Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology 1788016866, 9781788016865, 9781839160691, 9781839160776

Table of contents :
Cover
Half-title
Series information
Title page
Copyright information
Preface
Table of contents
Chapter 1 PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery?
References
Chapter 2 Structural and Biophysical Principles of Degrader Ternary Complexes
2.1 Introduction
2.1.1 Mechanistic Advantages of Targeted Protein Degradation
2.1.1.1 Immediate Advantages of Degradation Versus Inhibition
2.1.1.2 Differentiation of Degraders due to Their Mode of Action
2.1.2 History of PROTACs (2001–2010)
2.1.3 Small-molecule VHL- and CRBN-based PROTACs (2010–2015)
2.2 Structural Features of Ternary Complexes
2.2.1 Ternary Complex Equilibria and Definitions
2.2.2 Structural Elucidation of PROTAC Ternary Complexes
2.2.2.1 The First PROTAC Ternary Complex Crystal Structure: VHL:MZ1:Brd4BD2
2.2.2.2 Structure-guided design of SMARCA2/4 PROTACs
2.2.2.3 Ternary Structures of CRBN-based PROTACs
2.2.3 Degraders as Monovalent Molecular Glues
2.2.3.1 Cereblon-targeting Immunomodulatory Drugs
2.2.3.2 DCAF15-targeting Sulfonamide Drugs
2.2.4 Surface Areas Buried by PROTACs and Monovalent Glues
2.3 Ternary Assays
2.3.1 Can My PROTAC Form a Ternary Complex?
2.3.1.1 Pull-down Assays
2.3.1.2 Proximity-based Ternary Assays: AlphaScreen/LISA and TR-FRET
2.3.1.3 Surface Plasmon Resonance
2.3.2 How Tightly Does My Ternary Complex Bind?
2.3.2.1 Competition Assays
2.3.2.2 Direct Binding Assays
2.3.3 To What Extent Is My Ternary Complex Cooperative?
2.3.4 How Long Does My Ternary Complex Last?
2.3.5 Does the PROTAC Induce Ternary Complex Formation in Cells?
2.3.5.1 Separation of Phases-based Protein Interaction Reporter Assay (SPPIER)
2.3.5.2 Bioluminescence Resonance Energy Transfer (BRET)
2.4 Concluding Remarks
2.5 Acknowledgments
2.5.1 Funding
2.5.2 Conflict of Interest Statement
References
Chapter 3 Immediate and Selective Control of Protein Abundance Using the dTAG System
3.1 The Potential and Limitations of Targeted Protein Degradation
3.2 Chemical–Genetic Degradation Approaches
3.3 Development of the dTAG Platform
3.4 Genetic Methods to Express FKBP12F36V-fusions
3.4.1 Ectopic Expression of FKBP12F36V-fusions
3.4.2 Knock-in Strategies to Express FKBP12F36V-fusions
3.5 Strategies Towards Identification of a Lead dTAG Molecule
3.5.1 Biochemical Assays for FKBP12F36V and E3 Ligase Binding
3.5.2 Determining FKBP12F36V-specific Degradation in Cells
3.5.3 Requirement of E3 Ligase and Proteasome
3.5.4 Assessment of dTAG Molecule Selectivity
3.5.5 In Vivo Assessment of dTAG Molecule Activity
3.6 Case Studies Employing the dTAG Platform
3.6.1 Target Validation Using dTAG
3.6.2 Targeting Recalcitrant Oncoproteins Using dTAG
3.6.3 Targeting Essential Transcriptional Regulators Using dTAG
3.7 General Considerations for Employing Tag-based Strategies
3.8 Concluding Remarks
3.9 Abbreviations
3.10 Acknowledgments
3.10.1 Competing Financial Interests
References
Chapter 4 Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs
4.1 Introduction
4.2 The Importance of PK/PD Relationships and Additional Considerations for PROTACs
4.2.1 Building PK/PD Relationships
4.2.2 PK/PD Considerations for PROTACs
4.2.2.1 Catalytic Mechanism of PROTACs
4.2.2.2 Impact of Protein Half-life
4.2.2.3 Rate of PROTAC-mediated Degradation
4.2.2.4 Functional Consequences of PROTAC Binding to the Degradation Target
4.2.2.5 E3 Ligase and Target Protein Distribution
4.3 Developing PK/PD Relationships for a Series of RIPK2 PROTACs
4.3.1 Design of PK/PD Experiments for RIPK2 PROTACs
4.3.2 Results from PK/PD Experiments with RIPK2 PROTACs
4.4 PBPK/PD Models for PROTACs
4.5 Conclusions
4.6 Ethical Review
References
Chapter 5 New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs
5.1 Introduction
5.2 Targeted Protein Degradation Through Cereblon-CRL4
5.3 CRL4 Architecture
5.4 CELMoD Mechanism of Action
5.5 Identification of CELMoDs with Novel Activities
5.6 Molecular Basis for Substrate Recruitment
5.7 Identification of a Substrate Mediating Teratogenicity Through a Structural Degron Search
5.8 Expansion of Cereblon Neosubstrates
5.9 Further CELMoDs in Clinical Development
5.9.1 Avadomide
5.9.2 Iberdomide
5.9.3 CC-90009 and CC-92480
5.10 The Development of Cereblon-targeting Hetero-bifunctional Degraders
5.11 Differences Between Hetero-bifunctional and Scaffolded Protein–Protein Interaction Ubiquitin Ligase Modulators
5.12 Conclusions
References
Chapter 6 Structure-based PROTAC Design
6.1 Introduction
6.2 PROTAC Design – Differences to Small Molecules
6.3 Structure-based Linkerology
6.4 Learning from PPI Stabilization
6.5 VHL PROTAC Ternary Complexes
6.6 Cereblon PROTAC Ternary Complexes
6.7 Identifying the Right PPI to Target
6.8 Future Technologies for Structure-based PROTAC Design
6.9 Acknowledgments
References
Chapter 7 Plate-based High-throughput Cellular Degradation Assays to Identify PROTAC Molecules and Protein Degraders
7.1 Introduction
7.2 PROTACs
7.3 Plate-based Assays to Measure Protein Degradation
7.3.1 Immunofluorescent Target Imaging
7.3.2 ELISA or Derivative Assay Technologies
7.3.3 Protein Tagging
7.3.4 Validation of Assays to Measure Protein Degradation
7.4 Plate-based Assays to Understand Degrader Mechanism – PROTACS
7.4.1 Cellular Permeability and Target Engagement Assays
7.4.2 PROTAC Ternary Complex Formation
7.4.3 PROTAC-mediated Ubiquitination
7.4.4 Proteasome Recognition Assays
7.4.5 Modification of Degradation Assays to Assess POI Abundance and Turnover
7.4.6 Mechanistic Tools
References
Chapter 8 PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies
8.1 Introduction
8.2 Review of BTK Inhibitors
8.3 Drug Resistance and Side Effects of Ibrutinib
8.4 PROTAC Contributes to Overcome Drug Resistance of Ibrutinib
8.5 Summary
8.6 Acknowledgments
References
Chapter 9 An Efficient Approach Toward Drugging Undruggable Targets
9.1 Introduction
9.2 Target Identification and Validation
9.2.1 Use of AID Technology In Vivo
9.3 Efficient Drug Discovery Platform, RaPPIDS
9.3.1 Proprietary E3 Ligase Binders
9.3.2 A Strategy for Drug Candidate Discovery by RaPPIDS Platform
9.4 Case Study of RaPPIDS for 1st Program, IRAK-M Degrader
9.4.1 Background of IRAK-M
9.4.2 Drug Discovery of IRAK-M Degrader by RaPPIDS
9.5 Future Perspectives
9.6 Conclusion
9.7 Abbreviations
9.8 Acknowledgments
References
Chapter 10 E3-mediated Ubiquitin and Ubiquitin-like Protein Ligation: Mechanisms and Chemical Probes
10.1 Introduction
10.2 E3-dependent Conformational Regulation of E2~UB Intermediates
10.3 UBL Transfer to Substrates by Adaptor E3s Harboring RING and RING-like Domains
10.3.1 Cullin-RING Ligases (CRL)
10.3.2 CRL-dependent Ubiquitylation
10.3.3 CRL Modification by NEDD8
10.3.4 Small Molecules Manipulating Cullin Neddylation
10.3.5 CRL Assembly Cycle
10.3.6 RING-between-RING (RBR) Ligases
10.3.7 Unique Domains Specify Activation and Activity of RBR E3s
10.3.8 HECT E3 Ligases
10.3.9 Catalytic HECT Domain
10.3.10 HECT E3-mediated Polyubiquitylation
10.3.11 Modulation of NEDD4-family HECT E3 Activity by UB Binding to an N-lobe Exosite
10.3.12 HECT Domain Oligomerization
10.3.13 Mechanisms to Regulate HECT E3 Ubiquitylation Activity
10.4 Cysteine-reactive Probes
10.5 Chemical Biology Approach with Reactive E2~UB Conjugates Reveal RING-Cys-relay (RCR) Ligase
10.6 Conclusions and Future Perspectives
Note added on proof
References
Chapter 11 Plant E3 Ligases as Versatile Tools for Novel Drug Development and Plant Bioengineering
11.1 Introduction
11.2 The Four Classes of E3 Ligases in Higher Plants: A Brief Overview
11.2.1 Monomeric E3 RING-finger Ligases
11.2.2 Cullin-based RING E3 Ligases
11.2.3 U-box E3 Ligases
11.2.4 HECT E3 Ligases
11.3 Pathogens’ Use of the Ubiquitin Proteasome Pathway
11.4 The Ubiquitin Proteasome Pathway as an Opportunity
11.4.1 Novel Drug Development Utilizing the UPP
11.4.2 Synthetic Biology Using UPP Sensors
11.4.3 The N-degron Pathway as Bioengineering Tool
11.5 Future Perspectives
11.6 Acknowledgments
References
Chapter 12 Deubiquitinase Inhibitors: An Emerging Therapeutic Class
12.1 Introduction/Background
12.2 Biology and Clinical Opportunity for DUB Inhibition
12.2.1 USP7
12.2.2 USP22
12.2.3 OTULIN
12.2.4 USP30
12.3 Validating Inhibitors and Substrates of DUBs
12.3.1 Substrate Validation
12.3.2 Inhibitor Validation
12.4 Examples of Inhibitors
12.4.1 USP25/28
12.4.2 CSN5
12.4.3 Rpn11
12.4.4 USP7
12.5 Outlook and Future Directions
References
Chapter 13 Targeting Translation Regulation for the Development of Novel Drugs
13.1 Introduction
13.2 PSM, Discovery of Translation Regulators Using Pairs of Fluorescent tRNAs
13.3 Target Space for PSM: From Transcription to Translation
13.3.1 RNA Processing
13.3.2 RNA-binding Proteins
13.3.3 mRNA Localization
13.3.4 mRNA Translation
13.3.5 tRNA Modifications, Expression and Aminoacylation
References
Chapter 14 Classes, Modes of Action and Selection of New Modalities in Drug Discovery
14.1 Introduction
14.2 Nucleic Acid-based Modalities
14.2.1 Targetable Modes of Action
14.2.1.1 Protein Recognition
14.2.1.2 Direct and Indirect Downregulation of RNA Levels
14.2.1.3 Direct and Indirect Upregulation of RNA Levels
14.2.1.4 Genome Editing
14.2.2 Classes of Nucleic Acid-based Modalities
14.2.2.1 Antisense Oligonucleotide (ASO) and Small Interfering RNA (siRNA)
14.2.2.1.1 Chemical Modifications.
14.2.2.1.2 Design.
14.2.2.2 Modified mRNA (modRNA)
14.2.2.3 Aptamers
14.2.3 Strengths and Limitations of Nucleic Acid-based Modalities
14.2.3.1 Delivery
14.2.3.2 Tissue Distribution
14.2.3.3 Safety
14.3 Hyper-modified Peptides
14.3.1 Targetable Modes of Action
14.3.2 Classes of Hyper-modified Peptidic Modalities
14.3.2.1 Monocyclic Peptides Including Stapled Peptides and Other Protein Structure Mimetics
14.3.2.2 Polycyclic Peptides
14.3.3 Strength and Limitations of Peptide-based Modalities
14.4 Hybrid and Multivalent Modalities
14.4.1 Modality Fusion for Synergistic Binding and Polypharmacology
14.4.2 Modality Conjugation for Synergistic Binding
14.4.3 Enablement of Novel MOAs with Hybrid Modalities
14.4.4 Strength and Limitations of Hybrid Modalities
14.5 Selection of Modalities in Drug Discovery
14.5.1 Repertoire of Modes of Action and Modalities
14.5.1.1 Targeting at the DNA Level
14.5.1.2 Targeting at the RNA Level
14.5.1.3 Targeting at the Protein Level
14.5.2 Criteria and Perspective for Selecting Modalities
14.6 Conclusion
References
Chapter 15 Small-molecule Targeted Degradation of RNA
15.1 Introduction
15.2 Design Strategy for RNA Cleavers and Degraders
15.3 Cleaving r(CUG)exp via Photolysis of N-hydroxy-2(1H)-thione (HPT)
15.4 Harnessing the Cleavage Potential of the Bleomycin Family of Natural Products
15.4.1 Targeted Cleavage of r(CUG)exp
15.4.2 Targeted Cleavage of pri-miR-96
15.5 Harnessing the Potential of RNA Cellular Degradation Machinery
15.5.1 Recruitment of RNase L for Targeted Degradation of miR-96
15.5.2 Recruitment of RNase L for Targeted Degradation of miR-210
15.6 Outlook and Conclusions
References
Index

Citation preview

Drug Discovery

Protein Degradation with New Chemical Modalities Successful Strategies in Drug Discovery and Chemical Biology Edited by Hilmar Weinmann and Craig M. Crews

 i

Protein Degradation with New Chemical Modalities Successful Strategies in Drug Discovery and Chemical Biology

ii

Drug Discovery Series Editor-​in-​chief David Thurston, King’s College London, UK

Series editors: David Fox, Vulpine Science and Learning, UK Ana Martinez, Centro de Investigaciones Biologicas-​CSIC, Spain David Rotella, Montclair State University, USA Hong Shen, Roche Innovation Center Shanghai, China

Editorial advisor: Ian Storer, AstraZeneca, UK

Titles in the Series: 1: Metabolism, Pharmacokinetics and Toxicity of Functional Groups 2: Emerging Drugs and Targets for Alzheimer’s Disease; Volume 1 3: Emerging Drugs and Targets for Alzheimer’s Disease; Volume 2 4: Accounts in Drug Discovery 5: New Frontiers in Chemical Biology 6: Animal Models for Neurodegenerative Disease 7: Neurodegeneration 8: G Protein-​Coupled Receptors 9: Pharmaceutical Process Development 10: Extracellular and Intracellular Signaling 11: New Synthetic Technologies in Medicinal Chemistry 12: New Horizons in Predictive Toxicology 13: Drug Design Strategies: Quantitative Approaches 14: Neglected Diseases and Drug Discovery 15: Biomedical Imaging 16: Pharmaceutical Salts and Cocrystals 17: Polyamine Drug Discovery 18: Proteinases as Drug Targets 19: Kinase Drug Discovery 20: Drug Design Strategies: Computational Techniques and Applications 21: Designing Multi-​Target Drugs 22: Nanostructured Biomaterials for Overcoming Biological Barriers 23: Physico-​Chemical and Computational Approaches to Drug Discovery 24: Biomarkers for Traumatic Brain Injury 25: Drug Discovery from Natural Products 26: Anti-​Inflammatory Drug Discovery 27: New Therapeutic Strategies for Type 2 Diabetes: Small Molecules 28: Drug Discovery for Psychiatric Disorders 29: Organic Chemistry of Drug Degradation

 iii

30: Computational Approaches to Nuclear Receptors 31: Traditional Chinese Medicine 32: Successful Strategies for the Discovery of Antiviral Drugs 33: Comprehensive Biomarker Discovery and Validation for Clinical Application 34: Emerging Drugs and Targets for Parkinson’s Disease 35: Pain Therapeutics; Current and Future Treatment Paradigms 36: Biotherapeutics: Recent Developments using Chemical and Molecular Biology 37: Inhibitors of Molecular Chaperones as Therapeutic Agents 38: Orphan Drugs and Rare Diseases 39: Ion Channel Drug Discovery 40: Macrocycles in Drug Discovery 41: Human-​based Systems for Translational Research 42: Venoms to Drugs: Venom as a Source for the Development of Human Therapeutics 43: Carbohydrates in Drug Design and Discovery 44: Drug Discovery for Schizophrenia 45: Cardiovascular and Metabolic Disease: Scientific Discoveries and New Therapies 46: Green Chemistry Strategies for Drug Discovery 47: Fragment-​Based Drug Discovery 48: Epigenetics for Drug Discovery 49: New Horizons in Predictive Drug Metabolism and Pharmacokinetics 50: Privileged Scaffolds in Medicinal Chemistry: Design, Synthesis, Evaluation 51: Nanomedicines: Design, Delivery and Detection 52: Synthetic Methods in Drug Discovery: Volume 1 53: Synthetic Methods in Drug Discovery: Volume 2 54: Drug Transporters: Role and Importance in ADME and Drug Development 55: Drug Transporters: Recent Advances and Emerging Technologies 56: Allosterism in Drug Discovery 57: Anti-​aging Drugs: From Basic Research to Clinical Practice 58: Antibiotic Drug Discovery: New Targets and Molecular Entities 59: Peptide-​based Drug Discovery: Challenges and New Therapeutics 60: Drug Discovery for Leishmaniasis 61: Biophysical Techniques in Drug Discovery 62: Acute Brain Impairment Through Stroke: Drug Discovery and Translational Research 63: Theranostics and Image Guided Drug Delivery 64: Pharmaceutical Formulation: The Science and Technology of Dosage Forms 65: Small-​molecule Transcription Factor Inhibitors in Oncology 66: Therapies for Retinal Degeneration: Targeting Common Processes 67: Kinase Drug Discovery: Modern Approaches 68: Advances in Nucleic Acid Therapeutics 69: MicroRNAs in Diseases and Disorders: Emerging Therapeutic Targets

iv

70: Emerging Drugs and Targets for Multiple Sclerosis 71: Cytotoxic Payloads for Antibody–​Drug Conjugates 72: Peptide Therapeutics: Strategy and Tactics for Chemistry, Manufacturing, and Controls 73: Anti-​fibrotic Drug Discovery 74: Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology

How to obtain future titles on publication: A standing order plan is available for this series. A standing order will bring delivery of each new volume immediately on publication.

For further information please contact: Book Sales Department, Royal Society of Chemistry, Thomas Graham House, Science Park, Milton Road, Cambridge, CB4 0WF, UK Telephone: +44 (0)1223 420066, Fax: +44 (0)1223 420247, Email: [email protected] Visit our website at www.rsc.org/​books

 v

Protein Degradation with New Chemical Modalities

Successful Strategies in Drug Discovery and Chemical Biology

Edited by

Hilmar Weinmann Janssen Pharmaceutica N.V., Belgium Email: [email protected]

and

Craig Crews Yale University, USA Email: [email protected]

vi

Drug Discovery Series No. 74 Print ISBN:  978-​1-​78801-​686-​5 PDF ISBN:  978-​1-​83916-​069-​1 EPUB ISBN:  978-​1-​83916-​077-​6 Print ISSN: 2041–​3203 Electronic ISSN: 2041–​3211 A catalogue record for this book is available from the British Library © The Royal Society of Chemistry 2021 All rights reserved Apart from fair dealing for the purposes of research for non-​commercial purposes or for private study, criticism or review, as permitted under the Copyright, Designs and Patents Act 1988 and the Copyright and Related Rights Regulations 2003, this publication may not be reproduced, stored or transmitted, in any form or by any means, without the prior permission in writing of The Royal Society of Chemistry or the copyright owner, or in the case of reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency in the UK, or in accordance with the terms of the licences issued by the appropriate Reproduction Rights Organization outside the UK. Enquiries concerning reproduction outside the terms stated here should be sent to The Royal Society of Chemistry at the address printed on this page. Whilst this material has been produced with all due care, The Royal Society of Chemistry cannot be held responsible or liable for its accuracy and completeness, nor for any consequences arising from any errors or the use of the information contained in this publication. The publication of advertisements does not constitute any endorsement by The Royal Society of Chemistry or Authors of any products advertised. The views and opinions advanced by contributors do not necessarily reflect those of The Royal Society of Chemistry which shall not be liable for any resulting loss or damage arising as a result of reliance upon this material. The Royal Society of Chemistry is a charity, registered in England and Wales, Number 207890, and a company incorporated in England by Royal Charter (Registered No. RC000524), registered office: Burlington House, Piccadilly, London W1J 0BA, UK, Telephone: +44 (0) 20 7437 8656. For further information see our web site at www.rsc.org Printed in the United Kingdom by CPI Group (UK) Ltd, Croydon, CR0 4YY, UK

 vii

Preface The modulation of proteins within living cells plays a major role in the treatment of a multitude of severe diseases. Small-​molecule drugs can be used as the standard approach in many cases. Nevertheless, there are still some limitations related to small-​molecule inhibitors which restrict their utility in certain protein classes. Many of these compounds act through occupying active sites. However, many proteins, such as scaffolding proteins or transcription factors, lack such active sites and therefore are often considered to possess poor druggability. Other limitations of current small-​molecule drugs are induced target protein mutations resulting in drug resistance or compensatory feedback activation of downstream signaling while inhibiting one specific target. Long-​term target protein inhibition may lead to compensatory protein overexpression and protein accumulation which may lead to incomplete inhibition. Therefore, the search for innovative treatment methods and new modalities for intracellular modulation of proteins is a highly active field of current research. The aim of this book is to give an overview on promising recent approaches to overcome the limitations of classical small molecules with novel modalities for protein degradation and their applications in chemical biology and drug discovery. Proteolysis-​targeting chimeras (PROTACs) are an area of induced protein degradation, which makes use of hetero-​bifunctional molecules that induce a ligand to bind with the target protein, another ligand to recruit an E3 ubiquitin ligase, and a linker in between the two ligands. Once the ternary complex consisting of the target protein, the PROTAC and the E3 ligase is formed, the recruited E3 ligase will employ an E2 ubiquitin-​conjugating enzyme to transfer ubiquitin to the surface of the targeted protein. A polyubiquitination signal will be recognized by the proteasome to initiate the degradation of the Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

vii

viii

viii

Preface

targeted protein. PROTACs have been employed to induce the degradation of a variety of proteins, including kinases, nuclear receptors and transcriptional factors, as well as regulatory proteins. In the first chapter, Philipp M. Cromm, Craig M. Crews and Hilmar Weinmann describe the basic principles and summarize the historic development of PROTAC-​mediated target degradation as a potential paradigm changer in drug discovery. David Zollman and Alessio Ciulli describe the structural and biophysical principles of degrader ternary complexes. Degraders predominantly recruit a target protein to an E3 ubiquitin ligase and form with them a ternary complex, which triggers target ubiquitination and subsequent proteasomal degradation. The structural, thermodynamic and kinetic features of the ternary complexes determine the speed, potency, selectivity and durability of their cellular degradation activity. The authors illustrate how structural biology and biophysics are rapidly impacting the field and describe the main assays that are being developed and used to study PROTAC ternary complexes. In Chapter 3, Behnam Nabet and Nathanael S. Gray report on the immediate and selective control of protein abundance using the dTAG system. The dTAG technology platform uses hetero-​bifunctional small-​molecule degraders to co-​opt the endogenous cellular degradation machinery to rapidly and reversibly deplete FKBP12F36V-​tagged target proteins. The dTAG system and related tag-​based degradation strategies have the potential to become essential tools for pre-​clinical target validation and mechanistic biological investigation in cellular and mouse models of development and disease. Several examples of PROTACs with in vivo efficacy in pre-​clinical studies have been disclosed recently. While building pharmacokinetic (PK)/​pharmacodynamic (PD) relationships is recognized as a key activity in small-​molecule drug discovery to support translation from the research to clinical phases, there has been a lack of reports describing this for PROTACs despite their huge potential as therapeutics. John D. Harling, Paul Scott-​Stevens and Lu Gaohua discuss the unique mechanism of action of PROTACs and how this introduces additional factors which may need to be considered in the development of the PK/​PD relationship, including how PBPK/​PD modeling can be used to deliver human dose predictions with PROTACs. In the last few years interest in the potential of small-​molecule therapeutics capable of redirecting the cellular ubiquitin–​proteasome machinery to destroy specific proteins has increased drastically. While one approach utilizes hetero-​bifunctional ligands capable of simultaneously binding both an E3 ubiquitin ligase and the target protein, another class of molecules, termed CELMoDs (cereblon E3 ligase modulating drugs), are low molecular weight small molecules that induce the degradation of specific protein targets by binding to the cereblon-​CRL4 E3 ubiquitin ligase and scaffolding direct protein–​protein interactions to the target protein. These molecules are capable of recruiting undruggable and even unligandable targets through a so-​called “molecular glue” mechanism. Mary E. Matyskiela, Thomas Clayton, Joel W. Thompson, Christopher Carroll, Leslie Bateman, Laurie LeBrun and

 ix

Preface

ix

Philip P. Chamberlain focus on recent discoveries of CELMoD mechanism of action and discuss the future and breadth of this emerging class of molecules. Recent developments in structure-​based PROTACs design are summarized by Darryl B. McConnell. While small molecules and PROTACs both require the design of ligands which bind potently to the target protein, PROTAC design involves the additional step of stabilizing the interaction between two proteins, the E3-​ligase and the protein of interest. This second protein–​ protein interaction stabilization step of PROTAC design is where PROTACs gain their potency and selectivity advantages over classical small molecules. It is the structural insights gained from the practice of structure-​based PROTAC design which promise to accelerate the discovery of highly potent and selective PROTAC drugs. PROTAC design to subsequent degradation of the protein of interest is a multifaceted process, requiring cellular assays that interrogate and provide both structure–​activity relationship data at high throughput scale and mechanistic data to elucidate target binding, ternary complex formation, ubiquitination and ultimately degradation. Nikki Carter summarizes how this approach to PROTAC discovery and optimization has been developed in an industrial setting and describes the plate-​based cellular assays utilized to enable this evolving modality. Yonghui Sun and Yu Rao report on an application of PROTACs targeting BTK for the treatment of ibrutinib-​resistant B-​cell malignancies. In his chapter he covers the biology around the BTK signaling module and gives a general introduction of non-​Hodgkin’s lymphoma and the development of BTK inhibitors. The development and prospect of BTK PROTACs as novel treatment options are discussed. As the drug discovery of targeted protein degradation involves laborious time-​and cost-​consuming processes, it is important to establish a platform for not only the identification and validation of the targets but also rapid and efficient selection of promising drug candidates. Kanae Gamo, Naomi Kitamoto, Masato T. Kanemaki and Yusuke Tominari illustrate two processes to achieve these goals in Chapter 9. Target validation by a ligand-​induced genetic degradation system, especially Auxin-​Inducible Degron (AID) system and degrader drug discovery by Rapid Protein Proteolysis Inducer Discovery System (RaPPIDSTM) are discussed. E3 ligases drive the specificity of ubiquitin (UB) and UB-​like protein (UBL) ligation. Diverse E3 structures provide distinct mechanisms achieving timely and accurate formation of covalent bonds between targeted proteins and the C-​terminus of UB or a UBL. David T. Krist and Brenda A. Schulman give an insight into recent studies revealing remarkable mechanisms of action and regulation of major classes of eukaryotic UB ligases –​cullin-​RING, HECT and RBR E3s –​with additional focus on distinct activities of small-​molecule probes that have been developed to perturb these systems for pharmaceutical and biochemical discovery purposes. The ubiquitin proteasome pathway is a versatile regulatory mechanism that also plays a major role in plants and allows them to quickly react and

x

x

Preface

acclimatize to changing environmental conditions. E3 ligases are the key regulatory elements that provide specificity to the pathway. Raed Al-​Saharin, Sutton Mooney and Hanjo Hellmann provide a brief overview about the pathway and the specific classes of E3 ligases described in plants. Their main focus is on how the pathway provides opportunities to develop novel approaches and technological tools for research and agricultural industries. This is exemplified by how pathogens utilize the pathway, and recent novel technological developments, ranging from PROTACs, CRISPR/​CAS9, or the N-​degron pathway to develop new tools and novel applications towards improvement of plant growth and performance. Deubiquitinating enzymes (DUBs) control the removal of ubiquitin and ubiquitin-​like proteins from cellular proteins. There are approximately 100 DUBs in the human genome, and they regulate diverse biochemical, cellular and physiological processes. DUBs are known to control many pathways which are misregulated and affected in human diseases, such as cancer, immune diseases and neurodegeneration. Due to the broad scope of DUB biology, they are emerging as a target class for inhibitor development. Robert S. Magin, Laura M. Doherty and Sara J. Buhrlage describe the promise of targeting DUBs in different disease contexts, describe practices for identifying and validating small-​molecule inhibitors and physiologically relevant substrates of DUBs, and review recent examples of well-​characterized DUB inhibitors. Protein synthesis is the final step of gene expression, which includes a complex set of biochemical steps that are highly energy expensive. Selective and specific regulation of protein translation needs a complex regulatory machinery which offers multiple targets for therapeutic intervention in protein aggregation diseases and in targeting difficult proteins, such as structural proteins, transcription factors and scaffold and assembly proteins. Iris Alroy, Wissam Mansour and Yoni Sheinberger have developed a method that visualizes specific or global protein translation inside mammalian cells by monitoring the activity of ribosomes. This approach enables the discovery of small molecules which specifically regulate translation and can be used to identify novel targets for therapeutic intervention. Novel waves of innovation in drug discovery represented by an increasing range of drug modalities are providing scientists with an expanded repertoire of mode-​of-​actions and molecules for prosecuting these. Eric Valeur summarizes these “new” modalities, which include nucleic acid-​ based, hyper-​modified peptidic modalities as well as combinations of classical and new modalities. The targeted modes-​of-​action are reviewed along with their strengths and limitations in order to afford a perspective on the factors to consider for driving modality selection in drug discovery. Small-​molecule targeting of structural elements within disease-​causing RNAs has recently attracted a lot of interest both from academic research and the pharmaceutical industry. Andrei Ursu, Matthew G. Costales, Jessica L. Childs-​Disney and Matthew D. Disney describe advances in the targeted degradation of RNA by structure-​specific synthetic ligands that exploit natural products to cleave nucleic acids or compounds that locally recruit and activate

 xi

Preface

xi

endogenous ribonucleases to enzymatically cleave an RNA target. The assembly process of RNA degraders and their application to validate mode-​of-​action and profile on-​and off-​targets is discussed and the future challenges for RNA degraders and their therapeutic potential is outlined in the concluding chapter of this book. Protein Degradation with New Chemical Modalities covers a fascinating, active field of research and will enable a multitude of novel opportunities for applications in chemical biology and drug discovery. We hope that the readers of this book will share some of the excitement of this emerging field and hopefully will benefit for their own research by the overview provided by the individual chapters. As the editors of this book we are extremely grateful for the contributions of so many colleagues and leading researchers in this field. We would like to acknowledge the great enthusiasm of all chapter authors in supporting this book project and for the timely delivery of their high-​quality manuscripts. We would also like to express our gratitude to Katie Morrey and the whole staff of the Royal Society of Chemistry for triggering this project and for their extremely professional support in the production of this book. Craig M. Crews Hilmar Weinmann

xii

 xiii

Contents Chapter 1 PROTAC-​mediated Target Degradation: A Paradigm Changer in Drug Discovery? Philipp M. Cromm, Craig M. Crews and Hilmar Weinmann References

1

8

Chapter 2 Structural and Biophysical Principles of Degrader Ternary Complexes David Zollman and Alessio Ciulli 2.1 Introduction 2.1.1 Mechanistic Advantages of Targeted Protein Degradation 2.1.2 History of PROTACs (2001–​2010) 2.1.3 Small-​molecule VHL-​ and CRBN-​based PROTACs (2010–​2015) 2.2 Structural Features of Ternary Complexes 2.2.1 Ternary Complex Equilibria and Definitions 2.2.2 Structural Elucidation of PROTAC Ternary Complexes 2.2.3 Degraders as Monovalent Molecular Glues 2.2.4 Surface Areas Buried by PROTACs and Monovalent Glues

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

xiii

14

14 14 17 19 21 21 23 29 32

xiv

xiv

Contents

2.3 Ternary Assays 2.3.1 Can My PROTAC Form a Ternary Complex? 2.3.2 How Tightly Does My Ternary Complex Bind? 2.3.3 To What Extent Is My Ternary Complex Cooperative? 2.3.4 How Long Does My Ternary Complex Last? 2.3.5 Does the PROTAC Induce Ternary Complex Formation in Cells? 2.4 Concluding Remarks 2.5 Acknowledgments 2.5.1 Funding 2.5.2 Conflict of Interest Statement References Chapter 3 Immediate and Selective Control of Protein Abundance Using the dTAG System Behnam Nabet and Nathanael S. Gray 3.1 The Potential and Limitations of Targeted Protein Degradation 3.2 Chemical–​Genetic Degradation Approaches 3.3 Development of the dTAG Platform 3.4 Genetic Methods to Express FKBP12F36V-​fusions 3.4.1 Ectopic Expression of FKBP12F36V-​fusions 3.4.2 Knock-in Strategies to Express FKBP12F36V-fusions 3.5 Strategies Towards Identification of a Lead dTAG Molecule 3.5.1 Biochemical Assays for FKBP12F36V and E3 Ligase Binding 3.5.2 Determining FKBP12F36V-​specific Degradation in Cells 3.5.3 Requirement of E3 Ligase and Proteasome 3.5.4 Assessment of dTAG Molecule Selectivity 3.5.5 In Vivo Assessment of dTAG Molecule Activity 3.6 Case Studies Employing the dTAG Platform 3.6.1 Target Validation Using dTAG 3.6.2 Targeting Recalcitrant Oncoproteins Using dTAG

33 33 37 42 42 43 46 46 46 46 47 55

55 57 57 58 58 60 61 62 62 64 65 66 66 66 67

 xv

Contents

xv

3.6.3

Targeting Essential Transcriptional Regulators Using dTAG 68 3.7 General Considerations for Employing Tag-​based Strategies 68 3.8 Concluding Remarks 69 3.9 Abbreviations 70 3.10 Acknowledgments 70 3.10.1 Competing Financial Interests 70 References 70 Chapter 4 Developing Pharmacokinetic/​Pharmacodynamic Relationships With PROTACs John D. Harling, Paul Scott-​Stevens and Lu Gaohua

75

4.1 Introduction 75 4.2 The Importance of PK/​PD Relationships and Additional Considerations for PROTACs 76 4.2.1 Building PK/​PD Relationships 76 4.2.2 PK/​PD Considerations for PROTACs 78 4.3 Developing PK/​PD Relationships for a Series of RIPK2 PROTACs 82 4.3.1 Design of PK/​PD Experiments for RIPK2 PROTACs 82 4.3.2 Results from PK/​PD Experiments with RIPK2 PROTACs 84 4.4 PBPK/​PD Models for PROTACs 87 4.5 Conclusions 90 4.6 Ethical Review 91 References 91 Chapter 5 New Activities of CELMoDs, Cereblon E3 Ligase-​ modulating Drugs Mary E. Matyskiela, Thomas Clayton, Joel W. Thompson, Christopher Carroll, Leslie Bateman, Laurie LeBrun and Philip P. Chamberlain 5.1 Introduction 5.2 Targeted Protein Degradation Through Cereblon-​CRL4 5.3 CRL4 Architecture 5.4 CELMoD Mechanism of Action

94

94 95 96 98

xvi

xvi

Contents

5.5 Identification of CELMoDs with Novel Activities 99 5.6 Molecular Basis for Substrate Recruitment 99 5.7 Identification of a Substrate Mediating Teratogenicity Through a Structural Degron Search 102 5.8 Expansion of Cereblon Neosubstrates 104 5.9 Further CELMoDs in Clinical Development 106 5.9.1 Avadomide 107 5.9.2 Iberdomide 107 5.9.3 CC-​90009 and CC-​92480 108 5.10 The Development of Cereblon-​targeting Hetero-​bifunctional Degraders 108 5.11 Differences Between Hetero-​bifunctional and Scaffolded Protein–​Protein Interaction Ubiquitin Ligase Modulators 111 5.12 Conclusions 111 References 112 Chapter 6 Structure-​based PROTAC Design Darryl B. McConnell

115

6.1 Introduction 115 6.2 PROTAC Design –​Differences to Small Molecules 117 6.3 Structure-​based Linkerology 119 6.4 Learning from PPI Stabilization 120 6.5 VHL PROTAC Ternary Complexes 122 6.6 Cereblon PROTAC Ternary Complexes 126 6.7 Identifying the Right PPI to Target 128 6.8 Future Technologies for Structure-​based PROTAC Design 129 6.9 Acknowledgments 130 References 130 Chapter 7 Plate-​based High-​throughput Cellular Degradation Assays to Identify PROTAC Molecules and Protein Degraders Nikki Carter 7.1 Introduction 7.2 PROTACs 7.3 Plate-​based Assays to Measure Protein Degradation 7.3.1 Immunofluorescent Target Imaging

135

135 136 137 138

 xvi

Contents

xvii

7.3.2 7.3.3 7.3.4

ELISA or Derivative Assay Technologies 139 Protein Tagging 139 Validation of Assays to Measure Protein Degradation 140 7.4 Plate-​based Assays to Understand Degrader Mechanism –​PROTACs 141 7.4.1 Cellular Permeability and Target Engagement Assays 141 7.4.2 PROTAC Ternary Complex Formation 142 7.4.3 PROTAC-​mediated Ubiquitination 143 7.4.4 Proteasome Recognition Assays 143 7.4.5 Modification of Degradation Assays to Assess POI Abundance and Turnover 144 7.4.6 Mechanistic Tools 144 References 146 Chapter 8 PROTAC Targeting BTK for the Treatment of Ibrutinib-​ resistant B-​cell Malignancies Yonghui Sun and Yu Rao

147

8.1 Introduction 147 8.2 Review of BTK Inhibitors 150 8.3 Drug Resistance and Side Effects of Ibrutinib 152 8.4 PROTAC Contributes to Overcome Drug Resistance of Ibrutinib 153 8.5 Summary 164 8.6 Acknowledgments 166 References 166 Chapter 9 An Efficient Approach Toward Drugging Undruggable Targets Kanae Gamo, Naomi Kitamoto, Masato T. Kanemaki and Yusuke Tominari 9.1 Introduction 9.2 Target Identification and Validation 9.2.1 Use of AID Technology In Vivo 9.3 Efficient Drug Discovery Platform, RaPPIDS 9.3.1 Proprietary E3 Ligase Binders 9.3.2 A Strategy for Drug Candidate Discovery by RaPPIDS Platform

167

167 168 169 171 171 175

xvii

xviii

Contents

9.4 Case Study of RaPPIDS for 1st Program, IRAK-​M Degrader 176 9.4.1 Background of IRAK-​M 176 9.4.2 Drug Discovery of IRAK-​M Degrader by RaPPIDS 177 9.5 Future Perspectives 179 9.6 Conclusion 180 9.7 Abbreviations 180 9.8 Acknowledgments 181 References 181 Chapter 10 E3-​mediated Ubiquitin and Ubiquitin-​like Protein Ligation: Mechanisms and Chemical Probes David T. Krist and Brenda A. Schulman 10.1 Introduction 10.2 E3-​dependent Conformational Regulation of E2~UB Intermediates 10.3 UBL Transfer to Substrates by Adaptor E3s Harboring RING and RING-​like Domains 10.3.1 Cullin-​RING Ligases (CRL) 10.3.2 CRL-​dependent Ubiquitylation 10.3.3 CRL Modification by NEDD8 10.3.4 Small Molecules Manipulating Cullin Neddylation 10.3.5 CRL Assembly Cycle 10.3.6 RING-​between-​RING (RBR) Ligases 10.3.7 Unique Domains Specify Activation and Activity of RBR E3s 10.3.8 HECT E3 Ligases 10.3.9 Catalytic HECT Domain 10.3.10 HECT E3-mediated Polyubiquitylation 10.3.11 Modulation of NEDD4-​family HECT E3 Activity by UB Binding to an N-​lobe Exosite 10.3.12 HECT Domain Oligomerization 10.3.13 Mechanisms to Regulate HECT E3 Ubiquitylation Activity 10.4 Cysteine-​reactive Probes 10.5 Chemical Biology Approach with Reactive E2~UB Conjugates Reveal RING-​Cys-​relay (RCR) Ligase

184

184 187 188 189 189 190 192 193 194 195 197 198 199 199 200 201 202 203

 xix

Contents

xix

10.6 Conclusions and Future Perspectives 204 Note added on proof 204 References 204 Chapter 11 Plant E3 Ligases as Versatile Tools for Novel Drug Development and Plant Bioengineering Raed Al-​Saharin, Sutton Mooney and Hanjo Hellmann

212

11.1 Introduction 212 11.2 The Four Classes of E3 Ligases in Higher Plants: A Brief Overview 213 11.2.1 Monomeric E3 RING-​finger Ligases 213 11.2.2 Cullin-​based RING E3 Ligases 215 11.2.3 U-​box E3 Ligases 216 11.2.4 HECT E3 Ligases 216 11.3 Pathogens’ Use of the Ubiquitin Proteasome Pathway 217 11.4 The Ubiquitin Proteasome Pathway as an Opportunity 219 11.4.1 Novel Drug Development Utilizing the UPP 219 11.4.2 Synthetic Biology Using UPP Sensors 222 11.4.3 The N-​degron Pathway as Bioengineering Tool 222 11.5 Future Perspectives 225 11.6 Acknowledgments 225 References 225 Chapter 12 Deubiquitinase Inhibitors: An Emerging Therapeutic Class Robert S. Magin, Laura M. Doherty and Sara J. Buhrlage 12.1 Introduction/​Background 12.2 Biology and Clinical Opportunity for DUB Inhibition 12.2.1 USP7 12.2.2 USP22 12.2.3 OTULIN 12.2.4 USP30 12.3 Validating Inhibitors and Substrates of DUBs 12.3.1 Substrate Validation 12.3.2 Inhibitor Validation

234

234 237 237 239 239 239 240 240 243

xx

xx

Contents

12.4 Examples of Inhibitors 244 12.4.1 USP25/​28 244 12.4.2 CSN5 245 12.4.3 Rpn11 246 12.4.4 USP7 247 12.5 Outlook and Future Directions 248 References 249 Chapter 13 Targeting Translation Regulation for the Development of Novel Drugs Iris Alroy, Wissam Mansour and Yoni Sheinberger

254

13.1 Introduction 254 13.2 PSM, Discovery of Translation Regulators Using Pairs of Fluorescent tRNAs 256 13.3 Target Space for PSM: From Transcription to Translation 262 13.3.1 RNA Processing 262 13.3.2 RNA-​binding Proteins 266 13.3.3 mRNA Localization 268 13.3.4 mRNA Translation 269 13.3.5 tRNA Modifications, Expression and Aminoacylation 270 References 272 Chapter 14 Classes, Modes of Action and Selection of New Modalities in Drug Discovery Eric Valeur 14.1 Introduction 14.2 Nucleic Acid-​based Modalities 14.2.1 Targetable Modes of Action 14.2.2 Classes of Nucleic Acid-​based Modalities 14.2.3 Strengths and Limitations of Nucleic Acid-​ based Modalities 14.3 Hyper-​modified Peptides 14.3.1 Targetable Modes of Action 14.3.2. Classes of Hyper-​modified Peptidic Modalities 14.3.3 Strength and Limitations of Peptide-​based Modalities

277

277 279 279 282 287 290 291 291 295

newgenprepdf

 xxi

Contents

xxi

14.4 Hybrid and Multivalent Modalities 296 14.4.1 Modality Fusion for Synergistic Binding and Polypharmacology 297 14.4.2 Modality Conjugation for Synergistic Binding 299 14.4.3 Enablement of Novel MOAs with Hybrid Modalities 299 14.4.4 Strength and Limitations of Hybrid Modalities 300 14.5 Selection of Modalities in Drug Discovery 300 14.5.1 Repertoire of Modes of Action and Modalities 300 14.5.2 Criteria and Perspective for Selecting Modalities 302 14.6 Conclusion 305 References 306 Chapter 15 Small-​molecule Targeted Degradation of RNA Andrei Ursu, Matthew G. Costales, Jessica L. Childs-​Disney and Matthew D. Disney

317

15.1 Introduction 317 15.2 Design Strategy for RNA Cleavers and Degraders 319 exp 15.3 Cleaving r(CUG) via Photolysis of N-​hydroxy-​ 2(1H)-​thione (HPT) 321 15.4 Harnessing the Cleavage Potential of the Bleomycin Family of Natural Products 321 exp 15.4.1 Targeted Cleavage of r(CUG) 321 15.4.2 Targeted Cleavage of pri-​miR-​96 324 15.5 Harnessing the Potential of RNA Cellular Degradation Machinery 326 15.5.1 Recruitment of RNase L for Targeted Degradation of miR-​96 326 15.5.2 Recruitment of RNase L for Targeted Degradation of miR-​210 328 15.6 Outlook and Conclusions 330 References 332 Index 337

xxi

 1

CHAPTER 1

PROTAC-​mediated Target Degradation: A Paradigm Changer in Drug Discovery? PHILIPP M. CROMM,a CRAIG M. CREWSb,c,d AND HILMAR WEINMANNe* a

Research and Development, Pharmaceuticals, Bayer AG, 13353 Berlin, Germany b Department of Molecular, Cellular & Developmental Biology, Yale University, New Haven, CT 06511, USA c Department of Chemistry, Yale University, New Haven, CT 06511, USA d Department of Pharmacology, Yale University, New Haven, CT 06511, USA e Janssen Pharmaceutica N.V., Discovery Process Research, 2340 Beerse, Belgium *Corresponding author. Email: [email protected]

In the constant pursuit of medicinal chemistry to expand the druggable space and develop novel strategies to target essential disease driving proteins, small-​molecule-​induced protein degradation has by far created the most excitement in recent years. Controlling a protein’s fate and inducing its degradation allows one to not only target enzymes but also the targeting of adaptor and scaffolding proteins, which are notorious difficult to address due to the absence of defined active sites.1 To do so, scientists have discovered methods that make use of the cell’s native protein degradation Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

1

2

2

Chapter 1

machinery to mark selected target proteins and trigger their degradation. The ubiquitin proteasome system sits at the heart of this machinery and controls the protein’s fate via ubiquitination and subsequent proteasomal degradation.2,3 Proteolysis-​targeting chimeras (PROTACs) hijack this degradation machinery and recruit it for addressing target proteins that drive disease (Figure 1.1).4–​6 As a result, the target protein is irreversibly removed from the system in a proteasome-​dependent manner. This describes a fundamentally different mode of action relative to traditional small-​molecule inhibitors or even antibodies, which both act via high-​affinity target binding and by blocking protein binding (in case of protein–​protein interactions or PPIs) or active site access (enzymes). Induced target protein degradation allows one to drug previously unaddressable adaptor or scaffolding functions and expands the druggable space via connecting target engagement with a beneficial cellular event.7 Since the first proof-​of-​concept study in 2001, the PROTAC technology has matured as both a chemical biology approach and new therapeutic modality, in so far as the first clinical Phase I trials for an androgen and estrogen targeting PROTAC were initiated in 2019 (Figure 1.2).8 Early PROTACs were based on peptidic E3 ligase-​binding motifs, which were linked to different

Figure 1.1  Schematic representation of PROTAC mode of action. The PROTAC engages the target protein of interest (POI) and an E3 ligase to form a ternary complex. A ubiquitin is subsequently transferred onto the POI in a E2-​mediated manner. The poly-​ubiquitinated POI is recognized by the proteasome and degraded while the PROTAC is released to engage in an additional cycle.

 3

PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery?

3

Figure 1.2 Timeline of PROTAC development; scientific and industrial.

target-​binding moieties and suffered from limited cell permeability and low degradation efficacy. The breakthrough of the technology came with the development of a peptidomimetic ligand of the von Hippel Lindau (VHL) E3 ligase with more drug-​like physicochemical properties9,10 and the elucidation of the mode of action of thalidomide.11 These discoveries paved the way for the first drug-​like PROTACs for RIPK2/​ERRα12 and BRD413,14 described in 2015. Since these groundbreaking studies that significantly accelerated the field, interest in PROTAC-​induced protein degradation has increased greatly in academia as well as in industry.15 Since then, many scientific research groups from academia and industry have focused on exploring the strengths, opportunities, limitations and weaknesses of this technology. Since the launch of the pioneering PROTAC company Arvinas® in 2013, subsequent second-​generation companies such as C4 Therapeutics® (founded 2015) and Kymera Therapeutics® (founded 2017) were founded to explore the therapeutic potential of small-​molecule-​induced protein degradation. Given the relatively recent interest in this approach as a therapeutic modality, these three biotech companies have already landed multiple multimillion-​dollar collaborations with several large pharma companies. Almost every pharmaceutical company has begun to evaluate the PROTAC technology, either in collaboration with one of the aforementioned biotech companies or by building up their own protein degradation units, e.g., Novartis®, AstraZeneca® or GSK®. In 2019 Arvinas® launched the first Phase I clinical trials for orally available small-​molecule protein degraders targeting the androgen and estrogen receptor in metastatic castration-​resistant prostate cancer (mCRPC) and metastatic ER+ positive/​HER2–​ negative breast cancer, respectively. However, while these first companies have begun to generate orally available efficacious PROTACs, it is becoming apparent to others new to the field that the design and optimization of PROTACs is not a straightforward process (Figure 1.3A).16 While connecting two drug-​like small molecules with a linker in most cases yields compounds capable of degrading target proteins in cell culture, these compounds have physicochemical properties far beyond Lipinski’s rule of 5 chemical space, thus making oral bioavailability a challenging property to achieve.17 During PROTAC-​led optimization, the E3 ligase ligand, linker and target binder need to be studied and optimized.18

4

4

Chapter 1

Figure 1.3 Schematic PROTAC assembly. (A) A PROTAC comprises a protein ligand, a linker and a E3 ligase ligand. Representative targets (left) and E3 ligases (right) are named explicitly. (B) Structure of a FAK degrading VHL-​based PROTAC.7 The different elements of the PROTAC are colored accordingly.

Different exit vectors as well as linker compositions should be considered to obtain the most efficacious PROTAC. Linker structure and length play critical roles, because the linker can influence the overall PROTAC conformation and binding orientation as well as the formation of the ternary complex.19 Consequently, the linker does much more than simply connect two molecular entities; it directly influences PROTAC activity, selectivity and physicochemical properties.20 To support structure-​based optimization efforts, both the use of X-​ray co-​crystal structures of PROTAC ternary complexes and computational approaches as an additional tool to predict ternary complex models in the absence of X-​ray structural data have been successfully utilized thus far.19,21,22 Although the chemical space for linker design is potentially huge, most research groups prefer either alkyl or polyethylene glycol (PEG) chains as starting points to determine the optimal linker length. By selecting either alkyl or PEG linkers, the overall physicochemical properties (e.g., TPSA and cLog P) are modified accordingly. Rigidifying moieties (e.g., heterocyclic ring systems, alkyne groups) have also been applied in several studies.23,24 In order to rapidly determine the optimal linker length of a PROTAC, an accelerated approach for the efficient preparation of systematic PROTACs libraries comprising various linker compositions, length and different E3 ligase binders is of high value. Application of click chemistry for the preparation of such PROTACs libraries to rapidly identify starting points for further optimization has been described recently using BRD4 PROTACs as a model system.25 Furthermore, linkers can be also exploited as attachment points for targeting

 5

PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery?

5

modalities. In this context, the first PROTAC–​antibody conjugates delivering the PROTAC payload directly to the targeted tissues have been described.26 While some proteins like BRD4,27 the androgen receptor, or BTK are the targets of choice in many publications, reports of novel target proteins succumbing to PROTAC-​mediated degradation are surfacing almost on a daily basis. A non-​ comprehensive overview of most PROTAC targets and the corresponding E3 ligase can be found in Table 1.1 in chronological order. For most protein targets, PROTAC-​mediated degradation displays effects similar to traditional target inhibition, with the PROTAC outperforming the inhibitor with regards to potency or addressing certain resistance mechanisms.28 However, for focal adhesion kinase (FAK), PROTAC-​induced target degradation is able to shut down the scaffolding function of FAK, which is beyond the reach of small-​molecule Fak inhibitors (Figure 1.3B).7 The same holds true for PROTACs targeting SMARCA2/​4, which are part of the BAF complex and are notoriously difficult to target.22 Additionally, Tau degraders have been reported to clear accumulated protein in frontotemporal dementia (FTD) patient-​derived neuronal cell models expanding the therapeutic area of PROTACs to neurodegenerative diseases.29 Studies like these demonstrate the tremendous potential of PROTACs to expand the druggable space. It will be exciting to see which prominent hard-​to-​drug disease drivers will be tackled next by PROTAC-​induced protein degradation. Small-​molecule-​induced protein degradation has generated a lot of excitement in recent years, with many players in industry and academia joining in on the pursuit of the next big leap in drug discovery. While initially few medicinal chemists thought it possible, the first two PROTAC clinical candidates are administered p.o. and seem to exhibit favorable human PK profiles. Others have reportedly also developed orally available PROTACs and it appears that the long-​enshrined rule of 5 is beginning to crumble in certain cases, given the recent success of beyond rule of 5 clinical assets.17,80,81 Nevertheless, many aspects of PROTACs are still unknown and despite the already tremendous success of the technology thus far, its full potential has not yet been fully explored. Understandably, most companies and even academic groups avoid taking extensive risks when it comes to project selection. Most PROTAC targets are clinically validated and marketed drugs that are already available. As mentioned above, in these cases, the PROTACs impressively outperform the state-​of-​the-​art inhibitor but rarely touch on differential biology aspects out of reach of traditional, occupancy-​based modalities.4 The full potential of PROTACs will only be unleashed when adaptor and scaffolding proteins are being addressed, for which inhibitors are ineffective due to their orthogonal mode of action. However, a tremendous hurdle to achieve this goal will be the identification of suitable binders for these undruggable proteins, an endeavor drug discovery has already faced for decades. However, connecting the pharmacological benefit with pure binding and subsequent target degradation (event-​driven) instead of inhibition makes this challenge appear closer to reality than ever before. It will be exciting to see the field develop more towards the real undruggable proteome and help identifying, validating, devalidating and targeting highly interesting novel protein targets.

6

6

Chapter 1

Table 1.1 Selected proteins successfully degraded via a PROTAC’s ubiquitination approach. Year

Target

Target class

E3 ligase

Reference

2001 2003

MetAP2 Androgen receptor Estrogen receptor Androgen receptor Aryl hydrocarbon receptor Androgen receptor Estrogen receptor CRAPBPI and CRAPBPII RAR Androgen receptor Estrogen receptor FRS2α

Dimetallohydrolase Nuclear receptor Nuclear receptor Nuclear receptor Transcription factor

βTRCP (polypeptidic) βTRCP (polypeptidic) βTRCP (polypeptidic) VHL (polypeptidic) VHL (polypeptidic)

8 30 30 31 32

2004 2007 2008 2010 2011 2013

2014 2015

2016

2017 2018

Nuclear receptor Nuclear receptor Cellular retinoic acid-​ binding proteins Nuclear receptor Nuclear receptor Nuclear receptor Fibroblast growth factor receptor substrate 2 PI3K Kinase TACC3 Transforming acidic coiled-​coil-​ containing protein 3 BRD4 Bromodomain BET (BRD2, BRD3, Bromodomain BRD4) ERRα Nuclear receptor FKBP12 Peptidyl-​prolyl cis–​ trans isomerase RIPK2 Kinase AKT Kinase BCR-​ABL Kinase BET (BRD2, BRD3, Bromodomain BRD4) Tau Microtubule-​ associated protein tau CDK9 Kinase VHL TBK1 BTK

E3 ligase Kinase Kinase

TRIM24 PCAF/​GCN5

Bromodomain Bromodomain

ALK

Kinase Kinase

MDM2 (nutlin-​3a) 33 VHL (polypeptidic) 34 cIAP (small molecule) 35 cIAP (small molecule) cIAP (small molecule) cIAP (small molecule) VHL (polypeptidic)

36 36 36 37

VHL (polypeptidic) 37 cIAP (small molecule) 38 VHL (small molecule) CRBN (small molecule) VHL (small molecule) CRBN (small molecule) VHL (small molecule) VHL (polypeptidic) VHL (small molecule), CRBN (small molecule) VHL (small molecule)

39 13, 14

VHL (polypeptidic)

43

CRBN (small molecule) VHL (small molecule) VHL (small molecule) CRBN (small molecule) VHL (small molecule) CRBN (small molecule) CRBN (small molecule) VHL (small molecule)

44

12 13 12 40 41 42

45 46 47, 48 49 50 51, 52 53

 7

PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery? Table 1.1 (continued) Year

Target

Target class

E3 ligase

Reference

PI3K

Kinase

54

HDAC6

Histone deacetylase

CRBN (small molecule) CRBN (small molecule) CRBN (small molecule) CRBN (small molecule) CRBN (small molecule) Keap1 (peptide) CRBN (small molecule) VHL (small molecule) VHL (small molecule) VHL (small molecule) VHL (small molecule) CRBN (small molecule) VHL (small molecule), CRBN (small molecule) MDM2 (nutlin-​3)

63 64

CRBN (small molecule) VHL (small molecule) VHL (small molecule) CRBN (small molecule) VHL (small molecule) VHL (small molecule) CRBN (small molecule) CRBN (small molecule) VHL (small molecule)

67

RNF114 (nimbolide) VHL (small molecule) RNF4 (covalent binding small molecule) CRBN (small molecule)

76 77 78

BET (BRD2, BRD3, Bromodomain BRD4, BRDT) Sirt2 Lysine deacetylase BCL6 Tau CRBN

2019

7

Transcriptional regulator E3 Ligase

PTK2/​FAK FLT-​3 EGFR, HER2, and c-​Met IRAK4 Mcl-​1/​Bcl-​2

Kinase Kinase Kinase

PTK2/​FAK

Kinase

PARP1 CDK6

Poly (ADP-​ribose) polymerases Kinase

BRD9/​BRD7 CRBN EGFR

Bromodomain E3 ligase Kinase

Kinase Bcl-​2 family

Estrogen receptor Nuclear receptor Androgen receptor Nuuclear receptor MDM2 E3 ligase HDAC6

Histone deacetylase

SMARCA2, SMARCA4, PBRM1 BRD4 SGK3 BRD4

Bromodomain

HCV NS3/​4A

Protease

Bromodomain Kinase Bromodomain

55 56, 57 58 59 60 61 7 28 62

65 66

68 69, 70 71 72 73 23 74, 75 22

79

8

8

Chapter 1

Thus far, protein degradation has primarily been focused on the oncology field with some glimpses in neurodegenerative diseases (Tau) or inflammation/​immunology (IRAK4 and RIP2K).12,29,63 Given the generally poor prognosis associated with many cancer diagnoses, the huge medical need as well as the comparably higher tolerance of side effects in oncology make this therapeutic area the logical choice to begin with. To expand the indication spectrum towards other therapeutic areas including non-​malignant diseases will pose a challenge by itself. Not only does the target space comprising ion channels and G protein-​coupled receptors (GPCRs) in addition to nuclear hormone receptors or kinases look different, but the E3 ligases will have to withstand a closer examination, as well. Modulating the activity of a tumor suppressor such as VHL or the onco-​target CRBN might not be an option when it comes to indication expansion for PROTACs.11,82–​84 Inhibitor of apoptosis proteins (IAPs) might offer an alternative to VHL and CRBN and have already been utilized multiple times to induce protein degradation of various protein targets.85–​87 Nevertheless, the expansion to other therapeutic areas will increase the demand to utilize additional E3 ligases suitable for efficacious target degradation. In particular, identifying a suitable tissue-​or disease-​specific E3 ligase will be invaluable. Although a first report has been published recently,88 extracellular proteins and membrane proteins like ion channels and GPCRs are still out of reach for small-​molecule-​induced protein degradation. Inspired by PROTACs, it will be exciting to see how the field of artificially induced novel PPIs will develop.89 A first glimpse of what might be possible was already offered by the recruitment of a phosphatase to dephosphorylate specific kinases.90 Additionally, Oerth Bio®, the recently founded joint venture between Arvinas® and Bayer®, is trying to utilize the PROTAC technology to develop new agricultural products represents an exciting perspective of what targeted protein degradation might be capable of. Many challenges are left, and many additional hurdles need to be overcome, but the initial successes demonstrated thus far fuel hopes that PROTACs will have a bright future ahead, not only as an additional tool in the toolboxes of chemical biology and drug discovery but as a successful therapeutic modality. It certainly feels exciting to be at the forefront of such a powerful technology trying to lead the way to a new age of drug discovery. This book will provide a timely overview of the recent developments in the field of small-​molecule-​induced protein degradation by many of the leading scientists in the field. Contributions from academia as well as industry will provide insights in the field from all available angles.

References 1. 2. 3. 4. 5.

A. C. Lai and C. M. Crews, Nat. Rev. Drug Discovery, 2017, 16, 101–​114. M. Hochstrasser, Curr. Opin. Cell Biol., 1995, 7, 215–​223. A. Navon and A. Ciechanover, J. Biol. Chem., 2009, 284, 33713–​33718. P. M. Cromm and C. M. Crews, Cell Chem. Biol., 2017, 24, 1181–​1190. M. Pettersson and C. M. Crews, Drug Discovery Today: Technol., 2019, 31,  15–​27.

 9

PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery?

9

6. S.-​L. Paiva and C. M. Crews, Curr. Opin. Chem. Biol., 2019, 50, 111–​119. 7. P. M. Cromm, K. Samarasinghe, J. Hines and C. M. Crews, J. Am. Chem. Soc., 2018, 140, 17019–​17026. 8. K. M. Sakamoto, K. B. Kim, A. Kumagai, F. Mercurio, C. M. Crews and R. J. Deshaies, Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 8554–​8559. 9. D. L. Buckley, J. L. Gustafson, I. Van Molle, A. G. Roth, H. S. Tae, P. C. Gareiss, W. L. Jorgensen, A. Ciulli and C. M. Crews, Angew. Chem., Int. Ed., 2012, 51, 11463–​11467. 10. D. L. Buckley, I. Van Molle, P. C. Gareiss, H. S. Tae, J. Michel, D. J. Noblin, W. L. Jorgensen, A. Ciulli and C. M. Crews, J. Am. Chem. Soc., 2012, 134, 4465–​4468. 11. T. Ito, H. Ando, T. Suzuki, T. Ogura, K. Hotta, Y. Imamura, Y. Yamaguchi and H. Handa, Science, 2010, 327, 1345–​1350. 12. D. P. Bondeson, A. Mares, I. E. D. Smith, E. Ko, S. Campos, A. H. Miah, K. E. Mulholland, N. Routly, D. L. Buckley, J. L. Gustafson, N. Zinn, P. Grandi, S. Shimamura, G. Bergamini, M. Faelth-​Savitski, M. Bantscheff, C. Cox, D. A. Gordon, R. R. Willard, J. J. Flanagan, L. N. Casillas, B. J. Votta, W. den Besten, K. Famm, L. Kruidenier, P. S. Carter, J. D. Harling, I. Churcher and C. M. Crews, Nat. Chem. Biol., 2015, 11, 611. 13. G. E. Winter, D. L. Buckley, J. Paulk, J. M. Roberts, A. Souza, S. Dhe-​Paganon and J. E. Bradner, Science, 2015, 348, 1376–​1381. 14. J. Lu, Y. Qian, M. Altieri, H. Dong, J. Wang, K. Raina, J. Hines, J. D. Winkler, A. P. Crew, K. Coleman and C. M. Crews, Chem. Biol. (Oxford, U. K.), 2015, 22, 755–​763. 15. H. Pei, Y. Peng, Q. Zhao and Y. Chen, RSC Advances, 2019, 9, 16967–​16976. 16. H. J. Maple, N. Clayden, A. Baron, C. Stacey and R. Felix, MedChemComm, 2019. 17. S. D. Edmondson, B. Yang and C. Fallan, Bioorg. Med. Chem. Lett., 2019, 29, 1555–​1564. 18. S. L. Fisher and A. J. Phillips, Curr. Opin. Chem. Biol., 2018, 44,  47–​55. 19. M. S. Gadd, A. Testa, X. Lucas, K. H. Chan, W. Chen, D. J. Lamont, M. Zengerle and A. Ciulli, Nat. Chem. Biol., 2017, 13, 514–​521. 20. R. P. Nowak, S. L. DeAngelo, D. Buckley, Z. He, K. A. Donovan, J. An, N. Safaee, M. P. Jedrychowski, C. M. Ponthier, M. Ishoey, T. Zhang, J. D. Mancias, N. S. Gray, J. E. Bradner and E. S. Fischer, Nat. Chem. Biol., 2018, 14, 706–​714. 21. M. L. Drummond and C. I. Williams, J. Chem. Inf. Model., 2019, 59, 1634–​1644. 22. W. Farnaby, M. Koegl, M. J. Roy, C. Whitworth, E. Diers, N. Trainor, D. Zollman, S. Steurer, J. Karolyi-​Oezguer, C. Riedmueller, T. Gmaschitz, J. Wachter, C. Dank, M. Galant, B. Sharps, K. Rumpel, E. Traxler, T. Gerstberger, R. Schnitzer, O. Petermann, P. Greb, H. Weinstabl, G. Bader, A. Zoephel, A. Weiss-​ Puxbaum, K. Ehrenhöfer-​ Wölfer, S. Wöhrle, G. Boehmelt, J. Rinnenthal, H. Arnhof, N. Wiechens, M.-​Y. Wu, T. Owen-​Hughes, P. Ettmayer, M. Pearson, D. B. McConnell and A. Ciulli, Nat. Chem. Biol., 2019, 15, 672–​680.

10

10

Chapter 1

23. Y. Li, J. Yang, A. Aguilar, D. McEachern, S. Przybranowski, L. Liu, C.-​Y. Yang, M. Wang, X. Han and S. Wang, J. Med. Chem., 2019, 62, 448–​466. 24. C. P. Tinworth, H. Lithgow, L. Dittus, Z. I. Bassi, S. E. Hughes, M. Muelbaier, H. Dai, I. E. D. Smith, W. J. Kerr, G. A. Burley, M. Bantscheff and J. D. Harling, ACS Chem. Biol., 2019, 14, 342–​347. 25. R. P. Wurz, K. Dellamaggiore, H. Dou, N. Javier, M.-​C. Lo, J. D. McCarter, D. Mohl, C. Sastri, J. R. Lipford and V. J. Cee, J. Med. Chem., 2018, 61, 453–​461. 26. T. Pillow, J. Sadowsky, L. Staben, S. Staben, B. Wei, I. E. Wertz, P. Adhikari, N. Blaquiere, P. Dragovich and W. J. Fairbrother, WO 2017/​201449 A1, 2017, Roche. 27. C.-​Y. Yang, C. Qin, L. Bai and S. Wang, Drug Discovery Today: Technol., 2019, 31,  43–​51. 28. G. M. Burslem, J. Song, X. Chen, J. Hines and C. M. Crews, J. Am. Chem. Soc., 2018, 140, 16428–​16432. 29. M. C. Silva, F. M. Ferguson, Q. Cai, K. A. Donovan, G. Nandi, D. Patnaik, T. Zhang, H.-​T. Huang, D. E. Lucente, B. C. Dickerson, T. J. Mitchison, E. S. Fischer, N. S. Gray and S. J. Haggarty, eLife, 2019, 8, e45457. 30. K. M. Sakamoto, K. B. Kim, R. Verma, A. Ransick, B. Stein, C. M. Crews and R. J. Deshaies, Mol. Cell. Proteomics, 2003, 2, 1350–​1358. 31. J. S. Schneekloth, F. N. Fonseca, M. Koldobskiy, A. Mandal, R. Deshaies, K. Sakamoto and C. M. Crews, J. Am. Chem. Soc., 2004, 126, 3748–​3754. 32. H. Lee, D. Puppala, E.-​Y. Choi, H. Swanson and K.-​B. Kim, ChemBioChem, 2007, 8, 2058–​2062. 33. A. R. Schneekloth, M. Pucheault, H. S. Tae and C. M. Crews, Bioorg. Med. Chem. Lett., 2008, 18, 5904–​5908. 34. A. Rodriguez-​ Gonzalez, K. Cyrus, M. Salcius, K. Kim, C. M. Crews, R. J. Deshaies and K. M. Sakamoto, Oncogene, 2008, 27, 7201–​7211. 35. Y. Itoh, M. Ishikawa, M. Naito and Y. Hashimoto, J. Am. Chem. Soc., 2010, 132, 5820–​5826. 36. Y. Itoh, R. Kitaguchi, M. Ishikawa, M. Naito and Y. Hashimoto, Bioorg. Med. Chem., 2011, 19, 6768–​6778. 37. J. Hines, J. D. Gough, T. W. Corson and C. M. Crews, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, 8942–​8947. 38. N. Ohoka, K. Nagai, T. Hattori, K. Okuhira, N. Shibata, N. Cho and M. Naito, Cell Death Dis., 2014, 5, e1513–​e1513. 39. M. Zengerle, K.-​H. Chan and A. Ciulli, ACS Chem. Biol., 2015, 10, 1770–​1777. 40. R. K. Henning, J. O. Varghese, S. Das, A. Nag, G. Tang, K. Tang, A. M. Sutherland and J. R. Heath, J. Pept. Sci., 2016, 22, 196–​200. 41. A. C. Lai, M. Toure, D. Hellerschmied, J. Salami, S. Jaime-​Figueroa, E. Ko, J. Hines and C. M. Crews, Angew. Chem., Int. Ed., 2016, 55, 807–​810. 42. K. Raina, J. Lu, Y. Qian, M. Altieri, D. Gordon, A. M. K. Rossi, J. Wang, X. Chen, H. Dong, K. Siu, J. D. Winkler, A. P. Crew, C. M. Crews and K. G. Coleman, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, 7124–​7129. 43. T.-​T. Chu, N. Gao, Q.-​Q. Li, P.-​G. Chen, X.-​F. Yang, Y.-​X. Chen, Y.-​F. Zhao and Y.-​M. Li, Cell Chem. Biol., 2016, 23, 453–​461.

 11

PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery?

11

44. C. M. Robb, J. I. Contreras, S. Kour, M. A. Taylor, M. Abid, Y. A. Sonawane, M. Zahid, D. J. Murry, A. Natarajan and S. Rana, Chem. Commun., 2017, 53, 7577–​7580. 45. C. Maniaci, S. J. Hughes, A. Testa, W. Chen, D. J. Lamont, S. Rocha, D. R. Alessi, R. Romeo and A. Ciulli, Nat. Commun., 2017, 8, 830. 46. A. P. Crew, K. Raina, H. Dong, Y. Qian, J. Wang, D. Vigil, Y. V. Serebrenik, B. D. Hamman, A. Morgan, C. Ferraro, K. Siu, T. K. Neklesa, J. D. Winkler, K. G. Coleman and C. M. Crews, J. Med. Chem., 2018, 61, 583–​598. 47. A. D. Buhimschi, H. A. Armstrong, M. Toure, S. Jaime-​ Figueroa, T. L. Chen, A. M. Lehman, J. A. Woyach, A. J. Johnson, J. C. Byrd and C. M. Crews, Biochemistry, 2018, 57, 3564–​3575. 48. A. Zorba, C. Nguyen, Y. Xu, J. Starr, K. Borzilleri, J. Smith, H. Zhu, K. A. Farley, W. Ding, J. Schiemer, X. Feng, J. S. Chang, D. P. Uccello, J. A. Young, C. N. Garcia-​ Irrizary, L. Czabaniuk, B. Schuff, R. Oliver, J. Montgomery, M. M. Hayward, J. Coe, J. Chen, M. Niosi, S. Luthra, J. C. Shah, A. El-​Kattan, X. Qiu, G. M. West, M. C. Noe, V. Shanmugasundaram, A. M. Gilbert, M. F. Brown and M. F. Calabrese, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, E7285–​E7292. 49. L. N. Gechijian, D. L. Buckley, M. A. Lawlor, J. M. Reyes, J. Paulk, C. J. Ott, G. E. Winter, M. A. Erb, T. G. Scott, M. Xu, H.-​S. Seo, S. Dhe-​Paganon, N. P. Kwiatkowski, J. A. Perry, J. Qi, N. S. Gray and J. E. Bradner, Nat. Chem. Biol., 2018, 14, 405–​412. 50. Z. I. Bassi, M. C. Fillmore, A. H. Miah, T. D. Chapman, C. Maller, E. J. Roberts, L. C. Davis, D. E. Lewis, N. W. Galwey, K. E. Waddington, V. Parravicini, A. L. Macmillan-​Jones, C. Gongora, P. G. Humphreys, I. Churcher, R. K. Prinjha and D. F. Tough, ACS Chem. Biol., 2018, 13, 2862–​2867. 51. C. Zhang, X.-​R. Han, X. Yang, B. Jiang, J. Liu, Y. Xiong and J. Jin, Eur. J. Med. Chem., 2018, 151, 304–​314. 52. C. E. Powell, Y. Gao, L. Tan, K. A. Donovan, R. P. Nowak, A. Loehr, M. Bahcall, E. S. Fischer, P. A. Janne, R. E. George and N. S. Gray, J. Med. Chem., 2018, 61, 4249–​4255. 53. C. H. Kang, D. H. Lee, C. O. Lee, J. Du Ha, C. H. Park and J. Y. Hwang, Biochem. Biophys. Res. Commun., 2018, 505, 542–​547. 54. W. Li, C. Gao, L. Zhao, Z. Yuan, Y. Chen and Y. Jiang, Eur. J. Med. Chem., 2018, 151, 237–​247. 55. K. Yang, Y. Song, H. Xie, H. Wu, Y.-​T. Wu, E. D. Leisten and W. Tang, Bioorg. Med. Chem. Lett., 2018, 28, 2493–​2497. 56. B. Zhou, J. Hu, F. Xu, Z. Chen, L. Bai, E. Fernandez-​Salas, M. Lin, L. Liu, C.-​Y. Yang, Y. Zhao, D. McEachern, S. Przybranowski, B. Wen, D. Sun and S. Wang, J. Med. Chem., 2018, 61, 462–​481. 57. C. Qin, Y. Hu, B. Zhou, E. Fernandez-​ Salas, C.-​ Y. Yang, L. Liu, D. McEachern, S. Przybranowski, M. Wang, J. Stuckey, J. Meagher, L. Bai, Z. Chen, M. Lin, J. Yang, D. N. Ziazadeh, F. Xu, J. Hu, W. Xiang, L. Huang, S. Li, B. Wen, D. Sun and S. Wang, J. Med. Chem., 2018, 61, 6685–​6704.

12

12

Chapter 1

58. M. Schiedel, M. Moroglu, D. M. H. Ascough, A. E. R. Chamberlain, J. Kamps, A. R. Sekirnik and S. J. Conway, Angew. Chem., Int. Ed., 2019, doi:  10.1002/​anie.201812164. 59. W. McCoull, T. Cheung, E. Anderson, P. Barton, J. Burgess, K. Byth, Q. Cao, M. P. Castaldi, H. Chen, E. Chiarparin, R. J. Carbajo, E. Code, S. Cowan, P. R. Davey, A. D. Ferguson, S. Fillery, N. O. Fuller, N. Gao, D. Hargreaves, M. R. Howard, J. Hu, A. Kawatkar, P. D. Kemmitt, E. Leo, D. M. Molina, N. O’Connell, P. Petteruti, T. Rasmusson, P. Raubo, P. B. Rawlins, P. Ricchiuto, G. R. Robb, M. Schenone, M. J. Waring, M. Zinda, S. Fawell and D. M. Wilson, ACS Chem. Biol., 2018, 13, 3131–​3141. 60. M. Lu, T. Liu, Q. Jiao, J. Ji, M. Tao, Y. Liu, Q. You and Z. Jiang, Eur. J. Med. Chem., 2018, 146, 251–​259. 61. C. Steinebach, S. Lindner, N. D. Udeshi, D. C. Mani, H. Kehm, S. Kopff, S. A. Carr, M. Gutschow and J. Kronke, ACS Chem. Biol., 2018, 13, 2771–​2782. 62. G. M. Burslem, B. E. Smith, A. C. Lai, S. Jaime-​Figueroa, D. C. McQuaid, D. P. Bondeson, M. Toure, H. Dong, Y. Qian, J. Wang, A. P. Crew, J. Hines and C. M. Crews, Cell Chem. Biol., 2018, 25, 67–​77 e63. 63. J. Nunes, G. A. McGonagle, J. Eden, G. Kiritharan, M. Touzet, X. Lewell, J. Emery, H. Eidam, J. D. Harling and N. A. Anderson, ACS Med. Chem. Lett., 2019, 10, 1081–1085. 64. Z. Wang, N. He, Z. Guo, C. Niu, T. Song, Y. Guo, K. Cao, A. Wang, J. Zhu, X. Zhang and Z. Zhang, J. Med. Chem., 2019, 62, 8152–​8163. 65. J. Popow, H. Arnhof, G. Bader, H. Berger, A. Ciulli, D. Covini, C. Dank, T. Gmaschitz, P. Greb, J. Karolyi-​ Oezguer, M. Koegl, D. McConnell, M. Pearson, M. Rieger, J. Rinnenthal, V. Roessler, A. Schrenk, M. Spina, S. Steurer, N. Trainor, E. Traxler, C. Wieshofer, A. Zoephel and P. Ettmayer, J. Med. Chem., 2019, 62, 2508–​2520. 66. Q. Zhao, T. Lan, S. Su and Y. Rao, Chem. Commun., 2019, 55, 369–​372. 67. M. Brand, B. Jiang, S. Bauer, K. A. Donovan, Y. Liang, E. S. Wang, R. P. Nowak, J. C. Yuan, T. Zhang, N. Kwiatkowski, A. C. Müller, E. S. Fischer, N. S. Gray and G. E. Winter, Cell Chem. Biol., 2019, 26, 300–​306.e309. 68. V. Zoppi, S. J. Hughes, C. Maniaci, A. Testa, T. Gmaschitz, C. Wieshofer, M. Koegl, K. Riching, D. L. Daniels, A. Spallarossa and A. Ciulli, J. Med. Chem., 2019, 62, 699–​726. 69. M. Girardini, C. Maniaci, S. J. Hughes, A. Testa and A. Ciulli, Bioorg. Med. Chem., 2019, 27, 2466–​2479. 70. C. Steinebach, H. Kehm, S. Lindner, L. P. Vu, S. Köpff, Á. López Mármol, C. Weiler, K. G. Wagner, M. Reichenzeller, J. Krönke and M. Gütschow, Chem. Commun., 2019, 55, 1821–​1824. 71. G. Jaeschke, B. Kuhn, A. Ricci, D. Rueher, S. Steiner, Y. A. Nagel, M. duplessis, K. Lazarski and Y. Liang, WO 2019/​121562 A1, 2019, C4 Therapeutic/​Roche. 72. J. Hu, B. Hu, M. Wang, F. Xu, B. Miao, C.-​Y. Yang, M. Wang, Z. Liu, D. F. Hayes, K. Chinnaswamy, J. Delproposto, J. Stuckey and S. Wang, J. Med. Chem., 2019, 62, 1420–​1442.

 13

PROTAC-mediated Target Degradation: A Paradigm Changer in Drug Discovery?

13

73. X. Han, C. Wang, C. Qin, W. Xiang, E. Fernandez-​Salas, C.-​Y. Yang, M. Wang, L. Zhao, T. Xu, K. Chinnaswamy, J. Delproposto, J. Stuckey and S. Wang, J. Med. Chem., 2019, 62, pp 941–​964. 74. Z. An, W. Lv, S. Su, W. Wu and Y. Rao, Protein Cell, 2019, 10, pp 606–​609. 75. H. Wu, K. Yang, Z. Zhang, E. D. Leisten, Z. Li, H. Xie, J. Liu, K. A. Smith, Z. Novakova, C. Barinka and W. Tang, J. Med. Chem., 2019, 62, 7042–​7057. 76. J. N. Spradlin, X. Hu, C. C. Ward, S. M. Brittain, M. D. Jones, L. Ou, M. To, A. Proudfoot, E. Ornelas, M. Woldegiorgis, J. A. Olzmann, D. E. Bussiere, J. R. Thomas, J. A. Tallarico, J. M. McKenna, M. Schirle, T. J. Maimone and D. K. Nomura, Nat. Chem. Biol., 2019, 15, 747–​755. 77. H. Tovell, A. Testa, C. Maniaci, H. Zhou, A. R. Prescott, T. Macartney, A. Ciulli and D. R. Alessi, ACS Chem. Biol., 2019, 14, 882–​892. 78. C. C. Ward, J. I. Kleinman, S. M. Brittain, P. S. Lee, C. Y. S. Chung, K. Kim, Y. Petri, J. R. Thomas, J. A. Tallarico, J. M. McKenna, M. Schirle and D. K. Nomura, ACS Chem. Biol., 2019, DOI: 10.1021/​acschembio.1028b01083. 79. M. de Wispelaere, G. Du, K. A. Donovan, T. Zhang, N. A. Eleuteri, J. C. Yuan, J. Kalabathula, R. P. Nowak, E. S. Fischer, N. S. Gray and P. L. Yang, Nat. Commun., 2019, 10, 3468. 80. D. A. DeGoey, H.-​J. Chen, P. B. Cox and M. D. Wendt, J. Med. Chem., 2018, 61, 2636–​2651. 81. M. Tyagi, F. Begnini, V. Poongavanam, B. C. Doak and J. Kihlberg, Chem. –​Eur. J., DOI: 10.1002/​chem.201902716. 82. T. Asatsuma-​Okumura, H. Ando, M. De Simone, J. Yamamoto, T. Sato, N. Shimizu, K. Asakawa, Y. Yamaguchi, T. Ito, L. Guerrini and H. Handa, Nat. Chem. Biol., 2019, 15, 1077–1084. 83. K. A. Donovan, J. An, R. P. Nowak, J. C. Yuan, E. C. Fink, B. C. Berry, B. L. Ebert and E. S. Fischer, eLife, 2018, 7, e38430. 84. M. E. Matyskiela, S. Couto, X. Zheng, G. Lu, J. Hui, K. Stamp, C. Drew, Y. Ren, M. Wang, A. Carpenter, C.-​W. Lee, T. Clayton, W. Fang, C.-​C. Lu, M. Riley, P. Abdubek, K. Blease, J. Hartke, G. Kumar, R. Vessey, M. Rolfe, L. G. Hamann and P. P. Chamberlain, Nat. Chem. Biol., 2018, 14, 981–​987. 85. Y. Demizu, N. Shibata, T. Hattori, N. Ohoka, H. Motoi, T. Misawa, T. Shoda, M. Naito and M. Kurihara, Bioorg. Med. Chem. Lett., 2016, 26, 4865–​4869. 86. M. Naito, N. Ohoka and N. Shibata, Drug Discovery Today: Technol., 2019, 31, 35–42. 87. N. Ohoka, K. Okuhira, M. Ito, K. Nagai, N. Shibata, T. Hattori, O. Ujikawa, K. Shimokawa, O. Sano, R. Koyama, H. Fujita, M. Teratani, H. Matsumoto, Y. Imaeda, H. Nara, N. Cho and M. Naito, J. Biol. Chem., 2017, 292, 4556–​4570. 88. S. Banik, K. Pedram, S. Wisnovsky, N. Riley and C. Bertozzi, ChemRxiv, 2019, DOI: 10.26434/chemrxiv.7927061.v2. 89. C. Maniaci and A. Ciulli, Curr. Opin. Chem. Biol., 2019, 52, 145–​156. 90. S. Yamazoe and S. Staben, ChemRxiv, 2019, DOI: 10.26434/chemrxiv. 9177878.v1.

14

CHAPTER 2

Structural and Biophysical Principles of Degrader Ternary Complexes DAVID ZOLLMAN AND ALESSIO CIULLI* Division of Biological Chemistry and Drug Discovery, School of Life Sciences, University of Dundee, Dow Street, Dundee DD1 5EH, Scotland, United Kingdom *Corresponding author. Email: [email protected]

2.1 Introduction 2.1.1 Mechanistic Advantages of Targeted Protein Degradation Biological tools such as gene knockout or knockdown, e.g., via RNA suppression, are widely used to investigate protein function. Cellular responses or phenotypes based on such studies are typically used to infer the involvement of a particular gene, and so protein, into pathways that go awry in disease, qualifying novel targets for drug discovery.1 The outcome of this process ultimately motivates drug discovery programmes to address specific targets with pharmacological agents, either drug-​like small molecules or biologics, for example antibodies.2 Until recently, the predominant modality of pharmaceutical intervention has been occupancy-​driven, aimed at blocking the functionality of a protein by specifically binding to it with an agent. Conventional occupancy-​based Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

14

 15

Structural and Biophysical Principles of Degrader Ternary Complexes

15

small-​molecule agents are enzyme inhibitors, receptor antagonists or disruptors of protein–​protein interactions (PPIs). These agents work via saturating a binding site on the target, usually competing for binding with a natural substrate or partner molecules. While target blockade can drive a pharmacological response, this is mechanistically distinct from removing or depleting the target protein altogether, as achieved by genetic studies.3 As a result, a disconnect between these two approaches is often observed. Occupancy-​based agents are limited to targeting a functional site. They also need to fully saturate the binding site to exert the desired response, and so to sustain high concentrations well above the dissociation constant of the complex. This in turn imposes a high demand on potency, i.e., the concentration at which a cellular effect is observed. When dosing in vivo or in patients, high and frequent dosing are often required, which can impact safety. This has resulted in high attrition rates in the clinic.4

2.1.1.1 Immediate Advantages of Degradation Versus Inhibition Pharmacological target knockdown provides an alternative modality to occupancy-​based target blockade. Targeting RNA with antisense oligonucleotides (ASOs) and RNA interference (RNAi) have been a mainstream approach to target knockdown; however, delivery issues, poor pharmacokinetics profiles and limited on-​target kinetics and selectivity have long been a stumbling block to these approaches.5 More recently, the advent of small-​molecule degraders that induce targeted protein degradation (TPD) by binding to the target protein directly have opened a new era of pharmacological intervention.6–​8 Degrader molecules recruit a target protein to an E3 ubiquitin ligase, triggering protein ubiquitination and subsequent proteasomal degradation (Figure 2.1). Compared to occupancy-​based inhibitors or antagonists, degraders more closely phenocopy the outcome of genetic intervention because the target is also depleted from inside the cell.9 Moreover, the degrader may not need to bind to a functional site and could as well target any region on the surface of the protein of interest (POI). This opens an opportunity to target so-​called “undruggable” proteins.10 In addition, the degrader could in principle act catalytically, with each molecule turning over multiple protein molecules, as the protein gets degraded by the proteasome but the degrader itself is not.11 This is reflected by the observation that degraders work at much lower concentration than expected from their biophysical binding affinities in vitro, i.e., sub-​stoichiometrically (Figure 2.1). As an advantage compared to gene knockouts, protein degradation can be reversed (through protein resynthesis) and so can be used to investigate embryonically lethal or essential targets.12,13

2.1.1.2 Differentiation of Degraders due to Their Mode of Action Work in the degrader field over the past years has illuminated important non-​ obvious advantages of TPD beyond those described above. Firstly, the ability

16

newgenrtpdf

16 Chapter 2

Figure 2.1 Mechanism of degrader-​mediated mode of action and consequential advantages over occupancy-​based target blockade. Shown is a structural model obtained by superposing X-​ray crystallographic structures of full-​length Cullin RING ligase VHL (CRL2-​VHL, PDB code 5N4W),14 VCB:MZ1:Brd4BD2 complex (PDB code 5T35)15 and RING-​domain–​ubch5a–​ubiquitin heterotrimeric complex (PDB code 4AP4).16

 17

Structural and Biophysical Principles of Degrader Ternary Complexes

17

of degraders to be more selective than one might expect based on the inherent binding affinity of their cognate target binding portion. Degraders are able to discriminate amongst highly homologous targets in ways not possible with occupancy-​based inhibitors.15,17–​23 Secondly, the activity of degraders was shown to be less dependent on the binary binding affinity for their target and E3 ligase than the constituent binder alone, as degradation potency does not linearly disappear with loss of binding affinity.15,20,23–​25 This provides opportunities to revitalize already developed target protein binders that otherwise lack clinical utility either due to targeting a non-​functional binding site or because of insufficient potency to translate into a cellular effect.9,26 These remarkable features and mechanistic advantages are now understood to be fundamentally imparted by the distinct mode of action of degraders, which progresses via formation of a ternary complex with an E3 ubiquitin ligase (Figure 2.1). Structural and biophysical investigations have illuminated how proximity induced by the degraders within the ternary complexes allows for the formation of neo protein–​protein contacts and the burial of extensive solvent-​accessible surface area between the target to be degraded and the E3 ligase, which would not normally interact with each other inside the cell.27 The structural features of such interactions can drive the formation of highly stable, cooperative and long-​lived complexes. This in turn can underpin more efficient ubiquitination of the target protein and consequently faster target degradation, in a highly selective fashion and at lower degrader concentrations.9,28,29 As we strengthen our understandings of the mechanisms dictating the activities of degraders, we become more able to utilize the remarkable advantages of TPD. These learnings are being used by biologists, medicinal chemists and drug hunters alike to design molecules that meet crucial needs either in the form of new selective chemical tools to probe biological processes or novel differentiated molecular therapeutics. In this chapter, we intend to briefly recount the history of how degraders have come to the fore and how this modality has taken off in recent years, with a particular focus on bifunctional degraders, also known as proteolysis-​targeting chimeras (PROTACs) (Figure 2.2). We will then illustrate pioneering advances in structural biology and biophysical studies of the ternary complexes that underpin their mode of action. Finally, we describe some of the main assays that are being developed and used to monitor formation and properties of ternary species, an area of intense research within the field.

2.1.2 History of PROTACs (2001–​2010) For many years, small molecules offered the promise of artificially inducing ubiquitination and protein knockdown of therapeutic targets. Although mechanisms for natural intracellular protein degradation had been known for a long time, such as degrons30 and the E1–​E2–​E3–​proteasome cascade of the ubiquitin–​proteasome system (discoveries that were rewarded with the Nobel Prize for Chemistry in 2004),31–​33 chemists’ ability to co-​opt this

18

newgenrtpdf

18

Figure 2.2 Milestones of PROTAC development. (A) Early PROTAC proof-​of-​concept compounds. (B) Binding modes of high-​affinity ligands for E3 ligases VHL (PDB code 4W9H)51 and CRBN (PDB code 4CI3).56 (C) Representative PROTAC degraders made of VHL and CRBN ligands, disclosed in 2015. Chapter 2

 19

Structural and Biophysical Principles of Degrader Ternary Complexes

19

system for therapeutic benefit remained far-​fetched. In 1999, long before the mechanism of action of plant hormone auxin34 and infamous drug thalidomide35 were elucidated, an idea was postulated and even patented that a small molecule, perhaps double-​headed in nature, might allow recruitment of a target protein to the cell’s ubiquitin system.36 Later, in 2001, a first in vitro proof-​of-​concept of this idea was shown in a paper by the groups of Raymond Deshaies (Caltech) and Craig Crews (Yale).37 The groups developed a bifunctional molecule, that they dubbed “PROTAC”, comprising ovalicin, a high-​ affinity covalent ligand for methionine aminopeptidase (MetAP-​2), joined via an alkylic-​poly-​Gly-​based linker with a ~10-​amino acid phosphoserine peptide derived from IκBα, the natural substrate of the Skp1-​Cullin-​F box complex containing the F-​box substrate recognition subunit β-​TRCP (SCFβ-​TRCP).38 The PROTAC mediated ubiquitination of MetAP-​2 in in vitro ubiquitination assays, and promoted degradation of MetAP-​2 in cell extracts from unfertilized Xenopus laevis eggs.37 Later PROTACs incorporated an epitope peptide from the Hypoxia Inducible Factor 1-​alpha (HIF-​1α) (Figure 2.2a).39 Knowledge of the sequence was informed by co-​crystal structures of HIF-​1α peptides bound to von Hippel-​Lindau protein (VHL) in a complex with ElonginC and ElonginB (VCB),40,41 which form part of the Cullin RING E3 ubiquitin ligase with Cullin2 and Rbx1.14 These early PROTACs, while providing proof-​of-​concept, were significantly limited by low cellular activity due to their highly peptidic composition. Ubiquitination and co-​immunopricipation assays were being used to imply formation of a ternary complex,42,43 alongside degradation assays which mainly utilized immunoblotting techniques. However, biophysical or structural techniques, today widely applied in chemical biology and drug discovery,44,45 were not being used to help assess or develop the PROTACs beyond sporadic measurements of inhibition, e.g., IC50s.46,47 The peptidic nature of early PROTACs hampered their development, and so biological achievement of TPD remained challenging, let alone the prospect of turning these molecules into therapeutics. Today these promises have come to fruition thanks to important advances in the development of high-​quality E3 ligase ligands.

2.1.3 Small-​molecule VHL-​ and CRBN-​based PROTACs (2010–​2015) The emergence of high-​quality E3 ubiquitin ligase ligands allowed the field to devise much improved molecules in terms of cellular and in vivo potency and selectivity profiles (Figure 2.2B). Motivated by a desire to move away from the peptidic nature of early PROTACs, the Crews and Ciulli laboratories worked collaboratively to develop peptidomimetic ligands of the VHL E3 ligase. A hydroxyproline amino acid represents a crucial recognition element for VHL to HIF-​1α. Accordingly, a first-​series of drug-​like VHL ligands of crystallographically defined binding modes were elaborated in a fragment-​ based, structure-​guided fashion using hydroxyproline as the core anchor fragment.48–​50 Co-​crystal structures confirmed the binding mode of the VHL

20

20

Chapter 2

ligands, with the central hydroxyproline recapitulating its binding in the context of HIF-​1α, and the chemical groups around it fitting snugly into two distinct pockets on the surface of VHL, achieving sub-​micromolar dissociation constants (Kd). These developments gave the field the first non-​peptidic VHL ligand to work with. Further structure-​based optimization of this series led to the development of VH298, a selective, high-​affinity (Kd = 40 nM) and cell-​ active VHL inhibitor (Figure 2.2B).51–​53 In a separate approach, studies on the mode of action of the infamous immunomodulatory drug (IMiD) thalidomide led to the identification of the Cul4 E3 ubiquitin ligase cereblon (CRBN) as its biological target.35 Thalidomide and close analogues lenalidomide and pomalidomide were also found to display their anticancer activities by targeting CRL4CRBN via direct interaction with cereblon.54 Co-​crystal structures of IMiDs bound to CRBN revealed the ligand fitting snugly into a small hydrophobic pocket on the surface of CRBN defined by three tryptophan residues and a histidine (Figure 2.2B).55,56 Crucially, IMiDs can function as degraders in their own right, by “glueing” neo-​substrate proteins onto the surface of CRBN, described in more details in Section 2.2.3.1. These important developments opened the possibility to leverage small molecules of good physicochemical properties and crystallographically defined binding to E3 ligases for PROTACs, and so began to be conjugated to target protein ligands. In 2015, four independent studies disclosed VHL-​ based and CRBN-​based PROTACs against the BET proteins Brd2, Brd3 and Brd4 and protein kinases ERRa and RIPK2 (Figure 2.2C).17,57–​59 BET degraders, VHL-​based MZ1,17 CRBN-​based dBET159 and ARV-​82558 exhibited remarkable potency and selectivity in both cancer cell lines and in vivo. Consistent with their mechanism of action via the ternary complex, these PROTACs were shown to effectively induce reversible and long-​lasting protein degradation at very low concentrations, in a proteasomal and ligase-​dependent manner, and with remarkable target selectivity.17,57–​59 Of note, Zengerle et al. made the unexpected observation that the BET degrader MZ1 induced preferential degradation of one BET protein, Brd4, over its paralogues Brd2 and Brd3.17 This was not anticipated because the BET ligand (JQ1) used is a pan-​BET inhibitor that binds with comparable binding affinity to all BET bromodomains. This was a first observation of a now widely observed phenomenon in the field –​ that PROTAC degraders may confer selectivity over and above that of the constitutive ligands.15,17,18,20,21,60 The 2015 studies proved to be the tipping point for the PROTAC field because they demonstrated scope, potency and wide applicability in cells and in vivo of small-​molecule PROTAC degraders for the first time. Since then, the field has grown exponentially from no more than a handful of papers published per year to frequent publications every week, reviewed elsewhere.7,61–​65,66 However, the understanding of how the ternary complexes could mediate the different activities and selectivities observed and impact on PROTAC efficiency remained poorly understood. It was clear that the field needed to gain more insights into the structural and biophysical features of the ternary complex

 21

Structural and Biophysical Principles of Degrader Ternary Complexes

21

equilibria and how these impact on the PROTAC mode of action. The first crystal structure demonstrating how a PROTAC brings together its target protein to an E3 ligase in a ternary complex, solved in 2017 and described in Section 2.2.2.1, provided an important first step in this direction.15 The next section will highlight the key discoveries that have helped to shine light on the importance of ternary complexes in a PROTAC’s mode of action.

2.2 Structural Features of Ternary Complexes To understand the structural features of PROTAC ternary complexes it is important to consider the three-​body, two-​step equilibria that they undergo, and the key parameters that are defined (Figure 2.3). This is described in more detail in the next section.

2.2.1 Ternary Complex Equilibria and Definitions Ternary complexes are formed through two independent binding events: the sequential association of PROTAC to POI-​1 in a binary complex, followed by recruitment of POI-​2 resulting in a ternary complex.† The bifunctional nature of PROTACs mean that either POI could be recruited first, followed by ligase, or vice versa. Theoretically, the thermodynamics of the system is such that the nature of the most stable ternary complex species that can form should be independent of the pathway taken in the process (Figure 2.3A). The overall stability of the PROTAC ternary complex (ΔGcomplex) can therefore be defined as the sum of the free energy of binding for the two sequential steps, which should be the same irrespective of which POI binds first. For the purpose of this chapter, we will consider non-​covalent binding at either end of the PROTAC. PROTACs that react covalently at either POI or E3 ligase have been reported67–​74 and will not be covered here. The binding process between one end of the PROTAC molecule and POI-​1 can be affected by the presence of POI-​2 pre-​bound at the other end of the PROTAC. Recruitment of POI-1 to a PROTAC : POI-2 binary complex can therefore be more, or less, favored than expected based on POI-1 :  PROTAC binary affinity. This is reflected by a parameter called cooperativity (α), which is defined as the ratio between the dissociation constants (Kd) of a binary complex divided by the Kd of the corresponding ternary complex binding step (Figure 2.3A). A PROTAC is therefore said to exhibit “positive cooperativity” when it binds POI-​1 with higher binding affinity, i.e., lower Kd, when already engaged with POI-​2. Positive or negative cooperativity arise from either favourable or unfavourable cross-​interactions within the ternary complex. This so-​ called “cross-​energy”, when favourable, contributes extra stabilization energy to the complex, beyond the sum of the binary binding energies. †

For simplicity we will use the term “POI-​1” to refer to one protein component of the ternary complex (POI or ligase), and “POI-​2” to refer to the second protein component of the ternary complex (ligase or POI, respectively).

22

22

Chapter 2

Figure 2.3 Three-​body equilibrium of the PROTAC ternary complex. (A) Formation of a ternary complex proceeds through two sequential steps: formation of a binary complex followed by formation of a ternary complex. The overall free energy change (ΔG) of forming the ternary complex (ΔGcomplex) should be identical regardless of the chosen pathway, i.e., whether POI-​1 or POI-​2 binds to the PROTAC first. Cooperativity (α) is calculated by dividing the Kd of POI-​1 to PROTAC by the Kd of POI-​1 to the PROTAC in the presence of POI-​2. Theoretically, this should be equal to the cooperativity calculated from the Kd of POI-​2 to PROTAC ± POI-​1. (B) Graph depicting fraction of ternary complex as a function of total concentration of bifunctional molecule ([PROTAC]). Increases in PROTAC concentration initially result in an increase in ternary complex, followed by a plateau and then a decline in signal as the ligase and POI become saturated with PROTAC in binary complexes, so displaying the “hook” effect. The relative population and broadness of the pharmacological range at which ternary complex is formed is affected by the cooperativity of the PROTAC ternary complex. Profiles for a non-​cooperative PROTAC (α = 1, black line), a negatively cooperative PROTAC (α < 1, red line), a positively cooperative PROTAC (α > 1, green line) and a highly positively cooperative PROTAC (α >> 1, blue line) are shown.

 23

Structural and Biophysical Principles of Degrader Ternary Complexes

23

Cooperativity (α) and total stability (ΔGcomplex) are important features of ternary complex equilibria because they influence the population and distribution of the ternary complex. As bifunctional molecules, PROTACs can exhibit a well-​described biphasic pharmacology, also known as the “hook effect”, whereby binary engagement may compete with and ultimately overcome ternary complex formation. This results in a characteristic bell-​shaped curve (Figure 2.3B). Theoretical calculations suggest that forming cooperative and stable complexes is expected to alleviate the hook effect, and so enhances the extent to which the ternary complex is populated relative to unbound species and binary complexes, and enlarges the pharmacological range of its formation.75,76 Such a theoretical framework supports structural and biophysical studies to understand PROTAC ternary complex equilibria in more detail, which are covered next.

2.2.2 Structural Elucidation of PROTAC Ternary Complexes Classical approaches to structure-​based drug design (SBDD) focus on the binary interaction between a small molecule and a POI. While the binary affinities of the PROTAC to its protein partners are important, the inherent properties of the ternary complex, such as ternary binding affinity, cooperativity and ternary complex stability, matter to a great extent too. Cooperative effects are contributed by favourable interactions between the two proteins, and the PROTAC itself adopting an energetically allowed conformation within the complex. These features cannot be readily predicted by binary models alone; the ternary complex as a whole must therefore be investigated. Although computational docking programmes can be employed successfully for protein–​ ligand interactions (and even protein–​protein interactions), the application of this field to study and predict PROTAC-​induced ternary complexes is still in its infancy, and not yet able to achieve the resolution and accuracy required for PROTAC optimization.77 Experimentally derived complex structures are therefore critical for gaining insight into PROTAC-​mediated ternary complex binding. Protein X-​ray crystallography has been the primary technique employed so far to obtain this structural information.

2.2.2.1 The First PROTAC Ternary Complex Crystal Structure: VHL:MZ1:Brd4BD2 The PROTAC mode of action requires formation of a ternary complex, yet how the structural properties of the ternary complex might influence the ubiquitination activity and degradation potency and selectivity of PROTACs has remained elusive. The first reported PROTAC ternary complex crystal structure was solved by Gadd et al. at 2.7 Å resolution, comprising the second bromodomain of Brd4 (Brd4BD2), the BET degrader MZ1 and VCB.15 In the structure, MZ1 was fully defined within the electron density. The linker was not elongated in a linear conformation –​instead, it was seen coiled around itself, allowing MZ1 to adopt

24

24

Chapter 2

Figure 2.4 Views from ternary X-​ray crystal structures of the PROTAC MZ1 in complex with Brd4 and VHL. Top and left insert: VCB  : MZ1 :  Brd4BD2 ternary complex (PDB code 3T35)15 revealing linkage points for macrocylization. Right insert  : VCB : macroPROTAC-1 :  Brd4BD2 ternary complex (PDB 6SIS)24 confirms macroPROTAC-1 retains the binding mode of MZ1.

an S-​shaped conformation bound within a bowl-​shaped interface (Figure 2.4). This allowed formation of an extensive protein–​protein interface defined by multiple neo-​contacts between VHL and Brd4BD2 –​two proteins that do not interact with each other in the absence of the compound. MZ1-​induced neo-​ PPIs comprised of hydrophobic interactions at the “core” of the bowl-​shaped interface, including the stacking of VHL P71 with W374 from the “WPF shelf” of Brd4BD2; and hydrogen bonds and salt bridge interactions at the “rims” of the bowl, including a zipper of salt-​bridges between VHL R107 and R108 with D381 and E383 of Brd4BD2 at one end and R69 of VHL and E438 of Brd4BD2 at the other end. In addition, the linker is found sitting snugly at one end of the interface, forming a well-​defined hydrogen bond between an oxygen in the PEG linker and H437 in Brd4BD2. A total surface area of 2680 Å2 is buried, of which 680 Å2 is contributed by the VHL–​Brd4 interaction.

 25

Structural and Biophysical Principles of Degrader Ternary Complexes

25

The structure provided not only a first glimpse of how a PROTAC might work in action, but also allowed rationalizing the observed potency and intra-​BET degradation selectivity of MZ1 (described in Section 2.1.3).17 While binary binding affinities of MZ1 to the BET bromodomain were all similar, biophysical proximity and direct binding assays demonstrated that the ternary complexes formed between MZ1, VHL and the BET bromodomains all had very different population, stability and cooperativities. In particular, the ternary complex with the highest cooperativity (α = 23) and ternary complex stability (ΔGcomplex = –​22.2 kcal mol–​1) was indeed with Brd4BD2, the one that had successfully crystallized and yielded the structure. Such a large, positive cooperativity corresponds to about 2 kcal mol–​1 of energy, contributing about 10% of the entire complex stability. Crucially, many of the residues involved in the neo-​interactions formed were found to not be conserved amongst the members of the BET family. This contrasts with the strictly conserved residues found deep in the BET ligand binding pocket. This observation led the authors to hypothesize that isoform-​specific neo-​interactions must contribute to the “cross-​energy” and so cooperativity, and so might be responsible for the intra-​ BET selectivity observed with MZ1. Three non-​conserved residues between the most cooperative system, i.e., the C-​terminal bromodomain (BD2) of Brd4 (Brd4BD2), and the least cooperative, i.e., the N-​terminal bromodomain (BD1) of Brd2 (Brd2BD1), were identified, and site-​directed mutant swaps designed to test this hypothesis. As expected, the identified residue swaps led to either losses or gains of cooperativity, consistent with the induced PPI playing an important role in isoform-​selective recruitment.15 Knowledge of the ternary complex structure enabled the field to begin to move into the modern era of PROTAC SBDD, enabling the rational design of more optimal degraders. In a first example of this, guided by the crystal structure, Gadd et al. hypothesized that a new conjugation vector might be exploited to better discriminate against the BET proteins within the ternary complex. They subsequently designed a new series of PROTACs bearing a different conjugation vector out of the tert-​butyl group of the VHL ligand. This led to the identification of AT1, a degrader that exhibited enhanced selectivity for Brd4 compared to MZ1, as shown by western blots and unbiased proteomics.15 Based on mutagenesis data, AT1 likely maintains identical relative orientation between VHL and Brd4BD2, and because of a fivefold loss of binary affinity at VHL it preferentially discriminates against the less cooperative complexes with Brd2 and Brd3. Together, this work illustrated the value of structural data for PROTAC design, as AT1 might not have readily emerged through unguided medicinal chemistry efforts. More recently, Testa et al. hypothesized that a macrocyclic PROTAC could be designed based on the MZ1 ternary crystal structure, with the aim to “lock” the PROTAC in its bioactive conformation.24 MacroPROTAC-​1 was designed aided by in silico calculations and synthesized through the development of a bespoke trifunctional linker, allowing cyclization through a new linker connecting a phenolic position on the VHL ligand and a methylene group adjacent to the BET ligand’s amide bond (Figure 2.4). Despite a 12-​fold weaker

26

26

Chapter 2 BD2

binary binding affinity to Brd4 , macroPROTAC-​1 was found to induce potent, fast and selective degradation of Brd4 that fared comparably to that of its non-​cyclic progenitor MZ1. A ternary complex crystal structure showed that macroPROTAC-​1 bound as designed, validating the design hypothesis.24

2.2.2.2 Structure-​guided design of SMARCA2/​4 PROTACs The work by Gadd et al. and Testa et al. nicely illustrated how knowledge of the ternary complex structure could guide innovative ideas for PROTAC design, starting from the structure of an already potent degrader. This suggested that ternary crystal structures could be used to improve degraders during the design process. Farnaby et al. elegantly demonstrated the power of X-​ray structures to efficiently drive the design of potent PROTAC degraders of SWI/​ SNF Related, Matrix Associated, Actin Dependent Regulator Of Chromatin, Subfamily A (SMARCA)2 and SMARCA4.9 SMARCA2 and SMARCA4 are the ATPase subunits of the chromatin remodelling BAF/​PBAF complexes, which are highly mutated throughout various cancers. BAF complex proteins have been identified as promising intervention targets for exploiting cancer vulnerabilities, for example synthetically lethal relationships between SMARCA2 and SMARCA4 in SMARCA4-​mutant lung tumours.78–​81 Both proteins contain a highly conserved bromodomain which is ligandable.82,83 However, the bromodomain was shown to be non-​essential for the BAF ATPase function, so bromodomain inhibitors were ineffectual.84 These findings inspired Farnaby et al. to leverage the bromodomain ligands to develop bifunctional degraders of SMARCA2/​4. A small-​molecule binder of the SMARCA2/​4 bromodomain previously identified by a team at Genentech/​Constellation was selected as the POI ligand.85 An X-​ray crystal structure of the ligand bound to the bromodomain of SMARCA2 (SMARCA2BD) was solved and suggested promising exit vectors.9 A small number of VHL-​based PROTACs were synthesized leveraging recently described high-​affinity VHL ligands of known binding modes.53 Specifically, an analogue of VHL ligand VH03251,52 bearing a fluoro-​cyclopropyl group to cap the terminal tert-​Leu group was selected and conjugated via a phenolic exit vector.18,67 A small set of nine PROTACs with linkers comprising different numbers of polyethylene glycol (PEG) units were synthesized and profiled. Isothermal titration calorimetry (ITC), fluorescence polarization (FP) and time-​resolved fluorescence resonance energy transfer (TR-​FRET) experiments (expanded in Section 2.3.2) were used to identify PROTAC1 as able to form the most cooperative (α~10) and populated ternary complex out of the initial set. This positive cooperativity compensated a >10-​fold loss of binary binding affinity at SMARCA2BD. Degradation studies showed PROTAC1 only partially degraded SMARCA2 (65% of protein depleted).9 Motivated by the encouraging results of the biophysical studies, crystallographic efforts were directed at PROTAC1. These resulted in a SMARCA2BD : PROTAC1 : VCB crystal structure solved at 2.25 Å resolution

 27

Structural and Biophysical Principles of Degrader Ternary Complexes BD2

27 15

(Figure 2.5A). Similar to the Brd4  : MZ1 : VCB crystal structure, favourable PROTAC-​induced neo-​protein–​protein and protein–​PROTAC contacts were observed within the ternary complex interface. PPIs induced between SMARCA2BD and VHL include a bidentate hydrogen bond between VHL R69 and the main chain carbonyl groups of SMARCABD T1462 and F1463, as well as a potential hydrogen bond between the side chain hydroxyl of Y112 of VHL and SMARCABD N1464. These are likely potentiated by a helical dipole acting on SMARCABD T1462, F1463 and N1464. Additionally, the main chain carbonyl of SMARCABD F1463 could form polar contacts with the fluoro-​cyclopropyl group of the VHL portion of PROTAC1. Overall, a total surface area of 2330 Å2 was buried in the process, of which approximately 640 Å2 was contributed by the induced PPI. In contrast to these favourable interactions, the PEG linker is accommodated towards the surface of the VHL protein in an unfavourable conformation. Based on this observation, PROTAC2 was designed, incorporating a phenyl ring in an attempt to rigidify the linker while retaining the overall relative orientation of the ternary complex. The newly integrated phenyl ring had the added benefit of reducing the polarity of the linker while being poised to form π-​stacking interactions with VHL’s Y98. PROTAC2 was found to be a significantly better degrader than PROTAC1, achieving maximal (>90%) degradation of SMARCA2 with a DC50 of 70 nM while displaying increased ternary complex cooperativity (α = 19 by FP) and stability. A crystal structure of the SMARCA2BD : PROTAC2 : VCB ternary complex revealed a near identical binding mode, with the added phenyl ring accurately recapitulating the linker geometry observed in PROTAC1, while forming the predicted π-​stacking interaction (Figure 2.5B). Extending the linker of PROTAC2 through incorporation of an oxygen atom led to ACBI1, an even better degrader than PROTAC2, achieving >90% degradation of SMARCA2 with a DC50 of 6 nM. The high target degradation selectivity of ACBI1 was shown in unbiased mass spectrometer proteomics experiments. ACBI1 induced potent antiproliferative effects in leukaemia cell lines and SMARCA4-​deficient melanoma cell lines known to depend on intact BAF complexes. The on-​target specific cellular activity of ACBI1 was evidenced by the lack of activity of the VHL-​inactive cis-​ACBI1 or SMARCA bromodomain inhibitor alone, as well as the inactivity of ACBI1 in cancer cell lines deficient in both SMARCA2 and SMARCA4.9 Together this study reveals the vital role that knowledge of the ternary complex formed by a PROTAC can play in the drug design process. In particular, structural elucidation allowed for rapid guided optimization of the linker to further improve ternary complex thermodynamic parameters such as cooperativity and stability. It shows how the linker’s role is not merely to join two binding ligands, but to also generate new intermolecular interactions resulting in significantly enhanced degradation activity. Ultimately, structure-​based PROTAC design allowed the development of a new chemical tool to knockdown BAF complex ATPases, and chemically validated genetic vulnerabilities in cancers.

28

newgenrtpdf

28 Chapter 2

Figure 2.5 Structure-​based design of SMARCA2/​4 PROTAC ACBI-​1.9 Views from crystal structures of SMARCA2BD in complex with VCB and either PROTAC1 (left, PDB : 6HAY) or PROTAC2 (center, PDB : 6HAX), highlighting a site for linker optimization. Substitution of a flexible five-​atom chain within the linker of PROTAC1 with a phenyl ring resulted in rigidification of the linker and new potential T-​stacking interactions between PROTAC2 and Y98 in VHL. PROTAC2 displayed improved SMARCA2 degradation activity in A549 cells compared to PROTAC1. Lengthening the linker of PROTAC2 through addition of an oxygen atom resulted in ACBI1 with further enhanced SMARCA2 degradation activity (right). SMARCA2BD PROTAC affinity ± VCB was measured by TR-​FRET.

 29

Structural and Biophysical Principles of Degrader Ternary Complexes 15

24

29 9

While the studies by Gadd et al., Testa et al. and Farnaby et al. made use of VHL as the recruited E3 ligase, CRBN has also received attention as an E3 ligase for both monovalent ligands, e.g., molecular glues and bifunctional PROTAC degraders.86,87

2.2.2.3 Ternary Structures of CRBN-​based PROTACs Nowak et al. attempted to structurally characterize the ternary complexes formed by CRBN-​recruiting PROTACs (dBETs) comprising a thalidomide moiety linked to JQ1.88 Such dBETs appeared to form negatively cooperative complexes with a selection of different BET bromodomains. Degradation studies on isolated proteins revealed differences in cooperativity correlated with selectivity in the degradation activity of the PROTACs, with the more efficient degraders forming the less negatively cooperative complexes. Crystals structures of CRBN and Brd4BD1 in complex with four related PROTACs (dBET23, dBET6, dBET70 and dBET55) were solved at resolutions between 3.3 and 6.3 Å.88 As expected, the thalidomide and JQ1 moieties of dBETs adopted similar binding modes to CRBN and Brd4BD1 as in their respective binary structures. A large surface area of ~2600 Å2 was buried in the complex. Some of the PPIs induced in the ternary complex appeared unlikely to positively contribute to the binding energy, as illustrated by interactions formed by the Brd4BD1-​specific Asp145 residue (a Glu residue in Brd4BD2). The four dBETs sampled linker lengths from 10 (dBET70) to 34 (dBET55) atoms and included different Brd4 exit vectors. Surprisingly, despite the variation in conjugation points and linkers, all four crystal structures were solved in identical space groups with similar unit cell parameters and overall ternary structure. Nowak et al. acknowledged the possibility that the observed ternary binding modes could be influenced by crystal packing, and indeed, extensive crystal contacts between symmetry mates at the PROTAC binding site indicate this to be a plausible conclusion. Mutational studies replacing residues identified in the crystal structures as positively contributing to ternary complex PPIs resulted in decreased ternary complex formation as measured in a TR-​ FRET ternary complex proximity assay. Furthermore, when the inopportunely placed Brd4BD1 D145, located in a hydrophobic environment in the crystal structures, was replaced by alanine, ternary complex formation was seen to increase in the same TR-​FRET assay. Together, these mutational studies corroborated to some extent the validity of the CRBN : dBET : Brd4 structures.88

2.2.3 Degraders as Monovalent Molecular Glues 2.2.3.1 Cereblon-​targeting Immunomodulatory  Drugs Thalidomide’s infamous teratogenic side effects led to the birth of more than 10 000 deformed children in the 1950s. Despite this, thalidomide continues to be used clinically as an effective treatment for leprosy (since 1998) and multiple myeloma (since 2006).89–​91 The protein target and mechanism of action of

30

30

Chapter 2

thalidomide and its potent derivatives lenalidomide and pomalidomide, collectively referred to as immunomodulatory drugs (IMiDs), remained unknown until the early 2010s.35,92 Ito et al. discovered the primary target for IMiDs to be CRBN, a protein of unknown function at the time, later revealed to form a complex with the DNA damage-​binding adaptor protein (DDB1), Cullin-​4A and regulator of Cullins-​1 (Roc-​1), whereby CRBN acts as the substrate receptor subunit.35,92–​94 The physiological effects of IMiDs were found to be through altering the substrate specificity of CRBN to recognize and so ubiquitinate new protein targets including transcription factors Ikaros and Aiolos.54,95,96 IMiDs are distinct from PROTACs because they induce the intracellular degradation of proteins for which they show no significant binding affinity and contain no linker moiety as such. Because of their ability to induce two proteins to interact with each other, they have historically also been referred to as “molecular glues” to evoke a mode of action similar to that of the plant hormone auxin. Auxin binds to the plant Cullin RING ligase substrate receptor TIR1 to induce recruitment, ubiquitination and subsequent degradation of transcriptional repressor protein IAA as its neo-​substrate.97 Lenalidomide, alone amongst the IMiDs, is effective at treating 5q deletion-​associated myelodysplastic syndrome (del-​5q MDS) through induced degradation of casein kinase 1α (CK1α, also Csnk1a1), which is a negative regulator of the Wnt signalling pathway.98 CC-​885, a CRBN-​binding compound developed by Celgene, displays antiproliferative effects against acute myeloid leukaemia (AML) cell lines, the action of which has been attributed to the CC-​885-​mediated, CRBN-​ directed degradation of G1 to S Phase Transition 1 (GSPT1), also known as eukaryotic peptide chain release factor GTP-​binding subunit (ERF3A).99,100 Many of the IMiD mediated neo-​substrates of CRBN, including Ikaros and Aiolos, contain a C2H2 zinc finger (ZF) domain, but neither CK1α nor GSPT1 do. Structural studies by X-​ray crystallography and cryo-​electron microscopy (cryo-​EM) were deployed to elucidate the substrate recognition motifs of neo-​ substrates (Figure 2.6). A  ternary CRBN : DDB1 : lenalidomide : CK1α crystal structure was solved by Petzold et al., exposing a β-​hairpin loop between residues 35 and 41 in CK1α as the primary recognition element on the neo-​ substrate (Figure 2.6C).101 One of the residues on the CK1α β-​hairpin loop, G40, was found to be crucial for CRBN binding, with mutational studies at that position confirming it to be a binding hotspot.101 Around the same time, the solving of a second CRBN–​DDB1 ternary crystal structure, this time with CC-​ 885 and GSPT1 bound, revealed a specific folded β-​hairpin loop structure containing G575, analogous to the critical G40 residue in CK1α (Figure 2.6D).99 Indeed, substitution of G575 in GSPT1 to any other residue resulted in a loss of CRBN affinity.99 The importance of a glycine in the surface turn of IMiD-​ dependent CRBN substrates was further verified through identification of a glycine residue at the Ikaros ZF binding site that, when mutated, resulted in loss of affinity to CRBN in the presence of pomalidomide.102 Sievers et al. further expanded the characterization of the CRBN IMiD-​ mediated degron to the wider family of ZF-​containing proteins and defined two new ternary crystal structures containing CRBN : DDB1 : pomalidomide

 31

Structural and Biophysical Principles of Degrader Ternary Complexes

31

Figure 2.6  Structural features of CRBN–​ IMiD molecular glues. Views from CRBN : IMiD : neo-​ substrate ternary complexes illustrating how CRBN can be directed to bind a range of neo-​substrates that lack in sequence identity, but all contain a similar β-​hairpin loop including an equivalently placed glycine residue (red). PDB IDs : (A) 6H0F,102 (B) 6H0G,102 (C) 5FQD101 and (D) 5HXB.99

in complex with the ZF-​ containing proteins Ikaros and ZNF692 (Figures 2.6A,B).102 ZF specificity to IMiD–​CRBN was demonstrated to lack a sequence-​based degron, instead highlighting the importance of induced PPIs in the regions adjacent to the β-​hairpin  loop.

2.2.3.2 DCAF15-​targeting Sulfonamide  Drugs Small molecule-​induced proteasomal degradation has been shown to be possible for molecular glues targeting E3 Cullin RING ligases other than CRBN. The aryl-​sulfonamides E7820, indisulam and tasisulam have been found to induce degradation of splicing factor RBM39 through binding to the E3 ligase DCAF15.103,104 Unlike the neo-​substrates of the CRBN–​IMiD complex which display undetectable binding affinity for CRBN in the absence of IMiDs, Du et al. found DCAF15 to display moderate binding affinity to the RRM1–​ RRM2 domain of RBM39 even in the absence of the aryl-​sulfonamide compounds (Kd~5 µM).105 This binary interaction appears to be greatly potentiated through addition of sulfonamide ligands such as E7820, resulting in >1 000-​ fold improvement in binding affinity between the two proteins. However, in another study published around the same time, the DCAF15–​RBM39

32

32

Chapter 2

interaction was not detectable in the absence of a sulfonamide ligand acting as molecular glue.106 The ternary structure of DCAF15 to E7820 and RBM39 has been revealed both by X-​ray crystallography and cryo-​EM.105–​107 The structures revealed a large neo-​interface between the two proteins around the sulfonamide ligand binder, containing extensive PPIs, consistent with aryl-​sulfonamides playing a role as enhancers of the PPIs.

2.2.4 Surface Areas Buried by PROTACs and Monovalent Glues The emerging structural information on degrader-​mediated ternary complexes highlights the importance of forming tightly associated complexes, ideally displaying a high degree of cooperativity whereby constructive neo-​ interactions are formed relative to unbound molecules or binary complexes. Such interactions lead to significant burial of otherwise solvent-​accessible surface area (SASA), which can contribute to the increased affinity and stability of the ternary complexes. To compare and contrast the burial of SASA in the different ternary complexes, we used the proteins, interfaces, structures and assemblies (PISA) tool from the protein data bank in Europe (PDBe).108 We compared the protein–​protein and protein–​ligand interfaces of PROTAC and molecular glue complex structures currently available; the results of which are presented in Table 2.1. Comparisons between the POI-​ligase PPIs reveal a general trend for a higher buried surface area for molecular glues than for PROTACs (Table 2.1). This can be understood as in the case of molecular glues, neither the degrader Table 2.1 Comparison of surface areas buried on degrader-​mediated ternary binding. Buried surface areas were calculated from X-​ray structures using PDBePISA tool. One representative structure for each published PROTAC and molecular glue series is presented. Where multiple molecules are present in the asymmetric unit, analysis for the molecule displaying the greatest total surface area or highest structure resolution is presented. Ternary complexes

PROTACs VHL : MZ1 : BRD4BD2 VHL : PROTAC2 :​ SMARCA2BD CRBN : dBET6 : BRD4BD1 Monovalent glues CRBN : lenalidomide :​ CK1α CRBN : CC-​885 : GSPT1 DCAF15 : E7820 :​ RBM39

Buried surface area (Å2) PDB Resolution POI-​ (Å) degrader

Ligase-​ degrader

POI-​ligase Total (PPI)

5T35 2.70 6HAX 2.35

980 940

1020 700

680 640

2680 2280

6BOY 3.33

840

640

1100

2580

5FQD 2.45

460

200

1180

1840

5HXB 3.60 6PAI 2.90

620 640

500 260

1260 1860

2380 2760

 33

Structural and Biophysical Principles of Degrader Ternary Complexes

33

nor the targeted E3 ligase display measurable/​high binding affinity to the POI, so the action of the molecular glue is therefore dependent on the induced PPIs, which appears to be reflected in higher protein–​protein interface area for molecular glue structures than seen with PROTACs. A clear outlier to this trend is the CRBN : dBET : Brd4 case. Firstly, the buried SASA from the POI-​ ligase PPI (1100 Å2) is almost double that of the VHL complexes (~650 Å2). This is also much greater than expected considering that the CRBN Brd4 PROTACs appeared to be negatively cooperative, while the VHL Brd4 PROTACs are positively cooperative. Nonetheless, the total buried SASA is comparable between the two systems. Drawing informative conclusions on the effect of buried surface area on the degradation activity of PROTACs and molecular glues is limited by the small number of ternary complex structures currently accessible. As more ternary complex structures emerge, hopefully these trends will become clearer. Crystal structures are snapshots of the ternary complex, which could sample multiple different states, so may not give in all cases complete information about the complex processes at play. Additional biophysical methods are therefore needed to expand our understanding of ternary complex structural and dynamic features in solution.

2.3 Ternary  Assays It is important to gain insights into association and dissociation equilibria and kinetics of the PROTAC ternary complexes to better understand their role in the mode of action and guide the drug design process. Over the past few years there has been intense activity aimed at developing and applying assays to measure different aspects of this three-​body equilibrium. Here we will explore some of the techniques that have been developed and discuss the different questions the various assays can answer. In principle, all binding assays can be equally well applied to investigate one, or the other, protein component of the ternary complex (POI or ligase); however, in practice, protein-​specific issues (e.g., availability of binders/​ probes/​appropriately tagged constructs that behave well under assay conditions) may restrict the assay formats available. Furthermore, while this section focuses on the biophysical characterization of PROTAC complexes, many of the techniques outlined can be, and to varying degrees have been, used to study molecular glue systems too.

2.3.1 Can My PROTAC Form a Ternary Complex? Once binding ligands to a POI and a ligase have been selected and the first PROTACs designed and synthesized, one of the first questions that must be answered is “to what extent does the PROTAC induce the formation of a ternary complex?” Crystallography or cryo-​EM can give a strong indication of ternary complex formation while at the same time elucidating their molecular details (see Section 2.2). However, obtaining ternary structures is low

34

34

Chapter 2

throughput and often challenging, when at all possible, so validating ternary complexes in the solution phase remains important.

2.3.1.1 Pull-​down  Assays Pull-​down assays rely on the tagging of POI-​1 with an affinity tag, e.g., Flag, providing a handle by which the protein can be enriched from solution, e.g., cell lysates containing PROTAC and POI-​2. If POI-​1 becomes engaged in a ternary complex prior to pull-​down, those binding partners may be co-​separated, providing a strong indication of ternary complex formation. Repeating the pull-​down experiment in the absence of PROTAC or in the presence of a negative control compound, i.e., one that has significantly hampered binding to either POI-​1 or POI-​2, is important for verifying that a successful outcome of the pull-​down is indeed due to PROTAC-​mediated interactions. This technique can be applied to mixtures of purified proteins or cell lysates, enabling qualitative indication of PROTAC specificity. Examples of PROTAC-​mediated co-​immunoprecipitation of POI and E3 ligase include studies of VHL-​based p38 degraders,20,109 CRBN-​based Mcl1 PROTACs110 and BET-​targeting electrophilic PROTACs recruiting the E3 ligases DCAF1670 or RNF114.111 This method has particular utility with molecular glues where the neo-​substrate or other partnering subunits of the E3 ligase may not be known.99,103,104,112

2.3.1.2 Proximity-​based Ternary Assays: AlphaScreen/​LISA and TR-​FRET Proximity-​based ternary complex assays that exploit detection techniques such as AlphaScreen/​AlphaLISA or TR-​FRET all function via a conceptually similar mechanism.113 POI-​1 is associated to either a “donor” or an “acceptor” species, and POI-​2 is associated to the other. When illuminated by light under a particular wavelength, the donor species produces a secondary signal capable of exciting the acceptor species only when they are in close proximity. The excited acceptor species subsequently produces a fluorescent signal of a characteristic wavelength, providing a readout for the experiment. The method by which the donors excite the acceptors differ between these two assays: a short-​lived singlet oxygen species is produced by excitation of the donor beads in the case of AlphaScreen/​LISA, which can go on to excite the acceptor species; in contrast, FRET occurs between suitable fluorophores at different wavelengths. Both assays are highly sensitive to proximity; therefore, a strong signal is only observed when the donor and acceptor are bound within a short distance of each other, such as when associated in a ternary complex (Figure 2.7, top). A titration series of PROTAC will generally result in signal increasing with PROTAC concentration until a maximum is reached, after which the signal will decrease again. The decrease in signal at high PROTAC concentration (Figure 2.7, bottom) is due to the hook effect (see Section 2.2.1). As peak intensity and concentration are influenced by multiple factors, proximity assays can only deliver a semi-​quantitative readout of ternary complex

 35

Structural and Biophysical Principles of Degrader Ternary Complexes

35

Figure 2.7 Principles of proximity-​based ternary complex assays. Top: schematic depiction of the AlphaScreen/LISA concept. Fluorescent emission at a characteristic wavelength (520–​620 nm) is greatly amplified when donor and acceptor beads are within a short distance of each other, such as when associated to a PROTAC-mediated ternary complex. Bottom: representative AlphaScreen/LISA trace displaying the characteristic hook effect.

formation. Specifically in the case of AlphaScreen/​LISA, the bead-​based set-​ up with multiple derivatization sites can potentially influence signal intensity in unpredictable manners. Regardless, proximity-​based assays are amongst the highest throughput of the techniques so far applied to studying PROTAC ternary complexes, only requiring a very small amount of protein for each experiment. AlphaLISA proximity assays were used by Gadd et al. as an initial binding screen, resolving a stark difference in profiles of ternary complexes formed between VHL, a library of MZ-​based PROTACs and the BET bromodomains, despite highly similar binary affinities to those same bromodomains.15 Alpha assays have also been used to monitor formation of ternary complexes of CRBN-​based BET PROTACs,59,114,115 PROTACs targeting other bromodomain

36

36

Chapter 2 19,116

26

9

proteins such as Brd7/​9, TRIM24, SMARCA2/​4, protein kinases,20,109 117 and Bcl-​2/​Bcl-​xL. AlphaLISA has been used to demonstrate dimerization of recombinant VHL in vitro using VHL-​based homo-​PROTACs.18 AlphaLISA assays were also used to monitor ternary complex formation of SMARCA2/​4 PROTACs from cell lysates.9 TR-​FRET assays have been used extensively to characterize the differential recruitment of neo-​ substrates to CRBN by IMiDs,101 IMiD-​based 88,118 PROTACs, and more recently to monitor formation of ternary complexes between DCAF15, sulfonamide glue ligands and RBM39,105–​107,119 and VHL-​ based EED-​targeting PROTACs.120

2.3.1.3 Surface Plasmon Resonance Surface plasmon resonance (SPR) is another versatile biophysical method capable of directly observing both binary and ternary complex formation.44 An SPR chip is derivatized by immobilizing POI-​1 to the surface of one of the gold-​plated flow-​cells in the chip. POI-​1 immobilization is critical to the success of the SPR binding assay, so efforts must be made to maximize the activity and binding capacity of the immobilized protein. Direct immobilization is preferable, either by amine coupling (which leads to heterogeneity of POI-​1 orientations) or ideally via an affinity tag or biotin tag at a single site on POI-​1, e.g., the N-​terminus, thus allowing for a more homogeneous surface display. A solution containing PROTAC, or PROTAC pre-​incubated with equimolar or excess of POI-​2, is then flowed over the surface of the chip. The signal response observed is proportional to the increase in surface density of the chip, allowing differentiation between PROTAC alone associating to form a binary complex and PROTAC + POI-​2 forming a ternary complex with POI-​1 on the surface of the chip (Figure 2.8). The setup of this experiment means only the immobilized protein requires an affinity tag, whereas POI-​2 can remain tag-​free. Additionally, once immobilized, a POI-​1 surface can often be reused multiple times against a library of different PROTACs, POI-​2 proteins and combinations thereof. SPR is routinely used to get binary data, and is rapidly gaining in popularity in ternary complex studies since first described by Roy et al. for this purpose28 (covered in more detail in Section 2.3.4). SPR is unique amongst the biophysical techniques that have so far been applied to PROTACs as it is capable of quantitatively measuring the association and dissociation of the complexes under study in real time. The access to kinetic data afforded by SPR can become critical to the direction of PROTAC optimization projects and is discussed in greater detail later in this chapter. The biggest asset of the technique (real-​time kinetic measurements) also proves to be one of its drawbacks, as the timescales over which PROTACs dissociate, particularly for high-​affinity PROTACs and ternary complexes, can become quite lengthy, reducing throughput.28,121 As single cycle kinetic experiments only require the measurement of a single dissociation curve per analyte concentration series repeat, this method can greatly increase the throughput over the more traditional multi-​cycle kinetic approach. Examples of experimentally derived single-​cycle kinetic SPR traces are presented in Figure 2.12. Recent advances

 37

Structural and Biophysical Principles of Degrader Ternary Complexes

37

Figure 2.8 Schematic representation of binary and ternary SPR experiments. POI-​1 is immobilized to a certain surface density (blue arrow). When PROTAC is flowed over POI-​1, binary binding results in a small proportional increase in surface density (red arrow) relative to the POI-​1 immobilized baseline and a low maximal response (Rmax). When POI-​2 + PROTAC is instead introduced, and a ternary complex forms, a larger relative increase in surface density occurs (green arrow) leading to higher Rmax. Example binding profiles for a single injection of PROTAC and PROTAC + POI-​2 are presented (bottom left and right, respectively).

in newer, highly parallel SPR instruments can also help to increase assay turnover, but throughput inevitably lags behind the proximity-​based assays. Non-​ denaturing nano-​ electrospray ionization mass spectrometry has recently been shown by Beveridge et al. to be a promising, semi-​quantitative method for detecting ternary complexes under native conditions.122 This label-​free technique can detect and differentiate between multiple binding stoichiometries and binding partners within a single measurement and can detect even low-​populated species. These features highlight the utility of mass spectrometry both in in vitro studies or as a follow-​up method for investigating ternary complexes extracted from pull-​down assays.

2.3.2 How Tightly Does My Ternary Complex Bind? An increasingly well-​stocked toolbox of biophysical techniques is available for quantitatively assessing the affinities of PROTACs in the assembly of ternary complexes. These assays can be broadly grouped into “direct” and

38

38

Chapter 2

“competition” binding assays. They allow resolution of the Kd for an interaction, and therefore the ΔG for binary, or ternary interactions, leading to the total ΔG of complex formation, as discussed in Section 2.2.1.

2.3.2.1 Competition  Assays Competition assays utilize the displacement of a reporter species from POI-​1 by a PROTAC or PROTAC + POI-​2, resulting in a change in signal. The reporter species is based on a binder (e.g., small molecule, peptide, binding protein) to POI-​1 that is coupled to some manner of signal-​producing tag, the identity of which depends on the assay. AlphaScreen/​LISA and TR-​FRET can be used either in a ternary complex proximity mode as described in Section 2.3.1.3, or as a competition-​based binding assay. In the competition format the donor species is coupled to POI-​1, and the acceptor species is coupled to a POI-​1 binder (or vice versa). The POI-​1 binder can then be competed out with PROTAC or PROTAC + POI-​2, resulting in a decrease in signal as POI-​1: PROTAC or POI-​1: PROTAC : POI-​2 complexes form (Figure 2.9A). This competition mode therefore allows measurement of both binary and ternary binding in a side-​by-​side assay. This assay was used to compare and contrast binary and ternary Kd and so measure cooperativity of VHL-​based PROTACs targeting the SMARCA2/​4 bromodomains,9 and to measure binding of IMiDs and neo-​substrates to CRBN.102,123 Competition FP assays allow investigating PROTAC systems without requiring the tagging of either POI or ligase. In an FP experiment it is possible to measure the displacement of a small-​molecule probe that binds to POI-​1, upon increasing concentration either of PROTAC alone, or PROTAC + POI-​2. The small-​molecule probe contains a fluorophore that is able to depolarize plane-​polarized light when free in solution, but maintains polarization when bound to POI-​1.44 By measuring the intensity of light emitted perpendicular to the orientation of the incident light, an estimation of the proportion of probe displaced from binding to POI-​1 can be made (Figure 2.9B). Competition assays are generally high-​throughput but are more susceptible to context-​specific confounding issues which may impede direct comparisons between PROTACs, such as when compounds display high intrinsic absorbance or other unwanted properties such as fluorescent or singlet oxygen quenching. The observation of either left or right shifts in titration curves monitoring displacement of a POI-​1 binder by PROTAC in the absence or presence of POI-​2 suggests cooperativity, hence the presence of POI-​2 affects the equilibrium (Figure 2.10). Competition assays are not affected by the hook effect, allowing more quantitative estimation of binding than e.g., ternary proximity assays. Several recent studies have demonstrated the use of FP assays to monitor binary and ternary complexes and so measure cooperativity of PROTACs.9,28

 39

Structural and Biophysical Principles of Degrader Ternary Complexes

39

Figure 2.9 Schematic representation of competition-​based assays. (A) AlphaScreen-​ based competition assays measure a decrease in signal as a binder is displaced from POI-​1 by PROTAC or PROTAC + POI-​2. TR-​FRET assays work by a broadly similar mechanism. (B) In FP assays, rotated plane polarized light is increasingly detected as a fluorescent probe is displaced from POI-​1 by a PROTAC or PROTAC + POI-​2.

40

40

Chapter 2

Figure 2.10  Dose–​ response profiles from competitive ternary complex assays. Schematic of FP profiles measuring displacement of a fluorescence probe from POI-​1 by either PROTAC alone (black dotted line), or PROTAC : POI-​2 complex, in the case of a cooperative (green), non-​cooperative (cyan) or negatively cooperative PROTAC (red).

2.3.2.2 Direct Binding Assays Ternary proximity assays such as the previously discussed AlphaScreen/​LISA and TR-​FRET produce a readout on formation of ternary complexes. As a high signal is only obtained when POI and ligase are in close proximity, these methods are attractive for obtaining ternary complex affinity data but are not able to report on binary affinities. Binding assays are therefore the preferred choice to measure binary and ternary affinities directly. SPR is a robust method that can be used for these purposes (described in Section 2.3.1.3 and expanded on further in Section 2.3.4). Another gold standard, entirely label-​free technique to measure direct binding is isothermal titration calorimetry (ITC). By measuring minute changes in the temperature of a binding reaction on titrating a solution containing one binding partner into another, ITC is able to accurately determine binding stoichiometry (N), the Kd and the change in enthalpy (ΔH) (Figure 2.11). Because the free energy of binding (ΔG) can be calculated from Kd, the entropy (ΔS) of complex formation is measured indirectly too.44 ITC assays have been developed specifically for the purpose of studying PROTAC ternary complexes.15 The conventional approach for studying protein–​ ligand interactions by ITC involves titrating a solution of small molecules in the syringe into a solution of protein in the sample cell. While this approach can be used to study the binary binding of the PROTAC to either POI alone, it is not suitable to study ternary complex formation because of the hook effect. To circumvent this limitation, Gadd et al. established a sequential “reverse-​titration” approach.15 This involves first titrating POI-​1 into PROTAC (Figure 2.11A), followed by titration of POI-​2 into the binary complex POI-​1 : PROTAC previously formed in the sample cell upon completion of the first titration. By ensuring that the PROTAC is fully saturated by POI-​1 at the end of a first titration (or a separate sample preparation), any potential issues due to the hook effect are eliminated. As POI-​1 and POI-​2 typically do not interact with each other with any measurable affinity without the PROTAC bound, any stoichiometric excess of POI-​1 relative to PROTAC inside the sample cell is not expected to contribute to the heat signal measured during the second titration –​hence ensuring that only ternary complex formation is being monitored in the process. This approach allows for

 41

Structural and Biophysical Principles of Degrader Ternary Complexes

41

determination of the full thermodynamic picture of both binary and ternary PROTAC complexes. By comparing the ITC curve from the titration of POI-​1 into POI-​2 : PROTAC (Figure 2.11B) with that obtained from a separate experiment in which POI-​1 is titrated into PROTAC alone (Figure 2.11A), one can readily obtain the cooperativity of the PROTAC ternary complex system. Moreover, by summing up the free energies of the sequential binding processes (ΔG1 + ΔG2, or ΔG3 + ΔG4), one can obtain the stability of the PROTAC ternary complex. ITC thus allows the elucidation of the full thermodynamic picture of PROTAC ternary complex equilibria. This approach has been used to characterize VHL-​based PROTACs targeting the BET proteins Brd2, Brd3 and Brd4,15,23,24,124 Brd9,116 SMARCA2,9 as well as dimers of VHL formed by homo-​PROTACs.18

Figure 2.11 Illustrative results from a reversed ITC titration set-​up to monitor binary and ternary complexes and measure cooperativity. (A) Titration of POI-​1 into PROTAC alone; (B) titration of POI-​1 against PROTAC in the presence of a saturating amount of POI-​2. The latter sample can be obtained from a separate titration of POI-​2 into PROTAC, or upon pre-​mixing PROTAC with POI-​2 in slight stoichiometric excess. From these titrations, the following parameters of complex formation are measured: stoichiometry (N), association constants Ka, from which Kd and ΔG can be calculated; and binding enthalpy ΔH, allowing calculation of binding entropy (ΔS). The ratio between binary and ternary Kd determines cooperativity (α) values. Data collected by Dr Morgan Gadd, University of Dundee.

42

42

Chapter 2

ITC is a highly trusted technique that does not require any labelling of either protein or PROTAC, making it a gold-​standard technique for validating PROTAC systems or any biomolecular systems more generally. As a particularly protein-​ hungry technique of limited throughput, however, ITC tends to be better suited to high-​quality characterization of fewer systems of interest rather than as a primary screening assay. With the exception of the ternary proximity assays, all other biophysical techniques outlined in this section can be used to determine the affinity for a PROTAC to one of its protein binding partners in the presence or absence of the other. Comparing the results of these binary and ternary affinity experiments thus delivers highly valuable information that can shine a light on the relative cooperativity and stability of PROTAC ternary complexes.

2.3.3 To What Extent Is My Ternary Complex Cooperative? Within a ternary complex arise intermolecular interactions, not just between PROTAC and target proteins, but also between POI and the ligase in the form of de novo PPIs. Cooperative PROTACs can therefore gain potency over and above that of their constituent binders. Furthermore, PPIs can be highly specific to a large number of residues at the neo-​interface, opening the possibility of degrading specific target protein isoforms, or differentiating between close protein homologues or variants.15,99,101 These features mean that there is often a tight correlation between enhanced cooperativity and ternary complex stability and improved PROTAC specificity and target degradation potency and selectivity, which underline the importance of these as vital optimization parameters.9 We have frequently observed cases where a penalty to the binary affinity of a PROTAC to one of its POI is more than recouped through cooperativity.9,15,23,24 For the most comparable results, cooperativity should ideally be determined for a specific PROTAC system within the same assay, and confirmed by an orthogonal assay. Any assay that can be run in both binary and ternary modes, in which binding of a PROTAC to POI-​1 can be measured in the presence and absence of POI-​2, can be used to determine cooperativity. Schematic traces from, for example, an FP competition experiment are presented in Figure 2.10 displaying shifts in binding curves arising from a PROTAC inducing a positively, negatively and non-​cooperative ternary complex.

2.3.4 How Long Does My Ternary Complex Last? Work on PROTAC ternary complexes has revealed that long-​lived, more stable ternary complexes drive faster target degradation as a result of greater target ubiquitination.9,28 This highlights the importance of having reliable tools to probe the dissociation kinetics of PROTAC ternary complexes.28,29 SPR is unique in its ability to deconvolute the association and dissociation parameters from the ternary complex affinity, providing vital kinetic information about the systems under investigation. The value of measuring ternary complex half-​lives to probe PROTAC specificity was illustrated by Roy et al. in their SPR study of BET degraders.28 The well-​characterized PROTAC MZ1 was

 43

Structural and Biophysical Principles of Degrader Ternary Complexes

43

studied as a model system. MZ1 had previously been shown to preferentially induce degradation of Brd4 (described in Section 2.1.3).15,17 To enable further investigation of the MZ1 ternary complexes, a construct of VCB chemically biotinylated at the C-​terminus of ElonginB (biotin-​VCB) was designed and immobilized on a streptavidin-​loaded SPR chip. SPR experiments comparing the half-​life of ternary complexes comprising biotin-​VCB, MZ1 and either the BD1s or BD2s of Brd2, Brd3 and Brd4 revealed the BD1s of each BET protein formed much shorter lived complexes (t1/​2 < 1 s) than the BD2s.28 Significantly, the stability of ternary complexes containing Brd2BD2 and Brd4BD2 were much higher (t1/​2 = 70 and 130 s, respectively) than the VCB : MZ1 : Brd3BD2 ternary complex (t1/​2 = 6 s). Representative single-​cycle SPR traces of MZ1 binding to VCB in the presence of saturating concentrations of either Brd3BD2 or Brd4BD2 are presented in Figure 2.12. This was found to correlate with a greater initial rate of proteasomal degradation of Brd4 (and Brd2) compared to Brd3.28,29 The crystal structure of the VCB : MZ1 : Brd4BD2 complex suggested the observed difference in ternary complex stability could be explained by a single amino acid difference between bromodomains (E344 in Brd3BD2; G382 or G386 in Brd2BD2 and Brd4BD2, respectively). These observations were validated by mutational studies implementing reciprocal amino acid swaps in constructs of Brd3BD2 (E344→G) and Brd4BD2 (G386→E), followed by SPR analysis of the resultant VCB : MZ1 : bromodomain kinetic profiles. Satisfyingly, these amino acid swaps resulted in an exchange of ternary complex affinity, with Brd3BD2,E344G forming a longer-​lived species (t1/​2 = 59 s), and Brd4BD2,G386 forming a shorter-​lived species (t1/​2 = 3.5 s) than their respective native variants.28 A later example of ternary complex kinetics being monitored in the study of PROTACs was shown by Pillow et al. in the analysis of GNE-​987, a selective, picomolar degrader of Brd4. GNE-​987 was shown by SPR to form a ternary complex with Brd4BD1 with a t1/​2 of ~4000 s which equates to >1 h half-​life, likely underpinning its remarkable degradation potency.121 SPR is not the only biophysical technique that can be used to monitor binding kinetics. Bilayer interferometry (BLI) can also be used to this end. BLI has been applied to measure binary affinities of molecular glues.105 To the best of our knowledge, the dissociation kinetics accessible via this technique have not yet been explored in PROTACs.

2.3.5 Does the PROTAC Induce Ternary Complex Formation in Cells? The extent to which biophysical studies with recombinant protein or isolated domains are bona-​fide representations of the cellular environment, for example within the native proteins as part of multi-​subunit complexes, must ideally be validated experimentally. As the first layer of key PPIs are formed around the PROTAC binding site, studying the ligand binding domain is an appropriate first approximation.9,15,24 However, discrepancies have been observed too. For example, Khan et al. developed a VHL-​based PROTAC (DT2216) derived from the Bcl-​2/​Bcl-​xL inhibitor ABT263 and found it to be selective for Bcl-​xL over

44

44

Chapter 2

Figure 2.12 Single-​cycle ternary complex SPR assays. Real-​time kinetic traces from SA chip–​immobilized biotin–​VCB (~1000 RU). MZ1 was flowed over the chip at 0.8, 4, 20, 100 then 500 nM concentrations in the presence of saturating concentrations of either (A) Brd3BD2 or (B) Brd4BD2. SPR traces reveal that MZ1 induces more stable complexes between biotin-​VCB and Brd4BD2 (t1/​2 = 130 s, Kd = 1 nM) than between biotin-​VCB and Brd3BD2 (t1/​2 = 6 s, Kd = 8 nM). Single-​cycle data differ from more conventional multi-​cycle data in that subsequent injections are performed prior to full dissociation of the pre-​ bound species. Data collected by Dr Michael Roy, University of Dundee.

Bcl-​2 as a result of preferential formation of a VHL : DT2216 : Bcl-​xL ternary complex over the corresponding VHL : DT2216 : Bcl-​2 in NanoBRET ternary complex assays in live cells.117 This appeared inconsistent with the results of AlphaLISA assays with recombinant proteins that instead detected ternary complexes with both target proteins. Cellular assays are important for validating target engagement in cell, and can be further applied to monitor ternary complex formation, POI ubiquitination, and degradation in real time in living cells.29 Two recently developed cell-​based assays have opened up the possibility of investigating the real-​time

 45

Structural and Biophysical Principles of Degrader Ternary Complexes

45

kinetics of ternary complex formation under more native conditions; separation of phases-​based protein interaction reporter (SPPIER), and bioluminescence resonance energy transfer (BRET).29,125

2.3.5.1 Separation of Phases-​based Protein Interaction Reporter Assay (SPPIER) Fluorescence imaging can be applied to monitor induction of protein–​protein interactions in cells by small molecules, including PROTACs and molecular glues. Towards this goal, Chung et al. developed a technique that they dubbed SPPIER125 by repurposing a method they had previously described to study protein kinases.126 Establishing a SPPIER assay begins with the genetic fusion of POI-​1 to enhanced green fluorescent protein (EGFP), which in turn is fused to a 30-​amino acid homo-​oligomeric tag (HO-​Tag)3. Likewise, a fusion protein containing POI-​2, EGFP and the 33-​amino acid HO-​Tag6 are constructed, and a cell line overexpressing these two fusion proteins is prepared. In solution the HO-​Tag3 and HO-​Tag6 form hexameric and tetrameric complexes, respectively. On addition of a PROTAC, POI-​1 and POI-​2 form ternary complexes, which, due to the multi-​valency introduced by the HO-​Tags, result in networks of POI–​EGFP–​HO-​Tag proteins rapidly forming, causing dense droplets of fluorescent protein to become observable by fluorescence cell imaging.125 The reversible nature of this ligand-​induced phase separation means washing out the PROTAC and observing the dissipation of the droplets can potentially inform on the dissociation kinetics of the ternary complex.

2.3.5.2 Bioluminescence Resonance Energy Transfer (BRET) As developed by Promega, NanoBRET experiments require the tagging of POI-​1 with a short, 11-​amino acid “HiBiT peptide” which, when co-​expressed with an 18 kDa LgBiT protein, spontaneously forms NanoBiT luciferase.127 Tagging of POI-​2 with the hydrolase “Halo” protein is also required. Luciferase acts as an energy donor, and the Halo tag, when bound to a separately added substrate, acts as the energy acceptor, together producing a luminescent signal when in close proximity, such as when the ternary complex forms.128 Cellular nanoBRET/​nanoBIT assays have been applied to the development of VHL-​ based Brd7/​9 degrader VZ185,116 Bcl-​2/​Bcl-​xL PROTAC DT2216,117 as well as CRBN-​based Cdk4/​6 degraders.129 Cellular assays are limited by the requirements to tag both POI and ligase. Because of interference by potentially limiting factors of cellular permeability and intracellular stability of compounds, accurately quantifying the thermodynamic and kinetic features of ternary complexes down to the resolution required in the native environment remains a challenge. It is therefore still crucial to have an array of robust, quantitative assays for measuring the biophysics of ternary PROTAC complexes in vitro with recombinant protein which allow control of the purity and concentration of the assay components.

46

46

Chapter 2

2.4 Concluding Remarks This chapter outlined some of the formative ideas and pioneering proof-​ of-​concept experiments that paved the way in the early days of targeted protein degradation. The identification of high-​affinity ligase binders dramatically increased the viability of this exciting new modality, triggering an explosion of interest in the field. Experimentally derived ternary complex structures of PROTACs and molecular glues first illuminated the unpredictable de novo protein–​protein interactions, opening up the possibility of using SBDD to optimize these small-​molecule protein degraders. Finally we took stock of the cutting-​edge, evolving arsenal of biophysical techniques currently under use by researchers to tease apart the complexities of the binding events that dictate the degrader mode of action. Together, these molecular structural techniques represent powerful tools to direct the design and improvement of degraders capable of pushing forwards the limits of science in the form of novel small-​molecule probes able to provide new ways to investigate and modulate protein function. As we learn to turn PROTACs into drugs we may finally have the opportunity to eliminate a whole range of therapeutically critical target proteins deemed “undruggable” by conventional medicine.

2.5 Acknowledgments We are indebted to many colleagues, collaborators and laboratory members over the years for valuable input and discussions. We thank Will Farnaby for helpful discussions and reading of the manuscript. We apologize to the authors of many articles that could not be cited because of space restrictions.

2.5.1 Funding The Ciulli laboratory’s work on E3 ligase targeting small molecules and PROTACs has received funding from the European Research Council (ERC) under the European Union’s Seventh Framework Programme (FP7/​2007–​ 2013) as a Starting Grant to A.C. (grant agreement No. ERC-​2012-​StG-​311460 DrugE3CRLs). The Ciulli laboratory receives or has received sponsored research support from Amphista Therapeutics, Boehringer Ingelheim, Eisai, Nurix, and Ono Pharmaceuticals.

2.5.2 Conflict of Interest Statement A.C. is a scientific founder, shareholder, non-​executive director and consultant of Amphista Therapeutics, a company that is developing targeted protein degradation therapeutic platforms. The remaining author reports no competing interests.

 47

Structural and Biophysical Principles of Degrader Ternary Complexes

47

References 1. G. Koscielny, P. An, D. Carvalho-​Silva, J. A. Cham, L. Fumis, R. Gasparyan, S. Hasan, N. Karamanis, M. Maguire, E. Papa, A. Pierleoni, M. Pignatelli, T. Platt, F. Rowland, P. Wankar, A. P. Bento, T. Burdett, A. Fabregat, S. Forbes, A. Gaulton, C. Y. Gonzalez, H. Hermjakob, A. Hersey, S. Jupe, Ş. Kafkas, M. Keays, C. Leroy, F.-​J. Lopez, M. P. Magarinos, J. Malone, J. McEntyre, A. Munoz-​Pomer Fuentes, C. O’Donovan, I. Papatheodorou, H. Parkinson, B. Palka, J. Paschall, R. Petryszak, N. Pratanwanich, S. Sarntivijal, G. Saunders, K. Sidiropoulos, T. Smith, Z. Sondka, O. Stegle, Y. A. Tang, E. Turner, B. Vaughan, O. Vrousgou, X. Watkins, M.-​J. Martin, P. Sanseau, J. Vamathevan, E. Birney, J. Barrett and I. Dunham, Nucleic Acids Res., 2017, 45, D985–​D994. 2. E. Valeur, S. M. Guéret, H. Adihou, R. Gopalakrishnan, M. Lemurell, H. Waldmann, T. N. Grossmann and A. T. Plowright, Angew. Chem., Int. Ed., 2017, 56, 10294–​10323. 3. W. A. Weiss, S. S. Taylor and K. M. Shokat, Nat. Chem. Biol., 2007, 3, 739–​744. 4. J. W. Scannell, A. Blanckley, H. Boldon and B. Warrington, Nat. Rev. Drug Discovery, 2012, 11, 191–​200. 5. R. L. Setten, J. J. Rossi and S. P. Han, Nat. Rev. Drug Discovery, 2019, 18, 421–​446. 6. A. C. Lai and C. M. Crews, Nat. Rev. Drug Discovery, 2017, 16, 101–​114. 7. S. J. Hughes and A. Ciulli, Essays Biochem., 2017, 61, 505–​516. 8. A. Mullard, Nat. Rev. Drug Discovery, 2019, 18, 237–​239. 9. W. Farnaby, M. Koegl, M. J. Roy, C. Whitworth, E. Diers, N. Trainor, D. Zollman, S. Steurer, J. Karolyi-​Oezguer, C. Riedmueller, T. Gmaschitz, J. Wachter, C. Dank, M. Galant, B. Sharps, K. Rumpel, E. Traxler, T. Gerstberger, R. Schnitzer, O. Petermann, P. Greb, H. Weinstabl, G. Bader, A. Zoephel, A. Weiss-​ Puxbaum, K. Ehrenhöfer-​ Wölfer, S. Wöhrle, G. Boehmelt, J. Rinnenthal, H. Arnhof, N. Wiechens, M.-​Y. Wu, T. Owen-​Hughes, P. Ettmayer, M. Pearson, D. B. McConnell and A. Ciulli, Nat. Chem. Biol., 2019, 15, 672–​680. 10. T. I. Oprea, C. G. Bologa, S. Brunak, A. Campbell, G. N. Gan, A. Gaulton, S. M. Gomez, R. Guha, A. Hersey, J. Holmes, A. Jadhav, L. J. Jensen, G. L. Johnson, A. Karlson, A. R. Leach, A. Ma’ayan, A. Malovannaya, S. Mani, S. L. Mathias, M. T. McManus, T. F. Meehan, C. von Mering, D. Muthas, D.-​T. Nguyen, J. P. Overington, G. Papadatos, J. Qin, C. Reich, B. L. Roth, S. C. Schürer, A. Simeonov, L. A. Sklar, N. Southall, S. Tomita, I. Tudose, O. Ursu, D. Vidovic, A. Waller, D. Westergaard, J. J. Yang and G.  Zahoránszky-​Köhalmi, Nat. Rev. Drug Discovery, 2018, 19, A68. 11. P. M. Cromm and C. M. Crews, Cell Chem. Biol., 2017, 24, 1181–​1190. 12. T. Wang, K. Birsoy, N. W. Hughes, K. M. Krupczak, Y. Post, J. J. Wei, E. S. Lander and D. M. Sabatini, Science, 2015, 350, 1096–​1101. 13. A. Yesbolatova, Y. Tominari and M. T. Kanemaki, Drug Discovery Today: Technol., 2019, 31,  91–​98.

48

48

Chapter 2

4. T. A. F. Cardote, M. S. Gadd and A. Ciulli, Structure, 2017, 25, 901–​911.e3. 1 15. M. S. Gadd, A. Testa, X. Lucas, K.-​H. Chan, W. Chen, D. J. Lamont, M. Zengerle and A. Ciulli, Nat. Chem. Biol., 2017, 13, 514–​521. 16. A. Plechanovová, E. G. Jaffray, M. H. Tatham, J. H. Naismith and R. T. Hay, Nature, 2012, 489, 115–​120. 17. M. Zengerle, K.-​H. Chan and A. Ciulli, ACS Chem. Biol., 2015, 10, 1770–​1777. 18. C. Maniaci, S. J. Hughes, A. Testa, W. Chen, D. J. Lamont, S. Rocha, D. R. Alessi, R. Romeo and A. Ciulli, Nat. Commun., 2017, 8, 830. 19. D. Remillard, D. L. Buckley, J. Paulk, G. L. Brien, M. Sonnett, H.-​S. Seo, S. Dastierdi, M. Wühr, S. Dhe-​Paganon, S. A. Armstrong and J. E. Bradner, Angew. Chem., Int. Ed., 2017, 56, 5738–​5743. 20. D. P. Bondeson, B. E. Smith, G. M. Burslem, A. D. Buhimschi, J. Hines, S. Jaime-​Figueroa, J. Wang, B. D. Hamman, A. Ishchenko and C. M. Crews, Cell Chem. Biol., 2018, 25, 78–​87.e5. 21. H.-​ T. Huang, D. Dobrovolsky, J. Paulk, G. Yang, E. L. Weisberg, Z. M. Doctor, D. L. Buckley, J.-​H. Cho, E. Ko, J. Jang, K. Shi, H. G. Choi, J. D. Griffin, Y. Li, S. P. Treon, E. S. Fischer, J. E. Bradner, L. Tan and N. S. Gray, Cell Chem. Biol., 2018, 25, 88–​99.e6. 22. C. M. Olson, B. Jiang, M. A. Erb, Y. Liang, Z. M. Doctor, Z. Zhang, T. Zhang, N. Kwiatkowski, M. Boukhali, J. L. Green, W. Haas, T. Nomanbhoy, E. S. Fischer, R. A. Young, J. E. Bradner, G. E. Winter and N. S. Gray, Nat. Chem. Biol., 2018, 14, 163–​170. 23. A. Testa, X. Lucas, G. V. Castro, K.-​H. Chan, J. E. Wright, A. C. Runcie, M. S. Gadd, W. T. A. Harrison, E.-​ J. Ko, D. Fletcher and A. Ciulli, J. Am. Chem. Soc., 2018, 140, 9299–​9313. 24. A. Testa, S. J. Hughes, X. Lucas, J. E. Wright and A. Ciulli, Angew. Chem., Int. Ed., 2020, 59, 1727–​1734. 25. X. Han, L. Zhao, W. Xiang, C. Qin, B. Miao, T. Xu, M. Wang, C.-​Y. Yang, K. Chinnaswamy, J. Stuckey and S. Wang, J. Med. Chem., 2019, 62, 11218–​11231. 26. L. N. Gechijian, D. L. Buckley, M. A. Lawlor, J. M. Reyes, J. Paulk, C. J. Ott, G. E. Winter, M. A. Erb, T. G. Scott, M. Xu, H.-​S. Seo, S. Dhe-​ Paganon, N. P. Kwiatkowski, J. A. Perry, J. Qi, N. S. Gray and J. E. Bradner, Nat. Chem. Biol., 2018, 14, 405–​412. 27. C. Maniaci and A. Ciulli, Curr. Opin. Chem. Biol., 2019, 52, 145–​156. 28. M. J. Roy, S. Winkler, S. J. Hughes, C. Whitworth, M. Galant, W. Farnaby, K. Rumpel and A. Ciulli, ACS Chem. Biol., 2019, 14, 361–​368. 29. K. M. Riching, S. Mahan, C. R. Corona, M. McDougall, J. D. Vasta, M. B. Robers, M. Urh and D. L. Daniels, ACS Chem. Biol., 2018, 13, 2758–​2770. 30. A.  Varshavsky, Cell, 1991, 64,  13–​15. 31. A. Hershko and A. Ciechanover, Annu. Rev. Biochem., 1998, 67, 425–​479. 32. A.  Varshavsky, Protein Sci., 2006, 15, 647–​654. 33. A. Ciechanover, A. Orian and A. L. Schwartz, Bioessays, 2000, 22, 442–​451. 34. N. Dharmasiri, S. Dharmasiri and M. Estelle, Nature, 2005, 435, 441–​445.

 49

Structural and Biophysical Principles of Degrader Ternary Complexes

49

35. T. Ito, H. Ando, T. Suzuki, T. Ogura, K. Hotta, Y. Imamura, Y. Yamaguchi and H. Handa, Science, 2010, 327, 1345–​1350. 36. J. H. Kentel and S. F. Roberts, US. Pat. 6,306,663, 1999. 37. K. M. Sakamoto, K. B. Kim, A. Kumagai, F. Mercurio, C. M. Crews and R. J. Deshaies, Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 8554–​8559. 38. G. Wu, G. Xu, B. A. Schulman, P. D. Jeffrey, J. W. Harper and N. P. Pavletich, Mol. Cell, 2003, 11, 1445–​1456. 39. J. S. Schneekloth, F. N. Fonseca, M. Koldobskiy, A. Mandal, R. Deshaies, K. Sakamoto and C. M. Crews, J. Am. Chem. Soc., 2004, 126, 3748–​3754. 40. J.-​H. Min, H. Yang, M. Ivan, F. Gertler, W. G. Kaelin and N. P. Pavletich, Science, 2002, 296, 1886–​1889. 41. W.-​C. Hon, M. I. Wilson, K. Harlos, T. D. W. Claridge, C. J. Schofield, C. W. Pugh, P. H. Maxwell, P. J. Ratcliffe, D. I. Stuart and E. Y. Jones, Nature, 2002, 417, 975–​978. 42. K. M. Sakamoto, K. B. Kim, R. Verma, A. Ransick, B. Stein, C. M. Crews and R. J. Deshaies, Mol. Cell. Proteomics, 2003, 2, 1350–​1358. 43. A. Rodriguez-​ Gonzalez, K. Cyrus, M. Salcius, K. Kim, C. M. Crews, R. J. Deshaies and K. M. Sakamoto, Oncogene, 2008, 27, 7201–​7211. 44. A.  Ciulli, Methods Mol. Biol., 2013, 1008, 357–​388. 45. J.-​ P. Renaud, C.-​ W. Chung, U. H. Danielson, U. Egner, M. Hennig, R. E. Hubbard and H. Nar, Nat. Rev. Drug Discovery, 2016, 15, 679–​698. 46. D. Puppala, H. Lee, B. K. Kyung and H. I. Swanson, Mol. Pharmacol., 2008, 73, 1064–​1071. 47. K. Cyrus, M. Wehenkel, E.-​Y. Choi, H. Lee, H. Swanson and K.-​B. Kim, ChemMedChem, 2010, 5, 979–​985. 48. D. L. Buckley, I. Van Molle, P. C. Gareiss, H. S. Tae, J. Michel, D. J. Noblin, W. L. Jorgensen, A. Ciulli and C. M. Crews, J. Am. Chem. Soc., 2012, 134, 4465–​4468. 49. I. Van Molle, A. Thomann, D. L. Buckley, E. C. So, S. Lang, C. M. Crews and A. Ciulli, Chem. Biol., 2012, 19, 1300–​1312. 50. D. L. Buckley, J. L. Gustafson, I. Van Molle, A. G. Roth, H. S. Tae, P. C. Gareiss, W. L. Jorgensen, A. Ciulli and C. M. Crews, Angew. Chem., Int. Ed., 2012, 51, 11463–​11467. 51. C. Galdeano, M. S. Gadd, P. Soares, S. Scaffidi, I. Van Molle, I. Birced, S. Hewitt, D. M. Dias and A. Ciulli, J. Med. Chem., 2014, 57, 8657–​8663. 52. J. Frost, C. Galdeano, P. Soares, M. S. Gadd, K. M. Grzes, L. Ellis, O. Epemolu, S. Shimamura, M. Bantscheff, P. Grandi, K. D. Read, D. A. Cantrell, S. Rocha and A. Ciulli, Nat. Commun., 2016, 7, 13312. 53. P. Soares, M. S. Gadd, J. Frost, C. Galdeano, L. Ellis, O. Epemolu, S. Rocha, K. D. Read and A. Ciulli, J. Med. Chem., 2018, 61, 599–​618. 54. A. K. Gandhi, J. Kang, C. G. Havens, T. Conklin, Y. Ning, L. Wu, T. Ito, H. Ando, M. F. Waldman, A. Thakurta, A. Klippel, H. Handa, T. O. Daniel, P. H. Schafer and R. Chopra, Br. J. Haematol., 2014, 164, 811–​821. 55. P. P. Chamberlain, A. Lopez-​Girona, K. Miller, G. Carmel, B. Pagarigan, B. Chie-​Leon, E. Rychak, L. G. Corral, Y. J. Ren, M. Wang, M. Riley,

50

50

Chapter 2

S. L. Delker, T. Ito, H. Ando, T. Mori, Y. Hirano, H. Handa, T. Hakoshima, T. O. Daniel and B. E. Cathers, Nat. Struct. Mol. Biol., 2014, 21, 803–​809. 56. E. S. Fischer, K. Böhm, J. R. Lydeard, H. Yang, M. B. Stadler, S. Cavadini, J. Nagel, F. Serluca, V. Acker, G. M. Lingaraju, R. B. Tichkule, M. Schebesta, W. C. Forrester, M. Schirle, U. Hassiepen, J. Ottl, M. Hild, R. E. J. Beckwith, J. W. Harper, J. L. Jenkins and N. H. Thomä, Nature, 2014, 512,  49–​53. 57. D. P. Bondeson, A. Mares, I. E. D. Smith, E. Ko, S. Campos, A. H. Miah, K. E. Mulholland, N. Routly, D. L. Buckley, J. L. Gustafson, N. Zinn, P. Grandi, S. Shimamura, G. Bergamini, M. Faelth-​Savitski, M. Bantscheff, C. Cox, D. A. Gordon, R. R. Willard, J. J. Flanagan, L. N. Casillas, B. J. Votta, W. den Besten, K. Famm, L. Kruidenier, P. S. Carter, J. D. Harling, I. Churcher and C. M. Crews, Nat. Chem. Biol., 2015, 11, 611–​617. 58. J. Lu, Y. Qian, M. Altieri, H. Dong, J. Wang, K. Raina, J. Hines, J. D. Winkler, A. P. Crew, K. Coleman and C. M. Crews, Chem. Biol., 2015, 22, 755–​763. 59. G. E. Winter, D. L. Buckley, J. Paulk, J. M. Roberts, A. Souza, S. Dhe-​ Paganon and J. E. Bradner, Science, 2015, 348, 1376–​1381. 60. H. Tovell, A. Testa, H. Zhou, N. Shapiro, C. Crafter, A. Ciulli and D. R. Alessi, ACS Chem. Biol., 2019, 14, 2024–​2034. 61. I. Collins, H. Wang, J. J. Caldwell and R. Chopra, Biochem. J., 2017, 474, 1127–​1147. 62. I.  Churcher, J. Med. Chem., 2018, 61, 444–​452. 63. S. Gu, D. Cui, X. Chen, X. Xiong and Y. Zhao, Bioessays, 2018, 40, e1700247. 64. J. Qi and G. Zhang, Future Med. Chem., 2019, 11, 723–​741. 65. M. Pettersson and C. M. Crews, Drug Discovery Today: Technol., 2019, 31,  15–​27. 66. R. Verma, D. Mohl and R. J. Deshaies, Mol. Cell, 2020, 77, 446–​460. 67. D. L. Buckley, K. Raina, N. Darricarrere, J. Hines, J. L. Gustafson, I. E. Smith, A. H. Miah, J. D. Harling and C. M. Crews, ACS Chem. Biol., 2015, 10, 1831–​1837. 68. H. Tovell, A. Testa, C. Maniaci, H. Zhou, A. R. Prescott, T. Macartney, A. Ciulli and D. R. Alessi, ACS Chem. Biol., 2019, 14, 882–​892. 69. C. P. Tinworth, H. Lithgow, L. Dittus, Z. I. Bassi, S. E. Hughes, M. Muelbaier, H. Dai, I. E. D. Smith, W. J. Kerr, G. A. Burley, M. Bantscheff and J. D. Harling, ACS Chem. Biol., 2019, 14, 342–​347. 70. X. Zhang, V. M. Crowley, T. G. Wucherpfennig, M. M. Dix and B. F. Cravatt, Nat. Chem. Biol., 2019, 15, 737–​746. 71. C. C. Ward, J. I. Kleinman, S. M. Brittain, P. S. Lee, C. Y. S. Chung, K. Kim, Y. Petri, J. R. Thomas, J. A. Tallarico, J. M. McKenna, M. Schirle and D. K. Nomura, ACS Chem. Biol., 2019, 14, 2430–​2440. 72. G. Xue, J. Chen, L. Liu, D. Zhou, Y. Zuo, T. Fu and Z. Pan, Chem. Commun. (Cambridge, U. K.), 2020, 56, 1521–​1524. 73. W.-​H. Guo, X. Qi, Y. Liu, C.-​I. Chung, F. Bai, X. Lin, L. Wang, J. Chen, K. J. Nomie, F. Li, M. C. Wang, X. Shu, J. N. Onuchic, J. A. Woyach, M. L. Wang and J. Wang, bioRxiv, 2019, Preprint. https://​doi.org/​10.1101/​ 2019.12.30.873588

 51

Structural and Biophysical Principles of Degrader Ternary Complexes

51

74. R. Gabizon, A. Shraga, P. Gehrtz, E. Livnah, Y. Shorer, N. Gurwicz, L. Avram, T. Unger, H. Aharoni, S. Albeck, A. Brandis, Z. Shulman, B.-Z. Katz, Y. Herishanu and N. London, J. Am. Chem. Soc., 2020, 142, 11734–11742. 75. E. T. Mack, R. Perez-​Castillejos, Z. Suo and G. M. Whitesides, Anal. Chem., 2008, 80, 5550–​5555. 76. E. F. Douglass, C. J. Miller, G. Sparer, H. Shapiro and D. A. Spiegel, J. Am. Chem. Soc., 2013, 135, 6092–​6099. 77. M. L. Drummond and C. I. Williams, J. Chem. Inf. Model., 2019, 59, 1634–​1644. 78. A. F. Hohmann and C. R. Vakoc, Trends Genet., 2014, 30, 356–​363. 79. C. Kadoch and G. R. Crabtree, Sci. Adv., 2015, 1, e1500447–​e1500447. 80. R. St Pierre and C. Kadoch, Curr. Opin. Genet. Dev., 2017, 42,  56–​67. 81. E. R. McDonald, A. de Weck, M. R. Schlabach, E. Billy, K. J. Mavrakis, G. R. Hoffman, D. Belur, D. Castelletti, E. Frias, K. Gampa, J. Golji, I. Kao, L. Li, P. Megel, T. A. Perkins, N. Ramadan, D. A. Ruddy, S. J. Silver, S. Sovath, M. Stump, O. Weber, R. Widmer, J. Yu, K. Yu, Y. Yue, D. Abramowski, E. Ackley, R. Barrett, J. Berger, J. L. Bernard, R. Billig, S. M. Brachmann, F. Buxton, R. Caothien, J. X. Caushi, F. S. Chung, M. Cortés-​ Cros, R. S. deBeaumont, C. Delaunay, A. Desplat, W. Duong, D. A. Dwoske, R. S. Eldridge, A. Farsidjani, F. Feng, J. Feng, D. Flemming, W. Forrester, G. G. Galli, Z. Gao, F. Gauter, V. Gibaja, K. Haas, M. Hattenberger, T. Hood, K. E. Hurov, Z. Jagani, M. Jenal, J. A. Johnson, M. D. Jones, A. Kapoor, J. Korn, J. Liu, Q. Liu, S. Liu, Y. Liu, A. T. Loo, K. J. Macchi, T. Martin, G. McAllister, A. Meyer, S. Mollé, R. A. Pagliarini, T. Phadke, B. Repko, T. Schouwey, F. Shanahan, Q. Shen, C. Stamm, C. Stephan, V. M. Stucke, R. Tiedt, M. Varadarajan, K. Venkatesan, A. C. Vitari, M. Wallroth, J. Weiler, J. Zhang, C. Mickanin, V. E. Myer, J. A. Porter, A. Lai, H. Bitter, E. Lees, N. Keen, A. Kauffmann, F. Stegmeier, F. Hofmann, T. Schmelzle and W. R. Sellers, Cell, 2017, 170, 577–​592.e10. 82. O. Fedorov, J. Castex, C. Tallant, D. R. Owen, S. Martin, M. Aldeghi, O. Monteiro, P. Filippakopoulos, S. Picaud, J. D. Trzupek, B. S. Gerstenberger, C. Bountra, D. Willmann, C. Wells, M. Philpott, C. Rogers, P. C. Biggin, P. E. Brennan, M. E. Bunnage, R. Schüle, T. Günther, S. Knapp and S. Müller, Sci. Adv., 2015, 1, e1500723. 83. C. L. Sutherell, C. Tallant, O. P. Monteiro, C. Yapp, J. E. Fuchs, O. Fedorov, P. Siejka, S. Müller, S. Knapp, J. D. Brenton, P. E. Brennan and S. V. Ley, J. Med. Chem., 2016, 59, 5095–​5101. 84. B. Vangamudi, T. A. Paul, P. K. Shah, M. Kost-​Alimova, L. Nottebaum, X. Shi, Y. Zhan, E. Leo, H. S. Mahadeshwar, A. Protopopov, A. Futreal, T. N. Tieu, M. Peoples, T. P. Heffernan, J. R. Marszalek, C. Toniatti, A. Petrocchi, D. Verhelle, D. R. Owen, G. Draetta, P. Jones, W. S. Palmer, S. Sharma and J. N. Andersen, Cancer Res., 2015, 75, 3865–​3878. 85. B. K. Albrecht, et al., WIPO Pat. WO2016138114, 2016. 86. P. P. Chamberlain and B. E. Cathers, Drug Discovery Today: Technol., 2019, 31,  29–​34.

52

52

Chapter 2

87. P. P. Chamberlain, L. A. D’Agostino, J. M. Ellis, J. D. Hansen, M. E. Matyskiela, J. J. McDonald, J. R. Riggs and L. G. Hamann, ACS Med. Chem. Lett., 2019, 10, 1592–​1602. 88. R. P. Nowak, S. L. DeAngelo, D. Buckley, Z. He, K. A. Donovan, J. An, N. Safaee, M. P. Jedrychowski, C. M. Ponthier, M. Ishoey, T. Zhang, J. D. Mancias, N. S. Gray, J. E. Bradner and E. S. Fischer, Nat. Chem. Biol., 2018, 14, 706–​714. 89. W.  Lenz, Teratology, 1988, 38, 203–​215. 90. S. Singhal, J. Mehta, R. Desikan, D. Ayers, P. Roberson, P. Eddlemon, N. Munshi, E. Anaissie, C. Wilson, M. Dhodapkar, J. Zeldis, B. Barlogie, D. Siegel and J. Crowley, N. Engl. J. Med., 1999, 341, 1565–​1571. 91. S. K. Teo, K. E. Resztak, M. A. Scheffler, K. A. Kook, J. B. Zeldis, D. I. Stirling and S. D. Thomas, Microbes Infect., 2002, 4, 1193–​1202. 92. A. Lopez-​Girona, D. Mendy, T. Ito, K. Miller, A. K. Gandhi, J. Kang, S. Karasawa, G. Carmel, P. Jackson, M. Abbasian, A. Mahmoudi, B. Cathers, E. Rychak, S. Gaidarova, R. Chen, P. H. Schafer, H. Handa, T. O. Daniel, J. F. Evans and R. Chopra, Leukemia, 2012, 26, 2326–​2335. 93. S. Angers, T. Li, X. Yi, M. J. MacCoss, R. T. Moon and N. Zheng, Nature, 2006, 443, 590–​593. 94. J. Lee and P. Zhou, Mol. Cell, 2007, 26, 775–​780. 95. J. Krönke, N. D. Udeshi, A. Narla, P. Grauman, S. N. Hurst, M. McConkey, T. Svinkina, D. Heckl, E. Comer, X. Li, C. Ciarlo, E. Hartman, N. Munshi, M. Schenone, S. L. Schreiber, S. A. Carr and B. L. Ebert, Science, 2014, 343, 301–​305. 96. G. Lu, R. E. Middleton, H. Sun, M. Naniong, C. J. Ott, C. S. Mitsiades, K.-​ K. Wong, J. E. Bradner and W. G. Kaelin, Science, 2014, 343, 305–​309. 97. X. Tan, L. I. A. Calderon-​Villalobos, M. Sharon, C. Zheng, C. V. Robinson, M. Estelle and N. Zheng, Nature, 2007, 446, 640–​645. 98. J. Krönke, E. C. Fink, P. W. Hollenbach, K. J. MacBeth, S. N. Hurst, N. D. Udeshi, P. P. Chamberlain, D. R. Mani, H. W. Man, A. K. Gandhi, T. Svinkina, R. K. Schneider, M. McConkey, M. Järås, E. Griffiths, M. Wetzler, L. Bullinger, B. E. Cathers, S. A. Carr, R. Chopra and B. L. Ebert, Nature, 2015, 523, 183–​188. 99. M. E. Matyskiela, G. Lu, T. Ito, B. Pagarigan, C.-​ C. Lu, K. Miller, W. Fang, N.-​ Y. Wang, D. Nguyen, J. Houston, G. Carmel, T. Tran, M. Riley, L. Nosaka, G. C. Lander, S. Gaidarova, S. Xu, A. L. Ruchelman, H. Handa, J. Carmichael, T. O. Daniel, B. E. Cathers, A. Lopez-​Girona and P. P. Chamberlain, Nature, 2016, 535, 252–​257. 100. J. D. Hansen, K. Condroski, M. Correa, G. Muller, H. W. Man, A. Ruchelman, W. Zhang, F. Vocanson, T. Crea, W. Liu, G. Lu, F. Baculi, L. LeBrun, A. Mahmoudi, G. Carmel, M. Hickman and C.-​ C. Lu, J. Med. Chem., 2018, 61, 492–​503. 101. G. Petzold, E. S. Fischer and N. H. Thomä, Nature, 2016, 532, 127–​130. 102. Q. L. Sievers, G. Petzold, R. D. Bunker, A. Renneville, M. Słabicki, B. J. Liddicoat, W. Abdulrahman, T. Mikkelsen, B. L. Ebert and N. H. Thomä, Science, 2018, 362, eaat0572.

 53

Structural and Biophysical Principles of Degrader Ternary Complexes

53

103. T. Uehara, Y. Minoshima, K. Sagane, N. H. Sugi, K. O. Mitsuhashi, N. Yamamoto, H. Kamiyama, K. Takahashi, Y. Kotake, M. Uesugi, A. Yokoi, A. Inoue, T. Yoshida, M. Mabuchi, A. Tanaka and T. Owa, Nat. Chem. Biol., 2017, 13, 675–​680. 104. T. Han, M. Goralski, N. Gaskill, E. Capota, J. Kim, T. C. Ting, Y. Xie, N. S. Williams and D. Nijhawan, Science, 2017, 356, eaal3755. 105. X. Du, O. A. Volkov, R. M. Czerwinski, H. Tan, C. Huerta, E. R. Morton, J. P. Rizzi, P. M. Wehn, R. Xu, D. Nijhawan and E. M. Wallace, Structure, 2019, 27, 1625–​1633.e3. 106. T. B. Faust, H. Yoon, R. P. Nowak, K. A. Donovan, Z. Li, Q. Cai, N. A. Eleuteri, T. Zhang, N. S. Gray and E. S. Fischer, Nat. Chem. Biol., 2020, 16,  7–​14. 107. D. E. Bussiere, L. Xie, H. Srinivas, W. Shu, A. Burke, C. Be, J. Zhao, A. Godbole, D. King, R. G. Karki, V. Hornak, F. Xu, J. Cobb, N. Carte, A. O. Frank, A. Frommlet, P. Graff, M. Knapp, A. Fazal, B. Okram, S. Jiang, P.-​Y. Michellys, R. Beckwith, H. Voshol, C. Wiesmann, J. M. Solomon and J. Paulk, Nat. Chem. Biol., 2020, 16,  15–​23. 108. E. Krissinel and K. Henrick, J. Mol. Biol., 2007, 372, 774–​797. 109. B. E. Smith, S. L. Wang, S. Jaime-​ Figueroa, A. Harbin, J. Wang, B. D. Hamman and C. M. Crews, Nat. Commun., 2019, 10, 131. 110. J. W. Papatzimas, E. Gorobets, R. Maity, M. I. Muniyat, J. L. Maccallum, P. Neri, N. J. Bahlis and D. J. Derksen. J. Med. Chem., 2019, 62, 5522–​5540. 111. J. N. Spradlin, X. Hu, C. C. Ward, S. M. Brittain, M. D. Jones, L. Ou, M. To, A. Proudfoot, E. Ornelas, M. Woldegiorgis, J. A. Olzmann, D. E. Bussiere, J. R. Thomas, J. A. Tallarico, J. M. McKenna, M. Schirle, T. J. Maimone and D. K. Nomura, Nat. Chem. Biol., 2019, 15, 747–​755. 112. K. R. Simonetta, J. Taygerly, K. Boyle, S. E. Basham, C. Padovani, Y. Lou, T. J. Cummins, S. L. Yung, S. K. von Soly, F. Kayser, J. Kuriyan, M. Rape, M. Cardozo, M. A. Gallop, N. F. Bence, P. A. Barsanti and A. Saha, Nat. Commun., 2019, 10, 1402. 113. A. Yasgar, A. Jadhav, A. Simeonov and N. P. Coussens, Methods Mol. Biol., 2016, 1439,  77–​98. 114. G. E. Winter, A. Mayer, D. L. Buckley, M. A. Erb, J. E. Roderick, S. Vittori, J. M. Reyes, J. di Iulio, A. Souza, C. J. Ott, J. M. Roberts, R. Zeid, T. G. Scott, J. Paulk, K. Lachance, C. M. Olson, S. Dastjerdi, S. Bauer, C. Y. Lin, N. S. Gray, M. A. Kelliher, L. S. Churchman and J. E. Bradner, Mol. Cell, 2017, 67, 5–​18.e19. 115. R. P. Wurz, K. Dellamaggiore, H. Dou, N. Javier, M.-​C. Lo, J. D. McCarter, D. Mohl, C. Sastri, J. R. Lipford and V. J. Cee, J. Med. Chem., 2018, 61, 453–​461. 116. V. Zoppi, S. J. Hughes, C. Maniaci, A. Testa, T. Gmaschitz, C. Wieshofer, M. Koegl, K. M. Riching, D. L. Daniels, A. Spallarossa and A. Ciulli, J. Med. Chem., 2019, 62, 699–​726. 117. S. Khan, X. Zhang, D. Lv, Q. Zhang, Y. He, P. Zhang, X. Liu, D. Thummuri, Y. Yuan, J. S. Wiegand, J. Pei, W. Zhang, A. Sharma, C. R. McCurdy, V. M. Kuruvilla, N. Baran, A. A. Ferrando, Y.-​M. Kim, A. Rogojina,

54

54

Chapter 2

P. J. Houghton, G. Huang, R. Hromas, M. Konopleva, G. Zheng and D. Zhou, Nat. Med., 2019, 25, 1938–​1947. 118. A. Zorba, C. Nguyen, Y. Xu, J. Starr, K. Borzilleri, J. Smith, H. Zhu, K. A. Farley, W. D. Ding, J. Schiemer, X. Feng, J. S. Chang, D. P. Uccello, J. A. Young, C. N. Garcia-​ Irrizary, L. Czabaniuk, B. Schuff, R. Oliver, J. Montgomery, M. M. Hayward, J. Coe, J. Chen, M. Niosi, S. Luthra, J. C. Shah, A. El-​Kattan, X. Qiu, G. M. West, M. C. Noe, V. Shanmugasundaram, A. M. Gilbert, M. F. Brown and M. F. Calabrese, Proc. Natl. Acad. Sci. U. S. A., 2018, 115, E7285–​E7292. 119. T. C. Ting, M. Goralski, K. Klein, B. Wang, J. Kim, Y. Xie and D. Nijhawan, Cell Rep., 2019, 29, 1499–​1510.e6. 120. J. H.-​R. Hsu, T. Rasmusson, J. Robinson, F. Pachl, J. Read, S. Kawatkar, D. H. O’ Donovan, S. Bagal, E. Code, P. Rawlins, A. Argyrou, R. Tomlinson, N. Gao, X. Zhu, E. Chiarparin, K. Jacques, M. Shen, H. Woods, E. Bednarski, D. M. Wilson, L. Drew, M. P. Castaldi, S. Fawell and A. Bloecher, Cell Chem. Biol., 2020, 27, 41–​46.e17. 121. T. H. Pillow, P. Adhikari, R. A. Blake, J. Chen, G. Del Rosario, G. Deshmukh, I. Figueroa, K. E. Gascoigne, A. V. Kamath, S. Kaufman, T. Kleinheinz, K. R. Kozak, B. Latifi, D. D. Leipold, C. Sing Li, R. Li, M. M. Mulvihill, A. O’Donohue, R. K. Rowntree, J. D. Sadowsky, J. Wai, X. Wang, C. Wu, Z. Xu, H. Yao, S.-​F. Yu, D. Zhang, R. Zang, H. Zhang, H. Zhou, X. Zhu and P. S. Dragovich, ChemMedChem, 2020, 15,  17–​25. 122. R. Beveridge, D. Kessler, K. Rumpel, P. Ettmayer, A. Meinhart and T. Clausen, ACS Cent. Sci., 2020, DOI: 10.1021/acscentsci.0c00049 123. M. E. Matyskiela, W. Zhang, H. W. Man, G. Muller, G. Khambatta, F. Baculi, M. Hickman, L. LeBrun, B. Pagarigan, G. Carmel, C.-​C. Lu, G. Lu, M. Riley, Y. Satoh, P. Schafer, T. O. Daniel, J. Carmichael, B. E. Cathers and P. P. Chamberlain, J. Med. Chem., 2018, 61, 535–​542. 124. K.-​H. Chan, M. Zengerle, A. Testa and A. Ciulli, J. Med. Chem., 2018, 61, 504–​513. 125. C.-​I. Chung, Q. Zhang and X. Shu, Anal. Chem., 2018, 90, 14287–​14293. 126. Q. Zhang, H. Huang, L. Zhang, R. Wu, C.-​I. Chung, S.-​Q. Zhang, J. Torra, A. Schepis, S. R. Coughlin, T. B. Kornberg and X. Shu, Mol. Cell, 2018, 69, 334–​346.e4. 127. M. K. Schwinn, T. Machleidt, K. Zimmerman, C. T. Eggers, A. S. Dixon, R. Hurst, M. P. Hall, L. P. Encell, B. F. Binkowski and K. V. Wood, ACS Chem. Biol., 2018, 13, 467–​474. 128. D. L. Daniels, K. M. Riching and M. Urh, Drug Discovery Today: Technol., 2019, 31,  61–​68. 129. M. Brand, B. Jiang, S. Bauer, K. A. Donovan, Y. Liang, E. S. Wang, R. P. Nowak, J. C. Yuan, T. Zhang, N. Kwiatkowski, A. C. Müller, E. S. Fischer, N. S. Gray and G. E. Winter, Cell Chem. Biol., 2019, 26, 300–​ 306.e9.

 55

CHAPTER 3

Immediate and Selective Control of Protein Abundance Using the dTAG System BEHNAM NABETa,b* AND NATHANAEL S. GRAYa,b* a

Department of Cancer Biology, Dana-​ Farber Cancer Institute, Boston, Massachusetts, USA b Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts, USA *Corresponding author. Email: behnam_​[email protected], nathanael_​ ­[email protected]

3.1 The Potential and Limitations of Targeted Protein Degradation Targeted protein degradation (TPD) is a rapidly growing area of research predicated on using pharmacological agents to reduce protein abundance. TPD is highly complementary to well-​established genetic knockdown or knockout strategies, as it achieves rapid and selective protein-​level loss. Small-​ molecule degraders, including hetero-​bifunctional degraders (often referred to as PROteolysis-​ TArgeting Chimeras, or PROTACs)1 and non-​ chimeric “molecular glue” compounds,2,3 facilitate tunable control of protein abundance. Both types of degraders exert their effects by recruiting an E3 ubiquitin ligase to a target protein, resulting in its polyubiquitination and subsequent Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

55

56

56

Chapter 3 1–​4

proteasomal degradation. The immediacy, depth and selectivity of target protein loss combined with the sub-​stoichiometric catalytic-​like activity of degraders are important advantages over genetic loss-​of-​function approaches and small-​molecule inhibitors.4–​7 These features of degraders provide a powerful opportunity to study protein function at the post-​translational level. This is of particular importance when investigating essential proteins in which immediate time-​dependent evaluation is necessary. Hetero-​bifunctional degraders consist of a target-​binding warhead bridged to an E3 ligase-​binding moiety via a chemical linker. These bivalent compounds simultaneously bind to both the target protein and the E3 ubiquitin ligase to induce formation of a ternary complex. The downstream functional consequence of productive ternary complex formation is target protein polyubiquitination and proteasomal degradation of the target protein.1,4 The scope of diverse proteins degraded using hetero-​bifunctional small-​molecule degraders are summarized in recent excellent reviews.8–​11 However, the lack of small-​molecule ligands with sufficient selectivity and affinity for the majority of proteins in the proteome12 significantly restricts the broad applicability of this strategy. To overcome this limitation and to complement development of degraders of endogenous proteins, we and others developed generalizable chemical–​genetic platforms for rapid and selective protein removal. These versatile approaches enable functional evaluation of target protein loss for pre-​clinical target validation and biological investigation. In this chapter, we will focus on a chemical–​genetic strategy known as the degradation tag (dTAG) system,13 which has been employed to validate and invalidate drug targets and to elucidate the biological functions of diverse target proteins (Figure 3.1).

Figure 3.1  The dTAG system for target-​ specific protein degradation. Schematic depiction of the first-​generation dTAG system in which FKBP12F36V-selective dTAG molecules induce ternary complex formation between an FKBP12F36V-​ tagged protein and the CRBN E3 ligase complex. Rapid ubiquitination of an FKBP12F36V-​tagged protein leads to target-​specific proteasome-​mediated degradation. Reproduced from Ref. 13 with permission from Springer Nature, Copyright 2018.

 57

Immediate and Selective Control of Protein Abundance Using the dTAG System

57

3.2 Chemical–​Genetic Degradation Approaches Most drug discovery and development efforts in the post-​genomic era have been focused on target-​driven approaches.14,15 Genetic strategies, including RNA interference and CRISPR/​Cas9-​based methods, have been employed for pre-​clinical target discovery and validation through mechanistic and phenotypic evaluation upon target gene transcript loss. However, these approaches are limited by the temporal delay between target loss and functional assessment, which makes it challenging to study essential dependencies that require rapid evaluation.16 Furthermore, the potential off-​target effects of these genetic approaches and the irreversibility of CRISPR/​Cas9-​mediated gene editing are of concern.17–​19 Pharmacologic strategies such as small-​ molecule covalent and reversible inhibitors can acutely disrupt the activity of a target protein. However, the development of selective chemical matter remains non-​trivial and requires an extensive medicinal chemistry and validation campaign.14,15 In addition, small-​molecule inhibitors commonly only disrupt the activity of a single domain of a target, rather than perturb entire protein function. As such, small-​molecule inhibitors of the enzymatic activity of proteins that possess additional structural and scaffolding roles suffer from incomplete effects and can only parse out a subset of the targeted protein’s activity. Furthermore, small molecules are rarely exquisitely selective and therefore often exhibit off-​target liabilities.20–​22 To address these shortcomings, we and others developed hybrid chemical–​genetic technologies to directly control protein abundance through TPD. Tag-​based degradation systems require fusion of a target protein with a degron tag that is selectively responsive to a peptide or small molecule. Ligand binding to the degron tag induces rapid and reversible destabilization of the fusion protein. These strategies include: the dTAG system13 (tag : FKBP12F36V, ligand : dTAG molecule such as dTAG-​13); the ligand-​induced degradation (LID) approach23 (tag : FKBP12F36V with 19 amino acid cryptic degron, ligand : Shield-1); the auxin inducible degron (AID) strategy24,25 (tag  :  IAA17 or miniIAA7, ligand : small molecules of auxin class such as indole-​3-​acetic acid); HaloTag-​ directed approaches26,27 (tag : HaloTag, ligand : HaloPROTAC or HyT13); small molecule-​assisted shutoff (SMASh) strategy28 (tag : NS3pro-​NS4A, ligand : NS3 protease inhibitor); the IKZF3 degron system29 (tag : IKZF3 25mer degron, ligand: an immunomodulatory drug (IMiD) such as pomalidomide); and Trim-Away30 (an antibody-​based approach). Due to space limitations, we will focus most of our discussion on the dTAG system, with the goal of providing best-​practice guidelines and examples of use. For more detailed information on other platforms, we refer readers to the exceptional references noted above and recent reviews.16,31–​34

3.3 Development of the dTAG Platform Our efforts to develop the dTAG system were predicated on providing the broad biomedical research community with a versatile and facile tool for rapid, potent and selective protein degradation. We aimed to create a technology that would co-​opt the endogenous degradation machinery to degrade

58

58

Chapter 3

a tagged target protein, with wide utility across diverse protein targets and contexts in cell lines and mouse models. These properties would address limitations with existing tag-​based degradation approaches including the requirement of exogenous expression of the degradation machinery (AID and Trim-​Away), lack of in vivo activity (LID, AID, HaloPROTAC, SMASh and Trim-​ Away) and untested or apparent proteome-​wide off-​target activity (LID, AID, SMASh, IKZF3 degron and Trim-​Away). The dTAG system is a dual-​component platform that involves fusion of a ligandable tag to a protein-of-interest paired with a selective hetero-​bifunctional degrader that targets the protein tag (Figure  3.2A–​C). For the protein tag, we focused on FKBP12 (FKBP1A), a cytosolic prolyl isomerase,35 due to its extensive prior use in other functional tag-​based systems to induce protein dimerization,36 stabilization37 or destabilization.23 In addition, hetero-​bifunctional degraders (dFKBP-​1 and dFKBP-​2) derived from the FKBP12-​binding ligand SLF effectively depleted FKBP12,4 indicating that FKBP12 is a degradable target protein. Moreover, an FKBP12 bump-​hole strategy was previously described in which a point mutation in FKBP12 (F36V, herein referred to as FKBP12F36V) caused the formation of a compensatory cavity that was targeted by a synthetic binding ligand containing a steric bump (AP1867).38 Based on these studies, we hypothesized that hetero-​bifunctional small-​molecule degraders targeting FKBP12F36V (herein referred to as the dTAG molecule) would not induce degradation of FKBP12WT, permitting selective depletion of FKBP12F36V-​fusion proteins. The first-​generation dTAG molecule was developed to co-​opt the CUL4–​ RBX1–​DDB1–​CRBN (CRL4CRBN) E3 ubiquitin ligase by conjugating analogues of AP1867 to thalidomide, an IMiD that binds to the widely expressed substrate adaptor protein cereblon (CRBN).39 IMiDs have been extensively employed as E3 ligase-​ recruiting moieties in hetero-​ bifunctional degraders.4,6,7,9,40,41 We envisioned that recruitment of CRBN using thalidomide to FKBP12F36V-​fusions would enable target protein depletion across diverse cell lines and mouse models. Below, we will summarize optimized genetic strategies to express FKBP12F36V-​fusions and criteria for identifying selective dTAG molecules.

3.4 Genetic Methods to Express FKBP12F36V-​fusions Varied ectopic expression and knock-​in strategies have been employed to generate FKBP12F36V-​fusion cell lines. Optimized genetic strategies to fuse FKBP12F36V to the N-​or C-​terminus of a target-​of-​interest will be highlighted below.

3.4.1 Ectopic Expression of FKBP12F36V-​fusions To facilitate expression of a gene-​of-​interest as an N-​or C-​terminal FKBP12F36V-​ fusion, we developed lentiviral expression plasmids that are compatible with gateway recombination cloning technology (pLEX_​ 305-​ N-​ dTAG, Addgene #91797 and pLEX_​305-​C-​dTAG, Addgene #91798). These expression plasmids contain a PGK promoter, tandem HA-​tags for immunodetection of gene

 59

newgenrtpdf

Immediate and Selective Control of Protein Abundance Using the dTAG System 59

Figure 3.2 Chemical and genetic components of the dTAG system. (A) Chemical structure of dTAG-​13. (B) Schematic depiction of lentiviral strategy to express N-​or C-​terminal FKBP12F36V-​fusions. Lentiviral plasmids contain a PGK promoter, tandem HA tags and selectable resistance marker. (C) Schematic depiction of CRIS–​PITCh dual plasmid knock-​in strategy. One plasmid (not depicted) contains a gene-​specific and PITCh-​specific sgRNA that can cut the N-​or C-​terminus of the locus-​of-​interest and concurrently liberate a cassette containing FKBP12F36V, tandem HA tags and a selectable resistance marker from the second plasmid [depicted, pCRIS–​PITChv2–​BSD–​dTAG (BRD4)]. Microhomology-​mediated end joining leads to repair and insertion of the cassette at the locus of interest. Reproduced from Ref. 13 with permission from Springer Nature, Copyright 2018.

60

60

Chapter 3

products and a puromycin selectable marker (Figure 3.2B). A gateway entry clone containing the gene-​of-​interest with a stop codon (N-​terminal fusion) or without a stop codon (C-​terminal fusion) can be used to recombine into the expression plasmids. It is important to determine the effects of N-​and C-​terminal target tagging and to establish that the tag does not affect protein functionality and localization (see Section 3.7). We and others have employed these plasmids to degrade ENL,42 MELK,43 KRASG12V,13 MYC,13 BRD4,13 EZH2,13 HDAC1,13 PLK1,13 SMARCB144 and SHOC2,45 establishing the feasibility of degrading proteins in different cellular compartments and cell lines (see Section 3.6 for discussion of specific case studies). Control plasmids such as luciferase, LACZ or fluorescent FKBP12F36V-​fusions are valuable for evaluating degradation in an unexplored context and can serve as additional experimental controls.13 While ectopic expression of FKBP12F36V-​fusions is an accessible method for achieving tunable control of protein abundance of tagged proteins, elimination of endogenous protein expression may be necessary to accurately evaluate the functional consequences of target protein loss. Thus, genetic strategies such as CRISPR/​ Cas9-​ mediated gene editing to inactivate the endogenous locus can be employed in concert with ectopic FKBP12F36V-​fusion expression. As exemplified in the development of ENL-​FKBP12F36V; ENL–​/​–​ cell lines, employing an sgRNA that targets the intron/​exon junction at the endogenous genomic locus is an effective strategy to avoid disruption of the ectopically expressed FKBP12F36V-​fusion.42 Similarly, as shown in the development of FKBP12F36V-​MELK; MELK–​/​–​ cell lines, expression of an FKBP12F36V-​fusion with sgRNA-​resistant mutations, such as silent mutations in the PAM motif, can prevent CRISPR/​Cas9-​mediated disruption.43 Finally, isolation of single cell clones may be necessary to generate clonal lines with expression levels that match endogenous protein levels. However, depending on the effectiveness of CRISPR/​Cas9-​mediated gene editing and the ability to isolate single cell clones, pooled cellular populations can be employed for functional evaluation.

3.4.2 Knock-​in Strategies to Express FKBP12F36V-​fusions While ectopic expression is a facile approach to generate degradable FKBP12F36V-​fusions, particularly in difficult to manipulate cancer cell lines with gene amplifications, locus-​specific knock-​in approaches preserve endogenous transcriptional regulation of the gene-​of-​interest. Moreover, ectopically expressed targets controlled by constitutively active promoters can suffer from impaired depth and kinetics of degradation, as compared to targets under the control of native promoters.46 A growing number of approaches have been employed to knock-​in FKBP12F36V at specific loci including Precise Integration into Target Chromosome (PITCh)47 and homology-​directed repair (HDR).48 PITCh, which uses microhomology-​mediated end-​joining (MMEJ) and has the advantage of employing 20 base-​pair microhomology arms, was initially

 61

Immediate and Selective Control of Protein Abundance Using the dTAG System F36V

61

13

described for FKBP12 knock-​in at the BRD4 locus. This dual-​plasmid system requires a first plasmid (such as pX330A–​nBRD4/​PITCh, Addgene #91794) that contains both an sgRNA targeting the genomic locus-​of-​interest and a PITCh-​specific sgRNA that excises the FKBP12F36V-​containing cassette, and a second plasmid (such as pCRIS–​PITChv2–​BSD–​dTAG (BRD4), Addgene #91792) that contains the FKBP12F36V-​cassette flanked by the microhomology sites and PITCh sgRNA targeting sequence (Figure 3.2C). This strategy has been extended to generate CDK9 knock-​in cell lines.49 HDR knock-​in strategies are also an effective method for generating homozygous FKBP12F36V clones, as exemplified by studies tagging YY1,50 OCT451 and LMNA24 using plasmid-​based approaches or NPM1c52 using Cas9-​sgRNA ribonucleoprotein complexes (RNPs).53 Employing cassettes with multiple antibiotic13 or fluorescence50 selectable markers (such as pAW62.YY1.FKBP.knock-​in.mCherry, Addgene #104370 and pAW63.YY1. FKBP.knock-​in.BFP, Addgene #104371) can increase the efficiency of isolation of homozygous knock-​in clones. Fluorescence52 or luminescence31 markers can also be introduced in tandem with FKBP12F36V to facilitate monitoring of target protein localization and degradation kinetics. For example, generation of NPM1c–​FKBP12F36V–​GFP cell lines using Cas9-​sgRNA RNPs enabled temporal monitoring of NPM1c degradation and specific protein loss in the cytoplasm.52 Collectively, these studies highlight that diverse strategies can be employed to generate cell lines expressing FKBP12F36V-​ fusions. While this discussion focused on currently described strategies, other genetic strategies to generate FKBP12F36V-​fusions exist; investigators should determine the optimal approach for a given target and context. We expect that advances in CRISPR/​Cas9-​mediated gene editing will continue to facilitate and improve efficiency and speed of development of engineered cell lines and model systems.

3.5 Strategies Towards Identification of a Lead dTAG Molecule To identify a cell-​permeable, selective and in vivo-​compatible dTAG molecule, we screened a library of compounds using a suite of biochemical, cellular and mouse model assays. dTAG molecules were synthesized through conjugation of AP1867 or ortho-​AP1867 to thalidomide with linkers of varying composition, length and attachment points.13,42 Assays to assess biochemical protein engagement and ternary complex formation, cell penetration and target-​specific degradation, on-​target mechanism-​of-​action, global selectivity profile and effectiveness in mouse models are described below. In addition to identification of effective dTAG molecules, this assay workflow can be adapted towards evaluation and discovery of target-​specific small-​molecule degraders. We used this approach to identify dTAG-​13 as the lead CRBN-​recruiting dTAG molecule due to its immediate action, potent ligand performance, great depth of degradation, absolute selectivity, broad applicability for degrading targets in different cellular compartments and in vivo utility (Figure 3.2A).13

62

62

Chapter 3 F36V

3.5.1 Biochemical Assays for FKBP12 Binding

and E3 Ligase

In order to exert their effect, degraders not only have to bind to the target protein (for example, FKBP12F36V) and the E3 ligase (for example, CRBN) but need to effectively induce ternary complex formation (FKBP12F36V–​dTAG molecule–​ CRBN). The Amplified Luminescent Proximity Homogeneous Assay (AlphaScreen)54 is a proximity-​based biochemical assay that can be used to identify molecules that bind FKBP12F36V or CRBN. We developed AlphaScreens in which displacement of biotin-​conjugated SLF (bio-​SLF) from GST–​FKBP12WT or GST–​FKBP12F36V and biotin-​conjugated thalidomide (bio-​Thal) from HIS–​CRBN–​DDB1 indicated effective dTAG molecule binding (Figure 3.3A,B).13 For example, in the absence of dTAG molecule, bio-​SLF binds GST–​FKBP12WT or GST–​FKBP12F36V and bio-​Thal binds HIS–​CRBN–​ DDB1, bringing donor and acceptor beads into close proximity. Upon illumination, donor beads transfer a singlet-​state oxygen to the acceptor bead, leading to a measurable luminescent signal. An effective dTAG molecule binds GST–​FKBP12F36V and displaces bio-​SLF or binds HIS–​CRBN–​DDB1 and displaces bio-​Thal, leading to signal loss. Selective dTAG molecules do not bind GST-FKBP12WT and therefore do not displace bio-SLF. To evaluate effective ternary complex formation, AlphaScreens were further modified to measure hetero-​dimerization between GST–​FKBP12F36V and HIS–​CRB–​ DDB1.13 Here, effective dTAG molecules induce FKBP12F36V/​CRBN–​DDB1 complex formation, bringing donor and acceptor beads into close proximity and inducing luminescent signal upon illumination (Figure 3.3C). At high doses, the characteristic “hook effect” behavior55 may be observed upon saturation of FKBP12F36V and CRBN binding. Together, these biochemical assays led to the identification of dTAG-​13 as an FKBP12F36V-​selective, CRBN–​DDB1-​ binding ligand capable of acting as an FKBP12F36V–​CRBN hetero-​dimerizer.

3.5.2 Determining FKBP12F36V-​specific Degradation in Cells The high molecular weight of hetero-​bifunctional degraders is a liability and poses challenges for effective cell penetration and oral bioavailability.56 High-​throughput luminescence-​or fluorescence-​based screening assays can serve as an efficient strategy to identify cell-​penetrant degraders.3,13,40,46,57 Towards identification of potent, cell-​permeable dTAG molecules, we deve­ loped a dual luciferase assay by generating 293FT cells expressing nanoluciferase fusions (FKBP12WT-​Nluc or FKBP12F36V-​Nluc) and firefly luciferase (Fluc). Loss of Nluc signal serves as a readout of cellular degradation, while Fluc signal serves as a ratiometric control. These cells enabled evaluation of dose-​and time-​dependent FKBP12F36V-​specific degradation (Figure 3.4).13 Consistent with the AlphaScreen results, dTAG-​13 treatment resulted in dose-​dependent degradation of FKBP12F36V-​Nluc but not FKBP12WT-​Nluc. Thus, dTAG-​13 was a cell-​penetrant, effective FKBP12F36V-​selective degrader.

 63

Immediate and Selective Control of Protein Abundance Using the dTAG System

63

Figure 3.3  AlphaScreen assays for biochemical binding and ternary complex formation. (A) Schematic depiction of AlphaScreen assays to evaluate dTAG molecule binding to FKBP12F36V. bio-​ SLF binding to GST-​ FKBP12F36V bridges streptavidin donor beads and GSH acceptor beads, which upon illumination leads to singlet-​ state oxygen transfer and luminescent signal. dTAG molecules compete with bio-​SLF to reduce luminescent signal. (B) Schematic depiction of AlphaScreen assays to evaluate dTAG molecule binding to CRBN–​DDB1. bio-​Thal binding to HIS–​CRBN–​DDB1 bridges streptavidin donor beads and nickel chelate acceptor beads, which upon illumination leads to singlet-​state oxygen transfer and luminescent signal. dTAG molecules compete with bio-​Thal

64

64

Chapter 3

Figure 3.4 Dual luciferase assay for cell penetration and on-​target degradation. Schematic depiction of integrated dual luciferase construct containing FKBP12F36V-​Nluc and a Fluc internal control. dTAG molecules recruit the CRBN E3 ligase complex to induce degradation of FKBP12F36V-​Nluc, leading to loss of Nluc signal. Reproduced from Ref. 13 with permission from Springer Nature, Copyright 2018.

3.5.3 Requirement of E3 Ligase and Proteasome Chemical and genetic rescue experiments are necessary to confirm the requirement of E3 ligases and proteasome for the observed degradation in response to a degrader.4 We commonly perform chemical rescue experiments using proteasome inhibitors, such as carfilzomib or bortezomib, and inhibitors of E3 ligase activation such as MLN4924, as well as competition experiments with the E3 ligase binding warhead, such as lenalidomide. For active degraders, these types of compounds prevent target degradation. Complementing chemical rescue experiments, CRISPR/​ Cas9-​ mediated gene editing serves as a generalizable strategy to develop isogenic E3 ligase-​ deficient cell lines. For example, CRBN knockout (CRBN–​/​–​) cell lines have been generated to demonstrate the requirement of CRBN for degrader activity.3,4,6

Figure 3.3  (continued) to reduce luminescent signal. (C) Schematic depiction of AlphaScreen assays to evaluate ternary complex formation between FKBP12F36V–​dTAG molecule–​CRBN. dTAG molecules induce hetero-​dimerization between GST-​FKBP12F36V and HIS–​CRBN–​DDB1, bridging GSH donor beads and nickel chelate acceptor beads, which upon illumination leads to singlet-​ state oxygen transfer and luminescent signal. Reproduced from Ref. 13 with permission from Springer Nature, Copyright 2018.

 65

Immediate and Selective Control of Protein Abundance Using the dTAG System

65

To confirm the mechanism of action of dTAG molecules in the dual luciferase assay noted above, chemical and genetic rescue experiments were employed.13 Pre-​treatment of 293FT FKBP12F36V-​Nluc cells with carfilzomib, MLN4924 or lenalidomide prevented dTAG-​ mediated FKBP12F36V-​Nluc degradation, confirming the requirement of the proteasome, activated E3 ligases and CRBN for the observed degradation. To further demonstrate the requirement of CRBN in response to dTAG molecule treatment, FKBP12F36V-​Nluc was expressed in 293FT CRBN–​/​–​ cells. Genetic deletion of CRBN abrogated the effective degradation of Nluc observed upon dTAG molecule treatment, confirming the requirement of CRBN for the observed degradation.

3.5.4 Assessment of dTAG Molecule Selectivity Before degraders can be used as research tools, it is necessary to carefully evaluate their selectivity and specificity and to identify potential off-​targets.16,58 We employ strategies including reporter assays, immunoblotting, RNA-​ sequencing analysis and quantitative mass spectrometry (MS)-​based proteomics to monitor off-​target activity and confirm the selectivity of degraders. As noted above, AlphaScreen assays, dual luciferase assays and immunoblotting were first used to identify dTAG molecules with selectivity towards FKBP12F36V over FKBP12WT. Similar approaches were also employed to monitor degradation of known CRBN neo-​substrates. For example, IKZF1 and IKZF3 are plasma cell lineage transcription factors and CRBN neo-​substrates that are degraded upon IMiD treatment.2,3 Some hetero-​bifunctional degraders that incorporate IMiDs as their E3 ligase-​binding moiety retain the ability to induce CRBN-​dependent degradation of IKZF1 and IKZF3.59 As many cell lines do not express IKZF1 and IKZF3, reporter systems such as dual luciferase 293FT IKZF1-​Fluc and IKZF3-​Fluc cells, enable dose-​and time-​dependent monitoring of IKZF1 and IKZF3 levels.3 Notably, we used 293FT IKZF1-​Fluc cells to confirm that dTAG-​13 had little to no activity on IKZF1. Potential off-​targets of hetero-​bifunctional degraders can often be predicted based on the choice of target-​binding warhead and the E3 ligase-​ binding moiety. Nevertheless, quantitative MS-​based proteomics is the gold standard for evaluating degrader selectivity. The time-​point for proteomic analysis needs to be carefully selected to minimize detection of secondary, downstream consequences of target loss. Monitoring target degradation using immunoblotting can help define the most appropriate time frame for proteomics evaluation. However, evaluating both acute and prolonged time-points can aid in assessing degrader selectivity and key downstream effects, respectively. For example, we leveraged NIH/​3T3 FKBP12F36V-KRASG12V cells to evaluate the proteome-​ wide selectivity of dTAG-13 using time-​ dependent quantitative MS-​ based proteomics.13 Following one hour of dTAG-​ 13 treatment, FKBP12F36V-​KRASG12V was the only significantly degraded target protein in these cells. Following four hours of dTAG-​13 treatment, in addition to pronounced loss of FKBP12F36V-​KRASG12V, diminished

66

66

Chapter 3 G12V

60

levels of known KRAS downstream targets were also observed. dTAG molecule selectivity at acute time-points has been further confirmed by monitoring FKBP12F36V-ENL degradation upon dTAG-​ 13 treatment42 and F36V 43 FKBP12 -MELK degradation upon dTAG-​7 treatment. Transcriptional measurements such as RNA-​sequencing are also important as they report on effects on mRNA levels upon small-​molecule degrader treatment. As above, temporal evaluation of transcript level changes in parental or engineered cell lines can identify potential degrader off-​targets. For example, in parallel with evaluation of transcriptional consequences upon FKBP12F36V-KRASG12V degradation, parental NIH/​3T3 cells were treated with dTAG-​13 to monitor off-​target transcriptional effects.13 dTAG-​13 treatment did not significantly alter the expression of any mRNA transcripts in parental NIH/​ 3T3 cells, further confirming the selectivity of this hetero-​bifunctional degrader.

3.5.5  In Vivo Assessment of dTAG Molecule Activity A key limitation of tag-​based degradation approaches has been the lack of utility for in vivo studies. To identify a dTAG molecule with in vivo potential, we developed a system to evaluate the depth of degradation achieved upon dTAG molecule administration using non-​invasive bioluminescent measurements in mice.13 This system allows repeated monitoring of luciferase signal to report on degradation kinetics and recovery of luciferase-​FKBP12F36V. Tail vein injection of MV4;11 luciferase-​FKBP12F36V cells into immunodeficient mice led to bone marrow engraftment and establishment of leukemic burden. Mice were dosed three times and non-​invasive bioluminescent measurements were performed 4 h after the second and third dTAG-​13 administrations. Significant loss of luciferase signal was observed upon dTAG-​13 treatment at each time point. To evaluate reversibility of the dTAG system, mice were also evaluated 28 h after the third and final treatment, and recovery of the luciferase signal was observed. These data support the use of the dTAG system to evaluate target protein loss in mouse models and provide a platform to non-​invasively evaluate target protein degradation.

3.6 Case Studies Employing the dTAG Platform We and others have employed the dTAG platform to degrade diverse targets in both normal developmental and cancer contexts, including ENL,42 MELK,43 YY1,50 KRASG12V,13 MYC,13 BRD4,13 EZH2,13 HDAC1,13 PLK1,13 chimeric antigen receptor (CAR),61,62 SMARCB1,44 OCT4,51 LMNA,24 and SHOC2.45 Below, we describe several representative examples to illustrate utility and performance of the dTAG system as a strategy for target validation and biological investigation.

3.6.1 Target Validation Using dTAG ENL, a critical component of the super elongation complex,63 was identified as a dependency in genome-​wide CRISPR/​Cas9 screens in acute myeloid leukaemia

 67

Immediate and Selective Control of Protein Abundance Using the dTAG System

67

42

(AML). The pronounced cellular dependence on ENL made it difficult to perform mechanistic evaluation of the consequences of ENL loss via conventional genetic means. ENL-​FKBP12F36V; ENL–​/​–​ cellular systems were developed to validate the genetic loss-​of-​function screens and to perform global proteomic, transcriptional and epigenomic measurements upon acute and prolonged ENL loss. Rapid and selective ENL-​FKBP12F36V degradation led to a global collapse in the transcriptional program through suppression of RNA polymerase II activity and pronounced loss of viability, confirming that ENL is a targetable dependency in AML. This study provided the first demonstration of the utility of the dTAG system for target validation and genome-​scale mechanistic studies. This study also confirmed the proteome-​wide selectivity of dTAG-​13 and demonstrated effective degradation upon dTAG-​13 treatment in multiple cell lines. In contrast to validation of ENL as an essential protein in AML, the dTAG system has been used to invalidate target proteins in the context of drug discovery. Targeting MELK, a serine/​threonine kinase, has been of great interest due its proposed oncogenic activity in diverse cancers.64 Contrary to the prevailing literature, development of a highly selective MELK inhibitor (HTH-​01-​091) in concert with evaluation of MELK loss with RNA interference, CRISPR interference and dTAG-​mediated MELK degradation using FKBP12F36V-​MELK; MELK–​/​–​ cell lines, confirmed that MELK was dispensable for basal-​like breast cancer cell proliferation.43 This work and the work of others65 highlighted that tools previously used to study MELK, including small-​molecule inhibitors and shRNAs, likely had unrecognized off-​target activity that led to phenotypes attributed to MELK. Importantly, these studies provided a roadmap for thorough target validation through complementary applications of chemical, genetic and chemical–​genetic approaches.

3.6.2 Targeting Recalcitrant Oncoproteins Using dTAG Covalent inhibition of KRASG12C has emerged as a promising strategy to tackle this oncogenic driver.66–​68 However, other mutant forms of KRAS, a GTPase that coordinates cellular signaling cascades such as the MEK-​ERK pathway to promote proliferation and survival, currently lack targeted therapies.69,70 The dTAG system was used to model degradation of KRASG12V, a commonly occurring driver lesion in pancreatic and lung cancers,70 through development of NIH/​3T3 FKBP12F36V-​KRASG12V cell lines.13 FKBP12F36V-​KRASG12V expression led to aberrant transcriptional signaling and cell growth, confirming the functionality of the oncogenic fusion protein. Importantly, FKBP12F36V-​KRASG12V degradation rapidly reversed the aberrant phospho-​proteomic and transcriptional program, leading to pronounced antiproliferative effects. FKBP12F36V-​ KRASG12V degradation was also leveraged to confirm the mechanism-​of-​action and proteome-​wide selectivity of dTAG-​13 and illustrate that degradation may be a tractable strategy to drug this recalcitrant oncoprotein. In addition to mutant KRAS, the dTAG system has been used to evaluate the degradation of effectors in the oncogenic KRAS signaling network. Genome-​ scale CRISPR/​Cas9 screens identified SHOC2 as a key mediator of resistance

68

68

Chapter 3 45

to MEK inhibitors through reactivation of MEK-​ERK signaling. Generation of FKBP12F36V-​SHOC2; SHOC2–​/​–​ cell lines confirmed that SHOC2 degradation in combination with MEK inhibition disabled CRAF to suppress reactivation of MEK-​ERK signaling, leading to a pronounced loss of cell viability. While SHOC2 remains a challenging drug target, TPD is an attractive strategy to modulate SHOC2 activity in combination with MEK inhibition.

3.6.3 Targeting Essential Transcriptional Regulators Using dTAG The rapid degradation afforded by the dTAG system is powerful for evaluation of loss of proteins involved in transcriptional regulation. In addition to evaluation of ENL, the dTAG system was used to investigate the consequences of YY1, BRD4 and NPM1c loss using strategies to knock-​in FKBP12F36V. First, dTAG-​mediated degradation of YY1-​FKBP12F36V, a zinc finger transcription factor,71 confirmed that YY1 occupies both enhancers and promoters and is essential for enhancer–​promoter looping.50 Second, dTAG-​mediated degradation of FKBP12F36V-​BRD4 demonstrated that pan-​BET bromodomain protein (BRD2, BRD3 and BRD4) inhibition or degradation is required for maximal effects on cell proliferation and RNA polymerase II activity.13 Third, allele-​ specific dTAG-​mediated degradation of NPM1c-​FKBP12F36V, a protein that is mislocalized to the cytoplasm upon NPM1 mutation in AML,72 revealed that NPM1c coordinates HOX/​MEIS1 gene expression to control the growth and differentiation of AML cells.52

3.7 General Considerations for Employing Tag-​based Strategies When using tag-​based TPD strategies, it is important to carefully evaluate tagged protein functionality and ligand off-​target activity as well as control for the consequences of ectopic expression of the degradation machinery. First, it is critical to use functional assays to identify whether the N-​and/​or C-​terminus of a target can accommodate a tag without altering protein function or localization. Prior precedent from the literature can help identify which terminus is appropriate for tagging and ectopic expression plasmids (Section 3.4.1) can provide a rapid means to compare the consequences of N-​or C-​terminal tagging, prior to investment in knock-​in strategies. Comparison of cells expressing the target with and without the tag can also help ensure that tagging was not detrimental to protein functionality and activity. Second, small molecules, peptidic ligands or antibodies that induce protein degradation may have off-​ target activity and toxicity. For example, small-​molecule degraders may maintain the ability to recruit the E3 ligase to neosubstrates,58,59 inducing off-​target degradation and toxicity in certain cell lines. It is important to treat parental cells or cells expressing a degradable control such as tagged LACZ, fluorescence marker or luciferase in functional and phenotypic assays. These experiments

 69

Immediate and Selective Control of Protein Abundance Using the dTAG System

69

are necessary to ensure that there are no toxicities at doses employed to induce tagged fusion protein degradation. Third, if employing strategies where expression of the E3 ligase or degradation componentry is necessary, evaluation of proteome-​wide protein homeostasis should be monitored. Quantitative MS-based proteomics serves as an ideal platform to confirm both ligand specificity and consequences of degradation machinery expression.

3.8 Concluding Remarks Tag-​based degradation strategies that directly perturb protein homeostasis to achieve protein loss can facilitate target discovery and evaluation. These technologies are highly complementary to genetic approaches and we expect that their adoption will continue to rise. With advancing genetic and chemical screening technologies, we envision that tag-​based degradation systems will emerge as a first step in mechanistic dissection of the consequences of target protein loss. In addition, with the rapid advancement of degrader molecules towards clinical applications, we expect that these approaches will serve as an essential complement to degrader development campaigns to benchmark and confirm on-​target degradation. While the dTAG system and related tag-​based degradation strategies show much promise as research tools, it is important to keep in mind several current limitations of these strategies. In certain contexts and model organisms, structural features that mediate target and E3 ubiquitin ligase compatibility, E3 ligase availability, target localization, target function and target lysine accessibility may impact performance of these research tools. Advances in the field of TPD will continue to expand our understanding of the rules behind degrader activity and lead to improvements in tag-​based degradation strategies. As the dTAG platform continues to be leveraged across the biomedical community, we expect that future advances in the technology will include identification of dTAG molecules with superior in vitro and in vivo activity, improvement of genetic strategies that enhance efficiency of FKBP12F36V knock-​in, and development of transgenic model organisms to degrade target proteins in vivo. We also envision that development of orthogonal tag-​based degradation strategies will enable dual target control and creation of hetero-​ bifunctional dimerizing agents to co-​opt additional E3 ligases, enzymes and non-​proteasomal disposal pathways will enable modulation of target expression and activity through complementary means. In addition, we anticipate that the dTAG system may have clinical potential as a rheostat of CAR T-​ cell activity.61,62 Reversibly tuning CAR T-​cell activity may improve potential adverse inflammatory responses to this exciting therapeutic modality.73 Finally, we believe in an open-​source paradigm to facilitate scientific investigation and have made the dTAG platform available to the broad biomedical research community. Current users of our technology span continents as well as targets, and we are excited to see what new biological and pharmacological insights emerge from these efforts.

70

70

Chapter 3

3.9 Abbreviations AID auxin inducible degron AML acute myeloid leukaemia bio-​SLF biotin-​conjugated  SLF bio-​Thal biotin-​conjugated thalidomide CAR chimeric antigen receptor CRBN cereblon dTAG degradation tag Fluc firefly luciferase HDR homology-​directed  repair IMiD immunomodulatory drug LID ligand-​induced degradation MMEJ microhomology-​mediated end-​joining MS mass spectrometry Nluc nanoluciferase PITCh Precise Integration into Target Chromosome PROTAC PROteolysis-​TArgeting Chimera RNP ribonucleoprotein complex SMASh small molecule-​assisted shutoff TPD targeted protein degradation

3.10 Acknowledgments We thank M. Kostic, N. Kwiatkowski and E. Wang for critical reading of the manuscript and S. Nabet and members of the Gray laboratory for helpful discussions. This work was supported by : American Cancer Society Postdoctoral Fellowship PF-​17-​010-​01-​CDD (B.N.), Claudia Adams Barr Program in Innovative Basic Cancer Research Award (B.N.), Katherine L. and Steven C. Pinard Research Fund (B.N. and N.S.G.) and Hale Center for Pancreatic Cancer Research (N.S.G.).

3.10.1  Competing Financial Interests The authors claim the following competing financial interests : B.N. is an inventor on patent applications related to the dTAG system (WO/​2017/​024318, WO/​ 2017/​024319, WO/​2018/​148440 and WO/​2018/​148443). N.S.G. is a Scientific Founder, member of the Scientific Advisory Board and equity holder in C4 Therapeutics, Syros, Soltego (board member), B2S, Allorion, Gatekeeper and Petra Pharmaceuticals. The Gray laboratory receives or has received research funding from Novartis, Takeda, Astellas, Taiho, Janssen, Kinogen, Voroni, Arbella, Deerfield and Sanofi.

References 1. K. M. Sakamoto, K. B. Kim, A. Kumagai, F. Mercurio, C. M. Crews and R. J. Deshaies, Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 8554–​8559.

 71

Immediate and Selective Control of Protein Abundance Using the dTAG System

71

2. J. Kronke, N. D. Udeshi, A. Narla, P. Grauman, S. N. Hurst, M. McConkey, T. Svinkina, D. Heckl, E. Comer, X. Li, C. Ciarlo, E. Hartman, N. Munshi, M. Schenone, S. L. Schreiber, S. A. Carr and B. L. Ebert, Science, 2014, 343, 301–​305. 3. G. Lu, R. E. Middleton, H. Sun, M. Naniong, C. J. Ott, C. S. Mitsiades, K. K. Wong, J. E. Bradner and W. G. Kaelin, Jr., Science, 2014, 343, 305–​309. 4. G. E. Winter, D. L. Buckley, J. Paulk, J. M. Roberts, A. Souza, S. Dhe-​ Paganon and J. E. Bradner, Science, 2015, 348, 1376–​1381. 5. D. P. Bondeson, A. Mares, I. E. Smith, E. Ko, S. Campos, A. H. Miah, K. E. Mulholland, N. Routly, D. L. Buckley, J. L. Gustafson, N. Zinn, P. Grandi, S. Shimamura, G. Bergamini, M. Faelth-​Savitski, M. Bantscheff, C. Cox, D. A. Gordon, R. R. Willard, J. J. Flanagan, L. N. Casillas, B. J. Votta, W. den Besten, K. Famm, L. Kruidenier, P. S. Carter, J. D. Harling, I. Churcher and C. M. Crews, Nat. Chem. Biol., 2015, 11, 611–​617. 6. H. T. Huang, D. Dobrovolsky, J. Paulk, G. Yang, E. L. Weisberg, Z. M. Doctor, D. L. Buckley, J. H. Cho, E. Ko, J. Jang, K. Shi, H. G. Choi, J. D. Griffin, Y. Li, S. P. Treon, E. S. Fischer, J. E. Bradner, L. Tan and N. S. Gray, Cell Chem. Biol., 2018, 25, 88–​99 e86. 7. C. M. Olson, B. Jiang, M. A. Erb, Y. Liang, Z. M. Doctor, Z. Zhang, T. Zhang, N. Kwiatkowski, M. Boukhali, J. L. Green, W. Haas, T. Nomanbhoy, E. S. Fischer, R. A. Young, J. E. Bradner, G. E. Winter and N. S. Gray, Nat. Chem. Biol., 2018, 14, 163–​170. 8. A. C. Lai and C. M. Crews, Nat. Rev. Drug Discovery, 2017, 16, 101–​114. 9. M. Pettersson and C. M. Crews, Drug Discovery Today: Technol., 2019, 31, 15–​27. 10. I.  Churcher, J. Med. Chem., 2018, 61, 444–​452. 11. S. L. Fisher and A. J. Phillips, Curr. Opin. Chem. Biol., 2018, 44,  47–​55. 12. A.  Mullard, Nat. Rev. Drug Discovery, 2019, 18, 733–​736. 13. B. Nabet, J. M. Roberts, D. L. Buckley, J. Paulk, S. Dastjerdi, A. Yang, A. L. Leggett, M. A. Erb, M. A. Lawlor, A. Souza, T. G. Scott, S. Vittori, J. A. Perry, J. Qi, G. E. Winter, K. K. Wong, N. S. Gray and J. E. Bradner, Nat. Chem. Biol., 2018, 14, 431–​441. 14. F. M. Ferguson and N. S. Gray, Nat. Rev. Drug Discovery, 2018, 17, 353–​377. 15. T. Zhang, J. M. Hatcher, M. Teng, N. S. Gray and M. Kostic, Cell Chem. Biol., 2019. 16. C. Mayor-​Ruiz and G. E. Winter, Drug Discovery Today Technol., 2019, 31, 81–​90. 17. W. G. Kaelin, Jr., Science, 2012, 337, 421–​422. 18. A. L. Jackson, S. R. Bartz, J. Schelter, S. V. Kobayashi, J. Burchard, M. Mao, B. Li, G. Cavet and P. S. Linsley, Nat. Biotechnol., 2003, 21, 635–​637. 19. W. Putzbach, Q. Q. Gao, M. Patel, S. van Dongen, A. Haluck-​Kangas, A. A. Sarshad, E. T. Bartom, K. A. Kim, D. M. Scholtens, M. Hafner, J. C. Zhao, A. E. Murmann and M. E. Peter, eLife, 2017, 6. 20. C. H. Arrowsmith, J. E. Audia, C. Austin, J. Baell, J. Bennett, J. Blagg, C. Bountra, P. E. Brennan, P. J. Brown, M. E. Bunnage, C. Buser-​Doepner, R. M. Campbell, A. J. Carter, P. Cohen, R. A. Copeland, B. Cravatt,

72

72

Chapter 3

J. L. Dahlin, D. Dhanak, A. M. Edwards, M. Frederiksen, S. V. Frye, N. Gray, C. E. Grimshaw, D. Hepworth, T. Howe, K. V. Huber, J. Jin, S. Knapp, J. D. Kotz, R. G. Kruger, D. Lowe, M. M. Mader, B. Marsden, A. Mueller-​ Fahrnow, S. Muller, R. C. O’Hagan, J. P. Overington, D. R. Owen, S. H. Rosenberg, B. Roth, R. Ross, M. Schapira, S. L. Schreiber, B. Shoichet, M. Sundstrom, G. Superti-​ Furga, J. Taunton, L. Toledo-​ Sherman, C. Walpole, M. A. Walters, T. M. Willson, P. Workman, R. N. Young and W. J. Zuercher, Nat. Chem. Biol., 2015, 11, 536–​541. 21. W. G. Kaelin, Jr., Nat. Rev. Cancer, 2017, 17, 425–​440. 22. A. Lin, C. J. Giuliano, A. Palladino, K. M. John, C. Abramowicz, M. L. Yuan, E. L. Sausville, D. A. Lukow, L. Liu, A. R. Chait, Z. C. Galluzzo, C. Tucker and J. M. Sheltzer, Sci. Transl. Med., 2019, 11. 23. K. M. Bonger, L. C. Chen, C. W. Liu and T. J. Wandless, Nat. Chem. Biol., 2011, 7, 531–​537. 24. S. Li, X. Prasanna, V. T. Salo, I. Vattulainen and E. Ikonen, Nat. Methods, 2019, 16, 866–​869. 25. K. Nishimura, T. Fukagawa, H. Takisawa, T. Kakimoto and M. Kanemaki, Nat. Methods, 2009, 6, 917–​922. 26. D. L. Buckley, K. Raina, N. Darricarrere, J. Hines, J. L. Gustafson, I. E. Smith, A. H. Miah, J. D. Harling and C. M. Crews, ACS Chem. Biol., 2015, 10, 1831–​1837. 27. T. K. Neklesa, H. S. Tae, A. R. Schneekloth, M. J. Stulberg, T. W. Corson, T. B. Sundberg, K. Raina, S. A. Holley and C. M. Crews, Nat. Chem. Biol., 2011, 7, 538–​543. 28. H. K. Chung, C. L. Jacobs, Y. Huo, J. Yang, S. A. Krumm, R. K. Plemper, R. Y. Tsien and M. Z. Lin, Nat. Chem. Biol., 2015, 11, 713–​720. 29. V. Koduri, S. K. McBrayer, E. Liberzon, A. C. Wang, K. J. Briggs, H. Cho and W. G. Kaelin, Jr., Proc. Natl. Acad. Sci. U. S. A., 2019, 116, 2539–​2544. 30. D. Clift, W. A. McEwan, L. I. Labzin, V. Konieczny, B. Mogessie, L. C. James and M. Schuh, Cell, 2017, 171, 1692–1706. 31. D. L. Daniels, K. M. Riching and M. Urh, Drug Discovery Today: Technol., 2019, 31,  61–​68. 32. B. Z. Stanton, E. J. Chory and G. R. Crabtree, Science, 2018, 359. 33. A. Yesbolatova, Y. Tominari and M. T. Kanemaki, Drug Discovery Today: Technol., 2019, 31,  91–​98. 34. G. Vandemoortele, S. Eyckerman and K. Gevaert, Trends Biotechnol., 2019, 37, 1078–​1090. 35. M. W. Harding, A. Galat, D. E. Uehling and S. L. Schreiber, Nature, 1989, 341, 758–​760. 36. D. M. Spencer, T. J. Wandless, S. L. Schreiber and G. R. Crabtree, Science, 1993, 262, 1019–​1024. 37. L. A. Banaszynski, L. C. Chen, L. A. Maynard-​Smith, A. G. Ooi and T. J. Wandless, Cell, 2006, 126, 995–​1004. 38. T. Clackson, W. Yang, L. W. Rozamus, M. Hatada, J. F. Amara, C. T. Rollins, L. F. Stevenson, S. R. Magari, S. A. Wood, N. L. Courage, X. Lu, F. Cerasoli, Jr., M. Gilman and D. A. Holt, Proc. Natl. Acad. Sci. U. S. A., 1998, 95, 10437–​10442.

 73

Immediate and Selective Control of Protein Abundance Using the dTAG System

73

39. T. Ito, H. Ando, T. Suzuki, T. Ogura, K. Hotta, Y. Imamura, Y. Yamaguchi and H. Handa, Science, 2010, 327, 1345–​1350. 40. G. E. Winter, A. Mayer, D. L. Buckley, M. A. Erb, J. E. Roderick, S. Vittori, J. M. Reyes, J. di Iulio, A. Souza, C. J. Ott, J. M. Roberts, R. Zeid, T. G. Scott, J. Paulk, K. Lachance, C. M. Olson, S. Dastjerdi, S. Bauer, C. Y. Lin, N. S. Gray, M. A. Kelliher, L. S. Churchman and J. E. Bradner, Mol. Cell, 2017, 67, 5–​18 e19. 41. J. Lu, Y. Qian, M. Altieri, H. Dong, J. Wang, K. Raina, J. Hines, J. D. Winkler, A. P. Crew, K. Coleman and C. M. Crews, Chem. Biol., 2015, 22, 755–​763. 42. M. A. Erb, T. G. Scott, B. E. Li, H. Xie, J. Paulk, H. S. Seo, A. Souza, J. M. Roberts, S. Dastjerdi, D. L. Buckley, N. E. Sanjana, O. Shalem, B. Nabet, R. Zeid, N. K. Offei-​Addo, S. Dhe-​Paganon, F. Zhang, S. H. Orkin, G. E. Winter and J. E. Bradner, Nature, 2017, 543, 270–​274. 43. H. T. Huang, H. S. Seo, T. Zhang, Y. Wang, B. Jiang, Q. Li, D. L. Buckley, B. Nabet, J. M. Roberts, J. Paulk, S. Dastjerdi, G. E. Winter, H. McLauchlan, J. Moran, J. E. Bradner, M. J. Eck, S. Dhe-​Paganon, J. J. Zhao and N. S. Gray, eLife, 2017, 6, e26693. 44. A. M. Weissmiller, J. Wang, S. L. Lorey, G. C. Howard, E. Martinez, Q. Liu and W. P. Tansey, Nat. Commun., 2019, 10, 2014. 45. R. Sulahian, J. J. Kwon, K. H. Walsh, E. Pailler, T. L. Bosse, M. Thaker, D. Almanza, J. M. Dempster, J. Pan, F. Piccioni, N. Dumont, A. Gonzalez, J. Rennhack, B. Nabet, J. A. Bachman, A. Goodale, Y. Lee, M. Bagul, R. Liao, A. Navarro, T. L. Yuan, R. W. S. Ng, S. Raghavan, N. S. Gray, A. Tsherniak, F. Vazquez, D. E. Root, A. J. Firestone, J. Settleman, W. C. Hahn and A. J. Aguirre, Cell Rep., 2019, 29, 118–​134 e118. 46. K. M. Riching, S. Mahan, C. R. Corona, M. McDougall, J. D. Vasta, M. B. Robers, M. Urh and D. L. Daniels, ACS Chem. Biol., 2018, 13, 2758–​2770. 47. T. Sakuma, S. Nakade, Y. Sakane, K. T. Suzuki and T. Yamamoto, Nat. Protoc., 2016, 11, 118–​133. 48. F. A. Ran, P. D. Hsu, J. Wright, V. Agarwala, D. A. Scott and F. Zhang, Nat. Protoc., 2013, 8, 2281–​2308. 49. M. Brand and G. E. Winter, in Target Identification and Validation in Drug Discovery, ed. J. Moll and S. Carotta, Humana Press, New York, 2nd edn, 2019, vol. 1953, ch. 7, pp. 105–119. 50. A. S. Weintraub, C. H. Li, A. V. Zamudio, A. A. Sigova, N. M. Hannett, D. S. Day, B. J. Abraham, M. A. Cohen, B. Nabet, D. L. Buckley, Y. E. Guo, D. Hnisz, R. Jaenisch, J. E. Bradner, N. S. Gray and R. A. Young, Cell, 2017, 171, 1573–​1588 e1528. 51. A. Boija, I. A. Klein, B. R. Sabari, A. Dall’Agnese, E. L. Coffey, A. V. Zamudio, C. H. Li, K. Shrinivas, J. C. Manteiga, N. M. Hannett, B. J. Abraham, L. K. Afeyan, Y. E. Guo, J. K. Rimel, C. B. Fant, J. Schuijers, T. I. Lee, D. J. Taatjes and R. A. Young, Cell, 2018, 175, 1842–​1855 e1816. 52. L. Brunetti, M. C. Gundry, D. Sorcini, A. G. Guzman, Y. H. Huang, R. Ramabadran, I. Gionfriddo, F. Mezzasoma, F. Milano, B. Nabet, D. L. Buckley, S. M. Kornblau, C. Y. Lin, P. Sportoletti, M. P. Martelli, B. Falini and M. A. Goodell, Cancer Cell, 2018, 34, 499–​512 e499.

74

74

Chapter 3

3. S. Lin, B. T. Staahl, R. K. Alla and J. A. Doudna, eLife, 2014, 3, e04766. 5 54. J. M. Roberts and J. E. Bradner, Curr. Protoc. Chem. Biol., 2015, 7, 263–​278. 55. E. F. Douglass, Jr., C. J. Miller, G. Sparer, H. Shapiro and D. A. Spiegel, J. Am. Chem. Soc., 2013, 135, 6092–​6099. 56. G. F. Watt, P. Scott-​Stevens and L. Gaohua, Drug Discovery Today: Technol., 2019, 31,  69–​80. 57. R. P. Nowak, S. L. DeAngelo, D. Buckley, Z. He, K. A. Donovan, J. An, N. Safaee, M. P. Jedrychowski, C. M. Ponthier, M. Ishoey, T. Zhang, J. D. Mancias, N. S. Gray, J. E. Bradner and E. S. Fischer, Nat. Chem. Biol., 2018. 58. M. Ishoey, S. Chorn, N. Singh, M. G. Jaeger, M. Brand, J. Paulk, S. Bauer, M. A. Erb, K. Parapatics, A. C. Muller, K. L. Bennett, G. F. Ecker, J. E. Bradner and G. E. Winter, ACS Chem. Biol., 2018, 13, 553–​560. 59. D. Dobrovolsky, E. S. Wang, S. Morrow, C. Leahy, T. Faust, R. P. Nowak, K. A. Donovan, G. Yang, Z. Li, E. S. Fischer, S. P. Treon, D. M. Weinstock and N. S. Gray, Blood, 2019, 133, 952–​961. 60. B. Nabet, O. B. P, J. M. Reyes, K. Shieh, C. Y. Lin, C. M. Will, R. Popovic, T. Ezponda, J. E. Bradner, A. A. Golden and J. D. Licht, Cell Rep., 2015, 12, 1300–​1313. 61. J. Bradner, J. Roberts, B. Nabet, G. Winter, A. Phillips, T. Heffernan and D. Buckley, Patent WO/​2017/​024318, 2017. 62. J. Bradner, J. Roberts, B. Nabet, G. Winter, A. Phillips, T. Heffernan and D. Buckley, Patent WO/​2018/​148440, 2018. 63. T. R. Bishop, Y. Zhang and M. A. Erb, Trends Pharmacol. Sci., 2019, 40, 388–​402. 64. J. Settleman, C. L. Sawyers and T. Hunter, eLife, 2018, 7. 65. A. Lin, C. J. Giuliano, N. M. Sayles and J. M. Sheltzer, eLlife, 2017, 6. 66. M. R. Janes, J. Zhang, L. S. Li, R. Hansen, U. Peters, X. Guo, Y. Chen, A. Babbar, S. J. Firdaus, L. Darjania, J. Feng, J. H. Chen, S. Li, S. Li, Y. O. Long, C. Thach, Y. Liu, A. Zarieh, T. Ely, J. M. Kucharski, L. V. Kessler, T. Wu, K. Yu, Y. Wang, Y. Yao, X. Deng, P. P. Zarrinkar, D. Brehmer, D. Dhanak, M. V. Lorenzi, D. Hu-​ Lowe, M. P. Patricelli, P. Ren and Y. Liu, Cell, 2018, 172, 578–​589 e517. 67. S. M. Lim, K. D. Westover, S. B. Ficarro, R. A. Harrison, H. G. Choi, M. E. Pacold, M. Carrasco, J. Hunter, N. D. Kim, T. Xie, T. Sim, P. A. Janne, M. Meyerson, J. A. Marto, J. R. Engen and N. S. Gray, Angew. Chem., Int. Ed., 2014, 53, 199–​204. 68. J. M. Ostrem, U. Peters, M. L. Sos, J. A. Wells and K. M. Shokat, Nature, 2013, 503, 548–​551. 69. M. B. Ryan and R. B. Corcoran, Nat. Rev. Clin. Oncol., 2018, 15, 709–​720. 70. A. G. Stephen, D. Esposito, R. K. Bagni and F. McCormick, Cancer Cell, 2014, 25, 272–​281. 71. Y. Shi, E. Seto, L. S. Chang and T. Shenk, Cell, 1991, 67, 377–​388. 72. S. Grisendi, R. Bernardi, M. Rossi, K. Cheng, L. Khandker, K. Manova and P. P. Pandolfi, Nature, 2005, 437, 147–​153. 73. S. S. Neelapu, S. Tummala, P. Kebriaei, W. Wierda, C. Gutierrez, F. L. Locke, K. V. Komanduri, Y. Lin, N. Jain, N. Daver, J. Westin, A. M. Gulbis, M. E. Loghin, J. F. de Groot, S. Adkins, S. E. Davis, K. Rezvani, P. Hwu and E. J. Shpall, Nat. Rev. Clin. Oncol., 2018, 15,  47–​62.

 75

CHAPTER 4

Developing Pharmacokinetic/​ Pharmacodynamic Relationships With PROTACs JOHN D. HARLING,*a PAUL SCOTT-​STEVENSb AND LU GAOHUAc a

Medicinal Chemistry, Medicines Design, DMPK, In Vitro–​In Vivo Translation, c Systems Modelling & Translational Biology, Data and Computational Sciences, GlaxoSmithKline, Medicines Research Centre, Gunnels Wood Road, Stevenage, Hertfordshire SG1 2NY, United Kingdom *Corresponding author. Email: [email protected] b

4.1 Introduction While the first publication on PROTACs appeared in 2001,1 it was not until 2015 that the area reached an inflection point with the publication of two reports in close succession from the Crews2 and the Bradner laboratories.3 The Crews paper demonstrated that a recently developed binder to the E3 ligase VHL could be incorporated into PROTACs to potently degrade the nuclear receptor ERRa and the serine/​threonine kinase RIPK2, while the Bradner paper disclosed thalidomide-​based binders to the E3 ligase cereblon that delivered nanomolar protein degradation of the BET family proteins BRD2, BRD3 and BRD4 as well as the immunophilin protein, FKBP12. Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

75

76

76

Chapter 4

Significantly, both papers presented the first evidence of in vivo degradation in mice with PROTACs. The ERRa PROTAC 1 (PROTAC_​ERRa, Figure 4.1) was dosed intraperitoneally (IP) to mice with MDA-​MB-​231 xenograft tumors at 100 mg kg–​1, dosed 4 times at 8-​h intervals. At the end of the experiment ERRa levels were reduced ~40% in heart, kidney and the tumor in a clear demonstration of broad tissue distribution and in vivo efficacy. Similarly, the BET PROTAC 2 (dBET1, Figure 4.1) was dosed in a murine model of human MV4-​II acute myeloid leukemia at a daily dose of 50 mg kg–​1 IP, for 14 days. Reduced tumor progression (>70%) and tumor weight (>50%) post mortem were demonstrated, along with evidence of in vivo degradation of BRD4 in the tumor via immunohistochemistry. Associated PK data were consistent with these findings, with PROTAC 2 achieving a Cmax of 12.4 µg mL–​1, which is significantly higher than the concentration required to achieved 50% degradation of BRD4. Since these pivotal reports, further publications have demonstrated in vivo efficacy for other protein degradation targets such as BTK,4 the androgen receptor5 and MDM2.6 Recently, Sun et al.7 described an extensive package of in vivo data using PROTAC 3 (RC32, Figure 4.1) which potently degrades FKBP12 (DC50 0.27 nM in Jurkat cells) and PROTAC 4 (P13IS, Figure 4.1) which degrades BTK. Both PROTACs utilize a cereblon binder and were assembled using click chemistry. With these PROTACs, Sun et al. demonstrated in vivo protein degradation in mice, rats, Bama pigs and rhesus monkeys. Protein degradation was also demonstrated across a wide range of tissues including heart, liver, kidney, pancreas, adipose, skin and even brain when intracerebroventruclar dosing was employed. A key point emphasized in this paper was that PROTACs provide a powerful in vivo chemical knockdown approach that can readily be deployed with temporal control and which compares favorably in this regard to CRISPR/​Cas9 type approaches to protein knockdown. Collectively, these reports show that targeted in vivo degradation can be readily achieved and therefore it comes as no surprise that it has resulted in the pharma sector becoming increasingly interested in PROTACs as a new therapeutic modality. Indeed, the androgen receptor PROTAC ARV-​110 and the estrogen receptor PROTAC ARV-​471 developed by Arvinas have recently become the first PROTACs to enter clinical trials.

4.2 The Importance of PK/​PD Relationships and Additional Considerations for PROTACs 4.2.1 Building PK/​PD Relationships The development of pharmacokinetic (PK)/​pharmacodynamic (PD) relationships is a key activity in drug discovery to support the successful transition from the early drug research through to the clinic. A seminal paper in 20128 by a group at Pfizer set out their so-​called “3 Pillars of Survival,” which provided a blueprint to increase the “likelihood of candidate survival in Phase II trials and improve the chance of progression to Phase III.” Following this framework,

 77

newgenrtpdf

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

77

Figure 4.1 Structure of selected PROTACs with published in vivo efficacy.

78

78

Chapter 4

it is clear that PK/​PD studies using appropriate biomarkers will form a key part of any early clinical study and that these will be built from a sound PK/​ PD understanding developed during the pre-​clinical phase. The application of this for PROTACs will be the focus of this chapter. There have been a number of reviews providing frameworks to building PK/​PD understanding in the pre-​ clinical phase.9,10 It is widely acknowledged that this needs to be a multidisciplinary and iterative approach. Key goals include full characterization of the PK profile and the selection of an appropriate PD marker and its characterization over time to assess if the response is directly related to drug concentration or is indirect. A thorough understanding of the PD marker is also required and how it relates to target or pathway engagement and whether it is relevant to the disease setting. The PK/​PD data package will also need to encompass time course, repeat dosing and ideally multiple pre-​clinical species components.

4.2.2 PK/​PD Considerations for PROTACs The unique mechanism of action of PROTACs provides attractive opportunities for delivering impressive in vivo efficacy and highly differentiated medicines, although there are some additional factors (Figure 4.4) that need to be considered in building a PK/​PD understanding to support clinical translation.

4.2.2.1 Catalytic Mechanism of PROTACs Unlike small molecules where efficacy is usually occupancy-​driven, PROTACs promote protein degradation through the formation of a transient ternary complex. Once this event has successfully tagged the protein for recruitment to the proteasome via ubiquitination, a PROTAC molecule is free to form a new ternary complex, thus acting in a catalytic manner. Clearly this has the potential to profoundly affect PK/​PD relationships and will need to be accounted for. The first evidence to support the presumed catalytic nature of PROTACs came from the in vitro experiments reported by Bondeson et al.2 In these experiments, the E1–​E2–​E3 enzymatic cascade for VHL was reconstituted in vitro. Incubation with a radiolabeled version of the kinase RIPK2 and a RIPK2-​ VHL PROTAC revealed sub-​stoichiometric ubiquitination of RIPK2 by the PROTAC. Further evidence was published by Riching et al.,11 who described a strategy to introduce a short 11-​amino acid tag into a protein in cells via a CRISPR approach. This maintains endogenous expression levels but facilitates protein quantification when combined with NanoBRETTM technology using ectopically expressed Halo-​tag-​E3 ligase or Halo-​tag ubiquitin, to provide an in-​cell method for kinetic monitoring of the ternary complex and downstream ubiquitination. Additionally, Roy et al.12 have described a surface plasmon resonance (SPR) approach to characterizing the ternary complex between VHL and BET family proteins. Both of these reports highlight the importance of ternary complex and how effective PROTAC design can lead to cooperative binding in the ternary complex, which results in cellular degradation potencies that would not have been predicted from the biochemical

 79

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

79

potencies of the protein and E3 ligase binding moieties in the PROTAC. These optimization strategies can contribute to improved in vivo efficacy.

4.2.2.2 Impact of Protein Half-​life Modern proteomics techniques, such as dynamic stable isotope labeling by amino acids in cell culture (SILAC) labeling experiments, have provided a wealth of data on protein half-​lives.13 The median half-​life of proteins in cells is 46 h, but ranges from 500 h for some kinases such as RIPK2 in long-​lived immune cells.14 The duration of the PD response following protein degradation via a PROTAC is likely to be impacted by the protein half-​life. For intracellular proteins with a short half-​life that are rapidly synthesized, the duration of PD response would be expected to be closely related to the PK profile of the PROTAC in the absence of any biologically driven PD extension. In contrast, for proteins that have long half-​lives there is an expectation that extended PD might be observed beyond the detectable presence of the drug (see Section 4.3), driven by the slow synthesis of the protein, although the level of degradation required to drive the PD will also impact this. Figure 4.2 shows simulations of the recovery of protein levels following complete degradation, with protein half-​lives of 1 h, 24 h or 240 h in the absence of any PROTAC. While the absence of a PROTAC in the simulation is a simplification, these graphs are nevertheless illustrative of the effect on PD extension that might be expected with these different protein half-​lives. Clearly, the amount of protein degradation required to drive the desired pharmacology is an important additional consideration. If >90% protein degradation is required to deliver the desired level of pharmacology then the PD extension from protein synthesis rates will be relatively modest (~1 day) even for a protein with a half-​life of 240 h. If, however, a lower amount of protein degradation can still deliver the desired pharmacology, then significant PD

Figure 4.2 Simulated recovery of protein levels in the absence of a PROTAC for a protein with a half-​life of 1, 24 or 240 h.

80

80

Chapter 4

extensions might be expected. In practice, however, the extent of PD extension with a PROTAC will depend on the interplay between the protein synthesis rate and the PK of the compound. An additional potential complication is that protein half-​life differs among different cell types, with long-​lived cells such as T cells typically having extended half-​lives for proteins. If the PD response is driven largely from a single cell type, then it may be possible to take this into account, although for degradation targets that are distributed among many cell types with different half-​lives then pragmatically an average half-​life may need to be adopted.

4.2.2.3 Rate of PROTAC-​mediated Degradation It is evident from in vitro data that PROTAC mediated degradation occurs over a time course ranging from minutes to many hours. There may be many reasons for this, such as the half-​life of the protein, the kinetics of ternary complex formation as discussed above, or additional factors such as the extent of ubiquitination. These additional factors will not be considered in detail here. An example of a rapid in vitro PROTAC-​mediated degradation is from the work of Basi et al.,15 where approximately 90% PCAF degradation was achieved by PROTAC GSK983 in human peripheral blood mononuclear cells (PBMCs) in 30 min. In contrast, Figure 4.3 shows the time course for degradation of RIPK2 by PROTAC 716 (Table 4.1) in both human and rat PBMCs where it takes >12 h to reach high levels of degradation and the difference in DC50, the concentration required to degrade 50% of the target protein, at 6 and 24 h can be clearly seen. Thus, it would be anticipated that the DC50 of a PROTAC is likely to decrease as protein degradation proceeds depending on the relationship between the extent of degradation and the functional response. This may need to be accounted for in the building of the PK/​PD relationships for PROTACs.

Figure 4.3 (A) Time course of RIPK2 degradation in rat and human PBMCs following treatment with 30 nM of PROTAC 7. (B) DC50 of PROTAC 7 at 6 h (DC50 = 2 nM) and 24 h (DC50 = 0.4 nM). Reproduced from Ref. 16, https://​doi.org/​ 10.1038/​s42003-​020-​0868-​6, under the terms of a CC BY 4.0 license, https://​creativecommons.org/​licenses/​by/​4.0/​.

 81

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

81

4.2.2.4 Functional Consequences of PROTAC Binding to the Degradation Target In principle, PROTAC-​mediated protein degradation requires only an affinity binder to the target protein in order to recruit it to an E3 ligase complex, and this has been widely cited as an advantage of the PROTAC mechanism of action that should allow this technology to target disease-​causing proteins that were previously considered undruggable. In these instances, the binding of the PROTAC to the protein may be functionally silent and efficacy will be driven entirely by protein degradation, although the amount of degradation required to drive the functional response will likely vary from target to target. Nevertheless, almost all PROTACs reported in the literature to date bind at functional binding sites such as the ATP pocket in kinases or the ligand binding domain of a nuclear receptor. However, more than one pharmacological scenario might need to be considered in building PK/​PD relationships. One situation is where all the pharmacology of a target is expressed through the functional binding site. In this case the PROTAC may act as both an inhibitor and a degrader. This will impact both the rate of onset of in vivo efficacy and its duration. At very early time points, efficacy may come largely from direct inhibition, but as the protein is degraded and the compound cleared then efficacy will now largely be driven by degradation. PK/​PD models for such a PROTAC will need to consider the impact of the PK profile on both the inhibition and degradation characteristics, such as rate of degradation and protein half-​life as discussed in Sections 4.2.2.2 and 4.2.2.3 (see Section 4.3 for the discussion around RIPK2 PROTACs). A different scenario arises where there is a functional consequence of binding to the target and then additional biology is expressed through the degradation of the protein. An example of this is where the protein has scaffolding functions which are only perturbed on protein degradation. The contributions to the PD readout from both direct binding and degradation will need to be assessed and factored into PK/​PD modeling and how this impacts disease modulation.

4.2.2.5 E3 Ligase and Target Protein Distribution Protein degradation by PROTACs will be impacted by the protein levels of both the degradation target protein and E3 ligase complex. The E3 ligases which have seen the greatest incorporation into PROTACs are cereblon, VHL and, to a lesser extent, members of the inhibitors of apoptosis (IAP) family. All of these E3 ligases are broadly expressed in cells. This represents the simplest scenario in terms of building a detailed PK/​PD understanding. Nevertheless, it is clear that there is a huge potential opportunity in utilizing E3 ligases with restricted distribution to limit degradation to certain cell types or tissues and identify medicines with improved therapeutic index. While there is information regarding protein expression levels in cells, it is far from comprehensive, and this may add additional complexities in developing a detailed PK/​PD understanding, especially if this needs to support a hypothesis around improved therapeutic index.

82

82

Chapter 4

Figure 4.4 Summary of additional considerations for building PK/​PD understanding with PROTACs.

4.3 Developing PK/​PD Relationships for a Series of RIPK2 PROTACs As part of a drug discovery program exploring RIPK2 PROTACs we sought to develop a detailed PK/​PD understanding. RIPK2 (also known as RIP2, RICK) is a cytosolic serine–​threonine kinase which is activated by the pattern recognition receptors NOD1 and NOD2 as part of the innate immune system. NOD1 and NOD2 recognize bacterial cell wall components such as muramyl dipeptide (MDP), which leads to phosphorylation and activation of RIPK2. This ultimately leads to the production of inflammatory cytokines such as TNFα. Dysregulation of this pathway has been implicated in autoimmune diseases such as inflammatory bowel disease (IBD).17

4.3.1 Design of PK/​PD Experiments for RIPK2 PROTACs The following considerations went into the design of the PK/​PD experiments. (a) MDP represents a pathway-​specific activator of the NOD2–​RIPK2 pathway that can be used in human and rat whole blood assays measuring cytokine production, typically with inhibition of TNFα production as an endpoint. While this can be used in a purely in vitro setting to provide compound potency data, it can also be conveniently deployed as an ex vivo whole blood assay to provide a readout in both pre-​clinical and clinical studies. The advantage of this kind of functional readout is that it captures additional factors impacting efficacy beyond the cellular DC50, such as the impact of protein binding in the biophase. Any NOD2-​independent biology18 may not be captured by this PD measurement. (b) RIPK2 has been targeted via traditional kinase inhibitor programs19–21 which afforded access to potent and selective binders at the ATP pocket. As discussed in Section 4.2.2.4, this is a scenario where binding and degradation will elicit the same functional response from MDP stimulation. It is therefore to be expected that the PD cytokine response will

 83

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

83

Table 4.1 Structure of RIPK2 PROTACs and table showing biochemical potency, MDP challenge assay potency in both human and rat whole blood as well as rat intravenous PK parameters.

PROTAC hRIPK2 FP IC50 (nM)

hWB MDP IC50 (nM)

rWB MDP Rat clearance IC50 (nM) (mL min–​1 kg–​1)

Rat VdSS (L kg–​1)

Rat t½ (h)

6 7 8

20 3 1.5

20 3 —​

11 7.6 14

6.9 16 24

250 10 13

18 10 17

be a composite of direct inhibition and protein degradation, which will need to be elucidated. (c) RIPK2 has a long half-​life in immune cells in the blood ranging from ~50 h in monocytes to >500 h in T cells. As such, it was anticipated that RIPK2 PROTACs might drive extended PD responses that were disconnected from circulating drug levels and that the duration of the PK/​PD experiments would need to accommodate the recovery phase for the protein levels. As we were interested in the overall capacity of a PROTAC to inhibit TNFα production, no effort was made to consider the individual synthesis rates in the different cells types. Studies would be conducted in rat, collecting a data package measuring the PROTAC drug levels, the RIPK2 protein levels and inhibition of MDP-​stimulated ex vivo TNFα production. This would be carried out over an appropriate time course and dosing regimen to define both the rate of RIPK2 degradation and the synthesis rate of the protein, following clearance of PROTAC drug levels. These studies would also provide an understanding of the amount of RIPK2 degradation required to maintain a specific

84

84

Chapter 4

level of TNFα inhibition (e.g., >90%), analogous to target engagement information gathered in traditional small-​molecule programs.

4.3.2 Results from PK/​PD Experiments with RIPK2 PROTACs Initial medicinal chemistry optimisation work at GSK identified PROTAC 5 (Table 4.1) as a potent degrader of RIPK2 (DC50 0.2 nM) in human PBMCs and good human whole blood activity (TNFα inhibition IC50 = 10 nM). This compound recruited E3 ligases from the IAP family of E3 ligases and was a more potent degrader than the corresponding VHL-​or cereblon-​based PROTACs. Unfortunately, PROTAC 5 had very high microsomal intrinsic clearance in both rat and human microsomes (11 and 29 mL min–​1 g–​1 liver in rat and human microsomes, respectively), possibly as a result of its high lipophilicity. Further chemistry optimization identified PROTAC 6. While this compound incorporated a significantly lower affinity RIPK2 binder (IC50 = 250 nM) than 5, its human whole blood potency (human WB IC50 = 20 nM) was only marginally less potent than 5, possibly due to improved physicochemical properties resulting in a higher fraction unbound. Rat whole blood activity (rat WB pIC50 = 7.7), following MDP stimulation, was essentially identical to that in human whole blood, but significantly, this compound demonstrated low clearance in hepatocytes (60% degradation and >80% inhibition of TNFα production observed. Nevertheless, the next two doses increased the response such that a maximal response was observed for both protein degradation and TNFα inhibition. Dosing at 0.05 mg kg–​1 proved to be a more convenient dose to study the additive pharmacology as the first dose only resulted in ~30% degradation. At the 24 h time point, immediately before the second dose, TNFα inhibition had increased to >60% despite there being no measurable drug concentration at this time point. Following the third dose, inhibition of TNFα had increased to nearly 80% while degradation of RIPK2 approached 60% at 8 h. The gradual increase in PD response over the time course of the study occurred even though PROTAC concentrations could not be measured for the majority of the study. The contrast with a traditional small-​molecule PK/​PD profile driven by occupancy is dramatic. This PD phenomenon observed with PROTACs against targets with a slow synthesis rate has important implications in human dose modeling.

4.4 PBPK/​PD Models for PROTACs Within the pharmaceutical industry there has been a strong focus on establishing PK/​PD relationships to aid translation and to reduce drug attrition in later stage clinical development.22,23 A major challenge for clinical translation is how to utilize the knowledge gained from in vivo animal studies, both PK and PD, to allow development and parameterization of PK/​PD models to aid human efficacious dose prediction with high confidence. While various empirical methods can be used to predict the human distribution volume and clearance,24,25 depending on the approach taken a range of clinical doses will be predicted; as such, a more mechanistic approach was used within the RIPK2 PROTAC program. PBPK modeling is a widely established tool that can aid clinical PK translation and human dose prediction.26 Representing the body

88

88

Chapter 4

as a series of compartments, parameterized based on species-​specific physiology (e.g., organ blood flow and weights, tissue composition) and integrating with compound-​specific data (e.g., logP, pKa, blood to plasma ratio), PBPK models can be used to retrospectively simulate observed or prospectively predict concentration–​time profiles. A PBPK/​PD model can be developed in different platforms and is dependent on the availability and functionality of the commercial software (e.g., GastroPlus, Simcyp, SimBiology/​Matlab), as well as the familiarity of the model developer and user to the software. PBPK modeling of the IV pre-​clinical PK and toxicokinetic data in rats, mini-​pigs and dogs was performed on PROTAC 7 using GastroPlus v.9.6 (Simulations Plus). Considering the size of PROTAC 7 (molecular weight >1000 g mol–​1) no renal clearance was expected and systemic clearance was assigned to the liver alone and set to be a perfusion-​limited organ with fixed clearance, based on non-​compartmental PK analysis of respective IV data. The estimated systemic clearance of PROTAC 7 was found to be consistent across the pre-​clinical species, ~10% of liver blood flow; as such, the human systemic clearance was assumed to be the same. Additionally, the size of PROTACs would mean it was also reasonable to use a permeability-​limited tissue model to describe tissue distribution. For permeability-​limited tissues, the permeability–​surface area product of the tissue (SpecPStc), a key parameter which describes the drug penetration across the cell membrane of the tissue, was estimated from rat PK profiles using the Optimization module embedded in the GastroPlus software. With the exception of brain, which was not unexpected due to the blood–​brain barrier, the predicted tissue distribution agreed reasonably well with measured pseudo-​steady state rat IV infusion data. As such, the SpecPStc was applied to both the dog and mini-​pig PBPK models with the resulting simulated IV PK profiles showing very good accordance with measured concentration–​time profiles from the various dosing regimens, suggesting its applicability across all pre-​clinical species and subsequently for use in building the human IV PBPK model. To build the PK/​PD model, pre-​clinical rat single-​or multiple-​dose PK/​PD data consisting of SC PK profiles, RIPK2 protein levels and TNFα inhibition data were collated and modeled using SimBiology software. PK profiles were simultaneously fitted using a three-​compartment PK model, parameterized from the observed IV PK profile, together with additional transit compartments to describe kinetics of PROTAC 7 from SC administration to the central compartment. An indirect response model (eqn (4.1)), was applied to describe the RIPK2 inhibition driven by the unbound systemic concentration via stimulation of RIPK2 degradation.27



 Emax ⋅ fu,p ⋅ C p dRIP = Ksyn − K deg  1 +  dt  EC50 + fu,p ⋅ C p

  ⋅ RIP (4.1)  

 89

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

89

(a)

(b)

(c)

Figure 4.8  (A) Schematic representation of the developed (PB)PK/​ PD model. (B) Simulated human PK profile following a 1.25 mg QD (--) or QD × 3days (—​) IV slow bolus injection of PROTAC 7. (Dashed blue line represents a

90

90

Chapter 4

where RIP is the RIPK2 abundance in cells with arbitrary unit (AU). Emax (dimensionless) is the maximum stimulus effect. EC50 is the concentration at which half of the maximum effect is achieved. Ksyn and Kdeg are RIPK2 synthesis rate (AU h–​1) and degradation rate constant (1 h–​1), respectively. For TNFα inhibition, a clinically translatable ex vivo rat MDP challenge model was used. An empirical model was developed to establish a relationship between ex vivo TNFα inhibition and the unbound fraction of RIPK2 with respect to the basal RIPK2 level, as given in eqn (4.2). 2



 RIPunbound  TNF=  (4.2) ex vivo 0.042 + 0.93 ⋅   RIPo 

where RIPunbound is the unbound fraction of RIPK2 and RIP0 is the basal RIPK2. Integration of these models into the full PK/​ PD model adequately described the observed PK/​PD relationship in rat. Assuming that the PD responses (RIPK2 degradation and TNFα inhibition) in humans have the same parameters as those in rat, the human PD response following IV administration can be predicted using the SimBiology PK/​PD model when parameterized using a human three-​compartmental PK model that approximates the GastroPlus human IV permeability-​limited PBPK model. The developed (PB)PK/​PD model (Figure 4.8) illustrates how a potential IV slow bolus study with PROTAC 7 could be designed to target 80% inhibition of TNFα for more than 24 h after a single dose or multiple daily dosing for 3 days.

4.5 Conclusions It is now widely recognized that there is enormous potential to develop highly differentiated PROTAC medicines. While most of this attention has been focused on the pharmacological opportunity that arises from complete removal of a target protein that can recapitulate a knockout phenotype, the unique mechanism of action of PROTACs can also deliver attractive benefits on in vivo efficacy. In cases where protein synthesis is slow, extended PD responses allow consideration of infrequent dosing regimens or an ability to work off low drug concentrations. Given the correlation between low human dose and the relative success for the develop of new medicines, this represents a particularly attractive feature. To fully capitalize on the opportunity for PROTAC medicines, irrespective of the protein half-​life, a rigorous approach to building PK/​PD relationships with PROTACs that capture the unique Figure 4.8 (continued) LC-​MS/​MS lower limit of quantification of 1 ng mL–​1). (C) Corresponding simulated PD responses for RIPK2 degradation and TNFα inhibition following single (dashed) or multiple (solid) IV dose administration.

 91

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

91

manner in which they exert their pharmacology will be required in order to deliver high-​quality human dose predictions.

4.6 Ethical  Review All animal studies were ethically reviewed and carried out in accordance with Animals (Scientific Procedures) Act 1986 and the GSK Policy on the Care, Welfare and Treatment of Animals.

References 1. K. M. Sakamoto, K. B. Kim, A. Kumagai, F. Mercurio, C. M. Crews and R. J. Deshaies, Proc. Natl. Acad. Sci. U. S. A., 2001, 98(15), 8554–​8559. 2. D. P. Bondeson, A. Mares, I. E. Smith, E. Ko, S. Campos, A. H. Miah, K. E. Mulholland, N. Routly, D. L. Buckley, J. L. Gustafson, N. Zinn, P. Grandi, S. Shimamura, G. Bergamini, M. Faelth-​Savitski, M. Bantscheff, C. Cox, D. A. Gordon, R. R. Willard, J. J. Flanagan, L. N. Casillas, B. J. Votta, W. den Besten, K. Famm, L. Kruidenier, P. S. Carter, J. D. Harling, I. Churcher and C. M. Crews, Nat. Chem. Biol., 2015, 11(8), 611–​617. 3. G. E. Winter, D. L. Buckley, J. Paulk, J. M. Roberts, A. Souza, S. Dhe-​ Paganon and J. E. Bradner, Science, 2015, 348(6241), 1376–​1381. 4. A. Zorba, C. Nguyen, Y. Xu, J. Starr, K. Borzilleri, J. Smith, H. Zhu, K. A. Farley, W. Ding, J. Schiemer, X. Feng, J. S. Chang, D. P. Uccello, J. A. Young, C. N. Garcia-​Irrizary, L. Czabaniuk, B. Schuff, R. Oliver, J. Montgomery, M. M. Hayward, J. Coe, J. Chen, M. Niosi, S. Luthra, J. C. Shah, A. El-​Kattan, X. Qiu, G. M. West, M. C. Noe, V. Shanmugasundaram, A. M. Gilbert, M. F. Brown and M. F. Calabrese, Proc. Natl. Acad. Sci. U. S. A., 2018, 115(31), E7285–​E7292. 5. X. Han, C. Wang, C. Qin, W. Xiang, E. Fernandez-​Salas, C. Y. Yang, M. Wang, L. Zhao, T. Xu, K. Chinnaswamy, J. Delproposto, J. Stuckey and S. Wang, J. Med. Chem., 2019, 62(2), 941–​964. 6. Y. Li, J. Yang, A. Aguilar, D. McEachern, S. Przybranowski, L. Liu, C. Y. Yang, M. Wang, X. Han and S. Wang, J. Med. Chem., 2019, 62(2), 448–​466. 7. X. Sun, J. Wang, X. Yao, W. Zheng, Y. Mao, T. Lan, L. Wang, Y. Sun, X. Zhang, Q. Zhao, J. Zhao, R. P. Xiao, X. Zhang, G. Ji and Y. Rao, Cell Discovery, 2019, 5, 10. 8. P. Morgan, P. H. Van Der Graaf, J. Arrowsmith, D. E. Feltner, K. S. Drummond, C. D. Wegner and S. D. A. Street, Drug Discovery Today, 2012, 17(9), 419–​424. 9. T. Tuntland, B. Ethell, T. Kosaka, F. Blasco, R. X. Zang, M. Jain, T. Gould and K. Hoffmaster, Front. Pharmacol., 2014, 5, 174. 10. G. F. Watt, P. Scott-​Stevens and L.Gaohua, Drug Discovery Today: Technol., 2019, 31,  69–​80. 11. K. M. Riching, S. Mahan, C. R. Corona, M. McDougall, J. D. Vasta, M. B. Robers, M. Urh and D. L. Daniels, ACS Chem. Biol., 2018, 13(9), 2758–​2770.

92

92

Chapter 4

12. M. J. Roy, S. Winkler, S. J. Hughes, C. Whitworth, M. Galant, W. Farnaby, K. Rumpel and A. Ciulli, ACS Chem. Biol., 2019, 14(3), 361–​368. 13. B. Schwanhausser, D. Busse, N. Li, G. Dittmar, J. Schuchhardt, J. Wolf, W. Chen and M. Selbach, Nature, 2011, 473(7347), 337–​342. 14. D. F. Tough and J. Sprent, Immunol. Res., 1995, 14(1),  1–​12. 15. Z. I. Bassi, M. C. Fillmore, A. H. Miah, T. D. Chapman, C. Maller, E. J. Roberts, L. C. Davis, D. E. Lewis, N. W. Galwey, K. E. Waddington, V. Parravicini, A. L. Macmillan-​Jones, C. Gongora, P. G. Humphreys, I. Churcher, R. K. Prinjha and D. F. Tough, ACS Chem. Biol., 2018, 13(10), 2862–​2867. 16. A. Mares, A. H. Miah, I. E. D. Smith, M. Rackham, A. R. Thawani, J. Cryan, P. A. Haile, B. J. Votta, A. M. Beal, C. Capriotti, M. A. Reilly, D. T. Fisher, N. Zinn, M. Bantscheff, T. T. MacDonald, A. Vossenkamper, P. Dace, I. Churcher, A. B. Benowitz, G. Watt, J. Denyer, P. Scott-​Stevens and J. D. Harling, Commun. Biol., 2020, 3(1), 140. 17. J. Bertin, W. J. Nir, C. M. Fischer, O. V. Tayber, P. R. Errada, J. R.Grant, J. J. Keilty, M. L. Gosselin, K. E. Robison, G. H. Wong, M. A. Glucksmann and P. S. DiStefano, J. Biol. Chem., 1999, 274(19), 12955–​12958. 18. T. Watanabe, K. Minaga, K. Kamata, T. Sakurai, Y. Komeda, T. Nagai, A. Kitani, M. Tajima, I. J. Fuss, M. Kudo and W. Strober, Int. Immunol., 2019, 31(10), 669–683. 19. P. A. Haile, L. N. Casillas, M. J. Bury, J. F. Mehlmann, R. Singhaus, Jr., A. K. Charnley, T. V. Hughes, M. P. DeMartino, G. Z. Wang, J. J. Romano, X. Dong, N. V. Plotnikov, A. S. Lakdawala, M. A. Convery, B. J. Votta, D. B. Lipshutz, B. M. Desai, B. Swift, C. A. Capriotti, S. B. Berger, M. K. Mahajan, M. A. Reilly, E. J. Rivera, H. H. Sun, R. Nagilla, C. LePage, M. T. Ouellette, R. D. Totoritis, B. T. Donovan, B. S. Brown, K. W. Chaudhary, P. J. Gough, J. Bertin and R. W. Marquis, ACS Med. Chem. Lett., 2018, 9(10), 1039–​1044. 20. P. A. Haile, L. N. Casillas, B. J. Votta, G. Z. Wang, A. K. Charnley, X. Dong, M. J. Bury, J. J. Romano, J. F. Mehlmann, B. W. King, K. F. Erhard, C. R. Hanning, D. B. Lipshutz, B. M. Desai, C. A. Capriotti, M. C. Schaeffer, S. B. Berger, M. K. Mahajan, M. A. Reilly, R. Nagilla, E. J. Rivera, H. H. Sun, J. K. Kenna, A. M. Beal, M. T. Ouellette, M. Kelly, G. Stemp, M. A. Convery, A. Vossenkamper, T. T. MacDonald, P. J. Gough, J. Bertin and R. W. Marquis, J. Med. Chem., 2019, 62, 6482–6494. 21. P. A. Haile, B. J. Votta, R. W. Marquis, M. J. Bury, J. F. Mehlmann, R. Singhaus, Jr. A. K. Charnley, A. S. Lakdawala, M. A. Convery, D. B. Lipshutz, B. M. Desai, B. Swift, C. A. Capriotti, S. B. Berger, M. K. Mahajan, M. A. Reilly, E. J. Rivera, H. H. Sun, R. Nagilla, A. M. Beal, J. N. Finger, M. N. Cook, B. W. King, M. T. Ouellette, R. D. Totoritis, M. Pierdomenico, A. Negroni, L. Stronati, S. Cucchiara, B. Ziolkowski, A. Vossenkamper, T. T. MacDonald, P. J. Gough, J. Bertin and L. N. Casillas, J. Med. Chem., 2016, 59(10), 4867–​4880. 22. I. Kola and J. Landis, Nat. Rev. Drug Discovery, 2004, 3(8), 711–​715.

 93

Developing Pharmacokinetic/Pharmacodynamic Relationships With PROTACs

93

23. P. Morgan, P. H. Van Der Graaf, J. Arrowsmith, D. E. Feltner, K. S. Drummond, C. D. Wegner and S. D. Street, Drug Discovery Today, 2012, 17(9–​10), 419–​424. 24. P. Zou, Y. Yu, N. Zheng, Y. Yang, H. J. Paholak, L. X. Yu and D. Sun, AAPS J., 2012, 14(2), 262–​281. 25. P. Zou, N. Zheng, Y. Yang, L. X. Yu and D. Sun, Expert. Opin. Drug Metab. Toxicol., 2012, 8(7), 855–​872. 26. N. A. Miller, M. B. Reddy, A. T. Heikkinen, V. Lukacova and N. Parrott, Clin. Pharmacokinet., 2019, 58(6), 727–​746. 27. M. A. Felmlee, M. E. Morris and D. E. Mager, Methods Mol. Biol., 2012, 929, 583–​600.

94

CHAPTER 5

New Activities of CELMoDs, Cereblon E3 Ligase-​modulating Drugs MARY E. MATYSKIELA*, THOMAS CLAYTON, JOEL W. THOMPSON, CHRISTOPHER CARROLL, LESLIE BATEMAN, LAURIE LEBRUN AND PHILIP P. CHAMBERLAIN* Celgene Corporation, 10300 Campus Point Dr., suite 100, San Diego, CA, USA *Corresponding author. Email: [email protected], [email protected]

5.1 Introduction The last few years have seen an explosion of interest in the potential of small-​ molecule therapeutics capable of redirecting the cellular ubiquitin–​proteasome machinery to destroy specific proteins. While one approach utilizes hetero-​ bifunctional ligands capable of simultaneously binding both an E3 ubiquitin ligase and the target protein, another class of molecules, termed CELMoDs (cereblon E3 ligase-​modulating drugs), are low-​molecular-​weight small molecules that induce the degradation of specific protein targets by binding to the cereblon-​CRL4 E3 ubiquitin ligase and scaffolding direct protein–​protein interactions to the target protein.1 These molecules are capable of recruiting undruggable and even unligandable targets through a so-​called “molecular

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

94

 95

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

95

glue” mechanism. We focus on recent discoveries of CELMoD mechanism of action and discuss the future and breadth of this emerging class of molecules. Despite the massive interest and clinical successes now achieved by the thalidomide analogues, the history of this molecular class also provides a tragic case study in human developmental toxicology. Thalidomide was first marketed as an over-​the-​counter medicine in the 1950s by Chemie Grunenthal, with recommended uses including as a sedative, a cold remedy and in the treatment of morning sickness.2,3 It was subsequently realized that the use of thalidomide during pregnancy was the cause of a global epidemic of severe birth defects, with thousands of babies affected and many more miscarriages attributed to pre-​ natal thalidomide exposure.4,5 Birth defects included various symptoms including the characteristic alteration in limb development described as phocomelia. Thalidomide was withdrawn from use in the 1960s, and the lessons learned from this tragedy helped establish the modern era of pharmacovigilence. It was subsequently found that thalidomide was effective in treating erythema nodosum leprosum, a complication of leprosy.6 Thalidomide was also found to be effective in the treatment of multiple myeloma, and Celgene was granted clinical approval for this use in the 1990s. A strict risk management procedure was critical to preventing exposure of pregnant women to the drug. Celgene developed lenalidomide as a more potent analogue, and achieved clinical approvals including for the treatment of multiple myeloma and in myelodysplastic syndrome patients bearing the del(5q) genetic lesion. Pomalidomide was subsequently developed as a potent analogue with clinical approvals for the treatment of relapsed/​refractory myeloma. These drugs are now part of the standard of care for the treatment of both newly diagnosed and relapsed/​refractory multiple myeloma. Development of thalidomide, lenalidomide and pomalidomide was largely completed using phenotypic markers for activity. However in 2010, affinity purification experiments found the protein cereblon to be the direct molecular binder for thalidomide analogues.7 Cereblon had been previously annotated to be a DDB1 and CUL4 associated factor (DCAF), marking it as a component of the CRL4 E3 ubiquitin ligase.8 Further works showed that genetic knockdown of cereblon blocked the effects of thalidomide analogues, indicating that rather than being inhibitors of the E3 ligase, these drugs were imparting a gain-​of-​function activity.9,10 The discovery of a clinical class of molecules that operate by binding and repurposing an E3 ubiquitin ligase now presented an incredible opportunity for drug discovery.

5.2 Targeted Protein Degradation Through Cereblon-​CRL4 Targeted protein degradation harnesses and redirects the cell’s own protein disposal machinery, termed the ubiquitin–​proteasome system (UPS), toward new therapeutic targets of interest. To mediate protein degradation in the cell, the UPS covalently modified substrates destined for destruction with the small protein ubiquitin.11,12 Ubiquitination occurs through a coordinated cascade of

96

96

Chapter 5

three classes of enzymes with ubiquitin first being activated by an E1 (ubiquitin-​ activating enzyme), then transferred to an E2 (ubiquitin-​conjugating enzyme), and finally brought into close proximity with a substrate for ubiquitin transfer by an E3 (ubiquitin ligase).13 Specificity in the UPS is conferred by the E3, which binds directly to target substrates to facilitate the transfer of ubiquitin from an E2. Mono-​ubiquitinated substrates can undergo further rounds of ubiquitination whereby ubiquitin is added to lysines of ubiquitin itself, leading to formation of polyubiquitin chains and marking the substrate for proteasomal degradation. Targeted protein degradation utilizes small-​molecule ligands to induce the recruitment of a protein of interest to an E3, redirecting the UPS to ubiquitinate proteins that are not normally subjected to proteolysis. There are distinct classes of small-​molecule ligands capable of redirecting E3 ubiquitin ligases toward new proteins, with different applications and pharmacological properties.14 One class are the hetero-​bifunctional molecules that have two separate binding moieties, one that binds the target (substrate) protein and one that binds the E3 ubiquitin ligase. These separate functionalities are joined by a linker region to enable the simultaneous binding of the ligase and substrate. These molecules were named PROTACs (Proteolysis-​TArgeting Chimeras) by Craig Crews, Ray Deshaies and colleagues, who demonstrated that hetero-​ bifunctional peptide-​based ligands could cause protein ubiquitination.15 A second class of molecules is capable of driving ubiquitination and degradation of target proteins without the need for discrete binding moieties for substrate and ligase. For example, compounds may bind an E3 ligase and alter the protein surface to create a new protein–​protein interaction interface, as typified by the class of molecules termed CELMoDs, including the clinically approved thalidomide, lenalidomide and pomalidomide.7,16–​19 This mechanism has been termed “molecular glue” and as was first described for the plant hormone auxin.20. CELMoDs act to scaffold direct protein–​protein interactions between the cereblon-​CRL4 E3 ubiquitin ligase and substrate, and as such do not need separate moieties or linkers to form the complex.

5.3 CRL4 Architecture The cereblon-​CRL4 ubiquitin ligase is part of a large class of E3 ubiquitin ligases (∼400) called cullin-​ring ligases (CRLs).21 CRLs share a general architecture of a large cullin scaffolding subunit that binds a substrate receptor subunit at its N-​terminus and a ring subunit (Rbx1) at its C-​terminus that provides the site for E2 recruitment. Thus, CRLs provide a protein scaffold upon which a substrate is brought into sufficiently close proximity to a ubiquitin-​ conjugated E2 enzyme such that a substrate lysine can attack the E2–​ubiquitin bond for direct ubiquitin transfer between E2 and substrate. CRL-​mediated protein ubiquitination regulates numerous fundamental cellular processes, including the cell cycle,22,23 oxygen and iron homeostasis24,25 and the N-​and C-​end rules of protein stability.26,27 Cereblon is one of ~70 substrate receptors, or DCAFs (DDB1 and Cul4 associated factors), that bind the CRL4 (DDB1-​Cul4-​Rbx1) ubiquitin ligase to mediate

 97

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

97

Figure 5.1 Architecture of the cereblon-​CRL4 ubiquitin ligase. (A) Structural model of the cereblon-​CRL4 ubiquitin ligase. The Cul4 scaffolding subunit (light blue) brings together the E2 binding subunit Rbx1 (dark blue) and the DDB1 adapter subunit (royal blue). Cereblon (green) binds to DDB1 and recruits substrates, allowing a substrate lysine to attack the E2–​ubiquitin bond (E2 purple, ubiquitin yellow). Red arrow points to bound lenalidomide. (CRL4 model built using PDB IDs : 4TZ4, 2HYE, 4AP4.) (B) Domain architecture and crystal structure of cereblon–​DDB1 (PDB ID : 4TZ4). Cereblon consists of three domains, the lon-​like domain (LLD, orange), the thalidomide-​binding domain (TBD, green) and the DDB1-​binding domain (red) which binds to DDB1 (blue), with red arrow pointing to lenalidomide bound in the tri-​tryptophan pocket.

substrate recruitment.8,28–​30 DCAFs like cereblon bind to the adapter protein DDB1, which links them to the Cul4 scaffold (Figure 5.1A). While most CRL4 substrate adapters are WD40 domain-​containing proteins, cereblon does not contain WD40 repeats, and its fold was unknown until its structure was determined through X-​ray Crystallography studies.16,17 The structural studies elucidated that cereblon contains three domains, a lon-​like domain (LLD) which has homology to lon protease, a thalidomide binding domain (TBD) which contains a hydrophobic pocket for ligand binding, and a helical bundle region which makes extensive contacts with DDB1 and provides the attachment site to the rest of the CRL4 ligase (Figure 5.1B). Structural studies of DDB1-​bound Cul4 have demonstrated significant rotational freedom around beta propeller B of DDB1, which links DDB1 to Cul4. Rotation varying by as much as 60 degrees has been observed, in addition to flexibilities in the Cul4 and Rbx1 proteins.31,32 Such flexibility may explain how CRLs catalyze ubiquitin transfer to substrates that are highly diverse in both shape and size, as well as how these ligases can be hijacked to modify diverse targets recruited by small molecules. Multiple E2s have been identified to work with CRL4 for the drug-​dependent ubiquitination of cereblon neosubstrates, and Ube2D3 and Ube2G1 have been shown to act sequentially to prime and extend ubiquitin chains on neosubstrates, respectively.33,34

98

98

Chapter 5

Figure 5.2  Cereblon binding molecules for targeted protein degradation. (A) Representative CELMoDs compared to hetero-​bifunctional degrader dBet-​1.35 (B) Lenalidomide docked into the tri-​ tryptophan pocket of cereblon, forming hydrogen bonds with the backbone and sidechain of H378 and the backbone of W380.

5.4 CELMoD Mechanism of Action The CELMoD compounds shown are comprised of a glutarimide fused to either an isoindolinone moiety, as in lenalidomide, or a phthalimide moiety, as in thalidomide and pomalidomide (Figure 5.2A). Crystal structures demonstrated that these molecules bind cereblon by the docking of their glutarimide ring in a hydrophobic pocket formed by three tryptophan residues, W380, W386 and W40016,17 (Figure 5.2B). The glutarimide is anchored in the tri-​Trp pocket via hydrogen bonds with the backbone and sidechain of histidine 378 and the backbone of tryptophan 380. This binding mode allows the isoindolinone or phthalimide to protrude out of the tri-​Trp pocket and present a planar hydrophobic group above the cereblon surface. While CELMoDs were initially hypothesized to be cereblon inhibitors, key studies published in 2014 and 2015 established that CELMoD binding resulted in the recruitment and degradation of new proteins, termed neosubstrates. Via luciferase reporter assays, co-​immunoprecipitation, and proteomics experiments, Lu et al., Gandhi et al. and Kronke et al. demonstrated that the T-​and B-​cell zinc finger transcription factors Ikaros and Aiolos are recruited to cereblon upon treatment with lenalidomide.36–​38 The authors showed that while mRNA levels for Ikaros and Aiolos remained unchanged in compound treated samples, degradation of the two zinc fingers was compound-​, cereblon-​, cullin-​and proteasome-​dependent. In the absence of compound, Ikaros and Aiolos exert a dual effect, repressing expression of IL2 in T cells and stimulating expression of IRF4, an essential gene in multiple myeloma. Thus, these studies determined that the loss of Ikaros and Aiolos was a driver for lenalidomide’s therapeutic and immunomodulatory effects. A similar study published in 2015 by Kronke et al. attributed the efficacy of lenalidomide in del5q myelodysplastic syndrome (del(5q) MDS) to the ubiquitination and subsequent proteasomal degradation of the kinase casein kinase

 99

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

99

39

1A1 (CK1α). Ck1α is located on the section of the chromosome that is lost in del(5q) MDS, and aberrant cells are thus sensitized to further reduction of its levels by lenalidomide-​induced degradation. CK1α was identified by mass spectrometry as being induced to interact with cereblon upon lenalidomide treatment. As with Ikaros and Aiolos, no change in CK1α mRNA levels are observed following lenalidomide treatment and CK1α degradation was found to be cereblon-​, cullin-​and proteasome-​dependent, as expected for a cereblon neosubstrate. While degradation of Ikaros and Aiolos was induced with thalidomide, lenalidomide, pomalidomide and iberdomide, only lenalidomide was found to have a significant effect on levels of CK1α. This publication marked the first example of selectivity within this drug class, and suggested the possibility that additional targets could be susceptible to degradation via a CELMoD analogue

5.5 Identification of CELMoDs with Novel Activities The neosubstrate degradation activities of thalidomide, lenalidomide and pomalidomide presented key questions about whether it would be possible to develop differentiated CELMoDs that degrade additional protein targets. In a search for novel CELMoD activities, the molecule CC-​885 (Figure 5.2) was identified as an experimental CELMoD with highly differentiated activity, exhibiting broad antiproliferative effects across a panel of cancer cell lines, which are not observed with thalidomide, lenalidomide or pomalidomide.19 Acute myeloid leukemia (AML) cell lines were particularly sensitive to CC-​885 treatment, and this effect was confirmed in CC-​885-​treated patient-​derived AML tumor cells that showed increased sensitivity relative to matched normal lymphoid cells. Further studies indicated CC-​885 was enacting these effects through a new cellular target. Immunoprecipitation and mass spectrometry studies of cells treated with CC-​885 were used to identify the protein GSPT1 (G to S Phase Transition Protein 1) as a CC-​885-​dependent cereblon binding protein. GSPT1 is a translation termination factor that mediates stop codon recognition and release of nascent proteins from the ribosome, and GSPT1 protein levels were reduced with CC-​885 treatment. This reduction was found to be dependent upon the proteasome, cullin neddylation, and cereblon, as expected for a bona fide neosubstrate. CC-​885-​dependent ubiquitination of GSPT1 was observed both in cells and in a purified, reconstituted system, demonstrating direct cereblon binding. GSPT1 was not targeted by any of the previously reported CELMoD compounds, indicating that the elaborated chemical moieties of CC-​885 were allowing the recruitment of a differentiated protein target.

5.6 Molecular Basis for Substrate Recruitment Despite the identification of new activities, the molecular mechanisms of substrate recruitment by CELMoDs remained unclear. Based upon crystal structures showing the isoindolinone or phthalimide moiety exposed on the

100

100

Chapter 5

cereblon surface, it was hypothesized that the exposed planar hydrophobic group coupled with adjacent unsatisfied hydrogen bonds on the cereblon surface created a protein–​protein interaction hotspot for the recruitment of neosubstrates.16 This mechanism of bound ligand creating or extending a ubiquitin ligase substrate recruitment site was first described by Ning Zheng to explain the mechanism of action of the plant hormones auxin and jasmonate, termed “molecular glue.”20,40 Auxin and jasmonate bind F-​box protein substrate adapters for the SCF ubiquitin ligase in plants, providing a new complementary interaction surface for substrate binding. The lenalidomide-​bound cereblon crystal structures suggested that CELMoDs might exert their therapeutic effects through a similar molecular glue mechanism. A sequence and structural comparison of Ikaros/​Aiolos, Ck1α, and GSPT1 revealed the absence of a common sequence motif or functional domain between these cereblon neosubstrates. Ikaros and Aiolos are made up of C2H2 zinc finger domains. These small, simple (~30 amino acid) domains are highly prevalent in mammalian proteomes, often occurring in sequence as part of a DNA binding domain in which the alpha helix of the zinc finger domains bind the major groove of DNA. CK1α is a member of the casein kinase 1 serine–​threonine protein kinase family and is involved in inhibition of p53 and negative regulation of Wnt signaling. Structurally, the kinase domain of CK1α is comprised of a predominately beta sheet n-​lobe and an alpha helical c-​lobe. GSPT1 is a translation termination factor containing an N-​terminal GTP binding domain and two C-​terminal beta barrel protein–​protein interaction domains. Interaction of GSPT1 with eRF1 stimulates the GTP-​dependent release of nascent peptide from the ribosome. Taken together, the mechanism by which CELMoDs facilitated the direct interaction between cereblon and these three structurally distinct neosubstrates was not obvious. The molecular basis of neosubstrate recruitment was identified in 2016 by the publication of two ternary complex crystal structures of cereblon-​DBB1 bound to the neosubstrates GSPT1 and CK1α.18,19 Both ternary complex structures show a CELMoD-​mediated direct interaction between cereblon and neosubstrate. The interacting surface for both GSPT1 and CK1α is a structurally similar (RMSD 0.46) glycine-​containing beta hairpin loop (Figure 5.2A). The glycine residue in both loops sits adjacent to the exposed planar hydrophobic surface formed by the isoindolinone, and takes on torsion angles only accessible to glycine. The presence and position of the glycine in the beta hairpin loop is critical and mutational studies have shown that mutation of the glycine residue, even to alanine, blocks neosubstrate recruitment. At the time, no structures had been solved for Ikaros/​Aiolos; however, homology modelling and mutagenesis studies established that these two protein were recruited to cereblon through a structurally similar beta hairpin loop in their second C2H2 zinc finger domains.18,19 These findings were reinforced in 2018 when the ternary complex structure of cereblon-​DDB1 bound to Ikaros was solved41 (Figure 5.3A). A sequence comparison of the GSPT1, Ikaros and CK1α beta

 101

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

101

Figure 5.3  A beta-​ hairpin turn containing a glycine at a key position mediates neosubstrate recruitment to cereblon by CELMoDs. (A) GSPT1 (blue, PDB ID : 5HXB), Ck1α (orange, PDB ID : 5FQD) and Ikaros (red, PDB ID : 6H0F) bind cereblon through similar beta-​hairpin turns. (B) Amino acid sequence alignment of the GSPT1, Ck1α, and Ikaros structural degrons showing few conserved residues beyond the key glycine.

hairpin loops shows no residues in common besides the glycine (Figure 5.3 B). Recruitment of such sequence dissimilar loops is possible because, for all three neosubstrates, all of the hydrogen bond interactions with cereblon are formed through the main chain carbonyl oxygens. These hydrogen bonds are formed with the side chains of cereblon residues N351, H357 and W400 and mutation of these residues rescues CELMoD-​mediated neosubstate degradation.18,19 Thus, the role of the beta hairpin turn is to present the carbonyl oxygens at the right orientation for critical H-​bonding interactions, regardless of amino acid side chain composition. The ternary complex structures establish a structural mechanism for how CELMoDs are able to recruit structurally and functionally unrelated protein targets. In contrast to other ligases that bind a specific sequence, neosubstrate recruitment to the CELMoD-​cereblon surface is via a structural degron defined as a solvent-​exposed beta hairpin loop with a glycine in the correct register. Implications of these structural studies were twofold: first, the presence of the structural degron in structurally diverse proteins suggested the possibility that other proteins containing the structural degron could be degraded via a molecular glue mechanism, allowing for modulation of the undruggable proteome; and second, that modification of current CELMoDs could allow for a tunable degradation profile, avoiding the destruction of undesired neosubstrates while targeting proteins necessary for the desired therapeutic outcome.

102

102

Chapter 5

5.7 Identification of a Substrate Mediating Teratogenicity Through a Structural Degron Search Following the identification of the structural degron, it became possible to search for this feature in other proteins and potentially identify novel cereblon neosubstrates. Identification of new cereblon noesubstrates could help to further understand the mechanism of action of CELMoDs. A remaining unexplained activity of CELMoDs was the infamous teratogenicity of thalidomide observed in the 1950s and 1960s. When thalidomide was originally developed as a mild sedative, the drug was tested in rodents where no toxic effects were observed even at the highest doses.42 Regrettably, the drug was not tested in additional organisms, nor were the developmental effects of the drug investigated, before it was prescribed to pregnant women as a treatment for morning sickness. Subsequently it was identified that exposure to thalidomide during development caused embryopathies characterized by the loss or shortening of limbs, termed phocomelia, as well as specific ear, eye, heart and kidney defects.2,4,5 None of the previously characterized neosubstrates had known roles in development, and so a search began for established developmental regulators that might be candidates for cereblon degradation. Using knowledge of the structural degron, a candidate-​based approach was utilized to identify potential degron-​containing proteins with links to limb development. Through this profiling of proteins containing the structural degron feature, the embryonic transcription factor SALL4 was identified as a thalidomide-​ dependent cereblon substrate43 (Figure 5.4A). SALL4 contains four potential glycine-​containing degron zinc fingers, and when each was tested for recruitment to cereblon by thalidomide, zinc finger 2 was strongly ubiquitinated in an in vitro ubiquitination system, and ubiquitination was found to be dependent upon the critical glycine. SALL4 is expressed in human iPS cells, and thalidomide treatment induced the degradation of SALL4. This degradation was dependent on cereblon, proteasome and the critical glycine within zinc finger 2. Meanwhile, SALL4 was also identified as a thalidomide-​dependent cereblon neosubstrate in human stem cells from an unbiased proteomics-​based approach, using mass spectrometry to identify potential developmental substrates of cereblon from human embryonic stem cells.44 Loss-​of-​function mutations in SALL4 cause a spectrum of human syndromes, including Duane–​radial ray, Instituto Venezolano de Investiaciones Cientifìcas (IVIC) and acro-​ renal–​ ocular syndromes. These disorders are characterized by forearm malformations as well as ear, eye, kidney and heart defects.45 Importantly, mutations in SALL4 are haploinsufficient, such that even a partial reduction in SALL4 activity is associated with developmental defects. The effects of SALL4 loss-​of-​function mutations have been so similar to thalidomide embryopathy that patients carrying SALL4 mutation have been

 103

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

103

Figure 5.4  The SALL4 zinc finger degron contains species-​ specific amino acid differences. (A) Domain architecture of the SALL4 protein highlighting the zinc finger degron targeted by CELMoDs. (B) Amino acid sequence alignment of human, rabbit and mouse SALL4 zinc finger degrons showing the mouse zinc finger contains five amino acid differences from human and rabbit (conserved amino acids denoted by “*,” semi-​conserved by “:”). (C) Amino acid differences of the mouse cereblon (red) mapped onto the surface of the human cereblon crystal structure (PDB ID : 4TZ4). (D) Amino acid differences of the mouse SALL4 (red) mapped onto a model of the human SALL4 degron zinc finger (model based on PDB ID : 2MA7).

misdiagnosed with thalidomide exposure, only to later understand the genetic basis of the disorder when the syndrome was passed on to their children.46,47 Early toxicity studies established the species specificity of thalidomide teratogenicity, identifying mice and rats as resistant to thalidomide embryopathy while rabbits showed sensitivity and phenocopied the limb defects of humans.42 By showing that murine cereblon was unable to mediate antimyeloma activity, two sequence differences proximal to the thalidomide binding pocket, V388I and E377V, were identified as potential drivers of rodent resistance.16 These sequence variations were subsequently shown to prevent substrate recruitment, with residue V388 critical for IKZF1 and CK1α, and E377 critical for GSPT1 recruitment by CC88519,39 (Figure 5.4C). Rabbit cereblon on the other hand has amino acids that match the human protein at these positions, consistent with conserved sensitivity to teratogenicity. In order to investigate the role of these amino acid differences, a fully humanized cereblon transgenic mouse was developed. This mouse model showed complete resistance to thalidomide embryopathy, even at high

104

104

Chapter 5 –​1

doses of 1 000 mg kg , indicating that there may be species-​specific amino acid differences in the teratogenicity-​causing neosubstrate as well.43 A similar mouse study incorporating only a single amino acid change to the human residue, I391V, identified that this change conferred some sensitivity to thalidomide toxicity, specifically inducing fetal loss,48 although no limb defects were observed. The SALL4 structural degron is conserved between rabbits and humans, but contains amino acid differences in the mouse and rat (Figure 5.4B,D). The ability of thalidomide to induce degradation of SALL4 in animal models was found to correlate with species sensitivity to teratogenicity: SALL4 degradation was observed in the rabbit which is sensitive to teratogenicity, but no SALL4 degradation was observed in the mouse or humanized-​cereblon mouse, which do not exhibit teratogenicity. Finally, SALL4 degradation was observed in the rabbit embryo with thalidomide treatment during the sensitive window of development, and this degradation correlated with the observation of limb defects and gross developmental abnormalities.43 Importantly, when different CELMoDs were assessed for their SALL4 degradation efficiency, SALL4 degradation did not always correlate with degradation of the therapeutic target.43 Among cereblon modulators, molecules that most potently degrade Ikaros and Aiolos, iberomide (CC-​220) and avadomide (CC-​122), showed relatively reduced activity towards SALL4. This important structure–activity relationship (SAR) break between off-​target SALL4 activity and on-​target therapeutic activity provides proof of principle that it will be possible to develop molecules with altered activity spectrums. Structural studies on the cereblon-pomalidomide-SALL4 ternary complex have shown that there is a shift in conformation compared to Ikaros, and should enable rational design of safer degraders.76 Importantly, these studies do not demonstrate that SALL4 degradation is the only cause of thalidomide teratogenicity. Over the years, many hypotheses for teratogenic mechanisms have been put forward, including roles for oxidative DNA damage from thalidomide metabolites as well as thalidomide’s anti-​angiogenic effects, or p63 degradation, and these activities may also contribute in some way to embryopathy.49,50 In addition, as more cereblon neosubstrates are identified we may find additional proteins with developmental regulatory roles that may or may not have contributed to some extent to the thalidomide tragedy, but which may otherwise be worth avoiding in the development of future therapeutics.

5.8 Expansion of Cereblon Neosubstrates In addition to SALL4, multiple new cereblon neosubstrates have been reported over the last two years. These studies have expanded on the work of the three landmark papers published in 2014 that reported Aiolos and Ikaros to be the first examples of cereblon neosubstrates and key efficacy targets of lenalidomide and pomalidomide in multiple myeloma.36–​38 Following these zinc finger (ZnF) transcription factors, more cereblon

 105

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

105

neosubstrates were slowly reported over several years –​CK1α in 2015, GSPT1 in 2016, and ZFP91 in 2017.18,19,51 In 2018 there was a dramatic increase in the number of novel cereblon neosubstrates uncovered. In that year alone, three different papers were published that reported a total of 12 new cereblon neosubstrates.41,43,44 Seeing that Aiolos, Ikaros and ZFP91 were C2H2 zinc finger transcription factors, it was postulated that additional members of the ~800 C2H2 ZnF proteins could also be cereblon neosubstrates.41,43,44 Systematic identification of cereblon neosubstrates among this protein family has been conducted by implementing a high-​throughput fluorescence-activated cell sorting (FACS)-​ based reporter screen.41 Approximately 6500 single C2H2 ZnF domains that possessed a critical glycine residue to function as a CELMoD structural degron were fused to GFP, and a stable cell line expressing these C2H2 ZnF–​ GFP reporters were treated with thalidomide, pomalidomide or lenalidomide. Using an internal mCherry control for normalization, cells that had high GFP signal were then collected via flow cytometry and their DNA was analyzed by next-​generation sequencing. Putative cereblon neosubstrates were identified based on the cDNAs that were underrepresented and hence potentially degraded from the cellular pool following incubation with CELMoDs. This primary screen led to the identification of 11 single C2H2 ZnF domains that appeared to be degraded in the presence of CELMoDs. Next, the authors investigated whether the full-​length proteins from these single C2H2 ZnF domains could also serve as cereblon neosubstrates. Ultimately, this approach uncovered four C2H2 ZnF proteins –​ZNF276, ZNF653, ZNF692 and ZNF827 –​previously unreported to be cereblon neosubstrates.41 Mass spectrometry has also been used to identify a number of new cereblon neosubstrates. Four different human cell lines were treated with thalidomide, pomalidomide, lenalidomide or vehicle control DMSO and global proteome changes were measured using isobaric mass tags and mass spectrometry.44 Consistent with the screening of zinc finger transcription factors, the mass spectrometry experiments likewise suggested that the three C2H2 ZnF proteins ZNF653, ZNF692 and ZNF827 were cereblon neosubstrates. Additionally, these same studies indicated that eight previously unreported proteins –​ SALL4 (also reported by Matyskiela et al.43), RNF166, FAM83F, RAB28, GZF1, ZBTB39, ZNF98 and DTWD1 –​were cereblon neosubstrates. Five of these proteins are members of the C2H2 ZnF family.44 The three cereblon neosubstrate identification papers of 2018 accentuated that C2H2 zinc finger transcription factors are tractable cereblon neosubstrates. This appears to be a direct result of the structural features of the C2H2 zinc finger domain, which contains a β-​hairpin fold. That C2H2 zinc finger proteins are amenable to CELMoD-​mediated degradation is striking because this class of proteins was previously considered undruggable –​many have no small-​molecule binding sites and no catalytic activity. As more CELMoD chemical matter and cereblon neosubstrates are investigated, there may be protein families in addition to C2H2 zinc fingers that are also distinctively

106

106

Chapter 5

structured to be susceptible to CELMoD-​mediated degradation. Moreover, these cereblon neosubstrate studies suggest we may only be at the inception of our understanding of the number of targets tractable by CELMoD-​based targeted protein degradation and hence the disease indications treatable with CELMoD therapies. As the field delves further into new CELMoD targets, robust screening platforms are needed to identify potent and selective chemical matter for CELMoD neosubstrates. This platform has been enabled by knowledge of the mechanism of action of CELMoDs. Whereas in the past new CELMoDs were identified through phenotypic screens, now novel CELMoDs can also be identified via a target-​based approach. Similar to phenotypic screening, target-​based screens can be driven by cell-​based assays. While CELMoD-​ mediated neosubstrate degradation can be monitored by multiple cellular assays including ELISA, mesoscale, homogeneous time-​resolved fluorescence and proteomics, these methods can be expensive, laborious to develop or low-​ throughput. As an alternative to the aforementioned methods, enzyme fragment complementation (EFC) technologies can be implemented to screen CELMoD chemical libraries for compounds that degrade specific targets of interest. EFC assays incorporate two inactive enzyme fragments generally consisting of a small peptide from the reporter enzyme and a larger fragment containing the majority of the reporter enzyme. In building an EFC assay, a small peptide tag from the reporter enzyme is appended to the protein of interest and the larger fragment from the reporter enzyme is added to the cells via a homogeneous assay mixture. Once the two components associate they form an active reporter enzyme whose signal, which is typically luminescence, can serve as a surrogate readout for relative protein levels for a screening target.52,53 Degradation of the protein of interest decreases the amount of the small fragment available to bind to the larger enzyme fragment, leading to a proportional drop in the reporter enzyme signal. These assays can be used to determine both the depth of degradation and potency of CELMoDs for a target in order to profile for potency and selectivity.

5.9 Further CELMoDs in Clinical Development Just as the cadre of cereblon neosubstrates has continued to expand, so too has the number of CELMoDs in clinical stage development (Table 5.1). These newer CELMoDs are building from the impressive foundation established by the transformational early CELMoDs thalidomide, pomalidomide and lenalidomide. Some examples of CELMoDs currently being tested in the clinic are avadomide (CC-​122), iberdomide (CC-​220), CC-​90009 and CC-​92480.14 These newer-​generation compounds are being investigated in indications previously not approved for treatment with CELMoDs, such as systemic lupus erythematosus (SLE). These compounds possess distinct chemical properties from traditional CELMoDs, suggesting the depth and breadth that can be developed is merely in its infancy.

 107

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

107

Table 5.1 CELMoDs in the clinic. Name

Ligase

Principle substrate

CRL4CRBN Ikaros, Aiolos Lenalidomide CRL4CRBN Ikaros, Aiolos, CK1a Pomalidomide CRL4CRBN Ikaros, Aiolos Avadomide CRL4CRBN Ikaros, (CC-​122) Aiolos Iberdomide CRL4CRBN Ikaros, (CC-​220) Aiolos CC-​90009 CRL4CRBN GSPT1 Thalidomide

CC-​92480

CRL4CRBN Ikaros, Aiolos

Principle indication

Regulatory Class status

Myeloma

Approved

Myeloma, del(5q) MDS

Approved

Relapsed refractory Approved myeloma Lymphoma Phase 2 Relapsed refactory Phase 1b/​ myeloma, lupus 2a Acute myeloid Phase 1 leukemia Myeloma Phase 1

Molecular glue Molecular glue Molecular glue Molecular glue Molecular glue Molecular glue Molecular glue

5.9.1 Avadomide Avadomide (CC-​122) is currently being investigated as a treatment for lymphoma in the clinic. This CELMoD contains a cereblon-​binding glutarimide moiety similar to thalidomide, pomalidomide and lenalidomide. Differing from thalidomide, pomalidomide and lenalidomide, avadomide does not contain a phthalimide or isoindolinone core. Instead it has a quinazolinone core with an amino group at the 5 position and a methyl group at position 2.54 Despite its divergent core, avadomide is still a potent degrader of the B-​cell-​ specific zinc finger transcription factors Aiolos and Ikaros. In fact, avadomide shows faster kinetics and greater depth of Aiolos and Ikaros degradation in diffuse large B-​cell lymphoma (DLBCL) cell lines than lenalidomide. In agreement with more robust degradation of Ikaros and Aiolos, avadomide has more potent antiproliferative activity in B cells (~10×) and greater anti-​ angiogenic activity (~100×) compared to lenalidomide. Avadomide treatment causes marked upregulation of the interferon (IFN) response proteins IRF7, IFIT3 and DDX58, whereas lenalidomide treatment causes a minor increase. Upregulation of these IFN genes following avadomide treatment induces apoptosis in both activated B-​cell (ABC) and germinal center B-​cell (GCB) DLBCL cells, while lenalidomide predominantly has activity only in ABC DLBCL.54,55 Again, this more pronounced activity of avadomide in comparison to lenalidomide is predominantly believed to be caused by the depth and speed of Aiolos and Ikaros degradation with the new CELMoD.

5.9.2 Iberdomide Iberdomide (CC-​220) is currently being investigated for relapsed/​refractory multiple myeloma and SLE. This CELMoD contains the cereblon-​binding

108

108

Chapter 5

glutarimide moiety similar to thalidomide, pomalidomide and lenalidomide. It also possesses the same isoindolinone core as lenalidomide. In contrast to lenalidomide, iberdomide has a core extended O-​linked benzyl and morpholino moieties. Iberdomide displays ~20× higher affinity for cereblon in vitro than lenalidomide or pomalidomide and crystallographic experiments revealed that due to the extension off of its core, iberdomide forms more surface contacts with cereblon than lenalidomide. In particular, the phenyl moiety interacts with a groove on the cereblon surface while the morpholino ring is positioned toward a hydrophobic pocket. Structural studies also suggest that iberdomide does not alter the way in which cereblon interacts with neosubstrates. Congruent with these biochemical and structural results are cellular data showing that iberdomide is a more potent and deeper degrader of Aiolos and Ikaros than lenalidomide. Thus, the augmented CC-​220/​cereblon interaction is most likely increasing the fraction of CRL4cereblon able to bind to and ubiquitinate neosubstrates, resulting in enhanced proteasomal degradation.56 Aiolos and Ikaros perform key roles in B-​cell maturation and homeostasis, which is underscored by human genetic findings. Single-​nucleotide polymorphisms (SNPs) in the gene for Aiolos are associated with a higher risk of developing immunological malignancies including rheumatoid arthritis, ankylosing spondylitis and SLE.57–​60 SNPs in the gene for Ikaros are associated with Crohn’s disease, IBD and SLE.61–​63 A single Ikaros SNP has also been reported as an expression quantitative trait locus resulting in enhanced expression of type I IFN genes and downregulation of complement genes, which along with aberrant autoantibody production are hallmarks of SLE.64 These and other compelling data linking Aiolos and Ikaros to SLE precipitated the investigation of iberdomide as a novel therapeutic agent in SLE. Of note, iberdomide abrogates B-​cell precursor differentiation with concomitant downregulation of IgM and IgG production in vitro.65 Peripheral blood mononuclear cells from SLE patients treated with iberdomide resulted in decreased expression of antiphospholipid and anti-​dsDNA autoantibodies. In healthy volunteers, iberdomide dosing reduced B-​cell numbers, caused upregulation of IL-​2 by T cells and inhibited IL-​1β production in response to Lipopolysaccharides (LPS) ex vivo.66 Thus, there is continued interest in developing iberdomide for the treatment of SLE and other immunological indications.

5.9.3 CC-​90009 and CC-​92480 CC-​90009 is in phase 1 clinical trials for treatment of patients with relapsed or refractory AML, and CC-​92480 is being tested in clinical trials for treatment of relapsed, refractory or newly diagnosed multiple myeloma.77

5.10 The Development of Cereblon-​targeting Hetero-​bifunctional Degraders The potential for targeted protein degradation as a therapeutic modality was appreciated well before the mechanism of action of CELMoDs was

 109

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

109

understood. In 2001, a peptide from hypoxia inducible factor 1a, which binds to the VHL substrate recognition subunit of the Cul2-​Rbx1-​EloB/​C-​ VHL E3 ligase, was covalently linked to a high-​affinity target-​binding moiety for the androgen or estrogen receptors.15 When introduced into cells, these PROTACs recruited the androgen or estrogen receptors to the VHL E3 ligase for ubiquitination, demonstrating the hijacking of the UPS to direct the degradation of non-​native substrates using hetero-​bifunctional molecules to be a viable strategy. Since this initial proof-​of-​concept study, considerable energy has been invested in repurposing small molecules that interact with E3 ligase substrate adaptors, including VHL, MDM2 and cIAP, to support targeted protein degradation.67–​69 These efforts have led to the synthesis of numerous small-​molecule hetero-​bifunctional ligands with improved drug-​ like properties that catalyze the degradation of a broad range of therapeutically relevant proteins. The discovery that cereblon is the target of thalidomide expanded the set of known small molecules that bind E3 ligases and enabled the rapid development of hetero-​bifunctional ligands that utilize the cereblon-​CRL4 E3 ligase to drive targeted protein degradation. The first cereblon-​dependent hetero-​ bifunctional ligands contained a thalidomide-​based cereblon-​binding moiety linked to either JQ1 or OTX015, which are high-​affinity ligands for the Bromodomain and Extra-​Terminal domain (BET) family proteins BRD2, BRD3 and BRD4.35,70 These hetero-​bifunctional ligands drove efficient BET degradation and were more efficacious in cells than JQ1 and OTX015 alone, suggesting that targeted protein degradation offers improved efficacy relative to stoichiometric inhibition. Hetero-​ bifunctional molecules have since been used to direct the cereblon-​dependent degradation of numerous additional targets from diverse protein families. The use of cereblon-​ binding moieties may offer some advantages for hetero-​bifunctional drug development, because the glutarimide and isoindolinone or phthalimide moieties are smaller than the ligase-​binding moieties commonly utilized for other E3 enzymes. Expanding the toolbox of ligase-​binding moieties to include cereblon has also enabled the degradation of targets that may not be easily accessed using other E3s. Due to the potential for clash in ubiquitin ligase ternary complex formation, target binding profiles and target degradation profiles for a given hetero-​bifunctional molecule are not necessarily well-​correlated. This is perhaps best illustrated by studies comparing hetero-​bifunctional ligands harboring either a cereblon or VHL ligase binding moiety linked to a promiscuous kinase-​binding moiety, Foretinib.71 When the degradation profiles as determined by mass spectrometry were compared for each degrader, only a fraction of the protein kinases that bind Foretinib were effectively degraded, and degraders that contained VHL-​or cereblon-​binding moieties catalyzed the degradation of very different sets of protein kinases. Accumulating data indicate that linker length, flexibility and trajectory play a crucial role in driving the assembly of a ternary complex that contains the ligase, compound and target of interest. In its simplest form,

110

110

Chapter 5

the ternary complex is assembled through the independent formation of binary complexes between the ligase and its cognate binder on one side and the target and its binding moiety on the other. The proximity of target and ligase in the ternary complex affords the opportunity, however, for direct favorable protein–​protein interactions or steric clash between the ligase and target to shape ternary complex formation. For example, favorable protein–​protein interactions between VHL and BRD4 are suggested to form the basis for positive cooperativity in ternary complex formation, correlating with more effective target degradation in cells.72 Negative cooperativity has, however, been described as a feature of hetero-​bifunctional compounds that target the BET family proteins through cereblon.73 By systematically modulating the features of the linker connecting the cereblon and target binding moieties, selective BRD4 degraders were developed by minimizing negative cooperativity between cereblon and BRD4. Observations that protein–​protein interactions in the ternary complex shape the degradation profile of hetero-​bifunctional ligands in many ways muddies the distinction between molecular glue and hetero-​bifunctional ligands. While these approaches are conceptually different, the reality is that hetero-​bifunctional ligands lie on a continuum between two extremes. On one extreme, the ligase and substrate interact independently with different moieties of the hetero-​bifunctional ligand. On the other, typified by molecular glue, protein–​protein interactions between ligase and substrate form the basis for ternary complex formation, and the ligands have no affinity for the target on its own. However, as CELMoDs become longer and more elaborated, as exemplified by CC-​885 (Figure 5.2), they may pick up more affinity for the target protein alone, increasing their hetero-​ bifunctional nature. Similarly, hetero-​bifunctional degraders have been shown to enable protein–​protein interactions between target and ligase. Thus, CELMoDs and hetero-​bifunctional degraders may be seen on a mixed spectrum of ligand affinities for the target and protein–​protein interactions (Figure 5.5). While optimization of linker features in hetero-​bifunctional molecules is critical, care must be taken during this process when using cereblon as the E3 enzyme. Hetero-​bifunctional molecules using cereblon employ ligase moieties that are highly related to CELMoDs. As a result, neosubtrates like Ikaros, Aiolos, GSPT1 and SALL4 can be common off-​targets for cereblon-​ targeting hetero-​ bifunctional ligands. For example, hetero-​ bifunctional molecules targeting BRD4 that vary only in their linker composition have been shown to have significant differences in off-​target potency against SALL4.43 Furthermore, off-​target degradation of GSPT1 has been described for several hetero-​bifunctional ligands, driving defects in cell proliferation that could complicate data interpretation.74,75 Therefore, any development strategy for hetero-​bifunctional degraders incorporating cereblon binders should ascertain whether the molecules also have activity against cereblon neosubstrates.

 111

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

111

Figure 5.5  A continuum of targeted protein degraders. Hetero-​ bifunctional molecules can be viewed on a spectrum with molecular glue molecules that vary in the extent of protein–​protein interactions (PPI) and affinity of the ligand for the target alone.

5.11 Differences Between Hetero-​bifunctional and Scaffolded Protein–​Protein Interaction Ubiquitin Ligase Modulators While ultimately resulting in targeted protein degradation, the mode of action for CELMoDs or “molecular glue” are distinct from hetero-​bifunctional degraders. These differences are important to note while developing an optimization strategy and considering pharmacology. Setting a precedent for protein degradation, CELMoDs have shown their clinical relevance, and have the advantages of low molecular weight and drug-​like properties. The molecular glue approach, as typified by CELMoDs, offers the possibility of recruiting and degrading proteins for which there are no ligands, but requires specific complementarity between proteins and ligand. Hetero-​bifunctional molecules offer no end to the possibilities of compounds that can be linked, but these molecules may require optimization for substrate and ligase compatibility. Adding additional properties from the linker between them in hetero-​ bifunctional degraders adds extra challenges to this modality. Each of these approaches may provide strategic advantages depending on the specific target and available tools.

5.12 Conclusions CELMoD mechanism of action has been understood only recently, establishing a new paradigm in drug discovery with the potential to target proteins that lack traditional ligand-​binding sites. At the same time, understanding the CELMoD

112

112

Chapter 5

molecular mechanism has helped elucidate many historical aspects of early CELMoDs thalidomide, lenalidomide and pomalidomide. The structural degron identified in cereblon neosubstrates has enabled an expansion into additional neosubstrates that new CELMoDs will target for future therapies. Targeted protein degradation through cereblon-​CRL4 promises that the scope of targets previously considered to be “undruggable” will continue to expand.

References 1. P. P. Chamberlain and L. E. Hamann, Nat. Chem. Biol., 2019, 15, 937–944. 2. W. Lenz, Teratology, 1988, 38, 203–​215. 3. L. Lasagna, J. Chronic Dis., 1960, 11, 627–​631. 4. W. P. Lenz, R. A. Pfeiffer, W. Kosenow, D. J. Hayman, Lancet, 1962, 279(7219),  45–​46. 5. W. G. Mcbride, Lancet, 1961, 278(7216), 1358. 6. J. Sheskin and F. Sagher, Int. J. Lepr. Other Mycobact. Dis., 1971, 39, 585–​588. 7. T. Ito, et al., Science, 2010, 327, 1345–​1350. 8. S. Angers, et al., Nature, 2006, 443, 590–​593. 9. A. Lopez-​Girona, et al., Leukemia, 2012, 26, 2326–​2335 10. Y. X. Zhu, et al., Blood, 2011, 118, 4771–​4779. 11. N. Zheng and N. Shabek, Annu. Rev. Biochem., 2017, 86, 129–​157. 12. D. Finley, Annu. Rev. Biochem., 2009, 78, 477–​513. 13. C. E. Berndsen and C. Wolberger, Nat. Struct. Mol. Biol., 2014, 21, 301–​307. 14. P. P. Chamberlain and L. G. Hamann, Nat. Chem. Biol., 2019, 15, 937–​944. 15. K. M. Sakamoto, et al., Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 8554–​8559. 16. P. P. Chamberlain, et al., Nat. Struct. Mol. Biol., 2014, 21, 803–​809. 17. E. S. Fischer, et al., Nature, 2014, 512,  49–​53. 18. G. Petzold, E. S. Fischer and N. H. Thoma, Nature, 2016, 532, 127–​130. 19. M. E. Matyskiela, et al., Nature, 2016, 535, 252–​257. 20. X. Tan, et al., Nature, 2007, 446, 640–​645. 21. S. Jackson and Y. Xiong, Trends Biochem. Sci., 2009, 34, 562–​570. 22. D. M. Koepp, et al., Science, 2001, 294, 173–​177. 23. S. J. Elledge and J. W. Harper, Biochim. Biophys. Acta, 1998, 1377, M61–​70. 24. P. H. Maxwell, et al., Nature, 1999, 399, 271–​275. 25. A. A. Salahudeen, et al., Science, 2009, 326, 722–​726. 26. A. Varshavsky, Protein Sci., 2011, 20, 1298–​1345. 27. H. C. Lin, et al., Mol. Cell, 2018, 70, 602–​613. 28. J. Jin, E. E. Arias, J. Chen, J. W. Harper and J. C. Walter, Mol. Cell, 2006, 23, 709–​721. 29. L. A. Higa, et al., Nat. Cell Biol., 2006, 8, 1277–​1283. 30. Y. J. He, C. M. McCall, J. Hu, Y. Zeng and Y. Xiong, Genes Dev, 2006, 20, 2949–​2954. 31. D. M. Duda, et al., Cell, 2008, 134, 995–​1006.

 113

New Activities of CELMoDs, Cereblon E3 Ligase-modulating Drugs

113

2. E. S. Fischer, et al., Cell, 2011, 147, 1024–​1039. 3 33. G. Lu, et al., eLife, 2018, 7, doi:10.7554/​eLife.40958 34. Q. L. Sievers, J. A. Gasser, G. S. Cowley, E. S. Fischer and B. L. Ebert, Blood, 2018, 132, 1293–​1303. 35. G. E. Winter, et al., Science, 2015, 348, 1376–​1381. 36. G. Lu, et al., Science, 2014, 343, 305–​309. 37. J. Kronke, et al., Science, 2014, 343, 301–​305. 38. A. K. Gandhi, et al., Br. J. Haematol., 2014, 164, 811–​821. 39. J. Kronke, et al., Nature, 2015, 523, 183–​188. 40. L. B. Sheard, et al., Nature, 2010, 468, 400–​405. 41. Q. L. Sievers, et al., Science, 2018, 362, doi:10.1126/​science.aat0572 42. I. D. Fratta, E. B. Sigg and K. Maiorana, Toxicol. Appl. Pharmacol., 1965, 7, 268–​286. 43. M. E. Matyskiela, et al., Nat. Chem. Biol., 2018, 14, 981–​987. 44. K. A. Donovan, et al., eLife, 2018, 7, doi:10.7554/​eLife.38430. 45. J. Kohlhase, et al., Hum. Mutat., 2005, 26, 176–​183. 46. J. Kohlhase, et al., J. Med. Genet., 2003, 40, 473–​478. 47. J. Kohlhase and L. B. Holmes, Birth Defects Res. A Clin. Mol. Teratol., 2004, 70, 550–​551. 48. E. C. Fink, et al., Blood, 2018, 132, 1535–​1544. 49. T. Parman, M. J. Wiley and P. G. Wells, Nat. Med., 1999, 5, 582–​585. 50. T. Asatsuma-​Okumura, et al., Nat. Chem. Biol., 2019, 15, 1077–​1084. 51. J. An, et al., Nat. Commun., 2017, 8, 15398. 52. F. H. Buttner, Expert Opin. Drug Discovery, 2006, 1, 373–​378. 53. K. Oh-​Hashi, E. Furuta, K. Fujimura and Y. Hirata, Biochem. Biophys. Rep., 2017, 12,  40–​45. 54. P. R. Hagner, et al., Blood, 2015, 126, 779–​789. 55. L. H. Zhang, et al., Br. J. Haematol., 2013, 160, 487–​502. 56. M. E. Matyskiela, et al., J. Med. Chem., 2018, 61, 535–​542. 57. F. A. Kurreeman, et al., Am. J. Hum. Genet., 2012, 90, 524–​532. 58. C. J. Lessard, et al., Am. J. Hum. Genet., 2012, 90, 648–​660. 59. R. Qiu, et al., Scand. J. Rheumatol., 2013, 42, 469–​472. 60. J. Zhang, Q. Zhu, F. Meng, H. Lei and Y. Zhao, Gene, 2014, 533, 385–​388. 61. A. Franke, et al., Nat. Genet., 2010, 42, 1118–​1125. 62. J. W. Han, et al., Nat. Genet., 2009, 41, 1234–​1237. 63. L. Jostins, et al., Nature, 2012, 491, 119–​124. 64. H. J. Westra, et al., Nat. Genet., 2013, 45, 1238–​1243. 65. Y. Nakayama, et al., J. Immunol., 2017, 199, 2388–​2407. 66. P. H. Schafer, et al., Ann. Rheum. Dis., 2018, 77, 1516–​1523. 67. A. R. Schneekloth, M. Pucheault, H. S. Tae and C. M. Crews, Bioorg. Med. Chem. Lett., 2008, 18, 5904–​5908. 68. Y. Itoh, M. Ishikawa, M. Naito and Y. Hashimoto, J. Am. Chem. Soc., 2010, 132, 5820–​5826. 69. D. P. Bondeson, et al., Nat. Chem. Biol., 2015, 11, 611–​617. 70. J. Lu, et al., Chem. Biol., 2015, 22, 755–​763.

114

114

1. 7 72. 73. 74. 75. 76. 77.

Chapter 5

B. E. Smith, et al., Nat. Commun., 2019, 10, 131. M. S. Gadd, et al., Nat. Chem. Biol., 2017, 13, 514–​521. R. P. Nowak, et al., Nat. Chem. Biol., 2018, 14, 706–​714. M. Ishoey, et al., ACS Chem. Biol., 2018, 13, 553–​560. J. Yang, et al., J. Med. Chem., 2019, 62, 9471–​9487. M. E. Matyskiela, et al., Nat. Struct. Mol. Biol., 2020, 4, 319–322. J. D. Hansen, et al., J. Med. Chem., 2012, 13, 6648–6676.

 115

CHAPTER 6

Structure-​based PROTAC  Design DARRYL B. MCCONNELL* Boehringer Ingelheim RCV GmbH & Co KG, Doktor-​Boehringer-​Gasse 5–​11, 1120 Vienna, Austria *Email:  darryl.mcconnell@boehringer-​ingelheim.com

6.1 Introduction In his seminal manuscript conceptualizing structure-​ based drug design (SBDD) in the late 1980s, Ripka stated that “Conceptually, the idea of constructing a molecule that represents a geometric and electrostatic fit to a well-​ defined enzyme site would not necessarily appear to be a difficult problem.”1 Despite the solution to the problem being equally simply expressed, namely “a molecule with bond lengths and angles, defined within specific limits, whose surface is a close complement of the receptor defined surface,” the problem remains unsolved more than 30 years after the birth of SBDD. At the center of SBDD is the assumption that being able to see drug ligand binding to its target in three dimensions will enable better and faster design of drug molecules. Given the significantly increased size and complexity in designing PROTACs, being able to see how PROTACs bind to their targets, structure-​based PROTAC design (SBPD) should have an even more dramatic impact. In the ensuing 10 years, SBDD emerged as the combination of three technologies, X-​ray crystallography, computer graphics and computational methods.2–​5 Contrary to popular belief, over the last 30 years the deposition of 3D protein structures into the Protein Data Bank (PDB)6 have not increased exponentially but linearly,7 at a rate of around 400 structures per year since 1990 Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

115

116

116

Chapter 6

Figure 6.1  Linear increase in new protein structures and new publications on structure-​based drug design.

preceded by a phase of no significant growth (see Figure 6.1). This inflection point in 1990 in protein X-​ray crystallography is arguably attributed to the scientific committee flocking to the concept of SBDD following the publications of the HIV protease drugs Agenerase and Viracept8,9 and the neuraminidase inhibitor drug Relenza for influenza from von Itzstein et al.10,11 Proteolysis-​targeting chimeras (PROTACs) are an emerging class of drug molecules wherein a target-​binding ligand linked covalently to an E3 ligase binding ligand (Figure 6.2) stabilizes the protein–​protein interaction (PPI) between the target and the E3-​ligase.12–​14 This PPI stabilization leads to the catalytic ubiquitination of the target protein and subsequent degradation by the proteasome. The initial designing of binders for the target and E3-​ligase proteins rests in the domain of classical SBDD. The optimization of the linker to ideally stabilize the PPI and maximize the catalytic enhancement of target protein ubiquitin-​driven degradation represents a conceptually new step in small-​molecule drug design. In general, this entails choosing the attachment points for both ligands, the linker length and linker atomic composition, which has been coined linkerology.15 Although PROTACs are schematically often represented in a barbell depiction, binding to two spatially separated

 117

Structure-based PROTAC Design

117

Figure 6.2 General architecture of PROTACs.

proteins, PROTACs actually form a compact target–​PROTAC–​ligase ternary structure with the PROTAC occupying one pocket formed between the two proteins. Linkerology, which has the goal of optimizing the PPI stabilization and tuning physico-​chemical properties, can be dramatically accelerated using SBPD. However, as will be discussed later, this is not the only aspect of PROTAC design that can be accelerated with SBPD. Since the early successes with HIV protease and neuraminidase, SBDD has become an integral approach utilized in many drug discovery programs. Fischer’s key and lock analogy16 of an enzyme substrate (the key) fitting perfectly into a pocket on the surface of the enzymatic protein (the lock) remains as applicable to small-​molecule drug discovery today as it was to enzymology over 100 years ago. Being able to see the “key” fitting in to the “lock” in three dimensions at atomic resolution assists drug design in improving affinity, particularly for difficult to drug pockets, as well as optimizing selectivity over other related proteins. Similarly, structure-​based PROTAC design promises to expedite the optimization of potency and selectivity of PROTACs.

6.2 PROTAC Design –​Differences to Small Molecules Small-​molecule and PROTAC drug design both require the fashioning of “keys” for the “locks” of choice. PROTAC drug design requires the second step of bringing the two “locks” together (see Figure 6.3) to form a ternary complex. This stabilization of two proteins creates PPIs which if complementary can increase the affinity of the ternary complex versus the binary complexes. This is termed cooperativity and is the ratio of the binary over the ternary dissociation constants. Ternary complexes are particularly important because these PPIs play a key role in giving PROTACs the potential to be more potent and selective than classical small-​molecule inhibitors. The stability of the formed PPI following PROTAC binding is key to the efficient ubiquitination of a lysine on the protein of interest and subsequent proteasomal degradation. While the first step of designing binders to the E3-​ligase and protein of interest (POI) is common to small molecules and PROTACs, because of the additional affinity gained by positively cooperative ternary complex formation, less potent binders can be used for PROTAC design versus small-​molecule inhibitors. The design process of optimizing the neo-​PPI formed upon PROTAC binding has been coined linkerology.15 Linkerology aims to explore the space of

118

118

Chapter 6

Figure 6.3 The two steps of PROTAC design.

possible PPIs formed between a given protein of interest and E3-​ligase with the goal of optimizing affinity, selectivity and degradation. This exploration is achieved through varying the linker attachment points, linker length and linker atomic composition and is constrained by the binding pockets chosen on the POI and E3-​ligase. The optimization of high-​affinity ligands for proteins with hard-​to-​drug pockets such as that of the BCL2 PPI inhibitor Venetoclax is unlikely to have been possible without the application of SBDD and fragment-​based drug design.17–​19 The large, flat pocket on BCL2 spanning 756 Å2 is more akin to a surface rather than a pocket20 and reminiscent of the pockets into which PROTACs bind. The MZ1 pocket formed by BRD4BD2 and VHL spans 436 Å2 with a volume of 295 Å3. PROTACs have a lesser requirement for binding affinity when compared with inhibitors and as such ligand design can be completed earlier for hard-​ to-​ drug pockets, allowing cooperativity and degradation potency to be optimized with the second PROTAC design step linkerology. The optimization of selectivity of drug candidates is also a crucial design step in discovering well-​tolerated medicines. Nowhere is this more important than in the fruitful area of kinases drug discovery. The impressive number of kinase drugs approved by the FDA21 continues with nine kinase drugs approved by the FDA in 2018,22 all of which can be traced backed to SBDD. While potency is readily attained for kinase inhibitors binding to the ATP pocket, selectivity versus the human kinome of 520 kinases23 represents a formidable challenge. Also for selectivity, linkerology can be applied to obtain

 119

Structure-based PROTAC Design

119

24

selectivity for PROTACs rather than ligand optimization as is needed for classical small-​molecule inhibitors. Given that the second unique PROTAC design step of PPI stabilization through linkerology is key to discovering potent and selective PROTACs, it stands to reason that obtaining the crystal structures of these complexes will accelerate PROTAC drug discovery just as it continues to do for classical small-​ molecule drug discovery.

6.3

Structure-​based Linkerology

At the time of writing, almost all targets for which PROTACs have been described as well as the E3-​ligases leveraged for PROTACs were structurally enabled. This clearly eases the first design step of linkerology, namely the choice of attachment points for the linker. Solvent-​exposed atoms of the ligand in the bound state which avoid loss of binding to the target through steric clash can be easily selected in the presence of co-​crystal structures. In the absence of structural enablement, additional synthetic efforts are required to establish structure–​activity relationships for linker attachment points. In order to systematically explore the PPI stabilization space possibly formed by PROTACs it is important to use this structural information to choose attachment points which are spread across the solvent-​exposed cross-​ section of the binding site in question. In order to achieve this, modification of the ligand structure or the introduction of new structural classes of ligands is required. Thus far, this has been explored only sparsely for given POIs, likely limiting the sampling of the total PPIs possible for a given POI–​E3-​ ligase pair. For example, of the some 300 VHL-​based BRD4 PROTACs reported in the scientific and patent literature at the time of writing, only three linker attachment vectors using four classes of BRD4 ligands have been reported (see Figure 6.4A). This lack of exploration is most likely explained by the high organic synthesis efforts required to synthesize ligands with novel attachment points, but remains an important aspect in comprehensively exploring the PPIs stabilized by PROTACs. The efforts to explore a broad range of attachment points of binders for a given E3-​ligase are no less labor-​intensive than for the binders to the protein targeted for degradation, but as each target of interest is typically conjugated with available E3-​ligase binders, by definition significantly more synthetic and medicinal chemistry efforts are put into E3-​ligase binders. Hence the attachment points for E3-​ligase binders would be expected to be more comprehensive over time. However, at the time of writing from the 400 VHL-​ based PROTACs reported, only four attachment points have been utilized (Figure 6.4B).25–​28 Of these four attachments, by far the majority of VHL-​based PROTACs have been explored using the terminal acetyl capping group. Clearly with the expansion of attachment points on both ligands to both POIs and E3-​ ligases the exploration of the percentage of the total possible number of PPIs possible will increase dramatically.

120

120

Chapter 6

Figure 6.4 Precedented attachment points for (A) BRD4 binders and (B) VHL binders (co-​crystal structures used for depiction for A: JQ1 –​PDB ID 3mxf,29 IBET 295 –​PDB ID 4clb30 and CF53 –​PDB ID 6C7R31 and B: PDB ID 5NVX32).

6.4

Learning from PPI Stabilization

Protein–​protein interactions are important for essentially all biological processes, with the number of interactions in humans estimated to be >500 000.24 This large number of PPIs, which are at the same time highly specific, suggests that there is something different about how proteins interact with each other compared to small-​molecule “key and lock” interactions. While deep, lipophilic pockets are conducive to the tight binding of small molecules, PPIs contain smaller, shallower lipophilic “hot spots” that typically cluster at the center of the PPI, which have been suggested to be mainly responsible for binding energy.20,33 However, polar residues are increasingly being considered to be responsible for not only specificity of association but also complex stabilization. Charged amino acids at protein interfaces can also significantly increase the association rates of two proteins, which might be an important aspect in the study of PROTAC-​induced ubiquitination kinetics. The residue composition at PPI interfaces resembles that of protein surfaces,34,35 averaging approximately 1600 Å2 (800 Å2 per monomer) in size. The interfaces contain around 10 intermolecular H-​bonds with every third H-​bond involving a charged amino acid.36 Although a hitherto neglected field compared to PPI inhibition, PPI stabilization has been investigated prior to the rise of PROTACs.37–​42 Arguably the most prominent examples of a PPI stabilization is that of the natural product macrolides, FK506 which stabilizes the PPI interactions between FKBP12 and calcineurin and rapamycin which stabilizes the interaction between FKBP12 and mTOR.43,44 In analogy to PROTACs and of similar molecular weight, FK506 and rapamycin form a binary complex with FKBP12 which then associates in a ternary complex with calcineurin or mTOR, respectively. X-​ ray crystallography of the ternary FKBP12–​ FK506–​ calcineurin complex45,46 and the FKBP12–​rapamycin–​mTOR complex47 shows that one binding pocket is formed by the two proteins (Figure 6.5B). As we shall see later, this is the same situation for PROTACs, with PROTACs binding to one composite pocket that is formed at the interface between two proteins and not two separate, independent pockets as it is often depicted.

 121

Structure-based PROTAC Design

121

Figure 6.5  FKBP12–​ rapamycin–​ mTOR ternary X-​ ray structure depicting (A) the key interactions formed and (B) the composite pocket formed between FKBP12 and mTOR.

Interactions in such ternary complexes can be formed between the ligand and either protein partner and between the two protein partners themselves. Rapamycin binds to the FKBP–​rapamycin-​binding (FRB) domain of mTOR with a moderate binary affinity of 26 µM, but in the presence of FKBP12 the ternary affinity to mTOR is 12 nM, a 2 000-​fold improvement in potency.48 This additional factor in potency is termed cooperativity or alpha-​value49,50 and is an important aspect of SBPD in order to design stable ternary complexes. In the case of the FKBP12–​rapamycin–​mTOR complex,47 additional direct and water-​mediated H-​bonds between four amino acids from FKBP12 and four amino acids from mTOR close the top and bottom of the composite binding pocket (Figure 6.5A) are formed. These additional PPIs explain the significant increase in ternary complex potency but are not sufficient in the absence of rapamycin to be detected. This PPI stabilization with cooperative binding is also at the core of structure-​based PROTAC design. PPI stabilization is not limited to large natural product macrocycles. The immunomodulatory drugs (ImIDs) thalidomide and its analogues pomalidomide and lenalidomide not only stabilize PPIs but do so with considerably less molecular weight (lenalidomide MW = 259 Da versus rapamycin MW = 914 Da). In the case of the ImIDs this protein stabilization also induces protein degradation; this class of compounds is broadly referred to as molecular glues. ImIDs bind to cereblon (CRBN), the substrate-​recognition subunit of the CRL4ACRBN E3 ligase complex.51,52 ImIDs stabilize ternary complexes

122

122

Chapter 6

Figure 6.6 CRBN–​lenalidomide–​CK1α ternary structure depicting (A) the key interactions formed and (B) the composite pocket formed between CRBN and CK1α.

between CRBN and neo-​substrate proteins such as Ikaros (IKZF1), Aiolos (IKZF3), CK1α, GSPT1 and SALL4, leading to their degradation.51–​56 The ternary X-​ray crystal structure of the adapter protein DDB1 and CRBN in complex with lenalidomide and the neo-​substrate CK1α57 shows again a composite binding pocket (Figure 6.6B) with interactions between lenalidomide and both CRBN and CK1α as well as interactions between the two proteins themselves. As observed for the rapamycin-​induced ternary structure, the CRBN–​lenalidomide–​CK1α ternary structure shows a network of direct and water-​mediated polar interactions between five amino acids from CRBN and five amino acids from CK1α (Figure 6.6A).

6.5

VHL PROTAC Ternary Complexes

Gadd et al.28 provided the first structural insights into PROTAC ternary complexes with the X-​ray co-​crystal structure of the PROTAC MZ158 with VHL and the second bromodomain of BRD4. MZ1 is a nanomolar degrader of BET proteins which cooperatively binds to the BD2 domains of BRD3 and BRD4 (α = 18 for BRD4BD2) and forms stable complexes (ΔG(binary + ternary) = − 22.2 ± 0.1 kcal mol–​1 for BRD4BD2). MZ1 binds to a large, bowl-​shaped pocket with a total buried surface area for the ternary complex of greater than 2600 Å2 (Figure 6.7A). This is significantly larger than the total buried surface area in the CRBN–​ lenalidomide–​CK1α structure, which is estimated at around 1830 Å2.

 123

Structure-based PROTAC Design

123 BD2

Six PPIs between VHL and BRD4 are formed in the PROTAC ternary complex (Figure 6.7B). These interactions form the base of the bowl and are a major contributor to the cooperativity of MZ1. A cluster of three electrostatic interactions between R108 from VHL and E383 and D381 from BRD4 are most likely the major PPI contributors to stabilizing the ternary complex. Two salt bridges are formed between the carboxylic acids of D381 and E383 and the guanidine of R108. E383 forms an additional H-​bond with the backbone NH of R108 (Figure 6.8B). In addition, histidine 110 forms two interactions with BRD4. Further interactions are formed between the backbone NH of H110 and the backbone carbonyl of A384 and a CH–​π interaction between the one of the methyl groups of L385 and the imidazole C4 of H110 (Figure 6.8A). A second CH–​π interaction makes up

Figure 6.7 VHL–​MZ1–​BRD4BD2 ternary structure depicting (A) MZ1 binding to the composite pocket formed between VHL and BRD4 and (B) the key interactions formed and (B) the composite pocket formed.

Figure 6.8 Protein–​protein interactions formed between VHL and BRD4 in the VHL–​ MZ1–​BRD4BD2 ternary complex.

124

124

Chapter 6

the sixth PPI formed in the complex between the methylene group of P71 and benzene ring of W374 (Figure 6.8A). This pattern of largely polar interactions surrounding the PROTAC binding site is reminiscent of Wells’20 original characterization of how PPIs are principally composed. Numerous interactions between the PROTAC itself and the two proteins are formed. It is instructive to separate these interactions into those maintained from the individual binary ligand–​protein complexes and those interactions which are newly formed as a consequence of ternary complex formation. It is these newly formed interactions which are more pertinent to SBPD because they can contribute to increased ternary complex stability. Such additional interactions can come from the ligand portions of the PROTAC or the linker itself and can be iteratively designed when three-​dimensional structural data are available. Both the VHL-​binding portion of MZ1 and the BRD4-​binding element replicates the binding mode of the parent ligands VH032 and JQ1, respectively. While the arylthiazole moiety of the VHL binder is in close proximity to the BRD4 protein, no additional complementary interactions are formed between the VHL binder and BRD4. Using this structural information, future VHL-​based BRD4 PROTACs could be designed whereby the arylthiazole moiety of the VHL binder is modified to specifically interact cooperatively with BRD4. On the other hand, the BRD4 binder forms one additional interaction with VHL. A CH–​π interaction is formed between the thiophene methyl group of the JQ1-​ligand and C2 of histidine 110 (Figure 6.9A). The fact that only one additional interaction is formed by ligand portions of MZ1 highlights in general the opportunity to further increase ternary complex stability of PROTAC molecules using SBPD. Obtaining such ternary X-​ray structures thus allows the rational, structural modification of the E3-​ligase or protein of interest binder to form additional interactions to the respective partner protein, something rarely done in PROTAC design thus far. The solvent-​exposed ethylene glycol linker of MZ1 does not form any interactions with either protein. Two of the three oxygen atoms of the linker are,

Figure 6.9 Ternary complex induced PROTAC ligand interactions between (A) the BRD4 binding portion of MZ1 with VHL and (B) the linker with water molecules and the BRD4 binding portion of MZ1.

 125

Structure-based PROTAC Design

125

however, solvated by one water molecule (Figure 6.9B). Knowing whether the linker of a PROTAC bound in a given ternary complex is solvent-​exposed or -​buried is important, as it determines whether polar or lipophilic linkers, respectively, need to be designed. This is very much akin to classical SBDD where solvation of the solvent exposed portions of the ligand is pursued.59 It is also interesting to note that one CH2 group of the linker is involved in an intramolecular CH–​π interaction with the p-​chlorophenyl ring of the JQ1 portion of the PROTAC further stabilizing the linker conformation in the observed bound state (Figure 6.9B). Farnaby et al.60 provided the second VHL–​PROTAC example of SBPD with the PROTAC ACBI-​1 and the bromodomain-​containing proteins SMARCA2 and SMARCA4. ACBI-​1 is a potent degrader of SMARCA2, SMARCA4 and PBRM1 and induces antiproliferative effects and cell death caused by SMARCA2 depletion in SMARCA4 mutant cancer cells. The ternary X-​ray crystal structures of a close analogue of ACBI-​1 (PROTAC-​2), VHL and both SMARCA2 (PDB code 6HAX) and SMARCA4 (PDB code 6HR2) were solved. The structure of PROTAC-​2 with VHL and SMARCA2, as for MZ1 with BRD4, forms a bowl-​shaped pocket also forming six PPIs between the two proteins (Figure 6.10A,B). Five of these interactions are direct and one is water-​mediated. The majority of these interactions are again H-​bonds and are

Figure 6.10 Ternary X-​ray crystal structure of VHL, PROTAC-2 and the bromodomain of SMARCA2. (A) and (B) show the interactions between the two protein partners VHL and SMARCA2 and (C) and (D) highlight the newly formed interactions between the PROTAC and the protein partners.

126

126

Chapter 6

likely to be major contributors to ternary complex cooperativity and stability. Histidine 110 from VHL is again involved, this time with a water-​mediated interaction with N1464 of SMARCA2. All other PPIs utilize different VHL amino acids versus those observed for MZ1 and BRD4, highlighting the ability of VHL to ultilize different epitopes to form stable PROTAC ternary complexes. R69 of VHL forms two H-​bonds to the backbone carbonyls of F1463 and T1462. Y112 and Q73 from VHL also form H-​bonds to the backbone carbonyls of N1464 and G1467, respectively. F91 and E1420 are in ideal proximity for a T-​stacked π-​interaction. No additional interactions are formed in the ternary complex by the SMARCA2 binding portion beyond its established binary binding mode (Figure 6.10C), again highlighting the untapped potential in the thus far designed PROTACs. This is in contrast to the VHL-​binding motif of PROTAC-​2, which is involved in two additional interactions with SMARCA2 (Figure 6.10D). A methylene group from the fluorocyclopropyl amide group forms a CH-​π interaction with F1463, likely to be an important contributor to complex stability. In addition, the hydroxyproline amide carbonyl forms a water-​mediated interaction with N1464. The first X-​ray structure obtained by the authors was with a PROTAC containing a flexible aliphatic linker (PDB code 6HAY) which came into proximity of the aromatic residue Y98 on VHL. The authors utilized this information to introduce a benzene ring into the linker to form a T-​stacked PPI (Figure 6.10C), resulting in a fourfold improvement in ternary complex dissociation complex and a 35-​fold improvement in degradation as measured by DC50. This highlights how structural information can be utilized to efficiently optimize the potency for PROTACs.

6.6

Cereblon PROTAC Ternary Complexes

Two ternary X-​ ray crystal structures for cereblon-​ based BRD4 PROTACs (dBET27 : PDB code 6BN7 and dBET6 : PDB code 6BOY) by Nowak et al. have also been published.61 In order to gain insights into the general principles of SBPD it is interesting to compare both the VHL-​based BRD4 PROTAC structures and also the cereblon ImID ternary structures. It should be noted that only modest resolutions were obtained for the dBET27 and dBET6 structures (3.5 and 3.3 Å, respectively), which is not sufficient to observe the involvement of water molecules as seen for the aforementioned higher-​resolution structures. Cereblon has three domains: the N-​terminal domain (NTD), the helical-​ bundle domain (HBD) and the C-​terminal domain (CTD), which contains the thalidomide-​binding pocket.52 Despite having radically different linker attachment points, interestingly both small-​molecule degraders dBET6 and dBET23 form the same PPI epitope between cereblon and the first bromodomain of BRD4. As expected, the binding portions of the two degraders also occupy the canonical binding sites on CRBN and BRD4BD1. The αC-​helix of BRD4BD1 (residues 145–​161) interacts at two points with the NTD of CRBN. The

 127

Structure-based PROTAC Design

127

first is a polar interaction from D145 at the base of the αC-​Helix and H103 on CRBN. The second is a hydrophobic loop on CRBN (residues 147–​154) which interacts further up the αC-​Helix at A152 (Figure 6.11A). The ZA loop of BRD4 (residues 76–​104)62 also interacts with CRBN, forming a complex network of H-​bonds and π-​interactions (Figure 6.11B). Q78, F79 and W81 from BRD4 form the core of this network. Q78 forms two H-​bonds to K156 and Q100 of cereblon, F79 form a T-​stacked π-​interaction with F150 of CRBN and W81 forms a H-​bond to the imidazole of H353 on CRBN. It is intriguing that overall dBET6 and dBET23 show a negative cooperativity of around 0.5 despite numerous neo-​PPIs that are formed between CRBN and BRD4. This suggests that significant unfavorable interactions are also formed by the PROTACs themselves. Both PROTACs contain a highly flexible C8 methylene chain which would cause a significant loss of entropy upon binding. The first examples of linker rigidification based on ternary PROTAC X-​ray structures with a macrocyclic VHL-​based BRD4 PROTAC have now been published.63 In addition to the entropic penalty, the lipophilic linker of dBET6 is solvent-​exposed, which would be expected to suffer a significant solvation penalty. The phthalimide ring of dBET6 does form two T-​stack π-​ interactions with W81 of BRD4, which should contribute positively to cooperativity (Figure 6.12A). Due to a slight change in binding orientation, dBET23 no longer forms this π-​interaction with W81, and although the aliphatic linker adopts a completely different position, it remains largely solvent-​exposed

Figure 6.11 CRBN–​dBET6–​BRD4BD1 ternary X-​ray crystal structure highlighting the PPIs formed between CRBN and BRD4BD1.

128

128

Chapter 6

Figure 6.12 CRBN–​dBET6–​BRD4BD1 ternary X-​ray crystal structure of dBET6 (A) and dBET23 (B) with CRBN and BRD4BD1.

(Figure 6.12B). These examples highlight the need to not only monitor binding modes to ensure that positive interactions are maintained during optimization but also the need to design solvation elements in solvent-​exposed PROTAC linker regions.

6.7

Identifying the Right PPI to Target

Non-​specific interactions between any two proteins are widespread and are thought to occur with affinities of greater than 10 mM.64 Because of this relative low energy of formation of PPIs, PROTACs are able to induce multiple PPI orientations. This gives PROTACs the ability to achieve high levels of cooperativity and selectivity, but also presents a significant challenge to SBPD. Not only does the PROTAC need to be optimized for the target-​ligase pocket, but prior to this the “best” PPI orientation needs to be identified for which the PROTAC should be optimized. The “best” PPI orientation could be one which has the highest potential to maximize cooperative and hence potency or one which provides selectivity over an important off-​target. The various PPIs are accessed through variations in the linker and linker attachment point (linkerology) and not the binder portions of the PROTAC architecture. Different PPIs can provide differing levels of cooperativity, the amount of binding affinity gained (positive alpha values) or lost (negative alpha values) between the binary and ternary PROTAC complexes. This

 129

Structure-based PROTAC Design

129

Figure 6.13  Numerous PPIs observed for VHL–​ PROTAC–​ SMARCA2BD X-​ray structures. The VHL–​ElonginB/​C complex is depicted in gray and SMARCA2BD is depicted in orange.

varying cooperativity can provide both significant increases in potency and also increased selectivity.61,65 Nowak et al. observed different PPIs stabilized by cereblon-​based BRD4 PROTACs which they have attributed to differing selectivities between BRD4-​BD1 and BRD4-​BD2 due to the PPI interface for certain complexes overlapping with regions of high dissimilarity between the two domains. In our work on VHL-​based SMARCA2 PROTACs,60 we have observed numerous different PPI epitopes between VHL and the bromodomain of SMARCA2, which in some cases differ enormously from each other (unpublished work; Figure 6.13). These different orientations of SMARCA2 relative to VHL are caused largely by changes in the linker attachment points and linker composition and length. Given the current technical challenges to generate ternary PROTAC X-​ray structures in a high-​throughput fashion and the apparent high number of orientations possible, technological advancement is needed to identify the “best” ternary complex to be optimized early in the optimization process.

6.8 Future Technologies for Structure-​based PROTAC Design Negative-​staining electron microscopy (EM) has been used to explore the conformations of complexes formed with ImIDs and cereblon.66 This approach has the potential to not only provide a higher throughput than X-​ray crystallography but also the ability to observe lowly populated states. Given the high-​ resolution cryo-​EM revolution67,68 and the large size of many E3 ligases, surely

130

130

Chapter 6

it is just a matter of time before cryo-​EM takes center stage alongside X-​ray crystallography as the technologies to drive structure-​based PROTAC design. Through the use of nano-​electrospray ionization (nESI),69 protein complexes can retain their native topology and stoichiometry during transfer from solution into the gas phase,70 an approach referred to as “native mass spectrometry” (nMS)71. The first use of native mass spectrometry to measure the extent of PROTAC-​mediated ternary complex formation has been reported.72 Using nMS ternary complexes of the BRD4 PROTACs, AT1 and MZ128,58,73,74 were measured semi-​quantitatively and their specificities to multiple BET proteins were evaluated in a single measurement. Molecular modelling has been used to optimize cereblon-​ based SIRT2 PROTACs75 and in silico protein–​protein docking using MD simulations have been used to explain the differing selectivities observed for VHL-​based p38 PROTACs.76 Monte Carlo docking algorithms were also applied to sample low-​ energy PPI conformations in the absence of PROTAC molecules and were able to identify conformations that closely resembled that observed in PROTAC ternary X-​ray crystal structures.61 Particularly in combination with the aforementioned biophysical approaches, protein–​protein docking is expected to play an important future role in the identification of the right ternary complex to optimize. If continued technological advances can be made and sufficient practitioners emerge, SBPD is sure to play a central role in the emergence of this new therapeutic modality.

6.9 Acknowledgments The author would like to thank Alessio Ciulli, Dirk Kessler and Peter Ettmayer for their assistance in the preparation of this book chapter.

References W. C. Ripka, W. J. Sipio and J. Blaney, Lect. Heterocycl. Chem., 1987, 9, S95. T. L. Blundell, Nature, 1996, 384, 23–26. R. S. Bohacek, C. McMartin and W. C. Guida, Med. Res. Rev., 1996, 16,  3–​50. P. M. Colman, Curr. Opin. Struct. Biol., 1994, 4, 868–​874. I. D. Kuntz, Science, 1992, 257, 1078–​1082. H. M. Berman, P. E. Bourne, J. Westbrook and C. Zardecki, The Protein Databank, CRC Press, 2003, 394–​410. 7. M. Levitt, Proc. Natl. Acad. Sci. U. S. A., 2007, 104, 3183–​3188. 8. M. Miller, J. Schneider, B. Sathyanarayana, M. Toth, G. Marshall, L. Clawson, L. Selk, S. Kent and A. Wlodawer, Science, 1989, 246, 1149–​1152. 9. R. Lapatto, T. Blundell, A. Hemmings, J. Overington, A. Wilderspin, S. Wood, J. R. Merson, P. J. Whittle, D. E. Danley, K. F. Geoghegan, S. J. Hawrylik, S. E. Lee, K. G. Scheld and P. M. Hobart, Nature, 1989, 342, 342299a0. 1. 2. 3. 4. 5. 6.

 131

Structure-based PROTAC Design

131

10. M. von Itzstein, W.-​Y. Wu, G. B. Kok, M. S. Pegg, J. C. Dyason, B. Jin, T. V. Phan, M. L. Smythe, H. F. White, S. W. Oliver, P. M. Colman, J. N. Varghese, D. M. Ryan, J. M. Woods, R. C. Bethell, V. J. Hotham, J. M. Cameron and C. R. Penn, Nature, 1993, 363, 363418a0. 11. J. N. Varghese, Drug Develop. Res., 1999, 46, 176–​196. 12. S. J. Hughes and A. Ciulli, Essays Biochem., 2017, 61, 505–​516. 13. I. Collins, H. Wang, J. J. Caldwell and R. Chopra, Biochem. J., 2017, 474, 1127–​1147. 14. M. Toure and C. M. Crews, Angew. Chem., Int. Ed., 2016, 55, 1966–​1973. 15. D. P. Bondeson and C. M. Crews, Annu. Rev. Pharmacol. Toxicol., 2017, 57, 107–​123. 16. E.  Fischer, Ber. Dtsch. Chem. Ges., 1894, 27, 2985–​2993. 17. H. Jhoti, A. Cleasby, M. Verdonk and G. Williams, Curr. Opin. Chem. Biol., 2007, 11, 485–​93. 18. P. J. Hajduk and J. Greer, Nat. Rev. Drug Discovery, 2007, 6,  211–​9. 19. S. B. Shuker, P. J. Hajduk, R. P. Meadows and S. W. Fesik, Science, 1996, 274, 1531–​1534. 20. T. Clackson and J. Wells, Science, 1995, 267, 383–​386. 21. P. Wu, T. E. Nielsen and M. H. Clausen, Drug Discovery Today, 2016, 21,  5–​10. 22. A.  Mullard, Nat. Rev. Drug Discovery, 2019, 18,  85–​89. 23. G. Manning, D. B. Whyte, R. Martinez, T. Hunter and S. Sudarsanam, Science, 2002, 298, 1912–​1934. 24. M. P. H. Stumpf, T. Thorne, E. de Silva, R. Stewart, H. J. An, M. Lappe and C. Wiuf, Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 6959–​6964. 25. V. Zoppi, S. J. Hughes, C. Maniaci, A. Testa, T. Gmaschitz, C. Wieshofer, M. Koegl, K. M. Riching, D. L. Daniels, A. Spallarossa and A. Ciulli, J. Med. Chem., 2018, 62, 699–​726. 26. X. Han, C. Wang, C. Qin, W. Xiang, E. Fernandez-​Salas, C.-​Y. Yang, M. Wang, L. Zhao, T. Xu, K. Chinnaswamy, J. Delproposto, J. Stuckey and S. Wang, J. Med. Chem., 2019, 62, 941–​964. 27. Y. Wang, X. Jiang, F. Feng, W. Liu and H. Sun, Acta Pharm. Sin. B, 2020, 10, 207–​238. 28. M. S. Gadd, A. Testa, X. Lucas, K.-​H. Chan, W. Chen, D. J. Lamont, M. Zengerle and A. Ciulli, Nat. Chem. Biol., 2017, 13, 514–​521. 29. P. Filippakopoulos, J. Qi, S. Picaud, Y. Shen, W. B. Smith, O. Fedorov, E. M. Morse, T. Keates, T. T. Hickman, I. Felletar, M. Philpott, S. Munro, M. R. McKeown, Y. Wang, A. L. Christie, N. West, M. J. Cameron, B. Schwartz, T. D. Heightman, N. Thangue, C. A. French, O. Wiest, A. L. Kung, S. Knapp and J. E. Bradner, Nature, 2010, 468, 1067–​1073. 30. S. J. Atkinson, P. E. Soden, D. C. Angell, M. Bantscheff, C. Chung, K. A. Giblin, N. Smithers, R. C. Furze, L. Gordon, G. Drewes, I. Rioja, J. Witherington, N. J. Parr and R. K. Prinjha, MedChemComm, 2014, 5, 342–​351. 31. Y. Zhao, B. Zhou, L. Bai, L. Liu, C.-​Y. Yang, J. L. Meagher, J. A. Stuckey, D. McEachern, S. Przybranowski, M. Wang, X. Ran, A. Aguilar, Y. Hu,

132

132

Chapter 6

J. W. Kampf, X. Li, T. Zhao, S. Li, B. Wen, D. Sun and S. Wang, J. Med. Chem., 2018, 61, 6110–​6120. 32. P. Soares, M. S. Gadd, J. Frost, C. Galdeano, L. C. J. Ellis, O. Epemolu, S. Rocha, K. D. Read and A. Ciulli, J. Med. Chem., 2017, 61, 599–​618. 33. M. R. Arkin and J. A. Wells, Nat. Rev. Drug Discovery, 2004, 3, 301–​317. 34. S. Jones and J. M. Thornton, Prog. Biophys. Mol. Biol., 1995, 63,  31–​65. 35. C. Tsai, S. L. Lin, H. J. Wolfson and R. Nussinov, Protein Sci., 1997, 6,  53–​64. 36. L. L. Conte, C. Chothia and J. Janin, J. Mol. Biol., 1999, 285, 2177–​2198. 37. M. Bartel, A. Schäfer, L. M. Stevers and C. Ottmann, Future Med. Chem., 2014, 6, 903–​921. 38. S. A. Andrei, E. Sijbesma, M. Hann, J. Davis, G. O’Mahony, M. W. D. Perry, A. Karawajczyk, J. Eickhoff, L. Brunsveld, R. G. Doveston, L.-​G. Milroy and C. Ottmann, Expert Opin. Drug Discovery, 2017, 12, 1–​16. 39. L.-​G. Milroy, T. N. Grossmann, S. Hennig, L. Brunsveld and C. Ottmann, Chem. Rev., 2014, 114, 4695–​4748. 40. L. Milroy, M. Bartel, M. A. Henen, S. Leysen, J. M. C. Adriaans, L. Brunsveld, I. Landrieu and C. Ottmann, Angew. Chem., Int. Ed., 2015, 54, 15720–​15724. 41. Y. Pommier and C. Marchand, Nat. Rev. Drug Discovery, 2012, 11, nrd3404. 42. P. Thiel, M. Kaiser and C. Ottmann, Angew. Chem., Int. Ed., 2012, 51, 2012–​2018. 43. J. Liu, J. D. Farmer, W. S. Lane, J. Friedman, I. Weissman and S. L. Schreiber, Cell, 1991, 66, 807–​815. 44. E. J. Brown, M. W. Albers, T. B. Shin, K. ichikawa, C. T. Keith, W. S. Lane and S. L. Schreiber, Nature, 1994, 369, 369756a0. 45. J. P. Griffith, J. L. Kim, E. E. Kim, M. D. Sintchak, J. A. Thomson, M. J. Fitzgibbon, M. A. Fleming, P. R. Caron, K. Hsiao and M. A. Navia, Cell, 1995, 82, 507–​522. 46. C. R. Kissinger, H. E. Parge, D. R. Knighton, C. T. Lewis, L. A. Pelletier, A. Tempczyk, V. J. Kalish, K. D. Tucker, R. E. Showalter, E. W. Moomaw, L. N. Gastinel, N. Habuka, X. Chen, F. Maldonado, J. E. Barker, R. Bacquet and J. E. Villafranca, Nature, 1995, 378, 378641a0. 47. J. Choi, J. Chen, S. L. Schreiber and J. Clardy, Science, 1996, 273, 239–​242. 48. L. A. Banaszynski, C. W. Liu and T. J. Wandless, J. Am. Chem. Soc., 2005, 127, 4715–​4721. 49. A.  Whitty, Nat. Chem. Biol., 2008, 4, 435–​439. 50. E. F. Douglass, C. J. Miller, G. Sparer, H. Shapiro and D. A. Spiegel, J. Am. Chem. Soc., 2013, 135, 6092–​6099. 51. P. P. Chamberlain, A. Lopez-​Girona, K. Miller, G. Carmel, B. Pagarigan, B. Chie-​Leon, E. Rychak, L. G. Corral, Y. J. Ren, M. Wang, M. Riley, S. L. Delker, T. Ito, H. Ando, T. Mori, Y. Hirano, H. Handa, T. Hakoshima, T. O. Daniel and B. E. Cathers, Nat. Struct. Mol. Biol., 2014, 21, 803–​809. 52. E. S. Fischer, K. Böhm, J. R. Lydeard, H. Yang, M. B. Stadler, S. Cavadini, J. Nagel, F. Serluca, V. Acker, G. M. Lingaraju, R. B. Tichkule, M. Schebesta, W. C. Forrester, M. Schirle, U. Hassiepen, J. Ottl, M. Hild, R. E. J. Beckwith, J. W. Harper, J. L. Jenkins and N. H. Thomä, Nature, 2014, 512, 49.

 133

Structure-based PROTAC Design

133

53. G. Lu, R. E. Middleton, H. Sun, M. Naniong, C. J. Ott, C. S. Mitsiades, K.-K. Wong, J. E. Bradner and W. G. Kaelin, Science, 2014, 343, 305–​309. 54. J. Krönke, E. C. Fink, P. W. Hollenbach, K. J. MacBeth, S. N. Hurst, N. D. Udeshi, P. P. Chamberlain, D. R. Mani, H. W. Man, A. K. Gandhi, T. Svinkina, R. K. Schneider, M. McConkey, M. Järås, E. Griffiths, M. Wetzler, L. Bullinger, B. E. Cathers, S. A. Carr, R. Chopra and B. L. Ebert, Nature, 2015, 523, 183. 55. M. E. Matyskiela, S. Couto, X. Zheng, G. Lu, J. Hui, K. Stamp, C. Drew, Y. Ren, M. Wang, A. Carpenter, C.-​W. Lee, T. Clayton, W. Fang, C.-​C. Lu, M. Riley, P. Abdubek, K. Blease, J. Hartke, G. Kumar, R. Vessey, M. Rolfe, L. G. Hamann and P. P. Chamberlain, Nat. Chem. Biol., 2018, 14, 981–​987. 56. K. A. Donovan, J. An, R. P. Nowak, J. C. Yuan, E. C. Fink, B. C. Berry, B. L. Ebert and E. S. Fischer, eLife, 2018, 7, e38430. 57. G. Petzold, E. S. Fischer and N. H. Thomä, Nature, 2016, 532, 127. 58. M. Zengerle, K.-​ H. Chan and A. Ciulli, ACS Chem. Biol., 2015, 10, 1770–​1777. 59. J. E. Ladbury, Chem. Biol. (Oxford, U. K.), 1996, 3, 973–​980. 60. W. Farnaby, M. Koegl, M. J. Roy, C. Whitworth, E. Diers, N. Trainor, D. Zollman, S. Steurer, J. Karolyi-​Oezguer, C. Riedmueller, T. Gmaschitz, J. Wachter, C. Dank, M. Galant, B. Sharps, K. Rumpel, E. Traxler, T. Gerstberger, R. Schnitzer, O. Petermann, P. Greb, H. Weinstabl, G. Bader, A. Zoephel, A. Weiss-​ Puxbaum, K. Ehrenhöfer-​ Wölfer, S. Wöhrle, G. Boehmelt, J. Rinnenthal, H. Arnhof, N. Wiechens, M.-​Y. Wu, T. Owen-​Hughes, P. Ettmayer, M. Pearson, D. B. McConnell and A. Ciulli, Nat. Chem. Biol., 2019, 15,  1–​9. 61. R. P. Nowak, S. L. DeAngelo, D. Buckley, Z. He, K. A. Donovan, J. An, N. Safaee, M. P. Jedrychowski, C. M. Ponthier, M. Ishoey, T. Zhang, J. D. Mancias, N. S. Gray, J. E. Bradner and E. S. Fischer, Nat. Chem. Biol., 2018, 14, 706–​714. 62. P. Filippakopoulos, S. Picaud, M. Mangos, T. Keates, J.-​ P. Lambert, D. Barsyte-​Lovejoy, I. Felletar, R. Volkmer, S. Müller, T. Pawson, A.-​ C. Gingras, C. H. Arrowsmith and S. Knapp, Cell, 2012, 149, 214–​231. 63. A. Testa, S. J. Hughes, X. Lucas, J. E. Wright and A. Ciulli, Angew. Chem., 2019, 132, 1744–​1751. 64. J. Kuriyan and D. Eisenberg, Nature, 2007, 450, 983–​990. 65. D. P. Bondeson, B. E. Smith, G. M. Burslem, A. D. Buhimschi, J. Hines, S. Jaime-​Figueroa, J. Wang, B. D. Hamman, A. Ishchenko and C. M. Crews, Cell Chem. Biol., 2018, 25,  78–​87. 66. M. E. Matyskiela, G. Lu, T. Ito, B. Pagarigan, C.-​ C. Lu, K. Miller, W. Fang, N.-​ Y. Wang, D. Nguyen, J. Houston, G. Carmel, T. Tran, M. Riley, L. Nosaka, G. C. Lander, S. Gaidarova, S. Xu, A. L. Ruchelman, H. Handa, J. Carmichael, T. O. Daniel, B. E. Cathers, A. Lopez-​Girona and P. P. Chamberlain, Nature, 2016, 535, 252. 67. E.  Callaway, Nature, 2020, 578, 201. 68. M. Saur, M. J. Hartshorn, J. Dong, J. Reeks, G. Bunkoczi, H. Jhoti and P. A. Williams, Drug Discovery Today, 2020, 25, 485–​490.

134

134

Chapter 6

69. J. Fenn, M. Mann, C. Meng, S. Wong and C. Whitehouse, Science, 1989, 246,  64–​71. 70. M. Sharon and C. V. Robinson, Annu. Rev. Biochem., 2007, 76, 167–​193. 71. A. J. R. Heck, Nat. Methods, 2008, 5, 927–​933. 72. R. Beveridge, D. Kessler, K. Rumpel, P. Ettmayer, A. Meinhart and T. Clausen, ACS Cent. Sci., 2020, 6, 1223–1230. 73. K. M. Riching, S. Mahan, C. R. Corona, M. McDougall, J. D. Vasta, M. B. Robers, M. Urh and D. L. Daniels, ACS Chem. Biol., 2018, 13, 2758–​2770. 74. M. J. Roy, S. Winkler, S. J. Hughes, C. Whitworth, M. Galant, W. Farnaby, K. Rumpel and A. Ciulli, ACS Chem. Biol., 2019, 14, 361–​368. 75. M. Schiedel, D. Herp, S. Hammelmann, S. Swyter, A. Lehotzky, D. Robaa, J. Olah, J. Ovádi, W. Sippl and M. Jung, J. Med. Chem., 2018, 61, 482–​491. 76. B. E. Smith, S. L. Wang, S. Jaime-​ Figueroa, A. Harbin, J. Wang, B. D. Hamman and C. M. Crews, Nat. Commun., 2019, 10, 131.

 135

CHAPTER 7

Plate-​based High-​throughput Cellular Degradation Assays to Identify PROTAC Molecules and Protein Degraders NIKKI CARTER* Discovery Biology Cellular Assay Development, AstraZeneca, Cambridge, UK *Email: [email protected]

7.1 Introduction Targeted protein degradation as an evolving modality shows therapeutic promise and has the potential to target proteins beyond the small-​molecule space. PROTACs (PROteolysis-​TArgeting Chimeras) are one way to harness the ubiquitin proteasome degradation pathway to degrade target proteins. To date, clinical trials that target hormone-​resistant androgen and estrogen receptors have been initiated. Emerging research also shows examples of protein degradation beyond PROTACs, utilizing small molecule-​induced degradation of proteins such as IMiD molecular glues and protein destabilization via SERD and SARD molecules.5 Despite these successful examples, our understanding and optimization of induced protein degradation is limited. This is in part due to the dynamic and complex process, which requires the redirection of endogenous cascades to target proteins. This process is often complex, depending on Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

135

136

136

Chapter 7

Figure 7.1 Schematic of the processes required for targeted protein degradation by a PROTAC or degrader. PROTAC/​degrader uptake, does the molecule enter the cell? Does the PROTAC/​degrader engage the target protein? Specifically, for PROTAC mechanisms, does the PROTAC engage the E3? Does the PROTAC induce a temary complex and then facilitate ubiquitination of the target protein? Does protein degradation occur via proteasomal recognition? Is this degradation specifically mediated by the PROTAC/​degrader? What is the influence of protein turnover/​synthesis in the cell in the presence of a PROTAC/​degrader?

multiple enzymes with different kinetics and which are therefore challenging to study. One example is the molecular glue mechanism of lenalidomide that redirects the E3 ligase cereblon to degrade Icarus, which leads to its beneficial effect in blood-​borne malignancies and Sal4 which, at least in part, causes its teratogenic effects. Even though this class of drugs was discovered in the last century, the details of the molecular glue mechanism that leads to on-​target degradation was only described recently. Therefore, the rational development of targeted protein degraders requires assays in addition to affinity readouts common to small-​molecule cascades. Cellular assays that determine target degradation are crucial. To understand the mechanism of action of candidate drugs, ubiquitin transfer rate and, in the case of PROTACs, ternary complex formation assays which provide insights into the key biological processes that can be optimized for rational drug design are required to build protein degrader discovery cascades. Taken together, the successful development of targeted protein degraders requires a clear understanding of complex processes in patient-​relevant cells, preferably in a kinetic setup. Figure 7.1 illustrates the optimal points where and when degrader process understanding is required.

7.2 PROTACs A PROTAC molecule is comprised of three components: A POI-​ binding ligand, an E3 ligase-​binding warhead and a connecting linker. This allows

 137

Plate-based High-throughput Cellular Degradation Assays

137

a PROTAC to bring a POI in close proximity to the E3 ligase, resulting in ubiquitination and subsequent proteasomal degradation of the POI. Because of their hypothesized catalytic mode of action,1 PROTACs can degrade POIs in a sub-​stoichiometric ratio, therefore potentially reducing off-​target toxicity. Intrinsic to the PROTAC-​specific mechanism of action (MoA) is the so-​called “hook effect,” which arises when a high concentration of a PROTAC oversaturates all binding sites of the E3 ligase and the POI independently, thereby leading to binary complexes with either POI or E3 rather than producing ternary complexes. So far, the design of PROTAC molecules is still largely empirical, but there is a growing body of evidence that positive cooperativity, kinetics and stability of ternary complex formation and efficient ubiquitin transfer contribute to an efficient PROTAC. This chapter will focus on how cellular assays that measure protein degradation used by AstraZeneca can facilitate intermediate mechanistic understanding of PROTACS and small-​molecule protein degraders.

7.3 Plate-​based Assays to Measure Protein Degradation The primary output for measurement of protein degradation are cellular endpoint assays that quantify relative protein levels in the absence and presence of a PROTAC or protein degrader. Historically, western blots have been used to quantify relative protein levels. However, the utility of western blots as part of degrader hit finding and optimization cascades are limited by the low-​ throughput and semi-​quantitative nature of the methodology. Technologies that facilitate dynamic measurement of the endogenous POI with quantitative measures such as DC50 and Dmax are preferable. Label-​free proteomic technologies such as targeted mass spectrometry is routinely utilized for bespoke characterization of protein levels, protein turnover rates and global protein change in the presence of a degrader. In addition, they can inform on the absolute protein concentrations by using labelled peptide standards (AQUA) and turnover rates (stable isotope labelling by amino acids in cell culture, or SILAC). While these methods are the gold standard for understanding protein dynamics and can be applied in bespoke MoA studies, they are limited by their low throughput, time-​consuming nature and the requirement for specialized data analysis. Taken together, this means that both western blot and mass spectrometry can be used to analyze individual drug candidates, but do not provide the throughput required to support a protein degradation cascade. Plate-​based cellular protein degradation assays combined with novel protein tagging technologies provide the throughput and precision required for drug discovery and compound characterization. The following assay technologies are suitable to measure POI degradation and provide different approaches to develop a PROTACs or small-​molecule cascade. A summary of microtiter plate-​based technologies to interrogate changes in cellular protein levels is shown in Table 7.1.

138

138

Chapter 7

Table 7.1  Assay technology summary for measurement of cellular degradation. Antibody epiturbance is the phenomena whereby the small-​ molecule degrader or PROTAC perturbs the antibody binding to the POI, thus artifactually appearing as if they degrade the target. Technology

Endogenous Kinetically Comments cells enabled

Immunofluorescent ✓ target imaging ELISA or derivatives ✓

✗

Ectopic tagging (GFP/​RFP, HiBIT)

✗

✓

Endogenous tagging ✓ (GFP/​RFP, HiBIT) Targeted LCMS ✓

✓

✗

✗

Antibody epiturbance sometimes encountered Antibody epiturbance sometimes encountered High-​expression systems lack sensitivity to measure accurate degradation *CRISPR/​Cas9 knockin of tag required for “endogenous” endpoint Specialized equipment and analysis required.

7.3.1 Immunofluorescent Target Imaging The use of high-​throughput microscopy in combination with fluorescent antibodies to label and quantify the endogenous POI is compatible with microtiter plate-​based setups and commonly used to quantify target protein levels. The possibility to “multiplex” protein immunofluorescent antibodies and include DNA stains like DAPI to quantify cell numbers to normalise POI signal can be utilized to investigate degrader selectivity over “off-​target” proteins in a single readout. The versatility of the immunofluorescent affinity reagents themselves confer advantages across different technologies, cell and tissue types. This technology can be applied to target validation and translation studies at different stages of a degrader discovery cascade, thereby providing a uniform readout to interrogate POI levels. One limitation of antibody-​based technologies is that cells have to be fixed and permeabilized to allow POI detection. This means that while endpoint assays are possible, kinetic studies in live cells are impossible. In addition to that, antibody-​based technologies have unique risks in the context of protein degraders: In AstraZeneca we have observed a phenomenon known as “epiturbance,” where the small-​molecule degrader or PROTAC perturbs the antibody binding to the POI, thus artifactually appearing as if they degrade the target, resulting in a false positive read out. The requirement to capture the dynamic nature of protein degradation and protein recovery means that protocols requiring fixed cells need multiple time points to capture the multiphasic profiles of degradation. This is both intensive and tedious; comprehensive assay optimization work to determine the dynamic range and sensitivity of the antibody tools and optimal time points is required. For some degrader projects in AstraZeneca, this has resulted in the requirement to run orthogonal assay technologies alongside target imaging endpoints to confirm key parameters like Dmax and transient

 139

Plate-based High-throughput Cellular Degradation Assays

139

behavior of hook effects. In summary, when using an immunofluorescent imaging assay, the following criteria require consideration to build a fit-​for-​ purpose cellular assay. • Ectopic or endogenous system –​for quantitative data DC50 and Dmax, endogenous is advantageous where possible. • Is the antibody optimized for the target POI? Sensitivity is a key parameter. • Check for epiturbance –​requires tool degraders. • Is a dynamic read out required?

7.3.2 ELISA or Derivative Assay Technologies ELISA detection technology has long been used for the quantification of proteins in cellular samples. New-​generation immunoassays use selective capture or proximity energy transfer methodologies to eliminate the need for multiple wash steps, thus facilitating a mix-​and-​read assay amenable to high-​throughput sensitive detection of endogenous POI. Unlike the target immunofluorescent imaging assays described, the use of new-​generation ELISA technologies such as Perkin Elmer’s AlphaLisa™ require just a standard luminescent plate reader. In AstraZeneca the technology has been employed to interrogate both POI degradation and cellular target engagement via high-​throughput CETSA,6 highlighting the advantage of versatile tool antibody systems. Likewise, this assay technology has limitations, intracellular POIs require a lytic endpoint, making this technology unsuitable for dynamic degradation measurements. Using antibody-​based technologies, AstraZeneca has seen evidence of epiturbance for some targets resulting in false positives. Orthogonal read outs have been employed alongside this technology to compliment the data derived from this technology. In summary, when using an ELISA assay format, the following criteria require consideration to build a fit-​for-​purpose cellular assay. • Ectopic or endogenous system –​for quantitative data DC50 and Dmax, endogenous is advantageous where possible. • Is there a suitable antibody pair, and are they optimized for the target POI? Sensitivity is a key parameter. • Check for epiturbance –​requires tool degraders. • Is a dynamic read out required?

7.3.3 Protein Tagging Protein tags as tools to measure protein dynamics are broad and they can be applied in a cellular context both in a recombinant ectopic system, or more recently as tags to endogenous proteins via CRISPR/​Cas 9 modification. The application of ectopic protein tagging to generate a fluorescent or luminescent labelled protein is an instrumental technology for measuring protein localization and dynamics in a live cell system via standard recombinant protein expression techniques. Fluorescent protein tags such as GFP and mCherry

140

140

Chapter 7

were first utilized in targeted protein degradation publications combined with confocal microscopy to show the degradation efficiency of PROTACs in live cells.8 More recently, the Promega HiBit® or NanoLuc® tag systems are prevalent, both in AstraZeneca and in the literature. The advantage of the Promega system is flexibility; they can be used in a transient or stable system to quickly validate targeted degradation as a modality. Depending on the design of the fusion vector the system is amenable to both lytic and live cell assays and has the option to add HaloTag and use Halo-​PROTAC as a positive control for the assay where no tool molecules are available. As such, AstraZeneca positions this technology at several stages of a protein degradation cascade. First, as a “plug and play” assay system to replace laborious western blots as we validate protein degradation as a modality for a new target; second, as an endpoint or kinetic assay for degrader profiling to interrogate degrader dynamics and MoA in live cells. When employing a tag system to build an assay, consideration must be given to the size and position of the tag and the choice of ectopic or endogenous systems applied, as all ultimately impact on the data conclusions that can be drawn from the subsequent assays. In AstraZeneca we have used both ectopic and endogenous methodologies successfully in degrader projects. We have observed that ectopic tagged systems show differing POI degradation and recovery dynamics and right-​shifted DC50 to endogenously tagged systems. This phenomenon is particularly prevalent in systems where the fusion system promoter results in high expression of the tagged POI. Ultimately related to both the expression level and turnover rate of the POI fusion and the lack of native constitutive regulation, it is to be expected that these systems do not reflect the native protein homeostasis that you would gain in an endogenous system. However, they do have value in a protein degrader assay cascade when positioned appropriately and with due diligence to the respective caveats. Unlike many other ectopic cellular assay systems employed for small-​molecule inhibitor programs, whereby maximal overexpression to obtain a robust assay window may be desirable, this is not the case in a degrader system. In summary, when using a tagged system, the following criteria require consideration to build a fit-​for-​purpose cellular assay. • Ectopic or endogenous system –​for live cell or quantitative data DC50 and Dmax, endogenous tagging is advantageous where possible. • Size of tag –​can vary from large protein tags to smaller amino acid tags, but complementation partners are often large and high-​affinity. • Position of tag –​N or C terminal tagging can be employed; target biology and steric hindrance are key considerations; trying both orientations is valuable to optimize the best position.

7.3.4 Validation of Assays to Measure Protein Degradation Validation of an assay for cellular degradation should undertake the same robust validation criteria that would be applied to any cellular assay build for a hit finding or profiling cascade. In AstraZeneca we find that the assays built

 141

Plate-based High-throughput Cellular Degradation Assays

141

for degrader programs using the technologies described above are as robust in terms of statistical parameters (robust Z′, CIR, MDR) as those employed for other small-​molecule modalities. However, there are some caveats unique to degrader programs that are key considerations to good cellular assay design and validation. • Use of degrader tools like Halo-​PROTAC or dTAG to define degradation phenotype is valuable in the assay development phase. • Consider the likelihood of false positives derived from cellular toxicity, not targeted degradation. • Degrader controls for maximal degradation –​the use of Halo-​PROTAC to define relative Dmax controls in the absence of tool degraders has shown value in our programs. • Robust validation of sample variance for both Dmax and DC50 –​understanding the variance of both is critical for rationale design. • Physical properties of PROTACS –​poor aqueous solubility and cellular permeability. • Data analysis challenges –​standard four-​parameter fit logistics in current curve-​fitting packages struggle to resolve the biphasic curves seen with some early-​stage PROTAC molecules. Careful consideration must be given to fixing any fit parameters.

7.4 Plate-​based Assays to Understand Degrader Mechanism –​PROTACS Cellular assays that help understand or troubleshoot a degrader molecule’s MoA and cellular interactions are required to drive rationale molecule design and as a key criterion for selecting which molecules to progress at candidate selection stage. As shown in Figure 7.1, there are multiple stages in the proteasomal degradation pathway where the degrader molecule can fail to result in degradation of the POI. This section will focus specifically on cellular assay technologies that can be used to interrogate the MoA of a PROTAC-​based degrader. However, several aspects could be applied to other targeted protein degrader modalities.

7.4.1 Cellular Permeability and Target Engagement Assays Cellular target engagement assays are routinely used to confirm that the degrader molecule or its intermediates are cell-​permeable and engaged with either the target POI or E3 ligase. There are several established assay technologies that can be used to interrogate these key parameters. Thermal stabilization on degrader ligand binding is known to occur within the context of the intracellular environment. This phenomenon can be detected in a Cellular Thermal Shift Assay (CETSA®). Here, degrader molecules are incubated with live cells before subjecting cells to a transient heat-​shock, after which the remaining soluble protein is quantified. CETSA® experiments can be performed with a

142

142

Chapter 7

single concentration of compound tested across multiple temperatures or in a dose–​response format at a single temperature optimized for the protein of interest. We have utilized high-​throughput CETSA® (CETSA® HT), which uses AlphaScreen® technology to quantify the remaining soluble protein after heat shock. In this format, a pair of antibodies directed at orthogonal epitopes on the target protein are used to form a ternary complex that can be detected by suitable donor and acceptor AlphaScreen® beads. As with all antibody-​based systems, ensuring that you don’t encounter epiturbance at the warhead/​ligand binding site is required to eliminate artifactual results. The advantage of the CETSA® HT technology is the ability to combine this with the same assay reagents and tools utilized for an AlphaScreen® degradation assay. An alternative assay system for confirming cellular target engagement is NanoBRET®.3 NanoBRET® measures degrader binding at either the POI or warhead binding the E3 within intact cells based on an energy transfer technique designed to measure molecular proximity in living cells. NanoBRET® target engagement assays can be performed using transient, stable or CRISPR endogenously tagged NanoLuc® or HiBiT fusions to the specific POI or E3 ligase; the fluorescent tracers are compatible with lytic or live-​cell formats to assess cellular permeability. There are synergies with the cell-​line tools and reagents that could be utilized in a degradation assay. A key advantage of the NanoBRET® system is the ability to look at a dynamic live-​cell format. This methodology can be informative for assessing kinetics of cellular permeability to rank POI ligands and the intermediates of PROTAC molecules, considering the potential for permeability issues relating to molecule size and linker composistion.

7.4.2 PROTAC Ternary Complex Formation Assays that measure formation of the ternary complex may be utilized for both mechanistic understanding and measuring quantitative parameters driving rational PROTAC design. At AstraZeneca, we use a combination of cell-​free and biophysical methods to derive parameters such as POI and warhead affinity and cooperativity coefficient alpha for the components of ternary complex in a PROTAC degrader system. We have used cellular endpoints to develop ternary complex understanding when required in a cellular context. We have used NanoBRET® with the POI fused to NanoLuc® luciferase acting as an as the energy donor, and the N terminus of E3 ligase fused to HaloTag® as an energy acceptor. Although transient expression of both fusion partners is possible, we have favored stable expression of the energy acceptor in a cell line of choice with transient expression of the donor because the variety of POI targets is broader than the E3 ligases currently utilized. This methodology acts as a simple cellular ternary complex endpoint to confirm data seen in the cell-​free systems. Applied predominantly to projects as a trouble-​shooting assay where we can see cellular target engagement but don’t measure POI degradation in a cellular system with unoptimized PROTAC molecules, this technology can be applied quickly with minimal molecular biological burden. Where we have shown considerable impact above a cell-​free or biophysical system is with

 143

Plate-based High-throughput Cellular Degradation Assays

143

kinetic ternary complex measurements. Utilizing a comparable system with either stable ectopic expression or endogenously tagging with CRISPR/​Cas9 and stable Vivacine™ substrate you can measure live-​cell ternary complex formation. Thus, enabling qualitative ranking of PROTAC molecules based on the dynamics and stability of ternary complex formation in a cellular context. This information contributes to a detailed MoA package as projects progress to candidate selection decisions. The caveat to this methodology is that the assay development phase is not trivial, the plated cells are often exposed to a non-​environmentally controlled plate reader for a prolonged period, so consideration must be given to both the value of this assay to a cascade and the potential for artifactual data arising from this assay type.

7.4.3 PROTAC-​mediated Ubiquitination Assays that measure addition/​transfer of ubiquitin to complex may be utilized for both mechanistic understanding and measuring quantitative parameters driving rational PROTAC design. In situations where quantitative parameters are required, AstraZeneca has tended toward cell-​free or biophysical methods to derive values such as Kcat of the ubiquitination event with a chosen E3 ligase. In projects where lower-​throughput assays are acceptable, use of western blots raised against affinity-​labelled ubiquitin have been the assay of choice. Commercial kits from life sensors utilizing magnetic TUBE’s to enrich and isolate ubiquitinylated POI claim to increase throughput from standard western blots protocols. For high-​throughput applications, we have used HaloTag® ubiquitin in a NanoBRET® format to develop assays to understand ubiquitin transfer in a cellular context, with HaloTag® ubiquitin as an energy acceptor and NanoLuc® luciferase fused POI as the donor. Utility of this assay format in a live-​cell dynamic read out has allowed qualitative ranking of PROTAC molecules based on the dynamics and stability of ubiquitin transfer in a cellular context. Ectopic or CRISPR/​Cas9 endogenous expression of HiBiT protein fusions is an alternative system, although not one that we have utilized in AstraZeneca routinely. Cellular assays that interrogate ubiquitination add value to a degrader cascade as a trouble-​shooting assay where cellular target engagement occurs, but POI degradation is not detected. Also, at a later project stage, optimized PROTACS form part of a detailed mechanism of action package prior to candidate selection.

7.4.4 Proteasome Recognition Assays Proteasome recognition assays by ELISA have been utilized to look at recognition of the affinity-​labelled ubiquitinated protein in a cellular lysate for isolated proteasomes. While these assay formats can be configured to look at a specific POI, the frequent wash steps and low-​density formats do not make them amenable to higher-​ throughput dynamic endpoints. For high-​ throughput applications, the HaloTag® technology in a NanoBRET® format can be utilized to understand proteasome recognition in a dynamic cellular context, with

144

144

Chapter 7

HaloTag® PSMD3 proteasomal subunit as an energy acceptor and NanoLuc® luciferase fused POI as the donor. Again, ectopic and endogenous applications are possible. This is not a function of PROTAC MoA that AstraZeneca chooses to study, based on the application of other mechanistic assays in the cascade that we feel add greater value to our understanding and from which information about the series of events leading to effective degradation can be implied.

7.4.5 Modification of Degradation Assays to Assess POI Abundance and Turnover Quantification of protein abundance and turnover forms a critical part of the validation of targeted degradation as a modality to obtain a desired biological effect. Internal mathematical modelling and literature2 show that POI turnover impacts DC50 and Dmax of a protein degrader. Measurements of POI stability/​half-​life can be achieved by application of cycloheximide, which is an inhibitor of protein biosynthesis that acts by prevention of translational elongation. It is widely used in cell biology to determine the half-​life of a given protein by western blots at optimized cycloheximide concentrations and time points. Combining cycloheximide additions to the plate-​based dynamic assay technologies previously described to study POI degradation enables a higher-​ throughput read out to determine POI half-​life. Alternative techniques such as click chemistry or radiochemical pulse chase labelling or SILAC methodologies may also be employed and provide highly quantitative data, but these often require specialist equipment and are less amenable to the degradation assay methods described above. For bespoke mechanistic profiling for candidate selection, SILAC is a viable endpoint. Consideration must be given to the use of ectopic versus endogenous read outs and how amenable the technology is to live-​cell dynamics to derive the optimal assay for high-​throughput half-​life determination. Translation between cell lines and ultimately tissue cannot be implied due to intrinsic differences in E3 ligase and POI concentrations. Therefore, quantitative knowledge of this parameter is required for modality validation and quantitative comparison for translation between tool cellular assay systems.

7.4.6 Mechanistic  Tools There are many chemical and protein tools that can be utilized alongside the assays described above to interrogate targeted protein degrader mechanism. • MG132 –​cell-​permeable reversible proteasome inhibitor, routinely used to confirm that a degrader molecule acts via the proteasome. • Cycloheximide –​inhibitor of protein synthesis, utilized in pulse chase experiments to determine protein half-​life. • dTAG4 and derivatives –​dTAG is a cereblon-​based PROTAC targeting mutant FKBP12F36V fusion proteins. The fusion protein can be expressed

 145

Plate-based High-throughput Cellular Degradation Assays

145

as a fusion with a target protein of interest via ectopic or CRISPR/​Cas9 cellular engineering. dTAG induces targeted degradation of FKBP12F36V fusion proteins in vitro and in vivo. Routinely used in cereblon degrader programs as a tool degrader in cellular assay development. Its intrinsic catalytic degrader activity makes dTAG a valuable tool. • Halo-​PROTAC7 and derivatives –​a chloroalkane-​containing PROTAC, induces degradation of HaloTag fusion proteins. • Chloroquine –​cell-​permeable lysosome inhibitor that acts by altering the acidification of the lysosome, routinely used to confirm that a degrader molecule acts via the lysosome. • MLN4924 –​a small-​ molecule inhibitor of Nedd8-​ activating enzyme (NAE) with potential antineoplastic activity, routinely used to confirm neddylation. There are a vast selection of cellular plate-​based assays and mechanistic tools that enable the in-​depth understanding of targeted protein degradation profiles and the respective biological processes that underpin successful therapeutic degradation of a protein target. Assays that measure dynamic cellular degradation profiles are critical path assays for any degrader discovery program. Learnings from AstraZeneca programs have highlighted that assay technologies amenable to kinetic live-​cell measurements in endogenous or endogenously tagged systems are required to drive the quantitative parameters needed to develop rational degrader design. Technologies that can be applied synergistically to other stages of a degrader cascade result in flexible dynamic assay development application. Assay requirements for mechanistic understanding of the biological processes that underlie degradation specifically for PROTAC programs are largely led by either early-​stage project troubleshooting around PROTAC structure–​activity relationships or late-​stage deep characterization of pre-​candidate molecules and should be applied when required (Figure 7.2).

Figure 7.2 Profiling PROTACs mechanism of action –​summary of key assays. Target engagement, degradation (endpoint and kinetic), bespoke cell-​ based ubiquitination and ternary complex as required or for troubleshooting.

146

146

Chapter 7

References 1. D. P. Bondeson, A. Mares, I. E. Smith, E. Ko, S. Campos, A. H. Miah, et al., Nat. Chem. Biol., 2015, 11, 611–​617. 2. P. M. Cromm and C. M. Crews, Cell Chem. Biol., 2017, 24(9), 1181. 3. T. Machleidt, C. C. Woodroofe, M. K. Schwinn, J. Mendez, M. B. Robers, K. Zimmerman, et al., ACS Chem. Biol., 2015, 10(8),1797–​804. 4. B. Nabet, J. M. Roberts, D. L. Buckley, J. Paulk, S. Dastjerdi, A. Yang, et al., Nat. Chem. Biol., 2018, 14(5), 431. 5. M. Pettersson and C. M. Crews, Drug Discovery Today: Technol., 2019, 31, 15. 6. J. Shaw, I. Dale, P. Hemsley, L. Leach, N. Dekki, J. P. Orme, et al., SLAS Discovery, 2019, 24(2), 121. 7. H. Tovell, A. Testa, C. Maniaci, H. Zhou, A. R. Prescott, T. Macartney, et al., ACS Chem. Biol., 2019, 14(5), 882. 8. M. Zengerle, K. H. Chan and A. Ciulli. ACS Chem. Biol., 2015, 10(8), 1770.

 147

CHAPTER 8

PROTAC Targeting BTK for the Treatment of Ibrutinib-​ resistant B-​cell Malignancies YONGHUI SUN AND YU RAO* MOE Key Laboratory of Protein Sciences, School of Pharmaceutical Sciences, MOE Key Laboratory of Bioorganic Phosphorus Chemistry & Chemical Biology, Tsinghua University, Beijing 100084, P.R. China. *Corresponding author. Email: [email protected]

8.1 Introduction Bruton’s tyrosine kinase (BTK) belongs to non-​receptor tyrosine protein kinase, which demonstrates important functions in B-cell receptor (BCR) signal transduction and B-​cell activation (Figure 8.1). As a member of the Tec kinase family, BTK plays a critical role in the differentiation and development of B lymphocytes. BTK is expressed in hematopoietic cells except T lymphocytes and natural killer cells. In activated B-​cell-​like (ABC) diffuse large B-​cell lymphoma (DLBCL), BTK also displays an essential function in the immune escape of these cells.1 BTK is the first TEC family kinase found to be bifunctional in apoptosis, which promotes free radical-​induced apoptosis in B lymphocytes but inhibits Fas-​activated apoptosis.2,3 Three PTK families –​Tec, Src and Syk –​are activated after BCR binds to the antigen. The first kinases activated are Src family

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

147

148

148

Chapter 8

Figure 8.1 BTK signal transduction in BCR signaling pathway.

members, including Fyn, Blk and Lyn. Activated Lyn phosphorylates ITAM in the domain of Igα/​Igβ to generate a pTyr motif, allowing the SH2 domain of Syk to bind, resulting in multiple cases of fast autophosphorylation in Syk within its linker. Phosphorylation of Syk provides an docking site in the SH2 domain for recruiting and activating downstream signalings such as BTK and PLCγ2. PLC2 hydrolyzes PIP2 to provide two second messengers, diacylglycerol (DAG) and IP3. IP3 binds to a receptor (IP3R), triggers mobilization of Ca2+, and subsequently causes NF-​AT to be activated by calmodulin and calcineurin. DAG recruited members of the protein kinase C family, which then activated the CARMA1, Bcl10 and MALT1 (CBM) complexes, thereby activating NF-​ κB. Recruitment of a variety of guanine nucleotide exchange factors (GEFs), including RasGRP from SOS and DAG from Grb2 to BCR signaling bodies, activates small GTPase Ras and then activates MAP. BTK is also a kind of protein tyrosine

 149

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 149

kinase (PTK), which can be activated and recruited into the proximal signal body of the Syk to phosphorylate PLC2. Starting from the Syk membrane proximal signal body, BCR signal transduction is diversified, leading to the triggering of at least four different signal transduction cascades to the transcriptional activation or regulation of NF-​AT, NF-​κB, MAPK and FoxO pathways.4 In organisms, BTK plays a critical physiological role. In 1952, American pediatrician Ogdon Bruton found that lacking of an antibody in children led to X-​linked agammaglobulinemia (XLA), leading to recurrent bacterial infections and sepsis.5,6 In 1993, the gene that caused XLA was finally confirmed. BTK can be divided into five essential domains: a Pleckstrin homeodomain (PH), a Tec homeodomain (TH), a Src homeodomain (SH3), an SH2 domain and a C-​terminal domain with kinase activity. In humans, if the BTK gene is absent in the genome, the number of B cells will significantly decrease and the serum gamma protein level will decrease also.7 After BTK mutates, X-​ related immunodeficiency (xid) occurs in mice, which is less severe than XLA.8 Because B cells play crucial functions in immune responses, besides antitumor effects, BTK inhibitors can effectively treat autoimmune diseases such as rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) in mice model studies.9,10 The significant function evidence of B cells in immune dysregulation in RA was demonstrated by observing clinical improvement in RA patients receiving B-​cell depletion antibody therapy (e.g., rituximab). So far, no small-​molecule BTK inhibitor has been approved for the treatment of autoimmune diseases in patients. Non-​Hodgkin’s lymphoma (NHL) is a type of cancer derived from lymphocytes. It has a high incidence in the world and is located in the top ten of malignant tumors. In 2015, 4.3 million people worldwide suffered from the disease, and about 230 000 people died of it. According to the American Cancer Society, about 66 000 new cases of NHL appear each year in the United States.11 Figure 8.2 shows the incidence of lymphoma in different races; lymphoma patients can be found in almost all parts of the world.12 NHL is mainly divided into three basic types: B lymphocytes, T lymphocytes and natural killer cells, of which NHL derived from B lymphocytes accounts for about 80% of the total. As shown in Figure 8.3,16 DLBCL is the most common

Figure 8.2 Incidence of lymphoma in different races (five races are counted).

150

150

Chapter 8

Figure 8.3 Classification of non-​Hodgkin’s lymphoma.

type of NHL, accounting for about 30% of the total. In the United Kingdom and the United States, there are 7–​8 cases of DLBCL per 100 000 people. Also, follicular lymphoma accounts for a large proportion of NHL. This section will summarize original works of proteolysis-​targeting chimeras (PROTACs) with BTK degradation function, and the biological function evaluation of BTK PROTACs. (1) The newly prepared PROTAC small molecule can selectively bind to BTK and E3 ubiquitinated ligase, then the spatial distance between the two proteins will be closer. The purpose of ubiquitination is degradation of BTK. (2) To clarify the binding site of PROTAC small molecule and the mechanism of BTK degradation. (3) To explore the functional experiment of the PROTAC small molecule in inhibiting BTK kinase-​dependent lymphoma cancer cells. (4) The BTK PROTAC small molecule can be applied in treating ibrutinib-​resistant lymphoma cells and overcoming the side effects of ibrutinib. (5) Evaluation of antitumor function and pharmacokinetics of PROTACs targeting BTK at the animal level.

8.2 Review of BTK Inhibitors As mentioned above, BTK belongs to non-​receptor tyrosine protein kinase and plays a significant role in the upstream part of the BCR pathway. BTK is essential in immune escape in ABC-​like DLBCL. ATP binding sites of kinases are highly conserved, so it is quite difficult to develop highly selective ATP competitive inhibitors. During the efforts to develop BTK inhibitors, it was found that a class of covalent inhibitor with electron-​donating groups can significantly improve the selectivity. Ibrutinib is the first drug approved as BTK covalent inhibitor. Ibrutinib demonstrates high selectivity, strong activity, and good oral bioavailability. The most prominent

 151

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 151

feature and contribution of ibrutinib was the proposal of acrylamide as a “warhead.”13 A “warhead” can selectively enter the ATP-​binding pocket of BTK, and a Michael addition reaction occurs between the thiol group of 481 cysteine of BTK and the double bond of acrylamide. The purpose is to permanently inactivate the BTK. The IC50 value of ibrutinib was 0.5 nM at the kinase level.14 The earliest pharmacodynamic study of ibrutinib was performed in a mouse model with RA. The study found that ibrutinib was effective in treating rheumatoid arthritis at a dose of 3 mg kg–​1, although the pharmacodynamics of ibrutinib is mainly focused on the treatment of cancer later.13 In 2013, the US FDA approved ibrutinib for the treatment of refractory, repetitive mantle cell lymphoma (MCL). Ibrutinib was further approved for the treatment of chronic lymphocytic leukemia (CLL) and Waldenström macroglobulinemia (WM) in 2014 and 2015, respectively. In August 2017, the US FDA approved ibrutinib for the treatment of graft-​versus-​host disease (GVHD). Due to the tremendous success in clinical applications, global sales of ibrutinib were $6.205 billion in 2018.15 However, the shortcomings of ibrutinib are still obvious. Due to the off-​target effects, ibrutinib has been reported to cause a series of side effects including joint pain, muscle pain, atrial fibrillation, ecchymosis and major bleeding.17 In addition, the dose of ibrutinib is large, about 560 mg per day. Ibrutinib has a relatively short half-​life in vivo. Therefore, development of new BTK inhibitors is still necessary. In addition to ibrutinib, there are many different types of BTK covalent inhibitors with different skeletons that have entered clinical trials. As shown in Figure 8.4, spebrutinib (formerly AVL-​292) has a pyrimidine ring skeleton similar to ibrutinib. Diphenyl ether is replaced by ethylene glycol binding to the pocket of the BTK hydrophobic zone. The “warhead” portion of spebrutinib retains the same structure as that of ibrutinib. In terms of bioactivity, spebrutinib inhibits proliferation of B lymphocytes by inhibiting BTK, and its EC50 could reach 3 nM.18 Spebrutinib has the advantage of being capable of increasing selectivity (>1400-​fold) to other kinases, thereby reducing the toxicity. In addition to B lymphocytes, BTK proteins are also expressed in monocytes and play a crucial role in the activation of Fcγ receptors such as FcγRIII and FcγRIIa.19 In mast cells and basophils, BTK can also regulate the expression of inflammatory cytokines, chemokines and cell-​adhesion factors by activating the FcεRI signaling pathway, thereby regulating the inflammatory response.20 At present, spebrutinib has entered the clinical phase II stage, and the main indication is to treat rheumatoid arthritis.21 Acalabrutinib17 is a new generation of BTK covalent inhibitors developed by Acerta. Chemical structure of acalabrutinib is similar to that of ibrutinib, and the major difference is that the “warhead” portion has changed from “olefin” to “alkyne.” Compared with ibrutinib, the application of acalabrutinib is characterized by increased binding affinity and selectivity, which significantly reduces the incidence of side effects. In October 2017, the US FDA approved acalabrutinib for the treatment of MCL. Among the BTK covalent inhibitors, ONO-​405922 and BGB-​311123 developed by Beigene have been included in clinical trials. The structures retain

152

152

Chapter 8

Figure 8.4 Examples of BTK covalent inhibitor.

a “warhead” part. The interaction mechanism is to covalently cross-​link with the sulfhydryl group of BTK 481 cysteine, thereby inhibiting the catalytic function of BTK. Co-​crystals of BTK with various inhibitors have been reported. These co-​crystal structures include 5P9J, 5P9H, 5P9M, 5P9L, 4OTF/​5P9F, 5P9G, etc. These structures provide important information for understanding the structure and function of BTK.

8.3

Drug Resistance and Side Effects of Ibrutinib

As mentioned, a variety of covalent inhibitors have been developed for BTK, which are characterized by high potency and specificity compared to traditional small-​molecule inhibitors. However, since the approval of ibrutinib by FDA in 2013, drug-​resistant patients have frequently appeared in clinical cases. Many mechanisms can lead to ibrutinib resistance, the most important of which is that the BTK has a point mutation in the C481.24 In 2014, the New England Journal of Medicine reported the clinical manifestations and the mechanism of ibrutinib resistance. Resistance to ibrutinib has been found in a variety of lymphomas, including MCL and CLL. As ibrutinib could not covalently bind to the hydroxyl group of serine, the IC50 value

 153

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 153

Figure 8.5 Effect of C481S mutation of BTK on ibrutinib binding and the ability of ibrutinib to inhibit BTK phosphorylation.

of ibrutinib in inhibition of BTK phosphorylation decreased from 2.2 nM to 1 μM (Figure 8.5).25 Therefore, there is an urgent need to develop a new strategy to solve the resistance caused by C481S mutation. In addition, ibrutinib has been reported to cause a range of serious side effects including joint pain, muscle pain, atrial fibrillation, ecchymoses and major bleeding. Because ibrutinib has potent inhibitory effects on EGFR, ITK and TEC, the IC50 is in the range of 10–​100 nM. The appearance of side effects of ibrutinib has been shown to be associated with off-​target effects of ibrutinib.17

8.4 PROTAC Contributes to Overcome Drug Resistance of Ibrutinib Proteolysis-​targeting chimera (PROTAC) is a novel chemical technique for selective degradation of target proteins in cells.27 The PROTAC molecule is able to pull the target protein close to the E3 ligase, ultimately leading to degradation of the target protein (Figure 8.6). Such dual-​targets molecules are composed of three parts: a target protein-​binding region (targeting arm, TA), an E3 ligase-​binding region (degradation arm, DA), and a chain to combine the two parts. This ubiquitin–​proteasome system (UPS) is capable of recruiting E3 ligase to ubiquitinate the target proteins, which can be subsequently degraded by proteasome. Unlike traditional inhibitors, the PROTAC technology aims to eliminate whole functions of proteins, rather than merely inhibiting their enzymatic activity. Therefore, in theory, the resistance caused by the BTK C481S mutant can be overcome by a PROTAC. To date, five groups have reported the development of BTK PROTAC molecules independently, which will be introduced in the following portion of this chapter.

154

154

Chapter 8

Figure 8.6 Degradation of BTK via PROTAC technology.

In 2018, Rao and coworkers28 from Tsinghua University reported BTK degradation technology and successfully developed the BTK degrader. These PROTAC molecules were capable of efficiently degrading wild-​type BTK. Also, the newly designed PROTAC molecule was effective in degrading the ibrutinib-​resistant BTK (50% degradation at 30 nM). What’s more, the newly synthesized PROTAC molecule effectively inhibited colony formation and cell proliferation, but had no inhibitory or degrading activity on ITK, TEC or EGFR. These proteins are the main targets of ibrutinib causing side effects. These data demonstrated great potential of PROTAC technology as a new therapeutic strategy. To develop BTK degraders, the BTK “targeting arm” was linked to the E3 ligase “degradation arm” by chains with different lengths (Figure 8.6). Ibrutinib and spebrutinib were selected as BTK ligands, and pomalidomide and RG-​7112 were selected as E3 ligase ligands. This study found that CRBN-​based PROTAC molecules were generally more active than MDM2-​based molecules. Among these CRBN-​based PROTACs, compound P13I linked by ibrutinib and pomalidomide demonstrated the strongest degrading activity. In human Burkitt’s lymphoma RAMOS cells, P13I induced 73% degradation of BTK at 10 nM. Under the treatment of P13I, the half-​lives of BTK and C481S mutant BTK were 4 hours and 3 hours, respectively. The experimental result proved that P13I can significantly accelerate the degradation of BTK (see the supplemental information in the article for details). More importantly, P13I was able to effectively degrade BTK in other non-​Hodgkin’s lymphomas, such as MCLL (Mino) cells and multiple myeloma cells. The DC50 was around 10 nM (see Figure 8.7). Control experiments showed that ibrutinib, pomalidomide or PROTAC arms (“targeting arm” and “degrading arm”) did not trigger degradation. On the other hand, ibrutinib or pomalidomide could competitively inhibit the effect of PROTACs. Moreover, proteasome inhibitor was able to completely inactivate PROTAC. A similar competitive inhibition phenomenon was also observed with the involvement of MLN-​4924 (a NAE inhibitor). These findings clearly indicate that BTK degradation is mediated by the UPS. In order to

 155

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 155

Figure 8.7 Degradation of wild-​type BTK and C481S mutant BTK.

evaluate the effect of P13I in vitro, the HBL-​1 cell line (ABC-​DLBCL) was used in this study because survival and growth of the cells depend on the activation of BTK. The results showed that GI50 of P13I was 1.5 nM (Figure 8.7). IC50 value of ibrutinib for inhibition of BTK phosphorylation was 0.5 nM, which was much lower than that of P13I (95 nM). This showed that the effect of P13I was by degradation of BTK rather than inhibition. Because the survival and growth of RAMOS cells do not rely on BTK, P13I or ibrutinib did not show cytostatic effect, suggesting that P13I may not have any general toxicity. As previously mentioned, according to the clinical case reports, the C481S mutation of BTK is one of the main reasons accounting for ibrutinib resistance. In this study, a recombinant mutant BTK plasmid was transfected into 293T cells and HeLa cells. P13I effectively degraded BTK C481S mutant at a concentration below 30 nM. However, ibrutinib did not inhibit autophosphorylation

156

156

Chapter 8

of the BTK C481S mutant even at a concentration of 1 μM. It has been reported that the BTKCys481Ser DLBCL cell line can be used as a model to study the resistance of ibrutinib. Indeed, the growth of HBL-​1 cells with BTK C481S mutant could be inhibited by P13I with GI50 under 30 nM, while ibrutinib was ineffective on this cell line with GI50 of approximately 700 nM (Figure 8.7). These data strongly prove that the PROTAC targeting BTK might provide a new practical method for the treatment of ibrutinib-​resistant tumor. In addition to BTK, ibrutinib can inhibit EGFR, TEC and ITK kinases at low concentrations, leading to serious side effects. To examine the side effects of the BTK degrader, this study analyzed the interaction of P13I with these proteins in a variety of cell lines. Even at a concentration of 5 000 nM, P13I did not induce protein degradation. Moreover, P13I has been shown to have weak inhibitory activity (>1 000 nM) on EGFR or ITK, suggesting that BTK degraders are unlikely to cause the side effects of ibrutinib. These results demonstrated that the BTK degrader P13I has important clinical potential in the treatment of B-​cell malignancies. For further research on BTK degraders, the Rao research group collaborated with the Zhu group from Peking University Cancer Hospital and the Liu group from Tsinghua University in the field of lymphomas and in vitro and in vivo study results have been published.34 In this study, BTK degrader P13I was further optimized, and the E3 ligase ligand was replaced by lenalidomide. The solubility and permeability of the new degrader L18I were largely improved (Figure 8.8). L18I successfully and efficiently degraded a variety of clinically relevant BTK mutants. L18I can overcome the resistance of ibrutinib caused by mutation of BTK in NHL cell lines. More importantly, the effectiveness of the new strategy to fight against drug resistance has been validated in vivo (Figure 8.9). In this study, it was found that BTK degrader L18I could efficiently degrade different BTK mutants (C481S/​T/​A/​G/​W) in HeLa cells at a concentration less than 50 nM. The proliferation of DLBCL and MCL cells expressing C481S BTK in vitro was effectively inhibited by compound L18I, while ibrutinib showed

Figure 8.8 Development of a new BTK degrader.

 157

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 157

Figure 8.9 Degradation of BTK in vivo and antitumor results of L18I.

pretty weak efficacy. Remarkably, L18I caused a decrease in the level of C481S BTK in DLBCL tumors, thereby effectively inhibiting tumor growth in vivo. In contrast, ibrutinib had nearly no inhibitory effect in vivo (Figure 8.9). The new strategy overcame the tumor resistance to ibrutinib due to BTK C481 mutation. In addition, downstream signals of BTK were also tested in this study, such as PLC-​γ2, p38 and ERK1/​2. L18I strongly blocked downstream signals with an effective concentration below 10 nM. Therefore, the reasons for the antitumor effect of the new strategy were explained at the mechanism level. Additionally, compound L18I was found to be safe and well tolerated. In addition to the potential treatment of DLBCL, L18I also effectively inhibited the growth of MCL cell lines (Mino and Z138), and its effect was much better than that of ibrutinib. Another highlight of this study was the combination of PROTAC molecules with other small-​molecule inhibitors, such as copanlisib, dasatinib, etc., which greatly enhanced the antitumor effect of the PROTAC strategy. In summary, Rao and co-​authors firstly developed a new PROTAC strategy for the degradation of ibrutinib-​resistant BTK. The PROTAC molecule L18I

158

158

Chapter 8

is a highly potent antitumor compound in vitro for ibrutinib-​resistant lymphoma expressing C481S BTK. Furthermore, L18I significantly inhibited the proliferation and potential colony formation of C481S BTK type HBL-​1 tumor (ibrutinib resistance) in vivo. Moreover, the novel BTK degrader was almost ineffective against ITK, EGFR or TEC kinases, which are known off-​targets for ibrutinib causing serious side effects. These findings provide new opportunities to develop more effective therapies for ibrutinib-​resistant NHL. More importantly, the results suggest that future PROTAC strategies (protein degradation rather than inhibition) can serve as a common and powerful treatment strategy for other drug-​resistant tumors. In 2018, Crews’ team at Yale University elegantly achieved the inhibition of wild-​type and C481S mutant BTK via PROTAC strategy (Table 8.1).29 It was expected that PROTAC molecules could be superior to the inhibitors in a certain system and degraders were used in drug-​resistant CLL cell lines in this study. MT-​802 was superior to ibrutinib in inhibition of BTK activation. There have been reports of PROTAC molecules based on different ligands, so comparing these molecules in wild-​type and C481S mutant systems might help to understand which tertiary complexes were more favorable for degradation. The results indicated that only the PROTAC molecule with CRBN ligand could induce BTK degradation, but the reason remained unclear. There are many possibilities for CRBN to capture BTK, which causes the formation of the target protein and E3 ubiquitin complex. The possibility may be due to the kinetics of the tertiary complex or the stereoscopic orientation of lysine. A critical and interesting feature of MT-​802 was that it did not have a hook effect due to the formation of secondary complexes (BTK−MT-​ 802 and MT-​802−cereblon) at high doses. Currently reasonable principles are still lacking to predict when a hook effect will occur. It was proposed that a possible reason for delay of the hook effect might be that MT-​802 could well coordinate the interaction between BTK and CRBN. It also depended on the length of the linker and the coordination of the ligands. When the ligation position of pomalidomide was changed from C4 to C5, the degradation efficiency improved, as this angle may further coordinate the formation of the complex. This study reported the first PROTAC molecule constructed from Table 8.1 Inhibition of wild-​type and C481S BTK via PROTAC technology.

BTK Variant DC50 of MT-​802 (nM)

MT-​802

WT C481S

Inhibition and degradation Inhibition and degradation Inhibition and degradation No inhibition or degradation

14.6 14.9

Ibrutinib

 159

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 159

the C5 position of the pomalidomide, but whether this can be employed to other targets should be further confirmed. Another interesting question raised by the authors concerns potential resistance to PROTAC. The first possibility is the mutation of the ATP-​binding pocket of BTK, such as T474I, L528W or C481R, but the probability of these mutations is much lower than C481S. A decrease in the binding capacity of the binary complex should result in a decrease in the activity of PROTAC, but this hypothesis has not been well understood and could be dependent on the specific target. Therefore, PROTACs may remain effective in an inhibitor drug resistance system caused by point mutations. The UPS may also be mutated to resist the effect of PROTACs. But so far, there is no reported example of this. The reason might be because the UPS plays a central role in the metabolism and survival of cells. Further research on this aspect will help to understand the clinical application of PROTAC. Compound MT-​802 could efficiently induce BTK degradation at nanomolar level within 4 hours in immune cells. It was observed that PROTAC molecules also had fewer off-​target effects, while ibrutinib could bind almost all homologous proteins of BTK. Although MT-​802 could bind to CRBN, no degradation of IKZF1 or IKZF3 was observed. MT-​802 also achieved efficient degradation of BTK in patient-​derived CLL cells, and this degradation occurred within 2 hours after administration. C481S BTK mutant could also be efficiently degraded by MT-​802, thereby reducing activation of the BTK pathway, while ibrutinib had no effect on mutant cell lines. In summary, this study found that reducing BTK levels by PROTAC could effectively inhibit the activation of the BTK pathway in C481S mutant cell lines, which was common in ibrutinib treatment. To apply MT-​802 as a therapeutic candidate in vivo, further optimization might be needed to improve its pharmacodynamic parameters. Degradation in this work provides a viable strategy for the study of recurrent CLL. When a small-​molecule inhibitor is ineffective in occupying the pocket, it can be transformed into a PROTAC molecule. The target protein can be recruited and degraded by the instantaneous and weak binding force of the PROTAC molecule, which should be able to overcome the drug resistance of covalent inhibition. Pfizer also reported a PROTAC study for BTK degradation in the same year.30 Although significant advances have been made in the PROTAC field, there is a lack of thorough mechanism characterization for the biochemical determinants that support the effectiveness of PROTAC. Therefore, this study used a range of cellular and biochemical techniques to study the requirements for PROTAC design, and found that in the BTK/​CRBN PROTAC system, in the absence of thermodynamic cooperation, effective knockout is related to the mitigation of spatial conflict. These data broaden the scope of application of PROTAC and provide insights for the basic design criteria of PROTAC. For the potential therapeutic application of PROTAC, it is important for researchers to understand the parameters that determine corresponding effectiveness and selectivity. Therefore, design and optimization of linker is an essential feature that must be extensively explored. This study used BTK/​CRBN as a study case to illustrate these parameters. This work involved

160

160

Chapter 8

11 PROTAC small molecules as a library to form binary and ternary complexes with BTK and the E3 ligase CRBN. As a result, the synergistic effect of BTK-​CRBN and the degradation of BTK in vitro and in vivo were examined by authors. The data showed that in the absence of thermodynamic synergy, the spatial conflict between BTK and CRBN could be alleviated by adjusting the length of linker, then effective BTK degradation could be achieved. Firstly, it is found that efficient, rapid and long-​term knockout of BTK could be achieved via PROTACs, following the characteristics of triplet interactions. The PROTAC compound libraries were divided into two categories: short-​chain PROTACs 1–​4 that were ineffective in vitro, and long-​chain PROTACs 6–11 that were both effective and efficient. Surprisingly, such compounds were closely related to the structure of the triplet: shorter PROTAC did not appear to bind CRBN and BTK at the same time. Although triplet analysis allowed classification of compounds into active and inactive groups, it was not enough for strict titer ranking. Although compounds 9 and 11 seem to produce the highest peak and the most effective complex formation, compound 10 showed the strongest BTK knockout activity at the cellular level. This might show the limitations of fluorescence resonance energy transfer (FRET) analysis (Table 8.2). Although compounds 6–​11 appeared to be very effective, no significant positive synergy was observed in triplet by authors. The affinity of PROTAC 10 to CRBN was similar in the absence or presence of BTK. Furthermore, the steep nature of the hook effect was consistent with the low synergistic system as it indicated that the binary interaction had the ability to compete with ternary ones. Compared with short-​chain 1–​4, long-​chain 6–​11 PROTACs significantly improved the effectiveness, mainly because the negative synergy was Table 8.2 PROTAC molecules to degrade BTK.

Compound Length of linker as PEG (number of atoms)

Max. TR-​FRET (in vitro BTK:PROTAC:CRBN ternary complex formation)

1 2 3 4 5 6 7 8 9 10 11

0.7 >0.9 >0.8 >0.8

3 4 6 7 9 12 13 15 16 18 19

 161

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 161

alleviated as the linker expanded. In this hypothetical extreme test, a long-​ chain PROTAC 16 containing 29 linker atoms was demonstrated to remain the activity for BTK degradation. According to a report describing the BRD4 degradation via VHL ligase,31 a concise mechanism was proposed that new interactions between E3 and the target could be promoted in the triplet. The interactions produced substantial positive synergy (α value up to 17.6), which resulted in knockout of BRD family members with high validity and selectivity. Although positive synergies may affect the function of PROTAC in many aspects (including improved activity, expansion of hook effects, and possible selectivity), current work on BTK showed that it was not a fundamental requirement for achieving effectiveness. It is worth noting that a lack of positive synergy did not imply that the E3 ligase and the target don’t interact in the triplet. In fact, this interaction was likely to occur, but the energy obtained by the protein–​protein interaction was offset by the entropy cost of reduced ligand flexibility. In this study, hydrogen–​ deuterium exchange (HDX) experiments indicated that PROTAC was not likely to form a stable, single and rigid triplet, but perhaps a limited set of conformations, although CRBN and BTK might interact in ternary complexes. Apparently, the degradation efficiency was closely related to the type of E3 ligase. CRBN PROTACs were effective in BTK degradation in the case where other variables remained constant, such as the warhead of inhibitor, composition and length of linker, ligase cell colocalization. VHL PROTACs did not lead to degradation. One reason first proposed by Crews in a review32 was that the CRBN ligase backbone had greater flexibility and torsional. This feature not only allowed nesting of different substrates in the core of E3 ligase, but also provided a broader pathway for surface Lys to the E2. XIAP is a non-​cullin E3 loop ligase; compared with CRBN, it mimics VHL in terms of limited spatial sampling. Although PROTAC was not strictly related to positive synergy, the activity was still dependent on the environmental background where it was located, as Ramos cells were more sensitive than THP-​1s. Compared to the primary cells (PBMC), activity of PROTACs was between THP-​1 and Ramos with moderate sensitivity. The expansion of the data in vivo showed a similar question; a dose-​dependent BTK decrease was observed in the spleen rather than in the lung when subcutaneously dosing compound 10. Although the PROTACs achieved similar levels between the two tissues, apparent tissue selectivity still existed. The reason is as yet unclear, but might be due to differences in E3 or target expression, differences in activity of deubiquitinase, or other unknown determinants. Further study of the reasons is necessary and critical to predict and understand differences in activity of PROTAC between tissues. It is also quite significant for understanding the safety and efficacy profile of the degraders (Table 8.3). In 2019, the Gray research group in Harvard University reported developing BTK PROTAC technology for possible treatment of B-​cell lymphoma.33 They analyzed the degradable kinase group and found BTK as a target easy to be degraded. This study found that the CGI1746-​CRBN-​based assembly was effective in inducing BTK degradation in cells (Table 8.4). Therefore, BTK appeared to be a kinase that was susceptible to degradation. Gray and

162

162

Chapter 8

Table 8.3  In vivo distribution of PROTAC molecules targeting BTK.

PROTAC (10) Tissue Concentration (ng mL–​1) Spleen Lungs

0.35 mg kg–​1

35 mg kg–​1

175 mg kg–​1

~50 ~20

~700 ~ 600

~ 3000 ~ 3000

coworkers developed the lead compound DD-​03-​171, which was capable of triggering the degradation of IKZF1/​3 and BTK, three proven targets for treatment of B-​cell malignancies. Promising preliminary results have been reported for the combination of rituximab, lenalidomide and ibrutinib. In theory, targeting multiple proteins by one molecule could avoid drug interactions and provide more predictable parameters in terms of drug combinations and simplicity of administration to patients. This study envisaged that this multi-​pharmacological method of degrading IKZF1 and IKZF3 could be used as a strategy for a variety of cancers. To date, multiple BTK ligands have been transferred into BTK degraders. This study demonstrated that the newly developed compound DD-​03-​171 had acceptable pharmacokinetic parameters. According to publications, DD-​03-​ 171 was the first BTK degrader with in vivo activity against B-​cell malignancies. However, multiple factors should be confirmed when weighing the efficiency and risks of BTK degradation. Firstly, PROTAC has not achieved optimal pharmacokinetic parameters in vivo. According to a recent publication, although the compound achieved similar levels of exposure between lung and spleen, apparent tissue selectivity still existed, which might be caused by the difference in CRBN levels. Second, ibrutinib has various side effects that may contribute to its role in specific situations. When using a triple degrader such as DD-​03-​171, it is important to consider potential side effects. Third, B-​cell lymphoma can produce resistance to BTK degraders in a variety of ways, including reducing absorption, increasing efflux, mutations in BTK binding sites, activating PLCγ2, NF-​κB mutations, or affecting functions of E3 ligase/​protease. Finally, the development of normal B cell depends on BTK. This suggested

 163

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 163 Table 8.4 Structures of PROTAC molecules and BTK degradation effects in vivo.

BTK degradation (in vivo) 100% survival proportions days

Vehicle

Ibruntinib

Lenalidomide DD-​03-​171

–​ ~25

–​ ~25

–​ ~31

+ ~48

that this degrader might cause severe B-​cell lymphocyte reduction. These patients need long-​term intravenous supplementation of immunoglobulin and have risk of reactivation of HBV infection. In summary, this study developed novel small-​molecule BTK degraders that exhibited a few positive therapeutic features, including (1) long-​lasting degradation and effective inhibition of signaling and antiproliferative activity; (2) overcoming clinically relevant resistance of ibrutinib; and (3) the efficacy in PDX model. Therefore, BTK degradation is a prospective therapeutic method for the treatment of B-​cell malignancies. In 2019, the GSK group reported their PROTAC research about BTK degradation through covalent binding mode.35 A catalytic mechanism of PROTAC-​ mediated protein degradation was believed to be the basis for the degradation efficiencies observed in vitro and in vivo. This catalytic mode provided an opportunity to deliver new drugs at low doses. However, covalent PROTAC might also act as a stoichiometric degrader to achieve degradation of one protein molecule by one PROTAC molecule. To prove the hypothesis, GSK researchers tested the effect of covalent binding on PROTAC-​mediated degradation of BTK. This was important because if the binding affinity to the target protein could be enhanced by using covalent binding, designing PROTAC for a less tractable target would be a new solution to the “undruggable” proteomes. In this study, ibrutinib was used as a BTK ligand and a set of structurally similar covalent and non-​covalent BTK-​targeted PROTAC molecules were

164

164

Chapter 8

designed using the E3 ligase ligand of IAP : covalent PROTAC 2 and reversible PROTAC 3. The in vitro results showed that the covalent PROTAC molecule could not achieve the degradation of BTK, and the reversible PROTAC molecule could achieve the catalytic mechanism of BTK degradation. In addition, the authors examined the level of phosphorylated BTK activated by the BCR pathway in Ramos cells and the level of cIAP1. The results showed that covalent PROTAC 2 fully occupied the BTK kinase domain, and theoretically stoichiometric BTK degradation should be achieved. However, the observed fact was no degradation of BTK at all, indicating that the reason for non-​ degradation was not due to the lack of catalysis of PROTAC. Next, the authors used the ligand of CRBN E3 ligase to synthesize a set of CRBN-​mediated covalent and non-​covalent PROTAC molecules: covalent PROTAC 4 and reversible PROTAC 5. It was found that PROTAC 4 could not degrade either mature or newborn BTK, but PROTAC 5 could degrade both of them. Furthermore, the authors found that PROTAC 4 effectively degraded the mature and nascent proteins CSK and LYN lacking the conserved cysteine in the SRC kinase. Taken together, the authors found that the degradation of the target protein by the covalent PROTAC molecule could be affected by a variety of factors within the cell, such as the ubiquitin transfer blocking, or the proteasome recognition hindering (Figure 8.10).

8.5 Summary BTK has been proved to be an important drug target in the treatment of non-​ Hodgkin’s lymphoma. Ibrutinib, as a covalent inhibitor of BTK, is a first-​line drug for the treatment of various B-​cell malignancies. However, the emergence of the BTK C481S mutation has become a major challenge for current clinical treatments. Proteolysis-​targeting chimera (PROTAC) represents a novel technique for the selective degradation of target proteins with small molecules. Ibrutinib only loses covalent binding to the BTK C481S mutant and does not completely lose its binding capacity, so its skeleton can still be used in the construction of PROTAC molecules. Unlike traditional kinase

Figure 8.10 Covalent inhibitors block the effect of PROTAC.

 165

PROTAC Targeting BTK for the Treatment of Ibrutinib-resistant B-cell Malignancies 165

inhibitors, PROTAC technology aims to eliminate the whole function of proteins, rather than merely inhibit some of their activities. Therefore, in theory, the resistance caused by the BTK C481S mutant can be overcome by PROTAC technique. After strict screening, a variety of new BTK-​selective degraders have been successfully developed by different groups. The new PROTAC molecules could efficiently degrade a variety of clinically relevant BTK mutants (C481S, C481T, C481A, etc.) and overcome the clinical resistance to the first-​ line drug ibrutinib. More significantly, the effectiveness of the new strategy for overcoming tumor resistance has been verified in vivo. It has been demonstrated that the novel BTK degrader has safety for in vivo experiments, and the new BTK degrader could avoid the side effects of traditional covalent inhibitors (no degradation or inhibition on TEC, EGFR, ITK). As a newly developed small-​molecule degrader that simultaneously degrades BTK and IKZF1/​3, it has many advantages. Targeting multiple proteins by one molecule could avoid drug interactions and provide more predictable parameters in terms of drug combinations and simplicity of administration to patients. This multi-​ pharmacological method of degrading IKZF1 and IKZF3 could be used as a strategy for a variety of cancers. Although significant progress towards BTK-​targeting PROTACs has been achieved, many critical questions remain to be addressed in the future. These challenges are summarized as follows. (1) PROTAC has not achieved optimal pharmacokinetic parameters in vivo. Although the compound achieved similar levels of exposure between lung and spleen, apparent tissue selectivity still existed; (2) ibrutinib has multiple side effects. When using a triple degrader such as DD-​03-​171, it is important to consider potential side effects; (3) lymphoma can produce resistance to BTK degraders in a variety of ways, including reducing absorption, increasing efflux, mutations in BTK binding sites, activating PLCγ2, NF-​κB mutations, or affecting functions of E3 ligase/​protease; (4) development of a normal B cell depends on BTK. This suggested that this degrader might cause severe B-​cell lymphocyte reduction. Despite all these considerations, as a new treatment option, application prospects of BTK degradation strategy are quite optimistic. Until now, there has been no UPS mutation to resist the PROTAC molecule, probably because the UPS plays a central role in cell metabolism and survival. Research on this will be more conducive to the clinical application of PROTAC. In the future, through rigorous screening and chemical modification of BTK PROTAC molecules, the optimal physical and chemical properties can be achieved. Deficiencies of pharmacokinetic parameters will be overcome. Advantages of degraders relative to inhibitors will be fully shown. Not only B-​cell malignancies, PROTAC can also be used in the treatment of other drug-​resistant tumors, including the resistance of C797S EGFR mutation in non-​small-​cell lung cancer to three-​generation covalent inhibitors. When a small-​molecule inhibitor is ineffective in occupying the pocket, it can be transformed into a PROTAC molecule. The target protein can be recruited and degraded by the instantaneous and weak binding force of the PROTAC molecule, which can overcome the drug resistance of both covalent and non-​covalent inhibitions.

166

166

Chapter 8

8.6 Acknowledgments We thank Hongying Gao and Zimo Yang in Yu Rao Laboratory from Tsinghua University for their contributions in collecting original information.

References 1. X. Li, et al., J. Med. Chem., 2014, 57(12), 5112–​5128. 2. C. Liang, et al., Eur. J. Med. Chem., 2018, 151, 315–​326. 3. Z. Huang, et al., Bioorg. Med. Chem. Lett., 2016, 26, 1954–​1957. 4. Y. Xu, et al., Cell Res., 2014, 24, 651–​664. 5. J. Zhang, et al., Cancer Res., 2015, 75, 2599. 6. B. Wang, et al., Eur. J. Med. Chem., 2017, 137, 545. 7. J. J. Buggy and L. Elias, Int. Rev. Immunol., 2012, 31, 119–​132. 8. S. Pal Singh, F. Dammeijer and R. W. Hendriks, Mol. Cancer, 2018, 17(1),  57–​79. 9. L. J. Crofford, et al., Expert Rev. Clin. Immunol., 2016, 12(7), 763–​773. 10. D. Xu, et al., J. Pharmacol. Exp. Ther., 2012, 341, 90. 11. L. M. Morton, et al., Blood, 2006, 107, 265–​276. 12. https://​seer.cancer.gov/​archive/​csr/​1975_​2010/​ 13. Z. Pan, et al., ChemMedChem, 2007, 2(1),  58–​61. 14. L. A. Honigberg, et al., Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 13075–​13080. 15. http://​www.medsci.cn/​article/​show_​article.do?id=b7b516085428 16. The 2012 Oncology Nurse Hematology Conference. 17. J. C. Byrd, et al., N. Engl. J. Med., 2016, 374, 323–​332. 18. E. K. Evans, et al., J. Pharmacol. Exp. Ther., 2013, 346, 219–​228. 19. G. V. De Lucca, et al., J. Med. Chem., 2016, 59, 7915–​7935. 20. H. S. Kuehn, et al., J. Cell Sci., 2010, 123, 2576–​2585. 21. https:// ​ w ww.clinicaltrials.gov/ ​ c t2/ ​ s how/ ​ N CT01975610?term= cc+292&rank=2 22. T. Yasuhiro, et al., Blood, 2013, 122, 5151. 23. Y. Guo and Z. Wang, Preparation of fused heterocyclic compounds as protein kinase inhibitors: WO, 2014173289, 2014. 24. J. A. Woyach, et al., N. Engl. J. Med., 2014, 370(24), 2286–​2294. 25. R. R. Furman, et al., N. Engl. J. Med., 2014, 370, 2352–​2354. 26. A. T. Bender, et al., Mol. Pharmacol., 2017. 91(3), 208–​219. 27. K. M. Sakamoto, et al., Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 8554–​8559. 28. Y. Sun, et al., Cell Res., 2018, 28, 779–​781. 29. A. D. Buhimschi, et al., Biochemistry, 2018, 57(26), 3564–​3575. 30. A. Zorba, et al., Proc. Natl. Acad. Sci. U. S. A., 2018, 115(31), E7285–​E7292. 31. M. S. Gadd, et al., Nat. Chem. Biol., 2017, 13, 514–​521. 32. A. C. Lai and C. M. Crews, Nat. Rev. Drug Discovery, 2017, 16, 101–​114. 33. D. Dobrovolsky, et al., Blood, 2019, 133(9), 952–​961. 34. Y. Sun, et al., Leukemia, 2019, 33(8), 2105–​2110. 35. C. P. Tinworth, et al., ACS Chem. Biol., 2019, 14(3), 342–​347.

 167

CHAPTER 9

An Efficient Approach Toward Drugging Undruggable Targets KANAE GAMO,a NAOMI KITAMOTO,a,b MASATO T. KANEMAKIb,c AND YUSUKE TOMINARIa* a

FIMECS, Inc. 26–​1, Muraoka-​Higashi 2-​chome, Fujisawa, Kanagawa 251–0012, Japan b Department of Chromosome Science, National Institute of Genetics, Research Organization of Information and Systems (ROIS), Yata 1111, Mishima, Shizuoka 411–​8540,  Japan c Department of Genetics, The Graduate University for Advanced Studies (SOKENDAI), Yata 1111, Mishima, Shizuoka 411–​8540, Japan *Corresponding author. Email: [email protected]

9.1 Introduction Targeted protein degradation (TPD) has gained increasing attention as next-​generation drugs aim to target currently undruggable proteins. Hetero-​ bifunctional degraders, known as proteolysis-​targeting chimeras (PROTACs), are an attractive novel modality.1,2 These degraders consist of a protein of interest (POI) binder and an E3 ubiquitin ligase recruiting binder connected by a linker, which hijacks the ubiquitin–​proteasome system and leads to protein degradation that modulates the protein levels. The degraders induce the formation of a ternary complex with a POI and an E3 ligase for subsequent POI ubiquitination. The polyubiquitinated POI is then recognized and degraded

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

167

168

168

Chapter 9 3,4

by the proteasome. Based on this mode of action (MOA), the TPD has certain advantages over traditional drugs such as small molecule inhibitors. Owing to the catalytic nature, an effective concentration of protein degraders would be lower than that of conventional inhibitors. In addition, low-​affinity ligands might be sufficient for achieving pharmacological effects because the degraders’ efficiency depends on the cooperativity of the protein–​protein interaction (PPI) between the POI and the E3 ligase.5–​7 Therefore, TPD is an attractive modality to address currently undruggable targets, such as pseudokinases, scaffolding proteins, transcription factors and phosphatases. However, several challenges for therapeutic application of TPD are yet to be addressed. The bivalent molecules exhibit poor drug-​like properties, including cell permeability, metabolic stability, tissue distribution and bioavailability due to their high molecular weight, polar surface areas and a number of rotatable bonds and hydrogen bond acceptors and donors.8 Moreover, in addition to the importance of the length, direction and composition of the linker, the structure, anchor point and property of the E3 ligase binder play a major role in TPD efficiency. Intriguingly, the process of synthesis is prolonged due to the middle size of molecular weight. Therefore, as the drug discovery of TPD involves time-​consuming and costly processes, it is important to successfully establish a platform for not only target identification and validation but also rapid and efficient drug discovery. This could be achieved by protein–​protein docking; however, the in silico calculation has not been perfect even for structure-​based drug design (SBDD) for small molecules. Thus, we proposed two processes as follows: (1) target validation by a ligand-​induced genetic degradation system, especially the Auxin-​Inducible Degron (AID) system;9,10 and (2) degrader drug discovery by Rapid Protein Proteolysis Inducer Discovery System (RaPPIDSTM). AID is a new method of POI reduction based on a plant-​specific degradation pathway transplanted to non-​plant cells. In human cells expressing a plant E3 ligase component, TIR1, a degron-​fused endogenous protein, is degraded within 15–​45 min upon addition of phytohormone auxin. Herein, we employed AID to control the protein expression in vivo in a mouse xenograft study. By applying AID for target identification and validation of TPD, we can select highly successful targets based on the actual protein degradation phenotypes. The core technology of RaPPIDS is a semi-​automated synthesis to rapidly prepare divergent degrader molecules, including the types of not only linkers but also E3 ligase binders. Furthermore, the strategy including the effective technologies could drive the drug discovery of TPD drugs against novel target proteins.

9.2 Target Identification and Validation TPD drugs are expected to target undruggable proteins, such as transcription factors which do not have enzymatic activity but are crucial for cancer development. Currently, the proteins potentially important for the survival of cancer cells can be identified by whole-​genome screening, such as siRNA-​or CRISPR-​ Cas9-​based screenings. The candidate proteins can also be identified from

 169

An Efficient Approach Toward Drugging Undruggable Targets

169

the cancer genome databases (for example, the Cancer Genome Atlas: www. cancer.gov/​about-​nci/​organization/​ccg/​research/​structural-​genomics/​tcga). An obvious strategy is to develop TPDs to target these candidate proteins for the treatment of cancer. However, it is critical to identify the validated targets for subsequent processes because developing TPDs are costly and time-​ consuming, and hence, all candidates cannot be subjected to the strategy after the first screening. Moreover, an adequate positive control is essential during the development of the process because the potential TPDs might degrade not only the on-​target but also off-​targets. In order to validate the candidate proteins after the first screening and generate a positive control, we recently proposed the use of ligand-​induced genetic degradation systems.11 Hitherto, several such systems are available: AID, dTAG, Halo-​PROTAC, Halo-​ HyT, and DD. The common feature of these systems is that the POI fused with a defined polypeptide or degron could be efficiently induced for degradation by the addition of a defined ligand. Therefore, the challenges encountered during the designing and optimizing of TPDs can be circumvented by genetically fusing a degron to the candidate proteins. Importantly, the genetic degradation systems are efficient and highly specific to the degron-​fused target, which minimizes the potential off-​targets.

9.2.1 Use of AID Technology In Vivo Plants have unique degradation pathways that are controlled by phytohormones.12 Among them, auxin (known as indole-​3-​acetic acid or IAA) is known to activate the SCF-​TIR1 E3 ubiquitin ligase for the rapid degradation of the transcriptional inhibitory AUX/​IAA-​family proteins to express the downstream genes. This degradation pathway was transplanted into human cells to establish the AID technology so as to control the proteins fused with mini-​AID (mAID), a 7-​kDa degron, derived from the IAA17 protein (see Figure 9.1A).9,10 To generate conditional AID cell lines, we initially established human colorectal cancer HCT116 cells expressing TIR1 derived from rice (OsTIR1). Subsequently, the mAID tag was fused to the endogenous protein using CRISPR-​Cas9-​based genome editing (see Figure 9.1A).10,13 The AID technology has been successfully applied to mammalian cells, zebrafish, fruit fly and nematodes,14–​16 but not yet to mice. Therefore, the present study aimed to control the protein expression in mice using a xenograft model. A common issue is that IAA is metabolized in vivo and converted to indoxyl sulfate, which damages the kidneys.17. Hence, IAA is not an optimal choice for inducing degradation in mice. To avoid this problem, a synthetic auxin, 1-​naphthaleneacetic acid (NAA), was used for this experiment. Herein, the human HCT116 cell line, in which the RAD21 gene was modified to express RAD21-​mAID-​Clover (RAD21-​mAC),10 was used for a mouse xenograft study. RAD21 is a component of cohesin, which is essential for sister-​chromatid cohesion as well as the formation of chromatin loops in the interphase.18 Our previous study showed that the RAD21-​mAC cells-​expressing OsTIR1 attenuated the proliferation after the addition of IAA.10 This cell line was used for

170

170

Chapter 9

Figure 9.1 The AID technology for target validation. (A) CRISPR-​Cas9-​based tagging with mAID and the principle of the AID technology. The gene of interest can be modified to express a degron-​fused protein by introducing an mAID cassette either at the N-​or C-​terminus coding region by CRISPR-​Cas9-​ based knockin. IAA (natural auxin) and NAA (synthetic auxin) activate the SCF-​OsTIR1 E3, which recognizes and polyubiquitylates the mAID-​fused target for proteasome-​dependent degradation. (B) Experimental scheme of xenograft study. (C) Tumor weights formed by RAD21-​mAC (control) and RAD21-​mAID + OsTIR1 cells on day 14. ** denotes p < 0.01 (Dunnett’s test).

 171

An Efficient Approach Toward Drugging Undruggable Targets

171

xenograft study. We inoculated the control RAD21-​mAC cells (without OsTIR1) and the RAD21-​mAC + OsTIR1 cells into BALB/​c-​nu mice (female) according to the protocol shown in Figure 9.1B. Seven days after inoculation, NAA was injected at 0, 10, 30 or 100 mg kg–​1 with one-​day interval, and the tumor weights were recorded on day 14. The control experiments with the RAD21-​ mAC cells without OsTIR1 showed that there was no difference in the tumor weights after treatment under any conditions (see Figure 9.1C, left panel). On the other hand, the RAD21-​mAC + OsTIR1 cells showed a significant reduction in tumor size in a dose-​dependent manner when treated with NAA (see Figure 9.1C, right panel). These results represented that the AID technology can be used in mice, and similar xenograft studies would be useful in target validation for TPD development.

9.3 Efficient Drug Discovery Platform, RaPPIDS The extent of degradation of a target by a degrader might not correlate with its affinity for that target.6 Moreover, the type of proteins that can be degraded is dependent on E3 ligases. Therefore, to encompass the maximum number of target proteins and generate the best molecule, various E3 ligase binders should be identified. In addition, the cooperativity of PPI between the POI and the E3 ligase is critical for the degradation activity of the POI,7 despite the difficulty of accurate prediction of the ternary complex. Thus, the trial-​and-​error process is essential. RaPPIDS is utilized not only for linkers but also for E3 ligase binders including novel XIAP binders to obtain divergent degrader molecules by semi-​ automated synthesis to select the best molecule.

9.3.1 Proprietary E3 Ligase Binders Regarding the E3 ligase binders for TPD, the binders of VHL,19 CRBN,20 MDM221 and IAP22 were identified (see Figure. 9.2A). The degraders based on these binders degraded the POIs, such as BRD4 by MZ-​1,7,23 dBET-​1,24 A187425 and SNIPER(BRD)-​1,26 which are conjugated with a BRD binder, JQ1 (see Figure 9.2BC). Another alternative included novel XIAP binders 1–​327 (see Figure 9.3A). XIAP is one of the IAP family members. These binders have a different pharmacophore from the traditional IAP binders, i.e., Smac peptide and its mimetics, which facilitate different interfaces of the PPI to degrade POIs. In a crystal structure (PDB : 3EYL),28 Smac fragment, AVPI tetrapeptide occupy an area of the BIR3 domain of the XIAP protein, which is stained cyan; also, the Smac mimetics occupy the same area (see Figure 9.3B). Conversely, XIAP binder 1 occupies an area colored magenta (see Figure 9.3C). The superposition of them elucidated the difference of interfaces and linker directions (see Figure 9.3D). Therefore, proprietary XIAP binders can provide an alternative for the degradation of POIs. A homodimer of E3 ligase binders, known as homo-​PROTAC,29,30 degrades the E3 ligase itself as it can induce self-​directed ubiquitination and degradation.

172

Chapter 9

172

A

Me

Me

OH

Me

O N

H2N O

NH

N H

O

O N

O

N

Me N

O

O

N

F

N H

O

S

H N

Me

Me

S

VH032 (VHL binder) O Cl

O

N

Me

B

Me

Me

NH

LCL161 (IAP binder) N Me

H N

OMe N

HO

O Me

CN

N

O

Me Me

Me

Me

F F

O

Me

O

S

HN

O OMe

Me

Cl

Cl

Idasanutlin (MDM2 binder)

Nutlin 3 (MDM2 binder)

N

N

Cl

N

Cl

O

Thalidomide (CRBN binder)

JQ1 (BRD binder)

C Me Me

S

Me

N N

N

O

O

Me N

N H

O 3

Me N H

Me

OH

Me

Me

N O

O

S

N N

N

O

H N

Me N

NH

N H

O

O O O

Cl

S

MZ1

Cl N

Me

dBET1

N

O NH O

Figure 9.2 The compound structures of (A) E3 ligase binders for TPD, (B) a BRD binder, JQ1, and (C) BRD4 degraders.

Figure 9.3 (A) The compound structures of XIAP binders. (B) The crystal structure of XIAP protein with Smac fragment, AVPI tetrapeptide (PDB : 3EYL).28 (C) The crystal structure of XIAP protein with XIAP binder 1. (D) The superposition of XIAP proteins of AVPI tetrapeptide and XIAP binder 1. The linkers are elongated in the direction of the arrows.

 173

An Efficient Approach Toward Drugging Undruggable Targets

173

The homodimer molecules 4 and 5 were designed and prepared from XIAP binder 2 and weakly degraded the XIAP protein itself (see Figure 9.4). As a second trial, XIAP binder 3 was conjugated with dasatinib,31,32 an inhibitor of multiple tyrosine kinases including BCR-​Abl, which would generate a hetero-​bifunctional molecule 6. This showed a weak degradation activity of BCR-​Abl (see Figure 9.5). Finally, XIAP binder 1 was linked to a proprietary multiple kinase binder 727 by four types of PEG linkers to generate the kinase degraders 8–​11 (see Figure 9.6). Their kinase binding affinities and XIAP binding

Figure 9.4 (A) The compound structures of XIAP homo-​PROTAC (XIAP-​based XIAP degrader). (B) THP1 cells were treated with the indicated concentrations of degraders 4 and 5 for 24 h. Whole-​cell lysates were analyzed by western blotting.

174

174

Chapter 9

Figure 9.5 (A) The compound structure of a BCR-​Abl binder, dasatinib and a BCR-​ Abl degrader. (B) K562 cells were treated with the indicated concentrations of degrader 6 for 24 h. Whole-​cell lysates were analyzed by western blotting.

O NH NH N N N

N

N N

O N

N H

N N

MeO

N MeO MeO

7

MeO

N

N N

O O

nN H

N H

O

N

N

HN Me Me

O

N Me

OMe

8: n = 1 9: n = 2 10: n = 3 11: n = 4

Figure 9.6 The structure of multi-​kinase binder 7 and its degraders 8–​11 with four types of PEG linkers.

affinity (IC50 = 39.9 µM) of degrader 9 were evaluated. In addition, GSK3α/​ β and GCN2 proteins were degraded by degraders 8–​11 and the degradation activities of degrader 9 were rescued by co-​treatment with a proteasome inhibitor, epoxomicin (see Figure 9.7). Thus, XIAP-​based degraders demonstrated proteasome-​dependent degradation and the MOA of the TPD pathway. However, the degradation activities of each protein did not correlate with the binding affinities of POIs. In addition, the best linker length differed in each protein. Therefore, an efficient drug discovery platform is required for TPD drug discovery.

 175

An Efficient Approach Toward Drugging Undruggable Targets

175

Figure 9.7 (A) THP1 cells were treated with the indicated concentrations of degraders 8, 9, 10 and 11 for 24 h. Whole-​cell lysates were analyzed by western blotting. (B) Quantification of the remaining protein. The binding affinity (% control at 1 µM) of degrader 9 to each protein evaluated by ScanELECT® (Eurofines DiscoverX) is shown at the bottom of the graph. (C) THP1 cells were treated with 10 µM of degrader 9 in the presence or absence of 1 µM epoxomicin for 8 h.

9.3.2 A Strategy for Drug Candidate Discovery by RaPPIDS Platform In order to overcome the challenge of unpredictable PPI cooperativity, we propose the RaPPIDS platform based on the diversity-​oriented synthesis (DOS) and evaluation strategy. The drug candidates could be selected based on the real data without the prediction of PPI cooperativity. RaPPIDS includes two steps: lead generation and lead optimization. In the lead generation stage, the first step of RaPPIDS, the degradable E3 ligase and the appropriate linker length are selected rapidly (see Figure 9.8A). To achieve this goal, we have stocked a variety of “Ready to Conjugate” (R2C) probes encompassing the binders of several E3 ligase including multiple chemotypes of proprietary XIAP binders with several PEG linker lengths. The R2C probes combine with the target binders to provide 20–​50 degrader molecules. Based on the degradation activity of POI, the degradable E3 ligases and its appropriate linker length are identified as the promising chemical space. At the lead optimization stage, the second step of RaPPIDS, >100 degraders focused on the promising chemical

176

176

Chapter 9

Figure 9.8  (A) RaPPIDS 1st step, lead generation stage. The target binders are conjugated with R2C probes to prepare 20–​ 50 hetero-​ bifunctional molecules, which are evaluated to identify the lead degraders with degradable E3 ligases and appropriate linker lengths. (B) RaPPIDS 2nd step, lead optimization stage. More than 100 divergent degraders are prepared using the fragments of not only the linker but also the E3 binders by semi-​automated synthesis.

space are rapidly prepared by DOS (see Figure 9.8B). In particular, the degraders can be modified not only by the linker types but also the moiety of the E3 ligase binder because these are synthesized by a combination of fragmented linkers and E3 ligase binders utilizing semi-​automated synthesis. After the evaluation of these degraders, another promising chemical space is reidentified. Further DOS could be repeated as necessary, and consequently two or three cycles of this flow could deliver a drug candidate within a year.

9.4 Case Study of RaPPIDS for 1st Program, IRAK-​M Degrader 9.4.1 Background of  IRAK-​M IRAK-​M, also known as IRAK-​3, has a critical role in tightly controlling the innate immune response to preserve immune homeostasis by acting as a negative feedback regulator of the TLR/​IL-​1R signaling pathway. After the stimulation of TLR/​IL-​1R and the formation of Myddosome, IRAK-​M binds to MyD88 and IRAK-​2/​4 to inhibit the activation of IRAK-​1. This, in turn, prevents the dissociation of IRAK1 and IRAK4 from MyD88 and the formation of the IRAK1–​TRAF6 signaling complex, eventually interrupting the downstream activation of NFκB and the subsequent inflammatory signaling pathway. The expression of IRAK-​M is generally confined to the myeloid cells and induced during macrophage maturation.33,34 This myeloid-​specific expression pattern of IRAK-​M enabled a precision medicine strategy by

 177

An Efficient Approach Toward Drugging Undruggable Targets

177

potentially limiting the inflammation-​mediated adverse events against non-​ target tissues. IRAK-M was first described as a negative regulator of the host immune defense. Reportedly, macrophages from IRAK-​M-​deficient mice exhibit elevated levels of inflammatory cytokines upon TLR ligand challenge, and IRAK-M-deficient mice show a hyper-​inflammatory response to bacterial infection.34 Several studies document a similar role of IRAK-​M in determining the phenotype of macrophages. An in-​vivo Lewis lung carcinoma (LLC) syngeneic model demonstrated high IRAK-​M expression in the tumor-​associated macrophages (TAMs).35 Interestingly, TAMs isolated from IRAK-​M-​deficient mice displayed features of the classically activated phenotype. Moreover, evaluating the effect of IRAK-​M in chronic renal injury in vivo revealed that IRAK-​M-​deficient mice were protected from fibrosis and displayed a diminished number of alternatively activated macrophages.36 These pieces of evidence supported the hypothesis that IRAK-​M is responsible for promoting alternative activation of macrophages. In addition, IRAK-​M has an immunoregulatory role in dendritic cells (DCs). The removal of IRAK-​M from DCs results in enhanced migration activity compared to wild-​type DCs.37 Furthermore, the vaccination of syngeneic models with IRAK-​M-​deficient DCs resulted in enhanced tumor clearance. Thus, in terms of therapeutic applications, targeting IRAK-​M may offer some benefit to improve the innate immune function in immunosuppressed cancer patients. Based on the evidence for the role of IRAK-​M in innate immunosuppressive capacity of myeloid cells in the tumor microenvironment, we generated compounds targeting IRAK-​M as an effective cancer immunotherapy strategy by enhancing the inflammatory state of the tumors. Because IRAK-​ M is a pseudokinase, lacking the kinase activity, it is considered “undruggable” and could not be targeted pharmacologically by the conventional small molecule. Therefore, we developed hetero-​bifunctional degrader molecules comprising the IRAK-​M binding moiety linked to proprietary E3 binders to eliminate the IRAK-​M protein.

9.4.2 Drug Discovery of IRAK-​M Degrader by RaPPIDS A high-​throughput screening campaign identified >200 IRAK-​M selective binders; however, none modulated its function, NFκB activation. In the lead generation stage for degrader discovery, two binders were selected based on the selectivity against the other kinases and drug-​like property, especially cellular permeability. The binders combined with R2C probes to provide 19 hetero-​bifunctional molecules. In the case of IRAK-​M, VHL and CRBN binders did not work as E3 binders for IRAK-M degrader in the range of appropriate linker length. On the other hand, proprietary XIAP binders were optimal for IRAK-​M protein degradation (see Figure 9.9). Furthermore, the lead compound 12 was identified through the small modification of the XIAP-​based degrader, which showed IRAK-​M binding affinity (IC50 = 380 nM), XIAP binding affinity (IC50 = 1942 nM) and degradation activity (DC50 = 680 nM).

178

178

Chapter 9

Figure 9.9 (A) The degradation activities of VHL, VRBN, and XIAP-​based degraders conjugated with two types of IRAK-​M binders, which were evaluated by western blotting (THP1 cells were treated with 10 µM of each compound for 24 h). NT, not tested; –​, no degradation at 10 µM; +, degradation at 10 µM; ++, degradation at 3 µM; +++, degradation at 1 µM; ++++, degradation at 4500 residues), which harbors a C-​terminal RING domain followed by a cysteine-​rich region. By working with a truncant of the RING + Cys region, Virdee and coworkers were able to determine the crystal structure and provide the initial characterization for a mechanism that features two catalytic cysteine residues on MYCBP2: first, Cys4520 undergoes transthiolation with E2~UB before intramolecular transthiolation to Cys4572. MYCBP2 Cys4572~UB then reacts with threonine on its substrate NMNAT2 to form a UB oxyester. This is

204

204

Chapter 10

the first description of a RING-Cys-relay (RCR) ligase. Although bioinformatics did not indicate additional human homologues, orthologues of MYCBP2 are found throughout animals as this serves as an inspiring case study for the discovery of mechanistically diverse ligases.

10.6 Conclusions and Future Perspectives A remarkable array of chemical probes and protein variants have provided unprecedented insight into catalytic and structural mechanisms of E3 ligases. Mimicking unstable UB thioesters with UB oxyesters or isopeptide bonds has enabled many structural snapshots of UB transfer complexes that explain details of protein interactions, conformational changes, and enzymology. Serendipitous binding of small-​molecule or UB variant probes to unexpected sites have unearthed a trove of novel and surprising features of E3 ligases. Recently, UB decorated with small molecules has even recently identified a new class of E3 ligase with unprecedented targeting specificity. Even for the well-​characterized classes of E3 ligase, the range of reactions catalyzed, regulation of assembly and activity through intra-​and intermolecular protein interactions, and conformations involved in specifying different activities is extraordinary. It seems that with hundreds of uncharacterized E3 ligases, we can expect a bounty of equally novel mechanisms to be uncovered in the future, and a commensurate explosion of routes for modulating the UB system through chemical probes.

Note added on proof Baek and coworkers recently used a chemical biology approach to visualize UB transfer from a UBE2D E2 to a neddylated CRL substrate.143

References 1. A. J. McClellan, S. H. Laugesen and L. Ellgaard, Open Biol., 2019, 9, 190147. 2. S. Bhogaraju, S. Kalayil, Y. Liu, F. Bonn, T. Colby, I. Matic and I. Dikic, Cell, 2016, 167, 1636–​1649. 3. K.-​C. Pao, N. T. Wood, A. Knebel, K. Rafie, M. Stanley, P. D. Mabbitt, R. Sundaramoorthy, K. Hofmann, D. M. F. van Aalten and S. Virdee, Nature, 2018, 556, 381–​385. 4. W. Kim, E. J. Bennett, E. L. Huttlin, A. Guo, J. Li, A. Possemato, M. E. Sowa, R. Rad, J. Rush, M. J. Comb, J. W. Harper and S. P. Gygi, Mol. Cell, 2011, 44, 325–​340. 5. A. Hershko and A. Ciechanover, Annu. Rev. Biochem., 1998, 67, 425–​479. 6. C. M. Pickart, Annu. Rev. Biochem., 2001, 70, 503–​533. 7. K. N. Swatek and D. Komander, Cell Res., 2016, 26, 399–​422. 8. S. Jentsch and I. Psakhye, Annu. Rev. Genet., 2013, 47, 167–​186. 9. R. I. Enchev, B. A. Schulman and M. Peter, Nat. Rev. Mol. Cell Biol., 2015, 16,  30–​44.

 205

E3-mediated Ubiquitin and Ubiquitin-like Protein Ligation

205

10. J. M. Huibregtse, M. Scheffner, S. Beaudenon and P. M. Howley, Proc. Natl. Acad. Sci. U. S. A., 1995, 92, 5249. 11. R. J. Deshaies and C. A. Joazeiro, Annu. Rev. Biochem., 2009, 78, 399–​434. 12. L. Aravind and E. V. Koonin, Curr. Biol., 2000, 10, R132–​134. 13. K. Wu, J. Kovacev and Z. Q. Pan, Mol. Cell, 2010, 37, 784–​796. 14. D. M. Wenzel, A. Lissounov, P. S. Brzovic and R. E. Klevit, Nature, 2011, 474, 105–​108. 15. M. Rape, S. K. Reddy and M. W. Kirschner, Cell, 2006, 124, 89–​103. 16. Y. Lu, B. H. Lee, R. W. King, D. Finley and M. W. Kirschner, Science, 2015, 348, 1250834. 17. L. Cappadocia and C. D. Lima, Chem. Rev., 2018, 118, 889–​918. 18. L. Buetow and D. T. Huang, Nat. Rev. Mol. Cell. Biol., 2016, 17, 626–​642. 19. N. Zheng and N. Shabek, Annu. Rev. Biochem., 2017, 86, 129–​157. 20. J. N. Pruneda, P. J. Littlefield, S. E. Soss, K. A. Nordquist, W. J. Chazin, P. S. Brzovic and R. E. Klevit, Mol. Cell, 2012, 47, 933–​942. 21. K. K. Dove, J. L. Olszewski, L. Martino, D. M. Duda, X. S. Wu, D. J. Miller, K. H. Reiter, K. Rittinger, B. A. Schulman and R. E. Klevit, Structure, 2017, 25, 890–​900. 22. H. Dou, L. Buetow, G. J. Sibbet, K. Cameron and D. T. Huang, Nat. Struct. Mol. Biol., 2012, 19, 876–​883. 23. A. Plechanovova, E. G. Jaffray, M. H. Tatham, J. H. Naismith and R. T. Hay, Nature, 2012, 489, 115–​120. 24. Derek F. Ceccarelli, X. Tang, B. Pelletier, S. Orlicky, W. Xie, V. Plantevin, D. Neculai, Y.-​C. Chou, A. Ogunjimi, A. Al-​Hakim, X. Varelas, J. Koszela, Gregory A. Wasney, M. Vedadi, S. Dhe-​Paganon, S. Cox, S. Xu, A. Lopez-​ Girona, F. Mercurio, J. Wrana, D. Durocher, S. Meloche, David R. Webb, M. Tyers and F. Sicheri, Cell, 2011, 145, 1075–​1087. 25. H. Huang, D. F. Ceccarelli, S. Orlicky, D. J. St-​Cyr, A. Ziemba, P. Garg, S. Plamondon, M. Auer, S. Sidhu, A. Marinier, G. Kleiger, M. Tyers and F. Sicheri, Nat. Chem. Biol., 2014, 10, 156–​163. 26. K. M. Williams, S. Qie, J. H. Atkison, S. Salazar-​Arango, J. Alan Diehl and S. K. Olsen, Nat. Commun., 2019, 10, 3296. 27. A. Saha, S. Lewis, G. Kleiger, B. Kuhlman and R. J. Deshaies, Mol. Cell, 2011, 42,  75–​83. 28. H. B. Kamadurai, J. Souphron, D. C. Scott, D. M. Duda, D. J. Miller, D. Stringer, R. C. Piper and B. A. Schulman, Mol. Cell, 2009, 36, 1095–​1102. 29. B. C. Lechtenberg, A. Rajput, R. Sanishvili, M. K. Dobaczewska, C. F. Ware, P. D. Mace and S. J. Riedl, Nature, 2016, 529, 546–​550. 30. A. A. Yunus and C. D. Lima, Nat. Struct. Mol. Biol., 2006, 13, 491–​499. 31. K. E. Wickliffe, S. Lorenz, D. E. Wemmer, J. Kuriyan and M. Rape, Cell, 2011, 144, 769–​781. 32. D. C. Scott, V. O. Sviderskiy, J. K. Monda, J. R. Lydeard, S. E. Cho, J. W. Harper and B. A. Schulman, Cell, 2014, 157, 1671–​1684. 33. F. C. Streich, Jr. and C. D. Lima, Nature, 2016, 536, 304–​308. 34. J. R. Lydeard, B. A. Schulman and J. W. Harper, EMBO Rep., 2013, 14, 1050–​1061.

206

206

Chapter 10

35. J. R. Skaar, J. K. Pagan and M. Pagano, Nat. Rev. Mol. Cell. Biol., 2013, 14, 369–​381. 36. F. R. Schumacher, K. Siew, J. Zhang, C. Johnson, N. Wood, S. E. Cleary, R. S. Al Maskari, J. T. Ferryman, I. Hardege, Yasmin, N. L. Figg, R. Enchev, A. Knebel, K. M. O’Shaughnessy and T. Kurz, EMBO Mol. Med., 2015, 7, 1285–​1306. 37. R. G. Hibbert, F. Mattiroli and T. K. Sixma, DNA Repair, 2009, 8, 525–​535. 38. R. Mosadeghi, K. M. Reichermeier, M. Winkler, A. Schreiber, J. M. Reitsma, Y. Zhang, F. Stengel, J. Cao, M. Kim, M. J. Sweredoski, S. Hess, A. Leitner, R. Aebersold, M. Peter, R. J. Deshaies and R. I. Enchev, eLife, 2016, 5. 39. S. Cavadini, E. S. Fischer, R. D. Bunker, A. Potenza, G. M. Lingaraju, K. N. Goldie, W. I. Mohamed, M. Faty, G. Petzold, R. E. Beckwith, R. B. Tichkule, U. Hassiepen, W. Abdulrahman, R. S. Pantelic, S. Matsumoto, K. Sugasawa, H. Stahlberg and N. H. Thoma, Nature, 2016, 531, 598–​603. 40. K. Nakayama, H. Nagahama, Y. A. Minamishima, S. Miyake, N. Ishida, S. Hatakeyama, M. Kitagawa, S. Iemura, T. Natsume and K. I. Nakayama, Dev. Cell, 2004, 6, 661–​672. 41. B. Hao, N. Zheng, B. A. Schulman, G. Wu, J. J. Miller, M. Pagano and N. P. Pavletich, Mol. Cell, 2005, 20,  9–​19. 42. R. J. Davis, M. Welcker and B. E. Clurman, Cancer Cell, 2014, 26, 455–​464. 43. B. Hao, S. Oehlmann, M. E. Sowa, J. W. Harper and N. P. Pavletich, Mol. Cell, 2007, 26, 131–​143. 44. P. Jaakkola, D. R. Mole, Y. M. Tian, M. I. Wilson, J. Gielbert, S. J. Gaskell, A. von Kriegsheim, H. F. Hebestreit, M. Mukherji, C. J. Schofield, P. H. Maxwell, C. W. Pugh and P. J. Ratcliffe, Science, 2001, 292, 468–​472. 45. M. Ivan, K. Kondo, H. Yang, W. Kim, J. Valiando, M. Ohh, A. Salic, J. M. Asara, W. S. Lane and W. G. Kaelin, Jr., Science, 2001, 292, 464–​468. 46. K. Taguchi and M. Yamamoto, Front. Oncol., 2017, 7, 85. 47. G. Li, W. Ci, S. Karmakar, K. Chen, R. Dhar, Z. Fan, Z. Guo, J. Zhang, Y. Ke, L. Wang, M. Zhuang, S. Hu, X. Li, L. Zhou, X. Li, M. F. Calabrese, E. R. Watson, S. M. Prasad, C. Rinker-​Schaeffer, S. E. Eggener, T. Stricker, Y. Tian, B. A. Schulman, J. Liu and K. P. White, Cancer Cell, 2014, 25, 455–​468. 48. M. Zhuang, M. F. Calabrese, J. Liu, M. B. Waddell, A. Nourse, M. Hammel, D. J. Miller, H. Walden, D. M. Duda, S. N. Seyedin, T. Hoggard, J. W. Harper, K. P. White and B. A. Schulman, Mol. Cell, 2009, 36,  39–​50. 49. M. C. Jaramillo and D. D. Zhang, Genes Dev., 2013, 27, 2179–​2191. 50. H. Wada, E. T. Yeh and T. Kamitani, Biochem. Biophys. Res. Commun., 1999, 257, 100–​105. 51. J. K. Monda, D. C. Scott, D. J. Miller, J. Lydeard, D. King, J. W. Harper, E. J. Bennett and B. A. Schulman, Structure, 2013, 21,  42–​53. 52. D. C. Scott, J. K. Monda, E. J. Bennett, J. W. Harper and B. A. Schulman, Science, 2011, 334, 674–​678. 53. D. M. Duda, L. A. Borg, D. C. Scott, H. W. Hunt, M. Hammel and B. A. Schulman, Cell, 2008, 134, 995–​1006.

 207

E3-mediated Ubiquitin and Ubiquitin-like Protein Ligation

207

54. G. A. Cope, G. S. Suh, L. Aravind, S. E. Schwarz, S. L. Zipursky, E. V. Koonin and R. J. Deshaies, Science, 2002, 298, 608–​611. 55. R. I. Enchev, D. C. Scott, P. C. da Fonseca, A. Schreiber, J. K. Monda, B. A. Schulman, M. Peter and E. P. Morris, Cell Rep., 2012, 2, 616–​627. 56. J. E. Brownell, M. D. Sintchak, J. M. Gavin, H. Liao, F. J. Bruzzese, N. J. Bump, T. A. Soucy, M. A. Milhollen, X. Yang, A. L. Burkhardt, J. Ma, H. K. Loke, T. Lingaraj, D. Wu, K. B. Hamman, J. J. Spelman, C. A. Cullis, S. P. Langston, S. Vyskocil, T. B. Sells, W. D. Mallender, I. Visiers, P. Li, C. F. Claiborne, M. Rolfe, J. B. Bolen and L. R. Dick, Mol. Cell, 2010, 37, 102–​111. 57. T. A. Soucy, P. G. Smith, M. A. Milhollen, A. J. Berger, J. M. Gavin, S. Adhikari, J. E. Brownell, K. E. Burke, D. P. Cardin, S. Critchley, C. A. Cullis, A. Doucette, J. J. Garnsey, J. L. Gaulin, R. E. Gershman, A. R. Lublinsky, A. McDonald, H. Mizutani, U. Narayanan, E. J. Olhava, S. Peluso, M. Rezaei, M. D. Sintchak, T. Talreja, M. P. Thomas, T. Traore, S. Vyskocil, G. S. Weatherhead, J. Yu, J. Zhang, L. R. Dick, C. F. Claiborne, M. Rolfe, J. B. Bolen and S. P. Langston, Nature, 2009, 458, 732–​736. 58. R. T. Swords, H. P. Erba, D. J. DeAngelo, D. L. Bixby, J. K. Altman, M. Maris, Z. Hua, S. J. Blakemore, H. Faessel, F. Sedarati, B. J. Dezube, F. J. Giles and B. C. Medeiros, Br. J. Haematol., 2015, 169, 534–​543. 59. D. C. Scott, J. T. Hammill, J. Min, D. Y. Rhee, M. Connelly, V. O. Sviderskiy, D. Bhasin, Y. Chen, S. S. Ong, S. C. Chai, A. N. Goktug, G. Huang, J. K. Monda, J. Low, H. S. Kim, J. A. Paulo, J. R. Cannon, A. A. Shelat, T. Chen, I. R. Kelsall, A. F. Alpi, V. Pagala, X. Wang, J. Peng, B. Singh, J. W. Harper, B. A. Schulman and R. K. Guy, Nat. Chem. Biol., 2017, 13, 850–​857. 60. J. T. Hammill, D. Bhasin, D. C. Scott, J. Min, Y. Chen, Y. Lu, L. Yang, H. S. Kim, M. C. Connelly, C. Hammill, G. Holbrook, C. Jeffries, B. Singh, B. A. Schulman and R. K. Guy, J. Med. Chem., 2018, 61, 2694–​2706. 61. J. T. Hammill, D. C. Scott, J. Min, M. C. Connelly, G. Holbrook, F. Zhu, A. Matheny, L. Yang, B. Singh, B. A. Schulman and R. K. Guy, J. Med. Chem., 2018, 61, 2680–​2693. 62. H. S. Kim, J. T. Hammill, D. C. Scott, Y. Chen, J. Min, J. Rector, B. Singh, B. A. Schulman and R. K. Guy, J. Med. Chem., 2019, 62, 8429–​8442. 63. S. Wang, L. Zhao, X. J. Shi, L. Ding, L. Yang, Z. Z. Wang, D. Shen, K. Tang, X. J. Li, M. Mamun, H. Li, B. Yu, Y. C. Zheng, S. Wang and H. M. Liu, J. Med. Chem., 2019, 62, 2772–​2797. 64. H. Zhou, J. Lu, L. Liu, D. Bernard, C. Y. Yang, E. Fernandez-​Salas, K. Chinnaswamy, S. Layton, J. Stuckey, Q. Yu, W. Zhou, Z. Pan, Y. Sun and S. Wang, Nat. Commun., 2017, 8, 1150. 65. H. Zhou, W. Zhou, B. Zhou, L. Liu, T. R. Chern, K. Chinnaswamy, J. Lu, D. Bernard, C. Y. Yang, S. Li, M. Wang, J. Stuckey, Y. Sun and S. Wang, J. Med. Chem., 2018, 61, 1934–​1950. 66. W. Zhou, L. Ma, L. Ding, Q. Guo, Z. He, J. Yang, H. Qiao, L. Li, J. Yang, S. Yu, L. Zhao, S. Wang, H. M. Liu, Z. Suo and W. Zhao, J. Med. Chem., 2019, 62, 5382–​5403.

208

208

Chapter 10

67. A. Schlierf, E. Altmann, J. Quancard, A. B. Jefferson, R. Assenberg, M. Renatus, M. Jones, U. Hassiepen, M. Schaefer, M. Kiffe, A. Weiss, C. Wiesmann, R. Sedrani, J. Eder and B. Martoglio, Nat. Commun., 2016, 7, 13166. 68. E. J. Bennett, J. Rush, S. P. Gygi and J. W. Harper, Cell, 2010, 143, 951–​965. 69. S. J. Goldenberg, T. C. Cascio, S. D. Shumway, K. C. Garbutt, J. Liu, Y. Xiong and N. Zheng, Cell, 2004, 119, 517–​528. 70. E. S. Fischer, A. Scrima, K. Bohm, S. Matsumoto, G. M. Lingaraju, M. Faty, T. Yasuda, S. Cavadini, M. Wakasugi, F. Hanaoka, S. Iwai, H. Gut, K. Sugasawa and N. H. Thoma, Cell, 2011, 147, 1024–​1039. 71. S. Wu, W. Zhu, T. Nhan, J. I. Toth, M. D. Petroski and D. A. Wolf, Nat. Commun., 2013, 4, 1642. 72. N. W. Pierce, J. E. Lee, X. Liu, M. J. Sweredoski, R. L. Graham, E. A. Larimore, M. Rome, N. Zheng, B. E. Clurman, S. Hess, S. O. Shan and R. J. Deshaies, Cell, 2013, 153, 206–​215. 73. S. V. Faull, F. Beuron, H. Yebenes, A. Lau, A. Politis and E. P. Morris, Biophys. J., 2017, 112, 578a. 74. J. M. Reitsma, X. Liu, K. M. Reichermeier, A. Moradian, M. J. Sweredoski, S. Hess and R. J. Deshaies, Cell, 2017, 171, 1326–​1339. 75. X. Liu, J. M. Reitsma, J. L. Mamrosh, Y. Zhang, R. Straube and R. J. Deshaies, Mol. Cell, 2018, 69, 773–​786.e776. 76. Y. W. Zhang, J. Brognard, C. Coughlin, Z. You, M. Dolled-​ Filhart, A. Aslanian, G. Manning, R. T. Abraham and T. Hunter, Mol. Cell, 2009, 35, 442–​453. 77. D. E. Spratt, H. Walden and G. S. Shaw, Biochem. J., 2014, 458, 421–​437. 78. T. Kitada, S. Asakawa, N. Hattori, H. Matsumine, Y. Yamamura, S. Minoshima, M. Yokochi, Y. Mizuno and N. Shimizu, Nature, 1998, 392, 605–​608. 79. E. M. Valente, P. M. Abou-​Sleiman, V. Caputo, M. M. Muqit, K. Harvey, S. Gispert, Z. Ali, D. Del Turco, A. R. Bentivoglio, D. G. Healy, A. Albanese, R. Nussbaum, R. Gonzalez-​Maldonado, T. Deller, S. Salvi, P. Cortelli, W. P. Gilks, D. S. Latchman, R. J. Harvey, B. Dallapiccola, G. Auburger and N. W. Wood, Science, 2004, 304, 1158–​1160. 80. B. Stieglitz, R. R. Rana, M. G. Koliopoulos, A. C. Morris-​Davies, V. Schaeffer, E. Christodoulou, S. Howell, N. R. Brown, I. Dikic and K. Rittinger, Nature, 2013, 503, 422–​426. 81. H. Johansson, Y. C. I. Tsai, K. Fantom, C. W. Chung, S. Kumper, L. Martino, D. A. Thomas, H. C. Eberl, M. Muelbaier, D. House and K. Rittinger, J. Am. Chem. Soc., 2019, 141, 2703–​2712. 82. H. Walden and K. Rittinger, Nat. Struct. Mol. Biol., 2018, 25, 440–​445. 83. T. Wauer and D. Komander, EMBO J., 2013, 32, 2099–​2112. 84. J. F. Trempe, V. Sauve, K. Grenier, M. Seirafi, M. Y. Tang, M. Menade, S. Al-​Abdul-​Wahid, J. Krett, K. Wong, G. Kozlov, B. Nagar, E. A. Fon and K. Gehring, Science, 2013, 340, 1451–​1455.

 209

E3-mediated Ubiquitin and Ubiquitin-like Protein Ligation

209

85. T. Wauer, M. Simicek, A. Schubert and D. Komander, Nature, 2015, 524, 370–​374. 86. A. Kumar, V. K. Chaugule, T. E. C. Condos, K. R. Barber, C. Johnson, R. Toth, R. Sundaramoorthy, A. Knebel, G. S. Shaw and H. Walden, Nat. Struct. Mol. Biol., 2017, 24, 475–​483. 87. C. Gladkova, S. L. Maslen, J. M. Skehel and D. Komander, Nature, 2018, 559, 410–​414. 88. V. Sauve, G. Sung, N. Soya, G. Kozlov, N. Blaimschein, L. S. Miotto, J. F. Trempe, G. L. Lukacs and K. Gehring, Nat. Struct. Mol. Biol., 2018, 25, 623–​630. 89. A. Ordureau, J. A. Paulo, W. Zhang, T. Ahfeldt, J. Zhang, E. F. Cohn, Z. Hou, J. M. Heo, L. L. Rubin, S. S. Sidhu, S. P. Gygi and J. W. Harper, Mol. Cell, 2018, 70, 211–​227. 90. I. R. Kelsall, D. M. Duda, J. L. Olszewski, K. Hofmann, A. Knebel, F. Langevin, N. Wood, M. Wightman, B. A. Schulman and A. F. Alpi, EMBO J., 2013, 32, 2848–​2860. 91. D. C. Scott, D. Y. Rhee, D. M. Duda, I. R. Kelsall, J. L. Olszewski, J. A. Paulo, A. de Jong, H. Ovaa, A. F. Alpi, J. W. Harper and B. A. Schulman, Cell, 2016, 166, 1198–​1214. 92. R. Huttenhain, J. Xu, L. A. Burton, D. E. Gordon, J. F. Hultquist, J. R. Johnson, L. Satkamp, J. Hiatt, D. Y. Rhee, K. Baek, D. C. Crosby, A. D. Frankel, A. Marson, J. W. Harper, A. F. Alpi, B. A. Schulman, J. D. Gross and N. J. Krogan, Cell Host Microbe, 2019, 26, 86–​99.e87. 93. M. Scheffner, J. M. Huibregtse, R. D. Vierstra and P. M. Howley, Cell, 1993, 75, 495–​505. 94. J. Weber, S. Polo and E. Maspero, Front. Physiol., 2019, 10, 370. 95. B. Strack, A. Calistri, M. A. Accola, G. Palu and H. G. Gottlinger, Proc. Na tl. Acad. Sci. U. S. A., 2000, 97, 13063–​13068. 96. A. Upadhyay, V. Joshi, A. Amanullah, R. Mishra, N. Arora, A. Prasad and A. Mishra, Front. Mol. Neurosci., 2017, 10, 151. 97. O. Staub, S. Dho, P. Henry, J. Correa, T. Ishikawa, J. McGlade and D. Rotin, EMBO J., 1996, 15, 2371–​2380. 98. G. Wang, J. Yang and J. M. Huibregtse, Mol. Cell Biol., 1999, 19, 342–​352. 99. S. Sanchez-​ Tena, M. Cubillos-​ Rojas, T. Schneider and J. L. Rosa, Cell. Mol. Life Sci., 2016, 73, 1955–​1968. 100. L. Huang, E. Kinnucan, G. Wang, S. Beaudenon, P. M. Howley, J. M. Huibregtse and N. P. Pavletich, Science, 1999, 286, 1321–​1326. 101. M. A. Verdecia, C. A. P. Joazeiro, N. J. Wells, J. L. Ferrer, M. E. Bowman, T. Hunter and J. P. Noel, Mol. Cell, 2003, 11, 249–​259. 102. E. Maspero, E. Valentini, S. Mari, V. Cecatiello, P. Soffientini, S. Pasqualato and S. Polo, Nat. Struct. Mol. Biol., 2013, 20, 696–​701. 103. H. B. Kamadurai, Y. Qiu, A. Deng, J. S. Harrison, C. Macdonald, M. Actis, P. Rodrigues, D. J. Miller, J. Souphron, S. M. Lewis, I. Kurinov, N. Fujii, M. Hammel, R. Piper, B. Kuhlman and B. A. Schulman, eLife, 2013, 2, e00828.

210

210

Chapter 10

104. M. E. French, J. L. Klosowiak, A. Aslanian, S. I. Reed, J. R. Yates, 3rd and T. Hunter, J. Biol. Chem., 2017, 292, 10398–​10413. 105. Z. M. Eletr, D. T. Huang, D. M. Duda, B. A. Schulman and B. Kuhlman, Nat. Struct. Mol. Biol., 2005, 12, 933–​934. 106. Z. M. Eletr and B. Kuhlman, J. Mol. Biol., 2007, 369, 419–​428. 107. C. Purbeck, Z. M. Eletr and B. Kuhlman, Biochemistry, 2010, 49, 1361–​1363. 108. M. Wang and C. M. Pickart, EMBO J., 2005, 24, 4324–​4333. 109. V. P. Ronchi, J. M. Klein and A. L. Haas, J. Biol. Chem., 2013, 288, 10349–​10360. 110. H. T. Kim, K. P. Kim, F. Lledias, A. F. Kisselev, K. M. Scaglione, D. Skowyra, S. P. Gygi and A. L. Goldberg, J. Biol. Chem., 2007, 282, 17375–​17386. 111. M. A. Michel, P. R. Elliott, K. N. Swatek, M. Simicek, J. N. Pruneda, J. L. Wagstaff, S. M . Freund and D. Komander, Mol. Cell, 2015, 58, 95–​109. 112. M. K. Hospenthal, S. M. Freund and D. Komander, Nat. Struct. Mol. Biol., 2013, 20, 555–​565. 113. H. C. Kim and J. M. Huibregtse, Mol. Cell Biol., 2009, 29, 3307–​3318. 114. L. K. Ries, B. Sander, K. K. Deol, M. A. Letzelter, E. R. Strieter and S. Lorenz, J. Biol. Chem., 2019, 294, 6113–​6129. 115. M. E. French, B. R. Kretzmann and L. Hicke, J. Biol. Chem., 2009, 284, 12071–​12079. 116. H. C. Kim, A. M. Steffen, M. L. Oldham, J. Chen and J. M. Huibregtse, EMBO Rep., 2011, 12, 334–​341. 117. E. Maspero, S. Mari, E. Valentini, A. Musacchio, A. Fish, S. Pasqualato and S. Polo, EMBO Rep., 2011, 12, 342–​349. 118. A. A. Ogunjimi, S. Wiesner, D. J. Briant, X. Varelas, F. Sicheri, J. Forman-​ Kay and J. L. Wrana, J. Biol. Chem., 2010, 285, 6308–​6315. 119. S. Wiesner, A. A. Ogunjimi, H. R. Wang, D. Rotin, F. Sicheri, J. L. Wrana and J. D. Forman-​Kay, Cell, 2007, 130, 651–​662. 120. A. Herrador, S. Leon, R. Haguenauer-​Tsapis and O. Vincent, J. Biol. Chem., 2013, 288, 16206–​16211. 121. W. Zhang, K. P. Wu, M. A. Sartori, H. B. Kamadurai, A. Ordureau, C. Jiang, P. Y. Mercredi, R. Murchie, J. Hu, A. Persaud, M. Mukherjee, N. Li, A. Doye, J. R. Walker, Y. Sheng, Z. Hao, Y. Li, K. R. Brown, E. Lemichez, J. Chen, Y. Tong, J. W. Harper, J. Moffat, D. Rotin, B. A. Schulman and S. S. Sidhu, Mol. Cell, 2016, 62, 121–​136. 122. V. P. Ronchi, J. M. Klein, D. J. Edwards and A. L. Haas, J. Biol. Chem., 2014, 289, 1033–​1048. 123. D. R. Todaro, A. C. Augustus-​Wallace, J. M. Klein and A. L. Haas, J. Biol. Chem., 2018, 293, 18192–​18206. 124. I. Attali, W. S. Tobelaim, A. Persaud, K. Motamedchaboki, K. J. Simpson-​ Lavy, B. Mashahreh, O. Levin-​ Kravets, T. Keren-​ Kaplan, I. Pilzer, M. Kupiec, R. Wiener, D. A. Wolf, D. Rotin and G. Prag, EMBO J., 2017, 36, 425–​440.

 211

E3-mediated Ubiquitin and Ubiquitin-like Protein Ligation

211

125. C. Riling, H. Kamadurai, S. Kumar, C. E. O’Leary, K. P. Wu, E. E. Manion, M. Ying, B. A. Schulman and P. M. Oliver, J. Biol. Chem., 2015, 290, 23875–​23887. 126. S. Mari, N. Ruetalo, E. Maspero, M. C. Stoffregen, S. Pasqualato, S. Polo and S. Wiesner, Structure, 2014, 22, 1639–​1649. 127. Z. Chen, H. Jiang, W. Xu, X. Li, D. R. Dempsey, X. Zhang, P. Devreotes, C. Wolberger, L. M. Amzel, S. B. Gabelli and P. A. Cole, Mol. Cell, 2017, 66, 345–​357. 128. Z. Wang, Z. Liu, X. Chen, J. Li, W. Yao, S. Huang, A. Gu, Q.-​Y. Lei, Y. Mao and W. Wen, Nat. Commun., 2019, 10, 3162–​3162. 129. K. Zhu, Z. Shan, X. Chen, Y. Cai, L. Cui, W. Yao, Z. Wang, P. Shi, C. Tian, J. Lou, Y. Xie and W. Wen, EMBO Rep., 2017, 18, 1618–​1630. 130. S. A. Forbes, D. Beare, H. Boutselakis, S. Bamford, N. Bindal, J. Tate, C. G. Cole, S. Ward, E. Dawson, L. Ponting, R. Stefancsik, B. Harsha, C. Y. Kok, M. Jia, H. Jubb, Z. Sondka, S. Thompson, T. De and P. J. Campbell, Nucleic Acids Res., 2017, 45, D777–​D783. 131. D. Chen, M. Gehringer and S. Lorenz, ChemBiochem, 2018, 19, 2123–​2135. 132. S. G. Kathman, I. Span, A. T. Smith, Z. Xu, J. Zhan, A. C. Rosenzweig and A. V. Statsyuk, J. Am. Chem. Soc., 2015, 137, 12442–​12445. 133. S. G. Kathman, Z. Xu and A. V. Statsyuk, J. Med. Chem., 2014, 57, 4969–​4974. 134. S. Park, D. T. Krist and A. V. Statsyuk, Chem. Sci., 2015, 6, 1770–​1779. 135. D. T. Krist, S. Park, G. H. Boneh, S. E. Rice and A. V. Statsyuk, Chem. Sci., 2016, 7, 5587–​5595. 136. D. T. Krist, P. K. Foote and A. V. Statsyuk, Curr. Protoc. Chem. Biol., 2017, 9,  11–​37. 137. S. Park, P. K. Foote, D. T. Krist, S. E. Rice and A. V. Statsyuk, J. Biol. Chem., 2017, 292, 16539–​16553. 138. P. K. Foote, D. T. Krist and A. V. Statsyuk, Curr. Protoc. Chem. Biol., 2017, 9, 174–​195. 139. M. P. Mulder, K. Witting, I. Berlin, J. N. Pruneda, K. P. Wu, J. G. Chang, R. Merkx, J. Bialas, M. Groettrup, A. C. Vertegaal, B. A. Schulman, D. Komander, J. Neefjes, F. El Oualid and H. Ovaa, Nat. Chem. Biol., 2016, 12, 523–​530. 140. K.-​C. Pao, M. Stanley, C. Han, Y.-​C. Lai, P. Murphy, K. Balk, N. T. Wood, O. Corti, J.-​C. Corvol, M. M. K. Muqit and S. Virdee, Nat. Chem. Biol., 2016, 12, 324–​331. 141. T. Kurz, Y. C. Chou, A. R. Willems, N. Meyer-​Schaller, M. L. Hecht, M. Tyers, M. Peter and F. Sicheri, Mol. Cell, 2008, 29,  23–​35. 142. T. Kurz, N. Ozlu, F. Rudolf, S. M. O’Rourke, B. Luke, K. Hofmann, A. A. Hyman, B. Bowerman and M. Peter, Nature, 2005, 435, 1257–​1261. 143. K. Baek, D. T. Krist, R. J. Prabu, S. Hill, M. Klügel, L. -M. Neumaier, S. von Gronau, G. Kleiger and B. A. Schulman, Nature, 2020, 578, 461–466.

212

CHAPTER 11

Plant E3 Ligases as Versatile Tools for Novel Drug Development and Plant Bioengineering RAED AL-​SAHARIN, SUTTON MOONEY AND HANJO HELLMANN* Washington State University, Abelson 435A, Pullman, 99164, WA, USA *Corresponding author. Email: [email protected]

11.1 Introduction The twenty-​first century provides a variety of challenges in securing sustainable agriculture that will be able to provide enough food for the growing human population. It is estimated that by 2050 there will be a global population of nearly 10 billion people (https://​population.un.org/​wpp/​), which will require a significant increase in both food production and usage of arable land over the next 30 years. Unfortunately this is being countered by an estimated yearly soil loss of 359 mega tons worldwide through erosion (https://​esdac.jrc.ec.europa. eu/​themes/​global-​soil-​erosion),1 and even more problematic are the current and predicted global climate changes that will continue to impact growth and yield of crop plants worldwide.2

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

212

 213

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

213

As sessile organisms plants of course cannot avoid stress by independently moving to a different location, and therefore require some physiological flexibility, along with rapid response pathways, to cope with environmental changes. One rapid response mechanism that plants rely on is the ubiquitin proteasome pathway (UPP). The pathway marks specific target proteins with a small polypeptide chain of ubiquitin (UBQ) and facilitates their proteolytic degradation by the 26S proteasome.3 First discovered in yeast and then identified in other eukaryotic organisms, research on this system has uncovered its involvement across extremely diverse cellular activities, and it therefore represents a perfect tool for plants to quickly react to changing conditions and environmental threats. The UPP starts with the ATP-​dependent activation of UBQ through a UBQ-​ activating E1 enzyme (see Figure 11.1A). The UBQ is then transferred to a UBQ-​conjugating E2 protein that physically assembles with an UBQ E3 ligase. The E3 ligase provides close proximity between UBQ and a substrate to facilitate the covalent attachment of UBQ moieties onto the substrate protein (Figure 11.1A). This normally starts by using an ε-​amino group of a lysine residue from the substrate and a C-​terminal glycine residue from the UBQ. Upon building up a poly-​UBQ chain, the fate of the substrate is likely proteolytic degradation (Figure 11.1A). UBQ contains seven lysine residues that can be used to generate chains, and the E2 enzyme determines which of these will be used for the poly-​ubiquitylation.4 This in turn affects whether the protein is degraded or undergoes another process, such as translocation within the cell.4 This review article focuses on proteolytic degradation, which is classically connected with Lys48-​linked poly-​ubiquitylated proteins.5–​8 While higher plants have only one or two E1s, they normally encode for a larger number of E2s (e.g., 37 in Arabidopsis,9 56 in tomato,10 75 in corn11) that may specifically interact with certain E3s and direct the formation of specific poly-​UBQ chains.4 E3 ligases are by far the most diverse group and are encoded in high numbers in plant genomes, frequently exceeding 1 000 different members.3 Consequently, E3 ligase activities provide striking diversity and flexibility to the pathway.3 This review first gives a brief overview of the different classes of E3s and highlights their diversity in plants. It will then describe a few examples of how pathogens employ E3 ligases to by-​pass plant defense responses, to illustrate how powerful it can be to hijack the pathway. Lastly, discussion will focus on utilization of the UPP to modify certain traits for agricultural purposes and engineer novel biotechnological research tools in plants.

11.2 The Four Classes of E3 Ligases in Higher Plants: A Brief Overview 11.2.1 Monomeric E3 RING-​finger Ligases Monomeric Really Interesting New Gene (RING)-​finger E3 ligases represent a diverse class of E3 ligases in plants. They are characterized by their RING domain, which consists of one histidine and seven cysteine residues (C3HC4;

214

214

Chapter 11

Figure 11.1 The ubiquitin proteasome pathway and classes of E3 ligases in plants. (A) The ubiquitin proteasome pathway depends on the concerted activities of UBQ-​activating E1 proteins, UBQ-​conjugating E2 proteins, and UBQ E3 ligases. E3 ligases bind the E2/​UBQ and facilitate transfer of UBQ moieties onto substrate proteins. Once a substrate is poly-​ ubiquitylated it becomes a target for proteolytic degradation by the 26S proteasome. (B) Model of monomeric RING-​finger and U-​box E3 ligases. (C–​F) The four different subclasses of multimeric E3 ligases containing a cullin protein as scaffolding subunit. SCF E3s use F-​box proteins as substrate adaptors (C), while CRL3 E3 ligases utilize proteins with a BTB fold (D), and CRL4 E3s use proteins with a modified WD-​40 motif (E). APC/​ C complexes (F) have been shown to use Cell Division Cycle20 (CDC20) and CDC20 Homolog1 (CDH1) for substrate recognition. (G) HECT E3 ligases have an intermediate ubiquitylation step, in which a conserved cysteine residue within the HECT domain forms a thiolester with UBQ before attaching it to the substrate.

C, cysteine; H, histidine), that are arranged in a cross-​brace manner to chelate two zinc ions.12–​14 The RING domain facilitates binding of the activated UBQ-​E2 and brings the UBQ into close proximity of the bound substrate protein15,16 (Figure 11.1B). Although able to function independently as E3 ligases,

 215

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

215

they have been reported to self-​assemble to homo-​and hetero-​dimeric complexes.17–​20 There are 470 different monomeric RING-​finger E3s encoded in Arabidopsis (Arabidopsis thaliana), and while comparably few have been functionally described so far, it is clear that they broadly affect plant development and stress tolerance.21

11.2.2 Cullin-​based RING E3 Ligases Cullin-​based E3 ligases are by far the largest class with likely more than 1 000 different members in most, if not all, higher plant species. These multimeric E3s use the cullin as a scaffolding backbone to bind substrate receptors at its N-​terminus, and a RING-​finger protein at its C-​terminal region.22–​25 Similarly to the monomeric E3s, the RING-​finger is needed for binding of the E2-​UBQ.26 Four different subclasses of Cullin-​based E3s have been described in plants: SCF (Skp1, Cullin1 (CUL1), F-​box), CRL3 (Cul3 RING E3 Ligase), CRL4 (Cul4 RING E3 Ligase). and the APC/​C (Anaphase Promoting Complex/​ Cyclosome) complex (Figure  11.1C–​F).23 The substrate receptors of SCF E3s are F-​ box-​ containing proteins (Figure 11.1C).27 The F-​box motif comprises around 50 amino acids and facilitates binding to a SKP1 protein in order to come into complex with the CUL1. The second protein-​binding domain is variable and allows for substrate recognitions.28 Depending on the species, plants can express several hundred to a thousand different F-​box proteins.27,29–​32 SCF complexes were first described in context with the phytohormone auxin, and have now been linked to all the phytohormone pathways, which affect development and physiology of plants. Research has also demonstrated involvement in pathogen defense, cell cycle control and photoperiod-​related responses.33–​39 CRL3 E3 ligases require a protein with a Broad-​Complex, Tramtrack and Bric-​a-​Brac (BTB) motif as a substrate receptor (Figure 11.1D).22 The BTB fold facilitates assembly with CUL3, and allows homo-​and heteromeric assemblies with other BTB proteins.22,40–​43 BTB-​containing proteins also have a secondary protein–​protein interaction domain that is needed to recognize substrates.44 They are present in smaller numbers compared to F-​box proteins, but still comprise between 80 in Arabidopsis and nearly 150 proteins in rice (Oryza sativa).44,45 For most of these BTB-​containing proteins, a role as substrate receptor in a CRL3 E3 ligase has not been confirmed. However, the published examples demonstrate that this specific E3 ligase is widely involved in abiotic stress control, salicylic acid and abscisic acid signaling, ethylene biosynthesis, red light responses and developmental processes, often by directly controlling stability of transcription factors.46–​56 In CRL4 complexes the substrate adaptors contain a modified WD40 motif that is required for binding to a protein called Damaged DNA Binding 1 (DDB1), which in turn binds to CUL4 (Figure 11.1E).57 A typical DDB1 interacting protein contains seven WD40 domains. of which at least one is slightly modified, ending in an aspartate–​arginine motif (WDxR).58 These WDxR-​ containing proteins are referred to as DDB1 CUL4 Associated Factors (DCAF),

216

216

Chapter 11 58

and are as abundant in plants as BTB-​containing proteins. For example, 119 and 151 members have been identified in Arabidopsis and rice, respectively.58 CRL4 E3 ligases were originally brought into context with damaged DNA recognition and photomorphogenesis,57,59–​62 but have meanwhile been demonstrated to affect epigenetic processes, abiotic stress responses and diverse developmental aspects.63–​68 APC/​C E3 ligases are mainly involved in controlling the transition from metaphase to anaphase in mitotically active cells; they affect gametogenesis and embryogenesis, and have also been reported to be involved in auxin-​and ethylene-​dependent processes.69–​72 This E3 ligase is the most complex of the cullin-​based E3 ligase, comprising at least 11 core subunits. The subunit that represents the cullin is called APC2, while the RING-​finger protein, which binds the E2-​UBQ, is APC11.23 Other subunits, such as APC1, 4, 5, 6 and 8, serve as scaffolding,73,74 and two small protein families, called Cell Division Cycle20 (CDC20) and CDC20 Homolog1 (CDH1), have been identified as substrate adaptors (Figure 11.1F).75–​78

11.2.3 U-​box E3 Ligases Plant U-​box (PUB) E3 ligases contain a motif that is highly related to the RING-​ finger (Figure 11.1B). It comprises around 70 amino acids that fold into a RING-​finger-​like structure that similarly binds an E2 carrying an activated UBQ, but it lacks the zinc-​chelating histidine and cysteine residues present in “true” RING fingers.79,80 The family of PUBs is less diverse than RING-​finger E3s, but one can still find 61 and 77 proteins in Arabidopsis and rice, respectively,81 and 125 in soybean (Glycine max).82 In plants, around 40% of PUBs have ARMADILLO (ARM) repeats in the second domain for substrate recognition.81 U-​box E3 ligases have been reported to be involved in a broad range of processes such as abscisic acid signaling, plant immunity, self-​incompatibility, meristem maintenance and abiotic stress tolerance.83–​89

11.2.4 HECT E3 Ligases Homologous to the E6-​ AP Carboxyl Terminus (HECT) domain containing E3s are the least-​diverse group of E3 ligases in plants.90 For example, barley (Hordeum vulgare) encodes only five members, while Arabidopsis has seven, turnip (Brassica rapa) 10, and in soybean 14 members are annotated.90–​92 The HECT domain comprises 350 amino acids and is located in the C-​terminal region of the protein. In the N-​terminal part an additional domain, such as armadillo repeats or IQ domains, can be found, and these are specific for an individual HECT E3 ligase.90 The HECT domain is involved in binding the UBQ-​E2, but it also facilitates a step unique to this class of E3s: a conserved cysteine residue within the domain forms a thiolester with the UBQ, which serves as an intermediate step before the UBQ is attached to the substrate protein (Figure 11.1G). HECT E3s were first described in context with trichome development and leaf senescence,91,93 but recent work has also demonstrated a role in plant immunity.94

 217

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

217

11.3 Pathogens’ Use of the Ubiquitin Proteasome Pathway Hosts and their various pathogens are permanently developing ways to overcome the others’ attack and defense systems. Consequently, sophisticated machinery has evolved on both sides to allow either efficient infection by the pathogen or successful defense and immunity by the plant. The following three selected examples show how prokaryotic pathogens, which do not have the UPP, have developed effective strategies by specifically using E3 ligases to prevent or overcome host defense responses. These “natural bioengineers” demonstrate the power of how utilizing the UPP can successfully subvert responses and alter traits in plants. For a more comprehensive overview, the reader is referred to some excellent reviews on this topic.95–​97 Probably one of the earliest reported examples of a prokaryotic pathogen hijacking the UPP in plants in order to serve its purpose came from Agrobacterium tumefaciens. Now classical tools in plant biotechnology, these bacteria naturally transfer parts of their genetic material via a type IV secretion system into plant cells where it recombines with the genomic DNA in the nucleus and induces callus growth.98 The transferred DNA (T-​DNA) is coated and protected by VIP1 and VIPE2 bacterial proteins, which need to be removed to allow integration of the T-​DNA into the plant’s genomic DNA (Figure 11.2A).98 Together with the T-​DNA, the agrobacterium also transfers VirF, an F-​box protein, into the plant cell. VirF binds to VIP1 and VIPE2, integrates into an SCF E3 ligase complex from the plant and promotes degradation of the T-​DNA’s coating proteins in the nucleus.99 The naked DNA is then usable for integration into the nuclear DNA (Figure 11.2A). The bacterium Pseudomonas syringae pv. tomato DC3000 causes bacterial speck disease in tomato and Arabidopsis.100 It encodes for several effector proteins that are brought into plant cells by a type III secretion system, allowing the Pseudomonas bacteria to counteract the plant’s defense responses.100 One of these proteins, AvrPtoB, contains several functional domains that are in part used to bind to membrane-​standing and cytosolic receptor kinases.101–​ 103 At the C-​terminal region, AvrPtoB folds into a PUB E3 ligase that, in tomato (Solanum esculentum), can target the threonine/​serine kinase Fen and the flagellin receptor kinase FLS2.101,104 FLS2 is needed to cause pathogen-​ associate molecular patterns (PAMPs)-​triggered immunity (PTI) by binding with its extracellular receptor unit flagellin, while Fen is needed intracellularly for effector-​triggered immunity (ETI) by binding to pathogenic and injected Type III-​secreted effector proteins.105 AvrPtoB thus counteracts both PTI and ETI by causing FLS and Fen ubiquitylation and subsequent proteolytic degradation by the 26S proteasome (Figure 11.2B). The third example comes from Xanthomonas campestris pv. vesicatoria. This bacterium causes leaf spot disease in several plant species.106 X. campestris also introduces effector proteins into plant cells via a type III secretion system. Most interesting is XopL,107 which is discussed to be a novel E3 ligase that does not resemble any of the known plant E3s.107 However, it can bind to specific E2s

218

218

Chapter 11

Figure 11.2 Schematic models of pathogens hijacking the UPP to by-​pass defense mechanisms in plants. (A) A. tumefaciens bacteria transfer T-​DNA via a type IV secretion system into plant cells. The T-​DNA is protected and coated by Vip1 and VipE2 proteins. The proteins need to be removed in the nucleus to allow integration of the T-​DNA into the plant’s genomic DNA. This is accomplished by a third A. tumefaciens protein called VirF, an F-​box protein that binds to Vip1 and VipE2, and forms an SCF E3 ligase complex together with CUL1 and RBX1 to ultimately promote degradation of the two DNA-​coating proteins. (B) The bacterium P. syringae transfers AvrPtoB proteins into the plant cell via a type III secretion system. AvrPtoB has E3 ligase activity and recognizes both the cytosolic Fen kinase and the plasma membrane standing FLS receptor kinase. AvroPtoB causes their ubiquitylation and subsequent degradation via the 26S proteasome. Fen kinase and FLS receptor kinase play critical roles in immune defense responses and their degradation increases the plant’s susceptibility for P. syringae infection. (C) A third bacterium, X. campestris, transfers the protein XopL into plant cells, also via a type III secretory system. XopL represents a novel class of E3 ligase that is different from all known plant E3 ligases described so far. XopL substrates are still unknown, but the E3 ligase negatively affects the oxidative burst, pathogen-​related marker gene expression and callose deposition.

that generate Lys11-​linked polyubiquitylation chains, and it has demonstrated auto-​ubiquitylation E3 ligase activity.107 Although host plant target proteins are not known to date, recent work in Arabidopsis showed that it suppresses PTI in the plant. Interestingly, PTI often is activated via a MAPK cascade, but XopL appears to interfere with MAPK-​independent PTI signaling.108 The impacts of XopL activity are quite broad, ranging from affecting the oxidative burst, expression of the FLG22-​induced Receptor-​like Kinase 1 (FRK1) and pathogen-​ related marker gene expression, as well as callose deposition (Figure 11.2C).108

 219

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

219

11.4 The Ubiquitin Proteasome Pathway as an Opportunity As described in the previous section, pathogens will modify, utilize and even develop novel E3 ligases to effectively overcome host defense mechanisms. Because of the diversity of E3 ligases and the speed of the ubiquitylation and degradation reactions, the UPP provides great opportunities to bioengineer plants with new traits, to develop novel drugs and design new research tools. This last section will explore these different directions and suggest how some of the current developments in plant research may benefit society by addressing both agricultural challenges as well as human health problems.

11.4.1 Novel Drug Development Utilizing the UPP The UPP has been long known as a target for novel drug developments in human. For example, targets against E1 or E1-​like proteins such as PYR-​41109 or MLN4924,110 respectively, are available, and the latter appears to be a promising candidate for clinical treatment of renal cancer.111 Others target activities of specific E2s or E3s such as NSC697923112 or nutlins,113 respectively. NSC697923 inhibits the E2 UBE2N/​UBC13, while nutlins are cis-​imidazoline analogues that block activity of the monomeric RING-​finger E3 ligase MDM2. Both UBE2n/​UBC3 and MDM2 affect activity of p53, a central and prominent cell cycle regulator that is often a target for anticancer drug development.112–​115 In comparison, hardly any drugs are available for agricultural purposes that would affect the UPP in plants at any stage. Based on the conserved nature of E1 enzymes across eukaryotes, some, such as PYR-​41 and MLN4924, may actually be functional in plants. However, others that are designed against UPP targets, such as NSC697923 and nutlins, may have no impact on the plant’s physiology, if these are animal-​specific. Application of drugs to crops also has several considerations. In addition to cost and efficiency, there may be a risk that certain drugs could also affect other organisms, including farmers themselves. Nevertheless, the examples available from research on humans show the potency of the UPP as a drug target, and one may expect that this can also be a fruitful route for novel herbicide development in plants. One very good example of E3 ligase activity being efficiently and specifically affected comes from the herbicide 2,4-​dichlorophenoxyacetic acid (2,4-​ D). 2,4-​D was discovered in the mid-​1940s, and is a synthetic auxin. Auxins are phytohormones that broadly affect plant development116 through the activity of the SCFTIR1 complex. Transport Inhibitor Response1 (TIR1) is the F-​box substrate receptor in this E3 ligase,117 and it was shown some years ago to also be the receptor for auxin.118 Binding of auxin by TIR1 is necessary to promote degradation of AUX/​IAA proteins, which are negative auxin response regulators. Consequently, application of 2,4-​ D over-​ stimulates the auxin response pathway and kills the treated plants, especially eudicots.119 TIR1 belongs to a small family of Auxin F-​box (AFB) proteins with six members

220

220

Chapter 11 116

in Arabidopsis. Interestingly, another synthetic auxin, picloram (4-​amino-​ 3,5,6-​ trichloropyridine-​ 2-​ carboxylic acid) specifically binds two TIR1/​ AFB members, AFB4 and 5, with high affinity to exert its herbicidal effect.120 In this context work from the Torii group developed a novel orthogonal auxin-​TIR1 pair,121 in which a synthetic, convex auxin (cvxIAA) is recognized and bound by a bioengineered complementary, concave TIR1 (ccvTIR1).121 Treatment of plants with the cvxIAA only induces an auxin response if plants express ccvTIR1. This system likely allows better investigation of auxin action, and specifically that of individual auxin receptors, and may help for crop improvement in context with auxin-​related traits.121 It is also discussed as a potential complementary approach to already existing controllable protein depletion systems that can be triggered by auxin, and which have been applied and established in human, yeast or Caenorhabditis elegans.122–​125 One of the closest TIR1/​AFB family members is another E3 substrate adaptor, Coronatine Insensitive1 (COI1), which serves as the substrate receptor for the phytohormone jasmonoyl isoleucine (JA-​Ile).126,127 JA-​Ile is critical for normal wound and pathogen responses, but also affects pollen development and meristem activity in plants.128,129 Coronatine (COR) and COR-​like compounds are important virulence factors synthesized and used by Pseudomonas syringae and Xanthomonas campestris pv. phormiicolai pathovars, respectiv ely.34,130,131 In general, they stimulate JA-​signaling while suppressing salicylic acid defense responses via antagonistic crosstalk between the two plant hormones. COR presence in the cell prevents necrosis, promotes opening of stomata and chlorosis in leaves, but suppresses cell wall defense reactions, such as callose deposition, all of which helps facilitate bacterial infection.131,132 Most of these responses are considered to be COI1-​dependent; however, there is evidence suggesting that COR also acts in a COI1-​independent manner (for a review see Ref. 132). Although CORs are not used as commercial herbicides, the example demonstrates another effective utilization by pathogens to undermine the host UPP to their benefit. The third phytohormone shown to be directly perceived and bound by an E3 ligase is salicylic acid (SA). The substrate adaptors in this CRL3 are Nonexpressor of Pathogenesis-​Related (NPR) proteins, which are closely connected with pathogen-​response.54 The interplay between SA and the NPR proteins are critical for plant immune response and systemic acquired resistance.133 To our knowledge, CRL3NPR E3 ligases have not been reported as targets of pathogen effector proteins or bacterial toxins, nor have any commercial drugs been developed as potential herbicides. Nevertheless, this may provide a possible angle for developing drugs that modulate pathogen sensitivity in plants, for example, by developing novel SA-​derivatives that can be bound with different affinities by NPR proteins. Alternatively, a synthetic biology approach may also be suitable here as described for auxin and established by the Torii group.121 A very interesting approach that may allow development of novel plant-​related drugs comes from proteolysis-​targeting chimeric molecules or PROTACs, which were first reported by the Crews and Deshaies laboratories in 2001 for human-​ related research.134 The technique is based on synthetic, bifunctional molecules

 221

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering 135,136

221

with two head groups that are connected by a flexible linker. One head group represents a ligand for a specific target protein, while the second group is recognized by an E3 ligase. Originally SCF complexes were used as the E3 ligase partner, but the technique has also been applied for other CRL E3s, and in principle any E3 ligase might be suitable for this technology.134,137,138 What is needed for PROTAC development is a very solid understanding of the structure of the two head groups in order to model, for example, a small peptide ligand and how it folds to be recognized by an E3 substrate receptor,137,139 or how a small chemical group is recognized by the target protein, such as an estrogen receptor.140 Upon treatment of a cell with such a PROTAC, the target binds to one head group (e.g., the peptide ligand or a chemical group that is recognized by the target), while the second head group mediates interaction with an E3 ligase, and thereby promotes degradation of the target and its depletion from the cell. In principle this can also be applied to plants, and may open up a broad range of novel drugs that can be used as herbicides or fungicides. As outlined above, many effector proteins from pathogenic bacteria target and modify E3 ligase activities, and one can imagine designing PROTACs against these effector proteins that actually promote their degradation, thereby improving immune responses on the plant side. For example such a target could be the effector protein AvrPtoB from P. syringae, for which the crystal structure is known.101 By modelling a peptide recognized by AvrPtoB, a destabilizing PROTAC could be designed that turns AvrPtoB into an E3 ligase target. Likewise, it is probably feasible to develop herbicidal or pesticidal PROTACs that would cause cell death in a plant or a pathogen, respectively. Currently, some of the E3 ligases involved in auxin, JA-​Ile and SA perception may have sufficient structural information available to develop specific ligands recognized, e.g., by TIR1 or COI1.141–​146 Another set of interesting candidate proteins that have recently been described to function as substrate receptors to an SCF complex, are F-​box-​Nictaba proteins.147–​150 These lectin like-​proteins appear to be present in small families across different plant species, but currently published examples only come from Arabidopsis.147,148,150 Carrying an F-​box motif in their N-​terminal region, their C-​terminal shares homologies with the tobacco Nictaba protein, which binds to glycans.147 Binding of N-​ and O-​glycans that contain N-​acetyllactosamine structures (Galβ1-​3GlcNAc, Galβ1-​4GlcNAc and poly-​N-​acetyllactosamine) has been proven for one of these members.147,150 The protein is upregulated upon infection with P. syringae and its overexpression has a protective function against the pathogen. While the underlying molecular mechanisms are not yet understood,150 this system could be a valuable resource for PROTAC development using specific glycans as one head group. Looking at human research, as well as the provided examples in this section, designing highly specific and novel UPP-​dependent drugs for plants seems quite feasible. PROTACs, like any other drug, would have the distinct advantages that the generation of transgenic organisms is not required, and that they can be directly applied. On the other hand, large-​scale synthesis and application presents other challenges, such as ensuring efficient uptake by the intended target, be that plant tissue or pathogen.

222

222

Chapter 11

11.4.2 Synthetic Biology Using UPP Sensors The UPP also allows the development of novel tools that combine the ability of specific receptor proteins with the proteasomal activity of the pathway. Such tools can be very useful in controlling the activity of a certain protein upon a specific signal, and for analysis of subsequent cellular responses or developmental changes. In the following, examples are given that use a synthetic biology approach toward accomplishing such a goal. The first example is an optogenetic module that was designed to induce proteasomal degradation of target proteins upon blue light induction in mammalian cells, called B-​ LID for blue light-​ inducible degradation domain.151,152 To accomplish this, a light-​sensitive Light-​Oxygen-​Voltage2 (LOV2) domain from oat (Avena sativa) phototropin1 protein was used.152 The domain undergoes a structural change upon blue light illumination that leads to an unwinding of a C-​terminal Jα helix. Bonger and coworkers fused a small degron motif (RRRG) to this helix that becomes exposed after blue-​light treatment and causes proteasomal degradation (Figure 11.3A).152 Consequently, any protein translationally fused to the B-​LID will also be degraded, providing an elegant way to control protein levels in a light-​ dependent manner. Modifications to this strategy, such as plant-​specific degrons coupled to UV or high-​light domains, could provide alternative designs for improved tolerances in crops. An interesting extension on this idea recently developed for plants are Hormone Activated Cas9-​based Repressors (HACRs).153 These are synthetic transcriptional repressors that consist of a deactivated Cas9 (dCAS9) protein from Streptococcus pyogenes, which can still bind gRNA but does not cut target DNA.154 The dCAS9 is translationally fused to a destabilizing degron motif and a partial TOPLESS repressor (TPL) protein.153,155 Co-​expression of a specific gRNA targets the HACR to defined promoter regions and leads to repression of the respective gene based on the TPL. The group utilized three different degron motifs sensitive to either auxin,156 JA157 or gibberrellic acid (GA).158 By treatment with one of these phytohormones, degradation of the HACR is induced and the repressed gene can be expressed again (Figure 11.3B). Using this system, Khakhar and coworkers could show that the system is capable of controlling expression of an auxin transporter and thereby affecting apical dominance and shoot branching in Arabidopsis.153 The system may be very useful and widely applicable to plants by using different types of degrons and gRNAs to affect a variety of agriculturally relevant traits.

11.4.3 The N-​degron Pathway as Bioengineering Tool The N-​degron pathway has been long known as a regulatory mechanism that controls half-​life of proteins, ranging from seconds to days, based on their N-​terminal amino acid residues.159–​161 As the traditional “start” of protein sequences, methionine is not a destabilizing residue per se, so pro-​and

 223

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

223

Figure 11.3 Using the UPP in synthetic biology. (A) The first example shows a blue light-​inducible degradation domain. It consists of a blue light-​sensitive LOV domain from a phototropin, fused to a Jα-​helix and a degron (D) motif. In the dark the Jα-​helix is bound to the LOV domain, and masks the degron motif. Blue light illumination opens up the LOV/​Jα-​ helix interaction and makes the degron detectable. A protein of interest (POI) fused C-​terminally to the degron motif is active in the dark but becomes degraded by the 26S proteasome upon blue light exposure. (B) The second example is a Hormone Activated Cas9-​based Repressors (HACRs) system, which consists of a deactivated CAS9 (dCAS9) that can still bind to gRNA (red line), a phytohormone-​inducible degron (D) motif, and a partial TOPLESS repressor protein (rep) to repress gene expression. HACRs bind to a specific promoter region targeted by the gRNA, and cause transcriptional repression of the gene of interest (GOI). Treatment with specific phytohormones (PH) promote degradation of the HACR and release gene expression.

eukaryotes must undergo post-​translational modifications in order to have another amino acid present at the N-​terminus. This can occur either by a methionine aminopeptidase, for example, or by an endoprotease that cleaves a pre-​cursor protein into N-​terminal and C-​terminal portions.162,163 As a consequence, the resulting protein has now a different N-​terminal amino acid present and can be directly recognized by so-​called N-​recognins. The N-​terminal residue may need to undergo further modification before a protein is identifiable by an N-​recognin (for comprehensive reviews on this topic see 162 BIB-​ 162–​164 BIB-​164). In plants, the known N-​recognins are Proteolysis 1 (PRT1) and PRT6, two RING-​finger-​type E3 ligases that ubiquitylate bound proteins to promote their degradation via the 26S proteasome.162,165,166 The N-​degron pathway has been demonstrated to be critical for stress tolerance, senescence and development in plants,167–​171 but has also been established as a bioengineering tool over the last years.162,165,172,173

224

224

Chapter 11

To study the N-​degron pathway and how certain amino acid residues affect protein half-​lives, early experiments done in yeast used UBQ fused to C-​terminal reporter protein, also known as the UBQ-​fusion technique.174–​176 Upon expression, the UBQ gets cleaved off by a de-​ubiquitylation enzyme, exposing a new N-​terminal amino acid residue on the reporter protein. By testing a set of expression constructs that have different amino acids exposed after UBQ cleavage, it was possible to determine stabilizing and destabilizing residues by tracking either the presence of the reporter or its activity.175 This led to the finding that basic amino acids, such as Arg, Lys and His, along with bulky ones, such as Ile, Leu, Phe, Trp and Tyr, are destabilizing, while Met, Ala. or Pro result in comparably stable proteins (Figure 11.4).175 Because the N-​degron system is mostly conserved among yeast, animals and plants, the ubiquitin fusion approach has been used in very elegant ways in the last years to develop condition-​specific lines, and the following work published by the Dissmeyer group may give a taste of what can be done with the pathway in plants. The group started out with an earlier developed temperature-​dependent system for yeast177 that used three components: (1) a ubiquitin fused to (2) a thermolabile mutant dihydrofolate reductase (DHFRts) from mouse, and (3) a protein of interest (POI). The ubiquitin becomes cleaved off by a de-​ ubiquitylating enzyme generating a DHFRts:POI fusion protein with a destabilizing N-​terminal amino acid residue. However, this residue under permissive, low temperatures is buried inside the DHFRts protein, and therefore not recognized by the N-​degron system. Upon treatment to the temperature that the DHFRts looses conformation, the residue is exposed and the DHFRts:POI fusion protein becomes degraded.177,178 The original temperature that caused DHFRts to denature was around 37–​42°C, which was too high for many model organisms, including plants, to be efficiently used. The Dissmeyer group further mutated DHFR to generate a DHFRT39A/​E173D (DHFRK2) protein that could denature at around ~29°C, a temperature that is a mild heat stress for most plants (further referred to as low-​temperature-​controlled N-​terminal

Figure 11.4 The UBQ-​fusion technique. In the example shown a protein of interest (POI) is fused to a UBQ. In planta, the UBQ is cleaved off by a de-​ ubiquitylation enzyme (DUB). This leads to exposure of an amino acid residue (X) at the N-​terminus of the POI, and which can be individually engineered. Depending on the exposed amino acid residue after UBQ removal, the POI is quickly ubiquitylated via the N-​degron pathway and removed from the cell.

 225

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

225

degradation or lt-​degron). Using this novel lt-​degron, the group showed that they can control developmental pathways, such as trichome formation or flowering time, in Arabidopsis by expressing certain key regulatory proteins in a temperature-​dependent manner.173 In addition, the group also demonstrated that the lt-​degron is suitable to affect developmental programs by the conditional expression of a toxic bacterial BARNASE (BAR) gene under the control of a trichome-​specific promoter.172 BAR is a general RNAse that will unspecifically degrade RNA in the cell when present and therefore becomes toxic.179 Consequently, plants grown at permissive temperatures (14°C) did not develop trichomes, in contrast to plants exposed to restrictive temperatures (28°C).172 These results show that the ubiquitin fusion technique and the N-​degron system can be very powerful tools to use in plant biotechnology and may lead to further interesting applications in the future.

11.5 Future Perspectives As stated above, the ubiquitin proteasome pathway represents a broad range of opportunities that can be helpful for future plant research and the agricultural industry. Some areas are truly underexplored, such as specific drugs that could target receptors related to the UPP, or synthetic drugs such as PROTACs. They might provide powerful tools in times where classical herbicides, for example, are challenged and new ones are urgently needed. Synthetic biology, coupled with structural biology and deep understanding of regulatory pathway, are likely critical to utilize the UPP in a most efficient manner. The development of novel tools such as HACRS or the B-​LID may allow addressing some of the future challenges, and overall help to generate crop plants that secure food production.

11.6 Acknowledgments We apologize to everyone whose work has not been cited. This project was supported by the Agriculture and Food Research Initiative competitive grant 2019-​67013-​29160 of the USDA National Institute of Food and Agriculture (NIFA) to H.H. and a PhD fellowship from the Tafila Technical University, Jordan to R.A.S.

References 1. P. Borrelli, D. A. Robinson, L. R. Fleischer, E. Lugato, C. Ballabio, C. Alewell, K. Meusburger, S. Modugno, B. Schutt, V. Ferro, V. Bagarello, K. V. Oost, L. Montanarella and P. Panagos, Nat. Commun., 2017, 8, 2013. 2. M. Collins, R. Knutti, J. Arblaster, J.-​ L. Dufresne, T. Fichefet, P. Friedlingstein, X. Gao, W. J. Gutowski, T. Johns, G. Krinner, M. Shongwe, C. Tebaldi, A. J. Weaver and M. Wehner, in Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, ed. T. F. Stocker, D. Qin, G.-​K. Plattner, M. Tignor, S. K. Allen, J. Boschung,

226

226

3. 4. 5. 6.

7. 8. 9. 0. 1 11. 12. 3. 1 14. 5. 1 16. 17. 18. 19. 0. 2 21. 22. 23. 24. 5. 2 26.

Chapter 11

A. Nauels, A. Xia, V. Bex and P. M. Midgley, Cambridge University Press, Cambridge, 2013, pp. 1030–​1136. L. Chen and H. Hellmann, Mol. Plant, 2013, 6, 1388–​1404. Y. Ye and M. Rape, Nat. Rev. Mol. Cell Biol., 2009, 10, 755–​764. L. Zhang, M. Xu, E. Scotti, Z. J. Chen and P. Tontonoz, J. Lipid Res., 2013, 54, 1410–​1420. L. Petrucelli, D. Dickson, K. Kehoe, J. Taylor, H. Snyder, A. Grover, M. De Lucia, E. McGowan, J. Lewis, G. Prihar, J. Kim, W. H. Dillmann, S. E. Browne, A. Hall, R. Voellmy, Y. Tsuboi, T. M. Dawson, B. Wolozin, J. Hardy and M. Hutton, Human Mol. Genet., 2004, 13, 703–​714. Y. Yanagawa, J. A. Sullivan, S. Komatsu, G. Gusmaroli, G. Suzuki, J. Yin, T. Ishibashi, Y. Saijo, V. Rubio, S. Kimura, J. Wang and X. W. Deng, Genes Dev., 2004, 18, 2172–​2181. C. K. Walsh and A. Sadanandom, Front. Plant Sci., 2014, 5, 122. E. Kraft, S. L. Stone, L. Ma, N. Su, Y. Gao, O. S. Lau, X. W. Deng and J. Callis, Plant Physiol., 2005, 139, 1597–​1611. B. Sharma and T. K. Bhatt, Sci. Rep., 2017, 7, 8613. D. Jue, X. Sang, S. Lu, C. Dong, Q. Zhao, H. Chen and L. Jia, PLoS One, 2015, 10, e0143488. M. J. Bijlmakers, J. M. Teixeira, R. Boer, M. Mayzel, P. Puig-​Sarries, G. Karlsson, M. Coll, M. Pons and B. Crosas, Sci. Rep., 2016, 6, 29232. K. L. Borden and P. S. Freemont, Curr. Opin. Struct. Biol., 1996, 6, 395–​401. R. Lovering, I. M. Hanson, K. L. Borden, S. Martin, N. J. O’Reilly, G. I. Evan, D. Rahman, D. J. Pappin, J. Trowsdale and P. S. Freemont, Proc. Natl. Acad. Sci. U. S. A., 1993, 90, 2112–​2116. D. M. Wenzel and R. E. Klevit, BMC Biol., 2012, 10, 24. C. A. Joazeiro and A. M. Weissman, Cell, 2000, 102, 549–​552. C. W. Liew, H. Sun, T. Hunter and C. L. Day, Biochem. J., 2010, 431,  23–​29. G. Polekhina, C. M. House, N. Traficante, J. P. Mackay, F. Relaix, D. A. Sassoon, M. W. Parker and D. D. Bowtell, Nat. Struct. Biol., 2002, 9,  68–​75. G. Buchwald, P. van der Stoop, O. Weichenrieder, A. Perrakis, M. van Lohuizen and T. K. Sixma, EMBO J., 2006, 25, 2465–​2474. D. P. Satijn and A. P. Otte, Mol. Cell Biol., 1999, 19,  57–​68. S. L. Stone, H. Hauksdottir, A. Troy, J. Herschleb, E. Kraft and J. Callis, Plant Physiol., 2005, 137,  13–​30. H. Weber, A. Bernhardt, M. Dieterle, P. Hano, A. Mutlu, M. Estelle, P. Genschik and H. Hellmann, Plant Physiol., 2005, 137,  83–​93. C. M. Choi, W. M. Gray, S. Mooney and H. Hellmann, Arabidopsis Book, 2014, 12, e0175. A. Capron, O. Serralbo, K. Fulop, F. Frugier, Y. Parmentier, A. Dong, A. Lecureuil, P. Guerche, E. Kondorosi, B. Scheres and P. Genschik, Plant Cell, 2003, 15, 2370–​2382. A. Capron, L. Okresz and P. Genschik, Trends Plant Sci., 2003, 8,  83–​89. D. E. Spratt, K. Wu, J. Kovacev, Z. Q. Pan and G. S. Shaw, J. Biol. Chem., 2012, 287, 17374–​17385.

 227

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

227

27. J. M. Gagne, B. P. Downes, S. H. Shiu, A. M. Durski and R. D. Vierstra, Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 11519–​11524. 28. N. Zheng, B. A. Schulman, L. Song, J. J. Miller, P. D. Jeffrey, P. Wang, C. Chu, D. M. Koepp, S. J. Elledge, M. Pagano, R. C. Conaway, J. W. Conaway, J. W. Harper and N. P. Pavletich, Nature, 2002, 416, 703–​709. 29. G. Xu, H. Ma, M. Nei and H. Kong, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 835–​840. 30. X. Yang, U. C. Kalluri, S. Jawdy, L. E. Gunter, T. Yin, T. J. Tschaplinski, D. J. Weston, P. Ranjan and G. A. Tuskan, Plant Physiol., 2008, 148, 1189–​1200. 31. S. Gupta, V. Garg, C. Kant and S. Bhatia, BMC Genomics, 2015, 16, 67. 32. D. Bellieny-​Rabelo, A. E. Oliveira and T. M. Venancio, PLoS One, 2013, 8, e55127. 33. S. Maraschin Fdos, J. Memelink and R. Offringa, Plant J., 2009, 59, 100–​109. 34. J. Yan, H. Li, S. Li, R. Yao, H. Deng, Q. Xie and D. Xie, Plant Cell, 2013, 25, 486–​498. 35. O. Madmon, M. Mazuz, P. Kumari, A. Dam, A. Ion, E. Mayzlish-​Gati, E. Belausov, S. Wininger, M. Abu-​Abied, C. S. McErlean, L. J. Bromhead, R. Perl-​Treves, C. Prandi, Y. Kapulnik and H. Koltai, Planta, 2016, 243, 1419–​1427. 36. M. Gou, Z. Shi, Y. Zhu, Z. Bao, G. Wang and J. Hua, Plant J., 2012, 69, 411–​420. 37. J. Baute, S. Polyn, J. De Block, J. Blomme, M. Van Lijsebettens and D. Inze, Plant Cell Physiol., 2017, 58, 962–​975. 38. Y. He, C. Wang, J. D. Higgins, J. Yu, J. Zong, P. Lu, D. Zhang and W. Liang, Plant Cell, 2016, 28, 1879–​1893. 39. L. Liu, Y. Wu, Z. Liao, J. Xiong, F. Wu, J. Xu, H. Lan, Q. Tang, S. Zhou, Y. Liu and Y. Lu, Heredity, 2018, 120, 310–​328. 40. M. Zhuang, M. F. Calabrese, J. Liu, M. B. Waddell, A. Nourse, M. Hammel, D. J. Miller, H. Walden, D. M. Duda, S. N. Seyedin, T. Hoggard, J. W. Harper, K. P. White and B. A. Schulman, Mol. Cell, 2009, 36,  39–​50. 41. W. J. Errington, M. Q. Khan, S. A. Bueler, J. L. Rubinstein, A. Chakrabartty and G. G. Prive, Structure, 2012, 20, 1141–​1153. 42. D. J. Gingerich, J. M. Gagne, D. W. Salter, H. Hellmann, M. Estelle, L. Ma and R. D. Vierstra, J. Biol. Chem., 2005, 280, 18810–​18821. 43. R. Perez-​Torrado, D. Yamada and P. A. Defossez, BioEssays, 2006, 28, 1194–​1202. 44. D. J. Gingerich, K. Hanada, S. H. Shiu and R. D. Vierstra, Plant Cell, 2007, 19, 2329–​2348. 45. M. Juranic and T. Dresselhaus, Plant Signal. Beh., 2014, 9, e28242. 46. S. Mooney, R. Al-​Saharin, C. M. Choi, K. Tucker, C. Beathard and H. A. Hellmann, Cells, 2019, 8. 47. L. Chen, A. Bernhardt, J. Lee and H. Hellmann, Mol. Plant, 2015, 8, 242–​250. 48. L. Chen, J. H. Lee, H. Weber, T. Tohge, S. Witt, S. Roje, A. R. Fernie and H. Hellmann, Plant Cell, 2013, 25, 2253–​2264.

228

228

Chapter 11

49. H. Chahtane, B. Zhang, M. Norberg, M. LeMasson, E. Thevenon, L. Bako, R. Benlloch, M. Holmlund, F. Parcy, O. Nilsson and G. Vachon, New Phytol., 2018, 220, 579–​592. 50. B. Zhang, M. Holmlund, S. Lorrain, M. Norberg, L. Bako, C. Fankhauser and O. Nilsson, eLife, 2017, 6. 51. C. M. Ha, J. H. Jun, H. G. Nam and J. C. Fletcher, Plant Cell Physiol., 2004, 45, 1361–​1370. 52. M. J. Christians, D. J. Gingerich, M. Hansen, B. M. Binder, J. J. Kieber and R. D. Vierstra, Plant J., 2009, 57, 332–​345. 53. E. Lechner, N. Leonhardt, H. Eisler, Y. Parmentier, M. Alioua, H. Jacquet, J. Leung and P. Genschik, Dev. Cell, 2011, 21, 1116–​1128. 54. Z. Q. Fu, S. Yan, A. Saleh, W. Wang, J. Ruble, N. Oka, R. Mohan, S. H. Spoel, Y. Tada, N. Zheng and X. Dong, Nature, 2012, 486, 228–​232. 55. A. Thomann, V. Brukhin, M. Dieterle, J. Gheyeselinck, M. Vantard, U. Grossniklaus and P. Genschik, Plant J., 2005, 43, 437–​448. 56. M. Dieterle, A. Thomann, J. P. Renou, Y. Parmentier, V. Cognat, G. Lemonnier, R. Muller, W. H. Shen, T. Kretsch and P. Genschik, Plant J., 2005, 41, 386–​399. 57. A. Bernhardt, E. Lechner, P. Hano, V. Schade, M. Dieterle, M. Anders, M. J. Dubin, G. Benvenuto, C. Bowler, P. Genschik and H. Hellmann, Plant J., 2006, 47, 591–​603. 58. J. H. Lee, W. Terzaghi, G. Gusmaroli, J. B. Charron, H. J. Yoon, H. Chen, Y. J. He, Y. Xiong and X. W. Deng, Plant Cell, 2008, 20, 152–​167. 59. S. Biedermann and H. Hellmann, Plant J., 2010, 62, 404–​415. 60. X. Huang, P. Yang, X. Ouyang, L. Chen and X. W. Deng, PLoS Genet., 2014, 10, e1004218. 61. H. Chen, X. Huang, G. Gusmaroli, W. Terzaghi, O. S. Lau, Y. Yanagawa, Y. Zhang, J. Li, J. H. Lee, D. Zhu and X. W. Deng, Plant Cell, 2010, 22, 108–​123. 62. Y. Zhang, S. Feng, F. Chen, H. Chen, J. Wang, C. McCall, Y. Xiong and X. W. Deng, Plant Cell, 2008, 20, 1437–​1455. 63. M. Pazhouhandeh, J. Molinier, A. Berr and P. Genschik, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 3430–​3435. 64. E. Dumbliauskas, E. Lechner, M. Jaciubek, A. Berr, M. Pazhouhandeh, M. Alioua, V. Cognat, V. Brukhin, C. Koncz, U. Grossniklaus, J. Molinier and P. Genschik, EMBO J., 2011, 30, 731–​743. 65. K. I. Seo, J. H. Lee, C. D. Nezames, S. Zhong, E. Song, M. O. Byun and X. W. Deng, Plant Cell, 2014, 26, 695–​711. 66. J. H. Lee, W. Terzaghi and X. W. Deng, Plant Sci., 2011, 180, 352–​357. 67. D. Li, L. Zhang, X. Li, X. Kong, X. Wang, Y. Li, Z. Liu, J. Wang, X. Li and Y. Yang, Plant Cell Environ., 2018, 41, 231–​244. 68. A. Bernhardt, S. Mooney and H. Hellmann, Planta, 2010, 232, 555–​566. 69. D. L. Lindsay, P. C. Bonham-​Smith, S. Postnikoff, G. R. Gray and T. A. Harkness, Planta, 2011, 233, 1223–​1235. 70. N. Takahashi, T. Kajihara, C. Okamura, Y. Kim, Y. Katagiri, Y. Okushima, S. Matsunaga, I. Hwang and M. Umeda, Curr. Biol., 2013, 23, 1812–​1817.

 229

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

229

71. Y. Wang, Y. Hou, H. Gu, D. Kang, Z. L. Chen, J. Liu and L. J. Qu, J. Integr. Plant Biol., 2013, 55,  64–​74. 72. I. Blilou, F. Frugier, S. Folmer, O. Serralbo, V. Willemsen, H. Wolkenfelt, N. B. Eloy, P. C. Ferreira, P. Weisbeek and B. Scheres, Genes Dev., 2002, 16, 2566–​2575. 73. Z. Zhang, J. Yang, E. H. Kong, W. C. Chao, E. P. Morris, P. C. da Fonseca and D. Barford, Biochem. J., 2013, 449, 365–​371. 74. J. Van Leene, J. Hollunder, D. Eeckhout, G. Persiau, E. Van De Slijke, H. Stals, G. Van Isterdael, A. Verkest, S. Neirynck, Y. Buffel, S. De Bodt, S. Maere, K. Laukens, A. Pharazyn, P. C. Ferreira, N. Eloy, C. Renne, C. Meyer, J. D. Faure, J. Steinbrenner, J. Beynon, J. C. Larkin, Y. Van de Peer, P. Hilson, M. Kuiper, L. De Veylder, H. Van Onckelen, D. Inze, E. Witters and G. De Jaeger, Mol. Syst. Biol., 2010, 6, 397. 75. Z. Kevei, M. Baloban, O. Da Ines, H. Tiricz, A. Kroll, K. Regulski, P. Mergaert and E. Kondorosi, PLoS One, 2011, 6, e20618. 76. K. Fulop, S. Tarayre, Z. Kelemen, G. Horvath, Z. Kevei, K. Nikovics, L. Bako, S. Brown, A. Kondorosi and E. Kondorosi, Cell Cycle, 2005, 4, 1084–​1092. 77. C. Breuer, K. Morohashi, A. Kawamura, N. Takahashi, T. Ishida, M. Umeda, E. Grotewold and K. Sugimoto, EMBO J., 2012, 31, 4488–​4501. 78. R. Kasili, J. D. Walker, L. A. Simmons, J. Zhou, L. De Veylder and J. C. Larkin, Genetics, 2010, 185, 257–​268. 79. C. W. Vander Kooi, M. D. Ohi, J. A. Rosenberg, M. L. Oldham, M. E. Newcomer, K. L. Gould and W. J. Chazin, Biochemistry, 2006, 45, 121–​130. 80. M. D. Ohi, C. W. Vander Kooi, J. A. Rosenberg, W. J. Chazin and K. L. Gould, Nat. Struct. Biol., 2003, 10, 250–​255. 81. L. R. Zeng, C. H. Park, R. C. Venu, J. Gough and G. L. Wang, Mol. Plant, 2008, 1, 800–​815. 82. N. Wang, Y. Liu, Y. Cong, T. Wang, X. Zhong, S. Yang, Y. Li and J. Gai, Plant Cell Physiol., 2016, 57, 1189–​1209. 83. K. J. Dietz, I. Turkan and A. Krieger-​Liszkay, Plant Physiol., 2016, 171, 1541–​1550. 84. F. Yu, Y. Wu and Q. Xie, Mol. Plant, 2016, 9,  21–​33. 85. W. Liang, S. van Wersch, M. Tong and X. Li, New Phytol., 2019, 221, 2054–​2066. 86. M. Tong, T. Kotur, W. Liang, K. Vogelmann, T. Kleine, D. Leister, C. Brieske, S. Yang, D. Ludke, M. Wiermer, Y. Zhang, X. Li and S. Hoth, New Phytol., 2017, 215, 1516–​1532. 87. A. Kinoshita, M. Seo, Y. Kamiya and S. Sawa, Plant Signal. Beh., 2015, 10, e1028707. 88. A. Kinoshita, C. A. ten Hove, R. Tabata, M. Yamada, N. Shimizu, T. Ishida, K. Yamaguchi, S. Shigenobu, Y. Takebayashi, S. Iuchi, M. Kobayashi, T. Kurata, T. Wada, M. Seo, M. Hasebe, I. Blilou, H. Fukuda, B. Scheres, R. Heidstra, Y. Kamiya and S. Sawa, Development, 2015, 142, 444–​453. 89. E. Indriolo, P. Tharmapalan, S. I. Wright and D. R. Goring, Plant Cell, 2012, 24, 4607–​4620. 90. I. Marin, PLoS One, 2013, 8, e68536.

230

230

Chapter 11

91. B. P. Downes, R. M. Stupar, D. J. Gingerich and R. D. Vierstra, Plant J., 2003, 35, 729–​742. 92. I. Alam, D. L. Cui, K. Batool, Y. Q. Yang and Y. H. Lu, Genes, 2019, 10. 93. Y. Miao and U. Zentgraf, Plant J., 2010, 63, 179–​188. 94. J. J. Furniss, H. Grey, Z. Wang, M. Nomoto, L. Jackson, Y. Tada and S. H. Spoel, PLoS Pathog., 2018, 14, e1007447. 95. M. J. Banfield, Cell Microbiol., 2015, 17,  18–​25. 96. S. Magori and V. Citovsky, Front. Plant Sci., 2011, 2, 87. 97. J. P. Anderson, C. A. Gleason, R. C. Foley, P. H. Thrall, J. B. Burdon and K. B. Singh, Funct. Plant Biol., 2010, 37, 499–​512. 98. S. B. Gelvin, Methods Mol. Biol., 2006, 343,  77–​84. 99. T. Tzfira, M. Vaidya and V. Citovsky, Nature, 2004, 431,  87–​92. 100. C. R. Buell, V. Joardar, M. Lindeberg, J. Selengut, I. T. Paulsen, M. L. Gwinn, R. J. Dodson, R. T. Deboy, A. S. Durkin, J. F. Kolonay, R. Madupu, S. Daugherty, L. Brinkac, M. J. Beanan, D. H. Haft, W. C. Nelson, T. Davidsen, N. Zafar, L. Zhou, J. Liu, Q. Yuan, H. Khouri, N. Fedorova, B. Tran, D. Russell, K. Berry, T. Utterback, S. E. Van Aken, T. V. Feldblyum, M. D’Ascenzo, W. L. Deng, A. R. Ramos, J. R. Alfano, S. Cartinhour, A. K. Chatterjee, T. P. Delaney, S. G. Lazarowitz, G. B. Martin, D. J. Schneider, X. Tang, C. L. Bender, O. White, C. M. Fraser and A. Collmer, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 10181–​10186. 101. J. Dong, F. Xiao, F. Fan, L. Gu, H. Cang, G. B. Martin and J. Chai, Plant Cell, 2009, 21, 1846–​1859. 102. S. Gimenez-​Ibanez, D. R. Hann, V. Ntoukakis, E. Petutschnig, V. Lipka and J. P. Rathjen, Curr. Biol., 2009, 19, 423–​429. 103. J. Mathieu, S. Schwizer and G. B. Martin, PLoS Pathog., 2014, 10, e1004227. 104. V. Gohre, T. Spallek, H. Haweker, S. Mersmann, T. Mentzel, T. Boller, M. de Torres, J. W. Mansfield and S. Robatzek, Curr. Biol., 2008, 18, 1824–​1832. 105. B. P. Thomma, T. Nurnberger and M. H. Joosten, Plant Cell, 2011, 23,  4–​15. 106. A. Gibly, A. Bonshtien, V. Balaji, P. Debbie, G. B. Martin and G. Sessa, Mol. Plant–​Microbe Interact., 2004, 17, 1212–​1222. 107. A. U. Singer, S. Schulze, T. Skarina, X. Xu, H. Cui, L. Eschen-​Lippold, M. Egler, T. Srikumar, B. Raught, J. Lee, D. Scheel, A. Savchenko and U. Bonas, PLoS Pathog., 2013, 9, e1003121. 108. X. Yan, J. Tao, H. L. Luo, L. T. Tan, W. Rong, H. P. Li and C. Z. He, Res. Microbiol., 2019, 170, 138–​146. 109. Y. Yang, J. Kitagaki, R. M. Dai, Y. C. Tsai, K. L. Lorick, R. L. Ludwig, S. A. Pierre, J. P. Jensen, I. V. Davydov, P. Oberoi, C. C. Li, J. H. Kenten, J. A. Beutler, K. H. Vousden and A. M. Weissman, Cancer Res., 2007, 67, 9472–​9481. 110. T. A. Soucy, P. G. Smith, M. A. Milhollen, A. J. Berger, J. M. Gavin, S. Adhikari, J. E. Brownell, K. E. Burke, D. P. Cardin, S. Critchley, C. A. Cullis, A. Doucette, J. J. Garnsey, J. L. Gaulin, R. E. Gershman, A. R. Lublinsky, A.

 231

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

231

McDonald, H. Mizutani, U. Narayanan, E. J. Olhava, S. Peluso, M. Rezaei, M. D. Sintchak, T. Talreja, M. P. Thomas, T. Traore, S. Vyskocil, G. S. Weatherhead, J. Yu, J. Zhang, L. R. Dick, C. F. Claiborne, M. Rolfe, J. B. Bolen and S. P. Langston, Nature, 2009, 458, 732–​736. 111. S. Tong, Y. Si, H. Yu, L. Zhang, P. Xie and W. Jiang, Sci. Rep., 2017, 7, 5599. 112. J. Cheng, Y. H. Fan, X. Xu, H. Zhang, J. Dou, Y. Tang, X. Zhong, Y. Rojas, Y. Yu, Y. Zhao, S. A. Vasudevan, H. Zhang, J. G. Nuchtern, E. S. Kim, X. Chen, F. Lu and J. Yang, Cell Death Dis., 2014, 5, e1079. 113. E. Drakos, R. R. Singh, G. Z. Rassidakis, E. Schlette, J. Li, F. X. Claret, R. J. Ford, Jr., F. Vega and L. J. Medeiros, Leukemia, 2011, 25, 856–​867. 114. W. Liu, J. Feng and J. Xu, Asian J. Androl., 2012, 14, 799–​800. 115. M. H. Lee, R. Zhao, L. Phan and S. C. Yeung, Cell Cycle, 2011, 10, 3057–​3066. 116. L. I. Calderon Villalobos, S. Lee, C. De Oliveira, A. Ivetac, W. Brandt, L. Armitage, L. B. Sheard, X. Tan, G. Parry, H. Mao, N. Zheng, R. Napier, S. Kepinski and M. Estelle, Nat. Chem. Biol., 2012, 8, 477–​485. 117. M. Salehin, R. Bagchi and M. Estelle, Plant Cell, 2015, 27,  9–​19. 118. N. Dharmasiri, S. Dharmasiri and M. Estelle, Nature, 2005, 435, 441–​445. 119. A. H. C. Anesio, M. V. Santos, R. R. Silveira, E. A. Ferreira, J. B. D. Santos and L. D. D. Silva, An. Acad. Bras. Cienc., 2018, 90, 1717–​1732. 120. M. J. Prigge, K. Greenham, Y. Zhang, A. Santner, C. Castillejo, A. M. Mutka, R. C. O’Malley, J. R. Ecker, B. N. Kunkel and M. Estelle, G3, 2016, 6, 1383–​1390. 121. N. Uchida, K. Takahashi, R. Iwasaki, R. Yamada, M. Yoshimura, T. A. Endo, S. Kimura, H. Zhang, M. Nomoto, Y. Tada, T. Kinoshita, K. Itami, S. Hagihara and K. U. Torii, Nat. Chem. Biol., 2018, 14, 299–​305. 122. K. Nishimura, T. Fukagawa, H. Takisawa, T. Kakimoto and M. Kanemaki, Nat. Methods, 2009, 6, 917–​922. 123. A. J. Holland, D. Fachinetti, J. S. Han and D. W. Cleveland, Proc. Natl. Acad. Sci. U. S. A., 2012, 109, E3350–​3357. 124. M. Kanke, K. Nishimura, M. Kanemaki, T. Kakimoto, T. S. Takahashi, T. Nakagawa and H. Masukata, BMC Cell Biol., 2011, 12, 8. 125. L. Zhang, J. D. Ward, Z. Cheng and A. F. Dernburg, Development, 2015, 142, 4374–​4384. 126. L. B. Sheard, X. Tan, H. Mao, J. Withers, G. Ben-​Nissan, T. R. Hinds, Y. Kobayashi, F. F. Hsu, M. Sharon, J. Browse, S. Y. He, J. Rizo, G. A. Howe and N. Zheng, Nature, 2010, 468, 400–​405. 127. B. Thines, L. Katsir, M. Melotto, Y. Niu, A. Mandaokar, G. Liu, K. Nomura, S. Y. He, G. A. Howe and J. Browse, Nature, 2007, 448, 661–​665. 128. J. Kim, B. Dotson, C. Rey, J. Lindsey, A. B. Bleecker, B. M. Binder and S. E. Patterson, PLoS One, 2013, 8, e60505. 129. A. Gfeller, R. Liechti and E. E. Farmer, Science’s STKE, 2006, cm1. 130. X. Chen, H. Huang, T. Qi, B. Liu and S. Song, Plant Signal. Beh., 2016, 11, e1135280. 131. X. Geng, J. Cheng, A. Gangadharan and D. Mackey, Plant Cell, 2012, 24, 4763–​4774.

232

232

Chapter 11

132. X. Geng, L. Jin, M. Shimada, M. G. Kim and D. Mackey, Planta, 2014, 240, 1149–​1165. 133. R. Backer, S. Naidoo and N. van den Berg, Front. Plant Sci., 2019, 10, 102. 134. K. M. Sakamoto, K. B. Kim, A. Kumagai, F. Mercurio, C. M. Crews and R. J. Deshaies, Proc. Natl. Acad. Sci. U. S. A., 2001, 98, 8554–​8559. 135. S. An and L. Fu, EBioMedicine, 2018, 36, 553–​562. 136. S. Gu, D. Cui, X. Chen, X. Xiong and Y. Zhao, BioEssays, 2018, 40, e1700247. 137. M. Girardini, C. Maniaci, S. J. Hughes, A. Testa and A. Ciulli, Bioorgan. Med. Chem., 2019, 27, 2466–​2479. 138. S. Lindner, C. Steinebach, H. Kehm, M. Mangold, M. Gutschow and J. Kronke, J. Vis. Exp., 2019, 147, e59472. 139. Y. Jiang, Q. Deng, H. Zhao, M. Xie, L. Chen, F. Yin, X. Qin, W. Zheng, Y. Zhao and Z. Li, ACS Chem. Biol., 2018, 13, 628–​635. 140. K. Cyrus, M. Wehenkel, E. Y. Choi, H. Lee, H. Swanson and K. B. Kim, ChemMedChem, 2010, 5, 979–​985. 141. K. Hayashi, J. Neve, M. Hirose, A. Kuboki, Y. Shimada, S. Kepinski and H. Nozaki, ACS Chem. Biol., 2012, 7, 590–​598. 142. X. Tan, L. I. Calderon-​Villalobos, M. Sharon, C. Zheng, C. V. Robinson, M. Estelle and N. Zheng, Nature, 2007, 446, 640–​645. 143. G. O. Badescu and R. M. Napier, Trends Plant Sci., 2006, 11, 217–​223. 144. C. Wasternack and E. Kombrink, ACS Chem. Biol., 2010, 5,  63–​77. 145. E. Adams and J. Turner, J. Exp. Bot., 2010, 61, 4373–​4386. 146. J. Yan, C. Zhang, M. Gu, Z. Bai, W. Zhang, T. Qi, Z. Cheng, W. Peng, H. Luo, F. Nan, Z. Wang and D. Xie, Plant Cell, 2009, 21, 2220–​2236. 147. K. Stefanowicz, N. Lannoo, P. Proost and E. J. Van Damme, FEBS Open Bio, 2012, 2, 151–​158. 148. N. Lannoo, W. J. Peumans and E. J. Van Damme, Biochem. Soc. Trans., 2008, 36, 843–​847. 149. Y. Yoshida, F. Tokunaga, T. Chiba, K. Iwai, K. Tanaka and T. Tai, J. Biol. Chem., 2003, 278, 43877–​43884. 150. K. Stefanowicz, N. Lannoo, Y. Zhao, L. Eggermont, J. Van Hove, B. Al Atalah and E. J. Van Damme, BMC Plant Biol., 2016, 16, 213. 151. J. Baaske, P. Gonschorek, R. Engesser, A. Dominguez-​Monedero, K. Raute, P. Fischbach, K. Muller, E. Cachat, W. W. A. Schamel, S. Minguet, J. A. Davies, J. Timmer, W. Weber and M. D. Zurbriggen, Sci. Rep., 2018, 8, 15024. 152. K. M. Bonger, R. Rakhit, A. Y. Payumo, J. K. Chen and T. J. Wandless, ACS Chem. Biol., 2014, 9, 111–​115. 153. A. Khakhar, A. R. Leydon, A. C. Lemmex, E. Klavins and J. L. Nemhauser, eLife, 2018, 7. 154. L. A. Gilbert, M. H. Larson, L. Morsut, Z. Liu, G. A. Brar, S. E. Torres, N. Stern-​Ginossar, O. Brandman, E. H. Whitehead, J. A. Doudna, W. A. Lim, J. S. Weissman and L. S. Qi, Cell, 2013, 154, 442–​451. 155. E. Pierre-​Jerome, S. S. Jang, K. A. Havens, J. L. Nemhauser and E. Klavins, Proc. Natl. Acad. Sci. U. S. A., 2014, 111, 9407–​9412.

 233

Plant E3 Ligases as Versatile Tools for Drug Development and Bioengineering

233

156. B. L. Moss, H. Mao, J. M. Guseman, T. R. Hinds, A. Hellmuth, M. Kovenock, A. Noorassa, A. Lanctot, L. I. Villalobos, N. Zheng and J. L. Nemhauser, Plant Physiol., 2015, 169, 803–​813. 157. L. Katsir, H. S. Chung, A. J. Koo and G. A. Howe, Curr. Opin. Plant Biol., 2008, 11, 428–​435. 158. K. Murase, Y. Hirano, T. P. Sun and T. Hakoshima, Nature, 2008, 456, 459–​463. 159. A. Bachmair, D. Finley and A. Varshavsky, Science, 1986, 234, 179–​186. 160. A. Belle, A. Tanay, L. Bitincka, R. Shamir and E. K. O’Shea, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 13004–​13009. 161. H. C. Yen, Q. Xu, D. M. Chou, Z. Zhao and S. J. Elledge, Science, 2008, 322, 918–​923. 162. N. Dissmeyer, Annu. Rev. Plant Biol., 2019, 70, 83–​117. 163. A. Varshavsky, Proc. Natl. Acad. Sci. U. S. A., 2019, 116, 358–​366. 164. N. Dissmeyer, S. Rivas and E. Graciet, New Phytol., 2018, 218, 929–​935. 165. A. C. Mot, E. Prell, M. Klecker, C. Naumann, F. Faden, B. Westermann and N. Dissmeyer, New Phytol., 2018, 217, 613–​624. 166. H. Dong, J. Dumenil, F. H. Lu, L. Na, H. Vanhaeren, C. Naumann, M. Klecker, R. Prior, C. Smith, N. McKenzie, G. Saalbach, L. Chen, T. Xia, N. Gonzalez, M. Seguela, D. Inze, N. Dissmeyer, Y. Li and M. W. Bevan, Genes Dev., 2017, 31, 197–​208. 167. S. Yoshida, M. Ito, J. Callis, I. Nishida and A. Watanabe, Plant J., 2002, 32, 129–​137. 168. C. Schuessele, S. N. Hoernstein, S. J. Mueller, M. Rodriguez-​Franco, T. Lorenz, D. Lang, G. L. Igloi and R. Reski, New Phytol., 2016, 209, 1014–​1027. 169. C. C. Lin, Y. T. Chao, W. C. Chen, H. Y. Ho, M. Y. Chou, Y. R. Li, Y. L. Wu, H. A. Yang, H. Hsieh, C. S. Lin, F. H. Wu, S. J. Chou, H. C. Jen, Y. H. Huang, D. Irene, W. J. Wu, J. L. Wu, D. J. Gibbs, M. C. Ho and M. C. Shih, Proc. Natl. Acad. Sci. U. S. A., 2019, 116, 3300–​3309. 170. H. Zhang, L. Gannon, P. D. Jones, C. A. Rundle, K. L. Hassall, D. J. Gibbs, M. J. Holdsworth and F. L. Theodoulou, Sci. Rep., 2018, 8, 15192. 171. J. Vicente, G. M. Mendiondo, M. Movahedi, M. Peirats-​Llobet, Y. T. Juan, Y. Y. Shen, C. Dambire, K. Smart, P. L. Rodriguez, Y. Y. Charng, J. E. Gray and M. J. Holdsworth, Curr. Biol., 2017, 27, 3183–​3190. 172. F. Faden, S. Mielke and N. Dissmeyer, Plant Physiol., 2019, 179, 929–​942. 173. F. Faden, T. Ramezani, S. Mielke, I. Almudi, K. Nairz, M. S. Froehlich, J. Hockendorff, W. Brandt, W. Hoehenwarter, R. J. Dohmen, A. Schnittger and N. Dissmeyer, Nat. Commun., 2016, 7, 12202. 174. A. Varshavsky, Methods Enzymol., 2005, 399, 777–​799. 175. A. Bachmair and A. Varshavsky, Cell, 1989, 56, 1019–​1032. 176. A. Varshavsky, A. Bachmair, D. Finley, D. K. Gonda and I. Wunning, Biotechnology, 1989, 13, 109–​143. 177. R. J. Dohmen, P. Wu and A. Varshavsky, Science, 1994, 263, 1273–​1276. 178. R. J. Dohmen and A. Varshavsky, Methods Enzymol., 2005, 399, 799–​822. 179. A. M. Buckle and A. R. Fersht, Biochemistry, 1994, 33, 1644–​1653.

234

CHAPTER 12

Deubiquitinase Inhibitors: An Emerging Therapeutic Class ROBERT S. MAGIN,a,b LAURA M. DOHERTYa,b,c AND SARA J. BUHRLAGEa,b* a

Department of Cancer Biology, Dana-​Farber Cancer Institute, Boston, MA 02215, USA b Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, MA 02115, USA c Department of Systems Biology and Laboratory of Systems Pharmacology, Harvard Medical School, Boston, MA 02115, USA *Corresponding author. Email: saraj_​[email protected]

12.1 Introduction/​Background Ubiquitin is a post-​translational modification (PTM), which regulates myriad important cellular processes, including protein homeostasis, transcription, autophagy, cell signaling, cell cycle progression and DNA repair.1,2 Proteins can be modified with mono-​or poly-​ubiquitin, and poly-​ubiquitin chains can be assembled by bond formation between the C-​terminus of one ubiquitin and one of seven lysine residues or the N-​terminus of the adjacent ubiquitin.3 Depending on the number of ubiquitin moieties and the pattern of ubiquitin connectivity, the addition of ubiquitin to substrates can modulate protein stability, protein–​protein interactions, protein localization or enzyme activity.1

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

234

 235

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

235

Ubiquitin is covalently attached to substrates by a cascade of enzymes (activating E1, conjugating E2 and ligase E3 enzymes) and removed by deubiquitinating enzymes (DUBs). Thus far, researchers have identified 99 human DUBs spanning seven families (Figure 12.1): 56 ubiquitin-​specific peptidases (USPs), 12 Jab1/​Mov34/​Mpr1 Pad1 N-​terminal + domain proteases (JAMMs), 17 ovarian tumor proteases (OTUs), four Machado–​Josephin Domain proteases (MJDs), four ubiquitin C-​terminal hydroxylases (UCHs), five motif interacting with ubiquitin-​containing novel DUB family proteases (MINDYs), and zinc finger-​containing ubiquitin peptidase 1 (ZUP1), which is the only member of the most recently discovered family in the DUB phylogenetic tree.4,5 The role of ubiquitin in regulating protein stability via the ubiquitin–​ proteasome and autophagy–​lysosome systems is the most well-​characterized function of ubiquitin.6 Both of these systems use ubiquitin as a signal to target proteins for degradation, with K48-​linked chains targeting proteins to the proteasome and K-​63-​linked chains targeting proteins to the lysosome.7 By cleaving ubiquitin from specific substrates, DUBs act to rescue proteins from degradation in these two degradation pathways. Certain DUBs also serve broad roles in regulating protein flux through the proteasome and lysosome. For example, three DUBs (USP14, UCHL5 and PSMD14) are associated with the proteasome.8 USP14 and UCHL5 cleave ubiquitin from proteins that bind transiently to the proteasome before they are committed to degradation, thus rescuing them from degradation. By contrast, PSMD14 removes ubiquitin from proteins as they undergo proteasomal degradation; thus, it is critical for proper proteasome function and is not restricted to specific substrates. USP8 and STAMBP are two DUBs that are associated with endosomal sorting complexes required for transport (ESCRT), and they regulate endocytic trafficking and determine whether receptors that are internalized within endosomes are directed to lysosomes to be degraded or recycled back to the cellular surface.9 The central role of ubiquitin in many cellular processes that are relevant for human diseases has generated significant interest in targeting DUBs for potential therapeutic benefit. Numerous studies have established that many human diseases are characterized by dysregulation of the ubiquitin system, including different types of cancers as well as inflammatory and neurodegenerative diseases; thus, there is growing interest in targeting pathologically overactive DUBs in these disease areas.10–​13 The successful use of proteasome inhibitors in the clinic suggests that there could be clinical benefit to targeting the DUBs that associate with the proteasome, even though inhibiting these DUBs may have far-​reaching effects on protein abundance rather than substrate-​specific effects. Particularly, researchers have been interested in developing inhibitors of these DUBs to determine if they may have utility in overcoming resistance to proteasome inhibitors. Also, because these proteasome-​bound DUBs have distinct functions in the proteasomal degradation pathway, small-​molecule inhibitors of these DUBs may show more nuanced effects.14 Researchers have been particularly keen on the prospect of targeting DUBs that exhibit substrate specificity and thus only regulate one or a handful of

236

236

Chapter 12

Figure 12.1 DUB phylogenetic tree. There are approximately 100 identified human DUBs, which cluster by the sequence similarity of their respective catalytic domain into seven families: ubiquitin-​specific protease (USP), ubiquitin C-​terminal hydroxylase (UCH), Machado–​Josephin Domain (MJD), Jab1/​ Mov34/​ Mpr1 Pad1 N-​ terminal + (JAMM), ovarian tumor (OTU), zinc finger-​containing ubiquitin peptidase 1 (ZUP1), and motif interacting with ubiquitin-​containing novel DUB family (MINDY).

 237

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

237

therapeutically relevant proteins, such as oncoproteins. Inhibiting a particular DUB will increase the ubiquitination of specific substrates without having broad impacts on the proteasomal degradation pathway or global protein abundance. Identifying the DUBs for traditionally intractable protein targets lacking a druggable binding pocket, such as transcription factors, intrinsically disordered proteins, or proteins involved in protein–​protein interactions, would be especially impactful.15 DUB inhibitors could also potentially overcome drug resistance by degrading mutant, drug-​resistant enzymes that are no longer modulated by direct inhibition.16 In response to these exciting possibilities, there has been considerable effort in recent years to identify the DUBs that regulate disease-​related proteins and to develop specific small-​molecule inhibitors of DUBs that show particular promise as potential therapeutic targets.17–​20 The first reported DUB inhibitors date back to more than 15 years ago, but until quite recently, there were questions over the quality and reliability of both the reported substrates and inhibitors.21 However, over the last few years, a number of rigorous small-​molecule screening and medicinal chemistry efforts have successfully yielded the first truly selective DUB inhibitors. In this chapter, we provide examples of the biological function of DUBs, highlight their therapeutic potential in disease and cell states, review best practices for inhibitor development and substrate discovery, and describe recently discovered potent and selective DUB inhibitors. These advances highlight the promise of targeting DUBs within the strategy of precision medicine, and underscore that DUBs represent an exciting, emerging class of therapeutic targets for modulation of clinically beneficial protein degradation.

12.2 Biology and Clinical Opportunity for DUB Inhibition The large number of human DUBs represents a broad and largely untapped group of enzymes for potential therapeutic uses. The number of different pathways modulated by DUBs spans a wide array of disease states, such as oncology, immunity/​ inflammation and neurodegeneration.5 The DUBs in these pathways have different functional effects and modulate ubiquitin signaling in diverse ways, such as substrate stability, modulating ubiquitin as a non-​degradative PTM and controlling the level of specific ubiquitin chain linkages.22 In this section, we give a number of examples demonstrating the diverse biological effects of DUBs and examine their potential as therapeutic targets.

12.2.1 USP7 USP7 (also known as HAUSP) is one of the most well-​studied DUBs in the human proteome. Although it has been implicated in many different pathways and has many reported substrates, its most robustly characterized role is in the MDM2–​p53 axis.23 MDM2 is an E3 ligase which ubiquitinates and targets the transcription factor p53 for degradation. Although both p53 and

238

238

Chapter 12 24

MDM2 are reported to be substrates of USP7, the converging consensus, at least for many cell types and states, is that the physiologically dominant effect of USP7 is to deubiquitinate and stabilize MDM2, leading to p53 degradation and cell cycle progression (Figure 12.2A).17,19,25 Stabilization of p53 in cancer treatment has been pursued for many years. Approximately 50% of all cancers have wild-​type p53, and increasing the abundance of functional p53 in this broad set of cancer types induces cell cycle arrest and apoptosis.26 Nutlins, which are MDM2 inhibitors, work by inhibiting the E3 activity of MDM2 and are currently in clinical trials, and have shown promising initial results.27 USP7 inhibitors work to stabilize p53 levels by inhibiting the deubiquitination of MDM2.17,19,25 Procuring a potent USP7 inhibitor has been sought for this reason. In addition to the potential clinical benefits of a USP7 inhibitor, it would also help elucidate the biological function of USP7 in general, and help confirm and identify new substrates.

Figure 12.2 Clinical interest in DUBs: oncology, inflammation and neurodegenerative diseases. (A) USP7 deubiquitinates and stabilizes MDM2, the E3 ligase for p53. (B) OTULIN disassembles linear ubiquitin chains, which are assembled by LUBAC. Linear ubiquitin chain regulate NF-​κB signaling and increase inflammation. (C) Damaged mitochondria activate Parkin, an E3 ligase that ubiquitinates outer mitochondrial membrane proteins, leading to mitophagy and cell survival. USP30 removes ubiquitin from mitochondrial proteins and leads to the accumulation of damaged mitochondria.

 239

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

239

12.2.2 USP22 DNA is packaged into chromatin by wrapping around an octameric core of histone proteins: two copies each of H2A, H2B, H3 and H4. The control of DNA replication, repair, recombination and transcription is exquisitely controlled by histone PTMs, such as methylation, acetylation, phosphorylation and ubiquitination, among others.28 These PTMs regulate the accessibility of chromatin to other proteins in the nucleus, serve as binding sites for functional effectors of DNA regulation, and crosstalk with each other in complex ways to establish what is known as the “histone code.”29 As these modifications are crucial for proper gene expression and cell division, the proteins which write, read and erase these PTMs have been heavily investigated. The best characterized ubiquitination site on histone proteins is K120 on histone H2B (K123 in yeast), which is a hallmark of active chromatin. This mono-​ubiquitination is non-​degradative and has many functions, including serving as a binding site for histone chaperones and methyltransferases, promoting transcriptional elongation by RNA polymerase, and also has been reported to recruit DNA damage response elements.30,31 The SAGA (Spt–​Ada–​ Gcn5) acetyltransferase complex contains a DUB module with a catalytic subunit of USP22 in mammals and Ubp8 in yeast.32 This DUB module is a well-​characterized H2B K120 deubiquitinase, which has roles in neurological development, embryogenesis and cell lineage specificity.33 Importantly, this complex has been implicated in diseases such as cancer and other pathologies, underscoring the importance of this DUB in disease states.34 Although to our knowledge there are no reported USP22 inhibitors, it remains an attractive target for inhibitor development.

12.2.3 OTULIN In 2013, a DUB from the OTU family of DUBs, termed OTULIN, was discovered to exclusively deubiquitinate linear (M1-​linked) ubiquitin chains, and was among the first DUBs dedicated to linear ubiquitin identified.35,36 Linear ubiquitin chains are assembled by the linear ubiquitin chain assembly complex (LUBAC) and regulate nuclear factor κB (NF-​κB) signaling, which is critical for inflammation and the immune response (Figure 12.2B).37 OTULIN mutations in humans lead to OTULIN-​related autoinflammatory syndrome (ORAS), causing severe inflammatory symptoms.38 There is a complex interplay of OTULIN and LUBAC regulation which occurs due to auto-​ubiquitination of LUBAC with linear ubiquitin.39 The interest in developing OTULIN inhibitors lies in studying the biology of inflammation and immunity, but also in modulating linear ubiquitin levels in the cell, and studying the effects of this unique ubiquitin linkage.

12.2.4 USP30 Parkinson’s disease is a complex condition characterized by death of dopaminergic neurons in the substantia nigra. While the etiology of the disease

240

240

Chapter 12

is incompletely understood, genetic and biochemical studies have strongly implicated oxidative damage and impaired mitophagy in disease progression.40 When mitochondria are damaged from oxidative stress, the kinase PINK1 is activated at the outer mitochondrial membrane (OMM) surface, and recruits and activates the E3 ligase parkin to mitochondria.41 There, parkin ubiquitinates OMM proteins and the damaged mitochondrion is targeted for mitophagy (Figure 12.2C). Parkinson’s-​associated familial mutations in PINK1 and parkin ablate the pathway, which allow for damaged mitochondria to accumulate and cause cell death.42 Therefore, stimulating mitophagy is a potential strategy for treating Parkinson’s disease. In 2014, Sheng and colleagues reasoned that one or more DUBs could be responsible for countering OMM ubiquitination and therefore antagonize mitophagy.43 In cultured cells overexpressing parkin, they found that overexpression of USP30 rescued mitochondrial degradation after treating the cells with protonophores. USP30 is known to localize to the OMM via an N-​ terminal transmembrane helix. In neurons in culture, wild-​type USP30 suppressed mitochondrial degradation, but a catalytic mutant did not, showing the effect was due to its deubiqutinating activity. Excitingly, the authors also showed that USP30 knockdown rescues mitophagy defects in cells expressing mutant parkin, and that USP30 knockdown in a Drosophila model of Parkinson’s largely counteracted mitochondrial defect, raised dopamine levels and increased fly survival. These data show how USP30 represents a very promising enzyme for inhibition to treat neurodegeneration.

12.3 Validating Inhibitors and Substrates of DUBs Despite the opportunity in targeting DUBs, there are still considerable hurdles in identifying DUB substrates, and there are only a limited set of convincingly validated small-​molecule inhibitors despite the growing number of tools to develop them.21 While we described a number of well-​characterized DUB substrates in the preceding section, it remains quite challenging to identify substrates with confidence. Similarly, many first-​generation DUB inhibitors suffered from off-​target effects, general toxicity and hyper-​reactivity. In this section, we will discuss a summary of practices that are increasingly being adopted as field standards to increase the confidence in the substrate or inhibitor of interest (Figure 12.3). While not comprehensive, we hope this will provide a guide for researchers in their own work and while reading the literature.

12.3.1 Substrate Validation As there are 100 DUBs, they have a wide variety of substrate proteins, and a large number of putative substrates have been reported. However, despite the many reported DUB substrates, the list of consistently reproducible, validated substrates remains relatively sparse. This is due to a variety of factors. Significant variation in the spatiotemporal expression and regulation of DUBs across different cell types, redundancy among the DUBs, and pleiotropic effects and artifacts from whole-​protein knockout can confound results and

 241

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

241

Figure 12.3 (A) DUB substrate validation. DUB substrates can be carefully validated with genetic tools, pulldown assays, readouts of ubiquitination levels on the substrate and unbiased proteomics. (B) DUB inhibitor assays. There is a suite of useful assays available for the rigorous characterization of DUB inhibitors to ensure the highest-​quality probes are developed for investigating the biological function and therapeutic potential of DUBs.

frustrate attempts at replication. Below we discuss general issues in identifying DUB substrates, caveats to keep in mind when thinking about DUB targets, and assays to increase confidence in the identified substrate. While it is often difficult for a study to employ all of these techniques, using them

242

242

Chapter 12

in combination is often the best practice to identify reliable and validated substrates. DUBs bind to their substrates through diverse mechanisms.5 Unlike other large enzyme families, like kinases, which are often specific for short consensus sequences of peptides,44 DUBs are organized more like E3 ligases, which use accessory domains, complex formation and localization to achieve substrate specificity.45 For example, USP7 uses its TRAF-​like domain to interact with MDM2 interaction domain,24 USP22 is targeted to deubiquitinate histone H2B as part of the larger SAGA complex DUB module46 and USP30 is localized to the OMM.47 Therefore, understating these factors is critical to understanding physiologically relevant DUB substrates. In addition, although certain DUBs do target very specific lysines, like USP22, many do not target ubiquitin chains on specific lysine residues, as DUBs act through a proximity mechanism similar to many E3 ligases, which can preclude site directed mutagenesis of specific lysine on putative substrates. The first steps in DUB target validation is usually knockdown or knockout of the DUB of interest to see the putative substrate level decrease (assuming the ubiquitin moiety or chain on the substrate targets the protein for degradation). Any decrease in cellular protein level should be confirmed to be at the protein and not mRNA level, and a proteasome or lysosome inhibitor used to identify whether degradation is occurring through the ubiquitin–​proteasome system or the lysosome. Protein pulldown followed by western blot for ubiquitin is also important to detect direct effect of ubiquitination on the protein of interest. To ensure that the effects are caused by the catalytic activity of the DUB, and not the more dramatic loss of the entire protein, rescue experiments should be performed with WT enzyme, and catalytic point mutants, which are usually straightforward to make for both the cysteine protease DUB family members by mutating the nucleophilic cysteine and the zinc metalloproteases by mutating zinc-​chelating residues. Notably, a recent publication suggests that the commonly used cysteine to alanine mutation causes an increase in ubiquitin chain binding of the DUB, which can sequester ubiquitin in the cell, and suggests caution when interpreting in cellulo results from these mutants. The authors showed the cysteine to arginine mutation abrogated catalytic activity of DUBs without an increase in ubiquitin chain binding, so this mutant may be more suitable to cell-​based assays.48 However, even the above assays should be interpreted with caution. DUB knockout or inhibition often has complex pleiotropic effects due to the change in the potentially large number of native substrates of the DUB. Looking for the same effect across multiple cell lines will help mitigate this problem, as well as finding a direct interaction between the DUB and its substrate. However, co-​IP from lysates suffer from artifacts such as overexpression of proteins, lack of evidence of a direct interaction and inability to determine stoichiometry from western blotting. Showing a direct interaction with recombinant protein increases the confidence in the interaction. In addition, interaction mapping allows for mutation of the substrate to confirm deubiquitination by the DUB of interest. Ultimately, complete biochemical reconstitution with the

 243

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

243

ubiquitinated protein and the DUB is ideal to do kinetic analysis, but practically unattainable for all but the most optimized systems. Finally, a non-​biased screen of DUB inactivation using proteomics with either CRISPR or a well-​validated inhibitor is an important consideration to make for these types of questions. The value of potent and specific small-​ molecule inhibitors with matched controls is highlighted below. The ability to modulate the endogenous enzyme with a small molecule eliminated artifacts from overexpression and gives exquisite control in the dosage and kinetics of in cellulo inhibition. Along these lines, recently developed technologies that enable rapid protein degradation that models small-​molecule inhibition will be valuable to incorporate into substrate validation and discovery pipelines for DUBs.49

12.3.2 Inhibitor Validation The rigorous characterization of small-​molecule inhibitors using orthogonal assays is of the utmost importance for DUB inhibitors, as for any target class. The ability to use DUB inhibitors as probes to evaluate the therapeutic relevance or fundamental biological function of specific DUBs requires a high level of confidence in the on-​target binding and selectivity profile. Unfortunately, the earliest reported DUB inhibitors have often not held up to more rigorous characterization and show broad, non-​specific phenotypic effects, but tremendous effort in recent years has led to the development of a number of useful DUB-​specific assays using ubiquitin-​based probes for screening and evaluating the potency and selectivity of DUB inhibitors. For initial high-​throughput screening and DUB target validation, the most frequently used assay is fluorescence-​based. A ubiquitin-​fluorophore substrate such as ubiquitin-​7-​amino-​4-​methylcoumarin (Ub-​AMC) or ubiquitin-​ rhodamine (Ub-​Rho) enables the quantification of in vitro DUB enzymatic activity because active DUBs can cleave the fluorophore from the ubiquitin C-​terminus. This substrate cleavage in the presence of active enzyme causes unquenching of the fluorophore, leading to a robust increase in fluorescent signal which is ideal for screening. However, it is important to validate hits from a fluorescence-​based experiment in an orthogonal assay, such as using activity-​based probes (ABPs). ABPs, including ubiquitin-​vinylmethyl sulfone (Ub-​VS),50 ubiquitin-​vinylmethyl ester (Ub-​VME)51 and ubiquitin-​propargylic acid (Ub-​PA),52 are comprised of a C-​terminal electrophilic warhead. These probes form an irreversible covalent bond to the nucleophilic cysteine of DUBs and contain sufficient molecular weight to cause a shift on SDS–​PAGE, enabling visualization of DUB labeling by western blot. By comparing DUB labeling in the presence and absence of inhibitor, these probes make useful tools for confirming target engagement. Unlike the fluorescence-​based probes, which are only amenable to in-​vitro biochemical assays, these ABPs can be used in cellular lysates, enabling validation of target binding in more complex and physiologically relevant contexts, an important step in inhibitor characterization.

244

244

Chapter 12

In addition to validating that a small molecule binds or inhibits the DUB of interest in multiple assays, it is essential to profile the inhibitor against a panel of DUBs in order to assess selectivity. In-​vitro DUB panels have become more comprehensive and readily available, enabling standardized profiling of selectivity against recombinant DUBs. Additionally, activity-​based protein profiling (ABPP) enables selectivity profiling in lysates by leveraging ABPs modified with an affinity purification tag (e.g., HA tag, FLAG tag, biotin) in conjunction with quantitative mass spectrometry. The first report of DUB ABPP was in 2011 using HA-​Ub-​VME,53 but in subsequent years several groups have reported improved ABPP methods enabling profiling of more than 50 DUBs in cell lysates. The ability to probe DUB inhibition in cellular lysates is particularly important for DUBs which require binding partners or other physiological parameters. The development of ubiquitin-​based probes for fluorescence-​based and cellular-​based assays enable the characterization of inhibitors against the DUB target of interest as well as panels of DUBs, both in vitro and in situ. Incorporation of assays that measure direct binding are also valuable for validating DUB inhibitors against the target of interest. Researchers have used differential scanning fluorimetry (DSF), hydrogen–​deuterium exchange–​mass spectrometry (HDX-​MS), isothermal titration calorimetry (ITC), and surface plasmon resonance (SPR) to characterize DUB inhibitors.17,19

12.4 Examples of Inhibitors Although there are a large number of reported small-​molecule DUB inhibitors, many of these suffer from being non-​specific and highly reactive, as mentioned above. Below are a number of recent compounds which have been more rigorously characterized. These efforts have come out of a combination of research done by academic laboratories and industry working to develop better probes and starting material for future clinical use. Although we will largely limit our discussion to compounds published in the peer-​reviewed literature, a large number of patents describing small-​molecule inhibitors of DUBs have also been approved and a selection are cited for reference.54–​59

12.4.1 USP25/​28 USP28 was shown to be essential for c-​Myc stability in 2007,60 and USP28 inhibition has been proposed to be a way to limit tumor growth in many cancer cell lines.61 However, the high sequence and structural homology between USP28 and USP25, which regulates the stability of several TNF-​receptor associated factors in innate immune signaling, has made selective inhibitors difficult to find.62 In 2017, a research team from AstraZeneca reported a HTS targeting USP28, which produced three structurally similar lead compounds (AZ1, 2 and 4) and an inactive derivative (AZ3).63 AZ1, the most potent lead inhibitor (0.6 µM), was the result of an extensive medicinal chemistry campaign, which included controlling for non-​specific assay interference, testing

 245

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

245

against multiple substrates and biophysical validation by ITC and microscale thermophoresis (Figure 12.4A). The group characterized AZ1, 2 and 3 at 10 µM and against a panel of caspases and cathepsins at 30 µM, and demonstrated that the compounds were selective for USP28 and USP25. In cells, they confirmed that AZ1 (but not AZ3) inhibited DUB ABP binding to both USP25 and USP28 in native lysate. They further confirmed target engagement by demonstrating that AZ1 (but not AZ3) induced degradation of c-​Myc in multiple cell lines. They did not assess the functional impact of USP25 inhibition, or utilize any of their compounds as an in-​vivo probe. However, this scaffold remains a good candidate for cell biology investigation from the selectivity profiling performed, and the presence of an inactive analogue.

12.4.2 CSN5 The cullin-​RING ligases (CRLs) are modular E3 complexes that consist of a central cullin, an E2-​recruiting RING finger protein and a substrate receptor, and require conjugation of the ubiquitin-​like protein Nedd8 to the cullin subunit of the CRL for full activation.64 This Nedd8 conjugation is achieved in an analogous manner to ubiquitination, with an E1 analogue Nedd8-​activating enzyme (NAE1) responsible for Nedd8 conjugation. Interestingly, a single deNeddylase, the COP9 signalosome (CSN), appears to be necessary and sufficient for cleavage of Nedd8 from all cullins in humans.65 The catalytic subunit of this complex, CSN5, is a JAMM family member, and while not a DUB per se, the principles underlying CSN5 inhibition may have important implications for the JAMM family as a whole. From a therapeutic perspective, CSN5 has emerged as an intriguing target in oncology.

Figure 12.4 (A) Structure of the USP25/​28 inhibitor AZ1. (B) Structure of the CSN5 inhibitor CSN5i-​ 3. (C) Structure of the USP30 inhibitor capzimin. (D) Structure of the USP7 inhibitor GNE-​6776. (E) Development of the 4-​ hydroxypiperadine scaffold as USP7 small-​molecule inhibitors.

246

246

Chapter 12

Developing and inhibitor against CSN5 posed a biochemical challenge, because CSN5 is inactive as a monomer and does not cleave short extensions from Nedd8 such as Nedd8-​AMC. Therefore, researchers at Novartis developed a complex HTS screening program which assayed the activity of the CSN complex against a fluorescence-​labeled CRL substrate via time-​resolved fluorescence resonance energy transfer (TR-​FRET).66 This assay produced two hit scaffolds that successfully co-​crystallized with the CSN5 catalytic domain, allowing optimization to CSN5i-​3 (5.8 nM) and CSN5i-​6 (60 nM), respectively (Figure 12.4B).26,27 These compounds were assessed for selectivity, although CSN5i-​6 was only tested against a panel of matrix metalloproteases (MMPs), while CSN5i-​3 was also tested against the DUBs AMSH-​LP and Rpn11. CSN5i-​ 6 was rigorously characterized as a CSN5 binder (DSF, SPR, NMR and crystallography), while the biophysical characterization of CSN5i-​3 was limited to crystallography. However, CSN5i-​3 was characterized far more extensively than CSN5i-​6 in cell line and mouse model experiments. CSN5i-​3 was tested in cyto against a panel of cullins and substrate receptors, in a panel of 500 cancer cell lines (against an inactive analogue) for viability effects, and in vivo for pharmacokinetics (PK) and pharmacodynamics (PD), tumor growth and biomarkers in a xenograft model. While the authors did not characterize target engagement beyond biomarker validation, the work indicates CSN5i-​3 to be a potent and bioavailable chemical probe that may be utilized for additional characterization of CSN5 function.

12.4.3 Rpn11 Rpn11 (also known as PSMD14) is a member of the JAMM zinc metalloprotease family of DUBs and a catalytic subunit of the 19S regulatory particle of the proteasome.67 Prior to proteasomal degradation, Rpn11 cleaves ubiquitin from substrates to be recycled. Loss of Rpn11 impairs substrate degradation because of the remaining bulky ubiquitin moiety, and has potential to cause preferential apoptosis in neoplastic cells due to the increased dependence on the proteasome, similar to approved proteasome inhibitors like bortezomib.68 In 2007, Seth Cohen and Ray Deshaies reported capzimin as the first inhibitor of Rpn11.18 They tested two libraries: a library of ~250 metal-​binding pharmacophores and a 330 000 compound diversity set. The HTS resulted in the identification of quinoline-​8-​thiol (8TQ) as a hit compound with an IC50 of 2.4 µM. Kinetic experiments indicated an uncompetitive mode of inhibition and the IC50 was right-​shifted in the presence of a chelating agent, consistent with the prediction that 8TQ coordinates the catalytic Zn2+ ion in Rpn11. SAR efforts yielded capzimin, which inhibited Rpn11 (IC50 = 0.34 µM) with good to excellent selectivity over other zinc metalloprotease DUBs CSN5 (IC50 = 30 µM), AMSH (IC50 = 4.5 µM) and BRCC36 (IC50 = 2.3 µM) and high selectivity over a small panel of non-​DUB metalloenzymes (Figure 12.4C). Treatment of HeLa cells stably expressing the proteasome reporter UbG76V-​GFP with capzimin, or the proteasome inhibitor MG132 as a positive control, elicited strong accumulation of UbG76V-​GFP. Furthermore, exposure to capzimin

 247

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

247

caused build-​ up of high-​ molecular-​ weight ubiquitin conjugates and the well-​characterized proteasome substrates Nrf2, p53 and Hif1α. Moreover, treatment of A549 cells with capzimin induced formation of aggresomes, perinuclear storage depots for misfolded and aggregated proteins, phenocopying proteasome inhibitors. Consistent with disruption of protein homeostasis, capzimin also induced an unfolded protein response. Unsurprisingly, Rpn11 inhibition by capzimin caused cell death in a wide variety of cancer cell lines with the most sensitive lines exhibiting GI50 values in the high nanometer to low micrometer range. Encouragingly, the compound was equipotent against a set of bortezomib-​sensitive and -​resistant cell lines.

12.4.4 USP7 In the past few years, several groups have developed programs to identify USP7 inhibitors. In 2017, researchers at Genentech employed an extensive pipeline to find USP7 inhibitors using HTS and NMR fragment screening, and counter-​ screening against USP5 and USP47. They reported GNE-​6776 as a lead USP7 inhibitor (1.34 µM) and GNE-​2118 as an inactive analogue (Figure 12.4D). Similar to the other current generation of inhibitors, GNE-​6776 was characterized via selectivity assays and extensive biophysical assays. The co-​crystal structure of GNE-​6776 with USP7 revealed that it binds to a pocket 12 Å away from the catalytic triad at the interface of the USP7 palm, finger and thumb subdomains (for detailed discussion of USP domain structure, see Ref. 5). The GNE-​6776 scaffold was assessed by ABP displacement and by MDM2, p53 and p21 western blots in WT, USP7–​/​–​ and TP53–​/​–​ HCT116 cells, and was assessed in vivo. Ultimately, the scaffold was extensively characterized, but its relatively modest potency remains a drawback for future studies as an optimal USP7 chemical probe. Further work with this scaffold, especially in relation to the XL188/​ALM5/​FT671 scaffolds, could provide valuable insights into more nuanced USP7 biology. Independent of these efforts, in 2013, Hybrigenics filed a patent for a series of 4-​hydroxypiperidines as USP7 inhibitors. This scaffold has high homology to three inhibitors published by independent groups in 2017 and 2018: XL188 (90 nM),17 ALM5 (1.5 nM)69 and FT671 (52 nM) (Figure 12.4E).19 Interestingly, these three compounds were identified via different approaches (analoguing known hits from the patent literature, fragment based screening, and HTS, respectively), but converged on similar scaffolds that bind the same pocket on USP7. In each case, optimization through medicinal chemistry efforts led to potent inhibitors that were co-​crystallized with the USP7 catalytic core. The structures revealed that these compounds bound in the S5 pocket of the catalytic domain, where they sterically competed with ubiquitin binding, and also bind the enzyme in a conformation where the catalytic triad is misaligned, and therefore is not catalytically competent (Figure 12.5). Notably, this is a distinct pocket from GNE-​6776, indicating that selective compounds for USP7 may be achieved at multiple sites. Each scaffold was further characterized for its USP7 binding properties through a wide variety of biochemical and biophysical assays including

248

248

Chapter 12

Figure 12.5 Comparison of USP7 bound to ubiquitin in (on left, PDB: 1NBF) with USP7 bound to the inhibitor ALM2 (right, PDB: 5N9R). Ubiquitin in shown as a wheat surface, ubiquitin as a gray surface, and ALM2 as magenta sticks. Asterisks indicates the location of the catalytic triad.

SPR, ITC, mutagenic studies, HDX-​MS and SPR. The compounds were also assayed for selectivity which showed excellent preference for USP7 inhibtion, and target engagement was confirmed by displacement of a DUB ABP. The compounds showed stabilization of p53 and downstream targets in cellulo, although inactive analogues were only reported for XL188 and ALM5, not FT671. Inactive control analogues were used for XL188 and ALM5, and FT671-​resistant USP7 was used to show that the increase in p53 was dependent on the USP7 inhibitory activity of the compound. Moreover, these scaffolds were assessed for preliminary ADME properties, and FT671 was employed in vivo in a xenograft study. Thus, all three of these chemical series represent high-​quality chemical probes for cellular, and perhaps in vivo studies.

12.5 Outlook and Future Directions DUBs have only relatively recently emerged as a potential therapeutic class of enzymes. Their ability to modulate traditionally intractable targets and diverse cellular and physiological pathways that they affect has raised excitement to target these enzymes for therapeutic purposes. However, the relative dearth of validated substrates and the only recent identification of potent and selective small-​molecule inhibitors of DUBs characterize a field in its early days. The promise and potential of the field are sure to be increasingly studied in the future as more is learned about the biological function, clinical relevance, and high-​quality probe compounds are discovered. Despite encouraging work in the field, a number of challenges remain. It is important to underscore that, as with any class of small molecules, when selecting probes, it is important to consider the available target space data for a tool compound. Deubiquitinating enzymes are an emerging class of targets, so scientists are actively developing and refining small-​molecule inhibitors for various DUBs of interest, and as the field is relatively young, there are a number of unoptimized compounds with a high degree of polypharmacology. Such compounds have limited utility in pinning function or therapeutic relevance onto a particular DUB target. As researchers continue to publish more

 249

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

249

potent and selective DUB inhibitors, the questions around basic DUB biology and therapeutic potential can be addressed. DUB inhibitors are particularly attractive because a small-​molecule DUB inhibitor has the potential to induce degradation of traditionally “undruggable” targets such as MYC or KRAS, which are quite challenging to target directly with small molecules. So, excitingly, DUBs offer the ability to expand the druggable proteome to include many proteins of interest, such as oncogenic transcription factors or protein aggregates implicated in neurodegenerative diseases. A large number of DUBs have not been very thoroughly characterized so the development of high-​quality DUB inhibitors and the use of CRISPR to modulate DUBs will enable researchers to determine the DUB that regulates “undruggable” proteins of interest. Another exciting possibility is the ability to develop DUB inhibitors that preferentially degrade mutant proteins while sparing the wild-​type version. For example, USP10 inhibitors induce the degradation of mutant FLT3 but not wild-​type FLT3.70 Mutations may render proteins less stable and more easily degraded, or mutations may impact the interactions between DUBs and substrates. Regardless, it is very promising that a DUB inhibitor could be used to preferentially degrade a mutant oncogenic driver within a tumor while sparing the wild-​type protein present in somatic tissues. This observation from the USP10 and FLT3 story highlights the importance of conducting phenotypic screens in the particular cellular context of interest in order to identify the DUBs that regulate the mutant forms of proteins of interest. There are some possible pitfalls in the modulation of DUBs for therapeutic benefit which are important to keep in mind. It is possible that the presence of functional redundancy or substrate promiscuity of DUBs may allow other DUBs to compensate for one another, reducing the efficacy of targeting a single DUB. The development of small-​molecule inhibitors against different DUBs will help address these basic questions around functional redundancy because inhibitors have several advantages over genetic tools, including their ease of use across many disease models and their acute temporal response. The key examples we have highlighted in this chapter illustrate some of the DUBs that have shown the most clinical promise, and continued research efforts should continue to elucidate potential DUB targets of interest. The combined use of small molecules and genetic tools will enable researchers to continue to investigate the potential of this exciting, emerging class of therapeutic targets.

References 1. O. Kerscher, R. Felberbaum and M. Hochstrasser, Annu. Rev. Cell Dev. Biol., 2006, 22, 159–​180. 2. D. Komander and M. Rape, Annu. Rev. Biochem., 2012, 81, 203–​229. 3. M. Akutsu, I. Dikic and A. Bremm, J. Cell Sci., 2016, 129, 875–​880. 4. D. Kwasna, S. A. Abdul Rehman, J. Natarajan, S. Matthews, R. Madden, V. De Cesare, S. Weidlich, S. Virdee, I. Ahel, I. Gibbs-​Seymour and Y. Kulathu, Mol. Cell, 2018, 70, 150–​164.

250

250

Chapter 12

5. T. E. T. Mevissen and D. Komander, Annu. Rev. Biochem., 2017, 86, 159–​192. 6. G. Kleiger and T. Mayor, Trends Cell Biol., 2014, 24, 352–​329. 7. A. Ciechanover, Nat. Rev. Mol. Cell Biol., 2005, 6,  79–​87. 8. M. J. Lee, B. H. Lee, J. Hanna, R. W. King and D. Finley, Mol. Cell. Proteomics, 2011, 10, R110 003871. 9. M. H. Wright, I. Berlin and P. D. Nash, Cell Biochem. Biophys., 2011, 60,  39–​46. 10. D. Popovic, D. Vucic and I. Dikic, Nat. Med., 2014, 20, 1242–​1253. 11. G. Atkin and H. Paulson, Front. Mol. Neurosci., 2014, 7, 63. 12. D. Shi and S. R. Grossman, Cancer Biol. Ther., 2010, 10, 737–​747. 13. Y. Park, H. S. Jin, D. Aki, J. Lee and Y. C. Liu, Adv. Immunol., 2014, 124,  17–​66. 14. E. E. Manasanch and R. Z. Orlowski, Nat. Rev. Clin. Oncol., 2017, 14, 417–​433. 15. C. V. Dang, E. P. Reddy, K. M. Shokat and L. Soucek, Nat. Rev. Cancer, 2017, 17, 502–​508. 16. B. Mansoori, A. Mohammadi, S. Davudian, S. Shirjang and B. Baradaran, Adv. Pharm. Bull., 2017, 7, 339–​348. 17. I. Lamberto, X. Liu, H. S. Seo, N. J. Schauer, R. E. Iacob, W. Hu, D. Das, T. Mikhailova, E. L. Weisberg, J. R. Engen, K. C. Anderson, D. Chauhan, S. Dhe-​Paganon and S. J. Buhrlage, Cell Chem. Biol., 2017, 24, 1490–​1500. 18. J. Li, T. Yakushi, F. Parlati, A. L. Mackinnon, C. Perez, Y. Ma, K. P. Carter, S. Colayco, G. Magnuson, B. Brown, K. Nguyen, S. Vasile, E. Suyama, L. H. Smith, E. Sergienko, A. B. Pinkerton, T. D. Y. Chung, A. E. Palmer, I. Pass, S. Hess, S. M. Cohen and R. J. Deshaies, Nat. Chem. Biol., 2017, 13, 486–​493. 19. A. P. Turnbull, S. Ioannidis, W. W. Krajewski, A. Pinto-​Fernandez, C. Heride, A. C. L. Martin, L. M. Tonkin, E. C. Townsend, S. M. Buker, D. R. Lancia, J. A. Caravella, A. V. Toms, T. M. Charlton, J. Lahdenranta, E. Wilker, B. C. Follows, N. J. Evans, L. Stead, C. Alli, V. V. Zarayskiy, A. C. Talbot, A. J. Buckmelter, M. Wang, C. L. McKinnon, F. Saab, J. F. McGouran, H. Century, M. Gersch, M. S. Pittman, C. G. Marshall, T. M. Raynham, M. Simcox, L. M. D. Stewart, S. B. McLoughlin, J. A. Escobedo, K. W. Bair, C. J. Dinsmore, T. R. Hammonds, S. Kim, S. Urbe, M. J. Clague, B. M. Kessler and D. Komander, Nature, 2017, 550, 481–​486. 20. Y. Wang, Y. Jiang, S. Ding, J. Li, N. Song, Y. Ren, D. Hong, C. Wu, B. Li, F. Wang, W. He, J. Wang and Z. Mei, Cell Res., 2018, 28, 1186–​1194. 21. C. Ndubaku and V. Tsui, J. Med. Chem., 2015, 58, 1581–​1595. 22. J. A. Ronau, J. F. Beckmann and M. Hochstrasser, Cell Res., 2016, 26, 441–​456. 23. S. Bhattacharya, D. Chakraborty, M. Basu and M. K. Ghosh, Signal Transduct. Target. Ther., 2018, 3, 17. 24. M. Hu, L. Gu, M. Li, P. D. Jeffrey, W. Gu and Y. Shi, PLoS Biol., 2006, 4, e27. 25. L. Kategaya, P. Di Lello, L. Rouge, R. Pastor, K. R. Clark, J. Drummond, T. Kleinheinz, E. Lin, J. P. Upton, S. Prakash, J. Heideker, M. McCleland, M.

 251

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

251

S. Ritorto, D. R. Alessi, M. Trost, T. W. Bainbridge, M. C. M. Kwok, T. P. Ma, Z. Stiffler, B. Brasher, Y. Tang, P. Jaishankar, B. R. Hearn, A. R. Renslo, M. R. Arkin, F. Cohen, K. Yu, F. Peale, F. Gnad, M. T. Chang, C. Klijn, E. Blackwood, S. E. Martin, W. F. Forrest, J. A. Ernst, C. Ndubaku, X. Wang, M. H. Beresini, V. Tsui, C. Schwerdtfeger, R. A. Blake, J. Murray, T. Maurer and I. E. Wertz, Nature, 2017, 550, 534–​538. 26. J. Zawacka-​Pankau and G. Selivanova, J. Intern. Med., 2015, 277, 248–​259. 27. A. Gupta, K. Shah, M. J. Oza and T. Behl, Biomed. Pharmacother., 2019, 109, 484–​492. 28. G. D. Bowman and M. G. Poirier, Chem. Rev., 2015, 115, 2274–​2295. 29. T. Jenuwein and C. D. Allis, Science, 2001, 293, 1074–​1080. 30. W. A. Krajewski, J. Li and Y. Dou, Nucleic Acids Res., 2018, 46, 7631–​7642. 31. G. Fuchs and M. Oren, Biochim. Biophys. Acta, 2014, 1839, 694–​701. 32. K. W. Henry, A. Wyce, W. S. Lo, L. J. Duggan, N. C. Emre, C. F. Kao, L. Pillus, A. Shilatifard, M. A. Osley and S. L. Berger, Genes Dev., 2003, 17, 2648–​2663. 33. L. Wang and S. Y. Dent, Epigenomics, 2014, 6, 329–​339. 34. J. M. Espinosa, Genes Dev., 2008, 22, 2743–​2749. 35. B. K. Fiil, R. B. Damgaard, S. A. Wagner, K. Keusekotten, M. Fritsch, S. Bekker-​Jensen, N. Mailand, C. Choudhary, D. Komander and M. Gyrd-​ Hansen, Mol. Cell, 2013, 50, 818–​830. 36. K. Keusekotten, P. R. Elliott, L. Glockner, B. K. Fiil, R. B. Damgaard, Y. Kulathu, T. Wauer, M. K. Hospenthal, M. Gyrd-​Hansen, D. Krappmann, K. Hofmann and D. Komander, Cell, 2013, 153, 1312–​1326. 37. F. Tokunaga and K. Iwai, Microbes Infect., 2012, 14, 563–​572. 38. R. B. Damgaard, J. A. Walker, P. Marco-​Casanova, N. V. Morgan, H. L. Titheradge, P. R. Elliott, D. McHale, E. R. Maher, A. N. J. McKenzie and D. Komander, Cell, 2016, 166, 1215–​1230. 39. K. Heger, K. E. Wickliffe, A. Ndoja, J. Zhang, A. Murthy, D. L. Dugger, A. Maltzman, E. M. F. de Sousa, J. Hung, Y. Zeng, E. Verschueren, D. S. Kirkpatrick, D. Vucic, W. P. Lee, M. Roose-​Girma, R. J. Newman, S. Warming, Y. C. Hsiao, L. G. Komuves, J. D. Webster, K. Newton and V. M. Dixit, Nature, 2018, 559, 120–​124. 40. J. S. Park, R. L. Davis and C. M. Sue, Curr. Neurol. Neurosci. Rep., 2018, 18, 21. 41. B. Bingol and M. Sheng, Free Radic. Biol. Med., 2016, 100, 210–​222. 42. S. Miller and M. M. K. Muqit, Neurosci. Lett., 2019, 705,  7–​13. 43. B. Bingol, J. S. Tea, L. Phu, M. Reichelt, C. E. Bakalarski, Q. Song, O. Foreman, D. S. Kirkpatrick and M. Sheng, Nature, 2014, 510, 370–​375. 44. P. S. de Oliveira, F. A. Ferraz, D. A. Pena, D. T. Pramio, F. A. Morais and D. Schechtman, Sci. Signal., 2016, 9, re3. 45. M. Iconomou and D. N. Saunders, Biochem. J., 2016, 473, 4083–​4101. 46. M. T. Morgan, M. Haj-​Yahya, A. E. Ringel, P. Bandi, A. Brik and C. Wolberger, Science, 2016, 351, 725–​728. 47. N. Nakamura and S. Hirose, Mol. Biol. Cell, 2008, 19, 1903–​1911. 48. M. E. Morrow, M. T. Morgan, M. Clerici, K. Growkova, M. Yan, D. Komander, T. K. Sixma, M. Simicek and C. Wolberger, EMBO Rep., 2018, 19, e45680.

252

252

Chapter 12

49. B. Nabet, J. M. Roberts, D. L. Buckley, J. Paulk, S. Dastjerdi, A. Yang, A. L. Leggett, M. A. Erb, M. A. Lawlor, A. Souza, T. G. Scott, S. Vittori, J. A. Perry, J. Qi, G. E. Winter, K. K. Wong, N. S. Gray and J. E. Bradner, Nat. Chem. Biol., 2018, 14, 431–​441. 50. A. Borodovsky, B. M. Kessler, R. Casagrande, H. S. Overkleeft, K. D. Wilkinson and H. L. Ploegh, EMBO J., 2001, 20, 5187–​5196. 51. A. Borodovsky, H. Ovaa, N. Kolli, T. Gan-​Erdene, K. D. Wilkinson, H. L. Ploegh and B. M. Kessler, Chem. Biol., 2002, 9, 1149–​1159. 52. S. Sommer, N. D. Weikart, U. Linne and H. D. Mootz, Bioorg. Med. Chem., 2013, 21, 2511–​2517. 53. M. Altun, H. B. Kramer, L. I. Willems, J. L. McDermott, C. A. Leach, S. J. Goldenberg, K. G. Kumar, R. Konietzny, R. Fischer, E. Kogan, M. M. Mackeen, J. McGouran, S. V. Khoronenkova, J. L. Parsons, G. L. Dianov, B. Nicholson and B. M. Kessler, Chem. Biol., 2011, 18, 1401–​1412. 54. P. Farshi, R. R. Deshmukh, J. O. Nwankwo, R. T. Arkwright, B. Cvek, J. Liu and Q. P. Dou, Expert Opin. Ther. Pat., 2015, 25, 1191–​1208. 55. D. J. Maloney, A. S. Rosenthal, A. Jadhav, T. S. Dexheimer, A. Simeonov, Z. Zhuang, Q. Liang and D. K. Luci, World Pat., WO2014105952A2, 2017. 56. A. J. Buckmelter, S. Ioannidis, B. Follows, G. Gustafson, M. Wang, J. A. Caravella, Z. Wang, E. L. Fritzen and J. Lin, World Pat., WO2017087837A1, 2018. 57. S. Ioannidis, A. C. Talbot, B. Follows, A. J. Buckmelter, M. Wang, A.-​M. Campbell, D. R. Schmidt, D. J. Guerin, J. A. Caravella and R. B. Diebold, World Pat., WO2016109515A1, 2018. 58. A. Jones, M. I. Kemp, M. L. Stockley, K. R. Gibson, G. A. Whitlock and A. Madin, World Pat., WO2017158388A1, 2018. 59. K. R. Gibson, A. Jones, M. I. Kemp, A. Madin, M. L. Stockley, G. A. Whitlock, M. D Woodrow and K. A. Menear, World Pat., WO2017158388A1, 2019. 60. N. Popov, M. Wanzel, M. Madiredjo, D. Zhang, R. Beijersbergen, R. Bernards, R. Moll, S. J. Elledge and M. Eilers, Nat. Cell Biol., 2007, 9, 765–​774. 61. X. Wang, Z. Liu, L. Zhang, Z. Yang, X. Chen, J. Luo, Z. Zhou, X. Mei, X. Yu, Z. Shao, Y. Feng, S. Fu, Z. Zhang, D. Wei, L. Jia, J. Ma and X. Guo, Cell Death Dis., 2018, 9, 186. 62. B. Zhong, X. Liu, X. Wang, X. Liu, H. Li, B. G. Darnay, X. Lin, S. C. Sun and C. Dong, Sci. Signal., 2013, 6, ra35. 63. J. D. Wrigley, G. Gavory, I. Simpson, M. Preston, H. Plant, J. Bradley, A. U. Goeppert, E. Rozycka, G. Davies, J. Walsh, A. Valentine, K. McClelland, K. E. Odrzywol, J. Renshaw, J. Boros, J. Tart, L. Leach, T. Nowak, R. A. Ward, T. Harrison and D. M. Andrews, ACS Chem. Biol., 2017, 12, 3113–​3125. 64. R. J. Deshaies, E. D. Emberley and A. Saha, Subcell. Biochem., 2010, 54,  41–​56. 65. N. Wei, G. Serino and X. W. Deng, Trends Biochem. Sci., 2008, 33, 592–​600. 66. A. Schlierf, E. Altmann, J. Quancard, A. B. Jefferson, R. Assenberg, M. Renatus, M. Jones, U. Hassiepen, M. Schaefer, M. Kiffe, A. Weiss, C.

 253

Deubiquitinase Inhibitors: An Emerging Therapeutic Class

67. 68. 69. 70.

253

Wiesmann, R. Sedrani, J. Eder and B. Martoglio, Nat. Commun., 2016, 7, 13166. R. Verma, L. Aravind, R. Oania, W. H. McDonald, J. R. Yates, 3rd, E. V. Koonin and R. J. Deshaies, Science, 2002, 298, 611–​615. M. Gallery, J. L. Blank, Y. Lin, J. A. Gutierrez, J. C. Pulido, D. Rappoli, S. Badola, M. Rolfe and K. J. Macbeth, Mol. Cancer Ther., 2007, 6, 262–​268. G. Gavory, C. R. O’Dowd, M. D. Helm, J. Flasz, E. Arkoudis, A. Dossang, C. Hughes, E. Cassidy, K. McClelland, E. Odrzywol, N. Page, O. Barker, H. Miel and T. Harrison, Nat. Chem. Biol., 2018, 14, 118–​125. E. L. Weisberg, N. J. Schauer, J. Yang, L. I., L. Doherty, S. Bhatt, A. Nonami, C. Meng, A. Letai, R. Wright, H. Tiv, P. C. Gokhale, M. S. Ritorto, V. De Cesare, M. Trost, A. Christodoulou, A. Christie, D. M. Weinstock, S. Adamia, R. Stone, D. Chauhan, K. C. Anderson, H. Seo, S. Dhe-Paganon, M. Sattler, N. S. Gray, J. D. Griffin and S. J. Buhrlage, Nat. Chem. Biol., 2017, 13, 1207–​1215.

254

CHAPTER 13

Targeting Translation Regulation for the Development of Novel Drugs IRIS ALROY,a* WISSAM MANSOURb AND YONI SHEINBERGERb a

Anima Biotech, Inc., 75 Claremont Rd, Ste 102, Bernardsville, NJ 07924, USA Anima Biotech, Ltd., 10 Hanechoshet St., Tel-​Aviv, Israel 6971072 *Corresponding author. Email: [email protected] b

13.1 Introduction Protein synthesis is highly conserved across evolution, and all proteins, of all shapes and functions, are made by the same machinery, the ribosome. The central dogma that a gene is transcribed to one mRNA molecule which is then translated to one protein is no longer true as protein to mRNA ratio is greater or smaller than one for many proteins.1 Based on the diversity of cell types and tissues it was expected that the human genome will encode a much larger number of genes than finally identified in the Human Genome Project. Although alternative splicing gives an additional layer of complexity, it does not suffice to explain the complexity and the coordinated temporal and spatial cellular protein expression. mRNA transcription control was studied extensively and shown to be a complex and highly coordinated process, while neither the importance nor the mechanisms of translational regulation have been studied to the same extent. Indeed, recent studies have revealed

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

254

 255

Targeting Translation Regulation for the Development of Novel Drugs

255

that both the global protein synthesis rates and translational efficiencies of specific mRNAs are regulated in a cell-​and tissue-​specific manner.2–​5 Protein synthesis is shifting from seeing the process as a linear, mechanical translation of codons to viewing translation as an intricate regulatory system that controls where, when, how much and which proteins are synthesized in the cell, impacting gene expression no less than regulation of transcription. Concomitantly, regulatory elements of protein synthesis are emerging as valid novel drug targets, their prevalent house-​keeping roles notwithstanding. This convergence reveals new opportunities for drug discovery in areas as varied as cancer, viral infections, fibrosis and neurodevelopmental diseases. The target space for drug discovery in translation regulation spans diverse targets which act from mRNA transcription where proteins regulating mRNA fate bind to the newly synthesized mRNA emerging from RNA polymerase II (Figure 13.1). Regulation continues with enzymes which modify nucleic acids, proteins which bind to exon–​exon sites and proteins which induce secondary and tertiary RNA structures. RNA-​binding proteins also play a role in mRNA maturation, transport to the cytoplasm, localization of mRNA to specific cytoplasmic locations, and in determination of mRNA half-​life. Moreover, RNA binding proteins determine the temporal and spatial regulation of mRNA translation, by inhibiting or recruiting translation initiation complexes. An additional layer of regulation is given by enzymes which modify the activity of RNA binding proteins, and proteins which modify ribosome assembly during translation initiation. Modification of ribosomal proteins and their associated proteins influences ribosome processivity and rate of translation. The rate of translation is also governed by abundance of tRNAs which decode mRNA, whereby rare tRNAs cause ribosome pausing, slowing translation of a specific protein, but also affecting global translation by sequestering ribosome to slowly decoded mRNA.6 tRNAs are the most highly modified RNAs, and tRNA modifications regulate the ability of tRNA aminoacyl transferases to load tRNAs with amino acids and regulate the efficiency at which a tRNA is recruited to ribosomes and its ability to interact with the mRNA codon.

Figure 13.1 Translation regulation target space.

256

256

Chapter 13

Thus, protein translation regulation, from transcription to protein synthesis, offers a plethora of targets which modulation offers a novel and unique space for drug discovery. We have developed a technology that monitors endogenous protein translation of all proteins or a specific protein by monitoring the activity of ribosomes. Our technology, Protein Synthesis Monitoring (PSM), enables the identification of novel target space in drug discovery.

13.2 PSM, Discovery of Translation Regulators Using Pairs of Fluorescent tRNAs Protein synthesis is an iterative process of ribosome recruitment (translation initiation), translation elongation and translation termination. Amino acids are carried by tRNAs, specific for each of the codons, to ribosomes; tRNAs enter ribosome through the acceptor site (A-​site), move to the peptide generation site (P-​site), and exit the ribosome through the exit site (E-​site). Amino acids are encoded by more than one codon (except for methionine), and tRNAs which encode different codons for the same amino acid are called tRNA isoacceptors. PSM is a novel technology that monitors the process of protein synthesis, directly from active ribosomes in cells, using fluorescent labeled tRNAs7–​13. In PSM, tRNAs labelled as Förster resonance energy transfer (FRET) pairs, donor and acceptor, are transfected to cells and donor, acceptor and FRET signals are measured by automated microscopy. Capturing signals by microscopy enables signal measurements from single cells and population-​based analyses. A FRET signal is generated between donor-​and acceptor-​labeled tRNAs only when they are found in close contact inside translating ribosomes, while the peptide chain is elongated. The intensity of the FRET signal correlates with the number of ribosomes containing the tRNA FRET pair, providing a real-​time, cell-​based assay for monitoring protein synthesis (Figure 13.2). PSM can monitor overall protein synthesis, using bulk tRNA (all isoacceptor tRNAs labeled as donor and acceptor), or the synthesis of a specific protein, using one or more specifically selected pairs of tRNA enriched in the protein’s sequence. The 46 human tRNA isoacceptors (anticodon-​specific) yield 1080 possible pairs. The tRNA pairs used by human mRNAs are analyzed by computationally converting the transcriptome to “tRNAome,” i.e., tabulating the number of different pairs of tRNAs in each mRNA, and finally determining the frequency and enrichment factor (E-​factor) of each tRNA pair in each mRNA over the transcriptome (Figure 13.3). Subsequently, the best pair(s) that are specifically enriched the target protein(s) are selected. The selection is further fine-​tuned based on RNA sequencing and ribosome profiling of the relevant cells, as sites of ribosome stalling are identified for further fine-​tuning of pair selection (Figure 13.3). RNA sequencing is used to give a better resolution of signal to noise ratio, using the mRNA expression levels of the target protein and the expression level of background proteins which

 257

newgenrtpdf

Targeting Translation Regulation for the Development of Novel Drugs

Figure 13.2 Anima’s translation monitoring technology. (A) tRNAs are labeled with Cy3 or Cy5 fluorophores and transfected into cells. When tRNAs are found in close proximity, inside the ribosomes, the distance and orientation allow for FRET to occur. (B) Translation of collagen 1 in human lung fibroblasts. Isoacceptor-​specific tRNA-​Pro and tRNA-​Gly labeled with Cy3 or Cy5, respectively, were transfected into WI-​38, human lung fibroblasts. Forty-​eight hours post transfection, cells were fixed with 4% paraformaldehyde and nuclei stained with DAPI (gray). Donor (Cy3), acceptor (Cy5), DAPI and FRET (orange) were captured with an automated microscope (Operetta, Perkin Elmer). Images were visualized with Columbus (Perkin Elmer).

257

258

258

Chapter 13

Figure 13.3 Selection criteria for tRNA signature pair to visualize translation of a specific gene.

Figure 13.4 Most human genes contain a signature tRNA pair. The highest E-​ factor for a single signature tRNA pair of each human gene was plotted against all human genes. The bars were sorted from lowest (left) to highest (right) E-​factors. E-​factors higher than 20 are marked on the top of the graph, 83% of human genes have a tRNA-​pair with an E-​factor greater than 20.

use the same pair of tRNA to better define signal to noise ratio. Identification of stalling sites and selection of the codon pair at the site will result in longer resident time of a ribosome on two labeled tRNAs and in a stronger FRET signal (Figure 13.3). tRNA pairs enrichment factors analyses show that more than 83% of the mRNAs have enrichment factors higher than 20 (i.e., 20-​fold enrichment versus the background; Figure 13.4), indicating that PSM technology can monitor most of the human proteome. Moreover, there are signatures that are shared within a protein family, enabling the selection of a pair which monitors the translation of a family of proteins involved in a specific biological pathway. For example, 13 interleukins share a signature pair tRNA, with a combined enrichment factor of 248 (Figure 13.5); this pair can be used to visualize an interleukin response in target cells. All isoacceptor tRNAs (named bulk tRNA) or oligo-​affinity-​purified isoacceptor tRNAs are labeled with Cy3 and Cy5 fluorophores14 and introduced into cells, in a 384-​well format, with liposome-​mediated transfection. Twenty-​four

 259

Targeting Translation Regulation for the Development of Novel Drugs

259

Figure 13.5 A signature tRNA pair for a group of proteins with a similar function. The E-​factor for the pair of tRNAs, isoacceptor-​specific tRNA-​Arg and tRNA-​ Leu, is plotted against the gene name. The combined E-​factor for the 13 interleukins is 248.

to 40 hours post transfection, compounds are added (2–​24 hours incubation), and cells are fixed at 48 hours post transfection. Images are captured with an automated wide-​field fluorescent microscope (Operetta, Perkin Elmer) using four different channels –​donor channel (Cy3), acceptor channel (Cy5), FRET channel, and DAPI channel to visualize nuclei, which are used to identify single cells. The whole process is automated from cell seeding, transfection, fixation and image capture, and 100K compounds are screened within 3 weeks (Figure 13.6). Once a plate is imaged, images are uploaded automatically to AWS cloud, and the donor and acceptor bleedthrough-​cleaned FRET image is calculated automatically. Image and data analysis algorithms have been developed at Anima to support high throughput and robust analyses. Data are analyzed on the single cell level; thus, for each compound there are 500–​2000 data points per well (Figure 13.6F). Hits are identified using machine learning and unsupervised learning by scoring for difference in any of the cFRET features extracted from the images (intensity, size, morphology, distribution, etc.). Hits are re-​tested in the primary screening assay to confirm translation modulation activity, and subsequently validated for target-​specific activity by several secondary assays: (1) selection against general translation inhibitors by using bulk labeled tRNA in the cells used for the screen or in another cellular environment and (2) metabolic labeling with fluorescent methionine; (3) ensure that the target protein levels are reduced or increased, by using

260

260

Chapter 13

Figure 13.6 Anima’s screening platform. (A) cells are plated in 384-​well plates and transfected with gene-​specific signature tRNA pairs. (B) Compounds (100–​ 200 000, 30 μM) are added to individual wells for 4–​ 24 hours incubation, cells are fixed and nuclei stained. (C) Plates are imaged automatically with Operetta (Perkin Elmer), and images are uploaded to AWS cloud. During upload, an automated algorithm calculates the clean FRET image. (D) Automated image and data analysis algorithms detect cells and FRET signals and analyze cell-​level data to generate a score for each compound, relative to DMSO treated wells. Each dot in the cloud of dots represents a single cell in a three-​parameter space (FRET parameters generated from the images). The brown dots are DMSO-​treated cells, and the orange dots are cells in a hit-​treated well.

immunofluorescence assay specific for the protein. During assay development, the half-​life of the target protein mRNA and protein levels are determined. In the screen, compound incubation time is selected based on the mRNA or protein half-​life, to ensure we identify compounds which act between transcription to translation, and not upstream to transcription (Figure 13.7). The data analyses are based on change from control, and not on increase or decrease in FRET signal, which results in identification of compounds which increase or decrease translation (Figure 13.8). Moreover, more than one chemotype of compounds are identified, i.e., possibly binding to different protein targets. Thus, running one PSM screen will result in elucidating several molecular pathways which regulate translation. Validated compounds, compounds that are target-​specific and do not affect general translation, are reordered from powder, purity determined (compounds below 80% purity are discarded) and assayed at dose–​response and time course in a target-​specific labelled pair of tRNA assay, target-​specific immunofluorescence and metabolic labeling. Subsequently, cytotoxicity is evaluated in cells used for the screen and in human fibroblasts. Compounds are clustered by structure similarity using Tanimoto scoring,15 and activity is plotted against each series to test if a structure–​activity relationship (SAR) is developing. Additional compounds with similar structures are purchased to enrich the SAR.

 261

Targeting Translation Regulation for the Development of Novel Drugs

261

Figure 13.7 Time course study of inhibitor treated cells. A549 cells, lung cancer epithelial cell line, were plated in 384-​well plates and treated with transcription (actinomycin D, blue), mRNA processing (cordycepin, purple), translation (cycloheximide, yellow) and protein degradation (MG132, black) inhibitors for the indicated time points. At the indicated times, cells were fixed with 4% paraformaldehyde, nuclei were stained and cells processed for c-​Myc protein (A) or c-​Myc mRNA (B) detection, with an anti-​Myc antibody, or Myc-​specific FISH probes, respectively.

Compound specificity and selectivity are tested comprehensively. Compound effects in different cellular environments are assessed. Indeed, compounds discovered during a screen for collagen translation modulators in human lung fibroblasts reduced collagen 1 levels in these cells, but not in human dermal fibroblasts (Figure 13.9). Two of the series were specific to lung fibroblasts, while the third series was active also in liver stelleate cells, liver-​ resident fibroblasts (data not shown). This indicates that translation is regulated in a tissue-​specific manner and that it offers the opportunity to discover tissue-​specific compounds, which may alleviate toxicity. Indeed, the expression of many RNA binding proteins (RBPs), which regulate mRNA processing, stability and translation, is tissue-​specific.16 During hit validation, mode of action studies are initiated to enable target identification. Compound effect on transcription and translation is assessed by fluorescence in situ hybridization (FISH), time course studies of compound activity on mRNA and protein are compared to known inhibitors of transcription (actinomycin D), mRNA maturation (cordycepin), protein translation (cycloheximide) and protein degradation (MG132) (Figure 13.7). RNA from cells treated with active and inactive compounds from each series are sent to mRNA and tRNA sequencing to conduct pathway analysis. In addition, the compound’s effect on mRNA and tRNA modification is checked by sequencing and LC-​MS.17

262

262

Chapter 13

Figure 13.8 Identification of collagen 1 translation modulators in human lung fibroblasts. (A) The collagen 1 signature pair, isoacceptor tRNA-​ Pro and tRNA-​Gly, were transfected to WI38 cells with scrambled siRNA (left panel) or COL1A1-​ specific siRNA (right panel). The calculated FRET signal, corrected for bleedthrough between the donor and acceptor channels, was markedely reduced by the COL1A1-​specific siRNA. (B) Collagen 1 modulators were identified by Anima’s translation modulators screen using a collagen 1 signature isoacceptor-​specific tRNA pair, tRNA-​Gly and tRNA-​Pro. Activation of human lung fibroblasts, WI38, with an activation cocktail, induced collagen 1 translation (upper panel, compare healthy and activated cells, FRET signal, cyan), and collagen 1 protein (lower panel, green). Activation of collagen 1 translation is detected by PSM technology 24 hours post compound treatment (top panel, cyan). However, it takes 5 days until enough collagen 1 protein accumulates and is detectable by an anticollagen antibody (lower panel, green). Selected hits from the screen are collagen 1 inhibitors, as can be detected by a decreased FRET signal and a decrease in collagen 1 protein (upper and lower panels, respectively), or collagen 1 inducers, compounds which increase the FRET signal and protein level.

13.3 Target Space for PSM: From Transcription to Translation Hit compounds in our screens act between transcription and translation, as evident by the mode of action studies described above. Compounds may regulate mRNA transport to cytoplasm, steady-​state level in cytoplasm or act as translation regulators. Targets regulating each of these processes are described in the next sections.

13.3.1 RNA Processing The mRNA life cycle, from transcription to translation, has been well characterized at the mRNA processing levels, but less so at the translation regulation

 263

Targeting Translation Regulation for the Development of Novel Drugs

263

Figure 13.9 Collagen 1 translation modulators regulate collagen 1 translation in a tissue-​specific manner. A representative hit compound, compound X, inhibits collagen 1 protein accumulation in the cell line used for the screen, WI-​38 cells (upper panel) andin primary human lung fibroblasts (middle panel). However, the hit compound does not inhibit collagen 1 translation in human primary skin fibroblasts (lower panel).

level. Only recently, new regulatory proteins, enzymes and pathways in translation regulation were discovered.18–​21 Advances in characterization methods of RNA binding proteins (RBPs), such as RNA immunoprecipitation (RIP) and ribosomal profiling, increase our understanding of how this process is being regulated in different cellular contexts. Cytoplasmic mRNA levels, the population of mRNAs that are available for the translation machinery, are regulated during transcription, mRNA processing, splicing, nuclear export and half-​life in cytoplasm. mRNA processing and modifications affect mRNA steady-​state levels and play an important role in the regulation of translation. Indeed, the regulation of translation and mRNA steady-​state levels are intertwined; mRNA that is not translated may be targeted for degradation. The effect of hit compounds identified in a PSM screen on transcription and mRNA levels is assessed by FISH (Figure 13.10). The transcription site in the nucleus and mRNA transcripts in the cytoplasm are enumerated, and the effect on the cell population is analyzed by image analysis and data algorithms to identify the mode of action of the compound, inhibition of transcription (reduction of transcription sites in the nucleus), mRNA transport or steady-​state levels (reduction of mRNA signal in cytoplasm), or at translation (mRNA signal in cytoplasm is intact, but there is no protein). In addition, mRNA localization in the nucleus is also examined, because altered mRNA localization into nuclear speckles may indicate an alternative splicing perturbation. Once transcribed, the newly synthesized mRNA begins a series of transcription-​coupled processing events, including 5′ cap addition and intron removal (splicing).22,23 Then, the immature mRNA undergoes cleavage and polyadenylation to become the mature form of the transcript. These

264

264

Chapter 13

Figure 13.10 Compounds identified in screens act at the level of translation. Cells were treated with DMSO, actinomycin D and hit compounds from collagen or c-​Myc screens. (A) Collagen protein (upper panel, green) or mRNA (lower panel, red) were detected with a collagen 1-​specific antibody or FISH probes, respectively. While the transcription inhibitor actinomycin D reduces both mRNA and protein in the cytoplasm, and inhibits transcription sites in the nucleus, hit compounds from two different chemical chemotypes do not affect transcription sites in the nucleus or accumulation of mRNA in the cytoplasm. However, collagen 1 protein is absent in the cytoplasm in cells treated with these two compounds. (B) c-​Myc protein (upper panel, green) or mRNA (lower panel, red) were detected with a c-​Myc-​specific antibody or FISH probes, respectively. While the transcription inhibitor actinomycin D reduces both mRNA and protein in the cytoplasm, and inhibits transcription sites in the nucleus, hit compounds from two different chemotypes, series 2 and 13, do not affect transcription sites in the nucleus or accumulation of mRNA in the cytoplasm. However, c-​Myc protein is absent in the cytoplasm in cells treated with these two compounds. A compound from series 8 inhibits transcription sites and c-​Myc mRNA accumulation, indicating that it acts at the transcription level.

 265

Targeting Translation Regulation for the Development of Novel Drugs

265

processes are coupled to nuclear export and if one of these is carried out inaccurately it will cause the retention of damaged mRNA in the nucleus due to regulation at the nuclear pores which prevents defective mRNA from leaving the nucleus and possibly becoming disease-​causing proteins. However, occasionally certain genes escape this regulatory network, which leads to diseases.24 Following the correct processing of pre-​mRNA, the mature mRNA will then leave the nucleus and will make its way to its site of translation, the endoplasmic reticulum (ER), mitochondrial membrane, cytoplasm, etc. During influenza virus infection the non-​structural 1 protein (NS1) mediates viral mRNA export from the nucleus to cytosol. Knocking down NS1 prevents viral mRNA splicing at nuclear speckles and nuclear export.25 The involvement of splicing factors in mRNA export from the nucleus is well-​established and new players in nuclear export have been identified as RNA modifying enzymes. Fragile X mental retardation protein (FMRP) is an RNA-​binding protein known to be causative in Fragile X. It was recently shown that FMRP facilitates nuclear export of its target mRNAs using the crm-​1 pathway. Indeed, when FMRP or METTL14, an m6A writer protein, were deleted by siRNA, FMRP target mRNAs were retained in the nucleus. Additionally, mRNA quality-​control proteins (SRp20, SRSF1), which regulate mRNA export, when overexpressed cause nuclear retention of undamaged mRNA thus leading to disease.26–​28 Nuclear export of specific mRNAs has been described in the literature, which opens the door for selective regulation and another opportunity for therapeutic targets. For example, during heat shock or high salt concentrations, the cellular heat shock response kicks in, which among other regulatory events reduces quality control at the nuclear pore for transport of Heat Shock mRNAs, allowing a much faster response to stress.29 Cleavage at the 3′ end of pre-​mRNA and the addition of a poly A tail are the last step in the maturation process of mRNAs that precedes nuclear export. The discovery of multiple poly A cleavage sites in different genes uncovered another layer of gene expression regulation. The 3′-​UTR is a binding site for regulatory proteins and microRNAs which are involved in almost every aspect of the mRNA life cycle. Cells can undergo global lengthening of 3′-​UTRs, for example, in senescence30; and global shortening of 3′-​UTRs in cancer.31 MicroRNAs and RBPs that stimulate translation are in competition for 3′-​UTR binding sites and whenever the balance is favored for microRNAs, protein synthesis is first inhibited and later the mRNA is degraded. Methods that can detect differences in translation efficiency of mRNA isoforms are still under development; for example, ribosome profiling without using mRNA digestion, thus maintaining full-​length transcripts.32 Alternative polyadenylation (APA) occurs for specific transcripts and is associated with multiple diseases and disorders. In the endocrine congenital adrenal hyperplasia disease, the affected gene, steroidogenic acute regulatory protein (StAR) generates two transcripts which differ in length of their 3′-​UTRs and act differently upon activation of cholesterol

266

266

Chapter 13 33

metabolism. In Huntington’s disease (HD), which is caused by multiple repeats in the huntingtin gene, alternative polyadenylation yields three transcripts of different lengths. The shortest transcript can only be seen in HD patients.34 Altogether, we believe that small changes in existing methods will enable studying APA events in specific transcripts, thus allowing the unravelling of new regulatory factors. Epitranscriptomics, mRNA editing and mRNA modification, once considered to be merely RNA “decoration,” is now known to be prevalent in thousands of transcripts and to have a regulatory role in mRNA. There are several ways in which epitranscriptomics can directly impact the translation process: (1) induction of mRNA structural changes at the 5′-​UTR; (2) mRNA binding by RNA binding proteins; (3) modifications of tRNAs; and (4) changes in mRNA sequence, new start codons, new stop codons, stop codon readthrough and translation recoding. N1-​methyladenosine (m1A) was recently shown to be prevalent and conserved between species such as human, murine cells and yeast.35,36 Recent m1A-​seq data showed that m1A modifications within 5′-​UTRs and around alternative and canonical start codons occur in response to stress cues, such as nutrient deprivation and heat shock, and affect the translation or steady-​ state levels of these mRNAs. Additionally, it is known that m1A modifies the mRNA secondary structure using chemical and enzymatic probing37,38 and the importance of mRNA secondary structure to translation initiation is well established.39 Together with the finding that m1A modified mRNA are translated at higher rates, it was hypothesized that m1A facilitate a secondary structure surrounding start codons that increase the translation effectiveness of those modified transcripts.40 N6-​methyladenosine (m6A) is the most abundant internal mRNA modification41 which is crucial for the binding of heterogeneous nuclear ribonucleoprotein C (HNRNPC), an RNA-​ binding protein involved in alternative splicing, among other things. Under oxidative conditions, 8-​ oxoguanosine (8-​ oxoG) and O6-​methylguanosine 6 (m G) were shown to exist in certain transcripts and impact their translation efficiency. 8-​OxoG was shown to hinder tRNA selection in yeast resulting in ribosome stalling.42 The RNA editing event that creates pseudouridines has many impacts on translation. One very well-​known one is to stop codon readthrough by reducing recognition of release factors by the modified stop codon.43,44

13.3.2 RNA-​binding Proteins Once a gene is transcribed, RNA is never naked, as RNA-​binding proteins interact with RNA to protect it and to induce its processing. RNA interacts with many proteins to form mRNA ribonucleoproteins (mRNPs) complexes, which consist of sequence-​specific RNA-​binding proteins (RBPs). RBPs regulate mRNA modifications, splicing, polyadenylation, translocation, translation and decay. Dysregulation in these complexes is involved in various diseases including neuronal diseases and cancer.45,46

 267

Targeting Translation Regulation for the Development of Novel Drugs

267

Eukaryotes encode a high number of RNA-​binding proteins to ensure correct handling and metabolism of RNA in living cells. A variety of methods were used in recent years to identify RBPs and to date over 1700 have been identified.47,48 Methods to identify RNA-​ binding proteins can be divided into two main approaches: (1) mRNA-​centric, in which the mRNA of interest is isolated with the proteins that bind to it; (2) protein-​centric, in which the protein of interest is immunoprecipitated with the mRNA it interacts with. RBPs recognize target mRNAs in a sequence-​dependent or a structure-​ specific manner through their RNA-​binding domains (RBDs), which usually span 60–​100 amino acids.49,50 Currently, over 40 RBDs were identified as canonical and non-​canonical domains. Among the canonical RBD found in eukaryotes are RNA recognition motif (RRM),51 Zn finger (ZF) domain,52 K-​ homology (KH) domain53 and double-​ stranded RNA binding motif (dsRBM).54 mRNA is a highly labile molecule in cells and there are multiple nucleases which are found in the nucleus and cytoplasm which degrade naked mRNA. Thus, once mRNA is exported to the cytoplasm it has a limited lifetime. The half-​life of mRNA is determined by the length of its poly-​A tale and by RBPs which bind to it and protect it from decay. For example, the expression of c-​ Myc, an oncogene implicated in many human cancers, is highly regulated at transcription and post-​transcription level. c-​Myc mRNA stability was found to be increased by insulin-like growth factor 2 messenger RNA (mRNA)-binding protein 1 (IGF2BP1) in human liver cancer55 and by IGF2BP3 in leukemia, both of which bind the 3′-​UTR of c-​Myc mRNA56. Another example is Hu antigen R (HuR), an RBP that modulates stability and translation of many mRNAs. It was shown to regulate c-​Myc mRNA stability under specific conditions.57 c-​Myc mRNA is negatively regulated by ELAV-​like family member 1–​CELF1, which was found to bind to 3′-​UTR, competing with HuR, thereby reducing c-​Myc translation without affecting its mRNA levels.58 Apurinic/​apyrimidinic endonuclease 1 (APE1) is another c-​Myc mRNA regulator, and it regulates c-​Myc mRNA half-​life in vitro and in vivo.59 We performed a screen to identify c-​Myc translation regulators, identifying compounds which reduce the c-​Myc-​specific FRET signal 4 hours after compound treatment (Figure 13.11A, upper panel). Myc protein levels were reduced by compounds treatment, at 4 and 24 hours post treatment, while mRNA levels were not affected (Figure 13.11A and Figure 13.10B, respectively). mRNA sequence analysis of one of the compounds has detected upregulation of epithelial splicing regulatory protein 1 (ESRP1) mRNA, an RBP which regulates c-​Myc mRNA (Figure 13.11B). ESRP1 binds to c-​Myc 5′-​UTR and regulates c-​Myc and translation of other oncogenes.60 ESRP-​ 1-​depleted cells showed an increase in c-​Myc mRNA level and expression.61 Indeed, upregulation of ESPR-​1 will lead to reduction in translation of  c-​Myc. An RNA-​binding protein is a putative target of a compound identified in Anima’s collagen 1 translation regulation. Compound treatment of human

268

268

Chapter 13

Figure 13.11 A c-​Myc translation regulator. (A) One of the compounds identified in Anima’s c-​ Myc screen acts in a dose-​and time-​ dependent manner. A549 cells were treated with increasing concentrations of the compound, for 4 (upper panel) or 24 (lower panel) hours. c-​Myc protein was detected by an antibody in the nucleus (red). (B) A549 cells were incubated with one of the hit compounds identified in the c-​Myc screen for 1 hour, and mRNA expression was analyzed (RNA-​ seq). A 4.5-​fold increase in the expression of an RNA-​binding protein was detected by one of the compounds (left bar), while no effect was seen by actinomycin D, and a marginal one by another compound. This RBP was shown in the literature to bind to c-​Myc mRNA and inhibit its translation.

lung fibroblasts, WI38 cells, reduced the levels of poly(rC)-​binding protein 1 (PCBP1) (Figure 13.12). PCBP1 and other RNA binding proteins, such as PCBP2, AUF1, hnRNPK and poly(A)-​binding protein (PABP), are constituents of the α-​complex that binds to C-​rich motifs found in 3′-​UTR of target genes, and act to prevent mRNA degradation.62

13.3.3 mRNA Localization Protein translation is localized to specific cellular compartments. For example, coordinated translation is required for protein subunits and protein complexes. mRNA localization in the cells is mediated by RBPs, which bind to motor proteins.63 Many mRNA are targeted to translation on the ER membrane by RBPs.64 Collagen 1 mRNA is translated on the surface of the ER and translation is coordinated with translocation into the ER lumen. La ribonucleoprotein domain family, member 6 (LARP6) binds to collagen 1 mRNA and localizes it to the ER. LARP6 has several functional domains, La motif (LaM) and an RNA recognition motif (RRM) which binds the collagen α1(I) and α2(I)

 269

Targeting Translation Regulation for the Development of Novel Drugs

269

Figure 13.12 A potential RBP target for a collagen 1 regulator. Human lung fibroblasts, WI38, were incubated with one of the hit compounds for 96 hours. Cells were fixed, nuclei stained with DAPI (blue), and incubated with an antibody directed to PCBP1, an RNA-​binding protein involved in the stabilization of COL1A1 mRNA (green).

mRNA and transports both mRNAs from the nucleus to the ER.65 LARP6 also interacts with SEC61 translocon, which translocates nascent chains to the ER, and enables coordinated translation of collagen α1(I) and α2(I) polypeptides to ensure effective folding into the triple helix structure by high local concentration. Interruption of this coordinated translation will result in inhibition of collagen 1 translation and thus in reduction of FRET signal using PSM. Indeed, in our screen for modulators of collagen 1 translation, a compound was identified which reduces LARP6 levels in human lung fibroblasts (Figure 13.12). Defects in mRNA localization are involved in a range of diseases and development disorders.66 In neurons, mRNAs are transported from the cell body to synapses for local translation.67 Zipcode binding protein 1 (ZBP1) binds to the 3′-​UTR of β-​actin mRNA in the nucleus during transcription and blocks its translation initiation while mRNA is transported to nerve endings.68 Once reaching its destination, src protein kinase promotes translation by phosphorylating ZBP1, resulting in its dissociation from β-​actin mRNA and initiation of translation.69

13.3.4 mRNA Translation Beside the canonical translation machinery: eukaryotic initiation factors (eIFs), eukaryotic elongation factors (eEFs) and eukaryotic release factors (eRFs), many RBPs regulate mRNA translation by binding to specific motifs in 5′ and 3′ untranslated regions (UTR) or in the coding region. One of the roles of these RBPs is to recruit other accessory factors or enzymes needed to unwind the mRNA secondary structures. For example, LARP6 recruits RNA helicase A (RHA) to the 5′-​UTR of type 1 collagen mRNA to unwire secondary structures and facilitate its translation.70 In Fragile X syndrome (FXS) the absence of RBP Fragile X Mental Retardation Protein (FMRP) in neurons is linked with the disease.71,72 Beside the FMRP function described in the mRNA

270

270

Chapter 13

processing section, FMRP also regulates diacylglycerol kinase kappa (Dgkκ) mRNA in neurons. Dgkκ is an enzyme that phosphorylates diacylglycerol and converts it to phosphatidic acid.73 In the absence of FMRP, Dgkκ is reduced in neurons and causes dendritic spine abnormalities, synaptic plasticity alterations and behavior disorders. Another example for regulation at the level of translation is preproinsulin (Ins2), translation of which is regulated by human antigen D (HuD) RNP in pancreatic β cells. Upon glucose stimulation, HuD is phosphorylated and released from the 5′-​UTR of Ins2 to enables its translation.74 Glucose also induces stabilization of the Ins2 mRNA by promoting the binding of tract-​ binding protein (PTB), also known as hnRNP1, with its four RRMs to the polypyrimidine-​rich sequence on the 3′-​UTR after phosphorylation.75

13.3.5 tRNA Modifications, Expression and Aminoacylation Translational regulation occurs at all translation steps, initiation, elongation and termination. Elongation is the step at which mRNA decoding by transfer RNAs (tRNA) can play a key regulatory role in translation. tRNAs act as a shuttle of amino acids to ribosomes, decoding the mRNA codon. tRNA conjugation with amino acids is done by enzymes named aminoacyl-​tRNA synthetases (AARSs).76,77 AARSs are a family of essential enzyme classified into two subclasses which differ by their structure and kinetics. For each amino acid there is a unique synthetase and mammalian synthetases are found as free enzymes or in a large tRNA multi-​synthetase complex (MSC). AARSs discriminate specific tRNA mainly via the anticodon loop, except AlaRS, LeuRS and SerRS, and ligate the amino acid to its CCA-​3′ tail in an ATP-​dependent reaction.78,79 Beside the canonical function of the AARSs in translation, binding and charging the tRNA, other non-​canonical functions are studied, including binding to mRNAs and regulating translation.80 The yeast aminoacyltransferase for histidine, HisRS, binds its own mRNA and regulates its translation according to coding demands for histidine as increasing cellular levels of uncharged His tRNA lead to upregulation in the HisRS translation.81 The mammalian dual glu-​pro aminoacyltransferase, GluProRS, functions as a regulator of interferon response by participating in the gamma-​interferon-​activated inhibitor of translation complex (GAIT) which regulates translation of target mRNA by binding to 3′-​UTRs82. IFN-​γ treatment of cells induces phosphorylation at two sites in GluProRS, induces its release from the aminoacylase complex it is part of, and induces its association with the GAIT complex that blocks translation of inflammatory genes.83 Protein synthesis elongation rate is controlled by tRNA abundance, codon usage, abundance of wobble-​modified tRNAs and by tRNA modifications.84–​86 mRNA codon recognition by anticodon region in tRNA does not only include Watson–​Crick base pair interaction, but also enables wobble interactions by nucleotide modification at the third wobble nucleotide in tRNA. The third position in the tRNA anticodon (position 34), when modified can induce an interaction between G and U or U and G wobble pairing.87 Post-​transcription modification at position 34 allows for additional

 271

Targeting Translation Regulation for the Development of Novel Drugs

271

wobble pairing, in which adenosine is modified to inosine allowing pairing with residues U, C and A.88,89 Wobble interactions cause slower decoding of target codons, resulting in slower movement of ribosome along the mRNA.90 tRNA isoacceptor abundance also affects translation rates. Low abundant tRNA are less available during translation, resulting in slower elongation rates and overexpression of specific tRNAs can lead to cancer progression by increasing elongation rates.91,92 In humans, 610 tRNA genes have been identified, of which only 430 are predicted to encode viable tRNAs.93,94 Expression of individual tRNA genes varies between tissues and cell types, and this enables tissue-​selective regulation of mRNAs. A change in the abundance of tRNA in a specific tissue due to mutation or single-​nucleotide polymorphism (SNP) is causative in disease.95 A silent mutation in the cystic fibrosis transmembrane conductance regulator (CFTR) mRNA in which the ACT codon for Thr is substituted by ACG leads to alteration in CFTR protein levels only in human bronchial epithelial cells due to low expression of the tRNA which decodes the mutated codon in these cells, tRNA-​CGU.96 Increasing the cellular levels of tRNA-​CGU was shown to rescue CFTR protein expression. Thus, upregulation of tRNA (CGU) expression may be a target in cystic fibrosis that can be revealed by PSM technology. Mutation in a single tRNA gene can lead to ribosome stalling and deregulate protein synthesis or can lead to mis-​aminoacylation and mistranslation in case the anticodon sequence was mutated. tRNA Arg (UCU) is encoded by six genes in humans with one isodecoder different for the others. This isodecoder is mostly expressed in the central nervous system (CNS), and includes an additional N2-​dimethyl-​G (m22G) modification which may be important for its role in translation regulation in neurons.97 Mutation in the CNS tRNA together with a ribosome rescue factor, GTPBP2, was shown to induce ribosome stalling in nerve cells and lead to neurodegeneration in mice.98 tRNAs are the most modified non-​coding RNA, with an average of 13 modifications per molecule in humans.99,100 Beside codon recognition and translation fidelity, modifications are important for tRNA folding, stability and turnover. Together with the modifications at position 34, modification at positions 37 (the anticodon loop) are the most abundant modification sites on the tRNA and are important for codon recognition and to prevent frame shift.101,102 Impaired modification of adenosine 37 in tRNA Lys (UUU) by a mutation in modification enzymes which catalyze 2-​ methylthio-​ N6-​ 2 6 threonylcarbamoyladenosine (ms t A) was shown to reduce Lys incorporation in β cells and alter glucose-​induced proinsulin biosynthesis and folding.103,104 The level of modifications in tRNA change in response to nutrients and stress and impact translation. Under stress, the modification 5-​methyl-​cytosine (m5C) in wobble position of tRNA Leu (CAA) is elevated by Trm4, leading to enhanced translation of stress-​related genes in which there are high occurrences of Leu (TTG) codons.105,106 Unmodified tRNAs are rapidly decayed107 and modification on residues outside the anticodon loop are needed for stability and function. The modification m1A58, found in almost all tRNAs, was shown to increase the affinity of the tRNA to the elongation factor 1A (EF-​1A)

272

272

Chapter 13

that delivers the tRNA to the A-​site of the ribosome. The methyltransferase TRMT6/​TRMT61A8 and the demethylase ALKBH1 can control the translation speed by determining the m1A58 modification levels of the tRNAs.108 Mutations in modification enzymes have been associated with human diseases such as cancer and neurologic and metabolic disorders.109–​ 111 Microcephaly is linked with mutations in methyltransferase NSUN2, which leads to loss of modification m5C in tRNAs. The lack of this modification increases the angiogenin-​mediated endonucleolytic cleavage of tRNA, leading to an accumulation of tRNA fragments. These fragments reduce protein translation rates and activate stress pathways, leading to reduced cell size and increased apoptosis of cortical, hippocampal and striatal neurons.112 tRNA hypomodification and tRNA-​modifying-​enzyme deregulation are linked with many human diseases including cancer and neurologic disorders.113,114

References 1. D. Wang, B. Eraslan, T. Wieland, B. Hallström, T. Hopf, D. Zolg, et al., Mol. Syst. Biol., 2019, 15(2), e8503. 2. L. Kühn, Metallomics, 2015, 7, 232–​43. 3. C. Salat-​Canela, M. Muñoz, M. Sesé, Y. Ramón, S. Cajal and T. Aasen, Biochem. Soc. Trans., 2015, 43, 465–​470. 4. K. J. Blackinton and J. D. Keene, Semin. Cell Dev. Biol., 2014, 34,  44–​54. 5. Y. Chen, Y. Chang and Y. Huang, Dev. Neurobiol., 2019, 79,  60–​74. 6. A. Bazzini, F. Del Viso, M. Moreno-​Mateos, T. Johnstone, C. Vejnar, Y. Qin, et al., EMBO J., 2016, 35, 2087–​2103. 7. J. Liu, M. Pampillo, F. Guo, S. Liu, B. Cooperman, I. Farrell, et al., J Cell Physiol., 2014, 229, 1121–​1129. 8. S. Barhoom, I. Farrell, B. Shai, D. Dahary, B. Cooperman, Z. Smilansky, et al., Nucleic Acids Res., 2013, 41(18), e177. 9. G. Rosenblum, C. Chen, J. Kaur, X. Cui, H. Zhang, H. Asahara, et al., J. Am. Chem. Soc., 2013, 135, 11322–​11329. 10. B. Stevens, C. Chen, I. Farrell, H. Zhang, J. Kaur, S. Broitman, et al., PLoS One, 2012, 7(5), e38344. 11. C. Chen, B. Stevens, J. Kaur, Z.Smilansky, B. Cooperman and Y. Goldman, Proc. Natl. Acad. Sci. U. S. A., 2011, 108, 16980–​16985. 12. S. Barhoom, J. Kaur, B. Cooperman, N. Smorodinsky, Z. Smilansky, M. Ehrlich, et al., Nucleic Acids Res., 2011, 39(19), e129. 13. C. Chen, B. Stevens, J. Kaur, D. Cabral, H. Liu, Y. Wang, et al., Mol. Cell, 2011, 42, 367–​377. 14. D. Pan, H. Qin and B. Cooperman, RNA, 2009, 15, 346–​354. 15. D. Bajusz, A. Rácz and K. Héberger, J. Cheminform., 2015, 7, 20. 16. B. Eraslan, D. Wang, M. Gusic, H. Prokisch, B. Hallström, M. Uhlén, et al., Mol. Syst. Biol., 2019, 15(2), e8513. 17. K. Miyauchi, T. Ohara and T. Suzuki, Nucleic Acids Res., 2007, 35(4), e24.

 273

Targeting Translation Regulation for the Development of Novel Drugs

273

18. R. Queiroz, T. Smith, E. Villanueva, M. Marti-​Solano, M. Monti, M. Pizzinga, et al., Nat. Biotechnol., 2019, 37, 169–​178. 19. M. Jazurek, A. Ciesiolka, J. Starega-​Roslan, K. Bilinska and W. Krzyzosiak, Nucleic Acids Res., 2016, 44, 9050–​9070. 20. G. Hanson and J. Coller, Nat. Rev. Mol. Cell Biol., 2018, 19,  20–​30. 21. K. Leppek, R. Das and M. Barna, Nat. Rev. Mol. Cell Biol., 2018, 19, 158–​174. 22. L. Herzel, D. Ottoz, T. Alpert and K. Neugebauer, Nat. Rev. Mol. Cell Biol., 2017, 18, 637–​650. 23. F. Baralle and J. Giudice, Nat. Rev. Mol. Cell Biol., 2017, 18, 437–​451. 24. S. Kelly and A. Corbett, Traffic, 2009, 10, 1199–​1208. 25. M. Amir, W. Alexander, Z. Ke, T. Matthew, E. Matthew, M. Raquel, et al., Nat. Microbiol., 2016, 2016, 16069. 26. C. Corbo, S. Orrù and F. Salvatore, Biochem. Biophys. Res. Commun., 2013, 436,  1–​5. 27. S. Das and A. Krainer, Mol. Cancer Res., 2014, 12, 1195–​1204. 28. R. de Bruin, T. Rabelink, A. van Zonneveld and E. van der Veer, Eur. Heart J., 2017, 38, 1380–​1388. 29. G Zander, H. Krebber, RNA Biol., 2017, 14, 1642–​1648. 30. Z. Ji, J. Lee, Z. Pan, B. Jiang and B. Tian, Proc. Natl. Acad. Sci. U. S. A., 2009, 106, 7028–​7033. 31. C. Mayr and D. Bartel, Cell, 2009, 138, 673–​684. 32. Y. Fu, L. Chen, C. Chen, Y. Ge, M. Kang, Z. Song, et al., Genome Res., 2018, 28, 1656–​1663. 33. H. Duan, N. Cherradi, J. Feige and C. Jefcoate, Mol. Endocrinol., 2009, 23, 497–​509. 34. L. Romo, A. Ashar-​Patel, E. Pfister and N. Aronin, Cell Rep., 2017, 20, 3057–​3070. 35. X. Li, X. Xiong, K. Wang, L. Wang, X. Shu, S. Ma, et al., Nat. Chem. Biol., 2016, 12, 311–​316. 36. D. Dominissini, S. Nachtergaele, S. Moshitch-​Moshkovitz, E. Peer, N. Kol, M. Ben-​Haim, et al., Nature, 2016, 530, 441–​446. 37. M. Helm, R. Giegé and C. Florentz, Biochemistry, 1999, 38, 13338–​13346. 38. L. Lu, C. Yi, X. Jian, G. Zheng and C. He, Nucleic Acids Res., 2010, 38, 4415–​4425. 39. N. Sonenberg and A. Hinnebusch, Cell, 2009, 136, 731–​745. 40. D. Dominissini, S. Nachtergaele, S. Moshitch-​Moshkovitz, E. Peer, N. Kol, M. Ben-​Haim, et al., Nature, 2016, 530, 441–​446. 41. R. Desrosiers, K. Friderici and F. Rottman, Proc. Natl. Acad. Sci. U. S. A., 1974, 71, 3971–​3975. 42. C. Simms, B. Hudson, J. Mosior, A. Rangwala and H. Zaher, Cell Rep., 2014, 9, 1256–​1264. 43. J. Karijolich and Y. Yu, Nature, 2011, 474, 395–​398. 44. I. Fernández, C. Ng, A. Kelley, G. Wu, Y. Yu and V. Ramakrishnan, Nature, 2013, 500, 107–​110.

274

274

Chapter 13

45. K. Lukong, K. Chang, E. Khandjian and S. Richard, Trends Genet., 2008, 24, 416–​425. 46. E. Conlon and J. Manley, Genes Dev., 2017, 31, 1509–​1528. 47. S. Gerstberger, M. Hafner and T. Tuschl, Nat. Rev. Genet., 2014, 15, 829–​845. 48. M. Hentze, A. Castello, T. Schwarzl and T. Preiss, Nat. Rev. Mol. Cell Biol., 2018, 19, 327–​341. 49. B. Lunde, C. Moore and G. Varani, Nat. Rev. Mol. Cell Biol., 2007, 8, 479–​490. 50. The database of RNA-​ binding protein specificities. [Online.] 2012. Available from: http://​rbpdb.ccbr.utoronto.ca/​index.php. 51. C. Maris, C. Dominguez and F. Allain, FEBS J., 2005, 272, 2118–​2131. 52. M. Fu and P. Blackshear, Nat. Rev. Immunol., 2017, 17, 130–​143. 53. R. Valverde, L. Edwards, and L. Regan, FEBS J., 2008, 275, 2712–​2726. 54. K. Chang and A. Ramos, FEBS J., 2005, 272, 2109–​2117. 55. T. Gutschner, M. Hämmerle, N. Pazaitis, N. Bley, E. Fiskin, H. Uckelmann, et al., Hepatology, 2014, 59, 1900–​1911. 56. J. Palanichamy, T. Tran, J. Howard, J. Contreras, T. Fernando, T. Sterne-​ Weiler, et al., J. Clin. Invest., 2016, 126, 1495–​1511. 57. L. Simone and J. Keene, Curr. Opin. Genet. Dev., 2013, 23,  35–​43. 58. L. Liu, M. Ouyang, J. Rao, T. Zou, L. Xiao, H. Chung, et al., Mol. Biol. Cell, 2015, 26, 1797–​1810. 59. T. Barnes, W. Kim, A. Mantha, S. Kim, T. Izumi, S. Mitra, et al., Nucleic Acids Res., 2009, 37, 3946–​3958. 60. I. Leontieva, Cell Cycle, 2009, 8, 490–​497. 61. S. Fagoonee, C. Bearzi, F. Di Cunto, J. Clohessy, R. Rizzi, M. Reschke, et al., PLoS One, 2013, 8(8), e72300. 62. H. Choi, C. Hwang, K. Song, P. Law, L. Wei and H. Loh, Biochem. Biophys. Res. Commun., 2009, 380(3), 431–​436. 63. M. Müller-​ McNicoll and K. Neugebauer, Nat. Rev. Genet., 2013, 14, 275–​287. 64. D. Reid and C. Nicchitta, Nat. Rev. Mol. Cell Biol., 2015, 16, 221–​231. 65. L. Cai, D. Fritz, L. Stefanovic and B. Stefanovic, J. Mol. Biol., 2010, 395, 309–​326. 66. A. Brinegar and T. Cooper, Brain Res., 2016, 1647,  1–​8. 67. P. Sahoo, D. Smith, N. Perrone-​Bizzozero and J. Twiss, J. Cell Sci., 2018, 131(8), jcs196808. 68. K. Farina, S. Huttelmaier, K. Musunuru, R. Darnell and R. Singer, J. Cell Biol., 2003, 160,  77–​87. 69. S. Hüttelmaier, D. Zenklusen, M. Lederer, J. Dictenberg, M. Lorenz, X. Meng, et al., Nature, 2005, 438, 512–​515. 70. Z. Manojlovic and B. Stefanovic, RNA, 2012, 18, 321–​343. 71. D. Nelson, H. Orr and S. Warren, Neuron, 2013, 77, 825–​843. 72. E. W. Khandjian, F. Corbin, S. Woerly and F. Rousseau, Nat. Genet., 1996, 12,  91–​93.

 275

Targeting Translation Regulation for the Development of Novel Drugs

275

73. R. Tabet, E. Moutin, J. Becker, D. Heintz, L. Fouillen, E. Flatter, et al., Proc. Natl. Acad. Sci. U. S. A., 2016, 113, E3619–​E3628. 74. E. Lee, W. Kim, K. Tominaga, J. Martindale, X. Yang, S. Subaran, et al., Mol. Cell, 2012, 45, 826–​835. 75. L. Tillmar, C Carlsson and N. Welsh, J. Biol. Chem., 2002, 277, 1099–​1106. 76. B. Alberts, A. Johnson, J. Lewis, M. Raff, K. Roberts and P. Walter, Molecular Biology of the Cell, Garland Science, New York, 4th edn, 2002. 77. S. Kim, F. Suddath, G. Quigley, A. McPherson, J. Sussman, A. Wang, et al., Three-​dimensional tertiary structure of yeast phenylalanine transfer RNA. Science, 1974, 185, 435–​440. 78. R. Giegé, M. Sissler and C. Florentz, Nucleic Acids Res., 1998, 26, 5017–​5035. 79. M. Ibba and D. Soll, Annu. Rev. Biochem., 2000, 69, 617–​650. 80. M. Guo, X. Yang and P. Schimmel, Nat. Rev. Mol. Cell Biol., 2010, 11, 668–​674. 81. O. Levi and Y. Arava, PLoS Biol., 2019, 17(5), e3000274. 82. E. Smirnova, V. Lakunina, I. Tarassov, I. Krasheninnikov and P. Kamenski, Biochemistry (Moscow), 2012, 77,  15–​25. 83. A. Arif, J. Jia, R. Mukhopadhyay, B. Willard, M. Kinter and P. Fox, Mol. Cell, 2009, 35, 164–​180. 84. J. Parmley and M. Huynen, PLoS Genet., 2009, 5(7), e1000548. 85. M. Stadler and A. Fire, RNA, 2011, 17, 2063–​2073. 86. E. Novoa, M. Pavon-​Eternod, T. Pan and L. Ribas de Pouplana, Cell, 2012, 149, 202–​213. 87. F. Crick, J. Mol. Biol., 1966, 19, 548–​555. 88. P. Agris, F. Vendeix and W. Graham, J. Mol. Biol., 2007, 366,  1–​13. 89. T. Takemoto, K. Takeishi, S. Nishimura and T. Ukita, Eur. J. Biochem., 1973, 38, 489–​496. 90. M. Stadler and A. Fire, RNA, 2011, 17, 2063–​2073. 91. H. Goodarzi, H. Nguyen, S. Zhang, D. Dill, H. Molina and S. Tavazoie, Cell, 2016, 165, 1416–​1427. 92. A. Dana and T. Tuller, Nucleic Acids Res., 2014, 42, 9171–​9178. 93. P. Chan and T. Lowe, Nucleic Acids Res., 2009, 37, D93–​97. 94. P. P. Chan and T. L. Lowe, Nucleic Acids Res., 2016, 44, D184–​189. 95. S. Kirchner and Z. Ignatova, Nat. Rev. Genet., 2015, 16, 98–​112. 96. S. Kirchner, Z. Cai, R. Rauscher, N. Kastelic, M. Anding, A. Czech, et al., PLoS Biol., 2017, 15(5), e2000779. 97. G. Zheng, Y. Qin, W. Clark, Q. Dai, C. Yi, C. He, et al., Nat. Methods, 2015, 12, 835–​837. 98. R. Ishimura, G. Nagy, I. Dotu, H. Zhou, X. Yang, P. Schimmel, et al., Science, 2014, 345, 455–​459. 99. M. Saikia, Y. Fu, M. Pavon-​Eternod, C. He and T. Pan, RNA, 2010, 16, 1317–​1327. 100. M. Machnicka, K. Milanowska, O. Osman Oglou, E. Purta, M. Kurkowska, A. Olchowik, et al., Nucleic Acids Res., 2013, 41, D262–​267.

276

276

Chapter 13

101. F. Murphy, V. Ramakrishnan, A. Malkiewicz and P. Agris, Nat. Struct. Mol. Biol., 2004, 11, 1186–​1191. 102. E. Klassen, A. Bruch and R. Schaffrath, RNA Biol., 2017, 14, 1252–​1259. 103. F. Wei, T. Suzuki, S. Watanabe, S. Kimura, T. Kaitsuka, A. Fujimura, et al., J Clin Invest., 2011, 121, 3598–​3608. 104. F. Wei and K. Tomizawa, Endocr J., 2011, 58, 819–​825. 105. C. Chan, M. Dyavaiah, M. DeMott, K. Taghizadeh, P. Dedon and T. Begley, PLoS Genet., 2010, 6(12), e1001247. 106. C. Chan, Y. Pang, W. Deng, I. Babu, M. Dyavaiah, T. Begley, et al., Nat. Commun., 2012, 3, 937. 107. A. Alexandrov, I. Chernyakov, W. Gu, S. Hiley, T. Hughes, E. Grayhack, et al., Mol. Cell, 2006, 21,  87–​96. 108. F. Liu, W. Clark, G. Luo, X. Wang, Y. Fu, J. Wei, et al., Cell, 2016, 167, 816–​828. 109. V. Rodriguez, Y. Chen, A. Elkahloun, A. Dutra, E. Pak and S. Chandrasekharappa, Genes Chromosomes Cancer, 2007, 46, 694–​707. 110. A. Alazami, H. Hijazi, M. Al-​ Dosari, R. Shaheen, A. Hashem, M. Aldahmesh, et al., J. Med. Genet., 2013, 50, 425–​430. 111. F. Wei, T. Suzuki, S. Watanabe, S. Kimura, T. Kaitsuka, A. Fujimura, et al., J. Clin. Invest., 2011, 121, 3598–​3608. 112. S. Blanco, S. Dietmann, J. Flores, S. Hussain, C. Kutter, P. Humphreys, et al., EMBO J., 2014, 33, 2020–​2039. 113. A. Torres, E. Batlle and L. Ribas de Pouplana, Trends Mol. Med., 2014, 20, 306–​314. 114. M. Pereira, S. Francisco, A. Varanda, M. Santos and A. Soares, Int. J. Mol. Sci., 2018, 19, 3738.

 277

CHAPTER 14

Classes, Modes of Action and Selection of New Modalities in Drug Discovery ERIC VALEUR* Flagship Pioneering, 55 Cambridge Parkway, Suite 800E, Cambridge, MA 02142, USA *Corresponding author. Email: [email protected]

14.1 Introduction The pharmaceutical industry has delivered many therapeutic successes for decades based on small molecules therapies and subsequently biologics.1 However, many treatments have predominantly been targeted at symptoms rather than the root causes of diseases. Diabetes is one example where many drugs aim at either controlling glucose levels or some of the complications associated with the disease. Arguably, the market is “saturated” with those disease-​management options, which increases the need to invest into research for curative treatments. This situation is similar in many other diseases areas and has resulted in a progressive shift towards novel biology. The sought-​after mechanisms are often leading to challenging, hard-​to-​drug intracellular therapeutic targets, for which classical small molecules or biologics are less well suited (vide infra). Classical small molecules have successfully drugged a plethora of protein targets including enzymes such as kinases and some proteases, receptors,

Protein Degradation with New Chemical Modalities: Successful Strategies in Drug Discovery and Chemical Biology Drug Discovery Series No. 74 Edited by Hilmar Weinmann and Craig Crews © The Royal Society of Chemistry 2021 Published by the Royal Society of Chemistry, www.rsc.org

277

278

278

Chapter 14

including G protein-​coupled receptor (GPCRs), or ion channels, to name a few. Their ability to form specific interactions with protein amino acid residues within a minimal three-​dimensional footprint confers the major advantage of enabling oral bioavailability and moreover intracellular access through passive diffusion. This also results in potential broad tissue distribution, including to peripheral tissues such as the central nerval system or the lungs. Classical biologics, namely synthetic or recombinant proteins and antibodies, typically display an opposite profile to small molecules. Their large three-​ dimensional volume enables interactions with protein targets over large surface areas. Consequently, when pockets are absent or poorly formed, they offer the opportunity to bind to their target with very strong affinity. On the other hand, the significantly larger size renders these modalities refractory to cellular uptake, to date preventing intervention against hard-​to-​drug intracellular targets. In addition, this profile usually prevents oral dosing, although progress is being made. Importantly, classical biologics and most notably antibodies can have very long plasma half-​lives, which offer the advantage of more convenient dosing regimen, such as every few months. Tissue distribution is, however, somewhat limited, with poor penetration into large tissues. Looking at the accessible target space for these classical modalities and the overall protein coding genome illustrates the vast limitation of these modalities when addressing novel biology. Indeed, an analysis showed that about 10% of protein targets have pockets, often hydrophobic, suitable for small molecules.2 At the same time only 10% of targets are secreted or membrane-​ bound and therefore targetable with classical biologics. Overall, the intersect demonstrates that about 80% of proteins are not targetable with classical modalities, which creates the need to identify or further expand other modalities, which can to a large extent be called “new.”3 Other qualifiers for such modalities have been used, calling these other molecules “alternative” based on the argument that many are arguably not truly new and are simply coming back in fashion. Nevertheless, the term “new modalities” is now broadly accepted and used across the industry with many biotechnology companies exploiting this term for approaches such as those discussed in this book. Several large pharmaceutical companies have created units dedicated to these “new modalities,” which was initiated by AstraZeneca in 2013, with Novartis, Roche and others following shortly after. Academic institutions are also adopting this terminology explicitly, including the Max Planck Institute for Molecular Physiology, A*STAR and others. New modalities can be divided into different classes based on different criteria.3 A traditional medicinal chemistry-​centric view would limit it to synthetic modalities and eventually not cover additional approaches such as genome editing. Herein we adopt a progressist view of including all novel approaches in this vocabulary, regardless of molecular weight. Those of larger molecular weight or higher-​order structure will be discussed more succinctly, and the reader will be referred to appropriate reviews for further reading. A first category of new modalities is based on nucleic acids, leveraging oligomers either as a single entity or as part of a more complex system.

 279

Classes, Modes of Action and Selection of New Modalities in Drug Discovery

279

These include antisense oligonucleotides, small interfering RNA, aptamers, modified RNA and CRISPR/​CAS9, in which the oligonucleotide is one component of the genome editing strategy. Another type of new modalities originates from peptides, which are being hypermodified to circumvent some of the challenges faced with this arguably “classical” modality. Increased complexity achieved through single or multiple cyclic systems increases stability, binding affinity and sometimes even results in intracellular activity. Non-​peptidic macrocycles may also be considered a new modality, although they are strongly derived by small molecules.4,5 These classes of modalities can also be merged or combined among themselves or together with classical modalities to form a range of novel hybrid modalities, whose application spans well beyond classical modes of action (MOAs) or single target engagement. Finally, new modalities also encompass higher-​order structures, where entire biological systems are being developed as therapeutics, including cell therapy6–​8 and microbes.9–​11 In this chapter, the main classes of new modalities will be discussed and systematically analysed by providing a description of the chemical nature of these modalities, the MOAs they can be leveraged for, and their strengths and limitations for therapeutics. Higher-​order new modalities fall outside the classical target-​based drug discovery paradigm and are therefore outside the scope of this chapter.

14.2 Nucleic Acid-​based Modalities Nucleic acids are at the root of any biological life in the form of DNA and its transcriptional products, which include a range of protein-​coding and non-​ coding RNAs. The Watson–​Crick base pairing has led to the mimicking of these naturally occurring factors to modulate biology with synthetic molecules.12 Nucleotides are constituted of a nucleobase, a carbohydrate and a phosphate backbone, and their polymers, oligonucleotides (ONs), can therefore be synthetically produced to target a specific single-​stranded sequence, most notably RNA. Endogenous oligonucleotides are typically short-​lived and mimicking these without any chemical modifications is therefore not a viable route. Over the last three decades numerous modifications have been designed and tested to improve, in particular, the stability of oligonucleotides with the prospect of using them as therapeutics, and these modifications will be discussed later in this section. Progress has enabled the creation of a whole new class of medicines which can be aimed at targeting or mimicking RNA species, or used as interactors with proteins upon folding. In this section, the different MOAs that can be targeted with nucleic acid-​based therapeutics will first be covered and the various classes of this modality will then be discussed.

14.2.1 Targetable Modes of Action Nucleic acid-​based therapeutics can be leveraged for a broad range of MOAs.3 The main MOAs are summarised on Figure 14.1. Importantly, applications of

280

280

Chapter 14

Figure 14.1 Overview of mechanisms leveraged by nucleic acid-​based modalities.

ONs are moving beyond their original MOA and some of these new mechanisms will be covered in this section.

14.2.1.1 Protein Recognition Similar to how peptides can fold into complex structures and proteins, oligonucleotides can also form secondary and tertiary structures.13,14 These folded architectures can bind to proteins and therefore block protein functions either by hindering an active site or by disrupting protein–​protein interactions, therefore following the principle of antibodies. In addition, they can activate receptors as agonists. As will be discussed below, this MOA is currently limited to membrane-​bound or extracellular proteins due to the very limited cell permeability. Although many aptamer programs are in the clinic, this MOA has therefore remained arguably anecdotic in the grand scheme of what oligonucleotides can achieve. The most powerful MOAs for oligonucleotides reside indeed in their sequence-​based ability to target RNA or DNA, an approach which is currently unrivalled by other modalities.

14.2.1.2 Direct and Indirect Downregulation of RNA Levels Cells regulate RNA species through several means. Upon transcription, mRNA is translated by the ribosome to form proteins. mRNA can be degraded and therefore downregulated via the so-​called RISC complex, which is formed between the argonaute protein and a non-​coding RNA species named micro-​ RNA (miRNA). miRNAs are first transcribed as pri-​miRNA and translocated

 281

Classes, Modes of Action and Selection of New Modalities in Drug Discovery

281

to the cytoplasm as pre-​miRNA, then cleaved by the dicer endoribonuclase to form miRNA. These short, single-​stranded RNAs load into the argonaut protein, and then recognise their target RNA sequence. Upon formation of the ternary complex between argonaute, miRNA and the RNA, the RNA function is either repressed when sequence complementarity is incomplete, or cleaved and therefore degraded by nucleases in the event of full sequence complementarity. This RNA interference mechanism can be hijacked in the form of small interfering RNAs (siRNA) or with synthetic miRNAs leading to repression or downregulation of mRNA.15–​18 Another approach consists of repressing RNA function or processing by binding along a section of the sequence. This is achieved by antisense oligonucleotides (ASOs), which can typically block translation or modulate alternative splicing.19,20 ASOs can also be aimed at exploiting the principle of RNA degradation originating from DNA–​RNA duplexes. A specific class of ASOs named gapmers can induce RNAse H recruitment leading to degradation of the target RNA. Finally, RNA levels can be downregulated with ribozymes which are essentially nucleic acid-​based enzymes.21

14.2.1.3 Direct and Indirect Upregulation of RNA Levels While downregulating RNA levels has become ubiquitous through the development of ASOs and siRNAs, the ability to upregulate RNA levels is far more challenging. However, progress in chemical modifications has enabled the possibility to directly mimic mRNA. These synthetic mRNAs, also called modified mRNA (modRNA), can be transfected into cells and therefore represent a direct increase in mRNA levels, leading to increased protein levels.22,23 This is of particularly high interest for diseases where functional proteins are either downregulated or depleted due to genetic mutations or aberrant splicing. In addition to treating genetic diseases or replacing proteins, an additional application is using modRNAs as vaccines, where antigens can be directly synthesised by the cell translation machinery. Another approach is to activate the promoter region of genes, via so-​called small activating RNA.24,25 These double-​stranded RNAs are processed and loaded into the argonaute protein, therefore sharing similarities with siRNAs, albeit to achieve gene expression in this case. While this strategy is still emerging, progress has been made with the first molecule being in Phase I clinical trials. A different and indirect approach to upregulate RNA consists of targeting the endogenous downregulation mechanism. As discussed above, miRNAs are involved in the repression and degradation of RNA upon formation of the RISC complex. Thus, ASOs that bind and repress miRNAs prevent loading of the miRNA into the RISC complex, and therefore prevent downregulation of mRNA, leading to higher levels of mRNA and therefore protein levels.26–​28

282

282

Chapter 14

14.2.1.4 Genome Editing Genome editing offers the promise of permanently correcting genetic defects, which explains the surge in research, and the developments of this strategy. Many genome editing systems are being prosecuted, and the CRISPR-​ Cas9 technique illustrates the vital role of oligonucleotides.29–​31 Within this approach an oligonucleotide is used to guide the editing system to the ­targeted gene. To achieve this, the guide RNA is designed based on the complementarity of the targeted gene and associated with the CAS9 endonuclease gene.32 Expression of the system results in directing the Cas9 protein to the gene followed by cleavage of the DNA. Different versions now allow gene downregulation, but also repair by inserting a correct gene sequence, and the interested reader is directed to several recent reviews for more details.33–​35

14.2.2 Classes of Nucleic Acid-​based Modalities 14.2.2.1 Antisense Oligonucleotide (ASO) and Small Interfering RNA (siRNA) ASOs and siRNA currently represent the dominant class of nucleic acid-​ based therapeutics and will therefore be discussed in more detail. Both approaches have been primarily aimed at reducing RNA levels of a specific target, but differ in their mechanism. ASOs also offer a broader range of MOAs beyond RNA degradation, including splice switching. Differences also exist in their composition: ASOs are single-​stranded ONs typically containing 16–​20 nucleotides;36–​38 siRNAs are double-​stranded ONs made of 20–​25 nucleotides, with one strand targeting the RNA and the other strand, also named the “passenger strand,” being removed upon intracellular processing by dicer. Binding to the argonaute protein then forms the RISC complex.15,18,38 Noteworthy, another type of ON called short hairpin RNA (shRNA) follows the same principle and constitution as siRNA, with the two strands being connected via a loop. This modality is processed by dicer to generate siRNA in situ.39 14.2.2.1.1  Chemical Modifications.  Oligonucleotides based on naturally occurring nucleotides suffer from poor properties, most notably intra-​and extracellular stability and limited uptake. Thus, a wealth of modifications has been explored to address these limitations, in addition to modulating binding affinity to their target RNAs (Figure 14.2).36–​38,40 Modifications have been targeted at all three components of the ONs: the nucleobase, the carbohydrate and the backbone.41 Since an ON needs to retain its ability to form Watson–Crick-base pairing, modification of the nucleobase is challenging. An extensive set of modified nucleobases have been investigated but few are routinely used in ASOs or siRNAs.42–​44 One exception is 5-​methyl cytidine: C-​G motifs induce immune activation, and while one approach is to avoid this motif in an ON sequence, this strategy restricts the range of targetable sequences.45

 283

Classes, Modes of Action and Selection of New Modalities in Drug Discovery

283

Figure 14.2 Examples of major nucleotide modifications.

Thus, using 5-​methyl cytidine in place of cytidine is an effective way to retain the nucleotide identity while decreasing the risk for immune activation. Backbone modifications represent a significant means to increase resistance to endo-​and exonuclease cleavage. Replacing the phosphodiester backbone with a phosphorothioate (PS) moiety, in which a sulfur atom replaces one oxygen atom, dramatically increases nuclease stability.46 Interestingly, the PS linkage modifies the properties of ONs beyond stability. Indeed, this modification results in higher plasma protein binding leading to reduced renal clearance. It is also believed that protein binding results in increased uptake and thus intracellular exposure, and in addition promotes nuclear uptake, although the mechanisms supporting these findings are poorly characterised. Other modifications of the backbone include phosphoroboran, in which a boran atom replaces an oxygen atom in the phophorodiester.47 Overall, PS remains one of the most widely used backbone modifications. Notably, the replacement of a phosphorodiester by a phosphorothioate introduces a chiral center in the oligonucleotide at every single replacement position. Thus, a fully modified ON contains a large number of chiral centres resulting in a mixture of numerous diastereoisomers after synthesis. It is has been proposed that generating diastereopure ONs can result in ONs with superior properties,48–​50 but this hypothesis has been contested in other studies.51 The vast majority of chemical modifications has been targeted at the carbohydrate component. These strategies usually provide additional resistance to nucleases, and can also increase the binding affinity of the designed ON to its RNA target. For example, replacing the 2′-​hydroxy group of the ribose with a fluorine atom or with an O-​methyl or O-​methoxyethyl (MOE) group increases the steric hindrance and thus reduces cleavage by nucleases.52–​54 These modifications also increase affinity by pre-​organising the ON into an A-​form via induction of a 3′-​endo conformation of the ribose. Notably, these

284

284

Chapter 14

modifications display less protein binding, which therefore mitigates promiscuous binding, and thus can result in a better safety profile. The carbohydrate moiety has also been further modified and stabilised by introducing a bridge between the 2′-​oxygen atom and the 4′-​carbon atom. This approach has been exploited in many guises, forming the broad category of bridged nucleic acids, among which LNAs (locked nucleic acids) and cEt (constrained ethyl) are the most prominent examples.55–​58 Similar to the other 2′-​modifications, the introduction of the bridge locks the sugar in a 3′-​endo conformation resulting in significantly higher affinity. Other examples includes cMOE, which exploits the 2′-​MOE modification and bridges it to the 4′ position, and tricyclic DNA (tcDNA), which incorporates two bridges between the carbon atoms in the 3′ and 5′ positions on one hand, and the 5′ and 6′ positions on the other hand.59,60 Some researchers have also modified the backbone and carbohydrate components at the same time. Thus, replacement with a peptidic backbone has generated peptide nucleic acids (PNAs) which are entirely resistant to nuclease degradation and display strong hybridisation to DNA and RNA.61 However, the physico-​chemical properties are very problematic with poor solubility and negligible cellular uptake, rendering this approach unviable from a therapeutic standpoint. Another concomitant modification of the backbone and carbohydrate is the phosphoramidate morpholino oligomers (PMO).62,63 In this strategy, the phosphodiester bond is replaced by a phosphorodiamidate, while the sugar is replaced by a morpholine ring. This approach results in very stable ONs and maintains good hybridisation to RNA and DNA. However, cellular uptake is inferior to other modifications. 14.2.2.1.2 Design. Due to their direct hybridisation mechanism, ASOs tolerate broader modifications than siRNAs, in which modifications are at risk of sterically clashing with the argonaute protein within the RISC complex. Thus, the smaller 2′F, 2′OMe and sometimes 2′MOE have been the predominant modifications in siRNA concomitantly with the phosphorothioate backbone. In the case of ASOs, modifications and their place of incorporation directly depends on whether the ON is developed as a stoichiometric binder, for example to induce splice switching, or at promoting degradation of the target RNA. In the first scenario, any modification can be introduced to generate so-​called mixmers, although some are not compatible with each other from a synthetic standpoint. For example, PMOs are not compatible with classical oligonucleotide solid-​phase synthesis. In the second scenario, restrictions exist to enable degradation. Typically, these ASOs are termed “gapmer” and incorporate a DNA section that hybridises to the target RNA to form a DNA–​RNA duplex, which is then recognised by RNAse H resulting in cleavage of the target RNA. This DNA section is usually 10 nucleotides long (the gap) and is flanked by 3–​5 modified RNA nucleotides on each end (Figure 14.3). Strong affinity enhancers such as LNA and cEt require only 2–​3 nucleotides on each end, while previous generations such as F, OMe or MOE usually require 5 nucleotides in the flanked regions.

 285

Classes, Modes of Action and Selection of New Modalities in Drug Discovery

285

Figure 14.3 Differences in the design of stoichiometric and gapmer ASOs.

To date, seven ONs based on ASOs (fomivirsen, mipomersen, nusinersen, inotersen and eteplirsen) and siRNA (patisiran, givosiran) have been approved. Many others are currently undergoing clinical development across several disease areas including cancer, Duchenne muscular dystrophy, diabetes and hypercholesteremia. Nusinersen, a 2′MOE ON, represents an interesting example with its mechanism of splice switching, for which the ASOs stoichiometrically block a site on a pre-​mRNA.64 Spinal muscular atrophy (SMA) is a neuromuscular disease triggered by the lack of functional survival motor neuron (SMN) protein.65 Nusinersen targets the SMN2 gene by altering splicing of the SMN mRNA resulting in upregulation of functional SMN protein. Eteplirsen, a full PMO ON, follows a similar principle to treat Duchenne muscular dystrophy by inducing exon skipping to increase the production of functional dystrophin.66 Inotersen is a 2′MOE ON, which adopts a gapmer design and thus contain 10 DNA nucleotides flanked by 5 MOE nucleotides.67 The ON enables treatment of hereditary transthyretin amyloidosis, by reducing the levels of thransthyretin mRNA and therefore its protein levels, thus reducing deposition of this protein in the peripheral nervous systems as well as other tissues such as heart and kidney.

14.2.2.2 Modified mRNA (modRNA) The promise of using ONs to directly increase protein levels is an exciting paradigm considering the paucity of approaches that can achieve u ­ pregulation. In this respect, the development of modRNA represents a significant expansion of the arsenal of MOAs that drug discoverers can exploit.22,23 Since modRNAs encode for translation like any mRNA, they are typically very long, constituted of thousands of nucleotides. Like any other ON, modifications are required for increased stability, but also for mitigating immune response activation. In this case, however, the modifications need to be compatible with RNA polymerase for in vitro transcription to generate the ON, and with the ribosomal machinery for in situ translation. Thus, backbone and carbohydrate modifications are typically not tolerated, and efforts have therefore been

286

286

Chapter 14

directed at the nucleobase. For example, converting uridine to pseudouridine or cytidine to 5-​methylcytidine can result in up to 10-​fold reduced immune activation.68,69 Interestingly, modifications of the nucleobases affect protein translation in a cell-​type and codon-​specific manner.70 Other modifications of the nucleobases include N1-​methylpseudouridine and 5-​methoxyuridine. A representative example of the modified mRNA technology is an mRNA encoding for the VEGF-​A protein. Intramyocardial injection in minipigs following myocardial infarction resulted in improved heart function. The approach was then taken to humans, in the very first trial of this modality, by looking at VEGF-​A production in patients with type 2 diabetes.71 The experimental treatment was administered by intradermal injection in the forearm, and skin microdialysis, demonstrated increased VEGFA protein levels and improved skin blood flow, which allowed advancement into Phase II. Another example is a triple therapy developed to address the dysfunctional immune system of solid tumors through pro-​inflammatory cytokines.72 Cooperativity was achieved by modRNAs encoding for IL-​ 36, a cytokine involved in damaged tissues, and for IL23, another cytokine which coordinates immune responses centrally, in addition to a modRNA encoding for OX40L, a T-​cell costimulatory. Remarkably, upon intratumoral injection, the combination resulted in pronounced tumour regression in a cell line refractory to other treatments. Moreover, treatment resulted in long-​lasting antitumor memory, and achieved over 70% response in a model resistant to checkpoint blockade. These impressive results translated to human cell lines which paved the way for clinical development, with the approach currently being in Phase I.

14.2.2.3 Aptamers Aptamers are single-​stranded ONs which fold into secondary and ternary structures. Similar to protein and antibodies, binding affinity to a target protein is very high and with high specificity. As mentioned previously, cellular uptake is negligible and therefore applications are limited to extracellular targets. Their smaller size (