Frequency of Breast Cancer Attributable to BRCA1 in a Population-Based Series of American Women FREE

COMMERCIAL availability of genetic tests for BRCA1 and BRCA2 poses a dilemma. Many professional and advocacy groups currently recommend testing for breast cancer predisposition in the context of research protocols.^1- 3 However, when confronted by medical concerns and by the marketplace, clinicians and consumers may find the choice to test difficult to reject. Actual frequencies of inherited BRCA1 mutations in white and black patients from the general population (ie, not selected for age at diagnosis or family history) may help inform testing decisions for the general population of American women.

Most information about inherited genetic susceptibility to breast cancer has come from research on high-risk families, including the basis of proof for BRCA1's existence,⁴ its localization by genetic mapping,^4- 5 and its cloning.⁶ Informativeness of families for genetic analysis is strongly influenced by family size, number of affected relatives, vital status of relatives, and degree of relatedness among relatives with disease. Hence, for earlier research, families with multiple cases of breast cancer in multiple generations were needed, thus selection for families with young ages at diagnosis occurred. Now, testing for cancer predisposition due to inherited BRCA1 mutation is complicated by limited information about frequency of BRCA1 mutations in the general population. In the human BRCA1 gene, more than 100 distinct variants have been reported.⁷ Of those variants that are disease related, nearly all are frameshift or nonsense mutations leading to truncated proteins.

The purpose of this study was to assess the potential clinical and public health significance of inherited BRCA1 mutations in white and black women not selected for breast cancer family history or for age at diagnosis. A series of interrelated questions are addressed: What proportion of breast cancer cases in the general population of white and black women carry known disease-related variants at BRCA1? How do these frequencies differ by age at diagnosis, race, or family history? What are the implications of these data for genetic testing?

The Carolina Breast Cancer Study (CBCS) is a population-based, case-control study of breast cancer in women aged 20 to 74 years from a 24-county area of central and eastern North Carolina.⁸ Women with incident, primary, invasive breast cancer were identified from 26 hospitals using a rapid ascertainment mechanism.⁹ Potential comparison women, frequency matched on age and race, were identified from the North Carolina Division of Motor Vehicles lists for those younger than 65 years and from the US Health Care Financing Administration lists for those aged 65 to 74 years. Nurses made home visits to conduct interviews and to draw blood. To characterize family history, probands were asked to enumerate first-degree relatives and report history of any cancer, irrespective of site, and age at diagnosis. For more distant relatives, only history of breast and ovarian cancer was queried along with relationship to the proband.

Interviews for phase 1 of the study took place from May 1993 through December 1996, with recruitment of 890 cases and 841 controls. To increase statistical power for subgroups, younger women and black women were oversampled using a modification of randomized recruitment.^10- 11 Overall response rates were 77% for cases and 68% for controls, with blood samples obtained for more than 95% of participants. This report focuses on the first 211 patients with breast cancer and 188 control women, regardless of race, and the subsequent 99 cases and 108 controls of African American ancestry. The expense of fully genotyping BRCA1 precluded testing the entire series of cases and controls with current technology. The project was approved by the institutional review boards of the University of North Carolina School of Medicine and the University of Washington. In addition, a certificate of confidentiality was obtained from the US Department of Health and Human Services.

Using DNA extracted from peripheral blood lymphocytes, 211 cases were screened for germline mutations in the BRCA1 coding sequence, splice junctions and neighboring intronic regions, 5′ untranslated region (UTR), and 3′ UTR using multiplex single-strand conformation analysis, as described below. In addition, DNA aliquots from all cases were hybridized to allele-specific oligonucleotide probes for 8 frequent mutations in European populations. As a further check on sensitivity of testing, exon 11 was screened using a protein truncation test. All potential rare variants were evaluated by direct genomic sequencing, using bands from the single-strand conformation analysis gel, as well as original DNA, as template. To compare frequencies of polymorphisms and rare variants of unknown clinical significance in cases and controls, the 188 controls were genotyped by the same methods used for that variant in cases. The number of controls differed for some analyses because of restricted amounts of DNA.

Forty-seven primer pairs were used to amplify BRCA1 from genomic DNA using polymerase chain reaction (PCR). Most primers were defined previously.^12- 13 New primers were designed for a few regions and are defined using the notation of Table 1 in Friedman et al¹²: exon 5-reverse: 5′-ATG GTT TTA TAG GAA CGC TAT G-3′; exon 6-reverse: 5′-GGT CTT ATC ACC ACG TCA TAG-3′; exon 8-reverse: 5′-TTT GGC AAA ACT ATA AGA TAA GG-3′; exon 9-reverse: 5′-TGC ACA TAC ATC CCT GAA CC-3′; exon 10-reverse: 5′-AGG TCC CAA ATG GTC TTC AG-3′; exon 16A-forward: 5′-AAC AGA GAC CAG AAC TTT GTA ATT C-3′; exon 16A-reverse: 5′-TGC ATT ATA CCC AGC AGT ATC AG-3′; exon 16B-forward: 5′-CCA TCT TCA ACC TCT GCA TTG-3′; exon 16B-reverse: 5′-ACT CTT TCC AGA ATG TTG TTA AGT C-3′; exon 20-forward: 5′-GCC TTA AAT ATG ACG TGT CTG CTC-3′; exon 20-reverse: 5′-TGG AAT ACA GAG TGG TGG GGT G-3′; and exon 24-forward: 5′-AGT CGA TTG ATT AGA GCC TAG-3′. Conditions for PCR amplification and single-strand conformation analysis were as described.¹³

The multiplex single-strand conformation analysis scheme allowed simultaneous analysis of several exons or several regions of exon 11 in the same lane of an electrophoretic gel. The PCR products from the 47 primer pairs were run in 14 gel lanes according to band size and pattern. Within each lane, bands corresponding to different amplified products could be easily distinguished (Table 1). If a band shift was observed, the fragment involved was amplified again, and direct DNA sequencing was performed in both forward and reverse strand directions. In addition, the shifted band was cut out of the gel, suspended in distilled water, and sequenced. Multiplex single-strand conformation analysis detected all variants reported in the study.

The allele-specific oligonucleotides were designed for 8 mutations. Mutations 185delAG, 4184delTCAA, 4446C→T, and 5382insC were genotyped using primers and conditions previously described.¹³ For 4 other mutations, primers were designed to distinguish normal and mutant alleles as follows: exon 5: 300 T→G - normal: 5′-CCT TCA CAG TGT CCT TTA-3′; mutant: 5′-CCT TCA CAG GGT CCT TTA-3′; exon 5: 331(+1) G→A - normal: 5′-AAC CAA AAG GTA TAT AAT-3′; mutant: 5′-AAC CAA AAG ATA TAT AAT-3′; exon 6: 332(−11) T→G - normal: 5′-CTC AAA CAA TTT AAT TTC-3′; mutant: 5′-CTC AAA CAA GTT AAT TTC-3′; exon 11: 2457 C→T - normal: 5′-ATG GCA CTC AGG AAA GTA-3′; mutant: 5′-ATG GCA CTT AGG AAA GTA-3′.

Individual dot blots in 96-well formats contained positive and negative controls for each screened mutation. All positive dots were confirmed by directly sequencing both the PCR product from the analysis and an independent aliquot from the original sample. The allele-specific oligonucleotide hybridization independently (and blindly) confirmed the 3 protein-truncating mutations observed.

The BRCA1 exon 11 was amplified and then translated in 3 fragments using primers and methods as described.¹³ No additional variants were detected by protein truncation test.

Reamplified PCR product of each variant exon or region was analyzed by automated and/or manual direct DNA sequencing of both strands using the original single-strand conformation analysis–PCR primers. Automated sequencing was performed with a DNA sequencer, using dye-labeled dideoxy terminator chemistry (ABI Prism 377, Foster City, Calif: Applied Biosystems of Perkin Elmer). Manual sequencing was conducted with the Sequenase PCR Product Sequencing Kit (Cleveland, Ohio: United States Biochemical) following the manufacturer's instructions.

Primer pairs and conditions for genotyping at markers flanking BRCA1 were as described.¹² For those CBCS cases with a disease-related BRCA1 variant, genotypes at these markers were compared with haplotypes from high-risk families with the same variant to assess whether the mutations shared a common origin.

Proportions and 95% confidence intervals (CIs) were computed using SAS¹⁴ and SUDAAN¹⁵ software. Both cases and controls were sampled using known probabilities based on race and age; weighting was incorporated in statistical analyses to produce parameter estimates that reflect frequencies in the underlying population. Odds ratios (ORs) and CIs were estimated by logistic regression models using PROC GENMOD software,¹⁶ which allowed for age and race adjustments and provided unbiased estimates through inclusion of offset terms derived from the sampling probabilities.

Comparisons of cases and controls with respect to race, age, and family history of breast or ovarian cancer are shown in Table 2. Cases and controls were similar in distribution of age and race, for which they were matched. Cases were more likely than controls to have first-degree or more distant relatives with breast or ovarian cancer and to have multiple affected relatives. Four women reported male relatives with breast cancer, 2 cases and 2 controls. In subsequent tables, results are restricted to women who identified themselves as either white or black, since numbers of women reporting other races were small.

Three of the 211 genotyped cases carried protein-truncating mutations in BRCA1 (Table 3, Figure 1). All 3 were white, with breast cancer diagnosis at ages 45, 52, and 53 years. An intron 5 splicing mutation, 332(−11) T→G, was observed in 2 cases (NC1 and NC2). The family of NC1 includes a sister diagnosed as having breast cancer at age 41 years. The family of NC2 includes a niece with ovarian cancer (the age of the niece at diagnosis and which sibling was her parent are not known). This intron 5 mutation destroys an acceptor splice site, leading to aberrant messenger RNA splicing and insertion of 59 nucleotides from intron 5 into the messenger RNA. Translation stops at codon 81 of the mutant BRCA1 sequence.¹² This splice mutation was seen in a prior series of high-risk kindreds (family 82¹²), as well as in 4 other high-risk families of western European ancestry.¹⁷ Genotypes of NC1, NC2, and family 82 at markers flanking BRCA1 indicate that these families share the same ancestral chromosome 17q21 and hence probably a common origin for the mutation (Table 4). Although not closely related, NC1 and NC2 are from the same North Carolina county, and family 82 is from North Carolina as well.

The single nucleotide substitution in BRCA1 exon 11, 2457 C→T, is a nonsense mutation causing an immediate stop in translation at codon 780. This mutation was observed in NC3, whose family includes her mother with breast cancer, a sister with bilateral breast cancer diagnosed at a young age, another sister with breast and ovarian cancer, and a brother with bladder cancer. This mutation was found previously in an American kindred of Dutch ancestry with 7 cases of breast cancer in 3 generations (family 7¹³). The same mutation was subsequently seen in 3 families in the Netherlands, 1 in Germany, and 4 in the United States.^7,18 Genotypes of NC3 and family 7 at markers flanking BRCA1 suggest a founder effect with a recent mutation at marker D17S1322 (Table 4).

Three missense mutations and one 3–base pair deletion were observed in 5 patients (Table 3). The consequences of the variants for disease risk are unknown. Substitutions of histidine for glutamine in exon 11 at codon 1200 in 1 black case (NC6) and of glycine for arginine in exon 20 in 2 white cases (NC7, NC8) have not been reported elsewhere. The exon 20 variant was observed once in 148 controls in the CBCS. These cases (NC6 through NC8) reported no relatives with breast or ovarian cancer. The substitution of methionine for isoleucine at codon 379 was observed in 1 black case (NC5) and 1 black control from this series and was reported once elsewhere.⁷ One case, NC5, reported a paternal aunt with breast cancer and a father and brother with prostate cancer. The 3–base pair deletion in exon 11, leading to deletion of an aspartic acid, was observed in 1 white case (NC4) in this series, who reported a paternal aunt with ovarian cancer. This same deletion also was seen in a breast cancer patient of English ancestry (R. A. Eeles, written communication, December 1997).

In our study area of North Carolina, prevalence of inherited disease-related BRCA1 variants was 3.3% in white cases, 0% in black cases, and 2.6% for the population of breast cancer patients as a whole (Table 5). All estimates were adjusted for sampling probabilities. In white women, inherited BRCA1 predisposition was responsible for 6.6% of cases with any relative with breast or ovarian cancer and 22.8% with any relative with ovarian cancer. As expected, inherited mutations were more frequent in high-risk families. Again in white women, 13.4% of cases from high-risk families with either breast or ovarian cancer and 33% of cases from high-risk families with both breast and ovarian cancer had a BRCA1 mutation. Breast cancer cases diagnosed before age 50 years were no more likely to have inherited BRCA1 disease-related variants than cases diagnosed after age 50 years. Weighted mean ages at diagnosis were 51.5 years for women with disease-related variants, 55.8 years for women with other rare missense or in-frame deletion mutations, and 55.7 years for all other breast cancer cases.

Variants in introns 8, 16, 18, 22 and at coding nucleotides 710, 1186, 2430, 2731, 3232, 3238, 3667, 4158, 4801, and 4956 were observed in the cases. All of these are considered polymorphisms and have assigned PM identification numbers.⁷ Five of the variants (including those in intron 8, exon 11 at nucleotides 2430, 3232, 3667, and intron 16) are inherited as a unit because they are in linkage disequilibrium. Four novel variants (in exons 11, 12, 16 and intron 12) were detected in black women, all of which are silent (no amino acid change) or occur in noncoding regions (Table 6). A fifth variant in exon 3 also was observed in 1 black case, reported previously.⁷ One missense mutation, 3537 A→G in exon 11, was observed at low frequency in black cases and controls.

The single nucleotide substitution C→G at position 36 in the 3′ UTR was more common in black cases than in black controls: 18 of 86 cases and 5 of 74 controls had genotypes CG or GG. The rarer G allele also was observed in 1 white and 1 Native American case and 1 white control. In this series of black women, the age-adjusted OR for breast cancer and the G allele was 3.5 (95% CI, 1.2-10.0). To provide an independent test of association, the subsequent 99 cases and 108 controls in the CBCS who identified themselves as black were genotyped with an age-adjusted OR for breast cancer and the G allele of 1.7 (95% CI, 0.5-6.3). Combined, the overall, age-adjusted OR was 2.8 (95% CI, 1.3-6.2). In black women, the weighted mean age at diagnosis was 54.5 years for women with the rare G allele and 52.5 years for those homozygous for the C allele. The 3′ UTR site is in partial linkage disequilibrium with a variant in intron 22. The association of the intron 22 variant with breast cancer was not statistically significant.

In white women in this North Carolina population, the prevalence of known, disease-related BRCA1 mutations was 3.3%. In black women in the study described herein, there were no BRCA1 mutations of the types known to be related to disease, although such mutations have been identified in black families selected for high risk.^19- 20 Hence, BRCA1 mutations are rare among incident (ie, newly diagnosed) patients with breast cancer not selected for family history or age at diagnosis. These are the vast majority of patients encountered in a primary care setting.

The low prevalence of 1 in 30 cases attributable to BRCA1 in white patients is consistent with statistical projections from other population-based series of the proportion of breast cancer due to all susceptibility genes combined.^21- 22 Similar results were obtained in statistical analyses of families of ovarian cancer patients, although this approach also potentially includes susceptibility genes other than BRCA1.^23- 24 As expected, BRCA1 mutation prevalence in high-risk patient series is considerably higher. The proportion of breast cancer families attributable to inherited BRCA1 mutations was 60% to 75% for those reporting 3 or more cases of breast or ovarian cancer,^25- 27 and 45% for families with 3 or more cases of breast cancer in absence of ovarian cancer.²⁸ In breast cancer patients attending referral clinics (often because of family history and/or young age at diagnosis), 11% to 16% inherited disease-related variants.^29- 32 In contrast, 7.5% of surviving breast cancer patients from western Washington with very early onset (<35 years) were BRCA1 mutation carriers³³ and a population-based study of incident cases of breast cancer before the age of 40 years in Australian women found a prevalence of 3.6% (95% CI, 0.3% to 12.6%) for protein truncation mutations in BRCA1 (J. L. Hopper, written communication, December 1997).

The North Carolina population contributed 1 in-frame deletion and several missense mutations to a class of variants whose biological meaning is unclear. Screening large series of well-matched controls is necessary, but not sufficient, for clarifying the role of these variants in breast cancer because the variants are rare in both cases and controls.³⁴ Development of functional assays may help determine their biological activity and thereby clarify their clinical significance.

In black women in this series, a relatively common variant in the 3′ UTR was almost 3 times more common in cases than controls, suggesting that this putative polymorphism may confer a moderately increased risk of breast cancer. Although located in a noncoding region, critical sequences in the 3′ UTR do influence stability of messenger RNA and therefore phenotype.³⁵ In the future, using cultured cells from women with the 3′ UTR variant and those with wild-type sequence, it will be possible to test experimentally for differences in stability of the wild-type vs variant BRCA1 message.

Who is most likely to carry a BRCA1 variant that influences disease risk? First, ovarian cancer is an important marker of inherited risk: 23% of white patients with breast cancer who had a family history of ovarian cancer had an inherited mutation in BRCA1. Second, families with at least 4 cases of breast or ovarian cancer, regardless of age, frequently reflect inherited susceptibility: 13% of cases from high-risk families with breast or ovarian cancer and 33% of cases from high-risk families with both breast and ovarian cancer had inherited BRCA1 mutations. Because these results are based on few families at the highest levels of risk, any fluctuation in number of families with BRCA1 mutations would make considerable difference in the point estimates. Evidence that prostate cancer risk also is influenced by BRCA1^36- 38 suggests that it may be worthwhile to include this cancer in future definitions of family history to identify high-risk individuals.

None of the 43 women in the CBCS whose breast cancers were diagnosed when they were younger than age 40 years carried disease-related BRCA1 variants. A recent report of BRCA1 testing in women from high-risk families also noted the absence of mutations among probands younger than 30 years.³² Similarly, the majority of younger patients with breast cancer studied in various other series have not tested positive for BRCA1 mutations.^29- 31,33 Hence, young age at diagnosis in itself does not identify a group of breast cancer patients likely to have inherited susceptibility at BRCA1. However, younger average age at diagnosis in family members generally is a better predictor.^31- 32 This occurs because a history of younger ages at diagnosis in family members (including the proband) provides more information than does the age at diagnosis of the proband herself.

Three previous studies of BRCA1 found higher prevalences of disease-related variants in their very early-onset breast cancer patients.^29,31,33 Differences between prevalences of BRCA1 mutations in young patients in those populations compared with the North Carolina population could be due to differences in allele frequencies or to sampling variability (CIs frequently overlap). Alternatively, unrecognized selection for patients who were not only young, but also from high-risk families, may account for higher estimates in the other studies. Possible explanations cannot be distinguished because the previous studies did not include older women to evaluate overall mutation frequency in those series or the relative distribution of variants by age. The well-established link between BRCA1 and early-onset breast cancer frequently overlooks the fact that nearly all families inheriting BRCA1 or BRCA2 mutations include women whose conditions were diagnosed at a wide range of ages.^28,39

Interpretation of risk in the context of predictive genetic testing poses clinical dilemmas. First, negative test results (ie, no detected mutation in BRCA1 or BRCA2) can have any of several meanings. There may be no inherited predisposition to the disease. Alternatively, a proportion of mutations are large deletions or genomically complex alterations that will be missed by approaches based solely on genomic DNA.^12,40- 41 Although in our series of high-risk American families,^12- 13 mutations detectable only by analysis of complementary DNA constituted approximately 5% to 10% of BRCA1 mutations, such mutations are far more common in Holland,⁴¹ and hence potentially in other populations where founder effects have major influence. Also, not all inherited breast cancer can be attributed to BRCA1 or BRCA2,^25- 26,42 and the other susceptibility genes have not yet been identified. Hence, unless a specific mutation has been identified in other family member(s), a negative test result does not provide complete information. Second, positive test results cannot yet be interpreted precisely because risk of cancer associated with mutation (ie, penetrance) is not yet well characterized for BRCA1 and BRCA2. Resolution of the question of risk associated with BRCA1 in the general population awaits analysis of women at risk, not selected from high-risk families, but for whom individual genetic information has been obtained. No such analysis has been completed yet, although several such analyses are in progress.

In women with inherited BRCA1 disease-associated mutations, options for preventing breast or ovarian cancer remain limited. Minimally, more frequent clinical screening by physical breast examination is recommended. The benefits of mammographic screening in the general population of women younger than 50 years remains controversial,^43- 44 but what about its usefulness in a high-risk population? Since the sensitivity of mammography differs among women,⁴⁵ it seems most prudent to decide when to begin mammographic screening on an individualized basis. Prophylactic mastectomy and oophorectomy are clearly far more drastic alternatives, for which some data on effectiveness have recently appeared. In women electing prophylactic mastectomy (primarily due to family history or to prior nonmalignant breast conditions), the reduction in breast cancer was 90% of that expected over the succeeding 17 years.⁴⁶ Alternatives for women with inherited mutations that fall between the extremes of screening and surgery, such as preventive use of tamoxifen, are under investigation.⁴⁷

What should the policy for predictive genetic testing be? The social and legal issues surrounding genetic testing remain more challenging than the technical ones.^48- 53 However, the ongoing debate may become moot as genetic testing becomes increasingly integrated in clinical practice. Potential clinical benefits of genetic testing may be optimized under certain conditions. Testing women from the general population for the entire BRCA1 and BRCA2 sequences is of questionable value, because a large number of women would be tested, expensively, to detect few mutations, and the negative results cannot be definitive, except in the context of a family with a known disease-related mutation. For populations with relatively common founder mutations, selected screening of breast and ovarian cancer patients may be reasonable. For other American women, the guidelines suggested by the results of this study may be realistic: family history of both breast and ovarian cancer or 4 or more cases of breast cancer (at any age) in the family. In either event, clinical decision making is substantially benefited by the involvement of specialists familiar with the subtleties and uncertainties of genetic testing, including how to communicate effectively results that may be inconclusive.

Coverage for medical follow-up for women with BRCA1 or BRCA2 mutations is an integral part of any rational health care system. Currently, these costs are spread across the entire population, usually in association with treatment of end-stage disease. However, when possible, consequences of these predispositions will be best treated with preventive care. Shifting the financial burden to individuals as this becomes a prospect seems shortsighted. In the end, we all are predisposed to some condition(s) as the consequence of our genotypes, albeit most of the predisposing genes have not yet been identified.

In summary, prevalences of BRCA1 mutations among breast cancer patients in this population-based study (3.3% in white and 0% in black cases) may be generalizable to large portions of the United States. In North Carolina, most white residents trace their ancestry primarily to northern and western Europe (US Census Bureau data, unpublished findings, 1990). Less than 2% of North Carolina residents are of Jewish ancestry.⁵⁴ Because there are few mutation carriers in this North Carolina population, fluctuations in this number could make considerable difference in the point estimate. However, the upper limit of the CI for white patients (7.2%) excludes the higher estimates of prevalence observed in other studies from clinical settings (11%-16%).^29- 32 This suggests, for most of the United States, that widespread screening of breast cancer patients (or the general population) for BRCA1 is not warranted.