Genomic analysis of familial pancreatic cancers and intraductal papillary mucinous neoplasms: A cross‐sectional study

Abstract Environmental and genetic factors play a critical role in the pathogenesis of pancreatic cancer, which is likely to follow a multistep process that includes intraductal papillary mucinous neoplasm. The pathogenesis of familial pancreatic cancer has been reported; however, epidemiological characteristics and causative genes remain unclear. This study aimed to determine the relationship between the family history of pancreatic cancer and tumor malignancy and identify novel susceptible germline variants of pancreatic cancer. We performed an epidemiologic study at our institute on a cohort of 668 patients with intraductal papillary mucinous neoplasm and 242 with pancreatic cancer but without associated intraductal papillary mucinous neoplasm stratified by family history of pancreatic cancer. Whole‐exome sequencing was conducted for 10 patients from seven families with familial pancreatic cancer and intraductal papillary mucinous neoplasm. We found that patients who had intraductal papillary mucinous neoplasm with positive family history of pancreatic cancer within first‐degree relatives were more likely to develop malignancy in a shorter period than those without family history. Duplicate frameshift variants in TET2 c.3180dupG (p.Pro1061fs) and ASXL1 c.1934dupG (p.Gly646fs) in one family and POLN c.1194dupT (p.Glu399fs) in another were identified as pathogenic truncating germline variants which were previously recognised susceptibility genes. Moreover, PDIA2 c.1403C>T (p.Pro468Leu) and DPYSL4 c.926C>A (p.Pro309Gln) were shared in four and two patients, respectively. In particular, PDIA2 was identified as a novel candidate for one of the deleterious variants of familial pancreatic cancer.


| INTRODUC TI ON
Pancreatic cancer is a malignant tumor with a poor prognosis, partly because of difficulty in detecting the cancer in early stages.
Delineation of risk factors for pancreatic cancer could help initiate surveillance targeting a subpopulation at greater risk. Previous studies have shown that both environmental and genetic factors play important roles in the pathogenesis of pancreatic cancers. Known environmental risk factors include obesity, diabetes, smoking, and alcohol consumption. 1 Documentation of familial aggregation of pancreatic cancer in the literature points to the influence of genetic factors. A prospective epidemiological study of pancreatic cancer conducted by Klien et al. 2 in 2004 showed that the lifetime risk of pancreatic cancer was 6.4 times higher in family members with two firstdegree relatives with pancreatic cancer than in those without first-degree relatives with pancreatic cancer. In addition, the risk was 32 times higher in family members with three first-degree relatives with pancreatic cancer. Sequencing analysis of germline variants of families with pancreatic cancer unraveled the critical role of ATM, BRCA1, BRCA2, and CHEK2. [3][4][5][6] Besides studies on familial clusters, genome-wide association studies demonstrated that the single-nucleotide variants rs13303010 in NOC2L at 1p36.33 and rs78193826 in GP2 at 16p12.3 are associated with pancreatic cancer even among patients without apparent family histories. 7,8 Similar to other cancers, the pathogenesis of pancreatic cancers is likely a multistep process. The observation that familial pancreatic cancer was more likely to have earlier onset and mortality supports the multistep progression of pancreatic cancers. Intraductal papillary mucinous neoplasm (IPMN) has been regarded as the critical precancerous lesion of pancreatic cancers. A recent whole-exome analysis of 350 patients with IPMN showed that germline variants in ATM and BRCA2 are associated with the progression of IPMN to pancreatic cancer. 9 However, whether genetic factors play a role in the development of IPMN remains unclear.
In this study, we performed epidemiologic and genetic studies on 668 patients with IPMN and 242 patients with pancreatic cancer but without associated IPMN. Among these two groups, we identified 18 patients with two or more affected family members, excluding the proband. Among these 18 patients, 10 patients from seven families underwent genomic studies. Identification of new risk genes for pancreatic cancer may facilitate genomic screening for early detection and treatment of pancreatic cancer.

| Patients and data collection
We retrospectively reviewed the clinical characteristics and environmental factors of all patients who underwent endoscopic ultrasound (EUS) and endoscopic retrograde cholangiopancreatography between 2012 and 2020 and identified 668 patients with IPMN and 242 patients with non-IPMN-associated pancreatic cancer ( Figure S1). We defined non-IPMN-associated pancreatic cancer as pancreatic cancer with no evidence of pancreatic cysts or IPMN based on magnetic resonance cholangiopancreatography (MRCP) or EUS scan interpreted by the radiologist and endoscopist.
The patients' clinical backgrounds were collected from medical records, and a questionnaire was provided during the first visit. Data collected included age, sex, body mass index (BMI), history of smoking and habitual alcohol consumption, medical history, family history of diabetes and cancer within first-or seconddegree relatives, and imaging findings of pancreatic lesions.
Habitual alcohol consumption was defined as the consumption of more than 100 g of alcohol per week. History of diabetes was  Table S1.
Patients with IPMN were classified as high risk or low risk according to the IPMN International Clinical Practice Guidelines 2017. 10,11 The high-risk IPMN group included those who met the criteria for "worrisome features (imaging findings include cyst of ≥3 cm, enhancing mural nodule <5 cm, thickened enhanced cyst walls, main pancreatic duct [MPD] 5-9 mm, lymphadenopathy, an elevated serum level of carbohydrate antigen, and a rapid rate of cyst growth >5 mm per 2 years)" and "high-risk stigma (obstructive jaundice, enhanced mural nodule ≥5 mm, MPD ≥10 mm)." Furthermore, those with intraductal papillary mucinous carcinoma (IPMC)-, IPMN-concomitant, and IPMN-derived pancreatic cancer were included in the high-risk IPMN group. The low-risk IPMN group comprised patients with IPMN who did not meet the criteria defined above.

| Whole-exome sequencing of germline samples
After obtaining written informed consent, whole-exome sequencing was performed on peripheral blood samples of 10 patients from seven families using the NovaSeq platform (Illumina) and Sure Select XT Human All Exon V6 (Agilent Technologies). Mapping of the sequenced reads to the reference human genome (GRCh37) and variant calling were performed according to the best practice guidelines of the Burrows-Wheeler Aligner 12 and the Genome Analysis Tool Kit, 13 as packaged in the integrated analysis suite variant tools. 14 The variants were annotated with SnpEff. 15

| Annotation of variants
To characterize potential functional significance of the variants revealed by whole-exome analysis, the allele frequencies of the variants among Japanese patients were evaluated from epidemiological standpoints. Any variant for which the allele frequency among >7000 normal Japanese individuals was larger than 0.03 as per the ToMMo database was excluded from further consideration. 16 Thus, only variants for which the allele frequency was between 40% and 60% were retained. When two or more affected members including the proband were tested, only the shared variants were retained.
Among the 10 patients from seven families, 4148 variants were retained according to these criteria.

| Search for pathogenic variants in previously recognized susceptibility genes
We extracted frameshift, nonsense, and splicing variants and filtered them by previously reported pancreatic cancer-related genes (Table S2). Variants corresponding to a combined annotationdependent depletion (CADD) score of >20 were extracted. 17

| Comparison with molecular epidemiological data from Japan
We further evaluated these filtered (nonsynonymous) variants based on our previous work on whole-genome analysis of samples from Biobank Japan. We analyzed the

| Statistical analyses
Categorical variables were compared between the two groups using the chi-squared and Fisher's exact tests. The Mann-Whitney U test was used to compare quantitative variables. Statistical analyses were performed using SPSS for Mac (version 25.0; IBM). Statistical significance was set at p < 0.05, and all tests were two-sided. Kaplan-Meier analysis was used to assess differences in survival between cohorts.

| Clinical characteristics of patients with IPMN and pancreatic cancer with a strong family history of pancreatic cancer
Among the 668 patients with IPMN and the 242 with non-IPMNassociated pancreatic cancer, we identified 15 patients with IPMN and three patients with non-IPMN-associated pancreatic cancer who had two or more affected family members excluding the proband (i.e., two first-degree relatives or one first-degree and one second-degree relative with pancreatic cancer; Figure S2). These patients were arbitrarily defined as "patients with a strong family history." The clinical characteristics of patients with IPMN and pancreatic cancer with a strong family history of pancreatic cancer (n = 18) are shown in Table 1

| Malignant progression of IPMN with positive family history of pancreatic cancer
We compared the clinical backgrounds of the patients with IPMN and those with non-IPMN-related pancreatic cancer by categorizing them into those with positive family history of pancreatic cancer (more than one first-degree relative with pancreatic cancer) and those without ( Table 2). Patients with IPMN-related pancreatic cancer with positive family history of pancreatic cancer were more likely to be female and to have a personal history of cancer; however, this finding was not statistically significant ( Table 2).
Progression to malignant pancreatic cancer or high-risk IPMN (including high-grade IPMN, IPMC, and IPMN-related pancreatic cancer) was evaluated in these subgroups using Kaplan-Meier analysis ( Figure 1). We observed 242 patients with high-risk IPMN during the study period, of which 23 had a positive family history of pancreatic cancer and 219 had no family history of pancreatic cancer.
We found that patients with IPMN and positive family history were more likely to develop malignancy in a shorter period than those with no family history of pancreatic cancer (log-rank test, p = 0.006).

| Families with truncating variants in known pancreatic cancer-related genes
Of the 18 patients with a strong family history, 11 patients underwent genomic analysis, five patients died, and two did not wish to undergo testing.
One patient (#7) underwent commercial testing for multi- Whole-exome sequencing was performed for 10 patients from seven families. The summary of the whole-exome sequencing results is shown in Table 3. Truncating variants of genes already implicated in the pathogenesis of pancreatic cancer (Table S1)  c.1934dupG, p. Gly646fs) were identified in patient #16 (Table 3a and Figure 2A), and a frameshift variant of NM_181808.4 (POLN; c.1194dupT, p. Glu399fs) was identified in patient #10 (Table 3b and Figure 2B).

| Comparison with molecular epidemiological data from Japan
We then compared the 4148 variants extracted from the analyzed patients to the allele frequencies of the whole-genome sequencing data from cancer (n = 1521) and noncancer patients (n = 6206).
In most families, only the probands were tested. Intrafamilial segregation was confirmed in two families.

| DISCUSS ION
In this study, we found deleterious variants in cancer susceptibility genes, including TET2, ASXL1, and POLN, which are known susceptibility genes, and PDIA2 and DPSYL4, which are novel risk factors.
Premature truncating germline variants in TET2 and ASXL1 have been identified in patients with familial pancreatic cancer. 5,6 Tet methyl cytosine dioxygenase 2 (TET2) is involved in DNA demethylation, and its loss-of-function mutations result in hypermethylation. 21 Meanwhile, ASXL1 is a histone modifier, and gain-of-function mutations are associated with tumorigenesis. 22 Somatic variants in TET2 and ASXL1 represent poor prognostic indicators in hematological tumors. 23 It is notable that a patient with strong family POLN gene is a member of the DNA polymerase family and is responsible for repairing DNA damage. 24 Truncating variants of POLN have been previously reported as causative genes of germline variants in pancreatic cancer. To our knowledge, this is the first study to report that a truncating variant (i.e., a frameshift variant) of POLN could be associated with both IPMN and pancreatic cancer. 5,25 Two variants observed in patients with a strong family history were previously reported to have a significantly higher frequency in patients with various cancers than in the general population.
First, the protein disulfide-isomerase A2 (PDIA2), a protein expressed in the pancreas, is involved in the regulation of cellular levels and biological functions of estrogen and has been identified as an autoantibody for autoimmune pancreatitis in vivo. 26,27 Although the association between a pathogenic PDIA2 variant and pancreatic cancer has not yet been reported, all four patients in the two families with this variant were females with a strong family history, which could lead to pancreatic cancer or cystic changes. Second, DPYSL4 is an intracellular metabolic regulator induced by p53, a known tumor suppressor, and is found in mitochondria of mast cells. 28 However, its relatively high allele fre-  In conclusion, in our cohort of 668 patients with IPMN and 244 with pancreatic cancer, we identified 18 patients with a strong family history who are at high risk of progression to pancreatic cancer.
Genomic analysis of these patients identified three previously recognized susceptibility genes and a novel potential candidate, PDIA2.

ACK N OWLED G EM ENTS
This study was supported by the Kenko Kagaku Zaidan (grant number KenkoKagakuZaidan201902). We would like to thank Kazumasa