Rare germline mutations in PALB2 and breast cancer risk: A population-based study†
Communicated by Michael Dean
Abstract
Germline mutations in the PALB2 gene are associated with an increased risk of developing breast cancer but little is known about the frequencies of rare variants in PALB2 and the nature of the variants that influence risk. We selected participants recruited to the Women's Environment, Cancer, and Radiation Epidemiology (WECARE) Study and screened lymphocyte DNA from cases with contralateral breast cancer (n = 559) and matched controls with unilateral breast cancer (n = 565) for PALB2 mutations. Five pathogenic PALB2 mutations were identified among the cases (0.9%) versus none among the controls (P = 0.04). The first-degree female relatives of these five carriers demonstrated significantly higher incidence of breast cancer than relatives of noncarrier cases, indicating that pathogenic PALB2 mutations confer an estimated 5.3-fold increase in risk (95% CI: 1.8–13.2). The frequency of rare (<1% MAF) missense mutations was similar in both groups (23 vs. 21). Our findings confirm in a population-based study setting of women with breast cancer the strong risk associated with truncating mutations in PALB2 that has been reported in family studies. Conversely, there is no evidence from this study that rare PALB2 missense mutations strongly influence breast cancer risk. Hum Mutat 33:674–680, 2012. © 2012 Wiley Periodicals, Inc.
Introduction
PALB2, originally identified as a BRCA2-interacting protein, is crucial for key BRCA2 genome caretaker functions [Xia et al., 2006] and more recently has also been shown to interact with BRCA1 [Sy et al., 2009a,b; Zhang et al., 2009]. Germline mutations in PALB2 (MIM# 610355), are associated with increased risk of breast cancer [Rahman et al., 2007] and mutations have been identified in approximately 1% of hereditary breast cancer (HBC) families throughout the world, summarized in [Tischkowitz and Xia, 2010]. The first study showing an association with germline mutations in PALB2 and breast cancer estimated the relative risk to be 2.3 (95% confidence interval (CI): 1.4–3.9) [Rahman et al., 2007], but subsequent studies have suggested that the risk could be higher [Erkko et al., 2008; Southey et al., 2010]. In this investigation, we examine the breast cancer risk conferred by PALB2 mutations in a population-based case–control study of high-risk women, the Women's Environment, Cancer, and Radiation Epidemiology (WECARE) Study [Bernstein et al., 2004]. In this study, mutation frequencies in incident cases of contralateral breast cancer (CBC) are compared with the population from which CBC cases emerge, namely survivors of unilateral breast cancer (UBC). This study design is especially advantageous for studying rare genetic mutations, since the frequencies of true risk variants are higher in both cases and controls as both have breast cancer [Begg and Berwick, 1997]. Other HBC genes such as BRCA1/BRCA2 and the CHEK2*1100delC variant have been studied in this way and have demonstrated increased risk of CBC [Fletcher et al., 2009; Malone et al., 2010]. We screened the coding regions and adjacent intronic areas of the PALB2 gene in lymphocyte DNA from BRCA1/BRCA2-negative women participating in the WECARE Study to determine the mutation frequency in CBC cases compared to UBC controls.
Methods
Study Population
The WECARE Study is a population-based, nested case–control study of CBC, which has been described previously [Bernstein et al., 2004]. All WECARE Study participants were ascertained through one of five population-based cancer registries covering the country of Denmark along with the State of Iowa, Los Angeles County, Orange County and San Diego regions of California, and three Western Washington counties in the United States (the U.S. registries participate in the Surveillance, Epidemiology, and End Results [SEER] registry system). All participants were diagnosed before age 55 years between 1985 and 2000 with a first primary invasive breast cancer that had not spread beyond regional lymph nodes. Cases were diagnosed with a second primary invasive or in situ CBC at least one year after first primary diagnosis in the period 1986–2001; controls were women with UBC who did not develop a second primary breast cancer during the study period (until 2001). For this investigation, we selected cases and controls that had been previously screened for pathogenic BRCA1 and BRCA2 mutations by denaturing high-performance liquid chromatography [Malone et al., 2010] and excluded individuals with detectable pathogenic mutations in these genes according to the following criteria: (1) exon sequence changes predicted to truncate protein production; (2) splice site mutations located within two base pairs of an intron/exon boundary or shown to cause aberrant splicing; (3) missense changes with known pathogenic functional effects. Our final analytic group consisted of 559 CBC cases and 565 UBC matched controls and the characteristics of the study participants are given in Table 1.
| Cases (n = 559) | Controls (n = 565) | |||||
|---|---|---|---|---|---|---|
| Characteristic | N | % | N | % | ||
| Age at diagnosis, years | ||||||
| <30 | 6 | 1.1 | 7 | 1.2 | ||
| 30–34 | 14 | 2.5 | 15 | 2.7 | ||
| 35–39 | 51 | 9.1 | 50 | 8.9 | ||
| 40–44 | 138 | 24.7 | 137 | 24.3 | ||
| 45–49 | 178 | 31.8 | 180 | 31.9 | ||
| 50–54 | 172 | 30.8 | 176 | 31.1 | ||
| Median | 47 | 47 | ||||
| Range (interquartile) | 43–51 | 43–51 | ||||
| At-risk period, years | ||||||
| <5 | 318 | 56.9 | 320 | 56.6 | ||
| ≥ 5 | 241 | 43.1 | 245 | 43.4 | ||
| Race/ethnicity | ||||||
| Non-Hispanic White | 519 | 92.8 | 525 | 92.9 | ||
| Hispanic White | 14 | 2.5 | 17 | 3.0 | ||
| Black | 17 | 3.0 | 16 | 2.8 | ||
| Japanese | 1 | 0.2 | 1 | 0.2 | ||
| Chinese | 1 | 0.2 | 1 | 0.2 | ||
| Filipino | 5 | 0.9 | 5 | 0.9 | ||
| Other Asian/other | 2 | 0.4 | 0 | 0.0 | ||
| Jewish ancestry | ||||||
| No | 504 | 90.2 | 519 | 91.9 | ||
| Yes | 42 | 7.5 | 37 | 6.5 | ||
| Unknown | 13 | 2.3 | 9 | 1.6 | ||
| Family history of breast cancer | ||||||
| No first-degree relatives | 388 | 69.4 | 450 | 79.6 | ||
| Any first-degree relatives | 167 | 29.9 | 110 | 19.5 | ||
| Adopted or unknown | 4 | 0.7 | 5 | 0.9 | ||
| Family history of ovarian cancer | ||||||
| No first-degree relatives | 538 | 96.2 | 545 | 96.5 | ||
| Any first-degree relatives | 17 | 3.0 | 15 | 2.7 | ||
| Adopted or unknown | 4 | 0.7 | 5 | 0.9 | ||
| Geographic region | ||||||
| State of Iowa | 89 | 15.9 | 89 | 15.8 | ||
| Orange County and San Diego County, California | 93 | 16.6 | 94 | 16.6 | ||
| Los Angeles County | 150 | 26.8 | 149 | 26.4 | ||
| Three Western Washington counties | 85 | 15.2 | 86 | 15.2 | ||
| Country of Denmark | 142 | 25.4 | 147 | 26.0 | ||
| Menopausal status | ||||||
| Premenopausal | 203 | 36.3 | 185 | 32.7 | ||
| Postmenopausal, <45 years | 133 | 23.8 | 142 | 25.1 | ||
| Postmenopausal, ≥ 45 years | 222 | 39.7 | 235 | 41.6 | ||
| Unknown | 1 | 0.2 | 3 | 0.5 | ||
| Estrogen receptor status of first primary breast cancer | ||||||
| Positive | 290 | 51.9 | 310 | 54.9 | ||
| Negative | 126 | 22.5 | 113 | 20.0 | ||
| Other or unknown | 143 | 25.6 | 142 | 25.1 | ||
| Chemotherapy for first primary breast cancer | ||||||
| No | 322 | 57.6 | 273 | 48.3 | ||
| Yes | 237 | 42.4 | 292 | 51.7 | ||
| Adjuvant hormone therapy for first primary breast cancer | ||||||
| No | 394 | 70.5 | 369 | 65.3 | ||
| Yes | 165 | 29.5 | 195 | 34.5 | ||
| Unknown | 0 | 0.0 | 1 | 0.2 | ||
| Radiation therapy | ||||||
| No | 291 | 52.1 | 176 | 31.2 | ||
| Yes | 268 | 47.9 | 389 | 68.8 | ||
| Oophorectomy status | ||||||
| Yes | 35 | 6.3 | 50 | 8.8 | ||
| No | 523 | 93.6 | 514 | 91.0 | ||
| Unknown | 1 | 0.2 | 1 | 0.2 | ||
Molecular Methods
Genomic DNA was obtained from lymphocytes using standard extraction methods. The 13 coding exons of PALB2 (NCBI reference sequence NM_024675.3) were screened using high-resolution melt (HRM) analysis, a method with similar sensitivity to sequencing [Wittwer, 2009]. All work was carried out in a single laboratory using the conditions listed in Table 2. The PCR for HRM was performed in 5 µl total volume (2 µl mastermix, 0.5 µl LCGreen, 0.5 µl each of forward and reverse primers (2.5 µM stock concentration), 0.5 µl of double distilled sterile water, 1 µl of DNA (25 ng/µl). As indicated in Table 2, Q solution instead of water was used for the exon 1 PCR and 25 mM of MgCl2 was added to the PCRs for exons 5.1, 13.1, and 13.2. The PCR protocol was as follows (for 66°C annealing temperature): 95°C for 2 min, 95°C for 45 sec, 73°C 45 sec Touchdown (−1°C increment for seven cycles), followed by 95°C for 45 sec, 66°C for 45 sec for 35 cycles. The PCR products were then analyzed by HRM using the LightScanner instrument (Idaho Technologies, Salt Lake City, Utah) according to the manufacturer's protocol. Nucleotide numbering reflects cDNA numbering where +1 corresponds to the A of the ATG translation initiation codon in the reference sequence (NM_024675.3) and the initiation codon is codon 1. Variants were confirmed by Sanger sequencing and were classified as pathogenic using the same criteria as the BRCA1/BRCA2 study [Malone et al., 2010] mentioned above. Sequence changes that did not fulfill these criteria for being pathogenic were divided into three categories (1) missense variants, (2) exonic synonymous single nucleotide variants (“synonymous”), and (3) intronic variants. We further examined the effects of missense mutations, classified on the basis of bioinformatic tools that predict functional significance: Sorting Intolerant from Tolerant (SIFT) [Kumar et al., 2009], Polyphen [Adzhubei et al., 2010], and Align-GVGD [Tavtigian et al., 2008]. SIFT and Polyphen were used under default conditions and with the programs' internally generated alignments. Align-GVGD was used with a curated alignment in which the sequence furthest diverged from human was from the fish Danio rerio. This alignment, or updated versions thereof, is available at the Align-GVGD website http://agvgd.iarc.fr/alignments.
| Primer sequences | |||
|---|---|---|---|
| Exon | Forward | Reverse | Annealing temperature (°C) |
| 1 | GGATTTAATTGGCCGGAGTTTA | ACACAAAGCCAGGCCTAAAAC | 66a |
| 2+3 | CTGACTCCACCTTTCCACTTG | TACCTGGGAAATGAATAATAAAGCA | 66 |
| 4.1 | GCCTGAATGAAATGTCACTGATT | ATGTCCTCTTCTGCTGCTTCTTT | 64 |
| 4.2 | GGCCCAGGAGGATTACCTAT | TGGGCAGTTGGTGGAATTA | 66 |
| 4.3 | TAATTCCACCAACTGCCCA | GTTATTGTAGGTGAGTTCATTTAGAGAACA | 66 |
| 4.4 | TGTTCTCTAAATGAACTCACCTACAATAAC | AATTTCACTTTGGTCAGTTTCCTC | 64 |
| 4.5 | CAGAGTCATTTGGATGTCAAGAAA | TGTTAACAATCGACAGGCTAGAAGTT | 66 |
| 4.6 | CATCAGATCATTGTGAACCACT | GCCAGGCAAATAGTAATTGTTAACTT | 66 |
| 5.1 | TTTGTCATCAGTGAAACAGATTGTC | GACTCAGTTCCTCTGGAAAAATACA | 65b |
| 5.2 | TTAAGCTTGAAAAAGTGAAGTCCTG | GTTCGTCCAGCAACTTCTGTAGAT | 66 |
| 5.3 | GCTTTCCCCATCTTAGGTACTACTC | ACAGAGTCACAGTCACAGGTAGGTT | 66 |
| 5.4 | AACCTACCTGTGACTGTGACTCTGT | GCTGTTTACATTCACTAAGGCATTT | 64 |
| 6 | TAGACATTTTGATGAGTGGGTAATG | CCAAATCTGTTTTTCTGAATCTGTT | 64 |
| 7 | TCACTTTGCATACTTATGCTTT | GGTAAGCTGCCCATCTACATTATC | 64 |
| 8 | TTGGAAAATCTGGATTAAACAAAAA | ACTTAAAACCAGCTGACAGAGACAA | 64 |
| 9 | TTAAAAGGTTACTCCTCACATCAC | AACGAGATCCTAGTTACCCAACTTT | 66 |
| 10 | CCTAGAGACTGCTTTAGTGCAAAGT | TCTTCACAACAACCCTGTAAAATTAG | 66 |
| 11 | GTCACCTCCTAAGACATGCTATGAT | TGAGACCAACAGTAACACACAAAGT | 64 |
| 12 | TGTCTCTGCCAGATCTTTATTTTTC | TTTCCATTCTTCTAAGTGACACAAA | 64 |
| 13.1 | GTTTTGGGAACATGGTTTTGAC | TTTTAAGTGTCATTCAGATATTCTCC | 64b |
| 13.2 | CTGGATATATTGGGCCTCTTAGTA | AAAGAATGGGAAACAATAACATGC | 65b |
- aPlus Q solution (Qiagen, Valencia, California).
- bPlus 25 mM MgCl2.
Statistical Methods
First, we evaluated the broad case–control associations with respect to various, possibly overlapping mutation types using individuals who had wild-type PALB2 as the baseline and estimating the risk of carrying a deleterious, missense synonymous, or intronic mutation. Due to the small sample sizes in some of the analyses presented, we employed Fisher's exact test and exact (i.e., permutation based) methods for constructing confidence intervals using Cytel Studio© software (Cytel Inc., Cambridge, MA). The exact two-sided Cochran–Armitage trend test was employed to investigate the relationship between the case–control status and the bioinformatics tools classifications and the exact Kruskal–Wallis test was performed to test for differences in continuous patient characteristics by mutation type. We estimated the relative risk conferred by pathogenic PALB2 mutations by comparing the breast cancer incidence in their first-degree female relatives with the corresponding incidence in first-degree relatives of noncarrier CBC cases using conventional actuarial techniques, and then by transforming the resulting estimate [Saunders and Begg, 2003]. We further examined the relative risks of individual rare missense variants (MAF ≤ 1%) using a hierarchical modeling approach developed for this purpose [Capanu et al., 2011]. This model included adjustment for eight common polymorphisms (MAF > 1%), the presence of a truncation or splicing mutation, the presence of a rare intronic mutation, the presence of a rare synonymous mutation, as well as childbearing history as possible confounders. The effects of individual variants were adjusted on the basis of the average age at diagnosis of the carriers of the specific variant, family history of breast cancer in first-degree relatives for these carrier probands, as well as the three bioinformatic prediction tools.
Results
We screened the PALB2 gene in a total of 1,124 women with breast cancer and we identified five pathogenic mutations in the cohort of 559 women with CBC versus 0/565 women with UBC (P = 0.04), Table 3. Of the five mutations, three have been previously reported: c.1592delT is a known founder mutation in Finland [Erkko et al., 2007], while c.3113G>A and c.3549C>G have been reported by several groups (Table 4). Among the five carriers of these mutations, the median ages of the first and second breast cancers were 46 and 55 years, respectively, and all probands had at least one first-degree relative affected with breast cancer, although none had an additional young-onset case (under age 50 years) in the family. Nonetheless the breast cancer incidence in these 17 first-degree female relatives of PALB2 carriers was significantly higher than the corresponding incidence in first-degree relatives of noncarrier cases, and from this we derived a relative risk estimate for carriers of a pathogenic mutation of 5.3 (95% CI: 1.8–13.2). The majority of the breast cancers in these five women were estrogen receptor (ER) positive, progesterone receptor (PR) positive, infiltrating ductal type, with the notable exception of the carrier for the c.1592delT mutation where both tumors were ER negative, PR negative, the first being classified as medullary type (Table 4).
| All variants | Rare variants (MAF ≤ 1%) | ||||||
|---|---|---|---|---|---|---|---|
| Variant category | CBC cases | UBC controls | RR (95% CI) P-value | Variant category | CBC cases | UBC controls | RR (95% CI) P-value |
| Wild-type | 275 | 316 | Wild-type | 275 | 316 | ||
| Deleterious | 5 | 0 | P = 0.04* | Deleterious | 5 | 0 | P = 0.04* |
| Intronic | 255 | 211 | 1.4 (1.1–1.8) P = 0.01* | Intronic | 12 | 1 | 13.8 (2.0–591.3) P < 0.01* |
| Synonymous | 35 | 30 | 1.3 (0.8–2.3) P = 0.32* | Synonymous | 5 | 9 | 0.6 (0.2–2.2) P = 0.6* |
| Missense | 128 | 116 | 1.3 (0.9–1.7) P = 0.14* | Missense | 23 | 21 | 1.3 (0.7–2.4) P = 0.56* |
| SIFT | SIFT | ||||||
| Wild-type | 275 | 316 | 1.0 | Wild-type | 275 | 316 | 1.0 |
| Tolerated | 92 | 95 | 1.1 (0.8–1.6) | Tolerated | 18 | 18 | 1.1 (0.6–2.4) |
| Affect protein function | 35 | 21 | 1.9 (1.1–3.5) P = 0.04a | Affect protein function | 5 | 3 | 1.9 (0.4–12.4) P = 0.4a |
| Polyphen | Polyphen | ||||||
| Wild-type | 275 | 316 | 1.0 | Wild-type | 275 | 316 | 1.0 |
| Benign | 60 | 60 | 1.1 (0.8–1.7) | Benign | 6 | 10 | 0.7 (0.2–2.1) |
| Possibly damaging | 1 | 1 | Possibly damaging | 1 | 1 | ||
| Probably damaging | 66 | 55 | 1.4 (0.9–2.0) P = 0.1a | Probably damaging | 16 | 10 | 1.8 (0.8–4.6) P = 0.2a |
| GV | GV | ||||||
| Wild-type | 275 | 316 | 1.0 | Wild-type | 275 | 316 | 1.0 |
| GV > 0 | 95 | 96 | 1.1 (0.8–1.6) | GV > 0 | 21 | 19 | 1.3 (0.6–2.6) |
| GV = 0 | 32 | 20 | 1.8 (1.0–3.5) P = 0.05a | GV = 0 | 2 | 2 | 1.1 (0.1–15.9) P = 0.6a |
| GD | GD | ||||||
| Wild-type | 275 | 316 | 1.0 | Wild-type | 275 | 316 | 1.0 |
| GD < 65 | 97 | 96 | 1.1 (0.8–1.6) | GD < 65 | 22 | 20 | 1.3 (0.6–2.5) |
| GV ≥ 65 | 31 | 19 | 1.9 (1.0–3.6) P = 0.05a | GV ≥ 65 | 1 | 1 | 1.1 (0.0–90.5) P = 0.6a |
- *P-values using Fisher's exact test evaluating the risk of developing CBC among carriers of different mutation types relative to wild-type PALB2 subjects.
- aExact two-sided Cochran–Armitage trend test P-value. Missense variants were further characterized using SIFT [Kumar et al., 2009], Polyphen [Adzhubei et al., 2010], and Align-GVGD [Tavtigian et al., 2008].
- GV, or Grantham Variation, is a measure of range of variation at an amino acid position in a protein multiple sequence alignment. GV = 0 is indicative of an invariant position, and higher GVs are indicative of greater variability [Tavtigian et al., 2008].
- GD, or Grantham Deviation, is a measure of the severity of a missense substitution with respect to the range of variation at its position in a protein multiple sequence alignment. GV ≥ 65 is indicative of a substitution that is nonconservative with respect to the observed range of variation [Tavtigian et al., 2008].
| Tumor characteristics | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Exon | Mutation | Type | Age | Type | Grade | ER | PR | Family history of BC | Previously reported |
| 4 | c.1592delT, p.Leu531Cysfs*30 | Deletion frameshift | 47 | Med | 3/3 | Neg | Neg | Mother 70, | [Erkko et al., 2007] |
| 55 | IDC | 3/3 | Neg | Neg | Paternal aunt 63 | ||||
| 9 − 1 | c.2835-1G>C | Splice | 46 | IDC | 3/3 | Neg | Neg | Mother 78 | No |
| 53 | IDC | 2/3 | Pos | Pos | |||||
| 10 | c.3113G>A, p.Trp1038* | Stop/splice | 38 | IDC | U | Pos | Pos | Mother 75, Paternal aunt 78 | [Casadei et al., 2011; Garcia et al., 2009; Rahman et al., 2007; Southey et al., 2010; Wong et al., 2011] |
| 47 | IDC/Lob | U | Pos | Pos | |||||
| 11 + 1 | c.3202+1G>C | Splice | 46 | IDC | 2/3 | Pos | Pos | Sister 55 | No |
| 60 | IDC | U | Pos | Pos | |||||
| 13 | c.3549C>G, p.Tyr1183* | Stop | 53 | IDC | U | Pos | Pos | Sister 65, Mother 68 | [Hofstatter et al., 2011; Rahman et al., 2007; Reid et al., 2007] |
| 68 | IDC | 1/3 | Pos | Neg | |||||
- ER, estrogen receptor status; PR, progesterone receptor status; IDC, infiltrating ductal carcinoma; Med, medullary; Lob, lobular; U, unknown; Pos, positive; Neg, negative. (Reference sequence NM_024675.3.)
We found no statistically significant difference in the frequency of missense variants between cases and controls (these variants are listed in Supp. Table S1). However, when classified by their predicted degrees of pathogenicity using the three different models, SIFT [Kumar et al., 2009], Polyphen [Adzhubei et al., 2010], and Align-GVGD [Tavtigian et al., 2008], for which the criteria was set as Grantham Variation (GV) = 0 and Grantham Deviation (GD) ≥ 65, following the classification at http://agvgd.iarc.fr/classifiers.php), there were marginally significant trends for SIFT and align-GVGD, evidenced by the higher relative frequencies of cases in the highest risk categories (Table 3). Further analyses of the individual rare missense mutations using hierarchical modeling revealed no specific mutations as significantly associated with risk (data not shown).
The overall frequency of intronic variants was higher in cases than controls (Table 3), and we did observe a notably increased frequency of rare intronic variants (MAF < 0.1%) with 12 rare intronic variants seen in cases versus one variant in controls (RR = 13.7, 95% CI: 2.0–591.3) listed in Supp. Table S2. This finding remained significant even when rare and common intronic variants were combined (P = 0.01). Further in silico analysis of these intronic variants using Human Splicing Finder v2.4.1 [Desmet et al., 2009] did not predict that they would have significant effects on splicing. Furthermore, when we examined the incidence of breast cancer in first-degree relatives of carriers of rare intronic variants the rate was not elevated compared to noncarriers. Eleven of the 12 individuals were of non-Hispanic White ethnicity making an ethnic-specific effect unlikely. When the median age of onset of the second breast cancer was compared between the rare intronic group (51.8 years, n = 11), carriers of deleterious mutations (56.6 years, n = 5) and the rest of the cases (age 51.1, n = 543) no significant difference was observed (P = 0.3, Kruskal–Wallis test). There was no significant effect of having received radiation therapy among subsets of mutation carriers (number of events was too small to analyze deleterious variants independently; data not shown).
Discussion
A number of studies have implicated PALB2 gene mutations as a rare, but important, contributing factor to HBC [Tischkowitz and Xia, 2010]. The results presented here support these findings in a population-based setting, and confirm that women with germline PALB2 mutations also have an increased risk of CBC. Although only five clearly pathogenic mutations were identified in the group of 1,124 women, all of these occurred among the cases. The penetrance of PALB2 gene mutations was originally estimated to be associated with a 2.3-fold increased risk [Rahman et al., 2007] but subsequent studies suggested this could be higher, at least for specific mutations [Erkko et al., 2008; Southey et al., 2010]. The relative risk estimate of 5.3 from our study is indeed higher, although this estimate needs to be interpreted with caution given the small number of carriers and carrier relatives in our study and the correspondingly wide confidence intervals.
Despite a large number of published studies, to date no definitely pathogenic PALB2 missense variants have been identified, suggesting that they are nonexistent or rare. Our results are consistent with this literature. As an analogy, in the breast cancer predisposing genes BRCA1 and BRCA2 the vast majority of missense variants are neutral, but a few deleterious variants have been identified. The identification of isolated, individual deleterious mutations using case–control studies requires very large numbers of subjects, or evidence that particular types of missense variants are deleterious collectively, allowing aggregation of the data to create sufficient statistical power. Our hierarchical modeling (data not shown) was designed for this purpose but did not uncover any trends of this nature. Thus, our study does not rule out the possibility that isolated missense PALB2 mutations may be deleterious, but it offers no strong evidence that this is a likely possibility.
The clinical significance of our observation that rare intronic variants are more frequent in the CBC cohort is unclear as none of these variants were predicted to be pathogenic by affecting splicing and only one has been reported previously. Further investigations to address this would involve studying these variants in additional breast cancer cohorts and, at the cellular level, looking for possible functional effects such as diminished BRCA2-binding capacity and reduced homologous recombination efficiency [Tischkowitz et al., 2007]. It should also be noted that we did not see an increased incidence of breast cancer in first-degree relatives of carriers of rare intronic variants that would argue against these being pathogenic.
Our study was conducted using population-based series of breast cancers and as such it avoids the potential problem of ascertainment bias that can occur in cohorts derived from familial cancer clinics or other high-risk settings. Nevertheless, our study has some limitations. First, while a large number of women were screened for PALB2 mutations, the relative rarity of these mutations meant that only a small number of pathogenic mutations were identified. Second, because the study excluded women with synchronous cancers and women who underwent prophylactic contralateral mastectomy following initial breast cancer, and included only women who survived their initial breast cancer, the results may underestimate the overall mutation prevalence. Lastly, we used HRM analysis that, as with conventional sequencing, would miss detection of large exon deletions. However, based on our previous experience on the rarity of PALB2 exon deletions [Tischkowitz et al., 2009] it is unlikely that we would have missed a significant number of mutations.
In summary, we have confirmed that germline deleterious truncating PALB2 mutations are associated with breast cancer in a population-based case–control study. Although PALB2 mutations are a rare cause of breast cancer, their overrepresentation in the cohort of women with CBC is relevant to the clinical management of newly diagnosed women with breast cancer who are found to be PALB2 mutation carriers as it implies a significant risk of developing a second breast cancer. Our study also provides suggestive evidence that rare intronic variants may be associated with risk, but this finding requires careful validation.
Acknowledgements
The WECARE Study Collaborative Group—Memorial Sloan Kettering Cancer Center (New York): Jonine L. Bernstein, PhD (WECARE Study P.I.), Colin Begg, PhD, Marinela Capanu, PhD, Xiaolin Liang, MD, Anne S. Reiner, MPH, Irene Orlow, PhD; City of Hope (Duarte, California), Leslie Bernstein, PhD (Subcontract P.I.), Laura Donnelly-Allen (some work performed at University of Southern California, Los Angeles, California); Danish Cancer Society (Copenhagen, Denmark): Jørgen H. Olsen, MD, DMSc (Subcontract P.I.), Lene Mellemkjær, PhD; Fred Hutchinson Cancer Research Center (Seattle, Washington): Kathleen E. Malone, PhD (Subcontract P.I.), International Epidemiology Institute (Rockville, Maryland), and Vanderbilt University (Nashville, Tennessee): John D. Boice, Jr. ScD (Subcontract P.I.); Lund University (Lund, Sweden): Åke Borg, PhD (Subcontract P.I.), Theresa Sandberg, PhD; National Cancer Institute (Bethesda, Maryland): Daniela Seminara, PhD, MPH; New York University (New York): Roy E. Shore, PhD, Dr. PH (Subcontract P.I.); Northern California Cancer Center (Fremont, California): Esther John, PhD (Subcontract P.I.); Norwegian Radium Hospital (Oslo, Norway): Anne-Lise Børresen-Dale, PhD (Subcontract P.I.); Samuel Lunenfeld Research Institute, MSH (Toronto, Canada): Julia Knight, PhD (Subcontract P.I.), Anna Chiarelli, PhD (Co-Investigator); Translational Genomics Research Institute (T-Gen)(Phoenix, Arizona): David Duggan, PhD (Subcontract P.I.); University of California at Irvine (Irvine, California): Hoda Anton-Culver, PhD (Subcontract P.I.); University of Iowa (Iowa City, Iowa): Charles F. Lynch, MD, PhD (Subcontract P.I.); University of Southern California (Los Angeles, California): Robert W. Haile, Dr. PH (Subcontract P.I.), Daniel Stram, PhD (Co-Investigator), Duncan C. Thomas, PhD (Co-Investigator), Anh T. Diep (Co-Investigator), Shanyan Xue, MD, Nianmin Zhou, MD, Evgenia Ter-Karapetova; University of Texas, MD Anderson Cancer Center (Houston, Texas): Marilyn Stovall, PhD (Subcontract P.I.), Susan Smith, MPH (Co-Investigator); University of Virginia (Charlottesville, Virginia): Patrick Concannon, PhD (Subcontract P.I.), Sharon Teraoka, PhD (Co-Investigator).
