Biomarkers of the Tumor Microenvironment 3030989496, 9783030989491

Citation preview

Lars A. Akslen Randolph S. Watnick Editors

Biomarkers of the Tumor Microenvironment Second Edition

123

Biomarkers of the Tumor Microenvironment

Lars A. Akslen  •  Randolph S. Watnick Editors

Biomarkers of the Tumor Microenvironment Second Edition

Editors Lars A. Akslen Centre for Cancer Biomarkers CCBIO Department of Clinical Medicine University of Bergen Bergen, Norway

Randolph S. Watnick Department of Surgery Harvard Medical School Vascular Biology Program Boston Children’s Hospital Boston, USA

ISBN 978-3-030-98949-1    ISBN 978-3-030-98950-7 (eBook) https://doi.org/10.1007/978-3-030-98950-7 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2017, 2022 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

The editors would like to dedicate this book to the memory of the late Judah Folkman, who through his visionary research and mentoring inspired the concept that the tumor microenvironment is a critical component of tumor biology. The editors would also like to express their deepest appreciation to their families for their continuous patience and support. Dr. Akslen specifically thanks his wife Åse, daughter Heidi, son Andreas, and in particular his grandson Hans, for constant joy and inspiration. Dr. Watnick thanks his wife Jing, daughter Audrey, son Eytan, and father David for their unconditional love and encouragement.

Foreword

Professor Lars A.  Akslen (Director, Center for Cancer Biomarkers CCBIO, University of Bergen) and Professor Randolph S. Watnick (Boston Children’s Hospital, Harvard University) have succeeded in the formidable task of assembling an extensively revisited book on the tumor microenvironment. This new edition deals with the most important aspects of the tumor microenvironment, providing an in-depth analysis of the interactions that take place between normal and malignant cell types. This new edition provides a rather exhaustive scrutiny of the phenotypes and organizations of components of the vascular systems, decoding the roles enacted by stromal fibroblasts, inflammatory cells, and immune cells and offers noteworthy insight into innervation within the tumor microenvironment. This book also provides in-depth insight into the clinical relevance of strategically important signaling systems. The advent of new technologies is also discussed, with a topical reflection on how these technologies are being engineered for the discovery of new biomarkers toward the optimization of clinical trials. Such technological improvements have also unveiled the advanced topological features of the tumor bed and, most importantly, identified a broad swathe of relevant biomarkers that will be instrumental in discriminating responders from non-responders and in monitoring the early acquisition of refractoriness. The new edition of this book presents new concepts that can be exploited for the design and thoughtful optimization of new therapeutic strategies. Particularly attractive are the model systems described in this book, which reflect the N-dimensional complexities of tumors. Personalized and precise treatment will likely ensue from these forms of adaptive therapy. Indeed, it is hoped that these objectives will be facilitated soon with advanced bioinformatics approaches supported by artificial intelligence algorithms. Researching the tumor microenvironment is a daunting task: the array of disorganization allows for too many distinct topological organizations. Anatomists from the late 1800s and early 1900s were able to categorize most cell types and their three-dimensional organization within tissues and organs using the most rudimentary microscopes and staining techniques. The atlas of Mathias Duval of the chick embryo (Atlas d’embryologie (Masson, Paris, 1879)) has never been surpassed, and the remarkable treaties of the famous neurohistologist Santiago Ramon y Cajal cannot be matched today by even the most advanced imaging technologies (The cerebral cortex annotated translation of the complete writings DeFilipe and Jones, Oxford University Press, 1988). But these remarkable advances were only achievable in the early days because the analysis was performed on stereotyped structures. The latest studies seeking to revisit the anatomical organization and lineage tracing of such structures using single-cell sequencing have not revealed major new findings. Unlike such typical tissue architectures, the tumor microenvironment exhibits a formidable heterogeneity in terms of cell composition, phenotype, and 3D organization, with a complex extracellular matrix. Although clinical pathologists are able to recognize critical features that can distinguish tumor type and grade, the considerable variability in the organization of the tumor microenvironment is extremely challenging to describe in depth. How can we rationalize and conceptualize the tumor microenvironment to unravel the signaling networks and identify the Achille’s heel of these tumors for effective targeting? We know that the communication systems mediated in part by cytokines and growth factors cannot be modeled easily because their distribution patterns cannot be predicted by the physical laws of diffusion. Similarly, drugs and antibodies delivered into vii

viii

Foreword

the tumor bed are also largely affected by the strong local variation in stiffness and pressure. A tumor could be theoretically analyzed by soft matter physics; albeit, the heterogeneity of the tumor structure prevents the application of thermodynamic laws of polymers (Essentials of soft matter science CRC press, Brochard-Wyart, Nassoy and Puech, 2019). We hope that, to some extent, artificial intelligence will be able to rationalize these chaotic structures. Such a task, however, requires an international consortia to address the formidable issues we face. New mathematical models and algorithms will need to be developed, similar to those recently applied by particle physicists and cosmologists to decode the universe. What we need now is to identify another Jean Francois Champollion to decode this new Rosetta Stone. Jean Paul Thiery Guangzhou Regenerative Medicine and Health Guangdong Laboratory, Guangzhou, China Department of Clinical Medicine, Faculty of Medicine University of Bergen, Bergen, Norway CNRS UMR 7057, Matter and Complex Systems University Paris Denis Diderot, Paris, France Comprehensive Cancer Center Institut Gustave Roussy, Villejuif, France Chief Scientific Consultant, Chair SAB Biosyngen Pte. Ltd., Singapore, Singapore Scientific Founder and Chairman, BioCheetah Pte. Ltd., Singapore, Singapore

Preface

In 1889, Stephen Paget formally postulated his “seed and soil” hypothesis, building upon the previous observations of Ernst Fuchs. This seminal concept on tumor progression was based on the analogy that tumor cells were seeds and needed a proper soil to grow. However, the study of cancer for the next hundred years focused primarily on characterizing the morphologic and molecular aberrations unique to tumor cells. This field of study yielded remarkable insights into intrinsic tumor cell biology with dozens of oncogenes and tumor suppressors discovered. In contrast, very few papers dealt with the “tumor microenvironment” until the 1990s. However, the theory that tumors require a permissive environment or “tissue predisposition” to grow, both in the primary site and following dissemination to distant organs, is now commonly accepted. That being said, our understanding of the basic composition of tumors has gradually transformed from a collection of tumor cells that grow in an uncontrolled fashion to miniature tissues or “organs” comprised of a neovasculature and complex stroma consisting of both resident and bone marrow-derived cells. This conceptual shift has been extended by the observations that tumor cells interact with their microenvironment and, in so doing, affect and are affected by the reciprocal intercellular signaling. These signaling events were recognized as critical by Judah Folkman, who postulated in the late 1960s that tumor growth and spread were linked to the ability to induce angiogenesis. By relentlessly pursuing the underlying biology behind these findings, he helped foster the notion that the tumor microenvironment was not a passive bystander but an active collaborator in tumor progression. Today, pathologists commonly use microvessel density or vascular proliferation as biomarkers for determining the aggressiveness of tumors. Moreover, the first identified angiogenic factor, VEGF, and its growth promoting receptor VEGFR2, are targets for anti-cancer therapeutic agents used widely in the treatment of cancer patients. As the field matured, we discovered that many different cell types exist within the tumor microenvironment that possess both growth promoting and inhibitory roles in tumor progression. Among the most prominent cell types that affect tumor growth are immune cells. Harold Dvorak famously described tumors as “wounds that do not heal.” This seemingly simple statement has profound implications not only on how tumors are studied and perceived, but also on the manner in which they are treated. Advancements in immunotherapy have led to the development of therapeutic agents that re-engage the adaptive immune response and enable the immune system to attack tumors. Immune checkpoint inhibitors that block PD-1, PD-L1, and CTLA-4 have been approved for multiple indications and have shown potent and durable responses in a subset of patients. Therapeutic agents targeting CD47, a “do not eat me” signal that represents an inflammatory checkpoint protein blocking macrophage phagocytosis, are in clinical trials. It is now recognized that evasion of immune and inflammatory cells is not mediated solely by the expression of checkpoint inhibitors on the surface of tumor cells. Myeloid derived cells, including monocytes and macrophages, are potent suppressors of the immune and inflammatory response, as are cancer-associated fibroblasts (CAFs). Such cells might also repress the expression of thrombospondin-1, a potent anti-angiogenic, anti-inflammatory, and immunomodulatory protein. Recent findings have highlighted that angiogenesis and immune evasion ix

x

are co-regulated, further underscoring the complexity and significance of these interactions between tumor cells and the microenvironment. This textbook contains 34 chapters, divided in two main sections, that highlight the multi-­ dimensional and complex nature of the tumor microenvironment and its role in tumor progression, biomarker development, and therapeutic targeting. The first section deals with basic mechanisms and biomarkers of the tumor microenvironment (TME) and its various components. The second section delves into examples of organ directed biomarker studies and clinical applications. Following a masterful overview with integrated perspectives on this field by Robert Weinberg, Watnick (Chap. 1) sets the stage and focus on the importance of microenvironmental context in the regulation of tumor angiogenesis. It is argued that paracrine signaling should be considered in the search for tumor progression drivers and novel targets along with companion biomarkers of potential clinical importance. The identification of targets within the TME is envisioned to be complementary to the search for mutations and aberrant signaling in tumor cells. In Chap. 2, Akslen continues along the same theme to discuss how tissue-based markers of angiogenesis and vascular invasion can be defined and applied in studies of human cancers and how their aggressive behavior can be graded by such markers, while Chaps. 3 and 4 see Milosevic et al. and Wagner & Wiig delve into the roles of vasculature and the lymphatic system in tumor growth and progression. In Chaps. 5 and 6, Zeltz et al. and Legget & Nelson outline the dual functions of the extracellular matrix and tissue architecture as supportive or inhibitory with respect to cancer progression. Especially, the role of the insoluble extracellular matrix (ECM) is discussed, and how its components influence matrix remodeling, tumor metastasis, and even tumor heterogeneity. This theme is continued in Chap. 7 as Dongre and Costea explore the roles of cancer-­associated fibroblasts (CAFs), the major producers of ECM proteins in the TME, In Chap. 8, Strell and Östman describe the paracrine interactions between mesenchymal cells and epithelial or endothelial cells, with particular reference to the PDGF family of growth factors and receptors. By combining experimental and clinical studies, the PDGF signaling systems appear as critical regulators of tumor growth, metastasis, and drug efficacy. In Chap. 9, Gilligan et al. elaborate further on the function of lipid signaling and the dual role of inflammation in cancer. The complex interactions between various classes of immune cells and how these appear to be regulated by fatty acid-derived lipid mediators such as prostaglandin E2 are discussed. Chapter 10 sees Kushekhar et al. delineate the multiple roles of lymphocytes in immunity and immune evasion. They explore mechanisms of immunoediting, immunosuppressive cytokines, and the multiple types of lymphocytes, their markers and function in tumor progression and treatment. In Chap. 11, Brekken and Wnuk-Lipinska discuss the regulation and relationship of epithelial plasticity (EMT programs) and immune escape mechanisms. The authors focus on molecules that can drive the immunosuppressive state in the tumor microenvironment and potentially serve as biomarkers for poor prognosis. Continuing on the role of the immune and inflammatory system in Chap. 12, Corthay and Haraldsen comment on a range of inflammatory biomarkers in cancer such as cytokines and interleukins converging on STAT3 signaling. In particular, this chapter discusses the biology of IL-33, the most recently identified member of the interleukin family. In Chap. 13, Floris et al. describe some tissue-based biomarkers of the immune response in solid tumors, such as tumor infiltrating lymphocytes (TILs) and tertiary lymphoid structures (TLSs), and how these can be recorded in human tumor tissues. In Chap. 14, Ramchandani et al. explore the role of the bone marrow and bone marrow-­ derived cells on tumor progression and metastasis. They discuss how bone marrow-derived cells constitute a significant fraction of the primary tumor microenvironment and processes such as angiogenesis as well as metastasis and growth in distant sites. Chapter 15 introduces the role of platelets in tumor progression. Guo et al. describe the intricate composition of platelets into multiple types of granules and the signals that elicit release of pro- and anti-tumorigenic proteins from these granules.

Preface

Preface

xi

In Chaps. 16 and 17, Vethe et al. and Balasubbramanian et al. explore the nascent and exciting field of neurogenesis and neurogenic factors in tumors. Vethe et al. discuss the biology of nerve fiber infiltration in tumors and their role in influencing tumor growth and progression via specific tumor–nerve interactions. Balasubbramanian et al. focus on the roles of two proteins originally identified in neurogenesis, neuropilin 1 and 2. They outline the novel observations that these receptors are also expressed on tumor cells and discuss their functional role in tumor progression and as biomarkers. In Chap. 18, Lotsberg et al. revisit epithelial plasticity and EMT, focusing on roles of the Axl receptor tyrosine kinase in tumor cell plasticity and tumor progress. In particular, the authors focus on the relationship between EMT programs, immune evasive phenotypes, and drug resistance, and how this suggests a potential for anti-Axl combination therapy in a range of aggressive cancers. Continuing the theme of advances in cancer treatment, in Chap. 19, Kleinmanns et  al. provide an overview of the advances in patient-derived xenografts. They discuss their utility in modeling disease progression, identifying novel biomarkers, and testing clinical and pre-clinical stage therapeutic agents. Herdlevær et al. provide an introduction to imaging mass cytometry as a method for deep interrogation of tissue biomarkers in Chap. 20. Specifically, they explain how this relatively new technique can assay up to 40 discrete biomarkers at once using metal affinity tagged antibodies. They discuss how researchers and clinicians are using this technology to gain unprecedented insights into the composition of the TME in naïve and treated tumors. To conclude Section 1, Chap. 21 sees Pedersen et al. explore the exciting technology of artificial intelligence and how it is being implemented to discern molecular and pathological trends and patterns in malignant tumors. Kicking off Section 2, which focuses on organ related studies and clinical applications of TME biomarkers, in Chap. 22, Moses et al. discuss the exploration of the TME for therapeutic targets. Specifically, they delineate the utility of the detection of biomarkers: exosomes, proteins, nucleic acids, lipids, miRNA, and cells in bodily fluids and how the modulation of these markers can be prognostic for disease outcome. In Chap. 23, Wik et al. explore the use of gene expression signatures of the tumor microenvironment in breast cancer. The authors discuss how composite signatures can be used as biomarkers and clinical tools to capture and reflect the complexity in human tumors that are not attainable using individual markers. In Chap. 24, Kim and co-workers delineate the use of non-invasive magnetic resonance imaging (MRI) and spectroscopy (MRS) to measure dynamic biomarkers that can be used to characterize alterations in tumors during treatment and follow-up. Importantly, they discuss how contrast-enhanced MRI methods can aid in the evaluation of tumor vascularization and function. In Chap. 25, Jokela et al. propose a dominant role of the microenvironment in tumor progression. The authors explore the influence of tissue architecture on drug responses, by focusing on applications and analytic approaches used for functional cell-based exploration of combinatorial microenvironments using microarray technology. In Chap. 26, Azeem et al. discuss the establishment of novel prostate cancer models, their applications, and their critical role in understanding disease progression and therapeutic strategies. Biomarkers of prostate cancer are further explored in Chap. 27 by Magnussen and Mills. Specifically, they advocate for the use of biomarkers, and the necessity for informed patient consent, to monitor the progression of prostate cancer and explore LRG-1 as a promising marker. In Chap. 28, Rogers draws insightful comparisons between the microenvironment of endometriosis and cancer and explores the idea that similar or analogous biological mechanisms and pathways may mediate both diseases. He proposes that rigorous experimental analysis of endometriosis is necessary for the field and that tools employed in cancer research can be utilized for this purpose. In Chap. 29, Ramnefjell and Akslen make the provocative case that angiogenesis and neovascularization are underutilized as biomarkers for non-small cell lung cancer (NSCLC). They further advocate that based on biomarker data and improved patient stratification, anti-angiogenic therapy might be more widely used to treat NSCLC patients.

xii

Preface

In Chap. 30, Leiss and colleagues expound on the specific features of tumor–host interactions in malignant gliomas. They explore how interactions are shaped by the structural ­organization of the CNS and involve multiple cell types, extracellular matrix components, and host cell-derived soluble factors that are unique to the CNS. In Chap. 31, Dillekås et al. discuss the importance of tumor stress responses and the TME in progression of melanoma and other tumors to the metastatic stage. They focus on the dual nature of the melanoma TME as having both stimulatory and inhibitory properties, as well as the involvement of the immune system. In Chap. 32, Wang and Dudley return to the brain tumor microenvironment to shed light on the exciting and still poorly understood field of vessel cooption. They first provide clinical evidence of vascular cooption and then explore experimental models. The authors postulate that the underlying biology of brain cancers and their often diffuse invasion patterns necessitate vessel cooption over neovascularization. In Chap. 33, Jebsen and co-workers outline the most active pathways to therapy development: mutation driven drug development, immunomodulatory therapy, and evolution of conventional chemo- and radiotherapy. In this context, they promote the merits of personalized medicine and argue that successful implementation requires more precise biomarkers, not only to increase precision and enhance efficiency but also to avoid unnecessary toxicity for the patient, and costs for the society. Section 2 and this text book conclude with a masterful treatise of the requirement for and development of novel biomarkers of the TME (Chap. 34). Sim et al. provide an overview of the myriad technologies and “omics” that can be marshalled to enhance the identification of new biomarkers and explain why they are needed. Finally, they advance the notion that “real-time” biomarkers are the next major development and their utility in monitoring both disease progression and treatment efficacy. It would be impossible to provide an exhaustive and thorough analysis of this rapidly expanding field in a single volume. Still, we hope that readers find this second edition useful. As the field of identifying and utilizing biomarkers of the tumor microenvironment rushes forward at a breakneck pace, we will be presented with a myriad of exciting and novel treatment targets and companion biomarkers. We must remember, that as is the case with any rapidly growing field, the challenge is to integrate the newly generated knowledge into the evolving practice of medical oncology and precision medicine. This challenge demands that those whose work focuses on the basic sciences work diligently to translate their innovative approaches and discoveries with special focus on working closely with their colleagues in clinical practice and research to continually improve trial design and follow-up of patients. Finally, we thank Springer Nature for allowing us the opportunity to work together again on this project. As longtime collaborators and friends who share a common passion for unlocking the mysteries and therapeutic potential of the tumor microenvironment, it has been truly awe inspiring to curate the remarkable advances made in this field. As is the case with any undertaking of this size and complexity, we, the editors, could not have successfully completed it without the help of several important contributors. We would like to thank the staff at Springer Nature, as well as our own staff, for their assistance and valuable advice. Bergen, Norway Boston, MA, USA 

Lars A. Akslen Randolph S. Watnick

Biomarkers of the Tumor Microenvironment—A Prologue

Until a quarter century ago, a powerful reductionist paradigm held sway in the field of cancer research: those interested in studying the mechanisms of cancer pathogenesis embraced the notion that the biology of tumors could be understood by analyzing the biology of the constituent cancer cells. Moreover, the biology and pathophysiology of individual cancer cells could be understood, in turn, by studying their genomes, more specifically cancer-associated somatically mutated genomes. This was, to be sure, a powerful model, in that it led to the discovery of the genetic determinants of cancer pathogenesis, including oncogenes, tumor suppressor genes, and yet other genes involved in DNA repair and apoptosis. As useful as this paradigm was, it overlooked an important aspect of cancer pathogenesis: tumors are histologically complex structure composed of multiple distinct cell types, a reality recognized by pathologists for more than a century. Accordingly, beginning in the late 1990s, it became increasingly clear, even to the most committed reductionists, that tumors were functionally far more complex than aggregates of cancer cells. Thus, as tumors develop, it became apparent that neoplastic cells rely on recruited normal host cells for various types of cell-­ physiologic support. The latter cells had been termed stroma by the pathologists. More detailed characterization of the stroma revealed that it consists, at least in the case of common carcinomas, of a diverse collection of cells, virtually all of which are of mesenchymal origin. Included in the stroma are cells that often form its bulk, including fibroblasts and myofibroblasts, the latter often termed carcinoma-associated fibroblasts (CAFs). Interwoven among these stromal connective tissue cells are a variety of cells of hematopoietic origin, including endothelial cells, various subsets of lymphocytes, macrophages, and occasional granulocytes. As carcinomas develop and progress to higher grades of malignancy, the stroma usually changes in lockstep, becoming increasingly “reactive” and thus assuming a biological state that exists only transiently in the wound sites within normal tissues that are in the midst of healing. Indeed, such reactive stroma increasingly resembles “wounds that do not heal.” The coordinated changes of neoplastic cells together with adjacent stroma provided, on its own, a clear indication that the two groups of cells intercommunicate, doing so via processes that are often termed heterotypic signaling, i.e., communication between distinct types of cells. In principle, this signaling might be unidirectional, in that, as an example, the neoplastic cells within a carcinoma might release signals that recruited a diverse array of stromal cells to the growing tumor and thereafter orchestrated their behavior. In truth, however, the heterotypic signaling is bidirectional, in that recruited stromal cells release signals that impinge reciprocally on the carcinoma cells that previously recruited them. Hence, the co-evolution of neoplastic cells and the co-opted host cells is enabled by bidirectional signaling. Importantly, while the neoplastic cells undergo both genetic and epigenetic evolution, the evidence to date indicates that the recruited stromal cells—which together form the “tumor microenvironment”—undergo phenotypic changes that are not driven by somatic mutations. In fact, the histopathological appearance of islands of tumor cells is often strongly influenced by the signals that these cells receive from the tumor-associated stroma. Most prominent among the phenotypic changes experienced by carcinoma cells is the activation of a usually latent cell-biological program termed the epithelial–mesenchymal transition (EMT), which is normally operative during early embryogenesis, where it programs the interconversions of cell xiii

xiv

Biomarkers of the Tumor Microenvironment—A Prologue

types that are destined to form distinct tissues and organs; in addition, the EMT program operates transiently during wound healing. In the case of carcinomas, the EMT-inducing signals received by carcinoma cells from their microenvironment drive the acquisition of a complex spectrum of cell-biological changes involving the shedding of preexisting epithelial traits (inherited from normal cells-of-origin) and the acquisition of mesenchymal traits, such as motility, invasiveness, an elevated resistance to various types of therapeutic intervention, and an ability to disseminate to anatomical sites distant from the primary tumor. The resulting secondary tumor colonies—metastases—are responsible for 90% of cancer-associated mortality. These heterotypic signaling interactions between neoplastic cells and their stromal microenvironment are extremely complex. Each of the participating cell types releases a complex mixture of heterotypic signals that impinge upon and influence multiple other cell types. This multi-body problem dwarfs in its complexity the three-body problem that has thwarted the attempts by physicists to describe mathematically. As a consequence, we come to realize that the study of tumor microenvironments, which is already a highly active area of research, is still in its infancy, given the complexity of the cell–cell signaling networks that operate within the tumor-associated stroma and between this stroma and nearby cancer cells. The present volume lays out some of the more salient of these interactions. As complex as they are, the signaling networks described here still represent only a beginning. Thus, at present, the complexity of the signaling networks vastly outstrips our ability to understand them in their entirety, i.e., to understand how multiple heterotypic interactions conspire to create the complex biology of high-grade malignancies. Still, what is presented in this volume represents an interesting and exciting beginning! Enjoy what you read! Robert A. Weinberg Whitehead Institute for Biomedical Research Ludwig/MIT Center for Integrative Cancer Research MIT Department of Biology, Cambridge, MA, USA

Contents

Part I Basic Studies: Tumor Mechanisms and Tissue Biomarkers 1 The Role of the Tumor Microenvironment in Regulating Angiogenesis ���������������   3 Randolph S. Watnick 2 Tissue-Based Biomarkers of Tumor-­Vascular Interactions �����������������������������������  17 Lars A. Akslen 3 Molecular Phenotypes of Endothelial Cells in Malignant Tumors�������������������������  31 Vladan Milosevic, Reidunn J. Edelmann, Johanna Hol Fosse, Arne Östman, and Lars A. Akslen 4 Lymphatics in Malignant Tumors�����������������������������������������������������������������������������  53 Marek Wagner and Helge Wiig 5 Role of the Extracellular Matrix in Tumor Stroma: Barrier or Support? �����������  63 Cédric Zeltz, Roya Navab, Ning Lu, Marion Kusche-­Gullberg, Ming-Sound Tsao, and Donald Gullberg 6 Tissue Architecture in Cancer Initiation and Progression �������������������������������������  91 Susan E. Leggett and Celeste M. Nelson 7 Tumor-Fibroblast Interactions in Carcinomas ������������������������������������������������������� 109 Harsh Dongre and Daniela Elena Costea 8 Stromal PDGF Receptors; Impact on Prognosis and Response to Treatment����������������������������������������������������������������������������������������������� 125 Carina Strell and Arne Östman 9 Inflammation and Cancer: Lipid Autacoid and Cytokine Biomarkers of the Tumor Microenvironment ��������������������������������������������������������� 139 Molly M. Gilligan, Bruce R. Zetter, and Dipak Panigrahy 10 Role of Lymphocytes in Cancer Immunity and Immune Evasion Mechanisms��������������������������������������������������������������������������������������������������� 159 Kushi Kushekhar, Stalin Chellappa, Einar M. Aandahl, and Kjetil Taskén 11 Drivers of EMT and Immune Evasion��������������������������������������������������������������������� 183 Rolf A. Brekken and Katarzyna Wnuk-Lipinska 12 Inflammatory Biomarkers for Cancer��������������������������������������������������������������������� 195 Alexandre Corthay and Guttorm Haraldsen 13 Tumor Infiltrating Lymphocytes in Breast Cancer: Implementation of a New Histopathological Biomarker����������������������������������������� 207 Giuseppe Floris, Glenn Broeckx, Asier Antoranz, Maxim De Schepper, Roberto Salgado, Christine Desmedt, Dieter J. E. Peeters, and Gert G. G. M. Van den Eynden xv

xvi

14 Regulation of Tumor Progression and Metastasis by Bone Marrow-­Derived Microenvironments������������������������������������������������������������� 245 Divya Ramchandani, Tyler P. El Rayes, Dingcheng Gao, Nasser K. Altorki, Thomas R. Cox, Janine T. Erler, and Vivek Mittal 15 The Role of Platelets in the Tumor Microenvironment������������������������������������������� 267 Qiuchen Guo, Harvey G. Roweth, Kelly E. Johnson, Sandra S. McAllister, Joseph E. Italiano Jr., and Elisabeth M. Battinelli 16 Neurogenesis in the Tumor Microenvironment������������������������������������������������������� 283 Heidrun Vethe, Ole Vidhammer Bjørnstad, Manuel Carrasco, and Lars A. Akslen 17 Neuropilins as Cancer Biomarkers: A Focus on Neuronal Origin and Specific Cell Functions��������������������������������������������������������������������������� 295 Dakshnapriya Balasubbramanian, Yao Gao, and Diane R. Bielenberg 18 The Role of AXL Receptor Tyrosine Kinase in Cancer Cell Plasticity and Therapy Resistance����������������������������������������������������������������������������� 307 Maria L. Lotsberg, Kjersti T. Davidsen, Stacey D’Mello Peters, Gry S. Haaland, Austin Rayford, James B. Lorens, and Agnete S. T. Engelsen 19 Modeling the Tumor Microenvironment in Patient-Derived Xenografts: Challenges and Opportunities ������������������������������������������������������������� 329 Katrin Kleinmanns, Christiane Helgestad Gjerde, Anika Langer, Vibeke Fosse, Elvira García de Jalón, Calum Leitch, Mihaela Popa, Pascal Gelebart, and Emmet McCormack 20 Use of Imaging Mass Cytometry in Studies of the Tissue Microenvironment ����� 345 Ida Herdlevær, Lucia Lisa Petrilli, Fatime Qosaj, Maria Vinci, Dario Bressan, and Sonia Gavasso 21 Artificial Intelligence in Studies of Malignant Tumours����������������������������������������� 365 André Pedersen, Ingerid Reinertsen, Emiel A. M. Janssen, and Marit Valla Part II Clinical Applications: Organ Related Studies of Biomarkers and Therapy 22 The Tumor and Its Microenvironment as Complementary Sources of Cancer Biomarkers ����������������������������������������������������������������������������������������������������������������� 379 Roopali Roy, Emily Man, Rama Aldakhlallah, Emma Rashes, and Marsha A. Moses 23 Gene Expression Signatures of the Tumor Microenvironment: Relation to Tumor Phenotypes and Progress in Breast Cancer����������������������������� 401 Elisabeth Wik, Lise M. Ingebriktsen, and Lars A. Akslen 24 MR-Derived Biomarkers for Cancer Characterization ����������������������������������������� 425 Eugene Kim, Morteza Esmaeili, Siver A. Moestue, and Tone F. Bathen 25 The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial Microenvironment Platforms������������������������������������������������������������������������������������� 441 Tiina A. Jokela, Eric G. Carlson, and Mark A. LaBarge 26 Models of Tumor Progression in Prostate Cancer��������������������������������������������������� 453 Waqas Azeem, Yaping Hua, Karl-Henning Kalland, Xisong Ke, Jan Roger Olsen, Anne Margrete Oyan, and Yi Qu

Contents

Contents

xvii

27 Prostate Cancer Biomarkers: The Old and the New����������������������������������������������� 467 Anette L. Magnussen and Ian G. Mills 28 The Role of the Microenvironment in Endometriosis: Parallels and Distinctions to Cancer������������������������������������������������������������������������� 483 Michael S. Rogers 29 Tumor-Vascular Interactions in Non-­Small Cell Lung Cancer ����������������������������� 497 Maria Ramnefjell and Lars A. Akslen 30 Tumor–Host Interactions in Malignant Gliomas����������������������������������������������������� 509 Lina Leiss, Ercan Mutlu, Mohummad Aminur Rahman, Mette Hartmark Nilsen, and Per Øyvind Enger 31 The Role of the Microenvironment in Tumor Promoting Stress Responses��������������������������������������������������������������������������������������� 519 Hanna Dillekås, Cornelia Schuster, Kjersti T. Davidsen, and Oddbjørn Straume 32 Vascular Co-option in the Brain Tumor Microenvironment����������������������������������� 537 Sarah Wang and Andrew C. Dudley 33 Biomarker Panels and Contemporary Practice in Clinical Trials of Personalized Medicine��������������������������������������������������������������������������������� 549 Nina Louise Jebsen, Irini Ktoridou-Valen, and Bjørn Tore Gjertsen 34 Cancer Biomarkers: A Long and Tortuous Journey����������������������������������������������� 563 Wen Jing Sim, Kian Chung Lee, and Jean Paul Thiery Index�����������������������������������������������������������������������������������������������������������������������������������  581

Part I Basic Studies: Tumor Mechanisms and Tissue Biomarkers

1

The Role of the Tumor Microenvironment in Regulating Angiogenesis Randolph S. Watnick

Abstract

The tumor microenvironment plays a crucial role in cancer development and progression. Paracrine signaling between tumor cells and the normal cells that make up the microenvironment is a critical component influencing the progression of tumors from the in situ stage to metastatic disease. Despite the importance of these paracrine signaling mechanisms and factors, the vast majority of academic research and development in the pharmaceutical industry is still targeted toward mutations and aberrant signaling pathways within tumor cells. As a result, the intercellular signaling that between tumor cells and the microenvironment has not been as extensively studied with regard to the regulation of angiogenesis. In this chapter, we define the key players in the regulation of angiogenesis and examine how their expression is regulated in the microenvironment. The resulting analysis presents observations that at first glance may seem paradoxical. However, these nuances serve to underscore the complexity of these interactions and the need to better delineate and define the environmental context underlying these mechanisms.

Take-Home Lessons

• The cells in the tumor microenvironment (TME) play a major role in regulating angiogenesis. • Non-tumor cells, such as fibroblasts, inflammatory cells, and immune cells, normally repress angiogenesis.

R. S. Watnick (*) Department of Surgery, Harvard Medical School, Vascular Biology Program, Boston Children’s Hospital, Boston, USA e-mail: [email protected]

• Tumors secrete growth factors and cytokines that reprogram non-tumor cells in the TME to repress anti-angiogenic proteins and express pro-­angiogenic proteins. • Tumor-secreted modulators of the TME can act both locally via paracrine interactions and systemically to reprogram cells in both the primary tumor and potential sites of metastasis.

Introduction In the earliest stages of cancer, carcinomas (tumors derived from epithelial cells), are physically confined within the epithelial compartment of the tissue from whence they arise. These early lesions (carcinomas in situ) are separated from the tissue parenchyma by the basement membrane [1]. Opposite the basement membrane are a myriad of cells consisting of fibroblasts, myofibroblasts, immune/inflammatory cells, and endothelial cells [2]. In addition to these cell types are extracellular matrix proteins, secreted largely by fibroblasts, which serve as a substrate for tumor cells to attach [2]. In order for tumors to progress to a clinically relevant and potentially lethal disease, they must, in a sequential manner, acquire the capacity to escape the epithelial compartment, invade the local parenchyma, and disseminate systemically. To enable this process, tumor cells must degrade, or induce the degradation of the basement membrane that separates the epithelial compartment from the parenchyma. Invasion of the tissue parenchyma by the tumor, or, conversely invasion of stromal cells into the tumor, initiates a phase of tumor progression in which tumor growth becomes dependent on non-cell autonomous processes regulated by paracrine and juxtacrine signaling interactions between the tumor and its microenvironment [3–5]. In other words, in order to expand in size beyond the diffusion limit of oxygen in tissue, a new vasculature must form

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. A. Akslen, R. S. Watnick (eds.), Biomarkers of the Tumor Microenvironment, https://doi.org/10.1007/978-3-030-98950-7_1

3

4

R. S. Watnick

in and around the tumor. The ingrowth of this vasculature is driven by both soluble and matrix-bound growth factors and enzymes secreted by tumor cells and the stromal cells that comprise the microenvironment [6]. Strong evidence exists indicating that stromal cells play as central a role in matrix remodeling, invasion, and metastasis as the tumor cells themselves [7–9]. Critical to this process is the observation that carcinoma cells are able to reprogram and coopt the surrounding stromal cells to enhance tumor growth [10]. Specifically, tumor-stromal paracrine signaling pathways have been demonstrated to play a major role in the tumorigenesis and subsequent outgrowth of tumors in multiple sites [11–13]. For example, stromal fibroblasts from prostate tumors are able to stimulate tumor formation of immortal but non-transformed prostate epithelial cells when the mixture is injected orthotopically into nude mice [10]. As stated above, tumor angiogenesis is intricately linked to signaling between the tumor and microenvironment. In normal tissue architecture, the epithelial compartment is not vascularized as it is generally only 1–2 cell layers in thickness. The minimal thickness allows oxygen to diffuse readily across the basement membrane and nourish both epithelial cell layers. However, when tumors form and the lumen of epithelial ducts fills with tumor cells, the cells in the center become hypoxic due to their increased distance from existing blood vessels. Thus, in order for tumors to gain access to the vasculature, the basement membrane must be degraded allowing the blood vessels to grow into the epithelial compartment (Fig. 1.1). Degradation of the basement membrane Fig. 1.1  Schematic diagram of tissue architecture with regard to the special distribution of normal epithelial and epithelial-­ derived carcinoma cells, extracellular components, as well as resident and infiltrating stromal cells

and ingrowth of blood vessels requires tumor cells to secrete pro-angiogenic growth factors, turn off production of anti-­ angiogenic factors. Further dissemination and metastasis require a recapitulation of that process in the distant metastatic microenvironment. Though a myriad of pro- and anti-­ angiogenic factors have been discovered and studied, the initial attempts to delineate their regulation were examined in a cell autonomous fashion, with most of the attention paid to vascular endothelial growth factor (VEGF) [14–18] and Thrombospondin-1 (Tsp-1) [19, 20], two of the major positive and negative regulators of angiogenesis. However, as the importance of the tumor microenvironment became more apparent the study of the regulation of angiogenic factors in stromal cells also increased. Not only is the regulation of angiogenesis in the tumor microenvironment critical to primary tumor growth, but also for metastatic dissemination and growth in distant organs. It is well established that tumors arising in different sites preferentially metastasize to specific organs [21]. For example, prostate cancer metastasizes preferentially to bone and liver, while breast cancer metastasizes to brain, bone, and lung [22]. The ability of a tumor cell to survive and proliferate in a metastatic environment ultimately relies on its ability to augment the angiogenic output of its microenvironment. The tumor microenvironment can grossly be categorized into two types of cells: (1). Resident cells that are present in the tissue prior to tumor development and (2) Infiltrating cells that are recruited to the tumor from the circulation or bone marrow. The first group is mainly comprised of fibroblasts and endo-

Carcinoma cells

Epithelial cells

Basement membrane ECM

Stroma

Macrophage

MDSC Monocypte

Immune cell

Fibroblast Endothelium

1  The Role of the Tumor Microenvironment in Regulating Angiogenesis

thelial cells. While the second is comprised of immune/ inflammatory cells, which include B and T cells, neutrophils, mast cells, dendritic cells, and macrophages. In this chapter, we will explore the roles of these different cell types, as well as the growth factors and extracellular matrix proteins that contribute to tumor progression.

 ell Signaling Mechanisms and Factors C Influencing Stromal Angiogenesis To understand how resident and infiltrating cells contribute to tumor angiogenesis it is necessary to delineate and describe the major angiogenic factors that stimulate and inhibit vessel ingrowth. Distilled to the most basic principles, tumor cells secrete growth factors, cytokines, and enzymes that act on non-cancerous cells in the TME to induce the production of factors that stimulate blood vessel growth, i.e., angiogenesis (Fig. 1.2).

Basic Fibroblast Growth Factor (bFGF) Basic fibroblast growth factor (bFGF, FGF2) was the first tumor-secreted pro-angiogenic factor to be isolated and purified [23]. It is one of the most potent pro-angiogenic growth

5

factors [23–25]. One interesting oddity about bFGF is despite the presence of high affinity cell surface receptors [26] and the myriad of observations that bFGF stimulates endothelial cell proliferation and angiogenesis in vivo and in vitro, the protein lacks a signal sequence to direct its secretion [27, 28]. The paracrine regulation of bFGF in stromal cells and subsequent effect on tumor angiogenesis has been confounded by its ability to potently stimulate tumor cell proliferation through FGFR signaling via both autocrine and paracrine signaling [29–31]. Nevertheless, bFGF expression in the stroma of lung adenocarcinoma patients inversely correlates with disease progression and overall survival [32]. Additionally, bFGF production is stimulated by stem cell factor (SCF) and TGF-β in inflammatory cells, including macrophages, mast cells, and neutrophils [33]. The role of these cells in tumor angiogenesis will be detailed later in this chapter.

VEGF Vascular endothelial growth factor (VEGF) is another tumor-­ secreted pro-angiogenic factor that was initially identified by its ability to induce vessel permeability, as such, it was first called vascular permeability factor or VPF [34, 35]. The regulation of VEGF in tumor cells has been exhaustively

Blood vessels

Tumor

Tumor Microenvironment

Fig. 1.2  Schematic diagram of the interaction between tumor cells and their microenvironment as mediated by tumor-secreted factors paracrine acting factors that modulate angiogenesis

6

studied. Signal transduction pathways leading from receptor tyrosine kinases or oncogenic Ras and PI3 kinase via the MAPK or Akt pathway lead to increased transcription of VEGF and its subsequent secretion into the extracellular matrix [16]. However, carcinoma cells also secrete proteins into the extracellular space, which do not act directly on endothelial cells but rather modulate VEGF production and secretion by stromal cells in the microenvironment, such as TGF-β, PDGF, and bFGF [36, 37]. Stromal VEGF expression was first demonstrated to be regulated by carcinoma cells in a transgenic mouse model in which GFP, driven by the VEGF promoter, was inserted into the mouse genome [38]. In this model, activation of the VEGF promoter results in the expression of GFP.  Examination of tumor xenografts in these VEGF-GFP mice revealed, via fluorescence microscopy, that the stromal fibroblasts that had infiltrated the tumor fluoresced green, indicating that the VEGF promoter had been activated. Strikingly, in normal tissues there were no fluorescent cells, indicating the VEGF expression is not required for normal tissue homeostasis. These results indicated that tumors secrete factors that act on cells in the microenvironment to stimulate VEGF expression. However, while it was clear from these experiments that VEGF expression was being stimulated, it was not evident whether this stimulation was required for tumor growth, supportive of tumor growth or merely a physiological reaction to local tumor growth. The evidence that stromally produced VEGF was critical for tumor growth was obtained from studies designed to test the efficacy of a human-specific anti-­ VEGF antibody, bevacizumab (Avastin). In these experiments, human tumor cells were injected into immunocompromised mice, which were subsequently treated with the human-specific VEGF antibody [39]. While the antibody was able to dramatically inhibit tumor growth the tumors still grew. The authors of the study hypothesized that the continued growth of the treated tumors was due to a residual angiogenic stimulus driven by VEGF produced and secreted from the murine tumor microenvironment, which could not be inhibited by the human-specific antibody. To test their hypothesis, human tumor xenografts were treated with human-specific VEGF antibodies as well as a soluble version of murine VEGFR1 (mFlt) fused to IgG, which acts as a decoy receptor for VEGF [40]. The combination of these two treatment modalities resulted in the complete blockage of tumor growth, demonstrating the importance of the contribution of stromal-produced VEGF.

R. S. Watnick

bFGF in lung fibroblasts [41]. Additionally, in response to the results achieved with the human-specific VEGF antibody described above, it was demonstrated that stromal VEGF expression was stimulated by tumor-derived PDGF [42]. Within that context, inhibition of PDGF activity via a soluble version of PDGFR was able to block the stimulation of VEGF in the microenvironment and inhibit angiogenesis. Moreover, another member of the PDGF family, PDGF B is also able to upregulate VEGF expression in vascular smooth muscle cells [36]. These data indicated that tumor-derived PDGF is a potent inducer of VEGF expression in the microenvironment. The most logical conclusion to be drawn from the above study is that PDGF promotes angiogenesis via the induction of stromal VEGF. However, somewhat analogous to TGF-β, the activities of PDGF is not as straightforward as these results would indicate. In addition to stimulating VEGF and bFGF, PDGF also stimulates Tsp-1 expression [43]. PDGF stimulation of Tsp-1  in fibroblasts is mediated by the Raf-­ MAPK pathway in a manner analogous to the stimulation of Tsp-1 by serum [44]. Intriguingly, the PDGF-mediated stimulation of VEGF is also mediated by the Raf-MAPK pathway [45]. Thus, whether PDGF acts as an anti-angiogenic factor or a pro-angiogenic factor is most likely dependent on complimentary or inhibitory orthogonal signals that act to inhibit or stimulate VEGF or Tsp-1.

TGF-β

One growth factor with perhaps the most paradoxical role in tumor growth and angiogenesis is TGF-β. TGF-β has been documented to have potent pro-angiogenic activity in  vivo [46]. However, in vitro the effects of TGF-β on endothelial cells are in diametric opposition as in this context it is actually growth-inhibitory [47, 48]. These seemingly incongruous activities were resolved by the discovery that TGF-β stimulates the expression of VEGF in stromal fibroblasts, indicating that the pro-angiogenic effects of TGF-β were mediated by the induction of VEGF in the tumor microenvironment [36, 49]. Moreover, TFG-β potently stimulates the expression of bFGF in fibroblasts [41]. These results suggest that low levels of tumor-secreted TGF-β induce tumor-­ associated fibroblasts to express VEGF and bFGF thereby stimulating angiogenesis. Conversely, higher levels of TGF-β may act directly on endothelial cells inhibiting their proliferation thus having an anti-angiogenic effect. Adding to the paradox of TGF-β’s role in tumor angioPDGF genesis is its ability to stimulate the expression of the anti-­ angiogenic protein Tsp-1 (which will be discussed in detail Another growth factor that possesses both pro- and anti-­ later in this chapter). In a classic example of a feed-forward angiogenic characteristics is PDGF.  In 1991, Goldsmith loop, Tsp-1 then activates TGF-β from its latent form [50– et al. demonstrated that PDGF was able to potently stimulate 54]. TGF-β is activated by two discrete processes: via prote-

1  The Role of the Tumor Microenvironment in Regulating Angiogenesis

ases that cleave the latent associate peptide; and via undergoing a conformational change, that exposes the receptor-­binding region. Tsp-1 activates TGF-β via the latter mechanism. Moreover, TGF-β expression in fibroblasts is induced by hypoxia, which is most often a result of a lack of tumor vascularization [55].

Matrix Metalloproteases The ability of tumors to invade locally, across the basement membrane, is critical for tumor growth and ultimately metastasis. One of the most crucial steps in tumor invasion and migration is the remodeling of the extracellular matrix (ECM) by proteases. Some of the major players in this field are the matrix metalloproteases, or MMPs. For example, an experiment in which MCF7 breast cancer cells and fibroblasts were co-injected into mice resulted in the significant acceleration of tumor growth [56]. Moreover, in a parallel experiment in which fibroblasts were engineered to ectopically express an inhibitor of MMP activity, TIMP-2 (tissue inhibitor of metalloprotease 2), the tumor stimulating activity of the co-injected fibroblasts was abrogated [57]. Analogously, administration of a broad spectrum MMP inhibitor, batimastat, also abrogated the ability of fibroblasts to stimulate tumor formation by MCF7 cells [57]. The matrix remodeling mediated by MMPs not only facilitates tumor cell migration into the surrounding microenvironment but also stimulates the migration of endothelial cells into the tumor by facilitating the formation of the leading edge of new blood vessels. MMPs also liberate growth factors such as VEGF and bFGF that would otherwise be sequestered in the ECM. The ability of MMPs to stimulate angiogenesis was established in an elegant genetic experiment in which tumor prone RIP-TAG2 mice were crossed with various matrix protease knockout mice [58]. By crossing the RIP-TAG mice with MMP2 knockout mice the authors demonstrated that tumor growth was impaired but not due to any defect in angiogenesis [58]. Conversely, MMP9-/- RIP-TAG mice displayed delayed tumor growth kinetics and defective angiogenesis [58]. The conclusion drawn from these observations was that in addition to cleaving matrix proteins, MMP9 also cleaves the latent associated peptide from TGF-β, converting it to the active form, and thereby stimulating tumor growth in a mammary tumor model [59].

Hormones and Nuclear Receptors The studies described above indicate that two of the most potent inducers of stromal VEGF and, consequently, angiogenesis also possess the seemingly counterproductive ability

7

to stimulate Tsp-1. These divergent events downstream from TGF-β and PDGF ligation to their cognate receptors suggest that tumor-derived TGF-β and PDGF expression should have no net effect on angiogenesis. That being said, it has also been demonstrated that inhibition of PDGF activity inhibits tumor angiogenesis [42]. Also, as described above, despite its ability to stimulate Tsp-1, TGF-β is a potent stimulator of angiogenesis. One potential explanation for the observed pro-angiogenic activities of these two proteins is that the expression of Tsp-1 in the microenvironment is suppressed by an independent signaling mechanism. This suppression of Tsp-1 would result in the stimulation of only the pro-­ angiogenic factors VEGF and bFGF by these two growth factors and thus resolve the seemingly paradoxical observations. Two candidates for such a Tsp-1 repressing factor are the hormones estrogen and androgen, which have both been demonstrated to repress Tsp-1 expression [60, 61]. While these hormones both repress Tsp-1 expression the mechanisms utilized are different between them. Estrogenmediated inhibition of Tsp-1 is achieved via activation of ERK1/2 and JNK [60]. Additionally, Tsp-1 repression by estrogen is mediated via inhibition of both transcription and protein secretion. Conversely, androgen-mediated repression of Tsp-1 is solely mediated by inhibition of transcription, via an androgen responsive element (ARE) in the Tsp-1 promoter [61]. While hormone-mediated effects on tumor growth have been largely studied through their actions on hormone-­ responsive tumor cells, it has also been demonstrated that estrogen can regulate angiogenesis on a systemic level [62]. This elegant study revealed that estrogen receptor (ER) positive stromal cells stimulate angiogenesis and promote tumor growth in response to estrogen even when the tumor cells were ER-negative. It has also been demonstrated that another nuclear receptor family, the peroxisome proliferator-activator receptors (PPAR), can regulate both VEGF and Tsp-1 expression. Specifically, it has been demonstrated that when normal tumor are cells injected into PPARα-/- mice they remain dormant for a prolonged period of time [63]. Moreover, the dormancy of these tumors was found to be the result of increased Tsp-1 expression in the host stroma. Surprisingly, it was later determined that fenofibrate and WY14643, two agonists of PPARα, also stimulated the expression of Tsp-1 [64]. These seemingly discordant results suggest that in the absence of PPARα, another member or the PPAR family, perhaps, may compensate and stimulate the expression of Tsp-1. Of note, PPARγ also stimulates the expression of CD36 [65], a receptor for Tsp-1. In keeping with these observations, it was demonstrated that the PPARγ agonists rosiglitazone and pioglitazone inhibit bFGF- and VEGFmediated angiogenesis [66].

8

R. S. Watnick

Thrombospondin-1

Nonprotein Mediators of Angiogenesis

While much of the attention in the field of angiogenesis has been given to the identification and characterization of pro-­ angiogenic factors, the studies detailing the role of one of the most potent anti-angiogenic proteins, Tsp-1, should not be overlooked. Thrombospondin-1 (Tsp-1) is an endogenous anti-angiogenic protein that functions via a multimodal approach. Tsp-1 binds to cell surface receptors CD36 and CD47 on the endothelial cell surface and renders the cells insensitive to both VEGF and bFGF.  Tsp-1 also induces caspase-­ dependent apoptosis mediated by Srcfamily kinase Fyn signaling downstream from CD36 [67– 70]. Tsp-1 also binds to MMP-9 and functionally inactivates it [71, 72]. In tumor cells, Tsp-1 expression is repressed via a signal transduction cascade emanating from PI3-kinase via Rho GTPase to ROCK to Myc, which represses Tsp-1 in a phosphorylation-­dependent manner [73]. The Rho-Rock pathway has been shown to be active in several human breast cancer cell lines in which Tsp-1 expression was virtually silenced [73]. Accordingly, this pathway represents the first biochemical elucidation of a cell autonomous “angiogenic switch.” While the expression of VEGF in the tumor-associated stroma is widely accepted to have a positive correlation with tumor progression [38, 74, 75], the role of Thrombospondin-1 (Tsp-1) expression in the tumor-associated stroma is unclear. Tsp-1 expression by epithelial tumor cells is observed infrequently and ectopic expression of Tsp-1 is inhibitory to tumor growth [19, 73, 76]. Stromal Tsp-1, meanwhile, has been correlated with a desmoplastic response and increased invasiveness in a subset of breast cancers [74, 77, 78], while it has been demonstrated to be inhibitory to early stage breast cancers [79]. Expression of Tsp-1 by stromal fibroblasts has been shown to be inhibitory to tumor formation and growth [80]. Intriguingly, the same report demonstrated that tumors that arose in an environment high in Tsp-1 eventually overcame the inhibitory effects of this protein by increasing their production of VEGF.  Thus, the complex interrelationship between these two proteins and their relative expression ­levels in the tumor-associated stroma plays a critical role in the induction and maintenance of angiogenesis in human tumors. The work described above demonstrated that VEGF expression in the stroma is a critical component in tumor-­ mediated angiogenesis. Conversely, Tsp-1 expression in the tumor-associated stroma can be a potent inhibitor of tumor angiogenesis and growth. The question that arises then is how do tumors stimulate the expression of VEGF in the stroma while concomitantly repressing the expression of Tsp-1?

One largely understudied signaling mechanism in the regulation of tumor angiogenesis is lipid and phospholipid signaling. Significantly, two landmark studies demonstrated that these molecules could regulate the expression of Tsp-1. These independent studies showed that the generation of phospholipids and the resultant signaling pathways potently repress Tsp-1 expression in stromal fibroblasts. The first study demonstrated that platelet-mediated generation of phospholipids, specifically sphingosine 1 phosphate (S1P), downregulated Tsp-1 expression in dermal fibroblasts by activating the Gi-protein coupled S1P receptors [81]. The second study also implicated S1P as a repressor of Tsp-1 by demonstrating that secretion of a low molecular weight molecule ( 10,000 times in liquid-state NMR.  Proc Natl Acad Sci U S A. 2003;100(18):10158–63. 103. Day SE, Kettunen MI, Gallagher FA, Hu DE, Lerche M, Wolber J, et al. Detecting tumor response to treatment using hyperpolarized 13C magnetic resonance imaging and spectroscopy. Nat Med. 2007;13(11):1382–7. 104. Ward CS, Venkatesh HS, Chaumeil MM, Brandes AH, Vancriekinge M, Dafni H, et  al. Noninvasive detection of target modulation following phosphatidylinositol 3-kinase inhibition using hyperpolarized 13C magnetic resonance spectroscopy. Cancer Res. 2010;70(4):1296–305. 105. Gallagher FA, Woitek R, McLean MA, Gill AB, Manzano Garcia R, Provenzano E, et al. Imaging breast cancer using hyperpolarized carbon-13 MRI. Proc Natl Acad Sci U S A. 2020;117(4):2092–8. 106. Gallagher FA, Kettunen MI, Hu DE, Jensen PR, Zandt RI, Karlsson M, et  al. Production of hyperpolarized [1,4-13C2] malate from [1,4-13C2]fumarate is a marker of cell necrosis and treatment response in tumors. Proc Natl Acad Sci U S A. 2009;106(47):19801–6. 107. Bohndiek SE, Kettunen MI, Hu DE, Witney TH, Kennedy BW, Gallagher FA, et  al. Detection of tumor response to a vascular disrupting agent by hyperpolarized 13C magnetic resonance spectroscopy. Mol Cancer Ther. 2010;9(12):3278–88. 108. Plathow C, Weber WA. Tumor cell metabolism imaging. J Nucl Med. 2008;49(Suppl 2):43S–63S. 109. Testa C, Schiavina R, Lodi R, Salizzoni E, Corti B, Farsad M, et  al. Prostate cancer: sextant localization with MR imaging, MR spectroscopy, and 11C-choline PET/CT.  Radiology. 2007;244(3):797–806. 110. Tozaki M, Hoshi K. 1H MR spectroscopy of invasive ductal carcinoma: correlations with FDG PET and histologic prognostic factors. AJR Am J Roentgenol. 2010;194(5):1384–90. 111. Gutte H, Hansen AE, Henriksen ST, Johannesen HH, Ardenkjaer-­ Larsen J, Vignaud A, et  al. Simultaneous hyperpolarized (13) C-pyruvate MRI and (18)F-FDG-PET in cancer (hyperPET): feasibility of a new imaging concept using a clinical PET/MRI scanner. Am J Nucl Med Mol Imaging. 2015;5(1):38–45. 112. Esmaeili M, Tayari N, Scheenen T, Elschot M, Sandsmark E, Bertilsson H, et  al. Simultaneous (18)F-fluciclovine positron emission tomography and magnetic resonance spectroscopic imaging of prostate cancer. Front Oncol. 2018;8:516. 113. Gutte H, Hansen AE, Larsen M, Rahbek S, Henriksen S, Johannesen H, et al. Simultaneous hyperpolarized 13C-pyruvate MRI and 18F-FDG-PET (hyperPET) in 10 canine cancer patients. J Nucl Med 2015;56(11):1786–92 114. Semenza GL. Defining the role of hypoxia-inducible factor 1 in cancer biology and therapeutics. Oncogene. 2010;29(5):625–34. 115. Thomlinson RH, Gray LH.  The histological structure of some human lung cancers and the possible implications for radiotherapy. Br J Cancer. 1955;9(4):539–49. 116. Ragnum HB, Vlatkovic L, Lie AK, Axcrona K, Julin CH, Frikstad KM, et  al. The tumour hypoxia marker pimonidazole reflects a transcriptional programme associated with aggressive prostate cancer. Br J Cancer. 2015;112(2):382–90. 117. Rofstad EK, Galappathi K, Mathiesen B, Ruud EB.  Fluctuating and diffusion-limited hypoxia in hypoxia-induced metastasis. Clin Cancer Res. 2007;13(7):1971–8.

439 118. Semenza GL.  The hypoxic tumor microenvironment: a driving force for breast cancer progression. Biochim Biophys Acta 2015. 119. Ellingsen C, Hompland T, Galappathi K, Mathiesen B, Rofstad EK. DCE-MRI of the hypoxic fraction, radioresponsiveness, and metastatic propensity of cervical carcinoma xenografts. Radiother Oncol. 2014;110(2):335–41. 120. Huuse EM, Moestue SA, Lindholm EM, Bathen TF, Nalwoga H, Kruger K, et  al. In vivo MRI and histopathological assessment of tumor microenvironment in luminal-like and basal-like breast cancer xenografts. J Magn Reson Imaging. 2012;35(5):1098–107. 121. Ovrebo KM, Hompland T, Mathiesen B, Rofstad EK. Assessment of hypoxia and radiation response in intramuscular experimental tumors by dynamic contrast-enhanced magnetic resonance imaging. Radiother Oncol. 2012;102(3):429–35. 122. Stubbs M, Robinson SP, Rodrigues LM, Parkins CS, Collingridge DR, Griffiths JR. The effects of host carbogen (95% oxygen/5% carbon dioxide) breathing on metabolic characteristics of Morris hepatoma 9618a. Br J Cancer. 1998;78(11):1449–56. 123. Baudelet C, Gallez B.  How does blood oxygen level-dependent (BOLD) contrast correlate with oxygen partial pressure (pO2) inside tumors? Magn Reson Med. 2002;48(6):980–6. 124. McPhail LD, Robinson SP.  Intrinsic susceptibility MR imaging of chemically induced rat mammary tumors: relationship to histologic assessment of hypoxia and fibrosis. Radiology. 2010;254(1):110–8. 125. Burrell JS, Walker-Samuel S, Baker LC, Boult JK, Jamin Y, Halliday J, et al. Exploring DeltaR(2) * and DeltaR(1) as imaging biomarkers of tumor oxygenation. J Magn Reson Imaging. 2013;38(2):429–34. 126. O'Connor JPB, Robinson SP, Waterton JC.  Imaging tumour hypoxia with oxygen-enhanced MRI and BOLD MRI.  Br J Radiol. 2019;92(1095):20180642. 127. Gross MW, Karbach U, Groebe K, Franko AJ, Mueller-Klieser W.  Calibration of misonidazole labeling by simultaneous measurement of oxygen tension and labeling density in multicellular spheroids. Int J Cancer. 1995;61(4):567–73. 128. Raleigh JA, Chou SC, Bono EL, Thrall DE, Varia MA. Semiquantitative immunohistochemical analysis for hypoxia in human tumors. Int J Radiat Oncol Biol Phys. 2001;49(2):569–74. 129. Lee CP, Payne GS, Oregioni A, Ruddle R, Tan S, Raynaud FI, et  al. A phase I study of the nitroimidazole hypoxia marker SR4554 using 19F magnetic resonance spectroscopy. Br J Cancer. 2009;101(11):1860–8. 130. Gulaka PK, Rojas-Quijano F, Kovacs Z, Mason RP, Sherry AD, Kodibagkar VD. GdDO3NI, a nitroimidazole-based T1 MRI contrast agent for imaging tumor hypoxia in vivo. J Biol Inorg Chem. 2014;19(2):271–9. 131. Okamoto S, Shiga T, Yasuda K, Ito YM, Magota K, Kasai K, et  al. High reproducibility of tumor hypoxia evaluated by 18F-fluoromisonidazole PET for head and neck cancer. J Nucl Med. 2013;54(2):201–7. 132. Rasey JS, Koh WJ, Evans ML, Peterson LM, Lewellen TK, Graham MM, et  al. Quantifying regional hypoxia in human tumors with positron emission tomography of [18F]fluoromisonidazole: a pretherapy study of 37 patients. Int J Radiat Oncol Biol Phys. 1996;36(2):417–28. 133. Kikuchi M, Yamane T, Shinohara S, Fujiwara K, Hori SY, Tona Y, et  al. 18F-fluoromisonidazole positron emission tomography before treatment is a predictor of radiotherapy outcome and survival prognosis in patients with head and neck squamous cell carcinoma. Ann Nucl Med. 2011;25(9):625–33. 134. Rischin D, Hicks RJ, Fisher R, Binns D, Corry J, Porceddu S, et al. Prognostic significance of [18F]-misonidazole positron emission tomography-detected tumor hypoxia in patients with advanced head and neck cancer randomly assigned to chemoradiation with

440 or without tirapazamine: a substudy of trans-Tasman radiation oncology group study 98.02. J Clin Oncol. 2006;24(13):2098–104. 135. Souvatzoglou M, Grosu AL, Roper B, Krause BJ, Beck R, Reischl G, et al. Tumour hypoxia imaging with [18F]FAZA PET in head and neck cancer patients: a pilot study. Eur J Nucl Med Mol Imaging. 2007;34(10):1566–75. 136. Sakso M, Mortensen LS, Primdahl H, Johansen J, Kallehauge J, Hansen CR, et al. Influence of FAZA PET hypoxia and HPV-­ status for the outcome of head and neck squamous cell carcinoma (HNSCC) treated with radiotherapy: long-term results from the DAHANCA 24 trial (NCT01017224). Radiother Oncol. 2020;151:126–33. 137. Capitanio U, Pepe G, Incerti E, Larcher A, Trevisani F, Luciano R, et al. The role of 18F-FAZA PET/CT in detecting lymph node metastases in renal cell carcinoma patients: a prospective pilot trial. Eur J Nucl Med Mol Imaging 2020. 138. Estrella V, Chen T, Lloyd M, Wojtkowiak J, Cornnell HH, Ibrahim-­ Hashim A, et  al. Acidity generated by the tumor microenvironment drives local invasion. Cancer Res. 2013;73(5):1524–35. 139. Stubbs M, McSheehy PM, Griffiths JR, Bashford CL. Causes and consequences of tumour acidity and implications for treatment. Mol Med Today. 2000;6(1):15–9. 140. Fukumura D, Jain RK.  Tumor microenvironment abnormalities: causes, consequences, and strategies to normalize. J Cell Biochem. 2007;101(4):937–49. 141. Gillies RJ, Robey I, Gatenby RA.  Causes and consequences of increased glucose metabolism of cancers. J Nucl Med. 2008;49(Suppl 2):24S–42S. 142. Liu J, Huang Y, Kumar A, Tan A, Jin S, Mozhi A, et al. pH-­sensitive nano-systems for drug delivery in cancer therapy. Biotechnol Adv. 2014;32(4):693–710. 143. Wang L, Zhu X, Xie C, Ding N, Weng X, Lu W, et al. Imaging acidosis in tumors using a pH-activated near-infrared fluorescence probe. Chem Commun (Camb). 2012;48(95):11677–9. 144. Hashim AI, Zhang X, Wojtkowiak JW, Martinez GV, Gillies RJ. Imaging pH and metastasis. NMR Biomed. 2011;24(6):582–91. 145. Gillies RJ, Raghunand N, Garcia-Martin ML, Gatenby RA. pH imaging. A review of pH measurement methods and applications in cancers. IEEE Eng Med Biol Mag. 2004;23(5):57–64. 146. Soto GE, Zhu Z, Evelhoch JL, Ackerman JJ. Tumor 31P NMR pH measurements in vivo: a comparison of inorganic phosphate and intracellular 2-deoxyglucose-6-phosphate as pHnmr indicators

E. Kim et al. in murine radiation-induced fibrosarcoma-1. Magn Reson Med. 1996;36(5):698–704. 147. Garcia-Martin ML, Herigault G, Remy C, Farion R, Ballesteros P, Coles JA, et al. Mapping extracellular pH in rat brain gliomas in vivo by 1H magnetic resonance spectroscopic imaging: comparison with maps of metabolites. Cancer Res. 2001;61(17):6524–31. 148. Provent P, Benito M, Hiba B, Farion R, Lopez-Larrubia P, Ballesteros P, et al. Serial in vivo spectroscopic nuclear magnetic resonance imaging of lactate and extracellular pH in rat gliomas shows redistribution of protons away from sites of glycolysis. Cancer Res. 2007;67(16):7638–45. 149. Liu G, Li Y, Sheth VR, Pagel MD.  Imaging in  vivo extracellular pH with a single paramagnetic chemical exchange saturation transfer magnetic resonance imaging contrast agent. Mol Imaging. 2012;11(1):47–57. 150. Chen LQ, Howison CM, Jeffery JJ, Robey IF, Kuo PH, Pagel MD. Evaluations of extracellular pH within in vivo tumors using acidoCEST MRI. Magn Reson Med. 2014;72(5):1408–17. 151. Jones KM, Randtke EA, Yoshimaru ES, Howison CM, Chalasani P, Klein RR, et  al. Clinical translation of tumor acidosis measurements with AcidoCEST MRI.  Mol Imaging Biol. 2017;19(4):617–25. 152. High RA, Randtke EA, Jones KM, Lindeman LR, Ma JC, Zhang S, et al. Extracellular acidosis differentiates pancreatitis and pancreatic cancer in mouse models using acidoCEST MRI. Neoplasia. 2019;21(11):1085–90. 153. Akhenblit PJ, Hanke NT, Gill A, Persky DO, Howison CM, Pagel MD, et  al. Assessing metabolic changes in response to mTOR inhibition in a mantle cell lymphoma xenograft model using AcidoCEST MRI. Mol Imaging 2016;15. 154. Gallagher FA, Kettunen MI, Day SE, Hu DE, Ardenkjaer-­ Larsen JH, Zandt R, et  al. Magnetic resonance imaging of pH in  vivo using hyperpolarized 13C-labelled bicarbonate. Nature. 2008;453(7197):940–3. 155. Bergamaschi A, Hjortland GO, Triulzi T, Sorlie T, Johnsen H, Ree AH, et al. Molecular profiling and characterization of luminal-like and basal-like in vivo breast cancer xenograft models. Mol Oncol. 2009;3(5–6):469–82. 156. Bergamaschi A, Tagliabue E, Sorlie T, Naume B, Triulzi T, Orlandi R, et  al. Extracellular matrix signature identifies breast cancer subgroups with different clinical outcome. J Pathol. 2008;214(3):357–67.

The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial Microenvironment Platforms

25

Tiina A. Jokela, Eric G. Carlson, and Mark A. LaBarge

Abstract

Predicting how anticancer therapeutics will function in people based on preclinical studies remains a significant challenge. High rates of phase II clinical trial failures indicate that many candidate therapeutics that pass preclinical studies lack efficacy in patients. The discovery of oncogenes and tumor suppressors has led to vast investments into developing technologies that enable exploration of the total complexity of genomes and proteomes intrinsic to cells. These technologies seek to define how mutations contribute to cancer development and progression. An important and unexpected outcome of those massive investments to understand cancer as a cell-intrinsic problem is the undeniable conclusion that mutations do not explain everything. Indeed, the fact that frankly malignant cells can be phenotypically normal, when exposed to a normal tissue microenvironment (ME), suggests that

there is a dominant role of the ME. Tumor microenvironments modulate the malignant phenotypes of cells and impact drug responses. In most drug screens, conventional two-dimensional plastic dishes are the substrate of choice for cell culture and rodents are used as the primary in vivo model, but these modalities lack context in a way that is relevant to predicting drug activity. Alternatively, combinatorial microenvironment microarray platforms provide a high-throughput means of exploring cell-based functional responses in diverse microenvironmental milieus. Data from these techniques are single-cell resolution and encapsulate cell–cell heterogeneity, which provides direct linkages between cellular phenotypes, such as drug responses, and MEs. This chapter focuses on the applications and analytic approaches used for functional cell-based exploration of combinatorial MEs using microarray technology.

T. A. Jokela Department of Population Sciences, Beckman Research Institute, Duarte, CA, USA

M. A. LaBarge (*) Department of Population Sciences, Beckman Research Institute, Duarte, CA, USA

The Faculty of Sport and Health Sciences, University of Jyväskylä, Finland, Jyväskylä, Finland e-mail: [email protected]

Center for Cancer and Aging Research, City of Hope, Duarte, CA, USA

E. G. Carlson Department of Population Sciences, Beckman Research Institute, Duarte, CA, USA

Center for Cancer Biomarkers Research (CCBIO), University of Norway, Bergen, Norway e-mail: [email protected]

Irell and Manella Graduate School of Biological Sciences, Duarte, CA, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. A. Akslen, R. S. Watnick (eds.), Biomarkers of the Tumor Microenvironment, https://doi.org/10.1007/978-3-030-98950-7_25

441

442

F   unctional dissection of a tumor microenvironment. A microarray printer deposits combinations of TME-relevant components on a printing substrate that best mimics the in vivo microenvironment being studied. After relevant cancer cells are seeded on the simplified, printed

Take-Home Lessons

• Tumor microenvironments have dominant effects on cellular phenotypes and can impact anticancer drug responses but are often ignored during drug development. • MicroEnvironment MicroArrays (MEMAs) are functional, cell-based, high-throughput platforms to study cancer cell phenotypes and functions in combinatorial microenvironments. • MEMAs provide a fast and cost-effective in  vitro platform to predict microenvironment components that will modify anticancer drug performance. • MEMA experiments produce high-dimensional, single-­cell datasets that necessitate optimized automated workflows for ease of analysis. • Quality control and data analysis may benefit greatly from the application of both traditional machine learning and convolutional neural network algorithms.

T. A. Jokela et al.

TMEs, anticancer therapeutics are added to culture. Following culture, cells are fixed, stained, and imaged leading to the acquisition of highdimensional, single-cell data.

The Challenge of Predicting Efficacy In anticancer drug development, a suboptimal ability to predict how therapeutics will perform in humans based on preclinical drug screening often delays the progress of drug discovery. In the last few decades, tremendous resources have been invested into translating preclinical anticancer therapeutics to drugs used clinically. From 2000 to 2015, the probability of success for a new anticancer therapeutic to gain FDA approval was 3.4% [1], which has allowed manufacturers to justify skyrocketing drug prices [2]. Advancements in cell and molecular biology, as well as engineering, have ushered in the modern era of pharmacology. In this era, new therapeutics with potentially selective activity against tumors are identified by using cell-based high-throughput screenings, and then the efficacy of the selected therapeutics is further validated in animal model systems, with rodents being the most popular mammalian system. Increased knowledge about the molecular underpinnings of tumor biology has accelerated the development of

25  The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial…

novel therapeutics over the past two decades. For instance, the Cancer Genome Atlas program has identified a broad range of recurrent gene mutations and structural rearrangements that putatively drive tumorigenesis. A number of drugs have been selected to target protein changes resulting from those specific genetic mutations, but while a number of candidates show promising effects in small animals, they have had much less success in patients [3]. Almost 70% of new cancer drugs exhibit no efficacy in phase II despite meeting safety standards established in phase I trials [1]. Several potential explanations for why anticancer therapeutics fail at such high rates are the significant differences in expressed genomes of mice and men [4], as well as the significant differences that arise at the level of physiology and tissue architecture that can impact drug responses [5, 6]. The tumor microenvironment (TME), i.e., the sum of cell–cell, cell-extracellular matrix (ECM), cell-soluble factor interactions, and the physical properties and geometry of the tumor, has been shown to impact cancer progression, drug responses, and many other tumor properties [7, 8]. Thus, it is important to identify preclinical screening modalities that take microenvironment (ME) into account. Implementation of techniques that are more reflective of the relevant biology in human tissues may provide more predictable outcomes in cancer drug development.

 umors Are Heterogenous “Organs” T and Tumor Microenvironments Are Important Determinants in Therapeutic Responses Innate inter- and intra-tumor heterogeneity is thought to be a major contributor to acquisition of drug resistance. Tumors rarely are homogenous expansions of neoplastic cells and may be better framed as abnormal “organs” with multiple cell types and dynamic microenvironmental ecologies [9]. These organs interact with the body via unique vascular systems and changes to immune homeostasis that together contribute to hampered immunosurveillance as well as dampened efficacy of cancer treatments [10]. ECM, growth factors, cytokines, mechanical stress, and oxygen tension construct combinatorial and dynamic TMEs, which contribute to control the malignant progression, metastasis, and drug responses of cancer [8, 11–15]. The cancer stem cell (CSC) hypothesis offers attractive explanations for generation of heterogeneity within tumors, metastatic dissemination, and resistance to therapy. The underlying logic is modeled on normal developmental hierarchies that are delineated for several adult tissues. Undifferentiated stem cells give rise to less potent progenitors, which produce the most specialized cells of a given tissue. Analogously, only CSCs are thought capable of

443

self-renewal, of initiating tumors at primary and distant locations, and of giving rise to more differentiated daughters that are less capable of reestablishing the tumor. Normal stem cell activity is maintained in niches; therefore, employing the same logic used for developmental hierarchies, niches that maintain CSCs, should also exist [16–18]. Niches are specialized MEs wherein stem cells reside (reviewed in [19, 20]), which exert control over cell function. Niches are immensely instructive, as progenitors in both skin and skeletal muscle can adopt residency in vacated stem cell niches, where they reacquire stem cell traits [21–23]. Impressively, testicular and neural stem cells from male mice were shown to give rise to lactating mammary glands when transplanted into the mammary fat pads of female mice [24, 25]. And in true reductionist models that used defined ME components, embryonic and adult stem and progenitor cell fate decisions were shown to be quantifiably flexible in response to combinatorial MEs [26–29]. The ability of the niche to determine the functional spectrum of stem cell activities led us to hypothesize that stem cell niches beget stem cell functions [30, 31]. Due to their role in maintaining stem cell activity, disrupting CSC-niche interactions may be crucial for overcoming barriers to therapeutic resistance [32]. In fact, putative components of the CSC niche that increase drug tolerance in malignant breast cancer cells have been identified using combinatorial microenvironment microarrays (MEMAs, formerly known as MEarrays) [33]. Thus, understanding the interactions between TMEs and cancer cells is important for the identification of druggable mechanisms (e.g., proliferation, differentiation, and quiescence) that may improve drug efficacy in humans.

Deconstructing Tumor Microenvironments into Experimentally Tractable Combinations Tissues are collections of cells and ECM knit together into unique spatial configurations that cooperatively carry out specialized functions. Remarkably, tissue with an intact architecture can maintain many basic functions despite the presence of gene mutations that cause dysfunctions when introduced into cells cultured on tissue culture plastic (TCP) [34]. Disrupted MEs can unleash the malignant potential of transformed cells, which further demonstrates the principle that tissue architecture and composition confers normal function in the face of cell-intrinsic perturbations [35]. Organized asymmetry is therefore an important basic feature of metazoan tissues; there must be distinctive topologies on which receptors assemble to correctly integrate the signaling patterns associated with tissue-specific functions. TMEs should as well possess combinatorial signaling asymmetries, though the MEs may be less obviously organized. One

444

hypothesis is that the normal and tumor MEs integrate the signaling apparatuses differently. Accordingly, therapeutic targets could be identified to selectively harm the tumor cells, with ME composition functioning as a determinant of drug efficacy. Those potential differences in signal integration can be revealed by technologies that recapitulate aspects of in  vivo MEs, using defined physical, geometric, and molecular elements. This allows one to assess the contribution of each attribute of the ME to emergent properties of tissues. The complexity of MEs is a major impediment to understanding their impact on cells. A majority of our understanding of biological mechanisms in human cells has been built upon studies using two-dimensional (2D) plastic plates or dishes. Since the first human cell line, HeLa, was established on cell culture dishes, 2D cell culture has been a mainstay of biological research. However, as the dominant nature of the ME over physiological processes has become increasingly appreciated, engineered 2D and three-dimensional (3D) culture platforms that better recapitulate the molecular and physical nuances of tissues in vivo are being developed. It is an oversimplification to distinguish 2D and 3D culture platforms based on dimensionality. The details of the culture MEs need to be considered with care to understand how each definable property affects cell physiology. Although 2D TCP has been used extensively for biological research, it does not offer an accurate physiological representation of tissues. In addition to the synthetic polymer composition of the plastic, cells in conventional 2D culture systems adhere to surfaces that are nonphysiologically rigid (elastic modulus of >2 Gigapascals (GPa)) as opposed to the rigidity of normal tissue (hundreds of Pa in soft tissue to tens of thousands of Pa for stiffer tissues like cartilage and bone) [36, 37]. As the importance of ME in therapeutic response has become better understood, and more widely accepted, the urgency to identify tractable organotypic culture systems for studying human tissues in vitro has manifested. Matrigel, HuBiogel, and other commercially available laminin-rich ECM are widely used to provide 3D cell growth environments. These gels are used increasingly to study the impact of drugs on cells grown in 3D and have allowed for adoption of 3D culture to high-throughput systems where the rate-­ limiting steps are quality imaging and analysis [38]. Biopolymers used for 3D culture systems better mimic tissue than plastic as Matrigel has an elastic modulus between 400 Pa and 1 kPa and collagen gels can range from 500 Pa to >12  kPa depending on collagen concentration and attachment to the culture vessel. Matrigel, which is harvested from a rodent sarcoma cell line, is comprised of hundreds of proteins that can vary significantly in their exact composition and properties between production lots [39–41]. While Matrigel has made 3D high-throughput systems less daunting, placing human cells in an undefined rodent sarcoma 3D

T. A. Jokela et al.

context may not mimic the intended in vivo ME, and variability in the molecular components may confound interpretations and reproducibility of the results. Synthetic 3D culture hydrogels, such as polyethylene glycol-based systems, offer precision tunability of the elastic modulus, which covers a range similar to collagen gels while allowing for greater control over molecular composition [42]. Every in  vitro system for studying tissue ME sacrifices important aspects of the in vivo situation, but there is merit in studying microenvironmental properties in isolation. Although engineered and biopolymer-derived systems necessarily oversimplify TMEs, they can reveal important mechanistic elements of cellular responses by winnowing down the possible candidate pathways involved in a given functional response. The ME can be dissected into biophysical (e.g., rigidity, shear force), biochemical (e.g., ECM, growth factors, cytokines), and architectural (e.g., dimension and geometry) properties, and each property plays a role in regulating various cellular functions. For instance, by examining normal mammary epithelial cells in the context of matrix rigidity, in isolation from many other ME properties, we discovered age-dependent regulation of the mechanotransducing YAP and TAZ transcription factors [43]. Similarly, by using engineered polymer surfaces, we showed that substrate rigidity is a determinant for HER2-targeted therapeutic efficacy via YAP and TAZ signaling, both in vitro and in vivo [44]. In vitro screens of thousands of MEs comprised of unique combinations of ECM, growth factors, and cytokines revealed combinations that induced expression of CSC markers in malignant breast cancer cells that in turn led to increased resistance to chemotherapeutics [33]. In addition to rigidity and composition of the MEs being major determinants of cell fate, it has been found that shape and geometry contribute to producing diverse cellular responses [45, 46]. Kilian et al. found that similar shapes with pentagonal symmetry, but slightly different subcellular curvature, could bias differentiation of human mesenchymal stem cells toward either osteoblasts or adipocytes [45]. As MEs are deconstructed and different properties are studied individually or in defined combinations, the knowledge that we accrue over time allows us to form a portrait that models, and possibly explains, ME effects on cellular functions.

 ombinatorial Microenvironment Platforms C Mimic Diverse and Defined Milieus and Allow for High-Throughput Experimentation Established human cell lines and primary cells propagated in 2D culture are amenable to high-throughput experimentation. Potentially powerful tools for performing cancer drug design in TME contexts are being developed by merging the flexibil-

25  The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial…

ity of functional cell-based screening with the highly parallel nature of microarray-type experiments. A microarray is a tool that contains thousands of functionalized probes immobilized on a substrate. These tools provide both complexity and scalability and are used to explore diversity in various biological systems. Broadly speaking, the microarray technology can be classified into protein arrays, gene chips, or carbohydrate microarrays, depending upon what probes are immobilized on the substrate [47–49]. An interesting innovation in this technology space has been to fabricate microarrays in 2D and 3D contexts, printing substratum that supports adhesion of cultured cells. These types of combinatorial MEMAs facilitate highly parallel cell-­based functional screening. Indeed, using different ECM, soluble ligands, and pathway-blocking or pathway-­activating antibodies in various combinations as printed probes enables molecular dissection of complicated MEs and TMEs [26–29, 33, 42, 50–58]. While these array platforms create caricatures of in vivo MEs, they enable researchers to functionally define molecular components that maintain cell fate, thus revealing molecular regulators and pathways of cell states. We predict that this type of functional cell-based dissection of combinatorial MEs will have particular high impact in understanding normal and malignant human stem cells, because human in vivo experiments are essentially impossible. For instance, putative niche molecules and other tissue-specific ligands have been identified using MEMA, and validated in vivo in some cases, that were relevant to human embryonic [51], neural [28] and mammary stem cells [27, 29, 33]. MEMAs were also used to profile cell-ECM adhesion biases [59] and to optimize growth conditions of cultured cells [53, 60]. Taking a combinatorial approach, relative to a candidate-based approach, allows screening combinations of multiple tissue-­ specific ME molecules to identify extracellular cues that are the basis for emergent cell behaviors. Figure  25.1 summarizes several MEMA experiments that highlight the importance of including components of the TME when studying anticancer drug activity. TME context was found to produce dramatically different responses in various cancer cell lines even when using standard of care therapeutics. The successful application of MEMA requires managing many technical details that are, in many cases, on the edge of discovery themselves. The remainder of this chapter will elaborate on some of the issues that arise most often when producing MEMA on 2D substrates and provide some discussion of how we are managing them. There are relatively fewer examples of MEMA-type platforms in 3D, perhaps because some of the high-throughput liquid handling and 3D imaging requirements raise the barrier to entry; however, there are successful examples of 3D MEMAs [42, 61].

445

 electing the Printing Substrate: It Depends S on the Biological Questions Being Asked There are numerous materials used to immobilize ME molecules. But the primary objectives remain the same where MEMA fabrication is concerned: a suitable surface coating for printing molecules upon should provide high adsorption capacity, low cell attachment in areas not printed with molecules (i.e., non-fouling), and low spot-to-spot variation. Other important considerations include the capacity of the material to retain the structure, functionality, and binding sites of any given ME component. The most commonly used substrata for immobilizing ME components are untreated polystyrene plastic surfaces, chemically modified surfaces of glass slides (with aldehydes or epoxies), or polymer (e.g., polydimethylsiloxane (PDMS))-coated glass slides. Slides with these surfaces adsorb proteins through hydrophobic interactions, covalent bonds, or strong electrostatic interactions, respectively. Covalent modifications provide irreversible attachment; however, protein 3D structures may not be well maintained. Unintended cell attachment can also occur on all substrata. The addition of non-fouling coatings, like synthetic blocking copolymers or bovine serum albumin, can lessen the extent of nonspecific cell adherence. Another option is to coat glass surfaces with polyacrylamide (PA) or poly (ethylene glycol) (PEG) hydrogels. These hydrogels physically absorb proteins through relatively weak electrostatic interactions, which retain most of the native protein conformation at the cost of higher variation in protein-binding capacity [62]. One of the most convenient properties of PA and PEG gels is their native non-fouling character, which reduces nonspecific cell attachment. Rigidity of the substrate is another important property to consider. Untreated polystyrene and modified glass surfaces provide extremely stiff substratum (>2 GPa), which is not representative of the physical range of human tissues. PDMS is inexpensive, and its elastic modulus is easy to manipulate by altering the cure:polymer ratio, covering a range of elastic moduli similar to cartilage, skin, and tendon (0.6 to 3.5  MPa). PEG represents a range of elastic moduli from 0.5 to 1600  MPa. PA is another inexpensive substrate, which can be tuned from 150  Pa to 0.4  MPa, which is closer to the biological ME for soft tissues like brain and breast, and a reasonable range of elastic moduli to simulate normal and malignant breast, 100 to 4000  Pa [37, 63]. The substrate used for protein immobilization ultimately depends upon the characteristics of the cells used, the tissue being mimicked, and the outcomes being measured.

446

T. A. Jokela et al.

Cell type

Substrate and MEs Polyacrylamide hydrogel

A549 lung adenocarcinoma cells

HER2-amplified HCC1569 breast cancer cell line, HER2negative BT549 breast cancer cell line, A549 non-small cell lung cancer (NSCLC) cell line, PC3 prostate cancer cell line

AU565 breast cancer cell line (L-HER2+ subtype) and HCC1954 breast cancer cell line (HER2E subtype)

Isogenic human mammary epithelial cell malignant progression series: 184, 184A1, 184AA3

55 ME comprised of single and two-factor combinations of 10 ECM proteins Polyacrylamide hydrogel tuned to stiffness of 2.5 and 40 kPa 70 ME comprised of twofactor combinations of 7 ligands and 5 ECM proteins

Drugs Alkylating agent and 5 tyrosine kinase inhibitors: cisplatin, cabozantinib, nilotinib, gefitinib, sunitinib, vandetanib Lapatinib and Lapatinib + Verteporfin (YAP inhibitor)

Measured outcomes

Highlights

Reference

Cell number, BrdU, and cleaved caspase-3

Lung adenocarcinoma cells with known genetic drivers responded differently to therapeutics depending on ME context

(52)

Cellular morphology, HER2/pHER2 protein expression, and cell proliferation

Standard of care therapeutics in combination with drugs that target ME-imposed phenotypes can improve therapeutic response

(54)

TCP >2,500 ME comprised of combinations of 56 ligands and 46 ECM proteins Polyacrylamide hydrogel tuned to stiffness of 4.5 kPa 228 ME comprised of combinations 16 ligands, 4 cell surface and 13 ECM proteins

Lapatinib

Cell proliferation

ME-induced resistance is potent even in the face of combination therapy if the combination does not consider ME

(55)

Drug delivered in the wrong context can stimulate cancer cell growth

Paclitaxel (drug was added to specific ME and not to array screen)

Cellular morphology, AXL and c-KIT expression

Defined components of the CSC niche that produce drug resistant phenotypes and understand how a breakdown in the architecture of the tissue can contribute to the production of more potent cells capable of resisting chemotherapeutics

(33)

Fig. 25.1  Examples of MEMA platforms used to analyze anticancer drug response in various cancer cell lines

 EMA Data Analysis: Seeing the Forest M for the Trees The main goal of MEMA-type experiments is to provide causal links between cellular responses and specific MEs. Both inter- and intra-ME heterogeneity of cellular responses are to be expected and can be instructive about the continuum of phenotypic plasticity within the experimental system. Measuring heterogeneity of drug responses in a diversity of contexts may result in more realistic expectations of drug responses in  vivo. By incorporating sufficient numbers of

replicate features into the design of a MEMA, significant associations between MEs and cell phenotypes can be identified. Still, the high dimensionality of the data is a hindrance to the extraction of meaningful information. Most MEMA platforms use fluorescent probes to visualize biochemical and functional phenotypes and fluorescent and phase microscopy to capture colorimetric and morphological phenotypes. There are no specialized high-throughput imaging systems for this type of work currently available; however, microarray scanners and programmable, motorized epifluorescence or laser scanning confocal microscopes have been successfully used to acquire the necessary images [27, 42, 64].

25  The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial…

447

Fig. 25.2  Example of MEMA imaging and data analysis workflow

Imaging Quality Control When implementing automated high-throughput imaging, there is a need for practical quality control steps. The variability of substratum that is inherent to the MEMA platform demands an autofocusing step for each individual spot and therefore increases acquisition time and poses a challenge for high-content imaging. Failed autofocusing generates blurry images, which cause distortion in downstream analyses. Background artifacts and image saturation cause similar problems. Manual and automated approaches can be applied to exclude these poor quality images from the dataset [65]. For manual quality control inspection, we have developed a “montage”-image navigating tool, which allows quick browsing of the entire dataset using Fiji open-source software (Fig. 25.2) [64]. For an automated quality control system, machine learning algorithms work very efficiently because they can be trained to recognize images with artifacts and automatically exclude them from the dataset (Fig. 25.2) (described here [65]).

Cell Segmentation and Feature Extraction Flow cytometry and single-cell transcriptome sequencing analyses have shown that tissues and tumors contain substantially heterogeneous cell subpopulations [66], the significance of which is evident when identifying tumor cell subpopulations that play a significant role in initiating tumor metastasis, drug resistance, or relapse [66]. Likewise, tissue homeostasis and repair are driven by rare stem cell popula-

tions [67]. To achieve cellular resolution of cell-ME analyses, MEMA images must be analyzed with cell segmentation algorithms found in image analysis software, such as Fiji and CellProfiler (Fig. 25.2) [68–70]. Cell segmentation provides information about the morphological features of cells and provides further insight about ME-imposed changes to cellular phenotypes. In summary, cell segmentation and multi-­ parametric feature extraction in MEMA data analysis need to be part of an automated workflow that demands optimization and rigorous quality control to provide informative single-­ cell resolution, high-dimensional datasets.

Data Analysis Even in cases where a MEMA experiment is designed to have reasonably low complexity, e.g., 100 or fewer unique ME combinations, the analytical challenges are significant. The complexity of the information space generated by MEMA experiments increases rapidly when taking into consideration multiple ME properties such as rigidity, geometry, and molecular composition. And in addition to this, multi-­ parametric data collected at single-cell resolution provides many dependent variables that can make meaningful interpretations difficult. In practice, the statistical analysis of MEMA experiments is a rate-limiting step for this technology, and there are multiple solutions for addressing this challenge. The data processing workflow for MEMA experiments includes signal normalization, identification of functionally similar MEs by clustering, dimension reduction, data visualization, and pathway analysis (Fig. 25.2).

448

Data Normalization All microarray-like data contain some useful information and a significant degree of noise; thus, proper normalization is crucial. The data analysis begins with measuring fluorescence intensity or colorimetric density of each target protein in cells on each array feature. In this context, intensity typically reflects the relative abundance of the target protein. Intensities are impacted by factors such as the characteristics of the dye (antibody), spatial location, cell morphology, and uneven surfaces of the slides that cause inconsistent background [71]. Unlike DNA microarrays, which load the same amount of cDNA onto the array and then use total intensity as an internal reference, the number of cells attached to MEMA features varies by ME.  Thus, we may use the average of the entire signal from all cells on all or spatially defined array features as a reference for normalization of arrays of the same treatment condition. A signal emanating from cells on a control ME, which is known a priori to reproducibly bias toward a given phenotype, can be used as a reference [27, 57]. An alternative is to use whole cell staining dyes such as CellMask that fluorescently labels the entire cell, which can be used as an internal control to normalize fluorescent signal between cells.

Statistical Considerations The primary purpose of MEMA experiments is to identify the specific MEs that modulate certain cellular states by comparing cellular phenotypes between treatment and control groups. Compared to using a Student’s t-test, a Dunnett’s test is a better option for correcting false p-values due to multiple comparisons and identifying MEs that impose phenotypes that are significantly different from the control [27]. The Z-score standardization is a simple method used to identify meaningful groups that are distinct from the global mean. Z-scores have been used successfully to identify and optimize culture conditions for rare cell populations [72]. However, the Z-score has several limitations, like skewing of values due to outliers within a dataset as well as decreased accuracy when cell numbers are reduced. Moreover, the Z-score assumes that the data fit a Gaussian distribution, which is not the case in many biological systems. Thus, Guyon et  al. proposed the Φ-score as a cell-to-cell phenotypic scoring method for selecting the hit discovery in cell-­based assays (Table 25.1). The Φ-score ranks cells instead of averaging them and performs better than the Z-score despite the limitations previously mentioned. Indeed, the Φ-score can be more sensitive (more true hits) and more specific (fewer false positives) compared to other conventional methods [73]. Along with a large number of dependent variables, MEMA data often have many independent variables to contend with as well. It is biologically relevant to study the synergistic effects of ME factors and this is a greater challenge than studying the

T. A. Jokela et al. Table 25.1  Examples of data analysis and visualization techniques used with MEMA-type data Methods Φ-score

Type Advantages Normalization Overcomes the limitation of Z-score PCA Dimension A simple, reduction deterministic method to identify patterns due to variance tSNE, UMAP Visualization Nonlinear methods to identify patterns due to variance

Multiple Statistical test linear regression model, General Linear Model (GLM) Chi-Square Statistical test

Can analyze multiple independent variables

Suitable for categorical variables

Limitations

Only reflects linear relationships

Stochastic. Global trends are not accurately represented in mapping. Cannot handle incomplete data Data need to be linear and normally distributed

Sensitive to sample size

effect of each individual ME factor on cellular phenotypes. Multiple linear regression is a regression model capable of estimating the relationship between the dependent variable and two or more independent variables (Table 25.1). It expands upon the linear regression model and therefore makes many of the same assumptions as the linear regression model, so data need to be normally distributed and linear. With the multiple linear regression model, it is possible to estimate how specific ME components (independent variables) change cell phenotype variables (dependent variable) [33, 42, 54, 74]. When some independent variables are categorical variables, such as specific ME components, the appropriate regression model to employ is called the General Linear Model (GLM) (Table 25.1). If all variables are categorical, a Chi-squared test provides an effective way to evaluate whether there is a significant association between the variables (Table 25.1). The Chisquared test has been used to evaluate the co-occurrence of immune cell populations in patients with triple-negative breast cancer (TNBC) to define the coordinated immune response that occurs within the TNBC TME [75].

 lustering, Dimension Reduction, and Data C Visualization Clustering methods commonly used for DNA microarray datasets, such as hierarchical or k-means clustering, also are used with MEMA data to identify meaningful groups. Konagaya et  al. interrogated a relatively small number of

25  The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial…

growth factor combinations to optimize neural progenitor cell culture MEs and then used hierarchical cluster analysis to reveal three major clusters of ME combinations that facilitated growth versus astrocyte or neuron differentiation [60]. Although these analyses can reveal the meaningful groups within simple datasets, like traditional two-color DNA microarray data, they become ineffective as the complexity of the data increases which has been described as the “curse of dimensionality” [76]. To overcome the challenges of working with high-dimensional datasets, dimension reduction techniques have been developed. Improvements in computational processing power better facilitate analysis of high-dimensional data because they have enabled researchers to use algorithms that do not make painful compromises in the name of efficiency. Dimension reduction essentially distills vast amounts of information into snapshots that are emblematic of the underlying biology. Principal component analysis (PCA) is used for dimension reduction and can reveal the most variable factors that contribute to certain phenotypes (Table 25.1) [77]. However, not all biological questions are related to the variables with the highest variance in the dataset, and in such cases, PCA is less well-equiped to identify the contributing factors. Linear techniques such as PCA focus on separating dissimilar data points far away in low-dimensional representations after data transformation. However, biological data are often nonlinear, and for high-­dimensional data, it is usually more important to keep similar data points close together in low-dimensional representations, which is typically not feasible with linear mapping techniques. t-Distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) provide nonlinear approaches for high-dimensional data visualization (Table  25.1) [78, 79]. These two approaches consider the major and minor sources of variance within a dataset and represent them on a lower dimensional surface. Once the meaningful information has been extracted, the data collected from MEMA need to be connected to the existing body of knowledge to perform further biological validation.

Machine Learning In addition to feature extraction analysis, high content MEMA image sets provide excellent material for machine learning-based classification and clustering analysis. Deep learning algorithms, like convolution neural network (CNN), are useful and efficient ways to analyze MEMA image sets [80]. When supervised, the CNN model can be utilized during the image quality control step, where the algorithm can be trained to recognize and discard low-quality images. Another supervised approach is to use machine learning to classify MEMA images based on a provided training set that could include example images of cell morphology (differen-

449

tiated versus non differentiated) or by cell proliferation rate. Following training, the algorithms are able to classify the remaining images in the dataset automatically with varying levels of accuracy, specificity, and sensitivity. Unsupervised deep learning can be used to optimize single-cell segmentation and improve the accuracy of single-cell level analysis [81]. Unsupervised clustering of MEMA images provides the potential to compare how different MEs affect not only the morphology of a single cell, but also how specific MEs influence cell–cell contact and self-organization. Currently, most of the deep learning image analysis algorithms are trained on 2D images, but promising deep learning 3D image analysis algorithms have been published recently [82, 83], and in the future, these algorithms may be excellent tools used to study 3D-MEMA data.

Open Access MEMA Data MEMA is a core technology of the Microenvironment Perturbagen (MEP)—LINCS center (https://lincsproject. org/LINCS/centers/data-­and-­signature-­generating-­centers/ mep-­lincs), which is one of six Data and Signature Generating Centers of the NIH Library of Integrated Network-Based Cellular Signatures (LINCS) program [84]. The broad aim of the LINCS program is to advance the understanding of health and disease by identifying how a broad range of perturbations to different tissues and cells generate patterns of common networks and cellular responses. MEP-LINCS complements the LINCS program by elucidating how ME signals affect molecular networks to generate experimentally observable cellular phenotypes. To this end, MEP-LINCS employs MEMA experiments to measure how close to 3000 pairwise MEs impact the proliferation, differentiation status, and cell cycle of 22 cell lines. Using rigorous data analysis workflows, researchers addressed how ME-induced cellular phenotypes were regulated by specific molecular networks, whether certain ME subsets that produce similar phenotypic responses were due to modulation of common molecular networks, and if specific molecular networks were capable of producing multiple cellular phenotypes. Integration of MEP-LINCS generated datasets with other datasets from the LINCS program will allow researchers to determine whether ME-induced network changes bear similarity to genetic or chemically induced network changes in the same cell lines. MEP-LINCS generated data are freely available online (https://www.synapse.org/#!Synapse:syn2862345/wiki/), as are details of the data processing steps necessary to generate meaningful conclusions from MEMA experiments (data processing pipelines available at https://github.com/MEP-­ LINCS). Two of the several tools available on the website are the Virtual Lab and MEMA Board. The Virtual Lab tool allows user to interact with MEMA experiments by selecting cell lines, ECM proteins, and ligands and then observing

450

images for the selections along with selected data. The MEMA Board app allows for exploration of the datasets using dimension reduction methods and interactive heatmaps. The webpage is routinely updated with the most recent datasets which can be downloaded to a local device for further analysis.

Concluding Remarks/Summary One important, and wholly unexpected, outcome of the massive efforts to understand cancer as an entirely cell-intrinsic problem is the undeniable conclusion that mutations do not explain everything. Indeed, the fact that frankly malignant cells can be phenotypically normal, when held in check by a normal microenvironment, suggests that there is a dominant role of the microenvironment. New investments need to be made in technologies that facilitate the dissection and exploration of tissue MEs. MEMA-type platforms, and their successors, will provide opportunities to gain a comprehensive understanding of how the ME modulates drug responses in human cells and will provide functional cell-based data for preclinical drug screening that is ultimately more predictive of in vivo biology. These platforms are amenable to high-throughput scale­up using several imaging modalities for quantification. The main challenges of this approach are access to purified extracellular proteins, managing the combinatorial complexity to minimize cost and maximize the combinatorial space that is evaluated, data visualization, and statistical analysis to identify microenvironment components that contribute to a given outcome. An important component that is still in its infancy is robust network analysis that can provide a systematic understanding of how microenvironments are linked to activity in specific signaling pathways, which underlie cell phenotypes, and reveal candidates for further investigation. Tapping into the accumulated knowledge, represented in public databases and tools for pathway mapping like GO (Gene Ontology) enrichment analysis and KEGG (Kyoto Encyclopedia of Genes and Genomes), will increase the possibility that we can connect ME-imposed phenotypes to known signaling pathways and, hence, to cellular functions. Different ME components, such as ECM or substrate rigidity, are the input, and the measurements, such as morphometrics and other protein markers, are the output. The major object of pathway analysis is to delineate the relationship between input and output. There is an obvious need to improve preclinical drug discovery and evaluation. Overall, MEMAs are meant to address the shortcomings of experimentation that uses standard human cell culture models (i.e., the nonphysiological contexts), rodent models (i.e., the nonhuman context), and human beings (i.e., the intractable model). One of the approaches is to consider the microenvironmental impact on drug responses

T. A. Jokela et al.

during the earliest design stages of therapeutics. The combinatorial nature of MEMAs provides the advantages of exquisitely controlling microenvironmental properties and enabling high throughput. MEMA data are high-content, single-cell resolution and can capture cell-to-­cell heterogeneity; hence, it may provide a more realistic picture of drug performance. However, it remains to be seen whether data from MEMAtype experiments can build in vivo response models. Improved knowledge of microenvironmental impact on drug responses will economize and hasten drug development by making the preclinical stage more predictive and aid in the deployment of more precision therapeutics.

References 1. Wong CH, Siah KW, Lo AW.  Estimation of clinical trial success rates and related parameters. Biostatistics. 2018;20:273–86. 2. Light DW, Kantarjian H.  Market spiral pricing of cancer drugs. Cancer. 2013;119:3900–2. 3. Talmadge JE, Singh RK, Fidler IJ, Raz A. Murine models to evaluate novel and conventional therapeutic strategies for cancer. Am J Pathol. 2007;170:793–804. 4. Yue F, Cheng Y, Breschi A, Vierstra J, Wu W, Ryba T, Sandstrom R, Ma Z, Davis C, Pope BD, Shen Y, Pervouchine DD, Djebali S, Thurman RE, Kaul R, Rynes E, Kirilusha A, Marinov GK, Williams BA, Trout D, Amrhein H, Fisher-Aylor K, Antoshechkin I, DeSalvo G, See L-H, Fastuca M, Drenkow J, Zaleski C, Dobin A, Prieto P, Lagarde J, Bussotti G, Tanzer A, Denas O, Li K, Bender MA, Zhang M, Byron R, Groudine MT, McCleary D, Pham L, Ye Z, Kuan S, Edsall L, Wu Y-C, Rasmussen MD, Bansal MS, Kellis M, Keller CA, Morrissey CS, Mishra T, Jain D, Dogan N, Harris RS, Cayting P, Kawli T, Boyle AP, Euskirchen G, Kundaje A, Lin S, Lin Y, Jansen C, Malladi VS, Cline MS, Erickson DT, Kirkup VM, Learned K, Sloan CA, Rosenbloom KR, Lacerda de Sousa B, Beal K, Pignatelli M, Flicek P, Lian J, Kahveci T, Lee D, James Kent W, Ramalho Santos M, Herrero J, Notredame C, Johnson A, Vong S, Lee K, Bates D, Neri F, Diegel M, Canfield T, Sabo PJ, Wilken MS, Reh TA, Giste E, Shafer A, Kutyavin T, Haugen E, Dunn D, Reynolds AP, Neph S, Humbert R, Scott Hansen R, De Bruijn M, Selleri L, Rudensky A, Josefowicz S, Samstein R, Eichler EE, Orkin SH, Levasseur D, Papayannopoulou T, Chang K-H, Skoultchi A, Gosh S, Disteche C, Treuting P, Wang Y, Weiss MJ, Blobel GA, Cao X, Zhong S, Wang T, Good PJ, Lowdon RF, Adams LB, Zhou X-Q, Pazin MJ, Feingold EA, Wold B, Taylor J, Mortazavi A, Weissman SM, Stamatoyannopoulos JA, Snyder MP, Guigo R, Gingeras TR, Gilbert DM, Hardison RC, Beer MA, Ren B, The Mouse EC. A comparative encyclopedia of DNA elements in the mouse genome. Nature. 2014;515:355–64. 5. Uhl EW, Warner NJ.  Mouse models as predictors of human responses: evolutionary medicine. Curr Pathobiol Rep. 2015;3:219–23. 6. Weigelt B, Lo AT, Park CC, Gray JW, Bissell MJ. HER2 signaling pathway activation and response of breast cancer cells to HER2-­ targeting agents is dependent strongly on the 3D microenvironment. Breast Cancer Res Treat. 2010;122:35–43. 7. Baghban R, Roshangar L, Jahanban-Esfahlan R, Seidi K, Ebrahimi-­ Kalan A, Jaymand M, Kolahian S, Javaheri T, Zare P. Tumor microenvironment complexity and therapeutic implications at a glance. Cell Commun Signal. 2020;18:59. 8. Bissell MJ, Hines WC. Why don't we get more cancer? A proposed role of the microenvironment in restraining cancer progression. Nat Med. 2011;17:320.

25  The Influence of Tissue Architecture on Drug Response: Anticancer Drug Development in High-Dimensional Combinatorial… 9. Egeblad M, Nakasone ES, Werb Z.  Tumors as organs: complex tissues that interface with the entire organism. Dev Cell. 2010;18:884–901. 10. Junttila MR, de Sauvage FJ.  Influence of tumour micro environment heterogeneity on therapeutic response. Nature. 2013;501:346–54. 11. Mlecnik B, Bindea G, Kirilovsky A, Angell HK, Obenauf AC, Tosolini M, Church SE, Maby P, Vasaturo A, Angelova M.  The tumor microenvironment and Immunoscore are critical determinants of dissemination to distant metastasis. Sci Transl Med. 2016;8:327ra26. 12. Northcott JM, Dean IS, Mouw JK, Weaver VM. Feeling stress: the mechanics of cancer progression and aggression. Front Cell Dev Biol. 2018;6:17. 13. Mpekris F, Angeli S, Pirentis AP, Stylianopoulos T. Stress-mediated progression of solid tumors: effect of mechanical stress on tissue oxygenation, cancer cell proliferation, and drug delivery. Biomech Model Mechanobiol. 2015;14:1391–402. 14. Correia AL, Bissell MJ. The tumor microenvironment is a dominant force in multidrug resistance. Drug Resist Updat. 2012;15:39–49. 15. Son B, Lee S, Youn H, Kim E, Kim W, Youn B.  The role of tumor microenvironment in therapeutic resistance. Oncotarget. 2017;8:3933–45. 16. Flynn CM, Kaufman DS. Donor cell leukemia: insight into cancer stem cells and the stem cell niche. Blood. 2007;109:2688–92. 17. Ingangi V, Minopoli M, Ragone C, Motti ML, Carriero MV. Role of microenvironment on the fate of disseminating cancer stem cells. Front Oncol. 2019;9 18. Plaks V, Kong N, Werb Z. The cancer stem cell niche: how essential is the niche in regulating stemness of tumor cells? Cell Stem Cell. 2015;16:225–38. 19. Fuchs E, Tumbar T, Guasch G. Socializing with the neighbors: stem cells and their niche. Cell. 2004;116:769–78. 20. Scadden DT.  The stem-cell niche as an entity of action. Nature. 2006;441:1075–9. 21. Collins CA, Olsen I, Zammit PS, Heslop L, Petrie A, Partridge TA, Morgan JE.  Stem cell function, self-renewal, and behavioral heterogeneity of cells from the adult muscle satellite cell niche. Cell. 2005;122:289–301. 22. Nishimura EK, Jordan SA, Oshima H, Yoshida H, Osawa M, Moriyama M, Jackson IJ, Barrandon Y, Miyachi Y, Nishikawa S-I. Dominant role of the niche in melanocyte stem-cell fate determination. Nature. 2002;416:854–60. 23. Sacco A, Doyonnas R, Kraft P, Vitorovic S, Blau HM. Self-renewal and expansion of single transplanted muscle stem cells. Nature. 2008;456:502–6. 24. Bruno RD, Fleming JM, George AL, Boulanger CA, Schedin P, Smith GH. Mammary extracellular matrix directs differentiation of testicular and embryonic stem cells to form functional mammary glands in vivo. Sci Rep. 2017;7:40196. 25. Booth BW, Mack DL, Androutsellis-Theotokis A, McKay RD, Boulanger CA, Smith GH. The mammary microenvironment alters the differentiation repertoire of neural stem cells. Proc Natl Acad Sci U S A. 2008;105:14891–6. 26. Flaim CJ, Chien S, Bhatia SN. An extracellular matrix microarray for probing cellular differentiation. Nat Methods. 2005;2:119–25. 27. LaBarge MA, Nelson CM, Villadsen R, Fridriksdottir A, Ruth JR, Stampfer MR, Petersen OW, Bissell MJ.  Human mammary progenitor cell fate decisions are products of interactions with combinatorial microenvironments. Integr Biol (Camb). 2009;1:70–9. 28. Soen Y, Mori A, Palmer TD, Brown PO. Exploring the regulation of human neural precursor cell differentiation using arrays of signaling microenvironments. Mol Syst Biol. 2006;2:37. 29. Au-Lin C-H, Au-Lee JK, Au-LaBarge MA (2012) Fabrication and use of microenvironment microarrays (MEArrays). JoVE. e4152. 30. LaBarge MA, Petersen OW, Bissell MJ. Of microenvironments and mammary stem cells. Stem Cell Rev. 2007;3:137–46.

451

31. Jokela TA, LaBarge MA.  Integration of mechanical and ECM microenvironment signals in the determination of cancer stem cell states. Curr Stem Cell Rep. 2020. 32. LaBarge MA. The difficulty of targeting cancer stem cell niches. Clin Cancer Res. 2010;16:3121–9. 33. Jokela TA, Engelsen AST, Rybicka A, Pelissier Vatter FA, Garbe JC, Miyano M, Tiron C, Ferariu D, Akslen LA, Stampfer MR, Lorens JB, LaBarge MA. Microenvironment-induced non-sporadic expression of the AXL and cKIT receptors are related to epithelial plasticity and drug resistance. Front Cell Dev Biol. 2018;6:41. 34. Bissell MJ, LaBarge MA. Context, tissue plasticity, and cancer: Are tumor stem cells also regulated by the microenvironment? Cancer Cell. 2005;7:17–23. 35. Kenny PA, Bissell MJ.  Tumor reversion: correction of malig nant behavior by microenvironmental cues. Int J Cancer. 2003;107:688–95. 36. Engler AJ, Sen S, Sweeney HL, Discher DE.  Matrix elasticity directs stem cell lineage specification. Cell. 2006;126:677–89. 37. Paszek MJ, Zahir N, Johnson KR, Lakins JN, Rozenberg GI, Gefen A, Reinhart-King CA, Margulies SS, Dembo M, Boettiger D, Hammer DA, Weaver VM. Tensional homeostasis and the malignant phenotype. Cancer Cell. 2005;8:241–54. 38. Booij TH, Price LS, Danen EHJ. 3D cell-based assays for drug screens: challenges in imaging, image analysis, and high-content analysis. SLAS Discov. 2019;24:615–27. 39. Hansen KC, Kiemele L, Maller O, O’Brien J, Shankar A, Fornetti J, Schedin P. An in-solution ultrasonication-assisted digestion method for improved extracellular matrix proteome coverage. Mol Cell Proteomics. 2009;8:1648–57. 40. Soofi SS, Last JA, Liliensiek SJ, Nealey PF, Murphy CJ. The elastic modulus of Matrigel™ as determined by atomic force microscopy. J Struct Biol. 2009;167:216–9. 41. Hughes CS, Postovit LM, Lajoie GA. Matrigel: a complex protein mixture required for optimal growth of cell culture. Proteomics. 2010;10:1886–90. 42. Ranga A, Gobaa S, Okawa Y, Mosiewicz K, Negro A, Lutolf M. 3D niche microarrays for systems-level analyses of cell fate. Nat Commun. 2014;5:1–10. 43. Pelissier FA, Garbe JC, Ananthanarayanan B, Miyano M, Lin C, Jokela T, Kumar S, Stampfer MR, Lorens JB, LaBarge MA. Age-­ related dysfunction in mechanotransduction impairs differentiation of human mammary epithelial progenitors. Cell Rep. 2014;7:1926–39. 44. Lin C-H, Pelissier FA, Zhang H, Lakins J, Weaver VM, Park C, LaBarge MA.  Microenvironment rigidity modulates responses to the HER2 receptor tyrosine kinase inhibitor lapatinib via YAP and TAZ transcription factors. Mol Biol Cell. 2015;26:3946–53. 45. Kilian KA, Bugarija B, Lahn BT, Mrksich M. Geometric cues for directing the differentiation of mesenchymal stem cells. Proc Natl Acad Sci U S A. 2010;107:4872–7. 46. Yu S-M, Li B, Granick S, Cho Y-K.  Mechanical adaptations of epithelial cells on various protruded convex geometries. Cell. 2020;9:1434. 47. Sealfon SC, Chu TT.  RNA and DNA microarrays. Methods Mol Biol. 2011;671:3–34. 48. Sutandy FXR, Qian J, Chen C-S, Zhu H (2013) Overview of protein microarrays. Curr Protoc Protein Sci. Chapter 27: 2711–27.1. 49. Shin I, Park S, Mr L.  Carbohydrate microarrays: an advanced technology for functional studies of glycans. Chemistry. 2005;11:2894–901. 50. Flaim CJ, Teng D, Chien S, Bhatia SN.  Combinatorial signaling microenvironments for studying stem cell fate. Stem Cells Dev. 2008;17:29–39. 51. Brafman DA, Shah KD, Fellner T, Chien S, Willert K.  Defining long-term maintenance conditions of human embryonic stem cells with arrayed cellular microenvironment technology. Stem Cells Dev. 2009;18:1141–54.

452 52. Kaylan KB, Gentile SD, Milling LE, Bhinge KN, Kosari F, Underhill GH. Mapping lung tumor cell drug responses as a function of matrix context and genotype using cell microarrays. Integr Biol (Camb). 2016;8:1221–31. 53. Hou L, Kim JJ, Wanjare M, Patlolla B, Coller J, Natu V, Hastie TJ, Huang NF. Combinatorial extracellular matrix microenvironments for probing endothelial differentiation of human pluripotent stem cells. Sci Rep. 2017;7:6551. 54. Lin CH, Jokela T, Gray J, LaBarge MA.  Combinatorial microenvironments impose a continuum of cellular responses to a single pathway-­ targeted anti-cancer compound. Cell Rep. 2017;21:533–45. 55. Watson SS, Dane M, Chin K, Tatarova Z, Liu M, Liby T, Thompson W, Smith R, Nederlof M, Bucher E, Kilburn D, Whitman M, Sudar D, Mills GB, Heiser LM, Jonas O, Gray JW, Korkola JE.  Microenvironment-mediated mechanisms of resistance to HER2 inhibitors differ between HER2+ breast cancer subtypes. Cell Syst. 2018;6:329–42.e6. 56. Kaylan KB, Berg IC, Biehl MJ, Brougham-Cook A, Jain I, Jamil SM, Sargeant LH, Cornell NJ, Raetzman LT, Underhill GH. Spatial patterning of liver progenitor cell differentiation mediated by cellular contractility and Notch signaling. eLife. 2018;7:e38536. 57. Smith R, Devlin K, Kilburn D, Gross S, Sudar D, Bucher E, Nederlof M, Dane M, Gray JW, Heiser L, Korkola JE. Using microarrays to interrogate microenvironmental impact on cellular phenotypes in cancer. J Vis Exp. 2019. https://doi.org/10.3791/58957 58. Gaharwar AK, Arpanaei A, Andresen TL, Dolatshahi-Pirouz A. 3D biomaterial microarrays for regenerative medicine: current state-of-the-art, emerging directions and future trends. Adv Mater. 2016;28:771–81. 59. Kuschel C, Steuer H, Maurer AN, Kanzok B, Stoop R, Angres B. Cell adhesion profiling using extracellular matrix protein microarrays. BioTechniques. 2006;40:523–31. 60. Konagaya S, Kato K, Nakaji-Hirabayashi T, Arima Y, Iwata H.  Array-based functional screening of growth factors toward optimizing neural stem cell microenvironments. Biomaterials. 2011;32:5015–22. 61. Floren M, Tan W. Three-dimensional, soft neotissue arrays as high throughput platforms for the interrogation of engineered tissue environments. Biomaterials. 2015;59:39–52. 62. Angenendt P. Progress in protein and antibody microarray technology. Drug Discov Today. 2005;10:503–11. 63. Kim HN, Kang D-H, Kim MS, Jiao A, Kim D-H, Suh K-Y. Patterning methods for polymers in cell and tissue engineering. Ann Biomed Eng. 2012;40:1339–55. 64. Jokela T, Todhunter ME, LaBarge MA. High-throughput microenvironment microarray (MEMA) high resolution imaging. Biosens Biodetect. 2021; 65. Bray M-A, Carpenter AE.  Quality control for high-throughput imaging experiments using machine learning in CellProfiler. In: High content screening. Springer; 2018. p. 89–112. 66. Lawson DA, Kessenbrock K, Davis RT, Pervolarakis N, Werb Z.  Tumour heterogeneity and metastasis at single-cell resolution. Nat Cell Biol. 2018;20:1349–60. 67. Rué P, Martinez AA. Cell dynamics and gene expression control in tissue homeostasis and development. Mol Syst Biol. 2015;11:792. 68. Abramoff M, Magalhães P, Ram SJ.  Image processing with ImageJ. Biophoton Int. 2004;11:36–42. 69. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez J-Y, White DJ, Hartenstein V, Eliceiri K, Tomancak P, Cardona A. Fiji: an open-source platform for biological-image analysis. Nat Methods. 2012;9:676–82. 70. Kamentsky L, Jones TR, Fraser A, Bray MA, Logan DJ, Madden KL, Ljosa V, Rueden C, Eliceiri KW, Carpenter AE. Improved structure, function and compatibility for CellProfiler: modular high-throughput image analysis software. Bioinformatics. 2011;27:1179–80.

T. A. Jokela et al. 71. Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J, Speed TP. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res. 2002;30:e15. 72. Brafman DA, Chien S, Willert K. Arrayed cellular microenvironments for identifying culture and differentiation conditions for stem, primary and rare cell populations. Nat Protoc. 2012;7:703. 73. Guyon L, Lajaunie C, Fer F, Bhajun R, Sulpice E, Pinna G, Campalans A, Radicella JP, Rouillier P, Mary M. Φ-score: a cell-­ to-­cell phenotypic scoring method for sensitive and selective hit discovery in cell-based assays. Sci Rep. 2015;5:14221. 74. Edwards AL (1985) Multiple regression and the analysis of variance and covariance (WH Freeman/Times Books/Henry Holt & Co). 75. Keren L, Bosse M, Marquez D, Angoshtari R, Jain S, Varma S, Yang S-R, Kurian A, Van Valen D, West R, Bendall SC, Angelo M.  A structured tumor-immune microenvironment in triple negative breast cancer revealed by multiplexed ion beam imaging. Cell. 2018;174:1373–87.e19. 76. Bellman RE. Adaptive control processes: a guided tour. Princeton University Press; 1961. 77. Hilsenbeck SG, Friedrichs WE, Schiff R, O'Connell P, Hansen RK, Osborne CK, Fuqua SA. Statistical analysis of array expression data as applied to the problem of tamoxifen resistance. J Natl Cancer Inst. 1999;91:453–9. 78. McInnes L, Healy J, Melville J (2018) Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426. 79. Lvd M, Hinton G. Visualizing data using t-SNE. J Mach Learn Res. 2008;9:2579–605. 80. Moen E, Bannon D, Kudo T, Graf W, Covert M, Van Valen D.  Deep learning for cellular image analysis. Nat Methods. 2019;16:1233–46. 81. Yao K, Rochman ND, Sun SX. Cell type classification and unsupervised morphological phenotyping from low-resolution images using deep learning. Sci Rep. 2019;9:13467. 82. Haberl MG, Churas C, Tindall L, Boassa D, Phan S, Bushong EA, Madany M, Akay R, Deerinck TJ, Peltier ST, Ellisman MH.  CDeep3M—Plug-and-Play cloud-based deep learning for image segmentation. Nat Methods. 2018;15:677–80. 83. Tokuoka Y, Yamada TG, Hiroi NF, Kobayashi TJ, Yamagata K, Funahashi A.  Convolutional neural network-based instance segmentation algorithm to acquire quantitative criteria of early mouse development. bioRxiv. 2018. 324186. 84. Keenan AB, Jenkins SL, Jagodnik KM, Koplev S, He E, Torre D, Wang Z, Dohlman AB, Silverstein MC, Lachmann A, Kuleshov MV, Ma’ayan A, Stathias V, Terryn R, Cooper D, Forlin M, Koleti A, Vidovic D, Chung C, Schürer SC, Vasiliauskas J, Pilarczyk M, Shamsaei B, Fazel M, Ren Y, Niu W, Clark NA, White S, Mahi N, Zhang L, Kouril M, Reichard JF, Sivaganesan S, Medvedovic M, Meller J, Koch RJ, Birtwistle MR, Iyengar R, Sobie EA, Azeloglu EU, Kaye J, Osterloh J, Haston K, Kalra J, Finkbiener S, Li J, Milani P, Adam M, Escalante-Chong R, Sachs K, Lenail A, Ramamoorthy D, Fraenkel E, Daigle G, Hussain U, Coye A, Rothstein J, Sareen D, Ornelas L, Banuelos M, Mandefro B, Ho R, Svendsen CN, Lim RG, Stocksdale J, Casale MS, Thompson TG, Wu J, Thompson LM, Dardov V, Venkatraman V, Matlock A, Van Eyk JE, Jaffe JD, Papanastasiou M, Subramanian A, Golub TR, Erickson SD, Fallahi-Sichani M, Hafner M, Gray NS, Lin JR, Mills CE, Muhlich JL, Niepel M, Shamu CE, Williams EH, Wrobel D, Sorger PK, Heiser LM, Gray JW, Korkola JE, Mills GB, LaBarge M, Feiler HS, Dane MA, Bucher E, Nederlof M, Sudar D, Gross S, Kilburn DF, Smith R, Devlin K, Margolis R, Derr L, Lee A, Pillai A. The library of integrated network-based cellular signatures NIH Program: system-level cataloging of human cells response to perturbations. Cell Syst. 2018;6:13–24.

Models of Tumor Progression in Prostate Cancer

26

Waqas Azeem, Yaping Hua, Karl-Henning Kalland, Xisong Ke, Jan Roger Olsen, Anne Margrete Oyan, and Yi Qu

Abstract

Human prostate cancer is initiated in a benign prostate epithelial cell which gains the potential to progress to invasive and metastatic disease. The exact cell of origin of prostate cancer has not been finally determined. The plasticity of cell differentiation, the evolutionary potential of cancer cells, and differences between human and mouse prostate glands may underlie differences in the results from different experimental models. Numerous experimental models are available for the study of prostate cancer progression and include benign and transformed cells in monolayer or three-dimensional cultures, patient-­derived explant and organoid cultures, xenografted and genetically modified animals, and

animal models with spontaneous prostate cancer development. Technological developments, such as lab-on-a-chip and three-­dimensional tissue printing, are ongoing with the aim to achieve standardized, miniaturized, and high-capacity models that capture essential features of cancer development in vivo. Recently, the development of high-­resolution assays has opened new avenues to the understanding of the complexity of prostate carcinogenesis, including such techniques as single-cell sequencing and mass cytometry. The choice of model will depend on the exact question that will be investigated. Here we review different types of experimental models that are available for increasing insight into prostate carcinogenesis.

W. Azeem Department of Clinical Science, University of Bergen, Bergen, Norway Centre for Cancer Biomarkers CCBIO, University of Bergen, Bergen, Norway Y. Hua · J. R. Olsen Department of Clinical Science, University of Bergen, Bergen, Norway K.-H. Kalland (*) Department of Clinical Science, University of Bergen, Bergen, Norway Centre for Cancer Biomarkers CCBIO, University of Bergen, Bergen, Norway Department of Microbiology, Haukeland University Hospital, Bergen, Norway e-mail: [email protected]

X. Ke · Y. Qu Center for Chemical Biology, Institute of Interdisciplinary Integrative Medicine Research, Shanghai University of Traditional Chinese Medicine, Shanghai, People’s Republic of China A. M. Oyan Department of Clinical Science, University of Bergen, Bergen, Norway Department of Immunology and Transfusion Medicine, Haukeland University Hospital, Bergen, Norway

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. A. Akslen, R. S. Watnick (eds.), Biomarkers of the Tumor Microenvironment, https://doi.org/10.1007/978-3-030-98950-7_26

453

454

W. Azeem et al. Prostate cancer progression

Prostate cancer

Animal model Spontaneous Induced Xenografted Gene modified Immune competent

Primary Prostate cell

Prostate gland

2D cell culture 3D cell culture Xenograft

Stepwise tumorigenesis model

Co-culture 2D 3D

High capactiy & high resolution methodology Automation Miniaturization

 Heterogeneity and context dependency represent complicated challenges in prostate cancer research and therapy. Choice of appropriate experimental models is critical in experimental design and for thera-

Take-Home Lessons

• • • •

Large number of experimental models are available. Choice of model depends upon scientific question. Technological development is ongoing. High-resolution methods improve detailed insight.

Origin of the Prostate Cancer Cell The prostate gland consists of multiple small glandular elements embedded in a vascularized connective tissue stroma (Fig. 26.1). Each small glandular element is defined by an outer basement membrane on which a layer of basal epithelial cells is situated (Fig. 26.1). The master transcription factor, the androgen receptor (AR), is silenced in basal cells and expressed in luminal cells. One prevailing view regarding normal prostate epithelial differentiation is that when AR is induced by unknown mechanisms in the presence of androgen, basal cells differentiate into luminal (secretory) cells and a minor population of neuroendocrine cells. Early passages of primary prostate cells can recapitulate such differentiation, e.g. with androgen and fibroblast growth factor 7 (FG7) added to the growth medium [1]. Immortalized basal cells can be propagated indefinitely as transit amplifying (TA) cells in culture media with low calcium concentration [2], but we have found that such long-term cultures are very resistant to luminal differentiation [3]. The explanation could be that normal luminal cells become terminally differentiated, and consequently the selection pressure favors TA cells that do not differentiate. The lineage relationships between basal, luminal and neuroendocrine prostate cells, and in particular which one is the cell of origin of prostate cancer, and

peutic relevance. Single cell resolution, automation and bioinformatic analyses and an increasing number of available models promise advances ahead.

of putative prostate cancer stem cells (CSCs), have been vigorously debated [4, 5]. The bulk of prostate adenocarcinoma cells express mostly luminal cell expression patterns, but recently evidence has been provided to support all the different lineages of benign epithelial prostate cells as the cell of origin of prostate cancer [6–12]. These cumulative results underscore considerable plasticity of prostate cell differentiation, but also that there are differences between the human and mouse prostate and the experimental models used. There are, however, strong clues that the key regulatory mechanisms in normal prostate epithelial differentiation are retained in a perverted form in advanced prostate cancer. This notion is exemplified by the importance of the AR transcription factor during prostate cancer progression, including in castration resistant prostate cancer (CRPC) [10, 13], and by the neuroendocrine differentiation [14, 15] in end-­ stage prostate cancer. The unknown activation status of AR in putative prostate CSCs remains an important unresolved question with significant therapeutic consequences [10, 16].

 ell Culture Modeling of Prostate C Carcinogenesis Primary prostate epithelial cells (PrECs) can be obtained from biopsies and surgical material, as well as commercially, and can be propagated for a limited number of passages in monolayers until senescence ensues. PrECs have been immortalized using either hTERT (human telomerase reverse transcriptase) or the transforming elements of DNA viruses [17–19]. The 957E/hTERT cells [20, 21] and EP156T cells [22] were immortalized by exogenous expression of hTERT.  PZ-HPV7, CA-HPV10, and RWPE-1 cells were immortalized by human papilloma virus (HPV) transform-

26  Models of Tumor Progression in Prostate Cancer

455

ing elements [23]. Immortalization of PrECs has been achieved without exogenous gene expression [5, 24–26], but there is no model available of PrECs that spontaneously have transformed into malignant cell lines in vitro. Many attempts have been made to study the malignant transformation of benign prostate cells in culture, but the use of strong carcinogens or oncogenic viral elements was necessary to achieve transformation [27]. Forced transformation may be useful for many purposes, but is suboptimal when physiological mechanisms of transcriptional reprogramming during prostate carcinogenesis are investigated. Physiological selection pressure was applied to EP156T cells by keeping the cells in a confluent monolayer with regular replacement

of fresh growth medium. After several months, progeny EPT1 cells with reduced cell-to-cell contact inhibition dominated the culture. EPT1 cells had undergone EMT but were not tumorigenic [27]. EMT turned out to be the first step in the accumulation of malignant traits in a succession of progeny cells, eventually resulting in tumorigenic EPT3 cells (Fig. 26.2) [28, 29]. This model encompasses benign transit amplifying epithelial cells (EP156T), benign (EPT1) and pre-malignant (EPT2) mesenchymal type cells, tumorigenic (EPT3-N04/EPT3-PT1) and metastatic (EPT3-M1) cells in mice (Fig. 26.2). The very different phenotypes share a common genotype. Forensic-grade DNA microsatellite, karyotype and copy number break point analyses verified progeny

Fig. 26.1  Possible normal differentiation pathways from prostate stem cells to epithelial basal cells, luminal cells and neuroendocrine (NE) cells and possible transformation pathways to prostate cancer stem cells

and cancer cells (TA  =  Transit Amplifying cell). Stained histological sections of prostate benign (upper) and cancer (lower) tissue are from reference [142]

Epithelial

Phenotype Loss of contact inhibition EMT Postconfluent proliferation Apoptosis resistance Anchorage indep. growth Growth factor indep. growth Tumor formation in mouse Tumor metastases

EP156T

-

Mesenchymal

EPT1

+ + + -

Fig. 26.2 Overview of the EPT prostate stepwise tumorigenesis model. EPT1 cells were selected for loss of cell contact inhibition. EPT2 cells were selected from foci of confluent EPT1 cells and cloned in soft agar. EPT2-D5-HS were selected in protein-free medium. EPT3

Soft agar colonies

EPT2

+ + ++ + + + -

Prostaspheres

Tumor

EPT2-D5-HS

EPT3

NA

NA

+ +

+ + + + -

+ + +

NA NA

++ ++

cells were selected following subcutaneous injection. Cells were recovered from the EPT3 tumor. The progressive accumulation of malignant hallmarks is summarized [29]

456

authenticity [29]. Each of the different cell types can be passaged indefinitely and to high cell numbers in sub-confluent monolayers. Subpopulations of tumor initiating EPT3 cells (TICs) show activation of an autocrine IL6/STAT3 loop and show increased resistance to apoptosis and anoikis [29]. Genome-wide analyses revealed that epigenetic promoter patterns at different steps of the model corresponded strongly with coordinated expression changes of regulatory gene modules, such as HOX and microRNA genes, and structural gene modules, such as desmosome and adherens junction genes [27, 28, 30]. The model demonstrates, however, an absence of gene expression patterns characteristic of the bulk cellular population of prostate adenocarcinomas. Prostate luminal gene expression is strikingly absent, and the model is more likely to represent features of mesenchymal type cells in prostate cancer progression. In fact, evidence of EMT in the progression of primary prostate cancer has been shown in patient tissue [31]. However, the role of EMT, and its significance, in early prostate carcinogenesis, in metastasis and in the development of resistance to androgen deprivation treatment (ADT) and other prostate cancer therapy requires further investigation in available cell cultures, preclinical models and patient samples [32]. A particular pressing issue is the mounting evidence that ADT and highly potent inhibitors of AR function, such as enzalutamide, might induce EMT and more aggressive cancer, possibly involving prostate CSCs [3, 33, 34]. Alternative hypotheses have been discussed, such as the existence of a common progenitor prostate cancer stem cell that gives rise to both the neuroendocrine like and adenocarcinoma components and both these components continue to evolve and respond to selective pressures in parallel [34]. A very interesting feature of this EPT stepwise prostate tumorigenesis model is that exogenous expression of the AR in EP156T cells strongly induced the androgen-responsive AR target genes typical of prostate luminal cells, such as KLK3 (PSA), TMPRSS2, FKBP5, in EP156T epithelial cells, but endogenous AR was not induced [3]. One possible explanation of the very restricted expression of AR could be that its expression is selected against because of its ability to induce terminal luminal cell differentiation in the presence of androgen in this context. It is additionally noteworthy that in the mesenchymal context of EPT3 cells in this model, exogenous AR expression and androgen are unable to induce the AR target genes that are readily induced by this treatment in epithelial EP156T cells [3].

Prostate Cancer Cell Lines LNCaP, PC3, and DU145 and their metastatic derivatives are still the most widely used human prostate cancer cell lines despite the length of time these “classical” cell lines have

W. Azeem et al.

been in culture since isolated from human metastases [10, 19, 35–39]. It has proven to be difficult to establish stable cell lines from primary prostate cancer. LNCaP cells are androgen responsive in contrast to the AR negative PC3 and DU145 cell lines, but are less effective in forming tumors and metastatic colonies in mouse xenografts. Reviews summarize in  vitro models of AR signaling in prostate cancer [35] and useful cell lines for mouse xenografting [38]. Among additional prostate cancer cell lines, the VCaP and DuCaP cell lines express AR and the androgen-responsive TMPRSS-ERG fusion, and the 22Rv1 cell line is considered an in  vitro model of CRPC [35]. 22Rv1 cells express the androgen-independent AR splice variant denoted AR-V7 [40]. These and additional prostate cancer cell lines have provided important information on prostate cancer, but also have many limitations. With their origin in metastatic tissue and lack of exact passage history they cannot be used to recapitulate prostate carcinogenesis, and it is difficult to estimate which genetic changes are due to in vitro culture selection [10]. Thus, in one genome-wide ChIP-seq study only 3% overlap in AR binding sites were found between prostate cancer cell lines and prostate cancer tissue prior to treatment [41]. Since most cell lines were isolated from patients who had undergone treatment this could also be a factor in the differential gene expression.

I n Vitro Modeling of the Prostate Cancer Microenvironment and 3-Dimensional (3D) Growth Conditions Cancer cells develop, proliferate, and invade in crosstalk with a microenvironment consisting of fibroblasts, immune cells, vessels, and nerves embedded in a connective tissue matrix (Fig.  26.1). Both gene expression and functional properties have been shown to differ between the same cells cultured in 3-dimensional (3D) compared to 2D cultures, and with 3D conditions corresponding better to the in vivo features and with the advantage when models are sought for drug discovery and development. Many models recapitulate selected aspects of cancer growth in 3D or microenvironment conditions [42–45]. One simple experimental approximation to the in vivo situation is to co-culture prostate cancer cells and stromal cells in monolayer or double layers. Primary and stable cell lines can be embedded in a variety of matrix materials, e.g., collagen, fibronectin, vitronectin, or commercially available gels, such as Matrigel or Geltrex or alternatively synthetically bioengineered scaffolds that may support 3D growth of both benign and malignant prostate cells [46, 47]. Several techniques are available to support the 3D spheroid growth of prostate cells with or without extracellular matrices. When grown on surfaces with ultralow attachment,

26  Models of Tumor Progression in Prostate Cancer

prostate cells tend to form spheroids or prostaspheres resulting in the enrichment of cells with stem cell features [29]. Spheroids grown either in the extracellular matrix or in ultralow attachment plates or as hanging drops may all reproduce the nutrition, oxygen, and pH gradients that are found in cancer tissues that outgrow their blood supply [48–52]. Nanotechnology applied on bioreactors and lab-on-a-chip solutions is a developing field [43]. The potential advantages are to standardize experimental conditions, such as stiffness, shear stress, hydrostatic pressure, and concentration gradients, and by miniaturization to save on patient samples and reagents. Inbuilt sensor mechanisms could increase sensitivity and reduce analyses time [53, 54]. So far microfluidics, lab- or organ-on-a-chip and bioreactors have nevertheless not been widely applied in prostate cancer experimental models. Disadvantages include the cost and labor to establish such systems. A prostate-on-a-chip system to mimic the functional epithelial-stromal interface lining the ductal systems of a human prostate gland was recently reported [55]. 3D and even 4D bioprinting aiming to establish physiologically relevant tissue models promise future potential [56– 58]. 4D bioprinting infers that printed biomaterials are able to relevant change of their shape or function over time as a response to developmental or external stimuli [59, 60].

Organoid Cultures Versus Tissue Explants Both tissue culture explants and organoids are highly useful 3D experimental models whenever patient tissues are available. Ex vivo tissue explants are thin slices or sections of surgical cancer tissue in short-term cultures [48, 61–63]. It is possible to cut slices manually using a razor blade or scalpel, but commercial microslicers and microchoppers can prepare more standardized and thinner slices down to the micrometer scale [62]. Neighboring slices may then be processed for histology or immunohistochemistry to validate representative tissue. The advantage of tissue explants is that they preserve most of the features of the tissue architecture as well as its heterogeneity, although it is difficult to maintain and propagate the cultures for more than a few weeks. Ex vivo explants can be useful for hormone and drug testing. This model has potential usefulness in designing personalized medicine and in experiments that assay morphological or signal pathway changes in a tissue context. The basis of organoid cultures is the availability of a matrix that supports 3D growth in vitro and an essential cocktail of compounds that modulate defined signal transduction pathways. In this way, adult stem cells have been able to differentiate and self-organize into organoids that retain many features of the organ of origin [64, 65]. Organoid culture technology has significantly improved the success rate of establishing in  vitro cultures of cancer cells [66]. Organoid

457

technology has successfully generated benign epithelial prostate cultures [25, 67] and cultures that represent different subtypes of prostate cancer [68]. Prostate organoids have been established from metastatic cells, though the establishment of in  vitro cultures of primary prostate cancer cells remains a challenge [26, 68]. Organoid cultures have the additional advantage of being able to be propagated indefinitely and can be stored in liquid nitrogen as a living biobank. Compared to monolayer cultures of stable cancer cell lines, organoid cultures may recapitulate more features of original cancer although more experience needs to be gathered regarding the extent and for how many passages essential aspects of the original tumor can be preserved. The outcome will be very important for the use of organoids in personalized medicine in order to test drug sensitivity in vitro and to have an expandable antigen source that may be exploited in individualized immunoassays and dendritic cell-­based vaccine development in the immunotherapy field. In prostate cancer experimental research the availability of organoids established from distinct cancer subtypes should facilitate investigation of critical molecular signaling pathways. Tissue explants and organoids may find their place between traditional cell cultures and animal models [66, 69] (Fig. 26.3). In the design of animal experiments, it should be considered whether organoids could replace traditional cell lines for ethical, cost, and capacity reasons. The CRISPR-Cas9 genome editing system has transformed genome editing by its efficiency to knock out or knock in genes in cells and animals [70–72]. When used in combination with organoid technology, CRISPR-Cas9, and additionally induced pluripotent stem cell technology, may generate attractive experimental systems with systematic manipulation of single cancer-relevant genes or combinations of genes [25, 73–75]. The main disadvantage with organoids is their limited take rate. In particular for prostate cancer, the take rate may be very low and establishment may fail in most cases. Additionally, one should be aware that critical elements of the in vivo tissue microenvironment could be missing [44, 69].

Animal Models Animal models are useful for better understanding of how cancer cells interact with the tumor microenvironment and with the entire organism during metastasis. Spontaneous development of prostate cancer is relatively common in dogs and some rat strains, but less common in mouse strains. Mouse models can be broadly divided into xenograft models and genetically engineered models [76]. While immunodeficient mouse strains are critical for xenograft models, the current interest in immunotherapy has increased the demand for immunocompetent (syngeneic) and humanized mouse models [49, 77].

458 Fig. 26.3  Prostate cancer cell lines have most commonly been used for mouse xenograft models. Patient-derived xenografts (PDXs) have several advantages in retaining heterogeneity and features of original cancer tissue. Organoids can be established directly from prostate cancer tissues or via PDXs and vice versa. Patient derived organoids (PDOs) have advantages when it comes to capacity and biobanking and several experimental types. In the panel to the right are shown short-term organoids grown from a primary prostate cancer core biopsy obtained at our Haukeland University Hospital from a patient in a Phase I Clinical Trial of cryoimmunotherapy against metastatic castration resistant prostate cancer

W. Azeem et al.

Cancer cell lines

Patient-derived tumor xenografts Subcutaneous Organoids

Subrenal capsule

Intracardiac

Prostate gland

Mouse Xenograft Models Xenografts can be grown from any tumorigenic prostate cell culture or pieces of tissue. LNCaP, PC3, and DU145 are the three most commonly used prostate cancer cell lines in xenograft models and have provided insight into disease biology, but with limitations [10, 78]. Technically, the simplest approach is to inject tumorigenic cells subcutaneously with or without an intercellular matrix support, such as Matrigel. Tail vein injection or technically more demanding orthotopic injection into the prostate gland may be advantageous to answer questions related to metastasis and stromal invasion. Sub-renal capsular injection of cells that are otherwise difficult to graft may be successful in part due to the high vascularization at this site. Tissue recombination models in which dissociated adult prostate cells are combined with embryonic urogenital sinus mesenchymal cells and implanted under the renal capsule have been useful for cell differentiation studies and epithelial-stroma interactions [44, 79]. The choice of mouse strain, and in particular the extent of immunodeficiency, may also affect the efficiency of xenograft formation. The “nude” mouse was first established

Tail vein

more than 40 years ago, and the advancement of immunodeficient mice to model human tumor growth has been reviewed [44, 80].

Patient Derived Xenografts Although cell culture-based xenografts may provide useful information on cancer biology, these models have important limitations. In recent years the cancer research field has become highly aware of the importance of cancer cell heterogeneity which cannot be recapitulated by available cell culture-based xenografts. Patient-derived tumor xenografts (PDX) have emerged as a powerful technology: capable of retaining the molecular heterogeneity of their originating sample and have been shown to exhibit genomic clonal dynamics reminiscent of their originating tumor sample [81– 83]. In contrast to cell-based xenografts, PDXs have original tumor morphology. PDXs have the potential to improve basic research on cancer subtypes with specific genomic lesions and could provide mouse avatars in personalized medicine drug evaluation and co-clinical trials [44, 83–85]. The establishment of prostate cancer PDXs is still challeng-

26  Models of Tumor Progression in Prostate Cancer

ing, with a take rate of 10–40% and prolonged latency time, although immunodeficient mice have proved very helpful to establish serially transplantable PDXs [49, 74]. Relevant stromal or immune drivers of malignant progression could, however, be missing when immunodeficient mice are used as recipients [82].

 enetically Engineered Mouse Models G (GEMMS) In genetically engineered mouse models (GEMMs) selected genes are introduced or deleted in order to study the effect of defined pathways on carcinogenesis and tumor progression [39, 76, 86–91]. Advantages compared to xenografted mouse models are that GEMMs are compatible with intact immune systems and stromal microenvironments of the same species as the tumor. The limitation of GEMMs is related to differences between mouse and men, regarding prostate architecture, cancer propensity in rodents and the small size of mice compared to humans. The TRAMP model is one of the most commonly used early transgenic models [39, 92]. The model was generated by the introduction of a gene construct with the minimal rat probasin promoter driving expression of the SV40 virus early region. Androgen-responsive expression of the SV40 large T antigen inhibits p53 and Rb and the small t antigen inhibits protein phosphatase 2A [88]. C57BL/6 TRAMP mice develop prostatic intraepithelial neoplasia (PIN) by 3 months of age. PIN typically progresses to neuroendocrine carcinoma within half a year with lymph nodes and lungs as metastatic predilection sites. The model has been extensively used in preclinical testing and studies on carcinogenesis and tumor progression, but several limitations exist [39, 88, 89]. DNA virus oncoproteins, such as SV40 large T antigen or Human papilloma virus E6 or E7, have a special power to force cancer development, but these viruses have not been shown to induce prostate cancer. Furthermore, neuroendocrine differentiation is a feature of end-stage human prostate cancer and is seen in less than 2% of primary human cancers [14]. Furthermore, TRAMP mice rarely develop bone metastases, a common event in human patients. The LADY model provides a modification compared to the TRAMP model by using a larger region of the rat probasin promoter to drive the SV40 large T antigen expression without small t antigen. Thereby a panel of less aggressive tumor lines, collectively referred to as LADY, was generated to study cancer-preventing factors and the synergistic effects of different oncogenes [39, 76, 86–88, 90]. Transgenic mouse models have since been generated to study signal transduction pathways involved in prostate carcinogenesis and progression in humans. Overexpression of the transcription factor MYC is prevalent in early prostate

459

cancer [93], overexpression of which immortalizes primary prostate cells, induces PIN in normal prostate tissue, and stimulates the growth of both early-stage and CRPC [94, 95]. Mouse models of prostate cancer based on c-Myc have been reviewed [39, 76, 86–89, 96]. The most frequently mutated single genes in primary prostate cancers are SPOP, TP53, FOXA1 and PTEN [97]. In one analysis of 333 primary prostate carcinomas, 15% harbored homozygous deletions spanning the PTEN locus [98]. The homozygotic knockout of Pten is lethal in mouse embryos, while heterozygotic knockout results in a spectrum of prostate phenotypes that, combined with other genetic lesions, such as p27Kip1−/− or Nkx3.1−/− mice, result in the progression to PIN and invasive prostate cancer [86]. The health problems associated with Pten knockout mice and the value of this genetic background in the study of additional genes and pathways in prostate cancer have encouraged the development of several conditional Pten knockout mouse models [39, 76, 86–90, 96]. A system that introduces genes directly into the prostate glands of mice using tissue electroporation was used to study WNT signaling in prostate cancer metastasis [99].

Genomic Editing of Mouse Models Traditional GEMMs have exploited genetic engineering and homologous recombination of embryonic stem cells followed by injection of the manipulated stem cells into wild-­ type blastocysts. Selected chimeric mice are then crossed to generate single-gene knockout or double-mutant mice [100]. Genetic elements that allow inducible gene expression [101], such as tetracycline inducible element, or conditional knock-­ out [102], such as the Cre-Lox system, or knock-in [103], have further expanded the utility of GEMMs [104]. The generation of these useful models has, however, been costly and time-consuming. The CRISPR-Cas9 genome editing system may lead to a breakthrough in easy, fast and effective generation of precision mouse cancer models [75, 100, 105].

Prostate Neuroendocrine Tumor Models Small-cell neuroendocrine carcinoma is a rare form of primary prostate cancer [14]. When the common acinar adenocarcinoma has reached the stage of CRPC it can often still be efficiently targeted by AR inhibiting compounds, such as enzalutamide and darolutamide, and androgen synthesis inhibitor abiraterone. Tumor relapse is, however, the eventual outcome and often in the form of aggressive neuroendocrine cancer. Neuroendocrine trans-differentiation may provide important clues to the nature of putative prostate CSCs. Recently, high-resolution single-cell RNA sequenc-

460

ing has uncovered mechanisms of neuroendocrine differentiation [6, 11, 106]. Models of neuroendocrine prostate cancer, such as xenograft and genetically engineered mouse models, are of increasing importance [14, 15, 34, 107]. The TRAMP model described above generates neuroendocrine mouse tumors [39].

Bone Metastasis Models The bone is a predilection site for metastases of prostate cancer resulting in high morbidity associated with late stages of this disease [108–110]. Dogs spontaneously may develop benign prostate hyperplasia and prostate cancer with osteoblastic bone metastases similar to the natural course of prostate cancer in men [111]. DPC-1, Ace-1, Leo and Probasco represent four dog cell lines available for research on bone metastases, including xenograft models. The human prostate cancer cell line PC3 also forms bone metastases in xenograft models. A mouse model of bone metastasis was generated by grafting human lung and bone tissue followed by tail vein injection of LNCaP cells. The LNCaP cells preferentially metastasized to the human bone tissue [112]. A 3D in-vitro model to examine bone metastatic prostate cancer under dynamic conditions has been reported [113]. Available prostate cell lines and xenograft models of prostate bone metastases have been reviewed [42, 110, 114].

Spontaneous Cancer Development In general, mice do not develop spontaneous prostate cancer with an incidence that makes them useful prostate cancer models. ApcMin/+ mice, which were originally selected from randomly mutagenized mice, develop multiple intestinal neoplasia (min), presumably due to Apc gene inactivation and consequently β-catenin activation [115]. It has been shown that up to 40% of male ApcMin/+ mice developed histological features of both PIN and prostate carcinoma at 5 to 6 months of age, thus mimicking the early stages of prostate cancer in aging men [116] making this an interesting model not only for intestinal tumors, but also for prostate cancer [117]. Several rat models are prone to spontaneous or chemically induced prostate cancer [76, 114]. Almost one-third of Lobund-Wistar rats develop spontaneous androgen-sensitive metastatic prostate adenocarcinomas at a mean age of 26  months. These tumors subsequently become androgen independent and metastasize primarily to the lung. The model has been useful in studies of chemical and dietary effects on carcinogenesis [118].

W. Azeem et al.

Dog Models Many breeds of domestic dogs are prone to develop spontaneous age-dependent benign prostatic hyperplasia (BPH, high-grade prostatic intraepithelial neoplasia, and invasive prostate cancer [114, 119–121]. Development of bone metastases with mixed osteoblastic and osteolytic lesions, and the emergence of new woven bone in the later stages of prostate cancer are similarities shared between dogs and humans [122]. The use of next-generation sequencing and a greater level, and depth, of information generated by large-scale sequencing of human prostate cancers [98] could increase the utility of dog models in studies of the relevance of selected genes in cancer development and progression. Dog breeding records would facilitate association analysis and family-based linkage studies [119]. Pet dogs could be valuable in preparing for Phase I Clinical Trials of novel targeted therapies, immunotherapies, and personalized innovative combination therapies. When considering the dog model, potential advantages, such as animal size and the propensity of skeletal metastases, should be balanced against potential limitations. Specifically, in contrast to what has been observed in castrated humans, castration of dogs does not seem to protect against development of prostate cancer. Additionally, the effect of androgen deprivation therapy may differ between human and dog prostate cancers [114].

Model Organisms The zebrafish and fruit fly models have become useful in cancer research [123, 124]. These model organisms are particularly useful in the study of defined oncogenes and signal transduction pathways. The zebrafish has become a widely used model organism for prostate cancer research with several advantages regarding optical clarity, fecundity, rapid embryo development, and absence of immune system development until 14 days post-fertilization. Genomic tools have made possible disease modeling and large phenotype-based screens in zebrafish models [124]. The zebrafish model offers a rapid and inexpensive means of evaluating the metastatic potential of prostate cancer cells. By injection into the perivitelline space of 2  day old embryos, DU145 prostate cancer cells can be found throughout the body after only 24–48  h, and knockdown of WASF3 led to suppression of metastasis in zebrafish [125]. In addition, the zebrafish model can be used for identification of prostate tumor initiating cells from cultured cells and primary prostate cancer cells and shows advantages over mouse models in prediction of therapy response, because its translucent nature allows non-invasive observation of tumor progression in real time [126].

26  Models of Tumor Progression in Prostate Cancer

Zebrafish might also provide an excellent vertebrate tool to accelerate cancer drug discovery and development, including high-throughput screening, toxicology, and target identification. In our group, we have evaluated compounds that inhibit Wnt/β-catenin signaling in  vivo using a transgenic zebrafish harboring the Tcf/Lef-miniP:dGFP reporter [127]. Imaging of the fluorescent protein reporter allows real-time determination of drug potency, targeting specificity and body toxicity in vivo.

 ancer Immunotherapy and Co-Culture C Models Cancer immunotherapy represents a most encouraging breakthrough during the last decade. Prostate cancer is, however, not among the solid tumors that has responded best to immune checkpoint inhibitors or other immunotherapy and has been considered an immunologically “cold” cancer type. Experimental in  vitro models to investigate interactions between immune cells and cancers are improving [128]. Several published models could potentially be applied on prostate cancer. Co-cultures of autologous tumor organoids and peripheral blood lymphocytes were used to study invasion and attack of cytotoxic lymphocytes on cancer cells [129, 130]. Additional examples are provided by the effect on cancer stem cells in co-cultures with dendritic cells and immune cells [131] and assays of co-cultures between dendritic cells and keratinocytes [132]. Co-culture systems for exploring cancer immunology has been reviewed [133–135]. Animal models for experimental cancer immunotherapy pose challenges due to differences between human and mouse immune systems and because immunodeficient mice have been extensively used to establish xenografts, but relevant immune competent and humanized animal models are improving [136–139].

Future Perspectives Multiple experimental and preclinical models are available and under development for future prostate cancer research. Presently, no single experimental model can capture all features of the complicated biology and evolution that is ongoing in patient with prostate cancer. Mice represent the most common animal model in drug discovery and preclinical development [140], but the small size of the mouse may have its disadvantages as a tumor model. Even large tumor masses in mice could have a volume one thousand-fold smaller than a human tumor of the same stage. Consequently, the cancer cell number would be proportionately lower in mice tumors and the cancer cell heterogeneity problem could be underes-

461

timated. This could be one reason why mouse tumors often have a higher cure rate than what is found in subsequent clinical testing in patients. The trend nevertheless is that the repertoire and sophistication of experimental models are advancing, and so are high-resolution methods, such as single-­cell sequencing and mass cytometry, that may offer fresh looks if previously established models are revisited. In silico modeling is another developing field [135, 141]. Ultimately, the choice of model should be carefully evaluated during experimental design to secure optimal scientific results with due attention given to statistics, ethics, capacity, and costs.

References1 1. Lamb LE, Knudsen BS, Miranti CK.  E-cadherin-mediated survival of androgen-receptor-expressing secretory prostate epithelial cells derived from a stratified in vitro differentiation model. J Cell Sci. 2010;123(Pt 2):266–76. 2. Antony L, van der Schoor F, Dalrymple SL, Isaacs JT. Androgen receptor (AR) suppresses normal human prostate epithelial cell proliferation via AR/beta-catenin/TCF-4 complex inhibition of c-MYC transcription. Prostate. 2014;74(11):1118–31. 3. Olsen JR, Azeem W, Hellem MR, Marvyin K, Hua Y, Qu Y, et al. Context dependent regulatory patterns of the androgen receptor and androgen receptor target genes. BMC Cancer. 2016;16:377. 4. Lee SH, Shen MM. Cell types of origin for prostate cancer. Curr Opin Cell Biol. 2015;37:35–41. 5. Strand DW, Goldstein AS.  The many ways to make a luminal cell and a prostate cancer cell. Endocr Relat Cancer. 2015;22(6):T187–97. 6. Dong B, Miao J, Wang Y, Luo W, Ji Z, Lai H, et al. Single-cell analysis supports a luminal-neuroendocrine transdifferentiation in human prostate cancer. Commun Biol. 2020;3(1):778. 7. Karthaus WR, Hofree M, Choi D, Linton EL, Turkekul M, Bejnood A, et al. Regenerative potential of prostate luminal cells revealed by single-cell analysis. Science. 2020;368(6490):497–505. 8. Kwon OJ, Zhang L, Jia D, Zhou Z, Li Z, Haffner M, et  al. De novo induction of lineage plasticity from human prostate luminal epithelial cells by activated AKT1 and c-Myc. Oncogene. 2020;39(48):7142–51. 9. Li JJ, Shen MM. Prostate stem cells and cancer stem cells. Cold Spring Harb Perspect Med 2019;9(6). 10. Maitland NJ. Resistance to antiandrogens in prostate cancer: is it inevitable, intrinsic or induced? Cancers. 2021;13(2) 11. Park JW, Lee JK, Sheu KM, Wang L, Balanis NG, Nguyen K, et  al. Reprogramming normal human epithelial tissues to a common, lethal neuroendocrine cancer lineage. Science. 2018;362(6410):91–5. 12. Guo W, Li L, He J, Liu Z, Han M, Li F, et al. Single-cell transcriptomics identifies a distinct luminal progenitor cell type in distal prostate invagination tips. Nat Genet. 2020;52(9):908–18. 13. Watson PA, Arora VK, Sawyers CL.  Emerging mechanisms of resistance to androgen receptor inhibitors in prostate cancer. Nat Rev Cancer 2015. 14. Berman-Booty LD, Knudsen KE. Models of neuroendocrine prostate cancer. Endocr Relat Cancer. 2015;22(1):R33–49.

Space constraints allow only for selected references.

1 

462 15. Terry S, Beltran H. The many faces of neuroendocrine differentiation in prostate cancer progression. Front Oncol. 2014;4:60. 16. Rane JK, Pellacani D, Maitland NJ.  Advanced prostate cancer--a case for adjuvant differentiation therapy. Nat Rev Urol. 2012;9(10):595–602. 17. Rhim JS, Li H, Furusato B. Novel human prostate epithelial cell culture models for the study of carcinogenesis and of normal stem cells and cancer stem cells. Adv Exp Med Biol. 2011;720:71–80. 18. Sobel RE, Wang Y, Sadar MD. Molecular analysis and characterization of PrEC, commercially available prostate epithelial cells. In Vitro Cell Dev Biol Anim. 2006;42(1–2):33–9. 19. van Bokhoven A, Varella-Garcia M, Korch C, Johannes WU, Smith EE, Miller HL, et al. Molecular characterization of human prostate carcinoma cell lines. Prostate. 2003;57(3):205–25. 20. Litvinov IV, Vander Griend DJ, Xu Y, Antony L, Dalrymple SL, Isaacs JT.  Low-calcium serum-free defined medium selects for growth of normal prostatic epithelial stem cells. Cancer Res. 2006;66(17):8598–607. 21. Yasunaga Y, Nakamura K, Ewing CM, Isaacs WB, Hukku B, Rhim JS. A novel human cell culture model for the study of familial prostate cancer. Cancer Res. 2001;61(16):5969–73. 22. Kogan I, Goldfinger N, Milyavsky M, Cohen M, Shats I, Dobler G, et  al. hTERT-immortalized prostate epithelial and stromal-­ derived cells: an authentic in vitro model for differentiation and carcinogenesis. Cancer Res. 2006;66(7):3531–40. 23. Bello D, Webber MM, Kleinman HK, Wartinger DD, Rhim JS.  Androgen responsive adult human prostatic epithelial cell lines immortalized by human papillomavirus 18. Carcinogenesis. 1997;18(6):1215–23. 24. Jiang M, Strand DW, Fernandez S, He Y, Yi Y, Birbach A, et al. Functional remodeling of benign human prostatic tissues in vivo by spontaneously immortalized progenitor and intermediate cells. Stem Cells. 2010;28(2):344–56. 25. Karthaus WR, Iaquinta PJ, Drost J, Gracanin A, van Boxtel R, Wongvipat J, et  al. Identification of multipotent luminal progenitor cells in human prostate organoid cultures. Cell. 2014;159(1):163–75. 26. Palechor-Ceron N, Suprynowicz FA, Upadhyay G, Dakic A, Minas T, Simic V, et  al. Radiation induces diffusible feeder cell factor(s) that cooperate with ROCK inhibitor to conditionally reprogram and immortalize epithelial cells. Am J Pathol. 2013;183(6):1862–70. 27. Ke XS, Qu Y, Goldfinger N, Rostad K, Hovland R, Akslen LA, et  al. Epithelial to mesenchymal transition of a primary prostate cell line with switches of cell adhesion modules but without malignant transformation. PLoS One. 2008;3(10):e3368. 28. Ke XS, Li WC, Hovland R, Qu Y, Liu RH, McCormack E, et al. Reprogramming of cell junction modules during stepwise epithelial to mesenchymal transition and accumulation of malignant features in  vitro in a prostate cell model. Exp Cell Res. 2011;317(2):234–47. 29. Qu Y, Oyan AM, Liu R, Hua Y, Zhang J, Hovland R, et  al. Generation of prostate tumor-initiating cells is associated with elevation of reactive oxygen species and IL-6/STAT3 signaling. Cancer Res. 2013;73(23):7090–100. 30. Ke XS, Qu Y, Cheng Y, Li WC, Rotter V, Oyan AM, et al. Global profiling of histone and DNA methylation reveals epigenetic-­ based regulation of gene expression during epithelial to mesenchymal transition in prostate cells. BMC Genomics. 2010;11:669. 31. Gravdal K, Halvorsen OJ, Haukaas SA, Akslen LA.  A switch from E-cadherin to N-cadherin expression indicates epithelial to mesenchymal transition and is of strong and independent importance for the progress of prostate cancer. Clin Cancer Res. 2007;13(23):7003–11. 32. Xie D, Gore C, Liu J, Pong RC, Mason R, Hao G, et  al. Role of DAB2IP in modulating epithelial-to-mesenchymal transi-

W. Azeem et al. tion and prostate cancer metastasis. Proc Natl Acad Sci U S A. 2010;107(6):2485–90. 33. Jadaan DY, Jadaan MM, McCabe JP. Cellular plasticity in prostate cancer bone metastasis. Prostate Cancer. 2015;2015:651580. 34. Nouri M, Ratther E, Stylianou N, Nelson CC, Hollier BG, Williams ED.  Androgen-targeted therapy-induced epithelial mesenchymal plasticity and neuroendocrine transdifferentiation in prostate cancer: an opportunity for intervention. Front Oncol. 2014;4:370. 35. Sampson N, Neuwirt H, Puhr M, Klocker H, Eder IE.  In vitro model systems to study androgen receptor signaling in prostate cancer. Endocr Relat Cancer. 2013;20(2):R49–64. 36. Sobel RE, Sadar MD.  Cell lines used in prostate cancer research: a compendium of old and new lines--part 2. J Urol. 2005;173(2):360–72. 37. Sobel RE, Sadar MD.  Cell lines used in prostate cancer research: a compendium of old and new lines--part 1. J Urol. 2005;173(2):342–59. 38. Wu X, Gong S, Roy-Burman P, Lee P, Culig Z.  Current mouse and cell models in prostate cancer research. Endocr Relat Cancer. 2013;20(4):R155–70. 39. Saranyutanon S, Deshmukh SK, Dasgupta S, Pai S, Singh S, Singh AP. Cellular and molecular progression of prostate cancer: models for basic and preclinical research. Cancers. 2020;12(9). 40. Lu J, Van der Steen T, Tindall DJ.  Are androgen receptor variants a substitute for the full-length receptor? Nat Rev Urol. 2015;12(3):137–44. 41. Sharma NL, Massie CE, Ramos-Montoya A, Zecchini V, Scott HE, Lamb AD, et  al. The androgen receptor induces a distinct transcriptional program in castration-resistant prostate cancer in man. Cancer Cell. 2013;23(1):35–47. 42. Costard LS, Hosn RR, Ramanayake H, O'Briena BCF, Curtin CM. Influences of the 3D microenvironment on cancer cell behaviour and treatment responsiveness: a recent update on lung, breast and prostate cancer models. Acta Biomater 2021. 43. Fernandes DC, Canadas RF, Reis RL, Oliveira JM. Dynamic culture systems and 3D interfaces models for cancer drugs testing. Adv Exp Med Biol. 2020;1230:137–59. 44. Kato M, Sasaki T, Inoue T.  Current experimental human tissue-­derived models for prostate cancer research. Int J Urol. 2021;28(2):150–62. 45. Linxweiler J, Hammer M, Muhs S, Kohn M, Pryalukhin A, Veith C, et al. Patient-derived, three-dimensional spheroid cultures provide a versatile translational model for the study of organ-confined prostate cancer. J Cancer Res Clin Oncol. 2019;145(3):551–9. 46. Lopes D, Fernandes C, Nobrega JM, Patricio SG, Oliveira MB, Mano JF. Screening of perfused combinatorial 3D microenvironments for cell culture. Acta Biomater. 2019;96:222–36. 47. Unal AZ, West JL. Synthetic ECM: bioactive synthetic hydrogels for 3D tissue engineering. Bioconjug Chem. 2020;31(10):2253–71. 48. Ellem SJ, De-Juan-Pardo EM, Risbridger GP. In vitro modeling of the prostate cancer microenvironment. Adv Drug Deliv Rev. 2014;79-80:214–21. 49. Namekawa T, Ikeda K, Horie-Inoue K, Inoue S.  Application of prostate cancer models for preclinical study: advantages and limitations of cell lines, patient-derived xenografts, and three-­ dimensional culture of patient-derived cells. Cell 2019;8(1). 50. Kim SJ, Kim EM, Yamamoto M, Park H, Shin H.  Engineering multi-cellular spheroids for tissue engineering and regenerative medicine. Adv Healthc Mater. 2020:e2000608. 51. Fontana F, Raimondi M, Marzagalli M, Sommariva M, Gagliano N, Limonta P. Three-dimensional cell cultures as an in vitro tool for prostate cancer modeling and drug discovery. Int J Mol Sci 2020;21(18). 52. Rodriguez-Dorantes M, Cruz-Hernandez CD, Cortes-Ramirez SA, Cruz-Burgos JM, Reyes-Grajeda JP, Peralta-Zaragoza O, et al.

26  Models of Tumor Progression in Prostate Cancer Prostate cancer spheroids: a three-dimensional model for studying tumor heterogeneity. Methods Mol Biol. 2021;2174:13–7. 53. Han SJ, Park HK, Kim KS. Applications of microfluidic devices for urology. Int Neurourol J. 2017;21(Suppl 1):S4–9. 54. Egger D, Fischer M, Clementi A, Ribitsch V, Hansmann J, Kasper C. Development and characterization of a parallelizable perfusion bioreactor for 3D cell culture. Bioengineering (Basel) 2017;4(2). 55. Jiang L, Ivich F, Tahsin S, Tran M, Frank SB, Miranti CK, et al. Human stroma and epithelium co-culture in a microfluidic model of a human prostate gland. Biomicrofluidics. 2019;13(6):064116. 56. Pavlovich MJ, Hunsberger J, Atala A.  Biofabrication: a secret weapon to advance manufacturing, economies, and healthcare. Trends Biotechnol. 2016;34(9):679–80. 57. Dey M, Ozbolat IT. 3D bioprinting of cells, tissues and organs. Sci Rep. 2020;10(1):14023. 58. Gebeyehu A, Surapaneni SK, Huang J, Mondal A, Wang VZ, Haruna NF, et  al. Polysaccharide hydrogel based 3D printed tumor models for chemotherapeutic drug screening. Sci Rep. 2021;11(1):372. 59. Peng W, Unutmaz D, Ozbolat IT. Bioprinting towards physiologically relevant tissue models for pharmaceutics. Trends Biotechnol. 2016;34(9):722–32. 60. Shah Mohammadi M, Buchen JT, Pasquina PF, Niklason L, Alvarez LM, Jariwala SH. Critical considerations for regeneration of vascularized composite tissues. Tissue Eng Part B Rev 2020. 61. Centenera MM, Raj GV, Knudsen KE, Tilley WD, Butler LM. Ex vivo culture of human prostate tissue and drug development. Nat Rev Urol. 2013;10(8):483–7. 62. Rosales Gerpe MC, van Vloten JP, Santry LA, de Jong J, Mould RC, Pelin A, et al. Use of precision-cut lung slices as an ex vivo tool for evaluating viruses and viral vectors for gene and oncolytic therapy. Mol Ther Methods Clin Dev. 2018;10:245–56. 63. Tieu T, Irani S, Bremert KL, Ryan NK, Wojnilowicz M, Helm M, et al. Patient-derived prostate cancer explants: a clinically relevant model to assess siRNA-based nanomedicines. Adv Healthc Mater 2020:e2001594. 64. Jung P, Sato T, Merlos-Suarez A, Barriga FM, Iglesias M, Rossell D, et al. Isolation and in vitro expansion of human colonic stem cells. Nat Med. 2011;17(10):1225–7. 65. Sato T, Vries RG, Snippert HJ, van de Wetering M, Barker N, Stange DE, et  al. Single Lgr5 stem cells build crypt-­villus structures in  vitro without a mesenchymal niche. Nature. 2009;459(7244):262–5. 66. Sachs N, Clevers H. Organoid cultures for the analysis of cancer phenotypes. Curr Opin Genet Dev. 2014;24:68–73. 67. Chua CW, Shibata M, Lei M, Toivanen R, Barlow LJ, Bergren SK, et al. Single luminal epithelial progenitors can generate prostate organoids in culture. Nat Cell Biol 2014;16(10):951–61, 1–4. 68. Gao D, Vela I, Sboner A, Iaquinta PJ, Karthaus WR, Gopalan A, et al. Organoid cultures derived from patients with advanced prostate cancer. Cell. 2014;159(1):176–87. 69. Gleave AM, Ci X, Lin D, Wang Y. A synopsis of prostate organoid methodologies, applications, and limitations. Prostate. 2020;80(6):518–26. 70. Cho SW, Kim S, Kim JM, Kim JS.  Targeted genome engineering in human cells with the Cas9 RNA-guided endonuclease. Nat Biotechnol. 2013;31(3):230–2. 71. Cong L, Ran FA, Cox D, Lin S, Barretto R, Habib N, et  al. Multiplex genome engineering using CRISPR/Cas systems. Science. 2013;339(6121):819–23. 72. Mali P, Yang L, Esvelt KM, Aach J, Guell M, DiCarlo JE, et al. RNA-guided human genome engineering via Cas9. Science. 2013;339(6121):823–6. 73. Drost J, van Jaarsveld RH, Ponsioen B, Zimberlin C, van Boxtel R, Buijs A, et al. Sequential cancer mutations in cultured human intestinal stem cells. Nature. 2015;521(7550):43–7.

463 74. Hepburn AC, Sims CHC, Buskin A, Heer R. Engineering prostate cancer from induced pluripotent stem cells-new opportunities to develop preclinical tools in prostate and prostate cancer studies. Int J Mol Sci 2020;21(3). 75. Hazafa A, Mumtaz M, Farooq MF, Bilal S, Chaudhry SN, Firdous M, et al. CRISPR/Cas9: a powerful genome editing technique for the treatment of cancer cells with present challenges and future directions. Life Sci. 2020;263:118525. 76. Ittmann M, Huang J, Radaelli E, Martin P, Signoretti S, Sullivan R, et al. Animal models of human prostate cancer: the consensus report of the New York meeting of the mouse models of human cancers consortium prostate pathology committee. Cancer Res. 2013;73(9):2718–36. 77. Wang M, Yao LC, Cheng M, Cai D, Martinek J, Pan CX, et al. Humanized mice in studying efficacy and mechanisms of PD-1-­ targeted cancer immunotherapy. FASEB J. 2018;32(3):1537–49. 78. Toivanen R, Taylor RA, Pook DW, Ellem SJ, Risbridger GP.  Breaking through a roadblock in prostate cancer research: an update on human model systems. J Steroid Biochem. 2012;131(3–5):122–31. 79. Zong Y, Goldstein AS, Witte ON.  Dissociated prostate regeneration under the renal capsule. Cold Spring Harb Protoc 2015;2015(11):pdb prot078063. 80. Shultz LD, Goodwin N, Ishikawa F, Hosur V, Lyons BL, Greiner DL. Human cancer growth and therapy in immunodeficient mouse models. Cold Spring Harb Protoc. 2014;2014(7):694–708. 81. Eirew P, Steif A, Khattra J, Ha G, Yap D, Farahani H, et  al. Dynamics of genomic clones in breast cancer patient xenografts at single-cell resolution. Nature. 2015;518(7539):422–6. 82. Cassidy JW, Caldas C, Bruna A.  Maintaining tumor heterogeneity in patient-derived tumor xenografts. Cancer Res. 2015;75(15):2963–8. 83. Palanisamy N, Yang J, Shepherd PDA, Li-Ning-Tapia EM, Labanca E, Manyam GC, et al. The MD Anderson prostate cancer patient-­ derived xenograft series (MDA PCa PDX) captures the molecular landscape of prostate cancer and facilitates marker-driven therapy development. Clin Cancer Res. 2020;26(18):4933–46. 84. Malaney P, Nicosia SV, Dave V.  One mouse, one patient paradigm: new avatars of personalized cancer therapy. Cancer Lett. 2014;344(1):1–12. 85. Navone NM, van Weerden WM, Vessella RL, Williams ED, Wang Y, Isaacs JT, et  al. Movember GAP1 PDX project: an international collection of serially transplantable prostate cancer patient-­derived xenograft (PDX) models. Prostate. 2018;78(16): 1262–82. 86. Grabowska MM, DeGraff DJ, Yu X, Jin RJ, Chen Z, Borowsky AD, et al. Mouse models of prostate cancer: picking the best model for the question. Cancer Metastasis Rev. 2014;33(2–3):377–97. 87. Irshad S, Abate-Shen C. Modeling prostate cancer in mice: something old, something new, something premalignant, something metastatic. Cancer Metast Rev. 2013;32(1–2):109–22. 88. Jeet V, Russell PJ, Khatri A.  Modeling prostate cancer: a perspective on transgenic mouse models. Cancer Metastasis Rev. 2010;29(1):123–42. 89. Parisotto M, Metzger D. Genetically engineered mouse models of prostate cancer. Mol Oncol. 2013;7(2):190–205. 90. Saxena M, Christofori G.  Rebuilding cancer metastasis in the mouse. Mol Oncol. 2013;7(2):283–96. 91. van Marion DM, Domanska UM, Timmer-Bosscha H, Walenkamp AM. Studying cancer metastasis: Existing models, challenges and future perspectives. Crit Rev Oncol Hematol. 2015. 92. Greenberg NM, DeMayo F, Finegold MJ, Medina D, Tilley WD, Aspinall JO, et al. Prostate cancer in a transgenic mouse. Proc Natl Acad Sci U S A. 1995;92(8):3439–43. 93. Hawksworth D, Ravindranath L, Chen Y, Furusato B, Sesterhenn IA, McLeod DG, et  al. Overexpression of C-MYC oncogene in

464 prostate cancer predicts biochemical recurrence. Prostate Cancer Prostatic Dis. 2010;13(4):311–5. 94. Gil J, Kerai P, Lleonart M, Bernard D, Cigudosa JC, Peters G, et al. Immortalization of primary human prostate epithelial cells by c-Myc. Cancer Res. 2005;65(6):2179–85. 95. Iwata T, Schultz D, Hicks J, Hubbard GK, Mutton LN, Lotan TL, et  al. MYC overexpression induces prostatic intraepithelial neoplasia and loss of Nkx3.1 in mouse luminal epithelial cells. PLoS One 2010;5(2):e9427. 96. Mimeault M, Batra SK.  Animal models relevant to human prostate carcinogenesis underlining the critical implication of prostatic stem/progenitor cells. Biochim Biophys Acta. 2011;1816(1):25–37. 97. Barbieri CE, Rubin MA.  Molecular characterization of prostate cancer following androgen deprivation: the devil in the details. Eur Urol. 2014;66(1):40–1. 98. Cancer Genome Atlas Research Network. Electronic address scmo, Cancer Genome Atlas Research N. The Molecular Taxonomy of Primary Prostate Cancer. Cell 2015;163(4):1011–1025. 99. Leibold J, Ruscetti M, Cao Z, Ho YJ, Baslan T, Zou M, et  al. Somatic tissue engineering in mouse models reveals an actionable role for WNT pathway alterations in prostate cancer metastasis. Cancer Discov. 2020;10(7):1038–57. 100. Mou HW, Kennedy Z, Anderson DG, Yin H, Xue W.  Precision cancer mouse models through genome editing with CRISPR-­ Cas9. Genome Med 2015;7. 101. Saunders TL. Inducible transgenic mouse models. Methods Mol Biol. 2011;693:103–15. 102. Friedel RH, Wurst W, Wefers B, Kuhn R. Generating conditional knockout mice. Methods Mol Biol. 2011;693:205–31. 103. Roebroek AJ, Gordts PL, Reekmans S.  Knock-in approaches. Methods Mol Biol. 2011;693:257–75. 104. Kasper S.  Survey of genetically engineered mouse models for prostate cancer: analyzing the molecular basis of prostate cancer development, progression, and metastasis. J Cell Biochem. 2005;94(2):279–97. 105. Huijbers IJ.  Generating genetically modified mice: a decision guide. Methods Mol Biol. 2017;1642:1–19. 106. Yamada Y, Beltran H. Clinical and biological features of neuroendocrine prostate cancer. Curr Oncol Rep. 2021;23(2):15. 107. Yu C, Hu K, Nguyen D, Wang ZA. From genomics to functions: preclinical mouse models for understanding oncogenic pathways in prostate cancer. Am J Cancer Res. 2019;9(10):2079–102. 108. Coleman RE.  Clinical features of metastatic bone disease and risk of skeletal morbidity. Clin Cancer Res. 2006;12(20 Pt 2):6243s–9s. 109. Mehra R, Kumar-Sinha C, Shankar S, Lonigro RJ, Jing X, Philips NE, et  al. Characterization of bone metastases from rapid autopsies of prostate cancer patients. Clin Cancer Res. 2011;17(12):3924–32. 110. Simmons JK, Hildreth BE 3rd, Supsavhad W, Elshafae SM, Hassan BB, Dirksen WP, et al. Animal models of bone metastasis. Vet Pathol. 2015;52(5):827–41. 111. LeRoy BE, Thudi NK, Nadella MV, Toribio RE, Tannehill-Gregg SH, van Bokhoven A, et al. New bone formation and osteolysis by a metastatic, highly invasive canine prostate carcinoma xenograft. Prostate. 2006;66(11):1213–22. 112. Yonou H, Yokose T, Kamijo T, Kanomata N, Hasebe T, Nagai K, et al. Establishment of a novel species- and tissue-specific metastasis model of human prostate cancer in humanized non-obese diabetic/severe combined immunodeficient mice engrafted with human adult lung and bone. Cancer Res. 2001;61(5):2177–82. 113. Jasuja H, Kar S, Katti DR, Katti K. Perfusion bioreactor enabled fluid-derived shear stress conditions for novel bone metastatic prostate cancer testbed. Biofabrication 2021.

W. Azeem et al. 114. Winter SF, Cooper AB, Greenberg NM.  Models of metastatic prostate cancer: a transgenic perspective. Prostate Cancer Prostatic Dis. 2003;6(3):204–11. 115. Bruxvoort KJ, Charbonneau HM, Giambernardi TA, Goolsby JC, Qian CN, Zylstra CR, et al. Inactivation of Apc in the mouse prostate causes prostate carcinoma. Cancer Res. 2007;67(6):2490–6. 116. Poutahidis T, Rao VP, Olipitz W, Taylor CL, Jackson EA, Levkovich T, et al. CD4+ lymphocytes modulate prostate cancer progression in mice. Int J Cancer. 2009;125(4):868–78. 117. Valkenburg KC, Hostetter G, Williams BO. Concurrent Hepsin overexpression and adenomatous polyposis coli d­ eletion causes invasive prostate carcinoma in mice. Prostate. 2015;75(14):1579–85. 118. Pollard M, Suckow MA.  Dietary prevention of hormone refractory prostate cancer in Lobund-Wistar rats: a review of studies in a relevant animal model. Comp Med. 2006;56(6):461–7. 119. Davis BW, Ostrander EA. Domestic dogs and cancer research: a breed-based genomics approach. ILAR J. 2014;55(1):59–68. 120. Leroy BE, Northrup N. Prostate cancer in dogs: comparative and clinical aspects. Vet J. 2009;180(2):149–62. 121. Waters DJ, Bostwick DG.  The canine prostate is a spontaneous model of intraepithelial neoplasia and prostate cancer progression. Anticancer Res. 1997;17(3A):1467–70. 122. Rosol TJ, Tannehill-Gregg SH, LeRoy BE, Mandl S, Contag CH.  Animal models of bone metastasis. Cancer. 2003;97(3 Suppl):748–57. 123. White RM. Cross-species oncogenomics using zebrafish models of cancer. Curr Opin Genet Dev. 2015;30:73–9. 124. Amawi H, Aljabali AAA, Boddu SHS, Amawi S, Obeid MA, Ashby CR, Jr., et al. The use of zebrafish model in prostate cancer therapeutic development and discovery. Cancer Chemother Pharmacol 2021. 125. Teng Y, Xie X, Walker S, White DT, Mumm JS, Cowell JK.  Evaluating human cancer cell metastasis in zebrafish. BMC Cancer. 2013;13:453. 126. Bansal N, Davis S, Tereshchenko I, Budak-Alpdogan T, Zhong H, Stein MN, et al. Enrichment of human prostate cancer cells with tumor initiating properties in mouse and zebrafish xenografts by differential adhesion. Prostate. 2014;74(2):187–200. 127. Shimizu N, Kawakami K, Ishitani T.  Visualization and exploration of Tcf/Lef function using a highly responsive Wnt/ beta-catenin signaling-reporter transgenic zebrafish. Dev Biol. 2012;370(1):71–85. 128. Yunger S, Bar El A, Zeltzer LA, Fridman E, Raviv G, Laufer M, et al. Tumor-infiltrating lymphocytes from human prostate tumors reveal anti-tumor reactivity and potential for adoptive cell therapy. Onco Targets Ther. 2019;8(12):e1672494. 129. Cattaneo CM, Dijkstra KK, Fanchi LF, Kelderman S, Kaing S, van Rooij N, et al. Tumor organoid-T-cell coculture systems. Nat Protoc. 2020;15(1):15–39. 130. Dijkstra KK, Cattaneo CM, Weeber F, Chalabi M, van de Haar J, Fanchi LF, et  al. Generation of tumor-reactive T cells by co-­ culture of peripheral blood lymphocytes and tumor organoids. Cell. 2018;174(6):1586–98 e12. 131. Yang T, Zhang W, Wang L, Xiao C, Wang L, Gong Y, et  al. Co-culture of dendritic cells and cytokine-induced killer cells effectively suppresses liver cancer stem cell growth by inhibiting pathways in the immune system. BMC Cancer. 2018;18(1):984. 132. Thelu A, Catoire S, Kerdine-Romer S. Immune-competent in vitro co-culture models as an approach for skin sensitisation assessment. Toxicology In Vitro. 2020;62:104691. 133. Fitzgerald AA, Li E, Weiner LM. 3D culture systems for exploring cancer immunology. Cancers 2020;13(1). 134. Ye W, Luo C, Li C, Huang J, Liu F. Organoids to study immune functions, immunological diseases and immunotherapy. Cancer Lett. 2020;477:31–40.

26  Models of Tumor Progression in Prostate Cancer 135. Lee MW, Miljanic M, Triplett T, Ramirez C, Aung KL, Eckhardt SG, et al. Current methods in translational cancer research. Cancer Metastasis Rev 2020. 136. Chulpanova DS, Kitaeva KV, Rutland CS, Rizvanov AA, Solovyeva VV. Mouse tumor models for advanced cancer immunotherapy. Int J Mol Sci 2020;21(11). 137. Schachtschneider KM, Schwind RM, Newson J, Kinachtchouk N, Rizko M, Mendoza-Elias N, et al. The oncopig cancer model: an innovative large animal translational oncology platform. Front Oncol. 2017;7:190. 138. Yin L, Wang XJ, Chen DX, Liu XN, Wang XJ. Humanized mouse model: a review on preclinical applications for cancer immunotherapy. Am J Cancer Res. 2020;10(12):4568–84.

465 139. Morillon YM 2nd, Sabzevari A, Schlom J, Greiner JW. The development of next-generation PBMC humanized mice for preclinical investigation of cancer immunotherapeutic agents. Anticancer Res. 2020;40(10):5329–41. 140. Valkenburg KC, Pienta KJ.  Drug discovery in prostate cancer mouse models. Expert Opin Drug Discov. 2015;10(9):1011–24. 141. Wang Y, Xing J, Xu Y, Zhou N, Peng J, Xiong Z, et al. In silico ADME/T modelling for rational drug design. Q Rev Biophys. 2015;48(4):488–515. 142. Rostad K, Mannelqvist M, Halvorsen OJ, Oyan AM, Bo TH, Stordrange L, et al. ERG upregulation and related ETS transcription factors in prostate cancer. Int J Oncol. 2007;30(1):19–32.

27

Prostate Cancer Biomarkers: The Old and the New Anette L. Magnussen and Ian G. Mills

Abstract

Triggered by the urgent need for greater accuracy in predicting the biological behaviour of diseases, foremost cancer, the search for the ideal biomarker—a molecule, gene or any other characteristic that can be objectively measured and evaluated—is relentless. The search for exemplary candidates has been encouraged by the rapid progress in profiling and detection technologies which have enhanced throughput and accuracy of protein, transcript and mutation detection. These advances are reflected in the sheer number of publications reporting discoveries and clinical evaluation of new genes or proteins of biomarker potential.

SAMPLING tissue • body fluids circulating cancer• cells

clinical trials

feasibility

non-invasive cost effective

BIO MARKER IN THE PROSTATE

PROPERTIES specificity • sensitivity prostate specific psa

sampling bias

DATA proteomics • genomics ethics

psa• pten • lrg-I

gene signature

 The pictured Venn—diagram suggests, in a non-mathematical sense, the relation between the three major pre-sets that determine the potential of a biological feature to become a prognostic signpost for prostate cancer. In the centre where the circles overlap in the most favourable way, we can find the perfect biomarker.

Take-Home Lessons A. L. Magnussen (*) Nuffield Department of Surgical Sciences, University of Oxford, Oxford, UK e-mail: [email protected] I. G. Mills Nuffield Department of Surgical Sciences, University of Oxford, Oxford, UK Patrick G Johnston Centre for Cancer Research, Queen’s University of Belfast, Belfast, UK Centre for Cancer Biomarkers, University of Bergen, Bergen, Norway Department of Clinical Science, University of Bergen, Bergen, Norway e-mail: [email protected]

• Prostate cancer biomarkers can in principle have an impact in stratifying patients based on progression risk at diagnosis or to predict the risk of treatment relapse. • Very few biomarkers have progressed into routine clinical practice and kallikreins remain, so far, the prime examples. • Biomarker adoption and routine use requires independent validation and careful cost–benefit assessments. The latter should be a continuous process as detection technologies and data accumulate and this process therefore is costly in time and resources. This process is exemplified for PSA.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. A. Akslen, R. S. Watnick (eds.), Biomarkers of the Tumor Microenvironment, https://doi.org/10.1007/978-3-030-98950-7_27

467

468

A. L. Magnussen and I. G. Mills

• By contrast the discovery of candidate biomarkers occurs with much shorter timelines and the costs of discovery research fall as profiling and detection technologies evolve. • Transcript gene signatures detectable in tissue; mutation detection in circulating cell-free tumour DNA and the detection of circulation tumour cells exemplify biomarkers on the cusp of adoption in selected clinical contexts. • In considering secreted protein biomarkers we exemplify their challenges and potential by describing leucine-rich glycoprotein-1, LRG-1, and its role in prostate cancer and other cancer types. • We conclude by discussing the importance of informed patient and clinician choice in bringing biomarkers into clinical practice.

Introduction In a cancer setting, biomarkers are needed to enhance the precision of diagnosis and disease staging as well as to prognosticate diagnosed cases and monitor disease progression. A potent biomarker helps treatment selection for patients and to monitor treatment response. It may even function as a therapeutic target. At the same time, given the nearly 40,000 references for ‘cancer biomarkers’ from the early 1940s to present, the number of approved cancer biomarkers is relatively small (Fig. 27.1). The disparity is, on reflection, not very surprising as the ideal biomarker has to comply with a number of high demands such as:

–– –– –– ––

originates only from tumour tissue highly sensitive and highly specific able to detect cancer at an early stage discriminatory between indolent and clinically significant tumours –– easily detectable by non-invasive test –– inexpensive In reality no single biomarker will fulfil all of these criteria. Instead, the predictive value is determined by combining a series of biomarkers with other biological parameters and clinical risk factors.

Biomarkers for Prostate Cancer Prostate cancer is the second most commonly diagnosed cancer amongst men worldwide. Statistically, one in eight men will develop prostate cancer at some time in their lives. The rate is even higher, six in ten, for men above 65 and in men of colour [1]. At the same time the overall rate of men dying from prostate cancer in the European Union, for example, is on the decline [2]. This apparent paradox is due to the wide spectrum of clinical behaviour with men surviving long enough to die of other causes and the high frequency with which relatively indolent cases are diagnosed. Such heterogeneity of clinical behaviour poses challenges for early detection of the disease. When a man is concerned about prostate cancer and seeks medical advice, two major questions arise that frame a particularly important context in which to add complementary prostate cancer biomarkers to existing detection strategies:

18000 Cancer Biomarker

Prostate Cancer Biomarker

22500

15000

7500

1940

Numbers of references

Numbers of references

30000

Proteomics

Transcriptomics

13500

9000

4500

1960

1970

1980 Year

1990

2000

2010

Fig. 27.1  The number of references has risen continuously from the first two publications on fetuin as a potential biomarker in the early 1940s to the present. A keyword query of ‘cancer biomarkers’ yields nearly 350′000 hits. Especially in the last 15 years, a particular steep increase in publications addressing biomarkers in cancer is recorded

1960

1990

2000

2010

2020

Year

simultaneously with the advances made in proteomic and transcriptomic pattern recognition and the ability of analysing very large data sets. Despite those efforts, however, the number of reliable biomarkers for cancer remains low. [Source https://pubmed.ncbi.nlm.nih.gov February 2021]

27  Prostate Cancer Biomarkers: The Old and the New

469

Table 27.1  Table is a digest of FDA approved biomarkers for prostate cancer to be applied in combination with classic means of diagnosis Name of marker BRCA1/BRCA1

Type of marker Gene mutation

Circulating cancer cells cell search PAP

Protein detection on circulating cancer cells Protein (Prostatic acid phosphatase) Protein (Prostate specific antigen) RNA (Transcriptome) RNA (Transcriptome)

PSA Pca3 mRNA 17-Gene signature oncotype DX 46-Gene signature prolaris

RNA (Transcriptome)

Source Blood and bone marrow Blood Blood

Target Breast cancer Ovarian cancer Prostate cancer Breast cancer Prostate cancer Metastatic prostate cancer

Blood

Prostate cancer

Urine Tumour tissue

Prostate cancer Prostate cancer

Tumour tissue

Prostate cancer

Application Treatment specification for breast and ovarian cancer; Risk assessment Assist with clinical decision making Diagnosis of poorly differentiated tumours Diagnosis and monitoring of treatment response and/or recurrence Assessment if second biopsy is needed Stratification Monitor tumour progression and assist disease management

Those markers mainly serve clinicians in monitoring treatment response and disease management. Source: NCI.org April 2021 [3]

1. Should an individual be tested for prostate cancer and by what means? 2. If he is diagnosed with prostate cancer, what is the risk of potentially lethal versus clinically insignificant disease? The classification of high-risk prostate cancer is based on any one of the following: evidence of metastasis; a Gleason score of 8 to 10 representing poorly differentiated or undifferentiated (immature) cells which often grow and spread quickly; a clinical stage where the tumour is large or spread beyond the prostate; or a very high PSA level. Each of these factors indicate real potential for the cancer to develop into a fatal type and definitive treatment of highrisk prostate cancer is paramount. However, it is among patients with low- or intermediate-risk prostate cancer that the concern about choosing the most befitting treatment is greatest. All approved prostate cancer biomarkers are intended to assist with this choice as an addition to conventional diagnosis methods. They are either detected as circulating biomarkers in blood or urine or isolated from tumour tissues [Table 27.1].

Circulating Biomarkers One of the earliest discovered markers for prostate cancer is the increased blood level of prostatic acid phosphatase (PAP), an enzyme produced in the prostate gland. The clinical testing for PAP proved to be impractical, marred by the relative instability of the enzyme and the low sensitivity in early-stage disease. PAP, however, is surprisingly accurate in identifying high-risk patients and raised levels point to poorly differentiated tumours [4, 5]. In clinic PAP made way in favour for the most widely implemented diagnostic reference point: the change of prostate-specific antigen levels (PSA) in the blood.

 rostate-Specific Antigen Measurements P in the Blood Are the Standard Molecular Diagnostic and Prognostic Tools PSA Testing PSA testing to aid diagnosis has been used in clinic since the early 1980s. The approval as diagnostic maker by the FDA followed in 1994. Its diagnostic and disease monitoring properties have been decidedly beneficial for many patients, but it comes with a critical limitation: PSA is not a cancer-­ specific marker, it is an organ-specific marker. Measuring PSA levels can be deceptive as the concentrations in the blood also rise during infection/inflammation and through benign prostatic hyperplasia. If a malignancy is indeed present, the PSA level does not provide information about the stage of the disease nor its aggressiveness [6]. PSA, also known as kallikrein 3 or hK3, is a serine protease and a member of a family of glandular kallikrein-related peptidases. The genes for glandular kallikreins are clustered on chromosome 19 (Chr19q13-4) and the transcription of PSA is regulated by androgens. The function of PSA is to liquefy seminal fluid through its action on the gel-forming proteins semenogelin and fibronectin. A healthy prostate is surrounded by a continuous layer of basal cells and a basement membrane. Both act as a barrier and prevent PSA leaking from in the prostate into the circulation, hence the low normal PSA baseline. During abnormal processes accompanied by inflammation this barrier becomes weaker and blood vessels more permeable for PSA. Despite extensive research and testing it is very difficult to define the optimal threshold (or cut-off value) for PSA in a patient. Traditionally, it was set at the range between 4–10 ngml−1 providing a sensitive test, with a positive predictive value of 37% and a negative predictive value of 91%. In other words, three quarters of men with blood PSA within that range and who undergo biopsy are not diagnosed with a malignancy. Still, even with

470

a threshold as low as 4  ng/ml, the risk that a cancerous growth might exist cannot be ignored. The Prostate Cancer Prevention Trial (PCPT), for example, reported that 27% of men whose digital rectal examination (DRE) had been normal and the serum PSA value just below the 4 ng/ml mark did have prostate cancer at the time of examination. On the other hand, lowering the PSA cut-off below 4 ng/ml had no preventative effect in the long term and would only lead to more unnecessary biopsies of clinical insignificant prostate cancer.

 ifferentiation Between Benign and Malignant D Prostatic Hyperplasia by the Prostate Health Index Honing PSA testing is built on protein-complexation and isomeric variants of PSA and their ratio to each other. PSA can circulate in the blood in its native form aka ‘free PSA’ (fPSA) or in complex of PSA (cPSA) binding to α1-antichymotrypsin (ACT), α2-macroglobulin (A2M) and α1-protease inhibitor (API)). The relative amount of fPSA was found to be a decisive factor between prostate cancer and benign conditions and the percentage, PSA/fPSA  +  cPSA)  ×  100, is in clinical use as a stratification guide for patients whose PSA lies within the critical range. The lower the percentage the higher becomes the probability for a malignancy to be present. Native fPSA exists in multiple isoforms, including inactive precursor PSA (pro-PSA), benign prostatic hyperplasia associated PSA (BPSA) and intact PSA (iPSA), the un-cleaved enzymatically inactive form of PSA [7]. Several studies provided evidence that levels of pro-PSA are significantly higher in patients with prostate cancer, whereas the levels of BPSA and iPSA are decreased [8, 9]. Tests of blood samples for the ‘two amino acid’ truncated variant of pro-PSA (p2PSA) demonstratively out-perform the percentage-fPSA guide and classic PSA screening in both sensitivity and specificity. The presence of relatively elevated p2PSA is especially dominant in men with metastatic prostate [10]. By reference to this information the Prostate Health Index (PHI) is assessed. PHI combines the ratio of fPSA to p2PSA with the total PSA into a single score which is calculated by the mathematical formula :p2PSA/fPSA ×  √ total PSA. As mentioned above, patients with moderately high PSA and negative DRE results still may be at risk and the PHI was introduced by the FDA in 2012 to aid in the distinction between benign and malignant condition. The PHI is also helpful in monitoring men after interventions. The probability that cancer is present or has recurred rises proportional to the increase of the index, but impact of the PHI on the clinical decision making remained below expectations.

A. L. Magnussen and I. G. Mills

4 Kallikrein Predictive Score Model The 4Kscore® is a statistical test that has shown to accurately diagnose prostate cancer of Gleason grade >2 before biopsy and to prognose distant metastases [11]. The statistical power of 4Kscore® rests on the measurement of four prostate-­ specific kallikreins: tPSA, fPSA, iPSA and human kallikrein-­ related peptidase2 (hK2). A data-based analysis from 2015 pooled 10 independent studies into one cohort of 14,580 patients, whose levels of all four kallikreins were measured before biopsy. The results of an assessment of the discrimination potential assessed showed that iPSA and hK2 were key factors to the success of accurate prognosis of biopsies and the 4Kscore® consequentially outperformed the conventional age-related tPSA approach [12].

Circulating Tumour Cells A less commonly applied biomarker is the standardised assay CellSearch®, which captures circulating (epithelial) tumour cells (CTC) with an antibody against the epithelial adhesion molecule (anti-epCAM antibody) in the patient’s blood [13, 14]. The cells are further classified as cytokeratin+/CD45-/DAPI+ to exclude immune cells or erythrocytes. CTCs are considered a ‘liquid biopsy’ and are invaluable material to study tumour cell behaviour without taking tissues, especially after recurrence of a growth when yet another biopsy is considered unethical for a patient. The CellSearch® assay has a high prognostic power, meaning number of CTCs in the blood helps to forecast a patient’s outcome without treatment. CellSearch® seems to fare less well in its predictive power, meaning the number of CTCs is less helpful in the estimate of a likely benefit a patient would receive from treatment [15]. Extending the verification markers of the commercial CellSearch® by additional specific molecular markers that are also easy therapeutic targets enhances the predictive power of the assay [16]. For prostate cancer a suitable target is the androgen receptor, and more specifically the splice variant AR-V7, which could become a marker directing either to hormone or chemotherapy [17]. The detection of AR-V7 on CTCs is in general a sign for the presence of metastatic castration-resistant prostate cancer and the majority of those patients would not benefit from further androgen deprivation therapy. This group of prostate cancer patients are more likely to respond to chemotherapy or to the emerging category of novel small molecule inhibitors that directly bind to the intrinsically disordered N-terminus of the androgen receptors and its splice variants [18, 19].

27  Prostate Cancer Biomarkers: The Old and the New

Other Circulating Biomarkers

471

ers of BRCA2 mutations had higher incidences in developing prostate cancer. Within the carrier cohort prostate cancer was In addition to the search for protein biomarkers, germline also diagnosed at younger age and number of clinically siggenetic variants have been identified through genome-wide nificant tumours was elevated compared to the cohort of non-­ association studies. Genetic variation accounts for up to one-­ carriers [27]. The genetic screening of BRCA2 mutations third of the cumulative lifetime risk of being diagnosed with has been extended to men with suspected prostate cancer to prostate cancer and this has been based on genome-wide guide treatment specification. association studies (GWAS) focussing predominantly on common single nucleotide polymorphisms (SNPs). Usefully SNPs can be detected in germline DNA extracted from either Tissue-Based Biomarkers saliva or blood samples fulfilling the criterion of a minimally invasive test. The focus of many of these GWAS studies has Biopsy or surgical removal of the prostate gland provides been on European Caucasian cases and controls. More recent tissue samples for classic pathological evaluation, specific multi-ethnic studies suggest that distinct genetic risk land- immunohistochemistry, or used for molecular profiling. scapes underpin heritable risk aligned to differing ethnici- Grade scoring of tissue samples by proliferative markers, ties. Genetic testing can provide a cost-effective approach to cell adhesion molecules protein expressed by tumour-­ focussed diagnostic testing using PSA, multi-parametric suppressor gene are no official prostate biomarkers but are magnetic resonance imaging and other modalities. Whereas useful co-indicators. the germline genetic landscape associated with prostate cancer risk is rich enough to lend itself to the development of polygenic risk scores (PRS) [20], more work is needed to Proliferation Index: Ki67 offer risk stratification tools for people from all ethnic backgrounds [21]. As the costs of genetic testing fall further and The Ki67 protein is found in the nuclei of proliferating cells capacity increases it will be possible to identify rare genetic and can be detected by immunohistochemistry. Ki67 is particuvariants associated with poor prognosis, metastatic disease larly attractive due to its ease of interpretation of stained tissues risk rather than the lifetime risk of developing prostate can- and high reproducibility of observational results. For prostate cer [22]. Some of this work indicates that such variants will cancer samples the proportion of tumour cells staining positive associate with biological mechanisms that are known to be for Ki-67 was higher in malignant than in benign phenotype dysregulated in tumours, including DNA repair pathways and strongly related to the Gleason Scoring [28, 29]. [23]. Clinically that will provide the possibility of selecting bespoke imaging and biomarker panels aligned to these genetic features and also to actionable molecular targets and E-Cadherin improved treatment outcomes. The prostate cancer gene 3 (Pca3 mRNA) is upregulated in Cell adhesion molecules preserve the structure integrity in a all cancerous tissues of the prostate, in comparison with tissue, for example, ‘calcium dependent adhesion’ molecules benign prostatic hyperplasia, and hence a better discriminating called cadherins. E-cadherin is the epithelial cell–cell adhefactor than PSA. Pca3 mRNA can be detected in the urine after sion molecule that ensures cell polarity and epithelial integDRE (prostate massage). For patients with raised PSA levels rity. The loss of E-cadherin and/or the switch to N-Cadherin and negative biopsy the Pca3 mRNA score helps to decide signals the transition to a less differentiated and more aggreswhether a second biopsy is advisable or not [24, 25]. sive tumour cell type in prostate cancer and could therefore Prostate cancer is by nature a ‘male’ disease, but a heredi- pass as a disease progression marker [30, 31]. tary link between men and women exists: the BRCA1 and BRCA2 gene. BRCA1 and BRCA2 are co-regulators of the androgen/oestrogen receptor and a germline mutation in PTEN those genes increases the risk for breast cancer in women. Their presence is usually detected in blood and bone marrow Phosphatase and tensin homologue (PTEN) is the protein of samples. Inherited BRCA2 mutations via the female line the tumour-suppressor gene pten on chromosome 10q23. also cause an increase in risk for men to develop prostate PTEN antagonises the PI-3 K/Akt signalling pathway which cancer [26]. The results from the IMPACT study, a collabo- regulates cell cycle and cell motility and the loss of the proration between 65 centres in 20 countries, showed that carri- tein in (prostate) cancer cells leads to their increased prolif-

472

eration and cell survival [32]. The loss of PTEN is generally suggestive of adverse oncological outcome and resistance to a number of conventional therapies, which may be accountable for the high probability of recurrence in patients that show a low expression of PTEN [33]. The features of PTEN loss could be instrumental for prognosis, especially for distinction between indolent and malignant tumours, and predictive for the course of treatment. As mentioned earlier, it is unlikely that one single biomolecule will combine all mandatory criteria that make an ideal biomarker and compromises must be made. At the same time, even the most hopeful candidate must be abandoned in the face of conflicting data. Gene fusion genes are known to be driver of carcinogenesis. For prostate cancer, recurrent fusion of the 5′ untranslated region of androgen-regulated trans-membrane protease serine (TMPRSS2) and Oncogene ERG of the ETS transcription factor family was first reported in 2005 [34]. The true value of fusion gene TMPRSS2-ERG as diagnostic, prognostic and predictive biomarker, however, could not be established but the specificity to prostate cancer could make it a strong therapeutic target [35, 36]. Although the enhancer of zeste homologue 2 (EZH2) gene is upregulated in prostate and other cancer types, its credentials as a biomarker have not been shown to date [37]. EZH2 mediates trimethylation of histone H3 lysine 27 (H3K27), which leads to repression of transcription and silencing of gene expression including the E-cadherin gene, which is vital for maintaining the epithelial integrity [38, 39]. On its own EHZ2 remains a meaningful therapeutic target rather than a convincing tissue biomarker, but as part of a 14 gene panel the diagnostic and prognostic results are promising [40].

A. L. Magnussen and I. G. Mills

The 46 gene signature is an extension to the aforementioned 17 gene signature and is also based on tissue qRT-­ PCR. The 46 gene panel includes 31 cell cycle genes and 15 house-keeping genes and is also known as cell cycle progression (CCP) test. The CCP scale is an empirical score system of arbitrary units on a dynamic range normalised to the expression levels of the 15 house-keeping genes [44, 45]. In 16 independent studies the CCP score reliably paired the prediction of biochemical recurrence and disease-specific progression and mortality with conventional means of diagnosis and prognosis [46]. Both 17 and 46 gene signatures are prognostic tests and tools for an individualised risk assessment. The two molecular assays are commercially available and traded under the product name of Oncotype DX Genomic Prostate Score (17 gene signature; Genomic Health, Inc.) and Prolaris CCP (46 gene signature; Myriad Genetics, Inc.). Both are approved by the FDA for patient stratification, treatment monitoring and disease management.

 eucine-Rich Alpha2 Glycoprotein: L A Biomarker in Many Contexts

Biopsies and tissue samples will never totally be avoidable and a tissue section provides invaluable information about stage, grade and general pathology of a tumour. Paraffin embedded tissue samples are well preserved and easy to archive. However, collecting tissue samples comes at a price. Prostate cancer, for example, is an inter and intra-­heterogeneic tumour, a feature that adds to the difficulty of biopsy stratification. A needle biopsy can, involuntarily, introduce sampling bias by simply not piercing the ‘right part’. Prostate Approved Tissue Markers for Prostate Cancer cancer is also multifocal: several small islands of cancerous tissue appear within the organ and might be missed by needle The 17 gene signature is a qRT-PCR assay to measure the biopsy. If a patient is under active surveillance and moniexpression levels of 12 cancer related genes and five house-­ tored for treatment response and/or recurrence, it is ethically keeping genes (control) in needle biopsy tissue samples. A untenable to submit them repeatedly to a biopsy. One top total of three clinical studies were conducted to validate the breakthrough in cancer detection is the ‘liquid biopsy’: a cirgene panel. The first two studies consisted of a discovery culating biomarker in blood or other bodily fluids, indepencohort of 441 samples and another cohort of 167 samples dent of tumour location. Samples can be easily and frequently from patients that were at intermediate to low risk but subject collected by minimal invasive or even non-invasive proceto active surveillance. From the genetic data of those two dures. As discussed in the previous chapter, the standard biostudies a multigene-based signature, the genomic prostate marker for prostate cancer still is PSA which has the great score (GPS) on a scale from 0–100 was developed. A third benefit of being detectable in the blood but falls short in study was designed to validate the GPS. A logistic regression other ways. of 395 samples tested the GPS in correlation to pathological Over the last decades, elevated levels of leucine-rich stage and grade at the time of prostatectomy. The results alpha2 glycoprotein (LRG-1) were detected in multiple showed that the GPS accurately discriminated the aggres- studies scrutinising the transcriptome and proteome of siveness of prostate cancer, despite tumour heterogeneity cancer cells and in tissues of other proliferative diseases and multifocal occurrences and may aid in clinic to separate [47–51] (Fig.  27.2). In cancer elevated LRG-1 levels are patients for active surveillance from those that need immedi- present in tumour tissues but appealingly, also in blood ate treatment [41–43]. and other bodily fluids of the patients. The consistency of

27  Prostate Cancer Biomarkers: The Old and the New Fig. 27.2  Graphic that summarises the clinical studies conducted to explore LRG-1 as a marker for tumour progression, prognosis and risk stratification. Figure designed with BioRender

473

Glioblastoma: High LRG- I levels associated with tumour location and with longer overall survival for those patients − an exception for high LRG- I as indicator for favourbable outcome Hepatocellular carcinoma: LRG- I indicative for HCC occurence after hepatitis B infection. High LRG-I is associated with metastatis and shorter overall survival and with unfavourable outcome.

Ovarian and prostate cancer: LRG-I levels in blood sera correlate to tumour progression stage and tumour cell phenotype. High LRG- I was associated with shorter disease free survival and unfavourable outcome.

finding LRG-1 enrichment as a sign for abnormal processes makes a strong case for LRG-1 becoming a powerful and versatile biomarker facilitating diagnosis and prognosis in the future. LRG-1 is found in a number of normal cell types in a number of organs: in the alveolar cells of the lung, duct cells of the pancreas and the glandular cells of the prostate but it is particularly enriched in the hepatocytes of the liver. In the eye, heart muscle, skin, placenta and prostate, LRG-1 is also enriched in endothelial cells of the blood vessels and it is found in neutrophils and monocytes of the blood. LRG-1 is packaged into the granules of the neutrophiles and released upon activation [52, 53]. In cultured hepatocytes LRG-1 is enclosed in vesicles in the cytoplasm and believed to be excreted [54]. LRG-1 can therefore be detected in blood serum and to some extent in other bodily fluids such as urine.

 RG-1 a Diagnostic Biomarker for Cancers That L Are Hard-to-Detect Early detection is a problem for any cancer; however, some cancers are harder to discover than others. Liver, pancreatic

Breast cancer: High LRG-I associated with increased lymphatic metastasis and shorter overall survival for those women. Rising levels of LRG-I correlate with high tumour progression stage and unfavourable outcome

Pancreatic and colon cancer: High LRG- I is associated increased metastasis and increased neo-vessel density in colon cancer and correlates to unfavourable outcome

and kidney cancer diagnosis is made more difficult by the location of the organs that are either sheltered by the ribcage or hidden deep in the torso. Ovarian cancer is notoriously overlooked and early symptoms are mundane and attributed to ‘indigestion’ rather than to a deadly growth [55]. The same is true for brain tumours: protected by the skull tumours in the brain are not palpable and headaches are common without a serious condition and slight changes in personality can be explained by external factors [56]. By the time the symptoms are so severe to raise concern, it is often too late.

Hepatocellular Carcinoma Unsurprisingly, upregulation of LRG-1 was first detected on hepatocellular carcinoma (HCC) tissue, as the cells in the liver are naturally expressing the protein. One of the earliest publications reported a mass spectroscopy analysis of glycol-­ proteins isolated by lectin-based chromatography from the cancerous tissue, which could otherwise be masked when screening the proteome in whole. In the sub-selection, LRG-1 was one of the most prominent proteins in the tumour tissue [57]. A later study of sera from patients with HCC

474

A. L. Magnussen and I. G. Mills

came to similar results. In this study the aim had been to see if the distribution of bio markers in the serum, including LRG-1, could be used to identify the origin of the cancer, here after infection with Hepatis B virus (HBV). The result was that patients whose liver tumour correlated with earlier HBV infection showed significant higher levels of LRG-1 in the serum samples, than the patients that carried no virus [58]. Those results were picked up and expanded with a large cohort of 777 HCC patient samples (both tumour and adjacent non-tumour liver samples). In 51.4% of the 777 samples LRG-1 (based on mRNA and protein) levels were significantly raised and correlated with the pathological data of consistently larger late-stage tumours, poor tumour differentiation and higher rate of vascular invasion. For the individuals in that particular group the overall survival rate and disease-free time period were shortened, the recurrence probability and the tendency to develop metastases were increased (Fig. 27.3). The statistical difference was significant for all criteria and the diagnostic power of the screening found to be as high as 95% [59]. Almost simultaneously a second, very similar study was published which came to the exact opposite results, reporting that LRG-1 was downregulated in HCC tissue samples. Alas, no clinical patient data had been available. The study also showed that LRG-1 had no effect on HCC cell proliferation but could inhibit HCC cell migration and invasion concluding LRG-1 as a potential anti-metastatic factor [60].

were identified as glioblastoma, 27 as astrocytoma and 14 as oligodendroglioma was analysed for LRG-1 [61]. All samples showed increase of LRG-1 levels compared to control tissues from normal brain cortex samples. The assessment of the samples followed a scoring system from 0 (negative) to 4 (positive liver control). Within the glioma group the level of LRG-1 expression varied. The occurrences of high LRG-1 in the non-glioblastoma samples were significantly fewer than in the major group of glioblastomas. In the astrocytoma and oligodendroglioma subgroups only 3 (21%) and 1 (4%) case, respectively, had high levels of LRG-1. In the major group of glioblastomas, 47 (41%) cases showed high LRG-1 levels. As for clinical outcome: after surgery, patients that had tumours with high levels of LRG-1 had an advantage in their ‘overall survival’ but not in ‘progression free survival’ over patients whose samples showed lower LRG-1 levels (Fig.  27.3). Case samples for this study were matched in their ‘extension of resection’ (EOR) index. Yet, the tumours that showed high levels of LRG-levels had been located further away from the subventricular zone (SVZ) and were predominantly removed from the periphery of the brain, those with lower LRG-levels were closer to the SVZ.

Ovarian Cancer LRG-1 was found to be a novel protein marker after deep depletion of abundant proteins from ovarian cancer tissue samples and blood serum of women with ovarian cancer [62, 63]. In a cohort of 114 women, 58 diagnosed with ovarian cancer and 56 healthy women, LRG-1 concentrations were indicative to the stage the cancer had been at the time of sampling. In blood serum (and tissue samples) the mean LRG-1 level was consistently raised and significantly (2-fold) higher

Glioblastoma LRG-1 was found in patients diagnosed with glioma. A clinical study that had been conducted and results published in 2020. A cohort of 155 glioma tissue samples, in which 114

liver (n=1) pancreas (n=1) brain (n=1) prostate (n=1) 50% OS years

10

8

6

4

2

unfavourable high LRG-1

Fig. 27.3  Graph that shows qualitatively LRG-1 levels relating to overall survival. The bar-graph matches LRG-1 levels to the patient’s clinical outcome and is based on 50% death rate in the highest risks groups in each study. All patients had received maximum treatment depending on their diagnosis and circumstances. The term ‘high’ and ‘low’ LRG-1 in this graphic is a general term to describe the empirical threshold concentrations that vary for each study. Other variables are

0

2

6

4

8

10

favourable low LRG-1

the source of LRG-1 (circulating protein in blood vs tissue samples) and the methods with which LRG-1 was detected in the clinical samples. Liver: IHC and scoring system of patient samples [59]; Pancreas: LRG-1 plasma levels by ELISA [90]; Brain: IHC and scoring system of patient samples [61]; Prostate: LRG-1 plasma levels by ELISA [105]. n = number of clinical studies. Figure designed in BioRender

27  Prostate Cancer Biomarkers: The Old and the New

in cancer patients than in the healthy cohort. The highest concentrations were measured in women diagnosed with late-stage (III/IV) ovarian cancer. Compelling data, however, was obtained from pre-operative blood samples taken from a separate cohort of 193 women that underwent surgery for suspicious adnexal growth. In this study the mean LRG-­ levels in those patients that actually had a malignant growth (serous and clear ovarian cancer) were 1.7 fold higher than in the women whose tissue pathology report returned as ‘benign’, manifesting the hypothesis that LRG-1 could be a distinction marker between malignant and benign phenotypes [64, 65]. In a later (but much smaller study) the concentrations of LRG-1 in the urine of women diagnosed with early-stage ovarian were compared to an age-matched cohort of heathy volunteers. The women with early-stage ovarian cancer had a mean LRG-1 content in their urine that was significantly higher than in the control group [66, 67]. This result is insofar worth noting as many women with early-­ stage ovarian cancer show normal levels of cancer antigen 125 (CA125), one of the FDA approved marker of ovarian cancer to assist diagnoses and to monitor treatment response.

Breast Cancer High LRG-1 in tissues of breast cancer linked the patients to the malignant progression of the disease. In 2020, a tissue study of 330 breast cancer samples showed that the elevated LRG-1 levels went along with climbing numbers of lymph metastases and raised scores in the TNM pathological staging system. Those women went through shorter disease-free survival and shorter overall survival than the women whose tissue samples showed significantly lower LRG-1 concentrations [68].

Other Harder to Diagnose Cancer Types LRG-1 upregulation has also been observed in plasma samples from patients with pancreatic cancer and in patients with colon cancer [69, 70]. A different study on colon cancer patients made the case that not only the raised levels of LRG-1  in the tumour tissue are indicative for disease progression but also showed that the micro-vessel density in those tissues had been greater. Greater micro-vessel density is a nod to the angiogenic properties of LRG-1 and suggestive for increased risk that tumour cell invades the blood vessels [71]. LRG-1 protein was also upregulated in tissue samples, glycoprotein-enriched serum samples and urine samples from lung cancer patients. For this study no correlation between LRG-1 levels in those samples and the clinical outcome of the patients was drawn [72–74]. A study on patients diagnosed with non-small cell lung cancer (NSCLC) however, showed that high levels of LRG-1 in biopsies were

475

associated with the tendency for metastasis and shorter overall survival [75]. Similar results derived from clinical studies on patients with thyroid cancer [76].

The History of LRG-1 Protein In 1977, an unknown human serum protein was isolated from the by-product of the large-scale preparation of albumin and g-globulin: the 347 amino acids long leucine-rich alpha2 glycoprotein (LRG-1). The amino acid content is unusual in that the leucine content is almost 17%, which means that one fifth of the amino acid content is a leucine [77]. Crystal structure analysis shows the protein as a single polypeptide chain with one galactosamine and four glucosamine attached. Of the 312 amino acids, 66 are leucine. The peptide can be divided into 13 segments with 24 residues each. Eight residues present a periodic pattern of leucine, proline and asparagine. The protein contains domains capable of bipolar surface orientation and shows homology to segments of mitochondrial proteins, viral envelope proteins and oncogene proteins [78, 79]. The 3D protein model, based on NetrinG2 template, reveals a coiling amino acid ß-chain that forms a barrel or tube-like tertiary structure and features a short amino acid α-helix towards the C-terminus. The structure is stabilised by four cysteines that form di-sulphide bonds, one bond at either end [80, 81].

 RG-1: The Transforming Growth Factor L Connection For the longest time LRG-1 was classified as ‘a serum protein with unknown function’ and its true biological role is still not very well understood [77, 82]. First hints about role of LRG-1  in the cell cycle came from a report in 1995. A study of differentially expressed genes in human hepatoma cells was laid out to search for novel factors involved in the transforming growth factor b (TGF-ß) signalling pathway and identified mRNA for a gene, SB31, that was 100% homologous with LRG-1. This gene was co-ordinately expressed with mRNA of the TGF-ßR2 gene, which suggested an interaction between the both [83].

TGF-ß Signalling in a Nutshell What makes the function of LRG-1 hard to discern in its relation to TGF-ß, is the fact that TGF-ß itself has several roles with often opposite consequences in the grand scheme of cell signalling. In order to interpret the effect LRG-1 has on TGF-ß signalling we need to understand the role of TGF-ß first. TGF-ß is a multifunctional growth factor of, and name-­

476

giver to, the transforming growth factor superfamily (which in fact comprises a large number of proteins). Their downstream signalling pathways regulate the cell cycle, and can either suppress or stimulate cell proliferation and differentiation. Although downstream effects of TGF-ß pathways are heavily dependent on cellular context, the downstream signalling is, at least partially, conserved in many cell types. There are three known and highly homologues isomers of TGF-ß, numbered 1–3. All three act via the same receptors, which are single strand serine/threonine kinase receptors R1, R2 and R3. R3 possesses no kinase activity and is not directly involved in the TGF-ß signal transduction pathway but works as a reservoir for the growth factor. TGF-ß is a ligand to R2 and the TGF-ß/R2 complex recruits and phosphorylates R1. Depending on R1/R2 phosphorylation pattern a canonical (also called ‘SMAD’) signalling pathway or a non-canonical signalling pathway is triggered. In the SMAD signalling pathway, activated R1  in return phosphorylates receptor-­ regulated transcription factors SMAD (R-SMAD), that is then able to bind to coSMAD. The R-SMAD/coSMAD complex translocates to the nucleus and modulates gene expression depending on the cellular context. In the non-canonical signalling pathway TGF-ß–R1/R2 complex carries an additional phosphate and activates several other signalling pathways including Rho/Rock that regulates the tight junctions, MAPK and Erk pathways, both transcriptional regulators, and P13K/Akt pathway that controls protein synthesis. Both TGF-ß signalling pathways contribute to the regulation of proliferation, differentiation, apoptosis and epithelial–mesenchymal transition—key events in cell biology [84]. Upon secretion by various cell types, including macrophages, TGF-ß is present in its latent form bound mainly by two TGF-ß binding proteins (LTBP and LAP). Proteinases, for example, plasmin, can cleave the complex binding and release activated TGF-ß into the extracellular space. Other TGF-ß activating factors are integrins, reactive oxygen species (ROS) and thrombospondin. On the surface of macrophages, for example, the latent TGF-ß complex is bound to CD36 via thrombospondin-1 (TSB-1). In response to inflammatory stimuli plasmin levels rise and the release of activated TGF-ß1 is turned on, hence the regulatory role of TGF-ß in inflammation [85, 86].

TGF-ß: Cancers Double Agent In normal cells TGF-ß signalling arrests the cells in G1 phase of the cell cycle. In the early stages of tumour growth TGF-ß indeed exerts an anti-proliferative influence on the cells but in the later stages of tumour progression this effect wears off. There it is replaced by loss of control over proliferation and

A. L. Magnussen and I. G. Mills

apoptotic response while sustaining angiogenesis. The dual effect could be explained by significant downregulation of receptor expression on cancer cells as the disease moves ahead and with mutations of the growth and other downstream factors involved in cell cycle regulation [87, 88]. At the same time, many types of tumour cells and cells of the tumour microenvironment start to overexpress the growth factor, which skews the ligand/receptor balance. TGF-ß is also a potent immunosuppressor adding to the formidable ability of the cancer cells to evade attack by the immune system of their host [89].

LRG-1 and TGF-ß: Brothers in Arms LRG-1 and TGF-ß and TGF-ß receptor enrichment is often simultaneously observed in human tissues and fluids of various proliferate diseases. It was, for example, discovered in patients with idiopathic normal pressure hydrocephalus (INPH), a neuro-degenerative disease with excess cerebrospinal fluid (CSF) build-up in the brain ventricles. Analysis of the INPH-CFS showed a markedly higher concentration of LRG-1 and TGF-ß1 and TGF-ßR2 than in the CFS of healthy controls [49]. How the upregulation of two but not the third mandatory factor, TGF-ßR1, in the conserved three-­ factor activation pathway effects TGF-ß1 downstream signalling was not clear. In cultured NSCLC, LRG-1 affects TGF-ß1 dependent cell proliferation, migration and invasion and promotes TGF-ß1 dependent angiogenesis. A549 (lung cancer) cells were manipulated to either lose or to overexpress LRG-1. In the knockout variant proliferation and cell motility were significantly reduced and enhanced in LRG-1 overexpressing cells. LRG-1 rich exosomes isolated [75]. NSCLC patient blood samples and cells lines encouraged an angiogenic response in HUVECs and the effect could be negated by selective TGF-ß1 receptor blocker SM-16. The analysis of HUVECs for TGF-ß1 expression (on mRNA and protein level) before and after exosome exposure on mRNA revealed a significant increase of the growth factor, which the authors concluded, was due to the presence of LRG-1. In LRG-1 expressing clones of human pancreatic ductal adenocarcinoma (PDAC) cell line Panc1/LRG, TGF-ß1 induced transition from epithelial phenotype to mesenchymal phenotype took place and was observed in the change of the shape in the majority of cells [90]. Those cells showed a drop of phosphorylated Smad2, part of the Smad-transcription factor complex that control expression of cell–cell adhesions molecules such as E-cadherin, occludins and claudins (tight junction molecules) and tissue specific cytokeratins. The loss of E-cadherin in general and here in Panc1/LRG clones, is a distinctive (but not exclusive) sign that EMT commences.

27  Prostate Cancer Biomarkers: The Old and the New

LRG-1 Regulates Angiogenesis In 2013 a study was published, presenting data that linked LRG-1 upregulation tightly to TGF-ß induced angiogenesis in the mouse retina and for the first time reported a distinct function of LRG-1 [91]. Prominent Lrg1 gene upregulation in the transcriptome of three mouse models simulating retinal vessel architecture and retinal degeneration was observed in the first instance. LRG-1 protein in all three models was almost exclusively found in the blood vessels of the retina and to some lesser extent the choroid, which is the vascularised layer of tissue that separates the retina from the sclera. Upregulation of Lrg1 gene in the retina was also observed in angiogenic phase in mouse models of oxygen induced retinopathy (OIR) and choroidal neovascularisation (CNV) days after laser injury. The retinas of LRG-1 knockout mice are only slightly less developed than those of their wild-type littermates, but LRG1−/− did not display active angiogenesis in the OIR model or CNV after injury. Cultured endothelial cells (HUVEC) were responsive both after LRG-1 activation and LRG-1 inhibitions in angiogenesis assays. Immunoprecipitation of LRG-1 with TGF-ßR2 and endoglin (auxiliary receptor to TGF-ßR2 on endothelial cells), this time in primary mouse brain endothelial cells, suggested the direct interaction of LRG-1 with those receptors. Endoglin is not a direct receptor for the growth factor, but the interaction with LRG-1 promotes TGF-ß receptor configuration and that flips the angiogenic switch via the canonical signalling pathway. LRG-1 also promotes angiogenesis in normal wound healing. LRG-1 is a necessary angiogenic factor in the proliferative stage of the healing process, when fibroblasts are recruited to form the primary granulation tissue. The healing process is significantly delayed in LRG-1−/− mice. At the same time, it does not promote wound healing in diabetic mice, although the sites of wounding were infused with LRG-1 [50]. In kidneys of diabetic mice, LRG-1 is predominantly localised to the capillaries in glomeruli and contributed to adverse events such as glomerular angiogenesis, diabetic glomerulopathy and podocyte loss [92, 93]. Although the supportive role of LRG-1 in angiogenesis is amply cited, little is known about the role LRG-1 actually plays in tumour angiogenesis. A niche in vitro study that makes a link to angiogenic, long noncoding RNA, specifically taurine-upregulated 1 (TUC1) by way of LRG-1 expressing ovarian cancer cells, was published in 2019 [94]. In this study the authors report the various angiogenic effects of conditioned media taken from TUC1 knockdown ovarian cancer cells. They established a negative correlation between low LRG-1 and angiogenic factors (reduction in VEGF and Ang1 but not in TGF-ß) in the supernatant of TUC knockdown cells and the limited activation of HUVECs, observed by absence of typical vascular structure elements and immobilisation of endothelial cells assessed in classic angiogenesis and migration assays. HUVEC activa-

477

tion could be rescued by the addition of recombinant LRG-1 to the experimental mix, hence the conclusion that LRG-1 coming from the cancer cells regulates the angiogenic response of HUVECs.

The Cytochrome c Connection Alerted by the diminished sensitivity to cytochrome c (Cyt c) in ELISA assays in the presence of serum, a mass spectroscopy study was conducted in 2006 to find that in both human and bovine foetal serum the inhibitory factor was LRG-1 [95].

Cytochrome c in Apoptosis In response to stress (or developmental signals) the intrinsic, or mitochondrial, pathway of apoptosis can be triggered from within the cells. Cty c is a protein that normally resides in the mitochondria where it is part of the mitochondrial electron-transport chain. Upon release into the cytosol the role of Cyt c changes and it binds to an adaptor protein called apoptosis-protease activation factor (Apaf1). The binding causes Apaf1 to oligomerise into a hexamer called the apotosome. Apotosomes recruit initiator caspase-9 protein that sets off the caspase cascade leading to apoptosis [96, 97].

LRG-1 Is a Competitive Ligand to Cyt c The region on LRG-1 and Apaf1 that binds to Cyt c is very similar in the amino acid sequence and both proteins home to the same binding region on Cyt c. In controlled cytotoxicity experiments however, the apoptotic effect of Cyt c on lymphocytes was blocked by LRG-1, suggesting that despite LRG-1 being in competition with Apaf1 as a ligand to Cyt c, only Apaf1 is able to activate the caspase pathway of apoptosis [98]. Blocking the mitochondrial pathway of apoptosis could have critical consequences on the natural suppression of cancer cell proliferation. A study that investigated this hypothesis was published in 2021 [99]. A three-way cytotoxicity experiment was designed, taking breast cancer cells MCF-7 and additionally transfecting them: a) with the lrg1 gene to generate an LRG-1 overexpressing genotype and b) transfecting them with interfering hairpin RNA to generate an LRG-1 knock down genotype. In the parental MCF-7 cells LRG-1 was confined to the cytosol and was not secreted in detectable amounts. Apoptosis assays revealed a significant resistance of the transfected cells to enter the apoptotic cycle. The LRG-1 seems to instantly trap Cyt c that was released from the mitochondria before Apaf1 binding could occur. The affinity with which LRG-1 bound to Cyt c is extremely high, which, on the other hand, makes Cyt c an extremely sensitive substate for LRG-1 detection [100].

478

 RG-1: Prognosticator Prostate Cancer L Progression In 2007 and 2008, LRG-1 was identified in the proteome of PC3, LNCaP and 22Rv1 prostate cancer cell lines (supernatant and cell membrane), but at the time the researchers had focussed on different proteins and the potential of LRG-1 as a biomarker in patient samples had not been investigated [101, 102]. LRG-1 makes several appearances in supplemental data in publications analysing human serum proteins and mouse pten knock out prostate cancer models [103, 104]. It was in a comprehensive study published in 2020 where LRG-1 first took centre stage in its role as biomarker and with convincing results. Data had been generated suggesting LRG-1 measurements could be a new standard for improved risk stratification for patients with prostate cancer [105]. For this project, archival blood samples spanning three decades were analysed and classified into four independent cohorts of prostate cancer patients: discovery cohort, confirmation cohort (with follow-up data ≥10  years), and exploration cohort. The final validation cohort covered 451 blood samples. The study confirmed that LRG-1 levels were consistently elevated in cancer patients over those in healthy control volunteers (discovery cohort). The variations within cancer patients correlated with the progressions of the tumours, being highest in the group of patients that were diagnosed with metastases and those whose outcome was fatal, approximately two-fold increase in serum LRG-1 over the second highest group, that of indolent prostate cancer. The LRG-1 serum concentration in the group of men that were cancer free at the time of blood draw, but developed the disease later on however, were baseline (confirmation cohort), meaning that there was no evidence that serum LRG-1 could predict a pre-cancer status. A clear relation between rising LRG-1 concentration in the blood serum and increasing risk factor in those patients to emerge with metastasis and unfavourable outcome was found (Fig. 27.3). Few deaths were registered in the group of men with low LRG-1 levels, whereas men with high LRG-1 levels were more likely subject to disease progression, metastasis and ultimately, death (exploration cohort).

 he Challenges in Translating Novel T Biomarkers into Clinical Practice: An Epilogue  he Natural Function of LRG-1 Remains T a Mystery Good efforts—and progress—have been made to shed light on the function of LRG-1 in cell biology. In all the published clinical studies, LRG-1 levels are pronounced in tumours (of

A. L. Magnussen and I. G. Mills

various sources), in blood sera and sometimes also urine of cancer patients or patients with other proliferative diseases. Apart from very few exceptions, high-rise LRG-1 seems associated to an unfavourable outcome. Biologically, LRG-1 interferes with the TGF-ß signalling cascade and the effect is noticeable in both epithelial and endothelial cells. LRG-1 actively causes an angiogenic switch via the canonical TGF-ß signalling pathway and furthers endothelial cell activation in vitro and in vivo (mouse). Even so, many questions are waiting for an answer. Why is LRG-1 elevated in the first place? Do baseline levels of LRG-1 in healthy tissues in any way impact the upregulation of the protein under pathological conditions? Does LRG-1 inhibit TGF-ß anti-proliferative or support TGF-ß proliferative function in tumour cells? How important is LRG-1  in tumour angiogenesis? In the healthy prostate, for example, LRG-1 is expressed in basal glandular cells and urothelial cells but is highest in the endothelium. Would that mean, LRG-1 effect on tumour angiogenesis would be dominant and would prostate cancer tissue with high LRG-1 also be more vascularised? Does LRG-1 wield an inhibitory effect on mitochondrial programmed cell death in cancer [106]? Ideally, the biological function of LRG-1 should be completely understood to maximise the extent for LRG-1 as a clinical marker. Regardless of the perforated knowledge about LRG-1, sound circumstantial evidence that LRG-1 is a high-performing biomarker might be sufficient for clinicians, in unison with their patients, to make informed decisions in treatment. Unravelling the biological function of LRG-1 in all its particulars may not be strictly necessary. Testing for LRG-1 in blood or even urine samples is comparably simple and could work together with other markers, such as PSA. In the meantime, PSA stays as the most widely used biomarker for prostate cancer diagnosis, prognosis and prediction. Much research and efforts went and still go into the refinement of PSA testing. A protein produced only in the prostate exerts a strong attraction and the high specificity reconciles scientists and clinicians alike with the lack in sensitivity. Although various biomarkers for prostate cancer have been implemented in discrete clinical settings, none of them rivals the convenience of PSA. For progressive scientists and clinicians, the search for novel candidates continues and funding for clinical studies to do so has surged in the last decade, not the least due to the enormous societal impact a success promises. Alas, poor reproducibility taints the results from many of those studies. The more we ask of a biomarker, for example, as a prognosticator when the gap between diagnosis and outcome can be 10 years or more, the more biomarkers succumb to variables that may still validate their biological significance but may invalidate their clinical translation. We have moved into the era of precision technology and image-guided biopsy and as a result a much higher proportion of tumour biopsies taken presently are of intermedi-

27  Prostate Cancer Biomarkers: The Old and the New

ate to advanced stage (according to the Gleason score) than in the past. In other words, a substantial proportion of archival tissue samples are non-malignant. This sampling bias creates difficulties for direct performance comparison between tissue biomarkers assessed in samples from one era versus the other and some biomarkers will need re-­evaluation. Vital data about progression free survival, overall survival and quality of life are often missing from the documentation collected the past and the lesson has been taken no longer to neglect those aspects when building a tissue archive and planning future studies. Fundamentally the question of whether medical testing will improve patient outcome endures. Desirable and beneficial advances in screening and prognosis are overshadowed by the inherently ethical dilemma between the right of patients to full and accurate information about their own medical condition and the psychological impact this information, above all in the absence of symptoms, may have on the individual. This challenge is most acute when considering the use of germline genetics but applies for many other types of biomarker. Applications of testing and treatments must still be aligned to the needs of a patient and not to indulge scientific curiosity nor for corporate financial gain. It is hard to establish whether a particular test/biomarker actually influences the clinical decision making, since planning a route of treatment does not follow an algorithm but is influenced by many factors. Nor can the economic benefit be assessed objectively. Statistically, any healthcare system might save money otherwise spent on over-treatment but in reality, the human factor might offset this benefit. In 2017, the independent organisation Health Quality Ontario, Canada published a systematic review that addresses the questions of cost efficiency of genomic profiling of prostate cancer and whether such an approach should be funded under the existing health care plan. Based on a possible offer of a Prolaris CCP test evidence for clinical and economic benefit for patients with low-and intermediate-risk localised prostate cancer was sought [107]. Although the analysis was admittedly flawed by gaps in crucial data, the investigation found no evidence that additional Prolaris CCP testing has an impact on the patient-important outcome. Neither would a routine CCP test pay-off in the short term, rather it would load significant costs onto the local health care system. In interviews with prostate cancer patients, it transpired that many of the men were initially overwhelmed with the diagnosis and most of them put great value on the recommendations from their physicians. For all patients the emotional impact of the diagnosis took hold over their decision making and the fear of metastasis weighed heavily on most patients’ minds. The anxiety that the cancer may spread beyond the prostate led some men decide against active surveillance and opt for more drastic interventions. Some men had doubts that a CCP test result, which they feared may only further com-

479

plicate the decision-making process, had let them change the course of treatment, others would not exclude a change of mind if the CCP results were unambiguous. An evaluation of the Prolaris® CCP test from 2016 by the National Institute for Health and Care Excellence (NICE), UK came to similar results as to cost efficiency, although their conclusions were based on the questionnaires given to clinicians only and did not involve actual patients [108]. More recently in 2019, a clinical trial by the Leeds Centre for Personalised Medicine and Health was launched to collect data if a precise genetic profiling (Prolaris®) would help to avoid unnecessary surgery and radiotherapy in men that developed prostate cancer [109]. The design of the study, which is in partnership with Myriad, has placed an emphasis on how the test had influenced the participants in their decision making about treatment and their quality of life. The trial results are expected in 2022. Routinely consulting novel and intricate biomarkers, and complex tests such as gene profiling, are in its infancy, and the temptation to stick to time-proven methods like PSA testing and Gleason scoring is strong. In a matter of time, however, what is new and intimidating will become familiar and chances are that novel or complex biomarkers will be a habitual part of any anti-cancer regime in the near future.

Concluding Remarks/Summary With technological advances it has become possible to sensitively profile tissue, blood and urine samples from prostate cancer patients to identify candidate biomarkers reflecting metabolic, protein, RNA, DNA and epigenetic changes. Aligning those measurements, often arising from diagnostic or pre-diagnostic samples, to the ultimate outcome of the disease for a given patient remains a huge challenge. This is because a significant time-lag may exist between diagnosis and progression, but also because the disease is highly heterogenous and a given patient may present with an array of histopathologically, molecularly and spatially different alterations in the prostate gland. Furthermore, some informative biomarkers may reflect changes in the tumour micro-environment in response to these alterations. In this chapter, we have chosen to highlight some of this complexity for one biomarker, LRG1. The next phase in the translation of this and many other markers will need to focus on understanding their spatial expression and biological function, their relationship to imaging modalities now being incorporated into diagnostic pathways and their contribution alongside other markers in predicting treatment responses and outcome. To achieve this, new clinical and research environments will need to emerge to enhance data sharing, technology adoption and the development and application of data type-agnostic analytical workflows.

480

References

A. L. Magnussen and I. G. Mills

23. Darst BF, et al. Germline sequencing DNA repair genes in 5,545 men with aggressive and non-aggressive prostate cancer. J Natl Cancer Inst. 2020; 1. Rawla P.  Epidemiology of prostate cancer. World J Oncol. 24. Li M, et  al. Urine PCA3 mRNA level in diagnostic of prostate 2019;10(2):63–89. cancer. J Cancer Res Ther. 2018;14(4):864–6. 2. Carioli G, et  al. European cancer mortality predictions for 25. Shappell SB, et al. PCA3 urine mRNA testing for prostate carcithe year 2020 with a focus on prostate cancer. Ann Oncol. noma: patterns of use by community urologists and assay perfor2020;31(5):650–8. mance in reference laboratory setting. Urology. 2009;73(2):363–8. 3. https://www.cancer.gov/about-­c ancer/diagnosis-­s taging/ 26. López-Abente G, Mispireta S, Pollán M. Breast and prostate candiagnosis/tumor-­markers-­list. cer: an analysis of common epidemiological features in mortality 4. Taira A, et al. Reviving the acid phosphatase test for prostate cantrends in Spain. BMC Cancer. 2014;14(1):874. cer. Oncology (Williston Park). 2007;21(8):1003–10. 27. Page EC, et al. Interim results from the IMPACT study: evidence 5. Xu H, et  al. Prostatic acid phosphatase (PAP) predicts prostate for prostate-specific antigen screening in BRCA2 mutation carricancer Progress in a population-based study: the renewal of PAP? ers. Eur Urol. 2019;76(6):831–42. Dis Markers. 2019;2019:7090545. 28. Tretiakova MS, et al. Prognostic value of Ki67 in localized pros 6. Farshchi F, Hasanzadeh M.  Nanomaterial based aptasensing of tate carcinoma: a multi-institutional study of >1000 prostatectoprostate specific antigen (PSA): recent progress and challenges in mies. Prostate Cancer Prostatic Dis. 2016;19(3):264–70. efficient diagnosis of prostate cancer using biomedicine. Biomed 29. Verma R, et al. Significance of p53 and ki-67 expression in prosPharmacother. 2020;132:110878. tate cancer. Urol Ann. 2015;7(4):488–93. 7. Slawin KM, Shariat S, Canto E. BPSA: a novel serum marker for 30. Olson A, et al. The comprehensive role of E-cadherin in maintainbenign prostatic hyperplasia. Rev Urol. 2005;7(Suppl 8):S52–6. ing prostatic epithelial integrity during oncogenic transformation 8. Mikolajczyk SD, Rittenhouse HG. Pro PSA: a more cancer speand tumor progression. PLoS Genet. 2019;15(10):e1008451. cific form of prostate specific antigen for the early detection of 31. Tomita K, et al. Cadherin switching in human prostate cancer proprostate cancer. Keio J Med. 2003;52(2):86–91. gression. Cancer Res. 2000;60(13):3650. 9. Mikolajczyk SD, et  al. A truncated precursor form of prostate-­ 32. Jamaspishvili T, et al. Clinical implications of PTEN loss in prosspecific antigen is a more specific serum marker of prostate cantate cancer. Nat Rev Urol. 2018;15(4):222–34. cer. Cancer Res. 2001;61(18):6958–63. 33. Geybels MS, et al. PTEN loss is associated with prostate cancer 10. Sottile A, et  al. A pilot study evaluating serum pro-prostate-­ recurrence and alterations in tumor DNA methylation profiles. specific antigen in patients with rising PSA following radical Oncotarget. 2017;8(48):84338–48. prostatectomy. Oncol Lett. 2012;3(4):819–24. 34. Petrovics G, et  al. Frequent overexpression of ETS-related 11. Punnen S, Pavan N, Parekh DJ. Finding the wolf in sheep’s clothgene-1 (ERG1) in prostate cancer transcriptome. Oncogene. ing: the 4Kscore is a novel blood test that can accurately iden2005;24:3847–52. tify the risk of aggressive prostate cancer. Rev Urol. 2015;17(1): 35. Gasi Tandefelt D, et  al. ETS fusion genes in prostate cancer. 3–13. Endocr Relat Cancer. 2014;21(3):R143–52. 12. Vickers A, et  al. Value of intact prostate specific antigen and 36. Kumar-Sinha C, Tomlins SA, Chinnaiyan AM.  Recurrent gene human Kallikrein 2 in the 4 Kallikrein predictive model: an indifusions in prostate cancer. Nat Rev Cancer. 2008;8(7):497–511. vidual patient data meta-analysis. J Urol. 2018;199(6):1470–4. 37. Varambally S, et al. The polycomb group protein EZH2 is involved 13. Andree KC, van Dalum G, Terstappen LWMM.  Challenges in in progression of prostate cancer. Nature. 2002;419(6907):624–9. circulating tumor cell detection by the CellSearch system. Mol 38. Duan R, Du W, Guo W. EZH2: a novel target for cancer treatment. Oncol. 2016;10(3):395–407. J Hematol Oncol. 2020;13(1):104. 14. Galletti G, et al. Circulating tumor cells in prostate cancer diagno- 39. Yang YA, Yu J.  EZH2, an epigenetic driver of prostate cancer. sis and monitoring: an appraisal of clinical potential. Mol Diagn Protein Cell. 2013;4(5):331–41. Ther. 2014;18(4):389–402. 40. Guo J, et al. Establishing a urine-based biomarker assay for pros 15. Wang L, et al. Promise and limits of the CellSearch platform for tate cancer risk stratification. Front Cell Dev Biol. 2020;8:597961. evaluating pharmacodynamics in circulating tumor cells. Semin 41. Klein EA, et  al. A 17-gene assay to predict prostate cancer Oncol. 2016;43(4):464–75. aggressiveness in the context of Gleason grade heterogene 16. Sharp A, et al. Clinical utility of circulating tumour cell androgen ity, tumor multifocality, and biopsy undersampling. Eur Urol. receptor splice Variant-7 status in metastatic castration-resistant 2014;66(3):550–60. prostate cancer. Eur Urol. 2019;76(5):676–85. 42. Eggener S, et al. A 17-gene panel for prediction of adverse pros567es 17. Tagawa ST, et al. Expression of AR-V7 and ARv in circulating tate cancer pathologic features: prospective clinical validation and tumor cells correlates with outcomes to Taxane therapy in men utility. Urology. 2019;126:76–82. with metastatic prostate cancer treated in TAXYNERGY.  Clin 43. Cullen J, et al. The 17-gene genomic prostate score test as a preCancer Res. 2019;25(6):1880. dictor of outcomes in men with unfavorable intermediate risk 18. Antonarakis ES, et al. Androgen receptor variant-driven prostate prostate cancer. Urology. 2020;143:103–11. cancer: clinical implications and therapeutic targeting. Prostate 44. Crawford ED, et al. CCP score and risk stratification for prostate Cancer Prostatic Dis. 2016;19(3):231–41. cancer patients at biopsy. J Clin Oncol. 2014;32(4_suppl):47. 19. Sadar MD. Discovery of drugs that directly target the intrinsically 45. Cuzick J, et al. Validation of an RNA cell cycle progression score disordered region of the androgen receptor. Expert Opin Drug for predicting death from prostate cancer in a conservatively manDiscovery. 2020;15(5):551–60. aged needle biopsy cohort. Br J Cancer. 2015;113(3):382–9. 20. Lambert SA, Abraham G, Inouye M.  Towards clinical utility of 46. Sommariva S, et  al. Prognostic value of the cell cycle progrespolygenic risk scores. Hum Mol Genet. 2019;28(R2):R133–42. sion score in patients with prostate cancer: a systematic review 21. Huynh-Le MP, et  al. Polygenic hazard score is associated with and meta-analysis. Eur Urol. 2016;69(1):107–15. prostate cancer in multi-ethnic populations. Nat Commun. 47. Gillott DJ, et  al. Specific isoforms of leucine-rich alpha2-­ 2021;12(1):1236. glycoprotein detected in the proliferative endometrium of women 22. Saunders EJ, Kote-Jarai Z, Eeles RA. Identification of germline undergoing assisted reproduction are associated with spontaneous genetic variants that increase prostate cancer risk and influence pregnancy. Fertil Steril. 2008;90(3):761–8. development of aggressive disease. Cancers (Basel). 2021;13(4)

27  Prostate Cancer Biomarkers: The Old and the New 48. Kentsis A, et al. Detection and diagnostic value of urine leucine-­ rich α-2-glycoprotein in children with suspected acute appendicitis. Ann Emerg Med. 2012;60(1):78–83.e1. 49. Li X, et al. Expression of TGF-betas and TGF-beta type II receptor in cerebrospinal fluid of patients with idiopathic normal pressure hydrocephalus. Neurosci Lett. 2007;413(2):141–4. 50. Liu C, et al. A multifunctional role of leucine-rich α-2-glycoprotein 1 in cutaneous wound healing under Normal and diabetic conditions. Diabetes. 2020;69(11):2467–80. 51. Wang Y, et al. TNF-α-induced LRG1 promotes angiogenesis and mesenchymal stem cell migration in the subchondral bone during osteoarthritis. Cell Death Dis. 2017;8(3):e2715. 52. Ai J, et  al. LRG-accelerated differentiation defines unique G-CSFR signaling pathways downstream of PU.1 and C/ EBPepsilon that modulate neutrophil activation. J Leukoc Biol. 2008;83(5):1277–85. 53. Druhan LJ, et al. Leucine rich α-2 glycoprotein: a novel neutrophil granule protein and modulator of myelopoiesis. PLoS One. 2017;12(1):e0170261. 54. proteinatlas.org. 55. Lengyel E.  Ovarian cancer development and metastasis. Am J Pathol. 2010;177(3):1053–64. 56. Perkins A, Liu G. Primary brain tumors in adults: diagnosis and treatment. Am Fam Physician. 2016;93(3):211–7. 57. Chaerkady R, et al. O labeling for a quantitative proteomic analysis of glycoproteins in hepatocellular carcinoma. Clin Proteomics. 2008;4(3–4):137–55. 58. Sarvari J, et al. Comparative proteomics of sera from HCC patients with different origins. Hepat Mon. 2014;14(1):e13103. 59. Wang CH, et al. LRG1 expression indicates unfavorable clinical outcome in hepatocellular carcinoma. Oncotarget. 2015;6(39):42118–29. 60. Zhang Y, et  al. LRG1 suppresses the migration and invasion of hepatocellular carcinoma cells. Med Oncol. 2015;32(5):146. 61. Furuta T, et  al. The multipotential of leucine-rich α-2 glycoprotein 1 as a clinicopathological biomarker of glioblastoma. J Neuropathol Exp Neurol. 2020;79(8):873–9. 62. Lin B, et al. Deep depletion of abundant serum proteins reveals low-abundant proteins as potential biomarkers for human ovarian cancer. Proteomics Clin Appl. 2009;3(7):853–61. 63. Boylan KL, et al. Quantitative proteomic analysis by iTRAQ(R) for the identification of candidate biomarkers in ovarian cancer serum. Proteome Sci. 2010;8:31. 64. Andersen JD, et al. Leucine-rich alpha-2-glycoprotein-1 is upregulated in sera and tumors of ovarian cancer patients. J Ovarian Res. 2010;3:21. 65. Wu J, et al. Altered expression of sialylated glycoproteins in ovarian cancer sera using lectin-based ELISA assay and quantitative glycoproteomics analysis. J Proteome Res. 2013;12(7):3342–52. 66. Mu AK, et  al. Identification of O-glycosylated proteins that are aberrantly excreted in the urine of patients with early stage ovarian cancer. Int J Mol Sci. 2013;14(4):7923–31. 67. Smith CR, et al. Deciphering the peptidome of urine from ovarian cancer patients and healthy controls. Clin Proteomics. 2014;11(1):23. 68. Zhang YS, et  al. Prognostic value of LRG1  in breast cancer: a retrospective study. Oncol Res Treat. 2020:1–6. 69. Kakisaka T, et  al. Plasma proteomics of pancreatic cancer patients by multi-dimensional liquid chromatography and two-dimensional difference gel electrophoresis (2D-DIGE): up-regulation of leucine-­ rich alpha-2-glycoprotein in pancreatic cancer. J Chromatogr B Analyt Technol Biomed Life Sci. 2007;852(1–2):257–67. 70. Choi JW, et  al. Proteomic and cytokine plasma biomarkers for predicting progression from colorectal adenoma to carcinoma in human patients. Proteomics. 2013;13(15):2361–74.

481 71. Sun DC, et al. Leucine-rich alpha-2-glycoprotein-1, relevant with microvessel density, is an independent survival prognostic factor for stage III colorectal cancer patients: a retrospective analysis. Oncotarget. 2017;8(39):66550–8. 72. Heo SH, et  al. Identification of putative serum glycoprotein biomarkers for human lung adenocarcinoma by multilectin affinity chromatography and LC-MS/MS.  Proteomics. 2007;7(23):4292–302. 73. Li Y, et al. Proteomic identification of exosomal LRG1: a potential urinary biomarker for detecting NSCLC.  Electrophoresis. 2011;32(15):1976–83. 74. Liu Y, et al. Integrative proteomics and tissue microarray profiling indicate the association between overexpressed serum proteins and non-small cell lung cancer. PLoS One. 2012;7(12):e51748. 75. Li Z, et al. Exosomal leucine-rich-Alpha2-glycoprotein 1 derived from non-small-cell lung cancer cells promotes angiogenesis via TGF-β signal pathway. Mol Ther Oncolytics. 2019;14:313–22. 76. Ban Z, et  al. LRG-1 enhances the migration of thyroid carcinoma cells through promotion of the epithelial-mesenchymal transition by activating MAPK/p38 signaling. Oncol Rep. 2019;41(6):3270–80. 77. Haupt H, Baudner S.  Isolation and characterization of an unknown, leucine-rich 3.1-S-alpha2-glycoprotein from human serum (author’s transl). Hoppe Seylers Z Physiol Chem. 1977;358(6):639–46. 78. uniprot.org. 79. Takahashi N, Takahashi Y, Putnam FW.  Periodicity of leucine and tandem repetition of a 24-amino acid segment in the primary structure of leucine-rich alpha 2-glycoprotein of human serum. Proc Natl Acad Sci. 1985;82(7):1906. 80. swissmodel.expasy.org. 81. Brasch J, et al. Crystal structure of the ligand binding domain of netrin G2. J Mol Biol. 2011;414(5):723–34. 82. Schwick HG, Haupt H.  Purified human plasma proteins of unknown function. Jpn J Med Sci Biol. 1981;34(5):299–327. 83. Sun D, Kar S, Carr BI.  Differentially expressed genes in TGF-­ beta 1 sensitive and resistant human hepatoma cells. Cancer Lett. 1995;89(1):73–9. 84. Xu J, Lamouille S, Derynck R.  TGF-beta-induced epithelial to mesenchymal transition. Cell Res. 2009;19(2):156–72. 85. Vander Ark A, Cao J, Li X.  TGF-β receptors: in and beyond TGF-β signaling. Cell Signal. 2018;52:112–20. 86. Morikawa M, Derynck R, Miyazono K.  TGF-β and the TGF-β family: context-dependent roles in cell and tissue physiology. Cold Spring Harb Perspect Biol. 2016;8(5) 87. Haque S, Morris JC. Transforming growth factor-β: a therapeutic target for cancer. Hum Vaccin Immunother. 2017;13(8):1741–50. 88. Principe DR, et  al. TGF-β: duality of function between tumor prevention and carcinogenesis. JNCI: J Natl Cancer Inst. 2014;106(2) 89. Wojtowicz-Praga S.  Reversal of tumor-induced immunosuppression by TGF-β inhibitors. Investig New Drugs. 2003;21(1):21–32. 90. Otsuru T, et al. Epithelial-mesenchymal transition via transforming growth factor beta in pancreatic cancer is potentiated by the inflammatory glycoprotein leucine-rich alpha-2 glycoprotein. Cancer Sci. 2019;110(3):985–96. 91. Wang X, et al. LRG1 promotes angiogenesis by modulating endothelial TGF-β signalling. Nature. 2013;499(7458):306–11. 92. Hong Q, et al. LRG1 promotes diabetic kidney disease progression by enhancing TGF-β-induced angiogenesis. J Am Soc Nephrol. 2019;30(4):546–62. 93. Zhang A, et al. Role of VEGF-A and LRG1 in abnormal angiogenesis associated with diabetic nephropathy. Front Physiol. 2020;11:1064.

482 94. Fan M, et  al. Knockdown of long noncoding RNA-taurine-­ upregulated gene 1 inhibits tumor angiogenesis in ovarian cancer by regulating leucine-rich α-2-glycoprotein-1. Anti-Cancer Drugs. 2019;30(6):562–70. 95. Cummings C, et  al. Serum leucine-rich alpha-2-glycoprotein-1 binds cytochrome c and inhibits antibody detection of this apoptotic marker in enzyme-linked immunosorbent assay. Apoptosis. 2006;11(7):1121–9. 96. Alberts B, et al. The molecular biology of the cell. 2014;1025. 97. Garrido C, et al. Mechanisms of cytochrome c release from mitochondria. Cell Death Differ. 2006;13(9):1423–33. 98. Codina R, et al. Cytochrome c-induced lymphocyte death from the outside in: inhibition by serum leucine-rich alpha-2-glycoprotein­1. Apoptosis. 2010;15(2):139–52. 99. Jemmerson R, et  al. Intracellular leucine-rich alpha-2-­ glycoprotein-­1 competes with Apaf-1 for binding cytochrome c in protecting MCF-7 breast cancer cells from apoptosis. Apoptosis. 2021; 100. Shirai R, et  al. Autologous extracellular cytochrome c is an endogenous ligand for leucine-rich alpha2-glycoprotein and beta-type phospholipase A2 inhibitor. J Biol Chem. 2010;285(28):21607–14. 101. Sardana G, et al. Proteomic analysis of conditioned media from the PC3, LNCaP, and 22Rv1 prostate cancer cell lines: discov-

A. L. Magnussen and I. G. Mills ery and validation of candidate prostate cancer biomarkers. J Proteome Res. 2008;7(8):3329–38. 102. Sardana G, Marshall J, Diamandis EP.  Discovery of candidate tumor markers for prostate cancer via proteomic analysis of cell culture-conditioned medium. Clin Chem. 2007;53(3):429–37. 103. Cima I, et  al. Cancer genetics-guided discovery of serum biomarker signatures for diagnosis and prognosis of prostate cancer. Proc Natl Acad Sci. 2011;108(8):3342. 104. Murphy K, et  al. Integrating biomarkers across omic platforms: an approach to improve stratification of patients with indolent and aggressive prostate cancer. Mol Oncol. 2018;12(9):1513–25. 105. Guldvik IJ, et  al. Identification and validation of leucine-rich α-2-glycoprotein 1 as a noninvasive biomarker for improved precision in prostate cancer risk stratification. Eur Urol Open Sci. 2020;21:51–60. 106. Lopez J, Tait SW. Mitochondrial apoptosis: killing cancer using the enemy within. Br J Cancer. 2015;112(6):957–62. 107. Health Quality O. Prolaris cell cycle progression test for localized prostate cancer: a Health technology assessment. Ontario Health Technol Assess Ser. 2017;17(6):1–75. 108. https://www.nice.org.uk/advice/mib65/chapter/Technology-­overview. 109. https://personalisedhealthleeds.com/news-­events/new-­hope-­for-­ prostate-­cancer-­care-­as-­leeds-­launches-­uk-­first-­study/.

The Role of the Microenvironment in Endometriosis: Parallels and Distinctions to Cancer

28

Michael S. Rogers

Abstract

Take-Home Lessons

Phenotypes viewed as distinctive to cancer are often reca• Endometriosis is an estrogen-dependent inflammapitulated in benign disease and consideration of these distory disease that affects ~10% of women of childeases can inform our understanding of the cancer bearing age. microenvironment. Endometriosis is an estrogen-­ • Like cancer, endometriosis is proliferative, invadependent inflammatory disease characterized by the sive, and metastatic. presence of “metastatic” endometrium-like glands and • Endometriosis predisposes to “endometriosis assostroma, together with hemosiderin and (often) fibrosis ciated ovarian cancers.” outside the uterine lumen. It is most often diagnosed as a • Endometriosis shares key microenvironmental fearesult of pain and/or infertility and results in substantial tures with gynecologic malignancies, including economic and personal costs. However, in contrast to canactivated angiogenesis and an altered immune/ cer it is typically not dysplastic and rarely causes death, inflammatory milieu. though it increases the risk of several ovarian cancer sub• Available treatments for endometriosis (NSAIDs, types. Like cancers, the disease is angiogenesis-­dependent hormonal therapy, surgery) are frequently ineffecand genetic studies demonstrate that the VEGFR2 signaltive; new treatments are urgently needed. ing axis plays a key role in the disease. In addition, molecular studies demonstrate that the immune/inflammatory milieu of endometriosis lesions is more similar to that of endometriosis-associated ovarian cancers (EAOCs) than Endometriosis is an estrogen-dependent gynecological disit is to eutopic endometrium. This is consistent with the ease characterized by the presence of endometrium-like dysregulation of a host of immune/inflammatory cells and glands, stroma, and hemosiderin in locations other than the cytokines in disease tissue in ways that often resemble lumen of the uterus. While endometriosis prevalence has not dysregulation observed in ovarian cancer. However, in been clearly determined, the condition is estimated to affect contrast to EAOC, pain is often a key early symptom of ~10% of the general female population [1, 2], and is present endometriosis and can accompany even very small in >50% of women and teenage girls with chronic pelvic lesions. Another key contrast with cancers is the very lim- pain and up to 50% of infertile women [3]. The annual costs ited range of medical treatments available. This is par- of endometriosis in the USA have been estimated at $69.4 tially driven by the much more limited range of side billion during the peridiagnostic period; however, this numeffects that is acceptable for treatment of a non-life-­ ber is likely substantially higher once the ten years surroundthreatening illness in women of childbearing age, but is ing diagnosis are considered [4]. The disease incurs similar also a function of the limited study of endometriosis per-capita costs in Europe as well, emphasizing the high economic burden of the disease [5]. Current endometriosis therpathophysiology that has occurred thus far. apies include medical and surgical options, but the success of these treatments is often limited and recurrence of symptoms is common [6]. Pain scores frequently return toward baseline levels after discontinuation of medication [7–9] and about M. S. Rogers (*) half of patients report recurrence of pain by 12 months postVascular Biology Program, Boston Children’s Hospital, Harvard Medical School, Boston, MA, USA operatively [10]. e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2022 L. A. Akslen, R. S. Watnick (eds.), Biomarkers of the Tumor Microenvironment, https://doi.org/10.1007/978-3-030-98950-7_28

483

484

M. S. Rogers

 ndometrium as a Model of Cancer-Like E Microenvironment There are several microenvironmental alterations that are often described as key hallmarks of cancer. These include angiogenesis, immune dysregulation, inflammation, invasion, and metastasis [11]. Cancers are also often characterized by rapid cell growth. The intense study of malignancy over the last half century has sometimes obscured the extent to which non-cell-autonomous features of cancer are also exhibited by both normal physiology and benign pathologies. In this context a useful comparison can be made between cancer and the endometrium and the most common endometrial pathology, endometriosis (Fig. 28.1). Endometriosis is most commonly characterized by infertility and/or chronic pain, especially in the pelvic or abdominal region [3]. The disease is associated with the growth of proliferative, invasive, endometrium-like tissue, often located on sites such as the ovaries, posterior cul-de-sac, or

Proliferation

Metasta tis

mation Inflam

n sio va In

N rec eu ur ron uit m e

al nt

esis gen gio An

Imm un ity

Fig. 28.1  Key Hallmarks of Endometriosis. Endometriosis shares multiple microenvironmental hallmarks of cancer, including: Proliferation; eutopic endometrium is highly proliferative and endometriosis tissue reflects this. Some normal differentiation is lost as endometriosis only rarely decidualizes. Importantly, the disease is highly responsive to steroid hormone manipulation. Invasion; essentially all lesions invade the mesothelium, with deep invasion into organ structures in a significant subset of women. Metastasis; metastatic tissue implants are produced without oncogenic transformation. Distal metastases are observed. Angiogenesis; robust angiogenesis defines lesion color. Immunity; dysfunctional innate immunity that fails to clear shed tissue and no longer reflects cyclic recruitment of immune cell types to lesions. Inflammation; lesions and surrounding tissue are characterized by ongoing sterile inflammation. Neuronal recruitment is dramatically increased in a subset of individuals, potentially contributing to lesion growth via neuroimmune communication

bladder. The best supported hypothesis for the origin of these lesions is Sampson’s theory of retrograde menstruation [12]. It posits that endometriosis results when menstrual fluid flows through the Fallopian tubes into the abdominal and/or pelvic spaces where it seeds lesions. Endometriosis can lead to several cancers, but is not itself usually considered a precursor lesion. The fraction of endometriosis cases that lead to malignancy is relatively small (~1%). And though there are several “endometriosis-associated ovarian carcinomas” (EAOCs), the odds ratios for women with endometriosis being later diagnosed with these cancers is substantially smaller than with typical precursor lesions; for clear cell ovarian cancer (OR, 3.73), endometrioid ovarian cancer (OR, 2.32), and low-grade serous ovarian cancer (OR, 2.02) [13, 14]. Endometriosis is not usually dysplastic (though nuclear atypia is sometimes observed, especially in conjunction with EAOC). Importantly, although endometriosis is associated with individual cancer-associated mutations (in, e.g., KRAS, ARID1A, PIK3CA [15]), most lesions have no cancer-­ associated mutations and those that do have only a single cancer gene mutated. In addition, disease symptoms commonly manifest themselves in young women shortly after menarche, suggesting that the local microenvironment, rather than mutational processes (that take time) predominates in disease susceptibility. Thus, comparing endometriosis pathophysiology with that of EAOC can help differentiate mutation-driven processes from those that result from maladaptive results of normal biology.

 he Endometrium Is an Extraordinarily T Proliferative Tissue The endometrium is the lining of the uterus and its major function is to enable embryo implantation, placenta formation, and gestation. It is composed of two layers, the basalis, a ~ 0.5 mm thick [16] layer of compact stromal tissue with “rhizome-like” horizontal glands [17, 18] that is not shed during menstruation. Overlying and arising from the basalis is the functionalis, an often spongy layer containing a characteristic stroma with vertical glands and an overlying luminal epithelium. This layer ranges in thickness from 0 to >8 mm in thickness, depending on the menstrual phase. Menstruation is induced when a drop in progesterone causes the dense network of spiral arteries feeding the functionalist to constrict, resulting in tissue hypoxia/ischemia, apoptosis, and shedding of the vast majority, if not all, of the functionalis. The shedding process is followed by an extraordinarily rapid proliferation of endometrial epithelial and stromal cells, that rivals the fastest tumor growth rates. Within a few days of the onset of menstruation, epithelial cells from glands in the basalis and any residual functionalis proliferate and cover the newly denuded lumen of the endo-

28  The Role of the Microenvironment in Endometriosis: Parallels and Distinctions to Cancer

metrium, forming a new luminal epithelium by the end of the menstrual phase. Then, during the proliferative phase, the endometrium rapidly expands, more than doubling in size in a week [19]. The proliferative phase ends with ovulation. The next phase is called the secretory phase and is named for the secretion by the endometrial glands of histotroph, which nourishes the developing embryo prior to establishment of the placenta. During this phase, modest continued glandular proliferation is accompanied by decidualization of the stroma and continued vascular proliferation in preparation for embryo implantation. If these steps do not occur, progesterone drops and the cycle repeats. In modern humans this can occur >400 times in a lifetime.

Dissemination/Colonization: Endometriosis as “Metastatic” Endometrium Because the Fallopian tubes are open to the pelvic space, retrograde menstruation (flow of menstrual fluid through the Fallopian tubes, in addition to through the cervix) is common, being observed in ~90% of women. About 10% of women experience endometriosis, the presence of endometrium-­like tissue in a location other than the uterine lumen [20]. The most widely accepted hypothesis for the origin of endometriosis is that it represents metastatic dissemination of eutopic endometrium via retrograde menstruation [12]. The retrograde menstruation hypothesis is supported by the observation that risk of endometriosis is increased by anything that is likely to increase the amount of menstrual tissue deposited in the pelvic space (earlier menarche, decreased cycle length, heavier flow, obstructed flow) [20, 21]. In bilateral endometrioma, lesions in a given patient typically do not share mutations [22], suggesting that the capacity for dissemination and implantation is not rate limiting in lesion generation. Importantly, endometriosis is not limited to the pelvic and abdominal spaces. Lesions have been reported throughout the body, including lungs, brain, etc. It is commonly assumed that lesions arrive at these locations via lymphatic or hematogenous spread, but this has not been clearly demonstrated. Thus, in addition to very rapid proliferation, endometrium exhibits the ability to metastasize. In the context of EAOC, and especially cancers that arise from endometriosis lesions, this means that the tissue has metastasized to a new location before oncogenic transformation.

Invasion in Endometriosis In addition to high proliferative rates, endometriosis can also be invasive. (Adenomyosis, which consists of endometrium invading into the myometrium is typically considered a separate disease and will not be discussed here.) Endometriosis lesions are

485

commonly divided into 3 types according to location and invasivity. Most endometriosis lesions are of the superficial peritoneal type and invade the mesothelium, but only exhibit shallow (