Evolutionary Dynamics of Malignancy: The Genetic and Environmental Causes of Cancer 3031325729, 9783031325724

Table of contents :
Acknowledgments
Contents
Abbreviations
Chapter 1: Cancer as a Disease of Cell Proliferation
1.1 What Is Evolutionary Dynamics?
1.2 Cancer Enzymology
1.3 Oncogenes and Tumour Suppressor Genes
1.4 Signalling Pathways
1.5 Signalling Dynamics: ON/OFF Switches and Volume Controls
1.6 Dynamics of Tumour Growth
1.7 The Hallmarks of Cancer
1.8 Methotrexate: An Antiproliferative Drug
References
Chapter 2: Genetic and Chromosomal Instability
2.1 Cancer as a Disease of Ageing
2.2 Radiation and Chemical Carcinogenesis
2.3 Viral Carcinogenesis
2.4 Burkitt´s Lymphoma and Chromosomal Translocation
2.5 Aneuploidy and Duesberg´s Hypothesis
2.6 Differences Between Carcinomas and Lymphomas
2.7 Microsatellite Instability
2.8 Tumour Heterogeneity
2.9 Cancer Genome Projects: Driver and Passenger Mutations
2.10 Replicative Stress and Chromosomal Instability
2.11 Modelling Mutations with a Genetic Algorithm
2.12 Polyploidy
2.13 Mutation Rates and Lifetime Cancer Risk
2.14 Alkylating Agents: Drugs that Cross-Link DNA
References
Chapter 3: Cancer as a Disease of Defective Cell Cycle Checkpoint Function
3.1 Differentiation or Cell Division?
3.2 The Mammalian Cell Cycle
3.3 P53, the Guardian of the Genome
3.4 The G1:S Checkpoint
3.5 Mutations that Inactivate or Over-Ride the Checkpoint
3.6 Benign Hyperplasia and Premalignant Conditions
3.7 Inhibitors of Cyclin-Dependent Kinases
3.8 The CYCLOPS Model of the Cell Cycle
3.9 Palbociclib-Induced Cell Cycle Arrest
3.10 Selectivity of Signalling Pathway Inhibition Against Mutants
References
Chapter 4: The DNA Damage Checkpoint
4.1 DNA Repair Pathways
4.2 Selectivity of DNA-Damaging Drugs
4.3 Mutant or Abnormally Expressed DNA Repair Enzymes in Tumours
4.4 Cross-Talk Between the DNA Damage Checkpoint and the SAC
4.5 Replication Catastrophe
4.6 Modelling Inhibition of the DDR
4.7 Kinetics of the Checkpoint
4.8 Olaparib: An Inhibitor of the DDR
References
Chapter 5: Dynamics of the Spindle Assembly Checkpoint
5.1 The Mitotic Spindle
5.2 Mitotic index as a Measure of Cell Proliferation
5.3 The Spindle Assembly Checkpoint
5.4 The SAC as Guardian of the Karyotype
5.5 SAC Over-Ride and Cancer
5.6 Loss-of-Function Mutations in the SAC
5.7 Inhibitors of SAC Components as Anticancer Drugs
5.8 The SAC and Speciation
5.9 Modelling Checkpoint Mutations with a Finite State Machine
5.10 A Kinetic Model of the SAC
5.11 The Two Checkpoints Theory of Cancer
References
Chapter 6: Dynamics of Drug Resistance
6.1 Luria and Delbruck´s Fluctuation Equation
6.2 The Model of Goldie and Coldman
6.3 Intrinsic and Acquired Drug Resistance
6.4 In Vitro, In Vivo and Clinical Endpoints for Chemotherapy
6.5 Mechanisms of Acquired Drug Resistance
6.6 Evolutionary Dynamics of Drug Resistance
6.7 Model Validation and Interspecies Scaling
6.8 Principles of Combination Chemotherapy
6.9 Adaptive Cancer Therapy
6.10 Cisplatin: A Magic Bullet?
References
Chapter 7: Chronic Myeloid Leukaemia: A One-Hit Malignancy
7.1 CML as a Disease of Redox Imbalance
7.2 The Philadelphia Chromosome and the Bcr-Abl Translocation
7.3 CML as a Constitutively Activated Innate Immune Response
7.4 Mcl-1 and Myeloid Cell Immortalisation
7.5 Reactive Oxygen Species, Ageing, and Cancer
7.6 8-Oxoguanine as a Mutagen
7.7 Imatinib in the Treatment of CML
7.8 Modelling the Evolutionary Dynamics of CML
References
Chapter 8: Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy
8.1 Cancer as a Disease of Ageing: The Age Distribution of CMML
8.2 Epigenetic Changes in Malignancy
8.3 Epigenetic Gene Silencing and the Role of TET2
8.4 Loss-of-Function Mutations and Tumour Suppressor Genes
8.5 Hypomethylating Agents as Treatments for CMML
8.6 Evolutionary Dynamics of CMML
References
Chapter 9: The Cancer Stem Cell and Tumour Progression
9.1 Tumour Progression as a Process of Natural Selection
9.2 Origins of Cancer Stem Cells
9.3 Driver and Passenger Mutations in Tumour Progression
9.4 De-differentiation
9.5 Angiogenesis as an Aspect of Tumour Progression
9.6 The Warburg Effect
9.7 The Epithelial-Mesenchymal Transition
9.8 Metastasis as an Aspect of Tumour Progression
9.9 Modelling Metastasis
9.10 The Big Bang Model of Tumour Growth
9.11 Antiandrogens in Treatment of Prostate Cancer
References
Chapter 10: Evading the Antitumour Immune Response
10.1 Cells of the Immune System
10.2 The Major Histocompatibility Complex
10.3 Immunotherapy
10.4 Escape from Immune Surveillance
10.5 Immune Checkpoint Inhibitors
10.6 Cancer Vaccines in Treatment and Prevention
10.7 Cellular Immunotherapy
10.8 Modelling Antitumour Immunity
References
Chapter 11: Implications of Evolutionary Dynamics for Cancer Treatment and Prevention
11.1 Current Views of the Causes of Cancer
11.2 Implications of Evolutionary Dynamics for Diagnosis
11.3 Implications of Evolutionary Dynamics for Cancer Treatment
11.4 Implications of Evolutionary Dynamics for Tumour Prevention
11.5 Manipulating the Environment to Reduce Cancer Incidence
11.6 Using Drugs to Inhibit Tumour Progression
11.7 Artificial Intelligence Systems for Predicting Tumour Progression
11.8 In Silico Clinical Trial Modelling
11.9 Clinical Endpoints: Progression-Free Survival
11.10 Cytotoxicity and Cytostasis
11.11 The Potential Role of Cancer Vaccines
References
Chapter 12: In Science All Conclusions Are Provisional
12.1 Why Has Cancer Been an Exception in the March of Medical Progress?
12.2 Deterministic and Probabilistic Events in Models of Malignancy
12.3 One-Hit, Two-Hit, and Three-Hit Malignancies: Causes and Consequences
12.4 Major Ideas in the Development of the Evolutionary Dynamics of Malignancy
12.5 All Stages of Transformation and Progression Have Been Targeted by Anticancer Drugs
12.6 Future Progress Will Require a Deeper Understanding of Malignant Progression
12.7 Unanswered Questions that May Be Studied by Evolutionary Dynamics
12.8 Cancer as a Disease of Gene Expression
References
Appendix: Using the Online Supplements
Index

Citation preview

Robert C. Jackson

Evolutionary Dynamics of Malignancy The Genetic and Environmental Causes of Cancer

Evolutionary Dynamics of Malignancy

Robert C. Jackson

Evolutionary Dynamics of Malignancy The Genetic and Environmental Causes of Cancer

Robert C. Jackson Pharmacometrics Ltd Cambridge, UK

ISBN 978-3-031-32573-1 ISBN 978-3-031-32572-4 https://doi.org/10.1007/978-3-031-32573-1

(eBook)

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Acknowledgments

I have benefited from discussion and debate with many chemists, biologists, clinicians, and biomathematicians. Colleagues to whom I am especially grateful are Brian Adger, Robin Bannister, Duncan Jodrell, Hitesh Mistry, and Tom Radivoyevitch. All the articles cited in the references have provided valuable information and insights. Books that have particularly helped me are Martin Nowak's founding text, “Evolutionary Dynamics” and two previous monographs on cancer dynamics, those by Frank and by Wodarz and Komarova.

v

Contents

1

2

Cancer as a Disease of Cell Proliferation . . . . . . . . . . . . . . . . . . . . . 1.1 What Is Evolutionary Dynamics? . . . . . . . . . . . . . . . . . . . . . . 1.2 Cancer Enzymology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Oncogenes and Tumour Suppressor Genes . . . . . . . . . . . . . . . 1.4 Signalling Pathways . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.5 Signalling Dynamics: ON/OFF Switches and Volume Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.6 Dynamics of Tumour Growth . . . . . . . . . . . . . . . . . . . . . . . . 1.7 The Hallmarks of Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.8 Methotrexate: An Antiproliferative Drug . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1 2 3 5 6 12 14 17 20 22

Genetic and Chromosomal Instability . . . . . . . . . . . . . . . . . . . . . . . 2.1 Cancer as a Disease of Ageing . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Radiation and Chemical Carcinogenesis . . . . . . . . . . . . . . . . . 2.3 Viral Carcinogenesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Burkitt’s Lymphoma and Chromosomal Translocation . . . . . . . 2.5 Aneuploidy and Duesberg’s Hypothesis . . . . . . . . . . . . . . . . . 2.6 Differences Between Carcinomas and Lymphomas . . . . . . . . . 2.7 Microsatellite Instability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.8 Tumour Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.9 Cancer Genome Projects: Driver and Passenger Mutations . . . . 2.10 Replicative Stress and Chromosomal Instability . . . . . . . . . . . 2.11 Modelling Mutations with a Genetic Algorithm . . . . . . . . . . . . 2.12 Polyploidy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.13 Mutation Rates and Lifetime Cancer Risk . . . . . . . . . . . . . . . . 2.14 Alkylating Agents: Drugs that Cross-Link DNA . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25 26 27 28 29 30 31 32 32 33 35 36 37 38 41 42

vii

viii

3

4

5

6

Contents

Cancer as a Disease of Defective Cell Cycle Checkpoint Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1 Differentiation or Cell Division? . . . . . . . . . . . . . . . . . . . . . . 3.2 The Mammalian Cell Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 P53, the Guardian of the Genome . . . . . . . . . . . . . . . . . . . . . . 3.4 The G1:S Checkpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Mutations that Inactivate or Over-Ride the Checkpoint . . . . . . 3.6 Benign Hyperplasia and Premalignant Conditions . . . . . . . . . . 3.7 Inhibitors of Cyclin-Dependent Kinases . . . . . . . . . . . . . . . . . 3.8 The CYCLOPS Model of the Cell Cycle . . . . . . . . . . . . . . . . . 3.9 Palbociclib-Induced Cell Cycle Arrest . . . . . . . . . . . . . . . . . . 3.10 Selectivity of Signalling Pathway Inhibition Against Mutants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45 46 47 49 50 51 52 53 54 57 60 63

The DNA Damage Checkpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 DNA Repair Pathways . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Selectivity of DNA-Damaging Drugs . . . . . . . . . . . . . . . . . . . 4.3 Mutant or Abnormally Expressed DNA Repair Enzymes in Tumours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4 Cross-Talk Between the DNA Damage Checkpoint and the SAC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Replication Catastrophe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 Modelling Inhibition of the DDR . . . . . . . . . . . . . . . . . . . . . . 4.7 Kinetics of the Checkpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.8 Olaparib: An Inhibitor of the DDR . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

65 65 67

Dynamics of the Spindle Assembly Checkpoint . . . . . . . . . . . . . . . . 5.1 The Mitotic Spindle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Mitotic index as a Measure of Cell Proliferation . . . . . . . . . . . 5.3 The Spindle Assembly Checkpoint . . . . . . . . . . . . . . . . . . . . . 5.4 The SAC as Guardian of the Karyotype . . . . . . . . . . . . . . . . . 5.5 SAC Over-Ride and Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . 5.6 Loss-of-Function Mutations in the SAC . . . . . . . . . . . . . . . . . 5.7 Inhibitors of SAC Components as Anticancer Drugs . . . . . . . . 5.8 The SAC and Speciation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.9 Modelling Checkpoint Mutations with a Finite State Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.10 A Kinetic Model of the SAC . . . . . . . . . . . . . . . . . . . . . . . . . 5.11 The Two Checkpoints Theory of Cancer . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

85 86 87 88 90 91 92 92 93

68 69 70 70 72 82 82

93 97 98 99

Dynamics of Drug Resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 6.1 Luria and Delbrūck’s Fluctuation Equation . . . . . . . . . . . . . . . 104 6.2 The Model of Goldie and Coldman . . . . . . . . . . . . . . . . . . . . 104

Contents

7

8

9

ix

6.3 Intrinsic and Acquired Drug Resistance . . . . . . . . . . . . . . . . . 6.4 In Vitro, In Vivo and Clinical Endpoints for Chemotherapy . . . 6.5 Mechanisms of Acquired Drug Resistance . . . . . . . . . . . . . . . 6.6 Evolutionary Dynamics of Drug Resistance . . . . . . . . . . . . . . 6.7 Model Validation and Interspecies Scaling . . . . . . . . . . . . . . . 6.8 Principles of Combination Chemotherapy . . . . . . . . . . . . . . . . 6.9 Adaptive Cancer Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.10 Cisplatin: A Magic Bullet? . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

106 107 108 111 119 121 122 127 128

Chronic Myeloid Leukaemia: A One-Hit Malignancy . . . . . . . . . . . 7.1 CML as a Disease of Redox Imbalance . . . . . . . . . . . . . . . . . . 7.2 The Philadelphia Chromosome and the Bcr-Abl Translocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.3 CML as a Constitutively Activated Innate Immune Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4 Mcl-1 and Myeloid Cell Immortalisation . . . . . . . . . . . . . . . . . 7.5 Reactive Oxygen Species, Ageing, and Cancer . . . . . . . . . . . . 7.6 8-Oxoguanine as a Mutagen . . . . . . . . . . . . . . . . . . . . . . . . . . 7.7 Imatinib in the Treatment of CML . . . . . . . . . . . . . . . . . . . . . 7.8 Modelling the Evolutionary Dynamics of CML . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

131 132

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy . . . 8.1 Cancer as a Disease of Ageing: The Age Distribution of CMML . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.2 Epigenetic Changes in Malignancy . . . . . . . . . . . . . . . . . . . . . 8.3 Epigenetic Gene Silencing and the Role of TET2 . . . . . . . . . . 8.4 Loss-of-Function Mutations and Tumour Suppressor Genes . . . 8.5 Hypomethylating Agents as Treatments for CMML . . . . . . . . . 8.6 Evolutionary Dynamics of CMML . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

153 154 155 157 158 158 162 175

The Cancer Stem Cell and Tumour Progression . . . . . . . . . . . . . . . 9.1 Tumour Progression as a Process of Natural Selection . . . . . . . 9.2 Origins of Cancer Stem Cells . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Driver and Passenger Mutations in Tumour Progression . . . . . . 9.4 De-differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5 Angiogenesis as an Aspect of Tumour Progression . . . . . . . . . 9.6 The Warburg Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.7 The Epithelial-Mesenchymal Transition . . . . . . . . . . . . . . . . . 9.8 Metastasis as an Aspect of Tumour Progression . . . . . . . . . . . 9.9 Modelling Metastasis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.10 The Big Bang Model of Tumour Growth . . . . . . . . . . . . . . . . 9.11 Antiandrogens in Treatment of Prostate Cancer . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

179 179 181 184 184 185 187 191 192 194 195 196 200

132 133 137 137 138 139 139 149

x

Contents

10

Evading the Antitumour Immune Response . . . . . . . . . . . . . . . . . . 10.1 Cells of the Immune System . . . . . . . . . . . . . . . . . . . . . . . . . 10.2 The Major Histocompatibility Complex . . . . . . . . . . . . . . . . . 10.3 Immunotherapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.4 Escape from Immune Surveillance . . . . . . . . . . . . . . . . . . . . . 10.5 Immune Checkpoint Inhibitors . . . . . . . . . . . . . . . . . . . . . . . . 10.6 Cancer Vaccines in Treatment and Prevention . . . . . . . . . . . . . 10.7 Cellular Immunotherapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10.8 Modelling Antitumour Immunity . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Current Views of the Causes of Cancer . . . . . . . . . . . . . . . . . . 11.2 Implications of Evolutionary Dynamics for Diagnosis . . . . . . . 11.3 Implications of Evolutionary Dynamics for Cancer Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.4 Implications of Evolutionary Dynamics for Tumour Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.5 Manipulating the Environment to Reduce Cancer Incidence . . . 11.6 Using Drugs to Inhibit Tumour Progression . . . . . . . . . . . . . . 11.7 Artificial Intelligence Systems for Predicting Tumour Progression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.8 In Silico Clinical Trial Modelling . . . . . . . . . . . . . . . . . . . . . . 11.9 Clinical Endpoints: Progression-Free Survival . . . . . . . . . . . . . 11.10 Cytotoxicity and Cytostasis . . . . . . . . . . . . . . . . . . . . . . . . . . 11.11 The Potential Role of Cancer Vaccines . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

12

In Science All Conclusions Are Provisional . . . . . . . . . . . . . . . . . . . 12.1 Why Has Cancer Been an Exception in the March of Medical Progress? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.2 Deterministic and Probabilistic Events in Models of Malignancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.3 One-Hit, Two-Hit, and Three-Hit Malignancies: Causes and Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.4 Major Ideas in the Development of the Evolutionary Dynamics of Malignancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12.5 All Stages of Transformation and Progression Have Been Targeted by Anticancer Drugs . . . . . . . . . . . . . . . . . . . . . . . . 12.6 Future Progress Will Require a Deeper Understanding of Malignant Progression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

205 206 208 209 210 212 212 213 214 220 223 224 225 226 230 231 232 234 235 240 240 241 241 245 246 247 248 250 254 255

Contents

xi

12.7

Unanswered Questions that May Be Studied by Evolutionary Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255 12.8 Cancer as a Disease of Gene Expression . . . . . . . . . . . . . . . . . 259 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259 Appendix: Using the Online Supplements . . . . . . . . . . . . . . . . . . . . . . . . 261 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263

Abbreviations

AI AK-A AK-B ALL AML AMP-K APC AR AUC BMP BPH cdk CLL CML CMML CT DDR DNADC DNMT1 DSB EBV EGF EMT FAK GDP GIST GLK GPCR GSSG GTP

Artificial intelligence Aurora kinase-A Aurora kinase-B Acute lymphoblastic leukaemia Acute myeloid leukaemia AMP-dependent protein kinase Anaphase-promoting complex Androgen receptor Area under the concentration-time curve Bone morphogenic protein Benign prostatic hyperplasia Cyclin-dependent kinase Chronic lymphocytic leukaemia Chronic myeloid leukaemia Chronic myelomonocytic leukaemia Computerised tomography DNA damage response DNA damage checkpoint DNA methyltransferase 1 Double-strand DNA breaks Epstein-Barr virus Epidermal growth factor Epithelial-mesenchymal transition Focal adhesion kinase Guanosine 5′-diphosphate Gastric interstitial tumour Gross log kill G protein coupled receptor Oxidised glutathione Guanosine 5′-triphosphate xiii

xiv

HDAC HIV HPV HR HSC IC50 ILS IMPDH MCC MDS MHC MI MMEJ MRI MTD MTX NHEJ NSCLC PARP PCR PDGF PKC PK/PD PSA PLM RBC ROS RTK SAC SSB TCA cycle TCGA 6-MP 6-TG THU TKI TNFα TRAMP VEGF WBC

Abbreviations

Histone deacetylase Human immunodeficiency virus Human papilloma virus Homologous recombination Haematopoietic stem cells 50% inhibitory concentration Increase in lifespan (percent) Inosine 5′-phosphate dehydrogenase Mitotic checkpoint complex Myelodysplastic syndrome Major histocompatibility complex Mitotic index Microhomology-mediated end joining Magnetic resonance imaging Maximum tolerated dose Methotrexate Non-homologous end joining Non-small cell lung cancer Poly-ADP-ribose polymerase Polymerase chain reaction Platelet-derived growth factor Protein kinase C Pharmacokinetic/pharmacodynamic Prostate-specific antigen Percent labelled mitoses Red blood cells Reactive oxygen species Receptor tyrosine kinase Spindle assembly checkpoint Single-strand DNA breaks Tricarboxylic acid cycle The Cancer Genome Atlas 6-mercaptopurine 6-thioguanine Tetrahydrouridine Tyrosine kinase inhibitor Tumour necrosis factor alpha Transgenic adenocarcinoma of mouse prostate Vascular endothelial growth factor White blood cells

Chapter 1

Cancer as a Disease of Cell Proliferation

Abstract Evolutionary dynamics relates the growth and development of tumours to changes in rates of cell proliferation and cell death, and to their mutation rates. The cellular environment exerts selective pressure on the cellular variants in the tumour population. Tumours have multiple biochemical abnormalities, but the most consistent changes relate to increased cell division. Decreased spontaneous cell death is also often seen. Loss of differentiated tissue function is a frequent observation in advanced tumours. With rare exceptions, development of cancer is a multi-step process, explaining why it is primarily a disease of ageing, as the multiple mutations involved in tumour progression accumulate over time. Cancer is a progressive condition, such that developing tumours become increasingly independent of normal growth regulation. Oncogenes are dominant genes, involved in signal transduction pathways, that have mutated in malignant cells to become independent of their normal regulation. Tumour suppressor genes are recessive genes, often involved in signal transduction and apoptosis pathways, that carry loss-of-function mutations in tumour cells. Many anticancer drugs target the biochemical pathways of cell division. Such drugs are also toxic to normal proliferating cells. The enzymes of cell division do not necessarily differ between normal and malignant cells, though there are differences in their regulation. In some cases, the enzymes of DNA replication in tumour cells may carry mutations, but more often the mutations are in the signalling pathways and transcription factors that control the expression of the enzymes involved in celll division. Properties of cytotoxic drugs that influence their efficacy are their spectrum of activity against different tumours, their selectivity (i.e. activity against tumours compared with normal cells) and the incidence of acquired drug resistance. Evolutionary dynamics provides strategies for minimising or avoiding acquired resistance.

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_1. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_1

1

2

1

Cancer as a Disease of Cell Proliferation

Although there are many fields of biology that are essentially descriptive, with the application of information theory, theoretical biology can now take its place with theoretical physics without apology. Thus biology has become a quantitative and computational science. . . Hubert P. Yockey, Information Theory, Evolution, and the Origin of Life (2005)

1.1

What Is Evolutionary Dynamics?

Why is evolutionary dynamics important to an understanding of cancer? The subject originated as a quantitative description of evolutionary biology: can we provide a mathematical expression for how populations grow, change into new species, migrate into a new habitat, or become extinct? In particular, evolutionary dynamics describes how populations respond to selective pressures—Darwin’s “struggle for existence”, Spencer’s “survival of the fittest”. Cell biologists have adapted the techniques of evolutionary dynamics to describe the behaviour of the different cell populations within an organism, and in the present volume we focus on describing the events that occur when a normal cell transforms into a cancer cell, and the growth of that transformed cell into a life-threatening tumour. We shall explore the dynamics of cell division, of self-renewal and differentiation, and of apoptotic and non-apoptotic cell death. Selective pressures during tumour growth may arise from limitations in the availability of nutrients and oxygen, from the immune system of the host, and (in a tumour in a patient undergoing treatment) from the presence of drugs. An understanding of evolutionary dynamics can help us answer such quantitative questions as: do we need to kill every cancer cell, or is it sufficient to reduce the number of tumour cells to a tolerable level? Can early diagnostic data from asymptomatic individuals suggest if, or when, it is appropriate to start treatment? Is a course of treatment no longer providing a benefit—is it time to change the treatment? Two concepts from evolutionary biology have particular relevance to our understanding of cancer dynamics: natural selection, as an interaction of cellular variants with their environment, and chromosomal rearrangement, which in population biology can result in formation of new species, in cancer biology can mark the irreversible change from a normal to a transformed cell. In Chap. 2, we shall discuss two kinds of chromosomal rearrangement: translocation, which is often the cause of leukaemia or lymphoma, and aneuploidy, which is the cause of epithelial cell transformation. Evolutionary dynamics is an inherently mathematical subject, but many of its implications can be appreciated without needing to delve into the mathematical detail. This book can be read at two levels: the printed text covers the topic qualitatively. It asks how the changes that occur during the development of a tumour lead to the observed biological behaviour, and why it is important that they occur in a particular order. We shall consider the implications of the evolutionary dynamics of malignancy for the diagnosis, treatment, and prevention of cancer. The calculations, equations, and computer programs on which these conclusions are based are shown in the online supplement, so that those readers who want to understand, repeat, and

1.2

Cancer Enzymology

3

perhaps extend them can do so. Programs are written in the widely available (and free) R programming language (Ekstrøm 2012), and instructions for accessing the online supplement and for downloading R are found in the appendix. When we describe a complex system in mathematical terms—create a model—it is expected that the model will not only provide a concise description of the data on which it is based, but also that it will have predictive power: it should be able to describe the behaviour of the system in situations outside the original data. How good these predictions are likely to be will depend upon two things: first, the structure of the model—the relationship between its component parts, the various feedback loops, the sensitivity of its predicted output to changes in the input data— must accurately reflect the system being modelled. The complexity of biological systems is such that all mathematical models are necessarily simplifications. What degree of simplification is appropriate will depend upon the use to which the model is put, and ultimately reflects the skill and experience of the modeller. Second, the predictions of the model will only be as good as the input data. A model that is based upon data from a wide range of conditions is likely to be more reliable than a model of more limited input conditions. It is important when making decisions based upon the predictions of a computer model to be clear about the assumptions underlying that model.

1.2

Cancer Enzymology

It was evident from the earliest days of cancer research that malignancy was not only a phenomenon of excessive cell proliferation, but also of proliferation that was not subject to normal physiological constraints. Early biochemical studies attempted both to catalogue those abnormalities, and to determine how the normal growth constraints had been evaded. Harold Morris (1975) studied rat hepatomas induced by chemical carcinogens, and attempted, by comparing their biochemistry to that of normal liver, to identify the critical changes in those hepatomas that were least altered. These “minimal deviation hepatomas” had slow growth rates and retained much of the differentiated histology of normal liver. It was thus argued that biochemical differences between normal liver and minimal deviation hepatomas must be critical to the process of neoplastic transformation. George Weber (1977) examined a spectrum of rat hepatomas, differing in growth rate and degree of differentiation, and distinguished between three classes of biochemical change: first, there were changes in enzyme activity that correlated (either positively or negatively) with tumour growth rate. Weber argued that such changes were related to cell proliferation, rather than to neoplasia per se. He showed that similar changes were usually seen in regenerating or neonatal normal liver. Secondly, there was a group of biochemical changes that were seen in all hepatomas, regardless of growth rate. It was claimed that these changes must be closely linked to the transforming process, since they were not generally found in regenerating or neonatal liver. Finally, a third group of enzymes appeared to vary randomly across the hepatoma spectrum, without

4

1 Cancer as a Disease of Cell Proliferation

showing consistent correlation with either growth rate or malignancy. This third group foreshadowed the observation of “passenger mutations” to be discussed later. Weber’s analysis was extended to tumours other than hepatomas, and in general these conclusions appeared to apply across several tumour types. These studies accumulated a wealth of data, but were unable to identify the critical changes that triggered malignant transformation. An example of the proliferation-linked enzymes is inosine 5'-phosphate dehydrogenase (IMPDH), whose activity across a panel of eleven hepatomas was elevated at least 2.5-fold, with the increase reaching 13-fold in the most rapidly proliferating hepatomas (Jackson et al. 1975). In regenerating normal liver the elevation was fivefold. IMPDH was also elevated in foetal and neonatal liver (Weber et al. 1976). IMPDH was increased in other tumours, including renal cell carcinomas (Weber et al. 1978). The significance of IMPDH is that it is the rate-determining enzyme in the pathway of guanosine 5′-diphosphate (GDP) and guanosine 5′-triphosphate (GTP) biosynthesis. GDP is required for biosynthesis of DNA, and GTP for both RNA and protein. Importantly, GTP is also required for biosynthesis of adenylate nucleotides, so IMPDH appears to be a major control point for production of both branches of the purine nucleotide biosynthetic pathway, adenylates and guanylates. The purine catabolic enzyme, xanthine oxidase, had decreased activity, relative to normal liver, in a panel of twelve rat hepatomas. The decrease showed a rough negative correlation with growth rate, and was 10%, or less, of the normal liver value in the fastest growing tumours. In general, a negative correlation with tumour growth rate was seen with catabolic enzymes, and for enzymes of differentiated liver function (e.g. gluconeogenesis). In agreement with the earlier claims of Warburg, Morris and Weber found that all tumours had increased glycolysis (relative to normal liver) and decreased oxidative phosphorylation. An example of an enzyme that was elevated in hepatomas, relative to normal liver, is PRPP amidotransferase (Weber et al. 1978). It was increased two to threefold in the hepatoma panel, but did not correlate with tumour growth rate. It showed a modest (~50%) increase in regenerating normal liver. PRPP amidotransferase is probably not normally the rate-limiting enzyme in its pathway (that is likely to be the preceding enzyme, PRPP synthetase), and it is not clear what the regulatory significance is of these small changes. Weber’s conclusion that systematic changes are seen in cancer for “key enzymes” raises the question of what is meant by “key enzyme”? Presumably all the enzymes in the cell are essential, since they would otherwise be eliminated by natural selection. The key enzymes, then, are those whose expression level makes them rate-determining for their pathway. In summary, the work of Weber and Morris and their collaborators has direct relevance for our understanding of the pathways involved in cell proliferation. Morris’s “minimal deviation” concept made the important point that there is not a continuous spectrum of malignancy, ranging from completely normal, through “a little bit malignant” to “highly malignant”. Rather, there is a discontinuity: either a cell is untransformed or it is transformed, and all transformed cells, regardless of how slow growing or differentiated they may be, have the potential to progress to advanced, life-threatening tumours. Weber’s class of “transformation-linked”

1.3

Oncogenes and Tumour Suppressor Genes

5

enzymes reflects this insight, though he was not successful in identifying what those key changes are. Those changes that Weber described as transformation-linked are probably related to events that occur in tumour progression, such as de-differentiation, rather than the earliest changes in neoplastic transformation. Some of the enzymes that Weber classified as transformation linked are, in fact, enzymes in biosynthetic pathways, but not the rate-limiting enzymes in those pathways (e.g. PRPP amidotransferase). Activity of the purine biosynthetic pathway is increased in all tumours, but the regulation of the pathway is such that only the rate-limiting enzyme (PRPP synthetase in this instance) is correlated with growth rate. The search for a single biochemical change that resulted in neoplastic transformation was doomed to fail, because (according to our present understanding) transformation requires, minimally, two events: chromosomal or genetic instability (which will be discussed in Chap. 2), and a genetic or chromosomal change that confers a growth advantage.

1.3

Oncogenes and Tumour Suppressor Genes

An alternative approach to identifying the trigger that makes a normal cell cancerous emerged from studies with transforming viruses (Minden 1987). Several RNA viruses were shown to carry genetic information (oncogenes) that when inserted into a mammalian or avian genome (by reverse transcription) directly caused transformation (rapidly transforming viruses). Another group of viruses (slowly transforming viruses), after being inserted next to a cellular gene (a proto-oncogene) altered its transcription in a way that circumvented the normal regulatory mechanisms of the cell. These studies used in vitro transformation of normal fibroblasts. Bishop (1983) and Varmus (1984) suggested that cellular proto-oncogenes were probably normal growth promoters that were tightly regulated, but that this regulation could be over-ridden by viral infection. The rapidly transforming virus then originated by acquisition of cellular oncogenes, which after loss of non-coding regions became viral oncogenes that were capable of transforming normal cells. By the mid-1980s more than twenty proto-oncogenes were known, and attention focussed on determining their biochemical activity. Tony Hunter and his associates, working with polyoma virus, showed that the src oncogene had protein tyrosine kinase activity (Hunter and Cooper 1985). Several other oncogenes were subsequently shown to be protein kinases, some membrane-bound (often tyrosine kinases), others cytosolic, which could be either tyrosine kinases or serine/threonine kinases. Other oncogenes had GTPase activity (Minden 1987). Early oncogene studies suggested that they fell into two classes, or complementation groups (Tannock and Hill 1987). Some oncogenes, such as myc, were able to immortalise cells, without causing neoplastic transformation. Other oncogenes caused morphological transformation, but did not in themselves cause immortalisation. Complete neoplastic transformation required that a cell be transfected with one oncogene from each group. We shall return to the current

6

1 Cancer as a Disease of Cell Proliferation

interpretation of this observation in our discussion of the two-stage theory of carcinogenesis, in Chap. 3. The observation (discussed in Chap. 2) that Burkitt’s lymphoma resulted from abnormal activation of the oncogene, c-myc, raised expectations that cancers in general might be caused by abnormal expression of a single oncogene, but subsequent research indicates that Burkitt’s tumour is an outlier in this respect. In contrast to oncogenes, tumour suppressor genes are recessive genes whose loss of function results in hyperproliferation. For this to happen, it is necessary that both alleles be inactive, i.e. both copies of the tumour suppressor gene must have a lossof-function mutation. An important example is Apc, whose function is defective in about 85% of colon cancers. Studies in mice showed that deletion of Apc caused polyposis. These polyps were benign, but are considered premalignant as subsequent mutations in Kras and p53 resulted in invasive carcinoma (Dow et al. 2015). The price a tumour cell pays for becoming autonomous may be enhanced vulnerability to particular inhibitors, a phenomenon sometimes described as oncogene addiction (Joe and Weinstein 2011). An example is melanoma, which often has constitutively activated B-RAF (a component of the MAP kinase pathway, discussed below). This makes the cells able to proliferate independently of external growth factors, but highly sensitive to the B-RAF inhibitor sorafenib.

1.4

Signalling Pathways

An understanding of signalling pathways, also known as signal transduction pathways, is important in cancer dynamics because their control is invariably altered in cancer. Oncogenes and tumour suppressor genes appear to act on growth factors, or their receptors, or the signalling pathways that communicate between activated receptors and the transcription factors that activate expression of groups of genes. Of the many signalling pathways in the cell, six are particularly important for an understanding of cancer: the mitogen-activated protein kinase (MAPK) pathway, the PI3 kinase pathway, the Wnt signalling pathway, integrin signalling, cytokine signalling, and protein kinase C signalling (Nelson 2008; Hancock 2010). Certain aspects of signalling dynamics only emerge when multiple signalling pathways operate within a single cell—crosstalk, potentiation or antagonism, multistability. These pathways overlap and interact in multiple, complex ways, but all act as communication channels between receptors and transcription factors, all involve protein phosphorylation events, and, more generally, are composed of certain functional components described below. Despite the overlap, the MAP kinase pathway may be considered as primarily involved in cell division, the PI3 kinase pathway in regulation of metabolic activity, and Wnt signalling in morphogenesis. Cytokine signalling is of particular importance in control of the immune system. Before considering examples of particular signalling pathways, it is important to understand the functional components common to all the pathways (Aguda 2001). The function of a signalling pathway is to carry a message from a hormone or growth

1.4

Signalling Pathways

7

factor, the ligand, to the DNA. This involves, first, a receptor. Some receptors are intracellular: for example, the androgen receptor (AR) is in the cytosol. Androgens are lipophilic, and able to diffuse across the cell membrane. When the AR binds its ligand, it migrates to the nucleus. Many hormones and growth factors are peptides, and unable to cross the outer cell membrane. Their receptors are large proteins that span the cell membrane. They have an extracellular domain, which binds the ligand and an intracellular domain, that has catalytic activity, often a protein kinase, which is activated when the extracellular domain binds its ligand. The signal is often weak and transient, and a vital function of signalling pathways is to amplify the signal both in amplitude and in duration. Signal amplification is often achieved through a kinase cascade, in which kinase 1 phosphorylates and activates kinase 2, etc. Because each stage is catalytic, even a 2-stage kinase cascade can act as a high-gain amplifier. Another important component of signalling pathways is proteins that expand the signal in time, rather than amplitude. The most important family of these pulse expanders is the G-proteins. Many G proteins exist in close association with receptors that have seven transmembrane domains, the so-called G protein-coupled receptors (GPCRs). The small G proteins are a family of smaller (21 kD) proteins, including the three members of the Ras family. G proteins are bistable: when they bind GTP, they are active. When their downstream substrate (Raf, in the case of Ras) is activated, the bound GTP is hydrolysed to GDP. G proteins with bound GDP are inactive. The bound GDP dissociates extremely slowly. In fact, by analogy with electronic circuits, G proteins, although existing as active and inactive forms, are not really bistable, but can be considered as a monostable switch, which is normally OFF. However, another family of proteins, the guanine nucleotide exchange factors, stimulates the release of GDP from G proteins. In the case of the MAP kinase pathway (Fig. 1.1), the exchange factor is GRB2 (in mammalian cells; the corresponding protein is Drosophila is SOS). When the GDP is released, the G protein is free to bind another guanine nucleotide. Since the cellular concentration of GTP is about ten-fold higher than that of GDP, the net effect of a guanine nucleotide exchange factor is to replace most of the bound GDP with GTP, and thus reactivate the G protein. The pulse-expander role of Ras was demonstrated by a computer model of the MAP kinase pathway in which a five minute exposure to EGF resulted in elevation of cyclin D levels for over two hours (Jackson 2017). Signalling pathways act as logic gates. In digital electronic circuits, logic gates are circuits that provide outputs that are conditional upon particular combinations of inputs. An OR gate has an output that is OFF unless any of a number of inputs is active. An important example of OR gates in signal transduction is where a common signalling pathway can be driven by a number of receptors. The PI3 kinase pathway is shown in Fig. 1.2 as driven by PDGF, but it can also be activated by insulin, insulin-like growth factor 1 (IGF-1) and vascular endothelial growth factor (VEGF). In PC12 cells, transient activation of ERK by EGF triggers cell proliferation, but sustained activation by NGF triggers neuronal differentiation (Brightman and Fell 2000; Vaudry et al. 2002). AND gates provide for convergent branching. The output of an AND gate is only ON if both input signals are active. In biology, some genes should only be

8

1

Cancer as a Disease of Cell Proliferation

EGF

extracellular

EGFR

membrane

+

Shc

GRB2/SOS +

membrane/cytosol

+

PI3K

raf

+

PKC

MEK

cytosol cytosol/nucleus

ras

PP2A –

nucleus

ERK

+

elF-4E

c-fos cyclin D E2F

Fig. 1.1 The MAP kinase signalling pathway Fig. 1.2 The PI3 kinase signalling pathway

transcribed when two or more conditions are met. An example from developmental biology is that many cells can only enter cell division if they are in the right place. For these cells, cell division requires both a growth signal (from a growth factor receptor) AND an attachment signal (e.g. from an integrin). Multiplexers are components of electronic or signalling circuits where a single input can drive multiple outputs. They thus represent divergent branch points. A dozen or so cell signalling pathways determine the activity of a thousand or more transcription factors. It follows that the pathways are highly branched. The same pathways may activate different transcriptional networks in different cell types, or in the same cells at different stages of their life cycle. Which branches are active under various conditions are determined by a complex system of feedback effects.

1.4

Signalling Pathways

9

For the subset of signalling pathways that drive gene expression, the final stage is activation of one or more transcription factors, proteins that bind to specific regulator genes and turn on the synthesis of messenger RNA (Latchman 2008). In the case of the MAP kinase pathway the earliest transcription factor to be activated is c-Fos. Fos drives transcription of one or more of the six members of the E2F family of transcription factors. E2F in turn activates c-myc, which is responsible for transcription of the enzymes required for the cell to enter S phase of the cell cycle, such as ribonucleotide reductase. Like c-Fos, the transcription factor Jun is activated by ERK; Jun is also activated by integrin signalling. Jun and Fos combine to form the transcription factor AP1, which controls yet another group of genes. An important group of transcription factors that we shall discuss in connection with cytokine signalling are the STATs. Some oncogenes are mutated transcription factors. Constitutive activation of myc can result in unregulated cell division. Conversely, loss-of-function mutations in p53 result in a non-functional G1:S checkpoint. About 50% of human cancers have p53 mutations. Most of the components of signalling pathways have been explored as targets for anticancer drugs. Antibodies have been developed that target and inactivate the extracellular domains of growth factor receptors. An example is trastuzumab (Herceptin) that blocks the HER2 receptor, a member of the EGF receptor family that is expressed at high levels by many tumours. An alternative strategy is to inhibit the intracellular tyrosine kinase domain, thus blocking downstream signalling. This approach means that small molecule inhibitors can be used, which have pharmacokinetic advantages over antibodies (e.g. they may be orally active, rather than requiring intravenous injection). However, there are many tyrosine kinases in the cell, so obtaining a high degree of selectivity against a particular receptor kinase can be difficult. Other kinase inhibitors have been developed that are selective for the serine/threonine kinases that are involved in cell cycle control. Yet other kinase inhibitors (e.g. sorafenib) are active against both tyrosine and serine/threonine kinases. Inhibitors targeted against pulse expanders, such as K-ras, have been harder to develop, partly because their very high affinity for GDP makes it difficult to dislodge from its binding site, Sotorasib targets the G12C mutation in K-ras, forming a covalent bond with the cysteine; it is without effect on wild-type Ras. It has activity against non-small cell lung cancer. Before considering examples of particular signalling pathways, it is important to understand the functional components common to all the pathways. The function of a signalling pathway is to carry a message from a hormone or growth factor, the ligand, to the DNA. This involves, first, a receptor. Some receptors are intracellular: for example, the androgen receptor (AR) is in the cytosol. Androgens are lipophilic and able to diffuse across the cell membrane. When the AR binds its ligand, it migrates to the nucleus. Many hormones and growth factors are peptides and unable to cross the outer cell membrane. Their receptors are large proteins that span the cell membrane. They have an extracellular domain, which binds the ligand and an intracellular domain, that has catalytic activity, often a protein kinase, which is activated when the extracellular domain binds its ligand. As discussed above, a

10

1

Cancer as a Disease of Cell Proliferation

Fig. 1.3 The Wnt signalling pathway

Wnt

–

sFRP

Frizzled Dv1 –

GSK-3β Tcf

β-catenin

cyclin D LEF-1

vital function of signalling pathways is to amplify the signal both in amplitude and in duration. Signal amplification is often achieved through a kinase cascade, in which kinase 1 phosphorylates and activates kinase 2, etc. resulting in a high-gain amplifier. The PI3 kinase pathway (Fig. 1.2) is involved in regulation of a variety of metabolic processes, of which one of the most significant is protein synthesis. This pathway is activated by a wide range of ligands, one of which is platelet-derived growth factor (PDGF) an important driver of wound healing. The receptor for vascular endothelial growth factor (VEGF)—another growth factor involved in wound healing—also signals through this pathway. As with the MAP kinase pathway, a brief exposure to growth factor causes prolonged activation. This is probably attributable to the fact that PDGFR and VEGFR are G-protein coupled receptors. PI3K can also be activated by Ras, and in turn it can activate Grb2, an autocatalytic loop. The Wnt pathway is a developmental pathway, primarily involved in morphogenesis. It is reactivated in 85% of colon cancers. This reactivation is often associated with loss of the tumour suppressor, Apc, and restoration of Apc was shown to restore normal crypt organisation (Dow et al. 2015). Binding of the Wnt ligand to a receptor of the frizzled family activates a cytosolic protein, Dv1, which inhibits GSK-3. This decreases phosphorylation of β-catenin (slowing its proteosomemediated degradation) which otherwise translocates to the nucleus and stimulates transcription through Tcf/LEF transcription factors. Cyclin D2 and Jun are among the targets of one of these transcription factors, LEF-1 (Fig. 1.3). Integrin signalling provides an example of a signalling pathway that acts as an AND gate (Fig. 1.4). Epithelial cells normally require both a growth signal (e.g. PDGF) AND an attachment signal for activation of focal adhesion kinase (FAK). Epithelial tissues normally grow on a basement membrane, and cells that are not attached will die. If the integrin signal in a tumour cell becomes constitutively active (active in the absence of attachment to the basement membrane), the tumour becomes invasive and grows as a solid lump, rather than a membrane-attached sheet. This is a critical event in the process of tumour progression.

1.4

Signalling Pathways

Fig. 1.4 An example of integrin signalling

11 PDGF

integrin (e.g. αVβ3)

Shc FAK

–

PTEN

PI3K PKB/Akt – GSK-3β – Fos

Jun

+

+ AP1 cyclin D1

Fig. 1.5 Cytokine signalling

Cells of the immune system are triggered into proliferation when they recognise a non-self antigen. The immune system is coordinated by a network of activators and inhibitors that make it the most complex of the body’s organ systems, with the exception of the nervous system. Like the nervous system, the immune system is capable of learning, memory, and self/non-self recognition. A group of regulatory proteins, the cytokines (including the interferons and interleukins) bind to receptors in the membrane of B cells and T cells. Downstream signalling from cytokine receptors involves a family of tyrosine kinases, the Janus-acting kinases (JAKS), which phosphorylate and activate members of the family of signal transducer and activator of transcription (STAT) family (Fig. 1.5). There are four JAKS, though the picture is complicated by the fact that activation of JAKS causes them to dimerise, and heterodimers are possible. JAK2 is commonly mutated, and constitutively activated (i.e. activated in the absence of a receptor signal) in many leukaemias

12

1

Cancer as a Disease of Cell Proliferation

Fig. 1.6 Some reactions of PKC signalling

and lymphomas. There are seven members of the STAT family. We shall discuss the role of STAT3 and STAT5 in chronic myeloid leukaemia (CML) in Chap. 7. Other signalling pathways operate primarily in cells of the immune system: the NFAT and the NfκB pathway are examples, and are sometimes abnormally expressed in leukaemias or lymphomas. Both NFAT and NfκB act as transcription factors in cells of the immune system. NFAT is activated by dephosphorylation, by the protein phosphatase, calcineurin (also known as PP2B0 which is in turn activated by calcium ions and calmodulin. Calcineurin is the target for the transplant rejection drug, cyclosporin. Protein kinase C (PKC) is a family of serine/threonine kinases that appear to be primarily involved in the regulation of inflammatory responses (Gomperts et al. 2003). Their involvement in carcinogenesis has been suggested partly because PKC signalling (Fig. 1.6) interacts with Raf and Ras (by activation of the guanine nucleotide exchange factor Ras-GRP), and partly because the carcinogenic phorbol esters appear to act primarily by activation of PKC. PKC family members are also activated by diacylglycerol (DAG) produced by the action of phospholipase C, itself activated by a number of GPCRs. A subset of human thyroid tumours has been shown to express a mutated PKCα (Prevostel et al. 1995).

1.5

Signalling Dynamics: ON/OFF Switches and Volume Controls

Signalling pathways differ from metabolic pathways in that they are processing information, rather than material. All signal processing systems necessarily have a material component, but to understand their function requires that we focus on the signal, rather than on the chemistry of the communication channel. We could not have a piano concerto without a piano, but to understand what is unique about Rachmaninoff’s Second, it would not be helpful to take apart a concert grand. For the most part, signalling pathways involve proteins and lipids, or (at the final level of transcription factors) nucleic acids. In signalling pathways, the essential information is conveyed by the instantaneous signal level, whereas for biosynthetic or catabolic biochemical pathways, what matters is the total amount of material produced or

1.5

Signalling Dynamics: ON/OFF Switches and Volume Controls 0.09

13

cyclin D

0.08 0.07

activity

0.06 0.05 0.04 0.03 0.02 0.01 0 0.001

0.01

0.1

1

10

[EGF]

Fig. 1.7 Calculated dose–response curve for activation of the MAP kinase signalling pathway by a 20-minute pulse of EGF (reproduced from Jackson 2017)

degraded—the integrated signal. Protein kinases and phosphatases are prominent in signalling pathways because the transfer of a phosphate group, adding or removing a strong electrical charge, results in a conformational change in the donor or recipient that may activate or inactivate its catalytic function. The multiple positive and negative feedback loops involved in signalling pathways make it difficult to predict their quantitative behaviour, and this has been explored using computer models (Brightman and Fell 2000; Gilbert et al. 2006; Jackson 2017). A model of the MAP kinase pathway, based on Fig. 1.1 predicted that a brief extracellular pulse of EGF resulted in raised cyclin D levels for several hours (online appendix). According to this model, the system shows Boolean (all or none) kinetics, i.e. it acts as an ON/OFF switch. EGF levels less than 37pM had no effect, while concentrations of 50pM and above resulted in complete activation (Fig. 1.7). The modelling suggested that the steep saturation curve of the MAP kinase pathway was a consequence of the three-stage amplifier. Boolean kinetics in signalling pathways has also been described by Veliz-Cuba and Stigler (2011) and by Jenkins and Macauley (2017). Unlike the MAP kinase pathway, a model of the PI3 kinase/Akt pathway based on Fig. 1.2 showed conventional saturation (Michaelis–Menten) kinetics, i.e. it acted as a volume control rather than an ON/OFF switch (Fig. 1.8). Why the difference? In the case of the MAP kinase pathway, which primarily drives DNA synthesis, DNA must either be made in a fixed amount (to double the existing DNA content) or not at all. By contrast, the PI3K/Akt pathway primarily regulates protein synthesis, and protein may be required in varying amounts, depending on the nutritional status of the cell, and how large the cell is intended to be.

14

1 3.5

Cancer as a Disease of Cell Proliferation

protein

3

protein synthesis

2.5 2 1.5 1 0.5 0 0.01

0.1

0

10

100

[PDGF]

Fig. 1.8 Calculated dose–response curve for activation of PI3 kinase signalling by a 20-minute pulse of PDGF (reproduced from Jackson 2017)

Growth signals can be negative as well as positive. Early studies of growth regulation, carried out largely with cultures of mouse embryo fibroblasts, reported an effect described as contact inhibition, in which cells stopped dividing when they were completely surrounded by other cells. These cells, when grown in glass or plastic Petri dishes require attachment for growth, and when the growth surface is completely covered, growth stops. Although this is an artificial system, it probably reflects the in vivo behaviour of epithelial cells, which normally grow on a basement membrane formed of collagen, laminin, and other proteins or glycoproteins. Figure 1.4 shows an example of integrin signalling where, to divide, the cell requires both a growth signal, from a growth factor, and an attachment signal, from an integrin or cadherin. The absence of the attachment signal results in cessation of cell division. Other negative growth regulators interact directly with the signalling pathways. As shown in Fig. 1.4, the tumour suppressor protein, PTEN, opposes the activation of FAK. FAK is activated by phosphorylation, and PTEN, which is a protein phosphatase, removes the activating phosphate group from FAK.

1.6

Dynamics of Tumour Growth

One of the earliest aspects of cancer dynamics to be studied was the growth curve. Early-stage epithelial tumours usually grow as a thin sheet of cells, attached to a basement membrane. This growth pattern, described as surface growth results in the

1.6

Dynamics of Tumour Growth

15

1e+010

Nt

9e+010 8e+010

population

7e+010 6e+010 5e+010 4e+010 3e+010 2e+010 1e+010 0

0

108.281, -1.42105e+010

50

100

150

200

250

300

350

time (days)

Fig. 1.9 Example of a Gompertzian growth curve. Plotted by Gompertz.R with parameters N0 = 1, Ninf = 1.0e+11, k = 0.34657/day

diameter of the tumour increasing as a linear function of time (Wodarz and Komarova 2014). Once tumour cells become independent of attachment to extracellular matrix, they grow as a three-dimensional lump. In the earliest stages of solid tumour growth, it may be approximately exponential. However, exponential growth can never continue for very long; nutrient depletion, accumulation of toxic waste products, or physical constraints will result in slowing down of growth, and its eventual cessation. More realistic expressions of tumour growth have been described (Jackson 1992). A widely used description of tumour growth used by biologists is the Gompertz equation (Fig. 1.9). Growth that is initially almost exponential reaches a maximum (at the inflexion point) then decelerates, and eventually levels off (in mathematical terms, reaches an asymptote). Untreated human tumours normally become lethal before the asymptote is reached. In the example shown in Fig. 1.9, a single transformed cell is projected to reach its asymptotic size at about 13 years, though it would probably reach a lethal size of 1012 cells by four years. Equations for the Gompertz growth curve, and a number of alternative models are provided in the online supplement. These growth curves are continuous functions, but tumour growth may show discontinuities as a result of mutation (West et al. 2016). Human breast cancer has been shown to often follow a Gompertzian function (Norton 1988), but Speer et al. (1984) reported that breast cancers, while basically following Gompertzian growth, sometimes showed random, spontaneous changes in the growth parameters. This form of growth curve, described by evolutionary biologists as “punctuated”, has implications for the

16

1

Cancer as a Disease of Cell Proliferation

Fig. 1.10 Growth curve for the May equation with growth parameter = 1.0

optimal treatment of breast cancer, because Norton and Simon (1986) predicted that the optimal way to treat a large tumour with a Gompertzian growth curve was induction treatment followed by intensive consolidation treatment, whereas for the Speer model it was predicted that prolonged low dose treatment would be advantageous. Growth curves reflect the balance between cell division, cell stasis (which may result from reversible entry into a G0 state, or from irreversible senescence or terminal differentiation), and cell death. In evolutionary dynamics calculations, selective pressure from resource depletion may affect both cell growth and cell death. Additional causes of selection pressure may result from the tumour microenvironment, e.g. hydrostatic pressure, or hypoxia caused by from a tumour outgrowing its blood supply. The growth equations discussed so far assume that the population converges on a maximum value, the asymptote. In actual biological populations this is not necessarily the case. A rapidly growing population may temporarily exceed the carrying capacity of its environment, in which case the approach to a maximum is followed by a collapse. Such a system is described by the discrete logistic equation of May (discussed by Murray 1993). Depending upon the growth parameters, the May equation may be a smooth sigmoidal (S-shaped) curve, as in Fig. 1.10, or it may end in stable or unstable oscillations, or in chaos (Fig. 1.11). In biology, a population that outgrows the carrying capacity of its habitat will collapse because of starvation, disease, or environmental destruction, recovering later when the population has declined to a sustainable level. The equation, in describing population collapse, provides a quantitative description of the prophet Ezekiel’s four horsemen of the

1.7

The Hallmarks of Cancer

17

140

N

120

POPULATION

100 80 60 40 20 0

0

2

4

6

-4.36047, -19.8947

8

10 12 TIME (years)

14

16

18

20

Fig. 1.11 Growth curve for the May equation with growth parameter = 2.6

apocalypse—war, famine, wild beasts, and plague. In cancer biology, the discrete logistic equation could describe the growth of a tumour that outstrips its blood supply, resulting in necrosis, but which then resumes growth when it has shrunk to a size that is again sustainable.

1.7

The Hallmarks of Cancer

Hannahan and Weinberg (2000) provided a concise summary of the state of cancer biology in the final year of the twentieth century. They listed six essential alterations in cell physiology that dictate malignant growth: self-sufficiency in growth signals; insensitivity to growth-inhibitory (anti-growth) signals; evasion of programmed cell death (apoptosis); limitless replicative potential; sustained angiogenesis; and tissue invasion and metastasis. Of these factors, positive and negative growth signalling, and the tendency of tumours to become independent of these signals, have been discussed above. Apoptosis (formerly known as programmed cell death) is involved in normal development; normal function of immune cells; and quality control of cells with damaged DNA. It is driven by a two-stage cascade of proteolytic enzymes, caspases. Detailed kinetic models of apoptosis have been described (Fusenegger et al. 2000; Bentele et al. 2004; Hua et al. 2005) and a simplified depiction of a typical caspase pathway is shown in Fig. 1.13. In dividing cells, there is a very low baseline level of

1

Cancer as a Disease of Cell Proliferation

0.0e+00

0.5e+09

count 1.0e+10

1.5e+10

2.0e+10

18

0

50

100 time(days)

150

200

Fig. 1.12 Punctuated growth curve with a break point at 50 days, plotted by the program speer.R (Chap. 1 supplement)

active caspase 3, maintained by a low background level of activation of its inactive precursor, procaspase 3. A combination of pro-apoptotic proteins, such as BAX, and anti-apoptotic proteins (e.g. Mcl-1, Bcl2) maintains a delicate balance. Certain markers of cellular trauma trigger activation of procaspase 9, for example, the appearance of cytochrome C in cytosol; cytochrome C is normally confined to mitochondria. Active caspase 9 now catalyses activation of procaspase 3 to caspase 3. Up to this point, the process is reversible. When the level of caspase 3 passes a critical concentration, irreversible cell damage ensues. The dynamics of apoptosis are such that it takes several hours to reach the point of no return. In the DNA damage checkpoint (to be discussed in Chap. 4) DNA damage triggers the caspase pathway, but cell death only occurs if the DNA damage has not been repaired within 6 to 10 hours (depending on the cell type). The delayed action dynamics of the caspase pathway are an essential feature of the checkpoint: if the cell can repair the damage, it does, but rather than let a genetically damaged cell go into cell division, it is killed. In cancer, where the checkpoint is often defective, the proliferation of genetically altered cells contributes to the phenomenon of tumour progression. The pathway described in Fig. 1.13 is the intrinsic pathway of apoptosis. Apoptosis can also be triggered by extracellular events, through the extrinsic pathway (Hancock 2010). Some cells in the body are destined to carry out their function, then die. These cells express death receptors, a subgroup of the TNF receptor family, on

1.7

The Hallmarks of Cancer

Fig. 1.13 Control of caspase-dependent apoptosis. R1 indicates transcriptional activation

19

cytochrome C

survivin

+

+ procaspase 9

caspase 9 +

procaspase 3

+ caspase 3

+

+ Mcl-1

BAX +

+

r1

r1 apoptosis

their outer membrane. The intracellular domain of death receptors reacts with, and activates, initiator procaspases. An alternative route of cell death, necrosis, appears to be driven by hypoxia. Cells are able to withstand transient hypoxia: they go into stasis, i.e. do not progress in the cell cycle; this is probably because when oxidative phosphorylation ceases, ATP levels drop and they are unable to make the DNA precursors without which the cell cannot enter S phase. If hypoxia lasts too long, the ATP depletion means that the cell cannot carry out essential metabolism, and the cell dies. Solid tumours in which growth has outpaced new blood vessel formation (angiogenesis) frequently have a necrotic core. Other cell death pathways exist: autophagy (“self-devouring”) is a stress response in which non-essential cells may self-destruct to conserve energy (e.g. in conditions of starvation). It has been suggested that autophagy may sometimes play a role in cancer prevention (Li et al. 2020). Mitotic catastrophe is a cell death pathway resulting from inaccurate chromosome segregation in mitosis. It will be discussed in Chap. 5. Angiogenesis, invasion, and metastasis are events that occur after the initial transformation process, they are part of tumour progression, and will be discussed in Chap. 9. The other hallmark of cancer listed by Hanahan and Weinberg, immortalisation, is actually a property of germ cells, rather than being peculiar to cancer. Somatic cells (i.e. cells of the body other than the reproductive germ cells that produce ova and spermatozoa) have limited reproductive potential (Shay and Wright 2000).

20

1

Cancer as a Disease of Cell Proliferation

The molecular basis for this “Hayflick limit” is telomere shortening. Telomeres are non-transcribed regions at each end of chromosomes that become shorter after each successive cell division. Telomeres contain no genes, but consist of hundreds of repeats of the sequence TTAGGG. When the telomere becomes so short that the DNA can no longer replicate, the cell is said to be senescent. Though incapable of further cell division, the cell remains metabolically active and able to carry out its specialised functions. Germ cells, unlike normal somatic cells, express an enzyme, telomerase, that is able to lengthen their telomeres. Tumour cells often express telomerase, and are thus potentially immortal, and in the laboratory human tumour cell lines have been maintained for thousands of cell divisions. Hanahan and Weinberg’s landmark paper provided a valuable systematisation of the confusingly large literature on the biochemical changes found in tumour cells. It is a kind of descriptive natural history of cancer cells. The task of evolutionary dynamics is more mechanistic: it is to describe how the tumour cells acquired those changes, and to consider which of the changes are essential, and which are inevitable consequences of preceding events. The two decades since Hanahan and Weinberg’s review have added a few additional “hallmarks” which we shall consider later, but in general their analysis has held up well, and their argument provides a useful starting point for the present analysis. Early studies of tumour biochemistry noted that carcinogenesis was a multi-stage process, that certain biological motifs were invariably involved (essentially Hanahan and Weinberg’s six hallmarks), and that each of these could be brought about by any one of several alternative mechanisms. It was thus suggested that cancer, like a six-course meal, could be the result of a sequence of events involving one mutation that caused enhanced proliferation, one mutation that prevented apoptosis, one that caused invasiveness, another that resulted in metastasis, resulting finally in a fully transformed tumour cell (the “restaurant menu” theory of cancer). Vogelstein and his collaborators (1993), based upon studies of colon cancer, whose development often follows a progression from benign polyp, to localised carcinoma, to metastatic disease postulated a similar multi-step origin of cancer. This view is substantially correct, but is not necessarily the simplest interpretation of the data, and fails to capture the characteristic dynamics of malignant transformation. In particular, although Hanahan and Weinberg described genetic instability as an “enabling factor”, this does not make clear the central role of genetic/chromosomal instability as the defining characteristic of malignancy.

1.8

Methotrexate: An Antiproliferative Drug

Cancer being clearly a disease of cell proliferation, early anticancer drugs targeted cell proliferation. Since at that time the differences in control of cell division between normal and tumour cells were unknown, these early antitumour drugs inhibited cell division both in malignant and normal cells. For this reason, they are toxic to proliferating normal tissues, including bone marrow, gastrointestinal epithelium,

1.8

Methotrexate: An Antiproliferative Drug

21

Fig. 1.14 Methotrexate, an antifolate

and buccal mucosa. Methotrexate (Fig. 1.14) is an early example of the class of anticancer agents termed antimetabolites, which have close structural resemblance to normal metabolites, in this case the vitamin folic acid (Pratt and Ruddon 1979). The molecular target of methotrexate is the enzyme dihydrofolate reductase, which is required for the biosynthesis of thymidylate, needed for synthesis of DNA, and also for biosynthesis of the purine nucleotide inosine 5'-phosphate (IMP) required for synthesis of both DNA and RNA. The pharmacology of methotrexate illustrates many of the characteristic features of cytotoxic drugs. It has a limited spectrum of activity. Many years of clinical trials have shown that cytotoxic drugs, on average, will give responses in about 20% of cancer patients. Of course, the 20% that respond to drug A might not be the 20% that respond to drug B, and the details of which tumour will respond to which drug will depend upon the detailed cellular pharmacology of the drug. In the case of methotrexate, which is an anionic molecule, it requires active or facilitated transport across the cell membrane. Some tumours have high levels of the required carrier molecules, other do not. Once inside the cell, methotrexate is conjugated to a high molecular weight polyglutamate derivative, which traps it inside the cell. In cells that are unable to form polyglutamates, the methotrexate will rapidly leak out, before it has a chance to cause lethal damage. The selectivity of anticancer drugs is a measure of their effectiveness against tumour cells compared to their activity against susceptible normal cells (in the case of methotrexate, bone marrow stem cells or cells of the intestinal epithelium). A compound that was just as toxic to normal cells as to tumour cells would not be a drug, it would be a poison. Drug selectivity is often expressed as therapeutic index. In animals or human patients, therapeutic index is defined as the maximum tolerated dose divided by the minimum effective dose. Acquired resistance to drugs is the situation where the disease initially responds, but later, as a result of mutations, becomes less sensitive, or completely insensitive. Mutations that cause resistance to methotrexate may result in loss of membrane transport, or loss of the ability to form polyglutamates, or changes in the active site of dihydrofolate reductase that result in less potent binding of methotrexate. A frequent cause of methotrexate resistance is overproduction of the target enzyme, dihydrofolate reductase, sometimes by more than one hundred-fold. Because the binding of methotrexate by dihydrofolate reductase is almost stoichiometric, a hundred-fold increase in the level of the target

22

1

Cancer as a Disease of Cell Proliferation

enzyme means that it takes one hundred times as much methotrexate to inactivate it. Dihydrofolate reductase is not the rate-limiting enzyme in the pathway of thymidylate biosynthesis, so typically it has to be inhibited about 95% before the synthesis of DNA starts to slow down. Many antibacterial drugs are cytostatic, rather than cytotoxic, that is, they do not kill the bacteria, but they prevent them from replicating. This gives the immune system of the infected person or animal a chance to eliminate the infection. This strategy does not work for anticancer drugs: they must be cytotoxic rather that cytostatic. Any cancer cells that survive a treatment, even if they are temporarily unable to divide, will eventually recommence cell division, and tumour growth will resume. In fact, many tumours do elicit an immune response, but it is typically much weaker than that caused by infectious diseases. Methotrexate acts by starving the cell of precursors needed for synthesis of DNA and RNA, suggesting that it might be merely cytostatic. However, the imbalance in DNA precursors results in breaks in the growing DNA strands that the cell is unable to repair, resulting in cytotoxicity. Because of their toxicity to normal proliferating tissues, antimetabolites have limited anticancer selectivity, but methotrexate and other antimetabolites continue to play an important role in the treatment of human cancer.

References Aguda BD (2001) Kick-starting the cell cycle: from growth-factor stimulation to initiation of DNA replication. Chaos 11:269–276 Bentele M, Lavrik I, Ulrich M, Stosser S, Heermann DW, Kalthoff P, Krammer H, Eils R (2004) Mathematical modelling reveals threshold mechanism in CD95-induced apoptosis. J Cell Biol 166:839–851 Bishop JM (1983) Cellular oncogenes and retroviruses. Annu Rev Biochem 52:301–354 Brightman FA, Fell DA (2000) Differential feedback regulation of the MAPK cascade underlies the quantitative differences in EGF and NGF signalling in PC12 cells. FEBS Lett 482:169–174 Dow LE, O’Rourke KP, Simon J et al (2015) Apc restoration promotes cellular differentiation and reestablishes crypt homeostasis in colorectal cancer. Cell 161:1539–1552 Ekstrøm CT (2012) The R primer. CRC Press, Boca Raton, FL Fusenegger M, Bailey JE, Varner J (2000) A mathematical model of caspase function in apoptosis. Nat Biotechnol 18:768–774 Gilbert D, Fuss H, Gu X et al (2006) Computational methodologies for modelling, analysis and simulation of signalling networks. Brief Bioinform 7:339–353 Gomperts BD, Kramer LM, Tatham PER (2003) Signal transduction. Elsevier, New York, pp 71–105 Hancock JT (2010) Cell signalling, 3rd edn. Oxford University Press, Oxford, pp 321–324 Hannahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100:57–70 Hua F, Cornejo MG, Cardone MH et al (2005) Effects of bcl-2 levels on Fas signalling-induced caspase-3 activation: molecular genetic tests of computational model predictions. J Immunol 175:985–995 Hunter T, Cooper JA (1985) Protein-tyrosine kinases. Annu Rev Biochemist 54:897–930 Jackson RC (1992) The theoretical foundations of cancer chemotherapy introduced by computer models. Academic, New York, pp 259–271 Jackson RC (2017) The Boolean kinetics of signal transduction. Integr Cancer Sci Therap 4:1–8

References

23

Jackson RC, Weber G, Morris HP (1975) IMP dehydrogenase, an enzyme linked with proliferation and malignancy. Nature 256:331–333 Jenkins A, Macauley M (2017) Bistability and asynchrony in a Boolean model of the L-arabinose operon in Escherichia coli. Bull Math Biol 79:1778–1795 Joe AK, Weinstein IB (2011) Oncogene addiction. In: Encyclopedia of cancer. IntechOpen, pp 2616–2622. https://doi.org/10.1007/978-3-642.16483-5-4222 Latchman DS (2008) Eukaryotic transcription factors, 5th edn. Academic Li X, He S, Ma B (2020) Autophagy and autophagy-related proteins in cancer. Mol Cancer 19:12 Minden MD (1987) Oncogenes. In: Tannock IF, Hill RP (eds) The basic science of oncology. Pergamon Press, Oxford, pp 72–88 Morris HP (1975) Biological and biochemical characteristics of transplantable hepatomas. In: Grundman E (ed) Handbuch der allgemeinen Pathologie, vol 6, part 7. Springer, Berlin, pp 277–334 Murray JD (1993) Mathematical biology, 2nd edn. Springer, Berlin, p 41, 60 Nelson J (2008) Structure and function in cell Signalling. Wiley, Chichester Norton L (1988) A Gompertzian model of human breast cancer growth. Cancer Res 48:7067–7071 Norton L, Simon R (1986) The Norton-Simon hypothesis revisited. Cancer Treat Rep 70:163–169 Pratt WB, Ruddon RW (1979) The anticancer drugs. Oxford University Press, Oxford Prevostel C, Alvaro V, Bolvilliers F et al (1995) The natural protein kinase C alpha mutant is present in human thyroid neoplasms. Oncogene 11:669–674 Shay JW, Wright WE (2000) Hayflick, his limit, and cellular ageing. Nat Rev Mol Cell Biol 1:72– 76 Speer JF, Petrovsky VE, Retsky MW, Wardwell RH (1984) A stochastic numerical model of breast cancer growth that simulates clinical data. Cancer Res 44:4124–4130 Tannock IF, Hill RP (1987) The basic science of oncology. Pergamon Press, Oxford Varmus HE (1984) The molecular genetics of cellular oncogenes. Ann Rev Genet 18:553–612 Vaudry D, Stork PJ, Lazarovici P et al (2002) Signaling pathways for PC12 cell differentiation: making the right connections. Science 296:1648–1649 Veliz-Cuba A, Stigler B (2011) Boolean models can explain bistability in the lac operon. J Comput Biol 18:783–794 Vogelstein B, Kinzler KW (1993) The multistep nature of cancer. Trends Genet 9:138–141 Weber G (1977) Enzymology of cancer cells, parts 1 and 2. N Engl J Med 296(486–491):541–551 Weber G, Prajda N, Jackson RC (1976) Key enzymes of IMP metabolism. Transformation and proliferation-linked alterations in gene expression. Adv Enzym Regul 14:3–24 Weber G, Goulding FJ, Jackson RC, Eble JN (1978) Biochemistry of human renal cell carcinoma. In: Davis W, Harrap KR (eds) Characterization and treatment of human Tumours. Excerpta Medica, Amsterdam, pp 227–235 West J, Hasnain Z, Macklin P, Newton PK (2016) An evolutionary model of tumor cell kinetics and the emergence of molecular heterogeneity driving Gompertzian growth. SIAM Rev Soc Appl Math 58:716–736 Wodarz D, Komarova NL (2014) Dynamics of cancer: mathematical foundations of oncology. World Scientific, Singapore, p 37

Chapter 2

Genetic and Chromosomal Instability

Abstract All tumours have genetic or chromosomal abnormalities or both, and most cancers are diseases of ageing, suggesting the progressive accumulation of genetic damage. Tumours are genetically unstable, such that the number of mutations and chromosomal abnormalities increases with time. Genetic and/or chromosomal instability is now regarded as the defining characteristic of cancer. Cancer genome projects, based upon whole genome sequencing, have identified thousands of mutations in tumours, many of which, driver mutations, are linked to transformation or tumour progression, while others—passenger mutations—have no known causal link to cancer. Whether or not a particular tumour cell carries genetic mutations or karyotypic changes, inappropriate gene expression is the common determinant. Large tumours are highly heterogeneous, because different cells in the tumour have accumulated different combinations of mutations. The process of tumour progression results from Darwinian natural selection, with cells that have a competitive advantage eventually dominating the population. An important class of anticancer drugs, the alkylating agents, acts by causing DNA damage; their antitumour selectivity arises from the fact that most tumours are less efficient at repairing DNA damage than normal cells. However, DNA-active drugs are mutagenic which can result in further mutations in surviving tumour cells.

Human DNA has about three billion base pairs (per haploid set) and although we now know that only a fraction of these have to be replicated accurately, the error rate cannot be greater than about one in a hundred million (speaking very roughly) or the organism would be torpedoed in evolution by its own errors. Francis Crick, What Mad Pursuit (1988)

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_2. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_2

25

26

2.1

2

Genetic and Chromosomal Instability

Cancer as a Disease of Ageing

The age distribution of cancer indicates that it increases as the fourth to sixth power of age. The interpretation of this observation by epidemiologists has been that tumour development is a multi-step process. Interpreting “steps” as mutations, this means that four to six mutations are required. This argument makes a number of assumptions: that all the cells at risk are present for the lifetime of the individual, that cells with fewer than the required number of mutations have no growth advantage, that susceptible tissues have a large number of cells at risk, and that environmental carcinogens cause a large increase in the spontaneous mutation rate. Background mutation rates are estimated at 10-7 to 10-6 per cell per division. Mutagens may increase this by one or two orders of magnitude (Phillips 1987). Given the six “hallmarks of cancer” of Hanahan and Weinberg (2000) it is tempting to subscribe to the “restaurant menu” theory: an aspiring malignant cell must choose one mutation that confers immortality, one that gives a growth advantage, one for invasiveness, one for metastasis. This is a reasonable analogy, since, as with a four-course meal, the cell cannot have its pudding before the soup: the mutations have to take place in a particular order. Calculations with the FINITE-T model of malignant progression (see the online supplement) suggest that if the development of a cancer required three or more mutations at the background rate, the lifetime risk of cancer in humans would be less than 1% (in fact it is about 40%). Either we are constantly exposed to mutagens or some other factor is increasing the spontaneous mutation rate. The current consensus view is a nuanced version of the multi-stage origin of malignancy, and evolutionary dynamics calculations can suggest a sequence of events that is consistent with the cancer incidence data. Frank (2007) provides detailed incidence data for many cancer types and distinguishes age-specific incidence from acceleration, the rate at which incidence increases with age. He cites an equation (originally due to Peto et al. 2000): I = nbt n - 1 ,

ð2:1Þ

where I is the standard measure of age-specific incidence, b is a measure of carcinogen dosage, t is the number of years of carcinogen exposure until onset, and n is a scaling factor of incidence with time. Parameter b is specific to a particular carcinogen. This kind of analysis has been applied to tumours where the cause is relatively homogeneous, e.g., NSCLC in tobacco smokers, and used to predict the consequences of smoking cessation at different ages.

2.2

2.2

Radiation and Chemical Carcinogenesis

27

Radiation and Chemical Carcinogenesis

Perhaps the earliest indication that tumours had genetic abnormalities came from the observation that radiation and chemical carcinogens cause damage to DNA. Ultraviolet (UV) radiation is not capable of penetrating biological tissues to any great extent, and only skin tumours have been shown to result from UV irradiation. Experiments with mice showed that a single UV exposure, even at very high intensity, did not cause tumours, but that frequent exposure to low doses was an effective carcinogen. It is notable that malignant melanoma is most prevalent in areas that have a lot of sunshine, and where many people enjoy an outdoor lifestyle, such as California and Australia. UV irradiation causes formation of thymidine dimers in DNA, blocking DNA replication. These dimers can be removed by a photolyase repair enzyme (Butenandt et al. 2000). Following UV irradiation, these dimers can be detected in human urine (Le Curieux and Hemminki 2001). The resulting DNA strand breaks are then repaired by the nucleotide excision repair pathway, but if repair is incomplete, cells may either die or continue to replicate with damaged DNA. Tumours resulting from incompletely repaired UV damage, including melanomas, are often highly immunogenic. Ionising radiation, defined as photons with energy greater than 10 eV, has greater penetrating power than UV, and can cause tumours in internal organs. Ionising radiation can result in single- and double-strand DNA breaks. These breaks can be repaired by a number of different pathways, as described in Chap. 4 and, as with UV damage, incomplete repair can result in permanent DNA damage that may result in malignancy (Rauth 1987). It was first suggested in the late eighteenth century that the high incidence of scrotal cancer in chimney sweeps might be a result of their exposure over many years to soot and tar (Archer 1987). Subsequent work, from the early part of the twentieth century onwards, showed that polycyclic aromatic hydrocarbons present in coal tar caused tumours when painted on to rabbit’s ears. Many other, chemically unrelated, classes of compounds are now known to be carcinogenic: what they have in common is that they can form directly, or be metabolised to, reactive electrophilic forms (Miller and Miller 1981). At about the same time it was reported that DNA was the primary molecular target for these electrophiles, and that chemical carcinogens caused somatic mutations (Straus 1981). Moolgavkar and Knudson (1981) proposed a two-stage model for carcinogenesis, in which a first mutation in a stem cell increases the probability of self-renewal and decreases the probability of differentiation to produce a so-called initiated cell. A second mutation then eliminates the ability to differentiate and further increases the self-renewal capacity. Some chemical carcinogens, termed “complete carcinogens” caused both tumour initiation, and the second, promotion phase. Other compounds, or sometimes the same compounds at lower doses, caused only the initiation stage. Mechanistic work on the mechanism of tumour promotion focussed on a family of lipid-reactive molecules, the phorbol esters. However, establishing structure–activity relationships for tumour promotion was complicated by the fact that tumour

28

2 Genetic and Chromosomal Instability

promotion could be activated by a wide range of chemically unrelated structures. A property of many tumour-promoting compounds was their ability to stimulate production of reactive oxygen species (ROS). ROS are mutagenic by virtue of their ability to oxidise DNA guanine bases to 8-oxo-guanine, which base-pairs with T, rather than C, and thus is a point mutation. However, tumour promotion appears to be the effect of a number of unrelated molecular mechanisms: the phorbol esters, for example, activate protein kinase C, which in turn interacts with Ras and Raf (Chap. 1). Archer (1987) reviewed a number of examples of chemical carcinogens and oncogenes causing oncogene activation. The main conclusions to emerge from the very large body of research on chemical carcinogens was that most of them cause DNA damage, that this sometimes resulted in mutations, and that some of these mutations activated oncogenes. There is a continuing debate whether all mutagens are carcinogens. Obviously not all mutations result in cancer, but mutations to certain genes (such as components of cell cycle checkpoints) do result in cancer, and most mutagens are not specific to particular genes. Another debate concerns what, if any, is a safe level of exposure to the various sources of radiation, including diagnostic X-rays. This is not the place to review this topic, except to note that mammalian cells have extensive DNA repair capacity, meaning that there are safe levels; and secondly, that because of frequent polymorphisms in DNA repair genes, there may be considerable individual variation in those levels.

2.3

Viral Carcinogenesis

In addition to radiation and chemical carcinogenesis, there is evidence that some tumours have a viral aetiology. In 1908, Ellerman and Bang showed that an extract from the blood of leukaemic chickens induced leukaemia in normal chickens, and, in 1911, Peyton Rous showed that a “filterable agent” could transmit fibrosarcoma in chickens (Sheinin et al. 1987). These and other experiments in mice and rabbits established that viruses could cause—or transmit—cancer in birds and mammals. Both DNA viruses and RNA viruses were subsequently shown to play a causative role. Although for many years there was scepticism about whether human tumours ever had a viral aetiology, by the 1980s, it was shown that some strains of human papilloma virus (HPV) were linked with cervical carcinoma in women, and that certain retroviruses, human T-cell leukaemia viruses (HTLV) were able to transform human T-lymphocytes. The Rous sarcoma virus was a valuable tool for studying viral transformation because its genome contains only four genes. One of these, v-src appeared to be an altered form of a normal human gene, c-src. The src genes, both viral and cellular, code for cytosolic protein tyrosine kinases that in turn phosphorylate a number of signalling pathway components, including STAT transcription factors (Fig. 1.5). HTLV1 is a retrovirus associated with T cell leukaemias and lymphomas. Although the mechanism by which it causes transformation is unknown, it infects

2.4

Burkitt’s Lymphoma and Chromosomal Translocation

29

CD4+ cells, and causes increases in inflammatory cytokines, including TNFα, and causes oxidative stress (Brites et al. 2021). This combination of oxidative stress and activation of immune signalling resembles the pattern seen in chronic myeloid leukaemia (CML), discussed in Chap. 7 (Niazmand et al. 2022).

2.4

Burkitt’s Lymphoma and Chromosomal Translocation

Burkitt’s lymphoma, a B-cell malignancy, is an early example of a human malignancy associated with chromosomal translocation of a specific gene (Vockerodt et al. 2015). Initially found to be commonest in districts where malaria is prevalent, most cases were shown to be linked with infection by the Epstein–Barr virus (EBV). EBV is a herpes virus, commonly associated with infectious mononucleosis, an infection that causes B cell proliferation, but non-malignant. Burkitt’s lymphoma is an example of non-Hodgkin lymphoma, i.e., lymphomas other than Hodgkin’s disease; it is very sensitive to chemotherapy. The virally transformed cells express EBV antigens, making the tumour highly antigenic. Cytogenetic studies showed that Burkitt’s lymphoma was associated with chromosomal translocations that placed the growth-promoting gene c-myc under the control of a promoter associated with the immune response: in 85% of cases this was a t(8,14)(q24:q32) translocation that placed the c-myc gene under the control of the immunoglobulin heavy chain gene IGHα. Burkitt’s lymphoma is primarily a disease of children, though it does occur in adults, and the adult disease generally has a worse prognosis. The overall cure rate (in societies with modern health services) after chemotherapy is over 90%. Patients with poor prognosis usually have additional genetic abnormalities (mutations or chromosomal translocations) as well as the primary translocation caused by EBV. Another human malignancy, nasopharyngeal carcinoma, is also linked with EBV infection. The conclusion from this, and other virally-induced tumours, is that malignant transformation is the result, not of expression of viral genes, but of the chromosomal rearrangement resulting from the virus’s interaction with host cell DNA. Another example of a malignancy that has a complete complement of normal genes but a single chromosomal translocation is chronic myeloid leukaemia (CML). In this case, the Abl gene, which is normally activated to drive the innate immune response by the presence of bacterial or fungal cell wall components, is permanently activated by a chromosomal translocation that puts it under the control of the Bcr gene. Abl phosphorylates and activates the transcription factors STAT3 and STAT5. Thus, CML has a complete complement of normal genes (though two of them, after the translocation, are expressed as a single fusion protein), but one of them is abnormally expressed. This will be discussed in detail in Chap. 7. In general, carcinomas do not show consistent chromosomal translocations of that kind. Some melanomas have deletion of a tumour suppressor gene, CDKN2A, which, in conjunction with Ras activation, results in transformation (Chin et al. 1997).

30

2.5

2

Genetic and Chromosomal Instability

Aneuploidy and Duesberg’s Hypothesis

Peter Duesberg (2005, 2007) argued that aneuploidy, rather than mutations, was the primary driver of most, if not all, carcinomas. His hypothesis rested upon a number of observations. In general, cancer is not inherited (though there may be an inherited predisposition to some tumour types). Most carcinogens are mutagens, but some carcinogens are non-mutagenic. Carcinomas may develop years or even decades after exposure to carcinogens, as for example with mesothelioma, which may develop up to thirty years after asbestos exposure. Cancers may follow pre-neoplastic aneuploidy, and carcinomas in particular are almost always aneuploid. Aneuploid cells are karyotypically unstable, so their presence in tumours is a driver of cellular heterogeneity. The rate of chromosomal variations exceeds the genetic mutation rate by many orders of magnitude. Unlike Hanahan and Weinberg, who suggested that genetic instability was an “enabling factor”, in Duesberg’s view chromosomal instability is the essence of neoplasia, and the other “hallmarks of cancer” are consequences of this instability (Duesberg et al. 2005). Duesberg and his collaborators have likened carcinogenesis to a process of speciation (Hirpara et al. 2018). The progressive accumulation of aneuploidy results in new phenotypes, most of which will be lethal, or have a competitive disadvantage, resulting in their eventual extinction. A few of these altered phenotypes may have a comparative advantage, resulting in their survival, and eventual dominance of the tumour population, representing the next stage in the tumour progression process, analogous to emergence of a new species in population biology. They concluded that this speciation theory is consistent with the long latent period of most cancers. They further argued that their theory supports a single-step origin of cancers, “because karyotypic autonomy is all-or-nothing”. Part of the argument of Duesberg et al. is that the effects of chromosomal rearrangement are necessarily pleiotropic, since chromosomal changes may alter the expression of large numbers of genes. It follows that some changes will occur, and persist, which do not present an immediate selective advantage. As an example, Bloomfield and Duesberg (2015) cite metastasis. Metastasis is an almost invariable consequence of the tumour progression process, even though the genetic changes that drive it must first occur in the primary tumour where they offer no immediate selective advantage. The chromosomal changes that cause metastasis, in this view, are initially, an irrelevant consequence of pleiotropic changes that become significant when the selective environment changes. In summary, the Duesberg hypothesis accepts that neoplastic transformation is a consequence of genetic mutations, but argues that most mutations are irrelevant to transformation, or to tumour progression: the mutations that matter are those that result in aneuploidy: “cancer is a chromosomal, rather than a genetic disease” (Duesberg et al. 2006). He considers the possibility that mutations that are involved in DNA repair and in chromosomal segregation are the primary causes of chromosomal instability but suggests that there are no consistent correlations to support this view. We shall reconsider this point in Chap. 3.

2.6

Differences Between Carcinomas and Lymphomas

31

Since chromosomal rearrangements, including aneuploidy, do not alter the genome of the cell (unless chromosomes, or parts of chromosomes, are deleted) why should they affect gene expression? Venkitaraman (2007) considered the various ways in which chromosomal changes could alter gene expression levels. The commonest is ploidy changes, but more subtle effects are sometimes involved: for example, non-reciprocal chromosome translocations may result in gain, or loss, of genomic regions. Chromosomal rearrangements may result in a growthpromoting gene coming under the control of an inappropriate promoter (as in Burkitt’s lymphoma) or in being removed from its appropriate promoter or enhancers. However, consistent karyotypic changes with an obvious link to transformation were not seen. Venkitaraman (2007) concluded that “the contribution to carcinogenesis made by changes in ploidy is not wholly clear”. The observation that almost all carcinogens caused mutations suggested that mutations, rather than chromosomal rearrangements, might be the primary genetic abnormality in cancer. However, certain malignancies, particularly leukaemias and lymphomas, appear to have a complete complement of normal genes, but packaged in an abnormal karyotype. Duesberg and colleagues suggested that there were parallels between the origin of tumours and formation of new species in biological populations (Hirpara et al. 2018). Gatenby et al. (2009), discussing parallels between cancer evolution and cancer biology, point out that the selective advantage of cancer-causing mutations must be balanced against the phenotypic cost to the cell (in terms of increased cellular resource requirements). It is true that chromosomal rearrangements contribute to speciation and to cancer, but there are important differences. In particular, the primary source of chromosomal change in normal reproduction is the recombination that occurs in meiosis. There is no close parallel to that process in carcinogenesis.

2.6

Differences Between Carcinomas and Lymphomas

The observation that Burkitt’s lymphoma was caused by a translocation that placed the c-myc gene under the control of a promoter associated with the immune response was a pivotal discovery in the development of the oncogene concept. Originally it was suggested that constitutive activation of c-myc, or some other growth-promoting transcription factor, was sufficient to transform the cell. In the case of lymphomas and lymphoid leukaemias this is the case. For epithelial cells, constitutive activation of a cellular proto-oncogene will result in hyperproliferation, but not neoplastic transformation; epithelial cells require a second mutation or chromosomal rearrangement to generate a cancer stem cell. The explanation for this difference became clear with the growing realisation that malignancy is a Darwinian process that requires the two drivers of the genetic algorithm: heterogeneity, and a selective growth advantage. When a chromosomal rearrangement places a cellular oncogene under the control of an immune system promoter, and the cell is then activated by an immune response—an EB virus

32

2 Genetic and Chromosomal Instability

infection in the case of Burkitt’s lymphoma—the normal processes of somatic hypermutation and V(D)J recombination result in a genetically heterogeneous clone of lymphocytes. The two requirements for tumour formation are now in place: cellular heterogeneity and a cell that possesses a growth advantage over its neighbours. A lymphoma results. When a cellular oncogene in an epithelial cell is activated, that cell now possesses a growth advantage over its neighbours; it may form a carcinoma in situ, or a benign polyp or tumour, but it is not genetically unstable. Such lesions are premalignant in the sense that, if they subsequently have a mutation or chromosomal rearrangement that results in genetic instability or aneuploidy, they are then transformed. As we shall discuss in later chapters, in epithelial cells the instability often results from a mutation in the mitotic spindle assembly checkpoint, or the DNA damage checkpoint. Lymphoid cells are inherently unstable, it is part of how the immune system works.

2.7

Microsatellite Instability

Microsatellites are short repeated sequences (2–10 base pairs) that occur throughout the genome. In about 15% of colon carcinomas, and to a lesser extent in other carcinomas, microsatellite instability occurs as a result of defects in the DNA mismatch repair system (Alhopuro et al. 2012). Microsatellite instability often results in inactivation of tumour suppressors, particularly Apc. Apc is mutated or deleted in the hereditary condition, familial adenomatous polyposis, which causes a predisposition to intestinal polyps. Apc is a negative regulator of the Wnt signalling pathway (Fig. 1.3), and, significantly, interacts with BUB1, a component of the SAC (Chap. 5). For this reason, loss of Apc is associated with chromosomal instability. Tumours that have microsatellite instability are often diploid, as are certain experimental tumours induced by insertion of multiple altered proteins (Venkitaraman 2007). However, the great majority of human carcinomas are aneuploid. Microsatellite instability results in further genetic instability by causing frameshift mutations, often in genes involving DNA repair, cell signalling, apoptosis, epigenetic regulation, and micro-RNA processing (Yamamoto and Imai 2015). A major group of tumours with microsatellite instability appear to be caused by Lynch syndrome, which results from hypermethylation of genes involved in DNA mismatch repair. Another subgroup of tumours with microsatellite instability is associated with mutations in BRAF.

2.8

Tumour Heterogeneity

The number of mutations increases as a tumour increases in size. Individual carcinomas may contain 50 or more mutations. Most authors distinguish between “driver” mutations, whose expression is directly linked to the process of tumour

2.9

Cancer Genome Projects: Driver and Passenger Mutations

33

progression, and “passenger” mutations, apparently random changes that do not appear to affect the biological properties of the tumour. The complexity of the situation is increased by the fact that different cells within a tumour mass may contain different combinations of mutations. Some of these cells may have mutations that give them a growth advantage over their neighbours, and as the tumour grows, those cells will account for a progressively greater share of the total cell population. Tumour heterogeneity thus leads to tumour progression by a process of Darwinian natural selection. A recent review of intra-tumoral heterogeneity at the genome level (Dentro et al. 2021) sequenced 2658 human cancers, of 38 tumour types. 95% showed evidence of distinct subclonal expansion, and there were cancer type-specific patterns of driver gene mutations and copy number alterations. This study provided direct evidence that tumour progression proceeds by a process of selection among diverse subclones carrying driver mutations that confer selective advantage.

2.9

Cancer Genome Projects: Driver and Passenger Mutations

The cancer genome projects in human and mouse tumours have identified thousands of cancer-associated mutations. The majority of these mutations appear to be unrelated to proliferation or malignancy, and are termed “passenger mutations (Pon and Marra 2014; Kumar et al. 2020). A recent review listed passenger mutations in 2500 cancer genomes (Kumar et al. 2020). Other mutations, often in components of signalling pathways or in DNA repair enzymes are probably causally linked to transformation or progression, and are termed “driver” mutations (Greenman et al. 2007). A study in melanoma detected a dominant mutational signature reflecting DNA damage due to ultraviolet light exposure, a known risk factor for melanoma (Pleasance et al. 2010). In a group of 100 breast tumours, there was no single mutational signature, though about 10% of the samples showed mutations of cytosine at TpC dinucleotides. There was a strong correlation between mutation number, age at diagnosis, and cancer histological grade (Stephens et al. 2012). A study of over 7000 tumours from 30 cancer types (Iranzo et al. 2018) found common mutations in 198 previously known cancer-associated genes. 27% of the tumours could be assigned to clusters, most of which encompassed one or two cancer types. The clustering patterns were independent of the source tissue. A database of mouse and human driver gene mutations (Abbott et al. 2015) was used for a hierarchical clustering analysis that showed certain pathways were enriched in blood cancers compared to solid tumours, e.g. CDKN2A (which encodes p16 and p14ARF) suggesting that genomic instability was the result of oncogene-induced DNA replication stress (Negrini et al. 2010). One of the genes often mutated, ATM, can cause aneuploidy through cross-talk with the SAC.

34

2

Genetic and Chromosomal Instability

Some mutations appear to act as drivers in certain tumours, but as passengers in others. This is because whether a mutation confers a selective advantage or disadvantage is context-dependent. Selection is the result of the interaction of cells with their environment. A mutation that conferred resistance to hypoxia at the cost of a slightly slower growth rate in well-oxygenated conditions could be a driver mutation in hypoxic conditions, but a passenger, or completely eliminated, in conditions of adequate blood supply. Since environmental conditions vary over time, a particular mutation could be advantageous over a portion of the growth process, but become redundant later. The Cancer Genome Atlas (TCGA) was begun in 2006 as a whole genome sequencing initiative of the US National Cancer Institute, and is available as a publicly accessible database. In 2013, it reported conclusions from 12 tumour types. The database now includes 20,000 specimens, covering 33 cancer types. There have been frequent reviews and updates, both of the data and of the evolving methods (Ganini et al. 2021), and detailed analyses have been published for several individual tumour types (Ricketts et al. 2018). Other large databases of whole genome sequences exist, such as the International Cancer Genome Consortium (ICGC) and the Pan-Cancer Analysis of Whole Genomes (PCAWG) also contain about 1200 related transcriptomes (Ganini et al. 2021). The cancer genome databases have been used to compile summaries of cancerassociated chromosomal variations (Chinnalyan and Palanisamy 2010; Drews et al. 2022). Recurrent chromosome translocations, and gene fusions are described, and 17 copy number signatures. An analysis of copy number signatures in ovarian carcinoma claimed that copy number signatures at diagnosis could predict both overall survival and the probability of relapse due to acquired resistance to platinum drugs (Macintyre et al. 2018). The overwhelming impression to emerge from the cancer genome projects is one of great complexity and great heterogeneity. However, certain recurrent motifs are apparent. One of these is the mitogen-activated protein kinase (MAPK) pathway, which signals from receptor tyrosine kinases through RAS, RAF, MEK and ERK to the E2F and myc transcription factors (Way et al. 2018; Imperial et al. 2019). A review of conclusions from TCGA, ICGC, and PCAWG made several interesting generalisations. On average, cancer genomes contained 4–5 driver mutations. In about 5% of cases, no drivers were identified, suggesting that drivers that are as yet unidentified were involved. Chromothripsis was reported to be an early event in tumour evolution (Campbell et al. 2020). Chromothripsis is a simultaneous multiple chromosome breakage event, in which the fragments (tens to thousands) are misjoined, resulting in transformation. In acral melanoma this event precedes point mutations and affects several cancer-associated genes simultaneously. Cancers arising in tissues with low replicative activity often had mutations that resulted in maintenance of telomere levels. Timing analysis of a series of over 2600 cancers suggested that driver mutations often precede diagnosis by many years, even decades (Gerstung et al. 2020). Early oncogenesis is characterised by mutations in a limited set of driver genes, and copy number gains that may be specific to particular tumour types (e.g. trisomy 7 in

2.10

Replicative Stress and Chromosomal Instability

35

glioblastoma). With tumour progression the number of driver gene mutations increases up to fourfold, along with increased genomic instability in some cases. Since most cancer deaths result from metastatic disease, particular interest has been attached to the mutations that drive metastasis, a subject discussed in Chap. 9. It is generally believed that metastases arise from a single cell, though there is evidence that some prostate tumour metastases may result from polyclonal seeding, and metastasis to metastasis spread was seen (Gundem et al. 2015). Lesions affecting tumour suppressor genes were the result of single events, but mutations in genes involved in androgen receptor signalling commonly involved multiple convergent events in different metastases. Based upon analysis of The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC) it is possible to identify distinct subtypes of prostate cancer with respect to the genes involved in initiation, progression, and metastasis (Fraser and Rouette 2019). A similar analysis has been published for squamous cell tumours of the head and neck (Plath et al. 2020). It identified 7 main mutation signatures, one of which was associated with alcohol and tobacco consumption, an example of how tumour genome analysis may give clues to aetiology.

2.10

Replicative Stress and Chromosomal Instability

The kinetics of the G1:S checkpoint suggest that it acts as an ON/OFF switch, meaning that DNA synthesis is S phase is either fully active, or completely inactive. Even in proliferating epithelial cell populations, DNA replication is probably off more than on. Normal DNA replication and subsequent decatenation of the replicated strands involves making and rejoining strand breaks in DNA. In normal cells the capacity of the repair pathways is sufficient that cells do not progress into mitosis with unrepaired strand breaks. Since most, if not all, tumours have DNA synthesis permanently ON, it is argued that spontaneous generation of DNA damage may exceed the capacity of the repair enzymes, a situation termed replicative stress. It is suggested that, for this reason, chromosomal instability may be an unavoidable consequence of replicative stress, without needing to postulate mutations in the DNA damage checkpoint (DNADC) or the spindle assembly checkpoint (SAC). Negrini et al.(2010) cited the low incidence of mutations in spindle assembly checkpoint genes compared with the relatively high incidence of mutations in repair genes, including ATM. In fact, all human tumours that have been sequenced do have either SAC mutations or DNADC mutations, and because of cross-talk between the DNADC and the SAC it seems likely that all tumours lack a functional SAC. Benign hyperproliferative conditions, such as intestinal polyps, have a normal, diploid karyotype. However, at present the possibility that replicative stress may cause chromosomal instability in cells with intact checkpoints cannot be ruled out.

36

2.11

2

Genetic and Chromosomal Instability

Modelling Mutations with a Genetic Algorithm

The resemblance of malignant progression to a process of Darwinian selection has been remarked upon by many authors (reviewed by Cahill et al. 1999). In this view, cancer is the end result of progressive accumulation of somatic mutations, each of which confers a selective proliferative advantage on those cells that possess it. Darwin postulated three requirements for evolution: first, descent with modification, i.e. a source of variants; second, struggle for existence: if there are unlimited resources, there can be no competition between the variants; third, natural selection of the favoured variants. Although the term was not in use in Darwin’s day, he was describing an algorithm (Dennett 1995). When interpreting malignant progression as a Darwinian process, the source of variants is the cellular heterogeneity resulting from genetic instability. The competition for resources results from limitations in the supply of nutrients or oxygen, or in the case of epithelial tissues, space on the membrane on which the cells are attached. The “favoured variants” survive by having a faster doubling time or a lower rate of spontaneous cell death. These factors can all be modelled in a genetic algorithm (Fig. 2.1), an example of which is given in the online supplement. Proliferation of cells, and their competition for space on a portion of basement membrane fed by a capillary, corresponds to the competition of organisms in an ecosystem. Loss of function of the G1:S checkpoint gives cells a competitive advantage in this environment. Aneuploidy caused by chromosomal instability provides the cellular heterogeneity corresponding to Darwin’s “descent with modification”. As with natural selection, the malignant progression process is iterative, autocatalytic, and irreversible. The positive feedback in the algorithm results from the fact that the higher the proportion of a favoured variant, the greater the rate of progression (Jackson 2016). In the early stages of malignant progression, the

Fig. 2.1 The malignant progression algorithm. The dashed line indicates a positive feedback loop

dysfunctional spindle checkpoint

dysfunctional G1 checkpoint

aneuploidy

uncontrolled proliferation

+ malignant progression

2.12

Polyploidy

37

objective function of the algorithm is provided by the relative numbers of normal and transformed cells competing for space on the basement membrane. However, after a mutation that removes the need for anchorage dependence, the objective function changes, because cells can relocate to a range of alternative environments. These later stages of malignant progression are more analogous to the process of speciation in evolution than to adaptation to an unchanging environment. The discussion so far has emphasised the role of aneuploidy in the development of carcinomas, and the role of chromosomal translocation in the aetiology of lymphomas and of CML. However, chromosomal instability also occurs invariably in acute myeloid leukaemia (AML). There are three types of AML: de novo AML, which appears to present as an acute disease from the onset; secondary AML, in which acute disease arises from the progression of a less aggressive precursor, such as CML or CMML, and therapy-related AML, which is sometimes a late consequence of otherwise successful treatment of a different malignancy with mutagenic chemotherapy. Chromosomal instability is seen in all three types (Lisboa et al. 2021). When modelling mutations in replicating cell populations, mutation rates are usually expressed as number of mutations per cell division, and for mutations occurring through spontaneous miscoding this is appropriate. However, non-replicating cells can mutate, because of spontaneous cytosine deamination, or because of guanine 8-oxide formation. In summary, some malignancies have a complete set of normal genes, but packaged in an abnormal karyotype. Other tumours contain genetic mutations that are clearly related to abnormal growth control. Such tumours are invariably aneuploid. Three generalisations can be drawn: (i) all tumours have karyotypic abnormalities; either aneuploidy or translocations; (ii) whether or not a malignant cell has genetic mutations, inappropriate gene expression is the common determinant; and (iii) cancer is a process, not a phenotype. The malignant progression algorithm assumes that cells compete for resources, but a developing tumour cell does not compete with every other cell in the entire body. The area within which competition occurs is topographically restricted. When a skin cell is infected by papilloma virus, its G1:S checkpoint is overridden and a wart develops. The growth of the wart is typically limited to the area fed by a single afferent capillary. This area will usually contain from fewer than one hundred to a few hundred cells. In the algorithm, the area within which competition for space and nutrients takes place is termed a domain.

2.12

Polyploidy

A particular kind of chromosomal abnormality sometimes observed in both normal and malignant cells is polyploidy. Whereas the term “aneuploidy” refers to an abnormal number of chromosomes (i.e. in human cells a chromosome number other than 46 for somatic cells or 23 for germ cells), the term “polyploidy” refers

38

2 Genetic and Chromosomal Instability

to an abnormal number of complete sets of chromosomes; cells with 69 chromosomes are referred to as “triploid” and cells with 92 chromosomes are tetraploid. Polyploidy can arise when cells complete two rounds of DNA synthesis before entering mitosis, or when a normal mitosis is not followed by cytokinesis. It can occur when an ovum is fertilised by more than one sperm. An ovum fertilised by two spermatozoa will be triploid and is said to have karyotype 69, XXX (assuming both spermatozoa were X). Such embryos usually abort, but are occasionally carried to term in which case the neonate is likely to have birth defects as a result of abnormal gene dosage effects. Somatic polyploid cells are quite common, particularly in liver and other non-dividing normal cell populations and will have their normal differentiated function. Polyploid tumour cells are able to complete normal mitosis. Having to replicate more than the usual amount of DNA will constitute a selective disadvantage, and whether such cells can persist in the tumour will depend on the magnitude of any compensating growth advantages they may have. Drugs that disrupt normal function of the SAC may result in polyploidy, for example this has been reported in cells treated with inhibitors of aurora kinase B.

2.13

Mutation Rates and Lifetime Cancer Risk

Evolutionary dynamics calculations showed that if tumour development requires four mutations, and that the mutation rate was 2 × 10-6per cell division, and if the smallest palpable tumour in a mouse was 107 cells, the lifetime incidence of spontaneous tumours in mice would be 50%); it is thus regarded as a tumour suppressor gene. Functionally, it is a transcription factor, activating a large number of genes, including p21 and p27 which form an essential part of the G1:S checkpoint, discussed below. P53 thus forms an essential part of the G1:S checkpoint (Fig. 3.2), but it is not confined to G1 phase, and also occurs later in the cell cycle, where it again can cause cell cycle arrest in response to DNA damage. As well as responding to DNA damage, p53 is activated by other stress situations, including oxidative stress and ribonucleotide depletion. In non-dividing cells, p53 protein levels are kept low by binding to the protein MDM2 (also called HDM2 in humans); the MDM2:p53 complex migrates from the nucleus to the cytosol where the p53 is ubiquitinated which targets it for turnover by the proteosome. Other members of the p53 gene family, TP63 and TP73, are much less frequently mutated in cancer. However, Monti et al. (2017) reported that TP63 was mutated in about 15% of melanomas, and its mutation rate correlated with UV exposure. In AML, p53 was less often mutated (5–10%) (Hunter and Sallman 2019). Mutations in one allele of the p53 gene are often followed by the loss of the remaining wild-type p53 allele (Ghaleb et al. 2020). This phenomenon, loss of heterozygosity, is a frequent aspect of the chromosomal instability that occurs during tumour progression.

50

3

Cancer as a Disease of Defective Cell Cycle Checkpoint Function

Fig. 3.2 Components of the G1:S checkpoint

3.4

The G1:S Checkpoint

Another component of the G1:S checkpoint that was discovered as a result of a congenital cancer susceptibility is the retinoblastoma protein, pRb. Retinoblastoma is a tumour of infancy associated with a congenital mutation in the RB1 gene. Since both alleles of this tumour suppressor gene need to be affected for retinoblastoma to occur, it is believed that affected individuals inherit one mutant gene, and that a second, somatic mutation occurs during embryonic development. The pRB pathway is summarised in Fig. 3.2 (Aguda 2001). pRb binds and inactivates the transcription factor E2F. Activity of E2F is essential for the cell to progress from G1 phase into S phase of the cell cycle. If pRb is sequentially phosphorylated by the cyclin-dependent kinases cdk4 and cdk2, the phosphorylated pRb releases E2F, which is thus activated, and stimulates transcription of myc and other genes required for progression into S phase. When the checkpoint is activated, for example by DNA damage, activated p53 causes transcription of the tumour suppressor protein p21, which inhibits both cdk4 and cdk2, thus preventing phosphorylation of pRb, so that E2F remains inactive and cell cycle progression does not occur. Another inhibitor of cdk4 and cdk2, p27, acts similarly to p21, but is activated by different transcription factors, primarily FoxO, in response to cytokines. The p16 protein is primarily a cdk4 inhibitor; expression of p16 is driven by the transcription factors CTCF, Ets, and Sp1. High-level expression of p16 results in cellular senescence (Rayess et al. 2011). Conversely, loss of p16 function causes override of the G1:S checkpoint. In cultured fibroblasts from mice with the premature ageing disease, Werner’s syndrome, mutations in p16 resulted in apparent reversal of senescence (Wu et al. 2012). The requirement of RB for two cyclin-dependent kinases for complete phosphorylation means that the checkpoint acts as an AND gate. Lifting the checkpoint requires that both cyclin D (indicating the presence of a growth factor signal) AND cyclin E (indicating that the cell is in late G1 phase) are present.

3.5

Mutations that Inactivate or Over-Ride the Checkpoint

51

What is the biological role of the G1:S checkpoint? Early in embryological development, mammalian cells may divide every 24 hours or less. Clearly, such exponential growth cannot continue for very long. In adult tissues, cells divide much more slowly, and some (e.g. nerve cells) not at all. Epithelial tissues, such as skin, or the inner lining of the intestine, regularly slough dead cells from their surface, and these must be replaced. Some cell division takes place to repair tissue damage. Fibroblasts, found in connective tissue, do not normally divide, but in wound healing, where a rapid response is necessary, are capable of doubling every 12 hours. The G1:S checkpoint acts as a brake on cell division. Normally the brake is ON, so cell division does not occur. When necessary, to replace rapidly turning over cells, or to repair tissue damage, the brake is released (Aguda 2001). In tumours, the G1:S checkpoint is always either defective, or over-ridden by a sort of permanent wound-healing response.

3.5

Mutations that Inactivate or Over-Ride the Checkpoint

Mutation or abnormal expression of any of the components shown in Fig. 3.2 can inactivate the checkpoint. Mutant, inactive, P53 occurs in half of all human tumours. As noted above, biallelic p16 mutations result in an inactive checkpoint. P16 is frequently deleted or mutated in pancreatic and lung cancers, and in melanoma, and is often silenced by hypermethylation in oesophageal cancer. Many tumours have upregulation of cyclin D or cyclin E or both. This is probably the result of constitutive activation of one of the signalling pathways that control cyclin expression. Many human breast cancers have constitutively activated HER2, a member of the EGF receptor family that normally requires activation by its ligand, EGF, but which, when mutated, is active in absence of ligand binding. Another very common mutation in many human cancers is constitutive activation of Kras resulting in permanent activation of the MAP kinase signalling pathway (Fig. 1.1) and consequent upregulation of cyclin D. Constitutive expression of c-myc, as happens in Burkitt’s lymphoma, will also override the G1:S checkpoint. Loss of the G1:S checkpoint results in more rapid entry of cells into S phase, and an increased rate of DNA synthesis. However, one of the enzymes whose transcription is driven by normal p53 is the p53R2 subunit of ribonucleotide reductase, the rate-limiting enzyme for biosynthesis of deoxyribonucleoside triphosphates (dNTP) required for DNA synthesis. This could potentially oppose the growth-enhancing effect of p53 loss. However, Radivoyevitch et al. (2012) showed (from the Gene Expression Omnibus) that following p53 loss there were compensating increases in ribonucleotide reductase subunits R1 and/or R2 and in deoxycytidine kinase and thymidine kinase 1. These changes would compensate for loss of the p53R2 subunit of ribonucleotide reductase and enable p53-deficient cells to maintain DNA synthesis.

52

3.6

3

Cancer as a Disease of Defective Cell Cycle Checkpoint Function

Benign Hyperplasia and Premalignant Conditions

For most tissues, the only time that normal cell division proceeds continually without a brake is in the early stages of embryonic development. Certain tissues (skin, intestinal epithelium) require a baseline rate of cell division to replace normal wear and tear. Blood cells, both erythrocytes and leukocytes, turn over in the course of their normal function, and need constant topping up. For most tissues, most of the time, the default situation is for the G1:S checkpoint to be ON—that is, for cell division to be arrested. The checkpoint is maintained in the ON position by the various inhibitors of the cyclin-dependent kinases, p16, p21, and p27. In those tissues where regular cell division is required, the balance between activation and inhibition of cdk4 and cdk2 is fine-tuned by increasing the level of the activators, cyclin D and cyclin E, and this is done in response to growth factors or cytokines acting through the various signalling pathways discussed in Chap. 1. The brake is removed, or the growth stimulus increased, in various pathological conditions. Human papilloma viruses (HPV) code for two viral proteins that stimulate abnormal growth of skin cells: HPV-E6 binds and inactivates p53; HPV-E7 binds and inactivates pRb. In both cases, the result is an abnormal skin growth, a wart. Warts are an example of benign hyperplasia resulting from turning OFF the G1:S checkpoint. Warts are not malignant, though a subset of human papilloma viruses can result in premalignant lesions, that if further activated can result in cervical carcinoma. There are many other examples of G1:S checkpoint override causing benign hyperplasia. Mutation or deletion of the tumour suppressor gene, APC, causes benign colonic polyps. APC is a negative regulator of β-catenin formation, part of the Wnt signalling pathway (Fig. 1.3). Loss of APC activity thus elevates cyclin D, resulting in override of the G1:S checkpoint, and hyperproliferation of the colonic epithelium. These colon polyps are not malignant and can be removed surgically. However, they are regarded as premalignant lesions, because further mutations that result in genetic or epigenetic instability may result in transformation. It is believed that 85% of colon carcinomas result from this process (Lewis 2011). Benign hyperplasia caused by loss or override of the G1:S checkpoint is believed to be an essential step in all neoplasia: malignancy itself is the result of genetic or chromosomal instability (Chap. 2), but that instability (caused by loss or override of the SAC) on its own causes a lethal growth deficit unless it is balanced by a positive growth stimulus. Thus, formation of a malignant tumour requires both override of the G1:S checkpoint AND genetic instability, except in rare cases where there is a congenital deficiency in the G1:S checkpoint or (as discussed in Chap. 8) a single chromosomal translocation results in both genetic instability and checkpoint override. This two-stage origin of cancer should be distinguished from the early two-stage model of carcinogenesis of Moolgavkar and Knudson (1981) which distinguished initiation and progression stages of carcinogenesis. Both stages of the two checkpoint model of cancer are included in Moolgavkar and Knudson’s initiation stage;

3.7

Inhibitors of Cyclin-Dependent Kinases

53

initiation (loss of the two checkpoints) results in a tumour stem cell, which then requires further mutations to drive progression to an actual tumour.

3.7

Inhibitors of Cyclin-Dependent Kinases

Cyclin-dependent kinases (cdks) are present throughout the cell cycle, but their obligatory activator proteins, the cyclins, are highly cell cycle phase-dependent and show periodic oscillations (Fig. 3.3). Cdk4, and the closely related cdk6, when bound to cyclin D phosphorylate pRb, which is then further phosphorylated by cdk2/cyclin E. The fully phosphorylated pRb is no longer able to bind E2F which is now able to drive transcription of myc and other genes required for progression to S phase. Cdk2 can also be activated by cyclin A, which is active in S phase. Cdk1, bound to cyclin B is active in G2 phase and is essential for progression into M phase (discussed in Chap. 4). Cdk7 (bound is cyclin H) is an activator of other cyclins. Cdk9 (bound to cyclin T or cyclin K) is a component of the P-TEFb complex which is an elongation factor for RNA polymerase II-directed transcription; it phosphorylates the c-terminal domain of the large subunit of RNA polymerase II. Potent inhibitors have been discovered for many members of the cdk family. Palbociclib, discussed below, is an example of a cdk4/6 inhibitor. Seliciclib (R-roscovitine), once believed to be acting as a cdk2 inhibitor, is in fact more potent

Fig. 3.3 The CYCLOPS model of the cell cycle oscillator

54

3

Cancer as a Disease of Defective Cell Cycle Checkpoint Function

against cdk9. It, and other cdk9 inhibitors are potent blockers of RNA transcription. Selective inhibitors of cdk1 are also known. Cells in G1 phase of the cell cycle have a number of possible fates: if the G1:S checkpoint is OFF (i.e. if growth factors are present), they will progress into S phase, but in the absence of growth factors they may enter a stationary (G0) phase, in which they retain the ability to re-enter the cell cycle, or they may become senescent, or differentiate, in which case the ability to replicate is permanently lost. There have been trials of differentiating agents, such as hexamethylenebisacetamide, but, as yet, induction of tumour differentiation has not become an established method of cancer treatment. Based upon evolutionary considerations, it is to be expected that resistance to differentiation inducers would emerge faster than to most classes of anticancer drug, since cells that have not differentiated could still proliferate, and thus have a selective advantage even in the absence of the selecting agent. Differentiation is the fate of most normal cells, and part of normal maturation in developing tissues appears to be increased expression of p16 (Fig. 3.2), which by inhibiting cdk4/cyclin D, shifts the balance between proliferation and differentiation.

3.8

The CYCLOPS Model of the Cell Cycle

The prevailing view is that cancer is the result of chromosomal instability, accompanied or preceded by a mutation causing a growth advantage, and that these two changes result in a cancer stem cell. For a cancer stem cell to become a tumour requires further multiple genetic changes—malignant progression—to be discussed in detail in Chap. 9. Malignant progression is a process of cellular Darwinism, implying that we can capture its essence in a quantitative model, specifically a form of genetic algorithm, which we call the malignant progression algorithm. The algorithm is stated explicitly as software in the online supplement to Chap. 2, and can be used to make predictions, for example, of the outcome of treatment or prevention strategies. The first element of the genetic algorithm, competition for proliferation, is determined by loss of control of the G1:S checkpoint. The details of this checkpoint, and how control of it is lost in cancer, are discussed above. Loss of control of this checkpoint confers a selective reproductive advantage on cancer cells, and makes them autonomous agents, released from the constraints of physiological growth controls, but it does not make the cells cancerous, so long as the second checkpoint remains functional. Cells without a G1:S checkpoint but with an intact mitotic spindle checkpoint (SAC) may form benign tumours. They have uncontrolled growth, but remain diploid. Examples are warts, carcinoma in situ, and intestinal polyps. For the Darwinian process of malignant progression to occur, the growth must have genetic diversity. Earlier commentators on cancer as a Darwinian process have suggested that the required genetic diversity could arise by accumulation of somatic mutations. Somatic mutations certainly occur in cancer, and the various cancer genomics projects have catalogued thousands of mutations in human cancer, but they cannot tell the whole story, or even the main story, because

3.8

The CYCLOPS Model of the Cell Cycle

55

somatic mutations are compatible with the cell remaining diploid, but in fact all malignant tumours have some degree of aneuploidy; that is, they have a chromosomal number that differs from the canonical number (in human cells) of 23 pairs. They may have chromosomal deletions, rearrangements, duplications, but all malignant tumours have an abnormal chromosome complement. And this is their principal source of genetic diversity. Duesberg et al. (2006) argued that this aneuploidy was the consequence of loss of control of the SAC, and that it was the defining characteristic of cancer. The present position is that he was right, if we are concerned with the difference between a benign and a malignant tumour but that a more illuminating view is to regard cancer as a Darwinian process in which loss of the G1:S checkpoint provides the competition for proliferation, loss of control of the SAC provides the genetic diversity, the cell cycle provides the iteration with positive feedback, and the inevitable outcome is the cellular selection process that we call malignant progression. Abnormalities in both the G1:S checkpoint and the SAC are necessary and sufficient for malignant transformation. As discussed in Chap. 4, defects in the DNA damage checkpoint (DNADC) can over-ride the SAC. The DNADC and the SAC, although operating in different phases of the cell cycle, form a coordinated system for ensuring accurate chromosome segregation in dividing normal cells, and defects in any part of the system can result in aneuploidy. To express this concept mathematically, we need a model of the cell cycle that includes a detailed description of the G1:S checkpoint, its working in normal cells, and its deficiencies in tumour cells. A number of normal cell models have been reported (Novak and Tyson 2004; Novak et al. 2007). Our model of the cancer cell cycle also includes a detailed mechanistic description of the SAC (Jackson et al. 2017). The cell cycle module of CYCLOPS is based upon the model of Novak and Tyson (2004) which is a system of coupled oscillators. Several cell components oscillate; cyclin E peaks in late G1 phase, phosphorylating pRb and causing progression into S phase. Cyclin B builds up during S phase and peaks in G2. When bound to cdk1, the active cdk1/cyclin B causes progression into M phase by stimulating chromosome condensation and breakdown of the nuclear envelope. In this model, the anaphase-promoting complex (APC) is activated sharply in M phase and is degraded at the end of M. APC activation is driven by cdk1/cyclin B, which peaks in late G2 phase, and is broken down before cell division. This system is described by a system of 11 rate equations (details in the supplementary material to Jackson et al. 2017). The cell cycle in mammalian cells is not a free-running oscillator, but is interrupted by several checkpoints which must be inactivated or over-ridden for cells to progress through the cycle. The published version of CYCLOPS contains kinetic descriptions of two cell cycle checkpoints: the p53 and Rb-dependent G1/S checkpoint, and the mitotic spindle assembly checkpoint (SAC). In about half of human tumours (those with mutant p53) the G1/S checkpoint is normally OFF (i.e. progression into S phase can occur) because the E2F transcription factors required for synthesis of S phase proteins that are otherwise tightly bound to

56

3

Cancer as a Disease of Defective Cell Cycle Checkpoint Function

Fig. 3.4 Effect of a 5-minute pulse of EGF on passage of cells from G1 to S phase (v22) as modelled by SIGNAL7

pRB are released through phosphorylation of pRB by cdk4/cyclin D and cdk2/cyclin E. In tumours with normal p53, cells respond to radiation damage by turning ON the checkpoint until DNA damage has been repaired. Those tumours with normal p53 may still not have an effective G1/S checkpoint, because they usually have abnormally high levels of cyclin D or cyclin E which can over-ride an otherwise functional G1/S checkpoint. Irradiated p53-negative cells may still arrest because of activation of the DNA damage checkpoint. Cyclin D is not part of the cell cycle oscillator, but is produced by growth-factor-dependent signalling pathways, of which the MAP kinase pathway is modelled in CYCLOPS. Modelling the G1:S checkpoint makes it possible to explore the effects of inhibitors and mutations of this key regulator of the cell cycle. The program SIGNAL7 (online supplement) was used to simulate the effect of a 5 minute pulse of EGF (Fig. 3.4). The dynamics of the MAP kinase signalling pathway and the G1:S checkpoint are such that a 5-minute pulse of EGF released the checkpoint for a period of about 2.5 hours. As can be seen from Fig. 3.2, for RB to be fully phosphorylated, and thus release the checkpoint, requires the presence of both cdk4/cyclin D and cdk2/cyclin E. Cyclin E peaks in late G1 and is rapidly degraded when the cell enters S phase. Figure 3.5 shows a repeat of the simulation of Fig. 3.4, but with the cyclin E level at 2% of its peak value. The entry of cells into S phase (v22) is very low, showing that the checkpoint remains in place. For progression from G1 to S phase to occur

3.9

Palbociclib-Induced Cell Cycle Arrest

57

Fig. 3.5 Effect of a 5-minute pulse of EGF on passage of cells from G1 to S phase (v22) at low cyclin E level as modelled by SIGNAL7

requires the presence of both cyclin D (the downstream indicator of growth factor presence) AND cyclin E (showing that the cell is in late G1 phase). Figure 3.6 shows the rate of progression of cells from G1 into S phase in cells with constitutively activated Ras. Even in the absence of EGF, mutant, constitutively activated Ras results in permanently elevated levels of cyclin D. However, cyclin E is still required for passage of cells into S phase. Mutant Her2 has the same effect as mutant Ras, as it keeps Ras activated in the absence of EGF.

3.9

Palbociclib-Induced Cell Cycle Arrest

Palbociclib (Fig. 3.7) is a selective inhibitor of cdk4 and the closely related cdk6 (Fry et al. 2004). It is a potent inhibitor of proliferation of Rb + cells. It is used clinically in the treatment of breast cancer. When treated with palbociclib, cells are arrested at the G1:S boundary. If the inhibitor is removed after a few hours, most of the arrested cells will progress into S phase. However, prolonged arrest will trigger apoptosis. Cells that lack a functional G1:S checkpoint are relatively unaffected. It has been suggested that selective blockade of normal cells could be used to protect them from cytotoxic agents that act in S phase, an approach termed cyclotherapy (Blagosklonny and Pardee 2001). The CYCLOPS program was used to model the combination of palbociclib with the S/G2-phase agent, gemcitabine. It predicted that following

58

3

Cancer as a Disease of Defective Cell Cycle Checkpoint Function

Fig. 3.6 Progression of cells from G1 to S phase (v22) in cells with constitutively activated Ras as modelled by SIGNAL7 Fig. 3.7 Palbociclib, an inhibitor of cdk4 and cdk6

palbociclib treatments with delayed gemcitabine could increase the tumour/normal selectivity of gemcitabine by up to 17-fold. Figure 3.8 shows the calculated effect of palbociclib inhibition of cdk4/cyclin D in cells with wild-type and mutant Ras. The Ras mutants were inhibited, but much less than wild-type cells. Ras mutations do not directly affect cdk4, but result in production of more cyclin D. A newer cdk4/6 inhibitor abemaciclib is approved for treatment of “triple negative” breast cancer (ER-. PR-, HER2-). X-ray damage to DNA has three effects: it induces repair of the DNA strand breaks, it starts a clock that will trigger apoptosis if the DNA damage has not been repaired within several hours, and it activates p53. P53 induces p21, which inhibits cdk4 and cdk21, thus preventing RB phosphorylation required for activation of the E2F transcription factor. Figure 3.9 shows a simulation by signal7 of the effect of a

3.9

Palbociclib-Induced Cell Cycle Arrest

59

Fig. 3.8 Calculated effect of palbociclib on tumour cells with wild-type and mutant Ras. V[23] is the rate of passage of cells from G1 phase into S phase. The degree of inhibition is expressed as fractional activity, where 1,0 is fully active, and 0 is fully inhibited

Fig. 3.9 Effect of 4 Gray dose of X-irradiation on the G1 checkpoint, following a 5 min pulse of 1 nM EGF, modelled by the signal7 program. V23 is the rate of progression of cells from G1 into S phase

60

3

Cancer as a Disease of Defective Cell Cycle Checkpoint Function

4 Gray radiation dose. P21 is rapidly activated, and subsequently declines as the DNA damage is repaired. Passage of cells from G1 into S phase is inhibited, but starts to recover after about 4 hours, and is restored to normal by 12 hours. This effect is similar in cells with mutant Ras, but requires wild-type p53. In the presence of mutant p53, radiation does not trigger the G1:S checkpoint. P53 acts as a general stress sensor, detecting a number of conditions that indicate the cell is not ready to move into S phase, including oxidative stress, or insufficient levels of dNTPs (substrates for DNA polymerase) for DNA synthesis to begin.

3.10

Selectivity of Signalling Pathway Inhibition Against Mutants

The availability of a kinetic model of MAPK signalling makes it feasible to predict which inhibitors will have selectivity against particular mutants. Oncogenic mutations fall into three classes: – overexpression or gene amplification – constitutive activation – loss of negative feedback control These mutations may occur in any of the five types of functional components of signalling pathways (receptors, signal amplifiers, pulse expanders, logic gates, and transcription factors), meaning there are fifteen kinds of oncogene. Each of the five kinds of signalling modules may be inhibited at substrate and regulatory sites, meaning there are ten basic kinds of signalling inhibitor. This analysis was applied to the simplified model of the MAPK pathway illustrated in Fig. 3.10. The SIGNAL7 model was used to predict the relative activity, compared to wild type, of ten classes of signalling inhibitor against fifteen classes of oncogenic mutation. Results are summarised in Table 3.1, as ratios of IC50 values against the mutants compared to wild type. In many cases, the ratios are close to unity, meaning there is no antitumour selectivity. In some cases, the ratio is greater than one: for example, a tumour cell with constitutively activated Ras is unlikely to respond to an inhibitor of an upstream kinase. However, there are other cases where the ratio is 80 >80 >80 >80 0.22 0.37

M5 0.94 0.98 3.94 4.10 1.46 1.46 1.44 1.54 0.94 0.97

M6 0.93 0.95 1.04 4.53 1.00 1.00 1.00 1.00 0.92 0.94

M7 0.44 0.59 2.72 3.02 2.87 3.57 2.56 3.50 0.40 0.54

M8 0.32 0.49 2.68 3.07 3.71 3.78 2.67 3.56 0.28 0.42

M9 0.99 1.00 0.99 0.99 1.00 4.96 0.99 1.00 0.99 1.00

M10 0.27 0.44 2.85 3.27 >80 >80 10.4 15.8 0.23 0.38

M11 0.27 0.45 2.87 3.28 7.27 7.46 7.21 10.1 0.24 0.38

M12 0.61 0.71 0.75 0.78 0.85 0.85 0.85 3.93 0.58 0.68

M13 0.28 0.46 >80 >80 >80 >80 >80 >80 0.22 0.35

M14 1.06 1.04 1.06 1.06 1.02 1.02 1.02 1.04 5.55 5.51

M15 1.02 1.01 1.02 1.02 1.02 1.01 1.01 1.02 1.60 5.22

3

M1 has a constitutively activated transcription factor M2 has 5-fold upregulation of TF M3 has TF with 5-fold loss of feedback inhibition M4 has constitutively activated Ras M5 has 5-fold upregulation of Ras M6 has Ras with 5-fold weaker feedback inhibition M7 has constitutively activated RTK M8 has 5-fold upregulation of RTK M9 has RTK with 5× weaker feedback inhibition M10 has constitutively activated SOS M11 has 5-fold upregulation of SOS M12 has 5-fold upregulation of MEK M13 has MEK with 5× weaker feedback inhibition M14 has 5-fold upregulation of myc M15 has myc with 5-fold weaker feedback inhibition Drugs 1 and 2 are transcription factor inhibitors Drugs 3 and 4 are Ras inhibitors Drugs 5 and 6 are RTK inhibitors Drugs 7 and 8 are GTP exchange factor (SOS) inhibitors Drugs 9 and 10 are MEK inhibitors

IC50 (mutant)/IC50 (wild type) Drug M1 M2 M3 1 0.06 1.23 1.56 2 0.16 1.20 5.06 3 >80 1.02 1.01 4 >80 1.02 1.01 5 >80 1.06 1.06 6 >80 1.01 1.01 7 >80 1.01 1.01 1.01 8 >80 1.01 9 >80 1.02 1.01 10 >80 1.01 1.01

Table 3.1 Selectivity of inhibitors of MAPK signalling against different classes of mutants

62 Cancer as a Disease of Defective Cell Cycle Checkpoint Function

References

63

References Aguda BD (2001) Kick-starting the cell cycle: from growth-factor stimulation to initiation of DNA replication. Chaos 11:269–278 Blagosklonny MV, Pardee AB (2001) Exploiting cancer cell cycling for selective protection of normal cells. Cancer Res 61:4301–4305 Duesberg P, Li R, Fabarius A, Hehlman R (2006) Aneuploidy and cancer: from correlation to causation. Contrib Microbiol 13:16–44 Fry DW, Harvey PJ, Keller PR et al (2004) Specific inhibition of cyclin-dependent kinase 4/6 by PD0332991 and associated antitumor activity in human tumor xenografts. Mol Cancer Therap 3:1427–1438 Ghaleb A, Padellan M, Marchenko N (2020) Mutant p53 drives the loss of heterozygosity by the upregulation of Nek2 in breast cancer cells. Breast Cancer Res 22:133 Hunter AM, Sallman DA (2019) Current status and new treatment approaches in TP54 mutated AML. Best Pract Res Clin Haematol 32:133–144 Jackson RC, Di Veroli GY, Koh S-B et al (2017) Modelling of the cancer cell cycle as a tool for rational drug development: a systems pharmacology approach to cyclotherapy. PLOS Comp Biol 13:e1005529. https://doi.org/10.1371/journal.pcbi.1005529 Lewis R (2011) Human genetics: the basics. Routledge, Oxford, p 105 Monti P, Ghiorzo P, Menichini P et al (2017) TP63 mutations are frequent in cutaneous melanoma, support UV etiology, but their role in melanomagenesis is unclear. Oncol Rep 38:1985–1994 Moolgavkar SH, Knudson AG (1981) Mutation and cancer: a model for human carcinogenesis. J Natl Cancer Inst 66:1037–1052 Novak B, Tyson JJ (2004) A model for restriction point control of the mammalian cell cycle. J Theor Biol 230:563–579 Novak B, Tyson JJ, Gyorffy B, Csikasz-Nagy A (2007) Irreversible cell cycle transitions are due to systems-level feedback. Nat Cell Biol 9:724–728 Radivoyevitch T, Saunthararaja Y, Pink J et al (2012) dNTP supply gene expression patterns after P53 loss. Cancers (Basel) 4:1212–1224 Rayess H, Wang MB, Srivatsan ES (2011) Cellular senescence and tumor suppressor gene p16. Int J Cancer 130:1715–1725 Regev A, Teichmann SA, Lander ES et al (2017) The human cell atlas. Elife 6:e27041. https://doi. org/10.7554/elife.27041 Steel GG (1977) Growth kinetics of tumours. Oxford University Press, Oxford Wu X, Jia S, Zhang X et al (2012) Two mechanisms underlying the loss of p16(Ink4a) function are associated with distinct tumorigenic consequences for WS MEFs escaping from senescence. Mech Ageing Dev 133:549–555

Chapter 4

The DNA Damage Checkpoint

Abstract The DNA damage response (DDR), also known as the DNA damage checkpoint (DNADC) acts to prevent cells with damaged DNA from entering mitosis. Kinds of DNA damage that can trigger the checkpoint include singlestrand breaks, double-strand breaks, stalled replication forks, and premature chain termination. Many tumours have mutations or deletions of one or more components of the DDR. For this reason, tumours are often repair-deficient, so that DNA-damaging drugs tend to have antitumour selectivity. Inhibitors of components of the DDR, such as PARP, or chk-1, will potentiate the antitumour activity of DNA-damaging drugs, or of radiation. Where a tumour has a defective DDR, inhibitors of the pathway may result in synthetic lethality; e.g. olaparib, a PARP inhibitor, is highly selective against BRCA-deficient tumours. There is cross-talk between the DDR and the SAC, resulting in regulatory coordination. Even if a cell with unrepaired DNA damage enters mitosis, it may be unable to progress to anaphase, in which case it will undergo apoptosis. A model of the DDR can be used to design treatment with specific activity against tumours with particular mutation profiles.

Random nucleotide substitutions in the gene yield corresponding changes in anatomy, physiology, or behavior ... Genetic change is also initiated when genes shift position on the chromosomes, or when the number of chromosomes...is raised or lowered. Edward O. Wilson, The Diversity of Life (1992).

4.1

DNA Repair Pathways

It is because the effects of genetic changes or genetic damage are so profound that the cell has evolved multiple, complex pathways to prevent and correct them. During S phase of the mammalian cell cycle, the DNA of the cell must be replicated once, Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_4. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_4

65

66

4 The DNA Damage Checkpoint

and only once. The resulting helical coils are themselves coiled (supercoiling). In G2 phase the replicated, but tangled, strands of DNA are decatenated by topoisomerase enzymes The function of the DNA damage checkpoint (DNADC) is to ensure that the newly replicated DNA is complete and undamaged before the cell progresses into mitosis. The DNADC is usually referred to as the G2 checkpoint in the earlier literature, but it can also function in late S phase. The term “DNA damage response” (DDR) is also widely used; our present discussion uses the “checkpoint” terminology when necessary to emphasise its all-or-none kinetics. Various forms of DNA damage can be repaired by a normal cell and there are reviews by Löbrich and Jeggo (2007), Toledo et al. (2017), and Alkan et al. (2018). Single-strand breaks (SSB) occur naturally during the topoisomerisation process, do not affect the threedimensional structure of DNA, and can be readily and rapidly repaired. In addition to spontaneous DNA strand breaks formed as part of the replication process, DNA damage may be caused by radiation or by chemicals. An important marker of DNA damage is poly-(ADP ribose), produced by poly-(ADP ribose) polymerase (PARP). SSB repair involves XRCC1 (a scaffold protein without enzymatic activity) and DNA ligase 3, and is usually rapid and complete. Double-strand breaks, however, are potentially lethal. The damage sensor for DSB is a complex of proteins centred around NBS1, which activates ATM. There are several mechanisms of DSB repair: homologous recombination (HR) uses a sister chromatid or homologous chromosome as a template. It is the most faithful form of DSB repair. Non-homologous end joining (NHEJ) is involved in normal function of the immune system in recombination in B cell and T cell receptors and other situations where DSB must be repaired without access to a template; it requires DNA ligase IV. NHEJ is DNAPK dependent. Another route of DSB repair, microhomology-mediated end joining (MMEJ) consists of removal of nucleotides around the break by MRE11 nuclease until short homology regions (microhomologies) are found. PARP1 is required for MMEJ. It has been claimed that a single unrepaired DSB was sufficient to prevent cell cycle progression, though this view has come under challenge (Shaltiel et al. 2015). In addition to SSB and DSB, several other kinds of DNA damage can be repaired by the cell. Stalled replication forks are the situation where strand elongation has failed during DNA replication, and must be cleaved so that DNA elongation can be re-started. In the case of SRF, single-strand regions downstream from the replication fork, resulting from the activity of DNA helicases, activate a complex group of proteins centred on replication protein A (RPA). RPA activates 9-1-1 proteins, which in turn activate ATR. The chain-terminating nucleotide (e.g. gemcitabine) is removed by nucleotide excision, and the DNA is repaired by DNA polymerase β and DNA ligase 3. DNA cross-links are covalent chemical links between complementary DNA strands. If they cannot be removed before the cell enters mitosis, normal cell division will be impossible and the cell will die. Chemical substitutions that involve a single DNA strand are termed mono-adducts. Removal of chemical substitutions on the O-6 position of DNA guanine bases by guanine O6-methyltransferase was described in Chap. 2. There are other mechanisms for removal of mono-adducts at other sites. DNA bases with chemical damage, including oxidative damage, are removed by the process of base excision repair. The damaged bases are recognised

4.2

Selectivity of DNA-Damaging Drugs

67

and cleaved by DNA glycosylase enzymes, leaving an apurinic/apyrimidinic (AP) site. The sugar-phosphate backbone is then cleaved by an AP endonuclease, and the resulting breaks are patched by a specific polymerase, DNA polymerase beta in conjunction with DNA ligase 1 or DNA ligase 3. Some damaged DNA bases are not recognised by the DNA glycosylases, so cannot be removed by the base excision repair pathway. An example is the thymidine dimers caused by ultraviolet irradiation. An alternative pathway, nucleotide excision repair, is able to remove these bases. Nucleotide excision repair is particularly important in skin, and a genetic defect in this pathway results in the hereditary skin condition, xeroderma pigmentosum. Another pathway, DNA mismatch repair, recognises, not damaged bases, but incorrect Watson–Crick base-pairing. It operates specifically on the newly-synthesised DNA strand. Mismatch repair is thus a “proof-reading” system. Proliferating cell nuclear antigen (PCNA) is involved in the mismatch recognition. Exonucleases remove the mismatched nucleotide and a number of adjacent nucleotides. The resulting single-strand region is patched by DNA polymerase 3 and DNA ligases. Defects in DNA mismatch repair may result in microsatellite instability (discussed in Chap. 2).

4.2

Selectivity of DNA-Damaging Drugs

Depending on their detailed mechanism of action, DNA-damaging drugs can form different kinds of DNA damage, some of which are more easily repaired than others. In general, the normal tissues most affected by anticancer drugs are those with a rapid replication rate, particularly bone marrow and gastrointestinal epithelium. Most tumours do not have faster replication rates than replicating normal tissues. However, as discussed below, many tumours have defects in one or more of the DNA repair pathways, and this contributes to their selectivity. Some classes of DNA-active drugs, such as the nitrosoureas (Chap. 2) cause guanine O-6 crosslinking, and are thus most active against mer- cells. However, haematopoietic stem cells (the common precursor for red and white blood cells and platelets) are monoadduct excision repair negative (mer-), so nitrosoureas can cause dangerous depletion of platelets and other blood components. Other alkylating drugs, such as cisplatin, give guanine N-7 cross-links, and cause neutropenia, but with much lesser effects on platelets. The selectivity of DNA-damaging drugs can be largely understood in terms of the types of DNA lesions they cause, and the relative activity of the DNA repair pathways in tumours and normal tissues.

68

4.3

4

The DNA Damage Checkpoint

Mutant or Abnormally Expressed DNA Repair Enzymes in Tumours

It seems likely that the antitumour selectivity of DNA-damaging drugs is largely attributable to the fact that many, perhaps most tumours have mutations in components of the DDR. The tumour suppressor gene, p53, was discussed in Chap. 3 as a component of the G1:S checkpoint. However, p53 is present in later stages of the cell cycle,where it is also activated by DNA damage, and can cause G2 arrest in cells that lack a functional G1:S checkpoint. Its interactions in G2 phase are complex, but (as in the G1:S checkpoint) p21 expression is involved (Kim et al. 2004). Compounds that block activity of the DDR preferentially synergise with DNA-damaging agents in cells with mutant p53 (O’Connor 1997). BRCA1 and BRCA2 are proteins involved in HR, so required for DSB repair. Mutations in BRCA1 or BRCA2 result in increased risk of a number of cancers, particularly breast cancer. For this reason they are described as cancer susceptibility genes. In prostate cancer, BRCA2 mutations are often associated with particularly aggressive disease (Das et al. 2019). PARP inhibitors inhibit SSB repair, and BRCA1 or BRCA2 mutants, which have impaired repair of DSB, are thus particularly sensitive to PARP inhibitors such as olaparib, an example of synthetic lethality. DNA-dependent protein kinase (DNAPK) is involved in NHEJ. DNAPK promotes genomic stability and cell viability following exposure to replication stress. Artemis is a protein involved in generating sequence diversity in immune function. It is a substrate of DNAPK, and as such is necessary for NHEJ. NBS1, another protein involved in NHEJ, is over-expressed in many tumour types. It acts as a damage sensor for DSB. RAD51, required for HR, is abnormally expressed in many cancers; mutant RAD51 is another cancer susceptibility gene. Somatic mutations in ATM frequently occur in lymphoid malignancies, and germ-line ATM mutations result in the ataxia telangiectasia syndrome, which gives a 20%–30% lifetime risk of many tumour types (Choi et al. 2016). DNA ligase IV (lig4) prevents replication fork stalling and is frequently amplified in cancer (Joshi et al. 2019). DNA polymerase beta is mutated in about 30% of human cancers. Of these various components of the DDR that are frequently mutated or have expression changes in cancer, p53, BRCA1 and ATM are modelled in the checkD program (online supplement and Fig. 4.1). Other components of the DNA damage pathway that frequently show expression changes in cancer include ERCC1, MSH3, PAXIP1, PCNA, and RFC. The fact that a gene coding for a component of the DDR is mutated in cancer does not necessarily make it a driver mutation. The function of the DDR is to detect DNA damage, and when it is found, to cause cell cycle arrest. If the DNA damage is not repaired within a few hours, rather than allow the damaged cell to enter mitosis, it will undergo apoptosis. If, because of a mutation, the DNA damage is not promptly repaired, this need not contribute to tumour promotion so long as the damaged cell self-destructs. However, if the damaged cell does not enter apoptosis, but enters mitosis with unrepaired DNA damage, this can result in a clone of cells with

4.4

Cross-Talk Between the DNA Damage Checkpoint and the SAC

69

Fig. 4.1 Free RPA is regenerated when RPA-SRF are repaired

abnormal DNA content or chromosomal organisation, increasing cellular heterogeneity and possibly accelerating tumour progression. RAD51 mutations are frequently reported in cancer and are likely to result in inefficient HR. In pancreatic cancer, Goral (2015) reported mutations in several components of the DDR, including ATM and BRCA2. Ten percent of pancreatic cancer cases have a familial occurrence, suggesting germ-line mutations. In prostate cancer, BRCA1, BRCA2, CHK2, ATM, and PALB2 are frequently mutated, as well as DNA mismatch repair genes MLH1, MSH2, MSH6, and PMS2 (Das et al. 2019). Polymorphisms in DNA repair genes have been shown to influence the risk of colorectal cancer (Al-Shaheri et al. 2020).

4.4

Cross-Talk Between the DNA Damage Checkpoint and the SAC

The DDR functions to prevent cells with unrepaired damage from progressing into mitosis. If, because of mutations in DDR components (as in many tumours) the checkpoint fails, another quality control feature is that the mitotic spindle assembly checkpoint (SAC), discussed in Chap. 5, provides another opportunity for cells with damaged chromosomes to be eliminated. There are a number of points of cross-talk between the DDR and the SAC. For example, chk-1 phosphorylates aurora kinase B. Inhibition of BRCA1 causes down-regulation of multiple M phase enzymes, including BUB1, NEK2, and PLK (Ferrari 2006; Bae et al. 2005). In fact, the DDR and SAC, although active in different phases of the cell cycle, appear to act as a single coordinated system. DSB repair can also occur in M phase, and when it does, can result in genomic/ chromosomal instability (Terasawa et al. 2014). M phase-specific phosphorylation of XRCC4 (a regulatory subunit of DNA ligase IV) by PLK1 maintains genomic stability by preventing DSB repair in M phase. Partial inhibition of Cdk1/cyclin B also causes SAC over-ride (McCloy et al. 2014).

70

4.5

4

The DNA Damage Checkpoint

Replication Catastrophe

Toledo et al. (2017) reported the existence of a DNA replication checkpoint, whose function is to ensure that DNA replication is complete before the cell can move out of S phase. This checkpoint may be a function of the DDR, or at least involved several of the same components, including chk1 and ATR. Malfunction of this checkpoint results in catastrophic genome disruption, DNA breakage, and cell death. If SRF do not bind to RPA they may become DSB, which may overwhelm the repair pathways. Toledo et al. (2013) proposed that the checkpoint is triggered by exhaustion of replication protein A (RPA) which binds to stalled replication forks, and that ATR averts replication catastrophe by preventing RPA depletion (Fig. 4.1). Synthesis of RPA is activated by Chk1. Inhibitors of chk1 cause replication catastrophe by increasing the activation of cdk2 which increases the number of replication forks while reducing their stability (King et al. 2015). When cyclin A appears in S phase, it dislodges cyclin E from cdk2, terminates assembly of the pre-replication complex, and initiates DNA replication. Chk1 blocks initiation of DNA synthesis and ensures that DNA replication occurs only once per cell cycle. Chk1 inhibition can lead to increased initiation of DNA synthesis.

4.6

Modelling Inhibition of the DDR

Alkan et al. (2018) published a detailed (98 reaction) mechanistic model of the DNA damage response and used it to predict the response of osteosarcoma cells to DNA-damaging drugs, alone and in combination with inhibitors of the DDR. Their objective is to use the model to identify optimal drug combinations that can be matched to individual tumour genotypes. The Alkan model does not appear to include PARP or BRCA1. When modelling complex systems, the objective is usually to identify the simplest model that retains the essential behaviour of the system under study. Depamphilis et al. (2012) concluded that only eight protein kinases (of 557 protein kinases in the human genome) controlled mammalian DNA replication. These are cdk1, cdk2, cdk4, cdk6, cdk7, chk1, chk2, and cdc7 (the “octet”). In fact, they noted that only four of these enzymes are essential for mammalian development. Depending upon the environmental conditions, some cells are developmentally programmed to switch from cell division to endocycles, a process in which multiple rounds of DNA replication occur in the absence of mitosis or cytokinesis, resulting in non-proliferating polyploid cells. They proposed that cancer cells could be induced to begin a second round of DNA replication before mitosis is complete, resulting in non-proliferating cells. Of the octet, CYCLOPS (Chap. 3) already models cdk1, cdk2, cdk4, and chk1. Cdc7 (a serine/threonine kinase) is involved in initiation of replication. In

4.6

Modelling Inhibition of the DDR

71

Fig. 4.2 The Check D model of the DNA damage checkpoint

association with Dbf4 it acts as a helicase at the origin of replication. It has a constant protein level through the cell cycle but kinase activity increases in S phase. Cdc7 overexpression may be associated with neoplastic transformation. Loss of cdc7 function causes G2 arrest. Cdk7 acts both as a CAK (cdk-activating kinase, required for activation of cdk1, cdk2, and cdk4) and as a component of the transcription factor TFIIH, which activates the C-terminal domain of RNA polymerase II. TFIIH, which also contains cyclin H, is involved in nucleotide excision repair. The checkD model of DDR kinetics can be integrated into the CYCLOPS cell cycle model (Chap. 3) which already contains kinetic descriptions of the G1:S and spindle assembly checkpoints. The resulting cell cycle model thus describes all the checkpoints that function abnormally in tumours, and drugs that act on those checkpoints. Some of the reactions involved in repair of SSB, DSB, and stalled replication forks (SRF) are summarised in Fig. 4.2. These reactions are included in the checkD model of the DNA damage checkpoint (online supplement) which provides a quantitative description of DNADC dynamics. CheckD models the essential activities of the checkpoint which are (i) to detect DNA damage; (ii) if damage is detected to arrest cell cycle progression (iii) to start a clock; (iv) to repair the DNA damage, and (v) if the damage cannot be repaired within a period of about 8 hours, to activate apoptosis (Fig. 4.3). DNA containing

72

4

y DNA damage?

(cell cycle arrest) (inhibit survivin) (activate repair)

The DNA Damage Checkpoint

n caspase 3 activated?

n cycle progression

return to start

y apoptosis

Fig. 4.3 Overview of cell responses to DNA damage

SSB activates PARP, which polymerises poly-(ADP ribose) around the break. This attracts repair enzymes, including DNA polymerase-β and DNA ligase 3. Wee-1 is involved in the cell size checkpoint. It phosphorylates and inhibits cdk1/cyclin B (maturation promoting factor, MPF), thus inhibiting entry into mitosis and prolonging G2. Cdk1 is activated by the phosphatase cdc25C. Cdk1 then activates cdc25 and inactivates wee-1, creating a positive feedback loop. Cdc25C is activated by PLK-1 and inactivated by chk-1. Chk2 also inhibits cdc25c. The availability of a model of the checkpoint may assist in the process of optimising treatment of tumours with particular mutations or expression changes, and helping to make possible individualised treatment. With multiple changes in the DNADC often present in a single cell, and multiple feedback effects, both negative and positive, modelling provides a way of estimating the net effect of these multiple changes on checkpoint function.

4.7

Kinetics of the Checkpoint

Wee1 inhibits cdk1/cyclin B, and cdk1/cyclin B inhibits Wee1. This constitutes a negative autocatalytic loop. Cdc25C activates cdk1/cyclin B, and cdk1/cyclin B activates cdc25C. This constitutes a positive autocatalytic loop. ATR and chk1 form a 2-stage amplifier. These features suggest that the system is sensitive to very small amounts of DNA damage. The autocatalytic loops are a common feature of systems that show Boolean (ON/OFF) kinetics (Jackson 2017). If this is also the case for the DNA damage checkpoint, this simplifies modelling, since the checkpoint is either ON or OFF, and knowledge of detailed kinetic parameters is not required. One of the objectives of the DNADC study is to determine whether the DNA damage checkpoint follows Boolean kinetics. Preliminary data suggest that this is the case (Table 4.1). In the presence of the default activity of cdc25C, the APC is activated approximately 21 hours from the start of the cycle. Even 1% inhibition of cdc25C caused many hours of progression delay, and 3% inhibition caused essentially indefinite progression delay (Fig. 4.4). Depending upon the cell line, progression delay of 6–12 h causes apoptosis (in the presence of DNA damage), meaning that activation

4.7

Kinetics of the Checkpoint

Table 4.1 Effect of inhibition of cdc25C on cell cycle progression as modelled by checkD

73 [cdc25C] 1.01 1.00 0.99 0.985 0.98 0.97 0.96 0.95

Time to induction of APC (h) 14 21 34 50 77 >485 >485 >485

Fig. 4.4 Dependence of time to progression into M phase on activity of cdc25C. In this figure, times are measured from the beginning of G2 phase

Table 4.2 Effect of SRF on cell cycle progression as modelled by checkD

SRF at time 0 0 5 10 15 20 25 30 35 40 and higher

Time to progression into M (h) 2.0 3.0 4.2 5.0 5.4 5.8 6.2 6.6 Apoptosis at 6.6 h

of the checkpoint is Boolean. The very high order kinetics of inhibition of cdc25C is presumably the result of the positive feedback loop and the double-negative feedback loop in activation of cdc25C (Fig. 4.2). The checkD program was used to predict cell cycle progression delay caused by SRF (Table 4.2). Depending upon the parameter values for apoptosis and for DNA repair in the cell line being modelled, if the strand breaks have not been repaired by a

74

4

The DNA Damage Checkpoint

Table 4.3 Program checkD modelling the DNA damage checkpoint Time (h) 0.000 0.800 1.600 2.400 3.200 3.999 4.799 5.600 6.401 6.662

cyclinB 284.487 292.055 299.364 306.399 313.160 319.645 325.847 331.756 337.356 339.114

APC 0.042 0.040 0.040 0.040 0.041 0.041 0.041 0.041 0.042 0.042

cyclinE 0.069 0.069 0.069 0.069 0.069 0.069 0.069 0.069 0.070 0.070

SRF 40.0 26.0 16.9 10.9 7.1 4.6 3.0 1.9 1.3 1.1

CASP9 0.081 0.100 0.103 0.104 0.104 0.105 0.106 0.108 0.118 4.371

CASP3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 1.023

cdc25C 1.000 0.995 0.995 0.995 0.995 0.995 0.995 0.995 0.995 0.995

CASP3 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

cdc25C 1.000 0.995 0.995 0.995 0.995 0.995 0.995 0.995 0.995 1.000

APOPTOSIS Table 4.4 Program checkD modelling the DNA damage checkpoint Time (h) 0.000 0.800 1.600 2.400 3.200 4.000 4.800 5.600 6.400 6.619

cyclinB 284.487 292.055 299.364 306.399 313.160 319.645 325.847 331.756 337.356 337.974

APC 0.042 0.040 0.040 0.040 0.041 0.041 0.041 0.041 0.042 0.504

cyclinE 0.069 0.069 0.069 0.069 0.069 0.069 0.069 0.069 0.070 0.070

SRF 35.0 22.7 14.8 9.6 6.2 4.0 2.6 1.7 1.1 0.0

CASP9 0.081 0.100 0.103 0.104 0.104 0.105 0.106 0.108 0.118 0.159

Progression into M phase

critical time, the cell enters apoptosis. If the strand breaks are repaired before the critical time, the cell enters mitosis. The checkD model predicted that RPA depletion decreased the proportion of SRF entering repair. Table 4.3 shows an example of program output when unrepaired DNA damage remains at the critical time and the cell enters apoptosis. Caspase 9 reaches the threshold activity for activation of procaspase 3, and active caspase 3 appears. Table 4.4 shows an example of program output when DNA damage is repaired before the critical time is reached, and the cell re-enters the cell cycle. Modelling studies have been carried out with six inhibitors of the DNADC: MK1775 (inhibitor of Wee1), LY2603618 (inhibitor of chk1), NU6027 (inhibitor of ATR), BI2536 (inhibitor of PLK1), olaparib (inhibitor of PARP), and RO3306 (inhibitor of cdk1/cyclin B). Activation of cdk1/cyclin B by the Wee1 inhibitor MK1775 was modelled in the absence of DNA damage. In these simulations, Wee1 = 1.2, Vm[11] = 0.01. This slightly elevated activity of Wee1 caused indefinite G2 arrest, which was overcome

4.7

Kinetics of the Checkpoint

75

Fig. 4.5 Activation of cdk1/cyclin B by the Wee1 inhibitor, MK1775

Fig. 4.6 Effect of the chk1 inhibitor LY2603618 on time to mitosis and unrepaired DNA damage at mitotic entry

by the Wee1 inhibitor, MK1775, whose dose–response curve was almost vertical (Fig. 4.5). Modelling the chk1 inhibitor LY2603618 in the presence of DNA damage showed a concentration-dependent decrease in time to mitosis. At low concentrations of the inhibitor, cells progressed into mitosis with no unrepaired DNA damage, but above a threshold level of the inhibitor cells entered mitosis with unrepaired DNA strand breaks (Fig. 4.6). LY2603618 was predicted to have no effect in the absence of DNA damage. The presence of an IC50 concentration of the inhibitor increased the number of SSB required to give apoptosis by over six-fold. The IC50 level of Chk1 inhibitors is defined in the model as the concentration that halves the increase in time to progression to M phase; it is not a constant, but depends upon the amount of DNA damage.

76

4

The DNA Damage Checkpoint

Table 4.5 Effects of the ATR inhibitor NU6027 SRF at 0 time 0 0 35 35 35 5 35

NU6027 0 3.0 0 0.1 0.3 1.0 3.0

Table 4.6 Replicative catastrophe resulting from ATR inhibition, calculated by check D

Table 4.7 Effect of BI2536 in the absence of DNA damage

SRF 1 10 30 40 50 60 70 80 90 100

BI2536 0 0.01 0.03 0.05 0.1 0.2

Time to M phase (h) 2.19 2.19 6.67 4.44 3.35 2.67 2.25

SRF at mitotic entry 0 0 0 3.2 5.7 8.3 10.4

% repair NU6027 = 0 99.4 99.3 98.5 97.9 97.3 96.6 96.0 95.4 94.8 94.3

NU6027 = 1.0 99.3 99.1 89.8 68.7 38.6 32.0 27.3 23.9 21.2 19.0

Time to M phase (h) 2.19 2.45 3.81 5.14 8.52 16.58

As an example of an ATR inhibitor, the compound NU6027 was modelled. The effects were generally similar to those of chk1 inhibition: In the absence of DNA damage, NU6027 had no effect on the modelled time to M phase. In the presence of DNA damage, NU6027 shortened the time to progression to M, and cells had unrepaired DNA damage at all concentrations of NU6027 tested. As with Chk1 inhibitors, ATR inhibition leads to increased initiation of DNA synthesis. This lowers levels of RPA, resulting in greatly decreased repair of SRF, and increased replicative catastrophe (Tables 4.5 and 4.6). Inhibition of PLK1 by BI2536 was modelled (Tables 4.7 and 4.8). PLK1 inhibition prolonged G2. If there is DNA damage, the cell has more time to repair it, so is less likely to move into M with damaged chromosomes.

4.7

Kinetics of the Checkpoint

Table 4.8 Effect of BI2536 in the presence of DNA damage

77 SRF 0 25 35 40 0 25 35 40 0 25 35

BI2536 0 0 0 0 0.1 0.1 0.1 0.1 0.2 0.2 0.2

Time to M phase (h) 2.19 6.19 6.67 Apoptosis at 6.66 8.52 8.24 Apoptosis at 6.68 Apoptosis at 6.66 16.58 15.65 Apoptosis at 6.68

Fig. 4.7 Effect of the cdk1/ cyclin B inhibitor RO3306 in presence and absence of DNA damage

Simulating inhibition of Cdk1/cyclin B by RO3306 resulted in very long (>100 hour) delay in time to enter mitosis. In the presence of DNA damage cells were less sensitive to RO3306 (Fig. 4.7). McCloy et al. (2014) reported that partial inhibition of Cdk1 in G2 phase by RO3306 caused aberrant mitosis. This effect was prevented by okadaic acid, an inhibitor of the protein phosphatase, PP2A. The pyrimidine drug gemcitabine (as its 5′-triphosphate) is incorporated into growing DNA chains, in place of 2′-deoxycytidine, where it causes stalled replication forks. Results of modelling gemcitabine on time to mitosis with the checkD program are shown in Fig. 4.8. If the checkD program was used in conjunction with the CYCLOPS model of the cell cycle, it was able to predict the effects of gemcitabine on growth of tumour cells in culture (Fig. 4.9). Modelling effects on a single cell (or a synchronous culture), the effect was triphasic: very low concentrations had no effect, intermediate concentrations caused growth inhibition by prolonging the time to mitosis, and higher concentrations gave virtually complete

78

4

The DNA Damage Checkpoint

Fig. 4.8 Effect of gemcitabine on the DNADC

Fig. 4.9 Effect of gemcitabine on growth of MiaPaca tumour cells in culture. The red line shows the calculated effect on a single cell (or a synchronous culture) and the green line shows the effect on an asynchronous culture

inhibition by triggering apoptosis. If an asynchronous culture was modelled, the discontinuities evened out, giving a conventional dose–response curve. Table 4.9 shows the calculated effect of gemcitabine on distribution among the phases of the cell cycle. Gemcitabine was predicted to cause accumulation of cells in G2 phase, and depletion of M phase. RAD51 binds to RPA and is involved in HR. RAD51 mutations had only a minor effect on DSB, according to the model. If most of DSB repair is by NHEJ, this would explain this result, as NHEJ is not RAD51-dependent. Subsequent simulations with the checkD model focussed upon interactions of inhibitors of the DNADC with drugs that cause different kinds of DNA damage.

4.7

Kinetics of the Checkpoint

Table 4.9 Effect of gemcitabine on cell cycle distribution, as calculated by CYCLOPS

79 Phase G1 S G2 M

Control 50.8 31.2 12.0 5.2

Treated 41.5 34.9 22.7 0.8

Gemcitabine was used as an example of a drug whose major effect on DNA replication is SRF. The effect of gemcitabine in the presence of the chk1 inhibitor, LY2603618 was triphasic. At low doses there was no effect. Above concentrations that activated the checkpoint, there was extensive growth inhibition, and cells entered mitosis with unrepaired DNA damage. At high concentrations, cells entered apoptosis from G2 phase. When checkD was incorporated into the CYCLOPS model of the cell cycle (Chap. 3) it could be used to model effects of DNADC inhibitors in combination with DNA-damaging drugs on growth and apoptosis of MiaPaca2 pancreatic tumour cells. In this combined model, the Cdk1 inhibitor LY2603618 (2 μM) sensitised gemcitabine by about a factor of two, with the gemcitabine IC50 decreasing from 33 nM to 14 nM. In a similar modelling study, the ATR inhibitor, AZD6738, also sensitised MiaPaca2 cells by a factor of 2–3-fold (Fig. 4.10). In this case, the triphasic dose– response curve for gemcitabine was converted to a simple on/off relationship. If the unique potency parameter for gemcitabine was replaced by a normal distribution, a more typical dose–response curve was obtained. The PARP inhibitor, olaparib (Fig. 4.11) inhibited repair of SSB in a time- and concentration-dependent manner (Fig. 4.12). According to checkD, olaparib causes a dose-dependent inhibition of repair of SSB (Table 4.10). It has no effect on repair of SRF. In the model, olaparib has no effect in the absence of SSB. Since olaparib does, in fact, inhibit cell growth in the absence of drugs or radiation, it must be concluded that there is a background level of SSB formation in absence of DNA-damaging treatment. It is concluded that, in this version of the model, olaparib inhibits DSB repair, but through NHEJ (known to be PAR-dependent) rather than by inhibition of HR. This is consistent with experimental and clinical observation that BRCA1-deficient tumours (which are unable to carry out HR) show synthetic lethality to PARP inhibitors. This is interpreted as indicating that such tumours are totally dependent upon NHEJ for repair of DSB. HR-deficient cells can also repair DSB by singlestrand annealing, though the relative importance of single-strand annealing and NHEJ is unclear. As an example of a drug that acts primarily by causing SSB, irinotecan (SPT11), a camptothecin analogue that inhibits topoisomerase I, was used. Irinotecan had a mean IC50 of 55.6 μM against a panel of colon carcinoma cell lines, though it was as low as 1 μM in SW620 cells (Sharma and Smith 2008; Zhang et al. 2017). The calculated dose–response curve predicted that irinotecan concentrations up to 3 μM did not trigger the DNA damage response. Between 3 and 8 μM the checkpoint was triggered, causing G2 delay, and the SSB were repaired before progression into

80

4

The DNA Damage Checkpoint

Fig. 4.10 Predicted dose– response curves for the gemcitabine + AZD6738 combination as calculated by CYCLOPS

Fig. 4.11 Olaparib, a PARP inhibitor

Fig. 4.12 Inhibition of SSB repair by olaparib. In presence of olaparib, cells progress into M phase with unrepaired SSB. Progression into M occurs earlier in presence of olaparib, in a concentrationdependent manner

mitosis. Above 8 μM there was G2 delay, but cells progressed into mitosis with some unrepaired SSB. At 60 μM and above, most cells went into apoptosis in G2 phase. As an example of a drug believed to kill cells by DSB formation, carboplatin was modelled (Fig. 6.6). Predicted effects on single cells have four possible outcomes. Below a threshold of ~1.5 μM the checkpoint is not triggered. At higher carboplatin

4.7

Kinetics of the Checkpoint

81

Table 4.10 Effect of olaparib on DSB repair DSB 10 10 30 30

Olaparib 0 1 0 1

G2 duration 3.356 3.286 4.777 4.518

DSB → 0 3.2 3.2 4.4 4.4

SSB at transition 0.475 4.556 0.707 12.006

concentrations the checkpoint is triggered. Up to 2.8 μM the cell is able to repair the strand breaks before progressing into mitosis. Above 2.8 μM the cell progresses into mitosis with unrepaired DNA damage. Above 3.2 μM cells enter apoptosis in G2. The CYCLOPS model of the cell cycle also describes the action of topoisomerase inhibitors, which delay progression from G2 into M phase in the absence of DNA damage (Jackson et al. 2017). To summarise the behaviour of the checkD model of the DDR: – If there is no DNA damage (SRF, SSB, or DSB) cells progress from G2 into M phase after DNA decatenation is complete with no delay. – If there is DNA damage but no checkpoint inhibitor is present, there is either G2 delay (during which the DNA damage is repaired) followed by normal progression or (if repair is not complete after several hours) apoptosis. – If a checkpoint inhibitor is present but there is no DNA damage, cells progress into M phase with no delay. – If there is DNA damage and a checkpoint inhibitor is present, depending on the inhibitor concentration, cells either progress into M phase with unrepaired DNA damage or undergo apoptosis. What is the subsequent fate of cells that progress into M with unrepaired DNA damage? This may be either death from mitotic catastrophe, or cell division of an aneuploid cell. The latter clearly happens some of the time, because we know that tumours have a tendency to become increasingly aneuploid. We also know that aneuploid cells have a replicative disadvantage. This is factored into the cell cycle model by enabling aneuploid cells to have a higher cell loss factor. What are the possible applications for a checkpoint model such as checkD? In pharmacodynamics, such a model can be used to optimise dose and treatment schedules. It may suggest selective combinations. Modelling can be used for in vitro– in vivo and mouse–human extrapolation. A checkpoint model can be used to search for synthetic lethalities: e.g. do BRCA mutations increase sensitivity to chk1 inhibition? Modelling can be used to suggest the most productive use of experimental resources; e.g. does inhibition of multiple repair and/or checkpoint pathways (e.g. chk1 and chk2) increase efficacy or just increase toxicity? In personalised medicine, modelling can explore how mutations or polymorphisms in repair pathways, in apoptosis, in checkpoint function, are expected to influence response to DNA-active drugs. Modelling can be used to generate prior parameter estimates when selecting personalised treatments using Bayesian modelling. When personalised treatment is impractical, clustering based upon frequent mutations in

82

4

The DNA Damage Checkpoint

(e.g.) pancreatic cancer might be used to assign patients to treatment groups. Modelling could help to suggest the most appropriate cluster assignments. Do repair inhibitors increase selectivity of DNA-damaging drugs, or just potency? If the DNA repair pathways and the DDR in the tumour were normal, then inhibiting repair would confer no selective advantage. As we have seen, most, if not all, carcinomas do have one or more mutations in these pathways, so it is likely that inhibitors of repair will, in general, enhance selectivity. However, which inhibitors will increase selectivity, and how much, will depend upon the particular mutations in individual tumours. Moreover, the enhancement will depend upon the kinds of DNA damage produced by particular DNA-damaging drugs. To predict whether (for example) a Chk1 inhibitor will enhance gemcitabine selectivity will require knowledge of the control of the checkpoint in normal cells. A model of the DDR, which can predict the effect of the mutation profiles of individual tumours, will help to individualise and optimise the use of DNA-damaging drugs.

4.8

Olaparib: An Inhibitor of the DDR

Olaparib (Fig. 4.11) was the first PARP inhibitor approved for clinical use. It is active as a single agent in BRCA1- or BRCA2-mutated ovarian, breast, and prostate tumours. It is also used in treatment of BRCA-deficient pancreatic cancers.

References Alkan O, Schoeberl B, Shah M et al (2018) Modeling chemotherapy-induced stress to identify rational combination therapies in the DNA damage response pathway. Sci Signal 11(540): eaat0229. https://doi.org/10.1126/scisignal.aat0229 Al-Shaheri FN, Al-Shami KM, Garnal EH et al (2020) Association of DNA repair gene polymorphisms with colorectal cancer risk and treatment outcomes. Exp Mol Pathol 113:104364. https://doi.org/10.1016/j.yexmp.2019.104364 Bae I, Rih JK, Kim HJ et al (2005) BRCA1 regulates gene expression for orderly mitotic progression. Cell Cycle 4:1641–1666 Choi M, Kipps T, Kurzrock R (2016) ATM mutations in cancer: therapeutic implications. Mol Cancer Ther 15:1781–1791 Das S, Salami SS, Spratt DE et al (2019) Bringing prostate cancer germline genetics into clinical practice. J Urol 202:223–230 Depamphilis ML, de Renty CM, Ullah Z, Lee CY (2012) “The octet”: eight protein kinases that control mammalian DNA replication. Front Physiol 3:368 Ferrari S (2006) Protein kinases controlling the onset of mitosis. Cell Mol Life Sci 63:781–795 Goral V (2015) Pancreatic cancer: pathogenesis and diagnosis. Asian Pac J Cancer Prev 16:5619– 5624 Jackson RC (2017) The Boolean kinetics of signal transduction. Integr Cancer Sci Therap 4:1–8 Jackson RC, Di Veroli GY, Koh S-B et al (2017) Modelling of the cancer cell cycle as a tool for rational drug development: a systems pharmacology approach to cyclotherapy. PLoS Comp Biol 13:e1005529. https://doi.org/10.1371/journal.pcbi.1005529

References

83

Joshi RR, Ali SI, Ashley AK (2019) DNA ligase IV prevents replication fork stalling and promotes cellular proliferation in triple negative breast cancer. J Nucl Acids 9170341:2019–2011. https:// doi.org/10.1155/2019/9170341 Kim OH, Lim JH, Woo KJ et al (2004) Influence of p53 and p21Waf1 expression on G2/M phase arrest of colorectal carcinoma HCT116 cells to proteosome inhibitors. Int J Oncol 24(4): 935–941 King C, Diaz HB, McNeely S et al (2015) LY2606368 causes replication catastrophe and antitumor effects through chk1-dependent mechanisms. Mol Cancer Ther 14:2004–2013 Löbrich M, Jeggo PA (2007) The impact of a negligent G2/M checkpoint on genomic instability and cancer induction. Nat Rev Cancer 7:861–869 McCloy RA, Rogers S, Caldon CE et al (2014) Partial inhibition of Cdk1 in G2 phase overrides the SAC and decouples mitotic events. Cell Cycle 13:1400–1412 O’Connor PM (1997) Mammalian G1 and G2 phase checkpoints. Cancer Surv 29:151–182 Shaltiel IA, Krenning L, Bruinsma W, Medema RH (2015) The same, only different – DNA damage checkpoints and their reversal throughout the cell cycle. J Cell Sci 128:607–620 Sharma RI, Smith TAD (2008) Colorectal tumor cells treated with 5-FU, oxaliplatin, irinotecan, and cetuximab exhibit changes in 18F-FDG incorporation corresponding to hexokinase activity and glucose transport. J Nucl Med 49:1386–1394 Terasawa M, Shinohara A, Shinohara M (2014) Canonical non-homologous end joining in mitosis induces genome instability and is suppressed by M-phase-specific phosphorylation of XRCC4. PLoS Genet 10:e1004563. https://doi.org/10.1371/journal.pgen.1004563 Toledo LI, Altmeyer M, Rask M-B et al (2013) ATR prohibits replication catastrophe by preventing global exhaustion of RPA. Cell 155:1088–1103 Toledo L, Neelsen KJ, Lukas J (2017) Replication catastrophe: when a checkpoint fails because of exhaustion. Mol Cell 66:735–749 Zhang J, Liu J, Wei W (2017 Apr 3) “FEM1”nism controls SLBP stability during cell cycle. Cell Cycle 16(7):597–598

Chapter 5

Dynamics of the Spindle Assembly Checkpoint

Abstract The spindle assembly checkpoint (SAC) normally acts to prevent cells that have not fully completed separation of their replicated chromosomes from undergoing cytokinesis and completing cell division. In normal cells, prolonged cell cycle arrest by the SAC results in apoptosis. Most cancer cells have mutations or expression changes in components of the SAC that allow cells with abnormal chromosomes to escape the checkpoint, and because of over-ride of the SAC, most tumours become progressively more aneuploid. An inactive or defective SAC results in a high rate of cell death. For this reason, it does not, on its own, cause malignancy. However, if it is accompanied by other changes, such as a defective G1:S checkpoint, that confer a compensating growth advantage, the resulting aneuploid cells will be malignant. Some cancer cells with a normal SAC have mutations in the DNA damage response (DDR). Cross-talk between the DDR and the SAC means that mutations or expression changes in the DDR can over-ride a normal SAC and drive aneuploidy. Malignant transformation requires deletion or malfunction of both the G1:S checkpoint and either the DDR or the spindle assembly checkpoint. An important class of anticancer drugs, the spindle poisons, block cells in mitosis, resulting in apoptosis. Inhibitors of components of the SAC, such as aurora kinase A (AK-A) and aurora kinase B (AK-B), also have anticancer activity.

The effect of a gene on development is a function of its own structure as well as of its position in the chromosome. A change of the linear order of the genes in a chromosome may then leave the quantity of the gene unaffected, and yet the functioning of the genes may be changed. Theodosius Dobzhansky, Genetics and the Origin of Species (1937).

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_5. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_5

85

86

5.1

5

Dynamics of the Spindle Assembly Checkpoint

The Mitotic Spindle

The final event of G2 phase of the cell cycle is activation of CDK1/cyclin B, which activates the mitotic kinase MPS1, which in turn activates another kinase, MAD1, and thereby sets in motion assembly of the outer part of the kinetochore, the microtubule-binding region of the chromosomes (Hayward et al. 2019). The cell is now in M phase, mitosis. The microscopists who provided the earliest studies of mitosis described six phases of the process. Following the preceding phase of the cell cycle, G2, the replicated chromosomes, consisting of two identical chromatids, joined at a specialised region termed the centromere, are long and filamentous, and not visible to light microscopy. Note that the following description applies to animal cells: in plant and fungal cells the details of the process differ, though the end result is the same. During prophase the chromosomes condense into visible, compact structures. The condensation process is the result of the DNA being wound onto a series of protein structures. RNA transcription stops, because the RNA polymerases are no longer able to access the condensed DNA. The nucleolus dissolves. The centrosomes, which are organelles close to the nucleus that replicated in the previous S phase, begin the process of microtubule formation, which involves the polymerisation of the protein, tubulin. At this stage the microtubules grow out, in apparently random directions, from the centrioles, the core regions of the centrosomes (Fig. 5.1). In prometaphase, the nuclear membrane dissolves. The growing microtubules become attached to specialised structures on each centromere, the kinetochores. The result of this process is a structure termed the mitotic spindle, in which each member of a chromatid pair is attached to the centrosome at opposite ends (“poles”) of the cell, as shown in Fig. 5.1. The number of chromosome pairs varies between species. In humans there are 23 pairs. In metaphase, the centrosomes pull the chromosomes into the equatorial plane of the cell, i.e. the middle of the cell, equidistant from the poles. The force for this process comes from shortening of the Fig. 5.1 The mitotic spindle in metaphase

5.2

Mitotic index as a Measure of Cell Proliferation

87

microtubules by tubulin depolymerisation. Before the cell can proceed further, to anaphase, a checkpoint mechanism, the spindle assembly checkpoint (SAC) determines that each member of a chromatid pair is correctly attached to the opposite spindle poles (Sear and Howard 2006; Musacchio and Salmon 2007). This is a tension-sensing mechanism: correct attachments are in a state of tension, incorrect attachments are not. Anaphase: When the checkpoint turns off (because all chromatid pairs are in a state of tension) a group of proteins, the anaphase-promoting complex (APC) is activated. This causes cleavage of the cohesin links that join the chromatid pairs. The microtubules shorten and pull the separated chromatids (now referred to as “daughter chromosomes”) to opposite ends of the cell, a process termed chromosome segregation. In telophase, new nuclear membranes form, enclosing the replicated sets of chromosomes, which now begin to de-condense. Finally, the process of cytokinesis rearranges the cell membrane and cytosol so that the binucleate cell resulting from mitosis becomes two daughter cells, each containing one nucleus and one centrosome. One cell has become two.

5.2

Mitotic index as a Measure of Cell Proliferation

Chromosomes are not visible to light microscopy for most of the cell cycle, but on entry to M phase they become condensed—the DNA is tightly coiled around histones and other chromosomal proteins. As described above, the condensed chromosomes are then assembled at the equator of the mitotic spindle. In metaphase, the resulting mitotic figures, appropriately stained, are readily visible under the microscope. The number of visible mitoses, expressed as a fraction or percentage of the total cell number, is the mitotic index. For a human cell line growing in culture with a doubling time of about 24 hours, the mitotic index (MI) is about 6%. For normal human tissues in situ, the MI is usually 90% (Peckham et al. 1983; Williams et al. 1987). The reason for the dramatic difference between the cure rate using a two-drug combination and a three-drug combination can be explained by the evolutionary dynamics analysis of Bozic et al.(2013). Assuming typical mutation rates for carcinoma cells in excess of 10-5 per cell division, and a tumour size at diagnosis in excess of 1010 cells, Eq. (6.5) indicates a significant likelihood of pre-existing double mutants before the start of treatment. However, the probability of triple mutants, resistant to three drugs is very low. Cisplatin, though a very effective drug, is not a magic bullet, but three-drug combinations containing it do indeed have qualitatively different properties from two-drug combinations. Because of their superficial location, primary testicular tumours are likely to be detected while they are still small. However, the closely related ovarian carcinomas, with an intra-abdominal location, are usually detected much later, and (while they are highly responsive to initial chemotherapy) are much more likely to relapse after a few years with drug-resistant disease. Cisplatin is toxic to the kidneys (nephrotoxic), and the reason it can be combined effectively with bleomycin, whose dose-limiting toxicity is to the lungs, and etoposide, which is myelosuppressive (toxic to bone marrow) is because the toxicities of the three drugs do not overlap. An analogue development programme, in search of platinum compounds that were less nephrotoxic resulted in carboplatin (Fig. 6.11). However, the dose-limiting toxicity of carboplatin is myelosuppression, which complicates the design of combination regimens with other myelosuppressive drugs. Carboplatin and cisplatin are guanine N-7 cross-linkers, so (unlike the nitrososureas) are active against mer + tumours. Fig. 6.10 Cisplatin, a DNA cross-linking agent

128

6

Dynamics of Drug Resistance

Fig. 6.11 Carboplatin, a non-nephrotoxic cisplatin analogue

References Arnal A, Ujvari B, Crespi B et al (2015) Evolutionary perspective of cancer: myth, metaphors, and reality. Evol Appl 8:541–544 Bozic I, Reiter JG, Allen B et al (2013) Evolutionary dynamics of cancer in response to targeted combination therapy. Life 2:300747. https://doi.org/10.7554/eLife.00747 Brady-Nicholls R, Nagy JD, Gerke TA (2020) Prostate-specific antigen dynamics predict individual responses to intermittent androgen deprivation. Nat Commun 11:1750. https://doi.org/10.1038/ s41467-020-15424-4 Brady-Nicholls R, Zhang J, Zhang T et al (2021) Predicting patient-specific response to adaptive therapy in metastatic castration-resistant prostate cancer using prostate-specific antigen dynamics. Neoplasia 23:851–858 Clevers H, Loh KM, Nusse R (2014) Stem cell signaling. An integral program for tissue renewal and regeneration: Wnt signaling and stem cell control. Science 346:1248012. https://doi.org/10. 1126/science.1248012 Coldman AJ, Goldie JH (1986) A stochastic model for the origin and treatment of tumors containing drug-resistant cells. Bull Math Biol 48:279–292 Cunningham JJ, Gatenby RA, Brown JS (2011) Evolutionary dynamics in cancer therapy. Mol Pharm 8:2094–2100 Cunningham JJ, Brown JS, Gatenby RA, Stankova K (2018) Optimal control to develop therapeutic strategies for metastatic castrate resistant prostate cancer. J Theor Biol 459:67–78 De Kruif P (1926, reprinted 1954) Microbe hunters. Harcourt Brace Jovanovich, New York, p 323 Enriquez-Navas PM, Wojtkowiak JW, Gatenby RA (2015) Application of evolutionary principles to cancer therapy. Cancer Res 75:4675–4680 Enriquez-Navas PM, Kam Y, Das T et al (2016) Exploiting evolutionary principles to prolong tumor control in preclinical models of breast cancer. Sci Transl Med 8:327ra24. https://doi.org/ 10.1126/scitranslmed.aad7842 Gallaher JA, Enriquez-Navas PM, Luddy KA et al (2018) Spatial heterogeneity and evolutionary dynamics modulate time to recurrence in continuous and adaptive cancer therapies. Cancer Res 78:2127–2139 Gatenby RA, Brown JS (2020) The evolution and ecology of resistance in cancer therapy. Cold Spring Harb Perspect Med 10:a040972. https://doi.org/10.1101/cshperspect.a40972 Gatenby RA, Silva AS, Gillies RJ, Frieden BR (2009) Adaptive therapy. Cancer Res 69:4894–4903 Goldie JH, Coldman AH (1979) A mathematic model for relating the drug sensitivity of tumors to their spontaneous mutation rate. Cancer Treat Rep 63:1727–1733 Goldie JH, Coldman AH (1983) Quantitative model for multiple levels of drug resistance in clinical tumours. Cancer Treat Rep 67:923–931 Goldie JH, Coldman AJ, Gudauskas GA (1982) Rationale for the use of alternating non-crossresistant chemotherapy. Cancer Treat Rep 66:439–449 Jackson RC (1992) The theoretical foundations of cancer chemotherapy introduced by computer models. Academic, New York, pp 239–242 Jackson RC (1993) Amphibolic drug combinations: the design of selective antimetabolite protocols based upon the kinetic properties of multienzyme systems. Cancer Res 53:3998–4003 Jackson RC (1996) Computer techniques in preclinical and clinical drug development. CRC Press, Boca Raton, FL

References

129

Kam Y, Das T, Minton S, Gatenby RA (2014) Evolutionary strategy for systemic therapy of metastatic breast cancer: balancing response with suppression of resistance. Womens Health (Lond) 10:423–430 Lloyd HH (1977) Growth kinetics and biochemical regulation of normal and malignant cells. Williams & Wilkins, pp 445–469 Lloyd MC, Cunningham JJ, Bui MM et al (2016) Darwinian dynamics of intratumoral heterogeneity: not solely random mutations but also variable environmental selection forces. Cancer Res 76:3136–3144 Luria SE, Delbrück M (1943) Mutations of bacteria from virus sensitivity to virus resistance. Genetics 28:491–511 Meade MB, Gatenby RA, Dalton WS (2009) Environment-mediated drug resistance: a major contributor to minimal residual disease. Nat Rev Cancer 9:665–674 Peckham MJ, Barrett A, Liew KH et al (1983) The treatment of metastatic germ-cell testicular cancers with bleomycin, etoposide, and cisplatin (BEP). Br J Cancer 47:613–619 Rideout D (1986) Self-assembling cytotoxins. Science 233:561–563 Silva AS, Kam Y, Minton SE et al (2012) Evolutionary approaches to prolong progression-free survival in breast cancer. Cancer Res 72:6362–6370 Skipper HE (1983) The forty-year-old mutation theory of Luria and Delbrück and its pertinence to cancer chemotherapy. Adv Cancer Res 40:331–363 Skipper HE (1986) On mathematical modelling of critical variables in cancer treatment (goals: better understanding of the past and better planning in the future). Bull Math Biol 48:253–278 Skipper HE, Schabel FM, Wilson WS (1964) Experimental evaluation of potential anticancer agents XIII. On the criteria and kinetics associated with “curability” of experimental leukemia. Cancer Chemother Rep 35:1–111 Strobl MAR, West J, Viossat V et al (2020) Turnover modulates the need for a cost of resistance in adaptive therapy. Cancer Res 81:1135–1147 Wangari-Talbot J, Hopper-Borge E (2013) Drug resistance mechanisms in non-small cell lung cancer. J Can Res Updates 2:265–282 West JB, Dinh MN, Brown JS et al (2019) Multidrug cancer therapy in metastatic castrate-resistant prostate cancer: an evolution-based strategy. Clin Cancer Res 25:4413–4421 West J, You L, Zhang J et al (2020) Towards multidrug adaptive therapy. Cancer Res 80:1578– 1589 Williams SD, Birch R, Einhorn LH et al (1987) Treatment of disseminated germ-cell tumors with cisplatin, bleomycin and either vinblastine or etoposide. New Engl J Med 316:1435–1440 Zhang J, Cunningham JJ, Brown JS, Gatenby RA (2017a) Integrating evolutionary dynamics into treatment of metastatic castrate-resistant prostate cancer. Nat Commun 8:1816. https://doi.org/ 10.1038/s41467-017-01968-5 Zhang L-L, Kan M, Zhang M-M et al (2017b) Multi-region sequencing reveals the intratumor heterogeneity of driver mutations in TP53-driven non-small cell lung cancer. Int J Cancer 140: 103–108

Chapter 7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

Abstract Chronic myeloid leukaemia (CML) is caused by the Bcr-Abl translocation. It is otherwise genetically normal. CML is unusual in being a “single-hit” malignancy, i.e. although other genetic abnormalities appear as the disease progresses, the translocation is sufficient to result in neoplastic transformation. The t(9; 22) chromosomal translocation results in formation of the Bcr-Abl fusion oncogene in which constitutive activation of Abl by Bcr results in turning on an innate immune response. Untreated CML is genetically unstable and progresses in about three years to an acute leukaemia. As with normal neutrophils, this causes activation of the antiapoptotic protein Mcl-1, and production of high levels of reactive oxygen species (ROS). ROS are damaging to normal tissues, and in a normal innate immune response are turned off after a few days (the adaptive immune response takes over). In CML, the switch-off does not occur, and high levels of ROS are responsible for much of the morbidity of CML. Among the damaging effects of ROS is oxidation of DNA guanine bases to 8-oxo-guanine. This is mutagenic because 8-oxo-G miscodes, resulting in genetic instability. Inhibitors of Abl kinase activity, such as imatinib, block STAT signalling, turning off Mcl-1 and ROS. Thus, imatinib and its analogues block both cell proliferation and disease progression. As CML progresses, the expression level of the transcription factor, Nrf2, is increased. This is a frequent response to oxidative stress, and is also seen in solid tumours. Nrf2 upregulation is believed to promote cancer cell survival by inducing proteins that protect from oxidative damage.

Slow electron flux translates into lower oxygen consumption, limited proton pumping, falling membrane potential ... and collapsing ATP synthesis. The highly reduced state of complex I increases its reactivity with oxygen, forming free radicals such as superoxide. Nick Lane, The Vital Question (2015).

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_7. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_7

131

132

7.1

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

CML as a Disease of Redox Imbalance

A small abnormal chromosome, the Philadelphia chromosome (Ph), first described by Nowell and Hungerford (1960), is diagnostic for CML. In CML the number of white cells both in bone marrow and blood is greatly increased, and these are predominantly differentiated polymorphonuclear cells (Carella et al. 2001; Cortes and Deininger 2007). It was the first genetic defect to be linked with a specific human cancer. Before the nature of the Ph chromosome had been established, Harrap and Speed (Harrap and Speed 1964; Bergel and Harrap 1965) showed that the glutathione of CML neutrophils was largely in the oxidised state. In normal neutrophils, and indeed in all normal tissues, glutathione exists at millimolar concentrations, and less than 1% is in the oxidised form. Glutathione is the main redox buffer in normal cells, and its oxidation in CML results in the thiol groups of essential enzymes becoming oxidised. This was the first report of the profound oxidative imbalance in CML. CML is a progressive condition: after a time it enters an accelerated phase, then an acute phase where immature myeloid cells (myeloblasts) appear in the circulation. This disease progression is attributed to genomic instability (Jorgensen and Holyoake 2007). In cells from CML patients who had progressed to the acute phase, the glutathione had returned to the reduced form. Follow-up studies showed that the enzymes responsible for glutathione oxidation and reduction (glutathione peroxidase and glutathione reductase, respectively) had normal activity in CML, as had the cofactor for glutathione reductase, NADPH, and the enzymes responsible for maintaining NADPH in its reduced form (Harrap and Jackson 1969). It was concluded—by default—that since the reducing capacity of CML cells was unaltered, that the supply of oxidising capacity must be increased. Many years later, Nieborowska-Skorska et al.(2012) showed massively increased production of reactive oxygen species (ROS) in CML. ROS—a term that embraces peroxides, superoxides, and hydroxyl radicals—are chemically reactive, and cause oxidative damage to proteins, nucleic acids, and lipids, and are likely to be the primary factor in CML morbidity. It was concluded that the high levels of ROS in CML neutrophils were the reason for the abnormal oxidation state of glutathione.

7.2

The Philadelphia Chromosome and the Bcr-Abl Translocation

Rowley (1973) reported that the Ph chromosome resulted from a reciprocal translocation between chromosomes 9 and 22. Unlike most malignancies, CML is a “one hit” malignancy (Michor et al. 2006). The process by which the t(9,22) translocation was shown to produce a Bcr-Abl fusion protein was reviewed by Minden (1987). The Abl gene is a driver of the innate immune response. Following an infection, it is activated by TNFα, produced by macrophages that were in turn activated by bacterial or fungal cell wall components. Abl codes for a tyrosine kinase. In CML,

7.3

CML as a Constitutively Activated Innate Immune Response

133

the Abl domain of the Bcr-Abl fusion gene is constitutively activated by the Bcr domain. Normal Bcr codes for a serine/threonine kinase and also for a rho/rac family GTPase.

7.3

CML as a Constitutively Activated Innate Immune Response

CML can thus be understood as a permanently activated innate immune response. This is the primitive form of cellular immunity that is believed to have predated the development of B cells and T cells in evolution, and which remains the first line of defence in mammals (Fig. 7.1). Bone marrow stem cells (HSC) and myeloid progenitor cells reside in bone marrow, blood neutrophils, and monocytes in blood and in tissues. The presence of bacterial cell wall components in infected tissues activates macrophages, which respond by releasing inflammatory cytokines. One of these, interleukin 8 (IL-8), attracts blood neutrophils to the site of infection. Another cytokine, tumour necrosis factor alpha (TNFα) activates tissue neutrophils. This neutrophil activation process inhibits their spontaneous apoptosis by up-regulating the anti-apoptotic proteins, Mcl-1 and Bcl-XL. The primary bactericidal activity of activated neutrophils is formation of ROS, which in presence of the enzyme myeloperoxidase, convert chloride ions to hypochlorite (Fig. 7.1) (Jackson and Radivoyevitch 2013). The gatekeeper molecule for neutrophil activation is the Abelson proto-oncogene (Abl), a tyrosine kinase that phosphorylates and activates the transcription factors STAT3 and STAT5. STAT3 activates mitochondrial electron transport with increased electron back-pressure that causes leakage of ROS from the mitochondria. STAT5 drives transcription of multiple proteins, including Mcl-1 and cyclin D. The latter, as discussed in Chap. 3, stimulates cell proliferation, including that of myeloid progenitor cells. Pluripotent bone marrow stem cells, under the influence of IL-3, differentiate into myeloid progenitor cells. Self-renewal of myeloid progenitors requires the growth Fig. 7.1 The innate immune response to bacterial infection

134

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

Fig. 7.2 Neutrophil trafficking between bone marrow, blood, and tissues (from Jackson and Radivoyevitch 2013) +

M-CSFR

M-CSF

+

Tet-2

Bound M-CSF

+

Myeloid progenitors

monocytes

Actovated macrophages

Fig. 7.3 Positive feedback in the innate immune response

factor, GM-CSF, which is produced by macrophages. Another growth factor, G-CSF, stimulates the differentiation of myeloid progenitor cells into neutrophils, which are released into the blood (Fig. 7.2). Simultaneously, the growth factor M-CSF causes differentiation of myeloid progenitor cells into monocytes. The numbers of circulating neutrophils is maintained roughly constant by a feedback mechanism: GM-CSF receptors on blood neutrophils bind circulating GM-CSF and thus limit its access to bone marrow. The positive feedback loop that drives switching of the innate immune system from the baseline (off) state to the activated (on) state is summarised in Fig. 7.3. The cytokines CCL-2 and M-CSF, produced by

7.3

CML as a Constitutively Activated Innate Immune Response

135

Fig. 7.4 Response of the innate immune system to a bacterial infection

activated macrophages, stimulate the production of more macrophages (Fixe and Praloran 1997). Expression of M-CSF receptor, controlled by TET-2, acts as a negative regulator of monocyte production, by acting as a sink for free M-CSF, a feedback loop that regulates monocyte cell numbers. The anti-inflammatory cytokine, IL-10, mediates another feedback loop. IL-10, primarily produced by monocytes, inhibits production of TNFα and other inflammatory cytokines by activated macrophages. It has been described as the “master regulator” of immunity to infection (Couper et al. 2008). IL-10 production is PD-1 dependent (Haymaker et al. 2017). PD-1 is a cell surface receptor that prevents activation of T cells when it is bound to its ligand: it drives apoptosis of antigen-specific T cells in lymph nodes and reduces apoptosis in suppressor T cells. Its ligand, PD-L1, is upregulated on macrophages and dendritic cells in response to lipopolysaccharide (LPS) and GM-CSF. This protects macrophages from T cell-mediated cell killing during infections, and thus increases their counts (Tang et al. 2011). PD-1 mediates innate immune system communication with T cells and is targeted by new anticancer antibodies including nivolumab and pembrolizumab. Formation of ROS is thus an essential part of the elimination of bacterial or fungal infection by neutrophils, but once the infection is eliminated it is essential that the activated neutrophils are removed, through apoptosis, because prolonged exposure to ROS causes extensive normal tissue damage. Because of the complex, nonlinear dynamics of the innate immune response, and its multiple feedback loops, understanding of the system has been facilitated by computer models. There are more than thirty published models of innate immunity, one of which (myeloN6.R) is included in the online supplement. Figure 7.4 summarises the response of the system to a bacterial challenge of 103 cells/g of tissue.

136

7 Chronic Myeloid Leukaemia: A One-Hit Malignancy

Note that this figure only shows six of the forty system variables in the model. After infection the bacterial count continues to rise for about 40 hours. During that time there is an increase in activated tissue macrophages and neutrophils. The bacterial count then declines, and the infection is largely eliminated by 100 hours, and the innate immune system starts to turn off. Understanding of the dynamics of the innate immune response must reflect the fact that this system is potentially toxic to the host. Prolonged exposure to the ultimate mediators of antibacterial activity, reactive oxygen species (ROS) and hypochlorite, causes severe damage to cells of the host tissues. For this reason, prompt switching off of the innate immune response is as important as the on switch. The mammalian immune system, which involves both a high-capacity/low affinity system (innate immunity) and a low capacity/high affinity system (adaptive immunity), is an adaptation that has evolved to optimise the balance of efficacy and toxicity. When the innate immune system is triggered, a feed-forward signal activates B-cells, which, after a few days, start to produce antibodies, so that the adaptive immune system takes over. If the innate immune system does not switch off promptly, the result is chronic inflammatory disease. The off switch involves three main factors. First, when bacteria are removed by the system, macrophage activation ceases after which neutrophil invasion and activation, and production of ROS, decline. Second, even in the presence of a continuing signal of infection, the activation of macrophages is self-limiting: as the number of circulating monocytes increases, the expression of membrane growth factor receptors on these cells depletes the concentration of circulating growth factors, so that proliferation of monocyte precursors, and their differentiation to monocytes and macrophages, decreases. Third, as noted above, anti-inflammatory cytokines, principally IL-10, eventually block neutrophil activation. Inflammatory diseases are a heterogeneous group of conditions. Autoimmune diseases, in which normal body components are erroneously interpreted as foreign, are mediated by the adaptive immune system, and are outside the scope of our model. In a few cases, a pathogen may evoke an inflammatory response without being totally eliminated by it, so that the innate immune system remains chronically activated. Lyme disease is an example of this. It has been suggested that other inflammatory diseases, such as rheumatoid arthritis, may be triggered by pathogen infection, though definitive evidence is lacking. It seems likely that some inflammatory diseases may be caused by failure of the innate immune system to switch off appropriately. Our model suggests that one way this might happen is by failure of the feedback control of monocyte differentiation and migration, so that the system is operating as an open loop. Why should this be? Recent evidence has shown that monocyte and macrophage function are under epigenetic control (Carr and Patnaik 2020). This is a negative control, since hypermethylation increases the monocyte: neutrophil ratio (Hoeksema and de Winther 2016). The specific genes involved in this process are not yet identified, though there is evidence that enhancer regions of DNA control the expression of inflammatory genes in macrophages. Production of MIF (macrophage migration inhibitory factor) is increased by TET-2 downregulation (Pronier et al. 2022). It is likely that some inflammatory diseases

7.5

Reactive Oxygen Species, Ageing, and Cancer

137

are the result of epigenetic reprogramming of monocytes/macrophages, and this will provide new targets for design of anti-inflammatory drugs, as has been suggested for another macrophage-driven condition, atherosclerosis. TET-2 has been shown to be required for anti-inflammatory responses in experimental models of such conditions as endotoxin shock and colitis. The inability to turn off an innate immune response, probably because of epigenetic changes, results in inflammatory disease. If the innate immune system is permanently stimulated as a result of a genetic mutation, or a chromosomal rearrangement, the result is leukaemia. As discussed above, CML is the result of a chromosomal translocation causing constitutive activation of the Abl gene. This can be modelled using a version of the myelo computer model (Jackson and Radivoyevitch 2014).

7.4

Mcl-1 and Myeloid Cell Immortalisation

As shown in Fig. 7.5, Abl activates the transcription factors STAT3 and STAT5, STAT3 is active in mitochondria, where it stimulates oxidative phosphorylation, resulting in release of ROS into cytosol and nucleus. STAT5 is a nuclear transcription factor, and activates growth factors, including IL3 and GM-CSF, and thus stimulates proliferation of myeloid progenitor cells. c-Myc is a transcription factor that drives progression of cells from G1 phase into S. The other product of STAT5 is the anti-apoptotic protein Mcl-1. Unactivated tissue neutrophils have a short half-life (4–5 days), but expression of Mcl-1 effectively immortalises them. The Mcl-1 transcript has a very short half-life (about 20 minutes) so when the innate immune response turns off, tissue neutrophil turnover resumes rapidly.

7.5

Reactive Oxygen Species, Ageing, and Cancer

Production and turnover of ROS in myeloid cells are summarised in Fig. 7.6. The transcription factor STAT3 drives oxidative phosphorylation in mitochondria. Excess ROS, resulting from imbalance in the mitochondrial electron transport Fig. 7.5 Bcr-Abl signalling in myeloid cells, as modelled by the myeloL.R program. All components shown are modelled explicitly as dynamic system variables

138

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

Fig. 7.6 Formation and utilisation of ROS in myeloid cells

chain escape from mitochondria and enter the cytosol and nucleus. ROS cause oxidative damage, including lipid peroxide formation in membranes and 8-oxoguanine in DNA. Hydrogen peroxide (H2O2), a predominant form of ROS, is removed in peroxisomes by catalase, which oxidises one H2O2 to O2 as it reduces a second H2O2 to water, in the cytosol by peroxidases that use glutathione (GSH) as the electron source, in the mitochondria by peroxiredoxin 3 which uses thioredoxin as the electron source, and in neutrophil lysosomes by myeloperoxidase which converts H2O2 and chloride anion into hypochlorite (HOCl) to kill bacteria. ROS have been implicated as causative agents in ageing: an early theory of ageing was that it was the result of cumulative irreparable damage to cell membranes, caused, at least partly, by ROS-induced membrane damage. A more recent theory of ageing is that it is the result of progressive epigenetic gene silencing. The enzyme TET-2 (discussed in Chap. 8) removes methyl groups from DNA cytosine bases, and thus reverses one of the major forms of gene silencing. TET-2 requires the antioxidant ascorbic acid (vitamin C) as a substrate. This provides a possible mechanism for a putative anti-ageing effect of ascorbic acid, though evidence for this remains equivocal. Ascorbic acid, and other antioxidants, have been claimed to have anticancer activity. Again, a possible mechanism for this may be through the role of ascorbic acid in TET-2 activity. Other antioxidants could presumably act by removing ROS, so having a sparing effect on ascorbic acid. The plasma ascorbic acid concentration in well-nourished humans is about 8 mM. Supplementing the diet with high doses of ascorbic acid can raise the plasma concentration to about 30 mM, and this level is well tolerated. A modelling study of CML predicted that 30 mM ascorbic acid would have no effect on the neutrophil cell count but decreased ROS by about 75% (Jackson and Radivoyevitch 2016).

7.6

8-Oxoguanine as a Mutagen

One malignant disease where ROS are undoubtedly implicated is CML, where high levels of ROS cause genomic instability. This is caused by oxidation of DNA guanine bases to 8-oxo-guanine, resulting in point mutations because 8-oxo-guanine base-pairs with T instead of C (Cheng et al. 1992). Treatment of CML with inhibitors

7.8

Modelling the Evolutionary Dynamics of CML

139

of Abl kinase not only relieves the acute symptoms of CML, but also prevents progression of the disease to the acute phase.

7.7

Imatinib in the Treatment of CML

Imatinib (Glivec®, Fig. 7.7) was introduced into clinical practice for the treatment of CML in 2001, and also has activity in Ph+ CLL and certain solid malignancies, including gastric interstitial tumour (GIST) (Druker et al. 1996). Although the rate of acquired resistance of CML is low, it does occur, and has been shown to be caused by mutations in the Abl binding site, such as the T315I mutation. Second-generation Abl kinase inhibitors, such as nilotinib, dasatinib, and ponatinib often retain activity against resistant CML (Kantarjian et al. 2006; Quintas-Cardama et al. 2007; Cortes et al. 2012}. Compared with other malignant diseases treated with tyrosine kinase inhibitors (TKI), the resistance rate is remarkably low. Kalmanti et al. (2015) reported that progression-free survival ten years from the start of imatinib treatment was 82%, an annual failure rate of 2%. By contrast, when gastrointestinal stromal tumours were treated with imatinib, the 2-year progression-free survival was 59% (O’Hare et al. 2007), in line with the level of activity observed with other antitumour kinase inhibitors. There is no evidence that the baseline mutation rate is lower in CML than in other malignancies: untreated CML progresses to end-stage disease in 3 to 4 years, and when CML is treated with cytotoxic agents, acquired drug resistance occurs just as rapidly as with other leukaemias. Despite the effectiveness of imatinib in CML, and the durability of the responses, it does not appear to be curative, since sensitive PCR tests on the bone marrow of patients in remission show the persistence of Bcr-Abl-positive cells.

7.8

Modelling the Evolutionary Dynamics of CML

If the model of the innate immune system described above is modified to replace c-Abl with constitutively activated Bcr-Abl, the revised model can be used to study the dynamics of CML.

Fig. 7.7 Imatinib, an inhibitor of Abl tyrosine kinase

140

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

A version of this model, myeloL, is included in the online supplement. Running this model with all parameters set to their default values predicted a circulating neutrophil count that was elevated over ten-fold relative to baseline. The plasma survival time of neutrophils was only slightly greater than normal—5.4 days, compared with a normal value of 4.7 days—but the survival time of tissue neutrophils was significantly longer, 13.2 days. Mcl-1 increased ten-fold. ROS at steady state was 45-fold increased, and STAT3 was >100-fold higher. In the bone marrow, oxidised glutathione (GSSG) was 30% of total glutathione in CML, and undetectable in normal bone marrow. Figure 7.8 shows the calculated effect of assuming different levels of Bcr-Abl. The steady-state circulating neutrophil count was a linear function of Bcr-Abl expression, whereas ROS production followed a Hill equation saturation curve with a half-maximal effect at 4 nM (Fig. 7.8). There are no published data on absolute expression levels of Bcr-Abl in clinical CML samples. Subsequent calculations assumed a level of 13 nM, corresponding to that found in Bcr-Abltransformed cell culture lines (Jackson and Radivoyevitch 2014). The model was used to predict the effects of imatinib and its analogues, initially assuming a steady-state imatinib concentration of 3.4 μM. This is the steady-state trough level achieved in humans after sustained oral treatment with the maximal clinical dose of 800 mg twice daily. The effects were that expression of Mcl-1 fell to normal levels almost immediately, ROS levels fell by 96% with a half-maximal response in 6.5 days, circulating white cell count dropped to close to a normal level in 30 days, with a half-maximal response by 8 days, and after 30 days the normal myeloid progenitor cell population was still depressed, but starting to recover. The simulations described above explored the pharmacodynamics (PD) of inhibition of Bcr-Abl signalling by assuming constant inhibitor concentrations. Modelling actual in vivo or clinical treatments requires drug concentrations to be calculated at each

Fig. 7.8 Steady-state circulating neutrophil count and ROS dependence on Bcr-Abl expression level. Reproduced from Jackson and Radivoyevitch (2014)

7.8

Modelling the Evolutionary Dynamics of CML

141

Fig. 7.9 PK/PD modelling of 4 daily doses of 800 mg seliciclib. Reproduced from Jackson and Radivoyevitch (2014)

time point by a pharmacokinetic (PK) model. Using a PK/PD model of daily oral doses of 400 mg imatinib predicted a marked (87%) decline in circulating CML neutrophils by 30 days and a 97% decrease in ROS (Jackson and Radivoyevitch 2016). The myeloL model (online appendix) was used to predict the effect of drugs other than TKI on CML. Seliciclib, also known as R-roscovitine, is an inhibitor of cyclindependent kinases 2 and 9, and a primary regulator of Mcl-1 (De La Motte and Gianella-Borradori 2004; Jackson et al. 2008; Wang et al. 2012). Modelling four daily oral doses of 800 mg seliciclib predicted that ROS decreased 82% by the end of treatment, but circulating neutrophils decreased only 11% (Fig. 7.9). PK/PD modelling of anti-leukaemic drugs requires that we describe the kinetics of both the leukaemic cells and the normal myeloid cell populations. The normal myeloid cell kinetics can be described using the previously published model which includes the same features as the CML model (Fig. 7.2) except that c-Abl is modelled instead of Bcr-Abl (Jackson and Radivoyevitch 2013). The normal and CML populations interact at two levels: first, the proliferating progenitor cells compete for resources in the bone marrow, which is able to support a limited population of progenitor cells. Second, proliferation of both normal and leukaemic progenitor cells is activated by GM-CSF, and the free GM-CSF level is depleted by binding to receptors in circulating non-proliferating cells. This means that proliferation of both normal and leukaemic progenitors is subject to feedback inhibition by the total number of circulating cells, i.e. proliferation of normal progenitors is inhibited by circulating CML cells, and vice versa. The model predicts, in agreement with clinical observation, that as the number of circulating CML cells rises, the number of normal cells is depressed. In untreated chronic phase CML, with a total white blood cell count of 50,000/μL, only about 2% were predicted to be normal neutrophils.

142

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

Fig. 7.10 Effect of longterm treatment of CML with daily doses of imatinib (600 mg) on numbers of circulating CML cells and normal neutrophils. Reproduced from Jackson and Radivoyevitch (2014)

The time course of one-year CML treatment with imatinib (600 md/day) is shown in Fig. 7.10. A rapid exponential decline in Bcr-Abl positive cells, lasting about 3 months, is followed by a more gradual decline, in which leukaemic progenitor cells are slowly pushed out of the marrow by a recovering population of normal progenitors. This modelled sequence of events is in good agreement with the clinical data of Tang et al. (2011). The model predicted that after 12 months the GSSG content of circulating CML cells had declined from >30% to 8%, and was still decreasing. The in vitro version of the myeloL model was used to predict the effect of drug combinations. The effect of eight concentrations (including zero) of imatinib was examined in the presence of 8 concentrations of seliciclib. The simulated data were analysed by the method of Greco et al. (1990). The calculated interaction parameter, α, of Greco’s model was 2.8; positive values of α indicate synergy, zero, additivity, and negative values correspond to antagonism. The fact that Bcr-Abl TKI, such as imatinib, reverse the elevation of ROS in CML suggests that they should have an inhibitory effect on formation of 8-oxo-guanine in DNA, and thus delay progression of the disease to its acute “blast crisis” phase. Figure 7.11 is plotted from the 2015 release of the US NCI SEER database: the relative risk of mortality has plateaued at a value of about 2.5, with median age of onset of 60 for males and 62 for females (Radivoyevitch 2015). The SEER database does not provide information on treatment, but it may be assumed that most CML patients diagnosed since 2001 have been treated with imatinib or a similar Bcr-Abl TKI. Modelling studies using the myeloL program (online appendix) showed that to predict the clinical course of CML it was necessary to assume that the Bcr-Abl transformation resulted in increased proliferation of the transformed progenitor cells (because of GM-CSF upregulation) and in decreased apoptosis of circulating CML cells (because of upregulation of the anti-apoptotic protein Mcl-1). Making these assumptions the model predicted that, starting with a single cell carrying the Bcr-Abl translocation the condition progressed to clinical leukemia in ~7.7 years (Fig. 7.12).

7.8

Modelling the Evolutionary Dynamics of CML

143

Fig. 7.11 Trend of CML mortality relative risk. All 18 registries of the 1973–2012 SEER data were used. Reproduced from Jackson and Radivoyevitch (2016)

Fig. 7.12 Modelling CML progression from a single Bcr-Abl + cell at zero time. Reproduced from Jackson and Radivoyevitch (2016)

This is in agreement with the Japanese database of atom bomb survivors, which shows a peak incidence of CML diagnoses 8 years after radiation exposure. If either the proliferation rate of progenitor cells, or the apoptosis rate of circulating cells, was unchanged, leukaemia did not occur. The rate of progression from a single transformed cell to clinical leukaemia, predicted by the model, was parameterdependent. The magnitude of the growth advantage is irrelevant to the eventual outcome, but it affects time to onset of clinical disease, as predicted by previous models (Tang et al. 2011). Following a chronic phase of 3.9 years, blast cells appear in the blood. The model assumed that four types of mutation could occur: m1 mutations result in independence from exogenous growth factors; m2 mutations result in loss of differentiation. Since myeloid cell differentiation results in spontaneous apoptosis, loss of differentiation can result in a selective advantage. The observation that CML blast-phase cells, unlike chronic-phase cells, do not contain

144

7 Chronic Myeloid Leukaemia: A One-Hit Malignancy

GSSG is evidence that disease progression has resulted in loss of their differentiated phenotype. Mutations of type m3 result in downregulation of cell surface adhesion molecules (Jongen-Lavrencic 2005), so that bone marrow myeloblasts, which normally grow attached to bone marrow stroma, appear in the blood. Development of mutations for resistance to imatinib and other drugs for front-line treatment (m4) results in a growth advantage in the presence of drug. The myeloL model was used to ask the question: “how is CML progression affected if the mutation rate for independence from GM-CSF is set to zero? The latent period for progression from one Bcr-Abl+ cell to clinical disease was unaffected, but the system then was predicted to take an additional 50 years to reach acute phase. Though the mutations that cause accelerated growth (m1) and appearance of blasts in the circulation (m3) are independent, the accelerated phase appears to be a necessary step in the progression of CML. If the mutation rate m2 was set to zero, both latent period and progression time were unaffected. Mutations of type m2 describe changes that result in loss of ability of myeloid cells to differentiate. Loss of differentiation of circulating myeloid cells is invariably a feature of advanced CML, but according to the model, such mutations are not on the critical path of CML progression. The continued survival and proliferation of normal myeloid progenitor cells, and of CML chronic-phase progenitors, requires attachment to bone marrow stroma. In the absence of an attachment signal, an apoptotic pathway is triggered, and after a few hours the cells die. In advanced CML mutations that result in the survival signal becoming constitutive occur, so that progenitor cells survive in the absence of stromal attachment (Makishima et al. 2011). In the model, the total rate of these mutations is described as m3. When m3 was set to zero, the latent period was unaffected, but subsequent disease progression was very slow. The model thus predicted that time of progression of CML from chronic phase to blast phase was determined by m1 and m3. Busulfan (Myleran) is a DNA cross-linking agent formerly used to treat CML. It reduced numbers of circulating CML cells and gave symptomatic improvement but had only a modest effect on survival (Malpas 1986). The model simulated the effect of busulfan, 4 mg/day, starting at the time of diagnosis (Fig. 7.13). After an initial fall in circulating CML cells, the white blood cells count stabilised for about five years, after which the disease progressed to blast phase. This predicted effect of busulfan is rather better than was observed clinically. However, busulfan is a mutagen, and in a subsequent simulation busulfan was assumed to increase all mutation rates by a factor of ten. The predicted overall survival was now 4.3 years, compared with untreated survival of 3.9 years. The myeloL model thus suggested that the improvement in survival that was expected from the decreased numbers of CML progenitors in the bone marrow following busulfan treatment was largely cancelled out by the increased rate of progression to blast phase caused by the mutagenic effect of busulfan. Since it is believed that CML progression is driven by the mutagenic effect of ROS, it was of interest to model the effect of an antioxidant. The normal plasma concentration of ascorbate is about 8 mM. Supplementing the diet with high doses of

7.8

Modelling the Evolutionary Dynamics of CML

145

Fig. 7.13 Modelling CML progression assuming daily treatment with busulfan (4 mg). Reproduced from Jackson and Radivoyevitch (2016)

ascorbate can raise the plasma level to about 30 mM, and this level is well tolerated. The model predicted that ascorbate (30 mM) would have no effect on CML cell count, but decreased ROS by 75%. Maintaining ascorbate at 30 mM had no effect on the latent period, but was predicted to have a modest effect on disease progression, delaying inset of blast phase by six months. More potent antioxidants might be worth exploring. However, the relationship between ROS and CML disease progression, though well established in the early stages, is less clear in the later stages. When CML reaches blast phase, the redox potential is normalised. CML blast phase cells no longer contain GSSG, suggesting that ROS levels have been restored to normal. This may be because the antioxidant protein Nrf2 was shown to be upregulated in CML blast phase. NQO1, which is regulated by Nrf2 was also elevated in blast phase (Radich et al. 2006; Jackson and Radivoyevitch 2014). Evolutionary dynamics modelling predicted that a maximum tolerated daily dose of seliciclib would decrease Mcl-1 levels and induce apoptosis of CML cells, both progenitors and circulating cells. The ROS concentration was predicted to be markedly decreased. The onset of CML blast phase was delayed (in the model) by over five years. After one year, almost all the progenitors were drugresistant (Fig. 7.14). The effects of the TKI imatinib were studied by the myeloL model. If it was assumed that a maximum tolerated daily dose was commenced immediately after the event that triggered the translocation (such as radiation exposure) the predicted onset of clinical disease was delayed by about 40 years, suggesting that imatinib could be used prophylactically in this situation. Imatinib treatment normalised the circulating myeloid cell count and increased the number of normal myeloid progenitor cells in the bone marrow. That earlier study did not model the events occurring in CML progression. Continual imatinib treatment (3.4 μM), beginning at the time of diagnosis of CML chronic phase caused an immediate fall in circulating CML cells, after which the WBC count stabilised at about 110% of normal (Fig. 7.15). Blast cells did not appear

146

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

Fig. 7.14 Modelling CML progression assuming a steady-state seliciclib concentration of 7.0 μM. Reproduced from Jackson and Radivoyevitch (2016)

Fig. 7.15 Modelling CML chronic phase assuming a steady-state imatinib concentration of 3.4 μM. Reproduced from Jackson and Radivoyevitch (2016)

in the circulation for at least 20 years. If a high mutation rate to imatinib resistance was assumed, long-term control of CML was much less effective. Over 40 mutations in Bcr-Abl have been described that result in clinical resistance to imatinib (Talpaz et al. 2006). A Bayesian approach to modelling resistance has been adopted, in which prior assumptions about resistance rates, based upon preclinical data, were used to predict time to progression, and predicted outcomes were compared to observed outcomes to derive posterior estimates of mutation rates. The predicted time to progression of chronic phase CML was both dose-dependent and dependent upon the rate of mutation to imatinib resistance. These calculations suggested that

7.8

Modelling the Evolutionary Dynamics of CML

147

for CML chronic phase to progress to blast phase in the mean four-year progression time, the mutation rates must have been increased by about 130-fold over background. This increase is consistent with the elevated ROS levels in CML, and the resulting formation of 8-oxo-guanine in DNA. Why is TKI therapy of CML so much more effective than cytotoxic therapy? TKI potently block the ATP-binding site of both Bcr-Abl and c-Abl. Their selectivity arises from the fact that CML cells express active Bcr-Abl constitutively, whereas c-Abl kinase of normal neutrophils is only active under conditions of infection or inflammation. Inhibitors of Abl kinase achieve their anti-leukaemic activity at the cost of blocking the innate immune response. In CML cells, the inhibition of Bcr-Abl not only inhibits proliferation of CML progenitor cells but it also stimulates apoptosis by downregulating Mcl-1. The decrease in ROS caused by Bcr-Abl inhibition prevents morbidity caused by oxidative damage; importantly it also lowers the rates of mutations associated with progression of chronic phase CML to the accelerated phase of the disease. The one-hit nature of CML thus has profound consequences for targeted therapy. Bcr-Abl inhibitors both inhibit cell proliferation and block the progression process. An implication of this is that imatinib (unlike kinase inhibitors used to treat carcinomas) actually inhibits the development of acquired drug resistance (Jackson and Radivoyevitch 2016). Although TKI give long, durable remissions of CML, they are not curative. The Bcr-Abl transcript is used as a biomarker. In TKI-treated CML patients, Bcr-Abl shows a decline, initially a steep decline, followed by a slower phase, and often becomes undetectable, a state termed “molecular remission”. There is debate among haematologists whether it is safe to stop treatment of patients in molecular remission. The fact that Bcr-Abl reappears in most patients who have been taken off treatment is interpreted as indicating that TKI are not curative. When Bcr-Abl reappears in patients in remission, it often indicates the appearance of drug-resistant disease. Another factor in the reappearance of Bcr-Abl following cessation of treatment may be a decreased level of anti-CML immune response. Wodarz and Komarova (2014) present a mathematical model in which the intensity of the evoked immune response is proportional to the disease burden. The decline of CML neutrophils as the patient enters remission is followed by a decline in the anti-CML immune response. The rate of reappearance of Bcr-Abl transcripts following cessation of treatment appears to correlate negatively with the patients’ immune status (Mustjoki et al. 2020). The model of Wodarz and Komarova (2014) suggested that immune stimulation should lessen the incidence of treatment failure due to emergence of TKI-resistant cells, since CML-specific T cells should be equally active against TKI-sensitive and -resistant cells. Is the fact that CML is a one-hit malignancy incompatible with the conclusion of the previous chapters that there is no single biochemical change that can make a normal cell malignant? No, because the Bcr-Abl translocation has a pleiotropic effect, giving a growth stimulus, and inhibition of apoptosis, and genetic instability. Other apparent one-hit malignancies are known, but they are generally found in individuals with an inherited predisposition to cancer. For example, women with inherited BRCA-1 or BRCA-2 mutations have a high incidence of breast and ovarian

148

7 Chronic Myeloid Leukaemia: A One-Hit Malignancy

tumours. These are really two-hit tumours in which the first mutation is inherited rather than acquired. CML is commoner in males than in females at all ages above 25 years, despite there being no sex difference in ROS generation. This has been attributed to the higher numbers of haematopoietic stem cells at risk in males (Radivoyevitch et al. 2014). It has been proposed that radiation exposure may create both benign and malignant Bcr-Abl clones, and the benign clones are more likely to act as anti-CML vaccines in females than in males (Radivoyevitch 2020). The elevated activity of Nrf2 in CML is not peculiar to CML. It is frequently elevated in solid tumours and is believed to promote cancer cell survival by inducing proteins that protect from oxidative damage (Chartoumpekis et al. 2015; Leinonen et al. 2015). A n important target of Nrf2 is the stress-response protein haem oxygenase-1 (HO-1) which is also elevated in many tumour types (Na and Surh 2014). Nrf2 is targeted for degradation by Keap1, which thus acts as a tumour suppressor. Activation of AMK-activated protein kinase (AMPK) enhances Nrf2stimulated signalling of HO-1, demonstrating cross-talk between the central regulators of energy metabolism and redox metabolism (Zimmerman et al. 2015). Another link between Nrf2 expression and tumour progression was reported by Lignitto et al. (2019) who showed that Nrf2 promotes lung cancer metastasis by inhibiting the degradation of the pro-metastatic transcription factor, Bach1. Thirty percent of human lung cancers have mutations in either Keap1 or Nrf2, resulting in stabilisation of Nrf2, and activation of lung cancer metastasis. There are thus indications that oxidative stress contributes to the progression of a number of tumours. In the case of CML, however, it is the primary driver, resulting from point mutations caused by ROS-mediated oxidation of DNA guanine bases. Can we quantify this increase in the mutation rate? For untreated CML, assume that 3 mutations are required: (1) loss of ability to differentiate; (2) loss of dependence on external growth factors; (3) loss of attachment of myeloblasts to bone marrow stroma. Various scenarios can be modelled: (a) the mutations can occur in any order; (b) mutation 3 must occur last; (c) To model TKI-treated CML, it is necessary to add an additional mutation (4) which causes TKI resistance. These scenarios can be modelled using a version of the finite state machine described in the supplement to Chap. 2. The modified program, finiteBC.R is in the supplement to Chap. 7. It is possible to calibrate the 3-stage untreated model from old clinical data (pre-imatinib) on time to acute phase, and to calibrate the 4-stage model from data on time to progression (or 5-year progression-free survival) of imatinib-treated patients. If we assume that time to acute phase in the absence of treatment is 30 months, that there are 107 myeloid progenitor cells per patient, and that 3 mutations are necessary (after the t(9:22) translocation), this suggests a mean mutation rate of 5.3e-10 per cell division (34 h). This value seems low: perhaps because the number of cells at risk is actually less than 107, or because they are in a number of separate domains such that the transformed cells are not competing with the entire population. Bone marrow stem cells exist at many sites (at a rough guess, about 100), and progenitor cells that detach from bone marrow matrix will die, unless a mutation that disables attachment signalling has already occurred. However, we know that when the Bcr-Abl

References

149

translocation occurs, Bcr-Abl+ cells show up at multiple sites, so progenitor cells from one site must be able to translocate. The model will describe this situation by assuming that cells, after detaching, can survive for a short time (perhaps 8 hours) before entering apoptosis. Another way of estimating the relative mutation rates in CML and normal myeloid precursors is to regard acute phase as a form of AML. All forms of AML share the three genetic abnormalities discussed above—not necessarily the same specific mutations, but the same three phenotypes. Thus the incidence of AML gives us a guide to the background frequency of the three groups of mutations. This background rate is much lower than the almost 100% incidence in 34 months seen in untreated CML. Thus, the Bcr-Abl translocation results in a large increase in mutation rates. If the lifetime incidence of AML is 1.2% (.012), and three mutations are necessary, then the mean lifetime risk of each will be 3√.012 = 0.23, and the annual risk of each mutation = 0.23/80 = 0.003. For CML there is approximately 100% risk in 2.5 years, so the annual risk of each mutation is 3√(1 /2.5) = 0.4, about 130× the background incidence. This must reflect the genetic instability of CML stem cells.

References Bergel F, Harrap KR (1965) Future possibilities for the development of treatments for leukaemia. Cancer Res 25:1643–1848 Carella A, Daley G, Eaves C et al (2001) Chronic myeloid Leukaemia: biology and treatment. Martin Dunitz, London Carr RM, Patnaik MM (2020) Genetic and epigenetic factors interacting with clonal hematopoiesis resulting in chronic myelomonocytic leukemia. Curr Opin Hematol 27:2–10 Chartoumpekis DV, Wakabayashi N, Kensler TW (2015) Keap1/Nrf2 pathway in the frontiers of cancer and non-cancer cell metabolism. Biochem Soc Trans 43:639–644 Cheng KC, Cahill DS, Kasai H et al (1992) Hydroxyguanine, an abundant form of oxidative DNA damage, causes G→T and A→C substitutions. J Biol Chem 267:166–172 Cortes J, Deininger M (2007) Chronic myeloid leukemia. Informa, New York Cortes JE, Kantarjian H, Shah NP et al (2012) Ponatinib in refractory Philadelphia chromosomepositive leukemias. New Engl J Med 367:2075–2088 Couper KN, Blount DG, Riley EM (2008) IL-10: the master regulator of immunity to infection. J Immunol 180:5771–5777 De La Motte S, Gianella-Borradori A (2004) Pharmacokinetic model of R-roscovitine and its metabolite in healthy make subjects. Int J Pharmacol Ther 42:232–239 Druker BJ, Tamura S, Buchdunger E et al (1996) Effects of a selective inhibitor of the Abl tyrosine kinase on the growth of Bcr-Abl positive cells. Nat Med 2:561–566 Fixe P, Praloran V (1997) Macrophage colony-stimulating factor (M-CSF or CSF-1) and its receptor: structure-function relationships. Eur Cytokine Netw 8:125–136 Greco WR, Park HS, Rustum YM (1990) Application of a new approach to the quantitation of drug synergism to the combination of cis-diamminedichloroplatinum and 1-β-Darabinoduranosylcytosine. Cancer Res 50:5318–5327 Harrap KR, Jackson RC (1969) Some biochemical aspects of leukaemias: leucocyte gluathione metabolism in chronic granulocytic leukaemia. Eur J Cancer 5:61–67

150

7

Chronic Myeloid Leukaemia: A One-Hit Malignancy

Harrap KR, Speed DEM (1964) Some biochemical aspects of leukaemias: the appearance of a soluble disulphide in the blood in chronic granulocytic leukaemia. Br J Cancer 18:809–817 Haymaker CL, Kim D, Uemura M et al (2017) Metastatic melanoma patient has complete response with clonl expansion after whole brain radiation and PD-1 blockade. Cancer Immunol Res. https://doi.org/10.1158/2326-6066 Hoeksema MA, de Winther MPJ (2016) Epigenetic regulation of monocyte and macrophage function. Antioxid Redox Signal 25:758–774 Jackson RC, Radivoyevitch T (2013) Modelling c-Abl signaling in activated neutrophils: the antiinflammatory effect of seliciclib. Biodiscovery 2013(7):4–11 Jackson RC, Radivoyevitch T (2014) A pharmacodynamic model of Bcr-Abl signalling in chronic myeloid leukaemia. Cancer Chemother Pharmacol 74:765–776. https://doi.org/10.1007/ s00280-014-2556-z Jackson RC, Radivoyevitch T (2016) Evolutionary dynamics of chronic myeloid leukemia progression: the progression-inhibitory effect of imatinib. AAPS J 18:914–922. https://doi.org/10. 1208/s12248-016-9905-2 Jackson RC, Barnett AL, McClue SJ, Green SR (2008) Seliciclib, a cell cycle modulator that acts through inhibition of cyclin-dependent kinases. Expert Opin Drug Discov 3:131–143 Jongen-Lavrencic M (2005) Bcr/Abl-mediated downregulation of genes implicated in cell adhesion and motility leads to impaired migration towards CCR7 ligands CCL19 and CCL21 in primary Bcr/ABl-positive cells. Leukemia 19:373–380 Jorgensen HG, Holyoake TL (2007) Characterization of cancer stem cells in chronic myeloid leukaemia. Biochem Soc Trans 35:1347–1351 Kalmanti L, Saussele S, Lauseker M et al (2015) Safety and efficacy of imatinib in CML over a period of ten years: data from the randomised CML study IV. Leukemia 29:1123–1132. https:// doi.org/10.1038/leu.2015.36 Kantarjian H, Giles F, Wunderle L et al (2006) Nilotinib in imatinib-resistant CML and Philadelphia-chromosome-positive CLL. New Engl J Med 354:2542–2551 Leinonen HM, Kansanen E, Pölönen P et al (2015) Dysregulation of the Keap1-Nrf2 pathway in cancer. Biochem Soc Trans 43:645–649 Lignitto L, LeBoeuf SE, Homer H et al (2019) Nrf2 activation promotes lung cancer metastasis by inhibiting the degradation of Bach1. Cell 178:316–329 Makishima H, Jankowska AM, McDevitt MA et al (2011) CBL, CBLB, TET2, ASXL1 and IDH1/2 mutations and additional chromosomal aberrations constitute molecular events in chronic myeloid leukemia. Blood 117:198–206 Malpas JS (1986) Chemotherapy. In: Franks LM, Teich N (eds) Introduction to the cellular and molecular biology of cancer. Oxford University Press, Oxford, p 363 Michor F, Iwasa Y, Nowak MA (2006) The age incidence of CML can be explained by a one mutation model. Proc Natl Acad Sci U S A 103:14931–14934 Minden MD (1987) Oncogenes. In: Tannock IF, Hill RP (eds) The basic science of oncology. Pergamon Press, Oxford, pp 72–88 Mustjoki S, Jilg S, Jost PJ et al (2020) Model-based inference and classification of immunologic control mechanisms from TKI cessation and dose reduction in patients with CML. Cancer Res 80:2394–2406 Na HK, Surh YJ (2014) Oncogenic potential of Nrf2 and its principal target protein heme oxygenase-1. Free Radic Biol Med 67:353–365 Nieborowska-Skorska M, Kopinski PK, Ray R et al (2012) RAC2-MRC-cIII-generated ROS cause genomic instability in chronic myeloid leukemia stem cells and primitive progenitors. Blood 119:4253–4263 Nowell PC, Hungerford DA (1960) A minute chromosome in human chronic granulocytic leukemia. Science 132:1497 O’Hare T et al (2007) Bcr-Abl kinase domain mutations, drug resistance, and the road to a cure for chronic myeloid leukemia. Blood 110:2242–2248

References

151

Pronier E, Imanei A, Selimoglu-Buet D et al (2022) Macrophage migration-inhibitory factor is overproduced through EGR1 in TET2low resting momocytes. Commun Biol 5:110. https://doi. org/10.1038/s42003-022-03057-w Quintas-Cardama A, Kantarjian H, Jones D et al (2007) Dasatinib (BMS-354825) is active in Philadelphia chromosome-positive chronic myelogenous leukemia after imatinib and nilotinib (AMN107) therapy failure. Blood 109:497–499 Radich JP, DaiH MM et al (2006) Gene expression changes associated with progression and response in chronic myeloid leukemia. Proc Natl Acad Sci U S A 103:2794–2799 Radivoyevitch T (2015). http://epbi-radivot.cwru.edu/SEERaBOMB/SEERaBOMB.html Radivoyevitch T (2020) Parameter perturbations in a post-treatment chronic myeloid leukaemia model capture the essence of pre-diagnosis A-bomb survivor mysteries. Radiat Environ Biophys 60:41–47 Radivoyevitch T, Jankovic GM, Tiu RV et al (2014) Sex differences in the incidence of chronic myeloid leukemia in adults. Radiat Environ Biophys 53:55–63 Rowley JD (1973) A new consistent chromosomal abnormality in chronic myelogenous leukaemia identified by quinacrine fluorescence and Giemsa staining. Nature 243:290–293 Talpaz M, Shah NP, Kantarjian H et al (2006) Dasatinib in imatinib-resistant Philadelphia chromosome-positive leukemias. N Engl J Med 354:2531–2541 Tang M, Gonen M, Quintas-Cardama A et al (2011) Dynamics of chronic myeloid leukemia response to long-term targeted therapy reveal treatment effects on leukemic stem cells. Blood 118:1622–1631 Wang K, Hampson P, Hazeldine J et al (2012) Cyclin-dependent kinase 9 activity regulates neutrophil spontaneous apoptosis. PLoS One 7:e30128. https://doi.org/10.1371/journal.pone. 0030128 Wodarz D, Komarova NL (2014) Dynamics of cancer: mathematical foundations of oncology. World Scientific, Singapore, p 301–329, 449–457 Zimmerman K, Baldinger J, Mayerhofer B et al (2015) Activated AMPk boosts the Nrf2-HO-1 signaling axis – a role for the unfolded protein response. Free Radic Biol Med 88(Pt B):417–426

Chapter 8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Abstract Gene expression is epigenetically controlled, and the changes in gene expression that drive tumour progression are in some cases epigenetically determined. Those epigenetic changes may in turn be a result of genetic or chromosomal instability. Malignant progression caused by epigenetic silencing of function of tumour suppressor genes will often require inactivation of both alleles, though in some cases loss of a single allele may result in allelic insufficiency. CMML is a three-hit malignancy, caused by loss of function of a growthpromoter gene in conjunction with loss of function of both copies of an epigenetically controlled tumour suppressor gene. In many cases, this gene is TET2, though other genes involved in epigenetic regulation, such as IDH1 or IDH2, are sometimes responsible. The peak age at onset of CMML is 75; this age distribution of CMML is consistent with its three-hit aetiology. About 20% of CMML patients progress to a form of AML, but progression is slow, and most CMML patients succumb to other effects of the disease before it progresses to the acute stage. Hypomethylating agents such as decitabine are frequently effective treatment for CMML. Histone deacetylase inhibitors also have activity against some tumour types. Epigenetic changes alone cannot account for tumour progression, but in some cases they can be one factor in the multi-stage process.

The cells within a single human form a small population, tightly bound by their ecology. This population has low genetic variation. (It varies more in the marks established by “epigenetic” inheritance mechanisms . . .). Peter Godfrey-Smith, Darwinian Populations and Natural Selection (2009)

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_8. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_8

153

154

8.1

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Cancer as a Disease of Ageing: The Age Distribution of CMML

Chronic myelomonocytic leukaemia (CMML) is a condition that has features of both a myelodysplastic syndrome (MDS) and a myeloid leukaemia. It is characterised by an abnormally high count of circulating monocytes, with concomitant depletion of other myeloid cell types, as well as of erythrocytes and platelets. Approximately 20% of CMML patients progress to secondary AML, which has a very poor prognosis with a median overall survival of only ~6 months, and the remainder succumb to complications of bone marrow failure before such progression occurs (Issa 2013). The median age at onset of CMML is 75 years of age, the highest of any cancer in the Surveillance, Epidemiology and End Results (SEER) data (Fig. 8.1). Up to the age of 50, the incidence is negligible, rising exponentially thereafter, with the suggestion that it may reach a plateau somewhere after the age of 90. Incidence at all ages is higher in men than in women. The differentiation of myelomonocytic progenitor cells in the bone marrow is under epigenetic control suggesting that epigenetic changes may be involved in the aetiology of CMML (Hoeksema and de Winther 2016). All cells of the body contain the same DNA, but each of the 250 cell types in the body expresses a different sub-set of the approximately 23,000 genes that constitute the human genome. By “gene expression” is meant synthesis of messenger RNA, and thus synthesis of protein, from that particular gene. Some genes are expressed at some phases of the life cycle but not others; some genes are expressed in response to particular physiological or environmental conditions. Which genes are expressed, and when, is determined by chemical modification either of DNA itself or of the family of DNA-binding proteins known as the histones. The study of the control of gene expression by chemical modification of DNA and histones is termed epigenetics (Carey 2011). The epigenetic modifications to DNA consist of methylation of cytosine bases in promoter regions of DNA, which has the effect of silencing Fig. 8.1 Age distribution of CMML, plotted from the NCI SEER database (from Jackson and Radivoyevitch 2021)

8.2

Epigenetic Changes in Malignancy

155

expression of genes controlled by that promoter. The cytosines that are methylated occur in CpG sequences, and these sequences occur with high frequency in regulatory regions of DNA termed CpG islands. Histones are basic proteins that bind to DNA, and in so doing prevent access of RNA polymerases to that region of the DNA, blocking gene expression. Chemical modification (primarily acetylation) of basic groups on the histones makes them less basic, so more weakly attracted to DNA, and thus activates gene transcription. Why are there two mechanisms of control of gene expression? Yeast has acetylated histones, but not methylated DNA. DNA methylation appears to have evolved with multicellularity, suggesting that its primary function is cellular differentiation. Histone acetylation, on the other hand, may be primarily a way of adapting to changed metabolic circumstances (Carey 2011, p. 273). Ageing is generally considered to be an epigenetic phenomenon, a process of progressive gene silencing, though there are many competing theories. One of these is that ageing is a result of telomere shortening, which certainly correlates with age. It should be remembered that telomerase itself is epigenetically regulated. In many species, ranging from the fruit fly, Drosophila, to mice, nutritional deprivation seems to increase longevity. The mechanism of this effect remains obscure, though there appears to be a link between longevity and expression of the histone deacetylase (HDAC) enzyme, sir2 (Carey 2011, p. 274). Further research on this fascinating topic will undoubtedly shed further light on the relationship between epigenetics, ageing, and cancer.

8.2

Epigenetic Changes in Malignancy

All 250 cell types in the mammalian body contain the same DNA sequence, but the different cell types express different (though overlapping) sub-sets of their DNA. A particular cell type may express different genes at different stages in its life cycle— developing, mature, ageing. Epigenetic control of gene expression clearly evolved with the origin of multicellular organisms: it regulates the normal process of cellular differentiation. Mature tissues change their DNA expression pattern in response to physiological circumstances: the uterus of a pregnant mother must express different genes from a non-pregnant uterus; a lactating breast must produce milk proteins. The digestive system may change its gene expression in response to diet or nutritional status. Many disease states involve altered gene expression: in type I diabetes, the pancreatic beta cells have lost the ability to produce insulin. There are two mechanisms of epigenetic control of gene expression. The first of these involves chemical modification of DNA, specifically methylation of the 5-position of DNA cytosine bases. The methylation occurs only after the cytosine has been incorporated into DNA and is catalysed by DNA methyltransferases, DNMTs (of which there are three). DNMTs will only add a methyl group to a C that is followed by a G (a CpG sequence). CpG sequences are not randomly distributed throughout the genome, but tend to cluster just upstream of gene promoter regions. Methylation of these CpG sequences has the effect of blocking

156

8 Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

transcription of the adjoining gene—it has been epigenetically “silenced”. These patterns of DNA methylation are preserved after cell division. When a chromosome containing methylated cytosine is replicated, the corresponding cytosine base in the newly replicated DNA strand will be unmethylated; this will be recognised by DNMT1, which then methylates the appropriate cytosine in the newly replicated strand. In this way, when a differentiated cell replicates, the differentiated state is preserved. The other molecular mechanism of epigenetic regulation involves modification of histones, DNA-associated proteins, rather than DNA itself. Histones are basic (positively charged) proteins that can bind tightly to (negatively charged) DNA. Binding of histones prevents RNA polymerases and other proteins required for gene transcription from binding to DNA, thus silencing gene expression. When the lysine amino groups of histones are acetylated their positive charge is neutralised, and their attraction to DNA is weakened. A group of enzymes, histone deacetylases (HDACs) catalyse the removal of the acetyl groups from histones, increasing their binding to DNA, with resulting gene silencing. Why have we evolved two chemically distinct mechanisms of gene silencing? Apart from the general consideration that two ways of solving a problem confer greater flexibility, it seems likely that the two systems operate over different time scales. Tissue differentiation must be retained over the life of an organism and is handled by DNA methylation. Shorter-term adaptations, in response to altered nutritional or hormonal conditions, are more likely to be managed by histone acetylation. Early work on epigenetic changes in cancer was reviewed by Carey (2011). Sawan et al. (2008) demonstrated aberrant histone modification in a variety of tumour types. Some changes could be clearly related to tumorigenesis, such as silencing of tumour suppressor genes. Kalari and Pfeifer (2010) reviewed changes in DNA methylation patterns in cancer. Again, some reported changes were probably causally related to cancer, such as tumour suppressor gene silencing or methylation of homeobox genes, genes for developmental transcription factors. Other genes that were silenced were probably passenger genes. Epigenetic silencing of tumour suppressor genes, e.g. hypermethylation of the promoter region of VDL, was associated with clear-cell renal carcinoma (Herman et al. 1994). Interactions between cancer cells and surrounding stromal and immune cells, some of which are epigenetically controlled, can influence competition for survival (Zahir et al. 2020). Abnormal DNA methylation patterns can in some circumstances act as a driver of aneuploidy (Davidson 2014) indicating a possible link between epigenetic changes and chromosomal instability. Some DNA repair enzymes are epigenetically regulated, so that silencing of their genes can result in enhanced mutation rates (Jacinto and Esteller 2007; Lahtz and Pfeifer 2011; Bernstein et al. 2013). As discussed in Chap. 2, some cases of microsatellite instability in colorectal cancer are associated with hypermethylation of genes of the DNA mismatch repair system. The greatest degree of hypermethylation is seen in EB virus-positive gastric cancer (Usui et al. 2021). An analysis of breast cancer showed frequent changes in methylation patterns (Bernstein et al. 2013; Lin et al. 2015; Tanas et al. 2019); although

8.3

Epigenetic Gene Silencing and the Role of TET2

157

showing individual variation, the patterns of hypermethylated regions fell into six clusters. A computational analysis of breast cancer methylation patterns with clinical data from the Cancer Genome Atlas showed that DNA methylation could be used as a biomarker to predict invasiveness (Wang et al. 2020). Epigenetic changes have also been implicated in the onset of metastasis (Chattergee et al. 2018), and histone demethylation has been linked to hypoxic reprogramming in cancer metabolism (Chang et al. 2019).

8.3

Epigenetic Gene Silencing and the Role of TET2

In considering epigenetic changes as possible drivers of tumour progression, it is important to note that epigenetic change may be a consequence of genetic change. Tan et al. (2015) reported that DNMT3A mutations alone transformed haematopoietic stem cells into pre-leukaemic stem cells without causing fullblown leukaemia. Approximately 50% of CMML patients carry TET2 mutations (Ko et al. 2010). Other mutations commonly found in CMML include ASXL1 (40%), SRSF2 (40%), RUNX1 (20%), constitutively activated N-ras, JAK2, and mutant p53 (Adamson et al. 1995; Gur et al. 2018;. Kunimoto et al. 2018). Patel et al. (2017) reported that 59% of CMML patients had mutant TET2, SRSF2, or ASXL1. Progression to acute disease was associated with acquisition of biallelic TET2 mutations, Ras family, or spliceosomal gene mutations. The enzyme TET2 converts 5-methylcytosine (5mC) in DNA to 5-hydroxymethylcytosine (5hmC) and further oxidises 5hmC to 5-formylcytosine (5fC) and 5fC to carboxylic acid cytosine (5caC). This causes 5mC conversions to C either actively by base excision repair of 5fC/5caC or passively by DNA methyltransferase 1 (DNMT1) not recognising 5hmC after DNA replication (Cui et al. 2016). The net result is reactivation of the epigenetically silenced gene. TET2 mutations associated with myeloid malignancies compromise catalytic activity and result in DNA hypermethylation (Jackson and Radivoyevitch 2021). Small hairpin RNA (shRNA) targeted against TET2 in mouse haematopoietic precursors skewed their differentiation towards monocyte/macrophage lineages in culture, indicating that TET2 mutations could be causally linked with CMML (Ko et al. 2010). In AML, Molenaar et al. (2014) reported frequent mutations in isocitrate dehydrogenase 1 and 2 (IDH1, IDH2). These enzymes generate α-ketoglutarate, a substrate for TET2. The IDH inhibitor, ivosidenib, gave durable remissions in IDH1-mutated AML patients with relapsed and refractory disease (DiNardo et al. 2018). IDH1/2 mutations sensitise AML to PARP inhibition (Molenaar et al. 2018). Of 299 driver mutations listed in TCGA, 12 are involved in histone methylation and demethylation (Chang et al. 2019). Mutations in DNMT3A have been reported in AML and are predictive for shorter overall survival (Im et al. 2014; Metzeler et al. 2016).

158

8.4

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Loss-of-Function Mutations and Tumour Suppressor Genes

There is a long history in the cancer literature of explaining cancer as a multi-step process. The age distribution of most malignancies pointed to cancer being the end result of a series of mutations. Research on carcinogenesis argued for cancer being a two-step process, based upon conclusions that initiation and progression were separate processes (Moolgavkar and Knudson 1981; Phillips 1987; Wodarz and Komarova 2014). The discovery of tumour suppressor genes complicated the picture further, in that both alleles need to be inactivated or deleted to cause transformation. How does this observation fit with the conclusion of Chap. 3, that cancer is the result of inactivation or over-ride of two cell cycle checkpoints? If a tumour suppressor gene acts on the G1:S checkpoint (e.g. PTEN), then malignant transformation will require loss of both alleles of the tumour suppressor, plus a mutation that results in genetic or chromosomal instability. If, however, the tumour suppressor gene acts to cause genetic or chromosomal instability (or, as in the case of TET-2, epigenetic instability) for transformation to occur, loss of both alleles of the tumour suppressor must be accompanied, or preceded, by a mutation that causes G1:S checkpoint override. The result, in both cases, is a three-hit malignancy. CMML appears to be an example of this, consistent with its age distribution. The dynamics of CMML will be further explored using a cytokinetic/pharmacodynamic model, as discussed below.

8.5

Hypomethylating Agents as Treatments for CMML

Decitabine (2′-deoxy-5-aza-cytidine) is a close analogue of the DNA precursor, 2′-deoxycytidine (Fig. 8.4). As an antimetabolite, its biochemical behaviour closely resembles that of natural 2′-deoxycytidine. Inside the cell it is phosphorylated to the 5′-mono-, di-, and triphosphates, and the triphosphate can then be incorporated into DNA in place of dCTP. During DNA synthesis and mRNA synthesis it base-pairs with G, and (unlike many antimetabolites) it is not highly cytotoxic. However, having a nitrogen atom (in place of carbon) at the 5-position of the pyrimidine ring, it is unable to accept a methyl group. Thus, those genes in a decitabine-treated cell that would normally be silenced by methylation (or some fraction of them, depending on the dose) remain active. The DNA is said to be hypomethylated. Decitabine is used in the treatment of CMML, and the related nonmalignant condition, MDS. In a study of CMML and MDS combined, response rates were higher in mutant TET2 cases (Itzykson et al. 2011). This is consistent with CMML being a disease of abnormal DNA methylation, and with TET2 being a primary regulator of methylation patterns (Fig. 8.2). A disadvantage of decitabine is that, like cytidine and 2′-deoxycitidine, it is rapidly deaminated by the enzyme cytidine deaminase to its uridine derivative, 2′-deoxy-5-azauridine, which is inactive. Since intestine and liver contain high

8.5

Hypomethylating Agents as Treatments for CMML

159

Fig. 8.2 Cell populations described by the CMMLsim model. Haematopoietic stem cells (HSC) in bone marrow differentiate into four lineages. Progenitor cells in each lineage possess self-renewal capability and also undergo further differentiation into non-proliferating end cells that are released into the blood (from Jackson and Radivoyevitch 2021)

levels of cytidine deaminase, decitabine is not active when administered orally and must be given by intravenous injection. However, if administered with tetrahydrouridine (THU), a potent inhibitor of cytidine deaminase, decitabine is orally active. A close analogue of decitabine, 5-azacytidine, differs only in having ribose, instead of 2′-deoxyribose, attached to the 5-azacytosine base. Like decitabine, it has clinical activity against CMML, but requires conversion (by ribonucleotide reductase) of its diphosphate, and subsequently by nucleoside diphosphate kinase, in order to form the active 5′-triphosphate. Exposure to decitabine and 5-azacytidne depletes the epigenetic regulator DNMT1, which results in changes to the regulation of pyrimidine nucleotide metabolism, including downregulation of thymidylate synthase and ribonucleotide reductase (Gu et al. 2021). However, continuous exposure to decitabine or 5-azacytidine results in the pyrimidine regulatory network preventing DNMT1 depletion, causing resistance to the hypomethylating agents (Figs. 8.3 and 8.4). Is there any evidence that progression of carcinomas is epigenetically driven? Methylation mutants have been reported in some AML cases. We know that epigenetic changes are not normally associated with mutation, but may be the result of ageing, adaptation to a different environment, or lifestyle changes. Is it possible that some of these factors could influence epigenetic changes that drive tumour progression? There is evidence for epigenetic effects associated with type 2 diabetes.

160

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Fig. 8.3 Cytokines involved in monocyte proliferation, differentiation and trafficking (from Jackson and Radivoyevitch 2021) Fig. 8.4 Decitabine, a hypomethylating agent

Could the Warburg effect be linked to epigenetic changes? Might the inhibition of tumour progression by metformin be epigenetically mediated? TET2 mutant clones are sometimes seen in elderly individuals, and over 40% of these are ancestral (Hirsch et al. 2018). These TET2 mutations appear to cause a predisposition to myelodysplastic syndrome, and subsequent mutations may lead to leukaemias. It seems clear that epigenetic changes alone cannot cause malignancy, based upon the fact that, of thousands of tumours whose DNA has been sequenced, none have been found with a totally normal genotype. Myelodysplastic syndrome (MDS) is an example of a nonmalignant epigenetic disease (Issa 2013; Gerds 2014). The mutations found in MDS have been reviewed by Bartram (1992) and by Jin et al. (2018). Haploinsufficiency for these genes (i.e. loss or inactivation of a single allele)

8.5

Hypomethylating Agents as Treatments for CMML

161

Fig. 8.5 Regulation of the differentiation of myeloid precursor cells into monocytes and neutrophils, and its disruption in CMML. Feedback inhibition effects are shown with dashed lines. Upper panel: normal myeloid cells; lower panel: dysplastic cells. Note that the feedback inhibition of monocyte differentiation is missing in CMML. P = normal myeloid progenitors; M = normal monocytes; N = normal neutrophils; DPP = dysplastic progenitors; DPM = dysplastic monocytes; DPN = dysplastic neutrophils (from Jackson and Radivoyevitch 2021)

tends to result in MDS rather than CMML (Davids and Steensma 2010). MDS is a serious, life-threatening, condition, it is a condition that can progress in severity, but it lacks the defining quality of malignancy: it does not have the property of increasing cellular heterogeneity with time. Perhaps MDS can be considered as a premalignant condition, a benign dysplasia that requires a further mutation, or epigenetic change, to result in full-blown malignancy? Does the fact that only 20% of CMML patients progress to acute leukaemia mean that only those 20% of cases are actually malignant? Or would the other 80% have progressed to acute leukaemia had they not succumbed to the serious morbidities of MDS first? To answer this would require showing that the 20% and the 80% groups have distinct patterns of gene expression. This question has not yet been definitively answered, but the data so far do not support such a distinction. Cellular heterogeneity is a precondition for the process of Darwinian selection that drives the progression of malignant disease. Is the bone marrow population in CMML patients more heterogeneous than that of MDS patients? In the case of CMML, at least, epigenetic changes contribute to tumour progression: can progression ever be attributed to epigenetic changes alone? (Fig. 8.5) Epigenetic control of gene expression involves histone modification as well as DNA methylation. A family of 11 histone deacetylase (HDAC) enzymes (in mammals) is involved in removal of acetyl groups from histones which results in tighter binding of histone to DNA, with resulting shut-down of gene expression (Carey 2011). These HDAC enzymes have an active site zinc atom which is essential for activity. Another family of enzymes, the sirtuins, do not contain zinc and require

162

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Fig. 8.6 Vorinostat, an HDAC inhibitor

NAD+ for activity. A number of HDAC inhibitors have been shown to have anticancer activity. Vorinostat (suberanilohydroxamic acid, SAHA) acts as a zinc chelator (Fig. 8.6). It is approved for treatment for cutaneous T cell lymphoma and also has demonstrated clinical activity in MDS, and in Sézary syndrome (a T cell lymphoma), and (in combination with other agents) in glioblastoma and non-small cell lung cancer.

8.6

Evolutionary Dynamics of CMML

Biochemical modelling has suggested ways in which the properties of CMML can be explained by epigenetic changes (Jackson and Radivoyevitch 2021). A pharmacodynamic model of CMML treatment, termed CMMLsim.R, is described in the online appendix. Cell lineages modelled by the program, and their interconversions, are summarised in Fig. 8.2. Bone marrow haematopoietic stem cells (HSC) have four alternative routes of differentiation. Progenitor cells in each lineage have selfrenewal capability and can also further differentiate into non-proliferating end cells that are released into the circulation. Lymphoid cells and their precursors are not described in detail in this model. The other cell lineages result from HSC stimulated to differentiate by specific growth factors: erythroid differentiation is triggered by erythropoietin, and megakaryocyte differentiation by thrombopoietin. Differentiation of HSC into granulocyte-macrophage progenitor cells is caused by granulocyte-macrophage colony-stimulating factor (GM-CSF). There are then two alternative fates: in presence of granulocyte colony-stimulating factor (G-CSF) GM progenitor cells form neutrophils, and in presence of macrophage colony-stimulating factor (M-CSF) they form monocytes. Fig. 8.3 summarises the differentiation of HSC to macrophages, and the cytokines involved. This complex system is controlled by a network of feedback effects. Circulating neutrophils express surface receptors for G-CSF; this binding has no stimulatory effect, because mature neutrophils do not replicate. If the neutrophil count becomes abnormally high, binding of G-CSF to these neutrophil receptors depletes unbound G-CSF in blood, so differentiation of granulocyte-macrophage progenitors to neutrophils decreases. Similarly, circulating monocytes express M-CSF receptors. These feedback effects are shown in Fig. 8.5 (dashed lines). The upper part of the figure shows the control of differentiation of GM progenitors in normal subjects, and the lower part of the figure shows GM progenitor differentiation in patients with MDS or CMML. Note that the feedback inhibition of monocyte differentiation is defective or absent in MDS and CMML. It is epigenetically silenced: deletion of the negative feedback causes excess production of monocytes, depletion of HSC and progenitor populations, with resulting

8.6

Evolutionary Dynamics of CMML

163

anaemia and thrombocytopenia. The cause of this feedback silencing is believed to be loss of activity of the enzyme TET2, the master regulator of DNA methylation. Since TET2 acts by removing methyl groups from DNA cytosine bases, loss of TET2 causes excess methylation, silencing of the negative feedback, and thus activation of monocyte production. Table 8.1 shows output from the CMMLsim.R model, simulating the normal system in a steady state. Note that Table 8.1 shows only a small sub-set of the 27 state variables of the model (all variables can be displayed by the online model). Potential effects of various mutations were then explored. The mutations were assumed to occur either in haematopoietic stem cells (HSC), which rarely divide, or in proliferating progenitor cells. The effect of an inactivating TET2 mutation, occurring in both alleles of a single HSC or myeloid progenitor cell, TET2 -/-, on the population is modelled in Table 8.2. Despite the mutant cells being assumed to have the same growth rate as wild-type cells, the mutant cells were rapidly eliminated (Fig. 8.7). Similarly, in the absence of a compensating growth advantage, TET2 +/- progenitors were also competed out. N-ras mutations, which are frequently found in CMML, were assumed by the model to stimulate cell growth, which they do by elevating cyclin D levels and thereby over-riding the G1:S checkpoint. HSC or progenitor cells with mutant N-ras can proliferate independently of GM-CSF, which frees them from feedback inhibition by circulating monocytes and neutrophils. The effects of an N-ras mutation (again, assumed to occur in a single HSC or myeloid progenitor) are shown in Table 8.3. The model predicted a moderate (85%) increase in the total (normal +mutant) monocyte count, almost no change in total white blood cells, and a concomitant decline in erythrocytes and platelets. Thus, N-ras mutation can cause moderate hyperplasia of the myeloid population, but it does not meet the definition of CMML. The calculated effect of an N-ras mutation followed by a mutation in a single TET2 allele is shown in Table 8.4. Gradually (over a period of about two years), the double-mutant progenitors became the dominant bone marrow population, and the total circulating monocyte count increased about 2.5-fold. However, if the TET2 mutation occurred first, the mutant clone became extinct before an N-ras mutation could occur, and CMML did not result. Simulating the situation of a clone of cells with mutant N-ras where a TET2 mutation in one allele was followed by a TET2 mutation in the other allele is summarised in Table 8.5 and Fig. 8.8. The condition was predicted to progress from a single double-mutant cell to clinical disease in about 2 years: this combination of mutations resulted in CMML. There were minor decreases in erythrocytes and platelets, increased total white blood cells, and a large increase in total monocytes (over 15-fold), of which most were dysplastic. There is nothing specific about this effect of N-ras: constitutively activating Ras mutations results in elevated cellular levels of cyclin D, over-riding the G1:S cell cycle checkpoint (Jackson et al. 2017). Other mutations that result in G1:S checkpoint over-ride, such as JAK2 mutations (also found in some cases of CMML), are likely to have the same effect.

P 6.620e+006 6.211e+006 6.980e+006 6.823e+006 6.620e+006 6.649e+006 6.714e+006 6.717e+006 6.702e+006 6.700e+006 6.705e+006 6.707e+006 6.706e+006 6.705e+006 6.706e+006

WBC 8.457e+006 8.386e+006 8.492e+006 8.682e+006 8.610e+006 8.552e+006 8.572e+006 8.595e+006 8.595e+006 8.590e+006 8.591e+006 8.592e+006 8.593e+006 8.593e+006 8.593e+006

Platelets 2.950e+008 3.548e+008 2.898e+008 2.954e+008 2.935e+008 2.927e+008 2.942e+008 2.935e+008 2.932e+008 2.933e+008 2.934e+008 2.933e+008 2.933e+008 2.933e+008 2.933e+008

RBC 4.820e+009 5.128e+009 4.857e+009 4.788e+009 4.811e+009 4.827e+009 4.820e+009 4.815e+009 4.815e+009 4.817e+009 4.817e+009 4.816e+009 4.816e+009 4.816e+009 4.816e+009

Monocytes 3.570e+005 3.430e+005 3.456e+005 3.512e+005 3.459e+005 3.444e+005 3.461e+005 3.468e+005 3.466e+005 3.464e+005 3.465e+005 3.466e+005 3.466e+005 3.466e+005 3.466e+005

MKcytes 1.030e+005 1.084e+005 1.006e+005 1.045e+005 1.031e+005 1.032e+005 1.036e+005 1.033e+005 1.033e+005 1.033e+005 1.033e+005 1.033e+005 1.033e+005 1.033e+005 1.033e+005

Neutro 5.100e+006 5.043e+006 5.146e+006 5.330e+006 5.265e+006 5.208e+006 5.226e+006 5.248e+006 5.248e+006 5.244e+006 5.244e+006 5.246e+006 5.246e+006 5.246e+006 5.246e+006

Erythro 1.600e+008 1.641e+008 1.564e+008 1.580e+008 1.600e+008 1.597e+008 1.591e+008 1.590e+008 1.592e+008 1.592e+008 1.592e+008 1.591e+008 1.592e+008 1.592e+008 1.592e+008

8

P, myeloid progenitor cells; WBC, total leucocytes; RBC, circulating erythrocytes; Mkcytes, bone marrow megakaryocytes; Neutro, circulating neutrophils; Erythro, bone marrow erythroblasts Time in days. Data shown as cells/L of blood or bone marrow; from Jackson and Radivoyevitch (2021)

Time 0 120 240 360 480 600 720 840 960 1080 1200 1320 1440 1560 1680

Table 8.1 The uninhibited haematopoietic system as modelled by CMMLsim

164 Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

P 7.036e+006 7.302e+006 7.557e+006 7.292e+006 7.204e+006 7.257e+006 7.284e+006 7.272e+006 7.262e+006 7.264e+006 7.267e+006 7.267e+006 7.266e+006 7.266e+006 7.266e+006

MDP 1.000e+000 4.724e-001 2.165e-001 9.250e-002 4.019e-002 1.783e-002 7.895e-003 3.477e-003 1.531e-003 6.750e-004 2.977e-004 1.313e-004 5.787e-005 2.551e-005 1.125e-005

WBC 5.725e+006 6.057e+006 5.988e+006 5.959e+006 5.861e+006 5.839e+006 5.859e+006 5.866e+006 5.861e+006 5.859e+006 5.860e+006 5.860e+006 5.860e+006 5.860e+006 5.860e+006

Platelets 3.060e+008 3.552e+008 2.911e+008 3.064e+008 3.018e+008 3.019e+008 3.032e+008 3.024e+008 3.024e+008 3.026e+008 3.026e+008 3.025e+008 3.025e+008 3.025e+008 3.025e+008

RBC 4.920e+009 5.152e+009 4.888e+009 4.868e+009 4.893e+009 4.897e+009 4.891e+009 4.890e+009 4.891e+009 4.892e+009 4.892e+009 4.891e+009 4.891e+009 4.891e+009 4.891e+009

Monocytes 3.575e+005 3.715e+005 3.621e+005 3.603e+005 3.564e+005 3.566e+005 3.576e+005 3.576e+005 3.573e+005 3.573e+005 3.574e+005 3.574e+005 3.574e+005 3.573e+005 3.573e+005

MDM 7.000e-002 2.954e-001 1.925e-001 1.006e-001 4.756e-002 2.173e-002 9.792e-003 4.371e-003 1.939e-003 8.574e-004 3.787e-004 1.672e-004 7.374e-005 3.252e-005 1.434e-005

MDP, myelodysplastic progenitors in bone marrow; monocytes, total monocytes (normal + dysplastic); MDM, circulating dysplastic monocytes. Other abbreviations as defined in Table 8.1. Time in days. Data shown as cells/L of blood or bone marrow; from Jackson and Radivoyevitch (2021)

Time 0 120 240 360 480 600 720 840 960 1080 1200 1320 1440 1560 1680

Table 8.2 The effect of an inactivating TET2 mutation as modelled by CMMLsim

8.6 Evolutionary Dynamics of CMML 165

166

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Fig. 8.7 Modelling the consequences of a single cell with a TET2 mutation (from Jackson and Radivoyevitch 2021)

The CMMLsim.R model was also used to study effects of mutant p53. Cells with mutant p53 have accelerated proliferation, because of G1 checkpoint over-ride, but their proliferation still requires GM-CSF. Like N-ras mutations or JAK2 mutations, p53 mutations decrease the G1:S delay factor but they do so by different mechanisms. Normal p53 stimulates transcription of p21, which acts as an inhibitor of cdk2/cyclin E and cdk4/cyclin D, required for cell cycle progression from G1 into S phase. Because of this mechanistic difference, N-ras mutations and p53 mutations have independent effects on G1:S progression delay (Table 8.6). The calculations described in Tables 8.1–8.6 simulated the situation where one or more mutations have produced a single mutant progenitor cell, or a small clone of such cells, in the bone marrow. In the following simulations, the starting point was a post-diagnosis state of CMML, and the intent was to predict the pharmacodynamic effects of treatment. Initially the program was used to model pharmacodynamic effects of the hypomethylating agent, decitabine. Table 8.7 shows the calculated effect of several steady-state plasma concentrations of decitabine. In agreement with clinical observation, over several weeks of treatment, decitabine caused a dosedependent decline in circulating monocytes and in total white blood cells. High concentrations of decitabine, after prolonged treatment, reduced monocytes and total white cells below normal, probably reflecting the fact that in addition to its hypomethylating effect, decitabine can be myelosuppressive. All concentrations of decitabine modelled increased the erythrocyte count and the platelet count towards, but not above, their normal values. Simulated treatment of CMML with the DNA strand breaking agent, CNDAC, is summarised in Table 8.8. As with decitabine, this S phase-specific agent normalised the erythrocyte and platelet counts and decreased the total WBC and monocyte counts. The number of dysplastic progenitors in bone marrow was markedly reduced at the higher concentrations. The pharmacodynamic effects of CNDAC and the

P 6.620e+006 4.237e+006 5.444e+006 7.123e+006 6.554e+006 2.312e+006 8.192e+005 5.030e+005 4.503e+005 4.409e+005 4.332e+005 4.292e+005 4.292e+005 4.288e+005 4.275e+005

MDP 1.000e+000 6.504e+000 3.216e+002 3.256e+004 1.583e+006 1.294e+007 1.587e+007 1.377e+007 1.368e+007 1.479e+007 1.485e+007 1.449e+007 1.450e+007 1.466e+007 1.466e+007

WBC 1.630e+007 1.003e+007 7.825e+006 8.212e+006 9.048e+006 1.346e+007 1.748e+007 1.691e+007 1.598e+007 1.626e+007 1.663e+007 1.655e+007 1.646e+007 1.652e+007 1.657e+007

Platelets 2.950e+008 3.548e+008 3.155e+008 3.074e+008 2.923e+008 2.407e+008 1.401e+008 1.316e+008 1.306e+008 1.262e+008 1.235e+008 1.222e+008 1.211e+008 1.207e+008 1.206e+008

RBC 4.820e+009 5.128e+009 5.104e+009 4.905e+009 4.739e+009 4.329e+009 3.242e+009 3.217e+009 3.580e+009 3.499e+009 3.341e+009 3.363e+009 3.414e+009 3.399e+009 3.377e+009

Monocytes 3.570e+005 2.878e+005 2.859e+005 3.406e+005 3.755e+005 5.753e+005 6.721e+005 6.466e+005 6.317e+005 6.543e+005 6.645e+005 6.593e+005 6.574e+005 6.606e+005 6.621e+005

MDM 0.000e+000 1.385e-001 4.671e+000 4.230e+002 2.338e+004 3.334e+005 5.750e+005 6.035e+005 6.046e+005 6.318e+005 6.437e+005 6.391e+005 6.372e+005 6.405e+005 6.421e+005

Abbreviations as defined in Tables 8.1 and 8.2. Time in days. Data shows as cells/L of blood or bone marrow; from Jackson and Radivoyevitch (2021)

Time 0 120 240 360 480 600 720 840 960 1080 1200 1320 1440 1560 1680

Table 8.3 The effect of a constitutively active N-ras mutation as modelled by CMMLsim

8.6 Evolutionary Dynamics of CMML 167

P 6.620e+006 4.237e+006 5.444e+006 7.125e+006 6.656e+006 2.602e+006 8.841e+005 5.280e+005 4.635e+005 4.536e+005 4.450e+005 4.406e+005 4.406e+005 4.403e+005 4.389e+005

MDP 1.000e+000 6.139e+000 2.849e+002 2.723e+004 1.288e+006 1.202e+007 1.573e+007 1.372e+007 1.345e+007 1.456e+007 1.470e+007 1.433e+007 1.432e+007 1.448e+007 1.449e+007

WBC 1.630e+007 1.003e+007 7.825e+006 8.212e+006 9.007e+006 1.298e+007 1.742e+007 1.704e+007 1.606e+007 1.628e+007 1.668e+007 1.662e+007 1.651e+007 1.657e+007 1.663e+007

Platelets 2.950e+008 3.548e+008 3.155e+008 3.074e+008 2.926e+008 2.484e+008 1.470e+008 1.348e+008 1.346e+008 1.305e+008 1.279e+008 1.264e+008 1.254e+008 1.250e+008 1.249e+008

RBC 4.820e+009 5.128e+009 5.104e+009 4.905e+009 4.742e+009 4.403e+009 3.328e+009 3.232e+009 3.600e+009 3.544e+009 3.383e+009 3.396e+009 3.448e+009 3.436e+009 3.414e+009

Monocytes 3.570e+005 2.878e+005 2.859e+005 3.408e+005 3.847e+005 7.037e+005 9.109e+005 8.878e+005 8.646e+005 8.947e+005 9.113e+005 9.048e+005 9.015e+005 9.058e+005 9.081e+005

MDM 0.000e+000 2.400e-001 7.532e+000 6.511e+002 3.434e+004 4.711e+005 8.219e+005 8.510e+005 8.432e+005 8.775e+005 8.957e+005 8.897e+005 8.864e+005 8.907e+005 8.931e+005

8

Abbreviations as defined in Tables 8.1 and 8.2. Time in days. Data shows as cells/L of blood or bone marrow; from Jackson and Radivoyevitch (2021)

Time 0 120 240 360 480 600 720 840 960 1080 1200 1320 1440 1560 1680

Table 8.4 The effect of an N-ras mutation followed by mutation of a single TET2 allele

168 Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

P 6.620e+006 4.237e+006 5.444e+006 7.133e+006 7.051e+006 5.725e+006 2.699e+006 1.356e+006 9.281e+005 8.410e+005 8.049e+005 7.830e+005 7.800e+005 7.821e+005 7.805e+005

MDP 1.000e+000 3.465e+000 9.209e+001 4.998e+003 1.487e+005 2.477e+006 9.594e+006 9.761e+006 8.605e+006 9.318e+006 9.854e+006 9.610e+006 9.481e+006 9.592e+006 9.633e+006

WBC 1.630e+007 1.003e+007 7.825e+006 8.209e+006 8.817e+006 9.608e+006 1.482e+007 1.746e+007 1.640e+007 1.600e+007 1.659e+007 1.673e+007 1.656e+007 1.655e+007 1.663e+007

Platelets 2.950e+008 3.548e+008 3.155e+008 3.075e+008 2.939e+008 2.851e+008 2.453e+008 2.118e+008 2.249e+008 2.281e+008 2.289e+008 2.273e+008 2.265e+008 2.270e+008 2.272e+008

RBC 4.820e+009 5.128e+009 5.104e+009 4.905e+009 4.754e+009 4.745e+009 4.385e+009 3.998e+009 4.259e+009 4.372e+009 4.271e+009 4.240e+009 4.278e+009 4.282e+009 4.268e+009

Monocytes 3.570e+005 2.878e+005 2.860e+005 3.409e+005 3.789e+005 7.797e+005 3.566e+006 5.247e+006 5.087e+006 5.074e+006 5.377e+006 5.438e+006 5.365e+006 5.368e+006 5.402e+006

MDM 0.000e+000 8.224e-001 1.342e+001 7.444e+002 2.576e+004 5.125e+005 3.464e+006 5.216e+006 5.076e+006 5.069e+006 5.373e+006 5.434e+006 5.361e+006 5.364e+006 5.398e+006

Abbreviations as defined in Tables 8.1 and 8.2. Time in days. Data shows as cells/L of blood or bone marrow; from Jackson and Radivoyevitch (2021)

Time 0 120 240 360 480 600 720 840 960 1080 1200 1320 1440 1560 1680

Table 8.5 The effect of an N-ras mutation followed by mutation of both TET2 alleles

8.6 Evolutionary Dynamics of CMML 169

170

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Fig. 8.8 Progression from a single double-mutant cell to clinical disease, modelled by CMMLsim.R (from Jackson and Radivoyevitch 2021).

Table 8.6 Interaction of mutations in p53, N-ras, and TET2

p53 Normal Normal Normal Normal Normal Normal Mutant Mutant Mutant Mutant Mutant Mutant

N-ras Normal Normal Normal Mutant Mutant Mutant Normal Normal Normal Mutant Mutant Mutant

TET2 1.000 0.500 0.001 1.000 0.500 0.001 1.000 0.500 0.001 1.000 0.500 0.001

Total monocytes 3.48e+5 3.51e+5 3.48e+5 6.65e+6 9.12e+5 5.43e+6 7.38e+5 7.38e+5 6.65e+6 1.02e+7 1.41e+7 1.16e+7

Total WBC 8.63e+6 8.63e+6 8.63e+6 1.67e+7 1.67e+7 1.67e+7 1.86e+7 1.86e+7 1.93e+7 2.61e+7 2.63e+7 2.94e+7

Values of the G1:S checkpoint delay parameter: normal, 2.0; mutant N-ras, 0.5; mutant p53, 0.4 p53 and N-ras both mutated, 0.2. From Jackson and Radivoyevitch (2021)

hypomethylating agent, decitabine, act by unrelated mechanisms. If we define the in vivo 50% inhibitory concentration (IC50) of anti-CMML drugs as the concentration that reduces total monocytes by 50%, the IC50 of CNDAC, as calculated by the model is 85.7 nM, and the IC50 of decitabine is 28.6 nM. If we combined half IC50 levels of each drug, i.e. 42.85 nM CNDAC + 14.3 nM decitabine, the predicted result was a 52.2% decrease in the monocyte count, i.e. a combined effect slightly greater than additive. In a more rigorous analysis, the interaction of the CNDAC prodrug,

8.6

Evolutionary Dynamics of CMML

171

Table 8.7 Treatment of CMML with decitabine Decitabine (μM) 0 0.3 0.5 1.0 2.0 3.0

RBC 4.33e+9 4.50e+9 4.80e+9 5.21e+9 5.24e+9 5.24e+9

Platelets 4.54e+8 4.75e+8 5.13e+8 5.66e+8 5.70e+8 5.70e+8

Total WBC 1.70e+7 1.13e+7 9.46e+6 6.79e+6 4.11e+6 3.73e+6

Monocytes 5.64e+6 8.62e+5 6.07e+5 3.46e+5 1.28e+5 9.47e+4

Cells/L of blood, as calculated by CMMLhs4.R. Values are steady-state levels after three months of treatment. From Jackson and Radivoyevitch (2021) Table 8.8 Effect of plasma concentration of CNDAC on established CMML CNDAC (μM) 0 0.05 0.1 0.3 0.5 1.0

RBC 4.33e+9 4.77e+9 5.02e+9 5.24e+9 5.24e+9 5.24e+9

Platelets 4.54e+8 5.10e+8 5.42e+8 5.70e+8 5.70e+8 5.70e+8

Total WBC 1.70e+7 1.26e+7 1.01e+7 5.61e+6 4.14e+6 3.73e+6

Monocytes 5.64e+6 3.65e+6 2.55e+6 6.87e+5 1.52e+5 9.46e+4

Cells/L of blood, as calculated by CMMLhs4.R. Values are steady-state levels after three months of treatment. From Jackson and Radivoyevitch (2021)

sapacitabine, with decitabine was modelled. Results are shown in isobol format in Fig. 8.9. The isobol is slightly concave, indicating a minor degree of synergy. Analysis of the modelled data by the method of Greco et al. (1990) gave an alpha factor of +0.34. In this form positive alpha factors indicate synergy, alpha of zero corresponds to independence, and negative values indicate antagonism. Modelling the effect of TET2 haploinsufficiency, i.e. the loss of activity of one of the TET2 alleles by HSC or myeloid progenitors, indicated very little phenotypic effect. However, if the loss of one TET2 allele was preceded by a p53 mutation, the predicted result was a form of myelodysplasia, in which the erythrocyte and platelet counts were depressed, total white cells were moderately elevated (2.1-fold) and monocytes were 2.8-fold increased. Treatment with decitabine had very little effect on this condition (Table 8.9). At very high plasma concentrations there was a modest decrease in circulating monocytes, though this did not reach normalisation. Decitabine had no effect on the erythrocyte or platelet count. Interestingly, CNDAC was predicted to be active against TET2 haploinsufficiency, restoring erythrocyte and platelet counts to normal, and reducing the elevated total white blood cell and monocyte counts (Table 8.10). Decitabine acts both as a hypomethylating agent and as an S phase-specific cytotoxic agent. It was thus of interest to explore which of the two effects was the primary cause of its antileukaemic activity. This was done computationally by removing one or other effect of decitabine from the model. Figure 8.10 shows the dose-response curves for decitabine assuming either that both mechanisms were

172

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Fig. 8.9 Isobol of the sapacitabine + decitabine interaction (from Jackson and Radivoyevitch 2021) Table 8.9 Treatment of MDS resulting from TET2 haploinsufficiency with decitabine Decitabine (μM) 0 1.0 10.0 30.0

RBC 3.08e+9 3.07e+9 3.04e+9 3.04e+9

Platelets 8.34e+7 8.34e+7 7.95e+7 7.92e+7

Total WBC 1.86e+7 1.86e+7 1.85e+7 1.85e+7

Monocytes 1.01e+6 1.01e+6 7.70e+5 7.48e+5

Cells/L of blood, as calculated by CMMLsim.R. The model assumes mutant p53 and TET2 = 0.5. Values are steady-state levels after three months of treatment. From Jackson and Radivoyevitch (2021) Table 8.10 Treatment of MDS resulting from TET2 haploinsufficiency with CNDAC CNDAC (μM) 0 0.1 1.0

RBC 3.08e+9 4.80e+9 5.24e+9

Platelets 8.34e+7 2.91e+8 3.45e+8

Total WBC 1.86e+7 1.14e+7 4.02e+6

Monocytes 1.01e+6 3.04e+5 1.16e+5

Cells/L of blood, as calculated by CMMLsim.R. The model assumes mutant p53 and TET2 = 0.5. Values are steady-state levels after three months of treatment. From Jackson and Radivoyevitch (2021)

operative (“two targets”) or that it was acting purely as a hypomethylating agent (“one target”). Hypomethylation alone resulted in an extensive decrease in circulating monocytes. When the cytotoxic effect of decitabine was also included in the

8.6

Evolutionary Dynamics of CMML

173

Fig. 8.10 Dose-response curves for decitabine. One target: decitabine acts as a hypomethylating agent only; two targets: decitabine acts as a hypomethylating agent and an S-phasespecific cytotoxic agent (from Jackson and Radivoyevitch 2021)

model, its effect was greater, indicating that S-phase-specific cytotoxicity made a significant contribution to the antileukaemic effect of decitabine. Decitabine is usually administered by intravenous infusion. It is orally available, but following absorption from the intestine, enters the liver via the hepatic portal vein, and is rapidly deaminated by hepatic cytidine deaminase, so that its effective oral bioavailability is less than 10%. Co-administration of THU at 400 mg/m2 increased the oral bioavailability of decitabine to >80% (Lavelle et al. 2012). Calculations with a physiologically-based pharmacokinetic model (Table 8.11) predicted that co-administration of THU with oral decitabine resulted in area under the concentration-time curve (AUC), mean plasma concentration, and intracellular deaza-CTP comparable to levels obtained after IV infusion. Monocyte and macrophage differentiation and activation are known to be epigenetically regulated and loss-of-function mutation of the key epigenetic regulator, TET2, is the commonest genetic abnormality in CMML (Yamazaki et al. 2015; Mason et al. 2016). The CMMLsim model of multiple progenitor cell populations in bone marrow, their proliferation, differentiation, and mutual competition, makes possible prediction of the effects of mutations, singly and in combination. The model predicted that a single replicating bone marrow cell with a TET2 -/- mutation, or a small number of such cells, would eventually become extinct, despite the dysplastic progenitors having the same growth rate as wild-type myeloid progenitor cells (Fig. 8.6). Hypermethylation resulting from the TET2 mutation prevented or disrupted feedback inhibition of monocytic differentiation. The greater rate of differentiation of dysplastic (mutant) progenitors disrupted the balance between their proliferation and release into the circulation, so that they were out-competed by normal myeloid progenitors, and rapidly disappeared. N-ras mutations resulted in approximately two to fourfold increased circulating monocytes and total white cell count, and a small decrease in the count of erythrocytes and platelets. As with epithelial tissues, Ras mutations in the absence of other

174

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Table 8.11 Effect of THU on oral bioavailability of 20 mg dose of decitabine Treatment IV infusion Oral Oral + THU

AUC (μmoles/Lmin) 176

Mean plasma level (μM) 0.120

Peak aza-dCTP (μM) 0.443

Time of peak (min) 182

63 205

0.043 0.139

0.167 0.436

140 177

From Jackson and Radivoyevitch (2021)

mutations cause a nonmalignant hyperplasia, i.e. increased cell proliferation, but no genetic or epigenetic instability. However, a Ras mutation followed by a TET2 mutation resulted in the malignant condition CMML, with a large increase in the circulating monocyte count and total white blood cell count. The model predicted that increased self-renewal of dysplastic progenitors suppressed the normal progenitors (Table 8.5) as reported experimentally (Walenda et al. 2014). The same pair of mutations occurring in the opposite order would be unlikely, since a TET2 mutated clone would, according to the model, become extinct before the second mutation could occur. The cooperative effect of N-ras and TET2 mutations in driving myeloid cell transformation has been described experimentally (Kunimoto et al. 2018), and our model suggests a possible mechanism for this effect. Loss of TET2 activity is clearly a primary causative factor in CMML, so it seems possible that TET2 activators, if such can be identified, might be effective treatments for CMML. Increasing circulating ascorbic acid is a possible approach. The other TET2 substrate, α-ketoglutarate, is probably too rapidly metabolised in the tricarboxylic acid (TCA) cycle for this to be a practical approach, but analogues of α-ketoglutarate that are not substrates for the TCA cycle are worth exploring. Hypomethylating agents are known to normalise (at least partially) the count of circulating blood cells. Can they be used to restore bone marrow homeostasis and prevent or delay disease progression? The clinically approved hypomethylating agents, 5-azacytidine and decitabine decrease DNA methylation following their incorporation into DNA. Both suffer from the disadvantage that they are substrates for cytidine deaminase, particularly if they are administered orally. Co-administration of THU is an established method of increasing the oral bioavailability of decitabine and probably increases its conversion to the active 5′-triphosphate. Our model suggests that cellular aza-dCTP levels remain high for several hours after treatment; however, the important endpoint is DNA incorporation, which we are at present unable to model. Once incorporated, 5-azacytosine will remain in DNA until the cell dies, though it will be progressively diluted out as a result of DNA replication. The cytokinetic/pharmacodynamic models of CML described in Chap. 7 suggested ways in which treatment of that disease might be optimised to delay progression of CML to its terminal blast crisis stage. CMML has both similarities and differences from CML: like CML, CMML is associated with hyperproliferation of myeloid precursor cells, and like CML it sometimes progresses to a secondary

References

175

acute myeloid leukaemia (AML). Modelling CMML and its genetic and epigenetic abnormalities will deepen our understanding of the disease process and may suggest ways to optimise its treatment (Jackson 2012). Is CMML a malignant disease, or a premalignant form of MDS? If we regard genetic instability as the essential hallmark of malignancy, then the 20% of CMML cases that progress to secondary AML are unquestionably malignant. The other 80% die of other morbidities; if this were not the case, would they, too, have progressed to secondary AML? It would be interesting to know if the mutation profile of the 20% that progress is different from that of the 80% that do not. There is an apparent contradiction between the prediction of our model that TET2 loss or haploinsufficiency, in the absence of an additional mutation that provides a growth stimulus, made the mutant progenitors non-viable, and the report that in a mouse model of TET2 loss, the self-renewal capacity of myeloid precursors was increased (Moran-Crusio et al. 2011). This difference may be explained by our simplifying assumption (in the model) that TET2 loss only affects monocytic differentiation. In fact, the effects of TET2 loss are likely to be pleiotropic, and DNA methylation may be a negative regulator of myeloid progenitor self-renewal. Nevertheless, the fact that growth-stimulatory mutations are so common in CMML (N-ras, p53, JAK2) suggests that some cases of human CMML, at least, require an additional growth stimulus for transformation. In summary, we propose that CMML is a three-hit malignancy, requiring minimally (a) a growth-promoting mutation (or possibly loss of epigenetic silencing of a negative growth regulator); (b) increased expression of a normally silenced regulator of monocyte differentiation, often by loss of a single TET2 allele; and (c) loss of the second TET2 allele, resulting in increased methylation, and consequent silencing, of an enhancer region controlling the negative regulation of monocyte differentiation. The age distribution of CMML is strong evidence for the multi-step origin of this leukaemia (Jackson and Radivoyevitch 2021). The model correctly describes many features of CMML, but it does not predict progression to secondary AML, which is seen in about 20% of actual cases. This appears to be consistent with the conclusion that epigenetic changes alone cannot explain malignant progression, and indeed there are no reports of any malignant disease being caused by only epigenetic changes, in the absence of any genetic mutation. Future studies with the model may suggest changes that could account for CMML progression.

References Adamson DJ, Dawson AA, Bennett B et al (1995) p53 mutation in the myelodysplastic syndromes. Br J Haematol 89:61–66 Bartram CR (1992) Molecular genetic aspects of myelodysplastic syndromes. Hematol Oncol Clin North Am 6:557–570 Bernstein C, Nfonsam V, Prasad AR, Bernstein H (2013) Epigenetic field defects in progression to cancer. World J Gastrointest Oncol 5:43–49 Carey N (2011) The epigenetics revolution. Icon Books, London, pp 205–232

176

8

Chronic Myelomonocytic Leukaemia: A Three-Hit Malignancy

Chang S, Yim S, Park H (2019) The cancer driver genes IDH1/2, JARID1C/KDM5C, and UTX/KDM6A: crosstalk between histone demethylation and hypoxic reprogramming in cancer metabolism. Exp Mol Med 51:1–17 Chattergee A, Rodger EJ, Eccles MR (2018) Epigenetic drivers of tumourigenesis and cancer metastasis. Semin Cancer Biol 51:149–159 Cui Y, Tong H, Du X et al (2016) Impact of TET2, SRSF2, ASXL1 and SETBP1 mutations on survival of patients with chronic myelomonocytic leukemia. Exp Hematol Oncol 4:14. https:// doi.org/10.1186/s40164-015-0009-y Davids MS, Steensma DP (2010) The molecular pathogenesis of myelodysplastic syndromes. Cancer Biol Ther 10:309–319 Davidson J (2014) The epigenetic landscape of aneuploidy: constitutional mosaicism leading the way? Epigenomics 6:45–58 DiNardo CD, Stein EM, de Botton S et al (2018) Durable remissions with ivosidenib in IDH-1 mutated relapsed or refractory AML. N Engl J Med 378:2386–2398 Gerds AT (2014) I walk the other line: myelodysplastic/myeloproliferative neoplasm overlap syndromes. Curr Hematol Malig Rep 9:400–408 Greco WR, Park HS, Rustum YM (1990) Application of a new approach for the quantitation of drug synerism to the combination of cis-diammine dichloroplatinum and 1-β-Darabinofuranosylcytosine. Cancer Res 50:5318–5327 Gu X, Tohme R, Tomlinson BJ et al (2021) Decitabine- and 5-azacytidine resistance emerges from adaptive responses of the pyrimidine metabolism network. Leukemia 35:1023–1036 Gur HD, Loghavi S, Garcia-Manero G et al (2018) Chronic myelomonocytic leukemia with fibrosis is a distinct disease subset with myeloproliferatice features and frequent JAK2 V617F mutations. Am J Surg Pathol 42:799-806.u Herman JG, Latif F, Weng Y et al (1994) Silencing of the VHL tumor suppressor gene by DNA methylation in renal carcinoma. Proc Natl Acad Sci U S A 91:9700–9704 Hirsch CM, Nazha A, Kneen K et al (2018) Consequences of mutant TET2 on clonality and subclonal hierarchy. Leukemia 32:1751–1761 Hoeksema MA, de Winther MPJ (2016) Epigenetic regulation of monocyte and macrophage function. Antioxid Redox Signal 25:758–774 Im AP, Sehgal AR, Carroll MP et al (2014) DNMT3A and IDH mutations in acute myeloid leukemia and other myeloid malignancies: associations with prognosis and potential treatment strategies. Leukemia 28:1774–1783 Issa JP (2013) The myelodysplastic syndrome as a prototypical epigenetic disease. Blood 121: 3811–3817 Itzykson R, Kosmider O, Cluzeau T et al (2011) Impact of TET2 mutations on response rate to azacitidine in myelodysplastic syndromes and low blast count acute myeloid leukemias. Leukemia 25:1147–1152 Jacinto FV, Esteller M (2007) Mutator pathways unleashed by epigenetic silencing in human cancer. Mutagenesis 22:247–253 Jackson RC (2012) What can systems pharmacology contribute to drug development? Disease modelling as a predictive tool. BioDiscovery 4:4 Jackson RC, Radivoyevitch T (2021) A pharmacokinetic/pharmacodynamic model of the action of hypomethylating agents in chronic myelomonocytic leukaemia. Med Res Arch 9. https://doi. org/10.1186/s40164-015-0009-y412-420 Jackson RC, DiVeroli GY, Koh SB, Goldlust I, Richards FM, Jodrell DI (2017) Modelling of the cancer cell cycle as a tool for rational drug development: a systems pharmacology approach to cyclotherapy. PLoS Comput Biol 13:e1005529. https://doi.org/10.1371/journal.pcbi.1005529 Jin X, Qin T, Bailey N et al (2018) Oncogenic N-ras and Tet2 haploinsufficiency collaborate to dysregulate hematopoietic stem and progenitor cells. Blood Adv 2:1259–1271 Kalari S, Pfeifer GP (2010) Identification of driver and passenger DNA methylation in cancer by epigenomic analysis. Adv Genet 70:277–308

References

177

Ko M, Huang Y, Jankowska AM et al (2010) Impaired hydroxylation of 5-methylcytosine in myeloid cancers with mutant TET2. Nature 468:839–843 Kunimoto H, Meydan C, Nazir A et al (2018) Cooperative epigenetic remodeling by TET2 loss and NRAS mutation drives myeloid transformation and MEK inhibitor sensitivity. Cancer Cell 33: 44–59 Lahtz C, Pfeifer GP (2011) Epigenetic changes of DNA repair genes in cancer. J Mol Cell Biol 3: 51–58 Lavelle D, Vaitkus K, Ling Y et al (2012) Effects of tetrahydrouridine on pharmacokinetics and pharmacodynamics of oral decitabine. Blood 119:1240–1247 Lin I-H, Chen D-T, Chang YF et al (2015) Hierarchical clustering of breast cancer methylomes revealed differentially methylated and expressed breast cancer genes. PLoS One 10:e0118453. https://doi.org/10.1371/journal.pone.0118453 Mason CC, Khorashad JS, Tantravahi SK et al (2016) Age-related mutations and chronic myelomonocytic leukemia. Leukemia 30:906–913 Metzeler KH, Herold T, Rothenberg-Thurley M et al (2016) Spectrum and prognostic relevance of driver gene mutations in acute myeloid leukemia. Blood 128:686–698 Molenaar RJ, Radivoyevitch T, Maciejewski JP et al (2014) The driver and passenger effects of isocitrate dehydrogenase 1 and 2 mutations in oncogenesis and survival prolongation. Biochim Biophys Acta 1846:326–341 Molenaar RJ, Radivoyevitch T, Nagata Y et al (2018) IDH1/2 mutations sensitize acute myeloid leukemia to PARP inhibition and this is reversed by IDH1/2-mutant inhibitors. Clin Cancer Res 24:1705–1715 Moolgavkar SH, Knudson AG (1981) Mutation and cancer: a model for human carcinogenesis. J Natl Cancer Inst 66:1037–1052 Moran-Crusio K, Reavie L, Shih A et al (2011) Tet2 loss leads to increased hematopoietic stem cell self-renewal and myeloid transformation. Cancer Cell 20:11–24 Patel BJ, Przychodzen B, Thota S et al (2017) Genomic determinants of chronic myelomonocytic leukemia. Leukemia 31:2815–2823 Phillips RA (1987) The genetic basis of cancer. In: Tannock IF, Hill RP (eds) The basic science of oncology. Pergamon Press, Oxford, pp 26–27 Sawan C, Vaissière T, Murr M, Herceg Z (2008) Epigenetic drivers and genetic passengers on the road to cancer. Mutat Res 642:1–13 Tan Y, Liu H, Chen S (2015) Mutant DNA methylation regulators endow hematopoietic stem cells with the preleukemic stem cell property, a requisite of leukemia initiation and relapse. Front Med 9:412–420 Tanas AS, Sigin VO, Kalinkin AI (2019) Genome-wide methylotyping resolves breast cancer epigenetic heterogeneity and suggests novel therapeutic perspectives. Epigenomics 11:605–617 Usui G, Matsusaka K, Mano Y et al (2021) DNA methylation and genetic aberrations in gastric cancer. Digestion 102:25–32 Walenda T, Stiehl T, Braun H et al (2014) Feedback signals in myelodysplastic syndromes: increased self-renewal of the malignant clone suppresses normal hematopoiesis. PLoS Comput Biol 10:e1003599 Wang C, Zhao N, Yuan L, Liu X (2020) Computational detection of breast cancer invasiveness with DNA methylation biomarkers. Cell 9:326. https://doi.org/10.3390/cells9020326 Wodarz D, Komarova NL (2014) Dynamics of cancer: mathematical foundations of oncology. World Scientific, Singapore, pp 147–168. ISBN 978-981-4566-36-0 Yamazaki J, Jelinek J, Cesaroni M et al (2015) TET2 mutations affect non-CpG Island DNA methylation at enhancers and transcription factor-binding sites in chronic myelomonocytic leukemia. Cancer Res 75:2833–2843 Zahir N, Sun R, Gallahan D et al (2020) Characterizing the ecological and evolutionary dynamics of cancer. Nat Genet 52:759–767

Chapter 9

The Cancer Stem Cell and Tumour Progression

Abstract Tumour stem cells result from genetic or chromosomal changes in normal tissue stem cells, i.e. cells with self-renewal capability. Subsequent tumour progression is a consequence of tumour cell heterogeneity, which in turn is a result of genetic or chromosomal instability. In a heterogeneous tumour mass, Darwinian natural selection will result in preferential survival and reproduction of those cells with faster doubling times or loss of cell death pathways. Environmental factors also contribute to selective pressure. Aspects of tumour progression include de-differentiation, invasiveness, metastasis, metabolic changes that result in tolerance of hypoxia, neoangiogenesis, and insensitivity to immune checkpoints. Selective pressure in a growing tumour may result from limited availability of nutrients or oxygen, ability to survive the lack of cell contact signalling, or (in a treated tumour) presence of antitumour drugs. Evolutionary dynamics calculations suggest that the Warburg effect (aerobic glycolysis) confers a selective advantage on metastatic cells by downregulation of the intrinsic pathway of apoptosis. At least three mutations are required for a tumour stem cell to progress to a metastatic tumour.

Otto Warburg was the most remarkable person I have ever been closely associated with. Remarkable as a scientific genius of the highest calibre, as a highly independent, penetrating thinker, as an eccentric who shaped his life with determination and without fear, according to his own ideas and ideals. Hans Krebs, Reminiscences and Reflections (1981)

9.1

Tumour Progression as a Process of Natural Selection

Evolutionary dynamics calculations suggest that tumour development requires multiple genetic changes. The critical event, which may be a mutation or a chromosomal rearrangement, greatly increases the probability of further mutations. In almost all

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_9. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_9

179

180

9 The Cancer Stem Cell and Tumour Progression

cases, this transforming event must be accompanied or preceded by a mutation that stimulates growth, because chromosomal rearrangements tend to increase spontaneous cell death. This two-stage transformation process results in a tumour stem cell. The subsequent events that result in a tumour stem cell becoming a clinical cancer are termed tumour progression. The essential features of tumour progression can be explained by considering it as an example, at the cellular level, of Darwinian selection. The philosopher Daniel Dennett (1995) emphasised that Darwinism is more than a theory of the origin of species, it is an algorithm. This algorithm, the genetic algorithm (Holland 1992) has been applied to a wide range of problems. What they have in common is that all describe processes by which complex systems can arise from a set of simple rules. The original version of the genetic algorithm generated diversity by simulating the process of “crossing over” (sister chromatid exchange) during meiosis. In fact, a number of processes can contribute to genetic change between generations, such as germ-line mutations and chromosomal rearrangement. Darwinism has been glibly, but erroneously, described as a tautology: Darwin “explained” evolution as “survival of the fittest” where by “fittest” he meant those most likely to survive. The logical fallacy in this criticism is that it focuses on just one of the five elements that comprise the genetic algorithm. It sets up one of the elements of Darwinism, as a straw man, to represent the whole, and then claims that Darwin’s conclusion does not follow from his premise. It is worth describing in detail what the essential features of the genetic algorithm are. First, Darwin postulated a struggle for survival—in modern terms, the tendency of a population to grow beyond the carrying capacity of its environment, thus setting up competition for reproductive resources. This concept was not original to Darwin, it had been prefigured by Malthus, for example, but Malthus, like the “tautology” critics, emphasised just this one of the five elements required for evolution by natural selection. The second essential element Darwin described as descent with modification—in modern terminology, genetic (and thus phenotypic) diversity. In the pre-Mendelian era, Darwin was unable to suggest what the origin of this diversity might be, but the genetic algorithm is not concerned with mechanism. Other nineteenth century biologists, most famously Lamarck, had realised that diversity contributed in some way to the evolutionary process, but again this is just one of the essential elements. Darwin’s creative leap was grasping that the third element, “survival of the fittest” emerged as an inevitable consequence of the first two elements. The fourth essential element of the genetic algorithm is that it is an iterative process. In a single generation the change in composition of the population may be small, but when the process is iterated over many generations, a major population shift can result. The iterative nature of the genetic algorithm is a feature of all algorithms that generate complexity from simple rules (the Mandelbrot algorithm is another example of this). Finally, the genetic algorithm provides an element of autocatalysis—with each succeeding iteration (generation) the selected group forms a larger fraction of the total population. It is interesting to compare the genetic algorithm with the closely related Metropolis algorithm, in which the objective is to modify a set of input variables that

9.2

Origins of Cancer Stem Cells

181

generate an approximation to some desired endpoint so that a closer approximation is achieved. For example, the starting inputs could be a set of weak inhibitors of a particular enzyme. In the Metropolis algorithm, the ability of the inhibitors to bind to the target is calculated by a defined objective function (e.g. a free energy calculation) and a set of rules for generating diversity is applied. To prevent the system from converging on a local minimum, further diversity is injected into the system by a process of simulated annealing, essentially introducing large random perturbations that move the system far from its local minimum. Unlike the Metropolis algorithm, the genetic algorithm (at least in the original form conceived by Darwin) does not have an explicit objective function. By selecting for “fitness”, i.e. adaptation to a particular environment, the genetic algorithm is in fact selecting by an implicit objective function. Transformation does not necessarily involve rapid cell division. A minimal deviation hepatoma may have a growth rate much less than that of normal epithelial tissues. Uncontrolled cell proliferation is thus not, per se, an aspect of tumour formation. As demonstrated by the early studies with minimum deviation hepatomas, the requirement for transformation is release of cell proliferation from its physiological constraints. That is one of the two requirements for the transformation process. However, once that, and the other requirement for transformation—genetic/ chromosomal instability—is satisfied, the subsequent process of Darwinian selection will, of course, favour rapidly proliferating variants. Thus, an increased proliferation rate is an aspect of tumour progression. The selective advantage can be very small (Bozic et al. 2010). Two mutations or many? Once the two essentials for transformation have occurred, progression is inevitable. Given the cellular heterogeneity that is a consequence of chromosomal instability, the conditions for Darwinian selection have been met (descent with modification, and competition for resources). The increase in mutation rates is not essential for progression, but accelerates the process. Darwinian selection involves the interaction of variants with their environment: in the case of tumour progression, such factors as regional differences in blood flow can influence whether or not particular mutations will be selected (Lloyd et al. 2016).

9.2

Origins of Cancer Stem Cells

Frank (2007) discusses the kinetics of normal tissue stem cells, focussing on bone marrow, skin, and intestinal epithelium. Stem cells form a very small fraction of the total cell population and divide occasionally to give one stem cell and one progenitor cell (“transit cell” in epithelial tissues). The progenitor cells undergo multiple divisions, to produce differentiated cells. Since all the progeny of progenitor/transit cells are destined to be sloughed (into the blood, the intestinal lumen, or the bathwater) mutations in these populations are unlikely to produce a tumour. The separation of stem cells and progenitor cells thus minimises the lifetime risk of oncogenic mutations. Haematopoietic stem cells in mice replicate about every

182

9 The Cancer Stem Cell and Tumour Progression

14 days. Gastrointestinal stem cells replicate approximately every 24 hours (Bach et al. 2000). Study of intestinal cell dynamics has been facilitated by the discovery of the Lgr5 marker (Barker et al. 2007). Each intestinal crypt contains 4–7 stem cells (Clevers et al. 2014). In small intestinal crypts, each stem cell is in close proximity to an underlying differentiated cell, a Paneth cell. Paneth cells secrete antibacterial peptides, suggesting that they serve a protective function. When a small intestinal stem cell divides, the daughter cell that remains in contact with the underlying Paneth cell remains a stem cell, and the other daughter cell, which is pushed “up” the crypt (towards the intestinal lumen) is now a mortal transit cell. Those cells that do not remain stem cells differentiate to one of six different mature cell types and migrate to the villi where they are shed into the intestinal lumen after 3–5 days (van der Flier and Clevers 2009; Gehart and Clevers 2019). Colonic crypts do not contain Paneth cells. Differentiation of the transit cells is controlled by Wnt signalling (Clevers et al. 2014). What is a cancer stem cell? Do cancer stem cells arise from normal tissue stem cells? In normal epithelial tissues, only a small fraction of the cell population, the somatic stem cells, possesses both the property of self-renewal and the ability to differentiate into the various cell types in that tissue. It is argued that the most probable route to cancer is mutation of the somatic stem cells to become cancer stem cells (Martinez-Climent et al. 2006). Can normal tissue progenitor cells (which lack self-renewal capability) give rise to cancer stem cells? Presumably they are subject to the same mutations and chromosomal rearrangements as stem cells, but because of their rapid turnover the chances of them transforming in their short lifetime are small. In Chap. 3 it was argued that a cancer stem cell was a cell with self-renewal capacity that had undergone two changes that caused over-ride of cell cycle checkpoints: inactivation or over-ride of the G1:S checkpoint meant that cell division would occur regardless of whether there was “demand” for it. Loss of the SAC or its over-ride by components of the DDR resulted in genetic instability. Genetic or chromosomal instability resulted in cellular heterogeneity that set in train a process of Darwinian selection. Advanced tumours express many more changes than the two essentials, but these are an inevitable result of the selection process. Tumour progression is the phenotypic expression of the series of these changes. Many tumours harbour cancer stem cells in dedicated niches (Batlle and Clevers 2017). Are tumour stem cells always derived from normal tissue stem cells? In particular, are myeloid leukaemias derived from bone marrow stem cells, or can they arise from partially differentiated myeloid progenitor cells? All cells of the body contain the same genetic information, so any replicating cell has the potential to undergo oncogenic mutations. In epithelial tissues, the small fraction of stem cells, cells with self-renewal capability, are much more likely to transform than more differentiated cells. There is also a positional component to the probability of tumour formation. As the work of Clevers’ group has shown, when a stem cell in an intestinal crypt replicates, it produces one daughter cell that remains in the position of the parent cell and one more superficially positioned daughter cell. This second daughter cell is able to undergo further division, but ultimately all its progeny are

9.2

Origins of Cancer Stem Cells

183

destined to be sloughed into the intestinal lumen. Oncogenic mutations in this group of cells are possible, but much less likely to produce an established tumour than the same mutations in an intestinal stem cell. Tissue architecture not only influences cellular dynamics, but also response to drugs. Given that (in the mouse) an intestinal crypt has seven stem cells, if that mouse is treated with an anticancer drug that gives one log kill of normal cells, after treatment most of the crypts will have no surviving stem cells, and the mouse will die from intestinal toxicity. Probably there will be some crypts that have one or more surviving stem cells, which could in principle repopulate those crypts—but not the other crypts. In contrast, bone marrow stem cells are more mobile, and as we know from bone marrow transplantation, myeloid stem cells in one part of the body can repopulate bone marrow at distant sites. The suggestion that tumour cells are potentially immortal is based upon the laboratory observation that tumour cells in culture are able to undergo hundreds of successive cell divisions. In contrast, the process of cellular senescence was first explored in human embryo fibroblasts by Hayflick, who observed that the normal cell was capable of about sixty cell divisions, before becoming incapable of further division (Shay and Wright 2000). Cellular senescence was shown to be caused by the fact that every time a chromosome divides, it loses a few bases from structures at each end of the chromosomes, the telomeres (reviewed by Carey 2011). The DNA repair systems of the cell interpret loose ends in DNA as a sign of DNA damage, and the function of the telomere is to prevent the repair enzymes from attempting to modify the chromosome ends. Telomere length acts as a molecular clock, and when the telomere length falls below a critical value the cell becomes senescent, and no longer divides. An exception to this is the reproductive germ cells, which contain an enzyme, telomerase, that maintains telomere length. Germ cells are potentially immortal, as are the stem cells of rapidly dividing tissues, such as bone marrow and intestinal epithelium, and indeed stem cells also contain telomerase. Do tumour cells have to express telomerase? Cells without telomerase have the potential for about 60 cell divisions. This is more than enough for a single cell to grow to a tumour of lethal size. Of course, a cell that was transformed late in life would have fewer of its divisions remaining, so turning on telomerase might then confer a selective advantage. If cancer stem cells are derived from normal tissue stem cells, they already express telomerase. Rather than regarding immortalisation as part of tumour progression, it is perhaps more accurate to regard telomerase expression as a prerequisite for transformation that must be maintained during the progression process.

184

9.3

9

The Cancer Stem Cell and Tumour Progression

Driver and Passenger Mutations in Tumour Progression

The distinction between “driver” and “passenger” mutations was noted in Chap. 2: both tend to increase in number during tumour progression. Driver mutations are those that confer a selective advantage, either by increasing the growth rate, or by decreasing apoptosis, or by making the cell better adapted to its microenvironment— e.g., hypoxia-tolerant. A 2018 review of the Cancer Genome Atlas listed 299 driver mutations (Iranzo et al. 2018). Passenger mutations are what evolutionary theorists term “neutral mutations”, that is they cause neither a selective advantage nor disadvantage. Early biochemical studies, e.g. in chemically-induced rat hepatomas, noted that some changes from normal liver expression appeared to vary randomly between tumours, rather than correlating with transformation or growth rate. Mutations that cause a selective disadvantage may be transiently present, but will be eliminated by natural selection. The distinction between drivers and passengers may be somewhat elastic, because a mutation that causes an advantage in certain environmental conditions may cease to be advantageous if the microenvironment changes, for example, because of new blood vessel growth. Vermeulen et al. (2013) quantified the competitive advantage of Apc loss, Kras activation, and p53 mutations in mouse intestine. P53 mutations displayed a condition-dependent advantage, for example p53 mutations were especially favoured in colitis-affected intestine. Their work supported the previously theoretical idea that the tissue architecture of the intestine suppressed the accumulation of mutations.

9.4

De-differentiation

Early histopathological descriptions of tumour progression described progressive loss of differentiated characteristics. Early-stage tumours often retain much of the structural and functional character of their tissue of origin, though this tends to decrease with time. Early biochemical studies supported the histological observations, for example well-differentiated Morris hepatomas expressed enzymes of gluconeogenesis (a differentiated property of normal liver), but more rapidly growing hepatomas progressively lost this activity. The evolutionary explanation of de-differentiation is that expression of differentiated activity comes at a cost—in ATP, in precursors for synthesis of macromolecules, and so confers a selective disadvantage. As a result, advanced tumours tend to converge to a common appearance and protein expression pattern (the full-blown cancer phenotype). A late-stage lung tumour and a late-stage colon tumour resemble each other, histologically and biochemically, more than they resemble normal lung or colonic epithelium, respectively.

9.5

Angiogenesis as an Aspect of Tumour Progression

185

An aspect of de-differentiation described by the early histopathologists was the epithelial-mesenchymal transition (discussed in more detail below) in which an early-stage epithelial tumour, growing, in its site of origin, as a flat sheet of cells on a basement membrane, transforms into a solid lump, invading the surrounding tissues (Shin et al. 2010). This behaviour is caused by loss of contact dependence as a result of constitutive signalling of adhesion pathways. Normal differentiation involves Wnt signalling, and in mature tissues Wnt signalling is turned off. It is re-activated in most colon cancers, and this reactivation is usually associated with loss of the tumour suppressor, Apc, which is a negative regulator (Fig. 1.3). Restoration of Apc to transgenic mice with colon tumours restored normal crypt differentiation (Dow et al. 2015).

9.5

Angiogenesis as an Aspect of Tumour Progression

Once a tumour has become independent of extracellular matrix, it grows as a solid lump, instead of as a flat sheet. When this early solid tumour reaches a diameter of more than about 1 mm, the interior no longer receives enough oxygen to sustain cell division (since without a blood supply, the only source of oxygen is by passive diffusion from the extracellular fluid), and the hypoxic core of the tumour becomes quiescent, although most of the cells are still viable and can resume cell division if the oxygen supply is restored. Cells in the centre of the tumour may not receive enough oxygen to sustain life. This necrotic core is irreversibly damaged. Necrotic tumour cells are harmless, but viable hypoxic cells are dangerous. Hypoxic cells are resistant to X-rays and to some cytotoxic drugs. Other drugs are potentially active against hypoxic cells, but have difficulty reaching the cells because of the lack of a blood supply. In 1971, Folkman suggested that solid tumours could only grow beyond a size of 1–2 mm once they had established a blood supply (Folkman 2003). Hypoxia stimulates hypoxia-dependent transcription factors, HIF-1α and HIF-1β. The primary function of these is to activate transcription of proteins required for new blood vessel formation, angiogenesis. One of the HIF-1α transcripts is vascular endothelial growth factor (VEGF), which stimulates outgrowth of endothelial cells from existing blood vessels towards regions of hypoxia (Ramjiawan et al. 2017). Other growth factors such as fibroblast growth factor (FGF) and PDGF also stimulate angiogenesis. After a growing solid tumour has an established blood supply and becomes metastatic, the resulting micro-metastases, in turn, will only grow once they have developed their own blood supply. Tumour angiogenesis is thus an aspect of tumour progression for both primary and secondary tumours. Most normal tissues in mature organisms produce natural inhibitors of angiogenesis, such as angiostatin. Whether new blood vessels develop in a tissue depends upon the balance between pro- and anti-angiogenic factors. In general, in early development the pro-angiogenic factors will predominate, and in mature tissues the antiangiogenic factors predominate, except where tissue re-modelling occurs, as in wound healing. However, an early-stage tumour will generally be in a

186

9 The Cancer Stem Cell and Tumour Progression

predominantly antiangiogenic environment, and this will provide a natural limit to its growth (Wodarz and Komarova 2014). The tumour vasculature produced by the neoangiogenesis process is abnormal in a number of respects (Viallard and Larrivée 2017). The new vessels are unusually permeable and highly branched. This results in leakage of fluid into the tumour, causing increased hydrostatic pressure, which then limits blood flow and oxygenation. Angiogenesis inhibitors were evaluated as antitumour agents by Folkman and colleagues, who showed that corticosteroids had antiangiogenic activity. Corticosteroids could inhibit the growth of transplanted experimental tumours and metastatic growths, but they did not kill the tumours, and in fact did not even block tumour growth completely, because cells on the surface of tumours were still able to get oxygen and nutrients by diffusion from extracellular fluid. Antiangiogenic agents increase the quiescent cell fraction of tumours, so antagonise the cytotoxic effect of cell cycle-specific drugs, and cause drug-delivery problems for all classes of drugs. A more recent class of antiangiogenic agents are inhibitors of VEGF receptor tyrosine kinase, e.g. sunitinib. VEGFR is a family of three receptors of which VEGFR2 is expressed primarily in the endothelial cells that form the lining of blood vessels. Sunitinib is used in the treatment of renal cell carcinoma and gastric tumours. Note that this class of inhibitors block the proliferation of endothelial cells, not tumour cells. Sorafenib, though primarily a Raf kinase inhibitor, also inhibits VEGF-R kinases. Bevacizumab is a monoclonal antibody that binds to the growth factor VEGF, preventing its binding to, and activation of VEGFR. All these agents can cause adverse reactions related to inhibition of VEGF signalling in the endothelial cells of blood vessels in normal tissues. Unlike antiangiogenic drugs, which inhibit new blood vessel formation without affecting existing vasculature, vascular targeting agents (also known as vascular disrupting agents) damage existing blood vessels in tumours, causing extensive haemorrhagic necrosis. An example of this class of agent is flavone acetic acid (Mita et al. 2013). This class of drugs can kill the hypoxic, non-cycling cells in the core of large tumours through necrosis, rather than apoptosis—an advantage, because many tumours are resistant to apoptosis. Like antiangiogenic drugs, vascular targeting agents do not affect the outer proliferating rim of tumours, so they must be used in combination with conventional chemotherapy. Although angiogenesis is essential for sustained tumour growth, the clinical activity of antiangiogenic drugs, used as single agents, has been disappointing, perhaps because they cannot prevent continued cell division of the outer layer of tumours (Li et al. 2018). This necessitates that antiangiogenic agents be used in combination with other drugs, but, by disrupting tumour blood supply, the efficacy of these combining agents is limited. Mathematical models of angiogenesis have been developed (Chaplain and Anderson 1996) and used to suggest optimal scheduling of antiangiogenics with cytotoxic agents (Benzekry et al. 2012; Finley et al. 2014; Jackson et al. 2015). The Physiomics (2022) Virtual Tumour allows for modelling antiangiogenic drugs in combination with cytotoxics or targeted agents,

9.6

The Warburg Effect

187

Fig. 9.1 Inhibition of tumour blood vessel formation by the antiangiogenic agent, cortexolone, (d.4-18) modelled by the SKIPPER program

Use of cytotoxic agents first is predicted to give maximal cell kill, and antiangiogenic treatment should ideally be sustained for several days. The SKIPPER model, described in the online supplement to Chap. 6, can be used to model treatment with antiangiogenic agents. Figure 9.1 shows an example of modelling the effect of the antiangiogenic agent, cortexolone, in a mouse. This rapidly growing tumour reached a lethal size in 13 days, Treatment with cortexolone on days 4 to 18 prolonged survival to 20 days, Fig. 9.1 shows the effect of treatment on tumour blood vessel formation. During the treatment period, the rate of increase in tumour blood vessels was inhibited >90%. After treatment was discontinued there was a rapid rebound in tumour blood vessel numbers, partly due to resumption of proliferation of tumour endothelial cells, and partly to infiltration from surrounding tissues. Normal tissues in the mouse were unaffected by the treatment.

9.6

The Warburg Effect

In the 1920s, before the biochemical pathways of energy metabolism had been mapped in detail, it was already apparent that there were two ways in which eukaryotic cells could convert carbohydrates into energy: oxidative metabolism, carried out in mitochondria, and the more evolutionarily ancient glycolytic pathway, in the cytosol. Glycolysis (“sugar breaking”) is a process closely related to fermentation in yeast. Glycolysis produces (net) 2 ATP per glucose consumed, compared with an (additional) 34 ATP produced by oxidative phosphorylation. Warburg reported that tumour tissue contained fewer mitochondria per cell than normal tissues, and those mitochondria were structurally damaged. In 1930 Warburg reported that tumour tissue slices demonstrated aerobic glycolysis, i.e. they continued to use the glycolytic pathway as their primary energy source even in the presence

188

9 The Cancer Stem Cell and Tumour Progression

of oxygen. This appeared to be a qualitative difference from normal tissues. Exercised muscle, for example, generates the ATP required for contraction primarily by oxidative phosphorylation. When the demand for ATP outstrips the oxygen supply, glycolysis becomes the major energy provider, and lactic acid accumulates. This is normal, anaerobic glycolysis. When the exercised muscle is allowed to rest, oxidative metabolism resumes, and the lactic acid is cleared. Warburg suggested that the essential characteristic of malignancy was a reversion of cells to the evolutionarily ancient pattern of a world before the existence of photosynthetic algae and green plants, a world without oxygen. In support of this, he claimed that when healthy cells were starved of oxygen for several hours, they became cancerous. This claim has not been repeated using modern criteria of “cancerous” and it seems likely that what he was seeing was structural damage to oxygen-starved mitochondria. However, nearly a century on, Warburg’s central observation, that tumours have a change in energy metabolism that enables them to survive and proliferate in the absence of oxygen, and that this pattern of metabolism persists in presence of oxygen, has held up well, and appears to apply to all cancers. What is not universally accepted is Warburg’s claim that this is the universal cause of cancer, as distinct from an aspect of the tumour progression process. If we accept the modern interpretation, that aerobic glycolysis persists in tumours because it confers a survival advantage, we must explain what that advantage is. Bustamante and Pederson (1980) showed that hexokinase II was elevated in all tumours they examined. Unlike the predominant hexokinase I, hexokinase II is not subject to feedback inhibition by its product, glucose-6-phosphate. Hexokinase II is associated with the voltage-dependent anion channel (VDAC) of mitochondrial membranes. VDAC normally allows mitochondrial cytochrome c to enter the cytosol, where it triggers apoptosis. When hexokinase II is bound to VDAC, cytochrome c entry into cytosol (and thus apoptosis) is blocked. This decreased apoptosis could provide a selective growth advantage for tumour cells. Other interactions between tumour progression and glycolysis are known: for example, c-Myc activates transcription of genes involved in aerobic glycolysis. Activated ERK2 (part of the MAPK pathway) phosphorylates pyruvate kinase M2 which translocates to the nucleus and promotes glycolysis (Yang et al. 2012). The tumour suppressor, PTEN, lost in many tumours, is reported to regulate the balance of energy metabolism in favour of oxidative phosphorylation, so that its deletion would favour glycolysis (Garcia-Cao et al. 2017). An area of continuing investigation is how aerobic metabolism relates to the other biochemical attributes of malignancy, regardless of whether it is the primary cause of transformation or a late downstream effect. The current view is that the reprogramming of energy metabolism in tumours is part of the tumour progression process. Following the loss of growth dependence on basement membrane, what had been a monolayer of transformed cells now becomes a three-dimensional lump. The resulting increased distance from blood capillaries results in hypoxia. Hypoxia activates the transcription factor HIF1α, which stimulates new blood vessel formation—neoangiogenesis, as discussed above. HIF1α and HIF-1β also activate the glucose transporter GLUT-1.

9.6

The Warburg Effect

189

Epstein et al. (2014) argued that aerobic glycolysis is not necessarily a pathological phenomenon. Glycolysis responds more rapidly than oxidative phosphorylation to sudden changes in energy demand, so that aerobic glycolysis can be an effective adaptation to short-timescale events, such as membrane transport activity. Adaptive responses of carbohydrate metabolism often become permanent as a result of epigenetic changes—or if not permanent, very long-lasting. Gatenby and Gillies (2004) proposed that the Warburg effect is driven by tumour hypoxia (possibly transient) causing an increased AMP:ATP ratio which upregulates AMP-dependent protein kinase (AMP-K). This is a permanent response to transient hypoxia. It has been claimed that dietary restriction (which increases AMP-K activity) can inhibit tumour progression. AMP-K signalling activates uptake and metabolism of glucose and fatty acids. Inhibition of tumour progression by the antidiabetic drug, metformin, which increases activity of AMP-K, has been described in experimental systems, and patients with type 2 diabetes, who are often treated with metformin, have a lower incidence of cancer than the general population. It is possible that the antitumour effect of metformin against prostate cancer operates by a different, or additional mechanism, as metformin has been shown to downregulate androgen receptors (Wang et al. 2015). Another consistent metabolic change in tumour cells is the appearance of an alternatively spliced variant of pyruvate kinase M, termed PK-M2 (Christofk et al. 2008). Expression of the PK-M2 isoform is necessary for aerobic glycolysis. ERK2 (a member of the EGF-driven MAP kinase pathway) phosphorylates PK-M2, which then migrates to the nucleus and stimulates transcription of enzymes of the glycolytic pathway, including PK-M2 itself (Hitosugi et al. 2009; Yang et al. 2012; DeNicola and Cantley (2015). This occurs in normal replicating cells, as well as in tumour cells, and it is suggested that this is a way of diverting cellular resources to production of metabolic intermediates required for cell division, rather than energy production (Vander Heiden et al. 2009). The autocatalytic activation by PK-M2 of its own transcription is indicative of a switching mechanism. It suggests that there are two stable states of energy generation: a “housekeeping” state, largely used by non-proliferating cells, and driven by oxidative phosphorylation, and a “metabolitegenerating state”, in which glycolysis provides both ATP (albeit in smaller amounts) and metabolites required for cell division. Normal cells, that replicate occasionally as needed, are able to switch between these states. Malignant cells, it is suggested, are permanently dependent upon the high-glycolysis, replicating state. It would be interesting to know if tissues with benign hyperplasia, in which MAP-K signalling is activated, exhibit aerobic glycolysis. Specific activators of PK-M2 have been evaluated as anticancer agents (Boxer et al. 2010), and if the hypothesis of Vander Heiden et al. (2009) is correct, these compounds could also inhibit normal cell proliferation. It is not clear whether using glycolysis as the primary energy source confers a selective advantage in non-hypoxic conditions, after tumour neoangiogenesis has occurred. It is likely that the level of ROS in cells using glycolysis rather than oxidative phosphorylation may be lower, but there are no data to support this.

190

9

The Cancer Stem Cell and Tumour Progression

Evolutionary dynamics suggests an alternative explanation for the persistence of tumour cell glycolysis under aerobic conditions. Epithelial cells, to function appropriately, must stay in their proper place, their organ or tissue of origin. In this respect they differ from haematopoietic cells, or cells of the immune system, whose function requires them to have access to all parts of the body. Epithelial cells that stray outside their proper location are programmed to self-destruct by apoptosis. This apoptosis, by the intrinsic pathway, requires cytochrome c (Fig. 1.13). Cytochrome c is normally confined to the mitochondria, and its release into cytosol activates apoptosis. If cells contain fewer mitochondria, their capacity for intrinsic apoptosis will be lower, so wandering epithelial cells that rely upon glycolysis as their primary energy source will have a selective advantage. The hypoxic reprogramming of cancer cell energy metabolism has been shown to be epigenetically regulated (Chang et al. 2019). In summary, it is well established that all advanced tumour cells demonstrate aerobic glycolysis and decreased mitochondrial activity, and there are plausible hypotheses to explain why selective pressure should favour such cells. The controversial question is whether the Warburg effect is a primary cause of malignant transformation (Christopherson 2017), or whether it is an effect of malignant progression. Do cancer stem cells show a Warburg effect? The evidence is equivocal: the suggestion that cancer is a reversion to the evolutionarily primitive form of energy that prevailed before the evolution of mitochondria cannot be correct, since cancer cells that totally lack mitochondria have never been reported. Indeed, many tumour cells use glutamine, rather than glucose, as a primary energy source, and glutamine (after deamination to glutamate) is metabolised in the tricarboxylic acid cycle, an exclusively mitochondrial pathway (Smallbone et al. 2007; Dando et al. 2015). Cancer stem cells isolated from an ovarian carcinoma were heterogeneous with respect to energy metabolism, with some relying primarily on glycolysis, and others on mitochondria (Sancho et al. 2016; Sato et al. 2016; Danhier et al. 2017). Stem cells derived from a human osteosarcoma cell line did not exhibit a Warburg effect. (Koka et al. 2018). In glioblastoma, quiescent stem cells relied primarily on oxidative phosphorylation, and more differentiated cells had increased cytoplasmic glycolysis (Iranmanesh et al. 2021). The general consensus from the stem cell literature is that the Warburg effect is not a primary event in cell transformation, but occurs during the malignant progression process, and whether it is then an early or late event depends upon the particular microenvironment (Sun et al. 2018; Damaghi et al. 2021). It is possible that the occasional observations of aerobic glycolysis in tumour stem cells are an artefact resulting from their extraction in high-glucose tissue culture medium. Early or late, cancer cells clearly have profound changes in their energy metabolism; this affects pathways other than glycolysis, for example, Carracedo et al. (2013) reported increased fatty acid metabolism in cancer cells. Warburg, in fact, observed two related but distinct effects: aerobic glycolysis, which may be proliferation- rather than transformation-linked, and decreased mitochondrial activity (and numbers), which is selected as a later effect in malignant progression because fewer mitochondria result in decreased spontaneous apoptosis.

9.7

The Epithelial-Mesenchymal Transition

191

Tumour hypoxia has consequences for cancer treatment. Hypoxic cells are radiation-resistant, since radiation-induced cell killing is oxygen-dependent. Also, hypoxia makes cells resistant to the cytotoxic effects of many anticancer drugs, by increasing the proportion of non-proliferating cells in the tumour.

9.7

The Epithelial-Mesenchymal Transition

Histopathological examination of very early-stage carcinomas showed that tumours undergo a transition from an epithelial appearance to a more solid (mesenchymal) appearance. This change, the epithelial-mesenchymal transition (EMT) is a normal process in tissue repair and wound healing; when tissue repair is complete, cells undergo the opposite transition and revert to an epithelial state. The term “carcinoma in situ” is sometimes used to describe epithelial tumours that grow as a flat sheet without invading the underlying tissues. There is debate as to whether carcinoma in situ is malignant or premalignant, and this clearly depends upon the definition of malignancy. The definition used in the present work is based upon karyotype: a lesion that has abnormal growth, but remains diploid, is premalignant in the sense that a further mutation may make it malignant, but the mutation rate is not raised above its background level. Clinically, such a carcinoma in situ may safely be left alone, subject to follow-up examinations to check that it has not progressed. A lesion with an aneuploid karyotype, however, must be considered malignant, because it is genetically unstable. While it may, as yet, be non-invasive, its mutation rate is increased to such an extent that further progression is inevitable, and the next step in the progression process is likely to be invasion of the surrounding normal tissues. An aneuploid carcinoma in situ should be removed by surgery or radiation. The development and maintenance of normal tissues is controlled by multiple families of adhesion molecules, surface receptors that enable cells to attach to each other, or to a basement membrane. These adhesion molecules ensure that the various organs and tissues maintain their shape and correct position in the body. Which particular adhesion molecules are expressed in a cell varies from tissue to tissue, and most cells express multiple adhesion molecules. The integrins are a widely expressed family of adhesion molecules whose extracellular domain is involved in binding of cells to extracellular matrix, and whose intracellular domain signals to the nucleus whether or not the cell is attached. Integrins are thus involved both in a structural capacity and in information processing. In Chap. 1, an example was given of cells that, in order to undergo cell division, require both a proliferation signal (from a growth factor) and an attachment signal (from an integrin). In the example discussed in Chap. 1, the kinase that requires the dual signal is FAK (focal adhesion kinase). Mutations in FAK that result in it signalling in the absence of an attachment signal may result in the cell becoming invasive. Cells that would normally grow as a flat sheet on a basement membrane can now grow as a solid lump. The p21-activated kinases (PAKs) are a family of kinases that, when overexpressed, increase anchorage-independent growth and cell migration, by regulation of actin

192

9 The Cancer Stem Cell and Tumour Progression

cytoskeletal dynamics (Wells and Jones 2010; Li et al. 2010, 2021). PAK-4 inhibitors have been shown to inhibit invasion through modification of integrin signalling (Eswaran et al. 2009; Li et al. 2020, 2021). Sakai et al. (2018) showed that Apc mutations were sufficient to cause intestinal adenomas. Combination with p53 mutation or deletion of Tgfbr2 (the gene for TGFβ receptor 2) resulted in submucosal invasion. The addition of Kras mutations then gave EMT-like morphology. The combination of ApcΔ716 with KrasG12D and Fbxw7 mutation was insufficient for invasion, but still induced EMT-like histology. Fbxw7 is a controller of ubiquitination and subsequent turnover of a number of oncoproteins, including myc (Yeh et al. 2018). For cancer invasion to occur, it is not enough for cells to become independent of adhesion signalling: the potentially invasive cells must have a survival advantage in their environment. They must adapt to diminished oxygen levels, to a metabolic shift that increases the rate of glycolysis, and to tolerance of acidosis. These changes may present opportunities for novel anti-invasion strategies.

9.8

Metastasis as an Aspect of Tumour Progression

About 50% of human tumours can be cured by surgical removal, without the need for radiotherapy or chemotherapy. Surgery is a local treatment: tumours that have not spread from their site of origin (metastasized) will be curable by surgery, and those tumours that have metastasized will not be curable by surgery alone. Certain kinds of primary tumours have a predisposition to metastasize to particular sites: lung tumours to the brain, prostate tumours to bone, colon tumours to the liver. Some tumour types tend to metastasize early, so that by the time a primary tumour is clinically detectable, it will have already spread to multiple secondary sites (Hill 1987). Other tumours may become very large before they start to metastasize; in all aspects of metastasis, there is wide variation from tumour to tumour. Certain drugs can inhibit the process of metastasis, for example, aspirin. After metastatic cells detach from the primary tumour, and enter the circulation, they eventually attach to the wall of a blood vessel, and from there invade the tissues and form a secondary tumour. The attachment process requires that blood platelets surround the circulating tumour cell, and form a small, sticky clot. Drugs that inhibit platelet aggregation, including aspirin, can thus prevent metastasis. Experimental drugs have been tested as inhibitors of all stages of metastasis. Compounds that inhibit cell motility (e.g. microtubule inhibitors) block outward migration from the surface of the primary tumour to blood vessels—i.e., invasion; cells being transported in the blood are susceptible to the usual cytotoxic drugs; the process of thrombus formation and attachment of the thrombus to the endothelium can be inhibited by anticoagulants or platelet aggregation inhibitors. Migration from the endothelium to the secondary tumour site within the tissues is again a process of invasion. It has been suggested that antimetastatic therapy is either unnecessary or useless. The argument is that if a tumour has not metastasized at the time of diagnosis, then it should be

9.8

Metastasis as an Aspect of Tumour Progression

193

curable by local treatment (radiation and/or surgery) and antimetastatic drugs have nothing to add. If, on the other hand, the tumour has already metastasized at the time of diagnosis, then it is too late for an antimetastatic drug to be useful. This argument is probably correct for aspirin-like drugs. For some developmental antimetastatic drugs, this argument may not apply. Normal tissues have survival signals that require attachment either to basement membrane or to other cells: this is how the body controls organ and tissue formation and development. If a tumour cell can express the survival signal constitutively, independently of an attachment signal, then an inhibitor of that survival signal may prevent metastasis. The equations developed by Goldie and Coldman to relate curability of tumours to the rate of mutation to drug resistance (Eqs. 6.2, 6.3) can also be applied to metastasis, this time making the assumption that a primary tumour is curable, but a metastatic tumour is not. While cellular attachment to ECM is principally integrin-dependent, cell–cell attachment probably involves primarily another family of adhesion molecules, the cadherins. It is estimated that mutations that make E-cadherin signalling attachmentindependent give an 80% lifetime risk of gastric cancer (Black et al. 2014). Normal cells that wander off on their own are programmed to die (other than blood cells and cells of the immune system). In metastatic cancer cells, these mechanisms for eliminating antisocial cells are disabled. For example, in some metastatic cells the enzyme PAK-4 is upregulated, fooling a detached cell into behaving as if it is still attached, and it is therefore able to replicate. A number of inhibitors of mutated anchorage-dependent survival pathways, including PAK-4, are under investigation as antimetastatic drugs, and the expectation is that they should be active against tumour cells that have already metastasized (Eswaran et al. 2009). Nek-2 inhibitors (Chap. 5) are also reported to block metastasis (Xu et al. 2020). In the studies of Sakai et al. (2018) of genes involved in the progression of intestinal cancer (discussed above in connection with invasion) KrasG12D was critical for liver metastasis following Tgfbr2 deletion, and the incidence of metastasis was highest with the ApcΔ716 KrasG12D Tgfbr2-/- genotype. Activation of Wnt and Kras with suppression of TGFβ signalling was sufficient, in intestinal epithelial cells, for colorectal cancer metastasis. Must an epithelial cell lose extracellular matrix (ECM) or basement membrane attachment first, before it can lose cell–cell adhesion and become metastatic? Probably: a cell that loses cell–cell attachment but still requires ECM attachment would have no selective advantage. Although metastasis is clearly a multi-stage process, it has been argued that there may be a single key step that is rate-limiting for the process: Dujon et al. (2021), modelling breast cancer metastasis, suggest that the most critical parameter in clinical metastasis is the survival duration of circulating tumour cells. This is clearly related to constitutive expression of a survival signal in the absence of attachment, since in the absence of such a signal unattached cells are programmed to enter apoptosis. Epigenetic changes have been implicated as drivers of metastasis (Chattergee et al. 2018). It is likely that changes in expression levels of adhesion molecules, or their associated signalling pathways, could result in primary tumour cells becoming

194

9 The Cancer Stem Cell and Tumour Progression

metastatic. The adaptations in energy metabolism that occur as a result of hypoxia probably become permanent as a result of epigenetic reprogramming. Epigenetic activation of Nrf2, a transcription factor that regulates the expression of antioxidant proteins, is a frequent event in tumour progression, and may be an adaptation to oxidative stress.

9.9

Modelling Metastasis

Does metastasis have to occur after invasiveness? The mutations that lead to metastasis, such as constitutive cadherin signalling, can occur at any time. However, such mutations would be without effect in cells that require an integrin survival signal from basement membrane—they would confer no selective advantage. Also, tumours that have become invasive rapidly become greater, since their Gompertzian asymptote is now larger, so there will be more cells at risk of all mutations. The earliest stage of metastasis, migration of cells that are no longer anchoragedependent from the primary tumour to the nearest blood vessel, is in fact an aspect of invasion, and depends upon cathepsins and collagenases. The approach to modelling metastasis taken here thus assumes that the epithelial-mesenchymal transition must precede metastasis. The dynamics of metastasis then resemble the dynamics of drug resistance described in Chap. 6, with the number of new metastatic cells dependent upon the number of primary tumour cells at risk and a mutation rate. However, once a newly metastatic cell has become established at the secondary site, its growth kinetics may differ from that of the primary: for example, it will be at an earlier stage of the Gompertzian curve, and its growth will no longer be constrained by the Gompertzian asymptote of the primary tumour. Metastatic tumours may thus have shorter doubling times than the primary growth. This could result in metastatic tumours having enhanced sensitivity to cycle-specific drugs. On the other hand, metastases may be in a site that is less readily accessible to drugs than that of the primary. For this reason, brain metastases are particularly hard to treat. The brain is not the only one of these “pharmacological sanctuary” sites. In the early days of chemotherapy, it was observed that the relapse rate in boys treated for childhood leukaemia was higher than for girls. The difference was traced to the ability of the testicles to act as a pharmacological sanctuary site (Jackson 1992). Figure 9.2 shows treatment of a mouse tumour treated with methotrexate as modelled by the SKIPPER program (supplement). The number of metastatic cells (initially zero) increases faster than the primary tumour because their number is driven not only by replication of tumour cells already in the brain, but also by additional metastatic cells entering the brain. After the start of treatment, the number of metastatic cells declines very slightly, because fewer additional metastatic cells were entering the brain. By 48 days the number of primary tumour cells was still declining, but the number of brain metastases was again rising. Since treatment decisions often depend on whether a tumour is likely to be, or become, metastatic, it would be useful to be able to use biopsy data to predict

9.10

The Big Bang Model of Tumour Growth

195

Fig. 9.2 Growth of a primary subcutaneous mouse tumour and brain metastases after treatment with methotrexate, modelled by the SKIPPER program

metastatic potential. Albaradei et al. (2021) discuss the use of artificial intelligence (AI) techniques, based upon genomic and proteomic data for metastasis prediction.

9.10

The Big Bang Model of Tumour Growth

The widely accepted view that tumour progression involves a succession of mutations, each of which confers a selective advantage, and sets in train a process of Darwinian selection was challenged by Sottoriva et al. (2015) who proposed (initially for colon cancer) the “big bang” model: that early-stage, undetectable tumours already had all the mutations required for subsequent growth to an advanced tumour and that early subclones were not subject to stringent selection. They argued on statistical grounds that mutations in the growing tumour were neutral—presenting neither a selective advantage nor disadvantage. Genomic profiling of >300 individual glands showed high intratumoral heterogeneity and subclone mixing, a finding consistent with most heterogeneity arising in early tumour growth. A subsequent paper extended this conclusion to 14 tumour types (Williams et al. 2016). Their argument was criticised on various grounds, including the fact that mutations conferring a small, but significant selective advantage may be hard to distinguish from neutral mutations. Tung and Durrett (2021) reviewed the evidence for and against neutral evolution in cancer and presented modelling studies that neutral evolution in an exponentially growing tumour can be distinguished from a two-stage model in which both neutral and advantageous mutations are assumed to occur. Graham and Sottoriva (2017) suggest that both selective and neutral evolution feature prominently in carcinogenesis. Bollen et al. (2021) combined live cell imaging of tumour organoid outgrowths with whole genome sequencing

196

9 The Cancer Stem Cell and Tumour Progression

of each imaged cell and showed that karyotypic alterations could arise within a few cell generations. In contrast, gross genome-wide karyotype alterations were generated in a single erroneous cell division, providing support that aneuploid tumour genomes can evolve via punctuated evolution. The persistence of neutral mutations in the presence of advantageous mutations over many cell doubling times implies the lack of selective pressure; if there is no competition for resources, there can be no such thing as a competitive advantage. We know that advanced tumours can acquire new mutations, since drug resistance mutations appear and accumulate when the presence of an anticancer drug provides selective pressure. The frequency of a double-mutant resistant to two drugs is of the order of 10-12, so the likelihood that such cells were already present in a sub-clinical tumour of, say, 107 cells is vanishingly low, based on the equation of Luria and Delbrück (Eq. 6.1), and would seem to be inconsistent with a strict reading of the big bang theory. Is it possible that the multiple mutations required for a cancer stem cell to evolve into an advanced cancer are already present in an early clone? Given the very high mutation rates of chromosomally unstable cells, and if we accept that the only essential mutations after the cancer stem cell stage are those required for epithelial-mesenchymal transition and for metastasis, it cannot be ruled out. It must also be remembered that the environment of a growing tumour is not constant, so that a mutation that is neutral in one microenvironment may become a driver mutation when the microenvironment changes. The phenomenon of chromothripsis (Campbell et al. 2020), though driven by chromosomal fragmentation and mis-joining, rather than multiple mutations, appears to result in similar dynamics to the big bang, in that it causes multiple copy number changes that are likely to have a neutral selective effect.

9.11

Antiandrogens in Treatment of Prostate Cancer

The prostate gland, part of the male reproductive system, produces seminal fluid, which supports and nourishes spermatozoa. Human prostate cancer can be a lifethreatening metastatic disease, but a high proportion of prostate tumours follow a rather indolent course, progressing sufficiently slowly that a high proportion of men with prostate cancer die from other causes (Bassetto et al. 2016). Such cases, if diagnosed, are often left untreated. The fact that many cases of prostate cancer progress slowly makes it of interest in understanding the progression process, but as yet no factors have been identified that distinguish the slowly progressing cases from the more aggressive form of the disease, or distinguish prostate cancer from other tumours that usually progress more rapidly. A blood biomarker, prostatespecific antigen (PSA) is widely used to monitor progression and for prostate cancer screening. However, PSA is not tumour-specific, as it is also elevated in benign prostatic hypertrophy and prostate infection. An early treatment for prostate cancer was surgical castration. A sub-set of prostate tumours was found to be resistant to castration, indicating that some

9.11

Antiandrogens in Treatment of Prostate Cancer

197

Fig. 9.3 Abiraterone, an inhibitor of prostate cancer progression

advanced cases had become androgen-independent. This independence was often related to downregulation of the androgen receptor (AR) though it could also be related to mutations that enable AR to signal in the absence of ligand binding. Normal prostate tissue and early-stage prostate tumours express the androgen receptor. Unlike receptors for peptide hormones and growth factors, AR is not cell-membrane bound, but is cytosolic, since its ligands, testosterone and dihydrotestosterone, are lipophilic and readily cross cell membranes. When the AR binds testosterone, it migrates to the nucleus, where it acts as a transcription factor. Nuclear AR causes histone demethylation, resulting in a more open chromatin structure, which causes increased binding of several transcription factors containing a region termed the bromodomain (Latchman 2008). Why do some prostate tumours never progress? Or at least, many men with non-metastatic prostate cancers die of other causes before progression occurs. Is the mutation rate lower in these tumours, or do they retain sufficient differentiated function that they remain under tight physiological control? These non-progressing or slowly progressing tumours usually remain androgen-dependent. The more aggressive prostate tumours often progress to androgen independence: downregulation of AR is an instance of de-differentiation, loss of differentiated tissue function. Abiraterone (Fig. 9.3) is a drug that has been shown to be an inhibitor of prostate cancer progression (Feyerabend et al. 2018). It inhibits a cytochrome P450 isoform, CYP17A1, that is on the pathway of androgen biosynthesis, and thus depletes circulating testosterone levels. Other inhibitors of androgen synthesis include the 5-α-reductase inhibitors such as finasteride and dutasteride, which are used to treat benign prostatic hyperplasia (BPH). The 5-α-reductase inhibitors have been shown to inhibit progression of BPH to prostatic carcinoma in experimental systems. Another class of antiandrogens with clinical activity in prostate cancer (including sometimes activity against castrateresistant tumours) is the androgen-receptor antagonists, such as bicalutamide and enzalutamide. Drug resistance as an aspect of tumour progression has been discussed in Chap. 6. The emergence of drug resistance is an evolutionary phenomenon, and as such can be minimised by application of evolutionary principles, including drug exposure as a selective agent. In prostate cancer, as with other tumours, combination chemotherapy is used to combat drug resistance (Bozic et al. 2013). A transgenic adenocarcinoma of mouse prostate (TRAMP) model has been used to study factors that

198

9

The Cancer Stem Cell and Tumour Progression

Fig. 9.4 Treatment of advanced prostate cancer with combined abiraterone plus bicalutamide, modelled with the VIP program

correlate with progression (Gingrich et al. 1997; Hurwitz et al. 2001; Wickström et al. 2005). These transgenic male mice develop spontaneous prostate tumours whose biology has similarities to that of aggressively progressing human prostate tumours. The resulting lesions vary in severity from premalignant to metastatic. This model is thus used to study biochemical factors that correlate with tumour progression (Kido et al. 2019). A number of epigenetic changes correlated with progression in this model, including epigenetic activation of the antioxidant protein Nrf2 and downstream targets of the Nrf2 promoter. This activation of Nrf2 in progression is not specific to prostate tumours: it is also observed when CML progresses from the chronic phase to the acute phase (Chap. 7). Expression of genes related to angiogenesis (VEGF) and metastasis (MMP-2 and MMP-9) was increased as the TRAMP model progressed. Platelet-derived growth factor B was overexpressed in the more advanced tumours (Zhang et al. 2018). Progression of the TRAMP tumours is inhibited by antiinflammatory and antiangiogenic treatment. Development of these tumours is androgen-dependent and is inhibited by abiraterone and 5-α-reductase inhibitors such as finasteride (Opoku-Acheampong et al. 2017). Maintaining an alkaline environment (with sodium bicarbonate) inhibited tumour development in the TRAMP model (Ibrahim-Hashim et al. 2012). As discussed above, the SKIPPER program was originally written to model drug resistance with a genetic algorithm and later extended to describe other aspects of tumour progression, including angiogenesis, metastasis, and immunogenicity. The SKIPPER program described the growth and treatment of transplanted tumours in inbred mice. In spontaneous human tumours, growth parameters show much more individual variation. A development of the SKIPPER model, the Virtual Interactive Patient (VIP) (described in Chap. 11), models the growth of human tumours by replacing the unique values of growth parameters by stochastic variables. Figure 9.4 shows an example of a simulation of treatment of an advanced human prostate tumour with the combination of abiraterone (500 mg) + bicalutamide (25 mg) daily.

9.11

Antiandrogens in Treatment of Prostate Cancer

199

Fig. 9.5 Treatment of advanced prostate cancer with combined abiraterone, bicalutamide, and enzalutamide modelled with the VIP program

The predictions of the model are shown as a Kaplan-Meier plot, which plots the proportion of survivors as a function of time. Note that this simulation predicted 4/10 five-year survivors. Figure 9.5 shows the predicted outcome of adding a third drug to the combination. The proportion of long-term survivors has increased to 70%. Tumour progression is clearly a multi-stage process, and each stage involves one or more mutations or chromosomal rearrangements, but progression is not a random process. So long as the tumour is restricted to a small region on a basement membrane, it is unlikely to have sufficient cells for further mutations to be very likely, even with genetic instability. The epithelial-mesenchymal transition plays a gatekeeper role, because it allows the tumour to grow as a solid cell mass, instead of a flat sheet. The resulting increase in tumour cell number makes further mutations more probable. This, in turn, results in hypoxia, because most cells are not able to receive much oxygen by diffusion from the tumour surface. The next mutation to occur probably results in a change in energy generation from oxidative phosphorylation to glycolysis. Hypoxia triggers expression of HIF1α, resulting in stimulation of tumour angiogenesis. Tumour angiogenesis is an essential stage of tumour progression, but it does not need a separate mutation: it is an automatic response to hypoxia. Angiogenesis allows further tumour growth, making the next mutation more probable. Two further mutations are needed for the process to culminate in an advanced metastatic tumour, and it is not clear whether the order in which they occur is important. For metastasis to occur requires that cell–cell attachment is no longer essential for cell division, e.g. E-cadherin signalling may become constitutive. As the tumour grows larger and more diverse, it must become more antigenic, and at greater risk of being destroyed by the host immune system. To prevent this, the tumour triggers immune checkpoints, as discussed in the next chapter. Tumour progression is a phenomenon of Darwinian selection, and as such the outcome is determined by the interaction between the tumour and its

200

9

The Cancer Stem Cell and Tumour Progression

microenvironment. This includes not only the availability of nutrients and oxygen, physical factors such as pH and hydrostatic pressure, hormonal and immunological factors, but also the surrounding normal cells. Fibroblasts can both promote and inhibit tumour growth. Kobayashi et al. (2021) showed that genes involved in bone morphogenic protein (BMP) signalling influenced tumour growth: GREM1, produced by fibroblasts, was stimulatory, and ISL1, produced by a different fibroblast population, inhibited growth of liver metastases of colon cancer. GREM1 and ISLR, in turn, are reciprocally regulated by TGFβ. Lan et al. (2022) reported that in pancreatic ductal adenocarcinoma, the epithelial-mesenchymal transition was inhibited by GREM1.

References Albaradei S, Thafar M, Alsaedi A et al (2021) Machine learning and deep learning methods that use omics data for metstasis prediction. Comput Struct Biotechnol J 19:5008–5018 Bach SP, Renehan AG, Potten CS (2000) Stem cells: the intestinal stem cell as a paradigm. Carcinogenesis 21:469–476 Barker N, van Es JH, Kuipers J et al (2007) Identification of stem cells in small intestine and colon by marker gene Lgr5. Nature 449:1003–1007 Bassetto M, Feria S, Pertusati F et al (2016) Design and synthesis of novel bicalutamide and enzalutamide derivatives as antiproliferative agents for the treatment of prostate cancer. Eur J Med Chem 118:230–243 Batlle E, Clevers H (2017) Cancer stem cells revisited. Nat Med 23:1124–1134 Benzekry S, Chapuisat G, Ciccolini J et al (2012) A new mathematical model for optimizing the combination between antiangiogenic and cytotoxic drugs in oncology. C R Acad Sci Paris Ser I 350:23–28 Black MD, Kaneshiro R, Lai J, Shimizu DM (2014) Hereditary diffuse gastric cancer associated with E-cadherin germline mutation: a case report. Hawaii J Med Public Health 73:204–207 Bollen Y, Stelloo E, van Leenen P et al (2021) Reconstructing single-cell karyotype alterations in colorectal cancer identifies punctuated and gradual diversification patterns. Nat Genet. 53:1187– 1195 Boxer MB, Jiang J-K, Vander Heiden MG et al (2010) Evaluation of substituted N,N′-diarylsulfonamides as activators of the tumor cell specific M2 isoform of pyruvate kinase. J Med Chem 53:1048–1055 Bozic I, Antal T, Ohtsuki H et al (2010) Accumulation of driver and passenger mutations during tumor progression. Proc Natl Acad Sci U S A 107:18545–18550 Bozic I, Reiter JG, Allen B et al (2013) Evolutionary dynamics of cancer in response to targeted combination therapy. eLife 2:e00747. https://doi.org/10.7554/eLife.00747 Bustamante E, Pederson PL (1980) Mitochondrial hexokinase of rat hepatoma cells in culture: solubilization and kinetic properties. Biochemistry 19:4972–4977 Campbell PJ, Getz G, Korbel JO et al (2020) Pan-cancer analysis of whole genomes. Nature 578: 82–93 Carey N (2011) The epigenetics revolution. Icon Books Ltd, London, pp 269–272 Carracedo A, Cantley LC, Pandolfi PP (2013) Cancer metabolism: fatty acid oxidation in the limelight. Nat Rev Cancer 13:227–232 Chang S, Yim S, Park H (2019) The cancer driver genes IDH1/2, JARID1C/KDM5C, and UTX/KDM6A: crosstalk between histone demethylation and hypoxic reprogramming in cancer metabolism. Exp Mol Med 51:1–17

References

201

Chaplain MAJ, Anderson ARA (1996) Mathematical modelling, simulation and prediction of tumour-induced angiogenesis. Invasion Metastasis 16:222–234 Chattergee A, Rodger EJ, Eccles MR (2018) Epigenetic drivers of tumourigenesis and cancer metastasis. Semin Cancer Biol 51:149–159 Christofk HR, Vander Heiden MG, Harris MH et al (2008) The M2 splice isoform of pyruvate kinase is important for cancer metabolism and tumour growth. Nature 452:230–233 Christopherson T (2017) Tripping over the truth. Chelsea Green Publishing, London Clevers H, Loh KM, Nusse R (2014) Stem cell signalling. An integral program for tissue renewal and regeneration: Wnt signaling and stem cell control. Science 346:1248012. https://doi.org/10. 1126/science.1248012 Damaghi M, West J, Robertson-Tessi M et al (2021) The harsh microenvironment in early breast cancer selects for a Warburg phenotype. Proc Natl Acad Sci U S A 118:e2011342118. https:// doi.org/10.1073/pnas.2011342118 Dando I, Pozza ED, Biondani G et al (2015) The metabolic landscape of cancer stem cells. IUBMB Life 67:687–693 Danhier P, Banski P, Payen VL et al (2017) Cancer metabolism in space and time: beyond the Warburg effect. Biochim Biophys Acta Bioenerg 1858:556–572 DeNicola GM, Cantley LC (2015) Cancer's fuel choice:new flavors for a picky eater. Mol Cell 60: 514–523 Dennett DC (1995) Darwin’s dangerous idea: evolution and the meanings of life. Simon and Schuster, New York Dow LE, O’Rourke KP, Simon J et al (2015) Apc restoration promotes cellular differentiation and reestablishes crypt homeos tasis in colorectal cancer. Cell 161:1539–1552 Dujon AM, Capp J-P, Brown JS et al (2021) Is there one key step in the metastatic cascade? Cancers (Basel) 13:3693. https://doi.org/10.3390/cancers13153693 Epstein T, Xu L, Gillies RJ, Gatenby RA (2014) Separation of metabolic supply and demand: aerobic glycolysis as a normal physiological response to fluctuating energetic demands in the membrane. Cancer Metab 2(7):2014. https://doi.org/10.1186/2049-3002-2-7 Eswaran J, Soundararajan M, Knapp S (2009) Targetting group II PAKs in cancer and metastasis. Cancer Metastasis Rev 28:209–217 Feyerabend S, Saad F, Li T et al (2018) Survival benefit, disease progression and quality of life outcomes of abiraterone acetate plus prednisone versus docetaxel in metastatic hormonesensitive prostate cancer: a network meta-analysis. Eur J Cancer 103:78–87 Finley SD, Chu LH, Popel AS (2014) Computational systems biology approaches to antiangiogenic cancer therapeitics. Drug Discov Today 20:187–197. https://doi.org/10.1016/j. drudis.2014.09.026 Folkman J (2003) Fundamental concepts of the angiogenic process. Curr Mol Med 3:643–651 Frank SA (2007) Dynamics of cancer: incidence, inheritance, and evolution. Princeton University Press, Princeton, NJ Garcia-Cao I, Song MS, Hobbs RM et al (2017) Systemic elevation of PTEN induces a tumorsuppressive metabolic state. Cell 149:49–62 Gatenby RA, Gillies RJ (2004) Why do cancers have high aerobic glycolysis? Nat Rev Cancer 4: 891–899 Gehart H, Clevers H (2019) Tales from the crypt: new insights into intestinal stem cells. Nat Rev Gastroenterol Hepatol 16:19–34 Gingrich JR, Barrios RJ, Kattan MW et al (1997) Androgen-independent prostate cancer progression in the TRAMP model. Cancer Res 57:4697–4689 Graham TA, Sottoriva A (2017) Measuring cancer evolution from the genome. J Pathol 241:183– 191 Hill RP (1987) Metastasis. In: Tannock IF, Hill RP (eds) The basic science of oncology. Pergamon Press, Oxford, pp 160–191

202

9

The Cancer Stem Cell and Tumour Progression

Hitosugi T, Kang S, Vander Heiden MG et al (2009) Tyrosine phosphorylation inhibits PKM2 to promote the Warburg effect and tumor growth. Sci Signal 2:ra73. https://doi.org/10.1126/ scisignal.2000431 Holland J (1992) Adaptation in natural and artificial systems. MIT Press, Cambridge, MA Hurwitz AA, Foster BA, Allison JP et al (2001) The TRAMP mouse as a model for prostate cancer. Curr Protoc Immunol Chapter 20:Unit 20.5. https://doi.org/10.1002/0471142735.im2005s45 Ibrahim-Hashim A, Cornnell HH, Abrahams D et al (2012) Systemic buffers inhibit carcinogenesis in TRAMP mice. J Urol 188:624–631 Iranmanesh Y, Jiang B, Favour OC et al (2021) Mitochondria’s role in the maintenance of cancer stem cells in glioblastoma. Front Oncol 11:582694. https://doi.org/10.3389/fonc.2021.582694 Iranzo J, Martincorena I, Koonin EV (2018) Cancer-mutation network and the number and specificity of driver mutations. Proc Natl Acad Sci USA 115:E6010–E6019 Jackson RC (1992) The theoretical foundations of cancer chemotherapy introduced by computer models. Academic, New York, pp 185–192 Jackson RC, Fernandez E, Radivoyevitch T (2015) Bayesian systems for optimizing treatment protocols in oncology. In: Ur-Rahman A (ed) Frontiers in clinical drug research – anticancer agents. Bentham Science Publishers, pp 50–119 Kido LA, Lamas CA, Marostica MR (2019) Transgenic adenocarcinoma of the mouse prostate (TRAMP) model: a good alternative to study PCa progression and chemoprevention approaches. Life Sci 217:141–147 Kobayashi H, Gieniec KA, Wright JA et al (2021) The balance of stromal BMP signaling mediated by GREM1 and ISLR drives colorectal carcinogenesis. Gastroenterology 160:1224–1239 Koka P, Mundre RS, Rangarajan R et al (2018) Uncoupling Warburg effect and stemness in CD133+ve cancer stem cells from Saos-2 (osteosarcoma) cell line under hypoxia. Mol Biol Rep 45:1653–1662 Lan L, Evan T, Li H et al (2022) GREM1 is required to maintain cellular heterogeneity in pancreatic cancer. Nature 607:163–168 Latchman DS (2008) Eukaryotic transcription factors, 5th edn. Academic, New York Li Z, Lock JG, Olofsson H et al (2010) Integrin-mediated cell attachment induces a PAK-4dependent feedback loop regulating cell adhesion through modified integrin alpha v beta 5 clustering and turnover. Mol Biol Cell 21:3317–3329 Li S, Xu H-X, Wu C-T et al (2018) Angiogenesis in pancreatic cancer: current research status and clinical implications. Angiogenesis 22:15–36 Li Z-F, Yao Y-D, Zhao Y-Y et al (2020) Effects of PAK4/LIMK1/cofilin-1 signaling pathway on proliferation, invasion, and migration of human osteosarcoma cells. J Clin Lab Anal 34:e23362. https://doi.org/10.1002/jcla.23362. Epub 2020 May 28 Li M, Wang Y, Li M et al (2021) Integrins as attractive targets for cancer therapeutics. Acta Pharm Sin B 11:2726–2737 Lloyd MC, Cunningham JJ, Bui MM et al (2016) Darwinian dynamics of intratumoral heterogeneity: not solely random mutations but also variable environmental selection forces. Cancer Res 76:3136–3144 Martinez-Climent JA, Andreu EJ, Prosper F (2006) Somatic stem cells and the origin of cancer. Clin Transl Oncol 8:647–663 Mita MM, Sargsyan L, Mita AC, Spear M (2013) Vascular disrumpting agents in oncology. Expert Opin Investig Drugs 22:317–328 Opoku-Acheampong AB, Henningson JN, Lindshield BL (2017) The impact of finasteride, and dutasteride treatments on proliferation, apoptosis, androgen receptor, 5-α-reductase 1 and 5-α-reductase 2 in TRAMP mouse prostates. Heliyon 3:e00360. https://doi.org/10.1016/j. heliyon.2017.e00360 Physiomics plc (2022). www.physiomics-plc.com Ramjiawan RR, Griffioen AW, Duda DG (2017) Anti-angiogenesis for cancer revisited: is there a role for combinations with immunotherapy? Angiogenesis 20:185–204

References

203

Sakai E, Nakayama M, Oshima H et al (2018) Combined mutation of Apc, Kras, and Tgfbr2 effectively drives metastasis of intestinal cancer. Cancer Res 78:1334–1346 Sancho P, Barneda D, Heerschen C (2016) Hallmarks of cancer stem cell metabolism. Br J Cancer 114:1305–1312 Sato M, Kawana K, Adachi K et al (2016) Spheroid cancer stem cells display reprogrammed metabolism and obtain energy by actively running the tricarboxylic acid (TCA) cycle. Oncotarget 7:33297–33305 Shay JW, Wright WE (2000) Hayflick, his limit, and cellular ageing. Nat Rev Mol Cell Biol 1:72– 76 Shin S-Y, Rath O, Zebisch A et al (2010) Functional roles of multiple feedback loops in extracellular signal-regulated kinase and Wnt signalling pathways that regulate epithelial-mesenchymal transition. Cancer Res 70:6715–6724 Smallbone K, Gatenby RA, Gillies RJ et al (2007) Metabolic changes during carcinogenesis: potential impact on invasiveness. J Theor Biol 244:703–713 Sottoriva A, Kang H, Ma Z et al (2015) A big bang model of human colorectal tumor growth. Nat Genet 47:209–216 Sun L, Suo C, Li S-T et al (2018) Metabolic reprogramming for cancer cells and their microenvironment: beyond the Warburg effect. Biocheim Biophys Acta Rev Cancer 1870:51–66 Tung H-R, Durrett R (2021) Signatures of neutral evolution in exponentially growing tumours: a theoretical perspective. PLoS Comput Biol 17:e1008701 van der Flier LG, Clevers H (2009) Stem cells, self-renewal, and differentiation in the intestinal epithelium. Annu Rev Physiol 71:241–260 Vander Heiden M, Cantley LC, Thompson CB (2009) Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324:1029–1033 Vermeulen L, Morrissey E, van der Heiden M et al (2013) Defining stem cell dynamics in models of intestinal tumour initiation. Science 342:995–998 Viallard C, Larrivée B (2017) Tumor angiogenesis and vascular normalization: alternative therapeutic targets. Angiogenesis 20:409–426 Wang Y, Liu G, Tong D et al (2015) Metformin represses androgen-dependent and androgenindependent prostate cancers by targeting androgen receptor. Prostate 75:1187–1196 Wells CM, Jones GE (2010) The emerging importance of group II PAKs. Biochem J 425:465–473 Wickström P, Lindahl C, Bergh A (2005) Characyerization of the autochthonous transgenic adenocarcinoma of the mouse prostate (TRAMP) as a model to study effects of castration therapy. Prostate 62:148–164 Williams MJ, Werner B, Barnes CP et al (2016) Identification of neutral tumor evolution across cancer types. Nat Genet 48:238–244 Wodarz D, Komarova NL (2014) Dynamics of cancer: mathematical foundations of oncology. World Scientific, Singapore, pp 106–128 Xu T, Zeng Y, Shi L et al (2020) Targeting NEK2 impairs oncogenesis and radioresistance via inhibiting the Wnt1/β-catenin signaling pathway in cervical cancer. J Exp Clin Cancer Res 39: 183 Yang W, Zheng Y, Xia Y et al (2012) ERK1/2-dependent phosphorylation and nuclear translocation of PKM2 promotes the Warburg effect. Nat Cell Biol 14:1295–1304 Yeh C-H, Bellon M, Nicot C (2018) FBXW7: a critical tumor suppressor of human cancers. Mol Cancer 17:115. https://doi.org/10.1186/s12943-018-0857-2 Zhang Y, Wang D, Li M et al (2018) Quantitative proteomics of TRAMP mice combined with bioinformatics analysis reveals that PDGF-β regulatory network plays a key role in prostate cancer progression. J Proteome Res 17:2401–2411

Chapter 10

Evading the Antitumour Immune Response

Abstract The mutations that cause (or accompany) malignancy result in tumour cells being detected as “non-self” by the immune system and may evoke an antitumour immune response. As tumour progression proceeds, genetic instability results in the developing tumour becoming increasingly immunogenic. Immune surveillance may result in early-stage tumours being eliminated without developing into clinically evident disease. Less immunogenic tumours may be held in a state of dormancy in which cell proliferation is balanced by immune destruction. Immune checkpoints are negative regulators of the immune system that can prevent immune destruction of cells on which they are expressed. Immune checkpoints may be expressed by developing tumours and confer a survival advantage on those tumour cells that express them. Monoclonal antibodies have been developed against a number of cancer-associated targets, including growth factor receptors and immune checkpoints. Monoclonal antibodies are limited to extracellular targets. Cancer vaccines that target neo-epitopes found on tumour cells are under evaluation against multiple tumour types. To prevent the rapid emergence of resistance, vaccines usually target multiple mutated cell surface components. Cellular immunotherapy is a technique in which immune cells (usually T cells) either from the patient or from a donor are expanded ex vivo, often genetically engineered for greater efficacy, and infused into the patient. In general, the immune response retains full activity against tumour cells with acquired drug resistance, and thus provides a valuable treatment option for combating drug resistance.

CD4 cells are stimulated by the presence of foreign antigens. Once stimulated they divide and send activation signals to CD8 cells and B cells. CD8 cells recognise and kill virusinfected cells. B cells release antibodies that attack viruses and other infectious agents. Martin A, Nowak, Evolutionary Dynamics (2006)

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_10. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_10

205

206

10.1

10

Evading the Antitumour Immune Response

Cells of the Immune System

The immune system is, after the nervous system, the most complex organ system of the animal body. Its implementation involves multiple cell populations, communicating through a large number of cytokines, which in turn activate multiple overlapping signal pathways driving transcription networks. The immune system shows many of the same emergent properties that we see in the nervous system: learning, memory, pattern recognition, self/non-self recognition, and information processing between detector and effector arms. The modern immune system is a multi-layered system that bears traces of its evolutionary history. This results in some redundancy of function. Macrophages (MΦ) are probably the most primitive cells of the immune system. In some evolutionarily simple organisms, macrophages (“amoeboid cells”) may be the only mobile cells. Their original function was probably to act as garbage collectors, scavenging dead and damaged cells, and thus involved in tissue remodelling. Macrophage activation normally proceeds by them detecting molecules on the cell surface that should be inside the cell, indicating cellular damage. The immune system appeared early in evolution when macrophages acquired the ability to be activated by lipopolysaccharide (LPS) and related bacterial and fungal cell wall components. Activated macrophages phagocytose bacterial and fungal cells and also release cytokines (notably TNFα) that activate other components of the immune system. Note the positive feedback loop: GM - CSF → MΦ → TNFα → c - Abl → GM - CSF which confers ON/OFF switching kinetics on the innate immune system. In modern organisms, macrophages act as antigen-presenting cells, and thus activate the adaptive immune response. There are various sub-populations of macrophages, differing in the cytokines they release: M1 macrophages are pro-inflammatory, and M2 macrophages are anti-inflammatory. Neutrophils are the primary effector arm of the innate immune system. They are activated by inflammatory cytokines, primarily TNFα. The resulting activation of c-Abl kinase drives formation of reactive oxygen species (ROS) including peroxide, which in presence of the enzyme myeloperoxidase and chloride anion generates hypochlorite, a potent bactericidal agent. In conditions of infection or inflammation, c-Abl turns on the anti-apoptotic protein, Mcl-1, which greatly extends the survival of tissue neutrophils (Jackson and Radivoyevitch 2013). Neutrophils are normally short-lived, because their prolonged activation causes severe tissue damage. The tissue damage in inflammatory diseases such as rheumatoid arthritis is largely neutrophil-driven. Neutrophil depletion (neutropenia), e.g. after chemotherapy with myelosuppressive drugs, can result in septicaemia. Natural killer (NK) cells are non-B, non-T lymphocytes that are cytotoxic to virally infected cells by a mechanism that is not MHC-dependent. They also have activity against some malignant cells. They probably represent the first evolutionary

10.1

Cells of the Immune System

207

occurrence of true self/non-self recognition resulting in cellular cytotoxicity. Why do we need two immune systems? The innate immune system (neutrophil-driven) is a high-capacity/low-affinity system. It can eliminate many logs of infectious organisms, but it is not good at eliminating the last few logs of pathogens, because that requires prolonged activation, which causes severe damage to normal tissues. The evolutionary response to this has been to supplement the innate immune system with a low-capacity/high-affinity system for mopping up remaining pathogens after the initial burst of neutrophil activation. NK cytotoxicity becomes active about three days after an infection. NK cells have activity against some tumour cells in experimental systems, but infusions of autologous NK cells have not shown much clinical anticancer activity. Nude mice have no T cells or B cells, but they do have NK cells. Human tumour implants grow readily in nude mice, but seldom metastasize, suggesting that NK cells may have activity against circulating tumour cells. There have been clinical studies reporting activity of NK cells against human leukaemia (Fabian and Hodge 2021). NK cells do not develop a specific memory of infection in the way that B cells and T cells can, but there is evidence that repeated stimulation can increase the general level of NK cell sensitivity, a process called “trained immunity”, and this effect may contribute to the activity of general immunostimulants such as BCG (Lawton 2021). B lymphocytes: The adaptive immune response, in which self/non-self recognition is driven by genetic recombination of lymphocyte precursors in early life, followed by clonal expansion, is a later evolutionary development. In broad terms, the humoral immune response—antibody production by activated B cells—combats extracellular pathogens. As such, it is probably not usually involved in antitumour immune responses. However, there have been reports that spontaneous remissions of human melanomas have been accompanied by the appearance of melanoma-specific antibodies in the blood. On the occasions when antitumour antibodies arise, the tumour cell killing is usually complement-mediated. Immunity driven by B cells is long-lived, because of formation of a sub-set of B memory cells, in which apoptosis is suppressed by the anti-apoptotic protein Bcl-2. Tolerance is the process by which certain epitopes are no longer recognised as non-self. T lymphocytes are a population of lymphocytes whose primary function is to eliminate intracellular pathogens. They are the primary effectors of the cellular immune response. Their cytotoxic effect requires physical contact between the cytotoxic T cell and the target cell. T-helper cells (CD4+) produce cytokines that stimulate proliferation of other lymphocyte populations. TH1 cells produce IL-2 which stimulates proliferation of cytotoxic T cells, and TH2 cells stimulate B cell proliferation by secretion of interleukins-4, 5, and 6. Cytotoxic T cells (CD8+) attach to and destroy infected cells that display the epitope to which they are sensitised in association with MHC class I (see below). Activation of the T cell receptor by a cell that displays its specific epitope in association with MHC class I drives several signalling pathways including NFAT and MAP kinase signalling. Inhibitors of NFAT signalling (e.g. cyclosporin A) prevent transplant rejection. The dominant contribution to the antitumour immune response is mediated by cytotoxic T cells. There are additional T cell sub-sets with a regulatory function: regulatory T cells

208

10

Evading the Antitumour Immune Response

(Treg cells) also known as T suppressor cells prevent autoimmunity by triggering apoptosis of T cells that have become sensitised to self-antigens. They can also protect tumour cells from immune destruction, and it has been suggested that high levels of Treg cells in the vicinity of a tumour is a negative prognostic factor. Treg cells are stimulated to differentiate from naïve CD4+ cells by the growth factor TGFβ. Memory cells are a sub-set of activated T cells (and B cells) that maintain survival by up-regulation of the anti-apoptotic protein Bcl-2. Proliferation of sensitised T cells is not unlimited, but is controlled by negative regulatory mechanisms. The negative regulatory effect is driven by a variety of effectors, including FAS ligand and TGFβ. Dendritic cells are the most powerful antigen-presenting cells (APC), though macrophages also act as APC. Follicular dendritic cells, unlike all other cells of the immune system, are not derived from bone marrow stem cells, but are mesenchymal in origin. There is evidence that memory cells often remain in close contact with follicular dendritic cells, suggesting that their continued survival may require “booster” stimulation from dendritic cells. Follicular dendritic cells represent a significant site of continuing infection in HIV disease, but they do not appear to be a useful target for cancer immunotherapy. Dendritic cells are a major source of FAS ligand, which causes apoptosis of T cells.

10.2

The Major Histocompatibility Complex

When the immune system recognises an epitope (typically an 8–11 amino acid peptide sequence with a definite shape) it must determine whether it is self or non-self, and if non-self, whether it is being displayed by an infected cell, or another immune cell that is reacting to that infection. This is determined by the context: is the epitope associated with MHC class I (non-immune cells) or class 2 (immune system cells)? Early studies on allograft rejection showed that there were about fifty cell surface sites associated with weak rejection, and one locus associated with strong rejection. This locus is termed the major histocompatibility complex (MHC)—“complex” because it comprises several closely linked genes (Miller 1987). In humans, it is also known as the human leukocyte antigen (HLA) complex. MHC class I glycoproteins are found on the surface of all cells. MHC class II glycoproteins occur only on cells of the immune system. A third group, MHC class III is involved in the complement system. The MHC system was explored primarily as the self/non-self determinant of skin graft (allograft) rejection in mice. However, its primary biological function is the detection and elimination of virus-infected cells. When these are detected, a specific clone of cytotoxic T cells is generated that recognises a portion of viral protein in complex with class I MHC. T-helper cells, which are involved in activation of B cells to produce antibodies similarly recognise a portion of a foreign protein in association with class II MHC on the surface of an antigen-presenting cell (B cells, dendritic cells or macrophages). The cytotoxic T cells and helper T cells thus activated are said to be “MHC-restricted”.

10.3

10.3

Immunotherapy

209

Immunotherapy

Antitumour immunity has been attributed to macrophages, NK cells, and cytotoxic T cells. Small-molecule activators of macrophages and NK cells (biological response modifiers) show preclinical anticancer activity and there is some limited clinical data. Krestin, a fungal cell wall component, has been used as an anticancer agent in some countries, though clinical evidence for its activity is limited. BCG (used in bladder cancer) is a bacterial cell wall preparation. Levamisole (a macrophage activator used as an anthelmintic) has been used in combination with 5-fluorouracil in adjuvant treatment of colon cancer. Cytokines that activate cytotoxic T cells, notably IL-2, have been investigated as anticancer agents, but are very toxic. Adoptive immunotherapy, in which a patient’s T cells are expanded ex vivo, has shown clinical activity, but this is a complex and costly process. The latest version of this approach is chimeric antigen receptor T cell therapy (CAR-T), discussed below. Krestin, a fungal cell wall component, has been used as an anticancer agent in some countries, though clinical evidence for its activity is limited. BCG (used in bladder cancer) is a bacterial cell wall preparation. Activation of antigen-presenting cells was shown to enhance the response to immunotherapy of weakly antigenic tumours (Sancho-Araiz et al. 2021a, b). Flavone acetic acid and its analogue, dimethylxanthenone acetic acid (DMXAA), have potent activity against mouse tumours and against human tumour xenografts in nude mice. They cause haemorrhagic necrosis. Their mechanism of action is believed to be activation of STING (stimulator of interferon genes) suggesting that interferons are the ultimate effectors. Their clinical activity, however, is modest, probably because in humans, unlike mice, they are rapidly metabolised. Monoclonal antibodies, though they are immune system effectors, are not necessarily immunotherapy, in the sense of modulators of the host immune system. Early therapeutic antibodies were derived from mouse cells, but current antibody treatments use chimeric, humanised, or fully human products, which are non-antigenic, or less antigenic in humans than murine proteins. Many monoclonal antibodies are in fact targeted against immune cells and are indeed immunomodulators. Others are targeted against mutant oncogene proteins. Examples are Herceptin (trastuzumab) for HER2 receptor-positive breast cancer; cetuximab (anti-EGF receptor), used in the treatment of colorectal cancer, and rituximab, targeted to CD20, primarily a B cell antigen, which is active in non-Hodgkin lymphoma. Monoclonal antibodies are unable to penetrate cell membranes, so are necessarily limited to extracellular targets, of which growth factor receptors form an important class of oncology targets. Monoclonal antibodies cannot be administered orally, because they would be degraded by proteolytic enzymes in the digestive tract; they must be administered intravenously. Therapeutic antibodies are themselves immunogenic. There have been various approaches to minimise their antigenicity. It is possible to make human antibodies, but they are much more difficult and expensive to produce than mouse antibodies, especially at large scale. A more practical approach is “humanised” antibodies, in which certain amino acid residues

210

10

Evading the Antitumour Immune Response

or sequences of the mouse IgG molecule are replaced with the corresponding parts of the human molecule. “Chimeric” antibodies are mouse antibodies in which the mouse constant region has been replaced by the corresponding human sequence. Monoclonal antibodies represent the closest approach in medicine to “magic bullets”. They can be engineered to bind selectively to the mutant growth factor receptors found in some tumours without affecting the corresponding normal receptors. Bevacizumab is an antibody that binds and neutralises vascular endothelial growth factor (VEGF). Since VEGF is required for tumour angiogenesis, bevacizumab is used to treat solid tumours that are highly vascularised, particularly renal clear cell tumours. Pembrolizumab, targeted against PD-1, is an example of a monoclonal antibody that modulates the immune system. It is used in the treatment of melanoma. PD-1 is an immune checkpoint molecule whose expression can enable tumour cells to evade immune surveillance, as discussed below. Interferons (normally produced by the innate immune system) have been evaluated as immunotherapy. Recombinant interferon-α has shown activity in leukaemias, in cutaneous T cell lymphomas, and in recurrent melanoma. The major adverse event is flu-like symptoms. Interferon inducers such as flavone acetic acid (FAA) have activity against solid tumours in mice and appear to act as vascular targeting agents, destroying the endothelial cells of tumour vasculature, resulting in haemorrhagic necrosis of the tumour. FAA and its analogues have been disappointing in human medicine, probably because of rapid liver metabolism.

10.4

Escape from Immune Surveillance

The ability of the immune system to distinguish self from non-self is determined early in life by the process of clonal selection (Burnet 1957). For a long time it was believed that, since tumours did not have qualitative differences from normal cells, antitumour immunity could not exist. This view was modified with the discovery of virally induced tumours that expressed viral antigens. It was subsequently observed that somatic mutations that turned cellular proto-oncogenes into active oncogenes could make tumours immunogenic and that the degree of oncogenicity tended to increase as genetic instability caused further divergence from the host genotype. Certain tumour types were found to be more immunogenic than others: melanomas and renal cell carcinomas are examples. Detection and eradication of tumour cells probably evolved as a relatively late, additional function of a system whose primary function was detection of infection. Immune killing of tumour cells was shown to be primarily caused by tumourinfiltrating T cells, though activated macrophages and NK cells can also destroy tumour cells. Macrophage activation, however, is normally transient, whereas activated T cell populations can persist throughout life. Much of the early research on anticancer drug discovery was conducted with transplanted mouse tumours, and to

10.4

Escape from Immune Surveillance

211

minimise variability this usually involved using syngeneic tumours that could be transplanted without evoking an immune response. As little as one tumour cell could then give rise to a lethal tumour. It was observed that allogeneic transplants, i.e. transplants from related, but not genetically identical, animals were more easily curable than syngeneic transplants, and this was attributed to immune transplant rejection. In humans, the observation that organ transplant patients tend to have higher than normal incidence of malignant disease (particularly leukaemias and lymphomas) suggested that immune surveillance was acting to remove transformed cells and that the immunosuppressive drugs used to prevent transplant rejection were inhibiting that process. People with congenital immunodeficiency also have abnormally high incidence of tumours. Regulation of the immune system is complex, involving multiple feedback loops. It is important that immune attack can be rapidly switched off, for example if the system is damaging the bodies own tissues, as in autoimmune disease. In pregnancy the immune system of the mother must be prevented from attacking the tissues of the foetus. A number of the so-called immune checkpoints have evolved to protect the body from attack by its own immune system. For example, the checkpoint molecule CTLA-4 is expressed on a T cell sub-set, the regulatory T cells (Treg cells). Hiding from the immune system appears to be an almost universal attribute of advanced tumours, perhaps because tumours only get to be advanced if they can protect themselves from immune attack. As a growing tumour expresses an increasing number of mutations (some tumours may have more than fifty) they become increasingly immunogenic. At this stage, tumours must express immune checkpoint molecules, or be subject to immune attack. Hanahan and Weinberg (2011), in a ten-year follow-up to their “hallmarks of cancer” paper added two additional hallmarks to their list: “reprogramming of energy metabolism”, by which they meant aerobic glycolysis and its sequelae, and evasion of the host immune response. Some tumours protect themselves from immune attack by stimulating immune checkpoints. Solid tumours often develop immune-inhibitory barriers in the tumour microenvironment (Altvater et al. 2021). This provides cells with a selective advantage and is part of the process of tumour progression. Note that “checkpoint” is used in a different sense from cell cycle checkpoints, which have ON/OFF kinetics. Immune checkpoints are negative regulators of the immune response that may have “volume control” kinetics. Immune escape relies on direct inhibition of immune effector cells and induction of a suppressive microenvironment. The immune checkpoint HLA-G, whose physiological function is to protect a foetus from attack by the mother’s immune system, is involved at all stages of immune suppression. When a tumour expresses HLA-G, 10% of expressing tumour cells can protect the whole tumour.

212

10.5

10

Evading the Antitumour Immune Response

Immune Checkpoint Inhibitors

Ipilimumab inhibits CTLA-4 (first approved for melanoma). An inhibitor of CTLA4 has also been delivered by a recombinant herpes virus, RP2, delivered intralesionally. The construct was reported to have clinical activity, including a complete response in a phase 1 study (Harrington 2022). Nivolumab and pembrolizumab inhibit PD-1 (approved for numerous indications, including lung cancer). Nivolumab was most active against tumours with high mutation burdens, supporting the claim that tumours become more immunogenic as the mutation count increases. The second-generation PD-1 inhibitor, dostarlimab was evaluated in patients with colon cancers deficient in mismatch repair; all 18 patients in this preliminary study had complete responses. Atezolizumab binds to and inhibits the PD-1 ligand, PD-L1 (approved for bladder cancer). Patients treated with immune checkpoint inhibitors are at risk for autoimmune side effects. Intra-vesicular administration of PD-L1 gives activity against bladder cancer with lower risk of systemic side effects. Immune checkpoint inhibitors are unlikely to have an antitumour effect if the tumour is not immunogenic. However, a non-immunogenic tumour would not gain a selective advantage by up-regulating a checkpoint. To date, checkpoint inhibitors have increased 5-year survival in melanoma to about 50%, but have little activity in less immunogenic tumours, such as colorectal cancer. HLA-G is an immune checkpoint expressed by foetal tissue and placenta, that prevents the semi-allogenic foetus from being destroyed by the immune system of the mother. Following birth, expression of HLA-G is silenced epigenetically, but many tumours have reactivated expression of HLA-G, thus mimicking pregnancy, suggesting that it could be a target for immunotherapy (Komohara et al. 2007; Alboncini et al. 2020; Altvater et al. 2021; Anna et al. 2021; Jan et al. 2021). ILT4 is one of the receptors for HLA-G (Garcia et al. 2020) and acts as an immune checkpoint. Anti-ILT4 antibodies are in clinical trials as checkpoint inhibitors as monotherapy and in combination with pembrolizumab. It is reported that the ILT4 antibody can circumvent pembrolizumab resistance in patients with advanced solid tumours (Siu et al. 2022).

10.6

Cancer Vaccines in Treatment and Prevention

To date, use of cancer-preventing (prophylactic) vaccines has been confined to virally induced tumours, of which the best known is the anti-human papillomavirus (HPV) vaccine. This prevents cervical cancer, because this tumour expresses HPV antigens. Testing of vaccines against non-virally induced tumours has necessarily concentrated on treating established tumours. Early attempts to use antibody vaccines raised against neoantigens resulting from tumour mutations had limited success. By analogy with drug resistance, tumours that downregulate the targeted antigen have a selective advantage and result in vaccine resistance. Again by analogy

10.7

Cellular Immunotherapy

213

with drug resistance, vaccines that target three or more neoantigens have been more successful, because the probability of triple mutants causing multiple antibody resistance is low. Tumour cells can, however, become resistant to multiple vaccines by expressing immune checkpoints, and the addition of checkpoint inhibitors to multiple vaccines has given promising results. Recent research has used personalised vaccines: whole genome sequencing is run on samples of a patients tumour and 20 or more cancer-associated mutations are identified. Vaccines are then raised against the abnormal epitopes. A BioNTech study of personalised vaccines against pancreatic cancer showed that 8/16 patients were cancer-free 18 months after surgery after such treatment (WHO 2022). That study used RNA vaccines, in which, instead of administering antibody proteins, mRNA (formulated in a delivery form that enables cellular uptake) codes for synthesis of the antibodies within the tumour cells. Personalised vaccines have also shown preliminary clinical activity against melanoma and glioblastoma (Klein 2022; BioNTech 2022). This is a current growth area in clinical oncology, and these studies will be followed up by large-scale randomised clinical trials. Some of the most successful early trials of personalised vaccines have been in combination with the checkpoint inhibitor, pembrolizumab. As yet, it has not been rigorously demonstrated that results are better than obtained with pembrolizumab alone, though anecdotal reports suggest that remissions with the combinations occur more rapidly than would be expected using checkpoint inhibitors alone. A study in head and neck cancer showed that the Moderna personalised vaccine, in combination with a checkpoint inhibitor, delayed tumour progression by ten months, compared with progression delay of two months for the checkpoint inhibitor alone. Another unanswered question is how many individual vaccines are necessary in each personalised vaccine cocktail: BioNTech target 20 cancer markers, and Moderna, 34 (BioNTch 2022, WHO 2022). Since the rationale for targeting multiple mutations is that tumours can acquire resistance to vaccines by downregulating the abnormal epitopes, by analogy with drug resistance clearly a combination is necessary. For drugs, a triple combination is usually sufficient to reduce the probability of triple mutations to near zero. However, the dynamics of tumour cell killing by the immune system are different from those of drugs. It is believed that use of a large number of individual vaccines not only lowers the probability of multiple resistance, but also increases the total number of activated T cells.

10.7

Cellular Immunotherapy

Early attempts to collect tumour-infiltrating lymphocytes (mainly T cells) and expand them ex vivo, before infusing them back into the patient, had limited success, probably for the same reasons that early vaccine trials failed—tumour heterogeneity and high levels of acquired resistance. Recent attempts at cellular immunotherapy of cancer have concentrated on chimeric antigen receptor T cells (CAR-T cells), in which genetically modified effector cells combine antigen binding and T cell

214

10

Evading the Antitumour Immune Response

activation in a single receptor. CAR-T cells can be both CD4+ and CD8+. They can be autologous (derived from the patient’s own cells) or allogeneic (donor-derived). They are genetically engineered and expanded ex vivo, then infused into the patient, where they recognise their cognate antigen, bind to it, and proliferate. Early clinical studies with CAR-T cells used anti-CD19, which is expressed on B cell acute lymphoblastic leukaemia (ALL) and on B cell lymphomas. Other CAR-T cell targets include CD4, BCMA, and CD33 (June et al. 2018). CAR-T treatment must be individualised, making the approach inherently costly. As discussed above, many tumours express the immune checkpoint HLA-G. CAR-T cells directed against HLA-G are under clinical investigation against a number of solid tumours. A substantial “bystander effect” means that not all cells in the tumour need to express HLA-G. NK cells, including chimeric NK cells have also been investigated as cellular immunotherapy. In general, CAR-T therapy has shown more activity against leukaemias and lymphomas than against solid tumours. This could be attributed to poor tumour penetration, though preclinical studies have shown that T cells can penetrate solid tumours. It has been suggested that the tumour microenvironment (TME) may have a locally immunosuppressive effect (Al-Hosseini et al. 2021; Jenkins et al. 2021). This could be caused by immunosuppressive cytokines or nutrient depletion. Intratumoral IL-12 delivery activated CAR-T cells (targeted against EGF receptor variant III) in a mouse glioma (Agliardi et al. 2021). T cells rely primarily on mitochondrial oxidative phosphorylation as their energy source, unlike tumour cells that primarily utilise glycolysis, so it is possible that agents that modulate energy metabolism (such as metformin) could enhance the activity of CAR-T cells in solid tumours. No form of cancer treatment is entirely free from potential risks, and the commonest adverse effects of immunotherapy result from perturbing the complex regulation of the immune system. Autoimmune symptoms may result, or exacerbation of existing autoimmune disease. Chronic infectious diseases, such as tuberculosis, that are maintained in a dormant state by the immune system, may show flare-ups (Okwundu et al. 2021).

10.8

Modelling Antitumour Immunity

The complex, highly interactive regulation of the immune system complicates the process of modelling, but makes the system impossible to understand without modelling. Human tumours that are highly immunogenic are usually eliminated early by the process of immune surveillance, unless they originate in a privileged site (as with kidney tumours). The rare but well-documented cases of spontaneous remission of established tumours probably represent examples of highly immunogenic tumours that have evaded immune surveillance. For moderately immunogenic tumours in non-privileged sites, their elimination or progression will depend upon

10.8

Modelling Antitumour Immunity

215

the balance between tumour cell proliferation and immune killing of tumour cells. This balance will vary between individuals, depending upon such factors as MHC haplotype. The balance may be changed by immunomodulators. Some of these may up-regulate the immune system generally (e.g. macrophage activators) and others may have a more specific effect by stimulating T cell proliferation or by making particular tumours more immunogenic. Killing of tumour cells by the immune system is selective: it requires recognition of a specific epitope on the target tumour cell by a member of a specific T cell clone. An antitumour immune response may involve multiple epitopes. Antitumour immunity follows zero-order kill kinetics with respect to tumour cells. Killing of the target requires physical contact between a target tumour cell and a T cell, so kinetics are second-order, but the affinity of the interaction is so high that the T cell’s binding site will be effectively saturated. Unlike chemotherapy, the kinetics of immune killing of tumour cells are such that a particular “dose” of cytotoxic T cells will kill a given number of tumour cells, rather than a given fraction. A consequence of the kinetics of immune cell killing is that immune stimulants will be most effective in situations of low disease burden. This may occur (a) in prophylactic treatment where sub-clinical disease may be treated with tumour vaccines; (b) in prevention or minimisation of metastasis from inoperable primary tumours. Since minimal residual disease following chemotherapy is often enriched with drug-resistant cells, immune stimulation may be an effective means of preventing or delaying acquired drug resistance. Advanced tumours may downregulate their immunogenicity, resulting in acquired resistance to immune stimulants. The kinetics of immune killing mean that combining immunotherapy with chemotherapy will have some similarity to combination chemotherapy, but some differences. The ability to model many of the critical factors determining tumour response to immunomodulators makes possible the use of virtual clinical trials for optimising clinical trial design. A sensitised T cell expands into a clone by the normal process of cell division. This cell division is triggered by recognition of the cognate epitope by the T cell receptor (TCR) but also requires a second signal (from class I MHC). Certain sensitised T cells act as memory cells because their rate of spontaneous apoptosis is close to zero. Memory cells may require occasional booster signals from antigenpresenting cells. In general, immune killing will not distinguish between drugresistant tumour cells and the drug-sensitive cells from which they were derived, unless the resistant cells have accumulated additional mutations that affect their immunogenicity. Mutations that make a tumour cell sub-population less immunogenic will result in resistance to immune stimulants. These mutations can be modelled in the same way as drug resistance mutations. The dynamics of the antitumour immune response may be summarised as follows. – It is a low-capacity high-affinity system. The coupling of a high-capacity, low-affinity system with a low-capacity, high-affinity system is a frequent motif in nature. It is efficient and versatile. The low-affinity system of antitumour

216

– – – –

10

Evading the Antitumour Immune Response

immunity is probably represented by macrophages and NK cells. This system is fast-acting, but rather non-specific. The kinetics of tumour cell killing by T cells are such that a given number of effector T cells will kill a given number of tumour cells, not a given fraction as with drugs. Tumour cells stimulate the proliferation of the cognate population of effector T cells, and these T cells kill the tumour cells that stimulated them—a negative feedback loop. Expansion of a particular T cell clone is self-limiting—another negative feedback. Tumour progression tends to result in increasing antigenicity as a result of genetic instability.

Wodarz and Komarova (2014) explored a model that described the interactions between a tumour cell population and a tumour-specific immune response. CTL activation by dendritic cells and the ability of the tumour to counter the immune response (e.g. by downregulating immunogenic epitopes) gave rise to three possible outcomes, clearance of the tumour, tumour persistence at a stable equilibrium, and freedom of the tumour from immune control, leading to continued tumour growth. The model was used to explore immunotherapy approaches, including dendritic cell vaccination, which, in a situation where the immune system had become tolerant to the tumour, was able (temporarily) to break the tolerance. It was concluded that vaccination would need to be repeated until the last tumour cell had been eliminated. The VIP model described in the online supplement to Chap. 11 can model tumour immunogenicity. Evidence from take rates with transplanted mouse tumours showed that an immunocompetent mouse can eliminate about 105 allogeneic tumour cells. By contrast, a single cell of a syngeneic tumour transplanted into an inbred mouse can give rise to a lethal tumour. Scaling the mouse value of 105 allogeneic tumour cells to human body weight suggests that the immune system could eliminate about 3.5 × 108 highly immunogenic tumour cells. For one sensitised CD8 cell to reach this level would require 28 doublings. Since an autochthonous tumour takes more than 28 days to develop, the model assumes that patients with established tumours are fully sensitised. The VIP program was used to examine the effect of immunogenicity on tumour growth. Figure 10.1 compares the growth of a highly immunogenic tumour with an otherwise identical tumour that evoked only a weak immune response. The immune response was able to eliminate up to 80,000 cells of the highly immunogenic tumour: in this system, survival of >365 days indicates that no tumour cells survived. In contrast, the weakly immunogenic tumour was able to survive and grow to a lifethreatening tumour so long as more than 300 cells were present. Above 80,000 cells, growth of the two tumours was indistinguishable. The result of Fig. 10.1 supports the conclusion of earlier studies that immunotherapy is most effective when employed in situations of minimal residual disease. Figure 10.2 illustrates the output of a simulation in which the VIP program modelled the treatment of a weakly immunogenic prostate tumour with daily

10.8

Modelling Antitumour Immunity

217

Fig. 10.1 Effect of immunogenicity on tumour growth. IMF = 1.0: highly immunogenic; IMF = 0.02 slightly immunogenic

Fig. 10.2 Effect of abiraterone (15 mg/kg, daily) on a tumour with low immunogenicity

abiraterone. Within three weeks, 99% of the tumour cells were eliminated, but the remaining 1% of tumour cells were abiraterone-resistant (tumour/R1 in Fig. 10.2) and within 10 months, despite continued treatment had grown to a lethal tumour mass. In a similar simulation, summarised in Fig. 10.3, the abiraterone was supplemented with an immunostimulant. Once again, 99% of the tumour was eliminated within three weeks, and once again, the surviving cells at three weeks were abiraterone-resistant. However, the resistant tumour cells succumbed to the stimulated immune response, and by 39 days our virtual patient was tumour-free. The VIP program was also used to model the phenomenon of immune-induced tumour dormancy: depending on the degree of immunogenicity of a tumour, sub-curative treatment may result in a prolonged period of tumour stasis. Although this is often described as tumour dormancy, modelling suggests that this may not be

218

10

Evading the Antitumour Immune Response

Fig. 10.3 Effect of abiraterone plus an immunostimulant on a tumour with low immunogenicity

Fig. 10.4 Dormancy of an immunogenic tumour during 30-day treatment with an immunostimulant

true dormancy (which implies that all cells are quiescent) but may be a balance between tumour cell proliferation and immune cytotoxicity. In Fig. 10.4, 30-day treatment with an immune stimulant (as modelled by the VIP program) maintained stasis, but after cessation of treatment the tumour resumed growth. The VIP program was used to model the effect of inoculum size on growth of an immunogenic mouse tumour. In these simulations, the immune system was assumed to be previously sensitised to the tumour, i.e. memory cells were present. This models the situation of a spontaneous tumour, where the immune response is triggered as soon as “non-self” cells are detected. Table 10.1 shows the dependence of the survival of mice inoculated with cells of an immunogenic tumour on inoculum size. One thousand cells were eliminated by the host in three days, a situation described by experimental oncologists as a “no take”. Increasing the inoculum to

10.8

Modelling Antitumour Immunity

Table 10.1 Dependence of survival time on inoculum size of a transplanted mouse tumour

Inoculum (cells) 100 1000 2000 10,000 68,000 70,000 100,000 140,000 150,000 200,000 1000,000

219 Survival time (days)

Other outcome Cure at 1 day Cure at 3 days

35 32 39 Cure at 15 days Cure at 10 days Cure at 12 days 40 32 26

Tumour cell cycle time was 24 hours. Lethal tumour burden assumed to be 1 × 109 cells

2000 cells resulted in a lethal tumour burden at 35 days. A somewhat larger tumour inoculum, in the range 70,000 to 140,000 cells, again resulted in “no takes”. Inocula of 150,000 cells and above were able to outgrow the antitumour immune response and resulted in lethal tumour growth (Jackson et al. 2015). This paradoxical situation, where a small tumour inoculum is able to evade the immune response, but a larger tumour succumbs, has been described experimentally and is attributed to the fact that the larger inoculum results in a stronger and faster immune response. It is probably an artefact of transplanted tumours, since autochthonous human tumours start with a single cell, and only tumours that have outgrown the initial immune response will produce a clinically evident growth. The VIP program was also used to model the effect of the biological response modifier, levamisole, on a transplanted mouse tumour. Daily treatment was able to eliminate small tumour inocula, up to 50,000 cells. Increasing the tumour inoculum greatly attenuated the effect, and at inocula above 300,000 cells the levamisole was without effect. The clinical implication of this situation is that immune stimulants are most usefully employed in situations of minimal residual disease. An immunomodulator may give selective suppression of metastatic cells. If metastatic potential is assumed to result from a single somatic mutation, the appearance of metastatic cells may be modelled by the VIP program as for drug-resistant cells. The metastatic cells may be more accessible to the immune system than the primary tumour, and thus apparently more immunogenic (Melero et al. 2014). In this situation, modelling the unassisted immune response predicted elimination of metastatic cells even though there was very little effect on the primary tumour. In the same study, levamisole had a proportionately greater effect on metastatic cells than on growth of the primary tumour (Jackson et al. 2015). When combining immunotherapy with chemotherapy, it must be remembered that many chemotherapy drugs are immunosuppressive. Modelling has been used to predict the optimal balance of chemotherapy and immunotherapy (Park et al. 2019). Mathematical models of immuno-oncology can be used to identify predictive

220

10

Evading the Antitumour Immune Response

biomarkers, optimal dosing schedules, and rational combinations (Sancho-Araiz et al. 2021a, b). The VIP program is able to model up to four toxic side effects of drugs, and T cell proliferation can be one of these (the others usually include myelosuppression and intestinal crypt stem cell toxicity). Modelling can thus be used to explore optimal combination schedules that can minimise immunosuppression. The role of the innate immune system in eliminating tumour cells is less well understood. Part of the innate immune system is NK cell activation, and, as discussed above, NK cells may have activity against leukaemia cells and circulating tumour cells, but not against solid tumours. Some, but not all, tumours evoke an inflammatory response. Inflammation is a chronic activation of the innate immune response, mediated by macrophages and neutrophils, which are very effective at eliminating bacterial or fungal pathogens. The cytokines released by the innate immune response, tumour necrosis factor (TNF) and interferons, have been tested in cancer patients: they have rather marginal anticancer activity only at levels that cause severe host toxicity. Macrophages also release IL1 and IL2, but this is part of the feed-forward from innate to adaptive immunity. Following an infection, the innate immune system, the body’s first line of defence, generally switches off after a few days, which prevents serious toxicity, by which time the adaptive immune system has been triggered. When the innate system does not turn off, either because of chronic stimulation or because the negative feedback signal (e.g. by IL10) has been lost, inflammation will result. IL10 signalling is epigenetically regulated and may sometimes be inappropriately switched off. This raises the dilemma: is it desirable to treat inflammatory tumours with anti-inflammatory drugs or not? On the one hand, inflammatory cells may have some modest tumour cell killing effect, on the other hand, some tumours may be stimulated by inflammatory cytokines. There are drugs and antibodies that can stimulate the innate immune system, but they must be used with great caution, because over-activation of the system (“cytokine storm”) is extremely toxic and frequently lethal.

References Agliardi G, Liuzzi AR, Hotblack A et al (2021) Intratumoral IL-12 delivery empowers CAR-T cell immunotherapy in a pre-clinical model of glioblastoma. Nat Commun 12:444. https://doi.org/ 10.1038/s41467-020-20599-x Alboncini L, Benvenuto M, Focaccetti C et al (2020) PlGF immunological impact during pregnancy. Int J Mol Sci 21:8714. https://doi.org/10.3390/ijms21228714 Al-Hosseini RSM, Halpin JC, Moffaei M et al (2021) Metabolic and mitochondrial functioning in chimeric antigan receptor (CAR)-T cells. Cancers (Basel) 13:1229. https://doi.org/10.3390/ cancers13061229 Altvater B, Kailayangiri S, Pérez Lanuza LF et al (2021) HLA-G and HLA-E immune checkpoints are widely expressed in Ewing sarcoma but have limited functional impact on the effector functions of antigen-specific CAR-T cells. Cancers (Basel) 13:2857. https://doi.org/10.3390/ cancers13122857

References

221

Anna F, Elodie BR, LeMaoult J et al (2021) First immunotherapeutic CAR-T cells against the immune checkpoint protein HLA-G. J Immunother Cancer 9:e001998. https://doi.org/10.1136/ jitc-2020-001998 BioNTech’s second act: can it transform the fight against cancer? | Financial Times (ft.com) (2022) Burnet M (1957) Cancer – biological approach : I. The processes of control. II. The significance of somatic mutation. Br Med J 1(5022):779 Fabian KP, Hodge JW (2021) The emerging role of off-the-shelf engineered natural killer cells in targeted cancer immunotherapy. Mol Ther Oncol. 23:266–276 Garcia M, Palma MB, Verine J et al (2020) The immune checkpoint HLA-G/ILT4 is involved in the regulation of VEGF expression in clear cell renal carcinoma. BMC Cancer 20:624. https://doi. org/10.1186/s12885-020-07113-8 Hanahan D, Weinberg RA (2011) Hallmarks of cancer: the next generation. Cell 144:646–674 Harrington K (2022) Phase 1 study of Herpes RP2. 2022 ESMO Congress, Paris, September 2022 Jackson RC, Radivoyevitch T (2013) Modelling c-Abl signaling in activated neutrophils: the antiinflammatory effect of seliciclib. Biodiscovery 2013(7):4–11 Jackson RC, Fernandez E, Radivoyevitch T (2015) Bayesian systems for optimizing treatment protocols in oncology. In: Ur-Rahman A (ed) Frontiers in clinical drug research – anticancer agents. Bentham Science Publishers, pp 50–119 Jan C-I, Huang S-W, Canoll P et al (2021) Targeting human leukocyte antigen G with chimeric antigen receptors of natural killer cells convert immunosuppression to ablate solid tumors. J Immunother Cancer 9:e003050. https://doi.org/10.1136/jitc-2021-003050 Jenkins Y, Zabkiewicz J, Ottman O, Jones N (2021) Tinkering under the hood: metabolic optimization of CAR-T cell therapy. Antibodies (Basel) 10:17. https://doi.org/10.3390/antib10020017 June CH, O’Connor RS, Kawelekar OU et al (2018) CAR-T cell immunotherapy for human cancer. Science 359:1361–1365 Klein A (2022) Time to get personal. New Scientist 25(6/22):38–41 Komohara Y, Harada M, Ishihara Y et al (2007) HLA-G as a target molecule in specific immunotherapy againsy renal cell carcinoma. Oncol Rep 18:1463–1468 Lawton G (2021, April 3) Your first line of defence. New Scientist 2021:40–44 Melero I, Rouzaut A, Motz GT et al (2014) T cell and NK cell infiltration into solid tumors: a key limiting factor for efficacious cancer immunotherapy. Cancer Discov 4:522. https://doi.org/10. 1158/2159-8290.CD-13-0985 Miller RG (1987) Immunology related to cancer. In: Tannock IF, Hill RP (eds) The basic science of oncology. Pergamon Press, New York, pp 223–233 Okwundu N, Grossman D, Hu-Lieskovan S et al (2021) The dark side of immunotherapy. Ann Transl Med 9:1041. https://doi.org/10.21037/atm-20-4750 Park DS, Robertson-Tessi M, Luddy KA et al (2019) The Goldilocks window of personalized chemotherapy: getting the immune response just right. Cancer Res 79:5302–5315 Sancho-Araiz A, Mangas-Sanjuan V, Troconiz IF (2021a) The role of mathematical models in immuno-oncology: challenges and future perspectives. Pharmaceutics 13:1016. https://doi.org/ 10.3390/pharmaceutics13071016 Sancho-Araiz A, Zalba S, Garrido MJ et al (2021b) Semi-mechanistic model for the antitumor response of a combination cocktail of immuno-modulators in non-inflamed (cold) tumors. Cancers (Basel) 13:5049. https://doi.org/10.3390/cancers13205049 Siu LL, Wang D, Hilton J et al (2022) First-in-class anti-immunoglobulin-like transcript 4 myeloidspecific antibody MK-4830 abrogates a PD-1 resistance mechanism in patients with advanced solid tumors. Clin Cancer Res 28:57–70 WHO (2022) The Moderna COVID-19 (mRNA-1273) vaccine: what you need to know (who.int) Wodarz D, Komarova NL (2014) Dynamics of cancer: mathematical foundations of oncology. World Scientific, Singapore, pp 431–457

Chapter 11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

Abstract Understanding the evolutionary dynamics of malignancy has implications for prevention, diagnosis, and treatment of cancer. Cancer incidence can be partially prevented by minimising exposure to radiation and chemical carcinogens, but there is a baseline amount of spontaneous chromosomal damage and genetic mutation, meaning that cancer can never be completely prevented. Although it may never be possible to prevent cancer initiation, many of the deadly effects could be prevented or delayed by inhibiting tumour progression. The evolutionary dynamics of malignancy suggest that personalised cancer treatment is unlikely to be widely successful because widespread and ever-increasing heterogeneity make it almost impossible to eradicate all cells. Cancer diagnosis is being revolutionised by molecular diagnostics to identify mutations associated with cancer. In particular, tumour DNA released into the bloodstream is making possible diagnosis of tumours that are too small to cause overt symptoms. Appropriate design of drug combinations can delay or minimise the development of drug resistance. Early diagnosis raises the possibility of developing treatments that keep the tumour at an early stage. Approaches to preventing tumour progression include antimetastatics, drugs to prevent over-ride of the mitotic spindle assembly checkpoint, and immune checkpoint inhibitors. Although cancer vaccine development is at an early stage, it has the potential to stimulate the immune destruction of asymptomatic, early-stage tumours, before they have the opportunity to progress.

The slightest advantage in one being, at any age or during any season, over those with which it comes into competition, or better adaptation in however slight a degree to the surrounding physical conditions, will turn the balance. Charles Darwin, On the Origin of Species (1859).

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/978-3-031-32573-1_11. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_11

223

224

11.1

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

Current Views of the Causes of Cancer

Chemical carcinogens, ionising radiation, ultraviolet light, and certain viruses are all known to cause DNA damage. Do these environmental factors cause all cancers, or are there sources of spontaneous mutation? Deamination of DNA cytosine to uracil will result in miscoding. Oxidation of DNA guanine bases (which base-pair with cytosine) to 8-oxo-guanine (which can base-pair with thymine) will also cause a point mutation. One theory of the origin of life suggests that RNA preceded DNA in evolutionary history as the genetic material, but the relatively frequent deamination of cytosine (on evolutionary time scales) makes RNA unstable. The cell cannot distinguish a uracil that is supposed to be there from a uracil that is meant to be a cytosine. In DNA, uracil has been replaced by its methylated analogue, thymine, so a uracil base in DNA is definitely out of place. The cell has evolved an enzyme, uracil DNA glycosylase, to excise these bases, so that they can be replaced by thymine. Similarly, mechanisms have evolved for recognising and removing 8-oxo-guanine. However, there may have been a period in evolutionary history, after life evolved DNA, but before mechanisms for removing 8-oxo-guanine, when background mutation rates were much higher than at present. Some evolutionary biologists have claimed that the 3.7 billion years since life appeared on earth is not sufficient time for the number of mutations needed for the evolution of modern species to have occurred. This had led some biologists to postulate an extraterrestrial origin of life. If the background mutation rate was much higher in the past, there is no need to look elsewhere for the origin of terrestrial life. There is a further source of spontaneous mutation that does not involve chemical damage to DNA bases, which is that the enzymes of DNA biosynthesis, the DNA polymerases, have a theoretical limit to their selectivity. In the growing DNA chain, the polymerase will insert a cytosine opposite a guanine, and vice versa; it will insert a thymine opposite a cytosine, and vice versa. About once in every ten billion base pairs, it will use a thymine instead of a cytosine, or an adenine instead of a guanine (or vice versa). The thermodynamics of the reaction mean that perfect selectivity would require an infinite free energy change. As pointed out by Francis Crick (1988) the canonical base pairing of DNA base pairs, in which A forms two hydrogen bonds with T, and G forms three hydrogen bonds with C assumes that the bases are in their predominant tautomeric forms. If, for example, a particular guanine base happened to be in its alternative tautomeric form at a crucial moment in DNA replication, it would only be able to form two hydrogen bonds, so might base-pair with T instead of C. This means that replicating the 6 × 1010 bases of a human cell is likely to result in several spontaneous errors every time a cell divides. Since about 98% of human DNA is non-coding “junk” DNA, most of these mutations will have no effect. However, even in the complete absence of ionising radiation or chemical carcinogens, a background incidence of cancer is unavoidable. “Proof-reading” mechanisms, such as DNA mismatch repair, have evolved to identify and correct coding errors, but these, too, cannot be 100% effective. It seems likely that reducing the level of environmental carcinogens to zero would eliminate most cancers, but that an

11.2

Implications of Evolutionary Dynamics for Diagnosis

225

irreducible minimum of spontaneous tumours would remain. At present we do not know what that irreducible minimum is. Guanine oxidation depends upon the level of ROS, which, in turn, is influenced by environmental factors, so apparently spontaneous sources of carcinogenesis may in fact be environmentally determined. The unresolved debate about whether antioxidants reduce cancer incidence illustrates the complexity of the issue.

11.2

Implications of Evolutionary Dynamics for Diagnosis

The last decade has seen unprecedented growth in our knowledge of tumour genomics. Thousands of tumour-associated mutations have been described, mostly involving growth factor receptors and components of their associated signalling pathways and transcription factors. What is the diagnostic significance of these mutations? Tumour genetics is complex: there is no single mutation that causes transformation, rather a multiplicity of mutation patterns appears to drive an evolutionary process that leads from a cancer stem cell to an advanced metastatic tumour. New analytical techniques have made possible identification of tumourassociated mutations in small blood samples, either from circulating tumour cells or from DNA released into the blood. Early tests of this type were intended to detect particular cancers, but since many cancer-causing mutations are found in many tumour types, the focus has shifted to multi-cancer tests . A large-scale study, the PATHFINDER study (Nadauld et al. 2021) examined cell-free DNA from over 6600 subjects aged over 50. Of these, 92 were positive for cancer-associated mutations, and further examination found cancers or leukaemias in 35 of these. Some of the tumours found were types for which there is no existing screen. If the DNA sequencing is supplemented with DNA methylation analysis, it is often possible to possible to identify the tissue of origin of the tumour (Klein et al. 2021, 2022). This test, marketed under the name “Galleri test”, is now under evaluation in 165,000 subjects. However, the detection of oncogenic mutations in a person’s blood does not necessarily mean that that individual is destined to develop clinical cancer: the nascent lesion may spontaneously abort, or it may be destroyed by the host’s immune system. This technology has the potential to transform the early diagnosis of cancer, but at present it is not clear how (or if) to treat asymptomatic subjects who are shown to carry cancer-associated mutations. Agents that can prevent tumour progression will need to be developed. Cancer vaccines will also have a role to play. In parallel with these developments in tumour genetics, it has become clear that whether a normal cell can develop into a tumour is not determined by mutations alone. The microenvironment determines whether a cancer stem cell, a cell that has accumulated the complement of driver mutations that can result in transformation, will survive or die, whether it will remain as a local lesion or become a lifethreatening metastatic tumour. These environmental factors include the oxygen supply, the metabolic status of the cell and the surrounding tissues, and the interaction of the transformed cell with the immune system. It is estimated that 1 of every

226

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

600,000 cells in 50-year-old lung contains cancer-causing mutations, but these cells remain dormant unless stimulated. Micro-particulate air pollution, by stimulating IL-1β may trigger these dormant cells into proliferation (Pich et al. 2022). In cancer diagnosis, as in all complex relationships, correlation does not necessarily imply causation. Nevertheless, even tests for factors that do not have unequivocal causal status may have diagnostic value. An example is the use of the prostate-specific antigen (PSA) in diagnosis of prostate cancer. Elevated PSA may be the result of a prostate infection, or of benign prostate hypertrophy, as well as prostate carcinoma, but it is a risk factor that can influence a decision to carry out more definitive diagnostic tests. As well as the absolute level of PSA, the direction of change with time has diagnostic value. In the same way, detection of mutant K-ras in blood is not, in itself, a diagnosis of cancer, but it is certainly a risk factor. Although no single mutation provides a definitive diagnosis, the growing database of combinations of driver mutations, with associated outcomes data, is leading to a revolution in early diagnosis. For prediction of outcomes, it is not necessary to determine the quantitative effect of a mutation—a qualitative prediction of whether a mutation is positive (driving progression), negative, or neutral may suffice. Where the effect of a mutation is context-dependent, the database of outcomes may provide a clue. Evolutionary dynamics, by providing an approach to predicting the possible consequences of particular mutation patterns—spontaneous regression, stasis, or progression, will be a tool for evaluating the significance of molecular diagnostic data.

11.3

Implications of Evolutionary Dynamics for Cancer Treatment

The discussion of drug resistance in Chap. 6 gave one important example of how evolutionary dynamics has guided cancer treatment. Advanced tumours may require treatment with multiple drugs (a fact that had already been established empirically) and, less obviously, for maximum cell kill the component drugs of the combination should be used simultaneously, not sequentially. This in turn implies that the component drugs should have non-overlapping host toxicity, which usually requires that they have different molecular targets. The other obvious requirement is that component drugs of the combination should be non-cross-resistant. For some mechanisms of resistance, such as target enzyme overproduction or mutations at the target active site, different sites of action guarantee non-cross-resistance, but when drug resistance is caused by efflux pumps or by loss of membrane transport carriers, crossresistance between chemically and mechanistically unrelated drugs is common. Evolutionary approaches have been suggested to slow the development of drug resistance, such as using low doses of verapamil or 2-deoxyglucose to increase the energetic cost of resistance (Silva et al. 2012). If sufficient information is available on the population PK and PD of the component drugs in a combination, it may be possible to devise combination regimens

11.3

Implications of Evolutionary Dynamics for Cancer Treatment

227

that alternate the component drugs on a timescale that can be almost as effective as simultaneous combinations, while minimising the risk of overlapping toxicity. To optimise such regimens will require combining evolutionary dynamics modelling (to predict the emergence of drug resistance) with population PK/PD modelling. The concept of synthetic lethality (Kailin 2009) describes the situation where alternative biochemical pathways exist, such that loss of one or other of them may not have serious consequences, but loss of both will be lethal. In such cases, combining inhibitors of the two pathways may be a very effective way of killing tumour cells, though probably at the cost of serious host toxicity. A more useful, and selective, application of the concept is to identify tumours that have a genetic deficiency of one pathway, then treat it with an inhibitor of the other pathway. The treatment of BRCA1-deficient tumours with a PARP inhibitor is an example of this. The most important implication of evolutionary dynamics for cancer treatment is that early-stage tumours are much more likely to be curable than large, established tumours. Surgeons were of course well aware of this without the benefit of evolutionary dynamics; they are able to perform curative surgery before a tumour has metastasized, but metastatic tumours will usually require that surgery be followed by radiotherapy or chemotherapy. Invasion, metastasis, and drug resistance may be the most obvious aspects of tumour progression, but the loss of cell cycle checkpoints, the ability to tolerate hypoxia, or the ability to evade immune surveillance have equally grave implications for treatment outcomes. Tumour size may be regarded as a surrogate for heterogeneity, and tumour heterogeneity will always make a tumour harder to treat and limit the potential cure rate. Chapter 6 compared treatment outcomes of testicular and ovarian cancer as an example of this. Despite their similar origins in germinal epithelium, there are many differences between the two tumours, but these differences are often a direct consequence of tumour progression. Tumours that are usually diagnosed late in the progress of the disease, such as ovarian and pancreatic carcinomas, are notoriously hard to treat. We are more likely to make progress against such tumours by focussing resources on screening and early diagnosis. Tumour heterogeneity is a consequence of genetic and chromosomal instability. Consider the implications of tumour heterogeneity for cancer treatment. Accepting Hanahan and Weinberg’s six hallmarks of cancer, each of which may be driven by multiple alternative mutations, gives us an intimidating perspective: the full-blown malignant phenotype of advanced cancer is the result of at least a six-stage process. Of these six or more stages, the first—independence of external growth stimuli—is obligatory. Subsequent steps may occur in any order, e.g. some tumours become metastatic early, others late. Each of these steps may be driven by any of at least 20 oncogenes, so the number of potential tumour phenotypes is about 206, over 6 × 107 types of tumour. The oft-repeated saying that “cancer is more than 100 diseases” was based on histopathological criteria. Drug sensitivity, however, depends upon biochemical factors. This means that cancer is 60 million different diseases, and the optimal treatment for each will be different. Clearly, we cannot conduct 60 million clinical trials, so how are we to proceed? One approach that has been adopted is to use clustering analysis to reduce the huge number of different tumour

228

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

genotypes to a manageable number of related clusters. If all the cases in a particular cluster are treated with a particular protocol, the treatment for each particular individual within a cluster may not be optimal, but is likely to be better than basing treatment upon its histological type. One particular lung tumour may resemble a kidney tumour with a similar mutation profile more closely than it resembles other lung tumours. The clustering approach is sometimes dismissed as arbitrary—at one extreme, each individual tumour may be considered as its own cluster; at the other extreme, all transformed cells could be regarded as a single cluster. What criteria should be used to determine the appropriate number of clusters? This same criticism has been addressed by biochemical taxonomists. Their position is that taxonomic clusters reflect objective divergence in evolutionary history. As we have seen, tumour progression is also an evolutionary process, and it should be possible to use the mutation profile of advanced tumours to reconstruct their evolutionary history and assign them to a cluster that optimises their treatment. The treatment of cancers according to their mutation profile holds promise, but is still at an early stage. The ability to factor mutation profiles into optimal treatment will require a database of outcomes, and for the less common cancer genotypes it may take a very long time to assemble this. In such cases, in silico clinical trial modelling may help to reduce the very large number of possible trial designs to a manageable number for clinical evaluation. The extreme heterogeneity of tumours places severe constraints on the usefulness of personalised medicine (Gillies et al. 2012). Beckman et al. (2012) used modelling of over 3 million virtual “patients” to compare standard personalised medicine strategies (in which the best drug against sensitive cells was used until relapse, followed by a switch to a non-cross-resistant agent) with an evolutionary dynamicsbased approach that attempted to predict the appearance and overgrowth of resistant cells. They predicted that the evolutionary dynamics-based approach had the potential to produce significant therapeutic gains. Another approach to adaptive treatment was described by Gallaher et al. (2018) who showed that continuous fixed-dose treatment was effective in treating tumours consisting of mostly sensitive cells, but treatment regimens that incorporated dose modulation or treatment vacations gave better overall control when resistant cells were present. These adaptive approaches exploited the fact that resistant cells were often less well adapted to the local microenvironment than sensitive cells in the absence of drug. Large tumours must be assumed to contain drug-resistant cells and probably cells with pre-existing resistance to any two drugs. They have probably already metastasised at the time of diagnosis. Surgery and/or radiotherapy are normally used to minimise the primary tumour, and this will often be followed by adjuvant chemotherapy to treat the metastases. This improves survival in a number of tumour types; however, since the presence or absence of metastases generally cannot be detected at the time of surgery, this implies that some patients who are, in fact, cured by the initial treatment are exposed to potentially toxic drugs. This is being addressed by the development of more sensitive tumour markers, and molecular profiling of the tumour may identify those mutations that predispose a tumour to early metastasis.

11.3

Implications of Evolutionary Dynamics for Cancer Treatment

229

Sometimes patients with large tumours are given drug treatment before surgery, a strategy known as neoadjuvant chemotherapy. The rationale for this is that tumours may have diffuse, hard-to-detect infiltration of surrounding normal tissues, liable to be missed by surgery, and that shrinking the primary before surgery could result in more complete surgical removal. It is believed that the presence of drug-resistant cells should not compromise the outcome, since most of them should be removed at surgery, and the number of resistant cells in the part of the tumour that is not removed, while not reduced by the neoadjuvant chemotherapy, would not be increased by it. Conventional cancer treatment is based on the premise that the best treatment is the one that kills the most cancer cells (without unacceptable toxicity to normal tissues). This is true if the aim of treatment is curative, but once it is accepted that a particular patient has incurable disease, the most appropriate approach for that individual may be to maximise the time to progression (TTP). Evolutionary dynamics calculations show that this may sometimes be achieved, not by maximising cell kill, but by manipulating the environment so as to favour more slowly growing cells at the expense of more rapidly dividing ones. Anticancer drugs are part of the environment responsible for the selection process, and resistance to those drugs is achieved at a cost to the cell—in growth rate and in energy metabolism. If drugresistant tumour cells had a selective advantage over wild-type cells in the absence of drug, they would have become the wild-type. Adaptive therapy is an evolutionary dynamics-based approach that attempts to maximise TTP by manipulating the environment and the timing and frequency of drug exposure. The principles of the approach were worked out with single agent therapy, for example with antiandrogens used to treat prostate cancer, but adaptive regimens are being devised for combination treatment. Advanced prostate cancer is conventionally treated with continuous antiandrogen therapy, and adaptive regimens are necessarily more complex, and doses and treatment schedules must be individualised. This requires frequent clinical assessment, which is facilitated by the availability of a tumour biomarker, PSA assays in the case of prostate cancer. Optimal control strategies have been explored for treating advanced prostate cancer with abiraterone. Unlike conventional therapy, which seeks to maximise tumour cell kill by treating continually at the maximum tolerated dose until treatment failure, this study explored alternative objectives, such as maintaining a constant total tumour volume or minimising the fraction of drug-resistant cells (Cunningham et al. 2018). This study found that long-term control was achievable while providing acceptable patient quality of life. Another trend in cancer treatment whose objective is to improve palliative care of patients with advanced cancers is the “repurposing” movement (Sabale et al. 2020). This is the use of drugs that are marketed for unrelated therapeutic indications to prevent, or at least slow down, tumour progression. There is often preclinical data supporting efficacy of these agents, but they have not been tested in randomised clinical trials (with occasional exceptions, such as the use of aspirin as an antimetastatic; aspirin, as an antiinflammatory may slow tumour progression in those cases where it is driven by inflammatory cytokines). There is experimental

230

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

data showing that statins can inhibit Ras signalling, The rationale for several of the re-purposed agents is the change in metabolic regulation that is a universal aspect of tumour progression. The antitumour effect of metformin has already been discussed; metformin selectively reduces glucose uptake by tumour cells. The anthelmintic, mebendazole, has antiproliferative activity. The H2 antagonist, cimetidine (Tagamet®) is reported to have antiproliferative activity and to inhibit angiogenesis. The antifungal, itraconazole, also inhibits angiogenesis, as does the antibiotic, doxycycline. Because all these drugs are marketed for non-cancer indications, and they are well-tolerated, some oncologists are prepared to use them in the absence of controlled clinical trial data. One study is treating patients who have run out of other treatment options with a cocktail of metformin, mebendazole, doxycycline, and a statin in an uncontrolled trial. It is intended to compare survival data with historical values (Daily Telegraph 2015).

11.4

Implications of Evolutionary Dynamics for Tumour Prevention

Perhaps the first suggestion that evolutionary dynamics studies could suggest strategies for cancer prevention was the report by Gatenby et al. (2010) that tumour progression involved changes in the interaction of tumour cells with their environment, the normal epithelium. They pointed out that normal tissues were organised in a way that minimises cancer incidence. If cancer-causing mutations arise in cells of the epidermis, they will have no permanent effect because it is the fate of these cells to be sloughed. Only mutations in the stem cell population can cause tumours. For many cancers, it is likely that future research will focus on tumour prevention. Since, even in the absence of environmental carcinogens, mutations are inevitable, it may be more practical to try to detect tumours at a very early stage and prevent them from progressing. This will require new tests for early detection, and new treatments to prevent the later stages of tumour development. While not strictly “prevention”, this approach should certainly prevent, or reduce, the occurrence of clinically detectable cancers. Conventional cancer detection (for solid tumours) for the most part requires a tumour mass that can be detected either by physical examination or by imaging by mammography, CT scanning, or MRI. By the time the patient is experiencing symptoms, the tumour is likely to contain more than 109 cells and to be highly heterogeneous. At present a number of blood-based diagnostics are used, such as prostate-specific antigen (PSA) for prostatic cancer. The established blood-based diagnostics generally have limitations of sensitivity or selectivity that make them of limited use for mass screening of healthy subjects. Of the newer early detection technologies, the most promising is looking for cancer-linked mutations in plasma DNA (Oversoe et al. 2020; Al-Shaheri et al. 2021; Darrigues et al. 2021; Magbanua et al. 2021; Roosan et al. 2021). This exploits the enormous sensitivity and selectivity of polymerase chain reaction (PCR) assays. Cancer cells always have a low but

11.5

Manipulating the Environment to Reduce Cancer Incidence

231

non-zero turnover and release DNA into the blood. This technology is still in the exploratory stage, but has already become a valuable confirmatory technique. Extending it for very early diagnosis will present a number of questions that we are currently unable to answer. If a particular mutation is known to be cancerlinked—say a p53 mutation—is there likely to be a lower limit for it to be of diagnostic significance? It is possible that the process of immune surveillance is able to eliminate potential cancer cells, and only when their number exceeds the capacity of the immune system to remove them do these mutations become a threat. We believe this to be the case, because people with compromised immune function, such as organ transplant recipients, tend to have higher incidence of cancer than fully immunocompetent individuals. Certain cancer-linked mutations may only lead to tumours when other mutations or chromosomal rearrangements are present, as discussed in Chap. 3, so it may be necessary to look for multiple mutations. In cases when the blood-based DNA diagnostics make it certain that the body contains a number of transformed cells, there may be no way of telling where in the body those cells are located, so that follow-up examinations cannot be used to track tumour development. There are, however, recent studies showing that particular patterns of multiple mutations, or DNA methylation profiles, tend to be associated with particular tumour sites. When these sensitive diagnostics make it likely that a person will develop cancer, at some unknown organ site after what may be a considerable period of time (remember that the induction time for CML is about eight years), what is the oncologist able to do about it? Existing chemotherapy agents, which are often seriously toxic, would be hard to justify in a person with no symptoms. At present, the response to this situation is frequent check-ups, so that physical examination or repeated diagnostics will enable the growing tumour to be detected and treatment as early as possible. The obvious downside of this approach is that the tumour will inevitably be more heterogeneous by then. In this situation, drugs that prevent tumour progression, so long as they have minimal side effects, used as soon as the plasma DNA mutation profile shows that transformed cells are present, could radically reduce the incidence of cancer.

11.5

Manipulating the Environment to Reduce Cancer Incidence

Obviously the identification of carcinogens and their elimination from the environment is the most direct way to do this. There are environmental factors other than DNA-active chemicals that have been reviewed by epidemiologists for cancercausing potential. One of the most controversial is diet. Obesity is an established risk factor for some tumour types, but in the absence of a molecular mechanism, other than advising overweight individuals to go to the gym (and eat less) it is not clear how to intervene. There is evidence that caloric restriction lowers cancer

232

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

incidence in mice, and metformin, which mimics the metabolic effect of caloric restriction is associated with reduced cancer incidence in diabetic humans. There is early data suggesting that high levels of dietary folic acid promoted tumour growth in rats, but no indication that this applies to humans. Is folate normally a limiting factor for cell division? It certainly is in individuals with folate-deficiency anaemia. There is confusing and contradictory data on antioxidants, such as vitamin C and vitamin E, and this subject urgently needs clarification. In addition to the nutritional environment, the hormonal environment of tumours is a major factor in tumour progression. Antiandrogens have an established role in slowing the progression of prostate cancer, and there have been extensive trials with prophylactic tamoxifen in breast cancer.

11.6

Using Drugs to Inhibit Tumour Progression

Early experimental studies showed that cis-hydroxyproline (an inhibitor of extracellular matrix formation) was able to inhibit invasion (McAuslan et al. 1988). This interesting observation does not appear to have been extended to clinical trials, probably because at the time there were no diagnostic tests capable of detecting early tumours at the pre-invasive stage. With the advent of molecular diagnostics, the concept of anti-invasive agents is worth re-visiting. The demonstration that PAK (p21-activated) kinases are involved in cancer cell invasion (Whale et al. 2011) suggests a possible approach. If a tumour remains localised, it may be possible to cure it surgically, or if this is not possible or desirable, it may anyway be possible to live with it for a long time. This is often the case for human prostate tumours. It is when such tumours metastasize that they become life-threatening. A large, localised tumour will already be genetically diverse, and drugs targeted at metastatic cells could prevent or delay this aspect of tumour progression. Despite the wealth of preclinical research on antimetastatic drugs (Steeg and Theodorescu 2008; Chaffer and Weinberg 2011; Liew et al. 2020), there have been few clinical trials of antimetastatic agents. This has been partly because of the belief that if a tumour has not metastasized at the time of diagnosis it should be curable by surgery, so that antimetastatic drugs have nothing to add, whereas if a tumour has already metastasized then it is too late for an antimetastatic drug to be useful. For certain classes of antimetastatic agents, this argument may not apply: for example, if an unattached tumour cell has become able to survive because of constitutive expression of the cadherin signalling pathway, it may be susceptible to inhibition of kinases in that pathway. The other reason that drug developers have been reluctant to invest in clinical studies of antimetastatic agents is the long timelines to a clinical endpoint. The availability of new biomarkers (e.g. circulating tumour cells) may help to address this issue. One example of antimetastatic therapy that has had extensive clinical evaluation is the use of prophylactic aspirin. The rationale for this is that circulating tumour

11.6

Using Drugs to Inhibit Tumour Progression

233

cells, in order to invade tissues, attach themselves to the capillary wall in a fibrin clot, so that inhibiting clot formation should lead to fewer metastatic growths. Since the aneuploidy of many carcinomas appears to be the result of SAC override, compounds that restore SAC function might be worth exploring as inhibitors of tumour progression. Overexpression of aurora kinase A (AK-A) has been reported in many human tumours; in one report 60% of human breast tumours overexpressed AK-A. Potent, selective inhibitors of AK-A exist and have antitumour activity, but they do not appear to have been tested specifically as inhibitors of tumour progression. Similarly, NEK-2 overexpression also causes SAC over-ride, and NEK-2 is overexpressed in human tumours. Inhibitors of NEK-2 have antitumour activity in preclinical systems, but, again, have not been tested as progression inhibitors. As discussed in Chap. 7, inhibitors of Abl kinase, such as imatinib, inhibit progression of CML to the acute phase by decreasing ROS levels, which are mutagenic. The low incidence of acquired resistance to imatinib is evidence that mutation rates are decreased by this agent. This raises the possibility that progression of other tumours could be blocked by agents that prevent increase in genetic diversity. For those diseases that are epigenetically driven, such as CMML, it is possible that ascorbic acid could inhibit progression by activation of TET-2. There are numerous reports that dietary restriction has a cancer-preventive effect (Wu et al. 2013; Clifton et al. 2021; Ibrahim et al. 2021). The mechanism of this effect is unclear, but is widely believed to be related to the well-documented effect of caloric restriction on longevity. In C. elegans, the AMP-dependent kinase (AMP-K) homologue is required for glucose deprivation to give prolongation of lifespan. This is interesting, because the anti-diabetic agent, metformin, which is reported to have cancer-preventive activity in experimental systems, up-regulates AMP-K. AMP-K is a master switch, regulating glucose consumption, driving ERK and WNT5A signalling (via HNF4α) and acting as a negative regulator of androgen receptor expression. AMP-K phosphorylates raptor, which is a component of mTORC1 (Hardy and Ashford 2014). AMP-K is activated by both AMP and AICAR. It has been proposed that metformin may prevent tumour progression by blocking or inhibiting the Warburg effect (Christopherson 2017). Can AMP-K activation explain the antitumour activity of metformin? Is the mechanism related to that of dietary restriction? Does AMP-K downregulation confer a growth advantage? If so, is it due to an anti-apoptotic effect? Is there cross-talk between AMP-K signalling and ROS signalling? If so, could the activity of metformin be potentiated by antioxidants? By pro-oxidants (compounds that cause oxidative stress)? Are there biomarkers of the altered pattern of energy metabolism that might predict for metformin response? Is AMP-K downregulation an adaptive response to tumour hypoxia? If there is a relationship of AMP-K expression with hypoxia, what are the implications for combining metformin with radiotherapy? Whatever the answers to these questions, it is clear that tumours are able to adapt their metabolic status to their advantage. Hypoxia not only limits oxidative phosphorylation through oxygen depletion, but it also stimulates glycolysis by a change in the AMP/ATP ratio, activating phosphofructokinase. Hexokinase II is also activated by AMP-K.

234

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

Pyruvate kinase L (liver type) is phosphorylated and inactivated by AMP-K (muscle type, PK-M, is not affected). Pyruvate kinase M2 expression in tumour cells appears to be associated with aerobic glycolysis; PKM2 is allosterically activated by fructose 1,6-bisphosphate. PKM2 is oxidised and inactivated by ROS (Brimacombe et al. 2012). PKM2 in some tumours is upregulated by mutations in the Von Hippel-Lindau tumour suppressor gene. Note that high cytosolic AMP also activates protein kinase A (PKA, whose effects on carbohydrate metabolism antagonise those of AMP-K) and downregulates NfκB signalling. A possible explanation of aerobic glycolysis in tumours is that replacement of PK-M1 by PK-M2 maintains a high glycolytic rate even in presence of oxygen. PK-M2, i.e. the mostly inactive dimeric form occurs at high levels in some tumours, resulting in diversion of phospho-enolpyruvate into formation of metabolic precursors, rather than formation of pyruvate and lactate. Under these conditions, glutamine becomes the main energy source. Akt mobilises glucose transporters to the cell surface and activates hexokinase II (Kim and Dang 2006). Pyruvate decarboxylase kinase-1 (PDK1) phosphorylates and inactivates pyruvate decarboxylase, and thus reduces the supply of acetyl-CoA for TCA cycle activity. The TCA cycle feeds back to glycolysis through citrate inhibition of phosphofructokinase. How do damaged mitochondria cause cell proliferation? Seyfried et al. (2014) suggest that persistent damage to oxidative metabolism triggers an epigenetic signal; genes that respond include MYC, TOR, RAS, NfκB, and CHOP. The immune checkpoint inhibitors, anti-PD-1 and anti-PD-L1 were discussed in Chap. 10 as therapeutics: could they be used as tumour progression inhibitors? This implies chronic use which might make cost a limitation.

11.7

Artificial Intelligence Systems for Predicting Tumour Progression

As the database of cancer mutations grows, increasingly it will become possible to predict probable outcomes from particular mutation patterns. However, if progression to an advanced metastatic tumour requires a minimum of five mutations, and if there are at least twenty possible mutations that can cause each stage of tumour progression, there will be millions of possible cancer genotypes, and it will take a very long time to build a database capable of directly predicting outcomes and optimal treatments for each possible combination. The term “artificial intelligence” (AI) embraces a number of computational techniques, but often refers to deep learning, a method for extracting meaningful patterns from large, multivariate databases. With appropriate AI software, even though only a small fraction of the possible mutation profiles have been directly linked to particular outcomes, it should be possible to predict the course of development of occult tumours detected by mutation signatures discovered in DNA in blood from cancer patients and—of great significance for cancer prevention—of

11.8

In Silico Clinical Trial Modelling

235

asymptomatic individuals. The power of such databases will be increased if factors other than DNA profiles are included, such as lifestyle choices (e.g. smoking history), other test results (e.g. PSA), blood levels of nutrients (ascorbate, folates), and non-cancer-related medications (e.g. metformin). The database will grow with time, and the AI system will be Bayesian, so that it will learn with experience. A database of mutation signatures and other risk factors with an associated AI outcome prediction system would help resolve the dilemma of how, or if, to treat asymptomatic individuals who have cancer mutations detected in their blood. It could also be applied to clustering analysis of test results from cancer patients and make it more likely that they can be optimally treated.

11.8

In Silico Clinical Trial Modelling

The development of new pharmacodynamic biomarkers has greatly increased the information content of clinical trials and made possible the construction of PK/PD models (Jackson et al. 2015). A population PK/PD model, in conjunction with a disease model can then be used to simulate clinical trials in silico. Considering the great complexity of biological systems, and the inevitable simplifications that must be made in modelling biological systems, it may seem to a clinician or to an experimental oncologist almost pointless to attempt to predict the outcome of a clinical trial with a computer program. Indeed, no matter how well we understand the theory, clinical questions must be answered by an empirical trial. However, the design of a clinical trial always involves assumptions (based upon preclinical studies, and on prior clinical experience with the drug, or with related drugs) about what dose to use, how often, what the endpoint should be—in short, a model; no less a model for being a purely mental model. The advantage of making those assumptions mathematically explicit is that we can examine the predicted outcome of changing each of the variables, run the simulation hundreds of times, if necessary, with slight variations, get a feeling for which are the critical variables. In the models of treatment of tumours in mice that we have described, the modelling was deterministic: fixed values were assumed for the tumour growth parameters, and for the drug pharmacokinetic and pharmacodynamic parameters. This is a reasonable assumption to make when dealing with a transplanted tumour in inbred mice. Obviously, in a human population, with spontaneous tumours, there will be considerable variation in these parameters, and the model must reflect this. Population PK/PD models will give a measure of the patient-to-patient variability and make possible modelling of response rates. The problem is, population PK/PD parameter values generally only become available late in clinical development, when hundreds or even thousands of patients have been treated, and we want to use in silico clinical trials as a tool for early clinical development. At present, the best way to resolve this dilemma is to adopt a Bayesian approach: use the best available prior assumptions, or even guesses, in the beginning, and refine the model as more data becomes available.

236

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

In the case of oncology, the disease model must describe the cancer cell cycle and such aspects of tumour biology as growth, invasion, metastasis, angiogenesis, and interactions of tumour cells with the immune response. VIP is a modelling program that describes the effects of drugs on four normal cell populations and multiple tumour cell populations (Jackson et al. 2015). Growth parameters included in the model include cell cycle time, the quiescent cell fraction, and rate of spontaneous apoptosis. Information about the status of the various cell cycle checkpoints and signalling pathways can be included in the model, e.g. p53 (wild type or mutant), normal or mutant Ras, normal or mutant EGF receptor. Various treatment options can then be modelled, and the one predicted to be most active selected for clinical evaluation. Since, in practice, many of the tumour growth parameters will be unknown, a Bayesian approach is required: prior assumptions are made based upon preclinical data and historical precedent. The course of treatment, based on these prior assumptions, is modelled, and the difference fed back to drive model adjustments. The assumptions of the model are thus progressively refined—the system learns from experience. VIP can be used either as a deterministic model, in which growth parameters and PK/PD parameters are fixed, or it can treat doubling times, tumour cell count at diagnosis, and drug parameters as stochastic variables. For clinical trial modelling it is usually preferable to use stochastic mode. Table 11.1 shows population simulations where an advanced prostate tumour was treated with various combinations of the antiandrogen drugs abiraterone, enzalutamide, and the bicalutamide analogue, Catylix-03, three drugs that are mutually non-cross-resistant. Treatment was once daily; doses are in mg/kg/day and are at or near MTD. Survival in the absence of treatment was about four months, and each of the drugs, given as a single agent, was able to extend survival to about 12½ months. In each case the duration of response was limited by the emergence of acquired drug resistance. If treatment was switched at six months, about the time resistance started to appear, survival could be extended to 16 months (lines 5–7 of Table 11.1). Lines 8–10 show simultaneous combinations of two drugs; the dose of each drug was halved, so the total amounts of drug (relative to MTD) was constant. Simultaneous combinations of two drugs were not superior to the sequential use of the same two drugs, in fact they were slightly worse. Lines 11–16 of Table 11.1 show the calculated effects of sequential three-drug treatment; in each case, treatment was switched at 8 and 16 months. All the threedrug regimens were superior to two-drug regimens, with median survival now close to 2½ years. Line 17 shows the calculated effect of a simultaneous three-drug combination, in which each drug was given daily at 1/3 of its daily MTD. The response was markedly superior to sequential administration of the same drugs, at the same total dose, with 6/10 five-year survivors. Why is simultaneous treatment with three drugs predicted to be so much more effective than sequential treatment with the same drugs, when sequential and simultaneous two-drug regimens were approximately equivalent? The explanation is that an advanced tumour will almost certainly contain doubly-resistant cells (to any two drugs) at the start of treatment. However, the dynamics of the emergence and selection of drug-resistant mutants are such that triple mutants are unlikely to be

11.8

In Silico Clinical Trial Modelling

237

Table 11.1 Modelling combination treatment of prostate cancer with antiandrogens

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Abiraterone Dose Days 0 15 1–380 0 ” 0 ” 15 1–380 15 181–360 15 1–180 7.5 1–430 7.5 1–430 0 15 1–250 15 1–250 15 251–500 15 501–900 15 251–500 15 501–900 5.0 1–1826 15 q3da

Enzalutamide Dose Days 0 0 2.3 1–380 0 2.3 1–500 2.3 1–180 0 1.15 1–430 0 1.15 1–430 2.3 501–900 2.3 251–500 2.3 501–900 2.3 251–500 2.3 1–250 2.3 1–250 1.8 1–1826 2.3 q3db

Catylix-03 Dose Days 0 0 0 0.72 1–373 0 0 0.72 181–500 0 0.36 1–430 0.72 1–430 0.72 251–500 0.72 501–900 0.72 1–250 0.72 1–250 0.72 501–900 0.72 251–500 0.24 1–1826 0.72 q3dc

Median survival (days) 117 380 381 373 497 512 435 412 435 413 911 899 848 902 872 848 1826 1206

5-year survivors 0/10 0/10 0/10 0/10 0/10 0/10 0/10 1/10 0/10 0/10 0/10 0/10 0/10 0/10 0/10 0/10 6/10 5/10

a

Starting day 1 Starting day 2 c Starting day 3 b

present at the start of treatment, no matter how large the tumour. Sufficiently doseintensive simultaneous three-drug treatment will prevent their emergence, or at least delay the process for a long time. However, sequential treatment will allow for cells resistant to the third drug to appear during the time that the third drug is not present, so that by the time it is used, the seeds of resistance are already present (Fig. 11.1). Line 18 of Table 11.1 shows the calculated effect of alternating treatment in which the three drugs were used separately, at full doses, on successive days. The result was almost as effective as simultaneous treatment with 5/10 predicted fiveyear survivors, though the median survival was slightly shorter. Other simulations (not shown) explored less frequent periods of rotation between the three drugs: switching treatment weekly was less effective than daily rotation, and switching monthly less effective still, though still superior to the sequential regimen.

238

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

Fig. 11.1 Treatment of an established prostate tumour with abiraterone (15 mg/kg, d. 1–182), enzalutamide (2.3 mg/kg, d. 183–365), and Catylix-03 (0.72 mg/kg, d. 366 -). Overall survival time was 738 d. R1: resistance to abiraterone, R2: resistance to enzalutamide, R3:resistance to catylix-03 (from Jackson et al. 2015)

Kaplan-Meier plots provide a convenient means of visualising probability of survival as a function of time, and VIP can present data in this form. Figure 11.2 shows a Kaplan-Meier plot of VIP output for treatment of a large prostate tumour with abiraterone (15 mg/kg) on days 1–180, followed by enzalutamide (2.3 mg/kg). Figure 11.3 compares the result from using simultaneous abiraterone (7.5 mg/kg) and bicalutamide (0.36 mg/kg) daily for five years. Note the 4/10 five-year survivors. Clinical trials simulations have implications for both drug development strategy and for personalised medicine. An example of the former is modelling as an

11.8

In Silico Clinical Trial Modelling

239

Fig. 11.2 Kaplan-Meier plot of treatment of prostate cancer with successive abiraterone + enzalutamide

Fig. 11.3 Kaplan-Meier plot of treatment of prostate cancer with simultaneous abiraterone + enzalutamide

approach to optimising fixed ratio combination formulations, designed to minimise drug resistance. An example of the second is to use modelling to predict the correct time to switch to an alternative treatment in a patient who is beginning to relapse. Assuming that the first-line drug is the most effective, switching to a less effective (but non-cross-resistant) drug too soon will result in decreased response duration. Switching too late will increase the risk that cells resistant to the second drug are already present at the time of the switch. The clinical trial simulator, VIP (virtual interactive patient) is a model of this type. It is described in detail in the online supplement.

240

11.9

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

Clinical Endpoints: Progression-Free Survival

Overall survival is obviously the most important measure of drug effectiveness from the point of view of the patient, but it can have a complicated and unpredictable relationship to drug efficacy, depending on the exact site of the tumour, and how likely it is to invade a vital organ. For comparing drug protocols, progression-free survival (PFS) is a more consistent measure, and it is clinically relevant in that PFS is likely to confer a good quality of life. The VIP program, in its deterministic mode, will predict PFS as an output. Figure 11.4 shows calculated PFS for a prostate tumour treated with daily abiraterone (15 mg/kg). Progression of this tumour is assumed to be the result of a mutation that causes drug resistance. The figure shows TTP as a function of mutation rate. A two-log decrease in the mutation rate caused a 21% increase in TTP. However, a 4.8 log decrease totally prevented resistance. What is the cause of this discontinuity? The number of abiraterone-resistant cells present at the start of the simulation was calculated from the equation of Luria and Delbrück (eq. 6.1). For this particular tumour cell number, when the mutation rate was 1.2e-10 per cell division, there were no R1 cells present at the start of the simulation. When the mutation rate was 1.3e10, there was one R1 cell at the start of the simulation, and this eventually resulted in treatment failure.

11.10

Cytotoxicity and Cytostasis

In treatment of infectious disease, antibiotics that prevent bacteria from replicating, rather than actually killing them, are widely used. They give the body’s own defences time to eliminate the infection without allowing the pathogen to accumulate Fig. 11.4 Time to progression of a prostate tumour as a function of mutation rate

References

241

to dangerous levels. Most effective anticancer drugs are cytotoxic agents: they kill tumour cells, with greater or lesser degrees of selectivity. However, cytostatic agents have been explored in oncology, with rather limited success. Unlike the situation with antimicrobial therapy, the antitumour immune response is usually not strong enough for the approach to work. Now that we are learning to boost antitumour immunity, it may be time to re-visit cytostatic agents. Drugs that inhibit protein synthesis are usually cytostatic, rather than cytotoxic. An example is 8-azaguanine. It is too toxic for parenteral use, but has been used with some success as a topical treatment for skin tumours. VIP is able to model cytostatic agents and may be able to suggest effective combinations of cytostatic drugs with immunomodulators. Clinical trials of agents that might prevent cancer have historically been few and far between, largely because of their expense and the expected long timelines. This is an area where in silico clinical trials could provide an early, inexpensive way of optimising trial designs and prioritising agents for clinical testing. These models could be used to predict the effects of (for example) antioxidants or antimetastatic agents on progression of asymptomatic individuals to clinical cancer. Measurements of circulating tumour DNA could then be used to validate the models.

11.11

The Potential Role of Cancer Vaccines

Cancer vaccines, and cellular immunotherapy with CAR-T cells are discussed in Chap. 10. (Alboncini et al. 2020; Anna et al. 2021). The current clinical trials with cancer vaccines and cellular immunotherapy are testing their therapeutic efficacy— i.e., in established disease, but their long-term potential will be for cancer prevention. When cancer-causing mutations are detected in people with no symptoms, ethical and economic considerations probably rule out existing chemotherapy approaches, but a vaccine with no significant side effects, if it was affordable, would be justifiable. A cancer vaccine that was as safe and cheap as the flu vaccine could transform the future of oncology. It would probably need to be a cocktail of vaccines raised against multiple mutations, but that is already established practice with antiinfective vaccination. Circulating tumour DNA mutations, having provided the initial diagnosis, would provide an early biomarker for vaccine efficacy.

References Alboncini L, Benvenuto M, Focaccetti C et al (2020) PlGF immunological impact during pregnancy. Int J Mol Sci 21:8714. https://doi.org/10.3390/ijms21228714 Al-Shaheri FN, Alhamdani MSS, Bauer AS et al (2021) Blood biomarkers for differential diagnosis and early detection of pancreatic cancer. Cancer Treat Rev 96:102193. https://doi.org/10.1016/j. ctrv.2021.102193

242

11

Implications of Evolutionary Dynamics for Cancer Treatment and Prevention

Anna F, Elodie BR, LeMaoult J et al (2021) First immunotherapeutic CAR-T cells against the immune checkpoint protein HLA-G. J Immunother Cancer 9:e001998. https://doi.org/10.1136/ jitc-2020-001998 Beckman RA, Schemmann GS, Yeang C-H (2012) Impact of genetic dynamics and single-cell heterogeneity on development of nonstandard personalised medicine strategies for cancer. Proc Natl Acad Sci U S A 109:14586–14591 Brimacombe KR, Anastasiou D, Hong BS et al (2012) ML285 affects reactive oxygen species’ inhibition of pyruvate kinase M2. In: Probe reports from the NIH molecular libraries program [internet]. National Center for Biotechnology Information (US), Bethesda, MD Chaffer CL, Weinberg RA (2011) A perspective on cancer cell metastasis. Science 331:1559–1564 Christopherson T (2017) Tripping over the truth: how the metabolic theory of cancer is overturning one of medicine’s most entrenched paradigms. Chelsea Green Publishing, White River, VT Clifton KK, Ma CX, Fontana L, Peterson LL (2021) Intermittent fasting in the prevention and treatment of cancer. CA Cancer J Clin 71:527–546 Crick F (1988) What mad pursuit. A personal view of scientific discovery. Basic Books, New York, p 111 Cunningham JJ, Brown JS, Gatenby RA, Stankova K (2018) Optimal control to develop therapeutic strategies for metastatic castrate resistant prostate cancer. J Theor Biol 459:67–78 Daily Telegraph (2015, February 23) The Prof who ‘cured’ his own cancer Darrigues L, Pierga J-Y, Bernard-Tessier A et al (2021) Circulating tumor DNA as a dynamic biomarker of response to palbociclib and fulvestrant in metastatic breast cancer patients. Breast Cancer Res 23:31. https://doi.org/10.1186/s13058-021-01411-0 Gallaher JA, Enriquez-Navas PM, Luddy KA et al (2018) Spatial heterogeneity and evolutionary dynamics modulate time to recurrence in continuous and adaptive cancer therapies. Cancer Res 78:2127–2139 Gatenby RA, Gillies RJ, Brown JS (2010) Evolutionary dynamics of cancer prevention. Nat Rev Cancer 10:526–527 Gillies RJ, Verduzco D, Gatenby RA (2012) Evolutionary dynamics of carcinogenesis and why targeted therapy does not work. Nat Rev Cancer 12:487–493 Hardy DG, Ashford MLJ (2014) AMP-K: regulating energy balance at the cellular and whole body levels. Physiology (Bethesda) 29:99–107 Ibrahim EM, Al-Faheidi MH, Al-Mansoor MM (2021) Energy and caloric restriction, and fasting and cancer: a narrative review. Support Care Cancer 29:2299–2304 Jackson RC, Fernandez E, Radivoyevitch T (2015) Bayesian systems for optimizing treatment protocols in oncology. In: Ur Rahman A (ed) Frontiers in clinical drug research – anticancer agents, vol 2. Bentham Science Publishers, pp 60–145 Kailin WG (2009) Synthetic lethality: a framework for the development of wiser cancer therapeutics. Genome Med 1:99. https://doi.org/10.1186/gm99 Kim J-W, Dang CV (2006) Cancer’s molecular sweet tooth and the Warburg effect. Cancer Res 66: 8927–8930 Klein EA, Richards D, Cohn A et al (2021) Clinical validation of a targeted methylation-based multi-cancer early detection test using an independent validation set. Ann Oncol 32:1167–1177 Klein EA, Beer TM, Seiden EN (2022) The promise of multicancer early detection. Diagnostics (Basel) 12:1243. https://doi.org/10.3390/diagnostics12051243 Liew SK, Malagobadan S, Arshad NM, Nagoor NH (2020) A review of the structure-activity relationship of natural and synthetic antimetastatic compounds. Biomol Ther 10:138. https://doi. org/10.3390/biom10010138 Magbanua MJM, Swigart LB, Wu H-T et al (2021) Circulating tumor DNA in neoadjuvant-treated breast cancer reflects response and survival. Ann Oncol 32:229–239 McAuslan BR, Reilly W, Hannan GN et al (1988) Induction of endothelial cell migration by proline analogs and its relevance to angiogenesis. Exp Cell Res 176:248–257

References

243

Nadauld LD, McDonnell CH, Beer TM et al (2021) The PATHFINDER study: assessment of the implementation of an investigational multi-cancer early detection test into clinical practice. Cancers (Basel) 13:3501. https://doi.org/10.3390/cancers13143501 Oversoe SK, Clement MS, Pedersen MH et al (2020) TERT promoter mutated circulating tumor DNA as a biomarker for prognosis in hepatocellular carcinoma. Scand J Gastroenterol 55:1433– 1440 Pich O, Bailey C, Watkins TBK et al (2022) The translational challenges of precision oncology. Cancer Cell 40:458–478 Roosan MR, Mambetsariev I, Pharaon R et al (2021) Evaluation of somatic mutations in solid metastatic pan-cancer patients. Cancers (Basel) 13:2776. https://doi.org/10.3390/ cancers13112776 Sabale AS, Patel UM, Gorhe AA et al (2020) Prospective anticancer agents: a recent update on drug repurposing. Anticancer Agents Med Chem 20:1–21 Seyfried TN, Flores RE, Poff AM, D'Agostino DP (2014) Cancer as a metabolic disease: implications for novel therapeutics. Carcinogenesis 35:515–527 Silva AS, Kam Y, Khin ZP (2012) Evolutionary approaches to prolong progression-free survival in breast cancer. Cancer Res 72:6362–6370 Steeg PS, Theodorescu D (2008) Metastasis: a therapeutic target for cancer. Nat Clin Pract Oncol 5: 206–219 Whale A, Hashim FN, Fram S et al (2011) Signalling to cancer cell invasion through PAK family kinases. Front Biosci 16:849–864 Wu CA, Chao Y, Shiah SG, Lin WW (2013) Nutrient deprivation induces the Warburg effect through ROS/AMP-K-dependent activation of pyruvate dehydrogenase kinase. Biochim Biophys Acta 1833:1147–1156

Chapter 12

In Science All Conclusions Are Provisional

Abstract Cancer is best understood as a process of Darwinian selection, as an instance of a genetic algorithm that has two essential drivers: cellular heterogeneity, and competition for resources that results in those variants that have a selective advantage becoming the predominant population in an early-stage tumour. The essence of malignancy is genetic or chromosomal instability, and it is this instability that results in the cellular heterogeneity that provides the “descent with modification” required for Darwinian selection. Genetic instability in itself is usually either lethal or results in a reproductive disadvantage. For the variant cells to survive and multiply, the instability must be accompanied or preceded by a mutation that provides an offsetting growth advantage. When a normal tissue stem cell has acquired these two changes, it has become a tumour stem cell. Because of their genetic/ chromosomal instability, tumour stem cells have an effective mutation rate orders of magnitude greater than normal. This results in the phenomenon of tumour progression, in which several additional mutations cause accelerated growth, decreased spontaneous apoptosis, and independence from cellular adhesion to extracellular matrix. The resulting tumour can now grow as a three-dimensional lump, resulting in hypoxia, which in turn stimulates angiogenesis and the metabolic changes known as the Warburg effect. Subsequently, loss of the requirement for cell–cell contact means that tumour cells can leave the primary site and migrate to other parts of the body to form metastatic growths. Many tumours express immune checkpoint proteins, making them invisible to immune surveillance. Anticancer drugs have been developed that target most of these processes. Many early anticancer agents were alkylating agents that exploit the defective DNA damage response of most tumours. Antimetabolites deplete the nucleotides required for repair of damaged DNA. Many tumours carry mutations in the mitotic spindle assembly checkpoint, and another important family of antitumour drugs, the spindle poisons, target mitosis. A newer family of drugs, protein kinase inhibitors, block the signal transduction pathways that stimulate cell division by override of the G1:S cell cycle checkpoint. The multiple mutations acquired in the process of tumour progression make tumours immunogenic, and the immune system can often inhibit or eliminate them. Tumours can, however, thwart the immune surveillance process by expressing immune checkpoint proteins on their surface, and inhibitors of these checkpoints are becoming increasingly important in cancer treatment. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1_12

245

246

12

In Science All Conclusions Are Provisional

Because of their extreme heterogeneity, many tumours are able to develop resistance to all these classes of drugs, and for such tumours the emphasis is switching to early diagnosis. The future of treatment for many cancers is likely to involve agents that block the progression process, including cancer vaccines.

For every complex problem, there is an answer that is clear, simple—and wrong. (Paraphrased from HL Mencken)

A number of important insights have resulted in our present understanding of cancer. And despite being, in Mencken’s sense “wrong”, they were wrong in the sense that Darwin or Newton were wrong, and in the sense that all models of complex systems are wrong, because all models of complex systems are necessarily simplifications.

12.1

Why Has Cancer Been an Exception in the March of Medical Progress?

During my childhood, in the early 1950s, our next door neighbour, Billy, was diagnosed with tuberculosis and “sent away” to a sanatorium. My parents and the other neighbours regarded this as a death sentence. A few months later, Billy came home, cured. TB had joined streptococcal disease, and a number of other formerly fatal infectious diseases, in becoming curable. Not long after, the former scourge of childhood, polio, became preventable by vaccination. These successes of scientific medicine caused a wave of optimism, in the medical profession and in the general public. Surely cancer would be next? That was seventy years ago. A decade later, I was at Cambridge University, studying biochemistry. Cambridge in the early 60s was still enjoying the afterglow of Watson and Crick’s solving the structure of the DNA double helix. Several of my lecturers were present or future Nobel laureates. A number of diseases, beginning with sickle cell anaemia, were found to be caused by genetic mutations: “molecular medicine” was born. The MRC Laboratory of Molecular Biology was founded. Cancer continued to elude a simple genetic explanation. This is not to deny that growing understanding of cancer biology has been reflected in improved treatment of many malignant diseases. But for a patient with (for example) metastatic lung cancer the expectation that a cure may be just around the corner, as it was for tuberculosis seventy years ago, has not yet been realised. The frustratingly slow progress is because cancer is an order of magnitude more complex than infectious disease, it must be understood at multiple levels of biological organisation, molecular, cellular, and environmental. Theodosius Dobzhansky famously said that “nothing in biology makes sense except in the light of evolution”. The aim of the present work has been to try and make sense of cancer as an evolutionary process.

12.2

Deterministic and Probabilistic Events in Models of Malignancy

247

Cancer is not unique in having resisted the advances in molecular medicine. Much the same could be said of complex neurological diseases, or of autoimmune diseases: these, like cancer, are perturbations of complex systems. In fact, it is the infectious diseases that are the real outliers. They offer qualitative biochemical differences as drug targets, and they are strongly immunogenic.

12.2

Deterministic and Probabilistic Events in Models of Malignancy

There are two broad approaches to modelling biological systems. In deterministic models, the system parameters are assumed to have unique, known values. In population models the parameter values are random values determined by a population mean and a statistical distribution (often, but not necessarily, Gaussian) around that mean value. The two approaches are used for different applications (Figueredo et al. 2014). If a model is to be used to study mechanistic questions, deterministic modelling is appropriate. For example, if we have an experimental drug that is known to be a cell cycle-specific agent, and we know that in a tumour whose cell cycle time is 48 hours, drug resistance is observed after 6 months, we may wish to estimate the time to emergence of resistance in a tumour with a 96-hour cell cycle time, and a deterministic model will enable us to do this. Of course, in a population of mice or humans the actual doubling times will differ from one individual to the next. If the question we want to model is “what proportion of patients will still be responding after six months”, then the appropriate technique is population modelling. The VIP in silico clinical trial program described in the supplement to Chap. 11 is an example of this approach to modelling. Figure 12.1 shows an example of the use of the VIP program to model levels of PSA in five prostate cancer patients treated daily with abiraterone. Based upon PSA levels, one patient had a complete response, and three had partial responses. The fifth patient’s tumour continued to progress in the face of treatment. After about eight months, abiraterone resistance had appeared in all five patients, and PSA levels were rising rapidly (Jackson et al. 2015). Stochastic modelling, as outlined above, is a way of including inter-personal parameter variation within the population in our models. However, it has been argued that the critical events in tumour development may have a probabilistic component. Wodarz and Komarova (2014) devote a major section of their “Dynamics of Cancer” to stochastic models of both gain of function (oncogene-induced) and loss of function (tumour initiation through tumour suppressor genes). Vermeulen et al. (2013), discussing the fate of intestinal stem cells that have undergone potentially cancer-causing mutations, suggest that the outcome of such mutations may not be deterministic and that many mutant stem cells may be replaced in the intestinal epithelium by wild-type stem cells, after biased, but still stochastic events.

248

12

In Science All Conclusions Are Provisional

Fig. 12.1 Population PK/PD modelling of the effect of abiraterone on PSA levels in five prostate cancer patients

Tumour progression has been modelled as a Markov chain (Ludwig et al. 2021). A Markov process is an event whose probability depends upon the current state of the system and is not affected by its previous history. A Markov chain is a sequence of such events. The probability of, for example, an invasive but non-metastatic tumour becoming metastatic will depend on its population size, the mutation rate, and certain environmental factors (e.g. distance from a blood vessel). For R programmers, a package is available called POMP (Tools for data analysis with partially observed Markov process models, King et al. 2022).

12.3 One-Hit, Two-Hit, and Three-Hit Malignancies: Causes and Consequences There are two events essential to malignant transformation: proliferation that is independent of physiological constraints and chromosomal (or genetic) instability. Duesberg (2007) focussed attention on chromosomal chaos, and indeed essentially all carcinomas are aneuploid. However, aneuploidy results in loss of viability, since a substantial proportion of aneuploid cells may lack essential genes or have genes expressed at inappropriate levels. For aneuploid cells to survive the Darwinian selection process they require a compensating growth advantage. The mutation or chromosomal rearrangement that confers genetic or chromosomal instability must be

12.3

One-Hit, Two-Hit, and Three-Hit Malignancies: Causes and Consequences

249

accompanied or preceded by a growth-promoting change, such as a constitutive expression of an oncogene that overrides the G1:S checkpoint, or occasionally loss of an enzyme required for spontaneous apoptosis. This “two-hit” process results in a cancer stem cell. The other hallmarks of cancer—invasion, metastasis, loss of spontaneous apoptosis (if that was not part of the initial transformation event), de-differentiation, invisibility to the immune system—arise as an inevitable consequence of chromosomal or genetic instability which increases mutation rates by orders of magnitude. The age-incidence curves for most common human cancers (Frank 2007) are consistent with two low-frequency events producing transformation followed by three or four high frequency events that drive the Darwinian selection process of tumour progression. After the first two hits have produced a cancer stem cell, depending on how well the early-stage tumour is adapted to its environment, it will either become extinct or undergo the remaining three or four steps of tumour progression. Although both genetic or chromosomal instability AND a growth advantage are required for transformation, there are a few rare malignancies in which a single event causes both these changes. Chapter 7 discussed the example of CML, where the t (9,22) chromosomal translocation resulted in constitutive activation of the Abl tyrosine kinase proto-oncogene. Abl phosphorylates and activates STAT transcription factors, including STAT5, which activates multiple proteins, including the growth stimulatory cyclin D and the anti-apoptotic protein Mcl-1, and the mitochondrial transcription factor STAT3, which triggers generation of high levels of reactive oxygen species that cause genetic instability as a result of oxidation of DNA guanine bases. Thus, although CML is a one-hit malignancy, the two-armed nature of Abl signalling means that CML progression is consistent with an evolutionary explanation. Some virally-induced tumours may be single-hit malignancies; an example is the Abelson murine leukaemia virus, which transfects mouse cells with the viral transforming gene, v-abl. An example in humans is Burkitt’s lymphoma, discussed in Chap. 2. This B cell malignancy is linked with infection by Epstein-Barr virus (EBV), which can cause a chromosomal rearrangement that places the proto-oncogene c-myc under the control of the immunoglobulin heavy chain gene IGHα. The transformed B cells express EBV antigens, making the tumour highly antigenic, triggering B cell proliferation. In this case, as with CML, a single translocation causes chromosomal instability and cell proliferation. Burkitt’s lymphoma is commoner in children than in adults, making the age distribution compatible with a one-hit aetiology. In some cases, the growth-enhancing mutation may involve a tumour suppressor gene, that is, an autosomal recessive gene. Mutations in both alleles of the tumour suppressor gene will be required, together with a mutation that causes genetic instability, making three hits necessary for transformation. If one mutant tumour suppressor gene is inherited, the result is a tumour susceptibility syndrome. PTEN (a PI3 phosphatase) is an example of a tumour suppressor gene where bi-allelic loss in the absence of genetic instability results in nonmalignant hyperplastic disease, and in the presence of genetic instability, in malignancy.

250

12

In Science All Conclusions Are Provisional

Mutations in epigenetic regulators can have pleiotropic effects, resulting in epigenetic instability. CMML, discussed in Chap. 8, is an example of a three-hit malignancy; loss of both alleles of an epigenetic regulator results in a nonmalignant myelodysplastic syndrome, and in combination with activation of a growthenhancing oncogene, such as Ras, results in malignancy. The age distribution of CMML, whose median age at diagnosis is greater than that for any malignant disease, is consistent with its three-stage aetiology. Since it is theoretically possible that both growth promotion and genetic (or epigenetic) instability could be the result of recessive mutations, four-hit malignancies might exist. However, no examples of four-hit malignancies have been reported. The one-hit, two-hit, or three-hit nature of tumours has clear implications for optimal treatment strategies. Imatinib and its analogues work so well in CML because inhibition of Abl kinase blocks both cell proliferation and progression. When treating two-hit tumours it may be necessary to combine antiproliferatives with separate agents or immunological treatments to block progression. For three-hit tumours, the optimal treatment may depend on whether or not the recessive component shows allelic insufficiency.

12.4

Major Ideas in the Development of the Evolutionary Dynamics of Malignancy

Evolutionary dynamics is one of a number of computational techniques that are increasingly used to develop screening strategies to detect cancer earlier, predict response to treatment, and design individualised treatment plans (Rockne et al. 2019). Evolutionary dynamics has been based upon data from tumour biology, biochemistry, virology, immunology, genetics, epigenetics, and genomics. Each of the preceding chapters has attempted to show how these various disciplines have contributed to our present understanding of the subject. Chapter 1: Weber and Morris’s early studies on tumour biochemistry identified three classes of biochemical difference between normal and malignant tissues: changes that correlated with tumour growth rate, changes that correlated with transformation per se, and changes that appeared to be random. In fact, there is a fourth class: activities that correlate with tumour progression, rather than with transformation per se, and the loss of differentiated functions is an example of this. The first two classes reflect what cancer geneticists later described as “driver” mutations, and the third class was apparently the result of “passenger” mutations. “Transformation-linked” biochemical changes in cancer turned out to be much more elusive (e.g. PRPP amidotransferase). Harold Morris was surely right in concluding from his study of “minimum deviation” hepatomas that malignant tissues differ qualitatively from normal tissues or benign tumours. However, George Weber, despite amassing a wealth of biochemical data, was unable to establish molecular

12.4

Major Ideas in the Development of the Evolutionary Dynamics of Malignancy

251

correlates of that difference. There is no gene, or set of genes, that uniquely defines neoplasia. The discovery of oncogenes and tumour suppressor genes emphasised the role of signal transduction pathways in activation of cell division. However, oncogenes are not unique to tumours: they are either mutant variants of normal cellular genes or normal genes that are inappropriately expressed. Oncogenes, as originally conceived—genes that when constitutively activated cause cancer—do not exist. Lymphoid neoplasias, such as Burkitt’s lymphoma, have apparently been transformed solely by a translocation that puts c-myc under the control of an immunoglobulin promoter. For transformation to occur requires the subsequent activation of that promoter by a virus infection. When the same oncogene is activated, by mutation or translocation in a non-lymphoid cell, once again, transformation only ensues following a second event, such as chromosomal instability. Cancer is not a molecular disease, in the way that, say, severe combined immunodeficiency is. That is a category error. Cancer, like ageing, or dementia, or arthritis, is a biological abnormality at a higher level of organisation, that of gene expression. Hanahan and Weinberg’s hallmarks of cancer are descriptive, but not explanatory. Chromosomal instability, which they list as an enabling factor, is the essence of neoplasia, not just an accelerant. Chapter 2: Work on the mechanism of chemical and radiation carcinogenesis showed that induction of malignancy was associated with genetic damage. Genetic damage is usually lethal, but cells that survive it often carry mutations. Cytogenetic studies showed that all tumours or leukaemias have abnormal karyotypes. Some malignancies result from chromosomal translocations. Carcinomas have multiple karyotypic changes and are usually aneuploid. Duesberg, in defining malignancy as a chromosomal abnormality was right, but not completely right: chromosomal aberrations are usually lethal, unless they are accompanied by an offsetting growth advantage. Duesberg made the illuminating comparison between tumour progression and speciation. Once again, there is a good deal of truth in the comparison, but it is not the whole truth, for the same reason that Steve Jones and others have criticised Darwin: natural selection, in itself, does not explain the origin of species. It is a contributor, but without partial reproductive isolation, karyotypic changes (or growth-enhancing mutations) will not result in speciation. The fact that chromosomal translocation, in the absence of mutation, can transform normal cells to malignancy supports Duesberg’s contention that cancer is not a genetic disease, it is a disease of gene expression. Chromosomal instability results in cellular heterogeneity, a prerequisite for Darwinian selection. Tumour genomics, like tumour biochemistry, has concluded that there is no single mutation that causes cancer, and there is no particular mutation that is universal to all cancers. Chapter 3: All malignant cells have deleted or malfunctioning G1:S checkpoints. About 50% of cancers have deleted or mutant p53; the remainder have mutations causing constitutive activation of the pathways that turn off the G1:S checkpoint. David Lane, in describing p53, did indeed discover the “guardian of the genome”, but mutant p53, or other mutations that override the G1:S checkpoint, though always involved in malignancy, are not THE cause of cancer, they are a component part of a

252

12

In Science All Conclusions Are Provisional

higher-order pathology. Loss of the G1:S checkpoint in diploid, chromosomally stable cells causes benign hyperplasia. This may be regarded as a premalignant state, in that it increases the probability that aneuploid cells that may arise subsequently will survive as malignant cells. The importance of loss of the G1:S checkpoint to malignancy is that it contributes one of the two requirements for Darwinian selection to occur—a selective survival advantage. Chapter 4: Many tumours have dysfunctional DNA damage checkpoints, resulting in loss of ability to repair DNA damage and progression of cells with damaged DNA into mitosis. DNA strand breaks are part of normal cell division. Deletions or mutations in the process of DNA repair are normally harmful, even lethal. As part of a multi-step process, however, they can lead to malignancy; a number of such mutations are therefore regarded as driver mutations. Cross-talk between the DDR and the SAC means that inactivation of the SAC can enable cells with unrepaired DNA damage to enter cell division and establish a clone of transformed cells. Chapter 5: Carcinomas usually have mutations or expression changes in components of the mitotic spindle assembly checkpoint, resulting in progressive aneuploidy. The SAC normally functions by making errors of chromosome segregation lethal. As with errors in the DDR, if SAC malfunction is part of a broader pattern of genetic changes, or changes in gene expression, transformation may result. The dynamics of the SAC are such that the checkpoint is either ON (meaning that cells that have not correctly sorted their replicated chromosomes into two complete sets cannot undergo cell division) or OFF (resulting in aneuploidy). Aneuploid cells usually have an increased cell loss factor, resulting in their eventual extinction, but if aneuploidy is coupled with a growth advantage, resulting from loss or override of the G1:S checkpoint, aneuploid cells may survive as cancer stem cells. Chapter 6: Drug resistance is a frequent consequence of the elevated mutation rate of malignant cells. The model of Goldie and Coldman, and the model of Skipper and Lloyd suggested that drug-resistance mutations are the most frequent cause of treatment failure. The difficulty in extrapolating from Skipper’s work with syngeneic tumours in inbred mice is that in human (or with immune-competent mice) it is NOT necessary to eliminate every last tumour cell, and indeed chemotherapy drugs based on these studies with syngeneic mouse tumours (or human tumour xenografts in immunodeficient mice) often exaggerated the activity of immunosuppressive drugs. Magic bullets do not exist: To explain why adding cisplatin to the two-drug combination of bleomycin plus etoposide turned testicular cancer, a disease that was 90% lethal into 90% curable, we must invoke evolutionary dynamics, which describes biological phenomena, not at the molecular level, but at the level of population biology. The concept of adaptive therapy was devised using evolutionary principles to maximise the time to progression of advanced tumours. The acquisition of drug resistance by tumour cells incurs costs to the cell that can be exploited by manipulating the tumour’s environment, including the drug environment, so as to minimise progression.

12.4

Major Ideas in the Development of the Evolutionary Dynamics of Malignancy

253

Chapter 7: Experience with CML showed that a single chromosomal translocation could result in malignant transformation, if that translocation had pleiotropic effects that both resulted in genetic instability and a growth stimulus or inhibition of apoptosis. CML is a one-hit malignancy but because that single chromosomal rearrangement causes genetic instability AND a compensating growth advantage, it is a progressive condition. Imatinib and its analogues are extremely effective as single agents, because by targeting Abl kinase, they inhibit both proliferation and disease progression. Chapter 8: Experience with CMML showed that epigenetic gene silencing can contribute to malignant transformation and/or progression. Gene silencing can have the same effect on gene expression as a loss-of-function mutation. The common feature is a change in gene expression. Changes in gene expression that cause genetic or chromosomal instability, whether resulting from a genetic or epigenetic origin, can result in malignant progression. The enzyme TET2, which is part of the system that removes DNA methyl groups, can reactivate silenced genes. However, if, as with CMML, a tumour suppressor function has been epigenetically silenced, TET2 activation may have an antitumour effect. Chapter 9: A stem cell that has suffered both a chromosomal or genetic change resulting in chromosomal/genetic instability AND a mutation that confers a growth advantage sufficient to offset the growth disadvantage resulting from the genetic damage is now a tumour stem cell. Tumour stem cells are potentially immortal and genetically unstable, resulting in cellular heterogeneity and setting in train the process of Darwinian selection known as tumour progression. Warburg’s explanation of cancer as a reversion from oxidative metabolism to the more primitive anaerobic metabolism is no longer regarded as a primary cause of cancer, but the change in energy metabolism that upregulates glycolysis and decreases mitochondrial function is an almost universal feature of the tumour progression process. Some aspects of tumour progression, such as angiogenesis, are a normal physiological response to hypoxia. Metastasis, however, requires driver mutations that enable cells to survive in the absence of cell–cell contact. Chapter 10: The body is protected from many potentially oncogenic mutations by the process of immune surveillance. However, during tumour progression, cells may emerge that express negative feedback effects known as immune checkpoints. Such cells are now invisible to immune surveillance and have a selective survival advantage. Tumour checkpoint inhibitors are becoming an important component of treatment regimens for many tumours. Although it is widely accepted that the biology of tumour progression can be explained as a process of Darwinian natural selection, an opposing view argues that neutral evolution, i.e. accumulation of mutations that are neither advantageous or disadvantageous, may be consistent with the observed dynamics of tumour progression. The two contrasting views are not necessarily mutually exclusive, because neutral mutations that may be passengers when they arise may become drivers at a later stage of the progression process. Chapter 11: Our growing understanding of tumour immunology has suggested that we do not have to kill every last cancer cell to “cure” cancer. The multi-stage

254

12

In Science All Conclusions Are Provisional

process of tumour progression has multiple potential points of intervention. Some of these have been widely explored (such as metastasis), others (such as restoring activity of the SAC) hardly at all. “Cancer is a hundred different diseases”—according to histopathologists. “Every tumour is genetically unique”—according to molecular geneticists. What all the millions of possible tumours have in common is genetic (or chromosomal) instability that results in cellular heterogeneity that sets in train a process of Darwinian selection for the phenotype best adapted to survive and grow free of normal physiological constraints. The various explanations of cancer that we have discussed, though incomplete, have all contributed insights into what we now accept to be a phenomenon of complexity. Just as Newton’s mechanics is an almost complete explanation of physics, and Einstein’s relativity is an even closer explanation, and quantum mechanics fills a few more of the gaps, our understanding of tumour biology has proceeded by a series of successive approximations. Darwin didn’t quite explain the origin of species, and the modern synthesis of Darwinism and Mendelian genetics, and the Drosophila geneticists and Watson and Crick all added a piece of the puzzle. Sometimes a chromosomal translocation causes cancer, sometimes a chemical carcinogen, or a virus. When a complex system goes wrong, or changes its behaviour, the explanation must be sought in the altered interaction of its components.

12.5

All Stages of Transformation and Progression Have Been Targeted by Anticancer Drugs

Most of our existing anticancer drugs inhibit cell proliferation. Tumours do not necessarily have higher proliferation rates than normal tissues, and inhibition of bone marrow or intestinal stem cells, or of the immune system, or hair follicles, or all of these, is an almost invariable side effect of antiproliferative drugs. In the early days of anticancer drug discovery, it was believed that transformed and normal cells had only quantitative differences, so that—unlike antibacterial therapy, where there are qualitative biochemical differences to exploit—the selectivity of anticancer drugs must always be limited. In fact, as I hope the preceding chapters have made clear, tumour cells do have qualitative differences from the normal cells from which they are derived, but those differences are not solely at the molecular level, but at a more complex level of biological organisation, the control of gene expression. Attempts to make antiproliferative drugs more tumour-selective must address this fact. An ingenious example is the concept of cyclotherapy (Blagosklonny and Darzynkiewicz 2002) which attempts to exploit the fact that normal cells have a functional G1:S checkpoint, and tumour cells do not. If the checkpoint is triggered, the normal cells will arrest at the G1:S boundary, while the tumour cells will progress into S phase, where they can be selectively destroyed by an S phase-specific drug. An example

12.7

Unanswered Questions that May Be Studied by Evolutionary Dynamics

255

studied by Lane and colleagues (Rao et al. 2010) used the p53 activator, nutlin-3, to arrest normal cells, and thus protect them from a downstream cytotoxic agent.

12.6

Future Progress Will Require a Deeper Understanding of Malignant Progression

The two stages (usually) required to generate a cancer stem cell—release of cell proliferation from physiological constraint, and genetic/chromosomal instability, result in cellular heterogeneity that triggers Darwinian tumour progression. At least three subsequent driver mutations are then required for a cancer stem cell to progress to a metastatic tumour. If these processes can be inhibited, it may be possible to keep tumours localised. However, because there are multiple mutations that result in (for example) an invasive but non-metastatic tumour becoming metastatic, inhibiting tumour progression is likely to require multiple-drug treatment. Expression of immune checkpoints is a frequent aspect of progression, and checkpoint inhibitors are playing an increasingly prominent role in treating advanced cancers. If, as seems likely, progression results in tumours becoming more immunogenic, cancer vaccines will be a tool for inhibiting progression.

12.7

Unanswered Questions that May Be Studied by Evolutionary Dynamics

If, as argued by Gillies et al. (2012), the evolutionary nature of cancer development means that targeted therapy cannot work, can evolutionary dynamics be used to base personalised medicine on mutation profiles? Can we detect and treat tumours before the progression process has reached a point of no return? As discussed in Chap. 11, the detection of a cancer-associated mutation in a sample of blood from an asymptomatic person is not a diagnosis of cancer. A cell carrying such a mutation may undergo spontaneous apoptosis, it may be eliminated by the immune system, it may survive without proliferating, or it may progress to form an advanced tumour. In any case, no conclusions can be drawn from a single mutation. However, conclusions can be drawn from multiple mutation signatures. It is increasingly possible from particular mutation signatures and DNA methylation patterns to infer the probable body site of the otherwise undetectable tumour that has released cells or DNA into the blood. Is it possible to predict whether or not a particular early-stage tumour, detected by a mutation signature in blood-borne DNA will progress? In the long term, a database of outcomes can be created for the various signatures. However, the many different ways that a normal cell can become malignant make it impractical for such a database to contain outcomes for every possible mutation signature. We have seen that any of (at least) 30 mutations can result in oncogene activation that will

256

12

In Science All Conclusions Are Provisional

override the G1:S checkpoint. If any of those mutations is followed by one of (conservatively) 20 mutations in the DDR or the SAC that can result in genetic or chromosomal instability, there are at least 600 possible cancer stem cell genotypes. A cancer stem cell may then have one of 20 or more mutations resulting in epithelialmesenchymal transition to produce an invasive tumour. That invasive tumour can then undergo one of a number of mutations (say 20) that result in the tumour becoming metastatic. Either before or after becoming metastatic a growing tumour must, unless it is in an immunologically privileged site (such as under the kidney capsule), have a further mutation to activate an immune checkpoint; there are probably at least ten mutations that could have this effect. At a rough, but probably conservative, estimate there are 600 × 20 × 20 × 10 = 24 million possible mutation combinations that could lead to an advanced metastatic tumour. Some of these mutation signatures will be much commoner than others. This raises the possibility of creating a database of outcomes for the commonest signatures, and using that database to validate an evolutionary dynamics model for predicting probable outcomes. Gerstung et al. (2017) discuss such an approach to personalisation of treatment of AML. Genomic data from 1540 AML patients with matched clinical data can predict likelihood of remission, relapse, and mortality. Similar analyses for other malignancies will undoubtedly follow. How can evolutionary dynamics predict the likely fate of an early-stage malignancy from its mutation profile? For each mutation, this will require knowledge of that mutation’s effect on growth and death rates. As we have discussed in the preceding chapters, this information is rapidly accumulating. At a first approximation, at least, these effects will be independent of the tumour type, and they can be measured in vitro. This approximation ignores the fact that the microenvironment of a glioma cell in the brain may be very different from that of a carcinoma stem cell in the colon, but it is a place to start. Are cancer stem cells sufficiently foreign that they can be eliminated by immune surveillance? If a cancer stem cell can differ from a normal stem cell by as few as two mutations, it may be possible, in principle, to produce a vaccine that recognises those mutant epitopes. The problem is that there are (at least) twenty or thirty possible mutations for each of the two steps required to produce a cancer stem cell, meaning that even if the mutations were sufficiently immunogenic the number of cancer stem cell phenotypes will be very large. Aerobic glycolysis undoubtedly contributes to malignant progression. Can it also, in itself, initiate cancer? In certain environments (e.g. hypoxia) cells that use glycolysis as their primary energy pathway will have a selective advantage. Is that enough? Can we use drugs, or immunological treatments, to turn cancer into a chronic disease that we can live with? The fact that a substantial sub-set of prostate tumours do not progress, or progress sufficiently slowly that they are not life-threatening, gives us encouragement that a better understanding of the tumour progression process, and a wider range of drugs, or immunological treatments, that target tumour progression may enable us to make more malignancies into chronic diseases, as imatinib has done for CML. The adaptive therapy concept, based upon evolutionary

12.7

Unanswered Questions that May Be Studied by Evolutionary Dynamics

257

dynamics, represents an approach that argues that, while many advanced tumours may not be curable, they may be manageable as chronic diseases. How many steps are required for completion of the process of carcinogenesis? It is always a multi-step process. Even with the rare one-hit malignancies, such as CML, the translocation that transforms a myeloid stem cell is followed by the inevitable mutations that drive the disease into the accelerated phase and the acute phase. Hanahan and Weinberg described six hallmarks of cancer, and later added two more. However, this does not mean that eight mutations are necessary for malignant transformation and progression. Growth that is independent of physiological constraint and a decreased level of spontaneous cell death will both provide a competitive survival advantage, as will insensitivity to negative growth signals. It may well be that a particular tumour cell exhibits all these changes. In CML, the same translocation that upregulates the growth-promoting STAT5 also upregulates the anti-apoptotic protein Mcl-1. At the other extreme, Duesberg argues that all that is necessary for transformation to occur is chromosomal instability. Instability results in heterogeneity (“descent with variation”), which provides one of the essentials for Darwinian selection. The present consensus is that chromosomal or genetic instability is necessary, but not sufficient, for transformation. Evolutionary dynamics calculations suggest that chromosomal instability, in itself, incurs a selective disadvantage, and for aneuploid cells to persist and form a tumour, the chromosomal instability must be accompanied or preceded by a growth-promoting mutation. This provides the second essential for Darwinian selection: competition for resources, Darwin’s “struggle for existence”. At a minimum, then, two changes are necessary to make a cancer stem cell. Because transformation is a Darwinian process, it is likely that more than the minimum number of changes will have occurred in a particular cancer stem cell. A cell that has mutant p53 will have a competitive advantage over a normal stem cell, but a cell with both mutant p53 and constitutively activated Ras will have a competitive advantage over a cell with only one of these mutations. What changes must occur next will depend upon the lineage of the cell. For a leukaemia or a lymphoma subsequent mutations will tend to make the disease more aggressive, more rapidly proliferating, as we saw in the analysis of CML and CMML. Myeloid and lymphoid cells are mobile and able to spread to almost any part of the body. For an epithelial tumour, a carcinoma—and these are the commonest human cancers—the process of malignant progression is more complex. These cells are firmly anchored in their tissue of origin, and in many cases can only proliferate when attached to extracellular matrix. A carcinoma in situ, growing as a flat sheet of cells on a basement membrane—surface growth—will replicate only at the margins. For exponential growth to occur, a further mutation is required that enables the cell to replicate independently of extracellular matrix (ECM) attachment. This change, the epithelial-mesenchymal transition (EMT), means that the cell is now able to grow as a solid lump and is able to invade surrounding tissues. It is possible that the EMT involves more than one mutation, but the ability to replicate independently of attachment, involving perhaps constitutive integrin signalling, is a

258

12

In Science All Conclusions Are Provisional

minimal requirement. If the tumour is a sarcoma, it is already a mesenchymal cell and does not need to undergo the EMT. A tumour that has undergone the EMT remains localised at its site of origin. It will generally be curable surgically. Once it has grown to more than a couple of millimetres in diameter, most of its cells, certainly those in the interior, will become hypoxic. This triggers the activation of hypoxia-inducible transcription factors, which stimulate the growth of blood vessels into the growing tumour mass. This process of tumour angiogenesis does not require a further mutation. It occurs by the same process as the vascularisation of growing tissues and organs during normal development. As such, it is controlled epigenetically, rather than genetically. The same probably applies to the Warburg effect, the switch of energy metabolism from oxidative phosphorylation to glycolysis under conditions of hypoxia. What is different from normal tissues is the fact that glycolysis persists as the primary energy source when (because of angiogenesis) the oxygen supply is restored. Though an almost invariable part of tumour progression, the Warburg effect probably does not require an additional mutation. It is a consequence of the EMT and epigenetically regulated. The next stage in the process of tumour progression is the onset of metastasis. Cells in an invasive tumour no longer require ECM attachment to replicate, but in general they do require cell–cell attachment. This is part of the normal development programme: organs and tissues could not develop normally if cells were free to wander off and live independently. The integrity of normal tissues is maintained by a complex network of adhesion molecules. One group of these, the cadherins, is particularly involved in maintaining cell–cell contact, and for epithelial cells to replicate they require both a growth signal and an attachment signal. If, because of a mutation, the cell–cell attachment signal becomes constitutive (active in the absence of attachment), the cell is now potentially metastatic. Metastasis is a complex process and probably requires several mutations, but it clearly requires at least this one. It is possible that other mutations required for metastasis have accumulated as neutral mutations during the pre-metastatic phase. These four changes—growth advantage, chromosomal instability, EMT, and metastasis—are the minimum requirement for an epithelial cell to become a metastatic tumour. Leukaemias and lymphomas only require the first two, though they do accumulate further mutations and progress. Sarcomas do not require the EMT. However, in the process of acquiring this succession of mutations, the tumour has become genotypically and phenotypically less normal, and, as a result, increasingly immunogenic. At this point, some tumours may be destroyed by the host immune system. Others may enter a state of apparent dormancy, in which cell proliferation and immune cell killing are roughly in balance. At this point, tumours which are able to express immune checkpoint molecules gain a Darwinian advantage. This process is probably epigenetically regulated, though it is likely that the epigenetic change is itself triggered by a mutation (e.g. in TET-2). Some tumours, such as renal cell carcinomas, originate in a site that is inaccessible to cells of the immune system, and these tumours do not activate immune checkpoints.

References

259

Thus, the short answer to the question “how many mutations does it take for a normal cell to become an advanced cancer?” is five—with the many caveats listed above. This raises the follow-on question “ is the order of the mutations important?” Cancer stem cells already possess the two attributes required for Darwinian selection—chromosomal instability, which generates heterogeneity, and growth advantage, which enables them to compete for resources—and pass them on to their progeny. The process of metastasis requires that cell to be invasive, so probably the mutation that causes EMT must precede loss of cell–cell contact. Some tumours metastasise early, and others late, so it is possible that some of the mutations required for metastasis depend on the site of the primary tumour. It is not clear whether evasion of the immune response is an early or late property, and it may be that this can confer a survival advantage at any stage of the malignant progression process.

12.8

Cancer as a Disease of Gene Expression

The researchers cited in this survey of cancer biology have progressed from describing cancer as a disease of excessive cell division, to a disease of uncontrolled cell division, to a state of genetic instability, to a disease of complexity—a perturbation of chromosomal, genetic, and epigenetic control. It is all these things, but perhaps the closest description of our present state of knowledge can be found at the intersection of genetics and environment, in the control of gene expression. George Weber described the biochemical changes in cancer as resulting from “reprogramming of gene expression”. But was gene expression ever “programmed”? (If so, who programmed it?). An evolutionary perspective tells us, rather, that certain expression patterns are selected, and that when potentially cancer-causing mutations arise, it is the competition between those mutants and the normal stem cells in their particular microenvironment that will determine which cells will go on to proliferate.

References Blagosklonny MV, Darzynkiewicz Z (2002) Cyclotherapy: protection of normal cells and unshielding of cancer cells. Cell Cycle 1:375–382 Duesberg P (2007) Chromosomal chaos and cancer. Sci Am 296:35–41 Figueredo GP, Siebers P-O, Owen MR et al (2014) Comparing stochastic differential equations and agent-based modelling and simulation for early-stage cancer. PLoS One 9(4):e95150 Frank SA (2007) Dynamics of cancer: incidence, inheritance, and evolution. Princeton University Press, Princeton Gerstung M, Papaemmanuil E, Martincorena I et al (2017) Precision oncology for acute myeloid leukemia using a knowledge bank approach. Nat Genet 49:332–340 Gillies RJ, Verduzco D, Gatenby RA (2012) Evolutionary dynamics of carcinogenesis and why targeted therapy does not work. Nat Rev Cancer 12:487–493

260

12

In Science All Conclusions Are Provisional

Jackson RC, Fernandez E, Radivoyevitch T (2015) Bayesian systems for optimizing treatment protocols in oncology. In: Rahman AU (ed) Frontiers in clinical drug research – anticancer agents, vol 2. Bentham Science, pp 60–145 King AA, Ionides EL, Breto C, et al (2022). https://CRAN.R-project.org/package=pomp Ludwig R, Pouymayou B, Balermpas P, Unkelbach J (2021) A hidden Markov model for lymphatic tumor progression in the head and neck. Sci Rep 11:12261. https://doi.org/10.1038/s41598-02191544-1 Rao B, van Leeuwen IM, Higgins M et al (2010) Evaluation of an actinomycin D/VX680 aurora kinase inhibitor combination in p53-based cyclotherapy. Oncotarget 1:639–650 Rockne RC, Hawkins-Daarud A, Swanson KR et al (2019) The 2019 mathematical oncology roadmap. Phys Biol 16:041005. https://doi.org/10.1088/1478-3975/ab1a09 Vermeulen L, Morrissey E, van der Heijden M et al (2013) Defining stem cell dynamics in models of intestinal tumor initiation. Science 342:995–998 Wodarz D, Komarova NL (2014) Dynamics of cancer: mathematical foundations of oncology. World Scientific, Singapore, p 131

Appendix: Using the Online Supplements

How to download the R programming language. The R programs contained in the online supplement can be used with any operating system that R runs on: Windows, Mac, or Linux. R is a free software and can be downloaded from http://cran.r-project.org. Some of the R programs in the supplement require the package deSolve, which can be downloaded from a cran site, and must be installed on the user’s system before the program will run. Other programs are written in the C programming language. These are provided as C source code and must be compiled using a C or C++ compiler. The programs were written and tested using the free software Dev-c++ which can be downloaded from https://sourceforge.net. Minor adjustments may be necessary to the #include statements if you are using other compilers. To run the R programs in the appendix, simply copy the R text (using Notepad or a similar text editor) and copy into the R prompt. To run C programs, copy the source code into the C compiler. Graphical output is included in the R programs and will be displayed automatically. It can be saved by using R’s “copy to the clipboard” function. The C programs create text files that can then be used with graphics programs. The illustrations in the text were created using the free software “gnuplot” (https:// sourceforge.net/projects/gnuplot). Where gnuplot is called by programs, it will be necessary to install a short program in the working directory. For example, the appendix to Chap. 3 contains a program, “signal7A.gnu”. Inspecting this shows that the “plot” command requires an output file “table4.txt” which was created when the signal7X program was run. If you are using other graphics programs, you can identify the graphics output text file by inspecting the “plot” commands of the appropriate .gnu files. Accessing the online supplements: a link to the online supplement of each chapter can be found on the chapters’ opening pages.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1

261

Index

A Abelson proto-oncogene (Abl), 133 Abemaciclib, 58 Abiraterone, 119, 123, 197–199, 216–218, 229, 236–240, 247, 248 Acquired resistance, 21, 34, 106, 107, 109, 139, 213, 215, 233 Acral melanoma, 34 Actin, 191 Active transport, 109 Adaptive immune response, 206, 207 Adaptive therapy, 117, 123–126, 229, 252, 256 Adjuvant chemotherapy, 228 Adoptive immunotherapy, 209 Aerobic glycolysis, 91, 187–190, 211, 234, 256 Ageing, 26, 50, 137–138, 154–155, 159, 251 Alkylating agents, 41, 111, 117, 122 α-ketoglutarate, 157, 174 5-α-reductase, 197, 198 Amphitelic attachment, 89 AMP kinase, 233 Anaphase-promoting complex (APC), 52, 55, 72–74, 87, 208 AND gates, 7, 10, 50 Androgen, 7, 9, 35, 106, 189, 197, 233 Aneuploidy, 2, 30–33, 36, 37, 55, 90–92, 96– 99, 156, 233, 248, 252 Angiogenesis, 17, 19, 185–187, 198, 199, 210, 230, 236, 253, 258 Antiandrogens, 106, 196–200, 229, 237 Antifolates, 21, 110, 118 Antimetabolites, 21, 22, 41, 110, 118, 122, 158 Apc, 6, 10, 32, 92, 184, 185, 192

Apoptosis, 17–20, 32, 39, 48, 49, 57, 58, 68, 71–75, 77–81, 91, 92, 94, 95, 97, 126, 133, 135, 142, 143, 145, 147, 149, 184, 186, 188, 190, 193, 207, 208, 215, 236, 249, 253, 255 Artemis, 68 Artificial intelligence systems, 234–235 Asbestos, 30 Ascorbic acid, 138, 174, 233 Aspartate transcarbamylase, 108 Aspirin, 192, 229, 232 Ataxia telangiectasia, 68 Atezolizumab, 212 Atherosclerosis, 137 ATM, 33, 35, 66, 68, 69 ATR inhibitor, 76, 79 AUC, 173, 174 Aurora kinase, 38, 69, 89–92, 99, 233 Autoimmune disease, 136, 211, 214, 247 Autophagy, 19, 91 5-azacytidine, 159, 174

B Bach1, 148 Base excision repair, 66, 67, 157 Basement membrane, 10, 14, 36–39, 185, 191, 193, 194, 199, 257 BAX, 18 BCG, 207, 209 Bcl-2, 207, 208 Bcr-Abl translocation, 108, 132–133, 142, 147– 149 Benign hyperplasia, 52–53, 189, 252

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 R. C. Jackson, Evolutionary Dynamics of Malignancy, https://doi.org/10.1007/978-3-031-32573-1

263

264 β-catenin, 10, 52 Bevacizumab, 186, 210 Bicalutamide, 119, 197–199, 236, 238 Big bang theory, 196 Biological response modifier, 209, 219 Bladder cancer, 209, 212 B-lymphocytes, 207 Bone morphogenic proteins (BMP), 200 Boolean kinetics, 72, 98 Breast cancer, 15, 16, 51, 57, 58, 68, 92, 117, 122, 123, 156, 157, 193, 209, 232 BUB1, 32, 69, 92 Burkitt’s lymphoma, 6, 29, 31, 32, 51, 111, 249, 251 Busulfan, 111, 144, 145

C Cadherin, 14, 193, 194, 232, 258 Calcineurin, 12 Calmodulin, 12 Camptothecin, 79 Cancer enzymology, 3–5 Cancer Genome Atlas, The, 34, 35, 157, 184 Cancer genome projects, 33–35 Cancer stem cells, 31, 38, 54, 181–183, 190, 196, 225, 249, 252, 255–257, 259 Cancer vaccines, 212–213, 225, 241, 255 Carboplatin, 80, 127, 128 Carcinoma in situ, 32, 54, 191, 257 Carmustine, 41 CAR-T cells, 213, 214, 241 Caspases, 17 CD20, 209 CDKN2A, 29, 33 Cell cycle, 9, 19, 28, 47–50, 53–60, 65, 66, 68– 71, 73, 74, 77–79, 81, 86–88, 91, 97, 124–126, 158, 163, 166, 182, 211, 219, 227, 236, 247 Cell division, 6, 8, 9, 14, 16, 18, 20, 22, 37, 38, 46–49, 51, 52, 55, 66, 70, 81, 88, 90, 91, 95–97, 104, 105, 119, 127, 148, 156, 181–183, 185, 186, 189, 191, 196, 199, 215, 232, 240, 251, 252, 259 Cell loss factor, 39, 49, 81, 96, 105, 124, 126, 252 Cellular immunotherapy, 213–214, 241 Centriole, 86 Cervical carcinoma, 28, 52 Cetuximab, 209 Chemical carcinogenesis, 27–28 Chk1 inhibitor, 75, 79, 82

Index Chromosomal instability, 20, 30, 32, 35–37, 39, 49, 52, 54, 69, 92, 107, 122, 156, 158, 181, 182, 227, 248, 249, 251, 253, 255–259 Chromosomal rearrangement, 2, 29–32, 93, 98, 99, 137, 179, 180, 182, 199, 231, 248, 249, 253 Chromothripsis, 34, 196 Chronic myeloid leukaemia (CML), 12, 29, 37, 108, 111, 132–149, 174, 198, 231, 233, 249, 250, 253, 256, 257 Chronic myelomonocytic leukaemia (CMML), 37, 154–155, 157–175, 233, 250, 253, 257 Cimetidine, 230 Cisplatin, 67, 115, 116, 127, 128, 252 c-myc, 6, 9, 29, 31, 51, 137, 188, 249, 251 CNDAC, 166, 170–172 Cohesin, 87, 89, 92 Colcimide, 87, 90 Colitis, 137 Collagen, 14 Collateral sensitivity, 110 Colorectal cancer, 69, 156, 193, 209, 212 Combination chemotherapy, 106, 121–122, 197, 215 Complete response (CR), 108, 122, 212, 247 Contact inhibition, 14 Cortexolone, 187 CpG islands, 155 Cross-resistance, 109, 110, 226 c-src, 28 CTLA-4, 211 Cutaneous T-cell lymphoma, 162 Cyclin B, 53, 55, 69, 72, 74, 75, 77, 86 Cyclin D, 7, 13, 50–54, 56–58, 133, 163, 166, 249 Cyclin-dependent kinases (cdks), 47, 50, 52, 53, 141 CYCLOPS, 48, 53–57, 70, 71, 77, 79–81, 97 Cyclosporin A, 207 Cyclotherapy, 57, 254 CYP3A4, 121 Cytidine deaminase, 158, 159, 173, 174 Cytochrome c, 18, 188, 190 Cytochrome P450, 197 Cytokine signalling, 6, 9, 11 Cytokinesis, 38, 48, 70, 87, 90 Cytokinetics, 48, 121, 158, 174 Cytoxan, 106

D Dasatinib, 139 Death receptors, 18, 19

Index Decitabine, 158–160, 166, 170–174 Dendritic cells, 135, 208, 216 Deoxycytidine kinase, 51 2-deoxyglucose, 117, 226 Diacylglycerol (DAG), 12 Differentiation, 2, 3, 7, 16, 27, 46–48, 54, 134, 136, 143, 144, 154–157, 159–162, 173, 175, 182, 185 Dihydrofolate reductase (DHFR), 21, 22, 108, 109 Dimethylxanthenone acetic acid (DMXAA), 209 DNA cross-linking, 115, 127, 144 DNA damage response (DDR), 66, 68–72, 79, 81, 82, 182, 252, 256 DNA-dependent protein kinase (DNAPK), 66, 68 DNA glycosylase, 67 DNA helicase, 66 DNA ligase, 47, 66–69, 72 DNA methylation, 155–158, 161, 163, 174, 175, 225, 231, 255 DNA methyltransferase (DNMT), 155, 157 DNA mismatch repair, 32, 67, 69, 156, 224 DNA polymerase, 60, 66–68, 224 DNA repair, 28, 30, 32, 33, 65–69, 73, 82, 96, 99, 117, 156, 183, 252 Dostarlimab, 213 Double strand breaks, 66 Doxorubicin, 117, 121 Doxycycline, 230 Driver mutations, 33, 34, 61, 68, 157, 184, 196, 225, 226, 252, 253, 255 Drug resistance, 104–119, 122–125, 127, 139, 147, 193, 194, 196–198, 212, 213, 215, 226, 227, 236, 239, 240, 247, 252 Dutasteride, 197

E EGF, see Epidermal growth factor (EGF) EGF receptor, 9, 51, 214, 236 Endotoxin shock, 137 Enzalutamide, 119, 197, 199, 236–239 Epidermal growth factor (EGF), 7, 9, 13, 47, 51, 56, 57, 59, 189, 214, 236 Epigenetic changes in malignancy, 155–157 Epigenetic gene silencing, 138, 157, 253 Epithelial-mesenchymal transition (EMT), 185, 191–192, 194, 196, 199, 200, 256–259 Epstein-Barr virus (EBV), 29, 249 ERK2, 188, 189 Erythropoietin, 162 Etoposide, 127, 252 Extracellular matrix, 15, 185, 191, 193, 232, 257

265 F Facilitated diffusion, 109 FAS ligand, 208 Fibroblast growth factor, 47, 185 Fibrosarcoma, 28 Finasteride, 197, 198 Finite state machine, 39, 93–97, 148 Flavone acetic acid (FAA), 186, 209, 210 Fluctuation equation, 104 5-fluorouracil, 209 Focal adhesion kinase (FAK), 10, 14, 191 Folic acid, 21, 232 Follicular dendritic cells, 208 Folylpolyglutamate synthetase, 109 Fos, 9

G Gemcitabine, 57, 58, 66, 77–80, 82 Gene silencing, 46, 138, 155, 156, 253 Genetic algorithm, 31, 36–37, 54, 124, 125, 180, 181, 198 Glioblastoma, 34, 162, 190, 213 Glioma, 41, 214, 256 Gluconeogenesis, 4, 184 Glutathione, 132, 138, 140 Glycolysis, 4, 187–190, 192, 199, 214, 233, 234, 253, 256, 258 GM-CSF, 134, 135, 137, 141, 142, 144, 162, 163, 166 Gompertz equation, 15 G protein coupled receptors (GPCRs), 7, 12 Granulocyte colony-stimulating factor (G-CSF), 134, 162 GREM1, 200 G1:S checkpoint, 9, 35–39, 47, 49–52, 54–57, 60, 68, 96, 98, 99, 158, 163, 182, 249, 251, 252, 254, 256 Guanine O6-methyltransferase, 41, 66

H Haematopoietic stem cells (HSC), 67, 112, 133, 148, 157, 159, 162, 163, 171, 181 Hallmarks of cancer, 17–20, 26, 30, 211, 227, 249, 251, 257 Haploinsufficiency, 160, 171, 172, 175 Hayflick limit, 20 Hepatoma, 3, 4, 181, 184, 250 HER2, 9, 51, 209 Hexamethylenebisacetamide, 54 Hexokinase, 188, 233, 234 HIF-1α, 185

266 Histone, 87, 154–157, 161, 197 Histone deacetylase (HDAC), 155, 156, 161, 162 HIV disease, 119, 122, 208 HLA-G, 211, 212, 214 Homologous recombination (HR), 66, 68, 69, 78, 79 H-ras, 39 Human papilloma virus (HPV), 28, 52, 96, 212, 213 Human T-cell leukaemia viruses (HTLV), 28 Hypochlorite, 133, 136, 138, 206 Hypomethylating agents, 158–162, 166, 170–174 Hypoxanthine/guanine phosphoribosyltransferase (HGPRT), 108–110 Hypoxia, 16, 19, 34, 117, 122, 185, 188, 189, 191, 194, 199, 227, 233, 253, 256, 258

I IGHα, 29, 249 ILT4, 212 Imatinib, 139–142, 144–147, 233, 250, 253, 256 Immune checkpoints, 199, 210–214, 234, 253, 255, 256, 258 Immune surveillance, 111, 210–211, 214, 227, 231, 253, 256 Immunosuppression, 117, 220 Immunotherapy, 208–210, 212, 214–216, 219 Innate immune response, 29, 132–137, 147, 220 Inosine 5'-phosphate (IMP), 4, 21 Inosine 5'-phosphate dehydrogenase (IMPDH), 4 In silico clinical trials, 120, 228, 235–239, 241, 247 Integrins, 6, 8–11, 14, 191, 192, 194, 257 Interferons, 11, 209, 210, 220 Interleukin-2, 207, 209, 220 Interleukin-4, 207 Interleukin-6, 207 Interleukin-10, 220 Interleukins, 11, 133 Interspecies scaling, 119–121 Intrinsic resistance, 106 Invasion, 17, 19, 136, 191–194, 227, 232, 236, 249 Ionizing radiation, 27, 91, 224 Ipilimumab, 212 Irinotecan, 79

Index ISLR, 200 Isocitrate dehydrogenase (IDH), 157 Itraconazole, 230 Ivosidenib, 157

J Janus-acting kinases (JAKS), 11 Jun, 9, 10

K Kaplan-Meier plot, 199, 238, 239 Keap1, 148 Kinetochore, 86, 89–92, 97 K-ras, 9, 39, 226 Krestin, 209

L Laminin, 14 Leukaemia, 2, 11, 12, 28, 31, 37, 39, 88, 110, 111, 117, 118, 137, 139, 143, 154, 157, 160, 161, 175, 182, 194, 207, 210, 211, 214, 220, 225, 249, 251, 257, 258 Levamisole, 209, 219 Lipopolysaccharide (LPS), 135, 206 Logic gates, 7, 60 Lung cancer, 51, 120, 148, 212, 246 Lyme disease, 136 Lymphoma, 2, 6, 29, 31, 32, 51, 111, 162, 249, 251, 257 Lynch syndrome, 32

M Macrophages, 132–137, 157, 162, 173, 206, 208–210, 215, 216, 220 Major histocompatibility complex, 208 Malignant progression algorithm, 36, 37, 54 MAP kinase pathway, 6, 7, 9, 10, 13, 56, 189 Markov process, 248 May equation, 16, 17 Mcl-1, 18, 133, 137, 140, 141, 145, 147, 206, 249, 257 Mebendazole, 230 Melanoma, 6, 27, 29, 33, 49, 51, 207, 210, 212, 213 Memory cells, 207, 208, 215, 218 6-mercaptopurine (6-MP), 106, 108–110 Mesothelioma, 30 Metaphase, 86, 87, 90

Index Metastasis, 17, 19, 20, 26, 30, 35, 118, 148, 157, 192–196, 198, 199, 215, 227, 228, 236, 249, 253, 254, 258, 259 Metformin, 160, 189, 214, 230, 232, 233, 235 Methotrexate, 20–22, 41, 108–110, 118, 194, 195 MHC haplotype, 215 Microenvironment, 16, 184, 190, 196, 199, 211, 214, 225, 256, 259 Microhomology-mediated end joining (MMEJ), 66 Microsatellite instability, 32, 67, 156 Microtubules, 86–91, 192 Minimal residual disease, 118, 215, 216, 219 Mitosis, 19, 35, 38, 48, 66, 68–70, 72, 74, 75, 77, 79–81, 86, 87, 90–92, 97, 124, 252 Mitotic arrest, 87, 90, 97 Mitotic catastrophe, 19, 81, 90–92 Mitotic index (MI), 87–88, 90 Mitotic spindle, 32, 47, 48, 54, 55, 69, 86–88, 252 Monoclonal antibodies, 186, 209, 210 Monocytes, 133–137, 154, 157, 160–175 Monotelic attachment, 89 Multidrug resistance, 109 Mutagens, 26, 28, 30, 138–139, 144 Mutation rate, 26, 30, 37–39, 49, 94–96, 104–106, 111, 114, 116–119, 122, 123, 127, 139, 144, 146–149, 156, 181, 191, 194, 197, 224, 233, 240, 248, 249, 252 Myelodysplastic syndrome (MDS), 154, 158, 160–162, 172, 175, 250 Myeloperoxidase, 133, 138, 206 Myelosuppression, 41, 117, 118, 120, 127, 220

N Nasopharyngeal carcinoma, 29 Natural killer (NK) cells, 206, 207, 209, 214, 216, 220 NBS1, 66, 68 Necrosis, 17, 19, 123, 133, 186, 209, 210, 220 Neoadjuvant chemotherapy, 122, 229 Neutral evolution, 195, 253 Neutrophils, 41, 132–138, 140–142, 147, 161–164, 206, 207, 220 Neutrophil trafficking, 134 NFAT, 12, 207 NFκB, 12, 234 Nilotinib, 139 Nitrosoureas, 41, 67 NK cells, see Natural killer (NK) cells Non-Hodgkin lymphoma, 29, 209

267 Non-homologous end joining (NHEJ), 66, 68, 78, 79 Non-small cell lung cancer, 9, 162 N-ras, 39, 157, 163, 166–170, 173–175 Nrf2, 145, 148, 194, 198 Nucleotide excision repair, 27, 67, 71 Nutlin-3, 255

O Oesophageal cancer, 51 Okadaic acid, 77 Olaparib, 68, 74, 79–82 Oncogene addiction, 6 Oncogenes, 5–6, 9, 28, 31, 32, 60, 122, 209, 210, 227, 249–251, 255 Ovarian carcinoma, 34, 190 Oxidative phosphorylation, 4, 19, 137, 187–190, 199, 214, 233, 258 Oxidative stress, 29, 49, 60, 148, 194, 233 Oxidised glutathione, 140 8-oxo-guanine, 28, 138, 142, 147, 224

P P53, 49, 51, 58, 60, 184 Paclitaxel, 97, 111–117 PAK, 191, 232 PALA, 108, 112–116 Palbociclib, 53, 57–59 Pancreatic cancer, 69, 82, 213 Paneth cells, 182 Papilloma virus, 37 Partial response (PR), 108 Passenger mutations, 4, 184, 250 PD-1, 135, 210, 212 Pembrolizumab, 135, 210, 212, 213 Percent labelled mitosis (PLM), 87 Peroxiredoxin 3, 138 P170 glycoprotein, 109, 117 Philadelphia chromosome, 132–133 Phorbol ester, 12, 27, 28 Phosphofructokinase, 233, 234 PI3 kinase pathway, 6, 7, 10 PKC signalling, 12 Platelet-derived growth factor (PDGF), 7, 10, 14, 47, 185, 198 Poly-(ADP ribose), 66, 72 Poly-(ADP ribose) polymerase (PARP), 66, 68, 70, 72, 74, 79, 80, 82, 157, 227 Polyglutamates, 21, 109 Polymerase chain reaction (PCR), 108, 139, 230

268 Ponatinib, 139 pRb, 50, 52, 53, 55, 56 Premalignancy, 6, 32, 52–53, 96, 98, 161, 175, 191, 198, 252 Procaspases, 18, 19, 74 Prodrugs, 108, 170 Progenitor cells, 41, 133, 134, 137, 140–145, 147–149, 154, 159, 162–164, 166, 173, 181, 182 Progressive disease (PD), 97, 108, 120, 140, 141, 226, 227, 235, 236, 248 Proliferating cell nuclear antigen (PCNA), 67, 68 Prostate cancer, 35, 68, 69, 123, 189, 196–200, 226, 229, 232, 237, 239, 247, 248 Prostate-specific antigen (PSA), 123, 196, 226, 229, 230, 235, 247, 248 Protein kinase A (PKA), 234 Protein kinase C (PCA), 6, 12, 28 Proteosome, 49 Proto-oncogenes, 5, 31, 210, 249 PTEN, 14, 158, 188, 249 Pyruvate kinase, 188, 189, 234

R RAD51, 68, 69, 78 Radiation carcinogenesis, 251 Raf, 7, 12, 28, 34, 186 Raptor, 233 Ras genes, 95 Reactive oxygen species (ROS), 28, 132, 133, 135–138, 140–142, 144, 145, 147, 148, 189, 206, 225, 233, 234, 249 RECIST criteria, 108 Regulatory T cells, 207, 211 Renal cell carcinoma, 186 Replication catastrophe, 70 Replication protein A (RPA), 66, 69, 70, 74, 76, 78 Replicative stress, 35 Repurposing, 229 Retinoblastoma, 50 Revertants, 105, 116, 117 Ribonucleotide reductase, 9, 51, 159 Rituximab, 209 RNA polymerase, 53, 71, 86, 155, 156 Rous sarcoma virus, 28

S SAC, see Spindle assembly checkpoint (SAC) Salvage pathway, 110, 118

Index Sapacitabine, 170, 172 Sarcoma, 106, 258 SEER database, 142, 154 Seliciclib, 53, 141, 142, 145, 146 Sézary syndrome, 162 Signalling pathways, 6–14, 28, 32, 33, 47, 51, 52, 56, 60–61, 91, 122, 193, 207, 225, 232, 236 Single-strand breaks (SSB), 66, 68, 71, 75, 79–81 Sirtuins, 161 Sister chromatids, 66, 89, 180 Sorafenib, 6, 9, 186 Sotorasib, 9 Speciation, 30, 31, 37, 93, 99, 251 Spindle assembly checkpoint (SAC), 32, 33, 35, 38, 39, 47, 48, 52, 54, 55, 69, 87–93, 96–99, 182, 233, 252, 254, 256 Spindle poisons, 90, 91, 122 Stable disease (SD), 108 Stalled replication forks (SRF), 66, 70, 71, 73, 74, 76, 77, 79, 81 Statins, 230 STATs, 9, 11, 12, 28, 249 Stem cells, 21, 27, 38, 41, 46, 53, 94, 99, 112, 114, 118, 120, 133, 148, 149, 157, 180–183, 190, 208, 220, 230, 247, 253, 254, 256, 257, 259 STING, 209 Stochastic modelling, 247 Sunitinib, 186 Syntelic attachment, 89 Synthetic lethality, 68, 79, 81, 110, 227

T Taxanes, 87, 90, 98, 121 Telomerase, 20, 155, 183 Telomere, 20, 34, 155, 183 Tension sensing, 87, 89 Testicular cancer, 127, 252 TET2, 157, 158, 160, 163, 165, 166, 168–175, 253 Tetrahydrouridine (THU), 159, 173, 174 TGFβ, 192, 193, 200, 208 T-helper cells, 207, 208 Therapeutic index, 21, 116 6-thioguanine (6-TG), 110 Thioredoxin, 138 Thrombopoietin, 162 Thymidine dimers, 27, 67

Index Thymidylate, 21, 22, 118, 159 Time to progression (TTP), 73, 75, 76, 108, 123, 125, 126, 146, 148, 229, 240, 252 T lymphocytes, 28, 207 TNF (TNFα), 18, 29, 132, 133, 135, 206, 220 Tolerance, 118, 192, 207, 216 Topoisomerase, 47, 66, 79, 81 TRAMP model, 198 Transcription factor, 6, 8–10, 12, 28, 29, 31, 34, 47, 49, 50, 55, 58, 60, 62, 71, 92, 133, 137, 148, 156, 185, 188, 194, 197, 225, 249, 258 Transgenic adenocarcinoma of mouse prostate (TRAMP) model, 197 Transit cells, 181, 182 Trastuzumab, 9, 209 Tricarboxylic acid cycle (TCA cycle), 174, 190, 234 Trimetrexate, 110 T-suppressor cells, 208 Tubulin, 86, 87, 90, 97, 98 Tumour heterogeneity, 32–33, 213, 227 Tumour progression, 5, 10, 18, 19, 30, 32, 33, 35, 49, 69, 91, 94, 95, 111, 123, 148, 157, 159–161, 179–189, 192–195, 197–199, 211, 213, 216, 225, 227–235, 248–251, 253–256, 258 Tumour suppressor genes, 5–6, 29, 35, 49, 50, 52, 68, 92, 156, 158, 247, 249, 251 Two checkpoints model of cancer, 40, 52 Tyrosine kinase, 5, 9, 11, 28, 34, 132, 133, 139, 186, 249

269 U Ultraviolet (UV) radiation, 27 Uracil DNA glycosylase, 224

V Vascular endothelial growth factor (VEGF), 7, 10, 47, 185, 186, 198, 210 Vascular targeting agents, 186, 210 VEGF receptor, 186 Verapamil, 109, 117, 226 Vinca alkaloids, 87, 90 Vincristine, 91, 117 Viral carcinogenesis, 28–29 Virtual clinical trials, 215 Vitamin C, 138, 232 Von Hippel-Lindau tumour suppressor gene, 234 Vorinostat, 162 v-src, 28

W Warburg effect, 160, 187–191, 233, 258 Wart, 37, 52, 54, 96 Wee-1, 72 Werner’s syndrome, 50 Wnt signalling, 6, 10, 32, 52, 182, 185

X Xeroderma pigmentosum, 67 X-irradiation, 59 XRCC1, 66 XRCC4, 69