Novel Pharmacogenomic Markers of Irinotecan-Induced Severe Toxicity in Metastatic Colorectal Cancer Patients

Home , Bilirubin glucuronide, C8orf34, Drug tolerance

Novel pharmacogenomic markers of irinotecan-induced severe toxicity in metastatic colorectal cancer patients

Thèse

Siwen Sylvialin Chen

Doctorat en sciences pharmaceutiques Philosophiae doctor (Ph.D.)

Québec, Canada

Résumé

L’irinotécan est un agent de chimiothérapie largement utilisé pour le traitement de tumeurs solides, particulièrement pour le cancer colorectal métastatique (mCRC). Fréquemment, le traitement par l’irinotécan conduit à la neutropénie et la diarrhée, des effets secondaires sévères qui peuvent limiter la poursuite du traitement et la qualité de vie des patients. Plusieurs études pharmacogénomiques ont évalué les risques associés à la chimiothérapie à base d’irinotécan, en particulier en lien avec le gène UGT1A, alors que peu d’études ont examiné l’impact des gènes codant pour des transporteurs. Par exemple, le marqueur UGT1A1*28 a été associé à une augmentation de 2 fois du risque de neutropénie, mais ce marqueur ne permet pas de prédire la toxicité gastrointestinale ou l’issue clinique. L’objectif de cette étude était de découvrir de nouveaux marqueurs génétiques associés au risque de toxicité induite par l’irinotécan, en utilisant une stratégie d’haplotype/SNP-étiquette permettant de maximiser la couverture des loci génétiques ciblés. Nous avons analysé les associations génétiques des loci UGT1 et sept gènes codants pour des transporteurs ABC impliqués dans la pharmacocinétique de l’irinotécan, soient ABCB1, ABCC1, ABCC2, ABCC5, ABCG1, ABCG2 ainsi que SLCO1B1. Les profils de 167 patients canadiens atteints de mCRC sous traitement FOLFIRI (à base d’irinotécan) ont été examinés et les marqueurs significatifs ont par la suite été validés dans une cohorte indépendante de 250 patients italiens. Nous avons découvert dans la région intergénique en aval du gène UGT1, un nouveau marqueur (rs11563250G) associé à un moindre risque de neutropénie sévère (rapport des cotes (RC)=0.21; p=0.043 chez les canadiens, RC=0.27; p=0.036 chez les italiens, et RC=0.31 p=0.001 pour les deux cohortes combinées). De plus, le RC est demeuré significatif après correction pour multiples comparaisons (p=0.041). Par ailleurs, pour l’haplotype défini par les marqueurs rs11563250G et UGT1A1*1 (rs8175347 TA6), le RC était de 0.17 (p=0.0004). Un test génétique évaluant ces marqueurs permettrait d’identifier les patients susceptibles de bénéficier d’une augmentation de dose d’irinotécan. En revanche, une autre combinaison de marqueurs, ABCC5 rs3749438 et rs10937158 (T–C), a prédit un risque plus faible de diarrhée sévère dans les deux cohortes (RC = 0.43; p=0.001). La coexistence des marqueurs ABCG1 rs225440T et ABCC5 rs2292997A a prédit un risque accru de neutropénie (RC=5.93; p=0.0002), alors qu’une prédiction encore plus significative a été obtenue lorsque ces marqueurs sont combinés au marqueur de risque bien établi UGT1A1*28 rs8175347 (RC=7.68; p<0.0001). Enfin, les porteurs de l’allèle de protection

iii UGT1 rs11563250G en absence d’allèles de risque, ont montré une incidence réduite de neutropénie sévère (8.2% vs. 34.0%; p<0.0001). Nous concluons que ces nouveaux marqueurs génétiques prédictifs pourraient permettre d’améliorer l’évaluation du risque de toxicité et personnaliser le traitement à base d’irinotécan pour les patients atteints du cancer colorectal métastatique.

Abstract

Irinotecan is a cytotoxic agent widely used for the treatment of solid tumors, most particularly for metastatic colorectal cancers (mCRC). Treatment with this drug frequently results in severe neutropenia and diarrhea that can seriously impact the course of treatment and patients’ quality of life. Pharmacogenomic tailoring of irinotecan-based chemotherapy has been the subject of several investigations, especially for the UGT1A1 gene, but with limited data regarding transporter genes. In this study, we sought to discover toxicity-associated markers using a haplotype-tagging SNP (htSNP) strategy to maximize gene coverage. We examined the genetic association across the UGT1 locus, and in seven transporter genes participating in irinotecan pharmacokinetics involving the ABC transporter genes ABCB1, ABCC1, ABCC2, ABCC5, ABCG1, ABCG2 and the solute carrier organic anion transporter gene SLCO1B1. The profiles of 167 mCRC Canadian patients treated with FOLFIRI-based regimens were examined and findings were replicated in an independent cohort of 250 Italian patients. We found rs11563250G, located in the intergenic region downstream of UGT1, to be significantly associated with reduced risk of severe neutropenia (odds ratio (OR)=0.21; p=0.043 and OR=0.27; p=0.036, respectively, and OR=0.31 when combined; p=0.001), which remained significant upon correction for multiple testing in the combined cohort (p=0.041). For the two-marker haplotype rs11563250G and UGT1A1*1 (rs8175347 TA6), the OR was of 0.17 (p=0.0004). Genetic testing of this marker may identify patients who might benefit from increased irinotecan dosing. In combined cohorts, a two-marker ABCC5 rs3749438 and rs10937158 haplotype (T–C) predicted a lower risk of severe diarrhea (odds ratio (OR) of 0.43; p=0.001). The co-occurrence of ABCG1 rs225440T and ABCC5 rs2292997A predicted an increased risk of severe neutropenia (OR=5.93; p=0.0002), which was further improved when incorporating the well-known risk marker UGT1A1*28 rs8175347 (OR=7.68; p<0.0001). In contrast, carriers of one protective marker (UGT1 rs11563250G) but none of these risk alleles experienced significantly less severe neutropenia (8.2% vs. 34.0%; p<0.0001). This combination of predictive genetic markers could lead to better risk assessment and may thus enhance personalized treatment.

Table of Contents

Résumé ...... iii Abstract ...... v Table of Contents ...... vii List of Tables ...... ix List of Figures ...... xi List of Abbreviations ...... xiii Acknowledgements ...... xvii Foreword ...... xix Chapter I Introduction ...... 1 1. Colorectal cancer...... 1 1.1 Disease Prevalence and Incidence ...... 1 1.2 An overview of treatment options ...... 2 2. Irinotecan, an antineoplastic agent used in first-line treatment of mCRC . 4 2.1 Mechanism of Action ...... 6 2.2 Pharmacokinetic pathways of irinotecan: metabolism and transport ...... 7 2.3 Tissular & Cellular transport of irinotecan and its metabolites ...... 12 2.4 Pharmacodynamic pathways of irinotecan ...... 19 2.5 Toxicities induced by irinotecan ...... 22 3. Pharmacogenomics (PGx) ...... 25 3.1 Established pharmacogenomics tests : a few examples ...... 26 3.1.1 Warfarin ...... 26 3.1.2 Clopidogrel ...... 28 3.2 Irinotecan Pharmacogenomics ...... 29 3.2.1 Molecular and clinical impact of UGT1A1*28 ...... 29 3.2.2 Other UGT1A variants ...... 31 3.2.3 UGT1A1*28 genotyping test ...... 33 4. Hypothesis & Objectives ...... 39 4.1 Identify UGT1 markers to better predict risk of severe toxicity ...... 39 4.2 Identify novel markers to help predict toxicity induced by irinotecan in drug transport pathways ...... 40 5. Methodologies & Approaches ...... 41 5.1 Study cohorts: ...... 41 5.1.1 Discovery cohort ...... 41 5.1.2 Validation cohort ...... 42 5.2 Evaluation of toxicity outcomes ...... 43

vii 5.2.1 Severe neutropenia ...... 43 5.2.2 Severe diarrhea ...... 43 5.2.3 Bilirubin levels ...... 44 5.3 Selection of candidate genes and single nucleotide polymorphisms (SNPs) .. 45 5.4 DNA Sample preparation and Genotyping ...... 46 5.5 Statistical analysis ...... 46 5.5.1 Hardy-Weinberg equilibrium ...... 46 5.5.2 Genetic Associations ...... 47 5.6 Functional studies and in-silico analyses ...... 48 5.6.1 3’ Rapid Amplification of cDNA Ends (RACE) ...... 48 5.6.2 Resequencing of UGT1 locus ...... 49 5.6.3 In-silico predictive bioinformatics analysis ...... 49 Chapter II ...... 51 Identification of a novel genetic marker in UGT1A locus with better tolerance against severe neutropenia in metastatic colorectal cancer patients ...... 51 Chapter III ...... 81 Identification of ABCC5 and ABCG1 polymorphisms that predict irinotecan- induced severe toxicity in metastatic colorectal cancer patients ...... 81 Discussion ...... 125 Conclusion ...... 131 Bibliography ...... 133

viii

List of Tables

Table 1. Common treatment strategies used in the management of colorectal cancer...... 2 Table 2. Irinotecan-based regimens for treatment of metastatic colorectal cancer (mCRC) 5 Table 3. Dose-limiting toxicities experienced by irinotecan-receiving patients ...... 6 Table 4. UGT1A polymorphisms found to influence SN-38 glucuronidation rates in vitro .. 11 Table 5. Selected drug transporters and their substrates (non-exhaustive) ...... 12 Table 6. Major transporter gene polymorphisms and clinical outcomes studied previously in irinotecan-treated cancer patients ...... 16 Table 7. Pharmacodynamic genes associated with SN-38 pharmacokinetics, toxicity and response in irinotecan-receiving colorectal cancer patients...... 21 Table 8. Dosing recommendations according to current CPIC guidelines based on VKORC1 and CYP2C9 genotype ...... 27 Table 9. UGT1A gene polymorphisms and associations with clinical outcomes studied previously in irinotecan-treated cancer patients ...... 32 Table 10. Demographic and Clinical characteristics studied in Discovery and Validation cohorts ...... 42 Table 11. NCI-CTACE v3.0 criteria for severe neutropenia and diarrhea toxicities ...... 44 Table 12. Genes involved in the metabolism and transport of SN-38 studied in the discovery cohort ...... 45 Table 13. In silico bioinformatic databases and tools used to evaluate the potential functionality of novel markers ...... 49

List of Figures

Figure 1. Global cancer incidence, death and prevalence rates according to cancer types from 184 countries...... 1 Figure 2. Metabolism, biotransformation and inactivation reactions of CPT-11 ...... 7 Figure 3. UGT1A enzymes involved in the metabolism of CPT-11 and its metabolites ..... 10 Figure 4. Commonly known SNPs in the UGT1 locus ...... 11 Figure 5. Schematic representation of ABC and SLCO transporter members involved in the transport of irinotecan and its metabolites from hepatocytes to the intestine...... 13 Figure 6. Distribution of Irinotecan and its metabolites excreted in the bile and urine ...... 18 Figure 7. SN-38 mechanism of action on Top1-DNA complex and activation of pharmacodynamic pathways ...... 19 Figure 8. Timeline of Pharmacogenomics over the last decade...... 25 Figure 9. Schematic representation of UGT1A1*28 on UGT1 locus and its effect on SN-38 glucuronidation ...... 30

xii

List of Abbreviations

3’UTR 3’Untranslated region 5-FU 5-Fluorouracil ABC ATP-binding cassette transporters ADR Adverse Drug Reaction ANC Absolute Neutrophil Count APC 7-ethyl-10-[4-N-(5-aminopentanoic acid)-1-piperidino]carbonyloxycamptothecin AUC Area under the curve CAP Capecitabine CDC45L Gene encoding for CDC45-related protein (cell-division cycle 45) CES Carboxylesterase I or II CEU Caucasian Europeans from Utah ChIP-seq Chromatin Immunoprecipitation Sequencing CPIC Clinical Pharmacogenomics Implementation Consortium CPT Camptothecin CYP Cytochrome P450 enzyme DLT Dose-limiting toxicity DPWG Dutch Pharmacogenomics Working Group ECOG Eastern Cooperative Oncology Group EGAPP US Evaluation of Genomic Applications in Practice and Prevention EGFR Epidermal Growth Factor Receptor ENCODE Encyclopedia of DNA Elements ERCC1 Excision repair cross-complementation group I protein FDA United States Food and Drug Administration G-CSF Granulocyte-Colony Stimulating Factor GI Gastrointestinal GSTP1 Gene encoding for Glutathione-S-transferase P enzyme GWAS Genome-Wide Association Study htSNP Haplotype-Tagging SNP HWE Hardy-Weinberg Equilibrium INDEL Insertion/Deletion IRI Irinotecan IV Intravenous KRAS GTPase KRas protein LD Linkage Disequilibrium LV Leucovorin M4 Metabolite of irinotecan MAF Minor Allele Frequency mCRC metastatic Colorectal Cancer MMF Mycophenolate Mofetil MPA Mycophenolic Acid MPAG Mycophenolic Acid Glucuronide MS Mass Spectrometry MTD Maximum Tolerated Dose NCBI National Center for Biotechnology Information NCCN National Comprehensive Cancer Network NCI-CTCAE National Cancer Institute Common Toxicity Criteria for Adverse Events NGS Next Generation Sequencing NPC 7-ethyl-10-(4-amino-1-piperidino) carbonyloxycamptothecin

xiii OR Odds Ratio OS Overall Survival OX Oxaliplatin PARP1 Poly(ADP-ribose) polymerase enzyme PCR Polymerase Chain Reaction PD Pharmacodynamic PFS Progression-free Survival PGx Pharmacogenomics PK Pharmacokinetics RACE 3’ Rapid Amplification of cDNA Ends SLCO Solute-carrier organic anion transporter SNP Single Nucleotide Polymorphism SSB/DSB Single/Double-Strand Breaks TDP1 Tyrosyl-DNA phosphodiesterase I enzyme TOP1 Topoisomerase I UGT UDP-glucuronosyltransferase ULN Upper Limit Of Normal Range VEGF-A Vascular Endothelial Growth Factor A VKORC1 Vitamin K Epoxide Reductase Complex C1 XRCC1 X-ray repair complementation group I protein

xiv

May all sentient beings have happiness and its causes, May all sentient beings be free of suffering and its causes, May all sentient beings never be separated from bliss without suffering, May all sentient beings be in equanimity, free of bias, attachment and anger. - The Four Immeasurables

To all sentient beings, may they be free of diseases and sufferings

xvi

Acknowledgements

First, I would like to express my utmost gratitude and appreciation to my doctoral advisor, Dr. Chantal Guillemette, for giving me the opportunity to be mentored and trained in your laboratory. Despite a rocky start, I grew to understand and appreciate your motivation to expect the best from your students. I have learned excellent organization skills, the rigors of scientific investigations, developed critical thinking as well as how to function as a team player. My time under your mentorship has allowed me to evolve not only personally but also enhance my communication skills and work ethics professionally. Your ability to be firm and yet patient during moments of self-doubt and poor self-confidence has enabled me to persevere and complete my studies. This dissertation would not have been completed without your constant encouragement and critical feedback. My goal was to work and learn from the best in the field of Pharmacogenomics and I will cherish every little bit of my time here. I strive to use my training in this field into my future endeavors and make you proud!

Next, I thank all the members of the jury committee for spending time to review this dissertation. I am honored to be given the opportunity to present my work to you for your comments.

Naturally, every successful laboratory is made up of a team of cohesive members. I would like to thank especially Isabelle Laverdiere, an excellent fellow PhD student who helped make my transition into the lab a smooth one. Thank you for spending your precious time showing me the ropes, providing constructive comments and giving me pep talks. In part, this PhD would not have been successfully completed without your support. Another person who deserves many thanks is Lyne, who has been a delight to work with. Thank you for guiding me with my experiments, questioning my rationales and most importantly helping me with my Français! Merci beaucoup! Also, a special mention should be given to Anais, whose arrival brought much joy and vivacity into the lab. Thank you for your candor and positive spirit, you often made my frown into a smile! Thank you Michele for entertaining my questions, they helped me to put things in perspective, especially when writing this thesis; I thoroughly enjoyed our scientific conversations! To other members, former and present, thank you for your presence and friendship, and also tolerating my broken French! You all helped me to assimilate into Quebec fast and furious, without any official French courses!

xvii My cross-border move from Boston would not have been smooth-sailing if it were not for my group of dedicated friends who never failed to encourage and support me mentally and physically during times of need. They convinced me that I would overcome all challenges in front of me and never doubted that I would succeed. To Kelly, Rachel, Minu (and family), Casey (and family), Ieva, Joey (and Rory) and Myra, I am truly blessed to have you all in my life. Thank you for being there for me always.

To all the wonderful people I met here in Quebec who became my close friends, namely, Joannie, Adriana and Ricardo, thank you for making my life here not all about work. I am fortunate to explore some of Quebec’s beauty and culture in spite of my hectic student life with you. I express my sincere gratitude towards new friends, Aimee, Ewan, Fiona and Tessa, who took me into their family, cared and nurtured me during these past months of difficulties. I am eternally grateful for your love, generosity, kindness, support, keeping my spirits up, dispensing advice and helping me regain confidence. Thank you seems insufficient in expressing how much I am indebted to your family, but I hope to repay you some day.

Last but not least, I want to thank my family for their endless love and support, albeit from afar in Singapore. I was first inspired to study Pharmacogenomics as a result of my father’s condition and his medication, warfarin. I am proud to declare that I have succeeded in this goal and I plan to apply this knowledge acquired to benefit more people in the future. To my mother, who showered me with love and concern, always making sure that I focused and never strayed from my goal. I did it! Also, thank you to my sisters, Joy and Charlene, for holding the fort down at home and keeping me sane. I hope you both would achieve similar success in your careers.

Thank you to all the wonderful people around me, I am excited to explore the next stage in my life and cannot wait for you to be my witnesses!

xviii

Foreword

This thesis, entitled, “Novel pharmacogenomic markers of irinotecan-induced severe toxicity in metastatic colorectal cancer patients,” embodies the work achieved during my doctoral training under Dr. Chantal Guillemette’s mentorship in the Pharmacogenomics Laboratory at CHU de Quebec Research Center. The purpose of this thesis is for submission to Université Laval for the degree of Philosophiae doctor, encompassing two original articles, “A novel UGT1 marker associated with better tolerance against irinotecan-induced severe neutropenia in metastatic colorectal cancer patients” and “ABCC5 and ABCG1 polymorphisms predict irinotecan-induced severe toxicity in metastatic colorectal cancer patients”

The first article, “A novel UGT1 marker associated with better tolerance against irinotecan- induced severe neutropenia in metastatic colorectal cancer patients“ was completed in collaboration with multicenters in Canada and Italy, comprising of 167 and 250 Caucasian metastatic colorectal cancer patients. This article was recently published in the Pharmacogenomics Journal (Impact factor: 5.513) in March 2015. My role has been to participate in the selection of candidate genes and SNPs, together with another Ph.D. student Isabelle Laverdière. I also participated in sample preparation, statistical analysis, interpretation of results and further in-silico analyses while Isabelle Laverdière and Lyne Villeneuve performed and validated all statistical analyses, with support by Alan Tourancheau for bioinformatic scripts. Mario Harvey participated in the initial data analysis. I participated in the draft of the manuscript while Dr. Chantal Guillemette wrote the complete version of the paper with the help of Dr. Lévesque. All authors performed a critical review of the manuscript. All tables and figures presented in the article were ensembled by Isabelle and myself. Dr. Derek Jonker, Dr. Félix Couture, Dr. Èric Lévesque and their team recruited Canadian patients and provided clinical information while Dr. Erika Cecchin and Dr. Giuseppe Toffoli coordinated the Italian cohort. Dr. Michael H. Court provided the human liver tissue samples. The conception, design and supervision of the study was by Dr. Chantal Guillemette.

The second article, “ABCC5 and ABCG1 polymorphisms predict irinotecan-induced severe toxicity in metastatic colorectal cancer patients” has been accepted in the journal Pharmacogenetics and Genomics (Impact factor: 3.481) in July 2015. It was also performed with our Italian collaborators and their roles were the same as in the first study.

xix My role was essentially the same as in the first study with regards to SNPs selection, data analysis and drafting parts of the manuscript. Statistical analyses and their validation were performed by Lyne Villeneuve and Isabelle Laverdière. Dr. Chantal Guillemette wrote the complete version of the paper. All authors performed a critical review of the manuscript. The conception, design and supervision of the study was by Dr. Chantal Guillemette.

Chapter I Introduction 1. Colorectal cancer

1.1 Disease Prevalence and Incidence Colorectal cancer (CRC) is currently the third most common cancer in the United States and Canada and the third leading cause of death in men and women. As of 2014, the incidence rates projected amongst men and women in North America are approximately 51 per 100,000 in men and 40 per 100,000 in women. These North American trends are in line with the global cancer statistics whereby CRC was the third most diagnosed type of cancers (10%) in 2012, after breast and lung cancers while the mortality rate (9%) caused by CRC deaths was ranked fourth. The 5-year global prevalence rate of CRC remains at 11%, a drastic improvement from the mid-20th century (1930-1950s) as a result of improved treatment options and early detection. In particular, a 3% decrease in incidence rates per year during the period of 2001 and 2010 has been observed.1-3 (Figure 1)

Figure 1. Global cancer incidence, death and prevalence rates according to cancer types from 184 countries. A) Overall cancer incidences ranked by cancer types in 2012. B) Mortality rates segregated by cancer types in 2012. C) 5-year prevalence rates of cancer types. (Adapted from GLOBOCAN project 2012, http://globocan.iarc.fr accessed Oct 15, 2014)

1 These trends correlate with better early detection and stage diagnosis of CRC, improving the 5-year relative survival rate for patients with all tumor stages from 65% to 90% for those with localized tumors.1-3

1.2 An overview of treatment options The improvement of 5-year relative CRC survival rate is in part due to a greater adoption of colorectal screening during annual medical examinations, advances in surgical and radiation therapy and the use of efficacious chemotherapeutic agents. Treatment strategies of diagnosed CRC are largely dependent on type of tumor, location, disease stages, the extent of tumor progression, curative intention, genetic mutations and tolerance to toxicities. Table 1 lists the common clinical treatment options used in the management of CRC.

Table 1. Common treatment strategies used in the management of colorectal cancer.

Treatment Option Clinical Goal/Intention Disease Stage Surgery - Open Resect metastases and localized tumor, only I-III, neoadjuvant option of long-term survival therapy candidates - Laparoscopy Radiation therapy I-III, candidates who - short course Shrink tumor size and prevent recurrence, reject surgery relieve cancer symptoms (ex. pain) - long course IV, liver/lung Prolong survival and prevent tumor metastases, Chemotherapy progression candidates for resection - monotherapy 5-fluorouracil/leucovorin (5-FU/LV) irinotecan metastatic CRC oxaliplatin patients capecitabine - combination therapy FOLFIRI (5-FU/LV, irinotecan) (regimens) FOLFOX (5-FU/LV, oxaliplatin) metastatic CRC CAPOX (capecitabine, oxaliplatin) patients CAPIRI (capecitabine, irinotecan) FOLFOXIRI (5-FU/LV, irinotecan, oxaliplatin) - targeted therapy bevaxizumab (administered with cetuximab1 metastatic CRC another regimen) panitumumab1 patients with drug aflibercept resistance to sorafenib2 regimens regorafenib Abbreviations: IRI = irinotecan, 5-FU, 5-fluorouracil, LV, leucovorin, CAP, capecitabine, OX, oxaliplatin, CRC, colorectal cancer 1 Suitable for patients without KRAS mutations identified by FDA-approved genetic test; wild-type KRAS patients exhibited increased overall response and survival) 4-8

2 Currently in clinical trials for treatment of mCRC

Unfortunately for the majority of patients who are presented with metastases at time of diagnosis (almost 40%-50%)9,10, the main intention is to prolong survival and prevent further tumor progression. Hence, intensive treatment with chemotherapeutic agents becomes the best logical option. Patients who were not previously considered for resection may also benefit from neoadjuvant chemotherapy to become candidates for surgical removal options. Furthermore, the risk of local recurrence in Stage III patients is 15-50%, hence adjuvant chemotherapy is also employed in an effort to prevent regrowth of resected tumors.2

Current combination therapies have been favored over monotherapy strategies as a result of improved response rates, overall survival (OS) and progression-free survival (PFS) rates. The two most common combination regimens used in medical centers in North America and Europe are FOLFIRI (irinotecan-based) and FOLFOX (oxaliplatin-based). The preference to use one regimen over the other often depends on clinicians’ decisions, type of medical center, availability of chemotherapeutic agents, toxicity occurrence, tumor resistance and patients’ physical condition and economic circumstances.11-14 In fact, one major advantage irinotecan-containing regimens have over oxaliplatin-based ones is the absence of neurotoxicity after several months of therapy. In lieu of a dose-reduction, patients may choose to switch to an irinotecan-based treatment in order not to compromise therapeutic objectives. Irinotecan has also been observed to incur a 1.2-1.8 fold increase in PFS and OS when administered concurrently with 5-FU/leucovorin compared with non-irinotecan containing treatments.15-17 The most frequently used irinotecan-containing treatments are available in Table 2.

In recent years, novel biologics have been developed to improve cytotoxic targeting of tumors and are undergoing clinical trial testing in conjunction with existing regimens. Despite early promising results, the clinical implementation of targeted therapy is still nascent and a majority of medical centers continue to use either irinotecan-based or oxaliplatin-based regimens.12,18,19

Several trials have also assessed the risk-benefit characteristics of FOLFOXIRI (irinotecan+oxaliplatin+5-fluorouracil+leucovorin) but have yet to demonstrate a superior efficacy and decreased toxicities over existing FOLFIRI treatment. Until further evaluative

3 studies on the added benefits of oxaliplatin or novel biologics, FOLFIRI is still one of the best regimens for the treatment of metastatic colorectal cancer (mCRC).

2. Irinotecan, an antineoplastic agent used in first-line treatment of mCRC

The introduction of irinotecan for use as an anticancer drug has tremendously improved the current state of treatment options in metastatic colorectal cancer patients. The efficacy of irinotecan in the treatment of mCRC is far superior over 5-FU/LV treatment. Irinotecan- based regimens indicate a two-fold improvement in overall survival, progression-free survival and response rates, with less adverse toxic events. In fact, a 50% reduction in chemotherapy-induced gastrointestinal toxicity was observed with irinotecan monotherapy compared to 5-FU/leucovorin.16

Irinotecan (CPT-11) is a water-soluble camptothecin analog synthesized by Yakult Honsha Co Ltd (Tokyo, Japan), displaying antitumor activity in mouse and human carcinoma cell lines.20-24 The anticancer drug class, camptothecins are plant alkaloids first isolated from the bark of the Chinese tree Camptotheca acuminata and were found to have antileukemic properties in mice cell lines.25,26 Characterization studies later confirmed this cytotoxic effect was a result of specific inhibition of DNA topoisomerase I (Top1) enzyme activity, by induction of single-strand DNA breaks and premature termination of replication.27,28

Unlike traditional camptothecins, irinotecan was found to exhibit favorable physical properties, such as solubility, stability and reduced side-effects.29,30 Furthermore, as a prodrug, CPT-11 can be safely administered intravenously without drug delivery problems. Therefore, irinotecan has emerged as the best camptothecin-based drug currently used for treatment of colorectal and lung cancers.

Table 2. Irinotecan-based regimens for treatment of metastatic colorectal cancer (mCRC)

Regimen Irinotecan Leucovorin 5- Oxaliplatin Frequency Fluorouracil Once 350mg/m2 IV every 3 (90 min) Monotherapy weeks 125mg/m2 IV Once per

(90 min) week 400mg/m2 IV Once 400mg/m2 Bolus + FOLFIRI 180mg/m2 IV every 2 IV 2400mg/m2 IV weeks (46hrs) Once 200mg/m2 3200mg/m2 IV FOLFOXIRI1 165mg/m2 IV 85mg/m2 IV every 2 IV (48hrs) weeks 1 Recently completed in clinical trials, actual use in practice unknown Abbreviation: IV, intravenous

A major limitation to any irinotecan-based treatment is the occurrence of two dose-limiting toxicities (DLT) - diarrhea and severe myelosuppression (ex. neutropenia). Numerous studies have reported that neutropenia and late-stage diarrhea were the most common forms of toxicities experienced by patients receiving irinotecan. Approximately 10-54% patients were found with severe Grade 3-4 neutropenia while 9-30% had severe Grade 3-4 diarrhea.15-17,31,32 (Table 3)

The clinical implication for severe neutropenia is the inability to fight infections as a result of decrease neutrophil production and a compromised immune system. Similarly, gastrointestinal toxicities are a common adverse effect of chemotherapy. In particular, the most severe type of chemotherapy-induced diarrhea is late-onset at 24 hours-3 days after treatment with irinotecan. This event is often debilitating, life-threatening as well as socially-challenging to manage. Patients may not accurately report their bowel patterns, as a result of embarrassment, leading to ineffective prophylactic use. Overall, both toxicities could result in hospitalization and delay in chemotherapy treatment. Loperamide, an anti- diarrheal drug may be given to ameliorate late-onset diarrhea episodes. Almost half of the patients undergoing irinotecan-based regimens will experience a DLT. Even though a dose reduction can likely reduce these events, it is performed at the expense of chemotherapeutic tumor response. Therefore, the ability to predict patients who would be at risk for severe neutropenia and diarrhea would greatly benefit from dose-management decisions in mCRC treatment, without compromising therapeutic response.

5 Table 3. Dose-limiting toxicities experienced by irinotecan-receiving patients

Toxicity Clinical definition Measurement Consequence (frequency) Decreased granulocyte Risk of infection Absolute neutrophil Neutropenia production or damage; increases with low count (ANC) (14-31%) onset within 6 months of ANC, more severe <1.5x109/L chemotherapy with fever, fatal Dehydration, present Diarrhea - Cholinergic effect of Increase in # with other symptoms Early Stage irinotecan (within 30 stools/day compared (ex. cramping, (7-22%) mins) to before treatment1 rhinitis) manageable with atropine Debilitating, effect on Breakdown of intestinal Increase in # quality of life, Diarrhea – mucosa by irinotecan- stools/day compared precursor for other Late Stage cytotoxicity, poor to before treatment disorders, requires (14-31%) absorption of fluid by (≥7 stools/per day)1 hospitalization, life- epithelial cells threatening 1 It is necessary to determine the baseline levels of bowel patterns before evaluation

2.1 Mechanism of Action As a synthetic analog of camptothecin, irinotecan exerts antitumor activity by targeting Top1 to disrupt DNA replication. The prodrug (CPT-11) is converted into its active metabolite (SN-38) and other minor metabolites (M4, APC, NPC) by carboxylesterase (CES) and CYP3A4/5 enzymes respectively. (Figure 2) The active metabolite SN-38 is 100-1000x more potent than CPT-11 in part due to its specificity as an intercalating molecule between Top1-DNA single-strand interactions.33-36 Like other camptothecins, SN- 38 binds to Top1-DNA cleavage complexes (Top1-cc) formed momentarily during DNA unwinding and prevents re-ligation of the template strand.37-39 This binding is reversible upon changes in physiological pH and absence of active compound. Hence, the effect on tumor death and drug response is strongly dependent on the bioavailability of this active metabolite.40,41

Figure 2. Metabolism, biotransformation and inactivation reactions of CPT-11 (Adapted from Mathijssen RH et al, 200142)

In contrast, the prolonged exposure of toxic SN-38 in the systemic circulation and intestinal cells can result in onset of myelosuppression and gastrointestinal toxicities.43,44 These adverse events are likely to occur more frequently due to inadequate biotransformation (CES) and inactivation (UGT1A) caused by altered protein function.

2.2 Pharmacokinetic pathways of irinotecan: metabolism and transport As an antineoplastic agent, the metabolism of CPT-11 is indeed crucial in delivering therapeutic concentrations of SN-38 to tumor cells. The efficacy of CPT-11 on tumor response and induced toxicities relies on the efficiency of various pharmacokinetic pathways that transform, inactivate and metabolize CPT-11. The biotransformation of CPT-11 is performed by carboxylesterases, enzymes that hydrolyze the piperidino side chain to form active SN-38.45 (Figure 2) Human CES2 (hCES2) was discovered to have a 12.5-fold higher specificity for CPT-11 and is 26-fold more active than hCES1.46,47 hCES1 and hCES2 are commonly expressed in the liver, kidney and gut, amongst other tissues, where their biotransformation efficiencies are dependent on the tissue type. hCES2 is highly expressed and performs 99% of CPT-11 hydrolysis in the small intestine while hCES1 is present in the liver and is responsible for hydrolyzing at least 50% of CPT-11 levels.48,49

Many studies have also observed an interindividual variability of hCES2 expression in liver, colon and plasma specimens, indicating underlying phenotypic differences in CPT-11 conversion.49-52 A moderate significant correlation between CES2 mRNA expression and irinotecan-induced toxicities (severe neutropenia and diarrhea) was also confirmed, with 35% of CRC patients expressing high mRNA levels, who had experienced severe grade 3- 4 toxicities.53 Hence, it is clear that interindividual differences, most likely due to genetic variations, can account for the observed variability in tumor response and toxicities.

Other CPT-11-derived metabolites such as APC, NPC and M4 are derived from CYP3A- mediated metabolism. CYP3A4 and CYP3A5 catalyze the Phase I reactions of CPT-11 to form APC by hydrolyzing the terminal piperidine ring whereas the N-dealkylation of the distal piperidine ring forms NPC.54-56 These CYP450-converted metabolites do not display strong antitumor effect, unlike SN-38, and are negligible in predicting irinotecan efficacy or toxicity.29 In addition to irinotecan, the CYP3A family is responsible for the oxidative conversion of numerous drugs. Studies involving drug-drug interactions between irinotecan and CYP3A4 inhibitors have clarified that there was no change in the pharmacokinetic measures and induced-toxicities of irinotecan-receiving patients, in the presence of omeprazole.57 Conversely, ketoconazole was suggested to increase SN-38 exposure but failed to identify a significant effect on irinotecan clearance and SN-38 glucuronides.58 A haplotype-based CYP3A4 investigation also observed the absence of positive associations between CYP3A4 polymorphisms, clearance and irinotecan-induced toxicities.59 The contribution of concomitant medications towards these toxicities60 may indeed be important when considering irinotecan treatment efficacy. In fact, some have demonstrated a significant association between the administration of other drugs with irinotecan-related toxicities while others have clarified patients’ comorbidities, although indirectly indicate the use of multiple medications, do not affect toxicities.61,62 The mechanistic explanations of drug-drug interactions influencing irinotecan efficacy remain confounding. Differences in cohort size, methodological approaches and cancer types make it challenging to determine the impact of drug substrates of CYP3A4. For now, clinicians prefer to avoid co-administering irinotecan and CYP3A4 inhibitors.63

Interestingly, even though NPC was found to be 100x less potent than SN-38, it can be readily converted into active SN-38 by hCES2. The catalytic efficiency of hCES2 is more

pronounced for NPC than APC.64 Thus, it is no surprise that the majority of existing irinotecan-related pharmacokinetic and pharmacogenomic studies have investigated the effect of carboxylesterase-mediated formation of SN-38.

High circulating levels of toxic SN-38 often result in gradual degradation of epithelial cells in the intestine and liver. Histological and pharmacokinetic findings have correlated the increased exposure of SN-38 and CPT-11 to decreased absolute neutrophil counts (ANC)43,65,66, severe damage of intestinal mucosa 67,68 and higher biliary indexes.69,70

The inactivation of SN-38 is mediated by a Phase II reaction catalyzed by UDP- glucuronosyltransferase enzymes (UGTs) whereby the transfer of a glucuronic acid (G) moiety from uridine diphosphoglucuronic acid (UDPGA) forms a more polar SN-38 glucuronide (SN-38G). UGTs are widely expressed in drug-metabolizing tissues, glucuronidating numerous endogenous compounds (ex. steroid hormones) and xenobiotics (ex. drugs). The formation of these glucuronides enables compounds to be readily eliminated in the bile or urine.71

Human UGT enzymes are classified into four families, UGT1, UGT2, UGT3 and UGT8, with only UGT1 and UGT2 being involved in drug metabolism.72,73 Both UGT1 and UGT2 families possess distinct substrate specificity and account for the majority of glucuronidation activities. The UGT1A protein family consists of 13 members, which are derived from alternatively spliced first exons and 4 common exons (exons 2-5) on the gene UGT1 locus.74 In particular, in vitro studies have confirmed UGT1A proteins, namely UGT1A1, UGT1A6, UGT1A7 and UGT1A9 to be responsible for the inactivation of SN- 38.75-78 (Figure 3)

Figure 3. UGT1A enzymes involved in the metabolism of CPT-11 and its metabolites (from Guillemette, C et al, 201474)

Furthermore, strong interindividual variability of UGT1A1 mRNA expression detected in mice correlated with irinotecan-induced toxicities.79 As the only detoxification enzyme of SN-38, functional UGT1As are necessary to alleviate toxic effects in the liver, intestine and circulation. The UGT1A family is composed of transmembrane-bound proteins that are responsible for the detoxification of numerous drug substrates and exhibit distinct substrate specificity according to the type of drug metabolizing tissue.74 The tissue distribution of UGT1A enzymes emphasizes their important glucuronidating function and their ability to impact the clinical consequences of drug metabolism. The substrate specificity of multiple UGT1A enzymes in different tissues may infer a compensatory mechanism present in the event of a malfunction protein and/or suggest a cooperative effort in drug detoxification. Indeed, the co-occurrence of functional UGT1A variants was found to potentiate irinotecan-induced toxicity80 and indicate the need to systematically investigate these UGT1A enzymes together. The primary cellular location of UGT1A enzymes is in the endoplasmic reticulum where they are docked within the membrane with the substrate-binding domain exposed on the lumen.81 UGTs have also been detected in other subcellular compartments such as mitochondria, nucleus but their functions in these organelles have yet to be determined.82,83 Genetic variations in the UGT1 locus have been associated with changes in glucuronidation rates of SN-38 and inferred reduction in UGT1A function. These in vitro studies summarized in Table 4 demonstrate the impact of polymorphic changes on

UGT1A glucuronidating ability of SN-38. Figure 4 depicts the location of these commonly known UGT1 SNPs, of which several have been studied in relation to SN-38 metabolism.

Table 4. UGT1A polymorphisms found to influence SN-38 glucuronidation rates in vitro

UGT1 Variant rs # Effect on SN-38 Ref Gene glucuronidation UGT1A1 promoter TA repeat 7/7 rs8175347  SN-38G formation 84-86 (*28) c.-3156G>A (*93) rs10929302  SN-38G formation 84 c.-3279T>G (*60) rs4124874  SN-38G formation 84,78 G71R (*6) rs4148323  SN-38G formation 78,87 Y486D (*7) rs34993780  SN-38G formation 78,87 P229Q (*27) rs35350960  SN-38G formation 78,87 L233R (*35) rs72551344  SN-38G formation 78 UGT1A7 N129K, R131K (*2) rs17868323 - SN-38G formation 78,88 N129K,R131K,W208R (*3) rs17863778  SN-38G formation 88 W208R (*4) rs11692021  SN-38G formation 88 G115S (*5) rs61261057  SN-38G formation 88 E139D (*6) rs114052958  SN-38G formation 88 UGT1A9 M33T (*3) rs72551330  SN-38G formation 88 intronic C399T rs2741049  SN-38G levels 84 -118dT[9/10] (*22) rs3832043  SN-38G levels 84

Figure 4. Commonly known SNPs in the UGT1 locus

(*) Represents variants from Table 4 that have been previously identified to regulate SN-38 glucuronidation levels. (Modified from Guillemette, C et al 201474)

11 2.3 Tissular & Cellular transport of irinotecan and its metabolites The ATP-binding cassette (ABC) and solute-carrier (SLC) families of membrane transporter proteins are responsible for the uptake and efflux of CPT-11, SN-38 and its glucuronide, SN-38G. Together, they play a critical role in modulating the irinotecan pharmacokinetics and the bioavailability of the drug.89-95 The entire family of ABC transporter is divided into seven subfamilies of proteins, A-G. Each subfamily consists of transporters that have unique substrate specificity and structure. Currently, certain gene members of each subfamily, namely, ABCB1, ABCC1, ABCC2 and ABCG2 have been linked to drug transport and resistance to irinotecan. (Table 5) Similarly, the SLC superfamily of membrane-bound transporters facilitates the transport of a diverse group of molecules, including drug substrates. In particular, SLCO1B1 from solute-carrier organic anion transporter (SLCO or OATP) family is well characterized by its cellular disposition of drugs in the liver.96

Table 5. Selected drug transporters and their substrates (non-exhaustive)

Gene Drug substrate colchicine, doxorubicin, VP16, adriamycin, vinblastine, ABCB1 digoxin, saquinivir, paclitaxel, doxorubicin, daunorubicin, vincristine, indinavir, topotecan, CPT-11 ABCC1 VP16, colchicines, etoposide, rhodamine, adefovir, indinavir, CPT-11 ABCC2 vinblastine, cisplatin, sulfinpyrazone, SN-38, active metabolite of CPT-11 ABCC3 methotrexate, etoposide, tenoposide ABCC4 nucleoside monophosphates nucleoside monophosphates, mitoxantrone, topotecan, ABCC5 doxorubicin ABCG1 mitoxantrone, doxorubicin daunorubicin, doxorubicin, rosuvastatin, sulfasalazine, ABCG2 CPT-11, rhodamine atrasentan, atorvastatin, bosentan, ezetimibe, fluvastatin, glyburide, rosuvastatin, simvastatin acid, pitavastatin, SLCO1B1 pravastatin, repaglinide, rifampin, valsartan, olmesartan SN-38, active metabolite of CPT-11 Modified from 97 and U.S. Food and Drug Administration (FDA) website (accessed November 10, 2014)

In addition, each transporter member from the ABC and SLCO family is tissue-specific and tends to be localized on either the apical or basolateral side of the cellular membrane. As these drug carriers have either absorptive or secretory functions, the underlying

assumption is that clearance of drug metabolites is a result of a combined, cooperative effort of transporters present in a cellular location. Indeed, pharmacokinetic evidence depicting the role of various transporters in irinotecan disposition confirms that each transporter type has different substrate affinities.98 Figure 5 illustrates a schematic representation of drug transporters involved in active efflux and movement of irinotecan and its metabolites. As these transporters make up the entire cellular transport system, they are also speculated to work in sync with drug metabolizing enzymes to mediate overall drug transport.99-102

Figure 5. Schematic representation of ABC and SLCO transporter members involved in the transport of irinotecan and its metabolites from hepatocytes to the intestine. (Modified from Pancyzk, M. 2014103)

The adequate expression and function of these membrane-bound proteins is necessary for the movement of drugs across tissue compartments.97,104,105 It is unsurprising that these proteins are associated with drug resistance, often a result of impaired or enhanced transporter function, which can invariably change metabolite concentrations and impact tolerance levels. The interindividual variability of drug resistance in cancer treatments has

13 been suggested to be the result of variable expression of transporter proteins and Phase I, Phase II enzymes.106-108

The majority of this variability exists as polymorphisms in transporter genes that alter expression and function. The impact of these genetic variants on drug disposition has been widely evaluated in numerous in-vivo and in-vitro models, while associations to clinical outcomes in different diseases have been made in clinical studies.106,109,110 As a prototypic drug for mCRC treatment, the efficiency of irinotecan transport to its target tissue has largely been the focus of most investigations: outlining evidence of transporter dysfunction, tendency to acquire drug resistance and develop related adverse events. (Table 6)

The ABC transporter ABCB1, also known as multidrug resistance protein 1 (MDR1/p- glycoprotein) is located on the biliary canalicular side in the liver and the apical surface of gastrointestinal tract epithelial cells.111 Iyer et al. have found that ABCB1 has a stronger affinity for irinotecan, compared to its active metabolite, SN-38 or its glucuronide, SN- 38G.92 Numerous findings regarding polymorphisms in ABCB1 that influence irinotecan efflux and toxicities so far have been mixed and therefore necessitate further validations. The most frequently studied polymorphisms are three variants (1236>T, 2677G>A/T, 3435C>T) that are in strong linkage disequilibrium with one another.112 These variants have been linked to an increase in active metabolite concentration113, increase/decrease in tumor response114,115 and inconsistent associations with toxicity outcomes115-118 while others have yet to be verified in other patient cohorts. The potential mechanistic impact of these polymorphisms has also yet to be clarified. The variant, ABCB1 3435T correlated with decreased gene expression in normal human duodenum samples119 whereas another study found this variant associated with elevated ABCB1 mRNA levels in SN-38-resistant human colon tumor colonies.120

Compared to ABCB1, ABCC1, the multidrug resistance-related protein 1 (MRP1) transporter is less efficient in the secretory efflux of irinotecan and SN-38. Despite being located in the basolateral side of liver cell membrane, the ABCC1 transporter protein is more efficient in transporting SN-38 in an apical-to-basolateral transport manner than vice- versa.121 Since ABCC1 has also been detected in CPT-11-resistant human epidermoid carcinoma cells and SN-38-resistant HeLa cells,94,122 its function in irinotecan transport has

also been supported by two genetic studies linking SN-38-related pharmacokinetic and toxicity outcomes.114,123 More recently, an in vitro study confirmed that ABCC1 variant 2012G>T had no functional effect on overall ABCC1 expression in embryonic kidney cells, contradictory to clinical observations known.124

The ABCC2 protein, a canalicular multispecific organic anion transporter (cMOAT) also located in the hepatic biliary canicular membrane, transports SN-38 and its glucuronide SN-38G more efficiently than its prodrug, irinotecan.90,93 Lower metabolite uptake and clearance levels measured in ABCC2-deficient rats indicate the importance of this transporter in irinotecan metabolism. Several genetic reports have investigated at least two variants (-24C>T, -3972T>C) in multiple independent patient populations receiving irinotecan-based treatments without similar findings.116,123,125,126 Other ABCC2 variants evaluated against toxicities and response have not yet been validated elsewhere.

Amongst all the types of transporters responsible for irinotecan efflux, the breast cancer resistance protein (BCRP) or ABCG2 has been most studied. ABCG2 is located on the canalicular side and apical membrane of enterocytes, where it transports SN-38 and its glucuronide across the biliary membrane.127,128 Many have shown the ability of ABCG2 to induce irinotecan resistance in overexpressed cellular models89,129-132 or have increased activity in SN-38-resistant cells.122 Pharmacogenetic studies of mCRC and lung cancer patients have since identified five polymorphisms that are associated with irinotecan pharmacokinetics, toxicities and tumor response.114,126,133,134 In fact, the variant 421C>A was found to reduce protein expression and decrease irinotecan resistance in in vitro models135, indicating that ABCG2 is critical in the transport of irinotecan. This observation was confounding in Asian subjects who were heterozygous for this variant whereby an increase in ABCG2 activity was implied with greater SN-38G clearance levels.

The human solute carrier organic anion family member 1B1 (SLCO1B1) gene encodes for the protein SLCO1B1 that is predominately expressed on the basolateral membrane of hepatic cells and exclusively absorbs SN-38 from blood, instead of irinotecan and SN- 38G.95 Two most studied variants, 521T>C and 388A>G have been respectively linked to increases in severe neutropenia and gastrointestinal toxicity.126,136,137 These associations with clinical toxicities also appear to correlate with other pharmacokinetic studies measuring irinotecan, SN-38 concentrations and clearance.123,136,138

Even though ABCG1 and ABCC5 are not known to be involved in irinotecan transport, two variants, ABCC5 (3’UTR 1243T>C) and ABCG1 (c.974-898C>G) were associated with severe gastrointestinal toxicities in a small mCRC patient cohort.137 An increase in ABCC5 mRNA expression has also been observed in SN-38-resistant HeLa cells122, possibly implicating the role of ABCC5 in SN-38 efflux or clearance.

Table 6. Major transporter gene polymorphisms and clinical outcomes studied previously in irinotecan-treated cancer patients

Gene Variant Main findings Ref CPT-11 AUC SN-38 AUC 113 1236 C>T (G412G) SN-38 clearance Asthenia 117 2677G>A/T (A893T/S) PFS, OS 115 OS 114 Grade 4 neutropenia in T or A 116 carriers 3435C>T (I1145I) Early toxicities 115 CPT-11 plasma concentration ABCB1 134 in T carriers Diarrhea 117 Grade 3 diarrhea in TT carriers 116 Diarrhea, Combined neutropenia + diarrhea in TT 118 carriers IVS9 -44A>G SN-38 AUC 123 IVS4-25G>T OS in T carriers 114 IVS13+24T>C OS in T carriers 114 IVS14+38G>A OS in T carriers 114 IVS11 -48C>T ANC nadir levels 123 SN-38, SN-38G/SN-38 123 Nonhematological toxicity in T 1684T>C (L562L) carriers 114 ABCC1 Hematological toxicity in TT carriers 2012G>T (G671V) Response rate in T carriers 114 Hematological toxicity in CC IVS18-30C>G 114 carriers CPT-11 for T carriers 123 Response rate 125 -24C>T PFS CC carriers Response rate, PFS in TT 116 carriers ABCC2 Grade 3-4 neutropenia in CT 139 carriers SN-38G 123 APC for T carriers -3972T>C Response rate, PFS in TT 116 carriers

Grade 3 diarrhea in CC carriers 126 1249G>A (V417I) Response rate in A carriers 114 IVS23+56T>C Response rate in CC carriers 114 Nonhematological toxicity in A -1023G>A 114 carriers SN-38 AUC/dose -1774delG 138 Neutropenia (*1A/*1A) GI toxicity in CC carriers 137 ABCC5 3’UTR 1243T>C SN-38G in C carriers 139 ABCG1 c.974-898C>G GI toxicity in GG carriers 137 CPT-11 plasma concentrations 134 34G>A (V12M) Grade 3 diarrhea in A carriers 126 421C>A SN-38G in CA carriers 133 -1203_-1200 SN-38 AUC 134 ABCG2 delCTCA REC -15622C>T (Novel) Response rate in T carriers 114 Response rate in A carriers -15994G>A Non-hematological toxicity (first 114 cycle) SN-38 AUC/dose 138 Toxicity 140 CPT-11 AUC 123 SN-38 AUC, SN-38 clearance 521T>C (V174A) 136 Grade4 neutropenia Grade 4 neutropenia in C 126 carriers SN-38 AUC in C carriers 139 SN-38 AUC, SN-38 clearance SLCO1B1 in AA carriers 136 -11187G>A Grade4 neutropenia Response rate in A carriers 114 PFS ANC nadir 123 RRR, PFS, OS, TTF in A 141 carriers 388A>G (N130D) Grade3 diarrhea 136 GI toxicity in GA carriers 137 Oral mucositis in GA carriers 139 AUC, area under curve; ANC, Absolute Neutrophil Count; PFS, progression-free survival; REC, relative extent of conversion; OS, overall survival; GI, gastrointestinal; RRR, relative response rate; TTF, time to treatment failure.

Irinotecan, like many other xenobiotics, is excreted via the biliary and renal systems. Early studies conducted in rodents detected its active compound, SN-38 in bile, urine and feces while 33% and 58% of irinotecan was found in bile and urine, respectively. A majority of the dose excreted in the bile was detected in the form of inactivated SN-38.142,143 Studies in human patients revealed that over 90% of detected compounds in the bile and urine were composed of CPT-11 and its metabolites, SN-38, SN-38G and APC.42,144 High CPT- 11 and SN-38 concentrations in the bile and urine were identified analytically as inactive carboxylate forms.54,145-148 Specifically, CPT-11 accounts for 3-30% and 10-30% of the

17 dose administered found in the bile and urine respectively. SN-38 and SN-38G amounts excreted through both biliary and urinary systems are low, accounting for 1-3% of the effective dose. On the other hand, SN-38G, despite being most abundant in the blood plasma showed a wide interindividual variability across samples investigated;69,146,149 suggesting differential SN-38 metabolism rates may be implicated. (Figure 6)

Minor metabolites of irinotecan, APC and NPC that do not display any drug activity are also present in bile, urine and feces in trace concentrations.29,146,148 The cumulative low amounts of APC and SN-38G in the urine also confirm that the primary route of irinotecan metabolism is not mediated by the renal system.150,151, therefore reiterating that irinotecan elimination is conducted via biliary and fecal pathways.

Figure 6. Distribution of Irinotecan and its metabolites excreted in the bile and urine (Adapted from Slatter, JG. 2000144)

More importantly, conjugated SN-38 molecules can be subjected to deconjugation by intestinal beta-glucuronidase activity. These de-conjugated entities, observed as rebound peaks in pharmacokinetic AUC curves 30 min to 1 hour after drug administration have been presumed to undergo enterohepatic recycling.69,152-154 Enterohepatic recycling enables the prolongation of active compounds’ bioavailability as de-conjugated compounds re-enter the systemic circulation and liver from the intestine. Several factors, including genetic variations in transporters have been noted to have an impact on the efficiency of this recycling.155 Particularly, further exposure of cytotoxic SN-38 in the plasma, liver and intestine translates into increased efficacy and higher rates of toxicities such as severe neutropenia and diarrhea.

2.4 Pharmacodynamic pathways of irinotecan Current research has described the pharmacodynamic action of SN-38 to implicate cellular response from Top1-induced DNA damage. As the sole molecular target of SN-38, the Top1 protein acts in concert with the unwound single-strand DNA to form the protein-DNA complex and momentarily halts DNA replication. SN-38 specifically binds to this complex and disrupts further synthesis processes. This breakage also induces the activation of three distinct pathways - apoptosis, DNA damage repair and cell-cycle arrest. (Figure 7)

Figure 7. SN-38 mechanism of action on Top1-DNA complex and activation of pharmacodynamic pathways (Adapted from Hurley, LH et al, 2002 156 and Pharmagkb.org)

As the primary molecular target of SN-38, the cellular levels, gene expression and catalytic activity of Top1 protein are crucial in determining the efficacy of SN-38-induced cytotoxicity. So far, resistance to SN-38 has been attributed to genetic mutations, reduced quantities and activity of Top1.157-161 Further experiments describe the diminished expression of Top1 in SN-38-resistant cell lines157,158 while others have associated this resistance to copy number variations and point mutations.162-164 Accordingly, certain variants have been proposed to alter the binding affinity of SN-38 at the catalytic site of Top1, with clear evidence of their impact on the formation of Top1-cleavage complex and double-strand breakage.165

19 Even though these specific point mutations were not detected in CRC patients, intra- and inter-variation in Top1 activity and protein levels have been observed in different untreated tumor tissues.166-168 Therefore, other genetic polymorphisms may be involved in modulating Top1 expression. More importantly, colon tumor tissues were found to have a higher Top1 heterogeneous activity than other tissues evaluated (64-fold variation compared to 16-fold in cervix tumors).167 Together, these observations show that interindividual differences in Top1 can affect levels of unbound cytotoxic SN-38 but also indirectly influence SN-38-induced tumor response and toxicities.

The formation of Top1-cc-induced DNA single-strand breaks (SSB) triggers the response of DNA repair pathways, where proteins are recruited during the repair process. The presence of Top1-cc-induced SSB determines the state of DNA repair and its subsequent sensitivity to the Top1 inhibitor. For example, poly(ADP-ribose) polymerase (PARP1) is stimulated by DNA breakage events and activates its catalytic function by acting as an intermediary to destabilizes Top1-cc. At the same time, it works with tyrosyl-DNA phosphodiesterase I (Tdp1) to remove Top1 from the DNA strand. This allows base- excision repair proteins, including X-ray repair complementation group 1 (XRCC1) to access the region of damage.169-171

Genetic deficiencies of both PARP1 and TDP1 debilitate the DNA repair system and result in accumulation of unrepaired DNA breaks.172,173 PARP1 and excision repair cross- complementation group 1 (ERCC1) have also been implicated in alternative repair pathways of Top1-cc with endonucleases.174,175 Presence of XRCC1 gene also conferred a higher resistance to campthotecin while at the same time XRCC1 interacts closely with Tdp1 to activate the repair process.176 The breakdown of the DNA repair system will likely lead to unrepaired DNA lesions and eventual cell-cycle arrest and apoptosis. In addition to SSB, double-strand breaks (DSB) converted from single strand lesions occur during collision of Top1-cc and moving replication forks. DSB-mediated repairs incorporate two distinct processes – homologous recombination and non-homologous end joining. Variants in genes (ex. checkpoint proteins and tumor suppressor) from these signaling pathways can render the dysfunction of their supposed function and prevent tumor cell death or halt DNA synthesis. The overall impact of these downstream genes can affect intrinsic SN-38- induced cytotoxicity and thereby influence tumor response and accumulated toxicities.

Interestingly, further investigations with clinical samples have associated higher Top1 protein levels with longer OS and PFS in mCRC patients177 while TDP1 and ERCC1, amongst other genes, were upregulated in CRC samples.178 ERCC1 expression was also higher in responders of CPT-11 treatment but was less than that of the reference gene.179 Thus, the regulatory contributions of these genes in irinotecan-induced toxicities should be considered as well. The importance of these signaling pathways is obvious by differential gene expression detected in colorectal cancer samples. Transcriptional RNA/DNA profiling expression methods have identified the expression of certain genes with functions in DNA repair, cell cycle, apoptosis, to be higher in tumoral or SN-38-treated tissues.178,180 Other reports using proteomic methods also revealed levels of such proteins to vary in SN-38- resistant cell-lines.181,182

To date, few cohort studies have collectively studied the biological relevance of all genes in relation to CPT-11-mediated tumor response and toxicities. The majority of them have focused on few genes or only known polymorphisms. Table 7 outlines the genes and/or specific variants related to SN-38 pharmacodynamic pathways previously investigated in mCRC patients receiving irinotecan-based regimens. Unfortunately, these studies failed to clearly define the precise association of pharmacodynamics genes with predicting clinical outcomes. Moreover, the majority of these reports only studied specific putative causal variants from certain genes, not considering other variants and potentially functional SNPs present in the gene locus. Hence, in order to clarify their roles, it is necessary to perform a comprehensive study incorporating all known polymorphisms from both pharmacokinetic and pharmacodynamic genes.

Table 7. Pharmacodynamic genes associated with SN-38 pharmacokinetics, toxicity and response in irinotecan-receiving colorectal cancer patients.

Type of Cohort Genes studied Major findings Ref chemotherapy Size MLH1/MSH2, p53, irinotecan- 1313 High Top1 expression  OS, PFS TOP1, ERCC1, containing 177 mCRC pts (from immunohistochemistry) MGMT, COX2 regimens irinotecan- Higher ERCC1 expression in non- TOP1, ERCC1, 56 mCRC containing responders vs responders, no 179 GSTP1 pts regimens association with toxicities Specific Variants Type of Cohort Major findings Ref studied chemotherapy Size MLH1 -93G>A irinotecan- 982-1002 None significant with survival XRCC1 R339Q containing 177 mCRC pts outcomes GSTP1 I105V regimens

21 ERCC2 K751Q MLH1 -93G>A XRCC1 Toxicity-induced dose irinotecan- XRCC1 R339Q 633-1136 delay/reduction containing 183 GSTP1 I105V mCRC pts ERCC2 , GSTP1  irinotecan- regimens ERCC2 K751Q related Grade ≥3 toxicity HNF1A 79A>C Irinotecan 85 pts  CPT-11 AUC levels 123 ERCC2 -1989A>G, irinotecan- D711D, K751Q 520 None significant with survival containing 184 GSTP1 A114V, I105V mCRC pts outcomes or toxicities regimens XRCC1 R399Q 14 SNPs* from irinotecan- TOP1 and PARP1 diplotype  CDC45L, NFKB1, 89-107 containing neutropenia risk, TDP1 and XRCC1 185 PARP1, TDP1, mCRC pts regimens diplotype objective response TOP1, XRCC1 6 htSNPs from TOP1, None associated with ANC or PARP1, TDP1, irinotecan 85 pts 186 neutropenia risk XRCC1 irinotecan- XRCC1 R399Q 43 mCRC containing XRCC1 and ERCC2 OS, 187 ERCC2 K751Q pts regimens 10 SNPs from TOP1, irinotecan- 61 mCRC CDC45L IVS-24C>TPFS, Other TDP1, PARP1, containing 125 pts variants not significant XRCC1, CDC45L regimens ERCC1 N118N, XRCC1 R399Q irinotecan- 146 ERCC2 N312N, containing XRCC3 PFS 188 mCRC pts K751Q, regimens XRCC3 T214M irinotecan- 267 GSTP1 I105V containing Not associated with toxicities, PFS 189 mCRC pts regimens irinotecan- ERCC2 K751Q, ERCC2, GSTP1 not associated with containing 48 mCRC 190 GSTP1 1105V response, OS, PFS regimens *includes 10 haplotype-tagged SNPs (htSNPs) and 4 putative causative SNPs AUC, Area Under the Curve; ANC, Absolute Neutrophil Count; PFS, progression-free survival; OS, Overall Survival

2.5 Toxicities induced by irinotecan As alluded earlier, prolonged exposure to both CPT-11 and SN-38 is significantly associated with reduced absolute neutrophil counts (severe neutropenia) in a sigmoidal fashion.43,65,66 Early Phase I trials have identified a strong correlation between absolute neutrophil count, CPT-11 levels, clearance rates and SN-38 plasma concentrations, while similar observations were also reported for SN-38G associations with diarrhea events.69,152,191-193 Indeed, higher concentrations of circulating SN-38 may on one hand improve efficacy against rapidly dividing tumor cells but at the same time cause dose- limiting toxic events and prove to be fatal.

Hematological toxicities such as neutropenia occur due to the profound cytotoxic effect of active SN-38 on neutrophil production and indirect suppression of cellular differentiation in the bone marrow. Irinotecan-induced diarrhea events occur in two stages, acute (within 24 hours) and late (after 1-3 days). The former is caused by the cholinergic effects of the prodrug and manifests as a variety of GI symptoms including increased stool frequencies, cramps and emesis194,195 whereas the delayed onset of diarrhea results from the SN-38- mediated breakdown of intestinal mucosal lining and increased mucin secretion.67,68,196-198 SN-38 glucuronide conjugates can also be subjected to deconjugation (cleavage of the glucuronic acid moiety) into cytotoxic SN-38 by bacterial β-glucuronidase produced by microflora activity in the intestine. It has been speculated that this glucuronidase-mediated formation of SN-38 is responsible for the mucositis of intestinal membrane wall, shown by improved diarrhea conditions and histological changes in rats given antibiotics.199 Positive correlations between E. coli levels, severe diarrhea and β-glucuronidase expression have been detected in irinotecan-treated rats.198 Unfortunately, these observations were not replicated in mCRC patients who were given antibiotics (ex. neomycin) as a co-treatment with CPT-11.200

A more recent study also revealed that SN-38 formation from β-glucuronidase deconjugation may in fact be sequestered in the intestinal cell wall or bound to plasma proteins instead of freely circulating in the intestines as previously assumed.201 Thus, while it is well established that bacterial β-glucuronidases are indeed involved in the regeneration of toxic SN-38 in the gut, it is still uncertain whether this form of SN-38 induces intestinal toxicities.

Pharmacokinetics-driven studies correlating irinotecan-derived metabolites levels and toxicities have provided some insights into a biological explanation for these events. The inconsistencies reported however also highlight the inadequacies of using pharmacokinetic parameters (CPT-11 AUC, SN-38 AUC and SN-38G) as predictors of toxicity events.147,196,202 Moreover, some reports have noted the heterogeneous profile in patients with varying area under the curve (AUC) values of each metabolite,66,69,149 and might suggest that this variability was due to genetic variations in these enzymes, thereby influencing functions and modulating levels of free SN-38 to induce toxicity and response.203

Considering that neither physiopathological characteristics (ex. age, sex, weight)152,193 nor bacterial microflora levels are convincing predictors of toxicities196,198, indications of higher biliary indexes in patients with severe diarrhea69,70 and a 4.5-fold interindividual variability of Top1 enzymatic activity may be worthy of attention instead.196 Hence, attempts to explain these observed variations have led recent advances in irinotecan research toward a pharmacogenetic focus.

3. Pharmacogenomics (PGx)

Pharmacogenomics is the study of how variations in the human genome affect response to medications. Pharmacogenomics can play an important role in identifying responders and non-responders to medications, avoiding adverse events, and optimizing drug dose.

The completion of the Human Genome Project in the early 2000s has brought much anticipation and hopes of elucidating biological secrets held by the human genome. One of the major goals from this genomic breakthrough was to pave the way towards personalized treatment and individualized medicine. The past decade has since witnessed further advances in mapping of global populations with both International HapMap and 1000 Genomes Projects. (Figure 8)

Figure 8. Timeline of Pharmacogenomics over the last decade. (from di Lulio et al 2011204)

The ability to refine the prediction of drug responses was certainly welcomed in the medical field as clinicians often rely on empirical factors (disease, symptoms) to prescribe treatment and adopt a ‘wait and see’ approach towards adverse drug reactions (ADRs).205 Variations in drug efficacy and reactions have mostly been combated using close monitoring of dose-toxicity outcomes within narrow therapeutic windows. The large variability in response to treatment has been attributed to genetic, pharmacokinetic and pharmacodynamics factors.206 The advent of next-generation sequencing technology and improvements in bioinformatic applications has propelled interests into the field of pharmacogenomics, in hopes of identifying novel markers that can explain this variability.

Many pharmacogenetic markers relate to pharmacokinetic markers since the parameters are readily measurable, unlike pharmacodynamic ones. The most commonly studied genetic variants are predominantly involved in Phase I and II drug metabolism, whereby genetic alterations including polymorphisms of CYP450 enzymes have been discovered to modulate the efficacy of various drug classes (ex. anticancer, antipsychotics, anticoagulants).207 More importantly, the significance of these findings has led to the incorporation of pharmacogenomic biomarkers on US Food and Drug Administration (FDA)-approved drug labels (over 140 drugs have PGx information208) Two well-studied drugs are the anticoagulant warfarin and the antiplatelet clopidogrel used in the treatment of cardiovascular diseases; I use these drugs as examples below.

3.1 Established pharmacogenomics tests : a few examples

Pharmacogenomics is slowly evolving towards identifying potential variants that can modulate adverse effects and tumor response to efficiently incorporate genetic information into clinical decisions. Accordingly, numerous drug-gene combinations, including UGT1A- irinotecan, CYP2C9 and VKORC1-warfarin are now among the top ten most studied genes in the pharmacogenetic field.209

3.1.1 Warfarin Warfarin is one of the first drugs approved by the FDA to include genetic testing recommendations of two genes, CYP2C9 and VKORC1 (vitamin K epoxide reductase complex C1). The clinical implications of these tests are tremendous because, as an agent with an extremely narrow therapeutic window, patients receiving this drug require frequent measurements of the blood levels (INR-International Normalized Ratio) to monitor its activity. Studies have confirmed that genetic variations from CYP2C9 and VKORC1 account for approximately 40-50% of observed interindividual variability of warfarin response in the Caucasian European population.210 Furthermore, meta-analyses have uncovered underlying dose requirement differences amongst white Caucasians, African- American and Asian patients with CYP2C9*2 and CYP2C9*3 variations. Similar trends were also identified for VKORC1 polymorphisms.211-213

As of 2010, the FDA has included specific dose recommendations according to the combined genotypes of both VKORC1 -1639G>A and CYP2C9*1/*2/*3, thereby allowing clinicians to control bleeding or clotting events within the therapeutic range of INR 2.0- 3.0214 In fact, numerous randomized-controlled trials and cohort studies have started to incorporate genetic information into existing clinical algorithms. Despite certain study design differences, most findings are in accord with the overall benefits of a genotype- guided dose regimen.215-219 Notably, dose management with pharmacogenetic information was deemed to be more accurate than empirical dosing, corresponded to longer therapeutic durations and reduced adverse effects. The final decision to utilize a genotype- based dosing treatment ultimately lies within the clinicians and respective medical facilities. Several other factors such as regional differences in pharmacogenetic algorithms, starting dose amounts, patients’ age and ethnicities still need to be considered before warfarin pharmacogenetics can have full functional clinical utility.

The Clinical Pharmacogenomics Implementation Consortium (CPIC) guidelines are designed to help clinicians understand how available genetic test results should be used to optimize drug therapy, rather than whether tests should be ordered. These guidelines are peer-reviewed and published in leading journals. However, there is currently no equivalent organization in Canada. Based on evidence, current recommendations are based on combinations of CYP2C9 and VKORC1 genotypes.220 (Table 8)

Table 8. Dosing recommendations according to current CPIC guidelines based on VKORC1 and CYP2C9 genotype

CYP2C9 VKORC1

(-1639G>A) *1/*1 *1/*2 *1/*3 *2/*2 *2/*3 *3/*3

0.5-2 mg GG 5-7 mg 5-7 mg 3-4 mg 3-4 mg 3-4 mg 0.5-2 mg AG 5-7 mg 3-4 mg 3-4 mg 3-4 mg 0.5-2 mg 0.5-2 mg AA 3-4 mg 3-4 mg 0.5-2 mg 0.5-2 mg 0.5-2 mg

27 3.1.2 Clopidogrel The overall known variability of the CYP450 family of enzymes has led to the categorization of patient carriers into four groups – ultrarapid, extensive, intermediate and poor metabolizers. These groups are useful for CYP2D6, a polymorphic enzyme that converts 25% of all current drugs207,221, while distinctions for other P450 enzymes are less well defined. For example, CYP2C19 is responsible for the two-step conversion of the antiplatelet agent clopidogrel and its active metabolite. The CYP2C19*2 allele is implied to induce reduced enzymatic function, resulting in higher platelet accumulation levels and adverse drug effects.222-225 Two meta-analyses have recently confirmed that this variant was associated with greater ADRs in patients with coronary diseases.226,227 Accordingly, the FDA has also updated the drug label warnings declaring that poor metabolizers who carry two variant alleles, (*2 and *8) were likely to experience reduced anti-platelet therapy and were at a greater risk for adverse cardiovascular events.228,229

Only 12% of observed variability in clopidogrel response can be explained by CYP2C19 polymorphisms.230 Recent studies have emerged claiming that other genetic variants from drug transporters, receptors and other drug metabolizing enzymes may also contribute to this variability.230-232 These reports, however, remain inconclusive on the effects of ABCB1, PON1 (gene encoding for paroaoxonase 1) and P2RY12 (gene encoding for active site of clopidogrel P2Y12 receptor) variants and tended to be ethnic-centric hence warrant further larger-scale studies for clarifications. 233-236

In spite of the high availability of CYP2C19-specific genotyping tests in the market today, their clinical implementation has been somewhat limited.237 So far, only two reports of utilizing a CYP2C19 genotype-guided to predict antiplatelet therapy response exist: a single-patient study describes a switch in antiplatelet therapeutic agent after genotyping for CYP2C19*2 while the larger scale study of 2700 patients is still ongoing.238,239 Conclusively, the effect of pharmacogenomics on clopidogrel response appears to be multigenic and it is crucial that a greater emphasis be placed on these other genetic factors.231

In brief, CPIC recommendations for clopidogrel are categorized according to the phenotype groups - ultrarapid metabolizer (*1/*17 and *17/*17), extensive metabolizer (*1/*1), intermediate metabolizer (*1/*2, *1/*3, *2/*17) and poor metabolizer (*2/*2, *2/*3,

*3/*3). An increase in clopidogrel dose is implied for ultrarapid metabolizers while a reduction in treatment is recommended for those with intermediate and poor metabolizing status.229

3.2 Irinotecan Pharmacogenomics The successful translation of warfarin pharmacogenomics from discovery to the clinic brings the current field a step closer to achieving the goal of personalizing medicine. Unlike anticoagulant and antiplatelet therapeutic agents, the response and efficacy of chemotherapeutic drugs cannot be explained solely by germline mutations. It is widely considered that cancer pharmacogenomics is complicated by the influence of both germline and somatic mutations. On one hand, knowledge of specific germline mutations in drug-metabolizing enzymes enables better management of drug-induced toxicities and response; while somatic mutations lead to the development of novel drugs specifically targeting biomarkers (ex. targeted therapy). Drug-induced toxicity is one of the leading causes of chemotherapy discontinuation.194,240,241

The first-line mCRC drug, irinotecan, whether used as monotherapy or in combination with other drugs, has been the subject of multiple studies confirming that the promoter variant of UGT1A1 (UGT1A1*28) is associated with severe neutropenia.242 Homozygous carriers (*28/*28) tend to incur a greater risk of toxicity compared to heterozygous and wild type carriers (*1/*28, *1/*1). This variant translates to an additional TA repeat in the TATA box region and results in a 40% reduction in UGT1A1 glucuronidating function.86 These findings have since resulted in a FDA-recommended dose-reduction by one level of irinotecan for carriers of *28/*28 following a diagnostic UGT1A1 genotyping test in 2005. Due to differences in treatment schedules and impact of UGT1A1*28 at varying dose levels, current labeling have yet to specify exact dose-reduction amounts and depend entirely on the occurrence of dose-limiting toxicities.243

3.2.1 Molecular and clinical impact of UGT1A1*28

The best-studied UGT1A polymorphism so far is UGT1A1*28, which consists of a (TA)7TA insert instead of normal six TA-repeats. This insertion produces less UGT1A1 and results in reduced glucuronidation capacity of SN-38.86,244 Significant associations between severe neutropenia and carriers of TA7 repeat have been confirmed in clinical samples receiving irinotecan-based treatments.184,245-251 (Figure 9) Whereas others have suggested that this

29 polymorphism may or may not predict severe diarrhea events instead.183,245,250,252-260. Thus, the inconclusive relationship between UGT1A1*28 and diarrhea necessitates further clarifications.

promoter UGT1A1

...(TA)nTAA... 1 2 3 4 5b 3’UTR 5a 3’UTR

UGT1A1*28 UGT1A1*1

Mutant (TA)7TAA Wild type (TA)6TAA

~70% Expression Expression

Reduced SN-38 Glucuronidation Normal SN-38 Glucuronidation ~30-50%

Diarrhea Tolerance Neutropenia Figure 9. Schematic representation of UGT1A1*28 on UGT1 locus and its effect on SN-38 glucuronidation (Modified from Relling, MV et Dervieux, 2001261)

Subsequent meta-analyses confirmed that UGT1A1*28 homozygous carriers had a two- fold increase risk in neutropenia when compared to non-carriers. This dose-dependent association was observed at all doses (low, medium and high) of irinotecan administration242,249, perhaps implying that while a dose reduction was more effective at high doses (>250mg/m2) than at low doses (<150mg/m2), patients would likely benefit most from a lower starting dose.74,244 This conclusion also signifies that even though UGT1A1*28 is a good predictor of severe neutropenia occurrence, it is not the only genetic contributor or other non-genetic contributors involved.

Despite the strong evidence depicting the clinical validity of UGT1A1*28 in predicting hematological toxicity, there has been a lack of consensus in the community on the actual maximum tolerated dose (MTD) required. Accordingly, genotype-guided dose studies have since been conducted in hopes of clarifying this matter. The main findings from these reports demonstrate that the current approved dose of 180mg/m2 (in combined chemotherapy) or 150-350mg/m2 (monotherapy) is at least 15-30% higher than the MTD in patients with UGT1A1*28 genotype.262-264 Conversely, the conclusion from the three studies indicate that the current 180mg/m2 is much lower than the MTD tolerated by non- carriers (*1/*1 and *1/*28) of this variant. In addition, Innocenti et al. also found that this genotype-based dosing approach helped standardize the SN-38 AUC levels, clearly demonstrating its clinical utility in reducing adverse events.262

It is worth noting that UGT1A1*28 is present in varying frequencies amongst different ethnic groups - 2-13% in Caucasians, ≤3% in Asians and 16-19% in Africans.74 In Asians, the UGT1A1*6 has been predominantly associated with neutropenia risk instead.133,265-267 A subsequent genotype-directed study using irinotecan monotherapy at 150mg/m2 thus observed that Asian patients homozygous for UGT1A1*28 or UGT1A1*6 were associated with increased SN-38 AUC and experienced greater toxic events compared to those at lower doses (75-125mg/m2).268 Another Asian study recently confirmed the observations seen in Caucasians; the MTD in FOLFIRI-receiving mCRC patients with two or more variant alleles (*28/*28, *6/*6) was 150mg/m2 while carriers of one or no allele withstood 270-300mg/m2 of irinotecan dose respectively.269 Lastly, three meta-analyses have attempted to correlate UGT1A1*28 with response and overall survival and concluded that there were no significant associations.270-272 Overall, these findings bring new insights into the complexities of irinotecan-mediated toxicities and response – UGT1A1*28 genotyping is indeed useful but insufficient in predicting neutropenia risks, especially at low doses. It is probable that other variants may contribute in modulating diarrhea events and response rates in irinotecan-based regimens.

3.2.2 Other UGT1A variants

Currently, much focus in irinotecan pharmacogenomics has been on elucidating associations of UGT1A1*28 and clinical outcomes. The negative relationship between this variant and severe neutropenia risks has been established while other associations remain less clear. Based on prevailing evidence that SN-38 can be glucuronidated by other members of UGT1A, UGT1A6, UGT1A7 and UGT1A9, presented previously in Table 4, recent investigations have thus directed their attention on elucidating the impact of other UGT1A variants on irinotecan-induced toxicities. For instance, UGT1A1*93 has been evaluated against irinotecan-related clinical outcomes in multiple reports and displays an increasing trend against neutropenia or hematological risks. 80,184,273-275 The validity of this trend should be carefully considered because these studies were performed using different irinotecan-containing regimens, small cohort sizes and there were discrepancies in measuring toxicities. UGT1A9*22 has also been studied in numerous cohorts, including one in Asian patients. Non-carriers of this variant (UGT1A9*1/*1 -118dT9/9) appear to be associated with an increase risk of diarrhea and hematologic toxicities. However, some of these studies have associated this variant with an overall clinical outcome (hematologic

31 toxicity) and do not enable specific conclusions to be drawn from these results. Finally, these findings have yet to be validated in other independent cohorts. (Table 9)

Table 9. UGT1A gene polymorphisms and associations with clinical outcomes studied previously in irinotecan-treated cancer patients

Gene Variant Main findings Ref G71R (*6) Response and Gr3-4 Neutropeniaa 276 Hematological toxicity 273 Gr3-4 Neutropenia in GA carriers 274 Hematological toxicity in GA carriers; UGT1A1 c.-3156G>A (*93) 275 Response rate in AA carriers 184 Gr4 Neutropenia risk 80 Gr3-4 Neutropenia in A carriers c.-3279T>G (*60) Hematological toxicity in GG carriers 275 S7A Gr3-4 Neutropenia in G carriers 80 UGT1A6 T181A Gr3-4 Neutropenia in G carriers 80 R184S Gr3-4 Neutropenia in C carriers 80 W208R Gr3-4 Neutropenia in C carriers 80 Gr3-4 Diarrhea, Hematologic toxicity 252 in carriers (*3/*3)

N129K,R131K,W208R Gr3-4 Hematologic toxicity in (*3/*3) UGT1A7 275 (*3) Gr3-4 Diarrhea, response in 277 N129K, R131K (*2) carriers (*2/*3)

Early-stage diarrhea and 278 thrombocytopenia in *2 carriers Gr3-4 Diarrhea, non-hematologic 252 toxicity in *1 carriers Gr3-4 Toxicity, response in *1 277 -118dT (*1) [9/9] carriers -118dT (*22) [9/10] Overall Hematologic toxicityb in *1 275 UGT1A9 carriers Gr3-4 Diarrhea in *1 carriersa 276 c.-1212 Gr3-4 Neutropenia 80 c.-688 Gr3-4 Neutropenia 80 c.-440 Gr3-4 Neutropenia 80 a Study was performed in Asian patients with non-small-cell lung cancer tumors b Overall hematologic toxicity was measured throughout entire course of treatment

An in-depth exploration of polymorphisms in the UGT1A locus is warranted so as to identify causal single nucleotide polymorphisms (SNPs) that can predict clinical toxicities. In order to refine the role of UGT1A SNPs in irinotecan metabolism, several groups, including ours, have undertaken a haplotype-grouped association approach to assess the combined effect of UGT1A variants on irinotecan-induced clinical outcomes.80,266,275 This method not only increases statistical power over a single UGT1A1*28 test, but also allows other UGT1A variants to be considered in modulating toxicity risks. However, each study

had the tendency to combine different variants in the haplotype groups, therefore making it challenging to make cross-comparisons. Notably, our group previously investigated the same haplotype group (Haplotype II) reported by Cecchin et al. and found a significant association with higher severe neutropenia risk, which was not observed in the original cohort.80,275 This confounding result is likely due to the differences in which the toxicity measurements were categorized. Hence, in order to achieve more robust and interpretable results, pharmacogenetic studies should strive towards validating their findings in other independent cohorts with similar methodologies and regimens.

3.2.3 UGT1A1*28 genotyping test

Following the FDA-mandated recommendations for a dose-reduction in patients who are homozygous for UGT1A1*28 allele, an Invader® UGT1A1 Molecular Assay diagnostic test (Hologic Inc, USA) was subsequently developed to meet the needs of clinicians who wish to determine UGT1A1*28 genotype. The assay detects the TA-insertion by measuring the fluorescent probe signals that result from cleavage of unpaired oligonucleotides on each DNA strand. The ratios of the signals are then calculated to determine the heterozygousity of the SNP. When compared to the traditional direct DNA sequencing methods, it was 99%-100% successful in calling genotypes (analytical validity). The Invader test has several advantages over the older techniques including an extremely quick turnaround time (4-5 hours), high-throughput screening ability and 98% reproducibility amongst different medical centers.279

Despite the availability of such diagnostic tests, the clinical implementation rate has been lukewarm. A recent review by Gillis et al summarizes the differences in recommendations of routine UGT1A1 genotyping testing by various international regulatory bodies, including the Dutch Pharmacogenetics Working Group (DPWG), US Evaluation of Genomic Applications in Practice and Prevention (EGAPP) Working Group, National Comprehensive Cancer Network (NCCN) and the FDA. So far, only DPWG has officially suggested a 30% reduction in starting dose (>250mg/m2) for patients homozygous for the *28 allele.280 Other groups have not made a stance regarding its clinical application. The EGAPP Working Group in particular remained unconvinced by the clinical utility of this test, viewing the reduction in dose to alleviate toxicity is at the expense of achieving therapeutic response.281 The application of pharmacogenetic testing appears to be

33 variable across different countries. In France, an average of 3300 tests/year were ordered during the period of 2009-2011.282 In the United States, major hospitals such as Mayo Clinic and St-Jude’s Children’s Hospital have also initiated a pharmacogenetics-based testing for UGT1A1*28.283,284 In Canada, evidence of clinical implementation has been scarce whereby the most recent recommendation of the Eastern Canadian Colorectal Cancer Consensus Conference, dated 2007, does not recommend routine UGT1A1 testing.285

The lack of precise recommendations has resulted in these tests to be poorly adopted by clinicians to accurately interpret test results. In addition, inconclusive UGT1A1*28 associations with diarrhea and response are still unresolved and have to be managed clinically.286 As of now, the current UGT1A1*28 genotyping test, while in limited use, remains the best diagnostic tool to predict severe neutropenia risk in irinotecan-receiving patients and help identify under-dosed patients (*1/*1 and *1/*28) who would benefit from a dose increase.

3.3 Current state of PGx studies

Researchers in the field face tremendous challenges in trying to understand and make meaningful conclusions of their findings. For example, most of the PGx studies have been performed in small (<100), individual cohorts and are not powered to detect significantly reliable variants. The majority of findings from candidate gene-based studies are almost never replicated in other independent cohorts, likely due to differences in treatment dose, patients’ ethnicity and tumor types.287 Indeed, the same group of researchers attempted to validate their earlier findings185 in another cohort but were unable to observe the same phenotype.186 The necessity of replication cohort gives rise to other challenges such as maintaining similar study designs, measurement of same clinical outcomes, population ethnic makeup and statistical analysis.

As genetic sequencing costs become more affordable with the advent of new technologies, large-scale studies are being undertaken to maximize resources and time. The biological explanation for a single genetic variant to be the sole cause of adverse drug effects/phenotype is unlikely, given what we know about epistasis and systems biology. Hence, with improvements in next-generation sequencing (NGS), recent investigations are

utilizing a genome-wide association study (GWAS) approach to identify novel genes that can explain drug response or cancer phenotype.

In the subfield of irinotecan pharmacogenomics, only three studies so far have employed this method in cancer patients receiving irinotecan-based regimens.288-290 While these studies were generally small-sized (n=139, 101 and 168 respectively), they demonstrate difficulties of applying these markers into clinical use. A study investigating 440,094 SNPs in 101 non-small cell lung cancer (NSCLC) patients identified C8orf34 (Chromosome 8 open reading frame 34), FLJ41856 (similar to carcinoembryonic antigen-related cell adhesion molecule CEACAM5) and PLCB1 (Phospholipase C, beta 1) to have genome- wide significance (p<1x10-5) against grade 3 diarrhea occurrence after validation in a secondary cohort.289 These genes do not have known correlations with irinotecan so far and are unlikely to have any direct impact on clinical endpoints. Furthermore, numerous studies often fail to correct for multiple testing and lead to false positive markers that cannot be replicated elsewhere. The rate of false-positive varianta uncovered also tends to increase due to lack of stringent statistical criteria set in GWAS studies.

Ultimately, the goal of extending cohort-specific results to benefit larger patient populations requires the process of rigorous replication in studies with sample sizes of thousands in order to determine any true association outcome. Despite the low prevalence (1.2%) of replicated variants in the literature, replication remains a gold standard in the pharmacogenomics field.291 The feasibility of conducting these replication studies becomes contentious, as it is often logistically expensive and difficult to maintain larger clinical cohorts. Even by ensuring all specifications are identical across studies, other confounding factors such as penetrance, effect size and frequencies of outcome cannot be controlled.

To account for these unpredicted genetic characteristics, researchers have begun to redefine their candidate-gene approach to include entire gene loci; in hopes of identifying strongly linked functional variants. Bioinformatic tools and publicly available databases such as Haploview and International HapMap project have thus facilitated a more comprehensive approach for genetic markers. Using a haplotype-based strategy provides a cost-effective way to explore a larger genetic distance without needing to genotype all SNPs. SNPs that are in the same haplotype-block can be represented (‘tagged’) by a single SNP which is then genotyped and assessed.

In accordance with the genetic structure, SNPs from different loci in strong linkage disequilibrium with each other are inherited together and expected to have identical associations to the same phenotype.292 Following genetic association analyses, retrospective investigations of potential functional variants (ex. synonymous, non- synonymous) linked to tag SNPs can thus render new insights and lead to further validations. More importantly, this expansion of gene coverage may also identify novel ‘hotspot’ gene regions associated with the outcome.

The growing acceptance of pharmacogenetic testing in the clinical setting has also sparked discussions about costs involved and whether they are beneficial in the long run. A genotyping test with high analytical validity, clinical utility and validity can reduce treatment cost and time to both clinicians and patients. One of the ultimate goals of pharmacogenetic testing is to avoid ineffective treatment, including adverse drug effects and unnecessary poor outcomes. The promise of genetic testing prior to drug administration allows identification of patients who would be susceptible to incur drug- induced toxicities or those who would respond poorly to the drug. Indeed, the decreasing costs of these tests have persuaded certain regulatory bodies to allow for their implementation and establish specific guidelines.293

However, some of the underlying problems that clinicians and patients both face include ethical and social issues.294 The responsibilities of paying for genotyping tests appear to be inconsistent. Current genotyping kits can cost between US$80-$5000295 and tend not to be reimbursed by healthcare payers. The majority of insurance companies and hospitals preferred that reimbursement of drugs be limited if tests showed high efficacy while others questioned the practicality and usefulness of these tests.296 Even though there appears to be a superior advantage of performing these tests, the major issue is the lack of financial support from healthcare providers and health insurance companies who deem these tests as experimental, investigative and not medically necessary. The lack of clinical evidence demonstrating a direct health-related outcome and test result further discourages healthcare payers to embrace these additional costs.297 In Canada, genetic testing is usually performed at the discretion of hospitals, which are dependent on provincial funding. The biomarker tests for colorectal cancer, KRAS and EGFR are the only tests funded by both the province (Alberta) and pharmaceutical industries.298

Independent reports on the cost-effectiveness analysis of an UGT1A1-genotyping test (US$100-$250) have been encouraging whereby hypothetical patients who had undergone such tests and genotyped with *28/*28, received a 25% decrease in irinotecan dose and experienced an improved quality-adjusted life expectancy. A cost-savings of US$272/patient was derived for every 10,000 patients, taking into consideration cost of test, physicians’ fees, hospitalization due to severe neutropenia and chemotherapy treatment costs at varying doses. Surprisingly, approximately 84 cases of severe neutropenia and 4.4 deaths would have been also avoided.299 Thus, for 10,000 homozygous patients, this would translate to a total of US$2.7 millions in savings.

Other groups have utilized a prophylactic approach by using granulocyte-colony stimulating factor (G-CSF) treatments to prevent occurrence of neutropenia. Pichereau et al, have demonstrated the cost-beneficial use of genotyping to avoid one case of febrile neutropenia was approximately €1000 in cost to the hospital and for every 1000 patients, 96 febrile neutropenia incidents could be avoided. In the long run, from the patients’ perspective, hospitalization costs due to febrile neutropenia of €1500-€4000 may be saved.300 Instead of prophylactic interventions, a 20% dose reduction in a high dose regimen of irinotecan (300-300mg/m2 once per 3 weeks) was also determined to be more cost-effective and convenient compared to a weekly low dose (125mg/m2), particularly in Caucasian and African populations. In addition, there was a cost savings of US$110-$460 by those patients who benefited from this dose reduction while avoiding 1.6-2.8 severe neutropenia events per 100 patients.301 Hence, there is a pressing need for future genotyping tests not only to be clinically usable but also cost-effective to be fully incorporated in any drug management strategy. More importantly, a genotyping test that can account for the multigenic contribution towards drug response would not only allow the simultaneous assessment of important genetic markers but also minimize the need to perform multiple single-SNP genotyping tests. A well-validated multimarker test will be indeed cost-effective for healthcare providers and patients alike, improving prediction of overall clinical outcomes.

While genotyping tests can certainly be useful at the dose-management level, they can also help pharmaceutical drug companies avoid costly drug recalls and failed Phase 3-4 clinical trials. Genotyping can identify subgroups of patient populations who would be at

37 risk for severe toxicities and allow companies to better focus their target group, capitalizing on a smaller and reliable market share. There is also an added potential for failed drugs to be reanalyzed for pharmacogenetic biomarkers, which can speed-up the arduous drug development and regulatory processes.299

The growing availability of these tests also brings forward potential genetic discrimination by insurance companies. The Office of the Private Commissioner of Canada has called on the healthpayer industry to avoid requesting for genetic test results from its members, thus emphasizing the importance of genetic privacy protection.302

Ultimately, some hurdles still need to be overcome before pharmacogenetic testing can be adopted into clinical practices. In addition to assessing the robustness and clinical validity of these tests, its clinical utility needs to be clarified in a way where its benefits clearly outweigh the disadvantages. Factors that need to be carefully considered include an increase in quality of life (due to absence of debilitating toxicities), overall treatment costs, the ethical implications of tests, clinicians’ training in interpreting results and infrastructure of diagnostic services locally available.

4. Hypothesis & Objectives

Germline variability can be caused by a myriad of structural changes in the genome (ex. SNPs, INDEL (insertion/deletion) mutations, copy number variations, tandem repeats etc). These changes can alter the regulation of protein-coding genes and thus influence their intrinsic effect on protein function. In general, variants from genes involved in the pharmacokinetics (PK) and pharmacodynamics (PD) pathways of irinotecan metabolism, transport and action have the propensity to affect toxicities and drug response, respectively.

So far, the best-characterized genetic variant linked to irinotecan-induced severe neutropenia is the variant UGT1A1*28, as described previously. Regulatory agencies such as the US FDA and the DPWG recommend testing for this variation prior to irinotecan- based treatment and to reduce dosage by one level in homozygous carriers. Furthermore, an approved genetic test is also available (ex. Invader® UGT1A1 Molecular Assay). Meta- analyses indicate a dose-dependent effect of this allele on risk of severe diarrhea but no evidence support an impact on tumor response and survival.256,271

Therefore, a haplotype analysis of SNPs in UGT1 locus would likely better refine gene interactions involved in irinotecan metabolism. Other additional pathways related to drug metabolism, transport and action should also be considered in defining neutropenia and diarrhea risks. The overall goal here is to identify novel pharmacogenetic markers that can improve existing prediction of irinotecan-induced toxicities.

4.1 Identify UGT1 markers to better predict risk of severe toxicity

Since multiple members of the UGT1A family of proteins inactivate the active metabolite of irinotecan SN-38 76,84,88, the primary objective was to focus on identifying additional UGT1 markers to help better predict the risk of severe toxicity. Few studies have been conducted to include variants in other UGT1 such as UGT1A7 and UGT1A9 but none has performed a systematic survey in relation to severe toxicity. Furthermore, haplotype associations with pharmacogenomic markers are increasingly favored over the traditional single-SNP analysis since it allows for the further refinement of genetic etiology of drug-induced clinical outcomes and improved prediction of response and dose-management. Our group,

39 amongst others have clearly shown the advantages of this approach with specific combinations of functional SNPs from UGT1 locus over that of a single SNP. Hence, we argue that the use of combined UGT1 variants to determine genetic associations can add value to the current single-SNP genotyping analysis methods, thereby enhancing dosing strategies for irinotecan-based treatments as well as increase drug efficacy.

The first objective was thus to perform a comprehensive screening of the UGT1 locus including all known functional variants and a haplotype-tagging SNP strategy. This initial screening was conducted in a prospective discovery cohort of metastatic colorectal cancer Caucasian patients receiving irinotecan-based regimens, all recruited in Canada. Positive findings were then replicated in an independent validation cohort of patients of European ancestry from Italy. Overall this study improves our understanding and appreciation of UGT1 genetic markers as important modulators of severe neutropenia risk induced by irinotecan-based chemotherapeutic treatment.

4.2 Identify novel markers to help predict toxicity induced by irinotecan in drug transport pathways

We then sought to identify markers in other pathways relevant to irinotecan exposure (PK) and transport. Some reports in the literature have identified certain variants from ABC and SLC transporters genes that predict the risk of severe toxicity. Most of these studies were from smaller patient populations and remain un-validated in independent patient cohorts. Previously reported variants will also be included in an attempt to validate these findings in our cohorts. The primary endpoints were severe toxicities, neutropenia and diarrhea, analyzed separately. Furthermore, we endeavored to overcome previous limitations by utilizing a prospective assessment of GI toxicity through a daily journal self-completed by patients with follow-up by the research nurses. Positive markers were then validated in the independent validation cohort.

40 5. Methodologies & Approaches

This section outlines the methods and approaches that were used in the two original scientific papers presented in the core of this thesis. The relevance of using genotyping technologies complemented by bioinformatic analysis emphasizes the characteristics of clinical pharmacogenomic studies as well as the broader underlying concepts applied in translational medicine.

Treatment with irinotecan frequently results in severe neutropenia and diarrhea that can seriously impact the course of treatment and patients’ quality of life. Pharmacogenomic tailoring of irinotecan-based chemotherapy has been the subject of several investigations but with limited data regarding ABC and SLC transporter genes whereas the UGT1 has not been systematically covered in terms of genetic variability. Our general approach was to initially profile 167 mCRC Canadian patients treated with FOLFIRI-based regimens and replicate findings in an independent cohort of 250 Italian patients.

5.1 Study cohorts:

5.1.1 Discovery cohort The focus of this study was to scrutinize the association between germline mutations and adverse drug reactions in a patient population and subsequently validate these positive markers in an independent, ethnically similar cohort. The discovery cohort consists of 167 Caucasian patients diagnosed with advanced metastatic colorectal cancer recruited from three medical centers in eastern Canada over a 10-year period of 2003 to 2012: Hotel- Dieu de Québec in Québec City, Quebec; Hotel-Dieu de Lévis in Lévis, Québec; and The Ottawa Hospital in Ottawa, Ontario. Each ethics committee locally approved the clinical study protocol and all patients provided a signed written informed consent for participation in the genetic study. Patients aged 18 to 90 years old were deemed eligible based on several clinical criteria: presence of metastatic colorectal cancer confirmed by histological results, life expectancy of 3 months or greater and a performance status ≤2 which assesses the progress of the disease and its effect on daily living abilities. (Eastern Cooperative Oncology Group (ECOG).

All patients received a FOLIRI regimen consisting of 180mg/m2 irinotecan administered intravenously for two hours on the first day, followed by a 400mg/m2 bolus dose of 5-FU and a continuous infusion of 2400mg/m2 5-FU with 200mg/m2 leucovorin for 46 hours. This regimen was repeated every two weeks. 75 patients also received a co-treatment of either bevacizumab (Avastin®), a monoclonal antibody that inhibits vascular endothelial growth factor A (VEGF-A), an experimental drug, sorafenib (a tyrosine kinase receptor inhibitor), or a placebo.

5.1.2 Validation cohort A second patient cohort of 250 Caucasians was recruited from 2002 to 2005 by 13 clinical institutions in Northeast Italy. Similar eligible criteria were applied to the validation cohort: patients between ages 18 to 75 years with a life expectancy greater than 3 months, metastatic tumors verified histologically and a ECOG performance status between 0 and 2. The same irinotecan-based treatment regimen was used. The specific characteristics of both cohorts are presented in Table 10.

Table 10. Demographic and Clinical characteristics studied in Discovery and Validation cohorts

Discovery cohort Validation cohort (Canadians) (Italians)a Characteristics N (%) N (%) Overall 167 250 Sex Male 110 162 Female 57 88 Median age (years) 61.5 60.6 Primary tumor site Colon 122 (73.1) 179 (71.6) Rectum 42 (25.1) 71 (28.4) Unknown 3 (1.8) - Treatment regimen FOLFIRI 167 250 Co-treatment bevacizumab 69 (41.3) - Other drugs 6 (3.6) - Toxicity Neutropenia (Grade 3-4) 28 (16.8) 33 (13.0) Diarrhea (Grade 3-4) 24 (14.4) 21 (8.4) aDemographic characteristics have been published previously (Cecchin E et al, 2009275and Toffoli G et al, 2006303).

5.2 Evaluation of toxicity outcomes The clinical endpoints were severe neutropenia and diarrhea (Grade 3-4), indicative of hematological and gastrointestinal toxicities, respectively.

5.2.1 Severe neutropenia The occurrence of hematological toxicity was defined by the measurement of laboratory parameters, including neutrophil count per liter. Samples were collected prior to the start of each chemotherapy cycle and/or when the treatment was given. The severity of neutropenia was assessed according to the guide provided by the National Cancer Institute Common Toxicity Criteria for Adverse Events version 3.0 (NCI-CTCAE v3.0).

More precisely, patients with neutrophil count of 1.5-1.7 x 109/L were categorized under Grade 1, between 1.0 – 1.5 x 109/L in Grade 2, between 0.5 – 1.0 x 109/L in Grade 3 and <0.5 x 109/L in Grade 4. Patients with normal neutrophil count, i.e. >1.7 x 109/L were considered as Grade 0. The same criteria were applied for the Italian cohort. (Table 11)

5.2.2 Severe diarrhea Gastrointestinal toxic events were evaluated based on the number of stools per day over a baseline and categorized in the same manner using the NCI-CTCAE v3.0 criteria. Grade 1 was defined by <4 stools increase per day, Grade 2 with an increase of 4-6 stools per day, Grade 3 with an increase of ≥7 stools per day. Grade 4 was deemed as a life-threatening condition, requiring hospitalization. Grade 0 patients were defined as those with no increase in stool counts per day. (Table 11) A potential discrepancy from aforementioned studies was that these cohorts were retrospective in nature and thus limited in carefully addressing GI toxicities. In our study, all patients from the Canadian cohort completed a daily diary of GI toxic events during the first fortnight of each cycle where incidences and episodes of nausea, vomiting and diarrhea were recorded; except for 1 patient who died before toxicities could be assessed and another who did not complete the GI diary. In the Italian cohort, the same NCI-CTCAE v3.0 criterion was used to grade the GI toxicities. Specific questions about nausea, vomiting, mucositis, diarrhea, malaise and appetite were verbally asked during each treatment cycle. Overall, the most severe toxicity recorded was used for data analysis.

43 Table 11. NCI-CTACE v3.0 criteria for severe neutropenia and diarrhea toxicities

Grading criteria for toxicity and adverse events Grading of severe neutropenia events (neutrophil) count Grade 0 > 1.7 x 109/L Grade 1 1.5 – 1.7 x 109/L Grade 2 1.0-1.5 x 109/L Grade 3 0.5-1.0 x 109/L Grade 4 < 0.5 x 109/L Grading of gastrointestinal toxicity (stools increase/day) Grade 0 No increase Grade 1 < 4 Grade 2 4-6 Grade 3 ≥7 Grade 4 life-threatening, hospitalization required Taken from NCI-CTACE v3.0, last updated in 2006

5.2.3 Bilirubin levels Direct bilirubin levels measures bilirubin-glucuronide conjugates while indirect bilirubin consists of unconjugated bilirubin. Total bilirubin levels refer to the sum of both direct and indirect bilirubin amounts. In the discovery cohort, we measured the total, direct and indirect bilirubin levels with units (mg/dL), prior to the start of irinotecan treatment. The upper limit of normal range (ULN) was determined as 21mg/dL in the Canadian cohort.

5.3 Selection of candidate genes and single nucleotide polymorphisms (SNPs) A comprehensive Pubmed literature search identifying candidate genes associated with CPT-11 and SN-38 metabolism and transporters was initially performed. A total of eight candidate genes were chosen and are outlined in Table 12.

Table 12. Genes involved in the metabolism and transport of SN-38 studied in the discovery cohort

Gene ID Gene Name Gene Function UGT1 UDP-glucuronosyltransferase Phase II Metabolite Detoxification ATP-binding Cassette transporter, associated ABCB1 Cellular Transport with MDR1 protein ATP-binding Cassette transporter, associated ABCC1 Cellular Transport with MRP ATP-binding Cassette transporter, associated ABCC2 Cellular Transport with MRP ATP-binding Cassette transporter, sub-family ABCC5 Cellular Transport C gene, member 5 ATP-binding Cassette transporter, sub-family ABCG1 Cellular Transport G gene, member 1 ATP-binding Cassette transporter, sub-family ABCG2 G gene, member 2, breast cancer resistance Cellular Transport protein Solute carrier organic anion transporter Uptake of endogenous compounds SLCO1B1 family, member 1B1 and elimination of xenobiotics

For this study, we employed a haplotype-tagging SNP strategy during the discovery phase. SNPs in a haplotype block, i.e. a group of SNPs in the same gene locus in strong linkage disequilibrium (LD) with one another is represented by a ‘tag SNP’. This htSNP (haplotype-tagging SNP) is often used as a proxy during genotyping as it allows for the simultaneous genetic associations of multiple SNPs to be performed without the need to genotype every single variant in the block; thereby translating into cost and time savings.

Each candidate gene was individually queried on the International HapMap Project database release #28 (http://hapmap.ncbi.nlm.nih.gov/) (NCBI), which is a global initiative maintained by the National Center for Biotechnology Information to genotype and catalog SNPs from individuals of different ethnic groups. The objective was to iterate all SNPs from gene regions spanning between ±5kbp within the loci in the Caucasian European (CEU) population, so that all genetic variants located in the regulatory (promoter and 3’UTR), exonic and intronic regions were accounted for.

45 Subsequently, these SNPs were further investigated on the haplotype analysis software, HaploView 4.2 (Broad Institute) to determine their haplotype block associations and select htSNPs using the “Tagger” tool. The criterion for SNPs to be considered included a minor allelic frequency (MAF) of 0.05 or greater and a Hardy-Weinberg Equilibrium (HWE) p- value cut-off of 0.05. This was to ensure that all SNPs scrutinized in our study were significant in our patient populations and that they did not deviate from HWE due to other predisposed factors. The minimum correlation coefficient between the htSNP and its associated SNP is r2 ≥0.8. Positive markers and functional SNPs previously reported in the literature to correlate with irinotecan-induced clinical outcomes were also assessed in conjunction with the htSNPs. Overall, 238 htSNPs in 8 genes were genotyped using PCR (polymerase chain reaction) amplifications of genomic DNA in the discovery phase.

5.4 DNA Sample preparation and Genotyping Genomic DNA was extracted from patients’ blood samples using a DNA extraction Kit (QIAamp DNA Blood Mini Kit) and their respective concentrations were quantified by Nanodrop Spectrophotometer (Thermofisher) and diluted to the same concentrations (15ng/µl). A total of 167 samples were separated into two 96-well plates with each plate containing 5% random duplicates, 1 positive and 1 negative control. The samples were subsequently genotyped using Sequenom iPLEX matrix-assisted laser desorption/ionization time-of-flight MS technology at the CHU de Québec Genomics Research Center. PCR assays and associated extension reactions were designed using SpectroDESIGNER software (Sequenom®). The same workflow was used for replication of positive markers in the Italian population of 250 samples.

5.5 Statistical analysis

5.5.1 Hardy-Weinberg equilibrium

Hardy-Weinberg equilibrium (HWE) is defined by the constant inheritance of genetic variation from one generation to the next in a population, assuming natural factors such as mutations, natural selection, genetic drift and gene flow are not present. This principle allows for the detection of any changes in allelic or genotypic frequencies within a population by calculation using a mathematical equation p2 + 2pq + q2 = 1; where p and q represent major and minor alleles respectively and 2pq is the heterozygous genotype. In this study, deviations from HWE were determined by a Chi-Square test using a genetic association software PLINK and the p-value threshold used was p<0.05. The significance

of this test not only identifies which SNPs are not in HWE in our patient populations but also acts as a quality control to predict genotyping errors. We observed <1% genotyping errors in our samples.

5.5.2 Genetic Associations For each clinical outcome, patients with toxicity Grade 0,1 and 2 were categorized as less severe while those with Grade 3-4 were considered as more severe. To determine genetic associations between each htSNP and clinical outcome, a Fisher’s exact test was performed to compare the odds of exposure to the variant in predicting a more severe outcome. In general, the formula used to calculate the odds ratio -

Number of patients with more severe outcome (carriers of htSNP) Number of patients with less severe outcome (Non-carriers of htSNP)

In order to account for the allelic, dominant and recessive models of genetic expression, alleles of each variant were categorized accordingly and association tests were performed using the same formula described above. Multivariate analyses consisting of demographic factors (ex. age) that are known to correlate with the clinical outcomes in our cohort were carried out with logistic regression models on SPSS® software v21 (IBM, New York, NY). This method allows for the consideration of other environmental factors that may influence association of SNPs with a particular clinical outcome. For associations with qualitative variables (ex. total bilirubin levels), a Student’s t test was utilized to compare the mean values of patients carrying htSNP with that of non-carriers. All htSNPs that were deemed positive (p<0.05) or had a trend (p<0.10) by univariate and multivariate analyses were replicated in validation cohort.

Haplotype and co-occurrence analyses were performed to assess the genetic associations between combined positive variants and clinical outcomes. For analysis of SNPs in the same gene locus, pairwise measures of LD between polymorphisms were determined with an Expectation-Maximization (EM) algorithm on Haploview 4.2 software. LD is defined by the probability that closely linked variants in the same gene were most likely inherited together rather than by chance. The extent of disequilibrium, expressed by r2, denotes the correlation coefficient between two SNPs and ranges from 0 to 1; where r2 = 1 indicates a complete linkage. Even though the D’ measure, defined by the ratio of observed frequency

47 over expected maximum frequency, is widely used, it does not consider allelic frequencies nor rare SNPs.

One major step in haplotype analysis is statistical phasing of haplotypes in unrelated individuals. For a 2-marker haplotype with unique alleles, there are numerous combinations that can be inferred for each patient sample. For example, Patient 1 is heterozygous for both SNP A (A/T) and SNP B (T/C). When considering a haplotype of SNP A and B, this patient may have the genotype AT, TT, AC or TC. One way to determine the best combinations is to reconstruct individual haplotypes based on existing haplotype frequencies and likelihood ratios. Hence, all haplotypes were inferred using a Bayesian-based phasing software (Phase v2.1.1)304,305 Subsequently, a two-tailed Fisher’s exact test was employed to determine the association between each haplotype group and the clinical outcome. A similar approach (co-occurrence analysis) was used to calculate genetic associations in markers from different gene locus and dependent variable without the need to phase individual haplotypes. All statistical analyses were performed using add- in software, XLSTAT (AddinSoft Inc, Brooklyn, NY). False-discovery rates (q values) were calculated using the R QVALUE package (http://genomics.princeton.edu /storeylab/qvalue/).

5.6 Functional studies and in-silico analyses

5.6.1 3’ Rapid Amplification of cDNA Ends (RACE) To assess if positive markers were comprised in mRNA transcripts, we sought to perform a 3’ Rapid Amplification of cDNA Ends (RACE) experiment using normal hepatic tissue samples. Total RNA extracted from two liver samples previously genotyped as either major or minor homozygous for a SNP significantly associated with a clinical outcome was converted into cDNA, amplified using primers from the kit and sequenced on an ABI PRISM 3730XL DNA Analyzer (Applied Biosystems, Foster City, CA). Sequences were analyzed using the Staden package software version 2.0.0b9 (https://sourceforge.net/projects/staden) and compared with the GenBank reference sequence NG_002601.

5.6.2 Resequencing of UGT1 locus In an attempt to identify potential novel functional SNPs associated with positive markers, such as the germline marker located in the 3’ flanking region of the UGT1 gene, a resequencing approach was applied. PCR amplifications on germline DNA of homozygous liver tissue sample in the promoter regions, the first exons of UGT1A1, UGT1A7 and UGT1A9, the common exons 2, 3, 4, 5a and 5b, the intron-exon boundaries, and the 3'- UTR regions of exon 5a and exon 5b were performed and sequenced. Comparisons with the reference sequence NG_002601 were performed using Staden package and NovoSNP, a software used for identification of novel variants from sequencing data.

5.6.3 In-silico predictive bioinformatics analysis A series of in-silico bioinformatics predictive analyses was performed to assess the effect of positive variants on gene function by modulating regulatory factors of transcription. Depending on the location of the variant on the gene locus, appropriate databases and bioinformatics tools were used to identify potential elements that can interact with the positive SNPs. The UCSC genome browser was first used to visualize the gene structure defined by GenBank and RefSeq and regulatory tracks (ex. ChIP-seq Chromatin Immunoprecipitation sequencing, DNA methylation status, histone modifications) from ENCODE (Encyclopedia of DNA Elements) and other sources were added to identify potential regulatory functions. Further analyses were also undertaken to investigate the relationship between variants and regulatory factors such as transcription factors, microRNA, splicing factor binding sites, large intergenic non-coding RNA (lincRNA). A list of databases used for this purpose is shown in Table 13.

Table 13. In silico bioinformatic databases and tools used to evaluate the potential functionality of novel markers

Type of Name of regulatory Website database element Mapper2 http://genome.ufl.edu/mapper/mapper-main rSNPs sTRAP http://epicenter.immunbio.mpg.de/cgi-bin/chromos/sTrap.cgi pfSNP http://pfs.nus.edu.sg/ Transcription F-SNP http://compbio.cs.queensu.ca/F-SNP/ Factors (TF) SNP Function http://brainarray.mbni.med.umich.edu/Brainarray/Database/SearchSNP/snpfunc.aspx Portal HaploRegv2 http://www.broadinstitute.org/mammals/haploreg/haploreg.php FuncPred http://snpinfo.niehs.nih.gov/snpinfo/snpfunc.htm

49 ESEfinder 3.0 http://rulai.cshl.edu/cgi-bin/tools/ESE3/esefinder.cgi?process=home Exonic and ESRSearch http://esrsearch.tau.ac.il/ Intronic SFMap http://sfmap.technion.ac.il/index.html Splicing SpliceAid http://193.206.120.249/splicing_tissue.html Elements (ESE) Human Splicing Finder v3.0 http://www.umd.be/HSF3/HSF.html Large lincRNA, intergenic non- lincPoly coding RNA human 210.46.85.180:8080/LincSNP (lincRNA) database DNAse HaploRegv2 http://www.broadinstitute.org/mammals/haploreg/haploreg.php Hypersensitivity rSNPbase I experiments http://rsnp.psych.ac.cn/ http://epicenter.immunbio.mpg.de/services/chromos/ Chromatin State ChroMos microRNA http://epicenter.ie-freiburg.mpg.de/services/microsniper/index.php microSNiper binding site Histone protein RegulomeDB http://regulomedb.org/ modifications rSNPbase http://rsnp.psych.ac.cn/

Chapter II Identification of a novel genetic marker in UGT1A locus with better tolerance against severe neutropenia in metastatic colorectal cancer patients

51 Résumé

Le marqueur UGT1A1*28 est associé au le risque de neutropénie induite par le traitement du cancer colorectal métastatique (mCRC) avec l’irinotécan, créant une exposition accrue au métabolite actif, SN-38. Notre objectif était de découvrir de nouveaux marqueurs des loci UGT1 par une stratégie d’haplotype/SNP-étiquette couplé à la spectrométrie de masse. Les profils de 167 patients canadiens atteints de mCRC sous traitement FOLFIRI (à base d’irinotécan) ont été examinés et les marqueurs significatifs ont été validés dans une cohorte indépendante de 250 patients italiens. Nous avons découvert, dans la région intergénique en aval du gène UGT1, un nouveau marqueur (rs11563250G) associé à un moindre risque de neutropénie sévère (rapport des cotes (RC)=0.21, p=0.043 chez les Canadiens; RC=0.27, p=0.036 chez les Italiens; et RC=0.31, p=0.001 pour les deux cohortes combinées). Par ailleurs, pour l’haplotype défini par ce marqueur, rs11563250G combiné à UGT1A1*1 (rs8175347 TA6), le RC était de 0.17 (p=0.0004). Un test génétique évaluant ces marqueurs permettrait d’identifier les patients susceptibles de bénéficier d’une augmentation des doses d’irinotécan.

A novel UGT1 marker associated with better tolerance against irinotecan-induced severe neutropenia in metastatic colorectal cancer patients

Sylvia Chen*1, Isabelle Laverdiere*1, Alan Tourancheau1, Derek Jonker2, Félix Couture3, Erika Cecchin4, Lyne Villeneuve1, Mario Harvey1, Michael H. Court5, Federico Innocenti6, Giuseppe Toffoli4, Eric Lévesque1,3, and Chantal Guillemette1†

1Pharmacogenomics Laboratory, Centre Hospitalier Universitaire de Québec Research Center and Faculty of Pharmacy, Laval University, Québec, Canada 2Division of Medical Oncology, Department of Medicine, Ottawa Hospital, University of Ottawa, Ontario, Canada 3Centre Hospitalier Universitaire de Québec Research Center and Faculty of Medicine, Laval University, Québec, Canada 4Division of Experimental and Clinical Pharmacology, Department of Molecular Biology and Translational Research, National Cancer Institute and Cancer for Molecular Biomedicine, Aviano, Italy 5Individualized Medicine Program, Department of Veterinary Clinical Sciences, Washington State University College of Veterinary Medicine, Pullman, Washington 6Division of Pharmacotherapy & Experimental Therapeutics, Eshelman School of Pharmacy, University of North Carolina, Chapel Hill, North Carolina *These authors contributed equally.

Running head: Biomarker of irinotecan tolerance

†Correspondence: Chantal Guillemette, PhD; Canada Research Chair in Pharmacogenomics, Centre Hospitalier Universitaire de Québec Research Center, R4720, 2705 Boul. Laurier, Québec, Canada, G1V 4G2. Tel. (418) 654-2296 email: [email protected]

Keywords: Colorectal cancer, Genetic variants, Irinotecan, Neutropenia, UGT1

Pharmacogenomics Journal; advance online publication March 17, 2015

Impact Factor: 5.513

53 AUTHOR CONTRIBUTIONS

Designed Research and supervision: Chantal Guillemette

Patients’ recruitment, clinical data and provision of human livers: Derek Jonker, Félix Couture, Erica Cecchin, Giuseppe Toffoli, Eric Lévesque and Michael H. Court

Performed Research: Sylvia Chen, Isabelle Laverdiere, Lyne Villeneuve and Mario Harvey

Analyzed Data: Sylvia Chen, Isabelle Laverdiere, Alan Tourancheau, Lyne Villeneuve, Mario Harvey, Eric Lévesque and Chantal Guillemette

Wrote Manuscript: Sylvia Chen, Isabelle Laverdiere, Eric Lévesque and Chantal Guillemette

Critical revision of the manuscript for important intellectual content: All authors

Abbreviations: UGT, UDP-glucuronosyltransferase; PCR, polymerase chain reaction; UTR, untranscribed region; SNP, single nucleotide polymorphism; ECOG, Eastern Cooperative Oncology Group; LD, linkage disequilibrium; OR, Odds Ratio; RACE, RNA ligase-mediated rapid amplification of cDNA end; htSNP, haplotype-tagging SNP.

Abstract The risk of severe irinotecan-induced neutropenia has been shown to be related to the UGT1 variant UGT1A1*28, which increases exposure to the potent metabolite SN-38. Our goal was to identify a novel UGT1 marker(s) using 28 haplotype-tagged single nucleotide polymorphisms genotyped by mass spectrometry. By characterizing the UGT1 sequence from a cohort of 167 Canadian metastatic colorectal cancer (mCRC) patients and a validation cohort of 250 Italian mCRC patients, we found rs11563250G, located in the intergenic region downstream of UGT1, to be significantly associated with reduced risk of severe neutropenia (odds ratio (OR)=0.21; p=0.043 and OR=0.27; p=0.036, respectively, and OR=0.31 when combined; p=0.001), which remained significant upon correction for multiple testing in the combined cohort (p=0.041). For the two-marker haplotype rs11563250G and UGT1A1*1 (rs8175347 TA6), the OR was of 0.17 (p=0.0004). Genetic testing of this marker may identify patients who might benefit from increased irinotecan dosing.

55 INTRODUCTION Irinotecan is a chemotherapeutic agent used in combination with folinic acid (leucovorin), and 5-fluorouracil as a first-line treatment of metastatic colorectal cancer (mCRC) in a regimen denoted FOLFIRI. Irinotecan exerts its cytotoxicity by inhibiting topoisomerase I during DNA replication through its active metabolite SN-38. As an anticancer agent with a narrow therapeutic index, dose management of irinotecan is necessary to minimize associated toxicities, i.e., neutropenia and diarrhea. Hepatic and extrahepatic phase II UDP-glucuronosyltransferase (UGT) drug-metabolizing enzymes, i.e., UGT1A1, UGT1A7, and UGT1A9, convert SN-38 into an inactive form SN-38 glucuronide (SN-38G).1 Neutropenia is the most significant dose-limiting toxicity associated with irinotecan treatment and is directly related to the plasma SN-38 concentration, which, in addition to UGT1 activity, depends on biliary excretion and the activities of several transporter genes.2-6 In patients, germline information concerning the UGT1 pathway may help to optimize the chemotherapeutic agent dose and type of therapy.7-9 Several reports of distinct irinotecan-reaction profiles associated with common UGT1 variants have highlighted the relevance of characterizing UGT1 variants.2, 4, 5, 10, 11 Most tested biomarkers for UGT1 variants have been shown to be useful tools to identify patients more likely to experience severe neutropenia related to irinotecan-containing regimens. In particular, the variant UGT1A1*28, which contains seven, instead of six TA repeats in its promoter A(TA)nTAA region, is associated with significantly decreased glucuronidation activity, which results in reduced SN-38 clearance.12 This reduced clearance is consistently associated, in a dose-dependent manner, with an increased risk of severe neutropenia in patients homozygous for this allele.2, 13-15 More recently, it was established that UGT1 haplotypes, e.g., combination of variants in UGT1A1, UGT1A6, UGT1A7, and UGT1A9, are also associated with an increased risk of severe neutropenia.11, 16, 17 These findings demonstrate that in addition to the well-established UGT1A1 rs8175347 TATA box promoter variant, other UGT1 variants might be involved in irinotecan-induced toxicities. Through haplotyping, our group recently found that the presence of the variant rs8330 in the 3'–untranslated region (3’UTR) of the UGT1 locus improves the ability to predict the risk of severe irinotecan-induced neutropenia, which suggests variance in this region common to all UGT1A transcripts, may also participate in the toxic effect of irinotecan.16 Furthermore, the clinical relevance of rs8330 was recently demonstrated for acetaminophen-induced acute liver failure associated with modification of the exon 5a/5b splice variants mRNA ratio.18 These findings, therefore, support the

contribution of variants across UGT1 in irinotecan pharmacogenetics and the putative role of the 3'-region of the gene in the overall glucuronidation capacity and subsequent risk of severe toxicity. The study reported herein aimed to examine the genetic association across the UGT1 locus with the risk of developing severe neutropenia in mCRC patients treated with FOLFIRI-based regimens using a haplotype-tagging SNP (htSNP) strategy to maximize gene coverage and discover novel markers. We initially studied a prospective cohort of mCRC patients recruited in Canada (n = 167) treated with FOLFIRI-based regimens (discovery cohort) and replicated the main findings of that study in a similar, but independent, cohort of 250 Italian patients (validation cohort). The most significant and replicated finding of this work is the discovery that the variant allele rs11563250G in the 3'- flanking region of UGT1 is associated with a substantially reduced risk of irinotecan- induced neutropenia in both populations. This new marker may help refine our ability to predict the risk of severe neutropenia, optimize irinotecan dosage, and personalize treatment to improve clinical outcomes.

57 MATERIALS AND METHODS Patient cohorts and liver samples. One hundred and sixty-seven Eastern Canadian mCRC patients were recruited and then begun on a FOLFIRI regimen. All patients received a FOLFIRI regimen that included 180 mg irinotecan/m2 intravenously with 69 patients also receiving a co-treatment, i.e., bevacizumab, an experimental drug, or a placebo. Specific treatment modalities and eligibility criteria have been published.16 Participants provided written consent for genetic analysis. Each local research ethics committee approved the research protocol. Table 1 summarizes patient demographics (age and sex) and clinical information (treatment, toxicity, tumor site). The replication cohort consisted of 250 Northeastern Italian mCRC patients that are receiving a FOLFIRI treatment of the same dose and delivery method as described.9, 11 The severity of neutropenia was evaluated according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0. We studied 48 livers to assess the relationship between severe irinotecan-induced neutropenia and UGT1 genotypes (see below for the genotyping procedure). UGT1A1 expression levels and rates of bilirubin and SN-38 glucuronidation for these liver samples have been reported.19, 20 Genetic analysis, 3'-RACE, and re-sequencing. Single nucleotide polymorphisms (SNPs) were identified in UGT1A from the CEU population using International HapMap Project information (http://hapmap.ncbi.nlm.nih.gov). To maximize coverage, we included the ±5 kb flanking UGT1. htSNPs were found by Haploview v4.2 (Broad Institute, Cambridge, MA, USA). Markers that had previously been associated with irinotecan- related outcomes in the literature but not listed in the HapMap Project were also included. SNPs that could not be sequenced as the result of poor primer design or because they were located in duplicated regions were replaced with tagged SNPs in complete LD (r2 = 1.0). All selected htSNPs (n = 28) (Supplementary Table S1) were genotyped using an iPLEX matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (Sequenom, San Diego, CA). Negative controls and a 5% random sample duplicate population were used to ensure the robustness of the assay and genotyping reproducibility. A 3'-RACE study (Life Technologies, Burlington, ON, CA) was performed as described by the manufacturer using total RNA extracted from two liver samples that had been genotyped as homozygous for the rs11563250 variant (one variant was AA and the other was GG). PCR amplicons were subsequently sequenced on an ABI PRISM 3730XL DNA Analyzer (Applied Biosystems, Foster City, CA). Sequences were analyzed using the

Staden package software version 2.0.0b9 (https://sourceforge.net/projects/staden) and compared with the GenBank reference sequence NG_002601. Re-sequencing was performed using germline DNA from the homozygous rs11563250G carrier by PCR amplification of the promoter regions and first exons of UGT1A1, UGT1A7, and UGT1A9, the common exons 2, 3, 4, 5a and 5b, the intron-exon boundaries, and the 3'-UTR regions of exon 5a and exon 5b. Statistical analysis. All genetic association tests were assessed by logistic regression analysis using SPSS 21.0 software (SPSS Inc, Chicago, IL) with independent analyses to account for allelic, dominant, and recessive modes of transmission. ORs were adjusted for age and co-medication, as in our previous study.16, 21 Genetic variants with p < 0.10 were investigated in the replication cohort. The statistically significant threshold was fixed at p ≤ 0.05. Deviation from Hardy-Weinberg Equilibrium values was calculated using the PLINK v1.07 whole genome association analysis toolset.22 Haplotypes and pairwise LDs were inferred using Phase v2.1.1and Haploview v4.2, respectively.23, 24 To account for the false discovery rate associated with the combined cohort analysis, a Bonferroni correction was applied using R software (version 2.15.3). The statistical difference in bilirubin levels between carriers and non-carriers for a given genetic variation was assessed using the Student’s t-test. For studies using the human liver samples, analyses were performed by XLSTAT software version 2014.3.04 (AddinSoft Inc, Brooklyn, NY) using a one-way analysis of variance (ANOVA) with haplotypes, bilirubin-G, SN-38-G, and UGT1A1 expression as variables. A post-hoc Dunnett’s Test was applied with the haplotype 6A set as the reference.

59 RESULTS A total of 28 UGT1 htSNPs from the discovery cohort composed of 167 Canadians with mCRC (Table 1) were first assessed for their association with grade 3–4 severe neutropenia. This set of SNPs across the UGT1 gene had never been genotyped in this population. A linkage disequilibrium (LD) map representing these markers, along with the well-established UGT1A1 rs8175347 TATA box promoter variant, is depicted in Figure 1.

Eight novel markers, rs4663326, rs17863787, rs7583278, rs28899187, rs3771342, rs2302538, rs6717546 and rs11563250, were significantly associated with severe neutropenia in the Canadian derivation cohort. Four htSNPs located in the common region (minor alleles rs3771342A and rs2302538G; r2=0.62) or downstream of the 3'-UTR of UGT1 (rs6717546A and rs11563250G; r2=0.28) were significantly associated with a lesser risk of severe neutropenia (OR = 0.22–0.46; p < 0.05). In contrast, carriers of rs17863787G and rs7583278T (r2=0.75) in the first exon of UGT1A6 had an increased risk of severe neutropenia (OR = 2.04 and 1.86; p = 0.019 and 0.039, respectively) (Table 2). In addition, all Canadian carriers of the minor allele for rs4663326G in the first exon UGT1A6 and rs28899187A in the first exon UGT1A4 (r2=0.52) did not experience severe neutropenia (p < 0.02).

Positive markers were subsequently genotyped in the independent Italian cohort (n = 250 mCRC cases; Table 1). As observed in the discovery cohort, the minor 3'-flanking variant rs11563250G was also associated with a decreased occurrence of severe neutropenia in the replication cohort (OR = 0.27; CI 95% 0.08–0.91, p = 0.036). For the combined cohort, the OR value (0.31; p = 0.001) remained significant upon adjustment for multiple testing (p = 0.041). All other associations between htSNPs and risk of severe neutropenia found in the Canadian discovery cohort were not replicated in the Italian validation cohort (p>0.05; Table 3).

We then sought to evaluate the co-occurrence of the UGT1A1*28 risk allele (rs8175347; -54_-53insTA) and rs11563250 in a two-marker haplotype analysis. Compared with Canadian and Italian patients carrying the reference haplotype I denoted 6A in Figure 2 (UGT1A1*1, containing the reference six TA repeat in its promoter and the major rs11563250A allele), those carrying the UGT1A1*28 risk allele [a seven TA repeat in the promoter] and rs1156250A (haplotype II, 7A) tended to be at greater risk for severe neutropenia (OR = 1.44, p = 0.092; Figure 2). In contrast, haplotype III (6G) (UGT1A1*1

and the rs11563250G corresponding to alleles individually associated with a lower risk of neutropenia) was associated with a significantly decreased risk of severe neutropenia in both populations whether their risk was analyzed individually (Canadian cohort (14.1%), OR = 0.13, p = 0.021; Italian cohort (14.6%), OR = 0.21, p = 0.016) or in combination (OR = 0.17, p = 0.0004). In a second series of haplotype analysis, we tested a three-marker haplotype incorporating the rs8330 marker previously shown to be associated with reduced risk of severe neutropenia in haplotype analyses.16 Results revealed a comparable association with OR = 0.13 (95% CI= 0.02 - 0.69; p = 0.026) associated with haplotype 6GC (UGT1A1*1, the rs11563250G and the rs8330C) for the Canadian cohort (data not shown). Therefore, the observed protective effect cannot be attributed to this 3'-UTR variation and because there is no LD between rs11563250 and rs8330 in Canadian and Italian populations (r2 < 0.10). Consistent with the protective effect of rs11563250G, carriers of the G allele (AG + GG) exhibited a 17.5% decrease (p = 0.004) in total bilirubin compared with carriers of the AA genotype (Figure 3a), suggesting that the rs11563250G carriers had elevated UGT1A1 activity. When assessing only UGT1A1*1 carriers, total bilirubin was reduced in carriers of rs11563250G (p = 0.024; Figure 3b). In line with these data, the two-marker haplotype analysis also revealed a trend towards lower levels of unconjugated bilirubin for mCRC p = 0.059) in the patients for whom data were available (data not shown). Using a bank of 48 human livers previously studied for UGT1A1 expression levels and rates of bilirubin and SN-38 glucuronidation,19, 20 we also assessed the relationship with HI, HII, and HIII. No significant differences were found for these three endpoints in carriers of HI compared with HIII carriers. As expected, compared with carriers of HI, those with the UGT1A1*28 allele (HII) expressed less UGT1A1 protein and had decreased formation of bilirubin-G and SN- 38G (p = 0.0006, 0.003, and 0.0008, respectively; data not shown). With the aim of identifying additional markers across the UGT1 locus linked to rs11563250G (not in linkage with the UGT1A1 TATA box variant rs8175347; r2=0.35), we genotyped eight additional SNPs found in the International HapMap project and 1000 genomes from the CEU population in strong linkage disequilibrium (r2 ≥ 0.80), that corresponds to a LD block of SNPs is distributed over a 13.6-kb area. Four of these intergenic variants rs17862880, rs28900409, rs10199882, and rs7586006, all located between UGT1 and HEATR7B1 (MROH2A) are almost in complete LD with rs11563250 (r2

61 ≥ 0.92–1.00) but only the p-value for rs11563250 is significant in the replication cohort (Figure 4). For instance, the rs17826880 variant, which is closest to the 3'-UTR of the UGT1 gene and tightly linked to rs11563250 (r2 = 1.0 for Canadians and r2 = 0.87 for Italians,), is also associated with a reduced risk of neutropenia for the Canadian cohort (OR = 0.22, p = 0.044) but did not reach significance for the replication cohort (OR = 0.39; p = 0.076). Variants in high LD with rs11563250 are all located in the intergenic non-coding DNA region between UGT1 and HEATR7b1 (Figure 4). To determine whether these sequences are present in hepatic UGT1A transcripts, a series of 3'-RACE experiments were conducted using human liver mRNA from homozygous carriers of the rs11563250 A and G alleles. The results indicate that the region encompassing these variants, including the variant rs17826880 closest to the 3'-UTR of exon 5a (rs17826880 is 110 bp downstream this 3'-UTR), is not present in liver mRNAs encoded by the UGT1 locus (data not shown). Furthermore, sequencing of the first exons, promoter regions of the most active UGT1As towards SN-38 (1A1, 1A7, and 1A9),21 the common exons, intron-exon junctions, and the 3'-UTRs of exons 5a and 5b of germline DNA from a homozygous carrier of rs11563250G, did not identify new variants associated with this protective marker.

DISCUSSION Pharmacogenetic tailoring of irinotecan-based chemotherapy has been the subject of several investigations over the last several years. Despite these efforts, the most reliable predictor of severe neutropenia remains associated with the UGT1A1*28/28 promoter genotype related to decreased SN-38 glucuronidation, greater exposure to SN-38, and an approximately two-fold greater risk of toxicity, which helps to identify patients who would benefit from a reduced irinotecan dose.2 We report herein a novel marker (rs11563250G) located in the 3'-flanking region of UGT1 associated with a better tolerance against severe irinotecan-induced neutropenia in two independent cohorts of mCRC patients, which should allow certain patients to benefit from an increased irinotecan dose and potentially improved benefit from irinotecan-based therapy. Previous studies advised that patients with a favorable genetic profile might benefit from an increased irinotecan dose to maximize antitumor activity.7-9 Toffoli and collaborators demonstrated that the irinotecan recommended dose of 180 mg /m2 in FOLFIRI regimens is considerably less than the dose that can be tolerated by non-carriers of UGT1A1*28/*28.9 This dose-escalation study established that a dose of 370 or 310 mg/m2 of irinotecan can be safely administered every 2 weeks for mCRC patients, with the *1/*1 or *1/*28 genotype, respectively. In line with this study, Marcuello and colleagues showed that the recommended FOLFIRI dose of 180 mg/m2 irinotecan is ~two-fold lower than the dose that can be tolerated by patients with UGT1A1*1/*1 or *1/*28 (390 mg/m2 and 340 mg/m2, respectively), as part of FOLFIRI treatment.8 More recently, Innocenti and collaborators studied the effect of administering irinotecan once every 3 weeks and demonstrated that the predicted maximum tolerated dose is significantly superior for patients with UGT1A1*1/*1 or *1/*28, i.e., 470 mg/m2 and 390 mg/m2 respectively, compared with the recommended dose of 350 mg/m2.7 According to our findings, the

UGT1A1*1 (rs8175347 TA6)/rs11563250G two-marker haplotype significantly improves prediction of a decreased risk for severe neutropenia compared with an assessment of UGT1A1*1 alone, suggesting that patients carrying these two markers are currently being under dosed, further reinforcing the clinical relevance of our findings.7 Consistent with the protective effect of the rs11563250G allele, we found a reduction in total plasma bilirubin, suggesting that this polymorphism, or SNPs in high LD, might be associated with an enhanced glucuronidation capacity. However, we could not assess the relationship between genotype and exposure to active SN-38 and inactive SN-38 glucuronide which, therefore, represents a limitation of the study. However, a genome-

63 wide meta-analysis by Johnson and colleagues also reported an association between rs11563250 and total plasma bilirubin levels (p = 3.7  10–8) in three combined cohorts (n = 9,464).25 In line with a protective effect conferred by this allele, a second study found an interaction between a haplotype comprising the intergenic marker rs7586006 tightly linked to rs11563250 and frequent NSAID use that significantly decreased colorectal cancer risk.26 Our data indicate that the rs11563250 variant is located outside the UGT1 3'-UTR region, distant from the first exons of UGT1. In support, our 3'-RACE experiments did not capture 3'-mRNA sequences encompassing rs11563250. The rs11563250 variant is tightly linked to eight other variants (r2 ≥ 0.80). No additional variants located within the UGT1 gene were in significant LD with rs11563250 and sequencing of rs11563250G homozygous carriers did not allow for the identification of additional neutropenia-associated variants. It can be inferred that rs11563250 possesses a functional role in regulating UGT1 expression. However, no significant variation in UGT1A1 expression or bilirubin and SN-38 glucuronidation rates could be detected in our liver samples that contain the rs11563250G allele, possibly a consequence of the small sample size. The rs11563250 variant is within an intergenic zone that corresponds to an open chromatin region at chromosome 2 according to ENCODE project data. It is possible, therefore, that this variant, or closely related variants, display allelic differences in regulatory activity of the UGT1 locus, potentially affecting chromatin folding, epigenetic factors, or is part of cis-regulatory and complex long-range promoter-enhancer communication regulating transcription of UGT1. Such intergenic SNPs, co-localizing with transcription factors that bind to these sequences and act as positive regulators of gene expression, have been reported.27-29 The biological mechanism behind the association of rs11563250G with decreased risk of neutropenia also deserves additional in-depth functional investigations as it may potentially affect the overall conjugation capacity of UGT1A-targeted substrates. In support, a pharmacokinetic study has been undertaken in our laboratory 30 that has found that organ transplant patients receiving mycophenolate mofetil and carriers of rs11563250G present an overall greater mycophenolic acid glucuronide concentration after 2 h of drug administration compared with the levels found in non-carriers, suggesting that increased glucuronidation rates are associated with this allele as mycophenolic acid is a substrate of UGT1A enzymes (unpublished data). In conclusion, we report the identification of an intergenic variant, rs11563250G located in the 3'-flanking region of UGT1, which is associated with better tolerance to irinotecan-

induced neutropenia. rs11563250 genotyping may be clinically useful to identify patients who would better tolerate a greater irinotecan dose, especially those individuals with the UGT1A1*1/*1 genotype. We base these conclusions on our study of the two independent cohorts of the 406 FOLFIRI-treated mCRC patients and reiterate the need to validate the presence of a biomarker(s) in independent populations to obtain clinically meaningful findings with translational potential. The strengths of our study include the substantial plausibility of an association(s) given UGT1A enzymes involvement in irinotecan disposition, the extensive coverage of UGT1 htSNPs, replication in an independent population, and correction for multiple testing. Further investigations related to the function of this intergenic variant are required to decipher the molecular mechanism underlying its protective effect and potential role in affecting the metabolism of other substrates of UGT1A enzymes. We conclude that this relatively common variation (12%) influences irinotecan toxicity and should be considered to refine pharmacogenetic testing. The need to genotype the two markers rs11563250 and rs8175347 in the UGT1A1 promoter variant is our major conclusion as it may have clinical consequences in irinotecan-dosing management, especially in patients who are carriers of rs11563250G and might, therefore, tolerate, and likely benefit, from greater irinotecan dosing to maximize antitumor activity without increasing toxicity. Our study represents another step towards personalized and more precise FOLFIRI-related treatment of mCRC patients.

65 Acknowledgements: The authors thank all participants in this study and the research nurses from Québec and Ottawa hospitals for their contributions. The Canadian Institutes of Health Research (C.G.; grant MOP-42392) and the Canada Research Chair Program (C.G.) supported this work. S.C. was a recipient of the studentship award ‘Fonds de l’Enseignement et de la Recherche’ from Faculty of Pharmacy at Laval University. I.L. is a recipient of a Canadian Institutes of Health Research Frederick Banting and Charles Best studentship award and a Graduate Scholarship for clinician-scientist from FRQ-S. M.H.C. was supported by the United States National Institutes of Health, National Institute of General Medical Sciences [Grant GM-102130] and the William R. Jones Endowment at Washington State University. E.L. is a recipient of a Canadian Institutes of Health Research clinician-scientist phase II award and Prostate Cancer Canada Rising Star Award (RS2013-55). C.G. holds a Canada Research Chair in Pharmacogenomics (Tier I).

Conflict of interest disclosure: The authors have no conflicts of interest to disclose.

References

1. Gagne JF, Montminy V, Belanger P, Journault K, Gaucher G, Guillemette C. Common human UGT1A polymorphisms and the altered metabolism of irinotecan active metabolite 7-ethyl-10-hydroxycamptothecin (SN-38). Mol Pharmacol 2002; 62: 608-617. 2. Barbarino JM, Haidar CE, Klein TE, Altman RB. PharmGKB summary: very important pharmacogene information for UGT1A1. Pharmacogenet Genomics 2014; 24: 177-183. 3. de Jong FA, Kitzen JJ, de Bruijn P, Verweij J, Loos WJ. Hepatic transport, metabolism and biliary excretion of irinotecan in a cancer patient with an external bile drain. Cancer Biol Ther 2006; 5: 1105-1110. 4. De Mattia E, Toffoli G, Polesel J, D'Andrea M, Corona G, Zagonel V, et al. Pharmacogenetics of ABC and SLC transporters in metastatic colorectal cancer patients receiving first-line FOLFIRI treatment. Pharmacogenet Genomics 2013; 23: 549-557. 5. Marsh S, Hoskins JM. Irinotecan pharmacogenomics. Pharmacogenomics 2010; 11: 1003-1010. 6. Smith NF, Figg WD, Sparreboom A. Pharmacogenetics of irinotecan metabolism and transport: an update. Toxicol In Vitro 2006; 20: 163-175. 7. Innocenti F, Schilsky RL, Ramirez J, Janisch L, Undevia S, House LK, et al. Dose- finding and pharmacokinetic study to optimize the dosing of irinotecan according to the UGT1A1 genotype of patients with cancer. J Clin Oncol 2014; 32: 2328-2334. 8. Marcuello E, Paez D, Pare L, Salazar J, Sebio A, del Rio E, et al. A genotype- directed phase I-IV dose-finding study of irinotecan in combination with fluorouracil/leucovorin as first-line treatment in advanced colorectal cancer. Br J Cancer 2011; 105: 53-57. 9. Toffoli G, Cecchin E, Gasparini G, D'Andrea M, Azzarello G, Basso U, et al. Genotype-driven phase I study of irinotecan administered in combination with fluorouracil/leucovorin in patients with metastatic colorectal cancer. J Clin Oncol 2010; 28: 866-871. 10. Guillemette C, Levesque E, Rouleau M. Pharmacogenomics of human uridine diphospho-glucuronosyltransferases and clinical implications. Clin Pharmacol Ther 2014; 96: 324-339. 11. Cecchin E, Innocenti F, D'Andrea M, Corona G, De Mattia E, Biason P, et al. Predictive role of the UGT1A1, UGT1A7, and UGT1A9 genetic variants and their haplotypes on the outcome of metastatic colorectal cancer patients treated with fluorouracil, leucovorin, and irinotecan. J Clin Oncol 2009; 27: 2457-2465. 12. 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 U S A 1998; 95: 8170-8174. 13. Hu ZY, Yu Q, Pei Q, Guo C. Dose-dependent association between UGT1A1*28 genotype and irinotecan-induced neutropenia: low doses also increase risk. Clin Cancer Res 2010; 16: 3832-3842. 14. Hu ZY, Yu Q, Zhao YS. Dose-dependent association between UGT1A1*28 polymorphism and irinotecan-induced diarrhoea: a meta-analysis. Eur J Cancer 2010; 46: 1856-1865. 15. Dias MM, McKinnon RA, Sorich MJ. Impact of the UGT1A1*28 allele on response to irinotecan: a systematic review and meta-analysis. Pharmacogenomics 2012; 13: 889-899.

67 16. Levesque E, Belanger AS, Harvey M, Couture F, Jonker D, Innocenti F, et al. Refining the UGT1A haplotype associated with irinotecan-induced hematological toxicity in metastatic colorectal cancer patients treated with 5- fluorouracil/irinotecan-based regimens. J Pharmacol Exp Ther 2013; 345: 95-101. 17. Innocenti F, Liu W, Chen P, Desai AA, Das S, Ratain MJ. Haplotypes of variants in the UDP-glucuronosyltransferase1A9 and 1A1 genes. Pharmacogenet Genomics 2005; 15: 295-301. 18. Court MH, Freytsis M, Wang X, Peter I, Guillemette C, Hazarika S, et al. The UDP- Glucuoronosyltransferase (UGT) 1A polymorphism c.2042C>G (rs8330) is associated with increased human liver acetaminophen glucuronidation, increased UGT1A Exon 5a/5b Splice Variant mRNA ratio, and decreased risk of unintentional acetaminophen-induced acute liver failure. J Pharmacol Exp Ther 2013; 345: 297- 307. 19. Girard H, Thibaudeau J, Court MH, Fortier LC, Villeneuve L, Caron P, et al. UGT1A1 polymorphisms are important determinants of dietary carcinogen detoxification in the liver. Hepatology 2005; 42: 448-457. 20. Girard H, Villeneuve L, Court MH, Fortier LC, Caron P, Hao Q, 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. 21. Toffoli G, Cecchin E, Corona G, Russo A, Buonadonna A, D'Andrea M, 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. 22. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 2007; 81: 559-575. 23. Stephens M, Donnelly P. A comparison of bayesian methods for haplotype reconstruction from population genotype data. Am J Hum Genet 2003; 73: 1162- 1169. 24. Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet 2001; 68: 978-989. 25. Johnson AD, Kavousi M, Smith AV, Chen MH, Dehghan A, Aspelund T, et al. Genome-wide association meta-analysis for total serum bilirubin levels. Hum Mol Genet 2009; 18: 2700-2710. 26. Angstadt AY, Hartman TJ, Lesko SM, Muscat JE, Zhu J, Gallagher CJ, et al. The effect of UGT1A and UGT2B polymorphisms on colorectal cancer risk: haplotype associations and gene-environment interactions. Genes Chromosomes Cancer 2014; 53: 454-466. 27. Fogarty MP, Panhuis TM, Vadlamudi S, Buchkovich ML, Mohlke KL. Allele-specific transcriptional activity at type 2 diabetes-associated single nucleotide polymorphisms in regions of pancreatic islet open chromatin at the JAZF1 locus. Diabetes 2013; 62: 1756-1762. 28. Sebio A, Paez D, Salazar J, Berenguer-Llergo A, Pare-Brunet L, Lasa A, et al. Intergenic polymorphisms in the amphiregulin gene region as biomarkers in metastatic colorectal cancer patients treated with anti-EGFR plus irinotecan. Pharmacogenomics J 2014; 14: 256-262. 29. Stadhouders R, Aktuna S, Thongjuea S, Aghajanirefah A, Pourfarzad F, van Ijcken W, et al. HBS1L-MYB intergenic variants modulate fetal hemoglobin via long-range MYB enhancers. J Clin Invest 2014; 124: 1699-1710.

30. Levesque E, Delage R, Benoit-Biancamano MO, Caron P, Bernard O, Couture F, et al. The impact of UGT1A8, UGT1A9, and UGT2B7 genetic polymorphisms on the pharmacokinetic profile of mycophenolic acid after a single oral dose in healthy volunteers. Clin Pharmacol Ther 2007; 81: 392-400.

69 Table 1 Demographic and clinical characteristics of the study populations.

Canadian cohort Italian Cohorta (Discovery cohort) (Validation cohort) Characteristics N % N % Total number 167 250 Gender (male/female) 110/57 162/88 Median age (years) 61.5 60.6 Primary tumor site Colon 122 73.1 179 71.6 Rectum 42 25.1 71 28.4 Unknown 3 1.8 - Regimen FOLFIRI 167 250 Co-treatment: bevacizumab 69 41.3 - Other drug 6 3.6 - Toxicity Diarrhea (grade 3–4) 24 14.4 21 8.4 Neutropenia (grade 3–4) 28 16.8 33 13.0 aDemographic characteristics have been reported.9, 10

70 Table 2 UGT1 htSNPs significantly associated with severe neutropenia (grade 3–4) in the Canadian cohort.

Neutropenia Neutropenia Regions htSNPs Alleles OR1 (95% CI) P Genotypes OR1 (95% CI) 0–2 3–4 0–2 3–4 P C (Ref) 267 54 CC (Ref) 131 27 rs1597942 - 0.5953 - 0.5923 T 5 0 CT/TT 5 0 1A8 G (Ref) 214 44 GG (Ref) 85 17 rs1377460 0.84 (0.40–1.78) 0.649 1.02 (0.43–2.41) 0.973 A 58 10 GA/AA 51 10 C (Ref) 145 34 CC (Ref) 42 13 rs18238032 0.67(0.37–1.23) 0.199 0.48 (0.21–1.12) 0.088 T 131 20 CT/TT 96 14 A (Ref) 190 31 AA (Ref) 70 10 rs2741034 1.63 (0.89–2.98) 0.111 1.72 (0.73–4.06) 0.213 G 84 23 AG/GG 67 17 A (Ref) 248 53 AA (Ref) 113 25 rs11892031 0.56 (0.16–1.93) 0.358 0.60 (0.17–2.15) 0.429 C 26 3 AC/CC 24 3 1A10 T (Ref) 143 34 TT/TC (Ref) 101 23 rs7571337 0.68 (0.37–1.26) 0.220 0.54 (0.17–1.68) 0.287 C 129 20 CC 35 4 C (Ref) 198 44 CC (Ref) 71 18 rs1112310 0.64 (0.31–1.35) 0.239 0.57 (0.24–1.38) 0.213 A 74 10 CA/AA 65 9 C (Ref) 214 44 CC (Ref) 85 18 rs1113193 0.98 (0.48–1.99) 0.957 0.89 (0.38–2.10) 0.792 T 60 12 CT/TT 52 10 A (Ref) 243 48 AA (Ref) 107 20 rs6706988 1.22 (0.53–2.82) 0.646 1.34 (0.53–3.36) 0.539 G 33 8 AG/GG 31 8 C (Ref) 251 51 CC (Ref) 115 24 rs10176426 1.31 (0.46–3.67) 0.614 1.03 (0.32–3.32) 0.965 1A9 T 21 5 CT/TT 21 4 A (Ref) 174 38 AA (Ref) 57 14 rs2741048 0.74 (0.39–1.40) 0.352 0.69 (0.30–1.59) 0.379 C 100 16 AC/CC 80 13 C (Ref) 149 34 CC/CT (Ref) 102 23 rs26023812 0.78 (0.43–1.41) 0.417 0.63 (0.22–1.79) 0.385 T 125 22 TT 35 5 G (Ref) 257 54 GG (Ref) 120 26 1A7 rs17862859 0.52 (0.12–2.31) 0.390 0.54 (0.12–2.51) 0.434 A 19 2 GA/AA 18 2 G (Ref) 188 44 GG (Ref) 64 17 1A6 rs6751673 0.58 (0.29–1.17) 0.128 0.57 (0.25–1.30) 0.180 A 88 12 GA/AA 74 11

rs17863787 T (Ref) 188 28 2.04 (1.12–3.72) 0.019 TT/GT (Ref) 121 19 3.16 (1.18–8.50) 0.023

G 86 26 GG 16 8 A (Ref) 236 56 AA (Ref) 100 28 rs4663326 - <0.0013 - <0.0013 G 38 0 AG/GG 37 0 C (Ref) 170 25 CC/CT (Ref) 116 18 rs7583278 1.86 (1.03–3.37) 0.039 2.53 (1.00–6.41) 0.050 T 106 29 TT 22 9 C (Ref) 231 50 CC (Ref) 100 23 rs2011404 0.62 (0.25–1.54) 0.301 0.58 (0.20–1.65) 0.308 T 43 6 CT/TT 37 5 1A4 T (Ref) 247 54 TT (Ref) 113 27 rs28899187 – 0.0193 – 0.0273 A 23 0 TA/AA 22 0 T (Ref) 145 27 TT/TC (Ref) 106 20 1A3 rs4663965 1.20 (0.67–2.15) 0.537 1.36 (0.54–3.42) 0.509 C 131 29 CC 32 8 C (Ref) 258 54 CC (Ref) 121 26 1A1 rs11568318 0.64 (0.14–2.90) 0.564 0.63 (0.13–2.92) 0.553 A 16 2 CA/AA 16 2 G (Ref) 208 45 GG (Ref) 80 18 rs28946889 0.86 (0.42–1.77) 0.677 0.85 (0.36–2.01) 0.716 T 60 11 GT/TT 54 10 C (Ref) 237 52 CC (Ref) 99 25 Common rs3771342 0.24 (0.06–1.03) 0.054 0.21 (0.05–0.93) 0.040 A 39 2 CA/AA 39 2 region A (Ref) 236 52 AA (Ref) 98 25 rs2302538 0.22 (0.05–0.95) 0.043 0.19 (0.04–0.85) 0.030 G 40 2 AG/GG 40 2 C (Ref) 174 35 CC/CT (Ref) 118 24 rs4148328 0.95 (0.51–1.77) 0.877 0.85 (0.23–3.13) 0.802 T 98 19 TT 18 3 G (Ref) 171 44 GG (Ref) 54 18 rs6717546 0.46 (0.23–0.91) 0.026 0.37 (0.16–0.86) 0.021 A 103 12 GA/AA 83 10 3'- A (Ref) 235 54 AA (Ref) 99 26 Flanking rs11563250 0.23 (0.05–1.00) 0.050 0.21 (0.05–0.95) 0.043 G 41 2 AG/GG 39 2 region C (Ref) 189 39 CC (Ref) 66 13 rs6719561 0.93 (0.50–1.75) 0.829 1.03 (0.45–2.34) 0.948 T 87 17 CT/TT 72 15

Ref.: reference OR, odds ratio; 95% CI, 95% confidence interval. p-Values in bold type are <0.05. 1Adjusted OR and p-values. 2SNPs that were not found in the Hardy-Weinberg equilibrium (p < 0.05). 3P-values obtained by univariate analysis.

Table 3 Replication of positive UGT1 markers in the Italian cohort in relation to severe neutropenia (grades 3–4). Neutropenia Neutropenia Exons htSNPs Alleles OR1 (95% CI) P Genotypes OR1 (95% CI) P 0–2 3–4 0–2 3–4 A (Ref) 383 55 AA (Ref) 178 22 rs4663326 2.62 (1.26–5.64) 0.011 3.21 (1.40–7.36) 0.006 G 29 11 AG/GG 28 11 T (Ref) 276 42 TT 97 14 1A6 rs17863787 1.31 (0.74–2.30) 0.347 1.30 (0.61–2.75) 0.501 G 112 22 GT/GG 97 18 C (Ref) 264 42 CC (Ref) 82 13 rs7583278 1.02 (0.60–1.76) 0.934 1.02 (0.48–2.16) 0.961 T 148 24 CT/TT 124 20 T (Ref) 353 57 TT (Ref) 185 23 1A4 rs28899187 0.95 (0.44–2.01) 0.884 3.12 (1.24–7.89) 0.016 A 59 9 TA/AA 21 8 C (Ref) 391 54 CC (Ref) 168 24 rs3771342 2.80 (0.18–6.65) 0.020 1.65 (0.71–3.84) 0.244 Common A 21 8 CA/AA 38 9 region A (Ref) 369 57 AA (Ref) 153 25 rs2302538 1.35 (0.62–2.91) 0.449 0.93 (0.39–2.18) 0.859 G 43 9 AG/GG 53 8 G (Ref) 243 39 GG (Ref) 71 11 3'- rs6717546 1.01 (0.59–1.71) 0.984 1.06 (0.49–2.31) 0.889 A 169 27 GA/AA 135 22 Flanking A (Ref) 349 63 AA (Ref) 150 30 region rs11563250 0.26 (0.08–0.87) 0.028 0.27 (0.08–0.92) 0.036 G 63 3 AG/GG 56 3 Ref.: reference OR, odds ratio; 95% CI, 95% confidence interval. p-Values in bold type are <0.05. 1Adjusted OR and p-values.

73 Supplementary Table S1 UGT1 htSNPs genotyped in the Canadian cohort.

Chr. Amino acid OR Regions SNPs Allele Location position substitution (p-value1) rs1597942 234535850 C/T c.855+8642 1.00 (0.99) 1A8 rs1377460 234536637 A/G c.855+9429 0.82 (0.61) rs1823803 234539111 C/T c.855+11903 0.70 (0.21) rs2741034 234548814 A/G c.855+2791 1.58 (0.11) rs11892031 234565283 A/C c.855+19260 0.59 (0.39) 1A10 rs7571337 234566418 C/T c.855+20395 0.70 (0.22) rs1112310 234567916 A/C c.855+21893 0.63 (0.23) rs1113193 234569137 C/T c.855+23114 0.96 (0.91) rs6706988 234573623 A/G c.855+27600 1.17 (0.73) rs17868320 234578428 C/T c.-2152 0.28 (0.22) rs2741044 234579368 A/G c.-1212 1.96 (0.024) rs3806598 234579892 G/T c.-688 5.61 (0.022) rs10176426 234579915 C/T c.-665 1.29 (0.63) rs2741045 234580140 C/T c.-440 1.83 (0.037) 1A9 rs6714486 234580305 A/T c.-275 0.49 (0.34) c.-118_- 0.67 (0.21) rs3832043 234580454 -/T 117insT rs2741048 234581748 A/C c.855+313 0.73 (0.32) rs2741049 234581834 C/T c.855+399 0.69 (0.23) rs17862857 234582084 C/T c.855+649 0.59 (0.22) rs2602381 234584324 C/T c.855+2889 0.78 (0.39) rs7586110 234590527 G/T c.-57 1.25 (0.45) rs17868323 234590970 G/T c.387 N129K 0.68 (0.22) 1A7 rs17863778 234590974 A/C c.391 R131R 0.68 (0.21) rs11692021 234591205 C/T c.622 W208R 1.99 (0.021) rs17862859 234597842 A/G c.855+6404 0.53 (0.39) rs6759892 234601669 G/T c.19 S7A 1.94 (0.026) rs2070959 234602191 A/G c.541 T181A 2.20 (0.007) rs1105879 234602202 G/T c.554 R184S 2.16 (0.008) 1A6 rs6751673 234604903 A/G c.861+2392 0.59 (0.14) rs17863787 234611094 G/T c.861+8583 1.93 (0.025) rs4663326 234616571 A/G c.861+14060 - (0.001)2 rs7583278 234617407 C/T c.861+14896 1.80 (0.046) 1A4 rs2011404 234627937 C/T c.471 C157C 0.64 (0.31) rs2889918 - (0.019)2 234630967 A/T c.867+2634 7 1A3 rs4663965 234650604 C/T c.867+11965 1.23 (0.48) rs11568318 234665498 A/C c.868-10182 0.62 (0.54) 1A1 rs4124874 234665659 A/C c.-3279 1.20 (0.53)

rs10929302 234665782 A/G c.-3156 2.06 (0.011) rs8175347 1.77 (0.047) 234668880 TA6/TA7 -54_-53insTA TATA rs28946889 234671462 G/T c.864+1665 0.88 (0.73) Common rs3771342 234672663 A/C c.864+2866 0.21 (0.042) region rs2302538 234676413 A/G c.994-82 0.20 (0.036) rs4148328 234677659 C/T c.1304+574 0.96 (0.90) rs10929303 234681416 C/T c.*211 0.64 (0.26) rs1042640 234681544 C/G c.*339 0.89 (0.75) rs8330 234681645 C/G c.*440 0.69 (0.33) 3'-UTR3 rs6717546 234682119 A/G c.*914 0.46 (0.026) rs11563250 234683350 A/G g.189961 0.22 (0.042) rs6719561 234683763 C/T g.190374 0.94 (0.86) SNPs previously published by our group are italicized. 16SNPs replicated in Italian cohort are identified by bold type.1Values were adjusted for age. 2Indicates univariate p-values 3Relative to UGT1A1

75 FIGURE LEGENDS Figure 1. Linkage disequilibrium map of 28 htSNPs genotyped in the discovery cohort. The map illustrates the linkage disequilibrium for the 28 UGT1 htSNPs first assessed in the discovery cohort of 167 Canadian patients and resembles that from the CEU population. The rs8175347 corresponding to the well-known UGT1A1 promoter variant

(A(TA)6>7TAA region) has been included in the LD map but was previously reported for this cohort of patients.16 Values inside each square are those for r2 and are reported as percentages. The colors depict the strength of the LD between each pair of htSNPs.

Figure 2. Schematic of the two-marker haplotype comprising rs11563250 and the UGT1A1 promoter variant rs8175347. Yellow rectangles represent the reference nucleotide (with respect to the reference sequence, AF297093), whereas olive-green rectangles represent the variant allele. HI-HIII, Haplotype groups I–III; OR, odds ratio; 95% CI, 95% confidence intervals; H1 corresponds to the reference haplotype (OR= 1.0). Frequencies in studied populations are shown.

Figure 3. e presence of rs11563250. a) Data are presented for the discovery cohort patients and b) for the carriers of UGT1A1*1/*1 in that cohort. The p-value significance was determined by comparing the means of the logarithmic-transformed raw bilirubin values using the Student’s t-test. The red bars indicate the mean values for each group.

Figure 4. Schematic showing the positions of the 3'-flanking marker rs11563250 and eight LD markers genotyped in UGT1 from the Canadian and Italian mCRC patients. OR, odds ratio; 95% CI, 95% confidence intervals. R2 values between rs11563250 and each of the eight variants are provided.

Figure 1

77 Figure 2

Figure 3

Figure 4

Chapter III Identification of ABCC5 and ABCG1 polymorphisms that predict irinotecan-induced severe toxicity in metastatic colorectal cancer patients

81 Résumé

Objectif: Le traitement du cancer colorectal métastatique (mCRC) avec l’irinotécan provoque fréquemment la neutropénie et la diarrhée, limitant sa poursuite. Bien que plusieurs études pharmacogénomiques aient évalué les risques associés à cette chimiothérapie, l’impact des gènes des transporteurs (ABC et SLC) a été peu étudié. Approche: Par une stratégie d’haplotype/SNP-étiquette (htSNP), 210 htSNPs dans les sept gènes ABC et dans SLCO1B1 impliqués dans la pharmacocinétique de l’irinotécan furent séquencés chez 167 patients canadiens sous traitement FOLFIRI, puis validés dans une cohorte indépendante de 250 patients italiens. Résultats: La combinaison des marqueurs ABCC5 rs3749438 et rs10937158 (T–C) a prédit un risque réduit de diarrhée sévère dans les deux cohortes (RC = 0.43; p=0.001). Un risque accru de neutropénie est prédit par la coexistence des marqueurs ABCG1 rs225440T et ABCC5 rs2292997A (RC=5.93; p=0.0002) et plus significatif encore si combinés avec le marqueur de risque connu UGT1A1*28 rs8175347 (RC=7.68; p<0.0001). Enfin, les porteurs de l’allèle de protection UGT1 rs11563250G en absence d’allèles de risque ont montré une incidence réduite de neutropénie sévère (8.2% vs. 34.0%; p<0.0001). Conclusion: Ces nouveaux marqueurs génétiques prédictifs pourraient permettre d’améliorer l’évaluation du risque de toxicité et de personnaliser le traitement à l’irinotécan pour les patients atteints du mCRC.

ABCC5 and ABCG1 polymorphisms predict irinotecan-induced severe toxicity in metastatic colorectal cancer patients

Sylvia Chen1, Lyne Villeneuve1, Derek Jonker2, Félix Couture3, Isabelle Laverdière1, Erica

Cecchin4, Federico Innocenti5, Giuseppe Toffoli4, Eric Lévesque1,3 and Chantal

Guillemette1,6

1Pharmacogenomics Laboratory, Centre Hospitalier Universitaire de Québec (CHU de Québec) Research Center and Faculty of Pharmacy, Laval University, Québec, Canada. 2Department of Medicine, Ottawa Hospital Research Institute, University of Ottawa, Ontario, Canada. 3CHU de Québec Research Center and Faculty of Medicine, Laval University, Québec, Canada. 4Division of Experimental and Clinical Pharmacology, Department of Molecular Biology and Translational Research, National Cancer Institute and Cancer for Molecular Biomedicine, Aviano, Italy. 5Division of Pharmacotherapy et Experimental Therapeutics, Eshelman School of Pharmacy, University of North Carolina, Chapel Hill, North Carolina. 6Canada Research Chair in Pharmacogenomics.

Running Title: Transporter genes and irinotecan-induced toxicity

Correspondence: Chantal Guillemette, PhD; Canada Research Chair in Pharmacogenomics, CHU de Québec Research Center, R4720, 2705 Boul, Laurier, Québec, Canada, G1V 4G2. Tel. (418) 654-2296 Fax. (418) 654-2761 E-mail: [email protected]

Pharmacogenetics and Genomics; accepted July 6, 2015

Impact factor: 3.481

83 ABSTRACT

Objective: Irinotecan is a cytotoxic agent widely used for the treatment of solid tumors most particularly for metastatic colorectal cancers (mCRC). Treatment with this drug frequently results in severe neutropenia and diarrhea that can seriously impact the course of treatment and patients’ quality of life. Pharmacogenomic tailoring of irinotecan-based chemotherapy has been the subject of several investigations but with limited data regarding ABC and SLC transporter genes. Methods: In this study, we sought to discover toxicity-associated markers in seven transporter genes participating in irinotecan pharmacokinetics involving the ABC transporter genes ABCB1, ABCC1, ABCC2, ABCC5, ABCG1, ABCG2 and the solute carrier organic anion transporter gene SLCO1B1 and using a haplotype-tagging single-nucleotide polymorphisms (n=210 htSNPs) strategy. The profiles of 167 mCRC Canadian patients treated with FOLFIRI-based regimens were examined and findings were replicated in an independent cohort of 250 Italian patients. Results: In combined cohorts, a two-marker ABCC5 rs3749438 and rs10937158 haplotype (T–C) predicted lower risk of severe diarrhea (odds ratio (OR) of 0.43; p=0.001). The co-occurrence of ABCG1 rs225440T and ABCC5 rs2292997A predicted risk of severe neutropenia (OR=5.93; p=0.0002), which was further improved when incorporating the well-known risk marker UGT1A1*28 rs8175347 (OR=7.68; p<0.0001). In contrast, carriers of one protective marker (UGT1 rs11563250G) but none of these risk alleles experienced significantly less severe neutropenia (8.2% vs. 34.0%; p<0.0001). Conclusion: This combination of predictive genetic markers could potentially lead to better risk assessment and may thus improve personalized treatment.

Key words: ABC transporters, genetic variants, irinotecan, colorectal cancer, hematological toxicity, gastrointestinal toxicity

INTRODUCTION Irinotecan treatment is widely used in combination chemotherapy with folinic acid and fluorouracil (FOLFIRI) as one of the key regimens for metastatic colorectal cancer [1-3]. Irinotecan is a prodrug converted by carboxylesterases into its active metabolite, 7-ethyl- 10-hydroxy-camptothecin (SN-38), which acts to inhibit topoisomerase I activity during DNA replication. SN-38 is mainly detoxified by UDP-glucuronosyltransferases (UGTs) to form the inactive SN-38 glucuronide (SN-38G) that is eliminated by the biliary tract. Patients receiving irinotecan-containing regimens often experience severe neutropenia and diarrhea that can seriously impact the course of treatment and patient quality of life. In recent years, the vast majority of irinotecan-related pharmacogenomics studies have focused on the impact of UGT genetic variants. The most reliable predictor of severe neutropenia is the UGT1A1*28/*28 promoter genotype linked to decreased SN-38 glucuronidation, greater exposure to SN-38, and a 2-fold greater risk of toxicity as a result of impaired UGT1A1 function [4]. Other UGT variants such as the UGT1A1*6 variant affecting the enzyme sequence (Arg71Gly; rs4148323) predominates in Asians and is linked to an increased incidence of severe neutropenia [5, 6]. Starting with a lower dose of irinotecan may help reduced their risk of severe neutropenia [4, 7-10]. However, prediction of severe diarrhea based on the UGT1A1*28 is not consistent across studies [11-14]. Recently, our group reported a novel UGT1 marker (rs11563250G) located in the 3'- flanking region of UGT1 that is associated with lower incidence of severe irinotecan- induced neutropenia in two independent cohorts of mCRC patients [15]. Genotyping of this variant along with knowledge of the UGT1A1 TATA box promoter status (rs8175347) may have clinical consequences on irinotecan-dosing management because carriers of rs11563250G better tolerate the drug and might therefore benefit from greater irinotecan dosing to maximize antitumor activity without increasing toxicity. Other factors are likely to cooperate with common UGT1 genetic variants to influence the risk of severe toxicity. These factors include several membrane transporter proteins members of the ATP-binding cassette (ABC) and the solute carrier (SLC) families that participate in the efflux and uptake of irinotecan, SN-38, and SN-38G, which play critical roles in modulating the pharmacokinetics of the drug [16-22]. At least six reports of mCRC patients suggest that polymorphisms of transporter genes significantly impact the risk of irinotecan-associated toxicities (Table 1). Irinotecan and SN-38 may be transported out of the cell via the efflux transporter ABCB1 (MDR1/p-glycoprotein) and subsequently be excreted in bile and urine [19]. In a cohort of 140 patients, homozygous carriers of the

85 ABCB1 3435T variant had an increased risk of toxicities, whereas in a smaller study of 56 irinotecan-treated patients, it was associated with a lower risk of diarrhea [23, 24]. Irinotecan and its metabolites are also substrates for the efflux carrier ABCG2 (BCRP). One variant located in the 5' untranslated region of ABCG2 (−15994G>A) was linked to non-hematological toxicity in the first cycle of treatment [25]._ENREF_22 The ABCC1 multidrug resistance-related protein 1 (MRP1) transporter was first identified in irinotecan- resistant human epidermoid carcinoma cells [21, 26]. One study examined three ABCC1 variants linked to altered irinotecan pharmacokinetics and risk of hematologic and non- hematologic toxicity [25]. ABCC2, known as canalicular multispecific organic anion transporter (MRP2), is responsible for the export of molecules from hepatocytes and has a greater affinity for SN-38 and SN-38G substrates than irinotecan [17, 20]. The ABCC2 −1023G>A variant was inversely linked to toxicity of any type and non-hematological toxicity in a study of 250 mCRC patients.[25] SLCO1B1 encodes the solute carrier organic anion transporter, which mediates the hepatic influx of SN-38 from the blood [22]. One SLCO1B1 variant (*1b; 388A>G) was associated with an increased risk of gastrointestinal toxicity and an increased absolute neutrophil count in a small study of 26 patients [27]. In contrast, 521T>C was linked to decreased toxicity in 56 patients [28]. Although few reports have described the functions of ABCG1 and ABCC5 in irinotecan-related efflux, one study suggested a significant association of markers in ABCC5 (3′ untranslated region, UTR, 1243T>C) and ABCG1 (c.974–898C>G) with an increased risk of GI toxicity [27]. Overall, all studies have focused on a limited number of variants in one or only a few transporter genes in patients treated with different irinotecan-containing regimens (Table 2). However, significant findings have yet to be validated in independent cohorts. Therefore, the challenge is to refine the precise role of polymorphisms in irinotecan transporter genes in predicting toxicity. In the present work, the seven transporter genes discussed above (ABCB1, ABCG1, ABCG2, ABCC1, ABCC2, ABCC5, SLCO1B1) were systematically investigated by genotyping 210 haplotype-tagged single nucleotide polymorphisms (htSNPs) in a discovery cohort of 167 Canadian mCRC patients. Positive findings were subsequently replicated in an independent cohort of 250 Italian mCRC patients. The study endpoints were severe grade 3–4 neutropenia and diarrhea, analyzed separately. In addition, we also tested the co-occurrence of novel genetic markers in transporter genes with established UGT1 markers inversely associated with risk of grade 3–4 neutropenia (UGT1A1*28 rs8175347 and rs11563250G) [7, 8, 10, 15]. We found that a combination of

novel markers significantly associated with severe neutropenia and diarrhea could be validated in two independent populations, whereas their combined analyses with UGT1 markers further improved the prediction of risk.

87 MATERIALS AND METHODS Patient cohorts and severe toxicities In our prospective study, 167 mCRC white patients of European genetic ancestry from eastern Canada who received the FOLFIRI (FOL (folinic acid), F (fluorouracil or 5-FU), and IRI (irinotecan) regimen were profiled as the discovery cohort. All patients received a 180 mg/m2 intravenous dose of irinotecan every 2 weeks and 75 patients also received co- treatments such as bevacizumab. Detailed treatment modalities, eligibility, recruitment criteria and isolation of DNA from whole blood samples were documented previously [29]. Table 2 summarizes patient demographics (age, gender) and clinical information (treatment, toxicity, tumor site). The replication cohort consisted of 250 mCRC white patients of European genetic ancestry receiving first-line FOLFIRI regimen from northeast Italy previously described elsewhere [30, 31], and genomic DNA was available for 237. The severities of neutropenia and diarrhea were evaluated based on National Cancer Institute Common Terminology Criteria for Adverse Events version 3.0 criteria (NCI- CTCAE v3.0). One unique aspect of the discovery cohort included a daily journal completed by all patients who self-documented gastrointestinal toxicity events for the first two cycles of treatment. The most severe neutropenia and diarrhea events reported in that period were used for data analysis. Among all patients, one died before levels of toxicities could be assessed, and another patient did not fill out the journal. In total, 28 and 24 Canadian patients were characterized with grade 3–4 severe neutropenia and diarrhea, respectively, whereas in the Italian cohort 33 subjects had severe neutropenia and 21 patients had severe diarrhea. All participants provided written informed consent for genetic analysis, and local research ethics committees approved the research protocol.

Genetic analysis Seven candidate transporter genes involved were selected from literature findings related to transport of irinotecan and its metabolites. To maximize locus coverage, SNPs located in all gene regions, including regulatory, exonic, and intronic areas spanning ±5 kb in the 5′ and 3′ regions, were identified from the International HapMap Project Database (Release 28). htSNPs representing all SNPs in linkage disequilibrium (LD) with r2 > 0.8 were generated by performing haplotype association tests with Haploview v.4.2 software (Broad Institute, Cambridge, MA, USA). SNPs reported in the literature were included. All extension primers and PCR assays were designed using SpectroDESIGNER software (Sequenom, San Diego, CA, USA). htSNPs that could not be sequenced owing to poor

primer design or were in duplicate regions were replaced by tagged SNPs in complete LD (r2 = 1). A total of 210 htSNPs (n = 183 for ABC transporter genes comprising ABCB1, ABCC1, ABCC2, ABCC5, ABCG1, and ABCG2 and n = 27 for the SLC transporter gene SLCO1B1) were genotyped by Sequenom® iPLEX matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (Sequenom, San Diego, CA, USA) at the CHU de Quebec Research Center. Assay robustness and genotyping reproducibility were verified with negative controls and 5% random sample duplicates. Data analysis Allelic and genotype frequencies for each marker for all genes, including deviations from Hardy-Weinberg equilibrium, were determined with the PLINK v1.07 whole-genome association analysis toolset [32]. Genetic association tests between genotype and clinical outcome were performed with SPSS v22 software (SPSS Inc., Chicago, IL, USA) using logistic regression for both allelic and genotypic models. Covariates such as age and co- treatment were also added in the models, when applicable. Haplotype constructs were inferred with Phase v2.1.1 [33], and genetic associations—including co-occurrence against clinical outcomes—were calculated using Fisher’s exact test. Comparisons between frequencies of carriers and non-carriers of risk markers were assessed using a two-tailed t-test. False-discovery rates (q values) were calculated using the R QVALUE package (http://genomics.princeton.edu /storeylab/qvalue/).

89 RESULTS We screened 210 htSNPs from seven candidate genes that encode transporter proteins (ABCB1, ABCC1, ABCC2, ABCC5, ABCG1, ABCG2, and SLCO1B1) in the prospective discovery cohort composed of mCRC FOLFIRI-treated patients (n=167). Associations with grade 3–4 severe diarrhea and neutropenia were tested separately. All previously reported markers (Table 1) were either genotyped or tagged by htSNPs in our analysis. Positive markers were then replicated in an independent Italian cohort of FOLFIRI-treated patients (Table 2). Markers of severe diarrhea Six htSNPs in ABCC5 and ABCG1 were significantly associated with severe grade 3–4 diarrhea in the discovery cohort (Table 3). Two ABCC5 markers rs10937158 and rs3749438 were successfully replicated in the Italian cohort and were inversely associated with grade 3–4 severe diarrhea. rs10937158T was significantly associated with a 2.5-fold decreased risk for both the Canadian and Italian cohorts (OR = 0.40–0.45, p = 0.011– 0.040) and OR of 0.42 (p = 0.001) in the combined analysis. In contrast, rs3749438T was associated with OR values of 5.93 (p = 0.019) and 4.63 (p = 0.044) for the Canadian and Italian populations, respectively, and an OR of 2.04 (p = 0.002) when combined. All other associations between htSNPs and risk of severe diarrhea found in the Canadian discovery cohort were not replicated in the Italian validation cohort (p > 0.05). We then examined the combined effect of these two ABCC5 variants in a haplotype-based association test because these two htSNPs are in moderate LD with one another (r2 = 0.52). When compared with the most frequent haplotype HI (C-T) consisting of rs10937158C and rs3749438T, the variant haplotype HII (T-C) consisting of rs10937158T and rs3749438C was associated with a reduced risk of severe diarrhea with OR values of 0.37 (p = 0.010), 0.41 (p = 0.035), and 0.39 (p = 0.0006) for the Canadian, Italian, and combined cohorts, respectively (Figure 1). Markers of severe neutropenia Twelve htSNPs from the ABC transporter gene family, notably ABCC1, ABCC2, ABCC5, and ABCG2, and one htSNP in the solute carrier organic anion transporter gene SLCO1B1 were found to be associated with a significant increased risk of severe grade 3–4 neutropenia in the discovery cohort with ORs ranging from 2.36 to 7.63 (p < 0.05) (Table 4). Conversely, three markers in ABC transporter genes ABCC5 and ABCG1 and one SLCO1B1 marker were found to be associated with a significant reduced risk of grade 3–4 neutropenia risk with ORs ranging from 0.20 to 0.33 (p < 0.05) (Table 4). Positive markers

were subsequently genotyped in the Italian cohort (n = 237 mCRC cases), and two htSNPs in ABCG1 (rs225440) and ABCC5 (rs1016752) were replicated (Table 5). All other associations between htSNPs and risk of severe neutropenia found in the Canadian discovery cohort were not replicated in the Italian validation cohort (p > 0.05). As observed for the Canadian cohort (OR = 3.08, p = 0.033), the minor ABCG1 rs225440T allele was also significant in the Italian cohort with an increased risk of severe neutropenia (OR= 1.89; p = 0.019). When cohorts were combined, the OR value was 1.71 (p = 0.007). In addition, carriers of rs2292997 minor allele A in ABCC5 had a greater risk of severe neutropenia with ORs of 3.15 (p = 0.014) and 2.04 (p = 0.077) for the Canadian and Italian populations, respectively and OR of 1.89 (p = 0.018) when combined (Tables 4 and 5). We then sought to evaluate the co-occurrence of both markers. Compared with the reference group (CC-GG), the co-occurrence of both ABCG1-ABCC5 markers T-A (CT/TT- AA/GA) further increased the risk of severe neutropenia with OR values of 8.34 (p = 0.002) and 6.10 (p = 0.022) for the Canadian and Italian populations, respectively. When the two cohorts were combined, the risk was also significantly greater (OR = 5.93; p = 0.0002) (Table 6). Finally, a combined analysis of genetic markers in UGT1, ABCG1, and ABCC5 was undertaken in relation to grade 3–4 neutropenia because two genetic variants in UGT1 (rs8175347 and rs11563250) were previously linked to severe neutropenia in the same two mCRC patient cohorts [15, 29, 30]. The markers associated with increased risk were ABCG1 rs225440T, ABCC5 rs2292997A, and UGT1A1*28, whereas inversely, the UGT1 rs11563250G variant has been associated with a decreased risk of severe neutropenia in both populations [15]. Accordingly, carriers of one protective UGT1 rs11563250G allele and/or non-carriers of risk markers constituted the reference group for this analysis (Table 7). Patients carrying at least two risk alleles displayed a significantly increased risk with OR values of 9.64 (p = 0.001), 6.61 (p = 0.001), and 7.67 (p < 0.0001) for the Canadian, Italian, and combined cohorts, respectively. Actually, patients carrying two or more risk markers experienced more grade 3 or 4 neutropenia (62.3% versus 33.7%; p < 0.0001). In contrast, patients carrying no risk alleles experienced significantly less grade 3 or 4 neutropenia (8.2% versus 34.0%; p < 0.0001). A detailed analysis of tagged SNPs in strong LD (r2 = 0.80) with the intronic variants ABCC5 rs2292997, rs10937158, and rs3749438 and variant ABCG1 rs225440 and their genomic positions is provided in Figure 2 and Supplementary Figure 1. Notably, none of these variants were linked to non-synonymous variations.

91 DISCUSSION Through a systematic investigation of genetic variations in seven transporter genes, we provide evidence for the relationship between the drug efflux proteins ABCG1 and ABCC5 and irinotecan-induced toxicities in two independent FOLFIRI-treated mCRC patient cohorts comprising over 400 patients. Data support the role of ABCG1 and ABCC5 transporters in irinotecan clearance and further imply cooperation between these transport pathways as well as UGT-related drug metabolism in determining drug exposure and risk of severe toxicities. In addition, our findings underscore the need to validate the presence of a biomarker(s) in independent populations to obtain clinically meaningful results with translational potential. Our results represent the first evidence of a role for genetic variants in ABCC5 and ABCG1 in neutropenia toxicity of irinotecan and infer that the simultaneous alteration of both efflux transporters may result in higher efficiency of SN-38 transport to the systemic circulation, thereby leading to severe grade 3–4 neutropenia induced by excessive exposure to SN- 38. The minor ABCG1 rs225440T was significantly associated with an increased risk of severe neutropenia in both populations. A similar finding was observed for the minor ABCC5 rs2292997A allele, whereas their co-occurrence further increased the risk, suggesting that increased exposure to the toxic metabolite SN-38 is coupled to neutropenia. The assessment of these new transporter gene markers along with relevant UGT1 variants (UGT1A1*28 and rs11563250) [15, 29, 30], which have been proven to affect risk of severe neutropenia in the studied populations significantly improved risk prediction, suggesting that both metabolic and transport pathways cooperate to determine drug exposure and subsequent risk of neutropenia.

ABCG1 has been predominantly characterized as a sterol efflux transporter and modulates apoptosis and endothelial function by export of cholesterol [34, 35]. Other ABCG1 substrates are two anticancer drugs, mitoxantrone and doxorubicin [36]. Similar to the structure of cholesterol, these drugs have a sterol-like ring structure that may interact with ABCG1, whereas irinotecan also contains a pentacyclic ring structure that could potentially interact with this transporter. The variant ABCG1 rs225440 associated with severe neutropenia is located in intron 7 of ABCG1 and is genetically linked to 19 other variations in introns 6 and 7, according to public databases (Figure S1). A number of other ABCG1 isoforms that are generated by alternative splicing and/or promoter usage have been

identified [37-40], but none of these polymorphisms are predicted to affect splicing regulation of ABCG1. Our study reinforces emerging evidence that ABCC5 is involved in SN-38 transport. One report implicates ABCC5 as an efflux transporter of SN-38, showing an increase in ABCC5 mRNA expression in SN-38-resistant human cervical carcinoma cell lines [26]. ABCC5 is ubiquitously expressed in many tissues including liver, colon, muscle, heart, and brain, localizes to the basolateral membrane and is responsible for the transport of cyclic nucleotides and platinum-based and nucleoside-based analog drugs used in antiviral and anticancer treatment [41-43]. Studies of ABCC5 and its molecular mechanisms of drug transport are limited, and our findings suggest further investigation of the possible involvement of ABCC5 with irinotecan and its metabolites. The rs2292997 polymorphism is located in intron 2 of ABCC5 is linked to over 20 other SNPs across the gene, but none introduce an amino acid change or potentially alter the binding affinity of putative transcription factors (Figure S1). We further established the relationship between severe diarrhea and a haplotype defined by two intronic ABCC5 markers, rs3749438 and rs10937158, not previously reported in the literature. A variation in the expression or activity of this transporter could create a milieu predisposed to drug-induced damage of the intestinal mucosa caused by an imbalance in SN-38 secretion. Notably, ABCC5 rs3749438 and rs10937158 are strongly linked (r2 = 1.0) to respective variants rs3749440 and rs4148575 located in the 3′UTR of an ABCC5 variant transcript (NM_001023587), potentially affecting its expression. This 1933-bp-long truncated variant ABCC5 transcript has been previously detected with at least 2-fold greater expression in the distal colon and stomach compared with the full-length ABCC5 transcript and was found to negatively modulate the expression of native ABCC5 mRNA in retinal cells [44]. The change from A to G in rs3749440 induces the creation of a new acceptor splice site while the rs4148575 variation is also located in the 3’UTR of a variant ABCC5 transcript (GenBank AY754874) that undergoes nonsense-mediated decay [44]. In a study that included 27 mCRC patients, Di Martino et al. found that the ABCC5 marker rs562CC was strongly associated with a 32-fold increase risk in severe diarrhea.[27] This specific marker was not significant in our discovery cohort of 167 patients (OR = 1.69. p = 0.297), further reinforcing the need to include a larger number of patients and to replicate findings in independent cohort. We did not observed significant associations with most of the other variants of previous studies in relation to irinotecan-induced toxicities (summarized in Table 1). This is likely because these previous associations were derived

93 from a limited number of variants in one or only a few transporter genes and in a limited number of patients of diverse ethnicity, treated with different irinotecan-containing regimens, and with sometimes different designation of toxicity endpoints when defined. Strengths of the study include systematic coverage of common genetic variations in seven transporter genes that have not been simultaneously studied before, a replication cohort, in addition to haplotype and co-occurrence analysis along with established markers in UGT1. In addition, the use of a detailed journal of gastrointestinal toxicity in the prospective discovery cohort enhanced the accuracy of this phenotype and may explain why some of the associations were not validated in the replication cohort. A limitation of this study is that the relationship with irinotecan pharmacokinetic data could not be assessed and therefore we cannot draw causal conclusions. In addition, coverage of regulatory regions of candidate transporter genes, >5 kb beyond the 5' and 3' regions, were not tested. Finally, the implication of these findings in other ethnic groups remains unstudied and we cannot rule out that the observed associations are related to other drugs of the FOLFIRI regimens since patients did not just receive irinotecan herein. Additional studies are thus required to define the mechanisms underlying the observed associations.

Conclusion Our study supports for a contribution of ABC transporters to irinotecan pharmacogenomics and identified novel markers in the transporter gene ABCG1 associated with irinotecan- induced neutropenia, whereas ABCC5 markers were associated with both neutropenia and diarrhea. Our study represents another step towards personalized and more precise FOLFIRI-related treatment of mCRC patients.

95 Acknowledgements This work was supported by the Canadian Institutes of Health Research (CIHR) (C.G.; CIHR MOP-42392) and the Canada Research Chair Program (C.G.). S.C. was a recipient of a studentship award from ‘Fonds de l’Enseignement et de la Recherche (FER)’ from Faculty of Pharmacy at Laval University. I.L. is a recipient of a CIHR Frederick Banting and Charles Best studentship award and a Graduate Scholarship for clinician-scientist from FRQ-S. E.L. is a recipient of a CIHR clinical-scientist phase II award and Prostate Cancer Canada Rising Star Award (RS2013-55). C.G. holds a Canada Research Chair in Pharmacogenomics (Tier I).

References

1. Zafar SY, Malin JL, Grambow SC, Abbott DH, Kolimaga JT, Zullig LL, et al. Chemotherapy use and patient treatment preferences in advanced colorectal cancer: a prospective cohort study. Cancer. 2013;119:854-862. 2. Abrams TA, Meyer G, Schrag D, Meyerhardt JA, Moloney J, Fuchs CS. Chemotherapy usage patterns in a US-wide cohort of patients with metastatic colorectal cancer. J Natl Cancer Inst. 2014;106:djt371. 3. Inoue Y, Toiyama Y, Tanaka K, Miki C, Kusunoki M. A comprehensive comparative study on the characteristics of colorectal cancer chemotherapy. Jpn J Clin Oncol. 2009;39:367-375. 4. 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 U S A. 1998;95:8170-8174. 5. Han FF, Guo CL, Yu D, Zhu J, Gong LL, Li GR, et al. Associations between UGT1A1*6 or UGT1A1*6/*28 polymorphisms and irinotecan-induced neutropenia in Asian cancer patients. Cancer Chemother Pharmacol. 2014;73:779-788. 6. Cheng L, Li M, Hu J, Ren W, Xie L, Sun ZP, et al. UGT1A1*6 polymorphisms are correlated with irinotecan-induced toxicity: a system review and meta-analysis in Asians. Cancer Chemother Pharmacol. 2014;73:551-560. 7. Barbarino JM, Haidar CE, Klein TE, Altman RB. PharmGKB summary: very important pharmacogene information for UGT1A1. Pharmacogenet Genomics. 2014;24:177-183. 8. Dias MM, McKinnon RA, Sorich MJ. Impact of the UGT1A1*28 allele on response to irinotecan: a systematic review and meta-analysis. Pharmacogenomics. 2012;13:889-899. 9. Guillemette C, Levesque E, Rouleau M. Pharmacogenomics of human uridine diphospho-glucuronosyltransferases and clinical implications. Clin Pharmacol Ther. 2014;96:324-339. 10. Hu ZY, Yu Q, Pei Q, Guo C. Dose-dependent association between UGT1A1*28 genotype and irinotecan-induced neutropenia: low doses also increase risk. Clin Cancer Res. 2010;16:3832-3842. 11. Hu ZY, Yu Q, Zhao YS. Dose-dependent association between UGT1A1*28 polymorphism and irinotecan-induced diarrhoea: a meta-analysis. Eur J Cancer. 2010;46:1856-1865. 12. Li M, Wang Z, Guo J, Liu J, Li C, Liu L, et al. Clinical significance of UGT1A1 gene polymorphisms on irinotecan-based regimens as the treatment in metastatic colorectal cancer. Onco Targets Ther. 2014;7:1653-1661. 13. Shulman K, Cohen I, Barnett-Griness O, Kuten A, Gruber SB, Lejbkowicz F, et al. Clinical implications of UGT1A1*28 genotype testing in colorectal cancer patients. Cancer. 2011;117:3156-3162. 14. Marcuello E, Altes A, Menoyo A, Del Rio E, Gomez-Pardo M, Baiget M. UGT1A1 gene variations and irinotecan treatment in patients with metastatic colorectal cancer. Br J Cancer. 2004;91:678-682. 15. Chen S, Laverdiere I, Tourancheau A, Jonker D, Couture F, Cecchin E, et al. A novel UGT1 marker associated with better tolerance against irinotecan-induced severe neutropenia in metastatic colorectal cancer patients. Pharmacogenomics J. 2015; In Press. 16. Nakatomi K, Yoshikawa M, Oka M, Ikegami Y, Hayasaka S, Sano K, et al. Transport of 7-ethyl-10-hydroxycamptothecin (SN-38) by breast cancer resistance

97 protein ABCG2 in human lung cancer cells. Biochem Biophys Res Commun. 2001;288:827-832. 17. Chu XY, Kato Y, Niinuma K, Sudo KI, Hakusui H, Sugiyama Y. Multispecific organic anion transporter is responsible for the biliary excretion of the camptothecin derivative irinotecan and its metabolites in rats. J Pharmacol Exp Ther. 1997;281:304-314. 18. Chu XY, Kato Y, Ueda K, Suzuki H, Niinuma K, Tyson CA, et al. Biliary excretion mechanism of CPT-11 and its metabolites in humans: involvement of primary active transporters. Cancer Res. 1998;58:5137-5143. 19. Iyer L, Ramirez J, Shepard DR, Bingham CM, Hossfeld DK, Ratain MJ, et al. Biliary transport of irinotecan and metabolites in normal and P-glycoprotein-deficient mice. Cancer Chemother Pharmacol. 2002;49:336-341. 20. Sugiyama Y, Kato Y, Chu X. Multiplicity of biliary excretion mechanisms for the camptothecin derivative irinotecan (CPT-11), its metabolite SN-38, and its glucuronide: role of canalicular multispecific organic anion transporter and P- glycoprotein. Cancer Chemother Pharmacol. 1998;42 Suppl:S44-49. 21. Chen ZS, Furukawa T, Sumizawa T, Ono K, Ueda K, Seto K, et al. ATP-Dependent efflux of CPT-11 and SN-38 by the multidrug resistance protein (MRP) and its inhibition by PAK-104P. Mol Pharmacol. 1999;55:921-928. 22. Nozawa T, Minami H, Sugiura S, Tsuji A, Tamai I. 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. 23. Cortejoso L, Garcia MI, Garcia-Alfonso P, Gonzalez-Haba E, Escolar F, Sanjurjo M, et al. Differential toxicity biomarkers for irinotecan- and oxaliplatin-containing chemotherapy in colorectal cancer. Cancer Chemother Pharmacol. 2013;71:1463- 1472. 24. Glimelius B, Garmo H, Berglund A, Fredriksson LA, Berglund M, Kohnke H, et al. Prediction of irinotecan and 5-fluorouracil toxicity and response in patients with advanced colorectal cancer. Pharmacogenomics J. 2011;11:61-71. 25. De Mattia E, Toffoli G, Polesel J, D'Andrea M, Corona G, Zagonel V, et al. Pharmacogenetics of ABC and SLC transporters in metastatic colorectal cancer patients receiving first-line FOLFIRI treatment. Pharmacogenet Genomics. 2013;23:549-557. 26. Takara K, Kitada N, Yoshikawa E, Yamamoto K, Horibe S, Sakaeda T, et al. Molecular changes to HeLa cells on continuous exposure to SN-38, an active metabolite of irinotecan hydrochloride. Cancer Lett. 2009;278:88-96. 27. Di Martino MT, Arbitrio M, Leone E, Guzzi PH, Rotundo MS, Ciliberto D, et al. Single nucleotide polymorphisms of ABCC5 and ABCG1 transporter genes correlate to irinotecan-associated gastrointestinal toxicity in colorectal cancer patients: a DMET microarray profiling study. Cancer Biol Ther. 2011;12:780-787. 28. Rhodes KE, Zhang W, Yang D, Press OA, Gordon M, Vallbohmer D, et al. ABCB1, SLCO1B1 and UGT1A1 gene polymorphisms are associated with toxicity in metastatic colorectal cancer patients treated with first-line irinotecan. Drug Metab Lett. 2007;1:23-30. 29. Levesque E, Belanger AS, Harvey M, Couture F, Jonker D, Innocenti F, et al. Refining the UGT1A haplotype associated with irinotecan-induced hematological toxicity in metastatic colorectal cancer patients treated with 5- fluorouracil/irinotecan-based regimens. J Pharmacol Exp Ther. 2013;345:95-101. 30. Cecchin E, Innocenti F, D'Andrea M, Corona G, De Mattia E, Biason P, et al. Predictive role of the UGT1A1, UGT1A7, and UGT1A9 genetic variants and their

haplotypes on the outcome of metastatic colorectal cancer patients treated with fluorouracil, leucovorin, and irinotecan. J Clin Oncol. 2009;27:2457-2465. 31. Toffoli G, Cecchin E, Gasparini G, D'Andrea M, Azzarello G, Basso U, et al. Genotype-driven phase I study of irinotecan administered in combination with fluorouracil/leucovorin in patients with metastatic colorectal cancer. J Clin Oncol. 2010;28:866-871. 32. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559-575. 33. Stephens M, Donnelly P. A comparison of bayesian methods for haplotype reconstruction from population genotype data. Am J Hum Genet. 2003;73:1162- 1169. 34. Tarr PT, Tarling EJ, Bojanic DD, Edwards PA, Baldan A. Emerging new paradigms for ABCG transporters. Biochim Biophys Acta. 2009;1791:584-593. 35. Tarling EJ. Expanding roles of ABCG1 and sterol transport. Curr Opin Lipidol. 2013;24:138-146. 36. Cserepes J, Szentpetery Z, Seres L, Ozvegy-Laczka C, Langmann T, Schmitz G, et al. Functional expression and characterization of the human ABCG1 and ABCG4 proteins: indications for heterodimerization. Biochem Biophys Res Commun. 2004;320:860-867. 37. Kennedy MA, Venkateswaran A, Tarr PT, Xenarios I, Kudoh J, Shimizu N, et al. Characterization of the human ABCG1 gene: liver X receptor activates an internal promoter that produces a novel transcript encoding an alternative form of the protein. J Biol Chem. 2001;276:39438-39447. 38. Lorkowski S, Rust S, Engel T, Jung E, Tegelkamp K, Galinski EA, et al. Genomic sequence and structure of the human ABCG1 (ABC8) gene. Biochem Biophys Res Commun. 2001;280:121-131. 39. Langmann T, Porsch-Ozcurumez M, Unkelbach U, Klucken J, Schmitz G. Genomic organization and characterization of the promoter of the human ATP-binding cassette transporter-G1 (ABCG1) gene. Biochim Biophys Acta. 2000;1494:175- 180. 40. Engel T, Bode G, Lueken A, Knop M, Kannenberg F, Nofer JR, et al. Expression and functional characterization of ABCG1 splice variant ABCG1(666). FEBS Lett. 2006;580:4551-4559. 41. Kool M, de Haas M, Scheffer GL, Scheper RJ, van Eijk MJ, Juijn JA, et al. Analysis of expression of cMOAT (MRP2), MRP3, MRP4, and MRP5, homologues of the multidrug resistance-associated protein gene (MRP1), in human cancer cell lines. Cancer Res. 1997;57:3537-3547. 42. Wielinga PR, van der Heijden I, Reid G, Beijnen JH, Wijnholds J, Borst P. Characterization of the MRP4- and MRP5-mediated transport of cyclic nucleotides from intact cells. J Biol Chem. 2003;278:17664-17671. 43. Oguri T, Isobe T, Suzuki T, Nishio K, Fujiwara Y, Katoh O, et al. Increased expression of the MRP5 gene is associated with exposure to platinum drugs in lung cancer. Int J Cancer. 2000;86:95-100. 44. Stojic J, Stohr H, Weber BH. Three novel ABCC5 splice variants in human retina and their role as regulators of ABCC5 gene expression. BMC Mol Biol. 2007;8:42. 45. Akiyama Y, Fujita K, Ishida H, Sunakawa Y, Yamashita K, Kawara K, et al. Association of ABCC2 genotype with efficacy of first-line FOLFIRI in Japanese patients with advanced colorectal cancer. Drug Metab Pharmacokinet. 2012;27:325-335.

99 46. Huang L, Zhang T, Xie C, Liao X, Yu Q, Feng J, et al. SLCO1B1 and SLC19A1 gene variants and irinotecan-induced rapid response and survival: a prospective multicenter pharmacogenetics study of metastatic colorectal cancer. PLoS One. 2013;8:e77223.

100

Table 1. Genetic variations in ABC and SLC transporter genes associated with toxicity and response in metastatic colon cancer patients who received irinotecan-based regimens. Variant Main findings Cohort size Treatment Ref. ABCB1 Diarrhea (CC carriers) 56 mCRC IRI-based 23 3435C>T (I1145I) Toxicity1 (T allele) 140 mCRC FOLFIRI 24 Haplotype (*2) RR, OS 140 mCRC FOLFIRI 24 1236T/2677T/3435T 2677G>A/T (A893T/S) OS 250 mCRC FOLFIRI 25 ABCC1 IVS18–30G>C Hematological toxicity2 250 mCRC FOLFIRI 25 Hematological toxicity2 1684T>C (L562L) Non-hematological 250 mCRC FOLFIRI 25 toxicity3 2012G>T (G671V) RR 250 mCRC FOLFIRI 25 ABCC2 Toxicity, any type −1023G>A Non-hematological 250 mCRC FOLFIRI 25 toxicity 1249G>A (V417I) RR 250 mCRC FOLFIRI 25 IVS23+56T>C RR 250 mCRC FOLFIRI 25 61 mCRC −24C>T RR. PFS CC carriers FOLFIRI 45 Asians ABCC5 3′UTR 1243T>C GI toxicity4 26 mCRC IRI-based 27 ABCG1 c.974–898C>G GI toxicity4 26 mCRC IRI-based 27 ABCG2 RR −15994G>A Non-hematological 250 mCRC FOLFIRI 25 toxicity3 −15622C>T RR in T carriers 250 mCRC FOLFIRI 25 SLCO1B1 −11187G>A RR 250 mCRC FOLFIRI 25 IRI-based 521T>C Toxicity5 54 mCRC6 28 FOLFIRI RR, PFS, OS, 137 mCRC FOLFIRI. 46 388A>G (N130D) (*1b) TTF Asians mCapeIRI GI toxicity4 26 mCRC IRI-based 27

1Registered after one or two cycles, type of toxicity not specified; most common neutropenia and diarrhea. 2Hematological toxicity defined as neutropenia, anemia, leukopenia, and thrombocytopenia. 3Non-hematological toxicity defined as diarrhea, nausea, vomiting, asthenia,

101 alopecia, mucositis, anorexia, and non-neutropenic infections. 4Drug-metabolizing enzyme and transporter microarray genotyping method. 5Type of toxicity included GI (diarrhea, nausea, vomiting), neutropenia, sepsis, dyspnea, deep vein thrombosis, fatigue. 6Patient ethnicity includes Asian, Caucasian, Black, and Hispanic. Corresponding ‘rs’ numbers for genetic variations, when available: ABCB1 1236 C>T (rs1128503), 2677G>A/T (rs2032582), 3435C>T (rs1045642); ABCC1 IVS18–30C>G (rs2074087), 1684T>C (rs35605), 2012G>T (rs45511401); ABCC2 −1023G>A (rs7910642), 1249G>A (rs2273697), IVS23+56T>C (rs4148396), −24C>T (rs717620); ABCC5 3′UTR 1243A>G (rs562); ABCG1 c.974–898C>G (rs425215); ABCG2, −15994G>A (rs7699188); SLCO1B1 −11187G>A (rs4149015), 521T>C (rs4149056), 388A>G (rs2306283). Abbreviations: IRI, irinotecan; mCapeIRI, irinotecan plus capecitabine; mCRC, metastatic colorectal cancer; PFS, progression-free survival; OS, overall survival; GI, gastrointestinal; RR, response rate; TTF, time to treatment failure.

102

Table 2. Demographic and clinical characteristics of the study populations

Canadian cohort Italian cohort1 (discovery cohort) (validation cohort) Characteristics N % N % Total number 167 250 Gender (male/female) 110/57 162/88 Median age (years) 61.5 60.6 Primary tumor site Colon 122 73.1 179 71.6 Rectum 42 25.1 71 28.4 Unknown 3 1.8 - Regimen FOLFIRI 167 250 Co-treatment: Bevacizumab 69 41.3 - Other drug 6 3.6 - Toxicity Diarrhea (grade 3–4) 24 14.4 21 8.4 Neutropenia (grade 3–4) 28 16.8 33 13.0 1Demographic characteristics have been reported.31,32 Germline DNA was available for 237 patients.

103 Table 3. Haplotype-tagged SNPs associated with severe grade 3–4 diarrhea.

Canadian cohort Diarrhea Diarrhea Gene SNP Allele OR (95% CI) P Genotype OR (95% CI) P 0–2 3–4 0–2 3–4 C (ref) 160 37 CC (ref) 46 13 rs10937158 0.40 (0.19-0.81) 0.011 0.41 (0.17-1.00) 0.049 T 120 11 CT/TT 94 11 ABCC5 C (ref) 158 19 CC (ref) 48 2 rs3749438 2.08 (1.11-3.89) 0.022 5.93 (1.34-26.31) 0.019 T 116 29 CT/TT 89 22 C (ref) 161 37 CC (ref) 45 14 rs225440 0.40 (0.20-0.82) 0.012 0.34 (0.14-0.82) 0.016 T 119 11 TT/TT 95 10 G (ref) 199 28 GG/GA (ref) 125 17 rs3787997 1.90 (1.01-3.57) 0.047 4.29 (1.49-12.39) 0.007 A 75 20 AA 12 7 ABCG1 C (ref) 146 34 TT (ref) 38 12 rs3787970 0.45 (0.23-0.87) 0.018 0.37 (0.15-0.90) 0.028 T 134 14 TC/CC 102 12 G (ref) 188 39 GG (ref) 63 16 rs915847 0.46 (0.22-0.99) 0.048 0.40 (0.16-1.00) 0.051 A 94 9 GA/AA 78 8 Italian cohort Diarrhea Diarrhea Gene SNP Allele OR (95% CI) P Genotype OR (95% CI) P 0-2 3-4 0-2 3-4 C (ref) 244 29 CC (ref) 75 11 rs10937158 0.45 (0.21-0.96) 0.040 0.41 (0.16-1.07) 0.069 T 170 9 CT/TT 132 8 ABCC5 C (ref) 251 17 CC (ref) 73 2 rs3749438 1.90 (0.97-3.71) 0.060 4.63 (1.04-20.60) 0.044 T 163 21 CT/TT 134 17 C (ref) 228 24 CC (ref) 56 6 rs225440 0.63 (0.32-1.24) 0.182 0.74 (0.27-2.04) 0.560 T 212 14 TT/TT 164 13 G (ref) 266 24 GG/GA (ref) 182 16 rs3787997 0.91 (0.44-1.87) 0.800 0.95 (0.21-4.38) 0.955 A 146 12 AA 24 2 ABCG1 C (ref) 209 24 TT (ref) 48 6 rs3787970 0.53 (0.27-1.05) 0.068 0.60 (0.22-1.67) 0.333 T 231 14 TC/CC 172 13 G (ref) 267 27 GG (ref) 81 9 rs915847 0.63 (0.30-1.30) 0.211 0.65 (0.25-1.66) 0.365 A 173 11 GA/AA 139 10 Ref = reference. All SNPs were in Hardy-Weinberg equilibrium. Bold, p < 0.05; q values were all >0.05.

Table 4. Haplotype-tagged SNPs associated with severe grade 3–4 neutropenia in the discovery cohort.

Neutropenia Neutropenia Gene SNP Allele OR1 (95% CI) P Genotype OR1 (95% CI) P 0–2 3–4 0–2 3–4 G (ref) 245 45 GG (ref) 110 17 rs215064 1.88 (0.87-4.03) 0.107 2.45 (1.03-5.86) 0.044 A 31 11 GA/AA 28 11 C (ref) 260 47 CC (ref) 125 20 rs215074 2.94 (1.09-7.95) 0.034 3.64 (1.25-10.63) 0.018 T 12 7 CT/TT 11 7 ABCC1 C (ref) 240 42 CC (ref) 105 15 rs212083 1.85 (0.89-3.87) 0.101 2.50 (1.06-5.91) 0.037 T 36 12 CT/TT 33 12 C (ref) 237 44 CC (ref) 104 16 rs212088 1.71 (0.82-3.56) 0.152 2.36 (1.00-5.54) 0.049 T 37 12 CT/TT 33 12 G (ref) 221 43 GG/GA (ref) 135 24 ABCC2 rs2273697 1.28 (0.64-2.57) 0.485 7.63 (1.58-36.92) 0.012 A 55 13 AA 3 4 T (ref) 215 40 TT/TC (ref) 132 24 rs1554395 1.49 (0.75-2.96) 0.258 6.99 (1.28-38.28) 0.025 C 55 14 CC 3 3 G (ref) 187 33 GG/GC (ref) 126 21 rs1016752 1.43 (0.78-2.64) 0.251 3.63 (1.18-11.19) 0.025 C 85 21 CC 10 6 C (ref) 169 29 CC/CT (ref) 120 19 ABCC5 rs10937158 1.37 (0.76-2.48) 0.298 2.91 (1.10-7.73) 0.032 T 107 25 TT 18 8 C (ref) 144 35 CC/CT (ref) 104 24 rs3749438 0.53 (0.28-1.00) 0.048 0.26 (0.06-1.16) 0.077 T 128 17 TT 32 2 G (ref) 247 46 GG (ref) 114 18 rs2292997 2.39 (1.06-5.41) 0.037 3.15 (1.26-7.88) 0.014 A 23 10 AG/AA 21 10 C (ref) 170 29 CC (ref) 54 5 rs225440 1.60 (0.89-2.87) 0.115 3.08 (1.10-8.63) 0.033 T 104 27 CT/TT 83 23 C (ref) 236 54 CC (ref) 102 26 rs17767083 0.20 (0.05-0.85) 0.030 0.20 (0.04-0.88) 0.033 T 40 2 CT/TT 36 2 ABCG1 A (ref) 201 49 AA (ref) 73 22 rs2234718 0.39 (0.17-0.91) 0.028 0.31 (0.12-0.81) 0.017 G 73 7 AG/GG 64 6 G (ref) 195 33 GG/GA (ref) 122 21 rs3787997 1.81 (0.97-3.35) 0.062 3.13 (1.03-9.48) 0.044 A 75 21 AA 13 6

105 T (ref) 174 27 TT (ref) 57 6 ABCG2 rs12505410 1.90 (1.06-3.41) 0.031 2.70 (1.02-7.15) 0.046 100 G 29 TG/GG 80 22 A (ref) 153 25 AA/AG (ref) 112 17 rs2306283 1.66 (0.92-3.00) 0.090 3.09 (1.26-7.57) 0.014 G 125 31 GG 27 11 SLCO1B1 C (ref) 150 34 CC (ref) 37 13 rs2291076 0.66 (0.36-1.22) 0.184 0.33 (0.14-0.81) 0.015 T 120 20 CT/TT 98 14 Ref = reference. 1OR and p values were corrected for age. All SNPs were in HWE. Bold, p < 0.05; q values were all >0.05.

106

Table 5. Haplotype-tagged SNPs associated with severe grade 3–4 neutropenia in the validation cohort

Neutropenia Neutropenia Gene SNP Allele OR1 (95% CI) P Genotype OR1 (95% CI) P 0–2 3–4 0–2 3–4 G (ref) 336 53 GG (ref) 145 25 rs215064 1.12 (0.52-2.42) 0.771 0.71 (0.27-1.83) 0.473 A 50 9 GA/AA 48 6 C (ref) 357 61 CC (ref) 168 29 rs215074 0.70 (0.21-2.39) 0.568 0.76 (0.21-2.68) 0.665 T 25 3 CT/TT 23 3 ABCC1 C (ref) 318 57 CC (ref) 129 24 rs212083 0.73 (0.35-1.55) 0.415 0.76 (0.33-1.73) 0.511 T 70 9 CT/TT 65 9 C (ref) 315 57 CC (ref) 128 24 rs212088 0.70 (0.33-1.49) 0.358 0.74 (0.32-1.69) 0.473 T 71 9 CT/TT 65 9 G (ref) 292 45 GG/GA (ref) 180 30 ABCC2 rs2273697 1.37 (0.76-2.46) 0.300 1.23 (0.26-5.93) 0.795 A 88 19 AA 10 2 T (ref) 288 55 TT/TC (ref) 178 33 rs1554395 0.58 (0.29-1.16) 0.124 - 0.1363 C 98 11 CC 15 0 G (ref) 261 51 GG/GC (ref) 171 33 rs1016752 0.60 (0.32-1.11) 0.101 - 0.0313 C 127 15 CC 23 0 C (ref) 227 46 CC/CT (ref) 156 31 ABCC5 rs10937158 0.62 (0.35-1.09) 0.099 0.27 (0.06-1.19) 0.084 T 159 20 TT 37 2 C (ref) 235 33 CC/CT (ref) 166 27 rs3749438 1.44 (0.85-2.45) 0.179 1.10 (0.39-3.10) 0.856 T 153 31 TT 28 5 G (ref) 337 54 GG (ref) 149 21 rs2292997 1.63 (0.81-3.27) 0.173 2.04 (0.93-4.51) 0.077 A 47 12 GA/AA 43 12 C (ref) 226 26 CC (ref) 57 5 rs225440 1.89 (1.11-3.22) 0.019 2.21 (0.81-6.04) 0.121 T 186 40 CT/TT 149 28 C (ref) 353 59 CC (ref) 163 26 rs17767083 1.15 (0.49-2.73) 0.744 1.38 (0.55-3.47) 0.496 T 35 7 CT/TT 31 7 ABCG1 A (ref) 332 52 AA (ref) 133 21 rs2234718 1.12 (0.59-2.13) 0.722 1.05 (0.49-2.26) 0.896 G 80 14 AG/GG 73 12 G (ref) 252 38 GG/GA (ref) 170 28 rs3787997 1.20 (0.69-2.08) 0.528 0.77 (0.22-2.75) 0.688 A 134 24 AA 23 3 T (ref) 263 45 TT (ref) 83 17 ABCG2 rs12505410 0.82 (0.47-1.43) 0.485 0.63 (0.30-1.32) 0.220 G 149 21 TG/GG 123 16

107

A (ref) 223 39 AA/AG (ref) 155 27 rs2306283 0.99 (0.58-1.68) 0.955 0.94 (0.36-2.46) 0.904 G 163 27 GG 38 6 SLCO1B1 C (ref) 207 35 CC (ref) 58 11 rs2291076 0.83 (0.48-1.44) 0.513 0.73 (0.32-1.64) 0.442 T 181 27 CT/TT 136 20 Ref = reference. 1OR and p values were corrected for age; 2P-values of the univariate analysis. All SNPs were in Hardy-Weinberg equilibrium. Bold, p < 0.05; q values were all >0.05.

108

Table 6. Combined analysis of genetic markers in ABCG1 and ABCC5 in relation to severe grade 3–4 neutropenia.

Canadian cohort (n = 162) Italian cohort (n = 225) Combined cohort (n = 387)

Co-occurrence Frequency Frequency ABCG1 – ABCC5 OR (95% CI) p value OR (95% CI) p value OR (95% CI) p value (%) (%) rs225440 rs2292997 CC – GG 27.8 1.00 - 19.6 1.00 - 1.00 - CC – AA/AG 6.8 1.03 (0.14 – 7.34) 1.000 6.7 5.25 (0.92 – 29.95) 0.099 2.52 (0.69 – 9.13) 0.230 CT/TT– GG 53.1 1.99 (0.65 – 6.14) 0.295 56.0 3.73 (0.95 – 14.58) 0.107 2.55 (1.06 – 6.14) 0.039 CT/TT – AA/AG1 13.3 8.34 (2.29 – 30.71) 0.002 17.8 6.10 (1.41 – 36.45) 0.022 5.93 (2.25 – 15.59) 0.0002

1Predicted high-risk group based on individual transporter genotypes in the Canadian and Italian cohorts presented in Tables 4-5, respectively. OR, odds ratio; 95% CI, 95% confidence interval; Reference: OR = 1.00. Patients with missing genotype information were not included in the analysis.

109 Table 7. Combined analysis of genetic markers in UGT1, ABCG1, and ABCC5 in relation to severe grade 3–4 neutropenia.

Risk Canadian cohort (n = 162) Italian cohort (n = 225) Combined cohort (n = 387) markers UGT1A1*28 Neutropenia grade P Neutropenia grade P Neutropenia grade P value ABCG1 value value rs225440T ABCC5 0–2 3–4 0–2 3–4 0–2 3–4 rs2292997A (n = 134) (n = 28) (n = 192) (n = 33) (n = 326) (n = 61) 47 2 64 3 111 5 ‡ ‡ ‡ 01 (35.1%) (7.1%) 0.003 (33.3%) (9.1%) 0.005 (34.0%) (8.2%) <0.0001 OR = 1.00 OR = 1.00 OR = 1.00 48 10 57 8 105 18 (35.8%) (35.7%) 0.991‡ (29.7%) (24.2%) 0.487‡ (32.2%) (29.5%) 0.678‡ 1 OR = 4.90 0.036* OR = 2.94 0.128* OR = 3.81 0.008* (1.16-20.58) (0.81-10.73) (1.42-10.23) 39 16 71 22 110 38 (29.1%) (57.1%) 0.004‡ (37.0%) (66.7%) 0.001‡ (33.7%) (62.3%) <0.0001‡ ≥2 OR = 9.64 0.001* OR = 6.61 0.001* OR = 7.67 <0.0001* (2.39-38.92) (2.04-21.41) (3.02-19.46)

1Non-carriers of risk markers (UGT1A1*28, ABCG1 rs225440T and ABCC5 rs2292997) and/or carriers of at least one protective UGT1 rs11563250G marker. ‡T-test, *Fisher’s exact test. Reference group: OR = 1.00. Patients with missing genotype information were not included in the analysis.

Supplementary Table 1. List of htSNPs genotyped in the discovery cohort for seven transporter genes.

MAF Gene SNP Chr. Chr. position Alleles AA change (EUR 1000 genome) ABCC5 rs2176825 3 183634592 A/G 0.19 ABCC5 rs562 3 183637845 T/C 0.48 ABCC5 rs9838667 3 183664678 T/G 0.25 ABCC5 rs1554395 3 183664873 C/T 0.74 ABCC5 rs1016752 3 183665062 G/C 0.64 ABCC5 rs1464322 3 183666862 C/A 0.31 ABCC5 rs3792583 3 183683938 G/A 0.20 ABCC5 rs13093614 3 183691243 A/G 0.09 ABCC5 rs3749438 3 183705184 C/T 0.37 ABCC5 rs10937158 3 183708439 T/C 0.56 ABCC5 rs4148573 3 183714467 G/A 0.05 ABCC5 rs2292997 3 183724072 G/A 0.11 ABCG2 rs2725270 4 89007921 C/A 0.48 ABCG2 rs2231156 4 89020427 C/A 0.08 ABCG2 rs4693924 4 89023224 G/A 0.08 ABCG2 rs2725264 4 89026109 C/T 0.92 ABCG2 rs2231148 4 89028478 T/A 0.4 ABCG2 rs12505410 4 89030841 T/G 0.4 ABCG2 rs2622621 4 89030920 C/G 0.31 ABCG2 rs13120400 4 89033527 T/C 0.29 ABCG2 rs2725256 4 89050998 A/G 0.34 ABCG2 rs2231142 4 89052323 G/T Q141K 0.10 ABCG2 rs17731538 4 89055379 G/A 0.20 ABCG2 rs2231137 4 89061114 C/T V12M 0.06 ABCG2 rs2725252 4 89061910 C/A 0.53 ABCG2 rs3114018 4 89064581 A/C 0.50 ABCG2 rs17013881 4 89065076 C/T 0.07 ABCG2 rs2622626 4 89066715 C/A 0.47 ABCG2 rs2725248 4 89068007 C/A 0.72 ABCG2 rs2725246 4 89068498 G/A 0.40 ABCG2 rs9999111 4 89073197 A/C.G 0.07 ABCG2 rs2622604 4 89078924 T/C 0.77 ABCG2 rs2231135 4 89079994 A/G 0.07 ABCB1 rs6946119 7 87128865 T/C 0.24 ABCB1 rs7802783 7 87130397 C/T 0.13 ABCB1 rs3842 7 87133366 T/C 0.13 ABCB1 rs1882478 7 87137018 C/T 0.26 ABCB1 rs1045642 7 87138645 A/G.T I1145I 0.47 ABCB1 rs6949448 7 87141814 T/C 0.58

111 ABCB1 rs2032582 7 87160618 A/C.T S893T 0.57 ABCB1 rs11983225 7 87161520 T/C 0.13 ABCB1 rs12720066 7 87169702 A/C 0.06 ABCB1 rs4728701 7 87172053 A/G 0.95 ABCB1 rs4148735 7 87172881 C/T 0.43 ABCB1 rs2235013 7 87178626 C/T 0.48 ABCB1 rs2235033 7 87179143 A/G 0.48 ABCB1 rs2032588 7 87179443 G/A 0.07 ABCB1 rs1128503 7 87179601 A/G G412G 0.57 ABCB1 rs10276036 7 87180198 C/T 0.57 ABCB1 rs2235015 7 87199564 C/A 0.20 ABCB1 rs10259849 7 87200842 C/T 0.71 ABCB1 rs10264990 7 87202615 C/T 0.67 ABCB1 rs1202180 7 87203840 C/T 0.63 ABCB1 rs4148733 7 87213232 A/G 0.13 ABCB1 rs1202184 7 87213901 C/T 0.50 ABCB1 rs17327624 7 87216817 G/T 0.18 ABCB1 rs3789243 7 87220886 A/G 0.48 ABCB1 rs9282564 7 87229440 T/C N21D 0.09 ABCB1 rs2214102 7 87229501 T/C 0.90 ABCB1 rs4148732 7 87234049 T/C 0.10 ABCB1 rs13233308 7 87244960 C/T 0.48 ABCB1 rs10267099 7 87278760 G/A 0.75 ABCB1 rs6465118 7 87330423 G/A 0.06 ABCC2 rs1885301 10 101541053 A/G 0.58 ABCC2 rs2804402 10 101541583 A/G 0.42 ABCC2 rs717620 10 101542578 C/T 0.21 ABCC2 rs2804398 10 101558634 A/T 0.36 ABCC2 rs2756109 10 101558746 G/T 0.55 ABCC2 rs2273697 10 101563815 G/A V417I 0.20 ABCC2 rs2002042 10 101587931 C/T 0.27 ABCC2 rs4148397 10 101592320 A/G 0.56 ABCC2 rs17216177 10 101603522 T/C 0.06 ABCC2 rs3740066 10 101604207 C/T I1324I 0.36 ABCC2 rs3740063 10 101610723 A/G 0.43 ABCC2 rs8187710 10 101611294 G/A C1515Y 0.06 ABCC2 rs7067971 10 101616579 A/G 0.37 SLCO1B1 rs4149015 12 21283322 G/A 0.06 SLCO1B1 rs2010668 12 21294293 T/G 0.93 SLCO1B1 rs4149022 12 21295612 G/A 0.26 SLCO1B1 rs2417955 12 21296475 T/A 0.63 SLCO1B1 rs11045797 12 21309714 T/C 0.18 SLCO1B1 rs16923519 12 21311718 A/G 0.22

112

SLCO1B1 rs7953338 12 21314572 T/C 0.91 SLCO1B1 rs4149033 12 21317810 A/G 0.77 SLCO1B1 rs4149035 12 21318265 T/C 0.62 SLCO1B1 rs2199766 12 21319526 G/A 0.32 SLCO1B1 rs7139376 12 21321429 T/C 0.89 SLCO1B1 rs2306283 12 21329738 A/G N130D 0.40 SLCO1B1 rs4149056 12 21331549 T/C V174A 0.17 SLCO1B1 rs4149057 12 21331599 T/C L191L 0.60 SLCO1B1 rs2291076 12 21331987 C/T 0.46 SLCO1B1 rs11045821 12 21332423 G/A 0.14 SLCO1B1 rs6487213 12 21333266 C/T 0.43 SLCO1B1 rs991262 12 21334214 G/A 0.05 SLCO1B1 rs2900476 12 21336063 C/T 0.75 SLCO1B1 rs11045834 12 21341096 C/T 0.26 SLCO1B1 rs1564365 12 21354470 G/A 0.92 SLCO1B1 rs12427008 12 21357948 T/C 0.07 SLCO1B1 rs4363657 12 21368722 T/C 0.21 SLCO1B1 rs4149070 12 21369883 C/G 0.08 SLCO1B1 rs12830367 12 21388905 G/T 0.37 SLCO1B1 rs12371604 12 21391336 T/C 0.26 SLCO1B1 rs10841769 12 21395019 G/A 0.47 ABCC1 rs4148330 16 16041768 G/A 0.66 ABCC1 rs215108 16 16046257 C/T 0.95 ABCC1 rs4781699 16 16049633 T/G 0.67 ABCC1 rs215101 16 16052973 C/G 0.18 ABCC1 rs215100 16 16054650 G/A 0.66 ABCC1 rs215098 16 16054844 G/T 0.75 ABCC1 rs169985 16 16060881 C/T 0.66 ABCC1 rs12922404 16 16060994 C/T 0.17 ABCC1 rs215088 16 16065401 A/G 0.71 ABCC1 rs6498594 16 16073437 A/C 0.13 ABCC1 rs215064 16 16077972 A/G 0.84 ABCC1 rs17501011 16 16081123 G/A 0.04 ABCC1 rs215074 16 16082301 G/A 0.07 ABCC1 rs152023 16 16085236 C/T 0.66 ABCC1 rs17501331 16 16089441 A/G 0.10 ABCC1 rs246217 16 16090354 T/G 0.84 ABCC1 rs12927980 16 16090651 G/T 0.08 ABCC1 rs9635480 16 16094601 A/G 0.54 ABCC1 rs11075289 16 16098368 T/C 0.64 ABCC1 rs4781711 16 16100380 A/G 0.09 ABCC1 rs8187843 16 16101875 G/A 0.06 ABCC1 rs8058040 16 16107712 A/G 0.16

113 ABCC1 rs3784862 16 16110891 G/A 0.72 ABCC1 rs8059648 16 16111437 C/G 0.08 ABCC1 rs246214 16 16114892 C/T 0.18 ABCC1 rs17264736 16 16116556 G/T 0.55 ABCC1 rs152033 16 16117271 T/C 0.89 ABCC1 rs875740 16 16123048 C/A 0.63 ABCC1 rs924136 16 16126475 G/A 0.49 ABCC1 rs246233 16 16129251 T/G 0.11 ABCC1 rs903880 16 16130514 C/A 0.23 ABCC1 rs8054670 16 16132134 T/C 0.22 ABCC1 rs171583 16 16132332 A/G 0.32 ABCC1 rs246230 16 16132880 C/T 0.15 ABCC1 rs35588 16 16139878 A/G 0.32 ABCC1 rs17205859 16 16141738 C/T 0.06 ABCC1 rs35592 16 16141823 T/C 0.27 ABCC1 rs3765129 16 16149901 C/T 0.16 ABCC1 rs17287570 16 16155103 A/C 0.21 ABCC1 rs9932506 16 16161425 G/A 0.48 ABCC1 rs4148348 16 16161928 G/A 0.09 ABCC1 rs35605 16 16162019 T/C L562L 0.82 ABCC1 rs8187858 16 16162039 C/T Y568Y 0.1 ABCC1 rs4148350 16 16170477 G/T 0.06 ABCC1 rs35626 16 16170615 G/T 0.32 ABCC1 rs35628 16 16171106 A/G 0.09 ABCC1 rs35629 16 16171442 C/T 0.11 ABCC1 rs4148355 16 16174667 A/G 0.15 ABCC1 rs10852377 16 16176824 T/C 0.78 ABCC1 rs11075295 16 16177687 G/A 0.84 ABCC1 rs2889517 16 16181956 T/C 0.67 ABCC1 rs3888565 16 16183045 G/A 0.16 ABCC1 rs2074087 16 16184232 C/G 0.84 ABCC1 rs4148358 16 16187175 C/T 0.23 ABCC1 rs4148361 16 16187469 C/T 0.52 ABCC1 rs11864374 16 16201885 G/A 0.25 ABCC1 rs3784867 16 16203345 C/T 0.2 ABCC1 rs3887893 16 16205501 T/C 0.41 ABCC1 rs12927259 16 16208171 G/T 0.2 ABCC1 rs4148373 16 16212104 C/T 0.05 ABCC1 rs2238475 16 16214255 T/G 0.06 ABCC1 rs212079 16 16220126 C/T 0.06 ABCC1 rs2283512 16 16220202 A/C 0.69 ABCC1 rs212081 16 16225971 G/A 0.36 ABCC1 rs212083 16 16227943 G/A 0.17

114

S1334S S1219S ABCC1 rs2230671 16 16228242 G/A.C S1275S 0.29 S1278S S1204S ABCC1 rs212085 16 16229238 C/G 0.43 ABCC1 rs212088 16 16232433 G/A 0.16 ABCC1 rs4148380 16 16236431 G/A 0.06 ABCC1 rs212091 16 16236650 T/C 0.14 ABCC1 rs212093 16 16237754 A/G 0.43 ABCC1 rs4148382 16 16238494 G/A 0.10 ABCC1 rs8053988 16 16241252 T/G 0.16 ABCG1 rs1117640 21 43625274 T/G 0.13 ABCG1 rs221948 21 43633193 A/G 0.89 ABCG1 rs3787970 21 43651556 C/T 0.44 ABCG1 rs225440 21 43653053 C/T 0.39 ABCG1 rs915847 21 43657248 G/A 0.33 ABCG1 rs9978671 21 43660279 C/T 0.05 ABCG1 rs4148117 21 43666065 T/C 0.43 ABCG1 rs225374 21 43671633 C/G 0.42 ABCG1 rs225376 21 43674122 G/C 0.70 ABCG1 rs7281093 21 43674859 G/A 0.23 ABCG1 rs225378 21 43675694 A/G 0.53 ABCG1 rs7281684 21 43676586 A/T 0.17 ABCG1 rs183436 21 43677831 C/A 0.70 ABCG1 rs17767083 21 43684321 C/T 0.12 ABCG1 rs225391 21 43684668 G/A 0.09 ABCG1 rs225398 21 43688594 G/C 0.67 ABCG1 rs532345* 21 43689491 C/T 0.41 ABCG1 rs2234718 21 43697195 A/G 0.24 ABCG1 rs182694 21 43698815 A/G 0.64 ABCG1 rs3787995 21 43698914 G/C 0.09 ABCG1 rs3787997 21 43699096 G/A 0.32 ABCG1 rs225406 21 43699230 G/A 0.11 ABCG1 rs225410 21 43700907 C/T 0.52 ABCG1 rs3788007 21 43706776 G/A 0.20 ABCG1 rs450808 21 43706944 T/C 0.79 ABCG1 rs425215 21 43707101 C/G 0.64 ABCG1 rs4148137 21 43710052 C/T 0.38 ABCG1 rs914189 21 43710909 C/G 0.20 ABCG1 rs13050646 21 43712049 G/A.T 0.00 ABCG1 rs9975333 21 43714517 A/C 0.58 ABCG1 rs474142 21 43714579 G/A 0.03

115 ABCG1 rs1044317 21 43716901 A/G 0.55 ABCG1 rs1541290 21 43718483 G/A 0.51 ABCG1 rs6586304 21 43719965 G/A 0.58 *Not in Hardy-Weinberg equilibrium. MAF, Minor Allele Frequency

116

FIGURE LEGENDS

Figure 1. Schematic representation of a two-marker ABCC5 haplotype comprising variants linked to severe grade 3–4 toxicity (rs10937158C and rs3749438T; olive-green rectangles) compared with the reference allele (rs10937158T and rs3749438C; yellow rectangles) and their predictive values for severe diarrhea in the Canadian and Italian cohorts. HI-HIII, haplotype I-III; OR, odds ratio; 95% CI, 95% confidence interval.

Figure 2. Schematic representation of the four markers linked to severe diarrhea and neutropenia and replicated in both cohorts. Upper: Schematic representation of the ABCC5 locus (NM_005688) and location of the rs3749438 marker and rs10937158 marker associated with severe diarrhea and the rs2292997 marker associated with severe neutropenia. Lower: The ABCG1 locus (NM_207627–28) depicting variant (rs225540) associated with severe neutropenia.

Supplementary Figure 1. Schematic representation of haplotype-tagged SNPs of markers for which r2 > 0.8. For the ABCC5 locus (NM_005688), the markers include rs10937158 (Figure S1a), rs3749438 (Figure S1b), and rs2292997 (Figure S1c). For the ABCG1 locus (NM_207627–28), the marker is rs225440 (Figure S1d).

117 Figure 1

118

Figure 2

119 Supplementary Figure S1a

120

Supplementary Figure S1b

121 Supplementary Figure 1c

122

Supplementary Figure 1d

123

124

Discussion

As a chemotherapeutic agent, irinotecan possesses a narrow therapeutic window that often results in at least 25% of the patient population with dose-limiting toxicities. The body of work presented in this thesis demonstrates the importance of germline genetic variants in metabolism and transporter genes in predicting drug-induced toxicities with two independent patient cohorts.

The most widely studied irinotecan associated drug-metabolizing enzyme, UGT1A1 has been repeatedly shown to regulate SN-38 AUC levels and thereby predict clinical outcomes, especially severe neutropenia incidents.75-78 Existing explorations have investigated a handful of SNPs that induced an effect on SN-38 glucuronidation in vitro but none have successfully replicated their findings in patient cohorts. Replication of results in independent patient populations is essential as it eliminates false positives and allow for greater acceptance by regulatory bodies. In light of this limitation, we conducted a comprehensive search on the UGT1 locus, employing the use of a haplotype-tagging strategy to maximize genetic coverage and lower costs, in two cohorts. The results presented in the first article refines current knowledge about UGT1A1*28 and its ability to predict severe neutropenia. A novel variant, rs11563250, located in the 3’ flanking region of UGT1, was identified to induce elevated UGT1A1 activity. Carriers of the minor G allele exhibited a lower tendency to develop severe neutropenia and higher bilirubin clearance. When assessing its effect with a known neutropenia predictor, UGT1A1*28 in a two- marker haplotype, we observed that patients who were non-carriers of *28 and carriers of G allele were less likely to have neutropenia (OR=0.17, p=0.0004).

Current UGT1A1 genotyping clinical practices involve a dose-reduction in irinotecan by one level for patients found with the *28/*28 genotype and the maximum tolerated dose by non-carriers of *28 is actually higher than the usual 180mg/m2 dose. The two-marker 6G haplotype allows for an improved prediction of adverse events and also gives clinicians a greater flexibility in increasing dose to maximize tumor response. The inclusion of this novel marker, approximately 1404bp downstream from 3’UTR of UGT1 locus, in available UGT1A1*28 tests such as FDA-approved Invader Assay UGT1A1279,306 could help identify specific subpopulations that would benefit from a larger dose (non-carriers of the *28 allele + carriers of the rs11563250G allele).

125 It is worth noting that despite previous reports purporting the role of UGT1A1*28 in predicting irinotecan-induced diarrhea events, our analyses did not detect any significant associations between UGT1 variants and severe diarrhea risks in both cohorts studied. Therefore, we hypothesized that other genes involved in irinotecan pharmacokinetic pathways may have induced these toxicities instead.

Indeed, the inter-relationship between drug transporters (ABCB1) and drug-metabolizing enzymes (CYP3A) to affect drug response has been previously suggested.100 The substrate specificities of both the enzyme and transporter type can help predict the potential interactions that may exist in modulating drug outcome. At present, substantial but yet unclarified reports have suggested that genetic alterations in ABC transporter genes may in fact explain irinotecan-induced toxicities. (Table 6)

Using the same genotyping strategy and approaches, we identified several novel transporter variants that were linked to Grade 3-4 neutropenia and diarrhea. A diplotype of two ABCC5 variants (rs10937158T and rs3749438C) was associated with a greater tolerance towards severe diarrhea (OR=0.39, p=0.0006). This transporter type has previously been studied in relation to irinotecan-induced gastrointestinal toxicities124 and was most recently associated with decreased SN-38G hepatic efflux.139 Our overall results support evidence of ABCC5 as a key transporter of irinotecan-metabolites and its ability to predict severe diarrhea. We also observed from the co-occurrence analyses of ABCG1 and ABCC5 variants that a cooperative effect may be present during SN-38 transport, thereby increasing active metabolite exposure and elevating neutropenia risk. Carriers of ABCG1 rs225540T and ABCC5 rs2292997A exhibited a six-fold increase risk (OR=5.93, p=0.0002) compared to non-carriers.

From our findings, we inferred the possibility of multiple transporters overlapping in function to transport irinotecan and its metabolites and thus evaluated the synergistic impact of UGT1A1*28 with ABCG1 rs225540T and ABCC5 rs2292997A on predicting neutropenia. We noted a seven-fold increase in Grade 3-4 neutropenia risks in patients who carried two or more risks alleles (OR=7.67, p<0.0001) compared to non carriers of risk alleles and carriers of the protective UGT1 rs11563250G variant, indicating strong evidence of a UGT1A and ABC transporter cooperation.

126

So far, a pharmacological cell model has depicted interplay between UGT1A9 and ABCG2 functions in the elimination of flavonoids307 while the genetic transactivation of CYP2A13 and ABCB1 promoters by the same transcription factor resulted in overall decrease in lung cancer risks.308 This form of cooperation is further complicated by the overlap in substrates of different transporters and metabolizing enzymes, supporting the idea of multiple pathways interacting during drug disposition. Hence, it is also conceivable that interactions between drug transporter and metabolism pathways have the propensity to contribute towards predicting overall clinical outcomes together.

In terms of irinotecan-mediated pathways, the intrinsic functions of transporters can be more easily clarified with substrate-binding or cellular-transport experiments whereas the role of genes in pharmacodynamic pathways is less straightforward. The downstream signaling pathways of irinotecan, specifically from the effect of SN-38-mediated DNA breakage, DNA repair and apoptosis would more likely impact tumor response and, to a lesser extent, toxicities. For instance, a high expression of TDP1 was found to be protective against irinotecan-mediated cytotoxicity in colorectal cancer cell lines and may be useful in conjunction with TOP1 as a diagnostic biomarker for assessing tumor response.309 It is therefore unsurprising that we failed to observe any positive markers from these genes validated against severe neutropenia and diarrhea in our study, which is in accordance with existing reports (data not shown).

Consequently, the next step in this project would be to determine whether these pharmacodynamic genes are associated with clinical response and survival parameters such as overall survival and progression-free survival. The prevailing consensus on pharmacodynamic gene variants and their association with response rate appear to be limited while the definite impact of ABC transporter gene variants remains contradictory. This may be in part due to the differences in endpoint measurements, presence of co- treatments or modalities used. It is crucial that future genetic association studies take all these factors into account and guidelines should be placed to allow for better study comparisons.310

In order to make meaningful interpretations about complex biological interactions, researchers should consider utilizing a combination of pathway-based and multiple candidate genes approach in genetic association studies. Even though single-SNP

127 association tests are considered the gold standard in assessing implications of genetic variants in PGx and genomic studies, an alternative approach would be to combine SNPs from the same locus and analyze them as a haplotype block. Numerous studies have already documented the advantages of this approach over a single-SNP association analysis as it increases statistical power and accounts for genetic transmission via chromosomal recombination.311-313 Lastly, it allows for the combined effect of multiple functional SNPs interacting in cis to be evaluated against each independent outcome; which will be undetected in single-SNP association studies. Ultimately, haplotype-based analyses can prove to be useful in developing multi-marker genetic tests, as they tend to represent the predictive or causal variants that induce an overall phenotypic effect.

The importance of replicating these positive variants in other independent cohorts is to ensure a broader application of clinical tests for majority of the patient population. Healthcare agencies also require clear and validated results before implementing into clinical applications. A multi-marker genetic test that has high analytical validity measured in different ethnic populations could be expanded for use globally, thereby allowing manufacturers to justify costs and capture greater market shares. For instance, UGT1A1*28 genotyping test was found to be less cost-effective in Asians due to a lower allelic frequency (<16%) in this population.301,314 On the other hand, UGT1A1*6 occurs in higher frequencies (13-23%) in Asian samples and these patients would benefit from a test more specific to this variant.

Interestingly, even though the clinical populations studied herein have been composed of Caucasians, a novel diplotype UGT1A1 test (UGT1A1*28/rs11563250G) could be applicable for use in patients of Hispanic/Latino descent given that these populations tend to have higher MAFs for UGT1A1*28 (18%) and rs11563250 (18%).315,316 The current FDA-approved Invader Assay UGT1A1 test detects fluorescent signals based on different cleavage reactions in the presence of a SNP.279 The developers of this test confirmed that multiple sets of probes could be incorporated to detect other UGT1 variants. Hence, it is reasonable to assume that a two-marker UGT1A1 test can be develop to include the protective variant rs11563250G using this technology or other approaches.

While the two-marker UGT1 testing would indeed be useful in identifying patients more tolerant to irinotecan treatment, genotyping of both UGT1 and ABC transporter variants to predict a seven-fold neutropenia risk would be equally beneficial. The FDA-approved

128

Verigene® warfarin Metabolism Nucleic Acid Test determines warfarin sensitivity based on genotyping variants, CYP2C9*1/*2/*3 and VKORC1 -1639G>A. This molecular technique could be designed to specifically detect risk alleles of UGT1A1*28, ABCG1 rs225540T and ABCC5 rs2292997A and protective allele, UGT1 rs11563250G. Ultimately, the clinical uses of these genotyping tests, likely to be preemptive in the future317, would greatly improve clinicians’ ability to make dose-management decisions.

Even though the warfarin sensitivity tests has been generally accepted by most regulatory agencies, the European Organization for warfarin research (EU-PACT) uses a nonlinear mixed-effect (NLME) model to assess warfarin disposition parameters instead, as it was deemed to be more accurate than genotype-guided dosing in local populations. Accordingly, a complementary model combining genetic association with pharmacokinetic measurements has been derived to enhance stratification of patients based on their genetic and pharmacokinetic profiles.318 Since severe neutropenia is strongly correlated to the high levels of systemic SN-38152,319 and low SN-38G/SN-38 ratios are suggestive of a dysfunctional UGT1A-mediated metabolism320, it would be worthwhile for future clinical diagnostic tests to incorporate both genotyping and pharmacokinetic data, cross-validating both biological observations at the same time and providing a mechanistic explanation for the observed association.

In addition to irinotecan, the family of UGT1A enzymes are responsible for the detoxification of other drug substrates or compounds, such as simvastatin, mycophenolic acid and others;74,321 whereas the ABC transporters are involved in the influx and efflux of multiple drug molecules (Table 5). The protective marker, UGT1 rs11563250G mentioned here appears to correlate to distinct mycophenolic acid glucuronide (MPAG) concentration profiles in carriers and non-carriers (data not shown). As multiple UGT1A enzymes are involved in mycophenolic acid (MPA) conversion to MPAG322, one could make a similar inference about an increased drug tolerance in rs11563250G carriers with regards to hematological adverse events (ex. leukopenia) experienced by patients receiving mycophenolate mofetil (MMF); the prodrug of MPA.323 Separately, the ABCG1/ABCC5 two-marker haplotype may also be extended to detect other toxicities such as myelosuppression, which is induced by chemotherapy agents, mitoxantrone and doxorubicin. Since cellular transport of both antineoplastic drugs is mediated by ABCC5 and ABCG1, we could potentially predict the level of risks in hematological toxicities based

129 on the genotypes of ABCG1 rs225440 and ABCC5 rs2292997, and reduce drug dosage at the start of treatment. Our findings might thus be informative for other drugs.

Although the significant genetic associations reported here have important implications in predicting toxicity risk and tolerance to irinotecan, future functional validations are necessary to understand their mechanistic attributes. Preliminary investigations using in- silico bioinformatic tools were undertaken to determine the potential effects of these variants on enzyme/protein function. The ENCODE project data available depicts rs11563250 to be located in an intergenic region, closer to the UGT1 locus. A strong polycomb repressed state has been detected in liver cell lines (HepG2) while DNAse hypersensitivity I clusters were found in only two cell types, one of which was in hepatocytes. Taken together, this variant is in an open chromatin region (active), possibly a target site of a regulatory protein.324-326

Similar approaches could be made to assess the regulatory role of ABC transporter variants outlined; especially those located in the 3’UTR or promoter regions. Experiments measuring the allele-specific expression of these intronic SNPs may help clarify their putative regulatory functions. So far, annotation biotools, (ex. HaploReg) have suggested these variants may modulate functions of regulatory elements such as microRNAs, enhancers, silencers etc. Few techniques have emerged to aid in this line of investigation. Some have combined allele-specific expression (ASE) and expression quantitative trait loci (eQTL) in mice samples to determine inheritance patterns of genes and SNPs327 while others sought to identify regulatory elements responsible for inducing ASE328. The accelerated growth of NGS in genomics-based applications has fueled the use of ChIP- SNP technology, a method incorporating ChIP-seq with SNP genotyping, to identify allele- specific SNPs in enriched genomic samples.329,330

However, the vast information available in the ENCODE project also mandates researchers to exercise some caution when performing analyses and making interpretations. This is because only certain cell types have been studied and observations from ChIP-seq, DNAse I hypersensitivity experiments, transcription factor binding etc, are often reported in specific cell lines. Hence, it is crucial to identify the cell types that are the best representatives of one’s project. Finally, observations from cellular models may in fact contradict in silico predictions124 and should be carefully validated by other types of experiments before relating its clinical implications.

130

Conclusion

We have demonstrated the significance and clinical relevance of germline variations in the drug-metabolizing enzyme UGT1A and transporters ABCC5 and ABCG1 in predicting drug toxicity, more specifically neutropenia and diarrhea, in irinotecan-treated mCRC patients. The two cohorts studied here clearly depict the necessity of replication in association studies, though validations in other cohorts as well as mechanistic experiments are warranted. These observations highlight the importance of genetic variations in the modulation of SN-38 detoxification and transport. They also help explain the interindividual variability that results in adverse drug effects. The role of pharmacogenomics in improving dose-management decisions has allowed for a better stratification of patient populations who are most likely to benefit from drug response while reducing severe toxicity prior to initiation of treatment. This approach would eventually save time and hospitalization costs in the event of ineffective drug-response and dose-limiting toxicities. Lastly, it is evident from our work that irinotecan-induced drug toxicity can hardly be explained by a single genetic marker. Instead, variations at the level of metabolism and transport are evidently inferred to work in concert to modulate overall drug exposure. The identification of these markers has brought us a step closer towards personalizing irinotecan-based treatments.

131

Bibliography

1 Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2015. CA Cancer J Clin 65, 5-29 (2015). 2 Brenner, H., Kloor, M. & Pox, C. P. P. Colorectal cancer. The Lancet 383, 1490- 1502 (2014). 3 Canadian Cancer Society. Colorectal Cancer 5-year survival risks, (2014) Last accessed. Nov 20, 2014 4 Van Cutsem, E. et al. Cetuximab and chemotherapy as initial treatment for metastatic colorectal cancer. N Engl J Med 360, 1408-1417 (2009). 5 De Roock, W. et al. Effects of KRAS, BRAF, NRAS, and PIK3CA mutations on the efficacy of cetuximab plus chemotherapy in chemotherapy-refractory metastatic colorectal cancer: a retrospective consortium analysis. Lancet Oncol 11, 753-762 (2010). 6 Bokemeyer, C. et al. Fluorouracil, Leucovorin, and Oxaliplatin With and Without Cetuximab in the First-Line Treatment of Metastatic Colorectal Cancer. J Clin Oncol 27, 663-671 (2009). 7 Karapetis, C. S. et al. K-ras mutations and benefit from cetuximab in advanced colorectal cancer. N Engl J Med 359, 1757-1765 (2008). 8 Kohne, C. H. & Lenz, H. J. Chemotherapy with targeted agents for the treatment of metastatic colorectal cancer. Oncologist 14, 478-488 (2009). 9 Van Cutsem, E., Cervantes, A., Nordlinger, B., Arnold, D. & Group, E. G. W. Metastatic colorectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol 25 Suppl 3, iii1-9 (2014). 10 National Cancer Institute. Stage IV and Recurrent Colon Cancer Treatment, (2015) Last accessed. April 20, 2015 11 Nelson, M. A., Shetty, S., Kulakodlu, M., Harley, C. & Seal, B. A comparison of mortality and costs associated with FOLFOX versus FOLFIRI in stage IV colorectal cancer. J Med Econ 14, 179-186 (2011). 12 Abrams, T. A. et al. Chemotherapy usage patterns in a US-wide cohort of patients with metastatic colorectal cancer. J Natl Cancer Inst 106, djt371 (2014). 13 Zhao, Z. et al. Patterns of treatment with chemotherapy and monoclonal antibodies for metastatic colorectal cancer in Western Europe. Curr Med Res Opin 28, 221- 229 (2012). 14 Hess, G. P., Wang, P. F., Quach, D., Barber, B. & Zhao, Z. Systemic Therapy for Metastatic Colorectal Cancer: Patterns of Chemotherapy and Biologic Therapy Use in US Medical Oncology Practice. J Oncol Pract 6, 301-307 (2010). 15 Kohne, C. H. et al. Phase III Study of Weekly High-Dose Infusional Fluorouracil Plus Folinic Acid With or Without Irinotecan in Patients With Metastatic Colorectal Cancer: European Organisation for Research and Treatment of Cancer Gastrointestinal Group Study 40986. J Clin Oncol 23, 4856-4865 (2005). 16 Saltz, L. B. et al. Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer. Irinotecan Study Group. N Engl J Med 343, 905-914 (2000). 17 Douillard, J. Y. et al. Irinotecan combined with fluorouracil compared with fluorouracil alone as first-line treatment for metastatic colorectal cancer: a multicentre randomised trial. The Lancet 355, 1041-1047 (2000). 18 Zafar, S. Y. et al. Chemotherapy use and patient treatment preferences in advanced colorectal cancer. Cancer 119, 854-862 (2012).

133 19 Inoue, Y., Toiyama, Y., Tanaka, K., Miki, C. & Kusunoki, M. A Comprehensive Comparative Study on the Characteristics of Colorectal Cancer Chemotherapy. Jpn J Clin Oncol 39, 367-375 (2009). 20 Sawada, S., Yokokura, T. & Miyasaka, T. Synthesis of CPT-11 (irinotecan hydrochloride trihydrate). Ann N Y Acad Sci 803, 13-28 (1996). 21 Nitta, K. et al. [Antitumor activity of new derivatives of camptothecin]. Gan To Kagaku Ryoho 14, 850-857 (1987). 22 Tsuruo, T., Matsuzaki, T., Matsushita, M., Saito, H. & Yokokura, T. Antitumor effect of CPT-11, a new derivative of camptothecin, against pleiotropic drug-resistant tumors in vitro and in vivo. Cancer Chemother Pharmacol 21, 71-74 (1988). 23 Matsuzaki, T., Yokokura, T., Mutai, M. & Tsuruo, T. Inhibition of spontaneous and experimental metastasis by a new derivative of camptothecin, CPT-11, in mice. Cancer Chemother Pharmacol 21, 308-312 (1988). 24 Kawato, Y. et al. Antitumor activity of a camptothecin derivative, CPT-11, against human tumor xenografts in nude mice. Cancer Chemother Pharmacol 28, 192-198 (1991). 25 Wall, M. E. et al. Plant antitumor agents. I. The isolation and structure of camptothecin, a novel alkaloidal leukemia and tumor inhibitor from camptotheca acuminata1, 2. J Am Chem Soc 88, 3888-3890 (1966). 26 Gallo, R. C., Whang-Peng, J. & Adamson, R. H. Studies on the antitumor activity, mechanism of action, and cell cycle effects of camptothecin. J Natl Cancer Inst 46, 789-795 (1971). 27 Hsiang, Y. H. & Liu, L. F. Identification of mammalian DNA topoisomerase I as an intracellular target of the anticancer drug camptothecin. Cancer Res 48, 1722-1726 (1988). 28 Hsiang, Y. H., Hertzberg, R., Hecht, S. & Liu, L. F. Camptothecin induces protein- linked DNA breaks via mammalian DNA topoisomerase I. J Biol Chem 260, 14873- 14878 (1985). 29 Dodds, H. M., Haaz, M. C., Riou, J. F., Robert, J. & Rivory, L. P. Identification of a new metabolite of CPT-11 (irinotecan): pharmacological properties and activation to SN-38. J Pharmacol Exp Ther 286, 578-583 (1998). 30 Kunimoto, T. et al. Antitumor activity of 7-ethyl-10-[4-(1-piperidino)-1- piperidino]carbonyloxy-camptothec in, a novel water-soluble derivative of camptothecin, against murine tumors. Cancer Res 47, 5944-5947 (1987). 31 Tournigand, C. FOLFIRI Followed by FOLFOX6 or the Reverse Sequence in Advanced Colorectal Cancer: A Randomized GERCOR Study. J Clin Oncol 22, 229-237 (2003). 32 Maiello, E. et al. FOLFIRI regimen in advanced colorectal cancer: the experience of the Gruppo Oncologico dell'Italia Meridionale (GOIM). Ann Oncol 16, iv56- iv60 (2005). 33 Kawato, Y., Aonuma, M., Hirota, Y., Kuga, H. & Sato, K. Intracellular roles of SN- 38, a metabolite of the camptothecin derivative CPT-11, in the antitumor effect of CPT-11. Cancer Res 51, 4187-4191 (1991). 34 Kojima, A., Shinkai, T. & Saijo, N. Cytogenetic effects of CPT-11 and its active metabolite, SN-38 on human lymphocytes. Jpn J Clin Oncol 23, 116-122 (1993). 35 Tanizawa, A., Fujimori, A., Fujimori, Y. & Pommier, Y. Comparison of topoisomerase I inhibition, DNA damage, and cytotoxicity of camptothecin derivatives presently in clinical trials. J Natl Cancer Inst 86, 836-842 (1994). 36 Jaxel, C., Capranico, G., Kerrigan, D., Kohn, K. W. & Pommier, Y. Effect of local DNA sequence on topoisomerase I cleavage in the presence or absence of camptothecin. J Biol Chem 266, 20418-20423 (1991).

134

37 Hertzberg, R. P., Caranfa, M. J. & Hecht, S. M. On the mechanism of topoisomerase I inhibition by camptothecin: evidence for binding to an enzyme- DNA complex. Biochemistry 28, 4629-4638 (1989). 38 Jaxel, C., Kohn, K. W., Wani, M. C., Wall, M. E. & Pommier, Y. Structure-activity study of the actions of camptothecin derivatives on mammalian topoisomerase I: evidence for a specific receptor site and a relation to antitumor activity. Cancer Res 49, 1465-1469 (1989). 39 Covey, J. M., Jaxel, C., Kohn, K. W. & Pommier, Y. Protein-linked DNA strand breaks induced in mammalian cells by camptothecin, an inhibitor of topoisomerase I. Cancer Res 49, 5016-5022 (1989). 40 Pommier, Y. Camptothecins and topoisomerase I: a foot in the door. Targeting the genome beyond topoisomerase I with camptothecins and novel anticancer drugs: importance of DNA replication, repair and cell cycle checkpoints. Curr Med Chem Anticancer Agents 4, 429-434 (2004). 41 Kaneda, N., Nagata, H., Furuta, T. & Yokokura, T. Metabolism and pharmacokinetics of the camptothecin analogue CPT-11 in the mouse. Cancer Res 50, 1715-1720 (1990). 42 Mathijssen, R. H. et al. Clinical pharmacokinetics and metabolism of irinotecan (CPT-11). Clin Cancer Res 7, 2182-2194 (2001). 43 Ramchandani, R. P. et al. The Role of SN-38 Exposure, UGT1A1*28Polymorphism, and Baseline Bilirubin Level in Predicting Severe Irinotecan Toxicity. J Clin Pharmacol 47, 78-86 (2013). 44 Xie, R., Mathijssen, R. H., Sparreboom, A., Verweij, J. & Karlsson, M. O. Clinical Pharmacokinetics Of Irinotecan And Its Metabolites In Relation With Diarrhea. Clin Pharmacol Ther 72, 265-275 (2002). 45 Tsuji, T. et al. CPT-11 converting enzyme from rat serum: purification and some properties. J Pharmacobiodyn 14, 341-349 (1991). 46 Senter, P. D., Beam, K. S., Mixan, B. & Wahl, A. F. Identification and activities of human carboxylesterases for the activation of CPT-11, a clinically approved anticancer drug. Bioconjug Chem 12, 1074-1080 (2001). 47 Humerickhouse, R., Lohrbach, K., Li, L., Bosron, W. F. & Dolan, M. E. Characterization of CPT-11 hydrolysis by human liver carboxylesterase isoforms hCE-1 and hCE-2. Cancer Res 60, 1189-1192 (2000). 48 Hatfield, M. J. et al. Organ-specific carboxylesterase profiling identifies the small intestine and kidney as major contributors of activation of the anticancer prodrug CPT-11. Biochem Pharmacol 81, 24-31 (2011). 49 Xu, G., Zhang, W., Ma, M. K. & McLeod, H. L. Human carboxylesterase 2 is commonly expressed in tumor tissue and is correlated with activation of irinotecan. Clin Cancer Res 8, 2605-2611 (2002). 50 Hosokawa, M. et al. Interindividual variation in carboxylesterase levels in human liver microsomes. Drug Metab Dispos 23, 1022-1027 (1995). 51 Shingyoji, M. et al. In vitro conversion of irinotecan to SN-38 in human plasma. Cancer Sci 95, 537-540 (2004). 52 Sanghani, S. P. et al. Carboxylesterases expressed in human colon tumor tissue and their role in CPT-11 hydrolysis. Clin Cancer Res 9, 4983-4991 (2003). 53 Cecchin, E. et al. Carboxylesterase isoform 2 mRNA expression in peripheral blood mononuclear cells is a predictive marker of the irinotecan to SN38 activation step in colorectal cancer patients. Clin Cancer Res 11, 6901-6907 (2005). 54 Lokiec, F., du Sorbier, B. M. & Sanderink, G. J. Irinotecan (CPT-11) metabolites in human bile and urine. Clin Cancer Res 2, 1943-1949 (1996).

135 55 Santos, A. et al. Metabolism of irinotecan (CPT-11) by CYP3A4 and CYP3A5 in humans. Clin Cancer Res 6, 2012-2020 (2000). 56 Atsumi, R., Suzuki, W. & Hakusui, H. Identification of the metabolites of irinotecan, a new derivative of camptothecin, in rat bile and its biliary excretion. Xenobiotica 21, 1159-1169 (1991). 57 van der Bol, J. M. et al. Effect of omeprazole on the pharmacokinetics and toxicities of irinotecan in cancer patients: a prospective cross-over drug-drug interaction study. Eur J Cancer 47, 831-838 (2011). 58 Kehrer, D. F., Mathijssen, R. H., Verweij, J., de Bruijn, P. & Sparreboom, A. Modulation of irinotecan metabolism by ketoconazole. J Clin Oncol 20, 3122-3129 (2002). 59 Sai, K. et al. Impact of CYP3A4 haplotypes on irinotecan pharmacokinetics in Japanese cancer patients. Cancer Chemother Pharmacol 62, 529-537 (2008). 60 Scripture, C. D. & Figg, W. D. Drug interactions in cancer therapy. Nat Rev Cancer 6, 546-558 (2006). 61 Sasaki, T. et al. Concomitant polypharmacy is associated with irinotecan-related adverse drug reactions in patients with cancer. Int J Clin Oncol 18, 735-742 (2013). 62 Jehn, C. F., Boning, L., Kroning, H., Pezzutto, A. & Luftner, D. Influence of comorbidity, age and performance status on treatment efficacy and safety of cetuximab plus irinotecan in irinotecan-refractory elderly patients with metastatic colorectal cancer. Eur J Cancer 50, 1269-1275 (2014). 63 Cancer Care Ontario. Irinotecan Drug Monograph, (2015) Last accessed. July 3, 2015 64 Sanghani, S. P. et al. Hydrolysis of irinotecan and its oxidative metabolites, 7-ethyl- 10-[4-N-(5-aminopentanoic acid)-1-piperidino] carbonyloxycamptothecin and 7- ethyl-10-[4-(1-piperidino)-1-amino]-carbonyloxycamptothecin, by human carboxylesterases CES1A1, CES2, and a newly expressed carboxylesterase isoenzyme, CES3. Drug Metab Dispos 32, 505-511 (2004). 65 Mathijssen, R. H. et al. Irinotecan pharmacokinetics-pharmacodynamics: the clinical relevance of prolonged exposure to SN-38. Br J Cancer 87, 144-150 (2002). 66 Canal, P. et al. Pharmacokinetics and pharmacodynamics of irinotecan during a phase II clinical trial in colorectal cancer. Pharmacology and Molecular Mechanisms Group of the European Organization for Research and Treatment of Cancer. J Clin Oncol 14, 2688-2695 (1996). 67 Araki, E. et al. Relationship between development of diarrhea and the concentration of SN-38, an active metabolite of CPT-11, in the intestine and the blood plasma of athymic mice following intraperitoneal administration of CPT-11. Jpn J Clin Oncol 84, 697-702 (1993). 68 Chen, S. et al. Intestinal glucuronidation protects against chemotherapy-induced toxicity by irinotecan (CPT-11). Proc Natl Acad Sci U S A 110, 19143-19148 (2013). 69 Gupta, E. et al. Metabolic fate of irinotecan in humans: correlation of glucuronidation with diarrhea. Cancer Res 54, 3723-3725 (1994). 70 Gupta, E. et al. Pharmacokinetic and pharmacodynamic evaluation of the topoisomerase inhibitor irinotecan in cancer patients. J Clin Oncol 15, 1502-1510 (1997). 71 Tukey, R. H. & Strassburg, C. P. Human UDP-glucuronosyltransferases: metabolism, expression, and disease. Annu Rev Pharmacol Toxicol 40, 581-616 (2000).

136

72 Guillemette, C., Lévesque, E., Harvey, M., Bellemare, J. & Menard, V. UGT genomic diversity: beyond gene duplication. Drug Metab Rev 42, 24-44 (2010). 73 Mackenzie, P. I. et al. Nomenclature update for the mammalian UDP glycosyltransferase (UGT) gene superfamily. Pharmacogenet Genomics 15, 677- 685 (2005). 74 Guillemette, C., Levesque, E. & Rouleau, M. Pharmacogenomics of Human Uridine Diphospho-Glucuronosyltransferases and Clinical Implications. Clin Pharmacol Ther 96, 324-339 (2014). 75 Hanioka, N. et al. Human liver UDP-glucuronosyltransferase isoforms involved in the glucuronidation of 7-ethyl-10-hydroxycamptothecin. Xenobiotica 31, 687-699 (2001). 76 Ciotti, M., Basu, N., Brangi, M. & Owens, I. S. Glucuronidation of 7-ethyl-10- hydroxycamptothecin (SN-38) by the human UDP-glucuronosyltransferases encoded at the UGT1 locus. Biochem Biophys Res Commun 260, 199-202 (1999). 77 Iyer, L. et al. Genetic predisposition to the metabolism of irinotecan (CPT-11). Role of uridine diphosphate glucuronosyltransferase isoform 1A1 in the glucuronidation of its active metabolite (SN-38) in human liver microsomes. J Clin Invest 101, 847- 854 (1998). 78 Gagné, J.-F. et al. Common human UGT1A polymorphisms and the altered metabolism of irinotecan active metabolite 7-ethyl-10-hydroxycamptothecin (SN- 38). Mol Pharmacol 62, 608-617 (2002). 79 Ahowesso, C. et al. Relations between strain and gender dependencies of irinotecan toxicity and UGT1A1, CES2 and TOP1 expressions in mice. Toxicol Lett 192, 395-401 (2010). 80 Lévesque, E. et al. Refining the UGT1A haplotype associated with irinotecan- induced hematological toxicity in metastatic colorectal cancer patients treated with 5-fluorouracil/irinotecan (FOLFIRI)-based regimens. J Pharmacol Exp Ther 345, 95-101 (2013). 81 Rowland, A., Miners, J. O. & Mackenzie, P. I. The UDP-glucuronosyltransferases: their role in drug metabolism and detoxification. Int J Biochem Cell Biol 45, 1121- 1132 (2013). 82 Bostrom, M., Bjork, K., Nelson, B. D. & DePierre, J. W. Detailed analytical subcellular fractionation of non-pregnant porcine corpus luteum reveals peroxisomes of normal size and significant UDP-glucuronosyltransferase activity in the high-speed supernatant. Chem Biol Interact 147, 23-34 (2004). 83 Radominska-Pandya, A., Ouzzine, M., Fournel-Gigleux, S. & Magdalou, J. Structure of UDP-glucuronosyltransferases in membranes. Methods Enzymol 400, 116-147 (2005). 84 Girard, H. 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 34, 1220-1228 (2006). 85 Ando, Y. et al. UGT1A1 genotypes and glucuronidation of SN-38, the active metabolite of irinotecan. Ann Oncol 9, 845-847 (1998). 86 Iyer, L. et al. Phenotype-genotype correlation of in vitro SN-38 (active metabolite of irinotecan) and bilirubin glucuronidation in human liver tissue with UGT1A1 promoter polymorphism. Clin Pharmacol Ther 65, 576-582 (1999). 87 Jinno, H. 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 31, 108-113 (2003). 88 Villeneuve, L., Girard, H., Fortier, L. C., Gagné, J.-F. & Guillemette, C. Novel Functional Polymorphisms in the UGT1A7 and UGT1A9 Glucuronidating Enzymes

137 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 307, 117-128 (2003). 89 Nakatomi, K. et al. Transport of 7-ethyl-10-hydroxycamptothecin (SN-38) by breast cancer resistance protein ABCG2 in human lung cancer cells. Biochem Biophys Res Commun 288, 827-832 (2001). 90 Chu, X. Y. et al. Multispecific organic anion transporter is responsible for the biliary excretion of the camptothecin derivative irinotecan and its metabolites in rats. J Pharmacol Exp Ther 281, 304-314 (1997). 91 Chu, X. Y. et al. Biliary excretion mechanism of CPT-11 and its metabolites in humans: involvement of primary active transporters. Cancer Res 58, 5137-5143 (1998). 92 Iyer, L. et al. Biliary transport of irinotecan and metabolites in normal and P- glycoprotein-deficient mice. Cancer Chemother Pharmacol 49, 336-341 (2002). 93 Sugiyama, Y., Kato, Y. & Chu, X. Multiplicity of biliary excretion mechanisms for the camptothecin derivative irinotecan (CPT-11), its metabolite SN-38, and its glucuronide: role of canalicular multispecific organic anion transporter and P- glycoprotein. Cancer Chemother Pharmacol 42 Suppl, S44-49 (1998). 94 Chen, Z. S. et al. ATP-Dependent efflux of CPT-11 and SN-38 by the multidrug resistance protein (MRP) and its inhibition by PAK-104P. Mol Pharmacol 55, 921- 928 (1999). 95 Nozawa, T., Minami, H., Sugiura, S., Tsuji, A. & Tamai, I. 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 33, 434-439 (2004). 96 Zair, Z. M., Eloranta, J. J., Stieger, B. & Kullak-Ublick, G. A. Pharmacogenetics of OATP (SLC21/SLCO), OAT and OCT (SLC22) and PEPT (SLC15) transporters in the intestine, liver and kidney. Pharmacogenomics 9, 597-624 (2008). 97 Dean, M., Rzhetsky, A. & Allikmets, R. The human ATP-binding cassette (ABC) transporter superfamily. Genome Res 11, 1156-1166 (2001). 98 Chu, X. Y., Kato, Y. & Sugiyama, Y. Multiplicity of biliary excretion mechanisms for irinotecan, CPT-11, and its metabolites in rats. Cancer Res 57, 1934-1938 (1997). 99 Reese, M. J. et al. In vitro investigations into the roles of drug transporters and metabolizing enzymes in the disposition and drug interactions of dolutegravir, a HIV integrase inhibitor. Drug Metab Dispos 41, 353-361 (2013). 100 Kim, R. B. et al. Interrelationship between substrates and inhibitors of human CYP3A and P-glycoprotein. Pharm Res 16, 408-414 (1999). 101 Jiang, W. & Hu, M. Mutual interactions between flavonoids and enzymatic and transporter elements responsible for flavonoid disposition via phase II metabolic pathways. RSC Adv 2, 7948-7963 (2012). 102 Nies, A. T., Schwab, M. & Keppler, D. Interplay of conjugating enzymes with OATP uptake transporters and ABCC/MRP efflux pumps in the elimination of drugs. Expert Opin Drug Metab Toxicol 4, 545-568 (2008). 103 Panczyk, M. Pharmacogenetics research on chemotherapy resistance in colorectal cancer over the last 20 years. World J Gastroenterol 20, 9775 (2014). 104 Ho, R. H. & Kim, R. B. Transporters and drug therapy: implications for drug disposition and disease. Clin Pharmacol Ther 78, 260-277 (2005). 105 Gong, I. Y. & Kim, R. B. Impact of genetic variation in OATP transporters to drug disposition and response. Drug Metab Pharmacokinet 28, 4-18 (2013).

138

106 Lee, N. H. Pharmacogenetics of drug metabolizing enzymes and transporters: effects on pharmacokinetics and pharmacodynamics of anticancer agents. Anticancer Agents Med Chem 10, 583-592 (2010). 107 Prasad, B. et al. Interindividual variability in hepatic organic anion-transporting polypeptides and P-glycoprotein (ABCB1) protein expression: quantification by liquid chromatography tandem mass spectroscopy and influence of genotype, age, and sex. Drug Metab Dispos 42, 78-88 (2014). 108 Bruhn, O. & Cascorbi, I. Polymorphisms of the drug transporters ABCB1, ABCG2, ABCC2 and ABCC3 and their impact on drug bioavailability and clinical relevance. Expert Opin Drug Metab Toxicol, 1-18 (2014). 109 Cascorbi, I. Role of pharmacogenetics of ATP-binding cassette transporters in the pharmacokinetics of drugs. Pharmacol Ther 112, 457-473 (2006). 110 Yiannakopoulou, E. C. Pharmacogenomics of phase II metabolizing enzymes and drug transporters: clinical implications. Pharmacogenomics J, 1-5 (2012). 111 Thiebaut, F. et al. Cellular localization of the multidrug-resistance gene product P- glycoprotein in normal human tissues. Proc Natl Acad Sci U S A 84, 7735-7738 (1987). 112 Smith, N. F., Figg, W. D. & Sparreboom, A. Pharmacogenetics of irinotecan metabolism and transport: An update. Toxicol In Vitro 20, 163-175 (2006). 113 Mathijssen, R. H. et al. Irinotecan pathway genotype analysis to predict pharmacokinetics. Clin Cancer Res 9, 3246-3253 (2003). 114 De Mattia, E. et al. Pharmacogenetics of ABC and SLC transporters in metastatic colorectal cancer patients receiving first-line FOLFIRI treatment. Pharmacogenet Genomics 23, 549-557 (2013). 115 Glimelius, B. et al. Prediction of irinotecan and 5-fluorouracil toxicity and response in patients with advanced colorectal cancer. Pharmacogenomics J 11, 61-71 (2010). 116 Han, J.-Y. et al. Associations ofABCB1, ABCC2, andABCG2 polymorphisms with irinotecan-pharmacokinetics and clinical outcome in patients with advanced non- small cell lung cancer. Cancer 110, 138-147 (2007). 117 Cortejoso, L. et al. Differential toxicity biomarkers for irinotecan- and oxaliplatin- containing chemotherapy in colorectal cancer. Cancer Chemother Pharmacol 71, 1463-1472 (2013). 118 Lara, P. N. et al. Phase III Trial of Irinotecan/Cisplatin Compared With Etoposide/Cisplatin in Extensive-Stage Small-Cell Lung Cancer: Clinical and Pharmacogenomic Results From SWOG S0124. J Clin Oncol 27, 2530-2535 (2009). 119 Hoffmeyer, S. et al. Functional polymorphisms of the human multidrug-resistance gene: multiple sequence variations and correlation of one allele with P-glycoprotein expression and activity in vivo. Proc Natl Acad Sci U S A 97, 3473-3478 (2000). 120 Nishioka, C. et al. MDR1, MRP1 and MRP2 genotypes and in vitro chemosensitivity in Japanese patients with colorectal adenocarcinomas. Kobe J Med Sci 50, 181-188 (2004). 121 Luo, F. R., Paranjpe, P. V., Guo, A., Rubin, E. & Sinko, P. Intestinal transport of irinotecan in Caco-2 cells and MDCK II cells overexpressing efflux transporters Pgp, cMOAT, and MRP1. Drug Metab Dispos 30, 763-770 (2002). 122 Takara, K. et al. Molecular changes to HeLa cells on continuous exposure to SN- 38, an active metabolite of irinotecan hydrochloride. Cancer Lett 278, 88-96 (2009). 123 Innocenti, F. et al. Comprehensive Pharmacogenetic Analysis of Irinotecan Neutropenia and Pharmacokinetics. J Clin Oncol 27, 2604-2614 (2009).

139 124 Conseil, G. & Cole, S. P. Two polymorphic variants of ABCC1 selectively alter drug resistance and inhibitor sensitivity of the multidrug and organic anion transporter multidrug resistance protein 1. Drug Metab Dispos 41, 2187-2196 (2013). 125 Akiyama, Y. et al. Association of ABCC2 genotype with efficacy of first-line FOLFIRI in Japanese patients with advanced colorectal cancer. Drug Metab Pharmacokinet 27, 325-335 (2011). 126 Han, J.-Y., Lim, H.-S., Park, Y. H., Lee, S. Y. & Lee, J. S. Integrated pharmacogenetic prediction of irinotecan pharmacokinetics and toxicity in patients with advanced non-small cell lung cancer. Lung Cancer 63, 115-120 (2009). 127 Estudante, M., Morais, J. G., Soveral, G. & Benet, L. Z. Intestinal drug transporters: an overview. Adv Drug Deliv Rev 65, 1340-1356 (2013). 128 Köck, K. & Brouwer, K. L. R. A Perspective on Efflux Transport Proteins in the Liver. Clinical Pharmacology & Therapeutics 92, 599-612 (2012). 129 Schellens, J. H. et al. Transport of topoisomerase I inhibitors by the breast cancer resistance protein. Potential clinical implications. Ann N Y Acad Sci 922, 188-194 (2000). 130 Kawabata, S. et al. Breast cancer resistance protein directly confers SN-38 resistance of lung cancer cells. Biochem Biophys Res Commun 280, 1216-1223 (2001). 131 Candeil, L. et al. ABCG2 overexpression in colon cancer cells resistant to SN38 and in irinotecan-treated metastases. Int J Cancer 109, 848-854 (2004). 132 Bates, S. E. ABCG2 Mediates Differential Resistance to SN-38 (7-Ethyl-10- hydroxycamptothecin) and Homocamptothecins. J Pharmacol Exp Ther 310, 836- 842 (2004). 133 Jada, S. R. et al. Role of UGT1A1*6, UGT1A1*28and ABCG2c.421C>A polymorphisms in irinotecan-induced neutropenia in Asian cancer patients. Cancer Sci 98, 1461-1467 (2007). 134 Zhou, Q. et al. Pharmacogenetic profiling across the irinotecan pathway in Asian patients with cancer. Br J Clin Pharmacol 59, 415-424 (2005). 135 Imai, Y. et al. C421A polymorphism in the human breast cancer resistance protein gene is associated with low expression of Q141K protein and low-level drug resistance. Mol Cancer Ther 1, 611-616 (2002). 136 Han, J.-Y. et al. Influence of the organic anion-transporting polypeptide 1B1 (OATP1B1) polymorphisms on irinotecan-pharmacokinetics and clinical outcome of patients with advanced non-small cell lung cancer. Lung Cancer 59, 69-75 (2008). 137 Di Martino, M. T. et al. Single nucleotide polymorphisms of ABCC5 and ABCG1 transporter genes correlate to irinotecan-associated gastrointestinal toxicity in colorectal cancer patients: A DMET microarray profiling study. Cancer Biol Ther 12, 780-787 (2011). 138 Sai, K. et al. Additive effects of drug transporter genetic polymorphisms on irinotecan pharmacokinetics/pharmacodynamics in Japanese cancer patients. Cancer Chemother Pharmacol 66, 95-105 (2009). 139 Teft, W. A. et al. OATP1B1 and tumour OATP1B3 modulate exposure, toxicity, and survival after irinotecan-based chemotherapy. Br J Cancer 112, 857-865 (2015). 140 Rhodes, K. E. et al. ABCB1, SLCO1B1 and UGT1A1 gene polymorphisms are associated with toxicity in metastatic colorectal cancer patients treated with first- line irinotecan. Drug Metab Lett 1, 23-30 (2007). 141 Huang, L. et al. SLCO1B1 and SLC19A1 gene variants and irinotecan-induced rapid response and survival: a prospective multicenter pharmacogenetics study of metastatic colorectal cancer. PLoS One 8, e77223 (2013).

140

142 Kaneda, N. & Yokokura, T. Nonlinear pharmacokinetics of CPT-11 in rats. Cancer Res 50, 1721-1725 (1990). 143 Itoh, T. et al. Biliary excretion of irinotecan and its metabolites. J Pharm Pharm Sci 7, 13-18 (2004). 144 Slatter, J. G. et al. Pharmacokinetics, metabolism, and excretion of irinotecan (CPT-11) following I.V. infusion of [(14)C]CPT-11 in cancer patients. Drug Metab Dispos 28, 423-433 (2000). 145 de Forni, M. et al. Phase I and pharmacokinetic study of the camptothecin derivative irinotecan, administered on a weekly schedule in cancer patients. Cancer Res 54, 4347-4354 (1994). 146 Lokiec, F. et al. Pharmacokinetics of irinotecan and its metabolites in human blood, bile, and urine. Cancer Chemother Pharmacol 36, 79-82 (1995). 147 Rothenberg, M. L. et al. Phase I and pharmacokinetic trial of weekly CPT-11. J Clin Oncol 11, 2194-2204 (1993). 148 Sparreboom, A. et al. Irinotecan (CPT-11) metabolism and disposition in cancer patients. Clin Cancer Res 4, 2747-2754 (1998). 149 Rivory, L. P. et al. Pharmacokinetic interrelationships of irinotecan (CPT-11) and its three major plasma metabolites in patients enrolled in phase I/II trials. Clin Cancer Res 3, 1261-1266 (1997). 150 de Jong, F. A. et al. Renal function as a predictor of irinotecan-induced neutropenia. Clin Pharmacol Ther 84, 254-262 (2008). 151 Venook, A. P. et al. A phase I and pharmacokinetic study of irinotecan in patients with hepatic or renal dysfunction or with prior pelvic radiation: CALGB 9863. Ann Oncol 14, 1783-1790 (2003). 152 Chabot, G. G. et al. Population pharmacokinetics and pharmacodynamics of irinotecan (CPT-11) and active metabolite SN-38 during phase I trials. Ann Oncol 6, 141-151 (1995). 153 Catimel, G. et al. Phase I and pharmacokinetic study of irinotecan (CPT-11) administered daily for three consecutive days every three weeks in patients with advanced solid tumors. Ann Oncol 6, 133-140 (1995). 154 Abigerges, D. et al. Phase I and pharmacologic studies of the camptothecin analog irinotecan administered every 3 weeks in cancer patients. J Clin Oncol 13, 210-221 (1995). 155 Roberts, M. S., Magnusson, B. M., Burczynski, F. J. & Weiss, M. Enterohepatic circulation: physiological, pharmacokinetic and clinical implications. Clin Pharmacokinet 41, 751-790 (2002). 156 Hurley, L. H. DNA and its associated processes as targets for cancer therapy. Nat Rev Cancer 2, 188-200 (2002). 157 Sugimoto, Y., Tsukahara, S., Oh-hara, T., Isoe, T. & Tsuruo, T. Decreased expression of DNA topoisomerase I in camptothecin-resistant tumor cell lines as determined by a monoclonal antibody. Cancer Res 50, 6925-6930 (1990). 158 Kanzawa, F. et al. Establishment of a camptothecin analogue (CPT-11)-resistant cell line of human non-small cell lung cancer: characterization and mechanism of resistance. Cancer Res 50, 5919-5924 (1990). 159 Andoh, T. et al. Characterization of a mammalian mutant with a camptothecin- resistant DNA topoisomerase I. Proc Natl Acad Sci U S A 84, 5565-5569 (1987). 160 Takano, H., Kohno, K., Matsuo, K., Matsuda, T. & Kuwano, M. DNA topoisomerase-targeting antitumor agents and drug resistance. Anticancer Drugs 3, 323-330 (1992). 161 Burgess, D. J. et al. Topoisomerase levels determine chemotherapy response in vitro and in vivo. Proc Natl Acad Sci U S A 105, 9053-9058 (2008).

141 162 McLeod, H. L. & Keith, W. N. Variation in topoisomerase I gene copy number as a mechanism for intrinsic drug sensitivity. Br J Cancer 74, 508-512 (1996). 163 Tsurutani, J. et al. Point mutations in the topoisomerase I gene in patients with non-small cell lung cancer treated with irinotecan. Lung Cancer 35, 299-304 (2002). 164 Kubota, N. et al. Detection of topoisomerase I gene point mutation in CPT-11 resistant lung cancer cell line. Biochem Biophys Res Commun 188, 571-577 (1992). 165 Gongora, C. et al. New Topoisomerase I mutations are associated with resistance to camptothecin. Mol Cancer 10, 64 (2011). 166 Takatani, H. et al. Gene mutation analysis and quantitation of DNA topoisomerase I in previously untreated non-small cell lung carcinomas. Jpn J Cancer Res 88, 160- 165 (1997). 167 McLeod, H. L. et al. Topoisomerase I and II activity in human breast, cervix, lung and colon cancer. Int J Cancer 59, 607-611 (1994). 168 Husain, I., Mohler, J. L., Seigler, H. F. & Besterman, J. M. Elevation of topoisomerase I messenger RNA, protein, and catalytic activity in human tumors: demonstration of tumor-type specificity and implications for cancer chemotherapy. Cancer Res 54, 539-546 (1994). 169 Alagoz, M., Gilbert, D. C., El-Khamisy, S. & Chalmers, A. J. DNA repair and resistance to topoisomerase I inhibitors: mechanisms, biomarkers and therapeutic targets. Curr Med Chem 19, 3874-3885 (2012). 170 Pommier, Y. DNA topoisomerase I inhibitors: chemistry, biology, and interfacial inhibition. Chem Rev 109, 2894-2902 (2009). 171 Pommier, Y. Topoisomerase I inhibitors: camptothecins and beyond. Nat Rev Cancer 6, 789-802 (2006). 172 Park, S. Y. & Cheng, Y. C. Poly(ADP-ribose) polymerase-1 could facilitate the religation of topoisomerase I-linked DNA inhibited by camptothecin. Cancer Res 65, 3894-3902 (2005). 173 Perego, P. et al. Role of tyrosyl-DNA phosphodiesterase 1 and inter-players in regulation of tumor cell sensitivity to topoisomerase I inhibition. Biochem Pharmacol 83, 27-36 (2012). 174 Das, B. B. et al. PARP1-TDP1 coupling for the repair of topoisomerase I-induced DNA damage. Nucleic Acids Res 42, 4435-4449 (2014). 175 Zhang, Y. W. et al. Poly(ADP-ribose) polymerase and XPF-ERCC1 participate in distinct pathways for the repair of topoisomerase I-induced DNA damage in mammalian cells. Nucleic Acids Res 39, 3607-3620 (2011). 176 Plo, I. et al. Association of XRCC1 and tyrosyl DNA phosphodiesterase (Tdp1) for the repair of topoisomerase I-mediated DNA lesions. DNA Repair (Amst) 2, 1087- 1100 (2003). 177 Braun, M. S. et al. Predictive biomarkers of chemotherapy efficacy in colorectal cancer: results from the UK MRC FOCUS trial. J Clin Oncol 26, 2690-2698 (2008). 178 Yu, J., Shannon, W. D., Watson, M. A. & McLeod, H. L. Gene expression profiling of the irinotecan pathway in colorectal cancer. Clin Cancer Res 11, 2053-2062 (2005). 179 Vallböhmer, D. et al. Molecular determinants of irinotecan efficacy. Int J Cancer 119, 2435-2442 (2006). 180 Allen, W. L. et al. A Systems Biology Approach Identifies SART1 as a Novel Determinant of Both 5-Fluorouracil and SN38 Drug Resistance in Colorectal Cancer. Mol Cancer Ther 11, 119-131 (2012).

142

181 Peng, X.-C. et al. Proteomic analysis of cell lines to identify the irinotecan resistance proteins. J Biosci 35, 557-564 (2010). 182 Gong, F.-M. et al. Comparative proteomic analysis of irinotecan-sensitive colorectal carcinoma cell line and its chemoresistant counterpart. Anticancer Drugs 22, 500- 506 (2011). 183 Braun, M. S. et al. Association of Molecular Markers With Toxicity Outcomes in a Randomized Trial of Chemotherapy for Advanced Colorectal Cancer: The FOCUS Trial. J Clin Oncol 27, 5519-5528 (2009). 184 McLeod, H. L. et al. Pharmacogenetic Predictors of Adverse Events and Response to Chemotherapy in Metastatic Colorectal Cancer: Results From North American Gastrointestinal Intergroup Trial N9741. J Clin Oncol 28, 3227-3233 (2010). 185 Hoskins, J. M. et al. Irinotecan Pharmacogenetics: Influence of Pharmacodynamic Genes. Clin Cancer Res 14, 1788-1796 (2008). 186 Hoskins, J. M., Rosner, G. L., Ratain, M. J., McLeod, H. L. & Innocenti, F. Pharmacodynamic genes do not influence risk of neutropenia in cancer patients treated with moderately high-dose irinotecan. Pharmacogenomics 10, 1139-1146 (2009). 187 Artac, M. et al. The value of XPD and XRCC1 genotype polymorphisms to predict clinical outcome in metastatic colorectal carcinoma patients with irinotecan-based regimens. J Cancer Res Clin Oncol 136, 803-809 (2009). 188 Ruzzo, A. et al. Pharmacogenetic profiling in patients with advanced colorectal cancer treated with first-line FOLFIRI chemotherapy. Pharmacogenomics J 8, 278- 288 (2007). 189 Kweekel, D. M. et al. GSTP1 Ile105Val polymorphism correlates with progression- free survival in MCRC patients treated with or without irinotecan: a study of the Dutch Colorectal Cancer Group. Br J Cancer 99, 1316-1321 (2008). 190 Le Morvan, V. et al. Determination of ERCC2Lys751Gln and GSTP1Ile105Val gene polymorphisms in colorectal cancer patients: relationships with treatment outcome. Pharmacogenomics 8, 1693-1703 (2007). 191 Rowinsky, E. K. et al. Phase I and pharmacological study of the novel topoisomerase I inhibitor 7-ethyl-10-[4-(1-piperidino)-1-piperidino] carbonyloxycamptothecin (CPT-11) administered as a ninety-minute infusion every 3 weeks. Cancer Res 54, 427-436 (1994). 192 Ohe, Y. et al. Phase I study and pharmacokinetics of CPT-11 with 5-day continuous infusion. J Natl Cancer Inst 84, 972-974 (1992). 193 Sasaki, Y. et al. A pharmacokinetic and pharmacodynamic analysis of CPT-11 and its active metabolite SN-38. Jpn J Cancer Res 86, 101-110 (1995). 194 Stein, A., Voigt, W. & Jordan, K. Chemotherapy-induced diarrhea: pathophysiology, frequency and guideline-based management. Ther Adv Med Oncol 2, 51-63 (2010). 195 Gandia, D. et al. CPT-11-induced cholinergic effects in cancer patients. J Clin Oncol 11, 196-197 (1993). 196 Saliba, F. et al. Pathophysiology and therapy of irinotecan-induced delayed-onset diarrhea in patients with advanced colorectal cancer: a prospective assessment. J Clin Oncol 16, 2745-2751 (1998). 197 Gibson, R. J., Bowen, J. M., Inglis, M. R., Cummins, A. G. & Keefe, D. M. Irinotecan causes severe small intestinal damage, as well as colonic damage, in the rat with implanted breast cancer. J Gastroenterol Hepatol 18, 1095-1100 (2003).

143 198 Stringer, A. M. et al. Irinotecan-induced mucositis manifesting as diarrhoea corresponds with an amended intestinal flora and mucin profile. Int J Exp Pathol 90, 489-499 (2009). 199 Takasuna, K. et al. Involvement of beta-glucuronidase in intestinal microflora in the intestinal toxicity of the antitumor camptothecin derivative irinotecan hydrochloride (CPT-11) in rats. Cancer Res 56, 3752-3757 (1996). 200 Kehrer, D. F. et al. Modulation of irinotecan-induced diarrhea by cotreatment with neomycin in cancer patients. Clin Cancer Res 7, 1136-1141 (2001). 201 Takakura, A. et al. Rapid deconjugation of SN-38 glucuronide and adsorption of released free SN-38 by intestinal microorganisms in rat. Oncol Lett 3, 520-524 (2012). 202 Rothenberg, M. L. et al. Phase I dose-finding and pharmacokinetic trial of irinotecan (CPT-11) administered every two weeks. Ann Oncol 12, 1631-1641 (2001). 203 Di Paolo, A. B., Guido, Polillo, M., Del Re, M., Di Desidero, T. L., Marianna & Danesi, R. Pharmacokinetic and pharmacogenetic predictive markers of irinotecan activity and toxicity. Curr Drug Metab 12, 932 (2011). 204 di Iulio, J. & Rotger, M. Pharmacogenomics: what is next? Front Pharmacol 2, 86 (2011). 205 Pirmohamed, M. Acceptance of biomarker-based tests for application in clinical practice: criteria and obstacles. Clinical Pharmacology & Therapeutics 88, 862-866 (2010). 206 Pirmohamed, M. Personalized pharmacogenomics: predicting efficacy and adverse drug reactions. Annu Rev Genomics Hum Genet 15, 349-370 (2014). 207 Ingelman-Sundberg, M., Sim, S. C., Gomez, A. & Rodriguez-Antona, C. Influence of cytochrome P450 polymorphisms on drug therapies: Pharmacogenetic, pharmacoepigenetic and clinical aspects. Pharmacol Ther 116, 496-526 (2007). 208 US Food and Drug Administration. Table of Pharmacogenomic Biomarkers in Drug Labeling, (2015) Last accessed. April 23, 2015 209 Holmes, M. V. et al. Fulfilling the promise of personalized medicine? Systematic review and field synopsis of pharmacogenetic studies. PLoS One 4, e7960 (2009). 210 Fung, E. et al. Effect of genetic variants, especially CYP2C9 and VKORC1, on the pharmacology of warfarin. Semin Thromb Hemost 38, 893-904 (2012). 211 Yang, L., Ge, W., Yu, F. & Zhu, H. Impact of VKORC1 gene polymorphism on interindividual and interethnic warfarin dosage requirement--a systematic review and meta analysis. Thromb Res 125, e159-166 (2010). 212 Wadelius, M. et al. Common VKORC1 and GGCX polymorphisms associated with warfarin dose. Pharmacogenomics J 5, 262-270 (2005). 213 Jorgensen, A. L., FitzGerald, R. J., Oyee, J., Pirmohamed, M. & Williamson, P. R. Influence of CYP2C9 and VKORC1 on patient response to warfarin: a systematic review and meta-analysis. PLoS One 7, e44064 (2012). 214 US Food and Drug Administration. Warfarin Prescribing Information and Label (2011) Last accessed. Dec 01, 2014 215 Finkelman, B. S., Gage, B. F., Johnson, J. A., Brensinger, C. M. & Kimmel, S. E. Genetic warfarin dosing: tables versus algorithms. J Am Coll Cardiol 57, 612-618 (2011). 216 Mega, J. L. & Giugliano, R. P. Genotype-guided dosing of warfarin. Clin Chem 60, 920-922 (2014).

144

217 Pirmohamed, M. et al. A randomized trial of genotype-guided dosing of warfarin. N Engl J Med 369, 2294-2303 (2013). 218 Liao, Z., Feng, S., Ling, P. & Zhang, G. Meta-analysis of randomized controlled trials reveals an improved clinical outcome of using genotype plus clinical algorithm for warfarin dosing. J Thromb Thrombolysis (2014). 219 Xu, H. et al. Meta-analysis of efficacy and safety of genotype-guided pharmacogenetic dosing of warfarin. Int J Cardiol 177, 654-657 (2014). 220 Dean, L. in Warfarin therapy and the genotypes CYP2C9 and VKORC1. (2013) Medical Genetics Summaries. Last accessed. April 20, 2015 221 Owen, R. P., Sangkuhl, K., Klein, T. E. & Altman, R. B. Cytochrome P450 2D6. Pharmacogenet Genomics 19, 559-562 (2009). 222 Mega, J. L. et al. Reduced-function CYP2C19 genotype and risk of adverse clinical outcomes among patients treated with clopidogrel predominantly for PCI: a meta- analysis. JAMA 304, 1821-1830 (2010). 223 Shuldiner, A. R. et al. Association of cytochrome P450 2C19 genotype with the antiplatelet effect and clinical efficacy of clopidogrel therapy. JAMA 302, 849-857 (2009). 224 Pare, G. et al. Effects of CYP2C19 genotype on outcomes of clopidogrel treatment. N Engl J Med 363, 1704-1714 (2010). 225 Horenstein, R. B. et al. Effectiveness of clopidogrel dose escalation to normalize active metabolite exposure and antiplatelet effects in CYP2C19 poor metabolizers. J Clin Pharmacol 54, 865-873 (2014). 226 Mao, L. et al. Cytochrome CYP2C19 polymorphism and risk of adverse clinical events in clopidogrel-treated patients: a meta-analysis based on 23,035 subjects. Arch Cardiovasc Dis 106, 517-527 (2013). 227 Jang, J. S. et al. Meta-analysis of cytochrome P450 2C19 polymorphism and risk of adverse clinical outcomes among coronary artery disease patients of different ethnic groups treated with clopidogrel. Am J Cardiol 110, 502-508 (2012). 228 Society for Cardiovascular, A. et al. ACCF/AHA Clopidogrel clinical alert: approaches to the FDA "boxed warning": a report of the American College of Cardiology Foundation Task Force on Clinical Expert Consensus Documents and the American Heart Association. Circulation 122, 537-557 (2010). 229 Scott, S. A. et al. Clinical Pharmacogenetics Implementation Consortium guidelines for CYP2C19 genotype and clopidogrel therapy: 2013 update. Clin Pharmacol Ther 94, 317-323 (2013). 230 Notarangelo, M. F., Bontardelli, F. & Merlini, P. A. Genetic and nongenetic factors influencing the response to clopidogrel. J Cardiovasc Med (Hagerstown) 14 Suppl 1, S1-7 (2013). 231 Momary, K. M., Dorsch, M. P. & Bates, E. R. Genetic causes of clopidogrel nonresponsiveness: which ones really count? Pharmacotherapy 30, 265-274 (2010). 232 Yin, T. & Miyata, T. Pharmacogenomics of clopidogrel: evidence and perspectives. Thromb Res 128, 307-316 (2011). 233 Lewis, J. P. et al. Paraoxonase 1 (PON1) gene variants are not associated with clopidogrel response. Clin Pharmacol Ther 90, 568-574 (2011). 234 Tang, X. F. et al. Effect of the CYP2C19 2 and 3 genotypes, ABCB1 C3435T and PON1 Q192R alleles on the pharmacodynamics and adverse clinical events of clopidogrel in Chinese people after percutaneous coronary intervention. Eur J Clin Pharmacol 69, 1103-1112 (2013).

145 235 Kassimis, G. et al. CYP2C19*2 and other genetic variants affecting platelet response to clopidogrel in patients undergoing percutaneous coronary intervention. Thromb Res 129, 441-446 (2012). 236 Kang, Y. H. et al. Association of PON1 genotype and haplotype with susceptibility to coronary artery disease and clinical outcomes in dual antiplatelet-treated Han Chinese patients. Eur J Clin Pharmacol 69, 1511-1519 (2013). 237 Ned, R. M. Genetic testing for CYP450 polymorphisms to predict response to clopidogrel: current evidence and test availability. Application: pharmacogenomics. PLoS Curr 2 (2010). 238 Bergmeijer, T. O. et al. CYP2C19 genotype-guided antiplatelet therapy in ST- segment elevation myocardial infarction patients-Rationale and design of the Patient Outcome after primary PCI (POPular) Genetics study. Am Heart J 168, 16- 22 e11 (2014). 239 Rai, M. et al. CYP2C19 genotype-guided antiplatelet therapy in a patient with clopidogrel resistance. Conn Med 76, 267-272 (2012). 240 Cai, Z., Yang, J., Shu, X. & Xiong, X. Chemotherapy-associated hepatotoxicity in colorectal cancer. J BUON 19, 350-356 (2014). 241 Boussios, S., Pentheroudakis, G., Katsanos, K. & Pavlidis, N. Systemic treatment- induced gastrointestinal toxicity: incidence, clinical presentation and management. Ann Gastroenterol 25, 106-118 (2012). 242 Hu, Z. Y., Yu, Q., Pei, Q. & Guo, C. Dose-Dependent Association between UGT1A1*28 Genotype and Irinotecan-Induced Neutropenia: Low Doses Also Increase Risk. Clin Cancer Res 16, 3832-3842 (2010). 243 US Food and Drug Administration. Camptosar (irinotecan HCL) injection, (2010) Last accessed. Dec 10, 2014 244 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 U S A 95, 8170-8174 (1998). 245 Iyer, L. et al. UGT1A1*28 polymorphism as a determinant of irinotecan disposition and toxicity. Pharmacogenomics J 2, 43-47 (2002). 246 Innocenti, F. Genetic Variants in the UDP-glucuronosyltransferase 1A1 Gene Predict the Risk of Severe Neutropenia of Irinotecan. J Clin Oncol 22, 1382-1388 (2004). 247 Ando, Y. et al. Polymorphisms of UDP-Glucuronosyltransferase Gene and Irinotecan Toxicity: A Pharmacogenetic Analysis. Cancer Res 60, 6921-6926 (2000). 248 Rouits, E. Relevance of Different UGT1A1 Polymorphisms in Irinotecan-Induced Toxicity: A Molecular and Clinical Study of 75 Patients. Clin Cancer Res 10, 5151- 5159 (2004). 249 Hoskins, J. M., Goldberg, R. M., Qu, P., Ibrahim, J. G. & McLeod, H. L. UGT1A1*28 Genotype and Irinotecan-Induced Neutropenia: Dose Matters. J Natl Cancer Inst 99, 1290-1295 (2007). 250 Tziotou, M. et al. Polymorphisms of uridine glucuronosyltransferase gene and irinotecan toxicity: low dose does not protect from toxicity. Ecancermedicalscience 8, 428 (2014). 251 Shulman, K. et al. Clinical implications of UGT1A1*28 genotype testing in colorectal cancer patients. Cancer 117, 3156-3162 (2011). 252 Martinez-Balibrea, E. et al. UGT1A and TYMS genetic variants predict toxicity and response of colorectal cancer patients treated with first-line irinotecan and fluorouracil combination therapy. Br J Cancer 103, 581-589 (2010).

146

253 Marcuello, E. et al. UGT1A1 gene variations and irinotecan treatment in patients with metastatic colorectal cancer. Br J Cancer 91, 678-682 (2004). 254 Ferraldeschi, R. et al. UGT1A1*28 genotype predicts gastrointestinal toxicity in patients treated with intermediate-dose irinotecan. Pharmacogenomics 10, 733-739 (2009). 255 Li, M. et al. Clinical significance of UGT1A1 gene polymorphisms on irinotecan- based regimens as the treatment in metastatic colorectal cancer. Onco Targets Ther 7, 1653-1661 (2014). 256 Hu, Z.-Y., Yu, Q. & Zhao, Y.-S. Dose-dependent association between UGT1A1*28 polymorphism and irinotecan-induced diarrhoea: A meta-analysis. Eur J Cancer 46, 1856-1865 (2010). 257 Nagar, S. & Blanchard, R. L. Pharmacogenetics of Uridine Diphosphoglucuronosyltransferase (UGT) 1A Family Members and its Role in Patient Response to Irinotecan*. Drug Metab Rev 38, 393-409 (2006). 258 Massacesi, C. et al. Uridine diphosphate glucuronosyl transferase 1A1 promoter polymorphism predicts the risk of gastrointestinal toxicity and fatigue induced by irinotecan-based chemotherapy. Cancer 106, 1007-1016 (2006). 259 Cai, X. et al. Analysis of UGT1A1*28 genotype and SN-38 pharmacokinetics for irinotecan-based chemotherapy in patients with advanced colorectal cancer: results from a multicenter, retrospective study in Shanghai. J Cancer Res Clin Oncol 139, 1579-1589 (2013). 260 Hirasawa, A. et al. Polymorphisms in the UGT1A1 gene predict adverse effects of irinotecan in the treatment of gynecologic cancer in Japanese patients. J Hum Genet, 1-5 (2013). 261 Relling, M. V. & Dervieux, T. Pharmacogenetics and cancer therapy. Nature Reviews Cancer 1, 99-108 (2001). 262 Innocenti, F. et al. Dose-Finding and Pharmacokinetic Study to Optimize the Dosing of Irinotecan According to the UGT1A1 Genotype of Patients With Cancer. J Clin Oncol 32, 2328-2334 (2014). 263 Toffoli, G. et al. Genotype-Driven Phase I Study of Irinotecan Administered in Combination With Fluorouracil/Leucovorin in Patients With Metastatic Colorectal Cancer. J Clin Oncol 28, 866-871 (2010). 264 Marcuello, E. et al. A genotype-directed phase I–IV dose-finding study of irinotecan in combination with fluorouracil/leucovorin as first-line treatment in advanced colorectal cancer. Br J Cancer 105, 53-57 (2011). 265 Minami, H. et al. Irinotecan pharmacokinetics/pharmacodynamics and UGT1A genetic polymorphisms in Japanese: roles of UGT1A1*6 and *28. Pharmacogenet Genomics 17, 497-504 (2007). 266 Hazama, S. et al. UGT1A1*6, 1A7*3, and 1A9*22genotypes predict severe neutropenia in FOLFIRI-treated metastatic colorectal cancer in two prospective studies in Japan. Cancer Sci 104, 1662-1669 (2013). 267 Gao, J. et al. UGT1A1*6/*28 polymorphisms could predict irinotecan-induced severe neutropenia not diarrhea in Chinese colorectal cancer patients. Med Oncol 30, 604 (2013). 268 Satoh, T. et al. Genotype-directed, dose-finding study of irinotecan in cancer patients with UGT1A1*28 and/or UGT1A1*6 polymorphisms. Cancer Sci 102, 1868-1873 (2011). 269 Kim, K. P. et al. A Phase I Study of UGT1A1 * 28/ * 6 Genotype-Directed Dosing of Irinotecan (CPT-11) in Korean Patients with Metastatic Colorectal Cancer Receiving FOLFIRI. Oncology 88, 164-172 (2014).

147 270 Liu, X., Cheng, D., Kuang, Q., Liu, G. & Xu, W. Association between UGT1A1*28 Polymorphisms and Clinical Outcomes of Irinotecan-Based Chemotherapies in Colorectal Cancer: A Meta-Analysis in Caucasians. PLoS One 8, e58489 (2013). 271 Dias, M. M., McKinnon, R. A. & Sorich, M. J. Impact of the UGT1A1*28allele on response to irinotecan: a systematic review and meta-analysis. Pharmacogenomics 13, 889-899 (2012). 272 Dias, M. M. et al. The effect of the UGT1A1*28 allele on survival after irinotecan- based chemotherapy: a collaborative meta-analysis. Pharmacogenomics J, 1-8 (2014). 273 Cote, J. F. et al. UGT1A1 Polymorphism Can Predict Hematologic Toxicity in Patients Treated with Irinotecan. Clin Cancer Res 13, 3269-3275 (2007). 274 Lamas, M. J. et al. The value of genetic polymorphisms to predict toxicity in metastatic colorectal patients with irinotecan-based regimens. Cancer Chemother Pharmacol 69, 1591-1599 (2012). 275 Cecchin, E. et al. Predictive Role of the UGT1A1, UGT1A7, and UGT1A9 Genetic Variants and Their Haplotypes on the Outcome of Metastatic Colorectal Cancer Patients Treated With Fluorouracil, Leucovorin, and Irinotecan. J Clin Oncol 27, 2457-2465 (2009). 276 Han, J.-Y. 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 24, 2237-2244 (2006). 277 Carlini, L. E. et al. UGT1A7 and UGT1A9 polymorphisms predict response and toxicity in colorectal cancer patients treated with capecitabine/irinotecan. Clin Cancer Res 11, 1226-1236 (2005). 278 Lankisch, T. O. et al. Gilbert's Syndrome and Irinotecan Toxicity: Combination with UDP-Glucuronosyltransferase 1A7 Variants Increases Risk. Cancer Epidemiol Biomarkers Prev 17, 695-701 (2008). 279 Hasegawa, Y., Ando, Y. & Shimokata, K. Screening for adverse reactions to irinotecan treatment using the Invader UGT1A1 Molecular Assay. Expert Rev Mol Diagn 6, 527-533 (2006). 280 Gillis, N. K., Patel, J. N. & Innocenti, F. Clinical implementation of germ line cancer pharmacogenetic variants during the next-generation sequencing era. Clin Pharmacol Ther 95, 269-280 (2014). 281 Group, E. o. G. A. i. P. P. W. Recommendations from the EGAPP Working Group: can UGT1A1 genotyping reduce morbidity and mortality in patients with metastatic colorectal cancer treated with irinotecan? Genet Med 11, 15-20 (2009). 282 van Schaik, R. H. Clinical Application of Pharmacogenetics: Where are we now? J Int Fed Clin Chem 24 (2013). 283 Crews, K. R. et al. Development and implementation of a pharmacist-managed clinical pharmacogenetics service. Am J Health Syst Pharm 68, 143-150 (2011). 284 Skierka, J. M., Kotzer, K. E., Lagerstedt, S. A., O'Kane, D. J. & Baudhuin, L. M. UGT1A1 genetic analysis as a diagnostic aid for individuals with unconjugated hyperbilirubinemia. J Pediatr 162, 1146-1152, 1152 e1141-1142 (2013). 285 Jonker, D. J. et al. Consolidating gains in colorectal cancer. Oncology Exchange 7, 18-22 (2008). 286 Harada, T. et al. Clinical usefulness of testing for UDP glucuronosyltransferaseï¿½1 family, polypeptide A1 polymorphism prior to the inititation of irinotecan-based chemotherapy. Mol Clin Oncol (2014). 287 Ioannidis, J. P. To replicate or not to replicate: the case of pharmacogenetic studies: Have pharmacogenomics failed, or do they just need larger-scale evidence and more replication? Circ Cardiovasc Genet 6, 413-418; discussion 418 (2013).

148

288 Han, J.-Y. et al. A genome-wide association study of survival in small-cell lung cancer patients treated with irinotecan plus cisplatin chemotherapy. Pharmacogenomics J (2013). 289 Han, J.-Y. et al. A genome-wide association study for irinotecan-related severe toxicities in patients with advanced non-small-cell lung cancer. Pharmacogenomics J, 1-6 (2012). 290 Takahashi, H. et al. Application of a combination of a knowledge-based algorithm and 2-stage screening to hypothesis-free genomic data on irinotecan-treated patients for identification of a candidate single nucleotide polymorphism related to an adverse effect. PLoS One 9, e105160 (2014). 291 Aslibekyan, S., Claas, S. A. & Arnett, D. K. To replicate or not to replicate: the case of pharmacogenetic studies: Establishing validity of pharmacogenomic findings: from replication to triangulation. Circ Cardiovasc Genet 6, 409-412; discussion 412 (2013). 292 Greene, C. S., Penrod, N. M., Williams, S. M. & Moore, J. H. Failure to replicate a genetic association may provide important clues about genetic architecture. PLoS One 4, e5639 (2009). 293 Lam, Y. W. Scientific challenges and implementation barriers to translation of pharmacogenomics in clinical practice. ISRN Pharmacol 2013, 641089 (2013). 294 McGowan, M. L., Settersten, R. A., Jr., Juengst, E. T. & Fishman, J. R. Integrating genomics into clinical oncology: ethical and social challenges from proponents of personalized medicine. Urol Oncol 32, 187-192 (2014). 295 Goldstein, D. A., Shaib, W. L. & Flowers, C. R. Costs and effectiveness of genomic testing in the management of colorectal cancer. Oncology (Williston Park) 29, 175- 183 (2015). 296 Patel, H. N., Ursan, I. D., Zueger, P. M., Cavallari, L. H. & Pickard, A. S. Stakeholder views on pharmacogenomic testing. Pharmacotherapy 34, 151-165 (2014). 297 Deverka, P. A. & Dreyfus, J. C. Clinical integration of next generation sequencing: coverage and reimbursement challenges. J Law Med Ethics 42 Suppl 1, 22-41 (2014). 298 Butts, C. et al. Benefits, issues, and recommendations for personalized medicine in oncology in Canada. Curr Oncol 20, e475-483 (2013). 299 Gold, H. T., Hall, M. J., Blinder, V. & Schackman, B. R. Cost effectiveness of pharmacogenetic testing for uridine diphosphate glucuronosyltransferase 1A1 before irinotecan administration for metastatic colorectal cancer. Cancer 115, 3858-3867 (2009). 300 Pichereau, S. et al. Cost-effectiveness of UGT1A1*28 genotyping in preventing severe neutropenia following FOLFIRI therapy in colorectal cancer. J Pharm Pharm Sci 13, 615-625 (2010). 301 Obradovic, M., Mrhar, A. & Kos, M. Cost–effectiveness of UGT1A1genotyping in second-line, high-dose, once every 3 weeks irinotecan monotherapy treatment of colorectal cancer. Pharmacogenomics 9, 539-549 (2008). 302 The Office of the Privacy Commissioner of Canada. Office of the Privacy Commissioner of Canada issues statement on the use of genetic test results by life and health insurance companies, (2014) Last accessed. April, 23 2015 303 Toffoli, G. The Role of UGT1A1*28 Polymorphism in the Pharmacodynamics and Pharmacokinetics of Irinotecan in Patients With Metastatic Colorectal Cancer. J Clin Oncol 24, 3061-3068 (2006).

149 304 Stephens, M., Smith, N. J. & Donnelly, P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet 68, 978-989 (2001). 305 Stephens, M. & Donnelly, P. A comparison of bayesian methods for haplotype reconstruction from population genotype data. Am J Hum Genet 73, 1162-1169 (2003). 306 Petrides, A. K., Clarke, W. & Marzinke, M. A. Application and Utility of Pharmacogenetics in the Clinical Laboratory. J Analyt Molecul Tech. 307 Jiang, W., Xu, B., Wu, B., Yu, R. & Hu, M. UDP-glucuronosyltransferase (UGT) 1A9-overexpressing HeLa cells is an appropriate tool to delineate the kinetic interplay between breast cancer resistance protein (BRCP) and UGT and to rapidly identify the glucuronide substrates of BCRP. Drug Metab Dispos 40, 336-345 (2012). 308 Xiang, C. et al. Pulmonary expression of CYP2A13 and ABCB1 is regulated by FOXA2 and their genetic interaction is associated with lung cancer. FASEB J (2015). 309 Meisenberg, C. et al. Clinical and cellular roles for TDP1 and TOP1 in modulating colorectal cancer response to irinotecan. Mol Cancer Ther 14, 575-585 (2015). 310 Wheeler, H. E., Maitland, M. L., Dolan, M. E., Cox, N. J. & Ratain, M. J. Cancer pharmacogenomics: strategies and challenges. Nat Rev Genet 14, 23-34 (2013). 311 Ken-Dror, G., Humphries, S. E. & Drenos, F. The Use of Haplotypes in the Identification of Interaction between SNPs. Human Heredity 75, 44-51 (2013). 312 Chapman, J. M., Cooper, J. D., Todd, J. A. & Clayton, D. G. Detecting Disease Associations due to Linkage Disequilibrium Using Haplotype Tags: A Class of Tests and the Determinants of Statistical Power. Human Heredity 56, 18-31 (2003). 313 Balding, D. J. A tutorial on statistical methods for population association studies. Nat Rev Genet 7, 781-791 (2006). 314 Barbarino, J. M., Haidar, C. E., Klein, T. E. & Altman, R. B. PharmGKB summary. Pharmacogenet Genomics 24, 177-183 (2014). 315 Genomes Project, C. et al. An integrated map of genetic variation from 1,092 human genomes. Nature 491, 56-65 (2012). 316 Ribaudo, H. J. et al. Impact of UGT1A1 Gilbert variant on discontinuation of ritonavir-boosted atazanavir in AIDS Clinical Trials Group Study A5202. J Infect Dis 207, 420-425 (2013). 317 Bielinski, S. J. et al. Preemptive genotyping for personalized medicine: design of the right drug, right dose, right time-using genomic data to individualize treatment protocol. Mayo Clin Proc 89, 25-33 (2014). 318 Bertrand, J., De Iorio, M. & Balding, D. J. Integrating dynamic mixed-effect modelling and penalized regression to explore genetic association with pharmacokinetics. Pharmacogenet Genomics 25, 231-238 (2015). 319 Hirose, K. et al. Correlation between plasma concentration ratios of SN-38 glucuronide and SN-38 and neutropenia induction in patients with colorectal cancer and wild-type UGT1A1 gene. Oncol Lett 3, 694-698 (2012). 320 Hirose, K. et al. Usefulness of one-point plasma SN-38G/SN-38 concentration ratios as a substitute for UGT1A1 genetic information after irinotecan administration. Int J Clin Oncol 19, 397-402 (2014). 321 Stingl, J. C., Bartels, H., Viviani, R., Lehmann, M. L. & Brockmöller, J. Relevance of UDP-glucuronosyltransferase polymorphisms for drug dosing: A quantitative systematic review. Pharmacol Ther 141, 92-116 (2014). 322 Lévesque, E., Benoit-Biancamano, M.-O., Delage, R., Couture, F. & Guillemette, C. Pharmacokinetics of mycophenolate mofetil and its glucuronide metabolites in healthy volunteers. Pharmacogenomics 9, 869-879 (2008).

150

323 Prausa, S. E. et al. UGT genotype may contribute to adverse events following medication with mycophenolate mofetil in pediatric kidney transplant recipients. Clin Pharmacol Ther 85, 495-500 (2009). 324 Thurman, R. E. et al. The accessible chromatin landscape of the human genome. Nature 489, 75-82 (2012). 325 Grossniklaus, U. & Paro, R. Transcriptional silencing by polycomb-group proteins. Cold Spring Harb Perspect Biol 6, a019331 (2014). 326 Ernst, J. & Kellis, M. Discovery and characterization of chromatin states for systematic annotation of the human genome. Nature Biotechnology 28, 817-838 (2010). 327 Hasin-Brumshtein, Y. et al. Allele-specific expression and eQTL analysis in mouse adipose tissue. BMC Genomics 15, 471 (2014). 328 Fogarty, M. P., Panhuis, T. M., Vadlamudi, S., Buchkovich, M. L. & Mohlke, K. L. Allele-specific transcriptional activity at type 2 diabetes-associated single nucleotide polymorphisms in regions of pancreatic islet open chromatin at the JAZF1 locus. Diabetes 62, 1756-1762 (2013). 329 Maynard, N. D., Chen, J., Stuart, R. K., Fan, J. B. & Ren, B. Genome-wide mapping of allele-specific protein-DNA interactions in human cells. Nat Methods 5, 307-309 (2008). 330 Kadota, M. et al. Allele-specific chromatin immunoprecipitation studies show genetic influence on chromatin state in human genome. PLoS Genet 3, e81 (2007).

151