the evolvability of traits. Since we generally do not observe such strong pleiotropic constraints (Wagner & Zhang, 2011), we must either assume that network co-option is quite rare, or explain why repeated occurrences of network co-option do not hamper evolvability. The growing number of studies that implicate co-option suggest that this mechanism is common enough to warrant a discussion of the latter. We presume that networks as a whole regain or maintain specificity after network co-option, as we observe some degree of modularity for GRNs that have thus far been characterized (Davidson & Erwin, 2006; Sabarı´s, Laiker, Noon, & Frankel, 2019; Wagner et al., 2008). What we would like to provide here is a thorough analysis of how the specificity of network

Network co-option and subsequent evolution

379

nodes fluctuates over time after network co-option. We hope this will be informative for our understanding of co-option as a mechanism, and in particular, our understanding of how this mechanism might relate to evolvability. We will break down the phenomenon by first outlining the range of possible immediate outcomes for a given instance of network co-option. We will then describe the mechanisms by which network nodes may either retain or subsequently regain specificity of their cisregulatory information. Finally, we will draw on our outline of this mechanism to discuss the potential role(s) of network co-option in the evolution of organismal parts.

2. Immediate outcomes of network co-option Network co-option events can theoretically yield a wide range of outcomes. In most cases, there will exist many differences in the transregulatory landscape of the cells of the novel (differing in space or time) context and that of the context in which the network operated previously. Distinct regulatory information in the novel cellular context can intersect or interfere with the newly redeployed network at any point downstream of the initiating trans change, and thus individual cases of network co-option may differ in the number of network genes redeployed, as well as the identities of downstream targets. We can visualize the spectrum of possible outcomes at the initiation of network co-option by outlining four broad categories (Fig. 1): Wholesale co-option, partial co-option, functionally divergent co-option, and aphenotypic co-option.

2.1 Wholesale co-option One possibility is that the entire, or nearly the entire network downstream of the initiating trans change is redeployed in the novel tissue. We call such cases “wholesale co-option.” The result of wholesale co-option is that the same set of terminal effectors is activated in the novel location, and there will be a recapitulation or near-recapitulation in the novel location of the trait generated by the network downstream of the initiating trans change in the ancestral location (Fig. 1C). Gain-of-function homeotic transformations are an illustrative example of wholesale network reuse, where the initiating trans change may be the introduction of a Hox gene. For example, in Drosophila melanogaster the antennae can be transformed into legs by the overexpression of the homeobox gene Antennapedia (Schneuwly, Klemenz, & Gehring, 1987). Here, the

380

Eden McQueen and Mark Rebeiz

Fig. 1 Network Co-option results in a range of possible outcomes. (A) Ancestral location and outcome of network deployment, showing the phenotype in that context. (B) Prior to network co-option, the network is inactive in the second location. (C–F) Activation of the upstream “initiating trans factor” results in redeployment of some or all of this network in the second location. (C) Wholesale co-option involves the redeployment of the

Network co-option and subsequent evolution

381

addition of a single upstream factor results in the deployment of an entire leg formation network in a different location, with the terminal result being easily recognizable as the trait for which the network is generally employed in wild-type animals. Likewise, misexpression of the eyeless (ey) gene in Drosophila melanogaster is capable of generating ectopic eyes (Halder, Callaerts, & Gehring, 1995). Similar kinds of homeotic transformations involving changes to single factors have also been observed in ´ lvarez-Buylla et al., 2010; Coen & Meyerowitz, 1991). floral parts (A Wholesale co-option might also be common when repeated structures, such as neurons, muscles, epithelial appendages, and even seriallyhomologous body segments increase in number, as these networks have already been subject to recurrent reuse and therefore may possess nodes capable of “selector-like” (Garcı´a-Bellido, 1975; Mann & Carroll, 2002) or “input-output” (Stern & Orgogozo, 2009) function (i.e., largely sufficient to produce the phenotype). For example, Marcellini and Simpson (2006) showed that the expanded domain of a single enhancer of the gene scute was sufficient to explain the derived condition in which the number or dorsocentral bristles increased from two to four in the Drosophilid species Drosophila quadrilineata (Marcellini & Simpson, 2006). The D. quadrilineata enhancer was able to recapitulate the derived condition when used to drive scute expression in D. melanogaster (ordinarily possessing only two dorsocentral bristles), demonstrating that the existing downstream regulatory logic was used in the construction of the novel pair of bristles, consistent with wholesale co-option. We define wholesale co-option as an instance of network co-option in which the direct downstream consequences of the redeployment of the initiating trans factor are identical (or nearly identical) in the novel cellular context to those in the ancestral cellular context. This definition does not entire network in the novel context, resulting in the recapitulation of the phenotype that appears in the ancestral context. (D) Partial co-option, in which some of the downstream transcription factors and terminal effectors are not redeployed in the novel context. The phenotype in the novel context may share some features with the phenotype in the ancestral context. (E) Functionally-divergent co-option is similar to D, except that in the novel context, some of the downstream targets of the redeployed network are distinct from the ancestral context. The phenotype is not necessarily recognizable as being associated with the phenotype in the ancestral context. (F) Aphenotypic co-option, in which no terminal effectors are activated, although there are changes to the upstream developmental program. No phenotype is observed, apart from changes of expression that can be detected experimentally.

382

Eden McQueen and Mark Rebeiz

specify what qualifies as a recapitulated “trait.” A trait does not, for example, need be a discrete organ like a bristle, eye, or wing. It could instead be a characteristic such as pigmentation or a chemical signal, provided that the initiating trans change is sufficient to recapitulate the downstream effect. For instance, a possible example of wholesale co-option is in the exoskeletalization of the elytra (exoskeletalized forewings) of beetles. Experiments on Tribolium castaneum suggest that the derived state of elytral exoskeletalization may have evolved via redeployment of the entire exoskeletalization network of the body wall to the beetle forewing, involving novel upstream regulatory roles of existing wing-patterning elements (Tomoyasu, Arakane, Kramer, & Denell, 2009). The recapitulated “trait” in this case is the exoskeletal fate of cells in the forewing tissue.

2.2 Partial network co-option and functionally-divergent network co-option Wholesale co-option represents the most comprehensive case of network reuse. The striking differences between structures that we believe were built via co-option and the ancestral traits from which the networks were co-opted suggests that simple transplantation of entire networks will be rare. Rather, we anticipate that in the majority of co-option cases, only a portion of the network downstream of the initiating trans change will be redeployed (Erwin, 2020). We refer to cases wherein some subset, but not the entirety of a downstream network is redeployed as “partial co-option” (Fig. 1D). Network architecture can be highly context-dependent (Luscombe et al., 2004), and the fidelity of the network redeployment can range from substantial, in which case many features of the ancestral trait are identifiable, to quite minimal, such that the imported elements of the ancestral trait are unrecognizable, or nearly so, in the novel context. Many factors may come into play to prevent activation of some downstream nodes, including (but not be limited to): other tissue specific transcription factors, tissue-specific posttranscriptional modification (e.g., splicing, phosphorylation, protein cleavage), extrinsic signaling from adjacent tissues (Barolo & Posakony, 2002), and boundary conditions set up by developmental timing or mechanical constraints (Davidson, 2012; Green & Batterman, 2017; Womack, Metz, & Hoke, 2019). A similar phenomenon occurs in a case of what we will call “functionally divergent” co-option (Fig. 1E). In these cases, upstream network architecture is co-opted, but the terminal effectors activated by that upstream network differ in the novel context. The upstream nodes would be active in

Network co-option and subsequent evolution

383

both ancestral and novel settings, but the CREs of the distinct downstream nodes would not be. We sketch out the distinctions between wholesale, partial, and functionally divergent co-option largely for theoretical purposes here, to clearly outline the full range of possible implications for this mechanism. Empirically, these outcomes would only be definitively distinguishable from each other at the time the co-option was first initiated, and would require detailed knowledge of the network CREs, such that the activity of those CREs and the downstream targets of network genes could be compared across contexts. Any subsequent changes to these networks would obscure the distinction between categories of co-option. Many well-known examples of network co-option referenced in the literature likely represent instances of partial co-option, functionally-divergent co-option, or some combination of both. For example, in contrast to the ectopic eyes generated by Pax6 mis-expression, a fascinating example of a possible partial network co-option of an eye network was found in a study of extinct dipterans. Two species in the genus Eohelea possessed a structure on their wings that bore a remarkable resemblance to the compound eyes of individuals of those same species, leading the authors to conclude that this structure was likely built through network reuse. However, it appears that in the case of this novel wing structure, the novelty consisted only of the cuticular part of the eye, and was not an entire ectopic eye (Dinwiddie & Rachootin, 2011). We note that in this case because these are not extant species, it is not possible to distinguish between a partial network co-option and a wholesale co-option of a single independent part of the eye network. Functionally divergent co-option may often result from the co-option of signaling pathways, which are utilized throughout development and quite commonly implicated in the formation of novel traits (Cebra-Thomas et al., 2005; Harris, Fallon, & Prum, 2002; Harris, Williamson, Fallon, Meinhardt, & Prum, 2005; Loredo et al., 2001; Nakamura et al., 2015; Wasik & Moczek, 2011). For example, studies on butterfly eyespots suggest that the evolution of these novelties likely included functionally divergent co-option of a deeply conserved anterior-posterior boundary-forming network. The downstream consequences of this network in the novel context provides pattern information for the color phenotypes manifested in scales (Carroll et al., 1994; Keys et al., 1999). Thus, an important process was co-opted, that is, the formation of a particular transcription factor landscape pattern, yet the downstream targets differed in the novel location. We currently do not understand how these functionally divergent

384

Eden McQueen and Mark Rebeiz

outcomes became connected to the anterior-posterior boundary network, € and many hypotheses exist (Ozsu & Monteiro, 2017). A similar case has been suggested in plants, for which a abaxial-adaxial polarity gene network responsible for the flattening of organs such as leaves appears to have been co-opted to cause flattening of stamen filaments, a derived condition (de Almeida et al., 2014). A well-known example that appears to support partial co-option is the redeployment of the leg network in beetle head horns (Moczek & Nagy, 2005). Not every gene of the canonical leg network is actually required for horn development, as evidenced by the lack of a phenotype when the gene dachshund was knocked down (Moczek & Rose, 2009). There is also evidence for partial co-option in the case of the posterior lobe (a genital structure) of fruit flies in the Drosophila melanogaster clade. The evolution of this novel structure appears to have involved co-option of part, but not all, of a network responsible for the development of an ancestral larval structure (Glassford et al., 2015). Additional examples include tree-hopper helmets, the evolution of which appears to have involved co-opted elements of the wing-patterning network (Fisher, Wegrzyn, & Jockusch, 2019; Prud’homme et al., 2011), bilaterian appendages, which may have involved redeployment of an existing network for anteroposterior patterning (Lemons, Fritzenwanker, Gerhart, Lowe, & McGinnis, 2010), and the use of Hox genes in the evolution of paired vertebrate limbs (Zakany & Duboule, 2007). We imagine that these cases represent an amalgam of partial and functionally-divergent co-option, but we are still uncovering the full picture of how the ancestral networks were reused and rewired. Future work characterizing these networks more extensively will help us understand how and when nodes were lost and gained across contexts.

2.3 Co-option resulting in downstream regulatory expression only (aphenotypic co-option) Finally, network co-option may involve the introduction of an upstream regulator that results in the deployment of some of the upstream network nodes in the novel context, but causes no phenotype (defined as measurable change in morphology/behavior); only transcription factor expression patterns are altered. Such cases of “aphenotypic” co-option could ensue if there is a total lack of activation, or inadequate activation, of required terminal effector genes (Fig. 1F). Aphenotypic co-option is essentially an extreme case of partial network co-option. As the term “pleiotropy” is usually restricted to mean that changes to one locus can induce multiple

Network co-option and subsequent evolution

385

phenotypes (Wagner et al., 2008), aphenotypic co-option generates no new pleiotropy, although changes to the developmental program have occurred. Examples of aphenotypic co-option are lacking, but there are reasons to believe they exist, or at least we have evidence of the possibility in that we often observe expression patterns for which we can offer no functional explanation. For instance, RNAi screens sometimes find that knockdowns of some transcription factors expressed in the tissue of interest have no phenotype (e.g., Staller et al., 2013; Zattara, Busey, Linz, Tomoyasu, & Moczek, 2016). Similarly, a comparison of the expression of 20 genes in imaginal tissues of four very closely related species in the melanogaster clade uncovered striking differences in expression across species, many of which are not connected to any known phenotype (Rebeiz, Jikomes, Kassner, & Carroll, 2011). Many CREs for genes exhibit expression patterns outside the focal tissue of a given study, these patterns having no known functional role (e.g., late anterior expression driven by the minimal even-skipped stripe 2 enhancer in Drosophila embryos ( Janssens et al., 2006)). Besides the aforementioned examples, many more cases like these may suffer from the “file drawer problem” (Rosenthal, 1979). Such results are usually ignored, or interpreted as evidence of robustness, but may sometimes represent cases of aphenotypic network redeployment or non-functional nodes of partially co-opted networks. Aphenotypic co-option as an idea has generally received sparse attention, although it has been mentioned as a possibility in the past (True & Carroll, 2002). While considering such an outcome on its own may seem irrelevant, when considered in the light of long evolutionary periods, this phenomenon could nonetheless have some quite interesting implications, as we will discuss in Sections 3 and 4.

3. How do GRNs maintain or recover specificity after network co-option? Restoration of at least partial regulatory specificity of co-opted CREs is almost certain to be a pervasive phenomenon, considering the number of morphological novelties we have discussed here that arose through likely network co-options but now apparently evolve in a largely independent manner (e.g., treehopper helmets (Prud’homme et al., 2011), beetle horns (Emlen, Lavine, & Ewen-Campen, 2007; Moczek, 2009), butterfly wing spots (Brunetti et al., 2001; Oliver, Tong, Gall, Piel, & Monteiro, 2012), and feathers (Prum, 1999; Prum, 2005; Prum & Brush, 2002)).

386

Eden McQueen and Mark Rebeiz

The process of reestablishing CRE tissue specificity could happen in two ways: In cis, via changes to the co-opted CREs themselves, or in trans, via changes outside the network that introduce tissue-specific regulators of the co-opted CREs. We discuss these possibilities below. We note that although our discussion is centered on co-opted CREs, multiple studies have noted pleiotropy in enhancer sequences in the absence of network co-option events (Nagy et al., 2018; Rebeiz et al., 2011; Preger-Ben Noon et al., 2018), and the mechanisms we describe below apply broadly to the evolution of regulatory specificity.

3.1 Changes in trans Many genes outside of the co-opted GRN will likely have pre-existing roles in the tissue that predated the network co-option event. These genes may be available for genetic tinkering that yields tissue-specific modifications, or could contribute to an immediate plastic response that could modulate pleiotropy, and be genetically modified later (West-Eberhard, 2005). Novel expression domains of such genes beyond the co-opted GRN could also arise subsequently, after co-option has occurred, and be exploited to achieve tissue independence at that future time. A modification of this type can be inferred from the striking instance of wholesale co-option demonstrated for the embryonic skeleton of sea urchins (a derived trait) which employs a GRN co-opted from adult skeletogenesis (Gao & Davidson, 2008). In this case, while most of the genes in the co-opted network are shared across the two contexts, implicating wholesale co-option, the embryonic skeletogenesis network incorporates a small number of genes that are not part of the adult skeletogenesis network. One of these genes, tbrain (tbr), has known ancestral roles in endomesoderm specification in other echinoderms, leading the authors to suggest that the addition of this regulator was a modification to the embryonic skeletogenesis network that occurred after the initial co-option event. Direct regulators of the initiating trans factor could be targets for modification, if these differ between the novel and ancestral contexts. For instance, in the example discussed above concerning the exoskeletalization of elytra in beetles, the authors showed that the gene apterous (ap), which is part of the ancestral wing network, has gained a novel role in redeploying the exoskeleton network in the elytra, whereas ap is not a direct regulator of the exoskeleton network in the mesonotum (a cuticular part of a thoracic segment). Consequently, when RNAi was performed targeting ap,

Network co-option and subsequent evolution

387

defects were seen in exoskeletalization of the forewing, but not the mesonotum (Tomoyasu et al., 2009). This demonstrates that, in principle, exoskeletalization could be targeted independently in the elytra by modifying regulators further upstream of the co-opted network, even if the exoskeletalization network itself were pleiotropic.

3.2 Changes in cis If tissue specificity is to be regained via changes to CREs of nodes in the co-opted GRN, the first requirement is that the ancestral and novel tissues must have at least one qualitative or quantitative difference in cellular regulatory content (e.g., transcription factor identity, activity, or concentration) at the time the node in question is active. That is, there must exist some form of potentially exploitable tissue specificity, otherwise all changes to the CRE would necessarily affect both the ancestral and novel contexts. We suspect that this is usually the case, although in principle it is possible that the regulatory states of the ancestral and novel contexts would not differ at all after co-option (i.e., if the only difference between the two contexts prior to co-option was the presence/absence of the initiating trans factor itself ). In such cases, multiple mutations would be required to regain tissue specificity. There are two primary mechanisms that can mitigate or eliminate the potential for pleiotropic constraint at a co-opted CRE: regulatory input diversification (Fig. 2B) and context-specific redundant enhancers (Fig. 2D). The result of these mechanisms would be either a facilitation of independent regulation of the node in the novel and ancestral contexts, or a modulation of the CRE in question in one context (for instance inactivating it in one context). A critical note is that while these features that restore CRE tissue specificity may evolve subsequent to network co-option, they also may already be in place at the time of co-option. Indeed, the modifications to CREs we describe here would be causal explanations of why some nodes are not expressed in partial or aphenotypic co-option outcomes. Similarly, we note that although entirely novel mutations could be the causal changes in these mechanisms, novel genetic combinations of existing variants already segregating in the population could also alter pleiotropy or mask it through epistasis (Pavlicev et al., 2008). Given the pervasive nature of epistasis in natural populations (Mackay, 2014; Phillips, 2008) this may be a frequently employed path to recapturing lost specificity. It is also important to mention that tracing the process of how tissue specificity was restored may be very difficult when comparing species

Fig. 2 Mechanisms to retain or regain specificity of pleiotropic CREs. (A) A gene that is redeployed during network co-option possesses a pleiotropic cis-regulatory element that drives expression in contexts 1 and 2. The CRE is activated by the binding of transcription factors ii, iii, and iv, and the output of expression in both contexts is not independent. (B) Regulatory input diversification: evolution of a binding site for transcription factor v, which acts as a repressor, only affects expression in context 2, as v is not present in context 1. Further modification can occur to achieve greater or full independence via enhancer splitting (C), in which a single enhancer fragments into two enhancers employing context-specific activators. (D) Redundant enhancers: A second enhancer for the target gene affects expression in context 1 only, due to the fact that this redundant enhancer requires the binding of transcription factor i, which is not present in context 2. Further modification can occur via enhancer subfunctionalization (E), in which redundant enhancers that have partial or full overlap in their expression profiles gain or lose binding sites for context-specific factors to achieve complete independence.

Network co-option and subsequent evolution

389

that have diverged for long periods of time. Once sufficient time has passed, the footprints of this process will have been erased. As such, evidence for this mechanism will likely be found in cases where at least some of the ancestral pleiotropic and redundant enhancers are still detectable. 3.2.1 Regulatory input diversification The diversification of regulatory inputs (Fig. 2B) mitigates pleiotropy via binding sites at the pleiotropic enhancer that affect the regulatory outcome of the enhancer differentially across tissues. For example, this might be the gain of a binding site for a repressor that is only present in one tissue. Enhancer splitting (Fig. 2C) is an extension of the process above, and is related to the idea of enhancer sprawl (Rice & Rebeiz, 2019), in which it is understood that enhancers sometimes expand and contract due to the addition and removal of binding sites via turnover. In this case, tissuespecific binding sites accumulate such that a single enhancer that has some tissue specific and some pleiotropic binding sites may eventually split into two completely separate tissue-specific cis-regulatory elements in adjacent positions on the DNA. 3.2.2 Specificity conferred by redundant CREs Redundant CREs (also called “shadow enhancers”) are defined here as at least two CREs driving expression of the same target gene in redundant or semi-redundant expression patterns (Barolo, 2012; Hong, Hendrix, & Levine, 2008). Redundant CREs are a route to at least partial recovery of modularity in cases of co-option because if the CREs employ a unique set of regulators, one or both of the redundant CREs may drive expression differently across the two tissues. We already have several empirical examples of redundant enhancer pairs that display different regulatory logic (Vincent et al., 2018; Wunderlich et al., 2015), lending credence to this potential route to specificity. With redundant CREs, there are two possible conditions. First, a redundant CRE could drive expression in only one of the tissues (Fig. 2D). This could in fact provide an immediate mode of retaining specificity, as a redundant CRE of this type could already exist in the cis-regulatory region of a GRN node at the time of co-option. A redundant CRE of this kind could also evolve later and restore specificity (Rebeiz & Tsiantis, 2017). However, in the two cases just described, the redundant CRE from the co-opted network is still pleiotropic. CRE sub-functionalization (Fig. 2E), in which two redundant CREs of a single gene (i.e., a redundant CRE pair) each evolve

390

Eden McQueen and Mark Rebeiz

independent roles specific to one of their initial developmental contexts, would be required to erase all pleiotropic linkages between ancestral and novel contexts (e.g., the “cis-regulatory element duplication, degeneration, and complementation” model (Monteiro & Gupta, 2016)). This may or may not be favorable, as robustness via redundant enhancers is also considered to be potentially beneficial (Barolo, 2012; Frankel et al., 2010; Perry, Boettiger, Bothma, & Levine, 2010).

4. The action of selection on co-opted networks over time Co-option is often viewed as a potential mechanism for facilitating the origins of morphological novelties. In other words, the appeal of this mechanism rests in its possible explanatory value with regard to evolution. However, while network co-option is often invoked in this way (Monteiro, 2012; Olson, 2006; Peter & Davidson, 2015; Shubin et al., 2009; True & Carroll, 2002), in order to appreciate the full evolutionary implications of this mechanism, we must carefully consider the manner in which we expect natural selection and neutral processes to operate on co-opted networks after they occur. As with any mutation, in a simple sense there are three potential fitness effects of an initiating trans change (Fig. 3A): beneficial, neutral, or deleterious (Fig. 3B). When thinking about network co-option and evolution, the added element to consider is the concomitant reduction of tissue specificity, which may have long term consequences (Fig. 3C). Below we discuss the evolutionary implications of pleiotropy at co-opted network CREs, given each of the three possible fitness consequences of the initiating mutation.

4.1 Initiating trans change with positive fitness effects If network co-option generates a novel phenotype and the net effect on fitness is beneficial, we expect that the initiating trans change, would be under positive selection (Fig. 3B, i). If the beneficial phenotype is in the novel deployment context, this situation exemplifies what is imagined to be the major upshot of network co-option as a mechanism for evolutionary change that we discussed in the introduction: a novel, beneficial, phenotype produced in one or just a few evolutionary steps. Still, while the overall phenotype may be beneficial, the effects of pleiotropy resulting from the network co-option may be detrimental, either due to the lack of modularity between the ancestral and novel traits that limits adaptation, or because some nodes

B

Net fitness consequence of change to phenotype(s)

C Fitness

Anticipated evolutionary processes immediately following

implications of trait nonindependence beneficial

Anticipated evolutionary processes at pleiotropic nodes after co-option event

i

YES

i

Initiating trans change will be subject to POSITIVE SELECTION

ii

Initiating trans change will be subject to PURIFYING SELECTION

beneficial YES Does the phenotype itself have a fitness effect?

A Initiating trans change

YES Is there a phenotype?

NO

detrimental

Does lack of trait independence have fitness implications?

detrimental

ii

iii NO

neutral

POSITIVE SELECTION on pleiotropic CREs (Conservation of genetic correlation) Alleviation of pleiotropy will proceed via POSITIVE SELECTION on novel variants that reduce pleiotropy but preserve beneficial phenotypic effects of co-option.

Fixation of novel variants that change the level of pleiotropy while preserving beneficial phenotypic effects of network co-option,will proceed via DRIFT

NO

neutral

iii Intitiating trans

change will be subject to DRIFT only

iv

Intitiating trans change will be subject to DRIFT only

iv neutral

v neutral

Fixation of any novel variants that change the level of pleiotropy will be mediated by DRIFT

Fixation of any novel variants that change the level of correlated expression will be mediated by DRIFT

Fig. 3 Fitness effects of network co-option evolution at pleiotropic nodes. Figure depicts a decision tree that partitions the evolutionary ramifications of GRN co-option and subsequent modification of pleiotropic nodes. While co-option events that have immediately advantageous effects are often discussed, here we outline a wider range of possibilities that could result from network co-option, including those with neutral outcomes. Starting on the left side of the diagram with the initiating trans change (A), we may first anticipate the evolutionary processes that would follow that mutation based on its phenotypic effects (or lack thereof ) (B). After network co-option, the fitness effects of decreased tissue-specificity are considered as the network is subject to additional modifications (C). These would be the predictions for long term selection on the affected network, given the assumption that selection regimes do not change from the initial state. As discussed in the text, selection on these networks/phenotypes may indeed change over time and affect how co-opted networks evolve, adding further complexity to this picture.

392

Eden McQueen and Mark Rebeiz

have negative pleiotropic effects (Fig. 3C, ii). In such cases, selection should favor mutations that maintain the expression of beneficial co-opted nodes in both tissues but restore specificity, for example, inactivation of particular nodes that have negative pleiotropic effects. In spite of the popularity of this view of network co-option, we do not have empirical examples that explicitly demonstrate this sequence of events. In theory, it could be that when a co-option event occurs, the resulting pleiotropy is itself advantageous, in the sense that if the two traits usually experience selection together and in the same direction, the genetic correlation between them would allow selection to operate more efficiently. A simple example would be a trait that tracks environmental conditions, such as fur thickness or cryptic coloration, expanding via co-option to a new location on the body. If selection on the character in both contexts is uniform, the novel and ancestral traits would functionally amount to only one trait with respect to the co-opted network. In such cases, there could be selection against variants that broke up pleiotropy at CREs in the co-opted network, as trait independence would represent an unnecessary increase in complexity, possibly slowing the ability of the population to adapt overall (Orr, 2000; Welch & Waxman, 2003) (Fig. 3C, i). To our knowledge this particular outcome of network co-option does not currently have direct empirical support, although the idea of selectively maintaining beneficial pleiotropy has been suggested more generally (e.g., between interacting parts such as integrated skeletomuscular traits (Karasik & Kiel, 2010)). A possible scenario of this sort could also occur in plants, where it is known that male and female floral parts (androecium and gynoecium) of some species share much of their developmental toolkit (Dornelas, Patreze, Angenent, & Immink, 2011). A correlation of male and female floral structures could be favorable in some cases if it were required for efficient pollination. Alternatively, the network co-option could confer a fitness benefit in the novel location, and the existence of pleiotropic roles of any given CRE could be neutral, or nearly neutral (Fig. 3C, iii). This could happen if there is only selection on the beneficial novel trait and the ancestral trait is either completely neutral (i.e., has no function), or if the majority of phenotypic changes to the ancestral trait via mutations at co-opted nodes would be neutral such that the pleiotropy is nearly neutral. Over time, evolution of the pleiotropic cis-regulatory regions could erode the genetic correlation by chance if changes arose to increase regulatory independence without negatively impacting the phenotype(s). Otherwise, because the CRE confers a

Network co-option and subsequent evolution

393

functional benefit to the novel context, it may be conserved and the correlation could be maintained incidentally. A study of the genetic correlations among tetrapod limb developmental serial homologs suggests that covariance structures that result from reuse of networks (as is thought to be the case with hindlimbs and forelimbs (Sears, Capellini, & Diogo, 2015)) can persist for long periods (Young & Hallgrı´msson, 2005). The authors found that the correlation between the lengths of hind limbs and forelimbs is only broken in cases of extreme functional necessity, such as is observed in the extremely divergent limb and digit proportions that enabled flight in the lineage leading to bats. However, we do not know whether the covariance structure in this case was maintained actively or passively, and the authors conclude that stabilizing selection on such correlations may often be an important factor beyond genetic constraints (Hallgrı´msson et al., 2009; Young & Hallgrı´msson, 2005). Another important possibility to consider is that the mutation confers a benefit in the ancestral context and initiates network co-option neutrally elsewhere as a byproduct. In such cases, the initiating mutation could be subject to selection irrespective of the co-option per se (this causal mechanism for selectively neutral traits is discussed in Lovejoy, Meindl, Ohman, Heiple, & White, 2002). A modeling study showed that the addition of genes to a network generally improved the “fit” of the model to its target data, which suggests that recruitment of genes to already functioning networks could be common (Spirov, Sabirov, & Holloway, 2012), and might be a source of this type of “collateral” network co-option. This could be difficult to detect, as neutral phenotypes generated by network co-option could appear to be under selection if there is selection on the genetically correlated character (Lande & Arnold, 1983). We will discuss the implications of this potential outcome more in the section on neutral fitness outcomes below.

4.2 Initiating trans change with detrimental fitness effects If network co-option is deleterious, the initiating trans change should be lost due to purifying selection (Fig. 3B, ii) unless it is fixed by drift, which is more likely in small populations and when the fitness consequence is mild (Fisher, 1930). It is also possible that the phenotypic consequences of a given co-option event on the novel tissue were initially beneficial or neutral, and only later became detrimental (e.g., accompanying a change in environment that alters selective regime or developmental plasticity, epistatic changes that

394

Eden McQueen and Mark Rebeiz

reveal larger effects on phenotype, etc.). In such cases, the upstream mutation may have already been fixed in the population. In either of the above cases, the detrimental effects of network co-option could either be eliminated by another change at the upstream trans factor that reverts the co-option, or the effects could be reduced over time by evolving tissue specific repression of downstream network nodes individually. Any given case of this latter process would be indistinguishable from an initial state of partial or aphenotypic co-option, although in some cases a comparison across species or populations that diverged after the initiating trans change might reveal a history of modifications deactivating the co-opted network.

4.3 Initiating trans change with selectively neutral effects One possible outcome of an initiating trans change that incurs network co-option is the generation of a phenotype which is completely neutral with respect to fitness (Fig. 3B, iii). Another neutral outcome, which we discussed in section one, is the possibility that no phenotype is generated in the novel tissue at all (e.g., aphenotypic co-option) (Fig. 3B, iv). In both of these cases, the genetic correlation generated between the tissues is also likely neutral, unless future mutations alter the neutrality of the phenotype or induce a new phenotype via the previously aphenotypic network. Otherwise, both the initiating trans change (Fig. 3B, iii and iv), and any future mutations that alter the genetic correlation generated by co-option would be fixed only by drift (Fig. 3C, iv and v). We have no reason at all to believe that fixation of a mutation of this type this would be more uncommon than the stochastic fixation of any other neutral mutation. This scenario is therefore especially important to consider in small populations that are more heavily influenced by drift. Modeling and analysis of changes to gene expression across species of Heliconius butterflies (Catala´n, Briscoe, & H€ ohna, 2019), fish (Whitehead & Crawford, 2006), and primates (Chaix, Somel, Kreil, Khaitovich, & Lunter, 2008; Khaitovich, P€a€abo, & Weiss, 2005) all showed that the majority of changes to gene expression across species were consistent with neutral evolution, lending credibility to this possibility. With respect to latent expression generated by co-opted networks specifically, we do not currently have examples. However, a study on the evolution of Onthophagus beetle horns suggested that exploitation of an existing expression pattern in the beetle anterodorsal head tissue was important to the evolution of the novel horn structures. A key member of this gene network, an ortholog of the Drosophila gene orthodenticle (otd), was also found to be

Network co-option and subsequent evolution

395

expressed ancestrally in the anterodorsal head tissue of an outgroup species (Tribolium castaneum), which lacks horns. Interestingly, knockdown of otd in Tribolium does not induce detectable defects in the head, suggesting that this pattern is not functional in Tribolium (Zattara et al., 2016).

5. Network co-option and the origin and diversification of traits Our breakdown of the phenomenon of network co-option in the previous sections now puts us in a position to offer a few discussion points on the relationship between network co-option and the evolvability of traits. This is in no way a comprehensive list. Our comments here will hopefully serve as a jumping-off point for further conversation. First, as we discussed in Section 1, it is important to recognize that the comprehensive case, wholesale co-option, is likely not the most common outcome. We anticipate that many more instances of network co-option will be only partial, and therefore a degree of trait independence may be retained even at the onset of network co-option. It has been suggested that intermediate levels of pleiotropy maximize evolvability (Hansen, 2003), and thus many cases of co-option may be well within the range of pleiotropic effects that do not cause serious problems for evolvability. Nevertheless, in such cases that the pleiotropy generated by network co-option acts as a constraint, many routes exist to modify CREs directly in cis or via their regulators to regain specificity, as we discussed in Section 2. Second, we must keep in mind that the effects of pleiotropy are not always detrimental. Not all phenotypes generated by co-option will initially, or ever, affect fitness, and not all co-option events will have a phenotype. These neutral outcomes would still alter modularity in the strict sense that the co-opted CREs would have decreased potential to confer tissue specificity, however there would be no immediate visibility of these events in terms of selection, and thus the evolvability of ancestral traits would not be affected, at least initially. Models that allow for neutral pleiotropic effects of co-option would improve our understanding in this area. One model predicting the degree to which pleiotropy would act as a developmental constraint revealed that the level of constraint was sensitive to changing the fitness effect of pleiotropy (Otto, 2004). As has been pointed out previously in the case of gene pleiotropy (Stern & Orgogozo, 2008), concerns about pleiotropic constraint resultant from network co-option may be

396

Eden McQueen and Mark Rebeiz

mitigated by a clearer understanding of the forms pleiotropy can take as a result of this mechanism. Beyond simply failing to obstruct the evolvability of traits, we should also keep in mind that neutral or nearly neutral outcomes of network co-option that are retained stochastically (phenotypes, expression patterns) could provide a reservoir of cryptic genetic variation. Such variation may have phenotypic and selective consequences later if subsequent mutations activate processes, such as additional network co-option events, downstream of these nodes. Initially aphenotypic outcomes of co-option might therefore contribute positively to trait evolvability (e.g., in cases of preadaptation). This possibility has been noted before (True & Carroll, 2002), but we currently lack empirical examples to support this conjecture. However, there is a growing interest in understanding how cellular and morphological phenotypes may evolve neutrally (Ruths & Nakhleh, 2013; Wideman, Novick, Mun˜oz-Go´mez, & Doolittle, 2019; Zhang, 2018). A recent study on cryptic genetic variation demonstrated that neutral mutations accumulated at the level of a single protein facilitated subsequent adaptation of that protein (Zheng, Payne, & Wagner, 2019). This result might scale up to the level of networks. More examples such as these that examine multi-gene interactions, and especially comparative analyses of the network architecture of such cases will help us understand the role of network co-option in the generation of cryptic genetic variation. The implications of the observations above are magnified when we consider that the simple version of co-option that begins with GRN deployment in one tissue and expands to deployment in two tissues is probably not realistic. More extensive effects across many tissues are likely to be common. As networks evolve downstream of newly redeployed nodes after network co-option, a complex collage forms rather than a pre-made template which is simply “copy-pasted” to a new location. Indeed, this view is supported by a mathematical modeling study, which demonstrated that the construction of a novel expression domain is facilitated by reuse of multiple but distinct existing modules that contribute to that domain elsewhere (Espinosa-Soto & Wagner, 2010). Such a scenario, wherein CRE pleiotropy is spread out over multiple ancestral GRNs, might lend more flexibility to circumventing developmental constraints for both the ancestral and novel structures. To be sure, there are likely to be many cases where network co-option does result in constraint on some properties. For example, it has been suggested that limb outgrowth was constrained to have anteriorposterior polarity due to the fact that Hox genes were co-opted to initiate extension from the body wall (Tarchini, Duboule, & Kmita, 2006).

Network co-option and subsequent evolution

397

We are still in the process of discovery in the area of network co-option, and there are many ways forward. With respect to modeling, it would be very enlightening to incorporate network co-option into dynamical models (Alexander, Kim, Emonet, & Gerstein, 2009; Irons & Monk, 2007; Verd, Monk, & Jaeger, 2019), which take spatio-temporal information into account when designating modules. Models incorporating some of the neutral outcomes of co-option that we discussed here would also be very useful. Beyond model development, many more empirical examples of network co-option are needed. In particular, to gain insight into the evolution of co-opted networks, we need examples wherein the structure of known or suspected co-opted networks is compared across species. One study investigated the expression of genes in the network co-opted to generate eyespots across 21 species of Nymphalid butterflies (Oliver et al., 2012). They showed that the expression of some network members was highly conserved, whereas repeated losses of others suggests that these nodes were more evolutionarily labile or possibly not necessary in the first place. More examples like this one would greatly improve our understanding of how co-opted networks are incorporated into existing networks and change over time. We hope the framework that we have outlined highlights the scope of possibilities that accompany network co-option and inspires a wide range of research questions into this intriguing mechanism of developmental evolution.

6. Concluding remarks In writing this review, we were reminded of the way that general thinking has progressed with regard to genes. What began by attributing strict functional identities to individual genes (“a gene for function x”), eventually became more nuanced in light of empirical data that was inconsistent with a one-to-one view (Duboule & Wilkins, 1998). Considering network reuse as a mechanism of altering development similarly complicates our concept of GRN identity. GRNs are not tidy, self-contained “programs” for specific traits (DiFrisco & Jaeger, 2019; Nijhout, 1990), but are instead highly context-dependent and may therefore yield different outcomes in different developmental circumstances. This suggests that we must caution ourselves against falling into a “GRN for function x” trap, and instead recognize that the GRN that produces any given trait will be a haphazard assembly of parts, often with a few spare odds and ends, drawn from existing GRNs over evolutionary time. Like all products of evolution, GRNs will be the result of evolutionary “tinkering” ( Jacob, 1977): functional, but messy.

398

Eden McQueen and Mark Rebeiz

Acknowledgments We would like to acknowledge Aaron Novick, Gavin Rice, and Ben Vincent for their helpful comments on this manuscript. Our work on the specificity of co-opted networks is supported by the National Institutes of Health (GM112758 to MR).

Glossary Cis-regulatory element (CRE) A stretch of non-coding DNA that is physically upstream, downstream, or in the intron, of a given gene, and which influences the expression of that gene at some time(s) and location(s) during the development or adult life of the organism. Developmental context A temporal and spatial domain in a tissue during which specific developmental events, such as the activation of a GRN or morphogenetic process occurs. Epistasis The condition in which the phenotypic effect or effects of a particular allele at a particular genetic locus are influenced by the presence of one or more alleles elsewhere in the genome. Evolvability Evolvability has many definitions (for discussion see Pigliucci, 2008), but can be roughly defined as the ability of a system to change adaptively. Gene regulatory network (GRN) Semi-autonomous regulatory modules responsible for characters or phenotypes. This is by no means the only way to define GRNs, as we suggest in our discussion. However, a full discussion of gene regulatory network ontology is outside the scope of this work. Initiating trans change The mutation(s) that cause the deployment of a regulatory factor (e.g., transcription factor, signaling molecule) in a novel developmental context to activate its downstream network, initiating a co-option event. Initiating trans factor In network co-option, the regulatory factor whose activity in the novel context was triggered by the initiating trans change, and which is responsible for the co-option of downstream network. Modularity With respect to development, the phenomenon of partial independence of organismal parts during development (Bolker, 2000; Wagner & Altenberg, 1996). Phenotype A measurable morphological or behavioral character, trait, behavior, or quality. Pleiotropy The condition in which a single mutation causally affects (alters) two or more traits. Regulatory state The total set of regulatory factors (transcription factors, co-factors, and signaling molecules) present, and their concentrations, at a given time and in a given cell or specified location during development (Peter & Davidson, 2015). Terminal effector A gene that contributes directly, via its participation in control of cellular proteins, to the mechanical behavior or physical phenotype of a cell or a group of neighboring cells (Smith, Mark, & Davidson, 2018). Trans-landscape See Regulatory State, above.

References Alexander, R. P., Kim, P. M., Emonet, T., & Gerstein, M. B. (2009). Understanding modularity in molecular networks requires dynamics. Science Signaling, 2(81), 1–4. pe44. https://doi.org/10.1126/scisignal.281pe44. de Almeida, A. M. R., Yockteng, R., Schnable, J., Alvarez-Buylla, E. R., Freeling, M., & Specht, C. D. (2014). Co-option of the polarity gene network shapes filament morphology in angiosperms. Scientific Reports, 4(1), 1–9. https://doi.org/10.1038/srep06194.

Network co-option and subsequent evolution

399

´ lvarez-Buylla, E. R., Ambrose, B. A., Flores-Sandoval, E., Englund, M., Garay-Arroyo, A., A Garcı´a-Ponce, B., et al. (2010). B-function expression in the flower center underlies the homeotic phenotype of Lacandonia schismatica (Triuridaceae). The Plant Cell, 22(11), 3543–3559. https://doi.org/10.1105/tpc.109.069153. Barolo, S. (2012). Shadow enhancers: Frequently asked questions about distributed Cisregulatory information and enhancer redundancy. BioEssays: News and Reviews in Molecular, Cellular and Developmental Biology, 34(2), 135–141. https://doi.org/10.1002/ bies.201100121. Barolo, S., & Posakony, J. W. (2002). Three habits of highly effective signaling pathways: Principles of transcriptional control by developmental cell signaling. Genes & Development, 16(10), 1167–1181. https://doi.org/10.1101/gad.976502. Bolker, J. A. (2000). Modularity in development and why it matters to Evo-Devo. Integrative and Comparative Biology, 40(5), 770–776. https://doi.org/10.1093/icb/40.5.770. Brunetti, C. R., Selegue, J. E., Monteiro, A., French, V., Brakefield, P. M., & Carroll, S. B. (2001). The generation and diversification of butterfly eyespot color patterns. Current Biology, 11(20), 1578–1585. https://doi.org/10.1016/s0960-9822(01)00502-4. Carroll, S. B., Gates, J., Keys, D. N., Paddock, S. W., Panganiban, G. E., Selegue, J. E., et al. (1994). Pattern formation and eyespot determination in butterfly wings. Science (New York, N.Y.), 265(5168), 109–114. https://doi.org/10.1126/science.7912449. Carroll, S. B. (2008). Evo-Devo and an expanding evolutionary synthesis: A genetic theory of morphological evolution. Cell, 134(1), 25–36. https://doi.org/10.1016/j.cell.2008.06.030. ohna, S. (2019). Drift and directional selection are the evoCatala´n, A., Briscoe, A. D., & H€ lutionary forces driving gene expression divergence in eye and brain tissue of Heliconius butterflies. Genetics, 213(2), 581–594. https://doi.org/10.1534/genetics.119.302493. Cebra-Thomas, J., Tan, F., Sistla, S., Estes, E., Bender, G., Kim, C., et al. (2005). How the turtle forms its shell: A paracrine hypothesis of carapace formation. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, 304B(6), 558–569. https://doi.org/ 10.1002/jez.b.21059. Chaix, R., Somel, M., Kreil, D. P., Khaitovich, P., & Lunter, G. A. (2008). Evolution of primate gene expression: Drift and corrective sweeps? Genetics, 180(3), 1379–1389. https://doi.org/10.1534/genetics.108.089623. Coen, E. S., & Meyerowitz, E. M. (1991). The war of the whorls: Genetic interactions controlling flower development. Nature, 353(6339), 31–37. https://doi.org/10.1038/ 353031a0. Davidson, E. H. (2010). The regulatory genome: Gene regulatory networks in development and evolution. Elsevier. Davidson, E. H., & Erwin, D. H. (2006). Gene regulatory networks and the evolution of animal body plans. Science, 311(5762), 796–800. Davidson, L. A. (2012). Epithelial machines that shape the embryo. Trends in Cell Biology, 22(2), 82–87. https://doi.org/10.1016/j.tcb.2011.10.005. DiFrisco, J., & Jaeger, J. (2019). Beyond networks: Mechanism and process in Evo-Devo. Biology and Philosophy, 34(6), 54. https://doi.org/10.1007/s10539-019-9716-9. Dinwiddie, A., & Rachootin, S. (2011). Patterning of a compound eye on an extinct dipteran wing. Biology Letters, 7(2), 281–284. Dornelas, M. C., Patreze, C. M., Angenent, G. C., & Immink, R. G. H. (2011). MADS: The missing link between identity and growth? Trends in Plant Science, 16(2), 89–97. https:// doi.org/10.1016/j.tplants.2010.11.003. Duboule, D., & Wilkins, A. S. (1998). The evolution of ‘bricolage’. Trends in Genetics, 14(2), 54–59. https://doi.org/10.1016/S0168-9525(97)01358-9. Emlen, D. J., Lavine, L. C., & Ewen-Campen, B. (2007). On the origin and evolutionary diversification of beetle horns. Proceedings of the National Academy of Sciences of the United States of America, 104(Suppl. 1), 8661–8668. https://doi.org/10.1073/pnas. 0701209104.

400

Eden McQueen and Mark Rebeiz

Erwin, D. H. (2020). Evolutionary dynamics of gene regulation. Current Topics in Developmental Biology this issue. Espinosa-Soto, C., & Wagner, A. (2010). Specialization can drive the evolution of modularity. PLoS Computational Biology, 6(3), e1000719. https://doi.org/10.1371/journal. pcbi.1000719. Farley, E. K., Olson, K. M., & Levine, M. S. (2015). Regulatory principles governing tissue specificity of developmental enhancers. Cold Spring Harbor Symposia on Quantitative Biology, 80, 27–32. https://doi.org/10.1101/sqb.2015.80.027227. Faust, T. B., Binning, J. M., Gross, J. D., & Frankel, A. D. (2017). Making sense of multifunctional proteins: Human immunodeficiency virus type 1 accessory and regulatory proteins and connections to transcription. Annual Review of Virology, 4(1), 241–260. https://doi.org/10.1146/annurev-virology-101416-041654. Fisher, C. R., Wegrzyn, J. L., & Jockusch, E. L. (2019). Co-option of wing-patterning genes underlies the evolution of the treehopper helmet. Nature Ecology & Evolution, 4, 1–11. https://doi.org/10.1038/s41559-019-1054-4. Fisher, R. A. (1930). The genetical theory of natural selection. Oxford: Clarendon Press. Frankel, N., Davis, G. K., Vargas, D., Wang, S., Payre, F., & Stern, D. L. (2010). Phenotypic robustness conferred by apparently redundant transcriptional enhancers. Nature, 466(7305), 490–493. https://doi.org/10.1038/nature09158. Gao, F., & Davidson, E. H. (2008). Transfer of a large gene regulatory apparatus to a new developmental address in echinoid evolution. Proceedings of the National Academy of Sciences of the United States of America, 105(16), 6091–6096. https://doi.org/10.1073/ pnas.0801201105. Garcı´a-Bellido, A. (1975). Genetic control of wing disc development in Drosophila. Ciba Foundation Symposium, 0(29), 161–182. https://doi.org/10.1002/9780470720110.ch8. Glassford, W. J., Johnson, W. C., Dall, N. R., Smith, S. J., Yang, L., Boll, W., et al. (2015). Co-option of an ancestral Hox-regulated network underlies a recently evolved morphological novelty. Developmental Cell, 34(5), 520–531. https://doi.org/10.1016/j.devcel. 2015.08.005. Gompel, N., Prud’homme, B., Wittkopp, P. J., Kassner, V. A., & Carroll, S. B. (2005). Chance caught on the wing: Cis-regulatory evolution and the origin of pigment patterns in Drosophila. Nature, 433(7025), 481–487. https://doi.org/10.1038/nature03235. Green, S., & Batterman, R. (2017). Biology meets physics: Reductionism and multi-scale modeling of morphogenesis. Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences, 61, 20–34. https://doi.org/ 10.1016/j.shpsc.2016.12.003. Halder, G., Callaerts, P., & Gehring, W. J. (1995). Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science, 267(5205), 1788–1792. https:// doi.org/10.1126/science.7892602. Hallgrı´msson, B., Jamniczky, H., Young, N. M., Rolian, C., Parsons, T. E., Boughner, J. C., et al. (2009). Deciphering the palimpsest: Studying the relationship between morphological integration and phenotypic covariation. Evolutionary Biology, 36(4), 355–376. https:// doi.org/10.1007/s11692-009-9076-5. Hansen, T. F. (2003). Is modularity necessary for evolvability?: Remarks on the relationship between pleiotropy and evolvability. Biosystems, 69(2), 83–94. https://doi.org/10.1016/ S0303-2647(02)00132-6. Harris, M. P., Fallon, J. F., & Prum, R. O. (2002). Shh-Bmp2 signaling module and the evolutionary origin and diversification of feathers. Journal of Experimental Zoology, 294(2), 160–176. https://doi.org/10.1002/jez.10157. Harris, M. P., Williamson, S., Fallon, J. F., Meinhardt, H., & Prum, R. O. (2005). Molecular evidence for an activator–inhibitor mechanism in development of embryonic feather

Network co-option and subsequent evolution

401

branching. Proceedings of the National Academy of Sciences, 102(33), 11734–11739. https:// doi.org/10.1073/pnas.0500781102. Hinman, V. F., Nguyen, A., & Davidson, E. H. (2007). Caught in the evolutionary act: Precise cis-regulatory basis of difference in the organization of gene networks of sea stars and sea urchins. Developmental Biology, 312(2), 584–595. https://doi.org/10.1016/j. ydbio.2007.09.006. Hong, J.-W., Hendrix, D. A., & Levine, M. S. (2008). Shadow enhancers as a source of evolutionary novelty. Science (New York, N.Y.), 321(5894), 1314. https://doi.org/10.1126/ science.1160631. Irons, D. J., & Monk, N. A. M. (2007). Identifying dynamical modules from genetic regulatory systems: Applications to the segment polarity network. BMC Bioinformatics, 8(1), 413. https://doi.org/10.1186/1471-2105-8-413. Jacob, F. (1977). Evolution and tinkering. Science, 196(4295), 1161–1166. Smith, S. J., Mark, R., & Davidson, L. (2018). From pattern to process: Studies at the interface of gene regulatory networks, morphogenesis, and evolution. Current Opinion in Genetics & Development, 51, 103–110. https://doi.org/10.1016/j.gde.2018.08.004. Janssens, H., Hou, S., Jaeger, J., Kim, A.-R., Myasnikova, E., Sharp, D., et al. (2006). Quantitative and predictive model of transcriptional control of the drosophila melanogaster even skipped gene. Nature Genetics, 38(10), 1159–1165. https://doi.org/10.1038/ng1886. Karasik, D., & Kiel, D. P. (2010). Evidence for pleiotropic factors in genetics of the musculoskeletal system. Bone, 46(5), 1226–1237. https://doi.org/10.1016/j.bone.2010.01.382. Keys, D. N., Lewis, D. L., Selegue, J. E., Pearson, B. J., Goodrich, L. V., Johnson, R. L., et al. (1999). Recruitment of a hedgehog regulatory circuit in butterfly eyespot evolution. Science, 283(5401), 532–534. Khaitovich, P., P€a€abo, S., & Weiss, G. (2005). Toward a neutral evolutionary model of gene expression. Genetics, 170(2), 929–939. https://doi.org/10.1534/genetics.104.037135. Lande, R., & Arnold, S. J. (1983). The measurement of selection on correlated characters. Evolution, 37(6), 1210–1226. https://doi.org/10.2307/2408842. Lemons, D., Fritzenwanker, J. H., Gerhart, J., Lowe, C. J., & McGinnis, W. (2010). Co-option of an anteroposterior head axis patterning system for proximodistal patterning of appendages in early bilaterian evolution. Developmental Biology, 344(1), 358–362. https://doi.org/10.1016/j.ydbio.2010.04.022. Levine, M. (2010). Transcriptional enhancers in animal development and evolution. Current Biology: CB, 20(17), R754–R763. https://doi.org/10.1016/j.cub.2010.06.070. Loredo, G. A., Brukman, A., Harris, M. P., Kagle, D., Leclair, E. E., Gutman, R., et al. (2001). Development of an evolutionarily novel structure: Fibroblast growth factor expression in the carapacial ridge of turtle embryos. Journal of Experimental Zoology, 291(3), 274–281. https://doi.org/10.1002/jez.1103. Lovejoy, C. O., Meindl, R. S., Ohman, J. C., Heiple, K. G., & White, T. D. (2002). The Maka femur and its bearing on the antiquity of human walking: Applying contemporary concepts of morphogenesis to the human fossil record. American Journal of Physical Anthropology, 119(2), 97–133. https://doi.org/10.1002/ajpa.10111. Luscombe, N. M., Madan Babu, M., Yu, H., Snyder, M., Teichmann, S. A., & Gerstein, M. (2004). Genomic analysis of regulatory network dynamics reveals large topological changes. Nature, 431(7006), 308–312. https://doi.org/10.1038/nature02782. Mackay, T. F. C. (2014). Epistasis and quantitative traits: Using model organisms to study gene–gene interactions. Nature Reviews Genetics, 15(1), 22–33. https://doi.org/10.1038/ nrg3627. Mann, R. S., & Carroll, S. B. (2002). Molecular mechanisms of selector gene function and evolution. Current Opinion in Genetics & Development, 12(5), 592–600. https://doi.org/ 10.1016/S0959-437X(02)00344-1.

402

Eden McQueen and Mark Rebeiz

Marcellini, S., & Simpson, P. (2006). Two or four bristles: Functional evolution of an enhancer of scute in Drosophilidae. PLoS Biology, 4(12), e386. https://doi.org/10.1371/ journal.pbio.0040386. Minafra, L., Bravata`, V., Forte, G. I., Cammarata, F. P., Gilardi, M. C., & Messa, C. (2014). Gene expression profiling of epithelial–mesenchymal transition in primary breast cancer cell culture. Anticancer Research, 34(5), 2173–2183. Moczek, A. P. (2009). On the origins of novelty and diversity in development and evolution: A case study on beetle horns. Cold Spring Harbor Symposia on Quantitative Biology, 74, 289–296. https://doi.org/10.1101/sqb.2009.74.010. Moczek, A. P., & Nagy, L. M. (2005). Diverse developmental mechanisms contribute to different levels of diversity in horned beetles. Evolution & Development, 7(3), 175–185. https://doi.org/10.1111/j.1525-142X.2005.05020.x. Moczek, A. P., & Rose, D. J. (2009). Differential recruitment of limb patterning genes during development and diversification of beetle horns. Proceedings of the National Academy of Sciences of the United States of America, 106(22), 8992–8997. https://doi.org/10.1073/ pnas.0809668106. Monteiro, A., & Gupta, M. D. (2016). Identifying coopted networks and causative mutations in the origin of novel complex traits. Current Topics in Developmental Biology, 119, 205–226. https://doi.org/10.1016/bs.ctdb.2016.03.003. Monteiro, A. (2012). Gene regulatory networks reused to build novel traits. BioEssays, 34(3), 181–186. https://doi.org/10.1002/bies.201100160. Nagy, O., Nuez, I., Savisaar, R., Peluffo, A. E., Yassin, A., Lang, M., et al. (2018). Correlated evolution of two copulatory organs via a single cis-regulatory nucleotide change. Current Biology, 28(21), 3450–3457.e13. https://doi.org/10.1016/j.cub.2018.08.047. Nakamura, T., Klomp, J., Pieretti, J., Schneider, I., Gehrke, A. R., & Shubin, N. H. (2015). Molecular mechanisms underlying the exceptional adaptations of batoid fins. Proceedings of the National Academy of Sciences of the United States of America, 112(52), 15940–15945. https://doi.org/10.1073/pnas.1521818112. Nijhout, H. F. (1990). Metaphors and the role of genes in development. BioEssays: News and Reviews in Molecular, Cellular and Developmental Biology, 12, 441–446. Oliver, J. C., Tong, X.-L., Gall, L. F., Piel, W. H., & Monteiro, A. (2012). A single origin for nymphalid butterfly eyespots followed by widespread loss of associated gene expression. PLoS Genetics, 8(8), e1002893. https://doi.org/10.1371/journal.pgen.1002893. Olson, E. N. (2006). Gene regulatory networks in the evolution and development of the heart. Science (New York, NY), 313(5795), 1922–1927. https://doi.org/10.1126/ science.1132292. Orr, H. A. (2000). Adaptation and the cost of complexity. Evolution; International Journal of Organic Evolution, 54(1), 13–20. https://doi.org/10.1111/j.0014-3820.2000.tb00002.x. Otto, S. P. (2004). Two steps forward, one step back: The pleiotropic effects of favoured alleles. Proceedings of the Royal Society B: Biological Sciences, 271(1540), 705–714. € Ozsu, N., & Monteiro, A. (2017). Wound healing, calcium signaling, and other novel pathways are associated with the formation of butterfly eyespots. BMC Genomics, 18(October), 788. https://doi.org/10.1186/s12864-017-4175-7. Pavlicev, M., Kenney-Hunt, J. P., Norgard, E. A., Roseman, C. C., Wolf, J. B., & Cheverud, J. M. (2008). Genetic variation in pleiotropy: Differential epistasis as a source of variation in the allometric relationship between long bone lengths and body weight. Evolution, 62(1), 199–213. https://doi.org/10.1111/j.1558-5646.2007.00255.x. Perry, M. W., Boettiger, A. N., Bothma, J. P., & Levine, M. (2010). Shadow enhancers foster robustness of Drosophila gastrulation. Current Biology: CB, 20(17), 1562–1567. https:// doi.org/10.1016/j.cub.2010.07.043. Peter, I. S., & Davidson, E. H. (2015). Genomic control process: Development and evolution (1st ed.). London, UK; San Diego, CA, USA: Academic Press.

Network co-option and subsequent evolution

403

Phillips, P. C. (2008). Epistasis—The essential role of gene interactions in the structure and evolution of genetic systems. Nature Reviews. Genetics, 9(11), 855–867. https://doi.org/ 10.1038/nrg2452. Pigliucci, M. (2008). Is evolvability evolvable? Nature Reviews Genetics, 9(1), 75–82. https:// doi.org/10.1038/nrg2278. Noon, P.-B., Ella, G. S., Ortiz, D. M., Sager, J., Liebowitz, A., Stern, D. L., et al. (2018). Comprehensive analysis of a cis-regulatory region reveals pleiotropy in enhancer function. Cell Reports, 22(11), 3021–3031. https://doi.org/10.1016/j.celrep.2018.02.073. Prud’homme, B., Minervino, C., Hocine, M., Cande, J. D., Aouane, A., Dufour, H. D., et al. (2011). Body plan innovation in treehoppers through the evolution of an extra wing-like appendage. Nature, 473(7345), 83–86. https://doi.org/10.1038/nature09977. Prum, R. O. (1999). Development and evolutionary origin of feathers. The Journal of Experimental Zoology, 285(4), 291–306. Prum, R. O. (2005). Evolution of the morphological innovations of feathers. Journal of Experimental Zoology. Part B, Molecular and Developmental Evolution, 304(6), 570–579. https://doi.org/10.1002/jez.b.21073. Prum, R. O., & Brush, A. H. (2002). The evolutionary origin and diversification of feathers. The Quarterly Review of Biology, 77(3), 261–295. https://doi.org/10.1086/341993. Rebeiz, M., Jikomes, N., Kassner, V. A., & Carroll, S. B. (2011). Evolutionary origin of a novel gene expression pattern through co-option of the latent activities of existing regulatory sequences. Proceedings of the National Academy of Sciences of the United States of America, 108(25), 10036–10043. https://doi.org/10.1073/pnas.1105937108. Rebeiz, M., Patel, N. H., & Hinman, V. F. (2015). Unraveling the tangled skein: The evolution of transcriptional regulatory networks in development. Annual Review of Genomics and Human Genetics, 16(1), 103–131. https://doi.org/10.1146/annurev-genom-091212153423. Rebeiz, M., & Tsiantis, M. (2017). Enhancer evolution and the origins of morphological novelty. Current Opinion in Genetics & Development, 45(August), 115–123. https://doi. org/10.1016/j.gde.2017.04.006. Rice, G., & Rebeiz, M. (2019). Evolution: How many phenotypes do regulatory mutations affect? Current Biology, 29(1), R21–R23. https://doi.org/10.1016/j.cub.2018.11.027. Rosenthal, R. S. (1979). The file drawer problem and tolerance for null results. Psychological Bulletin, 86, 638–641. https://doi.org/10.1037/0033-2909.86.3.638. Ruths, T., & Nakhleh, L. (2013). Neutral forces acting on intragenomic variability shape the Escherichia coli regulatory network topology. Proceedings of the National Academy of Sciences of the United States of America, 110(19), 7754–7759. Sabarı´s, G., Laiker, I., Noon, E. P.-B., & Frankel, N. (2019). Actors with multiple roles: Pleiotropic enhancers and the paradigm of enhancer modularity. Trends in Genetics, 35(6), 423–433. https://doi.org/10.1016/j.tig.2019.03.006. Saeij, J. P. J., Coller, S., Boyle, J. P., Jerome, M. E., White, M. W., & Boothroyd, J. C. (2007). Toxoplasma co-opts host gene expression by injection of a polymorphic kinase homologue. Nature, 445(7125), 324–327. https://doi.org/10.1038/nature05395. Schneuwly, S., Klemenz, R., & Gehring, W. J. (1987). Redesigning the body plan of Drosophila by ectopic expression of the homoeotic gene Antennapedia. Nature, 325(6107), 816–818. https://doi.org/10.1038/325816a0. Sears, K. E., Capellini, T. D., & Diogo, R. (2015). On the serial homology of the pectoral and pelvic girdles of tetrapods. Evolution, 69(10), 2543–2555. Shah, S. N., Cope, L., Poh, W., Belton, A., Sujayita, R., Conover Talbot, C., Jr., et al. (2013). HMGA1: A master regulator of tumor progression in triple-negative breast cancer cells. PLoS One, 8(5), e63419. https://doi.org/10.1371/journal.pone.0063419. Shubin, N., Tabin, C., & Carroll, S. (2009). Deep homology and the origins of evolutionary novelty. Nature, 457(7231), 818–823. https://doi.org/10.1038/nature07891.

404

Eden McQueen and Mark Rebeiz

Spirov, A. V., Sabirov, M. A., & Holloway, D. M. (2012). In silico evolution of gene cooption in pattern-forming gene networks. The Scientific World Journal, 2012 (December), 1–19. https://doi.org/10.1100/2012/560101. Staller, M. V., Dong, Y., Randklev, S., Bragdon, M. D., Wunderlich, Z. B., Tao, R., et al. (2013). Depleting gene activities in early Drosophila embryos with the ‘maternal-Gal4– ShRNA’ System. Genetics, 193(1), 51–61. https://doi.org/10.1534/genetics.112.144915. Stern, D. L., & Orgogozo, V. (2008). The loci of evolution: How predictable is genetic evolution? Evolution; International Journal of Organic Evolution, 62(9), 2155–2177. https://doi. org/10.1111/j.1558-5646.2008.00450.x. Stern, D. L., & Orgogozo, V. (2009). Is genetic evolution predictable? Science, 323(5915), 746–751. Suryamohan, K., Hanson, C., Andrews, E., Sinha, S., Scheel, M. D., & Halfon, M. S. (2016). Redeployment of a conserved gene regulatory network during Aedes aegypti development. Developmental Biology, 416(2), 402–413. https://doi.org/10.1016/ j.ydbio.2016.06.031. Tarchini, B., Duboule, D., & Kmita, M. (2006). Regulatory constraints in the evolution of the tetrapod limb anterior–posterior polarity. Nature, 443(7114), 985–988. https://doi. org/10.1038/nature05247. Tomoyasu, Y., Arakane, Y., Kramer, K. J., & Denell, R. E. (2009). Repeated co-options of exoskeleton formation during wing-to-elytron evolution in beetles. Current Biology: CB, 19(24), 2057–2065. https://doi.org/10.1016/j.cub.2009.11.014. True, J. R., & Carroll, S. B. (2002). Gene co-option in physiological and morphological evolution. Annual Review of Cell and Developmental Biology, 18(1), 53–80. https://doi.org/ 10.1146/annurev.cellbio.18.020402.140619. Verd, B., Monk, N. A. M., & Jaeger, J. (2019). Modularity, criticality, and evolvability of a developmental gene regulatory network. Edited by Lee Altenberg and Patricia J Wittkopp, eLife, 8, e42832. https://doi.org/10.7554/eLife.42832. Vincent, B. J., Staller, M. V., Lopez-Rivera, F., Bragdon, M. D. J., Pym, E. C. G., Biette, K. M., et al. (2018). Hunchback is counter-repressed to regulate even-skipped stripe 2 expression in Drosophila embryos. PLoS Genetics, 14(9), e1007644. https:// doi.org/10.1371/journal.pgen.1007644. Wagner, G. P., & Altenberg, L. (1996). Perspective: Complex adaptations and the evolution of evolvability. Evolution, 50(3), 967–976. https://doi.org/10.2307/2410639. Wagner, G. P., Kenney-Hunt, J. P., Pavlicev, M., Peck, J. R., Waxman, D., & Cheverud, J. M. (2008). Pleiotropic scaling of gene effects and the ‘cost of complexity’. Nature, 452(7186), 470–472. https://doi.org/10.1038/nature06756. Wagner, G. P., & Zhang, J. (2011). The pleiotropic structure of the genotype–phenotype map: The evolvability of complex organisms. Nature Reviews Genetics, 12(3), 204–213. https://doi.org/10.1038/nrg2949. Wasik, B. R., & Moczek, A. P. (2011). Decapentaplegic (Dpp) regulates the growth of a morphological novelty, beetle horns. Development Genes and Evolution, 221(1), 17–27. https://doi.org/10.1007/s00427-011-0355-7. Welch, J. J., & Waxman, D. (2003). Modularity and the cost of complexity. Evolution, 57(8), 1723–1734. https://doi.org/10.1111/j.0014-3820.2003.tb00581.x. West-Eberhard, M. J. (2005). Developmental plasticity and the origin of species differences. Proceedings of the National Academy of Sciences of the United States of America, 102(Suppl. 1), 6543–6549. https://doi.org/10.1073/pnas.0501844102. Whitehead, A., & Crawford, D. L. (2006). Neutral and adaptive variation in gene expression. Proceedings of the National Academy of Sciences of the United States of America, 103(14), 5425–5430. https://doi.org/10.1073/pnas.0507648103.

Network co-option and subsequent evolution

405

Wideman, J. G., Novick, A., Mun˜oz-Go´mez, S. A., & Doolittle, W. F. (2019). Neutral evolution of cellular phenotypes. Current Opinion in Genetics & Development, 58–59, 87–94. https://doi.org/10.1016/j.gde.2019.09.004. Womack, M. C., Metz, M. J., & Hoke, K. L. (2019). Larger genomes linked to slower development and loss of late-developing traits. The American Naturalist, 194, 854–864. https:// doi.org/10.1086/705897. Wray, G. A. (2007). The evolutionary significance of cis-regulatory mutations. Nature Reviews. Genetics, 8, 206–216. Wunderlich, Z., Bragdon, M. D. J., Vincent, B. J., White, J. A., Estrada, J., & DePace, A. H. (2015). Kr€ uppel expression levels are maintained through compensatory evolution of shadow enhancers. Cell Reports, 12(11), 1740–1747. https://doi.org/10.1016/j.celrep. 2015.08.021. Young, N. M., & Hallgrı´msson, B. (2005). Serial homology and the evolution of mammalian limb covariation structure. Evolution, 59(12), 2691–2704. https://doi.org/10.1111/ j.0014-3820.2005.tb00980.x. Zakany, J., & Duboule, D. (2007). The role of Hox genes during vertebrate limb development. Current Opinion in Genetics & Development, 17(4), 359–366. https://doi.org/10.1016/j.gde. 2007.05.011. Zattara, E. E., Busey, H. A., Linz, D. M., Tomoyasu, Y., & Moczek, A. P. (2016). Neofunctionalization of embryonic head patterning genes facilitates the positioning of novel traits on the dorsal head of adult beetles. Proceedings of the Royal Society B: Biological Sciences, 283(1834), 1–8. https://doi.org/10.1098/rspb.2016.0824. Zhang, J. (2018). Neutral theory and phenotypic evolution. Molecular Biology and Evolution, 35(6), 1327–1331. Zheng, J., Payne, J. L., & Wagner, A. (2019). Cryptic genetic variation accelerates evolution by opening access to diverse adaptive peaks. Science, 365(6451), 347–353. https://doi. org/10.1126/science.aax1837.

CHAPTER THIRTEEN

Evolutionary dynamics of gene regulation Douglas H. Erwin∗ Department of Paleobiology, National Museum of Natural History, Washington, DC, United States ∗ Corresponding author: e-mail address: [email protected]

Contents 1. Conceptual approaches to understanding the evolution of gene regulation 2. The origin of metazoan regulatory novelties 3. The expansion and exploitation of combinatoric complexity 4. Evolutionary repatterning of gene regulatory networks 5. Micro- and macroevolutionary changes in gene regulatory networks 6. Do novel morphologies involve mechanistically distinct pathways? 7. Towards a general theory of GRN evolution 8. Concluding remarks Acknowledgments References Further reading

408 409 412 414 419 421 423 424 426 426 431

Abstract The long controversy over the importance of changes in the regulatory genome has been resolved with the recognition that such changes are a fundamental component of evolutionary dynamics. Comparative studies have revealed four dominant modes of change as the regulatory genome evolved: (1) the origin of regulatory novelties such as distal enhancers and new types of promoters at the origin of Metazoa; (2) the expansion of regulatory capacity, most notably with diversification of transcription factors. Together these changes expanded the available combinatoric complexity of regulatory interactions and allow an increase in the variety of cell types. There are two more common modes of regulatory evolution: (3) Repatterning of gene regulatory networks. Such repatterning largely involves the introduction of transposons, promoter switching, co-option of regulatory genes or subcircuits, recombination, and the de novo generation of new regulatory sequences. Finally, (4) changes in enhancer and promoter specificity enable fine-scale adaptive changes. One of the outstanding issues at the intersection of evolutionary and developmental biology is how these various modes of regulatory evolution translate to morphological change, and particularly macroand microevolutionary patterns and whether evolutionary novelties are associated with distinctive patterns of regulatory change.

Current Topics in Developmental Biology, Volume 139 ISSN 0070-2153 https://doi.org/10.1016/bs.ctdb.2020.02.006

2020 Published by Elsevier Inc.

407

408

Douglas H. Erwin

1. Conceptual approaches to understanding the evolution of gene regulation Changes to the regulatory genome, particularly those associated with cis-regulatory modules (CRMs), play a central role in morphological evolution ranging from the adaptive diversification of color patterns in the wings of flies and butterflies to the origin of novel morphological features. Sequencing and comparative developmental and genomic studies have illuminated the complexity of evolutionary dynamics in the regulatory genome. A variety of approaches have been pursued in exploring these evolutionary dynamics. These include focusing on particular components such as enhancers or transcription factors (Halfon, 2019; Villar et al., 2015), exploring the processes responsible for generating particular aspects of morphology, such as arthropod or vertebrate limbs (Petit, Sears, & Ahituv, 2017), neural crest (Green, Simoes-Costa, & Bronner, 2015; Sauka-Spengler, Meulemans, Jones, & Bronner-Fraser, 2007), or flowers (Irish & Litt, 2005). Particular processes of regulatory change have received attention, such as co-option of regulatory subcircuits (Peter & Davidson, 2015; True & Carroll, 2002), the evolutionary dynamics of chromatin (Malik & Henikoff, 2009; Yadav, Quivy, & Almouzni, 2018) and gene duplication (Holland, Marletaz, Maeso, Dunwell, & Paps, 2017). The regulatory network architecture is particularly important as it influences developmental outcomes so that changes at different nodes within the network will have different phenotypic effects (Davidson & Erwin, 2010; Erwin & Davidson, 2009; Peter & Davidson, 2015; Rebeiz, Patel, & Hinman, 2015). Each approach provides distinct perspectives on the evolution of the regulatory genome. These different approaches reflect the variety of components involved in the evolution of regulation. But we can discern four dominant modes of change in metazoan regulatory genomes: First, the introduction of new regulatory tools, such as distal enhancers in Metazoa, expands the regulatory and developmental capacity of a clade. Second, diversification of existing regulatory components, which has been particularly evident in the expansion of transcription factor families. Third, regulatory networks are repatterned, as with the co-option of subcircuits within gene regulatory networks (GRNs). Finally, there are also many fine-scale or microevolutionary changes in regulatory control, many of which are achieved through alterations of enhancer specificity and changes in promoters. Modifications in GRNs may have a variety of different effects: they may turn off

Evolutionary dynamics of gene regulation

409

expression, shift expression either temporally or spatially, or changes in expression levels. The recent increase in inter- and intra-specific studies of regulatory evolution have revealed many cases in which relatively stable morphology masks considerable tumult at the regulatory level. This suggests that considerable regulatory evolution may be adaptively neutral at the level of changes in enhancer and promoter specificity. Unfortunately the paucity of detailed, comparative studies of GRNs still constrains our understanding of the relative frequency of different types of regulatory change, and which of these most often translate into phenotypic evolution. This contribution focuses on the first three modes of regulatory change and the implications each has for different types of morphological evolution, drawing largely on examples from animals, although the conceptual framework presented here seems equally applicable to regulatory evolution in plants. The central question I address here is whether there is a mechanistic difference in the types of GRN evolution associated with larger-scale morphological evolution (macroevolution) versus small-scale adaptive or microevolutionary changes. Whether macroevolutionary patterns are driven by microevolutionary processes or by a distinct suite of macroevolutionary processes has been debated since the work of Goldschmidt and Simpson during the Modern Synthesis in the 1950s. As evolutionary novelties have received considerable attention particularly with respect to origins of new body plans, I consider this aspect of macroevolution separately before turning to the issue of a general framework for GRN evolution.

2. The origin of metazoan regulatory novelties The closes living relatives of Metazoa are, successively, choanoflagellates, filastreans and ichthysporeans (collectively these four groups comprise the Holozoa). Comparative studies of representatives of these groups have provided insight into the early evolution of the metazoan regulatory genome (Brunet & King, 2017; Richter, Fozouni, Eisen, & King, 2018; Richter & King, 2013; Sebe-Pedros et al., 2016; SebePedros, Degnan, & Ruiz-Trillo, 2017; Simakov & Kawashima, 2017), as studies of sponges, cnidarians and other taxa have illuminated the expansion of the metazoan regulatory genome. Gene loss has been common, however, particularly in clades that have experienced morphologic simplification. Thus using the genome of a single species as an exemplar for a large clade can be misleading. Fig. 1 details the progressive expansion of the metazoan regulatory genome in a phylogenetic context (see Erwin, 2020, for more detail).

Deuterostomia

Ecdysozoa

Lophotrochozoa

Cnidaria Placozoa Porifera

Protostomia Bilateria

Filastrea Ichthyosporea

Eumetazoa

Growth of GRNs through intercalation of spatial and temporal regulators; distal enhancers

Metazoa Proximal regulation via TF combinatorics

Holozoa

Fig. 1 The history of acquisition of important parts of the holozoan and metazoan regulatory genomes, plotted on a phylogenetic tree. Red bars show the independent co-option of patterning elements in deuterostomes, ecdysozoans and lophotrochozoans associated with the generation of more hierarchically structured GRNs and other regulatory novelties as animal body sizes increased during the EdiacaranCambrian metazoan radiation (
550–520 million years ago). Based on references cited in text and in Erwin, D. H. (2020). Origin of animal bodyplans: A view from the regulatory genome. Development 147, dev182899.

Evolutionary dynamics of gene regulation

411

Three noteworthy points emerge from such a compilation. First, many putatively “bilaterian” or “metazoan” elements have now been identified among the cousins of Metazoa and a plausible case has been advanced that they originally evolved to control temporal patterning in groups with complex life cycles (Arenas-Mena, 2017; Sebe-Pedros et al., 2017; Sogabe et al., 2019). Both complex life cycles and multicellular species are found within each of the major clades of holozoans, although the form of multicellularity differs. With the advent of animals, regulators of temporal differentiation were repurposed for spatial control. Second, during the early diversification of animals the evolution of the metazoan genome included both the introduction of novel regulatory elements, including distal enhancers, new developmental transcription factor families, a new type of promoter, and other elements, as well as continuing expansion of the developmental toolkit. Distal enhancers do not appear to have become widely deployed components of the regulatory genome until the diversification of Bilateria (Sebe-Pedros et al., 2018). (Distal enhancers have recently been reported from plants as well (Lu et al., 2019)). As genome size expanded, the complexities of regulatory control evidently increased as chromatin architecture played an increasingly important role in regulating transcription in bilaterians, particularly through the appearance of CTCF sequences as boundary elements for transcriptionally active domains (TADs) (Gaiti, Calcino, Tanurdzic, & Degnan, 2017). Finally, the phylogenetically early appearance of many deeply conserved regulatory elements reveals that the function of these elements changed over evolutionary time. In contrast to the early days of “evo-devo” it is now clear that even if genes are deeply homologous and serve similar functions in diverse living clades, this is not unambiguous evidence for the functions of these genes half a billion years ago. This final point bears elaboration for it provides critical insight into the nature of regulatory evolution in animals. Extensive co-option of GRN subcircuits initially involved in cell type specification and simple embryonic patterning placed these subcircuits at the top of extensive regulatory hierarchies responsible for regional patterning (Erwin, 2020; Erwin & Davidson, 2009). These considerations inform the following model (Erwin, 2020): temporal and spatial regulation is a common feature of Holozoa, and was co-opted for increased spatial differentiation within animals. The dominant form of regulatory control in basal metazoans (sponges, cnidarians and placozoa) is proximal via combinations of transcription factors at enhancers proximal to the coding sequence (Sebe-Pedros et al., 2018) (Fig. 1).

412

Douglas H. Erwin

Most GRNs in the earliest animals were likely relatively flat, at least in comparison to those developing among bilaterians. The initial diversification of bilaterians involved the generation of more hierarchically structured GRNs through intercalation of new spatial and temporal regulators into the initial, flatter GRNs (Davidson & Erwin, 2006; Erwin & Davidson, 2009; Peter & Davidson, 2015). Extensive co-option of subcircuits occurred, distal enhancers became more widespread, and new forms of control over chromatin were introduced. Finally, co-option leads to independent construction of many developmental processes, including segmentation, a tri-partite brain, appendages, regionalized gut and sensory systems. These processes generated new cell types, new developmental patterns and thus new phenotypes. Although the evolution of new cell types has been described as bifurcating divergence (Arendt, 2008) other processes may be involved as well (Arendt et al., 2016).

3. The expansion and exploitation of combinatoric complexity The complexities of regulatory evolution were first revealed by the deep homologies of many transcription factor genes, including Hox and Pax genes. These are parts of large families of genes whose expansion coincided with the diversification of animals through extensive gene duplication (Degnan, Vervoort, Larroux, & Richards, 2009; Larroux et al., 2008). This in turn has enabled an increase in the complexity of transcription factors combinatorics, as evidenced by proportionally larger increase in the number of TFs as genome size expands (Babu, Luscombe, Aravind, Gerstein, & Teichmann, 2004; Schmitz, Zimmer, & Bornberg-Bauer, 2016). In contrast to the regulatory novelties described in the previous section, these trends have occurred independently in many different metazoan clades. The details of TF combinatorics are beyond the scope of this paper, but the nature of the cooperativity in TF-TF interactions may influence the scope of the combinatorics and how they evolve (Reiter, Wienerroither, & Stark, 2017). At the end of this section I also note evolutionary trends such as multiple transcription start sites (TSS) and alternative splicing which also increase combinatoric complexity. A study of the five largest metazoan TF families (bHLH, bZip, Homeobox, Nuclear Receptor and C2H2 ZF) and the p53 family documented the effects of both single gene and whole genome duplications as well as domain rearrangements (Schmitz et al., 2016). The results illustrated

Evolutionary dynamics of gene regulation

413

expansions of TF families, largely due to single gene duplication. Bursts of accelerated expansion were also identified, some of which were associated with whole genome duplication. But the most important driver of functional diversification of these TF families was rearrangement of modular domains to generate new TF subfamilies. In turn, the new functions are often associated with rounds of expansion of TFs and adaptive fine tuning. The importance of domain rearrangements for functional diversification further emphasizes the importance of modular structure of TFs and their interactions. Significantly, this study also emphasized the contrasts between punctuated evolution of TF families and more continuous adaptive changes in enzyme function. To the extent that evolutionary theory and population genetics have emphasized the latter they have provided a biased view of evolutionary dynamics. TADs have continued to evolve among Bilateria. Multiple developmental functions in Hox genes depends in some instances also involved the evolutionary acquisition of multiple TADs separating these functions (Darbellay et al., 2019). Thus while the hox cluster in Amphioxus has a single TAD with most regulatory inputs 30 to the cluster, jawed vertebrates have a more complex architecture with one TAD upstream and another downstream (Acemel et al., 2016). Components of the anterior TAD interact with the anterior genes in the hox cluster while the posterior genes interact with the posterior TAD. The more complex architecture of vertebrate TADs relative to both Amphioxus and protostomes has extended the involvement of HOX genes to additional patterning functions, such as limbs. The expansion of TF families may be the most prominent trend in regulatory genome evolution, but a number of other have been identified as well. These trends include increased macro- and micro-synteny (Simakov & Kawashima, 2017; Zimmermann, Robert, Technau, & Simakov, 2019), such as the formation of the pharyngeal gene cluster in deuterostomes (Simakov et al., 2015) as well as expansion of microRNAs (Wheeler et al., 2009). There have also been increases in alternative splicing and the generation of multiple TSS (Gaiti et al., 2017). The generation of large TF families has evidently generated challenges in coordination that have been resolved by increasing clustering in many bilaterian lineages, suggesting that increased synteny, for example is a driven rather than a passive trend. An interesting challenge for future work is how many of these trends represent adaptive responses to the demands of increasing coordination of regulatory response versus neutral accumulation (e.g., Lynch, 2007).

414

Douglas H. Erwin

4. Evolutionary repatterning of gene regulatory networks A variety of different mechanisms change GRN structure and function: recruitment of new regulatory sequences via insertion of transposons, co-option of existing GRN subcircuits for expression in new spatial or temporal domains, gain and loss of regulatory inputs, remodeling of internal GRN connections, including changes in binding affinity, switching of promoters, and other changes, and the de novo origination of new regulatory genes (Fig. 2) (Carroll, 2008; Rebeiz, Jikomes, Kassner, & Carroll, 2011; Rebeiz et al., 2015; Wray et al., 2003; Chipman’s paper in this volume discusses the evolution of the gap gene network). If regulatory capacity of a clade at any point in time is largely controlled by the factors described in the previous two sections, this section focuses on those processes used by evolution for the repatterning and expansion of GRNs once that capacity is established.

Fig. 2 Mechanisms of change to GRN structure include insertion of transposable elements; co-option of regulatory information; repatterning of GRNs, here with the conversion of a coherent feed-forward subcircuit into an incoherent feed-forward subcircuit by the addition of a spatially restricted activator which causes repression of Gene C in contexts where the activator is expressed; and de novo origination of new regulatory sites, here via conversion of a formerly low-affinity binding site. Presumably some such low-affinity binding sites arise from sequences which fortuitously bind appropriate TF’s, with selection converting these through low-affinity binding sites to more robust binding sites.

Evolutionary dynamics of gene regulation

415

Long before the ubiquity of transposons was established, Barbara McClintock famously winkled out the importance of the process in maize (corn in the US) and such elements have since been shown to be nested within earlier transposable elements (TEs) like Russian dolls (Fedoroff, 2012). In their early work on regulatory evolution (Britten & Davidson, 1971) also posited a role in cis-regulatory evolution for what is now known to be transposon insertion. Eukaryotic cells have been fighting off the detrimental effects of TEs for over a billion years, and consequently TEs come equipped with promoter and enhancer sequences tuned for integration into the host genome. TEs have been hypothesized to be potent sources of transcription factor binding sites, thus provide important sources of change in GRNs, as the TE regulatory sequences are co-opted by adjacent CRMs. Because drift and selection obscures signatures of TEs with time, the extent of TE modification of GRNs remains uncertain, but there is considerable evidence for their activity (Chuong, Elde, & Feschotte, 2017; Sundaram & Wang, 2018). One study implicated TEs in 20% of TF binding sites (Sundaram et al., 2014) and endogenous retroviruses have been implicated in regulation of the mammalian immune system (Chuong, Elde, & Feschotte, 2016), the earliest stages of development (Chuong et al., 2017) and pregnancy (Lynch et al., 2015). A recent study of liver enhancers and promoters across six primate species showed that the majority of newly evolved and differentially expressed cis-regulatory elements had signatures of TEs (Trizzino et al., 2017). In contrast, in many cell types highly conserved promoters and enhancers showed little evidence of TE activity. In some ways the widespread involvement of TEs in areas requiring rapid adaptation to changing circumstances, such as immune response, should not be surprising. Co-option of singe genes or GRN modules is a widespread mode of change as it supplies already functional GRN subcircuits into new spatial or temporal positions (Davidson & Erwin, 2006; McQueen & Rebeiz, 2020; Monteiro & Gupta, 2016; Peter & Davidson, 2015). McQueen & Rebeiz, 2020 described different types of co-option, the implications for pleiotropy, and their potential evolutionary impact. Limb GRNs appear to have been co-opted from anterior-posterior patterning systems in early bilaterians (Lemons, Fritzenwanker, Gerhart, Lowe, & McGinnis, 2010) with the insect appendage formation genes distalless and homothorax co-opted for generating the novel adult horns in beetles (Moczek, 2009; Moczek & Rose, 2009). Paired fins in fish evolved through co-option from the median fin GRN (Freitas, Zhang, & Cohn, 2006), while the novel

416

Douglas H. Erwin

anterior component of the lateral fins of skates and rays involves co-option of parts of the GRN controlling formation of the Apical Ectodermal Ridge (Nakamura et al., 2015). The novel embryonic skeleton in modern sea urchins has been redeployed from the adult skeletogenic machinery (Gao & Davidson, 2008). The generation of horns on the head shield of adult beetles, particularly Onthophagus, has become a well-studied example of evolutionary novelty (Moczek, 2008; Zattara, Busey, Linz, Tomoyasu, & Moczek, 2016). Horns are used in head-to-head contact among males, and arise from the head shield through co-option of Otd1/2 activity from the embryo into the post-embryonic stages. These abundant examples of co-option associated with macroevolutionary transitions sometimes obscure the contribution of co-option to more restricted GRN changes. One of the best studied example comes from the origin of the posterior lobe in the Drosophila melanogaster clade (Glassford et al., 2015). The GRN controlling this male copulatory structure was co-opted from the posterior spiracle (a respiratory structure) in embryonic flies. Co-option does not necessarily involve an entire GRN, as in many of the previous examples. A clade of beetles, including Tribolium, contains a defensive epidermal structure on the embryo called a “gin-trap.” In this case a single gene was co-opted from a wing development GRN, but its function is essential to generating the structure (Hu et al., 2018) (Fig. 3). Limited co-option is also characteristic of signaling circuits known as “plug-ins” such as Delta-Notch, Hh and FGF, which are repeatedly co-opted as developmental switches (Andrikou & Arnone, 2015; Davidson & Erwin, 2006). Larger GRN subcircuits can act as “plug-ins” as well, co-opted as a single unit because of their network properties rather than because they control the same function in different areas. Martik and McClay identified a subcircuit of Pax6-Six3-Six1/2-Eya-Dach1 responsible for migration of micromeres in the developing sea urchin embryo which is identical in structure to the retinal eye determination subcircuit in Drosophila (Martik & McClay, 2015). Co-option has interesting consequences that distinguish it from other modes of GRN change. First, co-option allows evolution to redeploy already functional components in new contexts. Second, co-option will often generate pleiotropic enhancers—enhancers that operate in two or more circumstances. Pleiotropy is of great interest to evolutionary biologists because it is generally thought to constrain subsequent change because of the need to satisfy all existing functions; it thus facilitates modular structures in GRNs (Carroll, 2008; Davidson, 2006; Melo, Porto, Cheverud, & Marroig, 2016). Third, co-option necessarily generates more hierarchically structured GRNs (Sabaris, Laiker, Preger-Ben Noon, & Frankel, 2019).

Evolutionary dynamics of gene regulation

417

Fig. 3 Gene co-option plays a significant role in the repatterning of developmental networks. In this example genes active in two different developmental contexts are shown (A and B). A module of regulatory sites from Gene B has been co-opted by Gene A (purple square and blue triangle and distinguished by the double hash marks in the yellow gene CRM) leading the downstream circuit (here a coherent feed-forward circuit controlling expression of three differentiation genes) to be expressed in contexts where Gene A is activated in addition to the ancestral expression controlled by expression of Gene B. Upstream regulatory sites are shown by colored objects.

GRN remodeling also occurs through enhancer shuffling both within and between species. Wing patterning in Amazonian Heliconius provides an example where reshuffling of two CRMs controlling optix rapidly generated a stunning diversity of wing patterns (Wallbank et al., 2016). This example beautifully illustrates the impact of combinatorics, for the various combinations of these two elements create a greater variety of wing patterns than would be possible with a single CRM. Even more striking, however, is the evidence for introgression of each element between two species of Heliconius. Repatterning of GRNS can also include loss of function changes. A well-documented example of this mechanism comes from the loss of limbs in snakes through progressive functional degradation of the limb enhancer for Sonic hedgehog (Shh) (Kvon et al., 2016). By comparing enhancer activity in other vertebrates with basal snakes, which have vestigial limbs, with those of advanced snakes, which lack them, the study demonstrated the decreased sequence conservation for the Zone of Polarizing Activity regulatory sequence and progressive loss of enhancer activity.

418

Douglas H. Erwin

Finally, promoters, enhancers and other components of the regulatory genome may arise de novo. Rebeiz et al. (2011) included de novo origin of GRNs essentially as a null model, but many regulatory sequences are relatively short and transcription factors often bind to low-affinity sites. This provides a basis for both selection and drift to generate new regulatory sequences from non-functional sequences. Some data on rapid turnover of these sites, as well as the numbers of “orphan genes” in clades (those with no sequence similarity to closely related taxa), suggests a continuing flux of de novo originations. The extent to which these merely provide “churn” to regulatory interactions (see discussion) versus providing a substantial basis for evolutionary change remains unclear. One of the central findings of studies of regulatory evolution over the past several decades has been the deep conservation of many metazoan regulatory components and even interaction networks across hundreds of millions of years. Evidence from transcription factors provide the first insight into the extensive conservation of the metazoan regulatory genome. Unlike TFs, however, enhancers evolve much more rapidly with little evidence for sequence conservation. Surprisingly, however, sponge enhancers can drive cell types specific vertebrate enhancers, showing that some TF-enhancer interactions have been conserved since the origin of animals (Wong et al., 2019). But such deep homology raises critical issues about how regulatory interactions are restructured, how variability arises upon which selection can act and what permits both evolvability of regulatory interactions yet preserves the stability of body plans. These conundrums have led to tensions between those who believe that evolution of the regulatory genome operates in much the same way as other evolutionary processes, with substantial intra-specific variability providing a foundation for drift and selection, while others downplay the importance of population-level variability in cis-regulatory modules in favor of a mutation-driven process in which regulatory changes play a dominant role; a corollary of the later view is that many types of regulatory change may be relatively rare. One means of addressing this controversy is to examine several factors which control the evolutionary impact of GRN changes: the amount of variability present within a species, or between closely related species; epistasis, or interactions between different genes; and pleiotropy, or multiple effects from a single gene. Examining inter- and intra-specific variability in GRN structure remains challenging but some recent work suggests that “cryptic” variability has been under-appreciated. The endoderm GRN for C. elegans includes two key inputs: SKN-1/Nrf2 and MOM-2/Wnt, but natural

Evolutionary dynamics of gene regulation

419

populations exhibit variability in the their requirements: some require both inputs, while other populations exhibit a tradeoff where one input can substitute for low levels of the other (Torres Cleuren et al., 2019). Some of this variability may be effectively neutral, but nonetheless providing a critical foundation for subsequent evolutionary change. Epistasis can mask the effects of changes to GRNs. For example, Abd-B is required for abdominal pigmentation in Drosophila yakuba but changes to Abd-B expression in the sister species, D. santomea do not affect pigmentation. Evidently the epistatic effects of four genes downstream of Abd-B mask the effects of changes in expression (Liu et al., 2019). As with other recent work, this reveals GRN evolution via incremental network adjustments. This shifts the focus from specific genes to how GRN structures have evolved (Phillips, 2008). Pleiotropic effects are generally seen as an evolutionary constraint, since changes must satisfy all the uses of the gene, and thus pleiotropy can significantly influence changes across GRNs (Pavlicev & Wagner, 2012; Wagner & Zhang, 2011) and this has influenced discussions about the modularity of GRNs as a means of reducing pleiotropic interactions. But genome wide studies have now shown that pleiotropic enhancers are widespread in many metazoan genomes (Sabaris et al., 2019).

5. Micro- and macroevolutionary changes in gene regulatory networks Paleontologists have long proposed that macroevolutionary dynamics are decoupled from microevolutionary dynamics within species ( Jablonski, 2017a, 2017b) largely via differential sorting of species and clades and events such as mass extinctions. But long-standing interest in whether macroevolutionary patterns were underpinned by differences in the sources of variation (Erwin, 2017a; Godfrey-Smith, 2009; Gould, 2002; Jablonski, 2017a) has been strengthened by comparative studies of developmental mechanisms. Evolutionary developmental biologists have differed sharply over this issue. Carroll and his colleagues, for example, have argued strongly that cis-regulatory evolution is the dominant mode of evolutionary change (unlike many population geneticists who have argued for primacy of changes in coding regions), but that macroevolutionary patterns are the consequence of many small-scale changes in cis-regulatory elements (Carroll, 2008; Prud’homme, Gompel, & Carroll, 2007; Wray et al., 2003). Other developmental biologists have associated macroevolutionary dynamics with distinctive patterns of regulatory change. For example, Davidson

420

Douglas H. Erwin

and I suggested that microevolutionary changes may be more common in the periphery of a GRN while macroevolutionary changes may be largely confined to the core of a GRN or otherwise involve distinct mechanisms (Davidson, 2006; Davidson & Erwin, 2006). Peter and Davidson (2015) provide extensive discussion of these issues. Wagner has associated morphological novelty, a component of macroevolutionary patterns, with distinctive modes of changes in GRNs (Wagner, 2014). In 2017 I distinguished these two approaches as “developmental-push” with regulatory and developmental changes pushing macroevolutionary dynamics, versus “environmental-pull,” where ecological opportunities in the environment generate the opportunities for success of developmental or regulatory changes (Erwin, 2017a). In other words, are there characteristic differences between the processes that drive adaptive microevolutionary changes and those involved in larger-scale patterns such as the origin of new morphologies, or with new clades? There are abundant studies of the role of GRNs and regulatory evolution in microevolutionary patterns. Some of the more prominent of these involve rapid changes near the periphery of GRNs in wing pigmentation patterns in flies and in butterflies. In flies the Yellow locus controls wing pigmentation through a landscape of enhancers and promoters that combine in a variety of different ways to generate characteristic patterns. (Gompel, Prud’homme, Wittkopp, Kassner, & Carroll, 2005; Prud’homme et al., 2007; Wittkopp, Vaccaro, & Carroll, 2002). Although the specific mechanisms is different, the transcription factor optix is involved in generating the array of red pigmentation patterns in Heliconius butterflies, with striking patterns of mimicry, again in the distal portions of the regulatory network (Martin et al., 2014). In contrast, macroevolutionary changes in the organization and patterning of the developing embryo are more commonly associated with more central components of GRNs, as seen in the examples of co-option discussed previously. Convergence is a ubiquitous evolutionary phenomenon, producing similar morphological outcomes in sometimes distantly related taxa, and can reveal interesting insights into macroevolutionary processes. The loss of digits has occurred in many different tetrapod clades, and can involve a variety of developmental mechanisms (Cooper et al., 2014), but more commonly studies have shown repeated involvement of related regulatory genes to achieve similar morphological outcomes (Stern, 2013). For example, pigmentation in Drosophila wing, thorax and abdomen involves the same suite of genes (which also contribute to inter-specific divergences in pigmentation patterns) (Massey & Wittkopp, 2016). Loss of flight among ratites

Evolutionary dynamics of gene regulation

421

such as ostriches, kiwis, emus and the extinct moa and elephant birds happened at least three to six times (possibly more), and each case seems to involve a similar suite of regulatory genes involved with limb morphogenesis (Sackton et al., 2019). As Martin and Orgogozo emphasized in their compilation of hotspots of convergence, these cases illustrate that there are often “pathways of least resistance” which are repeatedly used by evolution to achieve similar ends (Martin & Orgogozo, 2013). This suggests a degree of predictability in patterns of change in GRN structure. The modes discussed here appear to differ in their frequency and in their evolutionary impact. Novelties in the regulatory genome are the least frequent, but may have the greatest impact as they generate new developmental opportunities. Evolutionary biologists often think in terms of the spaces in which evolution operates, and in this sense these novelties construct or expand the developmental opportunity space available for morphological change (Erwin, 2017b). Other modes of regulatory evolution, including expansion of TF families, may enlarge these opportunity spaces, while the most common regulatory changes drive evolutionary dynamics within existing developmental spaces.

6. Do novel morphologies involve mechanistically distinct pathways? Evolution happens at many levels, from genome sequences through cell types and physiology to morphology and behavior. These changes are often independent in the sense that changes in the genome do not necessarily change the phenotype. A critical question for comparative developmental biology has been whether the mechanisms underlying the generation of phenotypic, particularly morphologic, novelty are distinct from those associated with most adaptive evolution. The terms novelty and innovation are often used interchangeably, but within the evolutionary developmental biology community the term novelty is now generally defined as the origin of a newly individuated, homologous character, such as the origin of neural crest, feathers, or the vertebrate limb (Erwin, 2015; Moczek, 2008; Muller & Wagner, 1991; Wagner, 2014). New characters are recognized based on some degree of morphological discontinuity, and thus novelties are not distinguished by the mechanisms of formation, nor how generative or consequential the clade containing the novelty may become with time (Erwin, 2020). By introducing new characters, novelties are thus at least conceptually distinct from most adaptive evolutionary change.

422

Douglas H. Erwin

In understanding novelty, there is a critical distinction between the origin of a new character and subsequent changes in the nature, or state, of that character (Wagner, 2014). Thus vertebrate limbs arose once, but have many different states in frogs, turtles, birds, hippos and jaguars; feathers have variety of states both in different places on a bird, and in different clades, but all trace back to a single origin; and the variety of types of flowers almost defies description. The novelties associated with the origins of Metazoa and major clades suggests that at least some novelties may represent a distinct mode of evolutionary change closely linked to restructuring of the regulatory genome (Carroll, 2005, 2008; Davidson & Erwin, 2006; Wagner, 2014). If novelties represent a distinct evolutionary process from microevolutionary adaptive change, if they are more, or different than the summation of other changes over time (the traditional view) then advocates for novelty must identify either a unique mechanism of change (the “systemic mutations” of Goldschmidt’s hopeful monsters) or continuity at the level of development which is translated into discontinuous morphological change. Earlier sections of this paper have identified the disjunct origin of some regulatory or developmental novelties, including the origin of distal enhancers and TADs, and mechanisms such as transposition and co-option associated with morphological novelties. But the latter two mechanisms can generate a wide range of morphological effects, depending on where they act within a regulatory network (Erwin & Davidson, 2009; Peter & Davidson, 2015) and are not uniquely associated with the generation of novelty. The possibility of decoupling between development and morphology underlay the proposals that some morphological novelties were associated with the formation of highly conserved recursively wired GRN subcircuits. In 2006 Davidson suggested that such subcircuits, or termed “kernels” were responsible for establishing regional developmental patterning in endoderm and heart formation (Davidson, 2006; Davidson & Erwin, 2006). Wagner identified similar structures the same year underpinning cell-type specification and termed them character homology interaction networks, or CHiNs (Wagner, 2007). (The two concepts are topologically equivalent, but differ in the developmental components to which they have been applied.) The concept of kernels and CHiNs have the virtue of providing a mechanistic basis for the stability of morphological novelties over long spans of evolutionary time without invoking stabilizing selection. In both cases the recursive wiring serves to “lock-down” the morphology generated. A similar endoderm kernel has been identified in zebrafish

Evolutionary dynamics of gene regulation

423

(Tseng, Jang, Huang, & Yuh, 2011) but thus far kernels do not appear to be widespread features of GRNs, although ChiNs may be more common in cell-type specification GRNs. Wagner tied his view of evolutionary novelties to CHiNs (Wagner, 2014), but it might be more appropriate to view this as a hypothesis awaiting experimental confirmation. The origins of morphological novelty remain, and likely will remain, a topic of intense interest among developmental and evolutionary biologists. Some developmental novelties meet the criteria of newly individuated characters but their underlying, mechanistic basis is often unclear. Gene duplication has been a potent force driving the expansion of TF families and underlies some macroevolutionary patterns such as the diversification of vertebrate and arthropod limbs. There is little evidence, however, that TF expansion directly drives morphological changes. Co-option and transposition are both discontinuous mechanisms relative to single-base pair mutations, and certainly many co-options of regulatory genes or GRNs underlie macroevolutionary transformations. These are the cases that may come the closest to apparent macroevolutionary “discontinuities,” but where they have been studied in detail, the incorporation of co-options generally requires other developmental and regulatory accommodation.

7. Towards a general theory of GRN evolution Studies of regulatory genome evolution across metazoans has revealed similar patterns arising independently in different clades. These include increases in gene clustering through macro- and micro-synteny, increased intron density, the expansion of most families of metazoan transcription factors, which in turn has enabled the increased complexity of GRNs via growth in the number of promoters and transcription start sites. Since the growth in the size of genomes has been largely decoupled from increased developmental and morphological complexity, the later has been achieved by the greater complexity of regulatory interactions, with both genes and GRN subcircuits co-opted for new functions. Bilaterians exhibit a great expansion in both the roles of regulatory RNAs, and in structuring of chromatin through TADs. It is a plausible argument that these have been required to structure regulatory interactions. These similarities suggest that, at least to some extent, it may be possible to articulate a general theory of GRN evolution. In some sense GRN dynamics are a special case of network dynamics in general, with a specific set of rules. Application of network dynamics has

424

Douglas H. Erwin

already informed much of our understanding of GRNs and some general principles about the evolution of the regulatory genome are already emerging. But one of them is the importance of contingency and history in structuring networks, and in constraining the available opportunities from an existing configuration (Sorrells, Booth, Tuch, & Johnson, 2015) as is generally the case in evolution (Erwin, 2016; Ramsey & Pence, 2016). Thus the expansion of TFs is not surprising, but different families have expanded in plants versus animals, and even within animals there have been different patterns of expansion across different clades. Understanding the various forces underlying deep conservation of many genes, regulatory interactions and phenotypic structures (as body plans) contrasts with evidence for rapid turnover in enhancers. Drift in network components while preserving developmental and phenotypic outcomes has been documented as developmental systems drift: (True & Haag, 2001). Evidence for drift indicates that GRN structures are relatively resilient to some changes. Comparison of experimentally validated GRNs against simulations studies has already yielded insights into this problem. Sorrells and Johnson (2015) compared the network structure of transcription factors in biofilms of the yeast Candida albicans with embryonic stem cell networks in mice (Mus musculus). They concluded that the networks were not optimized by natural selection but preserved the history of their evolution, with some types of network structures being more likely to evolve than others (what the authors term “high-probability events”). The number of genes, overlapping connections and functional redundancy all exceed what would be expected from an “optimized” network. A purely qualitative overview of some of the networks discussed in this paper suggests that this may be generally true of bilaterian dGRNs. As with a study of protein-protein interactions across the tree of life (Zitnik, Sosic, Feldman, & Leskovec, 2019), network resilience may have increased through time, although more intensive studies will be required to test this hypothesis. These issues complicate a search for a general theory of GRN evolution, but they do not eliminate it, but they do suggest that mathematical structure of population genetics may not (or at least not yet) provide a sufficient theoretical framework for understanding the evolution of the regulatory genome.

8. Concluding remarks The regulatory and thus the developmental capacity of the genome has expanded significantly since the metazoan lineage split from other

Evolutionary dynamics of gene regulation

425

holozoans over half a billion years ago. Important attributes were inherited by this ancestral lineage, but the subsequent history of metazoan regulatory evolution has encompassed critical novelties, including distal enhancers, new types of promoters and elaboration of chromatin control systems such as TADs as well as independent trends in many lineages towards enhanced combinatoric complexity by diversification of transcription factor families. Regulatory evolution, however, has been dominated by mechanisms that alter GRN structure, principally co-options, transposon insertion, repatterning of promoter and enhancer connections and the de novo origination of new regulatory sites. The studies of the evolution of GRNs and of the regulatory genome raise a number of evolutionary questions that will be addressed in coming years as experimental studies reveal the details of developmental GRNs. As more GRNs are explicated our understanding of the relative importance of different types of regulatory change will be placed on a firmer foundation. What are the relative frequency of co-option verses other mechanisms of GRN repatterning, and is co-option, for example, more likely to be involved in macroevolutionary patterns than is enhancer and promoter switching? Are recursively wired subcircuits critical for regional patterning or cell-type formation (kernels: Davidson & Erwin, 2006; CHiNs: Wagner, 2014)? Placing information on regulatory evolution in a phylogenetic context will increasingly allow us to understand whether the importance of different types of regulatory rewiring has changed over time, or through the history of a clade. The possibilities for fruitful interactions between experimentalists and theoreticians to explore these issues will grow rapidly as new GRNs are elucidated. Such questions include: How many different network topologies would yield the same developmental and phenotypic outcome? How has evolution structured the topology of regulatory networks, and how does that topology influence future evolutionary change? In other words, have these networks become optimized over time by natural selection for robustness or other attributes? Does regulatory evolution evolve via paths of least resistance, or via “tinkering”? Is neutral evolution truly as widespread as suggested by recent studies, and why do the structures of GRNs appear to be so far from optimal? Each of these patterns suggest that selection may be less effective in crafting the regulatory genome (even in large populations) than for other parts of the genome. Conserved regulatory circuits and features such as kernels/CHiNs suggest that the need to preserve functions could be more powerful that selection for optimal circuit design, and some

426

Douglas H. Erwin

features may be “frozen accidents.” As always, one must be cautious about generalization. The number of large-scale, detailed and experimentally verified metazoan GRN remains small, and other information on the regulatory genome can be biased by inadequate taxon sampling (which obscures gene gain and loss, for example), and other factors. New complexities of the regulatory genome continue to be discovered regularly, particular as new techniques such as single-celled transcriptomics are applied more widely and to new clades. Just as the recognition of deep homologies among TFs startled developmental and evolutionary biologists in the 1990s, more recent work has documented the remarkable churn among enhancers and promoters in closely related species and even between within populations in the same species. As this trend expands in coming years, the richness of population-level information on the regulatory genome will continue to bring population genetics and developmental biology closer together, to the benefit of both and of evolutionary biology as a whole.

Acknowledgments I thank James Briscoe, Isabelle Peter, Ellen Rothenberg and the faculty and students of recent sessions of the Gene Regulatory Network course at the MBL for discussions of the issues raised in this manuscript. I imagine that Eric Davidson would have hated much of this, but it would not have happened without our long collaboration.

References Acemel, R. D., Tena, J. J., Irastorza-Azcarate, I., Marletaz, F., Gomez-Marin, C., de la Calle-Mustienes, E., et al. (2016). A single three-dimensional chromatin compartment in amphioxus indicates a stepwise evolution of vertebrate Hox bimodal regulation. Nature Genetics, 48, 336–341. Andrikou, C., & Arnone, M. I. (2015). Too many ways to make a muscle: Evolution of GRNs governing myogenesis. Zoologischer Anzeiger, 256, 2–13. Arenas-Mena, C. (2017). The origins of developmental gene regulation. Evolution & Development, 19, 96–107. Arendt, D. (2008). The evolution of cell types in animals: Emerging principles from molecular studies. Nature Reviews Genetics, 8, 868–882. Arendt, D., Musser, J. M., Baker, C. V. H., Bergman, A., Cepko, C., Erwin, D. H., et al. (2016). The origin and evolution of cell types. Nature Reviews Genetics, 17, 744–757. Babu, M. M., Luscombe, N. M., Aravind, L., Gerstein, M., & Teichmann, S. A. (2004). Structure and evolution of transcriptional regulatory networks. Current Opinion Structural Biology, 14, 283–291. Britten, R. J., & Davidson, E. H. (1971). Repetitive and non-repetitive DNA sequences and speculation on the origins of evolutionary novelty. Quarterly Review of Biology, 46, 111–138. Brunet, T., & King, N. (2017). The origin of animal multicellularity and cell differentiation. Developmental Cell, 43, 124–140.

Evolutionary dynamics of gene regulation

427

Carroll, S. B. (2005). Evolution at two levels: On genes and form. PLoS Biology, 3, 1159–1166. Carroll, S. B. (2008). Evo-devo and an expanding evolutionary synthesis: A genetic theory of morphological evolution. Cell, 134, 25–36. Chuong, E. B., Elde, N. C., & Feschotte, C. (2016). Regulatory evolution of innate immunity through co-option of endogenous retroviruses. Science, 351, 1083–1087. Chuong, E. B., Elde, N. C., & Feschotte, C. (2017). Regulatory activities of transposable elements: From conflicts to benefits. Nature Reviews Genetics, 18, 71–86. Cooper, K. L., Sears, K. E., Uygur, A., Maier, J., Baczkowski, K. S., Brosnahan, M., et al. (2014). Patterning and post-patterning modes of evolutionary digit loss in mammals. Nature, 511, 41–45. Darbellay, F., Bochaton, C., Lopez-Delisle, L., Mascrez, B., Tschopp, P., Delpretti, S., et al. (2019). The constrained architecture of mammalian Hox gene clusters. Proceedings of the National Academy of Sciences of the United States of America, 116, 13424–13433. https://doi. org/10.1073/pnas.1904602116. Davidson, E. H. (2006). The regulatory genome. San Diego: Academic Press. Davidson, E. H., & Erwin, D. H. (2006). Gene regulatory networks and the evolution of animal body plans. Science, 311, 796–800. Davidson, E. H., & Erwin, D. H. (2010). Evolutionary innovation and stability in animal gene networks. Journal of Experimental Zoology. Part B, Molecular and Developmental Evolution, 314B, 182–186. Degnan, B. M., Vervoort, M., Larroux, C., & Richards, G. S. (2009). Early evolution of metazoan transcription factors. Current Opinion in Genetics & Development, 19, 591–599. Erwin, D. H. (2015). Novelty and innovation in the history of life. Current Biology, 25, R930–R940. Erwin, D. H. (2016). Wonderful Life revisited: Chance and contingency in the Ediacaran-Cambrian radiation. In G. Ramsay & C. H. Pence (Eds.), Chance in evolution (pp. 279–298). Chicago, IL: University of Chicago Press. Erwin, D. H. (2017a). Developmental push or ecological pull? The causes of macroevolutionary dynamics. History and Philosophy of Life Science, 39, 36. Erwin, D. H. (2017b). The topology of evolutionary novelty and innovation in macroevolution. Philosophical Transactions of the Royal Society of London, Series B, 372, 20160422. Erwin, D. H. (2020). Origin of animal bodyplans: A view from the regulatory genome. Development, 147, dev182899, in press. Erwin, D. H., & Davidson, E. H. (2009). The evolution of hierarchical gene regulatory networks. Nature Reviews Genetics, 10, 141–148. Fedoroff, N. V. (2012). Transposable elements, epigenetics, and genome evolution. Science, 338, 758–767. Freitas, R., Zhang, G. J., & Cohn, M. J. (2006). Evidence that mechanisms of fin development evolved in the midline of early vertebrates. Nature, 442, 1033–1037. Gaiti, F., Calcino, A. D., Tanurdzic, M., & Degnan, B. M. (2017). Origin and evolution of the metazoan non-coding regulatory genome. Developmental Biology, 427, 193–202. Gao, F., & Davidson, E. H. (2008). Transfer of a large gene regulatory apparatus to a new developmental address in echinoid evolution. Proceedings of the National Academy of Sciences of the United States of America, 105, 6091–6096. Glassford, W. J., Johnson, W. C., Dall, N. R., Smith, S. J., Liu, Y., Boll, W., et al. (2015). Co-option of an ancestral hox-regulated network underlies a recently evolved morphological novelty. Developmental Cell, 34, 520–531. Godfrey-Smith, P. (2009). Darwinian populations and natural selection. Oxford: Oxford University Press.

428

Douglas H. Erwin

Gompel, N., Prud’homme, B., Wittkopp, P. J., Kassner, V. A., & Carroll, S. B. (2005). Chance caught on the wing: cis-regulatory evolution and the origin of pigment patterns in Drosophila. Nature, 433, 481–487. Gould, S. J. (2002). The structure of evolutionary theory. Cambridge: Harvard University Press. Green, S. A., Simoes-Costa, M., & Bronner, M. E. (2015). Evolution of vertebrates as viewed from the crest. Nature, 520, 474–482. Halfon, M. S. (2019). Studying transcriptional enhancers: The founder fallacy, validation creep, and other biases. Trends in Genetics, 35, 93–103. Holland, P. W. H., Marletaz, F., Maeso, I., Dunwell, T. L., & Paps, J. (2017). New genes from old: Asymmetric divergence of gene duplicates and the evolution of development. Philosophical Transactions of the Royal Society of London, Series B, 372, 20150480. Hu, Y. G., Schmitt-Engel, C., Schwirz, J., Stroehlein, N., Richter, T., Majumdar, U., et al. (2018). A morphological novelty evolved by co-option of a reduced gene regulatory network and gene recruitment in a beetle. Proceedings of the Royal Society of London B, 285, 20181373. Irish, V. F., & Litt, A. (2005). Flower development and evolution: Gene duplication, diversification and redeployment. Current Opinion in Genetics & Development, 15, 454. Jablonski, D. (2017a). Approaches to macroevolution: 1. General concepts and origin of variation. Evolutionary Biology, 44, 427–450. Jablonski, D. (2017b). Approaches to macroevolution: 2. Sorting of variation, some overarching issues, and general conclusions. Evolutionary Biology, 44, 451–475. Kvon, E. Z., Kamneva, O. K., Melo, U. S., Barozzi, I., Osterwalder, M., Mannion, B. J., et al. (2016). Progressive loss of function in a limb enhancer during snake evolution. Cell, 167(633–642), e611. Larroux, C., Luke, G. N., Koopman, P., Rokhsar, D., Shimeld, S. M., & Degnan, B. M. (2008). Genesis and expansion of metazoan transcription factor gene classes. Molecular Biology and Evolution, 25, 980–996. Lemons, D., Fritzenwanker, J. H., Gerhart, J., Lowe, C. J., & McGinnis, W. (2010). Co-option of an anteroposterior head axis patterning system for proximodistal patterning of appendages in early bilaterian evolution. Developmental Biology, 344, 358–362. Liu, Y., Ramos-Womack, M., Han, C., Reilly, P., Brackett, K. L., Rogers, W., et al. (2019). Changes throughout a genetic network mask the contribution of hox gene evolution. Current Biology, 29(2157–2166), e2156. Lu, Z., Marand, A. P., Ricci, W. A., Ethridge, C. L., Zhang, X., & Schmitz, R. J. (2019). The prevalence, evolution and chromatin signatures of plant regulatory elements. Nature Plants, 5, 1250–1259. Lynch, M. (2007). The origins of genome architecture. Sunderland, MA: Sinauer Assoc. Lynch, V. J., Nnamani, M. C., Kapusta, A., Brayer, K., Plaza, S. L., Mazur, E. C., et al. (2015). Ancient transposable elements transformed the uterine regulatory landscape and transcriptome during the evolution of mammalian pregnancy. Cell Reports, 10, 551–561. Malik, H. S., & Henikoff, S. (2009). Major evolutionary transitions in centromere complexity. Cell, 138, 1067–1082. Martik, M. L., & McClay, D. R. (2015). Deployment of a retinal determination gene network drives directed cell migration in the sea urchin embryo. eLife, 4, e08827. Martin, A., McCulloch, K. J., Patel, N. H., Briscoe, A. D., Gilbert, L. E., & Reed, R. D. (2014). Multiple recent co-options of Optix associated with novel traits in adaptive butterfly wing radiations. EvoDevo, 5, 7. Martin, A., & Orgogozo, V. (2013). The loci of repeated evolution: A catalog of genetic hotspots of phenotypic variation. Evolution, 67, 1235–1250. Massey, J. H., & Wittkopp, P. J. (2016). The genetic basis of pigmentation differences within and between Drosophila species. Current Topics in Developmental Biology, 119, 27–61.

Evolutionary dynamics of gene regulation

429

McQueen, E., & Rebeiz, M. (2020). On the specificity of gene regulatory networks: How does network co-option affect subsequent evolution? Current Topics in Developmental Biology, 139, 375–405. Melo, D., Porto, A., Cheverud, J. M., & Marroig, G. (2016). Modularity: Genes, development, and evolution. Annual Review of Ecology, Evolution, and Systematics, 47, 463–486. Moczek, A. P. (2008). On the origins of novelty in development and evolution. BioEssays, 30, 432–447. Moczek, A. P. (2009). The origin and diversification of complex traits through microand macroevolution of development: Insights from horned beetles. Current Topics in Developmental Biology, 86, 137–162. Moczek, A. P., & Rose, D. J. (2009). Differential recruitment of limb patterning genes during development and diversification of beetle horns. Proceedings of the National Academy of Sciences of the United States of America, 106, 8992–8997. Monteiro, A., & Gupta, M. D. (2016). Identifying coopted networks and causative mutations in the origin of novel complex traits. Current Topics in Developmental Biology, 119, 205–226. Muller, G. B., & Wagner, G. P. (1991). Novelty in evolution: Restructuring the concept. Annual Review of Ecology and Systematics, 22, 229–256. Nakamura, T., Klomp, J., Pieretti, J., Schneider, I., Gehrke, A. R., & Shubin, N. H. (2015). Molecular mechanisms underlying the exceptional adaptations of batoid fins. Proceedings of the National Academy of Sciences of the United States of America, 112, 15940–15945. Pavlicev, M., & Wagner, G. P. (2012). A model of developmental evolution: Selection, pleiotropy and compensation. Trends in Ecology & Evolution, 27, 316–322. Peter, I. S., & Davidson, E. H. (2015). Genomic control processes. In Development and evolution. London: Academic Press. Petit, F., Sears, K. E., & Ahituv, N. (2017). Limb development: A paradigm of gene regulation. Nature Reviews Genetics, 18, 245–258. Phillips, P. C. (2008). Epistasis—The essential role of gene interactions in the structure and evolution of genetic systems. Nature Reviews Genetics, 9, 855–867. Prud’homme, B., Gompel, N., & Carroll, S. B. (2007). Emerging principles of regulatory evolution. Proceedings of the National Academy of Sciences of the United States of America, 104(Suppl. 1), 8605–8612. Ramsey, G. R., & Pence, C. H. (2016). Chance in evolution. Chicago, IL: University of Chicago Press. Rebeiz, M., Jikomes, N., Kassner, V. A., & Carroll, S. B. (2011). Evolutionary origin of a novel gene expression pattern through co-option of the latent activities of existing regulatory sequences. Proceedings of the National Academy of Sciences of the United States of America, 108, 10036–10043. Rebeiz, M., Patel, N. H., & Hinman, V. F. (2015). Unraveling the tangled skein: The evolution of transcriptional regulatory networks in development. Annual Review of Genomics and Human Genetics, 16, 103–131. Reiter, F., Wienerroither, S., & Stark, A. (2017). Combinatorial function of transcription factors and cofactors. Current Opinion Genetics & Development, 43, 73–81. Richter, D. J., Fozouni, P., Eisen, M. B., & King, N. (2018). Gene family innovation, conservation and loss on the animal stem lineage. eLife, 7, e34226. Richter, D. J., & King, N. (2013). The genomic and cellular foundations of animal origins. Annual Review of Genetics, 47, 509–537. Sabaris, G., Laiker, I., Preger-Ben Noon, E., & Frankel, N. (2019). Actors with multiple roles: Pleiotropic enhancers and the paradigm of enhancer modularity. Trends in Genetics, 35, 423–433.

430

Douglas H. Erwin

Sackton, T. B., Grayson, P., Cloutier, A., Hu, Z. R., Liu, J. S., Wheeler, N. E., et al. (2019). Convergent regulatory evolution and loss of flight in paleognathous birds. Science, 364, 74–78. Sauka-Spengler, T., Meulemans, D., Jones, M., & Bronner-Fraser, M. (2007). Ancient evolutionary origin of the neural crest gene regulatory network. Developmental Cell, 13, 405–420. Schmitz, J. F., Zimmer, F., & Bornberg-Bauer, E. (2016). Mechanisms of transcription factor evolution in Metazoa. Nucleic Acids Research, 44, 6287–6297. Sebe-Pedros, A., Ballare, C., Parra-Acero, H., Chiva, C., Tena, J. J., Sabido, E., et al. (2016). The dynamic regulatory genome of Capsaspora and the origin of animal multicellularity. Cell, 165, 1224–1237. Sebe-Pedros, A., Chomsky, E., Pang, K., Lara-Astiaso, D., Gaiti, F., Mukamel, Z., et al. (2018). Early metazoan cell type diversity and the evolution of multicellular gene regulation. Nature Ecology & Evolution, 2, 1176. Sebe-Pedros, A., Degnan, B. M., & Ruiz-Trillo, I. (2017). The origin of Metazoa: A unicellular perspective. Nature Reviews Genetics, 18, 498–512. Simakov, O., & Kawashima, T. (2017). Independent evolution of genomic characters during major metazoan transitions. Developmental Biology, 427, 179–182. Simakov, O., Kawashima, T., Marletaz, F., Jenkins, J., Koyanagi, R., Mitros, T., et al. (2015). Hemichordate genomes and deuterostome origins. Nature, 527, 459. Sogabe, S., Hatleberg, W. L., Kocot, K. M., Say, T. E., Stoupin, D., Roper, K. E., et al. (2019). Pluripotency and the origin of animal multicellularity. Nature, 570, 519–522. Sorrells, T. R., Booth, L. N., Tuch, B. B., & Johnson, A. D. (2015). Intersecting transcription networks constrain gene regulatory evolution. Nature, 523, 361–365. Sorrells, T. R., & Johnson, A. D. (2015). Making sense of transcription networks. Cell, 161, 714–723. Stern, D. L. (2013). The genetic causes of convergent evolution. Nature Reviews Genetics, 14, 751–764. Sundaram, V., Cheng, Y., Ma, Z., Li, D., Xing, X., Edge, P., et al. (2014). Widespread contribution of transposable elements to the innovation of gene regulatory networks. Genome Research, 24, 1963–1976. Sundaram, V., & Wang, T. (2018). Transposable element mediated innovation in gene regulatory landscapes of cells: Re-visiting the “gene-battery” model. BioEssays, 40, 1700155. Torres Cleuren, Y. N., Ewe, C. K., Chipman, K. C., Mears, E. R., Wood, C. G., Al-Alami, C. E. A., et al. (2019). Extensive intraspecies cryptic variation in an ancient embryonic gene regulatory network. eLife, 8, e48220. Trizzino, M., Park, Y., Holsbach-Beltrame, M., Aracena, K., Mika, K., Caliskan, M., et al. (2017). Transposable elements are the primary source of novelty in primate gene regulation. Genome Research, 27, 1623–1633. True, J. R., & Carroll, S. B. (2002). Gene co-option in physiological and morphological evolution. Annual Review of Cell and Developmental Biology, 18, 53–80. True, J. R., & Haag, E. S. (2001). Developmental system drift and flexibility in evolutionary trajectories. Evolution & Development, 3, 109–119. Tseng, W. F., Jang, T. H., Huang, C. B., & Yuh, C. H. (2011). An evolutionarily conserved kernel of gata5, gata6, otx2 and prdm1a operates in the formation of endoderm in zebrafish. Developmental Biology, 357, 541–557. Villar, D., Berthelot, C., Aldridge, S., Rayner, T. F., Lukk, M., Pignatelli, M., et al. (2015). Enhancer evolution across 20 mammalian species. Cell, 160, 554–566. Wagner, G. P. (2007). The developmental genetics of homology. Nature Reviews Genetics, 8, 473–479.

Evolutionary dynamics of gene regulation

431

Wagner, G. P. (2014). Homology, genes, and evolutionary innovation. Princeton, NJ: Princeton University Press. Wagner, G. P., & Zhang, J. Z. (2011). The pleiotropic structure of the genotype-phenotype map: The evolvability of complex organisms. Nature Reviews Genetics, 12, 204–213. Wallbank, R. W. R., Baxter, S. W., Pardo-Diaz, C., Hanly, J. J., Martin, S. H., Mallet, J., et al. (2016). Evolutionary novelty in a butterfly wing pattern through enhancer shuffling. PLoS Biology, 14, e1002353. Wheeler, B. M., Heimberg, A. M., Moy, V. N., Sperling, E. A., Holstein, T. W., Heber, S., et al. (2009). The deep evolution of metazoan microRNAs. Evolution & Development, 11, 50–68. Wittkopp, P. J., Vaccaro, K., & Carroll, S. B. (2002). Evolution of yellow gene regulation and pigmentation in Drosophila. Current Biology, 12, 1547–1556. Wong, E. S., Tan, S. Z., Garside, V., Vanwallenghem, G., Gaiti, F., Scott, E., et al. (2019). Early origin and deep conservation of enhancers in animals. BioRxiv, 633651. Wray, G. A., Hahn, M. W., Abouheif, E., Balhoff, J. P., Pizer, M., Rockman, M. V., et al. (2003). The evolution of transcriptional regulation in eukaryotes. Molecular Biology and Evolution, 20, 1377–1419. Yadav, T., Quivy, J. P., & Almouzni, G. (2018). Chromatin plasticity: A versatile landscape that underlies cell fate and identity. Science, 361, 1332–1336. Zattara, E. E., Busey, H. A., Linz, D. M., Tomoyasu, Y., & Moczek, A. P. (2016). Neofunctionalization of embryonic head patterning genes facilitates the positioning of novel traits on the dorsal head of adult beetles. Proceedings of the Royal Society B, 283, 21060824. Zimmermann, B., Robert, N. S. M., Technau, U., & Simakov, O. (2019). Ancient animal genome architecture reflects cell type identities. Nature Ecology & Evolution, 3, 1289–1293. Zitnik, M., Sosic, R., Feldman, M. W., & Leskovec, J. (2019). Evolution of resilience in protein interactomes across the tree of life. Proceedings of the National Academy of Sciences of the United States of America, 116, 4426–4433.

Further reading Erwin, D.H., Submitted-a. A conceptual model of evolutionary novelty and innovation. Biological Reviews n.d., under review.