Inherited Predisposition to Pancreatic Cancer in the Era of Next Generation Sequencing

Jeremy Humphris

A thesis in fulfillment of the requirements for the degree of Doctor of Philosophy

St Vincent’s Clinical School

Faculty of Medicine

August 2015

PLEASE TYPE THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet

Surname or Family name:

First name: Other name/s:

Abbreviation for degree as given in the University calendar:

School: Faculty:

Title:

Abstract 350 words maximum: (PLEASE TYPE)

Declaration relating to disposition of project thesis/dissertation

I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

…………………………………………………………… ……………………………………..……………… ……….……………………...…….… Signature Witness Date

The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

ORIGINALITY STATEMENT

‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed ……………………………………………......

Date ……………………………………………......

‘I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation. I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstract International (this is applicable to doctoral theses only). I have either used no substantial portions of copyright material in my thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.'

Signed ……………………………………………......

Date ……………………………………………......

AUTHENTICITY STATEMENT

‘I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format.’

Signed ……………………………………………......

Date ……………………………………………...... Acknowledgements: ...... 4 List of Figures: ...... 7 List of Tables: ...... 9 Chapter 1. Introduction ...... 14 1.1 Epidemiology of PC...... 15 1.1.1 Non-genetic risk factors ...... 15 1.1.2 Genetic risk factors ...... 22 1.1.2.1 Hereditary Tumour Predisposition Syndromes ...... 23 1.1.2.2 Hereditary Pancreatitis ...... 26 1.1.2.3 Familial Pancreatic Cancer ...... 27 1.1.2.4 Low Penetrance Susceptibility Variants ...... 29 1.2 Precursor Lesions and Progression to PC ...... 34 1.2.1 Pancreatic Intra-epithelial Neoplasia...... 34 1.2.2 Cystic Neoplasms ...... 36 1.3 Predicting the risk of developing PC ...... 38 1.4 Early Detection of PC ...... 39 1.4.1 Imaging approach ...... 42 1.4.2 Biomarker approach ...... 45 Chapter 2: The prognostic and predictive value of serum CA19.9 in pancreatic cancer...... 54 2.1 Introduction ...... 56 2.2 Patients and Methods ...... 57 2.3 Results ...... 61 2.3.1 CA19.9 and disease stage ...... 61 2.3.2 Prognostic Value of CA19.9 ...... 61 2.3.3 Predictive Value of CA19.9 ...... 65 2.4 Discussion ...... 68 Chapter 3: Clinico-pathological features of familial PC ...... 72 3.1 Introduction ...... 73 3.2 Patients and methods ...... 74 3.3 Results ...... 76 3.3.1 Clinico-pathologic variables and outcome ...... 77 3.3.2 Previous extra-pancreatic malignancy ...... 79 3.3.4 Family history of extra-pancreatic malignancy ...... 80 3.3.5 Other PC risk factors ...... 80 3.4 Discussion ...... 82 Chapter 4: Hypermutation in pancreatic cancer ...... 85 4.1 Introduction ...... 87 4.2 Patients and Methods ...... 89 4.3 Results ...... 94 4.4 Discussion ...... 103 Chapter 5: Comprehensive Assessment of Germline Cancer Predisposition Genes in Pancreatic Cancer ...... 107 5.1 Introduction ...... 109 5.2 Patients and methods ...... 110 5.3 Results ...... 114 5.3.1 Clinical Cohort ...... 114

2 5.3.2 Pathogenic mutations in PC predisposition genes ...... 120 5.3.3 Pathogenic mutations in other cancer predisposition genes ...... 124 5.3.4 Germline copy number variants ...... 134 5.4 Discussion ...... 135 Chapter 6: Discussion...... 138 Appendix ...... 148 Chapter 2: Supplementary data ...... 149 Chapter 3: Supplementary data ...... 154 Chapter 4 Supplementary Data ...... 167 Chapter 5 Supplementary data ...... 170 Reference: ...... 181

3 Acknowledgements: Undertaking this PhD has been a privilege but would not have come to fruition without the support of several people.

Foremost, I would like to express sincere gratitude to my supervisor, Professor Andrew Biankin, whose vision for the Australian Pancreatic Genome Initiative (APGI) allowed me to undertake this study and who provided continual critical thinking, motivation and perspective. I would like to thank current and former members of the pancreatic cancer research group for their counsel, support and friendship. I would like to also thank all members and collaborators within the Australian Pancreatic Genome Initiative (www.pancreaticcancer.net.au). Furthermore, I would like to thank all the patients who altruistically contributed biospecimens to the APGI.

My wife, Mahila has been a constant and unwavering source of unconditional support, counsel and encouragement through the duration of this study, without this these studies would not have been possible. To my children, Harry and Aaron for their unconditional love and providing me with perspective. To my parents and siblings, for all their love and support during my formative years and beyond, allowing me to pursue my studies.

Several funding bodies provided me with scholarships which enabled me to undertake these studies and for this I would like to express my gratitude, these include the Gastroenterological Society of Australia, University of New South Wales and the Avner Pancreatic Cancer Foundation.

Abbreviations and Symbols: A Adenine AJCC American Joint Committee on Cancer APC Adenomatous polyposis coli APGI Australian Pancreatic Genome Initiative ATM Ataxia Telangiectasia mutated BMI Body mass index BRCA1 Breast cancer 1, early onset BRCA2 Breast cancer 2, early onset C Cytosine CA-125 Carbohydrate antigen 125 CA19.9 Carbohydrate antigen 19.9 cCA19.9 Corrected CA19.9 CDKN2A Cyclin dependent kinase inhibitor 2A CEA Carcino-embryonic antigen cfDNA cell-free DNA cfNA cell-free nucleic acids CI Confidence interval CONKO Charite Onkologie CP Chronic pancreatitis CT Computerised tomography CTC circulating tumour cells dbSNP Database of single nucleotide polymorphisms DM Diabetes mellitus DNA Deoxy-ribonucleic acid DSS Disease-specific survival ESPAC European study of adjuvant chemotherapy in resectable PC FA Fanconi anaemia FAMMM Familial atypical multiple mole melanoma FAP Familial adenomatous polyposis FDR First-degree relative FPC Familial pancreatic cancer G Guanine GWAS genome wide association study HBOC Hereditary breast-ovarian cancer HGMD Human Gene Mutation Database HNPCC Hereditary non-polyposis colorectal cancer HR Homologous recombination ICGC International Cancer Genome Consortium IHC Immunohistochemistry InSIGHT International Society for Gastrointestinal Hereditary Tumours IPMN Intraductal papillary mucinous neoplasm LCA Lobulocentric atrophy LFS Li-Fraumeni syndrome LOH Loss of heterozygosity MAF Minor allele frequency MCN Mucinous cystic neoplasm 5 mg/dL milligrams per deciliter miRNA MicroRNA MLH1 MutL homolog 1 MMR Mismatch repair MRCP Magnetic resonance cholangio-pancreatography MRI Magnetic resonance imaging mRNA Messenger RNA MSH2 MutS homolog 2 MSH6 MutS homolog 2 MSI Microsatellite instability MSI Microsatellite instability Mut/Mb Mutations per megabase NFPTR National familial pancreatic tumour registry NSWPCN New South Wales pancreatic cancer network OMIM Online Mendelian Inheritance in Man OR Odds ratio PALB2 Partner and localizer BRCA2 PanIN Pancreatic intra-epithelial neoplasia PARP Poly ADP ribose polymerase PC Pancreatic cancer PCR Polymerase chain reaction PJS Peutz-Jeghers syndrome PMS2 Post meiotic segregation 2 PSA Prostate specific antigen RNA Ribonucleic acid RTOG Radiation therapy oncology group SCN Serous cystic neoplasm SDR Second-degree relative SNP Single nucleotide polymorphism SNV Single nucleotide variant SPC Sporadic pancreatic cancer SPN Solid pseudopapillary neoplasm STK11 Serine/threonine kinase 11 T Thymine TP53 Tumour protein 53 U/mL International units per milliliter UTR Untranslated region VAF Variant allele fraction WES Whole exome sequencing WGS Whole genome sequencing WT Wild-type

6 List of Figures:

Chapter 1 Figure 1: Overview of the patient’s journey to the development of PC and the current diagnostic tests and opportunities to detect early disease.

Chapter 2 Figure 1A – F: Kaplan-Meier survival curves for the prognostic value of CA19.9 Figure 2A - F: Kaplan-Meier survival curves for the predictive value of CA19.9 Figure 3: Schematic representation of suggested time-points for CA19.9 measurements in clinical trials.

Chapter 3 Figure 1A - D: Kaplan-Meier survival curves for FPC and SPC patients according to resection status and presence of high grade PanIN

Chapter 4 Figure 1: Box and whisker plot showing the distribution of mutation burden in exome and whole genome sequenced groups Figure 2: Venn diagram showing the proportion of patients with personal or family history of malignancy in particular pancreatic (PC) and the colo-rectal (CRC) and endometrial cancer (EC). Figure 3: Overview of the methods used to characterize the cohort and define the mutational mechanisms producing increased mutation burden. Sequence data was analysed for 392 PC was performed. Hypermutated samples were detected and MMR was estimated from sequence data with the MSIsensor tool (MSIs) and IHC (immunohistochemistry). N normal, A abnormal. The cutoff used for an abnormal MSIsensor score was 3.5% Figure 4: Immunohistochemistry demonstrating MMR deficiency and the genomic mechanisms. 4A: Absent MSH2 staining in tumour cells but normal staining in stroma in ICGC_0076. 4B: Absent MSH2 in tumour cells in ICGC_0548. 4C: Tumours typically show concordant loss of the obligate binding MMR protein in this case MSH6 (ICGC_0090). 4D: Absent MLH1 in tumour cells in ICGC_0297. 4E: Somatic homozygous deletion involving MSH2 (ICGC_0076). 4F: Somatic foldback rearrangement between EPCAM and MSH2 (ICGC_0548). 4G: Somatic MSH2 splice site mutation (ICGC_0090). 4H: Beta values for methylation probes in the MLH1 promoter region

Chapter 5 Figure 1: Pie chart showing the distribution of germline mutations in cancer predisposition genes in patients with pancreatic cancer. Figure 2: Venn diagram showing the intersection of family history and personal of extra-pancreatic malignancy (EPM) with presence of a mutation in an

7 established PC predisposition gene (Pancreatic CPG) and other CPG. For the purpose of the figure, family history of PC and EPMs has been pooled together. Figure 3: Lolliplots of pathogenic mutations in established PC predisposition genes Figure 4: Lolliplots of pathogenic mutations in other CPGs

Appendix Chapter 3 Supplementary Figure 1A - E: Kaplan-Meier survival curves for FPC and SPC patients according to DM status. Supplementary Figure 2: Kaplan-Meier survival curves for FPC and SPC patients according to smoking and alcohol intake.

8 List of Tables:

Chapter 1 Table 1: Non-genetic risk factors for PC Table 2: Genetic risk factors for PC – hereditary cancer syndrome and moderate to high penetrance genes Table 3: Common low-penetrance genetic factors associated with increased risk of PC from GWAS Table 4: Summary of PC screening trials using a predominantly imaging based approach in high-risk individuals

Chapter 2 Table 1: Descriptive Statistics for Patients Resected for PC with CA19.9 values (n = 260) Table 2: Univariate Analysis of CA19.9 Values at Significant Time Points Table 3: Multivariate analysis

Chapter 3 Table 1: Distribution of relatives with PC and extra-pancreatic malignancy (EPM) Table 2: Comparison of clinico-pathological variables in resected PC patients Table 3: Risk factors for PC

Chapter 4 Table 1: Overview of patients with somatic hypermutation

Chapter 5 Table 1: Overview of the cohort demographics and personal and family history of malignancy Table 2: PC predisposition genes and other candidate genes with pathogenic variants

Appendix

Chapter 2 Supplementary Table 1: Clinico-pathological parameters and outcome (n = 904) Supplementary Table 2: Descriptive Statistics for the cohort who contributed CA 19.9 values at each time-point Supplementary Table 3: Pre-resection CA 19.9 and Pathologic Stage (non- expressors excluded) Supplementary Table 4: Post-resection CA 19.9 < 3 months and disease- specific survival

9 Supplementary Table 5: Post-resection CA 19.9 and disease-free survival in patients with an R0 resection Supplementary Table 6: Change in cCA 19.9 value after resection Supplementary Table 7: Pre-resection CA 19.9 and survival Supplementary Table 8: Pre-adjuvant CA 19.9 and benefit from adjuvant chemotherapy (49 of 67 completed  3 cycles)

Chapter 3 Supplementary Table 1: Clinico-pathological variables in resected PC patients Supplementary Table 2: Clinico-pathological variables in non-resected PC patients Supplementary Table 3: Previous extra-pancreatic malignancy Supplementary Table 4: Distribution of EPM in FDRs Supplementary Table 5: Risk factors for PC Supplementary Table 6: Mean age at diagnosis in resected patients Supplementary Table 7: Multivariate analysis in FPC and SPC resected patients

Chapter 4 Supplementary table 1: Clinico-pathologic features of patients with resected PC. Supplementary table 2: Clinico-pathologic features of patients with non- resected PC. Available on accompanying USB: Supplementary table 3: Overview of the 392 PC cohort and the classification by somatic hypermutation, MSIsensor and immunohistochemistry. Supplementary table 4: Hypermutated tumours, somatic readouts and proposed aetiology. Supplementary table 5: Mutational signatures in tumours that underwent WGS Supplementary table 6: All germline MMR (MSH2, MLH1, MSH6 and PMS2) and replicative polymerase (POLE and POLD1) single nucleotide variants and SNV and small indels in the cohort and their variant classification. Supplementary table 7: Somatic variants in the MMR genes (MSH2, MLH1, MSH6 and PMS2) and replicative polymerase (POLE and POLD1) Supplementary table 8: Methylation array for probes (beta-values) in the region of the MLH1 promoter in tumour DNA Supplementary table 9: Methylation array for probes (beta-values) in the region of the MLH1 promoter in adjacent non-tumour DNA Supplementary table 10: All patients in the cohort with a pathogenic germline or somatic mutation in BRCA1, BRCA2 and PALB2

10 Chapter 5 Supplementary Table 1: Complete list of candidate cancer predisposition genes. Supplementary Table 2: Criteria for the initial bioinformatic classification and the final classification after manual of germline variants Supplementary Table 3: Clinico-pathological features of resected PC patients Supplementary Table 4: Clinico-pathologic features of non-resected PC patients Available on accompanying USB: Supplementary Table 5: Overview of pathogenic germline variants, clinico- pathologic features and somatic readouts

Publications, abstracts and book chapters arising from work related to this thesis

Manuscripts:

Humphris JL, Chang DK, Johns AL, Scarlett CJ, Pajic M, Jones MD, Colvin EK, Nagrial A, Chin VT, Chantrill LA, Samra JS, Gill AJ, Kench JG, Merrett ND, Das A, Musgrove EA, Sutherland RL, Biankin AV; NSW Pancreatic Cancer Network. The prognostic and predictive value of serum CA19.9 in pancreatic cancer. Ann Oncol. 2012 Jul;23(7):1713-22.

Humphris JL, Johns AL, Simpson SH, Cowley MJ, Pajic M, Chang DK, Nagrial AM, Chin VT, Chantrill LA, Pinese M, Mead RS, Gill AJ, Samra JS, Kench JG, Musgrove EA, Tucker KM, Spigelman AD, Waddell N, Grimmond SM, Biankin AV; Australian Pancreatic Cancer Genome Initiative. Clinical and pathologic features of familial pancreatic cancer. Cancer. 2014;120(23):3669-75.

Humphris J, Chang DK, Biankin AV Inherited susceptibility to pancreatic cancer in the era of next-generation sequencing. Gastroenterology. 2015 Mar;148(3):496-8.

Johns AL, Miller DK, Simpson SH, Gill AJ, Kassahn KS, Humphris JL, Samra JS, Tucker K, Andrews L, Chang DK, Waddell N, Pajic M, Pearson JV, Grimmond SM, Biankin AV, Zeps N. Returning individual research results for genome sequences of pancreatic cancer. Genome Med. 2014 May 29;6(5):42.

Nagrial AM, Chang DK, Nguyen NQ, Johns AL, Chantrill LA, Humphris JL, Chin VT, Samra JS, Gill AJ, Pajic M; Australian Pancreatic Cancer Genome Initiative, Pinese M, Colvin EK, Scarlett CJ, Chou A, Kench JG, Sutherland RL, Horvath LG, Biankin AV.Adjuvant chemotherapy in elderly patients with pancreatic cancer. Br J Cancer. 2014 Jan 21;110(2):313-9.

Chou A, Waddell N, Cowley MJ, Gill AJ, Chang DK, Patch AM, Nones K, Wu J, Pinese M, Johns AL, Miller DK, Kassahn KS, Nagrial AM, Wasan H, Goldstein D, Toon CW, Chin V, Chantrill L, Humphris J, Mead RS, Rooman I, Samra JS, Pajic M, Musgrove EA, Pearson JV, Morey AL, Grimmond SM, Biankin AV. Clinical and molecular characterization of HER2 amplified-pancreatic cancer. Genome Med. 2013 Aug 31;5(8):78.

Chantrill LA, Nagrial AM, Watson C, Johns AL, Martyn-Smith M, Simpson S, Mead S, Jones MD, Samra JS, Gill AJ, Watson N, Chin VT, Humphris JL, Chou A, Brown B, Morey A, Pajic M, Grimmond SM, Chang DK, Thomas D, Sebastian L, Sjoquist K, Yip S, Pavlakis N, Asghari R, Harvey S, Grimison P, Simes J, Biankin AV; Australian Pancreatic Cancer Genome Initiative (APGI).

12 Precision Medicine for Advanced Pancreas Cancer: The Individualized Molecular Pancreatic Cancer Therapy (IMPaCT) Trial. Clin Cancer Res. 2015 May 1;21(9):2029-37.

Waddell N, Pajic M, Patch AM …. Humphris J, …. et al. Whole genomes redefine the mutational landscape of pancreatic cancer. Nature. 2015 Feb 26;518(7540):495-501.

Chang DK, Jamieson NB, Johns AL, Scarlett CJ, Pajic M, Chou A, Pinese M, Humphris JL, Jones MD, Toon C, Nagrial AM, Chantrill LA, Chin VT, Pinho AV, Rooman I, Cowley MJ, Wu J, Mead RS, Colvin EK, Samra JS, Corbo V, Bassi C, Falconi M, Lawlor RT, Crippa S, Sperandio N, Bersani S, Dickson EJ, Mohamed MA, Oien KA, Foulis AK, Musgrove EA, Sutherland RL, Kench JG, Carter CR, Gill AJ, Scarpa A, McKay CJ, Biankin AV. Histomolecular phenotypes and outcome in adenocarcinoma of the ampulla of vater. J Clin Oncol. 2013 Apr 1;31(10):1348-56

Biankin AV, Waddell N, Kassahn KS …. Humphris JL, …. et al. Pancreatic cancer genomes reveal aberrations in axon guidance pathway genes. Nature. 2012 Nov 15;491(7424):399-405.

Published Abstracts:

Jeremy Humphris, Amber Johns, Skye Simpson, David Chang ,Mark Cowley, Marina Pajic, Marc Jones, Scott Mead, Andrew Biankin. Risk factors for pancreatic cancer in patients with and without a family history of pancreatic cancer. Journal of Gastroenterology and Hepatology 2012; Volume 27, Issue Supplement s4, p127.

Jeremy Humphris, Jiang Tao, David Chang, Amber Johns, Mark Cowley, Marina Pajic, Marc Jones, Skye Simpson, Scott Mead, Andrew Biankin. Somatic mutations in pancreatic ductal adenocarcinoma using amplicon sequencing from formalin fixed paraffin embedded tissue. Journal of Gastroenterology and Hepatology 2012; Volume 27, Issue Supplement s4, p127.

Book chapter:

Humphris JL. Hereditary and Idiopathic Pancreatitis. Australian Handbook of Clinical Pancreatology. In press.

Chapter 1. Introduction

14 Pancreatic cancer (PC) is characterized by late stage at diagnosis with advanced local disease or metastases. Pancreatectomy offers the only potential for cure, but is only possible in a minority of patients. The current diagnostic algorithm for patients diagnosed with PC requires symptoms or incidental imaging findings suggestive of the disease. As a consequence few patients are diagnosed with early stage disease. Early detection of PC is an attractive strategy to improve the outcome for this highly aggressive disease. In order to achieve this robust screening tests are required which will likely need to focus on well-defined cohorts at increased risk. Even in those patients who undergo resection, most succumb to the disease due to progression of subclinical metastatic disease present at the time of diagnosis1. Systemic therapies are only modestly effective in advanced disease, but appear to have a greater impact in the adjuvant setting, with 5-Fluorouracil and Gemcitabine both having efficacy in a subgroup of patients and increasing five year survival from 10-15% with surgery alone, to 20-25%. 2-5 As a consequence, there is an urgent need to develop biomarkers for early detection, to better stratify patients for current treatment modalities, and for the development of novel therapeutic strategies.

In this chapter I will review the published literature on the genetic and non- genetic risk factors for PC, the precursor lesions and their progression to PC. Lastly, I will discuss the early detection of PC in individuals at genetic high risk.

1.1 Epidemiology of PC

1.1.1 Non-genetic risk factors Epidemiologic studies have identified several environmental and lifestyle risk factors for PC including age, race, sex, smoking, alcohol, diabetes mellitus, chronic pancreatitis, dietary components and obesity6-9. These frequently coexist and are likely to interact6.

15 Risk factor Estimated risk (95% CI)

Active Cigarette Smoking OR 2.20 (1.71-2.83)

Ceased Cigarette Smoking

> 1 but < 10 years OR 1.64 (1.36-1.97)

> 10 years OR 1.12 (0.86-1.44)

Diabetes Mellitus

< 2 years duration RR 7.94 (4.70-12.55)

>10 years duration OR 1.51 (1.16-1.96)

BMI (>35 vs. <25) OR 1.55 (1.16-2.07)

Heavy Alcohol (≥6 drinks/day) OR 1.46 (1.16-1.83)

Chronic Pancreatitis (>2 years) OR 2.71 (1.96-3.74)

Allergy (hay fever and animal allergy) OR 0.73 (0.64 – 0.74) Table 1: Non-genetic risk factors for PC

Age PC is largely a disease of advancing age which is rarely diagnosed before 40 years of age10. Only 5-10% of cases are diagnosed before 50 years but these are likely to be enriched with individuals with an inherited genetic predisposition6. The incidence increases exponentially in both sexes after age 40 from 2.3 cases per 100,000 for 40-44 year olds to 57 cases per 100,000 in those 70-74 years11.

Ethnicity The incidence of PC is higher in people of African American and Ashkenazi Jewish heritage compare to those of Caucasian, Hispanic and Asian descent6,12. This is likely the culmination of both genetic and non-genetic risk factors.

Gender In comparison to ethnicity, there are only small gender differences in the risk of developing PC. The lifetime risk of developing PC before age 75 for males is

16 0.9% and 0.6% for females11. This has been attributed to higher cigarette smoking rates in men as opposed to different gender-specific endocrine profiles6.

Cigarette Smoking Cigarette smoking is the best characterized environmental risk factor for the development of PC13. Analyses of data from 12 case-control studies showed that current smokers have a 2.2-fold (95% Confidence Interval (CI) 1.71–2.83) increased risk of PC compared with never smokers14. It is estimated that population attributable risk of PC due to cigarette smoking is 20% in a population with 30% smokers15,16. The risk is highest in current smokers with a 75% increased risk of developing PC and lower age exhibits dose-dependent effects15. This increased risk remains elevated for at least 10 to 15 years after cessation14,15,17. Risk estimates of 1.64 (Odds Ratio (OR), 95% CI 1.36–1.97) have been reported for those who ceased between 1 and 10 years ago and 1.12 (95% CI 0.86–1.44) for individuals who stopped > 10 years ago14. Studies in familial PC (FPC) have confirmed cigarette smoking as a major risk factor with a nearly 4-fold increased risk and the development of PC ten years before non-smokers18,19. This effect is more pronounced in males and subjects under the age of 50. As discussed in chapter 3 in our cohort patients with sporadic PC (SPC) were more likely to be active smokers at time of PC diagnosis and had higher exposure to cigarette smoke than FPC patients. This may suggest that individuals with a family history of PC require a reduced tobacco exposure for the development of PC20. Furthermore, active smokers were diagnosed with PC at a significantly younger age than non-smokers. Several germline PC risk genes are involved in the repair of DNA and there is evidence that polymorphisms in DNA repair genes can alter PC risk and show an interaction with smoking21,22. Tobacco smoke contains over 60 carcinogens which can directly or indirectly through metabolites form DNA adducts and produce characteristic mutation patterns23,24. The most well-described mutations are transversions (G>T/C>A)24 but double stranded DNA breaks are also seen due to oxygen free radicals25. These mutations require intact DNA repair pathways to correct, this includes nucleotide excision repair for bulky adducts and homologous recombination for DSBs23,25. Carriers of

17 polymorphisms in DNA repair pathway members such as XRCC126 (base- excision repair), XRCC225 (homologous recombination) and ERCC227 who smoke are at increased risk of developing PC. These findings suggest an interaction and potential synergism between environmental risk factors and inherited polymorphisms leading to increased risk of PC28. Increased risk of PC has been associated with other tobacco products such as pipes and cigars and smokeless tobacco e.g. oral chewing tobacco29. Furthermore environmental tobacco smoke exposure (passive or second-hand) and potentially in utero exposure from maternal smoking may predispose to development of PC.30 However, mutational signatures typical of direct tobacco exposure such as those seen in lung cancer are not a feature of PC31.

Alcohol Alcohol consumption has also been found to be related to increased risk of PC particularly in those with chronic pancreatitis but this is not a consistent finding with other studies reporting either no or a weak association32. Confounding this issue is the strong relationship between alcohol consumption with smoking. Intake of less than 3-4 standard drinks per day is not associated with mortality from PC but above this threshold there appears to be a dose-response relationship33,34. Consumption of > 3 and  6 standard drinks per day has an OR of 1.22 and 1.5 for the development of PC and similar to cigarette smoking this effect persists for 10 years after cessation35,36. This effect may be stronger with liquor consumption than beer and is independent of cigarette smoking with which it frequently co-exists33.

Diabetes Mellitus Diabetes mellitus (DM) frequently co-exists with PC (up to 40-50%) is likely to be both a risk factor for PC development and an early manifestation of PC 13,37,38. The risk of diagnosis of PC depends on the patients duration of DM with the highest risk in patients with recent onset DM. Approximately 1% of people over 50 years will be diagnosed with PC within 3 years of their diagnosis of DM39. Combined results from several studies have shown that longstanding DM (> 10 years) increases risk of PC development by 30-50%, whereas recent 18 onset DM (< 2 years) is associated with a 3-fold increased risk40-42. The latter finding likely represents reverse causality due to cancer-related endocrine dysfunction13, as evidenced by progressive hyperglycaemia with declining body mass index beginning 2 years before PC diagnosis43. Another study suggested the risk for PC was highest for DM duration between 2 to 8 years and found no association with DM duration over 9 years44. In the cohort described in Chapter 3 the prevalence of DM was 29% and was not different between FPC and SPC groups. However in both groups patients with long-standing DM (> 2 years) had significantly poorer post-resection survival. The aetiology of the coexistence of DM with PC is not uniform and there are likely to be multiple mechanisms involved. These include destructive changes as a direct result of the PC which reduces B cell mass and reduction in insulin secretion and peripheral insulin resistance45. Diabetic control can improve post- resection suggesting a para-neoplastic phenomenon rather than irreversible destruction of B cells in some cases. The ability to utilise DM as a marker to enrich for individuals with asymptomatic PC is limited, primarily due to the current inability to differentiate PC-associated DM from the much more common type II DM.45 Anti-diabetic therapy may further refine the risk of developing PC with metformin associated with a reduced risk and insulin and insulin secretagogues associated with increased risk46-48.

Obesity and Physical Inactivity Obesity as defined by increased body mass index (BMI) has been associated with an increased risk of developing PC independent of its association with DM49-51. Multiple cohort and case control studies have estimated that the risk of PC is increased 1.55-fold (95% CI=1.16 – 2.07) for individuals with a BMI >35 compared to individuals with a BMI < 258. The risk is particularly associated with central adiposity, which correlates with measures of insulin resistance52,53. Furthermore obesity is associated with an earlier age at diagnosis on average by 2-6 years7. There is no conclusive evidence for an independent role for physical inactivity with some showing a protective effect54 and others no change52,53.

Chronic Pancreatitis

19 Chronic pancreatitis (CP) is a pathological diagnosis reflecting progressive and irreversible fibro-inflammatory changes from sustained or repeated pancreatic insults from several different aetiologies55. Chronic pancreatitis elevates the risk of developing PC but similar to DM exhibits reverse causality with PC. An analysis of data from 10 case-control studies demonstrated that individuals with a history of chronic pancreatitis have a 2.71-fold (95% CI 1.96–3.74) increased risk of PC56. Estimation of the true risk is complicated by the frequent coexistence of confounding variables57,58. The risk of PC is particularly high in those with autosomal dominant hereditary pancreatitis due to gain-of-function PRSS1 mutations in predominantly due to the early age at first presentation of pancreatitis.

Infection Studies have implicated both past and current infection with hepatitis B or C as risk factors for PC, with the relationship stronger for hepatitis B infection59,60. Chronic helicobacter pylori infection has also been weakly linked albeit inconsistently with increased risk of developing PC61-63.

Allergies Epidemiologic studies have demonstrated reduced risk of developing PC in those with atopy. A meta-analysis of 14 PC studies showed self- reported history of any allergy was associated with a 30% reduced risk of developing PC whereas those who reported hayfever had a 45% reduction in risk64 . A more recent meta-analysis which pooled results from 10 case- control studies with 3,567 cases and 9,145 controls found an odds ratio for any allergy of 0.79 (95% confidence interval (95% confidence interval 0.62 - 1.00) but there was heterogeneity attributable to one study. When this study was excluded the pooled odds ratio was 0.73 (95% confidence interval 0.64 - 0.84)65. In particular, allergic rhinitis and animal allergy were related to lower risk of PC but not other allergies or asthma. The coexistence of asthma with respiratory allergies is frequent but not exclusive and subsequent studies have yielded mixed results66,67 Objective measures of an atopic state have also yielded mixed results but overall provide support that allergy sufferers have a lower risk of PC.

20 Specifically, positive skin prick tests for hay fever allergens, dust/mould and animal allergens are associatied with a statistically significant reduction on PC risk66,67 In contrast levels of total IgE and specific IgE to respiratory or food allergens bears no relationship with risk of PC68. In addition preliminary findings have demonstrated that polymorphisms in atopy related genes may be associated with PC risk69 There is no established biologic mechanism whereby allergies reduce risk of developing PC. It has been proposed that patients with allergies have primed immune systems with enhanced tumour surveillance for tumour- specific antigens but this while plausible remains speculative at this time64,70,71. Overall, these studies support a growing body of evidence that suggests certain allergies are associated with reduced PC risk.

Diet There is no strong dietary link with PC in part due to methodological problems in acquiring and assessing dietary information, such as inability to do randomized controlled studies thereby introducing recall bias, difficulty isolating individual components and in the case of vitamin D determining the actual total intake given it is derived from both diet and sunlight exposure. Dietary components which have been associated with a protective role in the development of PC include a “Mediterranean” diet,72 anti-oxidants especially flavonoids, and folic acid particularly in active smokers,73,74 75-77 vitamin D supplementation and status,78,79 methionine and vitamin B6.80

Periodontal disease Poor dental hygiene as reflected by tooth loss or periodontal disease has been associated with an increased PC risk in cohort studies and appears to be independent of confounders such as smoking81. While the aetiology of this link remains to be established the composition of the salivary microbiome may play a role82.

21 1.1.2 Genetic risk factors The first evidence of an inherited predisposition to PC came from several case reports describing familial aggregation in the 1970’s and 80’s83-87. Subsequent case-control and cohort studies supported this evidence and demonstrated that approximately 5-10% of patients diagnosed with PC have a family history of the disease88-91. Studies requiring histological confirmation have shown lower rates (1.9-2.7%) of familial aggregation92,93. Individuals with a close family member diagnosed with PC are at increased risk of developing PC with a relative risk of 1.5-1.794-96. Further refinement of the actual risk of PC development in relatives, received significant assistance from the establishment of prospective familial PC registries such as the National Familial Pancreatic Tumour Registry (NFPTR). It is now evident that the risk of close relatives developing PC increases with increasing number of affected first-degree relatives (FDR) from 2-fold with one affected FDR to 6-fold and 14-32-fold (up to 57-fold) with 2 and 3 affected FDRs respectively97,98. Inherited predisposition to PC manifests in 3 different settings 99: 1. hereditary tumour predisposition syndromes including Hereditary Breast Ovarian Cancer (HBOC), Peutz-Jegher Syndrome (PJS), Familial Atypical Multiple Mole Melanoma (FAMMM), Li Fraumeni, Hereditary Non-Polyposis Colo-rectal Cancer (HNPCC) and Familial Adenomatous Polyposis (FAP) which account for 15-20% of the burden of inherited disease 100 2. Hereditary pancreatitis due to mutations in PRSS1 and 3. Familial PC (FPC). FPC is defined as a family with at least 2 first-degree relatives with PC, which do not fulfill the diagnostic criteria for an inherited tumour syndrome101. The majority (80%) of hereditary PC is attributed to FPC with a pattern consistent with autosomal dominant inheritance in 50-80% of families102,103. Studies to date have delineated the underlying genetic basis in at best 25% of these families with mutations in BRCA2, PALB2 and ATM accounting for 6 -19%, 104-106 4.2% 107 and 3.6% 108 respectively109. A previous segregation analysis supports the existence of an unidentified, autosomal dominant gene(s) with reduced penetrance that confers susceptibility to FPC110. FPC is likely to be a heterogeneous syndrome with phenotype determined by the underlying genetic mechanism and modified by environmental risk factors. Not all families are likely to have inherited high penetrant cancer predisposing mutations, as familial clustering can occur by

22 phenocopying as a result of shared or common environmental exposures within families111.

Relative Risk (95% Estimated lifetime PC Clinical risk group Syndrome CI) risk (70 – 80 years) General Population NA 1 0.96 1 FDR PC NA 4.6 (0.5 - 16.4) 2 FDR PC Familial PC 6.4 (1.8 - 16.4)

≥3 FDR PC Familial PC 32 (10.2 - 74.7)

Genetic risk group

BRCA2 HBOC/FPC 3.51 (1.87-6.58) 3.36% Elevated but not Elevated but not PALB2 FPC defined defined BRCA1 HBOC 2.26 (1.26-4.06) 2.16% MSH2, MLH1, MSH6, HNPCC 8.6 (4.7-15.7) 3.68% (1.45%-5.88%) PMS2 Hereditary 30-40% in smokers, PRSS1 58 (23-105) pancreatitis 20% in non-smokers Peutz-Jeghers STK11 132 (44 - 261) 11%-32% syndrome

CDKN2A FAMMM 38 (10-97) 17%

Ataxia Elevated but not Elevated but not ATM (monoallelic) Telangiectasia (bi- defined defined allelic) Li Fraumeni Elevated but not Elevated but not TP53 syndrome defined defined Table 2: Genetic risk factors for PC – hereditary cancer syndrome and moderate to high penetrance genes

1.1.2.1 Hereditary Tumour Predisposition Syndromes Hereditary Breast-Ovarian Cancer Inherited pathogenic germline BRCA2 mutations place carriers at increased risk of cancers of the pancreas, prostate, gallbladder, bile duct, stomach and melanoma in addition to breast and ovarian cancer112-114. The prevalence of germline BRCA2 mutations in patients with PC depends on the ethnic ancestry of the population studied and is higher in groups with founder mutations such as

23 those of Ashkenazi Jewish descent. In an early report, Goggins et al found BRCA2 mutations in 7% of patients with apparent sporadic PC (3 of 41) of which one was the Ashkenazi founder mutation115. Studies have shown BRCA2 mutations in 5 – 10% of Ashkenazi Jews with PC116,117. In familial PC the mutation prevalence increase with rising number of affected relatives: 6-12% in families with two or more with PC and 16% from families in which 3 or more have PC118-120. The relative risk of developing PC in BRCA2 mutation carriers is approximately 3.5 - 6121-123. A substantial proportion of mutation-positive PC patients report neither a history of PC nor breast-ovarian cancer115,118. This is likely due to reduced penetrance for PC rather than there being PC specific genotype-phenotype correlation for BRCA2 mutations as has been seen in some breast-ovarian cancers124.

In contrast the role of BRCA1 mutations and predisposition to PC is less well studied. Initial studies in BRCA1 mutation positive families with young-onset breast or ovarian cancer suggested a 2.26-fold (95% CI = 1.26–4.06) increased risk of PC125-127. A high prevalence of the BRCA1 founder mutation was found in a series of resected PC in Ashkenazi patients.128 Other studies have reported no increase in the prevalence of BRCA1 mutations in patients with PC117,129. Furthermore BRCA1 mutations are uncommon in families with PC without a history of breast cancer130 and could represent incidental findings131. Other studies in BRCA1 mutation positive families with PC, the member who developed PC was a confirmed or obligate carrier132.

Familial Atypical Multiple Mole Melanoma Familial atypical multiple mole melanoma (FAMMM) is a syndrome characterized by predisposition to melanoma and PC. FAMMM is caused by germline mutations in CDKN2A (p16), which encodes the tumour suppressors ARF and INK4A. Pathogenic CDKN2A mutations have been identified in 21% of Australian melanoma kindreds133. Individuals with FAMMM have a 38-fold increased risk of developing PC compared to the general population, contributing to a lifetime risk of 17% by age 75134-136.

Peutz-Jegher Syndrome

24 Peutz-Jegher syndrome is an autosomal dominant disorder characterized by gastrointestinal tract hamartomatous polyps and mucocutaneous pigmentation.137 In 80 – 94% of individuals who meet the clinical criteria, pathogenic mutations in STK11 are identified138. These individuals have a 132-fold (95% CI = 44–261) increased risk of PC compared with the general population, and the lifetime risk of PC in these individuals has been estimated to be 11–32%139-141. It has been proposed that PJS patients who develop PC progress through IPMN given the frequent loss of STK11 protein expression in these lesions142.

Hereditary Non-Polyposis Colo-rectal Cancer Hereditary Non-Polyposis Colo-rectal Cancer (HNPCC or Lynch syndrome) is the result of germline mutations in the DNA mismatch repair genes MSH2, MLH1, MSH6 and PMS2. Patients are at increased lifetime risk for a wide range of tumour types but the predominant malignancies are colonic and endometrial cancer. The other associated tumour types are lower risk with < 5% lifetime risk and include PC, gastric, small bowel, ureteric and skin tumours. A recent study of 147 families containing a mutation in a mismatch gene reported a 8.6-fold (95% CI, 4.7–15.7) increased risk of PC compared with the general population.143 This corresponds to a 3.68% (95% CI, 1.45–5.88%) lifetime (by age 70) risk of PC143. A characteristic feature of Lynch-related tumours is microsatellite instability (MSI) with loss of the wild-type allele. The contribution of germline mismatch repair gene mutations and the prevalence of MSI is not well defined in part due to inconsistent definition and methodology for detecting MSI. Histologically MSI PC shows a medullary histology with syncytial growth pattern and indistinct pushing border but perhaps improved prognosis144,145. As shown in chapter 4 the prevalence of MSI in our predominantly sporadic PC cohort was 1.3%, which predominantly resulted from somatic inactivation of MMR genes. One pathogenic germline mutation was identified in PMS2 but the tumour did not display MSI, suggesting that in this cohort Lynch syndrome does not contribute to the development of PC.

Familial Adenomatous Polyposis

25 Familial adenomatous polyposis (FAP) is an autosomal dominant disorder characterized by the early development of hundreds to thousands of colonic adenomatous polyps. The natural history is that untreated nearly all affected patients will develop colorectal carcinoma by age 40146. In more than 70% of patients who meet the clinical criteria a germline mutation in the APC can be identified146-148. Patients with FAP are at increased risk for other neoplasms, including thyroid tumors, gastric, duodenal, and ampullary adenocarcinoma. The relative risk for the development of PC is 4.46 (95% CI 1.2 – 11.4) and again there is evidence suggesting precursor lesions progress through the IPMN pathway149-151.

Li-Fraumeni Syndrome Li Fraumeni syndrome (LFS) is an autosomal dominant highly penetrant cancer predisposition syndrome characterized by a variety of early onset tumors. The syndrome, described in 1969 by Li and Fraumeni based on a retrospective analysis of families with childhood rhabdomyosarcoma,152 was characterized by the presence of five cancers: sarcoma, adrenocortical carcinoma, breast cancer, leukemia, and brain tumors153,154. Several different clinical classification systems exist but these tumour types form the core clinical features. Li Fraumeni Syndrome is caused by germline mutations in the TP53 gene and is inherited in an autosomal dominant pattern. The risk of PC is increased but has not been quantified155,156. The prevalence of germline TP53 mutations and the contribution to sporadic PC has not been established.

1.1.2.2 Hereditary Pancreatitis Hereditary pancreatitis is a rare autosomal dominant form of inherited pancreatitis. This typically manifests as recurrent acute pancreatitis by age 10, chronic pancreatitis by age 20 and increased risk of PC after age 40157. In families meeting the clinical criteria mutations in the cationic trypsinogen gene (PRSS1) are found in around 80%158. Patients with hereditary pancreatitis have a 58-fold (95% CI 23–105) increased risk of developing PC and a lifetime risk (by age 70) of 30–40%159. Cigarette smoking increases the risk by 2-fold and brings the age at diagnosis forward 20 years160. The lifetime risk in non- smokers is estimated to be < 20%161.

1.1.2.3 Familial Pancreatic Cancer In keeping with the current study, a family history of PC is reported by approximately 5–10% of PC patients in the absence of a known predisposing variant162. The term familial pancreatic cancer (FPC) is defined as a kindred with at least two first-degree relatives with PC, which do not otherwise fulfill the diagnostic criteria for an inherited cancer syndrome101. In these families the relative risk of developing PC increases with increasing number of affected first- degree relatives: one FDR 4.6 (CI 0.5 – 6.4), two FDR 6.4 (CI 1.8 – 16.4), three FDR 32.0 (CI 10.2 – 74.7)97. The risk is higher in smokers, furthermore spouses show a two-fold increased risk which does not reach statistical significance, but suggests a role for shared environmental factors163,164. FPC kindreds also appear to be at increased risk of developing breast, ovarian and colorectal cancer particularly if the proband developed young onset (<50 years) PC164,165. This is in keeping with our current knowledge of PC predisposition genes and the inherited cancer syndromes where carriers are at risk of malignancy in multiple organs. Patients with a pathogenic germline BRCA2 mutation are at increase risk of cancers of the pancreas, prostate, gallbladder, bile duct, stomach and melanoma in addition to breast and ovarian cancer112-114. Also patients with PJS are at increased risk of cancers of the gastrointestinal tract (oesophagus, stomach, small bowel, colon and pancreas), lung, uterus, ovaries and breast140,166,167. As shown in chapter 4 in our cohort FPC patients (42.6%) were twice as likely as SPC patients (21.2%) to have at least one FDR with an extra-pancreatic malignancy. The most common malignancies in family members were breast, colorectal, melanoma, lung and prostate. FPC kindreds appeared to be at higher risk of breast cancer and melanoma in particular. This could be a result of chance given the high incidence in the Australian population or a result of an underlying inherited predisposition or both. This finding is consistent with previous reports where in 40% of FPC families PC was the sole tumour entity and in the remaining 60% other tumour types namely breast, colon and lung were seen168. In another study relatives of FPC probands were at increased risk of breast, ovarian and bile duct malignancies but this was not increased compared to SPC kindreds169. In SPC there is also increased risk of malignancy in close relatives, in addition to PC studies have found associations 27 with cancers of the colon, small intestine, breast, lung, testis and cervix 170 and lymphoma and ovarian cancer in ever smokers171. While others have reported that only mortality from PC is increased in relatives91,172. Other studies have found a family history of ovarian, colorectal, breast or prostate cancer is associated with increased risk of PC but this is not a consistent finding across studies90,173-177. Defining the precise organotypic distribution of tumours which cluster with PC is important because it (a) supports an underlying genetic predisposition or common environmental factor potentially even in the absence of multiple PC cases in the family, (b) allows a more precise definition and clinical recognition of the syndrome and (c) facilitates broader and more precise risk assessment and employment of risk reduction strategies in at risk family members169. These results highlight the importance of complete family history of all cancer types in clinical assessment of FPC pedigrees171. The underlying genetic basis of PC predisposition has been identified in less than 25% of such families, 105,107,108 despite 50-80% of families demonstrating an autosomal dominant inheritance pattern102,103. Overall, 0.6% of the general population is estimated to carry a mutation in a moderate-to high-risk PC predisposition gene with an attendant lifetime risk of developing PC (by the age of 85) of 32%110.

BRCA2 As discussed on pages 20-21, pathogenic BRCA2 variants predispose to PC and are the best characterised and most commonly described variants in FPC.

PALB2 PALB2 (Partner And Localizer of BRCA2) binds to BRCA2 and stabilizes it in the nucleus where they participate in the Fanconi anaemia/homologous recombination pathway a critical pathway for double-stranded DNA repair. Mutations in PALB2 have been attributed to cause approximately 1% of non- BRCA familial breast cancers and have been identified as PC susceptibility genes in FPC probands178. PALB2 was identified as PC susceptibility gene by Sanger exome sequencing of matched tumour: germline samples from 68 predominantly sporadic PC patients179. One patient who met criteria for FPC (15

28 461 germline variants) had germline truncating mutations with somatic LOH in 3 tumour suppressor genes (SERPINB12, RAGE, PALB2). PALB2 given its relationship with BRCA2 was deemed the most likely candidate and validation in 96 additional FPC probands found 3 more truncating mutations. Subsequent studies have confirmed that PALB2 mutations are found in 1-3% of familial PC probands particularly those families with an additional case(s) of breast cancer179-182.

ATM Loss-of-function ATM mutations have recently been implicated in predisposition to developing PC. In 2 FPC families truncating germline ATM mutations were found in a proband which seemed to segregate with disease. Further analysis in a further 160 FPC probands and 190 controls identified another 4 loss-of- function ATM variants in FPC cohort (P=0.046)183. The risk of developing PC due to pathogenic germline ATM mutations and their contribution to sporadic disease has not been defined.

Palladin In a large family with multiple affected members with apparent autosomal dominant inheritance and utilizing 373 microsatellite markers linkage was found with chromosome 4q32-34184. Further characterization of this region using microarrays and direct sequencing showed a missense mutation (p.P239S) in the palladin (PALLD) gene, which segregated with disease185. PALLD is a cytoskeletal protein required for the organization of the normal actin cytoskeleton with a role in establishing cell morphology, motility, cell adhesion and cell-extracellular matrix interactions in a variety of cell types. The mutant PALLD transcript is overexpressed in non-neoplastic stromal cells where it facilitates tumour invasion and metastasis186-188. Further studies failed to replicate this finding in other familial PC probands179,189.

1.1.2.4 Low Penetrance Susceptibility Variants In contrast to candidate gene studies a genome-wide association study (GWAS) is unbiased and can examine the association of several thousand to millions of

29 single nucleotide polymorphism (SNPs) in those with and without the trait. The identified SNP(s) mark a region of the genome, which is associated with the disease. Typically the SNP itself is not causal but in linkage disequilibrium with the variant(s) responsible for disease predisposition. These variants, which are relatively common as they are not subjected to the degree of purifying selection as high-penetrance variants and have low penetrance and relative risk. Additional functional and fine-mapping studies are necessary to identify the causal variants. To date 7 PC GWAS have been published which are summarized in Table 3.

30 Study Study Population Region Chromosome_ Reported Risk Allele Odds ratio 95% CI P Value reference position Gene(s) Cases Controls EUR EAS AFR EUR EAS AFR Wolpin BM et 7683 0 0 14397 0 0 5p15.33 1294086 TERT, MIR4457, rs2736098-T 0.8 0.76-0.85 1.00E-15 al Nature CLPTM1L Genetics 2014 PanScan III 7q32.3 130680521 LINC_PINT rs6971499-C 0.79 0.74-0.84 3.00E-12 16q23.1 75263661 BCAR1, CTRB1, rs7190458-A 1.46 1.30-1.65 1.00E-10 CTRB2 13q12.2 28493997 PDX1 rs9581943-A 1.15 1.10-1.20 2.00E-09 22q12.1 29300306 ZNRF3 rs16986825-T 1.18 1.12-1.25 1.00E-08 Wu C et al Gut 642 363 0 ? ? ? 11p15.4 9951515 SBF2 rs10500715-T 1.32 1.19-1.47 2.00E-07 2012 18p11.21 13366863 c18orf1 rs12456874- 1.38 1.22-1.57 6.00E-07 G 3p26.2 3314490 Intergenic rs9874556-A 1.31 1.17-1.46 4.00E-06 9q33.1 115288135 Intergenic rs7853844-A 1.43 1.23-1.65 2.00E-06 10q23.1 85980996 GRID1 rs10788473-T 1.27 1.15-1.41 3.00E-06 11p15.1 18363391 GTF2H1 rs9783347-A 1.36 1.19-1.56 9.00E-06 13q21.31 63567780 Intergenic rs1000589-T 1.27 1.15-1.39 3.00E-06 Willis JA et al 105 14 21 ? ? ? NR NR NR NR NR NS Clin Cancer Res 2012 Wu C et al 0 3584 0 0 4868 0 21q21.3 29345416 BACH1 rs372883-T 1.27 1.19-1.33 2.00E-13 Nature Genetics 2011 5p13.1 39394887 DAB2 rs2255280-T 1.23 1.15-1.32 4.00E-10 10q26.11 118519432 PRLHR rs12413624-T 1.23 1.16-1.31 5.00E-11 21q22.3 42358786 TFF1 rs1547374-A 1.27 1.19-1.35 4.00E-13 13q22.1 73334709 NR rs9573163-G 1.26 1.18-1.34 5.00E-13

31 22q13.32 48533757 FAM19A5 rs5768709-G 1.25 1.17-1.34 1.00E-10 1q43 238745053 MIPEPP2 rs2689154-G 1.20 1.11-1.29 6.00E-06 6q12 68432116 BAI3 rs9363918-A 1.27 1.15-1.39 1.00E-06 9p24.2 4426631 SLC1A1 rs10974531- 1.24 1.13-1.35 5.00E-06 A 6q25.3 155876368 ARID1B rs4269383-C 1.20 1.12-1.30 7.00E-07 3q29 196024759 TFRC rs4927850-A 1.24 1.14-1.35 2.00E-07 Low SK et al 0 991 0 0 5209 0 2q37.2 234706553 ARL4C rs6736997-A 1.57 1.29-1.91 6.00E-06 PLOS One 2010 7q36.2 153928758 DPP6 rs6464375-A 3.73 2.24-6.21 4.00E-07 8q24.13 123753462 FAM91A1 rs10088262- 1.40 1.21-1.61 4.00E-06 A 6p25.3 1339954 FOXQ1 rs9502893-G 1.29 1.17-1.43 3.00E-07 2q12.1 104762499 LOC284998 rs12615966- 3.15 1.91-5.21 7.00E-06 A 13q21.32 65907683 LOC387933 rs1585440-C 1.30 1.16-1.45 9.00E-06 13q22.1 73357977 LOC730242 rs1886449-A 1.51 1.26-1.80 9.00E-06 5p15.33 2109787 LOC731559 rs6879627-G 1.25 1.14-1.39 8.00E-06 17q11.2 32550640 MYO1D rs225190-G 1.26 1.14-1.39 6.00E-06 13q31.1 79725587 NDFIP2 rs2039553-A 1.73 1.36-2.19 7.00E-06 6q26 161815043 PARK2 rs3016539-A 1.50 1.26-1.79 7.00E-06 17p11.2 18850557 PRPSAP2 rs4924935-G 1.37 1.19-1.58 8.00E-06 2q11.2 101305708 RNF149 rs6711606-A 2.81 1.81-4.37 4.00E-06 17q22 58370936 RNF43 rs2257205-A 1.38 1.20-1.59 8.00E-06 8p11.22 38611785 RNF5P1 rs7832232-A 1.45 1.24-1.71 5.00E-06 2q22.1 136797654 THSD7B rs1427593-A 1.49 1.25-1.77 7.00E-06 12p11.21 32283475 BICD1 rs708224-A 1.32 1.19-1.47 3.00E-07 Diergaarde B 160 0 0 172 0 0 NR NR NR NR NR NS et al

32 Pancreatology 2010 Petersen GM 3851a 3934 13q22.1 73342491 KLF5, KLF12 rs9543325-C 1.26 1.18-1.35 3.00E-11 et al Nature Genetics 2010 PanScan II 1q32.1 200038304 NR5A2 rs3790844-T 1.30 1.19-1.41 2.00E-10 5p15.33 1321972 CLPTM1L, TERT rs401681-T 1.19 1.11-1.27 7.00E-07 Amundadottir 3891 0 0 3932 0 0 9q34.2 133273813 ABO rs505922-C 1.20 1.12-1.28 5.00E-08 L et al Nature Genetics 2009 PanScan I Table 3: Common low-penetrance genetic factors associated with increased risk of PC from GWAS. Adapted from the NHGRI-EBI GWAS catalog (www.ebi.ac.uk/gwas)190 a mixed ancestry, AFR African, EAS East Asian, EUR European

33 1.2 Precursor Lesions and Progression to PC PC develops from solid and cystic precursor neoplasms through the serial acquisition of mutations, which provide a selective advantage to the cells. The evolution of PC progresses through several stages from non-invasive precursor lesions such as pancreatic intraepithelial neoplasia (PanIN) or cystic neoplasms in particular mucin producing (intraductal papillary mucinous neoplasm (IPMN) and mucinous cystadenoma)191-193. Based on the genetic evolution of PC from sequencing of both primary and metastatic tumour it is estimated that it takes at least 10 years from time of the initiating mutation to the establishment of the founder non-metastatic cancer cell and a further 5 years for the development of metastatic potential194.

1.2.1 Pancreatic Intra-epithelial Neoplasia Pancreatic intra-epithelial neoplasia (PanIN) was first described more than a century ago and their role in PC is supported by their increasing frequency with age and in pancreata with an invasive cancer. Furthermore they frequently harbor the mutations seen in invasive cancer and they acquire these mutations in a step-wise fashion with increasing histologic features of dysplasia. PanINs are graded histologically as PanIN-1 (low-grade), PanIN-2 (intermediate-grade), or PanIN-3 (high-grade) based on the degree of architectural distortion progressing from flat lesions to papillary and increasing cellular atypia.195 Paralleling this histologic progression is a genetic progression. The lower grade lesions (PanIN-1 and PanIN-2) often harbor genetic alterations in the KRAS and CDKN2A, whereas the higher-grade PanIN-3 lesions and invasive adenocarcinomas, in addition have mutations in TP53, SMAD4 and BRCA2196- 201. Figure 1. PanIN1A and PanIn1B are often incidental findings in the presumed normal pancreata of adults particularly beyond 50 years of age202,203. In contrast, higher-grade PanINs are rare in the “normal” ageing pancreas unless there is an associated invasive PC or the patient has a strong family history of PC204,205. In adult pancreata without invasive carcinoma PanIN-1 lesions are seen in up to 40% of adult whereas PanIN-3 lesions are present in <5%195. Incorporation of these findings provides strong support for the

34 progression from normal ductal epithelium to low grade PanIN, high-grade PanIN, in situ carcinoma to invasive carcinoma196.

PanINs arise in the smaller ducts and the vast majority measure less than 5mm, as such they are difficult to detect with current imaging techniques206. However, it has been noted that PanIN even low-grade is frequently accompanied by parenchymal changes termed lobulocentric atrophy,207-209 which may be detected by sensitive imaging techniques notably endoscopic ultrasound (EUS)210. These parenchymal changes are characterized by a combination of acinar cell loss, proliferation of small ducts and fibrosis. When these changes are multifocal as is often the case in FPC kindreds the EUS appearance is similar to that of chronic pancreatitis210. The original postulate linking PanIN and LCA was that the PanIN lesion produced duct obstruction, injury and inflammation producing localised chronic pancreatitis. However findings in genetically engineered mouse models (GEMM) of PC suggest that in the presence of mutant KRAS the acinar cells undergo acinar-ductal metaplasia (ADM) which is accelerated by pancreatitis211. This suggests that ADM may responsible for the LCA and that PanIN is one of the possible results of this process212. This finding is supported by prominent acinar-ductal metaplasia (ADM) in the acini of lobulocentric atrophy207,213. Other knowledge gaps exist which also limit the usefulness and clinical applicability of identifying LCA +/- PanIN. The concordance between PanIN and LCA is incomplete, with LCA commonly identified in ageing pancreata with or without associated PanIN. The converse is also true and PanIN can occur in the absence of LCA and when present is typically focal making sampling errors likely if a fine needle aspirate is performed. In addition, there is no current evidence that the presence of LCA correlates with the higher grade PanIN, therefore its usefulness as a marker of a high grade PanIN is limited212. Lastly, the proportion and time-course over which progression to invasive cancer takes place remains unanswered. One estimate is 1% probability of a single PanIN lesion progressing to invasive cancer206,214.

35 1.2.2 Cystic Neoplasms Pancreatic cysts are common and their prevalence increases with age up to 10% in persons over 70 years of age215. Pancreatic cysts can be anxiety provoking to both medical professionals and patients resulting in multiple investigations due to the fear of harboring or developing in to PC. The clinical decision of which cysts need further assessment, follow-up and treatment, which is currently operative resection, remains a challenge.

Intraductal Papillary Mucinous Neoplasm Intraductal Papillary Mucinous Neoplasms (IPMN) are neoplastic proliferations within the main or branch pancreatic ducts and are defined by a variable degree of papillary architecture and mucin production. They characteristically produce obstruction and upstream dilatation of the duct system, which is detectable on imaging studies195. They are classified according to which duct is involved and categorized as main-duct, branch-duct or mixed-duct type and the epithelial subtype. Four epithelial subtypes are recognized on the basis of histologic appearance and immunohistochemical mucin (MUC 1, 2 and 5AC) staining profile: intestinal, gastric, oncocytic and pancreaticobiliary216. IPMNs in particular the main-duct type can progress to invasive carcinoma with around 50% being colloid (mucinous non-cystic), most of the remainder conventional tubular adenocarcinoma and rarely oncocytic195,217. The pancreatobiliary subtype shows a strong correlation with progression to invasive cancer and poorer survival216. The risk of cancer in IPMN patients ranges from 24% in IPMN involving the branch ducts to over 60% in lesions involving the main pancreatic duct218.

As is the case for PanINs there is a growing body of clinical and molecular evidence that support progression through low and high-grade dysplasia to invasive carcinoma. The mutation profile has recently been characterized using cyst fluid or tissue and shown that GNAS codon 201 mutations are relatively specific for IPMN can be identified in 40-66% while codon 12 KRAS mutations were seen in 47 - 74% with either mutation seen in 96% of IPMNs219,220. Inactivating mutations in RNF43 are seen in 75% of cases and mutations in CDKN2A, TP53 and SMAD4 seen in higher grade lesions221,222. Small

36 pancreatic cysts are difficult to characterize based on imaging criteria and attempts at fluid aspiration frequently yield inadequate specimens. In the absence of an alternative diagnosis these cysts are often labeled likely branch duct IPMN and serially monitored for changes. In a series of 21 small pancreatic cysts <1cm diameter with papillae and mucin termed incipient IPMNs, KRAS codon 12 mutations were seen in all 21 and GNAS codon 201 mutations seen in 7 (33%)223. Therefore it seems likely that a reasonable proportion of small pancreatic cysts, which are difficult to categorize on imaging, do represent IPMNs.

Mucinous Cystic Neoplasm (MCN) MCNs are mucin-producing neoplasms which are typically found in females (>95%) in the tail of the pancreas (>95%) and do not demonstrate communication with the pancreatic duct215. A proportion of MCN progress over years from low-grade to high grade dysplasia and invasive carcinoma with accumulation of genetic mutations221. MCN are characterized by the presence of ovarian stroma. The overall malignant potential of MCNs is low with <15% harbouring high-grade dysplasia or invasive carcinoma. These can be stratified clinically using imaging criteria of size > 4cm and the presence of mural nodules. In two studies no malignancy was seen in MCNs <4 cm without mural nodules224,225. The mutation profile of MCN includes KRAS mutations in 75%, inactivating RNF43 mutations in 40% and again CDKN2A, TP53 and SMAD4 with increasing dysplasia222,226.

Serous Cystic Neoplasm Serous cystic neoplasms (SCN) are female predominant (75%) but found throughout the pancreas and are composed of uniform glycogen-rich cuboidal cells which typically form multiple small cysts but an oligo-cystic variant is seen in 10%215. SCN are benign neoplasms with low malignant potential (<1%)227. The characteristic mutation profile for SCNs is loss of VHL which is seen in 50%222. This is consistent with patients with von Hippel Lindau syndrome frequently having multiple oligocystic SCNs228.

37 Solid-Pseudopapillary Neoplasm Solid pseudopapillary neoplasms (SPN) are an uncommon pancreatic neoplasm. They are composed of solid and cystic components and found throughout the pancreas in young females (median age 32-38 years) predominantly (>80%). Most SPNs are benign or have an indolent course229-231. SPNs typically (>90%) show activating mutations in beta-catenin (CTNNB1) and overexpression of its target genes232.

1.3 Predicting the risk of developing PC The primary goal of developing PC risk prediction models is to be able to personalize PC risk and in doing so inform genetic testing and potential screening options163. Multiple risk factors for PC have been identified, after increasing age the next major risk factor is a family history of the disease175,233. In those with a known mutation efforts have been made to quantitate this risk, but the majority of individuals at increased genetic risk do not have a known mutation and in effect each person presents with a unique risk factor profile. PancPro is a Bayesian model developed from pedigree data from the National Familial Pancreatic Tumour Registry (NFPTR) and calculates the risk that a person carries a high-penetrance PC gene and their risk by age of developing cancer163. The input variables required from each at-risk individual include personal and family history of cancer, current age, age of onset of cancer along with other model covariates. The model has been validated in an independent cohort and shown an observed to predicted PC ratio of 0.83 (95% CI, 0.52 to 1.20)234. PancPRO may be a useful strategy to rank families based on their PC risk and suitability for a screening program235. Given that most people who develop PC do not have a family history of the disease or known predisposing germline mutation a risk model has been developed which predicts risk utilizing genetic (low-penetrance SNPs identified in GWAS studies) and non-genetic PC risk factors (age, DM, alcohol intake, BMI and family history of PC)236. The ability of this model to detect individuals at high risk is limited, and it estimates fewer than 3 in 1000 individuals to have a > 5% predicted lifetime risk of PC.

38 1.4 Early Detection of PC There is currently no robust screening test suitable for testing the general population. As illustrated in Figure 1 the current clinically available tests for diagnosis and monitoring of PC are better at detecting later stage disease and there exists a gap in detecting early disease particularly PanINs. Furthermore if a precursor lesion is detected there are no tests shown to discriminate biologically relevant from indolent precursors. A screening test with high sensitivity and specificity will have a low positive predictive value i.e. majority of positive results will be false-positive, if the disease is uncommon in the population being studied237. Therefore defining a study population at higher risk than the general population, may derive benefit from screening. There are currently two situations in which PC screening might be considered: ‐ Individuals with increased genetic risk based on a known inherited predisposition mutation or strong family history of PC ‐ Individuals with a known precursor lesion which are currently predominantly pancreatic cystic lesions due to the current status of the resolution of pancreatic imaging Although screening has intuitive appeal with the potential benefit of early diagnosis and as a consequence improved treatment and prognosis, several issues need to be resolved before it can be incorporated into standard practice. First it needs to be established that PC screening is able to detect early disease and second that this translates into better outcomes. PCs diagnosed in screening trials have predominantly but not universally been resectable. However as with sporadic disease resectable does not always equate with early, as resected patients often have progressed to metastatic disease due to subclinical metastatic disease at diagnosis210,238-241. Screening can be associated with 2 forms of bias in particular, which can lead to false conclusions of benefit. Lead-time bias where screening leads to earlier diagnosis but does not prolong survival i.e. patients live longer with the knowledge they have cancer but there is zero-time shift237. Length-bias where people appear to live longer due to screening detecting biologically less-aggressive disease and miss those with rapidly progressive disease242. Second, the lesions which should be the aim of detection need to be established i.e. which lesions constitute a positive result in PC screening programs. Most clinical trials are based on

39 pancreatic imaging which provides limited or no information on the biology of the lesion. Therapeutic intervention if undertaken for precursor lesions in current clinical practice constitutes a pancreatic resection. Pancreatectomy for a precursor lesion with a low probability of progression is associated with significant morbidity and unlikely to change the outcome for the patient. Typically the long-term survivors after pancreatectomy for PC, are those with early stage tumours (<2cm and confined to the pancreas) and lymph node negative cancers243-246. However even in this small group a high rate of nodal metastases and poor prognosis has been described247. Currently, early stage cancers along with the high-grade precursor lesions (IPMN, MCN and PanIN 3 and CIS), represent the best opportunity to reduce mortality from PC as they are likely to progress and are potentially curable. Improving our understanding of the inherited predisposition to PC will lead to more precise risk assessment and potentially better selection of candidates who will benefit from screening.

Early detection tools typically fall into two categories: imaging or biomarker based (blood, pancreatic juice or intestinal contents including stool).

Figure 1: Overview of the patient journey from at risk individual to developing asymptomatic precursor lesions and early localized cancer to advanced symptomatic PC with elevated CA19.9248. The bottom panels show the opportunities for the detection of disease and potential interventions at each stage. The timing of diagnosis by the various modalities is illustrative, but depicts that most current investigations are better at detecting later stage disease but this may depend on the type of precursor lesion. PanIN = pancreatic intraepithelial neoplasia, mucin cyst = IPMN or mucinous cystadenoma, LGD = low-grade dysplasia, MGD = moderate-grade dysplasia, HGD = high-grade dysplasia

1.4.1 Imaging approach In recent years multiple PC surveillance programs have been established and initial findings reported. Table 4. The primary modalities used include endoscopic ultrasound (EUS), magnetic resonance imaging with/without magnetic resonance cholangiopancreatography (MRI/MRCP) and computerised tomography (CT). The sensitivity of these modalities to detect cystic pancreatic lesions is 93% with EUS, 81% with MRCP and 27% by CT249. Findings from the PC screening studies performed to date are difficult to consolidate due differences in the: (a) population studied and therefore absolute risk of PC, (b) imaging modalities used and (c) endpoints of the study. Overall, the studies demonstrate that precursor lesions or invasive cancers can be demonstrated in a variable but significant proportion of at-risk individuals but no study has shown better outcomes for patients238-241,249-254. The variable yield (1-50%) is partly dependent on the definition of the target lesions, which range from early cancer and high-grade dysplastic precursor lesions to IPMN with low-intermediate dysplasia to PanIN with any grade of dysplasia. The type of precursor lesion may have different malignant potential depending on the study population, with cystic lesions uncommon in CDKN2A mutant carriers but displaying higher malignant potential, whereas in FPC patients cystic lesions particularly IPMN are common255. Finding multiple cystic areas consistent with branch-duct IPMNs may be a marker of coexistent high- grade PanINs256. The prevalence of detectable neoplasia is also dependent on the risk in the population being studied, the modalities used, the duration of follow-up and the number that undergo definitive pathological assessment i.e. surgical resection. Screening brings with it the risk of overtreatment and additional controlled trials are needed to determine the risks, benefits and optimal approaches to PC screening. Prophylactic removal a precursor lesion in an individual at increased risk could prevent PC and improve outcome but this has not been established. There is no current consensus as to the inclusion criteria for a surveillance program, although guidelines suggest those with a 5-10 fold increased risk should be considered257,258. In addition there is no consensus for the age at

42 which surveillance should begin or the absolute indications for surgery. Most would agree that agree that a solid mass or cyst meeting the guidelines should be resected but patients frequently have widespread abnormalities on EUS complicating this decision. The Cancer of the Pancreas screening studies (CAPS) are the best-known PC screening trials of which 3 have been reported to date. The primary aim of these studies was to identify the utility of screening and the best and most sensitive method for detection of precancerous lesions in patients considered to be at high genetic risk. CAPS1 and CAPS2 utilised initial screening with EUS and CT to find pancreatic neoplasia in 5.3% and 10% respectively259,260. This includes only 2 PC IN 116 patients with the remainder of neoplastic lesions being predominantly IPMN. CAPS3 found a much higher rate of pancreatic neoplasms with 85 of 216 (39.4%) patients displaying pancreatic neoplasm (82 IPMN and 3 pancreatic neuroendocrine tumours)261. The detection of pancreatic neoplasms by EUS and MRI was superior to that of CT.

43 Patients Follow-up Study Risk category (n) (months) Imaging Modality Findings PanIN Yield PC IPMN 2 - 3 Other Brentnall et al Ann Int Med CT, EUS, ERCP +/- KRAS 1999 FPC 14 15 analysis in pancreatic juice 50% - - 7a - Kimmey et al Gastrointest Endosc 2002 FPC 46 EUS, ERCP 26% - - 12a - Canto et al Clin Gastro Hepatol 2004 FPC, PJS 38 22 EUS 5% 1 1 - - Canto et al Clin Gastro Hepatol 2006 FPC, PJS 78 12 EUS, CT 10% 1 6 1 - Poley et al Am J Gastro FPC, PJS, Other syndromes with ≥2 affected 2009 (BRCA, p16, p53, HP) 44 Initial finding EUS 23% 3 7 - - Langer et al Gut 2009 FPC, BRCA2 76 EUS, MR/MRCP 0.76% - 1 - - Verna et al Clin Cancer Res 2010 FPC, BRCA2, CDKN2A 51 Initial finding EUS, MRI 12% 2 4 - 4 EPMc Ludwig et al Am J Gastro 2011 FPC, BRCA 109 MRCP followed by EUS 8.30% 1 5 2 - Vasen et al Gastroenterology 2011 CDKN2A 79 48 MRI/MRCP 20% 7 9 - - Al-Sukhni et al J Gastrointest Surg 2011 FPC, BRCA, PJS, CDKN2A, HP 262 50 MRI/MRCP 7.30% 3 15 - 1 PNET Schneider et al Familial Cancer 2011 FPC, BRCA, PALB2 72 44 EUS, MR/MRCP 13% 1 7 2 - Canto et al Gastroenterology 2012 FPC, BRCA2, PJS 216 29 EUS, CT, MRI 43% 0 82 5b 3 PNET Table 4: Summary of PC screening trials using a predominantly imaging based approach in high-risk individuals. a widespread dysplasia, b 5 pancreatic resection and all had multi-focal IPMN and PanIN, c EPM extra-pancreatic malignancies - 2 ovarian cancers (in BRCA-1/2 mutation carriers), 1 retroperitoneal carcinoid, 1 papillary thyroid cancer

44 1.4.2 Biomarker approach In current clinical practice biomarkers have a limited role in the diagnosis and management of PC with carbohydrate antigen 19.9 (CA19.9) the only marker clinically utilized. More than 600 molecules have been shown to be overexpressed at the protein or RNA level in PC and represent potential diagnostic biomarkers262. This includes over 150, which are overexpressed and secreted and detectable in body fluids. Many of these have been studied in advanced PC but their significance and behavior is likely to be different in early PC. As PC spreads outside the pancreas abnormalities that are not produced by or specific to PC cells accumulate such as inflammatory markers248. These represent epiphenomena and are unlikely to provide prognostic or predictive value. Furthermore many markers have been identified using retrospective or archival samples and have not been validated prospectively. When they have some have been shown to not perform as well particularly in early disease263- 265. In view of recent large scale sequencing studies of PC, which highlight the significant heterogeneity of tumours, it brings into question whether it is possible to identify a “one-size fits all” biomarker of early PC. The following text will discuss the use and limitations of the only current clinically used biomarker CA19.9 and novel biomarkers in blood and/or pancreatic juice or cyst fluid for the diagnosis of early disease.

Blood based biomarkers Proteins Carbohydrate Antigen 19.9 (CA19.9) Carbohydrate antigen 19.9 (CA19.9) is the only clinically utilized biomarker in the management of patients with PC. CA19.9 was first isolated from a human colorectal cancer cell line in 1979. It was defined by a monoclonal antibody (1116 NS 19.9) produced by a hybridoma from mouse spleen immunized with a human CRC cell line 266,267. Since the development of a radio-immunometric assay in 1983268, CA19.9 has shown use as a diagnostic adjunct in PC with a meta-analysis of 22 studies showing a sensitivity of 79% and specificity 82% for the diagnosis of PC269. In patients undergoing pancreatic resection for PC the

45 pre-operative CA19.9 value has been shown to be associated with tumour stage, resectability, risk of recurrence and survival 270-274. Therefore CA19.9 is more likely to be elevated with advanced disease with one series showing only 65% of resectable patients with an elevated CA19.9 and an inability to distinguish PC from chronic pancreatitis (40% of CP patients had elevated CA19.9)275. Thus the sensitivity of CA19.9 is poor in early or small diameter pancreatic tumours with only 50% of tumours < 3cm having an elevated level276. The post-operative value and the change in CA 19.9 have been shown to correlate with survival and risk of recurrence 277 278,279. In non-resected patients the CA 19.9 at time of diagnosis is also a prognostic factor280. In patients undergoing systemic chemotherapy for PDAC the change in CA 19.9 correlates with objective (radiographic) response and survival281-284. Despite its acceptance as measure of PDAC tumour burden there exists caveats in the interpretation of CA 19.9 values. Firstly, CA 19.9 is now known to be a sialylated Lewis a (Lea) antigen. Lewis antigens are normal components of exocrine epithelial secretions (salivary, pancreatic +/- biliary), which are subsequently adsorbed onto the erythrocyte membrane. Like the ABH system they are carbohydrate structures that form epitopes on glycolipids and glycoproteins. They are formed from a type I precursors by the sequential addition of monosacharides by a set of glycosyl- and fucosyl- transferases.285 Two independent genes determine the Lewis phenotype: the Lewis gene or 1- 4 fucosyltransferase (also known as FUT3) and the secretor gene or 1-2 fucosyltransferase (FUT2).286 Individuals who lack a functional FUT3 allele are termed Lewis negative (Lea-b-) and are unable to synthesize CA 19.9. They comprise 7-10% of Caucasians but their incidence is higher in other populations e.g. Africans 22%.287 Individuals with at least one functional FUT3 allele are characterized by the red cell phenotype Lea+b- or La-b+ (rarely La+b+). Lack of a functional FUT2 allele leads to the non-secretor phenotype (as opposed to secretor) which is characterized by the absence of ABH in saliva and on various epithelial cell types.285 Non-secretors (se/se) however in general have higher serum and urine CA 19.9 levels than secretors.288 Secondly, CA 19.9 undergoes some degree of biliary excretion and may be produced by the biliary epithelial cells. Therefore in the setting of cholestasis (impaired biliary excretion) the CA 19.9 level is frequently elevated even in benign conditions289. In patients 46 with PC this is most often due to compression of the common bile duct due to tumours located in the head of the pancreas. Given the limitations of CA19.9 discussed above, its poor sensitivity at detecting early disease and the low prevalence of PC, several professional groups have recommended that it should not be used in the screening of asymptomatic individuals for PC290,291. Studies combining CA19.9 with other serum protein markers improves the ability to distinguish PC from benign pancreatic disease e.g. chronic pancreatitis292. Furthermore one study recently showed that at 95% specificity, CA19-9 (>37 U/mL) had a sensitivity of 68% up to 1 year, and 53% up to 2 years before diagnosis. Combining CA19-9 and CA125 improved sensitivity as CA125 was elevated (>30 U/mL) in ~20% of CA19-9-negative cases293. In chapter 2 I show that measurement of CA19.9 at clinical decision-making time points can have prognostic and predictive value in resected PC patients but its ability to detect early disease appears limited.

PAM4 The monoclonal antibody PAM4 has high specificity for discriminating between PC and its precursors and other cancers, normal pancreas and pancreatitis. The original target antigen was thought to be MUC1 but has recently been identified as MUC5AC294. Tissue expression using immunohistochemistry is seen in 87% of invasive pancreatic adenocarcinomas, including early stage 1 disease, 89% of PanIN-1 to – 3, and 86% of intraductal papillary mucinous neoplasia lesions295. Using a PAM4-based immunoassay the antigen was detected in the serum of 81% of PC patients overall, including 62% of stage 1 disease patients296. The serum level correlates with disease burden and discriminated between PC (stage 1 median value 4.53 units/mL) and chronic pancreatitis (median 1.28 units/mL) and healthy controls (median 1.18 units/mL)296. Combining the PAM4 immunoassay with CA19.9 improves sensitivity (84%) for the detection of PC without any loss of specificity (82%)297. These results provide a rationale for incorporating PAM4 testing into future or existing screening trials.

47 MIC-1 MIC-1 is a protein originally identified in association with macrophage activation and shows growth factor activity. MIC-1 is overexpressed and detectable in the blood of patients with colon and confers a poorer prognosis298. In resectable PC, MIC-1 performs significantly better than CA19-9 in differentiating patients with PC from healthy controls but not at distinguishing PC from chronic pancreatitis299.

Cell-Free Nucleic Acids All individuals contain circulating cell-free nucleic acids (cfNA: DNA, mRNA and miRNA) in their blood. In general cancer patients have higher levels of cfNA than non-cancer patients, which is thought to result from apoptosis and necrosis of neoplastic cells in the primary and metastatic tumour300. Detection and quantification of cfNA has potential utility as a biomarker for diagnosis, prognosis and response to therapy. Detecting cfNA in blood could serve as a minimally invasive assay or “liquid biopsy”300.

Micro-RNA MicroRNAs (miRNAs) are small non-coding RNA molecules, which consist of 19 to 25 nucleotides in their mature form. They are naturally occurring and their predominant role is the modulation of specific messenger RNA (mRNA) activity by inhibiting translation301. They occur in normal tissues but their expression is frequently deregulated in cancer, which may be partly due them frequently residing in regions of chromosomal fragility302. In body fluids including blood, miRNAs are typically stable owing to their presence in membranous vesicles e.g. exosomes, allowing them to withstand mRNA degradation303,304. Several miRNAs have been shown to be differentially expressed within PC305. In predominantly advanced PC, a panel of 4 miRNAs all previously implicated in the development of PC or critical cancer associated cellular pathways showed a sensitivity of 64% and specificity of 89%306. In 197 PC patients of whom 37% were resectable (stage I or II) 7 miRNAs were found to be differentially expressed from controls. The sensitivity and specificity in a validation cohort was 94% and 93% respectively307. Expression of miR-21 also correlated with survival. In the sera of 19 patients with resectable PC (3 stage I, 16 stage II) 18

48 of 735 assayed miRNAs were overexpressed compared to healthy controls and validated in an independent cohort. The best biomarker performance was seen with miR-1290 which accurately distinguished patients with early-stage PC or IPMN from healthy and disease controls (chronic pancreatitis, pancreatic neuroendocrine tumour) and higher serum miR-1290 levels were a marker of poor survival308.

Circulating cell-free DNA The proportion of blood cell-free circulating DNA (cfDNA) deriving from he tumour ranges from 0.01% to 93% in patients with cancer300,309. Sensitive molecular assays can detect tumour DNA aberrations at frequencies as low as 0.01% but a minimum concentration of cfDNA estimated at 30ng/mL (using Sequenom MassARRAY) in the plasma is required for optimal mutation detection310. Several cancers are known to have low average plasma cfDNA concentrations, limiting the ability of cfDNA to detect early stage cancers. In 21 patients with PC, mutant KRAS was identified in the tumour in 15 patients (71.4%) and the same mutation in the plasma in 9 patients (60%)311. Patients with KRAS mutations detected in plasma had significantly larger primary tumours than patients with KRAS mutant tumours but negative in the plasma (mean size 4.2cm vs. 2.6cm, P = 0.04)311. Furthermore, only 2 of the 9 with plasma KRAS mutations were considered resectable (6 had metastatic disease). In a subsequent study of 44 PC patients, plasma KRAS mutations were detected in 12 patients (27%) which correlated with advanced disease stage with 10 of these being stage IV disease and 1 each stage II and III312. In 39 patients tumour DNA was available and KRAS mutations identified in 28 (72%). Furthermore, plasma KRAS mutations were detected in 2 of 37 (5%) patients with chronic pancreatitis. Finally, in 21 PC patients plasma KRAS mutations were detected in 17 (81%) including 4 patients where detection of plasma KRAS mutation preceded the clinical diagnosis by 5-14 months313. However all 4 patients had clinical presentation and symptoms suggestive of PC and 3 had pancreatic masses identified on initial imaging suggesting they were not early stage at presentation. In summary, detection of plasma KRAS mutations with current techniques shows poor sensitivity for the diagnosis of

49 PC. Furthermore, in the majority with plasma KRAS mutations the disease is not early stage or amenable to surgical resection.

Circulating tumour cells Circulating tumour cells (CTC) can be identified using sensitive techniques in the peripheral blood of PC but their detection is more likely in advanced disease. These cells are proposed to lead to distant organ metastases and show up regulation of wnt signaling consistent with this314. In 16 patients with metastatic PC, CTCs were identified in 6 (37.5%) patients of which 4 had more than one CTC identified315. PC had one of the lowest rates of CTC of the tumour types studied. The use of CTCs as a screening test in high-risk individuals has not been evaluated.

Pancreatic Juice and Cyst Fluid Analysis As discussed above PC and its precursor lesion predominantly develop or communicate with the pancreatic ductal system making pancreatic secretions an attractive and logical place to look for early markers of PC. In the case of cystic neoplasms, tumour related molecules often collect and concentrate within the cyst fluid making this a site of potential biomarkers. The collection of these fluids however is invasive with pancreatic juice collection requiring endoscopic cannulation of the pancreatic duct with attendant risk of pancreatitis or aspiration of duodenal fluid with or without secretin mediated stimulation. Cyst fluid aspiration is typically performed under EUS-guidance with small risks of infection, bleeding and pancreatitis. These techniques again are therefore not well suited to population based screening but rather to high-risk groups typically with a precursor lesion identified on imaging but potentially also in those with suspected PanINs, which cannot be visualized.

Pancreatic Juice Over 90 - 95% of all PCs have mutations in KRAS typically codon 12 or 13 and this appears to occur early in the development of PC. Codon 12 KRAS mutations were found in the pancreatic juice of 77% (17/22) of PC patients and in 2 of these patients, KRAS mutations preceded clinical evidence of PC by 18

50 and 40 months respectively316. In this study KRAS mutations were not seen in healthy controls or patients with non-PC pancreatic disease. However mutant KRAS has been reported as a frequent finding in the aging pancreas, particularly in smokers and patients with chronic pancreatitis and PanIN1 lesions317,318. Determining the relative amount of mutant KRAS to wild type KRAS may be useful in discriminating mutant KRAS chronic pancreatitis from PC with the latter typically showing the mutant allele represents more than 0.5% of total KRAS319. The presence of mutant KRAS in chronic pancreatitis patients does not appear to be a significant predictor of PC development at least in the medium term (78 months)320. However in IPMN the detection of mutant GNAS predicts progression and may assist in risk stratification and surveillance of patients with pancreatic cysts321. The presence of mutant TP53 in pancreatic juice is a marker of high-grade dysplasia or invasive cancer in a high risk cohort322. It can be detected in 9% of intermediate-IPMN but 38% of high grade IPMN and 18%, 48% and 75% of PanIN2, PanIN3 and invasive carcinomas respectively. In comparison it was not found in healthy controls or study subjects without evidence of advanced lesions. The detection and quantification of aberrantly methylated promoter regions of several genes such as NPTX2 and mucins can discriminate PC patients from healthy and chronic pancreatitis controls and discriminate the types of IPMN323- 327.

Cyst Fluid Currently utilised methods predominantly aim to define pancreatic cysts into mucinous or serous and do not predict the subsequent behavior of the cyst. Standard techniques such as cytology are of limited benefit as FNA typically yields a pauci-cellular specimen. Cytological examination of the cystic fluid may be diagnostic if either glycogen-rich cells (serous cystadenoma) or mucin- containing cells (MCNs and IPMNs) are present, but the sensitivity is low (35%)121. The addition of cytology brushings may improve the detection of intracellular mucin and thereby increase the sensitivity of detection of mucinous cysts.328 The tumour marker carcinoembryonic antigen (CEA) is currently the

51 best clinically-used marker for classifying pancreatic cysts. It is the best studied and most accurate tumor marker (superior to CA19.9, CA 125, CA 15.3) for diagnosing a mucinous pancreatic cystic neoplasm (PCN), as suggested by a level above 110 to 192 ng/mL329,330. In general the higher the CEA the more likely the cyst is mucinous. However, there is no direct correlation of CEA concentration with malignancy 329,330. In addition the test is not useful for small cysts as it requires 0.2 to 1.0 mL of cyst fluid. More recent studies have evaluated molecular markers, cyst fluid DNA, and proteomic mucin profiling331- 335. The presence of mutant KRAS in cyst fluid has a sensitivity of 45 percent for diagnosing a mucinous cyst, with a specificity of 96 percent336. A high concentration of DNA with over 80% of the DNA molecules showing allelic loss of tumor suppressor genes is associated with malignancy, with a sensitivity of 70 percent and a specificity of 85 percent. Somatic cyst fluid mutations in GNAS (p.R201C or p.R201H) appear to be highly specific for IPMN (identified in 41 to 66 percent of cases) but are not associated with dysplasia grade or carcinoma. The detection of mutant GNAS in cyst fluid therefore may aid in the diagnosis of IPMN219,220. Cyst fluid mucin expression has been evaluated in a study of 79 patients with pancreatic cysts. This study found proteomic analysis was more accurate than cytology or cyst fluid CEA level for identifying lesions with malignant potential (98 versus 71 and 78 percent, respectively)335. In addition, the mucin profiling results were associated with the risk of malignant transformation.

Further prospective studies are required to validate these results and to determine the clinical utility of these markers.

In this dissertation I will initially address is the clinical utility of the only currently utilized PC biomarker, CA19.9, in particular its use in the diagnosis of resectable PC and its prognostic and predictive value. Next I define the current contribution of familial PC, which is currently defined based on family history of PC in a large cohort of 766 patients. Furthermore I examine the clinic- pathological differences from sporadic PC, which may need to be incorporated into the current clinical definition of familial PC to broaden the diagnosis, and

52 are of potential importance for screening programs. Next, I identity a hypermutation phenotype in a subset of PC and determine the underlying mechanism and the contribution of germline variants to this phenotype. Lastly, I define the prevalence of pathogenic mutations in established PC risk genes and candidate moderate-high penetrance cancer predisposition genes in a cohort of nearly 400 predominantly sporadic PC. Many of these would not have been detected using current genetic testing guidelines and several have direct clinical utility either therapeutic or established risk reduction guidelines. Finally I discuss the challenges in identifying new PC risk genes and the knowledge gaps that have developed with lowering genetic testing thresholds. I will discuss the strategies needed to overcome these challenges in future studies, which will eventually allow incorporation of these findings into the clinical management of patients at high risk of developing PC.

Chapter 2: The prognostic and predictive value of serum CA19.9 in PC.

54 BACKGROUND AND AIMS: Current staging methods for PC are inadequate, and biomarkers to aid clinical decision-making are lacking. Despite the availability of the serum marker CA19.9 for over 2 decades, its precise role in the management of PC is yet to be defined, and as a consequence it is not widely used. METHODS: We assessed the relationship between peri-operative serum CA19.9 levels, survival and adjuvant chemotherapeutic responsiveness in a cohort of 260 patients who underwent operative resection for PC. RESULTS: By specifically assessing the subgroup of patients with detectable CA19.9, we identified potential utility at key clinical decision points. Low postoperative CA19.9 at 3 months (median survival 25.6 vs. 14.8 months, p= .0052), and prior to adjuvant chemotherapy were independent prognostic factors. Patients with postoperative CA 19.9 levels >90 U/mL did not benefit from adjuvant chemotherapy (p= .7194) compared to those with a CA19-9 of ≤ 90 U/mL (median 26.0 vs. 16.7 months, p= .0169). Normalization of CA19.9 within 6 months of resection was also an independent favorable prognostic factor (median 29.9 vs. 14.8 months, p = .0004) and normal peri-operative CA19.9 levels identified a good prognostic group, which was associated with a 5-year survival of 42%.

CONCLUSIONS: Peri-operative serum CA19.9 measurements are informative in patients with detectable CA19.9 (defined by serum levels of >5 U/mL), and have potential clinical utility in predicting outcome and response to adjuvant chemotherapy. Future clinical trials should prioritize incorporation of CA19.9 measurement at key decision points to prospectively validate these findings and facilitate implementation.

55 2.1 Introduction PC is one of the most lethal solid organ malignancies. Pancreatectomy offers the only potential for cure, but is only possible in a minority of patients. Even in those patients who undergo resection, most succumb because occult extra- pancreatic metastatic disease was likely present at the time of diagnosis. 1 Systemic therapies are only modestly effective in advanced disease, but have a significant impact in the adjuvant setting, with 5-Fluorouracil and Gemcitabine both having efficacy in a subgroup of patients and increasing five year survival from 10-15% with surgery alone, to 20-25%. 2-5 As a consequence, there is an urgent need to develop biomarkers to better stratify patients for current treatment modalities, and for the testing of novel therapeutic strategies. Carbohydrate antigen 19.9 (CA19.9) first isolated in 1979 337,338, is the only available serum biomarker for PC, and has shown some utility as a diagnostic adjunct and a prognostic marker 339, but is not widely used in routine clinical practice340. Serum biomarkers such as PSA in prostate cancer, CEA in colorectal cancer and CA 125 in ovarian cancer, which have similar limitations and although sometimes controversial are routinely used and are an integral component of clinical trials in those diseases.

Despite its acceptance as a measure of PC tumor burden, caveats in the interpretation of CA 19.9 values limit application across the full spectrum of PC. First, CA19.9 is a sialylated Lewis a (Lea) antigen. Lewis antigens are normal components of exocrine epithelial secretions present on erythrocyte membranes formed from type I oligosaccharide precursors that undergo sequential addition of monosacharides by a set of glycosyl- and fucosyl- transferases 341. Two independent genes determine the Lewis phenotype: the Lewis gene or 1-4 fucosyltransferase (also known as FUT3) and the secretor gene or 1-2 fucosyltransferase (FUT2).342 Individuals who lack a functional FUT3 allele are termed Lewis negative (Lea-b-) and are unable to synthesize CA19.9.343 They comprise 7-10% of Caucasians, but their incidence is higher in other populations such as Africans (22%).342 Individuals with at least one functional FUT3 allele are characterized by the red cell phenotype Lea+b- or La-b+ (rarely La+b+). Lack of a functional FUT2 allele leads to the non-secretor phenotype (se/se) which is characterized by the absence of ABH determinants 56 in saliva and on some epithelial cell types.341 Non-secretors in general have higher serum and urine CA19.9 levels than secretors.344 Second, CA 19.9 can also be elevated in benign pancreatic diseases such as pancreatitis, which often coexist with PC.345 Third, CA19.9 undergoes some degree of biliary excretion and is produced by biliary epithelial cells. Therefore, in the setting of cholestasis, CA19.9 levels are frequently elevated even in benign conditions.346

Defining the subgroup of individuals where CA19.9 measurement is robust in predicting prognosis and chemotherapeutic responsiveness for PC, would improve current management and overall outcomes. Here we show that serum CA19.9 has distinct potential clinical utility in PC patients when assessed at distinct clinical decision points along the patient journey.

2.2 Patients and Methods Detailed clinico-pathologic and outcome data were collected for 260 consecutive patients who had a histopathologic diagnosis of pancreatic ductal adenocarcinoma and recorded CA19.9 levels from hospitals associated with the New South Wales Pancreatic Cancer Network, Australia, (NSWPCN; www.pancreaticcancer.net.au) as previously described.347,348 Ethical approval for the acquisition of data and biological material was obtained from the Human Research Ethics Committee at each participating institution. Clinical data were obtained from patients and family members, hospital notes and physician records. Outcome data including the date, and cause of death was sourced from the NSW Cancer Registry and treating clinicians.

Survival was measured from the date of histopathologic diagnosis until date of death or last follow-up. Patients with an R2 resection (macroscopically positive resection margins) were excluded from the analysis. All patients were staged based on the American Joint Committee on Cancer (AJCC) TNM classification 7th edition349. Recurrence was defined by pathologic confirmation by biopsy, cytology or clinical and radiographic findings consistent with metastatic disease.

Carbohydrate Antigen 19.9 Measurement

57 CA 19.9 measurements were performed at certified laboratories associated with the hospital where the patients were treated. The upper limit of normal used for CA19.9 was 37 U/mL. Red cell phenotyping for Lewis antigen status was not performed in the majority of patients, therefore patients whose CA 19.9 levels were persistently less than 5 (termed non-expressors) were deemed likely to be Lewis antigen (Lea-b-) negative and were analyzed as a separate group. Paired bilirubin and liver enzymes levels were also collected. CA19.9 levels may be artificially elevated in the setting of cholestasis and correcting it in relation to the bilirubin may improve its applicability. Significant impairment in biliary excretion for which bilirubin is a surrogate marker occurs when the bilirubin level is more than 1.5 times the upper limit of normal (or > 2mg/dL) 350. Analyses were performed using CA19.9 values in patients with normal bilirubin levels and in those with impaired biliary excretion (bilirubin > 2mg/dL). In patients with bilirubin > 2 mg/dL the corrected CA19.9 (cCA 19.9) was calculated by dividing the CA19.9 value by the bilirubin (in mg/dL) 350-352.

Patient Cohort The cohort consisted of 260 consecutive patients who had a pathologic diagnosis of pancreatic ductal adenocarcinoma who underwent pancreatic resection with curative intent (macroscopically clear margins). The clinico- pathologic characteristics are summarized in supplementary Table 1 in the appendix (page 94) . Clinico-pathologic factors associated with significantly better survival on univariate analysis (Table 1) included tumors of the pancreatic head (median survival 20.7 Vs. 11.9 months; P = .0053) compared with those of the body/tail, tumor size ≤ 20 mm (34.4 Vs. 16.7 months; P < .0001), absence of margin involvement (21.2 Vs. 12.8 months; P = .0019), absence of lymph node metastases (22.2 Vs. 16.8 months; P = .0261), and adjuvant chemotherapy (22.8 Vs. 16.4 months; P = .0007).

58 Resected patients with CA 19.9 values

Parameter n = 260 Median DSS (mo) P value (logrank) No. (%) Sex Female 123 (47.3) 18.1 Male 137 (52.7) 19.4 .9721 Age, y Mean 65.9 Median 67.0 Range 28-87 Outcome Follow-up (mo) 0 – 115.3 Median follow-up 11.7 30-day mortality 3 (1.2) Death PC 148 (56.9) Death other 17 (6.5) Alive 95 (36.5) Lost to follow-up 0 Stagea IA 7 (2.7) 25.2 IB 10 (3.8) 56.6 IIA 68 (26.2) 22.2 IIB 160 (61.5) 16.9 .3121b III 0 - IV 15 (5.8) 9.3 .0002c Differentiationd Well 23 (8.8) 22.8 Moderate 163 (62.7) 18.1 .8527 Poor 72 (27.7) 18.6 Missing 2 (0.8) Tumor locatione Head 211 (81.2) 20.7 Body/tail 49 (18.8) 11.9 .0053 Tumor sizef  20mm 44 (16.9) 34.4 > 20mm 164 (63.1) 16.7 .0019 Missing 52 (20) Margins Clear 152 (58.5) 22.8 Involved 107 (41.2) 13.2 .0002 Missing 1 (0.4) Lymph nodes Negative 89 (34.2) 22.2 Positive 171 (65.8) 16.8 .0261 Perineural invasion Negative 54 (20.8) 25.6 Positive 197 (75.8) 16.8 .2122 Missing 9 (3.5) Vascular invasion Negative 108 (41.5) 20.7 Positive 115 (44.2) 17.8 .1455 Missing 37 (14.2) Adjuvant Chemotherapy No adjuvant 95 (36.5) 16.4 Any adjuvant 107 (41.2) 22.8 .0007 - < 3 cycles 33 (12.7) -  3 cycles 74 (28.5) 34.3 <.0001 Missing 58 (22.3) Recurrence 166 (63.8) - R0 92 (35.4) - R1 74 (28.5) Median time to recurrence (mo) 9.6 - R0 12.1 - R1 8.0 Palliative chemotherapy 42 (16.2) Radiotherapy No radiotherapy 222 (85.4) 18.3 Any radiotherapy 38 (14.6) 19.8 .6596g - Adjuvant 29 (11.2) .8894h - Palliative 10 (3.8) CA19.9 Pre-resection N = 202 - Median time pre, (mo) 0.4 Post-resection < 3/12 N = 131 - Median time post < 3/12, (mo) 1.2 Post-resection lowest < 6/12 N = 153 - Median time post <6/12 (mo) 1.9 Pre and post < 3/12 N = 88 Table 1: Descriptive Statistics for Patients Resected for PC with CA19.9 values (n = 260).

59 a Staging based on AJCC TNM Staging System 7th Edition, 2010 b Stage I tumors Vs. Stage II for survival analysis c Stage I and 2 tumors Vs. Stage III and IV for survival analysis d Well and Moderately differentiated tumors grouped together for survival analysis. e Patients with tumors located in the head of the pancreas underwent Whipple pancreaticoduodenectomies, and those with tumors of the body/tail had left sided pancreatectomies. f Tumor size was also prognostic > 30 mm (p = 0.0015), and > 40 mm (p < 0.0001). g Analysis compares those patients who received radiotherapy at any time to all others. h Analysis compares those who received adjuvant radiotherapy versus those who did not.

Pre-resection CA19.9 values were available in 202 patients and were taken a median of 0.4 months prior to surgery. Of these, 10 patients (5%) had CA19.9 levels less than 5 pre and post-resection and in the absence of red cell phenotyping were deemed non-expressors. Post-resection CA19.9 values were available in 231 patients but only 131 and 162 patients had levels performed within 3 and 6 months of surgery respectively and of these 9 were excluded from the analysis due to being classified as non-expressors. Paired pre- and post-resection (within 3 months) CA19.9 values were available in 88 patients with 3 non-expressors excluded from the analysis.

Paired bilirubin assays at the time of CA19.9 measurement were available for all patients included in the analysis. Hyperbilirubinaemia, defined as serum level greater than 1.5 times the upper limit of normal (> 2mg/dL or 34.2mol/L) was used as a marker of impaired biliary excretion. In the pre-resection group the median bilirubin level was 2.0 mg/dL with hyperbilirubuinaemia present in 91 (45%). In the post-resection group the median bilirubin level was 0.5 mg/dL with hyperbilirubinaemia in 8 (6%). In the paired pre- and post-resection group hyperbilirubinaemia was present in 40 (45%) and 2 (2%) respectively.

Statistical Analysis Median survival was estimated using the Kaplan-Meier method and the difference tested using the logrank test. P values of less than .05 were considered statistically significant. Clinico-pathologic variables analyzed with a significant P value and those reported to be significant were entered into a Cox Proportional Hazard multivariate analysis and models resolved using backward elimination of redundant variables. Statistical analysis was performed using

60 Statview 5.0 Software (Abacus Systems, Berkeley, CA, USA). Disease-specific survival was used as the primary endpoint.

2.3 Results Clinically relevant time-points were examined to specifically assess a potential role in clinical decision making (i.e.: at 3 months post-resection, within 6 months; prior to adjuvant therapy, and after adjuvant therapy). We initially assessed the value of post-resection CA19.9 in patients who produced CA19.9 (>5 U/mL) in relation to disease specific survival, and subsequently response to adjuvant chemotherapy. In addition, we examined the potential value of preoperative CA19.9, and the change in level with surgery, with and without adjustment for hyperbilirubinaemia.

2.3.1 CA19.9 and disease stage CA19.9 has been proposed as a marker of tumour burden. There is evidence that a level >1000U/mL at diagnosis correlates with unresectable disease but there is limited evidence of lower levels of CA19.9 predicting resectable in particular early stage disease when it is confined to the pancreas (stage IA and IB). In patients with normal bilirubin values we found that 22.2% (2/9) with stage IA and IB disease had CA19.9 values less than 37U/mL. In addition 14.2% (4/28), 12.5% (8/64) and 33.3% (1/3) with stage IIA, IIB and IV disease respectively had normal CA19.9 values pre-resection. Overall, the median pre- resection CA19.9 in the setting of normal bilirubin level increased with disease but this was not significant (P = 0.6947). There was significant overlap in CA19.9 values between the stages and inter-individual levels of CA19.9 within each stage. Supplementary Table 3 (page 98).

2.3.2 Prognostic Value of CA19.9 Post-resection CA19.9 and Survival Of the 122 patients with CA19.9 values at or within 3 months of surgery, the majority (114; 93%) had bilirubin levels < 2mg/dL and did not require CA19.9 adjustment. The post-resection cCA19.9 co-segregated with disease specific

61 survival on univariate analysis into three prognostic groups: 1. those with less than 37 U/mL had the best outcome, 2. 37 – 120 U/mL, an intermediate outcome and 3. greater than 120 U/mL the worst outcome (median 25.6 vs. 20.7 vs. 14.8 months, P = .0052; Figure 1A). In the 37 – 120 U/mL group the CA19.9 decreased in 5 patients, normalized in 3 and increased in 5 patients at 6 months, suggesting that the intermediate group (37 – 120 U/mL) consisted of those that had elevated CA19.9 due to other causes as well as residual disease. Those within the normal range (< 37 U/mL) within 6 months of resection had a better outcome (median 29.9 vs. 14.8 months; P = .0004; Figure 1B),

Post-resection CA19.9 and Recurrence Of the 122 patients with post-resection CA 19.9 values, 76 had pathologically clear resection margins (R0), permitting assessment of disease free survival. The median time to recurrence was 9.0 months (range 0.4 – 57). Disease-free survival (DFS) paralleled disease specific survival (DSS). Patients with a CA 19.9 in the normal range within 3 months had significantly longer DFS than patients with CA19.9 > 37 U/mL (median disease free survival 22.0 Vs. 11.5 months, P = .0194; Figure 1C).

Pre- resection CA19.9 and Survival The ability to predict prognosis prior to pancreatectomy would significantly improve outcomes by better selecting patients for surgery. Preoperative measures of CA19.9 are often confounded by biliary obstruction, and as a consequence we assessed if serum CA19.9 adjusted for hyperbilirubinaemia could be used as a reliable measure of outcome. Of the 202 patients with available pre-resection CA19.9 values, 111 had paired bilirubin levels < 2mg/dL with a median CA19.9 of 138 U/mL (range 1 to 26,600). Those with a pre- resection CA19.9 < 120 U/mL were associated with better disease-specific survival (median 35.6 vs. 17.4, P = .0444; Figure 1D). In the remaining 91 patients with hyperbilirubinaemia the median uncorrected and corrected CA19.9 was 351 and 28 respectively (range 1 – 101,075 and 0.1 – 16,780). Combined uncorrected and corrected pre-resection cCA19.9 < 120 was also associated with a better disease-specific survival (22.2 Vs. 16.7, P = .0058; Figure 1E).

62 Table 2 summarizes the univariate analysis for all categories of CA19.9 measurement. In addition, pre-resection CA19.9 was associated with pathologic stage in patients with normal bilirubin levels but there was a wide range in CA19.9 values for each stage with overlap in the values between stages diminishing its clinical applicability.

PROGNOSTIC Variable Number Median DSS (months) P Value (logrank) Post-resection CA19.9 < 3/12 < 37 U/mL 61 25.6 37 – 120 U/mL 24 20.7 > 120 U/mL 37 14.8 .0052 Post-resection CA19.9 < 6/12 < 37 U/mL 77 29.9 > 37 U/mL 76 14.8 .0004 Pre-resection CA19.9 (bilirubin < 2mg/dL) Non-expressors 10 18.6 < 120 U/mL 42 35.6 > 120 U/mL 62 17.4 .0444 Pre-resection cCA19.9 < 120 U/mL 107 22.2 > 120 U/mL 85 16.7 .0058 Peri-operative change cCA19.9 < 37 pre and post 38 26.0 Any decrease 22 19.6 Any Increase 25 13.3 .0046 PREDICTIVE Variable Number Median DSS (months) P value (logrank) Pre-adjuvant CA19.9 < 90 U/mL 48 26.0 > 90 U/mL 19 16.2 .0190 Pre-adjuvant CA19.9 < 90 U/ml Adjuvant chemotherapy 48 26.0 No adjuvant chemotherapy 23 16.7 .0108 Pre-adjuvant CA19.9 > 90 U/mL Adjuvant chemotherapy 19 16.2 No adjuvant chemotherapy 16 9.0 .7194 Post-adjuvant CA19.9 < 37 U/mL 11 No patient died from PC > 37 U/mL 8 19.6 Not calculable Post-resection CA19.9 < 3/12 – surgery alone < 120 U/mL 40 20.7 > 120 U/mL 19 9.0 .0628 Table 2: Univariate Analysis of CA19.9 Values at Significant Time Points

Peri-operative change in CA19.9 and Survival Of the 85 patients with pre- and post-resection cCA19.9 values the cCA19.9 level remained less than 37 U/mL (but greater > 5) in 38 patients, decreased in 22 patients and increased in 25. These clustered into three corresponding prognostic groups: 1. < 37 U/ml pre- and post- resection, 2. a decrease in CA 19.9, and 3. an increase in CA 19.9. Patients whose CA 19.9 was always in the normal range had a better survival compared to the CA 19.9 decrease and increase groups (median survival 26.0 Vs. 19.6 Vs. 13.3 months, P = .0046; Figure 1F; and supplementary data).

Figure 1: Kaplan-Meier survival curves for: A. Post-resection CA19.9 < 3 months, B. Lowest post-resection CA19.9 < 6 months, C. Post-resection CA19.9 < 3 months and DFS, D. Pre-resection CA19.9: normal bilirubin, E. Pre- resection cCA19.9, F. Peri-operative change cCA19.9. (cCA19.9 = corrected CA19.9).

64 Multivariate Analysis A multivariate Cox Proportional Hazard model was developed using clinico- pathologic variables with P < .05 and those reported to be prognostic on univariate analysis (Tables 1 and 2). The resolved multivariate model after removal of redundant variables showed that post-resection CA19.9 > 120 U/mL, positive resection margins, and adjuvant chemotherapy were independent prognostic factors (Table 3A, B and C).

PROGNOSTIC

Models Variable Hazard Ratio (95% CI) P value A. Resected PC with Positive lymph nodes 1.69 (0.88 - 3.25) .1124 CA19.9 – Initial Model Size > 20 mm 1.45 (0.73 – 2.86) .2862 (n = 260) Poorly differentiated 0.94 (0.47 – 1.85) .8466 Vascular invasion 1.35 (0.71 – 2.54) .3596 Peri-neural invasion 1.03 (0.54 – 1.97) .9203 Involved margins 2.16 (1.12 – 4.15) .0207 Adjuvant chemotherapy 0.65 (0.37 – 1.15) .1346 Pre-resection cCA19.9 > 120 U/mL 1.27 (0.70 – 2.32) .4296 Post-resection CA19.9 > 120 U/mL (≤ 3 months) 2.47 (1.37 – 4.44) .0026 B. Resected PC with CA19.9 Positive lymph nodes 1.53 (0.99 – 2.35) .0553 – Resolved Model Size > 20 mm 1.61 (0.99 – 2.62) .0555 (n = 260) Involved margins 1.79 (1.15 – 2.77) .0090 Adjuvant chemotherapy 0.60 (0.39 – 0.92) .0198 Post-resection CA19.9 > 120 U/mL (≤ 3 months) 1.87 (1.20 – 2.92) .0056 C. Resected PC with Involved margins 2.19 (1.47 – 3.27) .0001 CA19.9 – Final Model (n = Adjuvant chemotherapy 0.61 (0.40 – 0.91) .0172 260) Post-resection CA19.9 > 120 U/mL (≤ 3 months) 1.90 (1.25 – 2.91) .0029 PREDICTIVE

Models Variable Hazard Ratio (95% CI) P value D. Pre-adjuvant CA19.9 < 90 Positive lymph nodes 1.09 (0.51 – 2.30) .8315 – Initial Model (n = 78) Size > 20 mm 2.52 (1.15 – 5.52) .0209 Poorly differentiated 1.11 (0.53 – 2.31) .7793 Vascular invasion 1.09 (0.53 – 2.22) .8125 Peri-neural invasion 2.20 (0.94 – 5.12) .0693 Involved margins 1.69 (0.85 – 3.36) .1334 Adjuvant chemotherapy 0.27 (0.13 – 0.55) .0004 E. Pre-adjuvant CA19.9 < 90 Size > 20 mm 2.56 (1.20 – 5.46) .0146 –Resolved Model Involved margins 1.94 (1.00 – 3.79) .0509 (n = 78) Adjuvant chemotherapy 0.28 (0.14 – 0.57) .0004 Table 3: Multivariate analysis

2.3.3 Predictive Value of CA19.9 Post-resection CA19.9 and benefit from adjuvant chemotherapy Adjuvant chemotherapy was associated with better survival overall (P = .0007; Table 1), however, only patients with pre-adjuvant chemotherapy CA19.9 values <90 U/mL appeared to benefit. Of 71 patients with a post-resection CA19.9 less than 90 U/mL, (measured within a month of commencement of adjuvant therapy, or at an equivalent time point for surgery only patients), 48

65 had adjuvant chemotherapy (45 received gemcitabine based regimens). In these patients, adjuvant chemotherapy was associated with a significant survival benefit (median survival 26.0 Vs. 16.7 months, P = .0108; Figure 2A and C). There were 35 patients with post-resection CA19.9 greater than 90 U/mL. Of these 19 received adjuvant chemotherapy, with no statistically significant survival benefit (median survival 16.2 Vs. 9.0 months, P = .7194; Table 2; Figure 2B, D and E). Multivariate analysis of variables within the CA19.9 < 90 subgroup showed that adjuvant chemotherapy and size were independent prognostic factors with involved margins of borderline significance (Table 3D and E). Although numbers were small, early data suggested that normal CA19.9 levels after completion of adjuvant therapy in those patients who produce CA19.9 may be associated with an excellent outcome (Figure 2F).

66 Figure 2: Kaplan-Meier survival curves for: A. Pre-adjuvant CA19.9 < 90, B. Pre-adjuvant CA19.9 > 90, C. Adjuvant chemotherapy CA19.9, D. No adjuvant chemotherapy CA19.9, E. Pre-adjuvant CA19.9 - stratified by chemotherapy, F. Post-adjuvant chemotherapy CA19.9.

67 2.4 Discussion Serum CA19.9 possesses many features of a robust or clinically useful biomarker and has been well studied to define its limitations. Despite this, there are no clear applications for CA19.9 in the management of PC as there are for other similar serum biomarkers such as PSA for prostate cancer, CEA for colorectal cancer and CA 125 in ovarian cancer, which are although sometimes controversial, used in routine practice and form integral components of clinical trials to further advance clinical management. We focused on defining potential roles for serum CA19.9 at key clinical decision-making time-points in patients who have demonstrable CA19.9 production. If an individual is identified to have the capacity to produce CA19.9, either through Lewis antigen testing or through a level over 5 U/mL at diagnosis as a surrogate, postoperative CA19.9 measurements have potential prognostic and predictive value. First, normalization of CA19.9 postoperatively is associated with a good prognosis. Second, postoperative CA19.9 levels >90 U/mL may be associated with a lack of response to adjuvant gemcitabine based chemotherapy, and third, although numbers are small, a normal CA19.9 after completion of adjuvant therapy is potentially associated with an excellent prognosis. These potential applications are summarized in Figure 3, which also proposes how CA19.9 levels could be incorporated in to trials to inform clinical decision-making.

Figure 3: Schematic representation of suggested time-points for CA19.9 measurements in clinical trials that would specifically address critical decision points and other applications such as surrogate endpoints to identify responders and non-responders early during therapy.

69 Prognostic Value of CA19.9 Postoperative CA19.9 measurements at 3 months (when hyperbilirubinaemia is uncommon), co-segregate into 3 prognostic groups. The intermediate prognostic group of 37 to 120 U/mL at that time likely includes a mix of those with progressive disease, and other causes of increased CA19.9 levels (biliary dysfunction and pancreatitis). This group declares itself by the 6-month stage to segregate into 2 prognostic groups dichotomized by the normal reference value (37 U/mL). In addition, a normal level both pre- and post-operatively identifies a group with the best prognosis, which has a 5-year survival of 42%.

The key element of the present study is that potential clinical utility was directly examined by assessing CA19.9 at specific decision-making time-points. Overall, although not directly comparable, our findings with regard to the prognostic value of CA19.9 are supported by evidence from retrospective cohorts and in clinical trials. In patients undergoing pancreatic resection for PC the preoperative CA19.9 value is associated with tumor stage, resectability, risk of recurrence and survival 350,353-357. Even with intercurrent biliary obstruction, which is present in about 50% of patients, adjustment of CA19.9 levels relative to the degree of hyperbilirubinaemia may still have potential prognostic value and further evaluation is encouraged. The post-operative level and the change in CA19.9 has also been correlated with survival and risk of recurrence 358 359- 362. Evidence is emerging that post-resection CA19.9 velocity akin to PSA doubling time in prostate cancer, may be a better predictor of recurrence and survival.363 In addition, for non-resectable patients the CA19.9 at the time of diagnosis is also a prognostic factor 364.

Predictive Value of CA19.9 In patients undergoing systemic chemotherapy for advanced PC, the change in CA19.9 correlates with objective (radiographic) response and survival. 365-368 The poor survival of the majority of patients even with clinically localized PC who undergo resection is likely due to occult metastatic disease at the time of diagnosis. Based on the rationale of adjuvant therapy targeting low volume disease and that serum CA19.9 is a measure of disease burden, CA19.9 may have predictive value in the adjuvant setting. A key finding in our study is that

70 patients with a post-resection CA19.9 > 90 U/mL did not achieve long-term benefit from adjuvant chemotherapy whereas those with CA 19.9 < 90 U/mL did. Previous studies have generally not addressed the relationship between CA19.9 and response to adjuvant therapy, however assumptions were made and selection criteria altered based on CA 19.9 levels. In CONKO-001, which compared adjuvant gemcitabine to observation, patients with post-operative CA19.9 values greater than 2.5 times the upper limit of normal ( 90 U/mL) were excluded.4 In contrast, no CA19.9 exclusion criteria were used in ESPAC- 1 and -3, or RTOG-9704.2,3,5 In the analysis of CA19.9 data from the RTOG- 9704 study of gemcitabine versus 5-FU before and after chemoradiation, Berger et al. identified that patients with a CA19.9 > 90 had a poor survival (23.0 vs. 10.4 months). 360 Our results suggest that patients with a post-resection CA19.9 > 90 do not obtain long-term benefit from current systemic chemotherapy regimens and support integration of CA19.9 testing pre- and post-adjuvant therapy to define its role as a predictive marker and a surrogate endpoint.

In conclusion, assessing serum CA19.9 at specific clinically relevant time-points whilst cognizant of its limitations has significant potential utility. These strategies should be incorporated into future clinical trials as a priority, which will define its precise role in the routine management of PC, and like PSA for prostate cancer, potentially become part of routine clinical practice.

Chapter 3: Clinico-pathological features of familial PC

ABSTRACT

72 BACKGROUND: Inherited predisposition to pancreatic cancer contributes significantly to its incidence, and presents an opportunity for the development of early detection strategies. The genetic basis of predisposition remains unexplained in a high proportion of familial PC (FPC). METHODS: Clinico-pathologic features were assessed in a cohort of 766 patients with a diagnosis of pancreatic ductal adenocarcinoma (PC). Patients were defined as FPC if they had 1 affected first-degree relatives (FDR), or otherwise classified as sporadic PC (SPC). RESULTS: The prevalence of FPC in this cohort was 8.9%. In FPC families with an affected parent-child pair, 71% were diagnosed 12.3 years younger in the subsequent generation. FPC patients had more FDRs with an extra- pancreatic malignancy (EPM) (42.6% vs. 21.2, P <0.0001), in particular melanoma and endometrial cancer, but not a personal history of EPM. SPC patients were more likely to be active smokers, have higher cumulative tobacco exposure and have fewer multi-focal precursor lesions, but these were not associated with differences in survival. Long-standing diabetes mellitus (> 2 years) was associated with poor survival in both groups. CONCLUSION: FPC represents 9% of PC and the risk of malignancy in kindred does not appear to be confined to the pancreas. FPC patients have more precursor lesions and fewer active smokers but other clinico-pathologic factors and outcome are similar to SPC patients. Furthermore some FPC kindreds may show anticipation. A better understanding of the clinical features of PC will facilitate efforts to uncover novel susceptibility genes and the development of early detection strategies.

3.1 Introduction PC is a lethal disease with a 5 year survival of less than 5%.369 The majority of patients present with locally advanced, or metastatic disease that is not

73 amenable to surgical resection, which currently offers the only chance of cure. Of the 10-20% of patients who undergo resection, most (80%) still succumb to the disease with a median survival of less than 2 years.370 Long-term survivors are usually those who had small non-metastatic tumors, clear lymph nodes, and were resected with negative surgical margins371 PC evolves through non- invasive precursor lesions, with the majority thought to develop from microscopic ductal lesions known as pancreatic intraepithelial neoplasia (PanIN). A small percentage arise from cystic lesions: intraductal papillary mucinous neoplasms (IPMN) or mucinous cystic neoplasms.191,192 Recent studies estimate that a period of 10 to 20 years is required from the time of an initiating mutation to the establishment of advanced disease, suggesting a prolonged period where intervention may be possible.194

Strategies that facilitate the early detection of PC or its precursors during this broad window are extremely attractive. Screening the general population is not feasible due to the low incidence of PC and the lack of a robust screening test. As a consequence, the focus has shifted to individuals considered to be at high- risk. Established risk factors for PC constitute both environmental and inherited influences and include age, ABO blood group, cigarette smoking, diabetes mellitus, obesity and a family history of PC.6 Inherited predisposition to PC manifests in 3 different settings 99: 1. Hereditary tumor predisposition syndromes which account for 15-20% of the burden of inherited disease such as Hereditary Breast Ovarian Cancer (HBOC) and Peutz-Jegher syndrome (PJS)100, 2. Hereditary Pancreatitis, and 3. Familial PC (FPC). FPC is defined as a kindred with at least two first-degree relatives with PC, which do not otherwise fulfill the diagnostic criteria for an inherited cancer syndrome. 101 The underlying genetic basis of PC predisposition has been identified in less than 25% of such families, 105,107,10898, 100, 10198, 100, 101 despite 50-80% of families demonstrating an autosomal dominant inheritance pattern102,103,97,99,100.

3.2 Patients and methods

Patient and Data acquisition

74 Detailed clinico-pathologic, treatment and outcome data for a cohort of 766 patients with a histopathologic diagnosis of Pancreatic Ductal Adenocarcinoma (PC) was accrued from twelve hospitals associated with the Australian Pancreatic Cancer Genome Initiative between 1994 and 2012 (APGI; www.pancreaticcancer.net.au). Recruitment was focused on biospecimen acquisition for genomic studies and as a consequence biased the cohort towards resected cases. Patients were defined as FPC if they had one or more first-degree relatives with a confirmed diagnosis of PC, with the remaining patients classified as sporadic PC (SPC). No patient had a known genetic predisposition or hereditary cancer syndrome at enrolment. Ethical approval was obtained from the Human Research Ethics Committee at each participating institution. All cases underwent central pathology review by at least one specialist pancreatic histopathologist blinded to the diagnosis and clinical outcome to verify the diagnosis of pancreatic ductal adenocarcinoma and to define histopathologic features in a standardized manner using a synoptic report developed for the purpose372. Tumours were staged according to the AJCC Cancer Staging Manual 7th edition 2009373.

Clinico-pathologic information was acquired initially retrospectively, but became prospective in 2006. Prospectively recruited participants underwent a structured interview by a trained interviewer using a validated questionnaire.374 Detailed baseline information included demography, cigarette smoking and alcohol consumption, personal and family history of malignancy and medical comorbidities including diabetes mellitus and pancreatitis. Cigarette smoking was stratified into 3 groups: active, prior and non-smokers. Active smoking was defined as ongoing use or cessation within 6 months of diagnosis. Prior smokers as having smoked >100 cigarettes, but ceased >6 months prior, and were further stratified based on duration of abstinence (6 months - 10 years and >10 years). Non-smokers had smoked fewer than 100 cigarettes in their lifetime. Cigarette smoking was quantified using pack-years with 1 pack-year representing smoking 20 cigarettes per day for 1 year. Alcohol consumption was classified on the basis of average consumption of all alcohol types for 12 months prior to PC diagnosis, using number of standard drinks (10 g ethanol) per day. Mild alcohol consumption represents 0-2, moderate 3-4 and heavy 5

75 standard drinks per day. Diabetes mellitus (DM) was based on physician diagnosis or treatment with insulin or oral hypoglycaemics. The duration of diabetes prior to diagnosis was stratified into 2 groups:  2 years and > 2 years. In both prospective and retrospective cases, additional clinical data was obtained from hospital notes, physician records and family members. The date and cause of death was obtained from cancer registries and treating clinicians.

Statistical Analysis

Disease-specific survival was used as the primary end point and calculated from the date of histopathologic diagnosis to the date of death or last clinical follow-up. Non-resected patients, and those with an R2 resection (macroscopically positive resection margins) were excluded from the survival analysis. Patients alive at the census date (1st June 2013) were censored. Univariate Kaplan-Meier analysis of patient, tumor and treatment variables compared median survival using the log-rank test. Chi-square and Fisher exact tests were used to compare categorical variables and the students t-test to compare continuous variables. Reported P values are two-sided and variables with a P < 0.05 were considered significant. Statistical analysis was performed using Statview 5.0 software (Abacus Systems, Berkeley, CA, USA).

3.3 Results The cohort consisted of 766 consecutive patients with a histopathologic diagnosis of PC, 698 of whom were sporadic (SPC) and 68 satisfied the criteria for FPC. The clinico-pathological characteristics are summarized in Supplementary Tables 1 and 2. The majority of patients (77.9%) underwent pancreatic resection with curative intent. In the FPC subset, 57 patients (83.8%) underwent pancreatic resection and 11 had a diagnostic biopsy only. The majority (77.9%) of FPC families had two affected FDRs and 8.8% had three affected FDRs. The remaining FPC families had combinations of affected first- and second-degree relatives (SDR) as described in Table 1.

Variable FPC SPC P value N = 68 (%) N = 698 (%) FDR with PC 2 FDR 6 (8.8) -

76 1 FDR 53 (77.9) - 1 FDR and 1 SDR 8 (11.8) - 1 FDR and 2 SDR 1 (1.5) - FDR with EPM Number of FDR with an EPM 1 20/68 (29.4) 115/698 (16.5) 2 4/68 (5.9) 27/698 (3.9) 3 4/68 (5.9) 6/698 (0.9) 4 1/68 (1.5) 0 Mean 1.52 1.26 0.0372 Total FDRs with an EPM 44 187 EPM Site - breast 7/68 (10.3) 34/698 (4.9) 0.0579 - colo-rectal 7/68 (10.3) 44/698 (6.3) 0.2077 - prostate 3/68 (4.4) 16/698 (2.3) 0.2834 - endometrial 2/68 (2.9) 4/698 (0.6) 0.0345 - ovarian 1/68 (1.5) 6/698 (0.9) 0.6133 - melanoma 6/68 (8.8) 4/698 (0.6) <0.0001 - gastric 3/68 (4.4) 12/698 (1.7) 0.1261 - lung 4/68 (5.9) 26/698 (3.7) 0.3813 TOTAL with EPM in 1 FDR 29/68 148/698 <0.0001 (42.6) (21.2) Table 1: Distribution of relatives with PC and extra-pancreatic malignancy.

3.3.1 Clinico-pathologic variables and outcome Patients from families with FPC represented 8.9% of all cases. There was no difference in overall outcome between SPC and FPC (Figure 1A and 1B), with median survival in resected patients of 19.8 and 17.4 months respectively (P = 0.1468). In addition, resected FPC and SPC patients had no differences in the distribution of prognostic clinico-pathological variables. (Table 2 and Supplementary Table 1) In both groups, patients with tumors located in the head of the pancreas and/or who received adjuvant chemotherapy had a better survival. Limited numbers in the FPC group likely influenced the statistical significance of other clinico-pathological variables such as size and nodal status.

Figure 1: Kaplan-Meier survival curves for: A. Survival post-resection for FPC and SPC patients, B. Survival in non-resected FPC and SPC patients, C. Survival in FPC patients who underwent localized resection, with and without PanIN 2 and/or 3, D. Survival in SPC who underwent localized resection, with and without PanIN 2 and/or 3.

The gender distribution of FPC and SPC patients was similar, as was the mean age at diagnosis (65.8 vs. 66.0 years, P = 0.8952). Furthermore there was no difference in the proportion of patients diagnosed at an early age (< 50 years)(Table 2). Of the 68 FPC patients, 40 were members of an affected parent-child pair. In 28 of these, the age at diagnosis was confirmed in the affected parent and child, and in 20 of these (71.4%) the age of the child was more than 5 years younger at diagnosis than the affected parent. For parent- child pairs, the mean age at diagnosis in parents was 72.9 years, and 60.6 years in affected offspring (P < 0.0001). The parent of origin did not appear to affect the age at diagnosis in the successive generation with children diagnosed

78 12.3 years earlier if the father was affected and 12.2 years for an affected mother (P = 0.9675).

Variable FPC (%) SPC (%) P value Mean age at diagnosis 65.8 66.0 0.8952

Age < 50 4/57 (7.0) 44/540 (8.1) 0.7653

Location – body/tail 10/57 (17.5) 97/540 (18.0) 0.9326

LN involved 38/57 (66.7) 356/540 (65.9) 0.9106

Differentiation poor 19/56 (33.9) 146/537 (27.2) 0.2841

Tumour size >20mm 50/57 (87.7) 427/540 (79.1) 0.6547

Margins involved 18/57 (31.6) 199/540 (36.9) 0.4312

Perineural invasion 43/53 (81.1) 397/522 (76.0) 0.4059

Vascular invasion 26/47 (55.3) 265/506 (52.4) 0.6987

Multifocal disease 21/57 (36.8) 129/540 (23.9) 0.0320 - PanIN2 or PanIN3 Pre-2004

Any adjuvant chemotherapy 7/15 (46.7) 68/290 (23.4) 0.0417

Adjuvant chemotherapy 3 cycles 5/15 (33.3) 34/290 (11.7) 0.0145

Post-2004

Any adjuvant chemotherapy 31/42 (73.8) 155/246 (63.0) 0.1761

Adjuvant chemotherapy 3 cycles 29/42 (69.0) 136/246 (55.3) 0.0956

Table 2: Comparison of clinico-pathological variables in resected PC patients

Resected FPC patients had more precursor lesions, specifically PanIN-2 and -3 distinct from the carcinoma in the resected specimen than SPC (36.8% vs. 23.9%, P = 0.0320)(Table 2). The presence of PanIN-2 and -3 was not associated with a difference in survival in either FPC or SPC (Figure 1C and D).

3.3.2 Previous extra-pancreatic malignancy There were 11 previously diagnosed extra-pancreatic malignancies (EPM) in 10 patients in the FPC group and 76 EPM in 72 patients in the SPC group. The proportion of FPC or SPC patients with a previously diagnosed malignancy was

79 similar (14.7% vs. 10.3%, P = 0.2636). The types of prior EPMs were similar in both groups with breast, colo-rectal, prostate and melanoma the most common. (Supplementary Table 3) A history of prior EPM was not related to survival in resected FPC (16.7 vs. 19.8 months, P = 0.5699), nor SPC patients (16.1 vs. 17.8 months, P = 0.9408).

3.3.4 Family history of extra-pancreatic malignancy FPC patients were significantly more likely to have at least one FDR with an extra-pancreatic malignancy (44.1% vs. 21.2%, P < 0.0001). Furthermore, they were more likely to have multiple FDRs with an EPM (mean 1.52 vs. 1.26 FDRs, P = 0.0372). (Table 1) The most common malignancies in both FPC and SPC were breast, colorectal, melanoma, lung and prostate. The distribution of malignancies in FDRs was similar in both groups, except that FPC kindreds were more likely to develop melanoma (8.8% vs. 0.6%, P < 0.0001) and endometrial cancer (2.9% vs. 0.6%, P = 0.0345). There was a trend towards higher rates of breast cancer in FPC kindreds (10.3% vs. 4.9%, P = 0.0579). (Table 1 and Supplementary table 4)

3.3.5 Other PC risk factors The prevalence of diabetes mellitus (DM) in FPC and SPC was 27.9% and 28.9% respectively (P = 0.8623). All diabetics in this study had type 2 or 3c DM. There was no difference between FPC and SPC patients with regard to mean duration of DM prior to PC diagnosis (mean 6.1 vs. 5.0 years, P = 0.6112) or the proportion diagnosed within 2 years of PC (58.3% vs. 53.0%, P = 0.7346) (Table 3). There was no association between the presence of DM, or duration, with age at PC diagnosis in resected FPC and SPC patients. (Supplementary Table 6) DM duration greater than 2 years was associated with a poor post- resection survival in both FPC and SPC groups (Supplementary Figure 1B – D). Multivariate analysis showed that positive lymph nodes, involved margins, size  20mm, adjuvant chemotherapy, post-resection CA19.9 >120 U/ml and DM greater than 2 years duration were independent prognostic factors (Supplementary Tables 5, 6 and 7). Multivariate analysis was not performed in FPC and SPC groups individually due to limited numbers in the FPC cohort.

Variable FPC SPC P value Diabetes Mellitus 19/68 (27.9) 202/698 (28.9) 0.8623 Missing date of diagnosis 6/19 (31.6) 98/202 (48.5) DM  2 years 7/19 (36.8) 47/202 (23.3) 0.5551 DM > 2 years 6/19 (31.6) 57/202 (28.2) 0.9327 Chronic Pancreatitis 6/68 (8.8) 37/698 (5.3) 0.2283 Alcohol Missing alcohol data 0 18/698 (2.6) Nil or Low Alcohol Intake ( 2 SD) 58/68 (85.3) 541/698 (77.5) 0.3000 Mod Alcohol Intake (3-4 SD) 5/68 (7.4) 68/698 (9.7) 0.4831 Heavy Alcohol Intake ( 5 SD) 5/68 (7.4) 71/698 (10.2) 0.4216 Cigarette Smoking Missing date ceased 2/21 (9.5) 31/165 (18.8) Never Smoked 41/68 (60.3) 318/698 (45.6) 0.0315 Prior Smoker 21/68 (30.9) 165/698 (23.6) 0.3314 - prior  10 years 5/21 (23.8) 50/165 (30.3) 0.9702 - prior > 10 years 14/21 (66.7) 84/165 (50.9) 0.0627 Active Smoker 6/68 (8.8) 197/698 (28.2) 0.0003 Mean smoke exposure (pack- 25.7 34.9 0.0479 years) Table 3: Risk factors for PC

A history of chronic pancreatitis was present at similar rates in both FPC and SPC (8.8% vs. 5.3%, P = 0.2283), and was not associated with an earlier age of diagnosis. There was no difference in post-resection survival between those with a history of pancreatitis and those without (median survival 18.1 vs. 17.9 months, P = 0.3481).

SPC patients were significantly more likely than FPC patients to be active smokers at the time of diagnosis (28.2% vs. 8.8%, P = 0.0003). Furthermore SPC patients (active and prior) had higher levels of smoke exposure with a mean of 34.9 pack-years of smoking versus 25.7 in FPC patients (P = 0.0479)(Table 3). Active smokers were diagnosed on average 9.8 years and 5.2 years younger in resected FPC (57.3 vs. 67.1 years, P = 0.0144) and SPC (62.4 years vs. 67.6 years, P < 0.0001) patients, respectively, compared to never smokers and prior smokers for greater than 10 years. Resected SPC

81 prior smokers  10 years were diagnosed 3.7 years earlier than never smokers and prior > 10 years (63.9 vs. 67.6 years, P = 0.0237), but this was not significant in the FPC group. There was no difference in age at diagnosis between FPC and SPC active smokers (57.3 years vs. 63.0 years, P = 0.2342) or never smokers and prior > 10 years (67.1 vs. 67.6, P = 0.7536). There was no difference in survival after resection between the 3 smoking classes (Supplementary Figure 1 E and 2A – B).

The majority of patients with both FPC and SPC had a low alcohol intake (nil or less than 2 SD per day) in the 12 months prior to diagnosis (85.3% vs. 77.5%, P = 0.3000), and only 7.4% and 10.2% were heavy drinkers respectively (P = 0.4216). There was no correlation between alcohol intake and age at diagnosis in FPC. (Supplementary Figure 2C – E)

3.4 Discussion The prevalence of familial PC in this cohort was 8.9%. There was no difference in age at diagnosis between FPC and SPC patients, but 71% of FPC families showed probable anticipation. FPC patients were more likely to have multifocal precursor lesions but fewer active smokers and lower smoke exposure. FPC patients were more likely to have one or more kindreds with an EPM, but not more likely to have a personal history of an EPM.

A prevalence of FPC of 8.9% is consistent with previous case-control and cohort studies, 88,90,91 although the requirement of histological confirmation in relatives lowers the rate of familial aggregation.92,93 Previous reports of younger age at diagnosis in FPC are inconclusive 102,375, and some suggest earlier onset by 5 years and a higher proportion (16%) of young-onset disease. 168,376 We found no difference in the age of diagnosis between FPC and SPC patients overall (mean 65.8 vs. 66.0 years), or the proportion with young-onset (<50 years) disease (9.6% vs. 8.6%). However, active smokers were diagnosed 9.8 years (FPC) and 5.2 years (SPC) earlier than never smokers, and those who ceased greater than 10 years previously. In 71% of affected parent-child pairs the child was diagnosed, on average, 12.3 years younger. This finding is unlikely to be due to environmental risk factors as the majority were non-

82 smokers. Anticipation has been reported in 32-85% of FPC families with successive generations developing PC 10-20 years earlier. 18,103,168 Age at diagnosis, anticipation and smoking has important implications for risk management, screening program development, and the identification of novel susceptibility genes.377

Those with FPC were twice as likely as SPC to have at least one FDR with an extra-pancreatic malignancy (42% versus 21%). In the majority of inherited cancer syndromes, the risk of malignancy is not confined to a single organ. In addition to pancreas, breast and ovarian cancer, BRCA2 mutation carriers are at increased risk of cancers of the prostate, gallbladder, bile duct, stomach and melanoma. 112 A personal history of extra-pancreatic malignancy was present in nearly 15% of FPC patients, which was not significantly higher than in SPC patients at 10%. This is consistent with previous reports of a 13 - 16% incidence of previous EPM in SPC. 378 Approximately 8.0% of cancer patients in the United States and Australia are expected to develop a second invasive malignancy and of these, an estimated 6% develop a second malignancy in a different organ.378-380 The occurrence of multiple primary malignancies in FPC kindreds is suggestive of an underlying genetic predisposition, with variable penetrance, interaction with other modifier alleles, and gene-environment factors.381 Understanding these complex phenotypes is important for the discovery of novel susceptibility loci, particularly at a time when advances in genomic sequencing have enabled the generation of large numbers of cancer genomes.

SPC patients were more likely to be active smokers at the time of PC diagnosis and had higher exposure to cigarette smoke than FPC patients. There was no difference in other risk factors such as alcohol consumption, DM and chronic pancreatitis. Recent data also support the notion that patients who smoke and have a family history of malignancy in a FDR require a reduced dose of tobacco exposure for the development of PC.20 A higher proportion of multifocal precursor lesions in FPC patients is consistent with previous findings.382 Importantly this did not affect outcome after localized resection.

83 Consistent with previous studies, there was no difference in survival between resected FPC and SPC patients. 19,375 Importantly we found that long-standing DM (> 2 years) was an independent prognostic variable in all patients who underwent resection. Its role as a prognostic marker is less well established, 13,37 and previous studies have yielded conflicting results.383,384

Our study has several potential limitations. Firstly, due to the nature of the data we were not able to adjust rates of EPM in close relatives for family size. Secondly, we utilized PC patients as proxy respondents to inform of cancer diagnoses in FDRs. Although proxy reporting has been shown to be accurate particularly for cancer diagnoses in FDRs there remains potential for recall bias.385 Furthermore, histologic confirmation of PC in family members was not possible in all cases due to often advanced stage at presentation without a tissue diagnosis, which was common practice at that time. Approximately 40% of the patients in this study were acquired retrospectively and as such is subject to bias associated with retrospective data. Finally our study is weighted toward resected patients due to minimum tissue requirements for additional studies.

In conclusion, in this cohort, FPC represents nearly 9% of all PC patients. FPC is likely to be a heterogeneous syndrome with phenotype determined by the underlying genetic variants and modified by environmental risk factors. Some familial clustering is likely to occur due to phenocopies from common environmental exposures. Robust clinical characterization of FPC is indispensible for ongoing efforts to identify susceptibility genes, particularly in the age of massively parallel genomic sequencing.

Chapter 4: Hypermutation in PC

85 ABSTRACT BACKGROUND: PC remains the fourth leading cause of cancer death with few effective therapeutic options. Mutation burden has recently been linked with response to immune checkpoint inhibitors in colon cancer and is a potential therapy for cancers of different organs with similar molecular characteristics.

METHODS: We assessed the mutation burden in the tumours from 392 patients with pancreatic adenocarcinomas using whole genome or exome sequencing and defined potential underlying causes using multiple assays.

RESULTS: Outlier analysis defined 20 hypermutated tumours (5.1%) with the greatest mutational burden, from those4 were mismatch repair (MMR) deficient based on analysis of microsatellites, immunohistochemistry, MMR gene mutation and mutational signatures. The remaining hypermutated tumours were enriched with tumours presenting defects in homologous recombination (n=10). Additionally, 2 tumours contained a high contribution of a mutational signature (T>G at TT sites) potentially associated with DNA oxidation.

CONCLUSION: Somatic hypermutation occurs in up to 5% of PC and is predominantly associated with defects in DNA maintenance.

86 4.1 Introduction The evolution of somatic genomic changes, which provide a selective growth advantage, is fundamental to the development and progression of cancer. These changes can take the form of simple somatic mutations (single nucleotide variants and small insertions and deletions), structural variants and epigenetic modifications, with individual tumours frequently displaying a unique combination of each type of variant reflecting the underlying process driving tumourigenesis386. The mutational load between tumours of the same pathological type can differ by several orders of magnitude. Hypermutation in tumours infers underlying mechanisms that may promote cancer development and progression. Recently elevated mutation burden has been linked to response to immune checkpoint inhibitors in melanoma387, lung388 and colon cancer389. Therefore defining the mutational mechanisms and heritable predispositions which contribute to hypermutation in different cancer types will have significant potential implications for clinical management. While the carcinogenic potential of hypermutation is still debated with many of the mutations likely passenger events, it is clear that hypermutated genomes do contain mutations in important driver genes. Patients with inherited bi-allelic mismatch repair (MMR) deficiency can rapidly develop tumours, which appear solely driven by simple somatic mutations with absence of structural variants. These tumours display ultra-hypermutation after somatic mutation in the proofreading domain of one of the replicative polymerases (POLE or POLD1), up to a threshold of 20,000 exonic mutations390.

Hypermutation can result from increased mutagenesis by exogenous or endogenous mutagens, or as a consequence of defective DNA repair. In addition features intrinsic to the DNA region can promote mutability such as proximity to structural variants, low transcription levels, late replication timing and heterochromatin formation391-393. These serve to preserve critical cellular functions as virtually all genes with essential cellular functions are early replicating. This is due to more effective MMR in euchromatic early replicating regions and actively transcribed regions have mutation rates 25% lower than non-transcribed regions due to transcription-coupled repair391,393,394. When DNA damage repair is deficient this protection to early replicating regions is lost393.

87 Characterisation of the exome- and genome- wide archive of mutations for patterns of mutated nucleotides and the immediate sequence context can inform about the mutational processes operating in each tumour395. This enables detection of mutation signatures which broadly result from endogenous e.g. APOBEC or exogenous mutagen exposure or defective DNA repair392. The underlying basis of some of these signatures has been determined such as ectopic APOBEC activity, DNA mismatch repair deficiency and BRCA deficiency31,396. The underlying basis of several mutation signatures remains obscure.

DNA mismatch repair (MMR) is an evolutionary conserved pathway whose predominant function is to increase the fidelity of DNA replication by correcting mismatched bases and insertion-deletion loops that escape the proofreading function of replicative DNA polymerases397,398. Lynch syndrome, which is characterized by the familial clustering of predominantly colon and/or endometrial cancer but also predisposes to PC, is caused by germline mutations in members of the MMR pathway (MLH1, MSH2, MSH6 and PMS2) 398,399. Characteristically, Lynch related tumours have bi-allelic loss of an MMR gene and show widespread hypermutation, particularly in regions of repetitive DNA sequence, termed microsatellite instability (MSI)398. Microsatellites are tandem repetitive DNA sequences of  2 iterations and 1-6 nucleotide repeats which are non-randomly distributed throughout the genome400. In the absence of effective MMR, DNA mutations accumulate, producing MSI and a ‘mutator phenotype’ which promotes and drives carcinogenesis398. Colorectal and other Lynch related tumours are typically categorized as microsatellite unstable on the basis of somatic allele repeat changes detected by PCR analysis of standardized microsatellites in the Bethesda panel, or absence of 1 or more MMR proteins on immunohistochemistry. However, it is unknown if the same criteria are applicable to other cancer types.

Microsatellite instability occurs in 15 - 20% of colorectal tumours, of which 5% harbour a pathogenic germline MMR mutation, predominantly in MSH2 or MLH1401. The majority of MSI in sporadic CRC is due to somatic hypermethylation of the MLH1 promoter402. However there remains a significant proportion with MMR deficiency but no germline MMR mutation or somatic

88 hypermethylation. Large cohort studies utilising deep exome sequencing have identified somatic hypermutation in 16% of colorectal cancers and up to 35% of endometrial cancers, which appears to be a marker of MMR deficiency with approximately three-quarters of hypermutated tumours showing MSI and pathogenic MMR mutations or silencing403-406. In cases with unexplained MMR deficiency, bi-allelic somatic mutation of MMR genes has recently been recognized as a frequent cause in colorectal, endometrial and advanced prostate cancer407-410. Furthermore subsets of tumours with extreme hypermutation are attributed to missense mutations in the proofreading domains of POLE and POLD1403,411.

Here we examine the prevalence and potential aetiology of hypermutation using in-depth genomic analysis of 392 pancreatic ductal adenocarcinomas (PC) and identify hypermutation in 5%. Most are associated with defects in homologous recombination, mismatch repair and a mutational signature T>G at TT sites previously associated with oesophageal carcinoma.

4.2 Patients and Methods

Patient and clinical data acquisition

The cohort consisted of 392 patients with histologically verified pancreatic ductal adenocarcinoma (PC), prospectively recruited between 2006 and 2013 through the Australian Pancreatic Cancer Genome Initiative (APGI; www.pancreaticcancer.net.au) as part of the International Cancer Genome Consortium (ICGC)412. Ethical approval was granted at all treating institutions and individual patients provided informed consent upon entry to the study.

Genome sequencing of tumour and matched normal Tumour and normal DNA were extracted after histological review from fresh frozen tissue samples collected at the time of surgical resection or biopsy as previously described374. Tumour cellularity was determined from SNP array data using qpure413. Tumours with a neoplastic content  40% were selected for whole genome sequencing (WGS), lower cellularity tumours underwent

89 whole exome sequencing. Exome and WGS were performed using paired 100bp on the Illumina HiSeq 2000 as previously described374,413.

Germline and somatic SNV and indel variant identification Single nucleotide variants were called using 2 variant callers: qSNP414 and GATK415. Mutations identified by both callers, or those that were unique to a caller but verified by an orthogonal sequencing approach were considered high confidence and used in all subsequent analyses. Small indels (<200 bp) were identified using Pindel416 and each indel was visually inspected in the Integrative Genome Browser.

Copy number analysis and identification of structural variations Regions of germline and somatic copy number change were detected using Illumina SNP BeadChips with GAP417 Somatic structural variants were identified from WGS reads using the qSV tool418. Structural variants and copy number data were visualised using circos419.

Identification of samples with somatic hypermutation

The total number of somatic mutations in the exome and WGS sequence data were analysed to define samples with hypermutation. Hypermutated samples were defined as outliers, with extreme outliers having a total number of mutations (coding and non-coding SNVs and indels) per megabase (mut/Mb) ≥ 75th centile + (3 x the interquartile range (IQR)), while outliers were defined as ≥ 75th centile + (1.5 x IQR). The threshold for extreme outliers in the exome and WGS groups was 5.40 and 5.87 and for outliers 3.54 and 4.16, respectively. Figure 1.

Figure 1: Box and whisker plot showing the distribution of mutation burden and the outliers in exome and whole genome sequenced groups

Identification of MMR deficiency

Two approaches were used to identify MMR deficiency:

(a) Bioinformatic analysis of sequence data

We used a previously validated tool, MSIsensor which directly compares microsatellite repeat lengths between paired normal and tumour sequencing data420. In the exome sequence the median number of microsatellite loci tested was 20,183.5 (range 0 to 33,752), while in the WGS group the median number

91 of microsatellite loci tested was 980,161.5 (range 12,167 to 1,261,702). We adopted an MSIsensor score of > 3.5% of somatic microsatellites with repeat length shifts as the threshold to indicate MSI as published for endometrial cancer, which correlated well with the 5 and 7 microsatellite panels recommended in the Bethesda guidelines420,421.

(b) Immunohistochemistry Tissue microarrays were constructed using at least three 1mm FFPE tumour cores. Immunohistochemistry for MSH6 and PMS2 was performed on TMA sections as a screen for MMR deficiency due to MMR proteins forming heterodimers with concordant mismatch repair loss (i.e., loss of MLH1 and PMS2 or loss of MSH2 and MSH6) 422. The IHC was performed as previously described and scored by a senior pathologist422. Normal tissue adjacent to the tumour was used as an internal control. The staining patterns were scored as follows: 0 - Negative staining in tumour, positive internal control (mismatch repair deficient), 1 - Equivocal staining (tumour is negative, but internal control is negative), 2 - positive staining (normal pattern of staining). IHC on whole sections for all 4 proteins was performed for all cores scoring 0 or 1.

Annotation of germline MMR variants

The genomic coordinates of germline variants were annotated for gene consequence with ensembl v75423. The effect of missense variants were predicted using polyphen2424, SIFT425, CADD426, VEST3427, alignGVGD428, MutationTaster429, phyloP430 and phastCons431. Allele frequency in the general population was obtained from the 1000 Genomes Project (www.1000genomes.org), Exome Variant Server (http://evs.gs.washington.edu/EVS) and dbSNPv141 (http://www.ncbi.nlm.nih.gov/SNP). The presence of variants in the published literature and inherited mutation databases including ClinVar (http://www.ncbi.nlm.nih.gov/clinvar), HGMD professional432, OMIM231 and InSiGHT433 was examined. The results were compiled and variants ranked using a 5-tiered schemata in accordance with the International Society for Gastrointestinal Hereditary Tumours (InSiGHT) guidelines433 (class 5 = pathogenic, class 4 = likely pathogenic, class 3 = uncertain significance, class 2

92 = probably no pathogenicity, 1 = no pathogenicity). Pathogenic and likely pathogenic variants were of the type expected to abrogate function of the transcript or protein including frameshift, nonsense and splice acceptor/donor variants or in the case of other variant types (missense, inframe, intronic, splice region variants) if published functional data demonstrated loss-of-function consistent with the disease phenotype. Class 5 variants are distinguished from class 4 by the presence of supporting functional data. Class 3 variants are rare (minor allele frequency (MAF) < 0.01), predicted to be deleterious (VEST3 provided the SIFT3 normalised score is >0.6) but represent variants of uncertain significance. Class 1 and 2 variants are predicted to be tolerated (benign), have a MAF  0.01 or synonymous variants.

Methylation arrays

Bisulfite converted whole-genome amplified DNA was hybridized to Infinium Human Methylation 450K Beadchips according to the manufacturers protocol (Illumina). Methylation arrays were performed on DNA from 265 PC which was compared to DNA from 29 adjacent non-tumoural pancreata. We examined the data for evidence of tumour specific hypermethylation of the promoter region of MLH1 and MSH2 genes. Specifically we looked at 48 CpG probes overlapping MLH1: 27 and 10 probes approximately 1500 base-pairs and 200 base-pairs respectively from the transcriptional start site and 8 probes in the 5’ untranslated region (UTR). For MSH2 there were 6 probes in the promoter region: 2 and 3 probes approximately 1500 base-pairs and 200 base-pairs respectively from the transcriptional start site and 1 probes in the 5’ untranslated region (UTR). No differential methylation was observed in the MSH2 promoter region between tumour and adjacent tissue. The methylation array data has been deposited into the ICGC data portal (dcc.icgc.org, project PACA-AU).

Mutational signature analysis

Mutational signatures were defined for genome-wide somatic substitutions, as previously described31.

93 4.3 Results

The cohort consisted of 392 consecutive patients, the majority of whom had sporadic PC and most (96.9%, n = 380) underwent operative resection. We examined the cohort for family and personal history of PC and the typical Lynch related cancers (colorectal and endometrial). Twenty eight patients (7.1%) had one first-degree relative (FDR) or more with PC and 26 (6.6%) and 4 of these (1.0%) had at least one FDR with colo-rectal (CRC) or endometrial cancer respectively. A family history of multiple kindreds with CRC was apparent in 5 patients and for 2 of these, the pedigree met the Amsterdam II criteria for Lynch syndrome421. A metachronous or synchronous malignancy (excluding non- melanoma skin cancer) was diagnosed in 54 patients and this included CRC and endometrial cancer in 8 (2.0%) and 2 (0.5%) patients respectively. Figure 2 and Supplementary tables 1 and 2.

94 Figure 2: Venn diagram showing the proportion of patients with personal or family history of Lynch syndrome related malignancy in particular PC, colorectal (CRC) and endometrial cancer (EC).

In total, 392 tumour-normal pairs underwent whole genome (WGS; n=183) and/or whole exome (WES; n=209) sequencing. We detected a median of 1.7 and 1.1 simple somatic mutations (substitutions and small insertions and deletions) per megabase (mut/Mb) in WGS and WES data respectively. (Supplementary Table 3) Hypermutated samples were identified by outlier analysis, whereby extreme outliers were defined as tumours with a total number of simple somatic mutations per megabase ≥ 75th centile + (3 x the interquartile range (IQR)), and outliers as ≥ 75th centile + (3 x IQR). The threshold for outliers 4.1 and 3.5 mutations per megabase, and extreme outliers in the WGS and exome groups was 5.8 and 5.4 mutations per megabase, respectively. Overall we detected 20 outliers (5.1%) with 10 extreme outliers (2.6%). (Figure 1 and 3, Table 1 and Supplementary Table 4).

Figure 3: Overview of the methods used to characterize the cohort and define the mutational mechanisms producing increased mutation burden. Sequence data was analysed for 392 PC was performed. Hypermutated samples were detected and MMR was estimated from sequence data with the MSIsensor tool (MSIs) and IHC (immunohistochemistry). N normal, A abnormal. The cutoff used for an abnormal MSIsenor score was 3.5%

96 ICGC_ID Seq type MMR readouts Proposed aetiology of hypermutation Dominant Muration MSIsensor KRAS TP53 mutation Structural variants(SV) IHC load(Mut/Mb)a (%)b signature Extreme SV total SV subtype outliers 0 MSH2 MMR deficiency: >280kb somatic homozygous ICGC_0076 WGS p.G12V p.G245S 37.3 MMR 55.0 131 Scattered 28.3 deletion over MSH2 0 MSH6 0 MLH1 MMR deficiency: Somatic MLH1 promoter ICGC_0297 WGS WT p.T125A 58.66 MMR 100 75 Scattered 27.33 0 PMS2 hypermethylation

0 MSH2 MMR deficiency: >27kb somatic foldback ICGC_0548 WGS WT p.S215N 29.16 MMR 49.8 49 Stable 17.47 rearrangement disrupting MSH2 0 MSH6 Deamination Deamination: age-related at CpG sites ICGC_0328 WGS p.G12D WT 16.09 110 Scattered Normal. 3.2 35.6 EAC signature: aetiology unknown. 0 MSH2 MMR deficiency: Somatic MSH2 splice site ICGC_0090 WES p.G12C p.R213* 12.9 NA (WES) NA NA (WES) 0.21 0 MSH6 c.2006G>A BRCA BRCA deficiency: No germline or somatic cause ICGC_0054 WGS p.G12V p.P153AfsTer28 6.85 310 Unstable Normal 0.01 deficiency 4.0 found. BRCA BRCA deficiency: Germline BRCA2 mutation ICGC_0290 WGS p.G12V p.C238Y 6.33 558 Unstable Not Available 0.07 deficiency 9.2 c.7180A>T, p.A2394*. Somatic CN-LOH. BRCA BRCA deficiency: Germline ATM mutation ICGC_0215 WGS p.G12V WT 6.07 111 Scattered Normal 0.01 deficiency 5.7 c.7539_7540delAT, p.Y2514*. Somatic CN-LOH. ICGC_0324 WES p.G12D p.C238* 6.24 NA (WES) NA NA (WES) Normal. 0 Undefined BRCA BRCA deficiency: Germline BRCA2 mutation ICGC_0034 WGS p.G12D WT 5.89 366 Unstable Normal. 4.02 deficiency 10.1 c.5237_5238insT, p.N1747*. Somatic CN-LOH.

Outliers

ICGC_0131 WGS p.G12D p.R175H 5.45 EAC 9.1 147 focal Normal 0 EAC signature: aetiology unknown. BRCA deficiency: Somatic BRCA2 c.5351dupA, ICGC_0006 WGS p.G12D p.V218CfsTer29 5.12 BRCA 3.7 211 unstable Normal 0.01 p.N1784KfsTer3. Somatic CN-LOH. BRCA deficiency: Germline BRCA2 c.6699delT, ICGC_0321 WGS p.G12D p.S183QfsTer64 4.63 BRCA 6.2 286 unstable Not Available 0 p.F2234LfsTer7. Somatic CN loss- 1 copy. ICGC_0309 WGS p.G12V p.C242AfsTer5 4.59 EAC 9.2 232 unstable Normal 0.03 EAC signature: aetiology unknown. BRCA deficiency: No germline or somatic cause ICGC_0005 WGS p.G12V p.Q165* 4.57 BRCA 3.2 95 focal Not Available 1 found. BRCA deficiency: Somatic RPA1 c.273G>T, ICGC_0016 WGS p.G12V p.V274F 4.46 BRCA 5.0 477 unstable Normal 3.03 p.R91S ICGC_0046 WES p.Q61H p.V272M 4.3 NA (WES) NA NA (WES) Normal 0 Undefined BRCA deficiency: Germline BRCA2 GARV_0668 WGS p.G12V p.C275Y 4.16 BRCA 4.9 464 Unstable Not Available 2.19 c.7068_7069delTC, p.L2357VfsTer2. Somatic CN loss - 1 copy. BRCA deficiency: Somatic BRCA2 c.7283T>A, ICGC_0291 WES p.G12R WT 3.84 NA (WES) NA NA (WES) Not Available 0.03 p.L2428* ICGC_0256 WES p.G12D p.R273H 3.72 NA (WES) NA NA (WES) Not Available 0.06 Undefined

97 Table 1: Overview of patients with somatic hypermutation aMut/Mb total number of mutations per megabase using 3100Mb as the denominator for genome and 50Mb for exome bthe MSIsensor threshold for significance was 3.5% of microsatellite with allelic repeat length shifts WGS whole genome sequencing, WES whole exome sequencing, WT wild-type

Defects in DNA damage repair, mismatch repair, signatures associated with esophageal carcinoma and APOBEC activity have distinct mutational signatures. We were able to define the predominant mutational signatures present in the 183 tumours that underwent WGS (WES sensitivity was insufficient to yield robust signatures). Using a combination of diagnostic immunohistochemistry used to identify MMR deficiency in colon cancer, point mutations of known causative genes, detection of microsatellites using MSIsensor, mutational signatures and number of structural variants to detect unstable genomes434 we were able to define a cause for hypermutation in 14 of the 20 hypermutated cases. An additional 3 had a T>G at TT sites mutational signature associated with oesophageal carcinoma, and is associated with hypermutation in that cancer394. Supplementary Tables 4 and 5. MMR, deficiencies in homologous recombination and the T>G at TT sites signature were significantly enriched in the hypermutated cases. (MMR 3/20 vs. 0/372, P = 0; BRCA 10/20 vs. 10/372, P < 0.0001; T>G at TT sites signature 3/15 vs. 0/166, P = 0).

MSIsensor analysis420 comparing germline and tumour microsatellite repeat lengths in the WGS group (n = 183) classified 4 cases as MSI high (score of >3.5). These included the 3 tumours with somatic mutation rates greater than 12mut/Mb, which is the level that defines hypermutation in colo-rectal cancer403 and a fourth tumour with 5.9Mut/Mb. None of the exome sequenced cases (n = 209) were MSI high. The maximum percentage of shifted microsatellites was 0.21 in a hypermutated tumour (12.9 mut/Mb), which was MMR deficient based on IHC. This score was obtained from a low purity primary tumour which can confound the detection of somatic mutations and result in lower mutant allele ratios414.

Immunohistochemistry (IHC) for MSH6 and PMS2 on matched FFPE tumour tissue microarray cores (n = 182) identified 5 (2.7%) cases with negative staining: 3 for MSH6 and 2 for PMS2 with concordant loss of the obligate binding MMR protein (Table 1, Figure 4A-D and Supplementary table 3).

99 Of the 20 hypermutated PDACs, 4 extreme outliers had MMR deficiency. This was further supported by a high MMR deficiency signature. The genomic events that led to MMR deficiency included: 1. a somatic homozygous deletion in MSH2 (also visible on SNP array)(Figure 4E); 2. a complex foldback rearrangement between the EPCAM and MSH2 genes disrupting the coding region of MSH2 (Figure 4F); 3. a somatic MSH2 splice mutation c.2006G>A. (Figure 4G) Previous reports have shown that a germline variant (c.2006G>T) at the same nucleotide is pathogenic, resulting in skipping of exon 13433. The tumour from this patient also harboured a somatic nonsense mutation in SETD2 (c.4022C>T, p.R1342*). SETD2 is essential for methylation of histone H3K36me3, which recruits MSH2-MSH6 onto chromatin435; and 4. Hypermethylation of the MLH1 promoter region. Somatic hypermethylation of the MLH1 promoter is the predominant mechanism of MSI in sporadic colon cancer. Furthermore deletions at the 3’ end of EPCAM can lead to transcriptional read-through and hypermethylation of the adjacent gene, MSH2. Methylation analysis was performed on 265 PC (Illumina 450K methylation array) and compared to 29 adjacent non-tumour pancreatic tissue. In total 48 CpG sites were assayed for the MLH1 gene locus including 27 sites located approximately 1500 bp from the transcriptional start site (TSS), 10 sites 200 bp from the TSS and 8 sites in the 5’ untranslated region (UTR). No significant difference was observed between tumour and adjacent non-malignant pancreas. However, 5 tumours showed evidence of MLH1 promoter hypermethylation compared to a cohort of non-tumour samples adjacent to tumour. (Supplementary tables 8 and 9) The highest level of MLH1 promoter methylation was seen in a hypermutated tumour (ICGC_0297). This tumour showed consistent methylation of probes across the region 1500 bp and 200 bp from the transcription start site and 5’ UTR. Figure 4H and Supplementary Tables 8 and 9. This finding was verified at the protein level with IHC showing absent MLH1 protein expression. The remaining four tumours showed methylation at only a minority of probes. One of these (ICGC_0215) showed hypermutation (6.07 mut/Mb) but MSIsensor (score 0.01) was not suggestive of MSI and IHC for MLH1 was normal. No tumour showed methylation in the promoter region of MSH2.

100

Figure 4: Immunohistochemistry demonstrating MMR deficiency and the genomic mechanisms. 4A: Absent MSH2 staining in tumour cells but normal staining in stroma in ICGC_0076. 4B: Absent MSH2 in tumour cells in ICGC_0548. 4C: Tumours typically show concordant loss of the obligate binding MMR protein in this case MSH6 (ICGC_0090). 4D: Absent MLH1 in tumour cells in ICGC_0297. 4E: Somatic homozygous deletion involving MSH2 (ICGC_0076). 4F: Somatic foldback rearrangement between EPCAM and MSH2 (ICGC_0548). 4G: Somatic MSH2 splice site mutation (ICGC_0090). 4H: Beta values for methylation probes in the MLH1 promoter region.

Germline mutations in MMR pathway members (MLH1, MSH2, MSH6 and PMS2) cause Lynch syndrome which predisposes to the development of colorectal, endometrial and other tumours including PC398,399. Germline variants in the proofreading domains of the replicative polymerases POLE and POLD1 predispose to colon cancer and tumours with pathogenic germline or somatic variants typically show ultra-hypermutation411. We screened and ranked all germline variants in the 4 key MMR members and the 2 replicative polymerases for pathogenic variants. (Supplementary table 6). One pathogenic (class 5)

101 MMR mutation was identified in PMS2 (c.1738A>T, p.Lys580*) in a patient with no personal or family history of malignancy. This mutation was validated with long-range PCR and bi-directional Sanger sequencing. The mutation is reported in the variant databases InSIGHT (PMS2-00090), HGMD and ClinVar as pathogenic. Tumour pathology showed a mucinous non-cystic carcinoma with focal signet ring cells. IHC was normal for PMS2 and WGS did not show hypermutation or MSI. There was no significant increase in the tumour variant allele frequency in the tumour to suggest loss of heterozygosity (germline VAF 51.9%, tumour VAF 60.7%). In patients with a personal or family history of colorectal or endometrial cancer no pathogenic variants (class 4 or 5) were found, including the two patients whose kindreds met the Amsterdam II criteria. In addition no germline CNV was detected in any of the MMR genes or the 5’ end of EPCAM. No pathogenic variants in POLE or POLD1 were identified. Three novel germline missense variants and one inframe deletion were identified within the POLE proofreading domain (amino acids 267 to 471: p.D287E, p.A430T, p.N336S and p.K391del) these were called variants of uncertain significance (class 3). The corresponding tumours did not show hypermutation. No germline missense variants were identified in the proofreading domain of POLD1.

In microsatellite unstable CRCs, poor differentiation and signet-ring cell component show a high level of agreement among pathologists whereas other MSI features show low level agreement421. There was no PC with medullary cancer or a syncytial growth pattern. Poor differentiation was found in 129 PC (32.9%) and 15 (3.8%) were undifferentiated and only 7 tumours (1.8%) showed a signet ring cell component. There was no difference in post-resection survival between patients with MMR deficient tumours and those with intact MMR (median survival 21.6 vs. 20.2 months, P = 0.6588).

Of the 16 hypermutated cases that were not microsatellite unstable, 10 showed evidence of homologous recombination (HR) deficiency: 9 underwent WGS and 8 had dominant BRCA mutational signatures. The BRCA mutational signature was >3 mut/Mb in all 9 cases and 7 of these 9 cases also had a high level of genomic instability with > 200 structural variants. The remaining two cases had a moderate level of genomic instability with 111 and 95 structural variants

102 respectively. Underlying mutations present in these 10 cases included 4 germline and 2 somatic pathogenic BRCA2 mutations, all of which showed bi- allelic inactivation and 1 truncating germline ATM mutation with evidence of 2nd hit in the tumour. In the remaining 3 hypermutated tumours with genomic instability; in one case, genomic instability and a BRCA mutational signature was speculated to be driven by a somatic mutation in RPA1; in the remaining 2, no known cause was identified. Across the 392 patient cohort we identified 20 patients with a pathogenic germline or somatic mutation in BRCA1, BRCA2 or PALB2. However, the 10 that were not within the outlier hypermutated group still had higher mean mutation rates compared to all patients (3.0 vs. 1.6 mut/Mb, P = 0.0014) (Supplementary table 10).

Of the 20 outliers, 6 cases had no obvious explanation for hypermutation. Two tumours exhibited a dominant T>G mutations at TT sites mutational signature previously associated with oesophageal carcinoma31,436. These two tumours had the highest oesophageal carcinoma signature in the cohort and only 3 other tumours showed a dominant and >3 mut/Mb signature. An additional hypermutated PC (ICGC_0328) exhibited borderline MSIsensor score (3.2%), but normal MMR IHC and no mutations in MMR genes. The mutational signature showed a dominant deamination signature previously attributed to aging31 and a T>G at TT sites signature of 4.7mut/Mb. For the remaining 3 cases, no potential causative events could be identified.

4.4 Discussion

In our cohort of 392 unselected predominantly sporadic PDAC, the median mutation burden was approximately 1 mut/Mb consistent with previous findings31. Outlier analysis defined relatively high mutation loads in 5.1% (20/392). MMR deficiency driven by somatic mutation or methylation changes was the underlying cause of hypermutation in 4 outliers (1% of total). Homologous recombination deficiency explained the hypermutation in an additional 10 tumours (50% of all HR pathway related PDAC and 2.6% of total). The remaining tumours showed either a dominant mutational signature

103 associated with T>G mutations at TT sites (n=2, 1% of total), an excessive deamination signature or no potential causative events.

We found that tumours with the highest level of hypermutation (>12mut/Mb) was a marker of MMR deficiency as defined by genomic readouts and IHC. This is similar to the mutation rate in MMR deficient colo-rectal cancer403. Low cellularity tumours reduces the confidence in calling somatic mutations, which is likely to affect mutation rate and MSIsensor calling. As is the case for the Bethesda PCR panel, identification of MMR deficiency using high-throughput sequencing in low cellularity tumours may require more than one readout437. Only one tumour had a mutation rate >12mut/Mb with an intact MMR pathway. This was a cell line with mutational signature showing that the hypermutation was driven by age-related deamination at CpG sites. There have been wide estimates of MSI prevalence in PC due to small cohort size and methodological inconsistency. Our finding of 1.0% (4/392) is consistent with recent studies which performed PCR of microsatellites recommended in the Bethesda panel438. This suggests that MMR deficiency plays a role in pancreatic carcinogenesis but only in a small proportion of cases.

In our PC cohort the MMR deficiency was driven primarily by somatic inactivation of MSH2. Two tumours had structural rearrangements which disrupted both copies of the MSH2 coding region and a further tumour had a somatic MSH2 splice site and a truncating SETD2 mutation. The tri-methylated histone mark, H3k36me3 is required to recruit the MSH2-MSH6 complex to chromatin. H3k36me3 is methylated by SETD2 and cells which lack SETD2 show microsatellite instability435. This contrasts with findings in colorectal and endometrial cancers where of the 15% with MMR deficiency, approximately two-thirds is driven by somatic MLH1 epigenetic silencing and one-third by germline MMR mutations403,404. We identified only one hypermutated tumour with significant methylation of the MLH1 promoter region which was MSI and had absent MLH1 on IHC. This tumour was initially called a PC but subsequent review of the pathology and genomic features raised the possibility of cholangiocarcinoma arising from a biliary intestinal-type intraductal papillary neoplasm. In addition, in our cohort Lynch syndrome does not play a major role in predisposition to PC with only 1 pathogenic germline mutation in PMS2 in a

104 tumour without hypermutation or MMR deficiency. Pathogenic or probable pathogenic variants were not seen in any patient with personal or family history of colorectal or endometrial cancer including the 2 patients whose pedigrees met the Amsterdam II criteria. This is in contrast to CRC where 14.1% (9/64) and 80% (4/5) with a first-degree relative or personal history of CRC or endometrial cancer respectively, had a pathogenic germline mutation in MLH1 or MSH2.437 Patients with colon cancer showing MMR deficiency but no identified germline mutation or somatic MLH1 methylation have been termed as having Lynch-like syndrome and in one study represented nearly one-third of MMR deficient CRCs439. Lynch-like patients show intermediate features between true Lynch and sporadic cases in terms of risk of colon cancer and age of diagnosis439. Often these families are recommended to undergo clinical screening as for Lynch syndrome 407-409,440,441. Our findings echo that of a recent study in advanced prostate cancer which found 12% (7/60) were hypermutated all of which were MSI, frequently due to somatic structural variants in MSH2 or MSH6410. The relative absence of MLH1 promoter hypermethylation and germline MMR mutations support an alternative mechanism for the acquisition of MMR deficiency in PC and highlight the importance of identifying somatic mechanisms of MSI to avoid potentially unnecessary screening for Lynch related tumours in kindreds.

Homologous recombination is an error-free pathway for the repair of double- stranded DNA breaks. Tumours with defective BRCA develop genomic instability with multiple structural rearrangements. There is also clustering of strand-coordinated mutations near the breakpoints of rearrangements which initially develop in transient hypermutable single-stranded DNA intermediaries formed as DSB sites442-444. Overall we identified 10 tumours with hypermutation driven by a defective homologous recombination, 8 contained a dominant BRCA mutation signature, 7 of which also showed genomic instability where available. The underlying cause is either germline or somatic mutation of genes involved in homologous recombination including germline BRCA2 mutations associated with somatic bi-allelic inactivation, somatic BRCA2 mutation or germline truncating mutation of ATM.

105 In conclusion, hypermutation is present in 5% of PC and is typically associated with defects in DNA maintenance and repair. Recent evidence linking hypermutation with response to immune checkpoint inhibitors in other cancer types implies that hypermutated PDAC may also be associated with similar responses. This will have a significant impact on patients with hypermutated PDAC and requires urgent exploration.

106

Chapter 5: Comprehensive Assessment of Germline Cancer Predisposition Genes in PC

107 ABSTRACT BACKGROUND: The genetic component of predisposition to PC remains largely unexplained. In the 5-10% of PC patients who report a family history of PC an inherited mutation has been identified in 20 – 25% of probands. Patients without a family history are presumed sporadic and the inherited genetic contribution to disease is even less well defined. Identification of individuals at increased risk of developing PC could facilitate early diagnosis and subsequently improve outcome. METHODS: We performed whole-genome (n = 184) and exome (n = 208) sequencing of matched tumour-normal DNA pairs from 392 patients with predominantly sporadic PC. We identified pathogenic mutations in candidate genes (n = 150) with established predisposition to PC or medium-high penetrance genes with well-defined cancer associated syndromes or phenotypes. RESULTS: We identified 73 pathogenic single nucleotide and small indel variants in 32 cancer predisposition genes in 65 patients (16.6%). Of these 20 (26.7%) have current direct clinical utility. A mutation in an established PC risk gene was identified in 5.9% (23/392) which was predominantly BRCA2 (9/23). A further 50 pathogenic mutations were identified in 25 moderate-high penetrance cancer predisposition genes (CPG) which have not previously been associated with PC. Overall, the presence of a pathogenic germline CPG mutation did not correlate with age at diagnosis or personal or family history of malignancy including PC but BRCA1-2 and PALB2 mutations were associated with increased family history of breast and/or ovarian cancer. CONCLUSION: A significant proportion of patients with predominantly sporadic PC have a pathogenic germline mutation in a cancer predisposition gene. These are frequently found in the absence of clinical predictors of cancer predisposition and explain only a minority of the familial clustering of PC. Further evidence of the role of these cancer predisposition genes in the development of PC is required.

108

5.1 Introduction PC is a lethal malignancy with an overall 5-year survival of less than 5%369. Surgical resection offers the only potential for cure but is only feasible in the 10 – 15% of patients diagnosed with early stage disease. In patients who undergo pancreatic resection the majority still succumb to the disease. The 10 – 15% of resected patients with survival beyond 5 years are usually those with small primary tumours that are resected with clear surgical margins and uninvolved lymph nodes. Increasing the proportion of patients diagnosed with early stage disease is an attractive strategy to improve outcomes. General population screening is not feasible due to low incidence and lack of a robust screening test. As a consequence, a central focus of early detection strategies is defining groups at increased risk. As genomic sequencing becomes more readily available, defining individuals with increased genetic risk based on a known or suspected inherited predisposition mutation has potential to advance current early detection strategies. In addition, germline mutations can potentially influence therapeutic selection, for example DNA damaging agents or PARP inhibitors in individuals with germline BRCA1 or BRCA2 mutations 445.

Based on family history, it is estimated that ~10% of PC has a significant heritable component. The clinical picture of patients with inherited predisposition typically manifests as one of three scenarios99: 1. hereditary tumour predisposition syndromes including hereditary breast ovarian cancer

(BRCA2), Peutz-Jegher Syndrome (STK11), Familial Atypical Multiple Mole

Melanoma (CDKN2A), Li Fraumeni (TP53), Hereditary Non-Polyposis Colo- rectal Cancer (MLH1, MSH2, MSH6, PMS2) and Familial Adenomatous

Polyposis (APC). Together these account for 15-20% of the burden of inherited disease100 2. Hereditary pancreatitis (PRSS1) which is uncommon; and 3. Familial PC (FPC). FPC is defined as a family with at least 2 affected

109 first-degree relatives who do not fulfill the diagnostic criteria for one of the above inherited tumour syndromes101. The majority (80%) of hereditary PC is attributed to FPC with a pattern consistent with autosomal dominant inheritance in 50-80% of families102,103. Studies to date have delineated the underlying genetic basis in at best 25% of these families with mutations in

BRCA2, PALB2 and ATM accounting for 6 - 19%, 104-106 4.2% 107 and 3.6% 108 respectively109. The proportion of individuals with sporadic PC who carry a deleterious germline tumour suppressor gene mutation is not defined.

Advances in genomic sequencing has allowed testing of cancer related genes in patients who do not meet the traditional testing criteria, and have identified significant proportions with deleterious mutations, often in medically actionable genes446,447. Multigene and/or whole-exome sequencing studies in ovarian cancer have shown that 20% of patients harbour a germline truncation or deletion in a Fanconi anaemia pathway gene446,448.

Here we comprehensively assess the landscape of deleterious germline variants in cancer predisposition genes in 392 patients with predominantly sporadic PC using next-generation sequencing. We identify: 1. The prevalence of deleterious germline mutations in established PC risk genes; 2. The prevalence of deleterious mutations in a set of candidate genes associated with a hereditary cancer syndrome or predisposition to solid organ or haematologic malignancy; 3. Evidence to support potential causality such as loss of the wild- type allele, a mutational signature, or other such evidence in the tumour.

5.2 Patients and methods

Patient and Clinical Data acquisition

110 The cohort consists of 392 prospectively recruited sequential patients with pancreatic ductal adenocarcinoma (PC) and its variants between 2006 and 2013 for the International Cancer Genome Consortium (ICGC) as part of the Australian Pancreatic Cancer Genome Initiative (APGI; www.pancreaticcancer.net.au). Ethical approval was granted at all treating institutions and individual patients consented for the research study including genomic analysis. All cases underwent central pathology review by at least one specialist pancreatic histopathologist blinded to the diagnosis and clinical outcome to verify the diagnosis of pancreatic ductal adenocarcinoma and to define histopathological features using a standardised synoptic report372. Tumors were staged according to the AJCC Cancer Staging Manual 7th edition 2009373. Detailed clinico-pathological, treatment and outcome data was collated from a direct interview with the patient and/or next of kin using a previously validated questionnaire, review of the medical records and correspondence from treating physicians and cancer registries449. We limited our analysis of family history of malignancy to first-degree and second-degree relatives as this has been demonstrated to be accurate when using proxy reporting385. All clinical information and biospecimens were logged and tracked using unique identifiers on a purpose built data and biospecimen information management system.

Candidate gene selection

Candidate genes were selected after review of the published literature and the Online Mendelian Inheritance in Man (OMIM) database450. The list (N = 130) includes genes with established risk for PC, genes with a cancer-related syndrome or well-characterised solid organ or haematological cancer phenotype, and all cancer-related genes in the ACMG reporting of incidental findings policy statement451. Supplementary table 1

Genomic sequencing of tumour and matched normal Tumour and normal DNA were extracted after histological review from fresh frozen tissue samples collected at the time of surgical resection or biopsy as

111 previously described374. A fresh sample of each resected tumour was implanted subcutaneously into six immunocompromised mice for generation of patient- derived – xenografts and cell lines. Tumour cellularity was determined from the SNP array data using qpure413. Matched tumour-normal pairs where the tumour showed a neoplastic content  40% were selected for whole genome sequencing (WGS), paired samples with lower cellularity tumours underwent whole exome sequencing. Exome and WGS were performed using paired 100bp on the Illumina HiSeq 2000 as previously described374,413. Sequence data was mapped to the Genome Reference Consortium (http://www.ncbi.nlm.nih.gov/projects/genome/assembly/grc/human/) GRCh37 assembly using BWA452. Multiple BAM files from the same sequence library were merged and the resulting final BAMs were used for variant calling.

Bioinformatic Analysis

Identification of substitutions and small insertion/deletions Single nucleotide variants were called using 2 variant callers: qSNP414 and GATK415. Mutations identified by both callers, or those that were unique to a caller but verified by an orthogonal sequencing approach were considered high confidence. Small indels (<200 bp) were identified using Pindel416 and each indel was visually inspected in the Integrative Genome Browser (IGV).

Copy number analysis and identification of structural variations Regions of germline and somatic copy number change were detected using Illumina SNP BeadChips using GAP417. Somatic structural variants were identified from WGS reads using the qSV tool418. Structural variants and copy number data were visualised using circos419.

Germline variant classification

The genomic coordinates of germline variants were annotated for gene consequence with ensembl v75423. Missense variants were further classified according to the predicted effect of the nucleotide and amino acid substitution including polyphen2424, SIFT425, CADD426, VEST3427, alignGVGD428,

112 MutationTaster429, phyloP430 and phastCons431. Allele frequency in the general population was obtained from the 1000 Genomes Project (www.1000genomes.org), Exome Variant Server (http://evs.gs.washington.edu/EVS) and dbSNPv141 (http://www.ncbi.nlm.nih.gov/SNP). The results were compiled and variants ranked using a 5-tiered schema in accordance with the ACMG guidelines for reporting of sequence variations453. Variants were ranked using the following criteria (class 5 = pathogenic, class 4 = likely pathogenic, class 3 = uncertain significance, class 2 = probably no pathogenicity, 1 = no pathogenicity). The presence of variants with a score  3 was examined in published literature and inherited mutation databases including ClinVar (http://www.ncbi.nlm.nih.gov/clinvar), HGMD professional432, OMIM231 and locus-specific databases (for example InSiGHT433 and Leiden open variation databases454 (http://www.lovd.nl/3.0/home)). Class 5 variants include variants predicted to abrogate the transcript (nonsense, frameshift, consensus splice site, initiator codon, non-stop) previously reported to be pathogenic and missense variants with functional characterisation demonstrating a functional effect relevant to the disease phenotype and multiple independent reports of pathogenicity. Class 4 variants include previously unreported variants predicted to lead to protein truncation and missense variants with supporting functional evidence, but lacking multiple independent reports of pathogenicity. Variants predicted to abrogate the transcript but occurring in the last exon were called class 3 unless a functional effect or pathogenicity had been previously demonstrated. Class 4 and 5 variants will be collectively termed pathogenic. Class 3 variants all have a minor allele frequency <0.01 and include in-frame indels, splice region variants and missense variants with predicted deleterious consequence according to VEST3 or conflicting reports of pathogenicity. Variants with a minor allele frequency  0.01 were limited to a maximum score of 2. Class 1 and class 2 variants are comprised of missense variants which either have a MAF < 0.01 but are predicted to be benign or a MAF 0.01, synonymous variants and non-coding variants (intronic, UTR, down- and upstream). Supplementary table 2

Bi-allelic Inactivation

113 In tumours with sufficient neoplastic content (> 40%) allowing confident mutation calling, we assessed all pathogenic germline variants for bi-allelic inactivation. Evidence to support a second hit in the tumour came from SNP arrays and/or a significant increase in the tumour variant allele fraction (VAF) supporting loss of heterozygosity (LOH). Briefly, non-silent heterozygous germline variants showing a >15% increase in allele frequency in the tumour after correction for purity and allele sequencing bias was considered supportive of LOH. In addition we assessed all genes with a pathogenic germline mutation for a distinct second somatic pathogenic variant.

Mutational signatures

Mutational signature analysis was performed on the cohort which underwent whole-genome sequencing as per Alexandrov et al455. A robust signature could not be obtained using the whole-exome cohort, therefore this is not analysed further.

Statistical Analysis

Survival analyses used disease-specific survival as the primary end point, which was calculated from the date of histopathologic diagnosis to the date of death or last clinical follow-up. Patients with an R2 resection (macroscopically positive resection margins) were excluded from the survival analysis performed for resected patients. Patients who were alive at the census date were censored. Reported P values are two-sided and variables with a P < 0.05 were considered statistically significant. Statistical analysis was performed using Statview 5.0 software (Abacus Systems, Berkeley, CA, USA). Patients with incomplete comorbidity or pedigree data (first and second degree relatives) were

5.3 Results

5.3.1 Clinical Cohort The 392 PC patients in this prospective cohort were not selected for age at diagnosis or enriched for personal or family history of malignancy. The cohort 114 includes one pair of siblings; all other patients were not known to be related to each other. No patient had an established hereditary cancer syndrome at enrolment. Ethnicity was self-reported as Caucasian in 92.3%, Asian 5.4%, African American 1.8% and Pacific Islander 0.5%. No patient reported Ashkenazi Jewish ancestry. The median age at diagnosis was 67 years and ranged between 33 and 90 years. The majority (96.9%) of patients underwent pancreatic resection with two-thirds having stage IIB disease. Of the 392, 54 patients (13.7%) had a personal history of malignancy (non-melanoma skin cancer excluded) including breast cancer (n = 13, 7.0% females), prostate cancer (n = 13, 6.3% males), melanoma (n = 7, 1.8%) and colorectal cancer (n = 6, 1.5%). A family history of malignancy in at least one first-degree relative (FDR) was reported by 107 patients (27.3%). Table 1. Of these, 28 reported PC in at least one family member fulfilling the criteria for familial PC. In the remaining 79 patients a FDR had an extra-pancreatic malignancy including colorectal cancer (n = 30, 7.7%), breast cancer (n = 23, 5.9%) and prostate (n = 11, 2.8%). The complete clinic-pathological profile of the cohort is shown in Supplementary Table 3 and 4.

115

Variable Cohort N = 392 (%) Median age (range) 67 (33 - 90) Female 185 (47.2) Ethnicity Caucasian 362 (92.3) Asian 21 (5.4) Black/African 7 (1.8) Pacific islander 2 (0.5) Personal history of cancer 54 (13.8) 1 prior malignancy 49 (12.5) 2 prior malignancies 5 (1.3) Site Breast 13 Prostate 13 Endometrial 2 Melanoma 7 Colorectal 6 Gastric 1 Lung 3 Myeloma 1 Other 8 Family history of cancer in ≥1 FDR 107 Number of FDR with cancer 1 59 2 35 3 9 4 3 5 1 Familial PC 28 (7.1) 1 FDR PC 26 (6.6) 2 FDR PC 2 (0.5) Site Breast 23 Prostate 11 Endometrial 4

116 Ovarian 4 Melanoma 4 Colorectal 30 Gastric 8 Lung 13 Haematological 5 Oesophageal 7 Other 34 Table 1: Overview of the cohort demographics and personal and family history of malignancy

Overview of germline variants We identified 38,840 germline single nucleotide variants and small insertions and deletions in the coding regions of 130 cancer predisposition genes in the 392 patient cohort. This represents 2302 unique variants including 1057 missense, 17 non-sense, 14 essential splice site variants and 24 frame-shift and 22 inframe indels. Overall, 73 pathogenic mutations were detected in 32 candidate genes in 65 patients (16.6%). Figure 1 and Table 2. Eight patients harboured 2 pathogenic mutations. The majority (56.5%) of patients with a pathogenic germline mutation did not have a personal or family history of malignancy. Figure 2.

Relative risk of PC HGNC (lifetime risk by age 70- Variant classification Number of patients with Symbol Cancer Syndrome 80) (SNV and small indels) pathogenic variant

Established PC predisposition genes Class 5 Class 4 BRCA2 Hereditary breast/ovarian cancer 3.5 - 10 (3.5%) 7 2 9 Familial atypical multiple mole CDKN2A melanoma (FAMMM) 20 - 34 (10-17%) 0 2 2 STK11 Peutz-Jeghers syndrome 132 (30 - 60%) 0 1 1 Autosomal dominant Hereditary PRSS1 pancreatitis 50 - 80 (25 - 40%) 0 0 0 ATM Ataxia-telangiectasia Elevated but not defined 1 3 4 PALB2 Fanconi anaemia N Elevated but not defined 2 1 3 Adenomatous polyposis coli; Turcot APC syndrome 4.5 (4.32% est) 0 0 0 Hereditary non-polyposis colorectal 8.6 (2 - 6%) MSH2 cancer 0 0 0

117 Hereditary non-polyposis colorectal MLH1 cancer, Turcot syndrome 0 0 0 Hereditary non-polyposis colorectal MSH6 cancer 0 0 0 Hereditary non-polyposis colorectal PMS2 cancer, Turcot syndrome 1 0 1 BRCA1 Hereditary breast/ovarian cancer Elevated but not defined 2 0 3 Other candidate cancer predisposition genes

Tumour suppressor genes

BLM Bloom Syndrome NA 2 1 3 BUB1B Mosaic variegated aneuploidy NA 0 1 1 CHEK2 familial breast cancer NA 2 0 2 CTR9 NA 0 1 1 ERCC2 Xeroderma pigmentosum (D) NA 1 1 2 ERCC5 Xeroderma pigmentosum (G) NA 0 1 1 FANCA Fanconi anaemia A NA 0 1 2 FANCC Fanconi anaemia C NA 1 0 1 FANCE Fanconi anaemia E NA 0 1 1 FANCI Fanconi anaemia NA 0 1 1 FANCL Fanconi anaemia NA 0 1 3 FANCM Fanconi anaemia NA 0 0 2 LIG4 LIG4 syndrome NA 0 1 1 MAP3K6 NA 0 2 2 MUTYH Adenomatous polyposis coli NA 3 0 3 NBN Nijmegen breakage syndrome NA 2 1 7 POLH Xeroderma pigmentosum, variant type NA 0 1 1 RAD51D Familial ovarian cancer NA 0 1 1 Shwachman-Bodian-Diamond SBDS syndrome NA 1 1 5 Familial melanoma, Dyskeratosis TERT congenita NA 0 2 5 TGFBR1 Loys Dietz syndrome NA 0 1 1 VHL von Hippel-Lindau syndrome NA 1 0 1 XPC Xeroderma pigmentosum (C) NA 1 0 1 Proto- oncogen es Multiple endocrine neoplasia 2A/2B, RET Familial medullary thyroid cancer NA 0 1 1 Table 2: PC predisposition genes and other candidate genes with pathogenic variants

118

Figure 1: Pie chart showing the distribution of germline mutations in cancer predisposition genes in patients with pancreatic cancer.

Figure 2: Venn diagram showing the intersection of family history and personal of extra-pancreatic malignancy (EPM) with presence of a mutation in an established PC predisposition gene (Pancreatic CPG) and other CPG. For the 119 purpose of the figure, family history of PC and EPMs have been pooled together.

5.3.2 Pathogenic mutations in PC predisposition genes We identified 23 (5.9%) pathogenic mutations (class 4 and 5) in 7 established PC risk genes including 9 BRCA2 (2.3%), 2 CDKN2A (0.5%), 3 PALB2 (0.77%), 4 ATM (1.0%), 3 BRCA1 (0.77%) and 1 each in PMS2 and STK11 (0.26%). Figure 3. Pathogenic variants were not seen in MSH2, MLH1 and MSH6. The variant callers initially identified 10 pathogenic missense variants in PRSS1 (9 p.N29I and 1 p.Cys139Ser). In view of the 9 trypsinogen genes/pseudogenes with significant homology 8 patients (all p.N29I) that had residual germline DNA were tested with Sanger sequencing using a regulatory approved reference test, which did not detect the variant in any of these samples. The remaining 2 PRSS1 variants which did not undergo validation are not considered further and not included in ongoing analyses.

Figure 3: Lolliplots of pathogenic mutations in established PC predisposition genes

Breast cancer 1, early onset (BRCA1) and breast cancer 2, early onset (BRCA2): Of the 9 pathogenic BRCA2 variants, 8 are disruptive (frameshift or nonsense mutations) and of these 7 are recorded as pathogenic by the BIC. 120 One frameshift deletion (c.6699delT, p.Phe2234LeufsTer7) has not been reported before, but was verified with Sanger sequencing and expected to be pathogenic based on the effect on the transcript. We identified one SNV (c.7976G>A, p.R2659K), which is predicted to produce a missense variant as likely pathogenic (class 4). This variant occupies the position of the last base in exon 17 and results in skipping of exon 17 (171bp corresponding to amino acids 2602 to 2659)456. This region is highly conserved and plays a role in DNA binding and interaction with DSS1. This variant localises to the cytoplasm, segregates with disease in breast cancer families and does not show significant homologous recombination activity457,458. Seven BRCA mutations were detected using WGS and the tumour DNA supported bi-allelic inactivation in 6 typically by loss of the wild type allele (5 copy-neutral loss of heterozygosity (CNLOH) or copy number loss). All tumours with bi-allelic inactivation of BRCA2 showed a BRCA mutational signature >3 mut/Mb and 4 showed a high number of structural variants (2 showed scattered subtype). The one WGS tumour without bi-allelic inactivation had a BRCA signature <1 mut/Mb and stable SV subtype. In the 2 cases with germline BRCA2 mutations detected by WES, one tumour harboured a somatic BRCA2 nonsense mutation (ICGC_0418, c.9382C>T, p.R3128*) but the phase could not be determined. In the remaining tumour BRCA CN-LOH was suggested by the SNP array but no change in the variant allele frequency was seen to support this.

Of the 3 pathogenic BRCA1 mutations, 2 are the missense Slavic founder mutation (c.181T>G, p.Cys61Gly) that has been extensively characterised459,460. The remaining variant is a frameshift deletion (c.3477_3480delAAAG, p.Ile1159MetfsTer50) previously reported as clinically important461. In all 3 patients the tumours were low cellularity (<40%) reducing the confidence of detection of second hits. Of these 3, the tumour with the highest cellularity (37%) suggested loss of the wild-type BRCA1 allele as supported by an increase in the tumour variant allele fraction (ICGC_0401). Mutational signatures and structural variants could not be assessed in these cases.

121 Cyclin-dependent kinase inhibitor 2A (CDKN2A): Germline inactivating mutations in CDKN2A predispose to the development of PC and melanoma462. The CDKN2A gene encodes 2 proteins through alternative splicing of the first 2 exons that regulate critical cell cycles pathways, p16 (exons 1, 2 and 3) which binds to and inhibits the cyclin D-dependent kinases Cdk4 and Cdk6, regulating the retinoblastoma (RB1) pathway and ARF (exons 1, 2 and 3) which is involved in TP53 function by binding the MDM2 proto-oncogene and stabilizing TP53463. The majority of pathogenic CDKN2A mutations are predicted to affect p16 function, but mutations affecting both proteins may have a higher penetrance than either alone.464 We identified 2 pathogenic CDKN2A mutations, one of which is a founder mutation seen in Italian and French familial melanoma kindreds (c.301G>T, p.Gly101Trp)465. Functional characterisation of this variant demonstrates thermo-sensitive loss of inhibition of Cdk4 and Cdk6466. The second pathogenic CDKN2A mutation occurred in exon 1 (c.146T>G, p.Ile49Ser) in a patient with a strong family history of melanoma and PC. Functional characterisation of this variant demonstrates loss of function.467 In both cases there was no change in the VAF to suggest LOH but these were both low cellularity tumours.

Partner and localizer of BRCA2 (PALB2): Germline PALB2 mutations predispose to the development of PC but the absolute risk has not been established179. PALB2 binds to and co-localizes with BRCA2 promoting stabilization and intranuclear localisation. We identified 3 pathogenic frameshift variants in PALB2, which have all been previously described in hereditary breast and or PC178,179. Only one tumour (ICGC_0149) had cellularity > 40% but tumour sequencing did not support bi-allelic inactivation. However this tumour did show a dominant BRCA mutational signature and high genomic instability. The tumour showed a somatic variant in BRCA2 (c.8487G>C, p.Gln2829His) that occupies the last nucleotide in exon 19 and the variant is listed as disease- causing when in the germline (HGMD.CS1213343). The remaining two tumours were low cellularity but did not show evidence of bi-allelic inactivation.

122 Ataxia telangiectasia mutated (ATM): ATM has a key role in regulating multiple signaling cascades in response to DNA damage. Biallelic inactivating mutations result in the ataxia telangiectasia syndrome characterized by progressive cerebellar ataxia, sensitivity to ionizing radiation and predisposition to malignancy468. Heterozygous loss-of-function mutations in ATM predispose to breast and pancreatic malignancies and are found in approximately 3% of FPC kindreds183. We identified 4 pathogenic mutations in ATM including one missense (c.7271T>G, p.Val2424Gly) and 3 truncating mutations. The missense variant increases the risk of breast cancer by 7-fold and demonstrates reduced kinase activity469,470. The 3 truncating variants to our knowledge have not been previously reported. Of the 4 ATM mutation carriers 3 had synchronous or metachronous extra-pancreatic malignancies. All 4 pathogenic ATM mutations showed evidence of a second hit in the tumour with a significant increase in tumour VAF and/or SNP array suggesting LOH; and one a somatic ATM splice donor mutation (ICGC_0045, c.5177+2T>G). One tumour underwent WGS and showed moderate genomic instability but a high BRCA mutational signature.

Post meiotic segregation 2 (PMS2): Mutations in PMS2 result in Lynch syndrome. As discussed in chapter 4 a single truncating mutation in PMS2 was identified in a patient with no clinical features of Lynch syndrome. This mutation was confirmed by long range PCR and bidirectional Sanger sequencing. The tumour was not hypermutated and did not show MMR deficiency using IHC or MSIsensor.

Serine/threonine kinase 11 (STK11): Peutz-Jeghers syndrome is an autosomal dominant disorder caused by mutations in STK11. The phenotype is characterized by mucocutaneous pigmentation, hamartomatous gastrointestinal polyposis and predisposition to malignancy including PC471. In patients meeting the clinical criteria for PJS, mutations in STK11 are present in 94%, including large deletions in 30%472. We identified one STK11 splice-region variant (c.465- 4G>A) in intron 3, which was previously identified in a patient who exhibited

123 typical PJS features472. The corresponding tumour did not show evidence of LOH but was low cellularity.

Clinical correlation in patients with mutations in familial syndrome predisposition genes: In resected patients there was no difference in average age (67.9 vs. 66.6, P = 0.5468) at diagnosis or proportion of those < 50 years (0% (0/23) vs. 7.1% (22/309), P = 0.3829) at diagnosis between those with and without a pathogenic germline PC risk mutation. Patients with germline PC risk mutations reported a non-significant increased personal history of malignancy (31.8% (7/22) vs. 15% (42/280), P = 0.0642) but not family history of malignancy (55.6% (10/18) vs. 40.5% (100/247), P = 0.2249). Patients with BRCA1, BRCA2 or PALB2 mutations were associated with a family history (first or second degree relatives) of breast or ovarian cancer (27.3% (3/11) vs. 6.3% (19/301), P = 0.0348) but not personal history (7.1% (1/14) vs. 2.7% (8/299), P = 0.3412) compared those without a pathogenic germline mutation. Post- resection survival was similar in the group with a germline PC risk mutation and those without (median survival 17.4 vs. 21.7 months, P = 0.4526). Excluding those with incomplete pedigree data (4/23), only 3 of 19 patients with a mutation in a PC predisposition gene met the criteria for FPC, which was not significantly different from the cohort without a germline mutation (15.8% (3/19) vs. 6.4% (21/326), P = 0.1369). Furthermore only 7 (7/19, 37%) patients would meet the current guidelines for consideration of genetic testing in PC patients implying 3 BRCA2, 2 PALB2 and one mutation in each of CDKN2A and ATM could have been detected using our current clinical testing strategies (www.nccn.org)473.

5.3.3 Pathogenic mutations in other cancer predisposition genes For those cancer predisposition genes without an established PC phenotype, 50 pathogenic mutations in 42 patients were identified in 25 of 117 genes. Many of these genes have roles in DNA maintenance including DNA damage repair, replication, recombination and chromosome segregation. In addition to BRCA1/2 and PALB2, 9 other Fanconi anaemia/homologous recombination genes had pathogenic mutations including FANCA x2, FANCC x1, FANCE x1,

124 FANCI x1, FANCL x3, FANCM x2, NBN x7 and RAD51D x1. In the non- homologous end-joining pathway there was 1 mutation in LIG4, nucleotide excision repair had 4 mutated genes (ERCC2 x2, ERCC5 x1, XPC x1) and base excision repair 3 (monoallelic MUTYH x3). Other pathways with pathogenic mutations include cell cycle regulation (CHEK2 x2), resolution of Holliday junctions and chromosome segregation (BLM 3) and telomere maintenance (TERT x5, RTEL1 x1). Pathogenic mutations in other genes included SBDS in 5, MAP3K6 in 2 patients and singletons in VHL, CTR9 and TGFBR1. Figure 40

125 126 Figure 4: Lolliplots of pathogenic mutations in other CPGs

Chromosomal Instability Syndromes: The group of inherited chromosomal instability syndromes includes Fanconi anaemia (FA), Nijmegen breakage syndrome, Bloom’s syndrome, ataxia-telangiectasia and LIG4 syndrome. These disorders result from bi-allelic inactivation of the causative gene and have several features in common, notably spontaneous chromosomal instability, immunodeficiency and predisposition to cancer but have distinct cytogenetic features and sensitivities to DNA damaging agents474.

Fanconi anaemia/Homologous recombination pathway: Fanconi anaemia (FA) members participate in repair of DNA interstrand crosslinks and stalled replication forks and some members in homologous recombination475. Some members of the FA complementation groups are known PC predisposition genes: BRCA2 (FANCD1) and PALB2 (FANCN), while others are implicated in breast and ovarian cancer risk (BRIP1 (FANCJ) and RAD51C (FANCO)). Other FA members specifically FANCC and FANCG have been implicated in predisposition to PC, particularly young-onset (<55 years)476. Overall we identified 10 patients with pathogenic mutations in FA genes excluding BRCA2 and PALB2. Two patients had a heterozygous FANCA missense mutation (c.1874G>C, p.Ser625Cys) previously identified as pathogenic for FA477. One patient harboured the Dutch founder FANCC frameshift deletion (c.67delG), which is implicated in familial breast cancer478. FANCE interacts with FANCD2 to promote DNA repair and resistance to mitomycin C479. A single truncating mutation in FANCE was identified. One patient had a frameshift deletion in FANCI, which has not been previously reported. We identified 3 patients with a FANCL duplication in exon 14/14 leading to a terminal series of 8 missense amino acids and a protein 5 amino acids longer than the wild-type. This variant is functionally hypomorphic showing only partial correction of chromosomal instability and mitomycin C sensitivity480. Two patients harboured a FANCM truncation (c.5791C>T, p.R1931*) resulting in loss of the terminal 117 amino acids. This mutation has been linked to familial breast and sporadic colorectal cancer481. Bi-allelic inactivation could be assessed for only 5 of the 10 FA variants, of which only one variant (FANCM) showed LOH in the tumour. Four

127 tumours underwent WGS but did not show high genomic instability or dominant BRCA mutation signature.

The HR recombinase RAD51 catalyzes DNA strand invasion and exchange in concert with several other proteins including five RAD51 paralogs. Germline inactivating mutations in RAD51D predispose carriers to the development of PARP inhibitor sensitive ovarian cancer with intermediate penetrance482. We identified one patient with a truncating RAD51 mutation (c.298C>T, p.Arg100Ter) who had a second-degree relative with PC. There was no second hit, high genomic instability or dominant BRCA mutation signature seen in the tumour.

Nibrin (NBN): NBN forms a complex with RAD50 and MRE11A, which interacts with ATM and participates in homologous recombination. Carriers are at increased risk of solid organ in particular breast cancer and haematologic malignancy483. It is frequently missense variants in NBN key functional domains that are more frequently associated with an increased cancer risk484. We identified pathogenic NBN variants in 7 patients. This included the missense variant c.643C>T, p.Arg215Trp in 3 patients. Functional studies show this variant is unstable and heterozygous cells show delayed homologous recombination485. One patient carried c.511A>G, p.Ile171Val, which exerts a dominant negative effect on the wild type protein reducing HR through loss of association with DNA repair protein MDC1486. Three patients harboured the founder frameshift mutation (c.657_661delACAAA, p.Lys219AsnfsTer16) which affects the BRCT repeat domains of NBN and impairs the DNA damage response but less than the p.R215W variant487. In four patients the tumour cellularity was adequate to allow assessment of bi-allelic inactivation. In all 4 there was no evidence of a second hit and in 2 which had WGS there was no high genomic instability or dominant BRCA signature.

Bloom syndrome, RecQ helicase-like (BLM): BLM is a RecQ helicase involved in maintaining stability of the genome. Furthermore BLM suppresses critical telomere shortening and interacts with several members of the HR/FA pathway488. Heterozygous BLM mutations have been implicated in

128 predisposition to breast and colorectal cancer, but these findings are not universal489. Mice with a heterozygous BLM mutation show enhanced tumour formation but require an additional insult490. Tumours arising in the setting of germline heterozygous BLM mutations typically do not show loss of the wild- type allele491. We identified 3 distinct heterozygous disruptive BLM mutations, each in single patients which are reported as disease-causing in the Bloom’s syndrome registry492. Bi-allelic inactivation was assessed but not evident in 2 patients tumours. Neither tumor showed genomic instability or BRCA signature. One tumour had a high MMR deficiency signature from somatic inactivation of MSH2.

DNA ligase IV (LIG4): LIG4 syndrome occurs due to bi-allelic mutations in LIG4, a member of the non-homologous end-joining pathway493. Affected cells show radiosensitivity but have preserved cell cycle checkpoints but impaired double stranded DNA repair493. In our cohort one patient harboured a truncating mutation in LIG4 (c.2184delA, p.Glu728AspfsTer16) which although occurs in the last exon (exon 2/2) is expected to result in loss of the final 183 amino acids or 20% of the protein. Bi-allelic inactivation could not be assessed due to low tumour cellularity.

Checkpoint control: CHEK2 is a serine/threonine kinase whose predominant role is to relay DNA damage signals from the proximal checkpoint kinases, in particular ATM and ATR494. Germline inactivating variants in CHEK2 have been identified as low-penetrance susceptibility variants for breast and prostate cancer495. The wild-type allele is frequently retained in CHEK2 mutation carriers491. We identified 2 pathogenic CHEK2 variants, one frameshift deletion in a patient with previous bilateral breast cancer and a splice site variant in a patient with FPC. A second hit was not seen in the one tumour where it could be assessed. This tumour showed scattered structural variants and no high BRCA signature.

Telomere and mitotic spindle genes: Telomeres are repetitive DNA sequences that cap the ends of chromosomes, which in association with the shelterin complex protect chromosomes from recombination, end-to-end fusion and

129 generation of a DNA damage response496. Mutations in telomeric genes leading to both long and short telomeres have been associated with cancer predisposition. Germline TERT promoter mutations leading to enhanced expression of TERT are associated with increased melanoma risk497. Loss of function mutations in TERT and other telomeric genes are associated with short telomeres and the syndrome dyskeratosis congenita, which predisposes to aplastic anaemia and acute myeloid leukaemia (AML)496. We identified 2 hypomorphic TERT variants in 5 patients. The p.His412Tyr variant was identified in 2 patients. This variant has been reported in aplastic anaemia patients with telomerase activity 50% of normal and telomere length 10th centile.496,498 Three patients harboured an in-frame deletion of amino acid 441 (c.1323_1325delGGA, p.441Edel). This variant has also been associated with development of AML and demonstrated short telomeres with approximately 40% of wild-type telomerase activity.496 Bi-allelic inactivation could only be assessed in one patient who harboured two TERT variants but neither variant showed evidence of LOH.

RTEL1 encodes a helicase, which is essential for telomere maintenance. Germline RTEL1 mutations produce autosomal recessive dyskeratosis congenital characterised by short telomeres, bone marrow failure and immunodeficiency. We identified one truncating mutation in RTEL1 in a patient with no personal history or family history of malignancy. The PC in this patient was low cellularity limiting further analysis for LOH.

SBDS plays a role in rRNA processing and assembly and stability of the mitotic spindle499. Bi-allelic SBDS mutations result in Shwachman-Diamond syndrome whose characteristics include pancreatic exocrine insufficiency and bone marrow failure with a high risk of progression to leukaemia showing complex chromosomal abnormalities and aneuploidy500. Recurring disruptive SBDS mutations result from gene conversion events with a nearby pseudogene, SBDSP which has 97% sequence homology501. We identified 5 patients with SBDS mutations, 4 with a splice site mutation (c.258 + 2T>C, p.C84fsX3) and one with a nonsense mutation (c.25C>T, p.Gln9Ter)501. The role of mono-allelic or bi-allelic SBDS mutations in the predisposition to solid organ malignancy has 130 not been elucidated. There are case reports of young-onset PC developing in patients with bi-allelic SBDS mutations502,503. Bi-allelic inactivation could be assessed but was not evident in 4 patients tumours. These tumours did not show a high number of structural variants or dominant mutational signature.

BUB1B forms part of the mitotic spindle checkpoint which helps ensure correct chromosome segregation during mitosis. Bi-allelic mutations in BUB1B lead to mosaic variegated aneuploidy (MVA) characterized by constitutional mosaicism for chromosomal gains and losses and cancer predisposition in particular gastro-intestinal neoplasia, sarcomas and, Wilm’s tumour and leukaemia504. Germline heterozygous mutations predispose to variegated aneuploidies in multiple tissues and young-onset colorectal cancer505. We identified one heterozygous missense BUB1B variant (p.Leu1012Pro) which lies in the kinase domain and has been previously reported in a MVA proband with a truncating BUB1B variant in the other allele504. Kinase domain mutants including this one lead to low BUB1B levels due to high protein turnover and can lead to chromosome missegregation506. Bi-allelic inactivation could not be assessed due to low tumour cellularity.

Other DNA repair pathways: We identified 8 pathogenic mutations in two related single strand DNA repair pathways: base; and nucleotide excision repair. Base excision repair functions throughout the cell cycle to eliminate small, non-helix distorting damaged bases generated by endogenous e.g. reactive oxygen species, deamination or exogenous insults eg alkylating agents507. We identified 3 monoallelic pathogenic mutations in MUTYH: 2 founder missense mutations (p.Tyr176Cys and p.Gly393Asp) and one splice site mutation (c.925- 2A>G). These mutations affect conserved residues and have been reported as pathogenic508,509. Bi-allelic MUTYH mutations result in increased G:C to T:A transversion frequencies and increased risk of cancer in particular colorectal508. Monoellelic MUTYH mutations are associated with a modest predisposition to CRC but these findings are inconsistent and LOH is only identified in around 50% of cases510. Bi-allelic inactivation was assessed but not seen in 2 patients tumours and in one where mutational signature was available there was no dominant signature.

131

Nucleotide excision repair is an important pathway for bulky DNA adducts which distort the helix structure and are commonly related to ultra-violet light exposure and cigarette smoking511. This is exemplified by the syndrome xeroderma pigmentosum, which is caused by inheritance of bi-allelic mutations in NER members and is characterized by sensitivity to sunlight and predisposition to non-melanoma skin cancer. Somatic mutations in NER members specifically ERCC2 correlate with response to cisplatin in invasive uro-epthelial malignancy512. We identified 4 pathogenic mutations in members of the NER pathway, specifically: 3 splice site mutations 2 of which were in ERCC2 (c.183+2T>A and c.594+2_594+5delTGAG) and one in ERCC5 (c.1955-2A>T) and a frameshift deletion (c.1103_1104delAA, p.Gln368ArgfsTer6) in XPC. Bi- allelic inactivation was assessable but not evident in 3 tumours (not assessed for ERCC5 due to low cellularity). Mutational signatures were available for the corresponding three tumours which showed a high (>3mut/Mb) deamination signature in one ERCC2 mutated tumour and was borderline in the other ERCC2 mutated tumour (2.6mut/Mb). The tumour from the patient with germline XPC mutation did not show any dominant signature.

POLH is a DNA polymerase involved in translesion DNA synthesis which has low fidelity as it lacks a proofreading domain513. The predominant role is bypassing DNA damage induced by ultra-violet light, which is in keeping with bi- allelic pathogenic variants in POLH causing variant type xeroderma pigmentosum. A role for inactivating POLH mutations resulting in increased spontaneous DNA mutations due to oxidative damage in internal cancers has been suggested514. We found one inactivating POLH mutation (c.1077_1078insG, p.Asp360GlyfsTer32) leading to loss of the terminal 371 wild-type amino acids. The C-terminal 120 amino acids are required for POLH to localise to the nucleus515. This variant has been previously reported in variant xeroderma pigmentosum cell lines516. Bi-allelic inactivation was not seen and there was no dominant mutational signature.

132 Other cancer predisposition genes Germline loss-of-function mutations in CTR9 have recently been show to predispose to Wilm’s tumour517. CTR9 forms part of the PAF1 complex, which through transcription regulation and histone modification can affect cell cycle control and DNA repair518. One FPC patient harboured a truncating CTR9 mutation (c.1743G>A, p.Trp581*). This patient also had a germline BRCA2 mutation. There was no increase in the tumour CTR9 VAF to suggest LOH.

The mitogen-activated protein kinase (MAPK) is a 3-kinase cascade, which relays signals to the nucleus to activate gene transcription. In response to cellular stress, MAP3K6 activates c-Jun N-terminal Kinase (JNK) and p38 signalling cascades to promote apoptosis519. Recently, inactivating mutations in MAP3K6 have been implicated in a small proportion of non-CDH1 familial gastric cancer with variable penetrance520. We identified two variants predicted to impact MAP3K6 function: one affecting an essential splice site (c.1256- 2A>G); and the other the initiation codon (c.1A>G). In both cases there was no evidence of second hit in the tumour.

Transforming growth factor receptor beta-1 (TGFRB1) is a membrane-bound serine-threonine kinase receptor which forms a heterodimeric complex with TGFBRII leading to downstream activation of SMAD2-4 which enter the nucleus and regulate the transcription of target genes521. Germline mutations in TGFBR1 produce the Loeys-Dietz syndrome characterized by skeletal and connective tissue defects with prominent vascular abnormalities. Polymorphic TGFBR1 variants have been implicated as low-penetrance risk factors for the development of breast, ovarian and colo-rectal cancers521. In addition reduced constitutional expression of TGFBR1 confers risk for colorectal cancer although this is not universally accepted522. One patient without any personal or family history of malignancy harboured a splice site mutation in TGFBR1, this patient also had a SBDS splice site mutation. No evidence of bi-allelic inactivation or dominant mutational signature was seen.

In two patients with histologically verified PC we identified two variants typically associated with endocrine neoplasms. Gain of function mutations in RET lead to

133 the development of multiple endocrine neoplasia type 2 characterised by predisposition to medullary thyroid cancer and phaeochromocytoma. We identified one patient with no family history of malignancy with the missense variant p.Val292Met. This variant has been reported in a patient with medullary thyroid cancer and shows gain-of-function alteration but 10-fold weaker than other pathogenic variants523. Von Hippel Lindau syndrome is a dominantly inherited syndrome due to mutations in VHL which predispose the carrier to several benign and malignant neoplasms (renal cell cancer, phaeochromocytoma). One patient without a family history of malignancy harboured a germline truncation in VHL (c.154G>T, p.Glu52Ter). Approximately 20% of patients develop Von Hippel Lindau syndrome as a result of a de novo mutation524.

Clinical correlation The mean age at diagnosis was 66.1 years and 66.6 years (P = 0.7680) in the germline mutation and no germline mutation groups respectively, similarly there was no difference in the proportion diagnosed at an early age (<50 years) (7.3% (3/41) vs. 7.4% (23/309), P > 0.999). There was no difference in the proportion with familial PC in the group with a non-PC related germline mutation and those without (11.6% (5/43) vs. 6.4% (21/326), P = 0.2073). Similarly there was no difference in the proportion of patients with a family history of extra-pancreatic malignancy (54.5% (18/33) vs. 39.3% (103/262), P = 0.1320) or previous extra- pancreatic malignancy (18.9% (7/37) vs. 14.6% (43/295), P = 0.4685). There was no difference in post-resection survival between the germline mutation and no germline mutation group (21.6 vs. 20.5 months, P = 0.7776).

5.3.4 Germline copy number variants No germline deletions were detected across the cohort in BRCA1, BRCA2, PALB2, ATM, STK11 or the mismatch-repair genes (MSH2, MLH1, PMS2, MSH6).

134 5.4 Discussion The established PC susceptibility genes account for a minority of familial PC and their contribution to apparent sporadic PC is even less well defined. In our cohort of predominantly sporadic PC we found 73 pathogenic mutations in 32 cancer predisposition genes in 65 patients (16.6%). This includes 5.9% (23/392) of patients with a pathogenic mutation in a known PC risk gene. The current known PC predisposition genes were mutated in only a minority (2/28, 7.1%) of FPC probands in this study, which was not different from patients with apparent sporadic PC. The predominant mutated PC predisposition gene was BRCA2 in 2.3% (9/392) which is lower than previous reports in apparent sporadic115 (7%) and familial PC (6%)525. Unlike previous studies our cohort did not have any patients of Ashkenazi Jewish ancestry in whom founder BRCA2 mutations are present in nearly 11% of those who develop PC526. We only identified 1 (0.26%) pathogenic mutation in a mismatch repair genes, which is lower than recently reported (4/290, 1.4%)527. Germline BRCA1-2 and PALB2 mutations were associated with increased family history of breast and ovarian cancer. Overall when combined there was no association with germline mutation and personal or family history of malignancy or age at diagnosis. Absence of clinical history of malignancy is likely to become more common as testing thresholds decrease and can be explained in part by small family size, variable penetrance, incomplete pedigrees and for BRCA1 and BRCA2 male carriers. Recent studies in unselected breast and ovarian cancer probands found no personal or family history of malignancy and in up to one half of carriers of BRCA1 and BRCA2446,528. Only 7 mutation carriers met the current ACMG/NCCN guidelines for consideration of genetic testing for PC, meaning that only 3 BRCA2, 2 PALB2 and one each of CDKN2A and ATM mutations could have been identified using current guidelines473.

A further 50 pathogenic mutations were identified in 25 moderate-high penetrance cancer predisposition genes. Overall there was no association between these mutations and personal or family history of malignancy, although for some individual patients the clinical history was striking for example CHEK2 c.1299delC in a PC patient with previous bilateral breast cancer. In the majority of patients with these mutations bi-allelic inactivation was not evident or could

135 not be assessed due to low tumour cellularity. In addition, most are not associated with a specific mutational signature to support a role in the tumour. Furthermore, some of the variants reported as pathogenic on the basis of previous case reports in a patient with the relevant phenotype are likely hypomorphic variants and do not lead to complete abrogation of protein function e.g. TERT, RET and FANCL variants. The prevalence of individual mutations is beginning to be unraveled with large-scale sequencing projects, but the prevalence of pathogenic variants across the gene is not known. Enrichment of pathogenic variants in PC patients compared to aged healthy controls would support a role for these genes in PC predisposition. The role of these mutated genes in predisposition to PC and the clinical applicability of these genes are currently unknown and require further study. As previously described medically actionable germline mutations have direct clinical utility of a diagnostic or therapeutic nature and established clinical management guidelines.529 Although there is some variation, the list of actionable genes generally include MLH1, MSH6, MSH2, PMS2, APC, BRCA1, BRCA2, CDKN2A, PALB2, MUTYH (bi- allelic), VHL, MEN1, RET, NF2, STK11 and TP53. Of the pathogenic germline mutations found 20 (26.7% of mutations or 5.1% of patients) were clinically actionable. Other mutated genes may become actionable in the future as evidence grows.

The discovery of causal variants in PC presents several challenges and highlights current knowledge gaps. Some of the missing heritability is likely to result from moderate-low penetrant genes which have variable heritability with a high rate of phenocopy and are difficult to detect with traditional methods530,531,532. To overcome this, large case-control sequence cohorts will need to be established and compared to aged cancer-free populations. Other strategies such as integrating germline with paired tumour sequence data allows a more comprehensive picture of the involvement of germline variants in tumourigenesis. Another strategy to reduce ambient genetic noise is to prioritize genes established as predisposition genes in other cancers or with putative roles in cancer pathways. Multigene or panel testing of candidate cancer predisposition genes has highlighted critical knowledge gaps limiting their clinical applicability. Traditionally genetic testing in oncology has suffered from

136 significant selection bias with only those with the most extreme phenotypes undergoing genetic testing e.g. young age, multiple primary cancers or multiple affected kindreds, which enrich for deleterious mutations. The interpretation of deleterious variants in the absence of a relevant phenotype, in conjunction with counseling and cancer risk estimation is very difficult and currently there is little evidence to guide clinical decision making.

This study has several limitations. Firstly our candidate gene list was developed based on published cancer predisposition genes but this list is likely incomplete. Furthermore our study lacked a cancer-free control group, instead we used publically available germline variant databases to determine allele frequencies but these databases do not allow assessment of phenotype or co-inheritance of variants. We classified variants as pathogenic only if the nucleotide change was predicted to abrogate the transcript or if previous characterization determined loss of function. It is likely that some variants particularly missense variants we classified as not pathogenic do result in significant loss of function.

In this predominantly sporadic PC cohort the prevalence of pathogenic variants in known PC predisposition genes is 5.9%. The predominant mutated gene is BRCA2 which typically shows bi-allelic inactivation, and a neoplastic role is supported by genomic instability and dominant BRCA deficiency mutational signature in the tumour. The established PC predisposition genes account for only a small proportion of the familial PC in this cohort. Germline PC predisposition gene mutations were frequently found in patients with no personal or family history of malignancy, although BRCA1-2 and PALB2 mutations were associated with breast or ovarian cancer in kindred’s. In addition, 11.0% (43/392) of PC patients were found to have mutations previously reported as pathogenic in one of 25 cancer predisposition genes. The role of these genes in the predisposition to PC requires further evaluation.

137

Chapter 6: Discussion

138

The outlook for the majority of patients diagnosed with pancreatic cancer is poor with an incidence to mortality that approaches one. This has not changed significantly over the last 20 years, despite significant improvements in cross- sectional imaging, pancreatic tissue sampling and surgical and perioperative care. The fact remains that most people diagnosed with PC have advanced disease which is not amenable to resection, currently the only potential for cure. Even in those with resectable disease the majority succumb to metastatic PC due to occult disease present at diagnosis2,533. Decisions regarding the appropriateness of pancreatectomy are currently based purely on imaging criteria, with the major clinico-pathologic prognostic factors not determined until microscopic examination of the resection specimen. One would expect a sense of entrenched therapeutic nihilism but despite these poor figures there are signals of good response. Firstly, although long-term survivors are rare they are most likely to be those with resected small tumours confined to the pancreas with clear margins and uninvolved lymph nodes. Traditionally these have been rare finds but identification of individuals with higher genetic risk based on known mutation or family history and those with precursor lesions presents an opportunity for screening to detect and early intervention on lesions likely to progress to PC. Screening has been facilitated by improvements in imaging modalities such as endoscopic ultrasound which also allows tissue sampling and magnetic resonance imaging. Secondly, as we usher in the era of precision or personalised medicine and with advances in genomic sequencing it is increasingly recognised that druggable driver mutations are present in PC albeit individually at low frequency. Targeted therapies offer the potential for higher response rates than traditional chemotherapeutics as has been described with DNA damaging agents and PARP inhibitors in the setting of BRCA2 and PALB2 mutations534. The identification of druggable mutations and initiation of targeted therapy has seen reports of long-term survival even in the setting of metastatic disease, typically a stage of rapid progression535.

In this dissertation I have shown that CA19.9 the only clinically used biomarker for PC has limited ability to detect early disease, therefore its use in screening is

139 limited. Serial monitoring in the perioperative period does provide independent prognostic and predictive value.

In chapter 3 I have shown that FPC accounts for nearly 9% of PC. The age at presentation and outlook for patients with familial PC is similar to those with sporadic disease but there were some discriminating features. These include in FPC, a higher proportion of FPC kindred’s with extra-pancreatic malignancies, multi-focal disease and lower proportion of active smokers and cumulative tobacco exposure.

Hypermutated tumours have been described for some time such as MMR deficient cancers but these have been defined on the basis of mutability at 5 – 7 loci. Advances in DNA sequencing now allow the mutation burden across the entire genome or coding region to be characterised. This enables accurate detection of hypermutated tumours and with paired tumour-normal sequencing and mutation signature analysis the aetiology of this can de determined. In chapter 4, I show that 5% of PCs are hypermutated resulting predominantly from defects in DNA maintenance and repair. MMR deficiency was seen in 1% of PC as a result of somatic inactivation of key MMR members. Homologous recombination deficiency was the cause of hypermutation in 2.6% of PCs which typically also show genomic instability. HR deficiency resulted from a combination of germline and somatic truncating mutations most frequently in BRCA2. Hypermutated PC may benefit from combination therapy with immune checkpoint inhibitors.

The established PC susceptibility genes account for only a minority of the genetic contribution to FPC development. The hereditary component of apparent sporadic PC is even less well defined. Recent successes in the identification of PC susceptibility genes such as ATM and PALB2 are the result of establishment of large cohorts of individuals at increased risk for PC and technological advancements in genomic sequencing. In the predominantly sporadic PC cohort described in Chapter 5 I have shown that 5.9% (23/392) harboured a pathogenic mutation in a known PC risk gene. Only 7 mutation carriers met the current ACMG/NCCN guidelines for consideration of genetic

140 testing for PC, meaning that potentially 3 BRCA2, 2 PALB2 and one each of CDKN2A and ATM mutations would have been identified (www.nccn.org)473. A further 50 pathogenic mutations were identified in 25 moderate-high penetrance cancer predisposition genes in 42 patients. The role of these mutations in predisposition to PC is currently unknown and requires further study. Actionable genetic mutations have direct prognostic or therapeutic utility and/or evidence- based risk minimisation guidelines. Of the pathogenic germline mutations found 20 (26.7% of mutations or 5.1% of patients) were potentially clinically actionable. Other mutated genes may become actionable in the future as evidence grows.

The last 10-15 years has seen significant methodological advances in particular in massively parallel sequencing, lower cost and increasing access to genomic sequencing. This has allowed assessment of a number of well-known cancer predisposition genes in a broader context, compared to testing in patients with a clinically high risk of carrier status, raises significant questions concerning our understanding of the molecular pathology of PC predisposition. It also begins to paint a picture of the challenges ahead to our understanding of inherited cancer risk and the implications for current clinical approaches as genomic sequencing becomes broadly available. The challenges to the discovery of causal variants in PC include some of the missing heritability of PC is likely due to genes with low – moderate penetrance with variable heritability with a high rate of phenocopy530,531. Although rare these low-moderate penetrance alleles can be identified in individuals without the phenotype. Traditional methods of gene discovery such as linkage analysis and genome-wide association studies are designed to detect rare high-penetrance and common but low-penetrance alleles, respectively.532 These methods present challenges in PC often difficult due to the usually limited number of affected individuals, unavailability of both normal and tumour DNA from affected members due to short survival, variable often low penetrance of variants, locus heterogeneity i.e. mutation in one of several genes can lead to the same hereditary cancer phenotype.163 These deficiencies have partially been overcome with new techniques such as next generation sequencing of paired tumour-normal DNA from probands with or without unaffected kindreds. This method can generate a large number of

141 potential candidates and requires high fidelity filters to eliminate non-causal variants.

These include selecting variants with predicted severe consequences on the transcript such as nonsense, frameshift and essential splice site mutations. Additional support for pathogenicity can be obtained with examination of tumour DNA or protein expression such as if there is evidence of bi-allelic inactivation (DNA second hit or absent protein expression on IHC) or if the implicated variants produce a characteristic mutational signature in the tumour DNA.455 Furthermore selecting rare variants (typically minor allele frequency < 0.01) and eliminating those which are seen in publically available genomic databases such as the Single Nucleotide Polymorphism database (dbSNP), 1000 Genomes database and the National Heart Lung Blood Institute Exome Sequencing Project. Potential caveats with this approach include the inclusion of disease causing variants in the databases, furthermore phenotypic data is not recorded and they do not allow assessment of the coexistence of variants at an individual level530. Further complicating this is that we all carry multiple rare mutations which significantly impact the function of the protein some of which plausibly play a role in oncogenesis536. Other strategies such as integrating germline data with paired tumour sequence data allows a more comprehensive picture of the contribution of germline variants to tumourigenesis.

Furthermore, many causative mutations, including those that lead to pancreatic cancer, are likely to be of only moderate penetrance. Therefore, it is possible that these mutations may be observed in databases that are used for control data. Furthermore given the lack of strong linkage signals it is likely that rather than there being only a few genes that explain the remainder of the aggregation of pancreatic cancer in families, there are instead many genes, each of which only explains a small proportion of familial pancreatic cancer. Alternatively patients may have complex genotypes with multiple susceptibility variants which interact with each other and environmental factors to increase risk and can mimic autosomal dominant inheritance as has been shown in other diseases for example hereditary pancreatitis.537 Finally the presence of phenocopies (sporadic cancers that occur in kindreds with familial cancer) can lead to the

142 exclusion of disease-causing variants, as not all cancer cases in a kindred will necessarily carry the disease-causing mutation.

The identification of candidate genes with established roles in cancer development or in pathways with putative pathways in cancer is another method to reduce the ambient genetic noise. This has been successful in other tumour types such as ovarian cancer where 23% of women harboured a pathogenic mutation in one or more of 12 genes446. In triple-negative breast cancer a BRCA1 or BRCA2 mutation was identified in 11.2% and a further 3.7% had a mutation in 15 additional genes447. Multigene or panel testing of candidate cancer predisposition genes has highlighted critical knowledge gaps limiting their clinical applicability. Traditionally genetic testing in oncology has suffered from significant selection bias with only those with the most extreme phenotypes e.g. young age, multiple primary cancers or multiple affected kindreds, which enrich for deleterious mutations undergoing genetic testing. As the threshold for genetic testing The interpretation of deleterious variants and estimation of risk in the absence of a related phenotype or family history is unknown and there is little evidence to guide counseling and clinical decision making. Even in the presence of a potentially related phenotype it may be difficult to assign causality to a deleterious variant and additional evidence may be required538. It needs to be determined where we can achieve equipoise in the set and number of genes to maximize clinically relevant information and minimize ambient noise528.

The substantial diversity of the human genome and the complexity of cancer genomes infer that our traditional approach to identifying predisposition genes and quantifying relative risk will require even larger numbers. We will be challenged by unraveling the contribution of multiple loci, including combinations of different genes, co-existent variants within genes and gene- environment interactions. Moreover, we are only utilizing limited endpoints for cancer predisposition: increased incidence and young age of onset. We are yet to examine other endpoints such as survival. Most genes we have identified have either well-known large effects such as Li-Fraumeni Syndrome and TP53 mutations, or those that occur early in carcinogenesis such as APC mutations.

143 Inherited deleterious variants may not substantially lower the age of onset, or dramatically increase the incidence of a particular cancer, but may lead to a poor prognosis cancer since the initiating mutation is still environmentally determined, but “progressor” mutations may already be present.

To circumvent these hurdles, we need to identify other ways to gather the evidence required to impact on clinical management 538. We can only estimate the likely extremely large cohort sizes we will need for case control studies, notwithstanding the challenge of defining low-risk controls. The concept of healthy controls of advanced age may be helpful, and requires assessment, however, it is likely that only large-scale “knowledge bank” approaches that track generations over time with well documented clinical histories will begin to unravel this complexity. We also need to define the role of other measures such as functional readouts, or surrogate measures of the consequences of specific genomic variants, or constellations of variants to better understand cancer predisposition, and explore the potential direct clinical utility of such approaches. An example is using whole genome sequences to identify surrogates of genetic defects in tumours. These include microsatellite instability and mutational signatures associated with defects in DNA maintenance. The latter is a specific signature of point mutations that are associated with defects in BRCA1 and BRCA2 function 539. Variants associated with such surrogate measures can then focus experimental approaches to demonstrate the functional significance of these variants.

As we accumulate more cancer genomes through large-scale international efforts such as the International Cancer Genome Consortium (ICGC)412 and The Cancer Genome Atlas (TCGA)239,540, the germline sequences that accompany these genomes will provide insights into the prevalence of known predisposition loci in the germ line, and perhaps point to novel candidates 541. These approaches will define hypotheses that could be tested using different methodologies and begin to define a framework for expanding the way we approach cancer predisposition. Other insights may be gleaned from a better understanding of precursor lesions, the mutations that are present within them, in particular, surrogate measures such as point mutation signatures that could

144 infer mechanisms operating during early carcinogenesis. Familial tumour registries such as the National Familial Pancreatic Cancer Tumour Registry at Johns Hopkins University 163 with detailed data and biospecimen acquisition provide an important resource for identification of candidate risk genes, clustering of related tumour types, the estimation of risk and the assessment of early detection strategies. Follow up and biospecimen acquisition (germ line DNA, and where appropriate, tumour DNA) of patients and their families for index cases with variants of unknown significance may also bear fruit in the longer term.

Summary and Conclusion In this thesis I have shown that CA19.9 the only clinically used biomarker at present has prognostic and predictive value in patients with resected PC. Its characteristics however limit its use in the detection of early disease in both the general population and high risk individuals. Secondly, in our cohort FPC accounts for a significant proportion of PC cases. FPC kindreds were twice as likely as SPC kindreds to have an extra-pancreatic malignancy. The assessment of EPM and incorporation into the risk prediction may assist in the identification of high risk individuals, the target organs at risk and allow a more precise estimation of risk. Thirdly, somatic hypermutation is found in 5% of PC with the tumours with the highest mutation burden displaying evidence of MMR deficiency. The mechanism of MMR deficiency was predominantly driven by somatic inactivation in the MMR genes. Only one pathogenic germline MMR mutation was identified but the tumour did not show MMR deficiency. Lynch syndrome did not contribute to the development of PC in this cohort. Most of the remaining hypermutated tumours show evidence of homologous recombination deficiency or a T>G at TT sites mutation signature seen in oesophageal cancer but of uncertain aetiology. Lastly, in a large cohort of predominantly sporadic PC I have shown that 5.9% of patients carry a pathogenic germline mutation in an established PC predisposition gene. A further 10.7% harbour a pathogenic mutation in an additional cancer predisposition gene. A minority (3/23) of those with a PC predisposition gene mutation met the current criteria for genetic testing. Approximately 5% of the pathogenic germline mutations identified

145 were clinically actionable. Further evidence is needed to determine the role of these other cancer predisposition genes in the development of PC.

146

147

Appendix

148 Chapter 2: Supplementary data In the resected subset (Table 1) there were 265 men (48.5%), the mean age at diagnosis was 65.6 years (range 28 to 87), with a median follow-up of 12.7 months (range, 0 to 169). One hundred and twenty-eight (23.4%) were alive at the census date; August 4, 2010. The 30-day mortality was 2.2%. Three hundred and seventy-six (68.9%) patients died from pancreatic cancer, and 26 (4.8%) of other causes. The median disease-specific survival was 16.8 months, with 3- and 5-year survival rates of 16.5 % and 7.0 % respectively. The majority of tumors (63.2%) were moderately differentiated, 26% were poorly differentiated, and only 9.3% of tumors were well differentiated. Most tumors were located in the head of the pancreas (81.9%) and were more than 20 mm in maximal diameter (77%). The most frequent T stage was T3 (78.2%) followed by T2, T1 and T4 in decreasing order (13.9%, 6.6% and 0.2% respectively). In terms of AJCC stage, the majority was stage IIB (59.3%) followed by IIA (26.7%) then I (9.0%) then IV (4.9%). Three hundred and seventeen patients (58.1%) had resections with clear surgical margins. Lymph node metastases were present in 348 (63.7%), perineural invasion in 391 (71.6%), and vascular space invasion in 226 patients (41.4%).

All Patients Resected Biopsy

Parameter n = 904 Median P value n = 546 Median P value n = 358 Median P value No. (%) DSS (mo) (logran No. (%) DSS (logran No. (%) DSS (logrank) k) (mo) k) (mo) Sex Female 434 (48.0) 265 (48.5) 169 (47.2) Male 470 (52.0) 281 (51.5) 189 (52.8) Age, y Mean 65.9 65.6 66.4 Median 67.0 67.0 68.0 Range 28-92 28-87 36-92 Specimen Resection 546 (60.4) 16.8 <0.0001 546 16.8 - Biopsy 350 (38.7) 6.9 - - 350 (97.8) 6.9 Post-mortem 8 (0.9) - - - 8 (2.2) Outcome Follow-up (mo) 0 - 242.5 0 - 230.6 0 – 242.5 Median follow-up 9.4 12.7 5.1

149 30-day mortality 13 (1.4) 12 (2.2) 1 (0.3) Death PC 669 (74.0) 376 (68.9) 293 (81.8) Death other 36 (4.0) 26 (4.8) 10 (2.8) Alive 180 (19.9) 128 (23.4) 52 (14.5) Lost to follow-up 7 (0.8) 4 (0.7) 3 (0.8) Stagea IA 26 (2.9) 25.2 21 (3.8) 27.9 5 (1.4) 6.3 IB 29 (3.2) 26.5 28 (5.1) 26.5 1 (0.3) - IIA 213 (23.6) 15.5 146 (26.7) 19.6 67 (18.7) 8.9 IIB 345 (38.2) 15.6 0.0070b 324 (59.3) 16.0 0.0058b 21 (5.9) 4.7 0.1058b III 114 (12.6) 7.7 0 - 114 (31.8) 7.7 IV 134 (14.8) 5.7 <0.0001 27 (4.9) 10.3 <0.0001 107 (29.9) 4.6 0.0207c Missing 43 (4.8) c 0 c 43 (12.0) Differentiationd Well 85 (9.4) 17.4 51 (9.3) 26.6 34 (9.5) 12.9 Moderate 475 (52.5) 12.5 345 (63.2) 16.7 130 (36.3) 11.1 Poor 254 (28.1) 9.3 0.0005 144 (26.4) 16.2 0.2787 110 (30.7) 4.3 <0.0001 Missing 90 (10.0) 6 (1.1) 84 (24.5) Tumor locatione Head 447 (81.9) 18.1 Body/tail 99 (18.1) 11.6 0.0007 Tumor sizef  20mm 99 (18.1) 31.0 <0.0001 > 20mm 342 (62.6) 15.6 Missing 105 (19.2) Margins Clear 317 (58.1) 21.2 <0.0001 Involved 221 (40.5) 12.8 Missing 8 (1.5) Lymph nodes Negative 197 (36.1) 20.4 0.0006 Positive 348 (63.7) 15.6 Missing 1 (0.2) Perineural invasion 125 (22.9) 22.0 0.0284 Negative 391 (71.6) 16.0 Positive 30 (5.5) Missing Vascular invasion Negative 241 (44.1) 19.6 0.0018 Positive 226 (41.4) 15.6 Missing 79 (14.5) Chemotherapy No therapy 244 (27.0) 9.9 162 (29.7) 13.5 82 (22.9) 4.5 Any therapy 317 (58.1) 16.0 0.0018 229 (41.9) 21.6 0.1560 88 (24.6) 8.7 0.0012 Missing 343 (37.9) 155 (28.4) 188 (52.5) Chemotherapy No adjuvant 444 (49.1) 11.1 278 (50.9) 15.6 0.0110 - Adjuvant 166 (18.4) 20.4 <0.0001 166 (30.4) 20.4 - Missing 294 (32.5) 102 (18.7) - Chemotherapy No adj or < 3 572 (63.3) 10.3 327 (59.9) 15.6 - cycles 102 (18.7) 25.3 0.0003 - Adjuvant ( 3 102 (18.7) 26.0 <0.0001 117 (21.4) - cycles) Missing 230 (42.1) Radiotherapy No radiotherapy 795 (87.9) 11.3 467 (85.5) 16.7 328 (91.6) 6.6 Any radiotherapyg 109 (12.1) 13.6 0.3340 79 (14.5) 18.0 0.5476 30 (8.4) 9.6 0.2041 No adjuvant 493 (90.3) 16.8 Adjuvant 53 (9.7) 16.9 0.5216 No palliative 519 (95.1) 16.7 Palliative 27 (4.9) 20.0 0.9718 CA 19.9 Pre-resection (or N = 239 N = 202 N = 37 biopsy) CA19.9 0.2 0.4 0.1 Median time pre, N = 268 N = 231 N = 37 (mo) N = 117 N = 131 N = 29 Post-resection CA 1.2 1.2 1.1 19.9 N = 107 N = 88 N = 21 Post-resection(or biopsy) < 3/12 15 10 5 Median time post < 3/12, mo Pre and post < 3/12 150 CA 19.9 < 5 U/mL (and post not > 5)

Supplementary Table 1: Clinico-pathological parameters and outcome (n = 904) a Staging based on AJCC TNM Staging System 7th Edition, 2010 b Stage I tumors Vs. Stage II for survival analysis c Stage I and 2 tumors Vs. Stage III and IV for survival analysis d Well, moderately and poorly differentiated tumors analysed individually for survival analysis. e Patients with tumors located in the head of the pancreas underwent Whipple pancreaticoduodenectomies, and those with tumors of the body/tail had left sided pancreatectomies. f Tumor size was also prognostic > 30 mm (p = 0.0015), and > 40 mm (p < 0.0001). g Analysis compares those patients who received radiotherapy at any time to all others.

Pre-resection Post- Pre and Post-resection CA 19.9 resection CA CA 19.9 <3/12 19.9 <3/12 Patient No. =202 % No. = % No. = 88 % 131 Characteristics Sex Female 94 46.5 59 45.0 39 Male 108 53.5 72 55.0 49 Age (years) Mean 65.8 65.2 64.8 Median 67.0 66.0 65.5 Range 28 - 87 30 - 87 43 - 87 Bilirubin, mg/dL Mean 6.0 0.9 - Median 2.0 0.5 - Range 0.6 – 0.1 – 8.5 - 49.3 Non-expressors N = 10 5.0 N = 9 6.9 N = 3 Normal bilirubin N = 111 55.0 N = 123 93.9 Pre Post CA 19.9, U/mL N = 48 N = 85 54.5/96.6 Mean 1188.6 237.1 500.2 280.8 Median 138.0 39.0 186.0 39.0 Range 1– 1 - 6150 1 - 6000 1 - 6150 26,600 Hyperbilirubinaemia N = 91 45.0 N = 8 6.1 N = 40 N = 3 45.5/3.4 CA 19.9 Mean 2415.4 121.0 1267.2 223.0 Median 351.0 54 392.0 43.0 Range 1 – 1 – 625 1 – 20000 1 - 625 Corrected CA 19.9 101,075 Mean 34.9 141.7 62.1 Median 347.2 10.0 31.8 7.2 Range 28.3 1 – 178 0.1 – 1666 1 – 178 0.1 – 16,780 T stage T1 10 5.0 11 8.4 5 5.7 T2 23 11.4 9 6.9 7 8.0 T3 168 83.2 110 84.0 75 85.2 T4 1 0.5 1 0.8 1 1.1 N stage N0 70 34.7 42 32.1 29 33.0 N1 132 65.3 89 67.9 59 67.0 151 AJCC stage IA 4 2.0 5 3.8 2 2.3 IB 10 5.0 4 3.1 4 4.5 IIA 54 26.7 31 23.7 23 26.1 IIB 126 62.4 81 61.8 55 62.5 III 0 0 0 IV 8 4.0 10 7.6 4 4.5 Tumor size  20mm 28 13.9 24 18.3 11 12.5 > 20mm 134 66.3 80 61.1 61 69.3 Missing 40 19.8 27 20.6 16 18.2 Differentiation Well/Moderate 146 72.3 96 73.3 68 77.3 Poor 54 26.7 34 26.0 19 21.6 Missing 2 1.0 1 0.8 1 1.1 Margins Clear 122 60.4 76 58.0 52 59.1 Involved 80 39.6 55 42.0 36 40.9 Perineural Invasion Absent 40 19.8 25 19.1 14 15.9 Present 154 76.2 101 77.1 70 79.5 Missing 8 4.0 5 3.8 4 4.5 Vascular Invasion Absent 82 40.6 50 38.2 30 34.1 Present 91 45.0 59 45.0 42 47.7 Missing 29 14.4 22 16.8 16 18.2 Recurrence 125 61.9 84 64.1 54 61.4 R0 resection - Yes 69 55.2 46 54.8 29 53.7 - No 56 44.8 38 45.2 25 46.3 Duration to recurrence N = 81 64.8 N = 59 70.2 N = 39 72.2 Mean 12.0 12.1 10.4 Median 8.7 9.0 7.3 Range 0.4 – 0.4 – 0.4 – 56.8 Missing recurrence 56.8 35.2 56.8 29.8 15 27.8 date 44 25 Adjuvant Therapy Chemotherapy 85 42.1 65 49.6 48 54.5 Chemotherapy  3 59 29.2 49 37.4 32 36.4 cycles 17 8.4 18 13.7 10 11.4 Radiotherapy Supplementary Table 2: Descriptive Statistics for the cohort who contributed CA 19.9 values at each time-point

Bilirubin <2mg/dL Stage Number Median CA 19.9 U/mL (range) I 9 76.0 (17 – 6000) II 92 169.0 (2 – 26,600) IV 3 200.0 (33 – 308) Bilirubin ≥ 2mg/dL Stage Number Median cCA 19.9 U/mL (range) I 3 11.2 (10.7 – 12.7) II 80 35.7 (0.2 – 16780.4) IV 5 31.7 (1.9 – 263.5) Combined Stage Number Median CA 19.9 U/mL (range) I 12 57.0 (10.7 – 6000) II 172 98.3 (0.2 – 26600) IV 8 35.5 (1.9 – 308) Supplementary Table 3: Pre-resection CA 19.9 and Pathologic Stage (non- expressors excluded)

152

Post-resection CA No. of Patients Median DSS (mo) P value 19.9 N = 122 < 37 57 26.0 > 37 65 14.8 0.0017 < 50 65 25.6 > 50 57 16.0 0.0034 < 90 83 25.2 > 90 39 16.0 0.0144 < 120 85 25.2 > 120 37 14.8 0.0020 < 180 97 22.4 > 180 25 14.8 0.0116 Supplementary Table 4: Post-resection CA 19.9 < 3 months and disease- specific survival

Post-resection CA No. of patients Median time to P value 19.9 N = 63 recurrence (logrank) < 37 31 22.0 > 37 32 11.5 0.0194 < 50 34 22.0 > 50 29 7.7 0.0049 < 90 41 20.6 > 90 22 11.5 0.0251 < 120 43 20.6 > 120 20 7.2 0.0299 < 180 48 19.5 > 180 15 7.2 0.0274 Supplementary Table 5: Post-resection CA 19.9 and disease-free survival in patients with an R0 resection

Decrease CA 19.9 (%) No. of Patients Median DSS P value N = 85 (months) Any decrease or remains 60 22.4 <37 Any increase 25 13.3 0.0017 Pre and post CA 19.9 < 37 38 26.0 Pre and/or post CA 19.9 > 47 14.8 0.0126 37 < 10 25 13.3 > 10 60 22.4 0.0017 < 20 26 13.3 > 20 59 26.0 0.0017 < 30 26 13.3 > 30 59 26.0 0.0017 < 40 27 13.5 > 40 58 22.4 0.0126 < 50 28 13.5 > 50 57 22.4 0.0237 < 75 35 13.5 > 75 50 22.4 0.0180 < 90 40 13.5 > 90 45 22.4 0.0177 Supplementary Table 6: Change in cCA 19.9 value after resection

153

Pre-resection CA 19.9 No. of Patients Median Survival P value Normal bilirubin (CA 19.9) N = 111 Non-expressors 7 < 37 15 35.6 > 37 89 20.4 0.5952 < 90 36 29.9 > 90 68 20.0 0.3542 < 120 42 35.6 > 120 62 17.4 0.0444 < 180 53 20.4 > 180 51 25.7 0.9029 < 1000 86 20.4 > 1000 18 25.7 0.7247 Hyperbilirubinaemia (cCA N = 91 19.9) Non-expressors 3 < 37 47 19.4 > 37 41 16.7 0.0686 < 90 62 19.8 > 90 26 16.0 0.0073 < 120 65 19.4 > 120 23 16.2 0.0255 < 180 72 17.8 > 180 16 16.7 0.1138 Combined (CA 19.9 + c- N = 202 CA 19.9) Non-expressors 10 18.6 < 37 62 19.8 > 37 130 17.8 0.2137 < 90 98 22.2 > 90 94 16.7 0.0474 < 120 107 22.2 > 120 85 16.7 0.0058 < 180 125 18.5 > 180 67 19.7 0.5028 < 1000 169 18.3 > 1000 23 21.0 0.6320 Supplementary Table 7: Pre-resection CA 19.9 and survival

Pre-adjuvant CA 19.9 No. of Patients Median DSS P value < 90 48 26.0 > 90 19 16.2 0.0190 < 120 49 26.0 > 120 18 16.2 0.0063 < 180 52 25.2 > 180 15 16.0 0.0095 Supplementary Table 8: Pre-adjuvant CA 19.9 and benefit from adjuvant chemotherapy (49 of 67 completed  3 cycles)

Chapter 3: Supplementary data In resected FPC patients there were 29 men (50.9%), the mean age at diagnosis was 65.8 years (range 44 to 82), with a median follow-up of 16.0 months (range 3.4 to 110). At the census date, 23 (40.4%) patients were alive.

154 Thirty three (57.9%) patients died from pancreatic cancer, and 1 (1.8%) of other causes. The median disease-specific survival was 19.8 months, with 3- and 5- year survival rates of 15.8% and 5.3 % respectively. In resected FPC patients the only clinico-pathologic factors associated with significantly better survival on univariate analysis (Supplementary Table 1) included tumors of the pancreatic head (median survival 23.7 Vs. 9.4 months; P = .0308) compared with those of the body/tail and adjuvant chemotherapy ≥ 3 cycles (51.0 Vs. 10.1 months; P = 0.0146).

In resected SPC patients there were 282 men (52.2%), the mean age at diagnosis was 66.0 years (range 28 to 90), with a median follow-up of 15.6 months (range 0.0 to 195.8). At the census date, 98 (18.1%) patients were alive. The 30-day mortality was 2.4%. Four hundred and three (74.5%) patients died from pancreatic cancer, and 25 (4.6%) of other causes. The median disease-specific survival was 17.4 months, with 3- and 5-year survival rates of 16.7% and 5.9% respectively. In resected SPC patients, the clinico-pathologic factors associated with significantly better survival on univariate analysis (Supplementary Table 1) included tumors of the pancreatic head (median survival 18.4 Vs. 13.9 months; P = .0190) compared with those of the body/tail, tumor size ≤ 20 mm (26.0 Vs. 16.2 months; P < .0001), absence of margin involvement (21.0 Vs. 12.7 months; P = < .0001), absence of lymph node metastases (21.7 Vs. 16.3 months; P = .0003), and adjuvant chemotherapy (22.8 Vs. 12.0 months; P = .0018).

In FPC and SPC the majority of tumors were moderately differentiated and there was no difference in the proportion of poorly differentiated tumors (39.2% vs. 27.0%, P = 0.2841). Most tumors were located in the head of the pancreas (82.5% vs. 82.0%, P = 0.9375) and were more than 20 mm in maximal diameter (80.8% vs. 78.1%, P = 0.6547). There was no difference in the rate of R0 resection between groups (68.4% vs. 61.3%, P = 0.4312). The distribution of lymph node metastases (67.3% vs. 64.9, P = 0.9106), perineural invasion (77.1% vs. 75.6%, P = 0.4059), and vascular space invasion (50.0% vs. 50.8%, P = 0.6987) was also similar in both groups. FPC patients diagnosed prior to

155 2004 were more likely than SPC patients to receive any adjuvant chemotherapy (61.5% vs. 37.0%, P = 0.0417) but this was not seen after 2004.

FPC SPC

N = 57 Median P value N = 540 Median P value Variable No. (%) DSS (mo) (logrank) No. (%) DSS (mo) (logrank)

Sex Male 29 (50.9) 19.8 282 (52.2) 18.3 Female 28 (49.1) 29.8 0.4377 258 (47.8) 16.7 0.8331 Age (years) Mean 65.8 66.0 Median 66.0 67.0 Range 44 - 82 28 - 90 Outcome Follow-up (months) 3.4 – 110.1 0.0 – 195.8 Median follow-up 16.0 15.6 Death PC 33 (57.9) 403 (74.6) Death other 1 (1.8) 25 (4.6) Surgical death 0 13 (2.4) Alive 23 (40.4) 98 (18.1) Lost to Follow-up 0 2 (0.4) Stage I 7 (12.3) ---- 45 (8.3) 27.9 II 48 (84.2) 19.8 474 (87.8) 16.9 IV 2 (3.5) 10.1 0.2149 21 (3.9) 12.0 <0.0001 T Stagea T1 2 (3.5) 19.8 33 (6.1) T2 10 (17.5) 51.0 76 (14.1) 20.4 T3 45 (78.9) 18.9 0.2868 430 (79.6) T4 0 ---- 1 (0.2) 16.9 0.1425 N Stage N0 19 (33.3) 15.1 184 (34.1) 21.7 N1 38 (66.7) 23.7 0.8783 356 (65.9) 16.3 0.0003 Differentiationb Well 2 (3.5) 49 (9.1) Moderate 35 (61.4) 19.8 342 (63.3) 18.3 Poor 19 (33.3) 19.8 0.2680 146 (27.0) 16.3 0.0672 Missing 1 (1.9) 3 (0.6) Tumour location Head 47 (82.5) 23.7 443 (82.0) 18.4 Body/Tail 10 (17.5) 9.4 0.0308 97 (18.0) 13.0 0.0190 Tumour size ≤ 20mm 7 (12.3) 51.0 113 (20.9) 26.0 > 20mm 50 (87.7) 18.9 0.3137 427 (79.1) 16.2 <0.0001 Margins Clear 39 (68.4) 19.8 341 (63.1) 21.0 Involved 18 (31.6) 29.8 0.7723 199 (36.9) 12.7 <0.0001 Perineural Invasion Negative 10 (17.5) ---- 125 (23.1) 25.7 Positive 43 (75.4) 19.8 0.1077 397 (73.5) 16.3 0.0037 Missing 4 (7.0) 18 (3.3) Vascular Invasion

156 Negative 21 (36.8) 23.7 241 (44.6) 20.7 Positive 26 (45.6) 19.8 0.5937 265 (49.1) 16.2 0.0006 Missing 10 (17.5) 34 (6.3) Chemotherapyc Before 2004 N = 15 N = 290 No Adjuvant 8 (53.3) 9.0 222 (76.6) 15.1 Any Adjuvant 7 (46.7) 9.4 0.5290 68 (23.4) 18.8 0.1997 No or Adjuvant < 3 10 (66.7) 9.4 256 (88.3) 15.0 cycles Adjuvant  3 cycles 5 (33.3) 9.3 0.4157 34 (11.7) 25.0 0.0345 After 2004 n = 42 n = 250 No Adjuvant 11 (26.2) 16.7 90 (35.9) 12.0 Any Adjuvant 31 (73.8) 51.0 0.0559 156 (62.2) 22.8 0.0018 No or Adjuvant < 3 13 (31.0) 10.1 109 (43.4) 12.5 cycles Adjuvant  3 cycles 29 (69.0) 51.0 0.0146 137 (54.6) 24.9 0.0002 Supplementary Table 1: Clinico-pathological variables in resected PC patients a T1/2 Vs. T3/4 for survival analyses based on AJCC TNM Staging System. b Well/Moderate vs. Poor for survival analysis. --- Median survival not reached. c Prior to 2004 adjuvant chemotherapy was not the standard of care in Australia

FPC SPC

N = 11 Median P value N = 158 Median P value Variable No. (%) DSS (mo) (logrank) No. (%) DSS (mo) (logrank)

Sex Male 6 (54.5) 6.0 96 (60.8) 7.9 Female 5 (45.5) 7.6 0.6869 62 (39.2) 5.0 0.0778 Age (years) Mean 58.5 66.5 Median 55.5 68.0 Range 40 - 81 36 - 92 Outcome Follow-up (months) 3.4 – 16.2 0.1 – 35.0 Median follow-up 7.2 6.4 Death PC 11 (100) 147 (93.0) Death other 0 4 (2.5) Alive 0 7 (4.4) Lost to Follow-up 0 0 Clinical Stage II 2 (18.2) 3.4 44 (27.8) 8.9 III 3 (27.3) 5.4 56 (35.4) 8.0 IV 6 (54.5) 7.2 0.8368 56 (35.4) 4.7 0.0060 Missing 2 (1.3) Tumour location Head 9 (81.8) 5.4 122 (77.2) 6.8 Body/Tail 2 (18.2) 8.4 0.4983 36 (22.8) 6.8 0.7682 Chemotherapy No Palliative 5 (45.5) 4.4 105 (66.5) 5.0 Palliative chemo 6 (54.5) 8.4 0.0224 53 (33.5) 8.9 0.0211

157 Supplementary Table 2: Clinico-pathological variables in non-resected PC patients

Variable Resected PC Non-resected PC Combined FPC SPC P value FPC SPC P value FPC SPC P value Previous 9/57 59/540 0.4794 1/11 13/158 0.9200 10/68 72/698 0.2636 malignancy (15.8) (10.9) (14.7) (10.3) EPM site n = 11 n = 76 - breast F 2/28 (7.1) 20/258 0.9085 0/5 3/62 (4.8) 0.6835 2/33 (6.1) 23/320 (7.2) 0.8595 M 0/29 (7.8) N/A 0/6 0/96 N/A 0/35 (0) 0/378 (0) N/A 0/282 - colo-rectal 3/57 (5.3) 9/540 (1.7) 0.0658 1/11 (9.1) 4/158 (2.5) 0.2145 4/68 (5.9) 13/698 (1.9) 0.0558 - prostate M 2/29 (6.9) 14/282 0.6538 0/6 0/96 N/A 2/35 (5.7) 14/378 (3.7) 0.5554 (5.0) - bladder 1/57 (1.8) 0/540 0.0955 0/11 3/158 (1.9) 0.6447 1/68 (1.5) 3/698 (0.4) 0.2557 - endometrial F 0/28 2/258 (0.8) 0.6401 0/5 0/62 N/A 0/33 2/320 (0.6) 0.6488 - melanoma 2/57 (3.5) 12/540 0.5416 0/11 0/158 N/A 2/68 (2.9) 12/698 (1.7) 0.4727 (2.2) - gastric 0/57 1/540 (0.2) 0.7451 0/11 0/158 N/A 0/68 1/698 (0.1) 0.7548 - lung 0/57 2/540 (0.4) 0.6453 0/11 0/158 N/A 0/68 2/698 (0.3) 0.6453 - testicular germ cell M 0/29 2/282 (0.7) 0.6491 0/6 0/158 N/A 0/35 2/378 (0.5) 0.6662 - renal 0/57 0/540 N/A 0/11 2/158 (1.3) 0.7074 0/68 2/698 (0.3) 0.6585 - lymphoma 0/57 0/540 N/A 0/11 1/158 (0.6) 0.7913 0/68 1/698 (0.1) 0.7548 Supplementary Table 3: Previous extra-pancreatic malignancy

158

EPM site FPC SPC P value N = 68 (%) N = 698 (%) - breast 7/68 (10.3) 34/698 (4.9) 0.0579 - colo-rectal 7/68 (10.3) 44/698 (6.3) 0.2077 - prostate 3/68 (4.4) 16/698 (2.3) 0.2834 - endometrial 2/68 (2.9) 4/698 (0.6) 0.0345 - ovarian 1/68 (1.5) 6/698 (0.9) 0.6133 - melanoma 6/68 (8.8) 4/698 (0.6) <0.0001 - gastric 3/68 (4.4) 12/698 (1.7) 0.1261 - lung 4/68 (5.9) 26/698 (3.7) 0.3813 - Unknown primary 3/68 (4.4) 15/698 (2.1) 0.2397 - Head and neck 0/68 5/698 (0.7) 0.4838 - brain 2/68 (2.9) 5/698 (0.7) 0.1219 - hepatocellular 2/68 (2.9) 3/698 (0.4) 0.0652 - renal tract 1/68 (1.5) 2/698 (0.3) 0.2437 - lymphoma 1/68 (1.5) 1/698 (0.1) 0.1698 - oesophageal 1/68 (1.5) 4/698 (0.6) 0.3803 - myeloma/leukaemia 0/68 3/698 (0.4) >0.9999 - sarcoma 0/68 2/698 (0.3) >0.9999 - gallbladder 0/68 1/698 (0.1) >0.9999 - testicular 1/68 (1.5) 0/698 0.0888 TOTAL with EPM in 1 FDR 29/68 148/698 <0.0001 (42.6) (21.2) Supplementary Table 4: Distribution of EPM in FDRs

159

Variable Resected PC Non-resected PC Combined FPC SPC (%) P FPC SPC P FPC SPC P (%) value value value Diabetes Mellitus 17/57 145/540 0.6312 2/11 57/158 0.2287 19/68 202/698 0.8623 (29.8) (26.9) (18.2) (36.1) (27.9) (28.9) DM  2 years 7/17 42/145 0.6164 0/2 5/57 0.5788 7/19 47/202 0.5551 (41.2) (29.0) (8.8) (36.8) (23.3) DM > 2 years 5/17 41/145 0.7628 1/2 16/57 0.7774 6/19 57/202 0.9327 (29.4) (28.3) (50.0) (28.1) (31.6) (28.2) Missing date of 5/17 62/145 1/2 36/57 6/19 98/202 diagnosis (29.4) (42.8) (50.0) (63.2) (31.6) (48.5) Chronic Pancreatitis 4/57 27/540 0.5138 2/11 10/158 0.1389 6/68 37/698 0.2283 (7.0) (5.0) (18.2) (6.3) (8.8) (5.3) Alcohol Nil or Low Alcohol 48/57 438/540 0.3183 10/11 103/158 0.7230 58/68 541/698 0.3000 Intake ( 2 SD) (84.2) (81.1) (90.9) (65.2) (85.3) (77.5) Mod Alcohol Intake 5/57 50/540 0.9004 0/11 18/158 0.2069 5/68 68/698 0.4831 (3-4 SD) (8.8) (9.3) (11.4) (7.4) (9.7) Heavy Alcohol Intake 4/57 51/540 0.5443 1/11 20/158 0.6373 5/68 71/698 0.4216 ( 5 SD) (7.0) (9.4) (9.1) (12.7) (7.4) (10.2) Missing alcohol data 0/57 1/540 0 17/158 0 18/698 (0.2) (10.8) (2.6) Cigarette Smoking Never Smoked 32/57 265/540 0.3165 9/11 53/158 0.0035 41/68 318/698 0.0315 (56.1) (49.1) (81.8) (33.5) (60.3) (45.6) Reformed Smoker 19/57 133/540 0.1389 2/11 32/158 0.3007 21/68 165/698 0.3314 (33.3) (24.6) (18.2) (20.3) (30.9) (23.6) - reformed  10 years 4/19 39/133 0.9544 1/11 11/32 0.9551 5/21 50/165 0.9702 (21.1) (29.3) (9.1) (34.4) (23.8) (30.3) - reformed > 10 years 13/19 75/133 0.0700 1/11 9/32 0.7997 14/21 84/165 0.0627 (68.4) (56.4) (9.1) (28.1) (66.7) (50.9) - missing date ceased 2/19 19/133 0/11 12/32 2/21 31/165 (10.5) (14.3) (37.5) (9.5) (18.8) Active Smoker 6/57 142/540 0.0083 0/11 55/158 0.0052 6/68 197/698 0.0003 (10.5) (26.3) (34.8) (8.8) (28.2) Missing smoking 0 0 0 18/158 0 18/698 data (11.4) (2.6) Mean smoke 25.4 32.6 0.0833 30.0 42.8 0.5456 25.7 34.9 0.0479 exposure (pack- years) Supplementary Table 5: Risk factors for PC

160 Variable FPC SPC Age (mean) Number P value Age (mean) Number P value 1. All patients a. Resected 65.8 57 66.0 540 b. Non-resected 57.3 11 a, b: 0.0105 66.5 158 a, b: 0.5865 2. Smoking status a. Never smoked and 67.1 46 a, b: 0.8048 67.6 340 a,b: 0.0243 reformed  10 years b. Reformed < 10 years 66.0 4 b, c: 0.0898 63.9 39 b,c: 0.4490 c. Active 57.3 6 a, c: 0.0144 62.4 142 a, c: <0.0001 3. Tobacco exposure a. < 20 pack-years 67.8 6 65.9 50 b. > 20 pack years 64.7 18 a, b: 0.4470 64.8 187 a, b: 0.5213 4. Alcohol a. Nil 65.1 28 a, b: 0.5966 66.6 283 a, b: 0.6367 a, c: 0.5906 a, c: 0.3830 a, d: 0.8676 a, d: 0.0630 b. Low ( 2 SD/day) 66.7 20 b, c: 0.7834 66.1 155 b, c: 0.6146 c. Moderate (3 – 4 SD/day) 67.8 5 c, d: 0.4013 65.2 50 c, d: 0.4397 d. Heavy ( 5 SD/day) 64.3 4 b, d: 0.5875 63.7 51 b, d: 0.1631 5. Pancreatitis a. No 66.3 53 66.0 512 b. Yes 60.0 4 a, b: 0.1832 65.7 28 a, b: 0.8530 6. Diabetes Mellitus a. No 63.9 40 65.8 395 b. Yes 70.5 17 a, b: 0.0096 66.6 145 a, b: 0.3934 b.1 DM  2 years 68.0 7 a, b1: 0.2689 66.2 42 a, b1: 0.8079 b.2 DM > 2 years 71.0 5 a, b2: 0.0928 66.0 41 a, b2: 0.9300 b1, b2: 0.5185 a1, b2: 0.8940 Supplementary Table 6: Mean age at diagnosis in resected patients

161 ALL RESECTED Models Variable Hazard Ratio (95% CI) P value A. Resected PC – Initial Model Positive lymph nodes 1.72 (1.16 – 2.56) 0.0074 (n = 597) Size > 20 mm 1.63 (1.06 – 2.50) 0.0260 Poorly differentiated 1.14 (0.75 – 1.74) 0.5303 Vascular invasion 1.29 (0.90 – 1.87) 0.1703 Peri-neural invasion 1.42 (0.93 – 2.18) 0.1068 Involved margins 1.76 (1.21 – 2.58) 0.0034 Adjuvant chemotherapy 0.56 (0.39 – 0.82) 0.0026 Post-resection CA19.9 > 120 U/mL (≤ 3 months) 1.52 (1.05 – 2.19) 0.0275 Diabetes Mellitus > 2 years 1.69 (0.93 – 3.06) 0.0829 C. Resected PC – Resolved Positive lymph nodes 1.94 (1.34 – 2.81) 0.0005 Model (n = 597) Size > 20 mm 1.77 (1.18 – 2.67) 0.0061 Involved margins 1.84 (1.29 – 2.62) 0.0007 Adjuvant chemotherapy 0.62 (0.44 – 0.88) 0.0069 Post-resection CA19.9 > 120 U/mL (≤ 3 months) 1.74 (1.23 – 2.48) 0.0019 Diabetes Mellitus > 2 years 1.87 (1.08 – 3.23) 0.0246 Size > 20 mm 2.52 (1.15 – 5.52) 0.0209 Poorly differentiated 1.11 (0.53 – 2.31) 0.7793 Vascular invasion 1.09 (0.53 – 2.22) 0.8125 Peri-neural invasion 2.20 (0.94 – 5.12) 0.0693 B. Resected PC – Final Model Positive lymph nodes 1.83 (1.26 – 2.66) 0.0015 (n = 597) Size > 20 mm 1.72 (1.14 – 2.60) 0.0099 Involved margins 1.80 (1.26 – 2.57) 0.0013 Adjuvant chemotherapy 0.58 (0.41 – 0.83) 0.0027 Post-resection CA19.9 > 120 U/mL (≤ 3 months) 1.66 (1.16 – 2.36) 0.0054 Diabetes Mellitus > 2 years 1.81 (1.05 – 3.13) 0.0331 Supplementary Table 7: Multivariate analysis in FPC and SPC resected patients. Clinico-pathologic variables analyzed with a significant P value and those reported to be significant were entered into a Cox Proportional Hazard multivariate analysis and models resolved using backward elimination of redundant variables.

162

Supplementary Figure 1: Kaplan-Meier survival curves for: A. Disease-free survival in resected FPC and SPC patients, B. Survival in all resected patients stratified by diabetes mellitus (DM) status: no DM, DM duration < 2 years, DM duration > 2 years, C. Survival in resected FPC patients stratified by diabetes mellitus (DM) status: no DM, DM duration < 2 years, DM duration > 2 years, D. Survival in resected SPC patients stratified by diabetes mellitus (DM) status: no

163 DM, DM duration < 2 years, DM duration > 2 years, E. Survival in all resected patients stratified by smoking status: never smoked or reformed more than 10 years prior to PC diagnosis, reformed smoker of less than 10 years but more than 6 months from PC diagnosis, active smoker at time of PC diagnosis or ceased within 6 months of diagnosis.

164

Supplementary Figure 2: Kaplan-Meier survival curves for: A. Survival in resected FPC patients stratified by smoking status: never smoked or reformed more than 10 years prior to PC diagnosis, reformed smoker of less than 10 years but more than 6 months from PC diagnosis, active smoker at time of PC diagnosis or ceased within 6 months of diagnosis, B. Survival in resected SPC patients stratified by smoking status: never smoked or reformed more than 10 years prior to PC diagnosis, reformed smoker of less than 10 years but more than 6 months from PC diagnosis, active smoker at time of PC diagnosis or ceased within 6 months of diagnosis, C. Survival in all resected patients stratified by alcohol intake within 12 months of PC diagnosis: nil-low (0 – 2 165 standard drinks per day), moderate (3 – 4 standard drinks per day), heavy ( 5 standard drinks per day), D. Survival in resected FPC patients stratified by alcohol intake within 12 months of PC diagnosis: nil-low (0 – 2 standard drinks per day), moderate (3 – 4 standard drinks per day), heavy ( 5 standard drinks per day), E. Survival in resected SPC patients stratified by alcohol intake within 12 months of PC diagnosis: nil-low (0 – 2 standard drinks per day), moderate (3 – 4 standard drinks per day), heavy ( 5 standard drinks per day).

166 Chapter 4 Supplementary Data N = 380 Median DSS P value Variable No. (%) (mo) (logrank)

Sex Male 199 (52.4) 19 Female 181 (47.6) 27.4 0.0821 Age (years) Mean 66.4 Median 67 Range 33 – 90 Missing 69 Outcome Follow-up (months) 0 – 94.9 Median follow-up 15.1 Death PC 224 (58.9) Death other 9 (2.4) Surgical death (<30 days) 14 (3.7) Alive 128 (33.7) - without disease 73 (19.2) - with disease 42 (11.1) - status unknown 13 (3.4) Lost to Follow-up 5 (1.3) Histopathological subtype PDAC 334 (87.9) IPMN with invasion 18 (4.7) Adenosquamous 14 (3.7) Mucinous non-cystic carcinoma 6 (1.6) Acinar cell carcinoma 2 (0.5) Undifferentiated 4 (1.1) Small cell carcinoma 2 (0.5)

Mixed ductal endocrine 1 (0.3) carcinoma Stagea Ia 8 (2.1) Ib 13 (3.4) 26.8 IIa 70 (18.4) IIb 265 (67.4) 21.7 III 3 (0.8) 21.8 IV 14 (3.7) 14.3 0.0369 Missing 7 (1.8) T Stageb T1 9 (2.4) T2 35 (9.2) 26.8 T3 327 (86.1) T4 5 (1.3) 21.3 0.1056 Missing 4 (1.1) N Stage N0 96 (25.3) 29.7 N1 278 (73.2) 20 0.0245 Missing 6 (1.6) Differentiationc Well 20 (5.3) 167 Moderate 208 (54.7) 29.7 Poor 129 (33.9) Undifferentiated 13 (3.4) 14.9 <0.0001 Missing 10 (2.6) Tumor location Head 307 (80.8) 23.2 Body/Tail 70 (18.4) 16.9 0.0313 Missing 3 (0.8) Tumor size ≤ 20mm 46 (12.1) 30 > 20mm 323 (85.0) 19.3 0.0342 Missing 11 (2.9) Marginsd Clear 268 (70.5) 24.1 Microscopically involved 94 (24.7) 16.9 0.0036 Macroscopically involved 9 (2.4)

Not assessed 1 (0.3) Missing 8 (2.1) Perineural Invasion Negative 41 (10.8) 40 Positive 309 (81.3) 20 0.0028 Missing 30 (7.9) Vascular Invasion Negative 131 (34.5) 31.4 Positive 223 (58.7) 17.2 <0.0001 Missing 26 (6.8) Clinical and Family History Previous malignancy 54 (14.2) 1 prior malignancy 49 (12.9) 2 prior malignancies 5 (1.3) Previous CRC 8 (2.1) Previous endometrial 3 (0.8) cancer Family history of cancer Familial PC 27 (7.1) 1 FDR PC 25 (6.6) 2 FDR PC 2 (0.5) Family history of CRC in ≥1 26 (6.8) FDR 2 FDR + 1 SDR 1 2 FDR 3 1 FDR + 3 SDR 1 1 FDR 21 Family history of endometrial 4 (1.1) cancer 1 FDR 4 Supplementary table 1: Clinico-pathological features of resected PC patients

168 N = 12 Variable No. (%) Median DSS (mo) P value (logrank)

Sex Male 9 (54.5) 6 Female 3 (45.5) 7.6 0.6869 Age (years) Mean 60.4 Median 61 Range 42 - 76 Outcome Follow-up (months) 4.7 – 36.8 Median follow-up 6 Death PC 9 (75.0) Death other 0 Alive with disease 2 (16.7) Lost to Follow-up 1 Histopathological subtype PDAC 10 (83.3) Undifferentiated 2 (16.7) Clinical Stage III 5 (41.7) 15.6 IV 6 (50.0) 6 0.4386 Missing 1 Clinical and Family History Previous malignancy 0 Family history of cancer Familial PC 1 (8.3) Family history of CRC 0 Family history of endometrial 0 cancer Table 2: Clinico-pathological variables in non-resected PC patients

169 Chapter 5 Supplementary data

HGNC Name Cancer Phenotype Cancer Syndrome Reference Symbol Established PC predisposition genes

BRCA2 familial breast/ovarian breast, ovarian, Hereditary breast/ovarian OMIM 600185 cancer gene 2 prostate, pancreas, cancer other CDKN2A cyclin-dependent kinase melanoma, Familial atypical multiple OMIM 600160 inhibitor 2A pancreatic, oro- mole melanoma laryngeal (FAMMM) STK11 serine/threonine kinase jejunal harmartoma, Peutz-Jeghers syndrome OMIM 602216 11 gene (LKB1) ovarian, testicular, pancreatic PRSS1 protease, serine, 1 pancreas Autosomal dominant OMIM 276000 (trypsin 1) Hereditary pancreatitis ATM ataxia telangiectasia leukemia, Ataxia-telangiectasia OMIM 607585 mutated lymphoma, medulloblastoma, glioma, pancreas PALB2 partner and localizer of Wilms tumor, Fanconi anaemia N OMIM 610355 BRCA2 medulloblastoma, AML ,breast, pancreas APC adenomatous polyposis colorectal, Adenomatous polyposis OMIM 611731 of the colon gene pancreatic, coli; Turcot syndrome desmoid, hepatoblastoma, glioma, other CNS MSH2 mutS homolog 2 (E. coli) colorectal, Hereditary non-polyposis OMIM 609309 endometrial, ovarian colorectal cancer MLH1 E.coli MutL homolog colorectal, Hereditary non-polyposis OMIM 120436 gene endometrial, colorectal cancer, Turcot ovarian, CNS syndrome MSH6 mutS homolog 6 (E. coli) colorectal, Hereditary non-polyposis OMIM 600678 endometrial, ovarian colorectal cancer PMS2 PMS2 postmeiotic colorectal, Hereditary non-polyposis OMIM 600259 segregation increased 2 endometrial, colorectal cancer, Turcot (S. cerevisiae) ovarian, syndrome medulloblastoma, glioma BRCA1 familial breast/ovarian breast, ovarian, Hereditary breast/ovarian OMIM 113705 cancer gene 1 pancreas cancer TP53 tumor protein p53 breast, sarcoma, Li-Fraumeni syndrome OMIM 191170 adrenocortical carcinoma, glioma, pancreas, multiple other tumour types Other candidate cancer predisposition genes

Tumour suppressor genes ATR ataxia telangiectasia and oropharyngeal Autosomal dominant OMIM 601215 Rad3 related oropharyngeal cancer syndrome, Seckel syndrome, Ciutaneous 170 telangiectasia and cancer syndrome AXIN1 AXIS INHIBITOR 1 colorectal PUBMED adenomas 15520370 AXIN2 AXIS INHIBITOR 2; colorectal Oligodontia-colorectal OMIM 604025 cancer syndrome BABAM1 BRISC and BRCA1 A Breast OMIM 612766, complex member 1 PUBMED 19305427 BAP1 BRCA1 associated mesothelioma, uveal Tumour predisposition OMIM 603089 protein-1 (ubiquitin melanoma, skin syndrome carboxy-terminal tumours hydrolase) BARD1 BRCA1-associated RING ovarian, breast OMIM 601593 domain 1 BLM Bloom Syndrome leukemia, Bloom Syndrome OMIM 604610 lymphoma, skin squamous cell , other cancers BMPR1A bone morphogenetic gastrointestinal Juvenile polyposis OMIM 601299 protein receptor, type IA polyps BRIP1 BRCA1 interacting AML, leukemia, Fanconi anaemia J, OMIM 605882 protein C-terminal breast breast cancer helicase 1 susceptiblity BUB1B BUB1 budding rhabdomyosarcoma Mosaic variegated OMIM 602860 uninhibited by aneuploidy benzimidazoles 1 homolog beta (yeast) CDC73 CELL DIVISION CYCLE Parathyroid adenoma and carcinoma OMIM 607393 PROTEIN 73, S. CEREVISIAE, HOMOLOG OF CDH1 cadherin 1, type 1, E- gastric Familial gastric carcinoma OMIM 192090 cadherin (epithelial) (ECAD) CDKN1B cyclin-dependent kinase Similar to MEN1 MEN type IV OMIM 600778, inhibitor 1B (p27, Kip1) PUBMED 24429628 CEBPA CCAAT/ENHANCER- AML OMIM 116897 BINDING PROTEIN, ALPHA CHEK2 CHK2 checkpoint breast familial breast cancer OMIM 604373 homolog (S. pombe) CTNNA1 ALPHA-E-CATENIN gastric cancer ?Herditary diffuse gastric OMIM 116805, cancer PUBMED 23208944 CTR9 Paf1/RNA polymerase II Wilms tumour OMIM 609366, complex component PUBMED 25099282 CYLD cylindromatosis Cylindroma, Cylindromatosis, familial, OMIM 605018 trichoepitheliioma Brooke-Spiegler syndrome DDB2 damage-specific DNA skin basal cell, skin Xeroderma pigmentosum OMIM 600811 binding protein 2 squamous cell, (E) melanoma DICER1 dicer 1, ribonuclease type Pleuropulmonary blastoma, rhabdomyosarcoma OMIM 606241 III DIS3L2 DIS3 like 3'-5' Wilms tumour Perlman syndrome OMIM 614184 exoribonuclease 2

171 DKC1 dyskeratosis congenita 1, X-linked dyskeratosis OMIM 300126, dyskerin congenita PUBMED 24429628 ERCC2 excision repair cross- skin basal cell, skin Xeroderma pigmentosum OMIM 126340 complementing rodent squamous cell, (D) repair deficiency, melanoma complementation group 2 (xeroderma pigmentosum D) ERCC3 excision repair cross- skin basal cell, skin Xeroderma pigmentosum OMIM 133510 complementing rodent squamous cell, (B) repair deficiency, melanoma complementation group 3 (xeroderma pigmentosum group B complementing) ERCC4 excision repair cross- skin basal cell, skin Xeroderma pigmentosum OMIM 133520 complementing rodent squamous cell, (F), Fanconi anaemia Q repair deficiency, melanoma complementation group 4 ERCC5 excision repair cross- skin basal cell, skin Xeroderma pigmentosum OMIM 133530 complementing rodent squamous cell, (G) repair deficiency, melanoma complementation group 5 (xeroderma pigmentosum, complementation group G (Cockayne syndrome)) EXT1 EXOSTOSIN exostoses multiple exostoses type I OMIM 608177 GLYCOSYLTRANSFERA SE 1 EXT2 EXOSTOSIN exostoses multiple exostoses type 2 OMIM 608210 GLYCOSYLTRANSFERA SE 2 FAM175A family with sequence breast OMIM 611143, similarity 175, member A PUBMED 22357538 FANCA Fanconi anemia, AML, leukemia Fanconi anaemia A OMIM 607139 complementation group A FANCB Fanconi anemia, AML, leukemia Fanconi anaemia OMIM 300515 complementation group B FANCC Fanconi anemia, AML, leukemia, Fanconi anaemia C OMIM 613899 complementation group C breast FANCD2 Fanconi anemia, AML, leukemia Fanconi anaemia D2 OMIM 613984 complementation group D2 FANCE Fanconi anemia, AML, leukemia Fanconi anaemia E OMIM 613976 complementation group E FANCF Fanconi anemia, AML, leukemia Fanconi anaemia F OMIM 613897 complementation group F FANCG Fanconi anemia, AML, leukemia Fanconi anaemia G OMIM 602956 complementation group G FANCI Fanconi anemia, AML, leukemia Fanconi anaemia OMIM 611360 complementation group I FANCL Fanconi anemia, AML, leukemia Fanconi anaemia OMIM 608111 complementation group L FANCM Fanconi anemia, AML, leukemia Fanconi anaemia OMIM 609644 complementation group M

172 FH fumarate hydratase Leiomyomatosis and renal cell cancer OMIM 136850 FLCN folliculin colorectal, medullary Birt-Hogg-Dube syndrome OMIM 607273 thyroid, GATA2 GATA binding protein 2 leukaemia, myelodysplastic syndrome OMIM 137295 KLLN killin, p53-regulated DNA breast Cowden syndrome type 4 OMIM 612105 replication inhibitor (germline hypermethylation) LIG4 ligase IV, DNA, ATP- leukaemia LIG4 syndrome OMIM 601837 dependent MAP3K6 mitogen-activated protein gastric PUBMED kinase kinase kinase 6 25340522 MAX MYC-ASSOCIATED phaeochromocytom OMIM 154950 FACTOR X a MEN1 multiple endocrine parathyroid Multiple Endocrine OMIM 613733 neoplasia type 1 gene adenoma, pituitary Neoplasia Type 1 adenoma, pancreatic islet cell, carcinoid MITF microphthalmia- melanoma, renal Waardenburg syndrome OMIM 193510 associated transcription cell cancer 2A factor MLH3 mutL homolog 3 colrectal, Colorectal cancer, OMIM 604395 endometrial hereditary nonpolyposis, type 7 MRE11A meiotic recombination 11 ovarian, breast Ataxia-telangiectasia like OMIM 600814, homolog A (S. cerevisiae) disorder PUBMED 19383352 MTAP methylthioadenosine histiocytoma Diaphyseal medullary OMIM 156540, phosphorylase stenosis with malignant PUBMED fibrous histiocytoma 24429628 MUTYH mutY homolog (E. coli) colorectal, Adenomatous polyposis OMIM 604933 endometrial coli NBN Nijmegen breakage NHL, leukaemia, Nijmegen breakage OMIM 602667 syndrome 1 (nibrin) glioma, syndrome medulloblastoma, rhabdomyosarcoma NF1 neurofibromatosis type 1 neurofibroma, Neurofibromatosis type 1 OMIM 162200 gene glioma NF2 neurofibromatosis type 2 meningioma, Neurofibromatosis type 2 OMIM 607379 gene acoustic neuroma PHOX2B PAIRED-LIKE Neuroblastoma +/- Hirschsprungs disease OMIM 603851 HOMEOBOX 2B POLD1 polymerase (DNA colorectal, OMIM 174761 directed), delta 1, endometrial catalytic subunit POLE polymerase (DNA colorectal OMIM 174762 directed), epsilon, catalytic subunit POLH polymerase (DNA Xeroderma pigmentosum, OMIM 603968 directed), eta variant type POT1 protection of telomeres 1 melanoma, glioma OMIM 606478 PRKAR1A protein kinase, cAMP- atrial myxoma, Carney myxoma- OMIM 188830 dependent, regulatory, various endocrine endocrine complex type I, alpha tumours PTCH1 Homolog of Drosophila skin basal cell, Nevoid Basal Cell OMIM 601309 Patched gene medulloblastoma Carcinoma Syndrome PTEN phosphatase and tensin harmartoma, glioma, Cowden Syndrome, OMIM 601728 homolog gene prostate, Bannayan-Riley- endometrial Ruvalcaba syndrome 173 RAD50 RAD50 homolog (S. ovarian, breast NBS-like disorder OMIM 604040, cerevisiae) PUBMED 24894818 RAD51C RAD51 homolog C (S. ovarian, breast, Fanconi anaemia O OMIM 602774, cerevisiae) leukaemia PUBMED 20400963 RAD51D RAD51 homolog D (S. ovarian, breast Familial ovarian cancer OMIM 602954 cerevisiae) RAD54L RAD54-like (S. ?breast OMIM 603615 cerevisiae) RB1 retinoblastoma gene retinoblastoma, Familial retinoblastoma OMIM 614041 sarcoma, breast, small cell lung RECQL4 RecQ protein-like 4 osteosarcoma, skin Rothmund-Thompson OMIM 603780 basal and sqamous Syndrome, RAPADILINO cell syndrome, Baller-Gerold syndrome RINT1 RAD50 interactor 1 breast, Lynch spectrum cancers OMIM 610089, PUBMED 25050558 RPS20 RIBOSOMAL PROTEIN colorectal cancer PUBMED S20 24941021 RTEL1 regulator of telomere SCC mouth, Dyskeratosis congenita OMIM 608833, elongation helicase 1 nasopharynx, oesoph, rectum, vagina, Hodgkins disease, adenoCa GI tract, glioma RUNX1 RUNT-RELATED AML OMIM 151385 TRANSCRIPTION FACTOR 1 SBDS Shwachman-Bodian- leukaemia, Shwachman-Bodian- OMIM 607444 Diamond syndrome myelodysplastic Diamond syndrome syndrome SDHA succinate dehydrogenase pituitary adenoma, phaeochromocytoma, OMIM 600857 complex, subunit A, paraganglionoma, GIST flavoprotein (Fp) SDHAF2 chromosome 11 open paraganglioma Familial paraganglioma OMIM 613019 reading frame 79 SDHB succinate dehydrogenase paraganglioma, Familial paraganglioma, OMIM 185470 complex, subunit B, iron pheochromocytoma, Cowden syndrome 2 sulfur (Ip) GIST SDHC succinate dehydrogenase paraganglioma, Familial paraganglioma OMIM 602413 complex, subunit C, pheochromocytoma, integral membrane GIST protein, 15kDa SDHD succinate dehydrogenase paraganglioma, Familial paraganglioma, OMIM 602690 complex, subunit D, pheochromocytoma Cowden syndrome 3 integral membrane protein SEMA4A sema domain, colorectal OMIM 607292, immunoglobulin domain PUBMED (Ig), transmembrane 25307848 domain (TM) and short cytoplasmic domain, (semaphorin) 4A SH2D1A SH2 domain containing lymphoma X-linked OMIM 300490, 1A lymphoproliferative PUBMED syndrome 1 24429628

174 SLX4 SLX4 structure-specific AML, leukemia Fanconi anaemia P OMIM 613278 endonuclease subunit homolog (S. cerevisiae) SMAD4 Homolog of Drosophila gastrointestinal Juvenile polyposis, OMIM 600993 Mothers Against polyps Hereditary haemorrhagic Decapentaplegic 4 gene telangiectasia SMARCA4 SWI/SNF related, matrix small cell ovarian Rhabdoid tumour OMIM 603254, associated, actin cancer, predispostion syndrome PUBMED dependent regulator of hypercalcaemic type 24429628, chromatin, subfamily a, PUBMED member 4 24658002, PUBMED 24658001 SMARCB1 SWI/SNF related, matrix malignant rhabdoid, Rhabdoid predisposition OMIM 601607 associated, actin schwannoma syndrome dependent regulator of chromatin, subfamily b, member 1 SMARCE1 SWI/SNF related, matrix meningioma Familial meningioma OMIM 603111, associated, actin PUBMED dependent regulator of 24429628 chromatin, subfamily e, member 1 SUFU suppressor of fused medulloblastoma, Medulloblastoma OMIM 607035 homolog (Drosophila) meningioma predisposition, Baseal cell naevus syndrome TERC telomerase RNA Dyskeratosis congenita OMIM 602322 component autosomal dominant 1 TERT telomerase reverse melanoma, Familial melanoma, OMIM 187270 transcriptase leukaemia Dyskeratosis congenita TGFBR1 transforming growth epithelioma Loys Dietz syndrome OMIM 190181, factor, beta receptor 1 PUBMED 24429628 TGFBR2 transforming growth colorectal Colorectal cancer, OMIM 190182 factor, beta receptor II hereditary nonpolyposis, (70/80kDa) type 6, Loeys-Dietz syndrome TMEM127 transmembrane protein Pheochromocytoma, susceptibility to OMIM 613403 127 TRIM37 tripartite motif containing Wilms tumour Mulibrey nanism OMIM 605073, 37 PUBMED 24429628 TSC1 tuberous sclerosis 1 gene hamartoma, renal Tuberous sclerosis 1 OMIM 605284 cell TSC2 tuberous sclerosis 2 gene hamartoma, renal Tuberous sclerosis 2 OMIM 191092 cell VHL von Hippel-Lindau renal, hemangioma, von Hippel-Lindau OMIM 608537 syndrome gene pheochromocytoma syndrome WAS Wiskott-Aldrich syndrome Wiskott-Aldrich syndrome OMIM 300392 WRN Werner syndrome osteosarcoma, Werner Syndrome OMIM 604611 (RECQL2) meningioma, others WT1 Wilms tumour 1 gene Wilms Denys-Drash syndrome, OMIM 607102 Frasier syndrome, Familial Wilms tumor XPA xeroderma pigmentosum, skin basal cell, skin Xeroderma pigmentosum OMIM 611153 complementation group A squamous cell, (A) melanoma XPC xeroderma pigmentosum, skin basal cell, skin Xeroderma pigmentosum OMIM 613208 complementation group C squamous cell, (C)

175 melanoma XRCC2 X-ray repair breast OMIM 600375, complementing defective PUBMED repair in Chinese hamster 22464251 cells 2 Proto-oncogenes

AKT1 V-AKT MURINE breast, thyroid, Cowden syndrome 6, OMIM 164730 THYMOMA VIRAL endoemtrial Proteus syndrome ONCOGENE HOMOLOG 1 ALK ANAPLASTIC neuroblastoma OMIM 105590 LYMPHOMA KINASE CBL Cbl proto-oncogene, E3 leukaemia Noonan syndrome-like OMIM 165360, ubiquitin protein ligase disorder +/- JMML PUBMED 24429628 CDK4 CYCLIN-DEPENDENT melanoma OMIM 123829 KINASE 4 EGFR EPIDERMAL GROWTH Non-small cell lung OMIM 131550 FACTOR RECEPTOR cancer GREM1 gremlin 1, DAN family colorectal Hereditary mixed OMIM 603054 BMP antagonist polyposis syndrome HRAS V-HA-RAS HARVEY RAT multiple tumour Costello synfrome OMIM 190020 SARCOMA VIRAL types ONCOGENE HOMOLOG KIT V-KIT HARDY- GIST, germ cell Familial GIST, Piebaldism OMIM 164920 ZUCKERMAN 4 FELINE tumours SARCOMA VIRAL ONCOGENE HOMOLOG MET MET proto-oncogene, hepatocellular, renal OMIM 164860 receptor tyrosine kinase cell PALLD palladin, cytoskeletal Pancreatic cancer, susceptibility to, 1 OMIM 608092 associated protein PDGFRA platelet-derived growth gastrointestinal Familial GIST OMIM 173490, factor receptor, alpha stromal tumour PUBMED polypeptide (GIST) 24429628 PIK3CA phosphatidylinositol-4,5- breast, thyroid, Cowden syndrome 5 OMIM 171834 bisphosphate 3-kinase, endoemtrial catalytic subunit alpha PTPN11 protein tyrosine Leukemia, juvenile Noonan syndrome, OMIM 176876, phosphatase, non- myelomonocytic LEOPARD syndrome PUBMED receptor type 11 24429628 RET ret proto-oncogene medullary thyroid, Multiple endocrine OMIM 164761 papillary thyroid, neoplasia 2A/2B, Familial pheochromocytoma medullary thyroid cancer RHBDF2 RHOMBOID 5, oesophageal OMIM 614404 DROSOPHILA, HOMOLOG OF, 2; RHBDF2

176 Supplementary Table 1: Complete list of candidate cancer predisposition genes. Red text = present in the ACMG guideline for reporting of secondary variants. Yellow highlight = present in Rahman N, Nature 2014

Variant ranking Initial bioinformatic Final classification classification Class 5 = Genomic coordinates fed to VEP Transcript abrogating Missense variants with pathogenic on ENSEMBL 75: variants (nonsense, with functional "splice_donor_variant" frameshift, consensus characterisation "splice_acceptor_variant" splice site) with previous demonstrating a "stop_gained" reports of pathogenicity functional effect relevant "frameshift_variant" to the disease phenotype and multiple independent reports of pathogenicity. Class 4 = likely "stop_lost" Previously unreported Missense variants with pathogenic "initiator_codon_variant" variants predicted to lead supporting functional "inframe_insertion" to protein truncation evidence, but lacking "inframe_deletion" (nonsense, frameshift, multiple independent “missense variant” and 1000 consensus splice site, reports of pathogenicity. genomes project combined MAF initiator codon, non-stop). < 0.01 and VEST3 call Variants predicted to deleterious abrogate the transcript but occurring in the last exon were called class 3 unless a functional effect or pathogenicity had been previously demonstrated. Class 3 = MAF < 0.01 and Class 3 variants all have a minor allele frequency uncertain "splice_region_variant" <0.01 and include in-frame indels, splice region significance "incomplete_terminal_codon_vari variants and missense variants with predicted ant" deleterious consequence according to VEST3 or “missense variant” and 1000 conflicting reports of pathogenicity. genomes project combined MAF < 0.01 and VEST3 call not available Class 2 = 1000 genomes combined MAF > Class 1 and class 2 variants are comprised of probable non- 0.01 and/or “missense variant” missense variants which either have a MAF < 0.01 pathogenic and VEST3 call tolerated but are predicted to be benign or a MAF 0.01, Class 1 = non- "synonymous_variant" synonymous variants and non-coding variants pathogenic "stop_retained_variant" (intronic, UTR, down- and up-stream). "coding_sequence_variant" "5_prime_UTR_variant" "3_prime_UTR_variant" "non_coding_exon_variant" "nc_transcript_variant" "intron_variant" "NMD_transcript_variant" "upstream_gene_variant" "downstream_gene_variant" "TFBS_ablation" "TFBS_amplification" "TF_binding_site_variant" "regulatory_region_variant" "regulatory_region_ablation" "intergenic_variant"

177 Supplementary Table 2: Criteria for the initial bioinformatic classification and the final classification after manual of germline variants

N = 380 Median DSS P value Variable No. (%) (mo) (logrank)

Sex Male 199 (52.4) 19 Female 181 (47.6) 27.4 0.0821 Age (years) Mean 66.4 Median 67 Range 33 – 90 Missing 69 Outcome Follow-up (months) 0 – 94.9 Median follow-up 15.1 Death PC 224 (58.9) Death other 9 (2.4) Surgical death (<30 days) 14 (3.7) Alive 128 (33.7) - without disease 73 (19.2) - with disease 42 (11.1) - status unknown 13 (3.4) Lost to Follow-up 5 (1.3) Histopathological subtype PDAC 334 (87.9) IPMN carcinoma with invasion 17 (4.5) IPMN carcinoma without invasion 1 (0.3) Adenosquamous 14 (3.7) Mucinous non-cystic carcinoma 6 (1.6) Acinar cell carcinoma 2 (0.5) Undifferentiated 4 (1.1) Small cell carcinoma 2 (0.5) Mixed ductal endocrine carcinoma 1 (0.3) Stagea Ia 8 (2.1) Ib 13 (3.4) 26.8 IIa 70 (18.4) IIb 265 (67.4) 21.7 III 3 (0.8) 21.8 IV 14 (3.7) 14.3 0.0369 Missing 7 (1.8) T Stageb

178 T1 9 (2.4) T2 35 (9.2) 26.8 T3 327 (86.1) T4 5 (1.3) 21.3 0.1056 Missing 4 (1.1) N Stage N0 96 (25.3) 29.7 N1 278 (73.2) 20 0.0245 Missing 6 (1.6) Differentiationc Well 20 (5.3) Moderate 208 (54.7) 29.7 Poor 129 (33.9) Undifferentiated 13 (3.4) 14.9 <0.0001 Missing 10 (2.6) Tumor location Head 307 (80.8) 23.2 Body/Tail 70 (18.4) 16.9 0.0313 Missing 3 (0.8) Tumor size ≤ 20mm 46 (12.1) 30 > 20mm 323 (85.0) 19.3 0.0342 Missing 11 (2.9) Marginsd Clear 268 (70.5) 24.1 Microscopically involved 94 (24.7) 16.9 0.0036 Macroscopically involved 9 (2.4)

179 N = 12 P value Variable No. (%) Median DSS (mo) (logrank)

Sex Male 9 (54.5) 6 Female 3 (45.5) 7.6 0.6869 Age (years) Mean 60.4 Median 61 Range 42 - 76 Outcome Follow-up (months) 4.7 – 36.8 Median follow-up 6 Death PC 9 (75.0) Death other 0 Alive with disease 2 (16.7) Lost to Follow-up 1 Histopathological subtype PDAC 10 (83.3) Undifferentiated 2 (16.7) Clinical Stage III 5 (41.7) 15.6 IV 6 (50.0) 6 0.4386 Missing 1 Supplementary Table 4: Clinico-pathologic features of non-resected PC patients

180 Reference: 1. Butturini G, Stocken DD, Wente MN, et al. Influence of resection margins and treatment on survival in patients with pancreatic cancer: meta- analysis of randomized controlled trials. Archives of Surgery. Jan 2008;143(1):75-83; discussion 83. 2. Neoptolemos JP, Stocken DD, Friess H, et al. A randomized trial of chemoradiotherapy and chemotherapy after resection of pancreatic cancer.[Erratum appears in N Engl J Med. 2004 Aug 12;351(7):726]. New England Journal of Medicine. Mar 18 2004;350(12):1200-1210. 3. Regine WF, Winter KA, Abrams RA, et al. Fluorouracil vs gemcitabine chemotherapy before and after fluorouracil-based chemoradiation following resection of pancreatic adenocarcinoma: a randomized controlled trial.[Erratum appears in JAMA. 2008 Apr 23/30;299(16):1902]. JAMA. Mar 5 2008;299(9):1019-1026. 4. Oettle H, Post S, Neuhaus P, et al. Adjuvant chemotherapy with gemcitabine vs observation in patients undergoing curative-intent resection of pancreatic cancer: a randomized controlled trial. JAMA. Jan 17 2007;297(3):267-277. 5. Neoptolemos JP, Stocken DD, Bassi C, et al. Adjuvant chemotherapy with fluorouracil plus folinic acid vs gemcitabine following pancreatic cancer resection: a randomized controlled trial. JAMA. Sep 8 2010;304(10):1073-1081. 6. Raimondi S, Maisonneuve P, Lowenfels A. Epidemiology of pancreatic cancer: an overview. Nat Rev Gastroenterol Hepatol. Dec 2009;6(12):699-708. doi 610.1038/nrgastro.2009.1177. Epub 2009 Oct 1036. 7. Li D, Morris JS, Liu J, et al. Body mass index and risk, age of onset, and survival in patients with pancreatic cancer. JAMA. Jun 24 2009;301(24):2553-2562. 8. Arslan AA, Helzlsouer KJ, Kooperberg C, et al. Anthropometric measures, body mass index, and pancreatic cancer: a pooled analysis from the Pancreatic Cancer Cohort Consortium (PanScan). Archives of internal medicine. May 10 2010;170(9):791-802. 9. Wolpin BM, Chan AT, Hartge P, et al. ABO blood group and the risk of pancreatic cancer. Journal of the National Cancer Institute. Mar 18 2009;101(6):424-431. 10. Ryan DP, Hong TS, Bardeesy N. Pancreatic adenocarcinoma. The New England journal of medicine. Sep 11 2014;371(11):1039-1049. 11. (AIHW). AIoHaW. Australian Cancer Incidence and Mortality (ACIM) books: Pancreatic cancer. Canberra: AIHW.

181 14. Bosetti C, Lucenteforte E, Silverman DT, et al. Cigarette smoking and pancreatic cancer: an analysis from the International Pancreatic Cancer Case-Control Consortium (Panc4). Annals of oncology : official journal of the European Society for Medical Oncology / ESMO. Jul 2012;23(7):1880-1888. 15. Iodice S, Gandini S, Maisonneuve P, Lowenfels AB. Tobacco and the risk of pancreatic cancer: a review and meta-analysis. Langenbeck's archives of surgery / Deutsche Gesellschaft fur Chirurgie. Jul 2008;393(4):535-545. 16. Blackford A, Parmigiani G, Kensler TW, et al. Genetic mutations associated with cigarette smoking in pancreatic cancer. Cancer research. Apr 15 2009;69(8):3681-3688. 17. Lynch SM, Vrieling A, Lubin JH, et al. Cigarette smoking and pancreatic cancer: a pooled analysis from the pancreatic cancer cohort consortium. Am J Epidemiol. Aug 15 2009;170(4):403-413. 18. Rulyak S, Lowenfels A, Maisonneuve P, Brentnall T. Risk factors for the development of pancreatic cancer in familial pancreatic cancer kindreds. Gastroenterology. May 2003;124(5):1292-1299. 19. James T, Sheldon D, Rajput A, et al. Risk factors associated with earlier age of onset in familial pancreatic carcinoma. Cancer. Dec 15 2004;101(12):2722-2726. 20. Lochan R, Daly AK, Reeves HL, Charnley RM. Family history of cancer and tobacco exposure in index cases of pancreatic ductal adenocarcinoma. Journal of oncology. 2011;2011:215985. 21. Li D, Suzuki H, Liu B, et al. DNA repair gene polymorphisms and risk of pancreatic cancer. Clinical cancer research : an official journal of the American Association for Cancer Research. Jan 15 2009;15(2):740-746. 22. Duell EJ, Bracci PM, Moore JH, Burk RD, Kelsey KT, Holly EA. Detecting pathway-based gene-gene and gene-environment interactions in pancreatic cancer. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Jun 2008;17(6):1470-1479. 23. Pfeifer GP, Denissenko MF, Olivier M, Tretyakova N, Hecht SS, Hainaut P. Tobacco smoke carcinogens, DNA damage and p53 mutations in smoking-associated cancers. Oncogene. Oct 21 2002;21(48):7435-7451. 24. Pleasance ED, Stephens PJ, O'Meara S, et al. A small-cell lung cancer genome with complex signatures of tobacco exposure. Nature. 01/ 2010;463(7278):184-190. 25. Jiao L, Hassan MM, Bondy ML, et al. XRCC2 and XRCC3 gene polymorphism and risk of pancreatic cancer. Am J Gastroenterol. Feb 2008;103(2):360-367. 26. Duell EJ, Holly EA, Bracci PM, Wiencke JK, Kelsey KT. A Population- based Study of the Arg399Gln Polymorphism in X-Ray Repair Cross- Complementing Group 1 (XRCC1) and Risk of Pancreatic Adenocarcinoma. Cancer Research. August 15, 2002 2002;62(16):4630- 4636. 27. McWilliams RR, Bamlet WR, Cunningham JM, et al. Polymorphisms in DNA Repair Genes, Smoking, and Pancreatic Adenocarcinoma Risk. Cancer Research. June 15, 2008 2008;68(12):4928-4935.

182 28. Yeo T, Hruban R, Brody J, Brune K, Fitzgerald S, Yeo C. Assessment of "gene-environment" interaction in cases of familial and sporadic pancreatic cancer. J Gastrointest Surg. Aug 2009;13(8):1487-1494. doi 1410.1007/s11605-11009-10923-11606. Epub 12009 May 11621. 29. Boffetta P, Hecht S, Gray N, Gupta P, Straif K. Smokeless tobacco and cancer. The Lancet Oncology. 2008;9(7):667-675. 30. Vrieling A, Bueno-de-Mesquita HB, Boshuizen HC, et al. Cigarette smoking, environmental tobacco smoke exposure and pancreatic cancer risk in the European Prospective Investigation into Cancer and Nutrition. International journal of cancer. Journal international du cancer. May 15 2010;126(10):2394-2403. 31. Alexandrov LB, Nik-Zainal S, Wedge DC, et al. Signatures of mutational processes in human cancer. Nature. Aug 22 2013;500(7463):415-421. 32. Raimondi S, Maisonneuve P, Lowenfels AB. Epidemiology of pancreatic cancer: an overview. Nature reviews. Gastroenterology & hepatology. Dec 2009;6(12):699-708. 33. Gapstur S, Jacobs E, Deka A, McCullough M, Patel A, Thun M. Association of alcohol intake with pancreatic cancer mortality in never smokers. Arch Intern Med. Mar 14 2011;171(5):444-451. doi 410.1001/archinternmed.2010.1536. 34. Lucenteforte E, La Vecchia C, Silverman D, et al. Alcohol consumption and pancreatic cancer: a pooled analysis in the International Pancreatic Cancer Case-Control Consortium (PanC4). Annals of oncology : official journal of the European Society for Medical Oncology / ESMO. Feb 2012;23(2):374-382. 35. Genkinger JM, Spiegelman D, Anderson KE, et al. Alcohol intake and pancreatic cancer risk: a pooled analysis of fourteen cohort studies. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Mar 2009;18(3):765-776. 36. Anderson M, Zolotarevsky E, Cooper K, et al. Alcohol and tobacco lower the age of presentation in sporadic pancreatic cancer in a dose- dependent manner: a multicenter study. Am J Gastroenterol. Nov 2012;107(11):1730-1739. doi 1710.1038/ajg.2012.1288. Epub 2012 Aug 1728. 37. Ben Q, Cai Q, Li Z, et al. The relationship between new-onset diabetes mellitus and pancreatic cancer risk: a case-control study. Eur J Cancer. Jan 2011;47(2):248-254. doi 210.1016/j.ejca.2010.1007.1010. Epub 2010 Aug 1013. 38. Chari ST, Leibson CL, Rabe KG, et al. Pancreatic Cancer–Associated Diabetes Mellitus: Prevalence and Temporal Association With Diagnosis of Cancer. Gastroenterology. 1// 2008;134(1):95-101. 39. Chari ST, Leibson CL, Rabe KG, Ransom J, de Andrade M, Petersen GM. Probability of pancreatic cancer following diabetes: a population- based study. Gastroenterology. Aug 2005;129(2):504-511. 40. Li D, Tang H, Hassan MM, Holly EA, Bracci PM, Silverman DT. Diabetes and risk of pancreatic cancer: a pooled analysis of three large case- control studies. Cancer causes & control : CCC. Feb 2011;22(2):189- 197.

183 41. Huxley R, Ansary-Moghaddam A, Berrington de Gonzalez A, Barzi F, Woodward M. Type-II diabetes and pancreatic cancer: a meta-analysis of 36 studies. Br J Cancer. Jun 6 2005;92(11):2076-2083. 42. Ben Q, Xu M, Ning X, et al. Diabetes mellitus and risk of pancreatic cancer: A meta-analysis of cohort studies. European journal of cancer. Sep 2011;47(13):1928-1937. 43. Pannala R, Leibson CL, Rabe KG, et al. Temporal association of changes in fasting blood glucose and body mass index with diagnosis of pancreatic cancer. Am J Gastroenterol. Sep 2009;104(9):2318-2325. 44. Elena J, Steplowski E, Yu K, et al. Diabetes and risk of pancreatic cancer: a pooled analysis from the pancreatic cancer cohort consortium. Cancer Causes Control. Jan 2013;24(1):13-25. doi 10.1007/s10552- 10012-10078-10558. Epub 12012 Oct 10531. 45. Pannala R, Basu A, Petersen GM, Chari ST. New-onset diabetes: a potential clue to the early diagnosis of pancreatic cancer. The Lancet Oncology. 1// 2009;10(1):88-95. 46. Li D, Yeung SCJ, Hassan MM, Konopleva M, Abbruzzese JL. Antidiabetic Therapies Affect Risk of Pancreatic Cancer. Gastroenterology. 8// 2009;137(2):482-488. 47. Bodmer M, Becker C, Meier C, Jick SS, Meier CR. Use of antidiabetic agents and the risk of pancreatic cancer: a case-control analysis. Am J Gastroenterol. Apr 2012;107(4):620-626. 48. Bowker SL, Majumdar SR, Veugelers P, Johnson JA. Increased Cancer- Related Mortality for Patients With Type 2 Diabetes Who Use Sulfonylureas or Insulin. Diabetes Care. February 1, 2006 2006;29(2):254-258. 49. Larsson SC, Orsini N, Wolk A. Body mass index and pancreatic cancer risk: A meta-analysis of prospective studies. International Journal of Cancer. 2007;120(9):1993-1998. 50. Renehan AG, Tyson M, Egger M, Heller RF, Zwahlen M. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. The Lancet. 2008;371(9612):569-578. 51. Silverman DT, Swanson CA, Gridley G, et al. Dietary and Nutritional Factors and Pancreatic Cancer: a Case-Control Study Based on Direct Interviews. Journal of the National Cancer Institute. November 18, 1998 1998;90(22):1710-1719. 52. Patel AV, Rodriguez C, Bernstein L, Chao A, Thun MJ, Calle EE. Obesity, Recreational Physical Activity, and Risk of Pancreatic Cancer In a Large U.S. Cohort. Cancer Epidemiology Biomarkers & Prevention. February 1, 2005 2005;14(2):459-466. 53. Berrington de González A, Spencer EA, Bueno-de-Mesquita HB, et al. Anthropometry, Physical Activity, and the Risk of Pancreatic Cancer in the European Prospective Investigation into Cancer and Nutrition. Cancer Epidemiology Biomarkers & Prevention. May 1, 2006 2006;15(5):879-885. 54. Calton BA, Stolzenberg-Solomon RZ, Moore SC, et al. A prospective study of physical activity and the risk of pancreatic cancer among women (United States). BMC cancer. 2008;8:63. 55. Braganza JM, Lee SH, McCloy RF, McMahon MJ. Chronic pancreatitis. The Lancet. 2011;377(9772):1184-1197.

184 56. Duell EJ, Lucenteforte E, Olson SH, et al. Pancreatitis and pancreatic cancer risk: a pooled analysis in the International Pancreatic Cancer Case-Control Consortium (PanC4). Annals of Oncology. November 1, 2012 2012;23(11):2964-2970. 57. Lowenfels AB, Maisonneuve P, Cavallini G, et al. Pancreatitis and the Risk of Pancreatic Cancer. New England Journal of Medicine. 1993;328(20):1433-1437. 58. Malka D, Hammel P, Maire F, et al. Risk of pancreatic adenocarcinoma in chronic pancreatitis. Gut. Dec 2002;51(6):849-852. 59. Hassan MM, Li D, El-Deeb AS, et al. Association Between Hepatitis B Virus and Pancreatic Cancer. Journal of Clinical Oncology. October 1, 2008 2008;26(28):4557-4562. 60. Huang J, Magnusson M, Torner A, Ye W, Duberg AS. Risk of pancreatic cancer among individuals with hepatitis C or hepatitis B virus infection: a nationwide study in Sweden. Br J Cancer. 11/26/print 2013;109(11):2917-2923. 61. Xiao M, Wang Y, Gao Y. Association between Helicobacter pylori Infection and Pancreatic Cancer Development: A Meta- Analysis. PLoS ONE. 2013;8(9):e75559. 62. de Martel C, Llosa AE, Friedman GD, et al. Helicobacter pylori infection and development of pancreatic cancer. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. May 2008;17(5):1188-1194. 63. Stolzenberg-Solomon RZ, Blaser MJ, Limburg PJ, et al. Helicobacter pylori Seropositivity as a Risk Factor for Pancreatic Cancer. Journal of the National Cancer Institute. June 20, 2001 2001;93(12):937-941. 64. Gandini S, Lowenfels AB, Jaffee EM, Armstrong TD, Maisonneuve P. Allergies and the risk of pancreatic cancer: a meta-analysis with review of epidemiology and biological mechanisms. Cancer Epidemiology Biomarkers & Prevention. 2005;14(8):1908-1916. 65. Olson SH, Hsu M, Satagopan JM, et al. Allergies and Risk of Pancreatic Cancer: A Pooled Analysis From the Pancreatic Cancer Case-Control Consortium. American Journal of Epidemiology. 07/02 08/17/received 03/01/accepted 2013;178(5):691-700. 66. Cotterchio M, Lowcock E, Hudson TJ, Greenwood C, Gallinger S. Association between allergies and risk of pancreatic cancer. Cancer Epidemiology Biomarkers & Prevention. 2014;23(3):469-480. 67. Gomez-Rubio P, Zock J-P, Rava M, et al. Reduced risk of pancreatic cancer associated with asthma and nasal allergies. Gut. 2015:gutjnl- 2015-310442. 68. Olson SH, Hsu M, Wiemels JL, et al. Serum immunoglobulin e and risk of pancreatic cancer in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Jul 2014;23(7):1414-1420.

185 69. Cotterchio M, Lowcock E, Bider-Canfield Z, et al. Association between variants in atopy-related immunologic candidate genes and pancreatic cancer risk. PloS one. 2015;10(5). 70. Rittmeyer D, Lorentz A. Relationship between allergy and cancer: an overview. International archives of allergy and immunology. 2012;159(3):216-225. 71. Vesely MD, Kershaw MH, Schreiber RD, Smyth MJ. Natural innate and adaptive immunity to cancer. Annual review of immunology. 2011;29:235-271. 72. Jiao L, Mitrou PN, Reedy J, et al. A combined healthy lifestyle score and risk of pancreatic cancer in a large cohort study. Archives of internal medicine. Apr 27 2009;169(8):764-770. 73. Bobe G, Weinstein SJ, Albanes D, et al. Flavonoid Intake and Risk of Pancreatic Cancer in Male Smokers (Finland). Cancer Epidemiology Biomarkers & Prevention. March 1, 2008 2008;17(3):553-562. 74. Nöthlings U, Murphy SP, Wilkens LR, Henderson BE, Kolonel LN. Flavonols and Pancreatic Cancer Risk: The Multiethnic Cohort Study. American Journal of Epidemiology. October 15, 2007 2007;166(8):924- 931. 75. Larsson SC, Giovannucci E, Wolk A. Folate Intake, MTHFR Polymorphisms, and Risk of Esophageal, Gastric, and Pancreatic Cancer: A Meta-analysis. Gastroenterology. 10// 2006;131(4):1271-1283. 76. Larsson SC, Håkansson N, Giovannucci E, Wolk A. Folate Intake and Pancreatic Cancer Incidence: A Prospective Study of Swedish Women and Men. Journal of the National Cancer Institute. March 15, 2006 2006;98(6):407-413. 77. Oaks BM, Dodd KW, Meinhold CL, Jiao L, Church TR, Stolzenberg- Solomon RZ. Folate intake, post–folic acid grain fortification, and pancreatic cancer risk in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. The American Journal of Clinical Nutrition. February 1, 2010 2010;91(2):449-455. 78. Skinner HG, Michaud DS, Giovannucci E, Willett WC, Colditz GA, Fuchs CS. Vitamin D Intake and the Risk for Pancreatic Cancer in Two Cohort Studies. Cancer Epidemiology Biomarkers & Prevention. September 1, 2006 2006;15(9):1688-1695. 79. Giovannucci E, Liu Y, Rimm EB, et al. Prospective Study of Predictors of Vitamin D Status and Cancer Incidence and Mortality in Men. Journal of the National Cancer Institute. April 5, 2006 2006;98(7):451-459. 80. Larsson SC, Giovannucci E, Wolk A. Methionine and Vitamin B6 Intake and Risk of Pancreatic Cancer: A Prospective Study of Swedish Women and Men. Gastroenterology.132(1):113-118. 81. Meyer MS, Joshipura K, Giovannucci E, Michaud DS. A review of the relationship between tooth loss, periodontal disease, and cancer. Cancer causes & control : CCC. Nov 2008;19(9):895-907. 82. Farrell JJ, Zhang L, Zhou H, et al. Variations of oral microbiota are associated with pancreatic diseases including pancreatic cancer. Gut. Apr 2012;61(4):582-588. 83. MacDermott R, Kramer P. Adenocarcinoma of the pancreas in four siblings. Gastroenterology. Jul 1973;65(1):137-139.

186 84. Ghadirian P, Simard A, Baillargeon J. Cancer of the pancreas in two brothers and one sister. Int J Pancreatol. Oct-Dec 1987;2(5-6):383-391. 85. Reimer R, Fraumeni J, Jr., Ozols R, Bender R. Pancreatic cancer in father and son. Lancet. Apr 23 1977;1(8017):911. 86. Danes B, Lynch H. A familial aggregation of pancreatic cancer. An in vitro study. Jama. May 28 1982;247(20):2798-2802. 87. Ehrenthal D, Haeger L, Griffin T, Compton C. Familial pancreatic adenocarcinoma in three generations. A case report and a review of the literature. Cancer. May 1 1987;59(9):1661-1664. 88. Ghadirian P, Boyle P, Simard A, Baillargeon J, Maisonneuve P, Perret C. Reported family aggregation of pancreatic cancer within a population- based case-control study in the Francophone community in Montreal, Canada. Int J Pancreatol. Nov-Dec 1991;10(3-4):183-196. 89. Fernandez E, La VC, D'Avanzo B, Negri E, Franceschi S. Family history and the risk of liver, gallbladder, and pancreatic cancer. Cancer Epidemiol Biomarkers Prev. Apr-May 1994;3(3):209-212. 90. Silverman D, Schiffman M, Everhart J, et al. Diabetes mellitus, other medical conditions and familial history of cancer as risk factors for pancreatic cancer. Br J Cancer. Aug 1999;80(11):1830-1837. 91. Del CM, Zerbi A, Falconi M, et al. Cancer risk among the relatives of patients with pancreatic ductal adenocarcinoma. Pancreatology. 2007;7(5-6):459-469. Epub 2007 Oct 2001. 92. Bartsch D, Kress R, Sina-Frey M, et al. Prevalence of familial pancreatic cancer in Germany. Int J Cancer. Jul 20 2004;110(6):902-906. 93. Hemminki K, Li X. Familial and second primary pancreatic cancers: a nationwide epidemiologic study from Sweden. Int J Cancer. Feb 10 2003;103(4):525-530. 94. Coughlin SS, Calle EE, Patel AV, Thun MJ. Predictors of pancreatic cancer mortality among a large cohort of United States adults. Cancer causes & control : CCC. Dec 2000;11(10):915-923. 95. Schenk M, Schwartz A, O'Neal E, et al. Familial risk of pancreatic cancer. J Natl Cancer Inst. Apr 18 2001;93(8):640-644. 96. Hemminki K, Dong C, Vaittinen P. Cancer risks to spouses and offspring in the Family-Cancer Database. Genet Epidemiol. Feb 2001;20(2):247- 257. 97. Klein AP, Brune KA, Petersen GM, et al. Prospective risk of pancreatic cancer in familial pancreatic cancer kindreds. Cancer research. Apr 1 2004;64(7):2634-2638. 98. Tersmette A, Petersen G, Offerhaus G, et al. Increased risk of incident pancreatic cancer among first-degree relatives of patients with familial pancreatic cancer. Clin Cancer Res. Mar 2001;7(3):738-744. 99. Bartsch D, Gress T, Langer P. Familial pancreatic cancer--current knowledge. Nat Rev Gastroenterol Hepatol. Aug 2012;9(8):445-453. doi 410.1038/nrgastro.2012.1111. Epub 2012 Jun 1035. 100. Hruban R, Canto M, Goggins M, Schulick R, Klein A. Update on familial pancreatic cancer. Adv Surg. 2010;44:293-311. 101. Brand RE, Lerch MM, Rubinstein WS, et al. Advances in counselling and surveillance of patients at risk for pancreatic cancer. Gut. 2007;56(10):1460-1469.

187 102. Lynch H, Fitzsimmons M, Smyrk T, et al. Familial pancreatic cancer: clinicopathologic study of 18 nuclear families. Am J Gastroenterol. Jan 1990;85(1):54-60. 103. McFaul C, Greenhalf W, Earl J, et al. Anticipation in familial pancreatic cancer. Gut. Feb 2006;55(2):252-258. Epub 2005 Jun 2021. 104. Hahn S, Greenhalf B, Ellis I, et al. BRCA2 germline mutations in familial pancreatic carcinoma. J Natl Cancer Inst. Feb 5 2003;95(3):214-221. 105. Murphy K, Brune K, Griffin C, et al. Evaluation of candidate genes MAP2K4, MADH4, ACVR1B, and BRCA2 in familial pancreatic cancer: deleterious BRCA2 mutations in 17%. Cancer Res. Jul 1 2002;62(13):3789-3793. 106. Couch F, Johnson M, Rabe K, et al. The prevalence of BRCA2 mutations in familial pancreatic cancer. Cancer Epidemiol Biomarkers Prev. Feb 2007;16(2):342-346. 107. Jones S, Hruban R, Kamiyama M, et al. Exomic sequencing identifies PALB2 as a pancreatic cancer susceptibility gene. Science. Apr 10 2009;324(5924):217. doi 210.1126/science.1171202. Epub 1172009 Mar 1171205. 108. Roberts N, Jiao Y, Yu J, et al. ATM mutations in patients with hereditary pancreatic cancer. Cancer Discov. Jan 2012;2(1):41-46. doi 10.1158/2159-8290.CD-1111-0194. Epub 2011 Dec 1129. 109. Lal G, Liu G, Schmocker B, et al. Inherited predisposition to pancreatic adenocarcinoma: role of family history and germ-line p16, BRCA1, and BRCA2 mutations. Cancer Res. Jan 15 2000;60(2):409-416. 110. Klein AP, Beaty TH, Bailey-Wilson JE, Brune KA, Hruban RH, Petersen GM. Evidence for a major gene influencing risk of pancreatic cancer. Genetic epidemiology. Aug 2002;23(2):133-149. 111. Klein A. Identifying people at a high risk of developing pancreatic cancer. Nat Rev Cancer. Jan 2013;13(1):66-74. doi 10.1038/nrc3420. Epub 2012 Dec 1036. 112. Cancer risks in BRCA2 mutation carriers. The Breast Cancer Linkage Consortium. Journal of the National Cancer Institute. Aug 4 1999;91(15):1310-1316. 113. van AC, Brohet R, Meijers-Heijboer E, et al. Cancer risks in BRCA2 families: estimates for sites other than breast and ovary. J Med Genet. Sep 2005;42(9):711-719. 114. Moran A, O'Hara C, Khan S, et al. Risk of cancer other than breast or ovarian in individuals with BRCA1 and BRCA2 mutations. Fam Cancer. Jun 2012;11(2):235-242. doi 210.1007/s10689-10011-19506-10682. 115. Goggins M, Schutte M, Lu J, et al. Germline BRCA2 gene mutations in patients with apparently sporadic pancreatic carcinomas. Cancer research. Dec 1 1996;56(23):5360-5364. 116. Ozcelik H, Schmocker B, Di Nicola N, et al. Germline BRCA2 6174delT mutations in Ashkenazi Jewish pancreatic cancer patients. Nature genetics. May 1997;16(1):17-18. 117. Ferrone CR, Levine DA, Tang LH, et al. BRCA germline mutations in Jewish patients with pancreatic adenocarcinoma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Jan 20 2009;27(3):433-438.

188 118. Murphy KM, Brune KA, Griffin C, et al. Evaluation of candidate genes MAP2K4, MADH4, ACVR1B, and BRCA2 in familial pancreatic cancer: deleterious BRCA2 mutations in 17%. Cancer research. Jul 1 2002;62(13):3789-3793. 119. Hahn SA, Greenhalf B, Ellis I, et al. BRCA2 germline mutations in familial pancreatic carcinoma. Journal of the National Cancer Institute. Feb 5 2003;95(3):214-221. 120. Couch FJ, Johnson MR, Rabe KG, et al. The prevalence of BRCA2 mutations in familial pancreatic cancer. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Feb 2007;16(2):342-346. 121. van Asperen CJ, Brohet RM, Meijers-Heijboer EJ, et al. Cancer risks in BRCA2 families: estimates for sites other than breast and ovary. Journal of Medical Genetics. September 1, 2005 2005;42(9):711-719. 122. Cancer risks in BRCA2 mutation carriers. Journal of the National Cancer Institute. Aug 4 1999;91(15):1310-1316. 123. Risch HA, McLaughlin JR, Cole DE, et al. Population BRCA1 and BRCA2 mutation frequencies and cancer penetrances: a kin-cohort study in Ontario, Canada. Journal of the National Cancer Institute. Dec 6 2006;98(23):1694-1706. 124. Thompson D, Easton D. Variation in cancer risks, by mutation position, in BRCA2 mutation carriers. American journal of human genetics. Feb 2001;68(2):410-419. 125. Thompson D, Easton DF, Consortium tBCL. Cancer Incidence in BRCA1 Mutation Carriers. Journal of the National Cancer Institute. September 18, 2002 2002;94(18):1358-1365. 126. Brose MS, Rebbeck TR, Calzone KA, Stopfer JE, Nathanson KL, Weber BL. Cancer Risk Estimates for BRCA1 Mutation Carriers Identified in a Risk Evaluation Program. Journal of the National Cancer Institute. September 18, 2002 2002;94(18):1365-1372. 127. Iqbal J, Ragone A, Lubinski J, et al. The incidence of pancreatic cancer in BRCA1 and BRCA2 mutation carriers. Br J Cancer. 12/04/print 2012;107(12):2005-2009. 128. Lucas AL, Shakya R, Lipsyc MD, et al. High Prevalence of BRCA1 and BRCA2 Germline Mutations with Loss of Heterozygosity in a Series of Resected Pancreatic Adenocarcinoma and Other Neoplastic Lesions. Clinical Cancer Research. July 1, 2013 2013;19(13):3396-3403. 129. Axilbund JE, Argani P, Kamiyama M, et al. Absence of germline BRCA1 mutations in familial pancreatic cancer patients. Cancer biology & therapy. Jan 2009;8(2):131-135. 130. Skudra S, Staka A, Pukitis A, et al. Association of genetic variants with pancreatic cancer. Cancer genetics and cytogenetics. Nov 2007;179(1):76-78. 131. Shi C, Hruban RH, Klein AP. Familial pancreatic cancer. Archives of pathology & laboratory medicine. Mar 2009;133(3):365-374. 132. Lynch H, Deters C, Snyder C, et al. BRCA1 and pancreatic cancer: pedigree findings and their causal relationships. Cancer Genet Cytogenet. Apr 15 2005;158(2):119-125.

189 133. Flores JF, Pollock PM, Walker GJ, et al. Analysis of the CDKN2A, CDKN2B and CDK4 genes in 48 Australian melanoma kindreds. Oncogene. Dec 11 1997;15(24):2999-3005. 134. Rutter JL, Bromley CM, Goldstein AM, et al. Heterogeneity of risk for melanoma and pancreatic and digestive malignancies. Cancer. 2004;101(12):2809-2816. 135. Vasen HFA, Gruis NA, Frants RR, van der Velden PA, Hille ETM, Bergman W. Risk of developing pancreatic cancer in families with familial atypical multiple mole melanoma associated with a specific 19 deletion of p16 (p16-Leiden). International Journal of Cancer. 2000;87(6):809-811. 136. Moskaluk CA, Hruban RH, Lietman A, et al. Novel germline p16INK4 allele (Asp145Cys) in a family with multiple pancreatic carcinomas. Human Mutation. 1998;12(1):70-70. 137. Beggs AD, Latchford AR, Vasen HFA, et al. Peutz–Jeghers syndrome: a systematic review and recommendations for management. Gut. July 1, 2010 2010;59(7):975-986. 138. McGarrity TJ AC, Frazier ML, et al. . . Peutz-Jeghers Syndrome. 2001 Feb 23 [Updated 2013 Jul 25]. In: Pagon RA, Adam MP, Ardinger HH, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993-2014. Available from: http://www.ncbi.nlm.nih.gov/books/NBK1266/. 2013. 139. Hearle N, Schumacher V, Menko FH, et al. Frequency and Spectrum of Cancers in the Peutz-Jeghers Syndrome. Clinical Cancer Research. May 15, 2006 2006;12(10):3209-3215. 140. Giardiello FM, Welsh SB, Hamilton SR, et al. Increased risk of cancer in the Peutz-Jeghers syndrome. The New England journal of medicine. Jun 11 1987;316(24):1511-1514. 141. Korsse SE, Harinck F, van Lier MGF, et al. Pancreatic cancer risk in Peutz-Jeghers syndrome patients: a large cohort study and implications for surveillance. Journal of Medical Genetics. January 1, 2013 2013;50(1):59-64. 142. Sahin F, Maitra A, Argani P, et al. Loss of Stk11/Lkb1 Expression in Pancreatic and Biliary Neoplasms. Modern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc. //print 0000;16(7):686-691. 143. Kastrinos F, Mukherjee B, Tayob N, et al. Risk of pancreatic cancer in families with Lynch syndrome. JAMA. Oct 28 2009;302(16):1790-1795. 144. Wilentz RE, Goggins M, Redston M, et al. Genetic, immunohistochemical, and clinical features of medullary carcinoma of the pancreas: A newly described and characterized entity. The American journal of pathology. May 2000;156(5):1641-1651. 145. Nakata B, Qing Wang Y, Yashiro M, et al. Prognostic Value of Microsatellite Instability in Resectable Pancreatic Cancer. Clinical Cancer Research. August 1, 2002 2002;8(8):2536-2540. 146. Vasen HFA, van der Meulen–de Jong AE, de Vos tot Nederveen Cappel WH, Oliveira J, Group ObotEGW. Familial colorectal cancer risk: ESMO Clinical Recommendations. Annals of Oncology. May 1, 2009 2009;20(suppl 4):iv51-iv53.

190 147. Groden J, Thliveris A, Samowitz W, et al. Identification and characterization of the familial adenomatous polyposis coli gene. Cell. Aug 9 1991;66(3):589-600. 148. Kinzler KW, Nilbert MC, Su LK, et al. Identification of FAP locus genes from chromosome 5q21. Science. Aug 9 1991;253(5020):661-665. 149. Giardiello FM, Offerhaus GJ, Lee DH, et al. Increased risk of thyroid and pancreatic carcinoma in familial adenomatous polyposis. Gut. Oct 1993;34(10):1394-1396. 150. Maire F, Hammel P, Terris B, et al. Intraductal papillary and mucinous pancreatic tumour: a new extracolonic tumour in familial adenomatous polyposis. Gut. September 1, 2002 2002;51(3):446-449. 151. Chetty R, Salahshor S, Bapat B, Berk T, Croitoru M, Gallinger S. Intraductal papillary mucinous neoplasm of the pancreas in a patient with attenuated familial adenomatous polyposis. Journal of Clinical Pathology. January 1, 2005 2005;58(1):97-101. 152. Li FP, Fraumeni JJF. Soft-Tissue Sarcomas, Breast Cancer, and Other NeoplasmsA Familial Syndrome? Annals of Internal Medicine. 1969;71(4):747-752. 153. Li FP, Fraumeni JF, Jr., Mulvihill JJ, et al. A cancer family syndrome in twenty-four kindreds. Cancer research. Sep 15 1988;48(18):5358-5362. 154. Garber JE, Goldstein AM, Kantor AF, Dreyfus MG, Fraumeni JF, Jr., Li FP. Follow-up study of twenty-four families with Li-Fraumeni syndrome. Cancer research. Nov 15 1991;51(22):6094-6097. 155. Birch JM, Alston RD, McNally R, et al. Relative frequency and morphology of cancers in carriers of germline TP53 mutations. Oncogene. 2001;20(34):4621-4628. 156. Ruijs MW, Verhoef S, Rookus MA, et al. TP53 germline mutation testing in 180 families suspected of Li–Fraumeni syndrome: mutation detection rate and relative frequency of cancers in different familial phenotypes. Journal of medical genetics. 2010;47(6):421-428. 157. Howes N, Lerch MM, Greenhalf W, et al. Clinical and genetic characteristics of hereditary pancreatitis in Europe. Clinical Gastroenterology and Hepatology.2(3):252-261. 158. Whitcomb DC, Gorry MC, Preston RA, et al. Hereditary pancreatitis is caused by a mutation in the cationic trypsinogen gene. Nature genetics. Oct 1996;14(2):141-145. 159. Lowenfels AB, Maisonneuve P, DiMagno EP, et al. Hereditary pancreatitis and the risk of pancreatic cancer. International Hereditary Pancreatitis Study Group. Journal of the National Cancer Institute. Mar 19 1997;89(6):442-446. 160. Lowenfels AB, Maisonneuve P, Whitcomb DC, Lerch MM, DiMagno EP. Cigarette smoking as a risk factor for pancreatic cancer in patients with hereditary pancreatitis. JAMA. Jul 11 2001;286(2):169-170. 161. Rebours V, Boutron-Ruault MC, Schnee M, et al. The natural history of hereditary pancreatitis: a national series. Gut. Jan 2009;58(1):97-103. 162. Klein AP. Genetic susceptibility to pancreatic cancer. Molecular carcinogenesis. Jan 2012;51(1):14-24. 163. Klein AP. Identifying people at a high risk of developing pancreatic cancer. Nat Rev Cancer. Jan 2013;13(1):66-74.

191 164. Brune KA, Lau B, Palmisano E, et al. Importance of Age of Onset in Pancreatic Cancer Kindreds. Journal of the National Cancer Institute. January 20, 2010 2010;102(2):119-126. 165. Wang L, Brune KA, Visvanathan K, et al. Elevated Cancer Mortality in the Relatives of Patients with Pancreatic Cancer. Cancer Epidemiology Biomarkers & Prevention. November 1, 2009 2009;18(11):2829-2834. 166. Giardiello FM, Brensinger JD, Tersmette AC, et al. Very high risk of cancer in familial Peutz-Jeghers syndrome. Gastroenterology. Dec 2000;119(6):1447-1453. 167. Hearle N, Schumacher V, Menko FH, et al. Frequency and spectrum of cancers in the Peutz-Jeghers syndrome. Clinical cancer research : an official journal of the American Association for Cancer Research. May 15 2006;12(10):3209-3215. 168. Schneider R, Slater E, Sina M, et al. German national case collection for familial pancreatic cancer (FaPaCa): ten years experience. Fam Cancer. Jun 2011;10(2):323-330. doi 310.1007/s10689-10010-19414-x. 169. Wang L, Brune K, Visvanathan K, et al. Elevated cancer mortality in the relatives of patients with pancreatic cancer. Cancer Epidemiol Biomarkers Prev. Nov 2009;18(11):2829-2834. doi 2810.1158/1055- 9965.EPI-2809-0557. Epub 2009 Oct 2820. 170. Hiripi E, Lorenzo BJ, Li X, Sundquist J, Hemminki K. Familial association of pancreatic cancer with other malignancies in Swedish families. Br J Cancer. Nov 17 2009;101(10):1792-1797. doi 1710.1038/sj.bjc.6605363. Epub 6602009 Oct 6605313. 171. Cote M, Schenk M, Schwartz A, et al. Risk of other cancers in individuals with a family history of pancreas cancer. J Gastrointest Cancer. 2007;38(2-4):119-126. doi 110.1007/s12029-12008-19022-12022. 172. McWilliams R, Rabe K, Olswold C, De AM, Petersen G. Risk of malignancy in first-degree relatives of patients with pancreatic carcinoma. Cancer. Jul 15 2005;104(2):388-394. 173. Jacobs E, Chanock S, Fuchs C, et al. Family history of cancer and risk of pancreatic cancer: a pooled analysis from the Pancreatic Cancer Cohort Consortium (PanScan). Int J Cancer. Sep 1 2010;127(6):1421-1428. doi 1410.1002/ijc.25148. 174. Jacobs E, Rodriguez C, Newton C, et al. Family history of various cancers and pancreatic cancer mortality in a large cohort. Cancer Causes Control. Oct 2009;20(8):1261-1269. doi 1210.1007/s10552- 10009-19339-10556. Epub 12009 Apr 10525. 175. Amundadottir LT, Thorvaldsson S, Gudbjartsson DF, et al. Cancer as a complex phenotype: pattern of cancer distribution within and beyond the nuclear family. PLoS medicine. Dec 2004;1(3):e65. 176. Hassan MM, Bondy ML, Wolff RA, et al. Risk factors for pancreatic cancer: case-control study. Am J Gastroenterol. Dec 2007;102(12):2696- 2707. 177. Poole C, Byers T, Calle E, Bondy J, Fain P, Rodriguez C. Influence of a family history of cancer within and across multiple sites on patterns of cancer mortality risk for women. Am J Epidemiol. Mar 1 1999;149(5):454- 462.

192 178. Rahman N, Seal S, Thompson D, et al. PALB2, which encodes a BRCA2-interacting protein, is a breast cancer susceptibility gene. Nature genetics. Feb 2007;39(2):165-167. 179. Jones S, Hruban RH, Kamiyama M, et al. Exomic sequencing identifies PALB2 as a pancreatic cancer susceptibility gene. Science. Apr 10 2009;324(5924):217. 180. Hofstatter EW, Domchek SM, Miron A, et al. PALB2 mutations in familial breast and pancreatic cancer. Familial cancer. Jun 2011;10(2):225-231. 181. Slater EP, Langer P, Niemczyk E, et al. PALB2 mutations in European familial pancreatic cancer families. Clinical genetics. Nov 2010;78(5):490-494. 182. Tischkowitz MD, Sabbaghian N, Hamel N, et al. Analysis of the gene coding for the BRCA2-interacting protein PALB2 in familial and sporadic pancreatic cancer. Gastroenterology. Sep 2009;137(3):1183-1186. 183. Roberts NJ, Jiao Y, Yu J, et al. ATM mutations in patients with hereditary pancreatic cancer. Cancer discovery. Jan 2012;2(1):41-46. 184. Eberle MA, Pfutzer R, Pogue-Geile KL, et al. A new susceptibility locus for autosomal dominant pancreatic cancer maps to chromosome 4q32- 34. American journal of human genetics. Apr 2002;70(4):1044-1048. 185. Pogue-Geile KL, Chen R, Bronner MP, et al. Palladin mutation causes familial pancreatic cancer and suggests a new cancer mechanism. PLoS Med. Dec 2006;3(12):e516. 186. Salaria SN, Illei P, Sharma R, et al. Palladin is overexpressed in the nonneoplastic stroma of infiltrating ductal adenocarcinomas of the pancreas, but is only rarely overexpressed in neoplastic cells. Cancer biology & therapy. Mar 2007;6(3):324-328. 187. Brentnall TA, Lai LA, Coleman J, Bronner MP, Pan S, Chen R. Arousal of cancer-associated stroma: overexpression of palladin activates fibroblasts to promote tumor invasion. PLoS One. 2012;7(1):e30219. 188. Goicoechea SM, Garcia-Mata R, Staub J, et al. Palladin promotes invasion of pancreatic cancer cells by enhancing invadopodia formation in cancer-associated fibroblasts. Oncogene. Mar 6 2014;33(10):1265- 1273. 189. Zogopoulos G, Rothenmund H, Eppel A, et al. The P239S palladin variant does not account for a significant fraction of hereditary or early onset pancreas cancer. Human genetics. Jun 2007;121(5):635-637. 190. Welter D, MacArthur J, Morales J, et al. The NHGRI GWAS Catalog, a curated resource of SNP-trait associations. Nucleic acids research. Jan 2014;42(Database issue):D1001-1006. 191. Hruban RH, Goggins M, Parsons J, Kern SE. Progression model for pancreatic cancer. Clin Cancer Res. 2000;6(8):2969-2972. 192. Iacobuzio-Donahue CA, Velculescu VE, Wolfgang CL, Hruban RH. Genetic basis of pancreas cancer development and progression: insights from whole-exome and whole-genome sequencing. Clinical cancer research : an official journal of the American Association for Cancer Research. 2012;18(16):4257-4265. 193. Vincent A, Herman J, Schulick R, Hruban RH, Goggins M. Pancreatic cancer. Lancet. 2011;378(9791):607-620.

193 194. Yachida S, Jones S, Bozic I, et al. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature. 2010;467(7319):1114-1117. 195. Hruban RH, Takaori K, Klimstra DS, et al. An illustrated consensus on the classification of pancreatic intraepithelial neoplasia and intraductal papillary mucinous neoplasms. The American journal of surgical pathology. Aug 2004;28(8):977-987. 196. Hruban RH, Goggins M, Parsons J, Kern SE. Progression Model for Pancreatic Cancer. Clinical Cancer Research. August 1, 2000 2000;6(8):2969-2972. 197. Hruban RH, Wilentz RE, Kern SE. Genetic progression in the pancreatic ducts. The American journal of pathology. Jun 2000;156(6):1821-1825. 198. Goggins M, Hruban RH, Kern SE. BRCA2 is inactivated late in the development of pancreatic intraepithelial neoplasia: evidence and implications. The American journal of pathology. May 2000;156(5):1767- 1771. 199. Moskaluk CA, Hruban RH, Kern SE. p16 and K-ras gene mutations in the intraductal precursors of human pancreatic adenocarcinoma. Cancer research. Jun 1 1997;57(11):2140-2143. 200. Wilentz RE, Geradts J, Maynard R, et al. Inactivation of the p16 (INK4A) tumor-suppressor gene in pancreatic duct lesions: loss of intranuclear expression. Cancer research. Oct 15 1998;58(20):4740-4744. 201. Wilentz RE, Iacobuzio-Donahue CA, Argani P, et al. Loss of expression of Dpc4 in pancreatic intraepithelial neoplasia: evidence that DPC4 inactivation occurs late in neoplastic progression. Cancer research. Apr 1 2000;60(7):2002-2006. 202. Kozuka S, Sassa R, Taki T, et al. Relation of pancreatic duct hyperplasia to carcinoma. Cancer. 1979;43(4):1418-1428. 203. Andea A, Sarkar F, Adsay VN. Clinicopathological correlates of pancreatic intraepithelial neoplasia: a comparative analysis of 82 cases with and 152 cases without pancreatic ductal adenocarcinoma. Modern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc. Oct 2003;16(10):996-1006. 204. Shi C, Klein AP, Goggins M, et al. Increased Prevalence of Precursor Lesions in Familial Pancreatic Cancer Patients. Clinical cancer research : an official journal of the American Association for Cancer Research. Dec 15 2009;15(24):7737-7743. 205. Cubilla AL, Fitzgerald PJ. Morphological lesions associated with human primary invasive nonendocrine pancreas cancer. Cancer research. Jul 1976;36(7 PT 2):2690-2698. 206. Hruban RH, Maitra A, Goggins M. Update on pancreatic intraepithelial neoplasia. International journal of clinical and experimental pathology. 2008;1(4):306. 207. Brune K, Abe T, Canto M, et al. Multifocal neoplastic precursor lesions associated with lobular atrophy of the pancreas in patients having a strong family history of pancreatic cancer. The American journal of surgical pathology. Sep 2006;30(9):1067-1076. 208. Detlefsen S, Sipos B, Feyerabend B, Kloppel G. Pancreatic fibrosis associated with age and ductal papillary hyperplasia. Virchows Archiv : an international journal of pathology. Nov 2005;447(5):800-805.

194 209. Meckler KA, Brentnall TA, Haggitt RC, et al. Familial fibrocystic pancreatic atrophy with endocrine cell hyperplasia and pancreatic carcinoma. The American journal of surgical pathology. Aug 2001;25(8):1047-1053. 210. Canto MI, Goggins M, Hruban RH, et al. Screening for early pancreatic neoplasia in high-risk individuals: a prospective controlled study. Clinical gastroenterology and hepatology : the official clinical practice journal of the American Gastroenterological Association. Jun 2006;4(6):766-781; quiz 665. 211. Morris JP, Wang SC, Hebrok M. KRAS, Hedgehog, Wnt and the twisted developmental biology of pancreatic ductal adenocarcinoma. Nat Rev Cancer. 10//print 2010;10(10):683-695. 212. Del Chiaro M, Segersvärd R, Lohr M, Verbeke C. Early detection and prevention of pancreatic cancer: Is it really possible today? World Journal of Gastroenterology : WJG. 09/14 12/28/received 01/23/revised 05/29/accepted 2014;20(34):12118-12131. 213. Zhu L, Shi G, Schmidt CM, Hruban RH, Konieczny SF. Acinar cells contribute to the molecular heterogeneity of pancreatic intraepithelial neoplasia. The American journal of pathology. Jul 2007;171(1):263-273. 214. Terhune PG, Phifer DM, Tosteson TD, Longnecker DS. K-ras mutation in focal proliferative lesions of human pancreas. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Jun 1998;7(6):515-521. 215. Farrell JJ, Fernandez-del Castillo C. Pancreatic cystic neoplasms: management and unanswered questions. Gastroenterology. Jun 2013;144(6):1303-1315. 216. Distler M, Kersting S, Niedergethmann M, et al. Pathohistological subtype predicts survival in patients with intraductal papillary mucinous neoplasm (IPMN) of the pancreas. Annals of surgery. Aug 2013;258(2):324-330. 217. Adsay NV. Intraductal papillary mucinous neoplasms of the pancreas: Pathology and molecular genetics. Journal of Gastrointestinal Surgery. 2002/10/01 2002;6(5):656-659. 218. Tanaka M, Fernandez-del Castillo C, Adsay V, et al. International consensus guidelines 2012 for the management of IPMN and MCN of the pancreas. Pancreatology : official journal of the International Association of Pancreatology (IAP) ... [et al.]. May-Jun 2012;12(3):183- 197. 219. Wu J, Matthaei H, Maitra A, et al. Recurrent GNAS Mutations Define an Unexpected Pathway for Pancreatic Cyst Development. Science Translational Medicine. July 20, 2011 2011;3(92):92ra66. 220. Furukawa T, Kuboki Y, Tanji E, et al. Whole-exome sequencing uncovers frequent GNAS mutations in intraductal papillary mucinous neoplasms of the pancreas. Sci. Rep. 11/18/online 2011;1. 221. Wolfgang CL, Herman JM, Laheru DA, et al. Recent progress in pancreatic cancer. CA: A Cancer Journal for Clinicians. 2013;63(5):318- 348.

195 222. Wu J, Jiao Y, Dal Molin M, et al. Whole-exome sequencing of neoplastic cysts of the pancreas reveals recurrent mutations in components of ubiquitin-dependent pathways. Proceedings of the National Academy of Sciences. December 27, 2011 2011;108(52):21188-21193. 223. Matthaei H, Wu J, Dal Molin M, et al. GNAS sequencing identifies IPMN- specific mutations in a subgroup of diminutive pancreatic cysts referred to as "incipient IPMNs". The American journal of surgical pathology. Mar 2014;38(3):360-363. 224. Reddy RP, Smyrk TC, Zapiach M, et al. Pancreatic mucinous cystic neoplasm defined by ovarian stroma: demographics, clinical features, and prevalence of cancer. Clinical gastroenterology and hepatology : the official clinical practice journal of the American Gastroenterological Association. Nov 2004;2(11):1026-1031. 225. Crippa S, Salvia R, Warshaw AL, et al. Mucinous cystic neoplasm of the pancreas is not an aggressive entity: lessons from 163 resected patients. Annals of surgery. Apr 2008;247(4):571-579. 226. Wolfgang CL, Herman JM, Laheru DA, et al. Recent progress in pancreatic cancer. CA: a cancer journal for clinicians. Sep 2013;63(5):318-348. 227. Galanis C, Zamani A, Cameron JL, et al. Resected serous cystic neoplasms of the pancreas: a review of 158 patients with recommendations for treatment. Journal of gastrointestinal surgery : official journal of the Society for Surgery of the Alimentary Tract. Jul 2007;11(7):820-826. 228. Neumann HP, Dinkel E, Brambs H, et al. Pancreatic lesions in the von Hippel-Lindau syndrome. Gastroenterology. Aug 1991;101(2):465-471. 229. Reddy S, Cameron JL, Scudiere J, et al. Surgical management of solid- pseudopapillary neoplasms of the pancreas (Franz or Hamoudi tumors): a large single-institutional series. Journal of the American College of Surgeons. May 2009;208(5):950-957; discussion 957-959. 230. Butte JM, Brennan MF, Gonen M, et al. Solid pseudopapillary tumors of the pancreas. Clinical features, surgical outcomes, and long-term survival in 45 consecutive patients from a single center. Journal of gastrointestinal surgery : official journal of the Society for Surgery of the Alimentary Tract. Feb 2011;15(2):350-357. 231. Nguyen NQ, Johns AL, Gill AJ, et al. Clinical and immunohistochemical features of 34 solid pseudopapillary tumors of the pancreas. Journal of gastroenterology and hepatology. Feb 2011;26(2):267-274. 232. Abraham SC, Klimstra DS, Wilentz RE, et al. Solid-pseudopapillary tumors of the pancreas are genetically distinct from pancreatic ductal adenocarcinomas and almost always harbor beta-catenin mutations. The American journal of pathology. Apr 2002;160(4):1361-1369. 233. Lennon AM, Wolfgang CL, Canto MI, et al. The early detection of pancreatic cancer: what will it take to diagnose and treat curable pancreatic neoplasia? Cancer research. Jul 1 2014;74(13):3381-3389. 234. Wang W, Chen S, Brune KA, Hruban RH, Parmigiani G, Klein AP. PancPRO: Risk Assessment for Individuals With a Family History of Pancreatic Cancer. Journal of Clinical Oncology. April 10, 2007 2007;25(11):1417-1422.

196 235. Leonardi G, Marchi S, Falconi M, et al. "PancPro" as a tool for selecting families eligible for pancreatic cancer screening: an Italian study of incident cases. Digestive and liver disease : official journal of the Italian Society of Gastroenterology and the Italian Association for the Study of the Liver. Jul 2012;44(7):585-588. 236. Klein AP, Lindström S, Mendelsohn JB, et al. An Absolute Risk Model to Identify Individuals at Elevated Risk for Pancreatic Cancer in the General Population. PLoS ONE. 2013;8(9):e72311. 237. Grimes DA, Schulz KF. Uses and abuses of screening tests. The Lancet. 3/9/ 2002;359(9309):881-884. 238. Verna EC, Hwang C, Stevens PD, et al. Pancreatic cancer screening in a prospective cohort of high-risk patients: a comprehensive strategy of imaging and genetics. Clinical cancer research. 2010;16(20):5028-5037. 239. Al-Sukhni W, Borgida A, Rothenmund H, et al. Screening for pancreatic cancer in a high-risk cohort: an eight-year experience. Journal of gastrointestinal surgery : official journal of the Society for Surgery of the Alimentary Tract. Apr 2012;16(4):771-783. 240. Poley JW, Kluijt I, Gouma DJ, et al. The yield of first-time endoscopic ultrasonography in screening individuals at a high risk of developing pancreatic cancer. Am J Gastroenterol. Sep 2009;104(9):2175-2181. 241. Vasen HF, Wasser M, van Mil A, et al. Magnetic resonance imaging surveillance detects early-stage pancreatic cancer in carriers of a p16- Leiden mutation. Gastroenterology. Mar 2011;140(3):850-856. 242. Barratt A, Mannes P, Irwig L, Trevena L, Craig J, Rychetnik L. Cancer screening. Journal of Epidemiology and Community Health. December 1, 2002 2002;56(12):899-902. 243. Jung KW, Kim M-H, Lee TY, et al. Clinicopathological aspects of 542 cases of pancreatic cancer: a special emphasis on small pancreatic cancer. Journal of Korean medical science. 2007;22(Suppl):S79-S85. 244. Egawa S, Takeda K, Fukuyama S, Motoi F, Sunamura M, Matsuno S. Clinicopathological aspects of small pancreatic cancer. Pancreas. Apr 2004;28(3):235-240. 245. Agarwal B, Correa AM, Ho L. Survival in pancreatic carcinoma based on tumor size. Pancreas. Jan 2008;36(1):e15-20. 246. Ishikawa O, Ohigashi H, Imaoka S, et al. Minute carcinoma of the pancreas measuring 1 cm or less in diameter--collective review of Japanese case reports. Hepato-gastroenterology. Jan-Feb 1999;46(25):8-15. 247. Franko J, Hugec V, Lopes T, Goldman C. Survival Among Pancreaticoduodenectomy Patients Treated for Pancreatic Head Cancer <1 or 2 cm. Ann Surg Oncol. 2013/02/01 2013;20(2):357-361. 248. Goggins M. Markers of Pancreatic Cancer: Working Toward Early Detection. Clinical Cancer Research. February 15, 2011 2011;17(4):635- 637. 249. Canto MI, Hruban RH, Fishman EK, et al. Frequent detection of pancreatic lesions in asymptomatic high-risk individuals. Gastroenterology. Apr 2012;142(4):796-804; quiz e714-795. 250. Brentnall TA, Bronner MP, Byrd DR, Haggitt RC, Kimmey MB. Early Diagnosis and Treatment of Pancreatic Dysplasia in Patients with a

197 Family History of Pancreatic Cancer. Annals of Internal Medicine. 1999;131(4):247-255. 251. Kimmey MB, Bronner MP, Byrd DR, Brentnall TA. Screening and surveillance for hereditary pancreatic cancer. Gastrointestinal Endoscopy.56(4):S82-S86. 252. Langer P, Kann PH, Fendrich V, et al. Five years of prospective screening of high-risk individuals from families with familial pancreatic cancer. Gut. Oct 2009;58(10):1410-1418. 253. Ludwig E, Olson SH, Bayuga S, et al. Feasibility and yield of screening in relatives from familial pancreatic cancer families. Am J Gastroenterol. May 2011;106(5):946-954. 254. Schneider R, Slater EP, Sina M, et al. German national case collection for familial pancreatic cancer (FaPaCa): ten years experience. Familial cancer. Jun 2011;10(2):323-330. 255. Potjer TP, Schot I, Langer P, et al. Variation in Precursor Lesions of Pancreatic Cancer among High-Risk Groups. Clinical Cancer Research. January 15, 2013 2013;19(2):442-449. 256. Bartsch DK, Dietzel K, Bargello M, et al. Multiple small “imaging” branch- duct type intraductal papillary mucinous neoplasms (IPMNs) in familial pancreatic cancer: indicator for concomitant high grade pancreatic intraepithelial neoplasia? Familial Cancer. 2013/03/01 2013;12(1):89-96. 257. Canto MI, Harinck F, Hruban RH, et al. International Cancer of the Pancreas Screening (CAPS) Consortium summit on the management of patients with increased risk for familial pancreatic cancer. Gut. Mar 2013;62(3):339-347. 258. Del Chiaro M, Zerbi A, Capurso G, et al. Familial pancreatic cancer in Italy. Risk assessment, screening programs and clinical approach: a position paper from the Italian Registry. Digestive and liver disease : official journal of the Italian Society of Gastroenterology and the Italian Association for the Study of the Liver. Sep 2010;42(9):597-605. 259. Canto M, Goggins M, Yeo C, et al. Screening for pancreatic neoplasia in high-risk individuals: an EUS-based approach. Clin Gastroenterol Hepatol. Jul 2004;2(7):606-621. 260. Canto M, Goggins M, Hruban R, et al. Screening for early pancreatic neoplasia in high-risk individuals: a prospective controlled study. Clin Gastroenterol Hepatol. Jun 2006;4(6):766-781; quiz 665. Epub 2006 May 2006. 261. Canto M, Hruban R, Fishman E, et al. Frequent detection of pancreatic lesions in asymptomatic high-risk individuals. Gastroenterology. Apr 2012;142(4):796-804; quiz e714-795. doi 710.1053/j.gastro.2012.1001.1005. Epub 2012 Jan 1012. 262. Harsha HC, Kandasamy K, Ranganathan P, et al. A Compendium of Potential Biomarkers of Pancreatic Cancer. PLoS Med. 2009;6(4):e1000046. 263. Brand RE, Nolen BM, Zeh HJ, et al. Serum biomarker panels for the detection of pancreatic cancer. Clinical cancer research. 2011;17(4):805- 816. 264. Nolen BM, Grizzle W, Brand R, Lokshin AE. Abstract 295: Urine and serum biomarkers of pancreatic cancer. Cancer Research. October 1, 2014 2014;74(19 Supplement):295.

198 265. Nolen BM, Brand RE, Prosser D, et al. Prediagnostic Serum Biomarkers as Early Detection Tools for Pancreatic Cancer in a Large Prospective Cohort Study. PLoS ONE. 2014;9(4):e94928. 266. Koprowski H, Herlyn M, Steplewski Z, Sears HF. Specific antigen in serum of patients with colon carcinoma. Science. 1981;212(4490):53-55. 267. Koprowski H, Steplewski Z, Mitchell K, Herlyn M, Herlyn D, Fuhrer P. Colorectal carcinoma antigens detected by hybridoma antibodies. Somatic cell genetics. 1979;5(6):957-971. 268. Del Villano B, Brennan S, Brock P, et al. Radioimmunometric assay for a monoclonal antibody-defined tumor marker, CA 19-9. Clinical chemistry. 1983;29(3):549-552. 269. Goonetilleke K, Siriwardena A. Systematic review of carbohydrate antigen (CA 19-9) as a biochemical marker in the diagnosis of pancreatic cancer. European Journal of Surgical Oncology (EJSO). 2007;33(3):266- 270. 270. Schlieman MG, Ho HS, Bold RJ. Utility of tumor markers in determining resectability of pancreatic cancer. Archives of Surgery. 2003;138(9):951- 956. 271. Forsmark CE, Lambiase L, Vogel SB. Diagnosis of pancreatic cancer and prediction of unresectability using the tumor-associated antigen CA19-9. Pancreas. 1994;9(6):731-734. 272. Nakao A, Oshima K, Nomoto S, et al. Clinical usefulness of CA 19 9 in pancreatic carcinoma. Paper presented at: Seminars in surgical oncology1998. 273. Barton JG, Bois JP, Sarr MG, et al. Predictive and prognostic value of CA 19-9 in resected pancreatic adenocarcinoma. Journal of Gastrointestinal Surgery. 2009;13(11):2050-2058. 274. Safi F, Schlosser W, Falkenreck S, Beger H. Prognostic value of CA 19-9 serum course in pancreatic cancer. Hepato-gastroenterology. 1997;45(19):253-259. 275. Rosty C, Goggins M. Early detection of pancreatic carcinoma. Hematology/oncology clinics of North America. Feb 2002;16(1):37-52. 276. Steinberg W. The clinical utility of the CA 19-9 tumor-associated antigen. Am J Gastroenterol. Apr 1990;85(4):350-355. 277. Montgomery RC, Hoffman JP, Riley LB, Rogatko A, Ridge JA, Eisenberg BL. Prediction of recurrence and survival by post-resection CA 19-9 values in patients with adenocarcinoma of the pancreas. Ann Surg Oncol. 1997;4(7):551-556. 278. Ferrone CR, Finkelstein DM, Thayer SP, Muzikansky A, Fernandez-del Castillo C, Warshaw AL. Perioperative CA19-9 levels can predict stage and survival in patients with resectable pancreatic adenocarcinoma. Journal of clinical oncology. 2006;24(18):2897-2902. 279. Berger AC, Garcia M, Hoffman JP, et al. Postresection CA 19-9 predicts overall survival in patients with pancreatic cancer treated with adjuvant chemoradiation: a prospective validation by RTOG 9704. Journal of Clinical Oncology. 2008;26(36):5918-5922. 280. Maisey N, Norman A, Hill A, Massey A, Oates J, Cunningham D. CA19-9 as a prognostic factor in inoperable pancreatic cancer: the implication for clinical trials. British journal of cancer. 2005;93(7):740-743.

199 281. Reni M, Cereda S, Balzano G, et al. Carbohydrate antigen 19 9 change during chemotherapy for advanced pancreatic adenocarcinoma. Cancer. 2009;115(12):2630-2639. 282. Wong D, Ko AH, Hwang J, Venook AP, Bergsland EK, Tempero MA. Serum CA19-9 decline compared to radiographic response as a surrogate for clinical outcomes in patients with metastatic pancreatic cancer receiving chemotherapy. Pancreas. 2008;37(3):269-274. 283. Ali CW, Kaye TF, Adamson DJ, Tait IS, Polignano FM, Highley MS. CA 19-9 and survival in advanced and unresectable pancreatic adenocarcinoma and cholangiocarcinoma. Journal of gastrointestinal cancer. 2007;38(2-4):108-114. 284. Hess V, Glimelius B, Grawe P, et al. CA 19-9 tumour-marker response to chemotherapy in patients with advanced pancreatic cancer enrolled in a randomised controlled trial. The lancet oncology. 2008;9(2):132-138. 285. PENDU J, MARIONNEAU S, CAILLEAU THOMAS A, ROCHER J, MOULLAC VAIDYE B, CLÉMENT M. ABH and Lewis histo blood group antigens in cancer. Apmis. 2001;109(1):9-26. 286. Vestergaard EM, Hein HO, Meyer H, et al. Reference values and biological variation for tumor marker CA 19-9 in serum for different Lewis and secretor genotypes and evaluation of secretor and Lewis genotyping in a Caucasian population. Clinical chemistry. 1999;45(1):54-61. 287. Tempero MA, Uchida E, Takasaki H, Burnett DA, Steplewski Z, Pour PM. Relationship of carbohydrate antigen 19-9 and Lewis antigens in pancreatic cancer. Cancer research. 1987;47(20):5501-5503. 288. Ørntoft TF, Vestergaard EM, Holmes E, et al. Influence of Lewis α1-3/4- L-fucosyltransferase (FUT3) gene mutations on enzyme activity, erythrocyte phenotyping, and circulating tumor marker sialyl-Lewis a levels. Journal of Biological Chemistry. 1996;271(50):32260-32268. 289. Mann D, Edwards R, Ho S, Lau W, Glazer G. Elevated tumour marker CA19-9: clinical interpretation and influence of obstructive jaundice. European Journal of Surgical Oncology (EJSO). 2000;26(5):474-479. 290. Locker GY, Hamilton S, Harris J, et al. ASCO 2006 update of recommendations for the use of tumor markers in gastrointestinal cancer. Journal of Clinical Oncology. 2006;24(33):5313-5327. 291. Duffy M, Sturgeon C, Lamerz R, et al. Tumor markers in pancreatic cancer: a European Group on Tumor Markers (EGTM) status report. Annals of oncology. 2009:mdp332. 292. Chan A, Prassas I, Dimitromanolakis A, et al. Validation of biomarkers that complement CA19.9 in detecting early pancreatic cancer. Clinical Cancer Research. September 19, 2014 2014. 293. O'Brien DP, Sandanayake NS, Jenkinson C, et al. Serum CA19-9 is significantly up-regulated up to 2 years prior to diagnosis with pancreatic cancer: implications for early disease detection. Clinical Cancer Research. June 17, 2014 2014. 294. Gold DV, Newsome G, Liu D, Goldenberg DM. Mapping PAM4 (clivatuzumab), a monoclonal antibody in clinical trials for early detection and therapy of pancreatic ductal adenocarcinoma, to MUC5AC mucin. Molecular cancer. 2013;12(1):143. 295. Gold DV, Karanjawala Z, Modrak DE, Goldenberg DM, Hruban RH. PAM4-reactive MUC1 is a biomarker for early pancreatic

200 adenocarcinoma. Clinical cancer research : an official journal of the American Association for Cancer Research. Dec 15 2007;13(24):7380- 7387. 296. Gold DV, Goggins M, Modrak DE, et al. Detection of early-stage pancreatic adenocarcinoma. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Nov 2010;19(11):2786-2794. 297. Gold DV, Gaedcke J, Ghadimi BM, et al. PAM4 enzyme immunoassay alone and in combination with CA 19-9 for the detection of pancreatic adenocarcinoma. Cancer. Feb 1 2013;119(3):522-528. 298. Mehta RS, Chong DQ, Song M, et al. Association Between Plasma Levels of Macrophage Inhibitory Cytokine-1 Before Diagnosis of Colorectal Cancer and Mortality. Gastroenterology. May 27 2015. 299. Koopmann J, Rosenzweig CNW, Zhang Z, et al. Serum Markers in Patients with Resectable Pancreatic Adenocarcinoma: Macrophage Inhibitory Cytokine 1 versus CA19-9. Clinical Cancer Research. January 15, 2006 2006;12(2):442-446. 300. Crowley E, Di Nicolantonio F, Loupakis F, Bardelli A. Liquid biopsy: monitoring cancer-genetics in the blood. Nature Reviews Clinical Oncology. 2013;10(8):472-484. 301. Schwarzenbach H, Hoon DSB, Pantel K. Cell-free nucleic acids as biomarkers in cancer patients. Nat Rev Cancer. 06//print 2011;11(6):426- 437. 302. Croce CM. Causes and consequences of microRNA dysregulation in cancer. Nature Reviews Genetics. 2009;10(10):704-714. 303. Kosaka N, Iguchi H, Ochiya T. Circulating microRNA in body fluid: a new potential biomarker for cancer diagnosis and prognosis. Cancer science. 2010;101(10):2087-2092. 304. Mitchell PS, Parkin RK, Kroh EM, et al. Circulating microRNAs as stable blood-based markers for cancer detection. Proceedings of the National Academy of Sciences. 2008;105(30):10513-10518. 305. Du Rieu MC, Torrisani J, Selves J, et al. MicroRNA-21 is induced early in pancreatic ductal adenocarcinoma precursor lesions. Clinical chemistry. 2010;56(4):603-612. 306. Wang J, Chen J, Chang P, et al. MicroRNAs in plasma of pancreatic ductal adenocarcinoma patients as novel blood-based biomarkers of disease. Cancer prevention research. 2009;2(9):807-813. 307. Liu R, Chen X, Du Y, et al. Serum microRNA expression profile as a biomarker in the diagnosis and prognosis of pancreatic cancer. Clinical chemistry. 2012;58(3):610-618. 308. Li A, Yu J, Kim H, et al. MicroRNA Array Analysis Finds Elevated Serum miR-1290 Accurately Distinguishes Patients with Low-Stage Pancreatic Cancer from Healthy and Disease Controls. Clinical Cancer Research. July 1, 2013 2013;19(13):3600-3610. 309. Jahr S, Hentze H, Englisch S, et al. DNA fragments in the blood plasma of cancer patients: quantitations and evidence for their origin from apoptotic and necrotic cells. Cancer research. Feb 15 2001;61(4):1659- 1665.

201 310. Perkins G, Yap TA, Pope L, et al. Multi-purpose utility of circulating plasma DNA testing in patients with advanced cancers. PloS one. 2012;7(11):e47020. 311. Yamada T, Nakamori S, Ohzato H, et al. Detection of K-ras gene mutations in plasma DNA of patients with pancreatic adenocarcinoma: correlation with clinicopathological features. Clinical cancer research : an official journal of the American Association for Cancer Research. Jun 1998;4(6):1527-1532. 312. Castells A, Puig P, Móra J, et al. K-ras Mutations in DNA Extracted From the Plasma of Patients With Pancreatic Carcinoma: Diagnostic Utility and Prognostic Significance. Journal of Clinical Oncology. February 1, 1999 1999;17(2):578. 313. Mulcahy HE, Lyautey J, Lederrey C, et al. A prospective study of K-ras mutations in the plasma of pancreatic cancer patients. Clinical cancer research. 1998;4(2):271-275. 314. Yu M, Ting DT, Stott SL, et al. RNA sequencing of pancreatic circulating tumour cells implicates WNT signalling in metastasis. Nature. 2012;487(7408):510-513. 315. Allard WJ, Matera J, Miller MC, et al. Tumor Cells Circulate in the Peripheral Blood of All Major Carcinomas but not in Healthy Subjects or Patients With Nonmalignant Diseases. Clinical Cancer Research. October 15, 2004 2004;10(20):6897-6904. 316. Berthelemy P, Bouisson M, Escourrou J, Vaysse N, Rumeau JL, Pradayrol L. Identification of K-ras mutations in pancreatic juice in the early diagnosis of pancreatic cancer. Annals of internal medicine. 1995;123(3):188-191. 317. Berger DH, Chang H, Wood M, et al. Mutational activation of K-ras in nonneoplastic exocrine pancreatic lesions in relation to cigarette smoking status. Cancer. 1999;85(2):326-332. 318. Watanabe H, Ha A, Hu Y-X, et al. K-ras mutations in duodenal aspirate without secretin stimulation for screening of pancreatic and biliary tract carcinoma. Cancer. 1999;86(8):1441-1448. 319. Shi C, Fukushima N, Abe T, et al. Sensitive and quantitative detection of KRAS2 gene mutations in pancreatic duct juice differentiates patients with pancreatic cancer from chronic pancreatitis, potential for early detection. Cancer biology & therapy. Mar 2008;7(3):353-360. 320. Furuya N, Kawa S, Akamatsu T, Furihata K. Long-term follow-up of patients with chronic pancreatitis and K-ras gene mutation detected in pancreatic juice. Gastroenterology. 1997;113(2):593-598. 321. Kanda M, Knight S, Topazian M, et al. Mutant GNAS detected in duodenal collections of secretin-stimulated pancreatic juice indicates the presence or emergence of pancreatic cysts. Gut. 2012:gutjnl-2012- 302823. 322. Kanda M, Sadakari Y, Borges M, et al. Mutant< i> TP53 in Duodenal Samples of Pancreatic Juice From Patients With Pancreatic Cancer or High-Grade Dysplasia. Clinical Gastroenterology and Hepatology. 2013;11(6):719-730. e715. 323. Matsubayashi H, Canto M, Sato N, et al. DNA methylation alterations in the pancreatic juice of patients with suspected pancreatic disease. Cancer research. 2006;66(2):1208-1217.

202 324. Hinton J, Callan R, Bodine C, et al. Potential epigenetic biomarkers for the diagnosis and prognosis of pancreatic ductal adenocarcinomas. 2013. 325. Sato N, Fukushima N, Maitra A, et al. Discovery of Novel Targets for Aberrant Methylation in Pancreatic Carcinoma Using High-Throughput Microarrays. Cancer Research. July 1, 2003 2003;63(13):3735-3742. 326. Yao F, Sun M, Dong M, et al. NPTX2 hypermethylation in pure pancreatic juice predicts pancreatic neoplasms. The American journal of the medical sciences. 2013;346(3):175-180. 327. Yokoyama S, Kitamoto S, Higashi M, et al. Diagnosis of Pancreatic Neoplasms Using a Novel Method of DNA Methylation Analysis of Mucin Expression in Pancreatic Juice. PLoS ONE. 2014;9(4):e93760. 328. Al-Haddad M, Raimondo M, Woodward T, et al. Safety and efficacy of cytology brushings versus standard FNA in evaluating cystic lesions of the pancreas: a pilot study. Gastrointestinal endoscopy. 2007;65(6):894- 898. 329. Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterology. 2004;126(5):1330-1336. 330. Cizginer S, Turner B, Bilge AR, Karaca C, Pitman MB, Brugge WR. Cyst fluid carcinoembryonic antigen is an accurate diagnostic marker of pancreatic mucinous cysts. Pancreas. 2011;40(7):1024-1028. 331. Khalid A, McGrath KM, Zahid M, et al. The role of pancreatic cyst fluid molecular analysis in predicting cyst pathology. Clinical gastroenterology and hepatology : the official clinical practice journal of the American Gastroenterological Association. Oct 2005;3(10):967-973. 332. Rockacy MJ, Zahid M, McGrath KM, Fasanella KE, Khalid A. Association between KRAS mutation, detected in pancreatic cyst fluid, and long-term outcomes of patients. Clinical gastroenterology and hepatology : the official clinical practice journal of the American Gastroenterological Association. Apr 2013;11(4):425-429. 333. Park WG, Wu M, Bowen R, et al. Metabolomic-derived novel cyst fluid biomarkers for pancreatic cysts: glucose and kynurenine. Gastrointestinal endoscopy. Aug 2013;78(2):295-302 e292. 334. Al-Haddad M, DeWitt J, Sherman S, et al. Performance characteristics of molecular (DNA) analysis for the diagnosis of mucinous pancreatic cysts. Gastrointestinal Endoscopy. 1// 2014;79(1):79-87. 335. Jabbar KS, Verbeke C, Hyltander AG, Sjövall H, Hansson GC, Sadik R. Proteomic Mucin Profiling for the Identification of Cystic Precursors of Pancreatic Cancer. Journal of the National Cancer Institute. February 1, 2014 2014;106(2). 336. Khalid A, Zahid M, Finkelstein SD, et al. Pancreatic cyst fluid DNA analysis in evaluating pancreatic cysts: a report of the PANDA study. Gastrointestinal endoscopy. May 2009;69(6):1095-1102. 337. Koprowski H, Herlyn M, Steplewski Z, Sears HF. Specific antigen in serum of patients with colon carcinoma. Science. Apr 3 1981;212(4490):53-55. 338. Koprowski H, Steplewski Z, Mitchell K, Herlyn M, Herlyn D, Fuhrer P. Colorectal carcinoma antigens detected by hybridoma antibodies. Somatic Cell Genetics. Nov 1979;5(6):957-971.

203 339. Goonetilleke KS, Siriwardena AK. Systematic review of carbohydrate antigen (CA 19-9) as a biochemical marker in the diagnosis of pancreatic cancer. European Journal of Surgical Oncology. Apr 2007;33(3):266- 270. 340. Boeck S, Stieber P, Holdenrieder S, Wilkowski R, Heinemann V. Prognostic and therapeutic significance of carbohydrate antigen 19-9 as tumor marker in patients with pancreatic cancer. Oncology. 2006;70(4):255-264. 341. Le Pendu J, Marionneau S, Cailleau-Thomas A, Rocher J, Le Moullac- Vaidye B, Clement M. ABH and Lewis histo-blood group antigens in cancer. APMIS. Jan 2001;109(1):9-31. 342. Vestergaard EM, Hein HO, Meyer H, et al. Reference values and biological variation for tumor marker CA 19-9 in serum for different Lewis and secretor genotypes and evaluation of secretor and Lewis genotyping in a Caucasian population. Clinical Chemistry. Jan 1999;45(1):54-61. 343. Tempero MA, Uchida E, Takasaki H, Burnett DA, Steplewski Z, Pour PM. Relationship of carbohydrate antigen 19-9 and Lewis antigens in pancreatic cancer. Cancer Research. Oct 15 1987;47(20):5501-5503. 344. Orntoft TF, Vestergaard EM, Holmes E, et al. Influence of Lewis alpha1- 3/4-L-fucosyltransferase (FUT3) gene mutations on enzyme activity, erythrocyte phenotyping, and circulating tumor marker sialyl-Lewis a levels. Journal of Biological Chemistry. Dec 13 1996;271(50):32260- 32268. 345. Ong SL, Sachdeva A, Garcea G, et al. Elevation of carbohydrate antigen 19.9 in benign hepatobiliary conditions and its correlation with serum bilirubin concentration. Digestive Diseases & Sciences. Dec 2008;53(12):3213-3217. 346. Mann DV, Edwards R, Ho S, Lau WY, Glazer G. Elevated tumour marker CA19-9: clinical interpretation and influence of obstructive jaundice. European Journal of Surgical Oncology. Aug 2000;26(5):474-479. 347. Biankin AV, Kench JG, Colvin EK, et al. Expression of S100A2 calcium- binding protein predicts response to pancreatectomy for pancreatic cancer. Gastroenterology. 2009 Aug;137(2):558-568. 348. Chang DK, Johns AL, Merrett ND, et al. Margin clearance and outcome in resected pancreatic cancer. Journal of Clinical Oncology. Jun 10 2009;27(17):2855-2862. 349. Edge SB BD, Compton CC, Fritz AG, Greene FL, Trotti A. AJCC Cancer Staging Manual. 7th Edition ed. New York: Springer; 2010. 350. Schlieman MG, Ho HS, Bold RJ. Utility of tumor markers in determining resectability of pancreatic cancer. Archives of Surgery. Sep 2003;138(9):951-955; discussion 955-956. 351. Kim YC, Kim HJ, Park JH, et al. Can preoperative CA19-9 and CEA levels predict the resectability of patients with pancreatic adenocarcinoma? Journal of Gastroenterology & Hepatology. Dec 2009;24(12):1869-1875. 352. Kang CM, Kim JY, Choi GH, et al. The use of adjusted preoperative CA 19-9 to predict the recurrence of resectable pancreatic cancer. Journal of Surgical Research. Jun 1 2007;140(1):31-35.

204 353. Safi F, Schlosser W, Falkenreck S, Beger HG. Prognostic value of CA 19-9 serum course in pancreatic cancer. Hepato-Gastroenterology. Jan- Feb 1998;45(19):253-259. 354. Forsmark CE, Lambiase L, Vogel SB. Diagnosis of pancreatic cancer and prediction of unresectability using the tumor-associated antigen CA19-9. Pancreas. Nov 1994;9(6):731-734. 355. Nakao A, Oshima K, Nomoto S, et al. Clinical usefulness of CA-19-9 in pancreatic carcinoma. Seminars in Surgical Oncology. Jul-Aug 1998;15(1):15-22. 356. Barton JG, Bois JP, Sarr MG, et al. Predictive and prognostic value of CA 19-9 in resected pancreatic adenocarcinoma. Journal of Gastrointestinal Surgery. Nov 2009;13(11):2050-2058. 357. Turrini O, Schmidt CM, Moreno J, et al. Very high serum CA 19-9 levels: a contraindication to pancreaticoduodenectomy? Journal of Gastrointestinal Surgery. Oct 2009;13(10):1791-1797. 358. Montgomery RC, Hoffman JP, Riley LB, Rogatko A, Ridge JA, Eisenberg BL. Prediction of recurrence and survival by post-resection CA 19-9 values in patients with adenocarcinoma of the pancreas. Annals of Surgical Oncology. Oct-Nov 1997;4(7):551-556. 359. Ferrone CR, Finkelstein DM, Thayer SP, Muzikansky A, Fernandez- delCastillo C, Warshaw AL. Perioperative CA19-9 levels can predict stage and survival in patients with resectable pancreatic adenocarcinoma. Journal of Clinical Oncology. Jun 20 2006;24(18):2897-2902. 360. Berger AC, Garcia M, Jr., Hoffman JP, et al. Postresection CA 19-9 predicts overall survival in patients with pancreatic cancer treated with adjuvant chemoradiation: a prospective validation by RTOG 9704. Journal of Clinical Oncology. Dec 20 2008;26(36):5918-5922. 361. Kondo N, Murakami Y, Uemura K, et al. Prognostic impact of perioperative serum CA 19-9 levels in patients with resectable pancreatic cancer. Ann Surg Oncol. Sep 2010;17(9):2321-2329. 362. Kinsella TJ, Seo Y, Willis J, et al. The impact of resection margin status and postoperative CA19-9 levels on survival and patterns of recurrence after postoperative high-dose radiotherapy with 5-FU-based concurrent chemotherapy for resectable pancreatic cancer. American Journal of Clinical Oncology. Oct 2008;31(5):446-453. 363. Hernandez JM, Cowgill SM, Al-Saadi S, et al. CA 19-9 velocity predicts disease-free survival and overall survival after pancreatectomy of curative intent. Journal of Gastrointestinal Surgery. Feb 2009;13(2):349- 353. 364. Maisey NR, Norman AR, Hill A, Massey A, Oates J, Cunningham D. CA19-9 as a prognostic factor in inoperable pancreatic cancer: the implication for clinical trials. British Journal of Cancer. Oct 3 2005;93(7):740-743. 365. Reni M, Cereda S, Balzano G, et al. Carbohydrate antigen 19-9 change during chemotherapy for advanced pancreatic adenocarcinoma. Cancer. Jun 15 2009;115(12):2630-2639. 366. Wong D, Ko AH, Hwang J, Venook AP, Bergsland EK, Tempero MA. Serum CA19-9 decline compared to radiographic response as a

205 surrogate for clinical outcomes in patients with metastatic pancreatic cancer receiving chemotherapy. Pancreas. Oct 2008;37(3):269-274. 367. Ali CW, Kaye TF, Adamson DJA, Tait IS, Polignano FM, Highley MS. CA 19-9 and survival in advanced and unresectable pancreatic adenocarcinoma and cholangiocarcinoma. Journal of Gastrointestinal Cancer. 2007;38(2-4):108-114. 368. Hess V, Glimelius B, Grawe P, et al. CA 19-9 tumour-marker response to chemotherapy in patients with advanced pancreatic cancer enrolled in a randomised controlled trial. Lancet Oncology. Feb 2008;9(2):132-138. 369. Hidalgo M. Pancreatic Cancer. New England Journal of Medicine. 2010;362(17):1605-1617. 370. Butturini G, Stocken DD, Wente MN, et al. Influence of resection margins and treatment on survival in patients with pancreatic cancer: meta- analysis of randomized controlled trials. Archives of surgery (Chicago, Ill. : 1960). Jan 2008;143(1):75-83; discussion 83. 371. Cleary S, Gryfe R, Guindi M, et al. Prognostic factors in resected pancreatic adenocarcinoma: analysis of actual 5-year survivors. J Am Coll Surg. May 2004;198(5):722-731. 372. Gill AJ, Johns AL, Eckstein R, et al. Synoptic reporting improves histopathological assessment of pancreatic resection specimens. Pathology. Feb 2009;41(2):161-167. 373. Edge SB, Compton CC. The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol. Jun 2010;17(6):1471-1474. 374. Biankin AV, Waddell N, Kassahn KS, et al. Pancreatic cancer genomes reveal aberrations in axon guidance pathway genes. Nature. Nov 15 2012;491(7424):399-405. 375. Barton J, Schnelldorfer T, Lohse C, et al. Patterns of pancreatic resection differ between patients with familial and sporadic pancreatic cancer. J Gastrointest Surg. May 2011;15(5):836-842. doi 810.1007/s11605- 11011-11417-x. Epub 12011 Feb 11626. 376. James TA, Sheldon DG, Rajput A, et al. Risk factors associated with earlier age of onset in familial pancreatic carcinoma. Cancer. 2004;101(12):2722-2726. 377. Canto M, Harinck F, Hruban R, et al. International Cancer of the Pancreas Screening (CAPS) Consortium summit on the management of patients with increased risk for familial pancreatic cancer. Gut. Mar 2013;62(3):339-347. doi 310.1136/gutjnl-2012-303108. Epub 302012 Nov 303107. 378. Hackert T, Tjaden C, Muller S, et al. Extrapancreatic malignancies in patients with pancreatic cancer: epidemiology and clinical consequences. Pancreas. Mar 2012;41(2):212-217. doi 210.1097/MPA.1090b1013e3182240602. 379. Mariotto AB, Rowland JH, Ries LA, Scoppa S, Feuer EJ. Multiple cancer prevalence: a growing challenge in long-term survivorship. Cancer epidemiology, biomarkers & prevention : a publication of the American Association for Cancer Research, cosponsored by the American Society of Preventive Oncology. Mar 2007;16(3):566-571.

206 380. Youlden DR, Baade PD. The relative risk of second primary cancers in Queensland, Australia: a retrospective cohort study. BMC cancer. 2011;11:83. 381. Gerdes B, Ziegler A, Ramaswamy A, Wild A, Langer P, Bartsch D. Multiple primaries in pancreatic cancer patients: indicator of a genetic predisposition? Int J Epidemiol. Dec 2000;29(6):999-1003. 382. Shi C, Klein AP, Goggins M, et al. Increased prevalence of precursor lesions in familial pancreatic cancer patients. Clinical Cancer Research. 2009;15(24):7737-7743. 383. Hwang A, Narayan V, Yang Y. Type 2 diabetes mellitus and survival in pancreatic adenocarcinoma: a retrospective cohort study. Cancer. Jan 15 2013;119(2):404-410. doi 410.1002/cncr.27731. Epub 22012 Aug 27731. 384. Chu CK, Mazo AE, Goodman M, et al. Preoperative diabetes mellitus and long-term survival after resection of pancreatic adenocarcinoma. Ann Surg Oncol. Feb 2010;17(2):502-513. 385. Love RR, Evans AM, Josten DM. The accuracy of patient reports of a family history of cancer. Journal of chronic diseases. 1985;38(4):289- 293. 386. Jia P, Pao W, Zhao Z. Patterns and processes of somatic mutations in nine major cancers. BMC medical genomics. 2014;7(1):11. 387. Hamid O, Robert C, Daud A, et al. Safety and tumor responses with lambrolizumab (anti–PD-1) in melanoma. New England Journal of Medicine. 2013;369(2):134-144. 388. Rizvi NA, Mazières J, Planchard D, et al. Activity and safety of nivolumab, an anti-PD-1 immune checkpoint inhibitor, for patients with advanced, refractory squamous non-small-cell lung cancer (CheckMate 063): a phase 2, single-arm trial. The Lancet Oncology. 2015;16(3):257- 265. 389. Le DT, Uram JN, Wang H, et al. PD-1 Blockade in Tumors with Mismatch-Repair Deficiency. New England Journal of Medicine. 2015;372(26):2509-2520. 390. Shlien A, Campbell BB, de Borja R, et al. Combined hereditary and somatic mutations of replication error repair genes result in rapid onset of ultra-hypermutated cancers. Nature genetics. 2015;47(3):257-262. 391. Jäger N, Schlesner M, Jones DT, et al. Hypermutation of the inactive X chromosome is a frequent event in cancer. Cell. 2013;155(3):567-581. 392. Roberts SA, Gordenin DA. Hypermutation in human cancer genomes: footprints and mechanisms. Nature Reviews Cancer. 2014;14(12):786- 800. 393. Supek F, Lehner B. Differential DNA mismatch repair underlies mutation rate variation across the human genome. Nature. 2015;521(7550):81-84. 394. Dulak AM, Stojanov P, Peng S, et al. Exome and whole-genome sequencing of esophageal adenocarcinoma identifies recurrent driver events and mutational complexity. Nature genetics. 2013;45(5):478-486. 395. Bolli N, Avet-Loiseau H, Wedge DC, et al. Heterogeneity of genomic evolution and mutational profiles in multiple myeloma. Nature communications. 2014;5. 396. Nik-Zainal S, Alexandrov LB, Wedge DC, et al. Mutational processes molding the genomes of 21 breast cancers. Cell. 2012;149(5):979-993.

207 397. Li GM. Mechanisms and functions of DNA mismatch repair. Cell research. Jan 2008;18(1):85-98. 398. Jiricny J. The multifaceted mismatch-repair system. Nature reviews. Molecular cell biology. May 2006;7(5):335-346. 399. Umar A, Boland CR, Terdiman JP, et al. Revised Bethesda Guidelines for hereditary nonpolyposis colorectal cancer (Lynch syndrome) and microsatellite instability. J Natl Cancer Inst. Feb 18 2004;96(4):261-268. 400. Frayling IM. Microsatellite instability. Gut. // 1999;45(1):1-4. 401. Kim GP, Colangelo LH, Wieand HS, et al. Prognostic and predictive roles of high-degree microsatellite instability in colon cancer: a National Cancer Institute-National Surgical Adjuvant Breast and Bowel Project Collaborative Study. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Mar 1 2007;25(7):767-772. 402. Kane MF, Loda M, Gaida GM, et al. Methylation of the hMLH1 promoter correlates with lack of expression of hMLH1 in sporadic colon tumors and mismatch repair-defective human tumor cell lines. Cancer research. Mar 1 1997;57(5):808-811. 403. Comprehensive molecular characterization of human colon and rectal cancer. Nature. Jul 19 2012;487(7407):330-337. 404. Kandoth C, Schultz N, Cherniack AD, et al. Integrated genomic characterization of endometrial carcinoma. Nature. May 2 2013;497(7447):67-73. 405. Kim T-M, Laird Peter W, Park Peter J. The Landscape of Microsatellite Instability in Colorectal and Endometrial Cancer Genomes. Cell. 11/7/ 2013;155(4):858-868. 406. Wang K, Kan J, Yuen ST, et al. Exome sequencing identifies frequent mutation of ARID1A in molecular subtypes of gastric cancer. Nature genetics. Dec 2011;43(12):1219-1223. 407. Mensenkamp AR, Vogelaar IP, van Zelst-Stams WA, et al. Somatic mutations in MLH1 and MSH2 are a frequent cause of mismatch-repair deficiency in Lynch syndrome-like tumors. Gastroenterology. Mar 2014;146(3):643-646 e648. 408. Haraldsdottir S, Hampel H, Tomsic J, et al. Colon and Endometrial Cancers With Mismatch Repair Deficiency Can Arise From Somatic, Rather Than Germline, Mutations. Gastroenterology. 12// 2014;147(6):1308-1316.e1301. 409. Geurts-Giele WR, Leenen CH, Dubbink HJ, et al. Somatic aberrations of mismatch repair genes as a cause of microsatellite-unstable cancers. The Journal of pathology. Dec 2014;234(4):548-559. 410. Pritchard CC, Morrissey C, Kumar A, et al. Complex MSH2 and MSH6 mutations in hypermutated microsatellite unstable advanced prostate cancer. Nat Commun. 09/25/online 2014;5. 411. Palles C, Cazier JB, Howarth KM, et al. Germline mutations affecting the proofreading domains of POLE and POLD1 predispose to colorectal adenomas and carcinomas. Nature genetics. Feb 2013;45(2):136-144. 412. Hudson TJ, Anderson W, Artez A, et al. International network of cancer genome projects. Nature. Apr 15 2010;464(7291):993-998. 413. Song S, Nones K, Miller D, et al. qpure: A tool to estimate tumor cellularity from genome-wide single-nucleotide polymorphism profiles. PLoS One. 2012;7(9):e45835.

208 414. Kassahn KS, Holmes O, Nones K, et al. Somatic Point Mutation Calling in Low Cellularity Tumors. PloS one. 2013;8(11):e74380. 415. McKenna A, Hanna M, Banks E, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. Sep 2010;20(9):1297-1303. 416. Ye K, Schulz MH, Long Q, Apweiler R, Ning Z. Pindel: a pattern growth approach to detect break points of large deletions and medium sized insertions from paired-end short reads. Bioinformatics. Nov 1 2009;25(21):2865-2871. 417. Popova T, Manié E, Stoppa-Lyonnet D, Rigaill G, Barillot E, Stern MH. Genome Alteration Print (GAP): a tool to visualize and mine complex cancer genomic profiles obtained by SNP arrays. Genome Biology. 11/11 02/02/received 09/24/revised 11/11/accepted 2009;10(11):R128-R128. 418. Quek K, Nones K, Patch AM, et al. A workflow to increase verification rate of chromosomal structural rearrangements using high-throughput next-generation sequencing. BioTechniques. 2014;57(1):31-38. 419. Krzywinski M, Schein J, Birol I, et al. Circos: an information aesthetic for comparative genomics. Genome Res. Sep 2009;19(9):1639-1645. 420. Niu B, Ye K, Zhang Q, et al. MSIsensor: microsatellite instability detection using paired tumor-normal sequence data. Bioinformatics. December 25, 2013 2013. 421. Umar A, Boland CR, Terdiman JP, et al. Revised Bethesda Guidelines for Hereditary Nonpolyposis Colorectal Cancer (Lynch Syndrome) and Microsatellite Instability. Journal of the National Cancer Institute. February 18, 2004 2004;96(4):261-268. 422. Hall G, Clarkson A, Shi A, et al. Immunohistochemistry for PMS2 and MSH6 alone can replace a four antibody panel for mismatch repair deficiency screening in colorectal adenocarcinoma. Pathology. 2010;42(5):409-413. 423. McLaren W, Pritchard B, Rios D, Chen Y, Flicek P, Cunningham F. Deriving the consequences of genomic variants with the Ensembl API and SNP Effect Predictor. Bioinformatics. August 15, 2010 2010;26(16):2069-2070. 424. Adzhubei I, Jordan DM, Sunyaev SR. Predicting functional effect of human missense mutations using PolyPhen-2. Current protocols in human genetics / editorial board, Jonathan L. Haines ... [et al.]. Jan 2013;Chapter 7:Unit7 20. 425. Kumar P, Henikoff S, Ng PC. Predicting the effects of coding non- synonymous variants on protein function using the SIFT algorithm. Nature protocols. 2009;4(7):1073-1081. 426. Kircher M, Witten DM, Jain P, O'Roak BJ, Cooper GM, Shendure J. A general framework for estimating the relative pathogenicity of human genetic variants. Nature genetics. Mar 2014;46(3):310-315. 427. Carter H, Douville C, Stenson PD, Cooper DN, Karchin R. Identifying Mendelian disease genes with the variant effect scoring tool. BMC genomics. 2013;14 Suppl 3:S3. 428. Mathe E, Olivier M, Kato S, Ishioka C, Hainaut P, Tavtigian SV. Computational approaches for predicting the biological effect of p53

209 missense mutations: a comparison of three sequence analysis based methods. Nucleic acids research. 2006;34(5):1317-1325. 429. Schwarz JM, Rodelsperger C, Schuelke M, Seelow D. MutationTaster evaluates disease-causing potential of sequence alterations. Nature methods. Aug 2010;7(8):575-576. 430. Pollard KS, Hubisz MJ, Rosenbloom KR, Siepel A. Detection of nonneutral substitution rates on mammalian phylogenies. Genome Res. Jan 2010;20(1):110-121. 431. Siepel A, Bejerano G, Pedersen JS, et al. Evolutionarily conserved elements in vertebrate, insect, worm, and yeast genomes. Genome Res. Aug 2005;15(8):1034-1050. 432. Stenson PD, Mort M, Ball EV, et al. The Human Gene Mutation Database: 2008 update. Genome medicine. 2009;1(1):13. 433. Thompson BA, Spurdle AB, Plazzer J-P, et al. Application of a 5-tiered scheme for standardized classification of 2,360 unique mismatch repair gene variants in the InSiGHT locus-specific database. Nature genetics. 02//print 2014;46(2):107-115. 434. Waddell N, Pajic M, Patch A-M, et al. Whole genomes redefine the mutational landscape of pancreatic cancer. Nature. 2015;518(7540):495- 501. 435. Li F, Mao G, Tong D, et al. The histone mark H3K36me3 regulates human DNA mismatch repair through its interaction with MutSalpha. Cell. Apr 25 2013;153(3):590-600. 436. Nones K, Waddell N, Wayte N, et al. Genomic catastrophes frequently arise in esophageal adenocarcinoma and drive tumorigenesis. Nat Commun. 10/29/online 2014;5. 437. Aaltonen LA, Salovaara R, Kristo P, et al. Incidence of hereditary nonpolyposis colorectal cancer and the feasibility of molecular screening for the disease. The New England journal of medicine. May 21 1998;338(21):1481-1487. 438. Laghi L, Beghelli S, Spinelli A, et al. Irrelevance of microsatellite instability in the epidemiology of sporadic pancreatic ductal adenocarcinoma. PloS one. 2012;7(9):e46002. 439. Rodriguez-Soler M, Perez-Carbonell L, Guarinos C, et al. Risk of cancer in cases of suspected lynch syndrome without germline mutation. Gastroenterology. May 2013;144(5):926-932 e921; quiz e913-924. 440. Sourrouille I, Coulet F, Lefevre JH, et al. Somatic mosaicism and double somatic hits can lead to MSI colorectal tumors. Familial cancer. Mar 2013;12(1):27-33. 441. Kang SY, Park CK, Chang DK, et al. Lynch like syndrome: Characterization and comparison with EPCAM deletion carriers. International Journal of Cancer. 2014. 442. Roberts SA, Sterling J, Thompson C, et al. Clustered mutations in yeast and in human cancers can arise from damaged long single-strand DNA regions. Molecular cell. 2012;46(4):424-435. 443. Yang Y, Sterling J, Storici F, Resnick MA, Gordenin DA. Hypermutability of Damaged Single-Strand DNA Formed at Double-Strand Breaks and Uncapped Telomeres in Yeast Saccharomyces cerevisiae. PLoS Genet. 2008;4(11):e1000264.

210 444. Drier Y, Lawrence MS, Carter SL, et al. Somatic rearrangements across cancer reveal classes of samples with distinct patterns of DNA breakage and rearrangement-induced hypermutability. Genome research. 2013;23(2):228-235. 445. Bryant HE, Schultz N, Thomas HD, et al. Specific killing of BRCA2- deficient tumours with inhibitors of poly(ADP-ribose) polymerase. Nature. 04/14/print 2005;434(7035):913-917. 446. Walsh T, Casadei S, Lee MK, et al. Mutations in 12 genes for inherited ovarian, fallopian tube, and peritoneal carcinoma identified by massively parallel sequencing. Proceedings of the National Academy of Sciences. November 1, 2011 2011;108(44):18032-18037. 447. Couch FJ, Hart SN, Sharma P, et al. Inherited Mutations in 17 Breast Cancer Susceptibility Genes Among a Large Triple-Negative Breast Cancer Cohort Unselected for Family History of Breast Cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Dec 1 2014. 448. Kanchi KL, Johnson KJ, Lu C, et al. Integrated analysis of germline and somatic variants in ovarian cancer. Nat Commun. 2014;5:3156. 449. Humphris JL, Johns AL, Simpson SH, et al. Clinical and pathologic features of familial pancreatic cancer. Cancer. Dec 1 2014;120(23):3669- 3675. 450. Rhim Andrew D, Mirek Emily T, Aiello Nicole M, et al. EMT and Dissemination Precede Pancreatic Tumor Formation. Cell. 1/20/ 2012;148(1–2):349-361. 451. Green RC, Berg JS, Grody WW, et al. ACMG recommendations for reporting of incidental findings in clinical exome and genome sequencing. Genet Med. Jul 2013;15(7):565-574. 452. Li H, Durbin R. Fast and accurate short read alignment with Burrows- Wheeler transform. Bioinformatics. Jul 15 2009;25(14):1754-1760. 453. Richards CS, Bale S, Bellissimo DB, et al. ACMG recommendations for standards for interpretation and reporting of sequence variations: Revisions 2007. Genet Med. 04//print 2008;10(4):294-300. 454. Fokkema IFAC, Taschner PEM, Schaafsma GCP, Celli J, Laros JFJ, den Dunnen JT. LOVD v.2.0: the next generation in gene variant databases. Human Mutation. 2011;32(5):557-563. 455. Alexandrov LB, Nik-Zainal S, Wedge DC, et al. Signatures of mutational processes in human cancer. Nature. 08/22/print 2013;500(7463):415- 421. 456. Hofmann W, Horn D, Huttner C, Classen E, Scherneck S. The BRCA2 variant 8204G>A is a splicing mutation and results in an in frame deletion of the gene. Journal of medical genetics. Mar 2003;40(3):e23. 457. Wu K, Hinson SR, Ohashi A, et al. Functional evaluation and cancer risk assessment of BRCA2 unclassified variants. Cancer Res. Jan 15 2005;65(2):417-426. 458. Farrugia DJ, Agarwal MK, Pankratz VS, et al. Functional assays for classification of BRCA2 variants of uncertain significance. Cancer Res. May 1 2008;68(9):3523-3531. 459. Friedman LS, Ostermeyer EA, Szabo CI, et al. Confirmation of BRCA1 by analysis of germline mutations linked to breast and ovarian cancer in ten families. Nature genetics. Dec 1994;8(4):399-404.

211 460. Towler WI, Zhang J, Ransburgh DJ, et al. Analysis of BRCA1 variants in double-strand break repair by homologous recombination and single- strand annealing. Hum Mutat. Mar 2013;34(3):439-445. 461. Montagna M, Santacatterina M, Corneo B, et al. Identification of seven new BRCA1 germline mutations in Italian breast and breast/ovarian cancer families. Cancer research. Dec 1 1996;56(23):5466-5469. 462. Borg Å, Sandberg T, Nilsson K, et al. High frequency of multiple melanomas and breast and pancreas carcinomas in CDKN2A mutation- positive melanoma families. Journal of the National Cancer Institute. 2000;92(15):1260-1266. 463. Robertson KD, Jones PA. Tissue-specific alternative splicing in the human INK4a/ARF cell cycle regulatory locus. Oncogene. Jul 1 1999;18(26):3810-3820. 464. Goldstein AM, Chan M, Harland M, et al. High-risk melanoma susceptibility genes and pancreatic cancer, neural system tumors, and uveal melanoma across GenoMEL. Cancer research. Oct 15 2006;66(20):9818-9828. 465. Hussussian CJ, Struewing JP, Goldstein AM, et al. Germline p16 mutations in familial melanoma. Nature genetics. Sep 1994;8(1):15-21. 466. Scaini MC, Minervini G, Elefanti L, et al. CDKN2A unclassified variants in familial malignant melanoma: combining functional and computational approaches for their assessment. Hum Mutat. Jul 2014;35(7):828-840. 467. Miller PJ, Duraisamy S, Newell JA, et al. Classifying variants of CDKN2A using computational and laboratory studies. Human Mutation. 2011;32(8):900-911. 468. Kastan MB, Zhan Q, el-Deiry WS, et al. A mammalian cell cycle checkpoint pathway utilizing p53 and GADD45 is defective in ataxia- telangiectasia. Cell. Nov 13 1992;71(4):587-597. 469. Goldgar DE, Healey S, Dowty JG, et al. Rare variants in the ATM gene and risk of breast cancer. Breast cancer research : BCR. 2011;13(4):R73. 470. Barone G, Groom A, Reiman A, Srinivasan V, Byrd PJ, Taylor AM. Modeling ATM mutant proteins from missense changes confirms retained kinase activity. Hum Mutat. Aug 2009;30(8):1222-1230. 471. Jenne DE, Reimann H, Nezu J, et al. Peutz-Jeghers syndrome is caused by mutations in a novel serine threonine kinase. Nature genetics. Jan 1998;18(1):38-43. 472. Aretz S, Stienen D, Uhlhaas S, et al. High proportion of large genomic STK11 deletions in Peutz-Jeghers syndrome. Human Mutation. 2005;26(6):513-519. 473. Hampel H, Bennett RL, Buchanan A, Pearlman R, Wiesner GL. A practice guideline from the American College of Medical Genetics and Genomics and the National Society of Genetic Counselors: referral indications for cancer predisposition assessment. Genet Med. 11/13/online 2014. 474. Varon R, Vissinga C, Platzer M, et al. Nibrin, a novel DNA double-strand break repair protein, is mutated in Nijmegen breakage syndrome. Cell. May 1 1998;93(3):467-476.

212 475. Wang W. Emergence of a DNA-damage response network consisting of Fanconi anaemia and BRCA proteins. Nature reviews. Genetics. Oct 2007;8(10):735-748. 476. van der Heijden MS, Yeo CJ, Hruban RH, Kern SE. Fanconi anemia gene mutations in young-onset pancreatic cancer. Cancer research. May 15 2003;63(10):2585-2588. 477. Castella M, Pujol R, Callen E, et al. Origin, functional role, and clinical impact of Fanconi anemia FANCA mutations. Blood. Apr 7 2011;117(14):3759-3769. 478. Thompson ER, Doyle MA, Ryland GL, et al. Exome sequencing identifies rare deleterious mutations in DNA repair genes FANCC and BLM as potential breast cancer susceptibility alleles. PLoS Genet. Sep 2012;8(9):e1002894. 479. Polito D, Cukras S, Wang X, et al. The carboxyl terminus of FANCE recruits FANCD2 to the Fanconi Anemia (FA) E3 ligase complex to promote the FA DNA repair pathway. The Journal of biological chemistry. Mar 7 2014;289(10):7003-7010. 480. Ali AM, Kirby M, Jansen M, et al. Identification and characterization of mutations in FANCL gene: a second case of Fanconi anemia belonging to FA-L complementation group. Hum Mutat. Jul 2009;30(7):E761-770. 481. Smith CG, Naven M, Harris R, et al. Exome resequencing identifies potential tumor-suppressor genes that predispose to colorectal cancer. Hum Mutat. Jul 2013;34(7):1026-1034. 482. Loveday C, Turnbull C, Ruark E, et al. Germline RAD51C mutations confer susceptibility to ovarian cancer. Nature genetics. May 2012;44(5):475-476; author reply 476. 483. Seemanova E, Jarolim P, Seeman P, et al. Cancer risk of heterozygotes with the NBN founder mutation. Journal of the National Cancer Institute. Dec 19 2007;99(24):1875-1880. 484. Damiola F, Pertesi M, Oliver J, et al. Rare key functional domain missense substitutions in MRE11A, RAD50, and NBN contribute to breast cancer susceptibility: results from a Breast Cancer Family Registry case-control mutation-screening study. Breast cancer research : BCR. 2014;16(3):R58. 485. Schroder-Heurich B, Bogdanova N, Wieland B, et al. Functional deficiency of NBN, the Nijmegen breakage syndrome protein, in a p.R215W mutant breast cancer cell line. BMC cancer. 2014;14:434. 486. Yamamoto Y, Miyamoto M, Tatsuda D, et al. A rare polymorphic variant of NBS1 reduces DNA repair activity and elevates chromosomal instability. Cancer Res. Jul 15 2014;74(14):3707-3715. 487. Mendez G, Cilli D, Berardinelli F, et al. Cleavage of the BRCT tandem domains of nibrin by the 657del5 mutation affects the DNA damage response less than the Arg215Trp mutation. IUBMB life. 2012;64(10):853-861. 488. Croteau DL, Popuri V, Opresko PL, Bohr VA. Human RecQ helicases in DNA repair, recombination, and replication. Annual review of biochemistry. 2014;83:519-552. 489. Gruber SB, Ellis NA, Scott KK, et al. BLM heterozygosity and the risk of colorectal cancer. Science. Sep 20 2002;297(5589):2013.

213 490. Goss KH, Risinger MA, Kordich JJ, et al. Enhanced tumor formation in mice heterozygous for Blm mutation. Science. Sep 20 2002;297(5589):2051-2053. 491. Suspitsin EN, Yanus GA, Sokolenko AP, et al. Development of breast tumors in CHEK2, NBN/NBS1 and BLM mutation carriers does not commonly involve somatic inactivation of the wild-type allele. Medical oncology (Northwood, London, England). Feb 2014;31(2):828. 492. German J, Sanz MM, Ciocci S, Ye TZ, Ellis NA. Syndrome-causing mutations of the BLM gene in persons in the Bloom's Syndrome Registry. Hum Mutat. Aug 2007;28(8):743-753. 493. O'Driscoll M, Cerosaletti KM, Girard PM, et al. DNA ligase IV mutations identified in patients exhibiting developmental delay and immunodeficiency. Molecular cell. Dec 2001;8(6):1175-1185. 494. Bartek J, Lukas J. Chk1 and Chk2 kinases in checkpoint control and cancer. Cancer cell. May 2003;3(5):421-429. 495. Meijers-Heijboer H, van den Ouweland A, Klijn J, et al. Low-penetrance susceptibility to breast cancer due to CHEK2(*)1100delC in noncarriers of BRCA1 or BRCA2 mutations. Nat Genet. May 2002;31(1):55-59. 496. Calado RT, Regal JA, Hills M, et al. Constitutional hypomorphic telomerase mutations in patients with acute myeloid leukemia. Proceedings of the National Academy of Sciences. January 27, 2009 2009;106(4):1187-1192. 497. Horn S, Figl A, Rachakonda PS, et al. TERT Promoter Mutations in Familial and Sporadic Melanoma. Science. February 22, 2013 2013;339(6122):959-961. 498. Yamaguchi H, Calado RT, Ly H, et al. Mutations in TERT, the gene for telomerase reverse transcriptase, in aplastic anemia. The New England journal of medicine. Apr 7 2005;352(14):1413-1424. 499. Stumpff J, Ghule PN, Shimamura A, Stein JL, Greenblatt M. Spindle Microtubule Dysfunction and Cancer Predisposition. Journal of cellular physiology. 2014;229(12):1881-1883. 500. Liu JM, Lipton JM, Mani S. Sixth International Congress on Shwachman- Diamond syndrome: from patients to genes and back. Annals of the New York Academy of Sciences. Dec 2011;1242:26-39. 501. Boocock GR, Morrison JA, Popovic M, et al. Mutations in SBDS are associated with Shwachman–Diamond syndrome. Nature genetics. 2003;33(1):97-101. 502. Dhanraj S, Manji A, Pinto D, et al. Molecular characteristics of a pancreatic adenocarcinoma associated with Shwachman-Diamond syndrome. Pediatric Blood & Cancer. 2013;60(5):754-760. 503. Nakaya T, Kurata A, Hashimoto H, et al. Young-age-onset pancreatoduodenal carcinoma in Shwachman-Diamond syndrome. Pathology international. Feb 2014;64(2):75-80. 504. Hanks S, Coleman K, Reid S, et al. Constitutional aneuploidy and cancer predisposition caused by biallelic mutations in BUB1B. Nature genetics. 2004;36(11):1159-1161. 505. de Voer RM, van Kessel AG, Weren RD, et al. Germline mutations in the spindle assembly checkpoint genes BUB1 and BUB3 are risk factors for colorectal cancer. Gastroenterology. 2013;145(3):544-547.

214 506. Suijkerbuijk SJ, van Osch MH, Bos FL, Hanks S, Rahman N, Kops GJ. Molecular causes for BUBR1 dysfunction in the human cancer predisposition syndrome mosaic variegated aneuploidy. Cancer research. 2010;70(12):4891-4900. 507. Krokan HE, Bjoras M. Base excision repair. Cold Spring Harbor perspectives in biology. Apr 2013;5(4):a012583. 508. Al-Tassan N, Chmiel NH, Maynard J, et al. Inherited variants of MYH associated with somatic G:C-->T:A mutations in colorectal tumors. Nature genetics. Feb 2002;30(2):227-232. 509. Nielsen M, Joerink-van de Beld MC, Jones N, et al. Analysis of MUTYH genotypes and colorectal phenotypes in patients With MUTYH- associated polyposis. Gastroenterology. Feb 2009;136(2):471-476. 510. Win AK, Dowty JG, Cleary SP, et al. Risk of Colorectal Cancer for Carriers of Mutations in MUTYH, With and Without a Family History of Cancer. Gastroenterology.146(5):1208-1211.e1205. 511. Friedberg EC. How nucleotide excision repair protects against cancer. Nature Reviews Cancer. 2001;1(1):22-33. 512. Van Allen EM, Mouw KW, Kim P, et al. Somatic ERCC2 mutations correlate with cisplatin sensitivity in muscle-invasive urothelial carcinoma. Cancer discovery. Oct 2014;4(10):1140-1153. 513. Matsuda T, Bebenek K, Masutani C, Hanaoka F, Kunkel TA. Low fidelity DNA synthesis by human DNA polymerase-eta. Nature. Apr 27 2000;404(6781):1011-1013. 514. Haracska L, Yu SL, Johnson RE, Prakash L, Prakash S. Efficient and accurate replication in the presence of 7,8-dihydro-8-oxoguanine by DNA polymerase eta. Nature genetics. Aug 2000;25(4):458-461. 515. Broughton BC, Cordonnier A, Kleijer WJ, et al. Molecular analysis of mutations in DNA polymerase η in xeroderma pigmentosum-variant patients. Proceedings of the National Academy of Sciences. January 22, 2002 2002;99(2):815-820. 516. Johnson RE, Kondratick CM, Prakash S, Prakash L. hRAD30 mutations in the variant form of xeroderma pigmentosum. Science. Jul 9 1999;285(5425):263-265. 517. Hanks S, Perdeaux ER, Seal S, et al. Germline mutations in the PAF1 complex gene CTR9 predispose to Wilms tumour. Nat Commun. 08/07/online 2014;5. 518. Tomson BN, Arndt KM. The many roles of the conserved eukaryotic Paf1 complex in regulating transcription, histone modifications, and disease states. Biochimica et biophysica acta. Jan 2013;1829(1):116-126. 519. Iriyama T, Takeda K, Nakamura H, et al. ASK1 and ASK2 differentially regulate the counteracting roles of apoptosis and inflammation in tumorigenesis. The EMBO Journal. 02/12 08/23/received 01/21/accepted 2009;28(7):843-853. 520. Gaston D, Hansford S, Oliveira C, et al. Germline Mutations in MAP3K6 Are Associated with Familial Gastric Cancer. PLoS Genet. 2014;10(10):e1004669. 521. Online Mendelian Inheritance in Man, OMIM®. McKusick-Nathans Institute of Genetic Medicine, Johns Hopkins University (Baltimore, MD), {date}. World Wide Web URL: http://omim.org/.

215 522. Recommendations from the EGAPP Working Group: genetic testing strategies in newly diagnosed individuals with colorectal cancer aimed at reducing morbidity and mortality from Lynch syndrome in relatives. Genet Med. Jan 2009;11(1):35-41. 523. Castellone MD, Verrienti A, Magendra Rao D, et al. A novel de novo germ-line V292M mutation in the extracellular region of RET in a patient with phaeochromocytoma and medullary thyroid carcinoma: functional characterization. Clinical endocrinology. Oct 2010;73(4):529-534. 524. Richards FM, Payne SJ, Zbar B, Affara NA, Ferguson-Smith MA, Maher ER. Molecular analysis of de novo germline mutations in the von Hippel- Lindau disease gene. Human Molecular Genetics. November 1, 1995 1995;4(11):2139-2143. 525. Couch FJ, Johnson MR, Rabe KG, et al. The Prevalence of BRCA2 Mutations in Familial Pancreatic Cancer. Cancer Epidemiology Biomarkers & Prevention. February 1, 2007 2007;16(2):342-346. 526. Lucas AL, Shakya R, Lipsyc MD, et al. High prevalence of BRCA1 and BRCA2 germline mutations with loss of heterozygosity in a series of resected pancreatic adenocarcinoma and other neoplastic lesions. Clinical cancer research : an official journal of the American Association for Cancer Research. Jul 1 2013;19(13):3396-3403. 527. Grant RC, Selander I, Connor AA, et al. Prevalence of Germline Mutations in Cancer Predisposition Genes in Patients with Pancreatic Cancer. Gastroenterology. Dec 2 2014. 528. Kurian AW, Hare EE, Mills MA, et al. Clinical Evaluation of a Multiple- Gene Sequencing Panel for Hereditary Cancer Risk Assessment. Journal of Clinical Oncology. July 1, 2014 2014;32(19):2001-2009. 529. Johns AL, Miller DK, Simpson SH, et al. Returning individual research results for genome sequences of pancreatic cancer. Genome medicine. 2014;6(5):42. 530. Kilpivaara O, Aaltonen LA. Diagnostic Cancer Genome Sequencing and the Contribution of Germline Variants. Science. March 29, 2013 2013;339(6127):1559-1562. 531. Gracia-Aznarez FJ, Fernandez V, Pita G, et al. Whole Exome Sequencing Suggests Much of Non-BRCA1/BRCA2 Familial Breast Cancer Is Due to Moderate and Low Penetrance Susceptibility Alleles. PLoS ONE. 2013;8(2):e55681. 532. Snape K, Ruark E, Tarpey P, et al. Predisposition gene identification in common cancers by exome sequencing: insights from familial breast cancer. Breast cancer research and treatment. 2012;134(1):429-433. 533. Michalski CW, Weitz J, Buchler MW. Surgery insight: surgical management of pancreatic cancer. Nature Clinical Practice Oncology. Sep 2007;4(9):526-535. 534. Tutt A, Robson M, Garber JE, et al. Oral poly(ADP-ribose) polymerase inhibitor olaparib in patients with BRCA1 or BRCA2 mutations and advanced breast cancer: a proof-of-concept trial. Lancet. Jul 24 2010;376(9737):235-244. 535. Villarroel MC, Rajeshkumar NV, Garrido-Laguna I, et al. Personalizing cancer treatment in the age of global genomic analyses: PALB2 gene mutations and the response to DNA damaging agents in pancreatic cancer. Molecular cancer therapeutics. Jan 2011;10(1):3-8.

216 536. MacArthur DG, Balasubramanian S, Frankish A, et al. A Systematic Survey of Loss-of-Function Variants in Human Protein-Coding Genes. Science. February 17, 2012 2012;335(6070):823-828. 537. LaRusch J, Barmada MM, Solomon S, Whitcomb DC. Whole exome sequencing identifies multiple, complex etiologies in an idiopathic hereditary pancreatitis kindred. JOP : Journal of the pancreas. May 2012;13(3):258-262. 538. MacArthur DG, Manolio TA, Dimmock DP, et al. Guidelines for investigating causality of sequence variants in human disease. Nature. 04/24/print 2014;508(7497):469-476. 539. Alexandrov LB, Nik-Zainal S, Wedge DC, Campbell PJ, Stratton MR. Deciphering signatures of mutational processes operative in human cancer. Cell reports. Jan 31 2013;3(1):246-259. 540. Integrated genomic analyses of ovarian carcinoma. Nature. Jun 30 2011;474(7353):609-615. 541. Stadler ZK, Schrader KA, Vijai J, Robson ME, Offit K. Cancer genomics and inherited risk. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. Mar 1 2014;32(7):687-698.

217