Contribution of Inherited Mutations in the BRCA2-Interacting Protein PALB2 to Familial Breast Cancer

Discovery and characterization of the breast cancer susceptibility genes BRCA1 and BRCA2 have led to major changes in both prevention and treatment. Genetic testing for inherited mutations offers the opportunity for risk-reducing intervention (1). Therapeutic approaches that exploit the biological function of BRCA1 and BRCA2 have been proposed (2) and are now showing promise in the clinic (3, 4). Given this experience, there has been a great deal of interest in the identification and characterization of other genes responsible for inherited breast cancer (5). Among these genes is PALB2, partner and localizer of BRCA2 (6), first characterized clinically in patients with Fanconi anemia, complementation group N (7, 8). It was soon discovered that heterozygosity for loss-of-function mutations at PALB2 increases risk of developing breast cancer 2- to 6-fold (5, 9). Inherited PALB2 mutations associated with increased risks of developing breast cancer have been identified in families from many parts of the world (10–18), but thus far, a heterogeneous American population has not been screened.

The purpose of the present project was to estimate the contribution of inherited mutations in PALB2 to familial breast cancer in a large series of patients from the United States and to characterize the spectrum of inherited breast cancer–associated mutations in PALB2 in this heterogeneous population. For this purpose, we sequenced the complete coding and flanking regulatory regions of PALB2 from constitutional DNA of 1,144 familial breast cancer patients, all previously determined to have wild-type sequences at BRCA1 and BRCA2.

Participants were patients with a primary diagnosis of invasive breast cancer at any age and of any histologic subtype and with at least 2 first- or second-degree relatives with invasive breast cancer (19). Two series of participants were identified: one series identifying their ancestry as Ashkenazi Jewish (AJ) and the other series not selected for any specific ancestry. The series of AJ patients was enrolled to evaluate the possibility of founder mutations in this population in either known or novel breast cancer genes. Enrollment of probands, ascertainment of families, and genomic analysis proceeded identically for the 2 series. Potential participants were referred by physicians, genetic counselors, or themselves. All potential participants were interviewed prior to enrollment by a certified genetic counselor (J.B.M.) and informed consent was obtained for those meeting these criteria and interested in enrolling. The consenting participant became the proband for her or his family. For each proband, all genetically informative adult relatives, whether or not affected with breast cancer, were also requested to enroll. Relatives were contacted first by the proband and then interviewed by the genetic counselor, and informed consent obtained for those interested. Enrollment began in 1988 and is ongoing. All families in this study have been followed from their time of enrollment to the present by the genetic counselor. For all probands, BRCA1 and BRCA2 had been determined to be wild type, in almost all cases, on the basis of commercial sequencing and BART analysis by Myriad Genetics (20). The 2 series combined included 1,144 familial breast cancer patients without mutations in BRCA1 or BRCA2, of whom 172 participants (all female) were of AJ ancestry and 972 participants (959 females and 13 males) were of other ancestries. The study was approved by the University of Washington Human Subjects Division (IRB protocol 34173).

PALB2 exons, flanking intronic regions (50–100 bp in length), and 5′- and 3′-untranslated regions (UTR) were evaluated by Sanger sequencing of constitutional genomic DNA from all subjects. Genomic DNA isolated from peripheral blood lymphocytes was amplified by PCR by using flanking intronic primers (Supplementary Table S1). Nested PCR and 4 overlapping amplicons were developed to fully cover the 1,473 bp of exon 4. Multiple internal primers were used to sequence exon 5 in both directions. Amplicons were sequenced in both forward and reverse directions except as follows: for exon 9, to overcome the chances of nonspecific products due to Alu sequences upstream of the splice site, the exon was amplified using nested primers from an outer product and then cycle sequenced from the reverse direction. Similarly, exon 13 was sequenced only in the reverse direction. PCR products were purified and cycle sequenced using the BigDye Terminator Cycle Sequencing chemistry (Applied Biosystems by Life Technologies) and analyzed on an ABI PRISM 3700 Genetic Analyzer (Applied Biosystems by Life Technologies). All sequence variants were confirmed by replicate Sanger sequence and then evaluated for cosegregation with breast cancer in the family of the proband.

RNA and cDNA were isolated and transcript lengths were evaluated as previously described (19). The possibility of nonsense-mediated decay was evaluated by comparing electropherogram peak heights of mutant and wild-type alleles from transcript sequences. Multiple splice variants resulting from single genomic mutations were quantified by cloning PCR-amplified cDNA products into pCR2.1-TOPO cloning vectors (Invitrogen), then by transforming competent Escherichia coli strains, and finally by sequencing individual clones. Cloning experiments were carried out in triplicate.

To genotype all informative relatives of each proband with a mutation, TaqMan assays were designed for PALB2 mutations c.168delTTGT, c.509delGA, c.757delCT, c.1240C>T, c.2386G>T, c.2559C>T, c.2686insT, c.2919delAA, and c.3022delC and restriction digests were designed for mutations c.196C>T (HinP1I), c.1653T>A (HpyCH4III), c.2718G>A (BsmAI), and c.3113G>A (BstEII). In addition to family members, 960 unrelated anonymous controls were evaluated for each mutation. For TaqMan assays, DNA was diluted to 20 ng, with all wells in the same assay containing the same amount of sample. Two controls without template and positive controls with the proband PALB2 mutation were run with every plate. After each real-time PCR amplification, an allelic discrimination plate was analyzed using Sequence Detection System software (Applied Biosystems), with the autocaller function enabled and 95% quality interval for allele calling. Fluorescence signals were plotted for test samples, for controls without template, and for positive controls, with x- and y-axes representing each allele. Genotypes were called on the basis of location on the scatterplot.

Genotypes of deceased family members and others not available for genotyping were called only if they could be reconstructed with certainty from surviving children, spouses, and siblings. No genotypes were imputed probabilistically.

To determine whether the mutation PALB2 c.3113 shared a common origin in the 5 patients in which it appears, all probands and their relatives carrying PALB2 c.3113 were genotyped with microsatellite markers on chromosome 16 developed for this purpose. The markers were selected to flank PALB2 at chr16:23,614,483-23,652,678 (hg19). The microsatellites were AC23 at chr16:21,732,554-21,732,600 (8 alleles in our series), 23GT at chr16:23,037,671-23,037,716 (9 alleles in our series), and TATG14 at chr16:23,749,512-23,749,567 (4 alleles in our series). Microsatellite sequences, PCR primers, and genotyping conditions are reported in Supplementary Table S2.

Tumor sections of 5 to 8 μm from paraffin-embedded blocks were microdissected with the automated LCM Veritas System (Arcturus Molecular Devices, Life Technologies Corp) as previously described (21). Briefly, for each initial cap placement, laser was focused with the 7.5-μm spot size setting and the 10× objective. The pulse of the capture laser was adjusted to a 1,000- to 1,500-μs duration and 70 mW power. Microdissection was done with a medium laser spot size of 12.43 μm and a 10× objective. LOH was analyzed by direct sequencing of reconstituted tumor DNA at the site of the mutation. For all the PALB2 mutations for which tumor specimens were available, sequencing primers were designed to span maximum amplicon sizes of 250 bp (Supplementary Table S3). Sequencing was carried out as described earlier. Basecalls with PHRED scores of more than 30 were included.

Frequencies of mutations in subgroups were compared by χ² tests or Fisher's exact tests, as appropriate. Risk ratios (RR) and 95% CIs were calculated. Cumulative risks were estimated by Kaplan–Meier methods.

From genomic DNA of all 1,144 participants, complete PALB2 coding sequences and flanking regulatory regions were evaluated for nonsense mutations, frameshift mutations, and splice mutations leading to out-of-frame message deletions. Of the 972 breast cancer probands of unselected ancestries, 33 probands (3.4%) carried a truncating mutation in PALB2 (Table 1). Of the 13 male probands, 2 (16%) carried a truncating mutation; of the 959 female probands, 31 probands (3.2%) carried a truncating mutation. Of the 172 breast cancer probands of AJ ancestry, none carried such a mutation. Given the mutation frequency among patients of unselected ancestries, 5.5 carriers would be expected among the AJ patients (χ² = 5.9, P = 0.015). Thirteen different truncating mutations in PALB2 were observed, leading to stops in exons 3 to 11 (Fig. 1).

Identification of aberrant splice mutations was done on the basis of analysis of transcripts. For each sample with a single bp genomic variant in coding sequence or in flanking intronic regions, cDNA was generated from lymphoblast RNA by reverse transcriptase-PCR of PALB2 message. Of the 19 such variants observed in the cohort, 2 substitutions, c.2559C>T and c.3113G>A, led to altered splicing, out-of-frame deletions in the PALB2 message, and hence premature stops (Fig. 2). At PALB2 c.3113G>A in exon 10, three different messages were produced. Cloning and sequencing multiple messages indicated that 56% of mutant messages included an in-frame 117-bp deletion, 40% of messages included an out-of-frame 31-bp deletion, and 4% of messages were an immediate stop at residue 1,038 due to the creation of a nonsense codon at this site. Other rare PALB2 variants in the 1,144 probands included silent and missense substitutions in coding sequence, variants in the 5′-UTR and flanking intronic sequences, and variants at splice sites, leading to in-frame deletions in transcripts (Supplementary Table S4 and Supplementary Fig. S1).

As each PALB2 mutation was identified in a breast cancer proband, all available affected and unaffected relatives were genotyped. For deceased relatives, genotypes were reconstructed as possible from surviving adult children, spouses, and siblings. Genotypes were assigned only if they could be reconstructed with certainty. For 83 female relatives, affected and unaffected, with unambiguous PALB2 genotypes, cumulative risks of developing breast cancer were compared for those carrying a PALB2 mutation versus those with wild-type PALB2 genotypes. Ratios of these risks provided estimates of relative risks, cumulative by age, associated with PALB2 truncating mutations. Relative risks were 2.3-fold (95% CI: 1.5–4.2) by age 55 and 3.4-fold (95% CI: 2.4–5.9) at age 85, consistent with those previously reported (4).

Of the 13 PALB2 mutations identified in this series, 5 have been previously reported and 8 were novel (Table 1). Ancestries of patients carrying PALB2 mutations, as proportions of all patients of each ancestry in our series, are indicated in Table 2. In some of the patients of German, French, and English origin, we detected mutations originally discovered in patients of the same ancestries (7–9). A mutation previously discovered in a French Canadian family (12) was not detected in patients of that ancestry in our series. Three mutations offered particularly interesting histories. PALB2 c.3113G>A appeared in 3 patients of English ancestry and 2 patients of African American ancestry. These 5 patients shared a haplotype of length more than 2 Mb defined by microsatellite markers on chromosome 16 flanking PALB2. Among the patients' relatives who were wild type at c.3113G>A, none shared this complete haplotype. The allele-specific haplotype suggests a single origin for this mutation, which has previously been reported in British families (9). A second historically interesting mutation was c.509-510delGA, carried by 7 patients, all of whom reported German ancestry on their familial lineages with histories of breast cancer. Further exploration of these patients' pedigrees revealed a common ancestor who immigrated to the United States from Germany in the 19th century. PALB2 c.509-510delGA has also been reported in families from Poland (17), suggesting that this mutation may have a central European origin. Finally, the 2 patients carrying PALB2 c.196C>T shared a distant ancestor who immigrated from Scotland.

The cancer profiles of PALB2 families (Table 3) were similar to cancer profiles of BRCA2 families. Male or female patients with PALB2 mutations were significantly more likely than patients without PALB2 mutations to have a relative with male breast cancer or with pancreatic cancer. In contrast, PALB2 genotype was not significantly associated with the likelihood of having a relative with ovarian cancer or with age at onset of female breast cancer (50.0 ± 11.9 years among PALB2 heterozygotes vs. 50.2 ± 6.8 years among patients not carrying PALB2 mutations). Patients with PALB2 mutations were more likely to be probands of families with at least 6 cases of female breast cancer, reflecting the possibility that some families in our series with relatively few cases of breast cancer do not carry mutations in any susceptibility genes. It should be noted that most of the patients with PALB2 mutations (17/33) were not probands of these extremely high-incidence families.

For 7 female breast cancer patients with PALB2 mutations, paraffin-embedded breast tumor specimens were available, enabling dissection of tumor cells by laser-capture microscopy and evaluation of LOH at PALB2. In each of these specimens, loss of the wild-type PALB2 allele was observed (Fig. 3). At least for these patients, LOH of the wild-type PALB2 allele suggests that PALB2 acts as a conventional inherited tumor suppressor gene.

In a series of familial breast cancer patients of unselected ancestry, with wild-type sequences of BRCA1 and BRCA2, the prevalence of inherited PALB2 mutations was 3.4%. Thirteen different protein truncating mutations were identified, all of which were individually rare. Five mutations had been previously reported; 8 were encountered for the first time in this series. Both the occurrence of PALB2 mutations among male and female breast cancer patients and the 4-fold higher frequency of male breast cancer among relatives of PALB2 heterozygotes suggest that PALB2 predisposes to both male and female breast cancers. Cancer profiles of these families are also consistent with the suggestion that PALB2 is a susceptibility gene for pancreatic cancer (22). Ovarian cancer was more common among relatives of PALB2 heterozygotes (55% of PALB2 families included a relative with ovarian cancer) than among relatives of other cases (41% of non-PALB2 families included a relative with ovarian cancer), but this difference was not significant given this sample size. Direct evaluation of large numbers of ovarian cancer patients will be useful in determining the role of inherited mutations in PALB2 in this malignancy. The distributions of ages at diagnosis of breast cancer patients with PALB2 mutations were very similar to those of the other familial breast cancer patients in the series, suggesting that in the context of familial breast cancer, PALB2 does not lead to significantly younger diagnosis. It is very unlikely that there is a founder allele of PALB2 among patients of AJ ancestry, but any family, of any ancestry, could harbor a private cancer-predisposing allele of PALB2.

The prevalence of inherited mutations in PALB2 in non-BRCA1, non-BRCA2 familial breast cancer patients is approximately the same as the prevalence of inherited mutations in CHEK2 in a similar cohort (19). It has been suggested that genetic testing for CHEK2 c.1100delC in the clinical setting is now timely, with targeted surveillance and medical follow-up for mutation carriers (23). We agree and join others (24) in suggesting that similar consideration be extended to PALB2. The breast cancer risk associated with protein-truncating mutations in PALB2 seems greater than the risk associated with CHEK2 c.1100delC. The challenge is that in the United States, no one mutation of PALB2 is sufficiently frequent to represent a substantial fraction of PALB2 mutations so that the entire gene must be sequenced to capture the responsible alleles. Simultaneous detection of all mutations in all known breast cancer genes will make such expanded screening more feasible (25).

Testing for PALB2 and other breast cancer susceptibility genes will add complexity to the clinical interpretation of results. In particular, variants of uncertain significance will be identified. This study focused on PALB2 mutations leading to protein truncation. Individual missense mutations in PALB2 will ultimately need to be evaluated, as they have been for BRCA1 and BRCA2 (26, 27).

The possibility of genetic testing for PALB2 provides another example of the complexity of the transition from research laboratory to clinic. The involvement of medical geneticists and genetic counselors in assessing risk is particularly critical when, as here, consequences of genetic testing can entail substantial medical and surgical decisions (28–30).

No potential conflicts of interest were disclosed.

Supported by the Breast Cancer Research Foundation, Komen Foundation for the Cure, and NIH grants 5R01ES013160 (M-C. King, S.C. Casadei, B.M. Norquist, T. Walsh, S. Stray, J.B. Mandell, and M.K. Lee) and 5U01ES017156 (J.A. Stamatoyannopoulos).

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.