Target Identification and Validation in Drug Discovery: Methods and Protocols [2nd ed.] 978-1-4939-9144-0;978-1-4939-9145-7

Table of contents :
Front Matter ....Pages i-xiii
Front Matter ....Pages 1-1
Using Functional Genetics in Haploid Cells for Drug Target Identification (Jennifer C. Volz, Nicole Schuller, Ulrich Elling)....Pages 3-21
Unbiased Forward Genetic Screening with Chemical Mutagenesis to Uncover Drug–Target Interactions (Moritz Horn, Franziska Metge, Martin S. Denzel)....Pages 23-31
Integration of RNAi and Small Molecule Screens to Identify Targets for Drug Development (Konstantinos Drosopoulos, Spiros Linardopoulos)....Pages 33-42
CellProfiler and KNIME: Open-Source Tools for High-Content Screening (Martin Stöter, Antje Janosch, Rico Barsacchi, Marc Bickle)....Pages 43-60
Front Matter ....Pages 61-61
In Silico Target Druggability Assessment: From Structural to Systemic Approaches (Jean-Yves Trosset, Christian Cavé)....Pages 63-88
In Silico Drug–Target Profiling (Jean-Yves Trosset, Christian Cavé)....Pages 89-103
Locus-Specific Knock-In of a Degradable Tag for Target Validation Studies (Matthias Brand, Georg E. Winter)....Pages 105-119
Expression of Human VH Single Domains as Fc Fusions in Mammalian Cells (Mahmoud Abdelatti, Peter Schofield, Daniel Christ)....Pages 121-136
Front Matter ....Pages 137-137
The Neurosphere Assay (NSA) Applied to Neural Stem Cells (NSCs) and Cancer Stem Cells (CSCs) (Rossella Galli)....Pages 139-149
3D-3 Tumor Models in Drug Discovery for Analysis of Immune Cell Infiltration (Annika Osswald, Viola Hedrich, Wolfgang Sommergruber)....Pages 151-162
Establishment and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Adenocarcinoma for Application in Drug Discovery (Julia C. Meier-Hubberten, Michael P. Sanderson)....Pages 163-179
Front Matter ....Pages 181-181
In Vivo Pharmacology Models for Cancer Target Research (Dawei Chen, Xiaoyu An, Xuesong Ouyang, Jie Cai, Demin Zhou, Qi-Xiang Li)....Pages 183-211
Use of CRISPR/Cas9 for the Modification of the Mouse Genome (Alexander Klimke, Steffen Güttler, Petric Kuballa, Simone Janzen, Sonja Ortmann, Adriano Flora)....Pages 213-230
Assessment of Gene Function of Mouse Innate Lymphoid Cells for In Vivo Analysis Using Retroviral Transduction (Cyril Seillet, Gabrielle T. Belz)....Pages 231-240
Creation of PDX-Bearing Humanized Mice to Study Immuno-oncology (Li-Chin Yao, Ken-Edwin Aryee, Mingshan Cheng, Pali Kaur, James G. Keck, Michael A. Brehm)....Pages 241-252
Immunophenotyping of Tissue Samples Using Multicolor Flow Cytometry (Martina M. Sykora, Markus Reschke)....Pages 253-268
Front Matter ....Pages 269-269
Combined MicroRNA In Situ Hybridization and Immunohistochemical Detection of Protein Markers (Boye Schnack Nielsen, Kim Holmstrøm)....Pages 271-286
Tangential Flow Filtration with or Without Subsequent Bind-Elute Size Exclusion Chromatography for Purification of Extracellular Vesicles (Joel Z. Nordin, R. Beklem Bostancioglu, Giulia Corso, Samir EL Andaloussi)....Pages 287-299
Chromogenic Tissue-Based Methods for Detection of Gene Amplifications (or Rearrangements) Combined with Protein Overexpression in Clinical Samples (Hiroaki Nitta, Brian Kelly)....Pages 301-314
Back Matter ....Pages 315-318

Citation preview

Methods in Molecular Biology 1953

Jürgen Moll Sebastian Carotta Editors

Target Identification and Validation in Drug Discovery Methods and Protocols Second Edition

METHODS

IN

MOLECULAR BIOLOGY

Series Editor John M. Walker School of Life and Medical Sciences University of Hertfordshire Hatfield, Hertfordshire, AL10 9AB, UK

For further volumes: http://www.springer.com/series/7651

Target Identification and Validation in Drug Discovery Methods and Protocols Second Edition

Edited by

Jürgen Moll Pharmacology and Translational Research, Boehringer Ingelheim RCV GmbH & Co KG, Vienna, Austria

Sebastian Carotta Department of Cancer Cell Signaling, Boehringer Ingelheim RCV GmBH & Co KG, Vienna, Austria

Editors Ju¨rgen Moll Pharmacology and Translational Research Boehringer Ingelheim RCV GmbH & Co KG Vienna, Austria

Sebastian Carotta Department of Cancer Cell Signaling Boehringer Ingelheim RCV GmBH & Co KG Vienna, Austria

ISSN 1064-3745 ISSN 1940-6029 (electronic) Methods in Molecular Biology ISBN 978-1-4939-9144-0 ISBN 978-1-4939-9145-7 (eBook) https://doi.org/10.1007/978-1-4939-9145-7 Library of Congress Control Number: 2019932163 © Springer Science+Business Media, LLC, part of Springer Nature 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Humana Press imprint is published by the registered company Springer Science+Business Media, LLC, part of Springer Nature. The registered company address is: 233 Spring Street, New York, NY 10013, U.S.A.

Preface Considering the incredible new developments in technologies over the last few years, this second edition of the book took care to cover breakthrough technologies in the field of drug target identification and validation. We are confident that the second edition will be as successful as the first, with the majority of chapters capturing recently emerging technologies. These new technologies include 3D cell culture and in particular revolutionary technologies such as CRISPR, which has a wide field of applications. Since new targets are still a valuable asset for drug development, an emphasis in this issue is on CRISPR-related screening technologies. Usage of haploid cell lines adds another breakthrough to exploit best CRISPR technologies. Other dynamic fields are “big data” and in silico approaches; hence we extended these topics compared to the previous edition. The in vivo applications of CRISPR and the best use of animal models in drug development complete the validation aspects covered by this book. The quality of target identification and validation is a first and critical indicator for attrition rate in drug development. In a wider sense, target validation also includes aspects of efficacy and target patient population, which together define the drug properties and commercial aspects that determine the success of a drug development program. Humanized in vitro and in vivo models are instrumental to judge the probability of success, both of which are covered in this book. This book contains a comprehensive collection of essential and state-of-the-art methods, contributed by internationally recognized experts in their specialized fields. The content of each chapter goes beyond pure protocol lists to also include useful hints, emphasizing the most critical steps and pinpointing typical pitfalls. The chapters are organized by major categories covering methods of early drug development related to target identification and validation but also translational aspects, such as animal models and biomarker development. This book is a valuable source of protocols for lab scientists; in addition, it represents a useful compendium for any “drug hunter” including molecular and cellular biologists, pharmacologists, pathologists, bioinformaticians, clinical researchers, or investigators, to name a few. Last but not least, any scientist who is not an expert in the field will get a quick overview on state-of-the-art technologies. Most importantly, we thank all the authors for their valuable contributions. It was a real pleasure to interact with them in a highly professional manner. The result of these efforts is the second edition of a book of which all contributors can be proud. ¨ rgen Moll Ju Sebastian Carotta

Vienna, Austria

v

Acknowledgments We thank all the authors for their dedicated efforts and Jordi Moll for help in editing.

vii

Contents Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

PART I

IDENTIFICATION OF NOVEL TARGETS USING HIGH THROUGHPUT SCREENING ASSAYS

1 Using Functional Genetics in Haploid Cells for Drug Target Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jennifer C. Volz, Nicole Schuller, and Ulrich Elling 2 Unbiased Forward Genetic Screening with Chemical Mutagenesis to Uncover Drug–Target Interactions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Moritz Horn, Franziska Metge, and Martin S. Denzel 3 Integration of RNAi and Small Molecule Screens to Identify Targets for Drug Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Konstantinos Drosopoulos and Spiros Linardopoulos 4 CellProfiler and KNIME: Open-Source Tools for High-Content Screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Martin Sto¨ter, Antje Janosch, Rico Barsacchi, and Marc Bickle

PART II

v vii xi

3

23

33

43

DRUG TARGET PROFILING AND VALIDATION

5 In Silico Target Druggability Assessment: From Structural to Systemic Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 Jean-Yves Trosset and Christian Cave´ 6 In Silico Drug–Target Profiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 Jean-Yves Trosset and Christian Cave´ 7 Locus-Specific Knock-In of a Degradable Tag for Target Validation Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 Matthias Brand and Georg E. Winter 8 Expression of Human VH Single Domains as Fc Fusions in Mammalian Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 Mahmoud Abdelatti, Peter Schofield, and Daniel Christ

PART III

THREE DIMENSIONAL CELL CULTURE TECHNIQUES MIMICKING THE TUMOR MICROENVIRONMENT IN VITRO

9 The Neurosphere Assay (NSA) Applied to Neural Stem Cells (NSCs) and Cancer Stem Cells (CSCs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 Rossella Galli

ix

x

10

11

Contents

3D-3 Tumor Models in Drug Discovery for Analysis of Immune Cell Infiltration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Annika Osswald, Viola Hedrich, and Wolfgang Sommergruber Establishment and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Adenocarcinoma for Application in Drug Discovery . . . . . . 163 Julia C. Meier-Hubberten and Michael P. Sanderson

PART IV ANIMAL MODELS FOR THE STUDY OF GENE FUNCTION IN VIVO 12

In Vivo Pharmacology Models for Cancer Target Research . . . . . . . . . . . . . . . . . . Dawei Chen, Xiaoyu An, Xuesong Ouyang, Jie Cai, Demin Zhou, and Qi-Xiang Li 13 Use of CRISPR/Cas9 for the Modification of the Mouse Genome . . . . . . . . . . . ¨ ttler, Petric Kuballa, Simone Janzen, Alexander Klimke, Steffen Gu Sonja Ortmann, and Adriano Flora 14 Assessment of Gene Function of Mouse Innate Lymphoid Cells for In Vivo Analysis Using Retroviral Transduction . . . . . . . . . . . . . . . . . . . . . . . . . Cyril Seillet and Gabrielle T. Belz 15 Creation of PDX-Bearing Humanized Mice to Study Immuno-oncology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Li-Chin Yao, Ken-Edwin Aryee, Mingshan Cheng, Pali Kaur, James G. Keck, and Michael A. Brehm 16 Immunophenotyping of Tissue Samples Using Multicolor Flow Cytometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Martina M. Sykora and Markus Reschke

PART V

183

213

231

241

253

TRANSLATIONAL METHODS TO VALIDATE BIOMARKERS

17

Combined MicroRNA In Situ Hybridization and Immunohistochemical Detection of Protein Markers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 Boye Schnack Nielsen and Kim Holmstrøm 18 Tangential Flow Filtration with or Without Subsequent Bind-Elute Size Exclusion Chromatography for Purification of Extracellular Vesicles . . . . . . 287 Joel Z. Nordin, R. Beklem Bostancioglu, Giulia Corso, and Samir EL Andaloussi 19 Chromogenic Tissue-Based Methods for Detection of Gene Amplifications (or Rearrangements) Combined with Protein Overexpression in Clinical Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301 Hiroaki Nitta and Brian Kelly Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

315

Contributors MAHMOUD ABDELATTI  Immunology Program, Garvan Institute of Medical Research, Sydney, NSW, Australia; Faculty of Medicine, St. Vincent’s Clinical School, University of New South Wales, Sydney, NSW, Australia XIAOYU AN  Crown Bioscience Inc., San Diego, CA, USA; State Key Laboratory of Natural and Biomimetic Drugs, Peking University, Beijing, China KEN-EDWIN ARYEE  Department of Molecular Medicine and the Diabetes Center of Excellence, University of Massachusetts Medical School, Worcester, MA, USA RICO BARSACCHI  Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany GABRIELLE T. BELZ  Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia; Department of Medical Biology, University of Melbourne, Melbourne, Australia MARC BICKLE  Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany R. BEKLEM BOSTANCIOGLU  Department of Laboratory Medicine, Clinical Research Center, Huddinge, Sweden MATTHIAS BRAND  CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences, Vienna, Austria MICHAEL A. BREHM  Department of Molecular Medicine and the Diabetes Center of Excellence, University of Massachusetts Medical School, Worcester, MA, USA JIE CAI  Crown Bioscience Inc., San Diego, CA, USA CHRISTIAN CAVE´  BioCIS UFR Pharmacie UMR CNRS 8076, Universite´ Paris Saclay, Orsay, France DAWEI CHEN  Crown Bioscience Inc., San Diego, CA, USA MINGSHAN CHENG  Department of In Vivo Pharmacology Services, The Jackson Laboratory, Sacramento, CA, USA DANIEL CHRIST  Immunology Program, Garvan Institute of Medical Research, Sydney, NSW, Australia; Faculty of Medicine, St. Vincent’s Clinical School, University of New South Wales, Sydney, NSW, Australia GIULIA CORSO  Department of Laboratory Medicine, Clinical Research Center, Huddinge, Sweden MARTIN S. DENZEL  Max Planck Institute for Biology of Ageing, Cologne, Germany KONSTANTINOS DROSOPOULOS  The Breast Cancer Toby Robins Research Centre, The Institute of Cancer Research, London, UK SAMIR EL ANDALOUSSI  Department of Laboratory Medicine, Clinical Research Center, Huddinge, Sweden; Evox Therapeutics Limited, Medawar Centre, Oxford, UK ULRICH ELLING  IMBA - Institute of Molecular Biotechnology of the Austrian Academy of Sciences, Vienna, Austria ADRIANO FLORA  PerkinElmer chemagen Technologie GmbH, Baesweiler, Germany ROSSELLA GALLI  Neural Stem Cell Biology Unit, Division of Neuroscience, San Raffaele Scientific Institute, Milan, Italy STEFFEN GU¨TTLER  Taconic Biosciences GmbH, Cologne, Germany VIOLA HEDRICH  Boehringer Ingelheim RCV GmbH & Co KG, Vienna, Austria

xi

xii

Contributors

KIM HOLMSTRØM  Bioneer A/S, Hørsholm, Denmark MORITZ HORN  Max Planck Institute for Biology of Ageing, Cologne, Germany ANTJE JANOSCH  Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany SIMONE JANZEN  Taconic Biosciences GmbH, Cologne, Germany PALI KAUR  Department of In Vivo Pharmacology Services, The Jackson Laboratory, Sacramento, CA, USA JAMES G. KECK  Department of In Vivo Pharmacology Services, The Jackson Laboratory, Sacramento, CA, USA BRIAN KELLY  Roche Tissue Diagnostics, Tucson, AZ, USA ALEXANDER KLIMKE  Taconic Biosciences GmbH, Cologne, Germany PETRIC KUBALLA  Taconic Biosciences GmbH, Cologne, Germany QI-XIANG LI  Crown Bioscience Inc., San Diego, CA, USA; State Key Laboratory of Natural and Biomimetic Drugs, Peking University, Beijing, China SPIROS LINARDOPOULOS  The Breast Cancer Toby Robins Research Centre, The Institute of Cancer Research, London, UK; Cancer Research UK, Cancer Therapeutics Unit, Division of Cancer Therapeutics, The Institute of Cancer Research, Sutton, UK JULIA C. MEIER-HUBBERTEN  Translational Innovation Platform Oncology, Merck KGaA, Darmstadt, Germany FRANZISKA METGE  Max Planck Institute for Biology of Ageing, Cologne, Germany BOYE SCHNACK NIELSEN  Bioneer A/S, Hørsholm, Denmark HIROAKI NITTA  Roche Tissue Diagnostics, Tucson, AZ, USA JOEL Z. NORDIN  Department of Laboratory Medicine, Clinical Research Center, Huddinge, Sweden; Evox Therapeutics Limited, Medawar Centre, Oxford, UK SONJA ORTMANN  Taconic Biosciences GmbH, Cologne, Germany ANNIKA OSSWALD  Boehringer Ingelheim RCV GmbH & Co KG, Vienna, Austria XUESONG OUYANG  Crown Bioscience Inc., San Diego, CA, USA MARKUS RESCHKE  Boehringer Ingelheim RCV GmbH & Co KG, Vienna, Austria MICHAEL P. SANDERSON  Translational Innovation Platform Oncology, Merck KGaA, Darmstadt, Germany PETER SCHOFIELD  Immunology Program, Garvan Institute of Medical Research, Sydney, NSW, Australia NICOLE SCHULLER  IMBA - Institute of Molecular Biotechnology of the Austrian Academy of Sciences, Vienna, Austria CYRIL SEILLET  Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia; Department of Medical Biology, University of Melbourne, Melbourne, Australia WOLFGANG SOMMERGRUBER  Boehringer Ingelheim RCV GmbH & Co KG, Vienna, Austria MARTIN STO¨TER  Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany MARTINA M. SYKORA  Boehringer Ingelheim RCV GmbH & Co KG, Vienna, Austria JEAN-YVES TROSSET  Bioinformation Research Laboratory, Sup’Biotech, Villejuif, France JENNIFER C. VOLZ  IMBA - Institute of Molecular Biotechnology of the Austrian Academy of Sciences, Vienna, Austria GEORG E. WINTER  CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences, Vienna, Austria

Contributors

xiii

LI-CHIN YAO  Department of In Vivo Pharmacology Services, The Jackson Laboratory, Sacramento, CA, USA DEMIN ZHOU  State Key Laboratory of Natural and Biomimetic Drugs, Peking University, Beijing, China

Part I Identification of Novel Targets Using High Throughput Screening Assays

Chapter 1 Using Functional Genetics in Haploid Cells for Drug Target Identification Jennifer C. Volz, Nicole Schuller, and Ulrich Elling Abstract Pooled genetic screens are a powerful tool to identify targets for drug development as well as chemogenetic interactions. Various complementary methods for mutagenesis are available to generate highly complex cell populations, including mRNA knockdown, directed genome editing, as well as random genome mutagenesis. With the availability of a growing number of haploid mammalian cell lines, random mutagenesis is becoming increasingly powerful and represents an attractive alternative, e.g., to CRISPR-based screening. This chapter provides a step-by-step protocol for performing haploid gene trap screens. Key words Functional genomics, Genetic screen, Haploid, Gene trap, Stem cell

1

Introduction Chemogenomic approaches can support drug development by rapidly uncovering functional interactions of small molecules and genes [1]. For example, overexpression of direct drug targets can result in partial resistance of cells to compounds and thus contribute to target deconvolution. Genetic screens can also shed light on resistance mechanisms of cytotoxic compounds such as chemotherapeutics. In addition, functional genomic studies can identify the genetic interactome of the drug target as well as the required enzymatic activity to activate prodrugs. The use of transcriptional reporters coupled to a fluorescent protein combined with fluorescence-activated cell sorting (FACS) or an antibiotic resistance allows to expand the screenable range of phenotypes from lethal assays to any transcriptional event. Furthermore, immunolabelling of signaling events such as phosphoepitopes followed by flow cytometry allows to deconvolute the genetic interactome of drug-induced signaling. Importantly, genetic phenotypic screens are also a powerful tool to identify novel drug targets for subsequent targeted drug development, for

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_1, © Springer Science+Business Media, LLC, part of Springer Nature 2019

3

4

Jennifer C. Volz et al.

example, for viral host factors or synthetic lethal interactions with tumorigenic mutations in positive or negative selection screening paradigms, respectively. This century has seen several true revolutions in functional genomic approaches including small RNA-based methods to perturb gene function such as siRNA, shRNA, and sgRNA screening. Alternative methods include random mutagenesis induced by chemical mutagens or mobile DNA elements. However, random mutagenesis in mammalian cells was until recently limited to dominant phenotypes, as all mammals and established mammalian cell lines are diploid, i.e., carry two alleles for all autosomal encoded genes. In diploidy, the remaining healthy allele will mask any recessive phenotype. Meanwhile, yeast geneticists immersed in the awesome power of yeast genetics (“APYG”) [2], where the power of haploid screens was already showcased in classic work half a century ago [3]. Chromosome loss and near-haploid chromosome sets are a peculiarity of acute myeloid and lymphoblastic leukemias [4, 5]. Based on the derivation of the near-haploid human myeloid leukemia cell line KBM-7 [6] with only chromosome 8 remaining diploid, seminal work by Carette, Brummelkamp, and co-workers established haploid genetic screens in mammalian cells in 2009 [7]. A partially reprogrammed adherent derivative of KBM-7, termed HAP-1, was developed shortly after [8]. These landmark studies led to the discovery of host factors of bacteriotoxins as well as viral life cycles [7, 9, 10]. As a direct consequence of sexual reproduction, all eukaryotes alternate between a haploid and a diploid stage in their life cycle [11]; however in most eukaryotic species including all mammals, the diploid stage is dominant, while the haploid stage is reduced to mature germ cells. Making use of the haploid stage of the life cycle in germ cells and induction of embryogenesis in the presence of only the maternal or paternal genome, the first fully haploid vertebrate cell lines were derived from frogs [12] and medaka fish [13]. In 2011, two groups reported derivation of the first truly haploid mammalian cell lines, mouse haploid embryonic stem (ES) cells from parthenogenic blastocysts [14, 15]. Shortly after, androgenic (male genome only) haploid mouse ES cells were reported [16, 17], followed by embryonic stem cells of rat [18], monkey [19], and recently human [20, 21]. Furthermore, haploid neural progenitor cells derived from rhesus monkey haploid embryonic stem cells extend the repertoire of haploid cell lines available for genetic screens [22]. Nevertheless, the availability of haploid cell lines still represents the major limitation of haploid screening technology. The fact that human haploid embryonic stem cells— unlike reported stem cell lines from other organisms—remain haploid also upon differentiation [20] opens the possibility of deriving a plethora of new haploid cell lines in the future.

Haploid Genetic Screens

5

Haploid screens are typically performed by targeting gene trap or polyA trap vectors randomly into the genome. This is accomplished by infection of cells with retro- or lentiviral vectors as well as transfection with transposons such as sleeping beauty, Tol2, or piggyBac. Integration of mobile elements can perturb gene expression by the presence of splice acceptors, which will trap mRNA transcripts if integrated into an intron. The advantage of insertional mutagenesis is the ease of identification of mutations in high throughput [9, 23]. However, hot and cold spots of mutagenesis, in particular of viral vectors, affect genome saturation, while transposons insert more evenly [23]. Thus, haploid insertional mutagenesis screens are mostly limited to positive selection and ideally performed based on at least 107–108 independently mutated cells. Under these conditions, positive selection screens identify often hundreds of biologically independent integrations within a gene of interest as causal for the phenotype of interest and can thus assign gene function with unprecedented certainty. Insertional mutagenesis in haploid cells is ideally suited to uncover genetic interactions with small molecules in an unbiased manner based on loss of function phenotypes. Advantages of this approach are that (1) it does not require the generation of small RNA libraries based on gene predictions, (2) mutations are directly identified as opposed to being inferred indirectly by presence of shRNAs or sgRNAs, and (3) results are based on hundreds of biologically independent mutations resulting in unambiguous identification of hits [24]. Moreover, new gene trap systems carrying transcriptional enhancers have been developed to activate gene expression of nearby genes [25]. Thus, enhanced gene trap insertions can uncover direct drug targets also in diploid cells due to increased resistance [26] by increased drug target abundance. Technically, a haploid genetic screen is similar to other pooled screens and consists of the generation of pools of mutagenized cells by infection or transfection with mutagenic elements followed by selection for gene trap insertions and subsequently phenotypes of interest. Mutations are then mapped to the genome by nextgeneration sequencing, and genes with clustered mutations are identified by bioinformatic pipelines. If mutated libraries of cells are already available, screens can be performed in less than 2 months, thereby making haploid insertion screens a powerful tool to identify chemogenetic interactions (Fig. 1). The following protocol provides a detailed step-by-step guide for successful performance of haploid genetic screens in positive selection paradigms.

6

Jennifer C. Volz et al.

Fig. 1 Flowchart and timeline of a haploid genetic drug screen

2 2.1

Materials Cell Culture

1. Embryonic stem cell medium (ESCM): combine 450 mL DMEM, 75 mL fetal calf serum (FCS; see Note 1), 5.5 mL penicillin-streptomycin, 5.5 mL MEM nonessential amino acid solution, 5.5 mL L-glutamine, 5.5 mL sodium pyruvate, and 0.55 mL 2-mercaptoethanol (1000
stock), and ESGRO® Leukemia Inhibitory Factors (according to manufacturers) are sterile filtered using a 0.2 μm sterile filter. 2. 2-Mercaptoethanol 1000
stock: 2-mercaptoethanol in 2.85 mL PBS.

dilute

10

μL

Haploid Genetic Screens

7

3. 1
freezing medium (10% DMSO): mix 25 mL ESCM with 20 mL FCS and 5 mL DMSO. 4. 2
Freezing medium (20% DMSO): mix 20 mL FCS with 5 mL DMSO. 5. Phosphate-buffered saline (PBS). 6. 1
Trypsin-EDTA. 7. 6-well plate, 10 and 15 cm cell culture-treated plates. 8. Centrifuge tubes. 9. 2 mL cryovials. 2.2 Hoechst33342 Staining and 1n FluorescenceActivated Cell Sorting (FACS)

1. Hoechst33342: In sterile conditions (tissue culture hood), dissolve the Hoechst33342 powder in sterile MonoQ to obtain a 10 mg/mL stock solution; aliquot in tinfoil wrapped tubes and store at 20  C.

2.3 Library Generation

1. Gene trap plasmid of your choice (see Note 2).

2. Flow cytometry tubes.

2. Fluorescence control plasmid (see Note 3). 3. Lentiviral (LV) or retroviral (RV) helper plasmids; Gag-Pol and VSV G. 4. Lenti-X 293T (Clontech Laboratories) or PlatE (Cell Biolabs). 5. DMEM. 6. Polyethylenimine (PEI): prepare a PEI stock of 1 mg/mL in MonoQ and sterile filtrate; aliquot and store at 20  C. 7. CaCl2 (1 M). 8. Sterile MonoQ. 9. 2
HBS: combine 49.2 g NaCl (280 mM final conc.), 35.76 g HEPES (50 mM final conc.), 0.63 g Na2HPO4 anhydrous (1.5 mM final conc.), 6 g dextrose or 6.6 g glucose (optional, 12 mM final conc.), and 2.22 g KCl (optional, 10 mM final conc.). Fill up to 2700 mL with sterile MonoQ and put on a magnetic stirrer until chemicals are completely dissolved. Aliquot solution into 6
450 mL and adjust pH with 0.5 N NaOH to 6.92, 6.96, 7.00, 7.04, 7.08, and 7.12. Subsequently, fill up to 500 mL each using sterile MonoQ. Filter sterilize using a 0.2 μm sterile filter and store the solutions at 4  C until test transfection is performed. Perform a CaPO4 transfection (see Subheading 3.3, step 2) for each different pH value using a retroviral plasmid containing a fluorescent marker. Change media the day after transfection and check fluorescence under the microscope. A fluorescence of >80% is desirable. Aliquot the 2
HBS with the best transfection rate (typically there are two different 2
HBS/pH values with

8

Jennifer C. Volz et al.

good transfection rates), and store at 20  C. Once an aliquot is thawed, do not refreeze. 10. 500 mL/250 mL 0.45 μm Rapid-Flow Bottle Top Filter. 11. Sekusept Plus. 12. Incidur Spray. 13. Polybrene (4 mg/mL). 14. Selection media: add the respective antibiotic substance— depending on the selection marker in your gene trap plasmid—to ESCM (see Note 4). 15. 4% paraformaldehyde in PBS or Histofix® Preservative. 2.4

Drug Screen

1. 96-well plate, 96U plates. 2. AlamarBlue cell viability reagent. 3. Genomic DNA lysis buffer (GDLB) without SDS: combine 5 mL 1 M Tris–HCl pH 8.0 (10 mM final conc.), 0.25 mL 1 M EDTA (0.5 mM final conc.), 50 mL 1 M NaCl (100 mM final conc.), and 44.75 mL MonoQ, and autoclave. 4. GDLB with SDS: combine 5 mL 1 M Tris–HCl pH 8.0 (10 mM final conc.), 0.25 mL 1 M EDTA (0.5 mM final conc.), 50 mL 1 M NaCl (100 mM final conc.), 10 mL 20% SDS (2% final conc.), and 34.75 mL MonoQ, and autoclave. 5. Proteinase K 10 mg/mL stock: combine 1 mL 1 M Tris–HCl pH 7.5 (10 mM final conc.), 2 mL 1 M CaCl2 (20 mM final conc.), 57.5 mL 87% glycerol (50% final conc.), and 39.5 mL MonoQ, and dissolve 1 g proteinase K in this solution; store at 20  C. 6. RNase A (100 mg/mL). 7. Roti®-phenol/chloroform/isoamyl alcohol (ratio 25:24:1; P/C/I; use the lower phase). 8. Chloroform/isoamyl alcohol (C/I): add 10 mL of isoamyl alcohol to 240 mL of chloroform. 9. Isopropanol. 10. 70% ethanol. 11. 1
Tris–EDTA (TE) buffer pH 8.

2.5 Mapping the Gene Trap: Inverse PCR

1. Respective restriction enzymes E1/E2 and buffer. 2. T4 DNA ligase and buffer. 3. PCR reagents (10 mM dNTP mix, 20
Taq polymerase, and 10
buffer). 4. DNA/PCR purification kit.

Haploid Genetic Screens

3

9

Methods This protocol describes library generation by viral infection of feeder-free mouse embryonic stem (mES) cells. To amplify the genomic region adjacent to the gene trap, several protocols are available, including linear-amplification-mediated PCR (LAM-PCR) [27], splinkerette PCR [28], and inverse PCR (iPCR) [29]. In this protocol inverse PCR is applied.

3.1

Cell Culture

1. ES cells need to be expanded for the fluorescence-activated cell sort (see Subheading 3.2) and afterward for infection (see Subheading 3.4). 2. To split cells, aspirate medium, wash dish once with PBS, add 1
Trypsin-EDTA, and incubate for 5 min at 37  C. Add ESCM (see Note 5) and dissociate cells to a single-cell suspension. Centrifuge cell suspension at 300
g for 5 min. Aspirate supernatant and resuspend pellet in ESCM according to the split ratio/number of dishes required. Seed cells and incubate at 37  C, 5% CO2 (see Note 6). 3. To freeze cells, trypsinize, stop with ESCM and pellet cells. Resuspend cells directly in 1
freezing medium. Cells are aliquoted into cryogenic vials and then transferred to a freezing container kept at 80  C overnight, before vials are transferred to liquid nitrogen for long-term storage.

3.2 Hoechst33342 Staining and 1n FluorescenceActivated Cell Sorting (FACS)

For library generation, it is essential to use a pure haploid mES cell population. This is achieved through performing a FACS sort, in which Hoechst33342-stained cells are separated by their DNA content, i.e., haploid and diploid state, respectively (Fig. 2). Since 520
106 cells are required for the infection, set the haploid sort on day 7 to day 4 prior to library generation (see Subheading 3.4) depending on the cells’ growth rate and commencing rate of haploidy (see Note 7). 1. Process 2
15 cm dishes at once to avoid prolonged periods of the cells kept on ice. 2. Prepare ESCM-containing Hoechst33342: mix 36 mL ESCM with 60 μL Hoechst33342 stock (10 mg/mL). 3. Aspirate media and wash the cells with PBS. 4. Trypsinize cells with 6 mL of 1
Trypsin-EDTA for 6 min at 37  C. 5. Add 6 mL ESCM and dissociate cells to a single-cell suspension (see Note 8). 6. Add 18 mL Hoechst33342 media (final concentration 10 μg/ mL Hoechst33342) per plate, and immediately swirl the plate to distribute the dye.

10

a

Jennifer C. Volz et al.

b

250K

150K

150K

DAPI-A

SSC-A

200K

Single Cells 96.1

100K

100K 50K 50K

Live 61.8

0

0 0

50K

100K

150K

200K

250K

50K

FSC-A

c 150 1n_1c 21.7

100K 150K 200K 250K 300K DAPI-W

d

2c 40.1

Count

100

2n_4c 28.0

50

0 0

50K

100K

150K

DAPI-A

Fig. 2 Example of a 1n FACS profile of a mouse embryonic stem cell line. (a) Viable cells are being selected in a forward and sideway scatter display. (b) Forward and sideway scatter display of Hoechst33342-stained cells in which single cells are selected. (c) A histogram charting the number of cells counted and the fluorescence intensity. The first peak contains the pure 1n/1c cell population, whereas the “2c” peak contains a mixture of 1n/2c cells and 2n/2c cells. The last gate contains the pure diploid 2n/4c population. (d) Summary displaying the percentages of different populations selected during FACS sort

7. Incubate for 30–40 min at 37  C. 8. To avoid settling of the cells during staining, occasionally mix the cells by gently swirling the plate. 9. After staining, pipet repeatedly to separate cells, collect the cell suspension, and centrifuge at 300
g for 5 min. 10. Aspirate the supernatant and resuspend the cell pellet in 500 μL of ESCM.

Haploid Genetic Screens

11

11. Pass the cells through the cell-strainer cap of a FACS tube and keep the cells on ice until FACS sorting. 12. Subject the cells to FACS sorting using a FACS sorter (e.g., BD FACSAria III sorter) equipped with a near-UV laser. 13. Collect the 1n peak (“1n_1c” in Fig. 2c) in ESCM (e.g., 8 mL of ESCM in a 15 mL centrifuge tube). 14. Centrifuge the collected cells at 300
g for 5 min and aspirate the supernatant. 15. Resuspend cells and plate them according to cell numbers. 16. Exchange ESCM after 4–8 h to increase the viability after FACS sorting. 17. Expand cells for infection (see Subheading 3.1). 3.3 Viral Supernatant Generation

1. Day 2 morning: seed PlatE (retrovirus) or Lenti-X (lentivirus) cells to 21
15 cm dishes 6 h prior to transfection (CaPO4 transfection for PlatE, PEI transfection for Lenti-X) at 60–70% confluency. 2. Day 2 evening, transfection: use 20
15 cm dishes for the transfection with the gene trap plasmid and 1
15 cm for a control transfection with a fluorescent plasmid. PEI transfection for Lenti-X: l

For transduction, mix per 15 cm plate 3 mL DMEM, 9 μg Helper1 Gag-Pol, 4.5 μg Helper2 VSV G, 18 μg LV plasmid with gene trap and vortex.

l

Add 94.5 μL PEI and immediately vortex again.

l

Incubate at room temperature for 15 min.

l

Add the transfection mix to the cells dropwise while swirling the plate.

l

Incubate at 37  C, 5% CO2 (see Note 9).

CaPO4 transfection for PlatE: l

l

Mix per 15 mL plate 60 μg RV plasmid with gene trap, 20 μg helper plasmid, and 375 μL 1 M CaCl2. Adjust to a total volume of 1500 μL with sterile MonoQ (solution A). In a separate tube, prepare 1500 μL of 2
HBS (solution B).

l

Dropwise add solution A to solution B while vortexing solution B. The mixture should turn turbid. Leave at room temperature for 15 min.

l

In case of precipitation, resuspend solution via pipetting or shearing through a 21 G needle.

l

Add mixture to cells in a dropwise manner while moving plate very gently.

l

Transfer plates to the incubator (37  C, 5% CO2).

12

Jennifer C. Volz et al.

3. Day 1: change media to 25 mL ESCM. Check fluorescence of control transfection under microscope. 4. Day 0, morning and evening: harvest viral supernatant and combine. Filter viral supernatant through a 0.45 μm bottletop filter with applied vacuum. Store on ice in between. 3.4 Library Generation

To integrate the gene trap into the haploid mES cells (sorted in Subheading 3.2), an infection with the virus supernatant generated in Subheading 3.3 is performed. 5. Day 0: plate 25
15 cm plates of haploid ESCs for infection and 1
15 cm plate for selection control with 20
106 cells each. Add polybrene to the pooled virus supernatant to a final concentration of 4 μg/mL. Infect the 25 dishes 4 h post seeding with the pooled virus supernatant. 6. Day 1, morning: feed cells to remove remaining viral particles. Collect with a pipet tip a streak of cells from 10
15 cm plates, and seed them onto 1
10 cm plate (test plate). 7. Day 1, evening: start selection on the infected and the selection control plates. 8. Day 2: trypsinize cells, count, spin and plate 20
106 cells per 15 cm plate. This will typically result in approximately 50 plates. Split and seed the same number of uninfected control cells onto 1
15 cm plate. Dissociate cells from the test plate; count and resuspend 11,000 cells in 33 mL. Distribute 30 mL and 3 mL onto 1
15 cm plates each to generate two plates with one containing 10,000 cells (10k test plate) and the second 1000 cells (1k test plate). Top up with ESCM. 9. Days 3–5: feed cells with selection media. Feed the 10k test plate with selection media and the 1k test plate with ESCM only. 10. Day 6: on the selection control plate, all cells should have died. Cryopreserve library: dissociate and count cells after pooling, and then aliquot into 50 mL centrifuge tubes for pelleting. Resuspend pellets in ESCM to freeze approximately 50–100
106 cells per vial, and add an equal amount of 2
freezing medium. Aliquot 2 mL per cryovial. Immediately transfer to freezing containers and to 80  C overnight. 11. Day 7: transfer cryovials to liquid nitrogen. Feed the 10k test plate with selection media and the 1k test plate with ESCM only. 12. Day 8: test thaw one vial of each batch of frozen library cells, count and seed. 13. Day 9: count test thawed cells 1 day post seeding and discard. Feed the 10k test plate with selection media and the 1k test plate with ESCM only.

Haploid Genetic Screens

13

14. Day 11: feed the 10k test plate with selection media and the 1k test plate with ESCM only. 15. Day 12: colonies of the 10k and the 1k test plate are washed, then fixed using 4% paraformaldehyde in PBS or Histofix® Preservative, and washed with MonoQ after 4 h incubation at room temperature. To estimate the infection efficiency and the number of insertions (equivalent number of actually mutagenized cells which corresponds to “complexity”), calculate as follows (see Note 10): l

Efficiency (%) ¼ a/(b * 10).

l

Complexity ¼ c * efficiency (%)/100.

l

Number of cells per insertion ¼ d/complexity.

l

3.5 Drug Titration and Screen

Number of vials that represent the whole library ¼ complexity/e

l

while using the following nomenclature:

l

a: number of colonies on the 10k test plate.

l

b: number of colonies on the 1k test plate.

l

c: number of cells used for infection (here: 500
106).

l

d: total number of cells frozen.

l

e: number of cells frozen per vial.

1. To determine the concentration of the drug applied in the screen, perform a drug titration assay. Use wild-type cells, and perform a serial dilution with the respective drug to assess cell viability. l

l

Seed cells into a 96-well plate at a density of 104 cells/well, perform a drug dilution series 4 h post seeding (leave some wells untreated for b), and monitor cell viability using a bright-field microscope over the next few days. Feed every day. After 3–4 days of incubation, perform an AlamarBlue cell viability test: feed cells with 150 μL AlamarBlue in ESCM according to manufacturer’s recommendation, and incubate cells and blank control for 3 h at 37  C, 5% CO2. Transfer 100 μL to a fresh 96 U plate and read the absorption (560:590 nm) on a multimode microplate reader. Identify the minimal drug concentration that results in complete cytotoxicity. If the concentration is physiologically relevant, continue with step 2.

2. Use mutagenized cells and the concentration determined in the drug titration assay to run a small-scaled pilot screen in 10 cm plates. Plate 1
106 cells per 10 cm dish, and maintain

14

Jennifer C. Volz et al.

cells at three drug concentrations around minimal determined concentration for 10–12 days. Identify a concentration resulting in almost complete cytotoxicity with few clearly defined and expanding colonies. For further success of the screen, it is essential to ensure that remaining colonies are expanding and do not represent remaining cellular debris. 3. For the large-scaled screen, make sure to thaw enough vials to have all insertions represented, and scale the screen up or down according to the number of cells surviving the small-scaled pilot screen. Incubate cells for 10–14 days with the identified drug concentration and trypsinize and pool all remaining cells. After reseeding, continue to expand cells under drug selection until cell number reaches at least 108 cells. 4. Harvest cells, and count and distribute cell suspension into two tubes. Pellet cells and cryopreserve cells from one tube in several cryotubes while using the cells from the second tube for genomic DNA extraction. 5. Genomic DNA extraction (see Note 11): l

l

l

l

l

l

Wash pellet once with PBS, centrifuge again and aspirate supernatant. Resuspend cell pellet in GDLB without SDS (2 mL per 107 cells). Add 250 μL proteinase K stock per mL GDLB with SDS and add the same amount of this mixture as GDLB without SDS, mix by returning the tube several times, and incubate at 55  C overnight. Add RNase A to a final concentration of 100 μg/mL and incubate 1 h at 37  C. Add 1 volume of P/C/I and mix by returning the tube several times. Spin 5 min at 2800
g and transfer the upper liquid phase into a new tube.

l

Add 1 volume of P/C/I and mix by returning the tube several times.

l

Spin again and transfer the upper phase liquid into a fresh tube.

l

Add 1 volume of C/I and mix by returning the tubes several times.

l

Spin 10 s at 2800
g and transfer the upper phase in a new tube.

l

Add 1 volume of isopropanol.

l

Collect DNA precipitate by using, e.g., a pipet, and transfer it to a fresh tube containing 70% ethanol in water.

Haploid Genetic Screens l

3.6 Mapping the Gene Trap: Inverse PCR

15

Centrifuge 15 min at 3000
g.

l

Remove ethanol briefly, spin again to remove remaining ethanol, and add TE buffer (2 mL/108 cells) to the pellet.

l

Leave the tube at room temperature at least for 12 h to fully dissolve genomic DNA.

The inverse PCR strategy (iPCR, Fig. 3) is used to map genomic integration sites of gene trap vectors by next-generation sequencing (NGS). The restriction enzyme E1 (a 4-cutter), which frequently cuts in the genome as well as in the terminal repeat of the gene trap vector, is used to digest the genomic DNA of the cells surviving the screen. By circularizing E1-digested DNA (ring ligation, RL) and subsequently amplifying and sequencing the adjacent genomic region by iPCR, each integration site can be mapped. We optimized all our mobile elements to carry universal primer binding sites (Fig. 3). To improve initial template binding by primers, a linearization step using the restriction enzyme E2 (an 8-cutter) is performed, which reopens the rings generated previously to allow for separation of strands during PCR (see Note 10). To determine the insertion frequencies in your mutagenized library, the original library is sequenced in parallel using the same conditions. It is recommended to upscale the protocol below to ideally map >106 insertions. Usually, we perform iPCR starting with 20 μg of genomic DNA (see Note 12). Adaption based on the number of “surviving” clones (¼independent mutations) may be required. 1. Primer design: the iPCR1 primer contains a sequence that binds to the oligo present on the NGS flow cell surface, an optional index of six to seven bases, and a sequence that binds to the gene trap between the restriction sites of E1 and E2. The index is a custom barcode that allows to “label” all PCR reactions from one complex sample. For NGS, samples can be labelled by using different iPCR1 primers and loaded together on one NGS flow cell. Primer iPCR2 binds between the long terminal repeats (LTR) site and the restriction site of E2. The Solexa flow cell primer binds to the very end of the gene trap sequence and thus is positioned next to the genomic sequence (Fig. 3). 2. Digest genomic DNA sample(s) with enzyme E1 (total volume 400 μL for 20 μg genomic DNA): l

200 μL DNA (100 ng/μL).

l

40 μL 10
enzyme buffer.

l

20 μL enzyme1 (10 U/μL).

l

140 μL dH2O.

16

Jennifer C. Volz et al.

Fig. 3 Schematic illustration of the iPCR strategy. (a) Gene trap insertion. An enhanced gene trap [25] consists of a selection cassette, i.e., β-Geo, conferring resistance to G418, and splice acceptor sites (SA) which are reversible using noncompatible loxP/lox5171 and FRT/F3 sites (in triangles). Osteopontin-enhancer (OPE) elements (in rectangles) enhance expression of β-Geo through Oct4 binding. LTR, long terminal repeats, contain primer-binding sites compatible with iPCR and subsequent Illumina sequencing. Two restriction enzymes (E1 and E2) are positioned adjacent to the LTR and an optional internal barcode (BC), which consists of 32 bases. (b) In the upper panel, a sequence representative of the 50 end of a gene trap is shown. The LTR site contains the Solexa flow cell primer-binding site for NGS. Restriction sites for 4 bp-cutters MseI and NlaIII

Haploid Genetic Screens

17

3. Purify restriction digest, e.g., using spin columns, and measure DNA concentration. 4. Ring ligation (total volume 3000 μL for 10 μg DNA, scale down if you use less DNA): l

10 μg E1-digested DNA.

l

300 μL 10
ligase buffer.

l

H2O to 3000 μL.

l

10 μL T4 DNA ligase (1 U/μL).

l

Incubate at 16  C overnight.

5. Heat-inactivate T4 DNA ligase at 65  C for 15 min (see Note 13). 6. Linearize samples with enzyme E2: add to RL-E1-DNA (3000 μL) 10 μL of E2 (10 U/μL). 7. Purify restriction digest, e.g., with spin columns, and measure DNA concentration. 8. Set up the iPCR reaction (see Note 14): l

10 μL E2-RL-E1-DNA (30–120 ng/μL).

l

0.1 μL primer iPCR1 (100 μM).

l

0.1 μL primer iPCR2 (100 μM).

l

1 μL 10 mM dNTP mix.

l

5 μL 10
Taq polymerase buffer.

l

3 μL 20
Taq polymerase.

l

30.8 μL dH2O.

9. Perform PCR using optimized cycling parameters. For our primer set, we use (1) 95  C 3 min, (2) 95  C 15 s, (3) 61  C 25 s, (4) 75  C 2 min, repeat steps (2)–(4) 37
, (5) 72  C 5 min.  Fig. 3 (continued) (E1) and 8 bp-cutter SbfI (E2) and primer-binding sites for iPCR1 and iPCR2 are positioned downstream. A black arrow indicates the position of an optional internal barcode. In the lower panel, full sequence of iPCR1 and iPCR2 primers is shown. They contain in their 50 end an adaptor sequence and in the 30 end the gene trap-binding site. The iPCR1 primer additionally contains an index sequence of six to seven bases (c) iPCR. Template preparation relies on an E1 digest (1) and subsequent ring ligation (2). For better iPCR efficiency, an additional digest with E2 is performed, which results in a linear fragment as template (3). (d) iPCR gel electrophoresis. Gel electrophoresis analysis of iPCR samples prepared from library and wild-type DNA samples. Here, library and wild-type DNA have been digested by two different E1 restriction enzymes and amplified with the respective iPCR1 primers. DNA smears ranging from 450 to 1200 bp in library DNA samples indicate high numbers of different integration sites. Distinctive bands in the smear indicate an enrichment of a certain insertion in the screen. Please note that the size of the E2-RL-E1 gene trap with attached primers and no genomic parts is approximately 300 bp. Since the wild-type DNA contains no gene trap, no amplicons are visible. Strong bands over 200 bp in the wild-type DNA lanes are indicative of contamination. No DNA bands should be visible in the negative controls (wild-type, negative/water control)

18

Jennifer C. Volz et al.

10. Confirm successful amplification using a fraction of amplicons for gel electrophoresis (1.5% agarose gel, 130–140 V, 45–60 min), and check for smear/bands (see Fig. 3d and Note 15). 11. Pool all PCR samples from one reaction condition. 12. Purify the PCR reaction (see Note 16) and measure the DNA concentration. 13. Submit your samples to NGS. We recommend using single read 50 runs as this is sufficient for successful mapping to the genome. 3.7 Sequence Analysis

4

Genomic integration sites of gene trap vectors are mapped by NGS to a specific gene to a particular region of a chromosome. This high-throughput approach allows for analysis of bulk cultures, i.e., control and selected pools, and thus identification of genes functionally contributing to the selected phenotype. Map reads to the genome (with a maximum of one mismatch) and count reads for each position. Insertion site is adjacent to the first sequenced base. We take a cutoff of ten reads per position to be counted as a true integration site and further filter out positions closer than 3 bp as these can result from PCR slippage. Intersect insertion sites with gene features such as exons, introns, UTR, etc. To evaluate enrichment of mutations within genes, count the number of independent mutagenic insertions. Mutagenic insertions are placed within exons or in introns in sense orientation relative to gene transcripts as only such integrations would result in trapping of mRNA. Integrations upstream of transcripts, e.g., within a 5 kb distance, but before the start of the next gene, are further assessed, as they often result in overexpression of genes due to the enhancers used for enhanced gene trapping. Such upstream integrations can, e.g., be indicative of a direct drug target, whereby overexpression results in drug resistance. The final evaluation of insertions on gene level is based on classes of integrations relative to expected probabilities, as quantified by sequencing of the mutagenized library prior to drug selection.

Notes 1. FCS must be tested for compatibility with mouse embryonic stem cells (mESCs) according to standard protocols such as proliferation and differentiation. 2. We use a gene trap vector as shown in Fig. 3. The gene trapping cassette consists of a selection or a fluorescence marker, a splice acceptor upstream, and a polyadenylation site downstream. Please note that in this protocol a gene trap vector with a

Haploid Genetic Screens

19

selection marker is used. Different LTRs will lead to different genomic integration patterns. Osteopontin enhancers lead to Oct4 binding and thus allow insertions into genes that show no detectable or minimal expression in stem cells. Optional elements of a gene trap vector are a 32 bp long internal barcode to ensure single-clone identification and loxP/lox5171 and FRT/F3 sites flanking the gene trapping cassette, which can be inverted/removed using Cre- or Flp-recombinase. 3. We use a MLP-GFP-puro plasmid. 4. Mutagens used for library generation typically carry a neomycin resistance cassette. Final antibiotic concentrations used on haploid mESCs are G418 250 μg/mL, puromycin 1 μg/mL, blasticidin 10 μg/mL, and hygromycin 250 μg/mL. 5. Always use prewarmed (37  C) media for the cells. 6. The split ratio of the cells can vary. We perform a split ratio of 1:3 for our mESCs if they need to be used the next day or a split ratio of 1:10 if they are used 48 h later. 7. We usually sort 14
15 cm dishes (approximately 800
106 cells, 6.7% haploid cells or 1n_1c as illustrated in Fig. 2). 8. Singularizing the cells is critical for the 1n FACS sort efficiency. 9. Any viral work must be performed according to safety regulations. Wear a coat, two pairs of gloves, mask, overshoes, and disposable sleeves. Have Incidur Spray at hand at all time, and flush/rinse everything with Sekusept after contact with viruscontaining liquid. Leave everything in the hood for UV decontamination. 10. Please note that these numbers are only estimates, since the test might not represent the whole library. The 1:10 dilution will lead to the loss of some insertions, and some mutations will lead to a growth advantage/disadvantage of cells. 11. Perform steps with P/C/I and C/I in a fume hood and wear gloves. The genomic DNA pellet must not dry completely as it will not dissolve optimally. The DNA may be left at room temperature for even a couple of days to mature before proceeding with the iPCR. 12. Always include wild-type genomic DNA as negative control to check for contamination during the process. We use two different E1 enzymes to digest the sample in parallel. The resulting iPCR template will be amplified by two different iPCR1 primers harboring two different indices. 13. We use an enzyme E2 that is compatible with the ligase buffer. Otherwise, purify the ligation reaction and set up a digest with the appropriate E2 buffer.

20

Jennifer C. Volz et al.

14. First run a small-scaled (one reaction) test PCR. Perform 5–20 parallel reactions per screen to ensure amplification of all insertion sites. No discrete band is expected as ring size varies due to distance of restriction site relative to insertion of the mutagen. 15. After the iPCR, you are dealing with very big molar amounts of amplicons. As NGS detects single molecules, there is a very high danger of contaminating other experiments with postPCR samples. Therefore, use different rooms, pipet sets, and all other equipment for post-PCR samples. With our primers, we often see bands of primer dimers running between 100 and 200 bp. Your purified samples intended for NGS submission must not contain these anymore. 16. We preferably use magnetic beads for sample purification as it removes primer dimers more efficiently than a gel extraction/ column kit. References 1. Hoepfner D, Helliwell SB, Sadlish H et al (2014) High-resolution chemical dissection of a model eukaryote reveals targets, pathways and gene functions. Microbiol Res 169:107–120. https://doi.org/10.1016/j. micres.2013.11.004 2. Forsburg SL (2001) The art and design of genetic screens: yeast. Nat Rev Genet 2:659–668. https://doi.org/10.1038/ 35088500 3. Hartwell LH, Culotti J, Pringle JR, Reid BJ (1974) Genetic control of the cell division cycle in yeast. Science 183:46–51. https:// doi.org/10.1126/science.183.4120.46 4. Pris J, Clement D, Kessous A, Colombies P (1980) Near haploid cell line in lymphoid blast crisis of ph1-positive chronic myeloid leukemia. Cancer Res 40:1354–1359 5. Brodeur G, Williams D, Look A et al (1981) Near-haploid acute lymphoblastic leukemia: a unique subgroup with a poor prognosis. Blood 58:14–19 6. Kotecki M (1999) Isolation and characterization of a near-haploid human cell line. Exp Cell Res 252:273–280. https://doi.org/10.1006/ excr.1999.4656 7. Carette JE, Guimaraes CP, Varadarajan M et al (2009) Haploid genetic screens in human cells identify host factors used by pathogens. Science 326:1231–1235. https://doi.org/10. 1126/science.1178955 8. Carette JE, Pruszak J, Varadarajan M et al (2010) Generation of iPSCs from cultured human malignant cells. Blood

115:4039–4042. https://doi.org/10.1182/ blood-2009-07-231845 9. Carette JE, Raaben M, Wong AC et al (2011) Ebola virus entry requires the cholesterol transporter Niemann-Pick C1. Nature 477:340–343. https://doi.org/10.1038/ nature10348 10. Carette JE, Guimaraes CP, Wuethrich I et al (2011) Global gene disruption in human cells to assign genes to phenotypes by deep sequencing. Nat Biotechnol 29:542–546. https://doi. org/10.1038/nbt.1857 11. Perrot V, Richerd S, Vale´ro M (1991) Transition from haploidy to diploidy. Nature 351:315–317. https://doi.org/10.1038/ 351315a0 12. Freed JJ, Mezger-Freed L (1970) Stable haploid cultured cell lines from frog embryos. Proc Natl Acad Sci U S A 65:337–344. https://doi. org/10.1073/pnas.65.2.337 13. Yi M, Hong N, Hong Y (2009) Generation of medaka fish haploid embryonic stem cells. Science 326:430–433. https://doi.org/10. 1126/science.1175151 14. Elling U, Taubenschmid J, Wirnsberger G et al (2011) Forward and reverse genetics through derivation of haploid mouse embryonic stem cells. Cell Stem Cell 9:563–574. https://doi. org/10.1016/j.stem.2011.10.012 15. Leeb M, Wutz A (2011) Derivation of haploid embryonic stem cells from mouse embryos. Nature 479:131–134. https://doi.org/10. 1038/nature10448

Haploid Genetic Screens 16. Li W, Shuai L, Wan H et al (2012) Androgenetic haploid embryonic stem cells produce live transgenic mice. Nature 490:407–411. https://doi.org/10.1038/nature11435 17. Yang H, Shi L, Wang BA et al (2012) Generation of genetically modified mice by oocyte injection of androgenetic haploid embryonic stem cells. Cell 149:605–617. https://doi. org/10.1016/j.cell.2012.04.002 18. Li W, Li X, Li T et al (2014) Genetic modification and screening in rat using haploid embryonic stem cells. Cell Stem Cell 14:404–414. https://doi.org/10.1016/j.stem.2013.11. 016 19. Yang H, Liu Z, Ma Y et al (2013) Generation of haploid embryonic stem cells from Macaca fascicularis monkey parthenotes. Cell Res 23:1187–1200. https://doi.org/10.1038/cr. 2013.93 20. Sagi I, Chia G, Golan-Lev T et al (2016) Derivation and differentiation of haploid human embryonic stem cells. Nature 532:107–111. https://doi.org/10.1038/nature17408 21. Zhong C, Zhang M, Yin Q et al (2016) Generation of human haploid embryonic stem cells from parthenogenetic embryos obtained by microsurgical removal of male pronucleus. Cell Res 26:743–746 22. Wang H, Zhang W, Yu J et al (2018) Genetic screening and multipotency in rhesus monkey haploid neural progenitor cells. Development 145. pii: dev160531. https://doi.org/10. 1242/dev.160531

21

23. Elling U, Wimmer RA, Leibbrandt A et al (2017) A reversible haploid mouse embryonic stem cell biobank resource for functional genomics. Nature 550:114–118. https://doi.org/ 10.1038/nature24027 24. Elling U, Penninger JM (2014) Genome wide functional genetics in haploid cells. FEBS Lett 588:2415. https://doi.org/10.1016/j.febslet. 2014.06.032 25. Schnu¨tgen F, Hansen J, De-Zolt S et al (2008) Enhanced gene trapping in mouse embryonic stem cells. Nucleic Acids Res 36:e133. https:// doi.org/10.1093/nar/gkn603 26. Horn M, Kroef V, Allmeroth K et al (2018) Unbiased compound-protein interface mapping and prediction of chemoresistance loci through forward genetics in haploid stem cells. Oncotarget. https://doi.org/10.18632/ oncotarget.24305 27. Schmidt M, Schwarzwaelder K, Bartholomae C et al (2007) High-resolution insertion-site analysis by linear amplification-mediated PCR (LAM-PCR). Nat Methods 4:1051–1057. https://doi.org/10.1038/nmeth1103 28. Potter CJ, Luo L (2010) Splinkerette PCR for mapping transposable elements in Drosophila. PLoS One 5:e10168. https://doi.org/10. 1371/journal.pone.0010168 29. Ochman H, Gerber AS, Hartl DL (1988) Genetic applications of an inverse polymerase chain reaction. Genetics 120:621–623

Chapter 2 Unbiased Forward Genetic Screening with Chemical Mutagenesis to Uncover Drug–Target Interactions Moritz Horn, Franziska Metge, and Martin S. Denzel Abstract The steadily increasing throughput in next-generation sequencing technologies is revolutionizing a number of fields in biology. One application requiring massive parallel sequencing is forward genetic screening based on chemical mutagenesis. Such screens interrogate the entire genome in an entirely unbiased fashion and can be applied to a number of research questions. CRISPR/Cas9-based screens, which are largely limited to a gene’s loss of function, have already been very successful in identifying drug targets and pathways related to the drug’s mode of action. By inducing single nucleotide changes using an alkylating reagent, it is possible to generate amino acid changes that perturb the interaction between a drug and its direct target, resulting in drug resistance. This chemogenomic approach combined with latest sequencing technologies allows deconvolution of drug targets and characterization of drug–target binding interfaces at amino acid resolution, therefore nicely complementing existing biochemical approaches. Here we describe a general protocol for a chemical mutagenesis-based forward genetic screen applicable for drug–target deconvolution. Key words Chemogenomics, Genetic screen, Unbiased chemical mutagenesis, Drug–target interaction site mapping, Amino acid resolution

1

Introduction Studying protein-small molecule interactions is of key importance to advance drug development strategies. Traditionally, biochemical and biophysical approaches have been successfully applied to unravel such interactions and to define binding interfaces [1]. However, most of these strategies require previous knowledge of interacting partners and thus come with a certain bias, are highly labor intensive, and rely on strong physical drug–target interactions. Moreover, these approaches also detect unspecific binding and do not necessarily reveal the functionally relevant interaction. More recently, such studies have been complemented by a variety of “chemogenomic” approaches.

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_2, © Springer Science+Business Media, LLC, part of Springer Nature 2019

23

24

Moritz Horn et al.

Chemogenomic approaches are based on “forward genetic screens” classically performed in model systems such as the fruit fly Drosophila melanogaster, the nematode Caenorhabditis elegans, the budding yeast Saccharomyces cerevisiae, and the house mouse Mus musculus [2–5]. Starting with a (randomly) mutagenized population, such screens can reveal genetic architectures underpinning a phenotype of interest including networks associated with drug resistance or sensitivity [6]. The most frequently applied techniques to perform genome-wide mutagenesis include gene-trap systems, transposons, or CRISPR/Cas9-mediated genome editing. While these methods allow the generation of highly complex mutagenized libraries, they do not target all genomic regions at equal efficiency and are largely limited to a gene’s loss of function, restricting such screens to nonessential genes. Finally, they fail to uncover protein/small molecule binding interfaces that involve only specific amino acids of the target protein. In contrast, alkylating agents such as N-ethyl-N-nitrosourea (ENU) modify DNA bases with lower “location bias” and result in single nucleotide variants (SNVs) upon replication [7]. This allows for the investigation of a broad range of functional consequences including loss-of-function, separation-of-function, and gain-of-function mutations probing literally every single amino acid [8, 9]. The alteration of individual amino acids enables highly specific protein changes, e.g., only affecting the active site or a binding pocket of a regulating molecule. Despite these advantages, the analysis of forward genetic screens using chemical mutagenesis has yet been challenging. While all other mutagenesis approaches are designed to include a marker that easily identifies the causative mutation site in positive screening hits, chemically induced alterations had to be mapped by labor-intensive methods. Only recently, with the rapid advancements in nucleic acid sequencing technologies, it has become possible to call causative mutations by parallel next-generation sequencing. Here we describe a forward genetic screening approach in mouse embryonic stem cells (mESC) based on chemical mutagenesis, selection by drug resistance, next-generation sequencing, and bioinformatic analysis that uncovers drug on- and off-targets at amino acid resolution [9]. Since drug resistance typically is a dominant phenotype, this approach can be applied to a variety of cellular systems and model organisms.

2

Materials

2.1 Full Medium for mESC

450 mL DMEM (Sigma, D1152). 75 mL FCS (Life Technologies).

Chemical Mutagenesis Screening

25

5 mL penicillin-streptomycin (Sigma, P0781). 5 mL nonessential amino acids (Sigma, M7145). 5 mL L-glutamine (Sigma, G7513). 5 mL sodium pyruvate (Sigma, S8636). 0.5 mL 2-mercaptoethanol (Merck) from a 0.35% (v/v) solution in PBS. ESGRO Leukemia Inhibitory Factor according to the manufacturer’s instructions (Millipore, ESG1107). 2.2 Buffer and Reagents

Phosphate-buffered saline (PBS) (Thermo Fisher). 0.5% Trypsin/EDTA (ThermoFisher)—dilute 1:10 in PBS. N-ethyl-N-nitrosourea (ENU) (Sigma). 66 mM disodium hydrogen phosphate (Na2HPO4*2H2O). 66 mM potassium dihydrogen phosphate (KH2PO4). M Sodium Hydroxide (NaOH). MG132 (Sigma, M7449).

2.3 Kits and Equipment

Cell Proliferation Kit II (XTT, Roche) or similar. Gentra Puregene Tissue Kit (Qiagen) or similar. SureSelectXT Mouse All Exon Kit or similar. High-flow chemical hood with a nutator mixer and a centrifuge for 15 mL Falcon tubes. Qubit (Thermo Fisher, optional). NanoDrop (Thermo Fisher). Agarose gel, running chamber, and gel documentation system.

3 3.1

Methods Cell Cultivation

Here, we exemplarily describe a chemical mutagenesis screen using mESC. They are highly suitable for this screening approach due to their rapid growth, ability to form single colonies, and their strong genomic integrity. mESC are cultivated on tissue culture-treated plates without a feeder layer. They are maintained at 10–80% confluence by regular passaging. Make sure to change the medium every day (latest every other day!) to prevent metabolic changes (glucose consumption!) and sporadic differentiation (see Note 1).

26

3.2

Moritz Horn et al.

ENU Mutagenesis

3.2.1 Safety Advice for Working with ENU

– ENU is highly genotoxic and should be handled only under a high-flow chemical hood. – The experimenter should be wearing safety gloves, lab coat, goggles, and a mask, particularly when working with ENU powder. – All solutions or items in touch with ENU need to be inactivated prior to disposal. Use 0.5 M NaOH or KOH for inactivation. – Central waste collection and waste incineration after ENU inactivation are recommended.

3.2.2 Prepare ENU Stock Solution

1. Prepare So¨rensen buffer with a pH of 6 by mixing 66 mM stock solutions of KH2PO4 and Na2HPO4 in a 1:8 ratio (control and adjust pH accordingly). 2. Dissolve 1 g ENU powder in 100 mL So¨rensen buffer under a chemical hood to obtain a 10 mg/mL ENU stock solution. 3. Store ENU aliquots at

3.2.3 Mutagenize a Cell Population

20  C.

1. Trypsinize cells, and prepare two 15 mL falcon tubes each containing about 50 million cells in 8 mL full medium. 2. Add 80 μL ENU stock to one vial (the other vial functions as non-mutagenized control) under a chemical hood, and incubate for 2 h at room temperature on a nutator mixer (see Note 2). 3. Wash cells 5 with 10 mL medium by centrifugation (3 min, 0.3 rcf). Collect supernatants in 50 mL Falcon tubes, and inactivate by adding 0.5 mM NaOH in a 1:1 ratio. 4. Transfer cell pellet to a fresh 15 mL Falcon containing 10 mL medium, and continue your work under a cell culture flow hood. The mutagenized and control cell pool can be frozen or directly used for a drug resistance screen (recommended) (see Note 3).

3.3 Drug Resistance Screen

The simplest application using the mutagenized cell library is a positive-selection screen using a toxic agent (described here). Further the screening procedure can be adjusted to other readouts, e.g., by using specific reporter systems combined with a fluorescence-activated cell sorting (FACS) readout [10].

3.3.1 Find the Adequate Drug Concentration

To separate nonresistant from resistant cells, it is key to find a suitable drug concentration using a cell viability assay. Ideally, one should select the lowest concentration that efficiently kills non-mutagenized wild-type (wt) cells. 1. Plate 2500 cells per well to a 96-well plate, and treat with a wide range of concentrations of your drug of interest. Start

Chemical Mutagenesis Screening

27

treatment 24 h post cell plating, and leave an untreated control. Performing at least duplicates is highly recommended. 2. Perform a cell viability assay (e.g., XTT or Alamar blue) according to the manufacturer’s instruction after 72 h of treatment (see Note 4). 3. Pick a concentration that kills >70% of the wt cells for the mutagenesis screen. 3.3.2 Test ENU Mutagenesis

It is recommended to control the mutagenesis protocol with a wellestablished drug resistance screen to validate proper ENU function. We therefore recommend to perform a small-scale MG132 resistance screen. Significant MG132 resistance is only caused by specific PSMB5 mutations in the MG132-PSMB5 binding interface [9]. This limited target space allows to estimate ENU efficiency by the ratio of resistant colonies over total mutagenized cells. 1. Set up 1  15 cm plates carrying about three million ENU mutagenized cells. When testing different ENU concentrations, set up one plate per condition. Include one control plate with about three million non-mutagenized cells. 2. For selection, apply MG132 treatment (for ESCs 0.5–0.75 μM) 24 h post mutagenesis by changing the medium (see Note 5). 3. Observe progression of cell death on a daily basis. Seventy-two h post selection, >99% of cells should appear dead. Adjust MG132 concentration if necessary. 4. Change medium containing MG132 every 48 h for the first week and then only once per week. 5. Maintain the screen for >14 days until small colonies emerge (visible by eye). 6. Compare and record number and appearance of colonies on the ENU-treated vs. control plates. We recommend an ENU concentration resulting in about 5–10 MG132-resistant colonies per million mutagenized cells, while 0–1 colonies per million cells should emerge in the non-mutagenized controls.

3.3.3 Select Resistant Mutants in a Large-Scale Screen

1. Set up 10  15 cm plates carrying three million ENU mutagenized cells, and include 2  15 cm non-mutagenized control plates (see Note 6). 2. Apply your drug of interest at the selected concentration 24 h post mutagenesis by changing the medium. 3. Observe the progression of cell death on a daily basis. Depending on the drug efficiency, >99% of cells should appear dead 72 h post selection. Adjust drug concentration if necessary.

28

Moritz Horn et al.

4. Change medium including the drug of interest every 48 h for the first week and then only once per week. 5. Maintain the screen for >14 days until small colonies emerge (visible by eye). 6. Compare and record number and appearance of colonies on the ENU-treated vs. control plates (see Note 7). 3.3.4 Pick and Validate Resistant Colonies

1. Aspirate the medium from plates containing drug-resistant colonies. Do not wash the plate with PBS to prevent loss of potentially loose colonies. 2. Aspirate the colony with the plastic tip of a 20 μL (or 100 μL) pipette and transfer it to 50 μL 0.5% trypsin/EDTA (in a 96-well plate or small tube). 3. Incubate at 37  C for 5 min and resuspend cell pellet with 100 μL medium. 4. Transfer cell suspension to a 24-well plate, and expand cells for 2–3 days to gain a sufficient number for a viability assay and further maintenance (>10,000 cells total). 5. Perform a viability assay of your choice (e.g., XTT) with wt cells and all cell clones isolated from your screen. Use a drug concentration that killed between 50 and 80% of your wt cells. 6. Maintain resistant colonies to uncover resistance-causing mutations by whole exome sequencing.

3.4 Identification of Causative Mutations 3.4.1 DNA Preparation and Whole Exome Sequencing

1. Trypsinize cells from the confirmed resistance colonies, and transfer about one million cells to a 1.5 mL Eppendorf tube in PBS (see Note 8). 2. Centrifuge (3 min, 0.3 g) and remove the supernatant (leave 20 μL). 3. Extract genomic DNA using the Gentra Pure Tissue Kit (Qiagen) according to the manufacturer’s instruction. 4. Assess DNA quantity and quality using the Qubit system (Thermo fisher) and by running DNA samples on a 1% agarose gel. Total DNA quantity should exceed 1 μg, appear as one high molecular weight band in the agarose gel, and show neither protein nor detergent contamination (OD 260/280 and 260/230 > 1.8). 5. Prepare Exome library using the SureSelect XT Mouse All Exon Kit according to the manufacturer’s protocol, and run the whole exome sequencing (see Note 9).

3.4.2 Data Analysis and Identification of Resistance-Causing Mutations

The following describes how the raw exome sequencing data is processed in order to identify single nucleotide variants potentially responsible for the resistance of the clone:

Chemical Mutagenesis Screening

29

1. Align the raw reads to the reference genome (e.g., with bwa-mem [11]). 2. Further process the raw alignment file for optimal variant calling following the GATK Best Practice pipeline (https://soft ware.broadinstitute.org/gatk/). 3. A pileup is generated from the processed alignment file. Keep point mutations only. 4. Cross-reference these potential variants with a wt sequence without mutagenesis to subtract background variants from the parental cell line. 5. Annotate the remaining variants using snpEff [12]. 6. Mutations with a high or moderate effect should be further filtered for properties of special interest (e.g., number of total reads >10, minor allele frequency >40% to identify also heterozygous mutations; see Note 10). The remaining mutations are candidate mutations potentially causing the clone’s drug resistance. To enrich true causal variants, we combine candidate lists from independent clones to detect allelism: genes carrying independent variants are strong candidates. We recommend to normalize the number of hits per gene to its cDNA size to exclude false positives with long coding sequences and to prioritize your candidates. 3.4.3 Validate Resistance-Causing Mutations

1. Select candidate genes and their specific mutations from your results lists. 2. Use CRISPR/Cas9 genome editing to introduce the single base pair changes obtained from the mutagenesis screen in a wt cell. Use a standard CRISPR approach with a singlestranded DNA repair template [13] or more recent single base pair editing technologies that do not introduce DNA double-strand breaks [14, 15]. 3. Repeat viability assay with your compound of interest comparing wt cells with your engineered candidate point mutants side by side. When two independent alleles in the same gene validate in the viability assay, one can infer that additional hits in that gene also cause drug resistance. All amino acid substitutions identified in that verified candidate gene should then be modelled into existing molecular structures or predicted structure models to gain insights into the potential drug–target interaction site (see Note 11).

30

4

Moritz Horn et al.

Notes 1. We use mouse embryonic stem cells isolated from F1 females from C57BL/6  129F1 crosses [16]. Any other stem cell line growing in feeder-free conditions might be suitable. 2. ENU concentration might need adjustment according to the cell type of your choice. For an initial test screen, we recommend to test 2–3 different ENU concentrations in the range of 0.01–0.25 mg/mL. 3. Please be aware that passaging of mutagenized cells leads to mixing of independent (and potentially resistant) clones. This might result in picking clones that stem from the same mutagenized cell, which hampers analysis. We thus recommend to directly plate cells after mutagenesis and proceed to selection without cell passaging. Alternatively, mutagenized cell pools can be frozen immediately after mutagenesis and thawed 1 day prior to selection. 4. If your drug of interest induces very slow cell death, cell viability can also be assessed at a later time point. Reduce plated number of cells accordingly. 5. It is recommended to start drug treatment 24 h post mutagenesis to allow one cell proliferation (required to establish a mutation from the alkylated base). 6. If your drug of interest induces very slow cell death, please consider to reduce plated cell number to about one million per plate. 7. The obtained number of resistant colonies allows estimation of the size of the “genomic target space.” A large number of colonies might indicate that, and the entire pathway is responsible for toxicity. 8. It is recommended to also sequence your wt cell population as a reference for subsequent bioinformatic analysis. 9. Whole exome sequencing was performed on an Illumina HiSeq X-ten sequencing platform using the SureSelect XT Mouse All Exon Kit for library preparation. Please note that it is possible to multiplex samples without barcoding. We successfully pooled four resistant clones of a diploid cell line resulting in 12.5% mutant reads per mutation (heterozygous), which can still be distinguished from sequencing noise. Bioinformatic analysis needs to be adjusted accordingly. 10. Filter criteria need to be adjusted to quality and depth of sequencing. This is particularly necessary when pooling several DNA clones prior to sequencing in a multiplex approach.

Chemical Mutagenesis Screening

31

11. Importantly, not every identified resistance locus necessarily constitutes the direct drug target but certainly is directly or indirectly involved in the drug’s mode of action. Direct interaction should be confirmed by biochemical methods. References 1. McFedries A, Schwaid A, Saghatelian A (2013) Methods for the elucidation of protein-small molecule interactions. Chem Biol 20:667–673 2. Jorgensen EM, Mango SE (2002) The art and design of genetic screens: caenorhabditis elegans. Nat Rev Genet 3:356–369 3. Forsburg SL (2001) The art and design of genetic screens: yeast. Nat Rev Genet 2:659–668 4. St Johnston D (2002) The art and design of genetic screens: Drosophila melanogaster. Nat Rev Genet 3:176–188 5. Kile BT, Hilton DJ (2005) The art and design of genetic screens: mouse. Nat Rev Genet 6:557–567 6. Hoepfner D, Helliwell SB, Sadlish H et al (2014) High-resolution chemical dissection of a model eukaryote reveals targets, pathways and gene functions. Microbiol Res 169:107–120 7. Beranek DT (1990) Distribution of methyl and ethyl adducts following alkylation with monofunctional alkylating agents. Mutat Res 231:11–30 8. Denzel MS, Storm NJ, Gutschmidt A et al (2014) Hexosamine pathway metabolites enhance protein quality control and prolong life. Cell 156:1167–1178 9. Horn M, Kroef V, Allmeroth K et al (2018) Unbiased compound-protein interface

mapping and prediction of chemoresistance loci through forward genetics in haploid stem cells. Oncotarget 9:9838–9851 10. Cheloufi S, Elling U, Hopfgartner B et al (2015) The histone chaperone CAF-1 safeguards somatic cell identity. Nature 528:218–224 11. Li H (2013) Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM 12. Cingolani P, Platts A, Wang Le L et al (2012) A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff. Fly 6:80–92 13. Ran FA, Hsu PD, Wright J et al (2013) Genome engineering using the CRISPR-Cas9 system. Nat Protoc 8:2281–2308 14. Gaudelli NM, Komor AC, Rees HA et al (2017) Programmable base editing of AlT to GlC in genomic DNA without DNA cleavage. Nature 551:464–471 15. Komor AC, Kim YB, Packer MS et al (2016) Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage. Nature 533:420–424 16. Elling U, Taubenschmid J, Wirnsberger G et al (2011) Forward and reverse genetics through derivation of haploid mouse embryonic stem cells. Cell Stem Cell 9:563–574

Chapter 3 Integration of RNAi and Small Molecule Screens to Identify Targets for Drug Development Konstantinos Drosopoulos and Spiros Linardopoulos Abstract Cellular models for siRNA and small molecule high-throughput screening have been widely used in the last decade to identify targets for drug discovery. As an example, we present a twofold readout approach based on cell viability and multipolar phenotype. To maximize the discovery of potential targets and at the same time reduce the number of false positives in our dataset, we have combined focused and rationally designed custom siRNA libraries with small molecule inhibitor libraries. Here we describe a cellular model for centrosome amplification as an example of how to design and perform a multiple readout/multiple screening strategy. Key words High-throughput screening, Centrosome amplification, Centrosome clustering, Highcontent screening, siRNA screening, Target discovery

1

Introduction Despite intensified efforts to identify new targets for drug development in cancer, in the last decades, only a very small fraction reaches the preclinical development phase. The most common complications reported are drug selectivity and lack of understanding of the underlying mechanisms and biomarkers. The recent advances in RNA interference (RNAi) and CRISPR-Cas9 technologies have permitted a wide adoption of high-throughput approaches for the discovery of new targets within their natural context. However, these strategies are still beset by selectivity problems due to offtarget effects, especially in mammalian cells [1, 2]. Network and protein connectivity analyses have contributed to shedding light on the potential therapeutic mechanism of a particular target, but these are still depended on high-quality input data in order to be successfully applied [3]. Several strategies have been suggested for improving the quality of data from high-throughput screens, including the use of isogenic cell lines (cell lines of genetically

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_3, © Springer Science+Business Media, LLC, part of Springer Nature 2019

33

34

Konstantinos Drosopoulos and Spiros Linardopoulos

identical background), use of multiple RNAi per gene, and multiplexing viability data with high-throughput microscopy data [4]. In order to improve the discovery of verifiable targets and facilitate identification of false positives, as well as the understanding of the underlying mechanisms, we use a twofold readout approach, also limiting our assay to a specific phase of the cell cycle where possible. In addition, in order to further reduce screening artifacts and false positives, we utilize isogenic cell lines with or without centrosome amplification that are screened under identical conditions and siRNA transfection efficiency. The centrosomes are microtubule-organizing centers which play an essential role in organizing the mitotic spindle and maintaining spindle bipolarity in mammalian cells [5]. Centrosome amplification is one of the most common abnormalities in cancer cells. In order to avoid the deleterious effects of multipolar mitotic spindles, cancer cells cluster supernumerary centrosomes in a bipolar fashion during mitosis, a process which can be exploited to selectively kill cancer cells [6]. The cellular model that we use consists of chromosomally stable human colon cancer DLD1 diploid (2N) and tetraploid (4N) cells, along with tetraploid centrosome-amplified (4NCA) cells. In order to generate 4NCA cells, we use dihydrocytochalasin B (DCB) to transiently block cytokinesis and induce tetraploidization and centrosome amplification in DLD1 cells. Centrosome amplification in 4NCA cells is transient, and therefore they need to be generated from 2N cells for each run of the assay. We then use this model to screen a custom siRNA library representing all known microtubule-binding proteins with the rationale that centrosomes, which are microtubuleassociated organelles, depend on microtubules and their associated proteins for their motility. In addition, we use a kinase-related siRNA library and a small molecule kinase inhibitor library in order to identify the regulatory networks that lie upstream of the microtubule-associated effectors. The siRNA screening is performed in duplicate as shown in Fig. 1. The first replica is used for viability readout and the second for phenotypic readout where we look at spindle structure by using a high-content screening platform. The small molecule library screening is also performed in duplicate with the first replica used for viability readout. In contrast to the siRNA libraries, small molecule libraries give us the opportunity to carefully time the treatment to include mostly one phase of the cell cycle, thus reducing the possibility for “noise” introduced by off-target inhibition of kinases that might be important in other phases of the cell cycle. For this reason, the second replica, which we use for the phenotypic readout, cells are synchronized in mitosis by treatment with the APC/C inhibitor pro-TAME [7] for 3.5 h and then placed at 4  C for 10 min to induce a reversible disorganization of the mitotic spindle. Finally, the cells are treated with the compound library in

Identification of Drug Targets for Centrosome Amplified Cells

NT

Negative/non-targeting Control

TC

Positive/Transfection Control

SC

Selective control

TC

NT

NT

SC

TC

NT

NT

SC

SC

NT

NT

TC

TC

NT

NT

SC

35

Fig. 1 Example 96-well plate with controls on columns 1 and 2. The rest of the plate is occupied by the samples (siRNA or small molecule). siRNA and small molecule libraries often come with columns 1 and 2 or 1 and 12 empty for the purpose of leaving space for the desired controls. Positioning the controls on columns 1 and 2 is preferable than the 1 and 12 option. Ideally, the controls should be dispersed throughout the plate if possible

the presence of MG132 for 3 h, to limit the effects of the treatment only in mitosis. Treatment with pro-TAME arrests the cells in metaphase without inducing mitotic spindle defects in the short term [7]. Incubation of the plates at 4  C aims to emulate the cells entering mitosis in the presence of the compounds in a uniform manner. MG132 is a proteasome inhibitor that prevents the degradation of spindle assembly checkpoint proteins and cyclin B; therefore, it helps to maintain the metaphase arrest. Metaphase is the optimal phase of mitosis where the number of spindle poles/cell can be quantified by automated analysis. Using this approach, we identify a number of hit genes that selectively kill cells with centrosome amplification validating at the same time both the underlying mechanism and the regulatory signaling involved in this process.

2

Materials The nature of this approach requires cell line-specific optimization and setting up of the assay. There is a variety of commercial lipids and siRNA controls that need to be tested for suitability for a particular cell line/cell model. It is desirable to achieve less than 10% toxicity of the lipid alone relative to growth media alone, less than 10% toxicity of the nontargeting (sequence does not match any known human mRNA) siRNA negative control relative to lipid alone, and more than 90% toxicity of the transfection efficiency control siTOX (highly toxic siRNA pool that will kill every transfected cell) relative to nontargeting siRNA negative control. In addition, cell plating should be optimized so that at the end of

36

Konstantinos Drosopoulos and Spiros Linardopoulos

the assay, cells are no more than 90–95% confluent for the viability assay and no more than 70–80% confluent for the phenotypic assay. It is important that every screen is set up individually by the person who is going to perform the screening since clonal variation of cell lines, available equipment, normal variation of different batches of transfection lipid, and personal technique can have significant effects in a high-throughput screen even if an exact protocol is followed. 2.1 Generation of Isogenic Populations

1. DLD1 colon adenocarcinoma cells grown in DMEM (41966029, Thermo Fisher) supplemented with 10% FBS. 2. Dihydrocytochalasin B (DCB, 250225, Merck Millipore), an inhibitor of actin polymerization, required to promote cytokinesis failure. 3. Hoechst 33342 (H3570, Thermo Fisher) for live cell sorting and isolation of diploid/tetraploid populations.

2.2

siRNA Screening

1. Transfection reagent: DharmaFECT 2 (T-2002, Dharmacon). 2. Opti-MEM I (31985, Thermo Fisher). 3. siRNA controls (Dharmacon): nontargeting siRNA Pool (D-001206-13), siTOX transfection control (D-001500-01), and siKIFC1 (M-004958-02). 4. Dharmacon kinase-related siRNA library or any custom/catalogue siRNA library.

2.3 Small Molecule Library Screening

1. DMSO for resuspension and dilution of small molecule library and for vehicle control. 2. Griseofulvin (G4753, Sigma): Used as a positive control (5 μM) that selectively kills cells with centrosome amplification. 3. pro-TAME (I-440, Boston Biochem): Used at 8 μM for synchronizing the cells in metaphase. Pro-TAME is a prodrug activated by intracellular esterases; optimal concentration should be determined for each cell line. 4. MG132 (M7449, Sigma): Used at 8 μM to maintain metaphase arrest during treatment. Most cell lines will maintain metaphase arrest for the duration of the assay at this concentration.

2.4 Viability and Phenotypic Readout

1. Opaque white-wall, clear-bottom 96-well tissue culture plates to be used with a luminescent cell viability reagent suitable for high-throughput screening assays (CLS3610-48EA, Corning). 2. CellTiter-Glo Luminescent Cell Viability Assay (G7572, Promega).

Identification of Drug Targets for Centrosome Amplified Cells

37

3. Black-wall, clear-bottom 96-well tissue culture assay plates to be read by a suitable high-content screening platform (6005182, PerkinElmer). 4. Ice-cold methanol. 5. Anti-Aurora-A (see Note 1) monoclonal antibody (610939, BD) and anti-histone H3 pS10 (pHH3) polyclonal antibody (ab47297, Abcam). 6. Alexa Fluor 488 goat anti-mouse and Alexa Fluor 555 goat anti-rabbit (both from Life Technologies). 7. Antibody solution: 1.5% FBS in PBS. 8. DAPI solution: 1:5000 DAPI (10 mg/mL stock, 40043, Biotium) in PBS. 9. Black adhesive film to seal the plates before reading. 10. Suitable HCS platform such as the IN Cell Analyzer (GE Life Sciences), Opera/Operetta (PerkinElmer), or ImageXpress (Molecular Devices).

3

Methods

3.1 Generation of Isogenic Cell Lines

1. Treat DLD1 diploid cells with normal centrosome number for 24 h with 2 μM DCB, and release for 96 h. 2. Prepare cell suspension of 500,000 cells/mL, and incubate with Hoechst 33342 (10 mg/mL stock) in PBS 1:5000 at 37  C for 20 min. 3. Cell sort populations according to cell cycle profile. Isolate diploid populations from diploid G1 and tetraploid populations from tetraploid G2 (see Note 2). 4. Grow isolated populations for 1 week, and verify DNA content by comparing isolated tetraploid populations to the original DLD1 cells. Repeat cell sorting if necessary.

3.2

siRNA Screening

All reagents must be at room temperature (RT). 1. Split, pellet, and resuspend 2 N, 4 N, and 4NCA cells at appropriate density (2000 cells/well for viability and 9000 cells/well for phenotypic) in growth media (DMEM) without antibiotics. 2. Add 80 μL PBS in the space between the wells of the 96-well assay plates (see Note 3). 3. Distribute library and control siRNAs (Fig. 1) in V-bottom 96-well plates (one siRNA pool/well). Add appropriate amount of DharmaFECT 2 in Opti-MEM (final concentration 0.1 μL per sample) using a 12-channel pipette and precision tips.

38

Konstantinos Drosopoulos and Spiros Linardopoulos

4. Gently pipette up and down 3–4 times, and immediately distribute to flat-bottom 96-well plates. 5. Place the lid on the plate, and incubate for 30 min at RT to allow the formation of siRNA-lipid complexes. 6. Add cell suspension on complexes by a single swift pipetting step, and shake on an orbital shaker for 2–3 s at 400 rpm to equally distribute the cells throughout the well. 7. Place the plates in the incubator, avoiding stacking if possible (see Note 3). 8. After 5–7 h, carefully add 200 μL of growth media, taking caution not to disturb the cells. This is done to minimize lipid toxicity. 9. (Optional) Next morning, remove half of the media in the plates using a multichannel pipette, and carefully add an equal amount of fresh media to further dilute the lipid. This will help to further reduce lipid toxicity effects and at the same time minimize disturbance of cells that will likely happen if the full amount of growth media is replaced. 10. For the phenotypic readout, 48 h after the transfection, remove media, and fix cells adding 150 μL of ice-cold methanol. 11. Incubate for 20 min at

20  C.

12. Leave for 2 min at RT before removing methanol, and immediately add 150 μL PBS to rehydrate the cells. 13. Remove PBS, and add 85 μL of antibody solution with AuroraA and pHH3 antibodies at 1:1000. 14. Incubate for 16–24 h at 4  C. 15. Wash 2  2 min with PBS taking care not to leave the cells dry for more than 30 s maximum. 16. Incubate with fluorescent secondary antibodies at 1:1000 and DAPI 1:5000 for 1 h at RT. 17. Wash 2  2 min with PBS. 18. Seal the plate with black adhesive film, and acquire images using a high-content platform. 19. For the viability assay, 96 h after the transfection, remove growth media by flicking the plate sideways above a sink, and add 100 μL of DMEM/Cell Titer Glo 1:1 solution. 20. Mix for 2–3 min on an orbital shaker and incubate for 20 min at RT. 21. Read using a luminescence plate reader.

Identification of Drug Targets for Centrosome Amplified Cells

3.3 Small Molecule Library Screening

39

1. Plate 12,500 4 N cells and 16,000 4NCA cells in black 96-well assay plates, one replica per concentration of the library (see Note 4). It is important that the cells are evenly dispersed within the wells. 2. Next day, aspirate media with a multichannel aspirator, and add 75 μL of 8 uM pro-TAME in each well with a 12-channel pipette. Aspirate and treat cells row by row to avoid drying out of cells. Incubate assay plates at 37  C, 5% CO2 for 3.5 h. 3. During the incubation, prepare 12 μM MG132 in DMEM for the serial dilutions of compounds at 1.5 the final concentration and appropriate DMSO controls. Distribute 350 μL/well in a standard 96-well plate in the same well positions as in the assay plates. Place the plate at 37  C. 4. Take the assay plates from the incubator, and place them at 4  C for 10 min. 5. In the hood at RT, using a 12-channel pipette, slowly (dropwise to avoid detachment of mitotic cells) add 150 μL from the serial dilution plates to the corresponding positions on the assay plates. Incubate at 37  C, 5% CO2 for 3 h. 6. Continue from Subheading 3.2, step 10 with the difference that the phenotypic replicas are treated for 3 h instead of 48 h.

3.4 Analysis and Interpretation of the Results

1. For the analysis of viability data, we use the MAD method selecting over one standard deviation [8]. 2. For the analysis of phenotypic results, we consider hits the siRNAs or small molecules that induce a multipolar spindle phenotype at a frequency which is at least two times over the one observed in the control-treated centrosome-amplified cells and at the same time do not induce the multipolar spindle phenotype in control cells (Fig. 2). 3. The common hits between the two different readouts represent our high confidence hits, which we validate further for their suitability to be used as targets for drug development for which both the target population and the underlying mechanisms involved are known. 4. Confirmed hits from the small molecule library serve three purposes: to identify tools that can be used at later stages of drug development, for determining the target population, and for providing potential models for structure–activity relationship (SAR). In addition, small molecules have the advantage of being suitable for dose-response experiments that are helpful in validating borderline hits from the kinome siRNA libraries. 5. Image acquisition and analysis can be performed with most major HCS systems. For imaging, 20 long working distance

40

Konstantinos Drosopoulos and Spiros Linardopoulos

Fig. 2 Example images acquired with IN Cell Analyzer 3000 (GE Healthcare). Control cells are 4 N DLD1 cells with no centrosome amplification; target cells are 4 N DLD1 cells with centrosome amplification. First set shows cells transfected with siControl nontargeting siRNA (negative), second set shows cells transfected with siKIFC1 (hit), and the third set shows cells transfected with siESPL1 (false positive) which is known to cause cytokinesis failure when knocked down. Note that in the first set, no multipolar spindles are observed; in the second set, multipolar spindles are observed only in the target cells; and in the third set, multipolar spindles are observed on both control and target cells

objective provides sufficient resolution for downstream analysis either with the respective proprietary software for each platform or with third-party software (e.g., CellProfiler). In principle: – Use the DAPI signal for the initial segmentation (“find nuclei”); this is optional, but it helps to quantify toxicity. – Select pHH3-positive nuclei (“find mitotic”) for further analysis. – Expand the borders of the pHH3 signal by 9–10 pixels to make sure the mitotic spindle (Aurora-A signal) is included in the region of interest (“find cytoplasm”). Cytoplasm border can also be determined from the Aurora-A signal. However, in general, outlining the cytoplasm border with

Identification of Drug Targets for Centrosome Amplified Cells

41

the pHH3 signal is more robust, especially in confluent samples. – Count the number of spindle poles/pHH3-positive nucleus. – Export single-cell data in CSV format, and calculate in Excel the percentage of multipolar mitoses relative to all mitoses (bipolar and multipolar).

4

Notes 1. Aurora-A staining has the advantage of localizing at the spindle poles (in contrast to α-tubulin staining that decorates the whole spindle) and at the same time producing a large enough area of signal that can be easily separated from background (in contrast to γ-tubulin or pericentrin staining). This protocol can be applied to any cell line as long as they express Aurora-A. Also, it allows visualization of both spindle structure and centrosomes with a single antibody. 2. Do not use more than 250,000 cell/mL as it will make it harder to separate the different phases of the cell cycle. Also, avoid keeping the cells with Hoechst for more than 3 h to avoid toxicity. The procedure will give stable diploid and tetraploid populations due the tendency of the cells that do not originally present centrosome amplification to reduce the centrosome number back to normal over time but maintaining 4 N DNA content. Cells with 4 N content and no centrosome amplification are used as a control cell line for the purpose of screening. 4 N cells with centrosome amplification are generated by 24 h incubation with DCB followed by a 24 h release. These cells will maintain abnormal centrosome numbers for a period of 6–10 days during which they can be used for screening. 3. When the plates are placed in the incubator, temperature and evaporation gradients start to form inside the plates which result in positional biases (edge effects) that introduce noise in the final data. The most affected samples are the ones in the periphery of the plate. Filling the spaces between the wells with PBS reduces the edge effect. Stacking the plates inside the incubator will also influence the evaporation gradient inside the plate, especially when the top plate of a stack is compared to the bottom plate of a stack. If the plates must be stacked, then the plates that are to be compared directly should be in the same position on different stacks. 4. Different small molecules have different efficacies. A very potent inhibitor can inhibit its target at nanomolar concentrations and start being toxic at micromolar concentrations.

42

Konstantinos Drosopoulos and Spiros Linardopoulos

Similarly, a less potent inhibitor might only work at micromolar concentrations. When screening small molecule libraries, it is best to do it at three different concentrations minimum (e.g., 100 nM, 1 μM, 10 μM) in order to maximize the production of useful data. References 1. Westwick JK, Lamerdin JE (2011) Improving drug discovery with contextual assays and cellular systems analysis. Methods Mol Biol 756:61–73 2. Schaefer KA, Wu WH, Colgan DF et al (2017) Unexpected mutations after CRISPR-Cas9 editing in vivo. Nat Methods 14:547–548 3. Dezso Z, Nikolsky Y, Nikolskaya T et al (2009) Identifying disease-specific genes based on their topological significance in protein networks. BMC Syst Biol 23:3–36 4. Ashworth A, Bernards R (2010) Using functional genetics to understand breast cancer biology. Cold Spring Harb Perspect Biol 2:a003327 5. Nigg EA, Stearns T (2011) The centrosome cycle: centriole biogenesis, duplication and

inherent asymmetries. Nat Cell Biol 13:1154–1160 6. Kwon M, Godinho SA, Chandhok NS et al (2008) Mechanisms to suppress multipolar divisions in cancer cells with extra centrosomes. Genes Dev 22:2189–2203 7. Zeng X, Sigoillot F, Gaur S et al (2010) Pharmacologic inhibition of the anaphase-promoting complex induces a spindle checkpointdependent mitotic arrest in the absence of spindle damage. Cancer Cell 4:382–395 8. Zhang XD, Yang XC, Chung N et al (2006) Robust statistical methods for hit selection in RNA interference high-throughput screening experiments. Pharmacogenomics 7:299–309

Chapter 4 CellProfiler and KNIME: Open-Source Tools for High-Content Screening Martin Sto¨ter, Antje Janosch, Rico Barsacchi, and Marc Bickle Abstract High-content screening (HCS) has established itself in the world of the pharmaceutical industry as an essential tool for drug discovery and drug development. HCS is currently starting to enter the academic world and might become a widely used technology. Given the diversity of problems tackled in academic research, HCS could experience some profound changes in the future, mainly with more imaging modalities and smart microscopes being developed. One of the limitations in the establishment of HCS in academia is flexibility and cost. Flexibility is important to be able to adapt the HCS setup to accommodate the multiple different assays typical of academia. Many cost factors cannot be avoided, but the costs of the software packages necessary to analyze large datasets can be reduced by using open-source software. We present and discuss the open-source software CellProfiler for image analysis and KNIME for data analysis and data mining that provide software solutions, which increase flexibility and keep costs low. Key words High-content screening, Image processing, Statistics, Open source, CellProfiler, KNIME, Distributed computing

1

Introduction In the 1960s, digital image processing was first applied at the Jet Propulsion Laboratory in California to enhance images from the lunar surface sent back by the Ranger 7 moon probe [1, 2]. In the 1970s, digital image processing became more widely applied mainly in fields such as medical imaging [3] or forensic science [4]. In the late 1980s, the pharmaceutical industry developed methods to increase the number of natural extract to screen for biological activity leading to the field of high-throughput screening (HTS) [5–7]. By the end of the 1990s, HTS and imaging were brought together, and high-content screening (HCS, also called highcontent analysis (HCA) or high-content imaging (HCI)) was born [8]. The motivation for the pharmaceutical industry to use the analytical power of microscopy was to obtain early in the drug development process information about compound action in a

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_4, © Springer Science+Business Media, LLC, part of Springer Nature 2019

43

44

Martin Sto¨ter et al.

cellular system. Typically, validated hits emerging from primary screening campaigns are tested in cell lines to assess gross toxicity and membrane permeability and specificity. In this secondary screening phase, many primary hit compounds need to be analyzed, and an increased throughput of microscopy analysis was desired. Therefore, automated microscopy coupled to automated image analysis was developed to obtain stable, user-friendly instruments by companies such as Cellomics. As HCS was primarily a technology used by the pharmaceutical industry and only few academic laboratories were equipped with automated microscopes, the hardware and software providers were mainly dealing with industrial clients with access to large budgets compared to academia. The market was limited and the development costs were high. All these factors contributed to the high costs for instruments, software, maintenance contracts, and yearly licensing fees. Consequently, access to the technology was limited to institutions with the required budget to maintain an HCS infrastructure. In the early 2000s, academic institutions started founding screening centers mainly due to two events. First, with the availability of several sequenced genomes and the discovery of RNA interference (RNAi), the systematic analysis of biological pathways by gene silencing became feasible. Second, academic institutions became more involved in biomedical research and began screening for novel drugs, especially since the launch of the NIH road map in 2004 (http://commonfund.nih.gov/aboutroadmap.aspx) [9]. Some of the newly created screening centers were equipped with automated imagers, and HCS entered the academic field. In general, these academic screening centers were well funded and could afford the prices demanded by vendors, perpetuating the tradition of cost-intensive screening. In recent years, R&D budgets in many pharmaceutical companies have been reduced, and many countries had to cut their research budgets because of the global financial crisis [10–13]. Concomitantly, the demand for screening small chemical compounds and RNAi libraries, especially in Europe, has been growing in academia. EU-OPENSCREEN (http://www.eu-openscreen.eu) was recently funded at the European level to provide screening capacity for researchers, and high-content screening services are offered. These counteracting trends have led academia to become an important customer for HCS providers. This trend might be further accentuated in coming years, as HCS has the potential to become a standard research tool and not solely a screening tool. There are several reasons why we believe that automated microscopy will be a widely adopted technology in normal research laboratories in the future. Firstly, many common biological experiments would benefit from the exploration of hundreds of experimental conditions to fully analyze a biological question. For instance,

Open Source Tools for HCS

45

combinations for optimizing antibody stainings or transfection conditions and time course experiments with several perturbing agents (i.e., drugs, viruses, growth factors) at various concentrations can easily fill numerous multiwell plates. Often the parameter space is not fully explored, simply because the manual acquisition of images and manual image analysis would be too time-consuming. The automation of this process with high-content imagers solves that problem and allows more thorough investigation of experimental conditions. Secondly, the automation of data acquisition and data analysis eliminates human bias. The large amount of data generated does not allow visual inspection but requires automated image analysis and quantification of the data. Thus, automated microscopy tends to improve results as they are based on larger datasets that underwent rigorous statistical analysis rather than biased human interpretation of sparse data. Thirdly, the miniaturization of assays results in significant time and cost savings. For HCS to be fully accepted by academia, several conditions will need to be fulfilled. Current HCS instruments are closed black boxes, and their expensive maintenance contracts do not allow any hardware or software modifications for adaptation to the diverse needs of academic research. Academic research is typically more diversified than pharmaceutical industry research, and the instruments need to be more customizable than they are now. In addition, the image and data formats need to be accessible and open. In academia, data is shared between collaborators and is analyzed with various, partly custom-made software. Therefore, the data needs to be accessible and open. Lastly, the yearly costs of maintenance contracts and licenses are particularly difficult to finance in academic research that relies heavily on grants. Grants typically do not cover licensing costs, or if they do, when the grant runs out, new sources of funding must be found. In reality, those costs must generally be covered by institutional funds. In order to reduce costs and increase the flexibility of HCS readers, it is to be expected that academic laboratories will assemble automated microscopes themselves. Hopefully this will be done with the open-source hardware (OSHW) model, allowing other laboratories to reproduce the design (http://www.openhardware. org) [14, 15]. Another expensive factor of HCS is software. All HCS instruments have onboard tools for image analysis and various levels of sophistication for data analysis and viewing. Often image and data analysis is done after the acquisition of the screening plates in order not to slow down the microscope. Vendors therefore offer offline licenses increasing the cost of software further. The image analysis software installed on instruments is often restricted in functionality to deliver easy-to-use software, although some vendors have expert mode tools allowing accessing many image analysis algorithms. For the diverse applications in academia, the push-button applications

46

Martin Sto¨ter et al.

found on the readers are in many cases not sufficient, and more advanced tools are required. Furthermore, academic laboratories need to publish their results in such a manner that other laboratories can reproduce their work. With commercial software, this is very difficult, and reviewers cannot readily assess the used algorithm or software. For the statistical analysis of screening data, visualization of results and images, and web mining, it is necessary to have more sophisticated software tools than typically found on HCS imagers. Many HCS vendors have entered partnerships with companies offering statistics software such as Pipeline Pilot, Genedata, or Spotfire. These software packages also have high licensing fees further increasing the costs. Fortunately, many powerful open-source software are available both for image analysis and statistical analysis (Table 1). Some of these tools have user-friendly GUIs and do not require any scripting skills. Using open-source software has three crucial advantages for academic laboratories: (1) flexibility to further develop the code within a community when needed, (2) reduced costs, and (3) the possibility to develop transparent analysis routines whose methods can be published. For screening facilities like ourselves, there is the added advantage that we can provide to our users the raw data, the processed data, and the analytical tools that were used to process it. The last point is particularly advantageous, as the collaborators are able to extend and modify easily the initial analysis on their own. In this chapter, we discuss two open-source software we use routinely in our screening facility to analyze high-content RNA interference and chemical compound screens. With these software, we are able to process several terabytes of data with relative ease. The aim of this chapter is not to give guidelines on how to analyze image-based screens; for this we refer to an excellent overview collated by several major laboratories in the field [16]. The tools presented here allow researchers to carry out all methods presented in the publication of Caicedo et al.

2

Software

2.1 CellProfiler and CellProfiler Analyst

CellProfiler and CellProfiler Analyst are free open-source software for automated image analysis, data visualization, and machine learning. Versions for Mac, Windows, and Linux are available, and both software can be downloaded at http://www.cellprofiler.org. CellProfiler was developed by Anne Carpenter and Thouis Jones in the laboratory of David Sabatini at the Whitehead Institute for Biomedical Research and by Polina Golland at the CSAIL of the MIT. CellProfiler is designed to analyze large amounts of images automatically [17, 18]. Since the analysis of thousands of images is computationally intensive, a version is available to run on a Linux

Open Source Tools for HCS

47

Table 1 List of open-source software for image analysis and statistics and their websites Software name

Main application

Octave

Image analysis/statistics Scripting

http://www.gnu.org/ software/octave/

Similar to MATLAB

ImageJ

Image analysis

GUI/ scripting

http://rsbweb.nih.gov/ ij/

Java

Fiji

Image analysis

GUI/ scripting

http://fiji.sc/wiki/index. php/Fiji

Java

CellProfiler

Image analysis

GUI

http://www.cellprofiler. org

BioImageXD Image analysis

GUI

http://www.bioimagexd. net

For 3D images

R

Statistics

Scripting

http://www.r-project.org

Similar to S

RapidMiner

Statistics

GUI

http://www.rapid-i.com

KNIME

Statistics

GUI

http://www.knime.org

SciPy

Statistics

Scripting

http://www.scipy.org

Python programming

Shogun

Statistics

Scripting

http://www.shoguntoolbox.org

Integrated into many languages

Weka

Statistics

GUI

http://www.cs.waikato. ac.nz/ml/weka/

μManager

Microscope control

GUI

http://www.micromanager.org/

Motmot

Image acquisition, storage, and analysis

GUI

http://code.astraw.com/ projects/motmot

GUI/scripting Website

Comment

Python and C

cluster allowing the parallelization of the analysis. The software is written in the Python programming language, and the source code can be downloaded and extended by the community (https:// github.com/CellProfiler/CellProfiler/wiki). CellProfiler has an intuitive user interface with point-and-click actions that allow assembling modules in a pipeline for processing images. Several features in CellProfiler make it an easy-to-use software. The website hosts many tutorials and examples of image processing pipelines together with the corresponding images. In the GUI of CellProfiler, every button- or drop-down box for adjusting parameters has a help button, allowing to quickly look up its functionality. Once a pipeline has been assembled, CellProfiler has a test mode to verify the settings of the modules in the pipeline and adjust parameters. With these helpful features, it is

48

Martin Sto¨ter et al.

possible for beginners to start producing quite complex image analysis pipelines in a very short time, provided they have basic knowledge in image processing. CellProfiler provides many advanced image processing algorithms, and the latest release supports 3D images allowing developing image analysis solutions for a wide variety of assays in mammalian cells, tissue slices, and small organisms. There are several modules for preprocessing images such as calculating illumination correction functions and applying them to images or measuring image quality [19]. Other modules detect objects in images using many different algorithms for distinguishing background from foreground and drawing separation lines between objects. There are several modules for performing various morphological operations for feature enhancement. Lastly, features such as intensity, texture, and morphology can be extracted either on a cellby-cell basis or on an image-by-image basis and exported to either CSV files, a MySQLite, or a MySQL database. Visual control of the performance of the image analysis is essential in large-scale screens to verify the segmentation quality of the algorithms. CellProfiler generates overlays of the detected object outlines on images, for instance, on RGB color mergers of the various channels created within CellProfiler or directly on the original images. Another useful feature of the software is the ability to annotate images and to extract metadata either from the path or the file name itself. The handling of the data in the statistical analysis is thereby made much easier and allows to quickly perform quality control checks of the images and the assay. When all the data and all the image file paths are exported to a MySQL database, a properties file can be created. This properties file can then be read by CellProfiler Analyst to access the data in the database [20, 21]. CellProfiler Analyst is a software for rapidly analyzing and visualizing large image-based datasets exported to a MySQL database using the “ExportToDatabase” module in CellProfiler. The measurements extracted from the images or segmented objects can be visualized with four different types of plots: histograms, scatterplots, density plots, and parallel coordinates. All plots are interconnected, and highlighting data points in one plot will highlight the same data points in the others. In this manner, it is possible to rapidly explore relationships among conditions or among parameters. From this, graphical representation data points or regions can be selected (gating), and either the images or the objects giving rise to the measurement can be displayed for visualization. This process is very important in large-scale imaging experiments, as it allows to quickly access data points standing out in the dataset and visually inspect the corresponding images. In this manner, novel phenotypes or artifacts stemming either from image analysis or plate preparation can be verified and sometimes be found as a distinct data cloud of points in a scatterplot. The software

Open Source Tools for HCS

49

CellProfiler Analyst also has a classifier tools and machine learning algorithms to classify cells. The user creates phenotype classes according to prior knowledge and trains the classifier. After a few iterations of testing and correcting, the software is able to classify whole screens and obtain enrichment scores for the various predefined phenotypes. 2.2

KNIME

KNIME is an open-source platform for data mining developed by the Chair for Bioinformatics and Information Mining at the University of Konstanz (http://www.knime.org). The software can be applied for many different types of analysis such as data mining, cheminformatics, image analysis, text mining, deep sequencing, and many more. KNIME provides a wide set of functionalities for many different common tasks such as input/output, data processing, statistics, data mining, and visualization. As KNIME is a rich client application based on the Eclipse platform, it benefits from the Eclipse plug-in concept. This concept provides an interface for software developers to easily extend the functionality of KNIME. New developments for KNIME can be for private use or published as community contribution. Especially in academic science, data analysis often starts as an exploratory creative process with evolving ideas of the data analysis flow and rapidly changing analysis parameters or conditions. Therefore, data analysis software has to be extremely flexible in order not to limit the exploration of data. Furthermore, it is important that the data analysis process is comprehensible and easily readable at all time points to ensure that scientists can share their approach with colleagues and to better prevent conceptual mistakes. A third requirement to data analysis software is the minimization of effort and time a scientist has to invest to implement various methods. KNIME provides a user-friendly interface to visually create workflows allowing a step-by-step data analysis flow. A node—as a single entity of such a workflow—provides a very confined analysis step with a set of parameter configurations. Workflows can branch at any point, which allows easily implementing multiple approaches. As KNIME is not focused on data analysis of high-throughput/high-content experiments, it lacks common methods necessary for screen mining. We have therefore developed and implemented a set of KNIME nodes providing various popular screen analysis methods available as community contribution in extension packages (https://www.knime.com/community). The first extension is called “HCS Tools” (https://github.com/ knime-mpicbg/HCS-Tools/wiki). The package includes readers for various file formats of image analysis software and plate readers, quality control metrics (Z-prime factors (Z0 ), strictly standardized mean difference (SSMD), coefficient of variance (CV)), common plate normalization methods (percent of control (POC), z-score,

50

Martin Sto¨ter et al.

B-score), an IC50 calculation and plotting tool, a very powerful plate heatmap viewer, as well as plate utilities for easy handling of barcodes, wells, and plate layouts. The second set of extensions is the open-source scripting integration framework for KNIME (https://github.com/knimempicbg/knime-scripting/wiki). Even for software developers, it is often a challenge to implement complicated statistic or data mining methods. Instead of investing time and resources in such tasks, it is more efficient to exploit methods that were already developed in software for statistical computing such as R or MATLAB. Although KNIME already implemented nodes to execute R scripts, we devised our own integration to simplify the use of R scripts for biologists without any scripting background. We implemented a new open-source scripting integration framework for KNIME, which is based on templates. Its main purpose is to hide the script complexity behind a user-friendly graphical interface (GUI). Furthermore, our approach goes beyond the existing integration of R as it provides better and more flexible graphics support, flow variable support, and an easy-to-extend server-based script template repository. The implementation is based on RGG (a R GUI generator, http://rgg.r-forge.r-project.org/) to provide a graphical user interface instead of a scripting interface. It allows us to further integrate the scripting languages Python, MATLAB, and Groovy. Using these tools, we have developed many statistical methods and plotting functions in KNIME.

3

Workflow of a HCS Analysis

3.1 Assay Development

Assay development is an iterative process whereby experimental conditions of the assay are optimized to obtain the greatest, reliable separation between positive and negative controls. In contrast to homogeneous assays that measure mostly one or two parameters, HCS can measure a plethora of parameters, some of which are useful for describing the phenotype under study and some of which are not. To complicate matters further, both the image acquisition and the image analysis are often optimized in parallel during assay development to better measure the observed phenotypes. Thus, assay development in HCS involves a combination of optimization of experimental conditions, parameters, acquisition settings, and data analysis methods. To carry out this complex task, CellProfiler and KNIME are ideally suited, thanks to their flexibility.

3.2 Loading and Preprocessing Images

Our facility has several high-content imagers such as the Yokogawa CV7000, the OPERA and Operetta from Perkin Elmer. These platforms produce images in various formats since standards are lacking in the imaging community. CellProfiler can read all of the

Open Source Tools for HCS

51

formats with a “LoadImages” module using the Bio-Formats library of LOCI. In this module, metadata such as plate name, date, well, or field can be extracted both from the file name and path using regular expressions and from the heard of the file. Once images are loaded, in a pipeline, they often require to be corrected for uneven illumination, which can be carried out with the “IlluminationCorrectionCalculate” and “IlluminationCorrectionApply” modules. Automated imagers focus automatically using either hardware or software focusing. There is generally a proportion of out-of-focus images in large screens, especially when using high-magnification lenses. The “MeasureImageQuality” module of CellProfiler can be used to flag out-of-focus images. Alternatively, data from out-of-focus images can be removed in subsequent analysis steps (see Subheading 3.9), as they often represent outliers. We analyze our images with CellProfiler installed on an in-house Linux cluster to speed up this calculation intensive process. The cluster is a 105-node cluster with 2520 cores with each 256 GB of RAM. We also run CellProfiler on 920 GPU nodes with 512 GB of memory. The cluster is connected to a 2.1 PB Lustre infrastructure. 3.3 Segmenting Objects

The most crucial and computer-intensive step in image analysis is object segmentation. A common task in most image analysis routines is to detect nuclei as primary objects and cytoplasm of cells as secondary objects around the nuclei. We generally stain nuclei with Hoechst and the cytoplasm with HCS CellMask Blue (Life Technologies). Nuclei are detected using Otsu’s three-class thresholding, and fused objects are separated by finding local maxima followed by watershed. These methods are available in the “IdentifyPrimaryObjects” module. Next, using the “DetectSecondaryObjects” module, the stained cytoplasms are detected by a propagation algorithm using local image similarity and the nuclei as seeds. The local thresholds are again determined by using multi-threshold Otsu. A further common task in many biological assays is to detect tertiary objects such as organelles in the cytoplasm of cells. For small structures with high background intensities, a white top-hat filter is applied using the “EnhanceOrSuppressFeatures” module to increase the contrast. The tertiary objects are then detected, using the “IdentifyPrimaryObject” module with multi-threshold Otsu and watershed seeded from local maxima. Finally, to obtain tertiary object data on a cell-by-cell basis, the objects are linked to their parent cell with a “RelateObjects” module.

3.4 Extracting Parameters

After segmenting all the objects of interest, features are extracted for the quantitative phenotypic description. Several measurement modules are available in CellProfiler for both imagebased measurements and object-based measurements. Using a

52

Martin Sto¨ter et al.

“MeasureObjectIntensity” module, many aspects of the intensity of objects can be captured such as their integral, average, or median intensity. Further intensity properties are the maximum, the minimum, the upper and lower quantile, or the standard deviation of the intensity within an object. Depending on the nature of the object and the possible phenotypes, any of these intensity attributes can be used to describe the phenotype of interest. Other parameters describe the morphology (“MeasureObjectSizeShape”), the texture (“MeasureTexture”), or the spatial distribution of objects (“MeasureObjectRadialDistribution,” “MeasureObjectNeighbors”). Extracting all these parameters can easily generate tables of several hundreds of columns. All these features can be used to either describe the objects for statistical analysis or directly within a pipeline to select objects for further processing and segmenting. For instance, in a mixed population of large, irregular cells and small round cells, it is possible to select with “FilterObject” module one population based on cellular feature to further segment tertiary objects. In general, we extract many parameters during initial image analysis to avoid a rerun of the analysis, since segmentation is a computationally intensive work. In this manner, it is possible to select relevant parameters later in the workflow of a screen analysis (see Subheading 3.8). 3.5 Control Images and Exporting Data

When analyzing large screens, it is impossible to visually control the quality of segmentation of all the images. A spot control needs to be carried out, and control images showing the outlines of the segmented objects are very useful for this. Furthermore, it is paramount to visualize all the images of hit wells to ensure that no spurious segmentation artifact leads to the result. In CellProfiler, it is possible to create color images of three- or four-channel images with the “GrayToColor” module. The outlines of each segmented object can be saved in the corresponding segmentation module and overlaid on the color merge image with the “OverlayOutlines” module. The image can then be saved with the “SaveImages” module in many different formats and the file name and path stored in the data output file. Thus, after identifying hits, the control images can be automatically retrieved within KNIME (see Subheading 3.13) for visualization. The last step of the analysis exports the data and associated metadata as comma-separated file (CSV file) using the “ExportToSpreadsheet” module or to a database using the “ExportToDatabase” module. If the image analysis of the entire screen is run on a cluster, the data is split into several jobs, each resulting in a separate CSV file; hence, the screen results are distributed among several CSV files.

3.6 Importing Data into KNIME

To read the exported CSV result files of CellProfiler into KNIME, we first capture the path of all files with a “List Files” node. The list of paths is then connected to an “Iterate List of Files” node to load

Open Source Tools for HCS

53

the data into a KNIME workflow. The barcode, plate row, and plate column metadata contained in the CSV files are used to associate a plate layout file (either a CSV file or a Microsoft Excel file) to an experimental condition for each line (lines representing either objects or images). This association can be carried out either with a “Joiner” or a dedicated “Join Layout” node developed by us. We have generated tables containing over ten million lines and hundreds of columns. KNIME is able to carry out computations with files of this size on a normally equipped desktop computer. The only crucial factor is that enough free disk space is available to store temporary files. 3.7 Parameter Selection

Parameter selection is a crucial process in multiparametric screening, and a full discussion on the subject is beyond the scope of this chapter. Briefly, parameters are sought that allow best separation between negative control and positive control. The parameters should describe various aspects of the phenotype and should therefore not be too much correlated to each other. Furthermore, the parameters should be robust. A classical method to estimate the discrimination strength of parameters is the Z-prime factor [22]. We have developed a KNIME node “Z-Primes” to calculate both robust Z-prime (with median and MAD) and non-robust Z-prime factors (mean and standard deviation). The node allows selecting several positive and negative controls, which is useful during assay development for determining the best controls to use during screening. In this manner, a matrix is obtained of Z-prime factors for all parameters and possible combination of positive and negative controls. Similarly, the strictly standardized mean difference (SSMD) [23], which is another value describing the strength of a parameter, can be calculated using the “SSMD” KNIME node. Lastly, we have developed two KNIME nodes to select parameters based on mutual information [24]. The “Parameter Mutual Information” computes the mutual information matrix for all parameters. The “Group Mutual Information” computes the mutual information between two reference populations for a set of selected parameters. In this manner, it is possible to select parameters and discover new phenotypes in a screen.

3.8 Binning of Population: Z-Score Profiles

In some assays, only a subpopulation shows a phenotypic response, and well averages are not able to discriminate between experimental conditions. Under these circumstances, we have found that dividing the negative control population into percentiles and comparing the population distribution in the corresponding parameter could be a useful analysis method for scoring hits. We have developed KNIME node for this subpopulation analysis. The script partitions the sorted population of a negative control into n equal-sized bins. Each bin covers a percentage of a parameter range. The script then

54

Martin Sto¨ter et al.

determines the count of objects in each parameter bin for each well. Objects whose values lie either below or above the range of the negative control are attributed to either the lowest or highest bin, respectively. The percentage of the population in each bin and each well is used to compute z-score values relative to the corresponding bin of the negative control wells (Fig. 1). This process is applied automatically for each parameter and each experimental condition in KNIME partitioning the population into a chosen number of bins. The final output of the template is a profile of n z-scores that we call “z-score profiles.” Eliminating z-scores between +2 and 2 using a KNIME node named “Range Filter” reveals the remaining parameters whose mean is at least two standard deviations away from the control population and has some discriminatory power. In this manner, it is possible to select parameters that score changes in population distributions. 3.9 Redundant Parameter Elimination

In order to eliminate parameters that are correlated to each other, we calculate either their Pearson correlation coefficients [25], their Spearman rho [26], or Kendall tau [27]. Uncorrelated parameters have coefficients close to zero and likely describe different aspects of the phenotype under study. We have developed a R template in KNIME to calculate correlation coefficients between parameters. Redundant parameters that yield correlation coefficients above 0.4 are eliminated. It is important to visually inspect the structure of the data using scatter matrices. A “Scatter Plot” and a “Scatter Matrix” node from KNIME exist that allow color-coding the controls for ease of viewing.

3.10 Robust Parameter Selection

Lastly, it is desirable that parameters are able to discriminate between positive and negative conditions in a variety of experimental conditions. In other words, they should be robust and reproducible. For this purpose, the Pearson correlation coefficient between all experimental repeats using control wells is calculated. Robust parameters have high Pearson correlation coefficients (above 0.7) in pairwise comparisons of experimental repeats. For this analysis, we have developed another R template in KNIME to calculate the Pearson correlation coefficient between experimental runs.

3.11

Removing outliers due to either out-of-focus images, debris, or segmentation errors is an important step in analyzing large datasets, as they can introduce errors and increase erroneously the estimated variance. It is important to carry out this analysis within the same and not between experimental conditions, as this might effectively eliminate hits. For instance, outliers from the entire set of images of the negative control wells can be removed to correctly estimate the location and spread of that population. We have developed a “Remove Outliers” node in KNIME that allows selecting either

Outlier Removal

Open Source Tools for HCS

A

Image based averages

Z Score

1 0 -1 -2 -3

Negative Experimental Control Point

100.0

Percentiles

C

100 80 60 40 20 0

Experimental Point Negative Control

Measurement

Negative control Percentiles 100.0% 80.0% 60.0% 40.0% 20.0%

80.0 60.0 40.0 20.0 0.0 100.0

Percentiles

B Object based distributions 140 120

2

55

Measurement

Experimental point Percentiles 100.0% 88.6% 74.4% 57.8% 33.4%

80.0 60.0 40.0 20.0 0.0 Measurement

D

Z Profile of experimental point Z Score

10 5 0 -5 Bins of the measurement

Fig. 1 Binning analysis allows detecting changes in subpopulations. Imagebased averages of repeats of an experimental population might show low z-scores for a chosen parameter as shown by the box plot in panel (a). The distribution of object-based values of the same parameter in experimental populations might still show significant differences as indicated by the density histogram in panel (b) or the cumulative distribution function in panel (c). The binning analysis subdivides the control population into a chosen number of equal-sized bins (panel (c), upper graph). The upper and lower parameter values of each bin divide all well populations. The population percentages in all bins are used for computing z-scores relative to the negative control wells. As shown in the box plot of the percentiles in panel (d), some subpopulations show significantly higher/lower z-score values than the average z-score in panel (a), demonstrating the superior discriminative power of binning analysis

of two methods for determining outliers from a chosen population: standard deviation and box plot statistics. In the standard deviation method, objects or images whose parameter values lie more than four standard deviations from the mean of the chosen population

56

Martin Sto¨ter et al.

can be considered outliers. In the box plot statistics method, objects or images lying outside the whiskers are removed. Useful parameters are, for instance, fluorescence intensity or surface of nuclei. Also, images with very few or too many cells could be removed, since the segmentation tends to be unreliable in such images. 3.12 Plate Quality Control (QC)

During a screening campaign, some plates and even entire batches can fail due to various reasons. To control the quality of plates, during a screen, we use the “Z-primes” and the “Multivariate Z-primes” nodes we have developed. Plates failing to fulfill a minimal criterion such as a Z-prime factor above 0.3 are eliminated from further analysis and scheduled for rescreening. A further useful node to verify the quality of plates is the “CV” node that calculates coefficients of variance for selected parameters. Lastly, it is important to visualize plates to obtain a graphical overview of the screen. To this end, we have developed a “Plate Viewer” to create heatmaps and a R template in KNIME to generate scatterplots of screening campaigns. These tools allow to visualize row and column artifacts and to compare the performance of various plate batches during a screening campaign.

3.13 Data Normalization

Due to plate-to-plate variations from different days or runs, a normalizing step is necessary to render the data comparable across entire screens. We have developed several KNIME nodes for popular normalization methods in HTS such as percent of control (POC), normalized percentage inhibition (NPI), standard score (z-score), and B-score [28]. For all nodes, robust statistics, grouping, negative control, and parameters can be chosen. The method chosen for normalization is dependent on the screening results and the normality of the data. A full discussion on this issue is beyond the scope of this article, and the reader is referred to excellent reviews [16, 29, 30].

3.14

Hit selection is the purpose of any screen and is therefore very important. Many different methods and criteria are applied in screening. Our method consists in calculating the separation of experimental wells from control in multiparametric space using either the Mahalanobis distance [31] or multivariate Z factor. We have implemented a R template in KNIME to calculate the Mahalanobis distance and a KNIME node for the multivariate Z-prime factor. Prior to applying this analysis, it is critical to verify that the data are approximately normally distributed. We have developed a R template in KNIME to produce quantile-quantile plots and compute a Shapiro-Wilk test for normality. A popular method to render data more normally distributed is to take the logarithm of the parameter. We have also developed a R template for Box Cox power transform in KNIME to attempt modifying data to become

Hit Selection

Open Source Tools for HCS

57

more normally distributed. To select a threshold when applying Mahalanobis distance, we either calculate the chi-square p value for the 99 percentile, select the top ~200 reagents, or set the threshold above the highest Mahalanobis distance of the negative control. 3.15

Clustering

3.16 QC of Hit Selection

Classification of hits into different phenotype classes is a debated issue in the field. Many clustering algorithms are implemented in KNIME, and a discussion about the choice of algorithm to use goes beyond the scope of this article. However, a clustering approach can yield insights into mode of action of hits and also help discover novel phenotypes. Our favorite approach uses cosine as distance measure, i.e., the angle between the n-dimensional vectors. In this manner, the amplitudes of the phenotypes are ignored, and only the directions of the vectors are of importance for clustering. To this end, we use a “Distance Matrix Calculate” node in KNIME and select cosine. To obtain a feel for the likely number of clusters in the dataset, we carry out a hierarchical clustering with the “Hierarchical Clustering” node. The result of the clustering is examined with the “Hierarchical Cluster View” node. Deciding on the number of clusters in the dataset is difficult and subjective. To help determine the appropriate number of clusters, it is helpful to have reagents (siRNA, CRISPR, or compounds) of known mode of action expected to yield different phenotypes. A good distribution should yield a random distribution of the negative control wells in each cluster, and known reagents should cluster together. Negative controls distribute randomly, since the data is normalized to them and their multiparametric vectors should therefore point randomly in space. Different clustering algorithms yield different results, and we therefore apply additional clustering algorithms such as k-Means and k-medoids or t-SNE to explore our screening data and compare the results. k-Means and k-medoids are implemented in KMINE nodes, and t-SNE is implemented in a Python snippet of our data integration. After classification of the data, hits are selected by applying a threshold on the amplitude of the phenotype as judged by the Mahalanobis distance (3.12). Deciding on an appropriate analysis is at this point subjective, but additional information such as the average Pearson correlation of profiles in a cluster or simply how positive controls cluster can be useful for judging the result of any clustering method. It is important to verify the results of the hit selection. Hits can be visualized as a heatmap in the “Plate Viewer” node to determine whether certain well positions, such as edge rows or edge columns, have a higher number of hits. It is also paramount to visualize the images of the identified hits. The path information and file name information of the control images generated by CellProfiler can be used to this end. In

58

Martin Sto¨ter et al.

KNIME, all the images of the hits can be opened using the “Picture Chooser” node. In this manner, all the control images with the corresponding object outlines can be quickly visualized. Typically, we rescreen selected hits in triplicate using the same assay, and, for validated hits, a dose-response relationship is generated. We have developed a “Dose Response” node in KNIME to plot dose-response curves and to calculate IC50 values. 3.17 Screen Annotation

4

Information of our chemical and siRNA libraries is stored in a PostgreSQL database. We annotate our validated hits using a “DatabaseReader” and a “Joiner” node to obtain either chemical structures for chemical screens or GeneID for RNAi screens. KNIME has many tools for cheminformatics, can visualize the molecule structures, and has tools to retrieve data from external public databases via web queries to further annotate hits and to allow clustering of either by chemical substructures or GO terms.

Conclusion With the availability of CellProfiler and KNIME, two very powerful open-source software tools have been created to analyze highcontent screens. These tools allow creating very complex analysis workflows, yet they can be handled by noncomputer scientists. A further advantage of using these software tools is that our analysis workflows become standardized, in spite of using several different automated imaging platforms. The imaging community has still not managed to agree on standards, which makes the comparison of data across laboratories difficult or impossible. By using the CellProfiler and KNIME, we have established a de facto standard for internal use and for our clients. Several conditions need to be fulfilled for image analysis and statistical software to be really useful for analyzing large datasets. For image analysis software, it is paramount that it can be run robustly on large computer clusters since calculating complex image analysis requires significant amount of processing power. The software must also have good visualization tools to allow verifying the performance of the segmentation. Image quality controls need to be available to correct for uneven illumination and out-of-focus images. All of these requirements are found in CellProfiler. For the statistical software, it is important that it can handle large datasets. Especially when analyzing screens at the individual cell level, the software must deal with millions of rows and hundreds of columns. For analyzing screens, special statistical methods are employed, and they need to be readily available. Visualization of the data at each single data processing step is also extremely important for handling the data properly and gaining insights into the data. The implementation of new statistical

Open Source Tools for HCS

59

methods and many visualization tools and their flexible extension is of great importance. KNIME with its plug-in concept and various extensions integrates all these requirements. Lastly, both CellProfiler and KNIME are easily extendable open-source software that both have a highly active and fastgrowing community since the release of their first versions in 2005 and 2006, respectively, a key for further development and implementation of new functions and state-of-the-art methods. Together CellProfiler and KNIME offer the possibility to apply advanced analysis tools to high-content screening data without paying expensive licensing fees. Especially in a time of reduced research budget, it is an important aspect for academic institutes interested in acquiring high-content imagers. If the costs of running high-content imagers will decrease, automated imaging will become more widely applied, and subsequently, new applications and modalities will be developed. With academia entering the field of automated imaging, smart microscopy with advance imaging modalities in automated mode implementing feedback loops between microscope, image analysis, and control of acquisition mode might become more widely applied. Thus, many applications will become automated in the future, freeing up time for researchers to carry out other tasks [32].

Acknowledgments This work was supported by the Max Planck Gesellschaft. References 1. Johnston AR, Powell RV (1970) Optics at the Jet Propulsion Laboratory. Appl Opt 9 (2):271–275 2. Harmon LD, KK C (1969) Picture processing by computer. Science 164(3875):19–29 3. Lipkin LE, Lipkin BS (1975) Computers in the clinical pathologic laboratory: chemistry and image processing. Annu Rev Biophys Bioeng 4(1):529–577. https://doi.org/10.1146/ annurev.bb.04.060175.002525 4. Blackwell RJ, Crisci WA (1975) Digital image processing technology and its application in forensic sciences. J Forensic Sci 20(2):17 5. Archer JR (2004) History, evolution, and trends in compound management for high throughput screening. Assay Drug Dev Technol 2(6):675–681. https://doi.org/10.1089/ adt.2004.2.675 6. Newman DJ, Cragg GM, Snader KM (2003) Natural products as sources of new drugs over the period 1981–2002. J Nat Prod 66

(7):1022–1037. https://doi.org/10.1021/ np030096l 7. Ortholand J-Y, Ganesan A (2004) Natural products and combinatorial chemistry: back to the future. Curr Opin Chem Biol 8(3):271–280. https://doi.org/10.1016/j.cbpa.2004.04. 011 8. Giuliano KA, DeBiasio RL, Dunlay RT, Gough A, Volosky JM, Zock J, Pavlakis GN, Taylor DL (1997) High-content screening: a new approach to easing key bottlenecks in the drug discovery process. J Biomol Screen 2 (4):249 9. Verkman AS (2004) Drug discovery in academia. Am J Physiol Cell Physiol 286(3): C465–C474 10. Cressey D (2011) Drug-maker plans to cut jobs and spending as industry shies away from drug discovery. Nature 470:154. https://doi. org/10.1038/470154a

60

Martin Sto¨ter et al.

11. Holt R (2011) Dueling visions for science. Science 333(6049):1549–1549 12. Gulledge J (2011) Debt crisis: crunch time for US science. Nature 477(7363):155–156 13. Hunter P (2010) Facing the credit crunch. EMBO Rep 11(12):924–926 14. D’Ausilio A (2011) Arduino: a low-cost multipurpose lab equipment. Behav Res Methods 44:1–9. https://doi.org/10.3758/s13428011-0163-z 15. Santos AF, Zaltsman AB, Martin RC, Kuzmin A, Alexandrov Y, Roquemore EP, Jessop RA, MGMv E, Verheijen JH (2008) Angiogenesis: an improved in vitro biological system and automated image-based workflow to aid identification and characterization of angiogenesis and angiogenic modulators. ASSAY Drug Dev Technol 6(5):693–710. https://doi.org/10.1089/adt.2008.146 16. Caicedo JC, Cooper S, Heigwer F, Warchal S, Qiu P, Molnar C, Vasilevich AS, Barry JD, Bansal HS, Kraus O, Wawer M, Paavolainen L, Herrmann MD, Rohban M, Hung J, Hennig H, Concannon J, Smith I, Clemons PA, Singh S, Rees P, Horvath P, Linington RG, Carpenter AE (2017) Dataanalysis strategies for image-based cell profiling. Nat Methods 14:849. https://doi. org/10.1038/nmeth.4397 17. Carpenter A, Jones T, Lamprecht M, Clarke C, Kang I, Friman O, Guertin D, Chang J, Lindquist R, Moffat J, Golland P, Sabatini D (2006) CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol 7(10):R100 18. Kamentsky L, Jones TR, Fraser A, Bray M-A, Logan DJ, Madden KL, Ljosa V, Rueden C, Eliceiri KW, Carpenter AE (2011) Improved structure, function and compatibility for CellProfiler: modular high-throughput image analysis software. Bioinformatics 27 (8):1179–1180 19. Bray M-A, Fraser AN, Hasaka TP, Carpenter AE (2011) Workflow and metrics for image quality control in large-scale high-content screens. J Biomol Screen 17(2):135–142 20. Jones R, Carpenter E, Lamprecht R, Moffat J, Silver S, Grenier K, Castoreno B, Eggert S, David R, Golland P, Sabatini D (2009) Scoring diverse cellular morphologies in image-based screens with iterative feedback and machine learning. PNAS 106(6):1826–1831

21. Jones T, Kang I, Wheeler D, Lindquist R, Papallo A, Sabatini D, Golland P, Carpenter A (2008) CellProfiler Analyst: data exploration and analysis software for complex image-based screens. BMC Bioinformatics 9(1):482 22. Zhang JH, Chung TD, Oldenburg KR (1999) A simple statistical parameter for use in evaluation and validation of high throughput screening assays. J Biomol Screen 4(67):67 23. Zhang XD (2007) A pair of new statistical parameters for quality control in RNA interference high-throughput screening assays. Genomics 89(4):552–561. https://doi.org/10. 1016/j.ygeno.2006.12.014 24. Moddemeijer R (1999) A statistic to estimate the variance of the histogram based mutual information estimator based on dependent pairs of observations. Signal Process 75 (1):51–63 25. Pearson K (1896) Mathematical contributions to the theory of evolution. III. Regression, heredity and panmixia. Philos Trans R Soc Lond A 187:253–318 26. Spearman C (1904) The proof and measurement of association between two things. Am J Psychol 15(1):72–101. https://doi.org/10. 2307/1412159 27. Kendall MG, Smith BB (1938) Randomness and random sampling numbers. J R Stat Soc 101(1):147–166. https://doi.org/10.2307/ 2980655 28. Brideau C, Gunter B, Pikounis B, Liaw A (2003) Improved statistical methods for hit selection in high-throughput screening. J Biomol Screen 8(6):634–647 29. Birmingham A, Selfors LM, Forster T, Wrobel D, Kennedy CJ, Shanks E, SantoyoLopez J, Dunican DJ, Long A, Kelleher D, Smith Q, Beijersbergen RL, Ghazal P, Shamu CE (2009) Statistical methods for analysis of high-throughput RNA interference screens. Nat Methods 6(8):569–575 30. Shun TY, Lazo JS, Sharlow ER, Johnston PA (2011) Identifying actives from HTS data sets. J Biomol Screen 16(1):1–14 31. Mahalanobis PC (1936) On the generalised distance in statistics. Proc Natl Inst Sci India 2 (1):49–55 32. Conrad C, Gerlich DW (2010) Automated microscopy for high-content RNAi screening. J Cell Biol 188(4):453–461

Part II Drug Target Profiling and Validation

Chapter 5 In Silico Target Druggability Assessment: From Structural to Systemic Approaches Jean-Yves Trosset and Christian Cave´ Abstract This chapter will focus on today’s in silico direct and indirect approaches to assess therapeutic target druggability. The direct approach tries to infer from the 3D structure the capacity of the target protein to bind small molecule in order to modulate its biological function. Algorithms to recognize and characterize the quality of the ligand interaction sites whether within buried protein cavities or within large proteinprotein interface will be reviewed in the first part of the paper. In the case a ligand-binding site is already identified, indirect aspects of target druggability can be assessed. These indirect approaches focus first on target promiscuity and the potential difficulties in developing specific drugs. It is based on large-scale comparison of protein-binding sites. The second aspect concerns the capacity of the target to induce resistant pathway once it is inhibited or activated by a drug. The emergence of drug-resistant pathways can be assessed through systemic analysis of biological networks implementing metabolism and/or cell regulation signaling. Key words Drug targets, Hot spots, Druggability, Flo-QXP, Pocket finder, Structure superimposition, Target promiscuity, Protein cavity

1

Introduction A “druggable” target is typically a receptor that interacts with a drug, i.e., a chemical compound or a protein that will trigger a biological response which will influence the evolution of the disease. In the context of this article, only proteins will be considered as a target and small molecular weight molecules as a drug. Identification of a disease-relevant target is a challenge on its own. This involves a significant number of information research [1–9] and validation studies to prove the influence of this specific target for the particular disease. Once the relevance of the target is found, the next step is to find a reliable biochemical or cellular assay to identify interacting compounds. With both a validated target and an efficient bioassay in hands, drug discovery can start, even without structural information. However, the increasing numbers of

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_5, © Springer Science+Business Media, LLC, part of Springer Nature 2019

63

64

Jean-Yves Trosset and Christian Cave´

experimental tridimensional structures have provided insight on the way drug interacts with protein targets. This understanding has contributed to the development of structure-based drug design techniques which have been adapted to assess the structural druggability of target, i.e., its capacity to bind a small drug-like molecule. Structure-based druggability assessment can be carried out independently of the biological relevance of the target or the existence of a suitable screening assay for high-throughput screening (HTS). Various aspects of target druggability can be assessed in silico [9–11] which we classify as direct and indirect approaches. Direct approach focuses on the capacity of the target to bind a small molecule, which can be assessed through structural analysis of surface of the target. The first step consists to check for the presence of a protein cavity that can be triggered by a small molecule. The second step is to check whether the physicochemical properties of this cavity can complement with the properties of a drug-like molecule through the existence of suitable pharmacophore site at the surface of the binding site. The shape and the existence of micro-cavities or narrow clefts that can become strong interacting hot spots [12–16] for the ligand is the third aspect of the assessment of the binding pocket. This particular case of protein-protein interactions (PPI) will be discussed in this context. A fourth aspect of the direct approach of the druggability assessment concerns the binding site flexibility [17]. Finding a cavity in a rigid target supposes that the cavity already exists. Certain pockets like inducible allosteric sites are revealed after protein conformational change. With the existence of multiple X-ray conformers for a specific target, it is possible to assess the relative variability of certain residues within the binding site. Certain techniques that assess the plasticity of the binding site will be presented. Indirect approaches deal with alternative issues that might challenge target druggability once a ligand-binding site is identified or even well characterized. It concerns target promiscuity, i.e., the degree of structural similarity of the target with other proteins of the organism. High similarity with unwanted targets or other members of the protein family may lead to toxicity problems at later stage of drug development. Such cross-reactivity of chemicals with unwanted targets can be identified using large-scale in silico structural comparison techniques [18]. The second type of difficulties that can be addressed through indirect approaches concerns systemic effects. Sometimes, targets may reveal as nonvalid after long preclinical studies even if the basic druggability criteria are met. This might be due to the emergence of alternating pathways induced by the action of the drug on the target. Such systemic effects can be analyzed through system biology approaches involving gene regulation network analysis. The systemic druggability assessment techniques will be reviewed at the last section of this chapter.

Target Druggability Assessment

2

65

Materials and Methods

2.1 Finding Protein Cavities

Geometry-based approach: from a geometrical point of view, a cavity is a concave empty space that can be described using 2D (surface) or 3D shape descriptors [19–21]. We consider three regions in the protein environment: the protein bulk, the bulk solvent, and the cavity space. The protein bulk is the space filled by the protein atoms. The bulk solvent is the space “outside” the protein which differentiates from the space “inside” the protein which defines the cavity where the drug-like molecule is supposed to bind. The identification of protein pockets by numerical methods supposes the capacity to discriminate first the protein bulk from the rest which is relatively easy and second to differentiate the bulk solvent from the cavity space which is more difficult and involves some 2D and 3D topological concepts that are used alternatively or in combination to distinguish these three regions. 2D approach: in this case, the delineation of the three regions is made through the calculation of a close surface surrounding either the protein alone or the protein including the cavities. A close surface defines by definition an inner and outer part of the protein. The inside region depends on the way the surface is calculated and in particular whether the interstice and micro-cavities are included in the interior of the surface. This surface is calculated by rolling a spherical probe on the protein atoms and by estimating for each of them the surface area accessible to this probe (see Note 1). The smaller the probe, the smaller will be the interstices accessible by the probe. For example, an infinitely small probe will be able to enter all interstices and cover the complete solvent accessible van der Waals surface of the protein (excluding overlap with protein atom). On the contrary, a very large probe will not be able to enter any of the protein interstices. This probe will describe a convex envelop of the protein including both the protein bulk and the cavity space. This surface would exclude nicely the bulk solvent from the rest (protein bulk + cavities). This surface can be considered as an upper limit surface that leaves the bulk solvent “outside” the surface. This surface is somehow equivalent to the convex hull of the set of protein atoms (see Note 2). The region outside the convex hull is exclusively made of bulk solvent. The reverse is not true: the inside region of the convex hull does not define exclusively the protein atoms and cavity space. Unfortunately, it may also include some bulk solvent area. The reason is that the protein envelop is not convex (see Note 2). Bended proteins like betacatenin [22], for example, have a large solvent region between the two protein lobs that cannot be considered as a cavity (see Fig. 1). Other cavities like the RNA-/DNA-binding region of the polymerases are much larger than the average volume of small molecular weight compounds. In the absence of DNA/RNA, this cavity is

66

Jean-Yves Trosset and Christian Cave´

Fig. 1 Geometry concepts for analyzing protein cavity. The protein bulk is represented in black, probes as little spheres. The convex hull of the protein is represented in dash line. The plain vectors emerging from the probe in gray are pointing toward the bulk solvent, whereas the dash vectors will encounter protein atoms within a radius of 8 A˚ (radius of the influence circle in dot line). The probe in front of the cleft defines the degree of precision in representing the molecular surface

filled by water that can be considered as part of the bulk solvent. For such type of targets, the convex hull is not a good protein envelop. It includes large region of bulk solvent [23]. Various surfaces can be calculated depending of the size of the probe. They can exclude the bulk solvent when using large probe and exclude also the protein cavities as the probe get smaller. Alphashape theory is an attempt to capture the imbrication of the various hierarchical levels into a single mathematical framework [24]. It provides a natural envelop that is in between the convex hull envelop of the protein and the protein surface that exclude all micro-cavities. Protein cavity surfaces are in between these two limits. In alpha-shape theory, the hierarchy of the various molecular surfaces is controlled by the size of the probe (1/alpha). For each alpha, there is a corresponding surface of the protein that envelops the cavities for which the entrance is smaller than the probe size (see white probe at the entrance of the cleft in Fig. 1). Alpha-shape theory is a tool to describe the surface of object like protein using a single continuous alpha parameter. There is no mathematical consideration concerning the best alpha parameter to use however. Topological concepts in the text are still necessary to choose the suitable protein surface that will delineate adequately the bulk solvent from protein cavities [24]. This hierarchical concept is discussed in detail by Kawabata et al. and implemented in pocket finder algorithms [25, 26]. 3D approach: here we consider the 3D environment of a probe. We investigate whether this probe is part of the protein bulk or within an empty space. It is relatively easy to see whether a probe is in the vicinity or overlapping with a protein atom. The difficulty is

Target Druggability Assessment

67

to distinguish between the cavity space and the bulk solvent. 3D neighborhood of a probe can easily be estimated numerically by immersing the protein into a three-dimensional grid. This 3D grid is made of nodes located at a center of a unit cube of a given size ˚ ). Each node can be considered as a (usually between 0.25 and 1 A reference point to which we can assign a spherical probe of a given size. The default value is often chosen to be that of a solvent water ˚ ). Once the probe is attached to a node, we molecule (radius ¼ 1.4 A can calculate various parameters such as the local atomic density, the energetic field, or the partial free energy contribution of all the protein atoms or any other physical or geometrical local parameters at that grid position [27]. The calculated information is then stored at the corresponding node (see Note 3). The center of the node can either be inside a protein atom or outside in the empty space. A simple pocket finder algorithm would consist to identify the nodes that are not overlapping with protein atoms and delineate the cluster of nodes in the empty space of suitable size to be considered as a pocket. As mentioned above, this empty space can itself be divided into bulk solvent and protein cavities. If this is very easy to localize the high density region (protein bulk), the real challenge is to distinguish between bulk solvent and protein cavity. In the 3D framework, the concept used to make this distinction is atom density. A solvent molecule in a cavity is surrounded by a lot of protein atoms, certainly more than when it is in the bulk solvent or close to a flat protein surface. The calculation of the number of surrounding protein atoms at an “empty” node of the 3D grid (node in the empty space) will tell whether the node is in a protein cavity or outside in the bulk solvent. This approach has been used by Weisel et al. Their PocketPicker algorithm [28] defines the degree of buriedness as the number of protein atoms which lies within a ˚ centered at the probe. A region with high buriedness sphere of 8 A index corresponds to a cavity region inside the protein. Such an approach is very pertinent to find micro-cavities or clefts that are surrounded by a lot of protein atoms. However, when the volume of the protein cavity increases or when the cavity broadly opens to solvent, the distinction between cavities and bulk solvent becomes more difficult. Cavity can always be defined theoretically as an empty space that is predominantly surrounded by protein atoms at a given shell radius. An equal exposition toward protein atoms or toward the bulk solvent corresponds to a “flat” protein surface. This can be estimated easily using a vector-based argument. Looking at all the directions around a given probe (node of 3D grid), a linear search can be carried out in each direction to check for the presence of a protein atom within the circle of influence (the radius of which is defined by the user). For a uniform distribution of vectors emerging from each node, two types of vectors will be identified: those pointing toward protein atoms (dashed vectors

68

Jean-Yves Trosset and Christian Cave´

in Fig. 1) and those pointing toward the bulk solvent (plain vectors in Fig. 1). The relative percentage of these two types of vectors will tell whether the probe is in the bulk solvent or inside the cavity. For example, the probe in the center of the dotted circle (on the right side of Fig. 1) is localized at the entrance of the cavity: 50% of the vectors are pointing toward the bulk solvent and the rest toward the protein bulk. This vector-based approach has been implemented into the program VICE by Tripathi and Kellogg [29]. These 2D and 3D concepts have been implemented into geometry-based pocket finder algorithms [21, 25, 28, 30–41] that are listed in Table 1. Energy-based approach: the concept of the energy-based pocket finder algorithms is to describe the atomic density at a

Table 1 Geometry-based pocket finder algorithms available online Cavity finder algorithm

Web references

APROPOS (Automatic PROtein Pocket Search) [30]

http://www.csb.yale.edu/userguides/datamanip/ apropos/manual.html

Binding response [31]

http://mackerell.umaryland.edu/CADD/CADD_ bindingresponse.html

CASTp [32]

http://sts.bioengr.uic.edu/castp/

CAVER [33]

http://loschmidt.chemi.muni.cz/caver/

Fpocket [34]

http://sourceforge.net/projects/fpocket/

GHECOM (Probe-based HECOMi finder) [25]

http://biunit.naist.jp/ghecom/

McVol [35]

http://www.bisb.uni-bayreuth.de/People/ullmann/ mcvol/mcvol.html

PASS (Putative Active Sites with Spheres) [36] http://www.ccl.net/cca/software/UNIX/pass/index. shtml PocketDepth [37]

http://proline.physics.iisc.ernet.in/pocketdepth/

PocketPicker [28]

http://gecco.org.chemie.uni-frankfurt.de/ pocketpicker/

SCREEN (Surface Cavity REcognition and EvaluatioN) [38]

http://interface.bioc.columbia.edu/screen/

SplitPocket [39]

http://pocket.uchicago.edu/

SURFNET [40]

http://www.biochem.ucl.ac.uk/_roman/surfnet/ surfnet.html

LIGSITE csc [41]

http://projects.biotec.tu-dresden.de/cgi-bin/index. php

VOIDOO [21]

http://xray.bmc.uu.se/usf/voidoo.html

Target Druggability Assessment

69

neighborhood of a probe using potential functions similar to the one used to calculate energy of molecular complexes. Various functions are suitable to estimate the presence of an atom at a given distance from a reference point. Depending on the degree of influence we want to assign to a protein atom, short-range or longrange potential functions can be designed. The step function, for example, mimics the hard-sphere geometrical view of atomic interactions: the value of the function is infinite or constant if two atoms overlap and zero otherwise. With this infinite short-range function, the probe cannot “feel” any protein atoms unless they touch the probe. Its first order derivative is also not continuous and therefore not adapted for gradient-based searching algorithm. The sigmoidal function is a continuous differentiable version of the step function. It includes a zone of influence where the probe can still feel surrounding protein atoms even it does not overlap with the probe. Finally, the third most common function is the hydrophobic van der Waals potential energy, which is implemented in most protein force field. The van der Waals energy function shows an infinite positive energy asymptote when the center of the van der Waal spheres of two atoms overlap and a negative energy minimum when the atoms are at optimal distance from each other. At longer interatomic distance, the energy tends to zero. Such a function is therefore suitable to check whether a solvent probe is in the protein bulk, close to the molecular surface, or far from any protein atom. The van der Waals energy will be positive in the first case, negative in the second, and close to zero in the third [42]. Note that the same difficulty as in the 2D approach remains: the energy value of the probe could not differentiate whether the probe is within the bulk solvent or located in the middle of a large cavity space. In both cases, the probe is far away from any protein atoms. Using energy-based approach, it is possible to implement the conformational sampling strategies that are implemented in the protein modeling package to sample the probe in the protein environment. Once the “cavity field” is implemented into a 3D grid, standard sampling techniques such as molecular dynamics [33] or Monte Carlo [35] can be used to search for protein cavities. Full force field as it is implemented in AutoLigand [43] could help in characterizing the hydrophobic or electrostatic properties of the surface. Energy-based algorithms available online [27, 43–48] are listed in Table 2. To conclude this section, we remind that the main purpose of the protein-finding algorithms is to localize the cavities within the protein and estimate its size (see Note 4). They do not give information on the druggability of the pocket itself. This will be the object of the next section.

70

Jean-Yves Trosset and Christian Cave´

Table 2 Energy-based pocket finder algorithms available online Cavity finder energy-based algorithms Web references SiteMap [44]

www.schrodinger.com/

ICM PocketFinder [45]

www.molsoft.com/

Q-SiteFinder [46]

www.modelling.leeds.ac.uk/qsitefinder/

SITEHOUND [47]

http://bsbbsinai.org/SHserver/SiteHound/download.html

AutoLigand [43]

http://mgltools.scripps.edu/downloads

GRID [27]

www.moldiscovery.com/soft_grid.php

Surflex protomol [48]

www.biopharmics.com

2.2 Druggability of Binding Site 2.2.1 Physicochemical Properties

Lipinski’s rule of five (RO5) connects the physicochemical properties of a drug with its pharmacokinetic properties [49]. Molecules of reasonable size ( 5) while capable to interact with polar solvent (number of H-bond donors > 5; sum of N’s and O’s > 10) will be characterized as drug-like molecules [49, 50]. The physicochemical properties of a druggable pocket should be the mirror image of the physical properties of the drug-like molecule itself [51, 52]. The concept of druggable pocket was born from this analogy. The complementary properties of the pocket therefore reflect the Lipinski’s rule of five of drug-likeness. Qualitatively, we should expect an adequate proportion of H-bond donors and acceptors as well as hydrophobic patches to accommodate both the polar and apolar characters of the drug. Graphical interfaces are valuable tools to inspect the proportion of polar versus hydrophobic area of the protein molecular surface (see Note 5). Druggability assessment can be made by visual inspection of the cavity with an adequate molecular surface. The relative number of H-bond donors and acceptors as well as the relative proportion of polar versus hydrophobic protein atoms at the surface of the cavity can easily be visualized using color-code surface: a given color for polar atoms (positive and negatively charged) and a third color for neutral atom like aliphatic or aromatic carbons (see Note 5). For an expert modeler, this is certainly the most precise assessment. Quantitatively, physicochemical features can be coded into in silico descriptors that can be used by a machine learning algorithm. Such target druggability prediction models [10] are useful to screen the whole PDB and to estimate the size of the druggable genome. These models supposed a pre-classification of the binding pockets into “druggable” and “undruggable” target set. For undruggable targets, it is always tempting to define them as such because of the absence of known ligands. This does not mean of course that the

Target Druggability Assessment

71

target is undruggable as it was demonstrated after the discovery of protein-protein interaction targets a decade ago [13]. Here there are two choices: either we want a good statistics, i.e., a good discrimination rate, and only very well-defined targets such as enzymes are entered in the druggable set or we are willing to “fish” exotic targets such as PPI at the risk of decreasing the statistical performance with normal targets. Hence, care should be taken when using statistical target druggability predictors. 2.2.2 Shape Properties

Not only are the physicochemical properties important to assess pocket druggability but also the geometry. The shape of the binding pocket and the existence of micro-cavities that could interact strongly with certain chemical groups of the drug can be crucial for suitable interactions with a small molecule. The shape of the binding site is always best described by visual inspection. It is indeed always difficult to code into a program all the geometrical aspect of a molecule that the eye can see. If statistical automated methods are useful for large-scale assessment of targets, visual inspection of the shape of the cavity using an adequate surface rendering is perfectly adapted when investigating a specific target of our choice. As seen above, atoms located at the contact energy surface, defined in the program FLO-QXP (see Note 5), give major contribution to the hydrophobic interaction (or contact energy) as compared to the atoms lying outside of the surface, i.e., within the bulk of the protein cavity. What ultimately counts for assessing the druggability of a target is the amount of protein surface that can wrap around the drug. In technical terms, this is the amount of van der Waals contact between the drug and the protein. Large spherical cavities, like the ones observed, for example, in RNA or DNA polymerases [53], are not very druggable pockets. The large cavity that hosts the double strands of RNA or DNA is well exposed to solvent and not really suitable for binding a small molecule. Cavities tend to have larger volume/surface ratio (in the spherical approximation) as their radius increases and tends to loose van der Waals interactions between ligand atoms and protein molecular surface as the cavity size increases. In this sense, large protein cavity tends to be more difficult to tackle with small molecules. A typical example for such a type of surface is found in kinase ATP-binding sites of protein kinases (Fig. 2). With a flat interface, each ligand atom contributes twice to the contact energy, from the face “below” and from the one “above.” Inducible pockets that specifically adapt the small molecules optimize this contact energy. A good example is given in the case of imatinib binding in the allosteric pocket of Bcr-Abl kinase [54]. Optimum atom ligand efficiency [55–57] is obtained when the atom is inside a micro-cavity (dot circle in Fig. 2). We define such micro-cavities as hot spots. They usually correspond to singularities of the molecular surface (see Note 6). The presence of such hot spots is characteristics of highly druggable targets, especially for

72

Jean-Yves Trosset and Christian Cave´

Fig. 2 QSURF protein surface from FLO-QXP around ligand in the hinge region of the ATP-binding site of a kinase. The scaffold makes two hydrogen bonds with the protein backbone (hinge). The micro-cavity on the right of the ligand (dot circle) is created by three atoms in the vicinity, the lines of singularity in the molecular surface results of the intersection between two protein atoms. The three color codes of the protein surface correspond to the hydrophobic area (yellow) and negative and positive polar areas (blue and red, respectively)

protein-protein interactions (see below). It is important to have graphical tools that can identify them. The QSURF program [58] has the advantage that it does not smooth out the surface and therefore does not remove these surface singularities. The hot spots are therefore easily identified with the FLO-QXP package. These singularities are the most informative elements when assessing target druggability. They are particularly important for the inhibition of protein-protein interactions as we will see in the following section. 2.2.3 Protein–Protein Interactions

The basic concept behind the inhibition of protein-protein interactions (PPI) with a small molecule is competition: David against Goliath, i.e., a small ligand competing with a big protein. The number of interactions is what makes the difference. For example, the competition between a kinase inhibitor and ATP is more or less equilibrated if not advantageous for the drug: the ATP is only a weak binder (Kd of the order of μM). It is enough for the kinase inhibitor to trigger additional interactions equivalent to one or two strong H-bonds to compete efficiently against ATP. Basically, most kinase inhibitors mimic ATP: a hydrophobic scaffold with two or three H-bonds interacting with the hinge of the kinase. Such ligand

Target Druggability Assessment

73

would not compete with a protein: the size of the ligand would have to be equivalent to estimated 30–50 ATPs. Therefore, a different strategy is necessary to identify useful inhibitors. If a small molecule cannot fulfill a large protein interface, the disruption of such interfaces by small drug-like molecule is still possible [13, 59, 60]. Particular features of the protein interface should be considered to make this possible [61]. As seen above, the presence of hot spots, i.e., the presence of micro-cavities in which ligand atoms can interact very strongly, is one of the most important criteria [62–67]. However, the presence of more than one anchoring points is important in order for the small molecule to be correctly “locked” within the binding site. When the ligand is well anchored in various hot spot positions, the more difficult it will be for a large competitor like protein, RNA or DNA strand, to displace this ligand. A PPI inhibitor can therefore be compared to a spider-shaped molecule: the “body” of the spider is a chemical moiety that connects the different “arms,” i.e., chemical side chains that interact with the anchors (Fig. 3). What contributes to the binding potency are mainly the “arms” of the molecule, not necessarily the “body” (i.e., the scaffold) as it is often the case in traditional binding sites like enzyme. Once the scaffold is selected, the affinity and the pharmacological properties of the compound are optimized using the decorating side chains (SSi). In the PPI inhibitors, the chemical groups anchored to hot spots are contributing most to the binding energy. The connector “C” plays the role of a linker between the different

Fig. 3 “Traditional” binding site model compared to the “spider” model of PPI. In traditional binding site, scaffold is part of the cavity and gives major contribution to the ligand binding affinity, whereas at the proteinprotein interface, the scaffold is more open to bulk solvent and plays the role of a connector (“C”) of highly interacting chemical groups anchoring to hot spots of the cavity of one of the protein partners

74

Jean-Yves Trosset and Christian Cave´

Fig. 4 Protein contact surface representation of the beta-catenin/Tcf4 protein complex interface near Tcf4 (D16). The QSURF color-coded surface highlights two protein micro-cavities that anchor the ligand PNU-74654 at two opposite sites (orange dot circle) involving a methyl-furan moiety and a benzene ring

anchors. It does not necessarily have a strong role in terms of affinity as the scaffold does in classical drugs. An example of a PPI inhibitor is shown in Fig. 4. The inhibitor PNU-74654 of the protein-protein complex β-catenin-Tcf4 is binding in the middle of a large groove of the PPI interface. This 350 nM inhibitor has been found by virtual screening using the above described spider concept [62]. It disrupts the β-catenin-Tcf4 interaction by targeting two anchoring points (marked by surrounded small circles in the upper panel of Fig. 4). Right- and left-hand cavities correspond to the groove of α-helices 23 and 30 of the armadillo repeat, respectively. The one from the right is hosting a benzene moiety that substitutes for the D16 aspartic side chain of Tcf4. On the left-hand side of the upper panel, the methyl on the furan moiety is contributing most of the activity of the compound. The same compound without this methyl group is about tenfold less potent. To illustrate the spider concept, the central benzamide of the PNU-74654 inhibitor of β-catenin can be considered as a linker that connects two anchoring points: the methyl-furan and the benzene ring that target the two hot spots of the protein receptor. The spider-type molecules are a general characteristic of the inhibitors of the protein-protein interactions [13, 15, 60]. It replaces the scaffold-containing molecules of the traditional

Target Druggability Assessment

75

binding sites. The design of protein–protein interaction inhibitors is mainly focused on the anchoring points of the protein. The challenge is to find a molecule that can target these hot spots simultaneously. Targeting these anchoring points is believed to be the key for successful design of PPI inhibitors. 2.2.4 Flexibility of Binding Sites

The intuitive concept discussed in this section is the ligand instability due to the protein conformational flexibility. From a dynamical point of view, the ligand interacts with its receptor through a type of “breathing” that can be modeled by calculating the normal modes of the protein-ligand complex. The normal modes of the protein can captures both short range and long range protein rearrangements which are often related to biological function of the enzyme. Drug binding can alter these protein motions. This can be adressed through normal mode analysis of the protein with and without the ligand. Those internal movements reflect also the function of the receptor, like the closing/opening of a catalytic region to accept the substrate and release the product of an enzymatic reaction, for example. Drugs usually interfere with these particular regions of the receptor that are usually localized at the hinge between movable domains and are therefore less perturbed by the internal motions of the receptor. Kinase hinge binding regions correspond to these zones of minimal movement amplitude. This hinge is at the intersection between the N- and C-terminal domains of the kinase, which move with respect to each other. The drugs are therefore in synchrony with the internal motions of the receptor. The study of the flexibility of the receptor-binding pocket is meant to see whether sheering movement of domains or protein loop movements can make certain region of the binding site unstable and make the binding with a ligand difficult. Thanks to the presence of various crystal structures for a same target with different ligand, it is possible to study the conformational variation of these protein-binding pockets [68]. When several X-ray structures are known for the same target, it is possible to detect the parts of the molecular surface of the binding site that move when transferring from one protein conformation to the other. The morphing technique is a simple way to analyze the variability of the binding site. The morphing describes the transition from one conformer to another using a predefined scheme of transformation. If the 3D coordinates of protein A and B are represented by XA and XB (an n  3 matrix where n is the number of atoms that are common to both proteins), then a linear morphing from protein A to protein B is a transformation that is defined by: X ðt Þ ¼ X A þ t ðX B  X A Þ ¼ tX B þ ð1  t ÞX A ¼ ½0; . . . 1

with t

76

Jean-Yves Trosset and Christian Cave´

where t is a parameter that runs from 0 to 1 which corresponds to the protein conformation A and B, respectively. It is possible to visualize the intermediate structures X(t) at every step t. Since all the intermediate structures are displayed sequentially like in a movie, the parameter t can be assimilated with “time” even if this morphing does not involve dynamical equations. The morphing technique can be applied to the protein surface as well. Differences in the shape and the physicochemical properties of the cavity surface inform on the hot spots that are not affected by the relative protein motions. Interaction with these hot spots will not be dependent on the change of binding site conformation. The morphing technique can be applied to the molecular surface as implemented in FLO-QXP. The moving molecular surface helps at identifying the flexible part of the protein-binding site. Prediction of the conformational change of the protein under binding is a challenge. New structural parameters based on highresolution X-ray structures have been presented recently [69]. Such studies open the door to a more extended analysis of conformational changes of protein cavity when binding with different ligands. 2.3 Target Promiscuity Assessment

Difficulty in developing a drug for a given target is not only related to the druggability of the target itself but also on the structural promiscuity of the target. Drugs that target one member of a large protein family, like kinases, are likely to cross-react with other family members due to the high degree of structural similarity within this family. If therapeutically promising multi-target effects can be identified [70, 71], high structural similarity with unwanted targets will represent a handicap for the drug development. The similarity of the binding sites could make the design of selective inhibitors difficult. The effective assessment of the structural landscape of the primary target binding site is therefore one of the ultimate steps in the target druggability assessment. One way to discover whether a compound can bind to different targets is to compare their binding sites [72–77]. Partial information can be obtained from sequence multi-alignment. When the degree of conservation between the two sequences is sufficiently high, identical amino acids in the sequence will likely correspond to identical binding site structure [78]. The contrary is not true. A change of amino acids in the sequence may or may not influence the binding mode of the inhibitor, depending whether the residue side chain is pointing toward the protein cavity. Finding a mapping between the atoms of the equivalent proteins is a preliminary step before the structural superposition. There are many ways of finding a mapping between a pair of macromolecular targets. I will describe two of them in the following section.

Target Druggability Assessment

77

2.3.1 Structural Alignment Based on 3D Grid Force Field

The field around the binding site can be calculated using various types of energy terms, including electrostatic, hydrophobic, hydrogen bonding, etc. This approach has been implemented into the GRID program [27]. Calculation of the molecular interaction fields (MIFs) has been used by Zamora et al. to study ADME properties of drugs [79, 80]. The field potentials can be calculated for each protein and used for comparing binding sites. Recently, a GRID potential approach has been implemented to compare protein-binding sites based on their GRID molecular interaction fields [81]. The structural similarity between a pair of proteins is studied by comparing the auto- and cross-correlation functions of the various MIFs of the two grids, respectively. Structure superimposition can also be obtained by using the fast Fourier transformation of the correlation functions as it is applied in the rigid body protein-protein docking problem [82–84] or by a global Monte Carlo minimization optimizer with a least square strategy on the two grids of the corresponding proteins [75].

2.3.2 Structural Alignment Based on Pharmacophore

Another approach consists of identifying chemical pharmacophore features that interact directly with the bound ligand. They can be summarized as hydrophobic centers, H-bond donors and acceptors, positive and negative charges, aromatic centers, etc. These surface chemical features (SCFs) are determined on the whole protein surface or on a chosen cavity. They form a 3D graph that reflects the shape and the size of the cavity of interest. Such surface chemical features (SCFs) can be calculated for all target structures of the Protein Data Bank (www.rcsb.org) as well as a query protein structure of interest. Triplets and quadruplets of SCF are then scanned against the pre-calculated SCF PDB database. This approach has been implemented in the commercial software MED-SuMo [85], an extension of the original free version SuMo [86]. It can be used either to find cavities from a query protein using a structural protein database [87] or to study the promiscuity of a target as shown below. The SCF of the protein-binding sites can be represented with colored bar codes for which each color bit represents a 3D SCF pharmacophore feature. Binding sites with the highest number of matches between these triplets and quadruplets of SCF have the highest similarity with the query binding site (Fig. 5). The larger the set of SCFs that match between the two proteins, the larger the similarity SuMo score. Such binding site comparison is a powerful tool to assess the specificity of the target, especially when it is part of a large protein family, like kinases, for example. Large-scale comparison has been carried out on the full PDB or on important protein families such as the purine-like target family [88]. This software is used in drug discovery to reveal potential off-targets or for drug repurposing [85].

78

Jean-Yves Trosset and Christian Cave´

Fig. 5 Structural binding site superposition using MED-Sumo. Surface chemical features (SCFs) are used to superimpose the protein-binding sites (right panel). Each type of SCF is associated to a give color. The SuMo score measures the quality of the 3D structural superimposition that is calculated using a bit-wise matching algorithm of the color-code bar fingerprint (left panel)

Methods for evaluating binding site similarities and related databases have been reviewed recently [18]. Table 3 lists available online structural comparative methods [86, 89–101]. It is worth to mention that binding site comparison relies on the entries of the structural protein database. The increasing number of available crystal structures will facilitate the identification of similar binding sites and potential off-targets and is most suitable for large protein families like kinases or proteins with purine containing binding sites [93]. 2.4 Systemic Druggability Assessment

Target inhibition by a drug may induce unexpected systemic effects such as the emergence of alternative pathways, which cancel the drug effect on the target. This is the case, for example, of the

Target Druggability Assessment

79

Table 3 Summary of available tools evaluating binding site similarities Binding site similarity algorithms

Web references

CavBase [89]

http://relibase.ccdc.cam.ac.uk relibase.rutgers.edu

CPASS (Comparison of Protein Active Site http://bionmr-c1.unl.edu/CPASS_OV/CPASS.htm Structures) [91] eF-seek [92]

http://ef-site.hgc.jp/eF-seek/top.do

FINDSITE [93]

http://cssb.biology.gatech.edu/skolnick/files/ FINDSITE/

MultiBind [94]

http://bioinfo3d.cs.tau.ac.il/MultiBind/

PROSURFER (PROteinSURFaceExploreR) [95]

http://www.tsurumi.yokohama-cu.ac.jp/fold/database. html

Query3d [96]

http://pdbfun.uniroma2.it/

SiteAlign [97]

http://bioinfo-pharma.u-strasbg.fr/template/jd/pages/ download/download.php

SiteBase [98]

http://www.modelling.leeds.ac.uk/sb

SiteEngine [101]

http://bioinfo3d.cs.tau.ac.il/SiteEngine/

SuMo [86]

http://sumo-pbil.ibcp.fr

@TOME-2 [100]

http://abcis.cbs.cnrs.fr/AT2

feedback activation of EGFR-mediated proliferation pathway after inhibition of the anticancer target BRAF-V600E [102]. Systemic effects can be defined as non-expected effects when only local interactions are considered. Such systemic effects result from the presence of positive/negative feed forward/backward regulation loops. Recent results from theoretical biology have related the positive loops (close circuits with an even number of inhibiting arrows) to the presence of multi-stationary states and the negative loops (close circuits with an odd number of inhibiting arrows) to the possible existence of oscillations (e.g., metabolic oscillations) for certain variables [103]. Target druggability assessment in this context consists of checking the longtime systemic impact of the inhibition or activation of a target to the disease-related phenotype. In silico construction of a biological network is only based on local interactions within proteins or genes (variables of the network), and systemic effects are studied by following the dynamic evolution of each components, and particular impact of target modulation can be characterized [104–106]. The methodology of a biological regulation model construction relies on four basic questions:

80

Jean-Yves Trosset and Christian Cave´

– Q0: What is the biological question the model is supposed to answer? (In pharmacology, models usually try to link the genotype with a disease-related phenotype). – Q1: What are the main actors with respect to the problem under study? – Q2: Which actors interact with which? – Q3: How each combination of inhibitions and activations modulate the state of a given variable? For well-studied therapeutic pathways, regulation signals of their components (Q2) are usually well known. The presence/ absence of the activation/inhibition signals pointing to a given variable (Q3) determines the status of that variable (e.g., active or inactive in Boolean modeling). Such regulatory signals (Q3) can be coded in various forms: 1. Inserted implicitly within time evolution ODE [107]. 2. Expressed through conditional probabilities, which in relation with causality axioms support Bayesian networks theory [108]. 3. Activation/inhibition rules can also be computationally coded through Boolean functions [109]. 4. Be incorporated explicitly through kinetic constants as in the generalized Rene´ Thomas’s regulation modeling framework [103, 110]. With these information in hands (Q3), the dynamics of the whole system can be studied. The time evolution of each variable (e.g., protein targets) can be analyzed as well as its influence on cellular phenotype [110].

3

Conclusion In silico target druggability can be assessed through direct and indirect approaches. The first one evaluates the capacity of the target to interact with a small drug-like molecule. By analyzing the 3D structures of proteins, it is possible to detect the presence of protein cavities that could complement the physicochemical properties of a drug-like molecule. Large conformational change of protein receptors may be involved upon drug binding, and molecular dynamic techniques associated with free energy calculation are often used to assess the presence of a stable-induced conformational state of the receptor upon ligand binding [111–114]. In the case of inhibition of protein-protein interactions, structural druggability assessment follows different rules and is more focused on the presence of interacting hot spots. As shown for

Target Druggability Assessment

81

β-catenin-TCF4 interactions, drug cavity within protein interface is changed to a spider-type of interaction for which chemical moieties make strong interaction with those hot spots without the presence of a well-defined protein cavity [62]. Even structurally well-validated targets may lead to difficulties in the early stages of drug development due to the close structural similarity of the target with unwanted targets or with other members of a large protein family as kinases. Thanks to the everincreasing number of X-ray protein 3D structures, large-scale comparison of ligand-binding sites can be made for entire protein families as well as between protein families, and common local structural features between protein can highlight which chemical classes will most likely highlight the promiscuity between a given set of protein targets [87]. Finally, techniques from system biology will soon be standardized for other types of target druggability assessment to consider the systemic effect of the inhibition of the target by a drug and in particular whether alternating pathways or regulation signals may have a counter effect on the action of the target on the phenotype. Modeling regulation of biological network is an emerging activity in system biology, which helps in understanding the systemic effect of drug targets.

4

Notes 1. When using pocket-finding programs, investigation of the probe size parameter is recommended. With increasing size of the probe, the hierarchy of the various pockets will be more apparent. On the other hand, with small probe size values, micro-cavities will appear. These micro-clefts might correspond to hot spots where strong interactions might take place with some atoms of the drug. It is left to the judgment of the user to discern whether the micro-pockets that appear correspond to errors in protein coordinates (which are often less than 0.3 A˚) or to a real protein cavity. With much caution, this parameter can be used to look for the existence of potential inducible binding sites. 2. The convex hull of a set of points is the volume surrounding these points such that any segment between any two of these points stay inside the volume. For protein, we might relax this definition to include “inside” the protein volume, the closed protein cavities that are not open to bulk solvent. A more adequate definition for protein would be any segment between any two points at the surface of the protein should stay inside the protein. Note that the protein example in Fig. 1 has been chosen such that the convex hull is not a good discrimination

82

Jean-Yves Trosset and Christian Cave´

between protein bulk and cavity space (even with the newly adapted definition pre´cised above). The large solvent area between the two lobs is considered as part of the bulk solvent. The cavity where drugs will interact corresponds to small clefts or hot spots within this protein-protein interface. This schematic representation of the protein mimics the 3D structure of β-catenin [22]. Note that docking packages have most of the ingredients for designing energy-based cavity-finding algorithms: energy force field, techniques for sampling the conformational space (molecular dynamics, Monte Carlo, systematic search) as well as the calculation of protein molecular surface around the ligand, etc. 3. The nodes represent a discrete sample of the protein space. Any kind of information can be stored on each node of a 3D grid such as the number of surrounding protein atoms, the van der Waals contribution of the all protein atoms at that point, the local atomic density, etc. It is possible to get an estimation of those quantities everywhere on the 3D grid using a linear or nonlinear interpolation scheme like B-splines. The B-spline interpolation scheme has the advantage to ensure the continuity of the function when crossing the cubic element. It is also possible to request the continuity of all the derivatives of this function. This allows some gradient-based optimization techniques to be implemented to find energy-optimum positions within the 3D grid. Such optima correspond to interesting hot spots of the protein cavity (region of lower energy). The B-spline interpolation scheme has been adapted by Oberlin et al. [115] for the estimation of van der Waals energies within 3D grids and implemented into the protein modeling/docking package PRODOCK [112–114]. The efficiency of the approach has been demonstrated in the case of rigid protein receptor [112]. 4. Note that it is possible for certain protein modeling software (especially docking) to display the surface of buried or closed protein cavities by locating a ligand inside this pocket. Indeed, most graphical interface of docking software can display the protein surface around a docked ligand. For a pocket finder algorithm, the ligand should play the role of a probe. A protein cavity-finding algorithm has been implemented into FLO-QXP by automatizing this concept: the protein is immersed into a 3D grid for which each node, considered as a single atom, is defined as part of a big ligand (the whole 3D grid). The QSURF program from FLO-QXP package [58] is calculating the solvent accessible surface of the protein around each grid node. All types of close cavities are then displayed for visual investigation. This has the advantage to not only locate the

Target Druggability Assessment

83

cavities but also to understand the shape and physical properties of this cavity. 5. When visualizing the interactions of the drug-receptor complex, it is very convenient to use the solvent accessible surface. In this case, the optimal van der Waals energy between a ligand atom and a protein atom corresponds very closely to the solvent accessible surface of the protein. Maximum energy contact is therefore obtained when a ligand atom is on the solvent accessible protein surface. Visually it is very easy to estimate whether an atom is on the surface or outside. The QSURF representation protein surface of FLO-QXP [58] is perfectly adapted for interactive drug design for this reason. The optimum interatomic distance that provides the best correlation with experimental drug potency and the contact energy term is about 1.4 A˚. This is also the size of the water probe used to calculate the accessible surface area. This contact surface has been designed by C. McMartin and R. Bohacek for intuitive drug design [116]. This surface is very similar to the solvent accessible surface. 6. There are basically two interesting singularities: on the intersecting line when two spheres overlap (contact energy surface of two protein atoms) or on the intersecting point when three spheres overlap. Of course, for this singularity the three atoms have to depart from each other to leave space for an atom of the ligand to stay in between. This little spherical hole of a size of an atom or chemical group like -CH3 or -SO3 represents a micro-cavity defined as a hot spot. These surface singularities are usually never represented in protein modeling packages. For esthetical constraints, the graphical software tends to smooth out the protein surface from all types of singularities (micro-cavities included). One of the main advantages of the molecular surface calculated by QSURF in FLO-QXP is to display the singularities of the surface. This is very informative as they correspond usually to positions in the protein where ligand atoms get maximum efficiency. References 1. Sakharkar MK, Sakharkar KR (2007) Targetability of human disease genes. Curr Drug Discov Tech 4:48–58 2. Wyatt PG, Gilbert IH, Read KD et al (2011) Target validation: linking target and chemical properties to desired product profile. Curr Top Med Chem 11:1275–1283 3. Taboureau O, Nielsen SK, Audouze K et al (2011) ChemProt: a disease chemical biology database. Nucleic Acids Res 39:D367–D372

4. Bender A, Young DW, Jenkins JL et al (2007) Chemogenomic data analysis: prediction of small-molecule targets and the advent of biological fingerprint. Combi Chem High Throughput Screen 10:719–731 5. Dixon SJ, Stockwell BR (2009) Identifying druggable disease-modifying gene products. Curr Opin Chem Biol 13:549–555 6. Caffrey CR, Rohwer A, Oellien F et al (2009) A comparative chemogenomics strategy to predict potential drug targets in the metazoan

84

Jean-Yves Trosset and Christian Cave´

pathogen, Schistosoma mansoni. PLoS One 4:e4413 7. Bredel M, Jacoby E (2004) Chemogenomics: an emerging strategy for rapid target and drug discovery. Nat Rev Genet 5:262–275 8. Li Q, Cheng T, Wang Y et al (2010) PubChem as a public resource for drug discovery. Drug Discov Today 15:1052–1057 9. Fauman EB, Rai BK, Huang ES (2011) Structure-based druggability assessmentidentifying suitable targets for small molecule therapeutics. Curr Opin Chem Biol 15:463–468 10. Schmidtke P, Barril X (2010) Understanding and predicting druggability. A highthroughput method for detection of drug binding sites. J Med Chem 53:5858–5867 11. Huang N, Jacobson MP (2010) Binding-site assessment by virtual fragment screening. PLoS One 5:e10109 12. Fuller JC, Burgoyne NJ, Jackson RM (2009) Predicting druggable binding sites at the protein-protein interface. Drug Discov Today 14:155–161 13. Sugaya N, Furuya T (2011) Dr. PIAS: an integrative system for assessing the druggability of protein-protein interactions. BMC Bioinformatics 9:12–50 14. Villoutreix BO, Bastard K, Sperandio O et al (2008) In silico-in vitro screening of protein–protein interactions: towards the next generation of therapeutics. Curr Pharm Biotechnol 9:103–122 15. Sperandio O, Reyne`s CH, Camproux A et al (2010) Rationalizing the chemical space of protein-protein interaction inhibitors. Drug Discov Today 15:220–229 16. Panjkovich A, Daura X (2010) Assessing the structural conservation of protein pockets to study functional and allosteric sites: implications for drug discovery. BMC Struct Biol 10:9–33 17. Che`ne P (2008) Challenges in design of biochemical assays for the identification of small molecules to target multiple conformations of protein kinases. Drug Discov Today 13:522–529 18. Pe´rot S, Sperandio O, Miteva M et al (2010) Druggable pockets and binding site centric chemical space: a paradigm shift in drug discovery. Drug Discov Today 15:656–667 19. Morris RRJ, Najmanovich R, Kahraman A et al (2005) Real spherical harmonic expansion coefficients as 3D shape descriptors for protein binding pocket and ligand comparisons. Bioinformatics (Oxford, England) 21:2347–2355

20. Mak L, Grandison S, Morris RJ (2008) An extension of spherical harmonics to regionbased rotationally invariant descriptors for molecular shape description and comparison. J Mol Graph Model 26:1035–1045 21. Kleywegt GJ, Jones TA (1994) Detection, delineation, measurement and display of cavities in macromolecular structures. Acta Crystallogr D Biol Crystallogr 50:178–185 22. Poy F, Lepourcelet M, Shivdasani RA et al (2001) Structure of a human Tcf4-beta-catenin complex. Nat Struct Biol 8:1053–1057 23. Tagami S, Sekine SI, Kumarevel T et al (2010) Crystal structure of bacterial RNA polymerase bound with a transcription inhibitor protein. Nature 468:978–982 24. Liang J, Edelsbrunner H, Woodward C (1998) Anatomy of protein pockets and cavities: measurement of binding site geometry and implications for ligand design. Protein Sci 7:1884–1897 25. Kawabata T, Go N (2007) Detection of pockets on protein surfaces using small and large probe spheres to find putative ligand binding sites. Bioinformatics 529:516–529 26. Kawabata T (2010) Detection of multiscale pockets on protein surfaces using mathematical morphology. Proteins 78:1195–1211 27. Goodford PJ (1985) A computational procedure for determining energetically favorable binding sites on biologically important macromolecules. J Med Chem 28:849–857 28. Weisel M, Proschak E, Schneider G (2007) PocketPicker: analysis of ligand binding-sites with shape descriptors. Chem Cent J 1:7–24 29. Tripathi A, Kellogg GE (2010) A novel and efficient tool for locating and characterizing protein cavities and binding sites. Proteins 78:825–842 30. Peters K, Fauck J, Fro¨mmel C (1996) The automatic search for ligand binding sites in proteins of known three-dimensional structure using only geometric criteria. J Mol Biol 256:201–213 31. Zhong S, MacKerell AD (2007) Binding response: a descriptor for selecting ligand binding site on protein surfaces. J Chem Inf Model 47:2303–2315 32. Binkowski TA, Naghibzadeh S, Liang J (2003) CASTp: computed atlas of surface topography of proteins. Bioinformatics 31:3352–3355 33. Petrek M, Otyepka M, Bana´s P et al (2006) CAVER: a new tool to explore routes from protein clefts, pockets and cavities. BMC Bioinformatics 7:316–325

Target Druggability Assessment 34. Guilloux VL, Schmidtke P, Tuffery P (2009) Fpocket: An open source platform for ligand pocket detection. BMC Bioinformatics 11:1–11 35. Till MS, Ullmann GM (2010) McVol—a program for calculating protein volumes and identifying cavities by a Monte Carlo algorithm. J Mol Model 16:419–429 36. Brady GP, Stouten PF (2000) Fast prediction and visualization of protein binding pockets with PASS. J Comput Aided Mol Des 14:383–401 37. Kalidas Y, Chandra N (2008) PocketDepth: a new depth based algorithm for identification of ligand binding sites in proteins. J Struct Biol 161:31–42 38. Nayal M, Honig B (2006) On the nature of cavities on protein surfaces: application to the identification of drug-binding sites. Proteins 63:892–906 39. Tseng YY, Dupree C, Chen ZJ et al (2009) SplitPocket: identification of protein functional surfaces and characterization of their spatial patterns. Nucleic Acids Res 37: W384–W389 40. Laskowski RA (1995) SURFNET: a program for visualizing molecular surfaces, cavities, and intermolecular interactions. J Mol Graph 13:323–30, 307–8 41. Huang B, Schroeder M (2006) LIGSITE csc: predicting ligand binding sites using the Connolly surface and degree of conservation. BMC Struct Biol 11:1–11 42. Kitchen DB, Decornez H, Furr JR et al (2004) Docking and scoring in virtual screening for drug discovery: methods and applications. Nat Rev Drug Discov 3:935–949 43. Harris R, Olson AJ, Goodsell DS (2008) Automated prediction of ligand-binding sites in proteins. Proteins 70:1506–1517 44. Halgren T (2007) New method for fast and accurate binding-site identification and analysis. Chem Biol Drug Des 69:146–148 45. An J, Totrov M, Abagyan R et al (2005) Pocketome via comprehensive identification and classification of ligand binding envelopes. Mol Cell Proteomics 4:752–761 46. Laurie AT, Jackson RM (2005) Structural bioinformatics Q-SiteFinder: an energybased method for the prediction of protein–ligand binding sites. Bioinformatics 21:1908–1916 47. Hernandez M, Ghersi D, Sanchez R (2009) SITEHOUND-web: a server for ligand binding site identification in protein structures. Nucleic Acids Res 37:W413–W416

85

48. Ruppert J, Welch W, Jain AN (1997) Automatic identification and representation of protein binding sites for molecular docking. Protein Sci 6:524–533 49. Lipinski C (2004) Lead- and drug-like compounds: the rule-of-five revolution. Drug Discov Today Technol 1:337–341 50. Zhang MQ, Wilkinson B (2007) Drug discovery beyond the “rule-of-five”. Curr Opin Biotechnol 18:478–488 51. Hopkins AL, Groom CR (2002) The druggable genome. Nat Rev Drug Discov 1:727–730 52. Costa PR, Acencio ML, Lemke N (2010) A machine learning approach for genome-wide prediction of morbid and druggable human genes based on systems-level data. BMC Genomics 11(Suppl 5):S9 53. Lesburg CA, Cable MB, Ferrari E et al (1999) Crystal structure of the RNA-dependent RNA polymerase from hepatitis C virus reveals a fully encircled active site. Nat Struct Biol 6:937–943 54. Salah E, Ugochukwu E, Barr AJ et al (2011) Crystal structures of ABL-related gene (ABL2) in complex with imatinib, tozasertib (VX-680), and a type I inhibitor of the triazole carbothioamide class. J Med Chem 54:2359–2367 55. Abad-Zapatero C, Metz JT (2005) Ligand efficiency indices as guideposts for drug discovery. Drug Discov Today 10:464–469 56. Ghersi D, Sanchez R (2009) Improving accuracy and efficiency of blind protein-ligand docking by focusing on predicted binding sites. Proteins 74:417–424 57. Hopkins AL, Groom CR, Alex A (2004) Ligand efficiency: a useful metric for lead selection. Drug Discov Today 9:430–431 58. Bohacek RS, McMartin C (1992) Definition and display of steric, hydrophobic, and hydrogen bonding properties of ligand binding sites in proteins using Lee and Richards accessible surface: validation of a high-resolution graphical tool for drug design. J Med Chem 35:1671–1684 59. Licata L, Briganti L, Peluso D et al (2011) MINT, the molecular interaction database: 2012 update. Nucleic Acids Res 40(Database issue):D857–D861 60. Bourgeas R, Basse M-J, Morelli X et al (2010) Atomic analysis of protein-protein interfaces with known inhibitors: the 2P2I database. PLoS One 5:e9598 61. Kozakov D, Hall DR, Chuang GY et al (2011) Structural conservation of druggable hot

86

Jean-Yves Trosset and Christian Cave´

spots in protein-protein interfaces. PNAS 108:13528–13533 62. Trosset J-Y, Dalvit C, Knapp S et al (2006) Inhibition of protein-protein interactions: the discovery of druglike beta-catenin inhibitors by combining virtual and biophysical screening. Proteins 64:60–67 63. Fasolini M, Wu X, Flocco M et al (2003) Hot spots in Tcf4 for the interaction with betacatenin. J Biol Chem 278:21092–21098 64. Clackson T, Wells JA (1995) A hot spot of binding energy in a hormone-receptor interface. Science (New York, N.Y.) 267:383–386 65. Kuttner YY, Engel S (2011) Protein hot spots—the islands of stability. J Mol Biol 415:419–428 66. Wanner J, Fry DC, Peng Z et al (2011) Druggability assessment of protein–protein interfaces. Future Med Chem 3:2021–2038 67. Metz A, Pfleger C, Kopitz H et al (2011) Hot spots and transient pockets: predicting the determinants of small-molecule binding to a protein-protein interface. J Chem Inf Model 52:120–133 68. Henrich S, Salo-Ahen OMH, Huang B et al (2010) Computational approaches to identifying and characterizing protein binding sites for ligand design. J Mol Recognit 23:209–219 69. McMartin C (2011) A geometry force field which converts low resolution X-ray models to structures with properties found at ultra high resolution. Protein Sci 21:75–83 70. Knight ZA, Lin H, Shokat KM (2010) Targeting the cancer kinome through polypharmacology. Nat Rev Cancer 10:130–137 71. Hopkins AL (2008) Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol 4:682–690 72. Chikhi R, Sael L, Kihara D (2010) Real-time ligand binding pocket database search using local surface descriptors. Proteins 78:2007–2028 73. Das S, Kokardekar A, Breneman CM (2009) Rapid comparison of protein binding site surfaces with property encoded shape distributions. J Chem Inf Model 49:2863–2872 74. Ivanisenko VA, Pintus SS, Grigorovich DA et al (2004) PDBSiteScan: a program for searching for active, binding and posttranslational modification sites in the 3D structures of proteins. Nucleic Acids Res 32: W549–W554 75. Totrov M (2011) Ligand binding site superposition and comparison based on Atomic Property Fields: identification of distant homologues, convergent evolution and

PDB-wide clustering of binding sites. BMC Bioinformatics 12:S35 76. Gold ND, Jackson RM (2005) A searchable database for comparing protein-ligand binding sites for the analysis of structure-function relationships. J Chem Inf Model 46:736–742 77. Schmidtke P, Souaille C, Estienne F et al (2010) Large-scale comparison of four binding site detection algorithms. J Chem Inf Model 50:2191–2200 78. Feldman HJ, Labute P (2010) Pocket similarity: are alpha carbons enough? J Chem Inf Model 50:1466–1475 79. Fontaine F, Pastor M, Zamora I et al (2005) Anchor-GRIND: filling the gap between standard 3D QSAR and the GRid-INdependent descriptors. J Med Chem 48:2687–2694 80. Crivori P, Zamora I, Speed B et al (2004) Model based on GRID-derived descriptors for estimating CYP3A4 enzyme stability of potential drug candidates. J Comput Aided Mol Des 18:155–166 81. Baroni M, Cruciani G, Sciabola S et al (2007) A common reference framework for analyzing/comparing proteins and ligands. Fingerprints for Ligands and Proteins (FLAP): theory and application. J Chem Inf Model 47:279–294 82. Ritchie DW, Venkatraman V (2010) Ultra-fast FFT protein docking on graphics processors. Bioinformatics (Oxford, England) 26:2398–2405 83. Katchalski-Katzir E (1992) Molecular surface recognition: determination of geometric fit between proteins and their ligands by correlation techniques. PNAS 89:2195–2199 84. Brenke R, Kozakov D, Chuang GY et al (2009) Fragment-based identification of druggable “hot spots” of proteins using Fourier domain correlation techniques. Bioinformatics (Oxford, England) 25:621–627 85. Doppelt-Azeroual O, Moriaud F, Adcock SA et al (2009) A review of MED-SuMo applications. Infect Disord Drug Targets 9:344–357 86. Jambon M, Andrieu O, Combet C et al (2005) The SuMo server: 3D search for protein functional sites. Bioinformatics (Oxford, England) 21:3929–3930 87. Moriaud F, Doppelt-Azeroual O, Martin L et al (2009) Computational fragment-based approach at PDB scale by protein local similarity. J Chem Inf Model 49:280–294 88. Doppelt-Azeroual O, Delfaud F, Moriaud F et al (2010) Fast and automated functional classification with MED-SuMo: an application on purine-binding proteins. Protein Sci 19:847–867

Target Druggability Assessment 89. Schmitt S, Kuhn D, Klebe G (2002) A new method to detect related function among proteins independent of sequence and fold homology. J Mol Biol 323:387–406 90. Bartlett S, Beddard GS, Jackson RM et al (2005) Comparison of the ATP binding sites of protein kinases using conformationally diverse bisindolylmaleimides, methods. J Am Chem Soc 127(33):11699–11708 91. Powers R, Copeland JC, Germer K et al (2006) Comparison of protein active site structures for functional annotation of proteins and drug design. Proteins 65:124–135 92. Kinoshita K, Furui J, Nakamura H (2002) Identification of protein functions from a molecular surface database, eF-site. J Struct Funct Genomics 2:9–22 93. Skolnick J, Brylinski M (2009) FINDSITE: a combined evolution/structure-based approach to protein function prediction. Brief Bioinform 10:378–391 94. Shulman-Peleg A, Shatsky M, Nussinov R et al (2008) MultiBind and MAPPIS: webservers for multiple alignment of protein 3D-binding sites and their interactions. Nucleic Acids Res 36:W260 95. Minai R, Matsuo Y, Onuki H et al (2008) Method for comparing the structures of protein ligand-binding sites and application for predicting protein–drug interactions. Proteins 72:367–381 96. Ausiello G, Via A, Helmer-Citterich M (2005) Query3d: a new method for highthroughput analysis of functional residues in protein structures. BMC Bioinformatics 6 (Suppl 4):S5 97. Schalon C, Surgand JS, Kellenberger E et al (2008) A simple and fuzzy method to align and compare druggable ligand-binding sites. Proteins 71:1755–1778 98. Brakoulias A, Jackson RM (2004) Towards a structural classification of phosphate binding sites in protein–nucleotide complexes: an automated all-against-all structural comparison using geometric matching. Proteins 56:250–260 99. Tomei L, Roussel A, Incitti I et al (1999) Crystal structure of the RNA-dependent RNA polymerase of hepatitis C virus. PNAS 96(23):13034–13039 100. Pons JL, Labesse G (2009) @TOME-2: a new pipeline for comparative modeling of proteinligand complexes. Nucleic Acids Res 37: W485–W491 101. Shulman-Peleg A, Nussinov R, Wolfson HJ (2004) Recognition of functional sites in protein structures. J Mol Biol 339:607–633

87

102. Berns K, Bernards K (2012) Understanding resistance to targeted cancer drugs through loss of function genetic screens. Nature 483:100–104 103. Richard A, Comet J-P, Bernot G (2006) Formal methods for modeling biological regulatory networks. Modern formal methods and applications. Springer, Dordrecht, pp 83–122 104. Le DH (2015) An approach based on boolean dynamics of biological networks for disease candidate gene prioritization. In: Le L, Pharm S (eds) Genomic medicine. Vietnam National University, Ho-Chi-Minh City 105. Choi M, Shi J, Zhu Y et al (2017) Network dynamics-based cancer panel stratification for systemic prediction of anticancer drug response. Nat Commun 8:1940–1104 106. Knaack SA, Siahpirani AF, Roy S (2014) A pan-cancer modular regulatory network analysis to identify common and cancer-specific network components. Cancer Inform 13:69–84 107. Peng Y, Scott P, Tao R et al (2017) Dissect the dynamic molecular circuits of cell cycle control through network evolution model. Biomed Res Int 2954351:1–9 108. Yu B, Xu JM, Li S et al (2017) Inference of time-delayed gene regulatory networks based on dynamic Bayesian network hybrid learning method. Oncotarget 8:80373–80392 109. Albert R, Thakar J (2014) Boolean modeling: a logic-based dynamic approach for understanding signaling and regulatory networks and for making useful predictions. Wiley Interdiscip Rev Syst Biol Med 6:353–369 110. Abou-Jaoude´ BW, Traynard P, Monteiro PT et al (2016) Logical modeling and dynamical analysis of cellular networks. Front Genet 7:94–114 111. Hantschel O, Grebien F, Superti-Furga G (2011) Targeting allosteric regulatory “Drugging the Undruggable” modules in oncoproteins. Cancer Res 2:829–830 112. Trosset J-Y, Scheraga HA (1999) Prodock: software package for protein modeling and docking. J Comput Chem 20:412–427 113. Trosset J-Y, Scheraga HA (1999) Flexible docking simulations: scaled collective variable Monte Carlo minimization approach using Bezier splines, and comparison with a standard Monte Carlo algorithm. J Comput Chem 20:244–252 114. Roisman LC, Piehler J, Trosset J-Y et al (2001) Structure of the interferon-receptor complex determined by distance constraints

88

Jean-Yves Trosset and Christian Cave´

from double-mutant cycles and flexible docking. PNAS 98:13231–13236 115. Trosset J-Y, Scheraga HA (1998) Reaching the global minimum in docking simulations: a Monte Carlo energy minimization approach using Bezier splines. PNAS 95:8011–8015

116. McMartin C, Bohacek RS (1997) QXP: powerful, rapid computer algorithms for structure-based drug design. J Comput Aided Mol Des 11:333–344

Chapter 6 In Silico Drug–Target Profiling Jean-Yves Trosset and Christian Cave´ Abstract Pharmacological science is trying to establish the link between chemicals, targets, and disease-related phenotypes. A plethora of chemical proteomics and structural data have been generated, thanks to the target-based approach that has dominated drug discovery at the turn of the century. There is an invaluable source of information for in silico target profiling. Prediction is based on the principle of chemical similarity (similar drugs bind similar targets) or on first principles from the biophysics of molecular interactions. In the first case, compound comparison is made through ligand-based chemical similarity search or through classifier-based machine learning approach. The 3D techniques are based on 3D structural descriptors or energy-based scoring scheme to infer a binding affinity of a compound with its putative target. More recently, a new approach based on compound set metric has been proposed in which a query compound is compared with a whole of compounds associated with a target or a family of targets. This chapter reviews the different techniques of in silico target profiling and their main applications such as inference of unwanted targets, drug repurposing, or compound prioritization after phenotypic-based screening campaigns. Key words Drug-target profile, Target identification, Chemical similarity search, Panel docking

1

Introduction A plethora of chemical proteomic data are now available to the scientific community. These data cover many aspects of drug development, ranging from large-scale crystallography [1] to parallel biochemical assays. The 3D structure of many members of protein-target families has been solved, and the large-scale comparisons of protein-binding sites gave lots of insights to understand drug specificity [2, 3]. In parallel to this, progress in assay miniaturization has promoted the development of high-throughput screening of compounds on large panels of protein targets [4] or protein-target families such as kinases [5, 6]. Millions of experimental data on drug–target interactions are now available in public databases such

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_6, © Springer Science+Business Media, LLC, part of Springer Nature 2019

89

90

Jean-Yves Trosset and Christian Cave´

as ChEMBL or PubChem [7–9] and that can be used for in silico drug–target profiling [10, 11]. The mining of structural and chemo-proteomics data has revealed that drugs usually bind to several targets, with an average range from 6 to 12 targets [12, 13] depending on the chemical class or the type of targets as protein families are more promiscuous than others [14]. This cross-reactivity is usually larger at early stages of drug discovery [15, 16]. This multiple drug–target interaction raised the concept of poly-pharmacology, which we will distinguish from the lack of specificity of a non-optimized compound. Drug polypharmacology certainly contributes to the success of many marketed drugs as multiple interactions may contribute to synergistic activations/inhibition of pathways or physiological functions. This synergy is the result of the complexity of the living system which is characterized by highly interrelated signaling network. Systemic effects of drugs are not always associated with toxicity but may be beneficial to treating the disease [17, 18]. The understanding and the prediction of the synergistic role of poly-pharmacology is one of the main applications of in silico target profiling. The second application in the list concerns drug repositioning. Drugs from different therapeutic areas may bind to similar targets. Finding such target is a common strategy to find new indication for a known drug which is another application of in silico drug–target profiling [8, 19–23]. Finally, the resurgence of phenotypic screening at the early stage of drug discovery is raising another challenge that can be approached by in silico target profiling. Compounds that emerge from phenotypic-based screening have usually a completely unknown mechanism of action (MoA). Prioritizing these compounds based on their predicted target profile is the third main application of in silico target profiling. This is likely the most difficult one as we are faced with the difficult task of differentiating between real poly-pharmacology and lack of specificity of compounds that is often the fate of initial hits at the early phase of drug discovery. In silico target prediction approaches are all based on two simple basic principles: the principle of similarity, which states that similar compounds bind to similar targets or have similar biological functions, and the first physics principles relying on biophysical molecular interactions. Certain approaches such as pharmacophore modeling make use of these two principles. In entity-based approach, target inference algorithms rely on pairwise similarity between a query compound and a known drug or on the docking energy of that query compound with a specific target [10, 19, 24–27]. More recent developments use a systemic view in which related compounds and/or related targets are taken into accounts in the prediction process. In this case, query

In Silico Target Profiling

91

compound is compared to a whole set of inhibitors associated with a target or a family of targets. The first family of techniques is well adapted for drug repositioning for which the target profile is well characterized. The second approach based on compound set metric is instead more adapted for prioritizing compounds that emerge from phenotypicbased screening. These initial hit compounds have usually a less well-defined target profile due to the lack of specificity. The techniques based on ensemble similarity metric are less dependent on computational artifacts associated with scoring similarity functions and have a better chance to capture weak signals within a given target profile. These two classes of approaches will correspond to the main two sections of this chapter. As they all use the similarity principle and first physics principles, the first section takes the occasion to review the classical way to introduce these principles in the target prediction techniques and the different contexts to which these techniques are applied. They correspond to (1) chemical similarity search in which full-ligand pairwise similarity is considered; (2) machine learning approach based on support vector machine, naı¨ve Bayesian, or fragment-based framework to identify the chemical descriptors that explain the specificity of a class of inhibitors associated with each target or family of targets or even to a whole therapeutic area; and (3) the 3D structure-based approaches. These correspond to the pharmacophore approach in which 3D conformation of the query compound is compared with a reference drug in its known 3D bound conformation within its target and the panel docking approach that uses energy-based physic principles to predict the binding of a compound onto each target for which the 3D structure is known. The second section will then focus on the compound set approaches based on the structure activity relationship (SAR) and the similarity ensemble approach.

2

Material and Methods

2.1 Single-Target Entity-Based Approaches

Approaches described in this section assess the comparison or the binding of a query compound with individual reference compounds or individual targets. Those reference compounds or reference targets are described using 2D or 3D chemical or pharmacophore descriptors, and a comparison with a query compound is made through the principle of similarity. This section describes the different contexts in which this principle applies.

2.1.1 Chemical Similarity Searching

The first group of algorithms uses chemical similarity searching with ligand-based 2D descriptions. The 2D chemical descriptors inform on the chemical contents of the molecules through

92

Jean-Yves Trosset and Christian Cave´

Table 1 List of databases for drug–target interactions

Name

Web site

PubMed ID

MDDR

http://accelrys.com/products/collaborative-science/databases/

15032555 [34]

CARLSBAD

http://sunsetmolecular.com/products/?id¼4

23794735 [35]

BioPrint

http://www.cerep.fr/cerep/users/pages/collaborations/bioprint. asp

12951810 [36]

AurSCOPE

http://www.aureus-pharma.com/Pages/Products/Aurscope.php

23495997 [37]

StARlite

http://www.inpharmatica.co.uk/StARLITe/Index.htm

17082834 [38]

GVK bio

http://gvkbio.com/informatics/dbprod.htm

21569515 [39]

DrugBank

https://www.drugbank.ca/targets

18048412 [40]

STITCH 5

http://stitch.embl.de

26590256 [41]

ChemProt 3.0

http://potentia.cbs.dtu.dk/ChemProt

26876982 [33]

Refs.

extended connectivity fingerprints [28] or path fingerprints known as feature-class fingerprints. Comparisons between these fingerprints have been shown recently [29]. Based on these chemical descriptors, the similarity distance between two molecules can be estimated using a Tanimoto type of similarity functions (see Note 1). There are many variants of similarity distance functions [30], and their common weak point is that the similarity score drops off very quickly even if the molecules remain related structurally from visual inspection (see Note 2). Consequently, the similarity searching approach is efficient at predicting targets if the query compound has a close similarity with existing drugs or reference compounds (of known MoA). Below a certain similarity threshold (e.g., usual range of 0.7–0.9), pair of compounds will be considered as non-similar. Failing at defining the relevant 2D or 3D descriptors with respect to the target may lead to false negative, resulting in missing the identification of the relevant target for that query compound. Chemical similarity search heavily depends on the existence of reference compounds (usually inhibitors) for known targets and therefore on the existence of data for existing drugs or preclinical drug candidates. These data were collected from literature, patents and assembled into specialized medicinal chemistry or chemical proteomics databases [31]. The major ones are listed in Table 1. It includes for example the MDDR database (MDL Information Systems, developed by Dassault System), which contains 180,000 biologically relevant compounds; the CARLSBAD database which contains 439,985 unique chemical structures, mapped onto

In Silico Target Profiling

93

1,420,889 unique bioactivities itself including the WOMBAT database (120,400 annotated molecules) and ChemProt [32, 33]; the AurSCOPE (Aureus Science, Elsevier) and StARlite databases have about 300,000 annotated molecules each (see Table 1 for references). Biochemical affinities of screening compound collections are another source of information that can be used for in silico target profiling. Such data available from public domains’ databases such as ChEMBL or PubChem were reassembled into drug–target databases to support fast in silico target profiling. Online servers associated with these databases are listed in Table 2. A major representative is the protein pharmacology interaction network (PhIN) database which has integrated data from ChEMBL and covers about 1.4 M compounds associated with 9400 targets totalizing 12 M measures of biochemical activity [42]. Private companies invest a lot of effort in compiling quality data into their own environment. SAFAN Bioinformatics, an Italianbased company, has compiled more than 3100 targets from 15 therapeutic diseases with 600K associated compound totalizing more than 1.4M of interactions. Similar efforts were made within Chemotarget, a Spanish-based company which compiled more than 1M compounds. 2.1.2 Machine Learning Approach

Machine learning (ML) approaches use a supervised learning concept, which “learn” the type of drug–target interactions based on known data. They try to relate classes of compounds with their associated targets. The goal is to find chemical descriptors that best discriminate target classes using various statistical approaches such as naı¨ve Bayesian [61–63] or support vector machine (SVM) [64–66]. Naı¨ve Bayesian determines the chemical descriptors most frequently seen in the compound class of interest and not seen in the complement compound set. SVM relies on kernel-based regression or discriminant algorithms that optimize the minimal distance that separates different classes, in this case between potential target binders from non-binders for a given target. The SVM is therefore a discriminant strategy that tries to enlarge this difference. A recent review on the use of SVM learning techniques has recently been applied in cancer chemical genomics [67]. Efforts have been made to validate the SVM approach using a challenging case such as mapping kinase inhibitors to correct kinase family members. As expected, the quality of prediction is not constant and depends on target promiscuity [65]. Besides inherent difficulties, Cichonska et al. recently made a computation-experimental validation test with 5 kinase inhibitors, each one targeting 20 kinases. Results show that a fairly good overall agreement (Pearson correlation of 0.774 with p < 0.0001) could be obtained between the modelpredicted (pKi) and the experimentally measured (pIC50)

94

Jean-Yves Trosset and Christian Cave´

Table 2 List of databases and server to query drug–target interactions Name

Web site

PubMed ID

Refs.

PhIN

https://www.ebi.ac.uk/chembl/faq#faq67

26225242

[42]

OpenTarget

www.targetvalidation.org/about

27899665

[43]

PharmMapper

http://59.78.96.61/pharmmapper/help.php

20430828

[44]

ChemMapper

http://lilab.ecust.edu.cn/chemmapper

23712658

[45]

TarPred

http://www.dddc.ac.cn/tarpred

25686637

[46]

SuperPred

http://prediction.charite.de

24878925

[47]

Pharos

https://pharos.nih.gov/idg/about

27903890

[48]

DePick

http://mips.helmholtz-muenchen.de/Depick/help.jsp

27667560

[49]

PanDrugs

https://www.pandrugs.org/#!/help

29848362

[50]

DINIES 2.0

https://www.genome.jp/tools/dinies/help.html

24838565

[51]

IUPHAR-DB

http://www.iuphar-db.org/index.jsp

23087376

[52]

SuperTarget

http://insilico.charite.de/supertarget/

17942422

[53]

Reaxys analysis view

https://www.reaxys.com

24598793

[54]

PPB

http://gdbtools.unibe.ch:8080/PPB/

28270862

[7]

iPHACE

http://cgl.imim.es/iphace/

20156991

[55]

Chemotarget

https://www.chemotargets.com

29026211

[56]

PASS

http://195.178.207.233/PASS/index.html

26305108

[57]

PharmaExpert

http://www.pharmaexpert.ru/passonline/

26305108

[57]

Spider

http://modlabcadd.ethz.ch/software/spider/

24591595

[58]

SwissTargetPrediction

www.swisstargetprediction.ch

24792161

[59]

SuperPred

http://prediction.charite.de/index.php

24878925

[47]

TarPred

www.dddc.ac.cn/tarpred

25686637

[46]

ChemProt

http://potentia.cbs.dtu.dk/ChemProt

26876982

[33]

TargetHunter

www.cbligand.org/TargetHunter

23292636

[60]

bioactivities. This validation was made with 100 compound-kinase pairs [68]. Popular online servers based on SVM machine learning are listed in Table 3. 2.1.3 3D StructureBased Approaches

Finally, the third type of methods uses 3D structural information concerning the target and/or the ligand bound to its targetbinding site. Various in silico target profiling techniques are associated with these structure-based approaches.

In Silico Target Profiling

95

Table 3 SVM-based online target profiling servers Name

Web site

PubMed ID

Refs.

SVM-Prot 2016

http://bidd2.nus.edu.sg/cgi-bin/svmprot/svmprot.cgi.

27525735

[69]

SELF-BLM

https://github.com/GIST-CSBL/SELF-BLM

28192537

[70]

PTS

http://www.rcdd.org.cn/PTS

PMC5750839

[71]

iDTarget

http://idtarget.rcas.sinica.edu.tw

29297315

[72]

Table 4 Dock-panel and pharmacophore-based online target profiling servers Name

Web site

PubMed ID

Refs.

TarFisDock

http://www.dddc.ac.cn/tarfisdock/

16844997

[78]

PharmMapper

http://59.78.96.61/pharmmapper/

28472422

[44]

ChemMapper

http://lilab.ecust.edu.cn/chemmapper/

23712658

[45]

Panel docking approaches: This reverse docking strategy screens a query compound to many different target-binding sites for which 3D structure is known. This inverse approach raises new challenges compared to the classical docking. The first one is the need of a validated docking model for each target constructed from the atomic coordinate of the experimental 3D structure (see Note 3). The second one is the capacity to compare docking energies of a compound with all targets. This assumes that the systematic error in the docking energy of a compound needs to be the same for all targets. Finally, as docking is associated with a very high production of false-positive results, more than 60% of top rank conformers at the very best and usually around 80–90% are in fact not active experimentally when running docking experimental validation [73, 74]. In practice, some targets are easier than others, and some compounds are easier to dock than others: A very high energy score for a target is always an indication for this target to be a potential candidate for this query compound. A last technical challenge is to create a high-throughput algorithm for which compounds can be tested on few thousand targets in a reasonable amount of time. TarFisDock is a web server that successfully addressed this problem and offers a docking panel protocol in which query compounds can be tested on 698 known targets. TarFisDock (references in Table 4) used rigid models for the protein, although with multiple copies of the same target structure if the protein is known to be highly flexible. Water molecules are not considered in the model, and a standard CoulombVan der Waals potential with the Amber force field serves as the

96

Jean-Yves Trosset and Christian Cave´

energy-scoring function to compare binding affinity of ligands between targets. Hui-Fang et al. made in 2010 a benchmark study on a set of 1594 protein targets using TarFishDock, GOLD, FlexX, TarFisDock, and two of their in-house target search schemes TarSearch-X and TarSearch-M. These last two were found to outperform the other algorithms for at least some targets. A similar energy-based approach using genetic algorithm for conformation sampling of the ligand structure has been implemented into idTarget server. The idTarget uses a contraction-and-expansion strategy to make other targets benefit the accumulated experience of finding identical protein environments in various proteinbinding sites. This algorithm has been validated by docking HIV-1 protease inhibitors, HDAC2, and kinase inhibitors on a set of variants for each target class. IFPTarget is another docking-based target identification method which implements an interaction fingerprinting (IFP) scheme to analyze target-specific interactions and a comprehensive index to improve binding pose prediction and target ranking [75]. A major difficulty of this panel docking approach concerns the over-granularity of the 3D structure models. A small error in the atomic coordinates of a rigid receptor model may jeopardize the prediction of the binding affinity of a validated ligand for its target. Special care in the smoothness of the energy docking function was considered to cancel this effect [76]. Another approach relies on the use of a shape description for the binding using filling spheres of different radii as implemented into the original DOCK algorithm by Kuntz [77]. This algorithm was implemented in the INVDOCK reverse docking algorithm and was reviewed recently by Huang et al. [50, 72]. Pharmacophore-based approaches: this pharmacophore approach tries to simplify the 3D structural description of drug–target interactions using its most basic elements; only the pharmacophore features of the ligand are considered in the first case. These universal features correspond to chemical moieties that are often seen to contribute to drug–receptor interactions or to ADMET properties. The target prediction is realized by sampling the conformational space of the query compound to find the best 3D pharmacophore matching between the query compounds and the reference inhibitors of the tested targets. The better the superposition of the pharmacophore is, the better is the probability for this compound to bind to the corresponding target. This is based on structural similarity functions which mix geometrical and energy-based components. As the 3D structures of bound ligand can be found in PDB, this pharmacophore approach can be used to profile any compounds on the target that have PDB entries. PharmMapper and ChemMapper are open servers using this pharmacophore or geometric description of compounds (see Table 4).

In Silico Target Profiling

97

2.2 Compound Set Metric-Based Approaches

Pairwise compound comparisons based on ligand-based chemical similarity have the main drawback to be sensitive to the type of distance similarity function used in the target profiling protocol [14, 79]. To attenuate this effect, two approaches have been considered which are reviewed below.

2.2.1 Similarity Ensemble Approach

This approach first describes a set of inhibitors that is associated with a given target or a target family. Compound set comparisons give an estimation of the structural distance between two targets: instead of considering a single distance between a pair of inhibitors of target A and target B, a distance distribution profile is calculated between all inhibitors of target A with all inhibitors of target B. In order to have a reference profile for which distance can be calculated from, the intra-set distance profile is calculated as well (using set of inhibitors associated with target A or target B). The difference of profiles using intra-set (A with A or B with B) and interset (A with B) pairwise compound similarities gives an estimation of how target A differs from B, i.e., the promiscuity of these two targets or target families [14]. In silico target profiling schemes can be derived easily by measuring the distance profile of a query compound with a set of inhibitors for each target. A universal measure can be constructed from each profile giving an optimal score for the set of inhibitors that is most similar to the query compound (see Note 4). These concepts are underlying the similarity ensemble approach (SEA) developed by Keiser et al. [14]. They compare targets from the inter-class chemical similarity distribution and show that the chemical similarity distance between any two inhibitors of target A and B can be very low ( 0.7 or 0.8) for finding “similar” compounds in broad imprecise terms. The distance-set metric alleviates potential failures on target membership prediction by averaging out local fluctuations of the pairwise similarity distances. Keiser et al. used a set of 65,000 ligands that were grouped to each target family, as annotated in the MDDR database. SEA-metric technique was applied to validate the primary target of a known drug and to infer new cross-reactivity with unexpected targets, which were validated experimentally. This SEA-metric is a general approach that is well suited for compound prioritization after cell-based screening.

2.2.2 Structure Activity Relationship Approach

A second well-known approach to predict the affinity of a compound to target from its 2D or 3D structure is to use structure activity relationship (SAR). It maps the 2D or 3D structure in an activity profile previously constructed with a set of known active and inactive compounds for that target. Testing a compound onto a SAR model is equivalent to comparing this compound to a whole set of active and inactive reference inhibitors for a given target. In

Jean-Yves Trosset and Christian Cave´

98

Table 5 List of online servers based on set-similarity metric and QSAR Name

Web site

PubMed ID

Refs.

SEA

http://sea.bkslab.org

17287757

[14]

Random Forest QSAR

http://rfqsar.kaist.ac.kr

29297315

[81]

TargetNet

http://targetnet.scbdd.com

27167132

[80]

this sense, the SAR approach is similar conceptually to the SEA-metric presented above. TargetNet is a web service application to predict drug–target interactions by testing compounds on multiple SAR models [80]. A random forest algorithm recently proposed by Lee et al. [81] aims at improving prediction power based on such QSAR models. Target-fishing servers based on SAR approach are listed in Table 5.

3

Conclusion Systematic target-based studies have provided a tremendous amount of data that makes possible the development of techniques to map chemical space onto target space. The fact that this mapping is not one-to-one makes this task difficult. A lot of cross-reactivity indeed exist for most compounds. This multiple target interaction suggests systemic mechanisms of action that may have unexpected effects on phenotype. Targets can be triggered directly through drug–target physical interactions or indirectly through systemic consequence along the signaling cascade. In both cases, targets modulated by a drug can be identified by mining experimental data such as chemical affinity for direct interactions or gene expression profiles [60] to deduce drug indirect from up- or downregulation. All these approaches use basic similarity principles to predict the targets of a compound. The 2D or 3D structure of a query compound is compared with existing drugs, and the target is predicted based on how similar two compounds are. Difficulty in capturing all levels of compound similarity at once has led to the emergence of new approaches for target profiling based on compound set distance-metric or QSAR for which query compounds are compared to a whole class of compound inhibitors at once. 3D structure-based approaches can also be used to complement the prediction of the target profile of a compound or to validate some unexpected secondary targets. These in silico techniques prepare the road of future research to understand the systemic action of drugs or the systemic effect of a target or combination of targets on a given disease-related

In Silico Target Profiling

99

phenotype. Understanding how chemistry impacts this phenotype through winning combination of targets is one of the great challenges of today’s pharmacology.

4

Notes 1. Tanimoto Similarity Score: Given two molecules A and B , and their vector representations A and B where each component corresponds to a chemical descriptor, then the Tanimoto similarity score between these two molecules is given by S ðA ; B ;Þ ¼

S ðA; B Þ ¼ ðA \ B Þ=ðA \ B Þ

where (A \ B) denotes the size of the intersection, i.e., the number of 1 bit common to A and B, and A \ B denotes the size of the union, i.e., the number of 1 bit in A or B as described by Baldi and Nasr [30]. An important point concerning similarity distance functions is the value of threshold below which a weak similarity between two molecules (from visual inspection) is not captured anymore by the similarity score. The significance of these similarity functions has been discussed by these authors [30]. 2. The similarity distance function uses descriptors as the arguments of the function. This function will therefore assess compounds’ similarity with respect to a given point of view, the one defined by the selected chemical descriptors. Some structural descriptors are more important than others to describe the interactions of a specific drug with its target. Those descriptors should be given a bigger weight to the similarity function. If not, prediction associated with this target will fail. Such artifact associated with the selection of chemical descriptors or with the choice of the similarity function is a recurrent problem in the prediction of target profile. Machine learning approach tackles directly this problem by identifying the relevant descriptors for a specific target. 3. Such 3D docking models are usually validated using an experimental set of active and inactive compounds for that target (a SAR series). These validation studies will ensure that any new compound will have a trustful predicted potential inhibition (pI) for that target. 4. Inter- or intra-compound set comparisons can be made by a full pairwise comparison of compounds within or between the two sets. These approaches were extensively used for estimating the chemical diversity of compound libraries [82]. Recently, new chemical database descriptors have been proposed to capture the diversity of compounds into a single vector and to compare compound set more easily [83].

100

Jean-Yves Trosset and Christian Cave´

References 1. Ng JT, Dekker C, Reardon P et al (2016) Lessons from ten years of crystallization experiments at the SGC. Acta Crystallogr D Struct Biol 72:224–235 2. Mulinari S (2014) The specificity triad: notions of disease and therapeutic specificity in biomedical reasoning. Philos Ethics Humanit Med 9:14–25 3. Milletti F, Vulpetti A (2010) Predicting polypharmacology by binding site similarity: from kinases to the protein universe. J Chem Inf Model 50(8):1418–1431 4. Fliri AF, Loging WT, Thadeio PF et al (2005) Biological spectra analysis: linking biological activity profiles to molecular structure. Proc Natl Acad Sci 102(2):261–266 5. Card A, Caldwell C, Min H et al (2009) Highthroughput biochemical kinase selectivity assays: panel development and screening applications. J Biomol Screen 14(1):31–42 6. Miduturu CV, Deng X, Kwiatkowski N et al (2011) High-throughput kinase profiling: a more efficient approach toward the discovery of new kinase inhibitors. Chem Biol 18 (7):868–879 7. Awale M, Reymond JL (2017) The polypharmacology browser: a web-based multi-fingerprint target prediction tool using ChEMBL bioactivity data. J Cheminform 9(1):1–10 8. Li Q, Cheng T, Wang Y et al (2010) PubChem as a public resource for drug discovery. Drug Discov Today 15(23–24):1052–1057 9. Gaulton A, Hersey A, Nowotka ML et al (2017) The ChEMBL database in 2017. Nucleic Acids Res 45(D1):D945–D954 10. Liu X, Xu Y, Li S et al (2014) In silico target fishing: addressing a “Big Data” problem by ligand-based similarity rankings with data fusion. J Cheminform 6(1):1–14 11. Schmidt F, Matter H, Hessler G et al (2014) Predictive in silico off-target profiling in drug discovery. Future Med Chem 6(3):295–317 12. Gregori-puigjane´ E, Mestres JA (2008) Ligand-based approach to mining the chemogenomic space of drugs. Comb Chem High Throughput Screen 11:669–676 13. Mestres J, Martı´n-Couce L, Gregori-Puigjane´ E et al (2006) Ligand-based approach to in silico pharmacology: nuclear receptor profiling. J Chem Inf Model 46(6):2725–2736 14. Keiser MJ, Roth BL, Armbruster BN et al (2007) Relating protein pharmacology by ligand chemistry. Nat Biotechnol 25 (2):197–206

15. Zhao Z, Xie L, Xie L et al (2016) Delineation of polypharmacology across the human structural kinome using a functional site interaction fingerprint approach. J Med Chem 59 (9):5b02041 16. Brylinski M, Skolnick J (2010) Cross-reactivity virtual profiling of the human kinome by X-React KIN: a chemical systems biology approach. Mol Pharm 206(1):1–8 17. Hopkins A (2008) Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol 4(11):682–690 18. Tang J, Aittokallio T (2014) Network pharmacology strategies toward multi-target anticancer therapies: from computational models to experimental design principles. Curr Pharm Des 20(1):23–36 19. Anighoro A, Bajorath J, Rastelli G (2014) Polypharmacology: challenges and opportunities in drug discovery. J Med Chem 57 (19):7874–7887 20. Phatak SS, Zhang S (2013) A novel multimodal drug repurposing approach for identification of potent ACK1 inhibitors. Pac Symp Biocomput 2013:29–40 21. Sekhon BS (2013) Repositioning drugs and biologics: retargeting old/existing drugs for potential new therapeutic applications. J Pharm Educ Res 4:1–15 22. Dubus E, Ijjaali I, Barberan O et al (2009) Drug repositioning using in silico compound profiling. Future Med Chem 1(9):1723–1736 23. Corsello SM, Bittker JA, Liu Z et al (2017) The Drug Repurposing Hub: a next-generation drug library and information resource. Nat Med 23(4):405–408 24. Jenkins JL, Bender A, Davies JW (2006) In silico target fishing: predicting biological targets from chemical structure. Drug Discov Today Technol 3(4):413–421 25. Drakakis G, Wafford KA, Brewerton SC et al (2017) Polypharmacological in silico bioactivity profiling and experimental validation uncovers sedative-hypnotic effects of approved and experimental drugs in rat. ACS Chem Biol 12(6):1593–1602 26. Chaudhari R, Tan Z, Huang B et al (2017) Computational polypharmacology: a new paradigm for drug discovery. Expert Opin Drug Discov 12(3):279–291 27. Hu Y, Gupta-Ostermann D, Bajorath J (2014) Exploring compound promiscuity patterns and multi-target activity spaces. Comput Struct Biotechnol J 9:e201401003

In Silico Target Profiling 28. Rogers D, Hahn M (2010) Extendedconnectivity fingerprints. J Chem Inf Model 50:742–754 29. O’Boyle NM, Sayle RA (2016) Comparing structural fingerprints using a literature-based similarity benchmark. J Cheminform 8 (1):1–14 30. Baldi P, Nasr R (2010) When is chemical similarity significant? The statistical distribution of chemical similarity scores and its extreme values. J Chem Inf Model 50(7):1205–1222 31. Nicola G, Liu T, Gilson MK (2012) Public domain databases for medicinal chemistry. J Med Chem 55(16):6987–7002 32. Taboureau O, Nielsen SK, Audouze K et al (2011) ChemProt: a disease chemical biology database. Nucleic Acids Res 39:D367–D372 33. Kringelum J, Kjaerulff SK, Brunak S et al (2016) ChemProt-3.0: a global chemical biology diseases mapping. Database 2016:1–7 34. Sheridan RP, Shpungin J (2004) Calculating similarities between biological activities in the MDL drug data report database. J Chem Inf Comput Sci 44(2):727–740 35. Mathias SL, Hines-Kay J, Yang JJ et al (2013) The CARLSBAD database: a confederated database of chemical bioactivities. Database 2013:1–8 36. Krejsa CM, Horvath D, Rogalski SL et al (2003) Predicting ADME properties and side effects: the BioPrint approach. Curr Opin Drug Discov Devel 6(4):470–480 37. Petitet F, Barberan O, Dubus E et al (2006) Development of an ADME and drug-drug interactions knowledge database for the acceleration of drug discovery and development. Expert Opin Drug Discov 1(7):737–751 38. Booth A (2006) “Brimful of STARLITE”: toward standards for reporting literature searches. J Med Libr Assoc 94(4):421–429, e205 39. Southan C, Boppana K, Jagarlapudi SA et al (2011) Analysis of in vitro bioactivity data extracted from drug discovery literature and patents: ranking 1654 human protein targets by assayed compounds and molecular scaffolds. J Cheminform 3(1):14 40. Wishart DS, Knox C, Guo AC et al (2008) DrugBank: a knowledgebase for drugs, drug actions and drug targets. Nucleic Acids Res 36(Suppl. 1):901–906 41. Szklarczyk D, Santos A, Von Mering C et al (2016) STITCH 5: augmenting proteinchemical interaction networks with tissue and affinity data. Nucleic Acids Res 44(D1): D380–D384

101

42. Wang Z, Li J, Dang R, Liang L et al (2015) PhIN: a protein pharmacology interaction network database. CPT Pharmacometrics Syst Pharmacol 4(3):160–166 43. Koscielny G, An P, Carvalho-Silva D et al (2017) Open targets: a platform for therapeutic target identification and validation. Nucleic Acids Res 45(D1):D985–D994 44. Wang X, Shen Y, Wang S et al (2017) PharmMapper 2017 update: a web server for potential drug target identification with a comprehensive target pharmacophore database. Nucleic Acids Res 45(W1):W356–W360 45. Gong J, Cai C, Liu X et al (2013) ChemMapper: a versatile web server for exploring pharmacology and chemical structure association based on molecular 3D similarity method. Bioinformatics 29(14):1827–1829 46. Liu X, Gao Y, Peng J et al (2015) TarPred: a web application for predicting therapeutic and side effect targets of chemical compounds. Bioinformatics 31(12):2049–2051 47. Nickel J, Gohlke BO, Erehman J et al (2014) SuperPred: update on drug classification and target prediction. Nucleic Acids Res 42 (W1):26–31 48. Nguyen DT, Mathias S, Bologa C et al (2017) Pharos: collating protein information to shed light on the druggable genome. Nucleic Acids Res 45(D1):D995–D1002 49. Liu X, Baarsma HA, Thiam CH et al (2016) Systematic identification of pharmacological targets from small-molecule phenotypic screens. Cell Chem Biol 23(10):1302–1313 50. Yue Q-X, Cao Z-W, Guan S-H et al (2008) Proteomics characterization of the cytotoxicity mechanism of ganoderic acid D and computerautomated estimation of the possible drug target network. Mol Cell Proteomics 7 (5):949–961 51. Yamanishi Y, Kotera M, Moriya Y et al (2014) DINIES: drug-target interaction network inference engine based on supervised analysis. Nucleic Acids Res 42(W1):39–45 52. Sharman JL, Benson HE, Pawson AJ et al (2013) IUPHAR-DB: updated database content and new features. Nucleic Acids Res 41 (D1):1083–1088 53. Gu¨nther S, Kuhn M, Dunkel M et al (2008) SuperTarget and Matador: resources for exploring drug-target relationships. Nucleic Acids Res 36:919–922 54. Shi SH, Cai YP, Cai XJ et al (2014) A network pharmacology approach to understanding the mechanisms of action of traditional medicine:

102

Jean-Yves Trosset and Christian Cave´

Bushenhuoxue formula for treatment of chronic kidney disease. PLoS One 9(3):e89123 55. Garcia-Serna R, Ursu O, Oprea TI et al (2010) iPHACE: integrative navigation in pharmacological space. Bioinformatics 26(7):985–986 56. Garcia-Serna R, Mestres J (2010) Anticipating drug side effects by comparative pharmacology. Expert Opin Drug Metab Toxicol 6 (10):1253–1263 57. Pogodin PV, Lagunin AA, Filimonov DA et al (2015) PASS Targets: ligand-based multi-target computational system based on a public data and naı¨ve Bayes approach. SAR QSAR Environ Res 26(10):783–793 58. Reker D, Rodrigues T, Schneider P et al (2014) Identifying the macromolecular targets of de novo-designed chemical entities through selforganizing map consensus. Proc Natl Acad Sci 111(11):4067–4072 59. Gfeller D, Grosdidier A, Wirth M et al (2014) SwissTargetPrediction: a web server for target prediction of bioactive small molecules. Nucleic Acids Res 42(W1):32–38 60. Wang L, Ma C, Wipf P et al (2013) TargetHunter: an in silico target identification tool for predicting therapeutic potential of small organic molecules based on chemogenomic database. AAPS J 15(2):395–406 61. Nigsch F, Bender A, Jenkins JL et al (2008) Ligand-target prediction using winnow and naive Bayesian algorithms and the implications of overall performance statistics. J Chem Inf Model 48(12):2313–2325 62. Asafu-Adjei JK, Betensky RA (2015) A pairwise Naı¨ve Bayes approach to Bayesian classification. Intern J Pattern Recognit Artif Intell 29 (7):1550023 63. Zhang Z (2016) Naı¨ve Bayes classification in R. Ann Transl Med 4(12):241–246 64. Zhang J, Han B, Wei X et al (2012) A two-step target binding and selectivity support vector machines approach for virtual screening of dopamine receptor subtype-selective ligands. PLoS One 7(6):e39076 65. Balfer J, Heikamp K, Laufer S et al (2014) Modeling of compound profiling experiments using support vector machines. Chem Biol Drug Des 84(1):75–85 66. Heikamp K, Bajorath J (2014) Support vector machines for drug discovery. Expert Opin Drug Discov 9(1):93–104 67. Huang S, Nianguang CAI, Penzuti Pacheco P et al (2018) Applications of support vector machine (SVM) learning in cancer genomics. Cancer Genomics Proteomics 15(1):41–51 68. Cichonska A, Ravikumar B, Parri E et al (2017) Computational-experimental approach to

drug-target interaction mapping: a case study on kinase inhibitors. PLoS Comput Biol 13 (8):1–28 69. Li YH, Xu JY, Tao L et al (2016) SVM-prot 2016: a web-server for machine learning prediction of protein functional families from sequence irrespective of similarity. PLoS One 11(8):1–14 70. Keum J, Nam H (2017) SELF-BLM: prediction of drug-target interactions via self-training SVM. PLoS One 12(2):1–16 71. Ding P, Yan X, Liu Z et al (2018) PTS: a pharmaceutical target seeker. Database 2017 2017:bax095 72. Huang H, Zhang G, Zhou Y et al (2018) Reverse screening methods to search for the protein targets of chemopreventive compounds. Front Chem 6:138 73. Cotesta S, Giordanetto F, Trosset J-Y et al (2005) Virtual screening to enrich a compound collection with CDK2 inhibitors using docking, scoring, and composite scoring models. Proteins 60(4):629–643 74. Trosset J-Y, Dalvit C, Knapp S et al (2006) Inhibition of protein-protein interactions: the discovery of druglike beta-catenin inhibitors by combining virtual and biophysical screening. Proteins 64(1):60–67 75. Li GB, Yu ZJ, Liu S et al (2017) IFPTarget: a customized virtual target identification method based on protein-ligand interaction fingerprinting analyses. J Chem Inf Model 57 (7):1640–1651 76. Trosset J-Y, Scheraga HA (1998) Reaching the global minimum in docking simulations: a Monte Carlo energy minimization approach using Bezier splines. Proc Natl Acad Sci U S A 95(14):8011–8015 77. Allen WJ, Balius TE, Mukherjee S et al (2015) DOCK 6: impact of new features and current docking performance. J Comput Chem 36 (15):1132–1156 78. Li H, Gao Z, Kang L et al (2006) TarFisDock: a web server for identifying drug targets with docking approach. Nucleic Acids Res 34:219–224 79. Chen B, Mcconnell KJ, Wale N et al (2011) Comparing bioassay response and similarity ensemble approaches to probing protein pharmacology. Bioinformatics 27(21):3044–3049 80. Yao ZJ, Dong J, Che YJ et al (2016) TargetNet: a web service for predicting potential drug–target interaction profiling via multitarget SAR models. J Comput Aided Mol Des 30(5):413–424 81. Lee K, Lee M, Kim D (2017) Utilizing random Forest QSAR models with optimized

In Silico Target Profiling parameters for target identification and its application to target-fishing server. BMC Bioinformatics 18(Suppl 16):75–86 82. Gonza´lez-Medina M, Prieto-Martı´nez FD, Owen JR et al (2016) Consensus diversity

103

plots: a global diversity analysis of chemical libraries. J Cheminform 8(1):1–11 83. Ferna´ndez-De Gortari E, Garcı´a-Jacas CR, Martinez-Mayorga K et al (2017) Database fingerprint (DFP): an approach to represent molecular databases. J Cheminform 9(1):1–9

Chapter 7 Locus-Specific Knock-In of a Degradable Tag for Target Validation Studies Matthias Brand and Georg E. Winter Abstract Small molecule-induced targeted protein degradation is a powerful approach for drug target validation given its selectivity, high kinetic resolution, dose dependency, and reversibility. Out of the several methods that have been reported so far, the 12 kDa degradation tag (dTAG) system has the advantage of hijacking a degradation machinery that is ubiquitously expressed in all human tissues. Therefore it is independent of expressing additional, exogenous factors and additionally permits target validation in vivo. Here, we describe the protocol for generation and validation of clones harboring knock-in of a selectable dTAG cassette at the endogenous locus of proteins of interest using the near-haploid cell line KBM7. Key words dTAG, PROTACs, Targeted protein degradation, Genome engineering, Target validation

1

Introduction A fundamental step and major bottleneck in drug development is to link genes to disease initiation or maintenance and thus deliver faithfully validated targets for therapeutic interference. Classically, the function of a candidate is studied by modulating its expression in vitro and in vivo either by overexpression (gain of function) or depletion (loss of function). This can be achieved on three different levels: (1) at the DNA level by genetic inactivation or transcriptional activation of the gene, (2) at the mRNA level by knockdown via RNA interference, or (3) directly at the protein level by pharmacologic stabilization, destabilization, or inhibition of the target. However, approaches acting on RNA or DNA level typically suffer from the lack of specificity or reversibility and are further limited by the time delay between perturbation and appearance of the phenotype. This can potentially trigger a confounding compensatory feedback that complicates mechanistic deconvolution. In particular, this has been limiting in mechanistically addressing the function of essential genes.

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_7, © Springer Science+Business Media, LLC, part of Springer Nature 2019

105

106

Matthias Brand and Georg E. Winter

Targeted protein degradation overcomes many of these limitations, resulting in an exquisitely selective, reversible, and dosedependent system to disrupt protein activity at similar kinetic resolution as small molecule inhibitors. Of note, degradation of a protein can yield additional effects compared to domain-specific inhibition due to the disruption of putative scaffolding functions and can therefore provide unique insights into target-specific biology. Several tag-based systems for small molecule-induced targeted protein degradation have been reported. In most cases, treatment with the small molecule degrader recruits an E3 ubiquitin ligase to the tagged protein of interest, leading to rapid ubiquitination and degradation of the target. Commonly used technologies include the auxin-inducible degron (AID), based on the plant SCF E3 ubiquitin ligase Tir1 [1, 2]; the degradation of Halo-tagged proteins via Halo-PROTACS, which recruit the E3 ubiquitin ligase VHL to the fusion proteins [3]; the hydrophobic tagging [4]; as well as the antibody-based Trim-Away method [5]. In the case of the small molecule-assisted shutoff (SMASh) approach, proteins are fused to a ligand-dependent degron that removes itself by proteolytic self-cleavage in the absence of drug [6]. Conversely, the SHIELD system is based on an instable mutant FKBP12-based tag that is constitutively degraded in the absence of compound, while treatment with a synthetic ligand stabilizes the tagged protein, shielding it from degradation [7, 8]. More recently, a novel degradation tag (dTAG) based on the F36V space-creating FKBP12 mutant was reported [9, 10]. Cellular treatment with bumped FKBP12 phthalimide conjugates recruits the endogenous E3 ubiquitin ligase substrate receptor CRBN to the tagged protein, leading to target ubiquitination and ensuing degradation by the proteasome. Since the degradation machinery exploited by the dTAG system is ubiquitously expressed in all tissues and the required ligands are compatible with in vivo use in mouse models, it represents an ideal strategy for target validation in a preclinical setting. Here we describe a protocol for the CRISPR/Cas9-mediated knock-in of the dTAG degradation cassette at the endogenous locus of a gene of interest via microhomology-mediated end joining [11]. In particular, we outline design of the knock-in strategy, cloning of the necessary plasmids, as well as generation, selection, and validation of clones harboring degradable alleles due to correctly inserted dTAG cassettes.

Small Molecule Induced Protein Degradation

2 2.1

107

Materials Plasmids

1. pCRIS-PITChv2-dTAG-Puro (addgene #91796). 2. pCRIS-PITChv2-Puro-dTAG (addgene #91793). 3. pCRIS-PITChv2-dTAG-BSD (addgene #91795). 4. pCRIS-PITChv2-BSD-dTAG (addgene #91792). 5. pX330A-1x2 (addgene #58766). 6. pX330S-2-PITCh (addgene #63670).

2.2

Primers

pPITChv2_seq fwd

50 -GGG TCA TTA GTT CAT AGC CC-30

pPITChv2_seq rev

50 -TAT TAG GAA AGG ACA GTG GG-30

pX330S_guide_seq

50 -GCT GGC CTT TTG CTC ACA TG-30

NPITCh_dTAG_seq_fwd 50 -CCA GAT GAG TGT GGG TCA GAG AGC-30 NPITCh_BSD_seq_rev

50 -GCA CCA CGA GTT CTG CAC AAG GTC-30

NPITCh_PURO_seq_rev 50 -GTC CGG ATC GAC GGT GTG G-30 CPITCh_BSD_seq_fwd

50 -CCT GAC TTG TAT CGT CGC GAT CGG-30

CPITCh_PURO_seq_fwd 50 -CCG CAA CCT CCC CTT CTA CG-30 CPITCh_dTAG_seq_rev1 50 -TCG GAT CAC CTC CTG CTT GCC-30

2.3 Enzymes and Buffers

MluI-HF (NEB R3198S). BbsI-HF (NEB R0539S). BsaI-HF (NEB R3535S). Q5 High-Fidelity DNA Polymerase (NEB M0491S). Quick Ligation Kit (NEB, M2200S). T4 DNA ligase (NEB M0202S). T4 PNK (NEB M0201L). rSAP (NEB M0371S). 10 CutSmart Buffer (NEB B7204S). Q5 Reaction Buffer (NEB B9027S). Q5 High GC Enhancer (NEB B9027S). 2 NEBuilder HiFi DNA Assembly Master Mix (E2621S).

2.4 Chemical Ligands

Heterobifunctional dTAG-ligands (dTAG-7, dTAG-13) have been characterized and reported previously [9, 10].

108

3

Matthias Brand and Georg E. Winter

Methods The following protocol exemplifies the generation of N-terminally tagged CDK9 in KBM7 cells; for adaptation of the protocol to other proteins or cell lines, see Notes 1 and 2, respectively.

3.1 Design of PITCh Vectors

As depicted in Fig. 1, the PITCh system is based on co-transfection of two plasmids: 1. A repair vector, which contains the dTAG cassette flanked by 20 bp microhomology regions and synthetic PITCh-sgRNA recognition sites. 2. A cutting vector that comprises Cas9 and two sgRNAs targeting (A) the genomic locus of interest and (B) the synthetic PITCh sites on the repair vector. Upon transfection, the genome is cut at the locus of interest, and the excised repair cassette is integrated via microhomology-mediated end joining.

Fig. 1 Schematic depiction of the (a) PITCh repair and (b) cutting vectors, (c) the composition of the dTAG cassette, and (d) the targeted genomic locus after a successful knock-in

Small Molecule Induced Protein Degradation

109

Fig. 2 (a) Example of an annotated genomic locus. The coding sequence of the protein of interest is marked in purple, the selected sgRNA and the corresponding PAM are in red, the Cas9 cutting site is indicated as a black arrow, and 20 bp microhomology regions upstream and downstream of the cutting site are colored in green. (b) Example of an annotated PITCh repair plasmid after insertion of the gene-specific microhomology regions. PITCh-sgRNA target sites are marked in purple, microhomology regions in green, and inserted base pairs to correct the frameshift in red. Primers necessary for amplification and cloning of the dTAG cassette are depicted as purple arrows

(a) Download the genomic region of interest from the NCBI GenBank (https://www.ncbi.nlm.nih.gov/genbank/) “gene” database (see Note 3). (b) Choose a terminus to knock in the tag and the preferred selection marker (puromycin or blasticidin). (c) Design sgRNA to the region of interest using the CRISPR design tool by the Zhang lab (http://crispr.mit.edu). Choose a sgRNA that cuts as close as possible to the start codon (see Note 4) with a good score (usually >60, optimally >80). Design primers for cloning the sgRNA into the PITCh cutting vector by appending the following adapters to the sense and antisense sgRNA sequence: sgRNA sense: 50 -CACCGNNNNNNNNNNNNNNN NNNNN-30 sgRNA antisense: 50 -AAACNNNNNNNNNNNNNN NNNNNNC-30 (d) On the genomic sequence of the locus, annotate the sgRNA and the cleavage site (3 bp immediately upstream of the PAM for spCas9), as well as microhomology regions of 20 bp up- and downstream of the cleavage site (Fig. 2a). (e) On the vector map of the PITCh repair plasmid initially used to tag BRD4 [10] (Addgene #91792), replace the BRD4 microhomology regions with the gene-

110

Matthias Brand and Georg E. Winter

specific regions annotated in step 3. Potentially occurring frameshifts need to be corrected (Fig. 2b). (f) Design primers for the amplification of the cassette via PCR as follows (see Note 5): PITCh fwd primer: 33 bp upstream to 22 bp downstream of the 50 microhomology region PITCh rev primer: 16 bp upstream to 34 bp downstream of the 30 microhomology region (g) Design primers for the amplification of the locus of interest for clone validation. Primers should bind approximately 100–200 bp upstream and downstream of the cleavage site and be designed according to standard primer design rules (Tm ¼ 60–65  C, 40–60% GC content, GC-clamp at the 30 end). (h) We strongly recommend to simulate the cloning steps detailed in Subheadings 3.2 and 3.3 in silico using specialized software, such as SnapGene, Geneious, CLC workbench, or Benchling, to verify the correct design of the cloning strategy. (i) Order the primers designed in Subheading 3.1, steps 3, 6, and 7 (desalted, 0.025 μmol, 100 μM). 3.2 Cloning of PITCh Repair Vector

1. Linearize a pCRIS-PITChv2 vector of choice (see Note 6) by digestion with MluI-HF. Mix 10 μg of pCRIS-PITChv2 vector and 10 μL of 10 CutSmart Buffer, bring to 98 μL with nuclease-free ddH2O and add 2 μL of MluI-HF. Split the reaction mix in 2  50 μL and incubate for 1 h at 37  C. Run on a 0.8% agarose gel, and purify the larger band (ca. 4500 bp) via gel extraction. 2. Amplify the dTAG cassette via PCR to introduce target genespecific microhomology arms. Prepare the PCR reaction mix as outlined in Table 1 on ice. Run the PCR on a thermocycler with the protocol outlined in Table 2. Run the product on a 1% agarose gel, and purify the PCR product (ca. 1000 bp) by gel extraction. 3. Insert the amplified dTAG cassette into the backbone via Gibson assembly. Mix 30 ng of MluI-digested vector backbone (Subheading 3.2, step 1) and 20 ng of amplified dTAG cassette (Subheading 3.2, step 2), and bring to 3 μL total volume. Then add 3 μL 2 NEBuilder HiFi Master Mix, mix thoroughly by pipetting up and down, and incubate for 30 min at 50  C. 4. Add 20 μL of nuclease-free ddH2O to the product, and transform 2 μL of the dilution in 50 μL competent bacteria (e.g., XL1Blue). Plate all of the transformation on ampicillin

Small Molecule Induced Protein Degradation

111

Table 1 PCR reaction for PITCh repair vector 10 μL

5 NEB Q5 buffer

10 μL

5 NEB GC enhancer buffer

1 μL

dNTP mix (10 mM each)

2.5 μL

Primer mix (10 μM each of PITCh fwd and PITCh rev)

X μL

pCRIS-PITChv2-BSD-dTAG (50 ng)

0.5 μL

Q5 polymerase

To 50 μL

Nuclease-free ddH2O

Table 2 PCR thermocycler protocol for PITCh repair vector 98  C 

30 s

5

98 C 55  C 72  C

10 s 30 s 45 s

25

98  C 68  C 72  C 72  C Store at 4  C

10 s 30 s 45 s 2 min

LB-Agar plates. Expected colony count is approximately 50–200. 5. Inoculate 1–2 colonies for miniprep, and submit for Sanger sequencing to verify the insertion of the cassette. We suggest the following primers: pPITChv2_seq fwd (50 -GGG TCA TTA GTT CAT AGC CC-30 ), pPITChv2_seq rev (50 -TAT TAG GAA AGG ACA GTG GG-30 ). 6. After sequence verification, prepare plasmid DNA by midi- or maxiprep (see Note 7). 3.3 Cloning of PITCh Cutting Vector

1. Clone the empty pX330A-sgX-sgPITCh vector by mixing 75 ng of pX330A-1x2 and 150 ng pX330S-2-PITCh vector, 2 μL of 10 T4 ligase reaction buffer, 1 μL of BsaI-HF, and 1 μL Quick Ligase, and bring to 20 μL with nuclease-free ddH2O. Perform Golden Gate assembly in a thermocycler with the program detailed in Table 3. Remove the tube from the thermocycler, add 1 μL of BsaI-HF, and incubate for 30 min at 37  C followed by 5 min at 80  C to digest the intact pX330A-1x2 plasmid. Transform 2 μL of the product into

112

Matthias Brand and Georg E. Winter

Table 3 PCR thermocycler protocol for PITCh cutting vector 25

37  C 16  C

5 min 10 min

Store at 4  C

Table 4 Ligase reaction X μL (100 ng)

BbsI digested plasmid

2 μL

10 NEB T4 ligation buffer

2 μL

Diluted annealed oligos

1 μL

T4 ligase (NEB)

To 20 μL

Nuclease-free ddH2O

50 μL competent bacteria (e.g., XL1Blue). Plate all of the transformation on ampicillin LB-agar plates. Expected colony count: 100–200. 2. Inoculate two colonies for miniprep, and submit for Sanger sequencing to verify the correct insertion of the PITCh sgRNA. After sequence verification, prepare plasmid DNA by midiprep. 3. Linearize pX330A-sgX-sgPITCh by digestion with BbsI-HF. Mix 10 μg of pX330A-sgX-sgPITCh vector and 10 μL of 10 CutSmart buffer, bring to 98 μL with nuclease-free ddH2O, and add 2 μL of BbsI-HF. Split the reaction mix in 2  50 μL and incubate for 1 h at 37  C. 4. Dephosphorylate the linearized plasmid by adding 2 μL rSAP to each tube, and incubate for 30 min at 37  C. Run on a 0.8% agarose gel, and purify the linearized plasmid by gel extraction. 5. Anneal and phosphorylate the sgRNA oligos by mixing 1 μL each of sgRNA sense and antisense oligo, 1 μL of 10 T4 ligase buffer (see Note 8), 6.5 μL nuclease-free ddH2O, and 0.5 μL T4 PNK. In a thermocycler, incubate for 30 min at 37  C followed by 5 min at 95  C. Ramp down to 25  C at 5  C/ min. Dilute the annealed oligos 1:200 in nuclease-free ddH2O. 6. Insert the annealed oligos into the digested backbone by preparing the ligation mix as outlined in Table 4. In order to assess ligation efficiency, also prepare a negative ligation reaction adding H2O instead of the annealed oligos. Incubate the reaction at room temperature for 1 h, and then heat inactivate the T4 ligase at 65  C for 10 min.

Small Molecule Induced Protein Degradation

113

7. Transform 2 μL of the ligation reaction into 50 μL competent bacteria. Plate all of the transformation on ampicillin LB-agar plates. Expected colony count: 100–200 (negative control: 10–20). 8. Inoculate two colonies for miniprep, and submit for Sanger sequencing to verify the insertion of the sgRNA. We suggest the following primer: pX330S_guide_seq (50 -GCT GGC CTT TTG CTC ACA TG-30 ). 9. After sequence verification, prepare plasmid DNA by midi- or maxiprep (see Note 7). 3.4 Knock-In Generation and Clone Selection

The following protocol is optimized for nucleofection of KBM7 using the Lonza Amaxa Kit V. The transfection method needs to be adapted to the cell line of choice, aiming for >50–70% of transfection efficiency to be successful. 1. The day before transfection, split the cells to 250,000 cells/mL to ensure exponential growth of the cells at the moment of transfection. 2. Pre-warm IMDM without antibiotics to 37  C. Prepare a 12-well plate with 1 mL IMDM without antibiotics per well. Warm the plate at 37  C. 3. In a 1.5 mL tube, mix 6 μg each of the PITCh cutting and PITCh repair vector (see Note 9). Prepare a positive control with 6 μg of a similarly sized plasmid for the expression of GFP or a fluorescent protein of choice. 4. Prepare the Amaxa Nucleofector V solution by adding 18 μL of supplement to 82 μL of Nucleofector. 5. Spin down 2  106 cells at 90 g for 10 min at room temperature. 6. Remove supernatant, and then resuspend the cells in 100 μL Nucleofector solution. Quickly transfer the cell suspension to the tube containing the DNA mix, pipette up and down twice, and then transfer the mix to the cuvette supplied in the kit. Make sure that no air bubbles are covering the bottom of the cuvette. Close the cuvette with the cap (see Note 10). 7. Insert the cuvette into the machine, select the “X-001” program, and start it by pressing the “X” button. 8. Immediately add 700 μL of pre-warmed IMDM without antibiotics, and transfer to the previously prepared 12-well plate using the supplied Pasteur pipette. 9. After 72 h, assess transfection efficiency by FACS of the positive control. In our experience, a transfection efficiency of 50–70% is sufficient for the successful generation of knock-ins.

114

Matthias Brand and Georg E. Winter

10. In addition, transfer the cells to T25 flasks, and bring to 5 mL total volume. Start selection by adding Blasticidin at a final concentration of 10 μg/mL (see Note 11). Also perform antibiotic selection on the GFP-transfected cells to monitor appearance of spontaneous resistance. 11. Continuously monitor the growth of the resistant clones. After they reach a size of ca. 20–30 cells/clone (after approximately 2–3 weeks), carefully resuspend the cells, dilute them 1:1 with trypan blue, and count viable cells. 12. For the generation of clonal populations, seed two 384-well plates with 50 μL per well of a cell suspension at a concentration of 2 cells/mL in IMDM þ 10 μg/mL of Blasticidin (see Note 12). 13. Grow and expand the clones. 3.5 Clone Characterization

1. Screen for positive clones by HA-immunoblot (see Note 13 and Fig. 3a). 2. To test degradation of the tagged subunit, treat 2  106 cells with DMSO or 500 nM dTAG-ligand for 6 h, and lyse and immunoblot for HA. If available, it is advisable to additionally immunoblot with a protein-specific antibody (Fig. 3b). 3. Verify the correct integration of the dTAG cassette via Sanger sequencing. Isolate genomic DNA from >0.5  106 cells, and then prepare PCR mixes as outlined in Table 5. For each clone, prepare three separate PCR reactions as depicted in Fig. 3c to amplify the 50 and the 30 integration site, as well as the complete cassette (see Note 14). Run PCR in a thermocycler with the protocol in Table 6 (Fig. 3d). Purify the PCR products and send for Sanger sequencing (see Note 15). 4. Characterize the degradation kinetics of successfully validated clones via dose- and time-resolved Western Blot. As a starting point, we recommend treating with 50, 100, 200, 500, and 1000 nM of dTAG-ligand for 4 h (dose range) and with 500 nM of ligand for 30, 60, and 90 min and 2, 3, and 4 h (Fig. 3e). 5. Depending on the scientific question, further validation steps might be needed (see Note 16).

4

Notes 1. When designing a knock-in strategy, an important consideration to make is which terminus to introduce the tag in. We select the terminus to be tagged based on reported fusions of the protein of interest to similarly sized or bigger tags (e.g., GFP) or on available structural information.

Small Molecule Induced Protein Degradation

115

Fig. 3 Basic workflow of dTAG knock-in validation. (a) Screening of positive clones by HA- and CDK9immunoblot at steady state. (b) HA-immunoblot upon 6 h treatment with 500 nM dTAG-ligand to verify degradation of the tagged protein. (c) Schematic representation of the genomic locus upon knock-in and the PCR reactions (50 , 30 , ext) to set up for validation by Sanger sequencing. (d) PCR result for wt cells and 6 screened clones. (e) Time- and dose-resolved HA-immunoblots in dTAG-ligand treated cells

2. The near-haploid karyotype of KBM7 cells is of advantage to ensure that all of the protein of interest is expressed fused to the degradation tag. However, it is possible to generate knock-ins via microhomology-mediated end joining in other cell lines as well, as long as sufficient transfection efficiency can be reached. To increase the odds of bi-allelic dTAG knock-ins, we advise using both puromycin and blasticidin resistance cassettes. 3. The full locus can be found under “Genomic regions, transcripts, and products” by clicking on “Go to nucleotide: GenBank.” Download the sequence in the GenBank format (.gb)

116

Matthias Brand and Georg E. Winter

Table 5 PCR reaction for clone characterization 5 μL

5 NEB Q5 buffer

5 μL

5 NEB GC enhancer buffer

0.5 μL

dNTP mix (10 mM each)

1.25 μL

Primer mix (10 μM each of fwd and rev)

X μL

gDNA

0.3 μL

Q5 polymerase

To 25 μL

Nuclease-free ddH2O

Table 6 PCR thermocycler protocol for clone characterization 98  C 35



98 C 58  C 72  C 72  C

30 s 10 s 30 s 30 s 2 min

Store at 4  C

by clicking on “send to” and choosing “file” as destination. The GenBank annotation includes all splice variants of the transcript, thereby allowing also variant-specific tagging (Fig. 4). 4. Since the dTAG cassette itself contains a start codon, it is also possible to generate expressed knock-ins using sgRNAs that cut in the 30 UTR of the target gene. For C-terminal tagging, the sgRNA should cut within the coding region, close to the stop codon. 5. In most cases, it is possible to find a sgRNA that cuts within the first 10 bp of the coding region, so only a few amino acids are lost in the tagged protein. If this loss should be prevented, it is possible to add back the missing base pairs between the microhomology region and the start of the cassette. If >20 bp are to be inserted this way, consider adding them sequentially through multiple PCR reactions. Note that adding back the sequence may restore the sgRNA recognition sequence, which should be modified to prevent cutting of the cassette. A commonly used strategy is incorporation of a silent mutation into the PAM, but it is also possible to introduce multiple silent mutations into the sgRNA sequence instead.

Small Molecule Induced Protein Degradation

117

Fig. 4 Screenshots from the NCBI GenBank database

6. Since all of the dTAG cassette is excised in the MluI digestion, any of the pCRIS-PITChv2 vectors can be used as backbone, irrespective of the chosen cloning strategy. It is therefore advisable to prepare a larger stock of digested plasmid and store it at 20  C for future projects. 7. The higher purity maxipreps can slightly improve the efficiency of the knock-in. 8. Use the T4 ligation buffer, as the supplied T4 PNK buffer does not contain ATP. Alternatively, it is possible to supplement the T4 PNK buffer to 1 mM ATP. 9. While the given indication of 6 μg per plasmid works well in KBM7 in our hands, it is advisable to titrate the plasmid amounts used for transfection. Higher DNA amounts can greatly improve knock-in efficiency but at the same time lead to higher toxicity of the transfection. For the cell lines we tested, the optimum DNA amount ranges from 1 to 6 μg of each plasmid. 10. It is crucial to act as quickly as possible during the nucleofection to reduce the stress on the cells to a minimum. We

118

Matthias Brand and Georg E. Winter

therefore advise to prepare all necessary equipment before starting the nucleofection. 11. It is important to titrate the antibiotic concentration for selection. To do this, seed 500,000 cells per well in a 12-well plate and increasing concentrations of antibiotic (1–10 μg/mL). Assess the effect on cell viability by trypan blue staining after 3, 5, and 7 days since treatment start, and select the lowest concentration at which >99% of cells are killed. 12. Alternatively, it is also possible to seed the cells in 384-well plates concomitantly to starting the antibiotic selection, as this reduces overall processing time and can increase success rates for cellular systems that don’t efficiently grow out clones from single cells. To do so, take 1:2, 1:4, and 1:8 of the cellular pool, fill them up to 20 mL total volume using culture medium supplemented with antibiotic, and seed 50 μL of cell suspension per well. Monitor outgrowth of clones and expand when necessary. 13. Prior to validating individual clones, it is possible to assess successful tagging of the locus in a pooled format via PCR. To do so, set up PCR reactions as described in Subheading 3.5, step 3 on genomic DNA (gDNA) isolated from the pool of cells after antibiotic selection. Comparison of the band intensity of the short (wild-type) and long (integrated cassette) external PCR product can also give a rough estimate of the percentage of positive clones. Furthermore, this PCR setup can be employed instead of HA-immunoblots as an alternative clone screening method. 14. Note that the external PCR does not always yield the long PCR product upon cassette integration. The loss of the short wildtype band is however a good indication for successful knock-in, and the product of the 50 and 30 boundary PCR can be used for sequence validation. Alternatively, nested PCR strategies can be applied. 15. In our experience, it is generally possible to identify clones harboring seamless integration of the dTAG cassette at the locus of interest. However, the integration of the linearized repair cassette can sometimes take place via nonhomologous end joining rather than microhomology-mediated end joining, resulting in the integration of 27 bp. As this integration does not shift the reading frame, in most cases it does not affect protein function. In addition, it is a good indicator to distinguish clones that arose from separate integration events and therefore be able to test for clonal variation. 16. If the tagged protein is essential in the used cell line, it is possible to verify if degradation of the protein recapitulates the genetic dependency by proliferation assays.

Small Molecule Induced Protein Degradation

119

Furthermore, it might be of interest to verify subcellular localization of the protein by immunofluorescence, integration of the subunit in higher-order molecular assemblies by size exclusion chromatography, or correct localization to gene regulatory elements via ChIP-PCR. References 1. Nishimura K, Fukagawa T, Takisawa H et al (2009) An auxin-based degron system for the rapid depletion of proteins in nonplant cells. Nat Methods 6:917–922 2. Natsume T, Kiyomitsu T, Saga Y et al (2016) Rapid protein depletion in human cells by auxin-inducible degron tagging with short homology donors. Cell Rep 15:210–218 3. Buckley DL, Raina K, Darricarrere N et al (2015) HaloPROTACS: use of small molecule PROTACs to induce degradation of HaloTag fusion proteins. ACS Chem Biol 10:1831–1837 4. Neklesa TK, Tae HS, Schneekloth AR et al (2011) Small-molecule hydrophobic tagginginduced degradation of HaloTag fusion proteins. Nat Chem Biol 7:538–543 5. Clift D, McEwan WA, Labzin LI et al (2017) A method for the acute and rapid degradation of endogenous proteins. Cell 171:1692–1706. e18 6. Chung HK, Jacobs CL, Huo Y et al (2015) Tunable and reversible drug control of protein

production via a self-excising degron. Nat Chem Biol 11:713–720 7. Banaszynski LA, chun CL, Maynard-Smith LA et al (2006) A rapid, reversible, and tunable method to regulate protein function in living cells using synthetic small molecules. Cell 126:995–1004 8. Zhou Q, Derti A, Ruddy D et al (2015) A chemical genetics approach for the functional assessment of novel cancer genes. Cancer Res 75:1949–1958 9. Erb MA, Scott TG, Li BE et al (2017) Transcription control by the ENL YEATS domain in acute leukaemia. Nature 543:270–274 10. Nabet B, Roberts JM, Buckley DL et al (2018) The dTAG system for immediate and targetspecific protein degradation. Nat Chem Biol 14 (5):431–441 11. Sakuma T, Nakade S, Sakane Y et al (2016) MMEJ-assisted gene knock-in using TALENs and CRISPR-Cas9 with the PITCh systems. Nat Protoc 11:118–133

Chapter 8 Expression of Human VH Single Domains as Fc Fusions in Mammalian Cells Mahmoud Abdelatti, Peter Schofield, and Daniel Christ Abstract Single-domain antibodies represent an emerging class of antibody fragments with promising therapeutic and diagnostic potential. As a result, multiple strategies have been developed in order to improve their biophysical and/or biological properties. In particular, the fusion of single-domain antibodies to the Fc part of an IgG molecule has become a common protein engineering approach toward this aim. Here, we describe a detailed protocol for a streamlined laboratory-scale production of VH single-domain antibodies as Fc fusions in mammalian cells. Firstly, DNA sequence encoding VH domain of interest fused to an IgG Fc is synthesized as a double-stranded gene fragment. Secondly, the DNA fragment is directly assembled into a restriction enzyme-digested vector in an assembly reaction. Finally, vector carrying the VH-Fc-fusion construct is introduced into suspension-adapted mammalian cells for transient expression of the Fc chimeric fusion. One-week post-transfection, the expressed Fc-fusion protein is purified using protein A/G affinity chromatography. Using this protocol, we were able to clone, express, and purify milligrams of isolated anti-HER2 VH domain as a mouse IgG2c Fc fusion in less than 2 weeks. This protocol can be readily modified to express proteins of interest other than VH domains as Fc fusions. Key words Human VH single domains, Fc fusions, Mammalian expression, Protein biotinylation, Bio-layer interferometry

1

Introduction Single-domain antibodies are the smallest antigen-binding fragments that can be derived from a full-length IgG antibody [1]. They comprise either the variable heavy domain (VH) or the variable light domain (VL) of an IgG. Antigen-binding of single domains is maintained by the presence of highly diversified, flexible loops called complementarity determining regions or CDRs. Due to their small size, which is approximately one tenth the size of an IgG molecule, single domains are characterized by superior tissue penetration properties and the ability to bind cryptic epitopes such as those found in viruses and enzyme active site. Thus, antibody single domains have the potential of competing with full-length

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_8, © Springer Science+Business Media, LLC, part of Springer Nature 2019

121

122

Mahmoud Abdelatti et al.

antibodies as therapeutic and diagnostic agents [2]. On the other hand, human single domains have short plasma half-lives and are generally known to display poor biophysical properties such as low solubility and poor expression yield rendering them unsuitable for therapeutic and diagnostic applications. Among the strategies that have been developed to overcome such drawbacks is fusion of single domains to the Fc region of IgG antibody. The Fc domain is a homodimeric protein that is comprised of the hinge, the heavy chain’s second constant domain (CH2), and the heavy chain’s third constant domain (CH3) regions of the IgG molecule. Fusion to the Fc part of an IgG endows the fused protein partner with a prolonged serum half-life and an augmented effector immune function through the interaction with neonatal Fc receptors (FcRn) [3] and Fc γ receptors (FcγRs) [4] on immune cells, respectively. Moreover, fusion to the Fc domain can improve binding affinity and therapeutic potency of the fused protein partner by increasing their avidity to target antigen conferred by the homodimeric nature of the Fc domain. Finally, Fc fusion significantly improves the solubility and expression yield of the recombinant protein and greatly reduces purification steps by using protein A/G affinity chromatography [5, 6].

2

Materials

2.1 Design and Assembly of VH-Fc Fragment into pCEP4 Mammalian Expression Vector

1. Synthesized gBlock gene fragment encoding the designed VH-Fc DNA sequence (Integrated DNA Technologies). 2. pCEP4 mammalian Fisher Scientific).

expression

vector

(Thermo

3. KpnI-HF and BamHI-HF restriction enzymes (New England Biolabs). 4. CutSmart buffer (New England Biolabs). 5. Calf intestinal alkaline phosphatase (CIP) (New England Biolabs). 6. Shaking incubator at 37
C, 250 rpm. 7. QIAquick PCR purification kit (Qiagen) or similar. 8. Milli-Q water. 9. NanoDrop spectrophotometer (Thermo Fisher Scientific). 10. Thermocycler. 11. NEBuilder HiFi DNA Assembly kit (New England Biolabs). 12. Chemically competent NEB 5-alpha Escherichia coli (E. coli) cells or similar. 13. 100 mg/mL carbenicillin solution: Dissolve 1 g of carbenicillin in 10 mL of Milli-Q water. Filter solution through a syringe-

Single Domain Antibodies

123

driven 0.22 μm filter. Make 1 mL aliquots and store at 20
C for long-term use. 14. Luria-Bertani (LB) liquid medium: 1% (w/v) tryptone, 0.5% (w/v) yeast extract, 0.5% (w/v) NaCl. Dissolve 10 g of tryptone, 5 g of yeast extract, and 5 g of NaCl in 900 mL deionized water. Adjust pH to 7.5 with 1 M NaOH, and make up to 1 L with Milli-Q water. Autoclave. 15. LB/carbenicillin agar plates: 1.5% (w/v) agar, 1% (w/v) tryptone, 0.5% (w/v) yeast extract, and 0.5% (w/v) NaCl. Dissolve 15 g agar, 10 g of tryptone, 5 g of yeast extract, and 5 g of NaCl in 900 mL of Milli-Q water. Adjust pH to 7.5 with 1 M NaOH, and make up to 1 L with Milli-Q water. Autoclave. Cool to 50
C, and add 1 mL of 100 mg/mL carbenicillin solution for a final concentration of 100 μg/mL carbenicillin. 16. GoTaq DNA polymerase (Promega). 17. 5 GoTaq reaction buffer (Promega). 18. 20% (w/v) formamide in Milli-Q water. 19. 10 mM dNTP (New England Biolabs). 20. 1 kb DNA ladder (Invitrogen). 21. pCEP4 forward primer (50 -AGCAGAGCTCGTTTAGTGAA CCG-30 ) and EBV reverse primer (50 -GATGAGTTTGGA CAAACCAC -30 ). 22. Tris–EDTA (TE) buffer: 10 mM Tris–HCl, 1 mM EDTA, pH 8. 23. TAE buffer: 40 mM Tris–acetate and 1 mM EDTA. 24. 1% (w/v) agarose in TAE buffer: Dissolve 1 g agarose in 100 mL TAE buffer, and melt agarose in microwave. 25. DNA gel electrophoresis system. 26. Bench-top vortex. 27. Bench-top minicentrifuge. 28. Laboratory centrifuge. 2.2 Preparation of Plasmid DNA for Transfection into Expi293F Cells

1. pCEP4 containing sequence-verified VH-Fc insert. 2. pCEP4 forward primer (50 -AGCAGAGCTCGTTTAGTGAA CCG-30 ) and EBV reverse primer (50 -GATGAGTTTGGA CAAACCAC-30 ). 3. 500 mL sterile Erlenmeyer flasks. 4. 12 mL bacterial culture tubes. 5. Sterile pipette tips. 6. Petri dishes. 7. Shaking incubator at 37
C, 250 rpm.

124

Mahmoud Abdelatti et al.

8. Ultrapure glycerol (Invitrogen). 9. Disposable spreading loop. 10. Cryotube vials. 11. ZymoPURE plasmid Maxiprep kit or similar. 12. 0.22 μm syringe filters. 2.3 Plasmid Transfection into Expi293F Cells and Transient Expression of VH-Fc Domain

1. Class II biosafety cabinet. 2. Sterile, disposable pipettes. 3. Expi293F cells (Thermo Fisher Scientific). 4. Expi293 expression medium (Thermo Fisher Scientific). 5. Opti-MEMI reduced Fisher Scientific).

serum

medium

(Thermo

6. ExpiFectamine293 (Thermo Fisher Scientific). 7. ExpiFectamine293 Fisher Scientific).

Enhancer

I

and

2

(Thermo

8. 80% ethanol. 9. 125 mL disposable, sterile Erlenmeyer flask. 10. Humidified tissue culture incubator at 37
C, 8% CO2 with orbital shaker, 125 rpm. 11. Hemocytometer. 12. Trypan blue solution. 13. Light microscope. 14. Sterile plastic dropper. 2.4 Purification of VH-Fc Fragment from Culture Supernatant Using Protein A/G Affinity Chromatography

1. 150 mL, 0.22 μm Stericup filter units. 2. Protein A or Protein G agarose resin. 3. Sterile 1.5, 15, and 50 mL tubes. 4. Sterile 1 PBS, pH 7.4 (Thermo Fisher Scientific). 5. Rolling shaker. 6. Disposable gravity flow columns (Thermo Fisher Scientific). 7. 0.1 M glycine, 0.1 M NaCl, pH 2.7: Dissolve 7.5 g of glycine and 5.8 g of NaCl in 900 mL Milli-Q water. Adjust pH to 2.7 using 5 M HCl, and make up to 1 L using Milli-Q water. Filter through a 0.22 μm (1 L) Stericup filter unit. 8. 1 M Tris–HCl, pH 7.6: Dissolve 24.2 g of Tris base in 180 mL Milli-Q water. Adjust pH to 7.6 using 5 M HCl, and make up to 200 mL using Milli-Q water. Filter through a 0.22 μm (500 mL) Stericup filter unit. 9. SnakeSkin dialysis tubing (3.5K MWCO). 10. 4 L beaker.

Single Domain Antibodies

125

11. 15 mL spin concentrator (3K MWCO). 12. Centrifuge. 13. UV-Vis spectrophotometer. 2.5 Biotinylation of Proteins

1. Human HER2 (ECD). 2. Sterile, 1 PBS (Thermo Fisher Scientific). 3. EZ-Link NHS-PEG4-Biotin, No-Weigh microtubes (Thermo Fisher Scientific). 4. Dimethyl sulfoxide (DMSO). 5. Zeba Spin desalting column (7 kDa MWCO). 6. 0.5 mL, 0.22 μm disposable centrifuge tube filters. 7. UV-Vis spectrophotometer.

2.6 Measuring the Affinity of AntiHER2 VH-Fc Using Bio-layer Interferometry (BLI)

1. V-bottom, 96-well plate. 2. Streptavidin (SA) biosensors (Dip and Read, ForteBio, USA). 3. Filtered, 1% (w/v) BSA in PBS. 4. BLItz instrument and BLItz Pro software (ForteBio, USA). 5. 0.5 mL black microcentrifuge tubes (Argos Technologies).

3 3.1

Methods Method Overview

This method describes the cloning, expression, purification, and affinity characterization of a VH-Fc raised against the extracellular domain of human epidermal growth factor 2, HER2, as a model antigen target. In this example, the anti-HER2 VH domain is fused to mouse IgG2c Fc domain. This method can be easily switched to use other mouse or human Fc domains using the sequences provided (see Note 1). As illustrated in Fig. 1, the DNA sequence encoding VH fragment is inserted upstream of the Fc domain of mouse IgG2c and is synthesized as a gBlock gene fragment (a costeffective, sequence-verified, double-stranded DNA fragment). The designed gBlock is assembled into restriction enzyme-digested pCEP4, an episomal mammalian expression vector, using a 15-minute NEBuilder HiFi DNA assembly reaction. The assembled construct is then transformed into competent Escherichia coli cells for plasmid propagation. Suspension-adapted Expi293F cells are then transfected with pCEP4 carrying the VH-Fc fusion. Following expression, the anti-HER2 VH-Fc fusion is purified from culture supernatant using protein A/G affinity chromatography. Characterization of the purified VH-Fc fusion is performed using SDS-PAGE analysis.

126

Mahmoud Abdelatti et al.

Fig. 1 Generation of anti-target VH single-domain-Fc chimeric fusion. (a) Illustration of cloning the VH-Fc fusion-encoding DNA fragment into KpnI-BamHI cut pCEP4 mammalian expression vector (top panel) and the corresponding VH-Fc chimeric fusion created (bottom panel). (b) Detailed illustration of the anatomy of VH-Fc fusion-encoding DNA fragment

Single Domain Antibodies

3.2 Design and Synthesis of VH-Fc DNA Fragment as gBlock Gene Fragment

127

1. Design the VH-Fc gBlock gene fragment using DNA sequences for VH of interest, IgG hinge region, CH2, and CH3 domains according to Fig. 1b. Final gene fragment should be codon-optimized for mammalian cell expression. 2. Check the designed gBlock for complexities using the provided online analysis tool (http://sg.idtdna.com/site/order/ gblockentry). Complexities such as the presence of hairpin secondary structures or low GC contents of the gene terminal ends may prevent synthesis of the gBlock. If complexities exist, modify the sequence accordingly, or proceed with standard PCR-based gene construction methods. 3. Order the designed gBlock from Integrated DNA Technologies, IDT (see Note 2). 4. Once gBlock product is received, centrifuge the tube for 1 min at 10,000  g in order to collect the fragment at the bottom of the tube. 5. Reconstitute the contents using TE buffer to a final concentration of 10 ng/μL. 6. Heat at 50
C for 15 min in order to linearize any formed secondary structure during manufacturing process. Briefly mix and centrifuge.

3.3 Assembly of VHFc gBlock Fragment into KpnI-BamHI Digested pCEP4 Vector

1. In a single 50 μL reaction, digest 3 μg of pCEP4 mammalian expression vector with kpnI-HF and BamHI-HF (30 units each) in CutSmart buffer (see Note 3). Incubate for 15 min at 37
C. 2. Add 10 units of calf intestinal alkaline phosphatase (CIP) directly into the kpnI-HF/BamHI-HF enzyme digestion reaction in order to reduce background colonies harboring selfligated vector. Mix and incubate for further 30 min at 37
C. 3. Purify the double-digested dephosphorylated pCEP4 using Qiagen QIAquick PCR purification kit (see Note 4). 4. Elute plasmid DNA in 30–50 μL of Milli-Q water, and measure DNA concentration using NanoDrop spectrophotometer. Store RE-cut plasmid at 20
C until use. 5. Using the NEBuilder HiFi DNA assembly kit, set up assembly reaction as outlined in Table 1 (see Note 5). 6. In a thermocycler, incubate the above reaction at 50
C for 15 min. Following assembly, the reaction can be used directly to transform bacteria, or it can be stored at 20
C until use. 7. Add 2 μL of the assembled product to 20 μL of chemically competent NEB 5-alpha cells (alternatively, use any other bacterial strain that is suitable for plasmid propagation), and mix gently by flicking the tube 4–5 times (do not vortex).

128

Mahmoud Abdelatti et al.

Table 1 NEBuilder HiFi DNA assembly reaction Reagents

Amount

VH-Fc gBlock DNA

X μL corresponding to 0.15 pmol

RE-cut pCEP4 vector

Y μL corresponding to 0.03 pmol

NEBuilder master mix

10 μL

Milli-Q water

10  (X + Y ) μL

Total

20 μL

8. Incubate transformation reaction on ice for 30 min and then heat shock at 42
C for 30 s. Incubate on ice for a further 5 min. 9. Add 250 μL of LB media to transformed cells, and incubate at 37
C for 1 h while shaking. 10. Spread cells on LB/carbenicillin agar plates, and incubate overnight at 37
C. 11. The following day, pick 4–8 individual colonies using sterile 10 μL pipette tips and transfer into 200 μL PCR tubes containing 5 μL LB medium, and mix by pipetting up and down several times. 12. Use 2 μL of the resuspended colony culture to perform colony PCR. 13. Use the remaining 3 μL to inoculate 4 mL LB/carbenicillin medium, and incubate overnight at 37
C while shaking at 250 rpm. 14. Prepare colony PCR master mix as outlined below and then add 48 μL aliquots into each PCR tube containing 2 μL of resuspended colony cultures from step 12 (Table 2). 15. Mix well then briefly spin down in a benchtop minicentrifuge. 16. Run the PCR program as indicated in Table 3 using a thermocycler. 17. Following colony PCR, verify the amplification of the cloned VH-Fc fragment by DNA gel electrophoresis. A single band corresponding to the calculated fragment size (~1300 bp) should appear after running 5 μL of PCR products on 1% (w/v) agarose gel. 18. Purify plasmids from overnight cultures (step 13) that resulted in the correct amplicon size in colony PCR (step 17). 19. Using pCEP4 specific forward and reverse primers, prepare and then send plasmids for sequencing.

Single Domain Antibodies

129

Table 2 Colony PCR reaction setup Promega GoTaq reaction buffer (5)

10 μL

dNTP mix (10 mM stock)

1 μL

pCEP4 forward primer (10 μM stock)

2 μL

pCEP4 reverse primer (10 μM stock)

2 μL

Promega GoTaq DNA polymerase (2.5 units)

0.25 μL

Formamide (20% v/v stock)

5 μL

Double distilled H2O add to

50 μL

Table 3 Colony PCR amplification protocol PCR step

Temperature (
C)

Duration

Initial denaturation

96

6 min

Denaturation Annealing Extension

96 51 72

30 s 30 s 90 sa

Final extension

72

10 min

Number of cycles

36 cycles

a

Extension time was calculated based on fragment size and amplification capacity of the GoTaq polymerase (1kb/min)

3.4 Preparation of Transfection-Grade Plasmid DNA

1. Transform 20 μL of chemically competent NEB 5-alpha cells with 100 ng of pCEP4 carrying the correct VH-Fc sequence (Subheading 3.3, step 19) as described in Subheading 3.3, steps 7–10. 2. Pick a single colony using a sterile pipette tip, and inoculate 4 mL of LB/carbenicillin starter culture and then grow for 8 h at 37
C, 250 rpm. Use the rest of transformation plate to prepare glycerol stock. To do so, add 3 mL of LB medium containing 16% (v/v) final concentration of glycerol, and gently detach bacterial colonies from the plate using a disposable spreading loop. Transfer 1 mL aliquots of the resuspended culture into labeled sterile cryotube vials. Snap freeze in liquid nitrogen and store at 80
C. 3. Use 150 μL of starter culture prepared above to inoculate 150 mL of LB/carbenicillin culture. Incubate overnight at 37
C, 250 rpm. 4. Pellet overnight culture by centrifugation at 3200  g, 4
C, and purify plasmid using large-scale, transfection-grade plasmid preparation kit such as ZymoPURE Maxiprep kit (Zymo Research).

130

Mahmoud Abdelatti et al.

5. Elute purified plasmid using 400 μL TE buffer, and measure DNA concentration using a NanoDrop spectrophotometer. 6. In a tissue culture cabinet, sterilize plasmid preparation by filtration using 0.22 μm syringe filter. 7. Store the endotoxin-free, sterile plasmid preparation at 20
C until use. 3.5 Transfection of Expi293F Cells and Transient Expression of VH-Fc Domains

1. As the case with all tissue culture protocols, handling of Expi293F cells and reagents should be performed in line with aseptic technique guidelines in a certified class II biosafety cabinet. 2. Thaw one vial of Expi293F cells in a 37
C tissue culture water bath for 1–2 min, and then decontaminate the vial with 80% ethanol before opening. 3. In a tissue culture cabinet, transfer the vial contents into a 125 mL disposable, sterile Erlenmeyer flask containing 29 mL of pre-warmed Expi293 expression medium. Incubate the flask while rotating at 125 rpm in a humidified incubator (37
C, 8% CO2) with orbital shaker. 4. Monitor cell count and viability every 2 days using a hemocytometer and trypan blue exclusion until cell density reaches 3  106 cells/mL, and then passage cells by seeding a 125 mL Erlenmeyer flask containing pre-warmed Expi293 expression medium to a final cell density of 3  105 cells/mL. 5. Subculture Expi293F cells at least three times before transfection. 6. Three days before transfection, prepare 30 mL of Expi293 expression medium containing 5  105 cells/mL in a 125 mL Erlenmeyer flask as described in step 4. 7. On day of transfection, determine cell count and viability as described in step 4. Adjust cell density at 3.5  106 cells/mL (see Note 6). 8. Transfer 25.5 mL into a sterile 125 mL Erlenmeyer flask, and incubate as in step 3. 9. Add 30 μg of transfection-grade pCEP4 containing VH-Fc fragment (Subheading 3.4, step 7) to 1.5 mL Opti-MEMI reduced serum medium (reaction 1). 10. Similarly, mix 81 μL of ExpiFectamine293 with Opti-MEMI reduced serum medium to a final volume of 1.5 mL (reaction 2). Incubate for 5 min at room temperature. 11. Prepare ExpiFectamine293-plasmid complexes by adding the contents of reaction 1 to the contents of reaction 2. Mix gently and incubate for 30 min at room temperature.

Single Domain Antibodies

131

12. After the 30-min incubation, add the formed ExpiFectamine293-plasmid complexes to the flask containing 25.5 mL of Expi293F cells (step 8) using a sterile dropper with constant swirling of flask contents. 13. Incubate flask as in step 3 for 20 h. 14. After 20 h, add 150 μL of ExpiFectamine293 Enhancer I and 1.5 mL of Enhancer 2 to the transfected cells (at this stage the total volume is approximately 30 mL). 15. Incubate transfected cells for a further 5–6 days as in step 3 (see Note 7). 3.6 Purification of VH-Fc Fragment from Culture Supernatant Using Protein A/G Affinity Chromatography

1. Following the 5- to 6-day incubation period, pellet cells by centrifugation at 3200  g, 4
C for 25 min. 2. Collect and filter culture supernatant using 0.22 μm 150 mL Stericup filter unit. Take 50 μL aliquot and store at 4
C for SDS-PAGE analysis. 3. Place 1 mL of Protein A or Protein G agarose resin (see Note 8) in a sterile 15 mL tube. 4. Centrifuge at 300  g for 2 min, and then carefully discard supernatant to remove preservative. 5. Remove traces of preservative by resuspending resin in 10 mL of sterile PBS, and mix and centrifuge at 300  g for 2 min. Discard supernatant and repeat once. 6. Resuspend protein A/G resin in 2 mL PBS, and then add to the filtered culture supernatant (step 2). 7. Incubate mixture for 1 h at room temperature or 2–4 h at 4
C while gently rotating on a rolling shaker. 8. After incubation, pour the mixture into a disposable gravity flow column, and collect flow-through (store at 4
C). 9. Wash resin with four-column volume of sterile PBS. Collect and store wash at 4
C. 10. Elute bound VH-Fc by four sequential elution steps each with 3 mL of 0.1 M glycine, 0.1 M NaCl, pH 2.7. Collect each elution fraction in 15 mL sterile tube containing 1 mL of 1 M Tris–HCl, pH 7.6 in order to neutralize the acidic pH. Mix gently. 11. Carefully combine elution fractions into SnakeSkin dialysis tubing (3.5K MWCO), and dialyze against 4 L of PBS at 4
C. Perform two buffer changes at 2- and 16-hour intervals. 12. Next day, carefully empty the contents of the dialysis tube into a sterile 50 mL falcon tube.

132

Mahmoud Abdelatti et al.

13. Centrifuge at 3200  g, 4
C to clear any protein aggregates, and then transfer the cleared supernatant into a sterile 15 mL spin concentrator (3K MWCO). 14. Concentrate the dialyzed VH-Fc protein by centrifugation at 3200  g, 4
C (see Note 9) to a final volume of 0.5–1 mL. 15. Filter the concentrated protein using 0.22 μm disposable centrifuge tube filter by centrifugation at 16,000  g, 4
C for 1 min in a fixed-angle microcentrifuge. 16. Determine VH-Fc concentration by measuring its A280nm using a UV-Vis spectrophotometer and by calculating the extinction coefficient using an online tool such as ExPASy ProtParam tool. 17. Analyze purity of concentrated VH-Fc by SDS-PAGE along with other fractions from steps 2, 8, and 9. 18. Aliquot and freeze at 80
C for long-term storage. 3.7 Determination of Anti-HER2 VH-Fc Affinity

1. In this protocol, the affinity of an expressed anti-HER2 VH-Fc for human HER2 extracellular domain is measured in vitro using bio-layer interferometry (BLI) instrument. BLI is a technique that characterizes biomolecular interactions by measuring the interference pattern between an incident and reflected beam of light passing through an optical fiber sensor on which the interactions take place. For this, biotinylated HER2 is prepared and immobilized on a streptavidin sensor. Various concentrations of the purified anti-HER2 VH-Fc are then allowed to interact with the immobilized antigen. Finally, the affinity is determined based on the calculated association and dissociation rates. 2. Prepare a 0.25 mg/mL working stock of HER2 in a total of 400 μL PBS. If your protein is in any buffer other than PBS, then perform buffer exchange by dialysis or by using desalting columns (see Note 10). 3. Calculate the number of moles of the 80 kDa HER2 present in 400 μL of the 0.25 mg/mL stock solution. This should correspond to approximately 1.04 nmol of HER2. 4. Prepare 100 mM stock solution of biotinylating agent by reconstituting the contents of one 2 mg EZ-Link NHS-PEG4-Biotin, No-Weigh microtube in 34 μL of moisture-free DMSO. Mix well and then store at 20
C in 5 μL aliquots until use (see Note 11). 5. Determine the amount of biotin needed in the biotinylation reaction. For the biotinylation of HER2, we use five-fold molar excess of biotin in the reaction (see Note 12). Based on calculations in step 3 of this section, this should correspond to 10.4 nmol of biotin.

Single Domain Antibodies

133

6. Immediately before use, prepare 1 mM (1 nmol/μL) working stock of EZ-Link NHS-PEG4-Biotin (step 4) in PBS, and then add 10.4 μL directly into HER2 solution (step 2). Mix well and incubate for 30 min at room temperature. 7. Following the incubation time, add the contents of biotinylation reaction into 2 mL PBS-equilibrated Zeba Spin desalting column (7 kDa MWCO) placed in clean 15 mL falcon tube. When the sample is fully absorbed into the column resin, add 40 μL stacker of PBS for maximum protein recovery. 8. Centrifuge column at 1000  g for 2 min, and collect the desalted biotinylated HER2. 9. Measure protein concentration (Subheading 3.6, step 16). Aliquot and freeze at 20
C for long-term storage. 10. In a biosensor rack with V-bottom, 96-well plate, incubate streptavidin (SA) biosensors (Dip and Read, ForteBio, USA) in 200 μL/well of 1% BSA in PBS for at least 10 min at room temperature in order to block non-specific binding sites on SA biosensors. 11. Open a new advanced kinetics protocol (see Note 13) in the Blitz Pro software (ForteBio, USA). Select the “tube” assay format and set up as indicated in Table 4. 12. Prepare 250 μL of biotinylated HER2 in PBS at 10 μg/mL in a black 0.5 mL microcentrifuge tube as well as five other black tubes containing 250 μL of anti-HER2 VH-Fc at 3, 1, 0.5, 0.25, and 0.125 μM in PBS. In cases where the affinity is unknown, start the assay at higher concentrations (e.g., start at 12 μM). 13. For each concentration of anti-HER2 VH-Fc, place the BSA-blocked biosensor in a BLItz reader, and run the protocol in step 11. Use PBS for baseline 1 and 2, and for the dissociation step (use new PBS for each step), use the tube containing biotinylated HER2 for the loading step (see Note 14) and the one containing VH-Fc for the association step.

Table 4 Blitz advanced kinetics protocol Step

Duration (s)

1

Baseline 1

30

2

Loading

240

3

Baseline 2

60

4

Association

120

5

Dissociation

120

134

Mahmoud Abdelatti et al.

14. As controls for specific binding, use, in separate reactions, biotinylated HER1 or a dummy VH-Fc during loading and association steps, respectively. 15. Analyze data and determine affinity by global fit using BLItz Pro software.

4

Notes 1. The following are amino acid sequences of selected examples of antibody hinge and Fc regions. The hinge region is denoted by italicized residues. Mouse IgG2a hinge þ Fc region EPRGPTIKPCPPCKCPAPNLLGGPSVFIFPPKIKDVLMISLSPIVTC VVVDVSEDDPDVQISWFVNNVEVHTAQTQTHREDYNSTL RVVSALPIQHQDWMSGKEFKCKVNNKDLPAPIERTISKPKGSVRAP QVYVLPPPEEEMTKKQVTLTCMVTDFMPEDIYVEWTNNGKTELN YKNTEPVLDSDGSYFMYSKLRVEKKNWVERNSYSCSVVHEGLHNHH TTKSFSRTPGK Mouse IgG2c hinge þ Fc region EPRVPITQNPCPPLKECPPCAAPDLLGGPSVFIFPPKIKDVLMISLSPM VTCVVVDVSEDDPDVQISWFVNNVEVHTAQTQTHREDYNSTL RVVSALPIQHQDWMSGKEFKCKVNNRALPSPIEKTISKPRGPVRAP QVYVLPPPAEEMTKKEFSLTCMITGFLPAEIAVDWTSNGRTEQNYKN TATVLDSDGSYFMYSKLRVQKSTWERGSLFACSVVHEGLHNHLTTK TISRSLGK Human IgG1 hinge þ Fc region EPKSCDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTC VVVDVSHEDPEVKFNWYVDGVEVHNAKTKPREEQYNSTYRVVSVL TVLHQDWLNGKEYKCKVSNKALPAPIEKTISKAKGQPREPQVYTLPP SRDELTKNQVSLTCLVKGFYPSDIAVEWESNGQPENNYKTTPPVLD SDGSFFLYSKLTVDKSRWQQGNVFSCSVMHEALHNHYTQKSLSL SPGK

2. Other custom-made, gene synthesis services such as GeneArt Strings provided by Thermo Fisher Scientific can be used instead. 3. In contrast to the high-fidelity version, the activity of the standard KpnI and BamHI enzymes is suboptimal in CutSmart buffer. Therefore, for the standard version of KpnI and BamHI, perform sequential digestion using NEBuffer1.1 and NEBuffer3.1, respectively. 4. For RE-digested plasmids larger than 10 kb, use QIAEX II Gel Extraction System instead. 5. NEB recommends using 0.03–0.2 pmol of total DNA in the reaction. It also recommends using 50–100 ng of cut vector and two-fold molar excess of insert. From our experience, we

Single Domain Antibodies

135

recommend using up to five-fold molar excess of DNA insert. This corresponds to 111 ng of VH-Fc gBlock (1200 bp) and 185 ng cut pCEP4 vector (10.1 kb). 6. For optimum transfection efficiency, Expi293F cell viability should be equal to or higher than 90%. 7. Monitor protein expression by collecting 250 μL of culture supernatant 2, 4, and 6 days post-transfection, and then evaluate binding on protein A or protein G sensor using Bio-Layer Interferometry or protein A/G ELISA using proper controls. 8. Protein A is known to bind members of human VH3 immunoglobulin germline family [7] as well as to the Fc domain of IgG antibody [8]. In this case, the use of protein A in the purification of VH-Fc fragment could improve purification yield by increasing binding avidity to the VH-Fc fragment. 9. In order to minimize protein loss due to aggregation, perform 5-min centrifugation cycles, and thoroughly mix sample by pipetting up and down after each cycle. 10. For successful biotinylation of your protein, avoid buffers containing primary amines such as Tris- or glycine-containing buffers as they will compete with your protein for binding to biotin. 11. The NHS ester is rapidly hydrolyzed in water. Therefore, use moisture-free DMSO or dimethylformamide (DMF) in order to prepare stable, concentrated stock that can be stored at 20
C for several months. 12. During biotinylation of proteins, primary amines of solventexposed amino acids are randomly targeted such as those of lysine (K) residues. Labeling of primary amines of key amino acids might have a negative impact on protein function. Therefore, molar ratio of biotin to protein should be optimized so that it does not interfere with binding activity of the protein of interest. 13. Loading step of biotinylated antigen on SA sensor can be performed offline by incubating the BSA-blocked sensor in 200 μL/well biotinylated HER2 for 5 min and then in 200 μL/well PBS until it is ready to perform the run. However, in this case, basic kinetics protocol consisting of baseline, association, and dissociation should be selected instead. 14. In order to ensure optimal binding signal for subsequent VH-Fc association step, it is important to achieve a saturating binding level of biotinylated HER2 on the SA sensor as demonstrated by a clear loading plateau. Therefore, the concentration of biotinylated HER2 and duration of loading step should be modified accordingly until adequate loading is achieved.

136

Mahmoud Abdelatti et al.

References 1. Holt LJ, Herring C, Jespers LS et al (2003) Domain antibodies: proteins for therapy. Trends Biotechnol 21:484–490 2. Holliger P, Hudson PJ (2005) Engineered antibody fragments and the rise of single domains. Nat Biotechnol 23:1126–1136 3. Rojas R, Apodaca G (2002) Immunoglobulin transport across polarized epithelial cells. Nat Rev Mol Cell Biol 3:944–955 4. Nimmerjahn F, Ravetch JV (2008) Fcγ receptors as regulators of immune responses. Nat Rev Immunol 8:34–47

5. Carter PJ (2011) Introduction to current and future protein therapeutics: a protein engineering perspective. Exp Cell Res 317:1261–1269 6. Czajkowsky DM, Hu J, Shao Z et al (2012) Fc-fusion proteins: new developments and future perspectives. EMBO Mol Med 4:1015–1028 7. Roben PW, Salem AN, Silverman GJ (1995) VH3 family antibodies bind domain D of staphylococcal protein A. J Immunol 154:6437–6445 8. Boyle MDP, Reis KJ (1987) Bacterial fc receptors. Nat Biotechnol 5:697–703

Part III Three Dimensional Cell Culture Techniques Mimicking the Tumor Microenvironment In Vitro

Chapter 9 The Neurosphere Assay (NSA) Applied to Neural Stem Cells (NSCs) and Cancer Stem Cells (CSCs) Rossella Galli Abstract The discovery of neural stem cells (NSCs) in the mammalian brain has raised many expectations as these unique cells might recapitulate different neurological diseases, including brain tumors, both from a functional and molecular perspective. Proper in vitro culturing of NSCs has emerged as a critical methodological issue, given that it should preserve the in vivo features of NSCs, with particular emphasis on cell heterogeneity. At the same time, the methodology for NSC culturing should allow the production of large amounts of cells to be exploited not only for prospective clinical applications but also for drug screening. Direct in vitro selection of NSCs and, very recently, cancer stem cells (CSCs) by means of defined serumfree conditions represents the most reliable methodology to obtain long-term expanding SC lines. Here we describe the methods currently employed to enrich for NSCs/CSCs based on the neurosphere assay (NSA) and their adaptation to specific assays for testing the efficacy of neuroactive compounds. Key words Neurosphere assay, Neural stem cell cultures, Cancer stem cell

1

Introduction The identification of self-renewing and multipotent somatic neural stem cells (NSCs) in the mammalian brain raised many expectations as these unique cells faithfully model different neurological diseases, including brain tumors, both from a functional and molecular standpoint. Likewise, the recent availability of induced pluripotent stem (iPS) cells opens new scenarios for disease-related basic research, as the advent of cell reprogramming enables the generation of NSCs from individual patients, thus leading to the establishment of cellular model systems of neurological diseases. Independent from their origin (somatic neural tissues or iPS cells), NSCs can differentiate in multiple types of neural cells, making cell targeting very effective, in particular for those neurodegenerative diseases that affect specific neural lineages.

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_9, © Springer Science+Business Media, LLC, part of Springer Nature 2019

139

140

Rossella Galli

Proper in vitro culturing of NSCs has emerged as a critical methodological issue, given that it implies the maintenance of the in vivo characteristics of NSCs while warranting the generation of large amount of cells (“biomass”) to be exploited for drug screening [1]. Two main methodologies are currently in use for isolation, enrichment, and propagation of NSCs: (a) the neurosphere assay [2] and (b) the adhesion-mediated enrichment assay [3]. Of note, the two in vitro methods have been successfully applied to the isolation of putative cancer stem cells (CSCs) from brain tumors [4] as well as from breast, colon, and prostate cancer [5, 6] The implementation and adaptation of the different technologies developed for the study of NSC proliferation and differentiation might be instrumental for the establishment of screening platforms to identify small molecules with neurogenic and/or antitumor activity. Some controversies were recently raised concerning the best method to be exploited for NSC isolation and propagation. The NSA was first developed in the early 1990s by the laboratory headed by Reynolds and Weiss [2] and relies on neural cell culturing in a serum-free medium supplemented with two growth factors, namely, epidermal growth factor (EGF) and fibroblast growth factor 2 (FGF2) [7]. By this assay, three-dimensional floating cell clusters, known as neurospheres, are detected in culture after a variable period of time, which depends on cell staging (embryonic, fetal, or adult) and on the species of origin. Neurospheres are highly heterogeneous entities that reproduce the same cellular composition of a typical NSC niche: indeed, they comprise a minority of bona fide stem cells with the remaining bulk of cells containing bipotent progenitors and, although scarcely represented, fully mature neurons, astrocytes, and oligodendrocytes. Through many subculturing passages, ideally more than seven, different cell types, which are mostly committed progenitors that persist in the primary culture as a remnant of the tissue of origin, are negatively selected. The neurosphere cell composition reaches finally a steady-state equilibrium, in which about 2.5% of the cells within a sphere are true NSCs [8]. Neurosphere cultures display an average amplification rate of 50 neurospheres deriving from the dissociation of one single neurosphere. As such, neurosphere expansion in culture might occur quite rapidly, producing relevant numbers of cells to be used for downstream applications. As per definition, NSCs are multipotent and, thus, can differentiate into the three main lineages of the central nervous system (CNS), i.e., neurons, astrocytes, and oligodendrocytes. Several protocols are available to enrich for specific lineages and to simulate in vitro the commitment and differentiation steps taking place in vivo [9]. The second assay currently used for NSC isolation was recently developed by the laboratories directed by Cattaneo and Smith and is based on the culture of putative NSCs as a monolayer, in the

Culturing NSC/CSCs by the NeuroSphere Assay

141

presence of EGF and FGF2 on an adhesive substrate such as laminin [3]. As opposed to neurospheres, the resultant neural stem (NS) cell cultures constitute homogenous cultures that can be propagated extensively without accompanying spontaneous differentiation. In fact, adherent NS lines obtained by propagation in the presence of mitogens do not show expression of neuronal or glial markers and uniformly express the precursor markers RC2 and nestin. As neurospheres, NS cells differentiate into neurons and astrocytes upon growth factor withdrawal and addition of serum. Notably, the choice of one of the two methodologies depends on the biological question addressed and on the rationale underlying the study. If the experimental requirement is the availability of highly enriched, homogeneous NSC/CSC cell cultures, the appropriate methods should be considered: the adhesion-mediated enrichment assay [3]. On the contrary, if the aim of the study entails the preservation of cellular heterogeneity, the NSA may be the optimal choice. In general, cell cultures derived from neural tissues are considered powerful tools to be exploited for a better understanding of the cellular and molecular mechanisms underlying neural development and function. Moreover, they can also be employed for the screening of potential neuroactive compounds. The development of optimal culture methods for neural and cancer stem cells is a critical foundation for their exploitation in drug screening. To date, very few studies focused on chemical genetic screens for molecules inhibiting neurosphere proliferation (i.e., the closest readout for self-renewal that can be translated in high-throughput settings), revealing that screen-recovered small molecules, in particular neuromodulators, had also potent inhibitory effects on cultures enriched for brain cancer stem cells [10]. Others have recently described methods for a simple high-content live-image-based platform, reporting the effect of a panel of several kinase inhibitors on NS cultures [11]. Similarly, heterogeneous CSCs isolated from glioblastoma multiforme by the NSA have been exploited for testing the activity of EGFR inhibitors [12]. Here we describe the methods for the isolation of NSCs and CSCs by the NSA and their adaptation to small-scale drug screening applications.

2

Materials Prepare all solutions using ultrapure cell culture grade water. Filter sterilize and store solutions as indicated below.

2.1 Reagents for Culture Media

1. 10
DMEM-F12 (1:1), Invitrogen. 2. GlutaMax, Invitrogen.

142

Rossella Galli

3. 0.2% heparin sodium salt, grade 1A. 4. 30% D(þ)glucose. 5. 1 M HEPES buffer, Invitrogen 6. 7.5% NaHCO3 buffer, Invitrogen. 7. 3 mM sodium selenite. 8. 2 mM progesterone. 9. Putrescine dihydrochloride. 10. Apo-transferrin. 11. Insulin. 12. Penicillin/streptomycin. Growth Factors 13. EGF, human recombinant, Peprotech, 100-15. 14. FGF2, human recombinant, Peprotech, 100-18B. 2.2 Reagents and Materials for Tissue Digestion

1. EBSS. 2. Papain, Worthington DBA. 3. L-Cysteine hydrochloride monohydrate. 4. EDTA. 5. DNase. 6. ACK (ammonium-chloride-potassium) (thermoscientific).

lysing

buffer

7. 70-μm nylon cell strainer, Falcon. 2.3

Other Reagents

1. MatrigelTM (growth factor-reduced), Becton Dickinson. 2. Diff-Quick stain set, Dade Behring, Inc. 3. 8-μm Transwell polycarbonate inserts. 4. Trypan blue. 5. 4% paraformaldehyde. 6. MTT. 7. Vacuum filtration system combining a filter unit with a receiver flask for processing and storing volumes from 150 to 1000 mL.

2.4 Media and Solutions

For cell media preparation, ultrapure cell culture tissue grade water should be used. Carry out all procedures in a sterile hood. All media and stock solutions should be prepared only in sterile disposable tubes and/or bottles. 1. 30% glucose. Dissolve 30 g glucose in 100 mL water. Heat the solution at 60  C until complete solubilization. Filter sterilize and store at 4  C.

Culturing NSC/CSCs by the NeuroSphere Assay

143

2. 0.2% heparin. Dissolve 100 mg heparin in 50 mL water. Filter sterilize and store at 4  C. 3. 10
stock solution of DMEM/F12. Dissolve the content of a powder box in 1 L of ultrapure cell culture tissue grade water under gentle continuous stirring for several hours until the solution is clear. Filter sterilize and store at 4  C. 4. 3 mM sodium selenite. Add 1.93 mL of ultrapure cell culture tissue grade water to a 1-mg vial of sodium selenite. Mix, aliquot into sterile tubes, and store at 20  C. 5. 2 mM progesterone. Add 1.59 mL of 95% ethanol to a 1-mg vial of progesterone. Mix, aliquot into sterile tubes, and store at 20  C. 6. 10
stock solution of hormone mix (for 400 mL). Mix 40 mL of 10
DMEM/F12, 8 mL 30% glucose, 6 mL 7.5% NaHCO3, 2 mL 1 M HEPES, and 284 mL of water. Dissolve 400 mg of apo-transferrin into the solution. Dissolve 100 mg insulin in 4 mL sterile 0.1 N HCl, mix in 36 mL of water, and add to the hormone mix solution. Dissolve 38.6 mg of putrescine in 20 mL water and add to hormone mix solution. Add 40 μL of 2 mM progesterone and 40 μL of 3 mM sodium selenite. Mix well and filter sterilize. Aliquot in sterile tubes and store at 20  C. 7. Growth factor stock. Reconstitute EGF and FGF2 in order to have a 500 μg/mL and a 100 μg/mL stock solution, respectively. Aliquot into sterile Eppendorf vials and store at 20  C. 8. Proliferation medium (for 500 mL). Mix 369 mL water, 50 mL 10
DMEM/F12, 10 mL 30% glucose, 7.5 mL 7.5% NaHCO3, 2.5 mL 1 M HEPES, 5 mL 200 mM GlutaMaxTM, 5 mL PenStrep solution, 50 mL 10
hormone mix, 1 mL 0.2% heparin, 20 μL EGF- and 50 μL FGF2-stock (final concentration: 20 ng/mL EGF and 10 ng/mL FGF2). 9. MatrigelTM stock and ready-to-use solution. Thaw a 10-mL vial of MatrigelTM overnight at 4  C. Aliquot into sterile tubes (0.5 mL/aliquot) using refrigerated plastic pipettes and store at 20  C. Dissolve a 0.5 mL aliquot into 150 mL DMEM to obtain the ready-to-use solution. 10. 0.1% DNase stock. Add 10 mL water to 10 mg DNase. Mix well and filter sterilize. Aliquot in sterile tubes (0.5 mL/aliquot) and store at 20  C.

144

3

Rossella Galli

Methods The neurosphere culture paradigm selectively enriches for the stem cell component within neural tissues, including tumors, by taking advantage of the intrinsic property of bona fide stem cells, i.e., their capacity to self-renew long term. When seeded in the presence of the two mitogens EGF and FGF2 with no adhesion substrate, cells from freshly dissociated neural tissues give rise to spherical clusters, which float in suspension and are named “neurospheres.” Of note, not all the cells within a neurosphere are true stem cells. This concept is highly relevant for CSCs isolated from brain tumors. In fact, multiple subpopulations of tumor-initiating cells have been identified in CSC neurospheres, based on the expression of surface markers, such as EGFR and others [12]. Neurospheres are routinely subcultured by mechanical dissociation and by replating under the same in vitro conditions. At every subculturing passage, committed cells are selected negatively, whereas NSCs/CSCs do self-renew and generate spheres, allowing the establishment of long-term expanding NSC lines.

3.1 Dissociation of Digested Neural Tissue and Primary Culture

1. Cut the dissected tissue of interest into small pieces (see Note 1). If the neural tissue is obtained from rodents from embryonic day (E) 12 up to postnatal day (P) 7, or from human fetuses, proceed directly to step 6. If the neural tissue is obtained from rodents after P7 or from tumors, proceed with enzymatic digestion of the tissues. To this end, dissolve 400 units of papain, 5 mg of cysteine, and 5 mg of EDTA into 25 mL of EBSS. Vortex until the solution is clear, add 0.5 mL of a 0.1% DNAse stock solution, and filter sterilize. 2. Transfer the pieces of tissues into the 15-mL tubes containing the papain mix solution (at least 12 mL) (see Note 2). Transfer the tubes to a rocking device. Incubate at 37  C for 20–60 min, depending on tissue consistence. 3. At the end of the enzymatic incubation, pellet tissues by centrifugation at 100 g for 10 min. 4. Remove almost all the supernatant overlaying the pellets with a pipette. 5. Add 1 mL of fresh EBSS to the tissue. 6. Dissociate by triturating 20–50 times using a sterile 1000-μL pipette tip. Add 10 mL of fresh EBSS to the cell suspension. 7. To get rid of nondigested fragments of tumor tissues, filter the cell suspension through a 70-μm nylon cell strainer positioned inside a 50-mL Falcon tube. Then, transfer the 10 mL suspension in a new 15-mL tube. 8. Pellet the cells by centrifugation at 100 g for 10 min.

Culturing NSC/CSCs by the NeuroSphere Assay

145

9. To eliminate red blood cells that may be retained within the cell pellet, remove the supernatant and dissolve the pellet in 0.5–1 mL of ACK (ammonium-chloride-potassium) lysing buffer for 10 min at room temperature. 10. Discard supernatant and resuspend cells in 200 μL of culture medium. Using a 200-μL pipette with the volume set at 180 μL, gently dissociate the pellet for 20–40 times. Dilute a 5 μL aliquot from each sample in 15 μL of Trypan blue and count in a Burker or Neubauer chamber. Seed cells at a density of 8
103 viable cells/cm2 in culture medium containing EGF and FGF2. 11. Incubate at 37  C, 5% CO2 in a humidified incubator. 12. Cells should proliferate giving rise to primary sphere-shaped clusters, which will float in suspension (see Notes 3 and 4). 3.2 In Vitro Propagation of Neurospheres by Subculturing

1. Pellet the neurosphere suspension by centrifugation at 100 g for 10 min (see Note 5). 2. Remove the supernatant leaving approximately a volume of 200 μL. Using a sterilized p200 pipetman set at a volume of 180–200 μL, gently triturate pellet (see Note 6). 3. Count viable cells by Trypan blue exclusion and seed cells at the appropriate density in culture medium in untreated tissue flasks.

3.3 Assays for Measuring the Effect of Pharmacologically Active Compounds on NSC/CSC Properties 3.3.1 Short-Term SelfRenewal Assay (Clonogenic Assay)

The clonogenic assay is normally exploited to analyze the effect of compounds on the symmetry of division in NSCs/CSCs. In this assay, the number of secondary spheres generated from the dissociation of a single sphere reflects the frequency of NSC/CSCs present in the original primary clone. This analysis also returns an estimate of the relative frequencies between symmetric proliferative (two SCs generated at each cycle) and symmetric differentiative (two differentiated and/or dead cells generated after cell division) divisions. 1. Seed the cells derived from the dissociation of clonal single neurospheres into uncoated 96-well plates in proliferation medium with or without the compound under testing. Count the cells in each well 2 h after plating, in order to obtain a “baseline” value to be used for establishing the efficiency of secondary sphere generation. 2. Count the neurospheres generated in each well, 3–10 days after plating, depending on the cell type. Normalize the number of secondary neurospheres to the number of cells originally seeded in each well.

146

Rossella Galli

3.3.2 Long-Term SelfRenewal Assay (Growth Curve/Population Assay)

To analyze the effect of compounds on long-term self-renewal in cultures of NSCs/CSCs, the appropriate method is the growth curve/population analysis. Given that NSCs/CSCs under mitogenic stimuli expand in number over serial subculturing, this assay monitors long-term self-renewal that relies on the generation of at least one cell identical to the original clonogenic stem cell at every cell division. Thus, stable stem cell lines can be established that share the same functional properties of the founder cell. In such a system, drug-dependent alterations in NSC/CSC properties will result in the modification of the overall growth rate of the NSC/CSC line under analysis. 1. For growth curve/population analyses, seed the cells at 8
103 cells/cm2 in proliferation medium with or without the compound under testing. 2. Collect the spheres when they reach the proper size. Spin them at 100 g for 10 min. Remove the supernatant leaving behind approximately 200 μL. Using a sterilized p200 pipetman set at a volume of 180–200 μL, gently triturate pellet. 3. Count viable cells by Trypan blue exclusion and seed cells at 8
103 cells/cm2 in proliferation medium with or without the compound under testing in new flasks. 4. Repeat steps 1 through 3 for at least 10 subculturing passages. At any passage, calculate the relative amplification rate by dividing the amount of counted cells by the number of cells originally seeded. Multiply the amplification rate for the number of cells originally seeded, and plot this number in correspondence of every subculturing passage. In this way, a growth curve will be generated, whose slope provides information as to the frequency of putative NSCs/CSCs in each experimental condition [8].

3.3.3 MTT Assay to Determine NSC Proliferation/Survival

The MTT assay allows to measure quantitatively the viability and, as such, the proliferation rate of cells upon drug administration. MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) is processed by metabolically active cells resulting in the generation of a purple formazan. Release of insoluble formazan by viable cells is promoted by using a solubilization agent as dimethyl sulfoxide (DMSO). The absorbance of the resulting solution is measured at 550-nm wavelength by a spectrophotometer, and it is directly related to cell viability/proliferation. 1. Coat the wells of 96-well plates with MatrigelTM overnight. 2. Generate a titration curve, by plating cells at different densities and by performing MTT assays at the same time point, e.g., 2 h after plating. The absorbance value for each cell density will

Culturing NSC/CSCs by the NeuroSphere Assay

147

allow estimating the number of cells survived/proliferated at the end of your experiment. 3. Seed the cells into the required number of 96-well plates at 2
104 cells/well in 180 μL of proliferation medium (see Note 7). Allow the cells to adhere to the plates for 2 h. At this point, (a) add MTT to “baseline” plates in order to obtain the absorbance value at baseline (see detailed description below); (b) add the vehicle/compound of interest to the remaining plates to be used for the assay. 4. After the appropriate period of time (normally ranging from 12 to 96 h), add 20 μL of a 5 mg/mL MTT solution in PBS (see Note 8) to each well, which corresponds to a 1:10 dilution. Incubate the cells with MTT for 1 h at 37  C. Next, gently aspirate the medium and add 50 μL of DMSO to each well. DMSO lyses the cells, allowing the insoluble formazan to be released and solubilized (see Note 9). Gently shake the plate on a horizontal shaker for 15 min. 5. Read the plate on a microplate spectrophotometer reader at 550 nm. Plot the data by exploiting the titration curve generated at the beginning of the experiment. 3.3.4 TranswellMediated MatrigelTM to Assess NSC Migration/ Invasion Potential

Cell migration and invasion are cardinal properties of NSCs and, most importantly, CSCs. Assessing the effect of drugs on these functional parameters is therefore of paramount preclinical relevance. A MatrigelTM invasion assay, performed in Transwell chambers, represents the most advantageous and efficient method. 1. Coat overnight with MatrigelTM the required number of 6-well Transwell inserts containing a polycarbonate 8.0-μm pore membrane. 2. Include an additional insert for any condition, treatment, or time point, which will serve as internal control, from which non-migrated cells won’t be removed, in order to assess the effect of the drug on the overall cell viability (see below for details). 3. Seed cells at a density of 2
104 cells/insert onto the layer of Matrigel in proliferation medium with or without the compound under testing for the appropriate time. 4. At the end of the experiment, mechanically remove the cells on the upper side of the inserts (which did not migrate through the membrane) by using a cell scraper or a cotton swab. Do not remove cells from the internal controls.Fix the inserts with 4% paraformaldehyde and then stain with hematoxylin and eosin (H&E) for 2 min or Diff-Quick staining solutions, according to the manufacturer’s instructions.

148

Rossella Galli

5. Cut the membrane out of the insert by using a scalpel and transfer the membrane on a glass microscope slide. Glue a coverslip to the microscope slide by using a solvent-free mounting agent or nail polish. 6. The extent of cell migration/invasion can be monitored by standard light microscopy and/or by densitometric analysis (e.g., using ImageJ software). Normalize the extent of migration in each sample to the one of the corresponding internal controls.

4

Notes 1. Dissection should be carried out as quickly as possible from tissue collection. If you estimate that more than 2 h is required, keep the tissues at 4  C. 2. The total amount of material to be digested in each tube should not cover more than half of the volume retained in the conic bottom of the tube. If the material is in excess, subdivide it in multiple 15-mL tubes. 3. Rodent NSCs from embryonic stages form neurospheres in about 2–3 days, whereas NSCs from postnatal stages or from human fetuses require 5–7 days to give rise to neurosphere formation. 4. Neurospheres are considered viable and healthy if single cell boundaries can be clearly observed within the cell cluster, if the overall morphology of the neurosphere is regular and round, and if neurosphere appears homogeneously bright. If the center of the neurospheres acquires a dark appearance, the cells within the core are dying, and, as such, the neurospheres need to be subcultured immediately. 5. Subculture neurospheres when their diameter is in the range of 200 and 500 μm. If neurospheres are too small when subcultured, the yield of cells will be very low; if neurospheres are too large, the number of dead cells inside the spheres (that give rise to the “dark core”) will be high, mechanical dissociation will be difficult, and viability of the culture will be very low. 6. For achieving maximal disaggregation of neurospheres, slightly tilt the pipetman and press tip against the bottom of the tube to generate a fair amount of resistance. Gently triturate pellet for 50–70 times for embryonic rodent cells and 60–80 times for adult rodent cells and cancer stem cells. 7. For MTT assays, include a “baseline” control plate for each condition/treatment.

Culturing NSC/CSCs by the NeuroSphere Assay

149

8. MTT solution should be protected from light, when stored at 4  C for up to 2 months. 9. Modified tetrazolium salts such as XTT have recently become available. Viable cells convert these modified compounds to a water-soluble formazan. As a result, the solubilization step can be eliminated from the procedure. References 1. Winquist RJ, Furey BF, Boucher DM (2010) Cancer stem cells as the relevant biomass for drug discovery. Curr Opin Pharmacol 10 (4):385–390. https://doi.org/10.1016/j. coph.2010.06.008 2. Reynolds BA, Weiss S (1992) Generation of neurons and astrocytes from isolated cells of the adult mammalian central nervous system [comment]. Science 255(5052):1707–1710 3. Conti L, Pollard SM, Gorba T, Reitanoi E, Toselli M, Biella G, Sun Y, Sanzone S, Ying QL, Cattaneo E, Smith A (2005) Nicheindependent symmetrical self-renewal of a mammalian tissue stem cell. PLoS Biol 3(9): e283. https://doi.org/10.1371/journal.pbio. 0030283 4. Galli R, Binda E, Orfanelli U, Cipelletti B, Gritti A, De Vitis S, Fiocco R, Foroni C, Dimeco F, Vescovi A (2004) Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma. Cancer Res 64(19):7011–7021 5. Ponti D, Costa A, Zaffaroni N, Pratesi G, Petrangolini G, Coradini D, Pilotti S, Pierotti MA, Daidone MG (2005) Isolation and in vitro propagation of tumorigenic breast cancer cells with stem/progenitor cell properties. Cancer Res 65(13):5506–5511. https://doi. org/10.1158/0008-5472.CAN-05-0626 6. Wang S (2009) Anchorage-independent growth of prostate cancer stem cells. Methods Mol Biol 568:151–160. https://doi.org/10. 1007/978-1-59745-280-9_9 7. Gritti A, Cova L, Parati EA, Galli R, Vescovi AL (1995) Basic fibroblast growth factor supports

the proliferation of epidermal growth factorgenerated neuronal precursor cells of the adult mouse CNS. Neurosci Lett 185 (3):151–154 8. Reynolds BA, Rietze RL (2005) Neural stem cells and neurospheres--re-evaluating the relationship. Nat Methods 2(5):333–336 9. Vescovi AL, Reynolds BA, Fraser DD, Weiss S (1993) bFGF regulates the proliferative fate of unipotent (neuronal) and bipotent (neuronal/ astroglial) EGF-generated CNS progenitor cells. Neuron 11(5):951–966 10. Diamandis P, Wildenhain J, Clarke ID, Sacher AG, Graham J, Bellows DS, Ling EK, Ward RJ, Jamieson LG, Tyers M, Dirks PB (2007) Chemical genetics reveals a complex functional ground state of neural stem cells. Nat Chem Biol 3(5):268–273. https://doi.org/10. 1038/nchembio873 11. Danovi D, Falk A, Humphreys P, Vickers R, Tinsley J, Smith AG, Pollard SM (2010) Imaging-based chemical screens using normal and glioma-derived neural stem cells. Biochem Soc Trans 38(4):1067–1071. https://doi.org/ 10.1042/BST0381067 12. Mazzoleni S, Politi LS, Pala M, Cominelli M, Franzin A, Sergi Sergi L, Falini A, De Palma M, Bulfone A, Poliani PL, Galli R (2010) Epidermal growth factor receptor expression identifies functionally and molecularly distinct tumor-initiating cells in human glioblastoma multiforme and is required for gliomagenesis. Cancer Res 70(19):7500–7513. https://doi. org/10.1158/0008-5472.CAN-10-2353

Chapter 10 3D-3 Tumor Models in Drug Discovery for Analysis of Immune Cell Infiltration Annika Osswald, Viola Hedrich, and Wolfgang Sommergruber Abstract The cross talk between tumor cells and other cells present in the tumor microenvironment such as stromal and immune cells highly influences the behavior and progression of disease. Understanding the underlying mechanisms of interaction is a prerequisite to develop new treatment strategies and to prevent or at least reduce therapy failure in the future. Specific reactivation of the patient’s immune system is one of the major goals today. However, standard two-dimensional (2D) cell culture techniques lack the necessary complexity to address related questions. Novel three-dimensional (3D) in vitro models—embedded in a matrix or encapsulated in alginate—recapitulate the in vivo situation much better. Cross talk between different cell types can be studied starting from co-cultures. As cancer immune modulation is becoming a major research topic, 3D in vitro models represent an important tool to address immune regulatory/modulatory questions for T, NK, and other cells of the immune system. The 3D systems consisting of tumor cells, fibroblasts, and immune cells (3D-3) already proved as a reliable tool for us. For instance, we made use of those models to study the molecular mechanisms of the cross talk of non-small cell lung cancer (NSCLC) and fibroblasts, to unveil macrophage plasticity in the tumor microenvironment and to mirror drug responses in vivo. Generation of those 3D models and how to use them to study immune cell infiltration and activation will be described in the present book chapter. Key words Spheroid, Floater, Matrigel/collagen, Bioreactor, 3D culture, Immune cell infiltration, NK cell, T cell, Alginate capsules, Non-small cell lung cancer (NSCLC)

Abbreviations 3D CAFs CFSE CMAC CTLA4 DMEM DO ECM EDTA EMT

Three-dimensional Cancer-associated fibroblasts Carboxyfluorescein succinimidyl ester 7-Amino-4-chloromethylcoumarin Cytotoxic T-lymphocyte-associated protein 4 (CD152) Dulbecco’s Modified Eagle’s Medium Dissolved oxygen Extracellular matrix Ethylenediaminetetraacetic acid Epithelial-mesenchymal transition

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_10, © Springer Science+Business Media, LLC, part of Springer Nature 2019

151

152

Annika Osswald et al.

FBS FCS GFP HDFs HEPES LAG3 MAPK mTOR NSCLC PBS PD1 PNKs RFP RPMI medium RT TAM TGF TIM3 T regs ZO-1

1

Fetal bovine serum Fetal calf serum Green fluorescence protein Human dermal fibroblasts 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid Lymphocyte activation gene 3 (CD223) Mitogen-activated protein kinase Mammalian target of rapamycin Non-small cell lung cancer Phosphate-buffered saline Programmed cell death protein 1 Primary NK cells Red fluorescence protein Roswell Park Memorial Institute medium Room temperature Tumor-associated macrophage Tumor growth factor T-cell membrane protein 3 (CD366) Regulatory T cells Zonula occludens, tight junction protein

Introduction Carcinomas are highly complex structures composed of genetically altered tumor cells, together with both normal and cancerassociated fibroblasts (CAFs), endothelial cells, pericytes, and inflammatory cells embedded in a matrix of extracellular proteins (ECM). It is exactly this molecular heterogeneity, which influences the way tumor cells migrate, proliferate, and survive during tumor progression [1–3]. As a consequence, this tumor stroma cross talk influences the tumor’s “immunologic behavior” finally leading to evasion of immune destruction by tumor cells that disrupt anticancer response pathways [4–6]. The term “cancer immunoediting” describes the fact that the immune system plays a kind of Dr. Jekyll and Mr. Hyde dual role in the complex interactions between tumors and the host immune system [4, 6]. The cross talk between tumor and stromal cells leads to a “reshaping” of the immunogenic tumor environment [7] and activates a tumor-promoting inflammation in an NF-κB-dependent manner [8, 9]. This finally leads to the negative regulation of T cells via checkpoint receptors such as PD-1, CTLA4, LAG3, and TIM3 to name but a few [10]. Immune checkpoint inhibition is an effective treatment strategy in multiple tumor types, including non-small cell lung cancer (NSCLC) [11] and melanoma [12]. However, only subpopulations of patients show responses to the anti-immune checkpoint treatment in particular in advanced stages [12–15]. The molecular mechanisms

3D-3 Tumor Models to Study Immune Cell Infiltration

153

leading to resistance to immune checkpoint blockade are – if at all – only partially understood [16]. Additionally, immune-related adverse events of immune checkpoint inhibitors have been reported [17, 18]. In order to learn more about the transcriptional repertoire of inactivated T and NK cells in the tumor microenvironment and to elucidate the underlying molecular mechanisms and cross talks contributing to resistance toward immune checkpoint inhibitors, novel 3D models are urgently needed. Cells grown in threedimensional (3D) scaffolds or as 3D aggregates (multicellular organotypic spheroids) embedded in an extracellular matrix much better recapitulate the in vivo situation of tissues (selected examples: [19–24]) or tumors (selected examples: [25–30]). Those cultures are now being also applied in drug discovery, drug delivery, drug resistance, and drug repositioning [29, 31–35]. Since the cancer immune modulation as a therapeutic concept becomes more and more important, there is also an increasing need in respective complex 3D organoid-like models to address questions such as the transcriptional requirements for infiltrating T and NK cells, the role of macrophages in immune modulation in tumors, the induction of immunogenic cell death, and drug screening for activation of Tregs to name but a few. Currently, only a limited number of models are available to be used for such studies [36–39]. Preferentially in vivo models would be most appropriate to address immune modulatory questions. However, due to the fact that mice and men differ quite substantially in the regulation of the immune system [40], complex 3D organoid-like human models are the models of choice. In general, organoid technology is the ultimate system to model various human pathologies in a tissue flask. Patient-derived organoids also allow predicting drug response [41–46]. To capture tumor complexity in vitro, we recently have compared 2D and 3D tumor models (including slice-based tumor models) for drug discovery clearly demonstrating that end points differ according to cell type, stromal co-culture, and culture format [47–49]. Furthermore, our findings that 3D co-cultures of human NSCLC cell lines with a high EMT-score and fibroblasts induce an NF-κB-dependent expression of cytokines and chemokines in fibroblasts led us to establish complex 3D triple cultures (3D-3). This culture platform is based on alginate microencapsulation and stirred culture systems containing tumor cell spheroids of non-small cell lung carcinoma (NSCLC) cells, fibroblasts, and monocytes. The cross talk between ECM and tumor, stromal, and immune cells in microencapsulated 3D-3-culture promotes the activation of monocytes into tumor-associated macrophages (TAM) (M2-like phenotype), thus recreating the TAM phenotype in vitro. Both the recruitment of human monocytes into tumor tissue and their polarization into an M2-like phenotype were

154

Annika Osswald et al.

possible without directed differentiation by the addition of exogenous cytokines. In addition, our 3D-3 model served as a tool to unveil macrophage plasticity in the tumor microenvironment [50]. In the present book chapter, we describe not only three different methods to cultivate 3D tumors but also another 3D triple model for studying the interaction between the tumor microenvironment and immune cells such as T and NK cells. This model system can then be used for interrogation of the molecular requirements of immune cell infiltration into tumor spheroids. In this complex model, 3D tumors are grown and embedded in Matrigel/collagen together with fibroblasts and immune cells mimicking the active infiltration of human immune cells to a 3D tumor. These organotypic 3D-3 models consisting of tumor spheroids, stromal fibroblasts, and NK or T cells represent valuable models to appropriately model infiltration and establish novel therapeutic concepts in the area of immune-oncology.

2

Materials

2.1 General Consumables

1. 1
phosphate-buffered saline (PBS). 2. 1
Roswell Park Memorial Institute—1640 (RPMI-1640) or Dulbecco’s Modified Eagle’s Medium (DMEM), phenol red-free if required (see Note 1), supplemented with 10% fetal calf serum (FCS), 2 mM Glutamax, 1
penicillin/streptomycin (P/S) (optional). 3. 0.25% trypsin-EDTA. 4. Plate reader (e.g., PE-Wallach, EnVision 2100).

2.2

3D Floater

1. 96-well ultra-low attachment plates with clear round bottom (Corning® Costar® Sigma-Aldrich)

2.3

3D Embedded

1. Black 96 wells μClear microtiter plate. 2. 2
(8 mg/mL) Matrigel solution (see Note 2, keep on ice!): Add 100
L-glutamine (2 mM final), 2.5 M D-glucose (12 mM final), FBS (2–10% final), and cell culture medium (without FBS; fill up to volume required to generate 8 mg/mL Matrigel) to a 10 mL vial of Matrigel stock (>9 mg/mL, phenol red and LDEV-free). 3. 2
(3 mg/mL) collagen solution (see Note 2, keep on ice!): Add 10
RPMI with L-glutamine (final 1
), 1 M HEPES (final 30 mM for 2
), 7.5% sodium bicarbonate (final 0.25% for 2
), 1 M NaOH (to neutralize collagen solution, approximately 18 μL/mL per 3 mg/mL), and FBS (2–10% final) to collagen I solution (rat tail). 4. Prepare 1:1 Matrigel/collagen mix by mixing equal volumes of 2
Matrigel and 2
collagen solutions.

3D-3 Tumor Models to Study Immune Cell Infiltration

2.4

3D Bioreactor

155

1. 1.1% (w/v) of ultrapure Ca2+ MVG alginate (UP MVG NovaMatrix, Pronova biomedical) in 0.9% (w/v) NaCl in Millipore H2O. 2. Cross-linking solution: 100 mM CaCl2 in 10 mM HEPES (pH 7.4). 3. Electrostatic bead generator (Nisco VarV1) and syringe pump. 4. DASbox® Mini Bioreactor System with BioBLU® Single-Use Vessels (Eppendorf). 5. 100 mM EDTA for alginate bead dissolution.

2.5 3D Tumor: Immune Cell Infiltration Assay

1. Black 96-well ultra-low attachment plates with clear round bottom for Envision measurement (Corning® Costar® SigmaAldrich). 2. 96-well ultra-low attachment plates with clear round bottom for imaging analysis (Corning® Costar® Sigma-Aldrich). 3. EasySep™ Human NK or T-Cell Isolation Kit (Stemcell). 4. CellTrace™ CFSE Cell Proliferation Kit (Thermo Fisher). 5. CellTracker™ Blue CMAC (Thermo Fisher).

3 3.1

Methods 3D Floater

1. Prepare tumor cells (and fibroblasts) for a desired concentration in the respective medium (500–10,000 cells per well, depending on cell line and assay conditions, see Note 3). 2. Seed cells in 200 μL per well in 96-well ultra-low attachment plates using the inner wells of the plate for analysis. Fill the outer wells with medium or 1
PBS to avoid the edge effect. 3. Incubate plates at 37  C and 5% CO2 in a humidified incubator.

3.2

3D Embedded

This section describes 3D-embedded cultures in either Matrigel or collagen only or in a 1:1 mix of Matrigel/collagen. 1. Prepare 8 mg/mL Matrigel and/or 3 mg/mL collagen solution (see Note 4). 2. Pre-coat 96-well plates with 30 μL of relevant matrix keeping final concentration of collagen at 1.5 mg/mL and Matrigel at 4 mg/mL. Examples: Collagen only: [3 mg Col I] + [RPMI 1
+ 2% FCS, Glutamax, P/S] ¼ 1:1. Matrigel/collagen mixture: [3 mg Col I] + [8 mg Matrigel]. Matrigel only: [8 mg Matrigel] + [RPMI 1
+ 2% FCS, Glutamax, P/S] ¼ 1:1.

156

Annika Osswald et al.

3. Directly centrifuge the plates at 974 g for 5 min at room temperature. 4. Place the plates in the incubator at 37  C for at least 30 min to allow gelling of the matrix. 5. Prepare cells of desired concentration in a falcon tube at room temperature (e.g., use 10,000 tumor and/or 1000 stromal cells per well). 6. Take the pre-coated 96-well plate out of the incubator on a plate heater in the laminar flow hood. 7. Resuspend cells in respective matrix, and add 60 μL per well to the plate. 8. Incubate the plate at 37  C for at least 1.5 h to allow complete gelling of the matrix. 9. Add 90 μL of warm 1
RPMI medium to the wells, and also fill empty wells to avoid the edge effect. 10. Keep the plates in the humidified incubator at 37  C and 5% CO2. For treatments, add 20 μL of diluted compounds to the well, and analyze fluorescence intensity using a plate reader (see Note 5). 3.3

3D Bioreactor

1. Collect tumor cells, fibroblasts, and/or immune cells in a falcon tube and centrifuge at 290 g, 5 min at room temperature. Use 6
106 cells per cell type per 1 mL of alginate. For co-cultures, use cell ratio 1:1. 2. Remove the supernatant, and resuspend the pellet in the calculated amount of alginate (see Note 6), and transfer the alginate suspension into a syringe (5 mL with some sterile air in it, see Note 7). 3. For cross-linking of alginate beads, place a gelling bath containing 100 mM CaCl2 solution with 10 mM HEPES (pH 7.4) with a magnetic stir bar underneath the metal needle of the electrostatic bead generator (VarV1). 4. Connect the syringe containing the alginate suspension with the metal needle by using silicon tubing. 5. Place the syringe in the syringe pump, and select an air flow rate of 10 mL/h. 6. For alginate bead generation, select the right voltage (5.4 V), and start electrostatic bead generator (VarV1). 7. Collect the capsules from the gelling bath with a pipet, and transfer them into a falcon tube. 8. Wash microcapsules two times with 0.9% NaCl solution (breaks off for centrifugation, 290 g, 5 min, see Note 8).

3D-3 Tumor Models to Study Immune Cell Infiltration

157

Fig. 1 3D tumors in alginate capsules cultivated in a bioreactor. (a) Red fluorescent (RFP) tumor cells (triplenegative breast cancer cell line BT-20) were cultivated in mono- or co-culture with green fluorescent (GFP) fibroblasts in alginate capsules in the bioreactor at 60 rpm with 30% DO and 5% CO2 at 37  C for up to 23 days. 3D tumor growth was monitored via fluorescence microscopy. From day 9 on, a glandular-like structure is built in the co-culture with fibroblasts. This might represent the basic structure of a mammary gland. (b) Mono-cultures of red fluorescent BT-20 tumor cells were treated with an mTOR inhibitor such as Dactolisib (mTORi, 5 μM). Cryosections were performed at Day 18 after treatment and stained for several markers in green; proliferation (Ki-67), cytoplasmic tight junction (ZO-1), and actin filament (Phalloidin). Treatment of 3D tumors in alginate capsules with mTORi resulted in destructed 3D tumor structures

9. Resuspend the microcapsules in sterile-filtered medium, and add them to the BioBLU Vessels (in total 150 mL of medium/ vessel for 1 mL alginate capsules). 10. Keep cultures at 60 rpm with 30% DO (dissolved oxygen) and 5% CO2, with 50% medium exchange every 2–3 days (see Note 9). 11. Take samples at different time points, and analyze alginate beads via fluorescence microscopy, measurement of fluorescence intensity, or staining of sections (see Fig. 1). For RNA or flow cytometry analysis, dissolve beads by incubation with 100 mM EDTA for 5 min. 3.4 3D Tumor: Immune Cell Infiltration Assay

Tumor cell lines were either transduced with a lentiviral construct (e.g., mKate2) or transfected with a plasmid (e.g., dTomato) to generate stably expressing fluorescent tumor cells. These modified tumor cell lines were grown as floating 3D spheroids and

158

Annika Osswald et al.

embedded in a co-culture with fibroblasts and immune cells in Matrigel/collagen. Cytokines or different compounds are added to activate or inactivate immune cells. The killing effect on tumor cells was analyzed via measuring the fluorescence intensity of tumor spheroids. For infiltration analysis, Z-stacks were made by confocal microscopy. Day 1 1. Seed tumor cells in 96-well floater plates (see Subheading 3.1): 5,000-10,000 cells/well, 3-4 days incubation time (both depending on cell line, start embedding when compact spheroids are formed). Day 2 2. Isolate immune cells from a buffy coat according to the kit protocol (see Subheading 2.5).

3. Pre-incubate NK cells with interleukins (IL-12: 5 ng/mL + IL18: 50 ng/mL) as positive control or with TGFβ (1 ng/mL TGFβ) as negative control. Activate T cells with CD3/CD28 beads. Respective compounds are added for 3 days: 200,000 NK or T cells per 6 well. Add IL-2 (10 ng/mL) to all samples containing NK cells [51]. Day 5 4. Prepare Matrigel/collagen solutions (see Subheading 3.2).

5. Prepare Matrigel/collagen mix for embedding of spheroids: 50% Matrigel +20% collagen (50 μL [8 mg Matrigel] per well + 20 μL [3 mg collagen]). 6. Detach fibroblasts from culture flask and wash in 1
RPMI (with 1
L-glut and P/S). 7. Stain immune cells green with carboxyfluorescein succinimidyl ester (CFSE) and fibroblasts with CellTracker Blue according to manufacturers’ instructions (1:1000 in 1
PBS for 20–30 min). 8. For the spheroid plate, remove 170 μL cell culture medium per well (leave 30 μL of medium in each well). 9. Mix and centrifuge fibroblasts and immune cells, aspirate supernatant, and resuspend the cells in desired volume of Matrigel/collagen mix. 10. Add IL-2 to all samples when using NK cells (10 ng/mL) and other cytokines for controls (see above) or compounds to the gel to the respective samples. 11. Load in six replicates 70 μL Matrigel/collagen mixed solution including cells or without cells for the background control (pipet slowly and dropwise to the edge of the well).

3D-3 Tumor Models to Study Immune Cell Infiltration

159

Co-culture: 1 tumor spheroid + 20,000 human dermal fibroblasts (HDFs) + 200,000 NK cells per well (see Note 10). Place the plate in 37  C incubator for at least 1.5 h. 12. Add 100 μL warm (37  C) 1
RPMI medium (with 2% FBS, 1
L-glut and P/S), add IL-2 to all NK samples and respective cytokines for controls and compounds. 13. Fill the outer wells with 1
PBS only to avoid the edge effect.

Tumor cells / Fibroblasts / Immune cells IL12 + IL18 activated NK cells

Calu-1

NCI-H1437

40

***

** **

30

** **

20

NCI-H157

**

10 n.s.

**

*

0

un

tr

ea

Unstimulated NK cells

b

te TG d Fb co m IL po 12 u un nd tr A ea te TG d Fb co m IL po 12 u un nd tr A ea te d TG Fb co m IL po 1 un 2 d A

a

% of infiltrated NK cells (PNK)

14. To analyze tumor cell killing, measure fluorescence intensity using a plate reader and generate Z-stacks with the confocal microscope for infiltration analysis (see Fig. 2).

Unstimulated T cells

CD3/CD28 activated T cells

c

Tumor cell killing

Viability [RFU mKate2]

200 150

Untreated PNK

100

PNK+IL12 50

0

24

48

72

96

Time [h]

Fig. 2 Immune cell infiltration and killing. Red fluorescent tumor cells were grown as floating 3D spheroids and embedded in a co-culture with fibroblasts (HDFs) and immune cells in Matrigel/collagen. In the case of NK cell activation, 1 ng/mL TGFß was used as a negative control [52–54], addition of IL-12 (5 ng/mL) alone or IL-12 (5 ng/mL) + IL-18 (50 ng/mL) served as a positive control [55, 56]. T cells pre-stimulated with CD3 and CD28 were used as positive control [57–59]. Compound A (2.5 μM) is a MAP-kinase inhibitor of TGFß downstream signaling. (a) Z-stack images (maximum intensity projection) show infiltration of activated primary NK or T cells to 3D tumor co-cultures (NCI-H157 dTomato) after 24 h. (b) Primary NK-cell (PNK) infiltration to 3D co-cultures was analyzed for three different NSCLC cell lines after 48 h: NCI-H1437, Calu-1, and NCI-H157. Statistical analysis was performed using Student’s t test (*0.05; **0.01; ***0.001). (c) Tumor cell killing of NCI-H157 mKate cells in co-culture with PNK cells was analyzed via measuring the fluorescence intensity of 3D tumors. IL-12-activated NK cells showed enhanced tumor cell killing compared to the control

160

4

Annika Osswald et al.

Notes 1. Phenol red-free medium is used when viability of tumor cells is measured via fluorescence intensity as phenol red might interfere with fluorescence measurements. 2. Matrigel and collagen are solid at 20  C, viscous liquids at 4  C, and solid gels at 37  C. Thaw Matrigel and collagen on ice at 4  C overnight or on ice at RT for 4 h. Maintain Matrigel and collagen on ice while working to prevent the gel from hardening, and use cold pipets if necessary. Solutions might be frozen at 20  C and refrozen if not fully used. 3. When a new cell line is tested, several tumor cell numbers (500–10,000 cells/well) should be analyzed for 3D tumor growth in ultra-low attachment plates. Use tumor cell medium for co-culture of tumor cells, fibroblasts, and immune cells. 4. Pipet slowly and do not blow out the pipet while mixing as this causes air bubble formation in Matrigel/collagen. 5. Test compounds also on monocultures to compare treatment effects with co-cultures. 6. Be careful while pipetting the alginate in order not to place too many bubbles in the suspension. 7. Sterile air is needed in the syringe to push the rest of the alginate cell suspension through the silicon tubing. 8. It is also possible to place the Falcon tube in a tube rack and wait for 5–10 min until alginate capsules are on the bottom of the tube. Carefully remove the supernatant without touching the alginate capsules. 9. Long-term cultivation of alginate capsules is possible up to 6–8 weeks, depending on the cell line behavior. 10. Plate six replicates (wells) and calculate with 10 wells per condition per cell line to have enough volume.

Acknowledgments We thank Nathalie Harrer for sharing her expertise on cultivation of immune cells and Martha Majewska for the technical support. References 1. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100:57–70 2. Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation. Cell 144:646–674

3. Fouad YA, Aanei C (2017) Revisiting the hallmarks of cancer. Am J Cancer Res 7:1016–1036

3D-3 Tumor Models to Study Immune Cell Infiltration 4. Dunn GP, Old LJ, Schreiber RD (2004) The three Es of cancer immunoediting. Annu Rev Immunol 22:329–360 5. Sukari A, Nagasaka M, Al-Hadidi A et al (2016) Cancer immunology and immunotherapy. Anticancer Res 36:5593–5606 6. Swann JB, Smyth MJ (2007) Immune surveillance of tumors. J Clin Invest 117:1137–1146 7. Ziani L, Chouaib S, Thiery J (2018) Alteration of the antitumor immune response by cancerassociated fibroblasts. Front Immunol 9:1–14 8. Erez N, Truitt M, Olson P et al (2010) Cancerassociated fibroblasts are activated in incipient neoplasia to orchestrate tumor-promoting inflammation in an NF-kappaB-dependent manner. Cancer Cell 17:135–147 9. Rudisch A, Dewhurst MR, Horga LG et al (2015) High EMT signature score of invasive non-small cell lung cancer (NSCLC) cells correlates with NFκB driven colony-stimulating factor 2 (CSF2/GM-CSF) secretion by neighboring stromal fibroblasts. PLoS One 10: e0124283 10. Pardoll DM (2012) The blockade of immune checkpoints in cancer immunotherapy. Nat Rev Cancer 12:252–264 11. Bianco A, Malapelle U, Rocco D et al (2018) Targeting immune checkpoints in non-small cell lung cancer. Curr Opin Pharmacol 40:46–50 12. Furue M, Ito T, Wada N et al (2018) Melanoma and immune checkpoint inhibitors. Curr Oncol Rep 20:29 ˜ o N, Reguart N (2018) 13. Remon J, Vilarin Immune checkpoint inhibitors in non-small cell lung cancer (NSCLC): approaches on special subgroups and unresolved burning questions. Cancer Treat Rev 64:21–29 ¨ ztop I (2017) Immune check14. Kazaz SN, O point inhibitors in advanced-stage non-small cell lung cancer. Turk Thorac J 18:101–107 15. Lugowska I, Teterycz P, Rutkowski P (2018) Immunotherapy of melanoma. Contemp Oncol (Pozn) 22:61–67 16. Rausch MP, Hastings KT (2017) Immune checkpoint inhibitors in the treatment of melanoma: from basic science to clinical application. In: Ward WH, Farma JM (eds) Cutaneous melanoma: etiology and therapy. Codon Publications, Brisbane. http://www.ncbi.nlm.nih. gov/books/NBK481851/ 17. Pillai RN, Behera M, Owonikoko TK et al (2018) Comparison of the toxicity profile of PD-1 versus PD-L1 inhibitors in non-small cell lung cancer: a systematic analysis of the literature. Cancer 124:271–277

161

18. Khoja L, Day D, Wei-Wu Chen T et al (2017) Tumour- and class-specific patterns of immune-related adverse events of immune checkpoint inhibitors: a systematic review. Ann Oncol 28:2377–2385 19. Gudjonsson T, Rønnov-Jessen L, Villadsen R et al (2002) Normal and tumor-derived myoepithelial cells differ in their ability to interact with luminal breast epithelial cells for polarity and basement membrane deposition. J Cell Sci 115:39–50 20. Pupovac A, Senturk B, Griffoni C et al (2018) Toward immunocompetent 3D skin models. Adv Healthc Mater 7(12):e1701405 21. Zhuang P, Sun AX, An J et al (2018) 3D neural tissue models: from spheroids to bioprinting. Biomaterials 154:113–133 22. Francipane MG, Lagasse E (2016) Towards organs on demand: breakthroughs and challenges in models of organogenesis. Curr Pathobiol Rep 4:77–85 23. Zscheppang K, Berg J, Hedtrich S et al (2018) Human pulmonary 3D models for translational research. Biotechnol J 13:1700268 24. Jackson EL, Lu H (2016) Three-dimensional models for studying development and disease: moving on from organisms to organs-on-achip and organoids. Integr Biol (Camb) 8:672–683 25. Sreepadmanabh M, Toley BJ (2018) Investigations into the cancer stem cell niche using in-vitro 3-D tumor models and microfluidics. Biotechnol Adv 36(4):1094–1110 26. Moreira L, Bakir B, Chatterji P et al (2017) Pancreas 3D organoids: current and future aspects as a research platform for personalized medicine in pancreatic cancer. Cell Mol Gastroenterol Hepatol 5:289–298 27. Rodenhizer D, Dean T, D’Arcangelo E et al (2018) The current landscape of 3D in vitro tumor models: what cancer hallmarks are accessible for drug discovery? Adv Healthc Mater 7 (8):e1701174 28. Hoarau-Ve´chot J, Rafii A, Touboul C et al (2018) Halfway between 2D and animal models: are 3D cultures the ideal tool to study cancer-microenvironment interactions? Int J Mol Sci 19(1):1–24 29. Caballero D, Blackburn SM, de Pablo M et al (2017) Tumour-vessel-on-a-chip models for drug delivery. Lab Chip 17:3760–3771 30. Weeber F, Ooft SN, Dijkstra KK et al (2017) Tumor organoids as a pre-clinical cancer model for drug discovery. Cell Chem Biol 24:1092–1100

162

Annika Osswald et al.

31. Langhans SA (2018) Three-dimensional in vitro cell culture models in drug discovery and drug repositioning. Front Pharmacol 9:6 32. Vanderburgh J, Sterling JA, Guelcher SA (2017) 3D printing of tissue engineered for in vitro modeling of disease progression and drug screening. Ann Biomed Eng 45:164–179 33. Jacobi N, Seeboeck R, Hofmann E et al (2017) Organotypic three-dimensional cancer cell cultures mirror drug responses in vivo: lessons learned from the inhibition of EGFR signaling. Oncotarget 8:107423–107440 34. Fang Y, Eglen RM (2017) Three-dimensional cell cultures in drug discovery and development. SLAS Discov 22:456–472 35. Osswald A, Sun Z, Grimm V et al (2015) Three-dimensional tumor spheroids for in vitro analysis of bacteria as gene delivery vectors in tumor therapy. Microb Cell Fact 14:199 36. Nyga A, Neves J, Stamati K et al (2016) The next level of 3D tumour models: Immunocompetence. Drug Discov Today 21:1421–1428 37. Hirt C, Papadimitropoulos A, Mele V et al (2014) In vitro 3D models of tumor-immune system interaction. Adv Drug Deliv Rev 79–80:145–154 38. Herter S, Morra L, Schlenker R et al (2017) A novel three-dimensional heterotypic spheroid model for the assessment of the activity of cancer immunotherapy agents. Cancer Immunol Immunother 66:129–140 39. Giannattasio A, Weil S, Kloess S et al (2015) Cytotoxicity and infiltration of human NK cells in in vivo-like tumor spheroids. BMC Cancer 15:351 40. Masopust D, Sivula CP, Jameson SC (2017) Of mice, dirty mice, and men: using mice to understand human immunology. J Immunol 199:383–388 41. Clevers H (2016) Modeling development and disease with organoids. Cell 165:1586–1597 42. Sachs N, Clevers H (2014) Organoid cultures for the analysis of cancer phenotypes. Curr Opin Genet Dev 24:68–73 43. Bartfeld S, Clevers H (2017) Stem cell-derived organoids and their application for medical research and patient treatment. J Mol Med (Berl) 95:729–738 44. Lancaster MA, Knoblich JA (2014) Organogenesis in a dish: modeling development and disease using organoid technologies. Science 345:1247125 45. Lancaster MA, Renner M, Martin CA et al (2013) Cerebral organoids model human brain development and microcephaly. Nature 501:373–379

46. Lancaster MA, Knoblich JA (2014) Generation of cerebral organoids from human pluripotent stem cells. Nat Protoc 9:2329–2340 47. Stock K, Estrada MF, Vidic S et al (2016) Capturing tumor complexity in vitro: comparative analysis of 2D and 3D tumor models for drug discovery. Sci Rep 6:28951 48. de Hoogt R, Estrada MF, Vidic S et al (2017) Protocols and characterization data for 2D, 3D, and slice-based tumor models from the PREDECT project. Sci Data 4:170170 49. Davies EJ, Dong M, Gutekunst M et al (2015) Capturing complex tumour biology in vitro: histological and molecular characterisation of precision cut slices. Sci Rep 5:17187 50. Rebelo SP, Pinto C, Martins TR et al (2018) 3D-3-culture: a tool to unveil macrophage plasticity in the tumour microenvironment. Biomaterials 163:185–197 51. Henney CS, Kuribayashi K, Kern DE et al (1981) Interleukin-2 augments natural killer cell activity. Nature 291:335–338 52. Bellone G, Aste-Amezaga M, Trinchieri G et al (1995) Regulation of NK cell functions by TGF-beta 1. J Immunol 155:1066–1073 53. Allan DS, Rybalov B, Awong G et al (2010) TGF-β affects development and differentiation of human natural killer cell subsets. Eur J Immunol 40:2289–2295 54. Zaiatz-Bittencourt V, Finlay DK, Gardiner CM (2018) Canonical TGF-β signaling pathway represses human NK cell metabolism. J Immunol 200(12):3934–3941 55. Alli RS, Khar A (2004) Interleukin-12 secreted by mature dendritic cells mediates activation of NK cell function. FEBS Lett 559:71–76 56. Tomura M, Zhou XY, Maruo S et al (1998) A critical role for IL-18 in the proliferation and activation of NK1.1+ CD3- cells. J Immunol 160:4738–4746 57. Jung G, Ledbetter JA, Mu¨ller-Eberhard HJ (1987) Induction of cytotoxicity in resting human T lymphocytes bound to tumor cells by antibody heteroconjugates. Proc Natl Acad Sci U S A 84:4611–4615 58. Turka LA, Ledbetter JA, Lee K et al (1990) CD28 is an inducible T cell surface antigen that transduces a proliferative signal in CD3+ mature thymocytes. J Immunol 144:1646–1653 59. Riddell SR, Greenberg PD (1990) The use of anti-CD3 and anti-CD28 monoclonal antibodies to clone and expand human antigenspecific T cells. J Immunol Methods 128:189–201

Chapter 11 Establishment and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Adenocarcinoma for Application in Drug Discovery Julia C. Meier-Hubberten and Michael P. Sanderson Abstract The high attrition rate of oncology drug candidates can be in part explained by the disconnect between the standard preclinical models (e.g., 2D culture, xenograft tumors) commonly employed for drug discovery and the complex multicellular microenvironment of human cancers. As such, significant focus has recently shifted to the establishment of preclinical models that more closely recapitulate human tumors, such as patient-derived xenografts, 3D spheroids, humanized mice, and mixed-culture models. For these models to be suited to drug discovery, they should optimally exhibit reproducibility, high-throughput, and robust and simple assay readouts. In this article, we describe a protocol for the generation of an in vitro 3D co-culture spheroid model that recapitulates the interaction of tumor cells with stromal fibroblasts in pancreatic adenocarcinoma. We additionally describe protocols relevant to the analysis of these spheroids in highthroughput drug discovery campaigns such as the assessment of spheroid proliferation, immunofluorescence and immunohistochemistry staining of spheroids, live-cell and confocal imaging and analysis of cell surface markers. Key words 3D co-culture, Confocal imaging, Drug discovery, Immunofluorescence, Immunohistochemistry, Live-cell analysis, Pancreatic cancer, Spheroids

1

Introduction The attrition rate for new molecular entities (NMEs) in oncology has been estimated at approximately 95% [1]. Given the large investment required for progression of NMEs to clinical development, and the mostly modest and incremental increases in patient benefit delivered by many approved new therapies, strategies to optimize the R&D process have received extensive attention [2–4]. A great deal of focus has been made in recent years to the establishment of preclinical models with improved translational predictivity for the response of human cancers to drug candidates. Historically, oncology drug discovery has typically employed cancer

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_11, © Springer Science+Business Media, LLC, part of Springer Nature 2019

163

164

Julia C. Meier-Hubberten and Michael P. Sanderson

cell lines in standard in vitro cellular assays (e.g., 2D proliferation) [5–7] and subcutaneous tumor xenografts in immunocompromised mice. While the reproducibility and throughput of these models are attractive for the rapid screening of compounds, it is generally agreed that these models do not adequately recapitulate the complex molecular evolution, multicellular microenvironment, and pathogenesis of human tumors. As such, the establishment of preclinical systems more closely mimicking the tumor microenvironment (TME) are of importance for identification of drug candidates with the greatest chance of delivering a meaningful benefit to patients. The last 5–10 years has seen a plethora of publications describing creative preclinical models claiming to recapitulate human tumors. Prominent and extensively reviewed examples include patient-derived xenografts (PDXs) [8], patient avatar models [9], and organoids [10]. However, the low throughput and heterogeneity of PDXs and tumoroids, compared to standard cell lines, complicate their application in commonly used experimental approaches such as phenotypic compound screens, highthroughput combinations, or genetic depletion screens. Ideally a preclinical model for drug discovery should recapitulate the human tumor while also being applicable to rapid and large-scale compound testing. Pancreatic adenocarcinoma has the lowest 5-year survival rate of all cancers in the USA, and most approved treatments are minimally efficacious [11]. The disease is the fourth leading cause of cancer-related mortality and is predicted to rise to the second leading cause before 2030 [12]. The lack of development of improved therapies for pancreatic adenocarcinoma can be partly explained by the poor translational utility of standard preclinical models of this disease [13]. Pancreatic adenocarcinomas are commonly characterized by the presence of a large stromal compartment containing cancer-associated fibroblasts (CAFs), immune cells, and a rich extracellular matrix (ECM), all of which interact closely with pancreatic cancer cells to modulate the pathogenesis, drug penetration, and chemotherapy sensitivity of the disease [14, 15]. Within the pancreatic adenocarcinoma TME, CAFs have been described to surround carcinomatous structures in a perineoplastic ring [16]. Stromal components of pancreatic cancer may also offer potential therapeutic targets [17, 18] and may predict patient outcome [19]. In this article, we describe protocols for generation and analysis of 3D co-culture spheroids of KP-4 pancreatic adenocarcinoma cells and NIH/3T3 fibroblasts in vitro. We focus on the 3D interaction of these cells, as opposed to 2D co-culture systems, for several reasons. Firstly, the 3D growth of cancer cells is commonly accepted to recapitulate the cell-cell interactions of human tumors more closely than cancer cells grown in 2D [5–7]. Secondly, the

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

165

drug sensitivity of tumor cells grown in 3D more accurately predicts their sensitivity in vivo [20, 21], potentially due to alterations in gene expression, signaling, cell polarization, and ECM components [22–24]. The protocols described in this article can be used in either high-throughput drug discovery campaigns or in-depth characterization of the mechanism of action of more advanced drug candidates. Furthermore, these protocols can be readily adapted for the establishment and analysis of 3D co-culture spheroid models from alternate pancreatic cancer cells and fibroblasts.

2

Materials All reagents should be stored at room temperature (RT), unless stated otherwise.

2.1 Preparation of 3D Co-Culture Spheroids and Compound Treatment

2.2 Assessing 3D Co-Culture Spheroid Proliferation

2.3 Non-AntibodyBased Staining to Monitor Cancer Cells and Fibroblasts in Spheroids

l

KP-4 human pancreatic adenocarcinoma cell line (JCRB #JCRB0182).

l

NIH/3T3 mouse fibroblast cells (ATCC #CRL-1658™).

l

Dulbecco’s Modified Eagle Medium (DMEM). Store at 4
C.

l

Fetal calf serum (FCS). Store at 20
C.

l

Dulbecco’s phosphate-buffered saline (DPBS). Store at 4
C.

l

175 cm2 cell culture flasks.

l

Trypsin-ethylenediaminetetraacetic acid (EDTA). Store at 4
C.

l

Cell counter device or hemocytometer.

l

Black ultra-low attachment (ULA) 96-well plates (Corning #4515).

l

Plate centrifuge.

l

CellTiter-Glo® 3D Cell Viability Assay (Promega #G9681). Store at 20
C.

l

Thermomixer with a microplate block and lid.

l

Microplate reader with a luminescence filter.

l

GraphPad Prism Software.

l

CellTracker™ Green CMFDA Dye (Thermo Fisher Scientific #C2925). Store at 4
C.

l

CellTracker™ Red CMTPX Dye (Thermo Fisher Scientific #C34552). Store at 4
C.

l

DAPI (40 ,6-diamidino-2-phenylindole). Store at 4
C.

l

Electronic multichannel pipette.

166

Julia C. Meier-Hubberten and Michael P. Sanderson

2.4 Antibody-Based Staining of Spheroids for Immunofluorescence (IF)

2.5 Antibody-Based Staining of Spheroids for Immunohistochemistry (IHC)

l

0.5% Triton-X™-100 in DPBS. Store at 4
C.

l

3% FCS/DPBS blocking buffer. Store at 4
C.

l

Primary and secondary antibodies suitable for IF assays of the researcher’s choice. We suggest using AlexaFlour®647 and AlexaFlour®488 secondary antibodies for IF.

l

4% paraformaldehyde (PFA) in DPBS. Store at 4
C.

l

DAPI, see Subheading 2.3.

l

4% PFA, see Subheading 2.4.

l

1.5 mL polypropylene tubes.

l

Prepare a fresh 1% eosin Y solution by dilution of eosin Y stock solution (Sigma-Aldrich #318906) in acetic acid. Filter through a membrane filter (e.g., 0.45 μm pore size).

l

30% sucrose in ultrapure water. Store at 4
C.

l

Isopentane.

l

Optimal cutting temperature (OCT) medium (Sakura #4583).

l

Tissue-Tek® Cryomolds® (Ted Pella Inc. #27181).

l

Cryotome.

l

Plate heater.

l

Prepare a 0.1% Tris-buffered saline with Tween® 20 (TBST) solution by diluting 10 TBS in ultrapure water. Add 0.1% Tween® 20. Store at 4
C.

l

PAP-Pen Liquid Blocker (Thermo Fisher #008899).

l

Peroxidase Blocking Solution (Dako #S202386-2). Store at 4
C.

l

Protein block serum-free solution (Dako #X0909). Store at 4
C.

l

Primary and secondary antibodies of the researcher’s choice. We suggest using EnVisionþ System- HRP-Labeled Polymer (Dako) reagents.

l

Liquid DABþ Substrate Chromogen System (Dako #K3468). Store at 4
C.

l

Mayer’s hemalum solution (Sigma-Aldrich #1092490500).

l

Ethanol absolute.

l

Isopropanol.

l

Xylene.

l

SuperFrost® Plus glass slides (Fisher Scientific #10149870).

l

NanoZoomer S60 Digital slide scanner (Hamamatsu Photonics K.K. #C13210-01).

l

HALO™ image #CLS141255).

analysis

software

(PerkinElmer

Inc.

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

2.6 Live-Cell Imaging of 3D Co-Culture Spheroids

l

CellTracker™ dyes, see Subheading 2.3.

l

DAPI, see Subheading 2.3.

l

IncuCyte® instrument (Essen BioScience).

2.7 Automatic Confocal Imaging

l

Gas mixer (HiTec Zang GmbH).

l

CQ1 Confocal Imaging Cytometer (Yokogawa).

2.8 Analysis of Cell Surface Markers from Spheroids

l

Trypsin-EDTA, see Subheading 2.1.

l

Fluorophore-labeled antibody of researcher’s choice.

l

3% FCS/DPBS blocking buffer.

l

DAPI, see Subheading 2.3.

l

FACSCanto™ II system (BD Biosciences).

l

FlowJo® analysis software (FlowJo, LLC).

3

167

Methods

3.1 Preparation of 3D Co-Culture Spheroids and Compound Treatment

We describe below the generation of 3D co-culture spheroids from KP-4 human pancreatic cancer cells and mouse NIH/3T3 fibroblasts. The spheroid should ideally display a compact round structure of 150–500 μm, potentially with a fibrotic capsule, nutrient and oxygen gradients and in some cases a lumen resulting from the differentiation of cellular layers [25]. All steps should be performed at RT, unless stated otherwise. Considerations for the use of other plate formats and the establishment of 3D co-culture spheroids from alternate pancreatic cancer cells and fibroblasts are discussed (see Notes 1–5). 1. Grow KP-4 and NIH/3T3 cells separately in 175 cm2 cell culture flasks in DMEM/10% FCS to a density of 70% in an incubator (37
C/5% CO2). 2. Aspirate the supernatants slowly using sterile aspirator tips. 3. Wash the cells once with DPBS (10 mL/flask) and aspirate. 4. Add 2.5 mL trypsin/EDTA, and place the flasks in the incubator for 5–10 min. 5. Add 7.5 mL DMEM/10% FCS, and centrifuge the solutions for 3 min at 500  g. 6. Aspirate the supernatants, and resuspend the cell pellets in 10 mL DMEM/10% FCS. 7. Count cells and then dilute in DMEM/10% FCS to the desired cell concentration. 8. Seed NIH/3T3 cells first into in black ULA 96-well plates 4 h before seeding KP-4 cells. Seed NIH/3T3 cells and KP-4 cells

168

Julia C. Meier-Hubberten and Michael P. Sanderson

at a 2:1 ratio and use between 750 and 1000 total cells per well in 100 μL DMEM/10% FCS. 9. Tap the plates, centrifuge for 1 min at 500  g, and incubate for 1 day. 10. Add compounds in triplicate (50 μL/well) to the desired dose range (e.g., 10 μM–1 nM, using 1:10 dilution steps) and incubate for 1–6 days. 3.2 Assessing 3D Co-Culture Spheroid Proliferation

Measurement of ATP concentrations is a straightforward approach for the assessment of spheroid proliferation. We describe below the use of the CellTiter-Glo® 3D Cell Viability Assay. Our experience has shown that this approach lyses the inner cells of the spheroids while leaving the outer capsule intact. Considerations and references to other methodologies for analysis of spheroid proliferation, apoptosis, and gene expression are discussed (see Notes 6 and 7). 1. Follow steps outlined in Subheading 3.1 to generate spheroids. 2. Remove 100 μL of the supernatant from each well, and equilibrate the plates for 10 min. 3. Add 100 μL of the CellTiter-Glo® 3D Cell Viability Assay reagent to each well and shake for 5 min on an orbital shaker at 20  g in the dark. 4. Incubate for a further 25 min in the dark without shaking, and measure luminescence using a microplate reader (500 ms/ well). 5. Calculate IC50 values using GraphPad Prism Software (nonlinear regression, four parametric).

3.3 Non-AntibodyBased Staining to Monitor Cancer Cells and Fibroblasts in Spheroids

Staining reagents such as DAPI, Hoechst 33342, SYTOX green, Calcein AM, and propidium iodide (PI) do not typically enable separate monitoring of the KP-4 and NIH/3T3 compartments of the spheroids. However, this can be achieved by separately staining the pancreatic cancer cells and fibroblasts with different CellTracker™ reagents prior to seeding for spheroid formation, as described below. An alternate approach for separate monitoring of the pancreatic cancer cell and fibroblast compartments of 3D co-culture spheroids is discussed (see Note 8). 1. Follow steps 1–7 of Subheading 3.1. 2. Add CellTracker™ Green CMFDA Dye (1:1000) to the NIH/3T3 cell stock and CellTracker™ Red CMTPX Dye (1:1000) to the KP-4 cell stock, and incubate each mixture for 30 min. 3. Centrifuge the cells for 3 min at 500  g, aspirate the supernatants, and resuspend the stained cell pellets in DMEM/ 10% FCS.

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

169

4. Follow steps 8–10 of Subheading 3.1 to generate spheroids. 5. Incubate the plates for the desired times, and perform IF and imaging as described below. 3.4 Antibody-Based Staining of Spheroids for Immunofluorescence

We describe below the protocol for staining of 3D co-culture spheroids for IF. Following this procedure, whole spheroids can be analyzed using the imaging techniques described in the following sections. 1. Follow steps outlined in Subheading 3.1 to generate spheroids. 2. Gently aspirate the supernatant from each well using an electronic multichannel pipette. Avoid disrupting or losing the floating spheroids. 3. Fix the spheroids in 4% PFA/DPBS (50 μL/well) for 20 min. 4. Gently aspirate and wash the spheroids once with DPBS (100 μL/well). At this stage the plates can be wrapped in parafilm and stored at 4
C until further use. 5. Permeabilize the spheroids in 0.5% Triton-X-100/DPBS (50 μL/well) for 20 min. 6. Gently aspirate and wash once with DPBS (100 μL/well). 7. Block for 3 h in 3% FCS/DPBS (50 μL/well). 8. Gently aspirate and incubate with the primary antibody diluted in 3% FCS/DPBS (50 μL/well) for 2 days at 4
C (see Note 9). 9. Gently aspirate and wash once with DPBS (100 μL/well). 10. Incubate the spheroids with the fluorescently labeled secondary antibodies diluted in 3% FCS/DPBS (50 μL/well) for 1 h. 11. Aspirate the supernatant and add DAPI (diluted 1:2500 in DPBS, 50 μL/well) and incubate for 20 min in the dark. 12. Gently aspirate and wash once with DPBS (100 μL/well). 13. Layer the stained spheroids with 200 μL DPBS at 4
C for confocal imaging (Subheading 3.7). Protect the plates from light sources.

3.5 Antibody-Based Staining of Spheroids for Immunohistochemistry

IHC can be used to analyze deep cell layers of spheroids. As spheroids are difficult to identify in paraffin-embedded or freezing medium blocks for subsequent sectioning for IHC, we suggest staining pooled spheroids prior to embedding, as described below. 1. Follow steps outlined in Subheading 3.1 to generate spheroids. 2. Gently aspirate the medium using an electronic multichannel pipette. 3. Fix the spheroids in 4% PFA/DPBS (50 μL/well) overnight at 4
C.

170

Julia C. Meier-Hubberten and Michael P. Sanderson

4. Carefully aspirate and resuspend the spheroids in DPBS (100 μL/well). 5. Carefully collect the spheroids using a 1000 μL pipette tip and transfer to 1.5 mL polypropylene tubes. Pool 3–5 spheroids per compound treatment into each tube. 6. Centrifuge the spheroids into a pellet for 1 min at 100  g using a benchtop centrifuge. 7. Aspirate all but approximately 25 μL DPBS from the tube, add 1% eosin solution (50 μL/tube), and incubate for 10 min. 8. Centrifuge for 1 min at 100  g, and wash the spheroids three times with 1 mL DPBS. 9. Aspirate most of the supernatant, and then gently mix the spheroids into a 30% sucrose solution (50 μL/tube) without generating air bubbles. Incubate overnight at 4
C. 10. Cool a reservoir of isopentane to 20
C by the addition of dry ice, and place a metal box with a flat surface in the reservoir. Perform in a fume hood. 11. Fill at least one third of a Tissue-Tek® Cryomold® with OCT medium, and place the mold on the cooled metal box surface. 12. Cool the spheroids in the tubes to 20
C and then centrifuge at 100  g for 1 min at 4
C. 13. Aspirate most of the sucrose solution, and transfer the spheroids to the top of the frozen OCT medium in the molds. 14. Cover the spheroids with fresh OCT medium and allow to freeze at 20
C. 15. Release the cryo blocks from the molds. At this stage the blocks can be wrapped in aluminum foil and stored at 80
C until further use. 16. Prepare 5–10 μm sections using a cryotome, and place the sections on SuperFrost® Plus glass slides. 17. Dry the sections for 30 min at 37
C on a plate heater. At this stage the slides can be stored at 80
C in slide boxes containing silica beads to absorb moisture. 18. Wash the slides three times for 2 min in TBST 0.1%, and mark the sections with a PAP-Pen Liquid Blocker. 19. Add Peroxidase Blocking Solution (70–100 μL per section) to block endogenous peroxidase activity and incubate for 10 min. 20. Add protein block serum-free solution (70–100 μL per section) to block unspecific IgG binding for 10 min. 21. The primary antibody of the researcher’s choice can be incubated with the section for either 1 h at RT or overnight at 4
C in a humid chamber. We have used the following antibodies on 3D co-culture spheroids from KP-4 and NIH/3T3 cells:

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

171

l

Biotinylated Ki-67 antibody (Dako #MIB-1) for assessing proliferation.

l

Cleaved-caspase-3 antibody (Cell Signaling Technology #Asp175) for assessing apoptosis.

l

Vimentin antibody (Abcam #ab92547) to stain NIH/3T3 fibroblasts.

l

Ras (mutated G12 D) antibody (Abcam #ab221163) to stain KP-4 cells.

22. Wash the slides three times for 2 min in TBST 0.1%. 23. Add the horseradish peroxidase (HRP)-labeled secondary antibodies for 30 min. 24. Wash the slides three times for 2 min in TBST 0.1%. 25. Incubate the slides with DABþ Substrate (1:50 in 3% FCS/DPBS) for 1–10 min under constant microscopic observation until the brown staining becomes visible. 26. Stop the HRP reaction with distilled water and counterstain with Mayer’s hemalum solution for 45 s. 27. Enable bluing of the spheroids under constant tap water flow for 10 min. 28. Sequentially dehydrate the slides in 60%, 70%, 80%, and 96% ethanol (2 min per step). Further dehydrate the slides by washing twice in isopropanol (2 min per step) and three times in xylene (5 min per step). 29. Cover the slides with mounting medium and cover slips, and dry the slides overnight. 30. Scan the slides with a digital slide scanner, and count positively stained cells using a quantitative imaging software such as HALO™ image analysis software. 3.6 Live-Cell Imaging of 3D Co-Culture Spheroids

An imager-like plate reader with an integrated non-confocal objective allows basic 2D measurements of different parameters of a 3D object such as area, perimeter, sphericity, and calculated volume as well as sophisticated fluorescence-based quantification. In general, these systems have a simple integrated image analysis platform for the bright-field-based quantification of spheroid area, counting of spheres, or quantification of fluorescence (e.g., from PI-stained spheroids). We describe below the live-cell imaging of spheroids using the IncuCyte® instrument. 1. Follow Subheading 3.3 to stain KP-4 and NIH/3T3 cells with CellTracker™ reagents, and then generate 3D co-culture spheroids as described in steps 7–10 of Subheading 3.1.

172

Julia C. Meier-Hubberten and Michael P. Sanderson

2. Place the plates into the IncuCyte® for bright-field and immunofluorescent time-lapse imaging for a maximum of 3–4 days (longer incubations can lead to phototoxicity). 3. Use 6 h intervals and 4 or 10 objectives depending on the spheroids size. The image analysis mode “mono-spheroid” enables assessment of spheroid area, perimeter, and IF-based readouts and the determination of dose-response curves. 4. Quantitate the CellTracker™ Green and CellTracker™ Red mean intensities to determine the composition of NIH/3T3 fibroblasts and the KP-4 cells, respectively, in each spheroid. 5. A bright-field confluency mask can be used in conjunction with DAPI staining (see Subheading 3.4) to evaluate spheroid size. 3.7 Automatic Confocal Imaging

Automated confocal imagers can be used for high-throughput image capture and on-the-fly image analysis. We describe here the use of the CQ1 Confocal Imaging Cytometer for automated image acquisition and high-content analysis of KP-4 and NIH/3T3 cells of 3D co-culture spheroids. Considerations for the use of confocal imagers are discussed (see Notes 10–13). Furthermore, an alternate imaging approach amenable to 3D co-culture spheroids is discussed (see Note 14). 1. Follow Subheading 3.3 to stain KP-4 and NIH/3T3 cells with CellTracker™ reagents, and then generate 3D co-culture spheroids as described in steps 7–10 of Subheading 3.1. 2. Fill the CQ1 water bath with 25 mL of distilled water, and set the gas mixer to 5% CO2. 3. Follow the software instructions to select the plate type and wells to be analyzed and the desired objective, number of fields of view, and time interval for image acquisition. 4. For imaging of spheroids of approximately 150 μm diameter, use eight z-stacks with 20 μm steps. An entire spheroid in one representative control well should be imaged. For larger spheroids, increase the steps to 30–40 μm. The fields of view should be distributed equally across the wells for a representative count of cells. 5. Select the image analysis module “spheroid” to quantify the CellTracker™ Green- and CellTracker™ red-stained cells in each well and z-stack. The total cell number, mean fluorescence intensity, diameter, circularity, and branch size can be automatically quantified.

3.8 Analysis of Cell Surface Markers from Spheroids

We describe below the protocol for dissociation of cells from 3D co-culture spheroids and subsequent analysis of surface protein expression by flow cytometry. Alternate approaches and considerations for cell dissociation are discussed (see Note 15).

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

173

1. Follow steps outlined in Subheading 3.1 to generate spheroids. 2. Process the spheroids into a single-cell suspension using stringent cell separation buffers. Depending on the density and thickness of the fibrotic capsule, dissociation reagents based on EDTA, collagenase, or trypsin can be used for between 15 and 60 min under gentle shaking (20  g) at 37
C. 3. Centrifuge the plates for 3 min at 500  g at 4
C. 4. Gently aspirate the dissociation reagent with an electronic multichannel pipette. 5. Add the desired fluorophore-labeled antibodies (50 μL/well, diluted in 3% FCS/DPBS) and incubate for 2 h at 4
C in the dark. 6. Centrifuge the plates for 3 min at 500  g at 4
C. 7. Gently aspirate and wash the spheroids once with ice-cold DPBS (100 μL/well). 8. Stain the spheroids with DAPI (1:2500 in DPBS, 50 μL/well) for 10 min in the dark at 4
C. 9. Perform flow cytometry using an instrument such as the Canto II (BD Biosciences) and quantify positively stained cells with FlowJo® analysis software.

4

Notes 1. Other scaffold-free plate formats such as hanging drop plates or rotation in bioreactors offer an alternative to ULA plates. These formats are amenable to the isolation of spheroids and their conditioned media for subsequent analyses (e.g., assessment of gene expression and mutations), IF-based high-content analysis, embedding for histology, and cryo-banking. Although bioreactors offer the advantage of encouraging continuous nutrient supply, they typically involve larger culture volumes and are thus less applicable to testing of compound libraries in drug discovery campaigns. Nevertheless, bioreactors have been used to model complex interactions between different cell types in the TME and are thus well suited for analysis of the mechanism of action of more advanced drug candidates. For example, an in vitro bioreactor co-culture model of MDA-MB-231 breast cancer cells and murine MC3T3E-1 pre-osteoblasts, separated by a dialysis membrane, was used to recapitulate the processes of bone metastasis [26]. 2. Scaffolds and scaffold-containing plate formats are also applicable to the generation of 3D co-culture spheroids. Scaffolds comprise mostly biocompatible, viscous, and inert media components which encourage 3D spheroid formation. They feature

174

Julia C. Meier-Hubberten and Michael P. Sanderson

the mechanical properties, growth factors, cytokines, and nutrients found in the ECM. Hydrogels are common scaffolds for 3D cultures and contain water and biomolecules such as collagen, agarose, hyaluronic acid, laminin, or synthetic or natural polymers (inert or biodegradable). The gelling properties of polymers such as polyethylene glycol make them amenable to manipulation. Additionally, non-gel polymers such as nanofibers can mimic specific tissue structures such as blood vessels. Nanofibers typically combine two different materials such as ceramics and collagen and can be either nontransparent or transparent to enable live-cell imaging. Reconstituted basement membranes such as Matrigel™ and Cultrex® offer an alternative to hydrogels. These materials are isolated from rodent tumor models and consist of collagen type IV, HSPG2, and laminin. However, it should be noted that these scaffolds display batch-to-batch variability, lack human factors, and have even been described to directly influence expression of miRNAs and noncoding RNAs [27]. For co-cultures embedded in scaffolds, pancreatic cancer cells and fibroblasts should be mixed together with the scaffold at different ratios prior to seeing on microtiter plates and layered with media. 3. Other sources of pancreatic cancer cells can also be employed for generation of 3D co-culture spheroids. The wealth of characterization of established cell lines provides an advantage for drug discovery in that cell lines with specific molecular features can be selected to match the desired mechanism of action of a drug (e.g., selection of KRAS G12C mutant cancer cells for testing of KRAS G12C inhibitors) [20]. Furthermore, the homogeneity of cell lines also provides the advantage of high assay reproducibility in drug discovery campaigns. In recent years, in vitro cellular models derived from PDX explants have been described for pancreatic adenocarcinoma [28] and other tumor types such as non-small cell lung carcinoma [29]. These offer the advantage of recapitulating the mutational heterogeneity and multicellular architecture of human tumors. However, the murine origin of stromal cells from PDX explants may complicate the testing of compounds specifically targeting human factors. Furthermore, some human proteins, such as the receptor tyrosine kinase c-Met, are not adequately activated by mouse ligands [30], thus complicating the analysis of paracrine interactions between mouse and human cellular compartments. Accordingly, several groups have described the shortterm drug treatment of 3D in vitro tumor models derived from primary patient tumor material [31, 32]. While such models are of high translational relevance for assessment of a drug’s mechanism of action, the limited and transient availability of the tumor material complicates their application in drug discovery campaigns.

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

175

4. Immortalized fibroblasts (e.g., HS5, TelCOFS02MA), fibroblast cell lines (e.g., MC-IXC, COLO 829), and primary human pancreatic fibroblasts (e.g., Creative Bioarray #CSC8246W) and pancreatic CAFs (e.g., Vitro Biopharma #CAF08) could be used in place of NIH/3T3 cells for establishment of a 3D co-culture spheroid model. The isolation of primary CAFs from pancreatic cancer patients has also been recently described [33] and would likely represent the most “disease-relevant” model. However, the limited availability of such cells is a barrier to application in drug discovery. 5. When establishing a new 3D co-culture spheroids model, we recommend testing different plate formats and media to achieve the desired spheroid architecture. Conditions leading to the formation of separate bulbs of tumor cells and fibroblasts or loss of cell viability should be avoided. Different ratios of pancreatic cancer cells and fibroblasts (e.g., 1:1, 1:1.5, 1:2, 1:4, etc.) should be tested to identify the optimal conditions for spheroid formation. It is important to examine the effects of parallel and sequential seeding of each pancreatic cancer and fibroblast cell type. In some cases, parallel seeding can lead to the undesirable formation of a central core of fibroblasts surrounded by tumor cells or the separate formation of bulbs of each cell type. 6. Non-lysis methods to measure ATP levels, lactate dehydrogenase (LDH), or the use of tetrazolium dyes such as 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) with supernatant samples are alternatives for assessment of 3D spheroid proliferation [34]. However, we have found that the dose-response activity of compounds can be more accurately determined from 3D spheroids when using lysis-based approaches. Lysis-based luminescence methods for assessment of apoptosis such as the Caspase-Glo® 3/7 reagent have also been described for 3D spheroid monocultures of either cancer cell lines [35] or mesenchymal cells [36]. The effects of compounds on hypoxia in 3D co-culture models can be assessed using several approaches. Hypoxia-inducible factor1α (HIF-1α) target genes such as CA9 and DDIT4 can be readily assessed in 3D spheroid extracts using quantitative RT-PCR. Meanwhile, oxygen consumption and extracellular acidification (ECA) in 3D spheroids can be assessed using porphyrin-based phosphorescent oxygen-sensitive and [Hþ] pH-sensitive probes, respectively. 7. In addition to direct assessment of the viability of spheroids, image-based analysis can be used to determine spheroid shrinkage upon compound treatment. The Celigo system (Nexelom Biosciences) allows bright-field-based quantification of spheroid area, diameter, and sphericity based on a 2D modus. This

176

Julia C. Meier-Hubberten and Michael P. Sanderson

instrument can be used in conjunction with immunofluorescent staining of spheroids with Hoechst 33342 (Thermo Fisher #H1399) and SYTOX™ green (Thermo Fisher #S7020) to quantify apoptosis within spheroids. This provides an alternative to the CQ1 or IncuCyte® devices. 8. Transfection of the individual cell types with expression vectors for red- or green-fluorescent protein is another wellestablished approach, which circumvents complications associated with heterogenous stain distribution while also allowing longer incubation times. Transfection with such markers is typically amenable to cell lines but may require optimization when using primary cells. However, photocytotoxicity associated with these approaches can complicate the establishment of a robust assay [37]. 9. This relatively long incubation time with primary antibodies is necessary due to limited diffusion through the deeper cellular layers of the spheroids. 10. A confocal microscope enables imaging to a depth of up to 250 μm for the visualization of small subcellular structures within at least the first 2–3 cell layers of the spheroid. Flexible adjustment of the z-axis (focus plane) is also possible and is important when analyzing floating spheroids and 3D objects embedded in scaffolds. Depending on the fibroblast-related collagen distribution, clearing with Visikol®, iDISCO [38], or ClearT [39] can be performed to reduce desmoplasticdependent light scattering and allow imaging up to 500 μm. However, it should be noted that these clearing agents can potentially alter cell morphology and surface markers [40]. 11. Image analysis and automatic plate scanning are not commonly installed features of a confocal imaging stage. This can be overcome by integration of a joystick station for automatic well imaging. Additionally, selected software packages from microscope providers or open-source cell image software such as CellProfiler [41] can be employed for sophisticated 3D analyses which are normally not installed by confocal microscope software analysis surfaces. 12. Algorithms designed for 3D applications are available to separate weak signals from cells of the deeper layers at 250–500 μm from stronger signals originating from surface cells at 20 μm. Furthermore, algorithms filling the holes of low fluorescence intensity are useful for nuclei segmentation in deeper cell layers (e.g., when using DAPI). However, setting of the optimal evaluation parameters can be complicated when 3D spheroids contain low intensity at cell layers deeper than 100–250 μm. 13. In general, imaging software algorithms average the objects counted in each field and generate a mean or median readout

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer

177

of the fluorescence intensity per pixel. Data for features of the spheroids such as the mean and median fluorescence intensity, sphericity, and branch-like patterns can be exported for the generation of dose-response curves of compound activity. In addition, the nucleus-to-cytoplasm ratio can be used to assess the cell cycle and apoptosis. Consideration should also be made as to whether each z-stack layer is analyzed individually or as a maximum intensity projection (MIP) image. An MIP image combines the z-stacks and generates average intensities. For accurate 3D analysis, we recommend analyzing each z-stack individually and summarizing the data afterward. This approach can generate very large data volumes. For example, analysis of a single 96-well plate with three z-stacks, four fields of view, and a 10 objective at 6 h intervals over 6 days with three colors will generate approximately 500 GB of data in a CQ1 imager platform. 14. Alternate imaging approaches can be used for 3D spheroids, the choice of which depends on the assay format (e.g., floating spheroids versus embedded spheroids in scaffolds) and spheroid size. Light sheet fluorescence microscopy (LSFM) has emerged as a powerful tool for 3D imaging of spheroids. This approach uses a thin plane of light to illuminate a sample orthogonally to the detection objective such that the axial resolution of the microscope is determined wholly or partly by the thickness of the light sheet. In contrast to the conventional confocal microscopy techniques described above, LSFM reduces the scattering of light to enable imaging of larger objects up to 1–2 mm. However, LSFM is a low-throughput approach, and thus we recommend its use for the detailed analysis of the mechanism of action of more advanced drug candidates. LSFM was used to generate complete images of BxPC-3 pancreatic cancer spheroids of 140 μm diameter [42]. In another report, the division of Capan-2 pancreatic cancer cells within spheroids of 400 μm diameter was imaged in real-time using LSFM [43]. 15. A recent publication described the use Accutase® to dissociate 3D mixed-culture spheroids containing human pancreatic cancer cells, MRC-5 fetal lung fibroblasts, and monocyte-derived macrophages (MDMs) for subsequent flow cytometry analysis of the surface markers EpCAM (pancreatic cancer cell-specific marker) and CD11b (macrophage-specific marker) and unstained fibroblasts [44]. It should be noted that extended treatment with dissociation reagents can reduce the viability of cells within the spheroids. An elegant recent publication described a novel in vitro method for the rapid isolation individual cell types from 3D mixed-culture structures for subsequent analysis, without the need for extensive chemical

178

Julia C. Meier-Hubberten and Michael P. Sanderson

dissociation [45]. Tumor cells were seeded onto a porous cellulose strip and rolled onto a small spindle to form a 3D structure containing different cell layers. Following incubation and treatment with compounds, unrolling of the sheet enabled the isolation of spatially defined cell populations for molecular and flow cytometry analyses. Potential Conflicts of Interest: Both authors are employees of Merck KGaA. References 1. Hutchinson L, Kirk R (2011) High drug attrition rates—where are we going wrong? Nat Rev Clin Oncol 8(4):189–190 2. Cook D, Brown D, Alexander R et al (2014) Lessons learned from the fate of AstraZeneca’s drug pipeline: a five-dimensional framework. Nat Rev Drug Discov 13(6):419–431 3. Pammolli F, Magazzini L, Riccaboni M (2011) The productivity crisis in pharmaceutical R&D. Nat Rev Drug Discov 10(6):428–438 4. Paul SM, Mytelka DS, Dunwiddie CT et al (2010) How to improve R&D productivity: the pharmaceutical industry’s grand challenge. Nat Rev Drug Discov 9(3):203–214 5. Wilding JL, Bodmer WF (2014) Cancer cell lines for drug discovery and development. Cancer Res 74(9):2377–2384 6. Selby M, Delosh R, Laudeman J et al (2017) 3D models of the NCI60 cell lines for screening oncology compounds. SLAS Discov 22 (5):473–483 7. Benien P, Swami A (2014) 3D tumor models: history, advances and future perspectives. Future Oncol 10(7):1311–1327 8. Byrne AT, Alferez DG, Amant F et al (2017) Interrogating open issues in cancer precision medicine with patient-derived xenografts. Nat Rev Cancer 17(4):254–268 9. Malaney P, Nicosia SV, Dave V (2014) One mouse, one patient paradigm: new avatars of personalized cancer therapy. Cancer Lett 344 (1):1–12 10. Clevers H (2016) Modeling development and disease with organoids. Cell 165 (7):1586–1597 11. Miller KD, Siegel RL, Lin CC et al (2016) Cancer treatment and survivorship statistics, 2016. CA Cancer J Clin 66(4):271–289 12. Rahib L, Smith BD, Aizenberg R et al (2014) Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid,

liver, and pancreas cancers in the United States. Cancer Res 74(11):2913–2921 13. Hwang CI, Boj SF, Clevers H et al (2016) Preclinical models of pancreatic ductal adenocarcinoma. J Pathol 238(2):197–204 14. Heinemann V, Reni M, Ychou M et al (2014) Tumour-stroma interactions in pancreatic ductal adenocarcinoma: rationale and current evidence for new therapeutic strategies. Cancer Treat Rev 40(1):118–128 15. Amrutkar M, Gladhaug IP (2017) Pancreatic cancer chemoresistance to gemcitabine. Cancers 9(11):E157 16. Lakiotaki E, Sakellariou S, Evangelou K et al (2016) Vascular and ductal elastotic changes in pancreatic cancer. APMIS 124(3):181–187 17. Olive KP, Jacobetz MA, Davidson CJ et al (2009) Inhibition of Hedgehog signaling enhances delivery of chemotherapy in a mouse model of pancreatic cancer. Science 324 (5933):1457–1461 18. Provenzano PP, Cuevas C, Chang AE et al (2012) Enzymatic targeting of the stroma ablates physical barriers to treatment of pancreatic ductal adenocarcinoma. Cancer Cell 21 (3):418–429 19. Heid I, Steiger K, Trajkovic-Arsic M et al (2017) Co-clinical assessment of tumor cellularity in pancreatic cancer. Clin Cancer Res 23 (6):1461–1470 20. Janes MR, Zhang J, Li LS et al (2018) Targeting KRAS mutant cancers with a covalent G12C-specific inhibitor. Cell 172(3):578–589 e517 21. Hirt UA, Waizenegger IC, Schweifer N et al (2018) Efficacy of the highly selective focal adhesion kinase inhibitor BI 853520 in adenocarcinoma xenograft models is linked to a mesenchymal tumor phenotype. Oncogene 7 (2):21 22. Luca AC, Mersch S, Deenen R et al (2013) Impact of the 3D microenvironment on phenotype, gene expression, and EGFR inhibition

Generation and Analysis of a 3D Co-Culture Spheroid Model of Pancreatic Cancer of colorectal cancer cell lines. PLoS One 8(3): e59689 23. Pampaloni F, Reynaud EG, Stelzer EH (2007) The third dimension bridges the gap between cell culture and live tissue. Nat Rev Mol Cell Biol 8(10):839–845 24. Tsunoda T, Takashima Y, Fujimoto T et al (2010) Three-dimensionally specific inhibition of DNA repair-related genes by activated KRAS in colon crypt model. Neoplasia 12 (5):397–404 25. Zanoni M, Piccinini F, Arienti C et al (2016) 3D tumor spheroid models for in vitro therapeutic screening: a systematic approach to enhance the biological relevance of data obtained. Sci Rep 6:19103 26. Dhurjati R, Krishnan V, Shuman LA et al (2008) Metastatic breast cancer cells colonize and degrade three-dimensional osteoblastic tissue in vitro. Clin Exp Metastasis 25 (7):741–752 27. Price KJ, Tsykin A, Giles KM et al (2012) Matrigel basement membrane matrix influences expression of microRNAs in cancer cell lines. Biochem Biophys Res Commun 427 (2):343–348 28. Roife D, Dai B, Kang Y et al (2016) Ex vivo testing of patient-derived xenografts mirrors the clinical outcome of patients with pancreatic ductal adenocarcinoma. Clin Cancer Res 22 (24):6021–6030 29. Ahonen I, Akerfelt M, Toriseva M et al (2017) A high-content image analysis approach for quantitative measurements of chemosensitivity in patient-derived tumor microtissues. Sci Rep 7(1):6600 30. Francone TD, Landmann RG, Chen CT et al (2007) Novel xenograft model expressing human hepatocyte growth factor shows ligand-dependent growth of c-Met-expressing tumors. Mol Cancer Ther 6(4):1460–1466 31. Finnberg NK, Gokare P, Lev A et al (2017) Application of 3D tumoroid systems to define immune and cytotoxic therapeutic responses based on tumoroid and tissue slice culture molecular signatures. Oncotarget 8 (40):66747–66757 32. Koerfer J, Kallendrusch S, Merz F et al (2016) Organotypic slice cultures of human gastric and esophagogastric junction cancer. Cancer Med 5(7):1444–1453 33. Hirakawa T, Yashiro M, Doi Y et al (2016) Pancreatic fibroblasts stimulate the motility of pancreatic cancer cells through IGF1/IGF1R signaling under hypoxia. PLoS One 11(8): e0159912

179

34. Stockert JC, Horobin RW, Colombo LL et al (2018) Tetrazolium salts and formazan products in Cell Biology: viability assessment, fluorescence imaging, and labeling perspectives. Acta Histochem 120(3):159–167 35. Jacobi N, Seeboeck R, Hofmann E et al (2017) Organotypic three-dimensional cancer cell cultures mirror drug responses in vivo: lessons learned from the inhibition of EGFR signaling. Oncotarget 8(64):107423–107440 36. Murphy KC, Fang SY, Leach JK (2014) Human mesenchymal stem cell spheroids in fibrin hydrogels exhibit improved cell survival and potential for bone healing. Cell Tissue Res 357(1):91–99 37. Magidson V, Khodjakov A (2013) Circumventing photodamage in live-cell microscopy. Methods Cell Biol 114:545–560 38. Renier N, Wu Z, Simon DJ et al (2014) iDISCO: a simple, rapid method to immunolabel large tissue samples for volume imaging. Cell 159(4):896–910 39. Costa EC, Moreira AF, de Melo-Diogo D et al (2018) Polyethylene glycol molecular weight influences the ClearT2 optical clearing method for spheroids imaging by confocal laser scanning microscopy. J Biomed Opt 23(5):1–11 40. Greenbaum A, Jang MJ, Challis C et al (2017) Q&A: How can advances in tissue clearing and optogenetics contribute to our understanding of normal and diseased biology. BMC Biol 15 (1):87 41. Carpenter AE, Jones TR, Lamprecht MR et al (2006) CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol 7(10):R100 42. Verveer PJ, Swoger J, Pampaloni F et al (2007) High-resolution three-dimensional imaging of large specimens with light sheet-based microscopy. Nat Methods 4(4):311–313 43. Lorenzo C, Frongia C, Jorand R et al (2011) Live cell division dynamics monitoring in 3D large spheroid tumor models using light sheet microscopy. Cell Div 6:22 44. Kuen J, Darowski D, Kluge T et al (2017) Pancreatic cancer cell/fibroblast co-culture induces M2 like macrophages that influence therapeutic response in a 3D model. PLoS One 12(7):e0182039 45. Rodenhizer D, Gaude E, Cojocari D et al (2016) A three-dimensional engineered tumour for spatial snapshot analysis of cell metabolism and phenotype in hypoxic gradients. Nat Mater 15(2):227–234

Part IV Animal Models for the Study of Gene Function In Vivo

Chapter 12 In Vivo Pharmacology Models for Cancer Target Research Dawei Chen, Xiaoyu An, Xuesong Ouyang, Jie Cai, Demin Zhou, and Qi-Xiang Li Abstract Experimental animal tumor models have been broadly used to evaluate anticancer drugs in the preclinical setting. They have also been widely applied for drug target discovery and validation, which usually follows four experimental strategies: first, assess the roles of putative drug targets using in vivo tumorigenicity and tumor growth kinetics assays of transplanted tumors, engineered through gain-of-function (GOF) by overexpressing transgene or knock-in (KI) or loss-of-function by gene silencing using knockdown (KD) or knockout (KO) or mutation via mutagenesis procedures; second, similarly genetically engineered mouse models (GEMM), through either germline or somatic cell procedures, are used to test the roles of potential targets in spontaneous tumorigenicity assays; third, patient-derived xenografts (PDXs), which most closely resemble patient genetics and histopathology, are used in tumor inhibition assays for evaluating target-/pathway-specific inhibitors, including large and small molecules, thus assessing the drug target; and fourth, the targets can be assessed in population-based trials, mouse clinical trials (MCT), so that the validation can be generally meaningful as performed in human clinical trials. This chapter outlines the commonly used protocols in cancer drug target research: the first four sections describe four sets of different, specific pharmacology protocols used in the respective cancer modeling stages, with the last section summarizing the common protocols applicable to all four pharmacology modeling steps. Key words Xenograft, PDX, Homograft, GEMM, Tumorigenesis, Tumor growth kinetics, Tumor growth inhibition, Transgene, Knock-in, Knockout, Knockdown

1

Introduction One of the straightforward strategies for assessing a gene target is to overexpress or silence the gene of interest in a human cell line. This can be performed through transgene overexpression or gene knock-in (KI) (gain-of-function, or GOF), by knockdown (KD or silencing)/knockout (KD/KO), or by other means of loss-of-function (LOF), thus impacting tumorigenicity (tumor formation in vivo) or tumor growth kinetics. Methods of cellular genetic

Dawei Chen and Xiaoyu An contributed equally to this work. Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_12, © Springer Science+Business Media, LLC, part of Springer Nature 2019

183

184

Dawei Chen et al.

manipulations including transgene, gene KI, gene KO, mutation introduction (e.g., mutagenesis or gene editing such as CRISPRCas9) [1, 2], and RNAi-mediated gene silencing (e.g., shRNA) [3] have been extensively described elsewhere and are described in more detail in other chapters of this book. Due to the fact that the majority of drugs are inhibitors of a given target, rather than activators, the more relevant experimental approaches to validate a cancer drug target are LOF of the gene target (usually “oncogenes”), instead of transgene expression. It is worth mentioning that when investigating a drug target for in vivo assessment, silencing many oncogenes may be embryonically lethal or lead to developmental defects, therefore hindering the evaluation of such targets in adulthood. Therefore, in many instances, a conditional (induced) KD/KO in vivo becomes necessary (e.g., by tet-on or tet-off system). A brief outline of tumorigenicity and tumor growth kinetics studies are summarized as follows: the engineered tumor cell lines are implanted subcutaneously (SC) or orthotopically into immunocompromised mice (e.g., nude, NOD/SCID, NSG®, NOG®) [4]. Tumor growth is monitored by cage-side observation and by measuring tumor volume (TV) using calipers and/or imaging methods. An experiment management software such as Studylog® can be utilized to directly measure and store the TV and body weight (BW) data. Tumor growth kinetics are determined by the TV and duration of the study. A negative impact on tumorigenicity and growth kinetics means that the targets are “oncogenes” and that they can be further investigated as potential drug targets (Fig. 1). When utilizing inducible expression or silencing vectors [5], the engineered cells are implanted as described above, followed by subsequently randomly grouping into treatment and control groups and subjecting to induction at desired time points in vivo. In many situations, the control vector-engineered cells are processed in parallel as additional controls. Tumor growth is then monitored and compared among the groups. Recently, libraries of shRNA using an RNAi approach or sgRNA using a CRISPR-Cas9 approach of genome-wide gene target validation have been introduced into cell populations [6–8], so that genes which negatively impact tumorigenicity/ growth kinetics could be identified as potential drug oncogene targets. It is worth mentioning that negative selection/identification is usually much more challenging than positive selection [9–12]. The cell lines used for engineering can be human or mouse cell lines. Engineered human cell lines can be xenografted into immunocompromised mice, while engineered mouse cell lines can be homografted into syngeneic mouse strains, most of which are immunocompetent, therefore enabling certain immuno-oncology (I/O) targeting.

Protocols for Cancer Target Pharmacology

185

Fig. 1 Different tumor animal modeling. A: Genetically engineered tumor cell culture
induction. B: SC inoculation, followed by induction if induced system is involved or
pharmacological treatment. TV is measured by caliper. C: Orthotopic implantation (e.g., liver, spleen, mammary fat pad) or intravenously. Tumor burden is measured by imaging or autopsy

For over three decades, genetically engineered mouse models (GEMMs), based on germline or somatic cell gene alterations of key oncogenic/tumor-suppressor pathways, have enabled numerous insightful mechanistic findings on tumor onset, progression, and metastasis and also enabled evaluation of potential drug targets [4, 13]. LOF of tumor suppressors or GOF of oncogenes are introduced to produce GEMMs [14] where conditional expressing/silencing is achieved in a constitutive manner or tissue-specific/inducible regulation, resulting in spontaneous tumor development. Utilizing CRISPR/Cas9 technology for direct in vivo targeting, or targeting embryonic stem cells, has greatly improved the efficiency of GEMM creation [15]. In particular, multiplex Cas9-mediated genome editing [16, 17] enables simultaneous modeling of a multigene tumorigenesis process to recapitulate the complex combinations of genetic lesions in patients. Compared to transplanted tumors, where fully developed tumors are implanted into a naı¨ve host, tumorigenesis in GEMMs represents de novo tumor onset and progression, mimicking tumorigenesis in patients and accompanied by escape from immune surveillance. Tumorigenesis in GEMMs can be used to evaluate potential drug targets according to the introduced genetic alterations in mice, similarly as described above for transplanted tumors [18, 19]. Gross-necropsy or imaging methods (e.g., micro MRI) are usually required to monitor tumorigenesis [20–22]. GEMM cancer models have an advantage over xenografts in that mouse

186

Dawei Chen et al.

immunity is intact, and therefore GEMM models are suitable for evaluating I/O targets [4]. These targets are only mouse surrogates of human targets, and therefore caution is needed for I/O target evaluation as mouse tumor/immunity may differ from human patients [23, 24]. On a separate note, genetic engineering techniques are also used to create chimeric GEMMs, which KI human targets, e.g., immune checkpoint genes, such as CTLA-4 or PD-1. This partially humanized GEMM, which we have named “HuGEMM™”, can have mouse tumor homografts from syngeneic cell lines transplanted for facilitating evaluation of human I/O targets and therapeutics [4]. The study design is similar to the efficacy studies described in Subheading 4. When a highly specific small molecule inhibitor or antibody is available against a potential gene target, it can directly be used to evaluate targeting by treating xenograft and homograft models. Patient-derived xenografts (PDX) are patient tumors grown in immunocompromised mice [4] which are known to maintain the patient tumor histopathology and molecular pathology [25, 26]. PDX are considered to be the model most predictive of the original patient response to treatment. PDX may not be readily used for I/O target evaluation due to its immunocompromised environment [4], but are particularly relevant for common target research. Selecting adequate PDX models per genetic characteristics, e.g., specific mutations or expression levels of one or a set of genes, is critical to testing hypotheses on potential drug targets [27–30]. The negative impact on tumor growth kinetics caused by specific agents can suggest that these genetic alternations might be oncogenic drivers or potential drug targets. Blood cancer xenografts are the “liquid version” of PDX. Patient leukemia cancer cells are systemically inoculated into immunocompromised mice, resulting in the development of leukemia [30], where leukemic tumor load is usually determined by measuring human CD45+ cells in peripheral blood using flow cytometry at various time points and in bone marrow, spleen, or other infiltrated organs at the terminal time point. Homografts of spontaneous or induced murine primary mouse tumors (the “mouse version of PDX”) [4], or simply syngeneic cell lines that grow in syngeneic immunocompetent mice, can be developed to evaluate mouse surrogate I/O therapeutics and therefore also their targets. Similar to PDX models, GEMM tumor-derived homografts recapitulate their original mouse disease significantly better than traditional, in vitro immortalized cell line-derived syngeneic models. This is observed for PDX vs. cell line-derived xenografts: maintaining heterogeneous histopathology as a typical cancer stem cell-driven disease [13]. Homografts derived from GEMM tumors with specific driver mutations are also a useful tool to assess potential targetability. The derived models also have

Protocols for Cancer Target Pharmacology

187

Fig. 2 PDX and mouse homograft tumor modeling. PDX or mouse primary tumors (e.g., spontaneous tumors from GEMM) are collected and cut into pieces of 2–3 mm in diameter. Chunks of these fragments are transplanted into mice using a trocar. Tumor growth is monitored and examined. Tumors are harvested when their volume reaches approximately 500–1200 mm3 for downstream processes, including efficacy studies, cryopreservation banking, histopathology, and molecular pathology characterization

superiority and simplicity in operational handling as compared to the original GEMM tumors — due to their short duration and synchronized tumor development. The methods and usage of allograft models are rather similar to PDX (as described above), including profiling, characterization, etc. The process for PDX or homograft modeling is summarized in Fig. 2. Alternatively, syngeneic cell lines derived from mouse tumor models are the most commonly explored models for preclinical I/O investigation. Most checkpoint inhibitors proof of concept (POC) was first confirmed in syngeneic models, e.g., PD-1 antibodies using MC38 tumors. Now, many syngeneic models have been extensively profiled genomically, immunologically, and for I/O agent efficacy by various laboratories. Furthermore, syngeneic tumor cell lines or mouse tumor homografts can be inoculated into humanized chimeric GEMM mouse to evaluate human-specific therapeutics directly [4]. This chapter provides some basic protocols that have been used in cancer pharmacology for animal models and which are applicable for cancer target research. These protocols intend to provide an elementary scope of the commonly used procedures for readers new to the field but are by no means comprehensive and detailed. Readers will need to generate more detailed protocols based around their own specific research objectives prior to conducting their investigations. See Table 1 for frequently used abbreviations.

188

Dawei Chen et al.

Table 1 Frequently used abbreviations Abbreviation

Definition

AFP

Alpha-fetoprotein

AUC

Area under curve

BME

Basement membrane extract

BW

Body weight

CRISPR

Clustered regularly interspaced short palindromic repeats

CV%

Standard deviation of TV/average TV

DOX

Doxycycline

ECM

Extracellular matrix

FFPE

Formalin-fixed paraffin-embedded

GEMM

Genetically engineered mouse model

GFP

Green fluorescence protein

GOF

Gain of function

IF

Immunofluorescence

IHC

Immunohistochemistry

I/O

Immuno-oncology

IP

Intraperitoneal(ly)

IV

Intravenous(ly)

IVC

Individual ventilated cages

IVIS

In vivo imaging system

KD/KO

Knockdown/knockout

KI

Knock-in

LOF

Loss of function

MCT

Mouse clinical trials

MRI

Magnetic resonance imaging

MTVR

Maximum tumor volume reduction

PD

Pharmacodynamic

PDX

Patient-derived xenograft

PO

Oral(ly)

POC

Proof of concept

QC

Quality control

RNAseq

RNA sequencing or transcriptome sequencing (continued)

Protocols for Cancer Target Pharmacology

189

Table 1 (continued) Abbreviation

Definition

SC

Subcutaneous(ly)

sgRNA

Single guide RNA

shRNA

Short hairpin RNA

SOC

Standard of care drug

SPF

Specific pathogen-free

STR

Short tandem repeats

TCGA

The Cancer Genome Atlas

TV

Tumor volume

2 Assay Tumorigenicity and Growth Kinetics of Transplanted Tumors Derived from GEMM Tumors or Engineered Cell Lines of a Gene(s) of Interest 2.1

Materials

2.1.1 Project-Specific Materials

The common and general materials and reagents are described in Subheading 6. Doxycycline, Sigma, MO, USA. Pancreatin, Sigma, MO, USA. Trypsin. DMEM. BioCoat™ Collagen 25 cm2 rectangular canted neck cell culture flask with vented cap.

2.1.2 Animals

Immunocompromised mice: NOG/NSG, NOD-SCID, B/C nude, or nu/nu mice. Immunocompetent mice: C57BL/6, Balb/c, FVB/N, and other strains. Mice are purchased from local suppliers in the major markets.

2.1.3 Cell Lines

2.2

Methods

2.2.1 Study Design

The cell lines used for engineering can be human cancer cell lines and mouse tumor cell lines, many of the cell lines can be obtained from the ATCC and other depositories. 1. Inducible vectors are transferred into the tumor cell line. 2. Inoculate the engineered cell line and vector control cell line into mice for tumor development.

190

Dawei Chen et al.

3. Monitor the tumor growth for tumorigenicity after inoculation, or after induction of the gene KD/KO, by cage-side observation. 4. After tumors become palpable, begin TV measurement. 5. Compare the difference in tumor growth kinetics between engineered and control cell lines. 2.2.2 Culture and Prepare Tumor Cells

The genetically engineered cancer cell lines are expanded in vitro via tissue culture prior to inoculation. As the cell number per injection site varies greatly between different injection methods or cell types (normally around 104 to 106 cells per site), the scale of culture will be considered accordingly. 1. Tumor cell lines are grown in complete medium and maintained as monolayer cultures. 2. Once the cell lines are growing in the exponential growth phase and reach 70% confluence, they are harvested for inoculation. 3. Remove the medium and rinse the cells briefly with PBS. 4. Digest the cells with 0.25% trypsin until they detach from the flask, shake the flask and stop trypsinization by adding FBS-containing medium, and then slowly pipette to resuspend cells. 5. Spin down the cells at 4  C by centrifugation at 1000  g for 5 min. 6. Resuspend the cell pellet in cold PBS by slowly pipetting the cell pellet using a 1 mL pipette. 7. Count the cells with a particle counter or hemocytometer, and assess the cell viability using trypan blue staining. Cell viability should be >95% to ensure successful inoculation. 8. The cell suspension is re-pelleted via centrifugation, followed by resuspending in serum-free medium to obtain the cell suspension in the required concentration. The volume of each inoculation should be around 100–200 μL.

2.2.3 Model Establishment

2.2.4 Induction of Gene Silencing or Transgene Expression

General methods for the establishment of transplant tumors will be described in Subheading 7 General Protocols, including animal housing, tumor cell inoculation, and BW/TV monitoring. 1. DOX (doxycycline) will be given via drinking water that contains 5% sucrose [5] or chow. 2. The treatment is initiated on the day the cells are inoculated [5, 31] or when tumor volume reaches a predefined size (e.g., ~100 mm3) [32], thus allowing assessment of the staged tumor response to treatment.

Protocols for Cancer Target Pharmacology

191

3. The animals are monitored for tumor growth and overall health during the induction period or until study endpoint. 4. Tumor-bearing mice are euthanized, and tumors are excised and weighed before sampling. 2.2.5 Efficacy Readout

3

The difference in growth kinetics is calculated based on the TV measurement, which is described in Subheading 7.

Assay Spontaneous Tumorigenesis of Mice with Target Gene KO/KI

3.1

Materials

The common and general materials and reagents are described in Subheading 7. GEMM and immunocompetent mouse strains can be obtained from various vendors in the different markets.

3.2

Methods

Spontaneously tumorigenic GEMMs develop tumors; however, this usually takes a long period of time, which varies from model to model (general range, 2–6 months). This leads to the requirement of a large number of study animals to ensure statistically significant results. The mice are randomized into different treatment groups per BW, followed by tumor formation monitored via clinical observation, necropsy, and/or MRI.

3.2.1 Study Design

3.2.2 Clinical Observation

GEMM mice spontaneously developing tumors, which may also be accompanied by certain clinical signs, which need to be closely monitored (daily). The method of clinical observation is described in Subheading 7.

3.2.3 Mouse MRI

In contrast to superficially implanted tumors (e.g., SC-transplanted) that can be monitored by standard caliper measurement, autochthonous tumor growth in GEMMs must be monitored by longitudinal imaging strategies, e.g., MRI. The MRI method is described in Subheading 7.

3.2.4 Necropsy

In many cases, tumor-bearing mice with certain clinical symptoms are sacrificed for necropsy. Tumors and organ tissues are collected as formalin-fixed paraffin-embedded blocks (FFPE) or snap frozen samples for pathology and genetic analysis. GEMM can have organ preference for tumor development. For instance, KRAS-G12D GEMM result in spontaneous tumorigenesis in the intestines or the lung [33]. Therefore, those involved organs are usually examined for gross and microscopic pathology. The necropsy and pathology evaluation is normally only focused on targeted organs. Necropsy and sample collection methods are described in Subheading 7.

192

Dawei Chen et al.

4 Assay Tumor Growth Kinetic Changes Caused by Target-Specific Agents: Small Molecular Inhibitors or Antibodies 4.1

Materials

4.1.1 Specific Reagents

1. RPMI-1640 with 2 antibiotic and antimycotic. 2. RPMI-1640. 3. Ficoll-Paque™ PREMIUM. 4. FITC antihuman CD45 antibody BioLegend®, CA, USA. 5. Red blood cell lysis buffer (generic). 6. EDTA routine blood tube 2 mL. 7. TaKaRa Taq™ (Taq DNA polymerase) Takara Bio. 8. dNTPs (generic).

4.1.2 Animals

Immunocompromised animals for xenografts: NOD-SCID, B/C nude, or nu/nu mice.

NOG/NSG,

Immunocompetent animals for homografts: C57/B, chimeric GEMM mouse. 4.2

Methods

4.2.1 Study Design

1. Usually 5–10 mice per arm are used in treatment studies. 2. Model selection can be based on the cancer type, pathology diagnosis, genetic profile, growth features, and/or standard of care treatment (SOC) efficacy. 3. Dosing is similar to that described for other models above. 4. For blood cancer, the tumor growth kinetics are determined by human CD45+ levels in blood.

4.2.2 Model Establishment Solid Tumor PDX

1. The patient tumor is collected by surgery or biopsy. 2. The collected patient tumor is kept in a sterile 50 mL tube containing 30 mL 4  C transfer media immediately postsurgery and transferred to an animal facility within 6 h. 3. Tumors are cut into chunks of 2 mm in diameter. 4. Inoculate the tumor chunk on the right flank site of immunocompromised mice (further details found within Subheading 7). The mice are observed each week for tumor development. 5. No palpable tumor developed within 120 days is considered unsuccessful engraftment. 6. Any developed tumor will be serially transplanted into new immunocompromised mice and preserved for banking and characterization.

Leukemia PDX

1. The patient bone marrow sample is collected into a sterilized 50 mL tube containing 4  C 30 mL transfer media and transferred to an animal facility within 6 h.

Protocols for Cancer Target Pharmacology

193

2. Mononuclear cells are isolated by Ficoll-Plaque per product instructions. 3. The isolated cells are IV injected into immunocompromised mice (NOG/NSG), per the method described in Subheading 7. 4. BW of mice is measured weekly. 5. The mice are sacrificed when CD45+ cells are >70%, and tumor (observed in some models), bone marrow, and spleen are collected. CD45+ cells are measured by flow cytometry; a detailed method is described in Subheading 7. 6. Tumor, bone marrow, and spleen are harvested at termination and digested to prepare single-cell suspensions. The spleen samples also need to be treated with the red blood cell lysis buffer. 7. The collected cells can be used for characterization, serially passed through engraftment, or cryopreserved. Mouse Tumor Homograft

1. GEMM mice are bred for 4–6 months to allow spontaneous tumor development. 2. Mice are observed daily for clinical symptoms of tumor development; the mouse is sacrificed for necropsy when such clinical symptoms become obvious. 3. The primary organ where the spontaneous tumor was expected to arise is collected and checked visually. 4. Tumor nodules are collected for transplantation, following the same protocol as PDX model establishment. The inoculation method is described in Subheading 7. 5. Part of the organ containing the spontaneous tumor is collected and embedded as FFPE for pathology examination.

Model Quality Control (QC)

The consistency of PDX, patient-derived blood cancer, and homograft models with their original tumors (patient or mouse) is critical. Thus QC of these models needs to be applied before their use. Short tandem repeat (STR) genotype genetic fingerprints, molecular pathology, and histopathology methods are often used for model QC.

STR Genotype Genetic Fingerprint QC of PDX Models

STR are regions with short repeat units (usually 2–6 base pairs in length), the number of which are highly variable among individual people. STR are therefore very useful in identifying the models in banks, different passages, and different studies, thereby reducing the risk of error. The simple protocol that we used is as follows: 1. 50 mg tumor tissue is collected for DNA extraction, as described before.

194

Dawei Chen et al.

Table 2 STR QC PCR reaction system Reagent

Volume (μL)

Final concentration

10 PCR buffer

2.5

1

10 M forward primer mixture

0.5

0.2 M

10 M reverse primer mixture

0.5

0.2 M

2.5 mM dNTPs

2

0.2 mM

TaqE (5 U/μL)

0.5

0.1 U

Template (DNA)

1

50–100 ng

Sterilizing water

19

Total

25

2. Primer pairs of eight chosen STR loci, D12S217, D7S820, TPOX, FGA, CSF1PO, D16S539, tyrosine hyderoxylase (TH01), and VWA, are prepared into three sets of mixtures per size for PCR reaction. 3. The PCR reaction system is detailed in Table 2. 4. Cycling program (steps detailed as below): Step 1: 95  C for 11 min. Step 2: 96  C for 1 min. Step 3: 94  C (temperature ramp rate 100%) for 30 s. Step 4: 60  C (temperature ramp rate 29%) for 30 s. Step 5: 70  C (temperature ramp rate 23%) for 45 s. Step 6: Go to Step 3, nine times. Step 7: 90  C (temperature ramp rate 100%) for 30 s. Step 8: 60  C (temperature ramp rate 29%) for 30 s. Step 9: 70  C (temperature ramp rate 23%) for 45 s. Step 10: Go to Step 7, 19 times. Step 11: 60  C for 30 min. 5. The PCR product from the original patient tumor is used as a control template. 6. PCR products of test tumors are compared alongside the control templates using 10% non-denaturing PAGE electrophoresis. 7. Mismatch with the patient tissue suggests mistakes in the model. Sometimes weak signaling might suggest high mouse contents.

Protocols for Cancer Target Pharmacology

195

Histology Pathology QC

Model histopathology is usually performed for each PDX and compared with the hospital-derived pathology diagnosis of the original patient, as another important QC procedure. The method of this QC is a routine pathology process: a model FFPE slide is prepared and H&E stained, followed by histopathology examination performed by pathologists. Additional immunohistochemistry (IHC) of different tissue markers is also needed to confirm disease type on a case-by-case basis.

Molecular Pathology QC

To further confirm the accuracy of PDXs’ pathology obtained from hospitals, we have recently developed a new molecular pathology algorithm for PDX diagnosis, which is able to accurately diagnose the pathology of PDX tumors using transcriptome sequences [26]. The RNAseq profiling data will be used for this QC process.

4.2.3 Tumor Implantation Methods

Transplanted tumors are the most commonly used animal models in cancer pharmacology, including target research. There are several common practices utilized in the laboratories per usages. A detail protocol is described in Subheading 7.

4.2.4 Grouping Methods

TV is the most commonly used parameter for randomization during grouping. Biomarkers whose levels are correlated to TV can sometimes also be used as a parameter for randomization, when tumor volume cannot be readily measured (e.g., AFP in orthotopic HCC models).

Randomized Enrollment of Study Mice

1. When there are sufficient mice with TV of approximately 100–200 mm3, the mice with close TV and BW value are selected and randomly assigned to respective treatment groups. 2. CV% of TV in each group should be less than 40%. 3. Treatment is initiated immediately after grouping. Each model should follow the same dosing regimen. 4. For blood cancer models, mice where the % hCD45+ cells reach 2–10% are enrolled for randomized grouping. Clinical Trial-Style Enrollment

Another grouping process mimics the enrollment method utilized in clinics, often used when there are insufficient animals within the close range of grouping parameters for simultaneous enrollment and synchronized dosing. 1. TV and BW of tumor-bearing mice are measured weekly. 2. Mice that reach the enrollment criteria (e.g., TV approximately 150–200 mm3, BW >20 g) are enrolled and randomized into treatment groups. 3. The treatment of an individual will be initiated immediately following enrollment, and the enrollment will continue until

196

Dawei Chen et al.

sufficient numbers of each group are enrolled at the set parameter. 4. All individual animals will follow respective dosing regimens, which are not synchronized, according to the protocol. 4.2.5 Tumor Growth Kinetics Readout

Tumor growth kinetics based on timely readouts of TV or tumor weight (TW) are the most common measurement reflective of tumor growth or tumor growth inhibition (efficacy). The detailed protocol is described in Subheading 7.

4.2.6 Pharmacodynamic (PD) Readouts, Beyond Tumor Response

Pharmacodynamic (PD) readouts are usually required to evaluate the inhibition of drug targets and to investigate the mechanism of antitumor activity via a putative target. A common PD analysis includes dephosphorylation of kinases. Particularly for tyrosine kinase inhibitors (TKIs), samples need to be harvested at 2, 6, 16 h posttreatment via Western blot and/or IHC. mRNA expression level changes are also evaluated by RNAseq at desirable time points. The sample collection method is described in Subheading 7.

4.2.7 I/O PD Readout

For I/O therapeutic target assessment, tumor baseline immunophenotype and/or PD change of immunophenotype resulting from I/O treatment is key to studying I/O targeting. The most common approach is to measure tumor-infiltrating immunological markers, reflective of common and key immune cells, e.g., different subtypes of T cells, macrophages, NK cells, etc. The common methods of these analyses are multicolor flow cytometry (FACS) [4]. Detailed methods are described in Subheading 7.

5 Evaluate Targets in Population-Based Mouse Clinical Trials (MCT) Using Specific Agents Used as avatars of cancer patients, PDXs provide the ideal experimental platform for clinical trial-like population-based investigations, also called mouse clinical trials (MCT) to evaluate drugs, but also for the discovery of new targets (hypothesis generation) or target validation (hypothesis testing) [28, 29, 34–36]. The principle of an MCT project is shown in Fig. 3. This type of trial has recently been enabled by the establishment of large PDX libraries of diverse cancer types, with full genomic annotations [26] based on characterization using some common assay types summarized in Table 3. The design of population-based studies and their data analysis have been previously described [36]. The protocol for running each mouse in an MCT is similar to that for single PDX model studies.

Protocols for Cancer Target Pharmacology

197

Fig. 3 Schematic illustration of mouse clinical trial (MCT). A large cohort of PDXs reflective of the heterogeneity of the patient population can be enrolled into a mouse clinical trial. The trial results can be potentially classified into two groups, responders and non-responders. The model efficacy data and genomic information are used to identify predictive biomarkers using machine learning methods

Table 3 Assays and applications commonly used in target research Assays Transcriptome sequencing (RNAseq)

Applications l l l

Whole exome sequencing (WES)

l

Real-time PCR

l

l

l l l

Immunohistochemistry (IHC)

l l l

Tissue microarray (TMA) Flow cytometry

Genome-wide expression profile. Mutation data at transcript level. Gene fusion. Exonic mutation data. Gene copy number. Target expression determination. GCN determination. SNP genotyping. Mutation allele frequency analysis. Protein expression. Phosphorylation. Microscopic localization.

An array of tumor tissues on the same chip, for biomarker screening l

l l

Hematological or immune cell phenotyping: surface/intracellular markers. Receptor density quantification. PD readouts: proliferation, apoptosis, cell cycle, differentiation.

198

6

Dawei Chen et al.

Preclinical Cancer Pharmacology General Protocols There are many common procedures that can be applied to different small animal modeling studies of cancer targeting. To avoid redundant description in each pharmacology section, here we summarize the common cancer pharmacology protocols that are used in the specific cancer pharmacology sections above.

6.1

Materials

6.1.1 Equipment

1. IVC system. 2. Biohazard hood. 3. Caliper, Sylvac S-Cal pro. 4. Portable liquid nitrogen tank. 5. IVIS bioluminescent imaging system. 6. Flow cytometer LSRFortessa X-20, BD, NJ, USA. 7. Centrifuge, Thermo ST16R. 8. Cell counter, Nexcelom Bioscience Cellometer® Auto T4. 9. gentleMACS™ Octo Dissociator, Miltenyi Biotec, Germany. 10. Studylog® Studylog Systems, Inc. CA, USA.

6.1.2 Consumables

1. 70 μm cell filters, one for each tissue. 2. Miltenyi C-tubes (one for each tumor). 3. Miltenyi gentleMACS with heater blocks (one for every eight tumors), Miltenyi Biotec, Germany. 4. EDTA routine blood tube 2 mL. 5. Trocar 20# (inner diameter 2 mm).

6.1.3 Specific Regents

1. Ficoll-Paque PREMIUM. 2. FITC antihuman CD45 antibody BioLegend, CA. 3. Cultrex® High Protein Concentration (HC20þ) BME, PathClear Trevigen MD or BD Matrigel™ Basement Membrane Matrix High concentration, BD Bioscience NY. 4. Red blood cell lysis buffer. 5. Tumor Dissociation Kit. 6. Phosphate-buffered saline (PBS) 50 mL. 7. RPMI-1640. 8. Iodophor swabs. 9. 10% formalin. 10. RNAlater® Thermo Fisher Scientific. 11. Murine tumor dissociation kit, Miltenyi Biotec, Germany. 12. Flow cytometry buffers:

Protocols for Cancer Target Pharmacology

199

Mincing buffer (for tissues such as tumor-draining lymph node or spleen). l l

5000 U/mL (10) DNAse I, Roche in RPMI-1640. Dilute immediately before use; store at 4  C for up to 1 month.

FACS wash buffer. l

10% FBS; azide-free

l

40 mM EDTA 0.5 M pH 7.4 stock

l

Sterile Ca þ and Mg þ free PBS (1, stock).

13. Brilliant staining buffer, BD, NJ. 14. Purified rat anti-mouse CD16/CD32, mouse BD Fc Block™ BD, NJ. 15. Foxp3 Fix/Perm kit, eBioscience Thermo Fisher Scientific. 16. PBS sterile, Hyclone. 17. UltraComp eBeads, eBioscience, Thermo Fisher Scientific.

7

Regulatory Compliance in Animal Experimentation All of the protocols and amendment(s) or procedures involving the care and use of animals need to be reviewed and approved by the local Institutional Animal Care and Use Committee (IACUC) prior to the conduct of studies. The care and use of animals will be conducted in accordance with AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care) International guidelines as reported in the Guide for the Care and Use of Laboratory Animals, National Research Council (2011). All animal experimental procedures will be under sterile conditions at SPF (specific pathogen-free) facilities and conducted in strict accordance with the Guide for the Care and Use of Laboratory Animals from different government institutions (e.g., the National Institutes of Health). The protocol will need to be approved by the Committee on the Ethics of Animal Experiments at the facility institution (e.g., institutional IACUC).

7.1

Methods

7.1.1 Animal Housing

All animal experiments are conducted in SPF facilities. 1. Mice are housed in individual ventilated cages. 2. Temperature: 20–26  C; humidity 30–70%; lighting cycle: 12-h light and 12-h dark. 3. Corncob bedding is used and changed weekly. 4. Diet: irradiation sterilized dry granule food during the entire study period. 5. Water: animals have free access to sterile drinking water.

200

Dawei Chen et al.

7.1.2 Tumor Inoculation Tumor Cell SC Inoculation

1. Mix the cell suspension with an equal volume of ECM (extracellular matrix, e.g., Matrigel® which can sometimes significantly increase tumor take rate [37]), and keep on ice during transportation to the vivarium. 2. Draw ECM-cell mixture slowly into a chilled 1 mL syringe, and keep the filled syringes on ice to avoid ECM solidification. 3. Immobilize the mouse, and inject 100–200 μL of ECM-cell mixture subcutaneously to the right flank of the recipient mouse; hold syringes for >3 s to allow complete ECM solidification. 4. Normally 5–10 mice are inoculated per group for tumor growth monitoring. When using tet-inducible cells, animals are randomly enrolled into two groups of 5–10 mice each and treated
induction agents (e.g., DOX). 5. Mice are monitored 24-h postinoculation for procedurerelated abnormalities.

Tumor Cell IV/IP Inoculation

For non-superficial implantation, e.g., IP and IV inoculation, tumors may not be readily and macroscopically visible. Therefore, IVIS or other imaging approaches are required to monitor tumorigenesis. Some of these imaging procedures require the engineered cell lines to have reporter gene expression (e.g., luciferase or GFP). 1. Luciferase- or GFP-engineered cells are suspended in serumfree DMEM and transported to the vivarium. 2. Gently mix the cells and then draw the mixture slowly into a 1 mL syringe. For IV injection, go to steps 3 and 4; for IP injection go to steps 5 and 6. 3. For IV injections: first immobilize the mouse, and then inject cells into the tail vein. 4. Gently constrict the injection point for 3–5 s to prevent bleeding or leakage. Move to step 7. 5. For IP injections: inject the cell suspension into the abdominal cavity of immobilized mice. 6. Rotate the needle 90 and retract to prevent leakage. 7. Check the mice 24-h postinjection for any procedure-related abnormalities.

Tumor Chunk Inoculation

1. Monitor BW and TV of tumor-bearing donor mice. When TV reaches 500–1200 mm3, the animal is euthanized in a biohazard hood as per protocol (shave the mouse around the tumor if necessary, e.g., when NOD-SCID mice are used). 2. Sterilize the skin around the tumor using iodophor swabs. 3. Surgically remove the tumor and place in a petri dish containing 20 mL PBS.

Protocols for Cancer Target Pharmacology

201

4. Cut the tumor in half, removing any extra skin, vessels, calcification, and/or necrosis. 5. If there is contaminating blood, transfer the tumor into another petri dish, and wash the tumor with PBS. 6. Cut approximately 50 mg of tumor tissue, and place it into a 2 mL cryopreserved tube. Snap freeze using liquid nitrogen for QC. 7. Put the rest of the tumor into a sterile 50 mL centrifuge tube, add 20 mL transport medium, and then transport the tube to a separate animal room for pharmacology studies. 8. Cut the tumors into 2 mm diameter pieces using a scalpel, putting 1 chunk into each trocar. 9. Sterilize the recipient mouse using iodophor swabs. 10. Inoculate the tumor chunk on the right flank site of the mouse. The mice will be kept for weekly observations of tumor development. Blood Cancer Model Inoculation and Dissociation of Solid Tumor to Single-Cell Suspension

1. Tumors, bone marrow, or the spleen of donor mice are collected and put in digestion media in one well of a sterile 6-well plate. 2. Hold the tissues in place with sterile tweezers/forceps, and slice them into small pieces (~1 mm3) with a scalpel. 3. Place tissue pieces into a C-tube, and use the remaining digestion buffer to wash the plate. Transfer the fluid into the C-tube, which is placed on ice until digestion. 4. Digest tumors using the desired gentleMACS program, and turn on the gentleMACS Octo Dissociator with Heaters. 5. Attach the tumor dissociation C-tubes upside down to the sleeves of the free tube positions, and alter the status of the tube positions from “free” to “selected”. 6. Select a dissociation program (37_c_m_TDK_1) [38], and press the Folder icon to select the required folder. The list of gentleMACS programs in the respective folder will be displayed. 7. After termination of the program, detach the C-tubes from the gentleMACS Dissociator, and perform a short spin down at 300 g to collect the sample at the bottom of the tubes. 8. Resuspend the samples, and apply the cell suspension to a cell strainer placed above a 50 mL tube. 9. Spleen cells need to be subjected to red cell lysis using the RBC lysis solution. 10. Wash the cells through the cell strainer with 10 mL wash buffer to obtain a single-cell suspension.

202

Dawei Chen et al.

11. Centrifuge the tubes at 300  g for 5 min, discard the supernatant, and resuspend the cells with 5 mL wash buffer. 12. Tumor cells are then inoculated IV following the instructions in the tumor cell IV/IP inoculation procedure as described above. 7.1.3 Tumor-Bearing Mice Health Monitoring

The health of the recipient mice is monitored daily. 1. Check the water and food consumption. 2. Check the mouse appearance for an ungroomed hair coat, lumps, thinness, abnormal breathing, and ascites. 3. Palpate the abdomen to check if there are spontaneous tumors on the liver or spleen. 4. Weigh mice weekly using a balance. 5. If any of the following clinical signs are observed, the mice are sacrificed for sample collection and necropsy: BW loss >20%; impaired mobility (not able to eat or drink); unable to move normally due to significant ascites and enlarged abdomen; effort respiration; and dying.

7.1.4 Tumor Burden Determination

1. The length and width of the tumor are measured by calipers after tumors become palpable.

SC Tumor TV Measurement

2. Immobilize the mouse; measure the longest diameter and the diameter perpendicular to the longest diameter using calipers. 3. TV is calculated using the formula TV ¼ a  b 2  6π , where a is the longest diameter of the tumor and b is the shorter diameter of the tumor. 4. Data are automatically analyzed and stored on the Studylog® database.

IVIS Bioluminescent Imaging System (Firefly Luciferase as an Example)

1. Mice are anesthetized using 5% isoflurane mixed with oxygen and their skin sterilized with 70% ethanol. 2. 150 mg/kg of luciferin is IP injected. 3. Place the mice on a disinfected imaging chamber of a commercial imaging system. 4. Maintain anesthesia using 1% isoflurane mixed with oxygen through a nose cone. 5. Take the first image by exposing for 5 min and check the signal intensity. Adjust the exposure time to optimal signal versus background.

Mouse MRI

1. Restrain animal, and induce anesthesia with isoflurane at 5%. Maintain anesthesia by nose cone at 1%.

Protocols for Cancer Target Pharmacology

203

2. Set up animal on animal holder, and place the animal in the correct position on the MRI. 3. Scan animals. 4. Remove the animals and stop the isoflurane, and then return the animals to cage. Blood Cancer Model Tumor Burden Determination

1. In blood cancer models, % human CD45+ (hCD45+) cells are measured to reflect tumor growth status. 2. Whole blood (50–100 μL) is collected by facial vein bleeding weekly in a BD EDTA routine blood tube, followed by lysis using red blood cell lysis buffer. 3. The mononuclear cells are stained by fluorescence-labeled huCD45+ antibody and analyzed using flow cytometry (for the flow cytometry method, see Subheading 7.1.7 within this section). % huCD45+ cells to total cells are calculated. 4. At the termination of the experiment, bone marrow and spleen are examined for huCD45+ ratio. For the cell digestion and flow cytometry methods, see Subheading 7.1.7 in this section.

Tumor Growth Kinetics

Tumor growth kinetics or its inhibition are the key readouts in preclinical tumor pharmacology. A frequently used analysis procedure is as follows. 1. The tumor growth curve is generated by entering TV data into a software tool; a typical tumor growth curve is shown in Fig. 4. 2. ΔT/ΔC% and tumor growth inhibition (TGI) are the most commonly used readouts/endpoints of preclinical cancer pharmacology efficacy assessment. In ΔT/ΔC%, T and C are the mean tumor volume of the treated and control groups, respectively, on a given day. TGI% ¼ 1ΔT/ΔC%. 3. We recently developed a new efficacy calculation endpoint called the median AUC ratio of a growth curve (Guo et al., unpublished): the median AUC ratio is calculated in three steps. First, the normalized AUC of each animal in the vehicle and treatment groups are calculated; second, each mouse in the treatment group is paired with any other mouse in the vehicle group, and the AUC ratio of each pair is calculated; third, the median of all AUC ratios is obtained. There are two advantages of the median AUC ratio: (1) It reflects the treatment efficacy over a whole period of treatment, whereas in the calculation of ΔT/ΔC% or TGI%, only the tumor volume on the selected day was considered; and (2) it is based on the median which is not skewed as much by the extremely large or small values that are frequently seen in PDX or I/O efficacy studies due to the tumor heterogeneity on growth rate and mouse immune status.

204

Dawei Chen et al.

Fig. 4 Typical tumor growth kinetic curves with and without treatment

7.1.5 Necropsy

At the termination of studies, mice are sacrificed for necropsy. Tumors and organ tissues are collected as FFPE or snap frozen tissue for pathology and genetic analysis. 1. Animal is anesthetized with 5% isoflurane and maintained by nose cone at 1%. 2. Visual and palpation inspection occurs for the presence of palpable tumors. 3. Animals are euthanized per approved protocol. 4. Open the abdomen and visually examine target organs for tumors. 5. Harvest tumor or target organ samples.

7.1.6 Sampling

Tumors or organs are collected per various purposes. Sample collection should follow the proper sequence: FFPE > snap frozen tissue for DNA/RNA (RNAlater) > tumor tissue for transplantation or cell suspension for other analysis, e.g., flow cytometry. 1. Animal is anesthetized with isoflurane at 5%, and anesthesia is maintained by nose cone at 1%. 2. The mice are euthanized per approved protocol, and tumors removed by surgery and added to cold PBS. 3. Cut ~300–500 mm3 of a tumor using a scalpel, and place in a 15 mL centrifuge tube with 10 mL room temperature 10% formalin. 4. Remove the blood, blood vessels, calcification, and necrosis. 5. Cut ~50 mg tumor and place into a 2 mL cryopreserve tube. Snap freeze by liquid nitrogen, and/or put the tumor into a 2 mL cryopreservation tube containing 1.5 mL RNA later.

Protocols for Cancer Target Pharmacology

205

6. The remaining tumor tissues are transferred into RPMI-1640 and kept on ice before transplantation or cell digestion. 7. Tissues should be collected immediately after mice are euthanized. If the tumor/mouse needs to be photographed, which is required at the same time as the mouse is euthanized, then it is recommended to not collect the tumor. 7.1.7 Flow Cytometry

Hematological markers (or immunological markers) are important for the characterization of hematological cancers and tumorinfiltrating immune cells and the immune components of the tumor microenvironment [38]. Multicolor FACS and/or immunofluorescence (IF) (or IHC) analysis are commonly performed to monitor these immune markers. Flow cytometry is most commonly used to monitor blood tumor growth or tumor-infiltrating immune cells in I/O animal modeling for I/O targets. Alternately, IF and IHC can reveal the location information of immune cells within tumors.

Tumor Dissociation

The first step for flow analysis is the dissociation of tumors into single-cell suspensions using various protocols, optimized for different tumors allowing the greatest yield of viable immune or leukemic cells [38]. However, the general simplified steps can be represented as follows. 1. Wash the tumor in PBS, and remove normal tissues attached to the tumor (e.g., blood vessel, fat, fascia, etc.). 2. Digest the tumor pieces as per the method described in Subheading 7 (Blood Cancer Model Inoculation). 3. Count cells using a cell counter, and adjust cell concentration to 1  106 cells per tube or per sample.

Staining

1. Fc block sample cells: resuspend cells in 200 μL staining buffer with 1 μg/mL Fc block (Mouse BD Fc Block™), followed by incubation on ice or 4  C refrigeration for 15 min in the dark. 2. Stain cells using the desired antibody/fluorescence panels (e.g., T-cell panel, macrophage panel, etc.): add the antibody mixture diluted in Fc blocking buffer to each sample; stain for at least 30 min on ice in the dark. 3. Add 1 mL of ice-cold PBS to each tube, and gently resuspend the cells, followed by centrifugation at 300  g for 5 min. Discard the resulting supernatant. 4. Repeat step 3 to wash the cells a total of twice. 5. Stain for intracellular markers if needed, following steps 6–10, otherwise proceed to step 11. 6. Resuspend the cell pellet by pulse vortex, and add 200 μL of prepared fixation/permeabilization working solution for each

206

Dawei Chen et al.

sample. Pulse vortex again, and then incubate at 4  C overnight (preferred) or 30 min at room temperature in the dark. 7. Pellet cells and remove the supernatant. 8. Wash twice by adding 1 mL of 1 permeabilization buffer (made from 10 permeabilization buffer, diluted with distilled H2O) followed by centrifugation and decanting of supernatant. 9. Add intracellular marker antibody in 1 permeabilization buffer, and incubate at room temperature for 30 min in the dark. 10. Wash cells twice with 1 mL of 1 permeabilization buffer. Centrifuge and decant supernatant. 11. Resuspend cells in 150 μL of staining buffer and analyze on cytometer. Due to the fixation and permeabilization procedure, the FSC (forward-light scatter)/SSC (sidelight scatter) distribution of the cell population will be different to live cells. Therefore, the gate and voltages will need to be modified. 12. The data are analyzed by Flowjo v10 software. FMO Controls (Fluorescence Minus One Control)

Flow Instrument Setup

Multicolor flow cytometry is centrally important for today’s I/O tumor-infiltrating immune cell (TILs) analysis, requiring a way to identify and gate cells in the context of data spread due to the multiple fluorochromes in a given panel. FMO controls or Fluorescence Minus One control is an important approach for this purpose. To this end, additional mice should usually be included in each Rx group for FMO controls (at least two per Rx) and processed individually for each tissue. After tissue dissociation, tissues should be pooled. For example, in a study with four Rx groups, eight additional tumors should be processed individually and then pooled into one sample for FMO’s. 1. Prepare compensation beads while the machine is warming up (at least 20 min). 2. Use BD’s CS&T beads to check performance. 3. Voltage and compensation settings: use eBioscience’s UltraComp beads, and vortex the UltraComp beads thoroughly before use. 4. Label a separate 12  75 mm sample tube (BD Falcon™) for each fluorochrome-conjugated antibody. 5. Add 100 μL of staining buffer (e.g., BD Pharmingen Stain (FBS)) to each tube. 6. Add 1 drop (approximately 60 μL) of the UltraComp eBeads to each tube.

Protocols for Cancer Target Pharmacology

207

7. Add antibodies, and perform the staining procedure as per the sample process stated in Subheading 7. 8. Resuspend the bead pellets by adding 0.5 mL of staining buffer to each tube. Vortex thoroughly. 9. Set the flow cytometer instrument PMT voltage settings using the target tissue for the given experiment. 10. Run each tube separately on the flow cytometer. Gate on the singlet bead population based on FSC and SSC characteristics. 11. Adjust flow rate to 200–300 events per second. 12. Create a dot plot for the given fluorochrome-conjugated antibody as appropriate (i.e., to set compensation for a fluorescein (FITC)-conjugated antibody, use an FL1 vs. FL2 dot plot). 13. Place a quadrant gate such that the negative bead population is in the lower left quadrant and the positive bead population is in the upper or lower right quadrant. Adjust the compensation values until the median fluorescence intensity (MFI) of each population (as shown in the quadrant stats window) is approximately equal (i.e., for FL2-%FL1, the FL2 MFI of both bead populations should be approximately equal when properly compensated). 14. Repeat steps 12 and 13 for each of the experimental tubes. 15. Proceed to acquiring the actual staining experiment. Run the compensation wizard, and save the settings with the format “date experiment your initials.” 7.2 General Consideration of Tumor Selection

8

Appropriate tumor model(s) are chosen dependent on the specific objectives of the target investigation (Table 4). Common parameters include species-specific targeting (e.g., human vs. mouse), conventional targeting vs. I/O targeting, tumor growth kinetics vs. tumorigenicity (e.g., initiation), etc.

Notes In order to use adequately different cancer models, one needs to fully understand the nature of these models, or in other words, the full-annotation of these models should be available for facilitating specific study design and model selection. The common annotations include genomic profile, pathology, driver mutations, as well as tumor microenvironment [38] particularly when an I/O study is planned. Using PDX as an example, certain criteria usually need to be met for optimal utilization of PDX, including (1) comprehensive genomic profiling (gene copy number (GCN), mRNA expression, mutation profile, gene rearrangements, etc.), where transcriptome

208

Dawei Chen et al.

Table 4 Different tumor models for target research Tumor model

Comments

Examples

Spontaneous No transplantation; suitable for mouse targeting tumor and I/O targeting; for studies involving tumor initiation

GEMM tumors

Autograft

Transplanted tumors from one location to another Skin transplantation of the same location of the same individual (identical MHC) person (similar to syngeneic transplantation)

Homograft

Donor and recipient are different individuals but Syngeneic tumors, MuPrime™ (murine version of PDX) belong to the same inbred of strain or identical twin (identical MHC); suitable for mouse targeting and I/O targeting

Allograft

Donor and recipient are different individuals (usually unmatched MHC)

Xenograft

Transplanted between species; for human targeting PDX, cell line-derived xenograft, rat tumor in mice

Common organ donor

sequencing (RNAseq) and whole exome sequencing (WES) have been commonly used; (2) adequate/stable tumor take rate and growth kinetics to support consistent and reliable results; and (3) sufficient banking of early passages to support relevant and reproducible investigations. Due to the potential changes during continuous passages, some models may drift in their growth properties (usually faster at later passage) [39], their genomic profile, and response to drugs. We therefore recommend to use passages after three, but fewer than ten, for relatively stable properties in these aspects. We also recommend, if possible, establishing a master bank of cryopreserved tumors at early passages. For xenograft model growth in mice, spontaneous mouse tumors occasionally occur and contaminate human tumor or completely take over, which would result in misleading observations. A QC process to quantify mouse/human DNA ratio could help to determine if there is mouse tumor contamination. Another key process vital to the success of cancer pharmacology program is the unique identification (ID) of each models, since the morphology of tumor or even growth will not be able to distinguish individual models. Most common technologies used for xenografts (PDX, cell line-derived xenografts) are STR described in this chapter and HLA typing, superior to genomic profile in cost and efficiency. However, mouse tumor ID cannot usually and readily use these methods, due to that most mouse tumors are derived from inbred and difference from those of human. To this end, usually certain genomic methods have been custom-developed (Cai et al., unpublished).

Protocols for Cancer Target Pharmacology

209

Acknowledgments The authors would like to thank all the members of the Translational Oncology Division, Crown Bioscience, Inc. for their dedicated work in cancer animal modeling over the last decade, which contributes to many of these protocols. The authors would also like to thank Dr. Jody Barbeau for careful reading and editing of this manuscript and Mr. Ralph Joseph Manuel for some of the artworks. Dawei Chen and Xiaoyu An contributed equally to this work. References 1. Hsu PD, Scott DA, Weinstein JA, Ran FA, Konermann S, Agarwala V, Li Y, Fine EJ, Wu X, Shalem O, Cradick TJ, Marraffini LA, Bao G, Zhang F (2013) DNA targeting specificity of RNA-guided Cas9 nucleases. Nat Biotechnol 31(9):827–832. https://doi.org/10. 1038/nbt.2647 2. Ran FA, Hsu PD, Wright J, Agarwala V, Scott DA, Zhang F (2013) Genome engineering using the CRISPR-Cas9 system. Nat Protoc 8 (11):2281–2308. https://doi.org/10.1038/ nprot.2013.143 3. Stovall DB, Wan M, Zhang Q, Dubey P, Sui G (2012) DNA vector-based RNA interference to study gene function in cancer. J Vis Exp 64:e4129. https://doi.org/10.3791/4129 4. Li QX, Feuer G, Ouyang X, An X (2017) Experimental animal modeling for immunooncology. Pharmacol Ther 173:34–46. https://doi.org/10.1016/j.pharmthera.2017. 02.002 5. Ke N, Zhou D, Chatterton JE, Liu G, Chionis J, Zhang J, Tsugawa L, Lynn R, Yu D, Meyhack B, Wong-Staal F, Li QX (2006) A new inducible RNAi xenograft model for assessing the staged tumor response to mTOR silencing. Exp Cell Res 312 (15):2726–2734. https://doi.org/10.1016/j. yexcr.2006.05.001 6. Liu G, Wong-Staal F, Li QX (2006) Recent development of RNAi in drug target discovery and validation. Drug Discov Today Technol 3:293–300. https://doi.org/10.1016/j. ddtec.2006.09.003 7. Yang JP, Fan W, Rogers C, Chatterton JE, Bliesath J, Liu G, Ke N, Wang CY, Rhoades K, Wong-Staal F, Li QX (2006) A novel RNAi library based on partially randomized consensus sequences of nuclear receptors: identifying the receptors involved in amyloid beta degradation. Genomics 88 (3):282–292

8. Zhou D, Wang C, Zhang J, Bliesath J, He QS, Ke N, Yu D, Li Q, Zhang LH, Wong-Staal F (2008) Generation of shRNA pool library: a revision of the biological technique from the viewpoint of chemistry. Chembiochem 9:1365–1367. https://doi.org/10.1002/ cbic.200800049 9. Gargiulo G, Serresi M, Cesaroni M, Hulsman D, van Lohuizen M (2014) In vivo shRNA screens in solid tumors. Nat Protoc 9:2880–2902. https://doi.org/10.1038/ nprot.2014.185 10. Zender L, Xue W, Zuber J, Semighini CP, Krasnitz A, Ma B, Zender P, Kubicka S, Luk JM, Schirmacher P, Mccombie WR, Wigler M, Hicks J, Hannon GJ, Powers S, Lowe SW (2008) An oncogenomics-based in vivo RNAi screen identifies tumor suppressors in liver cancer. Cell 135:852–864. https://doi.org/10. 1016/j.cell.2008.09.061 11. Baratta MG, Schinzel AC, Zwang Y, Bandopadhayay P, Bowman-Colin C, Kutt J, Curtis J, Piao H, Wong LC, Kung AL, Beroukhim R, Bradner JE, Drapkin R, Hahn WC, Liu JF, Livingston DM (2015) An in-tumor genetic screen reveals that the BET bromodomain protein, BRD4, is a potential therapeutic target in ovarian carcinoma. Proc Natl Acad Sci 112:232–237. https://doi.org/ 10.1073/pnas.1422165112 12. Murugaesu N, Iravani M, Van Weverwijk A, Ivetic A, Johnson DA, Antonopoulos A, Fearns A, Jamal-Hanjani M, Sims D, Fenwick K, Mitsopoulos C, Gao Q, Orr N, Zvelebil M, Haslam SM, Dell A, Yarwood H, Lord CJ, Ashworth A, Isacke CM (2017) An in vivo functional screen identifi es ST6GalNAc2 sialyltransferase as a breast cancer metastasis suppressor. Cancer Discov 4(3):304–317. https://doi.org/10.1158/2159-8290.CD13-0287 13. Talmadge JE, Singh RK, Fidler IJ, Raz A (2007) Murine models to evaluate novel and

210

Dawei Chen et al.

conventional therapeutic strategies for cancer. Am J Pathol 170:793–804. https://doi.org/ 10.2353/ajpath.2007.060929 14. Garraway LA, Lander ES (2013) Lessons from the cancer genome. Cell 153(1):17–37. https://doi.org/10.1016/j.cell.2013.03.002 15. Myers JN, Holsinger FC, Jasser SA, Bekele BN, Fidler IJ (2002) An orthotopic nude mouse model of oral tongue squamous cell carcinoma. Clin Cancer Res 8:293–298 16. Platt RJ, Chen S, Zhou Y, Yim MJ, Swiech L, Kempton HR, Dahlman JE, Parnas O, Eisenhaure TM, Jovanovic M, Graham DB, Jhunjhunwala S, Heidenreich M, Xavier RJ, Langer R, Anderson DG, Hacohen N, Regev A, Feng G, Sharp PA, Zhang F (2014) CRISPR-Cas9 knockin mice for genome editing and cancer modeling. Cell 159:440–455. https://doi.org/10.1016/j.cell.2014.09.014 17. Harms DW, Quadros RM, Seruggia D, Ohtsuka M, Takahashi G, Montoliu L, Gurumurthy CB (2014) Mouse genome editing using the CRISPR/Cas system. Curr Protoc Hum Genet 83:15.17.11–15.17.27. https:// doi.org/10.1002/0471142905.hg1507s83 18. Menezes ME, Shen X-N, Das SK, Emdad L, Guo C, Yuan F, Li Y-J, Archer MC, Zacksenhaus E, Windle JJ, Subler MA, Ben-David Y, Sarkar D, Wang X-Y, Fisher PB (2015) MDA-7/IL-24 functions as a tumor suppressor gene in vivo in transgenic mouse models of breast cancer. Oncotarget 6:36928–36942 19. Walter AO, Tjin R, Sjin T, Haringsma HJ, Ohashi K, Sun J, Lee K, Dubrovskiy A, Labenski M, Zhu Z, Wang Z, Sheets M, Martin TS, Karp R, Kalken DV, Chaturvedi P, Niu D, Nacht M, Petter RC, Westlin W, Lin K, Jaw-tsai S, Raponi M, Dyke TV, Etter J, Weaver Z, Pao W, Singh J, Simmons AD, Harding TC, Allen A, Sjin RTT, Haringsma HJ, Sun J, Ohashi K, Lee K, Dubrovskiy A, Labenski M, Wang Z, Zhu Z, Sheets M, Martin TS, Karp R, van Kalken D, Chaturvedi P, Niu D, Nacht M, Petter RC, Lin K, Westlin W, Jaw-tsai S, Raponi M, Van Dyke T, Etter J, Pao W, Weaver Z, Singh J, Simmons AD, Harding TC, Allen A (2013) Discovery of a mutant-selective covalent inhibitor of EGFR that overcomes T790M- mediated resistance in NSCLC. Cancer Discov 3:1404–1416. https://doi.org/10.1158/2159-8290.CD13-0314 20. Vooijs M, Jonkers J, Lyons S (2002) Noninvasive imaging of spontaneous retinoblastoma pathway-dependent tumors in mice. Cancer Res 62:1862–1867

21. Wang Y, Tseng J-C, Sun Y, Beck AH, Kung AL (2015) Noninvasive imaging of tumor burden and molecular pathways in mouse models of cancer. Cold Spring Harb Protoc 2015:135–144. https://doi.org/10.1101/ pdb.top069930 22. Buonincontri G, Methner C, Carpenter TA, Hawkes RC, Sawiak SJ, Krieg T (2013) MRI and PET in mouse models of myocardial infarction video link. J Vis Exp 82:e50806. https:// doi.org/10.3791/50806 23. Lute KD, May KF, Lu P, Zhang H, Kocak E, Mosinger B, Wolford C, Phillips G, Caligiuri MA, Zheng P, Liu Y (2005) Human CTLA4 knock-in mice unravel the quantitative link between tumor immunity and autoimmunity induced by anti–CTLA-4 antibodies. Blood 106:3127–3133. https://doi.org/10.1182/ blood-2005-06-2298 24. Peggs KS, Quezada SA, Chambers CA, Korman AJ, Allison aJP, Allison JP (2009) Blockade of CTLA-4 on both effector and regulatory T cell compartments contributes to the antitumor activity of anti–CTLA-4 antibodies. J Exp Med 206:1717–1725. https://doi.org/10. 1084/jem.20082492 25. Tentler JJ, Tan AC, Weekes CD, Jimeno A, Leong S, Pitts TM, Arcaroli JJ, Messersmith WA, Eckhardt SG (2012) Patient-derived tumour xenografts as models for oncology drug development. Nat Rev Clin Oncol 9 (6):338–350. https://doi.org/10.1038/ nrclinonc.2012.61 26. Guo S, Qian W, Cai J, Zhang L, Wery JP, Li QX (2016) Molecular pathology of patient tumors, patient-derived xenografts, and cancer cell lines. Cancer Res 76(16):4619–4626. https://doi.org/10.1158/0008-5472.CAN15-3245 27. Yang M, Shan B, Li Q, Song X, Cai J, Deng J, Zhang L, Du Z, Lu J, Chen T, Wery JP, Chen Y (2013) Overcoming erlotinib resistance with tailored treatment regimen in patient-derived xenografts from naive Asian NSCLC patients. Int J Cancer 132(2):E74–E84. https://doi. org/10.1002/ijc.27813 28. Yang M, Xu X, Cai J, Ning J, Wery JP, Li QX (2016) NSCLC harboring EGFR exon-20 insertions after the regulatory C-helix of kinase domain responds poorly to known EGFR inhibitors. Int J Cancer 139(1):171–176. https:// doi.org/10.1002/ijc.30047 29. Zhang L, Yang J, Cai J, Song X, Deng J, Huang X, Chen D, Yang M, Wery JP, Li S, Wu A, Li Z, Liu Y, Chen Y, Li Q, Ji J (2013) A subset of gastric cancers with EGFR amplification and overexpression respond to

Protocols for Cancer Target Pharmacology cetuximab therapy. Sci Rep 3:2992. https:// doi.org/10.1038/srep02992 30. An X, Liu J, Wang N, Wang D, Huang L, Zhang L, Cai J, Wery JP, Zhou D, Zhou J, Li QX (2017) AC220 and AraC cause differential inhibitory dynamics in patient-derived M5-AML with FLT3-ITD and, thus, ultimately distinct therapeutic outcomes. Exp Hematol 45:36–44 e32. https://doi.org/10. 1016/j.exphem.2016.09.004 31. Czauderna F, Santel A, Hinz M, Fechtner M, Durieux B, Fisch G, Leenders F, Arnold W, Giese K, Klippel A, Jr K (2003) Inducible shRNA expression for application in a prostate cancer mouse model. Nucleic Acids Res 31: e127. https://doi.org/10.1093/nar/gng127 32. Speers C, Zhao SG, Kothari V, Santola A, Liu M, Wilder-Romans K, Evans J, Batra N, Bartelink H, Hayes DF, Lawrence TS, Brown PH, Pierce LJ, Feng FY (2016) Maternal embryonic leucine zipper kinase (MELK) as a novel mediator and biomarker of radioresistance in human breast cancer. Clin Cancer Res 22(23):5864–5875. https://doi.org/10. 1158/1078-0432.CCR-15-2711 33. Ma L, Saiyin H (2017) p48-Cre mice are an applicable model for locally invasive and metastatic pancreatic. Cancer:1–17 34. Chen D, Huang X, Cai J, Guo S, Qian W, Wery J-P, Li Q-x, Cai J, Guo S, Qian W, Wery J-P, Li Q-x (2015) A set of defined oncogenic mutation alleles seems to better predict the response to cetuximab in CRC patient-derived xenograft than KRAS 12/13 mutations. Oncotarget 6:40815–40821. https://doi.org/10.18632/ oncotarget.5886 35. Guo S, Chen D, Huang X, Cai J, Wery JP, Li Q-X, Sheng Guo DC (2016) Cetuximab response in CRC patient-derived xenografts seems predicted by an expression based Ras pathway signature. Oncotarget

211

7:50575–50581. https://doi.org/10.18632/ oncotarget.10499 36. Gao H, Korn JM, Ferretti S, Monahan JE, Wang Y, Singh M, Zhang C, Schnell C, Yang G, Zhang Y, Balbin OA, Barbe S, Cai H, Casey F, Chatterjee S, Chiang DY, Chuai S, Cogan SM, Collins SD, Dammassa E, Ebel N, Embry M, Green J, Kauffmann A, Kowal C, Leary RJ, Lehar J, Liang Y, Loo A, Lorenzana E, Robert McDonald E 3rd, McLaughlin ME, Merkin J, Meyer R, Naylor TL, Patawaran M, Reddy A, Roelli C, Ruddy DA, Salangsang F, Santacroce F, Singh AP, Tang Y, Tinetto W, Tobler S, Velazquez R, Venkatesan K, Von Arx F, Wang HQ, Wang Z, Wiesmann M, Wyss D, Xu F, Bitter H, Atadja P, Lees E, Hofmann F, Li E, Keen N, Cozens R, Jensen MR, Pryer NK, Williams JA, Sellers WR (2015) Highthroughput screening using patient-derived tumor xenografts to predict clinical trial drug response. Nat Med 21(11):1318–1325. https://doi.org/10.1038/nm.3954 37. Fridman R, Benton G, Aranoutova I, Kleinman HK, Bonfil RD (2012) Increased initiation and growth of tumor cell lines, cancer stem cells and biopsy material in mice using basement membrane matrix protein (Cultrex or Matrigel) co-injection. Nat Protoc 7:1138–1144. https://doi.org/10.1038/nprot.2012.053 38. Xiaoyu An XO, Zhang H, Li T, Zhou D, Li Q-X (2018) Immunophenotyping of orthotopic homograft (Syngeneic) of murine primary KPC pancreatic 2 ductal adenocarcinoma by flow cytometry. J Vis Exp 140:e57460 39. Pearson AT, Finkel KA, Warner KA, No¨r F, Tice D, Martins MD, Jackson TL, No¨r JE (2016) Patient-derived xenograft (PDX) tumors increase growth rate with time. Oncotarget 7:7993–8005. https://doi.org/10. 18632/oncotarget.6919

Chapter 13 Use of CRISPR/Cas9 for the Modification of the Mouse Genome Alexander Klimke, Steffen Gu¨ttler, Petric Kuballa, Simone Janzen, Sonja Ortmann, and Adriano Flora Abstract The use of CRISPR/Cas9 to modify the mouse genome has gained immense interest in the past few years since it allows the direct modification of embryos, bypassing the need of labor-intensive procedures for the manipulation of embryonic stem cells. By shortening the overall timelines and reducing the costs for the generation of new genetically modified mouse lines (Li et al., Nat Biotechnol 31: 681–683, 2013), this technology has rapidly become a major tool for in vivo drug discovery applications. Key words CRISPR/Cas9, Mouse models, Knockout, Knock-in, Genome editing, In vivo drug discovery

1

Introduction Genome editing relies on the introduction of double-strand breaks (DSB) at defined genomic locations, followed by mutagenic events mediated by DNA repair pathways [1]. The genome editing CRISPR/Cas9 system comprises a ribonucleoprotein complex composed of a proteinaceous double-strand DNA nuclease (Cas9), a trans-activating RNA (tracrRNA) which binds the Cas9 protein and a CRISPR-RNA (crRNA) that defines the target specificity. The crRNA-tracrRNA hybrid molecule is defined as the guide RNA (gRNA), which can also be transcribed in vitro as a single RNA molecule (sgRNA) [2, 3]. The interaction between the guide RNA and its complementary genomic sequence mediates the recruitment of Cas9 on the specific target site. Once associated with the genomic DNA, Cas9 introduces a DSB, activating the endogenous DNA repair machinery of the cell [4]. The nature of the mutation introduced at the DSB site depends on the specific pathway triggered by the cell to repair the damage: the nonhomologous end-joining (NHEJ) DNA repair mechanism can introduce

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_13, © Springer Science+Business Media, LLC, part of Springer Nature 2019

213

214

Alexander Klimke et al.

deletions, whereas the homology-directed DNA repair machinery (HDR) can use exogenous DNA sequences as templates and introduce specific mutations. Generally, the NHEJ mechanism is much more efficient than the HDR pathway [5]. Due to its remarkable efficiency, the CRISPR/Cas9 system is well suited for the modification of the embryonic mouse genome. In the following sections, we will describe a robust methodology to generate constitutive knockout and knock-in alleles and provide some general guidelines to design optimized mouse models using the CRISPR/Cas9 approach. 1.1 Targeting Strategy Design

Generation of a new genetically engineered model always starts with the definition of the optimal strategy to introduce the desired mutation. Poor targeting design can result in a waste of time and resources on the characterization of suboptimal mouse models.

1.1.1 Design of a Knockout Allele

Knockout alleles are designed with the goal of ablating the function of the target gene. The first knockout alleles using CRISPR/Cas9 were generated by introducing a single deletion in an exonic sequence to promote the introduction of a small insertion or deletion (indel) during NHEJ that would render the downstream coding sequence out of frame [6]. Although very elegant, this approach has some limitations. The most important being that compensatory mechanisms (e.g., alternative splicing events) might lead to a partial or even full rescue of the function of the targeted allele. A much safer approach toward the generation of knockout alleles is introducing two independent lesions in the target gene by using two gRNAs cutting at distinct sites of the gene. When two DSBs are present in the same locus, the NHEJ machinery can efficiently join them, leading to the deletion of the entire intervening sequence (Fig. 1). Since the NHEJ-mediated deletion mechanism is very efficient over distances of tens of kilobases ([7] and our unpublished data), this strategy has now become the preferred approach for the generation of knockout alleles.

1.1.2 Design of a KnockIn Allele

The design of a knock-in allele is dictated by the position and nature of the desired point mutation (PM) in the mouse genome. The general approach is to co-inject the CRISPR/Cas9 complex with a single-stranded DNA donor molecule that can be used by the HDR system as a template to repair the DSB. The result of this process is the swap of wild-type sequence to the mutated donor sequence at the target site (Fig. 2). One of the most critical aspects of a successful CRISPR/Cas9 knock-in strategy is the distance between the Cas9-induced DSB and the mutation to be introduced. For the introduction of point mutations, this distance should be shorter than ten nucleotides, as the efficiency of introduction of the PM dramatically drops after this threshold (unpublished results). Thus, the limiting factor in the design of a successful knock-in strategy is

Mouse Model Generation using CRISPR/Cas9

215

Fig. 1 Preferred strategy for the generation of a constitutive knockout using CRISPR/Cas9. Top panel: two gRNAs mediate the introduction of DSBs flanking the genomic region to be deleted in the target gene. The NHEJ DNA repair mechanism mediates the fusion of the two free ends and the deletion of the intervening sequence. Bottom panel: an example of a knockout by deletion is depicted where 15 coding exons of a large gene have been eliminated to create a nonfunctional allele

the availability of an optimal gRNA sequence within the vicinity of the site to mutagenize. If no such gRNAs can be designed, it may be necessary to consider alternative technologies for the targeting. For maximal efficiency, a suitable donor oligonucleotide should have the same directionality as the gRNA used, i.e., if a given guide RNA has a sense orientation, the oligonucleotide should be oriented in sense as well to prevent binding of the gRNA to the donor oligo. Although mutagenesis rates may increase with increasing lengths of the two homology arms, homology arms with a length of 30–50 bases are sufficient to generate founder animals carrying the desired point mutation. The homology arms are synthesized 50 and 30 of the point mutation and may contain as well silent mutations that have to be introduced for analytical purposes. The purification of the oligo, e.g., via PAGE, may decrease unwanted mutations in the oligonucleotide, but, in our hands, the standard desalted oligos have been proven to be sufficient.

216

Alexander Klimke et al.

Fig. 2 Strategy to generate a knock-in allele. Top panel: a single-stranded DNA molecule is added to the microinjection mixture and used by the homology-driven DNA repair mechanism as a template molecule. Bottom panel: shown is an example of the knock-in of a point mutation 1.1.3 Design of Large Knock-In Alleles

Long synthesized oligonucleotides can serve as donor template for homologous recombination. While oligonucleotides of a size of up to 200 bp are commercially available (e.g., Ultramer Oligonucleotides from Integrated DNA technologies), synthesis of longer oligonucleotides with a sufficient yield and purity remains technically challenging. While size limitations can be resolved by the use of double-stranded DNA (dsDNA), the major disadvantage is that dsDNA integrates randomly into the genome with an efficiency that excels that of the desired homologous recombination [8] and HDR-positive founder animals will likely carry random insertions. These unwanted modifications need to be detected and can only be removed by extensive breeding and genotyping. Single-stranded DNA (ssDNA) can be used to overcome this issue. In addition to avoiding random integration, use of ssDNA results in higher recombination rates compared to dsDNA [9, 10]. Miura et al. describe the generation of ssDNA of approximately 0.5 kb generated by reverse transcription and its application as HDR donor. An alternative method for generation of large

Mouse Model Generation using CRISPR/Cas9

217

ssDNA molecules is the usage of nicking enzymes and nucleasemediated removal of the remaining strand [11]. Commercially available ssDNA was successfully used by Quadros et al. and Yang et al. reporting the insertion of up to 2 kb [10, 12]. In contrast to the design of knock-in models using oligonucleotides as donors, for which the length of the homology region flanking the region to be inserted is limited by the maximal size of the oligonucleotide, use of ssDNA allows more extended homology arms in the donor DNA. While a sequence of 50–100 bp seems sufficient for the insertion of up to 1000 bp [12, 13], elongation of the homology arms seems to have a positive impact on the recombination efficiency as seen for classical gene targeting [11]. 1.2 Guide RNA Selection

Various online tools are publicly available to identify gRNAs within a given target [14–18]. Many researchers show a preference for a handful of these programs, e.g., CHOPCHOP [19, 20] or the CRISPR design tool from the Zhang lab (www.crispr.mit.edu). While the choice of which tool to use is more or less dependent on personal preferences, it is good practice to use more than one of them and compare the results to identify the optimal gRNA. Different parameters such as the GC content and the number of potential off-targets should be considered when searching for a suitable gRNA. Guide RNAs with a GC content between 40 and 70% have been reported to correlate with a high cleavage efficiency of the Cas9 protein [16], while three or more total mismatches are expected to have a low off-target probability if two of these mismatches are in the seed region [4]. If only suboptimal gRNAs which do not suit the selection criteria can be designed, testing those in cell culture or by blastocyst analysis is recommended [21]. Also, if single nucleotide polymorphisms are annotated within the binding sequence of the gRNA, one should sequence this particular region beforehand, primarily when working with mouse strains different from the public database reference C57BL/6 N strain.

1.3 Genotyping Strategy and OffTarget Analysis

Mosaicism complicates the genotyping of founder animals. For knockout alleles, it is usually possible to design PCR primers around the region that will be deleted and confirm the generation of a knockout allele by amplification of a fragment of DNA of the expected size. Multiple similar knockout alleles can be present in the same founder and it is of paramount importance to characterize them, either by subcloning the DNA fragment corresponding to the knockout allele and sequencing multiple colonies or by performing next-generation sequencing directly on the PCR product. Knock-in alleles of point mutations are more difficult to identify by PCR compared to deletion alleles. To facilitate their detection, we suggest introducing a restriction enzyme recognition site

218

Alexander Klimke et al.

nearby the desired point mutation by inserting additional silent mutations in codons flanking the target sequence. When screening for potential founders, it will, therefore, be possible to perform enzymatic digestion of the genomic PCR with the newly introduced site and identify animals where the sequence of the oligonucleotide has been integrated into the target site. Also in this case, however, it is necessary to subclone the PCR product and sequence multiple colonies (or perform next-generation sequencing) to confirm the presence of the desired mutation on the founder genome. Larger fragment knock-in alleles allow a genotyping relying on insert-specific PCR primers combined with primers outside the modified region. PCRs can be designed in an overlapping manner to cover the entire knock-in sequences as well as the to ensure the correct integration. Sequencing of the PCR products (or subclones) is recommended to confirm that the insert does not contain mutations. In all cases, it is crucial to sequence all the progeny of founders, in order to verify germline transmission of the desired alleles. 1.4 Off-Target Effects

In the past, a significant concern regarding the use of CRISPR/ Cas9 to edit the mammalian genome has been the modification of genomic sequences sharing similarities with the target site. High frequencies of off-target effects have indeed been reported when using the CRISPR/Cas system in various human cell lines in vitro [22]. However, the analysis of mice generated by using CRISPR/ Cas9 suggests that the frequency of these events may be much lower in vivo than in vitro [6, 10]. Our data from the analysis of 505 genomic loci of 31 genome-edited ES cell clones and 310 loci of 555 heterozygous mice generated by embryonic manipulation indicates that less than 0.5% of gRNAs selected with the criteria described below show any activity on off-target sites.

1.5 Guide Selection and Preparation

To identify suitable guide RNAs, the sequence of interest can be analyzed with one of the tools mentioned in one of the previous chapters of this book or “CRISPR design” (www.crispr.mit.edu), which gives an output of guide RNAs ranked by inverse likelihood of off-target binding. Once the appropriate guide RNAs have been identified, they can be generated either by in vitro transcription as fusion molecules between the tracrRNA and the crRNAs (called single guide RNAs or sgRNAs) or by ordering a target-specific crRNA along with a generic tracrRNA. While the first approach requires cloning, sequencing, and in vitro transcription from a template, the latter system allows the use of the same tracrRNA together with locusspecific crRNAs, thus reducing costs and effort. Due to these advantages and the broad commercial availability, we routinely use the crRNA-tracrRNA system for all CRISPR applications.

Mouse Model Generation using CRISPR/Cas9

219

Once the gRNAs and, if applicable, the HDR oligonucleotide or ssDNA has been designed and received, the microinjection mix can be prepared.

2

Materials

2.1 Microinjection Mixture Preparation

RNase-free 1.5 mL Eppendorf tubes. RNase-free filter tips. RNase-free stripettes. RNaseZap (Life Technologies/Ambion, AM9780). Injection buffer (5 mM Tris–HCl, 0.1 mM EDTA, pH 7.4). Cas9 nuclease (protein), S. pyogenes, 20 μM (New England Biolabs, M0386 M). Synthetic tracrRNA Alt-R™ CRISPR-Cas9 tracrRNA, 100 nmol (IDT, 1072534) or. Synthetic tracrRNA (Dharmacon, U-002000-20). Locus-specific nucleotides (synthetic crRNA, HDR oligonucleotide, or ssDNA). Nanodrop. Thermoblock (Thermomixer).

2.2 Embryo Injection and Transfer Materials

Air compressor. Manipulator (Eppendorf, Transfer Man). Microinjector (Eppendorf, CellTram Air/oil/vario). Microscope (Leica, DM-IRBE). Holding needle. FemtoJet. Microloader. Borosilicate glass capillaries (Clark Electromedical GC120TF-10; 1.2 mm OD  0.94 mm ID). Retransfer pipettes L ¼ 160 mm).

2.3

Genotyping

(BioMedical

Instr.,

NEB-Q5 polymerase, order # M0491.

HB

1.8

Instr., 

1.2,

220

3

Alexander Klimke et al.

Methods

3.1 Concentrations and Calculation of the Mix Components 3.1.1 Concentrations Within the Final Mix 3.1.2 Calculating the Concentrations of the Pre-dilutions

The concentrations of Cas9 protein and crRNA-tracrRNA depend on the number of crRNAs used and not on the number of HDR oligos or plasmid DNA. Table 1 provides examples for the concentration of the different reagents in mixes designed for the use of one or two crRNAs. 1. Each component must be prepared as a pre-dilution. These pre-dilutions are then combined into the final mix. 2. Each pre-dilution should be set up in a volume of 20 μL. 3. When equal volumes of these pre-dilutions are combined, the following formula can be used to calculate the concentrations of the pre-dilutions: (Conc. within final mix)  (Number of pre-dilutions)  (Volume of pre-dilution). Table 2 shows examples of pre-dilutions for the use of one or two crRNAs for the generation of knock-in and knockout alleles, respectively.

3.2 Dissolving and Storing crRNA and tracrRNA

1. Lyophilized crRNA should be stored at 20  C until being dissolved. Note that the work with RNA reagents requires some important aspects to be considered (see Note 4.1). 2. If 20 nmol have ordered, add 500 μL injection buffer to the RNA (final conc. ~40 μM). 3. If two nmol have ordered, add 50 μL injection buffer to the RNA (final conc. ~40 μM). 4. Incubate for 30 min at room temperature. 5. Spin the tube shortly to ensure that the solution is collected at the bottom. 6. Measure the exact concentration by Nanodrop.

Table 1 Examples of concentration for the different reagents for reactions with one or two crRNAs 1 crRNA:

2 crRNAs:

Cas9 protein:

55 ng/μL

Cas9 protein:

73 ng/μL

crRNA:

9 ng/μL

crRNA 1:

6 ng/μL

tracrRNA:

15.5 ng/μL

crRNA 2:

6 ng/μL

tracrRNA:

21 ng/μL

Optional: HDR oligonucleotide:

5 ng/μL

or ssDNA:

100 ng/μL

Mouse Model Generation using CRISPR/Cas9

221

Table 2 Examples of pre-dilutions for one or two crRNAs reactions 4 pre-dilutions:

4 pre-dilutions:

Cas9 protein, 1 crRNA, 1 tracrRNA, 1 HDR oligonucleotide/ssDNA

Cas9 protein, 2 crRNAs, 1 tracrRNA

Cas9 protein:

55 ng/μL  4  20 μL ¼ 4400 ng Cas9 73 ng/μL  4  20 μL protein: ¼ 5840 ng

4400 ng Cas9 protein in injection buffer (total volume of 20 μL) crRNA:

9 ng/μL  4  20 μL ¼ 720 ng

720 ng crRNA in injection buffer (total volume of 20 μL) tracrRNA:

15.5 ng/μL  4  20 μL ¼ 1240 ng

5840 ng Cas9 protein in injection buffer (total volume of 20 μL) crRNA1:

6 ng/μL  4  20 μL ¼ 480 ng

480 ng crRNA in injection buffer (total volume of 20 μL) crRNA2:

6 ng/μL  4  20 μL ¼ 480 ng

1240 ng tracrRNA in injection buffer (total volume 20 μL) 480 ng crRNA in injection buffer (total volume of 20 μL) HDR oligonucleotide 5 ng/μL  4  20 μl ¼ 400 ng or ssDNA: 100 ng/μL  4  20 μL ¼ 8000 ng

tracrRNA: 15.5 ng/μL  4  20 μL ¼ 1240 ng

400 ng HDR oligonucleotide or 8000 ng ssDNA in injection buffer (total volume 20 μL)

1240 ng tracrRNA in injection buffer (total volume 20 μL)

7. To avoid more than two freeze-thaw cycles, prepare aliquots. 8. Store all aliquots at 80  C. 3.3 Pipetting Procedure on the Day of Injection

1. Injection buffer: l Aliquot the amount of injection buffer you need to prepare the injection mix in a 1.5 mL tube. l

Heat to 70  C for 1–2 min.

l

Let it cool down to room temperature.

2. HDR oligonucleotide or ssDNA: l If an HDR oligonucleotide or ssDNA is co-injected, dissolve it on the day of injection in injection buffer to a final concentration of 1000 ng/μL. l

Incubate for ~20 min at 50  C.

l

Spin the tube shortly to ensure that the solution is collected at the bottom.

l

To avoid more than one freeze-thaw cycle, prepare aliquots.

l

Store all tubes at 20  C.

222

Alexander Klimke et al.

3. crRNA and tracrRNA: l Take the aliquots out of the 80  C freezer, and store them on ice before use. l

Store used RNA aliquot immediately after use at 80  C.

4. Cas9 protein: store at 20  C. 3.3.1 Pipetting Procedure for One crRNA

1. Prepare a 1.5 mL tube for each pre-dilution. For proper handling prepare the pre-dilutions at room temperature. 2. In general, the concentrations of the nucleotide pre-dilutions have to be verified by Nanodrop measurement after pipetting. The concentration of the Cas9 protein pre-dilution is too low to be measured by Nanodrop. It is, therefore, necessary to ensure accurate pipetting and visual control of the volume pipetted. 3. crRNA pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Add the crRNA.

l

Mix by pipetting up and down 6–8 times. Measure concentration by Nanodrop; adjust if necessary.

4. tracrRNA pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. 5. Add the tracrRNA: l Mix by pipetting up and down 6–8 times. Measure concentration by Nanodrop; adjust if necessary. 6. In a new 1.5 mL tube mix: l 13 μL tracrRNA pre-dilution. l

Mix by pipetting up and down 6–8 times.

l

Incubate at room temperature for 10 min.

7. In the meantime, prepare Cas9 protein pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Add the Cas9 protein.

l

Mix by pipetting up and down 6–8 times.

8. To the tube containing the 26 μL crRNA-tracrRNA pre-dilution: l Add 13 μL of the Cas9 protein pre-dilution. l

Mix by pipetting up and down 6–8 times.

l

Incubate at 37  C for 15 min.

9. Prepare HDR oligonucleotide or ssDNA pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Add the HDR oligonucleotide or ssDNA.

Mouse Model Generation using CRISPR/Cas9

223

l

Mix by pipetting up and down 6–8 times.

l

Measure concentration by Nanodrop; adjust if necessary.

10. After 15 min incubation of the Cas9 protein-crRNA-tracrRNA mix, place the tube in a rack kept at room temperature, and prepare the final mix. 11. To the tube containing 39 μL Cas9 protein-crRNA-tracrRNA mix: l Add 13 μL of the HDR oligonucleotide or ssDNA pre-dilution. l

Mix by pipetting up and down 6–8 times.

12. Spin the final mix at room temperature for 2 min at 16,000  g. 13. Place it on ice until the injection. 3.3.2 Pipetting Procedure for Two crRNAs (see Note 4.1.8)

1. Prepare a 1.5 mL tube for each pre-dilution. For proper handling prepare the pre-dilutions at room temperature. 2. In general, the concentrations of the nucleotide pre-dilutions have to be verified by Nanodrop measurement after pipetting. Use a volume of 1.3 μL for measurement. Use separately kept filter tips for measurement. 3. The concentration of the Cas9 protein pre-dilution is too low to be measured by Nanodrop. It is, therefore, necessary to ensure accurate pipetting and visual control of the pipetted volume. 4. crRNA1 pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Add the crRNA1.

l

Mix by pipetting up and down 6–8 times.

l

Measure concentration by Nanodrop.

5. crRNA2 pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Add the crRNA2.

l

Mix by pipetting up and down 6–8 times.

l

Measure concentration by Nanodrop.

6. tracrRNA pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Mix by pipetting up and down 6–8 times.

l

Measure concentration; adjust if necessary.

7. In a new 1.5 mL tube, prepare the crRNA1-tracrRNA mix: l 13 μL crRNA1 pre-dilution. l

6.5 μL tracrRNA pre-dilution.

224

Alexander Klimke et al. l

Mix by pipetting up and down 6–8 times.

l

Incubate at room temperature for 10 min.

8. In a new 1.5 mL tube, prepare the crRNA2-tracrRNA mix: l 13 μL crRNA2 pre-dilution. l

6.5 μL tracrRNA pre-dilution.

l

Mix by pipetting up and down 6–8 times.

l

Incubate at room temperature for 10 min.

9. In the meantime, prepare Cas9 protein pre-dilution: l Pipette the injection buffer in a 1.5 mL tube. l

Add the Cas9 protein.

l

Mix by pipetting up and down 6–8 times.

10. Cas9 protein-crRNA1-tracrRNA mix: l Add 6.5 μL of the Cas9 protein pre-dilution to the tube containing 19.5 μL crRNA1-tracrRNA mix. l

Mix by pipetting up and down 6–8 times.

11. Cas9 protein-crRNA2-tracrRNA mix: l Add 6.5 μL Cas9 protein pre-dilution to the tube containing 19.5 μL crRNA2-tracrRNA mix. l

Mix by pipetting up and down 6–8 times.

12. Incubate both mixes at 37  C for 15 min. 13. After 15 min incubation of the Cas9 protein-crRNA-tracrRNA mixes, place the tubes in a tube rack kept at room temperature, and prepare the final mix: l Add 26 μL Cas9 protein-crRNA2-tracrRNA mix to the tube containing 26 μL Cas9 protein-crRNA1-tracrRNA mix. l

Mix by pipetting up and down 6–8 times.

14. Spin the final mix at room temperature for 2 min at 16,000  g. 15. Place it on ice until the injection. 3.4 Embryo Collection

Briefly, after administration of hormones, superovulated C57BL/ 6NTac females are mated with C5BL/6NTac males. One-cell stage fertilized embryos are isolated from the oviducts at day post-coitum (dpc) 0.5. For microinjection, the one-cell stage embryos are placed in a drop of the M2 medium under mineral oil. A microinjection pipette with an approximate internal diameter of 0.4 micrometers (at the tip) is used to inject the mixed nucleotide preparation into the pronucleus of each embryo. After recovery, 25–35 injected one-cell stage embryos are transferred to one of the oviducts of 0.5 dpc, pseudopregnant NMRI females.

Mouse Model Generation using CRISPR/Cas9

225

3.4.1 Mating of C57BL/6 NTac Mice; 5-Week-Old Females Are Superovulated with

1. PMS: 5 units/mouse at 50 U/mL (100 μL).

3.4.2 Embryo Isolation

1. At dpc 0.5 isolate fertilized eggs from ampulla in 1 mL M2 medium.

2. HCG: 5 units/mouse at 50 U/mL (100 μL).

2. Put eggs for 2–5 min in 300 μL M2 þ 50 μL hyaluronidase. 3.4.3 Embryo Washing

1. Wash the oocytes with at least six drops of 50 μL M2 using a new capillary. 2. Transfer oocytes in 20 μL KSOM drops under oil, and store in the incubator until injection.

3.5 Pronuclear Injection

1. Needles type 0.4 μm are pulled by Sutter instrument Micropipette Puller P-97. Heat: 650; Pull: 115; Velocity 50; Time 115.

3.5.1 Injection Needle Preparation

2. Fill the needle with three μL Cas9/crRNA/tracrRNA mix, diluted in EmbryoMax buffer (Millipore) using an Eppendorf Microloader. 3. Needle must be open; break the needle under microscope before injection.

3.5.2 Injection Process

1. Approx. 160 embryos are injected per session (150 survivors). 2. Injection is done in M2 or hCZB media under oil in Petri dish. 3. Load micro drop with approx. 80 embryos per injection round. 4. Parameters for the use of the FemtoJet. l Start with pc at 100 hPa, ti 0.5, pi 150 hPa. l

Flow to fast, turn down pc until flow ok; clean the needle by “clean function.”

l

Flow to slow, turn pc up.

l

Bad flow pc >200 rebreak needle; 25% in peripheral blood) NSGTM mice survive for 40 weeks or longer, while the majority of humanized NSGTM-SGM3 mice survive for 30 weeks or longer. Engrafted NSGTM-SGM3 cohorts have a slightly higher level of anemia development possibly due to macrophage activation syndrome [7]. The JAX mouse facility has a high health status standard, and all mice are observed regularly. It is important to monitor humanized mice weekly and only enroll healthy and non-anemic mice to studies. 1.2 HSC-Engrafted Mice Bearing PDX Tumors

PDX mouse models enable the study of human tumor biology, identification of therapeutic targets, and preclinical evaluation of cancer drugs [13]. Clinical tumor samples are expanded by implant into NSGTM mice without any in vitro manipulation and only used for PDX studies at a low passage number (see Note 2). The tumor characterization data and standard of care response data in NSGTM mice are available on the Mouse Tumor Biology Database (http:// tumor.informatics.jax.org/mtbwi/index.do). Solid tumor engraftment of mice is commonly performed with a 10–13 gauge trocar needle and plunger set [14]. A volume of approximately 30–40 mm3 of tumor is loaded into the bevel of the needle for injection. The trocar needle is cleaned with alcohol between mice from the same cage, and a new autoclaved needle is used for the next cage. Here we describe an improved trocar method using a

Fig. 1 Human immune cell reconstitution in peripheral blood of hu-NSGTM vs. hu-NSGTM-SGM3 engrafted using the same human HSC donor. 4–18 weeks after HSC injection, the cell counts (cells/μL) of hCD45+ cells (a), B cells (b), T cells (c), and myeloid cells (d) in the peripheral blood were determined by flow cytometry. The corresponding percentage of each indicated cell population is shown in (e–h)

244

Li-Chin Yao et al.

14 gauge  3/4 “disposable needle with 1-cc syringe for subcutaneous PDX implantation. This method is less invasive and enables the selection of a specific injection volume for each tumor model. Using a consistent injection volume minimizes tumor growth variation and ultimately synchronizes the timing of enrollment of mice with growing tumors into studies. Tumor-bearing hu-NSGTM and hu-NSGTM-SGM3 mice enable preclinical investigations of immuno-oncology therapies in mice. We have demonstrated that a wide range of PDX tumors grow in humanized mice without obvious indication of rejection (see Note 3) [15]. Human CD45+ immune cells infiltrate tumor tissues that are present in PDX-bearing humanized mice but not in PDX-bearing control NSGTM mice. In addition, PDX-bearing humanized mice respond to standard-of-care chemotherapeutics and to the checkpoint inhibitor Keytruda (pembrolizumab) [15]. Here we will describe standardized protocols for tumor processing and flow cytometry to examine tumor-infiltrating immune cells. The percentage of immune cell infiltrations in humanized NSGTM were compared to those in NSGTM-SGM3 mice engrafted with the same HSC donor.

2

Materials

2.1 PDX Tumor Preparation and Subcutaneous Implantation

1. PDX tumor donor mice: NSGTM mice engrafted with PDX. 2. HSC-engrafted NSG TM or NSGTM-SGM3 mice. 3. BSL2 cabinet. 4. Personal protective equipment (PPE) including lab coat, mask, safety glasses, and double gloves. 5. Autoclaved forceps and microdissecting scissors. 6. No. 21 disposable scalpels (Feather Safety Razor Co., Japan). 7. Falcon standard tissue culture dishes 150  15 mm (Thermo Fisher Scientific, Waltham, MA, USA), sterile. 8. Cold packs. 9. Anesthesia system: isoflurane vaporizer and attachment (VetEquip, Livermore, CA, USA). 10. Disposable 14 g  3/400 needles (Jorgensen Labs, Loveland, CO, USA). 11. Disposable 1-cc and 5-cc syringes (BD, Franklin lakes, NJ, USA). 12. Sterile alcohol pad (Fisherbrand). 13. Sterile 10 gauge reusable needle and plunger (Cadence Inc., Staunton, VA, USA), if injecting tumor growth supplement such as DHT or estradiol tablets (Innovative Research of America, Sarasota, FL, USA).

PDX-Bearing Humanized Mice

245

2.2 Tumor Dissociation and Flow Cytometry Analysis

1. No. 21 disposable scalpel (Feather safety Razor Co., Japan).

2.2.1 Tumor and Spleen Dissociation Reagents and Equipment

3. Microdissecting scissors: straight, sharp (small).

2. Falcon standard tissue culture dishes 100  20 mm (Thermo Fisher Scientific, Waltham, MA, USA), sterile. 4. RPMI 1640. 5. Falcon 50 mL conical centrifugation tubes (Thermo Fisher Scientific, Waltham, MA, USA). 6. Collagenase P (Roche CustomBiotech, Indianapolis, IN, USA). 7. DNase I recombinant, RNase-free (Thermo Fisher Scientific, Waltham, MA, USA). 8. 37  C water bath. 9. Variable Speed Nutator (VWR International, Radnor, PA, USA). 10. Fetal bovine serum (FBS, Atlanta Biologicals, Lawrenceville, GA, USA). 11. 70-μm nylon cell strainer (Thermo Fisher Scientific, Waltham, MA, USA). 12. Disposable 5-cc syringe (BD, Franklin Lakes, NJ, USA). 13. GentleMACSTM Dissociator (Miltenyi Biotec, Auburn, CA, USA). 14. GentleMACSTM C tubes (Miltenyi Biotec, Auburn, CA, USA).

2.2.2 Fluorescent Antibodies for Checking TILs

1. FACS Buffer: PBS supplemented with 1% FBS and 0.1% sodium azide. 2. Antibodies: Mouse CD45, 30-F11 clone (BD Biosciences, San Jose, CA, USA). Human CD45, 2D1 clone (BioLegend, San Diego, CA, USA). Human CD3, UCHT1 clone (BioLegend, San Diego, CA, USA). Human CD4, RPA-T4 clone (BioLegend, San Diego, CA, USA). Human CD8, RPA-T8 clone (BioLegend, San Diego, CA, USA). Human CD20, 2H7 clone (BioLegend, San Diego, CA, USA). Human CD33, WM53 clone (BioLegend, San Diego, CA, USA). Human CD14, HCD14 clone (BioLegend, San Diego, CA, USA).

246

3

Li-Chin Yao et al.

Methods

3.1 PDX Tissue Preparation and Subcutaneous PDX Implantation (Fig. 2)

All PDX tissue preparation and inoculation should be performed in a BSL2 cabinet using sterile tools and techniques (see Note 4). This cabinet should be decontaminated with 70% ethanol before use. Laboratory personnel should wear proper PPE to handle tumors. 1. Place the solid tumor collected from the donor mouse in a culture dish on a cold pack. Remove any necrotic tissue and cut the tumor into pieces with scalpel, and then mince them with a pair of fine scissors. Load all the finely minced tumor

Fig. 2 PDX tumor engraftment method. (a) Finely minced tumors in a sterile petri dish placed on an ice pack. (b) Comparison of different trocar needles: 10, 13 gauge needles with plunger (left, middle) and a 14  3/400 gauge disposable needle with 1-cc syringe (right). (c) Optimal tumor injection site on the right flank of the mouse

PDX-Bearing Humanized Mice

247

into a 5-cc syringe without air bubbles, and then transfer tumors into a 1 cc syringe. Place the 14 gauge needle on the 1-cc syringe tip. The needle is reused for up to five mice within the same cage. Save some tumor fragment for histological analysis (see Note 5). 2. Shave the implantation site prior to anesthesia. 3. Anesthetize mice by isoflurane inhalation until nonresponsive to toe-pinch. 4. Disinfect implantation site with an ethanol pad. Grasp and lift the skin below the right flank with forceps, and inject minced tumors subcutaneously with desired amount (20–40 μL). 5. For tumor models that are dependent upon hormones to grow (e.g., breast and prostate), mice will be subcutaneously injected with a DHT or estradiol tablet the day before tumor inoculation. Slow-release hormone tables will be implanted subcutaneously between the shoulder blades with a larger 10 gauge needle. 6. Mice are weighed at least twice a week and monitored daily post engraftment (see Note 6). 3.2 Evaluation of Immune Cell Infiltration by Flow Cytometry

All PDX tissue preparation and spleen collection should be performed in a BSL2 cabinet using sterile tools and techniques (see Note 4). This cabinet should be decontaminated with 70% ethanol before use. Personnel should wear proper PPE to handle tumors. For optimal evaluation of human immune cell infiltrates, the tumor tissue must be dissociated into a single-cell suspension and stained following a standard flow cytometry protocol as described below. 1. Using a scalpel, dissect tumor specimen into small pieces, and then mince completely using scissors in a 100-mm plastic tissue culture dish. 2. Using 10 mL of cold RPMI, transfer tumor cell suspension into a 50-mL conical centrifuge tube and centrifuge (400  g for 5 min at 4  C). Warm digest buffer (1 mg/mL collagenase P and 10 μL/mL DNase I in RPMI 1640) to 37  C in water bath (see Note 7). 3. Discard supernatant and estimate volume of tumor tissue within conical tube. Add digest buffer to achieve a 1:1 volume ratio with the estimated tumor volume (1 mL of digest buffer per 1 mL of tumor tissue). Place the 50-mL conical centrifuge tube with digesting tumor tissue on a Nutator in a 37  C incubator for 30 min (see Note 8). 4. Quench digestion by adding an equal volume (1:1 volume ratio to tumor sample volume) of cold RPMI supplemented with 2% FBS. Filter the cell suspension through a 70-μm nylon mesh cell strainer, and use the plunger of a 5-mL syringe to

248

Li-Chin Yao et al.

homogenize the tissue through the mesh 50-mL conical centrifuge tube. 5. Centrifuge the flow through cell suspension (400  g for 5 min at 4  C), and then decant supernatant. Wash cell pellet twice with 20 mL of cold RPMI supplemented with 2% FBS. After final wash, resuspend cell pellet in 5–10 mL of cold RPMI supplemented with 2% FBS and perform viability count. 6. For splenic tissue recovered from mice, prepare single-cell suspension of spleens by mechanical dissociation and perform viability count. 7. For staining of samples (tumor or spleen) with fluorescently conjugated antibodies, aliquot 1  106 cells into a tube or 96-well plates, and perform FACS analyses as previously described [16]. The results shown below compare different immune cell populations in spleens and melanoma PDX tumor PS4050 collected from hu-NSGTM and hu-NSGTM-SGM3 engrafted with the same HSC donor (Fig. 3a, b). Tissues were collected at 16 weeks post HSC engraftment and stained with mCD45, hCD45, hCD3, hCD20, hCD33, and hCD14 for flow analysis. Melanoma PDX showed more human CD45+ leukocyte infiltrates in hu-NSGTM-SGM3 than in hu-NSGTM . The proportion of CD3+T cells in the tumor was similar between two strains. The majority of TILs in hu-NSGTM-SGM3 mice were CD33+ myeloid cells, and about 50% of them were CD14+ macrophages. CD20+ B cells were most abundant in the tumor of hu-NSGTM mice. Similarly, higher levels of hCD45+ were detected in the spleen of hu-NSGTM-SGM3 mice. The proportion of CD33+ cells in the spleen was similar between two strains. The majority of immune cells in hu-NSGTM mice are CD3+ T cells, while CD20+ B cells are most abundant in the spleen of hu-NSGTM-SGM3 mice. A representative tumor flow profile data is shown in Fig. 3c.

4

Notes 1. The irradiation dose should be optimized based on the specific strain. The engraftment of NSGTM-SGM3 mice is very robust even when preconditioned at a lower dose when compared to NSGTM mice. For the experiments described above, NSG TM -SGM3 mice received 100 cGy and NSGTM mice received 140 cGy. 2. Early-passage PDX models retain the molecular characteristics of the corresponding patient tumor at both the genomic and

Fig. 3 Human immune cell infiltrations in PS4050-bearing hu-NSGTM or huNSGTM-SGM3 mice engrafted with the same human HSC donor. Tumors and spleens were processed into single-cell suspensions and stained for human CD45, CD3, CD20, CD33, and CD14. The percentage of each population was determined by flow cytometric analysis (a, b). Human CD3+, CD20+, CD33+, and CD33+CD14+ are presented as percentage of cells within hCD45+ cells. n ¼ 6 mice/group. Data were analyzed and graphs were prepared with GraphPad Prism 5. Differences between strains were assessed by unpaired student t test. (c) Representative flow data for cells expressing human CD45, CD3, CD20, CD33, CD14, and HLA-DR are shown for tumor-infiltrating immune cells recovered from a tumor growing in a hu-NSGTM-SGM3 mouse

250

Li-Chin Yao et al.

gene expression levels [17]. Each passage we confirm as human samples by immunohistochemically staining with a monoclonal antibody specific for human Ki67. Approximately 400 tumors in JAX’s PDX collection are available for research community (https://www.jax.org/jax-mice-and-services/in-vivo-pharma cology/oncology-services/pdx-tumors). 3. An important caveat for the use of HSC-engrafted NSGTM and NSGTM-SGM3 mice in immune-oncology studies is that the developing human immune systems and tumors are not genetically matched. Moreover, the majority of T cells developing in these models are educated in mouse thymus and are selected on mouse MHC [18]. Therefore, the growing tumor in most of these experimental models will be recognized as an allo- or xenograft by the immune response, unless the tumor and HSC are HLA matched and recipient mouse also expresses matched HLA molecules for human T cell development. 4. NSGTM mice are severely immunodeficient and susceptible to infections. They should be housed in pathogen-free condition and handled with aseptic techniques. They can become infected from PDX injection sites and any insult that punctures the skin. 5. Some reports have shown unanticipated formation of EpsteinBarr virus (EBV)-associated B cell lymphomas in PDXs [19, 20]. Reactivation of EBV residing in the PDX may develop unwanted growth of lymphoma after engrafting into immunodeficient mice. PDX tumors should be validated by frequently confirming the absence of CD45 cells and the presence of pan-cytokeratin staining. 6. Some aggressive tumors tend to become ulcerated. If any signs of tumor erosion or broken skin are observed, we consult a veterinarian and proactively apply antiseptic betadine or chlorhexidine to tumors. This protocol may help to prevent tumor ulceration. Mice will be euthanized if tumors ulcerate. 7. The detection of some cell surface molecules by flow cytometry will be sensitive to digestion protocols. The ability to detect specific markers should be confirmed following exposure of cells to the digestion buffers before conducting experiments. 8. As an alternative to the use of a Nutator for the digestion process, we use a gentleMACSTM Dissociator from Miltenyi Biotec. Dissociated tumor tissue is mixed with digestion buffer and added to gentleMACSTM C Tubes. The Tubes are then loaded on the Dissociator according to the manufacturers’ protocol.

PDX-Bearing Humanized Mice

251

Acknowledgments We thank Justin Jarvi and Marta Tewodros for trocar method development and technical assistance. This work was supported by The Jackson Laboratory. References 1. Shultz LD, Ishikawa F, Greiner DL (2007) Humanized mice in translational biomedical research. Nat Rev Immunol 7(2):118–130. https://doi.org/10.1038/nri2017 2. Tanaka S, Saito Y, Kunisawa J et al (2012) Development of mature and functional human myeloid subsets in hematopoietic stem cell-engrafted NOD/SCID/IL2rgammaKO mice. J Immunol 188(12):6145–6155. https://doi.org/10.4049/jimmunol. 1103660 3. Ishikawa F, Yasukawa M, Lyons B et al (2005) Development of functional human blood and immune systems in NOD/SCID/IL2 receptor {gamma} chain(null) mice. Blood 106 (5):1565–1573. https://doi.org/10.1182/ blood-2005-02-0516 4. Shultz LD, Lyons BL, Burzenski LM et al (2005) Human lymphoid and myeloid cell development in NOD/LtSz-scid IL2R gamma null mice engrafted with mobilized human hemopoietic stem cells. J Immunol 174(10):6477–6489 5. Brehm MA, Bortell R, Diiorio P et al (2010) Human immune system development and rejection of human islet allografts in spontaneously diabetic NOD-Rag1null IL2rgammanull Ins2Akita mice. Diabetes 59(9):2265–2270. https://doi.org/10.2337/db10-0323 6. Rongvaux A, Takizawa H, Strowig T et al (2013) Human hemato-lymphoid system mice: current use and future potential for medicine. Annu Rev Immunol 31:635–674. https://doi.org/10.1146/annurev-immunol032712-095921 7. Wunderlich M, Chou FS, Link KA et al (2010) AML xenograft efficiency is significantly improved in NOD/SCID-IL2RG mice constitutively expressing human SCF, GM-CSF and IL-3. Leukemia 24(10):1785–1788. https:// doi.org/10.1038/leu.2010.158 8. Billerbeck BWT, Mu K et al (2011) Development of human CD4þFoxP3þ regulatory T cells in human stem cell factor-, granulocytemacrophage colony-stimulating factor-, and interleukin-3-expressing NOD-SCID IL2Rgamma(null) humanized mice. Blood 117

(11):3076–3086. https://doi.org/10.1182/ blood-2010-08-301507 9. Jangalwe S, Shultz LD, Mathew A et al (2016) Improved B cell development in humanized NOD-scid IL2Rgamma(null) mice transgenically expressing human stem cell factor, granulocyte-macrophage colony-stimulating factor and interleukin-3. Immunity, inflammation and disease 4(4):427–440. https://doi. org/10.1002/iid3.124 10. Coughlan AM, Harmon C, Whelan S et al (2016) Myeloid engraftment in humanized mice: impact of granulocyte-colony stimulating factor treatment and transgenic mouse strain. Stem Cells Dev 25(7):530–541. https://doi. org/10.1089/scd.2015.0289 11. Miller PH, Cheung AM, Beer PA et al (2013) Enhanced normal short-term human myelopoiesis in mice engineered to express humanspecific myeloid growth factors. Blood 121(5): e1–e4. https://doi.org/10.1182/blood2012-09-456566 12. Hasgur S, Aryee KE, Shultz LD et al (2016) Generation of immunodeficient mice bearing human immune systems by the engraftment of hematopoietic stem cells. Methods Mol Biol 1438:67–78. https://doi.org/10.1007/9781-4939-3661-8_4 13. Lodhia KA, Hadley AM, Haluska P et al (2015) Prioritizing therapeutic targets using patientderived xenograft models. Biochim Biophys Acta 1855(2):223–234. https://doi.org/10. 1016/j.bbcan.2015.03.002 14. Dosch J, Ziemke E, Wan S et al (2017) Targeting ADAM17 inhibits human colorectal adenocarcinoma progression and tumor-initiating cell frequency. Oncotarget 8 (39):65090–65099. https://doi.org/10. 18632/oncotarget.17780 15. Wang M, Yao LC, Cheng M et al (2018) Humanized mice in studying efficacy and mechanisms of PD-1-targeted cancer immunotherapy. FASEB J Mar 32(3):1537–1549. https://doi.org/10.1096/fj.201700740R 16. Brehm MA, Cuthbert A, Yang C et al (2010) Parameters for establishing humanized mouse models to study human immunity: analysis of

252

Li-Chin Yao et al.

human hematopoietic stem cell engraftment in three immunodeficient strains of mice bearing the IL2rgamma(null) mutation. Clin Immunol 135(1):84–98. https://doi.org/10.1016/j. clim.2009.12.008 17. Whittle JR, Lewis MT, Lindeman GJ et al (2015) Patient-derived xenograft models of breast cancer and their predictive power. Breast Cancer Res 17:17. https://doi.org/10.1186/ s13058-015-0523-1 18. Brehm MA, Shultz LD, Luban J et al (2013) Overcoming current limitations in humanized mouse research. J Infect Dis 208(Suppl 2):

S125–S130. https://doi.org/10.1093/ infdis/jit319 19. John T, Yanagawa N, Kohler D et al (2012) Characterization of lymphomas developing in immunodeficient mice implanted with primary human non-small cell lung cancer. J Thorac Oncol 7(7):1101–1108. https://doi.org/10. 1097/JTO.0b013e3182519d4d 20. Chen K, Ahmed S, Adeyi O et al (2012) Human solid tumor xenografts in immunodeficient mice are vulnerable to lymphomagenesis associated with Epstein-Barr virus. PLoS One 7 (6):e39294. https://doi.org/10.1371/jour nal.pone.0039294

Chapter 16 Immunophenotyping of Tissue Samples Using Multicolor Flow Cytometry Martina M. Sykora and Markus Reschke Abstract Flow cytometry enables the measurement of single cells in a flowing system. Heterogeneous mixtures of cells or particles can be analyzed with respect to their morphology, surface and intracellular protein expression, DNA content, and cellular physiology at high speed and purity. A series of key technical developments and improvements in flow cytometry hardware, software, and dye chemistry made it possible to measure more than 20 parameters simultaneously. Here, we provide a stepwise protocol for the preparation of single cell suspension samples from different murine lymphoid or tumor tissues and a detailed description of a 17-color polychromatic flow cytometry analysis of tumor-infiltrating leukocytes. Key words Multicolor flow cytometry, Immunophenotyping, Murine mouse tissue, Single cell preparation, Surface marker staining, Intracellular staining, Compensation, FMO

1

Introduction Flow cytometry is a powerful tool for the analysis of immune cells on a single cell level. It enables the measurement of a number of different parameters, as cells cross the beams of laser light. In standard flow cytometers, detectors measure scattered light and fluorescence, emitted from cells or particles. Due to advances in protein detection, fluorophore chemistry, and flow cytometry instrumentation, it is now possible to simultaneously determine the levels of more than 20 surface and intracellular markers [1, 2]. Hence, polychromatic flow cytometry enables the detection of numerous cell types in a complex mixture at a reasonable throughput and in a cost-effective way, allowing for the comparison of multiple samples in a single experiment. This is especially important in the context of preclinical studies, when double or triple combinations of drugs are tested and the necessity of control groups can lead to a significant scale-up of an experiment.

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_16, © Springer Science+Business Media, LLC, part of Springer Nature 2019

253

254

Martina M. Sykora and Markus Reschke

Even though other methods like mass cytometry (CyTOF), single cell sequencing, or very recently CITEseq can provide a more detailed picture of the composition of immune cells in different diseases [2–4], it is difficult to implement those techniques in a routine workflow, due to the low throughput. However, especially in the field of immuno-oncology, it became apparent that changes in small subtypes of immune cells can be essential for effective therapy [5]. Often, combinations of numerous markers are required to clearly identify those cells, and thus sophisticated multicolor flow analyses are at the forefront of today’s oncology research. Unfortunately, the increase of information generated by multicolor flow cytometry due to the aforementioned complexity comes at a cost. As the number of fluorochromes increases, the cross-talk between the different fluorescence channels multiplies. The resulting detector spillover and the required compensation lead to data spreading and reduced resolution. For these reasons, proper panel design, usage of the right controls, and instrument settings become increasingly important [6, 7]. Here, we describe a state-of-the-art protocol for tumor, peripheral lymphoid organ, and peripheral blood processing with a stepby-step workflow. Moreover, we provide guidelines how to include important controls and how to stain key markers on the cell surface or intracellular targets, together with troubleshooting for compensation challenges in multicolor flow cytometry. This detailed staining and analysis procedure is tailored in particular to single cell suspensions from mouse tissue, but it can also be adapted to other specimen including patient samples. Our chosen example follows a 17-color polychromatic flow cytometry analysis, to identify subsets of immune cells within a complex tumor sample. An overview of key literature and technical notes are brought into a critical view to provide the reader with detailed insight into the power and potential pitfalls of multicolor flow cytometry.

2 2.1

Materials Lab Equipment

1. Cell counter (e.g., Vi-Cell XR cell viability analyzer, Beckman Coulter). 2. Refrigerated centrifuge. 3. GentleMACS Dissociator, Miltenyi Biotec. 4. MACS MultiStand, Miltenyi Biotec. 5. MiniMACS Separator, Miltenyi Biotec. 6. MidiMACS Separator, Miltenyi Biotec. 7. Flow Cytometer, e.g., BD LSR Fortessa (Becton Dickinson) or CytoFLEX LX (Beckman Coulter).

Multicolour Flow Cytometry

2.2 Tissue Dissociation

255

1. Tumor Dissociation Kit, mouse, Miltenyi Biotec. 2. Spleen Dissociation Kit, mouse, Miltenyi Biotec. 3. C-tubes, Miltenyi Biotec. 4. ACK lysis buffer composed of 150 mM ammonium chloride, 10 mM potassium bicarbonate, and 0.1 mM (EDTA) ethylenediaminetetraacetic acid. 5. 70 μm cell strainer. 6. PBS (phosphate buffered saline), 2% FCS (fetal calf serum).

2.3 TumorInfiltrating Leukocytes (TIL) Isolation

1. CD45 (TIL) MicroBeads, mouse, Miltenyi Biotec. 2. MS Columns, Miltenyi Biotec. 3. LS Columns, Miltenyi Biotec. 4. 30-μm cell strainer. 5. PBS, 2% FCS.

2.4 Preparation of Samples and Staining

1. PBS, 5 mM MgCl2 (magnesium chloride). 2. PBS, 5 mM MgCl2, 100 μg/mL DNase I (deoxyribonuclease I). 3. TruStain fcX™ (anti-mouse CD16/32) Antibody, BioLegend. 4. FACS (fluorescence-activated cell sorting) buffer: PBS, 2% FCS, 0.01% NaN3 (sodium azide), and 5 mM EDTA. 5. Brilliant Stain Buffer, BD Horizon. 6. Transcription Factor Buffer Set, BD Pharmingen. 7. Antibody capture beads/compensation beads (e.g., VersaComp Antibody Capture Beads, Beckman Coulter). 8. Fixable viability dye, e.g., Fixable Viability Stain 700, BD Horizon. 9. Fluorochrome-conjugated antibodies against markers of interest.

3

Methods

3.1 Preparation of Single Cell Suspensions from Murine Tissues

Samples should be kept cold on ice or at 4
C all the time, and buffers should be chilled before use, if not specified otherwise.

3.1.1 Tumor (Subcutaneous Murine Tumors from Different Tissue Origins)

1. Prepare enzyme mix (Tumor Dissociation Kit) as described by the manufacturer (2.35 mL DMEM (Dulbecco’s Modified Eagle Medium), 100 μL enzyme D, 50 μL enzyme R, and 12.5 μL enzyme A per tumor).

FCS must be heat inactivated at 56
C for 30 min prior to be used in flow cytometry procedures.

256

Martina M. Sykora and Markus Reschke

2. Cut tumors into small pieces (approximately 1–2 mm) using a scalpel, and transfer it together with the enzyme mix into a C-tube. 3. Close tube tightly and attach it upside down onto the sleeve of the gentleMACS Dissociator. 4. Since tumor dissociation programs are usually carried out at 37
C, set heaters onto the desired positions. 5. Run the appropriate gentleMACS mouse tumor program: 37C_m_TDK_1 for the dissociation of soft tumors and program 37C_m_TDK_1 for tough tumors. 6. To mince the remaining tumor pieces, run program m_impTumor-01. 7. After termination of the program, detach C-tube from the dissociator, perform a short spin at a maximum of 400  g, and filter cell suspension through a 70 μm cell strainer placed on a 15 mL tube. 8. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away supernatant (see Note 1). 9. To remove red blood cells, resuspend the cell pellet in 1 to 2 mL of ACK lysis buffer, and incubate the cell suspension for 2 min. 10. Add 10 mL of PBS containing 2% FCS to stop the reaction. 11. Filter cell suspension through a 70 μm filter placed on a 15 mL tube. 12. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away supernatant. 13. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells. Isolation of TILs from Tumor Cell Suspensions (Optional)

1. Centrifuge cell suspension at 400  g for 5 min and aspirate supernatant. 2. Resuspend cell pellet in PBS containing 2% FCS. Use 90 μL buffer for up to 1  107 total cells (for higher cell numbers, scale up buffer volume). 3. Add 10 μL TIL (CD45) MicroBeads per 1  107 cells, mix thoroughly, and incubate suspension for 15 min in the fridge. 4. Add PBS with 2% FCS to reach a final volume of 500 μL per 2  107 cells or 1 mL for up to 5  107 cells. For higher cell numbers, scale up volume accordingly, and split samples onto multiple columns. 5. Place columns on MACS separator. Use MS columns for up to 2  107 cells and LS columns for a maximum of 5  107 cells.

Multicolour Flow Cytometry

257

6. Prepare the column by rinsing it with either 500 μL (MS column) or 3 mL (LS column) PBS containing 2% FCS. 7. Place a 30 μm cell strainer onto the column and apply cell suspension. 8. Collect the flow through containing the unlabeled tumor cells if needed (see Note 2). 9. Wait until the column reservoir is empty. 10. Wash column either three times with 500 μL (MS column) or twice with 1 mL (LS column) PBS containing 2% FCS. 11. Remove the column from the separator and place it onto a 15 mL tube. 12. Add either 1 mL (MS column) or 3 mL (LS column) buffer into the column reservoir. 13. Immediately push through the fluid, using the plunger that was provided with the column. 14. Centrifuge the TIL containing cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant. 15. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells. 3.1.2 Lymph Node

1. Prepare enzyme mix (Spleen Dissociation Kit) as described by the manufacturer (2.4 mL buffer S, 50 μL enzyme D, and 15 μL enzyme A per lymph node). 2. Transfer the lymph node together with the enzyme mix into a C-tube. 3. Close tube tightly and attach it upside down onto the sleeve of the gentleMACS Dissociator. 4. Set the heaters onto the desired positions. 5. Run the program 37C_m_SDK_1. 6. After termination of the program, detach C-tube from the dissociator, perform a short spin at a maximum of 400  g, and filter cell suspension through a 70-μm cell strainer placed on a 15 mL tube. 7. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant. 8. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells.

258

Martina M. Sykora and Markus Reschke

If lymphocytes will be analyzed, a single cell suspension from lymph nodes can be prepared as follows: 1. Transfer lymph node onto a pre-wetted 70 μm cell strainer sitting either on a 15 mL tube or in a sterile petri dish. 2. Homogenize the tissue using the plunger of a 1 mL syringe. 3. Rinse the tissue thoroughly with PBS containing 2% FCS. 4. Transfer the cell suspension into a 15 mL tube, and spin at 400  g for 5 min before removing the supernatant. 5. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells. 3.1.3 Spleen

1. Prepare enzyme mix (Spleen Dissociation Kit) as described by the manufacturer (2.4 mL buffer S, 50 μL enzyme D, and 15 μL enzyme A per spleen). 2. Cut spleen into small pieces using a scalpel (approximately 4 mm), and transfer it together with the enzyme mix into a C-tube. 3. Close tube tightly and attach it upside down onto the sleeve of the gentleMACS Dissociator. 4. Put the heaters onto the desired positions. 5. Run the program 37C_m_SDK_1. 6. After termination of the program, detach C-tube from the dissociator, perform a short spin at a maximum of 400  g, and filter cell suspension through a 70 μm cell strainer placed on a 15 mL tube. 7. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant. 8. To remove red blood cells, resuspend the cell pellet in 1–2 mL of ACK lysis buffer, and incubate the cell suspension for 2 min. 9. Add 10 mL of PBS containing 2% FCS to stop the reaction. 10. Filter cell suspension through a 70 μm filter placed on a 15 mL tube. 11. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant. 12. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells. If lymphocytes will be analyzed, a single cell suspension from the spleen can be prepared as follows: 1. Transfer the spleen onto a pre-wetted 70 μm cell strainer sitting either on a 15 mL tube or in a sterile petri dish.

Multicolour Flow Cytometry

259

2. Homogenize the tissue using the plunger of a 1 mL syringe. 3. Rinse the tissue thoroughly. 4. Transfer the cell suspension into a 15 mL tube, and spin at 400  g for 5 min before removing the supernatant. 5. Proceed with step 8 from previous section (removal of red blood cells). 3.1.4 Peripheral Blood

1. Transfer blood samples (treated with an anticoagulant, e.g., EDTA) into a 15 mL tube. 2. Centrifuge at 400  g for 5 min and aspirate supernatant (see Note 3). 3. To remove red blood cells, resuspend the cell pellet in 1 to 2 mL of ACK lysis buffer, and incubate the cell suspension for 2 min. 4. Add 10 mL of PBS containing 2% FCS to stop the reaction. 5. Filter cell suspension through a 70 μm filter placed on a 15 mL tube. 6. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant. 7. If pellet is still reddish, repeat steps 3–6. 8. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells.

3.1.5 Bone Marrow

1. Free the bones from residual tissue (using a paper towel to rub off the tissue from the bones). 2. Transfer the bones into a mortar and add 5 mL PBS containing 2% FCS. 3. Grind the bones with a pestle and wash out the hematopoietic cells using a 5 mL pipette. 4. Filter cell suspension through a 70 μm filter placed on a 15 mL tube. 5. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant. 6. To remove red blood cells, resuspend the cell pellet in 1 to 2 mL of ACK lysis buffer, and incubate the cell suspension for 2 min. 7. Add 10 mL of PBS containing 2% FCS to stop the reaction. 8. Filter cell suspension through a 70 μm filter placed on a 15 mL tube. 9. Centrifuge cell suspension at 400  g for 5 min, and aspirate or pour away the supernatant.

260

Martina M. Sykora and Markus Reschke

10. Resuspend cell pellet in 1 mL PBS containing 2% FCS, and count cells using a Vi-Cell XR cell viability analyzer or a comparable method to assess the number of living cells. 3.2

DNase Digest

1. Wash cells once by resuspending cell pellets in 1 mL PBS containing 5 mM MgCl2 followed by a centrifugation at 400  g for 5 min. 2. Aspirate or pour away the supernatant. 3. Take pellets up in 0.1 mL buffer (PBS containing 5 mM MgCl2 and 100 μg/mL DNase I) per 1  106 cells. 4. Incubate at RT (room temperature) for 10 min.

3.3 Assessment of Cell Viability

1. Transfer approximately 1  106 cells per well into a 96-well plate. (If the population of interest is small, make sure to use enough cells. Adjust staining volume accordingly.) 2. Wash cells once with PBS. 3. Centrifuge samples at 400  g for 5 min, and discard supernatant by forcefully inverting the plate. 4. Dilute fixable viability dye in PBS, and protect dilutions from light exposure: FVS (fixable viability stain) 700 (1:10,000)

5. From this step on, samples have to be kept in the dark as much as possible (see Note 4). 6. Resuspend pellets in 0.1 mL fixable viability dye solution, and mix thoroughly by pipetting. 7. Incubate for 30 min. 8. Wash cells twice by resuspending cell pellets in 0.2 mL PBS followed by a centrifugation at 400  g for 5 min. Discard supernatant by forcefully inverting the plate. 3.4 Fc Receptor Blockade

1. Resuspend cell pellets in FACS buffer containing 1:100 diluted TruStain fcX (see Note 5). 2. Incubate for 10 min. 3. Centrifuge samples at 400  g for 5 min, and discard supernatant by forcefully inverting the plate.

3.5 Surface Marker Staining

1. Prepare surface marker staining panels in either FACS buffer or Brilliant Stain Buffer (see Note 6). A representative panel scheme is shown below.

Multicolour Flow Cytometry

261

Anti-CD4 BUV (brilliant ultraviolet) 395 Anti-B220 BUV 737 Anti-MHCII eFluor 450 Anti-Ly6G BV (brilliant violet) 510 Anti-CD11c BV 605 Anti-LAG3 BV 650 Anti-F4/80 BV 711 Anti-Thy1.2 BV 785 Anti-CD45 FITC (fluorescein isothiocyanate) Anti-CD11b PerCP-Cy5.5 (peridinin chlorophyll protein complex-cyanine 5.5) Anti-CD103 PE (phycoerythrin) Anti-PD-1 PE-CF594 (phycoerythrin-CF594) Anti-CD8a PE-Cy5 (phycoerythrin-cyanine 5) Anti-PDCA-1 PE-Cy7 (phycoerythrin-cyanine 7) Anti-Ly6C APC-Cy7 (allophycocyanin-cyanine 7)

2. Resuspend pellets in 50–100 μL antibody mixture, and mix thoroughly by pipetting. 3. Incubate samples for 30 min. 4. Wash cells twice by resuspending cell pellets in 0.2 mL FACS buffer followed by a centrifugation at 400  g for 5 min. Discard supernatant by forcefully inverting the plate. 5. If cells are stained for surface markers only, samples can be acquired immediately. Therefore, take cells up in desired amount of FACS buffer, and acquire samples using a flow cytometer. Alternatively, cells can be fixed (Subheading 3.6, steps 1 and 2) and stored in FACS buffer and protected from light at 4–8
C until acquisition. 3.6 Intracellular Marker Staining

1. Prepare Fixation/Permeabilization (Fix/Perm) buffer according to the manufacturer’s protocol (dilute 4 fixation/permeabilization buffer with diluent buffer). 2. Resuspend pellets in 100 μL Fix/Perm buffer, and incubate samples for 30 min in the fridge protected from light. 3. Prepare Perm/Wash buffer according to the manufacturer’s protocol (dilute 5 Perm/Wash buffer with deionized water). 4. Wash cells twice by resuspending cell pellets in 0.2 mL Perm/ Wash buffer followed by a centrifugation at 400  g for 5 min. Discard supernatant by forcefully inverting the plate.

262

Martina M. Sykora and Markus Reschke

5. Dilute antibodies detecting intracellular proteins in Perm/ Wash buffer (e.g., anti-Foxp3-APC). 6. Resuspend pellets in 50–100 μL antibody mixture, and mix thoroughly by pipetting. 7. Incubate samples for 30 min. 8. Wash cells twice by resuspending cell pellets in 0.2 mL Perm/ Wash buffer followed by a centrifugation at 400  g for 5 min. Discard supernatant by forcefully inverting the plate. 9. Wash cells once by resuspending cell pellets in 0.2 mL FACS buffer followed by a centrifugation at 400  g for 5 min. Discard supernatant by forcefully inverting the plate. 10. Take cells up in desired amount of FACS buffer, and acquire samples on a flow cytometer. 3.7 Controls (See Note 7)

1. Prepare control samples as described for staining samples until antibody staining (Subheading 3.5).

3.7.1 Fluorescence Minus One (FMO)

2. Prepare FMOs by adding all surface markers except one to either FACS buffer or Brilliant Stain Buffer. A representative scheme for the CD103-PE FMO is shown below. Anti-CD4 BUV395 Anti-B220 BUV737 Anti-MHCII eFluor 450 Anti-Ly6G BV510 Anti-CD11c BV605 Anti-LAG3 BV650 Anti-F4/80 BV711 Anti-Thy1.2 BV785 Anti-CD45 FITC Anti-CD11b PerCP-Cy5.5 No anti-CD103 PE Anti-PD-1 PE-CF594 Anti-CD8a PE-Cy5 Anti-PDCA-1 PE-Cy7 Anti-Ly6C APC-Cy7

3. In addition, an FMO plus isotype control can be prepared (see Note 7). A representative scheme for the CD103-PE FMO plus isotype is shown below. Make sure that the amount of

Multicolour Flow Cytometry

263

isotype antibody equals the amount of staining antibody used in the panel (see Note 8). Anti-CD4 BUV395 Anti-B220 BUV737 Anti-MHCII eFluor 450 Anti-Ly6G BV510 Anti-CD11c BV605 Anti-LAG3 BV650 Anti-F4/80 BV711 Anti-Thy1.2 BV785 Anti-CD45 FITC Anti-CD11b PerCP-Cy5.5 Isotype control PE Anti-PD-1 PE-CF594 Anti-CD8a PE-Cy5 Anti-PDCA-1 PE-Cy7 Anti-Ly6C APC-Cy7

4. Proceed with the protocol as described above (Subheading 3.5, step 2). 5. If intracellular staining will be performed, stain all surface markers first, and leave out the antibodies against intracellular proteins at Subheading 3.6, step 5, or replace them with isotype controls (see Note 8). 3.8

Compensation

1. Prepare single stains of each antibody with either cells or beads (see Note 9 and reference [7]). 2. Treat compensation samples like staining samples, because fixation, light exposure, etc. might have an influence on fluorophores. This is especially important for tandem dyes. Ideally prepare compensation controls together with the staining samples. 3. For the viability dye single stain control, label dead cells with the dye used in the experiment (see Note 10). 4. Finally add either unstained beads (see Note 11) or cells to the sample. 5. Acquire compensation samples with the same gain settings used in the experiment. If the positive population is off scale or out of channel linearity, dilute your staining reagent until the negative and positive population are in the accepted range.

264

Martina M. Sykora and Markus Reschke

6. Compute the spillover matrix with an analysis software like BD FACSDIVA™ or FlowJo. Alternatively, manual compensation can be performed as described in Ref. [7]. 3.9

Acquisition

1. Transfer samples into an appropriate plate or tube. 2. Perform a daily quality control (QC) to ensure proper performance of the flow cytometer (see Note 12). 3. Run an unstained sample to adjust scatter parameters if necessary. 4. Load application settings into the experiment. How to set application settings is briefly described (see Note 12). For more detailed instructions, see technical bulletin from BD “Standardizing Application Setup Across Multiple Flow Cytometers Using BD FACSDIVA™ Version 6 Software.” 5. Set the flow rate, and make sure that the data acquisition rate does not exceed the recommended one. 6. Set the events to be acquired and sample volume (when using the plate reader) to be analyzed. 7. Ideally, acquire samples and water alternatingly to minimize carryover and blockade of the sample injection tube.

3.10

4

Data Analysis

1. There are several data analysis software packages available for flow cytometry, including automated gating and clustering software (e.g., FlowJo, BD FACSDIVA™, CytExpert, tSNE, FlowSOM, flow cytometry suite, etc.). Analysis performed with the FlowJo software is shown in Fig. 1.

Notes 1. To pour away the supernatant, carefully invert the tube. Dry tube edges by tipping onto a paper towel, before putting it back in an upright position. 2. If tumor cells are not needed for further assays, it is recommended to check the completeness of the depletion. 3. Do not pour off the supernatant as the pellets of blood samples before red blood cell removal are rather loose. 4. Especially when using tandem dyes, make sure to cover plates immediately after pipetting, and protect all solutions containing antibodies or dyes from light exposure. 5. Fc receptors are found on numerous immune cells, for instance, on B-lymphocytes, natural killer cells, macrophages, neutrophils, and mast cells, and this can cause unwanted staining with antibodies, making Fc receptor blocking necessary.

Multicolour Flow Cytometry

265

Fig. 1 Flow cytometry gating scheme of a TIL analysis is shown. Samples were derived from subcutaneous breast cancer tumors (EMT6 mouse tumor model). Arrows highlight the parental populations (gates 8, 9, 10, 15, 16, 18, 19, 21). First CD45 was plotted against the time (1) to check the flow. After gating on CD45 (2), dead cells, duplets, and debris were removed from the analysis (3 to 6). Subsequently cells were evaluated based on their marker expression. The following cell populations are shown: gate 7, PMN-MDSCs (polymorphonuclear myeloid-derived suppressor cells); gate 11, CD8+ T cells, expression of PD-1 and LAG3 on this cell population was evaluated in the subsequent plot (26 to 29); gate 12, CD4+ T cells; gate 13, Tregs (regulatory T cells); gate 14, pDCs (plasmacytoid dendritic cells); gate 17, B cells; gate 20, MHCII-negative M-MDSCs (monocytic myeloid-derived suppressor cells); gate 22, TAMs (tumor-associated macrophages); gate 23, MHCII-positive M-MDSCs; gate 24, CD11b þ DCs (dendritic cells); gate 25, CD103þ DCs

6. It is recommended to use Brilliant Stain Buffer when using two or more antibodies conjugated to BD Brilliant dyes, as fluorescent dye interactions may cause staining artifacts. Brilliant Stain Buffer is compatible with traditional fluorochromes. If new antibody conjugates are used, the compatibility has to be checked beforehand. 7. FMOs are the most important controls when performing a multicolor flow experiment, as they are absolutely essential for proper gating [6, 7]. If possible, additional controls, like FMO plus isotype, can be prepared by replacing one antibody with the corresponding isotype. FMO plus isotype controls contain additional information about unspecific antibody or fluorophore binding and possible dye—dye interactions. Isotype controls alone (a mixture of all isotypes instead of staining

266

Martina M. Sykora and Markus Reschke

antibodies) are not very useful when analyzing complex tissue samples or performing multicolor experiments, because of two major reasons: firstly, tissue samples consist of different cell types with different autofluorescence levels, and secondly proper gating based on isotype controls is impossible due to compensation-dependent data spreading [6, 7]. 8. An appropriate isotype control antibody is characterized by being raised against an antigen not present on the cells analyzed, matching the host species, immunoglobulin subclass, and the fluorophore of the staining antibody. Usually, isotypes are recommended on the manufacturer’s webpage for each individual staining antibody. Unspecific antibody binding can be reduced by the blockade of Fc receptors (Subheading 3.4), by the addition of protein to the buffer (e.g., bovine serum albumin or fetal calf serum), and by the usage of the right antibody concentration. When titrating antibodies make sure to use the right sample, as marker expression can vary tremendously, for instance, between activated and inactivated cells, cells isolated from different tissues, treated and untreated samples, etc. 9. When preparing compensation controls, consider the three golden rules for compensation [7]: (a) Compensation controls must be as bright as or brighter than the marker stain. (b) Autofluorescence of positive and negative populations in compensation controls must be the same. (c) The fluorochromes used for compensation must match the experimental one. When compensating with cells, a reasonable amount of cells has to be stained. Especially when analyzing rare populations, this might not be possible using the experimental marker. A bright marker, e.g., CD45, CD8, or CD11b, can be used as a substitute. This might be problematic in the case of tandem dyes as their spectra can differ from batch to batch. If this approach is chosen, make sure to use substitute and experimental antibodies from the same company. Usually manufacturers have a quality controlled process to ensure that the spectra of different batches do not differ significantly. As described in rule 9b, the autofluorescence of the negative population has to match the stained one. Therefore, a scatter gate has to be set on the stained population (e.g., lymphocyte or monocyte gate) to remove all or most of the unwanted cells from the later compensation. Alternatively, specific cell types can be purified before preparing the compensation controls.

Multicolour Flow Cytometry

267

The majority of the problems, one can encounter when compensating with cells, can be solved by using commercially available antibody capture beads. Most antibody backbones are bound, and therefore the experimental antibodies can be used for compensation. Beads produce a very sharp and bright signal. However, very abundant marker signals can still exceed the brightness of the stained beads. To ensure proper compensation calculation, cellular samples have to be used for those markers. 10. An easy way to produce a homogenous sample for the viability dye control is to incubate the cell suspension at 56
C for 10 min. Alternatively cells can be treated with detergent (e.g., Perm/Wash buffer) to permeabilize the plasma membrane. 11. Most manufacturers provide negative beads without antibody capture molecules on the surface, which can be mixed with the positive microspheres before staining. 12. In the BD FACSDIVA™ software, a Cytometer Setup and Tracking (CS&T) module is implemented. It can be used for optimization and standardization of cytometer setups, performance tracking, as well as daily performance checks. Most importantly, the reports include information about channel linearity, detector efficiency (Q), optical background (B), and electronic noise (SDEN). When designing an experiment, keep in mind that important dim markers should be measured in detectors with high Q values, low B, and low SDEN (more information on optimal multicolor panel design can be found in Refs. [8, 9]. As a general rule, all signals have to be within the linear range of the detector, and the negative population has to be out of the electronic noise. The signal of the unstained sample is cell type specific; therefore it is highly recommended to generate application settings for each sample type. To do so, the negative population is set to 2.5  SDEN. Application settings can be saved and are automatically updated based on the daily performance check (for more detailed instructions, see technical bulletin from BD “Standardizing Application Setup Across Multiple Flow Cytometers Using BD FACSDIVA™ Version 6 Software”). QC parameter calculations and target value adjustments can also be done manually for any given cytometer [10]. Flow cytometer performance checks should include at least adjustment of laser delays and target values as well as the monitoring of tolerance range and optical background [10].

268

Martina M. Sykora and Markus Reschke

References 1. Perfetto SP, Chattopadhyay PK, Roederer M et al (2004) Seventeen-colour flow cytometry: unravelling the immune system. Nat Rev Immunol 4:648–655 2. Chattopadhyay PK, Roederer M (2015) A mine is a terrible thing to waste: high content, single cell technologies for comprehensive immune analysis. Am J Transplan 15:1155–1161 3. Bandura DR, Baranov VI, Ornatsky OI et al (2009) Mass cytometry: technique for real time single cell multitarget immunoassay based on inductively coupled plasma time-of-flight mass spectrometry. Anal Chem 81:6813–6822 4. Stoeckius M, Hafemeister C, Stephenson W et al (2017) Simultaneous epitope and transcriptome measurement in single cells. Nat Methods 14:865–868 5. Wei SC, Levine JH, Cogdill AP et al (2017) Distinct cellular mechanisms underlie AntiCTLA-4 and Anti-PD-1 checkpoint blockade. Cell 170:1120–1133

6. Maecker HT, Trotter J (2006) Flow cytometry controls, instrument setup, and the determination of positivity. Cytometry A 69:1037–1042 7. Roederer M (2002) Compensation in flow cytometry. Curr Protoc Cytom Chapter 1: Unit 1.14 8. Nguyen R, Perfetto S, Mahnke YD et al (2013) Quantifying spill- over spreading for comparing instrument performance and aiding in multicolor panel design. Cytometry A 83A:306–315 9. Baumgarth N, Roederer M (2000) A practical approach to multicolor flow cytometry for immunophenotyping. J Immunol Methods 243:77–97 10. Perfetto SP, Ambrozak D, Nguyen R et al (2012) Quality assurance for polychromatic flow cytometry using a suite of calibration beads. Nat Protoc 7:2067–2079

Part V Translational Methods to Validate Biomarkers

Chapter 17 Combined MicroRNA In Situ Hybridization and Immunohistochemical Detection of Protein Markers Boye Schnack Nielsen and Kim Holmstrøm Abstract MicroRNAs are short (18–23 nucleotides) noncoding RNAs involved in posttranscriptional regulation of gene expression through their specific binding to the 30 UTR of mRNAs. MicroRNAs can be detected in tissues using specific locked nucleic acid (LNA)-enhanced probes. The characterization of microRNA expression in tissues by in situ detection is often crucial following a microRNA biomarker discovery phase in order to validate the candidate microRNA biomarker and allow better interpretation of its molecular functions and derived cellular interactions. The in situ hybridization data provides information about contextual distribution and cellular origin of the microRNA. By combining microRNA in situ hybridization with immunohistochemical staining of protein markers, it is possible to precisely characterize the microRNA-expressing cells and to identify the potential microRNA targets. This combined technology can also help to monitor changes in the level of potential microRNA targets in a therapeutic setting. In this chapter, we present a fluorescence-based detection method that allows the combination of microRNA in situ hybridization with immunohistochemical staining of one and, in this updated version of the paper, two protein markers detected with primary antibodies raised in the same host species. Key words MicroRNA, Locked nucleic acid, In situ hybridization, Immunohistochemistry, Fluorescence multiplexing

1

Introduction MicroRNAs (miRNAs) constitute a group of short noncoding RNA, 18–23 nucleotides in length. Since the discovery of the small noncoding RNA in C. elegans, Lin-4 [1–3], miRNAs have been found to be an essential mechanism for gene expression in eukaryotic cells. In humans the number of annotated miRNAs has exceeded 1500 (miRBase 18). Individual miRNAs contribute to the regulation of protein expression typically by binding to the 30 untranslated region (UTR) of mRNAs [4–6]. miRNAs are considered to affect translation of more than 30% of all mRNAs and thereby affect cellular processes such as stem cell fate, differentiation and proliferation [7–9], cell migration [10, 11], and adhesion

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_17, © Springer Science+Business Media, LLC, part of Springer Nature 2019

271

272

Boye Schnack Nielsen and Kim Holmstrøm

[11]. miRNA genes are expressed as regular transcripts or as intronic transcripts containing regular short hairpin precursor structures. The sequential processing of the precursor transcript in the nucleus into mature single-stranded miRNAs in the cytoplasm may be unique for individual miRNAs; however, canonical pathways have been suggested [12, 13]. The mature miRNA is loaded into the RNA-induced silencing complex (RISC), which binds complementary regions in the 30 UTR of mRNA for regulation of the protein expression activity. The unique nucleic acid sequence of individual miRNA allows to some extent prediction of potential mRNA targets. Several databases based on target prediction algorithms, such as TargetScan (targetscan.org/vert_50/) and miRanda [14] (microrna.org/microrna/home.do), can be accessed online. The algorithms differ in particular in the criteria related to the pairing of the 30 -seed sequence resulting in different targets being identified. Critical use of these databases has recently been reviewed by Thomson et al. [15]. miRNA can be identified in most biological samples including paraffin-embedded tissue biopsies and blood samples. A surprisingly high stability of miRNA compared to mRNA is probably related to the high stability of the RISC/miRNA complex. Isolation of miRNA from tissue samples and body fluids has led to miRNA profiling at the level of microarray analysis [16, 17], PCR-based analyses [18], and deep sequencing [19, 20]. These quantitative miRNA expression analysis platforms have been extensively used to identify novel biomarker candidates among the miRNAs, including profiles in differential diagnosis [21, 22]. A common feature of these technologies is that the expression data represent a global average expression level in the tissue samples that is affected by the normalization procedure, which therefore must be chosen after detailed analysis and consideration. These technologies are strong quantitative tools with a high level of precision and reproducibility. However, to support and validate miRNA profiling as a biomarker discovery tool, more biological and molecular insight is needed to increase knowledge to the key mechanism regulated by the miRNA and to reach a better understanding of the cellular origin of the miRNA. For instance, knowing the precise characteristics of a cell population under investigation will narrow down the repertoire of potential targets to the biologically most relevant. Due to their functional involvement in specific molecular pathways, miRNAs have become interesting targets for therapeutic intervention. Examples of disease-related mechanisms affected by miRNAs include single nucleotide polymorphisms (SNPs) in 30 UTR sequences [23, 24] or in miRNA genes [25, 26] that may affect normal protein expression and have been identified in association with heart failure [27] and cancer [25, 28]. Chromosomal deletions and epigenetic changes involving miRNA also cause

MicroRNA In Situ Validation

273

abnormal protein expression that is associated with malfunctions and diseases [29, 30]. Aberrant protein expression processes may be normalized by therapeutic interference using either synthetic miRNA mimics [31], shRNAs [32–34], or miRNA inhibitors (anti-miR, antago-miR) [35–37]. The effect of introducing such a therapeutic strategy may be followed by monitoring if the protein expression returns to its normal level after treatment. Thus, validating the uptake of the miRNA-modulating drug and monitoring the efficacy of the targeting molecule may require thorough histological localization analysis of the therapeutic nucleic acid construct as well as the targeted protein(s). In situ hybridization (ISH) is a technology that enables characterization of the miRNA’s cellular origin [38–40]. For in situ detection of miRNAs, specific complementary probes consisting of DNA and locked nucleic acid (LNA) have been found to be a useful tool to achieve specific identification both in whole mount samples [41] and formalin-fixed and paraffin-embedded (FFPE) samples [38, 42, 43]. The results gained from an ISH study provide specific information of contextual expression within a tissue; the miRNA may be focally expressed in particular tissue compartments, for example, necrotic areas. For some miRNAs, expression is cell specific, such as miR-126 and miR-145, which are specifically expressed in endothelial cells and vascular smooth muscle cells, respectively [38]. In addition, simple histological analysis can often determine if a miRNA is expressed in the mesenchymal or epithelial compartment of a tissue. Specific identification of a cell group expressing a miRNA in cellular complex tissues like inflamed areas, tumor stroma, or regenerating tissue may require cell characterization using cell-type specific antibodies. We introduced a simple method combining miRNA ISH, using LNA-containing oligo probes, with immunohistochemical staining of a single protein in 2013 [44]. In the current updated paper, we have added a method option that combines miRNA ISH with sequential immunohistochemical staining of two protein markers, where the two primary antibodies are raised in the same host species, and that includes an antibody elution step [45]. This method is appealing in situations where, for example, two primary mouse mAbs are available for determining the cellular origin using a specific cell marker of interest and for the parallel monitoring of a potential miRNA target protein if it is absent or downregulated in the same cell population.

2 2.1

Materials Reagents

1. Double-labeled LNA™ probe (Exiqon/QIAGEN). In this study we used miRCURY™ double-FAM-labeled LNA™ probes specific for miR-21 and miR-205.

274

Boye Schnack Nielsen and Kim Holmstrøm

2. In situ hybridization buffer (microRNA ISH buffer set (FFPE), Exiqon/QIAGEN). 3. PBS 1, RNase-free quality. 4. Tween-20 and a 10% solution in RNase-depleted water. 5. RNase-depleted water, for example, RNase-free Milli-Q water (Millipore). 6. 20 SSC buffer, RNase-free quality. 7. Sheep anti-FITC-POD (Roche). Primary antibodies: in this study we used mouse monoclonal antibody against smooth muscle α-actin (sm α-actin, Dako), cytokeratin (Dako), laminin-5γ2 clone D4B5 (Merck Millipore, Billerica, MA, USA), and rabbit polyclonal antibody against programmed cell death 4 (PDCD4, Rockland Immunochemicals). 8. Cy3-conjugated goat anti-mouse, Cy3-conjugated goat antirabbit, and HRP-conjugated goat anti-mouse (Jackson ImmunoResearch). 9. Tyramine signal amplification (TSA) reagent, for example, TSA-FITC and TSA-Cy5 (Perkin Elmer). 10. Anti-fade Prolong Gold with DAPI (Invitrogen). 11. SuperFrost® Plus glass slides (Thermo Fisher Scientific). 12. Hybridizer (Dako). 13. Shandon Sequenza Slide racks (Thermo Fisher Scientific). 14. Horizontal humidifying chamber. 15. For RNase-depleting working tools and surfaces: RNase ZAP, RNase Away, or similar. 16. Hydrogen peroxide, 30% (VWR). 17. Glycine (Sigma). 18. 20% SDS (Sigma). 2.2 Buffers and Solutions

1. Proteinase-K buffer: to 900 mL Milli-Q water, add 5 mL of 1 M Tris–HCl (pH 7.4), 2 mL 0.5 M EDTA, and 0.2 mL 5 M NaCl. Adjust volume to 1000 mL. 2. Proteinase-K reagent: to 10 mL proteinase-K buffer, add 7.5 μL proteinase-K stock of 20 mg/mL. 3. 0.1 SSC buffer: to 995 mL Milli-Q water, add 5 mL 20 SSC. The SSC buffer should be autoclaved. 4. PBS-T: to 1 L of PBS (pH 7.4), add 1 mL Tween-20. 5. Blocking solution: to 37.5 mL of Milli-Q water, add 5 mL 1 M Tris–HCl (pH 7.5), 1.5 mL 5 M NaCL, 5 mL fetal bovine serum (FBS), and 100 μL diluted Tween-20 (10% solution). 6. 3% hydrogen peroxide. Add 5 mL to 45 mL of Milli-Q water for 50 mL solution.

MicroRNA In Situ Validation

275

7. 1% SDS: From the 20% SDS stock solution, add 2.5 mL to 47.5 mL Milli-Q water. 8. Antibody elution reagent: 0.94 g glycine is dissolved in 1% SDS. 2.3

3 3.1

Glassware

All glassware, including Coplin jars, glass-staining racks, and stacks of cover glass and bottles for buffers, should be heat treated in an oven at 180  C for 8 h. The items can be covered by aluminum foil before being placed in the oven in order to prevent contamination when removing the items afterward.

Methods Tissue Sections

3.2 In Situ Hybridization

Before starting cutting sections from the tissue blocks, the whole working place and tools (bench top, microtome, blade holder, brushes, tweezers, cooling plate, water bath, etc.) should be cleaned with RNase ZAP. Set the cooling plate to 15  C and place the FFPE block on the plate. Fill the warm water bath with RNase-free Milli-Q water, and set temperature to 40–50  C depending on the type of paraffin used for embedding. Another water bath is prepared as an RNase-free Coplin jar containing room-tempered RNase-free Milli-Q water. Insert a new disposable blade in the knife carrier, and place the block in the cassette clamp. Trim the block, in order to avoid the first couple of sections, which might be contaminated during previous handling such as from tissue preparation or embedding. After trimming the block, cut 6 μm sections and place them into a dry sterilized jar with room-tempered RNase-free Milli-Q water, where any folding can be reversed. Transfer the slide to a heated (40–50  C) water bath, where the tissue is stretched to avoid folds and overlaps in the structure and mount sections immediately on SuperFrost® Plus glass slides (glass slides taken from a new and untouched package are considered RNase-free). Mounting the section onto the slide should be done with care—allow water to slide away from in between the paraffin section and the glass slide— this process is important to avoid that sections fall off during deparaffinization. Let the slides dry for 2 h at room temperature and store at 4  C in a dry box containing silica gel. Melt paraffin at 60  C on the day prior to the in situ hybridization experiment. 1. Deparaffinize slides in xylene and ethanol solutions in Coplin jars ending up in PBS. In parallel, prepare a water bath and SSC buffer to be heated to 55  C (or the hybridization temperature). Place slides in xylene for 15 min (through two to three Coplin jars), and then hydrate through ethanol solutions 99% (three Coplin jars), 96% (two Coplin jars), and 70% (two

276

Boye Schnack Nielsen and Kim Holmstrøm

Coplin jars) to PBS (two Coplin jars). Each solution should include 5-min incubation. 2. Predigestion of tissue sections is done by applying 300 μL/ slide of the proteinase-K reagent at 25 μg/mL (see Note 1) directly on the tissue, and incubate for 40 min at 30  C in a horizontal humidifying chamber. 3. Discard the proteinase-K reagent and wash twice with PBS. 4. Apply 300 μL freshly prepared 3% hydrogen peroxide and incubate at room temperature for 10 min. 5. Discard the hydrogen peroxide and wash thrice with PBS. 6. Remove excess PBS and immediately apply 50 μL double-FAM LNA™ (see Note 2) probe solution, and gently shield with cover glass (see Note 3). The probe solution is prepared as follows: denature LNA™ probe and dilute the probe in Exiqon ISH buffer. For example, for 2 mL hybridization mix containing 20 nM double-FAM-labeled miR-21 LNA™ probe (from 25 μM probe stock), transfer 4 μL into the bottom of a 2-mL nonstick RNase-free tube, and place the tube at 90  C for 4 min. Spin down shortly using a table top centrifuge, and immediately add 2 mL ISH buffer into the tube. LNA™ probes for other miRNAs may require optimization of the concentration (see Note 4). 7. Place the slides in the hybridizer and start a preset hybridization program for 1 h at 55  C. 8. Place slides into 55  C pre-warmed 0.1 SSC in a Coplin jar. The cover slides will easily detach. Then transfer slides to another casket with 55  C pre-warmed 0.1 SSC. Wash slides thrice using 55  C pre-warmed 0.1 SSC. 9. Transfer slides to PBS-T and mount into Shandon Sequenza® slide racks. Avoid air bubbles during mounting. 10. Incubate with 200 μL blocking solution for 15 min at RT. 11. Detection of FAM probes is done by applying sheep anti-FAMPOD diluted 1:400 in blocking solution and incubated for 60 min at RT (preferably two times for 30 min). To 1000 μL blocking solution, add 2.5 μL sheep-anti-FAM POD. 12. Wash each slide with 300 μL PBS-T thrice for 3 min. 13. Incubate with 150 μL freshly prepared TSA-FITC or TSA-Cy5 reagent 7–15 min at RT. Protect from light during development. For the preparation of 1500 μL TSA-Cy5 substrate, add 30 μL TSA-Cy5 reagent into 1500 μL diluent (see Note 5). 14. Wash each slide with 300 μL PBS thrice for 3 min. 15. Incubate 150 μL primary antibody (here anti-sm α-actin diluted 1:200 or anti-PDCD4 diluted 1:300) (see Note 6)

MicroRNA In Situ Validation

277

diluted in PBS containing 1% BSA for 60 min at room temperature. 16. Wash each slide with 300 μL PBS thrice for 3 min. 17. If detection of a single protein marker antibody, then detect the primary antibody by incubating 150 μL Cy3-conjugated detecting antibody (anti-mouse or anti-rabbit) diluted 1:400 in PBS for 30 min at room temperature. Then go to step 29. 18. Or, if detection of two protein marker antibodies raised in the same host species (here, mouse mAbs), then incubate for 10 min in 3% hydrogen peroxide at RT. 19. Wash each slide with 300 μL PBS-T thrice for 3 min. 20. Detect the first primary antibody (step 15) by incubating 150 μL POD-conjugated detecting antibody (anti-mouse) diluted 1:400 in PBS for 30 min at RT. 21. Wash each slide with 300 μL PBS thrice for 3 min. 22. Incubate with 150 μL freshly prepared TSA-FITC reagent 7–15 min at RT. Protect from light during development. For preparation of 1500 μL TSA-FITC substrate, add 30 μL TSA-FITC reagent into 1500 μL diluent (see Note 5). 23. Wash each slide with 300 μL PBS thrice for 3 min. 24. To remove the first primary antibody from the tissue, place slides in a 50-mL Coplin jar with 50 mL glycine-SDS solution for 30–60 min at 50  C (see Note 7). Return the slides to the Shandon rack using PBS-T in the mounting process. 25. Wash each slide with 300 μL PBS-T thrice for 3 min. 26. Incubate 150 μL of the second primary mouse mAb for 1–2 h at room temperature or overnight at 4  C. 27. Wash each slide with 300 μL PBS thrice for 3 min. 28. Then, detect the second primary mouse mAb by incubating Cy3-conjugated detecting antibody (goat anti-mouse) diluted 1:400 in PBS for 30 min at room temperature. 29. Wash each slide with 300 μL PBS thrice for 3 min. 30. Mount slides directly with Anti-fade Prolong Gold containing DAPI. Store slides in the dark at 4  C. Evaluate staining result after 24 h. 31. Evaluate slides using a fluorescence microscope with filters allowing detection of FITC, Cy3, Cy5, and DAPI emission. 32. The evaluation of miRNA ISH results requires considerations on the specificity of the signal. Specificity analysis is essential in the development of molecular histology assays. For miRNA ISH analyses, there are several ways to address this question [40]. In particular for double fluorescence analyses, one key point to address is preventing cross-reaction between the two

278

Boye Schnack Nielsen and Kim Holmstrøm

parallel assays (see Note 8), and for the two mouse mAb stainings, it is necessary to ensure complete elution of the first antimouse antibody by including a control slide where the second primary mouse antibody is not added (step 26), but the Cy3 anti-mouse is (step 28) and no Cy3 fluorescence signal must be seen. 3.3 Microscopy Observations: Some Examples

In this methodology paper, we have combined (1) miR-21 ISH with sm α-actin IHC, (2) miR-205 ISH with sm α-actin IHC as examples to show a molecular characteristic of the miRNA-positive cells, and (3) miR-21 ISH with PDCD4 IHC as an example to compare a miRNA (miR-21) expression with one of its validated targets (PDCD4). Expression of miR-21 and miR-205 in human breast lesions has been described by others [42, 46]. sm α-actin is not expressed uniquely in one cell population in human breast tissue, but is expressed in different smooth muscle-differentiated cell populations: myoepithelial cells (the basal cells of the mammary gland and duct system), vascular smooth muscle cells (VSMC, surrounding the endothelial cells), and myofibroblasts (myo-differentiated fibroblastic cells present in tumor stroma). These three cell populations can be distinguished by their morphological characteristics. In the first example, miR-21 ISH signal is seen in some of the sm α-actin-positive tumor-associated myofibroblasts (Fig. 1). In the second example, miR-205 ISH signal is seen in sm α-actin-positive myoepithelial cells and is absent in VSMC and myofibroblasts (Fig. 2). The obvious differential expression of miR-205 and sm α-actin is noteworthy. miR-205 is likely not associated with smooth muscle differentiation in general since no expression is seen in the VSMC and myofibroblasts. miR-205 is more likely related to the basal cell characteristics also seen in skin [20]. The specific targets and role of miR-205 in differentiation of basal cells are not known [47]. The differential expression of miR-21 and sm α-actin is also interesting but is not as drastic as for miR-205. In addition to myofibroblasts, both smooth muscle cells in the colon [48] and myoepithelial cells in the breast [46] have been found to express miR-21. PDCD4 has been found by several groups to be a direct target of miR-21 [49, 50]. We combined miR-21 ISH and PDCD4 IHC in breast cancer samples and found that PDCD4 was more prevalent in the cancer cells than in adjacent stromal cells (Fig. 3). In contrast, miR-21 was most prevalent in breast cancer stromal cells. This differential expression should be expected if miR-21 downregulates its targets in the population where it is expressed, and the observation therefore strongly supports the many reports indicating that PDCD4 is a target of miR-21. In another application, we combined miR-21 ISH with immunohistochemical staining of cytokeratin and PDCD4 using a mouse mAb and the rabbit

MicroRNA In Situ Validation

279

Fig. 1 Combined miR-21 ISH and sm α-actin IHC for delineation of the cellular origin of miR-21 in breast cancer. miR-21 was detected with TSA-FITC substrate (green) and sm α-actin with Cy-3 (red). Section was counterstained with DAPI (blue). The miR-21 ISH signal is seen in sm α-actin-positive myofibroblast located in the breast cancer stroma (St) surrounding clusters of miR-21-negative cancer cells (Ca). The individual fluorophores are shown in black and white

polyclonal, respectively (see https://www.exiqon.com/ls/Pages/ BoyeSchnackNielsen.aspx). To present an example of using two antibodies raised in the same host species, we used an LNA probe to miR-21 in a colon cancer, and to document the presence of miR-21 in cancer cells in the tumor periphery, we first stained miR-21 using TSA-Cy5 substrate and then sequentially stained cytokeratin and laminin-5γ2 using two mouse mAbs (Fig. 4). A similar approach was used to co-stain miR-200b together with cytokeratin and laminin-5γ2 [51] and miR-21 together with CD45 and CD68 [52]. In this updated version of the combined microRNA ISH with IHC method paper [44], we have included the option of sequentially staining two different protein markers using two primary antibodies raised in the same host species. In a therapeutic setting, where a miRNA or its target sequence is obstructed, this method has an obvious advantage in validating the efficacy of the RNA interference drug, for example, after interfering with the

280

Boye Schnack Nielsen and Kim Holmstrøm

Fig. 2 Combined miR-205 ISH and sm α-actin IHC for delineation of the cellular origin of miR-205 in normal breast gland. miR-205 was detected with TSA-FITC substrate (green) and sm α-actin with Cy-3 (red). Section was counterstained with DAPI (blue). miR-205 ISH signal is seen in sm α-actin-positive myoepithelial cells surrounding the unstained luminal ductal epithelial cells

miR-21/PDCD4 pathway [49]. By adding more markers into the combined assay, the gained results can be highly informative and help to clarify molecular pathways. Beyond this application the technique might have a value in the future for differential diagnosis [53, 54].

4

Notes 1. Synchronous access to miRNAs and protein antigens requires that both probes (the LNA probe and the primary antibody) can bind to their targets. miRNA ISH has been developed on a

MicroRNA In Situ Validation

281

Fig. 3 Combined miR-21 ISH and PDCD4 IHC for delineation of regulatory targets of miR-21 in human breast cancer. miR-21 was detected with TSA-FITC substrate (green) and PDCD4 with Cy-3 (red). Section was counterstained with DAPI (blue). miR-21 ISH signal is seen in the stroma (St) surrounding clusters of cancer cells (Ca). Intense PDCD4 signal can be seen in cancer cells, which are miR-21 negative, whereas weaker PDCD4 signal is seen in miR-21-positive fibroblasts

platform where tissue sections are predigested with proteolytic enzymes, such as proteinase-K. This pretreatment procedure also allows antigen detection by many primary antibodies; however, today heat-induced epitope retrieval (HIER) has become a frequent standard for many antibodies, whereas the effect of proteolytic treatment is often not reported. Therefore, note that novel antibodies on the market may not have been tested in a protocol based on proteolytic antigen retrieval. On the other hand, miRNA ISH may still find a way to its target in non-proteolytic pretreatment procedures. In this study we performed proteinase-K treatment at 25 μg/mL at 30  C for 40 min. The optimal proteinase-K treatment is dependent on the tissue to be analyzed and the fixation protocol used. Routinely processed tissue samples, which are usually fixed overnight at room temperature in neutral buffered formalin, often show adequate performance. Denaturing fixatives should be avoided for use in ISH analyses. Proteinase-K treatment

282

Boye Schnack Nielsen and Kim Holmstrøm

Fig. 4 Combined staining of miR-21, cytokeratin, and laminin-5γ2 in human colon cancer. An FAM-labeled LNA probe was used for miR-21 ISH and detected with a POD-conjugated goat anti-FAM antibody and TSA-Cy5 substrate (white in merged image), followed by cytokeratin mouse mAb staining using a POD-conjugated anti-mouse secondary antibody and TSA-FITC substrate (green in merged image). After elution of the cytokeratin mAb, the laminin-5γ2 mAb was added and detected with Cy3-conjugated antimouse antibody (red in merged image); DAPI nuclear stain (blue in merged image) was added with the mounting medium. Arrows indicate cancer cells with miR-21 expression and some laminin-5γ2 staining. Reference added in proofs: Knudsen et al: Clin Exp Metastasis 2018, 35:819

efficiency is dependent on three conditional parameters: the proteinase-K concentration, the incubation temperature, and the duration of incubation. Fluorescence miRNA ISH staining of frozen sections [55] may reduce autofluorescence noise and provide a broader panel of useful primary antibodies. 2. LNA oligos can be differently labeled with haptens, such as carboxyfluorescein (6-FAM), digoxigenin (DIG), or biotin. Recently, we reported that FAM-labeled LNA probes work equally well as DIG-labeled LNA probes [56]. The choice of probe label depends on experimental design and the techniques available in the laboratory. The hapten label provides a template for crucial signal amplification since the FITC label on the oligo itself is not sufficient to allow detection in standard epifluorescence. In this study, the fluorescence signal was obtained with the TSA-FITC (or Cy5) substrate, which allowed detection of miR-21 and miR-205.

MicroRNA In Situ Validation

283

3. For the hybridization step, we employed a Dako hybridizer; however we also tested the protocol using a conventional hybridization oven and obtained similar ISH results. We applied 50 μL hybridization mix and sealed with a cover glass. In this experiment we did not seal the cover glass with rubber cement, which is proposed when incubating smaller hybridization reagent volumes [38, 40]. 4. In this study we used the miR-21 probe at 10–20 nM. The probe concentration range may vary from 10 to 100 nM. Attention should be paid to unspecific binding and crosshybridization when incubating probes at high-range probe concentrations. Use of a negative control probe (such as the scrambled probe, Exiqon/QIAGEN) often gives a good indication of unspecific probe binding, when incubated at the same concentration as the specific probe. Cross-hybridization cannot be evaluated using the scrambled probe and requires other specificity controls such as mismatch controls and positive controls [40]. 5. The intensity of the ISH signal can be regulated by the incubation time of TSA-FITC (or TSA-Cy5) substrate. We incubated the reagent for 5 min at room temperature, but the substrate reagent may be allowed to proceed for up to 15 min. A sufficient volume of the substrate reagent should be incubated to ensure that the reagent fully displaces the washing buffer in the slide chamber (at least 150 μL in the Shandon slide rack). 6. A series of antibody dilutions tests prior to the combined ISH/IHC can give a good hint of optimal antibody dilution. Best fluorescence images are obtained when the intensity of the individual fluorophores are balanced. 7. The duration of this antibody elution step was in some applications reduced to 30 min to prevent deterioration of the tissue for 5-μm-thick sections. Deterioration is evident when the DAPI nuclear stain becomes diffuse and the nuclei enlarged and will depend on how strong the tissue is fixed and how thick the tissue section is. The duration may be different for different antibodies, and the elution efficiency should be monitored by testing for remaining activity. 8. The experiment must be designed without possibility of crossreaction between the two parallel assays. For example, avoid using a goat primary antibody that needs detection with antisheep antibody since this may cross-react with the sheep antiFAM antibody. Also, the anti-FAM antibody should not be used after applying TSA-FITC reagent.

284

Boye Schnack Nielsen and Kim Holmstrøm

Acknowledgments We thank Trine Møller for excellent technical assistance and the Danish Ministry of Science, Innovation and Technology, for financial funding. References 1. Ambros V (2001) microRNAs: tiny regulators with great potential. Cell 107:823–826 2. Lee RC, Feinbaum RL, Ambros V (1993) The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell 75:843–854 3. Lagos-Quintana M, Rauhut R, Lendeckel W et al (2001) Identification of novel genes coding for small expressed RNAs. Science 294:853–858 4. Bartel DP (2004) MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 116:281–297 5. Pillai RS, Bhattacharyya SN, Artus CG et al (2005) Inhibition of translational initiation by Let-7 MicroRNA in human cells. Science 309:1573–1576 6. Liu J, Rivas FV, Wohlschlegel J et al (2005) A role for the P-body component GW182 in microRNA function. Nat Cell Biol 7:1261–1266 7. Ambros V (2011) MicroRNAs and developmental timing. Curr Opin Genet Dev 21:511–517 8. Cordes KR, Sheehy NT, White MP et al (2009) miR-145 and miR-143 regulate smooth muscle cell fate and plasticity. Nature 460:705–710 9. Makeyev EV, Zhang J, Carrasco MA et al (2007) The MicroRNA miR-124 promotes neuronal differentiation by triggering brainspecific alternative pre-mRNA splicing. Mol Cell 27:435–448 10. Yao Q, Cao S, Li C et al (2011) Micro-RNA-21 regulates TGF-beta-induced myofibroblast differentiation by targeting PDCD4 in tumorstroma interaction. Int J Cancer 128:1783–1792 11. Madhyastha R, Madhyastha H, Nakajima Y et al (2012) MicroRNA signature in diabetic wound healing: promotive role of miR-21 in fibroblast migration. Int Wound J 4:355–361 12. Medina PP, Slack FJ (2008) microRNAs and cancer: an overview. Cell Cycle 7:2485–2492 13. Winter J, Jung S, Keller S et al (2009) Many roads to maturity: microRNA biogenesis pathways and their regulation. Nat Cell Biol 11:228–234

14. Betel D, Wilson M, Gabow A et al (2008) The microRNA.org resource: targets and expression. Nucleic Acids Res 36:D149–D153 15. Thomson DW, Bracken CP, Goodall GJ (2011) Experimental strategies for microRNA target identification. Nucleic Acids Res 39:6845–6853 16. Ferdin J, Kunej T, Calin GA (2010) Non-coding RNAs: identification of cancerassociated microRNAs by gene profiling. Technol Cancer Res Treat 9:123–138 17. Sorensen KD, Orntoft TF (2010) Discovery of prostate cancer biomarkers by microarray gene expression profiling. Expert Rev Mol Diagn 10:49–64 18. Jensen SG, Lamy P, Rasmussen MH et al (2011) Evaluation of two commercial global miRNA expression profiling platforms for detection of less abundant miRNAs. BMC Genomics 12:435 19. Burnside J, Ouyang M, Anderson A et al (2008) Deep sequencing of chicken microRNAs. BMC Genomics 9:185 20. Joyce CE, Zhou X, Xia J et al (2011) Deep sequencing of small RNAs from human skin reveals major alterations in the psoriasis miRNAome. Hum Mol Genet 20:4025–4040 21. Schetter AJ, Leung SY, Sohn JJ et al (2008) MicroRNA expression profiles associated with prognosis and therapeutic outcome in colon adenocarcinoma. JAMA 299:425–436 22. Ralfkiaer U, Hagedorn PH, Bangsgaard N et al (2011) Diagnostic microRNA profiling in cutaneous T-cell lymphoma (CTCL). Blood 118:5891–5900 23. Clop A, Marcq F, Takeda H et al (2006) A mutation creating a potential illegitimate microRNA target site in the myostatin gene affects muscularity in sheep. Nat Genet 38:813–818 24. Richardson K, Lai CQ, Parnell LD et al (2011) A genome-wide survey for SNPs altering microRNA seed sites identifies functional candidates in GWAS. BMC Genomics 12:504 25. Permuth-Wey J, Thompson RC, Burton NL et al (2011) A functional polymorphism in the pre-miR-146a gene is associated with risk and

MicroRNA In Situ Validation prognosis in adult glioma. J Neuro-Oncol 105:639–646 26. Lei B, Gao S, Luo LF et al (2011) A SNP in the miR-27a gene is associated with litter size in pigs. Mol Biol Rep 38:3725–3729 27. Nossent AY, Hansen JL, Doggen C et al (2011) SNPs in microRNA binding sites in 30 -UTRs of RAAS genes influence arterial blood pressure and risk of myocardial infarction. Am J Hypertens 24:999–1006 28. Zhang L, Liu Y, Song F et al (2011) Functional SNP in the microRNA-367 binding site in the 30 UTR of the calcium channel ryanodine receptor gene 3 (RYR3) affects breast cancer risk and calcification. Proc Natl Acad Sci U S A 108:13653–13658 29. Lodygin D, Tarasov V, Epanchintsev A et al (2008) Inactivation of miR-34a by aberrant CpG methylation in multiple types of cancer. Cell Cycle 7:2591–2600 30. Calin GA, Ferracin M, Cimmino A et al (2005) A MicroRNA signature associated with prognosis and progression in chronic lymphocytic leukemia. N Engl J Med 353:1793–1801 31. Trang P, Wiggins JF, Daige CL et al (2011) Systemic delivery of tumor suppressor microRNA mimics using a neutral lipid emulsion inhibits lung tumors in mice. Mol Ther 19:1116–1122 32. Sioud M (2011) Promises and challenges in developing RNAi as a research tool and therapy. Methods Mol Biol 703:173–187 33. Wang Z, Rao DD, Senzer N et al (2011) RNA interference and cancer therapy. Pharm Res 12:2983–2995 34. Garofalo M, Croce CM (2011) microRNAs: master regulators as potential therapeutics in cancer. Annu Rev Pharmacol Toxicol 51:25–43 35. Gambari R, Fabbri E, Borgatti M et al (2011) Targeting microRNAs involved in human diseases: a novel approach for modification of gene expression and drug development. Biochem Pharmacol 82:1416–1429 36. Kasinski AL, Slack FJ (2010) Potential microRNA therapies targeting Ras, NFkappaB and p53 signaling. Curr Opin Mol Ther 12:147–157 37. Stenvang J, Silahtaroglu AN, Lindow M et al (2008) The utility of LNA in microRNA-based cancer diagnostics and therapeutics. Semin Cancer Biol 18:89–102 38. Jorgensen S, Baker A, Moller S et al (2010) Robust one-day in situ hybridization protocol for detection of microRNAs in paraffin samples using LNA probes. Methods 52:375–381 39. Soe MJ, Moller T, Dufva M et al (2011) A sensitive alternative for microRNA in situ

285

hybridizations using probes of 20 -O-methyl RNA þ LNA. J Histochem Cytochem 59:661–672 40. Nielsen BS (2012) MicroRNA in situ hybridization. Methods Mol Biol 822:67–84 41. Kloosterman WP, Wienholds E, de BE et al (2006) In situ detection of miRNAs in animal embryos using LNA-modified oligonucleotide probes. Nat Methods 3:27–29 42. Sempere LF, Preis M, Yezefski T et al (2010) Fluorescence-based codetection with protein markers reveals distinct cellular compartments for altered MicroRNA expression in solid tumors. Clin Cancer Res 16:4246–4255 43. Nuovo GJ (2010) In situ detection of microRNAs in paraffin embedded, formalin fixed tissues and the co-localization of their putative targets. Methods 52:307–315 44. Nielsen BS, Holmstrom K (2013) Combined microRNA in situ hybridization and immunohistochemical detection of protein markers. Methods Mol Biol 986:353–365 45. Pirici D, Mogoanta L, Kumar-Singh S et al (2009) Antibody elution method for multiple immunohistochemistry on primary antibodies raised in the same species and of the same subtype. J Histochem Cytochem 57:567–575 46. Rask L, Balslev E, Jorgensen S et al (2011) High expression of miR-21 in tumor stroma correlates with increased cancer cell proliferation in human breast cancer. APMIS 119:663–673 47. Greene SB, Herschkowitz JI, Rosen JM (2010) The ups and downs of miR-205: identifying the roles of miR-205 in mammary gland development and breast cancer. RNA Biol 7:300–304 48. Nielsen BS, Jorgensen S, Fog JU et al (2011) High levels of microRNA-21 in the stroma of colorectal cancers predict short disease-free survival in stage II colon cancer patients. Clin Exp Metastasis 28:27–38 49. Young MR, Santhanam AN, Yoshikawa N et al (2010) Have tumor suppressor PDCD4 and its counteragent oncogenic miR-21 gone rogue? Mol Interv 10:76–79 50. Ruan Q, Wang T, Kameswaran V et al (2011) The microRNA-21-PDCD4 axis prevents type 1 diabetes by blocking pancreatic beta cell death. Proc Natl Acad Sci U S A 108:12030–12035 51. Knudsen KN, Lindebjerg J, Nielsen BS et al (2017) MicroRNA-200b is downregulated in colon cancer budding cells. PLoS One 12: e0178564 52. Thorlacius-Ussing G, Schnack Nielsen B, Andersen V et al (2017) Expression and

286

Boye Schnack Nielsen and Kim Holmstrøm

localization of miR-21 and miR-126 in mucosal tissue from patients with inflammatory bowel disease. Inflamm Bowel Dis 23:739–752 53. Sempere LF (2011) Integrating contextual miRNA and protein signatures for diagnostic and treatment decisions in cancer. Expert Rev Mol Diagn 11:813–827 54. Soe MJ, Okkels F, Sabourin D et al (2011) HistoFlex-a microfluidic device providing uniform flow conditions enabling highly

sensitive, reproducible and quantitative in situ hybridizations. Lab Chip 11:3896–3907 55. Nielsen BS, Moller T, Holmstrom K (2014) Chromogen detection of microRNA in frozen clinical tissue samples using LNA probe technology. Methods Mol Biol 1211:77–84 56. Gould BR, Damgaard T, Nielsen BS (2017) Chromogenic in situ hybridization methods for microRNA biomarker monitoring of drug safety and efficacy. Methods Mol Biol 1641:399–412

Chapter 18 Tangential Flow Filtration with or Without Subsequent Bind-Elute Size Exclusion Chromatography for Purification of Extracellular Vesicles Joel Z. Nordin, R. Beklem Bostancioglu, Giulia Corso, and Samir EL Andaloussi Abstract Extracellular vesicles (EVs) have gained increased attention over the last decade due to their potential as biomarkers and therapeutic entities. However, the characterization and development of EV research has been hampered by the lack of sufficiently effective purification methods. Several concerns have been raised toward the gold standard purification method ultracentrifugation, such as operator-dependent yields, crushing and aggregation of vesicles, poor scalability, and relative lack of purity. Here, we describe, in details, the use of an alternative purification technique: tangential flow filtration with or without subsequent bind-elute size exclusion chromatography that we have previously shown to be reproducible and scalable for purification of EVs. Key words Extracellular vesicles, Exosomes, Microvesicles, Apoptotic bodies, Tangential flow filtration, Bind-elute size exclusion chromatography, Size exclusion chromatography, Purification

1

Introduction Extracellular vesicles (EVs) have gained increased attention over the last decade after a series of breakthrough articles [1–3]. EVs are secreted by most if not all cells in the body and can be detected in all body fluids. They can be divided into three main subgroups: exosomes of around 30–150 nm in diameter originating from the endolysosomal pathway, microvesicles 100–1000 nm in diameter that bud directly from the plasma membrane, and apoptotic bodies that are more heterogeneous ranging from 500 to 5000 nm and originate from apoptotic cells [4]. In this article the term EV will be used to describe exosomes and microvesicles. EVs contain bioactive lipids, proteins, and different RNA species (e.g., miRNA, mRNA, and rRNA) [4]. Due to their heterogeneous content, they can affect cells in a pleiotropic manner, such as directly interacting

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_18, © Springer Science+Business Media, LLC, part of Springer Nature 2019

287

288

Joel Z. Nordin et al. 15000

Particle count (x106)

CM TFF 10000

BE-SEC cBE-SEC

5000

0 0

200

400

600

800

1000

Particle size (nm) Fig. 1 NTA measurement after each step of purification of N2a cells in duplicate. Each line is the mean of two replicates. TFF and BE-SEC step renders high recovery of over 80% for each step. CM: conditioned media 0.22 μm filtrated, TFF: tangential flow filtration, BE-SEC: bind-elute size exclusion chromatography, cBE-SEC: concentrated bind-elute size exclusion chromatography

with cell surface receptors, deliver biologically active RNAs and proteins, and can subsequently, through various mechanisms, influence the phenotype of recipient cells [1–3]. However, both therapeutic and biomarker EV research has been hampered by the lack of stringent purification protocols. The gold standard in the field is ultracentrifugation (UC), which consists of sequential centrifugations with increasing speed where the last step is a UC step at 110,000–120,000
g [5]. However, several concerns have been raised regarding the technique including operator dependent yields, crushing and aggregation of vesicles, poor scalability and lack of purity [6, 7]. Hence, several purification methods have been developed including microfluidics, precipitation-based techniques, size exclusion chromatography (SEC) and immunocapture-based techniques [8, 9]. We have recently developed a scalable and robust purification method based on tangential flow filtration with or without subsequent bind-elute size exclusion chromatography (TFF/BE-SEC). TFF/BE-SEC is optimized for cell culture supernatants and produces consistently high EV yields between 50% and 80% (Fig. 1) [10]. The method is highly suitable for the processing of large media quantities to generate pure samples for subsequent preclinical and, potentially, clinical use.

2 2.1

Materials Reagents

1. RPMI 1640 medium, GlutaMAX™ Supplement, HEPES (Cat. No. 72400054, Thermo Scientific).

Tangential Flow Filtration

289

2. DMEM medium, high glucose, GlutaMAX™ Supplement, pyruvate (Cat. No. 31966021, Thermo Scientific). 3. Fetal bovine serum (FBS). 4. Opti-MEM, Reduced Serum Medium (Cat. No. 31985054, Thermo Scientific). 5. Antibiotic/Antimycotic (anti-anti) (Cat. No. 15240062, Thermo Scientific). 6. Hydrocortisone (Cat. No. HD888, Sigma). 7. Glycerol for molecular biology, 99% (G5516-1L, SigmaAldrich). 8. 1
DPBS, for cell culture. 9. Sodium hydroxide (NaOH), powder. 10. Human albumin, 100 mL 200 g/L (Albunorm, Octapharma). 11. 2-Propanol, EtOH, and Milli-Q water. 2.2

Other Materials

1. 0.22 μm cellulose acetate membrane bottle-top vacuum filters, low protein binding (Cat. No. CLS430769-12EA, SigmaAldrich). 2. 0.22 μm sterile syringe filters (Cat. No. 514-0061, VWR). 3. 50 mL sterile syringe. 4. Cell culture flasks, 175 cm2. 5. Petri dishes 150 mm
25 mm. 6. 50 mL polypropylene conical centrifuge tubes. 7. Low binding microtubes (maximum recovery—MCT-150-L-C). 8. 100 kDa cutoff hollow fiber filters (MidiKros 65 cm 100 kDa mPES 0.5 mm FLL
FLL, Cat. No. D06-E100-05-N, Spectrum Labs). 9. 300 kDa cutoff hollow fiber filters (MidiKros 65 cm 300 kDa mPES 0.5 mm FLL
FLL, Cat. No. D06-E300-05-N, Spectrum Labs). 10. 100 kDa cutoff hollow fiber filters (MicroKros, 20 cm2, 100 kDa mPES, Cat. No. C02-E100-05-N, Spectrum Labs). 11. 300 kDa cutoff hollow fiber filters (MicroKros, 20 cm2, 300 kDa mPES, Cat. No. C02-E300-05-N, Spectrum Labs). 12. Bind-elute size exclusion chromatography columns (HiScreen Capto Core 700 column, Cat. No. 17-5481-15, GE Healthcare Life Sciences). 13. Amicon Ultra-15 10 kDa (Cat. No. UFC901024) or 100 kDa (Cat. No. UFC910024) molecular weight cutoff spin filter (Millipore).

290

Joel Z. Nordin et al.

2.3 Recipes (See Note 1) 2.3.1 1
PhosphateBuffered Saline (PBS) (0.01 M)

Dissolve 8 g NaCl, 0.2 g KCl, 1.44 g Na2HPO4, and 0.27 g KH2HPO4 in 800 mL ddH2O. Adjust pH to 7.5 with HCl or NaOH and add ddH2O to 1 L.

2.3.2 20% Glycerin (1 L)

Add 200 mL of glycerol to 800 mL ultrapure water and mix it well. This solution can be stored at room temperature.

2.3.3 1 M NaOH (1 L)

Dissolve 40 g of NaOH in 1 L ultrapure water. Prepare 0.1 M NaOH solution by diluting 1 M NaOH 1:10 with ultrapure water. This solution can be stored at room temperature.

2.3.4 1 M NaOH in 30% Isopropanol (1 L)

Dissolve 40 g of NaOH in 700 mL ultrapure water and 300 mL of isopropanol and mix it well. This solution can be stored at room temperature.

2.3.5 20% EtOH (1 L)

Add 200 mL of 99.5% ethanol to 800 mL ultrapure water and mix it well. This solution can be stored at room temperature.

2.3.6 50% EtOH (1 L)

Add 500 mL of 99.5% ethanol to 500 mL ultrapure water and mix it well. This solution can be stored at room temperature.

2.3.7 0.3% Albumin (30 mL)

Add 18 μL of albumin in 30 mL of 0.01 M PBS. This solution can be stored at 4  C.

2.4

Equipment

1. Vacuum filtration units (Rapid-Flow™ filters MF 75, Nalgene®, Cat. No. 514-0024/NALG565-0020, Thermo Scientific). 2. Vacuum mini pump (Cat. No. Z288292EU, Sigma-Aldrich). 3. KR2i TFF system equipped with an automatic backpressure valve. The flow path was created using #16 tubings (Spectrum Labs). ¨ KTA start chromatography system (GE Healthcare Life 4. A Sciences). 5. Bench top centrifuge (Hettich Universal 320, Cat. No. 1401, Hettich Lab). 6. Cooling centrifuge (Hettich Rotina 420R, Cat. No. 4701, Hettich Lab).

3 3.1

Methods Cell Culturing

Immortalized, human bone marrow-derived mesenchymal stromal cells (hTertþ MSCs) [11, 12] were cultured in RPMI-1640 supplemented with 10% FBS, 106 mol/L hydrocortisone and 1%

Tangential Flow Filtration

291

anti-anti (full RPMI medium). N2a (mouse neuroblastoma cells) cells were cultured in DMEM supplemented with 10% FBS and 1% anti-anti (full DMEM). Cells were grown at 37  C, 5% CO2 in a humidified atmosphere. 1. Seed 2.5
106 MSC cells and/or 4
106 N2a cells into 150 mm petri dishes in full RPMI/DMEM medium. After 24 h, remove the medium, wash the cells once with warm DPBS and add 17–18 mL Opti-MEM per plate. After 48 h cells should reach at least 70–80% confluency; however, it should not be overgrown (see Notes 2–4). 2. After 48 h, harvest the conditioned medium (CM) and collect it in 50 mL centrifugation tubes. 3. Keep the 50 mL tubes on ice and transfer them to a pre-cooled centrifuge. Start a low-speed centrifugation step at 700
g for 5 min followed by a 2000
g spin for 10 min to remove larger particles and cell debris (see Notes 5–7). 3.2 Isolation of EVs from Cell Culture Supernatant 3.2.1 Tangential Flow Filtration (TFF)

0.01 M PBS, Milli-Q water (washing), 0.1 M NaOH (washing) and 20% glycerin (100 mL) are needed for this step. 1. Using 0.22 μm pore size filters, filter the sample to remove any larger particles, according to option A or B below (see Notes 8 and 9). Option A: For CM volumes up to 100 mL, we recommend using a 50 mL syringe and attach a 0.22 μm syringe filter. Manually push the sample through the filter. Keep the filtrate and discard the syringe (see Note 10). Option B: For CM volumes above 100 mL, we recommend bottle-top 0.22 μm vacuum filters. Connect the bottle-top filters to a vacuum source and let the CM pass through the filter. Keep the filtrate for subsequent steps. 2. The filtrated CM is now ready for TFF filtration. In the following steps, the settings and methods refer to TFF with the KR2i TFF system (Spectrum Labs) equipped with modified polyethersulfone (mPES) hollow fiber filters with 300 kDa membrane pore size (MidiKros, 370 cm2 surface area, Spectrum Labs) (see Note 11). In all the following TFF steps, the flow rate should be set at 100 mL/min and transmembrane pressure set at 3.0 psi (during priming, running the sample and washing, transmembrane pressure should not exceed 3.0) and shear rate at 3700 s1. 3. Start with turning on the machine (before turning on the machine, make sure all valves are set to open for the machine to calibrate the pressure). Be careful that the waste valve is open and that the reservoir and tubing are connected to the system.

292

Joel Z. Nordin et al.

Place the waste tube into waste container. Switch on the backpressure valve to “TMP” mode. 4. The TFF column is normally stored in 20% glycerin solution. Before loading the sample, revert the TFF pump direction counterclockwise to empty the column (pressing enter on the pump direction icon will change the pump direction). 5. Bring the pump direction back to clockwise to start the priming. 6. Prime the column first with 250 mL Milli-Q water, followed by 500 mL filtered PBS 1
or the buffer of choice. The flow must be set to clockwise direction with a speed of 100 mL/min and maximum transmembrane pressure set at 3.0 psi. 7. After priming, empty the column by running the flow in the counterclockwise direction. 8. Reset the pump direction to clockwise, load your sample and run at a flow rate of 100 mL/min (transmembrane pressure at 3.0 psi and shear rate at 3700 s1) (see Note 12). 9. Recommended: diafiltrate your sample with at least two times the initial volume with 0.01 M PBS or similar buffer of choice to remove protein contaminants and simultaneously make a buffer exchange. 10. Concentrate the sample to roughly 5 mL in the reservoir. 11. Revert the flow counterclockwise to recover the sample contained in the column and tubing (speed 100 mL/min and pressure at 3.0 psi). Final volume will be approximately 35 mL (with the column size described above and tubing length optimized). 12. Recommended: to further increase the recovery take 5–10 mL 0.22 μm filtered 0.01 M PBS and run it in the clockwise direction once through the column and recover the material in the same reservoir as before. Hence, after concentration/ diafiltration, recovery, and the second recovery step, the total final volume is around 40–50 mL (see Notes 13–15). 13. Keep the diafiltrated/concentrated material for downstream purification by BE-SEC if higher purity is needed otherwise go directly to the final concentration step. 3.2.2 TFF Column Washing

1. Rinse the reservoir and fill it with 250 mL Milli-Q water and run it at a flow rate of 100 mL/min (transmembrane pressure at 3.0 psi and shear rate at 3700 s1). After running the reservoir empty of Milli-Q water, revert the flow counterclockwise and empty the column of any leftover liquid. 2. Fill the reservoir with 250 mL 0.1 M NaOH and run it with the same settings in the clockwise direction. When about half the

Tangential Flow Filtration

293

volume is left, close the waste valve and let the buffer recirculate for 5–10 min. Subsequently, open the waste valve again and run leftover NaOH until all have gone to waste. Revert the flow counterclockwise to empty the column. 3. Rinse the reservoir and fill it with 1 L of Milli-Q water with the same settings as above. At the end of the washing with water, empty the column completely by reverting the flow counterclockwise. 4. The column is stored in 20% glycerin at RT for short-term and at 4  C for long-term storage (longer than 2 weeks). Fill the reservoir with 100 mL of 20% glycerin, let it run into the system and then close the waste valve to completely fill the column with the solution. 5. After 5 min stop the flow. 6. Remove and rinse the reservoir, and if used de-attach the feeding tube and rinse it with Milli-Q water. Empty the waste, open the backpressure valve and turn off the TFF machine. 3.2.3 Bind-Elute Size Exclusion Chromatography (BE-SEC)

For the following steps, make sure you have fresh and clean Milli-Q water, 0.22 μm filtrated 0.01 M PBS (for priming), 1 M NaOH in 30% isopropanol (for washing) and 20% EtOH (for storage), 1 L of each. The EV purification by BE-SEC has been performed on an ¨ KTA start chromatography system (GE Healthcare Life Sciences) A with UV absorbance set at 280 nm. The chromatography system is connected to a BE-SEC column (HiScreen Capto Core 700 column, GE Healthcare Life Sciences); firstly, the two stoppers at the end of the column have to be removed and the chromatography tubing should be connected with a drop-to-drop technique to avoid introducing air into the column. 1. Recommended: firstly, prime the tubing with water to avoid air going through the column (sample valve set to buffer, wash valve set to waste, outlet valve set to waste). To remove ethanol and avoid precipitation of buffer salts, wash the column with 2 column volumes (CVs) (HiScreen Capto Core column; 1 CV ¼ 4.7 mL) of Milli-Q water at a flow rate of 3 mL/min and pressure threshold set at 0.5 MPa (sample valve set to buffer, wash valve set to column and outlet valve set to waste). 2. Prime the column with 5 CVs of filtered 0.01 M PBS or in the buffer in which the sample is diluted in, with a flow rate of 2–3 mL/min. Sample valve set to buffer, wash valve set to column and outlet valve set to waste (see Notes 16 and 17). 3. Before loading the sample, make sure the UV absorbance baseline is stable; when it is stable, the baseline can be blanked (UV auto zero) before the sample is injected (see Note 18).

294

Joel Z. Nordin et al.

4. About 5 CVs of sample can be loaded onto the column with a flow rate of 1–1.5 mL/min, using the sample tubing attached to the sample valve. 5. When the sample is entirely loaded onto the system, the sample valve has to be switched to the buffer valve to avoid air going through the column (do the switch as soon as the sample is fully loaded into the sample tubing). 6. When the UV absorbance starts to increase, change the outlet valve to collection and start to collect your sample in a 50 mL tube. Since it is a BE-SEC column, impurities smaller than 700 kDa should bind to the inside of the column beads and the flow through corresponds to the purified sample; hence, collect all flow through for later concentration steps. 7. When the UV absorbance reaches the baseline, close the collection valve and keep your collected sample for the final concentration step as described below. 8. Pause the flow, revert the column upside down and proceed with the cleaning in place. Place one of the buffer tubing (A or B) into the bottle containing 0.1 M NaOH in 30% isopropanol and prime the tubing with the solution before running through the column (set sample valve to buffer (A or B), wash valve to waste and outlet valve to waste). 9. Run the cleaning in place for 30–60 min at a flow rate of 0.5 mL/min (set sample valve to buffer (A or B), wash valve to column and outlet valve to waste). 10. Revert the column (the flow should follow the orientation of the label indicated on the column). Change solution again to Milli-Q water and run with a flow rate of 0.5 mL/min for 60 min or until the UV absorbance reaches baseline (set sample valve to buffer (A or B), wash valve to column and outlet valve to waste). 11. Change solution to 20% EtOH and run 20–40 mL through the BE-SEC column. The column can thereafter remain connected to the chromatography system until the next run at room temperature, or for long-term storage, it should be tightly sealed and placed at 4  C. 3.2.4 Final Concentration Step

Option A: Tangential Flow Filtration (TFF) Concentration Step

To concentrate samples either TFF (100 kDa Cat. No. C02-E10005-N or 300 kDa Cat. No. C02-E300-05-N) filter or spin filters (10 or 100 kDa) can be utilized; see option A for TFF columns and option B for spin filters (see Notes 19–23). 1. Connect 50 mL syringes to the TFF column as shown in Fig. 2 and described in the next step.

Tangential Flow Filtration

295

Fig. 2 Schematic drawing of TFF column

2. Connect an empty 50 mL syringe to one end port (sample/ buffer port). Connect another 50 mL syringe to one of the side ports (waste port); this will be the waste outlet. Prime the TFF column with a 50 mL syringe containing 20–30 mL filtrated 0.01 M PBS by connecting it to the other end port (sample/ buffer port) and gently push the buffer through the column until 20–30 mL of sample is collected in the waste syringe. 3. Fill a 50 mL syringe with the sample and connect it to the sample end port. 4. Push the sample back and forth between the syringes connected to the sample end ports until the sample is concentrated down to 1–2 mL and aspirate what is left in the column. 5. The purified EVs can be stored for later or used immediately for subsequent experiments. 6. Optional: Flush the column with 5–10 mL of PBS from one side to remove any residual EVs stuck in the column (save in a separate tube at 4  C). TFF Column Washing

1. Connect clean 50 mL syringes to the TFF column as shown in Fig. 2. 2. Load 50 mL Milli-Q water into one of the syringes connected to the sample end port and push through from one side to the other side slowly. Continue until all solution is in the waste syringe. Be careful to not push too hard as it can damage the membrane. 3. Wash the column with 50 mL 0.1 M NaOH in the same manner. 4. Wash extensively with 100 mL Milli-Q water in the same manner. 5. Load 20 mL of 20% glycerin into the syringe and run it back and forth a few times.

296

Joel Z. Nordin et al.

6. De-attach the syringes, close the end and side port valves. Store at 4  C for later use. Option B: 10 kDa Spin Filter Concentration Step

Concentrate to a final volume of 200–500 μL by using an Amicon Ultra-15, 10 or 100 kDa molecular weight cutoff spin filter (Millipore) (see Notes 24–27). 1. Add your sample to the spin filter and centrifuge at 3000
g to a final volume of 200–500 μL. Start with 20 min centrifugation and if a smaller final volume is desired, extend the time further. 2. Optional: if a buffer exchange is needed for downstream analysis or for storage, this can be done now by adding the desired buffer and repeating the spinning procedure. For a complete buffer exchange, several spins may be required. 3. Transfer the sample to a clean low protein binding tube by pipetting. 4. The purified EVs can be stored for later or used immediately for subsequent experiments. 5. Optional: wash the membrane with clean/filtered PBS to collect any residual EVs from the membrane. Sample can be stored at 80  C.

4

Notes 1. All solutions are prepared with Milli-Q water. 2. For other cell sources, different seeding numbers are required and need to be optimized for different cell sources individually. 3. We recommend using low protein content medium for the 48 h incubation period to enhance any subsequent purification step. However, this needs to be optimized for each cell source. 4. For the immortalized MSCs, 100 mL of CM gives around 5
1010 EVs after the BE-SEC column purification. However, different cell sources produce different amounts of EVs; hence, the start volume needs to be adapted between different cell sources. 5. After the 700
g spin, pour the CM into new 50 mL tubes to avoid breakage of pelleted cells and cell debris that can affect the subsequent EV purification. 6. Stop at this step if necessary. The supernatant can be kept at 4  C for a week before proceeding to the next isolation step. These storage conditions have been validated for EV protein composition and EV numbers, however, not for RNA stability. 7. The subsequent EV purification steps have also been performed for other cell sources with similar results; however,

Tangential Flow Filtration

297

the initial cell seeding and cell medium may require optimization. 8. Always pre-clear your sample first by a low-speed centrifugation step (700
g for 5 min and 2000
g spin for 10 min) to avoid clogging of the system before filtration. 9. For step 3.2.1.1 (option A and B), instead of the 0.22 μm filtration step, a 10,000
g spin to pellet large vesicles can be performed for 10 min, giving similar results as the 0.22 μm filtration. 10. If a syringe and attached 0.22 μm filter are used for step 3.2.1.1 B, be careful to not apply excessive pressure on the syringe plunger and subsequently on the sample as it can cause vesicle rupture. The filter may need to be changed due to clogging. 11. For the TFF step, 100 kDa filters work equally well; however more proteins will be retained in the sample. Both 100 and 300 kDa filters give EV yields of nearly 100%. 12. Columns are designed for different filtration volumes. For example, 300 kDa MidiKros column is suitable for 100 mL to 3 L, while 300 kDa MicroKros column is suitable for 1–100 mL sample volumes. 13. To compare values of nanoparticles before and after TFF, save 100–1000 μL of CM sample and measure with nanoparticle tracking analysis (NTA). 14. You can stop after TFF if needed. Sample can be kept at 4  C overnight for further downstream steps. 15. The TFF filters are single use only. However, the filters can after washing be used again. The performance of the filters needs to be regularly checked if they are reused and discarded if more EVs are lost or if the protein amounts increase in the final product over time. 16. The column volume is 4.7 mL and bead height is 10 cm. If necessary, two columns or more can easily be connected in series to increase the binding capacity of the columns. 17. Priming overnight with a slow flow rate (e.g., 0.5 mL/min) can be done to make the column ready for usage the next morning. 18. If the absorbance spectrum starts to increase in the middle of the run, it is likely due to protein contamination because the columns’ binding capacity has been saturated. 19. 5–10 CVs (approx. 25–50 mL) of the sample can be loaded into the system using the sample valve inlet.

298

Joel Z. Nordin et al.

20. We recommend using the 100 kDa TFF filter here, since all proteins and debris should have been removed during earlier steps and this step is solely for concentrating the particles. 21. Remember to push gently, not to damage the column. 22. If the desired volume is larger than 1–2 mL, this step can be performed with the TFF machine, since the syringes are used to minimize the dead space. 23. During the concentration step either by TFF or spin filters the PBS can be exchanged to any buffer that is a suitable EV storage buffer or suitable for downstream analysis; thus this concentration step can also be utilized as a final buffer exchange step. 24. To compare values of particles before and after 100 kDa concentration step, save 10 μL of pre- and post-concentrated sample, dilute the sample accordingly in 0.01 M PBS (up to 1 mL) and analyze it with nanoparticle tracking analysis (NTA). 25. The spin filter can only accommodate 15 mL at a time; hence several spins may be needed. 26. Certain filters and EVs of some cell sources bind to the filters and are hard to remove. This can be avoided if the filters are pretreated with 0.3% albumin solution. 27. When recovering the sample from the filter, pipette up and down carefully to collect most of the EVs from the filter.

Conflict of Interest J.Z.N. and S.E.A. are consultants for and have equity interests in Evox Therapeutics Ltd. References 1. Ratajczak J, Miekus K, Kucia M et al (2006) Embryonic stem cell-derived microvesicles reprogram hematopoietic progenitors: evidence for horizontal transfer of mRNA and protein delivery. Leukemia 20(5):847–856. https://doi.org/10.1038/sj.leu.2404132 2. Valadi H, Ekstrom K, Bossios A et al (2007) Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat Cell Biol 9 (6):654–659. https://doi.org/10.1038/ ncb1596 3. Skog J, Wurdinger T, van Rijn S et al (2008) Glioblastoma microvesicles transport RNA and proteins that promote tumour growth and provide diagnostic biomarkers. Nat Cell Biol 10

(12):1470–1476. https://doi.org/10.1038/ ncb1800 4. El Andaloussi S, Mager I, Breakefield XO et al (2013) Extracellular vesicles: biology and emerging therapeutic opportunities. Nat Rev Drug Discov 12(5):347–357. https://doi. org/10.1038/nrd3978 5. Thery C, Amigorena S, Raposo G et al (2006) Isolation and characterization of exosomes from cell culture supernatants and biological fluids. Curr Protoc Cell Biol Chapter 3:Unit 3.22. https://doi.org/10.1002/ 0471143030.cb0322s30 6. Nordin JZ, Lee Y, Vader P et al (2015) Ultrafiltration with size-exclusion liquid chromatography for high yield isolation of extracellular

Tangential Flow Filtration vesicles preserving intact biophysical and functional properties. Nanomedicine 11 (4):879–883. https://doi.org/10.1016/j. nano.2015.01.003 7. Linares R, Tan S, Gounou C et al (2015) Highspeed centrifugation induces aggregation of extracellular vesicles. J Extracell Vesicles 4:29509. https://doi.org/10.3402/jev.v4. 29509 8. Momen-Heravi F, Balaj L, Alian S et al (2013) Current methods for the isolation of extracellular vesicles. Biol Chem 394(10):1253–1262. https://doi.org/10.1515/hsz-2013-0141 9. Li P, Kaslan M, Lee SH et al (2017) Progress in exosome isolation techniques. Theranostics 7 (3):789–804. https://doi.org/10.7150/thno. 18133

299

10. Corso G, Mager I, Lee Y et al (2017) Reproducible and scalable purification of extracellular vesicles using combined bind-elute and size exclusion chromatography. Sci Rep 7 (1):11561. https://doi.org/10.1038/ s41598-017-10646-x 11. Iwamoto S, Mihara K, Downing JR et al (2007) Mesenchymal cells regulate the response of acute lymphoblastic leukemia cells to asparaginase. J Clin Invest 117 (4):1049–1057. https://doi.org/10.1172/ JCI30235 12. Mihara K, Imai C, Coustan-Smith E et al (2003) Development and functional characterization of human bone marrow mesenchymal cells immortalized by enforced expression of telomerase. Br J Haematol 120(5):846–849

Chapter 19 Chromogenic Tissue-Based Methods for Detection of Gene Amplifications (or Rearrangements) Combined with Protein Overexpression in Clinical Samples Hiroaki Nitta and Brian Kelly Abstract Immunohistochemistry (IHC) is a well-established, tissue-based assay for the visualization of target proteins. For analysis of DNA targets, chromogenic in situ hybridization (CISH) applications have significant advantages over traditional fluorescence in situ hybridization (FISH). CISH slides can be analyzed using a regular light microscope, while FISH slides require the use of a specialized fluorescence microscope in a dark room. CISH slides allow observers to correlate the gene status (gene amplifications, gene rearrangements, and gene deletions) in the context of tissue morphology better than FISH slides. Recently, a combination of IHC and CISH assays (gene-protein assay, GPA) was developed to study the relationship between gene status and protein expression on the same tissue section. CISH and GPA applications can be optimized using an automated tissue slide processing system to generate reproducible results for a long and complex assay protocol. GPA applications are an ideal approach for tumor status and heterogeneity analyses for research and clinical investigations. Key words Gene amplification, Gene rearrangement, Protein overexpression, Gene-protein assay, Cancer, Personalized health care

1

Introduction Fluorescence in situ hybridization (FISH) methods are used for gene status analyses, such as gene amplification, gene rearrangement, and gene deletion in formalin-fixed, paraffin-embedded (FFPE) tissue sections. The observation of FISH slides requires a specialized fluorescence microscope in a dark room, and additional hematoxylin and eosin (H&E) slides are necessary to visualize the tissue morphology, define the tumor area, and interpret the gene signal. In order to overcome the cumbersome workflow of FISH assays, chromogenic in situ hybridization (CISH) applications were developed as clinical assays. Currently an automated HER2 dualcolor CISH assay (Roche Diagnostics, Indianapolis, IN, USA) is

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7_19, © Springer Science+Business Media, LLC, part of Springer Nature 2019

301

302

Hiroaki Nitta and Brian Kelly

available as a US Food and Drug Administration (FDA)-approved clinical test for selecting breast cancer patient candidates for the humanized monoclonal antibody trastuzumab (Herceptin®, Genentech, South San Francisco, CA, USA), while a manual dual-color HER2 CISH assay (ZytoVision GmbH, Bremerhaven, Germany) is available as an IVD/CE assay. Immunohistochemistry (IHC) assays are routinely utilized to visualize protein targets in tissue sections for research and clinical studies. Recently, we successfully combined IHC and CISH assays for the concurrent detection of HER2 [1] or anaplastic lymphoma kinase (ALK) protein [2] and gene targets on the same tissue sections to create a gene-protein assay (GPA). HER2 GPA studies of breast cancer cases demonstrated a previously unreported HER2-positive tumor cell population which is amplified for HER2 gene but lacks HER2 protein overexpression [3]. The HER2 GPA approach was very effective in determining HER2 status of double equivocal cases (HER2 IHC equivocal and HER2 FISH equivocal) [4]. Also, HER2 GPA testing resulted in the discovery of a new type of HER2 tumor heterogeneity in breast cancer [3] (Fig. 1). Furthermore, two retrospective clinical studies with breast cancer patients who were treated with trastuzumabbased chemotherapy showed significant correlation between HER2 intratumoral heterogeneity status and clinical outcome [5, 6]. Thus, the GPA technology is a new diagnostic tool for cancer biology investigations and is particularly suited to illuminate tumor heterogeneity. Protocols for CISH and GPA assays are long and convoluted. To obtain reproducible CISH and GPA assay results, the use of an

Fig. 1 Automated gene-protein assay (GPA) for human epidermal growth factor receptor 2 (HER2) gene, chromosome 17 centromere (CEN17), and HER2 protein (brown staining) of breast carcinomas. HER2 genetic heterogeneity (a) and HER2 nongenetic heterogeneity (b) cases are presented. HER2 genetic heterogeneity is a mixture of HER2-negative tumor cells and HER2-positive tumor cells (amplified HER2 gene and overexpressed HER2 protein), while HER2 nongenetic heterogeneity is a mixture of HER2-positive tumor cells and non-classic HER2-positive tumor cells (amplified HER2 gene and no HER2 protein). 60 objective

Chromogenic In Situ Hybridization

303

automated slide processing system is recommended to avoid human error. Based on our successful experiences developing CISH and GPA assays for gene amplifications [1, 7] and gene rearrangements [2, 8], we describe the importance of CISH and GPA assay optimization using an automated slide processing instrument and also for potential manual procedures in this chapter. However, because even today the pre-analytical tissue processing steps have not been standardized for FFPE samples, it is impossible to have one protocol that works on all clinical tissue samples. Therefore, this chapter should be used as a source of ideas to optimize a CISH or GPA assay in a laboratory.

2

Materials Formalin-fixed, paraffin-embedded xenograft tumors of wellcharacterized cell lines are ideal control samples for optimizing a CISH assay. Commercially available reagents are listed with specific vendors wherever possible. However, this is not a complete list of commercial reagents, and alternative reagents can be found from other vendors. Performance of laboratory-made reagents must be carefully evaluated with working protocols.

2.1

Clinical Samples

1. Formalin-fixed, Note 1).

paraffin-embedded

clinical

samples

(see

2.2 Primary Antibodies

1. Monoclonal or polyclonal antibodies for immunohistochemical assays from any vendors.

2.3 Hapten-Labeled DNA Probes (See Note 2)

1. ZytoDot2C SPEC probes (ZytoVision GmbH, Germany) (see Note 3).

2.4 Histology Reagents

1. Histology grade xylene from any vendors.

2. FISH probes (Abnova, Taiwan, and Empire Genomics, Inc., New York, USA) (see Note 4).

2. Histology grade alcohol from any vendors. 3. Mayer’s hematoxylin from any vendors. 4. Histomount mounting solution (HS-103, National Diagnostics, Atlanta, GA, USA).

2.5 Pretreatment Reagents (See Note 5)

1. Heat pretreatment solution Target retrieval solution (S1700, Agilent, Santa Clara, CA, USA). 2. SPOT-Light Tissue Pretreatment Kit (00-8401, Thermo Fisher Scientific, Waltham, MA, USA)

304

Hiroaki Nitta and Brian Kelly

This kit contains a ready-to-use heat pretreatment solution and a ready-to-use enzyme reagent. 3. Protease Proteinase K, ready-to-use (PRK-004, Spring Bioscience, Inc., Pleasanton, CA, USA, or similar products from other vendors). Pepsin, ready-to-use (S-99003, 42 Life Sciences GmbH, Germany, or similar products from other vendors). 2.6 Immunohistochemistry Detection

1. Appropriate commercially available horseradish peroxidase (HRP) or alkaline phosphatase (AP) detection system for immunohistochemistry assays (see Note 6).

2.7 Hybridization Buffer (See Note 7)

1. In situ hybridization buffer, 1.25 concentrate (ENZ-33808, Enzo Life Sciences, Farmingdale, NY, USA, or similar products from other vendors)

2.8 Stringency Wash Solution

1. SSC buffer, 20 concentrate (S6639-1L, Sigma-Aldrich, St. Louis, MO, USA, or similar products from other vendors).

2.9 CISH Immunodetection Reagents (See Note 8)

1. AP- and/or HRP-conjugated antibodies for the haptens used to label the probes (a direct detection method) or unconjugated antibodies for the haptens and AP- and/or HRP-labeled anti-species antibodies (an indirect detection method) (Vector Laboratories, Inc., Burlingame, CA, USA, and other vendors). 2. Antibody diluent (ADS-125, Spring Bioscience, or similar products from other vendors). 3. Rinse solution PBS with Tween 20 (PBT-999, Spring Bioscience, or similar products from other vendors). TBS with Tween 20 (TBT-999, Spring Bioscience, or similar products from other vendors). 4. Chromogen kits for AP- or HRP-based detection from various vendors of histology detection reagents (see Note 9). 5. ZytoDot2C CISH Implementation Kit (C-3044-40, ZytoVision GmbH, Germany) (see Note 10). 6. ZytoDot2C SPEC HER2/CEN17 Probe Kit (C-3022-40, ZytoVision) (see Note 11).

2.10 Instruments (See Note 12)

1. ThermoBrite System (07J91-010, Abbott, Abbott Park, IL, USA). 2. Hybridizer (Agilent). 3. PCR thermocycler with slide blocks from any vendors.

Chromogenic In Situ Hybridization

3

305

Methods Dual-color CISH assays allow detection of two DNA targets on the same tissue section. In general, manual dual-color CISH applications are labor-intensive 2-day protocols (Fig. 2). Paraffin sections must be treated to retrieve DNA targets so that probes may be hybridized effectively to the targets; while tissue morphology is preserved as much as possible. The most effective way to unmask the DNA targets is a combination of heat pretreatment and protease digestion (after removing paraffin from tissue sections). A pair GPA

Day 1

Deparaffinization

Antigen Retrieval

CISH Deparaffinization

Primary Antibody Incubation

Heat Treatment

IHC Detection with HRP or AP

Protease Digestion

Probe co-hybridization

Day 2 Stringency Wash

First CISH Detection with HRP

First CISH Detection with AP

First CISH Detection with AP

Second CISH Detection with AP

Second CISH Detection with HRP

AP Inactivation

Counterstaining

Second CISH Detection with AP

Air-drying

Coverslipping

Microscopy

Fig. 2 A simplified flowchart for dual-color chromogenic in situ hybridization (CISH) assays using horseradish peroxidase (HRP)- and/or alkaline phosphatase (AP)-based immunological signal detection and gene-protein assay (GPA) which is a combination of immunohistochemistry (IHC) and CISH assays. IHC assay is performed first followed by a CISH assay

306

Hiroaki Nitta and Brian Kelly

Fig. 3 Automated dual-color chromogenic in situ hybridization (CISH) application for human epidermal growth factor receptor 2 (HER2) gene and chromosome 17 centromere (CEN17) of breast carcinomas. Normal HER2 gene (a) and amplified HER2 gene (b) cases are presented. HER2 gene was detected with horseradish peroxidase (HRP)-based silver detection (as seen with black dots), and CEN17 was detected with alkaline phosphatase (AP)-based fast red detection (as seen with red dots). 100 objective

of DNA probes are labeled with two different haptens, such as digoxigenin (DIG) and 2,4-dinitrophenyl (DNP), and the haptens are then histochemically visualized with HRP- and/or AP-based detection systems. Typically, a combination of HRP- and AP-based detections is used for dual-color CISH assay. However, a combination of two AP-based detections can be achieved by inactivating the AP enzyme of the first CISH detection prior to the second AP-based CISH detection. A combination of two HRP-based detections is feasible; however, there are more possible colors commercialized for AP detections than HRP detections. A combination of two distinct colors should be selected for a dual-color CISH assay to facilitate easier analysis (Fig. 3), particularly for gene rearrangement CISH assays. Due to the lack of tissue processing standardization, the optimization of CISH assays must be vigorously evaluated. The heat pretreatment and protease digestion steps to retrieve the DNA targets and manufacturer recommended protocols for dual-color CISH assays need to be modified for further assay optimization. This is one of the main reasons why the use of an automated slide staining system is highly recommended to improve the reproducibility of dual-color CISH assays. An IHC assay and a CISH assay can be combined as a GPA on the same tissue section in order to analyze membrane or cytoplasmic target protein and gene status simultaneously at individual cell level. Immunohistochemical detection of a target protein must be conducted prior to a CISH protocol. The hybridization step can result in the antigenicity of a protein target being destroyed, and the detection of protein target would be compromised for accurate

Chromogenic In Situ Hybridization

307

IHC detection. When multiplex assays are performed, a selection of color combinations is significantly important, in addition to the sensitivity of assays. Good contrast of colors used in a multiplex assay helps to analyze several targets under the microscope without the use of image analysis software. 3.1

Deparaffinization

1. Bake the glass slide with tissue sections above the melting point of paraffin (approximately 5  C above the melting point) in the oven for 15 min to improve the tissue section adhesion. 2. Soak the slide in a Coplin jar of histology grade xylene for 5 min at room temperature. Transfer the slide to fresh xylene and soak for 5 min. 3. Rinse the slide to remove xylene in a Coplin jar of 100% histology grade alcohol for 2 min at room temperature. Transfer the slide to fresh 100% alcohol and soak for 2 min (see Note 13). 4. Transfer the slide to a Coplin jar of 80% alcohol and soak for 2 min at room temperature. 5. Rinse the slide in a Coplin jar of distilled water for 2 min at room temperature. Transfer the slide to fresh distilled water and soak for 2 min.

3.2 Heat Pretreatment for IHC (See Note 14)

1. Place the slide in preheated heat pretreatment solution in a heat-resistant container according to manufacturer recommendations. 2. Cool the slide according to manufacturer recommendations.

3.3 Primary Antibody Incubation (See Note 15)

1. Apply a primary antibody in a diluent at an appropriate antibody concentration, and incubate for an appropriate length in a humidity controlled chamber/box at room temperature. 2. Rinse the slide in a Coplin jar of a rinse buffer at room temperature. Repeat rinsing steps twice.

3.4 Immunohistochemical Detection (See Note 16)

1. Follow immunohistochemical detection steps according to manufacturer recommendations in a humidity controlled chamber/box at room temperature. 2. Apply a freshly prepared chromogen reagent mixture to the slide, and incubate in a humidity chamber/box at room temperature. Observe the color development under the light microscope, and terminate the reaction by soaking in a rinse buffer when it is appropriate. 3. Rinse the slide in a Coplin jar of a rinse buffer at room temperature. Repeat rinsing steps twice.

308

Hiroaki Nitta and Brian Kelly

3.5 Heat Pretreatment for CISH

1. Place the slide in preheated heat pretreatment solution in a heat-resistant container according to manufacturer recommendations. 2. Cool the slide recommendations.

3.6 Protease Digestion (See Note 17)

according

to

the

manufacturer

1. Apply a pepsin or proteinase K solution to the entire tissue section, and incubate for 3–5 min in a humidity controlled chamber/box at room temperature. 2. Rinse the slide in a Coplin jar of distilled water for 2 min at room temperature. 3. Rinse the slide in a Coplin jar of 100% alcohol for 2 min at room temperature. Transfer the slide to fresh 100% alcohol and soak for 2 min. 4. Remove the slide and air-dry completely.

3.7 Hybridization (See Note 18)

1. Apply appropriate amount of a probe cocktail in a hybridization buffer onto the tissue section. 2. Cover the tissue section with a coverslip, carefully avoiding air bubbles under the coverslip. 3. Seal the edge of coverslip with rubber cement to avoid evaporation during denaturing and hybridization steps. 4. Dry the rubber cement for 15 min or until it dries completely at room temperature. 5. Place the covered slides on a PCR thermal cycler with slide blocks or a hybridization instrument. 6. Denature the probe and target at 95  C for 5 min. 7. Hybridize overnight at 37  C.

3.8

Stringency Wash

1. Remove the rubber cement from the slide. 2. Soak the slides in a Coplin jar of 2 SSC at room temperature until the coverslip starts to fall off from the slide. 3. Transfer the slide to a Coplin jar of 1 SSC for 5 min at 45  C. Repeat two additional washes in fresh 1 SSC for 5 min each. 4. Soak the slide in a Coplin jar of a rinse solution for 5 min at room temperature.

3.9 CISH Signal Detection (See Note 19)

1. Remove the slide and apply an antibody for one of the haptens onto the slide, and incubate for approximately 15 min in a humidity chamber/box (see Note 20) at room temperature. If the first hapten to visualize is DNP, then an anti-DNP antibody is applied. For a direct detection with the use of an enzyme-conjugated anti-hapten antibody, skip steps 2 and 3.

Chromogenic In Situ Hybridization

309

2. Rinse the slides with a rinse solution in a Coplin jar, twice at room temperature. 3. Apply enzyme-conjugated anti-species antibody onto slide, and incubate for approximately 15 min in a humidity chamber/box at room temperature. If the primary antibody is a rabbit antibody, an enzyme-conjugated anti-rabbit antibody is applied. 4. Rinse the slides with a rinse solution in a Coplin jar, three times at room temperature (see Note 21). 5. Apply a freshly prepared chromogen reagent mixture to the slide, and incubate in a humidity chamber/box at room temperature. Observe the color development under a light microscope, and terminate the reaction by soaking in a rinse solution when it is appropriate (see Note 22). 6. For a dual CISH assay with two AP-based detections, incubate the entire tissue section with a hybridization buffer for 30 min in a humidity chamber/box at room temperature. 7. Rinse the slide very well with a rinse buffer in a Coplin jar for 5 min, three times at room temperature. 8. Remove the slide and apply an antibody for the other hapten onto the slide, and incubate for approximately 15 min in a humidity chamber/box at room temperature. For a direct detection with the use of an enzyme-conjugated antibody, skip steps 9 and 10. 9. Rinse the slides with a rinse solution in a Coplin jar, twice at room temperature. 10. Apply an enzyme-conjugated anti-species antibody onto the slide, and incubate for approximately 15 min in a humidity chamber/box at room temperature. If the second anti-hapten antibody is a mouse antibody, an enzyme-conjugated antimouse antibody is applied. 11. Rinse the slide with a rinse solution in a Coplin jar, three times at room temperature. 12. Apply a freshly prepared chromogen reagent mixture to the slide, and incubate in a humidity chamber/box at room temperature. Observe the color development under the light microscope as needed, and terminate the reaction by soaking in a rinse solution when it is appropriate. 3.10 Counterstaining and Mounting (See Note 23)

1. Transfer the slide to a Coplin jar of distilled water. 2. Soak the slide in a Coplin jar of Mayer’s hematoxylin solution for 5 s at room temperature. 3. Rinse the slides with running tap water in a Coplin jar for 2 min.

310

Hiroaki Nitta and Brian Kelly

4. Rinse the slides with distilled water in a Coplin jar for 2 min, twice at room temperature. 5. Air-dry the slides on a bench until completely dried or dry in an oven at 65  C for 15 min. 6. Coverslip the slides with a permanent mounting solution, such as Histomount.

4

Notes 1. Any archived formalin-fixed, paraffin-embedded tissue blocks can be used to prepare tissue sections. However, there are cases that would never demonstrate successful CISH (or FISH) signals. Since, in general, information of how tissue samples were processed for paraffin-embedding is very difficult to obtain, the exact cause of unsuccessful CISH assays cannot be clearly identified. The delay of placing tissue samples in a fixative is known to create difficulty for successful molecular histopathology assays. Tissue sections should be cut at 4–5 μm and placed onto Superfrost® Plus slides (Erie Scientific Company, Portsmouth, NH, USA) or similar glass slides. The tissue sections can be stored in a slide box at room temperature for a limited time. 2. Hapten-labeled DNA probes for FISH assays can be used for CISH assays, and those probes can be purchased from various vendors. If commercial hapten-labeled DNA probes are not available, DNA probes can be labeled with haptens by nick translation or by PCR probe labeling in a laboratory. Nick translation can be utilized for labeling DNA probes with haptens by using commercially available kits. New probes must be analyzed for their specificity by FISH or CISH assays using a comparative genomic hybridization metaphase control slide. 3. ZytoVision ZytoDot2C SPEC probes are labeled with DIG and DNP haptens, and the probe hybridization sites can be visualized with a CISH detection kit including anti-DIG and anti-DNP antibodies. 4. Abnova FISH probes are labeled with fluorescein isothiocyanate (FITC) and Texas Red (sulforhodamine 101 acid chloride) haptens, and the probe hybridization sites can be visualized with a CISH detection kit including anti-FITC and anti-Texas Red antibodies. 5. In order to accomplish GPA protocol, two pretreatment steps must be optimized: (1) for IHC and (2) for CISH. In general, pretreatment step for IHC assays is a heat-based antigen retrieval using a commercially available solution, while CISH assays require a combination of a heat pretreatment and

Chromogenic In Situ Hybridization

311

protease digestion. Using an automated slide processing system, we investigated tissue pretreatment conditions for CISH assays. We learned that the best pretreatment method is a combination of a heat pretreatment followed by a protease digestion with proteinase K or pepsin. Heat pretreatment alone or protease digestion alone is not adequate to retrieve DNA targets while preserving good tissue morphology. Unfortunately, similar to tissue processing protocols, the pretreatment conditions are not standardized. Because of unstandardized tissue processing protocols, a robust pretreatment condition must be established in each laboratory. Also, the pretreatment protocol may need to be tweaked for different tissue samples. 6. There are numerous commercial IHC detection kits available. The most important decision in selecting an IHC kit is the color of the detection system. Ideally it would provide a good color contrast to the color combination of the CISH detection. Also, the color of an IHC detection system should provide a good contrast to counterstaining. 7. In situ hybridization buffer is used for two purposes: (1) probe denaturing and hybridization and (2) AP enzyme inactivation between two sequential AP-based CISH detections. As we previously reported [8], we learned that an AP enzyme can be inactivated by incubating the tissue section with an in situ hybridization buffer containing formamide for 30 min at room temperature without losing probe-target hybridization. Thus, after the first AP-based CISH detection, an in situ hybridization buffer should be applied to inactivate prior to the second AP-based CISH detection. Since formamide can denature proteins, the slides must be washed well with a rinse solution before the second immunodetection step. 8. You can choose to build your own CISH detection kit from individual reagents sourced from various vendors or to purchase a commercial CISH kit. Selecting the right color combination with good color contrast is a key for successful CISH assay detection. Color blindness is an important factor in deciding a color combination. Common color combinations are (1) blue (AP-based detection) and red (AP-based detection), (2) black (HRP-based detection) and red (AP-based detection), and (3) green (HRP-based detection) and red (AP-based detection). Common haptens used to label DNA probes for CISH assays are biotin, DIG, DNP, FITC, and Texas Red. Based on the size of your DNA targets, you may choose from a direct detection or an indirect detection for CISH assays. In general, an indirect detection system can provide better sensitivity compared to a direct detection system. For an indirect detection, you need to select a combination of

312

Hiroaki Nitta and Brian Kelly

two antibodies raised with two different animal species, such as a mouse anti-DIG antibody and a rabbit anti-DNP antibody, so that enzyme-labeled anti-mouse antibody and anti-rabbit antibody can be applied for signal detection. If a direct CISH detection is going to be applied, anti-hapten antibodies raised in the same animal species that are labeled with either AP or HRP enzyme molecules can be used. Antibodies should be diluted according to manufacturer recommendations with an antibody diluent of your choice. 9. AP- and HRP-based chromogen kits can be obtained from various histology companies. Vector Laboratories, Inc. (Burlingame, CA, USA), carries the most extensive product line for both AP- and HRP-based chromogen kits that can be used for CISH detection. 42 Life Sciences GmbH (Germany) also carries both AP- and HRP-based chromogen kits. 10. ZytoVision ZytoDot2C CISH Implementation Kit contains all necessary CISH detection reagents for the detection of DIGand DNP-labeled DNA probes on formalin-fixed tissue sections. 11. ZytoVision ZytoDot2C SPEC HER2/CEN17 Probe Kit contains all necessary CISH detection reagents for the detection of HER2 and CEN17 DNA targets on formalin-fixed tissue sections. This is an ideal kit to explore manual dual-color CISH assay detection. 12. The listed in situ hybridization instruments are for manual CISH applications, particularly for denaturing and hybridization steps. Moisture pads can be placed to create humidity during hybridization, but we found the best way to avoid evaporation is to seal coverslips with rubber cement. Automated slide processing systems that allow hybridization steps are also available from several manufactures, such as Ventana Medical Systems, Inc. (Tucson, AZ, USA) and Leica Microsystems GmbH (Germany). Ventana offers software and reagents for automated CISH assays. 13. Instead of rehydrating tissue sections by using a series of alcohol solutions, they can be simply air-dried prior to a heattreatment step. 14. An important thing is to find an appropriate preheating step without destroying the tissue sections and to cool the slides. 15. Primary antibody dilutions must be carefully titrated for a maximum signal to noise ratio using appropriate positive and negative tissue sections. A diluent selection is also an important factor for successful IHC assay optimization with each primary antibody.

Chromogenic In Situ Hybridization

313

16. A humidity chamber/box can be assembled by placing wet paper towels in a tray and covering it with a plastic wrap sheet. However, Slide Master (ScyTek Laboratories, Inc., Logan, UT, USA) and Stain Tray Slide Staining System (Simport Scientific, Inc., Quebec, Canada) are ideal moisture chambers for reagent incubation steps as well as the washing steps of an IHC detection. A hydrophobic pen known as a PAP pen is useful for reagent incubation steps and saving reagents in IHC assays. The hydrophobic marking is removed with xylene after completing IHC detection. 17. Heat pretreatment alone is not adequate to retrieve DNA targets in tissue sections for CISH assays. Protease digestion must be carefully controlled for the best CISH signal and wellpreserved tissue morphology. Pepsin or proteinase K is commonly used for digesting tissue sections for DNA target retrieval: over-digestion of tissue sections will result in poor tissue morphology and less CISH signals. The use of non-tumor cells near a tumor cell population as an internal assay performance control is the best way to evaluate the success of CISH assay. Non-tumor cells should show one to two copies of a target DNA. 18. Hybridization step is a technically challenging step in a manual protocol. The amount of probe solution needs to be adjusted according to the size of the coverslip. Appropriate probe volumes are 10 μL for 22  22 mm coverslips, 20 μL for 22  30 mm coverslips, and 30 μL for 22  50 coverslips. Sealing the coverslip with rubber cement is not an easy task around the edge of coverslips with excess probe solution. As a larger size of coverslip is used, there are more chances to create air bubbles under the coverslip. In order to obtain consistent CISH signals, air bubbles must be avoided as much as possible. When the rubber cement does not seal the edge of coverslips, air bubbles will be created during the denaturing step. Air bubbles tend to disappear once denaturing temperature goes down to hybridization temperature. However, areas of tissue sections with air bubbles during hybridization will not have CISH signals. 19. A hydrogen peroxide (H2O2) blocking step is often used for immunohistochemical applications to inactivate endogenous peroxidases, but we have not encountered the necessity of a H2O2 blocking for CISH assays. Also, a protein blocking step is often incorporated prior to antibody incubation steps for immunohistochemical assays, but we found that it is not so important for CISH assays as long as the antibody concentrations are appropriate. The background staining caused by antibodies can be reduced or eliminated by adding casein in an

314

Hiroaki Nitta and Brian Kelly

antibody diluent. However, protein blocking may also reduce CISH signal. 20. Tissue sections should not be dried during immunological CISH detection (as described in Note 16) for immunohistochemical detection. However, because the use of xylene should be avoided for AP-based CISH detection, a PAP pen is not recommended in order to preserve CISH signal. 21. Washing steps prior to a chromogen reaction are more important than washing steps between antibody incubation steps. Use plenty of a rinse solution to remove excess enzymeconjugated antibodies to lower background staining. Gentle washing is important to preserve better tissue morphology. 22. Wear disposable gloves while handling slides covered with chromogen, particularly 3,3-diaminobenzidine (DAB) which is a well-known mutagen. 23. The purpose of counterstaining for CISH slides is to visualize the nuclei so that an observer can correlate the tissue morphology and CISH signal. Very weak counterstaining is ideal to avoid obscuring CISH staining. Also, bluing reagent for hematoxylin counterstaining might eliminate CISH staining dependent on which chromogen was used for immunodetection. References 1. Nitta H, Kelly BD, Padilla M, Wick N et al (2012) A gene-protein assay for human epidermal growth factor receptor 2 (HER2): brightfield tricolor visualization of HER2 protein, the HER2 gene, and chromosome 17 centromere (CEN17) in formalin-fixed, paraffin-embedded breast cancer tissue sections. Diagn Pathol 7:60. https://doi.org/10.1186/1746-1596-7-60 2. Nitta H, Tsuta K, Yoshida A et al (2013) New methods for ALK status diagnosis in non-smallcell lung cancer: an improved ALK immunohistochemical assay and a new brightfield dual ALK IHC-in situ hybridization assay. J Thorac Oncol 8:1019–1031. https://doi.org/10.1097/JTO. 0b013e31829ebb4d 3. Nitta H, Kelly BD, Allred C et al (2016) The assessment of HER2 status in breast cancer: the past, the present, and the future. Pathol Int 66:313–324. https://doi.org/10.1111/pin. 12407 4. Hou Y, Nitta H, Li Z (2017) HER2 gene protein assay is useful to determine HER2 status and evaluate HER2 heterogeneity in HER2 equivocal breast cancer. Am J Clin Pathol 147:89–95. https://doi.org/10.1093/ajcp/ aqw211 5. Hou Y, Nitta H, Wei L et al (2017) HER2 intratumoral heterogeneity is independently

associated with incomplete response to antiHER2 neoadjuvant chemotherapy in HER2positive breast carcinoma. Breast Cancer Res Treat 166:447–457. https://doi.org/10. 1007/s10549-017-4453-8 6. Horii R, Matsuura M, Nitta H et al (2018) Clinical significance of HER2 intratumoral heterogeneity, determined by simultaneous gene and protein analysis, in HER2-positive breast cancer. Eur J Cancer 92:S150. https://doi. org/10.1016/S0959-8049(18)30675-0 7. Nitta H, Hauss-Wegrzyniak B, Lehrkamp M et al (2008) Development of automated brightfield double in situ hybridization (BISH) application for HER2 gene and chromosome 17 centromere (CEN 17) for breast carcinomas and an assay performance comparison to manual dual color HER2 fluorescence in situ hybridization (FISH). Diagn Pathol 3:41. https://doi. org/10.1186/1746-1596-3-41 8. Nitta H, Zhang W, Kelly BD et al (2010) Automated brightfield break-apart in situ hybridization (ba-ISH) application: ALK and MALT1 genes as models. Methods 52:352–358. https://doi.org/10.1016/j.ymeth.2010.07. 005

INDEX A Affinity chromatography............122, 124, 125, 131, 132 Alginate................................................153, 155–157, 160 Amino acid resolution..................................................... 24 Antago-miR ................................................................... 273 Antibody ....................................................... 45, 106, 121, 166, 186, 233, 245, 255, 273, 302 Anti-HER2 ...................................................125, 132–134 Anti-miR ........................................................................ 273 Apoptotic bodies ........................................................... 287

B Bind-elute size exclusion chromatography (BE-SEC).................................................. 287–298 Bioassay............................................................................ 63 Bio-layer interferometry (BLI)..................................... 135 Bioreactor .....................................................155–157, 173 Bone marrow.............................................. 186, 192, 201, 203, 232, 235–237, 239, 259, 260, 290

C Cancer..................................................... 34, 92, 139–149, 152, 153, 157, 163–165, 168, 169, 174, 175, 177, 183–208, 231, 242, 265, 272, 276, 279, 281, 282, 302 Cancer-associated fibroblasts (CAF) .......... 152, 164, 175 Cancer stem cell (CSC) ...............................139–149, 186 Capsules ............................. 156, 157, 160, 167, 168, 173 Carboxyfluorescein succinimidyl ester (CFSE) .. 155, 158 Cas9 .......................................................... 24, 29, 33, 106, 108, 109, 184, 185, 213–229 Catenin ........................................................ 65, 74, 81, 82 CDK9 ................................................................... 108, 115 Cellomics ......................................................................... 44 CellProfiler ................................................................43–59 CellProfiler Analyst ......................................................... 46 CellTracker Blue................................................... 155, 158 Centrosome amplification ................................. 34–36, 40 Centrosome clustering.................................................... 34 Chemical mutagenesis ......................................... 4, 23–31 Chemogenomic .................................................... 3, 23, 24 Chromogenic in situ hybridization (CISH) ..................................... 301–306, 308–314

Clonogenic assay ........................................................... 145 Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)....................................v, 24, 29, 33, 57, 106, 109, 184, 185, 213–229 Co-culture .................................. 153, 156–160, 163–178 Collagen............................. 154, 155, 158–160, 174, 176 Common lymphoid progenitor (CLP) ............... 232–239 Compensation ..................................................... 206, 207, 254, 255, 263, 264, 266 Confocal imaging ................................167, 169, 172, 176 CRISPR-RNA (crRNA) ............ 213, 218, 220–224, 228 Culture supernatant ............................................ 124, 125, 131, 132, 135, 288, 291–296 Cytotoxic T-lymphocyte-associated Protein 4 (CTLA4) ............................................................ 152

D Degradation tag (dTAG) ..................................... 105–119 DNA-binding .................................................................. 65 DNA digest ...............................................................14, 17 Drug screening............................................ 140, 141, 153 Drug target............................................................. v, 3–20, 23–31, 89–99, 184–186, 196 Drug target interaction site mapping ............................ 29 Drug target profile ....................................................89–99 Druggability ..............................................................63–83

E Embryo injection .......................................................... 219 Embryonic stem (ES) cells ............ 4, 9, 10, 30, 185, 218 Exosomes....................................................................... 287 Extracellular matrix (ECM)................................ 152, 153, 164, 165, 174, 200 Extracellular vesicles............................................. 287–298

F Fetal calf (FC) fusion ........................................... 121–135 Floater................................................................... 154, 155 FLO-QXP ..............................................71, 72, 76, 82, 83 Flow cytometry ................................. 3, 7, 157, 172, 173, 177, 178, 186, 193, 196–198, 203–207, 236–238, 242–244, 247–250, 253–267

Ju¨rgen Moll and Sebastian Carotta (eds.), Target Identification and Validation in Drug Discovery: Methods and Protocols, Methods in Molecular Biology, vol. 1953, https://doi.org/10.1007/978-1-4939-9145-7, © Springer Science+Business Media, LLC, part of Springer Nature 2019

315

TARGET IDENTIFICATION

316 Index

AND

VALIDATION

IN

DRUG DISCOVERY: METHODS

Fluorescence activated cell sorting (FACS) ................3, 7, 9, 10, 26, 112, 196, 199, 205, 233, 235, 239, 245, 248, 255, 260–262 Fluorescence minus one (FMO) ........206, 262–263, 265 Functional genomic ...................................................... 3, 4

G gBlock ..................................................122, 125, 127, 128 Gene amplifications.............................................. 301–314 Gene expression ................................................5, 98, 165, 168, 173, 200, 231, 248, 271 Gene protein assay (GPA) ..................302, 305, 306, 310 Gene rearrangement ...........................207, 301, 303, 306 Genetic engineered mouse model (GEMM)..... 185–187, 189–193, 208 Genetic screen ......................................... 3–5, 23–31, 141 Gene trap ................................5, 7–10, 12, 14, 16, 17, 24 Genome ...........................................................4, 5, 14, 17, 18, 24, 29, 44, 70, 108, 184, 185, 213–229 Geometry based pocket finding ..................................... 68 Grid force field ................................................................ 77 Growth factors ...................................................... 45, 125, 140–143, 174, 302, 306

H Haploid .................................................................... v, 3–20 Hematopoietic stem cells (HSC) ....... 241–244, 248–250 High-content analysis (HCA) ............................. 172, 173 High-content screening (HCS) ........................ 34, 43–59 High-throughput screening (HTS) ....................... 33, 36, 43, 55, 64, 89 Hot spots .........................................64, 71, 73–76, 81–83 Humanized mouse model .................................................v

I Image processing................................................ 43, 47, 48 Immune cell........................................122, 152–160, 164, 196, 205, 206, 242–244, 247–249, 253, 254, 264 Immunodeficient..........................................241–242, 250 Immunofluorescence (IF)...................119, 166, 169, 205 Immunohistochemistry (IHC)....................................166, 169–171, 195, 197, 276, 279–281, 302, 304–307, 310–312 Immunophenotyping........................................... 253–267 Infiltrations ................................. 152–160, 244, 247–249 Inhibitor ............................................................34, 36, 40, 72–76, 91, 92, 96–98, 106, 141, 153, 157, 159, 174, 184, 186, 187, 192–196, 244, 273 Innate lymphoid cells (ILC)................................ 231–239 In silico ............................................ v, 63–83, 89–99, 110 In situ................................. 271–283, 301, 304, 311, 312 Interactome ....................................................................... 3 Intracellular staining ................................... 233, 238, 263

AND

PROTOCOLS

In vitro ........................................................ 105, 132, 140, 144, 145, 153, 164, 173, 174, 177, 186, 190, 213, 218, 232, 242 In vivo .............................................................. v, 105, 106, 140, 153, 165, 183–208, 218, 231–239 Isogenic cell lines ......................................................33, 37

K KBM7 ..................................................108, 112, 115, 117 Ki67 ...................................................................... 171, 250 KNIME......................................................................43–59 Knock-in (KI)...................................................... 105–119, 183, 186, 188, 191, 214–218, 220, 227 Knockout (KO) ................................................... 183, 184, 188, 190, 191, 226–228

L Ligands .............................................................. 64, 70–74, 76, 77, 79, 81–83, 94, 96, 97, 106, 107, 114, 174 Live-cell analysis ............................................................ 170 Locked nucleic acid (LNA) ................273, 276, 279, 282 Long-term self-renewal assay ....................................... 146

M Machine learning (ML) .......................................... 46, 49, 70, 91, 92, 99, 197 Magnetic resonance imaging (MRI).................. 185, 188, 191, 202, 203 Mammalian expression ....................... 122–123, 125–127 Mammalian target of rapamycin (mTOR)................... 157 MAP-kinase ................................................................... 159 Matrigel ............................................................... 142, 143, 146–148, 154, 155, 158–160, 174, 198, 200 Microinjection ............................................. 216, 219, 224 MicroRNAs (miRNAs) ........................................ 271–283 Microvesicles ................................................................. 287 Mitosis ............................................................................. 34 Monoclonal antibody (mAb) ..............37, 250, 273, 274, 276–279, 282, 302 Mouse models ............................................. 106, 214, 242 Multicolour flow cytometry ................................ 253–267

N Necropsy ..............................................191, 193, 202, 204 Neural stem cells (NSCs)..................................... 139–149 Neural stem cells (NSCs) cultures .............................. 141 Neurospheres..............................140, 141, 144, 145, 148 NK cells................................................................ 153, 154, 158, 159, 196, 233, 238 Non-obese diabetic (NOD) mice ...............184, 241–242 Non-small cell lung cancer (NSCLC)........ 152, 153, 159 NSG-mice .......................... 184, 189, 192, 193, 244, 248

TARGET IDENTIFICATION

AND

VALIDATION

O Off-targets ..................................24, 33, 77, 78, 217, 218 Open-source software ................................. 46, 47, 58, 59

P Pancreatic cancer ................................................. 164, 165, 167, 168, 174, 175, 177 Panel docking .................................................................. 91 Patient derived xenograft (PDX) ....................... 164, 174, 186–188, 192–197, 203, 207, 208, 241–250 Peripheral blood.........................186, 242, 243, 254, 259 Pharmacodynamic (PD) ............................. 188, 196, 197 Pharmacokinetic (PK)..................................................... 70 Pharmacology.......................................... 80, 99, 183–208 Pharmacophore .............................. 64, 77, 90, 91, 95, 96 Phenotypic readout ............................................ 34, 36, 38 Physicochemical properties......................... 64, 70, 71, 79 Plasmids ............................................................... 7, 10, 19, 106–109, 111–113, 117, 123–125, 127–130, 133, 157, 220, 234 Pocket finder ...................................................... 66–68, 70 Polybrene...........................................8, 12, 234, 236–237 Primary culture............................................ 140, 144, 145 Primers............................................... 14, 17, 19, 20, 107, 109–111, 113, 123, 128, 129, 194, 217, 218, 226 Programmed cell death protein 1 (PD1) ...................152, 186, 187, 265 PROTACS ..................................................................... 106 Protein cavity........................................................... 64, 66, 67, 71, 76, 81, 82 Protein degradation ............................................. 106, 118 Protein-protein interaction (PPI) ...............64, 71–75, 79 Proteomics.................................................................89, 92

Q QSAR ............................................................................... 98 Quality control (QC).............................................. 47, 48, 55, 57, 58, 188, 193–195, 201, 208, 264, 267

R Retroviral transduction ........................................ 231–239 RNA interference (RNAi) .......................................33–42, 44, 46, 58, 105, 184, 279

S Single domain antibody ....................................... 121–135 Single guide RNA (sgRNA) ........................................4, 5, 108, 109, 112, 113, 116, 184, 213, 218, 228 siRNA screening.............................................4, 34, 36, 37 Size exclusion chromatography...................119, 287–298 Small molecule library........................................ 34, 36, 39

IN

DRUG DISCOVERY: METHODS

AND

PROTOCOLS Index 317

Software ...................................................... 40, 44–47, 50, 51, 58, 77, 82, 83, 110, 125, 133, 134, 148, 165–168, 171–173, 176, 184, 203, 206, 264, 267, 307, 312 Spheroids ............................................................. 153, 154, 157–159, 163–178 Spleen.......................................................... 185, 186, 193, 199, 201–203, 238, 245, 247–249, 255, 257–259 Statistics ....................................................... 47, 50, 55, 71 Stem cells ...................................................... 4, 10, 19, 30, 140, 141, 144, 146, 148, 185, 186, 271 Structural alignment ....................................................... 77 Structure activity relationship (SAR) ......... 39, 91, 97, 99 Structure superimposition ........................................77, 78 Surface markers .......................................... 144, 167, 172, 173, 176, 177, 235, 238, 260–263

T Tangential flow filtration (TFF) .......................... 287–298 Target discovery ........................................ 33, 34, 71, 196 Targeted protein degradation....................................... 106 Target identification..........................................v, 3–20, 96 Target profiling .........................................................89–99 Target promiscuity ............................................. 64, 76, 92 T cells ................................................. 152, 154, 155, 158, 159, 196, 205, 232, 242, 243, 248, 250, 265 Three-dimensional (3D)............................................ v, 48, 65–69, 73, 77–79, 82, 89, 91, 92, 94, 96–99, 114, 140, 152–160, 163–178 Tissue ............................................................ 7, 25, 36, 48, 106, 121, 139, 153, 166, 185, 231, 244, 254, 272, 301 Tissue digestion............................................................. 142 Tissue sections..................................................... 275, 276, 281, 283, 301, 302, 305–314 Trans-activating RNA (tracrRNA) ............ 213, 218–220, 222–225, 228 Transduction ..................................................10, 231–239 Transfection........................................................... 5, 7, 10, 12, 34–36, 38, 45, 108, 112, 115, 117, 123, 124, 129–131, 135, 176, 232, 234, 238 Transgene ............................................183, 184, 190, 191 Transgene expression .................................. 184, 190, 191 Transwell chambers....................................................... 147 Trocar..........................................187, 198, 201, 242, 246 Trypan blue ......................................................... 114, 118, 124, 130, 142, 145, 146, 190 Tumor associated macrophages (TAM).............. 153, 265 Tumor growth..................................................... 157, 160, 183–187, 190–197, 200, 203–205, 207, 244 Tumor implantation...................................................... 195 Tumor-infiltrating leukocytes (TILs)..........................245, 248, 255–257, 265

TARGET IDENTIFICATION

318 Index

AND

VALIDATION

IN

DRUG DISCOVERY: METHODS

Tumor inoculation .......................................200–202, 247 Tumor microenvironment (TME) ..............................153, 154, 164, 173, 205, 207 Two-dimensional (2D) ........................................... 65, 68, 69, 91, 92, 97, 98, 153, 164, 170, 175

V Validation....................................................... v, 63, 92, 95, 99, 105–119, 184, 196

AND

PROTOCOLS

Vectors .......................................................... 5, 14, 17, 57, 66, 67, 91, 92, 99, 108, 109, 111, 112, 117, 122, 125–127, 133, 176, 184, 189, 229, 231, 234, 235, 238, 304, 312 VH single domains............................................... 121–135 Virus..........................................12, 19, 45, 121, 236, 250

Z Z-scores .............................................................. 47, 53–55