Triple-negative breast cancer and PTEN (phosphatase and tensin homologue)loss are predictors of BRCA1 germline mutations in women with early-onset and familial breast cancer, but not in women with isolated late-onset breast cancer

The recruitment of breast cancer patients into the Malaysian Breast Cancer Genetic Study (MyBrCa) started in January 2003 at the University Malaya Medical Centre in Kuala Lumpur. All were histopathology-proven breast carcinomas. Histomorphologic and biomarker parameters were retrieved from the histopathology reports. Blood, demographic, and family-history data were collected from breast cancer patients who consented to participate in this study. This study was approved by the ethics committee of University Malaya Medical Centre.

From January 2003 to February 2012, 1,454 breast cancer patients were recruited into the MyBrCa study. Germline DNA samples were screened for BRCA1 and BRCA2 mutations for all women with (a) early-onset breast cancer (≤35 years of age) (35 with and 96 without family history of breast and ovarian cancer); (b) family history of breast or ovarian cancer in first- and second-degree relatives (193 women); or (c) isolated triple-negative breast cancer diagnosed at between 36 and 50 years old in the absence of family history (47 women). In addition, of the 432 women who were diagnosed aged 36 to 50 years with non-TNBC, 60 women with the highest risk were analyzed (bilateral breast cancer, breast and ovarian cancer in the index patient, family history of breast and ovarian cancer in third-degree or isolated breast cancer (≤45 years of age)). Mutation detection for germline BRCA1 and BRCA2 mutations was conducted by using direct DNA sequencing and multiple ligation-dependent probe amplification (MLPA), as previously described [12, 13].

For this cohort, patients were classified as having TNBC when we found <10% ER^-, <10% PR^-, and 0 or ≤2⁺ HER2 staining with immunohistochemistry. HER2 was not routinely tested for in all patients prior to 2006, and therefore, in 259 patients, HER2 status was unavailable. Of the 110 women for whom germline BRCA1 and BRCA2 mutations were analyzed, adequate archived paraffinized invasive tumor tissue for evaluation was available from 32 index patients and two relatives. All cases were tested with immunohistochemistry for cytokeratin 5/6 (D5/16B4; Dako Ltd.), cytokeratin 14 (LL02; Novocastra Labs), and PTEN (6H2.1; Dako). Cytokeratin 5/6 or cytokeratin 14 was defined as positive if >10% of invasive tumor cells showed cytoplasmic staining [14], and PTEN was defined as negative if PTEN staining was undetectable in tumor cells in contrast to adjacent normal stromal cells [6‐8]. Description of other pathologic features was conducted as previously described [14].

Unconditional multiple linear logistic regression was used to model the probability that the proband was a mutation carrier as a function of her personal and family history and age at diagnosis, as described in [15].

Manchester score was calculated as previously described [16]. In brief, this system assigns scores depending on the type of cancer and age at diagnosis and developed such that a score of 15 was equivalent to a 10% chance of identifying a BRCA1 or BRCA2 mutation. The modified method [17‐19] includes upward adjustments for TNBC (+4), estrogen-negative (+1), and high-grade invasive cancers (+2), and downward adjustments for human epidermal growth factor receptor 2 (HER2)-positive (-4) or estrogen-receptor positive (-1), lobular (-2), and low-grade noninvasive cancers (-2). For individuals in whom tumor tissue was available for PTEN staining, upward (+1) and downward (-1) adjustments were made for the absence or presence of PTEN staining, respectively.

Of the 1,454 breast cancer patients in the MyBrCa Study, 177 (12.2%) had TNBC based on pathology reports. Of these 177 women, 50 older than 50 years developed breast cancer and did not have any family history of breast or ovarian cancer in first- or second-degree relatives and were therefore excluded from the study. Of the remaining 127 women, 63 had already been analyzed because of family history of either breast and/or ovarian cancer (38 individuals, 13 BRCA1 and three BRCA2 carriers, mutation prevalence, 42.1%) or early-onset breast cancer (25 individuals, seven BRCA1 and no BRCA2 carriers; mutation prevalence, 28.0%]. Of the remaining 64 women diagnosed at ages 36 through 50 years, but with no family history of breast or ovarian cancer, 47 women were screened for germline mutations in BRCA1 and BRCA2 genes, and three BRCA1 and one BRCA2 carriers were identified (mutation prevalence, 8.5%). Overall, of the 110 women who developed TNBC and were analyzed (63 with and 47 without family history of breast and ovarian cancers), 23 BRCA1 and 4 BRCA2 carriers were identified, giving a mutation prevalence of 24.5% (Table 1).

In total, 321 women with non-TNBC were analyzed. This included all women with a family history of breast and/or ovarian cancer (190 individuals, 10 BRCA1 and 17 BRCA2 carriers, mutation prevalence 14.2%) and women diagnosed at 35 years or younger with no family history of breast and ovarian cancer (71 individuals, three BRCA1 and four BRCA2 carriers; mutation prevalence, 9.9% (Table 2)). In addition, of the 432 women who were diagnosed aged 36 through 50 years with non-TNBC, 60 women with the highest risk were analyzed (bilateral breast cancer, breast and ovarian cancer in the index patient, family history of breast and ovarian cancer in third degree or isolated breast cancer (aged 45 years or younger)). Of these 60 women, one BRCA1 and three BRCA2 carriers were found [mutation prevalence, 6.7%]. Overall, of the 321 non-TNBC women analyzed, 14 BRCA1 and 24 BRCA2 carriers were identified, giving a mutation prevalence of 11.8%.

We compared the prevalence of BRCA1 and BRCA2 mutations in the women who developed TNBC and non-TNBC. Of the women with low familial risk of breast and ovarian cancer (diagnosed 36 to 50 with no family history of breast or ovarian cancer], 6.4% and 1.7% of women with TNBC and non-TNBC were found to be BRCA1 carriers, respectively, whereas 2.1% and 5.0% were BRCA2 carriers. Overall, 8.5% of women with TNBC and 6.7% of women with non-TNBC were found to have germline BRCA1 or BRCA2 mutations. This suggests that, regardless of the TNBC status, in the absence of family history of breast or ovarian cancer in this age group, a low (<10%) probability exists of having a BRCA1 or BRCA2 mutation.

By contrast, in two other groups of women, a significantly higher prevalence of BRCA1 mutations was found in women who developed TNBC versus non-TNBC. First, of the women in whom breast cancer developed at 35 years or younger, 28.0% of women with TNBC were BRCA1 or BRCA2 carriers compared with only 9.9% among those who developed non-TNBC (P = 0.045). Notably, all 28.0% of women diagnosed with TNBC at younger than 35 years were BRCA1 carriers, compared with 4.2% BRCA1 and 5.6% BRCA2 among women diagnosed with non-TNBC.

Second, of women who had a family history of breast and/or ovarian cancer, 42.1% of women with TNBC were BRCA1 or BRCA2 carriers compared with 14.2% of those with non-TNBC (P = < 0.0001). This is largely because of a difference in prevalence of BRCA1 mutations in TNBC compared with non-TNBC (34.2% compared with 5.3%; P = < 0.0001), as no significant difference occurred in the prevalence of BRCA2 in both subsets of women (7.9% compared with 8.9%; P = 1.00). Notably, no significant difference was noted in average age at onset of breast cancer in the index patient, or the mean number of first-degree relatives of women in whom TNBC developed compared with the non-TNBC (seven relatives), but the mean number of affected relatives in the non-TNBC group was higher than that in the TNBC group (0.8 compared with 0.6; P = 0.003). This suggests that the higher prevalence of BRCA1 mutations in the women in whom TNBC developed compared with women in whom non-TNBC developed is not due to a difference in the age at onset or strength of family history. Taken together, the results suggest that early onset and familial breast cancer patients in whom TNBC develops are more likely to have mutations in the BRCA1 genes compared with those in whom non-TNBC develops (24.5% versus 11.8%; P = 0.001).

In addition, in this cohort, a marked difference appears in prevalence of BRCA1 and BRCA2 in the TNBC and non-TNBC patients. TNBC patients are more likely to have BRCA1 than BRCA2 mutations (20.9% and 4.4%; p < 0.0001), whereas no statistically significant difference is present in BRCA1 and BRCA2 mutations in non-TNBC patients (3.6% and 7.5%, respectively; P = 0.158).

Given the possible confounding between the age of diagnosis, subtype of breast cancer, and family history of cancer, it has been demonstrated that the predictive effects of factors should be based on an analysis that takes into account all factors simultaneously [15]. By using binary logistic regression analysis, we found that age at diagnosis of breast cancer, having any family history of breast or ovarian cancer, and having triple-negative breast cancer were associated with a 2.6-fold (confidence interval (CI), 1.4 to 4.8; P < 0.003), 3.5-fold (CI, 1.9 to 6.8; P < 0.0001), and 3.5-fold (CI, 1.91 to 6.3; P < 0.0001) increase in risk of being a BRCA1 or BRCA2 carrier, respectively. No evidence of interaction effects between these factors was seen. With multiple linear logistic regression analysis, the strongest predictors of mutation status were breast cancer in the proband if age at diagnosis was younger than 35 years (P = 0.043), with bilateral breast cancer (P = 0.025), with triple-negative breast cancer (P < 0.0001), with breast cancer in a first-degree relative if age at diagnosis was younger than 60 years (P < 0.01), or ovarian cancer was present in a first- or second-degree relative (P < 0.025) (Table 3).

The probability that a proband with a given set of personal and family characteristics is a mutation carrier can be estimated from the model fit shown in Table 4. Starting with the log odds score of θ (the baseline coefficient), add the respective regression coefficient (β) for each personal characteristic and the respective regression for each family characteristic multiplied by the number of affected relatives. In this model, an overall score of -1.5 is equivalent to a 15% probability of being a BRCA1 or BRCA2 carrier. Notably, the overall scores for women with isolated TNBC who were diagnosed aged ≤35 years old, 36 to 39 years old, or 40 to 49 years old, are -1.1, -1.6, and -2.1, respectively, which correlates with a 29%, 14%, and 9% probability of being a BRCA1 or BRCA2 carrier. The overall scores for women with isolated non-TNBC who were diagnosed at younger than 35 years, 36 to 39 years old, and 40 to 49 years old, are -2.4, -2.9, and -3.4, respectively, which correlates with a 9%, 4%, and 4% probability of being a BRCA1 or BRCA2 carrier. These results show that women with isolated TNBC have a higher probability of being carriers compared with women with isolated non-TNBC and that, of the women with isolated breast cancers, only women with isolated TNBC diagnosed at younger than 40 years have a greater than 10% probability of having germline mutations in BRCA1 or BRCA2.

We analyzed the pathology reports of BRCA1 carriers and compared them with those in BRCA2 and non-BRCA carriers. Of the 31 BRCA1 index carriers and affected relatives where pathology reports were available, 26 (83.9%) developed TNBC. By contrast, of the 270 BRCA2 carriers and non-BRCA carriers, 88 (32.6%) developed TNBC.

Of the 110 TNBC patients included in this study and for whom germline status of BRCA1 and BRCA2 has been characterized (see earlier), 34 had adequate invasive tumor tissue for evaluation. Slides were cut and immunostained for several markers that have been reported to be useful in defining the basal-like phenotype, including basal cytokeratins CK5/6, CK14, and PTEN. Although some higher grade were present, higher basal cytokeratin, loss of PTEN, high grade of pleomorphism, presence of pushing margins, solid sheets, necrosis, and mitosis in BRCA1 carriers compared with non-BRCA1 carriers, these differences were not statistically significant (see Additional file 1).

To determine whether the addition of pathologic features can define a subset of women who are likely to have germline mutations in BRCA1 or BRCA2, we calculated the Manchester score for each individual based on the original Manchester score [16] and on the updated Manchester score where pathology was included [18]. Upward adjustments in BRCA1 mutation-prediction scores were made for grade 3 ductal cancers, estrogen receptor (ER), and triple-negative tumors, and downward adjustments in the score were made for grade 1 tumors, lobular cancer, ductal or lobular carcinoma in situ, noninvasive breast cancer, and ER/HER2 positivity, as described previously [18].

Of the 431 women in this study, 86 were excluded because pathology reports were incomplete. Without adjustment, 72 of 345 women had a Manchester score of ≥15, and this included only 12 of the 28 BRCA1 carriers (sensitivity, 42.9%; specificity, 81.1%; PPV, 16.7%). With the adjustment, 82 of the 345 women had a Manchester score of ≥15, and this included 18 of the 28 BRCA1 carriers (sensitivity, 64.3%; specificity, 79.8%; PPV, 22.0%) (Table 5). These results show that adjustment in this cohort resulted in 14% increase in the number of tests (72 to 82) and 21% increase in sensitivity (43% to 64%).

In addition, given that PTEN loss was associated with BRCA1 germline mutations [8], we made a further upward adjustment for PTEN loss (+1 point) for all patients for whom PTEN status was available. Without adjustment, five of 26 women had a >10% probability of BRCA1 and BRCA2 mutations, and this included only three of the seven BRCA1 carriers (sensitivity, 42.9%; specificity, 89.5%; PPV, 60.0%). With the adjustment for PTEN loss, 10 of the 26 women had a Manchester score ≥15, and this included six of the seven BRCA1 carriers (sensitivity, 85.7%; specificity, 78.9%; PPV, 60.0%; Table 5). Although PTEN results were available for only a small subset of patients, these results suggest that upward adjustment for PTEN may aid the identification of BRCA1 carriers.