BRCA1 Mutations and Breast Cancer in the General Population:  Analyses in Women Before Age 35 Years and in Women Before Age 45 Years With First-Degree Family History FREE

ABOUT 5% to 10% of breast cancer patients carry germline mutations that predispose them to inherited disease.^1- 2 Through genetic linkage studies, BRCA1, a breast and ovarian cancer susceptibility gene,^3- 4 was identified on chromosome 17q. Cloning and sequencing of BRCA1 revealed 24 exons and 5592 nucleotides.⁵ Over 130 germline mutations in the coding region are reported in the Breast Cancer Information Core database to date.⁶ The majority of these result in truncation of the message, leading to absence of functional protein.⁶

Chromosomal mapping of BRCA1 relied exclusively on the analysis of rare, high-risk families, each of which had multiple members affected with breast and/or ovarian cancer.^3- 4 Subsequent studies of the number and type of germline BRCA1 mutations or mutation-related breast and ovarian cancer risk also almost exclusively involved high-risk families.^7- 11 Two analyses, each involving over 100 families, suggest that 80% of high-risk breast and/or ovarian cancer families and nearly half of high-risk breast cancer families have inherited mutations in BRCA1,^12- 13 although a recent review of high-risk families ascertained worldwide suggests that proportions of such families may be more modest than prior estimates suggest.¹¹ The one exception has been the mutation screening of women of Ashkenazi Jewish descent, where analyses suggest 0.9% of Ashkenazi Jewish women have BRCA1 germline mutations.^14- 16

The question of how well data derived from relatively rare high-risk families helps to predict BRCA1 mutation status in women in the general population poses an obstacle to the development of comprehensive screening strategies. Many of the families used to generate the frequency and penetrance estimates have as many as 10 affected members. We wished to assess whether the type and frequency of mutations in the general population differ significantly from those seen in women from high-risk families and the degree to which family history profiles of the type seen in the general population are predictive of BRCA1 status.

In a prior analysis of 80 breast cancer cases diagnosed in women younger than 35 years from a population-based breast cancer case-control study and unselected for family history status, 6 (7.5%) had germline BRCA1 mutations.¹⁷ Two of the 6 women reported little or no breast cancer family history and the other 4 mutation carriers had relatively modest family history profiles. Similar results were found elsewhere in 30 women diagnosed as having breast cancer before age 30 years.¹⁸ In this study, we examined the frequency and spectrum of BRCA1 germline mutations in 2 categories of women drawn from 2 population-based studies hypothesized to be at increased risk of carrying BRCA1 mutations: those with a first-degree (mother or sister) breast cancer family history and breast cancer cases diagnosed in women younger than 35 years.

Cases (women diagnosed as having breast cancer before age 45 years) and controls (similarly aged women without breast cancer) were drawn from 2 population-based case-control studies of breast cancer conducted in western Washington State. The methods of the studies were similar and have been reported.^19- 20 Cases were identified via the Cancer Surveillance System of western Washington, a population-based registry that participates in the National Cancer Institute Surveillance, Epidemiology, and End Results Program. Controls were identified through random-digit dialing.²¹ The first study identified all incident first primary breast cancer cases diagnosed between January 1, 1983, and April 30, 1990, in white women born after 1944 living in the 3-county metropolitan Seattle area at the time of diagnosis. Interviews were completed on 845 cases (83.6% of eligible cases) and 961 controls(75.5% of those eligible). Supplemental funding allowed for subsequent blood collection in 563 cases and a subset of controls (n=113).

The second study identified all incident first primary breast cancer cases diagnosed before age 45 years, from May 1, 1990, to December 31, l992, in women of all races living in the 3-county area. Interviews were completed on 648 (87.0%) of eligible cases and 610 (78.7%) of eligible controls. As part of the original study and of a subsequent investigation, blood was collected from 523 cases and 450 controls.

In both studies, data regarding potential breast cancer risk factors were obtained via structured in-person interviews. Each subject was asked to enumerate all first-degree (mothers, sisters, daughters) and second-degree (aunts and grandmothers) female blood relatives. For each relative enumerated, the interviewer asked for birth year, vital status, death year, history and type of cancer, and laterality (if breast cancer). Follow-up with all previously interviewed women (or proxies) for updated pedigree information was initiated as part of a large, ongoing genetic-epidemiologic study. Information on vital status was obtained from the Cancer Surveillance System and from our ongoing follow-up tracing activities. The institutional review board at the Fred Hutchinson Cancer Research Center approved the study, and all subjects signed an informed consent document before participation.

Three subgroups from the 2 case-control studies were targeted for initial molecular analyses: cases with first-degree breast cancer family history (ie, affected mother and/or sister), cases diagnosed before age 35 years, and controls with first-degree breast cancer family history. We are also screening controls from the same 2 population-based case control studies who were not selected on the basis of family history; this ongoing work is part of larger population-based mutation analyses. Thus, we briefly report results of testing 1 subset of the controls: all women aged 40 to 44 years from our second case-control study for whom blood was available.

Genomic DNA was purified from either frozen buffy coats or immortalized lymphoblastoid cell lines. The polymerase chain reaction (PCR) and subsequent single-strand conformation polymorphism analyses were carried out as described with only minor modifications, including the addition of several new primer pairs constructed to allow for redundant screening.¹⁷ Results were visualized by autoradiography. Variant bands were cut from the gel, rehydrated, and all alterations seen in less than 10% of subjects were sequenced as described.¹⁷

Haplotypes were constructed for the BRCA1 region using microsatellites D17S855 (AFM248yg9), D17S1322 (s754), D17S1323 (s975), and D17S1327 (ED2) by inspection of segregation patterns.²²

Because of potentially limiting numbers, we used the Fisher exact test to assess whether age, stage, vital status, and family history varied between those tested and not tested for BRCA1 mutations. In tested women, the Fisher exact test was used to compare whether proportions of those with a BRCA1 mutation varied by age group, stage, vital status, and family history. We calculated the binomial exact 95% confidence interval (CI) for proportion of those with a BRCA1 mutation. P values for the Fisher exact tests and the 95% CIs and 97.5% CIs (where there were zero mutations) were attained using Stata statistical software (Stata Corp, College Station, Tex).

BRCA1 mutation status was determined for women in 3 categories hypothesized to have increased probability of carrying a mutation: 208 cases with first-degree breast cancer family history, 193 cases diagnosed before age 35 years (80 of whom were previously reported¹⁷ ), and 71 controls with first-degree family history. Also, 199 controls aged 40 to 44 years (not selected on basis of family history) were tested. Groups were not mutually exclusive. Thirty-eight cases studied had both positive first-degree family history and breast cancer diagnosis before age 35 years; similarly, 18 controls with first-degree family history were also members of the age 40 to 44 years control group. Of the total 363 cases studied, 350 (96.4%) were white, 4 (1.1%) African American, 6 (1.7%) Asian, and 1 (0.3%) American Indian; of 251 controls studied, 227 (90.4%) were white, 10 (4%) African American, 8 (3.2%) Asian, and 7 (2.8%) American Indian.

Table 1, Table 2, and Table 3 present characteristics of all eligible women from the population-based studies in each of the 3 potentially high-risk categories targeted. In both case groups studied, proportions available for molecular analysis did not vary substantially by age of diagnosis or family history (Table 1 and Table 2). Compared with all eligible cases, the subgroup of cases tested was more likely to be alive at last follow-up and to have been diagnosed as having earlier-stage breast cancer. This was largely attributable to a lag between interviewing and blood collection (median, 42 months in women tested) in our first case-control study that resulted from initial lack of funding for blood collection. The controls with first-degree family history who were available for analysis were slightly older than the entire group of controls with first-degree family history (Table 3). In controls aged 40 to 44 years, higher proportions were available for analysis in those with a positive family history (Table 4).

In the 208 breast cancer cases aged 21 through 44 years with a first-degree family history, 15 (7.2%, 95% CI, 4.1%-11.6%) had germline BRCA1 mutations (Table 5). Mutation frequency decreased by age of diagnosis, from 23.1% in cases diagnosed before age 30 years to 3.4% in the cases diagnosed at age 40 to 44 years. Although mutation frequency did not differ substantially by vital status, there was variation by disease stage. Specifically, the highest proportion of mutations was seen in women with regional or distant disease (11.3%).

There were 12 (6.2%, 95% CI, 3.2%-10.6%) germline BRCA1 mutations detected in 193 cases diagnosed before age 35 years (Table 6). Similar to the findings in cases with first-degree family history, mutation frequency was higher in this group in those diagnosed at a younger age (11.4% in cases diagnosed before age 30 years vs 4.7% in those diagnosed at ages 30 to 34 years) and in those diagnosed as having more advanced disease (12.7% in cases with regional or distant disease, 3.9% in those with local disease, and none in 27 in situ cases). All 12 mutations in cases diagnosed before age 35 years were in the 179 cases still alive at last follow-up; none were detected in 14 cases now deceased.

In both groups of cases, mutation frequency varied to some extent according to features of family history. In 208 cases studied because of at least 1 first-degree relative with breast cancer, mutations were seen in 10.0% of cases with an affected sister and 6.8% of cases with an affected mother (Table 5). There were 2 mutations (11.1%) in 18 cases with an affected sister and mother. Mutation frequency in cases with a mother and/or sister diagnosed as having breast cancer before age 45 years was 13.2% vs 4.4% in those with an affected mother and/or sister diagnosed at or after age 45 years. A similar pattern was seen regarding early age of diagnosis in aunts and/or grandmothers.

In cases diagnosed before age 35 years and not selected on the basis of family history, those with a mother and/or sister with breast cancer had a greater proportion of mutations (13.2%) than did cases with only an aunt and/or grandmother affected (7.2%) (Table 6). Of 104 women diagnosed before age 35 years with no breast cancer family history, 3 (2.9%) had mutations, one of whom, although she did not have breast cancer family history, did have a maternal aunt with ovarian cancer. Similar to findings in cases selected for first-degree family history, in cases diagnosed before age 35 years, those with a mother and/or sister diagnosed as having breast cancer before age 45 years had a mutation frequency of 16.7% vs those with an affected mother and/or sister diagnosed at age 45 years and older (10.5%). Mutations were also more common in cases with an aunt and/or grandmother diagnosed before age 45 years (15.4%) vs cases with an affected aunt and/or grandmother diagnosed at age 45 years or older (5.3%).

Family history of ovarian cancer was also indicative of BRCA1 mutation status among cases, although there were few subjects with such history. A high proportion of those with first-degree breast cancer family history and ovarian cancer history in a mother and/or sister or in an aunt and/or grandmother had a BRCA1 mutation (25.0% and 23.1%, respectively). In cases diagnosed as having breast cancer before age 35 years, 2 (25.0%) of 8 women with an aunt and/or grandmother with ovarian cancer had BRCA1 mutations.

Mutation frequency varied by number of relatives affected in each case group (Table 5 and Table 6) and in the aggregate of all cases tested. In all 363 cases tested, there were 3 mutations (2.9%) in 104 women with no breast cancer–affected relatives, 15 mutations (6.4%) in 235 cases with 1 to 3 such relatives, and 4 mutations (20.0%) in 20 cases with 4 or more affected relatives. Results were similar when we considered both breast and ovarian cancer in the families.

Mutation frequency did not vary noticeably by presence or absence of bilateral breast cancer family history (data not shown). In all cases with at least 1 relative with breast cancer (n=255), mutations were seen in 3 (4.8%) of 63 cases with at least 1 relative with bilateral breast cancer, 8 (6.3%) of 126 cases with no relatives affected with bilateral disease, and 8 (12.1%) of 66 cases unsure of relatives' laterality status.

No disease-associated mutations were observed in the 71 controls with first-degree breast cancer family history or in the 199 controls aged 40 to 44 years.

Molecular mutations observed in this study are listed in Table 7 and mutation location and frequency are shown in Figure 1. The most common mutation seen was the 185delAG mutation, which was seen in 6 cases and no controls. DNA from the 6 cases with this mutation was tested with markers spanning the BRCA1 region to determine haplotype. All 6 shared the same haplotype in this region that has been observed repeatedly in women of Ashkenazi Jewish descent. For markers D17S855, D17S1322, and D17S1323, all women carried alleles of 146, 128, and 157 base pairs (bp), respectively. For marker 17S1327, varying allele sizes were observed, although a 133-bp allele was seen in 4 of the 6 cases, as is commonly reported in Ashkenazi women.²³

Of the 25560 PCR reactions required for this study, 25558 (99.99%) were completed and scored. For 2 women, however, there was 1 reaction that never yielded a product, likely reflecting polymorphism within the PCR primers for both alleles.

Twenty of the 22 disease-associated mutations observed have been seen previously in studies of high-risk families, and are listed in the Breast Cancer Information Core database.⁶ Two mutations were novel: a deletion of 4 bp at nucleotide 2911 causing a stop codon at amino acid 999 and a 28-bp deletion at nucleotide 5489 causing a stop codon at amino acid 1825.

Missense mutations, which change only a single amino acid, were seen in cases and controls. Rare polymorphisms occurring in less than 5% of cases and controls were seen that did not change the protein in substantial ways and are not likely of molecular significance. The distribution of genetic alterations (disease-associated mutations, missense mutations of unknown significance, and rare polymorphisms) is given in Table 8.

Much of what is known about the frequency and distribution of germline BRCA1 mutations derives from studies of rare families with large numbers of affected women and many living family members available for sampling.^6- 8,13,24 While detailed analysis of these mapping families has proven invaluable in studies aimed at determining the initial mutation spectrum, additional studies are now needed to clarify the role of germline BRCA1 mutations outside the selected venue of high-risk families.

Hospital-based case series partially meet that need; for instance, Krainer et al²⁵ recently reported 9 (12.3%) BRCA1 germline mutations in 73 women diagnosed as having breast cancer before age 32 years. There was no information regarding family history, but this was part of a larger cohort originally collected because of early-onset breast cancer. Similarly, 12.8% of 798 women who presented at high-risk clinics because of their presumed elevated risk were found to be BRCA1 positive.²⁶ Mutation status was related to several factors, including early age of onset, number of relatives with breast or ovarian cancer, Ashkenazi ancestry, and bilateral disease. Similar results were reported by Couch et al²⁷ in 94 women diagnosed as having breast cancer before age 40 years identified from general oncology practices, and in 169 women with breast cancer referred to a breast cancer clinic because of family history. These studies provide useful data but have limited generalizability to those with more modest family histories for 2 reasons: the average number of affected women was substantial, with some families having as many as 11 affected women, and as many as 25% of families had both breast and ovarian cancer. Studies of broader scope are therefore needed to provide data for the large number of women with a modest to minimal family history of disease.

The analysis described here has several strengths. First, all women were drawn from the general population as part of 2 population-based case-control studies, the proportion of eligible cases interviewed was high, and no specific ethnic groups were targeted. Second, 2 factors believed relevant to BRCA1 mutation status, family history, and early age of onset were examined separately and together. Also, because detailed breast and/or ovarian cancer family history was available, we could assess contributions of additional family history characteristics that might affect probability of being a carrier. For example, mutation frequency was not noticeably different for those with a bilateral breast cancer family history vs those without such history. However, our data suggest that having relatives diagnosed as having breast cancer early in life, larger numbers of relatives affected with breast cancer, and ovarian cancer family history may be important factors in assessing risk. Also, as in the study by Couch et al,²⁷ mutation frequency was greater in those with both breast and ovarian cancer family history vs those with only breast cancer family history.²⁷ In all cases,BRCA1 mutations were found in 2 (2.1%) of 97 cases with no family history of breast or ovarian cancer, 14 (6.1%) of 230 cases with breast but not ovarian cancer family history, 1 (14.3%) of 7 cases with ovarian but not breast cancer family history, and 5 (20.0%) of 25 cases with both breast and ovarian cancer family history.

Also, our analysis provides information about the number of BRCA1 mutations in the general population in groups of women whose risk of being a carrier is uncertain but could be expected to be elevated. In the absence of population-based molecular studies, guidance regarding mutation frequency in autosomal dominant genes in those with fewer than 4 affected relatives or those diagnosed in specific age ranges has been largely drawn from statistical analyses of interview data from population-based studies. Initial estimates based on a large population-based case-control breast cancer study suggested that mutations in autosomal dominant genes were responsible for 36% of breast cancer in women younger than 30 years, decreasing to 1% in cases age 80 years or older.² By comparison, Whittemore et al²⁸ using data pooled from 3 population-based case-control studies of ovarian cancer estimated that in young women diagnosed before age 40 years, 11% of breast cancer and 18% of ovarian cancer are due to autosomal dominant germline mutations. Finally, data from 2 population-based studies of cancer mortality in relatives of breast and ovarian cancer patients were used to estimate the proportion of breast cancer cases in the general population due to autosomal dominant breast cancer susceptibility genes and the results showed it to be 5.3% in those younger than age 40 years, 2.2% between ages 40 and 49 years, and 1.1% between ages 50 and 70 years.²⁹

Collectively, these 3 studies offered upper-boundary estimates of BRCA1 mutation frequency in women in the general population. Variation among these studies is likely due, at least in part, to distinctions in the data sets used (eg, extent to which women with both breast and ovarian cancer were included). Also, contributions of BRCA1 could not be distinguished from those of other autosomal dominant breast cancer genes (eg, BRCA2).

The frequency of BRCA1 mutation in this study in women younger than 35 years (6.2%) approximates those predicted,^28- 29 with 11.4% of women diagnosed as having breast cancer younger than 30 years and 4.7% of women diagnosed from 31 to 34 years carrying germline mutations. Mutation frequency increased to 23.1% among cases with a first-degree breast cancer history diagnosed before age 30 years and was 8.0% in cases with such history diagnosed from 30 to 34 years, suggesting that a first-degree breast cancer family history along with young age at diagnosis may increase the risk of being a carrier.

Since the family history profiles seen in our series of cases are considerably less pronounced than those in studies of high-risk families,^27,30- 31 the BRCA1 mutation frequency is, as expected, much lower. It has been suggested that cancer in most families with fewer than 4 cases of breast and no ovarian cancer is not due to highly penetrant genes like BRCA1.²⁷ Excluding the study participants, the mean number of relatives affected with breast cancer in this series was 1.7, and the mean number with breast or ovarian cancer was 1.85. Although the number of families in our study with 4 or more affected women is small, as in Ford et al,²⁹ we did find that mutation frequency varied by combination of 4 or more relatives with breast cancer and ovarian cancer family history. There were mutations in 14 (4.5%, 95% CI, 2.5%-7.4%) of 312 women with fewer than 4 affected relatives and no ovarian cancer family history, 4 (14.8%, 95% CI, 4.2%-33.7%) of 27 women with fewer than 4 affected relatives and ovarian family history, 2 (10.5%, 95% CI, 1.3%-33.1%) of 19 women with more than 4 affected relatives and no ovarian cancer family history, and 2 (40.0%, 95% CI, 5.3%-85.3%) of 5 women with 4 affected relatives and this history (for difference in groups, P=.005).

Of 22 women with disease-associated BRCA1 mutations, 6 (27.8%) had the 185delAG mutation and associated haplotype, frequently seen in those of Ashkenazi descent.^6,14- 16 Data regarding ethnic background were not collected in the original interview; thus, no data are available regarding the ancestry of the 6 women with this mutation or its frequency in western Washington. Of originally interviewed controls (n=1571) and cases (n=1489), 1.5% and 2.4%, respectively, indicated their religion as Jewish.

Those with modest breast cancer family history present the biggest challenge for development of BRCA1 testing strategies in women from the general population. This will be further complicated by lack of knowledge about the BRCA1 protein and uncertainty regarding most rare missense mutations (sometimes referred to as "rare variants" or "missense changes of unknown consequence"). Only the missense mutation characterized by a Cys→Gly change at amino acid 61, which destroys the protein RING finger, and which was seen in 1 woman in this study, is known to be disease associated as it has been seen to segregate with affected women in many high-risk families.^7- 8 Functional assays will be useful for addressing significance of missense mutations such as the Ile→Met at nt 1256 or Met→Thr at nt 5002 described herein (thus far seen only in unaffected persons); such changes may represent rare events of no significance or disease-associated mutations of varying penetrance. Characteristics that suggest that a missense mutation may predispose to disease include absence in a control group of sufficient size, cosegregation with disease in some families, or occurrence in a highly conserved area of the protein or in a putative functional domain. Previously, because it was seen only in cases, we considered the 12-bp insertion in intron 20 likely to be disease associated.¹⁷ However, as it is seen in controls herein and as proposed,³² it is likely to be a polymorphism.

Studies of the relationship between first-degree breast cancer family history and survival and between BRCA1 mutation status and histopathologic prognostic factors are suggestive, but not conclusive.^33- 36 Our finding of higher proportions of mutations in women with more advanced disease and no mutations in women with in situ disease is provocative but needs replication.

Generalizability of our findings may be influenced by the extent to which the women whose samples were available for testing are not representative of the otherwise eligible women in the categories studied, and, if there were an association between BRCA1 mutations and increased (or decreased) survival from breast cancer, our study could overestimate (underestimate) overall mutation frequency due to inclusion of women more (less) likely to survive. It is reassuring that in cases younger than 35 years, there is no evidence of bias in blood collection by family history (Table 2). Although we cannot directly assess potential bias for blood collection by family history in women selected for testing on the basis of first-degree family history, the overall proportion of women aged 35 to 44 years with a first-degree family history (19.4%) interviewed in original case-control studies is similar to that (20.5%) having blood drawn.

The overall small number of mutations seen in 363 cases makes interpretation of their relationship with features of family history preliminary; some differences in frequency observed according to specific characteristics could be due to chance, as demonstrated by the wide CIs accompanying observed mutation frequencies. No mutations were observed in the 252 controls studied. Estimates of general population BRCA1 mutation frequency range from 1 in 500 to 1 in 1500 women.^2,28- 29 Thus, we would not necessarily have expected to detect even a single mutation in the control subgroup studied. Larger population-based studies will be needed to elucidate the role of specific family history features in predicting mutation status and for assessment of factors influencing risk of disease.

In addition, sensitivity of our mutational screening strategy could have influenced findings. When performed correctly, single-strand conformation polymorphism and techniques like it can detect over 80% of mutations.³⁷ To further optimize our ability to detect mutations, most primers were designed to produce PCR fragments that were no larger than 280 bp, gels were run as long as 24 hours to maximize detection of subtle variants, and additional primer pairs were constructed to allow for redundant screening of parts of the gene, particularly exon 11. However, a few mutations, likely missense changes, that tend to produce more subtle gel pattern perturbations will be missed even in an optimal setting. One additional limitation that is not unique to this or other commonly used screening methods is that mutations affecting expression, splicing, or stability of the transcript are difficult to detect. Yet, they likely account for as much as 15% to 20% of BRCA1 mutations.¹¹

Several issues raised in this report have implications for women and their clinicians who are considering genetic counseling. First, many women in the general population with a positive family history of breast cancer who may believe they are at high risk for carrying a mutation in an autosomal dominant breast cancer susceptibility gene are not BRCA1 carriers. Second, some women with family histories may have mutations in genes other than BRCA1. Third, some women may be BRCA1 mutation carriers in the absence of strong family history, perhaps reflecting paternal origin, few assessable family members, or variability in mutation penetrance; a mechanism for identifying these women is needed. Finally, the role of missense mutations or unclassified variants in disease causation is unclear and testing strategies targeting only known or common mutations will miss novel or rare mutations.

In light of such complexities, it is critical that genetic testing be undertaken only after careful counseling by skilled health care providers. While the information presented here suggests that overall the risk of carrying a BRCA1 mutation for women with modest breast cancer family history is much lower than that for women with extreme family histories, sufficient data do not yet exist to devise the optimal stratifications for deciding who, specifically, should be offered carrier testing. In the face of such uncertainties, it may be recalled that breast cancer is a disease of high frequency in Western women. While the research community continues to probe questions of genetic susceptibility, there remains no substitute for access to high-quality, affordable health care that provides for vigilant screening and early diagnosis for all women.