Predictive Factors for BRCA1 and BRCA2 Genetic Testing in an Asian Clinic-Based Population

Purpose The National Comprehensive Cancer Network (NCCN) has proposed guidelines for the genetic testing of the BRCA1 and BRCA2 genes, based on studies in western populations. This current study assessed potential predictive factors for BRCA mutation probability, in an Asian population. Methods A total of 359 breast cancer patients, who presented with either a family history (FH) of breast and/or ovarian cancer or early onset breast cancer, were accrued at the National Cancer Center Singapore (NCCS). The relationships between clinico-pathological features and mutational status were calculated using the Chi-squared test and binary logistic regression analysis. Results Of 359 patients, 45 (12.5%) had deleterious or damaging missense mutations in BRCA1 and/or BRCA2. BRCA1 mutations were more likely to be found in ER-negative than ER-positive breast cancer patients (P=0.01). Moreover, ER-negative patients with BRCA mutations were diagnosed at an earlier age (40 vs. 48 years, P=0.008). Similarly, triple-negative breast cancer (TNBC) patients were more likely to have BRCA1 mutations (P=0.001) and that these patients were diagnosed at a relatively younger age than non-TNBC patients (38 vs. 46 years, P=0.028). Our analysis has confirmed that ER-negative status, TNBC status and a FH of hereditary breast and ovarian cancer (HBOC) are strong factors predicting the likelihood of having BRCA mutations. Conclusions Our study provides evidence that TNBC or ER-negative patients may benefit from BRCA genetic testing, particularly younger patients (<40 years) or those with a strong FH of HBOC, in Asian patients.


Results
Of 359 patients, 45 (12.5%) had deleterious or damaging missense mutations in BRCA1 and/or BRCA2. BRCA1 mutations were more likely to be found in ER-negative than ERpositive breast cancer patients (P=0.01). Moreover, ER-negative patients with BRCA mutations were diagnosed at an earlier age (40 vs. 48 years, P=0.008). Similarly, triple-negative breast cancer (TNBC) patients were more likely to have BRCA1 mutations (P=0.001) and

Introduction
The National Comprehensive Cancer Network (NCCN) has recommended various guidelines for the genetic testing of BRCA1 and BRCA2, which include specific criteria on the age at diagnosis of the patients and family members; the occurrence of breast, ovarian, pancreatic or prostate cancer in close relatives; and the diagnosis of triple-negative breast cancer (TNBC) [1]. Notably, TNBC patients have higher incidence rates of BRCA1 and BRCA2 mutations of up to 30% and 17% respectively [2][3][4], with younger TNBC patients (aged below 40 years) having an even higher incidence of 36% compared to those diagnosed below 50 years of 27% [5]. Most of these studies were based on Caucasian populations. It is unclear if these guidelines may also be adopted in Asian populations.
Next-generation sequencing (NGS) techniques enable the mutation screening of a larger set of samples in parallel, in a cost effective and accurate manner [6,7]. Recently, the emergence of NGS techniques has played an important role in the simultaneous screening of multiple cancer susceptibility genes including the BRCA1 and BRCA2 genes [8,9]. NGS technology has also been widely used in identifying novel genes with mutations related to HBOC [10,11].
Here, we studied 359 breast cancer patients to determine the prevalence of BRCA mutations in an Asian clinic-based population, using next-generation sequencing and Sanger sequencing. In addition, we evaluated the predictive value of ER-, PR-and HER2-receptor status, age at diagnosis, FH, and histological type for determining the likelihood of mutations in the BRCA1 and BRCA2 genes.

Methods Patients
Peripheral blood samples were obtained from 359 breast cancer patients attending a risk assessment clinic at the National Cancer Centre Singapore (NCCS). Subjects were eligible if they had a FH of breast and/or ovarian cancer in first-and/or second-degree relatives (n = 176), or if they had early-onset breast cancer in the absence of FH ( 40 years of age) (n = 183). Patients were accrued from 2002 till 2013. Samples from two earlier studies (accrual from 1992 to 1996 and 2002 to 2006) were also included in this current study [12,13]. Of the 359 breast cancer patients, 321 (89.4%) were Chinese, 16 (4.5%) were Malays, 6 (1.7%) were Indians and 16 (4.5%) were of other Asian ethnicities. ER, PR and HER2 statuses were obtained from clinical databases, and were scored as positive or negative according to previously published criteria [14][15][16]; ER and PR were considered positive when nuclear staining was present in !1% of tumour cells. Her2 was considered as positive when >10% of tumour cells had strong (3+) cell membrane staining. The information for ER and TNBC status were available for 281 and 206 patients respectively. Written informed consent was obtained from all patients and the study was approved by the SingHealth Centralised Institutional Review Board.
Mutational screening of BRCA1 and BRCA2 S1 Fig shows a flow chart of the strategy used to detect mutations in the BRCA1 and BRCA2 genes, to predict damaging mutations and to identify driver/passenger mutations. Frameshift and nonsense mutations were considered to be deleterious.
Sanger sequencing of the BRCA1 and BRCA2 genes was performed as described previously [13], using the CEQ 8000 System (Beckman Coulter, Inc, CA, USA) or the ABI 3130 Genetic Analyzer (AB-Life Technologies; Thermo Fisher Scientific Corporation, MA, USA). The sequenced data were analyzed using the SeqMan Pro v. 8

Bioinformatic Analysis
For samples sequenced by NGS, reads were aligned to the UCSC human reference genome (hg 19) using the BWA aligner (version 0.5.6). Variant calling was done using the GATK Unified Genotyper [17], and CRISP pipelines [18] (for HiSeq).
All mutations identified from Sanger sequencing or NGS were annotated using the ANNO-VAR tool, which provides tools such as SIFT, PolyPhen-II HDIV, PolyPhen-II HVAR, LRT and Mutation Taster to predict the effect of amino acid substitution for each missense mutation. Every missense mutation was scored as damaging or benign with each of the five prediction tools. If the missense mutation was scored as damaging by three or more of the prediction tools, the mutation was classified as a 'Damaging' mutation and if less than three, the mutation was classified as 'Benign'. S1 Table shows the scores for the predictions from the various tools. All missense mutations were also checked against the BIC (http://research.nhgri.nih.gov/bic/), HGMD (http://www.hgmd.org/) and ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) databases, and were regarded as 'pathogenic' if classified as such in two or more databases. All deleterious or pathogenic mutations detected were confirmed by re-sequencing the samples by conventional Sanger sequencing, as described above.

Multiplex Ligation-dependent Probe Amplification (MLPA)
All DNA samples were screened for large genomic rearrangements by MLPA using the SALSA MLPA P002-C2 BRCA1 and SALSA MLPA P045-BRCA2 test kits, and validated using the MLPA P087 and P077 confirmation kits (MRC-Holland, Amsterdam, Netherland), respectively. The MLPA analyses were done by DNA fragment analysis on the ABI 3130 Genetic Analyzer and comparative analysis of samples using the Coffalyser freeware v.131123.1303 (MRC-Holland, AM, Netherland).

Statistical Analysis
Statistical analysis was done using SPSS version 18.0.2 (SPSS IBM, Armonk, NY). The nonparametric test, i.e., Mann-Whitney U-test was used to compare the median age of the carriers and non-carriers. The Fisher's exact test was used to determine significant associations between clinico-pathologic features and the BRCA mutation status. Binary logistic regression analysis was used to estimate the predictive effects of the significantly associated factors for predicting the probability of BRCA mutations. P-values of <0.05 were considered statistically significant.

Mutations in the BRCA1 and BRCA2 genes
Deleterious mutations detected in BRCA1 and BRCA2 are listed in S1 and S2 Tables. Frameshift and nonsense mutations, splice-site errors and large genomic rearrangements were classified as deleterious (n = 33). S3 Table shows the list of damaging missense mutations identified. Eleven of 68 missense mutations were predicted to be damaging.

Clinico-pathological characteristics and mutational status
Among 359 patients, 43 (12%) had a FH of HBOC, 132 (37%) had a FH of breast cancer, 1 (0.3%) had a FH of ovarian cancer and 183 (50.9%) were early-onset breast cancer patients without a FH (Table 1). BRCA mutation carriers were more likely to have a FH of HBOC than non-carriers (39.4% vs 9.2%). Conversely, BRCA carriers were less likely to have early-onset breast cancer in the absence of FH as compared to non-carriers (21.2% vs 54%).
The most common histological type of breast cancer in our study was infiltrating ductal carcinoma (IDC), at 72.2%, followed by infiltrating lobular carcinoma (ILC) (3.3%) and medullary cancer types (3.3%) ( Table 1). Only 1 patients with ILC had BRCA mutations and none of the medullary cases had BRCA mutations. In patients with IDC, the percentage of BRCA mutation carriers was higher at 57.6%, as compared to other histological types of breast cancer.

Associations between clinical characteristics with ER or TNBC status
The median age at diagnosis for ER-positive and ER-negative patients was 40 years and 39 years respectively (Table 3). In addition, the age at diagnosis for ER-negative patients with BRCA mutations was significantly younger than for ER-positive patients (40 vs 48, P = 0.008). When stratified by BRCA1 and BRCA2 mutational status independently, age at diagnosis for ER-negative patients with BRCA1 and BRCA2 mutations was significantly younger than for ER-positive patients (39.5 vs 50, P = 0.053 and 40 vs 48, P = 0.031, respectively) ( Table 3).
The median age at diagnosis for TNBC patients was younger than for non-TNBC patients although not statistically significant (38 vs 40) ( Table 3). The median age at diagnosis for TNBC patients with BRCA mutations was significantly younger than for non-TNBC patients with BRCA mutations (38 vs 47, P = 0.03). When stratified by BRCA1 or BRCA2 mutational status independently, age at diagnosis for TNBC patients with BRCA1 and BRCA2 mutations was significantly younger than for non-TNBC patients (38 vs 46 and 38.5 vs 48, respectively, P = 0.028). Furthermore, the percentage of TNBC patients with a FH of HBOC was higher than for non-TNBC (28.6% vs 11.8%, P = 0.035). However, there was no statistical difference between the TNBC and non-TNBC patients for patients with other FH. Table 4 shows the potential predictive factors for BRCA1 and BRCA2 mutation carriers determined by binary logistic regression analysis. The analyses showed that of all the clinico-pathological characteristics (ER status, age at diagnosis, pedigree diagnosis and histological data), ER-negative status and a FH of HBOC were the strongest predictors for BRCA1 or BRCA2 mutations. The likelihood of patients with HBOC having BRCA1/2 mutations was higher than for other patients (Odds ratio [OR] 3.898; 95% confidence interval [CI] 1.518-10.011; P = 0.005). The odds of having a BRCA1 or BRCA2 mutation in ER-positive patients was 0.390 times (95% CI 0.172-0.887; P = 0.025) less than ER-negative patients (OR 2.562; 95% CI 1.127-5.826; P = 0.025). Neither of the beta coefficients of both factors exceeded the absolute constant value (1.890), indicating that a single factor was insufficient to predict the mutation status. Both ER negative status and a FH of HBOC are required to predict the likelihood of having a BRCA1/2 mutation. Similar analyses were performed to investigate the potential contribution of PR status, as a predictive factor, to predict the likelihood of having BRCA1/2 mutations. However, no statistical significance was found (data not shown).

Predictive factors for BRCA1 and BRCA2 mutations in ER and TNBC patients
A similar analysis was done to evaluate the potential predictive factors for BRCA1 and BRCA2 mutations carriers in TNBC patients ( Table 4). The analyses showed that of all the clinico-pathological characteristics (including age of diagnosis, family history and histology), TNBC status and a FH of HBOC were the strongest predictors for BRCA1 or BRCA2 mutations. The likelihood of patients with HBOC being diagnosed with BRCA1/2 mutations was 3.164 times higher than for other patients (OR 3.164; 95% CI 1.080-9.268; P = 0.036). The odds of TNBC patients being diagnosed with BRCA1/2 mutations was 4.651 higher compared to non-TNBC patients (OR 4.651; 95% CI 1.643-13.163; P = 0.004). Similarly, neither of the beta coefficients of both factors exceeded the absolute constant value (2.835), indicating that a single factor was insufficient to predict the mutation status. Thus, both TNBC status and a FH of HBOC are required to predict the probability of having BRCA1/2 mutations. Predictive factors for BRCA1 and BRCA2 mutations, with the inclusion of damaging missense mutations Binary logistic regression analysis was performed as before but with deleterious BRCA mutations as well as damaging missense mutations (S4-S7 Tables). However, unlike the previous analyses, the median ages did not show any significant difference. Moreover, the analysis showed that only the FH of HBOC is necessary in patients with known ER status, to predict the likelihood of BRCA mutations (S7 Table). For patients with known TNBC status, having a TNBC status and a FH of HBOC are required as predictive factors for BRCA mutation testing (S7 Table).

Discussion
There are few Asian studies that have evaluated the association of BRCA mutation status and clinical characteristics. This current Singapore study, based on 359 Asian breast cancer patients prospectively accrued from a risk-assessment clinic, has identified ER-negativity, TNBC status and a FH of HBOC as predictive factors to increase the likelihood of detecting BRCA1 and BRCA2 mutations.
Approximately 70-80% of BRCA1-associated breast cancer cases are ER-negative [25][26][27][28][29]. We found that BRCA1 carriers are more likely to be ER-negative as has been reported previously in western populations [30]. ER-negative status has been suggested to be intrinsic to BRCA1-related cancer as it has been found that the proportion of ER-negative patients with BRCA1 mutations was significantly higher than for ER-positive patients [31].
Patients with BRCA mutations were diagnosed at an earlier age in this study. BRCA1-associated breast cancers have been shown to be more likely ER-negative for each age group (<45, 45-54, and 55-64 years), with an increase in ER-positive breast cancers with increasing age [31]. Our data concurs with these findings. Furthermore, we provide evidence that ER-negative patients with either BRCA1 or BRCA2 mutations were significantly younger than ER-positive patients.
Our logistic regression analyses indicated that the odds ratio of TNBC patients with HBOC having BRCA1/2 mutations was 3.164 (95% CI 1.080-9.268; P = 0.036), highlighting the importance of FH when estimating BRCA mutations prevalence. This is consistent with another study from the US that reported that TNBC patients with a FH of breast cancer or ovarian cancer had a higher probability of having BRCA mutations as compared to those without any FH of breast cancer or ovarian cancer (57% vs 29%; p<0.001 and 77% vs 41%; p<0.001, respectively) [34].
The NCCN guidelines have proposed the inclusion of TNBC patients aged 60 years or younger for BRCA mutation testing. Recently, a Korean study demonstrated that TNBC patients are more likely to be diagnosed at a younger age than non-TNBC patients in the cohort (42 vs 44.1) although the association was not statistically significant [35]. Nevertheless, in the mutation carriers, the mean age at diagnosis of TNBC patients was older than for the non-TNBC patients (39.2 vs 34.6 for BRCA1 and 51.5 vs 44.0 for BRCA2). Our data, however, are in contrast to these findings. We showed that the median age at diagnosis for TNBC patients with either BRCA1 or BRCA2 is younger than for non-TNBC patients (38 vs 46 for BRCA1 and 38.5 vs 48 for BRCA2), suggesting that BRCA1and BRCA2-associated breast cancer is most likely early-onset. A study from Malaysia showed that TNBC patients aged below 35 years had a higher prevalence of BRCA1 and BRCA2 mutations compared to non-TNBC patients (28% vs 9.9%) [32]. However, additional studies in larger populations from Asia are warranted to verify these findings from Malaysia and Singapore.
Collectively, and confirming previous findings in western populations, our results showed that the likelihood of TNBC patients being diagnosed with BRCA mutations was higher compared to non-TNBC patients, and the inclusion of additional criteria like a FH of HBOC may increase the probability of identifying BRCA1/2 mutations. A study from Malaysia showed an improvement in the sensitivity and specificity of the Manchester scoring system with the combination of negative ER status, FH and TNBC status [32].
In conclusion, our data showed that almost half of our BRCA mutation carriers in our cohort, are ER-negative. We also found that 29% (8/28) of TNBC patients are BRCA mutation carriers, with the majority being BRCA1 mutation carriers. In addition, we have shown that our TNBC patients with either BRCA1 or BRCA2 mutations were diagnosed at an earlier age. The discovery of the predictive factors, ER-negative status, TNBC status and HBOC, in our study, warrants confirmation in additional Asian populations.