Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response [1 ed.] 1588296466, 9781588296467

Citation preview

Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response

Cancer Drug Discovery and Development™ Beverly A. Teicher, Series Editor Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response, edited by Federico Innocenti, 2008 Checkpoint Responses in Cancer Therapy, edited by Wei Dai, 2008 Cancer Proteomics: From Bench to Bedside, edited by Sayed S. Daoud, 2008 Antiangiogenic Agents in Cancer Therapy, Second Edition, edited by Beverly A. Teicher and Lee M. Ellis, 2007 Apoptosis and Senescence in Cancer Chemotherapy and Radiotherapy, Second Edition, edited by David A. Gerwitz, Shawn Edan Holtz, and Steven Grant, 2007 Molecular Targeting in Oncology, edited by Howard L. Kaufman, Scott Wadler, and Karen Antman, 2007 In Vivo Imaging of Cancer Therapy, edited by Anthony F. Shields and Patricia Price, 2007 Transforming Growth Factor- in Cancer Therapy, Volume II: Cancer Treatment and Therapy, edited by Sonia Jakowlew, 2008 Transforming Growth Factor- in Cancer Therapy, Volume 1: Basic and Clinical Biology, edited by Sonia Jakowlew, 2008 Microtubule Targets in Cancer Therapy, edited by Antonio T. Fojo, 2007

Cytokines in the Genesis and Treatment of Cancer, edited by Michael A. Caligiuri, Michael T. Lotze, and Frances R. Balkwill, 2007 Regional Cancer Therapy, edited by Peter M. Schlag and Ulrike Stein, 2007 Gene Therapy for Cancer, edited by Kelly K. Hunt, Stephan A. Vorburger, and Stephen G. Swisher, 2007 Deoxynucleoside Analogs in Cancer Therapy, edited by Godefridus J. Peters, 2006 Cancer Drug Resistance, edited by Beverly A. Teicher, 2006 Histone Deacetylases: Transcriptional Regulation and Other Cellular Functions, edited by Eric Verdin, 2006 Immunotherapy of Cancer, edited by Mary L. Disis, 2006 Biomarkers in Breast Cancer: Molecular Diagnostics for Predicting and Monitoring Therapeutic Effect, edited by Giampietro Gasparini and Daniel F. Hayes, 2006 Protein Tyrosine Kinases: From Inhibitors to Useful Drugs, edited by Doriana Fabbro and Frank McCormick, 2005 Bone Metastasis: Experimental and Clinical Therapeutics, edited by Gurmit Singh and Shafaat A. Rabbani, 2005 The Oncogenomics Handbook, edited by William J. LaRochelle and Richard A. Shimkets, 2005

Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response

Edited by

Federico Innocenti, MD, PhD University of Chicago, Chicago, IL, USA

Editor Federico Innocenti, MD PhD University of Chicago Department of Medicine Section of Hematology/Oncology 5841 South Maryland Avenue Chicago IL 60637-1470 USA [email protected]

Series Editor Beverly A. Teicher, PhD Genzyme Corporation Framingham, MA

ISBN: 978-1-58829-646-7

e-ISBN: 978-1-60327-088-5

Library of Congress Control Number: 2008938265 ©Humana Press, a part of Springer Science+Business Media, LLC 2008 All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Humana Press, 999 Riverview Drive, Suite 208, Totowa, NJ 07512 USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. While the advice and information in this book are believed to be true and accurate at the date of going to press, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein. Printed on acid-free paper 987654321 springer.com

P REFACE Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response provides the most comprehensive body of knowledge available on the role of genetic and genomic variation in the individualization of drug therapies in cancer patients. As a consequence of the intrinsic chromosomal and genetic instability of the tumor genome, it is generally believed that tailoring of chemotherapy in cancer patients might be achieved by molecular analysis of patient tumor DNA. In addition, to reduce the toxicity risk of patients, the tumor DNA information should be integrated with the available data on polymorphic drug-metabolizing enzyme and transporter genes mediating the exposure of patients to active drugs and/or their active metabolites. The chapters of this book clearly show how DNA information from both the host (germline) and the tumor should be taken into account for rational selection of drug therapies in cancer patients, an aspect that received little attention, despite its importance. The availability of new molecular approaches to the selection of drug therapy is an emerging need, because the traditional approach based on the evaluation of patient and tumor characteristics is clearly far from optimal. Many treated patients do not experience significant benefits from the treatment, while they often experience moderate to severe toxicities. In addition, the development and clinical use of novel molecularly targeted agents (alone or in combination with classical cytotoxic therapy) requires the understanding of the molecular features of the tumors and the identification of tumor markers of response. In this book, the readers will find a series of chapters addressing the role of genomic information in cancer therapy and in drug development. Several books on pharmacogenomics are currently available, but this book represents a unique source, as it describes experimental approaches, statistical strategies, and clinical examples of the application of genomic medicine in oncology. Many outstanding scientists in the field of cancer pharmacogenomics have been invited to contribute, and I am grateful to have had the opportunity to work with them, learning a great deal from reading their chapters. I have approached this book from both a basic and an applied perspective. Among three different sections, six chapters in the first section are focused on up-to-date genomic experimental approaches in oncology, including genome-wide phenotyping (microarray and proteomics) and genotyping methods, as well as novel cell-based models used for the identification of genetic markers of drug response. The second section shows how genetic and genomic information is currently applied to treatment individualization and optimization: Eleven chapters describe some of the most elegant examples of genetic and genomic markers that are predictive of the survival and toxicity risk of cancer patients. Finally, in the third section, readers will find four chapters that address the role of pharmacogenomics in drug development in oncology, including an industry perspective on this subject, as well as statistical aspects related to the discovery of pharmacogenomic biomarkers during drug development. v

vi

Preface

Because the discovery of genetic markers of response is of high relevance in oncology, we perceived the need that a collection of multidisciplinary topics be gathered together to discuss this important aspect of pharmacogenomics applied to cancer patients. I believe that this book represents the first reference for researchers in the field of cancer pharmacogenomics and clinicians, from both the academia and industry. Federico Innocenti

C ONTENTS

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Part I

Genomic Experimental Approaches in Oncology

1 Toward the Realization of the Promise of Microarrays in Oncology . . . . . . . . . 3 Natalie Stickle and Neil Winegarden 2 Cell-Based Models to Identify Genetic Variants Contributing to Anticancer Drug Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 M. Eileen Dolan and Howard McLeod 3 Proteomic Analysis in Cancer Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Yasuhiro Kuramitsu and Kazuyuki Nakamura 4 MicroRNAs and Discovery of New Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Soken Tsuchiya, Yasushi Okuno, and Gozoh Tsujimoto 5 Pharmacogenomics of the National Cancer Institute’s 60-Tumor Cell Panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 Anders Wallqvist, Ruili Huang, and David G. Covell 6 Use of Single-Nucleotide Polymorphism Array for Tumor Aberrations in Gene Copy Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 Kwong-Kwok Wong Part II Pharmacogenomics of Toxicity and Response Of Chemotherapy 7 Concordance Between Tumor and Germline DNA . . . . . . . . . . . . . . . . . . . . . . . 91 Sharon Marsh

vii

viii

Contents

8 Epidermal Growth Factor Receptor Mutations and Sensitivity to Selective Kinase Inhibitors in Human Lung Cancer . . . . . . . . . . . . . . . . . . . 103 Anurag Singh, Sreenath V. Sharma, and Jeffrey Settleman 9 BCR-ABL Mutations and Imatinib Resistance in Chronic Myeloid Leukemia Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 Mark R. Litzow 10 Role of Thymidylate Synthase Gene Variations in Colorectal Cancer Patients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Georg Lurje and Heinz-Josef Lenz 11 Thiopurines in the Treatment of Childhood Acute Lymphoblastic Leukemia and Genetic Variants of the Thiopurine S-Methyltransferase Gene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173 Martin Stanulla, Elke Schaeffeler and Matthias Schwab 12 Impact of Polymorphisms on the Clinical Outcomes of Monoclonal Antibody Therapy Against Hematologic Malignancies . . . . . . . . . . . . . . . . 203 Dong Hwan Kim 13 DNA Repair and Mitotic Checkpoint Genes as Potential Predictors of Chemotherapy Response in Non-Small-Cell Lung Cancer . . . . . . . . . . . 231 Rafael Rosell, Miquel Taron, Mariacarmela Santarpia, Fernanda Salazar, Jose Luis Ramirez, and Miguel Angel Molina 14 Dihydropyrimidine Dehydrogenase (Dpyd) Gene Polymorphism: Portrait of a Serial Killer . . . . . . . . . . . . . . . . . . . . . . . . . . . 249 Joseph Ciccolini, C´edric Mercier, and G´erard Milano 15 Impact of UDP-Glucuronosyltransferase 1A Haplotypes on Irinotecan Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 Kimie Sai, Hironobu Minami, Yoshiro Saito, and Jun-ichi Sawada 16 Microarray Profiling in Breast Cancer Patients . . . . . . . . . . . . . . . . . . . . . . . . . 287 Yong Qian, Xianglin Shi, Vincent Castranova, and Nancy L. Guo 17 Role of the Folate-Pathway and the Thymidylate Synthase Genes in Pediatric Acute Lymphoblastic Leukemia Treatment Response . . . . . . . . . 299 Lea Cunningham and Richard Aplenc

Contents

ix

Part III Pharmacogenomics in Clinical Drug Development in Oncology 18 Pharmacogenomics in Drug Development: A Pharmaceutical Industry Perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313 Tal Zaks 19 Identification of Pharmacogenomic Biomarker Classifiers in Cancer Drug Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327 Richard Simon 20 Toxicogenomics Application to Oncology Drug Development . . . . . . . . . . . . 339 Luigi Calzolai and Teresa Lettieri 21 Strategies to Identify Pharmacogenomic Biomarkers: Candidate Gene, Pathway-Based, and Genome-Wide Approaches . . . . . . 353 Xifeng Wu, Jian Gu, and Margaret R. Spitz Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371

“This page left intentionally blank.”

C ONTRIBUTORS R ICHARD A PLENC , MD, MSCE • Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania, USA L UIGI C ALZOLAI , P H D • Medway School of Pharmacy, University of Kent, Kent, UK V INCENT C ASTRANOVA , P H D • The Pathology and Physiology Research Branch, Health Effects Laboratory Division, National Institute of Occupational Safety and Health, Morgantown, West Virginia, USA J OSEPH C ICCOLINI , P HARM D, P H D • Pharmacokinetics—Medical Oncology, Universit´e de la M´editerran´ee, Marseille, France DAVID G. C OVELL , P H D • Developmental Therapeutics Program, National Cancer Institute, Frederick, Maryland, USA L EA C UNNINGHAM , MD • Department of Pediatrics, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania, USA M. E ILEEN D OLAN , P H D • Section of Hematology/Oncology, University of Chicago, Chicago, Illinois, USA J IAN G U , P H D • Department of Epidemiology, The University of Texas, M.D. Anderson Cancer Center, Houston, Texas, USA NANCY L. G UO , P H D • Mary Babb Randolph Cancer Center, Department of Community Medicine, West Virginia University, Morgantown, West Virginia, USA RUILI H UANG , P H D • Developmental Therapeutics Program, National Cancer Institute, Frederick, Maryland, USA D ONG H WAN K IM , MD, P H D • Department of Hematology/Oncology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea F EDERICO I NNOCENTI , MD, P H D • Department of Medicine, Section of Hematology/ Oncology, University of Chicago, Chicago, Illinois, USA YASUHIRO K URAMITSU , MD, P H D • Department of Biochemistry and Functional Proteomics, Yamaguchi University, Graduate School of Medicine, Yamaguchi, Japan H EINZ -J OSEF L ENZ , MD, FACP • Division of Medical Oncology, University of Southern California, Norris Comprehensive Cancer Center, Keck School of Medicine, Los Angeles, California, USA T ERESA L ETTIERI , P H D • European Commission-DG Joint Research Centre, Institute for Environment and Sustainability, Ispra (VA), Italy M ARK R. L ITZOW, MD • Division of Hematology Research, Mayo Clinic, Rochester, Minnesota, USA G EORG L URJE , MD • Division of Medical Oncology, University of Southern California, Norris Comprehensive Cancer Center, Keck School of Medicine, Los Angeles, California, USA xi

xii

Contributors

S HARON M ARSH , P H D • Washington University, Division of Molecular Oncology, St. Louis, Missouri, USA H OWARD L. M C L EOD , P H D • University of North Carolina, Chapel Hill, Chapel Hill, North Carolina, USA C EDRIC M ERCIER , MD, P H D • Medical Oncology, Universit´e de la M´editerran´ee, Marseille, France G ERARD M ILANO , P H D • Laboratoire d’Oncopharmacologie, Centre Antoine Lacassagne, Nice, France H IRONOBU M INAMI , MD • Division of Oncology/Hematology, National Cancer Center Hospital East, Kashiwa, Japan M IGUEL A NGEL M OLINA , P H D • Catalan Institute of Oncology, Hospital Germans Trias i Pujol, Badalona (Barcelona), Spain K AZUYUKI NAKAMURA , P H D • Department of Biochemistry and Functional Proteomics, Yamaguchi University, Graduate School of Medicine, Yamaguchi, Japan YASUSHI O KUNO , P HARM D • Department of PharmacoInformatics, Graduate School of Pharmaceutical Sciences, Kyoto University, Kyoto, Japan YONG Q IAN , P H D • The Pathology and Physiology Research Branch, Health Effects Laboratory Division, National Institute of Occupational Safety and Health, Morgantown, West Virginia, USA J OSE L UIS R AMIREZ , P H D • Catalan Institute of Oncology, Hospital Germans Trias i Pujol, Badalona (Barcelona), Spain R AFAEL ROSELL , MD • Catalan Institute of Oncology, Hospital Germans Trias i Pujol, Badalona (Barcelona), Spain K IMIE S AI , P H D • Division of Biosignaling, National Institute of Health Sciences, Tokyo, Japan YOSHIRO S AITO , P H D • Division of Biochemistry and Immunochemistry, National Institute of Health Sciences, Tokyo, Japan F ERNANDA S ALAZAR , P H D • Catalan Institute of Oncology, Hospital Germans Trias i Pujol, Badalona (Barcelona), Spain M ARIACARMELA S ANTARPIA , MD • Department of Clinical Oncology and Innovative Therapy, University of Messina, School of Medicine, Messina, Italy J UN - ICHI S AWADA , P H D • Division of Biochemistry and Immunochemistry, National Institute of Health Sciences, Tokyo, Japan E LKE S CHAEFFELER , P H D • Margarete-Fischer-Bosch, Institute of Clinical Pharmacology, Stuttgart, Germany M ATTHIAS S CHWAB , MD • Department of Clinical Pharmacology, University Hospital Tuebingen, Tubingen, Germany

Contributors

xiii

J EFFREY S ETTLEMAN , P H D • Massachusetts General Hospital Cancer Center, Charlestown, Massachusetts, USA S REENATH V. S HARMA , P H D • Massachusetts General Hospital Cancer Center, Charlestown, Massachusetts, USA X IANGLIN S HI , P H D • The Pathology and Physiology Research Branch, Health Effects Laboratory Division, National Institute of Occupational Safety and Health, Morgantown, West Virginia, USA R ICHARD S IMON , DS C • Biometric Research Branch, National Cancer Institute, Bethesda, Maryland, USA A NURAG S INGH , P H D • Massachusetts General Hospital Cancer Center, Charlestown, Massachusetts, USA M ARGARET R. S PITZ , MD, MPH • Department of Epidemiology, The University of Texas, M.D. Anderson Cancer Center, Houston, Texas, USA M ARTIN S TANULLA , MD, MS C • Pediatric Hematology Oncology, Hannover Medical School, Hannover, Germany NATALIE S TICKLE , MS C • UHN Microarray Centre, University Health Network, T. Robert Beamish Family Convergence Centre of Medical Discovery, Toronto, Ontario, Canada M IQUEL TARON , P H D • Catalan Institute of Oncology, Hospital Germans Trias i Pujol, Badalona (Barcelona), Spain S OKEN T SUCHIYA , P HARM D • Department of Genomic Drug Discovery Science, Department of PharmacoInformatics, Graduate School of Pharmaceutical Sciences, Kyoto University, Kyoto, Japan G OZOH T SUJIMOTO , MD, P H D • Department of Genomic Drug Discovery Science, Graduate School of Pharmaceutical Sciences, Kyoto University, Kyoto, Japan A NDERS WALLQVIST, P H D • Laboratory of Computational Technologies, SAICFrederick, Inc., Frederick, Maryland, USA N EIL W INEGARDEN , MS C • UHN Microarray Centre, University Health Network, T. Robert Beamish Family Convergence Centre of Medical Discovery, Toronto, Ontario, Canada K WONG -K WOK W ONG , P H D • Department of Gynecologic Oncology, The University of Texas, M.D. Anderson Cancer Center, Houston, Texas, USA

xiv

Contributors

X IFENG W U , MD, P H D • Department of Epidemiology, The University of Texas, M.D. Anderson Cancer Center, Houston, Texas, USA TAL Z AKS , MD, P H D • Translational Medicine-Oncology, GlaxoSmithKline, Collegeville, Pennsylvania, USA

I

G ENOMIC E XPERIMENTAL A PPROACHES IN O NCOLOGY

“This page left intentionally blank.”

1

Toward the Realization of the Promise of Microarrays in Oncology Natalie Stickle, MSc, and Neil Winegarden, MSc CONTENTS M ICROARRAYS FOR C ANCER R ESEARCH : H ISTORICAL P ERSPECTIVES T UMOR H ETEROGENEITY S TEM C ELL T HEORY OF C ANCER P ROFILING S MALL S AMPLES WITH M ICROARRAYS : A MPLIFICATION P ROSPECTIVE V ERSUS R ETROSPECTIVE S TUDIES D IAGNOSTICS AND P ROGNOSTICS C ONCLUSIONS R EFERENCES

S UMMARY Microarrays have long been promised to be a tool that might one day revolutionize oncology research and drug development. Despite the tremendous potential, however, there have been few breakthroughs that can be directly attributed to microarray-based profiling. While many researchers now say that microarrays have been over-hyped, it is more likely that early experiments were simply conducted in a na¨ıve manner. Many believe that as technology and our understanding of experimental design improves, so

From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 1, 

3

4

Part I / Genomic Experimental Approaches in Oncology

too will the end results. We believe that with new approaches, particularly the use of pure cell populations potentially coupled with new and improved RNA amplification methodologies, the promise of microarrays for oncology finally will be realized. Key Words: Microarrays; gene expression; RNA amplification; laser capture microdissection; cancer stem cell theory; pure cell populations; diagnostics; prognostics

1. MICROARRAYS FOR CANCER RESEARCH: HISTORICAL PERSPECTIVES Gene expression profiling in cancer has become routine in recent years and to date encompasses the largest category of research using DNA microarrays. Microarray analysis has been used to assess transcriptome-level (expression analysis) (1) and genome-level (single-nucleotide polymorphism (2), amplification/deletion (3)) differences in cancerous versus normal cell samples. Furthermore, microarray analysis has been used to identify putative diagnostic and prognostic markers that will hopefully translate into clinical tests that can be used to stratify patients and determine treatment regimens (4). Comparison of treatment response at both the RNA and DNA levels has been conducted in an effort to further direct the most optimal therapy for an individual patient. Classically, many large-scale studies have revealed groups of genes differentially expressed between cancerous and normal cells, including genes known to be important for neoplastic transformation. However, despite the vast amounts of data generated to date, there has been limited translation from the research bench to the clinic. There are few drugs currently in the pipelines of the pharmaceutical industry that were developed because of a microarray-identified target (microarrays have been used to validate and test many potential pharamcophores, but few targets have actually been identified by microarrays). One of the greatest early promises of microarrays was the ability to identify signatures of biomarkers (panels of genes), which would be able to diagnose or prognose complex diseases such as cancer. Despite some notable exceptions, however, this potential benefit has largely gone unmet. As researchers have looked to capitalize on their investments in these high-throughput technologies, there has been some shift recently toward genotyping rather than transcript analysis in cancer biology. While these techniques are promising, there is still little sign at this early stage that suggests that this shift in focus will lead to revolutionary changes in the type of data obtained. All of these developments have led to the discouragement of many researchers who have begun to re-evaluate their desire to use high-throughput technologies for oncology research. Despite this, it is our belief that the technology in-and-of itself is not inherently flawed, and rather it is the way that microarrays have typically been applied to cancer research that has led to a decrease in the potential power of the analyses. Recent advances in microarray design, labeling, and amplification technologies—and even our theory of the origin of cancerous lesions—are leading to changes in the way microarray experiments are conducted. These developments are likely to finally fulfil the realization of the promise of microarrays in oncology.

Chapter 1 / Microarrays and Oncology

5

2. TUMOR HETEROGENEITY Many of the anti-neoplastic drugs currently available on the market specifically target the cancerous tumor and do not account for the overall heterogeneity of the disease. It is reasonable to assume that due to the complex make-up of a tumor, combined therapies may be necessary to treat the different cell types involved. One major hurdle to understanding tumor cells at the gene expression level is our limited ability to isolate pure populations of diseased cells. Tumor cells can be extracted from blood with relative ease, but with solid tumors, such as those found in prostate or breast cancer, tumor cells cannot be isolated so easily. Indeed, these cell populations often show very large degrees of heterogeneity. Solid tumors expand by infiltrating within the normal tissue scaffold, making it difficult but necessary to dissect the malignant cells away from the normal tissue structures (5). For those samples that are dissectible in theory, some cells outside the area of interest may remain attached, resulting in a sub-pure sample that confounds analysis ( 5). Even in cases where the tumor can be distinguished easily from the surrounding normal tissue, the tumors themselves remain heterogeneous. Tumor heterogeneity refers to the existence of distinct subpopulations of tumor cells with specific characteristics within a single neoplasm. Breast cancer is a classic example of a heterogeneous disease. First, the term breast cancer does not itself refer to a single disease. Breast cancers include many different diseases including (but not limited to) adenomas, papillomas, invasive ductal carcinoma, ductal carcinoma in situ (DCIS), and lobular carcinoma in situ (LCIS) (6). Breast tissue itself is highly heterogeneous, containing stroma, epithelial, ductal, lymphocytic, fibroblastic and other structures which further complicate issues. Even in morphologically pure populations, there remains a great deal of molecular heterogeneity. This is true also of breast cancer, which includes alterations in ER, Her2 BRACA1, and BRCA2 status among others (4,7,8,9). In addition to the well-recognized clonal mutations that afford a survival advantage to cells within the tumor microenvironment (i.e., areas of hypoxia, acid pH, and poor nutrient supply) (10,11), there is evidence of many randomly distributed unselected mutations that contribute to the heterogeneity of cells in a tumor (12,13). Overall, heterogeneity is likely the major hurdle in the usefulness of microarray technology in tumor biology. Important cell-specific signatures, particularly in rare cell populations, would be lost in a microarray analysis as to date, many microarray analyses used heterogeneous tumor samples for profiling. More recently, researchers have begun to turn their attention to microdissected, relatively pure cell populations for microarray analysis.

2.1. Tumor Microdissection Microdissection of a tumor was initially carried out using a standard syringe ( 5). The tumor was viewed under a microscope so that it could be separated from the surrounding normal tissues. For this method to be effective, tumors needed to be easily defined under the microscope, which limited the number of samples that were compatible. Microdissection techniques have further evolved to include techniques such as laser microbeam microdissection (LMM) and laser capture microdissection (LCM). These

6

Part I / Genomic Experimental Approaches in Oncology

techniques afford laser precision and the possibility to isolate single cells ( 14). As a result, these methods have gained importance as tools to obtain purified cell populations from tissue (15). LCM has been used as a tool for the purification and stratification of tumor samples that are to be used in subsequent microarray analyses (16,17,18,19,20,21). Separation of the tumor from the surrounding tissue provides cleaner populations of cells that can be used for the subsequent analysis of gene expression profiles or chromosomal aberrations. Furthermore, by purifying individual cell types, distinct signatures for each cell type can also be obtained using microarray profiling ( 22, 23). Such analyses demonstrate that the cancers are not homogenous at the molecular level. Yang et al. (23) demonstrated that the signature obtained from bulk tumors can include many common genes found in more pure populations of epithelial cells, but that both the purified epithelia and the bulk tumor have characteristic sets of genes that are unique as well. In analysing the results of a gene expression profiling on purified epithelia from ER-␣ positive and ER-␣ negative cells, a total of 146 ER-related genes were identified. When the authors then compared the expression profiles to those of the bulk tumors from which the purified cell populations were obtained, 61 of these genes were identified as being in-common. Thus, 85 of the genes identified (58%) could only be identified by using the pure cell population (23). As LCM becomes more commonly used upstream of microarray analysis, the quality of data as well as the usefulness of the signatures obtained will likely increase. With a greater understanding of the molecular heterogeneity at the transcriptome level for example, it will be possible to develop combinatorial therapies that use multiple different drugs to target the several different cell types involved in the disease. Very few other technologies hold the promise of microarrays when it comes to understanding cancers at this level of detail, and thus it is likely that these new and more focused approaches will lead to new therapeutic strategies.

3. STEM CELL THEORY OF CANCER While natural heterogeneity of cancer specimens is well documented, and the ability to isolate hundreds to thousands of cells with techniques such as LCM have allowed microarray analysis of more pure populations, the recent re-introduction of the cancer stem cell theory (24,25) has presented an extreme example of the need to isolate rare populations of cells. Cancer stem cells appear to exist at a frequency on the order of 1 in 1000 cells (26). While obtaining completely pure populations of stem cells remains difficult, highly enriched populations of these cells can be isolated using flow cytometry separating on the basis of the presence of certain specific cell markers. One hypothesis is that cancer stem cells are relatively immune to current therapeutic methodologies, due in large part to the fact that these cells divide only infrequently and are generally quiescent. As these cells also have the potential to cause the formation of a new cancer, therapeutics targeted to these specific cells would be of immense benefit. The fact that cancer stem cells occur in such low frequency dictates that what may arguably be the most important signature to identify within the tumor is completely lost in most microarray analyses.

Chapter 1 / Microarrays and Oncology

7

While laser capture microdissection and other related techniques may allow for the generation of relatively homogenous cell populations, most such purification steps will largely eliminate any potential cancer stem cells, because these techniques necessitate the identification of differentiated cell types for isolation. As such, cancer stem cells would frequently be eliminated from analysis, and any cancer stem cells that may be accidentally included in the analysis would be in such low occurrence that their signature would most certainly be missed. Assuming that the cancer stem cell theory is in fact correct, then we can see how it is possible that the reason we have not yet learned a great deal about cancer using microarrays is not due to the technology itself, but more due to its application and the sample that is being used to profile the cancer. As techniques are introduced that will allow the profiling of single cells, it may be possible to obtain specific gene expression signatures for these cancer stem cells. Drugs developed specifically to target these genes will revolutionize the means by which the disease is treated. Elimination of these tumor progenitors would not only help eradicate the initial disease, but would also essentially prevent recurrence as well. The problem, however, is that it is very difficult to profile individual cells, and techniques to do so have only recently been introduced.

4. PROFILING SMALL SAMPLES WITH MICROARRAYS: AMPLIFICATION With advances in tumor microdissection and flow sorting, the current trend in assessment of malignancies is toward evaluation of ever-smaller samples, in an attempt to separate the effects of tumor heterogeneity from the tumor-specific genetic signature. Furthermore, as diagnostic methodologies improve, the overall tumor, when biopsied, is becoming smaller and smaller. In many cases, much of the tumor material obtained during a biopsy is required for pathology, leaving very small amounts of tissue for the molecular researcher. There is also a trend toward performing minimally invasive biopsy techniques such as needle/core biopsies and fine needle aspirates (FNAs). The amount of material available from such techniques, even if the bulk (non-microdissected) sample were to be used, can be very small. Traditional DNA microarray analysis requires a fairly substantial amount of material for analysis; some 5–10 ␮g of total RNA is usually required for analysis in the absence of amplification. In order to obtain this quantity of RNA from typical epithelial cells, as many as 5 ×105 or 1 ×106 cells are required. Clearly, this total RNA requirement poses a challenge when studying microdissected samples, but it also presents a challenge when using small, unique clinical samples, such as FNA biopsies. Assersohn et al. (27) reported that the median recovery of breast FNAs was 202,500 cells, which translated to between 0.81 and 1.42 ␮g of total RNA, well below the typical requirement for a microarray experiment. Consequently, many researchers routinely incorporate amplification methods into their gene expression screens. There is currently no consensus in the literature with regard to the best amplification strategy to utilize. Indeed, even the fidelity of the various amplification methods available is the subject of dispute. Some researchers report that amplified RNA can substitute for total RNA in gene expression experiments (28,29,30) while others have shown that amplification of RNA introduces bias to the results ( 31, 32). It is quite possible, however, that this

8

Part I / Genomic Experimental Approaches in Oncology

apparent disagreement is caused by a difference in what the researchers consider to be an acceptable level of fidelity. It is well accepted that any amplification technique will introduce some degree of bias ( 33). What is more up for debate is to what level this bias can be introduced before it causes a detrimental effect on the analysis. Some would argue that no level of bias is acceptable; however, it is likely that as long as the amplification methodology is highly reproducible, statistical methods can be used to model and remove or correct for the bias.

4.1. T7-(“Eberwine”)–Type Amplification The most popular methods of RNA amplification for microarray-based transcription profiling are modified from the classic Eberwine protocol ( 34). The basic premise of this approach is that mRNA is converted into cDNA using an oligo-dT primer that has an additional sequence encoding the T7-promoter appended to it. After performing a second strand cDNA synthesis, an artificial gene under the control of a T7-promoter is created. This gene can then be transcribed by T7-polymerase creating multiple RNA copies of each gene. This basic technique is utilized by many researchers and is a standard part of the Affymetrix, Agilent, and Illumina labeling protocols. Proponents of this technique point to the fact that it is linear in nature, while providing great increases in sensitivity. This protocol can be performed in tandem (multiple rounds of amplification) to obtain even greater amounts of material; however, each successive round of amplification does introduce more bias (35). With multiple rounds of amplification, as little as 10 ng of total RNA can be used to create labeled material for microarray analysis ( 35, 36). While this thousand-fold increase in sensitivity is impressive, the requirement for a minimum of 1000 cells can still be problematic for many researchers using LCM techniques for tumor profiling. Another drawback of this method is the time requirement for amplification: a single round of amplification adds between 1.5 and 2 days of time to the labeling procedure, whereas two rounds of amplification can add 3 days (37).

4.2. Isothermal Amplification Strategies While T7-mediated amplification has shown a great deal of promise and has been employed in numerous studies (including nearly all microarray experiments to date using LCM isolated cells), other methodologies are being sought that may provide both greater sensitivity and fidelity while reducing the time requirement. One such method, Ribo-SPIA from NuGen Technologies, allows for amplification from as little as 5 ng of total RNA in as little as 5 hr (37,38). Ribo-SPIA utilizes a chimeric RNA/DNA primer that links a known RNA sequence tag to the 5 -end of an oligo-dT cDNA primer. After creating double-stranded cDNA from the mRNA, RNaseH is used to digest only the short RNA portion of the primer, which enables another chimeric primer to bind, and a new cDNA strand then displaces the old one. This process is repeated over many cycles creating thousands of single-stranded cDNA all in one tube (37,38). The major advantage of this technique over T7-based amplification is that it can be completed in a matter of hours, rather than days. A novel signal amplification method based on the rolling circle amplification (RCA) technology has also been reported. RCA involves many rounds of isothermal enzymatic

Chapter 1 / Microarrays and Oncology

9

synthesis where DNA polymerase extends a hybridized primer (consisting of a DNA minicircle) by continuously progressing around the circular DNA probe to replicate the sequence over and over again, producing a single amplified product that remains linked to the DNA primer. Nallur et al. (39) demonstrated that RCA is a sensitive method for detecting nucleic acid signals of microarrays. They reported that for DNA targets there was at least an 8000-fold increase in detection sensitivity over standard hybridization under the same conditions. Although this indicates a great potential to amplify signal from very small amounts of hybridized material, it is necessary to alter the hybridization protocol to account for the corresponding hybridization kinetics.

4.3. Signal Amplification Post-Hybridization Another approach employed to increase sensitivity of microarrays is signal amplification at the post-hybridization stage. One such technique is tyramide signal amplification (TSA) which requires 20–100 times less RNA than direct cDNA labeling ( 40). This method was originally used in immunohistochemistry and has been an important tool for immunofluorescence microscopy (41,42). The TSA technique involves the incorporation of a biotin- or fluorescein-labeled nucleotide and its subsequent detection with conjugate reporter molecules after hybridization. The enzyme portion of the conjugate is horseradish peroxidase (HRP), which catalyzes the breakdown of tyramide that results in the deposition of numerous Cy3 labels adjacent to the immobilized HRP. Karsten et al. (40) evaluated the use of the TSA method with archived samples on microarray gene expression analysis. Although they report that the TSA method worked well to amplify the signal from frozen tissue with goodquality RNA, there were still issues of poor reproducibility and reliability especially with formalin-fixed tissue samples. RNA from fixed tissue is generally of poor quality. Partial degradation and modification by proteins or chemicals usually results in disrupted cDNA synthesis with a high rate of misincorporations and short product length. These factors influence the specificity of hybridization, and the TSA method did not reduce their effects (40). Alternatively, dendrimers (highly branched molecules) can be used to increase the amount of label per nucleotide, and subsequently per labeled cDNA/cRNA molecule (www.genisphere.com). This technology involves dendrimers, each carrying hundreds of fluorescent tags, binding to each hybridized molecule. Yu et al. (43) evaluated both the TSA and Genisphere 3DNA dendrimer labeling systems and reported that the 3DNA system was less time consuming and produced superior and more consistent results. However, hybridizations with less than 2 ␮g of RNA produced low and variable spot intensities (43).

4.4. Exponential Amplification Each of the aforementioned amplification strategies is of potential use in studying smaller, purer cell populations; however, few if any of these techniques are amenable to the study of single cells, which may be necessary if analyzing cancer stem cells. In order to study individual cells, it is necessary to have a technique that is sensitive down to the

10

Part I / Genomic Experimental Approaches in Oncology

low picogram range ( 10%) in the population and are exonic resulting in nonsynonymous coding changes, or are located in known regulatory regions, are considered to be the “smoking gun.” Most scientists would argue that synonymous SNPs that do not produce altered coding sequences, and therefore are not expected to change the function of the protein in which they occur, would not be worth follow-up studies. However, recent results demonstrate that a synonymous SNP in P-glycoprotein with similar mRNA and protein levels resulted in altered conformation of the protein ( 29). This SNP was associated with altered drug and inhibitor function, illustrating the importance of an unbiased approach.

2.4.4. I NFERENCE S TUDIES A powerful means to identify genetic variants associated with any phenotype including cytotoxicity is to perform association studies with denser markers. The HapMap trios (parents and grandparents of some large pedigrees) have millions of SNPs, but the children within those large pedigrees have microsatellites and some SNPs. One way to obtain denser markers in the offspring of the HapMap trios is to use a genotype inference method. This method combines sparse marker data from the children with high-resolution SNP genotypes from the parents to infer genotypes for the offspring. Burdick et al. inferred over 53 million SNP genotypes for 78 children in the CEPH families (30). This ultimately leads to high-density genotypes in large families and can be used in mapping studies with cytotoxicity or apoptosis as the phenotype.

2.5. Expression Studies Gene expression, a determinant of a cell’s characteristics, is another phenotype that can be studied using lymphoblastoid cell lines. Studies have shown that gene expression levels in humans differ not only among cell types within an individual, but also among individuals (16,31). As a result, there have been several recent studies that have identified polymorphic genetic variants that influence gene expression levels (15,25,32). R ArExpression data on 233 CEPH cell lines using the Human Genome Focus ray representing over 8,500 verified human sequences from the NCBI RefSeq database is publicly available (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE1485) and can be used for correlation and regression analysis with cytotoxicity and apoptosis. One use of this data to obtain information on gene expression that is important in drug sensitivity/resistance is to use vitro phenotype (e.g., cytotoxicity, apoptosis) as a continuous variable and use correlation and/or regression analyses to determine the effect of gene expression on phenotype for each probe set (4,276 probe sets after filtering). One limitation is that the Focus array does not represent the entire genome

28

Part I / Genomic Experimental Approaches in Oncology

(www.affymetrix.com/products/arrays/specific/focus.affx) and there is a 3 bias in probe design. There has also been data generated on 90 CEPH and 90 Yoruban cell lines R array. Containing roughly at the baseline using the Human Exon 1.0 ST GeneChip 1.4 million probe sets designed to represent all known exon regions within the human R genome (Build 34), this exon chip overcomes the two limitations of the earlier Focus Array (8,500 probesets and the 3 bias in probe design). This array allows exon-level profiling and the ability to interrogate each exon across the genome on a single chip. Gene expression data for the 180 HapMap lines can be correlated to phenotypes (IC50 for cytotoxicity and apoptosis). Differences among individuals has led to several studies demonstrating expression differences among populations. Recently, ethnic differences in gene expression as a complex quantitative phenotype and its regulation by polymorphic genetic variants have not been investigated comprehensively. Spielman et al. utilized a subset of human genes (∼4,200 expressed in LCLs) with samples derived from unrelated individuals from the CEU (CEPH from Utah, USA) and the CHB/JPT (Han Chinese in Beijing and Japanese in Tokyo) samples to demonstrate that cis-acting regulators may account for some of the differences in gene expression between the populations (33). Using the same platform, Storey et al. showed that 17% of genes are differentially expressed between CEU and YRI (Yoruban from Ibadan, Nigeria) in a limited set of 16 samples (34).

2.6. Integrating Different Approaches Thus, the use of complementary approaches including heritability analysis, linkage analysis, expression studies, and association studies can be used to identify and characterize novel genes important in sensitivity and resistance to chemotherapy. Ultimately, one must consider validating the genes identified in an appropriate cell or tissue (Fig. 3). Among the advantages of using linkage analysis and association in cell-based models to identify genetic variants important in sensitivity to drugs are: (1) The use of cell lines from pedigrees allows genetic strategies such as linkage analysis to be used to identify regions of the genome important in sensitivity to chemotherapy without any assumptions

Fig. 3. Identifying genes/genetic variants contributing to susceptibility to anticancer agent cytotoxicity.

Chapter 2 / Pharmacogenetic Models of Drug Cytotoxicity

29

about the nature of the genes likely to be involved; (2) if heritability is low or linkage does not result in high lod scores, global association and expression studies can be performed on the CEPH and Yoruban HapMap samples; (3) chemotherapy does not have to be given to non-affected family members for a genetic study; (4) the phenotypic effects (e.g., cytotoxicity, apoptosis) are protected from confounding variables that exist in vivo; (5) the possibility of identifying genes that were previously unknown or unrecognized is realized.

3. RELEVANCE TO DRUG TOXICITIES IN HUMANS While the use of immortalized cell lines has the obvious benefits of convenience for high-throughput analysis, ability to replicate, and avoiding exposure of toxic molecules to patients or volunteers, cell lines are not human patients. The cells only represent one of many cellular constituents of a human (B-lymphocytes), and one that is not a major concern for life-threatening toxicity. This cell lineage may not adequately represent GI tract, nerves, and the complete bone marrow milieu. The cell lines do not have a liver to evaluate pharmacokinetics nor an immune system or other dynamic processes that are important for the in vivo effects of many drugs. Indeed, most CYPs and transporters are down-regulated in CEPH cell lines and therefore will not be evaluable in the toxicity screens. However, the cell lines do offer a cell autonomous model of drug effect, in that any mechanism that is common to multiple cell lineage is likely to be represented in the CEPH cells. The argument against any given cell lineage can be made for hepatocytes and/or human cancer cell lines, both of which are a major part of the drug development process. There is also an emphasis on the pharmacodynamic mechanisms of the drug effect, because of the absence of CYP/transport members. Because a drug’s pharmacokinetics are usually adequately characterized as part of the preclinical and early clinical development, pharmacodynamics remains an area of needed research.Specific steps have been taken to try to understand how relevant the CEPH cell lines are to other cancer cell line systems. The observed CEPH population mean IC50 for both docetaxel and 5-fluorouracil was similar to IC50 values observed across the NCI60 panel of human tumor cell lines (http://dtp.nci.nih.gov) ( 17). In addition, docetaxel- and 5-fluorouracil-induced cell death is mediated by caspase-3 cleavage, similar to that observed in tumor cells ( 17). These data are encouraging for the use of CEPH pedigrees as a discovery tool. However, the ultimate proof of the value of the cell-based models will be the human validation of markers derived from this process. These studies lie ahead and will help position cell-based models in their correct place in the drug development process.

ACKNOWLEDGEMENTS The authors are supported by the Pharmacogenetics of Anticancer Agents Research (PAAR) and the Comprehensive Research on Expressed Alleles in Therapeutic Evaluation (CREATE) groups within the NIH Pharmacogenetics Research Network (GM63340, GM61393).

30

Part I / Genomic Experimental Approaches in Oncology

REFERENCES 1. Spitz MR, Wu X, Mills G. Integrative epidemiology: from risk assessment to outcome prediction. J Clin Oncol 2005;23:267–275. 2. Walgren RA, Meucci MA, McLeod HL. Pharmacogenomic discovery approaches: will the real genes please stand up? J Clin Oncol 2005;23:7342–7349. 3. Shukla SJ, Dolan ME. Use of CEPH and non-CEPH lymphoblast cell lines in pharmacogenetic studies. Pharmacogenomics 2005;6:303–310. 4. Meucci MA, Marsh S, Watters JW et al. CEPH individuals are representative of the European American population: implications for pharmacogenetics. Pharmacogenomics 2005;6:59–63. 5. Dausset J, Cann H, Cohen D et al. Centre d’Etude du Polymorphisme Humain (CEPH): collaborative genetic mapping of the human genome. Genomics 1990;6:575–577. 6. Sugimoto M, Tahara H, Ide T et al. Steps involved in immortalization and tumorigenesis in human B-lymphoblastoid cell lines transformed by Epstein–Barr virus. Cancer Res 2004;64:3361–3364. 7. Miller G, Lipman M. Release of infectious Epstein–Barr virus by transformed marmoset leukocytes. Proc Natl Acad Sci USA 1973;70:190–194. 8. Henle W, Diehl V, Kohn G et al. Herpes-type virus and chromosome marker in normal leukocytes after growth with irradiated Burkitt cells. Science 1967;157:1064–1065. 9. Gipps EM, Kidson C. Cellular radiosensitivity: expression of an MS susceptibility gene? Neurology 1984;34:808–811. 10. Imray FP, Smith PJ, Relf W et al. Wilms’ tumour: association with cellular sensitivity to mitomycin C in patients and first-degree relatives. Lancet 1984;1:1148–1151. 11. Poot M, Gollahon KA, Rabinovitch PS. Werner syndrome lymphoblastoid cells are sensitive to camptothecin-induced apoptosis in S-phase. Hum Genet 1999;104:10–14. 12. Cloos J, Reid CB, van der Sterre ML et al. A comparison of bleomycin-induced damage in lymphocytes and primary oral fibroblasts and keratinocytes in 30 subjects. Mutagenesis 1999;14:87–93. 13. Jen KY, Cheung VG. Transcriptional response of lymphoblastoid cells to ionizing radiation. Genome Res 2003;13:2092–2100. 14. Schork NJ, Gardner JP, Zhang L et al. Genomic association/linkage of sodium lithium countertransport in CEPH pedigrees. Hypertension 2002;40:619–628. 15. Morley M, Molony CM, Weber TM et al. Genetic analysis of genome-wide variation in human gene expression. Nature 2004;430:743–747. 16. Cheung VG, Conlin LK, Weber TM et al. Natural variation in human gene expression assessed in lymphoblastoid cells. Nat Genet 2003;33:422–425. 17. Watters JW, Kraja A, Meucci MA et al. Genome-wide discovery of loci influencing chemotherapy cytotoxicity. Proc Natl Acad Sci USA 2004;101:11809–11814. 18. Dolan ME, Newbold KG, Nagasubramanian R et al. Heritability and linkage analysis of sensitivity to cisplatin-induced cytotoxicity. Cancer Res 2004;64:4353–4356. 19. Huang RS, Kistner EO, Bleibel WK et al. Effect of population and gender on chemotherapeutic agentinduced cytotoxicity. Mol Cancer Ther 2007;6:31–36. 20. Cheung VG, Spielman RS, Ewens KG et al. Mapping determinants of human gene expression by regional and genome-wide association. Nature 2005;437:1365–1369. 21. Correa CR, Cheung VG. Genetic variation in radiation-induced expression phenotypes. Am J Hum Genet 2004;75:885–890. 22. The International HapMap Consortium. A haplotype map of the human genome. Nature 2005;437:1299–1320. 23. Brem RB, Yvert G, Clinton R et al. Genetic dissection of transcriptional regulation in budding yeast. Science 2002;296:752–755. 24. Lo HS, Wang Z, Hu Y et al. Allelic variation in gene expression is common in the human genome. Genome Res 2003;13:1855–1862. 25. Cheung VG. Polymorphic landscape of the human genome. Eur J Hum Genet 2005;13:133–135. 26. Lewis CM. Genetic association studies: design, analysis and interpretation. Brief Bioinform 2002;3:146–153.

Chapter 2 / Pharmacogenetic Models of Drug Cytotoxicity

31

27. Cardon LR, Bell JI. Association study designs for complex diseases. Nat Rev Genet 2001;2:91–99. 28. Spielman RS, Ewens WJ. The TDT and other family-based tests for linkage disequilibrium and association. Am J Hum Genet 1996;59:983–989. 29. Kimchi-Sarfaty C, Oh JM, Kim IW et al. A “silent” polymorphism in the MDR1 gene changes substrate specificity. Science 2007;315:525–528. 30. Burdick JT, Chen WM, Abecasis GR et al. In silico method for inferring genotypes in pedigrees. Nat Genet 2006;38:1002–1004. 31. Schadt EE, Monks SA, Drake TA et al. Genetics of gene expression surveyed in maize, mouse and man. Nature 2003;422:297–302. 32. Stranger BE, Forrest MS, Clark AG et al. Genome–wide associations of gene expression variation in humans. PLoS Genet 2005;1:e78. 33. Spielman RS, Bastone LA, Burdick JT et al. Common genetic variants account for differences in gene expression among ethnic groups. Nat Genet 2007;39:226–231. 34. Storey JD, Madeoy J, Strout JL et al. Gene-expression variation within and among human populations. Am J Hum Genet 2007;80:502–509.

“This page left intentionally blank.”

3

Proteomic Analysis in Cancer Patients Yasuhiro Kuramitsu, MD, PhD, and Kazuyuki Nakamura, MD, PhD CONTENTS Introduction Materials Methods Res ults from Our Study Res ults from the Literature Potential Application in Oncology References

S UMMARY The term proteome means the total protein complement of a genome, and proteomics means the analysis for proteome. The combination of two-dimensional gel electrophoresis (2-DE) and mass spectrometry (MS) is a proteomic method of highthroughput analysis of protein expression. By using this 2-DE and MS, proteomic studies have identified many proteins that may be involved in the pathogenic mechanism of cancers. These studies analyzed cancer cell lines, as well as cancer tissues or serum from patients. In the present study, we analyzed proteome in hepatocellular carcinoma (HCC), esophageal cancer, and pancreatic cancer tissues. We identified many proteins whose expression in cancer tissues was different from corresponding non-cancerous tissues by using 2-DE and MS. Furthermore, we identified some auto-antibodies reacting to proteins in HCC cancer tissues. In this chapter, we will describe the method, our experimental result, and reports from other researchers about proteomic analysis in cancer patients. From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 3, 

33

34

Part I / Genomic Experimental Approaches in Oncology

Key Words: Proteomics; hepatocellular carcinoma; HCC; esophageal cancer; pancreatic cancer; auto-antibody

1. INTRODUCTION The term proteome comes from the study of gene-product mapping performed by the Ian Humphery-Smith and his group (1). They defined proteome as the total protein complement of a genome. Proteomics thus refers to the analysis of the total protein complement of a genome. Proteomics was the first of several succeeding developments of various studies such as transcriptomics, glycomics, and so on. The combination of two-dimensional gel electrophoresis (2-DE) and mass spectrometry (MS) is an effective method of high-throughput proteomics. The technique of 2-DE is able to differentiate proteins according to both their charges in isoelectoric focusing (IEF) gels and their size in sodium dodecyl sulfate (SDS) gels (2). The 2-DE technique has unique advantages for examining the expressions of hundreds of proteins simultaneously and also examining post-translational modifications of the protein spots. There are many reports of proteomics of diseases including cancer obtained by means of 2-DE (3,4,5). Recently, MS has become the first choice for determining sequences of proteins instead of the Edman method. By means of MS we can determine the masses of peptides with much accuracy, and from the huge database we can identify the protein (peptide mass fingerprinting) (6). Many proteomic studies have identified diverse proteins that may be involved in the pathogenic mechanism of disease and which may be disease markers. Most of the applications use expression proteomics to determine expression profiles of proteins in cells and tissues in normal or disease states. In the present study, we analyzed protein expression in cancer tissue samples and corresponding non-cancerous tissue samples to find proteins that might be involved in carcinogenesis or pathogenesis. Hepatocellular carcinoma (HCC) is the fifth most common and the third most deadly cancer. One million patients with HCC die each year. One of the major causes of HCC is infection with the hepatitis C virus (HCV), and the pathogenesis of HCV-related HCC in its incipient stage from the infection to the onset of cancer is being researched. Esophageal cancer is the sixth leading cause of cancer death in Japan with a very high mortality rate. Many patients die within 1 year after diagnosis, and the 5-year survival rate is less than 10%. Pancreatic cancer is a cancer with a poor prognosis, having the lowest 5-year survival rate (7). To find tumor biomarkers or therapeutic drugs, many investigators are performing proteomic studies of cancer tissues and cancer cell lines, and the data are being accumulated (8,9). In this chapter, we will introduce the proteomics for HCC, pancreatic cancer, and esophageal cancer tissues performed by us, and also show the result of proteomic analysis of auto-antibodies in patients.

2. MATERIALS 2.1. Tissue Samples Pairs of cancer tissues and adjacent non-cancerous tissues were obtained from patients who were diagnosed with cancer and underwent surgical organ resection. The histological diagnosis of cancer was made by formalin-fixed, paraffin-embedded tissues,

Chapter 3 / Proteomic Analysis in Cancer Patients

35

according to the World Health Organization criteria after surgery in all cases. The study protocol was approved by the Institutional Review Board for Human Use at the Yamaguchi University School of Medicine.

2.2. Serum Samples The autologous serum samples were obtained from each patient whose tissue was applied in this study. Sera from patients with normal liver tissues were also obtained as control sera.

3. METHODS 3.1. Overview of Experimental Procedure 3.1.1. P ROTEOMIC A NALYSIS OF C ANCER T ISSUES To analyze the differences of protein expression profiles between cancer tissues and corresponding non-cancerous tissues, the proteomic differential display method was used. In this method, 2-DE and MS were used to identify the proteins. We first separated proteins from cancer tissues and corresponding non-cancerous tissues by 2-DE. Then the protein spots of the samples from cancer tissues were compared to the spots of the samples from corresponding non-cancerous tissues by using software for proteomic differential display. Particular protein spots for which expression was different between cancerous and non-cancerous tissues were cut out from the gels, and were identified by means of MS. Figure 1 shows a workflow of these methods. The proteomic differential display method is a common method to analyze the profiling of protein expression. Usually 2-DE and MS have been used. Recently, however, other new developments in protein detection technologies such as fluorescence two-dimensional difference gel electrophoresis (2-D-DIGE), isotope-coded affinity tags (ICAT), and isobaric tag for relative and absolute quantitation (iTRAQ) have been used. The 2-DE makes it possible to separate proteins according to both their charge in isoelectoric focusing (IEF) gels and their molecular weight in sodium dodecyl sulfate (SDS) gels (10,11). The 2-DE technique has advantageous characteristics to compare the expression of a huge number of proteins simultaneously and to examine post-translational modifications of the protein spots. The technique of 2-D-DIGE makes it possible to run two or three differently labeled protein samples on the same gel simultaneously. It can exclude variability existing among gels to run two or three protein samples on the same gel ( 12, 13, 14, 15). Isotope-coded affinity tag (ICAT) labeling, a new quantitative method, was recently performed ( 16). ICAT makes it possible to comprehensively analyze two comparable samples immediately. After labeling protein samples with isotope tags of 12 C or 13 C, they are separated by HPLC and identified by MS. iTRAQ is a recently developed LC-based protein quantitation technique that utilizes four isobaric amine-specific tags. The principal advantages of iTRAQ are that

36

Part I / Genomic Experimental Approaches in Oncology

Fig. 1. Workflow of proteomic differential display methods.

four samples can be analyzed simultaneously, thereby reducing the amount of MS time needed for analysis, and it is more sensitive than other techniques (17).

3.1.2. P ROTEOMIC A NALYSIS OF AUTO - ANTIBODIES IN PATIENTS To detect auto-antibodies expected as cancer biomarkers for HCC, we analyzed serum auto-antibodies immunoreacting to proteins in cancer tissue obtained from patients with HCC. Tissue proteins were separated by 2-DE, transferred onto PVDF membranes, and immunoblotted with autologous sera. By comparing each immunoblot pattern, we identified immunoreactive spots with stronger staining intensity in cancer tissues than in corresponding non-cancerous tissues. Matched proteins on 2-DE gels were identified by MS. Figure 1 shows a workflow of these methods.

3.2. Preparation of Tissue Samples Tissues were homogenized in a lysis buffer [50 mM Tris-HCl, pH 7.5, 165 mM NaCl, 10 mM NaF, 1 mM sodium vanadate, 1 mM phenylmethylsulfonyl fluoride (PMSF), 10 mM EDTA, 10 ␮g/ml aprotinin, 10 ␮g/ml leupeptin, and 1% NP40]. After centrifugation with 15,000 g for 30 min at 4◦ C, the supernatants were taken and stored at –80◦ C until use.

Chapter 3 / Proteomic Analysis in Cancer Patients

37

3.3. Two-Dimensional Gel Electrophoresis 3.3.1. I SOELECTRIC F OCUSING Three-hundred micrograms of each tissue sample were applied for the firstdimensional isoelectric focusing (IEF) on 7-cm, immobilized, pH 3–10 linear gradient strips (Bio-Rad, Hercules, CA) at 20◦ C and a 50 ␮A/strip. Isoelectric focusing (IEF) was run in three steps: 500 V for 1 hr, 1000 V for 1 hr, and 8000 V for 3 hr. 3.3.2. S ODIUM D ODECYL S ULFATE -P OLYACRILAMIDE G EL E LECTROPHORESIS (SDS-PAGE) The second-dimensional run was performed on precast polyacrylamide gels (2-D homogeneous 12.5; GE Healthcare) in two steps: (i) 600 V, 20 mA for 30 min, and (ii) 600 V, 50 mA for 70 min. After electrophoresis, the blotting gels were used for 2-D immunoblotting, and the staining gels were stained with CBB R250 (Nacalai Tesque, Kyoto, Japan) for 24 hr. Gels were destained with 10% acetic acid in water containing 30% methanol for 30 min and then with 7% acetic acid and used for in-gel digestion. 3.3.3. T WO -D IMENSIONAL I MMUNOBLOT A NALYSIS For the blotting-gels, fractionated proteins were transferred electrophoretically onto PVDF membranes (Immobilon-P, Millipore Corporation, Bedford, MA), and the membranes were blocked for 1 hr at 4◦ C with TBS containing 1% skim milk. They were subsequently incubated overnight at 4◦ C with autologous serum (1:50 dilution), washed four times with TBS containing 0.05% Tween 20, and incubated for 1 hr at 4◦ C with horseradish peroxidase-conjugated secondary antibody (1:1000, #55256; ICN Pharmaceuticals, Aurora, OH). The reaction was visualized with a chemiluminescence reagent (ECL; Amersham Biosciences, Buckinghamshire, UK).

3.4. Mass Spectrometry Analysis 3.4.1. I N - GEL D IGESTION For staining gels, CBB dye was removed by rinsing three times in 60% methanol, 50 mM ammonium bicarbonate, and 5 mM dithiothreitol (DTT) for 15 min and twice in 50% acetonitrile, 50 mM ammonium bicarbonate, and 5 mM DTT for 10 min. Gel pieces were dehydrated three times in 100% acetonitrile for 30 min and then rehydrated in an ingel digestion reagent containing 10 ␮g/ml sequencing-grade modified trypsin (Promega, Madison, WI) in 30% acetonitrile, 50 mM ammonium bicarbonate, and 5 mM DTT. In-gel digestion were performed overnight at 30◦ C. The samples were rinsed in 30% acetonitrile, 50 mM ammonium bicarbonate, and 5 mM DTT for 2 hr and lyophilized overnight at –30◦ C. 3.4.2. A MINO ACID S EQUENCING BY L IQUID C HROMATOGRAPHY –TANDEM M ASS S PECTROMETRY (LC-MS/MS) Lyophilized samples were dissolved in 20 ␮l 0.1% formic acid and centrifuged at 15,000 g for 5 min. Peptide sequencing of identified protein spots was performed by

38

Part I / Genomic Experimental Approaches in Oncology

LC-MS/MS with a Spectrum Mill MS Proteomics Workbench (Agilent Technologies, Santa Clara, CA).

4. RESULTS FROM OUR STUDY 4.1. Proteomics for HCC Tissues from Patients The differential expression of paired cancerous and non-cancerous tissues was visually compared; 12 up-regulated and 11 down-regulated spots were identified. The up-regulated proteins were identified as GRP75, GRP78, HSC71, HSP70.1, HSP60, glutamine synthetase (GS), triosephosphate isomerase (TIM), ATP synthetase beta chain, and alpha-enolase. The three protein spots of Mr 42,000 and pI 6.4–6.8 were identified as GS. The down-regulated proteins were identified as aldolase, arginase 1, enoyl-CoA hydratase, ketohexokinase, smoothelin, tropomyosin beta chain, ferritin light chain, serum albumin (18,19,20,21).

4.2. Proteomics for Esophageal Cancer Tissues from Patients The differential expression of paired cancerous and non-cancerous tissues was visually compared, and three up-regulated and eight down-regulated spots were identified. The up-regulated proteins were identified as tropomyosin alpha-4 chain, transgelin, and pyruvate kinase. The down-regulated proteins were identified as serum albumin precursor, annexin A1 (2 spots), tropomyosin beta chain, 14-3-3 protein sigma, and serotransferrin precursor (3 spots)(22).

4.3. Proteomics for Pancreatic Cancer Tissues from Patients The differential expression of paired cancerous and non-cancerous tissues was visually compared, and 11 spots were up-regulated in cancerous tissues: alphaenolase, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) 2 (spots), triosephosphate isomerase, transgelin, calmodulin, MnSOD, PDI A3, cyclophilinA, GST-P, and apolipoprotein A-I precursor (23).

4.4. Proteomics for Auto-antibodies in Serum from HCC Patients Immunoreactivity of autologous serum auto-antibodies to tissue proteins was assessed in samples of HCC cancer tissues and corresponding non-cancerous tissues. Four immunoreactive spots were detected that showed increased intensity in cancer tissues compared to that of non-cancerous tissues. Each of the spots was matched to an equivalent spot on staining gels. To identify these four immunoreactive proteins, the spots were digested, and were identified by MS as HSP70, glyceraldehyde 3-phosphate dehydrogenase (GAPDH), peroxiredoxin, and Mn-SOD (24).

5. RESULTS FROM THE LITERATURE 5.1. Proteomics for HCC Tissues from Patients Table 1 shows the proteins up-regulated or down-regulated in HCC tissues. The upregulated proteins in HCC tissues have been identified. Park et al. identified aldehyde

Chapter 3 / Proteomic Analysis in Cancer Patients

39

Table 1 Increased or Decreased Proteins in HCC Tissues from Patients Increased

Decreased

hARLP-1, -2, -3, -4 Aflatoxin B1 aldehyde reductase 2 Aldehyde dehydrogenase 3 Alpha-enolase 4-Aminobutyrate aminotransferase Annexin A2 Annexin V APC-binding protein EB1 ATP synthase ATP synthase beta chain Betaine-himocysteine S-methyltransferase eIF-5a Chloride intracellular channel protein I Dihydropipoamide dehydrogenase Elongation factor 2 Enolase 1 E-FABP Fatty acid binding protein Fibrinogen beta chain Galactokinase 1 Glutamine synthetase GRP 75 GRP 78 GRP 94 HSC70 HSC 71 HSP 27 HSP 60 HSP 70 RY HSP 70.1 HSP 70.5 HSP 90-alpha HSP gp96 hnRNP K Lactoylglutathione lyase Lamin B1

Albumin Aldehyde dehydrogenase Aldehyde dehydrogenase 2 Aldolase Aldolase B Alpha-enolase Annexin V Arginase Argininosuccinate synthase Carbamoyl-phosphate synthase Catalase Cathepsin B1 Cathepsin D Cytochrome B5 Cytosol aminopeptidase Enoyl-CoA hydratase Fatty acid-binding protein Ferritin light chain Formininotransferase-cyclodeaminase Fructose-bisphosphatase Fumarylacetoacetase GAPDH Galectin-1 Glutamate dehydrogenase 1 Glutathione peroxidase Growth factor receptor-bound protein 2 Guanidinoacetate N-methyltransferase HSP 27 Ketohexokinase Liver carboxyesterase Mn-SOD Nucleophosmin Peroxiredoxin 3 Phenol-sulfating phenol sulfotransferase 1 Phosphatidylethanolamine-binding protein (PEBP) Protein disulfide isomerase P5C dehydrogenase Ribosome binding protein

Nucleophosmin Nucleoside diphosphate kinase A PCNA

(Continued)

40

Part I / Genomic Experimental Approaches in Oncology Table 1 (Continued)

Increased

Decreased

Phosphoglycerate mutase 1 Plasma retinal-binding protein precursor Protein disulfide isomerase A3 Protein disulfide isomerase ER-60 Receptor of activated protein kinase C 40S ribosomal protein SA Stathmin Thiosulfate sulfurtransferase Transitional endoplasmic reticulum ATPase Triosephosphate isomerase (TIM) Tropomyosin 3 Tubulin beta-1 chain Ubiquitin carboxyl-terminal hydrolase 5 Ubiquinol-cytochrome C reductase complex core protein I Vimentin

Sarcosine dehydrogenase Senescence marker protein-30 Serotransferrin precursor Smoothelin SOD 1 Tropomyosin beta chain Triosephosphate isomerase (TIM) Vimentin

dehydrogenase 3 ( 25). Kim et al. identified protein disulfide isomerase A3 ( 26). Lim et al. identified HSP gp96, HSP 90-alpha, and others (27). Li et al. identified stathmin, PCNA, and others (28). Fujii et al. identified APC-binding protein, EB1 and others (29). Zeindl-Eberhart et al. identified human aldose reductase–like protein isoforms (30). Kim et al. identified GRP 94, nucleophosmin, and others (31). The down-regulated proteins in HCC tissues have been identified. Park et al. identified aldehyde dehydrogenase 2 (25) and ferritin light chain (32). Kim et al. identified HSP 27, cathepsin D, and others ( 26). Lim et al. identified cytochrome B5, liver carboxyesterase, and others (27). Li et al. identified SOD 1, aldolase B, and others (28). Fujii et al. identified galectin-1 ( 29). Kim et al. identified argininosuccinate synthase, carbamoyl-phosphate synthase, and others (31). Table 1 shows the summary of the proteins whose expression was different between HCC cancer tissues and non-cancerous tissues.

5.2. Proteomics for Esophageal Cancer Tissues from Patients Table 2 shows the proteins up-regulated or down-regulated in ESCC tissues. Zhou et al. found that the expression of gp96, a tumor rejection antigen, increased and that the expression of annexin I decreased in ESCC tissues (33). Emmert-Buck et al. found that cytokeratin 1 increased and annexin 1 decreased in ESCC tissues (34). Zhang et al. reported loss of clustering, both in sera and tissues, which correlated with the tumorigenesis of ESCC (35). Zhou et al. also observed increases in the expressions of PCNA,

Chapter 3 / Proteomic Analysis in Cancer Patients

41

Table 2 Increased or Decreased Proteins in Esophageal Cancer Tissues from Patients Increased

Decreased

Actin (cytoplasmic 1) AKR family 1 Cytokeratin 1 Alpha enolase Elongation factor Tu eIF-1A Fascin GAPDH GST M2 gp96 Isocitrate dehydrogenase Keratin 1 Mn-SOD Neuronal protein PCNA Peroxiredoxin 1 Prohibitin Prosomal protein p30-33 k Proteasome subunit beta type 4 14-3-3 protein sigma Pyruvate kinase, M1 isozyme RNA binding motif protein 8A Reticulocalbin Thioredoxin peroxidase Transgelin (SM22-alpha) Transmembrane protein 4 Tropomyosin alpha 4 chain Tubulin alpha-1 chain Tubulin beta-5 chain Ubiquitin C-terminal esterase

Annexin I Alpha B crystalline Clusterin Desmin Galectin-7 Fatty acid-binding protein HSP27 Keratin 8 Keratin 13 Peroxiredoxin 2 isoform Proteinase inhibitor, Clade B Proteasome subunit beta type 9 S100 A9 SCCA1 Serotransferrin precursor Serum albumin precursor Stratifin TGase 3 Transgelin Tropomyosin Tropomyosin beta chain Tropomyosin isoform 1

RNA binding motif protein 8A, and others in ESCC tissues. Table 2 shows the summary of the proteins whose expression was different between esophageal cancer tissues and non-cancerous tissues.

5.3. Proteomics for Pancreatic Cancer Tissues from Patients Table 3 shows the proteins up-regulated or down-regulated in pancreatic cancer tissues. Shen et al. reported that, in pancreatic cancer tissues, expressions of Mn-SOD, S100A8, annexin A4, cathepsin D, 14-3-3 zeta, tropomyosin 2, actin, ferritin light chain, alpha-enolase, galectin-1, and cyclophilin A increased, and that those of peroxiredoxin

Part I / Genomic Experimental Approaches in Oncology

42

Table 3 Increased or Decreased Proteins in Pancreatic Cancer Tissues from Patients Increased

Decreased

Actin, beta and gamma Annexin A4 Apo A-I precursor Alpha-enolase Calmodulin Cathepsin D Cyclophilin A Ferritin light subunit GAPDH Galectin-1 Glutathione S-transferase MnSOD Protein disulfide isomerase A3 precursor S100A8 (Calgranulin A) Transgelin Triosephosphate isomerase Tropomyosin 2 14-3-3 zeta

Carboxypeptidase A1 Carboxypeptidase A2 DJ-1 HSC54 Neuropolypeptide h3 Peroxiredoxin II

II, DJ-1, HSC 54, carboxypeptidase A1, carboxypeptidase A2, and neuropolypeptide h3 decreased ( 36). Table 3 shows the summary of the proteins whose expression was different between pancreatic cancer tissues and non-cancerous tissues.

5.4. Proteomics for Autoantibodies in Serum from Cancer Patients Fujita et al. identified auto-antibodies reacting to peroxiredoxin VI in the sera of esophageal cancer patients (37). Hong et al. identified auto-antibodies reacting to calreticulin isoforms in the sera of pancreatic cancer patients (38). Le Naour et al. identified auto-antibodies reacting to Crt 32 in the sera of pancreatic cancer patients (39).

6. POTENTIAL APPLICATION IN ONCOLOGY The final goal of clinical cancer proteomics is to discover biomarkers for early small cancers, to clarify pathogenesis and carcinogenesis of cancers, and to tailor the therapy for the cancer patients. Although a huge number of reports about the discovery of biomarkers by proteomics has been published, still no biomarker is being used practically in hospitals. However, some kinds of proteins detected in sera from cancer patients are expectable as a biomarker (24,37,38,39).

Chapter 3 / Proteomic Analysis in Cancer Patients

43

Although results from the attempt at biomarker-discovery have not been entirely successful, the proteomic approach for the clarification of pathogenesis and carcinogenesis has contributed to our knowledge ( 8, 19, 20, 21, 33, 34). In fact, our proteomic analysis for HCC tissues has suggested an increased glycolytic pathway and decreased lipid metabolism from the results, showing an increased expression of phosphoglycerate mutase 1, triosephosphate isomerase, and alpha-enolase, and decreased expression of aldolase, enoyl-CoA hydratase, and ketohexokinase (8,19,20,21). Pharmacogenomic studies have led to many results that can be used to tailor therapy for cancer patients. Many reports have showed the identification of proteins concerned with drug-resistance proteomically (40,41,42,43,44). In fact, we identified HSP27 as a protein related to resistance to gemcitabine, which is the only effective anti-cancer drug against pancreatic cancer (45,46)

REFERENCES 1. Wasinger VC, Cordwell SJ, Cerpa–Poljak A et al. Progress with gene–product mapping of the Mollicutes: Mycoplasma genitalium. Electrophoresis 1995;16:1090–1094. 2. O’Farrell PH. High resolution two-dimensional electrophoresis of proteins.J Biol Chem 1975; 250:4007–4021. 3. Simpson RJ, Connolly LM, Eddes JS et al. Proteomic analysis of the human colon carcinoma cell line (LIM 1215): development of a membrane protein database. Electrophoresis 2000;21:1707–1732. 4. Ha GH, Lee SU, Kang DG et al. Proteome analysis of human stomach tissue: separation of soluble proteins by two-dimensional polyacrylamide gel electrophoresis and identification by mass spectrometry. Electrophoresis 2002;23:2513–2524. 5. Steel LF, Mattu TS, Mehta A et al. A proteomic approach for the discovery of early detection markers of hepatocellular carcinoma. Dis Markers 2001;17:179–189. 6. Aebersold R, Mann M. Mass spectrometry–based proteomics. Nature 2003;422:198–207. 7. Vital Statistics of Japan. Statistics and Information Department, Ministers Secretariat, Ministry of Health, Labor and Welfare of Japan, 2003:25. 8. Kuramitsu Y, Nakamura K. Current progress in proteomic study of hepatitis C virus–related human hepatocellular carcinoma. Expert Rev Proteomics 2005;2:589–601. 9. Kuramitsu Y, Nakamura K. Proteomic analysis of cancer tissues: shedding light on carcinogenesis and possible biomarkers. Proteomics 2006:6:5650–5661. 10. O’Farrell PH. High-resolution two-dimensional electrophoresis of proteins. J Biol Chem 1975: 250:4007–4021. 11. Latner AL, Marshall T, Gambie M. A simplified technique of high-resolution two-dimensional electrophoresis: serum immunoglobulins. Clin Chim Acta 1980;103:51–59. 12. Unlu M., Morgan ME, Minden JS. Difference gel electrophoresis: a single-gel method for detecting changes in protein extracts. Electrophoresis 1997:18:2071–2077. 13. Tonge R, Shaw J, Middleton B et al. Validation and development of fluorescence two-dimensional differential gel electrophoresis proteomics technology. Proteomics 2001:1:377–396. 14. Shaw J, Rowlinson R, Nickson J et al. Evaluation of saturation labelling two-dimensional difference gel electrophoresis fluorescent dyes. Proteomics 2003:3;1181–1195. 15. Kondo T, Seike M, Mori Y et al. Application of sensitive fluorescent dyes in linkage of laser microdissection and two-dimensional gel electrophoresis as a cancer proteomic study tool. Proteomics 2003;3:1758–1766. 16. Gygi SP, Rist B, Gerber SA et al. Quantitative analysis of complex protein mixtures using isotopecoded affinity tags. Nat Biotechnol. 1999:17;994–999. 17. Wu WW, Wang G, Baek SJ et al. Comparative study of three proteomic quantitative methods, DIGE, cICAT, and iTRAQ, using 2D gel- or LC-MALDI TOF/TOF. J Proteome Res 2006;5:651–658.

44

Part I / Genomic Experimental Approaches in Oncology

18. Kuramitsu Y, Harada T, Takashima M et al. Increased expression, and phosphorylation of liver glutamine synthetase in well-differentiated hepatocellular carcinoma tissues of patients infected with hepatitis C virus. Electrophoresis 2006;27:1651–1658. 19. Takashima T, Kuramitsu Y, Yokoyama Y et al. Overexpression of alpha-enolase in hepatitis C virus– related hepatocellular carcinoma: association with tumor progression as determined by proteomic analysis. Proteomics 2005;5:1686–1692. 20. Yokoyama Y, Kuramitsu Y, Takashima T et al. Proteomic profiling of proteins decreased in hepatocellular carcinoma from patients infected with hepatitis C virus. Proteomics 2004:4:2111–2116. 21. Takashima T, Kuramitsu Y, Yokoyama Y et al. Proteomic profiling of heat shock protein 70 family members as biomarkers for hepatitis C virus–related hepatocellular carcinoma. Proteomics 2003:3: 2487–2493. 22. Harada T, Kuramitsu Y, Makino A et al. Expression of tropomyosin alpha-4 chain is increased in esophageal squamous cell carcinoma as evidenced by proteomic profiling by two-dimensional electrophoresis and liquid chromatography–mass spectrometry/mass spectrometry. Proteomics Clinical Applications 2007;1:215–223. 23. Mikuriya K, Kuramitsu Y, Ryozawa S et al. Expression of glycolytic enzymes is increased in pancreatic cancerous tissues as evidenced by proteomic profiling by two-dimensional electrophoresis and liquid chromatography–mass spectrometry/mass spectrometry. Int J Oncol 2007;30:849–855. 24. Takashima T, Kuramitsu Y, Yokoyama Y et al. Proteomic analysis of auto-antibodies in patients with hepatocellular carcinoma. Proteomics 2006;6:3894–3900. 25. Park KS, Cho SY, Kim H et al. Proteomic alterations of the variants of human aldehyde dehydrogenase isozymes correlate with hepatocellular carcinoma. Int. J. Cancer 2002;97:261–265. 26. Kim J, Kim SH, Lee SU et al. Proteome analysis of human liver tumor tissue by two-dimensional gel electrophoresis and matrix-assisted laser desorption/ionization–mass spectrometry for identification of disease-related proteins. Electrophoresis 2002;23:4142–4156. 27. Lim SO, Park SJ, Kim W et al. Proteome analysis of hepatocellular carcinoma. Biochem Biophys Res Commun 2002;291:1031–1037. 28. Li C, Hong Y, Tan YX et al. Accurate qualitative and quantitative proteomic analysis of clinical hepatocellular carcinoma using laser capture microdissection coupled with isotope-coded affinity tag and two-dimensional liquid chromatography mass spectrometry. Mol Cell Proteomics 2004;3:399–409. 29. Fujii K, Kondo T, Yokoo H et al. Proteomic study of human hepatocellular carcinoma using two-dimensional difference gel electrophoresis with saturation cysteine dye. Proteomics 2005;5: 1411–1422. 30. Zeindl-Eberhart E, Haraida S, Liebmann S et al. Detection and identification of tumor-associated protein variants in human hepatocellular carcinomas. Hepatology 2004;39:540–549. 31. Kim W, Oe Lim S, Kim JS et al. Comparison of proteome between hepatitis B virus– and hepatitis C virus–associated hepatocellular carcinoma. Clin. Cancer Res 2003;9:5493–5500. 32. Park KS, Kim H, Kim NG et al. Proteomic analysis and molecular characterization of tissue ferritin light chain in hepatocellular carcinoma. Hepatology 2002;35:1459–1466. 33. Zhou G, Li H, DeCamp D et al. 2D differential in-gel electrophoresis for the identification of esophageal scans cell cancer-specific protein markers. Mol Cell Proteomics 2002;1:117–123. 34. Emmert-Buck MR, Gillespie JW, Paweletz CP et al. An approach to proteomic analysis of human tumors. Mol Carcinog 2000;27:158–165. 35. Zhang L, Ying W, Mao Y et al. Loss of clusterin both in serum and tissue correlates with the tumorigenesis of esophageal squamous cell carcinoma via proteomics approaches. World J Gastroenterol 2003;9:650–654. 36. Shen J, Person MD, Zhu J et al. Protein expression profiles in pancreatic adenocarcinoma compared with normal pancreatic tissue and tissue affected by pancreatitis as detected by two-dimensional gel electrophoresis and mass spectrometry. Cancer Res 2004;64:9018–9026. 37. Fujita Y, Nakanishi T, Hiramatsu M et al. Proteomics-based approach identifying auto-antibody against peroxiredoxin VI as a novel serum marker in esophageal squamous cell carcinoma. Clin Cancer Res 2006;12:6415–6420.

Chapter 3 / Proteomic Analysis in Cancer Patients

45

38. Hong S, Misek DE, Wang H et al. An autoantibody-mediated immune response to calreticulin isoforms in pancreatic cancer. Cancer Res 2004;64:5504–5510. 39. Le Naour F, Brichory F, Misek DE et al. A distinct repertoire of autoantibodies in hepatocellular carcinoma identified by proteomic analysis. Mol Cell Proteomics 2002;1:197–203. 40. Hasegawa N, Mizutani K, Suzuki T et al. A comparative study of protein profiling by proteomic analysis in camptothecin-resistant PC3 and camptothecin-sensitive LNCaP human prostate cancer cells. Urol Int 2006;77:347–354. 41. Strong R, Nakanishi T, Ross D et al. Alterations in the mitochondrial proteome of adriamycin-resistant MCF-7 breast cancer cells. J Proteome Res 2006;5:2389–2395. 42. Le Moguen K, Lincet H, Deslandes E et al. Comparative proteomic analysis of cisplatin-sensitive IGROV1 ovarian carcinoma cell line and its resistant counterpart IGROV1–R10. Proteomics 2006;6:5183–5192. 43. Smith L, Lind MJ, Welham KJ et al. Cancer proteomics and its application to discovery of therapy response markers in human cancer. Cancer 2006;107:232–241. 44. Liu Y, Liu H, Han B et al. Identification of 14-3-3sigma as a contributor to drug resistance in human breast cancer cells using functional proteomic analysis. Cancer Res 2006;66:3248–3255. 45. Mori-Iwamoto S, Kuramitsu Y, Ryozawa S et al. A proteomic profiling of gemcitabine resistance in pancreatic cancer cell lines. Molecular Medicine Reports 2008;1:429–434. 46. Mori-Iwamoto S, Kuramitsu Y, Ryozawa S et al. Proteomics finding heat shock protein 27 as a biomarker for resistance of pancreatic cancer cells to gemcitabine. International Journal of Oncology 2007;31:1345–1350.

“This page left intentionally blank.”

4

MicroRNAs and Discovery of New Targets Soken Tsuchiya, PharmD, Yasushi Okuno, PharmD, and Gozoh Tsujimoto, MD CONTENTS Introduction Biogenes is of miRNAs The Role of miRNAs in Cancer: Diagnos is and Drug Dis covery Pers pective Acknowledgements References

S UMMARY MicroRNAs are endogenous short non-coding RNAs that regulate gene expression mainly at the post-transcriptional level by base pairing to the 3 untranslated region of target messenger RNAs. At present, hundreds of microRNAs have been identified in humans, and some of them have been revealed to play a critical role especially in the initiation, progression, and malignant potential of various cancers. In this chapter, we discuss the role of microRNAs in cancer and its potential application for cancer therapy. Key Words: MicroRNA; non-coding RNA; translational suppression; cancer; oncogene; tumor suppressor gene; diagnosis; antisense oligonucleotide; drug discovery

From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 4, 

47

48

Part I / Genomic Experimental Approaches in Oncology

1. INTRODUCTION MicroRNAs (miRNAs) are evolutionarily conserved “non-coding RNA” molecules (∼22 nucleotides). miRNAs regulate various physiological pathways such as differentiation, proliferation, and apoptosis by cleavage or translational suppression of target messenger RNAs (mRNAs) ( 1, 2, 3). Currently, over 400 human miRNAs have been identified and registered in the miRNA database miRBase ( 4), and they are predicted to regulate 30% of protein-encoding transcripts (5,6). Computational analysis estimates the presence of up to 1,000 miRNAs (7). Recently, miRNAs have been reported to work as oncogenes or tumor suppressor genes and be directly involved in the initiation, progression, and metastasis of various cancers (8,9,10). Therefore, this chapter focuses on the role that miRNAs play in cancer, and the use of miRNAs in drug discovery. Collection of evidence suggests that miRNAs can be potentially useful for understanding tumorigenesis and discovering novel strategies for cancer diagnosis and therapy.

2. BIOGENESIS OF MIRNAS The majority of miRNA genes are located in the introns of protein-coding genes or outside genes ( 11). Unlike Drosophila, most human miRNA genes exist sporadically, although some miRNAs are found as clusters (12,13,14). miRNAs are generated in multiple steps (Fig. 1). Initially, miRNAs are transcribed by RNA polymerase II as long RNA precursors (pri-miRNAs) ( 15, 16, 17). Pri-miRNAs are usually several kilobases in length, and contain a 7-metyl guanosine cap structure and a poly(A) tail similar to protein-coding mRNAs. The transcribed pri-miRNAs are processed into precursors of approximately 70 nucleotides (pre-miRNAs) with a hairpin-shaped stem-loop secondary structure, a 5 phosphate and a two-nucleotide 3 overhang by the RNase III enzyme, Drosha, and a double-stranded-RNA-binding protein, DGCR8/Pasha ( 18, 19, 20). The pre-miRNAs are then transported to the cytoplasm by a member of the Ran transport receptor family, Exportin-5, in a Ran guanosine triphosphate-dependent manner ( 21, 22). Pre-miRNAs exported in the cytoplasm are further processed by another RNase III enzyme, Dicer, and unwound by a helicase (23). Finally, only one mature miRNA strand (guide strand) is incorporated into an RNAinduced silencing complex (RISC) that mediates cleavage or translational inhibition of target mRNAs, while the other strand (passenger strand) is quickly degraded (24,25,26). The stability of the base pairs at the 5 end of the duplex determines which strand is incorporated in RISC (27,28). RISC is composed of Dicer, Argonaute2 (Ago2), and the double-strand RNA binding protein, TRBP ( 26, 29), and cleaves target mRNAs more efficiently by using pre-miRNAs rather than the duplex RNAs that do not have the stemloop structure, suggesting that processing by Dicer may be coupled with assembly of the mature miRNA into RISC (26). The incorporated miRNA guides the RISC to the complementary sequence in the 3 untranslated region (UTR) of target mRNAs. miRNAs base-pair to the 3 UTR of the target mRNA with perfect or near perfect complementarity, leading to the target mRNA degradation by Ago2, a component of RISC (30). On the contrary, partial base pairing between a miRNA and a target mRNA leads to translational silencing of a target mRNA

Chapter 4 / MicroRNAs in Cancer

49

Fig. 1. Mechanism of biogenesis and function in miRNAs. A schematic diagram of miRNA biogenesis is shown. miRNAs are transcribed by RNA polymerase II and sequentially processed by drosha/DGCR8 and dicer. miRNA-loaded RISC causes the cleavage or translational silencing of target mRNAs.

without RNA degradation (31). In partial base pairing, the binding of some nucleotides in the 5 region of miRNAs has been indicated to be functionally important by systematic mutation experiments (32,33).

3. THE ROLE OF MIRNAS IN CANCER: DIAGNOSIS AND DRUG DISCOVERY miRNAs have distinct expression patterns among tissues and cells in different differentiation stages ( 34, 35). Lim et al. ( 36) showed that over-expression of miR-124, a brain-specific miRNA, shifted the gene expression profile of HeLa cells toward that of the brain. Similarly, over-expression of muscle-specific miR-1 shifted the expression profile toward that of muscle. These results indicate that miRNAs play important roles in cell differentiation and characterization. Therefore, miRNAs are considered to have a significant influence on various disorders. Recently, it has been reported that the expression of several miRNAs are altered in a variety of human cancers, suggesting potential roles of miRNAs in tumorigenesis (37). Calin et al. (38) showed that more than 50% of miRNAs were located in cancerassociated genomic regions or in fragile sites. In fact, miR-15a and miR-16 genes exist

50

Part I / Genomic Experimental Approaches in Oncology

as a cistronic cluster at 13q14, which is deleted or down-regulated in most cases (∼68%) of B-cell chronic lymphocytic leukemias (39). Cimmino et al. (40) found that both these miRNAs negatively regulate the expression of B-cell lymphoma 2 (BCL2), which inhibits apoptosis and is present in many types of cancers including leukemias. In fact, overexpression of miR-15 and miR-16 in the MEG-01 cell line induces apoptotic cell death. Alterations in the gene copy number of miRNAs are detected in a variety of human cancers (41,42,43). Zhang et al. (41) showed that miRNAs exhibited high-frequency genomic alterations in human ovarian, breast cancer, and melanoma using high-resolution array-based comparative genomic hybridization. Hayashita et al. (42) found that the expression and gene copy number of the miR-17– 92 cluster—composed of seven miRNAs—is increased in lung cancer cell lines, especially with small-cell lung cancer histology. Enforced expression of miRNAs included in this polycistronic cluster enhances cell proliferation in a lung cancer cell line. The increase in expression and gene copy number of miR-17–92 cluster was also found in B-cell lymphomas ( 43). The expression of miRNAs in this cluster is upregulated by c-Myc, whose expression and/or function is one of the most common abnormalities in human cancers, and miR-17-5p and miR-20a in this miR-17–92 cluster negatively regulate the expression of the transcriptional factor E2F1 (44). Furthermore, it was indicated that miR-17–19b cluster included in miR-17–92 cluster inhibited apoptotic cell death, and accelerated c-Myc–induced lymphomagenesis in mice reconstituted with miR-17–19b cluster-over-expressed haematopoietic stem cells (43). In addition, the miR-17–92 cluster has been reported to augment angiogenesis in vivo by down-regulation of anti-angiogenic thrombospondin-1 and connective tissue growth factor in Ras-transformed colonocytes (45). miR-155 was identified as a miRNA whose copy number and expression were upregulated in several types of B-cell lymphomas ( 46). The miR-155 gene is located in the final exon of the B-cell integration cluster (BIC) non-coding gene, which is shown to accelerate the pathogenesis of c-Myc–associated lymphomas and leukemias, suggesting potential roles of miR-155 in B-cell lymphomas (47). Indeed, transgenic mice with miR-155 driven by the B-cell-specific E␮ enhancer rapidly develop a polyclonal B-cell malignancy (48). The up-regulated expression of miR-155 is also reported in breast, lung, colon, and thyroid cancer (49,50). However, the actual molecular mechanism of miR-155 remains unknown, although it is reported that miR-155 down-regulates the expression of the angiotensin II type I receptor (51). As an antiapoptotic miRNA, miR-21 was recently identified to be up-regulated in human breast tumor tissues, glioblastoma tumor tissues, and malignant cholangiocytes ( 52, 53, 54). Inhibition of miR-21 by antisense oligonucleotides causes activation of caspases and induction of apoptotic cell death in a human breast cancer cell line and glioblastoma cell line (52,53). Furthermore, miR-21 inhibits gemcitabine-induced apoptotic cell death in cholangiocarcinoma cell lines by down-regulation of PTEN (phosphatase and tensin homolog deleted on chromosome 10), which positively regulates apoptosis via inhibition of PI 3-kinase signaling activation ( 54). The expression of miR-141 and miR-200b are also up-regulated in malignant cholangiocytes. Inhibitions of these miRNAs using miRNA-specific antisense oligonucleotides decreased proliferation

Chapter 4 / MicroRNAs in Cancer

51

of a malignant cholangiocyte cell line. The up-regulated expression of miR-21 is also detected in human colon, lung, pancreas, prostate, and stomach cancer (49,55), suggesting the possibility that miR-21 inhibits apoptotic cell death in these cancers. Interestingly, it has been reported that the expression level of let-7 is reduced in human lung cancers (56). This result suggests that let-7 might act as a tumor suppressor gene in lung cancer. In fact, regardless of disease stage, lung cancer patients with downregulation of let-7 had shortened post-operative survival ( 56). Furthermore, Johnson et al. ( 57) found that let-7 negatively regulated the expression of human RAS family members, which possess potent oncogenic activity. Actually, RAS protein levels are inversely correlated with let-7 expression levels in human lung cancers, suggesting a possible mechanism for let-7 in lung cancer. To identify novel miRNAs involved in cellular transformation, Voorhoeve et al. (58) performed functional genetic screens using a library of vectors expressing human miRNAs and in vitro neoplastic transformation assays. They showed that miR-372 and miR-373 accelerate proliferation and tumorigenic development in primary human cells that express oncogenic RAS and tumor suppressor p53, possibly through suppression of p53-mediated CDK inhibition by down-regulation of large tumor suppressor homolog 2 (LATS2) (58,59). Furthermore, miR-372 was found to be exclusively over-expressed in most human testicular germ cell tumors that rarely exhibit loss of p53 function, suggesting contribution of miR-372 to the development of human testicular germ cell tumors by inhibition of the p53 pathway (58). Recent evidence indicates that polymorphisms and genetic variation in germ line as well as somatic cells have a critical role in cancer predisposition and malignancy (60,61). However, in spite of comprehensive scanning of protein coding genes, the molecular basis of familial cancers remains largely unknown. Recently, a germ line mutation of the miR-16-1–miR-15a primary precursor, which impaired mature miRNA expressions, was identified in B-cell chronic lymphocytic leukemia patients (62). Furthermore, germ line or somatic mutations of miRNAs were found in 11 of 75 patients with B-cell chronic lymphocytic leukemia, but none of these mutations were found in 160 persons without cancer (62). These results suggest that genetic variation of miRNAs in a germ line may play important roles in cancer predisposition and malignancy. In addition, germ line mutation in miRNA-target sites of mRNA 3 UTR were found in KIT and slit and trk-like family member 1 (SLITRK1), suggesting genetic variation of miRNA-target sites in a germ line may also play significant roles in disease predisposition (50,63). Human cytochrome P450 (CYP) 1B1, which is abundantly expressed in malignant tumor tissues, is a member of drug-metabolizing enzymes and catalyzes the metabolic activation of various procarcinogens. Recently, it was found that CYP1B1 expression was post-transcriptionally inhibited by miR-27b ( 64). Furthermore, decrease of miR27b expression and increase of CYP1B1 expression in most breast cancer tissues was detected (64). These results indicate that miRNAs may play important roles in not only physiologic events but also drug metabolism and production of carcinogens. Global expression profiling analysis of protein coding genes is known to be useful for cancer diagnoses and prognosis predictions (65). Recently, Lu et al. (37) indicated that miRNA expression profiles can successfully classify poorly differentiated cancers that cannot be classified by mRNA expression profiles. Accordingly, miRNA expression profiles are more accurately correlated with clinical severity of cancer malignancy than

Part I / Genomic Experimental Approaches in Oncology

52

Table 1 Canser-Associated miRNAs miRNA Oncogene miR-17–92 miR-21

Cancer types

Targetsa

References

BCL, lung breast, cholangiocyte, colon, glioblastoma, lung, pancreas, prostate, stomach BCL, breast, colon, lung, thyroid testicular germ cell

CTGF, E2F1,Tsp1 PTEN

42–45 49, 52–55

AT1R LATS2

46, 49–51 58

RAS BCL2

55–57 39, 40

miR-155 miR-372/373 Tumor suppressor gene let-7a breast, lung miR-15a/16 B-CLL a

Target genes identified by the biological experiments are listed. Abbreviations: AT1R, angiotensin II type I receptor; B-CLL, B-cell chronic lymphocytic leukemia; BCL, B-cell lymphoma; BCL2, B-cell lymphoma 2; CTGF, connective tissue growth factor; LATS2, large tumor suppressor homolog 2; PTEN, phosphatase and tensin homolog deleted on chromosome 10; Tsp1, thrombospondin-1.

protein-coding gene expression profiles. This result indicates the potential of miRNA expression profiles in cancer classification and prognosis prediction (37). Because miRNAs act as oncogenes or tumor suppressor genes (Table 1), miRNAs are potential targets of therapeutic strategies. Recently, Krutzfeldt et al. (66) indicated that chemically engineered oligonucleotides, called antagomirs, efficiently inhibited miRNAs in vivo. Additionally, it is reported that introduction of 2 -O-methoxyethyl phosphorothioate antisense oligonucleotide of miR-122 (abundant in the liver and regulates cholesterol and fatty-acid metabolism) decreases plasma cholesterol levels and improves liver steatosis in mice with diet-induced obesity (67). These findings indicate that antisense oligonucleotides are also potential targets for drug discovery, suggesting the possibility that intractable cancers may become curable by over-expression and/or inhibition of miRNAs. However, for miRNAs to be used in gene therapy, further improvement is required to make miRNAs more effective and less toxic than other cancer therapy.

4. PERSPECTIVE It has been established that miRNAs play critical roles in cell differentiation, proliferation, and apoptosis, and the abnormalities of specific miRNA expression contribute to tumorigenesis. Additionally, recent studies show that polymorphisms or genetic variation of miRNAs and miRNA-target sites of mRNAs in a germ line may play important roles in cancer predisposition and malignancy (50,62). Therefore, miRNAs are expected to be powerful tools for cancer classification, diagnosis, and prognosis prediction, as well as to be potential targets of cancer therapy. Furthermore, identification of target mRNAs regulated by miRNAs, elucidation of the oncogenic or tumor suppressive molecular mechanisms by miRNAs, and identification

Chapter 4 / MicroRNAs in Cancer

53

of genetic variation in miRNAs and miRNA-target sites of mRNAs may lead to the discovery of new molecular targets related to oncogenesis. Bioinformatics approaches have predicted that a single miRNA may have hundreds of target genes ( 5, 6, 68, 69, 70, 71, 72, 73, 74), although detailed experimental validation has yet to be done. Development of a comprehensive assay to rapidly identify target mRNAs would greatly assist our understanding of miRNAs and lead to novel therapeutic approaches against cancer.

ACKNOWLEDGEMENTS We would like to thank Dr. N. Hirota for her invaluable advice. We apologize for the incompleteness of the referencing due to space limitations and timing.

REFERENCES 1. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 2004;116:281–297. 2. Tsuchiya S, Okuno Y, Tsujimoto G. MicroRNA: biogenetic and functional mechanisms and involvements in cell differentiation and cancer. J Pharmacol Sci 2006;101:267–270. 3. Pasquinelli AE, Reinhart BJ, Slack F et al. Conservation of the sequence and temporal expression of let-7 heterochronic regulatory RNA. Nature 2000;408:86–89. 4. Griffiths-Jones S, Grocock RJ, van Dongen S et al. miRBase: microRNA sequences, targets and gene nomenclature. Nucleic Acids Res 2006;34:D140–144. 5. Lewis BP, Burge CB, Bartel DP. Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets. Cell 2005;120:15–20. 6. Xie X, Lu J, Kulbokas EJ et al. Systematic discovery of regulatory motifs in human promoters and 3 UTRs by comparison of several mammals. Nature 2005;434:338–345. 7. Berezikov E, Guryev V, van de Belt J et al. Phylogenetic shadowing and computational identification of human microRNA genes. Cell 2005;120:21–24. 8. Esquela-Kerscher A, Slack FJ. Oncomirs: microRNAs with a role in cancer. Nat Rev Cancer 2006;6:259–269. 9. Calin GA, Croce CM. MicroRNA-cancer connection: the beginning of a new tale. Cancer Res 2006;66:7390–7394. 10. Calin GA, Croce CM. MicroRNA signatures in human cancers. Nat Rev Cancer 2006;6:857–866. 11. Rodriguez A, Griffiths-Jones S, Ashurst JL et al. Identification of mammalian microRNA host genes and transcription units. Genome Res 2004;14:1902–1910. 12. Lagos-Quintana M, Rauhut R, Lendeckel W et al. Identification of novel genes coding for small expressed RNAs. Science 2001;294:853–858. 13. Lim LP, Glasner ME, Yekta S et al. Vertebrate microRNA genes. Science 2003;299:1540. 14. Yu J, Wang F, Yang GH et al. Human microRNA clusters: genomic organization and expression profile in leukemia cell lines. Biochem Biophys Res Commun 2006;349:59–68. 15. Lee Y, Jeon K, Lee JT et al. MicroRNA maturation: stepwise processing and subcellular localization. EMBO J 2002;21:4663–4670. 16. Cai X, Hagedorn CH, Cullen BR. Human microRNAs are processed from capped, polyadenylated transcripts that can also function as mRNAs. RNA 2004;10:1957–1966. 17. Lee Y, Kim M, Han J et al. MicroRNA genes are transcribed by RNA polymerase II. EMBO J 2004;23:4051–4060. 18. Lee Y, Ahn C, Han J et al. The nuclear RNase III Drosha initiates microRNA processing. Nature 2003;425:415–419. 19. Gregory RI, Yan KP, Amuthan G et al. The microprocessor complex mediates the genesis of microRNAs. Nature 2004;432:235–240. 20. Han J, Lee Y, Yeom KH et al. Molecular basis for the recognition of primary microRNAs by the Drosha-DGCR8 complex. Cell 2006;125:887–901.

54

Part I / Genomic Experimental Approaches in Oncology

21. Yi R, Qin Y, Macara IG et al. Exportin-5 mediates the nuclear export of pre-microRNAs and short hairpin RNAs. Genes Dev 2003;17:3011–3016. 22. Lund E, Guttinger S, Calado A et al. Nuclear export of microRNA precursors. Science 2004;303: 95–98. 23. Hutvagner G, McLachlan J, Pasquinelli AE et al. A cellular function for the RNA-interference enzyme Dicer in the maturation of the let-7 small temporal RNA. Science 2001;293:834–838. 24. Matranga C, Tomari Y, Shin C et al. Passenger-strand cleavage facilitates assembly of siRNA into Ago2-containing RNAi enzyme complexes. Cell 2005;123:607–620. 25. Rand TA, Petersen S, Du F et al. Argonaute2 cleaves the anti-guide strand of siRNA during RISC activation. Cell 2005;123:621–629. 26. Gregory RI, Chendrimada TP, Cooch N et al. Human RISC couples microRNA biogenesis and posttranscriptional gene silencing. Cell 2005;123:631–640. 27. Khvorova A, Reynolds A, Jayasena SD. Functional siRNAs and miRNAs exhibit strand bias. Cell 2003;115:209–216. 28. Schwarz DS, Hutv´agner G, Du T et al. Asymmetry in the assembly of the RNAi enzyme complex. Cell 2003;115:199–208. 29. Chendrimada TP, Gregory RI, Kumaraswamy E et al. TRBP recruits the Dicer complex to Ago2 for microRNA processing and gene silencing. Nature 2005;436:740–744. 30. Meister G, Landthaler M, Patkaniowska A et al. Human Argonaute2 mediates RNA cleavage targeted by miRNAs and siRNAs. Mol Cell 2004;15:185–197. 31. Hutv´agner G, Zamore PD. A microRNA in a multiple-turnover RNAi enzyme complex. Science 2002;297:2056–2060. 32. Doench JG, Sharp PA. Specificity of microRNA target selection in translational repression. Genes Dev 2004;18:504–511. 33. Kiriakidou M, Nelson PT, Kouranov A et al. A combined computational–experimental approach predicts human microRNA targets. Genes Dev 2004;18:1165–1178. 34. Lagos-Quintana M, Rauhut R, Yalcin A et al. Identification of tissue-specific microRNAs from mouse. Curr Biol 2002;12:735–739. 35. Liu CG, Calin GA, Meloon B et al. An oligonucleotide microchip for genome-wide microRNA profiling in human and mouse tissues. Proc Natl Acad Sci USA 2004;101:9740–9744. 36. Lim LP, Lau NC, Garrett-Engele P et al. Microarray analysis shows that some microRNAs downregulate large numbers of target mRNAs. Nature 2005;433:769–773. 37. Lu J, Getz G, Miska EA et al. MicroRNA expression profiles classify human cancers. Nature 2005;435:834–838. 38. Calin GA, Sevignani C, Dumitru CD et al. Human microRNA genes are frequently located at fragile sites and genomic regions involved in cancers. Proc Natl Acad Sci USA 2004;101:2999–3004. 39. Calin GA, Dumitru CD, Shimizu M et al. Frequent deletions and down-regulation of micro- RNA genes miR15 and miR16 at 13q14 in chronic lymphocytic leukemia. Proc Natl Acad Sci USA 2002;99:15524–15529. 40. Cimmino A, Calin GA, Fabbri M et al. miR-15 and miR-16 induce apoptosis by targeting BCL2. Proc Natl Acad Sci USA 2005;102:13944-13949. Erratum in: Proc Natl Acad Sci USA 2006;103: 2464–2565. 41. Zhang L, Huang J, Yang N et al. MicroRNAs exhibit high-frequency genomic alterations in human cancer. Proc Natl Acad Sci USA 2006;103:9136–9141. 42. Hayashita Y, Osada H, Tatematsu Y et al. A polycistronic microRNA cluster, miR-17–92, is over-expressed in human lung cancers and enhances cell proliferation. Cancer Res 2005;65: 9628–9632. 43. He L, Thomson JM, Hemann MT et al. A microRNA polycistron as a potential human oncogene. Nature 2005;435:828–833. 44. O’Donnell KA, Wentzel EA, Zeller KI et al. c-Myc-regulated microRNAs modulate E2F1 expression. Nature 2005;435:839–843.

Chapter 4 / MicroRNAs in Cancer

55

45. Dews M, Homayouni A, Yu D et al. Augmentation of tumor angiogenesis by a Myc-activated microRNA cluster. Nat Genet 2006;38:1060–1065. 46. Eis PS, Tam W, Sun L et al. Accumulation of miR-155 and BIC RNA in human B cell lymphomas. Proc Natl Acad Sci USA 2005;102:3627–3632. 47. Tam W, Hughes SH, Hayward WS et al. Avian bic, a gene isolated from a common retroviral site in avian leukosis virus-induced lymphomas that encodes a noncoding RNA, cooperates with c-myc in lymphomagenesis and erythroleukemogenesis. J Virol 2002;76:4275–4286. 48. Costinean S, Zanesi N, Pekarsky Y et al. Pre-B cell proliferation and lymphoblastic leukemia/highgrade lymphoma in E(mu)-miR155 transgenic mice. Proc Natl Acad Sci USA 2006;103:7024–7029. 49. Volinia S, Calin GA, Liu CG et al. A microRNA expression signature of human solid tumors defines cancer gene targets. Proc Natl Acad Sci USA 2006;103:2257–2261. 50. He H, Jazdzewski K, Li W et al. The role of microRNA genes in papillary thyroid carcinoma. Proc Natl Acad Sci USA 2005;102:19075–19080. 51. Martin MM, Lee EJ, Buckenberger JA et al. MicroRNA-155 regulates human angiotensin II type 1 receptor expression in fibroblasts. J Biol Chem 2006;281:18277–18284. 52. Si ML, Zhu S, Wu H et al. miR-21-mediated tumor growth. Oncogene 2007;26:2799–2803. 53. Chan JA, Krichevsky AM, Kosik KS. MicroRNA-21 is an antiapoptotic factor in human glioblastoma cells. Cancer Res 2005;65:6029–6033. 54. Meng F, Henson R, Lang M et al. Involvement of human micro-RNA in growth and response to chemotherapy in human cholangiocarcinoma cell lines. Gastroenterology 2006;130:2113–2129. 55. Iorio MV, Ferracin M, Liu CG et al. MicroRNA gene expression deregulation in human breast cancer. Cancer Res 2005;65:7065–7070. 56. Takamizawa J, Konishi H, Yanagisawa K et al. Reduced expression of the let-7 microRNAs in human lung cancers in association with shortened postoperative survival. Cancer Res 2004;64:3753–3756. 57. Johnson SM, Grosshans H, Shingara J et al. RAS is regulated by the let-7 microRNA family. Cell 2005;120:635–647. 58. Voorhoeve PM, le Sage C, Schrier M et al. A genetic screen implicates miRNA-372 and miRNA-373 as oncogenes in testicular germ cell tumors. Cell 2006;124:1169-1181. 59. Hahn WC, Counter CM, Lundberg AS et al. Creation of human tumour cells with defined genetic elements. Nature 1999;400:464–468. 60. Hunter K. Host genetics influence tumour metastasis. Nat Rev Cancer 2006;6:141–146. 61. Hunter KW, Crawford NP. Germ line polymorphism in metastatic progression. Cancer Res 2006;66:1251–1254. 62. Calin GA, Ferracin M, Cimmino A et al. A MicroRNA signature associated with prognosis and progression in chronic lymphocytic leukemia. N Engl J Med 2005;353:1793–1801. 63. Abelson JF, Kwan KY, O’Roak BJ et al. Sequence variants in SLITRK1 are associated with Tourette’s syndrome. Science 2005;310:317–320. 64. Tsuchiya Y, Nakajima M, Takagi S et al. MicroRNA regulates the expression of human cytochrome P450 1B1. Cancer Res 2006;66:9090–9098. 65. Ramaswamy S, Tamayo P, Rifkin R et al. Multiclass cancer diagnosis using tumor gene expression signatures. Proc Natl Acad Sci USA 2001;98:15149–15154. 66. Krutzfeldt J, Rajewsky N, Braich R et al. Silencing of microRNAs in vivo with “antagomirs.” Nature 2005;438:685–689. 67. Esau C, Davis S, Murray SF et al. miR-122 regulation of lipid metabolism revealed by in vivo antisense targeting. Cell Metab 2006;3:87–98. 68. Enright AJ, John B, Gaul U et al. MicroRNA targets in Drosophila. Genome Biol 2003;5:R1. 69. John B, Enright AJ, Aravin A et al. Human microRNA targets. PLoS Biol 2004;2:e363. Erratum in: PLoS Biol 2005;3:e264. 70. Kiriakidou M, Nelson PT, Kouranov A et al. A combined computational–experimental approach predicts human microRNA targets. Genes Dev 2004;18:1165–1178. 71. Lewis BP, Shih IH, Jones-Rhoades MW et al. Prediction of mammalian microRNA targets. Cell 2003;115:787–798.

56

Part I / Genomic Experimental Approaches in Oncology

72. Krek A, Grun D, Poy MN et al. Combinatorial microRNA target predictions. Nat Genet 2005;37: 495–500. 73. Sethupathy P, Megraw M, Hatzigeorgiou AG. A guide through present computational approaches for the identification of mammalian microRNA targets. Nat Methods 2006;3:881–886. 74. Miranda KC, Huynh T, Tay Y et al. A pattern-based method for the identification of microRNA binding sites and their corresponding heteroduplexes. Cell 2006;126:1203–1217.

5

Pharmacogenomics of the National Cancer Institute’s 60-Tumor Cell Panel Anders Wallqvist, PhD, Ruili Huang, PhD, and David G. Covell, PhD CONTENTS Introduction National Cancer Ins titute (NCI)’S 60-Cell Screening and Gene Expres s ion Data Pathway Analys is of Gene Expres s ion in the NCI’S 60 Cell Lines Drug Mechanis m of Action Probed by Pathway Gene Expres s ions and Growth Inhibition Res pons e Targeting Cancer Pathways Conclus ion Acknowledgements References

S UMMARY One of the important goals of cancer research is to understand the nature of gene expression regulation and biological pathways and to apply this knowledge to find the mechanism by which small drug molecules interfere with the biological system through interactions with gene products and pathways. We have utilized the gene expression and small molecule screening data available at the National Cancer Institute (NCI) for 60 immortalized cell lines representing a range of major cancers. This exten-

From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 5, 

57

58

Part I / Genomic Experimental Approaches in Oncology

sive data set potentially contains the complete information necessary to understand and target cancer cells. In our experience it is most fruitful to adopt systems biology and pharmacogenomic approaches to deconvolute the necessary chemistry and biology in order to conduct a rational anti-cancer drug design effort. In this undertaking, existing biological pathway and gene expression information is merged with drug chemosensitivity data to both elucidate a drug’s mechanism of action and to find cancer-specific targets. This framework offers a rational design strategy to mine novel anti-cancer candidates that are both potent and show specificity to targets in cancer pathways. Key Words: Bioinformatics; chemosensitivity; tumor cell; cytotoxicity; NCI60; gene expression; compound screen

1. INTRODUCTION The highly complex cellular regulatory networks and their interactions with small molecules present challenges to our mechanistic understanding of drug action. Deeper insights into the fundamental mechanisms of cellular functions and pathway regulations are likely to be critical for the development of rational approaches directed at the identification of molecular targets and candidate inhibitors. While the anti-tumor activity of current anti-cancer drugs is reflected in cell killing, mechanism-based studies attempt to specifically associate a drug’s effect to one or many cellular regulation mechanisms ( 1). Individual protein targets of a small molecule may be involved in diverse cellular processes, some or all of which may contribute to the killing potential of a compound. Furthermore, environmental factors such as temperature, radiation, hypoxia, and nutrients, as well as drugs, stimulate an adaptive sensory and signaling machinery of the cell, and therefore may influence drug sensitivity, cell survival, and apoptosis. The normal network structures of this system may be perturbed in diseases through genetic mutations and/or by pathological environmental cues such as infectious agents or chemical carcinogens. Cancer is believed to arise from multiple spontaneous and/or inherited mutations functioning in networks that control vital cellular events ( 2, 3, 4), which is partly reflected by genetic alterations in intracellular signaling pathways that normally control the developmental programs and the cellular response to extrinsic factors ( 5). The evolving states of certain cancers are reflected in dynamically changing expression patterns of genes and proteins within the cells (6). An important challenge to associating gene expression in the context of biological processes involves the formulation of effective strategies to relate drug action to precise molecular targets. The notion of pathways (7) is a convenient abstraction that can both be considered in isolation and has been found useful in describing and understanding the inner workings of cellular biology (8,9). Because the biological response of a cell to a compound represents the whole organism readout of the drug’s interaction within a cellular milieu, one approach is to utilize drug–gene pathway relationships to propose novel drug targets or target-specific drugs. For example, if a drug interacts with one gene product, the entire pathway or pathways having this gene may be disturbed, and a direct correlation between the drug and its target may not be apparent. Alternatively, the

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

59

consequence of a drug action may be reflected by correlations of the drug response to other gene expressions within the pathways containing the putative target. Cases where single drug-gene correlations are not directly apparent may be revealed by this broader examination of related genes within a pathway. The use of pathways also provides a central reference to a more systematic view of biological processes (7,10,11). Because cancer is a disease closely tied to genetic instability, the relative stability of pathways or pathway gene expression regulation is of great interest. The tendency for some pathways to change their behavior in cancer tissues compared to normal tissues may be a reflection of instability in the regulation of these pathways and represent potentially important and specific drug targets.

2. NATIONAL CANCER INSTITUTE (NCI)’S 60-CELL SCREENING AND GENE EXPRESSION DATA The National Cancer Institute’s (NCI’s) 60-cell line drug discovery panel (NCI60 ) was developed as a tool to assess anti-cancer activity of compounds against a range of cell lines derived from different cancers, including lung, renal, colorectal, ovarian, breast, prostate, central nervous system, melanoma, and hematological malignancies ( 12). Chemicals that reduce the viability of the cell are tagged as potential leads for affecting particular pathways characteristic of each tumor cell’s biology. The biological response of such a cell-based assay is rich in information because, in principle, the complete systems biology information is encoded in the assay. Even though such data is far more complicated and harder to interpret than a noncellular, direct molecular binding assay, the ability to monitor a complete biological system over a number of genetically different cell lines offers advantages when studying mechanisms of drug action. The difference in cellular response to a drug is reflected in this biological response profile, ultimately encoding the underlying genetic difference responsible for the varied response. The screening data consists of concentration values (GI50 ) for each cell line at which the drug results in a 50% reduction in the net protein increase relative to untreated control cells. In this assay the cells are inoculated onto a standard microtiter plate (typically 20,000 cells/well) and then pre-incubated for 24 hr in the absence of drug. Test agents are then added in five 10-fold dilutions starting from the highest soluble concentration, and incubated for a further 48 hr. Finally the cells are fixed in situ, washed, dried and sulforhodamine B is added. After further washing and drying the stained cells are solubilized and measured spectrophotometrically on automated plate readers interfaced to a computer. From the relative growth inhibition compared to non-exposed cells, a dose–response curve is constructed, and the concentration at which 50% growth inhibition (GI50 ) is achieved is recorded. The log-transformed concentration for each cell line is recorded and can then be retrieved or visually inspected as a mean graph. Other end-points such as TGI and LC50 are also retrievable from the data. Because an individual GI50 concentration for a particular cell line may not be informative in itself, it is the differential pattern of the log-transformed concentrations across all cell lines that carries characteristic information about possible mechanisms of drug action. This pattern of GI50 measurements

60

Part I / Genomic Experimental Approaches in Oncology

across the tumor cell lines, the biological response profile, has proven useful for identifying mechanisms of action for drug classes, and aids in the classification of novel drugs submitted to the NCI’s tumor screen (13,14,15). The NCI60 screen has been operational since 1990 and has screened over 100,000 compounds.

2.1. Gene Expression The microarray data available for the NCI60 offers a rich source of information for examining gene expressions in conjunction with existing knowledge of chemotypes and cytotoxicity. This data is available for a number of different experimental platforms. Constitutive gene expression within the NCI60 reflects immortalized cell growth under specific environmental conditions. The common approach is to identify the strongest correlations between constitutive differential gene expression profiles and cytotoxicity measures across the NCI60 . While gene expression profiles lend themselves to many interpretations, we have adopted the simple working premise that differential expression reflects cellular maintenance and growth, albeit through complex processes. Thus, all cellular processes necessary for survival ultimately derive some dependency on gene expression, or the transcriptional control of gene expression. This premise can be naturally extended to include the view that gene expression also contributes to a cytotoxic response. Albeit an overly simplified premise for this complicated system, this perspective provides a reference point for evaluating whether constitutive gene expressions are different from “average,” and hence are an indication of some dependency on survival or cell death in response to drug exposure.

2.2. Self-Organizing Map In order to have an overall measure of similarities between all GI50 data vectors we have used a self-organizing map (SOM)(16) to organize cellular growth inhibition data derived from the NCI60 tumor cell panels ( 14). The SOM algorithm identifies cluster vectors in the 60-dimensional data space by minimizing the deviation between the GI50 data vectors and the cluster vectors. Regions in the GI50 -space that are dense with data vectors attract many cluster vectors and regions with few data vectors attract fewer cluster vectors, resulting in a division of response space that mimics information content. An advantage of SOM reordered data is the ability to visualize the global clustering results in an interpretable manner. Our preferred method of display is the uniform projection of SOM clustering from high-dimensional space to a 2D-map. This mapping is both simple and retains a great deal of the original high-dimensional information. Additional details regarding the creation and access to the GI50 SOM are given in the references (14). SOM clustering of the GI50 data segregates compounds into nine major response categories: mitosis (M), membrane function (N), nucleic acid metabolism (S), metabolic stress and cell survival (Q), kinases/phosphatases and oxidative stress (P), and four unexplored regions R, F, J, and V (17,18,19). Each of these regions is further divided into a total of 80 clades [a clade is a group of clusters (nodes) that share similar cytotoxic responses] or sub-regions: M1 –M8 , N1 –N13 , P1 –P8 , Q1 –Q7 , R1 –R7 , S1 –S13 , F1 –F8 , J1 –J8 , V1 –V8 . The current SOM extends our previously published analysis to include the existing complement of newly screened compounds (14).

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

61

3. PATHWAY ANALYSIS OF GENE EXPRESSION IN THE NCI’S 60 CELL LINES Numerous efforts have been dedicated to the task of deconvoluting the function and regulation of biological networks, especially with the increasing availability of RNA expression microarrays, which provide large amounts of data for analysis of individual genes within predefined pathways or elucidation of gene regulation networks (20,21,22,23,24). Deriving gene regulation networks or pathways on the basis of expression data is based on the general premise that co-regulated genes function in the same pathway, or, in other words, functionally related genes are co-regulated or co-expressed. Co-expression of genes has been observed using pathways annotated by the KEGG (Kyoto Encyclopedia of Genes and Genomes) based on gene expression data in colon and liver cancer cells and normal tissue samples ( 24) and in the Arabidopsis genome (25). Neighboring genes, that is, genes that are immediately adjacent on chromosomes, have been found to be co-expressed in humans (26,27), Drosophila (28,29,30), yeast (28), Caenorhabditis elegans (31), and Arabidopsis (25). The requirement for co-regulation of functionally related genes has been proposed as a possible cause for the observed co-expression of genes. In addition, the number of interactions between proteins has been implied as an important predictor for the degree of co-expression between their corresponding genes—a result that has been offered as an explanation for particularly high degrees of co-expression in genes encoding proteins that are known to function in multi-component complexes, which often contain a large number of protein–protein interactions (32,33). In yeast, genes that encode interacting proteins tend to be co-expressed (32,34,35). In contrast, the degree of co-expression for genes that encode enzymes in metabolic pathways has been found to be generally low (24,25), despite the observation of similarities in gene expression patterns for some metabolic pathways in embryonic and adult mouse tissues (36). Functionally linked interacting proteins have been observed to share a higher proportion of shared transcription factor–binding sites regulating transcription of genes for enzymes catalyzing the conversion of adjacent substrates to products in a collection of metabolic pathways (37). This finding has led to the hypothesis that genes encoding a set of interacting proteins can be transcribed using a common set of regulatory signals, whereas substrate concentration and enyzme–substrate interactions may exact regulatory control of metabolic pathways, as distinct from explicit transcriptional control (38). Our analysis focuses on the constitutive gene expression data measured across the NCI’s 60-tumor cell screen and gene expression regulation patterns within predefined pathways or functional annotations by KEGG, BioCarta, and gene categories defined by GO (gene ontology). One measure of pathway cohesiveness is based on correlations of gene expression patterns across the NCI60 by comparing the intra- versus interpathway gene expression correlations using the Kruskal–Wallis procedure ( 39). The cohesiveness of a pathway is represented by the coherence of its gene expressions as measured by their correlation strength. Pathways can further be organized according to similarity of gene occurrence patterns in each pathway. This is used to hypothesize

62

Part I / Genomic Experimental Approaches in Oncology

interactions between pathways either through gene sharing or co-expression of genes, which may be indicative of a higher level of regulation and coherence within cellular processes. Our study provides a comprehensive evaluation of the level of co-regulation in pathway gene expressions measured across the NCI60 . We confirm that the level of gene co-expression is significantly higher in pathways or functionally related gene groups than a randomly selected set of genes. Approximately 20% of pathways or functionally related gene groups analyzed have statistically significant, coherent gene expressions. Based on the type of pathways found to be cohesive versus non-cohesive, we can infer that pathway gene expression cohesiveness is probably on a “need-to-be” basis, that is, genes in the same pathway are only co-expressed when there is a need for it. Pathways are thus designed by nature to be robust and flexible enough to adapt to environmental changes in order to ensure cell survival; that is, alternate mechanisms can take over if parts of a pathway fail to function. However, pathways with genes encoding parts of a large protein complex need to be cohesive because the co-presence and close physical interaction of the proteins required for the proper function of the protein complex demand the co-expression and tight regulation of their corresponding genes. The same may be true for genes that are expressed in the same cellular location/component, which are found to be more cohesive than genes participating in the same biological process or engaging in the same molecular function. Pathways involved in genetic information processing (e.g., replication and repair, sorting and degradation, transcription, translation) and in the cell cycle tend to be cohesive as well, probably because they are responsible for the most vital processes in a biological entity, and therefore need to be tightly regulated at the transcriptional level to ensure precisely synchronized action of their constitutive genes to minimize error, or because these processes are important for growth and proliferation, and thus can be expected to be cohesive in constantly proliferating tumor cells. On the other hand, pathways involved in environmental information processing (signaling pathways) and most metabolic pathways are generally found to be not cohesive, presumably because these pathways need to be more robust to respond to environmental changes. The simultaneous presence of all these genes may not be necessary, because genes are turned on or off sequentially when needed. Therefore these pathways do not need to be tightly regulated at the transcriptional level and they are more likely to be regulated by, for example, substrate concentration and enzyme–substrate interactions. This is also reflected by the fact that these pathways are much less modular than the cohesive pathways. There is a high degree of crosstalk (gene sharing) between the signal transduction pathways as opposed to the very cohesive pathways responsible for genetic information processing, which are highly modular. The few metabolic pathways that are cohesive show that pathways responsible for vital life processes such as nucleotide metabolism, energy metabolism, and isoprenoid and cholesterol biosynthesis need to have tighter regulated gene expressions than generic metabolic pathways, such as carbohydrate metabolism, metabolism of cofactors and vitamins, fatty acid metabolism, protein amino acid glycosylation, and lipid catabolism.

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

63

4. DRUG MECHANISM OF ACTION PROBED BY PATHWAY GENE EXPRESSIONS AND GROWTH INHIBITION RESPONSE One general approach to identify target-specific agents as a basis for understanding a drug’s mechanism of action (MOA) is to relate gene expression patterns measured across a diverse set of tumor cell models to drug-induced chemosensitivity of these same cells (17,40,41,42,43,44,45,46,47,48,49,50,51,52,53). Previous efforts using this strategy have focused mainly on finding causes of drug resistance ( 40, 41, 42, 49, 54, 55). Gene expression signatures have also been used as surrogate markers of cellular states, for example, to identify agents that induce the differentiation of acute myeloid leukemia cells (56). However, nearly all of these investigations have been based on single gene expression–drug response relationships, whereas complex interactions between a drug and highly interconnected biological networks may not be reflected solely by the state of any one gene. Moreover, quantitative assessments that associate significant correlations between gene expression levels and drug sensitivity as a basis for validating a biologically significant connection are not yet a standard practice. Gene expression patterns across the NCI60 can be organized in terms of predefined pathways or functional categories annotated by KEGG, BioCarta, and Gene Ontology. These gene annotations are used to link pathways to drug responses through correlations between pathway gene expression patterns and drug GI50 response profiles clustered in SOM clades. Implicit in this design is the assumption that cytotoxicity profiles most strongly associated with gene expression profiles for genes within a defined pathway are valid indications of a test compound perturbing the pathway; conversely, the genes in the pathway would play major roles in dictating the cytotoxic activity of the compound. Our approach associates drug responses in each clade with subsets of pathways. Assignment of putative MOAs for agents clustered in each SOM response region are then postulated to involve pathways that can be significantly correlated with these agents.

4.1. Mapping Pathways to Drug Mechanism of Actions A hallmark of targeted molecular therapies is over-expression of the drug’s molecular target. Strong support for this behavior can be found within the NCI60 screen as seen by the positive correlations between gene expressions of the proteasome and heat shock proR R and Geldanamycin , respectively (17). Extending these cytotoxicity– teins to Velcade gene expression correlations to pathways is an attempt to establish a pathway-centric perspective to a drug’s MOA. Instead of examining each individual drug–gene correlation, correlations between the GI50 SOM clades (clusters of drugs with similar GI50 profiles) and pathways are evaluated as a more general approach. For each pathway, correlations with GI50 clades are compared between genes “on a pathway” with genes “off a pathway” and the Kruskal– Wallis H statistic is calculated (57,58). Each pathway in KEGG, BioCarta, and each GO term is then mapped onto the GI50 SOM, where each clade has an H-score, representing the strength of correlation between the pathway and the compounds in that clade. The most significantly and specifically correlated pathways are proposed as the most likely targets of the drugs within a clade. Collectively, signal transduction pathways are among the least cohesive pathways; their correlations with the GI50 SOM regions are generally

64

Part I / Genomic Experimental Approaches in Oncology

diffuse, lacking assignment to any one clade or region in GI50 response space. Conversely, this may also imply that it will be difficult to find drugs that specifically target signaling pathways. One of the primary goals of the drug–pathway analysis is to find and interpret drug targets and MOA. The biological pathways that are potentially perturbed by the drugs in the nine SOM response regions can be postulated. Conversely, each pathway may be associated with one or more response regions. We have previously established the general MOA of the agents in some of these SOM regions: mitosis (M), membrane function and oxidative stress (N), nucleic acid metabolism (S), and metabolic stress and cell survival (Q), oxidative metabolism (R), and kinases/phosphatases and oxidative stress (P), via other methods (17,18,19). The pathway mapping results provide additional support for the annotation of some of the SOM regions: For example, the pathways mitotic checkpoint, cytokinesis, kinetochore, and cell cycle are associated with the M-region; the mitochondrial inner membrane, response to oxidative stress, and oxidoreductase activity are associated with the N-region; the granzyme A mediated apoptosis pathway, the DNA topological change, and DNA topoisomerase (ATP-hydrolyzing) activity, are associated with the S-region; the pathways relating to glutamate metabolism, xenobiotic metabolism, cysteine metabolism, and glutathione biosynthesis are associated with the Q-region; the pathways of fatty acid metabolism and oxidative phosphorylation and NADH dehydrogenase (ubiquinone) activity, NADH dehydrogenase activity, oxidoreductase activity, and mitochondrion are associated with the R-region; and the pathways in signaling of hepatocyte growth factor receptor, Erk1/Erk2 MAPK signaling pathway, ATM signaling pathway, FAS signaling pathway (CD95), and oxidoreductase activity, cell–cell signaling and DNA damage response (signal transduction resulting in induction of apoptosis) are associated with the P-region. The additional pathways that are associated with each SOM region through this global pathway analysis provide valuable information and new insights into the MOA for similarly clustered drug molecules. Of interest are the associations of apoptosis with the M-region; cell adhesion and immune response signaling pathways with regions N and P; transport with the N-region; hypoxia and angiogenesis with the P-region; DNA replication, regulation of DNA repair, and translation with the Q-region; and cytoskeleton with the R-region. Finally, the MOA of the agents clustered in the three regions—F, J and V—can be postulated by examining the pathways associated with these regions: For example, the amino acid metabolism pathways and the Wnt signaling pathway map to the F-region, cell cycle and DNA damage related pathways map to the J-region, and urea cycle and metabolism of amino groups and pyruvate metabolism map to the V-region, which seems for the latter to share pathways with its neighboring regions. In fact, many pathways are shared among different SOM regions, and conversely each region is usually associated with multiple pathways. This is expected because any one biological process can be perturbed by many drugs but to different degrees. The agents that can most effectively disrupt a process can now be found by looking at the most significantly correlated sets. Moreover, each SOM region contains the GI50 profiles of thousands of compounds; therefore, it is not surprising that multiple processes, even though usually related, are associated with these compounds. To gain more specific

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

65

information on the MOA of one compound or a small cluster of related compounds, detailed drug–pathway analysis, as described earlier, is required. The region-wide analysis of biological pathways and drug response, however, provides a global view of biological activities or features shared by large groups of compounds. As discussed previously, pathways responsible for vital cellular processes or processes that are related to growth or proliferation, specifically in cancer cells, such as those engaged in genetic information processing, cell cycle, energy metabolism, and nucleotide metabolism, are found to have significantly more coherent gene expressions than most signaling and regular metabolic pathways. The cohesive pathways are also found here, in general, to have stronger pathway–drug correlations than non-cohesive pathways because the high level of gene co-expression in cohesive pathways makes it more likely for genes in the pathway to have similar correlation patterns with, or act coherently toward, a drug. This may imply that cohesive pathways are easier to target, because many drugs seem to be able to significantly disrupt these pathways. Conversely, the correlations of the least cohesive pathways with the GI50 SOM regions are generally diffuse and not strong or specific to any one clade or region. This may be an indication that it will be hard to find drugs that can target non-cohesive pathways or the relationship between drugs, and these pathways are not reflected or easily interpretable by gene–drug correlations. Therefore, instead of looking at non-cohesive pathways that do not correlate significantly with any drugs, it may be more interesting to examine those non-cohesive pathways that can act coherently toward certain drugs, that is, how correlation or interaction with drugs changes their intrinsic cohesiveness. Taking this one step further, in addition to looking at “drug-coupled” pathway cohesiveness through correlation, insight may be obtained by examining “drug-exposed” pathway cohesiveness, that is, to analyze and compare gene expression cohesiveness within a pathway prior to and after drug exposure. The number of pathways significantly correlated with each GI50 SOM clade, on the other hand, represents the number of biological processes the drug agents in the clade are potentially perturbing. This number can be used as an indicator of the level of target specificity or promiscuity of these drugs. High pathway correlation promiscuity is indicated for some drugs. Although this may seem undesirable because of multiple targets and thus the potential of detrimental side effects implicated for the drug, this may represent cases where assaulting a single target by the drug can cause multiple intracellular effects, as reflected by correlations with multiple pathways. On the other hand, this can be deemed to be a desirable property of the drug, because it presents the potential of overcoming the insufficiency of single target inhibition caused by the inherent ability of heterogeneous tumor populations to activate alternative or redundant pathways (59). Based on this premise, drugs with many significantly correlated pathways can be advanced for further investigation.

5. TARGETING CANCER PATHWAYS Cancer is essentially a disease arising from an accumulation of genetic abnormalities ( 60, 61, 62), which are thought to participate in neoplastic development and in some cases the development of chemotherapeutic resistance (63,64,65). Many genes have been

66

Part I / Genomic Experimental Approaches in Oncology

implicated in the genesis of various cancers (60,66). In the process of carcinogenesis, some are found to be mutated, whereas others tend to exhibit dysregulated levels of expression (67,68,69,70). Both mutation status and RNA or protein expression levels have proven valuable for the development of cancer diagnostic assays, particularly for prediction of prognosis. However, a diagnostic gene expression pattern does not necessarily have a causative role in carcinogenesis (71,72,73,74,75,76). A focused analysis on changes in the expression patterns of specific cellular pathways can reveal biological insights that are not easily apparent from variations in individual genes. We analyze gene expression data obtained from human tumor and normal tissue samples and evaluate these observations for the purpose of assessing which pathways are deregulated in cancer, and then apply these results toward the development of a rationale for selecting pathway-specific chemo interventions. The ability to identify and disrupt targets that are characteristic of cancer cells without affecting normal cells is crucial for successful anti-cancer therapy. This analysis focuses on the oligonucleotide microarray samples publicly available at the Whitehead site (www.broad.mit.edu/cancer), which encompass gene expression data measured in 190 patient tumor samples spanning 14 common tumor types (18 subtypes) and 90 normal samples including 12 tissue types (13 subtypes) (77). We have previously applied this data to successfully classify tumor tissue samples ( 78), with the seminal finding that gene expression profiles alone were sufficient to correctly classify most of the tumor tissues according to cancer type. By applying the pathway perspective we can compare and contrast pathway features using gene expression profiles obtained from normal and tumor tissues. We have organized the tissue gene expression patterns in terms of the previously employed gene annotations (pathways or functional categories) defined by KEGG, BioCarta, and Gene Ontology (GO). Co-expression of genes has been observed in certain pathways; however, it is not clear whether any difference exists in the level of pathway gene co-expression between cancer and normal cells. If we assume that co-expression is reflective of coordinated gene regulation, then any change in the level of co-expression in a pathway can be viewed as a change in that pathway’s regulation; and the propensity of a pathway to changes in regulation indicate a level of instability. The degree of gene expression coherence can be evaluated for each pathway following the same procedures as previously discussed ( 39, 58). Briefly, the Kruskal–Wallis H statistic is computed to compare gene–gene expression correlations within a pathway to those between pathways, and used as a measure of pathway gene expression coherence or cohesiveness (PGEC). We consider pathways with significantly stronger intra- than inter-pathway gene–gene correlations, characterized by a large and positive H-score (p < 0.05), as cohesive, and not cohesive otherwise.

5.1. Pathway Cohesiveness in Tumor Versus Normal Tissues As an initial assessment of PGEC in tumor versus normal tissues, all 190 tumor and 90 normal tissue samples are included in computing the H-scores for all three pathway collections. Genes in vetted pathways are found significantly more coherently expressed than a random set of genes, regardless of whether the tissue is normal or tumor bearing. However, a more interesting question is whether differences exist in measures of

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

67

cohesiveness between normal and tumor tissues for individual pathways. Hence, we examine the extent to which each individual pathway has changed its cohesiveness, what types of pathways they are, and whether they have become more or less cohesive in tumor tissues as compared to normal tissues. We will refer to pathways that have become significantly cohesive in tumor tissues as “up” pathways, and those that have lost their coherence in tumor tissues as “down” pathways. Random permutation tests to examine the probability of observing “up” and “down” pathways by chance show that our observation of “down” pathways is significantly higher than random (p = 0.002). These results indicate that pathways tend to become less cohesive in tumor tissues than in normal tissues. We have used the predefined pathway categories to establish the type of pathways that are prone to variations in their pattern of gene regulation. In most pathway categories, the number of “up” pathways is not significantly different from the number of “down” pathways; however, significantly more “down” pathways are found in the BioCarta pathway categories “cell signaling” and “cytokines/chemokines,” and the KEGG pathway category “environmental information processing,” which consists primarily of signaling pathways. This implies that signaling pathways, when compared to other pathways, are more likely to lose their cohesiveness in tumor-bearing tissues. This is consistent with our previous results where signaling pathways were mostly found not to be cohesive (39). The results obtained here utilize both normal and tumor tissue gene expressions and show that signaling pathways in normal tissues are not inherently incoherent. They are, in fact, significantly coherently expressed in normal tissues, but this coherence is lost for tumor tissues. The KEGG category of “genetic information processing” is the pathway category that shows the least change in their coherence level. Pathways in this category are among the most cohesive pathways, whereas the cancer pathway analysis additionally reveals that these pathways are inherently cohesive in normal tissues, and have maintained their cohesiveness in tumor tissues.

5.2. Pathway Stability as Reflected by Changes in Gene Expression Coherence Because cancer is a disease closely tied to genetic instability, the relative stability of pathways or pathway gene expression regulation is of great interest. The tendency for some pathways to change their cohesiveness may be a reflection of instability in the regulation of these pathways, which can be either the cause or consequence of cancer, and hence warrants further investigation. As a measure of tissue-related pathway stability, we have calculated for each pathway category the average number of tissue types in which a pathway is found to have changed its cohesiveness, either “up” or “down.” A pathway category is considered to be generically unstable if its pathways have changed their cohesiveness significantly (t-test: p < 0.05) above the average for all tissue types, and specifically unstable if their pathways change their cohesiveness only in specific tissue types. For the five KEGG pathway categories, “environmental information processing” is found to be the most generically unstable. This result is consistent with our earlier findings that this pathway category, which consists of mostly signaling

68

Part I / Genomic Experimental Approaches in Oncology

pathways, is also significantly enriched in “down” pathways, that is, they tend to lose their cohesiveness in cancer. On the other hand, the pathways categorized as metabolic are mostly specifically unstable when compared to the other KEGG pathway categories. The stability of other KEGG pathway categories is not significantly different from average. “Cell cycle regulation,” a BioCarta pathway category, is the most specifically unstable of all, with half of its pathways exhibiting no cohesiveness change and each pathway demonstrating change in only one tissue type on average. On the other hand, “adhesion” appears to be the most generically unstable BioCarta pathway category and “cell signaling” is second. These results are consistent with what we find with the KEGG pathways.

5.3. Stability of “Cancer Pathways” Cancer is a genetic disease and genes operate through pathways. Variations in the gene regulation patterns in a pathway, as reflected by the change of cohesiveness in the pathway, may be indicative of pathway instability, which itself may be a result (or cause) of genetic abnormalities. Approximately half of the 962 analyzed pathways contain at least one of the 346 known “cancer genes” according to a recent census of human cancer genes (60). A change in cohesiveness has been found in 25% of all the pathways analyzed, including both “down” and “up” pathways. The question is then whether the pathways that have shown a cohesiveness change in tumor compared to normal tissues can be considered to be cancer pathways based on the likelihood of them containing cancer genes, or whether cancer pathways in general are more likely to change their cohesiveness. The answer to the first questions is “yes,” because cancer genes are found in 57% of the pathways that have shown a change in their cohesiveness, which is significantly higher than the average probability (45%) of a pathway to contain cancer genes. (All comparative statements have been qualified using Fisher’s exact test and have p-values of less than 10−3 .) Moreover, a “down” pathway is found to be much more likely to contain cancer genes than an average pathway. Conversely, we have also found that cancer pathways are more likely to change their cohesiveness, that is, a significantly higher percentage of cancer pathways shows a cohesiveness change than the pathways that do not contain cancer genes (32% vs. 20%). Moreover, the cancer pathways are especially enriched in “down” pathways (22% vs. 11%), indicating that these pathways tend to become deregulated or dysfunctional in cancer. The fact that cancer pathways tend to become less cohesive in tumor-bearing tissues may be exploited to find more cancer pathways or cancer genes; that is, other pathways that are shown to lose their cohesiveness in cancer, but are not known to contain cancer genes, may be additionally interesting pathways to be considered as therapeutic targets, for example, pathways involving fatty acid metabolism, basal transcription factors, and glycolysis/gluconeogenesis. The instability of cancer pathways is also reflected by the number of tissue types in which they have shown a coherence change. Cancer pathways containing genes with germline or somatic mutations ( 66) exhibit a coherence change in significantly more tissue types than an average pathway (t-test: germline: p = 1.86 × 10–7 ; somatic:

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

69

p = 4.96 × 10–5 ), indicating that cancer pathways are generically less stable than other pathways. Interestingly, regardless of the fact that individuals with a germline mutation carry that mutation in every cell of their body—whereas somatic mutations are found only in an individual’s cancer cells (61,62) the tissue promiscuity/specificity of pathway instability is found to be not significantly different between pathways containing germline cancer genes and those containing somatic cancer genes This offers an explanation of the tissue specificity of gene defects, that is, even though the vast majority of inherited cancer genes appear to be expressed in most adult tissues, a germline mutation in these genes is manifested in only a limited spectrum of cancers. We have found that most pathways, including the ones containing cancer genes, only show a coherence change in a small fraction of tissue types (15%–20%), indicating that probably not all mutations will be translated into pathway instability, and only in the tissue types where the stability of pathways is compromised by a mutation will cancer arise. Just as mutated proteins may render a pathway unstable or deregulated, it is equally probable that pathway instability itself may be the contributing factor to certain gene mutations. Unstable pathways are probably more susceptible to environmental influences, which can trigger or facilitate mutations. The final result will be a malignant cycle that promotes uncontrolled growth. The implication that pathway instability may be the cause of genetic instability makes pathways as a system of interactions, rather than individual genes, an interesting, albeit diffuse, target in itself for therapy considerations.

5.4. Strategies to Target Change in Anti-Cancer Therapies by Finding Agents That Can Potentially Perturb an Unstable Pathway Pathways that tend to change their cohesiveness in tumor-containing versus normal tissues represent interesting targets for anti-cancer therapy, making it highly desirable to locate agents with the potential to specifically disrupt these pathways. The previously discussed relationship between compound cytotoxic response to pathways and MOAs through gene expression patterns can now be brought to bear on specific cancer pathway targets ( 58). Compound clusters derived from growth inhibition data (GI50 ) that are significantly correlated with a pathway can potentially perturb that pathway. Therefore, for each of the pathways with a change in gene expression coherence, we have found the SOM cluster that is associated with the pathway (58). The compounds that are the most significantly correlated with these pathways are mostly clustered in the SOM regions F, P, V, and N. This is not surprising because we have found previously that the P-region contains agents with kinase/phosphotase targeting as their putative MOA (18,79), and many of the pathways that tend to change their cohesiveness are signaling pathways. Some clades contain compounds that can potentially perturb both “up” and “down” pathways, such as clades F2 , P1 , and P2 , whereas other clades contain compounds that appear to perturb specifically “up” or “down” pathways. Most notably, compounds clustered in clades P4 and M3 are associated only with “up” pathways, and those in clades N9 , P6 , F6 , F8 , and V1 are associated predominantly with “down” pathways and with very few or no “up” pathways. Most clinically used compounds in anti-cancer therapies are located in regions M and S, essentially targeting the proliferation stage of a cancer cell. Our analysis, however, points to a much more

70

Part I / Genomic Experimental Approaches in Oncology

diverse set of compounds different from traditional cancer therapy and with the potential to disrupt a wide spectrum of cellular processes that are characteristic of cancer cells.

6. CONCLUSION In the pharmacogenomic study of NCI60 we have developed strategies to analyze pathway stability by comparing pathway gene expression coherence levels in tumor tissues to their normal counterparts. Changes occurring in cancer as reflected by variations in pathway coherence are considered to be indicative of pathway instability and possibly genetic instability. We have identified pathways that show a significant change in their coherence level in tumor tissues in general, as well as specific changes in certain tumor types. These pathways may represent good targets for developing novel anti-cancer therapies. Significantly more pathways are found to lose their coherence in tumor tissues. Signal transduction represents the most unstable pathway category, whose coherence in gene regulation is largely lost in tumor tissues. In contrast, pathways responsible for vital cellular processes are mostly able to maintain their gene expression coherence in tumor tissues and are among the most stable. The combination of homeostatic control over critical pathways to ensure survival, and altered regulation of signaling to allow excessive proliferation, forms the foundation for the selective growth advantage of cancer cells over normal cells. The instability of metabolic and cell cycle regulating pathways appear to be the most tissue-specific. The function of these pathways and their unstable tissue type may provide important clues for finding the molecular mechanisms underlying specific cancer types. We have examined the particular pathways that contain known cancer genes and compared their behavior with other pathways. Cancer pathways are found more likely to lose their coherence and thus show a greater level of instability than an average pathway. Finally, we have proposed strategies to target these changes, that is, to find new agents that can specifically target the unstable pathways that may be relevant in cancer.

ACKNOWLEDGEMENTS This project has been funded in whole or in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract NO1-CO-12400. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organization imply endorsement by the U.S. Government. This research was supported in part by the Developmental Therapeutics Program in the Division of Cancer Treatment and Diagnosis of the National Cancer Institute.

REFERENCES 1. Capranico G. A rational selection of drug targets needs deeper insights into general regulation mechanisms. Curr Med Chem Anti-Canc Agents 2004;4:393–394. 2. Klein CA. Gene expression signatures, cancer cell evolution and metastatic progression. Cell Cycle 2004;3:29–31. 3. Covitz PA. Class struggle: expression profiling and categorizing cancer. Pharmacogenomics J 2003;3:257–260.

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

71

4. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000, 100, 57–70. 5. Rennefahrt U, Janakiraman M, Ollinger R et al. tress kinase signaling in cancer: fact or fiction? Cancer Lett 2005;217:1–9. 6. Halvorsen OJ, Oyan AM, Bo TH et al. Gene expression profiles in prostate cancer: association with patient subgroups and tumour differentiation. Int J Oncol 2005;26:329–336. 7. Cary MP, Bader GD, Sander C. Pathway information for systems biology. FEBS Lett 2005;579: 1815–1820. 8. Apic G, Ignjatovic T, Boyer S et al. Illuminating drug discovery with biological pathways. FEBS Lett 2005;579,:1872–1877. 9. Huang S. Back to the biology in systems biology: what can we learn from biomolecular networks? Brief Funct Genomic Proteomic 2004;2:279–297. 10. Wolkenhauer O, Ullah M, Wellstead P et al. The dynamic systems approach to control and regulation of intracellular networks. FEBS Lett 2005;579:1846–1853. 11. Khalil IG, Hill C. ystems biology for cancer. Curr Opin Oncol 2005;17:44–48. 12. Monks A, Scudiero D, Skehan P et al. Feasibility of a high-flux anti-cancer drug screen using a diverse panel of cultured human tumor cell lines. J Natl Cancer Inst 1991;83:757–766. 13. Keskin O, Bahar I, Jernigan RL et al. Characterization of anti-cancer agents by their growth inhibitory activity and relationships to mechanism of action and structure. Anti-Cancer Drug Design 2000;15:79–98. 14. Rabow AA, Shoemaker RH, Sausville EA et al. Mining the National Cancer Institute’s tumor– screening database: Identification of compounds with similar cellular activities. J Med Chem 2002;45:818–840. 15. Paull KD, Shoemaker RH, Hodes L et al. Display and analysis of patterns of differential activity of drugs against human tumor cell lines: development of mean graph and COMPARE algorithm. J Natl Cancer Inst 1989;81:1088–1092. 16. Kohonen T. Self-Organizing Maps; Springer-Verlag: New York, 1995. 17. Covell DG, Wallqvist A, Huang R et al. Linking tumor cell cytotoxicity to mechanism of drug action: an integrated analysis of gene expression, small-molecule screening and structural databases. Proteins 2005;59:403–433. 18. Rabow AA, Shoemaker RH, Sausville EA et al. Mining the National Cancer Institute’s tumorscreening database: identification of compounds with similar cellular activities. J Med Chem 2002;45:818–840. 19. Huang R, Wallqvist A, Covell DG. Anti-cancer metal compounds in NCI’s tumor-screening database: putative mode of action. Biochem Pharmacol 2005;69:1009–1039. 20. Ihmels J, Bergmann S, Barkai N. Defining transcription modules using large-scale gene expression data. Bioinformatics 2004;20:1993–2003. 21. Ihmels J, Levy R, Barkai N. Principles of transcriptional control in the metabolic network of Saccharomyces cerevisiae. Nat Biotechnol 2004;22:86–92. 22. Li Z, Chan C. Inferring pathways and networks with a Bayesian framework. Faseb J 2004;18:746–748. 23. Li Z, Chan C. Integrating gene expression and metabolic profiles. J Biol Chem 2004;279: 27124–27137. 24. Yang HH, Hu Y, Buetow KH et al. A computational approach to measuring coherence of gene expression in pathways. Genomics 2004;84:211–217. 25. Williams EJ, Bowles DJ. Coexpression of neighboring genes in the genome of Arabidopsis thaliana. Genome Res 2004;14:1060–1067. 26. Caron H, Peter M, van Sluis P et al. Evidence for two tumor suppressor loci on chromosomal bands 1p35–36 involved in neuroblastoma: one probably imprinted, another associated with N-myc amplification. Hum Molec Genet 1995;4:535–539. 27. Lercher MJ, Urrutia AO, Hurst LD. Clustering of housekeeping genes provides a unified model of gene order in the human genome. Nature Gen 2002, 31, 180–183. 28. Cohen BA, Mitra RD, Hughes JD et al. A computational analysis of whole-genome expression data reveals chromosomal domains of gene expression. Nature Genet 2000;26:183–186.

72

Part I / Genomic Experimental Approaches in Oncology

29. Boutanaev AM, Kalmykova AI, Shevelyov YY et al. Large clusters of co-expressed genes in the Drosophila genome. Nature (UK) 2002;420:666–669. 30. Spellman PT, Rubin GM. Evidence for large domains of similarly expressed genes in the Drosophila genome. J Biol 2002;1:5. 31. Lercher MJ, Blumenthal T, Hurst LD. Co-expression of neighboring genes in Caenorhabditis elegans is mostly due to operons and duplicate genes. Genome Res 2003;13:238–243. 32. Ge H, Liu, Z, Church GM et al. Correlation between transcriptome and interactome mapping data from Saccharomyces cerevisiae. Nature Genet 2001;29:482–486. 33. Staudt LM, Brown PO. Genomic views of the immune system. Annu Rev Immunol 2000;18:829–859. 34. Grigoriev A. A relationship between gene expression and protein interactions on the proteome scale: analysis of the bacteriophage T7 and the yeast Saccharomyces cerevisiae. Nucleic Acids Res 2001;29:3513–3519. 35. Jansen R, Greenbaum D, Gerstein M. Relating whole-genome expression data with protein-protein interactions. Genome Res 2002;12:37–46. 36. Miki R, Kadota K, Bono H et al. Delineating developmental and metabolic pathways in vivo by expression profiling using the RIKEN set of 18,816 full-length enriched mouse cDNA arrays. Proc Natl Acad Sci USA 2001;98:2199–2204. 37. Hannenhalli S, Levy S. Transcriptional regulation of protein complexes and biological pathways. Mamm Genome 2003;14:611–619. 38. Ptashne M, Gann A. Imposing specificity by localization: mechanism and evolvability. Curr Biol 1998;8:R812–822. 39. Huang R, Wallqvist A, Covell DG. Comprehensive analysis of pathway or functionally related gene expression in the National Cancer Institute’s anti-cancer screen. Genomics 2006;87:315–328. 40. Butte AJ, Tamayo P, Slonim D et al. Discovering functional relationships between RNA expression and chemotherapeutic susceptibility using relevance networks. Proc Natl Acad Sci USA 2000;97:12182–12186. 41. Szakacs G, Annereau JP, Lababidi S et al. Predicting drug sensitivity and resistance: profiling ABC transporter genes in cancer cells. Cancer Cell 2004;6:129–137. 42. Huang Y, Anderle P, Bussey KJ et al. Membrane transporters and channels: role of the transportome in cancer chemosensitivity and chemoresistance. Cancer Res 2004;64:4294–4301. 43. Blower PE, Yang C, Fligner MA et al. Pharmacogenomic analysis: correlating molecular substructure classes with microarray gene expression data. Pharmacogenomics J 2002;2:259–271. 44. Zhou Y, Gwadry FG, Reinhold WC et al. Transcriptional regulation of mitotic genes by camptothecininduced DNA damage: microarray analysis of dose- and time-dependent effects. Cancer Res 2002;62:1688–1695. 45. Lee JK, Scherf U, Smith LH et al. Analysis of gene expression data of the NCI 60 cancer cell lines using Bayesian hierarchical effects model. Proceedings of SPIE–The International Society for Optical Engineering 2001;4266:228–235. 46. Scherf U, Ross DT, Waltham M et al. A gene expression database for the molecular pharmacology of cancer. Nature Genet 2000;24: 236–244. 47. Wosikowski K, Schuurhuis D, Johnson K et al. Identification of epidermal growth factor receptor and c-erbB2 pathway inhibitors by correlation with gene expression patterns. J Natl Cancer Inst 1997;89:1505–1515. 48. O’Connor PM, Jackman J, Bae I et al. Characterization of the p53 tumor suppressor pathway in cell lines of the National Cancer Institute anti-cancer drug screen and correlations with the growthinhibitory potency of 123 anti-cancer agents. Cancer Res 1997;57:4285–4300. 49. Alvarez M, Paull K, Monks A et al. Generation of a drug resistance profile by quantitation of mdr-1/ P-glycoprotein in the cell lines of the National Cancer Institute Anti-cancer Drug Screen. J Clin Invest 1995;95:2205–2214. 50. Li KC, Yuan S. A functional genomic study on NCI’s anti-cancer drug screen. Pharmacogenomics J 2004;4:127–135.

Chapter 5 / Strategies for Mining the Chemistry-Genetics-Biology Interface

73

51. Wallqvist A, Rabow AA, Shoemaker RH et al. Linking the growth inhibition response from the National Cancer Institute’s anti-cancer screen to gene expression levels and other molecular target data. Bioinformatics 2003;19:2212–2224. 52. Freije JMP, Lawrence JA, Hollingshead MG et al. Identification of compounds with preferential inhibitory activity against low-NM23-expressing human breast carcinoma and melanoma cell lines. Nature Med 1997;3:395–401. 53. Ficenec D, Osborne M, Pradines J et al. Computational knowledge integration in biopharmaceutical research. Brief Bioinformatics 2003;4:260–278. 54. Huang Y, Blower PE, Yang C et al. Correlating gene expression with chemical scaffolds of cytotoxic agents: ellipticines as substrates and inhibitors of MDR1. Pharmacogenomics J 2005;5:112–125. 55. Nakatsu N, Yoshida Y, Yamazaki K et al. Chemosensitivity profile of cancer cell lines and identification of genes determining chemosensitivity by an integrated bioinformatical approach using cDNA arrays. Mol Cancer Therapeutics 2005;4:399–412. 56. Stegmaier K, Ross KN, Colavito SA et al. Gene expression–based high-throughput screening (GE-HTS) and application to leukemia differentiation. Nat Genet 2004;36:257–263. 57. Huang R, Wallqvist A, Covell DG. Comprehensive analysis of pathway or functionally related gene expression in the National Cancer Institute’s anti-cancer screen. Genomics 2006;87:315–328. 58. Huang R, Wallqvist A, Thanki N et al. Linking pathway gene expressions to the growth inhibition response from the National Cancer Institute’s anti-cancer screen and drug mechanism of action. Pharmacogenomics J 2005;5:381–399. 59. Westwell AD, Stevens MF. Hitting the chemotherapy jackpot: strategy, productivity, and chemistry. Drug Discov Today 2004;9:625–627. 60. Futreal PA, Coin L, Marshall M et al. A census of human cancer genes. Nat Rev Cancer 2004;4: 177–183. 61. Knudson AG. Cancer genetics. Amer J Med Genet 2002;111, 96–102. 62. Fearon ER. Human cancer syndromes: clues to the origin and nature of cancer. Science (Washington, DC) 1997;278:1043–1050. 63. Loeb LA, Loeb KR, Anderson JP. Multiple mutations and cancer. Proc Natl Acad Sci USA 2003;100:776–781. 64. Rajagopalan H, Nowak MA, Vogelstein B et al. The significance of unstable chromosomes in colorectal cancer. Nat Rev Cancer 2003;3, 695–701. 65. Sieber OM, Heinimann K, Tomlinson IP. Genomic instability: the engine of tumorigenesis? Nat Rev Cancer 2003;3:701–708. 66. Vogelstein B, Kinzler KW. Cancer genes and the pathways they control. Nat Med 2004;10:789–799. 67. Feinberg AP, Tycko B. The history of cancer epigenetics. Nat Rev Cancer 2004;4:143–153. 68. Jones PA, Baylin SB. The fundamental role of epigenetic events in cancer. Nat Rev Genet 2002;3: 415–428. 69. Polyak K, Riggins GJ. Gene discovery using the serial analysis of gene expression technique: implications for cancer research. J Clin Oncol 2001;19:2948–2958. 70. Brown PO, Botstein D. Exploring the new world of the genome with DNA microarrays. Nat Genet 1999;21:33–37. 71. Schadt EE, Lamb J, Yang X et al. An integrative genomics approach to infer causal associations between gene expression and disease. Nat Genet 2005;37:710–717. 72. Rosenblatt KP, Bryant-Greenwood P, Killian JK et al. erum proteomics in cancer diagnosis and management. Annu Rev Med 2004;55:97–112. 73. Ma XJ, Wang Z, Ryan PD et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell 2004;5:607–616. 74. Collins FS, Green ED, Guttmacher AE et al. A vision for the future of genomics research. Nature 2003;422:835–847. 75. Schadt EE, Monks SA, Friend SH. A new paradigm for drug discovery: integrating clinical, genetic, genomic and molecular phenotype data to identify drug targets. Biochem Soc Trans 2003;31:437–443.

74

Part I / Genomic Experimental Approaches in Oncology

76. Wallqvist A, Connelly J, Sausville EA et al. Differential gene expression as a potential classifier of 2-(4-amino-3-methylphenyl)-5-fluorobenzothiazole–sensitive and –insensitive cell lines. Mol Pharmacol 2006;69:737–748. 77. Ramaswamy S, Tamayo P, Rifkin R et al. Multiclass cancer diagnosis using tumor gene expression signatures. Proc Natl Acad Sci USA 2001;98:15149–15154. 78. Covell DG, Wallqvist A, Rabow A et al. Molecular classification of cancer: unsupervised selforganizing map analysis of gene expression microarray data. Mol Cancer Ther 2003;2:317–332. 79. Wallqvist A, Monks A, Rabow AA et al. Mining the NCI screening database: explorations of agents involved in cell cycle regulation. Prog Cell Cycle Res 2003;5:173–179.

6

Use of Single-Nucleotide Polymorphism Array for Tumor Aberrations in Gene Copy Numbers Kwong-Kwok Wong, PhD CONTENTS Introduction Challenges of Performing an SNP Array Analys is on Tumor Samples Software to Vis ualize and Es timate Copy Number Variations from SNP Array Data Validation of SNP Array Data Future Pers pective References

S UMMARY The single nucleotide polymorphism (SNP) array was originally developed to determine the genotypes of a study population for linkage analysis or individual genetic variation analysis. Over the last few years, the number of SNP loci that can be evaluated in a single assay has increased from approximately 1500 to more than 500,000, covering all 22 autosomes and the X chromosome. Because the hybridization signal of each oligonucleotide on the SNP array depends on the amount of target DNA, various statistic algorithms have been developed to estimate the copy number of each SNP locus. Several studies have demonstrated the utility of the SNP array analysis in the detection of gene copy number aberrations in tumor DNA. In this review, we discuss the use of the SNP array analysis in determining tumor aberrations in gene copy numbers, the challenge of using tumor samples in the analysis, and improvements in From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 6, 

75

76

Part I / Genomic Experimental Approaches in Oncology

software for copy number measurement using SNP arrays. The unique ability of the SNP array to determine both the genotype and copy number has uncovered novel DNA amplification events that involve only a single allele. Key Words: single nucleotide polymorphism array; allelic imbalance; copy number variation; whole-genome amplification; formalin-fixed and paraffin-embedded tissue

1. INTRODUCTION Malignant tissues frequently exhibit chromosomal aberrations and altered gene expression. The altered transcript levels in cancer genomes are often related to changes in the number of gene copies, which may result from amplification of oncogenes or inactivation of tumor suppressor genes (TSGs), as detected by homozygous deletion or loss of heterozygosity (LOH). LOH and DNA copy number changes have been associated with prognosis (1,2,3), treatment responses ( 4, 5, 6, 7) and cancer progression ( 8, 9). Genes located in chromosomal regions with aberrations can be therapeutic targets, because they often play important roles in multiple genetic pathways that regulate cell growth, proliferation, apoptosis, and metastasis ( 10). Many tumor suppressor and oncogene loci have been identified in recurrently deleted or amplified chromosomal regions. For example, many TSGs, including RB1 ( 11), p16 ( 12) and PTEN ( 13) have been identified in regions of recurrent homozygous deletion. Myc (14), EGFR (15), ERBB2, and ERBB1 ( 16) have also been found in regions of chromosome amplification. Thus, identifying cancer-specific copy number alterations not only provides new insight into the molecular basis of tumorigenesis, but also facilitates the discovery of new TSGs and oncogenes. In the past, LOH patterns were detected by whole-genome allelotyping using either restriction fragment length polymorphism (17) or microsatellite markers (18). Restriction fragment length polymorphism assays have largely replaced by the use of polymorphic microsatellite markers. When using polymorphic microsatellite markers, the resolution for whole-genome scanning is limited to 5 cM. Moreover, the analysis is often long and tedious. Genome-wide scanning of chromosomes by metaphase comparative genomic hybridization (CGH) ( 19) with a resolution of only 20 Mb, has revealed many chromosome copy number aberrations in tumor cells. For example, a study has shown an association among 1p loss, 17q gain, and N-Myc amplification in stage IV neuroblastoma (20). Both of these techniques also require micrograms of genomic DNA, which are rarely available from small tissue samples. With the discovery of more than 5 million single nucleotide polymorphisms (SNPs) distributed throughout the human genome at a mean density of one SNP per kilobase of DNA sequence ( 21), high-resolution genome-wide allelotyping and copy number measurement has become a reality. The use of the high-density SNP array has led to substantial improvements in chromosomal resolution and throughput over the past few years, and parallel genotyping of more than 500,000 SNPs using a one-primer assay is now feasible ( 22, 23, 24). More than 1000 cell lines and tumor samples

Chapter 6 / Single-Nucleotide Polymorphism Array

77

have undergone whole-genome copy number variation analysis with the SNP array (www.sanger.ac.uk/genetics/CGP).

1.1. SNP Array Analysis SNPs are variable positions in a genome; they have different allele types that differ by a single base. There is an estimated one SNP every 300 nucleotides in the approximately 3 billion base pairs of the human genome ( 25, 26). They can be used as measures of genetic diversity in humans. Presently, more than 5 million SNPs have been mapped (21). An individual can have two alleles of the same sequence (homozygous) or alleles of different sequences (heterozygous) at each SNP locus. The biotechnology company, Affymetrix, Inc., has been developing commercialized SNP arrays for whole-genome genotyping since 1999. Preselected SNPs spaced throughout the genome are typed. For each SNP locus, two alleles of the highest heterozygosity in a tested population are selected. Each allele is analyzed by 20 oligonucleotides (10 perfect-match probes and 10 probes with a single-base mismatch to the allele) that are 25 bases long. On the basis of the hybridization signals of each allele, the genotypes of thousands of SNP loci are then determined using different algorithms (27,28). The possible genotype calls are AB (heterozygous), AA (homozygous), BB (homozygous), or No call (Table 1). The number of SNP loci that can be simultaneously genotyped across the human genome with an SNP array has increased from approximately 1500 to 500,000 (22,23,24,29). A 1 million-SNP array may be available in the near future. By comparing the genotypes of a patient’s blood and tumor DNA, we can easily detect SNP loci with LOH. A genotype that changes from heterozygous in the blood DNA to homozygous in the tumor DNA indicates an LOH event (Table 1). For tumor samples without matching normal DNA, the SNP loci with LOH can be inferred from the probability that a contiguous stretch of homozygous calls will happen by chance, as described previously (30). Recently, a hidden Markov model-based method was developed to identify LOH from tumor samples alone, taking into account SNP intermarker distances, SNP-specific heterozygosity rates, and the haplotype structure of the human genome ( 31) to filter out false-positive LOH. When both parents share the same haplotype, the children will inherit a long stretch of homozygous geneotypes (32). Table 1 LOH by Genotype Call in Blood and Tumor DNA Blood genotype

Tumor genotype Heterozygous

Homozygous

No genotype call

Heterozygous

No change

LOH

Homozygous

Conflict

No genotype call

Conflict

No change or loss of one allele Conflict

Partial loss of one allele or both alleles Partial loss of one allele or both alleles Consistent

78

Part I / Genomic Experimental Approaches in Oncology

Because the hybridization signal of each oligonucleotide on the SNP array depends on the copy number of the target DNA, various statistic algorithms have been developed to estimate the copy number for each SNP locus ( 33, 34, 35, 36). Thus, in addition to genotyping, SNP arrays have been used extensively to measure tumor aberrations in copy numbers. The high density, relatively uniform distribution, and allele specificity of SNPs in these arrays make them attractive for high-resolution analyses of LOH and copy number alterations in cancer genomes (33,37,38,39,40,41,42,43).

1.2. Advantages of the SNP Array The SNP array has several advantages over traditional methods such as karyotyping and allelotyping with microsatellite markers. First, it requires no cell cultures, eliminating some of the karyotyping problems that arise from passaging cells. Cultures from primary tumors are often contaminated with normal cells, which can proliferate and thus complicate the karyotyping results. Second, the SNP array has a high SNP density for genotyping and copy number measurements. For example, on the GeneChip human mapping 500 K array, the median physical distance between SNPs is 2.5 kb, and the mean distance is 5.8 kb. Eighty-five percent of the human genome is within 10 kb of a SNP, which provides a mapping resolution that is 100- to 300-fold higher than the current panels of ∼400 short tandem repeat markers (18). Moreover, the mean heterozygosity of these SNPs is 0.30, which generates many heterozygous informative genotype calls for inferring the LOH region. The SNP array has been shown to be highly accurate (99.5%) and reproducible (99.9%) and has a high call rate (95.0%) in genotyping ( 23). It can also be used to detect changes in the chromosome copy number at a single SNP locus resolution, as each SNP locus is analyzed with 40 SNP probes. The hybridization signals of these SNP probes depend on the copy numbers of DNA at each SNP locus. A statistical model has been developed that calculates the likelihood of determining LOH, SNP copy number changes, and chromosome copy number changes in one analysis without using normal DNA as the reference (43,44). This model can reliably detect amplifications and homozygous deletions that extend over regions of less than 1 Mb. In several cancer studies, it has been used to successfully determine LOH in newly identified and previously known chromosome regions (2,43,45,46,47,48,49). Figure 1 illustrates a simultaneous detection of copy number changes and LOH using the 10 K SNP array, as presented in our recent paper (49). The ability of the SNP array analysis to detect copy number changes and LOH simultaneously from a tumor sample has an added advantage over CGH. In several studies, different LOH mechanisms (i.e., LOHs with and without copy number changes) have been identified using SNP arrays. LOH detection with an SNP array analysis is based on genotyping calls and can thus detect the loss of one allele, followed by the reduplication of the remaining allele as LOH. We have found that amplified EGFR and PDGFR␣ genes are located in regions of LOH (Fig. 2), which implies that a specific allele is amplified ( 49). A sequencing analysis of the EGFR transcript in the tumor confirmed that the amplified EGFR allele carried a type III mutation and was the only amplified allele in the LOH region. This may represent a new mechanism of tumor progression ( 49). Similarly, another recent SNP array study of 100 cases of lung cancer revealed

Chapter 6 / Single-Nucleotide Polymorphism Array

79

Fig. 1. Side-by-side whole-genome LOH patterns and copy number changes in 28 pediatric gliomas. Top: The sample clustering tree is based on LOH data in the significant regions. LOH regions in each tumor are highlighted in dark gray color. Bottom: A darker color indicates a higher copy numbers at the corresponding regions. The graph was generated by dChip2006 software. (Reprinted with permission of the American Association for Cancer Research from Wong et al., Cancer Res 2006; 66(23): 11172–11178).

Fig. 2. An enlarged region of Fig. 1 showing amplification of specific alleles in LOH regions (49).

80

Part I / Genomic Experimental Approaches in Oncology

that DNA amplification is derived from a single allele or the remaining allele located at LOH region (50). In contrast, CGH is based on the ratio of the DNA copy number to a reference DNA and thus cannot detect reduplication of the remaining allele in the case of LOH, which will be detected as no change.

2. CHALLENGES OF PERFORMING AN SNP ARRAY ANALYSIS ON TUMOR SAMPLES Archived tumor samples are available as either frozen or formalin-fixed, paraffinembedded (FFPE) tissue. FFPE samples are usually more readily available than are frozen samples. Tumor samples are usually contaminated with normal stromal cells, resulting in a dilution effect, with chromosomal abnormalities of tumor cells masked by the presence of normal cells. As a result, the genotype’s accuracy and sensitivity to copy number changes will be affected by the percentage of normal cells contaminating the tumor tissue. Laser capture microdissection circumvents this problem by allowing for visualization and thus procurement of a homogeneous population of tumor cells, increasing the assay’s sensitivity to chromosomal abnormalities ( 51). On the other hand, a whole-genome copy number SNP array analysis using PPFE samples still need optimization.

2.1. Use of Whole Genome-Amplified DNA for SNP Array Analysis The amount of DNA available from tissue samples, especially samples from tumor biopsies, is usually limited. Fortunately, whole-genome amplification has been developed to generate micrograms of DNA for further analysis ( 52, 53). We found similar genotype calls and copy number changes in whole-genome amplified DNA with that using unamplified DNA, as determined by a 10 K SNP array analysis (30). Figure 3 shows a comparison between DNA extracted directly from tumor tissue without microdissection and whole-genome amplified DNA from microdissected tumor cells. LOH was detected in microdissected samples only. Thus, an increase in the sensitivity to copy number aberrations can be achieved using whole genome-amplified DNA from microdissected samples for SNP array analysis.

2.2. Use of Paraffin Embedded Material for SNP Array Analysis DNA fragments extracted from FFPE tissues are usually degraded or small. However, aberrations in copy numbers have been reported using FFPE-derived DNA with the firstgeneration SNP arrays—GeneChip HuSNP (40,45,54) or GeneChip mapping 10 K array (55). The assay for the HuSNP array uses 24 multiplex polymer chain reactions (PCRs) to generate labeled targets for hybridizing with 1494 SNP probe sets. Multiplex PCR amplification can apparently tolerate degraded DNA because the size of PCR products generated for subsequent hybridization is 100–450 base pairs. However, the 10 K mapping array and the 100 K and 500 K arrays use a specific restriction enzyme to digest the genomic DNA for subsequent ligation with a specific DNA linker. The DNA linkers then act as binding sites for the specific primer to initiate PCR amplification

Chapter 6 / Single-Nucleotide Polymorphism Array

81

Fig. 3. The use of microdissected material increased the sensitivity of the SNP array to LOH in tumor tissues. Regions with LOH are highlighted in black color. B, DNA extracted from bulk tumor tissue without microdissection; M, DNA was extracted from microdissected tumor cells and amplified by the multiple displacement amplification method (52).

and labeling. Most DNA fragments extracted from FFPE samples are less than 0.5 kb and will thus lack two XbaI (for the 10 K array) or HindIII (for the 50 K HindIII SNP array) restriction sites. As a result, many SNP loci will not be amplified as labeled targets, thus tremendously decreasing the genotype call rate and the level of hybridization signals. From our experience, we can obtain a call rate as high as 85%, but most of the time is around 50% when using DNA extracted from most FFPE samples (unpublished data). Further improvements to the SNP array analysis of FFPE samples are still needed.

3. SOFTWARE TO VISUALIZE AND ESTIMATE COPY NUMBER VARIATIONS FROM SNP ARRAY DATA The utility of SNP arrays for determining genome-wide tumor aberrations in gene copy numbers depends, in part, on the progress made in software development. The copy number analysis tool (33), developed by Affymetrix, is an integrated component of the GeneChip operating system and is used to identify genome-wide chromosome copy number gains and losses. The copy number estimation is based on a reference file generated from more than 100 samples for each type of SNP arrays. The reference file contains the log intensity data for each SNP and genotype. The log intensity of

82

Part I / Genomic Experimental Approaches in Oncology

each SNP in the reference set will determine a copy number of two. By comparing the normalized intensity values of an SNP locus from a tumor sample with the corresponding SNP intensity data in the reference file, we can determine whether copy number changes exist (Fig. 4). However, the normalized SNP intensity data are computed from the 20 perfect-match probes used to evaluate both alleles; thus, allele-specific copy number changes will not be detected.

Fig. 4. The copy numbers of individual SNPs in chromosome 6, as detected with amplified and unamplified tumor DNA from case OST197. The copy number results are plotted using light gray color for values above the threshold (2) and dark gray color for values below the threshold. Included in the graph is a representation of the genotype calls associated with the SNPs (small color bars to the right of the ideogram). Light gray color represents heterozygous calls, and dark gray color represents homozygous calls. An ideogram showing the corresponding cytoband locations of each SNP on chromosome 6 is shown below the graph. (Reprinted with permission of the Oxford University Press for Nucleic Acids Research from Wong et al., Nucleic Acids Res 2004;32:e69).

Chapter 6 / Single-Nucleotide Polymorphism Array

83

Other commonly used software developed by the academic community for the analysis of SNP array data include dChip ( 44), CNAG ( 34), and PLASQ ( 35, 50). dChip was originally developed as an analytic and visualization tool for gene expression array data. Using a similar model-based method for estimation of hybridization signals, dChip can infer the copy number changes in tumor samples by comparing the signal intensity of each SNP in the tumor DNA to the corresponding mean signal intensity from a set of reference samples. The reference samples are derived from normal tissues and are signed to have a ploidy of two. Using the genotype information associated with each tumor sample, LOH and copy number changes can be visualized simultaneously (Fig. 1). CNAG uses a robust algorithm to improve the signal-to-noise ratios and help select an optimal reference. Accounting for the length and GC content of the PCR products using quadratic regressions has resulted in substantial improvements to the raw signal-tonoise ratios (34). When matched normal DNA is available, an allele-based copy number can also be determined in cancer genomes. On the other hand, when matched normal DNA is not available, the optimal selection of multiple normal references with the highest signal-to-noise ratios and lowest standard deviations has been shown to improve the accuracy of estimations of the range and magnitude of copy number aberrations (24,34,56). Another software program, PLASQ (probe-level allele-specific quantitation) (35,50), is used in the statistical environment R (www.r-project.org/) to extract both copy number and allelotype information from SNP array data to determine allele-specific copy numbers across the genome. Using an expectation–maximization algorithm for a model derived from a novel classification of SNP array probes, the copy number of each parental chromosome across the genome can be accurately determined. With this, one can not only identify chromosome regions with amplifications and deletions but also the haplotype of the regions being amplified or deleted. PLASQ was used to determine that DNA amplification is essentially monoallelic (49), suggesting that a specific parental chromosome may be targeted for amplification because of germline or somatic variation (49).

4. VALIDATION OF SNP ARRAY DATA LOH and copy number changes at a specific SNP locus can be validated by direct sequencing of PCR products derived from SNP loci and quantitative PCR using a primer that targets the SNP locus, respectively (49). The DNA sequences of SNP loci can be retrieved from dbSNP databases ( 21) using the reference SNP identification number. The primers that flank the SNP locus can be designed, using the software program Primer3 (57) to generate PCR products of 100–150 base pairs, which covers the SNP locus. The PCR product is sequenced, and the presence of two alleles in the sequencing chromatogram of the blood DNA but only one allele in the sequencing chromatogram of tumor DNA confirms an LOH event at the corresponding SNP locus (Fig. 5). When tissue sections are available, copy number changes can be confirmed by fluorescent in situ hybridization (56,58).

84

Part I / Genomic Experimental Approaches in Oncology

Fig. 5. Validation of SNP loci with LOH by PCR amplification and direct sequencing of both blood and tumor DNA. Both alleles were present in the blood DNA as heterozygous. However, only one allele remained in the tumor DNA.

5. FUTURE PERSPECTIVE Extraordinary advances in SNP array technology and copy number analysis tools have been made over the past few years. SNP array technology is fast becoming an indispensable tool for discovering new tumor aberrations in gene copy numbers. More importantly, determining the relationship between gene copy number aberrations and drug responses, tumor progression, and other clinical data will provide a basis for individualized medicine, as has already been demonstrated in a few studies (59,60,61). The integration of large-scale copy number aberrations with gene expression profiles will facilitate the interpretation of gene expression data. Whether a gene is considered up-regulated or down-regulated in tumor samples frequently depends on the reference chosen (62). The relationship between the copy number gain or loss and gene expression will thus provide a molecular basis for the interpretation of up-regulation or down-regulation of gene expression in tumors. However, large-scale copy number variations must be considered in the interpretation of tumor aberrations in gene copy numbers. Common deletion polymorphisms ( 63, 64) and large-scale copy number variations ( 65, 66, 67, 68, 69) have been identified in normal human genomes. As a result, when normal DNA from a patient is not available, gene copy number aberrations detected in tumor DNA may include gene copy number variations that already exist in the normal population. In the future, the integration of gene expression data and DNA copy number aberrations will facilitate the identification of prognostic, diagnostic, and therapeutic targets for various tumors. The use of SNP arrays to analyze low-quality DNA from FFPE tissue is still challenging and will continue to be an important area of ongoing research. An alternative

Chapter 6 / Single-Nucleotide Polymorphism Array

85

high-throughput genotyping method that uses Illumina BeadArrays to detect tumor aberrations in gene copy numbers is currently being developed (70,71,72,73,74) and seems promising. A recent study demonstrated the utility of BeadArrays to generate identical genotypes as well as LOH/allelic imbalances from both FFPE and frozen tumor tissues (71). Furthermore, the LOH profiles of the BeadArrays were identical to those obtained by Affymetrix GeneChip 10 K arrays. Currently, three high-density SNP genotyping BeadChips are available: the Sentrix Human-1 Genotyping BeadChip, which contains more than 109 K exon-centric SNPs; the HumanHap300 BeadChip, which contains more than 317 K tagged SNPs; and the HumanHap550 BeadChip, which contains more than 550 K tagged SNPs ( 73). Until whole-genome sequencing of the 3 billion base pairs of tumor genomes becomes more affordable, high-throughput, and of high quality, SNP arrays or the BeadChip will remain the most cost-effective ways to detect tumor samples with LOH and copy number aberrations in a genome scale.

REFERENCES 1. Adams J, Williams SV, Aveyard JS et al. Loss of heterozygosity analysis and DNA copy number measurement on 8p in bladder cancer reveals two mechanisms of allelic loss. Cancer Res 2005;65: 66–75. 2. Bergamaschi A, Kim YH, Wang P et al. Distinct patterns of DNA copy number alteration are associated with different clinicopathological features and gene-expression subtypes of breast cancer. Genes Chromosomes Cancer 2006;45:1033–1040. 3. Jeon YK, Sung SW, Chung JH et al. Clinicopathologic features and prognostic implications of epidermal growth factor receptor (EGFR) gene copy number and protein expression in non-small cell lung cancer. Lung Cancer 2006;54:387–398. 4. Candiotti KA, Birnbach DJ, Lubarsky DA et al. The impact of pharmacogenomics on postoperative nausea and vomiting: do CYP2D6 allele copy number and polymorphisms affect the success or failure of ondansetron prophylaxis? Anesthesiology 2005;102:543–549. 5. Ouahchi K, Lindeman N, Lee C. Copy number variants and pharmacogenomics. Pharmacogenomics 2006;7:25–29. 6. Cappuzzo F, Varella-Garcia M, Shigematsu H et al. Increased HER2 gene copy number is associated with response to gefitinib therapy in epidermal growth factor receptor-positive non-small-cell lung cancer patients. J Clin Oncol 2005;23:5007–5018. 7. Endo K, Sasaki H, Yano M et al. Evaluation of the epidermal growth factor receptor gene mutation and copy number in non-small cell lung cancer with gefitinib therapy. Oncol Rep 2006;16:533–541. 8. Yang SH, Seo MY, Jeong HJ et al. Gene copy number change events at chromosome 20 and their association with recurrence in gastric cancer patients. Clin Cancer Res 2005;11:612–620. 9. Dimova I, Yosifova A, Zaharieva B et al. Association of 20q13.2 copy number changes with the advanced stage of ovarian cancer tissue microarray analysis. Eur J Obstet Gynecol Reprod Biol 2005;118:81–85. 10. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;100:57–70. 11. Friend SH, Bernards R, Rogelj S et al. A human DNA segment with properties of the gene that predisposes to retinoblastoma and osteosarcoma. Nature 1986;323:643–646. 12. Kamb A, Gruis NA, Weaver-Feldhaus J et al. A cell cycle regulator potentially involved in genesis of many tumor types. Science 1994;264:436–440. 13. Li J, Yen C, Liaw D et al. PTEN: a putative protein tyrosine phosphatase gene mutated in human brain, breast, and prostate cancer. Science 1997;275:1943–1947. 14. Little CD, Nau MM, Carney DN et al. Amplification and expression of the c-Myc oncogene in human lung cancer cell lines. Nature 1983;306:194–196.

86

Part I / Genomic Experimental Approaches in Oncology

15. Lin CR, Chen WS, Kruiger W et al. Expression cloning of human EGF receptor complementary DNA: gene amplification and three related messenger RNA products in A431 cells. Science 1984;224: 843–848. 16. Gilbertson RJ, Hill DA, Hernan R et al. ERBB1 is amplified and over-expressed in high-grade diffusely infiltrative pediatric brain stem glioma. Clin Cancer Res 2003;9:3620–3624. 17. Ueda M, Hung YC, Terai Y et al. Glutathione S-transferase GSTM1, GSTT1 and p53 codon 72 polymorphisms in human tumor cells. Hum Cell 2003;16:241–251. 18. Wang VW, Bell DA, Berkowitz RS et al. Whole-genome amplification and high-throughput allelotyping identified five distinct deletion regions on chromosomes 5 and 6 in microdissected early-stage ovarian tumors. Cancer Res 2001;61:4169–4174. 19. Kallioniemi A, Kallioniemi OP, Sudar D et al. Comparative genomic hybridization for molecular cytogenetic analysis of solid tumors. Science 1992;258:818–821. 20. Lastowska M, Cotterill S, Pearson AD et al. Gain of chromosome arm 17q predicts unfavourable outcome in neuroblastoma patients. U.K. Children’s Cancer Study Group and the U.K. Cancer Cytogenetics Group. Eur J Cancer 1997;33:1627–1633. 21. Sherry ST, Ward MH, Kholodov M et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res 2001;29:308–311. 22. Matsuzaki H, Dong S, Loi H et al. Genotyping over 100,000 SNPs on a pair of oligonucleotide arrays. Nat Methods 2004;1:109–111. 23. Matsuzaki H, Loi H, Dong S et al. Parallel genotyping of over 10,000 SNPs using a one-primer assay on a high-density oligonucleotide array. Genome Res 2004;14:414–425. 24. Przeworski M, Hudson RR, Di Rienzo A. Adjusting the focus on human variation. Trends Genet 2000;16:296–302. 25. Reich DE, Schaffner SF, Daly MJ et al. Human genome sequence variation and the influence of gene history, mutation and recombination. Nat Genet 2002;32:135–142. 26. Di X, Matsuzaki H, Webster TA et al. Dynamic model based algorithms for screening and genotyping over 100 K SNPs on oligonucleotide microarrays. Bioinformatics 2005;21:1958–1963. 27. Liu WM, Di X, Yang G et al. Algorithms for large-scale genotyping microarrays. Bioinformatics 2003;19:2397–2403. 28. Kennedy GC, Matsuzaki H, Dong S et al. Large-scale genotyping of complex DNA. Nat Biotechnol 2003;21:1233–1237. 29. Komura D, Shen F, Ishikawa S et al. Genome-wide detection of human copy number variations using high-density DNA oligonucleotide arrays. Genome Res 2006;16:1575–1584. 30. Wong KK, Tsang YT, Shen J et al. Allelic imbalance analysis by high-density single-nucleotide polymorphic allele (SNP) array with whole genome amplified DNA. Nucleic Acids Res 2004;32:e69. 31. Beroukhim R, Lin M, Park Y et al. Inferring loss-of-heterozygosity from unpaired tumors using highdensity oligonucleotide SNP arrays. PLoS Comput Biol 2006;2:e41. 32. Li LH, Ho SF, Chen CH et al. Long contiguous stretches of homozygosity in the human genome. Hum Mutat 2006;27:1115–1121. 33. Huang J, Wei W, Zhang J et al. Whole genome DNA copy number changes identified by high density oligonucleotide arrays. Hum Genomics 2004;1:287–299. 34. Nannya Y, Sanada M, Nakazaki K et al. A robust algorithm for copy number detection using high-density oligonucleotide single-nucleotide polymorphism genotyping arrays. Cancer Res 2005;65:6071–6079. 35. Laframboise T, Harrington D, Weir BA. PLASQ: a generalized linear model–based procedure to determine allelic dosage in cancer cells from SNP array data. Biostatistics 2007;8:323–336. 36. Lai Y, Zhao H. A statistical method to detect chromosomal regions with DNA copy number alterations using SNP-array–based CGH data. Comput Biol Chem 2005;29:47–54. 37. Janne PA, Li C, Zhao X et al. High-resolution single-nucleotide polymorphism array and clustering analysis of loss of heterozygosity in human lung cancer cell lines. Oncogene 2004;23:2716–2726. 38. Bignell GR, Huang J, Greshock J et al. High-resolution analysis of DNA copy number using oligonucleotide microarrays. Genome Res 2004;14:287–295.

Chapter 6 / Single-Nucleotide Polymorphism Array

87

39. Hoque MO, Lee J, Begum S et al. High-throughput molecular analysis of urine sediment for the detection of bladder cancer by high-density single-nucleotide polymorphism array. Cancer Res 2003;63:5723–5726. 40. Lieberfarb ME, Lin M, Lechpammer M et al. Genome-wide loss of heterozygosity analysis from laser capture microdissected prostate cancer using single nucleotide polymorphic allele (SNP) arrays and a novel bioinformatics platform dChipSNP. Cancer Res 2003;63:4781–4785. 41. Lindblad-Toh K, Tanenbaum DM, Daly MJ et al. Loss-of-heterozygosity analysis of small-cell lung carcinomas using single-nucleotide polymorphism arrays. Nat Biotechnol 2000;18:1001–1005. 42. Mei R, Galipeau PC, Prass C et al. Genome-wide detection of allelic imbalance using human SNPs and high-density DNA arrays. Genome Res 2000;10:1126–1137. 43. Zhao X, Li C, Paez JG et al. An integrated view of copy number and allelic alterations in the cancer genome using single nucleotide polymorphism arrays. Cancer Res 2004;64:3060–3071. 44. Lin M, Wei LJ, Sellers WR et al. dChipSNP: significance curve and clustering of SNP-array–based loss-of-heterozygosity data. Bioinformatics 2004;20:1233–1240. 45. Lam CW, To KF, Tong SF. Genome-wide detection of allelic imbalance in renal cell carcinoma using high-density single-nucleotide polymorphism microarrays. Clin Biochem 2006;39:187–190. 46. Koed K, Wiuf C, Christensen LL et al. High-density single nucleotide polymorphism array defines novel stage and location-dependent allelic imbalances in human bladder tumors. Cancer Res 2005;65:34–45. 47. Zhou X, Mok SC, Chen Z et al. Concurrent analysis of loss of heterozygosity (LOH) and copy number abnormality (CNA) for oral premalignancy progression using the Affymetrix 10 K SNP mapping array. Hum Genet 2004;115:327–330. 48. Pfeifer D, Pantic M, Skatulla I et al. Genome-wide analysis of DNA copy number changes and LOH in CLL using high-density SNP arrays. Blood 2007;109:1202–1210. 49. Wong KK, Tsang YT, Chang YM et al. Genome-wide allelic imbalance analysis of pediatric gliomas by single nucleotide polymorphic allele array. Cancer Res 2006;66:11172–11178. 50. LaFramboise T, Weir BA, Zhao X et al. Allele-specific amplification in cancer revealed by SNP array analysis. PLoS Comput Biol 2005;1:e65. 51. Rook MS, Delach SM, Deyneko G et al. Whole genome amplification of DNA from laser capturemicrodissected tissue for high-throughput single nucleotide polymorphism and short tandem repeat genotyping. Am J Pathol 2004;164:23–33. 52. Dean FB, Hosono S, Fang L et al. Comprehensive human genome amplification using multiple displacement amplification. Proc Natl Acad Sci USA 2002;99:5261–5266. 53. Paez JG, Lin M, Beroukhim R et al. Genome coverage and sequence fidelity of phi29 polymerase– based multiple strand displacement whole genome amplification. Nucleic Acids Res 2004;32:e71. 54. Wang ZC, Buraimoh A, Iglehart JD et al. Genome-wide analysis for loss of heterozygosity in primary and recurrent phyllodes tumor and fibroadenoma of breast using single nucleotide polymorphism arrays. Breast Cancer Res Treat 2006;97:301–309. 55. Thompson ER, Herbert SC, Forrest SM et al. Whole genome SNP arrays using DNA derived from formalin-fixed, paraffin-embedded ovarian tumor tissue. Hum Mutat 2005;26:384–389. 56. Walker BA, Leone PE, Jenner MW et al. Integration of global SNP–based mapping and expression arrays reveals key regions, mechanisms, and genes important in the pathogenesis of multiple myeloma. Blood 2006;108:1733–1743. 57. Rozen S, Skaletsky H. Primer3 on the WWW for general users and for biologist programmers. Methods Mol Biol 2000;132:365–386. 58. Garraway LA, Widlund HR, Rubin MA et al. Integrative genomic analyses identify MITF as a lineage survival oncogene amplified in malignant melanoma. Nature 2005;436:117–122. 59. Stordal B, Peters G, Davey R. Similar chromosomal changes in cisplatin- and oxaliplatin-resistant sublines of the H69 SCLC cell line are not associated with platinum resistance. Genes Chromosomes Cancer 2006;45:1094–1105. 60. Wang Y, Makedon F, Pearlman J. Tumor classification based on DNA copy number aberrations determined using SNP arrays. Oncol Rep 2006;15 Spec no.:1057–1059.

88

Part I / Genomic Experimental Approaches in Oncology

61. Yuan E, Haghighi F, White S et al. A single nucleotide polymorphism chip–based method for combined genetic and epigenetic profiling: validation in decitabine therapy and tumor/normal comparisons. Cancer Res 2006;66:3443–3351. 62. Zorn KK, Jazaeri AA, Awtrey CS et al. Choice of normal ovarian control influences determination of differentially expressed genes in ovarian cancer expression profiling studies. Clin Cancer Res 2003;9:4811–4818. 63. Conrad DF, Andrews TD, Carter NP et al. A high-resolution survey of deletion polymorphism in the human genome. Nat Genet 2006;38:75–81. 64. McCarroll SA, Hadnott TN, Perry GH et al. Common deletion polymorphisms in the human genome. Nat Genet 2006;38:86–92. 65. Feuk L, Carson AR, Scherer SW. Structural variation in the human genome. Nat Rev Genet 2006; 7:85–97. 66. Freeman JL, Perry GH, Feuk L et al. Copy number variation: new insights in genome diversity. Genome Res 2006;16:949–961. 67. Goidts V, Cooper DN, Armengol L et al. Complex patterns of copy number variation at sites of segmental duplications: an important category of structural variation in the human genome. Hum Genet 2006;120:270–284. 68. Sebat J, Lakshmi B, Troge J et al. Large-scale copy number polymorphism in the human genome. Science 2004;305:525–528. 69. Sharp AJ, Cheng Z, Eichler EE. Structural variation of the human genome. Annu Rev Genomics Hum Genet 2006;7:407–442. 70. Oliphant A, Barker DL, Stuelpnagel JR, Chee MS. BeadArray technology: enabling an accurate, costeffective approach to high-throughput genotyping. Biotechniques 2002;Suppl 56–58:60–61. 71. Lips EH, Dierssen JW, van Eijk R et al. Reliable high-throughput genotyping and loss-ofheterozygosity detection in formalin-fixed, paraffin-embedded tumors using single-nucleotide polymorphism arrays. Cancer Res 2005;65:10188–10191. 72. Shen R, Fan JB, Campbell D et al. High-throughput SNP genotyping on universal bead arrays. Mutat Res 2005;573:70–82. 73. Gunderson KL, Kuhn KM, Steemers FJ et al. Whole-genome genotyping of haplotype tag single nucleotide polymorphisms. Pharmacogenomics 2006;7:641–648. 74. Peiffer DA, Le JM, Steemers FJ et al. High-resolution genomic profiling of chromosomal aberrations using Infinium whole-genome genotyping. Genome Res 2006;16:1136–1148.

II

P HARMACOGENOMICS OF T OXICITY AND R ESPONSE O F C HEMOTHERAPY

“This page left intentionally blank.”

7

Concordance Between Tumor and Germline DNA Sharon Marsh, PhD CONTENTS Introduction Alterations of The Cancer Genome Genotype Concordance Conclus ions Acknowledgements References

S UMMARY Germline DNA (e.g., blood, mouthwash) is the most readily accessible source of material for the identification of pharmacogenetic markers for therapy selection. However, the cancer genome is altered by many processes that could affect the expression of functional alleles in the tumor. Consequently the utility of the germline genome to predict the tumor genome is under question. Studies have suggested strong concordance between the germline and tumor genotype profiles for pharmacogenetic markers. However, genotype is only one factor involved in tumor response to chemotherapy and mechanisms such as chromosome amplification and loss, copy number variation, microsatellite instability, chromosome instability and epigenetic variation (methylation) need to be taken into account. Key Words: Concordance; chromosome amplification; genotype; genome; germline; tumor

From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 7, 

91

92

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

1. INTRODUCTION Assault on the cancer genome can come from a range of processes (1) (Fig. 1), consequently there are many intra-individual differences between the cancer and germline genomes (1). In addition to differences at the DNA level, alterations in RNA expression between tumor and normal genomes have been well characterized (2,3). Pharmacogenetic markers of variability in the type and extent of toxicity experienced by patients receiving chemotherapy can be assessed using germline DNA; for example, assessment of UGT1A1*28 to predict irinotecan toxicity can be performed using whole blood (4). However, markers of response to chemotherapy may depend heavily on the tumor genome (5). Alterations in the tumor genome may have an effect on the presence of functional alleles (e.g., through gene amplification/loss), or their expression (e.g., through epigenetic regulation). Consequently, the utility of genotyping the germline genome for single-nucleotide polymorphisms (SNPs), indels, or tandem repeats to predict the tumor genotype is still unclear.

1.1. Source of Samples DNA from any source can be utilized for pharmacogenetics research, including germline (e.g., blood, mouthwash, frozen normal tissue, formalin-fixed normal tissue, formalin-fixed paraffin-embedded normal tissue) and tumor (e.g., frozen tumor tissue, formalin-fixed tumor tissue, formalin-fixed paraffin-embedded tumor tissue) sources. In addition, circulating DNA can be extracted from plasma and serum. Circulating DNA is mainly tumor DNA, although some germline DNA will be present. Studies have suggested a strong correlation between epigenetic markers in circulating DNA compared with matched tumor tissue, suggesting that this is a useful source of tumor DNA (6,7).

Fig. 1. The tumor genome can be altered in many ways. CIN = chromosome instability; MSI = microsatellite instability.

Chapter 7 / Tumor/Germline Concordance

93

Immortalized cell lines can provide an almost infinite DNA resource; however, although they are a powerful tool for genetics and pharmacogenetics these are not a perfect resource. Genotyping of EBV-transformed immortalized lymphoblastoid cells has demonstrated similar allele frequencies to equivalent populations (8), but the EBV transformation can lead to problems with using the cell lines for pharmacogenetics studies. Some commonly used drugs cause EBV lytic replication, and consequently the cause of cell death cannot be distinguished between this effect and the toxicity of the drug (9). Using cancer cell lines removes the problems of EBV transformation, but many cancer cell lines have chromosome instability (CIN), and multiple passages of CIN cell lines lead to different degrees of chromosome losses and gains (10), which might affect subsequent genotype information or pharmacogenetic outcomes. DNA from formalin-fixed tissue, paraffin-embedded tissue, and plasma or serum is typically highly fragmented and low yield, limiting the types and number of assays that can be performed. Germline DNA represents by far the most easily accessible source material ( 11, 12). In addition, good-quality, high molecular weight DNA can be obtained from the majority of germline sources, allowing a greater range of assays to be performed. The problem remains whether germline DNA is an accurate enough representation of the cancer genome to allow genotype information to be applicable in cancer pharmacogenetics.

1.1.1. M ICRODISSECTION One of the major problems with using tissue from formalin-fixed or paraffinembedded samples is the presence of normal tissue in the sample. Contamination from normal cells can be up to 95% ( 13). This can lead to difficulties with analyzing subsequent assays. For example, detecting loss of heterozygosity can be highly dependant on the amount of normal cell contamination in the sample ( 14), and gene copy number differences between tumor and germline may be impossible to accurately determine. The use of techniques such as laser-assisted microdissection can significantly reduce the problem of normal tissue contamination (13,15,16). The precision of these techniques can lead to the isolation of single cells from paraffin-embedded slides. This allows for a great deal of accuracy. For example, one study identified a mutation in the E-cadherin gene not present in the germline genome by isolating single tumor cells from the surrounding normal tissue (16). The use of microdissection is an essential step for utilizing fixed tumor tissues for pharmacogenomics studies.

2. ALTERATIONS OF THE CANCER GENOME 2.1. Polymorphism versus Mutation In addition to the various factors that can affect the presence of, or expression of, functional polymorphisms, the cancer genome can also acquire mutations. These may also be a factor in tumor response to chemotherapy and cannot be identified by screening the germline genome. Recently, mutations in EGFR have been identified in non-small cell

94

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

lung cancer, which correlate with response to EGFR inhibitors (17,18). These mutations are not present in the germline and consequently might be missed if screening blood or normal tissue.

2.2. Chromosome Amplification/Loss Cytogenetic abnormalities are frequent in most tumor types and are often associated with tumor characteristics or patient outcome. Altered DNA ploidy (copy number changes) in the cancer genome was described in the 1960s when it was demonstrated that bladder and prostate cancer patients with diploid or tetraploid tumor nuclei had a longer survival rate than patients with triploid or hexaploid tumor nuclei (19). In an analysis of 2210 solid tumors (27 different tumor types) every chromosome region demonstrated some level of both loss and amplification ( 20). However, loss and amplification of the same chromosome region were rarely seen in the same tumor type, and some chromosome regions were tumor specific; for example, the short arm of chromosome 12 (12p) is amplified in 96.3% of testicular cancers and 0% of renal cancers (20). The clinical implications of both amplification and loss in tumors have been studied. In a study of 29 Dukes’ C colorectal cancer patients, those with tumors that had two or more chromosomal regions of gain or loss had significantly better prognosis than patients with less (p = 0.02) (21). Loss of chromosome 5q and lack of 8q amplification in serous ovarian cancer (n = 96) is associated with improved prognosis (5-year survival of 75% versus 0% with no loss on 5q and amplification on 8q; p = 0.0007) (22). In childhood ALL the amplification of specific chromosomes, chromosome regions, and genes has been associated with chemotherapy resistance and clinical outcome (23,24).

2.3. Gene Amplification Gene amplification as a resistance mechanism has been widely studied in cell lines. In 1979, methotrexate-resistant cell lines were found to have amplification of DHFR (25). Topoisomerase I (TOP1) is amplified in cancer cell lines ( 26), and this amplification is associated with altered protein expression and resistance to camptothecin chemotherapy (27). Amplification of thymidylate synthase, a target for 5-fluorouracil, was found in drug-resistance cancer cell lines ( 28), and was subsequently observed in colorectal cancer patients as a resistance mechanism to 5-fluorouracil (5-FU) (29). More recently, multiple gene copies of the epidermal growth factor receptor (EGFR) were significantly associated with response to EGFR inhibitors in non-small cell lung cancer patients (30), and amplification of Her2/neu (ERBB2) in breast cancer is a prognostic maker and a predictor for response to the anti-Her2 antibody trustuzamab (herceptin) (31). Gene amplification is not unique to cancer cells. Recently, a map of copy number variations (CNVs) in the human genome was published (32). Approximately 12% of the human genome has copy number differences—covering more nucleotides per genome than SNPs—making this a huge source of interindividual variability (32,33). The impact of germline CNVs on pharmacogenetics remains to be elucidated.

Chapter 7 / Tumor/Germline Concordance

95

2.4. Microsatellite Instability Microsatellite instability (MSI) in the tumor caused by defective DNA mismatch repair has been associated with outcome to chemotherapy (34). For example, a study of 320 Dukes’ B2 and C colon cancers treated with 5-FU therapy showed that MSI status was correlated with improved survival (p = 0.01) (35). MSI is also associated with a mutator phenotype that may lead to mutations in pharmacogenetic markers, which would not be picked up by screening the germline genome.

2.5. Chromosome Instability (CIN) Aberrations in the mitotic spindle checkpoint in tumor cells can lead to incorrect separation of chromosomes during mitosis. This leads to the chromosome instability (CIN) phenotype. The incidence of CIN varies depending on tumor type. In colorectal cancer a high incidence of CIN occurs (10). In fact, the majority of microsatellite stable colorectal cancers are CIN and also have a significant level of loss of heterozygosity (36). Swanton et al. speculate that this is the basis for taxane (chemotherapy agents that target microtubules) inactivity in colorectal cancer ( 37). If further evidence confirms this, identifying patients with CIN tumors may be useful in determining whether taxane therapy is appropriate.

2.6. DNA Methylation Methylation at CpG islands silences transcription. This is one of the common mechanisms for inactivating tumor suppressor genes in tumors. There is evidence to suggest the role of methylation in the outcome to chemotherapy [reviewed in (38)]. In 70 ovarian cancer patients, methylation status of DNA repair/detoxification genes was significantly associated with taxane therapy outcome (p = 0.013) (39). Methylation of GGH in acute lymphoblastic leukemia patients was associated with significantly reduced GGH expression (and consequently increased methotrexate polyglutamation) in 34 patients with wild-type germline GGH (40). It is clear that patients without functional polymorphisms in pharmacogenetically relevant genes can still have altered gene expression leading to chemoresistance/sensitivity. Within germline cells, methylation may also be a factor influencing gene expression. Recent evidence suggests that there are age-related differences in gene methylation patterns in the germline genome (41).

3. GENOTYPE CONCORDANCE Despite the differences between germline and tumor genomes, differences at the single-nucleotide level have not been extensively studied. Several small studies have been performed in breast cancer samples (Table 1). In a study assessing the utility of paraffin-embedded material for polymorphism analysis, no genotype discrepancy between germline and tumor samples for ABCB1, CYP2C8, and CYP2D6 in breast tumors were identified (n = 10) ( 42). In 17 breast cancer patients assessed for CYP2D6*4, CYP2D6*6, and CYP3A5*3 to identify markers of outcome from tamoxifen treatment there was complete concordance between tumor and germline genotype (43).

96

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy Table 1 Concordance Genotype Concordance Frequencies

Polymorphism

n

Tumour Type

Concordance

Reference

ABCB1 3435 ABCC2 -24 ABCG2 421 CCND1 870 CDKN2A 494 CES1 -40 CES1 IVS4+10 CES1 1970 CES1 2480 CES2 -363 CES2 IVS10-88 CES2 IVS11-160 CES2 3071 E2F1 1177 ERCC2 -9164 ERCC2 -1989 ERCC2 -516 ERCC2 468 ERCC2 2251 ERCC2 2133 FDXR 768 FDXR IVS9+15 FDXR 368 MPO -463 TYMS TSER XRCC1 580 XRCC1 839 XRCC1 1196 ABCB1 3435 CYP2C8*3 CYP2D6*4 CYP2D6*6 CYP2D6*4 CYP2D6*6 CYP3A5*3 NOS3 -786 NOS3 Glu298Asp VEGF 936 ERCC1 exon 4 ERCC2 exon 6 ERCC2 exon 22

44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 44 10 10 10 10 17 17 17 21 21 21 16 16 21

Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Breast tumor Breast tumor Breast tumor Breast tumor Breast tumor Breast tumor Breast tumor Breast tumor Breast tumor Breast tumor Colon tumor Colon tumor Colon tumor

95% 92% 94% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 97% 100% 100% 100% 100% 100% 100% 97% 91% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100%

(46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (46) (42) (42) (42) (42) (43) (43) (43) (44) (44) (44) (45) (45) (45) (Continued)

Chapter 7 / Tumor/Germline Concordance

97

Table 1 (Continued) Polymorphism

n

Tumour Type

Concordance

Reference

ERCC2 exon 23 XPF exon 11 XPG exon 15 XRCC1 exon 17 XRCC3 exon 7 NQO1 609

17 16 13 22 28 49

100% 100% 100% 100% 97% 55%a

(45) (45) (45) (45) (45) (49)

ERBB2 amplification

23

50%

(55)

ERBB2 amplification

60

Colon tumor Colon tumor Colon tumor Colon tumor Colon tumor Lung tumor (cDNA) Breast tumor versus bone matastasis Breast tumor versus other metastatic sites

98%

(56)

a

Estimated from reference (49).

In breast cancer patients, genotypes for polymorphisms in the angiogenesis genes NOS3 (–786T > C and Glu298Asp) and VEGF (936C > T) correlated 100% among breast tumor tissue, normal lymph nodes, and histologically involved lymph nodes (n = 21) (44). A couple of studies have been performed in colorectal cancer samples to identify the correlation between tumor and normal tissue (Table 1). Despite the high incidence of MSI and CIN in colorectal tumor there is strong concordance between germline and tumor genotype. A study assessing polymorphisms in the DNA repair genes ERCC1, ERCC2, XPF, XPG, XRCC1, and XRCC3 in colorectal cancer tumor and matched normal samples (n = 13–22) found that only one heterozygous sample for the XRCC3 exon 7 polymorphism demonstrated different genotypes between tumor and normal tissue, indicating a possible loss of heterozygosity at this locus (45). A study of 44 colorectal tumor samples and adjacent normal tissue identified strong concordance between paired samples for 28 polymorphisms in 13 genes (ABCB1, ABCC2, ABCG2, CCND1, CDKN2A, CES1, CES2, E2F1, ERCC2, FDXR, MPO, TYMS, and XRCC1). Forty-one paired samples (93%) had one or zero genotype discrepancies in the 28 polymorphisms between tumor and germline DNA. The majority of samples (77%) had complete concordance between tumor and normal tissue. A maximum of two discrepant polymorphisms were observed in individual tumor DNA samples (46). These studies suggest that the variability between tumor and germline genotype at the SNP level is minimal and the germline genome can be used as a surrogate to assess polymorphism status in the tumor. However, these studies were all performed on small sample numbers and do not represent all tumor types.

98

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

3.1. Genotype Discordance Loss of the short arm of chromosome 18 is common in colorectal cancer (occurs in ∼15% of patients). The TYMS gene is located in this region and a repeat polymorphism in this gene (TSER) is commonly studied in terms of response to 5-FU therapy (47). Several colorectal cancer studies have identified differences in TSER genotype between tumor and normal DNA. The frequency of the discrepancies range from 4% to 77.3% of TSER*2/*3 heterozygotes [median 56%; reviewed in (48)] and no allele bias is apparent (i.e., one repeat is not favored over the other). Curiously none of the studies have identified genotype discrepancies at a frequency of ∼15%, which would be consistent with the frequency of 18p loss of heterozygosity (LOH) in colorectal cancer. There are multiple possible reasons for the wide variability in the results of these studies. Some studies only looked at heterozygous samples; the population of the patients may cause discrepancies; some studies used microsatellite markers to validate the presence of LOH; and all studies were performed on small sample sizes. To date the extent of genotype discrepancy for TYMS TSER in colorectal cancer remains unclear.

3.2. Allelic Imbalance A study of 50 non-small cell lung cancer patients the NQO1 609C > T polymorphism was assessed in cDNA from matched tumor and normal tissue. Significant differences were seen for genotypes between the tumor and normal (p = 0.0043). It was suggested that this discordance was caused by using cDNA rather than genomic DNA from the tumor samples, and that the genotypes identified in the tumor cDNA reflected an imbalance of allele expression (49). This raises an interesting point that studying polymorphisms in tumor cDNA may provide a more useful representation of expressed functional markers than genomic DNA. Allelic bias in expression in tumor cells is of particular importance if gene amplification has taken place. A recent study identified patients with ALL who had amplification of the TPMT gene. Patients with three gene copies of wild-type TPMT had increased TPMT expression compared to patients with two wild-type gene copies or multiple copies of variant alleles ( 50). Consequently, although genotype can be consistent between tumor and germline cells, the level of expression may be altered based on the gene copy number and the selection of the amplified allele.

3.3. Primary Tumor versus Metastasis Concordance between the primary tumor genome and DNA from metastatic cells is even less defined than the concordance between germline and tumor genomes. It is clear that genetic variation and gene expression in primary tumor is not always predictive of genotype or expression in metastases. Genetic alterations, including allelic imbalance, may be present only in metastatic disease, and would be missed by screening primary tumor alone (51). Thymidylate synthase protein expression in the primary tumor is not a useful predictor for either thymidylate synthase protein expression or response to 5-fluorouracil therapy in metastatic disease (52,53). Similarly, concordance rates for expression of HER family members ranges from 79% to 56% between primary breast cancer and matched distant

Chapter 7 / Tumor/Germline Concordance

99

metastases (54), which has implications for using primary tumor to predict the usefulness of anti-HER2 therapy in metastatic disease. Concordance of ERBB2 gene amplification status, the marker for selection of antiHER2 therapy, varies between primary and metastatic cells depending on the location of the metastases (Table 1). In bone metastasis, 50% of patients with primary ERBB2 amplification had lost their amplification status in the bone metastases (55). However, in other studies, ERBB2 amplification in primary breast carcinoma was retained in 98% of metastases (sites varied and included lung, liver, pleural, and bone) (56). Consequently, the site of metastasis, as well as the genetic marker, is an important factor to take into account if using primary tissue with the aim of predicting response to therapy in metastases.

4. CONCLUSIONS Several studies have shown minimal genotype discordance between the tumor and germline genomes. However, these studies have been small and in limited tumor types. Other factors including gene copy number and biased allele expression need to be taken into account when using the germline as a marker for tumor genotype status. A threedimensional pharmacogenetics approach involving assessment of several factors that could affect tumor response to chemotherapy needs to be performed in multiple tumor types before a conclusive argument for or against the use of germline DNA can be made.

ACKNOWLEDGEMENTS The author is supported by UO1 GM63340 and R21 CA113491.

REFERENCES 1. McLeod HL, Marsh S. Pharmacogenetics goes 3D. Nat Genet 2005;37:794–795. 2. Kidd EA, Yu J, Li X et al. Variance in the expression of 5-fluorouracil pathway genes in colorectal cancer. Clin Cancer Res 2005;11:2612–2619. 3. Yu J, Shannon WD, Watson MA et al. Gene expression profiling of the irinotecan pathway in colorectal cancer. Clin Cancer Res 2005;11:2053–2062. 4. Ratain MJ. From bedside to bench to bedside to clinical practice: an odyssey with irinotecan. Clin Cancer Res 2006;12:1658–1660. 5. Hoskins JM, Mcleod HL. Cancer pharmacogenetics: the move from pharmacokinetics to pharmacodynamics. Curr Pharmacogenomics 2006;4:39–46. 6. Widschwendter A, Muller HM, Fiegl H et al. DNA methylation in serum and tumors of cervical cancer patients. Clin Cancer Res 2004;10:565–571. 7. Taback B, Giuliano AE, Lai R et al. Epigenetic analysis of body fluids and tumor tissues: application of a comprehensive molecular assessment for early-stage breast cancer patients. Ann NY Acad Sci 2006;1075:211–221. 8. Meucci MA, Marsh S, Watters JW et al. CEPH individuals are representative of the European American population: implications for pharmacogenetics. Pharmacogenomics 2005;6:59–63. 9. Feng WH, Hong G, Delecluse HJ et al. Lytic induction therapy for Epstein–Barr virus-positive B-cell lymphomas. J Virol 2004;78:1893–1902. 10. Lengauer C, Kinzler KW, Vogelstein B. Genetic instability in colorectal cancers. Nature 1997; 386:623–627. 11. Lenz HJ. The use and development of germline polymorphisms in clinical oncology. J Clin Oncol 2004;22:2519–2521.

100

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

12. Savage SA, Chanock SJ. Using germline genetic variation to investigate and treat cancer. Drug Discov Today 2004;9:610–618. 13. Becker I, Becker KF, Rohrl MH et al. Laser-assisted preparation of single cells from stained histological slides for gene analysis. Histochem Cell Biol 1997;108:447–451. 14. Tomlinson IP, Lambros MB, Roylance RR. Loss of heterozygosity analysis: practically and conceptually flawed? Genes Chromosomes Cancer 2002;34:349–353. 15. Pinzani P, Orlando C, Pazzagli M. Laser-assisted microdissection for real-time PCR sample preparation. Mol Aspects Med 2006;27:140–159. 16. Becker I, Becker KF, Rohrl MH et al. Single-cell mutation analysis of tumors from stained histologic slides. Lab Invest 1996;75:801–807. 17. Paez JG, Janne PA, Lee JC et al. EGFR Mutations in lung cancer: correlation with clinical response to gefitinib therapy. Science 2004;304:1497–1500. 18. Pao W, Miller V, Zakowski M et al. EGF receptor gene mutations are common in lung cancers from “never smokers” and are associated with sensitivity of tumors to gefitinib and erlotinib. Proc Natl Acad Sci USA 2004;101:13306–13311. 19. Tavares AS, Costa J, de Carvalho A et al. Tumour ploidy and prognosis in carcinomas of the bladder and prostate. Br J Cancer 1966;20:438–441. 20. Rooney PH, Murray GI, Stevenson DA et al. Comparative genomic hybridization and chromosomal instability in solid tumours. Br J Cancer 1999;80:862–873. 21. Rooney PH, Boonsong A, McKay JA et al. Colorectal cancer genomics: evidence for multiple genotypes which influence survival. Br J Cancer 2001;85:1492–1498. 22. Staebler A, Karberg B, Behm J et al. Chromosomal losses of regions on 5q and lack of high-level amplifications at 8q24 are associated with favorable prognosis for ovarian serous carcinoma. Genes Chromosomes Cancer 2006;45:905–917. 23. Ito C, Kumagai M, Manabe A et al. Hyperdiploid acute lymphoblastic leukemia with 51 to 65 chromosomes: a distinct biological entity with a marked propensity to undergo apoptosis. Blood 1999;93: 315–320. 24. Raimondi SC, Zhou Y, Mathew S et al. Reassessment of the prognostic significance of hypodiploidy in pediatric patients with acute lymphoblastic leukemia. Cancer 2003;98:2715–2722. 25. Kaufman RJ, Brown PC, Schimke RT. Amplified dihydrofolate reductase genes in unstably methotrexate-resistant cells are associated with double minute chromosomes. Proc Natl Acad Sci USA 1979;76:5669–5673. 26. Boonsong A, Marsh S, Rooney PH et al. Characterization of the topoisomerase I locus in human colorectal cancer. Cancer Genet Cytogenet 2000;121:56–60. 27. McLeod HL, Keith WN. Variation in topoisomerase I gene copy number as a mechanism for intrinsic drug sensitivity. Br J Cancer 1996;74:508–512. 28. Wang W, Marsh S, Cassidy J et al. Pharmacogenomic dissection of resistance to thymidylate synthase inhibitors. Cancer Res 2001;61:5505–5510. 29. Wang TL, Diaz LA, Jr., Romans K et al. Digital karyotyping identifies thymidylate synthase amplification as a mechanism of resistance to 5-fluorouracil in metastatic colorectal cancer patients. Proc Natl Acad Sci USA 2004;101:3089–3094. 30. Cappuzzo F, Hirsch FR, Rossi E et al. Epidermal growth factor receptor gene and protein and gefitinib sensitivity in non-small-cell lung cancer. J Natl Cancer Inst 2005;97:643–655. 31. Leyland-Jones B. Trastuzumab: hopes and realities. Lancet Oncol 2002;3:137–144. 32. Redon R, Ishikawa S, Fitch KR et al. Global variation in copy number in the human genome. Nature 2006;444:444–454. 33. Shianna KV, Willard HF. Human genomics: in search of normality. Nature 2006;444:428–429. 34. Jo WS, Carethers JM. Chemotherapeutic implications in microsatellite unstable colorectal cancer. Cancer Biomark 2006;2:51–60. 35. Sinicrope FA, Rego RL, Halling KC et al. Thymidylate synthase expression in colon carcinomas with microsatellite instability. Clin Cancer Res 2006;12:2738–2744.

Chapter 7 / Tumor/Germline Concordance

101

36. Goel A, Nagasaka T, Arnold CN et al. The CpG island methylator phenotype and chromosomal instability are inversely correlated in sporadic colorectal cancer. Gastroenterology 2007;132:127–138. 37. Swanton C, Tomlinson I, Downward J. Chromosomal instability, colorectal cancer and taxane resistance. Cell Cycle 2006;5:818–823. 38. Stebbing J, Bower M, Syed N et al. Epigenetics: an emerging technology in the diagnosis and treatment of cancer. Pharmacogenomics 2006;7:747–757. 39. Teodoridis JM, Hall J, Marsh S et al. CpG island methylation of DNA damage response genes in advanced ovarian cancer. Cancer Res 2005;65:8961–8967. 40. Cheng Q, Cheng C, Crews KR et al. Epigenetic regulation of human gamma-glutamyl hydrolase activity in acute lymphoblastic leukemia cells. Am J Hum Genet 2006;79:264–274. 41. Flanagan JM, Popendikyte V, Pozdniakovaite N et al. Intra- and interindividual epigenetic variation in human germ cells. Am J Hum Genet 2006;79:67–84. 42. Rae JM, Cordero KE, Scheys JO et al. Genotyping for polymorphic drug metabolizing enzymes from paraffin-embedded and immunohistochemically stained tumor samples. Pharmacogenetics 2003;13:501–507. 43. Goetz MP, Rae JM, Suman VJ et al. Pharmacogenetics of tamoxifen biotransformation is associated with clinical outcomes of efficacy and hot flashes. J Clin Oncol 2005;23:9312–9318. 44. Schneider BP, Skaar TC, Sledge GW et al. Analysis of angiogenesis genes from paraffin-embedded breast tumor and lymph nodes. Breast Cancer Res Treat 2006;96:209–215. 45. Mort R, Mo L, McEwan C et al. Lack of involvement of nucleotide excision repair gene polymorphisms in colorectal cancer. Br J Cancer 2003;89:333–337. 46. Marsh S, Mallon MA, Goodfellow P et al. Concordance of pharmacogenetic markers in germline and colorectal tumor DNA. Pharmacogenomics 2005;6:873–877. 47. Marsh S. Thymidylate synthase pharmacogenetics. Invest New Drugs 2005;23:533–537. 48. Kawakami K. Thymidylate synthase gene in pharmacogenetics. current Pharmacogenomics 2004;2:137–147. 49. Kolesar JM, Pritchard SC, Kerr KM et al. Evaluation of NQO1 gene expression and variant allele in human NSCLC tumors and matched normal lung tissue. Int J Oncol 2002;21:1119–1124. 50. Cheng Q, Yang W, Raimondi SC et al. Karyotypic abnormalities create discordance of germline genotype and cancer cell phenotypes. Nat Genet 2005;37:878–882. 51. Takahashi K, Kohno T, Matsumoto S et al. Clonal and parallel evolution of primary lung cancers and their metastases revealed by molecular dissection of cancer cells. Clin Cancer Res 2007;13:111–120. 52. Findlay MP, Cunningham D, Morgan G et al. Lack of correlation between thymidylate synthase levels in primary colorectal tumours and subsequent response to chemotherapy. Br J Cancer 1997; 75:903–909. 53. Marsh S, McKay JA, Curran S et al. Primary colorectal tumour is not an accurate predictor of thymidylate synthase in lymph node metastasis. Oncol Rep 2002;9:231–234. 54. Fuchs IB, Siemer I, Buhler H et al. Epidermal growth factor receptor changes during breast cancer metastasis. Anticancer Res 2006;26:4397–4401. 55. Lorincz T, Toth J, Badalian G et al. HER-2/neu genotype of breast cancer may change in bone metastasis. Pathol Oncol Res 2006;12:149–152. 56. Gong Y, Booser DJ, Sneige N. Comparison of HER-2 status determined by fluorescence in situ hybridization in primary and metastatic breast carcinoma. Cancer 2005;103:1763–1769.

“This page left intentionally blank.”

8

Epidermal Growth Factor Receptor Mutations and Sensitivity to Selective Kinase Inhibitors in Human Lung Cancer Anurag Singh, PhD, Sreenath V. Sharma, PhD, and Jeffrey Settleman, PhD CONTENTS Introduction The Biochemis try and Signaling Properties of EGFR The Dis covery of Oncogenic EGFR Mutations The Biochemical and Signaling Properties of EGFR Mutants Primary and Acquired Res is tance to EGFR TKIs Alternative Therapeutic Strategies to Target EGFR Function Conclus ions References

S UMMARY The epidermal growth factor receptor (EGFR) is a receptor tyrosine kinase (RTK) with pleiotropic developmental functions in metazoans. It is the prototypical member of the ErbB family of RTKs, which can form homodimers or heterodimers with other From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 8, 

103

104

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

ErbB family members upon ligation with a variety of EGF-related extracellular ligands to facilitate transphosphorylation and downstream signaling. EGFR, as well as other ErbB family members and their cognate ligands, is dysregulated during the development of a substantial fraction of solid tumors, most notably cancers of the breast and lung. The selective ATP-competitive tyrosine kinase inhibitors (TKIs), gefitinib and erlotinib, which target the EGFR kinase domain, have been introduced in the clinic as a treatment for chemotherapy-refractory non-small cell lung cancer (NSCLC). Initial trials of these compounds revealed a narrow efficacy profile in NSCLC patients, with approximately 10% of patients experiencing significant clinical response. It was later found that the vast majority of these responsive patients exhibited somatic activating mutations within the kinase domain of EGFR. This chapter outlines recent insights into the biochemical and signaling properties of mutant EGFR proteins, as well as the molecular basis for the sensitivity of tumors harboring these mutants to gefitinib and erlotinib. The challenges associated with primary and acquired resistance to EGFR TKIs in NSCLC are also discussed. Finally, alternative strategies to target the activities of EGFR and other ErbB proteins in lung cancer are described. Key Words: EGFR; Non-small cell lung cancer; Oncogenes; Receptor tyrosine kinases; Gefitinib; Erlotinib; Tyrosine kinase inhibitors; Drug resistance

1. INTRODUCTION The EGF receptor (EGFR), also known as ErbB1 or HER1, along with three other members of the ErbB family (ErbB2/neu, ErbB3, and ErbB4) are prototypical or representative models for receptor tyrosine kinase (RTK) function ( 1, 2). EGFR and its downstream signaling pathways are evolutionarily conserved in higher eukaryotes, playing central and non-redundant roles during metazoan development ( 2). Its broad and diverse developmental functions are reflected in such processes as photoreceptor cell fate specification in the Drosophila eye (3) and placental and epithelial development in mammals (4). Oncogene products often promote cell cycle progression as well as cellular survival via the induction of antiapoptotic mediators and suppression of proapopotic or growthinhibitory mediators. The activities of EGFR, in terms of cellular signaling outputs, bear some similarities to those of other well-characterized oncogenes, such as Ras. Aberrant EGFR activity has now been established as critical in the initiation and progression of a substantial number of human cancers (1). Early evidence for EGFR’s oncogenic potential came via studies of the avian erythroblastosis virus, whose transforming genes were found to be highly homologous to the chicken c-ErbB or human EGFR genes ( 5, 6). Later studies identified examples of gene amplification and/or overexpression of ErbB family members in many solid tumors, which correlated to poor clinical prognoses (7). These findings subsequently prompted pharmacological endeavors to target the aberrant activity of ErbB proteins in cancer, culminating in the development of specific EGFR tyrosine kinase inhibitors (TKIs), gefitinib and erlotinib. This chapter explores and analyzes recent studies of somatic activating mutations of the EGFR gene, from their identification to the biochemical and mechanistic basis for the exquisite sensitivity of tumors that harbor such mutations to EGFR TKIs. The

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

105

mechanisms underlying primary and acquired resistance to these drugs is also discussed, as well as alternative strategies to target the EGFR signaling pathway. The identification of EGFR mutations has now firmly established the receptor as a key oncogenic determinant in lung cancers and potentially other solid tumor types of endodermal origin.

2. THE BIOCHEMISTRY AND SIGNALING PROPERTIES OF EGFR ErbB receptors exhibit conserved structural similarity and domain topography (Fig. 1). They contain an extracellular ligand-binding region, which consists of two cysteine-rich domains, and is responsible for ligand-induced receptor dimerization. This

Fig. 1. The domain topography of EGFR. The receptor contains cysteine-rich domains (CRDs) in the extracellular region, which contribute to ligand binding. The transmembrane region contains hydrophobic amino acids, as is the case with other cell surface receptors. Within the intracellular region lies the kinase domain, which facilitates transphorylation of the receptor upon ligand-induced dimerization, at distinct tyrosine residues within the C-terminal tail. Phosphorylation of these residues leads to the recruitment of SH2 domain-containing adaptor molecules and enzymes that mediate downstream signaling events. This can lead to cellular growth and proliferation, as well as cell survival via antiapoptotic signaling. EGFR mutations found in NSCLC cluster within the kinase domain, and affect the phosphorylation status of the receptor.

106

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

is followed by a hydrophobic transmembrane region and an intracellular region that consists of the kinase domain as well as a C-terminal tail that, upon phosphorylation of key tyrosine residues, serves as a docking site for canonical interacting partners.

2.1. Ligand-Induced Activation of EGFR EGFR, along with the other ErbB family members, is activated by a diverse array of extracellular growth factor ligands, such as EGF, transforming growth factor ␣ (TGF␣) and amphiregulin (AR) (1,2) as well as betacellulin, epiregulin, heparin binding-EGF, and the neuregulins ( 8). Upon ligation, ErbB proteins form homo- or hetero-dimers resulting in activation of their tyrosine kinase functionality and subsequent transphosphorylation of the receptors on multiple tyrosine residues in the cytoplasmic tail (Fig. 1), facilitating the recruitment of docking partners (9). No known ligand for ErbB2 has thus far been identified. It has been proposed that ErbB2 must form heterodimers with other ErbBs in order to undergo ligand-dependent activation. ErbB2 is also the preferred dimerization partner for all ErbB proteins (10). In a similar vein, ErbB3 lacks kinase activity and thus requires heterodimerization with and transphosphorylation by another ErbB member in order to become activated (11). EGFR, in contrast, can become activated by cognate ligands in homodimer form. The biochemical basis for EGFR activation upon ligand binding and dimerization has recently been elucidated ( 12). Recent crystallographic studies have shown that EGFR can adopt one of two conformations, an auto-inhibited (inactive) state, or an activated state. Upon ligand binding, an asymmetric homodimer forms between these two states, such that the C-terminal lobe of the “active” EGFR monomer causes relief of the autoinhibition in the “inactive” EGFR monomer. This mechanism is reminiscent of cyclin-mediated activation of cyclin-dependent kinases (CDKs) (13) and results in potent activation of EGFR’s kinase activity followed by autophosphorylation of C-terminal tyrosine residues. This mode of activation suggests that EGFR does not require prior tyrosine phosphorylation, but does require dimerization, in order to become activated, and that it may activate other ErbB family members in asymmetric heterodimers, independently of its own intrinsic kinase activity. Upon ligation, EGFR is rapidly endocytosed into clathrin-coated vesicles, along with bound ligand ( 14). In early endosomes, c-Cbl, an E3 ubiquitin ligase is recruited to EGFR, resulting in ubiquitination of the receptor (15). Ubiquitinated EGFR, localized within early endosomes, maintains mitogenic signaling capacity (16). Activated EGFR in endosomes is then sorted towards the lysosomal degradation pathway, resulting in signal attenuation or, in rare circumstances, is recycled to the cell surface following vesicular fusion with the plasma membrane.

2.2. EGFR Interacting Proteins and Their Functions There are six tyrosine sites in the C-terminal tail of EGFR (Y992, Y1045, Y1068, Y1086, Y1148, and Y1173), which when phosphorylated, act as docking sites for SH2domain containing adaptor molecules and enzymes (Fig. 1). A seventh site, Y845, is situated within the activation loop of EGFR and is phosphorylated by the tyrosine kinase Src, although its phosphorylation is not required for EGFR’s kinase activity nor for its ability to transform cells, in vitro (17,18,19).

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

107

The major EGFR docking proteins have enzymatic activity (Fig. 1). These include phospholipase C␥ (PLC␥), the Janus kinases, the phosphatases SHP1 and PTP1 and the E3 ubiquitin ligase c-Cbl. Additional docking partners serve as adaptors for the recruitment of other proteins with catalytic function, such as Shc and Grb2. Shc and Cbl are inducibly tyrosine-phosphorylated by EGFR (20), resulting in subsequent recruitment of multimolecular complexes, such as Grb2-SOS and the p85 regulatory subunit of phosphatidylinositol 3-kinase (PI3 K) (Fig. 2) (21). The p85 PI3 K subunit in complex with the p110 catalytic subunit may also be recruited to the membrane by EGF via EGFR-ErbB3 heterodimers (22). This recruitment of catalytically active protein complexes to plasma membrane microdomains represents a common paradigm of EGFR signal transduction, in that enzymes are brought into close proximity with their substrate counterparts. This ultimately results in the activation of mitogenic and pro-survival signaling pathways mediated by proto-oncogene products such as the small GTPase Ras and PI3 K (21,23,24,25). The SOS guanine–nucleotide exchange factor mediates GTP loading of Ras (Fig. 2), leading to activation of evolutionarily conserved effector proteins, such as the Raf serine/threonine kinase, PI3 K, and Ral-GDS. Raf activation triggers a cascade of phosphorylation events that lead to mitogen-activated protein kinase (MAPK) activation.

Fig. 2. EGFR-mediated signaling pathways. Activated EGFR can heterodimerize with and activate ErbB3 to facilitate the recruitment of the PI3 K complex, the p85 regulatory subunit, and the p110 catalytic subunit. PI3 K then phosphorylates phosphoinositides in the plasma membrane to generate PIP3. PIP3 accumulation results in the membrane recruitment and activation of the AKT serine/threonine kinase, which can promote growth via mTOR activation or cell survival via inhibition of the FOXO class of transcription factors. EGFR can also recruit the guanine-nucleotide exchange factor SOS, which causes Ras activation, resulting in MAPK signaling, as well the activation of a number of other effector proteins. This leads to changes in gene expression via up-regulation of transcription factors such as Elk-1.

108

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

MAPK mediates cellular proliferation via activation of transcriptions factors such as Elk-1, ETS, and the AP-1 complex (26). Membrane-recruitment of PI3 K by EGFR or Ras leads to the formation of phosphatidylinositol-trisphosphates (PIP3) in the plasma membrane, which has pleiotropic cell biological outcomes. PIP3 accumulation leads to the recruitment and activation of proteins containing Plekstrin-homology (PH) domains, such as the Akt serine/threonine kinase, which promote cellular growth via activation of the mTOR pathway, morphogenetic and cellular motility changes via Rho GTPase activation, and anti-apoptotic signaling via modulation of NF-␬B and FOXO transcription factors ( 27). In addition, the Src-mediated phosphorylation site in EGFR, Y845, serves as a critical docking site for mediators of EGFR-dependent cell survival, namely, Stat5b and cytochrome-C oxidase II ( 28, 29). Many of these EGFR-triggered signal transduction pathways are aberrantly regulated during cellular transformation and cancer progression.

3. THE DISCOVERY OF ONCOGENIC EGFR MUTATIONS The role of growth factor signaling in the pathophysiology of cancer progression is underscored by the notion that transformed cells in culture exhibit a reduced requirement for serum-borne growth factors in media, which is due to autocrine or paracrine secretion of high levels of various growth factor molecules ( 30, 31). Many of these factors are ligands for EGFR, such as transforming growth factor-alpha (TGF-␣), implicating EGFR as a key regulator of cancer cell growth and proliferation. In addition, both EGF and TGF-␣ have been found to be expressed at high levels in a number of lung cancer cell lines and tumor samples (32,33). These paracrine signaling loops provided a rationale for the design of monoclonal antibodies that target the extracellular ligand-binding domain of EGFR, preventing receptor dimerization and activation (34).

3.1. Dysregulation of EGFR in Lung Cancer From early studies into EGFR function, and the identification of ErbB transforming genes in the avian erythroblastosis virus, there was a clear consensus that aberrant EGFR activity likely contributes to the etiology of a large subset of cancers (1). Focusing on lung cancer in particular, there had been early suggestions that abnormal EGFR gene regulation was a common phenomenon in non-small cell lung cancer (NSCLC), with approximately 50% of such cancers exhibiting EGFR over-expression at the protein level (35,36), with more recent studies placing the number at around 62% (37). In addition, frequent amplifications of the chromosomal 7p12 region, in which the EGFR locus is contained, had been documented (38). The clinical outlook for lung cancer is bleak. Worldwide, deaths due to the disease account for a third of all cancer-related deaths, making it the deadliest of all cancers. Lung cancer encompasses a number of diseases of diverse histological subtypes, the etiologies of which are somewhat varied. There are two broad classifications; small cell lung cancer (SCLC), representing only 20% of all cases and non-small cell lung cancer (NSCLC), which accounts for the remaining 80% (39). NSCLC originates from the lung epithelia and can be further divided based on histology into adenocarcinoma, bronchoalveolar, squamous, anaplastic, and large cell

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

109

carcinoma. Most patients present with advanced metastatic disease, with a median survival after diagnosis of 4–5 months, if no treatment is given ( 40). Combination chemotherapy is often administered with very limited efficacy, providing limited overall survival benefit to patients. The need for targeted therapeutics against specific oncogenic determinants, such as EGFR, has been clearly apparent.

3.2. Selective EGFR TKIs and the Discovery of Sensitizing EGFR Mutations There have been frenzied efforts to develop anti-EGFR agents that might provide benefit to cancer patients with progressive and advanced disease who were refractory to traditional chemotherapeutic agents. Prior to these efforts, the use of a specific Abl tyrosine kinase inhibitor, imatinib, had proved very effective in the treatment of BCR-Abl positive chronic myelogenous leukemia (CML) (41). This target-based drug discovery strategy, that is, blocking the aberrant activity of a specific oncogenic driving-force, provided a rationale for the design of anti-EGFR agents to treat patients with advanced NSCLC. This eventually led to the development and introduction of two compounds that specifically target the tyrosine kinase activity of EGFR, gefitinib or iressa (Astra Zeneca) and erlotinib or tarceva (OSI Pharmaceuticals/Genentech). These TKIs received fasttrack FDA approval in 2003 for gefitinib, and 2004 for erlotinib, as last resort treatments for advanced NSCLC in patients who were refractory to traditional chemotherapeutic agents (42). Gefitinib and erlotinib are quinazolinamine compounds that act as classical reversible competitive ATP analogs. They inhibit the kinase activity of EGFR by competing with ATP at the ATP-binding pocket within the kinase domain of the receptor, preventing tyrosine autophosphorylation and activation of EGFR, as well the phosphorylation of other EGFR substrates (43). These specific anti-EGFR agents, when used as monotherapy, showed compelling efficacy in early clinical trials, but only in a small subset of NSCLC patients, the response fraction being approximately 10% (44). This narrow margin of response to gefitinib and erlotinib led to even more compelling observations. The majority of responders seemed to fall into three major population classes: persons of East-Asian origin, non-smokers and women (45,46). Thus, the drugs had potentially revealed a unique pharmacogenomic profile in NSCLC. Previous to the introduction of EGFR TKIs for NSCLC treatment, an EGFR mutant designated EGFR variant III (EGFRvIII) had been identified in some glioblastomas, which encodes for a truncated receptor lacking the ligand-binding region. These EGFRvIII mutants display ligand-independent constitutive activity (47). Intuitively, several groups began the arduous task of sequencing the entire EGFR coding region in gefitinib and erlotinib-sensitive NSCLC tumors to identify polymorphisms or mutations R R and Tarceva . This approach led to the that may confer a hypersensitivity to Iressa exciting discovery of a different set of somatic EGFR mutations than those in EGFRvIII, within exons 18 to 21 of the gene. The mutations cluster within the kinase domain of the receptor, the predominant ones being an L858R substitution, accounting for 40% of all EGFR mutations and small 12–18 base pair in-frame deletions of exon 19, which codes for the LREA stretch of amino acids (e.g., ⌬E746-750), accounting for a further 45% (Table 1) (45,46,48).

Table 1 Somatic Mutations in the Kinase Domain of EGFR Identified in NSCLC Patients Mutation

Frequency In NSCLC

Effect on TKI Sensitivity

Exon

Mutation

18

G719C G719S G719A E709 K/Q V689 M N700D S720P G719C + E709H L688P P694L/S E709 V/A/G I715S S720F G719S+L861Q ⌬E746-A750 ⌬E746-T751 ⌬E746-T751 (A/I ins) ⌬E746-S752 (A/V ins) ⌬L747-A750 (P ins) ⌬L747-T751 ⌬L747-T751 (P/S ins) ⌬L747-S752 ⌬L747-S752 (Q ins) ⌬L747-P753 (S ins)

5%

Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing

20

T790M Ins 770 (ASV) ) Ins 761 (EAFQ) Ins 770 (CV) Ins 770 (Y/NPG) Ins 771(G) Ins 773(NP), H775Y Ins 774(H) Ins 774(PH) Ins 774(NPH) Ins 775(HV) DA767-V769 S768I S768I + V769L A763 V V765A T783A D770G H773L H775Y R776C G779F S784F L792P

110

Exon

19

25−31%

Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing Sensitizing

Frequency In NSCLC

Effect on TKI Sensitivity Resistance

3%

111

⌬S752-I759 ⌬E746-T751 (V ins) ⌬E746-A750 (V/RP ins) ⌬E746-S752 (D ins) ⌬E746-P753 (LS ins) ⌬E746-P754 (VASS ins) ⌬L747-E749 ⌬L747-E749 (P ins) ⌬L747-T751 (Q/D ins) ⌬L747-S752 (S/H ins) ⌬749-751 ⌬751-I759 (S/N ins) W731STOP InsA743(KIPVAI) L730F P733L G735S V742A E746 K T751I S752Y K754R DE746-T751 (ins VA) + R803 W

Sensitizing 21

L798F G810S L858R R776 + L858R L861Q G863D L858R + T790M L858R + D761Y L858R + S768I L858R + R776C N826S H835L K846R T847I H850 N V851I/A I853T L858 M A859T L861Q A864T E866 K A871G

39–46%

Sensitizing Sensitizing Sensitizing Sensitizing Resistance Resistance

G873E Compiled from Lynch et al., 2004; Paez et al., 2004; Pao et al., 2004; Shigematsu et al., 2005; Kosaka et al., 2004; Janne et al., 2005; Han et al.,2005; Marchetti et al., 2005; Mitsudomi et al., 2005; Cappuzzo et al., 2005; Tsao et al., 2005; Chou et al., 2005; Tokumo et al., 2006.

112

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Two other amino-acid substitutions, G719S and L861Q, as well as small in-frame insertions, within exon 20, account for the remaining 5% of mutations. Furthermore, it has become evident that the frequency of EGFR mutations within the NSCLC patient population correlates very closely with the 10% response rate to EGFR TKIs (49). This provides some credence to the notion that these mutations confer hypersensitivity to the EGFR inhibitors. EGFR mutations have subsequently been identified in some pancreatic ductal adenocarcinomas (PDA), which may explain the partial efficacy of erlotinib in the treatment of this highly malignant disease (50). Erlotinib in combination with the DNA synthesis inhibitor gemcitabine has now received FDA approval for the treatment of PDA, with a median survival benefit of approximately 2 weeks. Subsequent studies of NSCLC patient responses revealed that different classes of EGFR mutations were associated with distinct clinical responses to EGFR TKIs. Thus, patients with tumors harboring the EGFR exon 19 deletion mutants exhibited a markedly better response to TKIs than those with the L858R mutation ( 51, 52, 53). Interestingly, lung tumors harboring the rare exon 20 insertion mutants of EGFR do not display a marked sensitivity to either gefitinib or erlotinib. Recent clinical studies have suggested that glioblastoma patients harboring EGFRvIII mutations exhibit a statistically significant better response to gefitinib than those with wild-type EGFR ( 54). Thus, sensitivity to EGFR TKIs appears to be a complex pharmacological conundrum. An additional point worth noting is that EGFR mutations do not confer a phenotypic sensitivity to EGFR TKIs with 100% penetrance. A small subset of patients harboring these mutations is refractory to the inhibitors. By the same token, approximately 18% of patients that respond to gefitinib or erlotinib do not harbor any mutations in EGFR. Taken together, these observations suggest that other genetic lesions or molecular determinants may modulate sensitivity to TKIs. This remains a pressing issue for clinicians seeking to identify biomarkers of insensitivity or resistance in patients. In theory, this will allow for better decisions to be made about which patients might benefit most from these therapies.

4. THE BIOCHEMICAL AND SIGNALING PROPERTIES OF EGFR MUTANTS Oncogenic mutations in genes such as Ras cause hyperactivation of the gene product, resulting in constitutive and dysregulated activity of the protein. This had been shown to be the case with the EGFRvIII mutant expressed in glioblastomas. Thus, upon identification of mutations in the kinase domain of EGFR, it was hypothesized that they would confer hyperactivation on the receptor. Indeed, EGFR kinase domain mutants exhibit enhanced EGF-stimulated autophosphorylation and activation of a subset of downstream effectors that mediate cellular survival signaling, namely, Stat3/5 and AKT ( 45, 55). Hyperactivation of these signaling pathways by mutant EGFRs is likely to drive the growth and survival of NSCLC cells harboring the mutations, thereby explaining their oncogenic role.

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

113

4.1. Structural Insights into Hyperactivation of EGFR Mutants Structural studies of isolated EGFR kinase domains, as mentioned above, have provided some insight into how EGFR mutations may confer constitutive activation (12). Activation proceeds via the formation of an asymmetric dimer, in which one monomer adopts an open or active confirmation facilitating the induction of the kinase activity of the other partner in the dimer, which adopts a closed confirmation, under basal conditions. It has been suggested that leucines 858 and 861 in the EGFR protein, which are mutated in NSCLC, interact strongly with hydrophobic side-chains in the so-called N-lobe of the kinase domain in the structural model of the inactive EGFR confirmation. Thus, it is likely that the L858R and L861Q mutations that result in replacement with hydrophilic or polar side-chains perturb the hydrophobic interactions and promote a switch to the open or active confirmation, thus rendering the receptor hyperactive. In addition, erlotinib binding is predicted to be incompatible with the closed or inactive EGFR confirmation, which may explain why tumors from NSCLC patients harboring L858R missense mutations are more sensitive to gefitinib, in comparison (45,48).

4.2. The Molecular Basis for Sensitivity of EGFR Mutants to Selective TKIs A central and as yet poorly understood concept in the arena of gefitinib and erlotinib pharmacology is the cellular and biochemical basis for sensitivity to these inhibitors. Early functional analyses of the mutant receptors in cell-based kinase assays had suggested they exhibit an approximately ten-fold increased sensitivity to inhibition by gefitinib relative to wild-type EGFR (45,55). Recent in vitro biochemical analyses utilizing purified isolated mutant EGFR kinase domains have verified these results, suggesting that the mutant EGFRs exhibit a higher Km (binding constant) for ATP and an increased affinity for the TKIs, or a lower Ki value, than wild-type EGFR (56). Does this hypersensitivity of EGFR mutants to inhibition by gefitinib and erlotinib really correspond to an increased affinity for these compounds? The question has raised some controversy. Another study in which a novel phage-display methodology was employed to calculate Kd values for various kinase inhibitors with 113 different kinases found that the ratios of Kd for gefitinib and erlotinib between mutant and wild-type EGFR proteins were not significantly greater than 3 or less than 0.33 (57). The limitation of these analyses was that the assay was performed by competition with immobilized staurosporine rather than ATP. Thus, the biochemical basis for the hypersensitivity of EGFR mutants to inhibition by gefitinib and erlotinib remains to be fully elucidated. Recent studies that have attempted to identify biomarkers for sensitivity to EGFR TKIs, have revealed some interesting results. Firstly, it appears that the up-regulation of ErbB3 is seen in a significant proportion of gefitinib-sensitive cell lines, which has been suggested to be a mechanism allowing for the coupling of EGFR to the PI3 K-AKT signaling pathway, resulting in EGFR-dependent survival signaling (58,59). Secondly, it has been suggested that cells that have undergone epithelial to mesenchymal transition (EMT), are inherently resistant to gefitinib (60,61). EMT is a process that occurs during normal development and during tissue homeostasis, which essentially involves a transdifferentiation mechanism allowing epithelial cells to migrate from attached substrata to distal sites within tissues. EMT occurs as

114

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

a result of down-regulation of epithelial markers such as E-cadherin and up-regulation of mesenchymal proteins such as vimentin. The process is thought to contribute to the malignancy of tumor cells, allowing them to invade into the vasculature and subsequently metastasize. The molecular basis for resistance to gefitinib in cells that have undergone EMT remains to be fully elucidated, although the expression level of E-cadherin seems to be one determinant (61).

4.3. Sensitivity of EGFR Mutants to TKIs: The Oncogenic Shock Model EGFR mutants are able to promote in vitro cellular transformation when expressed ectopically ( 62). The growth of these mutant EGFR-transformed cells can be potently inhibited with gefitinib and erlotinib. By a similar token, mice harboring human EGFR mutant transgenes, whose expression is under the control of doxycyclinedependent transcription, rapidly develop lung tumors with features reminiscent of human brochioalveolar adenocarcinoma, when given oral doses of doxycycline (63,64). When doxycycline administration is discontinued, the lung tumors regress, a phenomenon that can be mimicked with multiple EGFR antagonists, including TKIs and monoclonal antibodies. Thus, NSCLC cells expressing mutant EGFRs may rely solely on the oncogenic activity of the receptors for growth and viability. At this point, it may be prudent to introduce the theory of “oncogene-addiction” (65). This model postulates that cancer cells, upon transformation by a single hyperactivated oncogene, become heavily reliant or dependent on its activity for growth and survival. Essentially, the cancer cell is in a state of equilibrium in which pro-growth and antiapoptotic signals emanating from the transforming oncogene—EGFR, for example— counterbalance pro-apoptotic signals that are invariably triggered when a cell becomes transformed in efforts to facilitate a self-destruct mechanism. When the oncogenic stimulus is removed—for instance by pharmacologic intervention—the counterbalance is disturbed and the equilibrium shifts. Pro-apoptotic signals eventually predominate, resulting in programmed cell death of the cancer cells. This has now become entrenched as a central dogma in the rationale for target-based drug discovery efforts and has implications for the design of kinase inhibitors to target the activities of oncogenic kinases, such as B-Raf, EGFR, and other receptor tyrosine kinases. The theory also helps to explain the efficacy of proven target-based anti-cancer drugs such as imatinib for BCR-Abl and gefitinib or erlotinib for EGFR. Another interesting point to note is that NSCLC cell lines harboring EGFR mutations display gene amplifications of the mutant alleles which suggests that there may be a great selective advantage for elevated EGFR activation in these cancer cells. This amplification of the locus seems somewhat counterintuitive in the context of hypersensitivity of mutant versus wild-type EGFR to EGFR TKIs, because one may predict that a higher concentration of drug would be required to inhibit the higher number of total receptors expressed. However, this would suggest that limited drug bioavailability is not a significant issue. A variation of the oncogene addiction model, referred to as “oncogenic shock” (66,67) may explain the paradox of hypersensitivity to EGFR TKIs despite amplification of the gene (Fig. 3). This model suggests that transforming oncogenes, such as EGFR mutants, concomitantly activate both pro-survival and pro-death signaling pathways. Upon removal of the oncogenic stimulus, there are differences in the temporal attenuation of

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

115

Fig. 3. The oncogenic shock model. Tumor cells exhibit an equilibrium between pro-survival and pro-apoptotic signals, such that pro-survival predominates. Upon disruption of the oncogenic drivingforce, the pro-survival signals dissipate at a more rapid rate than the pro-apoptotic signals, such that there is a period during which pro-apoptotis predominates. Thus, tumor cells undergo tumor cell death.

the counterbalancing growth and death signals. In simple terms, the pro-survival signals dissipate very rapidly upon disruption of the oncogenic stimulus, but pro-death signals decay at a slower rate, such that they eventually outweigh the survival signals, resulting in apoptotic cell death. Thus, in NSCLC cells harboring amplified EGFR mutants, gefitinib- or erlotinib-induced cell death may be due to a large pro-death output emanating from activated EGFR. This mechanism could certainly explain why cells over-expressing the mutants are exquisitely sensitive to EGFR TKIs. Untransformed or normal cells are not sensitive to these drugs because they do not exhibit this fine balance between pro-survival and pro-death exhibited by tumor cells that are heavily dependent on a single oncogenic stimulus. Thus, target-based anti-cancer agents such as imatinib, gefitinib, and erlotinib display relatively low toxicity indices. The oncogenic shock model also has implications for combined chemotherapeutic regimens with targeted therapies (such as erlotinib) and conventional DNA damaging agents (such as gemcitibine). For instance, chemotherapeutic agents that trigger a DNA damage-induced checkpoint, resulting in mild growth arrest or apoptosis, may attenuate the effects of target-based therapies that attack the oncogenic driving force directly, which trigger a delayed but sustained pro-death response. Indeed, in clinical practice, the combination of gefitinib or erlotinib with traditional chemotherapeutic agents has not yielded a statistically significant increase in survival benefit (68).

116

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Finally, the oncogenic shock model suggests that in future drug-discovery endeavors, time-courses of drug action as well as traditional dose response analyses may need to be analyzed. In the case of EGFR, it is clear that anti-EGFR monoclonal antibodies exhibit a different toxicity profile than TKIs, which may be related to the rate at which EGFR signaling is “shut-off” with the two modalities. Thus, drugs that are faster acting in terms of attenuation of pro-survival signals may prove to be more efficacious in treating certain cancers.

5. PRIMARY AND ACQUIRED RESISTANCE TO EGFR TKIS Although the use of gefitinib and erlotinib in the setting of EGFR-driven cancers has revealed some clinical efficacy, these agents are limited in their beneficial scope. If one examines NSCLC as a representative model for targeting EGFR activity, the overall response rate of patients to gefitinib or erlotinib in published studies is approximately 10% in unselected NSCLC patients in U.S. and European populations. For patients harboring EGFR-activating mutations, the rate of response is between 78% and 100% (69). Approximately 90% of NSCLC patients with advanced metastatic disease exhibit little or no response to the drugs, a phenomenon referred to as primary or de novo resistance, since it is inherent to the genotypic profile of the patient prior to treatment and does not develop due to selective pressures imposed by the drug. This latter situation is representative of acquired resistance, and will be discussed below.

5.1. Primary or De Novo Resistance to EGFR TKIs The genetic basis for primary resistance to EGFR TKIs has been partly explained by clinical genotyping studies of NSCLC patients who are refractory to treatment. In one particular study of 38 patients who were deemed to be refractory to either gefitinib or erlotinib, 9 patients were found to harbor mutations in the K-Ras locus (70). K-Ras is mutated in approximately 30% of all NSCLC cases, and K-Ras mutation-positive tumors are associated with very poor prognoses, with such patients being highly refractory to conventional chemotherapeutic agents (71). Thus, many believe that K-Ras mutational status can be used as a clinical predictor of response to EGFR TKIs, although the present data sets do not achieve statistical significance. Oncogenic K-Ras exhibits many of the same signaling properties as EGFR, including the ability to drive growth and survival via the ERK and Akt pathways, respectively. In tumors expressing mutant EGFR proteins, since K-Ras activation lies downstream of receptor activation, it may partially mediate EGFR-dependent neoplasia. It has been observed that mutations in EGFR and K-Ras are invariably mutually exclusive and never coexist. This is thought to be due to functional redundancy but instead may be due to a phenomenon known as a oncogene-induced senescence, in which the oncogenic stimulus, rather than driving cellular growth, induces exit from the cell cycle into the G0 phase, a process thought to be due to activation of the p16/Ink4A/ARF tumor suppressor locus (72). Thus, mutational activation of EGFR and K-Ras within the same tumor cell may result in senescence when p16 or ARF activities have not been lost. Alternatively, EGFR and K-Ras mutations within the same tumor cell may trigger a strong apoptotic response

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

117

that cannot be counteracted by the oncogenic stimulus. These scenarios are consistent with the oncogenic shock model, in that constitutively active variants of EGFR and Ras drive pro-death signaling pathways that may function in an additive manner, such that the coexistence of the two in a tumor may result in a selective disadvantage. The molecular and cellular basis for primary resistance to EGFR TKIs in patients with K-Ras mutations is unclear. It has been suggested that K-Ras driven activation of the PI3 K-AKT-mTOR signaling pathway may be critical in establishing an EGFRindependent cellular survival framework. Thus, even though gefitinib and erlotinib may inhibit the kinase activity of EGFR in cells harboring oncogenic K-Ras, the drugs cannot elicit a pro-death response because K-Ras is driving cell survival. An alternative and more complex hypothesis, however, may be the following. As outlined above, it has been suggested that cells that have undergone EMT display an inherent or de novo resistance to gefitinib and erlotinib (61). It has been known for some time that oncogenic K-Ras has the potent ability to drive EMT in cells (73). Thus, it is possible that by inducing the transdifferentiation of epithelial cells to a mesenchymal state, oncogenic K-Ras facilitates the primary resistance to EGFR TKIs seen in the clinic. The vast majority of NSCLC patients who display primary resistance do not harbor K-Ras mutant alleles. The tumors from these patients may either exhibit alternative EGFR or K-Ras-independent survival signaling, or they may have undergone EMT independently of K-Ras activity, which are possibilities that remain to be validated. A small subset of NSCLC patients that are refractory to treatment with EGFR TKIs, paradoxically harbor activating EGFR mutations (45). It has been shown that in mouse models of tumorigenesis, such as those driven by oncogenic Myc or K-Ras, most tumors regress upon removal of the oncogenic stimulus, which is consistent with the oncogeneaddiction model. However, a small subset of tumors fails to regress if the initial oncogene remains active for sustained periods. This is thought to be due to the acquisition of secondary genetic lesions, which result in the oncogenic activation of other genes or the loss of tumor suppressor function. This may explain why some EGFR-mutation positive NSCLC cells display primary resistance to EGFR TKIs. For instance, it is possible that loss of expression of the PTEN tumor suppressor, via gene dysregulation or epigenetic mechanisms such as promoter hypermethylation ( 74) could lead to hyperactivation of PI3 K signaling, because PTEN is a negative modulator of PI3 K outputs ( 75). PTEN loss in the context of EGFR mutations can lead to EGFR-independent PI3 K signaling, resulting in a lack of response to drugs such as gefitinib and erlotinib.

5.2. Acquired Resistance to EGFR TKIs A disturbing facet of cancer pharmacology is the emergence of acquired resistance to drugs in patients following long-term treatment. This was shown to be the case with BCR-Abl positive CML patients who had been treated with imatinib. Some patients went into relapse, which seemed to be due to the acquisition of a secondary mutation in BCR-Abl, T315I. This has subsequently been referred to as the “gatekeeper residue,” because it sterically hinders binding of the competitive inhibitor, remaining active in the presence of high concentrations of the drug (76). In the case of NSCLC, a significant proportion of patients treated with gefitinib or erlotinib, who had displayed promising initial responses, have now gone into re-

118

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

lapse. Understanding the molecular determinants that underlie resistance to EGFR TKIs and other targeted therapies is crucial for developing agents that are efficacious in the treatment of recurrent disease. The molecular basis for acquired resistance to TKIs has partially been explained by genotyping analyses. Similar to the situation with imatinibresistant CML, a secondary T790M mutation in the kinase domain of EGFR was identified that resembled the gatekeeper residue in BCR-Abl (77,78). The T790M mutation occurs in approximately 50% of patients who develop acquired resistance to EGFR-TKIs although the frequency of the T790M-containing alleles does not occur in a 1:1 ratio with that of the original activating mutation. This suggests that the T790M mutation may exist in a subset of cancer cells. Alternatively, the T790M mutant allele may be expressed in every drug-resistant cell but is subject to allelic dilution, which has been observed during in vitro modeling of gefitinib resistance (79). It is unclear why this would be the case from a teleological point of view, although the oncogenic shock model could again be evoked. An EGFR double mutant that contains an activating mutation and the T790M mutation is significantly more active than either single mutant and displays a degree of ligand-independent activation (unpublished data). Therefore, an overly hyperactive double EGFR mutant could potentially trigger prodeath signaling and cells over-expressing the double mutant may be at a selective disadvantage in a tumor. Recent genetic studies have also identified familial single T790M EGFR mutations that confer a predisposition to lung cancer, lending credence to the idea that these mutants are hyperactivated (80). The remaining 50% of cases of acquired resistance, which do not harbor T790M mutations, fall into an ill-defined class that requires further characterization. The molecular determinants of resistance in these cases have only recently begun to be identified. Modeling of gefitinib resistance in vitro has been possible via the use of cell lines that are highly sensitive to treatment with the inhibitors. These cells can be grown in high concentrations of gefitinib, and limited dilution cloning of drug-resistant cells has been employed to molecularly define acquired resistance. As mentioned above, rare occurrences of T790M mutations have been observed by this approach as well as dysregulated EGFR endosomal trafficking and degradation (79,81). A number of alternative EGFR TKI resistance mechanisms can be postulated. These include EGFR gene over-amplification, such that equilibrium plasma drug concentrations are insufficient for kinase inhibition, or down-regulation of phosphatase genes that are responsible for pro-survival signal attenuation either via epigenetic or posttranslational mechanisms. The existence of these mechanisms has not been definitively tested in vitro or in vivo. Lastly, resistance may arise via physiological variations that lead to altered intracellular drug concentrations, such as changes in lysosomal pH or the up-regulation of multi-drug resistance (MDR) proteins, such as ATP-binding cassette (ABC) transporters. Some controversy has been raised as to the importance of drug efflux pumps in the emergence of gefitinib resistance. For instance, it has been suggested that the ABCG2 protein displays a high affinity for gefitinib and actively extrudes the drug from the cell (82,83,84). However, conflicting reports suggest that gefitinib actually inhibits the activity of MDR proteins such as ABCG2 and P-glycoprotein ( 85, 86). Thus, it seems that resistance mechanisms other than T790M mutations have yet to be definitively validated.

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

119

On a final note, an emerging idea in the field of cancer biology is the existence of the cancer stem cell, which may give rise to the phenotypically heterogeneous cellular sub-populations that are often seen in solid tumors ( 87, 88). It has been observed that certain cancer cells exhibit the hallmarks of stem cell–like behavior, with the ability to self-renew and the exhibition of multipotency, allowing for differentiation into multiple cell types. In addition, characteristics of normal stem cells, which have been acquired by the cancer stem cell, are the loss of gap-junctional communication and the up-regulation of MDR proteins that can extrude xenobiotic compounds. Thus, an intriguing possibility to explain EGFR TKI resistance may be the preexistence of lung cancer stem cells that are inherently resistant to the drugs. Small clones of the cancer stem cells may then be afforded an inadvertent selective growth advantage in tumors treated with gefitinib and erlotinib. This scenario of acquired resistance has yet to be investigated, but if correct may suggest that the resistant tumor might be clonally divergent from the original tumor, which could be tested via gene expression profiling. Such profiling could lead to the identification of biomarkers of resistance. Unfortunately, analyses of these kinds are confounded by the paucity of pre- and post-treatment patient biopsy samples.

6. ALTERNATIVE THERAPEUTIC STRATEGIES TO TARGET EGFR FUNCTION The emergence of resistance to gefitinib and erlotinib in patients who were initially sensitive to the drug has dealt a large blow to target-based drug discovery efforts. In addition, the narrow pharmacogenomic profile of responders to EFGR-targeted therapy leaves a vast majority of patients with advanced NSCLC in the lurch. Tumors from this population that display primary resistance to gefitinib and erlotinib are likely to be genetically diverse and not dependent on EGFR function for growth and viability. As a corollary, these tumors may display addiction to other oncogenic proteins, possibly kinases, which might lead to the development of novel classes of kinase inhibitors that are specifically efficacious in subpopulations of NSCLC patients.

6.1. Irreversible Inhibitors of EGFR and Other ErbB Proteins Returning to the problem of acquired resistance to EGFR TKIs, a number of promising alternative treatments to antagonize EGFR function are receiving some attention in early clinical trials. The predominant and most characterized mechanism of resistance to gefitinib and erlotinib is the presence of the T790M “gatekeeper” mutation in EGFR. Analogous mutations have been identified for other TKI targets that have become resistant to corresponding TKI, such as BCR-Abl, c-Kit and the PDGF receptor, suggesting that these mutations reflect a common problem for the use of the TKI class of targeted therapeutics and may be overcome by a common therapeutic alternative (76,89,90). Because the T790M mutation affects the ability of reversible TKIs to bind to EGFR, it follows that irreversible inhibitors may circumvent this problem. Irreversible EGFR inhibitors such as HKI-272, HKI-357, and a compound developed by Wyeth Research, EKB-569, have shown some early efficacy ( 81, 91, 92). These compounds form covalent bonds with the EGFR protein at a distinct cysteine residue C773, within the kinase

120

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

domain (93,94). The irreversible inhibitors are also able to target C751 of ErbB2, resulting in inhibition of its kinase activity, suggesting that the potency of these compounds may be due to dual inhibition of EGFR and ErbB2.

6.2. Pan ErbB Inhibitors Sporadic mutations in ErbB2 have been identified in NSCLC, suggesting that its activity may be critical in the etiology of EGFR-driven NSCLC (95). From a biochemical point of view, it is known that EGFR/ErbB2 heterodimers are endocytosed less rapidly than EGFR homodimers because ErbB2 cannot recruit the E3 ubiquitin ligase c-Cbl (96). Thus, EGFR/ErbB2 heterodimers exhibit a higher degree of constitutive activation. In addition mutated ErbB2/HER2 can confer resistance of lung cancer cells to EGFR TKIs, an observation that has clinical implications for EGFR targeted therapy in lung cancers that exhibit HER2 mutations (97). Dual or pan-erbB reversible kinase inhibitors have been developed such as lapatinib (GlaxoSmithKline) and PKI-166 (Novartis) that are currently being tested in clinical trials ( 98, 99). Such compounds may prove to be efficacious in the treatment of T790M-negative gefitinib-resistant tumors that depend on other ErbB proteins for viability.

6.3. Other Strategies to Target EGFR function Other classes of compounds to target various aspects of EGFR biology and biochemistry have been proposed. Clinically, EGFR over-expression can occur in NSCLC upon treatment with EGFR TKIs and this may represent an acquired resistance mechanism. Certain compounds such as vitamin D and retinoic acid, which promote changes in gene expression via ligation of steroid-like nuclear receptors, can cause decreased EGFR gene transcription due to the presence of response elements in the promoter region of the gene (100,101). Thus, vitamin D and retinoic acid analogues could be used to reduce EGFR expression in NSCLC where long-term treatment with EGFR TKIs has lead to its over-expression. Another potential “Achille’s heel” for mutant EGFR proteins, which it shares with several other oncoproteins, is that it requires the activity of chaperone proteins such as hsp90 to maintain a correct tertiary fold ( 102). In the absence of hsp90 activity, EGFR becomes misfolded and is rapidly degraged via the ubiquitin/proteosome machinery. Geldanamycin, an ansamycin antibiotic, can potently block the activity of the ATP-dependent hsp90 and has been shown to facilitate the rapid degradation of mutant EGFR in NSCLC cell lines harboring such mutants, including NCI-H1975, which harbors a T790M TKI-resistance mutation (102). Thus, geldanamycin or its derivatives may be useful agents in the treatment of NSCLC patients that have become refractory to EGFR TKIs. A mechanism by which ErbB receptor ligands can be released to engender ErbB activation, is via proteolytic cleavage of cell-surface tethered proligands (103). This is mediated by extracellular matrix metalloproteases, such as those belonging to the ADAM family. In some gefitinib-resistant NSCLC cell lines, it appears that an autocrine signaling loop involving the proteolytic release of heregulin by the ligand sheddase ADAM17 leads to ErbB2/ErbB3 hyperactivation (104). Thus, an ADAM inhibitor, INCB3619 can inhibit ErbB signaling in these cell lines and sensitize them to growth inhibitory

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

121

effects of gefitinib. Such studies have not been validated in vivo but may provide some promise as a means to target EGFR in the 90% of NSCLC cases that are unresponsive to gefitinib.

7. CONCLUSIONS The promise of target-based therapies for treatment of cancer is a tantalizing one. Ideally, the hope is to identify a unitary oncogenic driving force in a given patient and treat the disease with a potent and highly selective inhibitor. Such ideals have come close to fruition with the introduction of selective EGFR TKIs such as gefitinib and erlotinib which exhibit remarkable and specific efficacy, albeit for a small subset of NSCLC patients. These TKIs have not only aided in the treatment of NSCLC, but have also helped to reveal a pharmacogenomic profile of response leading to the identification of oncogenic EGFR mutations, approximately 20 years after the avian erythroblastosis virus was found to express transforming genes homologous to EGFR. Thus, it appears that ErbB family proteins should be attractive targets for drug discovery efforts, in NSCLC as well as other solid tumors. Primary and acquired resistance to EGFR TKIs has forced researchers to identify alternative targets or therapeutic strategies that may aid in targeting EGFR mutants in cancer. The major challenges that are manifest in this arena include the full understanding of the molecular and cellular basis for primary and acquired resistance, such as the identification of oncogenic determinants that may drive EGFR-independent cancer cell growth and survival. Such challenges are only now being addressed at the basic science level, and it is only a matter of time until these findings are translated into clinical practice.

REFERENCES 1. Normanno N, De Luca A, Bianco C et al. Epidermal growth factor receptor (EGFR) signaling in cancer. Gene 2006;366:2–16. 2. Singh AB, Harris RC. Autocrine, paracrine, and juxtacrine signaling by EGFR ligands. Cell Signal 2005;17:1183–1193. 3. Hafen E, Basler K, Edstroem JE et al. Sevenless, a cell-specific homeotic gene of Drosophila, encodes a putative transmembrane receptor with a tyrosine kinase domain. Science 1987;236:55–63. 4. Sibilia M, Wagner EF. Strain-dependent epithelial defects in mice lacking the EGF receptor. Science 1995;269:234–238. 5. Vennstrom B, Bishop JM. Isolation and characterization of chicken DNA homologous to the two putative oncogenes of avian erythroblastosis virus. Cell 1982;28:135–143. 6. Lin CR, Chen WS, Kruiger W et al. Expression cloning of human EGF receptor complementary DNA: gene amplification and three related messenger RNA products in A431 cells. Science 1984;224: 843–848. 7. Nicholson RI, Gee JM, Harper ME. EGFR and cancer prognosis. Eur J Cancer 2001;37 Suppl 4: S9–15. 8. Hynes NE, Lane HA. ERBB receptors and cancer: the complexity of targeted inhibitors. Nat Rev Cancer 2005;5:341–354. 9. Schlessinger J. Ligand-induced, receptor-mediated dimerization and activation of EGF receptor. Cell 2002;110:669–672. 10. Yarden Y, Sliwkowski MX. Untangling the ErbB signalling network. Nat Rev Mol Cell Biol 2001;2:127–137.

122

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

11. Guy PM, Platko JV, Cantley LC et al. Insect cell-expressed p180erbB3 possesses an impaired tyrosine kinase activity. Proc Natl Acad Sci USA 1994;91:8132–8136. 12. Zhang X, Gureasko J, Shen K et al. An allosteric mechanism for activation of the kinase domain of epidermal growth factor receptor. Cell 2006;125:1137–1149. 13. Jeffrey PD, Russo AA, Polyak K et al. Mechanism of CDK activation revealed by the structure of a cyclinA-CDK2 complex. Nature 1995;376:313–320. 14. Dikic I. Mechanisms controlling EGF receptor endocytosis and degradation. Biochem Soc Trans 2003;31:1178–1181. 15. Levkowitz G, Waterman H, Zamir E et al. c-Cbl/Sli-1 regulates endocytic sorting and ubiquitination of the epidermal growth factor receptor. Genes Dev 1998;12:3663–3674. 16. Benmerah A. Endocytosis: signaling from endocytic membranes to the nucleus. Curr Biol 2004;14:R314–316. 17. Gotoh N, Tojo A, Hino M et al. A highly conserved tyrosine residue at codon 845 within the kinase domain is not required for the transforming activity of human epidermal growth factor receptor. Biochem Biophys Res Commun 1992;186:768–774. 18. Tice DA, Biscardi JS, Nickles AL et al. Mechanism of biological synergy between cellular Src and epidermal growth factor receptor. Proc Natl Acad Sci USA 1999;96:1415–1420. 19. Sato K, Sato A, Aoto M et al. c-Src phosphorylates epidermal growth factor receptor on tyrosine 845. Biochem Biophys Res Commun 1995;215:1078–1087. 20. Carpenter G. Receptor tyrosine kinase substrates: Src homology domains and signal transduction. Faseb J 1992;6:3283–3289. 21. Soltoff SP, Cantley LC. p120cbl is a cytosolic adapter protein that associates with phosphoinositide 3-kinase in response to epidermal growth factor in PC12 and other cells. J Biol Chem 1996;271: 563–567. 22. Soltoff SP, Carraway KL, 3rd, Prigent SA et al. ErbB3 is involved in activation of phosphatidylinositol 3-kinase by epidermal growth factor. Mol Cell Biol 1994;14:3550–3558. 23. Park OK, Schaefer TS, Nathans D. In vitro activation of Stat3 by epidermal growth factor receptor kinase. Proc Natl Acad Sci USA 1996;93:13704–13708. 24. Hu P, Margolis B, Skolnik EY et al. Interaction of phosphatidylinositol 3-kinase-associated p85 with epidermal growth factor and platelet-derived growth factor receptors. Mol Cell Biol 1992;12:981–990. 25. Li N, Batzer A, Daly R et al. Guanine-nucleotide-releasing factor hSos1 binds to Grb2 and links receptor tyrosine kinases to Ras signalling. Nature 1993;363:85–88. 26. Denhardt DT. Signal-transducing protein phosphorylation cascades mediated by Ras/Rho proteins in the mammalian cell: the potential for multiplex signalling. Biochem J 1996;318:729–747. 27. Cantley LC. The phosphoinositide 3-kinase pathway. Science 2002;296:1655–1657. 28. Kloth MT, Laughlin KK, Biscardi JS et al. STAT5b, a mediator of synergism between c-src and the epidermal growth factor receptor. J Biol Chem 2003;278:1671–1679. 29. Boerner JL, Demory ML, Silva C et al. Phosphorylation of Y845 on the epidermal growth factor receptor mediates binding to the mitochondrial protein cytochrome c oxidase subunit II. Mol Cell Biol 2004;24:7059–7071. 30. Todaro GJ, De Larco JE. Growth factors produced by sarcoma virus-transformed cells. Cancer Res 1978;38:4147–4154. 31. de Larco JE, Todaro GJ. Growth factors from murine sarcoma virus-transformed cells. Proc Natl Acad Sci USA 1978;75:4001–4005. 32. Putnam EA, Yen N, Gallick GE et al. Autocrine growth stimulation by transforming growth factoralpha in human non-small cell lung cancer. Surg Oncol 1992;1:49–60. 33. Rusch V, Baselga J, Cordon-Cardo C et al. Differential expression of the epidermal growth factor receptor and its ligands in primary non-small cell lung cancers and adjacent benign lung. Cancer Res 1993;53:2379–2385. 34. Mendelsohn J. Epidermal growth factor receptor as a target for therapy with antireceptor monoclonal antibodies. J Natl Cancer Inst Monogr 1992:125–131.

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

123

35. Veale D, Ashcroft T, Marsh C et al. Epidermal growth factor receptors in non-small cell lung cancer. Br J Cancer 1987;55:513–516. 36. Veale D, Kerr N, Gibson GJ et al. Characterization of epidermal growth factor receptor in primary human non-small cell lung cancer. Cancer Res 1989;49:1313–1317. 37. Hirsch FR, Varella-Garcia M, Bunn PA, Jr. et al. Epidermal growth factor receptor in non-smallcell lung carcinomas: correlation between gene copy number and protein expression and impact on prognosis. J Clin Oncol 2003;21:3798–3807. 38. Testa JR, Siegfried JM. Chromosome abnormalities in human non-small cell lung cancer. Cancer Res 1992;52(9 Suppl):2702s–2706s. 39. Brambilla E, Travis WD, Colby TV et al. The new World Health Organization classification of lung tumours. Eur Respir J 2001;18:1059–1068. 40. Rapp E, Pater JL, Willan A et al. Chemotherapy can prolong survival in patients with advanced non-small-cell lung cancer: report of a Canadian multicenter randomized trial. J Clin Oncol 1988;6: 633–641. 41. Deininger M, Buchdunger E, Druker BJ. The development of imatinib as a therapeutic agent for chronic myeloid leukemia. Blood 2005;105:2640–2653. 42. Blackhall F, Ranson M, Thatcher N. Where next for gefitinib in patients with lung cancer? Lancet Oncol 2006;7:499–507. 43. Herbst RS, Bunn PA, Jr. Targeting the epidermal growth factor receptor in non-small cell lung cancer. Clin Cancer Res 2003;9:5813–5824. 44. Kris MG, Natale RB, Herbst RS et al. Efficacy of gefitinib, an inhibitor of the epidermal growth factor receptor tyrosine kinase, in symptomatic patients with non-small cell lung cancer: a randomized trial. JAMA 2003;290:2149–2158. 45. Lynch TJ, Bell DW, Sordella R et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med 2004;350: 2129–2139. 46. Pao W, Miller V, Zakowski M et al. EGF receptor gene mutations are common in lung cancers from “never smokers” and are associated with sensitivity of tumors to gefitinib and erlotinib. Proc Natl Acad Sci USA 2004;101:13306–13311. 47. Frederick L, Wang XY, Eley G et al. Diversity and frequency of epidermal growth factor receptor mutations in human glioblastomas. Cancer Res 2000;60:1383–1387. 48. Paez JG, Janne PA, Lee JC et al. EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy. Science 2004;304:1497–1500. 49. Haber DA, Bell DW, Sordella R et al. Molecular targeted therapy of lung cancer: EGFR mutations and response to EGFR inhibitors. Cold Spring Harb Symp Quant Biol 2005;70:419-426. 50. Kwak EL, Jankowski J, Thayer SP et al. Epidermal growth factor receptor kinase domain mutations in esophageal and pancreatic adenocarcinomas. Clin Cancer Res 2006;12:4283–4287. 51. Jackman DM, Yeap BY, Sequist LV et al. Exon 19 deletion mutations of epidermal growth factor receptor are associated with prolonged survival in non-small cell lung cancer patients treated with gefitinib or erlotinib. Clin Cancer Res 2006;12:3908–3914. 52. Mitsudomi T, Kosaka T, Endoh H et al. Mutations of the epidermal growth factor receptor gene predict prolonged survival after gefitinib treatment in patients with non-small-cell lung cancer with postoperative recurrence. J Clin Oncol 2005;23:2513–2520. 53. Riely GJ, Pao W, Pham D et al. Clinical course of patients with non-small cell lung cancer and epidermal growth factor receptor exon 19 and exon 21 mutations treated with gefitinib or erlotinib. Clin Cancer Res 2006;12:839–844. 54. Mellinghoff IK, Wang MY, Vivanco I et al. Molecular determinants of the response of glioblastomas to EGFR kinase inhibitors. N Engl J Med 2005;353:2012–2024. 55. Sordella R, Bell DW, Haber DA et al. gefitinib-sensitizing EGFR mutations in lung cancer activate anti-apoptotic pathways. Science 2004;305:1163–1167.

124

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

56. Carey KD, Garton AJ, Romero MS et al. Kinetic analysis of epidermal growth factor receptor somatic mutant proteins shows increased sensitivity to the epidermal growth factor receptor tyrosine kinase inhibitor, erlotinib. Cancer Res 2006;66:8163–8171. 57. Fabian MA, Biggs WH, 3rd, Treiber DK et al. A small molecule-kinase interaction map for clinical kinase inhibitors. Nat Biotechnol 2005;23:329–336. 58. Engelman JA, Cantley LC. The role of the ErbB family members in non-small cell lung cancers sensitive to epidermal growth factor receptor kinase inhibitors. Clin Cancer Res 2006;12:4372s–4376s. 59. Fujimoto N, Wislez M, Zhang J et al. High expression of ErbB family members and their ligands in lung adenocarcinomas that are sensitive to inhibition of epidermal growth factor receptor. Cancer Res 2005;65:11478–11485. 60. Coldren CD, Helfrich BA, Witta SE et al. Baseline gene expression predicts sensitivity to gefitinib in non-small cell lung cancer cell lines. Mol Cancer Res 2006;4:521–528. 61. Witta SE, Gemmill RM, Hirsch FR et al. Restoring E-cadherin expression increases sensitivity to epidermal growth factor receptor inhibitors in lung cancer cell lines. Cancer Res 2006;66:944–950. 62. Walker F, Kato A, Gonez LJ et al. Activation of the Ras/mitogen-activated protein kinase pathway by kinase-defective epidermal growth factor receptors results in cell survival but not proliferation. Mol Cell Biol 1998;18:7192–7204. 63. Ji H, Li D, Chen L et al. The impact of human EGFR kinase domain mutations on lung tumorigenesis and in vivo sensitivity to EGFR-targeted therapies. Cancer Cell 2006;9:485–495. 64. Politi K, Zakowski MF, Fan PD et al. Lung adenocarcinomas induced in mice by mutant EGF receptors found in human lung cancers respond to a tyrosine kinase inhibitor or to down-regulation of the receptors. Genes Dev 2006;20:1496–1510. 65. Weinstein IB. Addiction to oncogenes: the Achilles heal of cancer. Science 2002;297:63–64. 66. Sharma SV, Fischbach MA, Haber DA et al. “Oncogenic shock”: explaining oncogene addiction through differential signal attenuation. Clin Cancer Res 2006;12:4392s–4395s. 67. Sharma SV, Gajowniczek P, Way IP et al. A common signaling cascade may underlie “addiction” to the Src, BCR-ABL, and EGF receptor oncogenes. Cancer Cell 2006;10:425–435. 68. Lynch T, Jr., Kim E. Optimizing chemotherapy and targeted agent combinations in NSCLC. Lung Cancer 2005;50:S25–S32. 69. Sequist LV, Haber DA, Lynch TJ. Epidermal growth factor receptor mutations in non-small cell lung cancer: predicting clinical response to kinase inhibitors. Clin Cancer Res 2005;11:5668–5670. 70. Pao W, Wang TY, Riely GJ et al. KRAS mutations and primary resistance of lung adenocarcinomas to gefitinib or erlotinib. PLoS Med 2005;2:e17. 71. Slebos RJ, Kibbelaar RE, Dalesio O et al. K-ras oncogene activation as a prognostic marker in adenocarcinoma of the lung. N Engl J Med 1990;323:561–565. 72. Serrano M, Lin AW, McCurrach ME et al. Oncogenic ras provokes premature cell senescence associated with accumulation of p53 and p16INK4a. Cell 1997;88:593–602. 73. Janda E, Lehmann K, Killisch I et al. Ras and TGF[beta] cooperatively regulate epithelial cell plasticity and metastasis: dissection of Ras signaling pathways. J Cell Biol 2002;156:299–313. 74. Soria JC, Lee HY, Lee JI et al. Lack of PTEN expression in non-small cell lung cancer could be related to promoter methylation. Clin Cancer Res 2002;8:1178–1184. 75. Hay N. The Akt-mTOR tango and its relevance to cancer. Cancer Cell 2005;8:179–183. 76. Gorre ME, Mohammed M, Ellwood K et al. Clinical resistance to STI-571 cancer therapy caused by BCR-ABL gene mutation or amplification. Science 2001;293:876–880. 77. Haber DA, Settleman J. Overcoming acquired resistance to Iressa/Tarceva with inhibitors of a different class. Cell Cycle 2005;4:1057–1059. 78. Pao W, Miller VA, Politi KA et al. Acquired resistance of lung adenocarcinomas to gefitinib or erlotinib is associated with a second mutation in the EGFR kinase domain. PLoS Med 2005;2:e73. 79. Engelman JA, Mukohara T, Zejnullahu K et al. Allelic dilution obscures detection of a biologically significant resistance mutation in EGFR-amplified lung cancer. J Clin Invest 2006;116:2695–2706. 80. Bell DW, Gore I, Okimoto RA et al. Inherited susceptibility to lung cancer may be associated with the T790M drug resistance mutation in EGFR. Nat Genet 2005;37:1315–1316.

Chapter 8 / Targeting Mutant EGFR in Lung Cancer

125

81. Kwak EL, Sordella R, Bell DW et al. Irreversible inhibitors of the EGF receptor may circumvent acquired resistance to gefitinib. Proc Natl Acad Sci USA 2005;102:7665–7670. 82. Elkind NB, Szentpetery Z, Apati A et al. Multidrug transporter ABCG2 prevents tumor cell death induced by the epidermal growth factor receptor inhibitor Iressa (ZD1839, gefitinib). Cancer Res 2005;65:1770–1777. 83. Ozvegy-Laczka C, Cserepes J, Elkind NB et al. Tyrosine kinase inhibitor resistance in cancer: role of ABC multidrug transporters. Drug Resist Updat 2005;8:15–26. 84. Ozvegy-Laczka C, Hegedus T, Varady G et al. High-affinity interaction of tyrosine kinase inhibitors with the ABCG2 multidrug transporter. Mol Pharmacol 2004;65:1485–1495. 85. Kitazaki T, Oka M, Nakamura Y et al. Gefitinib, an EGFR tyrosine kinase inhibitor, directly inhibits the function of P-glycoprotein in multidrug resistant cancer cells. Lung Cancer 2005;49:337–343. 86. Nakamura Y, Oka M, Soda H et al. Gefitinib (“Iressa”, ZD1839), an epidermal growth factor receptor tyrosine kinase inhibitor, reverses breast cancer resistance protein/ABCG2-mediated drug resistance. Cancer Res 2005;65:1541–1546. 87. Pardal R, Clarke MF, Morrison SJ. Applying the principles of stem-cell biology to cancer. Nat Rev Cancer 2003;3:895–902. 88. Reya T, Morrison SJ, Clarke MF et al. Stem cells, cancer, and cancer stem cells. Nature 2001;414: 105–111. 89. Blencke S, Zech B, Engkvist O et al. Characterization of a conserved structural determinant controlling protein kinase sensitivity to selective inhibitors. Chem Biol 2004;11:691–701. 90. Daub H, Specht K, Ullrich A. Strategies to overcome resistance to targeted protein kinase inhibitors. Nat Rev Drug Discov 2004;3:1001–1010. 91. Rabindran SK, Discafani CM, Rosfjord EC et al. Antitumor activity of HKI-272, an orally active, irreversible inhibitor of the HER-2 tyrosine kinase. Cancer Res 2004;64:3958–3965. 92. Wissner A, Overbeek E, Reich MF et al. Synthesis and structure-activity relationships of 6,7disubstituted 4-anilinoquinoline-3-carbonitriles. The design of an orally active, irreversible inhibitor of the tyrosine kinase activity of the epidermal growth factor receptor (EGFR) and the human epidermal growth factor receptor-2 (HER-2). J Med Chem 2003;46:49–63. 93. Fry DW, Bridges AJ, Denny WA et al. Specific, irreversible inactivation of the epidermal growth factor receptor and erbB2, by a new class of tyrosine kinase inhibitor. Proc Natl Acad Sci USA 1998;95:12022–12027. 94. Singh J, Dobrusin EM, Fry DW et al. Structure-based design of a potent, selective, and irreversible inhibitor of the catalytic domain of the erbB receptor subfamily of protein tyrosine kinases. J Med Chem 1997;40:1130–1135. 95. Shigematsu H, Takahashi T, Nomura M et al. Somatic mutations of the HER2 kinase domain in lung adenocarcinomas. Cancer Res 2005;65:1642–1646. 96. Haslekas C, Breen K, Pedersen KW et al. The inhibitory effect of ErbB2 on epidermal growth factorinduced formation of clathrin-coated pits correlates with retention of epidermal growth factor receptorErbB2 oligomeric complexes at the plasma membrane. Mol Biol Cell 2005;16:5832–5842. 97. Wang SE, Narasanna A, Perez-Torres M et al. HER2 kinase domain mutation results in constitutive phosphorylation and activation of HER2 and EGFR and resistance to EGFR tyrosine kinase inhibitors. Cancer Cell 2006;10:25–38. 98. Johnston SR, Leary A. Lapatinib: a novel EGFR/HER2 tyrosine kinase inhibitor for cancer. Drugs Today (Barc) 2006;42:441–453. 99. Langer CJ. Emerging role of epidermal growth factor receptor inhibition in therapy for advanced malignancy: focus on NSCLC. Int J Radiat Oncol Biol Phys 2004;58:991–1002. 100. McGaffin KR, Acktinson LE, Chrysogelos SA. Growth and EGFR regulation in breast cancer cells by vitamin D and retinoid compounds. Breast Cancer Res Treat 2004;86:55–73. 101. McGaffin KR, Chrysogelos SA. Identification and characterization of a response element in the EGFR promoter that mediates transcriptional repression by 1,25-dihydroxyvitamin D3 in breast cancer cells. J Mol Endocrinol 2005;35:117–133.

126

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

102. Shimamura T, Lowell AM, Engelman JA et al. Epidermal growth factor receptors harboring kinase domain mutations associate with the heat shock protein 90 chaperone and are destabilized following exposure to geldanamycins. Cancer Res 2005;65:6401–6408. 103. Zhou BB, Fridman JS, Liu X et al. ADAM proteases, ErbB pathways, and cancer. Expert Opin Investig Drugs 2005;14:591–606. 104. Zhou BB, Peyton M, He B et al. Targeting ADAM-mediated ligand cleavage to inhibit HER3 and EGFR pathways in non-small cell lung cancer. Cancer Cell 2006;10:39–50.

9

BCR-ABL Mutations and Imatinib Resistance in Chronic Myeloid Leukemia Patients Mark R. Litzow, MD CONTENTS Introduction Clinical As pects Imatinib Res is tance Overcoming Imatinib Res is tance Conclus ion Acknowledgment References

S UMMARY The discovery of a translocation between the Abelson oncogene (ABL1) on the long arm of chromosome 9 and the breakpoint cluster region (BCR) on chromosome 22 resulting in the BCR-ABL1 gene mutation was a landmark discovery in the pathogenesis of human leukemia. The decades of research into the structure and function of this aberrant gene culminated in development of imatinib mesylate, which has had an astounding benefit in the treatment of this disease. Despite its resounding success, a minority of patients developed resistance to imatinib. The unraveling of the multiple resistance mechanisms and the discovery of the dominant mechanism of point mutations in the ABL1 kinase portion of BCR-ABL1 has led to the development of second and subsequent generation agents that are active against these mutations. Dasatinib R ) has rapidly achieved FDA approval for imatinib-resistant and intolerant (Sprycel From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 9, 

127

128

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

R CML. Nilotinib (Tasigna ) is undergoing FDA review. A host of other kinase inhibitors are in earlier stages of pre-clinical and clinical development. The structure– function relationships of imatinib in complex with BCR-ABL1 and the development of these new agents will certainly result in the use of combination therapy for CML and likely result in excellent long-term disease control and a probable cure. These developments have fueled the development of targeted therapy in multiple other malignancies and diseases and represent the beginnings of a golden age in the treatment of human disease.

Key Words: Chronic myeloid leukemia; imatinib resistance; BCR-ABL1; Mutations; dasatinib; nilotinib; tyrosine kinase inhibitors

1. INTRODUCTION In 1960, a minute acrocentric chromosome was noted in cells from seven patients with chronic myeloid leukemia (CML) (1). The subsequent identification of this abnormal chromosome 22, which came to be referred to as the Philadelphia chromosome, has become the basis for an explosion in knowledge over the past 40-plus years that culminated in the development of imatinib mesylate (IM), a highly effective targeted therapy of CML that is producing long-term disease control and a possible cure (2). In 1973, it was recognized that the Philadelphia chromosome was actually the result of a reciprocal translocation between chromosomes 9 and 22 that has come to be designated as t(9;22) (q34;q11) (3). This reciprocal translocation resulted in the juxtaposition of the Abelson (ABL1) oncogene, a tyrosine kinase on chromosome 9, with a gene of unknown function on the long arm of chromosome 22, referred to as the breakpoint cluster region (BCR) (4). The identification of this hybrid abnormal gene, BCR-ABL1, was of major significance for multiple reasons. Identification of this gene rearrangement is the sine qua non for the diagnosis of CML, and its presence in the hematopoietic cells of virtually all patients with CML made it an ideal target for the development of therapeutic agents. Proof that BCR-ABL1 was crucial for the development of CML was shown when the hybrid gene was transfected into bone marrow cells of mice which were subsequently transplanted into an irradiated syngeneic recipient and shown to lead to the development of a CML-like syndrome (5). The breakpoint in the ABL1 gene that results in the BCR-ABL1 translocation is remarkably consistent from patient to patient and usually occurs in exon 2. In contrast, the breakpoint on the BCR gene can occur in multiple regions, and it is this variation that results in different-sized hybrid aberrant tyrosine kinase genes. The two most common breakpoints in BCR occur either in the “major” breakpoint cluster region (M-BCR) of the BCR gene between exons 12 and 16 (also known as b1 to b5) or in a “minor” BCR (m-BCR) region that exclude exons e1 and e2 of the BCR gene. Two chimeric proteins result from these breakpoints including a 210-kd gene known as the p210 BCR-ABL1 resulting from the break in the M-BCR and a smaller 190-kd BCR-ABL1 known as p190 BCR-ABL. The p210 BCR-ABL occurs most frequently in cases of CML whereas the 190-kd BCR-ABL is more commonly seen in Philadelphia chromosome-positive acute lymphoblastic leukemia (Ph±ALL) (6). Other less common breakpoints have also been seen in the BCR gene resulting in other chimeric proteins

Chapter 9 / Imatinib Resistance in CML

129

with the most common being a break in the e19-e20 exons (known as ␮-BCR) and producing a 230-kd protein known as p230 BCR-ABL. This disorder most often results in a chronic neutrophilic leukemia (6) (Fig. 1A,B). A primitive hematopoietic progenitor cell is thought to be the cell from which CML arises. Whether the BCR-ABL1 mutation is the initial mutation that occurs or is a subsequent mutation is unclear, but these abnormal cells are able to gain a growth advantage over normal hematopoietic cells with resultant suppression of normal hematopoiesis ( 7). Multiple mechanism are responsible for the dominance of the CML cells in hematopoiesis in afflicted patients and are not only likely a result of a proliferative advantage, but also lilely result in prolonged survival as a consequence of decreased apoptosis and altered adherence to marrow stromal elements. These latter mechanisms enhance the release of CML progenitors into the peripheral blood (8,9).

Fig. 1. The translocation of t(9;22)(q34;q11) in CML. The Philadelphia (Ph) chromosome is a shortened chromosome 22 that results from the translocation of 3 (toward the telomere) ABL segments on chromosome 9 to 5 BCR segments on chromosome 22. Breakpoints (arrowheads) on the ABL gene are located 5 (toward the centromere) of exon a2 in most cases. Various breakpoint locations have been identified along the BCR gene on chromosome 22. Depending on which breakpoints are involved, different-sized segments from BCR are fused with the 3 sequences of the ABL gene. This results in fusion messenger RNA molecules (e1a2, b2a2, b3a2, and e19a2) of different lengths that are transcribed into chimeric protein products (p190, p210, and p230) with variable molecular weights and presumably variable function. The abbreviation m-bcr denotes minor breakpoint cluster region, M-bcr major breakpoint cluster region, and ␮-bcr a third breakpoint location in the BCR gene that is downstream from the M-bcr region between exons e19 and e20. (Reprinted with permission from Ref. (6)).

130

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

In normal cells, the ABL1 protein is found in both the nucleus and cytoplasm and can shuttle between these two locations ( 10). In contrast, the BCR-ABL1 protein is exclusively found in the cytoplasm where it is constitutively activated. The ABL1 protein has three SRC-homology domains: SH1, SH2, and SH3 (Fig. 1C). The SH1 domain is a kinase domain and is organized into an N-lobe and C-lobe with the adenosine triphosphate (ATP)-binding catalytic site positioned between these two lobes. In the BCR portion of the molecule, the coil-motif encoded by the first BCR exon is responsible for dimerization of the BCR-ABL1 (11). The normal c-ABL protein is structurally regulated by auto-inhibition in a fashion similar to that of c-SRC. This auto-inhibition depends on intramolecular interactions between the SH3 and SH2 domains and a linker between SH3 and SH2 known as the catalytic domain linker (CD-linker) and the myristoyl group from the cap or N-terminus region that is upstream of the SH3 domain (12). A “switch-clamp-latch mechanism” is critical for the quiescent state of c-ABL normally. The switch refers to a flip that occurs in an activation loop from the C-lobe that goes between opened and closed conformations and represent active and inactive states of the enzyme, respectively (Fig. 2). This “clamp” or loop limits the access of ATP and other substrates to the active binding site of the ABL1 protein. A “latch” stabilizes

Fig. 2. Structure of the ABL1 kinase portion of the BCR-ABL1 protein. The activation loop (blue) is in the closed (inactive) conformation on the left and in the open (active) conformation on the right. A molecule of imatinib is positioned in the ATP-binding site and is in green. (Reprinted with permission from U.S. Healthcare Communications, LLC. Litzow MR, Tefferi A. Chronic myeloid leukemia: problems propel progress. Amer J Hematol/Oncol, 2007; 6(5) supplement 7:19–22).

Chapter 9 / Imatinib Resistance in CML

131

Fig. 3. A simplified illustration of BCR-ABL and SRC family kinase involvement in oncogenic signaling pathways. The inhibitory effect is indicated by the upside-down T’s. ABL = Abelson tyrosine kinase; BCR = breakpoint cluster region; FAK = focal adhesion kinase; Grb-2 = growth factor receptor–bound protein 2; HcK = hematopoietic cell kinase; JNK = Jun amino-terminal kinase; P = phosphate group; PI3 K = phosphatidylinositol-3–kinase; SFK = SRC family kinases; Stat5 = signal transducer and activator of transcription 5. (Reprinted with permission from Ref. (123)).

the interaction with a myristoyl chain from the cap region. If any of these modules are disrupted, tyrosine kinase activity of the ABL is activated leading to cell growth (13). Thus, the BCR-ABL1 protein is able to phosphorylate tyrosine on a large number of substrate molecules and enhance the proliferative potential of the malignant cell (Fig. 3).

2. CLINICAL ASPECTS The natural history of ineffectively treated CML is one of a triphasic clinical course. There is an initial chronic phase characterized by mature leukocytosis and/or thrombocytosis with some immature myeloid elements in the blood and frequent basophilia that is usually 56 years in duration. Subsequently, CML will transform into an acute leukemic phase that can have either a lymphoid or myeloid phenotype. Although transiently controllable with chemotherapy, the development of this blast crisis usually heralds the demise of the patient. Some patients will develop an accelerated phase after the chronic phase and prior to the blast phase that is characterized by decreased disease control and/or additional chromosomal abnormalities (6).

132

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Up until the 1970s, CML was incurable. The development of allogeneic hematopoietic stem cell transplantation (HSCT) was subsequently shown to provide long-term disease eradication with prolonged disease-free survival (14). However, allogeneic HSCT is available to only a minority of patients because of the necessity of finding a suitable donor for transplantation and the toxicity of the procedure. In the 1980s, the availability of interferon-␣ (IFN-␣) and a demonstration that it had anti-proliferative and immunoregulatory effects led to clinical trials in patients with CML. Although IFN-␣ treatment is relatively toxic, much excitement was generated by the observation that patients who demonstrated a major (< 33% Ph-positive cells) or a complete cytogenetic response (0% Ph-positive cells) in their bone marrow had prolonged survival compared to patients with lesser cytogenetic responses (15,16). Randomized trials subsequently demonstrated the superiority of IFN-␣ compared to conventional chemotherapy including busulfan and hydroxyurea and it became the standard of care treatment for patients with CML who were not eligible for BMT (17). The development of IM dramatically changed the therapeutic landscape for CML. Imatinib mesylate, originally designated “signal transduction inhibitor 571” (STI-571), was developed through a random and time-consuming process of screening of a large number of compounds. Imatinib mesylate is a 2-phenyl-amino-pyrimidine that, in preclinical studies, emerged as a potent inhibitor of the ABL1 protein, although it was found to inhibit other tyrosine kinases such as c-kit and platelet-derived growth factor receptors (PDGFR) (2). A lack of significant toxicity in animal models and a favorable oral bioavailability profile led to its testing in large phase I and II trials. The initial phase I trial of IM was reported in 2001 and showed that the dose of imatinib could be steeply escalated from 25 mg/day up to as high as 1000 mg/day without a dose-limiting toxicity. Overall, adverse side effects were minimal with the most common being edema, myalgias, diarrhea, and nausea. The initial phase I trial included 83 patients with CML in chronic phase who had failed treatment with IFN-␣. A high rate of complete hematologic responses (CHR) was seen in 53 of 54 patients who received 300 mg/day or more of imatinib. Cytogenetic responses were observed in 29 patients, with 17 achieving a major cytogenetic response (MCR) and 7 a complete cytogenetic response (CCR) ( 18). Responses were also demonstrated in a second phase I trial in patients with myeloid or lymphoid blast crisis of CML, although the responses were at a lower rate and less robust (19). A large phase II trial of 532 patients with late chronic phase CML who had failed prior IFN-␣ treated patients with 400 mg of imatinib daily. A CHR was seen in 95% of these patients with MCR in 60%. Nearly 90% of the patients had not progressed to a more advanced phase of CML after 18 months of follow-up and only 2% of patients discontinued treatment because of adverse effects (20). Phase II studies in patients with accelerated and blast phase also demonstrated responses albeit at a lower level than in chronic phase patients (21,22,23). The definitive landmark trial that tested IM in patients with CML was the International Randomized trial of IFN-␣ and STI-571 (IRIS). In this study, 1106 patients, 553 in each arm, were randomized to receive IM at 400 mg daily or IFN-␣ with low-dose cytarabine. Crossover to the other arm of treatment was allowed for intolerance or lack of efficacy. The initial report with a median follow-up of 19 months showed an MCR

Chapter 9 / Imatinib Resistance in CML

133

of 87% in the IM group and 35% in the interferon-␣ plus cytarabine group. The cytogenetic responses were complete in 76% of the IM patients but only in 15% of the IFN-␣ plus cytarabine group. In the IFN-␣ plus cytarabine group, 89% of the patients crossed over to IM but only 15% of the IM group crossed over to IFN-␣ plus cytarabine. Intolerance of the IFN-␣ plus cytarabine combination was the most common reason for crossover (24). In this study, response assessment was conducted not only with bone marrow cytogenetic analysis but also with assessment of molecular responses using a quantitative real-time polymerase chain reaction (PCR) assay. Utilizing this assay, it was demonstrated that 57% of the patients in the IM group who achieved a CCR had at least a three-log reduction in BCR-ABL1 transcript, compared with 24% of the patients with a CCR in the IFN-␣ plus cytarabine group. Those patients achieving a CCR and at least a three-log reduction in BCR-ABL1 transcript at 12 months had a 100% probability of remaining progression-free at 24 months compared with 95% of patients with less than a three-log reduction and 85% of those patients not achieving a CCR at 12 months (p < 0.001) (24,25). A 5-year follow-up of patients on this trial has recently been published. At this time point, 69% of the IM group and only 3% of the IFN-␣ plus cytarabine group continued with their initially assigned treatment. By this follow-up, 98%, 92%, and 80%, respectively, of the patients receiving IM therapy initially achieved a CHR, MCR, or CCR. There were 124 patients who had a CCR and had blood samples available at 1 and 4 years of follow-up where BCR-ABL1 transcripts by PCR could be measured. In these patients, a three-log reduction was seen at 1 year in 53% of the patients and at 4 years in 80%. The estimated event-free survival was 83%, and an estimated 93% of patients had not shown progression of disease to accelerated or blast phase. Contrary to expectations, there was a declining rate of treatment failure after the start of imatinib, being 3.3% in the first year, 7.5% in the second year, 4.8% in the third, 1.5% in the fourth, and only 0.9% in the fifth year. Patients who achieved a CCR by 12 months or a complete molecular response (CMR) by 18 months had risks of progression to accelerated or blast phase of 3% and 0%, respectively. The estimated overall survival rate at 5 years was 89%. An allogeneic HSCT was carried out in 44 patients, and if these patients were censored, the estimated overall survival rate was 92% (26). Though excellent, these results indicate that some patients fail imatinib, with 6% of patients progressing to accelerated or blast phase, 3% experiencing a hematologic relapse, 5% sustaining a loss of MCR, and 2% dying from causes unrelated to CML (26). Thus, 14% of patients lost their response to IM and would be considered to have resistance.

3. IMATINIB RESISTANCE Resistance to IM can be defined at different levels and different time points. These categories are outlined in Table 1. Intrinsic or primary resistance to IM can be defined when initial IM therapy does not result in a hematologic or cytogenetic response. Secondary or acquired resistance or relapse is defined as a loss of IM efficacy after an initial response to therapy. Each of these types of resistance can be defined at the hematologic,

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

134

Table 1 Definitions of Resistancea Primary (intrinsic) Secondary (acquired) Subcategories of resistance Hematologic Chronic phase Advanced phases Cytogenetic Molecular a

See text for detailed explanations.

cytogenetic, or molecular level. Hematologic resistance must also be defined by the phase of the disease the patient is in. Hematologic resistance in the chronic phase would refer to a worsening of peripheral blood counts or the white blood cell differential or inability to reduce splenomegaly. In the accelerated or blast phase, hematologic resistance would refer to a lack of return to the chronic phase or a hematologic relapse following an initial response to therapy. At the cytogenetic level, resistance can be defined as a loss or lack of MCR or CCR. At the molecular level, a complete molecular response is defined as undetectable BCRABL1 transcripts by PCR or a ≥ 3-log reduction, a BCR-ABL1 transcripts which represents a major molecular remission. Loss of either of these responses would be defined as resistance at the molecular level. However, standardization of measurement of BCRABL1 transcripts by PCR has not yet been accomplished, and results in one lab may not necessarily be equivalent to those of another lab. Also, a standardized definition of resistance at the molecular level is not well defined, although many experts consider a half log or one-log increase which is confirmed in a second sample as a reasonable benchmark for loss of molecular response (27). An international effort is underway to harmonize methodologies for detecting BCRABL1 transcripts and kinase domain mutations and adjusting the results so they are standardized from one lab to another (28,29). The availability of a reproducible real-time quantitative PCR (RQ-PCR) has shown in one study that patients with a major molecular response do not have evidence of cytogenetic abnormalities in their bone marrow, and therefore a policy of performing bone marrow biopsies only in patients who have not achieved or have lost a major molecular response would allow many patients to forego the discomfort and expense of multiple bone marrow biopsies (30). A panel of experts has recently published a consensus opinion based on a literature review as to when to consider patients to have failed IM and to then be considered for alternate therapies. In this study, failure was defined as lack of a hematologic response after 3 months of IM, less than a complete hematologic response or no cytogenetic response (>95% positive PH chromosome) after 6 months of IM, less than an MCR after 12 months, and less than a CCR after 18 months. The development of mutations, loss of a CCR, or loss of a complete hematologic response at any time was also considered a failure. Definitions for a suboptimal response and “warning” situations such as additional chromosome abnormalities were also listed (Table 2) (31).

Chapter 9 / Imatinib Resistance in CML

135

Table 2 Operational Definition of Failure and Suboptimal Response for Previously Untreated Patients in ECP CML Who Are Treated with 400 Mg IM Daily Time

Failurea

Suboptimal responseb

Warningsc

Diagnosis

NA

NA

3 mo after diagnosis

No HR (stable disease or disease progression) Less than CHR, no CgR (Ph+ > 95%) Less than PCgR (Ph+ > 35%) Less than CCgR Loss of CHRd , loss of CCgRf , mutationg

Less than CHR

High riskd , del9qe , ACAs in Ph+ cells NA

6 mo after diagnosis 12 mo after diagnosis 18 mo after diagnosis Anytime

Less than PCgR (Ph+ > 35%) Less than CCgR

NA Less than MMolR

Less than MMolR NA ACA in Ph+ cellsh , Any rise in transcript loss of MMolRh , level; other chromosome mutationi abnormalities in Ph– cells

a

Failure implies that the patient should be moved to other treatments whenever available. Suboptimal response implies that the patient may still have a substantial benefit from continuing IM treatment but that the long-term outcome is not likely to be optimal, so the patient becomes eligible for other treatments. c Warnings imply that the patient should be monitored very carefully and may become eligible for other treatments. The same definitions can be used to define the response after IM dose escalation. d By Sokal or Hasford score. HR = hematologic response; ECP = early chronic phase; CHR = complete HR; CgR = cytogenetic response; PCgR = partial CgR (Ph + 1–35%); CCgR = complete CgR (Ph + 0%); MMolR = major molecular response (≤ 0.10 BCR-ABL1 gene ratio); ACA = additional chromosome abnormalities; NA, not applicable. e To be confirmed on two occasions unless associated with progression to AP/BC. f To be confirmed on two occasions, unless associated with CHR loss or progression to AP/BC. g High level of insensitivity to IM. h To be confirmed on two occasions, unless associated with CHR or CCgR loss. i Low level of insensitivity to IM. Reprinted with permission from: Baccarani M, Saglio G, Goldman J et al. Evolving concepts in the management of chronic myeloid leukemia: recommendations from an expert panel on behalf of the European Leukemia Net (31). b

Multiple mechanisms of resistance to IM have been defined in recent years. Four broad mechanisms have been characterized (Table 3). In vitro models that studied resistance predicted several of these mechanisms. CML cell lines that are BCR-ABL1 positive and murine hematopoietic cells that have been transformed with a BCR-ABL1 gene and then exposed to IM have subsequently developed resistance and have been used to predict several mechanisms (32). In a mouse model of IM resistance, it was demonstrated that in vivo tumors were resistant to IM but retained in vitro sensitivity (33). Several clinical studies have suggested

136

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy Table 3 Mechanisms of Imatinib Resistance

Decreased intracellular drug levels ◦ plasma binding by alpha 1-acid glycoprotein ◦ differential expression of transporter proteins (MDR-1, hOCT1) ◦ pharmacologic interactions Increased expression of BCR-ABL kinase from genomic amplification BCR-ABL independent mechanisms Clonal evolution (non- BCR-ABL-dependent mechanism) ◦ Clonal evolution ◦ Aneuploidy ◦ Over-expression of SRC family kinases Mutations in ABL kinase of BCR-ABL affecting drug interaction or kinase activity. Reprinted with permission from U.S Healthcare Communications, LLC. Litzow MR, Tefferi A. Chronic Myeloid Leukemia: Problems Propel Progress. The American Journal of Hematology/Oncology, 2007; 6(5) supplement 7:19–22.

that plasma binding of ␣-1 acid glycoprotein correlates with clinical responses to IM, although none of these studies have clearly distinguished a cause-and-effect relationship (34,35). Transporter proteins have been shown in in vitro models to contribute to IM resistance (32,36). This is known to be mediated in some instances by the multi-drug resistant gene (MDR1) producing the p-glycoprotein, where p-glycoprotein expression was associated with a decrease in intracellular IM levels and development of resistance (37). Recently, other transporter proteins mediating influx, the organic cation transporters (OCT), have been identified. It has been shown that the OCT-1 influx protein mediates transport of IM into cells and reduced OCT-1 activity appears to be a cause for low in vitro sensitivity of CML cells to IM (38,39). Down-regulation of T-cell protein tyrosine phosphatase in IM-resistant cells may represent a novel mechanism for IM resistance (40). A recently published analysis of data from the IRIS trial demonstrated that trough blood levels of IM and its active major metabolite, CGP74588, correlated with achievement of CCR and MMR (41). Gene amplification of the BCR-ABL1 kinase may be associated with development of IM resistance. The presence of multiple copies of the BCR-ABL1 gene in interphase nuclei can be demonstrated by fluorescence in situ hybridization (FISH) (40,42). In one series, more than half of the patients with IM resistance had evidence of clonal evolution with the development of additional chromosome abnormalities. Paired cytogenetic analyses performed at the beginning of IM therapy and at the time of resistance demonstrated this evolution. Chromosomal abnormalities included the presence of aneuploidy, a second Ph chromosome, and trisomy 8. The loss of one p53 allele by an alteration of the short arm of chromosome 17 was seen in seven patients, and new reciprocal translocations were seen in two patients. In eight cases, multiple cytogenetic abnormalities were also present (40).

Chapter 9 / Imatinib Resistance in CML

137

Loss of the p53 tumor suppressor gene has been shown to impede the anti-leukemic response to BCR-ABL1 inhibition ( 43). Another mechanism for resistance that is independent of BCR-ABL1 and that has been demonstrated in vitro is over-expression of SRC-related kinases such as LYN ( 44). It appears that kinases from the SRC family mediate signaling of BCR-ABL1. This has been demonstrated in vitro with the use of SRC kinase inhibitors and SRC mutants that are kinase defective (45,46,47). If one takes CML cells and cultures them in the presence of IM, or obtains cells from patients who have progressed while on IM therapy, a decrease in BCR-ABL1 mRNA or protein levels is seen with a concomitant increase in SRC family kinases ( 44). If one inhibits LYN expression by RNA interference in IM-resistant CML cells, survival and proliferation of these cells is inhibited (48). Additional support for the role of SRC kinases in some patients with IM resistance is that IM is unable to directly inhibit SRC kinases but does so only through its effect on BCR-ABL1 (49). The most frequent and increasingly best-studied mechanism of resistance to IM in patients with CML are gene mutations in the ABL1 (tyrosine kinase) domain of the BCR-ABL1 gene. In 2001, a single amino acid substitution at a threonine residue of ABL1 kinase domain was described and resulted in substitution of isoleucine for threonine at position 315 (T315I) of c-ABL1 (33). This amino acid substitution interfered with a hydrogen bond that formed between the ABL1 kinase and IM. The T315I mutation has turned out to be one of the most frequent mutations seen in patients with CML with resistance to imatinib. This altered binding of IM to the BCR-ABL1 kinase appears to confer significant resistance. A basic understanding of the structure of the BCR-ABL1 chimeric protein is important to understanding how the multiple mutations that have been described in the ABL1 kinase domain cause resistance. The c-ABL1 protein is expressed in two splice forms that are known as 1a and 1b (Fig. 1C). The 1a form is 19 residues shorter than 1b. The 1b form contains a myristoylation site on its second residue. The second residue is a glycine that appears to help regulate enzymatic activity, and its mutations to alanine prevents myristoylation and results in an activated kinase (50). The 1b form also contains a “cap” region that is believed to stabilize the inactive confirmation of the kinase (51). The numbering system used to identify the amino acid residues where mutations occur is based on the shorter 1a form. Within the c-ABL1 protein, there are three SRC homology domains as described briefly earlier. These include SH1, the kinase domain, which encodes for catalytic function; the SH2 domain, which binds phosphotyrosine-containing peptides; and SH3, which is a negative regulator of kinase activity (Fig. 1C). Toward the N-terminal end of the ABL1 kinase is the adenosine triphosphate-binding portion that is highly conserved with glycine-rich sequences and known as the P-loop. It interacts with IM through hydrogen and van deer Waals bonds. This area spans amino acids 248–256. At the carboxy- or C-terminal end of ABL1 kinase is a flexible activation loop. It is critical for the control of catalytic activity and, as its name implies, changes confirmation depending on whether the molecule is in the inactive or active state. It begins at amino acid 381. See Fig. 2 for details. Between the P-loop and the activation loop is the catalytic site of ABL1 kinase. It is located in a cleft where IM and other small molecule tyrosine kinase inhibitors bind. The shift of the activation loop between an inactive or closed confirmation and a catalytically

138

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

active or open confirmation seems to be regulated by the kinase itself in a process known as “auto-inhibition.” In the active state, the activation loop flips away from the catalytic region, whereas in the inactive state it is inward toward the catalytic region and appears to serve as a support for substrate binding (52) (Fig. 2). Further details on these structural details have been recently reviewed (51,53,54). The number of mutations within the ABL1 kinase domain of BCR-ABL1 continues to grow rapidly, and it has recently been estimated that 73 distinct point mutations causing substitutions in 50 amino acids have been found in cells from imatinib-resistant CML patients, with some being more frequently found than others (Fig. 4). Melo and Chuan have categorized these mutations into four groups, including (1) those which directly impair imatinib binding, (2) those occurring in the P-loop, (3) those within the activation loop, which prevent IM binding (IM is able to bind BCR-ABL1 kinase only when it is in the closed or inactive confirmation, and (4) those within the catalytic domain (Fig. 2) (52). Substitutions in these single nucleotides that lead to mutations change the amino acids that will, as a result, have varying effects on the confirmation of the ABL1 portion of the BCR-ABL1 protein and its binding to drugs or other substrates. In a study where the crystal structure of IM complexed to the catalytic domain of BCR-ABL1 predicted that resistance would occur with the T315I mutation because the threonine residue at position 315 forms a crucial hydrogen bond between IM and ABL1 kinase (55). In vitro random mutation of the BCR-ABL1 molecule followed by screening for IM resistance demonstrates many of the major mutations that have been identified in patients but also revealed other mutations that illustrate potentially novel mechanisms of acquired IM resistance (12). Two mutations that confer IM resistance including the T315I and the E255 K appear to enhance the activity of the BCR-ABL1 kinase through an enhanced ability to induce auto-phosphorylation. This observation would suggest that these mutations may

Fig. 4. Map of BCR-ABL kinase domain mutations associated with clinical resistance to imatinib. Abbreviations: P, P-loop; B, imatinib binding site; C, catalytic domain; A, activation loop. Aminoacid substitutions in green indicate mutations detected in 2–10% and in red in >10% of patients with mutations. (Reprinted with permission from Ref. (52)).

Chapter 9 / Imatinib Resistance in CML

139

confer a growth advantage for CML cells, even in the absence of selective pressure from treatment with IM (56). It is currently believed that mutations are not induced by IM, but rather exist before therapy is initiated, albeit at a low level, and then emerge as drug-sensitive cells are eliminated. This observation is reminiscent of the emergence of antibiotic-resistant bacteria. Indeed, multiple studies have now shown that mutations in the BCR-ABL1 gene that confer IM resistance are detectable prior to treatment with IM (57,58,59,60). Prior to therapy, these mutant clones are sometimes difficult to detect, but in other instances are more readily apparent. This again raises the question of whether these higher level mutations may confer a growth or survival advantage for the leukemic clone, even without therapy (61). However, some studies have shown that not all mutations detected before initiation of therapy were subsequently expanded by therapy with IM (60). This suggests that other factors may be necessary for development of IM resistance or that mutations may be present in cells that are unable to differentiate. The presence of a mutation in a patient may also not be the sole cause of resistance because multiple mechanisms may be contributing to resistance. Therefore, the presence of a mutation in a patient should not automatically be assumed to be the cause of resistance and needs to be interpreted within the overall clinical context (62). Additionally, not all mutations confer resistance to IM. One study found that 5 of 17 BCR-ABL1 kinase domain mutants remained sensitive to IM (63). Several studies have suggested that patients with IM resistance secondary to mutations in the P-loop have a poorer prognosis than patients with mutations in other portions of the ABL1 kinase domain. A study from Australia detected mutations in 27 of 144 patients at 17 different residues. Twenty-four of these 27 patients (89%) developed acquired resistance. Thirteen of these 24 had mutations in the P-loop, and 12 of these patients died with a median survival of only 4.5 months after mutation detection. In the 14 patients with mutations outside the P-loop, there were only three deaths, and the median follow-up was 11 months (64). Studies from France and Italy have confirmed these findings of a significantly poorer outcome in patients with P-loop mutations (65,66). These same studies showed a worse outcome for patients with the T315I mutation as well. One study, however, was unable to demonstrate a worse prognosis for patients with P-loop mutations and suggested that the prognosis of patients who fail IM is multifactorial in nature (67). Not surprisingly, mutations are more frequently found in patients in the late chronic, accelerated, or blast phase of CML and also in patients with additional chromosome abnormalities (60). Oligonucleotide microarray analysis of bone marrow samples from patients with IM-sensitive and IM-resistant CML have shown the ability to distinguish IM sensitivity and resistance. Differential gene expression patterns have the potential to identify new gene and protein targets for treatment and may have the potential to be used as a screening tool to identify patients with resistance prior to therapy (68,69,70,71,72). Mutation detection in the BCR-ABL1 gene has varying sensitivity based on the technique utilized. Methodologies vary from direct sequencing of the kinase domain (64), use of denaturing high-performance liquid chromatography (73), a flow cytometric measurement of downstream targets of BCR-ABL (74), restriction fragment length polymorphism (RFLP)-based assays (75), and various polymerase chain reaction–based

140

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

assays including those involving allele-specific oligonucleotides ( 57), peptide nucleic acid (PNA)-based clamping techniques ( 76), identification of single nucleotide polymorphisms utilizing microarrays (77), and real-time quantitative PCR (78). In this last study, a two-fold rise in BCR-ABL1 expression by RQ-PCR predicted a mutation in a kinase domain of BCR-ABL1 in 61% of patients either at the time of the rise in BCR-ABL1 expression or within 3 months of the rise ( 78). However, another study did not predict the presence of mutations in patients who had a single two-fold or greater rise in BCR-ABL1 transcripts. Rather, the development of rising BCR-ABL1 transcript levels was necessary to reliably pick up a mutation. The authors concluded that a serial rise was more reliable than a single rise (79). Eleven mutations were detected in 10 out of 82 patients in this study.

4. OVERCOMING IMATINIB RESISTANCE The understanding of mechanisms of IM resistance and of the overall biology of the BCR-ABL1 gene and protein has stimulated interest in developing therapeutic strategies to overcome resistance. Because it remains difficult to define the precise mechanism of resistance in the vast majority of patients on a routine clinical basis, the choice of alternative therapy in patients who are failing IM remains empirical. The commonly used strategy is to escalate the dose of IM since no dose-limiting toxicity was seen in the phase 1 trial of IM (18). A strategy of dose-escalating IM was also based on the efficacy of higher doses of IM in patients with accelerated and blast phase of CML, as previously described (21,22). In a trial of twice-daily administration of an IM dose of 400 mg (800 mg total), a complete hematologic response was seen in 65% of patients with cytogenetic remission, and patients with cytogenetic resistance achieved a complete cytogenetic response 56% of the time (80). The same group reported the outcome of 114 patients with newly diagnosed chronic-phase CML who were given 400 mg of IM twice daily. A MCR was achieved in 96% of patients with 90% of patients having a CCR. After a median follow-up of 15 months, no patient had progressed to accelerated or blastic phase, and the estimated 2-year survival rate was 94%. In 63% of patients, a MMR was achieved; and in 28%, a CMR was achieved. These responses were significantly better than those seen in a retrospective cohort of patients given 400 mg of IM daily. More frequent myelosuppression was seen with the high-dose regimen, but 82% of patients were able to continue to received 600 mg or more of IM daily (81). Different approaches have been considered to overcome some of the resistance mechanisms outlined earlier in this chapter. Many of these have not been tested clinically but could include blockage of p-glycoprotein–mediated drug efflux or administering agents such as erythromycin that compete for ␣-1 acid glycoprotein binding of IM (82). There is an anecdotal report of a patient who went into blast crisis of CML while taking IM and then reverted to chronic phase when IM was withdrawn (83). The rationale behind this approach is thought to be that withdrawal of IM might allow the re-emergence of unmutated leukemic clones that suppress the mutant clone by removing the competitive advantage the mutant clone has. A patient with CML resistance with a Y253H P-loop mutation discontinued IM and a reduction in a number of clones bearing the Y253H mutation was noted (84).

Chapter 9 / Imatinib Resistance in CML

141

Another approach to the management of IM resistance is to combine agents that are individually active against CML but have differing mechanisms of action that may allow either additive or synergistic effects in a non-cross-resistant manner. This approach has been extensively studied in the literature and will not be reviewed in detail here. An excellent discussion of combination therapy is found in the reviews by Hochhaus and La Rosee (52,85). Some combination approaches have utilized farnesyl transferase inhibition such as lonafarnib in combination with IM, inhibitors of the mammalian target of rapamycin (mTOR) in combination with IM and combining mycophenolic acid, an inhibitor of the JAK-STAT pathway. Because the dominant mechanism of IM resistance is the development of mutations, this area has attracted the most interest in developing therapeutic strategies. The focus has been to develop small molecules that can inhibit the BCR-ABL1 kinase protein in alternate ways compared to IM. The new class of compounds, pyridopyrimidines, which are inhibitors of SRC, have been shown to inhibit wild-type ABL1 at nanomolar concentrations (86). Several of these compounds were tested for activity against IM-resistant BCR-ABL1 mutants and demonstrated activity (87,88,89). However, these derivatives of pyridopyrimidine were predicted to have unsatisfactory pharmacokinetic profiles and further clinical development was aborted. Fortunately, an alternative compound was found to be efficacious (90). The two compounds in the most advanced stages of development included dasaR R ) and nilotinib (AMN107, Tasigna ) (Fig. 5). Dasatinib tinib (BMS354825, Sprycel

Fig. 5. Chemical formula of imatinib, the second-generation ABL kinase inhibitor, nilotinib, and the dual SRC/ABL kinase inhibitors. (Reprinted with permission from Ref. (52)).

142

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

is a synthetic, small-molecule carboxamide derivative that inhibits the SRC family of kinases. It is orally bio-available, and was recently reported to inhibit 14 of 15 IMresistant BCR-ABL1 mutants and had a two-log increased potency relative to IM (91). The T315I mutation has been consistently shown to retain resistance to both dasatinib and nilotinib, and to the other pyridopyrimidine derivatives previously described. These recently developed small-molecule inhibitors have a differential activity compared to IM that relates to several key structural elements of ABL1. The ABL1 kinase domain is bound by IM only in its inactive confirmation (with the activation loop in the closed position) (55). Because the inactive confirmations of ABL1 and SRC are distinct, IM is able to inhibit ABL1 but not SRC. As discussed previously, the activation loop can also flip into an active state, and the pyridopyrimidine derivatives and dasatinib are able to bind ABL1 whether the activation loop is in the closed or open position (inactivated or activated) (92). Thus, binding is not affected by the activation state. Dasatinib and related compounds are also smaller in size than IM, so the P-loop must undergo major conformational changes on binding with IM, whereas only minimal changes occur with dasatinib and related compounds. This dual activity of dasatinib also raises the question as to whether its broader activity may have broader effects, including potentially adverse effects in the treatment of patients. A phase I trial of dasatinib in IM-resistant Ph+ leukemias was reported last year and dosed over a range of 15–240 mg per day in once- or twice-daily doses. Complete hematologic responses were seen in 37 of 40 patients with chronic phase CML while major hematologic responses were noted in 31 of 44 patients with Ph+ ALL and blast crisis or accelerated-phase CML. Rates of MCR were 45% in chronic phase CML and 25% in the more advanced phase group. Ninety-five percent of the patients with chronic phase and 82% of the patients with accelerated phase disease had their responses maintained for a median of 12 and 5 months, respectively; whereas, virtually all patients with Ph+ ALL or lymphoid blast crisis had relapsed within 6 months. Only patients with the T315I mutation were resistant to dasatinib. The most common toxicity was myelosuppression, but this was not dose-limiting (93). Subsequent phase II trials in different phases and types of Ph+ disease are being reported. A phase II study of 186 patients with IM-resistant or -intolerant chronic phase CML with the standard dose of 70 mg orally BID has been reported. A CHR was achieved in 90% of these patients, and 52% achieved an MCR. Only 2% of patients achieving MCR progressed or died. Molecular responses were also seen with reductions in BCR-ABL1/ABL1 transcript ratios declining from 66% at baseline to 2.6% by 9 months of therapy (94). In accelerated-phase CML patients who are IM resistant or intolerant, a major hematologic response was found in 63% of patients with 43% of patients achieving a CHR and 20% showing no evidence of leukemia. An MCR was documented in 37% of patients and was a CCR in 28% and partial in 9%. The estimated progression-free survival at 9 months was 70%. Up to 80% of patients experienced grade 3 to 4 cytopenias, but non-hematologic toxicities are generally mild or moderate including diarrhea, headache, fatigue, fever, and pleural effusions (95). In IM-resistant or intolerant blast phase CML, a phase II trial demonstrated major hematologic responses in 34% of myeloid blast crisis and 31% of lymphoid blast crisis with MCR in 31% and 50% of these patients, respectively. Of the MCR achieved, 86%

Chapter 9 / Imatinib Resistance in CML

143

were CCR, and the responses were durable in 88% and 46% of myeloid blast crisis and lymphoid blast crisis patients, respectively ( 96). In patients with chronic phase CML who are resistant to IM in doses of 400–600 mg per day, a randomized comparison of dasatinib 70 mg orally BID to 800 mg per day of IM demonstrated a CHR of 92% for dasatinib vs. 82% for IM and MCR of 48% with dasatinib vs. 33% with IM. The CCR was 35% with dasatinib and 16% with IM suggesting that dasatinib may be more effective in achieving MCR than high-dose imatinib (97). Dasatinib is also active in patients who have failed both IM and nilotinib where in a small trial of 23 patients who were mostly in accelerated or blastic phase achieved a CHR in 43% with some form of cytogenetic response in 32% (98). Nilotinib was developed by scientists at Novartis by altering the N-methylpiperazine group and thus substantially increasing the selectivity and binding affinity of nilotinib for the ABL1 kinase compared with IM (99). In vitro nilotinib was found to be 20-fold more potent than IM against cells expressing wild-type or mutated BCR-ABL1 (dasatinib is 325-fold more potent than IM in the same system) (100). Similar to dasatinib, nilotinib also lacks activity against CML cells expressing the T315I mutation. In phase I testing of nilotinib against IM-resistant CML or ALL in doses ranging from 50 to 1200 mg once daily or 400 to 600 mg orally BID, 39% of blast phase CML patients had a hematologic response with 18% achieving an MCR. In 17 patients with chronic phase of CML, CHR was seen in 11 of 12 with active disease. Six of 17 CCR were seen (101). Phase II studies of nilotinib in different phases of Ph+ leukemias are also being reported. In 132 patients with IM-resistant or intolerant chronic phase CML, 69% of patients achieved a CHR. An MCR was observed in 42% of patients, and in 25% of patients a CCR was obtained. Cytopenias and elevated lipase were some of the most common toxicities. The median time to MCR was 2.6 months (102). In accelerated-phase disease, hematologic responses were observed in 44% of patients. Of these, 17% were complete. An MCR occurred in 31% of patients, and in 17% was complete (103). Patients resistant to IM and dasatinib can also respond to nilotinib with a CHR rate of 45% in chronic phase patients. An MCR was seen in 31% of patients (104). The results with dasatinib were impressive enough that the U.S. Food and Drug Administration (FDA) approved dasatinib for IM-resistant or intolerant CML in June 2006. The rapidity with which dasatinib and nilotinib were developed is striking because IM was approved by the FDA only in 2001. As Druker has pointed out, the understanding of the crystal structure of ABL1 when complexed with IM and the rapid understanding of mechanisms of relapse allowed development of modifications of IM or alternate agents (105). The development of multiple drugs that inhibit ABL1 raised the specter of utilizing drug combinations, and, indeed, in vitro studies have suggested that combining any of these three agents in pairs leads to maximal suppression of the outgrowth of resistant clones and that this can be achieved with lower concentrations of drug compared to any of the single agents (106,107,108,109). An additional concern has been whether any of the BCR-ABL1 tyrosine kinase inhibitors in development can inhibit or eliminate the leukemic stem cell. Dasatinib does appear to be active against an earlier progenitor population in the CD34+ CD38– population compared to IM, but is still not capable of eliminating the most primitive quiescent CML cells (110).

144

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

The development of these drugs has fueled an explosion of research into the discovery of alternate CML inhibitors, particularly those that may be active against the T315I mutation. The inhibitor ON012380 blocks the substrate-binding site rather than the ATP-binding site of ABL1, and in vitro studies have shown inhibition of multiple IM-resistant mutations including the T315I. This compound also demonstrates a 10-fold stronger inhibition of wild-type BCR-ABL1 compared to IM (111). Similarly, adaphostin is a tyrphostin that also alters the binding site of peptide substrates rather than that of ATP. However, adaphostin-mediated cytotoxicity is actually dependent on oxygen production and does not require BCR-ABL1, indicating—not surprisingly—that it may interact with multiple targets (112). Some examples of other agents active against BCR-ABL1 include INNO-406 and NS-187, a novel BCR-ABL1/LYN dual tyrosine kinase inhibitors ( 113, 114), and SKI-606, another dual inhibitor of SRC and ABL kinases (115). The aurora kinases are important in the regulation of mitotic chromosome segregation and cytokinesis. They show aberrant activity in a variant of human tumors. Their attractiveness in CML lies in their ability to inhibit the T315I mutation. Two of these agents, MK-0457 (formally VX-680) and VE-465, are in pre-clinical or early clinical development and show activity against the T315I mutation ( 116, 117, 118). Targeting BCR-ABL1–dependent signaling pathways required for transformation including RAS or the PI3 K pathway may also be important avenues of drug development. They are reviewed by Walz and Sattler (119).

5. CONCLUSION Since the identification of the minute Philadelphia chromosome in 1960 and its subsequent identification as a translocation with chromosome 9, our evolving understanding of the structure and function of the BCR-ABL1 aberrant tyrosine kinase has led to incredible advances in our understanding of the pathogenesis and treatment of CML. This understanding culminated in the development of IM, and this has been rapidly followed by second- and subsequent-generation agents that have built on the understanding of the action of IM and the development of resistance to it. The complex of imatinib and other small-molecule ABL1 inhibitors as determined by analysis of crystal structures will continue to assist in the development and optimization of inhibitors that are active against mutations conferring resistance in CML ( 120). The rapidity with which these developments have been occurring is truly astounding and underlies the continued optimism that the treatment of CML and disorders related to it and other disorders where the molecular mechanisms of disease are being unraveled, will become more and more treatable in the future. There is little doubt that combination therapy of CML and related disorders will be a reality in the not-distant future (121).

ACKNOWLEDGMENT The author gratefully acknowledges Mrs. Denise Chase for transcription and development of this manuscript.

Chapter 9 / Imatinib Resistance in CML

145

REFERENCES 1. Nowell PC, Hungerford DA. A minute chromosome in human chronic granulocytic leukemia. Science 1960;132:1497. 2. Druker BJ, Tamura S, Buchdunger E et al. Effects of a selective inhibitor of the ABL tyrosine kinase on the growth of BCR-ABL positive cells. Nat Med 1996;2:561–566. 3. Rowley JD. Letter: A new consistent chromosomal abnormality in chronic myelogenous leukaemia identified by quinacrine fluorescence and Giemsa staining. Nature 1973;243:290–293. 4. Shtivelman E, Lifshitz B, Gale RP et al. Fused transcript of ABL and BCR genes in chronic myelogenous leukaemia. Nature 1985;315:550–554. 5. Daley GQ, Van Etten RA, Baltimore D. Induction of chronic myelogenous leukemia in mice by the P210BCR/ABL gene of the Philadelphia chromosome. Science 1990;247:824–830. 6. Faderl S, Talpaz M, Estrov Z et al. The biology of chronic myeloid leukemia. N Engl J Med 1999;341:164–172. 7. Eaves CJ, Eaves AC. Progenitor cell dynamics. In: Carella AM, Daley GQ, Eaves CJ eds. Chronic myeloid leukaemia: biology and treatment. London: Martin Dunitz; 2001:73–100. 8. Gordon MY, Dowding CR, Riley GP et al. Altered adhesive interactions with marrow stroma of haematopoietic progenitor cells in chronic myeloid leukaemia. Nature 1987;328:342–344. 9. Verfaillie CM, Hurley R, Zhao RC et al. Pathophysiology of CML: do defects in integrin function contribute to the premature circulation and massive expansion of the BCR/ABL positive clone? J Lab Clin Med 1997;129:584–591. 10. Vigneri P, Wang JY. Induction of apoptosis in chronic myelogenous leukemia cells through nuclear entrapment of BCR-ABL tyrosine kinase. Nat Med 2001;7:228–234. 11. Goldman JM, Melo JV. Chronic myeloid leukemia: advances in biology and new approaches to treatment. N Engl J Med 2003;349:1451–1464. 12. Azam M, Latek RR, Daley GQ. Mechanisms of autoinhibition and STI-571/imatinib resistance revealed by mutagenesis of BCR-ABL. Cell 2003;112:831–843. 13. Azam M, Daley GQ. Anticipating clinical resistance to target-directed agents: the BCR-ABL paradigm. Mol Diagn Ther 2006;10:67–76. 14. Thomas ED, Clift RA, Fefer A et al. Marrow transplantation for the treatment of chronic myelogenous leukemia. Ann Intern Med 1986;104:155–163. 15. Guilhot F, Roy L, Guilhot J et al. Interferon therapy in chronic myelogenous leukemia. Hematol Oncol Clin North Am 2004;18:585-603, viii. 16. Guilhot F. Sustained durability of responses plus high rates of cytogenetic responses result in longterm benefit for newly diagnosed chronic-phase chronic myeloid leukemia (CML-CP) treated with imatinib (IM) therapy: update from the IRIS study. (abst #21). Blood 2004;104:10a. 17. Interferon alfa versus chemotherapy for chronic myeloid leukemia: a meta-analysis of seven randomized trials: Chronic Myeloid Leukemia Trialists’ Collaborative Group. J Natl Cancer Inst 1997;89:1616–1620. 18. Druker BJ, Talpaz M, Resta DJ et al. Efficacy and safety of a specific inhibitor of the BCR-ABL tyrosine kinase in chronic myeloid leukemia. N Engl J Med 2001;344:1031–1037. 19. Druker BJ, Sawyers CL, Kantarjian H et al. Activity of a specific inhibitor of the BCR-ABL tyrosine kinase in the blast crisis of chronic myeloid leukemia and acute lymphoblastic leukemia with the Philadelphia chromosome. N Engl J Med 2001;344:1038–1042. 20. Kantarjian H, Sawyers C, Hochhaus A et al. Hematologic and cytogenetic responses to imatinib mesylate in chronic myelogenous leukemia. N Engl J Med 2002;346:645–652. 21. Talpaz M, Silver RT, Druker BJ et al. Imatinib induces durable hematologic and cytogenetic responses in patients with accelerated phase chronic myeloid leukemia: results of a phase 2 study. Blood 2002;99:1928–1937. 22. Sawyers CL, Hochhaus A, Feldman E et al. Imatinib induces hematologic and cytogenetic responses in patients with chronic myelogenous leukemia in myeloid blast crisis: results of a phase II study. Blood 2002;99:3530–3539.

146

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

23. Ottmann OG, Druker BJ, Sawyers CL et al. A phase 2 study of imatinib in patients with relapsed or refractory Philadelphia chromosome-positive acute lymphoid leukemias. Blood 2002;100:1965–1971. 24. O’Brien SG, Guilhot F, Larson RA et al. Imatinib compared with interferon and low-dose cytarabine for newly diagnosed chronic-phase chronic myeloid leukemia. N Engl J Med 2003;348:994–1004. 25. Hughes TP, Kaeda J, Branford S et al. Frequency of major molecular responses to imatinib or interferon alfa plus cytarabine in newly diagnosed chronic myeloid leukemia. N Engl J Med 2003;349:1423–1432. 26. Druker BJ, Guilhot F, O’Brien SG et al. Five-year follow-up of patients receiving imatinib for chronic myeloid leukemia. N Engl J Med 2006;355:2408–2417. 27. Mauro MJ. Defining and managing imatinib resistance. Hematology Am Soc Hematol Educ Program 2006:219–225. 28. Hughes T, Deininger M, Hochhaus A et al. Monitoring CML patients responding to treatment with tyrosine kinase inhibitors: review and recommendations for harmonizing current methodology for detecting BCR-ABL transcripts and kinase domain mutations and for expressing results. Blood 2006;108:28–37. 29. Branford S, Cross NC, Hochhaus A et al. Rationale for the recommendations for harmonizing current methodology for detecting BCR-ABL transcripts in patients with chronic myeloid leukaemia. Leukemia 2006;20:1925–1930. 30. Ross DM, Branford S, Moore S et al. Limited clinical value of regular bone marrow cytogenetic analysis in imatinib-treated chronic phase CML patients monitored by RQ-PCR for BCR-ABL. Leukemia 2006;20:664–670. 31. Baccarani M, Saglio G, Goldman J et al. Evolving concepts in the management of chronic myeloid leukemia: recommendations from an expert panel on behalf of the European Leukemia Net. Blood 2006;108:1809–1820. 32. Mahon FX, Deininger MW, Schultheis B et al. Selection and characterization of BCR-ABL positive cell lines with differential sensitivity to the tyrosine kinase inhibitor STI571: diverse mechanisms of resistance. Blood 2000;96:1070–1079. 33. Gorre ME, Mohammed M, Ellwood K et al. Clinical resistance to STI-571 cancer therapy caused by BCR-ABL gene mutation or amplification. Science 2001;293:876–880. 34. Hochhaus A, Kreil S, Corbin AS et al. Molecular and chromosomal mechanisms of resistance to imatinib (STI571) therapy. Leukemia 2002;16:2190–2196. 35. Gambacorti-Passerini C, Barni R, le Coutre P et al. Role of alpha1 acid glycoprotein in the in vivo resistance of human BCR-ABL(+) leukemic cells to the ABL inhibitor STI571. J Natl Cancer Inst 2000;92:1641–1650. 36. Larghero J, Leguay T, Mourah S et al. Relationship between elevated levels of the alpha 1 acid glycoprotein in chronic myelogenous leukemia in blast crisis and pharmacological resistance to imatinib (Gleevec) in vitro and in vivo. Biochem Pharmacol 2003;66:1907–1913. 37. le Coutre P, Kreuzer KA, Na IK et al. Determination of alpha-1 acid glycoprotein in patients with Ph+ chronic myeloid leukemia during the first 13 weeks of therapy with STI571. Blood Cells Mol Dis 2002;28:75–85. 38. Thomas J, Wang L, Clark RE et al. Active transport of imatinib into and out of cells: implications for drug resistance. Blood 2004;104:3739–3745. 39. White DL, Saunders VA, Dang P et al. OCT-1-mediated influx is a key determinant of the intracellular uptake of imatinib but not nilotinib (AMN107): reduced OCT-1 activity is the cause of low in vitro sensitivity to imatinib. Blood 2006;108:697–704. 40. Shimizu T, Miyakawa Y, Iwata S et al. A novel mechanism for imatinib mesylate (STI571) resistance in CML cell line KT-1: role of TC-PTP in modulating signals downstream from the BCR-ABL fusion protein. Exp Hematol 2004;32:1057–1063. 41. Larson RA, Druker B, Guilhot F et al. Correlation of pharmacokinetic data with cytogenetic and molecular response in newly diagnosed patients with chronic myeloid leukemia in chronic phase (CML-CP) treated with imatinib-An analysis of IRIS study data (Abstract 429). Blood 2006;108:131a.

Chapter 9 / Imatinib Resistance in CML

147

42. Illmer T, Schaich M, Platzbecker U et al. P-glycoprotein-mediated drug efflux is a resistance mechanism of chronic myelogenous leukemia cells to treatment with imatinib mesylate. Leukemia 2004;18:401–408. 43. Wendel HG, de Stanchina E, Cepero E et al. Loss of p53 impedes the antileukemic response to BCRABL inhibition. Proc Natl Acad Sci USA 2006;103:7444–7449. 44. Donato NJ, Wu JY, Stapley J et al. BCR-ABL independence and LYN kinase over-expression in chronic myelogenous leukemia cells selected for resistance to STI571. Blood 2003;101:690–698. 45. Lionberger JM, Wilson MB, Smithgall TE. Transformation of myeloid leukemia cells to cytokine independence by BCR-ABL is suppressed by kinase-defective HCK. J Biol Chem 2000;275: 18581–18585. 46. Warmuth M, Simon N, Mitina O et al. Dual-specific SRC and ABL kinase inhibitors, PP1 and CGP76030, inhibit growth and survival of cells expressing imatinib mesylate-resistant BCR-ABL kinases. Blood 2003;101:664–672. 47. Wilson MB, Schreiner SJ, Choi HJ et al. Selective pyrrolo-pyrimidine inhibitors reveal a necessary role for SRC family kinases in BCR-ABL signal transduction and oncogenesis. Oncogene 2002;21:8075–8088. 48. Ptasznik A, Nakata Y, Kalota A et al. Short interfering RNA (siRNA) targeting the Lyn kinase induces apoptosis in primary, and drug-resistant, BCR-ABL1(+) leukemia cells. Nat Med 2004;10:1187–1189. 49. Donato N, Wu J, Kong LY et al. Constitutive activation of SRC-family kinases in chronic myelogenous leukemia patients resistant to imatinib mesylate in the absence of BCR-ABL mutations: a rationale use of SRC/ABL dual kinase inhibitor-based therapy (Abstract 1087). Blood 2005;106:316a. 50. Hantschel O, Nagar B, Guettler S et al. A myristoyl/phosphotyrosine switch regulates c-Abl. Cell 2003;112:845–857. 51. Nagar B, Hantschel O, Young MA et al. Structural basis for the autoinhibition of c-Abl tyrosine kinase. Cell 2003;112:859–871. 52. Melo JV, Chuah C. Resistance to imatinib mesylate in chronic myeloid leukaemia. Cancer Lett2007;249:121–132. Review. 53. Nardi V, Azam M, Daley GQ. Mechanisms and implications of imatinib resistance mutations in BCRABL. Curr Opin Hematol 2004;11:35–43. 54. Pluk H, Dorey K, Superti-Furga G. Autoinhibition of c-Abl. Cell 2002;108:247–259. 55. Schindler T, Bornmann W, Pellicena P et al. Structural mechanism for STI-571 inhibition of abelson tyrosine kinase. Science 2000;289:1938–1942. 56. Yamamoto M, Kurosu T, Kakihana K et al. The two major imatinib resistance mutations, E255 K and T315I, enhance the activity of BCR/ABL fusion kinase. Biochem Biophys Res Commun 2004;319:1272–1275. 57. Roche-Lestienne C, Soenen-Cornu V, Grardel-Duflos N et al. Several types of mutations of the ABL gene can be found in chronic myeloid leukemia patients resistant to STI571, and they can pre-exist to the onset of treatment. Blood 2002;100:1014–1018. 58. Hofmann WK, Komor M, Wassmann B et al. Presence of the BCR-ABL mutation Glu255Lys prior to STI571 (imatinib) treatment in patients with Ph+ acute lymphoblastic leukemia. Blood 2003;102: 659–661. 59. Shah NP, Nicoll JM, Nagar B et al. Multiple BCR-ABL kinase domain mutations confer polyclonal resistance to the tyrosine kinase inhibitor imatinib (STI571) in chronic phase and blast crisis chronic myeloid leukemia. Cancer Cell 2002;2:117–125. 60. Willis SG, Lange T, Demehri S et al. High-sensitivity detection of BCR-ABL kinase domain mutations in imatinib-naive patients: correlation with clonal cytogenetic evolution but not response to therapy. Blood 2005;106:2128–2137. 61. Griswold IJ, MacPartlin M, Bumm T et al. Kinase domain mutants of BCR-ABL exhibit altered transformation potency, kinase activity, and substrate utilization, irrespective of sensitivity to imatinib. Mol Cell Biol 2006;26:6082–6093. 62. Deininger M. Resistance to imatinib: mechanisms and management. J Natl Compr Canc Netw 2005;3:757–768.

148

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

63. Corbin AS, La Rosee P, Stoffregen EP et al. Several BCR-ABL kinase domain mutants associated with imatinib mesylate resistance remain sensitive to imatinib. Blood 2003;101:4611–4614. 64. Branford S, Rudzki Z, Walsh S et al. Detection of BCR-ABL mutations in patients with CML treated with imatinib is virtually always accompanied by clinical resistance, and mutations in the ATP phosphate-binding loop (P-loop) are associated with a poor prognosis. Blood 2003;102:276–283. 65. Soverini S, Martinelli G, Rosti G et al. ABL mutations in late chronic phase chronic myeloid leukemia patients with up-front cytogenetic resistance to imatinib are associated with a greater likelihood of progression to blast crisis and shorter survival: a study by the GIMEMA Working Party on Chronic Myeloid Leukemia. J Clin Oncol 2005;23:4100–4109. 66. Nicolini FE, Corm S, Le QH et al. Mutation status and clinical outcome of 89 imatinib mesylate– resistant chronic myelogenous leukemia patients: a retrospective analysis from the French intergroup of CML (Fi(phi)-LMC GROUP). Leukemia 2006;20:1061–1066. 67. Jabbour E, Kantarjian H, Jones D et al. Frequency and clinical significance of BCR-ABL mutations in patients with chronic myeloid leukemia treated with imatinib mesylate. Leukemia 2006;20: 1767–1773. 68. Hofmann WK, de Vos S, Elashoff D et al. Relation between resistance of Philadelphia-chromosomepositive acute lymphoblastic leukaemia to the tyrosine kinase inhibitor STI571 and gene-expression profiles: a gene-expression study. Lancet 2002;359:481–486. 69. McLean LA, Gathmann I, Capdeville R et al. Pharmacogenomic analysis of cytogenetic response in chronic myeloid leukemia patients treated with imatinib. Clin Cancer Res 2004;10:155–165. 70. Villuendas R, Steegmann JL, Pollan M et al. Identification of genes involved in imatinib resistance in CML: a gene-expression profiling approach. Leukemia 2006;20:1047–1054. 71. Chung YJ, Kim TM, Kim DW et al. Gene expression signatures associated with the resistance to imatinib. Leukemia 2006;20:1542–1550. 72. Frank O, Brors B, Fabarius A et al. Gene expression signature of primary imatinib-resistant chronic myeloid leukemia patients. Leukemia 2006;20:1400–1407. 73. Soverini S, Martinelli G, Amabile M et al. Denaturing-HPLC-based assay for detection of ABL mutations in chronic myeloid leukemia patients resistant to Imatinib. Clin Chem 2004;50:1205–1213. 74. Jacobberger JW, Sramkoski RM, Frisa PS et al. Immunoreactivity of Stat5 phosphorylated on tyrosine as a cell-based measure of BCR-ABL kinase activity. Cytometry A 2003;54:75–88. 75. Liu WH, Makrigiorgos GM. Sensitive and quantitative detection of mutations associated with clinical resistance to STI-571. Leuk Res 2003;27:979–982. 76. Kreuzer KA, Le Coutre P, Landt O et al. Pre-existence and evolution of imatinib mesylate–resistant clones in chronic myelogenous leukemia detected by a PNA-based PCR clamping technique. Ann Hematol 2003;82:284–289. 77. Cazzaniga G, Corradi B, Piazza R et al. Highly sensitive mutations detection in BCR-ABL positive leukemia prior and during imatinib treatment (Abstract 1985). Blood 2004;104:548a. 78. Branford S, Rudzki Z, Parkinson I et al. Real-time quantitative PCR analysis can be used as a primary screen to identify patients with CML treated with imatinib who have BCR-ABL kinase domain mutations. Blood 2004;104:2926–2932. 79. Wang L, Knight K, Lucas C et al. The role of serial BCR-ABL transcript monitoring in predicting the emergence of BCR-ABL kinase mutations in imatinib-treated patients with chronic myeloid leukemia. Haematologica 2006;91:235–239. 80. Kantarjian HM, Talpaz M, O’Brien S et al. Dose escalation of imatinib mesylate can overcome resistance to standard-dose therapy in patients with chronic myelogenous leukemia. Blood 2003;101: 473–475. 81. Kantarjian HM, Cortes JE, O’Brien S et al. Long-term survival benefit and improved complete cytogenetic and molecular response rates with imatinib mesylate in Philadelphia chromosome– positive chronic-phase chronic myeloid leukemia after failure of interferon-alpha. Blood 2004;104: 1979–1988. 82. Gambacorti-Passerini CB, Rossi F, Verga M et al. Differences between in vivo and in vitro sensitivity to imatinib of Bcr/Abl+ cells obtained from leukemic patients. Blood Cells Mol Dis 2002;28:361–372.

Chapter 9 / Imatinib Resistance in CML

149

83. Liu NS, O’Brien S. Spontaneous reversion from blast to chronic phase after withdrawal of imatinib mesylate in a patient with chronic myelogenous leukemia. Leuk Lymphoma 2002;43:2413–2415. 84. Muller MC, Lahaye T, Hochhaus A. [Resistance to tumor specific therapy with imatinib by clonal selection of mutated cells]. Dtsch Med Wochenschr 2002;127:2205–2207. 85. Hochhaus A, La Rosee P. Imatinib therapy in chronic myelogenous leukemia: strategies to avoid and overcome resistance. Leukemia 2004;18:1321–1331. 86. Dorsey JF, Jove R, Kraker AJ et al. The pyrido[2,3-d]pyrimidine derivative PD180970 inhibits p210Bcr-Abl tyrosine kinase and induces apoptosis of K562 leukemic cells. Cancer Res 2000;60:3127–3131. 87. La Rosee P, Corbin AS, Stoffregen EP et al. Activity of the BCR-ABL kinase inhibitor PD180970 against clinically relevant BCR-ABL isoforms that cause resistance to imatinib mesylate (Gleevec, STI571). Cancer Res 2002;62:7149–7153. 88. Huron DR, Gorre ME, Kraker AJ et al. A novel pyridopyrimidine inhibitor of ABL kinase is a picomolar inhibitor of BCR-ABL-driven K562 cells and is effective against STI571-resistant BCR-ABL mutants. Clin Cancer Res 2003;9:1267–1273. 89. O’Hare T, Pollock R, Stoffregen EP et al. Inhibition of wild-type and mutant BCR-ABL by AP23464, a potent ATP-based oncogenic protein kinase inhibitor: implications for CML. Blood 2004;104: 2532–2539. 90. Deininger MW, Druker BJ. SRCircumventing imatinib resistance. Cancer Cell2004;6:108–110. 91. Shah NP, Tran C, Lee FY et al. Overriding imatinib resistance with a novel ABL kinase inhibitor. Science 2004;305:399–401. 92. Tokarski JS, Newitt JA, Chang CY et al. The structure of Dasatinib (BMS-354825) bound to activated ABL kinase domain elucidates its inhibitory activity against imatinib-resistant ABL mutants. Cancer Res 2006;66:5790–5797. 93. Talpaz M, Shah NP, Kantarjian H et al. Dasatinib in imatinib-resistant Philadelphia chromosome– positive leukemias. N Engl J Med 2006;354:2531–2541. 94. Hochhaus A, Kantarjian HM, Baccarani M et al. Dasatinib induces notable hematologic and cytogenetic responses in chronic-phase chronic myeloid leukemia after failure of imatinib therapy. Blood 2007;109:2303–2309. 95. Cortes J, Guilhot F, Rosti Get al. Dasatinib (SPRYCEL) in patients (pts) with chronic myelogenous leukemia in accelerated phase (AP-CML) that is imatinib-resistant (IM-R) or intolerant (IM-I): updated results of the CA180-005 “START-A” phase II study (Abstract 2160). Blood 2006;108:613a. 96. Cortes J, Rousselot P, Kim DW et al. Dasatinib induces complete hematologic and cytogenetic responses in patients with imatinib-resistant or -intolerant chronic myeloid leukemia in blast crisis. Blood 2007;109:3207–3213. 97. Shah N, Pasquini R, Rousselot P et al. Dasatinib (SPRYCEL) vs. escalated dose of imatinib (IM) in patients (pts) with chronic phase chronic myeloid leukemia (CP-CML) resistant to imatinib: results of the CA180-017 START-R randomized study (Abstract 167). Blood 2006;108:53a. 98. Quintas-Cardama A, Kantarjian H, Jones Det al. Dasatinib (BMS-354825) is active in Philadelphia chromosome–positive chronic myelogenous leukemia after imatinib and nilotinib (AMN107) therapy failure. Blood 2007;109:497–499. 99. Weisberg E, Manley PW, Breitenstein W et al. Characterization of AMN107, a selective inhibitor of native and mutant BCR-ABL. Cancer Cell 2005;7:129–141. 100. O’Hare T, Walters DK, Stoffregen EP et al. In vitro activity of BCR-ABL inhibitors AMN107 and BMS-354825 against clinically relevant imatinib-resistant ABL kinase domain mutants. Cancer Res 2005;65:4500–4505. 101. Kantarjian H, Giles F, Wunderle L et al. Nilotinib in imatinib-resistant CML and Philadelphia chromosome–positive ALL. N Engl J Med 2006;354:2542–2551. 102. le Coutre P, Bhalla K, Giles F et al. A phase II study of nilotinib, a novel tyrosine kinase inhibitor administered to imatinib-resistant and intolerant patients with chronic myelogenous leukemia (CML) in chronic phase (CP) (Abstract 165). Blood 2006;108:53a.

150

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

103. Kantarjian H, Gattermann N, Hochhaus A et al. A phase II study of nilotinib: a novel tyrosine kinase inhibitor administered to imatinib-resistant or intolerant patients with chronic myelogenous leukemia (CML) in accelerated phase (AP) (Abstract 2169). Blood 2006;108:615a. 104. Giles F, le Coutre P, Bhalla K et al. A phase II study of nilotinib, a novel tyrosine kinase inhibitor administered to patients with imatinib resistant or intolerant chronic myelogenous leukemia (CML) in chronic phase (CP), accelerated phase (AP), or blast crisis (BC) who have also failed dasatinib therapy (Abstract 2170). Blood 2006;108:615a. 105. Druker BJ. Circumventing resistance to kinase-inhibitor therapy. N Engl J Med 2006;354:2594–2596. 106. Bradeen HA, Eide CA, O’Hare T et al. Comparison of imatinib mesylate, dasatinib (BMS-354825), and nilotinib (AMN107) in an N-ethyl-N-nitrosourea (ENU)-based mutagenesis screen: high efficacy of drug combinations. Blood 2006;108:2332–2338. 107. Burgess MR, Skaggs BJ, Shah NP et al. Comparative analysis of two clinically active BCR-ABL kinase inhibitors reveals the role of conformation-specific binding in resistance. Proc Natl Acad Sci USA 2005;102:3395–3400. 108. Weisberg EL, Catley L, Wright RD et al. Beneficial effects of combining nilotinib and imatinib in preclinical models of BCR/ABL+ leukemias. Blood 2007;109:2112–2120. 109. O’Hare T, Walters DK, Stoffregen EP et al. Combined ABL inhibitor therapy for minimizing drug resistance in chronic myeloid leukemia: SRC/ABL inhibitors are compatible with imatinib. Clin Cancer Res 2005;11:6987–6993. 110. Copland M, Hamilton A, Elrick LJ et al. Dasatinib (BMS-354825) targets an earlier progenitor population than imatinib in primary CML but does not eliminate the quiescent fraction. Blood 2006;107:4532–4539. 111. Gumireddy K, Baker SJ, Cosenza SC et al. A non-ATP-competitive inhibitor of BCR-ABL overrides imatinib resistance. Proc Natl Acad Sci USA 2005;102:1992–1997. 112. Orsolic N, Golemovic M, Quintas-Cardama A et al. Adaphostin has significant and selective activity against chronic and acute myeloid leukemia cells. Cancer Sci 2006;97:952–960. 113. Yokota A, Kimura S, Masuda S et al. INNO-406, a novel BCR-ABL/Lyn dual tyrosine kinase inhibitor, suppresses the growth of Ph+ leukemia cells in the central nervous system, and cyclosporine A augments its in vivo activity. Blood 2007;109:306-314. 114. Kimura S, Niwa T, Hirabayashi K et al. Development of NS-187, a potent and selective dual BCRABL/LYN tyrosine kinase inhibitor. Cancer Chemother Pharmacol 2006;58 Suppl 7:55–61. 115. Golas JM, Arndt K, Etienne C et al. SKI-606, a 4-anilino-3-quinolinecarbonitrile dual inhibitor of SRC and ABL kinases, is a potent antiproliferative agent against chronic myelogenous leukemia cells in culture and causes regression of K562 xenografts in nude mice. Cancer Res 2003;63:375–381. 116. Young MA, Shah NP, Chao LH et al. Structure of the kinase domain of an imatinib-resistant ABL mutant in complex with the Aurora kinase inhibitor VX-680. Cancer Res 2006;66:1007–1014. 117. Giles FJ, Cortes J, Jones D et al. MK-0457, a novel kinase inhibitor, is active in patients with chronic myeloid leukemia or acute lymphocytic leukemia with the T315I BCR-ABL mutation. Blood 2007;109:500–502. 118. Tauchi T, Akahane D, Nunoda K et al. Activity of a novel aurora kinase inhibitor, VE-465, against T315i mutant form of BCR-ABL: in vitro and in vivo studies (Abstract 1358). Blood 2006;108:396a. 119. Walz C, Sattler M. Novel targeted therapies to overcome imatinib mesylate resistance in chronic myeloid leukemia (CML). Crit Rev Oncol Hematol 2006;57:145–164. 120. Cowan-Jacob SW, Fendrich G, Floersheimer A et al. Structural biology contributions to the discovery of drugs to treat chronic myelogenous leukaemia. Acta Crystallogr D Biol Crystallogr 2007;63(Pt 1):80–93. 121. Van Etten RA. Mechanisms of transformation by the BCR-ABL oncogene: new perspectives in the post-imatinib era. Leuk Res 2004;28 Suppl 1:S21–28. 122. Deininger MW. Basic science going clinical: molecularly targeted therapy of chronic myelogenous leukemia. J Cancer Res Clin Oncol 2004;130:59–72. 123. Kantarjian HM, Talpaz M, Giles F et al. New insights into the pathophysiology of chronic myeloid leukemia and imatinib resistance. Ann Intern Med 2006;145:913–923.

10

Role of Thymidylate Synthase Gene Variations in Colorectal Cancer Patients Georg Lurje, MD, and Heinz-Josef Lenz, MD CONTENTS Introduction Molecular Biology of 5-FU Metabolis m Current Standard of Care for Metas tatic Colorectal Cancer Pharmacogenomics and Prognos tic and Predictive Markers Thymidylate Synthas e Polymorphis ms TS Polymorphis ms as Prognos tic and Predictive Factors Other Polymorphis ms of the 5-FU Metabolis m Conclus ion References

S UMMARY Colorectal cancer (CRC) is the third most common cause of cancer-related death in women and men in the United States. The current therapeutic options for patients with metastatic CRC (mCRC) are 5-fluorouracil (5-FU) based chemotherapy regimens with the addition of irinotecan (CPT-11) or oxaliplatin. It still remains a challenge for oncologists to evaluate the reasons for a wide variation in response and toxicity among patients undergoing systemic 5-FU based chemotherapy. Pharmacogenomics From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 10, 

151

152

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

has emerged as a useful tool to address the variations by evaluating the interplay between genotype and drug efficacy. Identifying a reliable panel of prognostic and predictive markers is critical in selecting an individualized chemotherapy regimen based on a particular tumor genotype. A substantial body of evidence has been accumulated in recent years, demonstrating that the level of thymidylate synthase (TS) mRNA and protein expression significantly correlates with sensitivity and resistance to TS-targeted 5-FU based chemotherapy regimens. However, the cause of the variability in TS expression still remains unclear, even though several molecular mechanisms have been identified that seem to regulate the expression of TS, which may have an impact on the response to 5-FU based chemotherapy. TS gene expression is associated with the presence of polymorphic tandem repeats (double or triple) in the 5 -UTR region (thymidylate synthase enhancer region, TSER) of the TS gene (1,2). Patients with colon cancer who are homozygous for the triple tandem repeat (TSER-3R/3R) had significantly higher levels of intratumoral TS compared with those with double tandem repeats (3). Furthermore, Mandola et al. identified a G to C single nucleotide polymorphism (SNP) within the second repeat of the 5 -UTR TSER, which may be responsible for the transcriptional up-regulation of TS. A third polymorphic change has been reported in the 3 UTR region of TS at position 1494, a 6bp repeat. These three different polymorphisms in the same gene obviously complicate efforts aimed at understanding the effects of each individual polymorphism. TS expression levels, tumor response, and toxicity are functions of multiple TS gene alterations, rather than the result of one single polymorphism. This review will provide an update of the most recent data on 5-FU metabolism and TS gene variations in CRC. Key Words: Thymidylate synthase; pharmacogenomics; polymorphism; chemotherapy; TSER; SNP; TS 1494–6bp/–6bp

1. INTRODUCTION Colorectal cancer (CRC) is the third most common cause of cancer-related death in women and men in the United States ( 4). Although surgical resection is the primary therapy for CRC, prognosis remains poor and not all patients are candidates for surgery. The current therapeutic options for patients with metastatic CRC (mCRC) are 5-fluorouracil (5-FU) based therapy regimens in combination with irinotecan (CPT-11) or oxaliplatin (5,6). While these traditional chemotherapeutic options have improved the median overall survival from 12 to 18–21 months, limitations with this treatment remain ( 7, 8). Poor response rates (RR) and sometimes life-threatening toxicity appear to be the most limiting factors in 5-FU based chemotherapy regimens (6,7,8). In recent years a number of new drugs and drug combinations have been evaluated for safety and efficacy in patients with metastatic CRC. Targeted agents such as cetuximab (monoclonal antibody against the epidermal growth factor receptor) and bevacizumab (monoclonal antibody against the vascular endothelial growth factor receptor) have significantly increased efficacy of chemotherapeutic regimens (9,10).

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

153

Despite these advances, it remains a challenge for oncologists to understand the wide variation in response and toxicity among patients undergoing 5-FU based chemotherapy. For example, one patient may have a significant response to a chemotherapeutic regimen while another may have progression of disease. The same dose and administration schedule of a drug may result in severe dose-limiting toxicity in one patient and not in another of the same age and sex. Pharmacogenomics has emerged as a useful tool to address the variations by evaluating the interplay between genotype and drug efficacy and toxicity (11,12). The possibility of individualizing cancer treatment is gaining wide acceptance. Numerous germline polymorphisms that influence enzyme function or expression, which may predict clinical outcome and toxicity to chemotherapy, have been identified. Preclinical and clinical studies have shown that gene variations in key enzymes of the 5-FU metabolism, such as thymidylate synthase (TS), thymidine phosphorylase (TP), and dihydopyrimidine dehydrogenase (DPD) are most likely responsible for variations in response, toxicity, and overall survival in 5-FU based chemotherapy regimens (Tables 2a and 2b) (13,14). The goal of this review is to provide an update of the most recent data on 5-FU metabolism and TS gene variations in CRC.

2. MOLECULAR BIOLOGY OF 5-FU METABOLISM Since its introduction over 50 years ago by Heidelberger et al. (15), 5-FU has been the drug of choice for systemic chemotherapy regimens in patients with CRC, although optimal administration schedules have not yet been fully established. 5-Fluorouracil (5-FU) or its prodrug capecitabine play a central role in the treatment of CRC. Thymidine phosphorylase (TP) and thymidine kinase (TK) produce fluorodeoxyuridine monophosphate (FdUMP), which forms a stable ternary complex with thymidylate synthase (TS), the sole de novo source of thymidine in the cell. Inhibition of TS rapidly shuts off DNA synthesis and triggers apoptosis and other cell death processes (16,17). The mechanisms of action of 5-FU include inhibiting DNA synthesis by incorporating as fluorodeoxyuridine triphosphate (FdUTP) and inhibiting RNA synthesis by incorporating as fluorouridine triphosphate (FUTP) leading to transcriptional errors and arrest (Fig. 1) (18). This mechanism provides not much selectivity for cancer cells, and thus TS-inhibitors also cause considerable toxicity as a side effect of 5-FU based chemotherapy regimens. Dihydropyrimidine dehydrogenase (DPD) is a key enzyme in 5-FU metabolism and responsible for inactivation and elimination of 5-FU in the liver (> 80%) (Fig. 1). Orotate phosphoribosyltransferase (OPRT) catalyses the conversion of 5-FU to fluorouridine monophosphate (FUMP), which is subsequently phosphorylated to activated fluorouridine triphosphate (FUTP). FUTP incorporates into RNA and thereby compromises RNA processing and function (Fig. 1). Methylentetrahydrofolate reductase (MTHFR) is another key enzyme of 5-FU metabolism, alternating 5-FU sensitivity indirectly by folate pool variations. MTHFR plays an important role in the action of 5-FU, an inhibitor of TS, by converting 5,10methylenetetrahydrofolate, a substrate of TS, to 5-methyltetrahydrofolate (19).

154

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Fig. 1. 5-FU metabolism. Abbreviations : 5-FU (5-fluorouracil); FdUMP (fluorodeoxyuridine monophosphate); TS (thymidylate synthase); FUMP (fluorouridine monophosphate); DPD (dihydropyrimidine dehydrogenase); FUTP (fluorouridine triphosphate); FU H2 (dihydrofluorouracil); FBAL (fluoro-␤-alanine).

3. CURRENT STANDARD OF CARE FOR METASTATIC COLORECTAL CANCER Until the year 2000, the only systemic therapy for metastatic CRC was a bolus or infusional 5-FU regimen, mostly in combination with leucovorin (5-FU/LV), achieving a median survival of 12 month (20). However, over the last 6 years many new agents have been developed and approved by the Federal Drug and Food Administration (FDA) for the treatment of mCRC (7). A summary of the current guidelines for treating colorectal cancer is shown in Table 1. Irinotecan (CPT-11), a prodrug of SN 38, achieves its anti-carcinogenic effect through inhibiting topoisomerase I, an enzyme involved in the relaxation of supercoiled DNA (21). Oxaliplatin is a platinum analog that inhibits DNA synthesis through the formation of intra-strand DNA adducts and is currently approved by the FDA for use both in adjuvant and metastatic settings (22). CPT-11 and oxaliplatin, when combined with 5-FU/LV based chemotherapy regimens, have boosted response rates (RR) and overall-survival (OS) to 50% and at least 16 months, respectively, compared to 20% and 10–12 month with 5-FU/LV mono-chemotherapy (23).

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

155

Table 1 Current Standard of Care for Colorectal Cancer (CRC) Early Stage

Adjuvant

Neoadjuvant

Metastatic

Resectable: Surgery and surveillance Adjuvant therapy Unresectable: Neoadjuvant therapy plus surgery

Capecitabine or FOLFOX

FOLFOX or FOLFIRI with or without bevacizumab

Bevacizumab Irinotecan or plus: FOLFOX irinotecan or FOLFIRI or plus CapeOx cetuximab, FOLFOX, or FOLFIRI

Rectal: Radiation with 5-FU or capecitabine

Second Line

From Ref. (85). Abbreviations : 5-FU/LV (5-fluorouracil plus leucovorin); CapeOx (capecitabine plus oxaliplatin); FOLFIRI (infusional 5-FU/LV plus irinotecan); FOLFOX (infusional 5-FU/LV plus oxaliplatin).

3.1. Administration Schedules Enormous efforts have been made recently to improve efficacy of 5-FU treatment, either by changing the infusion schedule or changing its biochemical modulation in combination with novel cytotoxic agents. In view of the short plasma half-life of 5-FU, most authors believe that continuous administration of 5-FU (CIFU) is the superior 5-FU schedule compared to bolus regimens. Continuous infusion of 5-FU seems to inhibit predominantly DNA synthesis, whereas bolus administration of 5-FU inhibits RNA splicing and DNA synthesis, resulting in different toxicity and efficacy profiles (24,25). Several randomized studies comparing CIFU versus bolus 5-FU with or without leucovorin (LV) suggest the superiority of CIFU over bolus 5-FU in terms of improved RR and reduced toxicity profile ( 26, 27, 28). However, no improvements in survival were noted ( 26, 27). A meta-analysis from seven randomized studies with a total of 1,219 patients confirmed superior response of CIFU over bolus regimens (22% vs. 14%, p = 0.0002), with statistically significant superior OS in patients treated with CIFU compared with those patients who received a bolus 5-FU regimen (12.4 months vs. 11.4 months p = 0.04) (28).

3.2. 5-FU Analogs and Prodrugs A number of fluoropyrimidines other than 5-FU have been synthesized, most of which R ) is an orally administered fluoropyrimact as prodrugs for 5-FU. Capecitabine (Xeloda idine that is converted to 5-FU by the enzyme thymidine phosphorylase (TP), which is often over-expressed in malignant compared to normal tissues ( 29). Capecitabine has been shown to be equivalent to 5-FU/LV and was the first oral agent to be approved by the FDA for the treatment of mCRC (30).

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

156

Another prodrug, ftorafur (FT/1-tetrahydrofuranyl-5-fluorouracil) is metabolized to 5-FU either by hepatic P-450 microsomal enzymes or ubiquitous cytosolic enzymes ( 31). The bioavailability is improved by co-administration of uracil in a 4:1 molar ratio (“UFT”), which blocks the degradation of DPD and therefore leads to a more prolonged concentration of 5-FU in tumor tissues ( 32). Another 5-FU prodrug is S-1, a fourth-generation oral fluoropyrimidine that is a formulation of tegafur (FT) and two modulators: 5-chloro-2,4-dihydroxypyridine (CDHP), and potassium oxonate (Oxo), causing decreased drug incorporation into cellular RNA. FT is a prodrug of cytotoxic fluorouracil (FU), and CDHP prevents its degradation by inactivation of DPD (200 times more potent than uracil), which allows its higher concentrations to enter the anabolic pathways (33). In animal models, Oxo is protective against FT-induced diarrhea, primarily reducing intestinal phosphorylation of FU by inhibiting OPRT ( 34, 35). Thus, one component of S-1, CDHP, reduces the degradation of cytotoxic FU, thereby prolonging its half-life (36). Oxo, another component, potentially reduces gastrointestinal (GI) toxicity in humans. A number of studies have investigated S-1 in patients with metastatic gastric cancer and reported response rates ranging from 26−45% ( 37, 38, 39). One Japanese study reported a 76% RR in patients with gastric cancer treated with S-1 and cisplatin combination chemotherapy (40). Interestingly, the maximum tolerable dose (MTD) of S-1 is higher in Japanese patients compared with Americans. Ajani et al. showed that the dose of S-1 tolerated by Western patients is lower than the dose tolerated by Japanese patients, with a MTD of 50 mg/m2/d and 80 mg/m2/d, respectively (41). One possible explanation for the differences in MTD is the significant discrepancy in frequencies of polymorphisms in the TS gene reported in Asians compared with Caucasians (42).

4. PHARMACOGENOMICS AND PROGNOSTIC AND PREDICTIVE MARKERS Pharmacogenomics has emerged as a useful tool to address the inter-individual and intratumoral gene variations by analyzing the interplay of genotype and drug-efficacy and toxicity (12). Having a reliable panel of prognostic and predictive markers will be critical in selecting an individualized and tailored chemotherapy regimen based on the particular tumor and host genotype (Table 2a). Table 2a Gene Expression Levels Associated with Outcome (5-FU Chemotherapy) Gene

Factor

Clinical Significance

TS DPD TP

mRNA expression mRNA expression mRNA expression

Prediction of RR and OS in first line and second line Prediction of RR and OS in mCRC Prediction of RR and OS in mCRC

Abbreviations : TS (thymidylate synthase); DPD (dihydropyrimidine dehydrogenase); TP (thymidine phosphorylase); RR (response rate); OS (overall survival); mCRC (metastatic colorectal cancer).

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

157

5-FU and its metabolites achieve their cytotoxicity by inhibiting TS, which is the sole de novo source of thymidine in the cell (43,44). Inhibition of TS rapidly shuts off DNA synthesis and triggers apoptosis and other cell death mechanisms ( 16). A substantial body of evidence has been accumulated over the years, demonstrating that the level of TS, TP, and DPD mRNA and protein expression highly correlates with sensitivity and resistance to TS-targeted 5-FU based chemotherapy regimens (45,46). Leichmann et al. were the first to demonstrate the significant inverse relationship of intratumoral TS gene expression and response to 5-FU based chemotherapy (47). In a retrospective study, 42 tissue samples of patients with CRC were analyzed for TS mRNA gene expression using quantitative real time RT-PCR. Patients with low TS gene expression levels had a significant higher response rate and showed a superior median survival of 13.6 months, compared to 8.2 months in patients whose tumors had an increased TSlevel (p = 0.02) (3,14,48,49). Based on these results, further clinical investigation showed that TS polymorphisms predicted response and survival (50,51). Most studies have consistently agreed that both, TS mRNA and TS protein expression do vary considerably among tumors and that the respond rate of various tumors towards 5-FU based chemotherapy regimens are related to intratumoral TS mRNA and protein expression. Higher TS expression levels are generally associated with lower response rates and shorter overall survival (3,52,53). However, the cause of the variability in TS expression remains unclear, even though several molecular mechanisms seem to regulate the expression of TS (Fig. 2), some of which have been found to have an impact on the probability of response to 5-FU based chemotherapy.

Fig. 2. Regulation of thymidylate synthase (TS) gene expression. Abbreviations : TS (thymidylate synthase); USF-1 (upstream stimulating factor 1); USF-2 (upstream stimulating factor 2); SNP (single nucleotide polymorphism). (Reprinted with permission of El-Khoueiry et al. Pharmacogenomics and molecular biology of gastrointestinal cancers, Atlas of Gastrointestinal Cancers, Current Medicine Group, LLC, Philadelphia, 2007).

158

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

5. THYMIDYLATE SYNTHASE POLYMORPHISMS 5.1. Thymidylate Synthase Promoter Enhancer Region (TSER) Polymorphism Although the intracellular TS mRNA and TS protein levels seem to be crucial for 5-FU based chemotherapy regimens, little is known about the mechanism by which TS expression is regulated ( 14, 45). TS expression seems to be regulated by a highly polymorphic tandem repeat in the TS promoter enhancer region (TSER). Horie et al. were the first to describe a 28-bp sequence at the 5 UTR region of the TS gene to be polymorphic in either double (2R) or triple tandem repeats (3R) (1) (Figs. 2 and 3). The authors demonstrated that the TS gene contains either homozygous double tandem repeats (2R/2R), homozygous triple tandem repeats (3R/3R), or heterozygous tandem repeats (2R/3R), and that the frequencies of these alleles are significantly different among ethnic populations (1,54,55). Even though allele frequency is similar among Caucasians and southwest Asians, homozygous triple repeat subjects are nearly twice as common in Chinese subjects (67%) compared with Caucasian subjects (38%) (55). Pullarkat et al. showed for the first time that the triple repeat yields four-fold (9.42) higher TS mRNA levels in tumor-tissue obtained from patients with mCRC in comparison with patients who carry the 2R variant (2.60, p < 0.004) (3). This polymorphism

Fig. 3. Polymorphisms in the TS gene. Abbreviations : TSER (thymidylate synthase enhancer region); 2R (two-tandem repeats); 3R (three-tandem repeats); TSER 3G (single nucleotide polymorphism of TSER 3R); -6bp (TS 3 -UTR 6bp deletion polymorphism). (Adapted to Marsh et al. Thymidylate synthase pharmacogenetics, Invest New Drugs 2005).

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

159

is of clinical significance because greater in vitro enzyme activity occurs with the triple repeat than is observed for the double repeat (1,56). Patients with metastatic colon cancer homozygous for the triple tandem repeat (TSER-3R/3R) had significantly higher intratumoral TS gene expression compared with those with double tandem repeats (TSER-2R/2R) within the 5 -UTR region (3,57). In another study of 70 patients with colorectal adenocarcinoma, total TS mRNA concentration was higher in patients with homozygous triple tandem repeats (3R/3R), compared with those individuals with heterozygous tandem repeats (3R/2R) or homozygous double tandem repeats (2R/2R), although these data did not reach statistical significance (57). Clinical analysis of the TSER polymorphism among chemo-naive patients with mCRC who received 5-FU based chemotherapy showed that 5 UTR variations in the TS gene differentiate chemotherapy responders from nonresponders (3). There are some data to suggest that patients homozygous for the 2R allele have better overall response rates to 5-FU than patients with the homozygous 3R variant (50% vs. 9%). A recent study by Etienne et al. confirmed this data by demonstrating that 2R TS genotype patients with CRC receiving 5-FU based chemotherapy have a more favorable prognosis for survival (58). Interestingly, the probability of downstaging after neoadjuvant radiochemotherpy in patients with advanced rectal cancer is correlated with the TSER genotype. As demonstrated by Villafranca et al., downstaging effect, observed in patients homozygous for TSER double repeats (2R/2R) and heterozygous for double repeats (2R/3R), was 3.7 times higher compared with triple repeat homozygotes (3R/3R) (48). Thus, 22% of the patients with the 3R/3R genotype experienced a downstaging effect, compared with 60% in the 2R/3R and 2R/2R group (p = 0.002) (48). Patients harboring the 2R allele showed a significantly longer progression-free survival (PFS) compared with patients of the homozygous 3R/3R group (41% vs. 81%, p = 0.17), even though no significant influence on OS could be observed (48).

5.2. TSER 3R G to C Single Nucleotide Polymorphism (SNP) In a recent study, Mandola et al. showed that the 28 bp TSER tandem repeats contain elements that bind upstream stimulating factor (USF), and also that ligand binding by USF-1 and USF-2 enhances the transcriptional activity of the TS gene (Fig. 2) ( 42). Electrophoretic mobility shift analysis has shown that the presence of a G-to-C single nucleotide polymorphism (SNP) within the second repeat of the 3R allele leads to decreased ability of upstream stimulatory factor (USF) to bind within the repeat and therefore sequentially result in decreased transcriptional activity of the 3R TS gene variant (42). The authors demonstrated that these polymorphisms alter mRNA stability and therefore enzyme activity. They showed that whereas phosphorylated USF-1 bound the normal consensus sequence, the G-to-C substitution abolished the binding (42,59). In vitro transcription analysis showed that the TSER 3RC allele caused a lower transcription rate than the TSER 3RG variant, which was comparable to the TSER 2R genotype. Interestingly, the frequency of the 3RC allele among all 3R alleles showed a variation of 56%, 47%, 28%, and 37% for Whites, Hispanics, African-Americans, and Singapore

160

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

and Chinese, respectively. Although the overall frequency was similar to that reported by Mandola et al., Japanese females were noted to have lower frequency of the 3RG allele than males (60).

5.3. Thymidylate Synthase 3 -UTR 6bp Deletion A third polymorphism of the TS gene is a 6bp deletion in the 3 -UTR region of the gene (Fig. 3). This polymorphism was identified by aligning expressed sequence tag (EST) databases ( 61). The deletion occurs at an allele frequency of 27%–29% in Caucasians (+6bp/+6bp, 48%; +6bp/–6bp, 44%; –6bp/–6bp, 7%) (61,62). The authors proposed that alterations in the 3 -UTR region of the TS gene could alter RNA stability and therefore influence TS mRNA and TS protein expression (61). Preliminary data from 43 patients analyzed for TS mRNA expression and 3 -UTR 6bp deletion suggest that patients homozygous for 6bp deletion (–6bp/–6bp) express threefold less TS mRNA than patients homozygous for the presence of the 6bp (+6bp/+6bp) (p = 0.017) ( 42, 59, 62). A recent study by Dotor et al. illustrated that in series of homogenously 5-FU treated patients, the presence of homozygous 3 -UTR 6bp deletion (–6bp/–6bp) appears to be a strong prognostic factor, which may be of benefit for at least 20% of the study population (63).

5.4. Loss of Heterozygosity (LOH) A potential advantage of using germline polymorphisms as predictive markers of tumor response is because it can be assayed in normal tissues without the necessity of obtaining tumor biopsy samples. However, possible genetic alteration at the tumorous TS locus may complicate the use of genotyping of TS polymorphisms in noncancerous tissue for prediction of tumor response. The TS gene is localized to the short arm of chromosome 18 at chromosome band 18p11.32 ( 64). Chromosome 18 is generally known to be a site of frequent deletions in colorectal cancer tissues (64). Therefore it is highly probably that allelic imbalance occurs at the TS locus in some colorectal tumors. Zinzindohoue et al. were the first to report on the idea of LOH at the TS locus. The authors showed that the TS genotype from 2R/3R heterozygotes differed in ratio between 2R and the 3R bands. The observed LOH frequency at the TS locus was 63% (31 of 50) (65). If LOH occurs at the TS locus, the tumor genotype of homozygous patients will be the same as that in normal tissue, but in the case of heterozygous 2R/3R individuals, the tumor will have either a 2R/loss or 3R/loss genotype. According to clinical outcome, 2R/3R heterozygous patients with a 2R/loss genotype in their tumor will have significantly better RRs and OS than patients with the 3R/loss genotype ( 66). These data additionally illustrate, that 3R negatively influences tumor response, whereas 2R having a positive effect on clinical outcome ( 66). Uchida et al. demonstrated that the TS genotype is modulated by LOH. Heterozygous 2R/3R tumor genotype had shorter OS, similar to those of 3R/3R patients, while deletion of the 3R allele resulted in a 2R/loss tumor genotype with high RRs and OS similar to homozygous 2R/2R genotypes (66).

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

161

6. TS POLYMORPHISMS AS PROGNOSTIC AND PREDICTIVE FACTORS A number of studies have demonstrated that TS expression is associated with polymorphic tandem repeats in the 5 -UTR region of the TS gene (3,42,67). A recent study by Dotor et al. showed that patients undergoing adjuvant chemotherapy showed significantly better OS (p = 0.02) when carrying homozygous TSER-3R compared with those patients homozygous or heterozygous for TSER-2R (2R/2R, 2R/3R). Patients with 3R/–6bp haplotype, which represented 22.8% of the study population analyzed, showed a benefit (p = 0.012) to adjuvant chemotherapy. The strength of this association was stronger for haplotypes harboring the TS 1494–6bp/–6bp allele, suggesting a prominent role of the 3 -UTR polymorphism. These results are in line with observations by Jakobsen et al. for patients in the metastatic setting (68). A different study by Hitre et al. verified better OS (p = 0.009) and PFS (p = 0.048) for patients receiving adjuvant 5-FU based chemotherapy with the 3R variant compared to other genotype combinations (69). A prospective study by Etienne et al. (58) among 103 colorectal cancer patients who received 5-FU based adjuvant chemotherapy identified the 2R/3R TS genotype as the most favorable for survival (58). The authors clearly demonstrated that the TS genotype does not necessarily correlate with TS activity, because highest TS activity was observed with 2R/3R heterozygotes and not as previously described with 3R/3R homozygotes (1,2,3). It was suggested that polymorphisms other than the 5 -UTR tandem repeat have an additional impact on translational and post-transcriptional regulation of the TS gene (58). However, in the adjuvant setting, the majority of reports associate homozygous triple repeats (TSER 3R/3R) with poor response and overall survival to colorectal tumors (Table 2b) (2,3,48,49). Additional studies are needed to identify the regulatory factors by which the 5 -UTR polymorphism alters TS expression, which might help to interpret these conflicting data. Mandola et al. shed further light on our understanding of TS expression and regulation by identifying a new G-to-C SNP within the 28 bp repeat Table 2b TS, DPD, MTHFR Polymorphisms Associated with Outcome and Toxicity Protein

Polymorphism

28-bp tandem repeat (5 -UTR) TS 1494+6bp/+6bp (3 -UTR) TS 1494–6bp/–6bp (3 -UTR) Exon 14 skipping mutation DPD MTHFR C677T transition TS

Function

Therapy

Clinical Significance

TS-activity ↑ TS-activity ↑

5-FU 5-FU

Response/toxicity ↓ Response/toxicity ↓

TS-activity ↓

5-FU

Response/toxicity ↑

DPD-activity ↓ Folate Imbalance

5-FU 5-FU, MTX

Toxicity ↑ Response/toxicity ↑

Abbreviations : 5-FU (5-fluorouracil); MTX (mMethotrexate); TS (thymidylate synthase); DPD (dihydropyrimidine dehydrogenase); MTHFR (methylenetetrahydrofolate reductase); TS 1494–6bp/–6bp (thymidylate synthase 3 -UTR 6bp deletion polymorphism).

162

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

polymorphism that solely appears within the 3R variant of the TS gene and disrupts a binding site for the transcriptions factor USF-1 (42). It has been suggested that this recently discovered SNP within the 5 -UTR 28 bp region of the TS gene may account for some of the discrepant study results. The combination of SNP and 5 -UTR TSER polymorphism allows the definition of three TS alleles: *2, *3 G, and *3C. Marcuello et al. performed TS genotyping on 89 mCRC patients who received palliative 5-FU based chemotherapy. The clinical outcome was evaluated according to the genotype (high expression genotype: *2R/*3 G; *3C/*3 G; *3 G/*3 G and low expression genotype: *2R/*2R; *2R/*3C; *3C/*3C). A higher overall RR (p = 0.04), PFS (p = 0.07) and OS (p = 0.03) was observed in the group of patients with a low expression genotype (70). A recent case-control study by Kawakami et al. showed similar results for patients with gastric adenocarcinoma. Ninety high-risk patients were treated with adjuvant 5-FU based chemotherapy after they had undergone R0 surgical resection. The clinical outcome was evaluated according to the genotype (high expression genotype 5 -UTR-*2R/*3 G; -*3C/*3 G; -*3 G/*3 G, 3 -UTR +6bp/+6b and low expression genotype 5 -UTR-*2R/*2R; -*2R/*3C; -*3C/*3C, 3 -UTR –6bp/–6bp, -+6bp/–6bp). Patients in the low-expression group had significantly better OS (p = 0.001) and PFS (p = 0.003) compared with patients in the high-expression group or patients at least carrying one high-expression allele (71). In an abstract from the 2006 ASCO meeting, Tan et al. reported that it seems feasible to direct neoadjuvant radiochemotherapy by using TS genotyping ( 72). The authors investigated two arms of treatment for patients with advanced rectal cancer. One arm, the “good risk” group (2R/2R, 2R/3R, 2R/4R) received standard 5-FU based radiochemotherapy, whereas the second arm, the “bad risk” group (3R/3R, 3R/4R), received 5-FU based radiochemotherapy with the addition of irinotecan (CPT-11). Preliminary data from this study showed that co-administration of CPT-11 in the 3Rgroup boosted downstaging (DS) and pathologic complete response (pCR) to 73.7% (DS) and 52.6% (pCR), respectively. Reported efficacy for the “good risk” group was similar to those predicted in prior studies, with 59.8% DS and 19.6% pCR, respectively (72). The investigators suggested that using TS genotyping prior to neoadjuvant radiochemotherapy might help direct patients to a more optimized and individualized chemotherapy schedule (72). Although most studies have reported poorer OS and PFS with tumors expressing high levels of TS mRNA and TS protein, estimates of the prognostic value of TS expression between those studies have differed widely. A comprehensive study by Popat et al. reviewed 20 recently published studies by a standard multi-analysis technique ( 73). In summary, the investigators concluded that tumors expressing high levels of TS appeared to have a poorer OS compared with tumors expressing low levels (73). To try to account for the discrepant and sometimes contradictory data in the literature, one can consider many factors such as small study sample number, lack of standardized methodologies for measuring protein and gene expression, suboptimal samples consisting of different mixtures of cells, tissue-specific differences, and study populations with different allele distributions. For example, measuring higher levels of TS mRNA or TS protein can be ascribed to the usage of laser-captured microdissection, purifying tumor cells from adjacent stroma cells, whereas mRNA analysis of non-microdissected tumor tissue can lead to false positive results.

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

163

Beside these technical issues, TS expression and its regulation by various types of TS gene polymorphisms seem to be crucial in predicting response and overall clinical outcome. The possibility of three different polymorphisms in the same gene obviously complicates efforts aimed at understanding the functional significance of each individual polymorphism. In the case of TS, there are 18 different allele combinations possible, all of which may theoretically influence clinical outcome. Thus, it is likely that the observed TS expression levels, tumor response, and toxicity may be complicated functions of multiple TS gene alterations, rather than the result of one single polymorphism. A summary of these studies is presented in Table 3.

7. OTHER POLYMORPHISMS OF THE 5-FU METABOLISM DPD represents another key enzyme of the 5-FU metabolism. This rate-limiting enzyme deactivates >80% of 5-FU in the liver ( 74). Retrospective analyses of gene expression has revealed that low DPD levels are associated with superior response to fluoropyrimidines-based chemotherapy regimens (14,74). It has been suggested that low levels of DPD may increase bioavailability of the drug, thereby improving response and toxicity (75,76,77). Several different mutations of the DPD gene have been identified to be associated with a decreased activity of the enzyme. Consequently, this decreased enzyme activity leads to accumulation of active 5-FU metabolites, which may be the cause for increased heamatopoetic and gastrointestinal toxicity (74,75, 76,77). A G-to-A substitution in the invariant GT splice donor site flanking exon 14 (IVS14+1 G>A) was reported for Caucasians (78), which appeared to be the most common known DPD variant ( 75). Patients being hetero- or homozygous carriers of this DPD gene variation have been shown to experience severe and even lethal toxicities of 5-FU based regimens (Table 2b). Salonga et al. demonstrated that intratumoral gene expression of DPD is associated with response and survival in patients with mCRC treated with infusional 5-FU chemotherapy (14). Recently, Isshi et al. demonstrated that high levels of orotate phosphorybosyl transferase (OPRT) may be associated with increased sensitivity to 5-FU based chemotherapy ( 79). OPRT catalyzes the reduction of FUDP to the actively TS-inhibiting metabolite FdUMP, indicating a role for chemosensitivity to 5-FU. A recent study by Ichikawa et al. indicated that a newly identified SNP of OPRT exon 3 (G-to-A substitution) may be critical to predict toxicity to 5-FU based chemotherapy (80). For some time it has been known that low folate levels are associated with greater toxicity to TS inhibitors and that co-administration of folic acid can significantly reduce this toxicity while preserving anti-tumorous activity of the drug ( 81, 82). Methylentetrahydrofolate reductase (MTHFR) is a key enzyme of 5-FU metabolism, alternating 5-FU sensitivity indirectly by folate pool variations. MTHFR converts 5,10methylenetetrahydrofolate, a substrate of TS, to 5-methyltetrahydrofolate (19,83). The relationship of TS and MTHFR gene polymorphisms on 5-FU sensitivity was tested on 19 human cancer cell lines (head and neck, breast, digestive tract) in the absence and presence of folinic acid supplementation. The C677T and A1298C MTHFR polymorphisms were determined by melting curve analyses (LightCycler). There was a marked trend for a greater FU efficacy in mutated A1298C variants (C/C+A/C) as compared to homozygous cell lines (A/A) (p = 0.055 and 0.085 without and with folic

Table 3 TS Polymorphisms as Predictive and Prognostic Markers in Adjuvant and Metastatic Chemotherapy Setting

164

Study Design

Results

Prognostic (OS, PFS)

Predicitve (RR, Toxicity)

Reference

129 colorectal cancer patients homogeneously treated with FU plus levamisole or leucovorin in the adjuvant setting were included → Tumor genotyoing of TS polymorphisms (5 -UTR TSER, A>G SNP, TS 1494–6bp/–6bp) 166 colorectal cancer patients with Dukes B2 and Dukes C underwent radical resection and received adjuvant 5-FU based chemotherapy → Genotyping of TS polymorphisms in peripheral blood mononuclear cells (5 -UTR TSER, A>G SNP, TS 1494–6bp/–6bp)

1) 5 -UTR 3R/3R genotype of TSER showed better outcome 2) A>G SNP added no prognostic information TS 1494–6bp/–6bp was protective 3) 3R/-6bp haplotype showed survival benefit compared to 2R/+6bp haplotype

1) TSER 3R/3R:OS ↑ (p = 0.020) PFS n/a 2) A>G SNP: OS – (p = 0.18) PFS n/a 3) TS 1494–6bp/–6bp: OS ↑(p= 0.012) PFS n/a

RR n/a

Dotor et al. (63)

1) 5 -UTR 3R/3R genotype of 1) TSER 3R/3R:OS ↑(p = 0.009) PFS ↑ (p = 0.048) TSER showed significantly 2) A>G SNP:OS n/a longer DFS and OS PFS n/a 2) 5 -UTR 2R/3R & +6b/6bp 3) 5 -UTR 2R/3R & combination showed significantly longer DFS and OS 3 -UTR+6b/+6bp 3) TS genotypes and their OS ↑(p = 0.043) combinations, which have been PFS ↑(p = 0.049) reported earlier with high TS expression, show longer OS/PFS (3R/3R with any 3 -UTR genotype and 2R/3R with +6bp/+6bp)

RR n/a

Toxicity n/a

Toxicity n/a

Hitre et al. (69)

165

1) Low TS expression group (A) 89 mCRC patients received 5-FU 1) Patients with TS low → OS ↑ (p = 0.03) based chemotherapy expression genotype → PFS ↑ (p = 0.07) → Tumor genotyping of TS (A-group), had an overall 2) High TS expression group (B) polymorphisms better RR, DFS and OS → OS ↓ (p = 0.03) A / low expression group = than the TS high → PFS ↓ (p = 0.07) expression genotype *2R/*2R; *2R/*3C; *3C/*3C → TS genotype as B / high expression group = *2R/*3 G;*3C/*3 G; *3 G/*3 G independent predictor of RR, PFS, OS 50 mCRC patients received 5-FU 1) Patients with homozygous n/a based chemotherapy TSER-3R polymorphisms → Tumor genotyping of TS had a 3.6-fold higher polymorphisms mRNA expression  (5 -UTR TSER) 2) Patients homozygous for TSER 3R had worse RRs, than patients homo- or heterozygous for TSER-2R 3) Inverse relationship of TSER-3R and toxicity

1) Low TS expression Marcuello et al. (70) group (A) → RR ↑ (p = 0.04) → Toxicity n/a 2) High TS expression group (B) → RR ↓ (p = 0.04) → Toxicity n/a Pullarkat et al. (3) 1) TSER 3R/3R → RR ↓ (p = 0.041) → Toxicity ↓ (p = 0.008) 2) TSER 2R/2R & TSER 2R/3R → RR ↑ (p = 0.041) → Toxicity ↑ (p = 0.008)

(Continued)

Table 3 (Continued) Study Design

Results

Prognostic (OS, PFS)

166

1) Low TS expression group (A) 90 patients with advanced gastric 1) The presence of at least → OS ↑ (p < 0.001–0.1) cancer received adjuvant 5-FU one high TS expression → PFS ↑ (p = 0.003) based chemotherapy after genotype showed 2) High TS expression group (B) surgical resection independent adverse → OS ↓ (p < 0.001–0.1) → Tumor genotyping of TS prognostic role in → PFS ↓ (p = 0.003) polymorphisms multivariant analysis A / low expression group = 2) Patients of group A showed better overall 5 -UTR *2R/*2R; *2R/*3C; survival and PFS in *3C/*3C  comparison to patients of 3 -UTR del6/del6 and del6/+6 group B, or patients at B / high expression group = least carrying one 5 -UTR *2R/*3 G;*3C/*3 G; “high-expression” allele *3 G/*3 G 3 -UTR +6/+6

Predicitve (RR, Toxicity)

Reference

RR n/a Toxicity n/a

Kawakami et al. (71)

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

167

acid supplementation, respectively). These results suggest a potential role of A1298C MTHFR polymorphism on fluoropyrimidine sensitivity and toxicity (74,84).

8. CONCLUSION Despite significant advances in the treatment of CRC, variations in response and toxicity among patients continues to pose an ongoing challenge for oncologists. Having a reliable panel of prognostic and predictive markers is critical for selecting optimal treatment for a patient based on his or her individual tumor genotype. Although some markers have been identified and investigated in CRC, none of them have been routinely used outside of clinical trials. In this chapter we discussed several molecular predictive and/or prognostic markers in the 5-FU pathway that are important in “tailoring” individual chemotherapy. To try to account for the discrepant and sometimes contradictory data in the literature, one can consider many factors such as small study sample number, lack of standardized methodologies for measuring protein and gene expression, suboptimal samples consisting of different mixtures of cells, tissue-specific differences, and study populations with different allele distributions. For example, measuring higher levels of TS mRNA or TS protein may be ascribed to the use of laser-captured microdissection, purifying tumor cells from adjacent stroma cells, whereas mRNA analysis of non-microdissected tumor tissue may lead to false positive results. Beside these technical issues, TS expression and its regulation by various types of TS gene polymorphisms seem to be critical to predicting response and overall clinical outcome. The possibility of three different polymorphisms in the same gene obviously complicates efforts aimed at understanding the functional significance of each individual polymorphism. In the case of TS, there are 18 different allele combinations possible, all of which may theoretically influence clinical outcome. Thus, it is likely that the observed TS expression levels, tumor response and toxicity may be complicated functions of multiple TS gene alterations, rather than the result of one single polymorphism. Expression of selected elements of the 5-FU metabolic pathway is predictive of response to 5-FU based chemotherapy regimens. High levels of TS, TP, and DPD are independent predictors of decreased response; vice versa, lower levels of TS, TP, and DPD correlate with higher sensitivity to 5-FU. High expression levels of one of these genes, even in the presence of down-regulation of the other two, has an adverse effect on response to 5-FU. In a few years, microarray technology may be the preferred method of genotyping, with the advent of customizable chips, in addition to chips to test polymorphisms and gene expression. The introduction of new therapeutic agents and the discovery and validation of predictive and prognostic markers along with new screening tools will enable oncologist to tailor patient-specific chemotherapy regimens by maximizing drug efficacy and minimizing adverse and possibly severe side effects. Much work, however, remains to be done. Ongoing and future clinical trials hold promise for further improvements in optimizing and specifying chemotherapy individually, not only prolonging lives, but also in augmenting quality of life.

168

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

REFERENCES 1. Horie N, Aiba H, Oguro K et al. Functional analysis and DNA polymorphism of the tandemly repeated sequences in the 5’-terminal regulatory region of the human gene for thymidylate synthase. Cell Struct Funct 1995;20:191–197. 2. Marsh S, McLeod HL. Thymidylate synthase pharmacogenetics in colorectal cancer. Clin Colorectal Cancer 2001;1:175–178; discussion 179–181. 3. Pullarkat ST, Stoehlmacher J, Ghaderi V et al. Thymidylate synthase gene polymorphism determines response and toxicity of 5-FU chemotherapy. Pharmacogenomics J 2001;1:65–70. 4. Jemal A, Siegel R, Ward E et al. Cancer statistics, 2006. CA Cancer J Clin 2006;56:106–130. 5. de Gramont A, Figer A, Seymour M et al. Leucovorin and fluorouracil with or without oxaliplatin as first-line treatment in advanced colorectal cancer. J Clin Oncol 2000;18:2938–2947. 6. Douillard JY, Cunningham D, Roth AD et al. Irinotecan combined with fluorouracil compared with fluorouracil alone as first-line treatment for metastatic colorectal cancer: a multicentre randomised trial. Lancet 2000;355:1041–1047. 7. Grothey A, Sargent D, Goldberg RM et al. Survival of patients with advanced colorectal cancer improves with the availability of fluorouracil-leucovorin, irinotecan, and oxaliplatin in the course of treatment. J Clin Oncol 2004;22:1209–1214. 8. Tournigand C, Andre T, Achille E et al. FOLFIRI followed by FOLFOX6 or the reverse sequence in advanced colorectal cancer: a randomized GERCOR study. J Clin Oncol 2004;22:229–237. 9. Giantonio BJ, Levy DE, O’Dwyer et al. A phase II study of high-dose bevacizumab in combination with irinotecan, 5-fluorouracil, leucovorin, as initial therapy for advanced colorectal cancer: results from the Eastern Cooperative Oncology Group study E2200. Ann Oncol 2006;17:1399–1403. 10. Hurwitz H, Fehrenbacher L, Novotny W et al. Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N Engl J Med 2004;350:2335–2342. 11. McLeod HL, Papageorgio C, Watters JW. Using genetic variation to optimize cancer chemotherapy. Clin Adv Hematol Oncol 2003;1:107–111. 12. McLeod HL, Yu J. Cancer pharmacogenomics: SNPs, chips, and the individual patient. Cancer Invest 2003;21:630–640. 13. Ichikawa W, Uetake H, Shirota Y et al. Combination of dihydropyrimidine dehydrogenase and thymidylate synthase gene expressions in primary tumors as predictive parameters for the efficacy of fluoropyrimidine-based chemotherapy for metastatic colorectal cancer. Clin Cancer Res 2003;9: 786–791. 14. Salonga D, Danenberg KD, Johnson M et al. Colorectal tumors responding to 5-fluorouracil have low gene expression levels of dihydropyrimidine dehydrogenase, thymidylate synthase, and thymidine phosphorylase. Clin Cancer Res 2000;6:1322–1327. 15. Heidelberger C, Chaudhuri NK, Danneberg P et al. Fluorinated pyrimidines: a new class of tumourinhibitory compounds. Nature 1957;179:663–666. 16. Danenberg PV. Thymidylate synthetase: a target enzyme in cancer chemotherapy. Biochim Biophys Acta 1977;473:73–92. 17. Danenberg PV. Pharmacogenomics of thymidylate synthase in cancer treatment. Front Biosci 2004;9:2484–2494. 18. Aschele C, Sobrero A, Faderan MA et al. Novel mechanism(s) of resistance to 5-fluorouracil in human colon cancer (HCT-8) sublines following exposure to two different clinically relevant dose schedules. Cancer Res 1992;52:1855–1864. 19. Cohen V, Panet-Raymond V, Sabbaghian N et al. Methylenetetrahydrofolate reductase polymorphism in advanced colorectal cancer: a novel genomic predictor of clinical response to fluoropyrimidinebased chemotherapy. Clin Cancer Res 2003;9:1611–1615. 20. Thirion P, Michiels S, Pignon JP et al. Modulation of fluorouracil by leucovorin in patients with advanced colorectal cancer: an updated meta-analysis. J Clin Oncol 2004;22:3766–3775. 21. Mathijssen RH, van Alphen RJ, Verweij J et al. Clinical pharmacokinetics and metabolism of irinotecan (CPT-11). Clin Cancer Res 2001;7:2182–2194.

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

169

22. Simpson D, Dunn C, Curran M et al. Oxaliplatin: a review of its use in combination therapy for advanced metastatic colorectal cancer. Drugs 2003;63:2127–2156. 23. Venook A. Critical evaluation of current treatments in metastatic colorectal cancer. Oncologist 2005;10:250–261. 24. Hoshino S, Yamashita Y, Maekawa T et al. Effects on DNA and RNA after the administration of two different schedules of 5-fluorouracil in colorectal cancer patients. Cancer Chemother Pharmacol 2005;56:648–652. 25. Kubota T, Watanabe M, Otani Y et al. Different pathways of 5-fluorouracil metabolism after continuous venous or bolus injection in patients with colon carcinoma: possible predictive value of thymidylate synthetase mRNA and ribonucleotide reductase for 5-fluorouracil sensitivity. Anticancer Res 2002;22:3537–3540. 26. O’Connell MJ, Martenson JA, Wieand HS et al. Improving adjuvant therapy for rectal cancer by combining protracted-infusion fluorouracil with radiation therapy after curative surgery. N Engl J Med 1994;331:502–507. 27. Sobrero AF, Aschele C, Bertino JR. Fluorouracil in colorectal cancer: a tale of two drugs: implications for biochemical modulation. J Clin Oncol 1997;15:368–381. 28. Efficacy of intravenous continuous infusion of fluorouracil compared with bolus administration in advanced colorectal cancer. Meta-analysis Group In Cancer. J Clin Oncol 1998;16:301–308. 29. Schuller J, Cassidy J, Dumont E et al. Preferential activation of capecitabine in tumor following oral administration to colorectal cancer patients. Cancer Chemother Pharmacol 2000;45:291–297. 30. Van Cutsem E, Twelves C, Cassidy J et al. Oral capecitabine compared with intravenous fluorouracil plus leucovorin in patients with metastatic colorectal cancer: results of a large phase III study. J Clin Oncol 2001;19:4097–4106. 31. El Sayed YM, Sadee W. Metabolic activation of R,S-1-(tetrahydro-2-furanyl)-5-fluorouracil (ftorafur) to 5-fluorouracil by soluble enzymes. Cancer Res 1983;43:4039–4044. 32. Tuchman M, Ramnaraine ML, O’Dea RF. Effects of uridine and thymidine on the degradation of 5-fluorouracil, uracil, and thymine by rat liver dihydropyrimidine dehydrogenase. Cancer Res 1985;45:5553–5556. 33. Tatsumi K, Fukushima M, Shirasaka T et al. Inhibitory effects of pyrimidine, barbituric acid, and pyridine derivatives on 5-fluorouracil degradation in rat liver extracts. Jpn J Cancer Res 1987;78: 748–755. 34. Shirasaka T, Shimamoto Y, Fukushima M. Inhibition by oxonic acid of gastrointestinal toxicity of 5-fluorouracil without loss of its anti-tumor activity in rats. Cancer Res 1993;53:4004–4009. 35. Takechi T, Nakano K, Uchida J et al. Anti-tumor activity and low intestinal toxicity of S-1, a new formulation of oral tegafur, in experimental tumor models in rats. Cancer Chemother Pharmacol 1997;39:205–211. 36. van Groeningen CJ, Peters GJ, Schornagel JH et al. Phase I clinical and pharmacokinetic study of oral S-1 in patients with advanced solid tumors. J Clin Oncol 2000;18:2772–2779. 37. Chollet P, Schoffski P, Weigang-Kohler K et al. Phase II trial with S-1 in chemotherapy-naive patients with gastric cancer: a trial performed by the EORTC Early Clinical Studies Group (ECSG). Eur J Cancer 2003;39:1264–1270. 38. Koizumi W, Kurihara M, Nakano S et al. Phase II study of S-1, a novel oral derivative of 5fluorouracil, in advanced gastric cancer. for the S-1 Cooperative Gastric Cancer Study Group. Oncology 2000;58:191–197. 39. Sakata Y, Ohtsu A, Horikoshi N et al. Late phase II study of novel oral fluoropyrimidine anticancer drug S-1 (1 M tegafur-0.4 M gimestat-1 M otastat potassium) in advanced gastric cancer patients. Eur J Cancer 1998;34:1715–1720. 40. Koizumi W, Tanabe S, Saigenji K et al. Phase I/II study of S-1 combined with cisplatin in patients with advanced gastric cancer. Br J Cancer 2003;89:2207–2212. 41. Ajani JA, Faust J, Ikeda K et al. Phase I pharmacokinetic study of S-1 plus cisplatin in patients with advanced gastric carcinoma. J Clin Oncol 2005;23:6957–6965.

170

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

42. Mandola MV, Stoehlmacher J, Muller-Weeks S et al. A novel single nucleotide polymorphism within the 5’ tandem repeat polymorphism of the thymidylate synthase gene abolishes USF-1 binding and alters transcriptional activity. Cancer Res 2003;63:2898–2904. 43. Berger SH, Jenh CH, Johnson LF et al. Thymidylate synthase overproduction and gene amplification in fluorodeoxyuridine-resistant human cells. Mol Pharmacol 1985;28:461–467. 44. Kundu NG, Heidelberger C. Cyclopenta(f)isoquinoline derivatives designed to bind specifically to native deoxyribonucleic acid. 3. Interaction of 6-carbamylmethyl-8-methyl-7 H-cyclopenta(f)isoquinolin-3(2 H)-one with deoxyribonucleic acids and polydeoxyribonucleotides. Biochem Biophys Res Commun 1974;60:561–568. 45. Shirota Y, Stoehlmacher J, Brabender J et al. ERCC1 and thymidylate synthase mRNA levels predict survival for colorectal cancer patients receiving combination oxaliplatin and fluorouracil chemotherapy. J Clin Oncol 2001;19:4298–4304. 46. Stoehlmacher J, Park DJ, Zhang W et al. A multivariate analysis of genomic polymorphisms: prediction of clinical outcome to 5-FU/oxaliplatin combination chemotherapy in refractory colorectal cancer. Br J Cancer 2004;91:344–354. 47. Leichman L, Lenz HJ, Leichman CG et al. Quantitation of intratumoral thymidylate synthase expression predicts for resistance to protracted infusion of 5-fluorouracil and weekly leucovorin in disseminated colorectal cancers: preliminary report from an ongoing trial. Eur J Cancer 1995;31A: 1306–1310. 48. Villafranca E, Okruzhnov Y, Dominguez MA et al. Polymorphisms of the repeated sequences in the enhancer region of the thymidylate synthase gene promoter may predict downstaging after preoperative chemoradiation in rectal cancer. J Clin Oncol 2001;19:1779–1786. 49. Iacopetta B, Grieu F, Joseph D et al. A polymorphism in the enhancer region of the thymidylate synthase promoter influences the survival of colorectal cancer patients treated with 5-fluorouracil. Br J Cancer 2001;85:827–830. 50. Carlini LE, Meropol NJ, Bever J et al. UGT1A7 and UGT1A9 polymorphisms predict response and toxicity in colorectal cancer patients treated with capecitabine/irinotecan. Clin Cancer Res 2005;11:1226–1236. 51. Gibson TB. Polymorphisms in the thymidylate synthase gene predict response to 5-fluorouracil therapy in colorectal cancer. Clin Colorectal Cancer 2006;5:321–323. 52. Kralovanszky J, Koves I, Orosz Z et al. Prognostic significance of the thymidylate biosynthetic enzymes in human colorectal tumors. Oncology 2002;62:167–174. 53. Nakagawa T, Tanaka F, Otake Y et al. Prognostic value of thymidylate synthase expression in patients with p-stage I adenocarcinoma of the lung. Lung Cancer 2002;35:165–170. 54. Marsh S. Thymidylate synthase pharmacogenetics. Invest New Drugs 2005;23:533–537. 55. Marsh S, Collie-Duguid ES, Li T et al. Ethnic variation in the thymidylate synthase enhancer region polymorphism among Caucasian and Asian populations. Genomics 1999;58:310–312. 56. Kawakami SD, Omura K, Park J et al. Effects of polymorphic tandem repeat sequence on the in vitro translation of messenger RNA. Proc Am Assoc Cancer Res (Abstract) 1999;40::436–437. 57. Kawakami K, Omura K, Kanehira E et al. Polymorphic tandem repeats in the thymidylate synthase gene is associated with its protein expression in human gastrointestinal cancers. Anticancer Res 1999;19:3249–3252. 58. Etienne MC, Chazal M, Laurent-Puig P et al. Prognostic value of tumoral thymidylate synthase and p53 in metastatic colorectal cancer patients receiving fluorouracil-based chemotherapy: phenotypic and genotypic analyses. J Clin Oncol 2002;20:2832–2843. 59. Mandola MV, Stoehlmacher J, Zhang W et al. A 6bp polymorphism in the thymidylate synthase gene causes message instability and is associated with decreased intratumoral TS mRNA levels. Pharmacogenetics 2004;14:319–327. 60. Kawakami K, Watanabe G. Identification and functional analysis of single nucleotide polymorphism in the tandem repeat sequence of thymidylate synthase gene. Cancer Res 2003;63:6004–6007.

Chapter 10 / TS Gene Variation in Colorectal Cancer Patients

171

61. Ulrich CM, Bigler J, Velicer CM et al. Searching expressed sequence tag databases: discovery and confirmation of a common polymorphism in the thymidylate synthase gene. Cancer Epidemiol Biomarkers Prev 2000;9:1381–1385. 62. Lenz H-J, Zhang W, Zahedy S et al. A 6 base-pair deletion in the 3 UTR of the thymidylate synthase (TS) gene predicts TS mRNA expression in colorectal tumors: a possible candidate gene for colorectal cancer risk. Proc Am Assoc Cancer Res (Abstract) 2002;43. 63. Dotor E, Cuatrecases M, Martinez-Iniesta M et al. Tumor thymidylate synthase 1494del6 genotype as a prognostic factor in colorectal cancer patients receiving fluorouracil-based adjuvant treatment. J Clin Oncol 2006;24:1603–1611. 64. Vogelstein B, Fearon ER, Kern SE et al. Allelotype of colorectal carcinomas. Science 1989;244: 207–211. 65. Zinzindohoue F, Ferraz JM, Laurent-Puig P. Thymidylate synthase promoter polymorphism. J Clin Oncol 2001;19:3442. 66. Uchida K, Hayashi K, Kawakami K et al. Loss of heterozygosity at the thymidylate synthase (TS) locus on chromosome 18 affects tumor response and survival in individuals heterozygous for a 28-bp polymorphism in the TS gene. Clin Cancer Res 2004;10:433–439. 67. Morganti M, Ciantelli M, Giglioni B et al. Relationships between promoter polymorphisms in the thymidylate synthase gene and mRNA levels in colorectal cancers. Eur J Cancer 2005;41:2176–2183. 68. Jakobsen A, Nielsen JN, Gyldenkerne N et al. Thymidylate synthase and methylenetetrahydrofolate reductase gene polymorphism in normal tissue as predictors of fluorouracil sensitivity. J Clin Oncol 2005;23:1365–1369. 69. Hitre E, Budai B, Adleff V et al. Influence of thymidylate synthase gene polymorphisms on the survival of colorectal cancer patients receiving adjuvant 5-fluorouracil. Pharmacogenet Genomics 2005;15:723–730. 70. Marcuello E, Altes A, del Rio E et al. Single nucleotide polymorphism in the 5 tandem repeat sequences of thymidylate synthase gene predicts for response to fluorouracil-based chemotherapy in advanced colorectal cancer patients. Int J Cancer 2004;112:733–737. 71. Kawakami K, Graziano F, Watanabe G et al. Prognostic role of thymidylate synthase polymorphisms in gastric cancer patients treated with surgery and adjuvant chemotherapy. Clin Cancer Res 2005;11:3778–3783. 72. Tan RM, Zehnbauer B, Picus J et al. TYMS genotype polymorphism directed neoadjuvant chemoradiation for patients (pts) with T3/T4 adenocarcinoma of the rectum (Abstract). 2006 Gastrointestinal Cancers Symposium 2006. 73. Popat S, Matakidou A, Houlston RS. Thymidylate synthase expression and prognosis in colorectal cancer: a systematic review and meta-analysis. J Clin Oncol 2004;22:529–536. 74. Heggie GD, Sommadossi JP, Cross DS et al. Clinical pharmacokinetics of 5-fluorouracil and its metabolites in plasma, urine, and bile. Cancer Res 1987;47:2203–2206. 75. Van Kuilenburg AB, Vreken P, Abeling NG et al. Genotype and phenotype in patients with dihydropyrimidine dehydrogenase deficiency. Hum Genet 1999;104:1–9. 76. van Kuilenburg AB, Muller EW, Haasjes J et al. Lethal outcome of a patient with a complete dihydropyrimidine dehydrogenase (DPD) deficiency after administration of 5-fluorouracil: frequency of the common IVS14+1 G>A mutation causing DPD deficiency. Clin Cancer Res 2001;7:1149–1153. 77. van Kuilenburg AB. Dihydropyrimidine dehydrogenase and the efficacy and toxicity of 5-fluorouracil. Eur J Cancer 2004;40:939–950. 78. Diasio RB, Beavers TL, Carpenter JT. Familial deficiency of dihydropyrimidine dehydrogenase: biochemical basis for familial pyrimidinemia and severe 5-fluorouracil-induced toxicity. J Clin Invest 1988;81:47–51. 79. Isshi K, Sakuyama T, Gen T et al. Predicting 5-FU sensitivity using human colorectal cancer specimens: comparison of tumor dihydropyrimidine dehydrogenase and orotate phosphoribosyl transferase activities with in vitro chemosensitivity to 5-FU. Int J Clin Oncol 2002;7:335–342. 80. Ichikawa W, Nihei Z, Uetake H et al. UFT plus leucovorin for metastatic colorectal cancer: Japanese experience. Oncology (Williston Park) 2000;14:41–43.

172

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

81. Alati T, Worzalla JF, Shih C et al. Augmentation of the therapeutic activity of lometrexol -(6-R)5,10dideazatetrahydrofolate- by oral folic acid. Cancer Res 1996;56:2331–2335. 82. Morgan SL, Baggott JE, Vaughn WH et al. The effect of folic acid supplementation on the toxicity of low-dose methotrexate in patients with rheumatoid arthritis. Arthritis Rheum 1990;33:9–18. 83. Ulrich CM, Yasui Y, Storb R et al. Pharmacogenetics of methotrexate: toxicity among marrow transplantation patients varies with the methylenetetrahydrofolate reductase C677T polymorphism. Blood 2001;98:231–234. 84. Sohn KJ, Croxford R, Yates Z et al. Effect of the methylenetetrahydrofolate reductase C677T polymorphism on chemosensitivity of colon and breast cancer cells to 5-fluorouracil and methotrexate. J Natl Cancer Inst 2004;96:134–144. 85. Engstrom P. Update: NCCN colon cancer Clinical Practice Guidelines. J Natl Compr Canc Netw 2005;3 Suppl 1:S25–28.

11

Thiopurines in the Treatment of Childhood Acute Lymphoblastic Leukemia and Genetic Variants of the Thiopurine S-Methyltransferase Gene Martin Stanulla, MD, MSc, Elke Schaeffeler, PhD, and Matthias Schwab, MD CONTENTS Treatment of Childhood Acute Lymphoblas tic Leukemia Thiopurines in the Treatment of Childhood Acute Lymphoblas tic Leukemia Metabolis m of 6-Mercaptopurine and 6-Thioguanine Monitoring of Thiopurine Therapy Thiopurine S-Methyltrans feras e Conclus ions and Further Pers pectives References

S UMMARY The thiopurines 6-mercaptopurine (6-MP) and 6-thioguanine (6-TG) are essential components of treatment protocols for childhood acute lymphoblastic leukemia From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 11, 

173

174

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

(ALL). In the past 25 years, considerable insights into thiopurine pharmacology have been gained through continuing research efforts which have led to the development of strategies for improving efficacy and reducing toxicity associated with 6-MP and 6-TG application. One important route of metabolism for thiopurines is methylation by the enzyme thiopurine S-methyltransferase (TPMT). The gene coding for TPMT is subject to phenotypically relevant genetic variation, with heterozygous individuals having intermediate TPMT activity, and homozygous variant individuals having low TPMT activity. In this chapter, we review the role of thiopurines in the treatment of childhood ALL and provide an overview of strategies aimed at optimization of thiopurine application by therapeutic drug monitoring of thiopurine metabolites and genoor phenotyping of TPMT. Key Words: Thiopurines; mercaptopurine; thioguanine; thiopurine S-methyltransferase; TPMT; leukemia; acute lymphoblastic leukemia; childhood; treatment; genetic variation; genetic polymorphism

1. TREATMENT OF CHILDHOOD ACUTE LYMPHOBLASTIC LEUKEMIA Treatment results in childhood acute lymphoblastic leukemia (ALL) are one of the true success stories of modern clinical oncology, with overall cure rates of 80% achieved by application of intensive multiagent chemotherapeutic regimens (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15). Modern regimens consist of at least four elements: (1) an induction phase aiming at an initial remission induction within approximately 4–6 weeks through the use of multiple cancer chemotherapeutic agents; (2) a consolidation segment to eradicate residual leukemic blasts in patients who are in remission by morphologic criteria; (3) extra-compartment therapy such as central nervous system (CNS)-directed therapy; and (4) a maintenance period to further stabilize remission by suppressing re-emergence of a drug-resistant clone through continuing reduction of residual leukemic cells (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15). In the second half of the 1970s, an additional treatment element was introduced, a so-called “re-induction” or “delayed re-intensification” phase (16,17). For certain patient populations, CNS-directed therapy may include preventive or therapeutic cranial radiotherapy as an additional treatment component to enhance specific targeting of leukemic cells in the CNS ( 18, 19). Depending on the protocol, the overall treatment duration varies from 2 to 3 years. Adjustment of therapy according to the risk of treatment failure has become a common feature in the clinical management of childhood ALL. Continuing research on the clinical and biological aspects of leukemias has indentified numerous features with prognostic potential, several of which have been extensively evaluated in large patient populations, although not uniformly in all subgroups. These factors mostly include clinical (e.g., gender, initial absolute leukocyte count, and age at diagnosis), immunological (e.g., immunophenotype), and genetic characteristics (e.g., non-random recurrent chromosomal aberrations) that are assessable at diagnosis (20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39). In addition, several study

Chapter 11 / Thiopurines and TPMT in Childhood ALL

175

groups have evaluated a variety of estimates of early response to treatment as a prognostic factor for treatment allocation (40,41,42,43,44,45,46,47,48,49, 50,51,52,53,54,55,56). The various risk assessment procedures applied by different study groups mainly translate into therapy stratification using two or three risk groups (e.g., standard/low, intermediate, high). As an example of a risk-adapted modern clinical protocol, Fig. 1 shows an outline of the treatment strategy applied in the ALL-BFM 95 trial conducted by the Berlin–Frankfurt–M¨unster (BFM) Study Group from 1995 to 2000 (57). In this trial patients were assigned to standard-risk (SR), intermediate-risk (MR), and highrisk (HR) subgroups. The main criteria for stratification were the early response to glucocorticoid treatment as assessed by quantifying the leukemic blast cell reduction in the peripheral blood on treatment day eight (prednisone response), the initial absolute leukocyte count, and age at diagnosis. Additional criteria included the presence of T-cell immunophenotype, and detection of the chromosomal translocations t(9;22) or t(4;11) or their respective molecular equivalents, the fusion transcripts BCR/ABL or MLL/AF4. Treatment strategies differ among study groups and it is well recognized that the significance of prognostic factors cannot easily be channelled into a uniform statement for all study groups ( 58, 59, 60). Additional differences between study groups (e.g., eligibility criteria and ethnic or racial composition of study populations) have to be taken into account when the relevance of specific prognostic factors is discussed. Also, virtually all prognostic factors are associated with the type and the intensity of treatment

Fig. 1. Outline of treatment strategy applied in the ALL–Berlin–Frankfurt–M¨unster (BFM) 95 study (1995–2000). Patients were assigned to standard-risk (SR), intermediate-risk (MR) and high-risk (HR) subgroups. Elements containing 6-mercaptopurine (6-MP) or 6-thioguanine (6-TG) are indicated in grey. Abbreviations : PRED-GR, prednisone good response; PRED-PR, prednisone poor response; DEXA, dexamethasone; VCR, vincristine; DNR, daunorubicin; HD-MTX, high-dose methotrexate; LD-ARA-C, low-dose cytarabine; SCT, stem cell transplantation.

176

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

administered. Thus, therapeutic changes in consecutive treatment protocols for childhood ALL may lead to a loss of prognostic impact of previously relevant factors, which underscores the importance of ongoing assessment. The majority of large study groups assess the most relevant and promising of these factors on a regular basis in prospective trials in order to evaluate their prognostic strength in association with their corresponding patient populations and current treatment strategies. This approach allows valid comparisons of different treatments for specific patient subgroups defined by the prognostic factors in question, and therefore may help to define which treatment components are beneficial for which subgroup. Clearly, the success of these cooperative ventures strongly depends on the quality of interaction among the study groups and the ability to arrive at a consensus to generate the tools needed to develop a more detailed uniform approach for risk assessment in childhood ALL.

2. THIOPURINES IN THE TREATMENT OF CHILDHOOD ACUTE LYMPHOBLASTIC LEUKEMIA The thiopurines 6-mercaptopurine (6-MP) and 6-thioguanine (6-TG) were synthesized by Elion and Hitchings in the 1950s, and they play an important part in treatment protocols for leukemia (61,62,63,64). For no other than historical reasons, 6-MP is used in ALL while 6-TG is mainly used in acute myeloid leukemia (AML) or relapsed ALL. First-line treatment for childhood ALL usually includes several cycles of 6-MP at doses of 50–75 mg/m2 /d, starting as early as in consolidation/early intensification treatment until up to 36 months after diagnosis (1,2,3). To elucidate the extensive administration of thiopurines in the treatment of childhood ALL, using the ALL-BFM 95 trial as an example, the elements containing either 6-MP or 6-TG are indicated (57); see Fig. 1. After induction of remission, thiopurines are used almost throughout the entire therapy. The most extensive application of thiopurines occurs during the maintenance phase. Maintenance treatment aims at a further stabilization of remission by suppressing the re-emergence of a drug-resistant clone through consistently reducing the pool of residual leukemic cells. The current standard of maintenance therapy consists of up to 2–3 years of treatment (from initial time of diagnosis) with daily oral 6-MP and weekly oral methotrexate (1,2,3). The combination of 6-MP with methotrexate acts synergistically as methotrexate inhibits purine de novo synthesis, leading to a higher intracellular availability and increased incorporation of phosphorylated thiopurines in DNA and RNA (65,66,67,68). Several studies in childhood ALL compared parental to intravenous 6-MP application with none of them proving advantageous for parental administration (69,70,71). In contrast to administration in earlier treatment elements applied in childhood ALL protocols (e.g., consolidation or extra-compartment therapy) where thiopurines are given at fixed doses, in maintenance both 6-MP and methotrexate doses are adjusted according to absolute leukocyte or neutrophil and platelet counts. Current BFM dose modification guidelines for maintenance treatment in childhood ALL call for an absolute leukocyte count in a target range of 2–3 × 109 /L ( 2, 57). Minimal requirements for starting maintenance treatment are an absolute leukocyte count of ≥ 1 × 109 /L with at least 0.2 × 109 /L neutrophils and 50 × 109 /L thrombocytes (counts not decreasing).

Chapter 11 / Thiopurines and TPMT in Childhood ALL

177

The starting dose as well as dose adjustment in therapy are made according to guidelines fixed in the treatment protocols. For pneumocystis carinii pneumonia prophylaxis, trimethoprim-sulfamethoxazole is administered on three consecutive days per week, with the largest possible interval in reference to the weekly methotrexate application. This is done to account for the theoretical enhancement of the antifolic activity of methotrexate by co-administered trimethoprim-sulfamethoxazole (72,73). Because several reports have suggested an improved outcome with bedtime administration, 6-MP is commonly administered in the evening hours (74,75). Also, 6-MP should not be given in combination with milk because the xanthine oxidase (XO) activity contained in milk decreases the bioavailability of 6-MP (76,77,78). Relling and colleagues at St Jude Childrens Research Hospital have demonstrated that maintaining the highest tolerable dose of daily 6-MP in maintenance therapy is an important prognostic factor in childhood ALL (79). However, when results are compared on efficacy and toxicity of thiopurines from clinical trials of different study groups, it must be considered that recommendations on the exact schedule of 6-MP, methotrexate, and other thiopurine- and/or methotrexate-interacting medications regularly administered during maintenance (e.g., trimethoprim-sulfamethoxazole for pneumocystis carinii pneumonia prophylaxis) may differ (72,73,80,81). Of particular importance, dose modification guidelines may differ between clinical trials and study groups. In some childhood ALL treatment protocols, other cytotoxic chemotherapeutic drugs may be administered during the maintenance phase as well, such as pulsed applications of vincristine and a glucocorticoid (8284). The reduction of maintenance below two years (from the date of initial diagnosis) was associated with an increased frequency of leukemic relapses (82,85). Although it has been proven disadvantageous to shorten maintenance treatment, whether or not extended maintenance of up to 3 years is offering any beneficial effect in the context of different treatment strategies remains to be evaluated, particularly in males patients. Other issues that must be resolved in the future include differences in requirements for maintenance therapy in specific childhood ALL patient subsets (e.g., those defined immunophenotypically or genetically). These questions are still under investigation. With regard to the debate about the better thiopurine, there are three published randomized studies comparing the toxicity and efficacy of 6-TG with 6-MP in interim maintenance and maintenance therapy of childhood ALL. The first published trial randomized 474 patients with childhood ALL to either 6-TG or 6-MP in maintenance and was conducted by the German COALL study group (86). The COALL-92 trial results showed no difference in the primary outcomes, but observed a tendency of higher CNS relapse (3.4% vs. 0.8%, p = 0.053) and prolonged myelosuppression with marked thrombocytopenia in the 6-TG treatment arm. In this trial, the incidence of thrombocytopenia (< 100 × 109 /L) without evidence of leukocytopenia (< 1 × 109 /L) was 7.5-fold higher in the 6-TG compared with the 6-MP arm. Treatment interruptions occurred at a 1.7-fold higher rate in the 6-TG arm (4.7% of maintenance treatment weeks vs. 2.8% for 6-MP). A recent British trial, UK MRC ALL 97, randomized 1498 children to receive either 6-TG or 6-MP (87). After a median follow-up of 6 years, no differences in event-free survival were detected between the two treatment arms. A large reduction of isolated disease recurrence in the CNS by 6-TG [odds ratio (OR) = 0.53, 95% confidence interval

178

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

(CI) = 0.30 – 0.92] was offset by an increased risk of death in remission (OR = 2.22, 95% CI = 1.20 – 4.14), mainly due to infections. Strikingly, 11% of patients in the 6-TG arm compared to less than 2% in the 6-MP arm developed non-fatal hepatic toxicity with features of veno-occlusive disease (VOD) characterized by symptoms including tender hepatomegaly, hyperbilirubinaemia with elevated aminotransferases, thrombocytopenia out of proportion to neutropenia, and portal hypertension. In 85% of affected 6-TG recipients, these symptoms were observed during maintenance or interim maintenance. Of interest, in patients randomized to 6-MP, hepatic toxicity was associated with intensification elements in which both treatment arms received exclusively 6-TG. A third, not yet completely published study randomizing 6-TG and 6-MP in childhood ALL, is the U.S. trial CCG-1952 including more than 2000 patients with lower-risk features at diagnosis (NCI standard-risk ALL) ( 88). This trial suggests for the 6-TG arm a significantly better event-free survival (85.1% vs. 77.1% at 5 years, p = 0.02), less CNS and also bone marrow and testicular relapses without differences in remission deaths between the arms. Of importance, 6-TG administration led to hepatic VOD-like symptoms in approximately 20% of patients. Looking at 6-TG-related efficacy and toxicity in the three trials, it is obvious that the CCG-1952 trial shows the greatest benefit for the 6-TG arm at the highest rate of hepatotoxic side effects. This trial also used the highest starting dose of 6-TG in maintenance (60 mg/m2 /d, reduced to 50 mg/m2 /d after noting 6-TG-related hepatotoxicity). The above-mentioned British trial used a starting dose of 40 mg/m2 /d; the COALL-92 trial initially used 50 mg/m2 /d which was reduced to 40 mg/m2 /d upon recognition of marked thrombocytopenia. Although the latter two trials used comparable starting doses, differences between COALL-92 and UK MRC ALL 97 with regard to risk of CNS relapse and toxicity can be identified. As a possible explanation, it may be speculated that either different treatment-adjustment procedures may have led to different 6-TG doses, or different background treatments in the trials may have contributed differently to the development of hepatotoxic side effects associated with 6-TG. Of importance, data from patients with inflammatory bowel disease suggest a critical threshold dose for development TG-related hepatotoxicity (89). Unfortunately, for UK MRC ALL 97, 6-TG doses and information on treatment interruptions are not available for an adequate comparison with the COALL-92 trial. Thus, this discussion remains speculative until detailed information on 6-TG doses, dose adjustment procedures, and pharmacologically relevant parameters (e.g., 6-thioguanine nucleotide [6-TGN] levels, thiopurine S-methyltransferase [TPMT]) is available in association with well-defined clinical endpoints. The latter will also require a consensus agreement on diagnostic procedures for measuring hepatotoxic side effects. When these goals are met, carefully designed randomized controlled trials aimed at identifying the best thiopurine for childhood ALL, with adaptation to the underlying causes, will become even more informative.

3. METABOLISM OF 6-MERCAPTOPURINE AND 6-THIOGUANINE Anabolic and catabolic pathways are involved in the metabolism of thiopurines (Fig. 2) (90,91,92). The enzyme hypoxanthine guanine phosphoribosylferase, responsible for bio-activation of thiopurines, converts 6-MP into 6-thioinosine monophosphate,

Chapter 11 / Thiopurines and TPMT in Childhood ALL

179

Fig. 2. Metabolic scheme of thiopurine metabolism. 6-MP, 6-mercaptopurine; 6-TG, 6-thioguanine; 6-MMP, 6-methyl-mercaptopurine; XO, xanthine oxidase; TPMT, thiopurine S-methyltransferase; HPRT, hypoxanthine guanine phosphoribosyl transferase; IMPDH, inosine monophosphate dehydrogenase; GMPS, guanosine monophosphate synthetase; 6-TX, 6-thioxanthine; 6-TIMP, 6-thioinosine 5 -monophosphate; 6-MMPR, 6-methyl-mercaptopurine ribonucleotides; 6-TXMP, 6-thioxanthine monophoshate; 6-TGMP, 6-thioguanosine 5 monophosphate; 6-TGDP, 6-thioguanosine 5 diphosphate; 6-TGTP, 6-thioguanosine 5 triphosphate; 6-MTGN, 6-methyl-thioguanine nucleotides. The different size of some letters of TGDP and TGTP indicates that isolated 6-thioguanosine 5 phosphates are responsible for a particular mode of action (e.g., only TGDP is converted to 2 -deoxy TGDP).

which is metabolized stepwise by several other intracellular enzymes (e.g., inosine monophosphate dehydrogenase, guanosine monophosphate synthetase) into 6-TGN. 6-TGN represents the sum of 6-thioguanosine monophosphate (6-thio-GMP), -diphosphate (6-thio-GDP) and -triphosphate (6-thio-GTP). In contrast, both TPMT and XO are the predominant catabolic enzymes in the metabolism of thiopurines. TPMT catalyses the S-adenosyl-l-methionine dependent S-methylation of 6-MP and its metabolites into 6-methyl-mercaptopurine (6-MMP), 6-methyl-mercaptopurine ribonucleotides (6-MMPR) such as 6-methylthioinosine monophosphat (meTIMP), and 6-methyl-thioguanine nucleotides (6-MTGN) (93). Thiouric acid is formed by XO and is excreted renally. Whereas the cytosolic enzyme TPMT is expressed ubiquitously in humans [e.g., in the intestine, liver, red blood cells (RBC) and white blood cells], XO is not expressed in hematopoietic tissue (94). Therefore, TPMT-dependent methylation is critical in white blood cells, leading to an enhanced cytotoxic effect in patients with low TPMT activity. 6-TGNs have a half-life of approximately one week with large variability ( 91, 92). Several cytotoxic and immunosuppressant mechanisms of thiopurine action have been described ( 95, 96, 97, 98, 99, 100). However, the major underlying mode of action in

180

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

treatment of leukemia is suggested to be the incorporation of 6-TGN into DNA via 2 deoxy-TGDP and into RNA, thereby inhibiting replication, DNA repair mechanisms, and protein synthesis. The intermediate metabolite, me-TIMP, inhibits the purine de novo synthesis, thus interfering with replication and contributing to the cytotoxic effects of 6-MP ( 101, 102, 103). Thus, it becomes clear that thiopurines are likely to function at least as “two-in-one drugs,” and that different mechanisms of thiopurine action are associated to various extents with the clinically observed overall treatment effect of thiopurines. Recently, a novel mechanism was described. In a first approach, Tiede et al. ( 104) revealed that 6-thio-GTP bound to the GTPase Rac1 led to a mitochondrial pathway of apoptosis in CD3 and CD28 co-stimulated T-cells. These data are supported by the finding that patients with Crohn’s disease responding to azathioprine had an increased rate of apoptotic cells in intestinal lamina propria cells in contrast to non-responders. In a clinical approach, a ratio of > 0.85 for 6-thio-GTP to 6-thio-GDP and 6-TGN levels of > 100 pmol per 8 × 108 RBC (ratio > 0.85: response in 81%; ratio < 0.85: response in 36%) turned out to be predictive of a response in Crohn’s disease patients treated with azathioprine (105). Subsequently, Poppe et al. (106) found that 6-thio-GTP primarily blocked the guanosine exchange factor Vav1, leading to accumulation of inactive 6-thio-GDP-loaded Rac1, because 6-thio-GTP did not inhibit Rac effector coupling. Inhibition of Rac proteins further led to suppression of CD28 lamellipodia formation and interferon-gamma production. In the absence of apoptosis, Vav1 blockade leads to suppression of the conjugation of T-cells with antigen-presenting cells. In conclusion, these findings provide an explanation for azathioprine-mediated suppression of T-cell–dependent pathogenic immune responses. However, it is unclear so far whether this mechanism appears to be important for cytotoxicity of 6-MP/6-TG therapy in childhood leukemia.

4. MONITORING OF THIOPURINE THERAPY The half-lives of 6-MP and 6-TG are 21 min and 90 min, respectively, and their plasma areas under the curve are highly variable and not strongly predictive of therapeutic response or side effects ( 107, 108, 109, 110, 111). It has been suggested that the major cytotoxic effect of thiopurine drugs is dependent on the intracellular accumulation of 6-TGN and 6-MMPR where on average a steady-state concentration is reached within 1–4 weeks after initiation of treatment (112). In individuals deficient of TPMT activity, 6-TGN accumulate more rapidly in RBC (113) with an 8- to 15-fold increase of 6-TGN concentration compared with wild-type patients, subsequently leading to an exaggerated cytotoxic effect (114). RBCs have been shown to act as a useful and simple surrogate for quantifying the relevant thiopurine metabolites by various high-performance liquid chromatography (HPLC) methods ( 115, 116, 117, 118, 119, 120, 121). In RBC, 6-TGN concentrations correlate with 6-TGN incorporation into the DNA of leukocytes, the target tissue of thiopurine therapies. It is important to note that available HPLC methods are not directly comparable with regard to 6-TGN concentrations because up to 2.6-fold differences have been described (122).

Chapter 11 / Thiopurines and TPMT in Childhood ALL

181

Whether these differences also apply to other metabolites such as 6-MMPR has not been systematically analyzed. Generally, thiopurine metabolite measurements show large intra- and inter-individual variation ( 79, 110, 112, 123, 124). Targeted therapeutic ranges for 6-TGN levels under 6-MP therapy are described between 225 and 500 pmol/ 8 × 108 RBC. For 6-MMPR, levels lower than 5700 pmol/8 × 108 RBC are aimed at. For appropriate interpretation of metabolite measurements in patients on thiopurine therapy, it appears to be helpful to consider both 6-TGN and 6-MMPR levels. For instance, metabolite measurements for 6-TGN and 6-MMPR deviating similarly in the same direction suggest either under-dosing/non-compliance (at low levels) or overdosing (at high levels). High 6-TGN levels in addition to absent 6-MMPR suggest deficiency in TPMT. In contrast, low 6-TGN levels in conjunction with high 6-MMPR suggest high/very high TPMT activity. An expected ratio of 6-MMPR to 6-TGN is considered a value between 5 and 25. When 6-TG is applied, 6-TGN levels in children using a median 6-TG dose of 41.8 mg/m2 /d have been described more than 7-fold higher in comparison to those on a median 6-MP dose of 54.3 mg/m2 /d (111). Thus, expected 6-TGN concentrations under 6-TG exposure range from approximately 500 to 5000 pmol/8 × 108 RBC (111).This discrepancy may be explained by the fact that in RBC, 6-TGN are generated far more efficiently from 6-TG than from 6-MP. However there are several open questions as to whether thiopurine metabolite levels in RBC are comparable with leukocytes under treatment of childhood leukemia (125) or whether a substantial part of 6-TGN measured in RBC under 6-MP therapy derives from the metabolism in other tissues via a subsequent uptake in RBC (126). Altogether, no consensus recommendations on therapeutic ranges for 6-MP and 6-TG metabolite monitoring in treatment of childhood ALL are currently available. Moreover, 6-TGN levels may be biased in individuals who have received repeated RBC transfusions due to high variability in TPMT activity of transfused donor RBC. Besides monitoring of patient compliance, therapeutic drug monitoring of 6-TG therapy has not proved to be helpful yet in predicting treatment efficacy or toxicity. In contrast, 6-TGN concentrations in RBC have been associated with efficacy and toxicity of 6-MP treatment in children with ALL (79,110,112,123,124). In their pioneering study of 172 children with ALL on 75 mg/m2 /d oral 6-MP, Lilleyman and Lennard reported on median 6-TGN concentrations of 284 pmol/8 × 108 RBC (range 113–1340 pmol/ 8 × 108 RBC) ( 123). Elevated 6-TGN levels were associated with neutropenia and patients with 6-TGN levels below the median had a 2.5-times higher risk of relapse compared to those with values above the median. Similar observations on 6-TGN levels alone or in combination with methotrexate polyglutamate levels have been made in other studies ( 124, 127). In contrast, a study from St. Jude Childrens’ Research Hospital by Relling and colleagues did not find a significant association of 6-TGN levels with treatment outcome (79). In this study, the observed median 6-TGN level of 401 pmol/8 × 108 RBC (range 0–1498 pmol/8 × 108 RBC) was substantially higher than in the study of Lilleyman and Lennard. As an explanation for that, Relling and colleagues suggested that either the majority of patients in the St. Jude study achieved active thiopurine metabolite concentrations which exceeded a minimum threshold value for efficacy or the previously described association of 6-TGN with outcome may have been more indicative of good compliance with chemotherapy

182

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

than with a minimum required level of exposure to thiopurines. Also, a higher dose of methotrexate (40 IV vs. 20 mg oral/m2 /week) used in the St. Jude study may have been more effective at inhibiting purine de novo synthesis, resulting in an increased availability of intracellular phosphorylated thiopurines. Overall, only sparse experience with prospective evaluations on 6-TGN-guided thiopurine therapy in treatment of childhood ALL exists. Based on the observation of 6-TGN and methotrexate polyglutamate levels being related to the cure rate of childhood ALL in their previous trial, Schmiegelow and colleagues in the Scandinavian NOPHO ALL-92 trial randomly assigned 538 children with ALL to have their oral 6-MP and methotrexate maintenance therapy adjusted by leukocyte counts, RBC 6-TGN and methotrexate polyglutamate levels (pharmacology group), or by leukocyte counts only (control group) (128). In this study, RBC 6-TGN levels were only of significance for patients who relapsed while receiving therapy. Unexpectedly, girls in the pharmacology group had a significantly increased relapse risk (19% vs. 5% in the control group) because of an increased relapse hazard during the first year off therapy. Therefore, Schmiegelow and colleagues suggested that merely securing high RBC 6-TGN levels may not be sufficient to reduce the risk of relapse. This is in line with the above-described randomized trials on 6-MP vs. 6-TG for the treatment of childhood ALL, COALL-92, and UK MRC ALL 97, which were based on the observations of higher RBC 6-TGN levels being positively associated with outcome and higher 6-TGN levels being achievable by the application of 6-TG in comparison with 6-MP but failing to detect an advantage for 6-TG (86,87). When interpreting these results, it should be taken into consideration that RBC 6-TGN levels can act only as a surrogate marker for the situation in the leukemic blast and may not ideally reflect incorporation of 6-TGN into the DNA and RNA of target cells. This process is likely to be influenced by other thiopurine metabolites such as 6-MMPR and their effects on incorporation and cytotoxicity through inhibition of de novo purine synthesis (101,103). Final results from larger studies on childhood ALL that have accessed 6-TGN and other metabolites (e.g., 6-MMPR) are not available yet. The need of more comprehensive evaluations of thiopurine metabolites is underscored by Schmiegelow and colleagues’ finding of TPMT activity being the most significant predictor of relapses among girls on the pharmacology arm in the NOPHO ALL-92 trial (128). A potentially relevant association of 6-MP dosing and its metabolite distribution was suggested recently by data from the Pediatric Oncology Group 9605 trial (129). In this trial, daily 6-MP was compared with twice-daily administration during maintenance therapy of standard-risk childhood ALL. RBC 6-TGN and methylated metabolite levels were measured in a randomly selected subset of patients. One hundred eighteen patients received mercaptopurine 75 mg/m2 /d and 108 received 37.5 mg/m2 /twice daily. In analyses adjusted for actual dose, days on study, age at diagnosis, absolute leukocyte count, gender, and race, daily dosing resulted in significantly higher average methylated mercaptopurine metabolites and a trend toward higher average erythrocyte 6-TGN. These results indicate that dosing may provide a strategy for optimization of 6-MP treatment in maintenance, and it will be interesting to see these data in association with treatment outcome after an adequate follow-up has been reached for the analyzed patient population.

Chapter 11 / Thiopurines and TPMT in Childhood ALL

183

5. THIOPURINE S-METHYLTRANSFERASE Another approach to optimize pharmacotherapy of leukemia with thiopurines is phenotyping and/or genotyping of TPMT. TPMT is a ubiquitously expressed cytosolic methyltransferase that catalyzes the S-methylation of aromatic and heterocyclic sulfhydryl compounds ( 130, 131). Neither the biological function of TPMT nor its endogenous substrates are known. In thiopurine-treated patients, the activity of TPMT determines the balance between 6-TGN and 6-MMPR. As already described above, high 6-TGN levels in combination with low 6-MMPR suggest low TPMT activity, while low 6-TGN levels in combination with high 6-MMPR suggest high TPMT activity. In 1980, Weinshilboum and Sladek first reported on a trimodal frequency distribution of TPMT activity in 298 randomly selected blood donors ( 132). In subsequent segregation analysis of selected families, the same authors showed that 66% of the observed total variance in RBC TPMT activity was mono-genetically inherited ( 133). In Weinshilboum and Sladek’s study, 88.6% of the randomly selected study subjects were homozygous for the trait of high TPMT activity (TPMTH /TPMTH ), 11.1% were heterozygous (TPMTH /TPMTL ), and one out of 298 was homozygous for the trait with low or undetectable TPMT activity (TPMTL / TPMTL ) (132). This distribution of TPMT activity in Caucasian populations has been subsequently confirmed by other studies (134,135). However, TPMT activity was also shown to differ between different ethnic populations (136). For example, African Americans were demonstrated to have lower enzyme activity in comparison with Americans of Caucasian descent. Other factors that have been variably associated with TPMT activity include age (higher in children, especially neonates), gender (slightly higher in men), smoking (higher in smokers), and thiopurine treatment (137,138,139,140). The increased TPMT activity during thiopurine administration in children with ALL decreased to levels in the normal range again after treatment was stopped ( 141). Of importance, three additional issues with regard to TPMT activity and thiopurine therapy in leukemic patients have to be addressed here. First, children with leukemia frequently may require RBC transfusions due to bone marrow insufficiency inherent with the disease. Transfused donor RBCs may bias TPMT activity measurements, and therefore TPMT phenotyping should be performed earliest after 8 weeks upon receiving RBC tranfusions. Second, in patients with leukemia, TPMT activity may also be biased by an older age of RBC. In the course of leukemia, RBC production decreases, leading to relatively older RBC populations with lower TPMT activity. Third, physicians should be aware that some drugs that may be co-administered may have a potential influence on TPMT activity. Such interactions with TPMT activity have been described in ex vivo experiments for aminosalicylates and furosemide ( 142, 143,144). TPMT activity can be analyzed by different phenotyping assays including the radiochemical method developed by Weinshilboum and more recent non-radioactive HPLC methods using either 6-MP or 6-TG as substrates ( 145, 146, 147, 148, 149,150,151,152,153). The radiochemical and HPLC assays have been shown to lead to comparable results with 6-MP as a substrate (146,147,148,149,150). However, when using 6-TG, TPMT activity was measured at 30% higher levels (152).

184

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

6-TG as a substrate offers the advantage of more sensitive and specific quantitation of TPMT activity through measurements of the highly fluorescent TPMT product 6-MTG (140,151). In RBC, TPMT activity correlates well with TPMT activity in lymphoblasts (154). TPMT activity shows large intra- and inter-individual variations. Using an HPLC assay and 6-TG as a substrate, the most comprehensive observation so far on individuals of similar ethnic origin (1214 healthy German Caucasians), identified cut-off values of > 22, 3–22, and ≤ 2 nmol/(g–1 hemoglobine × h–1 ), respectively, to distinguish TPMT wild-type (TPMTH /TPMTH ), heterozygous (TPMTH /TPMTL ), and homozygous mutant individuals (TPMTL /TPMTL ) from each other (140). Within individuals characterized as TPMTH /TPMTH , a subgroup of ultra-rapid metabolizers can be described by a cut-off TPMT activity level of > 50 nmol/(g–1 hemoglobine × h–1 ). However, reported values in the literature may differ depending on the method used for measuring TPMT activity and the population under investigation. In analogy with RBC 6-TGN levels, the comparison of TPMT activity between independent studies is complicated by different units of measurement [e.g., nmol/(mL-1 RBC × h–1 ) or nmol/(g-1 hemoglobine × h1 ] (146,147,148,149,150,151). Weinshilboum and Sladek’s observation of TPMT activity being monogenetically inherited was based on phenotypic measurement of TPMT and was subsequently confirmed by cloning and characterizing the gene coding for TPMT ( 155). The TPMT gene has been localized to chromosome 6p22.3 and encodes a 245 amino acid protein with a predicted molecular mass of 35 kD ( 155). The first report of the TPMT gene described a 34 Kb gene with 10 exons and a start codon in exon 3 ( 156). However, in the literature, exon 2 has only been reported once in human cDNA, and two later studies described a smaller-sized gene of 25 kb and 27 kb, respectively, containing only 9 exons ( 157, 158). To date, 24 mutant alleles (TPMT*2, *3A, *3B, *3C, *3D, *4, *5, *6, *7, *8, *9, *10, *11, *12, *13, *14, *15, *16, *17, *18, *19, *20, *21, *22) responsible for variation in TPMT enzyme activity have been described (Table 1) (159,160,161,162,163,164,165,166,167,168,169,170,171). Most of these alleles are characterized by one or more single nucleotide polymorphisms (SNPs) in the coding sequence of the TPMT gene (non-synonymous SNP), resulting in decrease or loss of enzyme activity. Moreover two variants that influences mRNA splicing (TPMT*4 and *15) are described. The distribution of TPMT mutant alleles differs significantly among ethnic populations. In the Caucasian population, TPMT*3A, *2, and *3C are the most frequently observed variant alleles which, altogether, account for more than 95% of variant alleles (Table 2) (64,170). In Asian and African populations, TPMT*3C is the most frequent variant allele (171,172). Comprehensive work was done to investigate the functional consequences of TPMT variants and recently alleles including 13 non-synonymous TPMT SNPs were transiently expressed in COS-1 cells and enzyme activity and protein quantity were determined (173). Previously, laboratory experiments demonstrated that the expression of TPMT*3A, *3B, or *3C transfected into COS-1 or yeast cells results in a decrease of enzyme activity and protein expression (156,160). Loss of activity in these experiments was highest for TPMT*3A, followed by TPMT*3B and *3C. Although there were changes in substrate kinetics, the functional effects resulted primarily from alterations in level of enzyme

Chapter 11 / Thiopurines and TPMT in Childhood ALL

185

Table 1 Phenotypically Relevant Variant TPMT Alleles Reported in the Literature Alleles

Exon

Mutation

Amino acid change

Reference

*1S *2 *3A

210a 159 156, 160

*5 *6 *7 *8 *9 *10 *11 *12 *13 *14 *15

4 8 10 10 5 7 6 6 3 3 —

*16 *17 *18 *19 *20 *21 *22

7 3 4 5 10 4 7

474T>C 238 G>C 460 G>A 719A>G 460 G>A 719A>G 292 G>T 460 G>A 719A>G G>A transition that disrupts the intron/exon acceptor splice junction at the final 3 nucleotide of intron 9 146T>C 538A>T 681T>G 644 G>A 356A>C 430 G>C 395 G>A 374C>T 83A>T 1A>G G>A transition that disrupts the intron/exon acceptor splice junction at the final 3 nucleotide of intron 7 488 G>A 124C>G 211 G>A 365A>C 712A>G 205C>G 488 G>C

— Ala80 Pro Ala154 Thr Tyr240 Cys Ala154 Thr Tyr240 Cys Glu98 X Ala154 Thr Tyr240 Cys

*4

7 5 7 10 7 10 5 7 10 —

*3B *3C *3D

a

156, 160 156, 160 161

161

Leu49 Ser Tyr180 Phe His227 Gln Arg215 His Lys119 Thr Gly144 Arg Cys132 Tyr Ser125 Leu Glu28 Val Met1 Val

161 161 162 163 140 164 165 166 166 167 167

Arg163 His Gln42 Glu Gly71 Arg Lys122 Glu Lys238 Glu Leu69 Val Arg163 Pro

140 140 140 168 169 169 169

Frequently observed silent change, reported allele frequency 0.215.

protein. Using pulse chase experiments performed with cultured mammalian cells as well as experiments performed with rabbit reticulocyte lysate, it was demonstrated for example that the common TPMT*3A variant allozyme was degraded much more rapidly than the TPMT wild-type enzyme through a ubiquitin-proteasome-mediated process

186

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Table 2 Population Frequencies of Clinically Relevant TPMT Variant Alleles in Different Ethnic Groups (64,170,171,172) Population

Caucasian South American African African American Asian

TPMT Allele Frequencies (%) *2

*3A

*3C

0.00–0.50 0.30–2.20 0.00 0.40 0.00

2.50–5.70 1.50–3.60 0.00 0.80 0.00–1.00

0.00–3.80 0.00–2.54 7.60–10.10 2.40 0.00–3.00

( 174, 175, 176). Moreover, it was shown that chaperone proteins, especially hsp90, are involved in targeting TPMT*3A ( 176) and very recently the assumption that the TPMT*3A polymorphisms might result in misfolding and protein aggregation with aggresome formation was elucidated (177). Besides the TPMT*2, *3A, *3B, and *3C alleles, most other phenotypically relevant variant alleles have only been reported in single patients. In addition, several intronic mutations and mutations outside of the open reading frame have been described ( 140, 178). Furthermore, a variable number tandem repeat (VNTR) within a GC-rich area in the 5 -flanking region of the TPMT gene has been reported to modulate levels of TPMT activity (179,180). TPMT VNTR length was described to vary between three and nine repeats (VNTR*V3 to *V9) with VNTR*V4 and *V5 being the most frequent of these17-bp or 18-bp repeats. However, neither strong nor consistent associations between the number of tandem repeats and TPMT activity can be demonstrated yet (181,182,183). TPMT genotyping is complicated by a processed pseudogene located on chromosome 18q21.1 and shows 96% homology to the TPMT coding sequence ( 184). Because this pseudogene is due to genomic integration of TPMT mRNA upon reverse transcription, it is important to choose primers for TPMT genotyping that extend throughout the exon boundaries ( 178). TPMT genotyping can be performed by different assays including restriction fragment length polymorphism analysis after polymerase chain reaction (RFLP-PCR), denaturing HPLC, real-time PCR analysis, molecular haplotyping, a multiplex amplification refractory mutation system (ARMS) strategy, arrayed primer extension (APEX), pyrosequencing, DNA microarray, and so on (185,186,187,188,189,190,191,192,193). The RFLP-PCR method is especially prone to analytical pitfalls due to potentially incomplete restriction enzyme digestion and requires strict quality control measures (194). Of importance, the TPMT*3A allele is conventionally analyzed by separate genotyping of the two involved SNPs, 460 G>A and 719A>G, and therefore it cannot distinguish if the two variant nucleotides are present in cis or trans (on two different alleles). Thus, this approach is inherent with the risk of misclassifying an individual compound heterozygous for TPMT*3B (460 G>A) and TPMT*3C (719A>G), and therefore is deficient in TPMT activity, as heterozygous for a TPMT*3A allele. Although the TPMT*3B allele

Chapter 11 / Thiopurines and TPMT in Childhood ALL

187

is rare, misclassification can be overcome by applying techniques for TPMT genotyping that take this problem into account (e.g., molecular haplotyping by long-range genomic PCR to cover the approximately 9 kb difference between the two SNPs and intramolecular ligation) (190). TPMT genotype can be used as a surrogate marker of TPMT activity. In several independent studies, TPMT genotype showed excellent concordance with TPMT phenotype. For example, Yates and colleagues analyzed the TPMT phenotype in 282 unrelated Caucasian Americans (170). Subsequently, all individuals phenotypically deficient or with intermediate TPMT activity and a randomly selected sample of individuals with high TPMT activity were TPMT genotyped (TPMT*2 and *3A, *3B, *3C). In this study, 21 patients had a heterozygous TPMT phenotype. With a frequency of 85%, TPMT*3A was the most prevalent variant allele, followed by TPMT*2 and TPMT*3C with about 5% each. All 6 patients who phenotypically displayed TPMT deficiency had two mutant alleles; 20 of the 21 patients with intermediate TPMT activity had one variant allele; and all of the selected 21 patients with high activity did not carry one of the tested TPMT variant alleles. Thus, the major inactivating TPMT variants can be detected reliably by a PCR-based method and demonstrated an excellent concordance with TPMT phenotype. Coulthard and colleagues analyzed the relationship between TPMT phenotype measured directly in the target of drug action, the leukemic cell, and TPMT genotype (195). They demonstrated that the TPMT activity in lymphoblasts from 38 children and adults homozygous for TPMT*1 was significantly higher than that in the five genotypically heterozygotes. Of interest, in the same study, a comparison of activity in blasts from AML and ALL showed a higher level in AML, and therefore suggests that factors other than genotype may also have an influence on TPMT expression. The most comprehensive analysis of TPMT phenotype versus genotype published to date was conducted by Schaeffeler et al. ( 140). In their study, RBC TPMT activity and genotype (TPMT*2 and *3 alleles) were analyzed in 1214 healthy Caucasian blood donors. Discordant cases between phenotype and genotype were systematically sequenced. The frequencies of the mutant alleles were 4.4% for TPMT*3A, 0.4% for TPMT*3C, and 0.2% for TPMT*2. All seven TPMT-deficient subjects identified by Schaeffeler and colleagues were homozygous or compound heterozygous carriers for these alleles. In 17 individuals with intermediate TPMT activity discordant to TPMT genotype, four novel genetic variants were identified (TPMT*9, *16, *17, and *18) leading to amino acid changes. Taking these additionally discovered variants into consideration, the overall concordance rate between TPMT phenotype and genotype was 98.4%. Specificity, sensitivity, and the positive and negative predictive power of the genotyping test were estimated to be higher than 90%. Thus, solid data support that TPMT phenotype can be predicted by molecular diagnostics. This information is of particular importance to leukemia patients, as phenotypic assessment of TPMT activity in these patients may be rendered unreliable by prior RBC transfusion. Most studies in leukemia patients have used RBC as a surrogate tissue for phenotyping TPMT activity, and several of them have shown that childhood leukemia patients with TPMTL /TPMTL phenotypes are at high risk of developing severe hematologic toxicity after treatment with standard doses of thiopurines (196,197,198). Thus, dose adjustment

188

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

at least in TPMT deficient patients is required with an initial dose reduction to 10%–15% of the standard dose of 6-MP as exemplarily shown for ALL patients (199,200) as well as patients with Crohn’s disease (201). However, it was also demonstrated that TPMT phenotype or genotype influences the effectiveness of therapy. Low TPMT activity has been associated with higher 6-TGN levels and improved survival, while high TPMT activity has been associated with lower 6-TGN concentrations and an increased relapse risk (123,127,141). The most comprehensive studies on thiopurines and their metabolism in the field of childhood ALL have been conducted since the early 1990s at St. Jude Childrens’ Research Hospital (67,68,69,79,198,202,203,204,205). As a consequence of their findings on the prognostic importance of 6-MP dose intensity in childhood ALL (e.g., avoidance of thiopurine treatment interruption resulted in fewer relapses), Relling and colleagues at St. Jude used measurements of TPMT activity and thiopurine metabolites, as well as clinical tolerance to maintenance therapy, to guide treatment. They selectively decreased the dose of 6-MP without modification of concurrent chemotherapeutic co-medication in patients with low or intermediate TPMT activity or increased doses in patients demonstrating persistently high absolute leukocyte counts. Following such a strategy in the St. Jude study, Total XIIIB, TPMT genotype was no longer predictive of hematologic relapse (5-year cumulative incidences of 13.2% vs. 6.7% among patients homozygous for TPMT*1 vs. TPMT genotypes conferring low enzyme activity, respectively) (205). This important finding of Relling and colleagues at St. Jude points to the potential of individualized thiopurine dosing strategies for improving outcome in childhood ALL. In the above-mentioned NOPHO ALL-92 study on treatment of children with precursor B-cell ALL, 6-MP and methotrexate maintenance therapy was adjusted by leukocyte counts, RBC 6-TGN, and methotrexate levels (pharmacology group), or by leukocytes only (control group) (128). Unexpectedly, girls in the pharmacology group had a significantly increased relapse risk (19% vs. 5% in the control group) because of an increased relapse hazard during the first year off therapy. TPMT activity was the strongest predictor of risk of relapse for girls in the pharmacology group, and girls who relapsed off therapy had higher TPMT activity than those who did not relapse, although this was not the case for girls relapsing on therapy. Schmiegelow and colleagues speculated that dose escalation of 6-MP may lead to increased intracellular levels of methylated metabolites, such as 6-MMPR, which inhibit purine de novo synthesis in leukemic lymphoblasts of patients with high TPMT activity and therefore, lead to a delay of these cells in S phase under maintenance treatment and potential regrowth after treatment discontinuation. With relevance to the administration of thiopurines in the early course of childhood ALL, the BFM Study Group reported on the association of TPMT genotype and minimal residual disease (MRD) in 810 children with childhood ALL enrolled into their trial ALL-BFM 2000 (206). In this trial, DNA-based MRD analysis after induction and after consolidation treatment was used for risk-adapted treatment stratification. A 4-week cycle of 6-MP was applied in-between these two MRD measurements. In patients homozygous for the TPMT*1 allele or those heterozygous for a variant TPMT allele, MRD levels on treatment day 33 were equally distributed between the groups. However, when MRD levels were assessed on treatment day 78, after administration of consolidation

Chapter 11 / Thiopurines and TPMT in Childhood ALL

189

treatment, including a 4-week cycle of 6-MP at a dose of 60 mg/m2 /d, significant differences with regard to clearance of MRD were observed between TPMT wild-type and heterozygous patients. For heterozygous patients, this distribution translated into a 2.9-fold reduction in risk of having measurable MRD after induction consolidation treatment (relative risk = 0.34; 95% CI = 0.13 – 0.86; p = 0.02). This point estimate did not significantly change in multivariate analysis including variables known to be associated with treatment response. Besides supporting an important role for 6-MP dose intensity, the BFM data demonstrate a substantial impact of TPMT genotype on MRD after administration of 6-MP in the early course of childhood ALL and may provide a rationale for genotype-based adaptation of 6-MP dosing in the early course of childhood ALL, provided that the described observations translate into an improved long-term outcome. With regard to long-term toxicity, some reports in the literature have suggested a relationship of secondary malignancies after treatment of childhood ALL with TPMTH /TPMTL and TPMTL /TPMTL phenotypes. In a study at St. Jude Children’s Research Hospital, SJCRH Total XIIIHR, patients with lower TPMT activity showed a trend towards a higher incidence of AML associated with application of the topoisomerase II inhibitor etoposide (203). A second study conducted at St. Jude, Total Therapy Study XII, reported on a higher incidence of brain tumors in childhood ALL patients with lower TPMT activity who had received cranial radiotherapy concurrent with 6-MP treatment in the initial maintenance phase (204). Six out of 52 patients receiving cranial radiotherapy with concurrent 6-MP treatment developed a brain tumor. Of these 6 patients, 4 had RBC 6-TGN levels above the 70th percentile for the entire cohort of 52 patients, and 3 patients were heterozygous for TPMT. The 8-year cumulative incidence of brain tumors among children with low TPMT activity was 42.9% (standard error 20.6) versus 8.3% (standard error 4.7) in TPMT wild-type patients. In the Scandinavian NOPHO ALL-92 trial, Thomsen and colleagues reported on a significantly higher risk of therapy-associated AML or myelodysplastic syndrome in patients with low TPMT activity and high erythrocyte 6-TGN levels and/or 6-MMPR (207). In contrast, successful TPMT genotyping of 72 patients out of a total of 115 patients with subsequent secondary malignant neoplasms after treatment for childhood ALL on seven consecutive BFM protocols (ALL-BFM 79, 81, 83, 86, 90, 95, and 2000) did not reveal a higher frequency of TPMT alleles associated with lower TPMT activity among these patients (208). Also, in stratified analyses by entities of secondary malignant neoplasms, no significant associations with TPMT alleles conferring lower enzyme activity have been observed. The main contrast between the St. Jude Children’s Research Hospital protocol and NOPHO protocols in comparison to BFM protocols for treatment of childhood ALL are that 6-MP starting doses in maintenance are lower in the latter ones (50 vs. 75 mg/m2/d). In addition, on BFM protocols the dose of 6-MP given concurrent with high-dose methotrexate is significantly lower (25 vs. 75 mg/m2/d). Whereas on the NOPHO ALL-92 protocol patients did not regularly receive cranial radiotherapy, 6-MP application concurrent with cranial irradation during early maintenance was lower on BFM compared to St. Jude protocols (50 vs. 75 mg/m2/d). Other differences of potential importance to this issue, which only apply to the comparison of

190

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

BFM and St. Jude protocols, are that on BFM protocols no topoisomerase II inhibitors are given in close association with thiopurines and, finally, that on BFM protocols no intrathecal triple therapy (methotrexate, cytarabine, and a glucocorticoid) are given concurrent with cranial radiotherapy and 6-MP. Another important toxicity issue associated with TPMT status relates to the above-described VOD-like symptoms of the liver in childhood ALL patients treated with 6-TG on the British MRC ALL97 trial ( 209). In this trial, TPMT activity was significantly lower in children in whom VOD developed while no differences in RBC 6-TGN levels were described. This information in association with ongoing research efforts will help to develop a better understanding of 6-TG-associated liver toxicity and may help to identify those individuals upfront who should not be administered 6-TG.

6. CONCLUSIONS AND FURTHER PERSPECTIVES Our review of thiopurines in the treatment of childhood ALL and genetic variants of the TPMT gene summarizes the current status and gives some clues in addition to the many and varied questions that remain to be answered. Despite considerable insights into thiopurine pharmacology that have been gained lasting recent decades leading to the development of novel strategies for improving efficacy and reducing toxicity of 6-MP as well as 6-TG administration in several institutional and collaborative ALL study groups, the validation of this information and its potential translation into clinical practice is still an ongoing process. Most of the major institutional and collaborative study groups on treatment of childhood ALL extensively assess prognostic factors on a regular basis in prospective trials. It can be extremely helpful for advancing our current knowledge on thiopurines in childhood ALL, if the evaluation of data on thiopurine response would be included in these studies on a broader basis. Initially, collection and comparison of information from different large clinical trial populations of children with ALL exposed to different background as well as thiopurine treatment may truly help to more clearly define the critical issues associated with thiopurine therapy. As a consequence, relevant results can be evaluated prospectively in future clinical trials. Clearly, the success of such potential cooperative ventures depends on the quality of interaction among the leading institutional and collaborative study groups. Moreover, it is necessary to arrive at a consensus for generating tools needed to develop a more detailed uniform approach for an improved assessment of thiopurine treatment in childhood ALL.

REFERENCES 1. Pui CH, Evans WE. Treatment of childhood acute lymphoblastic leukemia. N Engl J Med 2006;354:166–178. 2. Schrappe M, Reiter A, Zimmermann M et al. Long-term results of four consecutive trials in childhood ALL performed by the ALL-BFM study group from 1981 to 1995. Leukemia 2000;14:2205–2222. 3. Silverman LB, Sallan SE. Newly diagnosed childhood acute lymphoblastic leukemia: update on prognostic factors and treatment. Curr Opin Hematol. 2003;10:290–296. 4. Pinkel, D. Five-year follow-up of “Total Therapy” of childhood lymphocytic leukemia. JAMA 1971;216:648–652.

Chapter 11 / Thiopurines and TPMT in Childhood ALL

191

5. Riehm H, Gadner H, Henze G et al. The Berlin childhood acute lymphoblastic leukemia therapy study, 1970–1976. Am J Pediatr Hematol Oncol 1980;2:299–306. 6. Rivera GK, Raimondi SC, Hancock ML et al. Improved outcome in childhood acute lymphoblastic leukaemia with reinforced early treatment and rotational combination chemotherapy. Lancet 1991;337:61–66. 7. Gaynon PS, Steinherz PG, Bleyer WA et al. Improved therapy for children with acute lymphoblastic leukemia and unfavorable presenting features: a follow-up report of the Childrens Cancer Group Study CCG-106. J Clin Oncol 1993;11:2234–2242. 8. Reiter A, Schrappe M, Ludwig WD et al. Chemotherapy in 998 unselected childhood acute lymphoblastic leukemia patients: results and conclusions of the multicenter trial ALL-BFM 86. Blood 1994;84:3122–3133. 9. Chessells JM, Bailey C, Richards SM. Intensification of treatment and survival in all children with lymphoblastic leukaemia: results of UK Medical Research Council trial UKALL X. Medical Research Council Working Party on Childhood Leukaemia. Lancet 1995;345:143–148. 10. Conter V, Arico M, Valsecchi MG et al. Intensive BFM chemotherapy for childhood ALL: interim analysis of the AIEOP-ALL 91 study. Haematologica 1998;83:791–799. 11. Evans WE, Relling MV, Rodman JH et al. Conventional compared with individualized chemotherapy for childhood acute lymphoblastic leukemia. N Engl J Med 1998;338:499–505. 12. Gustafsson G, Kreuger A, Clausen N et al. Intensified treatment of acute childhood lymphoblastic leukaemia has improved prognosis, especially in non-high-risk patients: the Nordic experience of 2648 patients diagnosed between 1981 and 1996. Nordic Society of Pediatric Haematology and Oncology (NOPHO). Acta Paediatr 1998;87:1151–1161. 13. Kamps WA, Bokkerink JP, Hahlen K et al. Intensive treatment of children with acute lymphoblastic leukemia according to ALL-BFM-86 without cranial radiotherapy: results of the Dutch Childhood Leukemia Study Group protocol ALL-7 (1988–91). Blood 1999;94:1226–1236. 14. Sackmann-Muriel F, Felice MS, Zubizarreta PA et al. Treatment results in childhood acute lymphoblastic leukemia with a modified ALL-BFM 90 protocol: lack of improvement in high-risk group. Leuk Res 1999;23:331–340. 15. Schrappe M, Reiter A, Ludwig WD et al. Improved outcome in childhood acute lymphoblastic leukemia despite reduced use of anthracyclines and cranial radiotherapy: results of trial ALL-BFM 90. Blood 2000;95:3310–3322. 16. Henze G, Langermann HJ, Kaufmann U et al. Thymic involvement and initial white blood count in childhood acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1981;3:369–376. 17. Riehm H, Gadner H, Henze G. Acute lymphoblastic leukemia: treatment results in three BFM studies (1970–1981). In: Leukemia research: advances in cell biology and treatment. (Murphy SB, Gilbert JR, eds.), Amsterdam: Elsevier Science Publishing, 1983:251–163. 18. Pinkel D, Woo S. Prevention and treatment of meningeal leukemia in children. Blood 1994;84: 355–366. 19. Clarke M, Gaynon P, Hann Is et al. CNS-directed therapy for childhood acute lymphoblastic leukemia: childhood group collaborative group overview of 43 randomized trials. J Clin Oncol 2003;21: 1798–1809. 20. Bene MC, Castoldi G, Knapp W et al. Proposals for the immunological classification of acute leukaemias. European group for the immunological characterisation of leukaemias (EGIL). Leukemia 1995;9:1783–1786. 21. Hiddemann W, Wormann B, Ritter J et al. Frequency and clinical significance of DNA aneuploidy in acute leukemia. Ann NY Acad Sci 1986;468:227–240. 22. Pui CH, Williams DL, Raimondi SC et al. Hypodiploidy is associated with a poor prognosis in childhood acute lymphoblastic leukemia. Blood 1987;70:247–253. 23. Fletcher JA, Lynch EA, Kimball VM et al. Translocation t(9;22) is associated with extremely poor prognosis in intensively treated children with acute lymphoblastic leukemia. Blood 1991;77:435–439.

192

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

24. Trueworthy R, Shuster J, Look T et al. Ploidy of lymphoblasts is the strongest predictor of treatment outcome in B-progenitor cell acute lymphoblastic leukemia of childhood: a Pediatric Oncology Group Study. J Clin Oncol 1992;10:606–613. 25. Harris MB, Shuster JJ, Carroll A et al. Trisomy of leukemic cells chromosomes 4 and 10 identifies children with B-progenitor cell acute lymphoblastic leukemia with a very low risk of treatment failure: a Pediatric Oncology Group study. Blood 1992;79:3316–3324. 26. Harbott J, Ritterbach J, Ludwig W-D et al. Clinical significance of cytogenetic studies in childhood acute lymphoblastic leukemia: experience of the BFM trials. Recent Results in Cancer Research 1993;131:123–132. 27. Pui CH, Rivera GK, Hancock ML et al. Clinical significance of CD10 expression in childhood acute lymphoblastic leukemia. Leukemia 1993;7:35–40. 28. Ludwig W-D, Harbott J, Bartram CR et al. Incidence and prognostic significance of immunophenotypic subgroups in childhood acute lymphoblastic leukemia: experience of the BFM study 86. In: Recent Advances in Cell Biology of Acute Leukemia: Impact on Clinical Diagnosis and Therapy (Ludwig W-D, Thiel E, eds.), Berlin: Springer Verlag, 1993:269–282. 29. Rubnitz JE, Link MP, Shuster JJ et al. Frequency and prognostic significance of HRX rearrangements in infant acute lymphoblastic leukemia: a Pediatric Oncology Group study. Blood 1994;84:570–573. 30. Behm FG, Raimondi SC, Frestedt JL et al. Rearrangement of the MLL gene confers a poor prognosis in childhood acute lymphoblastic leukemia, regardless of presenting age. Blood 1996;87:2870–2877. 31. Schlieben S, Borkhardt A, Reinisch I et al. Incidence and clinical outcome of children with BCR/ABLpositive acute lymphoblastic leukemia (ALL): a prospective RT-PCR study based on 673 patients enrolled in the German pediatric multicenter therapy trials ALL-BFM 90 and CoALL-05-92. Leukemia 1996;10:957–963. 32. Borkhardt A, Cazzaniga G, Viehmann S et al. Incidence and clinical relevance of TEL/AML1 fusion genes in children with acute lymphoblastic leukemia enrolled in the German and Italian multicenter therapy trials. Associazione Italiana Ematologia Oncologia Pediatrica and the Berlin–Frankfurt– Munster Study Group. Blood 1997; 90:571–577. 33. Trka J, Zuna J, Hrusak O et al. Impact of TEL/AML1-positive patients on age distribution of childhood acute lymphoblastic leukemia in Czech Republic. Leukemia 1998;12:996–1007. 34. Loh ML, Silverman LB, Young ML et al. Incidence of TEL/AML1 fusion in children with relapsed acute lymphoblastic leukemia. Blood 1998;15:4792–4797. 35. Uckun FM, Sensel MG, Sun L et al. Biology and treatment of childhood T-lineage acute lymphoblastic leukemia. Blood 1998;91:735–746. 36. Heerema NA, Nachman JB, Sather HN et al. Hypodiploidy with less than 45 chromosomes confers adverse risk in childhood acute lymphoblastic leukemia: a report from the Childrens’ Cancer Group. Blood 1999;94:4036–4045. 37. Pui CH, Boyett JM, Relling MV et al. Sex differences in prognosis for children with acute lymphoblastic leukemia. J Clin Oncol 1999;17:818–824. 38. Gajjar A, Harrison PL, Sandlund JT et al. Traumatic lumbar puncture at diagnosis adversely affects outcome in childhood acute lymphoblastic leukemia. Blood 2000;96:3381–3384. 39. Breit S, Stanulla M, Flohr T et al. Activating NOTCH1 mutations predict favorable early treatment response and long-term outcome in childhood precursor T-cell lymphoblastic leukemia. Blood 2006;108:1151–1157. 40. Riehm H, Reiter A, Schrappe M et al. Die Corticosteroid-abh¨angige Dezimierung der Leuk¨amiezellzahl im Blut als Prognosefaktor bei der akuten lymphoblastischen Leuk¨amie im Kindesalter (Therapiestudie ALL-BFM 83). (The in vivo response on corticosteroid therapy as an additional prognostic factor in childhood acute lymphoblastic leukemia (therapy study ALL-BFM 83). Klin P¨adiatr 1987;199:151–160. 41. Gaynon PS, Bleyer WA, Steinherz PG et al. Day 7 marrow response and outcome for children with acute lymphoblastic leukemia and unfavorable presenting features. Med Pediatr Oncol 1990;18: 273–279.

Chapter 11 / Thiopurines and TPMT in Childhood ALL

193

42. Janka-Schaub GE, St¨uhrk H, Kort¨um B et al. Bone marrow blast count at day 28 as the single most important prognostic factor in childhood acute lymphoblastic leukemia. Haematol Blood Transfus 1992; 34:233–237. 43. Arico M, Basso G, Mandelli F et al. Good steroid response in vivo predicts a favorable outcome in children with T-cell acute lymphoblastic leukemia. Cancer 1995;75:1684–1693. 44. Gajjar A, Ribeiro R, Hancock ML et al. Persistence of circulating blasts after 1 week of multiagent chemotherapy confers a poor prognosis in childhood acute lymphoblastic leukemia. Blood 1995;86:1292–1295. 45. Steinherz PG, Gaynon PS, Breneman JC et al. Cytoreduction and prognosis in acute lymphoblastic leukaemia: the importance of early marrow response: report from the Childrens Cancer Group. J Clin Oncol 1996;14:389–398. 46. Schrappe M, Reiter A, Riehm H. Cytoreduction and prognosis in childhood acute lymphoblastic leukemia. J Clin Oncol 1996;14:2403–2406. 47. Gaynon PS, Desai AA, Bostrom BC et al. Early response to therapy and outcome in childhood acute lymphoblastic leukemia: a review. Cancer 1997;80:1717–1726. 48. Kaspers GJ, Pieters R, Van Zantwijk CH et al. Prednisolone resistance in childhood acute lymphoblastic leukemia: vitro–vivo correlations and cross-resistance to other drugs. Blood 1998;92:259–266. 49. Brisco MJ, Condon J, Hughes E et al. Outcome prediction in childhood acute lymphoblastic leukaemia by molecular quantification of residual disease at the end of induction. Lancet 1994;343:196–200. 50. Cave H, van der Werff ten Bosch J, Suciu S et al. European Organization for Research and Treatment of Cancer–Childhood Leukemia Cooperative Group. Clinical significance of minimal residual disease in childhood acute lymphoblastic leukemia. N Engl J Med 1998;339:591–598. 51. van Dongen JJ, Seriu T, Panzer-Gr¨umayer ER et al. Prognostic value of minimal residual disease in acute lymphoblastic leukaemia in childhood. Lancet 1998;352:1731–1738. 52. Ciudad J, San Miguel JF, L´opez-Berges MC et al. Prognostic value of immunophenotypic detection of minimal residual disease in acute lymphoblastic leukemia. J Clin Oncol 1998;16:3774–3781. 53. Foroni L, Harrison JC, Hoffbrand AV et al. Investigation of minimal residual disease in chidhood and adult acute lymphoblastic leukemia by molecular analysis. Br J Haematol 1999;105:7–24. 54. Coustan-Smith E, Sancho J, Hancock ML et al. Clinical importance of minimal residual disease in childhood acute lymphoblastic leukemia. Blood 2000;96:2691–2696. 55. Dworzak MN, Fr¨oschl G, Printz D et al. Austrian Berlin–Frankfurt–M¨unster Study Group. Prognostic significance and modalities of flow cytometric minimal residual disease detection in childhood acute lymphoblastic leukemia. Blood 2002;99:1952–1958. 56. Nyvold C, Madsen HO, Ryder LP et al. Nordic Society for Pediatric Hematology and Oncology. Precise quantification of minimal residual disease at day 29 allows identification of children with acute lymphoblastic leukemia and an excellent outcome. Blood 2002;99:1253–1258. 57. B¨urger B, Zimmermann M, Mann G et al. Diagnostic cerebrospinal fluid (CSF) examination in children with acute lymphoblastic leukemia (ALL): significance of low leukocyte counts with blasts or traumatic lumbar puncture. J Clin Oncol 2003;21:184–188. 58. Mastrangelo R, Poplack D, Bleyer A et al. Report and recommendations of the Rome workshop concerning poor-prognosis acute lymphoblastic leukemia in children: biologic bases for staging, stratification, and treatment. Med Pediatr Oncol 1986;14:191–194. 59. van der Does-van den Berg A, Bartram CR, Basso G et al. Minimal requirements for the diagnosis, classification, and evaluation of the treatment of childhood acute lymphoblastic leukemia (ALL) in the “BFM Family” Cooperative Group. Med Pediatr Oncol 1992;20:497–505. 60. Smith M, Arthur D, Camitta B et al. Uniform approach to risk classification and treatment assignment for children with acute lymphoblastic leukemia. J Clin Oncol 1996;14:18–24. 61. Elion GB, Burgi E, Hitchings GH. Studies on condensed pyrimidine systems, IX: the synthesis of some 6-substituted purines. J Am Chem Soc 1952;74:411–414. 62. Burchenal JH, Murphy ML, Ellison RR et al. Clinical evaluation of a new antimetabolite, 6-mercaptopurine, in the treatment of leukemia and allied diseases. Blood 1953;8:965–999.

194

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

63. Bertino JR. Improving the curability of acute leukemia: pharmacologic approaches. Semin Hematology 1991;28:9–11. 64. McLeod HL, Krynetski EY, Relling MV et al. Genetic polymorphism of thiopurine methyltransferase and its clinical relevance for childhood acute lymphoblastic leukemia. Leukemia 2000;14:567–572. 65. Bokkering J, Damen F, Huylscher M et al. Biochemical evidence for synergistic combination treatment with methotrexate and 6-mercaptopurine in childhood acute lymphoblastic leukemia. Haematol Blood Transfus 1990;33:110–117. 66. Giverhaug T, Loennechen T, Aarbakke J. Increased concentrations of methylated 6-mercaptopurine metabolites and 6-thioguanine nucleotides in human leukemic cells in vitro by methotrexate. Biochem Pharmacol 1998; 55:1641–1646. 67. Dervieux T, Brenner TL, Hon YY et al. De novo purine synthesis inhibition and antileukemic effects of mercaptopurine alone or in combination with methotrexate in vivo. Blood 2002;100:1240–1247. 68. Dervieux T, Hancock ML, Evans WE et al. Effect of methotrexate polyglutamates on thioguanine nucleotide concentrations during continuation therapy of acute lymphoblastic leukemia with mercaptopurine. Leukemia 2002;16:209–212. 69. Dervieux T, Hancock ML, Pui CH et al. Antagonism by methotrexate on mercaptopurine disposition in lymphoblasts during up-front treatment of acute lymphoblastic leukemia. Clin Pharmacol Ther 2003;73:506–516. 70. van der Werff Ten Bosch J, Suciu S, Thyss A et al. Value of intravenous 6-mercaptopurine during continuation treatment in childhood acute lymphoblastic leukemia and non-Hodgkin’s lymphoma: final results of a randomized phase III trial (58881) of the EORTC CLG. Leukemia 2005;19:721–726. 71. Tolar J, Bostrom BC, La MK, Sather HN. Intravenous 6-mercaptopurine decreases salvage after relapse in childhood acute lymphoblastic leukemia: a report from the Childrens’ Cancer Group study CCG 1922. Pediatr Blood Cancer 2005;45:5–9. 72. Ferrazzini G, Klein J, Sulh H et al. Interaction between trimethoprim-sulfamethoxazole and methotrexate in children with leukemia. J Pediatr 1990;117:823–826. 73. Beach BJ, Woods WG, Howell SB. Influence of cotrimoxazole on methotrexate pharmacokinetics in children with acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1981;3:115–119. 74. Rivard GE, Infante-Rivard C, Dresse MF et al. Circadian time-dependent response of childhood lymphoblastic leukemia to chemotherapy: a long-term follow-up study of survival. Chronobiol Int 1993;10:201–204. 75. Schmiegelow K, Glomstein A, Kristinsson J et al. Impact of morning versus evening schedule for oral methotrexate and 6-mercaptopurine on relapse risk for children with acute lymphoblastic leukemia. Nordic Society for Pediatric Hematology and Oncology (NOPHO). J Pediatr Hematol Oncol 1997;19:102–109. 76. Rivard GE, Lin KT, Leclerc JM et al. Milk could decrease the bioavailability of 6-mercaptopurine. Am J Pediatr Hematol Oncol 1989;11:402–406. 77. Riccardi R, Balis FM, Ferrara P et al. Influence of food intake on bioavailability of oral 6mercaptopurine in children with acute lymphoblastic leukemia. Pediatr Hematol Oncol 1986;3: 319–324. 78. Sofianou-Katsoulis A, Khakoo G, Kaczmarski R. Reduction in bioavailability of 6-mercaptopurine on simultaneous administration with cow’s milk. Pediatr Hematol Oncol 2006;23:485–487. 79. Relling MV, Hancock ML, Boyett JM et al. Prognostic importance of 6-mercaptopurine dose intensity in acute lymphoblastic leukemia. Blood 1999;93:2817–2823. 80. Rees CA, Lennard L, Lilleyman JS et al. Disturbance of 6-mercaptopurine metabolism by cotrimoxazole in childhood lymphoblastic leukaemia. Cancer Chemother Pharmacol 1984;12:87–89. 81. Burton NK, Aherne GW. The effect of cotrimoxazole on the absorption of orally administered 6-mercaptopurine in the rat. Cancer Chemother Pharmacol 1986;16:83–84. 82. Childhood ALL Collaborative Group. Duration and intensity of maintenance chemotherapy in acute lymphoblastic leukaemia: overview of 42 trials involving 12 000 randomised children. Lancet 1996;347:1783–1788.

Chapter 11 / Thiopurines and TPMT in Childhood ALL

195

83. Mitchell CD, Richards SM, Kinsey SE. Medical Research Council Childhood Leukaemia Working Party. Benefit of dexamethasone compared with prednisolone for childhood acute lymphoblastic leukaemia: results of the UK Medical Research Council ALL97 randomized trial. Br J Haematol 2005;129:734–745. 84. Conter V, Valsecchi MG, Silvestri D et al. Pulses of vincristine and dexamethasone in addition to intensive chemotherapy for children with intermediate-risk acute lymphoblastic leukaemia: a multicentre randomised trial. Lancet 2007;369:123–131. 85. Riehm H, Gadner H, Henze G. Results and significance of six randomized trials in four consecutive ALL-BFM trials. In: Haematology and Blood Transfusion, Vol. 33, Acute Leukemias II (B¨uchner T, Schellong G, Hiddemann W et al., eds.), Berlin: Springer-Verlag, 1990:439–450. 86. Harms DO, Gobel U, Spaar HJ et al. COALL Study Group. Thioguanine offers no advantage over mercaptopurine in maintenance treatment of childhood ALL: results of the randomized trial COALL92. Blood 2003;102:2736–2740. 87. Vora A, Mitchell CD, Lennard L et al. Medical Research Council. National Cancer Research Network Childhood Leukaemia Working Party. Toxicity and efficacy of 6-thioguanine versus 6-mercaptopurine in childhood lymphoblastic leukaemia: a randomised trial. Lancet 2006;368:1339–1348. 88. Stork LC, Sather H, Hutchinson RJ et al. Comparison of mercaptopurine (MP) with thioguanine (TG) and IT methotrexate (ITM) with IT “triples” (ITT) in children with SR-ALL: results of CCG-1952. Blood 2002;100:156a. 89. de Boer NK, Reinisch W, Teml A et al. Dutch 6-TG Working Group. 6-Thioguanine treatment in inflammatory bowel disease: a critical appraisal by a European 6-TG working party. Digestion 2006;73:25–31. 90. Aarbakke J, Janka-Schaub G, Elion GB. Thiopurine biology and pharmacology. Trends Pharmacol Sci 1997;18:3–7. 91. Lennard L. The clinical pharmacology of 6-mercaptopurine. Eur J Clin Pharmacol 1992;43:329–339. 92. Teml A, Schaeffeler E, Herrlinger KR et al. Thiopurine treatment in inflammatory bowel diseases: clinical pharmacology and implication of pharmacogenetically guided dosing. Clin Pharmacokinet 2007;46:187–208. 93. Elion GB. The George Hitchings and Gertrude Elion Lecture. The pharmacology of azathioprine. Ann NY Acad Sci 1993;685:400–407. 94. Lennard L, Van Loon JA, Lilleyman JS et al. Thiopurine pharmacogenetics in leukemia: correlation of erythrocyte thiopurine methyltransferase activity and 6-thioguanine nucleotide concentrations. Clin Pharmacol Ther 1987;41:18–25. 95. Maybaum J, Mandel HG. Differential chromatid damage induced by 6-thioguanine in CHO cells. Exp Cell Res 1981;135:465–468. 96. Christie NT, Drake S, Meyn RE et al. 6-Thioguanine-induced DNA damage as a determinant of cytotoxicity in cultured Chinese hamster ovary cells. Cancer Res 1984;44:3665–3671. 97. Pan BF, Nelson JA. Characterization of the DNA damage in 6-thioguanine-treated cells. Biochem Pharmacol 1990;40:1063–1069. 98. Bodell WJ. Molecular dosimetry of sister chromatid exchange induction in 9 L cells treated with 6-thioguanine. Mutagenesis 1991;6:175–177. 99. Somerville L, Krynetski EY, Krynetskaia NF et al. Structure and dynamics of thioguanine-modified duplex DNA. J Biol Chem 2003;278:1005–1011. 100. Swann PF, Waters TR, Moulton DC et al. Role of postreplicative DNA mismatch repair in the cytotoxic action of thioguanine. Science 1996;273:1109–1111. 101. Vogt MH, Stet EH, De Abreu RA et al. The importance of methylthio-imp for methylmercaptopurine ribonucleoside (Me-MPR) cytotoxicity in molt F4 human malignant T-lymphoblasts. Biochim Biophys Acta 1993;1181:189–194. 102. Tidd DM, Kim SC, Horakova K et al. A delayed cytotoxic reaction for 6-mercaptopurine. Cancer Res 1972;32:317–322. 103. Dervieux T, Blanco JG, Krynetski EY et al. Differing contribution of thiopurine methyltransferase to mercaptopurine versus thioguanine effects in human leukemic cells. Cancer Res 2001;61:5810–5816.

196

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

104. Tiede I, Fritz G, Strand S et al. CD28-dependent Rac1 activation is the molecular target of azathioprine in primary human CD4+ T lymphocytes. J Clin Invest 2003;111:1133–1145. 105. Neurath MF, Kiesslich R, Teichgr¨aber Ul et al. 6-Thioguanosine diphosphate and triphosphate levels in red blood cells and response to azathioprine therapy in Crohn’s disease. Clin Gastroenterol Hepatol 2005;3:1007–1014. 106. Poppe D, Tiede I, Fritz G et al. Azathioprine suppresses ezrin-radixin-moesin-dependent T-cellAPC conjugation through inhibition of Vav guanosine exchange activity on Rac proteins. J Immunol 2006;176:640–651. 107. Loo TL, Luce JK, Sullivan MP et al. Clinical pharmacologic observations on 6-mercaptopurine and 6-methylthiopurine ribonucleoside. Clin Pharmacol Ther 1968;9:180–194. 108. LePage GA, Whitecar JP. Pharmacology of 6-thioguanine in man. Cancer Res 1971;31:1627–1631. 109. Konits PH, Egorin MJ, Van Echo DA et al. Phase II evaluation and plasma pharmacokinetics of highdose intravenous 6-thioguanine in patients with colorectal carcinoma. Cancer Chemother Pharmacol 1982;8:199–203. 110. Balis FM, Holcenberg JS, Poplack DG et al. Pharmacokinetics and pharmacodynamics of oral methotrexate and mercaptopurine in children with lower risk acute lymphoblastic leukemia: a joint Childrens’ Cancer Group and Pediatric Oncology Branch study. Blood 1998;92:3569–3577. 111. Erb N, Harms DO, Janka-Schaub G. Pharmacokinetics and metabolism of thiopurines in children with acute lymphoblastic leukemia receiving 6-thioguanine versus 6-mercaptopurine. Cancer Chemother Pharmacol 1998;42:266–272. 112. Lennard L, Lilleyman JS. Variable mercaptopurine metabolism and treatment outcome in childhood lymphoblastic leukemia. J Clin Oncol 1989;7:1816–1823. Erratum in: J Clin Oncol 1990;8:567. 113. Lennard L, Lewis IJ, Michelagnoli M et al. Thiopurine methyltransferase deficiency in childhood lymphoblastic leukaemia: 6-mercaptopurine dosage strategies. Med Pediatr Oncol 1997;29:252–255. 114. Lennard L, Van Loon JA, Weinshilboum RM. Pharmacogenetics of acute azathioprine toxicity: relationship to thiopurine methyltransferase genetic polymorphism. Clin Pharmacol Ther 1989;46: 149–154. 115. Lavi LE, Holcenberg JS. A rapid and sensitive high-performance liquid chromatographic assay for 6-mercaptopurine metabolites in red blood cells. Anal Biochem 1985;144:514–521. 116. Lennard L. Assay of 6-thioinosinic acid and 6-thioguanine nucleotides, active metabolites of 6mercaptopurine, in human red blood cells. J Chomatogr 1987;423:169–178. 117. Lennard L, Singleton HJ. High-performance liquid chromatographic assay of the methyl and nucleotide metabolites of 6-mercaptopurine: quantitation of red blood cell 6-thioguanine nucleotide, 6-thioinosinic acid, and 6-methylmercaptopurine metabolites in a single sample. J Chromatogr 1992;583:83–90. 118. Keuzenkamp-Jansen CW, De Abreu RA, Bokkerink JP et al. Determination of extracellular and intracellular thiopurines and methylthiopurines by high-performance liquid chromatography. J Chromatogr B Biomed Appl 1995;672:53–61. 119. Dervieux T, Boulieu R. Simultaneous determination of 6-thioguanine and methyl 6-mercaptopurine nucleotides of azathioprine in red blood cells by HPLC. Clin Chem 1998;44:551–555. 120. Dervieux T, Chu Y, Su Y et al. HPLC determination of thiopurine nucleosides and nucleotides in vivo in lymphoblasts following mercaptopurine therapy. Clin Chem 2002;48:61–68. 121. Stefan C, Walsh W, Banka T et al. Improved HPLC methodology for monitoring thiopurine metabolites in patients on thiopurine therapy. Clin Biochem 2004;37:764–771. 122. Shipkova M, Armstrong VW, Wieland E et al. Differences in nucleotide hydrolysis contribute to the differences between erythrocyte 6-thioguanine nucleotide concentrations determined by two widely used methods. Clin Chem 2003;49:260–268. 123. Lilleyman JS, Lennard L. Mercaptopurine metabolism and risk of relapse in childhood lymphoblastic leukaemia. Lancet 1994;343:1188–1190. 124. Schmiegelow K, Schroder H, Gustafsson G et al. Risk of relapse in childhood acute lymphoblastic leukemia is related to RBC methotrexate and mercaptopurine metabolites during maintenance

Chapter 11 / Thiopurines and TPMT in Childhood ALL

125.

126. 127. 128.

129.

130. 131. 132. 133. 134. 135.

136. 137.

138. 139.

140.

141. 142.

143.

144.

197

chemotherapy. Nordic Society for Pediatric Hematology and Oncology. J Clin Oncol 1995;13: 345–351. Lancaster D, Patel N, Lennard L et al. Leucocyte versus erythrocyte thioguanine nucleotide concentrations in children taking thiopurines for acute lymphoblastic leukaemia. Cancer Chemother Pharmacol 2002;50:33–36. Rowland K, Lennard L, Lilleyman JS. In vitro metabolism of 6-mercaptopurine by human liver cytosol. Xenobiotica 1999;29:615–628. Bostrom B, Erdmann G. Cellular pharmacology of 6-mercaptopurine in acute lymphoblastic leukemia. Am J Pediatr Hematol Oncol 1993;15:80–86. Schmiegelow K, Bjork O, Glomstein A et al. Intensification of mercaptopurine/methotrexate maintenance chemotherapy may increase the risk of relapse for some children with acute lymphoblastic leukemia. J Clin Oncol 2003;21:1332–1339. Bell BA, Brockway GN, Shuster JJ et al. Pediatric Oncology Group study (now The Children’s Oncology Group). A comparison of red blood cell thiopurine metabolites in children with acute lymphoblastic leukemia who received oral mercaptopurine twice daily or once daily: a Pediatric Oncology Group study (now The Childrens’ Oncology Group). Pediatr Blood Cancer 2004;43:105–109. Weinshilboum R. Thiopurine pharmacogenetics: clinical and molecular studies of thiopurine methyltransferase. Drug Metab Dispos 2001;29:601–605. Krynetski E, Evans WE. Drug methylation in cancer therapy: lessons from the TPMT polymorphism. Oncogene 2003;22:7403–7413. Weinshilboum RM, Sladek SL. Mercaptopurine pharmacogenetics: monogenic inheritance of erythrocyte thiopurine methyltransferase activity. Am J Hum Genet 1980;32:651–662. Vuchetich JP, Weinshilboum RM, Price RA. Segregation analysis of human red blood cell thiopurine methyltransferase activity. Gen Epidemiol 1995;12:1–11. Coulthard SA, Rabello C, Robson J et al. A comparison of molecular and enzyme-based assays for the detection of thiopurine methyltransferase mutations. Br J Haematol 2000;110:599–604. Alves S, Amorim A, Ferreira F et al. Influence of the variable number of tandem repeats located in the promoter region of the thiopurine methyltransferase gene on enzymatic activity. Clin Pharmacol Ther 2001;70:165–174. McLeod HL, Lin JS, Scott EP, Pui CH, Evans WE. Thiopurine methyltransferase activity in American white subjects and black subjects. Clin Pharmacol Ther 1994;55:15–20. McLeod HL, Krynetski EY, Wilimas JA et al. Higher activity of polymorphic thiopurine S-methyltransferase in erythrocytes from neonates compared to adults. Pharmacogenetics 1995;5: 281–286. Klemetsdal B, Wist E, Aarbakke J. Gender difference in red blood cell thiopurine methyltransferase activity. Scand J Clin Lab Invest 1993;53:747–749. Szumlanski CL, Honchel R, Scott MC et al. Human liver thiopurine methyltransferase pharmacogenetics: biochemical properties, liver–erythrocyte correlation and presence of isozymes. Pharmacogenetics 1992;2:148–159. Schaeffeler E, Fischer C, Brockmeier D et al. Comprehensive analysis of thiopurine S-methyltransferase phenotype-genotype correlation in a large population of German-Caucasians and identification of novel TPMT variants. Pharmacogenetics 2004;14:407–417. Lennard L, Lilleyman JS, Van Loon J et al. Genetic variation in response to 6-mercaptopurine for childhood acute lymphoblastic leukaemia. Lancet 1990;336:225–229. Szumlanski CL, Weinshilboum RM. Sulphasalazine inhibition of thiopurine methyltransferase: possible mechanism for interaction with 6-mercaptopurine and azathioprine. Br J Clin Pharmacol 1995;39:456–459. Xin H-W, Fischer C, Schwab M et al. Effects of aminosalicylates on thiopurine S-methyltransferase activity: an ex vivo study in patients with inflammatory bowel disease. Aliment Pharmacol Ther 2005;21:1105–1109. Xin H-W, Fischer C, Schwab M et al. Thiopurine S-methyltransferase as a target for drug interaction. Eur J Clin Pharmacol 2005;61:395–398.

198

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

145. Weinshilboum RM, Raymond FA, Pazmino PA. Human erythrocyte thiopurine methyltransferase: radiochemical microassay and biochemical properties. Clin Chim Acta 1978;85:323–333. 146. Anglicheau D, Sanquer S, Loriot MA et al. Thiopurine methyltransferase activity: new conditions for reversed-phase high-performance liquid chromatographic assay without extraction and genotypicphenotypic correlation. J Chromatogr B Analyt Technol Biomed Life Sci 2002;773:119–127. 147. Menor C, Fueyo JA, Escribano O et al. Determination of thiopurine methyltransferase activity in human erythrocytes by high-performance liquid chromatography: comparison with the radiochemical method. Ther Drug Monit 2001;23:536–541. 148. Lennard L, Singleton HJ. High-performance liquid chromatographic assay of human red blood cell thiopurine methyltransferase activity. J Chromatogr B Biomed Appl 1994;661:25–33. 149. Boulieu R, Sauviat M, Dervieux T et al. Phenotype determination of thiopurine methyltransferase in erythrocytes by HPLC. Clin Chem 2001;47:956–958. 150. Indjova D, Shipkova M, Atanasova S et al. Determination of thiopurine methyltransferase phenotype in isolated human erythrocytes using a new simple nonradioactive HPLC method. Ther Drug Monit 2003;25:637–644. 151. Kroplin T, Weyer N, Gutsche S et al. Thiopurine S-methyltransferase activity in human erythrocytes: a new HPLC method using 6-thioguanine as substrate. Eur J Clin Pharmacol 1998;54:265–271. 152. Kroplin T, Iven H. Methylation of 6-mercaptopurine and 6-thioguanine by thiopurine Smethyltransferase: a comparison of activity in red blood cell samples of 199 blood donors. Eur J Clin Pharmacol 2000;56:343–345. 153. Khalil MN, Erb N, Khalil PN et al. Interference free and simplyfied liquid chromatography-based determination of thiopurine S-methyltransferase activity in erythrocytes. J Chromatogr B Analyt Technol Biomed Life Sci 2005;821:105–111. Erratum in: J Chromatogr B Analyt Technol Biomed Life Sci 2005;824:348–350. 154. McLeod HL, Relling MV, Liu Q et al. Polymorphic thiopurine methyltransferase in erythrocytes is indicative of activity in leukemic blasts from children with acute lymphoblastic leukemia. Blood 1995;85:1897–1902. 155. Honchel R, Aksoy IA, Szumlanski C et al. Human thiopurine methyltransferase: molecular cloning and expression of T84 colon carcinoma cell cDNA. Mol Pharmacol 1993;43:878–887. 156. Szumlanski C, Otterness D, Her C et al. Thiopurine methyltransferase pharmacogenetics: human gene cloning and characterization of a common polymorphism. DNA Cell Biol 1996;15:17–30. 157. Krynetski EY, Fessing MY, Yates CR et al. Promoter and intronic sequences of the human thiopurine S-methyltransferase (TPMT) gene isolated from a human Pac1 genomic library. Pharm Res 1997;14:1672–1678. 158. Seki T, Tanaka T, Nakamura Y. Genomic structure and multiple single-nucleotide polymorphisms (SNPs) of the thiopurine S-methyltransferase (TPMT) gene. J Hum Genet 2000;45:299–302. 159. Krynetski EY, Schuetz JD, Galpin AJ et al. A single point mutation leading to loss of catalytic activity in human thiopurine S-methyltransferase. Proc Natl Acad Sci USA 1995;92:949–953. 160. Tai HL, Krynetski EY, Yates CR et al. Thiopurine S-methyltransferase deficiency: two nucleotide transitions define the most prevalent mutant allele associated with loss of catalytic activity in caucasians. Am J Hum Genet 1996;58:694–702. 161. Otterness D, Szumlanski C, Lennard L et al. Human thiopurine methyltransferase pharmacogenetics: gene sequence polymorphisms. Clin Pharmacol Ther 1997;62:60–73. 162. Spire-Vayron de la Moureyre C, Debuysere H, Sabbagh N et al. Detection of known and new mutations in the thiopurine S-methyltransferase gene by single-strand conformation polymorphism analysis. Hum Mutat 1998;12:177–185. 163. Hon YY, Fessing MY, Pui CH et al. Polymorphism of the thiopurine S-methyltransferase gene in African-Americans. Hum Mol Genet 1999;8:371–376. 164. Colombel JF, Ferrari N, Debuysere H et al. Genotypic analysis of thiopurine S-methyltransferase in patients with Crohn’s disease and severe myelosuppression during azathioprine therapy. Gastroenterology 2000;118:1025–1030.

Chapter 11 / Thiopurines and TPMT in Childhood ALL

199

165. Schaeffeler E, Stanulla M, Greil J et al. A novel TPMT missense mutation associated with TPMT deficiency in a 5-year-old boy with ALL. Leukemia 2003;17:1422–1424. 166. Hamdan-Khalil R, Allorge D, Lo-Guidice JM et al. In vitro characterization of four novel non-functional variants of the thiopurine S-methyltransferase. Biochem Biophys Res Commun 2003;309:1005–1010. 167. Lindqvist M, Haglund S, Almer S et al. Identification of two novel sequence variants affecting thiopurine methyltransferase enzyme activity. Pharmacogenetics 2004;14:261–265. 168. Hamdan-Khalil R, Gala JL, Allorge D et al. Identification and functional analysis of two rare allelic variants of the thiopurine S-methyltransferase gene, TPMT*16, and TPMT*19. Biochem Pharmacol 2005;69:525–529. 169. Schaeffeler E, Eichelbaum M, Reinisch W et al. Three novel thiopurine S-methyltransferase allelic variants (TPMT*20, *21, *22)-association with decreased enzyme function. Hum Mutat 2006;27:976. 170. Yates CR, Krynetski EY, Loennechen T et al. Molecular diagnosis of thiopurine S-methyltransferase deficiency: genetic basis for azathioprine and mercaptopurine intolerance. Ann Intern Med 1997;126:608–614. 171. McLeod HL, Pritchard SC, Githang’a J et al. Ethnic differences in thiopurine methyltransferase pharmacogenetics: evidence for allele specificity in Caucasian and Kenyan individuals. Pharmacogenetics 1999;9:773–776. 172. Collie-Duguid ES, Pritchard SC, Powrie RH et al. The frequency and distribution of thiopurine methyltransferase alleles in Caucasian and Asian populations. Pharmacogenetics 1999;9:37–42. 173. Salavaggione OE, Wang L, Wiepert M et al. Thiopurine S-methyltransferase pharmacogenetics: variant allele functional and comparative genomics. Pharmacogenet Genomics 2005;15:801–815. 174. Tai HL, Krynetski EY, Schuetz EG et al. Enhanced proteolysis of thiopurine S-methyltransferase (TPMT) encoded by mutant alleles in humans (TPMT* 3A, TPMT* 2): mechanisms for the genetic polymorphism of TPMT activity. Proc Natl Acad Sci USA 1997;94:6444–6449. 175. Tai HL, Fessing MY, Bonten EJ et al. Enhanced proteasomal degradation of mutant human thiopurine S-methyltransferase (TPMT) in mammalian cells: mechanism for TPMT protein deficiency inherited by TPMT* 2, TPMT* 3A, TPMT* 3B or TPMT* 3C. Pharmacogenetics 1999;9:641–650. 176. Wang L, Sullivan W, Toft D et al. Thiopurine S-methyltransferase pharmacogenetics: chaperone protein association and allozyme degradation. Pharmacogenetics 2003;13:555–564. 177. Wang L, Nguyen TV, McLaughlin RW et al. Human thiopurine S-methyltransferase pharmacogenetics: variant allozyme misfolding and aggresome formation. Proc Natl Acad Sci USA 2005;102: 9394–9399. 178. Seki T, Tanaka T, Nakamura Y. Genomic structure and multiple single-nucleotide polymorphisms (SNPs) of the thiopurine S-methyltransferase (TPMT) gene. J Hum Genet 2000;45:299–302. 179. Fessing MY, Krynetski EY, Zambetti GP et al. Functional characterization of the human thiopurine S-methyltransferase (TPMT) gene promoter. Eur J Biochem 1998;256:510–517. 180. Spire-Vayron de la Moureyre C, Debuysere H, Fazio F et al. Characterization of a variable number tandem repeat region in the thiopurine S-methyltransferase gene promoter. Pharmacogenetics 1999;9:189–198. 181. Alves S, Amorim A, Ferreira F et al. Influence of the variable number of tandem repeats located in the promoter region of the thiopurine methyltransferase gene on enzymatic activity. Clin Pharmacol Ther 2001;70:165–174. 182. Yan L, Zhang S, Eiff B et al. Thiopurine methyltransferase polymorphic tandem repeat: genotype– phenotype correlation analysis. Clin Pharmacol Ther 2000;68:210–219. 183. Marinaki AM, Arenas M, Khan ZH et al. Genetic determinants of the thiopurine methyltransferase intermediate activity phenotype in British Asians and Caucasians. Pharmacogenetics 2003;13: 97–105. 184. Lee D, Szumlanski C, Houtman J et al. Thiopurine methyltransferase pharmacogenetics: cloning of human liver cDNA and a processed pseudogene on human chromosome 18q21.1. Drug Metab Dispos 1995;23:398–405.

200

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

185. Schaeffeler E, Lang T, Zanger UM et al. High-throughput genotyping of thiopurine Smethyltransferase by denaturing HPLC. Clin Chem 2001;47:548–555. 186. Hall AG, Hamilton P, Minto L et al. The use of denaturing high-pressure liquid chromatography for the detection of mutations in thiopurine methyltransferase. J Biochem Biophys Methods 2001;47:65–71. 187. Davison JE, McMullin MF, Catherwood MA. Genotyping of thiopurine methyltransferase in patients with acute leukemia using LightCycler PCR. Leuk Lymphoma 2006;47:1624–1628. 188. Lindqvist M, Almer S, Peterson C et al. Real-time RT-PCR methodology for quantification of thiopurine methyltransferase gene expression. Eur J Clin Pharmacol 2003;59:207–211. 189. Lu Y, Kham SK, Tan PL et al. Arrayed primer extension: a robust and reliable genotyping platform for the diagnosis of single gene disorders: beta-thalassemia and thiopurine methyltransferase deficiency. Genet Test 2005;9:212–219. 190. McDonald OG, Krynetski EY, Evans WE. Molecular haplotyping of genomic DNA for multiple single-nucleotide polymorphisms located kilobases apart using long-range polymerase chain reaction and intramolecular ligation. Pharmacogenetics 2002;12:93–99. 191. Ford L, Graham V, Berg J. Whole-blood thiopurine S-methyltransferase activity with genotype concordance: a new, simplified phenotyping assay. Ann Clin Biochem 2006;43:354–360. 192. Haglund S, Lindqvist M, Almer S et al. Pyrosequencing of TPMT alleles in a general Swedish population and in patients with inflammatory bowel disease. Clin Chem 2004;50:288-295. Erratum in: Clin Chem 2004;50:788. 193. Nasedkina TV, Fedorova OE, Glotov AS et al. Rapid genotyping of common deficient thiopurine S-methyltransferase alleles using the DNA-microchip technique. Eur J Hum Genet 2006;14:991–998. 194. Brouwer C, Marinaki AM, Lambooy LH et al. Pitfalls in determination of mutant alleles in the thiopurine methyltransfrase gene. Leukemia 2001;15:1792–1793. 195. Coulthard SA, Howell C, Robson J et al. The relationship between thiopurine methyltransferase activity and genotype in blasts from patients with acute leukemia. Blood 1998;92:2856–2862. 196. Lennard L, Gibson BE, Nicole T et al. Congenital thiopurine methyltransferase deficiency and 6-mercaptopurine toxicity during treatment for acute lymphoblastic leukaemia. Arch Dis Child 1993;69:577–579. 197. McLeod HL, Coulthard S, Thomas AE et al. Analysis of thiopurine methyltransferase variant alleles in childhood acute lymphoblastic leukaemia. Br J Haematol 1999;105:696–700. 198. Relling MV, Hancock ML, Rivera GK et al. Mercaptopurine therapy intolerance and heterozygosity at the thiopurine S-methyltransferase gene locus. J Natl Cancer Inst 1999;91:2001–2008. 199. Andersen JB, Szumlanski C, Weinshilboum RM et al. Pharmacokinetics, dose adjustments, and 6-mercaptopurine/methotrexate drug interactions in two patients with thiopurine methyltransferase deficiency. Acta Paediatr 1998;87:108–111. 200. Evans WE, Horner M, Chu YQ et al. Altered mercaptopurine metabolism, toxic effects, and dosage requirement in a thiopurine methyltransferase-deficient child with acute lymphocytic leukemia. J Pediatr 1991;119:985–989. 201. Kaskas BA, Louis E, Hindorf U et al. Safe treatment of thiopurine S-methyltransferase deficient Crohn’s disease patients with azathioprine. Gut 2003;52:140–142. 202. Evans WE, Hon YY, Bomgaars L et al. Preponderance of thiopurine S-methyl-transferase deficiency and heterozygosity among patients intolerant to mercaptopurine or azathioprine. J Clin Oncol 2001;19:2293–2301. 203. Relling MV, Yanishevski Y, Nemec J et al. Etoposide and antimetabolite pharmacology in patients who develop secondary acute myeloid leukemia. Leukemia 1998;12:346–352. 204. Relling MV, Rubnitz JE, Rivera GK et al. High incidence of secondary brain tumours after radiotherapy and antimetabolites. Lancet 1999;354:34–39. 205. Relling MV, Pui CH, Cheng C et al. Thiopurine methyltransferase in acute lymphoblastic leukemia. Blood 2006;107:843–844. 206. Stanulla M, Schaeffeler E, Flohr T et al. Thiopurine methyltransferase (TPMT) genotype and early treatment response to mercaptopurine in childhood acute lymphoblastic leukemia. JAMA 2005;293:1485–1489.

Chapter 11 / Thiopurines and TPMT in Childhood ALL

201

207. Thomsen BJ, Schroder H, Kristinsson J et al. Possible carcinogenic effect of 6-mercaptopurine on bone marrow stem cells: relation to thiopurine metabolism. Cancer 1999;86:1080–1086. 208. Stanulla M, Schaeffeler E, Moricke A et al. Thiopurine methyltransferase genotype is not a risk factor for secondary malignant neoplasias after treatment for childhood acute lymphoblastic leukemia on Berlin FrankfurtM¨unster protocols. Blood 2006;108:48a. 209. Lennard L, Richards S, Cartwright CS et al. UK MRC/NCRI Childhood Leukaemia Working Party. The thiopurine methyltransferase genetic polymorphism is associated with thioguanine-related venoocclusive disease of the liver in children with acute lymphoblastic leukemia. Clin Pharmacol Ther 2006;80:375–383. 210. Alves S, Prata MJ, Ferreira F et al. Thiopurine methyltrarisferase pharmacogenetics: alternative molecular diagnosis and preliminary data from Northern Portugal. Pharmacogenetics 1999; 9:257–261.

“This page left intentionally blank.”

12

Impact of Polymorphisms on the Clinical Outcomes of Monoclonal Antibody Therapy Against Hematologic Malignancies Dong Hwan Kim MD, PhD CONTENTS I NTRODUCTION R ITUXIMAB AND ITS M ECHANISM OF ACTION OTHER M ONOCLONAL A NTIBODIES IN THE T REATMENT OF H EMATOLOGIC M ALIGNANCIES T HE F C␥ R ECEPTOR G ENE P OLYMORPHISM AND IN V ITRO DATA T HE F C␥ R ECEPTOR G ENE P OLYMORPHISM AND C LINICAL DATA P OLYMORPHISM OF THE G ENES IN THE C OMPLEMENT PATHWAYS AND C LINICAL DATA F UTURE D IRECTIONS TO I MPROVE E FFICACY OF M ONOCLONAL T HERAPY R EFERENCES

S UMMARY Monoclonal antibodies that target various specific antigens can be used to kill the tumor cells expressing specific antigens, especially in hematologic cancers. Rituximab, one of the commonly used monoclonal antibodies, was suggested to mediate its action mechanism via antibody-dependent cellular cytotoxicity (ADCC), From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 12, 

203

204

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

complement-dependent cytotoxicity (CDC), or a direct pro-apoptotic effect. It has been proposed that the inter-individual variation of gene expression is a consequence of single nucleotide polymorphisms (SNPs) and that the response to a monoclonal antibody can be affected by the SNPs in the host genes corresponding to the drug binding to the target cells or the metabolism of the drug. This chapter reviews the current understanding of the mechanism of action of monoclonal antibodies (especially rituximab), the role of Fc␥ receptor and Fc␥ receptor gene polymorphisms, and their impact on treatment outcomes in hematologic malignancies including follicular lymphoma (FL), diffuse large B-cell lymphoma (DLBCL), Waldenstrom’s macroglobulinemia (WM), and chronic lymphocytic leukemia (CLL). In addition, we discuss the approaches augmenting its clinical activity, especially focusing on Fc␥ receptor re-engineered monoclonal antibody. Key Words: Polymorphisms; monoclonal antibody; hematologic malignancies; rituximab

1. INTRODUCTION Monoclonal antibodies that target various specific antigens can be used to kill the tumor cells expressing specific antigens. To our knowledge, five kinds of monoclonal antibodies have been approved by the Food and Drug Administration (FDA) of the R United States for the treatment of hematologic malignancies: rituximab (Rituxan  R or MabThera ; IDEC pharmaceuticals Corp, San Diego, CA, USA, and Genentech, R R or MabCAMPATH , Genzyme, San Fransisco, CA, USA), alemtuzumab (Campath Cambridge, MA, USA, and Schering AG, Berlin, Germany), ibritumomab tiuxtan R R (Zevaline , Biogen IDEC, San Diego, CA, USA), tositumomab (Bexxar , Corixa and R , Wyeth, GlaxoSmithKline, Seattle, WA, USA) and gemtuzumab ozogamicin (Mylotarg Madison, NJ, USA). A major breakthrough in the treatment of lymphoid malignancies was the discovery of monoclonal antibody activity, especially that of rituximab. Rituximab was the first monoclonal antibody approved by the U.S., FDA for the treatment of relapsed follicular lymphoma (1), and it has now been extensively used for the treatment of various lymphoid neoplasm which express CD20 antigen. Its efficacy has been also demonstrated against diffuse large B-cell lymphoma when administered as a combination regimen such as rituximab plus CHOP (R-CHOP) chemotherapy (2). The precise mechanism of rituximab, as well as other monoclonal antibodies, is still incompletely understood despite extensive investigations. To our current knowledge, the mechanism of rituximab activity includes antibody-dependent cellular cytotoxicity (ADCC), complement dependent cytotoxicity (CDC). and a direct pro-apoptotic effect (3,4). There are two major factors predisposing to resistance to monoclonal antibody therapy. One is a tumor-related factor such as antigen loss, complement resistance antigen expression, intrinsic resistance, or tumor burden. Besides tumor-related factors, growing evidence has indicated that patient-related factors may account for the different responses of the patients to monoclonal antibody therapy. For example, differences in ADCC or CDC function according to individuals may increase our understanding of

Chapter 12 / SNPs in Monoclonal Antibody Therapy

205

resistance to monoclonal antibody therapy. The affinity of host effector cells to monoclonal antibody via Fc␥ receptor III (Fc␥RIII; CD16) or Fc␥ receptor II (Fc␥RII; CD32) has been known to mediate the ADCC activity of effector cells (5). The importance of the Fc receptor to the action mechanism of monoclonal antibody is derived from the following facts. Immunoglobulin has two binding sites: binding sites to antigen and to Fc␥ receptor or complements. Although the antigenbinding site is important for binding to tumor cells expressing a specific antigen, the binding affinity to Fc␥ receptor is known to be correlated with the efficacy of monoclonal antibody, especially rituximab. Growing evidence suggests that the single nucleotide polymorphisms of Fc␥ receptor genes are associated with various binding capacities and different clinical responses to monoclonal antibody therapy (5,6). An understanding of Fc␥ receptor gene polymorphism explains why some patients do not respond to monoclonal antibody, and it may broaden our understanding of monoclonal antibody activity and improve treatment outcomes in the future. Specific strategies include modulation of Fc␥ receptor affinity or introduction of Fc␥ receptor-reengineered monoclonal antibody designed to enhance binding to the Fc␥ receptor. This chapter reviews our current understanding of the mechanism of action of monoclonal antibody (especially rituximab), as well as the role of Fc␥ receptor and Fc␥ receptor gene polymorphisms, and their impact on treatment outcomes in hematologic malignancies including follicular lymphoma (FL), diffuse large B-cell lymphoma (DLBCL), Waldenstrom’s macroglobulinemia (WM), and chronic lymphocytic leukemia (CLL). We will discuss the approaches augmenting the clinical activity of monoclonal antibody, especially focusing on Fc␥ receptor re-engineered monoclonal antibody. A better understanding of how monoclonal antibody acts in vivo will lead to the development of new, more effective therapeutic strategies.

2. RITUXIMAB AND ITS MECHANISM OF ACTION Rituximab is an IgG␬ chimeric monoclonal antibody derived from murine antibody against CD20. It has a fragment ab (Fab) domain that binds to the CD20 antigen on B lymphocytes, and a fragment c (Fc) domain that recruits immune effector cells to rituximab-coated tumor cells. CD20 antigen has been found on various lymphoid hematologic malignancies including B-cell non-Hodgkin’s lymphoma (NHL), B-cell CLL. and some acute lymphoblastic leukemia. The function of CD20 antigen has not been fully elucidated. However, it is generally accepted that CD20 plays an important role in regulating cell cycle progression in B cells (3). Several previous investigations showed that intracellular signals following binding of rituximab to CD20 antigen on lymphoma cells lead to apoptosis (4,7,8,9). Several mechanisms have been proposed for the action mechanism of rituximab including ADCC, CDC, complement dependent cellular cytotoxicity (CDCC), and a direct pro-apoptotic effect (3,4,10). However, the mechanism of action of rituximab has yet to be fully elucidated.

206

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

2.1. Induction of Apoptosis/Anti-proliferation Apoptosis is a process of programmed cell death by which cells undergo organized self-destruction. Classical apoptosis is caused by the activation of caspases, a family of intracellular cysteine proteases, although apoptosis may occur without activation of caspases in some situations (3). There are at least two major pathways of caspase activation in caspase-dependent apoptosis. The intrinsic, or mitochondrial-mediated pathway induces apoptosis through the bcl-2 family of proteins and activation of caspase 9. The extrinsic, or death receptor-mediated pathway induces activation of caspase 8. These two caspases (caspase 8 and 9) subsequently activate caspase 3. The intrinsic pathway is dependent on activated Bak/Bax and is inhibited by bcl-2 proteins (Fig. 1) ( 11, 12). Although there have been some controversial results in recent investigations, bcl-2 proteins appear to block the intrinsic apoptosis pathway by inhibiting caspase activity (3). In an in vitro model, exposure of lymphoma cells to rituximab resulted in the activation of the Src-family of protein tyrosine kinases (13), leading to the phosphorylation of PLC␥2, which induces calcium influx and activates caspase 3, resulting in promotion of apoptotic cell death ( 8, 14). Another in vitro model showed that exposure to rituximab resulted in the sustained phosphorylation of p38-MAPK, JNK, and ERK kinases

Fig. 1. Proposed mechanism of action of rituximab associated with the apoptosis pathway. Binding of rituximab with the CD20 antigen up-regulates the production of interleukin-10 (IL-10). The IL10 autocrine loop down-regulates the expression of the bcl-2 protein, which inhibits the intrinsic pathway (or mitochondrial mediated pathway) of apoptosis. The mitochondrial pathway is induced by intracellular stress signals. The translocation of the bcl-2 protein into the mitochondria leads to the activation of caspase 9 via release of cytochrome c and apoptotic protease-activating factor 1. The other pathway, the extrinsic pathway (or death receptor mediated pathway) activates caspase 8. Subsequently, caspase 8 or 9 activates caspase 3, leading to programmed cell death (apoptosis).

Chapter 12 / SNPs in Monoclonal Antibody Therapy

207

in lymphoma cells (15). Also, rituximab-induced translocation of CD20 into lipid rafts is known to be crucial calcium influx, resulting in apoptosis (8,13,16). Recent studies showed that rituximab down-regulates interleukin-10 (IL-10) in some lymphoma cell lines (17,18,19,20,21,22,23,24,25,26,27). IL-10 has a stimulatory function as an autocrine or paracrine growth factor for lymphoma cells (28). IL-10 is a known promoter of BCL2 expression in hematopoietic cells (29) as well as lymphoma cells (28). Rituximab is known to induce down-regulation of IL-10 expression and consequently of bcl-2 protein expression, making B-cells more susceptible to apoptotic signals (Fig. 1) (30,31,32). However, the implication of IL-10 on clinical outcome after rituximab therapy is still unclear.

2.2. Antibody-Dependent Cellular Cytotoxicity (ADCC) Antibody consists of fragment ab (Fab) domain that binds to the antigen and fragment c (Fc) domain that recruits immune effector cells. Accordingly, receptors for the constant region (Fc) of antibodies provide critical links between the cellular and humoral parts of the immune system (6). Fc receptors belong to the immunoglobulin superfamily members. Several types of Fc receptors have been revealed including Fc␥RI (CD64), Fc␥RII (CD32), and Fc␥RIII (D16). The encoding genes are located on chromosome 1q21–24. The structures and properties of Fc␥ receptors are shown in Fig. 2 and Table 1 ( 5). The ligand-binding chain of most Fc␥R consists of two or three extracellular immunoglobulin-like domains, a transmembrane domain, and intracellular cytoplasmic domains. As shown in Fig. 2, all Fc receptors have activating action on corresponding effector cells except for the Fc␥RIIIb which has immunoreceptor tyrosine-based inhibitory motif (ITIM, Fig. 2) in the cytoplasmic domain. Fc␥RIa, IIa, IIc, and IIIa are activating receptors due to the immunoreceptor tyrosine-based activation motif (ITAM, Fig. 2) either in the cytoplasmic domain (Fc␥RIIa, IIc) or in the accessory signaling ␥-chain (Fc␥RIa, IIc). On effector cells, the activating Fc receptor starts downstream signaling by tyrosine phosphorylation of ITAM, and results in subsequent immune responses including ADCC, phagocytosis, or cytokine production (5,6). The expression profiles of Fc receptors are summarized in Table 1 according to the type of effector cell. Monocytes/macrophages express Fc␥RI, IIa, IIb, and IIIa. Natural killer (NK) cells express Fc␥RIIc and IIIa, while neutrophils express RI, IIa, and IIIb. Cells that express both inhibitory and activating receptors, such as macrophages, may have a balanced activation by Fc receptor. The affinity of host effector cells to rituximab has been known to mediate the ADCC activity of effector cells (5). In addition, the Fc␥ receptor polymorphism has been found to affect the binding capacities and clinical responses to rituximab (5,6). This issue will be discussed later in this chapter.

2.3. Complement-Dependent Cytotoxicity (CDC) It has been proposed that rituximab also functions through the complement-dependent cytotoxicity (CDC) pathway, specifically the classical pathway, which requires

208

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Fig. 2. Organization of the Fc receptor gene cluster and the structure of Fc␥ receptors. A. The Fc receptor genes, the so-called Fc receptor cluster (FRC), are located on human chromosom 1q21–24 region. The FRC gene for Fc␥RI is located proximal to the centromeric end, while the Fc␥RII/III gene is located near the telomeric end of FRC. B. The Fc␥RI consists of ␣ chain and a dimer of the ␥ chain. The Fc␥RIIIA consists of ␣ chain and mainly ␥ chain dimer except for natural killer cell, where the Fc␥RIIIA contains ␨ chain dimer. The Fc␥RIIB, inhibitory Fc␥ receptor, has a cytoplasmic immunoreceptor tyrosine-based inhibitory motif (ITIM), while Fc␥RI, Fc␥RIIA and Fc␥RIIIA contains an immunoreceptor tyrosine-based activation motif (ITAM) on their cytoplasmic domains. Black circles indicate three principal polymorphism sites on Fc␥RIIA, Fc␥RIIIA, and Fc␥RIIIB. (Reprinted after modification from Journal of Allergy and Clinical Immunology, Vol 111, Binstadt BA, Geha RS, and Bonila FA., IgG Fc receptor polymorphisms in human disease: Implications for intravenous immunoglobulin therapy, 697–703 (2003) with permission from the American Academy of Allergy, Asthma, and Immunology).

immunoglobulin for activation (Fig. 3). The binding of the C1q component to Fc portions of IgG or IgM triggers a proteolytic cascade. The C3b fragment, which is generated as a result of the cascade, not only acts as opsonins, but also binds to C3 convertase to form C5 convertase, thereby generating a membrane attack complex that disrupts the target cell membrane and results in cell death. In addition, binding of C3b to complement receptors expressed on effector cells such as NK cells, macrophages, or neutrophils induces phagocytosis or cell-mediated lysis—so-called complement-dependent cellular cytotoxicity (CDCC) (3,4). In addition to the positive control for activation of complement lysis, complement inhibitory proteins such as CD55 or CD59 exert a negative control mechanism of complement lysis (3,4). Tumor cells over-expressing CD55 and CD59, such as CLL cells, are less likely to achieve a response to rituximab therapy. In other words, a blockade of CD55 and CD59 may enhance rituximab-mediated CDC.

Table 1 Properties and Distribution of Fc␥ Receptors Name

209

Fc␥RI (CD64) Fc␥RII (CD32) Fc␥RIIA Fc␥RIIB Fc␥RIIC Fc␥RIII (CD16) Fc␥RIIIA Fc␥RIIIB

Structure

Alleles

␣␥2

IgG Subclass Specificity

␣ ␣ ␣

R/H-131

IgG1=IgG3>IgG4>>IgG2 IgG1=IgG3>>IgG2,IgG4 R is lower affinity

␣␥2 ␣ (GPI)

F/V-158 NA1/NA2

IgG1, IgG3>>IgG2, IgG4 F is lower affinity NA2 is lower affinity

Distribution Mo/Mc

NK

PMN

B

DC

Mast

Plt

+

—

±

—

+

±

—

+ + —

— — +

+ — —

— + —

+ + —

+ + —

+ — —

+ —

+ —

— +

— —

+ —

— —

— —

Mo, monocytes; Mc, macrophage; NK, natural killer cell; PMN, polymorphonuclear leukocyte; B, B-cells; DC, dendritic cells; Plt, platelet; GPI, glycosylphosphatidylinositol-linked; NA, neutrophil antigen. Reprinted after modification from Journal of Allergy and Clinical Immunology, Vol. 111, Binstadt BA, Geha RS, and Bonila FA, IgG Fc receptor polymorphisms in human disease: Implications for intravenous immunoglobulin therapy, 697–703 (2003) with permission from the American Academy of Allergy, Asthma, and Immunology.

210

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Fig. 3. Mechanism of complement-dependent cytotoxicity. Binding of rituximab with the CD20 antigen leads to initiation of the classical pathway of complement activation. The binding of the Fc portions with C1q triggers a proteolytic cascade. The activated C1q subsequently activates C4 and C2. The complex of C1q, C4 and C2 subsequently forms the C3 convertase complex. The C3 convertase generates large amounts of C3b. The C3b molecule form C5 convertase after binding with C3 convertase. C5b, C6, C7, C8 and C9 associate to generate the membrane attack complex (MAC), which kills the target cells. The CD55 (decay accelerating factor, DAF), inhibits both C3 and C5 convertase, while CD59 (membrane inhibitor of reactive lysis) inhibits the activity of the MAC. In addition, released C3a and C5a molecules induce chemotaxis and the activation of immune cells.

3. OTHER MONOCLONAL ANTIBODIES IN THE TREATMENT OF HEMATOLOGIC MALIGNANCIES 3.1. Alemtuzumab R R Campath-1 H (alemtuzumab, Campath or MabCAMPATH ) is another monoclonal antibody targeting the CD52 antigen on tumor cells. It is a recombinant monoclonal antibody of the IgG1 isotype generated by cloning the hypervariable regions of the parent murine Campath-1 G into the framework regions of normal human IgG1. The CD52 antigen, a 21- to 28-kDa glycopeptide, is expressed at high density on the surface of B- and T-lymphocytes, macrophages, and monocytes, but not hematopoietic stem cells (33). It is also expressed on all CLL cells and indolent B-cell lymphoma cells. The entire structure of the CD52 molecule has been determined. However, its function still remains to be revealed. Although the precise mechanism of alemtuzumab activity is still under investigation, alemtuzumab exerts its therapeutic effects through binding to the CD52 antigen on target cells. CDC and ADCC occur by the activation of NK cells and macrophages through their Fc␥ receptors with subsequent clearance of malignant lymphocytes from

Chapter 12 / SNPs in Monoclonal Antibody Therapy

211

the peripheral blood and marrow of patients (10). In some studies, alemtuzumab-induced apoptosis has been reported to be caspase-independent (34). Alemtuzumab has shown impressive results in refractory or relapsed CLL as well as up-front therapy for untreated CLL (35). The use of alemtuzumab to eradicate minimal residual disease of CLL is also reasonable (36). The role of Fc␥ receptor polymorphisms on the clinical activity of alemtuzumab for CLL has not been extensively studies and will be discussed later.

3.2. Ibritumomab Tiuxetan R Ibritumomab tiuxetan (Zevalin , Biogen IDEC, San Diego, CA, USA) is a murine anti-CD20 antibody from which rituximab, a humanized antibody, was engineered. Zevalin was derived from ibritumomab by binding with tiuxetan, which chelates the radionuclide 90 Yttrium through carboxyl groups. 90 Yttrium-ibritumomab tiuxetan can selectively deliver therapeutic radiation dose to CD20 antigen expressing tumors, and can prevent significant radiation exposure to normal tissues overlying the tumor mass (37,38). Zevalin has been approved by the U.S. FDA for radioimmunotherapy in B-cell lymphoma ( 39). The activity of ibritumomab tiuxetan had been evaluated or is under investigation for rituximab-refractory lymphoma patients, relapsed diffuse large B-cell or mantle cell lymphoma patients, or as a part of high-dose chemotherapy with autologous hematopoietic stem cell support (40). The results of phase I, II, and randomized trials showed that the single doses of Zevalin are safe and effective in the treatment of patients with relapsed B-cell NHL with a response rate of 60%–80% ( 40). Unfortunately, no data are available regarding the impact of polymorphisms on the clinical outcomes after Zevalin therapy.

3.3. Tositumomab Tositumomab is a monoclonal antibody that binds to CD20 antigen, and can be R , Corixa and labeled with idodine-131 to yield 131 I-labelled tositumomab (Bexxar 131 GlaxoSmithKline, Seattle, WA, USA) ( 10). The activity of I-tositumomab depends on several mechanisms, including ionizing radiation from 131 I. It has been approved by the U.S. FDA for the treatment of patients with CD20 antigen-positive relapsed or refractory, low-grade, follicular or transformed lymphomas including rituximab-refractory lymphoma (10). The influence of polymorphisms on clinical outcomes following Bexxar therapy has not been investigated.

4. THE FC␥ RECEPTOR GENE POLYMORPHISM AND IN VITRO DATA Three major functionally relevant polymorphisms have been identified for the lowaffinity Fc␥R’s: Fc␥RIIa, Fc␥RIIIa, and Fc␥RIIIb. The Fc␥RI (CD64) is a high-affinity Fc␥R (5). The genes encoding Fc␥Rs are located on chromosome 1q21–24 (Fig. 2). In this chapter, we focus on the polymorphisms of the low-affinity Fc␥R’s such as Fc␥RIIa, Fc␥RIIIa, and Fc␥RIIIb. Each of the polymorphisms is located in the extracellular

212

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

domain of Fc␥R as shown in Fig. 2. The extracellular domain of Fc␥R corresponds to the Fc binding portion of Fc␥R, and affects the affinity to rituximab and other monoclonal antibodies (6).

4.1. FCGR2A Gene Polymorphism Fc␥RIIa is expressed on macrophages, neutrophils, dendritic cells, and some mast cells. The major Fc␥RIIa receptor polymorphism is a single nucleotide change affecting amino acid position at 131, coding for either arginine (R131) or histidine (H131) (41). FCGR2A-H131 has a higher affinity for human IgG3. Functionally, phagocytes obtained from individuals with homozygous HH 131 of FCGR2A showed much more effective phagocytic activity compared with those with RR 131 of FCGR2A. The FCGR2A gene contains a second polymorphic site. CA-to-GA mutation results in a glutamine or tryptophan at amino acid position 27 in the membrane distal Ig-like domain. However, this substitution does not affect the Fc␥RIIa affinity for IgG (5,6).

4.2. FCGR3A Gene Polymorphism Fc␥RIIIa is expressed on macrophages, natural killer cells, and some dendritic cells as a transmembrane receptor. The major polymorphism of Fc␥RIIIa gene is a single nucleotide substitution at position 559 resulting in the presence of either valine (V158) or phenylalanine (F158) at position 158 in the extracellular domain (41,42). It is known that valine (V) allele of the Fc␥ RIIIa gene has a higher affinity to human IgG1 and IgG3 than the phenylalanine (F) allele. Cells bearing the Fc␥ RIIIa V allele mediate ADCC more effectively (42,43). Dall’Ozzo et al. (44) found that NK cells from subjects with the VV genotype for FCGR3A-158 showed a higher affinity for Fc␥R, and mediated ADCC at a lower concentration of monoclonal antibody than did NK cells with the FF genotype. Hatjiharissi et al. (45) presented an important in vitro data regarding the functional assay results according to the FCGR3A genotype. They isolated peripheral blood NK cells from 52 healthy donors who were genotyped for FCGR3A-158. Higher transcript levels were noted in the VV group (23.2 ng/mL) compared to the VF (6.7 ng/mL) or FF group (6.2 ng/mL; p < 0.0001) using allele-specific quantitative RT-PCR. The number of CD16 receptors per NK cell was higher in a group carrying the V allele compared with the FF group (105,947 in VV; 94,863 in VF; and 69,130 in FF group; p = 0.033) by quantitative flow cytometry. Moreover, after incubation of NK cells with rituximab at a concentration of 10–200 mcg/mL, binding of rituximab to NK cells was higher among donors expressing at least one valine (72% in VV; 53% in VF; and 37% in FF group; p = 0.017). Finally, in an assay of rituximab-dependent NK cell–mediated cytotoxicity, a higher level of ADCC killing was observed in the VV (82%) or VF (80%) groups versus the FF group (23%). This study consistently showed that individuals who express at least one valine at FCGR3A-158 have increased Fc␥RIIIA receptor expression on NK cells, rituximab binding to NK cells, and ADCC-mediated killing of lymphoma cells. This may explain better clinical outcomes to rituximab therapy in patients with V-allele at FCGR3A 158 loci (45). The FCGR3A gene contains a second polymorphic site ( 6). A non-functional triallelic polymorphism in FCGR3A at amino acid position 48 within the extracellular

Chapter 12 / SNPs in Monoclonal Antibody Therapy

213

domain has been linked to the FCGR3A-158. Further study will be needed to reach definite conclusion on the role of this polymorphism in rituximab therapy.

4.3. FCGR3B Gene Polymorphism Neutrophils are the only effector cells expressing the GPI-anchored Fc␥RIIIb. This Fc␥ receptor plays an important role in the uptake of immune complexes. Fc␥RIIIb bears the neutrophil antigen polymorphism in its membrane-distal Ig-like domain (41). A four amino acid substitution causes alleleic differences in receptor glycosylation resulting in two isoforms of Fc␥RIIIb, Fc␥RIIIb-NA1, and Fc␥RIIIb-NA2. It is known that Fc␥RIIIb-NA1 has a higher affinity for IgG1- and IgG3-opsonized particles and immune complexed IgG2 than does Fc␥RIIIb-NA2 (5,6). The FCGR3B gene containes another polymorphism site at amino acid position 266 resulting in a substitution of asparatate to alanine. However, its functional relevance has not been demonstrated (6).

4.4. FCGR2B Gene Polymorphism A polymorphic change of nucleotide T-to-C specifying an isoleucine (I) or threonine (T) at amino acid position 232 has been found in the transmembrane region of Fc␥RIIb (6). The functional relevance of this polymorphism is not completely understood. Seven additional polymorphisms are described in either the ligand-binding domain or intronic regions (5,6).

4.5. Linkage Between FCGR3A and FCGR2A Gene Polymorphism Higher response rates after rituximab therapy were observed in FL patients with the HH genotype at FCGR2A-131 and the VV genotype at FCGR3A-158 (46). Therefore, a genetic link between FCGR3A and FCGR2A polymorphisms has been investigated. Hatjiharissi et al. ( 47) recently reported that there is strong linkage disequilibrium between FCGR3A and FCGR2A gene polymorphisms in a healthy population. They evaluated two loci of the FCGR3A gene (at position 158 and 48) and two loci of the FCGR2A gene (at position 131 and 27). Significantly, strong associations were noted between FCGR3A-158 and 48, between FCGR2A-131 and 27, between FCGR3A-48 and FCGR2A-131, between FCGR3A-48 and FCGR2A-27, and between FCGR3A-158 and FCGR2A-27. This finding suggested that significant linkage disequilibrium may account for some overlapping role of FCGR3A and FCGR2A gene polymorphism to predict the response to rituximab therapy in FL patients (47).

5. THE FC␥ RECEPTOR GENE POLYMORPHISM AND CLINICAL DATA 5.1. Follicular Lymphoma Follicular lymphoma (FL) is the most extensively investigated subtype of lymphoid neoplasm regarding the influence of Fc␥Rs polymorphism on treatment. Depending on the clinical situation and patients’ circumstances, there are various options for the treatment for FL. In this chapter, we summarize the impact of Fc␥R gene polymorphisms not

214

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

only on the outcome of rituximab therapy, but also on the outcomes of other treatment modalities such as chemotherapy, cytokine therapy, and idiotype vaccination therapy (Table 2).

5.1.1. R ITUXIMAB M ONOTHERAPY Previous studies that have examined single nucleotide polymorphisms of the FCGR3A gene have consistently demonstrated that FCGR3A polymorphisms influence response to rituximab or survival after rituximab therapy in FL patients (Table 2) (46,48). The first study of the impact of Fc␥R gene polymorphisms on the therapeutic activity of rituximab was reported by Cartron et al. (Table 2) (48). Out of 49 patients with untreated FL, the group with the VV genotype at FCGR3A-158 showed a higher response rate and clearance of the BCL2-JH gene rearrangement from peripheral blood or marrow following rituximab monotherapy (375 mg/m2 weekly for 4 weeks), compared to the group with VF or FF genotypes. The objective response rate at 2 months after rituximab therapy was 100%, 70%, and 64% in groups with VV, VF, and FF genotypes at FCGR3A-158, respectively (p = 0.09). A significant difference was found in the response rate at 2 months of the VV genotype group (100%) and that of the VF or FF genotype groups (67%; relative risk 1.5, p = 0.03). The objective response rate at 12 months was 90%, 59%, and 45% in the VV, VF, and FF genotype groups, respectively (p = 0.06). Also, a significant difference in response rate was noted between individuals with the VV genotype and F allele (90% vs. 51%, p = 0.03). In terms of the molecular response following rituximab therapy, a clearance of BCL2JH rearrangement in marrow or blood samples at 12 months was higher with the VV genotype (5 of 6 patients) compared to others (5 of 17 patients, p = 0.03). In this study, the group with the VV genotype seemed to have a longer progression free survival than those with the VF or FF genotype, but a difference did not reach statistical significance perhaps due to the limitation of sample size (n = 49). The survival benefit and superior response rate in the V/V genotype group has been demonstrated by the work of Weng et al (Table 2) (46). They analyzed the treatment outcomes of 87 patients with FL treated with rituximab (375 mg/m2 weekly for 4 weeks), according to the FCGR3A-158 and FCGR2A-131 genotypes. Objective response at 3, 6, 9, and 12 months following rituximab therapy and progression-free survival were significantly different depending on the FCGR3A genotype. The objective responses in the VV genotype group versus the VF or FF genotype groups were 92% vs. 59% at 1–3 months ( p= 0.027), 85% vs. 45% at 6 months (p = 0.013), 75% vs. 36% at 9 months (p = 0.023), and 75% vs. 26% at 12 months (p = 0.002). The time to progression was significantly longer in the VV genotype group (534 days) than in the VF or FF genotype groups (170 days, p = 0.023). Weng et al. (46) also analyzed the treatment outcomes of rituximab therapy for FL according to FCGR2A genotype. Although the response rates was not significantly different at 1–3 months, the objective response rates were significantly different between the HH genotype group and the HR or RR genotype groups at 6 months (80% vs. 43%, p = 0.005), 9 months (70% vs. 32%, p = 0.004) and 12 months (55% vs. 26%, p = 0.027).

Table 2 The Influence of Single Nucleotide Polymorphism on the Treatment Outcomes for Follicular Lymphoma Patients Reference (year)

Treatment Modality

FCGR3A-158 polymorphism Cartron (2002) Rituximab in untreated FL (n = 49)

Rituximab in untreated FL (n = 87)

Ghielmini (2003)

Rituximab in untreated FL (n = 144)

Maloney (2004)

Sequential CHOP + rituximab (n = 87)

Carlotti (2005)

Sequential CHOP + rituximab in untreated FL (n = 88) Chemo alone for FL (n = 158)

215

Weng (2003)

Weng (2004)

Milan (2004)

IL-2 + rituximab in rituximab-resistant low grade NHL (n = 57)

Outcomes

Significances

References

Response at 2 mo, VV (100%), VF (70%), FF (64%), F allele (67%) Response at 12 mo, VV (90%), VF (59%), FF (45%), F allele (51%) Response at 1–3 mo, VV (92%), VF (53%), FF (68%), F allele (59%) PFS at 2 yrs, VV (45%), VF (12%), FF (16%), F allele (14%) Response, VV (63%), VF or FF (49%) EFS duration, VV (33.3 mo), VF or FF (14.1 mo) Response, no difference PFS at 2 yrs, VV (55%), VF (80%), FF (75%) Complete Response, VV (76%), VF or FF (68%)

p = 0.03 p = 0.03

48

p = 0.027 p = 0.023

46

p = NS p = 0.0034

49

p = NS p = NS

51

p = NS

52

p = NS p = NS

53

Response, no difference PFS at 2 yrs, VV (35%), VF (40%), FF (38%) Complete response, VV (0%), VF (0%), FF (24%)

55

(Continued)

Table 2 (Continued) Reference (year)

Treatment Modality

Weng (2004)

Idiotype vaccination after chemotherapy for FL (n = 136) FCGR2A-131 polymorphism Weng (2003) Rituximab in untreated FL (n = 87)

Maloney (2004)

216

Sequential CHOP + rituximab in untreated FL (n = 87) Weng (2004) Chemo alone in FL (n = 158) Weng (2004) Idiotype vaccination after chemotherapy for FL (n = 136) FCGR2B-232 polymorphism Weng (2004) Chemo alone in FL (n = 158) C1qA polymorphism Racila (2005) Rituximab in untreated FL (n = 90) ∗

Outcomes

Significances

References

PFS at 5 yrs, VV (77%), VF (38%), FF (48%), F allele (41%)

P = 0.009

56

Response at 6 mo, HH (80%), HR (45%), RR (26%), R allele (43%) PFS duration, HH (445 days), HR or RR (158 days) Response, no difference PFS at 2 yrs, HH or HR (71%), RR (78%)

p = 0.005 p = 0.011

46

p = NS p = NS

51

Response and PFS, no difference

p = NS

53

PFS at 5 yrs, HH (49%), HR (48%), RR (43%)

p = NS

56

Response and PFS, no difference

p = NS

53

Response, no difference Duration of CR or PR, AA or AG (830 days), GG (250 days)

p = NS

72

Abbreviations: PFS, progression-free survival; FL, follicular lymphoma; EFS, event-free survival; CR, complete reponse; PR, partial response; NS, not significant

Chapter 12 / SNPs in Monoclonal Antibody Therapy

217

In addition, those with the HH genotype showed a longer period of time to progression compared with those with the R allele, (445 days in the HH genotype group vs. 158 days in the HR or RR genotype groups). Consistent with a previous result by Weng et al., Ghielmini et al. (49) reported that the genotype of FCGR3A is significantly associated with event-free survival (EFS) and response to rituximab monotherapy (375 mg/m2 weekly for 4 weeks). The median duration of EFS was 33.3 months in the VV genotype group, while that of the VF or FF genotype groups was 14.1 months (Table 2). Based on the above results, FCGR3A polymorphism influences the treatment outcome of rituximab monotherapy for FL in terms of response and progression-free survival. The Fc␥RIIb molecule is a potent regulator of ADCC in vivo by modulating the activity of Fc␥RIII on effector cells. The polymorphism of Fc␥RIIb may have an affect on the response of FL patients to rituximab therapy. However, regarding the polymorphism of inhibitory IgG Fc␥ receptor gene, FCGR2B, Weng et al. ( 50) reported no influence on treatment outcome by the FCGR2B-232 polymorphism. Out of 92 FL patients, the response rate of patients with I/I genotype and T-carrier (those with I/T or T/T genotypes) at FCGR2B-232 was similar: 65% vs. 60% at 1–3 months, 53% vs. 54% at 6 months, 41% vs. 54% at 9 months, and 34% vs. 38% at 12 months, respectively. Also, the progression-free survivals of both groups at 2 years were similar with 18% for the I/I genotype group and 36% for the T-carriers (p = 0.739).

5.1.2. S EQUENTIAL CHOP T HERAPY AND R ITUXIMAB As a frontline treatment for FL, the addition of rituximab to chemotherapy has significantly improved the clinical outcome of FL patients. Two studies were performed to evaluate the impact of FCGR3A polymorphism on the response to sequential CHOP and rituximab therapy in FL (Table 2). However, these two studies consistently reported that the FCGR3A polymorphism did not correlate with treatment outcome. Maloney et al. ( 51) reported that FCGR3A and FCGR2A polymorphism did not influence progression free survival of 87 FL patients treated with CHOP followed by rituximab (375 mg/m2× 4 weekly cycles). In addition, Carlotti et al. (52) also reported that the complete response rate in 88 previously untreated FL patients was not significantly different between groups with the VV genotype (76%) versus those with the VF or FF genotype (68%). These two reports suggest that the encouraging efficacy of CHOP therapy followed by rituximab for FL patients may be independent of Fc␥ receptor polymorphism (51,52). 5.1.3. C HEMOTHERAPY A LONE The potential implications of Fc␥ receptor polymorphism have been investigated by Weng et al. in terms of response to chemotherapy alone in FL patients (Table 2). In contrast with the results of rituximab monotherapy, no association of 3 Fc␥ receptor polymorphisms (i.e., FCGR3A-158, FCGR2A-131, and FCGR2B-232) has been shown with response to chemotherapy or time to progression after chemotherapy. Based on this finding, the impact of Fc␥ receptor polymorphism on clinical outcome after rituximab therapy for FL has to be justified (53).

218

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

5.1.4. I NTERLEUKIN -2 P LUS R ITUXIMAB T HERAPY Several trials have been performed to augment the activity of ADCC due to its role in rituximab activity. Interleukin-2 (IL-2) has been known to induce expansion and activation of Fc␥ receptor expressing effector cells, resulting in enhancement of ADCC (54). In addition, Milan et al. (55) reported that the FCGR3A gene polymorphism correlates with the response to rituximab plus IL-2 treatment (10 million IU, three times a week for 6–9 weeks or 14 million IU, three times a week for 2–5 weeks) in rituximab-resistant low-grade lymphoma patients. The role of IL-10 was suggested to induce expansion/activation of Fc receptor – bearing cells, resulting in enhancement of ADCC. In contrast to previous outcomes in rituximab monotherapy setting, the best response was observed in the FF FCGR3A158 genotype group in which 4 out of 17 patients responded. There were no responders in the VF or VV genotype groups. The administration of IL-2 together with rituximab seemed to restore the response to rituximab by achieving a critical threshold to drive ADCC in patients with the FF genotype at FCGR3A-158. 5.1.5. I DIOTYPE VACCINATION T REATMENT In addition to monoclonal antibody or chemotherapy, immunotherapy has been developed to target tumor antigen for the treatment of FL. The unique sequence of the protein, so-called idiotype (Id) protein, can be a target of immunotherapy. Because Id protein can induce humoral and cellular immune responses against idiotype protein, vaccination with Id protein can decrease the risk of progression. However, the mechanism of anti-Id response has been unclear. Recently, Weng et al. (56) demonstrated that patients exhibiting a humoral immune response after Id vaccination showed prolonged survival, and that a cellular immune response was not associated with better outcome. In addition, improved treatment outcome (especially in terms of progression-free survival) was noted in FL patients with the VV genotype of FCGR3A-158 in comparison to the VF or FF genotype groups (Table 2). This finding suggests that anti-Id immune response may be mediated through the ADCC mechanism. Based on this finding, efficient killing of FL cells by ADCC effector cells via Fc␥ receptor seemed to be affected by the binding capacity of effector cells to Fc␥RIIIa (56).

5.2. Diffuse Large B-cell Lymphoma The introduction of rituximab plus CHOP chemotherapy (R-CHOP) has significantly improved the treatment outcome of diffuse large B-cell lymphoma (DLBCL) (57). The benefit of R-CHOP therapy has been demonstrated in a randomized trial of R-CHOP versus CHOP in elderly patients with DLBCL (57). It has been proposed that the impact of R-CHOP on the outcomes for DLBCL may be derived from overcoming bcl-2 mediated chemoresistance (58). Chow et al. (59) also reported that combining rituximab with different cytotoxic drugs such as doxorubicin or mitoxantrone significantly decreased the dose of chemotherapeutic drugs necessary for the induction of apoptosis. In addition to overcoming bcl-2 related chemoresistance, it is likely that rituximab carries an anti-DLBCL effect by enhancing cellular cytotoxicity (60). Ansell et al. reported a high response rate in lymphoma trials

Chapter 12 / SNPs in Monoclonal Antibody Therapy

219

when rituximab was combined with non-chemotherapeutic agents such as IL-12 which can enhance cellular cytotoxicity. In this trial, the overall response rate was 64% (7 out of 11 patients) with 5 CRs in CD20 positive DLBCL patients. Recently, we reported that the FCGR3A gene polymorphism correlates with response to frontline R-CHOP therapy for DLBCL (61). The response rate correlates well with FCGR3A genotype (Table 3). The overall response rate was 98% in patients with the VV genotype for FCGR3A-158, while that in patients with VF and FF genotypes were 90% and 50%, respectively (p < 0.001; Fig. 4A). Complete response (CR) rate was also significantly different in favor of VV genotype group: 88% in the VV group, 79% in the VF group, and 50% in the FF group (p = 0.002). In addition, the group with the VV genotype for FCGR3A achieved CR or overall response faster than those with the VF or FF genotypes (Fig. 5). However, in terms of the FCGR2A gene polymorphism, no influence was noted in the response to R-CHOP in our study (Fig. 4B) (61). In contrast to the predictive impact of FCGR3A polymorphism on the response to R-CHOP therapy, the predictive value of FCGR3A gene polymorphism has been abolished in the setting of CHOP chemotherapy ( 61). This suggests that the mechanism mediated by the Fc␥ receptor may be involved in the action of R-CHOP chemotherapy for DLBCL. The overall (complete) response rates after CHOP chemotherapy were 71% for VV (59% of CR), 69% for VF (46%) and 70% (50%) for FF genotypes (p = 0.635). Accordingly, the additive effect of rituximab to CHOP chemotherapy can be explained by an effect of the FCGR3A genotype on ADCC activity. The NK cell–mediated ADCC may augment the chemotherapeutic effect of CHOP by binding of rituximab to NK cells through the Fc␥RIIIa receptor, resulting in release of granzyme and activation of caspases.

5.3. Waldenstrom’s Macroglobulinemia Waldenstrom’s macroglobulinemia (WM) is an uncommon lymphoid malignancy characterized by IgM monoclonal gammopathy and intertrabecular marrow infiltration by small lymphocytes expressing CD20 antigen. Currently, rituximab can be used for the treatment of WM with response rates between 20% and 50% (62,63). The impact of FCGR3A polymorphism on the response to rituximab for WM has been demonstrated by the work of Treon et al. (Table 3) (64). They reported that two polymorphisms, FCGR3A-43 and FCGR3A-158, correlate with response to rituximab therapy (375 mg/m2 weekly infusion), but not with time to progression (TTP) or progression-free survival (PFS). The response rate after rituximab therapy was significantly different in favor of the VV genotype of FCGR3A-158 polymorphism as follows: 40% in VV, 35% in VF, and 9% in FF genotype groups (p = 0.03). In addition, the response rate was also in favor of the LR or LH genotype of FCGR3A-48 polymorphism as follows: 38.5% in LH, 25% in LR, and 22% in LL genotype groups (p = 0.057). However, no significant difference in PFS or TTP was observed according to the FCGR3A-48 or FCGR3A-158 polymorphisms. To reach clear conclusion on the issue of whether the Fc␥ receptor gene polymorphism influences on treatment outcome of rituximab therapy for WM, further studies with larger number of patients are warranted.

Table 3 The Influence of Single Nucleotide Polymorphism on the Treatment Outcomes for Other Lymphoid Malignancies Besides Follicular Lymphoma Reference (year)

220

Treatment Modality

Polymorphism

Outcomes

Significances References

Diffuse large B-cell lymphoma Kim (2006) CHOP (n = 85) R-CHOP (n = 113)

FCGR3A-158 FCGR3A-158

p = NS p = 0.002 p = NS p = NS p = NS

61 61

FCGR2A-131

Response, VV (71%), VF (69%), FF (70%) Response, VV (88%), VF (79%), FF (50%) PFS, VV (81%), VF or FF (84%) OS, VV (94%), VF or FF (92%) Response, HH (95%), HR (92%), RR (75%)

FCGR3A-158

Response, VV (50%), VF (33%), FF (41.6%)

p = NS

69

FCGR2A-131 FCGR3A-158 FCGR2A-131

Response, HH (67%), HR (29%), RR (57%) Response, VV (25%), VF (40%), FF (32%) Response, HH (33%), HR (32%), RR (40%)

p = NS p = NS p = NS

70

FCGR3A-158

Response, VV (40%), VF (35%), FF (9%) PFS, TTP, no difference Response, LH (38.5%), LR (25%), LL (22%) PFS, TTP, no difference

p = 0.03

64

Chronic lymphocytic leukemia Farag (2004) Rituximab in untreated or pretreated CLL (n = 30) Lin (2005) Alemtuzumab in relapsed CLL (n = 36) Waldenstrom’s macroglobulinemia Treon (2005) Rituximab in untreated or pretreated WM (n = 58)

FCGR3A-48

Chapter 12 / SNPs in Monoclonal Antibody Therapy

221

Fig. 4. Response to frontline R-CHOP therapy for diffuse large B-cell lymphoma patients according to FCGR3A-158 and FCGR2A-131 polymorphisms. A higher rate of complete response was observed in the group with VV genotype (88%) compared to those with the VF (79%) or FF genotype (50%) for FCGR3A (p = 0.002). The difference in overall response rate was also significant favoring the patients with the VV genotype for FCGR3A (98%) versus those with the VF (90%) or FF genotype (50%; p < 0.001). B. According to the FCGR2A-131 polymorphism, however, a significant difference in response rate was not observed. The overall response rate of patients with the HH, HR, and RR genotype was 95%, 92%, and 75%, respectively (p = 0.137). (Originally published in Blood. Kim DH, Jung HD, Kim JG et al, eds. FCGR3A gene polymorphisms may correlate with response to frontline c Copyright R-CHOP therapy for diffuse large B-cell lymphoma. Blood, 2006; 108(8): 2720–2725.  American Society of Hematology).

222

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Fig. 5. The cumulative incidence of complete response to frontline R-CHOP therapy for diffuse large B-cell lymphoma according to the FCGR3A polymorphism. The time to achieve a complete response after R-CHOP was shorter in the group with the VV genotype (median 70 days, 95% C.I. [52–88 days]) compared to those with the VF or FF genotype for FCGR3A-158 (median, 99 days, 95% C.I. [60–138 days]; p = 0.020 by Wilcoxon test). This research was originally published in Blood. Kim DH, Jung HD, Kim JG et al. FCGR3A gene polymorphisms may correlate with response to frontline R-CHOP therapy for diffuse large B-cell lymphoma. Blood, 2006; 108(8): 2720–2725. c Copyright American Society of Hematology. 

5.4. Chronic Lymphocytic Leukemia Chronic lymphocytic leukemia (CLL) is a disorder of morphologically mature but immunologically immature lymphocytes expressing various degree of CD20 antigen. CLL is characterized by a progressive accumulation of CLL cells in the peripheral blood, marrow, and lymphoid tissues. Several kinds of chemotherapy have been adopted for the treatment of CLL including chlorambucil, fludarabine, and combination treatment with fludarabine and cyclophosphamide. Recently, two monoclonal antibodies have been introduced for the treatment of CLL, rituximab and alemtuzumab. The precise mechanisms of action remain unclear (35,65,66). In contrast to excellent response in other NHLs, the early outcome of rituximab treatment for CLL was disappointing ( 65). This lower response rate was attributed to the lower density of CD20 antigen on the surface of CLL cells, and a faster clearance rate of antibody from patients with CLL compared to other lymphoma patients (67,68). A substantial amount of soluble CD20, due to rapid turnover of CLL cells, may contribute

Chapter 12 / SNPs in Monoclonal Antibody Therapy

223

to increased clearance of rituximab from plasma by immune complex formation between soluble CD20 antigen and rituximab. In contrast to the good correlation between Fc␥R polymorphism and treatment outcomes in FL and WM, the influence of Fc␥R polymorphism on the response to rituximab or alemtuzumab in CLL has not been demonstrated (69,70).

5.4.1. R ITUXIMAB Farag et al. ( 69) reported that no differences in infusion toxicity or response were found according to FCGR3A or FCGR2A polymorphism after rituximab monotherapy (rituximab, stepped-up administration of rituximab from 100 mg to 375 mg/m2 during the first week and then given as described previously, three times a week for 4 weeks) in untreated or previously treated CLL patients (n = 30; Table 3). The response rate of the patients with VV, VF, and FF genotypes at FCGR3A-158 was 50%, 33%, and 42% (p = 0.78), while that of patients with HH, HR. and RR genotypes at FCGR2A-131 was 67%, 29%, and 57% (p = 0.70), respectively. The investigators concluded that unlike FL, a mechanism of tumor clearance other than ADCC may be more important in the action of rituximab toward CLL cells (69). Further studies with larger number of CLL patients are warranted to reach a clear understanding of the precise mechanism of rituximab activity on CLL cells. 5.4.2. A LEMTUZUMAB Similar to the result of FCGR polymorphism with rituximab therapy for CLL, Lin et al. ( 70) reported that the FCGR3A and FCGR2A polymorphism did not show any difference in response rate of relapsed CLL patients when treated with alemtuzumab (stepped-up dose schedule from 3 mg to 30 mg during the first week and then given at 30 mg three times a week for 12 weeks). The response rates of the patients with VV, VF, and FF genotypes at FCGR3A158 were not significantly different at 25%, 40%, and 32%, respectively (Table 3). Similar to the outcomes in rituximab therapy setting, there was also no difference in response rate according to the FCGR2A-131 polymorphism. The response rates of patients with HH, HA, and AA genotypes were 33%, 32%, and 40%, respectively (p = 0.70). In summary, the mechanism of monoclonal antibody activity for CLL might be different from that for lymphoma. Besides ADCC, other pathways such as caspase-dependent apoptosis or CDC are more likely to contribute to rituximab or alemtuzumab-induced clearance of CLL cells in vivo.

5.5. Rituximab Induced Neutropenia After Autologous Hematopoietic Stem Cell Transplantation for Lymphoma Recently, rituximab has been used as maintenance therapy after high-dose chemotherapy following autologous stem cell support. It can be given safely on a 4-week schedule. One of the side effects of maintenance rituximab therapy after autologous transplantation for lymphoma is severe neutropenia (neutrophil less than 1× 109 /L), which responds well to G-CSF support and has an incidence of 30%50%. Interestingly, the incidence of grade 3/4 neutropenia is well correlated with the FCGR3A gene polymorphism (Table 3).

224

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Weng et al. (71) reported that that out of 33 patients with NHL, 75% of patients with the VV genotype at FCGR3A-158 experienced severe neutropenia, while 57% and 33% of patients with the VF and FF genotype experienced severe neutropenia. The incidence of rituximab-induced neutropenia was significantly higher in the group with the VV genotype (71%) versus those with the VF or FF genotype (28%, p = 0.035). It may be possible that B-cell depletion by rituximab may affect cytokine production, which may influence the development of neutropenia (71).

6. POLYMORPHISM OF THE GENES IN THE COMPLEMENT PATHWAYS AND CLINICAL DATA There is increasing evidence that complement plays a role in the clinical response to rituximab. However, little data is available regarding the polymorphism of genes involved in the complement-mediated pathway. C1q (complement component 1, subcomponent q), generally associates with C1r and C1s to yield the first component of the complement system. C1q is composed of 18 polypeptide chains including 6 A-chains, 6 B-chains, and 6 C-chains. The A-, B-, and C-chains are arranged in the order A-C-B on chromosome 1. This gene encodes the A-chain polypeptide of C1q, and is located on chromosome 1p36.12. Racila et al. (72) recently identified a non-coding polymorphism in the C1qA component of complement, which may result in a post-translational splice variant of the C1qA protein. When analyzing the treatment outcome of rituximab monotherapy in 90 FL patients, no difference in response rate was noted according to C1qA gene polymorphism (Table 2). However, a trend of prolonged remission was noted in the group with the AA or AG genotype at the C1qA 276 locus among patients who achieved complete or partial response on rituximab therapy (p = 0.12). When confined to the group that achieved CR to rituximab monotherapy, patients with the AA or AG genotype at C1qA 276 locus showed a median of 830 days of time to progression (TTP), while those with GG genotype showed 250 days of TTP (p = 0.007). This data suggests that the polymorphism in C1qA may impact on the remission duration of FL patients after rituximab therapy (72). Further study regarding this polymorphism is strongly warranted.

7. FUTURE DIRECTIONS TO IMPROVE EFFICACY OF MONOCLONAL THERAPY Based on the findings we review in this chapter, how can we improve the clinical activity of rituximab against tumor cells? Given that ADCC activity is an important action of rituximab, several approaches have been proposed. First, higher-dose infusions of rituximab may overcome low binding affinity to rituximab, although clinical trials have not been done yet. Second, concurrent administration of cytokines or growth factors has been proposed to increase the activity of ADCC effector cells. Based on the finding that IL-2, IL-12, granulocyte colony-stimulating factor (GCSF), and granulocyte-macrophage colony-stimulating factor (GM-CSF) can enhance the ADCC activities of effector cells (73), several trials have demonstrated improved activity of effector cells by administration of cytokines or growth factor (54,74). However,

Chapter 12 / SNPs in Monoclonal Antibody Therapy

225

with regard to clinical response, it is difficult to make conclusions about the efficacy of this approach without further clinical trials. Another approach is to increase the binding affinity of the Fc portion in rituximab for the Fc␥ receptor on effector cells. Several kinds of genetically reengineered monoclonal antibodies have been developed and are under investigation. Modification of IgG glycosylation, which is necessary for the interaction of IgG with Fc␥ receptor, has been studied (43,75,76). Removal of fucose residues from the oligosaccharides of human IgG1, which makes a low-fucose version of rituximab, has been shown to improve binding to human Fc␥RIII (up to 50–100 fold) and ADCC activity in an in vitro model (43,75). Based on the requirement of N-Glycans at Asn-297 in the Fc domain of IgG for Fc receptor-mediated effector functions, N-glycan remodeling in vitro has been shown to increase ADCC activity of rituximab 10-fold (76). These two approaches of modifying glycosylation of rituximab Fc portion are promising. Recently, two groups reported a genetically reengineered version of rituximab to have enhanced affinity for the Fc␥ receptor. Bowles et al. (77) recently developed rituximab with varying affinity for Fc receptor and for CD20 antigen using a directed evolution approach. The reengineered version of rituximab with higher affinity to CD20 antigen and Fc␥ receptor was more effective at modulating CD16 (Fc␥RIII), activating NK cells and ADCC even in a group expressing lower affinity Fc␥R. Another approach is to use rituximab variants with re-engineered Fc portions (MacroGenics, Inc, Rockville, MD, USA) using single amino acid substitution by screening a yeast library containing randomly mutated Fc. Weng et al. (78) reported that the rituximab variants with re-engineered Fc showed increased interaction with Fc␥R on effectors and mediated ADCC more effectively than rituximab even with effectors of low-affinity genotypes of FCGR3A. Accordingly, these two types of reengineered versions of rituximab can be more effective toward CD20 (+) hematologic malignancies than rituximab, especially for patients with the FF genotype of FCGR3A-158. However, further clinical study to determine safety and clinical efficacy is strongly warranted.

REFERENCES 1. McLaughlin P, Grillo-Lopez JA, Link BK et al. Rituximab chimeric anti-CD20 monoclonal antibody therapy for relapsed indolent lymphoma: half of patients respond to a four-dose treatment program. J Clin Oncol 1998;16:2825–2833. 2. Habermann TM, Weller E, Morrison VA et al. Rituximab-CHOP versus CHOP with or without maintenance rituximab in patients 60 years of age or older with diffuse large B-cell lymphoma (DLBCL): an update. Blood 2004;104:127. 3. Cartron G, Watier H, Golay J et al. From the bench to the bedside: ways to improve rituximab efficacy. Blood 2004;104:2635–2642. 4. Smith MR. Rituximab (monoclonal anti-CD20 antibody): mechanisms of action and resistance. Oncogene 2003;22:7359–7368. 5. Binstadt BA, Geha RS, Bonilla FA. IgG Fc receptor polymorphisms in human disease: implications for intravenous immunoglobulin therapy. J Allergy Clin Immunol 2003;697–703. 6. van Sorge NM, van der Pol WL, van de Winkel JG. FcgammaR polymorphisms: Implications for function, disease susceptibility and immunotherapy. Tissue Antigens 2003;6:189–202. 7. Golay J, Zaffaroni L, Vaccari T et al. Biologic response of B lymphoma cells to anti-CD20 monoclonal antibody rituximab in vitro: CD55 and CD59 regulate complement-mediated cell lysis. Blood 2000;95:3900–3908.

226

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

8. Shan D, Ledbetter JA, Press OW. Apoptosis of malignant human B-cells by ligation of CD20 with monoclonal antibodies. Blood 1998;91:1644–1652. 9. Hofmeister JK, Cooney D, Coggeshall KM. Clustered CD20 induced apoptosis: src-family kinase, the proximal regulator of tyrosine phosphorylation, calcium influx, and caspase 3-dependent apoptosis. Blood Cells Mol Dis 2000;26:133–143. 10. Villamor N, Montserrat E, Colomer D. Mechanism of action and resistance to monoclonal antibody therapy. Semin Oncol 2003;30:424–433. 11. Jazirehi AR, Gan XH, De Vos S et al. Rituximab (anti-CD20) selectively modifies Bcl-xL and apoptosis protease activating factor-1 (Apaf-1) expression and sensitizes human non-Hodgkin’s lymphoma B cell lines to paclitaxel-induced apoptosis. Mol Cancer Ther 2003;2:1183–1193. 12. Jazirehi AR, Vega MI, Chatterjee D et al. Inhibition of the Raf-MEK1/2-ERK1/2 signaling pathway, Bcl-xL down-regulation, and chemosensitization of non-Hodgkin’s lymphoma B-cells by Rituximab. Cancer Res 2004;64:7117–7126. 13. Shan D, Ledbetter JA, Press OW. Signaling events involved in anti-CD20-induced apoptosis of malignant human B-cells. Cancer Immunol Immunother 2000;48:673–683. 14. Deans JP, Schieven GL, Shu GL et al. Association of tyrosine and serine kinases with the B-cell surface antigen CD20: induction via CD20 of tyrosine phosphorylation and activation of phospholipase C-gamma 1 and PLC phospholipase C-gamma 2. J Immunol 1993;151:4494–4504. 15. Pedersen IM, Buhl AM, Klausen P et al. The chimeric anti-CD20 antibody rituximab induces apoptosis in B-cell chronic lymphocytic leukemia cells through a p38 mitogen activated protein-kinasedependent mechanism. Blood 2002;99:1314–1319. 16. Janas E, Priest R, Wilde JI et al. Rituxan (anti-CD20 antibody)-induced translocation of CD20 into lipid rafts is crucial for calcium influx and apoptosis. Clin Exp Immunol 2005;139:439–446. 17. Aydin F, Yilmaz M, Ozdemir F et al. Correlation of serum IL-2, IL-6 and IL-10 levels with International Prognostic Index in patients with aggressive non-Hodgkin’s lymphoma. Am J Clin Oncol 2002;25:570–572. 18. Bohlen H, Kessler M, Sextro M et al. Poor clinical outcome of patients with Hodgkin’s disease and elevated interleukin-10 serum levels: clinical significance of interleukin-10 serum levels for Hodgkin’s disease. Ann Hematol 2000;79:110–113. 19. Cortes J, Kurzrock R. Interleukin-10 in non-Hodgkin’s lymphoma. Leuk Lymphoma 1997;26: 251–259. 20. el-Far M, Fouda M, Yahya R et al. Serum IL-10 and IL-6 levels at diagnosis as independent predictors of outcome in non-Hodgkin’s lymphoma. J Physiol Biochem 2004;60(4):253–258. 21. Ozdemir F, Aydin F, Yilmaz M et al. The effects of IL-2, IL-6, and IL-10 levels on prognosis in patients with aggressive non-Hodgkin’s lymphoma (NHL). J Exp Clin Cancer Res 2004;23:485–488. 22. Salgami EV, Efstathiou SP, Vlachakis V et al. High pretreatment interleukin-10 is an independent predictor of poor failure-free survival in patients with Hodgkin’s lymphoma. Haematologia (Budap) 2002;32:377–387. 23. Sarris AH, Kliche KO, Pethambaram P et al. Interleukin-10 levels are often elevated in serum of adults with Hodgkin’s disease and are associated with inferior failure-free survival. Ann Oncol 1999;10: 433–440. 24. Stasi R, Zinzani PL, Galieni P et al. Prognostic value of serum IL-10 and soluble IL-2 receptor levels in aggressive non-Hodgkin’s lymphoma. Br J Haematol 1994;88:770–777. 25. Vassilakopoulos TP, Nadali G, Angelopoulou MK et al. Serum interleukin-10 levels are an independent prognostic factor for patients with Hodgkin’s lymphoma. Haematologica 2001;86:274–281. 26. Visco C, Vassilakopoulos TP, Kliche KO et al. Elevated serum levels of IL-10 are associated with inferior progression-free survival in patients with Hodgkin’s disease treated with radiotherapy. Leuk Lymphoma 2004;45:2085–2092. 27. Viviani S, Notti P, Bonfante V et al. Elevated pretreatment serum levels of Il-10 are associated with a poor prognosis in Hodgkin’s disease, the Milan cancer institute experience. Med Oncol 2000;17: 59–63.

Chapter 12 / SNPs in Monoclonal Antibody Therapy

227

28. Voorzanger N, Touitou R, Garcia E et al. Interleukin (IL)-10 and IL-6 are produced in vivo by nonHodgkin’s lymphoma cells and act as cooperative growth factors. Cancer Res 1996;56:5499–5505. 29. Weber-Nordt RM, Henschler R, Schott E et al. Interleukin-10 increases bcl-2 expression and survival in primary human CD34+ hematopoietic progenitor cells. Blood 1996;88:2549–2558. 30. Alas S, Bonavida B. Rituximab inactivates signal transducer and activation of transcription 3 (STAT3) activity in B-non-Hodgkin’s lymphoma through inhibition of the interleukin 10 autocrine/paracrine loop and results in down-regulation of Bcl-2 and sensitization to cytotoxic drugs. Cancer Res 2001;61:5137–5144. 31. Alas S, Bonavida B. Inhibition of constitutive STAT3 activity sensitizes resistant non-Hodgkin’s lymphoma and multiple myeloma to chemotherapeutic drug-mediated apoptosis.Clin Cancer Res 2003;9:316–326. 32. Alas S, Emmanouilides C, Bonavida B. Inhibition of interleukin-10 by rituximab results in downregulation of bcl-2 and sensitization of B-cell non-Hodgkin’s lymphoma to apoptosis. Clin Cancer Res 2001;7:709–723. 33. Treumann A, Lifely MR, Schneider P et al. Primary structure of CD52. J Biol Chem 1995;270: 6088–6099. 34. Mone AP, Cheney C, Banks AL et al. Alemtuzumab induces caspase-independent cell death in human chronic lymphocytic leukemia cells through a lipid raft-dependent mechanism. Leukemia 2006;20:272–279. 35. Keating M, Coutre S, Rai K et al. Management guidelines for use of alemtuzumab in B-cell chronic lymphocytic leukemia. Clin Lymphoma 2004;4:220–227. 36. Montillo M, Schinkoethe T, Elter T. Eradication of minimal residual disease with alemtuzumab in B-cell chronic lymphocytic leukemia (B-CLL) patients: the need for a standard method of detection and the potential impact of bone marrow clearance on disease outcome. Cancer Invest 2005;23:488– 496. 37. Hernandez MC, Knox SJ. Radiobiology of radioimmunotherapy with 90Y ibritumomab tiuxetan (Zevalin). Semin Oncol 2003;30:6–10. 38. Witzig TE. Efficacy and safety of 90Y ibritumomab tiuxetan (Zevalin) radioimmunotherapy for nonHodgkin’s lymphoma. Semin Oncol 2003;30:11–16. 39. Weigert O, Illidge T, Hiddemann W et al. Recommendations for the use of Yttrium-90 ibritumomab tiuxetan in malignant lymphoma. Cancer 2006;107:686–695. 40. Wiseman GA, Witzig TE. Yttrium-90 (90Y) ibritumomab tiuxetan (Zevalin) induces long-term durable responses in patients with relapsed or refractory B-cell non-Hodgkin’s lymphoma. Cancer Biother Radiopharm 2005;20:185–188. 41. van der Pol W, van de Winkel G. IgG receptor polymorphisms: risk factors for disease. Immunogenetics 1998;48:222–232. 42. Wu J., Edberg JC, Redecha PB et al. A novel polymorphism of FcgammaRIIIa (CD16) alters receptor function and predisposes to autoimmune disease. J Clin Invest 1997;100:1059–1070. 43. Shields RL, Lai J, Keck R et al. Lack of fucose on human IgG1 N-linked oligosaccharide improves binding to human Fcgamma RIII and antibody-dependent cellular toxicity. J Biol Chem 2002;277:26733–26740. 44. Dall’Ozzo S, Tartas S, Paintaud G et al. Rituximab-dependent cytotoxicity by natural killer cells: influence of FCGR3A polymorphism on the concentration–effect relationship. Cancer Res 2004;64: 4664–4669. 45. Hatjiharissi E, Santo DD, Xu L et al. Individuals expressing Fcgamma-RIIIA-158 V/V and V/F show increased NK cell surface expression of FcgRIIIA (CD16), rituximab binding, and demonstrate higher levels of ADCC activity in response to rituximab. Blood 2005;106:776. 46. Weng WK, Levy R. Two immunoglobulin G fragment C receptor polymorphisms independently predict response to rituximab in patients with follicular lymphoma. J Clin Oncol 2003;21:3940–3947. 47. Hatjiharissi E, Hansen M, Verselis S et al. Polymorphisms in Fcgamma-RIIIA are genetically linked to Fcgamma-RIIA and may account for the primary predictive role ascribed to polymorphisms in Fcgamma-RIIIA-158 in determining rituximab responses. Blood 2005;106:684a.

228

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

48. Cartron G, Dacheux L, Salles G et al. Therapeutic activity of humanized anti-CD20 monoclonal antibody and polymorphism in IgG Fc receptor FcgammaRIIIa gene. Blood 2002;99:754–758. 49. Ghielmini M, Schmitz SFH, Leger-Falandry C et al. The genotype of the IgG Fc receptor is predictive of event-free survival after treatment with rituximab in patients with follicular lymphoma participating in study SAKK 35/98. Blood 2003;102:409a. 50. Weng WK, Levy R. Genetic polymorphism of the inhibitory IgG Fc receptor FcRIIb is not associated with clinical outcome of rituximab treated follicular lymphoma patients. Blood 2005;106:683a. 51. Maloney DG, Pender-Smith B, Unger JM et al. Fc receptor polymorphisms do not influence progression-free survival (PFS) of follicular NHL patients treated with CHOP followed by rituximab (SWOG 9800). Blood 2004;104:170a. 52. Carlotti E, Palumbo GA, Oldani E et al. Bone marrow BCL2/IgH+ cells at diagnosis and not FCGRIIIA polymorphism predict response in follicular non-Hodgkin’s lymphoma patients treated with sequential CHOP and rituximab. Blood 2005;106:289–290a. 53. Weng WK, Rosenberg A, Levy R. Immunoglobulin G Fc receptor polymorphisms and clinical course in follicular lymphoma patients. Blood 2004;104:887a. 54. Gluck WL, Hurst D, Yuen A et al. Phase I studies of interleukin (IL)-2 and rituximab in B-cell nonhodgkin’s lymphoma: IL-2 mediated natural killer cell expansion correlations with clinical response. Clin Cancer Res 2004;10:2253–2264. 55. Milan S, Wilson SE, Kahn KD et al. Investigation of FcgammaR polymorphisms and response to R ) and rituximab treatment in rituximab-resistant NHL patients: importance of the IL-2 (Proleukin F/F polymorphism at position 158 of the Fcgamma. Blood 2004;104:239b. 56. Weng WK, Czerwinski D, Timmerman J et al. Clinical outcome of lymphoma patients after idiotype vaccination is correlated with humoral immune response and immunoglobulin G Fc receptor genotype. J Clin Oncol 2004;22:4717–4724. 57. Feugier, P, Van Hoof A, Sebban C et al. Long-term results of the R-CHOP study in the treatment of elderly patients with diffuse large B-cell lymphoma: a study by the Groupe d’Etude des Lymphomes de l’Adulte. J Clin Oncol 2005;23:4117–4126. 58. Mounier N. Briere J, Gisselbrecht C et al. Rituximab plus CHOP (R-CHOP) overcomes bcl-2— associated resistance to chemotherapy in elderly patients with diffuse large B-cell lymphoma (DLBCL). Blood 2003;101:4279–4284. 59. Chow KU, Sommerlad WD, Boehrer S et al. Anti-CD20 antibody (IDEC-C2B8, rituximab) enhances efficacy of cytotoxic drugs on neoplastic lymphocytes in vitro: role of cytokines, complement, and caspases. Haematologica 2002;87:33–43. 60. Ansell SM, Witzig TE, Kurtin PJ et al. Phase 1 study of interleukin-12 in combination with rituximab in patients with B-cell non-Hodgkin lymphoma. Blood 2002;99:67–74. 61. Kim DH, Jung HD, Kim JG et al. FCGR3A gene polymorphisms may correlate with response to frontline R-CHOP therapy for diffuse large B-cell lymphoma. Blood 2006;108:2720–2725. 62. Johnson SA, Birchall J, Luckie C et al. Guidelines on the management of Waldenstrom macroglobulinaemia. Br J Haematol 2006;132:683–697. 63. Johnson SA. Advances in the treatment of Waldenstrom’s macroglobulinemia. Expert Rev Anticancer Ther 2006;6:329–334. 64. Treon SP, Hansen M, Branagan AR et al. Polymorphisms in FcgammaRIIIA (CD16) receptor expression are associated with clinical response to rituximab in Waldenstrom’s macroglobulinemia. J Clin Oncol 2005;23:474–481. 65. Ferrajoli A, Keating MJ. Current guidelines in defining therapeutic strategies. Hematol Oncol Clin North Am 2004;18:881–983, ix. 66. Oscier D, Fegan C, Hillmen P et al. Guidelines Working Group of the UK CLL Forum. British Committee for Standards in Haematology. Guidelines on the diagnosis and management of chronic lymphocytic leukaemia. Br J Haematol 2004;125:294–317. Review 67. van Meerten T, van Rijn RS, Hol S et al. Complement-induced cell death by rituximab depends on CD20 expression level and acts complementary to antibody-dependent cellular cytotoxicity. Clin Cancer Res 2006;12:4027–4035.

Chapter 12 / SNPs in Monoclonal Antibody Therapy

229

68. Kennedy AD, Beum PV, Solga MD et al. Rituximab infusion promotes rapid complement depletion and acute CD20 loss in chronic lymphocytic leukemia. J Immunol 2004;172:3280–3288. 69. Farag, SS, Flinn IW, Modali R et al. Fc gamma RIIIa and Fc gamma RIIa polymorphisms do not predict response to rituximab in B-cell chronic lymphocytic leukemia. Blood 2004;103:1472–1474. 70. Lin TS, Flinn IW, Modali R et al. FCGR3A and FCGR2A polymorphisms may not correlate with response to alemtuzumab in chronic lymphocytic leukemia. Blood 2005;105:289–291. 71. Weng WK., Horning SJ, Negrin RS et al. Immunoglobulin G Fc polymorphism is correlated with rituximab-induced neutropenia following autologous hematopoietic cell transplantation. Blood 2004;104:129a. 72. Racila E, Weng WK, Wooldridge JE et al. A polymorphism in the C1qA component of complement correlates with prolonged complete remission following rituximab therapy of follicular lymphoma. Blood 2005;106:229a. 73. Sashida G, Takaku TI, Honda S et al. Granulocyte colony-stimulating factor (G-CSF) could enhance Fcgamma receptor expression in neutrophils of patients with B-cell lymphoma treated with rituximab. Leuk Lymphoma 2005;46:789–791. 74. Niitsu N, Hayama M, Okamoto M et al. Phase I study of Rituximab-CHOP regimen in combination with granulocyte colony-stimulating factor in patients with follicular lymphoma. Clin Cancer Res 2004;10:4077–4082. 75. Niwa R, Hatanaka S, Shoji-Hosaka E et al. Enhancement of the antibody-dependent cellular cytotoxicity of low-fucose IgG1 Is independent of FcgammaRIIIa functional polymorphism. Clin Cancer Res 2004;10:6248–6255. 76. Hodoniczky J, Zheng YZ, James DC. Control of recombinant monoclonal antibody effector functions by Fc N-glycan remodeling in vitro. Biotechnol Prog 2005;21:1644–1652. 77. Bowles JA, Wang SY, Link BK et al. Anti-CD20 monoclonal antibody with enhanced affinity for CD16 activates NK cells at lower concentrations and more effectively than rituximab. Blood 2006;108:2648–2654. 78. Weng WK, Stavenhagen J, Koenig S et al. Rituximab variants with re-engineered Fc with higher affinity to activating Fc R eliminate the functional difference between Fc R genotypes. Blood 2005;106:105a.

“This page left intentionally blank.”

13

DNA Repair and Mitotic Checkpoint Genes as Potential Predictors of Chemotherapy Response in Non-Small-Cell Lung Cancer Rafael Rosell, MD, Miquel Taron, PhD, Mariacarmela Santarpia, MD, Fernanda Salazar, PhD, Jose Luis Ramirez, PhD, and Miguel Angel Molina, PhD CONTENTS Introduction BRCA1: A Potential Predictive Marker for Cis platin and Docetaxel BRCA1 in the Trans criptional Regulation of Spindle Checkpoint Genes Clinical Trials of Cus tomized Chemotherapy References

S UMMARY Metastatic stage IV non-small-cell lung cancer (NSCLC) has a grim outcome. The median survival does not exceed 11 months, with no differences according to different cisplatin-based regimens. However, this clinical observation is spurious from the molecular point of view, bewcausse DNA repair genes involved in several pathFrom: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 13, 

231

232

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

ways act as differential modulators of chemosensitivity. Low expression of ERCC1 or BRCA1 predicts higher sensitivity to cisplatin and other DNA-damaging agents, such as etoposide and mitomycin, while conferring resistance to antimicrotubule drugs, such as vinblastine and paclitaxel. In contrast, high expression of these genes confers resistance to cisplatin and sensitivity to antimicrotubules. Although this principle has yet to be demonstrated in clinical studies, it can be postulated that patients with low expression of nucleotide excision repair (NER) or NER-related genes may be good candidates for cisplatin plus either gemcitabine, etoposide, or mitomycin plus ifosfamide for optimal survival. On the same grounds, patients whose tumors have relatively high expression of these genes can still benefit from cisplatin combined with antimicrotubule drugs. The same premises of customization can be applied to early NSCLC. Various clinical observations indicate that when these defense NER genes are overexpressed in resected NSCLCs, they reflect a lower risk of relapse and increased cisplatin resistance. Several layers of evidence show that BRCA1 can be the most important predictive marker for customizing chemotherapy. In addition, the clinical value of several mitotic checkpoint genes that are dysfunctional in NSCLC and regulated by BRCA1 should also be investigated. Key Words: Non-small-cell lung cancer; DNA repair; ERCC1; BRCA1; Mad2; BubR1; MZF1; mitotic checkpoint

1. INTRODUCTION Cisplatin doublets remain the standard treatment for stage IV non-small-cell lung cancer (NSCLC) patients. Large randomized studies have demonstrated the equivalence of several cisplatin doublets, including gemcitabine and docetaxel, with meager response rates and grim survival outcomes. The median survival is commonly less than one year (1). Intriguingly, responses varied significantly among individual patients (2), highlighting the need for molecular predictive markers for cisplatin and docetaxel response and survival in NSCLC. Tobacco accounts for about 90% of lung cancers. Some tobacco carcinogens induce DNA adducts that are repaired by the nucleotide excision repair (NER) pathway ( 3). The lowest DNA repair capacity (DRC), measured in peripheral blood lymphocytes by the host cell reactivation assay (HCR), was observed in lung cancer patients who were less than 60 years old, female, or lighter smokers, and in those with a family history of cancer (4,5). Growing evidence shows the association between DRC phenotype, genetic polymorphisms of the NER genes, and the risk for tobacco-related cancers, including NSCLC (6,7,8,9). There are some hints that polymorphisms in DNA repair genes can modulate survival in NSCLC patients treated with cisplatin-based therapy ( 10,11). However, in order to validate customized cisplatin treatment based on DNA repair gene polymorphisms, patient age and cumulative cigarette smoking history need to be considered for a correct interpretation of the influence of the polymorphic variants on DRC (9).

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

233

1.1. Biological Background NER, a highly versatile pathway for DNA damage removal, is often dysfunctional in NSCLC and might therefore be the Achilles’ heel for customizing chemotherapy. NER removes numerous types of DNA helix-distorting lesions, including cisplatin- and ultraviolet-induced photo products (12). Inherited defects in the NER process cause serious repair disorders: xeroderma pigmentosum (XP), with extreme risk of ultraviolet-induced skin cancer, and Cockayne syndrome. NER functions by a “cut-and-paste” mechanism in which cisplatin damage recognition, local opening of the DNA helix around the lesion, damage excision, and gap-filling occur in successive steps ( 12) (Fig. 1). NER is composed of two subpathways: global genome NER (GG-NER) and transcription-coupled NER (TC-NER) share the same core mechanism but differ in the way lesions are recognized (13). The first step in GG-NER is damage recognition by the heterodimer XPC/hHR23B, which binds with higher affinity to helix-distorting DNA lesions than to non-damaged double stranded DNA ( 12). GG-NER of lesions that only mildly disturb the helical structure is greatly enhanced by the damaged DNA binding complex (DDB). Various NER factors, including transcription factor IIH (TFIIH, a general transcription factor for RNA polymerase II), XPA, replication protein A (RPA), and XPG work together to repair DNA damage (Fig. 1). The structure-specific endonuclease excision repair cross-complementing 1 (ERCC1) performs an essential late step in the NER process, where it nicks the damaged DNA strand at the 5 site of the helix-distorting cisplatin lesion. A significant contribution of the ERCC1 subunit to NER is interaction with the XPA protein that binds the lesion in conjunction with the strand-opening helicase complex TFII and the single strandbinding protein RPA (12,13). ERCC1 directs its XPF partner to a site of NER action and contributes to the correct positioning of the strand incisions (14). [See Wijnhoven et al. (13) for a detailed review.] NER-related mouse models of lung cancer have been identified ( 13,15). DDB2deficient mice also developed lung tumors (16). DDB2, a subunit of DDB, which binds to cisplatin-damaged DNA, has been linked to XPE, and mice targeted for knockout of XPC also showed an enhanced frequency of lung tumors and lung adenocarcinomas (17). ERCC1-targeted knockout mice display growth defects and premature senescence and an accumulation of endogenously generated DNA interstrand crosslinks which are normally repaired by ERCC1 (18). Importantly, the ERCC1/XPF structure-specific nuclease has an additional role in the repair of cisplatin adducts besides its function in NER: the recombinational repair of interstrand crosslinks (19). ERCC1- or XPF-deficient hamster mutant cell lines are hypersensitive to DNA crosslink agents, much more so than to ultraviolet-induced pyrimide dimers, the classical substrates for NER (20,21). Moreover, co-localization of ERCC1 foci and RAD51 foci in response to cisplatin treatment has recently been found and may represent recruitment of ERCC1/XPF to sites of recombination repair ( 22). Previous studies have shown that BRCA1, involved in homologous recombination repair, also plays a major role in the repair of cisplatin DNA damage (23). High tumor tissue levels of ERCC1 mRNA in ovarian and gastric cancer patients have been associated with cisplatin resistance (24,25). Similarly, inhibition of ERCC1 expression has been significantly associated with reduced HCR of cisplatin-treated cells

234 Fig. 1. Description of nucleotide excision repair pathways.

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

235

and their increased cisplatin sensitivity (26,27). Cisplatin resistance in NSCLC cell lines has also been related to the increase of HCR ( 28), and significant differences in survival were observed in cisplatin-treated NSCLC patients according to their DRC (29). When intratumoral ERCC1 mRNA derived from paraffin-embedded tumor specimens was measured by real-time reverse transcriptase polymerase chain reaction (RT-PCR) in metastatic colon cancer patients treated with oxaliplatin and 5-fluorouracil (5-FU), high levels of ERCC1 significantly correlated with poor response and shorter survival (30). We observed longer survival and a trend toward improved response in gemcitabine/ cisplatin-treated stage IV NSCLC patients with low ERCC1 mRNA levels (31). A higher response rate was observed in locally advanced NSCLC patients with low levels of ribonucleotide reductase subunit M1 (RRM1) and ERCC1 treated with induction gemcitabine/carboplatin (32). ERCC1 protein expression by immunostaining was associated with survival in NSCLC patients treated with cisplatin-based adjuvant therapy (33). However, GG-NER, may not correctly detect cisplatin DNA adducts, because it has been shown to possess a low affinity for these adducts (34,35). On the other hand, defects in TC-NER (Fig. 1) render cells markedly hypersensitive to cisplatin ( 36). BRCA1deficient cells are hypersensitive to cisplatin (37). Unlike ERCC1, BRCA1 is involved in TC-NER (38,39) and could be a better predictive marker of cisplatin response. BRCA1 may influence TC-NER via its ability to act as an ubiquitin ligase through its association with BARD1 (40). BRCA1 and BARD1 heterodimers also regulate centrosome function (41) (Fig. 2). The function of BRCA1 in GG-NER differs in that BRCA1 is thought to control the transcriptional up-regulation of a number of key GG-NER proteins, such as Growth Arrest and DNA Damage 45 (GADD45), XPC and DDB2 (42) (Fig. 3). Interestingly, BRCA1 is a component of the core transcriptional machinery and can act either as a coactivator or corepressor of transcription involving proteins implicated in chromatin remodeling such as the histone deacetylases HDAC1 and HDAC2 ( 40). BRCA1 is also a transcriptional regulator of several genes associated with the G2/M checkpoint. It regulates GADD45, which in turn regulates the cyclinB-cdc2 complex. By using inducible expression of GADD45, it was shown that BRCA1 also plays a role in regulating the G2/M checkpoint in response to antimicrotubule drugs (43) (Fig. 3). BRCA1 also regulates the chaperone protein 14-3-3␴, which targets cdc25C and sequesters it in the cytoplasm following DNA damage to prevent it from activating the cyclinB-cdc2 kinase complex ( 40) (Fig. 3). 14-3-3␴ was found methylated in the circulating serum DNA of one-third of gemcitabine/cisplatin-treated stage IV NSCLC patients, and methylation was strongly associated with significantly longer survival (44).

1.2. Clinical Findings Related to NER Genes in NSCLC Several retrospective analyses in stage IV gemcitabine/cisplatin-treated NSCLC show that patients with NER-defective tumors (e.g., low ERCC1 or RRM1 mRNA levels) have a median survival of 15 months. However, the predictive value of low ERCC1 mRNA levels found in our original study ( 31) was not borne out by our second study with Italian patients, in which low ERCC1 levels showed a non-significant trend towards better survival (unpublished data).

236

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Fig. 2. BRCA1 involved in regulation of mitotic checkpoint genes. (Reprinted with permission from Mullan et al. BRCA1: A good predictive marker of drug sensitivity in breast cancer treatment? Biochim Biophys Acta, in press,with kind permission from Elsevier).

Conversely, in these same Italian patients, low RRM1 mRNA levels were significantly associated with improved survival (15.5 vs. 6.8 months; p = 0.002) (45). No differences were found according to RRM1 status in patients treated with paclitaxel/carboplatin or vinorelbine/cisplatin as part of the original phase III randomized trial (46). While no differences between the three different cisplatin doublets were found in the original trial, RRM1 identified patients with better survival in the gemcitabine/cisplatintreated arm, which can be attributed to the effect of RRM1 on gemcitabine metabolism and on the NER pathway. Low RRM1 levels were associated with a significantly better survival (47) in Spanish NSCLC patients treated with gemcitabine/cisplatin as part of a large phase III randomized trial (48). However, the predictive value of RRM1 was not evident in the group of patients who received gemcitabine/cisplatin/vinorelbine triplets, raising the hypothesis that antimicrotubule drugs act on the NER pathway in a different way, as explained below. In another study (49) of patients treated with neoadjuvant gemcitabine/cisplatin followed by surgery, the lowest RRM1 mRNA levels (bottom quartile) predicted significantly better survival in comparison with those with higher levels of RRM1, while no differences were observed according to ERCC1 or XPD mRNA levels. Furthermore, in this group of patients, survival was significantly better for those in the bottom quartile of BRCA1 mRNA expression ( 50). In all these studies, the quantification of the gene transcripts was performed with mRNA isolated from paraffin-embedded tumor tissue.

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

237

Fig. 3. BRCA1, a master gene involved in DNA repair, transcriptional regulation and activation of mitotic checkpoints. (Reprinted with permission from Mullan et al. BRCA1: A good predictive marker of drug sensitivity in breast cancer treatment? Biochim Biophys Acta, in press, with kind permission from Elsevier).

It has also been documented that NER or NER-related genes can be prognostic and predictive markers of response to cisplatin-based chemotherapy. In the adjuvant chemotherapy setting, negative ERCC1 by immunohistochemical analysis predicted shorter survival in the control group but longer survival in patients treated with adjuvant cisplatin-based chemotherapy. In contrast, patients with positive ERCC1 expression had better survival in the control group and slightly worse survival when treated with adjuvant chemotherapy (33). Arsenic exposure contributes to cancer risk and has been associated with decreased ERCC1 expression in isolated lymphocytes at the mRNA and protein levels (51). Along the same lines, increased expression of defense genes, like ERCC1, has been observed in premalignant lesions, such as colorectal adenomas (52). In early-stage NSCLC, high ERCC1 expression has been observed in almost half of the patients, indicating both a lower risk of relapse and a lower probability of benefit from adjuvant cisplatin-based chemotherapy. In completely resected chemonaive NSCLC patients, risk of relapse increased as ERCC1 mRNA levels decreased ( 53). The same investigators found that low RRM1 levels in resectable NSCLC patients significantly increased the risk of relapse ( 54). These important clinical observations establish the bases for customizing adjuvant treatment in early NSCLC. Patients whose tumors have low NER gene expression have a higher risk of relapse and can be exquisitely sensitive to combinations such as

238

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

gemcitabine/cisplatin or etoposide/cisplatin, as further explained below. Clinical trials of customized chemotherapy in stage IB following these principles are highly warranted.

2. BRCA1: A POTENTIAL PREDICTIVE MARKER FOR CISPLATIN AND DOCETAXEL BRCA1 plays multiple roles not only in DNA damage repair but also in cell cycle regulation, transcriptional control, ubiquitination and apoptosis. (See Kennedy et al. (55) and Mullan et al. (40) for complete reviews.) BRCA1-deficient breast cancer cells are hypersensitive to a wide range of DNA damaging agents, such as cisplatin and mitomycin C (23). This hypersensitive phenotype can be reversed by expression of wild-type BRCA1. In contrast, BRCA1 deficiency confers resistance to antimicrotubule agents, such as paclitaxel and vinorelbine ( 56, 57, 58). This resistant phenotype is associated with a defective apoptotic response to these drugs in BRCA1-deficient cells, suggesting that BRCA1 could regulate apoptotic pathways (57). BRCA1 is required for paclitaxelinduced activation of the mitogen-activated protein kinase kinase 3 (MEKK3), providing further support for a role of BRCA1 in regulating apoptosis (59) (Fig. 2). Intriguingly, MEKK3 overexpression confers resistance to paclitaxel by decreasing the apoptotic response (60).

2.1. BRCA1, DNA Damaging Chemotherapy, and Antimicrotubule Drugs Reconstitution of full-length BRCA1 into mouse embryonic fibroblast cells with a disrupted BRCA1 led to an increase in resistance to several DNA damaging agents, including the platinum compounds carboplatin and oxaliplatin, the topoisomerase I drugs irinotecan and topotecan, and the topoisomerase II drugs doxorubicin and etoposide (61). HCC1937 breast cancer cells with a single copy of mutated BRCA1 were much more sensitive to a range of DNA damaging agents in comparison with HCC1937 cells reconstituted with wild-type BRCA1. In the same cell line model, BRCA1 also induced sensitivity to the antimicrotubule drugs paclitaxel and vinorelbine (57). These data suggest that BRCA1 functions as a differential modulator of chemotherapy-induced apoptosis depending on the nature of the cellular stress ( 40). BRCA1 dysfunction might be an Achilles heel for testing new drugs. Inhibitors of poly(ADP-ribose) polymerase (PARP) highly sensitize tumor cells harboring BRCA1 dysfunction (62,63). Breast cancer patients carrying BRCA1 mutations attain better response to anthracycline-based chemotherapy than patients without mutations (64). Along the same lines, we determined BRCA1 mRNA levels by RT-QPCR in 55 surgically resected NSCLC patients who had received neoadjuvant gemcitabine/cisplatin and divided the gene expression levels into quartiles. Median survival was not reached for the 15 patients in the bottom quartile, whereas for the 28 in the two middle quartiles, it was 37.8 months and for the 12 patients in the top quartile, it was 12.7 months. We concluded that low levels of BRCA1 confer high sensitivity to gemcitabine/cisplatin and hypothesized that those patients in the top quartile of BRCA1 expression with such dismal survival might have a better prognosis if treated with antimicrotubule agents instead of gemcitabine/cisplatin ( 50). In contrast to what occurs in breast and ovarian cancer, BRCA1 is rarely methylated in NSCLC. We found methylation in 4% of

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

239

NSCLC patients, mainly in adenocarcinomas, the same frequency as had been previously reported (65). Preclinical evidence indicates that knockdown of BRCA1 by small interference RNA (siRNA) in T47D cells leads to a 50-fold increase in resistance to paclitaxel. The increased sensitivity to paclitaxel observed in BRCA1 reconstituted cells correlated with the activation of the mitotic checkpoint and an increase in apoptosis (40,55).

3. BRCA1 IN THE TRANSCRIPTIONAL REGULATION OF SPINDLE CHECKPOINT GENES Several studies have reported that BRCA1 is involved in the transcriptional regulation of several mitotic checkpoint genes. The precise segregation of chromosomes during mitosis is mediated by the attachment of mitotic spindles to each pair of similar chromosomes at their kinetochores. Sister chromatid separation initiates after the activation of APC/C (anaphase-promoting complex/cyclosome) by cdc20 (cell division cycle 20) during metaphase to anaphase transition. Degradation of securin releases the inhibition on separase, a protease, which in turn cleaves cohesin, a protein complex that holds together the two sister chromatids (66) (Fig. 4). Destruction of cohesin allows the spindle microtubules to pull the separated chromatids to opposite poles of the cell. Failure of spindle attachment to a single kinetochore activates the SAC (spindle assembly checkpoint), which arrests cells at metaphase until corrections are effected and equal distribution of chromosomes has been ensured. A sensory mechanism initiates the “wait anaphase” signal from an unattached kinetochore and triggers the accu mulation of the checkpoint components that comprise the Bub (budding uninhibited by benomyl)-Mad (mitotic arrest deficient) families of proteins. These proteins form complexes with cdc20, thereby sequestering it from activating the APC/C complex ( 66) (Fig. 4). The spindle checkpoint is activated in response to various spindle poisons, such as nocodazole, a drug that depolymerizes microtubules and thus prevents the attachment of microtubules to the kinetochores. On the other hand, taxanes inhibit the dynamic instability of the spindle and allow microtubule attachment but prevent the generation of tension across kinetochores. Treatment of cells with antimicrotubule drugs inhibits chromosome alignment and results in a mitotic arrest before anaphase, which is followed by the induction of apoptosis (67). According to the catalytic model for the generation of “wait anaphase” signal, Mad1 and Bub1 are mainly resident at the kinetochore, whereas Mad2 (free of Mad1), BubR1 (Bub related 1), Bub3 (free of Bub1), cdc20 and Mps1 dynamically change as part of the diffuse “wait anaphase” signal (66,68) (Fig. 4). Along with Mad2, other MCCs (mitotic checkpoint components) such as cdc20, BubR1, and Bub3 assemble at the unattached kinetochore and then release the signal in the form of Mad2-cdc20 and BubR1-Bub3-cdc20 complexes, which subsequently inhibit APC function (Fig. 4). In this regard, breast cancer MCF-7 cells exhibiting reduced levels of BRCA1 cannot sustain mitotic arrest when exposed to paclitaxel, and BRCA1 down-regulation leads to premature onset of anaphase by activating the APC/C (69). The spindle assembly checkpoint induced by paclitaxel is partially impaired when BRCA1 expression is repressed in MCF-7 cells. BRCA1 up-regulates BubR1 transcription, and BubR1 transcription and expression are significantly down-regulated in MCF-7

240 Fig. 4. A graphic representation of the phases of mitosis and the key mitotic checkpoints involved. (Reprinted with permission from Baker DJ, Chen J, van Deursen JMA. The mitotic checkpoint in cancer and aging: what have mice taught us? Curr Opin Cell Biol 2005;17:583–589, with kind permission from Elsevier).

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

241

cells exhibiting low levels of BRCA1 (69). Mouse embryo fibroblasts with a homozygous deletion of BRCA1 exon 11 exhibit a defect in the mitotic spindle checkpoint and show reduced expression of several mitotic spindle checkpoint genes, including Mad2, Bub1, and BubR1 (70). Down-regulation of Bub1 and BubR1 due to knockdown of endogenous BRCA1 in prostate (DU-145) and breast (MCF-7) cancer cells was identified by microarray analysis (71) (Table 1). Mitotic checkpoint defects have been found in many human cancers (72,73,74,75,76). While Mad2 gene mutations occur rarely in many kinds of cancer ( 74,77), aberrant reduction of Mad2 is often observed (72,74,78). The T47D breast cancer cell line that is sensitive to paclitaxel and nocodazole had reduced Mad2 expression and failed to arrest in mitosis after nocodazole treatment (72). Mitotic checkpoint-impaired human lung cancer cell lines are highly resistant to nocodazole ( 79,80). In adult T-cell leukemia, HTLV-1 Tax protein affects subcellular localization of Mad2 proteins, leading to failure of response to mitotic checkpoint and chemoresistance to microtubule inhibitors ( 81). A dominant-negative Bub1 impaired the mitotic checkpoint, allowing cells to escape from cell death induced by nocodazole (82). By Western blot analysis, T47D cells had less than one-third the amount of Mad2

Table 1 Mitotic Checkpoint Defects in Cancer Cell Lines and Chemotherapy Sensitivity Gene

Genetic Aberration

Phenotype

Sensitivity

BubR1 (Sudo 2004) BRCA1 (Bae 2005, Chabalier 2006) MAD2 (Li and Benezra 1996, Sudo 2004, Kienitz 2005) BRCA1 (Wang 2004)

Partial down-regulation Partial down-regulation by BRCA1 abrogation Partial down-regulation

CIN

Resistance (T) Resistance (T)

CIN

Resistance (T)

CIN

Resistance (cisplatin) Resistance (?)

MAD2 (Cheung 2005, Feung 2006) MAD2 (Hernando 2004) BRCA1 (Mullan 2001, Lafargue 2001, Quinn 2003, Tassone 2003, Taron 2004) CHFR (Scolnick and Halazonetis 2001, Chaturvedi 2002, Toyota 2003, Satoh 2003, Ogi 2005)

Partial down-regulation by BRCA1 deletion Partial down-regulation Overexpressed by inactivated RB gene Partial down-regulation

Silencing

Resistance (T) and sensitivity (Cisplatin) Sensitivity (T, Tax)

242

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

that was present in nocodazole- and paclitaxel-resistant cell lines (72). Surprisingly, decreased expression of spindle checkpoint genes has recently been associated with resistance to antimicrotubule drugs, such as paclitaxel (80,81,83,84), and to DNA damaging agents, such as etoposide and doxorubicin (85). Suppression of Mad2 and BubR1 in paclitaxel-treated cancer cells results in paclitaxel resistance ( 84). Reduced expression of Mad2, but not of Mad1, also led to paclitaxel resistance (86). In addition, decreased expression of Mad2 correlated with cisplatin resistance in nasopharyngeal carcinoma cell lines compared to cell lines with high levels of Mad2 (87). Along the same lines, high levels of Mad2 conferred sensitivity to cisplatin in testicular germ cell tumor cell lines (88) (Table 1). An abundance of preclinical data indicates that Mad2, and possibly BubR1, mRNA levels may be predictive markers of chemotherapy response related to BRCA1 expression (Fig. 2), a hypothesis that warrants examination in clinical studies. Mad2 is significantly overexpressed in cancers in which the retinoblastoma gene is inactivated and E2F is constitutively activated (77). Similarly, Mad1 is up-regulated, rather than downregulated, in cancer and activated in gain-of-function p53 mutants (89). Importantly, not only the aberrant reduction but also the increase of these mitotic checkpoint proteins seems to be a major cause for mitotic abnormality or result from the loss-of-mitotic checkpoint control.

4. CLINICAL TRIALS OF CUSTOMIZED CHEMOTHERAPY ERCC1 has been associated with cisplatin resistance, providing a testable hypothesis for customizing therapy. From August 2001 to October 2005, 444 stage IV NSCLC patients were enrolled in a Spanish Lung Cancer Group randomized trial of customized chemotherapy based on ERCC1 mRNA levels. RNA was isolated from pretreatment biopsies, and quantitative real-time reverse transcriptase PCR assays were performed to determine ERCC1 mRNA expression. Results were available in 8 days. Patients in the control arm received docetaxel plus cisplatin. Patients in the genotypic arm received treatment based on ERCC1 mRNA levels: those with low levels received docetaxel plus cisplatin; those with high levels received non-cisplatin-based treatment (docetaxel plus gemcitabine) (90). The primary endpoint was the overall objective response rate. Of the 444 patients enrolled in the study, 78 patients (17.6%) went off-study prior to receiving one cycle of chemotherapy. The main reason for withdrawal was insufficient tumor tissue for ERCC1 mRNA assessment (42 patients [9.4%]). Of the 366 patients who completed the study, 346 were evaluable for response and 366 for progression-free and overall survival and safety. Objective response was observed in 53 patients (39.3%; 95% confidence interval [CI], 31.4%–47.8%) in the control arm and 107 patients (50.7%; 95%CI, 44%–57.5%; p = .019) in the genotypic arm (90). This study shows that assessment of ERCC1 mRNA expression in patient tumor tissue is feasible in the clinical setting and predicts response to docetaxel plus cisplatin. Further studies are warranted to optimize methodologies for ERCC1 analysis in small tumor samples and to refine a multi-biomarker profile predictive of patient outcome. Other

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

243

Fig. 5. Diagram of a potential customized chemotherapy trial based on the expression of NERrelated genes. Patients with low levels of BRCA1 expression (Q1) are assigned to receive gemcitabine/cisplatin; patients with intermediate levels of BRCA1 expression (Q2&3) will receive docetaxel/cisplatin; patients with high levels of BRCA1 expression will receive docetaxel alone.

genes should be examined in conjunction with ERCC1, because cisplatin treatment can lead to increased ERCC1 mRNA expression (91). It has been demonstrated that the human myeloid zinc finger (MZF1) (92) functions as a transcription repressor regulating ERCC1 transcription in response to cisplatin-induced DNA damage. It was found that MZF1 mRNA is constitutively expressed in human ovarian cancer cells. Quantitative PCR in the same cells showed that MZF1 mRNA was decreased upon cisplatin exposure (93). Therefore, decreased expression of MZF1 is a mechanism to be further investigated. The Spanish Lung Cancer Group will initiate a second randomized customized phase III trial in which simultaneous quantitative detection of ERCC1, MZF1, BRCA1, Mad2, and BubR1 will be performed. Patients randomized to the genotypic arm will be divided into quartiles based on levels of ERCC1 and BRCA1. Patients with low levels will receive gemcitabine plus cisplatin; patients with high levels will receive docetaxel alone; and patients with intermediate levels will receive docetaxel plus cisplatin. This study will help to confirm the preclinical data on BRCA1 regulation of mitotic checkpoint genes (Fig. 5).

REFERENCES 1. Schiller JH, Harrington D, Belani CP et al. Eastern Cooperative Oncology Group. Comparison of four chemotherapy regimens for advanced non-small-cell lung cancer. N Engl J Med 2002;346:92–98. 2. Hoang T, Xu R, Schiller JH et al. Clinical model to predict survival in chemonaive patients with advanced non-small-cell lung cancer treated with third-generation chemotherapy regimens based on eastern cooperative oncology group data. J Clin Oncol 2005;23:175–183. 3. Neumann AS, Sturgis EM, Wei Q. Nucleotide excision repair as a marker for susceptibility to tobaccorelated cancers: a review of molecular epidemiological studies. Mol Carcinog 2005;42:65–92. 4. Wei Q, Cheng L, Hong WK et al. Reduced DNA repair capacity in lung cancer patients. Cancer Res 1996;56:4103–4107. 5. Wei Q, Cheng L, Amos CI et al. Repair of tobacco carcinogen-induced DNA adducts and lung cancer risk: a molecular epidemiologic study. J Natl Cancer Inst 2000;92:1764–1772. 6. Spitz MR, Wu X, Wang Y et al. Modulation of nucleotide excision repair capacity by XPD polymorphisms in lung cancer patients. Cancer Res 2001;61:1354–1357. 7. Zhou W, Liu G, Miller DP et al. Gene-environment interaction for the ERCC2 polymorphisms and cumulative cigarette smoking exposure in lung cancer. Cancer Res 2002;62:1377–1381. 8. Wu X, Zhao H, Wei Q et al. XPA polymorphism associated with reduced lung cancer risk and a modulating effect on nucleotide excision repair capacity. Carcinogenesis 2003;24:505–509.

244

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

9. Zhou W, Liu G, Miller DP et al. Polymorphisms in the DNA repair genes XRCC1 and ERCC2, smoking, and lung cancer risk. Cancer Epidemiol Biomarkers Prev 2003;12:359–365. 10. Gurubhagavatula S, Liu G, Park S et al. XPD and XRCC1 genetic polymorphisms are prognostic factors in advanced non-small-cell lung cancer patients treated with platinum chemotherapy. J Clin Oncol 2004;22:2594–2601. 11. de las Penas R, Sanchez-Ronco M, Alberola V et al. Spanish Lung Cancer Group. Polymorphisms in DNA repair genes modulate survival in cisplatin/gemcitabine-treated non-small-cell lung cancer patients. Ann Oncol 2006;17:668–675. 12. de Laat WL, Jaspers NG, Hoeijmakers JH. Molecular mechanism of nucleotide excision repair. Genes Dev 1999;13:768–785. 13. Wijnhoven SW, Hoogervorst EM, de Waard H et al. Tissue-specific mutagenic and carcinogenic responses in NER defective mouse models. Mutat Res 2007;614:77–94. Review. 14. Tripsianes K, Folkers G, Ab E et al. The structure of the human ERCC1/XPF interaction domains reveals a complementary role for the two proteins in nucleotide excision repair. Structure 2005;13: 1849–1858. 15. Ide F, Iida N, Nakatsuru Y et al. Mice deficient in the nucleotide excision repair gene XPA have elevated sensitivity to benzo[a]pyrene induction of lung tumors. Carcinogenesis 2000;21:1263–1265. 16. Yoon T, Chakrabortty A, Franks R et al. Tumor-prone phenotype of the DDB2-deficient mice. Oncogene 2005;24:469–478. 17. Hollander MC, Philburn RT, Patterson AD et al. Deletion of XPC leads to lung tumors in mice and is associated with early events in human lung carcinogenesis. Proc Natl Acad Sci USA 2005;102: 13200–13205. 18. Weeda G, Donker I, de Wit J et al. Disruption of mouse ERCC1 results in a novel repair syndrome with growth failure, nuclear abnormalities and senescence. Curr Biol 1997;7:427–439. 19. Niedernhofer LJ, Odijk H, Budzowska M et al. The structure-specific endonuclease Ercc1-Xpf is required to resolve DNA interstrand cross-link-induced double-strand breaks. Mol Cell Biol 2004;24:5776–5787. 20. De Silva IU, McHugh PJ, Clingen PH et al. Defects in interstrand cross-link uncoupling do not account for the extreme sensitivity of ERCC1 and XPF cells to cisplatin. Nucleic Acids Res 2002;30: 3848–3856. 21. Prasher JM, Lalai AS, Heijmans-Antonissen C et al. Reduced hematopoietic reserves in DNA interstrand crosslink repair-deficient Ercc1–/– mice. Embo J 2005;24:861–871. 22. Cummings M, Higginbottom K, McGurk CJ et al, Masters JR. XPA versus ERCC1 as chemosensitising agents to cisplatin and mitomycin C in prostate cancer cells: role of ERCC1 in homologous recombination repair. Biochem Pharmacol 2006;72:166–175. 23. Bhattacharyya A, Ear US, Koller BH et al. The breast cancer susceptibility gene BRCA1 is required for subnuclear assembly of Rad51 and survival following treatment with the DNA cross-linking agent cisplatin. J Biol Chem 2000;275:23899–23903. 24. Dabholkar M, Vionnet J, Bostick-Bruton F et al. Messenger RNA levels of XPAC and ERCC1 in ovarian cancer tissue correlate with response to platinum-based chemotherapy. J Clin Invest 1994;94: 703–708. 25. Metzger R, Leichman CG, Danenberg KD et al. ERCC1 mRNA levels complement thymidylate synthase mRNA levels in predicting response and survival for gastric cancer patients receiving combination cisplatin and fluorouracil chemotherapy. J Clin Oncol 1998;16:309–316. 26. Selvakumaran M, Pisarcik DA, Bao R et al. Enhanced cisplatin cytotoxicity by disturbing the nucleotide excision repair pathway in ovarian cancer cell lines. Cancer Res 2003;63:1311–1316. 27. Chang IY, Kim MH, Kim HB et al. Small interfering RNA-induced suppression of ERCC1 enhances sensitivity of human cancer cells to cisplatin. Biochem Biophys Res Commun 2005;327:225–233. 28. Zeng-Rong N, Paterson J, Alpert L et al. Elevated DNA repair capacity is associated with intrinsic resistance of lung cancer to chemotherapy. Cancer Res 1995;55:4760–1764. 29. Bosken CH, Wei Q, Amos CI et al. An analysis of DNA repair as a determinant of survival in patients with non-small-cell lung cancer. J Natl Cancer Inst 2002;94:1091–1099.

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

245

30. Shirota Y, Stoehlmacher J, Brabender J et al. ERCC1 and thymidylate synthase mRNA levels predict survival for colorectal cancer patients receiving combination oxaliplatin and fluorouracil chemotherapy. J Clin Oncol 2001;19:4298–4304. 31. Lord RV, Brabender J, Gandara D et al. Low ERCC1 expression correlates with prolonged survival after cisplatin plus gemcitabine chemotherapy in non-small cell lung cancer. Clin Cancer Res 2002;8:2286–2291. 32. Bepler G, Kusmartseva I, Sharma S et al. RRM1-modulated in vitro and in vivo efficacy of gemcitabine and platinum in non-small cell lung cancer. J Clin Oncol 2006;24:4731–4737. 33. Olaussen KA, Dunant A, Fouret P et al. IALT Bio Investigators. DNA repair by ERCC1 in non-small cell lung cancer and cisplatin-based adjuvant chemotherapy. N Engl J Med 2006;355:983–991. 34. Laine JP, Egly JM. Initiation of DNA repair mediated by a stalled RNA polymerase IIO. Embo J 2006;25:387–397. 35. Tremeau-Bravard A, Riedl T, Egly JM et al. Fate of RNA polymerase II stalled at a cisplatin lesion. J Biol Chem 2004;279:7751–7759. 36. Furuta T, Ueda T, Aune G et al. Transcription-coupled nucleotide excision repair as a determinant of cisplatin sensitivity of human cells. Cancer Res 2002;62:4899–4902. 37. Husain A, He G, Venkatraman ES et al. BRCA1 up-regulation is associated with repair-mediated resistance to cis-diamminedichloroplatinum (II). Cancer Res 1998;58:1120-1123. 38. Le Page F, Randrianarison V, Marot D et al. BRCA1 and BRCA2 are necessary for the transcriptioncoupled repair of the oxidative 8-oxoguanine lesion in human cells. Cancer Res 2000;60:5548–5552. 39. Abbott DW, Thompson ME, Robinson-Benion C et al. BRCA1 expression restores radiation resistance in BRCA1-defective cancer cells through enhancement of transcription-coupled DNA repair. J Biol Chem 1999;274:18808–18812. 40. Mullan PB, Gorski JJ, Harkin DP. BRCA1-A good predictive marker of drug sensitivity in breast cancer treatment? Biochim Biophys Acta 2006;1766:205–216. Review. 41. Sankaran S, Starita LM, Simons AM et al. Identification of domains of BRCA1 critical for the ubiquitin-dependent inhibition of centrosome function. Cancer Res 2006;66:4100–4107. 42. Hartman AR, Ford JM. BRCA1 induces DNA damage recognition factors and enhances nucleotide excision repair. Nat Genet 2002;32:180–184. 43. Mullan PB, Quinn JE, Gilmore PM et al. BRCA1 and GADD45 mediated G2/M cell cycle arrest in response to antimicrotubule agents. Oncogene 2001;20:6123-6131. 44. Ramirez JL, Rosell R, Taron M et al. The Spanish Lung Cancer Group.. 14-3-3sigma methylation in pretreatment serum circulating DNA of cisplatin-plus-gemcitabine-treated advanced non-small-cell lung cancer patients predicts survival. J Clin Oncol 2005;23:9105–9112. 45. Rosell R, Scagliotti G, Danenberg KD et al. Transcripts in pretreatment biopsies from a three-arm randomized trial in metastatic non-small-cell lung cancer. Oncogene 2003;22:3548–3553. 46. Scagliotti GV, De Marinis F, Rinaldi M et al. Italian Lung Cancer Project. Phase III randomized trial comparing three platinum-based doublets in advanced non-small-cell lung cancer. J Clin Oncol 2002;20:4285–4291. 47. Rosell R, Danenberg KD, Alberola V et al. Spanish Lung Cancer Group. Ribonucleotide reductase messenger RNA expression and survival in gemcitabine/cisplatin-treated advanced non-small cell lung cancer patients. Clin Cancer Res 2004;10:1318–1325. 48. Alberola V, Camps C, Provencio M et al. Cisplatin plus gemcitabine versus a cisplatin-based triplet versus nonplatinum sequential doublets in advanced non-small-cell lung cancer: a Spanish Lung Cancer Group phase III randomized trial. J Clin Oncol 2003;21:3207–3213. 49. Rosell R, Felip E, Taron M et al. Gene expression as a predictive marker of outcome in stage IIBIIIA-IIIB non-small cell lung cancer after induction gemcitabine-based chemotherapy followed by resectional surgery. Clin Cancer Res 2004;10:4215s–4219s. 50. Taron M, Rosell R, Felip E et al. BRCA1 mRNA expression levels as an indicator of chemoresistance in lung cancer. Hum Mol Genet 2004;13:2443–2449.

246

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

51. Andrew AS, Burgess JL, Meza MM et al. Arsenic exposure is associated with decreased DNA repair in vitro and in individuals exposed to drinking water arsenic. Environ Health Perspect 2006;114: 1193–1198. 52. Saebo M, Skjelbred CF, Nexo BA et al. Increased mRNA expression levels of ERCC1, OGG1, and RAI in colorectal adenomas and carcinomas. BMC Cancer 2006;6:208. 53. Simon GR, Sharma S, Cantor A et al. ERCC1 expression is a predictor of survival in resected patients with non-small cell lung cancer. Chest 2005;127:978–983. 54. Bepler G, Sharma S, Cantor A et al. RRM1 and PTEN as prognostic parameters for overall and disease-free survival in patients with non-small-cell lung cancer. J Clin Oncol 2004;22:1878–1885. 55. Kennedy RD, Quinn JE, Johnston PG et al. BRCA1: mechanisms of inactivation and implications for management of patients. Lancet 2002;360:1007–1014. 56. Lafarge S, Sylvain V, Ferrara M et al. Inhibition of BRCA1 leads to increased chemoresistance to microtubule-interfering agents, an effect that involves the JNK pathway. Oncogene 2001;20: 6597–6606. 57. Quinn JE, Kennedy RD, Mullan PB et al. BRCA1 functions as a differential modulator of chemotherapy-induced apoptosis. Cancer Res 2003;63:6221–6228. 58. Tassone P, Tagliaferri P, Perricelli A et al. BRCA1 expression modulates chemosensitivity of BRCA1defective HCC1937 human breast cancer cells. Br J Cancer 2003;88:1285–1291. 59. Gilmore PM, McCabe N, Quinn JE et al. BRCA1 interacts with and is required for paclitaxel-induced activation of mitogen-activated protein kinase kinase kinase 3. Cancer Res 2004;64:4148–4154. 60. Samanta AK, Huang HJ, Bast RC, Jr. et al. Overexpression of MEKK3 confers resistance to apoptosis through activation of NFkappaB. J Biol Chem 2004;279:7576–7583. 61. Fedier A, Steiner RA, Schwarz VA et al. The effect of loss of Brca1 on the sensitivity to anticancer agents in p53-deficient cells. Int J Oncol 2003;22:1169–1173. 62. Farmer H, McCabe N, Lord CJ et al. Targeting the DNA repair defect in BRCA mutant cells as a therapeutic strategy. Nature 2005;434:917–921. 63. Bryant HE, Schultz N, Thomas HD et al. Specific killing of BRCA2-deficient tumours with inhibitors of poly(ADP-ribose) polymerase. Nature 2005;434:913–917. 64. Chappuis PO, Goffin J, Wong N et al. A significant response to neoadjuvant chemotherapy in BRCA1/2 related breast cancer. J Med Genet 2002;39:608–610. 65. Marsit CJ, Liu M, Nelson HH et al. Inactivation of the Fanconi anemia/BRCA pathway in lung and oral cancers: implications for treatment and survival. Oncogene 2004;23:1000–1004. 66. Baker DJ, Chen J, van Deursen JM. The mitotic checkpoint in cancer and aging: what have mice taught us? Curr Opin Cell Biol 2005;17:583–589. 67. Jordan MA, Wilson L. Microtubules as a target for anticancer drugs. Nat Rev Cancer 2004;4:253–265. 68. Mondal G, Baral RN, Roychoudhury S. A new Mad2-interacting domain of Cdc20 is critical for the function of Mad2-Cdc20 complex in the spindle assembly checkpoint. Biochem J 2006;396:243–253. 69. Chabalier C, Lamare C, Racca C et al. BRCA1 down-regulation leads to premature inactivation of spindle checkpoint and confers paclitaxel resistance. Cell Cycle 2006;5:1001–1007. 70. Wang RH, Yu H, Deng CX. A requirement for breast-cancer-associated gene 1 (BRCA1) in the spindle checkpoint. Proc Natl Acad Sci USA 2004;101(49):17108–17113. 71. Bae I, Rih JK, Kim HJ et al. BRCA1 regulates gene expression for orderly mitotic progression. Cell Cycle 2005;4:1641–1666. 72. Li Y, Benezra R. Identification of a human mitotic checkpoint gene: hsMAD2. Science 1996;274: 246–248. 73. Cahill DP, Lengauer C, Yu J et al. Mutations of mitotic checkpoint genes in human cancers. Nature 1998;392:300–303. 74. Takahashi T, Haruki N, Nomoto S et al. Identification of frequent impairment of the mitotic checkpoint and molecular analysis of the mitotic checkpoint genes, hsMAD2 and p55CDC, in human lung cancers. Oncogene 1999;18:4295–4300. 75. Wang X, Jin DY, Wong YC et al. Correlation of defective mitotic checkpoint with aberrantly reduced expression of MAD2 protein in nasopharyngeal carcinoma cells. Carcinogenesis 2000;21:2293–2297.

Chapter 13 / DNA Repair and Mitotic Checkpoint Genes in NSCLC

247

76. Shichiri M, Yoshinaga K, Hisatomi H et al. Genetic and epigenetic inactivation of mitotic checkpoint genes hBUB1 and hBUBR1 and their relationship to survival. Cancer Res 2002;62:13–17. 77. Hernando E, Nahle Z, Juan G et al. Rb inactivation promotes genomic instability by uncoupling cell cycle progression from mitotic control. Nature 2004;430:797–802. 78. Wang X, Jin DY, Ng RW et al. Significance of MAD2 expression to mitotic checkpoint control in ovarian cancer cells. Cancer Res 2002;62:1662–1668. 79. Weitzel DH, Vandre DD. Differential spindle assembly checkpoint response in human lung adenocarcinoma cells. Cell Tissue Res 2000;300:57–65. 80. Masuda A, Maeno K, Nakagawa T et al. Association between mitotic spindle checkpoint impairment and susceptibility to the induction of apoptosis by anti-microtubule agents in human lung cancers. Am J Pathol 2003;163:1109–1116. 81. Kasai T, Iwanaga Y, Iha H et al. Prevalent loss of mitotic spindle checkpoint in adult T-cell leukemia confers resistance to microtubule inhibitors. J Biol Chem 2002;277:5187–5193. 82. Taylor SS, McKeon F. Kinetochore localization of murine Bub1 is required for normal mitotic timing and checkpoint response to spindle damage. Cell 1997;89:727–735. 83. Anand S, Penrhyn-Lowe S, Venkitaraman AR. AURORA-A amplification overrides the mitotic spindle assembly checkpoint, inducing resistance to Taxol. Cancer Cell 2003;3:51–62. 84. Sudo T, Nitta M, Saya H et al. Dependence of paclitaxel sensitivity on a functional spindle assembly checkpoint. Cancer Res 2004;64:2502–2508. 85. Vogel C, Kienitz A, Muller R et al. The mitotic spindle checkpoint is a critical determinant for topoisomerase-based chemotherapy. J Biol Chem 2005;280:4025–4028. 86. Kienitz A, Vogel C, Morales I et al. Partial downregulation of MAD1 causes spindle checkpoint inactivation and aneuploidy, but does not confer resistance towards taxol. Oncogene 2005;24:4301–4310. 87. Cheung HW, Jin DY, Ling MT et al. Mitotic arrest deficient 2 expression induces chemosensitization to a DNA-damaging agent, cisplatin, in nasopharyngeal carcinoma cells. Cancer Res 2005;65: 1450–1458. 88. Fung MK, Cheung HW, Ling MT et al. Role of MEK/ERK pathway in the MAD2-mediated cisplatin sensitivity in testicular germ cell tumour cells. Br J Cancer 2006;95:475–484. 89. Iwanaga Y, Jeang KT. Expression of mitotic spindle checkpoint protein hsMAD1 correlates with cellular proliferation and is activated by a gain-of-function p53 mutant. Cancer Res 2002;62:2618–2624. 90. Cobo M, Isla D, Massuti B et al. Customizing cisplatin based on quantitative excision repair crosscomplementing 1 mRNA expression: a phase III trial in non-small-cell lung cancer. J Clin Oncol 2007;25:2747–2754. 91. Li Q, Gardner K, Zhang L et al. Cisplatin induction of ERCC-1 mRNA expression in A2780/CP70 human ovarian cancer cells. J Biol Chem 1998;273:23419–23425. 92. Gaboli M, Kotsi PA, Gurrieri C et al. Mzf1 controls cell proliferation and tumorigenesis. Genes Dev 2001;15:1625–1630. 93. Yan QW, Reed E, Zhong XS et al. MZF1 possesses a repressively regulatory function in ERCC1 expression. Biochem Pharmacol 2006;71:761–771.

“This page left intentionally blank.”

14

Dihydropyrimidine Dehydrogenase (Dpyd) Gene Polymorphism: Portrait of a Serial Killer Joseph Ciccolini, PharmD, PhD, Cédric Mercier, MD, PhD, and Gérard Milano, PhD CONTENTS 5-FU P HARMACOKINETICS AND D ISPOSITION : DPD, THE M ETABOLIC G ATE -K EEPER DPD D EFICIENCY AND T REATMENT-R ELATED T OXICITIES U PON 5-FU A DMINISTRATION T UMORAL DPD E XPRESSION AND T REATMENT E FFICACY DPD I MPAIRMENT: I MPLICATION OF DPYD G ENE P OLYMORPHISM AND G ENETIC D OWN -R EGULATION C ATCH M E I F YOU C AN : D ETERMINING DPD S TATUS IN C ANCER PATIENTS P ERSPECTIVES R EFERENCES

S UMMARY Fluoropyrimidines (e.g., 5-FU, oral capecitabine) remain unchallenged as reference drugs for treating numerous solid tumors in adults, including digestive, head and neck, and breast cancers. The wide inter-patient variability observed in the pharmacokinetic profiles of these drugs is mainly caused by the erratic activity of dihydropyrimidine From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 14, 

249

250

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

dehydrogenase (DPD), the enzyme responsible for fluoropyrimidines catabolism. Beside circadian variations, DPD activity can be affected as well by DPYD gene polymorphisms, and genetic and epigenetic regulations, both in the liver and at the tumor site, which can have a strong impact on drug-induced toxicities and/or treatment outcome. In this respect, studying the causes of DPD deficiency allows for better understanding of the pharmacodynamics of 5-FU and capecitabine. Consequently, several strategies have been proposed to predict and anticipate the impact of variations in DPD on the clinical outcome of cancer patients receiving fluoropyrimidine chemotherapy. Key Words: Fluoropyrimidines; pharmacogenetics; pharmacokinetics

1. 5-FU PHARMACOKINETICS AND DISPOSITION: DPD, THE METABOLIC GATE-KEEPER Fluoropyrimidine drugs (e.g., 5-FU, capecitabine) are a mainstay in the treatment of numerous solid tumors, including digestive, head and neck, and breast cancers. Although 5-FU has been in use for nearly 50 years, it remains an unchallenged reference drug in the setting of numerous chemotherapeutic protocols, and nowadays millions of cancer patients worldwide depend on this drug (1,2). Both 5-FU and oral capecitabine display narrow therapeutic indexes combined with high inter-patient pharmacokinetics variability. Consequently, severe toxicities often limit or delay the administration of successive, optimal chemotherapeutic courses, resulting in unfavorable clinical outcomes in cancer patients. Variations in fluoropyrimidine disposition are a major cause for the erratic PK profile observed in most cancer patients. The catabolic pathway of 5-FU has been extensively studied, both on the bench and at the bedside. Liver catabolism accounts for about 95% of the elimination of all fluoropyrimidine drugs. The rate-limiting enzyme responsible for their deactivation is known as dihydropyrimidine dehydrogenase (DPD, Fig. 1A and 1B). 5-FU is immediately dehydrogenated in the liver by DPD to yield dihydro-fluorouracil. This catabolite is subsequently converted to ␤-fluoro-␤ureido-propionic acid and subsequently to fluoro-␤ alanine (F␤AL) by dihydropyrimidinase and ␤-ureidopropionase, respectively. F␤AL derivatives are finally eliminated in urine as conjugated forms (3,4,5,6). Consequently, 5-FU pharmacokinetics is characterized by an extremely short half-life (A splice-site mutation has been reported as the most common one (>50% of all mutations) in Western populations ( 23, 33, 68). Several other SNPs and deletions have been documented with mixed, when not contradictory, effects reported

Table 1 Main Variant Alleles of the DPYD Gene. Reported Frequencies Are Collected from Caucasian Population Studies

254

Polymorphism

Location

Effects of Genotype

DPD Expression

DPD Activity

Allele Frequency

Ref

IVS14+1 G>A 85 T>C 496 A>G 1627 A>G 2194 GT delTCAT295-298

GT splice site EXON2/21 EXON6 EXON13 EXON18 EXON 22 EXON4

Exon 14 Skipping cys29arg met166 val ile543 val val732ile asp949 val DEL

Abolished/Decreased Decreased Decreased Decreased Decreased Decreased Decreased

Reduced/Unchanged Reduced/Unchanged Reduced/Unchanged Reduced Reduced/Unchanged Reduced Reduced

1 T and delTCAT295-298 are the most reported mutations affecting the DPYD gene, as summarized in Table 1. Other variations/deletions such as 611 C>T, 62 G>A, 257 C>T, 601 A>C, 632 A>G, 703 C>T, 1896 C>T, 1003 G>T, 2933 A>G, 1475 C>T, 1590T>C, 1601 G>A, 1679 T>G, delTG1039–1042, delC1897, and delT812 have been reported (51,69,70,71,72,73,74,75,76). It has been suggested recently that besides DPYD sequence variations, promoter hypermethylation and other tissue-specific epigenetic regulations could have an impact on DPD activity as well (77,78,79). Consequently, predicting the actual impact of DPYD gene polymorphisms on DPD impairment and the subsequent clinical outcome with fluoropyrimidine drugs remains a difficult task. For instance, the IVS14+1 G>A mutation has been identified as being the most critical variation associated with life-threatening toxicities upon 5-FU administration (24). This polymorphism affects the splice recognition sequence of intron 14, thus leading to exon skipping and subsequent 165-bp deletion in the mRNA. However, further clinical studies failed to find a clear implication of this SNP in patients displaying severe/lethal toxicities after 5-FU or capecitabine intake, thus raising serious concern about the relevance of basing determination of DPD status in cancer patients on this sole mutation. For instance, retrospective screening for this polymorphism showed that only 2 out of 144 patients with severe toxicities upon 5-FU administration were bearing the IVS14+1 G>A mutation (35). In another retrospective study, it was shown that none of the 6 patients with lethal toxicities after 5-FU or capecitabine administration displayed this polymorphism either, despite marked impairment in DPD activity (20,28,39). It is worth noting that significant differences in DPD activities have been demonstrated between liver and surrogate tissues (80), and single point mutations do not always impact systematically on DPD activity or expression. Such conflicting observations may be explained by the fact that in addition to genetic mutations, the importance of genetic and epigenetic regulations of the DPYD gene may be critical. Strong correlations have been found between DPD activity and mRNA levels, thus suggesting that transcriptional regulation should be an important mechanism leading to marked variations in DPD activity (81). SP1 and, to a lesser extent, SP3 proteins have been identified as transcription activators of the DPYD gene, thus suggesting that assessment of the SP1/SP3 status might be used as a marker for DPD expression (82). Finally, it has been shown that methylation of the DPYD promoter region can be associated with impaired DPD activity (77), although conflicting data related to a possible tissue-specificity have been reported ( 79). Thus the exact mechanisms associating methylation with DPD down-regulation remain to be elucidated.

5. CATCH ME IF YOU CAN: DETERMINING DPD STATUS IN CANCER PATIENTS Due to the critical role of the DPD status in 5-FU disposition and clinical outcome in cancer patients eligible for 5-FU-based chemotherapy, numerous methods have been proposed to identify DPD-deficient patients. A first strategy consists of describing

256

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

sequence variations associated with DPYD gene polymorphisms or down-regulation, whereas some other methods aim at establishing DPD activity as an ultimate, pharmacological endpoint. It is worth mentioning that none of these approaches have stood out as a standard, due to a wide range of biases, technical limitations, or inadequacies with routine clinical practice requiring simple, rapid, and inexpensive detection methods.

5.1. Genotypic Approaches Extensive studies have been undertaken for years to screen the mutations affecting the DPYD gene, and many methods are available now to detect the most common variations in sequence. Most are based upon Real-Time PCR methods (59,83,84,85,86,87) or more complex denaturating HPLC techniques (88,89,90). The most frequently used methods for detecting SNPs are based upon automated Real-Time-PCR approaches, or alternatively PCR-RFLP (restriction fragment length polymorphism) assays using appropriate restriction enzymes. RT-PCR is based on different melting temperatures of fluorescent-labelled oligonucleotide hybridization probes using a single-step assay combining both fluorescence PCR and melting curve analysis. However, only a priori known mutations can be found using such approaches. Although more complicated and with a longer run time, denaturing-HPLC is capable of detecting all known mutations, but can also identify unknown variations in sequence and promoter hypermethylations ( 77, 90). The rise of high-resolution melting curves generated with new DNA-binding dyes should allow in the near future a total specificity in the screening of mutations with a simple and time-effective RT-PCR-based method.

5.2. Phenotypic Approaches DPD is a ubiquitous enzyme. Several reports have shown that evaluating DPD activity in easily available surrogate tissues such as lymphocytes or fibroblasts can be used as a marker for the actual liver function, despite mixed correlations between activities (91,92). Admittedly, PBMC DPD levels below 150 pmol/min/mg protein are associated with a partial deficiency syndrome, and patients displaying DPD A mutation in patients developing FU-related toxicities: an updated analysis based on ten-year recruitment across multiple French institutions. ASCO Annual Meeting Proceedings. J Clin Oncol 2005;23. 36. Ciccolini J, Evrard A, Cuq P. Thymidine phosphorylase and fluoropyrimidines efficacy: a Jekyll and Hyde story. Curr Med Chem Anticancer Agents 2004;4:71–81. 37. Tsukamoto Y, Kato Y, Ura M et al. A physiologically based pharmacokinetic analysis of capecitabine, a triple prodrug of 5-FU, in humans: the mechanism for tumor-selective accumulation of 5-FU. Pharm Res 2001;18:1190–1202. 38. Hooiveld EA, van Kuilenburg AB, Haanen JB et al. Severe toxicity after treatment with capecitabine and fluorouracil due to partial dihydropyrimidine dehydrogenase deficiency. Ned Tijdschr Geneeskd 2004;148:626–628. 39. Mercier C, Yang C, Ciccolini J et al. Determination of uracil/UH2 ratio as potential surrogate for DPD status in cancer patients presenting with severe toxicities during fluoropyrimidine treatment. Asco Annual Meeting Proceedings. J Clin Oncol 2006;24. 40. Shintani Y, Ohta M, Hirabayashi H et al. Thymidylate synthase and dihydropyrimidine dehydrogenase mRNA levels in tumor tissues and the efficacy of 5-fluorouracil in patients with non-small-cell lung cancer. Lung Cancer 2004;45:189–196. 41. Matsuyama R, Togo S, Shimizu D et al. Predicting 5-fluorouracil chemosensitivity of liver metastases from colorectal cancer using primary tumor specimens: three-gene expression model predicts clinical response. Int J Cancer 2006;119:406–413. 42. Okumura K, Shiomi H, Mekata E et al. Correlation between chemosensitivity and mRNA expression level of 5-fluorouracil-related metabolic enzymes during liver metastasis of colorectal cancer. Oncol Rep 2006;15:875–882. 43. Takechi T, Okabe H, Fujioka A et al. Relationship between protein levels and gene expression of dihydropyrimidine dehydrogenase in human tumor cells during growth in culture and in nude mice. Jpn J Cancer Res 1998;89:1144–1153. 44. Ciaparrone M, Quirino M, Schinzari G et al. Predictive role of thymidylate synthase, dihydropyrimidine dehydrogenase and thymidine phosphorylase expression in colorectal cancer patients receiving adjuvant 5-fluorouracil. Oncology 2006;70:366–377. 45. Jensen SA, Vainer B, Sorensen JB. The prognostic significance of thymidylate synthase and dihydropyrimidine dehydrogenase in colorectal cancer of 303 patients adjuvantly treated with 5fluorouracil. Int J Cancer 2007;120:694–701. 46. Ishikawa T, Sekiguchi F, Fukase Y et al. Positive correlation between the efficacy of capecitabine and doxifluridine and the ratio of thymidine phosphorylase to dihydropyrimidine dehydrogenase activities in tumors in human cancer xenografts. Cancer Res 1998;58:685–690. 47. Okuda H, Nishiyama T, Ogura K et al. Lethal drug interactions of sorivudine, a new antiviral drug, with oral 5-fluorouracil prodrugs. Drug Metab Dispos 1997;25:270–273.

Chapter 14 / DPD Deficiency and Fluoropyrimidine Drugs

261

48. Nozoe Y, Ogata Y, Araki Y et al. Up–regulation in dihydropyrimidine dehydrogenase activity by raltitrexed causes antagonism in combination with 5-fluorouracil. Anticancer Res 2003;23: 4663–4669. 49. Milano G, Chamorey AL. Clinical pharmacokinetics of 5-fluorouracil with consideration of chronopharmacokinetics. Chronobiol Int 2002;19:177–189. 50. Grem JL, Yee LK, Venzon DJ et al. Inter- and intraindividual variation in dihydropyrimidine dehydrogenase activity in peripheral blood mononuclear cells. Cancer Chemother Pharmacol 1997;40: 117–125. 51. Yamaguchi K, Arai Y, Kanda Y et al. Germline mutation of dihydropyrimidine dehydrogenese gene among a Japanese population in relation to toxicity to 5-Fluorouracil. Jpn J Cancer Res 2001;92: 337–342. 52. Sohn DR, Cho MS, Chung PJ. Dihydropyrimidine dehydrogenase activity in a Korean population. Ther Drug Monit 1999;21:152–154. 53. Ridge SA, Sludden J, Brown O et al. Dihydropyrimidine dehydrogenase pharmacogenetics in Caucasian subjects. Br.J Clin Pharmacol 1998;46:151–156. 54. Morsman JM, Sludden J, Ameyaw MM et al. Evaluation of dihydropyrimidine dehydrogenase activity in Southwest Asian, Kenyan, and Ghanaian populations. Br J Clin Pharmacol 2000;50:269–272. 55. Mattison LK, Fourie J, Desmond RA et al. Increased prevalence of dihydropyrimidine dehydrogenase deficiency in African-Americans compared with Caucasians. Clin Cancer Res 2006;12:5491–5495. 56. Tuchman M, Roemeling RV, Hrushesky WA et al. Dihydropyrimidine dehydrogenase activity in human blood mononuclear cells. Enzyme 1989;42:15–24. 57. Fleming RA, Milano GA, Gaspard MH et al. Dihidropyrimidine dehydrogenase activity in cancer patients. Eur J Cancer 1993;29A:740–744. 58. Lenz HJ. Pharmacogenomics and colorectal cancer. Adv Exp Med Biol 2006;587:211–231. 59. Meinsma R, Fernandez-Salguero P, Van Kuilenburg AB et al. Human polymorphism in drug metabolism: mutation in the dihydropyrimidine dehydrogenase gene results in exon skipping and thymine uracilurea. DNA Cell Biol 1995;14:1–6. 60. Robert J. Pharmacogenetics and pharmacogenomics as new tools to optimise cancer chemotherapy. J Chemother 2004;16 Suppl 4:22–24. 61. Wadman SK, Berger R, Duran M et al. Dihydropyrimidine dehydrogenase deficiency leading to thymine-uraciluria: an inborn error of pyrimidine metabolism. J Inherit Metab Dis 1985;8 Suppl 2:113–114. 62. Au KM, Lai CK, Yuen YP et al. Diagnosis of dihydropyrimidine dehydrogenase deficiency in a neonate with thymine-uraciluria. Hong Kong Med J 2003;9:130–132. 63. Al Al-Sanna’a NA, Van Kuilenburg AB, Atrak TM et al. Dihydropyrimidine dehydrogenase deficiency presenting at birth. J Inherit Metab Dis 2005;28:793–796. 64. Yokota H, Fernandez-Salguero P, Furuya H et al. cDNA cloning and chromosome mapping of human dihydropyrimidine dehydrogenase, an enzyme associated with 5-fluorouracil toxicity and congenital thymine uraciluria. J Biol Chem 1994;269:23192–23196. 65. Johnson MR, Wang K, Tillmanns S et al. Structural organization of the human dihydropyrimidine dehydrogenase gene. Cancer Res 1997;57:1660–1663. 66. van Kuilenburg AB, De Abreu RA, Van Gennip AH. Pharmacogenetic and clinical aspects of dihydropyrimidine dehydrogenase deficiency. Ann Clin Biochem 2003;40:41–45. 67. Hasegawa T, Kim HS, Fukushima M et al. Sequence analysis of the 5’-flanking regions of human dihydropyrimidine dehydrogenase gene: identification of a new polymorphism related with effects of 5-fluorouracil. Nucleosides Nucleotides Nucleic Acids 2005;24:233–242. 68. Van Kuilenburg AB, Meinsma R, Zoetekouw L et al. High prevalence of the IVS14+1 G>A mutation in the dihydropyrimidine dehydrogenase gene of patients with severe 5-fluorouracil-associated toxicity. Pharmacogenetics 2002;12:555–558. 69. Vreken P, Van Kuilenburg AB, Meinsma R et al. Dihydropyrimidine dehydrogenase (DPD) deficiency: identification and expression of missense mutations C29R, R886H and R235 W. Hum Genet 1997;101:333–338.

262

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

70. A Van Kuilenburg AB, Vreken P, Riva D et al. Clinical and biochemical abnormalities in a patient with dihydropyrimidine dehydrogenase deficiency due to homozygosity for the C29R mutation. J Inherited Metab Dis 1999;22:91–192. 71. Vreken P, Van Kuilenburg AB, Meinsma R et al. Identification of a four-base deletion (delTCAT296– 299) in the dihydropyrimidine dehydrogenase gene with variable clinical expression. Hum Genet 1997;100:263–265. 72. Kouwaki M, Hamajima N, Sumi S et al. Identification of novel mutations in the dihydropyrimidine dehydrogenase gene in a Japanese patient with 5-fluorouracil toxicity. Clin Cancer Res 1998;4: 2999–3004. 73. E Collie-Duguid ES, Johnston SJ, Powrie RH et al. Cloning and initial characterization of the human DPYD gene promoter. Biochem Biophys Comp 2000;271:28–35. 74. van Kuilenburg AB, Baars JW, Meinsma R et al. Lethal 5-fluorouracil toxicity associated with a novel mutation in the dihydropyrimidine dehydrogenase gene. Ann Oncol 2003;14:341–342. 75. van Kuilenburg AB, Dobritzsch D, Meinsma R et al. Novel disease-causing mutations in the dihydropyrimidine dehydrogenase gene interpreted by analysis of the three-dimensional protein structure. Biochem J 2002;364:157–163. 76. Seck K, Riemer S, and Kates R. Analysis of the DPYD gene implicated in 5-fluorouracil catabolism in a cohort of Caucasian individuals. Clin Cancer Res 2005;11:5886–5892. 77. Ezzeldin HH, Lee AM, Mattison LK et al. Methylation of the DPYD promoter: an alternative mechanism for dihydropyrimidine dehydrogenase deficiency in cancer patients. Clin Cancer Res 2005;11:8699–8705. 78. Sato K, Kitajima Y, Miyoshi A et al. Deficient expression of the DPD gene is caused by epigenetic modification in biliary tract cancer cells, and induces high sensitivity to 5-FU treatment. Int J Oncol 2006;29:429–435. 79. Yu J, McLeod HL, Ezzeldin HH et al. Methylation of the DPYD promoter and dihydropyrimidine dehydrogenase deficiency. Clin Cancer Res 2006;12:3864. 80. Collie-Duguid ES, Etienne MC, Milano G et al. Known variant DPYD alleles do not explain DPD deficiency in cancer patients. Pharmacogenetics 2000;10:217–223. 81. Takabayashi A, Iwata S, Kawai Y et al. Dihydropyrimidine dehydrogenase activity and mRNA expression in advanced gastric cancer analyzed in relation to effectiveness of preoperative 5-fluorouracilbased chemotherapy. Int J Oncol 2000;17:889–895. 82. Zhang X, Li L, Fourie J, Davie JR et al. The role of Sp1 and Sp3 in the constitutive DPYD gene expression. Biochim Biophys Acta 2006;1759:247–256. 83. Wei X, McLeod HL, McMurrough J et al. Molecular basis of the human dihydropyrimidine dehydrogenase deficiency and 5-fluorouracil toxicity. J Clin Invest 1996;98:610–615. 84. Vreken P, Van Kuilenburg AB, Meinsma R et al. Dihydropyrimidine dehydrogenase (DPD) deficiency: identification and expression of missense mutations C29R, R886H and R235 W. Hum Genet 1997;101:333–338. 85. Nauck M, Gierens H, Marz W et al. Rapid detection of a common dihydropyrimidine dehydrogenase mutation associated with 5-fluorouracil toxicity and congenital thymine uraciluria using fluorogenic hybridization probes. Clin Biochem 2001;34:103–105. 86. Johnston SJ, Ridge SA, Cassidy J et al. Regulation of dihydropyrimidine dehydrogenase in colorectal cancer. Clin Cancer Res 1999;5:2566–2570. 87. Johnson MR, Wang K, Smith JB et al. Quantitation of dihydropyrimidine dehydrogenase expression by real-time reverse transcription polymerase chain reaction. Anal Biochem 2000;278:175–184. 88. Gross E, Seck K, Neubauer S et al. High-throughput genotyping by DHPLC of the dihydropyrimidine dehydrogenase gene implicated in (fluoro)pyrimidine catabolism. Int J Oncol 2003;22:325–332. 89. Fischer J, Schwab M, Eichelbaum M et al. Mutational analysis of the human dihydropyrimidine dehydrogenase gene by denaturing high-performance liquid chromatography. Genet Test 2003;7:97–105. 90. Ezzeldin H, Okamoto Y, Johnson MR et al. A high–throughput denaturing high-performance liquid chromatography method for the identification of variant alleles associated with dihydropyrimidine dehydrogenase deficiency. Anal Biochem 2002;306:63–73.

Chapter 14 / DPD Deficiency and Fluoropyrimidine Drugs

263

91. Chazal M, Etienne MC, Renee N et al. Link between dihydropyrimidine dehydrogenase activity in peripheral blood mononuclear cells and liver. Clin Cancer Res 1996;2:507–510. 92. Milano G, Etienne MC. Dihydropyrimidine dehydrogenase (DPD) and clinical pharmacology of 5fluorouracil (review). Anticancer Res 1994;14:2295–2297. 93. Lu Z, Zhang R, Diasio RB. Dihydropyrimidine dehydrogenase activity in human peripheral blood mononuclear cells and liver: population characteristics, newly identified deficient patients, and clinical implication in 5-fluorouracil chemotherapy. Cancer Res 1993;53:5433–5438. 94. Saif MW, Mattison L, Carollo T et al. Dihydropyrimidine dehydrogenase deficiency in an Indian population. Cancer Chemother Pharmacol 2006;58:396–401. 95. Milano G, Etienne MC. Potential importance of dihydropyrimidine dehydrogenase (DPD) in cancer chemotherapy. Pharmacogenetics 1994;4:301–306. 96. Milano G, Etienne MC. Individualizing therapy with 5-fluorouracil related to dihydropyrimidine dehydrogenase: theory and limits. Ther Drug Monit 1996;18:335–340. 97. Tuchman M, Roemeling RV, Hrushesky WA et al. Dihydropyrimidine dehydrogenase activity in human blood mononuclear cells. Enzyme 1989;42:15–24. 98. Van Kuilenburg AB, Van Lenthe H, Tromp A et al. Pitfalls in the diagnosis of patients with a partial dihydropyrimidine dehydrogenase deficiency. Clin Chem 2000;46:9–17. 99. Katsumata K, Tomioka H, Sumi T et al. Correlation between clinicopathologic factors and kinetics of metabolic enzymes for 5-fluorouracil given to patients with colon carcinoma by two different dosage regimens. Cancer Chemother Pharmacol 2003;51:155–160. 100. Johnson MR, Yan J, Shao L et al. Semi-automated radioassay for determination of dihydropyrimidine dehydrogenase (DPD) activity : screening cancer patients for DPD deficiency, a condition associated with 5-fluorouracil toxicity. J Chromatogr B Biomed Sci Appl 1997;696:183–191. 101. Etienne MC, Milano G, Fleming RA et al. Dihydropyrimidine dehydrogenase activity in lymphocytes: predictive factor for 5-fluorouracil clearance. Bull Cancer 1992;79:1159–1163. 102. Deporte-Fety R, Picot M, Amiand M et al. High-performance liquid chromatographic assay with ultraviolet detection for quantification of dihydrofluorouracil in human lymphocytes: application to measurement of dihydropyrimidine dehydrogenase activity. J Chromatogr B Biomed Sci Appl 2001;762:203–209. 103. Di Paolo A, Danesi R, Ciofi L et al. Improved analysis of 5-fluorouracil and 5,6-dihydro-5-fluorouracil by HPLC with diode array detection for determination of cellular dihydropyrimidine dehydrogenase activity and pharmacokinetic profiling. Ther Drug Monit 2005;27:362–368. 104. Sumi S, Kidouchi K, Ohba S et al. Automated screening system for purine and pyrimidine metabolism disorders using high performance liquid chromatography. J Chromatogr B Biomed Sci Appl 1995;672:233–239. 105. Kuhara T, Ohdoi C, Ohse M et al. Rapid gas chromatographic–mass spectrometric diagnosis of dihydropyrimidine dehydrogenase deficiency and dihydropyrimidinase deficiency. J Chromatogr B Analyt Technol Biomed Life Sci 2003;792:107–115. 106. Kuhara T, Ohdoi C, Ohse M. Simple gas chromatographic–mass spectrometric procedure for diagnosing pyrimidine degradation defects for prevention of severe anticancer side effects. J Chromatogr B Biomed Sci Appl 2001;758:61–74. 107. Jiang H, Jiang J, Hu P et al. Measurement of endogenous uracil and dihydrouracil in plasma and urine of normal subjects by liquid chromatography–tandem mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci 2002;769:169–176. 108. Remaud G, Boisdron-Celle M, Hameline C et al. An accurate dihydrouracil/uracil determination using improved high-performance liquid chromatography method for preventing fluoropyrimidines-related toxicity in clinical practice. J Chromatogr B Analyt Technol Biomed Life Sci 2005;823:98–107. 109. Garg MB, Sevester JC, Sakoff JA et al. Simple liquid chromatographic method for the determination of uracil and dihydrouracil plasma levels: a potential pretreatment predictor of 5-fluorouracil toxicity. J Chromatogr B Analyt Technol Biomed Life Sci 2002;774:223–230.

264

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

110. Gamelin E, Boisdron-Celle M, Larra F et al. A simple chromatographic method for the analysis of pyrimidines and their dihydrogenated metabolites. J Liq Chromatogr Relat Technol 1997;20: 3155–3172. 111. Deporte R, Amiand M, Moreau A et al. High-performance liquid chromatographic assay with UV detection for measurement of dihydrouracil/uracil ratio in plasma. J Chromatogr B Analyt Technol Biomed Life Sci 2006;834:170–177. 112. Ciccolini J, Mercier C, Blachon MF et al. A simple and rapid high-performance liquid chromatographic (HPLC) method for 5-fluorouracil (5-FU) assay in plasma and possible detection of patients with impaired dihydropyrimidine dehydrogenase (DPD) activity. J Clin Pharm Ther 2004;29: 307–315. 113. Bi D, Anderson LW, Shapiro J et al. Measurement of plasma uracil using gas chromatography–mass spectrometry in normal individuals and in patients receiving inhibitors of dihydropyrimidine dehydrogenase. J Chromatogr B Biomed Sci Appl 2000;738:249–258. 114. Sparidans RW, Bosch TM, Jorger M et al. Liquid chromatography–tandem mass spectrometric assay for the analysis of uracil, 5,6-dihydrouracil and beta–ureidopropionic acid in urine for the mesurement of the activities of the pyrimidine catabolic enzymes. J Chromatogr B Analyt Technol Biomed Life Sci 2006;839:45–53. 115. Jiang H, Lu J, Jiang J, Hu P. Important role of the dihydrouracil/uracil ratio in marked interpatient variations of fluoropyrimidine pharmacokinetics and pharmacodynamics. J Clin Pharmacol 2004;44:1260–1272. 116. Gamelin E, Boisdron-Celle M, Guerin-Meyer V et al. Correlation between uracil and dihydrouracil plasma ratio, fluorouracil (5-FU) pharmacokinetic parameters, and tolerance in patients with advanced colorectal cancer: a potential interest for predicting 5-FU toxicity and determining optimal 5-FU dosage. J Clin Oncol 1999;17:1105. 117. Mattison LK, Ezzeldin H, Carpenter M et al. Rapid identification of dihydropyrimidine dehydrogenase deficiency by using a novel 2-13C-uracil breath test. Clin Cancer Res 2004;10:2652–2658. 118. Mattison LK, Fourie J, Hirao Y et al. The uracil breath test in the assessment of dihydropyrimidine dehydrogenase activity: pharmacokinetic relationship between expired 13CO2 and plasma [2-13C]dihydrouracil. Clin Cancer Res 2006;12:549–555. 119. Maring JG, Oonq BN, de Vries EG. Plasma pharmacokinetics of uracil after an oral uracil challenge dose for dihydropyrimidine dehydrogenase (DPD) phenotyping. Asco Annual Meeting Proceedings J Clin Oncol 2004;22:2120. 120. Maring JG: DPD phenotyping in human volunteers and a DPD-deficient patient by assessing uracil pharmacokinetics after an oral oral uracil test dose: a preliminary report. In: New Insight of Pyrimidine Antagonists Chemotherapy, Chapter 5, pp. 102–111, Thesis, University of Gronigen, NL, 2005. 121. van Kuilenburg AB. Screening for dihydropyrimidine dehydrogenase deficiency: to do or not to do, that’s the question? Cancer Invest 2006;24:215–217. 122. Mercier C, Ciccolini J, Dupuis C et al. Prospective phenotypic screening for DPD deficiency in patients upon fluoropyrimidines administration: Impact on the reduction of drug-induced toxicities. Asco Annual Meeting Proceedings. J Clin Oncol. 2007;25. 123. Salgado J, Zabalegui N, Gil C et al. Polymorphisms in the thymidylate synthase and dihydropyrimidine dehydrogenase genes predict response and toxicity to capecitabine–raltitrexed in colorectal cancer. Oncol Rep 2007;17:325–328. 124. Saif MW, Ezzeldin H, Vance K et al. DPYD*2A mutation: the most common mutation associated with DPD deficiency. Cancer Chemother Pharmacol 2007;60:503–507. 125. Largillier R, Etienne-Grimaldi MC, Formento JL et al. Pharmacogenetics of capecitabine in advanced breast cancer patients. Clin Cancer Res 2006;12:5496–502. 126. Morel A, Boisdron-Celle M, Fey L et al. Identification of a novel mutation in the dihydropyrimidine dehydrogenase gene in a patient with a lethal outcome following 5-fluorouracil administration and the determination of its frequency in a population of 500 patients with colorectal carcinoma. Clin Biochem 2007;40:11–17.

Chapter 14 / DPD Deficiency and Fluoropyrimidine Drugs

265

127. Schneider HB, Becker H. Dehydropyrimidine dehydrogenase deficiency in a cancer patient undergoing 5-fluorouracil chemotherapy. Anticancer Res 2004;24:1091–1092. 128. Gross E, Ullrich T, Seck K et al. Detailed analysis of five mutations in dihydropyrimidine dehydrogenase detected in cancer patients with 5-–fluorouracil-related side effects. Hum Mutat 2003;22:498. 129. Saeki H, Ito S, Futatsugi M et al. Role of dihydropyrimidine dehydrogenase activity in patients with esophageal cancer. Anticancer Res 2002;22:3789–3792. 130. Van Kuilenburg AB, Meinsma R, Zoetekouw L et al. Increased risk of grade IV neutropenia after administration of 5-fluorouracil due to a dihydropyrimidine dehydrogenase deficiency: high prevalence of the IVS14+1 g>a mutation. Int J Cancer 2002;101:253–258. 131. Maring JG, van Kuilenburg AB, Haasjes J et al. Reduced 5-FU clearance in a patient with low DPD activity due to heterozygosity for a mutant allele of the DPYD gene. Br J Cancer 2002;86:1028–1033. 132. Di Paolo A, Danesi R, Falcone A et al. Relationship between 5-fluorouracil disposition, toxicity, and dihydropyrimidine dehydrogenase activity in cancer patients. Ann Oncol 2001;12:1301–1306. 133. van Kuilenburg AB, Muller EW, Haasjes J et al. Lethal outcome of a patient with a complete dihydropyrimidine dehydrogenase (DPD) deficiency after administration of 5-fluorouracil: frequency of the common IVS14+1 G>A mutation causing DPD deficiency. Clin Cancer Res 2001;7:1149–1153.

“This page left intentionally blank.”

15

Impact of UDP-Glucuronosyltransferase 1A Haplotypes on Irinotecan Treatment Kimie Sai, PhD, Hironobu Minami, MD, Yoshiro Saito, PhD, and Jun-ichi Sawada, PhD CONTENTS I NTRODUCTION G ENETIC P OLYMORPHISMS AND E THNIC D IFFERENCES OF UGT1A S A SSOCIATION OF UGT1A H APLOTYPES WITH P HARMACOKINETICS AND A DVERSE R EACTIONS OF I RINOTECAN R ELEVANCE TO A NTI -T UMOR R ESPONSES R EGULATORY S TATUS OF I RINOTECAN T HERAPY C ONCLUDING R EMARKS R EFERENCES

S UMMARY Irinotecan, an antineoplastic-prodrug, is widely used for the treatment of colorectal, lung and other cancers, and is one of model pharmaceuticals for personalized medicine. The active metabolite, SN-38, is a topoisomerase I inhibitor generated by hydrolysis of irinotecan by carboxylesterases. SN-38 is subsequently glucuronidated From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 15, 

267

268

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

by uridine diphosphate glucuronosyltransferase 1As (UGT1As) to form an inactive metabolite, SN-38 G. A reduction in SN-38 G formation is closely related to severe irinotecan toxicities (diarrhea and neutropenia). Therefore, the association of UGT1A1 polymorphisms with irinotecan toxicities has been intensively studied. A number of recent irinotecan-pharmacogenetic studies have revealed significant associations between UGT1A1*28 and severe irinotecan toxicities, leading to the introduction of the clinical application of *28 genetic testing in the United States. This paper provides an overview of recent progress in irinotecan pharmacogenetics, focusing on updated findings on the haplotype structures of UGT1As and addressing the clinical significance of UGT1A1 genotypes/haplotypes to severe irinotecan toxicities. Issues that must be addressed for improvement of personalized irinotecan therapy are also discussed. Key Words: UGT1A1; UGT1A7; UGT1A9; pharmacogenetics; irinotecan; SN38; Haplotype

1. INTRODUCTION There is a wide inter-individual variability of responses to a certain drug, and unexpected severe toxicities are very serious problems for anti-cancer chemotherapy. Genetic mutations, such as single nucleotide polymorphisms (SNPs), in genes encoding drug metabolizing enzymes that regulate the systemic concentration of certain drugs have been recognized as one of the factors responsible for inter-individual differences in drug response. The recent completion of the Human Genome Project and post-genome projects, such as the ENCODE project and International HapMap project, have provided pharmacogenetics/genomics researchers with information that could lead to the realization of “personalized medicine,” a specialty aimed at selecting the appropriate dosage of a suitable drug based on individual genetic factors. Through the efforts of researchers and regulatory agencies, the clinical application of genetic testing has just begun for certain anticancer drugs. Irinotecan, an anticancer prodrug, is one of the targets of personalized medicine in many countries. This drug is widely used to treat a broad range of carcinomas, including colorectal and lung cancers. The active metabolite SN-38 (7-ethyl-10hydroxycamptothecin), a topoisomerase I inhibitor (1), is generated by hydrolysis of the parent compound by carboxylesterases ( 2). SN-38 is subsequently glucuronidated by uridine diphosphate glucuronosyltransferases (UGT1As), such as 1A1, 1A7, 1A9 and 1A10, to form an inactive metabolite, SN-38 glucuronide (SN-38 G) (3,4,5,6) (Fig. 1). SN-38 G is then excreted into the bile and urine via the action of ATP-binding cassette transporters, such as breast cancer resistance protein (BCRP), multiple resistanceassociated protein 2 (MRP2), and p-glycoprotein (P-gp) ( 7). Incorporation of SN-38 from the plasma into the liver is mediated by the organic anion transporter C (OATP-C) (8). Irinotecan is also inactivated by CYP3A4 to produce APC and NPC (9). The dose-limiting toxicities in irinotecan therapy are severe diarrhea and leukopenia ( 10). An increased plasma level of SN-38 may be an important mechanism for severe irinotecan toxicities. Biliary SN-38 G excreted into the small intestine is cleaved by bacterial glucuronidases in the colon to regenerate SN-38; therefore, this process is also

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

269

Fig. 1. The role of UGT1As in the inactivation of an active metabolite of irinotecan (SN-38).

assumed to be one of the mechanisms of late-onset diarrhea ( 11). Among the several biological molecules responsible for irinotecan metabolisms and disposition, the clinical relevance of lowered UGT activity to severe irinotecan toxicities is well known (12). Among the UGT isoforms, UGT1A1 is thought to contribute predominantly to SN-38 G formation ( 3, 13). To date a number of studies have shown a significant association between UGT1A1*28, a repeat polymorphism in the TATA box [-54 -39A(TA)6 TAA>A(TA)7 TAA or -40 -39insTA], and reduced values of areas under the concentration-time curves (AUCs) of SN-38 G and/or severe neutropenia/diarrhea (14,15,16,17,18). Based on these findings, the Food and Drug Administration (FDA) of the United States approved changes on the label of Camptosar (irinotecan HCl) (NDA 20-571/S-024/S-027/S-028), and a genetic diagnostic kit for the *28 allele is available for clinical use in the United States. Nevertheless, several issues must be considered in clinical application of the genetic test in each country: for example, ethnic differences of genetic polymorphisms, possible contribution of the 1A7 and 1A9 genetic polymorphisms, and a balance between anti-tumor and adverse effects. In the meantime, recent pharmacogenomic studies have suggested an advantage to the use of linked combinations of SNPs (haplotypes), rather than individual SNPs, to investigate the associations between genotypes and phenotypes (19). Therefore, information on the haplotype structures covering the UGT1A gene complex in each ethnic population is particularly important for irinotecan pharmacogenetic research, and a number of recent studies with combinatorial haplotypes among 1A1, 1A7, and 1A9 have been instrumental in understanding irinotecan pharmacogenetics. This chapter reviews recent progress in irinotecan pharmacogenetics, including novel functional genetic variations and the haplotype structures of UGT1As, with particular focus on the cumulative data showing the clinical relevance of UGT1A1

270

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

genotypes/haplotypes to the severe toxicities. The current regulatory status and other issues relevant to the realization of personalized irinotecan therapy in each country are also discussed.

2. GENETIC POLYMORPHISMS AND ETHNIC DIFFERENCES OF UGT1AS 2.1. UGT1A Gene Structure UDP-glucuronosyltransferases (UGTs) catalyze the transfer of a glucuronoic acid from uridine diphosphoglucuronic acid to a variety of endogenous and exogenous compounds. The major function of these enzymes is to change hydrophobic compounds into soluble derivatives and thereby facilitate their detoxification and excretion into the bile or urine. Among four subfamilies of UGTs identified in humans (i.e., UGT1, UGT2, UGT3, and UGT8) (20), UGT1A isozymes, such as 1A1, 1A7, 1A9, and 1A10, are known to glucuronidate SN-38 (3,4,5,6,13). The human UGT1A gene complex spans approximately 200 kb on chromosome 2q37, and consists of nine active (1A8, 1A10, 1A9, 1A7, 1A6, 1A5, 1A4, 1A3, and 1A1) and four inactive (1A12P, 1A11P, 1A13P, and 1A2P) exon 1 segments and common exons 2–5 (Block C) (Fig. 2). Each UGT1A gene transcript is formed by splicing one of the first exons with the common exons 2–5 (Block C). The first exon of each UGT1A isoform encodes the N-terminal domain of the enzyme, which determines the substrate-binding specificity. The common exons 2–5 (Block C) encode the C-terminal domain that is essential for the binding of UDP-glucuronoic acid (21). UGT1A1, 1A3, 1A4, 1A6, and 1A9 are expressed in the liver as well as in extrahepatic tissues, including the gastrointestinal tract, while UGT1A7, 1A8, and 1A10 are detected in the extrahepatic tissues (22,23,24). The 5 -flanking region of each first exon is presumed to regulate expression of each UGT1A isoform, whereas the mechanisms for its basal expression and induction by endogenous or exogenous compounds have not yet been fully elucidated. There are large interindividual differences in UGT1A content and activities in various tissues, and the UGT1A genetic polymorphisms in the promoter or coding regions have been implicated as one of the sources of these variations. The following sections describe the genetic

Fig. 2. Structure of the UGT1A gene complex and the major UGT1A1, 1A7, and 1A9 polymorphisms.

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

271

polymorphisms and haplotypes of UGT1A1, 1A7, 1A9 and 1A10 that are responsible for glucuronidation of SN-38 (4,5,6).

2.2. Genetic Variations of UGT1A1, 1A7, 1A9, and 1A10 UGT1A1 is one of the abundantly expressed UGTs in the liver and is responsible for bilirubin glucuronidation in humans. Genetic defects cause hyperbilirubinemia, such as Gilbert’s syndrome and Crigler–Najjar syndrome types I and II. Much information on the genetic variations of UGT1A1 has been accumulated with relevance to hyperbilirubinemia, and more than 60 variations have so far been reported (http://www.pharmacogenomics.pha.ulaval.ca/sgc/ugt alleles/site/pharmacogenomics). In this Chapter, the term “allele” represents a genetic variation at one site, and the term “haplotype” refers to a combination of genetic variations on a chromosome. The most extensively investigated polymorphism is a variation of the number of TA repeats (A(TA)n TAA, n = 5–8) in the promoter region. The wild-type allele contains six [(TA)6] repeats that are located –53 to –42 from the translational start codon. UGT1A1*28 [(TA)7], a common variation in Gilbert’s syndrome ( 25, 26), has an in vitro translational activity that is 63% of the wild-type allele (27). Minor TA variations include *36 (n = 5) and *37 (n = 8), which result in enhanced and reduced, respectively, transcriptional activity in vitro (Table 1). The UGT1A1*60 allele (–3279T>G), located in the distal enhancer region [phenobarbital-responsive enhancer module (PBREM)], is another genetic variation that reportedly reduces transcriptional activity and has been associated with increased plasma bilirubin levels (28). UGT1A1*6 [211 G>A (G71R)] was originally found in Japanese patients with Crigler–Najjar type II and Gilbert’s syndrome ( 29, 30) and results in reduced SN-38 glucuronidation activity ( 5, 31). UGT1A1*27 [686C>A (P229Q)] is a rare nonsynonymous polymorphism with reduced or marginal in vitro glucuronidation of SN-38 that has also been detected in Asians with Gilbert’s syndrome ( 5, 30, 31). 1A1*7 [1456T>G (Y486D)] is another rare nonsynonymous variation, in exon 5, detected in Asians with Crigler–Najjar type II ( 29) and shown to have reduced SN38 glucoronidation activity (5,31). Among the common functional 1A7 polymorphisms, 1A7*2 [387T>G, 391C>A (N129K, R131K)], 1A7*3 [387T>G, 391C>A, 622C>T (N129K, R131K, W208R)], and 1A7*4 [622C>T (W208R)] (32), 1A7*3 and 1A7*4 were shown to cause reduced SN-38 G formation in vitro ( 5). In contrast, a common variation in the 1A9 promoter region, 1A9*1b (originally named *22) [–118 (T)9>10)] enhances transcriptional activity in vitro ( 33). Some rare genetic variations, 1A9*3a [98T>C(M33T)] ( 34), 1A9*5 [766 G>A(D256N)] ( 35), and 1A10*6 [605C>T(T202I)] (originally named *3) ( 36), were also shown to result in reduced SN-38 G formation in vitro (Table 1).

2.3. Haplotype Structures Covering the UGT1A Gene and Ethnic Differences As the importance of haplotype analysis in pharmacogenetic studies has been recognized, the analysis of UGT1A genes, particularly the 1A1 segment, has been conducted for irinotecan-pharmacogenetics in several ethnic populations. The first haplotype analysis of the UGT1A1 enhancer (PBREM)/promoter region was conducted by Innocenti et al. using hepatic samples from 55 Caucasians and 37 African-Americans (37). This

Table 1 UGT1A Genetic Variations Described in this Chapter Allelesa

UGT1A1

Region

Nucleotide change

Amino acid change

Phenotype

Enzymatic activity in vitro

in vivo

*6

Exon 1

211 G>A

G71R

CN2, Gilbert

Reduced

*7

Exon 5

1456T>G

Y486D

CN2

Reduced

*27

Exon 1

686C>A

P229Q

Gilbert

*28

(TA)6 > (TA)7

*60

Promoter A(TA)nTAA Promoter A(TA)nTAA Promoter A(TA)nTAA PBREM

Reduced or No change Reduced

-3279T>G

–

PBREM

-3156 G>A

Reduced

–

3’-UTR

1813C>T/ 1941C>G/2042 C>G (*IB)

Unknown

Reference No.

5, 29-31 5, 29, 31

272

*36 *37

Gilbert

5, 30, 31 25, 27

(TA)6 > (TA)5

Increased

(TA)6 > (TA)8

Reduced

27 27 Gilbert

Reduced 28 16 38, 50

UGT1A7

*2

Exon 1

387T>G/ 391C>A N129K/R131K

*3

Exon 1

*4

Exon 1

387T>G/ 391C>A/ N129K/R131K/ 622T>C W208R 622T>C W208R

*1b(*22)b

Promoter -118(T)n

Reduced or No change Reduced

5, 32 5, 32

Reduced 5, 32

UGT1A9

(T)9 > (T)10

Increased 33

273

No change, Increased or Reduced *3a

Exon 1

98T>C

M33T

Reduced

*5

Exon 1

766 G>A

D256N

Reduced

–

Intron

I399C>T

*6 (*3)b

Exon 1

605C>T

44, 45, 51

34 35 Increased 45 UGT1A10

T202I

Reduced 36

a UGT Alleles Nomemclature (http://www.pharmacogenomics.pha.ulaval.ca/sgc/ugt alleles/site/pharmacogenomics; as of August 15, 2007). b Original names are described in paretheses. PBREM, phenobarbital-responsive enhancer module; Gilbert, Gilbert syndrome; CN2, Crigler-Najjar syndrome type II.

274

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

study revealed a close linkage between the (TA)7 polymorphism (*28), –3279T>G (*60) and –3156 G>A. The frequencies of the 10 haplotypes identified in the study differed between the two ethnic groups (Fig. 3). Haplotype analysis of UGT1A1 in 195 Japanese subjects defined four haplotype groups in the 1A1 segment according to the *1, *60, *6, and *28 marker alleles (38) (Fig. 3). Close linkages among –3279T>G (*60), –3156 G>A and (TA)7 (*28) were also observed in the Japanese subjects (haplotype *28b), but the TA repeat variations were less frequent than in Caucasians and African-Americans ( 38). This study also revealed that the *6 allele [211 G>A (G71R)], specific for East Asians, was independent of the *28 allele, while the *27 allele [686C>A(P229Q)], another rare non-synonymous variation in Asians, was completely linked to the *28 allele; therefore, this haplotype was designated as a *28 group member (*28c). Haplotypes with the *60 allele (–3279T>G) in the absence of the *28 allele were defined as the *60 haplotype group (Fig. 3). Racial differences in the occurrence of UGT1A1 haplotypes were investigated among Caucasians, African-Americans, and Japanese for each 150 subjects (39). The *6 haplotype was detected only in the Japanese, while the frequencies of the *28 group in Caucasians and African-Americans were 3.9and 4.5-fold higher, respectively, than in the Japanese population. The other TA variant haplotypes, *36b [including (TA)5] and *37b [including (TA)8], were observed only in Caucasians and African-Americans (Fig. 3). Regarding the haplotypes covering exons 25 (Block C), the haplotypes harboring three linked SNPs in the 3 -UTR (1813C>T, 1941 C>G and 2042 C>G) were originally identified in the Japanese population and named *IB (38). The frequency of *IB was much higher in African-Americans (18.3%) and Caucasians (15.7%) subjects than in the Japanese (9.7%) (39). Comprehensive haplotype analysis throughout the UGT1A gene complex was conducted for 196 Japanese subjects ( 40), and the combinatorial haplotypes among 1A9, 1A7 and 1A1 were estimated. This study revealed close linkages between 1A9*1b(*22) [-118(T)10] and 1A7*1; between 1A7*2 and the 1A1*60 haplotype (without the *28 allele); and between 1A7*3 and 1A1*6 or 1A1*28 in the Japanese population. The five major combinatorial haplotypes (1A9-1A7-1A1) consisted of *1b(*22)-*1-*1, *1-*3-*6, *1-*2-*60, *1b(*22)-*1-*28 and *1-*3-*28 at frequencies of 58%, 13%, 11%, 6%, and 6%, respectively (Table 2A). An investigation of the haplotype structure among the 1A9, 1A7 and 1A1 polymorphisms in 81 Korean cancer patients found comparable structures between the two Asian populations (40,41). Linkages among the functional polymorphisms across the UGT1A segments were also shown in Caucasians and other ethnic populations. Kohle et al. investigated the linkages among the polymorphisms of 1A7, 1A6, and 1A1 for 100 Caucasians and 50 Egyptians, and reported a close linkage between the UGT1A1*28 [(TA)7] and UGT1A7*3 alleles (42). The frequencies of the combinatorial haplotype 1A7*3-1A1*28 were 28.5% in Caucasians and 21.8% in Egyptians. Perfect linkage of the low-activity 1A7 alleles (1A7*2 and 1A7*3) with the 1A9*1 allele [–118(T)9], and vice versa 1A7*1 with 1A9*1b(*22) [–118(T)10], were also found in a group of 66 cancer patients consisting of 55 Caucasians (43). In this study, the frequencies of the combinatorial haplotypes of 1A9*1 [–118(T)9]1A7*2, 1A9*1 [–118(T)9]-1A7*3, and 1A9*1b(*22) [–118(T)10]-1A7*1 were 25.8%,

275 Fig. 3. UGT1A1 haplotype frequencies in three ethnic populations. (a) UGT allele nomenclature (http://www.pharmacogenomics.pha.ulaval.ca/sgc/ ugt alleles/site/pharmacogenomics; as of August 15, 2007). (b) Definition by Kaniwa et al. (2005) ( 39). (c) The *28b and *28c haplotypes harbor –3156 G>A in addition to –3279 T>G. (d) Data in the literature were used to determine frequencies according to the haplotype definition of Kaniwa et al. (2005) (39). (e) Innocenti et al. (2002) (37). One subject with *36b was excluded from the haplotype analysis. (f) Kaniwa et al. (2005) (39). (g) Sai et al. (2004) (38). (h) 211 G>A (G71R) and 686C>C (P229Q) were not genotyped. N: The number of subjects is described in parentheses.

Table 2 Combinatorial Haplotypes Among UGT1A1, 1A7 and 1A9 (A) Combinatorial haplotypes of 1A9, 1A7 and 1A1: Comparison between Koreans and Japanese Combinatorial haplotype 1A9

-

*1b(*22)f -

Haplotype frequency a

Combinatorial haplotype

Haplotype frequencya

1A7

-

1A1

Korean b (N=81)

Japanese c (N=196)

1A9

-

1A1

Caucasiand (N=133)

Asiand (N=150)e

*1

-

*1

47.5

58.3

*1b(*22)

-

*1

36.8

45.3

51.8 other combinations *1

-

*2

-

*1

4.3

*60

14.8

59.5 1.2

276

other combinations

1.4

*1b(*22)

5.8 5.6 1.1

*1 *1b(*22)

*1

*28 *28 *28

0.6 4.9 1.8

*1

-

*6

20.4

7.3

3.1

12.5

13.1 23.5

*6

-

19.0

*60

9.1

22.3 23.3

*60

1.1

1.0

-

*28 *28

32.5 1.5

4.3 6.7

-

*6

0.0

15.3 2.2

52.6 7.3

10.2

2.4

-

other combinations

*1

*1

12.7

*60

*1 - *3 *1b(*22) - *1 other combinations *3

55.8 *1

11.3 17.2

-

(B) Combinatorial haplotypes of 1A9 and 1A1: Comparison between Caucasians and Asians

34.0

12.7 0.0

*1b(*22)

*6

0.0

N: The number of subjects is described in parentheses. a Data in the literature were used to determine frequencies according to the haplotype definition of Saeki et al.(2006) (40). b Han et al. (2006) (41). c Saeki et al.(2006) (40). d Innocenti et al.(2005) (44). e A mixed population that included 110 Chinese, 20 Tiwanese and 10 Japanese. f 1A9*1bwas originally named 1A9*22.

11.0

13.0 0.3

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

277

31.1%, and 43.2%, respectively. Innocenti et al. investigated the combinatorial haplotypes for 1A9 and 1A1 in 132 Caucasians and 150 Asians, and showed the distinct haplotype structures among two ethnic groups ( 44). The most common three haplotypes were 1A9*1b (*22)[–118(T)10]-1A1*1 (36.8%), 1A9*1[–118(T)9]-1A1*28b (32.5%) and 1A9*1[–118(T)9]-1A1*1 (19.0%) in Caucasians, and 1A9*1b(*22) [–118(T)10]-1A1*1 (45.3%), 1A9*1[-118(T)9]-1A1*60 (22.3%) and 1A9*1[-118(T)9]1A1*6 (12.7%) in Asians. (Table 2B). Girard et al. also characterized 1A9-1A1 haplotype structures in 42 Caucasians and identified a novel 1A9 intronic (3 -flanking) tagging variation, 1A9 I399C>T (IVS1+399C>T), with a frequency of 49% ( 45). Maitland et al. re-sequenced evolutionally conserved UGT1A segments (a total of 47.1 kb) for 24 African-American, 24 European-American and 24 Asians, and analyzed linkage disequilibrium among the hepatically expressed UGT1A segments (1A1, 1A6, and 1A9). This study also clearly demonstrated large ethnic variability in allele frequencies and linkages of genetic variations in these regions, emphasizing the importance of consideration for inter-population difference in phenotype-genotype association studies (46). Haplotype analyses across the UGT1A gene have revealed close linkages among the functional polymorphisms of 1A9, 1A7, and 1A1, and that their combinatorial haplotype structures vary by ethnicity. The data so far reported demonstrate that most 1A7*3containing haplotypes are linked to 1A1*28 in Caucasians and to either 1A1*6 or 1A1*28 in East Asians. The reported findings also indicate that observations based on the 1A9 or 1A7 polymorphisms might reflect phenotypes of the 1A1 genotypes, and that the distinction between the effects of 1A1 and 1A9 or 1A7 polymorphisms may be difficult. Thus, information on haplotype structures covering these 1A segments in each ethnic population is integral for precise evaluation and selection of genetic markers for irinotecan therapy.

3. ASSOCIATION OF UGT1A HAPLOTYPES WITH PHARMACOKINETICS AND ADVERSE REACTIONS OF IRINOTECAN 3.1. Roles of UGT1A1 Genotypes/Haplotypes Until now, much attention has focused on the possible relevance of UGT1A1 polymorphisms to irinotecan toxicities (Table 3). The first clinical evidence showing a role for UGT1A1*28 in irinotecan toxicities was reported by Ando et al. (14). In this study, associations of UGT1A1 genotypes with irinotecan severe toxicities (grade 4 leucopenia and/or grade 3 or 4 diarrhea) were retrospectively evaluated in 118 Japanese cancer patients who received irinotecan therapy with various regimens. Among 26 patients who experienced severe toxicities, 4 homozygous (15%) and 8 heterozygous (31%) *28-bearing patients were observed, while 3 homozygous (3%) and 10 heterozygous (11%) *28 patients were detected in the 92 patients who did not experience severe toxicities. Multivariate analysis also suggested that the genotype of UGT1A1*28 was a risk factor for the irinotecan toxicities (odds ratio, 7.23, confidence interval, 2.52–22.3). Although no statistically significant relevance of 1A1*6 or *27 alone to toxicity was obtained, an additive contribution of these alleles to the toxicities was suggested when

Table 3 Association of UGT1A Polymorphisms with Severe Irinotecan Toxicities: Depencency on Ethnicity and Combination Therapy

278

Reference (No.)

Toxicity

Responsible UGT1A polymorphisms

Ando Y et al..(2000) (14) Iyer L et al. (2002) (15)

Neutropenia and/or ↑ UGT1A1*28 Diarrhea Neutropenia ↑ UGT1A1*28

Innocenti F et al. (2004) (16)

Neutropenia

Ethnicity

Tumor type

Combination therapy

Japanese (N=118)

Various

Monotherapy and Combination (various) Monotherapy

Marcuello E et al. (2004) (17) Diarrhea

↑ UGT1A1*28

Caucasian (N=18) and Various Other ethnics (N=2) Caucasian (N=50) and Various Other ethnics (N=16) Caucasian (N=95) CRC

Rouits E et al. (2004) (18) Carlini LE et al. (2005) (43)

Neutropenia Diarrhea

Han JY et al. (2006) (41) Minami H et al. (2007) (47)

Neutropenia Neutropenia

↑ UGT1A1*28 ↓ UGT1A7(*2,*3) -1A9*1[-118(T)9] ↑ UGT1A1*6 ↑ UGT1A1*6 and *28

Caucasian (N=75) Caucasian (N=55) and Other ethnics (N=11) Korean (N=81) Japanese (N=55)

↑ UGT1A1*28

CRC, colorectal cancer; NSCLC, non-small cell lung cancer; ↑, increases toxicity; ↓, decreases toxicity.

CRC CRC NSCLC Various

Monotherapy Monotherapy and Combination (5-FU or raltitrexed) Combination (5-FU) Combination (capecitabine) Combination (cisplatin) Monotherapy

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

279

they were combined with *28. However, recent UGT1A1 haplotype analysis revealed that the *27 allele is linked to the *28 allele (38,40). Clinical significance of UGT1A1*28 in irinotecan therapy was further demonstrated in a prospective study in 20 patients (including 18 Caucasians) who received irinotecan monotherapy (15). This study showed that patients with the *28 allele had significantly lower rates of SN-38 G formation (AUC ratio of SN-38 G to SN-38) than those with the wild-type (TA)6/6, and that grade 3 or 4 diarrhea and neutropenia were observed only in the patients bearing *28. A significant correlation was found between the absolute neutrophil counts at nadir and the number of the *28 alleles present. Innocenti et al. also showed that the 1A1*28 allele was an important factor in grade 4 neutropenia in 66 cancer patients (including 50 Caucasians) who received irinotecan monotherapy (16). This study revealed that grade 4 neutropenia was much more frequent in the *28 homozygotes [(TA)7/7] (50%) than in heterozygotes [(TA)7/6](12.5%) and wild-type [(TA)6/6] (0%) subjects, and that the relative risk for homozygotes was 9.3-fold higher than that in the other genotypes. In addition, the –3156 G>A mutation, a variation closely linked to the *28 allele, was suggested as a better predictor for UGT1A1 status than the *28 allele in Caucasians (16), but further clinical studies are needed. The clinical importance of *28 was also demonstrated in Caucasian patients who received other regimens. Marcuello et al. investigated the relevance of 1A1*28 to severe toxicities in 95 Caucasian colorectal cancer patients who were treated with irinotecancontaining regimens (5-FU or raltitrexed) ( 17). Grade 3 or 4 diarrhea was more frequently observed in the *28 homozygous (70%) and heterozygous (33%) patients than in those with the wild-type genotype (17%) (p = 0.005). An association with severe neutropenia was also observed in the *28 homozygous patients, but it was not statistically significant. Rouits et al. investigated the relationship of 1A1*28 to severe toxicities in 75 Caucasian patients with colorectal cancer who received irinotecan plus 5-FU and showed high incidences of *28-dependent grade 3/4 neutropenia (18). While the majority of the evidence implicating the clinical importance of *28 in irinotecan treatment has been obtained in Caucasian patients, recent studies for Asian patients show involvement of the low-activity 1A1 allele *6, which is specific to the East Asian population. Sai et al. ( 38) reported that the effect of *6 on the AUC ratio was comparable and additive to that of *28, suggesting its clinical importance ( 38). Han et al. demonstrated the clinical significance of 1A1*6 for irinotecan pharmarcokinetics and pharmacodynamics in 81 Korean patients with non-small cell lung cancer (NSCLC) who received irinotecan plus cisplatin (41). In this population, the allele frequency of 1A1*6 (23.5%) was much more prevalent than 1A1*28 (7.3%). This study also showed a significant association between the homozygous 1A1*6 haplotype and grade 4 neutropenia, and between the homozygous 1A9*1-1A7*3-1A1*6 diplotype and lowered AUC ratios (SN-38 G/SN-38). Due to the close linkage of 1A7*3 to 1A1*6 and/or 1A9*1 [–118(T)9], the benefit of genotyping of 1A1*6 was suggested for Korean patients. The clinical significance of *28 was not shown in this Korean population, which may be partially due to its low frequency in this study (Table 2A). The clinical impact of *6 haplotype as well as *28 was also demonstrated in Japanese patients where the frequencies of *6 and *28 were almost equivalent (13%–16%). Multivariate analysis confirmed a significant contribution of the genetic marker “*6 or *28” to the altered AUC ratio (Fig. 4) and severe neutropenia (47). A different pharmacokinetic study also

280

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

Fig. 4. Association of the UGT1A1 genetic marker “*6 or *28” with a reduced AUC ratio (SN-38 G/ SN-38) in 176 Japanese cancer patients who received irinotecan. A significant gene dose-dependent decrease in the AUC ratio was observed for “*6 or *28” (p < 0.0001, Jonckheere–Terpestra test). The AUC ratios of the “*6 or *28” homozygotes were mostly distributed below the mean value of all patients.

suggested the clinical importance of 1A1*6 in addition to *28 for irinotecan treatment of Japanese patients (48). The *60 allele (–3279 T>G) in PBREM is partially linked with the *28 allele; namely, the *60 haplotype without the *28 allele is also found ( 37, 38, 39) (Fig. 3). Trends of *60 haplotype-dependent decrease in SN-38 G formation (38,47) and increase in total bilirubin levels (38) have been observed in Japanese; however, contribution to irinotecan severe toxicities of the *60 allele alone was not clearly demonstrated (41,47,49). As for the Block C (exons 2–5) haplotype *IB, its significant associations with reduced AUC ratios and neutropenia were not clearly demonstrated, but a possible contribution under co-occurrence with haplotype *60 was suggested ( 38, 47). In fact, the combinatorial haplotype *60 -*IB increased bilirubin levels in healthy Japanese subjects ( 50). Because the frequency of *IB homozygotes is rather low in Asians ( 39), further studies in other ethnic populations would clarify the role of *60-*IB in irinotecan treatment.

3.2. Roles of UGT1A7, 1A9, and 1A10 for Irnotecan Pharmacogenetics UGT1A7, 1A9, and 1A10, which are expressed in the gastrointestinal tract, are assumed to more or less contribute to irinotecan PK/PD. The common polymorphism 1A9*1b(*22) [-118(T)10] was shown to enhance in vitro transcription (33). However, a correlation between 1A9*1b(*22) and hepatic 1A9 protein levels was not observed (51), and association of 1A9*1b(*22) with hepatic SN-38 glucuronidation was inconsistent between two studies ( 44, 45). Recently, the 1A9 3 -flanking SNP I399C>T has been associated with increased enzymatic activities of 1A1 and 1A9 in hepatic samples in Caucasians ( 45), but its clinical significance requires further evaluation. As for rare variations of 1A9 and 1A10, no significant effects of heterozygous 1A9*3 [98 T>C (M33T)] ( 52), 1A9*5 [766 G>A (D256N)] and 1A10*6(*3) [605C>T (T202I)] ( 46) were observed on the AUC ratio (SN-38 G/SN-38) or severe toxicities.

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

281

Regarding 1A7 polymorphisms, no associations of 1A7*2 and *3 with severe toxicities (grade 4 leukopenia or grade 3 diarrhea) were observed in 118 Japanese patients who received irinotecan therapy, suggesting a minor role of 1A7 polymorphisms in SN-38 G formation in vivo (53). The analysis of combinatorial haplotypes covering 1A9, 1A7, and 1A1 in Japanese cancer patients has also suggested that the alterations in the AUC ratio (SN-38 G/SN-38) and the incidence of neutropenia were independent of the 1A9 or 1A7 haplotypes (47). On the other hand, unexpected effects of UGT1A7 and 1A9 polymorphisms on severe intestinal toxicity were reported in 66 colorectal cancer patients (including 55 Caucasians) who received capecitabine/irinotecan therapy ( 43). In this regimen, a major toxicity was grade 3 or 4 diarrhea (29%), while the incidences of grade 3 or 4 neutropenia (1%) and of co-occurrence of both severe toxicities (3%) were very low. Among the genetic polymorphisms of 1A1, 1A6, 1A7, and 1A9, a reduced incidence of severe diarrhea was observed in patients homozygous for low-activity 1A7 alleles (*2 and *3) which were linked to 1A9*1[–118(T)9], but not for the 1A1 or 1A6 polymorphisms. These findings suggest that it is the reduced UGT1A7/1A9 status that confers protection from severe diarrhea. The authors interpreted that the protection occurred through reduced SN-38 G excretion into the gut where SN-38 is regenerated from SN-38 G by bacterial ␤glucuronidase. This finding also raised a caution that higher intestinal levels of SN-38 G can promote diarrhea, while hepatic glucuronidation offers protection from neutropenia. The recent pharmacogenetic studies with UGT1A haplotypes have collectively revealed the primary importance of UGT1A1 genotypes/haplotypes for prediction of severe toxicities and identified ethnic-specific genotypes (*28 for Caucasians, and *6 and/or *28 for Asians). Although there still remains a possible contribution of 1A9*1[–118(T)9]-1A7*3 to irinotecan pharmacogenetics, the close linkages of 1A9*1 [–118(T)9]-1A7* -1A1*28 and 1A9*1[–118(T)9]-1A7*3-1A1*6 make it difficult to distinguish a role of each genetic variation. However, regarding neutropenia, it is likely that severity of neutropenia depends on the activity of UGT1A1, which is a major UGT1A enzyme in the liver and plays a primary role in regulating the plasma level of SN-38. As for intestinal toxicity, it is possible that the genetic polymorphisms of 1A7 or 1A9, which are expressed in the gastrointestinal tract, affect the severity as reported in the capecitabine/irinotecan regimen. There have been reports in which no significant (54) or only transient associations of *28 with severe hematologic toxicities were documented (55). Although these discrepancies may be partially due to the differences in the irinotecan regimens or population size, further confirmation of *28 genotyping benefits is needed.

4. RELEVANCE TO ANTI-TUMOR RESPONSES The clinical relevance of UGT1A polymorphisms to the anti-tumor activity of irinotecan was investigated in several reports. However, the results provide substantial inconsistencies among the studies performed under a variety of regimens. Some studies have shown that UGT1A polymorphisms worsen anti-tumor responses with more severe toxicities. Marcuello et al. reported a trend of poor overall survival for patients with the heterozygous and homozygous *28 alleles [(TA)6/7 and (TA)7/7] in a study of 95 Caucasians with metastatic colorectal cancers who received

282

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

irinotecan-containing therapy ( 17). The reason for the poor responses was interpreted as reduced irinotecan doses in the patients with *28 who showed severe diarrhea after irinotecan administration. The association of 1A1*6 with worse anti-tumor response rates, shorter progression-free period and overall survival ratios as well as a higher incidence of severe neutropenia was observed in 81 Korean NSCLC patients who received irinotecan/cisplatin therapy (41). In contrast, other studies have revealed an association of low-activity UGT1A polymorphisms with better anti-tumor responses and no-enhanced toxicities. Font et al. investigated the relationships between 1A1*28 and anti-tumor parameters (response rate, time to progression, and median survival time) in 47 NSCLC patients treated with an irinotecan/docetaxel regimen (54). In their study, a tendency toward better responses, improved time to progression (4 vs. 3 months) and median survival time (11 vs. 8 months), was observed in the patients with (TA)6/7 and (TA)7/7 compared with (TA)6/6 diplotypes, whereas no significant difference in the incidence of toxicities was observed among the genotypes. Toffoli et al. reported higher response rates and a trend of longer survival in patients with (TA)7/7, compared with (TA)6/6 in 71 Caucasian patients with metastatic colorectal cancer who received irinotecan containing therapy (55). Since the association of *28 with severe toxicity was detected only after the first cycle of irinotecan treatment in this study, the authors suggested that toxicities in (TA)7/7 patients could be well-managed during the entire course of treatment without reduction of irinotecan dosage. In the case of 66 colorectal cancer patients treated with capecitabine/irinotecan, homozygous patients with the low-activity 1A7 alleles (*2 and *3) which are linked to 1A9*1[–118(T)9] exhibited a better response rate with a lower incidence of severe diarrhea (43). These conflicting reports seemingly make it difficult to draw general conclusions on whether low-activity UGT1A polymorphisms enhance or exacerbate anti-tumor responses, or whether better anti-tumor responses parallel severer toxicities. Rather, the balance of toxicity/efficacy appears to vary depending on the regimen. Moreover, the appearance of severe toxicities depends on the exposure levels of SN-38 in the tissues, but the anti-tumor responses can be influenced by additional factors related to properties of target tumors, such as the tumor stage, acquisition of resistant factors, and sensitivity to other chemotherapeutic agents when combined. Therefore, evaluation of the relationship between UGT1A polymorphisms and anti-tumor responses should be carefully performed considering other background factors related to the tumor types, previous chemotherapy, and co-administered drugs, etc. The combination of haplotype analysis with other diagnostic approaches (e.g., transcriptomics or proteomics) to the tumor tissues and/or the plasma samples would be the most effective method for precise prediction of anti-tumor response.

5. REGULATORY STATUS OF IRINOTECAN THERAPY In 2005, based on cumulative evidence supporting the clinical significance of UGT1A1*28, the FDA in the United States approved a label amendment for irinotecan R ). The amendment recommended “a reduction in the starting dose by at (Camptosar least one level of irinotecan for the UGT1A1*28 homozygous patients” [NDA 20-571/S024/S-027/S-028 Camptosar (irinotecan HCl)]. However, because of the still limited

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

283

scales of clinical investigations, the genetic test is not mandatory and the appropriate dose adjustment has not yet been established. Currently, clinical trials using novel SN-38 delivery systems are also underway in several countries. Because approved irinotecan-containing regimens and ethnicity vary among countries, evaluation of the effectiveness of pharmacogenetic approaches should be based on the practical status of each irinotecan-containing regimen. Further prospective studies are needed to establish safer and more effective dosing strategies for irinotecan treatment to patients with UGT1A1*28 and *6.

6. CONCLUDING REMARKS Recent research on the haplotype structures covering the UGT1A gene complex has revealed the linkage patterns of the functional polymorphisms within the complex and their large ethnic variations. To date, cumulative data indicate the primary importance of UGT1A1 genetic polymorphisms, in particular the *28 promoter polymorphism, as a risk factor for severe irinotecan toxicities. Therefore, the clinical application of genetic testing for *28 has started in the United States. Recent haplotype-phenotype studies have also indicated that the genetic relevance to toxic events varies depending on ethnicity and regimen. On this point, it must be noted that 1A1*6 in East-Asians is also an important risk factor for severe neutropenia. In contrast, the effects of UGT1A genotypes on the anti-tumor responses to irinotecan treatment remain unclear. Therefore, further clinical studies are needed to evaluate the benefits of the genotyping of *28 and *6 or other markers in terms of irinotecan efficacy. In addition to the UGT1A genotypes, the genetic variations of other biomolecules responsible for irinotecan metabolism and disposition, such as ABCB1 ( 56) and OATP-C ( 57), should be evaluated for their clinical impact. Information of public databases available in several web sites, such as the ENCODE project data (www.ensembl.org/Homo sapiens/encode.html) and the Pharmacogenetics and Pharmacogenomics Knowledge Base (PharmaGKB) (www.pharmgkb.org), would facilitate further irinotecan-pharmacogenetic studies. Other diagnostic approaches including transcriptomics, proteomics, and metabolomics would provide additional effective biomarkers to more precisely predict toxic events and anti-tumor responses.

REFERENCES 1. Garcia-Carbonero R, Supko JG, Current perspectives on the clinical experience, pharmacology, and continued development of the camptothecins. Clin Cancer Res 2002; 8: 41–661. 2. Slatter JG, Su P, Sams JP et al. Bioactivation of the anticancer agent CPT-11 to SN-38 by human hepatic microsomal carboxylesterases and the in vitro assessment of potential drug interactions. Drug Metab Dispos 1997;25:1157–1164. 3. Iyer L, King CD, Whitington PF et al. Genetic predisposition to the metabolism of irinotecan (CPT11). Role of uridine diphosphate glucuronosyltransferase isoform 1A1 in the glucuronidation of its active metabolite (SN-38) in human liver microsomes. J Clin Invest 1998;15:847–854. 4. Ciotti M, Basu N, Brangi M et al. Glucuronidation of 7-ethyl-10-hydroxycamptothecin (SN-38) by the human UDP-glucuronosyltransferases encoded at the UGT1 locus. Biochem Biophys Res Commun 1999;260:199–202.

284

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

5. Gagne JF, Montminy V, Belanger P et al. Common human UGT1A polymorphisms and the altered metabolism of irinotecan active metabolite 7-ethyl-10-hydroxycamptothecin (SN-38). Mol Pharmacol 2002;62:608-617. 6. Oguri T, Takahashi T, Miyazaki M et al. UGT1A10 is responsible for SN-38 glucuronidation and its expression in human lung cancers. Anticancer Res 2004;24:2893–2896. 7. Sparreboom A, Danesi R, Ando Y et al. harmacogenomics of ABC transporters and its role in cancer chemotherapy. Drug Resist Updat 2003;6:71–84. 8. Nozawa T, Minami H, Sugiura S et al. Role of organic anion transporter OATP1B1 (OATP-C) in hepatic uptake of irinotecan and its active metabolite, 7-ethyl-10-hydroxycamptothecin: in vitro evidence and effect of single nucleotide polymorphisms. Drug Metab Dispos 2005;33:434–439. 9. Santos A, Zanetta S, Cresteil T et al. Metabolism of irinotecan (CPT-11) by CYP3A4 and CYP3A5 in humans. Clin Cancer Res 2000;6:2012–2020. 10. de Forni M, Bugat R, Chabot GG et al. Phase and pharmacokinetic study of the camptothecin derivative irinotecan, administered on a weekly schedule in cancer patients. Cancer Res 1994;54:4347–4354. 11. Kehrer DF, Sparreboom A, Verweij J et al. Modulation of irinotecan-induced diarrhea by cotreatment with neomycin in cancer patients. Clin Cancer Res 2001;7:1136–1141. 12. Gupta E, Lestingi TM, Mick R et al. Metabolic fate of irinotecan in humans: correlation of glucuronidation with diarrhea. Cancer Res 1994;54:3723–3725. 13. Hanioka N, Ozawa S, Jinno H et al. Human liver UDP-glucuronosyltransferase isoforms involved in the glucuronidation of 7-ethyl-10-hydroxycamptothecin. Xenobiotica 2001;31:687–699. 14. Ando Y, Saka H, Ando M et al. Polymorphisms of UDP-glucuronosyltransferase gene and irinotecan toxicity: a pharmacogenetic analysis. Cancer Res 2000;60:6921–6926. 15. Iyer L, Das S, Janisch L et al. UGT1A1*28 polymorphism as a determinant of irinotecan disposition and toxicity. Pharmacogenomics J 2002;2:43–47. 16. Innocenti F, Undevia SD, Iyer L et al. Genetic variants in the UDP-glucuronosyltransferase 1A1 gene predict the risk of severe neutropenia of irinotecan. J Clin Oncol 2004;22:1382–1388. 17. Marcuello E, Altes A, Menoyo A et al. UGT1A1 gene variations and irinotecan treatment in patients with metastatic colorectal cancer. Br J Cancer 2004;91:678–682. 18. Rouits E, Boisdron-Celle M, Dumont A et al. Relevance of different UGT1A1 polymorphisms in irinotecan-induced toxicity: a molecular and clinical study of 75 patients. Clin Cancer Res 2004;10:5151–5159. 19. Judson R, Stephens JC, Windemuth A. The predictive power of haplotypes in clinical response. Pharmacogenomics 2000;1:15–26. 20. Mackenzie PI, Bock KW, Burchell B et al. Nomenclature update for the mammalian UDP glycosyltransferase (UGT) gene superfamily. Pharmacogenet Genomics 2005;15:677–685. 21. Radominska-Pandya A, Czernik PJ, Little JM et al. Structural and functional studies of UDPglucuronosyltransferases. Drug Metab Rev 1999;31:817–899. 22. Tukey RH, Strassburg CP. Human UDP-glucuronosyltransferases: metabolism, expression, and disease. Annu Rev Pharmacol Toxicol 2000;40:581–616. 23. Finel M, Li X, Gardner-Stephen D et al. Human UDP-glucuronosyltransferase 1A5: identification, expression, and activity. J Pharmacol Exp Ther 2005;315:1143–1149. 24. Basu NK, Ciotti M, Hwang MS et al. Differential and special properties of the major human UGT1encoded gastrointestinal UDP-glucuronosyltransferases enhance potential to control chemical uptake. J Biol Chem 2004;279:1429–1441. 25. Bosma PJ, Chowdhury JR, Bakker C et al. The genetic basis of the reduced expression of bilirubin UDP-glucuronosyltransferase 1 in Gilbert’s syndrome. N Engl J Med 1995;333:1171–1175. 26. Monaghan G, Ryan M, Seddon R et al. Genetic variation in bilirubin UPD-glucuronosyltransferase gene promoter and Gilbert’s syndrome. Lancet 1996;347:578–581. 27. Beutler E, Gelbart T, Demina A. Racial variability in the UDP-glucuronosyltransferase 1 (UGT1A1) promoter: a balanced polymorphism for regulation of bilirubin metabolism? Proc Natl Acad Sci USA 1998;95:8170–8174.

Chapter 15 / UGT1A Pharmacogenetics of Irinotecan

285

28. Sugatani J, Yamakawa K, Yoshinari K et al. Identification of a defect in the UGT1A1 gene promoter and its association with hyperbilirubinemia. Biochem Biophys Res Commun 2002;292:492–497. 29. Aono S, Yamada Y, Keino H et al. Identification of defect in the genes for bilirubin UDP-glucuronosyltransferase in a patient with Crigler–Najjar syndrome type II. Biochem Biophys Res Commun 1993;197:1239–1244. 30. Aono S, Adachi Y, Uyama E et al. Analysis of genes for bilirubin UDP-glucuronosyltransferase in Gilbert’s syndrome. Lancet 1995;345:958–959. 31. Jinno H, Tanaka-Kagawa T, Hanioka N et al. Glucuronidation of 7-ethyl-10-hydroxycamptothecin (SN-38), an active metabolite of irinotecan (CPT-11), by human UGT1A1 variants, G71R, P229Q, and Y486D. Drug Metab Dispos 2003;31:108–113. 32. Guillemette C, Ritter JK, Auyeung DJ et al. Structural heterogeneity at the UDP-glucuronosyltransferase 1 locus: functional consequences of three novel missense mutations in the human UGT1A7 gene. Pharmacogenetics 2000;10:629–644. 33. Yamanaka H, Nakajima M, Katoh M et al. A novel polymorphism in the promoter region of human UGT1A9 gene (UGT1A9*22) and its effects on the transcriptional activity. Pharmacogenetics 2004;14:329–332. 34. Villeneuve L, Girard H, Fortier LC et al. Novel functional polymorphisms in the UGT1A7 and UGT1A9 glucuronidating enzymes in Caucasian and African-American subjects and their impact on the metabolism of 7-ethyl-10-hydroxycamptothecin and flavopiridol anticancer drugs. J Pharmacol Exp Ther 2003;307:117–128. 35. Jinno H, Saeki M, Saito Y et al. Functional characterization of human UDP-glucuronosyltransferase 1A9 variant, D256N, found in Japanese cancer patients. J Pharmacol Exp Ther 2003;306:688–693. 36. Jinno H, Saeki M, Tanaka-Kagawa T et al. Functional characterization of wild-type and variant (T202I and M59I) human UDP-glucuronosyltransferase 1A10. Drug Metab Dispos 2003;31:528–532. 37. Innocenti F, Grimsley C, Das S et al. Haplotype structure of the UDP-glucuronosyltransferase 1A1 promoter in different ethnic groups. Pharmacogenetics 2002;12:725–733. 38. Sai K, Saeki M, Saito Y et al. UGT1A1 haplotypes associated with reduced glucuronidation and increased serum bilirubin in irinotecan-administered Japanese patients with cancer. Clin Pharmacol Ther 2004;75:501–515. 39. Kaniwa N, Kurose K, Jinno H et al. Racial variability in haplotype frequencies of UGT1A1 and glucuronidation activity of a novel single nucleotide polymorphism 686C> T (P229L) found in an African-American. Drug Metab Dispos 2005;33:458–465. 40. Saeki M, Saito Y, Jinno H et al. Haplotype structures of the UGT1A gene complex in a Japanese population. Pharmacogenomics J 2006;6:63–75. 41. Han JY, Lim HS, Shin ES et al. Comprehensive analysis of UGT1A polymorphisms predictive for pharmacokinetics and treatment outcome in patients with non-small-cell lung cancer treated with irinotecan and cisplatin. J Clin Oncol 2006;24:2237–2244. 42. Kohle C, Mohrle B, Munzel PA et al. Frequent co-occurrence of the TATA box mutation associated with Gilbert’s syndrome (UGT1A1*28) with other polymorphisms of the UDPglucuronosyltransferase-1 locus (UGT1A6*2 and UGT1A7*3) in Caucasians and Egyptians. Biochem Pharmacol 2003;65:1521–1527. 43. Carlini LE, Meropol NJ, Bever J et al. UGT1A7 and UGT1A9 polymorphisms predict response and toxicity in colorectal cancer patients treated with capecitabine/irinotecan. Clin Cancer Res 2005;11:1226–1236. 44. Innocenti F, Liu W, Chen P et al. Haplotypes of variants in the UDP-glucuronosyltransferase1A9 and 1A1 genes. Pharmacogenet Genomics 2005;15:295–301. 45. Girard H, Villeneuve L, Court MH et al. The novel UGT1A9 intronic I399 polymorphism appears as a predictor of 7-ethyl-10-hydroxycamptothecin glucuronidation levels in the liver. Drug Metab Dispos 2006;34:1220–1228. 46. Maitland ML, Grimsley C, Kuttab-Boulos H et al. Comparative genomics analysis of human sequence variation in the UGT1A gene cluster. Prmacogenomics J. 2006;6:52–62.

286

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

47. Minami H, Sai K, Saeki M et al. Irinotecan pharmacokineitcs/pharmacodynamics and UGT1A genetic polymorphisms in Japanese: Roles of UGT1A1*6 and *28. Pharmacogenet Genomics 2007;17: 497–504. 48. Araki K, Fujita K, Ando Y et al. Pharmacogenetic impact of polymorphisms in the coding region of the UGT1A1 gene on SN-38 glucuronidation in Japanese patients with cancer. Cancer Sci 2006;97: 1255–1259. 49. Kitagawa C, Ando M, Ando Y et al. Genetic polymorphism in the phenobarbital-responsive enhancer module of the UDP-glucuronosyltransferase 1A1 gene and irinotecan toxicity. Pharmacogenet Genomics 2005;15:35–41. 50. Saeki M, Saito Y, Sai K et al. A combinatorial haplotype of the UDP-glucuronosyltransferase 1A1 gene (#60–#IB) increases total bilirubin levels in Japanese. Clin Chem 2007;53:356–358. 51. Girard H, Court MH, Bernard O et al. Identification of common polymorphisms in the promoter of the UGT1A9 gene: evidence that UGT1A9 protein and activity levels are strongly genetically controlled in the liver. Pharmacogenetics 2004;14:501–515. 52. Paoluzzi L, Singh AS, Price DK et al. Influence of genetic variants in UGT1A1 and UGT1A9 on the in vivo glucuronidation of SN-38. J Clin Pharmacol 2004;44:854–860. 53. Ando M, Ando Y, Sekido Y et al. Genetic polymorphisms of the UDP-glucuronosyltransferase 1A7 gene and irinotecan toxicity in Japanese Cancer Patients. Jpn J Cancer Res 2002;93:591–597. 54. Font A, Sanchez JM, Taron M et al. Weekly regimen of irinotecan/docetaxel in previously treated non-small cell lung cancer patients and correlation with uridine diphosphate glucuronosyltransferase 1A1 (UGT1A1) polymorphism. Invest New Drugs 2003;21:435–443. 55. Toffoli G., Cecchin E, Corona G et al. The role of UGT1A1*28 polymorphism in the pharmacodynamics and pharmacokinetics of irinotecan in patients with metastatic colorectal cancer. J Clin.Oncol 2006;24:3061–3068. 56. Sai K, Kaniwa N, Itoda M et al. Haplotype analysis of ABCB1/MDR1 blocks in a Japanese population reveals genotype-dependent renal clearance of irinotecan. Pharmacogenetics 2003;13:741–757. 57. Xiang X, Jada SR, Li HH et al. Pharmacogenetics of SLCO1B1 gene and the impact of *1b and *15 haplotypes on irinotecan disposition in Asian cancer patients. Pharmacogenet Genomics 2006;16:683–691.

16

Microarray Profiling in Breast Cancer Patients Yong Qian, Xianglin Shi, Vincent Castranova, and Nancy L. Guo CONTENTS C URRENT C HALLENGES IN B REAST C ANCER T REATMENT T RADITIONAL P ROGNOSTIC FACTORS FOR B REAST C ANCER DNA M ICROARRAYS FOR T RANSCRIPTION P ROFILING T UMOR T ISSUE M ICROARRAY AND P ROTEOMIC P ROFILING O F P ROTEIN K INASES M OLECULAR P ROFILING FOR C ANCER T HERAPEUTICS N EW G UIDELINES F OR R EPORTING T UMOR M ARKER S TUDIES C ONCLUSIONS R EFERENCES

S UMMARY Breast cancer is the most common cancer among women. It arises from a variety of genetic, epigenetic, and chromosomal alterations. The traditional prognostic and predictive factors in breast cancer mainly focus on the clinical–pathological parameters, which are unable to reveal the diverse molecular alterations of breast cancer and are imprecise in predicting breast cancer progression and clinical outcomes. In recent years, the advances in microarray profiling, including both DNA microarrays and Disclaimer: The findings and conclusions in this report are those of the author(s) and do not necessarily represent the views of the National Institute for Occupational Safety and Health.

From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 16, 

287

288

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

tumor tissue microarrays, provide an unprecedented screen technique to systemically study the pathogenesis of breast cancer. In this chapter, we mainly summarize the progress of our group in microarray profiling for breast cancer. In the first project, we present a population-based study to predict recurrence and metastases of breast cancer using the public gene expression profiles and associated clinical data. In the second project, we develop an integrative model for breast cancer survival and treatment response predictions, which is composed of the expression profiles of several major activated protein kinases as well as traditional clinical–pathological parameters. In the third project, we create a predictive model system to explore proteomic contributions to drug sensitivity, including breast cancer drugs, based on the NCI-60 cell line-related databases. Finally, we discuss the new guidelines for reporting tumor biomarkers in cancer prognostic studies. We believe that an integrated approach combining gene expression profiles, protein expression profiles, as well as clinical information will lead to more informed clinical decision making in breast cancer intervention. Key Words: Breast cancer prognosis; transcriptional profiling; proteomic profiling; tissue array; chemosensitivity prediction

1. CURRENT CHALLENGES IN BREAST CANCER TREATMENT Breast cancer is the most common cancer among women. There were an estimated 240,510 new breast cancer cases and an estimated 40,910 breast cancer deaths in the United Stated in 2007 (1). For the past decade, the overall risk of mortality due to breast cancer has been declining with the development of advanced therapies as well as advance in early detection (2). However, the survival rate has not been substantially improved for patients with recurrent or metastatic breast cancer (3). The main reason is that the current criteria to predict breast cancer progression and clinical outcome are unable to accurately classify breast cancer patients into subgroups of good prognosis and poor prognosis, reflecting a different probability of disease recurrence and survival after therapy ( 4). In this regard, another unsolved challenge is to create standards to guide clinicians in deciding which combination of treatments is most suitable for each individual patient. It has been proven that breast cancer, as with other cancers, arises from a variety of genetic, epigenetic, and chromosomal alterations ( 5). It is the accumulation of those alterations that contributes ultimately to the progressive conversion of normal human breast tissue into breast cancer. The traditional prognostic and predictive factors in breast cancer mainly focus on age, the size of the tumor, axillary-node status, the histologic type and pathological grade of cancer, and hormone-receptor status ( 4). However, they are unable to reveal the diversified molecular alterations of breast cancer and are imperfect in predicting breast cancer progression and clinical outcomes. In recent years, the advances in microarray profiling, including both DNA microarray and tumor tissue microarray, have provided an unprecedented screen technique to systemically study the pathogenesis of breast cancer. These techniques have enabled researchers to simultaneously study the expression patterns of thousands of genes and hundreds of proteins in breast cancer patients. In this chapter, we will mainly discuss recent progress of our group in microarray profiling for breast cancer. There are several

Chapter 16 / Molecular Prognosis of Breast Cancer

289

recent comprehensive reviews addressing the current progress of microarray profiling for breast cancer from different perspectives (6,7,8,9).

2. TRADITIONAL PROGNOSTIC FACTORS FOR BREAST CANCER Substantial efforts have been made to establish the predictive factors for patients with breast cancer during the last two decades. The traditional predictive factors are age, lymph node status, tumor size, histologic type, tumor grade, lymphatic vessel invasion, and hormone receptor status (10). With the development of molecular biology and cell biology, many new predictive factors have been created, including markers regulating cell cycle and cell death, Her2/neu, markers of metastasis or metastatic processes, lymph node micrometastases, bone marrow micrometastases, and markers of angiogenesis (11). Although the traditional predictive factors lack information about the biological diversity of breast cancer and do not reflect the complexity of molecular mechanisms of these diseases, they are still the most valuable criteria for clinicians to decide the relevant therapies (12). For instance, Adjuvant! Online (www.adjuvantonline.com) is a prognostic system based on traditional pathological features, including age, ER expression, and grade. It has been independently validated as a reliable aid to clinical decision making for average breast cancer patients (13). Although lymph node status, tumor size, histologic type, tumor grade, and tumor stage are the most powerful traditional pathologic prognostic and predictive factors for breast cancer patients, methods for measurement of these traditional pathologic factors have been difficult to standardize (10). Differences in these measurements might lead to the misclassification of some breast cancer patients. Even under the same standards, the inter-observer variability still occurs frequently among different examiners. Therefore, these traditional factors are inclined to be subjective. A recent report by a Breast Task Force serving the American Joint Committee on Cancer did not add tumor grade to its staging criteria because of the inter-observer variability problem in the current grading system (14). Breast cancer patients with the same stage of disease can have markedly different clinical outcomes. Traditional diagnostic and prognostic factors may stratify patients with molecularly distinct diseases into the same group based on morphological assessments. It remains a critical issue to reliably identify specific breast cancer patients at high risk for recurrent and metastatic disease. Molecular prediction is the future direction of personalized cancer care.

3. DNA MICROARRAYS FOR TRANSCRIPTION PROFILING Genomic-based microarray technologies have fostered tremendous advances in molecular diagnosis and prognosis of breast cancer. They enable clinicians to analyze thousands of gene expression patterns simultaneously to reveal the distinct nature of each individual’s breast cancer. Comparative genomic hybridization (CGH) is an array technology for detecting aberrations in DNA copy numbers. This technology is valuable in exploring genomic instability in breast cancer development and progression (15,16,17). Affymetrix chips across different generations and cDNA microarrays are common platforms used to identify gene expression-based signatures for breast cancer prognosis

290

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

(12,18,19,20,21,22,23,24,25,26,27). Based on transcriptional profiling, Oncotype DX and MammaPrint, which will be described in details later in this section, have been developed and used in clinics. It has been shown that the gene expression–based microarray profiling offers tremendous potential to define subcategories of breast cancer, to predict disease relapse, to predict chemotherapy response, and to predict progression of ductal carcinoma in situ (7). Microarray analyses of breast cancer have identified unique gene expression profiles associated with patient survival. Sorlie et al. ( 23) found the expression profiles to distinguish ER+ from ER– tumors with distinct outcomes in a cohort with locally advanced breast cancer treated with primary chemotherapy. Van’t Veer et al. (18) established a 70-gene signature to predict metastatic potential in an untreated, node-negative cohort. Sotiriou et al. ( 24) showed the concordance with these previous analyses in node-positive and node-negative patients with the majority receiving adjuvant treatment. The U.S. FDA recently approved the first gene test for cancer, MammaPrint (the 70-gene signature) of Agendia (Amsterdam, the Netherlands) ( 18), for node-negative women under 61 years of age with primary breast cancer tumors. Oncotype DX (a 21-gene signature) of Genomic Health (Redwood City, CA) is another commercially available gene test for breast cancer. It is currently applied in clinics to predict recurrence of tamoxifen-treated, node-negative, and estrogen-receptor-positive breast cancer (28). A recent study (29) evaluated the concordance of multiple gene signatures for breast cancer prognosis, including MammaPrint ( 18), Oncotype DX ( 28), a wound healing signature ( 30), a two-gene signature ( 31), as well as an intrinsic subtype signature ( 23, 32, 33). There is no common overlap in these evaluated gene models. This study concluded that all these gene sets except the 2-gene model showed significant agreement in individualized outcome prediction (29). These available signatures were based on specific subgroups of breast cancer patients. There were two population-based molecular signatures identified by Sotiriou et al. ( 24) and Naderi et al. ( 34), which comprised 93 genes and 70 genes, respectively. In these two studies, the association of the expression signatures with patients’ survival was evaluated using statistical analysis of hazard ratios or Kaplan–Meier plots. Nevertheless, the new NIH guidelines stated that the evaluation of various markers should be based on true positive and false positive, instead of odds ratios or relative risks, to determine the predictive power of biomarkers in individual patients (35). According to these guidelines, our group recently presented an accurate 28-gene signature for predicting an individual breast cancer patient’s risk for recurrence across population-based, heterogeneous patient cohorts (36). In this chapter, we present a population-based study that has predicted recurrence and metastases of breast cancer using gene expression profiles and clinical data from Sotiriou ( 24). Expression profiles of 7,091 genes were investigated on an unselected group of 99 node-negative and node-positive breast cancer patients to identify a prognostic gene signature of recurrence and metastases. The results showed that a unique 28-gene signature was able to achieve highly accurate prediction of disease-free survival and overall survival. Importantly, our identified 28-gene signature didn’t overlap with previously published breast cancer survival signatures ( 18, 23, 24). Most of these identified

Chapter 16 / Molecular Prognosis of Breast Cancer

291

genes are either related to tumorigenesis, or are directly involved in breast cancer pathogenesis. It was also found that our identified 28-gene signature was associated with the clinical variables, such as tumor size, tumor grade, ER status, PR status, and HER2/neu. The breast cancer patients were first divided into different risk groups based on 28-gene expression profile and then integrated with the clinical variables. The results demonstrated that the gene expression–defined risk groups were strongly associated with the clinical pathologic variables, confirming the clinical relevance of our identified 28-gene signature. To further elucidate the molecular pathogenesis of breast cancer, we also identified 14-gene predictors of breast cancer nodal status and 9-gene predictors of breast cancer tumor grade. This study established a population-based approach to predicting breast cancer outcomes at the individual level exclusively based on gene expression patterns and provided a general guideline for developing robotic prognosis toward personalized medicine. The results of this study are another successful example showing the tremendous potential of microarray analyses of breast cancer in identifying gene expression profiles associated with breast cancer patient survival.

4. TUMOR TISSUE MICROARRAY AND PROTEOMIC PROFILING OF PROTEIN KINASES DNA-based microarray profiling provides detailed genomic information, whereas protein-based microarrays reveal insights into cell functional units as well as cell signaling events ( 37). The gene expression pattern doesn’t necessarily correlate with the protein expression pattern, because the route from gene induction to protein expression involves several biological processes including translation and post-translational modification. Therefore, proteomics holds tremendous information related to cancer formation. It is the protein that ultimately plays a functional role in breast cancer development and progression. It has been well-established that the development and progression of breast cancer involve the activation of numerous protein kinases ( 38), which is characterized by an increase in protein phosphorylation. Furthermore, those activated protein kinases interact each other at multiple levels leading to the progression of breast cancer. Through collaboration with Clinomics Biosciences, Inc., we obtained access to their breast cancer patient database containing clinical information and immunohistochemistry tissue array data. We developed an integrative profile for breast cancer survival and treatment response predictions, which composed the expression profile of several major activated protein kinases as well as several traditional clinical parameters (39). We first created a set of predictors for breast cancer survival. The gain ratio attribute evaluation algorithm was applied to rank the importance of clinical parameters in breast cancer survival. The results showed that the optimal subset of clinical parameters is histology, positive lymph node status, pT (AJCC), pN (AJCC), and smoking. We further identify the best subset of seven antibody scores of activated protein kinases from 42 antibody scores in predicting breast cancer survival. These seven antibody scores contained the measured levels of six activated protein kinases: phospho-EGFR, phospho-ER, phospho-Her2/neu, phospho-IGF-IR/In, phospho-MAPK, and phospho-p70S6 K . We

292

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

then employed the Randon Committee algorithm to retrospectively predict breast cancer survival with different settings: (1) expression profiles of all 42 antibody scores; (2) the 42 antibody scores plus the selected clinical factors; and (3) the top seven ranked antibody scores plus the selected clinical factors. Our results demonstrated that the overall accuracy of survival prediction reached to 92.3% with the use of identified expression profiles of phospho-EGFR, phospho-ER, phospho-Her2/neu, phosphoIGF-IR/In, phospho-MAPK, and phospho-p70S6 K in connection with selected clinical information. Using the same algorithms, we further utilized the existing data to identify important clinical factors and activated protein kinase expression profiles for individualized breast cancer treatment response prediction. We identified the nine most informative clinical factors in treatment response prediction, including metastasis site, smoking, ER, histology, PR, surgery procedure, chemotherapy, stage, and pN (the sequence represents the order of the ranking). The most informative activated protein kinase measurements were phospho-ER, phospho-EGFR, phospho-Her2/neu, and phospho-p70S6 K . Further analysis demonstrated that, with the integration of the identified clinical factors and antibody measurements, the overall accuracy for breast cancer treatment response prediction was 92.6%, which was higher than using the clinical prediction factors or antibody measurements alone. Our results demonstrated that the identified subsets of the activated protein kinases significantly increased the accuracy of clinical outcome predictions. Most notably in the study, we evaluated protein phosphorylation levels instead of total protein expression levels. Protein phosphorylation and dephosphorylation are well-characterized biochemical processes for protein kinases to conduct cellular signal transduction. Phosphorylation at certain tyrosine, serine, or threonine residues in kinases is a key step for their activation, and the measurement of these phosphorylations reflects their functional status in vivo. Thus, the protein kinase phosphorylation-based tissue microarray more accurately reveals the molecular mechanisms of breast cancers, and more accurately predicts the individualized survival and treatment response. In addition, protein kinase phosphorylation-based tissue microarray potentially provides clues at the molecular level for the selection of an optimal therapy for breast cancers. It should be noted that one of the major challenges for this technique is the specificity and availability for the antibodies. Each antibody should be evaluated for specificity with Western blot analysis prior to being applied to tissue microarray assays.

5. MOLECULAR PROFILING FOR CANCER THERAPEUTICS Assessment of an individual patient’s predisposition to drugs is essential to achieve the goal of personalized medicine in cancer therapy. Such an approach is needed for clinicians to decide which chemotherapeutic agents would be effective for a given patient and to avoid including ineffective agents (and the entailed side effects) in treatment options. This decade has witnessed significant advances in pharmacogenomics research to predict drug sensitivity by DNA copy number (40), DNA mutation (41), transcriptional profiling (19,42,43,44,45), and proteomic profiling (46). It is especially challenging to

Chapter 16 / Molecular Prognosis of Breast Cancer

293

predict chemosensitivity in the clinical context, because drug responses reflect properties intrinsic to both the target cells and host metabolism (43). The National Cancer Institute’s Developmental Therapeutics Program compiled an extensive database for a panel of 60 human cancer cell lines (the NCI-60). The NCI-60 set includes cell lines derived from leukemias, melanomas, and carcinomas of ovarian, renal, breast, prostate, colon, lung, and central nervous system (CNS) origin. These cell lines have been screened for the drug activity of a broad range of chemical compounds. A sulphorhodamine B assay was applied to examine growth inhibition by measuring the total cellular protein changes upon the stimulation with a particular chemical compound. The drug activities were assessed based on the pattern of growth inhibition within 48 hours. Research focused on a 118-drug subset whose mechanisms of action was putatively known (42). Some of these drugs are currently in routine clinical use for cancer treatment, while others are either in clinical trials or in late stages of drug development. These data are publicly available from the National Cancer Institute’s Discover website (http://discover.nci.nih.gov/datasets.jsp). The NCI-60 panel was originally from clinical cancers. Generally speaking, they represent the biological properties of the corresponding cancer types. Analysis using these cell lines allows for the generation of reproducible and stable experiment results, which may guide clinical research in future. We recently created a predictive model system to explore proteomic contributions to drug sensitivity, including breast cancer drugs, based on the NCI-60 cell line-related data bases (46). Both the proteomic profiles and the drug activity database of the 118 agents were generated by the National Cancer Institute and are available from the National Cancer Institute’s Discover web site. The database of protein expression levels was generated by proteomic assays with a 52-antibody reverse-phase protein lysate microarray in each individual cell line ( 47). The proteomic assays were conducted using reverse-phase protein lysate microarrays ( 47, 48). The protein samples were robotically planted on the chips followed by measurement with 52 antibodies. Each of the 52 antibodies is a specific antibody that recognizes a specific protein. The data and detailed information are available to the public. We sought to identify important protein markers, which predict drug response of each individual cell line to the 118 anticancer agents, including some anti-breast cancer drugs. In this study, a machine learning model system was developed to classify cell line chemosensitivity exclusively based on proteomic profiling. Using reverse-phase protein lysate microarrays, protein expression levels were measured by 52 antibodies in a panel of 60 human cancer cell (NCI-60) lines. The model system combined several well-known algorithms, including Random forests, Relief, and the nearest neighbor methods, to construct the protein expression-based chemosensitivity classifiers. The classifiers were designed to be independent of the tissue origin of the cells. A total of 118 classifiers of the complete range of drug responses (sensitive, intermediate, and resistant) were generated for the evaluated anticancer drugs, one for each agent. The accuracy of chemosensitivity prediction of all the evaluated 118 agents was significantly higher (P < 0.02) than that of random prediction. Our results demonstrated that it was feasible to accurately predict chemosensitivity by reverse-phase protein lysate microarray-based proteomic approaches.

294

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

6. NEW GUIDELINES FOR REPORTING TUMOR MARKER STUDIES The high dimensionality of microarray data puts a premium on innovative analytical methods. Current DNA and protein microarray technology generate thousands of data points for a single sample. However, microarray experiments are expensive and usually exhibit high noise within each experiment and low reproducibility across different datasets (49,50). Therefore, it is essential to identify relevant and important biomarkers by appropriate data mining approaches. This step is of great significance for increasing prediction accuracy and reducing the costs, time, and labor of clinical tests. Traditional methods select a set of biomarkers based on statistical tests, such as the t-test, correlation coefficient, or Cox model, neglecting interactions among the markers (49). It has been found that top-ranked markers showing strong association with clinical outcome are not necessarily good classifiers (35,51,52). To evaluate the predictive power of various markers in individual patients, the new NIH guidelines state that the accuracy of the biomarkers should be based on true positive and false positive, instead of odds ratios or relative risks ( 35). Because the disease progression and clinical outcome vary in the time course, time-dependent ROC analysis ( 53, 54) should be used to assess the predictive power of prognostic biomarkers in accordance to the new NIH guidelines. In an alternative REMARK (REporting recommendations for tumor MARKer) system, frameworks and guidelines were proposed for data design and data analysis in tumor marker prognostic studies (55,56,57,58,59,60,61,62). According to REMARK, reports of a tumor marker study should include an introduction of the marker examined, the study objective, and any prespecified hypothesis. The report of materials and methods should include patients, specimen characteristics, assay methods, study design, and statistical analysis methods. Specifically, the report should specify the characteristics of patients and treatment received, as well as the characteristics of specimens (including control samples) and methods of preservation and storage. The description of assay methods should specify the assay method used and provide (or reference) a detailed protocol. It should also specify whether and how assays were performed blinded to the study endpoint. The study design should include the method of case selection and describe whether prospective or retrospective and whether stratification or matching (e.g., by disease stage or age) design were used. The study design should specify the follow-up period and the median follow-up time, define all clinical endpoints examined, list all candidate variables for inclusion in models, and give a rationale for sample size or target power with effect size. The report of statistical analysis methods should specify all computational methods, including details of any variable selection methods and other model construction issues, how model assumptions were verified, and how missing data were handled. It should also clarify how marker measurements were handled and/or methods for cutoff determination. The report of results should include data, analysis, and presentation. Specifically, the data report should include the description of the flow of the patients through the study, as well as the number of patients included in each stage of the analysis and/or subgroup examined. The data report should provide distributions (including missing values) of basic demographic characteristics (e.g., sex and age), standard (disease-specific) prognostic

Chapter 16 / Molecular Prognosis of Breast Cancer

295

variables, and tumor markers. The guidelines for analysis and presentation include the relation of the marker to standard prognostic variables, the relation of the marker and outcome, and sensitivity analysis. In the REMARK system, estimated effects (e.g., hazard ratios) with confidence intervals for the marker were recommended for univariant and key multivariable analyses, and a Kaplan–Meier plot was recommended to represent the effect of a tumor marker in a time-to-event outcome. The discussion section should interpret the results in the context of the pre-specified hypothesis and describe limitations of the study as well as implications for future research and clinical value. These guidelines were advocated for reporting of tumor marker studies in breast cancer research and treatment (63).

7. CONCLUSIONS Breast cancer is a complex and heterogeneous disease, encompassing a wide range of pathologic entities and molecular profiles. It is crucial for the physicians to accurately define a patient’s risk of developing metastatic and recurrent diseases at diagnosis. This will determine the clinical course for the given patient, that is, which patient should receive expensive and toxic adjuvant therapy and which patient should avoid over-treatment. Furthermore, it is critical for physicians to determine which combination of treatment is most suitable for each individual patient. However, it still remains challenging to make an accurate predictive assessment of a patient’s risk or response to certain treatment regimens. Advances in microarray technology promise breakthroughs in personalized medicine for breast cancer treatment. To date, the prediction models based on microarray technology for breast cancer have focused mainly on either transcriptional profiles or proteomic profiles, instead of the integrated transcriptional and proteomic profiles. The molecular mechanisms of breast cancer pathogenesis involve multiple levels, including transcription, translation, and post-translational modification. An approach that integrates gene expression profile and protein expression profiles as well as clinical information will more accurately reveal the intrinsic nature of breast cancer progression and development. Such an approach holds great promise to guide clinicians in treating breast cancer in the new era of personalized medicine.

REFERENCES 1. Cancer Facts & Figures 2007. Atlanta, GA: American Cancer Society, 2007. 2. Peto R, Boreham J, Clarke M et al. UK and USA breast cancer deaths down 25% in year 2000 at ages 20–69 years. Lancet 2000;355:1822. 3. Giordano SH, Buzdar AU, Smith TL et al. Is breast cancer survival improving? Cancer 2004;100: 44–52. 4. Kallioniemi A. Molecular signatures of breast cancer: predicting the future. N Engl J Med 2002;347:2067–2068. 5. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;100:57–70. 6. Abramovitz M, Leyland-Jones B. A systems approach to clinical oncology: focus on breast cancer. Proteome Sci 2006;4:5. 7. Cleator S, Ashworth A. Molecular profiling of breast cancer: clinical implications. Br J Cancer 2004;90:1120–1124. 8. Gruvberger-Saal SK, Cunliffe HE, Carr KM et al. Microarrays in breast cancer research and clinical practice: the future lies ahead. Endocr Relat Cancer 2006;13:1017–1031.

296

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

9. Paik S. Molecular profiling of breast cancer. Curr Opin Obstet Gynecol 2006;18:59–63. 10. Schnitt SJ. Traditional and newer pathologic factors. J Natl Cancer Inst Monogr 2001;22–26. 11. Hayes DF, Isaacs C, Stearns V. Prognostic factors in breast cancer: current and new predictors of metastasis. J Mammary Gland Biol Neoplasia 2001;6:375–392. 12. Ludwig JA, Weinstein JN. Biomarkers in cancer staging, prognosis, and treatment selection. Nat Rev Cancer 2005;5:845–856. 13. Olivotto IA, Bajdik CD, Ravdin PM et al. Population-based validation of the prognostic model ADJUVANT! for early breast cancer. J Clin Oncol 2005;23:2716–2725. 14. Singletary SE, Allred C, Ashley P et al. Revision of the American Joint Committee on Cancer staging system for breast cancer. J Clin Oncol 2002;20:3628–3636. 15. Chin K, de Solorzano CO, Knowles D et al. In situ analyses of genome instability in breast cancer. Nat Genet 2004;36:984–988. 16. Pinkel D, Albertson DG. Array comparative genomic hybridization and its applications in cancer. Nat Genet 2005;37 Suppl:S11–S17. 17. Rodriguez V, Chen Y, Elkahloun A et al. Chromosome 8 BAC array comparative genomic hybridization and expression analysis identify amplification and overexpression of TRMT12 in breast cancer. Genes Chromosomes Cancer 2007;46:694–670. 18. van’t Veer LJ, Dai H, van de Vijver MJ et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415:530–536. 19. Bild AH, Yao G, Chang JT et al. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature 2006;439:353–357. 20. Glinsky GV, Higashiyama T, Glinskii AB. Classification of human breast cancer using gene expression profiling as a component of the survival predictor algorithm. Clin Cancer Res 2004;10:2272–2283. 21. Huang E, Cheng SH, Dressman H et al. Gene expression predictors of breast cancer outcomes. Lancet 361;2003:1590–1596. 22. Murphy N, Millar E, Lee CS. Gene expression profiling in breast cancer: towards individualising patient management. Pathology 2005;37:271–277. 23. Sorlie T, Perou CM, Tibshirani R et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA 2001;98:10869–10874. 24. Sotiriou C, Neo SY, McShane LM et al. Breast cancer classification and prognosis based on gene expression profiles from a population-based study. Proc Natl Acad Sci USA 2003;100:10393–10398. 25. van de Vijver MJ, He YD, van’t Veer LJ et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002;347:1999–2009. 26. West M, Blanchette C, Dressman H et al. Predicting the clinical status of human breast cancer by using gene expression profiles. Proc Natl Acad Sci USA 2001;98:11462–11467. 27. Zhao H, Langerod A, Ji Y et al. Different gene expression patterns in invasive lobular and ductal carcinomas of the breast. Mol Biol Cell 2004;15:2523–2536. 28. Paik S, Shak S, Tang G et al. A multigene assay to predict recurrence of tamoxifen-treated, nodenegative breast cancer. N Engl J Med 2004;351:2817–2826. 29. Fan C, Oh DS, Wessels L et al. Concordance among gene-expression–based predictors for breast cancer. N Engl J Med 2006;355:560–569. 30. Chang HY, Nuyten DS, Sneddon JB et al. Robustness, scalability, and integration of a wound– response gene expression signature in predicting breast cancer survival. Proc Natl Acad Sci USA 2005;102:3738–3743. 31. Ma XJ, Wang Z, Ryan PD et al. A two-gene expression ratio predicts clinical outcome in breast cancer patients treated with tamoxifen. Cancer Cell 2004;5:607–616. 32. Perou CM, Sorlie T, Eisen MB et al. Molecular portraits of human breast tumours. Nature 2000;406:747–752. 33. Sorlie T, Tibshirani R, Parker J et al. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci USA 2003;100:8418–8423. 34. Naderi A, Teschendorff AE, Barbosa-Morais NL et al. A gene-expression signature to predict survival in breast cancer across independent data sets. Oncogene 2007;26:1507–1516.

Chapter 16 / Molecular Prognosis of Breast Cancer

297

35. Baker SG, Kramer BS, Srivastava S. Markers for early detection of cancer: statistical guidelines for nested case-control studies. BMC Med Res Methodol 2002;2:4. 36. Ma Y, Qian Y, Wei L et al. Population-based molecular prognosis of breast cancer by transcriptional profiling. Clin Cancer Res 2007;13:2014–2022. 37. Espina V, Geho D, Mehta AI et al. Pathology of the future: molecular profiling for targeted therapy. Cancer Invest 2005;23:36–46. 38. Rosen JM. Hormone receptor patterning plays a critical role in normal lobuloalveolar development and breast cancer progression. Breast Dis 2003;18:3–9. 39. Guo L, Abraham J, Flynn DC et al. Individualized survival and treatment response predictions for breast cancers using phospho-EGFR, phospho-ER, phospho-HER2/neu, phospho-IGF–IR/In, phospho-MAPK, and phospho–p70S6 K proteins. Int J Biol Markers 2007;22:1–11. 40. Bussey KJ, Chin K, Lababidi S et al. Integrating data on DNA copy number with gene expression levels and drug sensitivities in the NCI-60 cell line panel. Mol Cancer Ther 2006;5:853–867. 41. Ikediobi ON, Davies H, Bignell G et al. Mutation analysis of 24 known cancer genes in the NCI-60 cell line set. Mol Cancer Ther 2006;5:2606–2612. 42. Scherf U, Ross DT, Waltham M et al. A gene expression database for the molecular pharmacology of cancer. Nat Genet 2000;24:236–244. 43. Staunton JE, Slonim DK, Coller HA et al. Chemosensitivity prediction by transcriptional profiling. Proc Natl Acad Sci USA 2001;98:10787–10792. 44. Szakacs G, Annereau JP, Lababidi S et al. Predicting drug sensitivity and resistance: profiling ABC transporter genes in cancer cells. Cancer Cell 2004;6:129–137. 45. Solit DB, Garraway LA, Pratilas CA et al. BRAF mutation predicts sensitivity to MEK inhibition. Nature 2006;439:358–362. 46. Ma Y, Ding Z, Qian Y et al. Predicting cancer drug response by proteomic profiling. Clin Cancer Res 2006;12:4583–4589. 47. Nishizuka S, Charboneau L, Young L et al. Proteomic profiling of the NCI-60 cancer cell lines using new high-density reverse-phase lysate microarrays. Proc Natl Acad Sci USA 2003;100:14229–14234. 48. Paweletz CP, Charboneau L, Bichsel VE et al. Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene 2001;20:1981–1989. 49. Jiang H, Deng Y, Chen HS et al. Joint analysis of two microarray gene-expression data sets to select lung adenocarcinoma marker genes. BMC Bioinformatics 2004;5:81. 50. Nimgaonkar A, Sanoudou D, Butte AJ et al. Reproducibility of gene expression across generations of Affymetrix microarrays. BMC Bioinformatics 2003;4:27. 51. Emir B, Wieand S, Su JQ et al. Analysis of repeated markers used to predict progression of cancer. Stat Med 1998;17:2563–2578. 52. Pepe MS, Janes H, Longton G et al. Limitations of the odds ratio in gauging the performance of a diagnostic, prognostic, or screening marker. Am J Epidemiol 2004;159:882–890. 53. Heagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics 2000;56:337–344. 54. Heagerty PJ, Zheng Y. Survival model predictive accuracy and ROC curves. Biometrics 2005;61: 92–105. 55. McShane LM, Altman DG, Sauerbrei W et al. Reporting recommendations for tumor MARKer prognostic studies (REMARK). Breast Cancer Res Treat 2006;100:229–235. 56. McShane LM, Altman DG, Sauerbrei W et al. Reporting recommendations for tumor marker prognostic studies (remark). Exp Oncol 2006;28:99–105. 57. McShane LM, Altman DG, Sauerbrei W et al. REporting recommendations for tumor MARKer prognostic studies (REMARK). Nat Clin Pract Urol 2005;2:416–422. 58. McShane LM, Altman DG, Sauerbrei W et al. Reporting recommendations for tumor marker prognostic studies. J Clin Oncol 2005;23:9067–9072. 59. McShane LM, Altman DG, Sauerbrei W et al. REporting recommendations for tumor MARKer prognostic studies (REMARK). Nat Clin Pract Oncol 2005;2:416–422.

298

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

60. McShane LM, Altman DG, Sauerbrei W et al. REporting recommendations for tumour MARKer prognostic studies (REMARK). Br J Cancer 2005;93:387–391. 61. McShane LM, Altman DG, Sauerbrei W et al. Reporting recommendations for tumor marker prognostic studies (REMARK). J Natl Cancer Inst 2005;97:1180–1184. 62. McShane LM, Altman DG, Sauerbrei W et al. REporting recommendations for tumour MARKer prognostic studies (REMARK). Eur J Cancer 2005;41:1690–1696. 63. Hayes DF, Ethier S, Lippman ME. New guidelines for reporting of tumor marker studies in breast cancer research and treatment: REMARK. Breast Cancer Res Treat 2006;100:237–238.

17

Role of the Folate-Pathway and the Thymidylate Synthase Genes in Pediatric Acute Lymphoblastic Leukemia Treatment Response Lea Cunningham, MD, and Richard Aplenc, MD, MSCE CONTENTS Introduction Methotrexate Mechanis m of Action Methylenetetrahydrofolate Reductas e (MTHFR) Methylenetetrahydrofolate Dehdrogenas e (MTHFD1) Thymidylate Synthas e Gene Reduced Folate Carrier (RFC) Gene Dis cus s ion References

S UMMARY Acute lymphoblastic leukemia (ALL) is the most common childhood cancer with a cure rate of approximately 80%. However, despite the generally favorable outcome of ALL treatment, some children relapse or experience severe treatment side effects. Recent research efforts have focused on understanding the patient genetic characteristics that underlie treatment response. To date, several studies have demonstrated that From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 17, 

299

300

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

polymorphic genetic variation in genes in the folic acid cycle are associated with altered risks of ALL relapse or treatment toxicity. This chapter summarizes and reviews these results with a particular focus on the methylene tetrahydrofolate and thymidylate synthase genes. While providing intriguing data that germline genetic variation may determine ALL treatment response, adequately powered prospective trials are necessary to translate these findings into clinical practice. Key Words: Acute lymphoblastic leukemia; pharmacogenetics; single nucleotide polymorphisms; outcome; English; folate pathway; methotrexate

1. INTRODUCTION Acute lymphoblastic leukemia is the most common childhood cancer with cure rates of approximately 80% of all patients (1). This success rate largely stems from collaborative clinical trials that have developed risk stratification groups. Treatment intensity is proportional to relapse risk, with patients at relapse risks receiving more intensive, and therefore toxic, therapy. Thus, the overall goal of risk stratification is to balance successful treatment against toxicity. While substantial improvements have been made in risk stratification, current risk stratification algorithms still fail to identify prospectively the 20% of patients at risk for relapse. Likewise, these risk stratification methods imperfectly predict treatment toxicities, particularly less common or dose independent ones. The use of germline genetic data in risk stratification procedures has the potential to improve risk prediction and may ultimately allow for even more individualized treatment choices. This chapter focuses on genetic polymorphisms found in somatic DNA, particularly in the enzymes of the folate pathway, which is primarily targeted by methotrexate (MTX). Pediatric ALL treatment is comprised of multiple chemotherapy agents given in varied combinations over a several year time course. Figure 1 illustrates the interplay between methotrexate and other agents that in combination determine treatment response. Moreover, these pharmacogenetic factors are also modified by other factors, including patient compliance, concurrent medications, dietary factors, and environmental exposures. Thus, treatment response is clearly a complex outcome determined by the complex interplay of multiple mediators. The overarching purpose of the studies described below is to determine the role of germline polymorphisms amongst these other factors in determining treatment response.

2. METHOTREXATE MECHANISM OF ACTION Methotrexate inhibits folate metabolism by preventing methylenetetrahydrofolate reductase from converting 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate thus inhibiting thymidylate synthase conversion of dUMP to dTMP. DNA replication is effectively decreased by the diminution of dTMP availability. As shown in Fig. 2, multiple enzymes mediate the folate cycle. Thus, genetic variation in these enzymes may

Chapter 17 / Folate Pathway Pharmacogenetics and Pediatric

Fig. 1. Genetic factors impacting chemotherapy efficacy and toxicity.

Fig. 2. Folic acid cycle and methotrexate.

301

302

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

modify the clinical impact of methotrexate therapy. The peer-reviewed studies describing the impact of these genetic variants are summarized below.

3. METHYLENETETRAHYDROFOLATE REDUCTASE (MTHFR) Methylenetetrahydrofolate reductase is an approximately 19,301 base pair gene with 11 exons and located on chromosome 1p36.3 (2). Multiple polymorphic sites have been described, with the C677T and A1298G most often studied. As expected, allele frequency data varies by ethnicity: The MTHFR C677T variant allele is present in 34% of Caucasians, 20% of Italians and Hispanics, 14% of African-Americans, and 10 years (Total XIV). Aside from age (OR 24.2, p = 0.0001, 95% CI, 4.8–122.1) and Caucasian race (OR 11.1, P = 0.037, 95% CI, 1.2–105.6), children with the TS low activity 2R/2R wild-type polymorphism were at significantly increased risk (OR 7.2, p = 0.044; 95% CI, 1.05–48.9) for osteonecrosis of the hips (22). The authors note that osteonecrosis may have been over-represented in their sample due to the study design.

5.2. Thymidylate Synthase Confounding Issues As previously noted, difference in MTX dosing may confound associations between the 28 bp insertion and the clinical endpoints of interest. Likewise, biases in case ascertainment may also make interpretation of the studies more difficult. Relling et al. note this concern given the different indications for osteonecrosis screening used in Total Therapy XIIIB and XIV trials (22). Finally, the limited sample size of the Lauten et al. report makes interpretation of the reported negative result difficult (21).

Chapter 17 / Folate Pathway Pharmacogenetics and Pediatric

307

6. REDUCED FOLATE CARRIER (RFC) GENE The reduced folate carrier is the major transporter of methotrexate into cells. The RFC gene is comprised of 40,719 base pairs, has 12 exons, and is located on 21q22.3. The primary polymorphisms studied to date is the RFC1 G80A variant, which is associated with a decrease in MTX transport into cells. Rady et al. noted similar rates of variation in several ethnic groups: variant homozygosity rates were 28.7% in individuals of Ashkenazi-Jewish decent, 20.8% in African-American, 29% in Caucasians, and 26% in Hispanics (23).

6.1. RFC Studies Three studies to date have evaluated RFC polymorphisms and ALL treatment response (Table 3). Laverdiere et al. reported that patients homozygous for the RFC G80A variant had a decreased EFS compared to homozygous wild-type patients. Specifically, in Cox models correcting for age, WBC count, protocol, and risk stratification, the variant genotype had an increased hazard ratio of 2.8, 95% CI, 1.0–8.1, p = 0.05 ( 24). Interestingly, this corresponded to statistically significant higher MTX levels in patients homozygous for the 80A allele. The authors hypothesized that reduced intracellular transport of MTX may have resulted in an increased relapse rate and higher serum MTX level. A similar result was reported by Shimaski et al. who retrospectively evaluated toxicity in 15 Japanese children on four different protocols receiving methotrexate 3 g/m2 for 2 3 courses in leukemia or lymphoma therapy. In this study, increasing G allele copy number was significantly associated with an increased risk of emesis. However, hematologic and hepatic toxicity risks were not altered ( 12). Unlike these finding, the report by Kishi et al. did not find an association between neuro-toxicty and RFC G80A genotypes (7).

6.2. RFC Confounding Issues In addition to the frequent concerns of sample size and MTX dosing, clinical variation in the use of leucovorin may also confound these analyses. Furthermore, the Shimaski et al. study may be subject to treatment period effects as anti-emetic practices change over treatment eras (12). As noted by Laverdiere, prospective studies evaluating multiple folate pathway polymorphisms will be necessary to define clearly the relationship between the RFC G80A polymorphism and ALL treatment response (24).

7. DISCUSSION There has been great progress in our understanding of folate pharmacogenetics involved in pediatric acute lymphoblastic leukemia treatment response. Several studies have shown positive association between certain polymorphisms or haplotypes and treatment outcome. MTHFR C677T is associated with increased risk of relapse ( 6) while RFC A80G ( 24), TS 3R/3R (18), and MTHFD1 ( 5) are associated with decreased EFS. Associations have also been reported with treatment toxicity. Specifically, the RFC A80G polymorphism is associated with more vomiting (12) and children with TS 2R/2R have increased risk of osteonecrosis (22).

Table 3 RFC Polymorphisms Implicated in ALL Treatment Response 308

Gene

Polymorphism

Event

Case

Control

Or

95% CI

P

Ref.

RFC RFC RFC

A80 variant 80 G/A 80 G/A

EFS Vomiting (NCI-CTC grade 2 or higher) Neurotoxicity

35

169 15 53

2.8 3.13

1.1–8.1 1.081–9.072

0.03 0.036 NS

(24) (12) (7)

14

Chapter 17 / Folate Pathway Pharmacogenetics and Pediatric

309

While these positive associations provide promising evidence that genetic variation influences ALL treatment response, many of these studies share important limitations. Many of these studies are retrospective with modest sample sizes and varied inclusion criteria. Moreover, few studies consider the impact of multiple comparisons the reported positive associations. Further advances in pediatric ALL pharmacogenetics will be made through large-scale cooperative group studies that prospectively determine both the genotype and outcome of interest and then validate these associations in subsequent trials. As data from these studies matures, host genotype information may be used to direct the clinical care of patients.

REFERENCES 1. Pui CH, Sandlund JT, Pei D et al. Improved outcome for children with acute lymphoblastic leukemia: results of Total Therapy Study XIIIB at St Jude Children’s Research Hospital. Blood 2004;104: 2690–2696. 2. Goyette P, Sumner JS, Milos R et al. Human methylenetetrahydrofolate reductase: isolation of cDNA mapping and mutation identification. Nat Genet 1994;7:551. 3. Botto LD, Yang Q. 5,10-Methylenetetrahydrofolate reductase gene variants and congenital anomalies: a HuGE review. Am J Epidemiol 2000;151:862–877. 4. Robien K, Ulrich CM. 5,10-Methylenetetrahydrofolate reductase polymorphisms and leukemia risk: a HuGE minireview. Am J Epidemiol 2003;157:571–582. 5. Krajinovic M, Lemieux-Blanchard E, Chiasson S et al. Role of polymorphisms in MTHFR and MTHFD1 genes in the outcome of childhood acute lymphoblastic leukemia. Pharmacogenomics J 2004;4:66–72. 6. Aplenc R, Thompson J, Han P et al. Methylenetetrahydrofolate reductase polymorphisms and therapy response in pediatric acute lymphoblastic leukemia. Cancer Res 2005;65:2482–2487. 7. Kishi S, Griener J, Cheng C et al. Homocysteine, pharmacogenetics, and neurotoxicity in children with leukemia. J Clin Oncol 2003;21:3084–3091. 8. Jazbec J, Kitanovski L, Aplenc R et al. No evidence of association of methylenetetrahydrofolate reductase polymorphism with occurrence of second neoplasms after treatment of childhood leukemia. Leuk Lymphoma 2005;46:893–897. 9. Chiusolo P, Reddiconto G, Casorelli I et al. Preponderance of methylenetetrahydrofolate reductase C677T homozygosity among leukemia patients intolerant to methotrexate. Ann Oncol 2002;13: 1915–1918. 10. Ulrich CM, Yasui Y, Storb R et al. Pharmacogenetics of methotrexate: toxicity among marrow transplantation patients varies with the methylenetetrahydrofolate reductase C677T polymorphism. Blood 2001;98:231–234. 11. Costea I, Moghrabi A, Laverdiere C et al. Folate cycle gene variants and chemotherapy toxicity in pediatric patients with acute lymphoblastic leukemia. Haematologica 2006. 12. Shimasaki N, Mori T, Samejima H et al. Effects of methylenetetrahydrofolate reductase and reduced folate carrier 1 polymorphisms on high-dose methotrexate-induced toxicities in children with acute lymphoblastic leukemia or lymphoma. J Pediatr Hematol Oncol 2006;28:64–68. 13. Hol FA, van der Put NM, Geurds MP et al. Molecular genetic analysis of the gene encoding the trifunctional enzyme MTHFD (methylenetetrahydrofolate-dehydrogenase, methenyltetrahydrofolatecyclohydrolase, formyltetrahydrofolate synthetase) in patients with neural tube defects. Clin Genet 1998;53:119–125. 14. Kaneda S, Nalbantoglu J, Takeishi K et al. Structural and functional analysis of the human thymidylate synthase gene. J Biol Chem 1990;265:20277–20284. 15. Marsh S, Collie-Duguid ES, Li T et al. Ethnic variation in the thymidylate synthase enhancer region polymorphism among Caucasian and Asian populations. Genomics 1999;58:310–312.

310

Part II / Pharmacogenomics of Toxicity and Response of Chemotherapy

16. Mandola MV, Stoehlmacher J, Zhang W et al. A 6 bp polymorphism in the thymidylate synthase gene causes message instability and is associated with decreased intratumoral TS mRNA levels. Pharmacogenetics 2004;14:319–327. 17. Krajinovic M, Costea I, Chiasson S. Polymorphism of the thymidylate synthase gene and outcome of acute lymphoblastic leukaemia. Lancet 2002;359:1033–1034. 18. Krajinovic M, Costea I, Primeau M et al. Combining several polymorphisms of thymidylate synthase gene for pharmacogenetic analysis. Pharmacogenomics J 2005;5:374–380. 19. Costea I, Moghrabi A, Krajinovic M. The influence of cyclin D1 (CCND1) 870 A>G polymorphism and CCND1-thymidylate synthase (TS) gene–gene interaction on the outcome of childhood acute lymphoblastic leukaemia. Pharmacogenetics 2003;13:577–580. 20. Rocha JC, Cheng C, Liu W et al. Pharmacogenetics of outcome in children with acute lymphoblastic leukemia. Blood 2005;105:4752–4758. 21. Lauten M, Asgedom G, Welte K et al. Thymidylate synthase gene polymorphism and its association with relapse in childhood B-cell precursor acute lymphoblastic leukemia. Haematologica 2003;88:353–354. 22. Relling MV, Yang W, Das S et al. Pharmacogenetic risk factors for osteonecrosis of the hip among children with leukemia. J Clin Oncol 2004;22:3930–3936. 23. Rady PL, Szucs S, Matalon RK et al. Genetic polymorphism (G80A) of reduced folate carrier gene in ethnic populations. Mol Genet Metab 2001;73:285–286. 24. Laverdiere C, Chiasson S, Costea I et al. Polymorphism G80A in the reduced folate carrier gene and its relationship to methotrexate plasma levels and outcome of childhood acute lymphoblastic leukemia. Blood 2002;100:3832–3834.

III

P HARMACOGENOMICS IN C LINICAL D RUG D EVELOPMENT IN O NCOLOGY

“This page left intentionally blank.”

18

Pharmacogenomics in Drug Development: A Pharmaceutical Industry Perspective Tal Zaks, MD, PhD CONTENTS I NTRODUCTION BASIC P REMISES : P IPELINES AND T IMELINES G ERMLINE P HARMACOGENETICS IN O NCOLOGY D RUG D EVELOPMENT: L EARNING FROM L APATINIB P HARMACOGENOMICS P REDICTORS OF E FFICACY IN D RUG D EVELOPMENT T HE R EGULATORY E NVIRONMENT M ARKET S EGMENTATION C ONCLUSIONS R EFERENCES

S UMMARY The elucidation of the human genome sequence and the advent of genomic technologies have the potential to facilitate drug discovery and development as well as to define individual risks and benefits associated with specific therapeutic interventions. This chapter focuses on the application of this knowledge within the pharmaceutical industry, by providing current examples of the relevance of both germline and somatic genotypic variations to adverse event and efficacy profiles of recently developed anti-cancer drugs. These examples, discussed within the scientific, regulatory, and economic frameworks that shape the industry, highlight both the potential benefits and the emerging challenges to the application of post-genomic science to “real-life” drug development.

From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 18, 

313

314

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

Key Words: Single nucleotide polymorphisms; CYP2C19; lapatinib; ADME; pharmacogenetics

1. INTRODUCTION The completed elucidation of the human genome sequence has led to expectations that discovery of novel drug targets would be rapidly translated into effective therapeutic interventions; and that understanding the genomic basis for physiological and pathophysiological variation would lead to an era of tailored drug and dose selection based on this knowledge (1). From an economic perspective it can be argued that pharmacogenomic strategies would best benefit drugs that have narrow therapeutic indices and a high degree of variability in responses between individual patients; and would be most practical where there are limitations in methods for monitoring adverse effects (AEs) and treatment responses, or where there are relatively few treatment options (2); it is thus no surprise that oncology drug development is at the vanguard of pharmacogenomic approaches. Drug discovery and development occur in the western world almost exclusively within the biotechnology and pharmaceutical industries. With the availability of rapidly maturing technologies that enable genome-wide elucidation of gene sequences and alterations as well as characterization of gene products (RNA and protein), linking this knowledge with clinical parameters of safety and efficacy that are measured in clinical trials has become feasible. While these technologies can and are being applied to the processes of target, compound, and patient selection, this chapter will focus only on the latter, specifically the industry’s approach to a genomic understanding of the two components of the “risk–benefit” equation. The term pharmacogenetics is used here to specificy the relationship between inherited DNA variation and drug response, whereas pharmacogenomics will be used in the broader sense to encompass the relationship of DNA as well as other biological constituents (e.g., RNA, proteins) to both clinical toxicity and efficacy.

2. BASIC PREMISES: PIPELINES AND TIMELINES The pharmaceutical industry’s goal of developing beneficial drugs is achieved through a well-organized process termed the pharmaceutical pipeline (Fig. 1). An outline of this process facilitates an understanding of both opportunities and challenges presented by pharmacogenomics. Compounds are selected (i.e., candidate selection, CS) based on promising chemical and biological properties that predict developability, a process that lasts anywhere from 12 months to several years. Following this, candidates are tested for toxicity in preclinical animal models, resulting in the filing of an IND (Investigational New Drug application) with the regulatory agencies to obtain authorization to initiate the evaluation of the drug’s safety and efficacy in man; this phase usually takes 12–18 months and follows well-defined regulatory guidelines. The next phase involves testing in humans, with “first time in human” (FTIH)/phase I studies aimed at evaluating

Chapter 18 / Pharmacogenomics – An Industry Perspective

315

Fig. 1. The Pharmaceutical Pipeline. The different phases of drug development are outlined, starting with a lead compound and moving forward through preclinical and clinical testing towards market approval. (CS: candidate selection; IND: investigational new drug application; POC: proof of concept; NDA: new drug application; FTIH: first time in human).

the safety of the new compound and selecting an appropriate dose, phase II studies geared at providing a “proof of concept” (POC) regarding the drug’s efficacy, and, finally, phase III studies where the magnitude of clinical benefit over the standard care is defined and safety is assessed on a large number of patients. This latter part takes several years, as it depends on accumulating enough event rates, sufficient treatment duration, and length of follow up to ascertain efficacy and safety. Furthermore, in oncology the process of phase II–IV testing invariably continues in parallel in multiple types of cancer and in combination with other therapies even after achieving an initial marketing approval. Given the overall time required to successfully develop a compound into a marketed drug, this uniform drug development approach has been successful at defining tangible milestones and providing a framework for assessing an overall return on investment. The return on investment then becomes a function of the value to the patient and society within the context of a limited drug patent life. As has been recognized, this process resides in its entirety only within the pharmaceutical industry ( 3), which has the resources and expertise required to accomplish these tasks and the metrics needed to make transition-point decisions on the potential progression of new compounds to market. This assembly-line paradigm, even superimposed on a limited scientific understanding of disease etiology, has been remarkably successful in bringing new and effective drugs to market. The ever-increasing pace of scientific discovery and the deciphering of the human genome had been expected to yield a plethora of novel targets and means to realize their benefit (4,5); paradoxically however, the overall efficiency of the drug development process has been declining (6). In so far as pharmacogenomics is the study of the relationship between the effects of a drug and the genomic makeup of the host (or tumor), new discoveries and technologies should enable a more productive discovery and development pipeline by maximizing the benefit–risk ratio of therapy—and doing so at earlier stages of drug discovery, thus reducing (expensive) attrition later in development. While this is a promising vision shared by the pharmaceutical industry, the regulatory agencies, and academia (7), many practical hurdles need to be overcome for this vision to be realized within acceptable costs, risks, and timeframes.

316

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

3. GERMLINE PHARMACOGENETICS IN ONCOLOGY DRUG DEVELOPMENT: LEARNING FROM LAPATINIB In developing drugs that target cancer cells, one example of the influence of germline genetic polymorphisms on the risk–benefit profile of a drug effect may be derived from differences in toxicity between individuals that relate to individual differences in drug metabolism and transport. Nearly all of the examples of such associations have been derived retrospectively by academic centers outside the pharmaceutical industry (such as the relationship between the UGT1A1 polymorphism and irinotecan (8) and reviewed elsewhere in this volume). However, such a retrospective analysis, even if it unambigiously defines a subpopulation at increased risk, cannot easily determine the relationship to drug efficacy. This then translates into a clinical conundrum: Does one reduce the dose in susceptible patients or provide the “full” dose while supporting the patient through manageable AEs (i.e., with granulocyte growth factors to reduce drug-induced neutropenia)? Conceptually, several criteria need to be fulfilled for a pharmacogenetic association to be clinically meaningful: 1. The adverse event needs to be clinically significant enough that the risk-benefit of therapy is percieved to be altered. In this regard, the acceptable risk–benefit profile of anti-cancer therapies is markedly different than, for example, that of anti-hypertension therapy. 2. The genetic association needs to be robust enough such that the magnitude of attributable risk (the percent of events associated with a given genotype) and penetrance (the chance that a person with the given genotype will have an event) can account for the majority of adverse events (i.e., sufficient positive and negative predictive power). There is a considerable statistical gap between a relative risk (RR) ratio that defines a “risk factor,”and one that can be used for prospectively selecting patients for a given intervention (9). 3. An analytically validated assay needs to be available in the clinic. 4. The clinical impact of dose-modification based on a genetic test needs to be prospectively assessed both in terms of toxicity and efficacy; retrospective association analyses cannot by definition provide either an optimally safe or optimally efficacious dose in test-positive individuals.

The availability of both comprehensive SNP databases (10) and a plethora of technologies available to determine DNA variants at ever-decreasing costs has enabled the pharmaceutical industry to begin to incorporate germline DNA collection and testing into clinical trials ( 11). This allows for hypotheses to be developed and tested from the start of phase I testing in humans, when a direct correlation can be made between toxicity, efficacy, and pharmacokinetic variables. Furthermore, this allows the sponsor to pool data from several studies to significantly increase the statistical power. The technicalities of calculating the necessary sample size to form and test a hypothesis have been reviewed elsewhere (12), but can be summed up in principle: (i) There is a fundamental relationship between the magnitude of attributable effect, the number of genes (and/or SNPs) analyzed, and the sample size required. (ii) Hypotheses for associations can be derived from relatively small numbers of patients, especially if genes are preselected, but then require testing (“validation”) on substantial numbers of additional patients. (iii) Using genome-wide data to form and test hypotheses will require more patients than are required for testing prespecified hypotheses.

Chapter 18 / Pharmacogenomics – An Industry Perspective

317

These principles can be illustrated by an ongoing pharmacogenetic experiment supR ) at GlaxoSmithKline. Lapatinib is a porting the development of lapatinib (Tykerb small molecule oral dual-kinase inhibitor of EGFR (ERBB1) and HER2/neu (ERBB2). Early in its development, diarrhea and skin rash were anticipated class adverse effects R R ; AstraZeneca), erlotinib (Tarceva , OSI Pharmaceuticals), [seen with gefitinib (Iressa  R and cetuximab (Erbitux ; Bristol-Myers Squibb) and attributable to EGFR inhibition] and were indeed seen in FTIH (healthy volunteer) and phase I studies, albeit at mild grades (13). Lapatinib had been shown ex vivo to be primarily metabolized by CYP3A4/5 with a minor contribution from CYP2C19, and to be a substrate of the transporters MDR1(ABCB1) and BCRP(ABCG2). A hypothesis was proposed whereby polymorphisms in these genes would be associated with a propensity to develop rash and/or diarrhea. Retrospective exploratory analyses were conducted to identify any associations between pharmacokinetic (PK) parameters or incidence of diarrhea or rash with single nucleotide polymorphisms (SNPs) in the genes encoding the enzymes CYP3A4, 3A5, 2C19, and transporters ABCB1 and ABCG2. This may potentially lead not only to an ability to predict which patient may be at risk, but also, by inference, to better understand the relevance of these genes and pathways to lapatinib metabolism in vivo—and thus to predict potential drug–drug interactions with other agents. Two hundred and eighty-four single nucleotide polymorphisms (SNPs) in these genes were genotyped in a cohort of 107 Caucasian subjects from eight phase I lapatinib studies (there were an additional 10 Black and 9 Hispanic subjects genotyped, but to prevent spurious associations resulting from different allele frequencies in different ethnic groups, analyses were limited to Caucasian subjects only). Of note, these were comprised of 34 cancer patients plus 73 healthy volunteers, and PK data was available for 100 of them. These cases were analyzed in both a casecontrolled manner (defined as cases with either any grade rash or diarrhea, or those at the >90% exposure as measured by dose-normalized AUC), and by quantitative trait analyses (QTL). The most intriguing results showed that 18 different SNPs in CYP2C19 were associated with Tmax (p < 0.01). SNPs in this gene from the same linkage disequilibrium group (which includes the non-functional CYP2C19*2 allele) showed significant association with incidence of rash and diarrhea (p < 0.01, unadjusted). Of note, all three CYP2C19*2/*2 individuals in this cohort (two of which were healthy volunteers) developed some grade of rash and diarrhea. In addition, QTL analyses suggested that SNPs in ABCB1 were associated with lapatinib PK (AUC, Cmax, Tmax) (14). These preliminary observations demonstrate the feasibility of applying pharmacogenetics from the start of clinical development and the potential to elucidate relevant hypotheses from a relatively small sample size, although the results require replication and validation. Testing these hypotheses will require confirmation in a larger patient population (i.e., from phase II/III studies), which should also enable an assessment of the magnitude of effect and attributable risk. However, since the inception of these experiments the clinical profile of lapatinib has also become clearer. In the pivotal phase III study of lapatinib in combination with capecitabine (15), rates of diarrhea and skin rash were higher on the combination arm

318

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

than on the capecitabine-alone arm; however, by and large these were of low grade I/II severity that were clinically manageable without requiring dose reductions. Whether further genetic analyses to elucidate a genetic predisposition to rash and diarrhea will be warranted will ultimately depend on the clinical magnitude of the problem. There have been relatively few germline pharmacogenetic analyses of other AEs induced by EGFR tyrosine kinase inhibitors. One recently published report evaluated 4 SNPs (of known functional consequences) in EGFR ( 2), ABCB1 ( 1), and ABCG2 ( 1) in 178 Cacausian patients treated with gefitinib (both erlotinib and gefitinib are known substrates of ABCB1 and ABCG2). Seven (44%) of the 16 patients with at least one variant ABCG2 allele developed grade 1 or 2 diarrhea, whereas only 13 (12%) of 108 patients carrying the wild-type sequence for both alleles did (p = .0046). It is noteworthy that the one patient with the homozygous variant genotype had no noticeable toxicity (16). These results, while statistically significant, are unlikely to lead to pharmacogenetically determined dosing for the same reasons alluded above; the observation is not likely to confer a high enough predictive power to be clinically useful, and the severity of the adverse events are unlikely to warrant a dose reduction of an efficacious anti-cancer drug. Finally, while the above-mentioned associations occur with germline polymorphisms, there is a potential relationship between intra-tumoral expression of drug transporters and anti-cancer efficacy. Thus, expression of ABCG2 protects EGFR signaling-dependent A431 tumor cells from death after gefitinib treatment. A germline SNP associated with reduced efflux in the gut (and thus a proponsity toward diarrhea) could also be associated with reduced efflux from tumor cells elsewhere in the body. and an increased likelihood of tumor response to therapy (17). In the case of lapatinib, in addition to rash and diarrhea, cardiotoxicity was anticipated R ; Genetech), an antibody based on the known toxicity profile of trastuzumab (Herceptin that targets HER2/neu. This led to a rigorous assessment of cardiac function at baseline and at regular intervals in all patients treated with lapatinib. The current incidence of asympomatic declines in ejection fraction in all patients treated with lapatinib is approximately 1.4%. Some of these events may represent testing errors (as some cardiac function tests had returned to baseline despite contiued drug therapy) and the incidence of sympomatic heart failure is in the order of 0.2%. Nevertheless, given the obvious potential severity of this side effect and following a report linking the Ile655Val SNP in HER2/neu to trastuzumab-induced cardiotoxicity (18), an attempt was made to analyze this SNP (as well as a few other SNPs in candidate genes such as MMP2, MMP9, and ERBB4) in a case-controlled study of lapatinib treated patients. Asymptomatic decline in cardiac function was seen in 41 of 3127 patients, of whom 21 had provided samples and consents for DNA analyses. Eighteen of these (the Caucasian cases) were compared to 210 matched controls, and no associations with cardiotoxicity were found. These results are due to either (i) the sample size being underpowered to detect a weak assocation, or (ii) the lack of a genetic basis in these genes for lapatinib-mediated cardiotoxicity. These results illustrate the difficulties inherent to elucidating a genetic basis of severe adverse events. Because drug development is designed to ensure that these events are rare, it is likely that only very strong associations (and concomitantly high relative risks) can be determined from the sample sizes usually available in clinical trials. However, the ability to

Chapter 18 / Pharmacogenomics – An Industry Perspective

319

establish even negative findings can be of obvious interest where data on comparative agents exist. The above-mentioned experiments could not have been done without the routine collection of germline DNA samples in clinical trials, which should be a part of all modern clinical trials (3). Moreover, when such a resource exists, additional analyses can be performed as either more cases accumulate or novel genetic associations worth testing are described elsewhere.

4. PHARMACOGENOMICS PREDICTORS OF EFFICACY IN DRUG DEVELOPMENT While a comprehensive review of the pharmaceutical industry’s adjustment to the post-genomic era of pharmacogenetic predictors is beyond the scope of this chapter, several salient concepts will be discussed below, including determining the right marker, finding the right technology by which to measure it, and the move toward multiple markers derived in parallel from genome-wide analyses.

4.1. Measuring the Right Pharmacogenomic Predictor: Single Gene Markers One of the earliest examples of predicting efficacy has come from studies of breast cancer, where anti-estrogen therapy is clearly effective in patients whose tumors express the estrogen receptor, and ineffective when no such receptors are present. However, even in this well-studied case, current consensus statements recognize an intermediate “endocrine response uncertain” state, where “the exact boundary between ‘endocrine responsive’ and ‘endocrine response uncertain’ is undecided, and may well be different in different clinical settings” (19). The most-often cited recent example is probably that of the development of trastuzumab by Genetech. Post-development economic analyses have estimated that by selecting the patient population that overexpresses the target (HER2/neu), a putative response rate of 10% was increased to 50%, enabling clinical efficacy to be established with much smaller trials and shorter follow-up, saving millions of dollars in direct clinical trial costs and accelerating time-to-market by 8 years, thus enabling the treatment of 120,000 patients who might otherwise not have had access to this drug ( 20). The development of trastuzumab has thus become the “poster child” for patient selection R ; Novartis), which was in oncology, followed closely by imatinib mesylate (Gleevac tested initially in cancers that harbored a mutated oncogenic target (i.e., BCR-ABL in CML and c-KIT in gastrointestinal stromal tumors) (21). The targeted development of these two drugs has been enabled by straightforward and well-validated tests for the underlying genomic aberration that they target: immunohistochemistry and later fluorescent in situ hybridization in the case of HER2/neu, and the Philadelphia-chromosome translocation in the case of CML. Following the success of imatinib, mutations in the BCR-ABL kinase domain associated with resistance were discovered in patients and served as the basis for the in vitro development of dasatinib R , Bristol-Myers Squibb). This drug was developed very rapidly in patients (Sprycel with leukemias refractory to imatinib, and was recently approved by the FDA for that indication.

320

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

It is worth noting that while dasatinib is active against most imatinib-resistance mutations, it remains inactive against the T315I point mutation (22). However, because the drug showed sufficient efficacy in patients with the Philadelphia chromosome-positive leukemia who had progressed on imatinib, the patients were not tested specifically for the mutations against which the drug is active. Even in the era of known oncogenic kinases inhibited by specific drugs, it remains unclear how to best predict for clinical efficacy. Gefitinib, an ERBB1 inhibitor was initially approved in the United States and Japan for use in non-small cell lung cancer based on promising results of early phase studies. It was later found that responses (tumor shrinkage) occurred preferentially in tumors harboring ERBB1 mutations. Subsequent phase III studies done without selection for ERBB1 mutations failed to show efficacy when combined with chemotherapy. As a result, the drug was removed from the U.S. market. However, it appears that gefitinib may be more active in Asians, and whether this is due solely to the increased frequency of ERBB1 mutations in lung cancers in Asians remains to be determined. Currently, the drug is being tested in U.S. lung cancer patients whose tumors harbor these mutations. In this case, the decision not to incorporate a pharmacogenetic predictor for a response likely resulted in a negative pivotal trial and loss of a potentially active drug. However, it should be acknowledged that despite several years of intensive research and multiple clinical trials involving hundreds of patients with gefitinib and a similar drug, erlotinib, the best way to predict for clinical efficacy remains uncertain. In the case of erlotinib, significant benefit also may be seen in patients without ERBB1 mutations. Additional genomic markers predictive of benefit may include ERBB1 amplifications, ERBB2 amplifications, and absence of KRAS mutations (reviewed in Refs. 23 and 24). The opposite case, of “over-selecting” a patient population, has also been seen recently, and with another ERBB1 inhibitor—the monoclonal antibody cetuximab. Initial studies were conducted in patients whose tumors had “immunohistochemical evidence of EGFR expression” (25). This ultimately led to inclusion of this criterion in the drug label. It was subsequently found that benefit from cetuximab may also occur in the absence of ERBB1 expression (26). Recent trials in head and neck cancer have shown efficacy in unselected patients, with the role of ERBB1 expression relegated to post-hoc analysis (27). It remains to be determined whether this is the results of inaccurate measurement of the target protein ( 28) or that of measuring the wrong analyte, as responses may correlate better with target gene amplification or an oncogenic mutation elsewhere in the signal transduction pathway (29) rather than with protein overexpression. As a final example, trastuzumab has recently been shown effective against ERBB2+ breast cancer in the adjuvant setting, leading to a ∼50% reduction of relative risk of recurrence (19,30). What was unexpected was that most (though not all) of this benefit appeared limited to tumors that have a co-amplified c-MYC oncogene, on top of the amplified ERBB2 (31). Given that some efficacy is also seen in tumors without c-MYC amplification, these results are not likely to change clinical practice. There are emerging genomic predictive markers for trastuzumab resistance (e.g., high IGF-1R and loss of PTEN (32,33), that do not appear to preclude clinical response to lapatinib (34). The ability to distinguish patients who benefit from one drug vs. another, based on PGX markers, is of obvious clinical interest. These examples illustrate that

Chapter 18 / Pharmacogenomics – An Industry Perspective

321

while the pharmaceutical industry is embracing predictive genomic testing in the way it is developing current anti-cancer therapies, our scientific knowledge is still the limiting factor even for well-defined aberrations that are thought to be readily measurable.

4.2. Genome-Wide Markers The ability to assay in one experiment the entire genome via microchip (and other) technologies has magnified both the opportunities (e.g., defining a population likely to benefit) and challenges (e.g., coming up with the best diagnostic test) described above. Currently, cost considerations and the technical hurdles of deriving such data from paraffin-embedded tumor tissue have limited the utility of such approaches to post-hoc analyses in selected cases, where they are used in an effort to glean biological understanding of mechanisms of efficacy and resistance. Potentially, these data can then be deconvoluted into simpler qPCR-based assays that are more amendable to routine clinical use, as in the case of the prognostic assays for breast cancer ( 35, 36). However, as these technologies mature they are being incorporated into an increasing number of clinical trials (for recent reviews, see Refs. 37,38,39). Furthermore, novel clinical trial designs have been proposed that integrate knowledge learned from these platforms as interim endpoints to facilitate more efficient adaptive trials. Appreciation of the relevance of genomic aberrations to drug development coupled with the availability of these technologies are driving the pharmaceutical industry toward molecular pathology as the focal point between cancer biology and drug efficacy (40), a domain traditionally occupied by academic labs. In the coming years, the expanding academic knowledge base of genomic profiles that mark a different biological phenotype as measured by prognosis ( 41, 42, 43), activation of a specific pathway (which can be directly relevant to the drug being studied) (44), or known markers of clinical response to other therapeutic agents (45) will allow genome-wide data from clinical trials to be analyzed against a priori hypotheses, significantly limiting the number of samples needed for a meaningful result. Use of our scientific understanding, however, is constrained by the current clinical knowledge and infrastructure. Elucidating clinically relevant genomic markers will require much more comprehensive databases where the depth and quality of the clinical data can match that of the molecular data. At the present, resource limitations, issues regarding patient privacy rights, and the public’s fear of genetic testing are recognized impediments (46). The vision outlined above is not merely a futuristic one. The Molecular Profiling Institute has recently reported a test that assays relevant genomic pathways and that can suggest clinically relevant targets for patients who have progressed on prior chemotherapy; this test is currently being prospectively validated (47). Clearly, as more and different targeted agents become available, such approaches will become even more effective. The challenge, however, will be of designing clinical trials that can test the clinical efficacy of such approaches in ever smaller groups of patients. The examples cited to date have succeeded largely because the target population was either a significant subpopulation or readily identifiable by pre-existing clinical and diagnostic criteria. However, once the prevalence of a genomic aberration becomes less than

322

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

∼10% in the population, testing those patients under current clinical trial paradigms becomes a significant practical hurdle. For example, it has not been possible to test for the efficacy of trastuzumab in either ERBB2+ lung (48) or prostate (49) cancer.

5. THE REGULATORY ENVIRONMENT The declining economic efficiency of drug development led the FDA to try to identify root causes and potential solutions to this problem in its Critical Path Initiative, in which pharmacogenomics was recognized as “one of the technologies that will lead to innovation in the pharmaceutical industry” (6). This has been followed by several advisories to industry, such as an opportunity for “safe-harbor” voluntary submission of genomic data by sponsors (50) and a series of collaborative workshops (7). These above efforts have been very effective in understanding challenges and opportunities. For example, they have contributed to an understanding for the need for codevelopment of drugs and the diagnostic assays that predict their individual effectiveness (or toxicity) (51,52). The importance and success of these efforts to the pharmaceutical industry cannot be over-emphasized and are often under-appreciated in the academic world. In a highly regulated environment where success depends on both clinical efficacy and regulatory acceptability, these guidelines ultimately enable more rationale risk-management decisions. Much work remains to be done not only in forging the necessary regulatory frameworks, but also in harmonizing as much as possible these frameworks between the different continents to enable the drug development process to proceed efficiently across borders.

6. MARKET SEGMENTATION One cannot discuss an industry perspective on pharmacogenomics without addressing the traditional concern of segmenting the market for a given drug such that sales and revenues will be significantly decreased. This has implications for both investment and drug pricing, as the cost of developing and marketing a drug with a narrow target population may be similar to that of developing a “blockbuster” drug (53). It has become clear within oncology, however, that not only can this approach be profitable (54), but developing drugs specifically targeting a genomic aberration that is understood from the outset can also result in significantly reduced development times and costs, as in the above-cited case of dasatinib. Identification of patients who would not have otherwise qualified by virtue of a novel test, as has been the case for the use of imatinib in gastrointestinal stromal tumors that express a mutated KIT receptor, and increased treatment duration and compliance linked to prolongation of survival or efficacy in the adjuvant setting. In the final analysis, a significant clinical benefit can be expected to lead to higher adoption rates.

7. CONCLUSIONS The pharmaceutical industry is adopting and developing the knowledge and technologies afforded by the completion of the human genome sequence in the post-genomic era to develop drugs with better benefit–risk ratios in a more efficacious manner.

Chapter 18 / Pharmacogenomics – An Industry Perspective

323

Increasingly, blood samples for germline DNA analyses and tumor tissue for somatic DNA analyses are routinely requested as part of industry sponsored clinical trials (3), the results of which are beginning to bear fruit (55). Many challenges remain, however, in realizing the vision of individualized drug selection and optimization, and it should be acknowledged that the timelines for gains in clinical knowledge and “real-life” applicability will always lag behind the pace of basic scientific discovery. Success in this endeavor will continue to require a close collaboration between academic, regulatory, and pharmaceutical partners.

REFERENCES 1. Guttmacher AE, Collins FS. Realizing the promise of genomics in biomedical research. JAMA 2005;294:1399–1402. 2. Flowers CR, Veenstra D. The role of cost–effectiveness analysis in the era of pharmacogenomics. PharmacoEconomics 2004;22:481–493. 3. Roses AD. Pharmacogenetics and drug development: the path to safer and more effective drugs. Nat Rev Genet 2004;5:645–656. 4. Collins FS. Genetics: an explosion of knowledge is transforming clinical practice. Geriatrics 1999;54:41–47; quiz 8. 5. Lindsay MA. Target discovery. Nat Rev Drug Discov 2003;2:831–838. 6. Challenges and Opportunities on the Critical Path to New Medical Products. In: Food and Drug Administration, 2004. http://www.fda.gov/oc/initiatives/criticalpath/whitepaper.html.Accessed 17 May 2008. 7. Salerno RA, Lesko LJ. Three years of promises, proposals, and progress on optimizing the benefit–risk of medicines: a commentary on the 3rd FDA-DIA-PWG-PhRMA-BIO pharmacogenomics workshop. Pharmacogenomics J 2006;6:78–81. 8. Innocenti F, Ratain MJ. Irinotecan treatment in cancer patients with UGT1A1 polymorphisms. Oncology (Williston Park) 2003;17:52–55. 9. Ware JH. The limitations of risk factors as prognostic tools. N Engl J Med 2006;355:2615–2617. 10. Conrad DF, Jakobsson M, Coop G et al. A worldwide survey of haplotype variation and linkage disequilibrium in the human genome. Nat Genet 2006;38:1251–1260. 11. Gibson N, Jawaid A, March R. Novel technology and the development of pharmacogenetics within the pharmaceutical industry. Pharmacogenomics 2005;6:339–356. 12. Cardon LR, Idury RM, Harris TJ et al. Testing drug response in the presence of genetic information: sampling issues for clinical trials. Pharmacogenetics 2000;10:503–510. 13. Burris HA, 3rd, Hurwitz HI, Dees EC et al. Phase I safety, pharmacokinetics, and clinical activity study of lapatinib (GW572016), a reversible dual inhibitor of epidermal growth factor receptor tyrosine kinases, in heavily pretreated patients with metastatic carcinomas. J Clin Oncol 2005;23:5305–5313. 14. Zaks T, Akkari A, Briley L et al. Role of pharmacogenetic studies in early clinical development: Phase I studies with lapatinib. J Clinl Oncol 2006;24:3029. 15. Geyer CE, Forster J, Lindquist D et al. Lapatinib plus capecitabine for HER2-positive advanced breast cancer. N Engl J Med 2006;355:2733–2743. 16. Cusatis G, Gregorc V, Li J et al. Pharmacogenetics of ABCG2 and adverse reactions to gefitinib. J Natl Cancer Inst 2006;98:1739–1742. 17. Elkind NB, Szentpetery Z, Apati A et al. Multidrug transporter ABCG2 prevents tumor cell death induced by the epidermal growth factor receptor inhibitor Iressa (ZD1839, Gefitinib). Cancer Res 2005;65:1770–1777. 18. Milano G, Lescaut W, Formento J et al. HER2 genetic polymorphism and pharmacodynamics of trastuzumab-based treatment in breast cancer patients. J Clin Oncol 2005;23:501. 19. Piccart-Gebhart MJ, Procter M, Leyland-Jones B et al. Trastuzumab after adjuvant chemotherapy in HER2-positive breast cancer. N Engl J Med 2005;353:1659–1672.

324

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

20. Press M, Seelig S. Lesson learned from the development of a diagnostic to predict response to herceptin. In: Targeted Medicine, 2004. 21. Druker BJ. Imatinib as a paradigm of targeted therapies. J Clin Oncol 2003;21:239s–245. 22. Shah NP, Tran C, Lee FY et al. Overriding imatinib resistance with a novel ABL kinase inhibitor. Science 2004;305:399–401. 23. Amler LC, Goddard AD, Hillan KJ. Predicting clinical benefit in non-small-cell lung cancer patients treated with epidermal growth factor tyrosine kinase inhibitors. Cold Spring Harb Symp Quant Biol 2005;70:483–488. 24. Mendelsohn J, Baselga J. Epidermal growth factor receptor targeting in cancer. Semin Oncol 2006;33:369–385. 25. Cunningham D, Humblet Y, Siena S et al. Cetuximab monotherapy and cetuximab plus irinotecan in irinotecan–refractory metastatic colorectal cancer. N Engl J Med 2004;351:337–345. 26. Chung KY, Shia J, Kemeny NE et al. Cetuximab shows activity in colorectal cancer patients with tumors that do not express the epidermal growth factor receptor by immunohistochemistry. J Clin Oncol 2005;23:1803–1810. 27. Bonner JA, Harari PM, Giralt J et al. Radiotherapy plus cetuximab for squamous–cell carcinoma of the head and neck. N Engl J Med 2006;354:567–578. 28. Penault-Llorca F, Cayre A, Arnould L et al. Is there an immunohistochemical technique definitively valid in epidermal growth factor receptor assessment? Oncol Rep 2006;16:1173–1179. 29. Lievre A, Bachet JB, Le Corre D et al. KRAS mutation status is predictive of response to cetuximab therapy in colorectal cancer. Cancer Res 2006;66:3992–3995. 30. Romond EH, Perez EA, Bryant J et al. Trastuzumab plus adjuvant chemotherapy for operable HER2positive breast cancer. N Engl J Med 2005;353:1673–1684. 31. Kim C, Bryant J, Horne Z et al. Trastuzumab sensitivity of breast cancer with coamplification of HER2 and cMYC suggests pro-apoptotic function of dysregulated cMYC in vivo. Breast Cancer Res Treat 2005;94:S6. 32. Lu Y, Zi X, Zhao Y et al. Insulin-like growth factor–I receptor signaling and resistance to trastuzumab (Herceptin). J Natl Cancer Inst 2001;93:1852–1857. 33. Nagata Y, Lan KH, Zhou X et al. PTEN activation contributes to tumor inhibition by trastuzumab, and loss of PTEN predicts trastuzumab resistance in patients. Cancer Cell 2004;6:117–127. 34. Gomez HL, Chavez MA, Doval DC et al. Biomarker results from a phase II randomized study of lapatinib (GW572016) as first-line treatment for patients with ErbB2 FISH-amplified advanced or metastatic breast cancer. Breast Cancer Res Treat 2005;94:S63. 35. Glas AM, Floore A, Delahaye LJ et al. Converting a breast cancer microarray signature into a highthroughput diagnostic test. BMC Genomics 2006;7:278. 36. Fan C, Oh DS, Wessels L et al. Concordance among gene-expression–based predictors for breast cancer. N Engl J Med 2006;355:560–569. 37. Burczynski ME, Oestreicher JL, Cahilly MJ et al. Clinical pharmacogenomics and transcriptional profiling in early phase oncology clinical trials. Curr Mol Med 2005;5:83–102. 38. Dracopoli NC. Development of oncology drug response markers using transcription profiling. Curr Mol Med 2005;5:103–110. 39. Strand KJ, Khalak H, Strovel JW et al. Expression biomarkers for clinical efficacy and outcome prediction in cancer. Pharmacogenomics 2006;7:105–115. 40. Campbell DA, Carmichael J, Chopra R. Molecular pathology in oncology: the AstraZeneca perspective. Pharmacogenomics 2004;5:1167–1173. 41. Perou CM, Sorlie T, Eisen MB et al. Molecular portraits of human breast tumours. Nature 2000;406:747–752. 42. van de Vijver MJ, He YD, van’t Veer LJ et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002;347:1999–2009. 43. Potti A, Mukherjee S, Petersen R et al. A genomic strategy to refine prognosis in early-stage nonsmall-cell lung cancer. N Engl J Med 2006;355:570–580.

Chapter 18 / Pharmacogenomics – An Industry Perspective

325

44. Bild AH, Yao G, Chang JT et al. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature 2006;439:353–357. 45. Chang JC, Wooten EC, Tsimelzon A et al. Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 2003;362:362–369. 46. Dalton WS, Friend SH. Cancer biomarkers: an invitation to the table. Science 2006;312:1165–1168. 47. Von Hoff DD, Penny RS, Shack S et al. Frequency of potential therapeutic targets identified by immunohistochemistry (IHC) and DNA microarray (DMA) in tumors from patients who have progressed on multiple therapeutic agents. J Clin Oncol 2006;24:18S. 48. Langer CJ, Stephenson P, Thor A et al. Trastuzumab in the treatment of advanced non-small-cell lung cancer: is there a role? Focus on Eastern Cooperative Oncology Group Study 2598. J Clin Oncol 2004;22:1180–1187. 49. Lara PN, Jr., Chee KG, Longmate J et al. Trastuzumab plus docetaxel in HER-2/neu–positive prostate carcinoma: final results from the California Cancer Consortium Screening and Phase II Trial. Cancer 2004;100:2125–2531. 50. Frueh FW, Rudman A, Simon K et al. Experience with voluntary and required genomic data submissions to the FDA: summary report from track 1 of the third FDA-DIA-PWG-PhRMA-BIO pharmacogenomics workshop. Pharmacogenomics J 2006;6:296–300. 51. Goodsaid F, Frueh F. Process map proposal for the validation of genomic biomarkers. Pharmacogenomics 2006;7:773–782. 52. Gutman S, Kessler LG. The U.S. Food and Drug Administration perspective on cancer biomarker development. Nat Rev Cancer 2006;6:565–571. 53. DiMasi JA, Hansen RW, Grabowski HG. The price of innovation: new estimates of drug development costs. J Health Econ 2003;22:151–185. 54. Million RP. Impact of genetic diagnostics on drug development strategy. Nat Rev Drug Discov 2006;5:459–462. 55. Risner ME, Saunders AM, Altman JF et al. Efficacy of rosiglitazone in a genetically defined population with mild-to-moderate Alzheimer’s disease. Pharmacogenomics J 2006;6:246–254.

“This page left intentionally blank.”

19

Identification of Pharmacogenomic Biomarker Classifiers in Cancer Drug Development Richard Simon, DSc CONTENTS I NTRODUCTION P HARMACOGENOMIC B IOMARKER C LASSIFIERS T YPES OF P HARMACOGENOMIC B IOMARKER C LASSIFIERS D EVELOPING E MPIRICAL P HARMACOGENOMIC C LASSIFIERS U SING G ENE E XPRESSION D EVELOPMENTAL AND VALIDATION S TUDIES E STIMATES OF P REDICTIVE ACCURACY IN D EVELOPMENTAL S TUDIES U SE OF P HARMACOGENOMIC C LASSIFIERS IN N EW D RUG D EVELOPMENT C ONCLUSIONS ACKNOWLEDGEMENTS R EFERENCES

S UMMARY Physicians need improved tools for selecting treatments for individual patients. Many syndromes traditionally viewed as individual diseases are heterogeneous in molecular pathogenesis and treatment responsiveness. This results in treatment of many patients with ineffective drugs and leads to the conduct of large clinical trials to identify small average treatment benefits for heterogeneous groups of patients. From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 19, 

327

328

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

New genomic and proteomic technologies provide powerful tools for the selection of patients likely to benefit from a therapeutic program without unacceptable adverse events. In this chapter we attempt to clarify how pharmacogenomic biomarker classifiers of the patients most likely to benefit from a drug can be identified and utilized during clinical development. Key Words: Pharmacogenomics; biomarker; genomics; DNA microarray; clinical trial design; validation

1. INTRODUCTION Physicians need improved tools for selecting treatments for individual patients. For example, many cancer treatments benefit only a minority of the patients to whom they are administered. Being able to predict which patients are most likely to benefit would not only save patients from unnecessary toxicity and inconvenience, but might facilitate their receiving drugs that are more likely to help them. In addition, the current over-treatment of patients results in major expense for individuals and society, an expense that may not be indefinitely sustainable. In this discussion we will address some key issues in the validation of pharmacogenomic classifiers.

2. PHARMACOGENOMIC BIOMARKER CLASSIFIERS Much of the discussion about disease biomarkers is in the context of markers that measure some aspect of disease status, extent, or activity. Such biomarkers are often proposed for use in early detection of disease or as a surrogate endpoint for evaluating prevention or therapeutic interventions. The validation of such biomarkers is difficult for a variety of reasons, but particularly because the molecular pathogenesis of many diseases is incompletely understood, and hence it is not possible to establish the biological relevance of a measure of disease status. A pharmacogenomic biomarker is any pre-treatment measurable quantity that can be used to select treatment; for example, the result of an immunohistochemical assay for a single protein, the abundance of a protein in serum, the abundance of mRNA transcripts for a gene in a sample of disease tissue, or the presence/absence status of a specified germline polymorphism or tumor mutation. A pharmacogenomic biomarker classifier is a mathematical function that translates the biomarker values to a set of prognostic categories. These categories generally correspond to levels of predicted clinical outcome. With the advent of gene expression profiling, it is increasingly common to define composite pharmacogenomic biomarker classifiers based on the levels of expression of dozens of genes. For a fully specified classifier, however, all of the parameters and cutpoints are specified for determining how to weight the different components and how to map the multivariate data into a defined set of categories. A completely defined classifier can be used to select patients and stratify patients for therapy in clinical trials that enable the clinical value of the classifier to be evaluated. Specifying only the genes involved does not enable one to structure prospective clinical validation experiments in which patients are assigned or stratified in prospectively well-defined ways.

Chapter 19 / Pharmacogenomic Biomarker Classifiers

329

3. TYPES OF PHARMACOGENOMIC BIOMARKER CLASSIFIERS Pharmacogenomic biomarkers can either be defined based on the known molecular target of the drug or empirically developed by comparing responders versus nonresponders with regard to whole genome tumor characterizations such as transcript expression profiling. The former approach is preferable for a variety of reasons. First, a biomarker with a strong biological rationale linked to the mechanism of action of the drug is more satisfying than a “black box” classifier. Second, development time is likely to be shorter, and finally a reproducible assay for a single gene/protein biomarker may be implementable on a platform that does not require fresh/frozen tumor. Mutation-targeted drugs are drugs that are specific to a mutated gene that is driving the growth of a tumor. For example, there is currently considerable interest in developing drugs specific for mutated B-raf , which is present in approximately 60 percent of human melanomas. Such drugs have automatic pharmacogenomic biomarkers based on assaying for the presence of the mutation. In general, the mutation can either be a point mutation as in B-raf, gene amplification, or deletion. Many current cancer drugs are selected to inhibit oncogenes that are mutated in some tumors. Although the drugs are not designed to be specific for the mutated forms of the associated proteins, presence or absence of a mutation can be used as a natural pharmacogenomic biomarker of the patients most likely to benefit from the drugs. Papadopoulous et al. review the experience with molecularly targeted drugs of this type (1). Although all chemotherapeutic drugs are “molecularly targeted” in the sense that they interact with specific intracellular components ( 1), in some cases the true target is not known when the drug is developed; in some cases there are multiple targets; and in many cases there is no obvious assay for determining the extent to which the target is driving tumor growth. In these cases one must generally use empirical methods to develop pharmacogenomic biomarker classifiers of the patients most likely to benefit from the drug. With this approach a training set of tumor specimens from patients who have responded to the drug is assayed and compared to a training set of specimens from patients who have not responded. The specimens are assayed, using either whole-genome technology, such as expression profiling, or using assays based on candidate genes, and a predictive classifier is developed for identifying the tumors most likely to respond. In the next section, we will describe some aspects of the development of such classifiers using whole-genome transcript expression profiling.

4. DEVELOPING EMPIRICAL PHARMACOGENOMIC CLASSIFIERS USING GENE EXPRESSION There are three components to the empirical approach of developing a predictive classifier. The first component is determining which genes to include in the predictor. This is generally called “feature selection.” Including too many “noise variables” in the predictor usually reduces the accuracy of prediction. The second component is specification of the mathematical function that will provide a prediction for any given expression vector.

330

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

The third component is parameter estimation. Most predictors have parameters that must be assigned values before the predictor is fully specified. For many kinds of predictors there is also a cut-point that must be specified for translating a quantitative predictive index into a prediction (e.g., 0 or 1) for binary class prediction problems. Feature selection is usually based on identifying the genes that are differentially expressed among the classes when considered individually. For example, if there are two classes, one can compute a t-test or a modified t-test in which a hierarchical variance model is used for increasing the degrees of freedom for estimation of the gene-specific within-class variances ( 2). The logarithm of the expression measurements are used as the basis of the statistical significance tests. The genes that are significantly differentially expressed at a specified significance level are selected for inclusion in the class predictor. The stringency of the significance level that is used controls the number of genes that are included in the model. Although many computationally complex methods have been published to identify optimal sets of genes that together provide good discrimination, little compelling evidence currently exists that the computational effort of these methods is warranted. Many algorithms have been used effectively with DNA microarray data for predicting of a binary outcome, e.g., response versus non-response. Dudoit et al. ( 3) compared several algorithms using several publicly available data sets. A linear discriminant is a function  l(x) = wi xi (1) i∈F

where xi denotes the logarithm of the expression measurement for the ith gene, wi is the weight given to that gene, and the summation is over the set F of features (genes) selected for inclusion in the class predictor. For a two-class problem, there is a threshold value d, and a sample with expression profile defined by a vector x of values is predicted to be in class 1 or class 2 depending on whether l(x) as computed from equation (1) is less than the threshold d or greater than d, respectively. Many types of classifiers are based on linear discriminants of the form shown in ( 1). They differ with regard to how the weights are determined. The oldest form of linear discriminant is Fisher’s linear discriminant. To compute the weights for the Fisher linear discriminant, one must estimate the correlation between all pairs of genes that were selected in the feature selection step. The study by Dudoit et al. indicated that Fisher’s linear discriminant did not perform well unless the number of selected genes was small relative to the number of samples. The reason is that in other cases there are too many correlations to estimate and the method tends to be unstable and over-fit the data. Diagonal linear discriminant analysis is a special case of Fisher linear discriminant analysis in which the correlation among genes is ignored. By ignoring such correlations, one avoids having to estimate many parameters, and obtains a method that performs better when the number of samples is small. Golub’s weighted voting method (4) and the Compound Covariate Predictor of Radmacher et al. (5) are similar to diagonal linear discriminant analysis and tend to perform very well when the number of samples is

Chapter 19 / Pharmacogenomic Biomarker Classifiers

331

small. They compute the weights based on the univariate prediction strength of individual genes and ignore correlations among the genes. Support vector machines are very popular in the machine learning literature. Although they sound very exotic, linear kernel support vector machines do class prediction using a predictor of the form of equation (1). The weights are determined by optimizing a misclassification rate criterion, however, instead of a least-squares criterion as in linear discriminant analysis ( 6). Although there are more complex forms of support vector machines, they appear to be inferior to linear kernel SVM’s for class prediction with large numbers of genes (7). In the study of Dudoit et al. (3), the simplest methods, diagonal linear discriminant analysis, and nearest neighbor classification, performed as well or better than the more complex methods. Nearest neighbor classification is defined as follows. It depends on a feature set F of genes selected to be useful for discriminating the classes. It also depends upon a distance function d(x, y)w which measures the distance between the expression profiles x and y of two samples. The distance function utilizes only the genes in the selected set of features F. To classify a sample with expression profile y, compute d(x, y)f for each sample x in the training set. The predicted class of y is the class of the sample in the training set thst is closest to y with regard to the distance function d. A variant of nearest neighbor classification is k-nearest neighbor classification. For example with 3-nearest neighbor classification, you find the three samples in the training set that are closest to the sample y. The class that is most represented among these three samples is the predicted class for y. Tibshirani et al. (8) developed a variant called shrunken centroid classification that combines the gene selection and nearest centroid classification components. Dudoit et al. also studied some more complex methods such a classification trees and aggregated classification trees. These methods did not appear to perform any better than diagonal linear discriminant analysis or nearest neighbor classification. Ben-Dor et al. ( 7) also compared several methods on several public datasets and found that nearest neighbor classification generally performed as well or better than more complex methods.

5. DEVELOPMENTAL AND VALIDATION STUDIES It is important to distinguish the studies that develop parmacogenomic classifiers from those that utilize such classifiers for targeting treatment selection or for evaluating the clinical utility of such classifiers. The vast majority of published prognostic marker studies are developmental. Developmental studies are often based on a convenience sample of patients for whom tissue is available but who are heterogeneous with regard to treatment and stage. Although there is a large literature on prognostic markers, few such factors are used in clinical practice. Prognostic markers are unlikely to be used unless they are therapeutically relevant, and most developmental studies are not based on a cohort medically coherent enough to establish therapeutic relevance. The patients included in a developmental study of a pharmacogenomic biomarker to be used in drug development should be appropriate to enable identification of patients who are most likely to benefit from the new drug in a pivotal study. For example, suppose

332

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

that the pivotal study involves advanced disease patients who have failed first-line treatment and involves comparing survivals for patients receiving the new drug to survivals for patients receiving palliative care. Patients from single arm phase II trials of the new drug can be used to develop a pharmacogenomic biomarker classifier of those patients likely to respond to the new drug. Dobbin and Simon (9) have studied sample size considerations for developmental studies of predictive binary classifiers and have indicated that generally at least 20 cases in each class are required. Consequently, a phase II database containing at least 20 responders and 20 non-responders would be needed for the development of a pharmacogenomic classifier to be used in the subsequent pivotal trials. This may require a larger phase II developmental program than is conventional. If the pivotal study involves comparison of outcome for patients receiving a standard regimen C versus those receiving C plus the new drug, then development of a gene expression–based classifier is more complex. The classifier can be developed based on phase II studies of patients receiving C plus the new drug, but unless one also studied patients receiving C without the new drug one would not know whether prediction was drug specific or just reflected general responsiveness of the tumors. It is possible to develop pharmacogenomic predictors of risk of tumor progression rather than tumor response. Even if the patients are receiving the investigational drug as a single agent, however, it may not be clear to what extent the predictor reflects drug effect rather than non-specific disease pace. As indicated in the previous paragraphs, there are limitations to the adequacy of a conventional phase II database for empirically developing a pharmacogenomic classifier for use in a pivotal study. In many ways the best resource for developing a pharmacogenomic biomarker classifier for use in a pivotal trial is a collection of pre-treatment tumor specimens from patients enrolled in such a pivotal trial. For example, archived material from a “failed” pivotal trial of the drug can be used to develop a biomarker classifier of patients most likely to benefit from the drug compared to the control. The classifier can be based on the actual endpoint used in the clinical trial or upon an intermediate endpoint such as progression-free survival for which there may be more events available. By “failed” pivotal trial, we mean a trial for the same target population of patients that did not establish a statistically significant benefit for the drug for the randomized patients as a whole. The classifier developed based on archived material in a failed pivotal trial should be considered to have the same status as a classifier based on a phase II database. That is, the classifier should be used to design a new pivotal trial that establishes the clinical benefit of the drug in a prospectively specified subset of patients. Using the same pivotal trial to develop a pharmacogenomic classifier and to test treatment effects in subsets determined by the classifier is generally not valid. Freidlin and Simon ( 10) have shown, however, how one pivotal trial can be used potentially for both purposes—if the set of patients used to develop the classifier is kept distinct from the set of patients used to evaluate treatment benefit. Generally, however, the studies should be kept separate. Developmental studies are exploratory, though they should result in completely specified binary classifiers. Studies on which claims of drug benefit are based should be non-exploratory, but should instead test prospectively defined hypotheses about treatment effect in a pre-defined patient population.

Chapter 19 / Pharmacogenomic Biomarker Classifiers

333

6. ESTIMATES OF PREDICTIVE ACCURACY IN DEVELOPMENTAL STUDIES Developmental studies are analogous to phase 2 clinical trials. They should include an indication of whether the pharmacogenomic classifier is promising and worthy of phase 3 evaluation. There are special problems, however, in evaluating whether classifiers based on high-dimensional genomic or proteomic assays are promising. The difficulty derives from the fact that the number of candidate features available for use in the classifier is much larger than the number of cases available for analysis. In such situations, it is always possible to find classifiers that accurately classify the data on which they were developed even if there is no relationship between expression of any of the genes and outcome (5). Consequently, even in developmental studies, some kind of validation on data not used for developing the model is necessary. This “internal validation” is usually accomplished either by splitting the data into two portions, one used for training the model and the other for testing the model, or some form of cross-validation based on repeated model development and testing on random data partitions. This internal validation should not, however, be confused with external truly independent validation of the classifier. The most straightforward method for estimating the prediction accuracy is the splitsample method of partitioning the set of samples into a training set and a test set. Rosenwald et al. ( 11) used this approach successfully in their international study of prognostic prediction for large B-cell lymphoma. They used two-thirds of their samples as a training set. Multiple predictors were studied on the training set. When the collaborators of that study agreed on a single fully specified prediction model, they accessed the test set for the first time. On the test set there was no adjustment of the model or fitting of parameters. They merely used the samples in the test set to evaluate the predictions of the model that was completely specified using only the training data. In addition to estimating the overall error rate on the test set, one can also estimate other important operating characteristics of the test such as sensitivity, specificity, and positive and negative predictive values. The split-sample method is often used with so few samples in the test set, however, that the validation is almost meaningless. One can evaluate the adequacy of the size of the test set by computing the statistical significance of the classification error rate on the test set or by computing a confidence interval for the test set error rate. Because the test set is separate from the training set, the number of errors on the test set has a binomial distribution. Michiels et al. (12) suggested that multiple training-test partitions be used, rather than just one. The split-sample approach is mostly useful, however, when one does not have a well-defined algorithm for developing the classifier. When there is a single training set-test set partition, one can use biological insight on the training set to develop a classifier and then test that classifier on the test set. With multiple training-test partitions, however, that type of flexible approach to model development cannot be used. If one has an algorithm for classifier development, it is generally better to use one of the cross-validation or bootstrap resampling approaches to estimating error rate (see below) because the split-sample approach does not provide as efficient a use of the available data (13).

334

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

Cross-validation is an alternative to the split-sample method of estimating prediction accuracy ( 5). Molinaro et al. describe and evaluate many variants of cross-validation and bootstrap re-sampling for classification problems where the number of candidate predictors vastly exceeds the number of cases (13). The cross-validated prediction error is an estimate of the prediction error associated with application of the algorithm for model building to the entire dataset. A commonly used invalid estimate is called the re-substitution estimate. You use all the samples to develop a model. Then you predict the class of each sample using that model. The predicted class labels are compared to the true class labels and the errors are totaled. It is well known that the re-substitution estimate of error is highly biased for small data sets and the simulation of Simon et al. (14) confirmed that, with a 98.2% of the simulated data sets resulting in zero misclassifications even when no true underlying difference existed between the two groups. Simon et al. (14) also showed that cross-validating the prediction rule after selection of differentially expressed genes from the full data set does little to correct the bias of the re-substitution estimator: 90.2% of simulated data sets with no true relationship between expression data and class still result in zero misclassifications. When feature selection was also re-done in each cross-validated training set, however, appropriate estimates of mis-classification error were obtained; the median estimated misclassification rate was approximately 50%.

7. USE OF PHARMACOGENOMIC CLASSIFIERS IN NEW DRUG DEVELOPMENT With a pharmacogenomic classifier for predicting which patients are likely to benefit from an available treatment regimen, the emphasis should be on validation of the clinical utility of using the classifier. With an experimental therapy, however, the emphasis should be on demonstrating effectiveness of the drug in a population identified by the classifier. Simon and Maitournam (15,16) demonstrated that use of a genomic classifier for focusing a clinical trial in this manner can result in a dramatic reduction in required sample size, depending on the sensitivity and specificity of the classifier for identifying such patients. Not only can such targeting provide a huge improvement in efficiency in phase III development, it also provides an increased therapeutic ratio of benefit to toxicity and results in a greater proportion of treated patients who benefit. Simon and Maitournam consider use of the targeted design shown in Fig. 1. During pre-clinical and phase I/II clinical development one identifies a fully specified classifier of which patients have a high probability of responding to the experimental drug. That classifier is then used to select patients for phase III trial. This is a form of enrichment design. Table 1 shows the approximate number of events required in order to have 80% statistical power for comparing exponential survival times using the design of Fig. 1. The 196 events shown in Table 1 is compared to the number of events required in a standard clinical trial if the classifier is not used to select patients for randomization (Table 2). The table assumes that the treatment is not effective for the classifier negative patients. More extensive results on relative efficiency of the targeted and untargeted designs are described by Simon and Maitournam (15,17).

Chapter 19 / Pharmacogenomic Biomarker Classifiers

335

Fig. 1. Targeted clinical trial design for evaluating a new experimental therapy. A biomarker classifier is developed for identifying those patients most likely to respond to the new treatment (E). Only those patients are randomized to E versus the control treatment. The patients predicted less likely to respond (marker negative) are off-study. The targeted design is most useful in cases where the biomarker classifier has a strong biological rationale for identifying responsive patients and where it may not be ethically advisable to expose marker negative patients to the new treatment.

For many molecularly targeted drugs, however, the appropriate assay for selecting patients is not known and development of a classifier based on comparing expression profiles for phase II responders versus phase II non-responders may be the best approach. In such instances, one may not have sufficient confidence in the genomic classifier developed in phase II to use it for excluding patients in phase III trials as in Fig. 1. It may be better in this case to accept all conventionally eligible patients, and use the classifier in the pre-defined analysis plan. Figure 2 shows the marker by treatment interaction design discussed by Sargent et al. (18) and by Pusztai and Hess (19). Both marker positive and marker negative patients are randomized to the experimental treatment or control. The analysis plan either calls for separate evaluation of the treatment difference in the two-marker strata or for testing the hypothesis that the treatment effect is the same in both marker strata. When this design is used for development of an experimental drug, an appropriate analysis plan might be to utilize a preliminary test of interaction; if the interaction is Table 1 Approximate Number of Events Required for 80% Power with 5% Two-Sided Log-Rank Test for Comparing Randomized Arms of Design Shown in Fig. 1. Only Marker + Patients Are Randomized. Treatment Hazard Ratio for Marker + Patients Is Shown in First Column. Time-To-Event Distributions Are Exponential Hazard Ratio for Marker + Patients

Number of Events Required

0.5 0.67

66 196

336

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

Table 2 Approximate Number of Events Required for 80% Power with 5% Two-Sided Log-Rank Test for Comparing Treatment Versus Control arm of Design in which Marker is not Measured. Randomized Arms Are Mixtures of Marker – and Marker + Patients. Hazard Ratio For Marker – Patients Is 1 for the Two Treatment Groups and 0.67 For Marker + Patients. % of Patients Marker

Approximate Number of Events Required

20 33 50

5200 1878 820

not significant at a pre-specified level, then the experimental treatment is compared to the control overall. If the interaction is significant, then the treatment is compared to the control within the two strata determined by the marker. The sample size planning for such a trial and determination of the appropriate significance level for the preliminary interaction test are discussed by Simon (20). Simon and Wang (21) proposed an alternative analysis plan for the design of Fig. 2. They suggested that the overall null hypothesis for all randomized patients is tested at the 0.04 significance level. A portion, e.g., 0.02, of the usual 5 percent false positive rate is reserved for testing the new treatment in the subset predicted by the classifier to be responsive. The analysis starts with a test of the overall null hypothesis, without a preliminary test of interaction. If the overall null hypothesis is rejected, then one concludes that the treatment is effective for the randomized population as a whole and that the classifier is not needed. If the overall null hypothesis is not rejected at the 0.03 level, then a single subset analysis is conducted; comparing the experimental treatment to the control in the subset of patients predicted by the classifier as being most likely to be responsive to

Fig. 2. Stratified analysis design for evaluating a new experimental treatment (E) relative to a control (C). The status of a biomarker based classifier of the likelihood of responding to E is utilized in a prospectively specified analysis plan. The biomarker classifier is not just used for stratifying the randomization. Alternative analysis plans are described in the text.

Chapter 19 / Pharmacogenomic Biomarker Classifiers

337

the new treatment. If the null hypothesis is rejected, then the treatment is considered effective for the classifier determined subset. This analysis strategy provides sponsors an incentive for developing genomic classifiers for targeting therapy in a manner that does not unduly deprive them of the possibility of broad labeling indications when justified by the data.

8. CONCLUSIONS Physicians need improved tools for selecting treatments for individual patients. The genomic technologies available today are sufficient to develop such tools. There is not broad understanding of the steps needed to translate research findings of correlations between gene expression and prognosis into robust diagnostics validated to be of clinical utility. This paper has attempted to identify some of the major steps needed for such translation.

ACKNOWLEDGEMENTS Thanks to Dr. Wenyu Jiang for the computing of Table 2.

REFERENCES 1. Papadopoulos N, Kinzler KW, Vogelstein B. The role of companion diagnostics in the development and use of mutation-targeted cancer therapies. Nat Biotechnol 2006;24:985–995. 2. Wright GW, Simon R. A random variance model for detection of differential gene expression in small microarray experiments. Bioinformatics 2003;19:2448–2455. 3. Dudoit S, Fridlyand J, Speed TP. Comparison of discrimination methods for classification of tumors using gene expression data. J Am Stat Assoc 2002;97:77–87. 4. Golub TR, Slonim DK, Tamayo P et al. Molecular classification of cancer: class discovery and class prediction by gene expression modeling. Science 1999;286:531–537. 5. Radmacher MD, McShane LM, Simon R. A paradigm for class prediction using gene expression profiles. J Computat Biol 2002;9:505–511. 6. Ramaswamy S, Tamayo P, Rifkin R et al. Multiclass cancer diagnosis using tumor gene expression signatures. Proc Natl Acad Sci USA 2001; 98:15149–15154. 7. Ben-Dor A, Bruhn L, Friedman N et al. Tissue classification with gene expression profiles. J Comput Biol 2000;7:559–583. 8. Tibshirani R, Hastie T, Narasimhan B et al. Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci USA 2002;99:6567–6572. 9. Dobbin K, Simon R. Sample size planning for developing classifiers using high-dimensional DNA expression data. Biostatistics 2007;8:101–117. 10. Freidlin B, Simon R. Adaptive signature design: An adaptive clinical trial design for generating and prospectively testing a gene expression signature for sensitive patients. Clin Cancer Res 2005;11:7872–7878. 11. Rosenwald A, Wright G, Chan WC et al; Lymphoma/;eukemia molecular profiling project. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 2002;346:1937–1947. 12. Michiels S, Koscielny S, Hill C. Prediction of cancer outcome with microarrays: a multiple random validation strategy. Lancet 2005;365:488–492. 13. Molinaro AM, Simon R, Pfeiffer RM. Prediction error estimation: A comparison of resampling methods. Bioinformatics 2005;21:3301–3307.

338

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

14. Simon R, Radmacher MD, Dobbin K et al. Pitfalls in the analysis of DNA microarray data for diagnostic and prognostic classification. J Natl Cancer Inst 2003;95:14–18. Review. 15. Simon R, Maitournam A. Evaluating the efficiency of targeted designs for randomized clinical trials. Clin Cancer Res 2004;10:6759–6763. 16. Simon R, Maitnourim A. Evaluating the efficiency of targeted designs for randomized clinical trials: Supplement and Correction. Clin Cancer Res 2006;12:3229. 17. Maitournam A, Simon R. On the efficiency of targeted clinical trials. Stat Med 2005;24:329–339. 18. Sargent DJ, Conley BA, Allegra C et al. Clinical trial designs for predictive marker validation in cancer treatment trials. J Clin Oncol 2005;23:2020–2027. 19. Pusztai L, Hess KR. Clinical trial design for microarray predictive marker discovery and assessment. Ann Oncol 2004;15:1731–1737. Review. 20. Simon R. Designs and adaptive analysis plans for pivotal clinical trials of therapeutics and companion diagnostics. Expert opinion on Medical Dignostics (in press). 21. Simon R, Wang SJ. Use of genomic signatures in therapeutics development. The Pharmacogenomics Journal 2006;6:166–173.

20

Toxicogenomics Application to Oncology Drug Development Luigi Calzolai, PhD, and Teresa Lettieri, PhD CONTENTS I NTRODUCTION OVERVIEW OF G ENE E XPRESSION A NALYSES H OW G OOD A RE T HE DATA AND H OW S HOULD T HEY B E A NALYZED ? A PPLICATIONS OF T OXICOGENOMIC TO O NCOLOGY D RUG D EVELOPMENT A PPLICATIONS OF G ENE E XPRESSION P ROFILING TO D RUG R ESISTANCE AND C ANCER T REATMENT C ONCLUSIONS ACKNOWLEDGEMENTS R EFERENCES

S UMMARY Toxicogenomics tries to evaluate the toxicity and safety of chemical compounds by analyzing gene expression changes detected by measuring mRNA levels using DNA microarrays. The measurement of gene expression levels upon exposure to toxicants can give information about the mechanism of action of the toxic compound, and also predict the toxicity of unknown compounds, thus providing a valuable tool in the investigation of lead compounds for drug development. There are also a few examples of its use to identify safety issues of drugs that were not evident from preclinical studies. Gene expression profiling can greatly improve cancer treatment by providing predictive markers for drug response, drug resistance, and disease prognosis, and guide physicians in the use of chemotherapeutics. One of the major issues facing this field From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 20, 

339

340

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

physicians in the use of chemotherapeutics. One of the major issues facing this field is the quality and the comparability of the experimental data acquired using different technological platforms, together with the public accessibility of the toxicological data. The formation of international consortia would probably be the most effective way to provide the critical mass of high-quality, highly annotated, gene expression data in the public domain. Toxicogenomics and gene expression profiling technologies have the potential to provide safer and more specific drugs, and also to make available more effective and personalized treatments to oncology patients. Key Words: Toxicogenomics; microarray; gene expression profile; genetic signature; drug development; biomarker; drug resistance

1. INTRODUCTION Toxicogenomics tries to assess the toxicity and safety of chemical compounds by analyzing gene expression changes detected by measuring mRNA levels using DNA microarrays (1,2). The analyses of gene expression is a potent way to characterize biological processes of cells and organisms and to analyze how they react to changes to their “normal state” such as in the presence of toxicants or during a disease. The measurement of gene expression levels, upon exposure to a chemical or a drug, can be used both to provide information about the mechanism of action of toxicants, and also to form a sort of “genetic signature” from the pattern of gene expression changes it elicits both in vitro (3) or in vivo (4). Such genetic signatures can be used for the fast screening of unknown or suspected toxicants based on their similarity to known toxicants, and they would be extremely useful in the preclinical phase of drug discovery. Gene expression profiles can help in different ways in providing better treatments to cancer patients: (i) they can be used in preclinical studies to analyze the mechanism of action and predict the toxicity of lead compounds, thus potentially producing safer and more effective drugs ( 5); (ii) they can be used to develop predictive markers for drug response including those involved in drug resistance (6,7); (iii) they can provide molecular subtypes information for disease prognosis and guide the choice of the most appropriate treatment regime (8,9,10). The goal of this review is to describe the application of microarray technology for the development of new drugs in oncology. Examples of the contribution of toxicogenomics in drug development in oncology will be described, as well as problems associated with the technique. We will also briefly analyze the potential use of gene expression profiling in improving cancer treatment and drug resistance.

2. OVERVIEW OF GENE EXPRESSION ANALYSES DNA microarray technology was developed in the early 1990s. Its refinements, together with great improvements during these years, defined the technology as the platform of choice for the application of gene expression profiling in many fields, from toxicology and ecotoxicology (11), to drug development and diagnostic tools (8).

Chapter 20 / DNA MicroArray and Oncology Drug Development

341

Recently, the U.S. Food and Drug Administration approved the “Amplichip,” the first DNA microrarray application in clinic (12) to help physicians to test the dosage of drugs that are differently metabolized by cytochrome P450 enzyme variants. The field of DNA microarray has evolved from the key insight of Ed Southern (13), who showed that it is possible to attach nucleic acid to solid support. The resulting Southern blot can be viewed has the first DNA array ( 14). It was only a small step to improve the technique to filter-based screening of clone libraries, which introduced a one-to-one correspondence between the clone and the hybridization signal (15) in a fixed position; in this way the clone could be uniquely identified and information about it accumulated. The subsequent explosion of array technologies has been sparked by two key innovations. The first is the use of non-porous solid support, such as glass, which has facilitated the miniaturization of the array and the development of fluorescence-hybridization detection ( 16, 17, 18). The second critical innovation has been the development of methods for high-density spatial synthesis of oligonucleotides, which allows the analysis of thousands of genes at the same time. Because DNA cannot bind directly to the glass, the surface is first treated with silane to covalently attach reactive amine, aldehyde, or epoxies groups that allow stable attachment of DNA, proteins, and other molecules. The nucleic acid microarrays use short oligonucleotides (15–25 nt), long oligonucleotides (50–120 nt), and PCR-amplified cDNAs (100–3,000 bp) as array elements. The short oligonucleotides are primarily used for the detection of single nucleotide polymorphisms (SNPs). Indeed the destabilization, caused by mispairing of only one mismatch, is maximized by the short oligonucleotide ( 18). The long nucleotides and the PCRamplified cDNAs produce strong signals and high specificity (19), unambiguous sample identification, and affordability. The cDNA elements are readily obtained from cDNA libraries. They are typically used when only a limited part of genome is known. With this technology cells or tissues are exposed to toxicants, drugs, and then gene expression is measured by collecting mRNA, converting mRNA to labeled cDNA, hybridizing it to the DNA array, staining it with an appropriate dye, and visualizing the hybridized genes using a fluorometer (16,19,20) (see Fig. 1). The raw data are analyzed using bioinformatics software and databases. The aim is to obtain meaningful biological information such as patterns of relative induction/repression levels of gene expression, participation in biochemical pathways, and (in the most favorable cases) “genetic signatures.” Several exhaustive reviews are available both on the practical aspects of DNA microarrays and the analysis of data (21,22,23,24).

3. HOW GOOD ARE THE DATA AND HOW SHOULD THEY BE ANALYZED? One of the early concerns about the application of DNA microarray to toxicology has been how to properly compare experiments that use a wide variety of commercial and proprietary platforms, protocols, and analyses methods. The Health and Environmental Sciences Institute (HESI) of the International Life Sciences Institute (ILSI) has

342

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

Fig. 1. Gene expression analyses by microarray. (A) One-color expression analysis uses a single fluorescent label and two arrays to generate expression profiles for two cell or tissue samples (test and reference samples). Activated and repressed genes are obtained by superimposing images obtained by the two arrays. (B) Two-color expression analysis uses two different fluorescent labels and a single array to generate expression profiles for the test and reference samples. Activated and repressed genes are obtained by superimposing images generated in different channels on a single array. In both cases, the monochrome images from the scanner are imported into software in which the images are pseudo-colored and merged. Data analysis includes quality control, normalization, filtering, statistical analysis, and then annotation to obtain information about the biological knowledge. A gene will be indicated as “activated” if the level of gene expression relative to the reference sample will be more than 1 (1 means no change) or “repressed” if it will be less than 1.

Chapter 20 / DNA MicroArray and Oncology Drug Development

343

coordinated an international study involving more than 30 pharmaceutical companies, governmental, and academic institutions, to evaluate the harmonization of gene expression data and analyses (25). In the ILSI Genomic Project, common pools of RNA were analyzed in more than 30 different laboratories using both similar and different technical platforms. A number of groups performed several inter-laboratory evaluation of the following toxicant-induced gene expression changes: clofibrate-induced gene expression changes in rat liver (26); rat hepatic gene expression changes induced by methapyrilene (27); the effects of nephrotoxicants (28); the effects of genotoxic chemicals (29); and the effects of hepatotoxicants (30) on gene expression. The experimental programs have shown that ( 25): (a) patterns of gene expression relating to biological pathways are robust enough to allow insight into mechanisms of toxicity; (b) gene expression data can provide meaningful information on the physical location of the toxicity; and (c) dose-dependent changes can be observed. To make full use of microarray data it is necessary that data of published microarray experiments be made available to other researchers for comparison purposes. To this end, the Minimum Information About a Microarray Experiment (MIAME) ( 31) guidelines have been developed at the European Bioinformatics Institute (EBI). This standard describes the minimum information required to ensure that microarray data can be easily interpreted, and that results derived from its analysis can be independently verified. Public repositories of microarray gene expression data have been developed to store the results of array experiments: ArrayExpress ( 32) in Europe, Gene Expression Omnibus (GEO) in the United States ( 33), and the Center for Information Biology Gene Expression Database (CIBEX) (34) in Japan. Many journals already require an accession number (indicating that a data set has been submitted to one of these public databases) prior to publication. Several initiatives aim to extend the scope of public databases of microarray data to incorporate toxicology and biological end-points. These toxicogenomic databases are being developed with the aim of creating a knowledge base that can be used to support genomic applications in hazard identification (35). Two international consortia are developing public toxicogenomic databases with extensive cross-links to existing biological information and annotation: Tox-MiamExpress is being developed at EBI, while the Chemical Effects in Biological Systems (CEBS) database ( 36) is being developed at NCT (see Table 1). Commercial toxicogenomic databases have also been developed by private companies. In general, commercial databases try to address the issue of data quality by a high degree of standardization of input data and also include additional information of clinical chemistry and histopathology. Commercial toxicogenomic databases include (see Table 1) GeneLogic’s (MD, USA) ToxExpressTM Module (37), Curagen’s (CT, USA) PredicR itve Toxicogenomics Screen (PTSTM ) system and Iconix’s (CA, USA) DrugMatrix database (38). One of the most pressing concerns in the field is that the lack of correct data format and comparability of microarray experiments will limit future integrative microarray research and the ability of performing meta analysis of microarray data available in the public domain (39,40).

344

Part III / Pharmacogenomics in Clinical Drug Development in Oncology Table 1 List of Cited Databases and Repository Services.

Acronym

Full Name

Web Site

ArrayExpress GEO CIBEX

ArrayExpress at EBI Gene Expression Omnibus Center for Information Biology Gene Expression Database Toxicogenomics MIAMExpress Chemical Effects in Biological Systems NCTR’s Center for ToxicoinformaticsArrayTrack

www.ebi.ac.uk/arrayexpress www.ncbi.nlm.nih.gov/geo cibex.nig.ac.jp

Gene Logic Inc.: ToxExpress Module CuraGen Corporation: Predictive Toxicogenomic Screen Iconix Biosciences, Inc.: GeneMatrix

www.genelogic.com

Tox-MIAMExpress CEBS ArrayTrack

ToxExpress PTS GeneMatrix

www.ebi.ac.uk/tox-miamexpress cebs.niehs.nih.gov www.fda.gov/nctr/science/centers/ toxicoinformatics/ArrayTrack

www.curagen.com/rnd/pts.asp www.iconixbiosciences.com

4. APPLICATIONS OF TOXICOGENOMIC TO ONCOLOGY DRUG DEVELOPMENT The number of new chemical entities (NCEs) approved by the U.S. FDA has dropped in the last decade (41) and the average success rate, from the first-in-human studies to registration, is only 11% (42). The lack of drug efficacy and safety account for around 30% of the failures in the clinic ( 42). Thus, the ability to determine drug safety and efficacy early in the discovery process should help in reducing the failure rate during the costly development studies, and in the end it would produce better and safer drugs (43). Increasing the knowledge about a compound to include molecular mechanism of toxicity, in addition to molecular mechanisms of action, should lead to improved pre-clinical and clinical lead selection and facilitate improved clinical trial designs with the ultimate outcome of obtaining more selective and less toxic drugs. A number of examples of the application of toxicogenomic toward drug discovery have been published. Publications in the toxicogenomic field have evolved from evaluating the potential of the technology ( 44, 45) to illustrating the practical use of gene expression profiling in toxicology (4,46). For example, Hamadeh and coworkers (4) analyzed the patterns of gene expression in liver tissue taken from rats exposed to different chemicals. Their analysis revealed similarities in gene expression profiles among animals treated with different chemicals belonging to the same class of compounds (peroxisome proliferators). On the contrary, animals treated with a different class of compounds (enzyme inducers) showed a very distinctive gene expression profile.

Chapter 20 / DNA MicroArray and Oncology Drug Development

345

In the United States, the National Center for Toxicogenomics (NCT) has conducted some proof-of-principle experiments to establish signature profiles of known toxicants, and to link the pattern of altered gene expression to specific parameters of conventional indices of toxicity (4,47,48). These studies have shown that it is possible to identify a signature of expressed gene patterns after exposure to a given toxicant (49). DNA microarrays have also been used to develop a much deeper insight into the mechanism of chemical toxicity at the molecular level. Andrew and colleagues at Dartmouth College ( 50) used cDNA microarrays to compare the effects of arsenic, nickel, chromium, and cadmium on the expression of 1,200 human genes in human bronchial BEAS-2B cells. Cells were exposed both to low doses of the different metals and also to a cytotoxic dose of sodium arsenite. Metal exposure modified only a small subset of the 1,200 genes, and each metal modified the expression of a largely unique set of genes; thus these results could provide the basis for the development of metal-specific biomarkers. Exposure to low or high concentrations of sodium arsenite resulted in quite different expression profiles. The researchers suggested that such change in gene expression profiles represents a switch from a survival-based biological response at the lower dose to a cell death-inducing apoptotic response at the higher dose. Gene expression profiling has been used to identify the expression profiles of two different chemotherapeutic drugs in breast cancer cell lines (51). The researchers cultured separate breast cancer cell lines that are known to have distinct responses to two chemotherapeutic drugs: doxorubicin (DOX), and 5-fluorouracil (5-FU). Different cell lines (two basal-like and two luminal epithelium) were treated with toxic concentrations of DOX and 5-FU and then mRNA was extracted and analyzed. Gene expression profiling identified which genes had been up- or down-regulated and showed a characteristic pattern of gene expression in response to DOX and 5-FU in each cell type. Detailed analyses identified a subset of 100 genes that could be used to differentiate between DOX and 5-FU treated samples. Testicular toxicology is of particular interest in drug discovery and development because testicular changes are usually very small in early stages and thus difficult to detect (52). A number of studies have demonstrated that gene expression profiling can elucidate the molecular basis of testicular toxicity and provide rapid methods to develop screens for backup compounds ( 53, 54). Several studies have demonstrated that it is possible, using in vitro systems, to identify small subsets of genes that could be used to screen compounds for toxicity. For example Morgan et al. treated HepG2 cells with various agents that cause oxidative stress and using DNA microarray identified a subset of seven genes that correlated with oxidative stress (55). Some preliminary studies also suggest that the combination of DNA microarray analysis in isolated primary human cells may, in some cases, identify potential safety issues in candidate drugs that are not evident from preclinical studies. For example Kier and colleagues (56) treated isolated primary human hepatocytes with three different thiazolidinedione compounds used for the treatment of type II non-insulin-dependent diabetes. One of the tested compounds, troglitazone, showed expression changes in a large number of genes that were not observed with the other two compounds. The drug containing troglitazone can cause hepatotoxicity in a small percentage of patients and it has been removed from the market (57).

346

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

A recent paper clearly highlighted the limitations of in vitro systems in modeling whole-organism responses, which should be considered when developing biomarkers of in vivo toxicity. Dere and colleagues ( 58) compared the temporal gene expression profiles of Hepa1c1c7 mouse hepatoma cells and of the mice liver after treatment with a dioxin. The analysis revealed that Hepa1c1c7 cells were able to model the induction of xenobiotic metabolism in vivo. On the other hand, responses associated with cell cycle progression and proliferation were unique to the in vitro system, while lipid metabolism and immune responses were not replicated effectively in the Hepa1c1c7 cells. One quite new way of employing gene expression profiling in drug development consists in a chemical genetic approach that link gene profile signatures to drug activity patterns and mechanisms of action. A recent paper by Lamb and colleagues ( 40) describes a “connectivity map” that uses gene expression signatures to link small molecules to each other and to disease. The researchers treated breast cancer cell lines with 164 different small molecules, representing a range of FDA-approved drugs and non-drug bioactive compounds. They exposed the cells for 6 hr with a 10 ␮M concentration of each compound, collected the cell RNA and obtained the mRNA expression profile of each compound in the library using Affimetrix oligonucleotide chips. The resulting database of expression profiles, the “connectivity map,” was formatted so that a query signature could be compared to it, and a score was obtained for the similarity to each signature in the database. After validating the database they used the connectivity map to obtain new information; they queried the database with expression signatures from patients and animal models to find small molecules that were already in the database that may mimic or suppress particular diseases. In subsequent studies, they identified geduin, an HSP90 inhibitor, as a new lead compound for the treatment of prostate cancer ( 59) and rapamycin as a potential useful drug for treatment of acute lymphocytic leukemia resistant to dexamethasone (60).

5. APPLICATIONS OF GENE EXPRESSION PROFILING TO DRUG RESISTANCE AND CANCER TREATMENT Drug resistance in cancer therapy is still a major obstacle in chemotherapy to improve the cancer’s outcome. Because at the moment there are no proven predictors of a patient’s response to chemotherapy, all cancer patients receive the same treatment. Resistance to chemotherapy can be observed either at the onset of treatment, when a patient fails to show clinical response, or later when the disease recurs despite an initial successful response. Drug resistance in cancer arises from a complex range of biochemical and molecular events, which at the end results in tumor cells escaping death. The global genomic approach can be used to attack the problem of anti-cancer drug resistance in different ways. The identification of key genes and metabolic pathways involved in molecular mechanisms of resistances can establish new drug target or new strategies to overcome drug resistance can be envisioned by developing compounds that could administered alongside traditional treatment to limit or modulate resistance. The

Chapter 20 / DNA MicroArray and Oncology Drug Development

347

most attractive way would be the discovering of markers for drug resistance, allowing then the possibility to verify, before starting the therapy, if the cancer patient is a responder or a not responder to the drug. Many immortalized cancer cell lines have been exposed to increasing concentrations of an anticancer drug of interest, to isolate the resistant clones. Then, the global expression profiling obtained from the resistant-clone cells are compared with the gene expression profiling generated with drug-sensitive cells. Amundson et al. (61), Scherf et al. (62), and the Japanese Foundation for Cancer’s Takao Yamori (6) established global gene expression patterns of 60 cell lines and 39 well-characterized human cancer cell lines, respectively; they then attempted to correlate these patterns with cell lines’ sensitivity to chemotherapeutic drugs. For example, analysis of 5-fluorouracil-gene expression profile in breast cancer cell lines has identified genes associated with drug resistance and response (63). The in vitro system has some limitations and it can be used only as a model to address acquired drug resistance, because cell lines may not accurately reflect the in vivo situation of patients treated with anticancer drugs. In vivo analysis using patient samples to study anti-cancer drug resistance are described in the published literature. Chang and coworkers ( 64) showed that the gene-expression patterns from mRNA derived from needle biopsies of breast cancer patients treated with the anticancer drug, docetaxel, could be predictive of response to the therapy. Moreover gene-expression analysis of leukemia cells from B-lineage acute lymphoblastic leukemia pediatric patients has identified sets of differentially expressed genes that are associated specifically with sensitivity or resistance to chemotherapy. Such information can also be used in diagnostic assay to stratify the patients in sub-categories, to develop new drugs and later on to personalize the anti-cancer therapy. Transcription profiling has been essential for better understanding of the timing and sequential molecular events of tumor transformation. Global transcription profiling has revealed the possibility to identify individual genes and profiles (65) that predict metastasis and poor clinical outcome as well as profiles that predict tissue-specific metastasis (66,67). These data showed that somatic events required for metastatic spread had occurred early in the development of the disease, and that they are present in the bulk of the cells in primary tumor biopsies, suggesting that it is possible to develop sensitive and specific prognostic markers, and markers that would predict response to the available therapies ( 64, 68). For example a microarray of 70 genes can predict the likelihood of distant metastasis within 5 years in breast cancer patients (68). Gene espression profiles have also the potential to provide physicians with clinically useful biomarkers that will improve patient care, thanks to increased individualization of diagnosis and treatment of oncology diseases ( 8). In a very recent article, Potti et al. ( 10) used in vitro drug sensitivity data coupled with microarray data to develop gene expression signatures that predict sensitivity to individual chemotherapeutic drugs. They have been able to show that many of those signatures can predict clinical response in individuals treated with those drugs. Gene expression signatures of an individual’s tumor that predict response to various cytotoxic chemotherapeutic agents will allow the optimization of those drugs, thus balancing the relative benefits with risk.

348

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

6. CONCLUSIONS The application of gene expression analysis to toxicology is now a mature science. The field has rapidly progressed from the proof-of-principle phase to actual applications, and gene expression profiling is already being used in screening for toxicity of lead compounds in drug discovery. There are already a few examples of this technology being used to identify safety issues of drugs that were not evident from pre-clinical studies. To make the best use of the technology one of the prerequisite is the availability of extensive databases of toxicogenomixcs data, and there are already several databases, both public and commercial, that incorporate gene expression data with toxicology and biological end-points. The availability of highly annotated databases in the public domain would be extremely important to realize the full potential of such technology, and the formation of international consortia to harmonize the work would be a very effective way to move the field forward. Gene expression profiling has been also shown to be an important tool in addressing the problem of drug resistance in cancer treatment and as a predictor of disease outcome. This opens up the possibility of better detection of tumor development, a more accurate diagnosis and prognosis, and, above all, a vision of personalized oncology treatment. All complications noted, the application of gene expression profiling to toxicology and disease analysis has clearly the potential for providing better, safer, and more effective treatments to oncology patients.

ACKNOWLEDGEMENTS The authors thank Una Cullinan for her help with the English revision of the manuscript. The authors have no competing financial interests.

REFERENCES 1. Nuwaysir EF, Bittner M, Trent J et al. Microarrays and toxicology: the advent of toxicogenomics. Mol Carcinog 1999;24:153–159. 2. Lovett RA. Toxicogenomics. Toxicologists brace for genomics revolution. Science 2000;289:536–537. 3. Burczynski ME, McMillian M, Ciervo J et al. Toxicogenomics-based discrimination of toxic mechanism in HepG2 human hepatoma cells. Toxicol Sci 2000;58:399–415. 4. Hamadeh HK, Bushel PR, Jayadev S et al. Gene expression analysis reveals chemical-specific profiles. Toxicol Sci 2002;67:219–231. 5. Khor TO, Ibrahim S, Kong AN. Toxicogenomics in drug discovery and drug development: potential applications and future challenges. Pharm Res 2006;23:1659–1664. 6. Dan S, Tsunoda T, Kitahara O et al. An integrated database of chemosensitivity to 55 anticancer drugs and gene expression profiles of 39 human cancer cell lines. Cancer Res 2002;62:1139–1147. 7. Holleman A, Cheok MH, den Boer ML et al. Gene–expression patterns in drug–resistant acute lymphoblastic leukemia cells and response to treatment. N Engl J Med 2004;351:533–542. 8. Koch WH. Technology platforms for pharmacogenomic diagnostic assays. Nat Rev Drug Discov 2004;3:749–761. 9. Bild A, Yao G, Chang J et al. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature 2006;439:353–357. 10. Potti A, Dressman HK, Bild A et al. Genomic signatures to guide the use of chemotherapeutics. Nat Med 2006;12:1294–1300.

Chapter 20 / DNA MicroArray and Oncology Drug Development

349

11. Lettieri T. Recent applications of DNA microarray technology to toxicology and ecotoxicology. Environ Health Perspect 2006;114:4–9. 12. Jain KK. Applications of AmpliChip CYP450. Mol Diagn 2005;9:119–127. 13. Southern EM. Detection of specific sequences among DNA fragments separated by gel electrophoresis. J Mol Biol 1975;98:503–517. 14. Southern EM Blotting at 25. Trends Biochem Sci 2000;25:585–588. 15. Grunstein M, Hogness DS. Colony hybridization: a method for the isolation of cloned DNAs that contain a specific gene. Proc Natl Acad Sci USA 1975;72:3961–3965. 16. Schena M, Shalon D, Davis RW et al. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995;270:467–470. 17. Schena M, Shalon D, Heller R et al. Parallel human genome analysis: microarray–based expression monitoring of 1000 genes. Proc Natl Acad Sci USA 1996;93:10614–10619. 18. Lockhart DJ, Dong H, Byrne MC et al. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 1996;14:1675–1680. 19. DeRisi J, Penland L, Brown PO et al. Use of a cDNA microarray to analyse gene expression patterns in human cancer. Nat Genet 1996;14:457–460. 20. Lashkari DA, DeRisi JL, McCusker JH et al. Yeast microarrays for genome-wide parallel genetic and gene expression analysis. Proc Natl Acad Sci USA 1997;94:13057–13062. 21. Schena, M. (1999). DNA Microarrays: A Practical Approach. The Practical Approach Series; 205, Oxford University Press; New York. 22. Schulze A, Downward J. Navigating gene expression using microarrays: a technology review. Nat Cell Biol 2001;3:E190–195. 23. Schena M (2003). Microarray Analysis, Wiley–Liss, Hoboken, NJ. 24. Knudsen S (2004). Guide to Analysis of DNA Microarray Data. 2nd ed. Wiley–Liss, Hoboken, N.J. 25. Pennie W, Pettit SD, Lord PG. Toxicogenomics in risk assessment: an overview of an HESI collaborative research program. Environ Health Perspect 2004;112:417–419. 26. Baker VA, Harries HM, Waring JF et al. Clofibrate–induced gene expression changes in rat liver: a cross-laboratory analysis using membrane cDNA arrays. Environ Health Perspect 2004;112:428–438. 27. Waring JF, Ulrich RG, Flint N et al. Interlaboratory evaluation of rat hepatic gene expression changes induced by methapyrilene. Environ Health Perspect 2004;112:439–448. 28. Kramer JA, Pettit SD, Amin RP et al. Overview on the application of transcription profiling using selected nephrotoxicants for toxicology assessment. Environ Health Perspect 2004;112:460–464. 29. Newton RK, Aardema M, Aubrecht J. The utility of DNA microarrays for characterizing genotoxicity. Environ Health Perspect 2004;112:420–422. 30. Ulric RG, Rockett JC, Gibson GG et al. Overview of an interlaboratory collaboration on evaluating the effects of model hepatotoxicants on hepatic gene expression. Environ Health Perspect 2004;112: 423–427. 31. Brazma A, Hingamp P, Quackenbush J et al. Minimum information about a microarray experiment (MIAME): toward standards for microarray data. Nat Genet 2001;29:365–371. 32. Brazma A, Parkinson HS, Sarkans U et al. ArrayExpress: a public repository for microarray gene expression data at the EBI. Nucleic Acids Res 2003;31:68–71. 33. Edgar R, Domrachev M, Lash AE. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res 2002;30:207–210. 34. Ikeo K, Ishi-i J, Tamura T et al. CIBEX: center for information biology gene expression database. C R Biol 2003;326:1079–1082. 35. Mattes WB, Pettit SD, Sansone SA et al. Database development in toxicogenomics: issues and efforts. Environ Health Perspect 2004;112:495–505. 36. Waters M, Boorman G, Bushel P et al. Systems toxicology and the Chemical Effects in Biological Systems (CEBS) knowledge base. EHP Toxicogenomics 2003;111:15–28. 37. Castle AL, Carver MP, Mendrick DL. Toxicogenomics: a new revolution in drug safety. Drug Discov Today 2002;7:728–736.

350

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

38. Ganter B, Tugendreich S, Pearson CI et al. Development of a large-scale chemogenomics database to improve drug candidate selection and to understand mechanisms of chemical toxicity and action. J Biotechnol 2005;119:219–244. 39. Larsson O, Sandberg R. Lack of correct data format and comparability limits future integrative microarray research. Nat Biotechnol 2006;24:1322–1323. 40. Lamb J, Crawford ED, Peck D et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science 2006;313:1929–1935. 41. Service RF. Surviving the blockbuster syndrome. Science 2004;303:1796–1799. 42. Kola I, Landis J. Can the pharmaceutical industry reduce attrition rates? Nat Rev Drug Discov 2004;3:711–715. 43. Ulrich R, Friend SH. Toxicogenomics and drug discovery: will new technologies help us produce better drugs? Nat Rev Drug Discov 2002;1:84–88. 44. Burchiel SW, Knall CM, Davis JW 2nd et al. Analysis of genetic and epigenetic mechanisms of toxicity: potential roles of toxicogenomics and proteomics in toxicology. Toxicol Sci 2001;59:193–195. 45. Fielden MR, Zacharewski TR. Challenges and limitations of gene expression profiling in mechanistic and predictive toxicology. Toxicol Sci 2001;60:6–10. 46. Waring JF, Ciurlionis R, Jolly RA et al. Microarray analysis of hepatotoxins in vitro reveals a correlation between gene expression profiles and mechanisms of toxicity. Toxicol Lett 2001;120:359–368. 47. Bartosiewicz MJ, Jenkins D, Penn S et al. Unique gene expression patterns in liver and kidney associated with exposure to chemical toxicants. J Pharmacol Exp Ther 2001;297:895–905. 48. Bulera SJ, Eddy SM, Ferguson E et al. RNA expression in the early characterization of hepatotoxicants in Wistar rats by high-density DNA microarrays. Hepatology 2001;33:1239–1258. 49. Tennant RW. The National Center for Toxicogenomics: using new technologies to inform mechanistic toxicology. Environ Health Perspect 2002;110:A8–10. 50. Andrew AS, Warren AJ, Barchowsky A et al. Genomic and proteomic profiling of responses to toxic metals in human lung cells. Environ Health Perspect 2003;111:825–835. 51. Troester MA, Hoadley KA, Parker JS et al. Prediction of toxicant-specific gene expression signatures after chemotherapeutic treatment of breast cell lines. Environ Health Perspect 2004;112:1607–1613. 52. Yang Y, Blomme EA, Waring JF. Toxicogenomics in drug discovery: from preclinical studies to clinical trials. Chem Biol Interact 2004;150:71–85. 53. Richburg JH, Johnson KJ, Schoenfeld HA et al. Defining the cellular and molecular mechanisms of toxicant action in the testis. Toxicol Lett 2002;135:167–183. 54. Adachi T, Koh KB, Tainaka H et al. Toxicogenomic difference between diethylstilbestrol and 17betaestradiol in mouse testicular gene expression by neonatal exposure. Mol Reprod Dev 2004;67:19–25. 55. Morgan KT, Ni H, Brown HR et al. Application of cDNA microarray technology to in vitro toxicology and the selection of genes for a real-time RT-PCR-based screen for oxidative stress in Hep-G2 cells. Toxicol Pathol 2002;30:435–451. 56. Kier LD, Neft R, Tang L et al. Applications of microarrays with toxicologically relevant genes (tox genes) for the evaluation of chemical toxicants in Sprague Dawley rats in vivo and human hepatocytes in vitro. Mutat Res 2004;549:101–113. 57. Tolman KG, Chandramouli J. Hepatotoxicity of the thiazolidinediones. Clin Liver Dis 2003;7: 369–379. 58. Dere E, Boverhof DR, Burgoon LD et al. In vivo–in vitro toxicogenomic comparison of TCDDelicited gene expression in Hepa1c1c7 mouse hepatoma cells and C57BL/6 hepatic tissue. BMC Genomics 2006;7:80. 59. Hieronymus H, Lamb J, Ross KN et al. Gene expression signature-based chemical genomic prediction identifies a novel class of HSP90 pathway modulators. Cancer Cell 2006;10:321–330. 60. Wei G, Twomey D, Lamb J et al. Gene expression–based chemical genomics identifies rapamycin as a modulator of MCL1 and glucocorticoid resistance. Cancer Cell 2006; 10:331–342. 61. Amundson SA, Myers TG, Scudiero D et al. An informatics approach identifying markers of chemosensitivity in human cancer cell lines. Cancer Res 2000;60:6101–6110.

Chapter 20 / DNA MicroArray and Oncology Drug Development

351

62. Scherf U, Ross DT, Waltham M et al. A gene expression database for the molecular pharmacology of cancer. Nat Genet 2000;24:236–244. 63. Boyer J, Maxwell PJ, Longley DB et al. 5-Fluorouracil: identification of novel downstream mediators of tumour response. Anticancer Res 2004;24:417–423. 64. Chang J, Wooten E, Tsimelzon A et al. Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 2003;362:362–369. 65. Ramaswamy S, Ross KN, Lander ES et al. A molecular signature of metastasis in primary solid tumors. 2003;33:49–54. 66. Hynes RO. Metastatic potential: generic predisposition of the primary tumor or rare, metastatic variants—or both? Cell 2003;113:821–823. 67. Kang Y, Siegel PM, Shu W et al. A multigenic program mediating breast cancer metastasis to bone. Cancer Cell 2003;3:537–549. 68. Van’t Veer L, Dai H, van de Vijver M et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002;415:530–536.

“This page left intentionally blank.”

21

Strategies to Identify Pharmacogenomic Biomarkers: Candidate Gene, Pathway-Based, and Genome-Wide Approaches Xifeng Wu, MD, PhD, Jian Gu, PhD, and Margaret R. Spitz, MD, MPH CONTENTS Introduction Candidate Gene Approach Pathway-Bas ed Approach Genome-Wide Scanning Approach Conclus ions References

S UMMARY Genetic inheritance plays a significant role in inter-individual variation of drug response (efficacy and toxicity). Pharmacogenetics and pharmacogenomics are the studies of using genetic variations to predict drug response. The study methodology can be grouped into three approaches: a candidate gene approach, a pathway-based analysis, and a genome-wide analysis. The candidate gene approach has produced some striking examples of pharmacogenetic markers, but it also has resulted in numerous inconsistent associations. The pathway-based approach, which assesses the combined effects of a panel of polymorphisms that interact in the same pathway and/or different pathways, may amplify the effects of individual polymorphisms and enhance From: Cancer Drug Discovery and Development: Genomics and Pharmacogenomics in Anticancer Drug Development and Clinical Response c Humana Press, Totowa, NJ Edited by: F. Innocenti, DOI: 10.1007/978-1-60327-088-5 21, 

353

354

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

the predictive power. Both the candidate gene approach and the pathway-based approach are hypothesis-driven, relying on a priori knowledge of gene and polymorphism functions. In contrast, the genome-wide scan approach is a hypothesis-generating and discovery-driven approach, and no prior knowledge of the disease or drug action mechanisms is required. In this chapter, we highlight the major accomplishments and challenges of these three approaches and discuss the pros and cons of each. Key Words: pharmacogenetics; pharmacogenomics; pharmacokinetics; pharmacodynamics; polymorphisms; candidate gene approach; pathway analysis; genomewide scan

1. INTRODUCTION It is well known that multiple patients receiving the same dosage of a drug exhibit different outcomes. Some recover successfully, some fail to respond, and others experience adverse drug responses (ADR) that may be fatal. The favorable response rate for most drugs is typically 95% of the low-activity phenotypes. The U.S. FDA has added TPMT genetic information to the package inserts of thiopurine drugs, which recommends TPMT testing (genotypic testing for TPMT*3A, *3C, and *2 and/or phenotypic testing of thiopurine nucleotides and TPMT activity in erythrocytes) if patients exhibit clinical or laboratory evidence of severe toxicity, particularly myelosuppression. 2.1.2. I RINOTECAN AND UGT1A1 Irinotecan is an analogue of camptothecin, a topoisomerase I inhibitor that is used for the treatment of advanced colorectal cancer and a few other solid tumors. Irinotecan is first converted to an active metabolite, SN-38, which is further conjugated into the inactive SN-38 glucuronide (SN-38 G), mainly by UGT1A1, and then eliminated via the bile. A common dinucleotide (TA) repeat polymorphism (containing 5, 6, 7, or 8 copies of TA repeat) has been identified in the UGT1A1 promoter TATA box. There is an inverse relationship between the number of TA repeats and the promoter activity (16). The (TA)6 TAA allele is the most common wild-type allele, and the (TA)7 TAA (UGT1A1*28) is the major variant allele in Caucasians. Several retrospective

356

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

and prospective studies have linked the UGT1A1*28 allele with irinotecan-related severe toxicity, especially neutropenia (17,18,19,20). In 2005, the U.S. FDA approved the inclusion of UGT1A1 genotyping information in the irinotecan package insert and recommended reducing the starting dose of irinotecan for homozygous UGT1A1*28 carriers. Shortly thereafter, in August 2005, the FDA approved the first pharmacogenetic test, the Invader UGT1A1 Assay by Third Wave Technologies for the UGT1A1*28 allele. Recently, Han et al. (21) found that another common variant, UGT1A1*6 (Gly71Arg), present almost exclusively in Asians, was associated with a higher incidence of severe neutropenia in Koreans. No homozygous UGT1A1*28 allele was found in this study, suggesting that UGT1A1*28 testing may not be adequate to predict severe toxicity in Asians. Both UGT1A1*6 and UGT1A1*28 testing are likely needed in Asians. The results of UGT1A1 variants with irinotecan treatment efficacy were inconsistent. Evaluation of genes affecting treatment efficacy is far more complex than toxicity. Although UGT1A1 apparently plays the most significant role in irinotecan metabolism, other UGT1A isozymes are also involved. For example, UGT1A7 and 1A9 have been reported to predict irinotecan toxicity ( 22). In addition, multiple pharmacodynamic genes contribute to drug efficacy. Somatic changes may play an equally important role in determining treatment efficacy. A more comprehensive examination of UGT1A1 and other UGT1A subfamily gene polymorphisms may give a more consistent and powerful prediction of irinotecan response.

2.1.3. P HARMACOGENETICS OF 5-F LUOROURACIL 5-Fluorouracil (5-FU) is the drug most commonly used to treat gastrointestinal cancers and several other solid cancers. Thymidylate synthase (TS) is the main target for 5-FU. In the 5 untranslated enhancer region of the TS gene, there is a 28-bp variable number of tandem repeats (VNTR) polymorphism (23). In Caucasians, 2R and 3R alleles are predominant. The 3R allele has been associated with higher TS protein expression than the 2R allele. Earlier pharmacogenetic studies evaluating this VNTR polymorphism suggested that the 3R allele conferred a low response rate (24,25), which led to the first genotype-guided clinical trial in North America based on this TS VNTR polymorphism ( 26, 27). However, some recent studies have provided somewhat conflicting clinical results in advanced colorectal cancer, with one null result and another showing a significantly higher response rate for patients with the 3R/3R genotype (28,29). These inconsistent results again highlight the problem facing pharmacogenetic studies using the candidate gene approach. For treatment efficacy, it is unlikely that any single polymorphism would have a dramatic effect. By and large, the functional impact of any individual polymorphism is negligible or modest, and there might be other polymorphisms that may enhance or counteract this specific polymorphism. This complex scenario is best illustrated by the TS polymorphisms. There is an additional G/C SNP within the 3R VNTR, and the 3 G allele has been associated with higher transcriptional and translational activity than the 3C allele ( 30). Ruzzo et al. ( 31) recently showed that the TS 5 UTR 3 G genotypes (2R/3 G, 3C/3 G, 3 G/3 G) predicted worse outcome, suggesting that a double assessment of VNTR and G/C SNP might be more relevant than the simple VNTR analysis.

Chapter 21 / Approaches in Pharmacogenetic Studies

357

The analysis of the predictive role of TS polymorphisms may be more complex if the 6-bp deletion/insertion (del/ins) polymorphism in the 3 UTR is added for consideration. The 6-bp deletion allele may cause mRNA instability and reduce intratumoral TS mRNA levels (32). The improved study design on the pharmacogenetic effect of TS polymorphisms should include a global evaluation of the combined VNTR, G/C, and 6-bp del/ins polymorphisms (33). The same global analysis applies to other genes, especially when the functional polymorphism is unclear. Haplotypes may be more informative than individual genotypes. Haplotype tag SNPs (htSNPs) based on the international HapMap project may represent the next wave of pharmacogenomic studies.

2.1.4. P HARMACOGENETICS OF C ISPLATIN Cisplatin is one of the most commonly used chemotherapy drugs, and its pharmacogenetics has attracted wide interest. Platinum agents form intra- and inter-strand DNA adducts that result in bulky distortion of DNA and inhibition of DNA replication. The level of platinum-DNA adducts in the circulation is correlated with clinical outcome and resistance to platinum agents has been linked to enhanced tolerance and repair of DNA damage. Therefore, the major focus of cisplatin pharmacogenetics has been on DNA repair genes (especially nucleotide excision repair genes) as well as glutathionetransferases (GSTs), which are the major detoxifying enzymes for platinum agents. There are some positive associations, but there are also a number of null or opposite results (26,31,34,35,36,37,38,39). The discrepancy and lack of major candidate markers in cisplatin pharmacogenetics are not surprising, because platinum agents do not target a specific protein, and none of the numerous polymorphisms in DNA repair genes have shown major functional impact experimentally. In addition, both DNA repair and GST systems are highly complex and are probably tightly regulated and balanced in vivo due to redundant or alternative mechanisms. It is unlikely that a single polymorphism in a single gene would have a strong effect on treatment response. For the expected low impact of individual polymorphisms, we need to move beyond the candidate gene approach and apply a pathway-based polygenic approach, in order to identify clinically relevant pharmacogenomic markers.

2.2. Statistical Consideration and Power Calculation Statistical analyses in pharmacogenetic studies using the candidate gene approach are fairly simple and straightforward. Several generic methods, such as logistic regression, Cox proportional hazards model, Kaplan–Meier estimates and log-rank test, are commonly used to determine the association between individual genetic polymorphism and treatment response, toxicity, and survival. The major concern for the candidate gene approach in pharmacogenetic study is whether it has sufficient power to detect the expected effect. Pharmacogenetic studies in the literature have generally been underpowered, that is, the sample size is not large enough, which is one of the main reasons for the lack of reproducibility. The sample size requirement for a pharmacogenetic study depends on the prevalence of risk factor (e.g., variant allele frequency), the event (toxicity, treatment response, survival, etc) rate, the magnitude of effect, the significance level, and the desired power (generally >80%). The power calculation for binary variables (toxicity or treatment response) is relatively

358

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

simple. For example, we used the statistical program PS (40) to calculate the minimum power for a sample size of 300 patients to detect a range of given odds ratios (ORs) associated with toxicity or tumor response (Fig. 1) with a two-sided alpha level of 5% where event (toxicity or response) rate ranges from 10% to 50% and variant allele frequency ranges from 5% to 50%. With this given sample size (300 patients), which is at the high end of sample sizes in the literature, ORs lower than 2.0 can only be detected with more than 80% power for situations where both event rate and/or prevalence of risk factor (i.e., variant allele frequency) are adequate (top five lines in Fig. 1). Power calculations for survival data are more complex due to the nature of the analyses as well as factors that are involved in the accrual of participants (i.e., follow-up time, prevalence of risk factor, etc.). The following example is based on the method discussed by Simon and Altman (41) using an 18-month overall survival rate of 40%, two-sided alpha level of 5%, and no attrition for varying levels of risk factor prevalence and hazard ratios. Figure 2 presents the minimum power to detect the given hazard ratios (1.1–3.0), where the variant allele frequency ranges from 1% to 50% for a sample size of 300 participants. With this given sample size, hazard ratios below 2.0 can only be detected with sufficient power (>80%) for those polymorphisms with higher variant allele frequencies (>15%).

2.3. Pros and Cons of Candidate Gene Approach Besides the examples listed above, there are numerous exploratory association studies that have identified many potential polymorphism biomarkers for treatment response in membrane transporter, drug-metabolizing enzyme, and drug target genes. The methodology and statistical analysis for the candidate gene approach are simple; the results are easy to interpret.

Fig. 1. An example of power calculation to detect the indicated odds ratio for a range of risk factor prevalence and event rate with a sample size of 300 patients.

Chapter 21 / Approaches in Pharmacogenetic Studies

359

Fig. 2. An example of power calculation to detect the indicated hazard ratio for a range of variant allele frequency with a sample size of 300 patients.

The limited number of candidate polymorphisms reduces the probability of type I errors (false positives). The sample size requirement is smaller and the cost is lower. However, the results from the candidate gene approach have been inconsistent and contradictory, due to the fact that the candidate gene approach was originally intended to identify a large, independent main effect for a single locus (polymorphisms) based on the assumption of a monogenic trait. The best scenario for the success of the candidate gene approach would be if, for a certain drug, there exists a major drug metabolism or major target gene, there is a polymorphism in that gene that confers significant changes to its function, and the frequency of this polymorphism is relatively high. This would result in an acceptable cost–benefit ratio for pharmacogenetic testing. Unfortunately, it is rare to find such a strong, monogenic candidate polymorphism for currently used chemotherapeutic drugs. Instead, most candidate gene studies have used the candidate gene approach to investigate polygenic traits. In these studies, the hypotheses were not particularly strong, the effects were not expected to be large, and there was no consideration of gene–gene interactions. For polygenic traits, the small associations between individual polymorphisms and outcomes are expected, because treatment response is usually a multigenic step, and because any single genetic polymorphism usually does not have a dramatic effect on outcome. Therefore, it is not surprising that the literature of pharmacogenetic studies using the candidate gene approach is littered with so many inconsistent results. This situation is unlikely to change, due to the small number of patients evaluated, patient and tumor heterogeneity, different treatment regimens and schedules, and failure to evaluate the effect of multiple pathophysiologically related genes. Large studies with homogeneous patient populations and uniform treatment regimens are needed to obtain consistent data of small effects. However, even if confirmed by large studies, these modest response-predicting polymorphisms are less likely to be clinically relevant single predictors. On the other hand, the real predictor polymorphism may be left out at first place due to the limited available information.

360

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

3. PATHWAY-BASED APPROACH A pathway based genotyping approach, which assesses the combined effects of a panel of polymorphisms that interact in the same pathway and/or different pathways, may amplify the effects of individual polymorphisms and enhance the predictive power. An illustrative example of the importance of taking a multigenic approach to a pharmacogenomic study is warfarin. Warfarin is an anticoagulant drug that targets vitamin K epoxide reductase complex 1 (VKORC1). The drug is metabolized mainly by CYP2C9. To avoid serious adverse reaction (bleeding), an appropriate warfarin maintenance dose is critical. VKORC1 genotypes account for about a quarter of the variance in the warfarin maintenance dose. CYP2C9 accounts for 6%–10%. However, genotyping both genes can account for >50% of the variability in the maintenance dose of warfarin ( 42, 43, 44). Therefore, analysis of the combination of VKORC1 and CYP2C9 genotypes should identify warfarin-sensitive patients who would require a lower maintenance dose of the drug. The same multigenic approach can apply to cancer pharmacogenomic research, which may be especially relevant to cancer therapies that do not have an obvious candidate polymorphism in metabolism genes and do not have a major protein target, such as cisplatin-based chemotherapy and radiation therapy as well as combinatorial therapy with multiple agents. We will use cisplatin-based chemotherapy as a prototype to illustrate the application of using a pathway-based approach in pharmacogenomic studies. As we described above, there have been many pharmacogenetic studies of cisplatinbased therapy. The GST family of enzymes and DNA repair genes are the most commonly investigated. Several potential candidate polymorphisms in the GSTP1, ERCC1, XPD, and XRCC1 genes have been reported, but with many contradictory reports for each significant association reported (26,31,34,35,36,37,38,39). None of the evaluated candidate polymorphisms in these genes have particularly strong evidence of significant functional impact. The associations were generally weak and not clinically usable as single predictors of treatment response, even if confirmed in larger prospective studies. In order to be relevant clinically, new statistical tools are needed to amplify the modest individual effects.

3.1. Statistical Consideration of Pathway-Based Approach 3.1.1. S UMMATION OF A DVERSE A LLELES Theoretically, a large number of unfavorable alleles, each contributing in a small yet important proportion of treatment response, should collectively enhance the predictive power to a level that may be clinically relevant. Several recent “proof-of-principle” studies have demonstrated the enhanced power of this pathway-polygenic approach. Stoehlmacher et al. ( 28) jointly analyzed the polymorphisms of four genes (XPD, ERCC1, GSTP1, TS) in colorectal cancer patients treated with 5-FU/oxaliplatin, and found that an increasing number of favorable alleles was associated, in a stepwise manner, with a significantly longer survival time. Quintela-Fandino et al. ( 45) recently determined the associations between four DNA repair gene SNPs (XPD-Asp312Asn, Lys751Gln, ERCC1-C8092A, and XRCC1Arg399Gln) and clinical outcomes in advanced squamous cell carcinoma of head and

Chapter 21 / Approaches in Pharmacogenetic Studies

361

neck (SCCHN) patients with cisplatin-based induction chemotherapy. In Cox’s multivariate analysis, each variant allele reduced the risk of death by 2.1-fold and patients with seven variant alleles exhibited a staggering 175-fold decrease in risk of death compared to those with all common alleles (p < 0.001). The probability of achieving a complete response increased 2.94-fold per additional variant alleles (p = 0.041). We have evaluated the role of 9 SNPs of eight nucleotide excision repair (NER) genes in esophageal cancer patients treated with 5-FU/platinum chemoradiation (34). NER is the major cellular system for repairing platinum-induced DNA adducts. We did not find significant individual associations with clinical outcomes; however, there was a significant trend for a decreasing risk of death with a decreasing number of adverse alleles (p for trend = .0008) (34). In the same study, we also tallied the number of adverse alleles from other cisplatinrelated pathway genes, including drug transporter, detoxifying and apoptotic pathways, and found that there were significant trends for increasing recurrence and decreasing survival with increasing number of adverse alleles. For overall survival, compared to individuals with ten or more adverse alleles, individuals with seven or fewer adverse alleles had an almost 60% reduced risk of death (HR = 0.41, 95% CI: 0.19–0.88).

3.1.2. S OPHISTICATED M ACHINE L EARNING AND A NALYSIS T OOLS In addition to simply summing the number of adverse alleles, other sophisticated machine learning and analysis tools, such as classification and regression tree (CART) and multifactor dimentionality reduction (MDR) algorithms, have been used to analyze the combination of multiple polymorphisms in drug-related pathway genes to account for gene–gene interactions and to provide a more powerful predictive capability. The traditional approach to modeling the relationship between discrete predictors (e.g., genotypes) and discrete clinical outcomes is logistic regression. Logistic regression is a parametric statistical approach for relating one or more independent or explanatory variables (genotype) to a dependent or outcome variable (phenotype) that follows a binomial distribution. However, because the number of possible interaction terms grows exponentially as the number of genotypes increases, logistic regression is limited in its ability to deal with multiple higher-order interactions. Several non-parametric statistical methods are available to model gene–gene interactions. CART has been used to identify homogenous subgroups and to model high-order interactions in cancer risk assessment (46,47). CART uses a binary recursive partitioning method to identify subgroups of patients with worse or better clinical outcomes. The method generates a tree-structured node with binary splits and identifies optimal cut points at each node for the covariate. The recursive procedure is continued to yield subsequent nodes that are more homogeneous (with respect to the response variable) than the original node. Gordon et al. (48) recently performed a pathway-based pharmacogenomic study on rectal cancer treated with chemoradiation in which they evaluated 21 polymorphisms in 18 genes involved in the critical pathways of cancer progression (drug metabolism, tumor microenvironment, cell cycle control, and DNA repair). They applied the CART analysis and found that a classification tree with four genes (IL-8, ICAM-1, TGF-␤, and

362

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

FGFR4), as well as the TNM classification, was predictive of tumor recurrence. A large prospective trial is being conducted to validate this preliminary finding. Multifactor dimensionality reduction (MDR) is a non-parametric statistical approach developed specifically to improve the power to detect gene–gene interactions (49,50,51). The MDR approach is nonparametric and is free of any assumed genetic model. The major advantage of MDR over traditional logistic regression modeling is that it greatly reduces the degree of freedom necessary for modeling. MDR defines a single variable that incorporates information from several loci and/or environmental factors that can be divided into high-risk and low-risk combinations. This new variable can be evaluated for its ability to classify and predict outcome status using cross-validation and permutation testing. The combination of cross-validation and permutation testing reduces the chances of making a type I error due to multiple testing. The MDR approach has been applied to a variety of settings to assess gene–gene interactions as well as gene–environment interactions ( 52, 53). Wilke et al. ( 54) provided an excellent review of the potential application of MDR analysis in combinatorial pharmacogenetics.

3.1.3. M ULTIPLE T ESTING The individual associations from the pathway-based approach should be interpreted with caution due to multiple testing. The significant level for individual polymorphisms needs to be adjusted for multiple comparisons. The traditional Bonferroni correction may be overly conservative. Westfall and Young ( 55) suggested that the permutation re-sampling approach be applied, which shuffles the phenotype values among the study subjects hundreds and thousands of times to create permutated datasets that have random phenotype–genotype associations. It could be used in a pathway-based genotyping study with a medium number of polymorphisms evaluated. Alternatively, the controlling of the false discovery rate (FDR), using the method of Benjamini and Hochberg (56), could also be considered.

3.2. Pros and Cons of the Pathway-Based Approach The pathway-based approach is an extension of the candidate gene approach. It is hypothesis driven and uses a priori knowledge of potentially functional SNPs and gene functions. It gives a more complete picture of the roles for genes and pathways in affecting treatment response. By evaluating the combined impact of multiple polymorphisms, it may be possible to identify minor associations that would not have been detected with the candidate gene approach. By using sophisticated machine-learning and analysis tools, it may be possible to identify high-order gene–gene interaction and provide a clinically relevant prediction value based on distinct genotype profiles. However, this approach is still limited by our current knowledge of the function of selected genes and polymorphisms. By tallying the number of adverse alleles, this approach assumes an equal weight for each allele, which may not be true for all genes and polymorphisms. It is arbitrary to assign which allele is the adverse allele because there may be no functional data to show that one specific allele is functionally inferior. The minor alleles may not necessarily result in reduced gene expression or protein function. Data analyses using machine-learning tools become more complex with increasing numbers of polymorphisms, and the analyses require larger sample sizes. Finally, the

Chapter 21 / Approaches in Pharmacogenetic Studies

363

validation of identified genotype combinations by association studies is more demanding, and their biological plausibility is difficult to assess experimentally.

4. GENOME-WIDE SCANNING APPROACH Both the candidate gene approach and the pathway-based approach rely on a priori knowledge of SNP and gene functions and biological plausibility. Although these hypothesis-driven association studies have given and will continue to provide us with very valuable information, continued efforts to exhaustively search and genotype all identified SNPs with potential functional significance in so many genes are costly and impractical. As comprehensive and inclusive as it can be, the pathway-based approach may leave out the real predictor polymorphisms. With the rapid advance in genotyping technology and cost reduction, coupled with the progress of the International Hapmap project, genome-wide scanning becomes possible in association studies. A genome-wide scan using a high-density SNP array has been used successfully to identify genes that contribute to disease susceptibility, which would not be possible with the hypothesis-driven candidate gene approach (57,58,59,60,61,62). Most of these studies used a family-based design in which the sample size requirement was considerably smaller. However, a genome-wide scan of unrelated subjects (e.g., a case-control study of cancer risk, and a case series outcome study) requires a substantially larger sample size and rigorous validation. Therefore, the study design generally follows a multi-stage design.

4.1. Study Design The initial goal of genome-wide scan association studies is to suggest a list of candidate genes or regions with a high probability of association with disease risk or treatment outcomes. Because there are hundreds of thousands of individual SNPs, the vast majority of the nominally significant results from the initial scan are false positives. Validation is the key for a genome-wide scan study. A multistage genotyping strategy, in which the initial high-density SNP scanning is followed by successive validation steps to eliminate false positives, has been the current standard for genome-wide scan association studies (63,64,65,66). There are some variations in the stage design and SNP selection strategy of each stage. Satagopan and Elston (64) and Wang et al. (65) proposed a two-stage design, which uses a stringent significant ␣ level to select SNPs for the second validation stage. The first major whole-genome scanning projects in cancer, the NCI’s Cancer Genetic Markers of Susceptibility (CGEMS) project, planned to identify genetic susceptibility markers for prostate and breast cancer using Illumina’s 550 K high-density SNP array. The CGEMS project study design includes the initial scanning step in 1200 cases and 1200 controls followed by four successive independent validation steps, each of which will genotype 2000 cases and 2000 controls (http://cgems.cancer.gov). SNPs that pass a lenient significance level (p < 0.05) in each stage will be carried to the next stage for validation. This design offers a balance between keeping true positives and eliminating false positives. SNPs that do not meet this generous threshold are unlikely to achieve significance throughout the whole sample and can therefore be safely discarded.

364

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

Although the vast majority of the statistically significant SNPs at this ␣ level will be false positives in the initial stages, most of these false positives will be eliminated in the following validation steps. The SNP selection for next step validation is primarily based on statistical significance rather than on potential functional significance. The reason is that Illumina’s High-density array SNPs are derived mainly from the htSNPs identified by the International HapMap Project. Most of the identified significant SNPs would not be causative loci, but through indirect association with true causative loci. Therefore, the selection of SNPs based on functional impact would miss most of the indirectly associated loci. However, for a small number of SNPs, a rational selection based on bioinformatics may be used to supplement these SNPs for validation. For example, an empirical Bayes approach–based hierarchical modeling (67) incorporating functional prediction information from gene location and computational algorithium (e.g., nsSNPs passing SIFT and PolyPhen screening, promoter SNPs affecting consensus sites, and splice site SNPs) and previous genotype–phenotype correlation data. Presently, it is premature to launch such a multistage study in pharmacogenomics by any single institution. For a pharmacogenomic study, the sample size is much smaller than for a case-control study. A high-density SNP array in pharmacogenomic studies would require huge international collaborations in terms of patient population, tumor characteristics, and treatment regimens. The more realistic, immediate genome-wide approach (with fewer patients) may be to first perform low-resolution genomic screening to reduce multiple testing and sample size requirements. For identified candidate regions, htSNP-based fine mapping may be used to pinpoint specific genetic loci and SNPs to predict response.

4.2. Statistical Considerations Analyzing data derived from a genome-wide scan is a huge statistical endeavor and is beyond the scope of this chapter. There are a few specific analytic issues, however, that are worth mentioning. First, departure from Hardy–Weinberg equilibrium (HWE) would raise concerns about the validity of the observed significant associations and needs to be dealt with. It could be the result of genotyping error, unrecognized population stratification (due to inbreeding or stratified mating), or merely due to chance. Markers showing significant departure (p < 0.01) in genotyping scans need to be flagged. Those markers showing departure in both stage 1 and 2 should be excluded for further analyses. Second, because the underlying model for predicting risk may follow dominant, recessive, additive, or other models, and no prior knowledge about the actual mode of action of potential prediction factors is available, a log-additive model that is fitted with the number of minor alleles as predictors through unconditional logistic regression may be used. The log-additive model can detect effects from both recessive and dominant models, and also any other models in between. Third, multiple loci are often in strong linkage disequilibrium. When one locus is found to be significant, all other tightly linked loci are likely to be significant. If significant effects are identified from two tightly linked loci, both of them may be retained for further study. If multiple tightly linked loci are identified in the same haplotype

Chapter 21 / Approaches in Pharmacogenetic Studies

365

block, the two loci showing the highest level of significance may be kept, along with any additional loci that remain significant in a multiple logistic regression that includes the first two most significant loci. Fourth, population stratification is a potential source of bias in studies of unrelated individuals. With the large number of genotypes that will be produced, population stratification will result in the inflation of the test statistics under the null hypothesis. Several approaches have been proposed to address the issue of population stratification, including genomic control ( 68, 69) and structured association ( 70). Devlin and Roeder ( 69) proposed a genomic control approach in coping with the anti-conservative test statistics resulting from population stratification. This approach rescales the test statistic by the test statistic value and is not affected by the inclusion of a small proportion of true positive findings. The corrected p-values at any given locus can then be obtained using an adjusted distribution that accounts for any inflation observed. The structured association approach differs from the genomic approach in that it estimates the population structure while genomic control assumes a particular parametric distribution of the value of the test statistic (70). Compared with the structured association, the genomic control approach is computationally simple and can be applied to both scanning and validation stages.

4.3. Pros and Cons of the Genome-Wide Scan Approach There are several distinct advantages of the genome-wide scanning approach. It is a hypothesis-generating and discovery-driven approach and no prior knowledge of the disease or drug action mechanisms is necessary. It can identify novel disease susceptibility genes or outcome-predicting genes and also confirm previously described genes. It gives a global assessment of each individual gene and may uncover the role of gene–gene interactions in complex genetic traits. The major limitations are the huge numbers of false positives that require successive validation steps to eliminate. The sample size requirement is enormous for studies with unrelated subjects, the cost could be prohibitively expensive, and the data management and statistical analyses are challenging.

5. CONCLUSIONS In the past half-century, pharmacogenetics and pharmacogenomics have rapidly incorporated advances in modern biology, genetics, and genomics. The field has evolved from focusing on monogenic traits to emphasizing polygenic pathways. These pathways encompass genes in all relevant drug-metabolizing enzymes, drug transporters, and drug targets, as well as signaling cascades that are upstream or downstream from the targets. Hypothesis-driven approaches have resulted in some striking examples of pharmacogenetics that have elicited strong interest and raised expectations for the clinical application of pharmacogenetics and personalized therapy. The SNP-based genome-wide scan, although a hot topic in disease association studies, has yet to be implemented in a pharmacogenomic study, but will certainly be applied in the near future with major collective efforts. However, other whole-genome platforms, such as array-CGH, gene expression profiling and proteomics, have produced potentially informative predictive profiles for drug response, underlining the importance of taking a global approach for

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

366

pharmacogenomic research. Each approach we described in this chapter has its pros and cons (Table 1) and we will see the application of each in pharmacogenetic studies for a long time. Despite the progress of exploratory research, the translation of pharmacogenomics from research to bedside has not been as rapid as we hoped. The major hurdle seems to be a lack of reproducibility and validation. The inconsistent results that currently exist in the literature are astonishing. These inconsistent results need to be confirmed by larger retrospective studies. Then, results from retrospective studies need to be verified in prospective studies. The path to individualized therapy is complex and will ultimately require a combination of genomic and proteomic analyses, which will incorporate tumor characteristics and somatic events (mutation, deletion and amplification) in order to provide a complete Table 1 Comparisons Between the Three Approaches in Pharmacogenomic Studies Pros

Cons

Candidate gene approach

Hypothesis-driven study and the results are easy to interpret; small sample size requirement; simple analysis; no multiple comparison issue; low cost; easy to validate

Pathway-based approach

Higher predictive power than candidate gene approach; taking consideration of gene-gene interactions in polygenic traits

Genome-wide scanning

Hypothesis-generating study which is not dependent on current knowledge and may identify completely novel markers; non-biased global assessment of genes and polymorphisms which may uncover gene–gene interactions

Limited by our current knowledge of gene and polymorphism function and drug pathways which may miss the true candidate genes and polymorphisms; low predictive power for polygenic traits; no consideration of gene–gene interaction Limited by our current knowledge; requiring larger sample size; assuming equal contribution for each gene and allele; the resulting gene–gene interactions are difficult to test biologically; validation is more difficult Huge number of false-positives; very large sample size requirement; sophisticated statistical analysis; high cost

Chapter 21 / Approaches in Pharmacogenetic Studies

367

picture of treatment response. For adverse reactions, germline genetic polymorphisms play a decisive role. However, for treatment efficacy, somatic aberrations may play a more significant role. Clinical validation of the polygenic models requires a large homogeneous patient population, a uniform treatment regimen, and sophisticated statistical analyses. It should be noted that the current drugs are designed to downplay the importance of individual variation in drug response, in contrast to the purpose of pharmacogenomics. Future pharmacogenomic studies should be incorporated into the early-phase clinical trials during the development of drugs. Such pharmacogenomic information may help to reduce the size and cost of phase III trials and improve the chance of developing effective novel targeted drugs. The prospect of personalized chemotherapy based on a pharmacogenomic testing in the next half-century is indeed bright.

REFERENCES 1. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients: a meta-analysis of prospective studies. JAMA 1998;279:1200–1205. 2. Weinshilboum R. Inheritance and drug response. N Engl J Med 2003; 348:529–537. 3. Evans WE, McLeod HL. Pharmacogenomics: drug disposition, drug targets, and side effects. N Engl J Med 2003;348:538–549. 4. Weinshilboum RM, Wang L. Pharmacogenetics and pharmacogenomics: development, science, and translation. Annu Rev Genomics Hum Genet 2006;7:223–245. 5. Eichelbaum M, Ingelman-Sundberg M, Evans WE. Pharmacogenomics and individualized drug therapy. Annu Rev Med 2006;57:119–137. 6. Wang Y. Gene expression–driven diagnostics and pharmacogenomics in cancer. Curr Opin Mol Ther 2005;7:246–250. 7. Abramovitz M, Leyland-Jones B. A systems approach to clinical oncology: focus on breast cancer. Proteome Sci 2006;4:5. 8. Smith L, Lind MJ, Welham KJ et al. Cancer Biology Proteomics Group. Cancer proteomics and its application to discovery of therapy response markers in human cancer. Cancer 2006;107:232–241. 9. Weinshilboum RM, Sladek SL. Mercaptopurine pharmacogenetics: monogenic inheritance of erythrocyte thiopurine methyltransferase activity. Am J Hum Genet 1980;32:651–662. 10. Lennard L, Van Loon JA, Weinshilboum RM. Pharmacogenetics of acute azathioprine toxicity: relationship to thiopurine methyltransferase genetic polymorphism. Clin Pharmacol Ther 1989;46: 149–154. 11. Lennard L, Van Loon JA, Lilleyman JS et al. Thiopurine pharmacogenetics in leukemia: correlation of erythrocyte thiopurine methyltransferase activity and 6-thioguanine nucleotide concentrations. Clin Pharmacol Ther 1987;41:18–25. 12. Lennard L, Lilleyman JS, Van Loon JA et al. Genetic variation in response to 6-mercaptopurine for childhood acute lymphoblastic leukemia. Lancet 1990;336:225–229. 13. Evans WE, Horner M, Chu YO et al. Altered mercaptopurine metabolism, toxic effects and dosage requirement in a thiopurine methyltransferase–deficient child with acute lymphoblastic leukemia. J Pediatr 1991;119:985–989. 14. Lennard L, Lewis IJ, Michelagnoli M et al. Thiopurine methyltransferase deficiency in childhood lymphoblastic leukaemia: 6-mercaptopurine dosage strategies. Med Pediatr Oncol 1997;29:252–255. 15. Relling MV, Hancock ML, Boyett JM et al. Prognostic Importance of 6-mercaptopurine dose intensity in acute lymphoblastic leukemia. Blood 1999;93:2817–2823. 16. Beutler E, Gelbart T, Demina A. Racial variability in the UDP-glucuronosyltransferase 1 (UGT1A1) promoter: a balanced polymorphism for regulation of bilirubin metabolism? Proc Natl Acad Sci USA 1998;95:8170–8174.

368

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

17. Ando Y, Saka H, Ando M, Sawa T et al. Polymorphisms of UDP-glucuronosyltransferase gene and irinotecan toxicity: a pharmacogenetic analysis. Cancer Res 2000;60:6921–6926. 18. Rouits E, Boisdron-Celle M, Dumont A et al. Relevance of different UGT1A1 polymorphisms in irinotecan-induced toxicity: a molecular and clinical study of 75 patients. Clin Cancer Res 2004;10:5151–5159. 19. Marcuello E, Altes A, Menoyo A et al. UGT1A1 gene variations and irinotecan treatment in patients with metastatic colorectal cancer. Br J Cancer 2004;91:678–682. 20. Innocenti F, Undevia SD, Iyer L et al. Genetic variants in the UDP-glucuronosyltransferase 1A1 gene predict the risk of severe neutropenia of irinotecan. J Clin Oncol 2004;22:1382–1388. 21. Han JY, Lim HS, Shin ES et al. Comprehensive analysis of UGT1A polymorphisms predictive for pharmacokinetics and treatment outcome in patients with non-small-cell lung cancer treated with irinotecan and cisplatin. J Clin Oncol 2006;24:2237–2244. 22. Carlini LE, Meropol NJ, Bever J et al. UGT1A7 and UGT1A9 polymorphisms predict response and toxicity in colorectal cancer patients treated with capecitabine/irinotecan. Clin Cancer Res 2005;11:1226–1236. 23. Marsh S. Thymidylate synthase pharmacogenetics. Invest New Drugs. 2005;23:533–537. 24. Pullarkat ST, Stoehlmacher J, Ghaderi V et al: Thymidylate synthase gene polymorphism determines response and toxicity of 5-FU chemotherapy. Pharmacogenomics J 2001;1:65–70. 25. Villafranca E, Okruzhnov Y, Dominguez MA et al: Polymorphisms of the repeated sequences in the enhancer region of the thymidylate synthase gene promoter may predict downstaging after preoperative chemoradiation in rectal cancer. J Clin Oncol 2001;19:1779–1786. 26. Marsh S, McLeod HL. Pharmacogenomics: from bedside to clinical practice. Hum Mol Genet 2006;15:R89–93. 27. McLeod HL, Tan B, Malyapa R et al. Genotype-guided neoadjuvant therapy for rectal cancer. Proc Am Soc Clin Oncol 2005;23:197. 28. Stoehlmacher J, Park DJ, Zhang W et al. A multivariate analysis of genomic polymorphisms: prediction of clinical outcome to 5-FU/oxaliplatin combination chemotherapy in refractory colorectal cancer. Br J Cancer 2004;91:344–354. 29. Jakobsen A, Nielsen JN, Gyldenkerne N et al. Thymidylate synthase and methylenetetrahydrofolate reductase gene polymorphism in normal tissue as predictors of fluorouracil sensitivity. J Clin Oncol 2005;23:1365–1369. 30. Kawakami K, Watanabe G. Identification and functional analysis of single nucleotide polymorphism in the tandem repeat sequence of thymidylate synthase gene. Cancer Res 2003;63:6004–6007. 31. Ruzzo A, Graziano F, Kawakami K et al. Pharmacogenetic profiling and clinical outcome of patients with advanced gastric cancer treated with palliative chemotherapy. J Clin Oncol 2006;24:1883–1891. 32. Mandola MV, Stoehlmacher J, Zhang W et al. A 6 bp polymorphism in the thymidylate synthase gene causes message instability and is associated with decreased intratumoral TS mRNA levels. Pharmacogenetics 2004;14:319–327. 33. Graziano F, Kawakami K: Studying the predictive/prognostic role of thymidylate synthase genotypes in patients with colorectal cancer: Is one polymorphism enough? J Clin Oncol 2005;23:7230–7231. 34. Wu X, Gu J, Wu TT et al. Genetic variations in radiation and chemotherapy drug action pathways predict clinical outcomes in esophageal cancer. J Clin Oncol 2006;24:3789–3798. 35. Park DJ, Stoehlmacher J, Zhang W et al. A Xeroderma pigmentosum group D gene polymorphism predicts clinical outcome to platinum-based chemotherapy in patients with advanced colorectal cancer. Cancer Res 2001;61:8654–8658. 36. Stoehlmacher J, Park DJ, Zhang W et al. Association between glutathione S-transferase P1, T1, and M1 genetic polymorphism and survival of patients with metastatic colorectal cancer. J Natl Cancer Inst 2002;94:936–942. 37. Gurubhagavatula S, Liu G, Park S et al. XPD and XRCC1 genetic polymorphisms are prognostic factors in advanced non-small-cell lung cancer patients treated with platinum chemotherapy. J Clin Oncol 2004;22:2594–2601.

Chapter 21 / Approaches in Pharmacogenetic Studies

369

38. Zhou W, Gurubhagavatula S, Liu G et al. Excision repair cross-complementation group 1 polymorphism predicts overall survival in advanced non-small-cell lung cancer patients treated with platinumbased chemotherapy. Clin Cancer Res 2004;10:4939–4943. 39. Goekkurt E, Hoehn S, Wolschke C et al. Polymorphisms of glutathione S-transferases (GST) and thymidylate synthase (TS): novel predictors for response and survival in gastric cancer patients. Br J Cancer 2006;94:281–286. 40. Dupont WD, Plummer WD. PS power and sample size program available for free on the Internet. Controlled Clin Trials 1997;18:274. 41. Simon R, Altman DG. Statistical aspects of prognostic factor studies in oncology. Br J Cancer 1994;69:979–985. 42. Rieder MJ, Reiner AP, Gage BF et al. Effect of VKORC1 haplotypes on transcriptional regulation and warfarin dose. N Engl J Med 2005;352:2285–2293. 43. Sconce EA, Khan TI, Wynne HA et al. The impact of CYP2C9 and VKORC1 genetic polymorphism and patient characteristics upon warfarin dose requirements: proposal for a new dosing regimen. Blood 2005;106:2329–2233. 44. Wadelius M, Chen LY, Downes K et al. Common VKORC1 and GGCX polymorphisms associated with warfarin dose. Pharmacogenomics J 2005;5:262–270. 45. Quintela-Fandino M, Hitt R, Medina PP et al. DNA-repair gene polymorphisms predict favorable clinical outcome among patients with advanced squamous cell carcinoma of the head and neck treated with cisplatin-based induction chemotherapy. J Clin Oncol 2006;24:4333–4339. 46. Garzotto M, Beer TM, Hudson RG et al. Improved detection of prostate cancer using classification and regression tree analysis. J Clin Oncol 2005;23:4322–4329. 47. Wu X, Gu J, Grossman HB et al. Bladder cancer predisposition: a multigenic approach to DNA-repair and cell–cycle–control genes. Am J Hum Genet 2006;78:464–479. 48. Gordon MA, Gil J, Lu B et al. Genomic profiling associated with recurrence in patients with rectal cancer treated with chemoradiation. Pharmacogenomics 2006;7:67–88. 49. Ritchie MD, Hahn LW, Roodi N et al. Multifactor-dimensionality reduction reveals high0order interactions among estrogen–metabolism genes in sporadic breast cancer. Am J Hum Genet 2001;69: 138–147. 50. Moore, J.H. Computational analysis of gene–gene interactions using multifactor dimensionality reduction. Expert. Rev Mol Diagn 2004; 4:795–803. 51. Moore JH, Boczko EM, Summar ML. Connecting the dots between genes, biochemistry, and disease susceptibility: systems biology modeling in human genetics. Mol Genet Metab 2005;84:104–111. 52. Andrew AS, Nelson HH, Kelsey KT et al. Concordance of multiple analytical approaches demonstrates a complex relationship between DNA repair gene SNPs, smoking and bladder cancer susceptibility. Carcinogenesis 2006;27:1030–1037. 53. Xu J, Lowey J, Wiklund F et al. The interaction of four genes in the inflammation pathway significantly predicts prostate cancer risk. Cancer Epidemiol Biomarkers Prev 2005;14:2563–2568. 54. Wilke RA, Reif DM, Moore JH. Combinatorial pharmacogenetics. Nat Rev Drug Discov 2005;4: 911–918. 55. Westfall PH, Young SS. Resampling-Based Multiple Testing: Examples and Methods for P-Value Adjustment. John Wiley & Sons, New York, 1993. 56. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Royal Stat Soc Ser B 1995; 57:289–300. 57. Klein RJ, Zeiss C, Chew EY et al. Complement factor H polymorphism in age-related macular degeneration. Science 2005;308:385–389. 58. Horvath A, Boikos S, Giatzakis C et al. A genome–wide scan identifies mutations in the gene encoding phosphodiesterase 11A4 (PDE11A) in individuals with adrenocortical hyperplasia. Nat Genet 2006;38:794–800. 59. Smyth DJ, Cooper JD, Bailey R et al. A genome-wide association study of nonsynonymous SNPs identifies a type 1 diabetes locus in the interferon-induced helicase (IFIH1) region. Nat Genet 2006;38:617–619.

370

Part III / Pharmacogenomics in Clinical Drug Development in Oncology

60. Sellick GS, Webb EL, Allinson R et al. A high-density SNP genomewide linkage scan for chronic lymphocytic leukemia-susceptibility loci. Am J Hum Genet 2005;77:420–429. 61. Chiang AP, Beck JS, Yen HJ et al. Homozygosity mapping with SNP arrays identifies TRIM32, an E3 ubiquitin ligase, as a Bardet–Biedl syndrome gene (BBS11). Proc Natl Acad Sci USA 2006;103: 6287–6292. 62. Maraganore DM, de Andrade M, Lesnick TG et al. High-resolution whole-genome association study of Parkinson disease. Am J Hum Genet 2005;77:685–693. 63. Thomas DC, Haile RW, Duggan D. Recent developments in genome-wide association scans: a workshop summary and review. Am J Hum Genet 2005;77:337–345. 64. Satagopan JM, Elston RC. Optimal two-stage genotyping in population-based association studies. Genetic Epidemiology 2003;25:149–157. 65. Wang H, Thomas DC, Pe’er I et al. Optimal two-stage genotyping designs for genome-wide association scans. Genet Epidemiol 2006;30:356–368 . 66. Hirschhorn JN, Daly MJ. Genome-wide association studies for common diseases and complex traits. Nature Reviews Genetics 2005;6:95–108. 67. Hung RJ, Brennan P, Malaveille C et al. Using hierarchical modeling in genetic association studies with multiple markers: application to a case-control study of bladder cancer. Cancer Epidemiol Biomarkers Prev 2004;13:1013–1021. 68. Devlin B, Roeder K. Genomic control for association studies. Biometrics 1999;55:997–1004. 69. Devlin B, Bacanu SA, Roeder K. Genomic control to the extreme. Nature Genetics 2004;36: 1129–1130. 70. Pritchard JK, Donnelly P. Case-control studies of association in structured or admixed populations. Theor Popul Biol 2001;60:227–237

Index Abelson oncogene (ABL1), in human leukemia, 127–144 See also BCR-ABL mutations Aberrations, tumor in gene copy numbers, SNP array for, 75–85 Acquired resistance to EGFR TKIs, 117–119 Acute lymphoblastic leukemia (ALL) treatment, 174–190 therapy adjustment, 174–175 thiopurines in, 173–190 See also under Thiopurines Affymetrix arrays, 11 Alemtuzumab, 204, 210–211 for CLL, 223 ALL, see Acute lymphoblastic leukemia (ALL) Allelic imbalance in tumor cell, 76–85, 98 Amino acid sequencing by LC-MS/MS, 37–38 Amplification, for microarray-based profiling, 7–10 exponential amplification, 9–10 isothermal amplification strategies, 8–9 rolling circle amplification (RCA) technology, 8–9 signal amplification post-hybridization, 9 T7-type amplification, 8 Antibody-dependent cellular cytotoxicity (ADCC), of Rituximab action, 207 Anticancer agents evaluation cell-based models in, 19–29 See also under Cell-based models Anticancer agents study, challenges, 21 Antimicrotubule drugs, 238–239 Antisense oligonucleotide, 47–53 Apoptosis pathway, of Rituximab action, 206–207 Array technology, 340–341 See also Microarray Association studies using candidate genes, 26–27 Auto-antibodies in cancer patients, proteomic analysis, 36, 38, 42 BCR-ABL mutations and imatinib resistance in chronic myeloid leukemia patients, 127–144 clinical aspects, 131–133 Dasatinib for, 127 directly impairing imatinib binding, 138 in oncogenic signaling pathways, 131 in P-loop, 137–139 translocation of t(9;22)(q34;q11), 129 within the activation loop, 138 within the catalytic domain, 138 Biomarkers biomarkers signatures microarrays identifying, 4 classifiers, see Pharmacogenomic biomarker classifiers

BRCA1 as potential predictive marker for cisplatin, 238–239 for docetaxel, 238–239 in transcriptional regulation of spindle checkpoint genes, 239–242 Breakpoint cluster region (BCR), in human leukemia, 127–144 See also BCR-ABL mutations Breast cancer patients, 5 challenges in, 288–289 microarray profiling in, 287–295 traditional prognostic factors for, 289 c-ABL1 protein, 137 Campath-1H, see Alemtuzumab Candidate gene approaches in pharmacogenomic biomarkers identification, 355–359 6-mercaptopurine, 355 irinotecan, 355–356 pros and cons of, 358–359 TPMT, 355 UGT1A1, 355–356 Capecitabine, 251 for metastatic CRC treatment, 155 Cell-based models, in genetic variants identification of chemotherapeutic toxicities, 19–29 CEPH cell lines, 20 integrating different approaches, 28–29 International HapMap cell lines, 20 interpatient variability in response to drugs and toxicity, 21 LCLs in, 21–22, See also Lymphoblastoid cell lines limitations, 20 relevance to drug toxicities in humans, 29 Centre d’ Etude du Polymorphisme Humain (CEPH) pedigrees in chemotherapy-induced toxicity evaluation, 20 Docetaxel concentration–viability curves, 24–25 Cetuximab, 320 Chemosensitivity, 58–70 Chemotherapy-induced toxicities, genetic variants in, 19–29 Childhood acute lymphoblastic leukemia treatment, 174–176 thiopurines in, 173–190 See also under Thiopurines Chromosome amplification/loss, 94 Chromosome instability (CIN), 95

371

372 Chronic lymphocytic leukemia (CLL), 220, 222–223 Alemtuzumab, 223 Rituximab, 223 Rituximab induced neutropenia, 223–224 Chronic myeloid leukemia patients BCR-ABL mutations in, 127–144 imatinib resistance in, 127–144 Cisplatin, 238–239 BRCA1 as potential predictive marker, 238–239 pharmacogenetics of, 357 Classification and regression tree (CART), 361–362 CLL, see Chronic lymphocytic leukemia (CLL) Clofibrate-induced gene expression, 343 CNAG software, 83 Cohesiveness, pathway, in tumor versus normal tissues, 66–67 Colorectal cancer (CRC) 5-fluorouracil (5-FU) for, 151–152 irinotecan (CPT-11), 151–152 thymidylate synthase gene variations in, 151–167 See also Metastatic CRC Comparative genomic hybridization (CGH), 289 Complement-dependent cytotoxicity (CDC), of Rituximab action, 207–210 Compound screen, 58–70 See also National Cancer Institute’s 60 (NCI60) Concordance between tumor and germline DNA, 91–99 See also Germline DNA Continuous administration of 5-FU (CIFU), for metastatic CRC treatment, 155 Copy number variation, SNP array for, 76–85 CRC, see Colorectal cancer (CRC) CYP2C19, 314–323 Cytochrome P450 (CYP) 1B1, 51 Cytotoxicity, 58 Dasatinib, 127 chemical formula, 141 for imatinid-resistant and intolerant CML, 127–144 DASL (cDNAmediated annealing, selection, extension, and ligation) technique, 11–12 dChip software, 83 Diagonal linear discriminant analysis, 330 Diffuse large B-cell lymphoma (DLBCL), 218–220 Dihydopyrimidine dehydrogenase (DPD) polymorphism, 153, 161–162, 164–166, 249–258 in Fluoropyrimidine drugs disposition, 250–252 impairment, 253–255 status, determining, 255–257 genotypic approaches, 256 phenotypic approaches, 256–257 tumoral expression and treatment efficacy, 253 DNA damaging chemotherapy, 238–239 DNA microarrays for transcription profiling of breast cancer, 289–291 DNA repair genes as predictors, 231–243 in NSCLC chemotherapy response, 231–243

Index Docetaxel, 238–239 BRCA1 as potential predictive marker, 238–239 DPD, see Dihydopyrimidine dehydrogenase (DPD) polymorphism Drug development pharmacogenomics in, 313–323 biomarker classifiers identification, 327–337 efficacy predictors, 319–322 genome-wide markers, 321–322 germline pharmacogenetics, 316–319 market segmentation, 322 pipelines, 315–316 regulatory environment, 322 timelines, 315–316 Drug mechanism of action (MOA) probe GI50 SOM clades and, 63–65 mapping pathways, 63–65 of similarly clustered drug molecules, 64 SOM region annotation, 64 by pathway gene expressions, 63–65 EBV-transformed LCLs generation, 22 Epidermal growth factor receptors (EGFR) alternative therapeutic strategies to target, 119–121 biochemistry, 105–108 domain topography, 105 EGFR mutants biochemical properties, 112–116 hyperactivation, structural insights, 113 sensitivity to selective TKIs, molecular basis for, 113–114 sensitivity to selective TKIs, oncogenic shock model, 114–116 signaling properties, 112–116 interacting proteins, 106–108 irreversible inhibitors of, 119–120 pan ErbB Inhibitors, 120 ligand-induced activation of EGFR, 106 mutations and sensitivity to TKIs in human lung cancer, 103–121 acquired resistance to EGFR TKIs, 117–119 dysregulation of EGFR, 108–109 primary or de novo resistance to EGFR TKIs, 116–117 somatic mutations, 110–111 T790M mutation, 118 oncogenic EGFR mutations, 108–112 selective EGFR TKIs and sensitizing EGFR mutations, 109–112 signaling properties of, 105–108 Erlotinib, in setting EGFR-driven cancers, 104–121 Esophageal cancer, proteomic analysis, 38, 40–41 Ethnic differences, haplotype structures covering, 271–277 Excision repair cross-complementing 1 (ERCC1) in NER process, 233 Exponential amplification, 9–10 Expression studies, 27–28

Index Familial genetics, 20–29 Family based genetic association studies, 26–27 Fc receptor gene polymorphism, 205, 209 clinical data and, 213–224 Chronic lymphocytic leukemia (CLL), 220, 222–223 diffuse large B-cell lymphoma (DLBCL), 218–220 Follicular Lymphoma, 213–218 Waldenstrom’s macroglobulinemia (WM), 219–222 Follicular Lymphoma (FL) chemotherapy alone, 217 idiotype vaccination treatment, 218 interleukin-2 plus rituximab therapy, 218 rituximab monotherapy, 214–217 sequential chop therapy and rituximab, 217 in vitro data and, 211–213 FCGR2A, 212 FCGR2B, 213 FCGR3A and FCGR2A, linkage between, 213 FCGR3A, 212–213 FCGR3B, 213 Fine needle aspirates (FNAs), 7 Fisher linear discriminant analysis, 330 Fluorodeoxyuridine monophosphate (FdUMP), 153 Fluorodeoxyuridine triphosphate (FdUTP), 153 Fluoropyrimidines, 249–258 See also Dihydropyrimidine dehydrogenase 5-Fluorouracil (5-FU) anabolic pathways, 251 catabolic pathways, 251 in CRC chemotherapy, 151–167 gene expression profile, 347 molecular biology of, 153–154 pharmacogenetics, 356–357 Fluorouridine triphosphate (FUTP), 153 Folate-pathway genes in pediatric ALL treatment response, 299–309 methylenetetrahydrofolate dehdrogenase (MTHFD1), 304 methylenetetrahydrofolate reductase (MTHFR), 302–304 reduced folate carrier (RFC) gene, 307 thymidylate synthase (TS) gene, 304–306 See also Methotrexate Follicular Lymphoma (FL) Fc receptor gene polymorphism, 213–218, See also under Fc receptor gene polymorphism SNP influence, 215–216 Formalin fixation and paraffin embedding (FFPE), 10–12, 80–85 Ftorafur (FT/1-tetrahydrofuranyl-5-fluorouracil), 156 Gefitinib, 320 in setting EGFR-driven cancers, 104–121 Gene amplification, 94 Topoisomerase I (TOP1), 94

373 Gene expression analyses/profiling, 3–15, 340–341 in drug resistance and cancer treatment, 346–347 by microarray, 342 Genetic variants, in chemotherapy-induced toxicities, 19–29 Genome, cancer, alterations, 93–95 chromosome amplification/loss, 94 chromosome instability (CIN), 95 gene amplification, 94 genome scan, 20–29 microsatellite instability (MSI), 95 polymorphism versus mutation, 93–94 Genome-wide approaches in drug development, 321–322 in pharmacogenomic biomarkers identification, 363–365 analytic issue, 364 Germline DNA and tumor, concordance between, 91–99 genotype concordance, 95–99 allelic imbalance, 98 genotype discordance, 98 primary tumor versus metastasis, 98–99 laser-assisted microdissection, 93 microdissection, 93 tumor genome alteration, 92 See also Genome, cancer, alterations Germline pharmacogenetics in oncology drug development, 316–319 Haplotype structures covering ethnic differences, 271–277 covering UGT1A Gene, 271–277 HapMap cell lines, 20 genomic information on, 23–27 heritability studies, 23–26 inference studies, 27 linkage analysis, 26 whole genome association, 26–27 Hematologic malignancies monoclonal antibody therapy, See also individual entry polymorphisms impact on, 203–225 Hematopoietic stem cell transplantation (HSCT), 132 Hemosensitivity prediction, breast cancer patients, 288–295 Hepatocellular carcinoma (HCC) proteomic analysis, 38–40 Heritability studies, HapMap cell lines in, 23–26 Hidden Markov model-based method, 77 Horseradish peroxidase (HRP), 9 Ibritumomab Tiuxetan, 204, 211 Idiotype vaccination treatment, follicular lymphoma (FL), 218 Imatinib mesylate (IM)/Imatinib resistance, in CML patients, 133–140 definition, 134 mechanism, 135–136

374 overcoming imatinib resistance, 140–144 combination therapy, 141 dasatinib in, 141 nilotinib in, 141 T-cell protein tyrosine phosphatase down-regulation, 136 In-gel digestion, in proteomic analysis, 37 Interferon- (IFN-) treatment, for CML, 132 Interleukin-2 (IL-2) plus rituximab therapy, 218 Irinotecan/Irinotecan treatment, 355–356 for metastatic CRC treatment, 154 regulatory status, 282–283 UDP-glucuronosyltransferase 1A haplotypes impact on, 267–283 See also Uridine diphosphate glucuronosyltransferase 1As Isobaric tag for relative and absolute quantitation (iTRAQ), 35–36 Isoelectric focusing (IEF), in two-dimensional gel electrophoresis, 37 Isothermal amplification strategies, 8–9 Isotope-coded affinity tag (ICAT) labeling, 35 Kinase inhibitors in human lung cancer EGFR mutations and sensitivity to, 103–121 See also Epidermal growth factor receptors Labeling technologies in microarray experiments, 10 Lapatinib, 316–319 Laser capture microdissection (LCM), 3–15 LCLs, see Lymphoblastoid cell lines (LCLs) Let-7 expression, 51 Leucovorin, for metastatic CRC treatment, 154–156 Ligand-induced activation of EGFR, 106 Linkage analysis in anticancer agents study, 20–29 advantages, 28 Liquid chromatography–tandem mass spectrometry (LC-MS/MS) amino acid sequencing by, 37–38 Loss of heterozygosity (LOH) patterns detection, 76–77, 82–83, 160 whole-genome LOH patterns, 79 Lymphoblastoid cell lines (LCLs), 21–29 in anticancer agents pharmacogenomics evaluation, 21–22 LCLs in, 21–22 for pharmacogenomic discovery, 22–29 anticancer agent response, measures of, 23 EBV-transformed LCLs generation, 22 expression studies, 27–28 HapMap cell lines, genomic information on, 23–27 LCLs from targeted populations, 22–23 Mapping pathways, to drug mechanism of actions (MOA), 63–65

Index Mass spectrometry (MS), in proteomic analysis, 37–38 amino acid sequencing, 37–38 in-gel digestion, 37 6-Mercaptopurine (6-MP), 355 in childhood ALL treatment, 173, 176–190 metabolism of, 178–180 Metastasis, primary tumor versus, 98–99 Metastatic CRC 5-FU analogs and prodrugs, 155–156 5-FU and leucovorin combination, 154–156 administration schedules, 155 Capecitabine, 155 continuous administration of 5-FU (CIFU), 155 current standard of care for, 154–156 ftorafur (FT/1-tetrahydrofuranyl-5-fluorouracil), 156 Irinotecan (CPT-11), 154 oxaliplatin, 151 Methapyrilene, 343 Methotrexate, action mechanism, 300–302 Methylenetetrahydrofolate dehdrogenase (MTHFD1), in pediatric ALL treatment response, 304 Methylenetetrahydrofolate reductase (MTHFR) polymorphism, 153, 161–167 in pediatric ALL treatment response, 302–304 confounding issues, 306 Microarray technology/profiling in breast cancer patients, 3–15, 287–295 Affymetrix arrays/chips, 11, 289 comparative genomic hybridization (CGH), 289 DASL technique, 11–12 diagnostics, 12–14 using informatics, 14 using peripheral blood of patients, 14 DNA microarray, 288 population-based molecular signatures, 290 for transcription profiling, 289–291 gene expression analyses by, 342 one-color expression analysis, 342 two-color expression analysis, 342 genomic-based microarray technologies, 289 heterogeneity hurdle, 5, See also Tumor heterogeneity historical perspectives, 4 in identifying biomarkers signatures, 4 labeling technologies in, 10 microdissection of tumor, 5–6 molecular profiling, 292–293 prognostics, 12–14 randon committee algorithm, 292 reporting guidelines, 294–295 in small samples profiling, 7–10, See also Amplification stem cell theory of cancer, 6–7 tumor tissue microarray, 288, 291–292 X3P arrays, 11 Microdissection of tumor laser capture microdissection (LCM), 5–6 microarray technology in, 5–6

Index MicroRNAs (miRNAs) in cancer, 47–53 biogenesis of, 48–49 cytochrome P450 (CYP) 1B1, 51 diagnosis and drug discovery, 49–52 generation, 48 let-7 expression, 51 miR-155, 50 miR-15a, 49–50 miR-16, 49–50 miR-17–92 cluster, 50 miR-21, 50–51 Microsatellite instability (MSI), 95 Mitotic checkpoint genes as predictors, 231–243 in NSCLC chemotherapy response, 231–243 BRCA1 involved in, 236–237 in cancer cell lines, 241 in chemotherapy sensitivity, 241 customized chemotherapy trials, 242–243 phases in, 240 Molecular profiling for cancer therapeutics, 292–293 machine learning model system, 292 NCI-60 panel, 292 reverse-phase protein lysate microarrays, 292 sulphorhodamine B assay, 292 Monoclonal antibody therapy, 210–211 Alemtuzumab, 210–211 Fc receptors in, 205 Ibritumomab Tiuxetan, 211 polymorphisms impact on against hematologic malignancies, 203–225 resistance to, 204 factors predisposing to, 204–205 patient-related factors, 204–205 tumor-related factors, 204 Tositumomab, 211 See also Rituximab Multifactor dimentionality reduction (MDR) algorithms, 361–362 Multiple training-test, 333

National Cancer Institute’s 60 (NCI60)-tumor cell panel gene expression, 59–60 pathway analysis, 61–62, See also Pathway analysis pharmacogenomics, 57–70 screening, 59–60 self-organizing map (SOM), 60 Needle/core biopsies, 7 NER, see Nucleotide excision repair (NER) Nilotinib, 128 chemical formula of, 141 for imatinid-resistant and intolerant CML, 127–144 Non-coding RNA, 47–53 Non-small cell lung cancer (NSCLC), 231 chemotherapy response in DNA repair and mitotic checkpoint genes as predictors, 231–243 kinase inhibitors in

375 EGFR mutations and sensitivity to, 104–121, See also Epidermal growth factor receptors Nucleotide excision repair (NER) pathway, 232 in NSCLC, 232–238 clinical findings, 235–238 global genome NER (GG-NER), 233 transcription-coupled NER (TC-NER), 233 Oncogenic shock model, for sensitivity of EGFR mutants to TKIs, 114–116 Oncology drug development, toxicogenomics application to, 339–348 See also Toxicogenomics Orotate phosphoribosyltransferase (OPRT), 153 Oxaliplatin, for metastatic CRC treatment, 151 Pancreatic cancer, proteomic analysis, 38, 41–42 Paraffin embedded material for SNP array analysis, 80–81 Pathway analysis/Pathway gene expression of gene expression in NCI’s 60 cell lines, 61–62 co-expression, 61 cohesiveness, 61–62, 65 co-regulation level, evaluation, 62 drug mechanism of action (MOA) probe, 63–65 generically unstable pathway, 67 in environmental information processing, 62 neighboring genes, 61 pathway cohesiveness in tumor versus normal tissues, 66–67 specifically unstable pathway, 67 stability of pathway, 67–68 targeting cancer pathways, 65–70 pathway gene expression coherence or cohesiveness (PGEC), 66–67 in pharmacogenomic biomarkers identification, 360–363 adverse alleles summation, 360–361 pros and cons, 362–363 sophisticated machine learning and analysis tools, 361–362 statistical consideration, 360–362 Pediatric ALL treatment response, folate-pathway and the thymidylate synthase genes in, 299–309 Pharmacogenomic biomarker classifiers, 328 empirical classifiers development using gene expression, 329–331 feature selection, 329–330 mathematical function specification, 329–330 parameter estimation, 330 identification strategies, 353–367 5-fluorouracil, 356–357 candidate gene approaches, 353–367 cisplatin, 357 genome-wide approaches, 353–367 pathway-based approaches, 353–367

376 power calculation, 357–358 statistical consideration, 358–358 identification, in cancer drug development, 327–337 developmental and validation studies, 331–332 in new drug development, 334–337 predictive accuracy estimation in developmental studies, 333–334 cross-validation, 334 multiple training-test, 333 re-substitution estimate, 334 split-sample method, 333 Pipelines, pharmaceutical, 315–316 PLASQ (probe-level allele-specific quantitation) software, 83 Polymorphisms impact on monoclonal antibody therapy against hematologic malignancies, 203–225 versus mutation, 93–94 See also Fc receptor gene polymorphism; Rituximab Population based genetic association studies, 26–27 of breast cancer using gene expression profiles, 290 Predictive accuracy estimation in biomarker classifiers, 333–334 Primary or de novo resistance to EGFR TKIs, 116–117 Primary tumor versus metastasis, 98–99 Proteome/Proteomic analysis, 33–43 of auto-antibodies, 36 in breast cancer patients, 288, 291–292 definition, 34 mass spectrometry (MS), 37–38 oncology application of, 42–43 serum samples, 35 tissue samples, 34–35 two-dimensional gel electrophoresis (2-DE), 37 Pure cell populations, 3–15 Randon Committee algorithm, 292 Reduced folate carrier (RFC) gene in pediatric ALL treatment response, 307 REMARK (REporting recommendations for tumor MARKer) system, 294–295 Rituximab, 203–225 action mechanism, 205–210 antibody-dependent cellular cytotoxicity (ADCC) pathway, 207 apoptosis pathway, 206–207 complement-dependent cytotoxicity (CDC) pathway, 207–210 for CLL, 223 See also under Follicular Lymphoma (FL) RNA amplification, 3–15 Rolling circle amplification (RCA) technology, 8–9 Selective EGFR TKIs and the of sensitizing EGFR mutations, 109–112 Self-organizing map (SOM) algorithm, 60, 64 Signal amplification post-hybridization, 9

Index Single gene marks pharmacogenomic predictor measurement, 319–321 Single-nucleotide polymorphism (SNP) array challenges of SNP array analysis, 80–81 in CRC patients, 159–160 paraffin embedded material for array analysis, 80–81 for tumor aberrations in gene copy numbers, 75–85 advantages, 78–80 analysis, 77–78 hidden Markov model-based method, 77 software to visualize, 81–83 validation of array data, 83–84 whole genome-amplified DNA for array analysis, 80 Sodium dodecyl sulfate-polyacrilamide gel electrophoresis (SDS-PAGE), 37 Software for copy number variations estimation from SNP array data, 81–83 CNAG, 83 dChip, 83 PLASQ, 83 Sophisticated machine learning and analysis tools, in pharmacogenomic biomarkers identification, 361–362 Spindle checkpoint genes transcriptional regulation, BRCA1 in, 239–242 Split-sample method, 333 Stability of “cancer pathways”, 68–69 generically unstable pathway, 67 specifically unstable pathway, 67 unstable pathway, anti-cancer therapy for, 69–70 Stem cell theory of cancer, 6–7 T790M mutation in NSCLC, 118–119 T7-type amplification, 8 6-Thioguanine (6-TG) in childhood ALL treatment, 173, 176–190 metabolism of, 178–180 Thiopurine S-methyltransferase (TPMT) genotyping, 183–190, 355 in African population, 184–186 in Asian population, 184–186 in Caucasian populations, 183 genetic variants, 183–190 HPLC assays, 183, 186 radiochemical assays, 183 RFLP-PCR method, 186 Thiopurines in childhood ALL treatment, 173–190 choice of, trial results COALL-92 trial, 177–178 UK MRC ALL 97, 177–178 U.S. trial CCG-1952, 178 6-mercaptopurine (6-MP), 173, 176–178 monitoring of, 180–182 6-thioguanine (6-TG), 173, 176–178 Thymidine kinase (TK), 153 Thymidine phosphorylase (TP), 153

Index Thymidylate synthase (TS) gene variations/Thymidylate synthase enhancer region (TSER) in colorectal cancer patients, 151–167 pharmacogenomics and prognostic and predictive markers, 156–157 TS gene expression, regulation, 157 DPD polymorphism, 153 MTHFR polymorphism, 153 orotate phosphoribosyltransferase (OPRT), 153 in pediatric ALL treatment response, 304–306 confounding issues, 306 thymidylate synthase polymorphisms, 158–160 loss of heterozygosity (LOH), 160  thymidylate synthase 3 -UTR 6bp deletion, 160 TSER 3R G to C SNP, 159–160 TS polymorphisms as prognostic and predictive factors, 161–163 See also Methotrexate Timelines, pharmaceutical, 315–316 Tissue array, breast cancer patients, 288, 291–292 Topoisomerase I (TOP1) amplification, 94 Tositumomab, 204, 211 Toxicogenomics application to oncology drug development, 339–348 data analysis, 341–344 DNA microarrays, 345 gene expression analyses, 340–341, 345 transcription profiling, 347 TPMT, see Thiopurine S-methyltransferase (TPMT) genotyping Transcription profiling/regulation, 347 BRCA1 in, 239–242 breast cancer patients, 288–295 DNA microarrays for, 289–291 Translational suppression, 47–53 Transmission disequilibrium tests (TDT), in family based genetic association studies, 26–27 Trastuzumab, 318 TS, see Thymidylate synthase (TS) Tumor heterogeneity, 5–6 breast cancer, 5

377 Two-dimensional gel electrophoresis (2-DE), 35–43 in proteomic analysis, 33–43 2-D-DIGE, 35 isoelectric focusing (IEF), 37 SDS-PAGE, 37 two-dimensional immunoblot analysis, 37 Tyramide signal amplification (TSA) technique, 9 Tyrosine kinase inhibitors (TKIs), 103–121 in human lung cancer EGFR mutations and sensitivity to, 104–121, See also Epidermal growth factor receptors Uridine diphosphate glucuronosyltransferase 1As (UGT1As), 268, 355–356 in anti-tumor responses, 281–282 ethnic differences of, 270–277 genetic polymorphisms, 270–277 haplotype structures covering, 271–277 in irinotecan pharmacokinetics and adverse reactions, 277–281 UGT1A1 Genotypes/Haplotypes, 277–280 on irinotecan treatment, 267–283 UGT1A1, genetic variations, 271–272 UGT1A10, genetic variations, 271 in irnotecan pharmacogenetics, 280–281 UGT1A7, genetic variations, 271, 273 in irnotecan pharmacogenetics, 280–281 UGT1A9, genetic variations, 271, 273 in irnotecan pharmacogenetics, 280–281 Variable number of tandem repeats (VNTR) polymorphism, 356 Waldenstrom’s macroglobulinemia (WM), 219–222 Whole-genome amplification, 76–85 for SNP array analysis, 80 X3P arrays, 11 Zevalin, see Ibritumomab Tiuxetan