Prevalence of Pathogenic Mutations in Cancer Predisposition Genes among Pancreatic Cancer Patients

Pancreatic ductal adenocarcinoma (PDAC), which constitutes the vast majority of pancreatic cancer, has a poor prognosis with an overall 5-year survival rate of 6% (1). In the United States, pancreatic cancer is the fourth most common cause of cancer death (2). Pancreatic cancer is a component of hereditary breast–ovarian cancer syndrome (HBOC; refs. 3, 4), Lynch syndrome (5, 6), familial adenomatous polyposis (7), familial atypical multiple mole melanoma syndrome (8), hereditary pancreatitis (9), Peutz–Jeghers syndrome (10), and Li–Fraumeni syndrome (11). These syndromes have been well-defined in guidelines for care from the National Comprehensive Cancer Network (NCCN). Epidemiologic studies have suggested that 10% to 20% of pancreatic cancers are associated with an inherited component, with familial pancreatic cancer (FPC), defined as kindreds containing at least two affected first-degree relatives, an established entity of inherited disease (12). Although germline studies have focused on single cancer predisposition genes (13), the prevalence of pathogenic mutations in BRCA1, BRCA2, PALB2, and CDKN2A has recently been estimated at 8% among patients with FPC and 3.5% among families with less pancreatic cancer history (14). The first panel-based study of 13 cancer predisposition genes among patients with pancreatic cancer identified 11 pathogenic mutations (3.8%) in ATM, BRCA1, BRCA2, MLH1, MSH2, MSH6, and TP53 (15). Thus, determination of the contributions of predisposition genes to pancreatic cancer has important ramifications for patients and their relatives. Here, we report on mutation screening of 96 patients with PDAC, recruited to the Mayo Clinic pancreatic cancer patient registry over a 12-month period, with a 22-gene panel to determine the prevalence of mutations in these genes.

Patients with pancreatic cancer were identified at time of first presentation with a possible pancreatic cancer, often prior to a documented diagnosis through pancreatology, surgery, and oncology clinics and consented to a Mayo Clinic Institutional Review Board–approved prospective pancreas patient registry (13). PDAC was confirmed by medical record review and pathologic diagnosis. Patients with other diagnoses or nonadenocarcinoma histologies were excluded. Blood samples, risk factor, family history questionnaires, and access to medical records were requested. Approximately 80% of patients with confirmed PDAC participated in the registry. A sequential series of 96 patients with PDAC enrolled between June 1, 2013, and June 1, 2014, was included in this study.

Germline DNA samples underwent custom capture (Agilent eArray) for all coding sequences, intron/exon boundaries, and partial nonrepetitive intronic regions from 21 cancer predisposition genes (BRCA1, BRCA2, PALB2, ATM, BARD1, BRIP1, RAD51C, RAD51D, CHEK2, MRE11A, NBN, RAD50, MLH1, MSH2, MSH6, PMS2, CDH1, TP53, PTEN, STK11, and FANCM). CDKN2A was also screened because the gene has been implicated in pancreatic cancer (16). Products from each capture were sequenced on a HiSeq2000 (Illumina) and likely pathogenic variants were validated by Sanger sequencing.

Paired end reads (100 bp) were aligned to the hg19 reference human genome using Novoalign (Novocraft Technologies). Realignment and recalibration were performed with GATK (https://www.broadinstitute.org/gatk). Germline variants were called with GATK Unified Genotyper. Copy-number variation (CNV) was assessed by “PatternCNV” (17). Coverage was evaluated by Integrative Genomics Viewer (IGV; ref. 18). The χ² test, t test, and Fisher exact test were used for association studies.

Table 1 displays the baseline characteristics of the 96 patients in the study. Ninety-three of 96 (97%) were non-Hispanic white. No patients reported Ashkenazi Jewish ancestry. The mean age of diagnosis of PDAC was 66.7 years (range, 41–90 years), and 53% of patients reported a history of smoking. Of the 96 patients, approximately 5% had a personal history of breast, 2% colorectal, and 2% endometrial cancer, whereas 7% had first-degree relatives with pancreatic cancer and met the criteria for FPC. Another 10% had second-degree relatives with the disease. Breast cancer was observed in first-degree relatives for 14% and first- or second-degree relatives of 21%. Colorectal (8%), ovarian (3%), or endometrial cancers (6%) were also reported among first-degree relatives.

A total of 14 likely pathogenic mutations were identified and confirmed by Sanger sequencing in 13 patients (14%; Table 2). Two frameshift and two nonsense mutations were detected in ATM. Two truncating mutations were observed in BRCA2, but the K3326X (c.9976A>T) variant that is associated with a low risk of cancer was excluded (19). One frameshift and one splicing defect were detected in MSH6, and two 1100delC mutations were found in CHEK2. Protein truncating mutations were also detected in BRCA1, BARD1, and FANCM. Two mutations had not previously been reported (ATM c.6012_6013insA and FANCM c.2586_2589delAAAA). One patient had mutations in both NBN and CHEK2 (Table 2). Nine of 96 (9%) patients had mutations in established pancreatic cancer predisposition genes. Interestingly, no mutations were identified in the PALB2 pancreatic cancer susceptibility gene (14). The youngest age at diagnosis for patients with likely pathogenic mutations was 51 years (BRCA2: c.6373dupA). There was no difference in mean age at diagnosis between 83 noncarriers [66.2 years; 95% confidence interval (CI), 63.9–68.3] and either the 13 mutation carriers (69.5 years; 95% CI, 63.9–74.9; P = 0.27) or carriers of nine mutations in established predisposition genes (P = 0.93).

Mutation carriers displayed a marginal association with FPC (P = 0.056) and a significant association with breast cancer in first-degree relatives (P = 0.018; Table 3). Including first- and second-degree relatives with cancer yielded associations with family history of breast (P = 0.002) and pancreatic cancers (P = 0.023; Table 3). Furthermore, breast and pancreatic cancers combined (P = 0.003); breast/pancreatic/ovarian/colorectal combined (P = 0.037); and breast/pancreatic/ovarian/colorectal/endometrial cancer combined (P = 0.05) all showed associations with mutation status (Table 3). However, there was no significant association between mutations and personal history of these cancers (P = 0.167; Table 3).

No large deletions or rearrangements were identified in the 22 genes. However, 161 missense variants and in-frame deletions were identified (Supplementary Table S1). Each variant was evaluated for potential pathogenic effects with Veste3 (score > 0.5), MetaSVM (score > 0.5), and MetaLR (score > 0.5) in silico prediction models that exhibit approximately 90% sensitivity and specificity for known pathogenic missense variants in ClinVar. Fourteen variants were consistently predicted to disrupt protein activity, but two of these were classified as benign (Supplementary Table S2). Carriers of 10 of the 12 remaining variants of uncertain significance (VUS) did not have a family history of breast or pancreatic cancer. However, CDKN2A c.350T>C was associated with a family history of breast cancer and CDKN2A c.272T>A from the same patient, as MSH6 c.3040_3042delAAG was associated with a family history of pancreatic cancer. In addition, the MLH1 c.1832T>C and MSH2 c.905T>C carriers both had a family history of colorectal cancer and the CDH1 variant carrier had a family history of endometrial cancer. In addition, the MLH1 c.1832T>C and MSH2 c.905T>C carriers had a family history of colorectal cancer and the CDH1 variant carrier had a family history of endometrial cancer. Thus, up to 12 additional missense variants may predispose to pancreatic and other cancers.

Panel-based mutation screening of 22 known cancer predisposition genes identified 14 likely pathogenic mutations in eight genes among 13.5% of patients with PDAC. Of these, nine (9.4%) were in the established ATM, PALB2, BRCA1, BRCA2, and MSH6 high- and moderate-risk pancreatic cancer genes. Another 12 missense variants were predicted pathogenic by in silico models in 11 patients. These results suggest that 9.4% to 25% of patients with pancreatic cancer in the Mayo Clinic registry may carry predisposing alleles. This contrasts with 11 (3.8%) mutations in 13 genes identified by the Ontario Pancreas Cancer Study (OPCS) of PDAC with family history of breast, ovarian, or pancreatic cancer and a subset without a family history of these cancers (15). The enrichment of mutations in the Mayo Clinic patient registry may reflect the ascertainment of the patients, although the rapid ascertainment approach coupled with screening of a sequential series of patients enrolled from June 2013 to June 2014 ensures that the study population is representative of PDAC cases seen at Mayo Clinic. In addition, the 96 selected cases were not enriched for family history of breast or pancreatic cancer relative to OPCS (21% vs. 19.7% breast; 12% vs. 13.9% pancreas). However, 33% of the Mayo cases with family history of breast cancer had mutations, whereas only 10.7% of similar cases in OPCS had mutations. The reason for the higher mutation rate is unclear.

The finding that mutations in the Mayo Clinic registry were significantly associated with family history of breast/pancreatic cancer and breast/pancreatic/colorectal/ovarian/endometrial cancer in first- and second-degree relatives strongly suggests that the current definition of FPC, which focuses on pancreatic cancer among first-degree relatives only, excludes many of the individuals with mutations in cancer predisposition genes that account for a significant proportion of pancreatic cancer cases. Identification of other pancreatic cancer predisposition genes may be better served by studies of families with a broader, clinically relevant definition of FPC and a wide spectrum of cancers.

BRCA2 is an important pancreatic cancer predisposition gene with 6% to 10% of patients with FPC carrying BRCA2 mutations (13). The relative risk of pancreatic cancer among BRCA2 mutation carriers has been estimated at 3.5 to 5 (20). However, the two patients with BRCA2 mutations in this study did not fit FPC criteria. Few studies have implicated BRCA1 in pancreatic cancer (3, 4). Here, one mutation in BRCA1 (c.212+1G>A) that disrupts normal splicing was observed. ATM pathogenic mutations are frequent (1%) in the general population (21) and among patients with FPC (2.4%; ref. 22). Four of 14 mutations (31%) in the 96 sequential cases were likely pathogenic ATM mutations. Three carriers had a family history of breast, pancreatic, or colorectal cancer. Consistent with the link between predisposition genes and double-strand DNA break repair signaling, single mutations were also identified in BARD1 and NBN. Neither gene has been established as a pancreatic cancer predisposition gene, but both have been implicated in breast and ovarian cancer. Further studies are needed to define the role of these genes in pancreatic cancer.

DNA mismatch repair (MMR) genes have been associated with an elevated risk of pancreatic cancer (23). Here, two likely pathogenic mutations in MSH6 and seven predicted pathogenic missense VUS in MMR genes were identified, suggesting that germline MMR gene mutations may be as common as ATM mutations among pancreatic cancer cases. CHEK2 mutations have been implicated in multiple cancers and may contribute to FPC (24). Neither of the CHEK2 mutation carriers identified in this study, one with an NBN inactivating mutation, had a family history of breast, pancreatic, ovarian, or colorectal cancer. Thus, the contribution of CHEK2 to pancreatic cancer risk needs further exploration. Fanconi anemia complementation gene M (FANCM) was recently implicated in cancer susceptibility (25). Whether the FANCM c.2586_2589delAAAA mutation detected here confers an increased risk of cancer remains to be determined.

Overall, a high proportion of patients with pancreatic cancer with a family history of a variety of solid tumors carry likely pathogenic mutations in known cancer predisposition genes. The suggestion is that pancreatic cancer is part of a constellation of tumors resulting from mutations in these genes and that studies focused on rare families predominantly enriched for pancreatic cancer may underestimate the contribution of predisposition genes to this disease. Importantly, relatives of individuals with pancreatic cancer who are found to carry mutations in high-risk predisposition genes such as BRCA1, BRCA2, MMR genes, ATM, and PALB2 may also benefit from clinical testing for family mutation/s and subsequent awareness, screening, and perhaps prevention options for several types of cancer.

No potential conflicts of interest were disclosed.

Conception and design: G.M. Petersen, R.R. McWilliams, F.J. Couch

Development of methodology: S.N. Hart, Y.K. Lee, F.J. Couch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Hu, W.R. Bamlet, K. Nandakumar, B.W. Eckloff, G.M. Petersen, F.J. Couch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Hu, S.N. Hart, W.R. Bamlet, R.M. Moore, K. Nandakumar, G.M. Petersen, R.R. McWilliams, F.J. Couch

Writing, review, and/or revision of the manuscript: C. Hu, S.N. Hart, W.R. Bamlet, G.M. Petersen, R.R. McWilliams, F.J. Couch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Hu, R.M. Moore, F.J. Couch

Study supervision: F.J. Couch

This work was supported by an NCI Specialized Program of Research Excellence (SPORE) in Pancreatic cancer (P50-CA102701) and Mayo Clinic Comprehensive Cancer Center Support Grant (P30 CA15083) to G.M. Petersen, NIH grant CA116167 to F.J. Couch.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.