Letter to the editor

ATM and PALB2 variants were recently associated with familial pancreatic cancer (FPC) using exome sequencing (ES). Roberts et al. identified germline ataxia-telangiectasia (AT)-associated ATM variants in two FPC kindreds through ES, then in an additional 4/166 FPC probands by Sanger sequencing [1]. Jones et al. identified a frameshift variant in PALB2 with somatic loss of heterozygosity in a patient with FPC by ES. Inactivating PALB2 variants were found in an additional 3/96 FPC kindreds [2].

We sequenced the germline exomes of 11 affected individuals from five FPC families to search for predisposing mutations (Additional file 1: Table S1). Among genes previously associated with FPC, we identified a nonsense single nucleotide variant (SNV) in ATM and a nonsense SNV in PALB2. These SNVs were absent in 47 FPC probands, 97 in-house controls, over 6,000 control exomes, and dbSNP135 (Additional file 2).

The proband in family 1 carried the nonsense SNV in ATM (c.C1931A; p.S644X), but her younger brother with FPC did not (Figure 1A; Additional file 1: Table S2). Sanger sequencing confirmed the exome calls and demonstrated loss of the variant allele in DNA from a metastasis (Figure 2A,B,C; Additional file 2). This SNV has not been previously associated with AT.

Figure 1
figure 1

Pedigrees of familial pancreatic cancer families, noting cancers, ages at diagnoses, and sources of diagnoses. (A) Family 1 (ATM c.C1931A; p.S644X). (B) Family 2 (PALB2 c.C3256T;p.R1086X). HX diagnosis from history, PA diagnosis from pathology reports, PSU primary site unknown, SCC squamous cell carcinoma, BCC basal cell carcinoma.

Full size image
Figure 2
figure 2

Sequencing confirmed the exome calls and demonstrated loss of the variant alleles in matched tumors. (A) ATM c.C1931A in the proband of family 1. (B) ATM c.C1931A in tumor of the proband of family 1, demonstrating loss of the variant allele. (C) Wild-type ATM c.C1931A in the brother with PC of the proband of family 1. (D) PALB2 c.C3256T in the proband of family 2. (E) PALB2 c.C3256T in the tumor from the proband of family 2, demonstrating loss of the variant allele. (F) Wild-type PALB2 c.C3256T in the cousin with PC of the proband of family 2.

Full size image

The proband in family 2 carried the nonsense SNV in PALB2 (c.C3256T; p.R1086X), but his younger cousin with PC did not (Figures 1B and 2D,E,F; Additional file 1: Table S2, columns d,f). Sanger sequencing of the primary tumor demonstrated loss of the variant allele (Figure 2E). A third cousin with PC was an obligate carrier of the SNV (Figure 1B), as was this cousin’s sister and mother who died from metastatic adenocarcinoma from unknown primary sites (Figure 1B). This SNV was previously associated with FPC [2].

Inactivating variants in ATM or PALB2 are carried by approximately 0.2% of the population [3], so finding inactivating variants in two of the five kindreds supports that these genes are associated with FPC. Since the original ES studies [1, 2], inactivating PALB2 mutations were found in small proportions of FPC cohorts (for example, Tischkowitz et al. [4]), and somatic ATM mutations were found in 8% of sporadic pancreatic cancers [5].

However, our findings question whether ATM and PALB2 predispose to FPC as ‘two hit’ tumor-suppressor genes. Both probands had somatic loss of the variant allele, not the wild-type allele. Moreover, both families had a relative with FPC who did not carry the mutation, with a younger age of onset than the affected carriers. These non-carrier relatives may be phenocopies, which is an important consideration for future ES studies. Alternatively, other factors may cause the familial clustering of pancreatic cancer in these kindreds.

ES is a promising technique to interrogate the genome in search of causes of complex diseases. However, ES generates thousands of candidates, and care is needed to avoid false associations. Alternative and complementary candidate gene discovery technologies such as whole-genome sequencing, copy-number analysis, and methylome analysis generate even more candidates. Since the potential for spurious findings is higher with more candidate variants, confirming associations with replication and functional studies is especially important.

Functional studies are yet to elucidate the roles of ATM and PALB2 in FPC and large prospective studies assessing their associations with FPC do not exist. These considerations, combined with our findings, currently limit the interpretability and utility of clinical sequencing of ATM and PALB2 in FPC.