Assessment of homologous recombination deficiency phenotype in breast cancers in adolescents and young adults in the clinical setting

Background Homologous recombination deficiency (HRD), which may be associated with high efficacy of PARP inhibitor- and platinum agent-based therapies, is a prevalent phenotype of breast cancer diagnosed in adolescents and young adults (AYAs; 15–39 years old). HRD score, indicating HRD status, is not routinely assessed in the oncology clinic due to the need for genome-wide analyses. Methods Subjects were a Japanese cohort of 46 AYA breast cancer patients, whose HRD scores were calculated from whole-exome sequencing data, and two existing breast cancer cohorts (US and European) for which HRD scores were available. Genetic and clinicopathological factors associated with the HRD-high phenotype, defined as HRD score ≥42, were selected based on the criterion that they be assessible by routine examinations qualifying for insurance reimbursement. A model for prediction of the HRD-high phenotype was constructed and validated using data from the three cohorts. Results In the Japanese AYA cohort, as in the US and European cohorts, HRD-high phenotype (13/46, 28.3%) was preferentially observed in cases with any or combination of germline BRCA1/2 mutations, somatic TP53 mutations, triple-negative subtype, and higher tumor grades. Because these four factors can be assessed by routine examination that qualifies for insurance reimbursement, we developed a model based on these factors to judge whether a case is HRD-high, using the US cohort (n = 744; Area under the curve AUC = 0.85). The predictive power of the model was validated in the Japanese (n = 46; AUC = 0.90) and European (n = 58; AUC = 0.96) AYA cases. A model developed using the European cohort (n = 477; AUC = 0.89) had similar predictive power in Japanese (AUC = 0.89) and US (n = 54; AUC = 0.87) AYA cohorts. Conclusions The HRD-high phenotype of AYA breast cancer can be deduced based on genomic and pathological factors that are routinely examined in the

beta values for probes of tumor tissues were greater than 0.3. A sample was defined as hypermethylated when it had more than four outlier probes for a specific gene promoter [36]. Based on a previous study [34], hypermethylation of the BRCA1 and RAD51C genes in a US cohort was evaluated based on methylation status at the cg04658354 and cg14837411 loci, respectively. For the European cohort, only the hypermethylation status of BRCA1 was available [39].

Mutational signature analysis
Mutational signatures of the breast cancer genome from the Japanese cohort samples was obtained by decomposing somatic mutations into four major mutational signatures (catalogue of somatic mutations in cancer (COSMIC) signature 1, 2, 3, 6), defined previously [37], to minimize the Kullback-Leibler divergence. A heat map based on signature profiles was generated using the R command regHeatmap. The Pearson's correlation coefficient between the percentage of BRCA signatures (COSMIC signature 3) and the HRD score was calculated.

Model for predicting HRD status using factors assessed in clinical setting
A prediction model for judging the HRD-high phenotype (i.e., HRD score ≥42) was constructed based on logistic regression of four or five variables: i) presence or absence of pathogenic germline or somatic BRCA1/2 mutations, ii) presence or absence of nonfunctional somatic TP53 mutations, iii) TNBC subtype or not, iv) high tumor grade (grade III) or not, and v) hypermethylation (BRCA1 and RAD51C) or not (Table 1). To construct and validate the model using these four or five factors, only cases with information for all four or five variables were included in the analysis. First, a HRD prediction model was constructed using data from all cases in the US cohort, irrespective of age, as a development cohort (n = 744). Then, the constructed model was applied to the AYA cases of the Japanese (n = 37 or 46) and European (n = 58) cohorts, respectively (i.e., validation cohorts). When all cases of the European cohort (n = 477) were used as the development cohort, AYA cases of the Japanese (n = 37 or 46) and US (n = 54) cohorts were used as validation cohorts. The area under the receiver operating characteristic curve (AUC) was calculated to evaluate the predictive power of each cohort. Cutoff values of the prediction model were defined using Youden's index where the sum of sensitivity and specificity was maximal. Based on the cutoff value, positive and negative predictive values were calculated in the validation cohorts.

Statistical analyses
Associations among clinicopathological and genetic factors were examined by Fisher's exact test, Chi-squared test, Pearson's correlation, Kruskal-Wallis test, and Mann-Whitney U-test.

Three AYA breast cancer cohorts with HRD scores
The characteristics of three AYA breast cancer cohorts, i.e., Japanese (n = 46), US (n = 70), and European (n = 70) cohorts are shown in Table 1. Pathological stage and histology did not differ significantly among the three cohorts (P > 0.05), whereas the European cohort contained more TNBC and high tumor grade cases than other two cohorts (P < 0.01) ( Table 1). The US and European AYA subjects were members of pan-age cohorts including 1,044 and 560 subjects, respectively.
HRD scores were calculated for 46 subjects of the Japanese cohort using whole-exome sequencing data ( Figure 1A). The fraction of cases with the HRD-high phenotype, defined as HRD score ≥42 (13/46, 28.3%), was similar to that of the US cohort but considerably lower than that of the European cohort ( Table 1). The calculated HRD scores of the Japanese cohort subjects were strongly correlated with the fraction of cases with the COSMIC mutational signature 3 (Pearson r = 0.78), which is predominant in BRCA1/2mutated and/or HRD-high tumors [47] ( Figure 1B). Hierarchical clustering analysis revealed that most HRD-high subjects were co-clustered in a group of cases in which the COSMIC mutational signature 3 was predominant ( Figure 1C). This result confirmed that the HRD scores of the Japanese cohort samples were tightly linked to BRCA1/2 and other homologous recombination repair deficiencies. hypermethylation were also preferentially (4/5, 80%) observed in HRD-high cases. In addition, somatic mutations in the TP53 gene were more abundant in HRD-high cases than in HRD-low cases (6/13 [46.2%] vs 3/33 [9.1%]; P < 0.01 by Fisher's exact test; Table S2 in Additional file 1). In terms of clinicopathological factors, TNBC included many HRD-high cases (5/6, 83.3%) but other categories, such as luminal subtype (6/35, 17.1%), also included HRD-high cases ( Figure 2A). In addition, high tumor grades (i.e., grade III) were prevalent in HRD-high cases (11/13, 84.6%) ( Table S2 in Additional file 1). The same tendency was also observed in all cases in the US and European cohorts and in AYA cases in these cohorts ( Figure S1 in Additional file 2, Table S2 in Additional file 1, and Figure 2B-

Genetic and clinicopathological factors associated with the HRD-high phenotype
HRD-high judgement was more prevalent in AYA cases than in non-AYA cases ( Figure S2 in Additional file 3), indicating that HRD is a common feature of AYA breast cancers. On the other hand, among the AYA cases, younger age was not associated with the HRD-high phenotype in the Japanese, US, and European AYA subjects (Table S2 in Additional file 1).  Table S3A in Additional file 1). Thus, high predictive power for the HRD-high phenotype was achieved using only the four factors that are routinely assessed in oncological practice. When the cutoff value of the prediction model from the US cohort was tentatively set at 1.0 based on optimal Youden's index (sensitivity: 79.7%; specificity: 80.4%), positive predictive values in the Japanese AYA and European AYA cohorts were 78.6% and 82.5%, respectively, whereas negative predictive values were 93.8% and 94.4%, respectively.

Discussion
Due to the high rate of germline BRCA1/2 mutations, as well as the high fraction of TNBC [6, 48-50], HRD has been considered to be a major phenotype of AYA breast cancer.
Indeed, this study confirmed that HRD-high cases are a major fraction not only of European and US cases (25-50%) [34, 36], but also of Japanese cases (28%). Therefore, PARP inhibitor-and platinum agent-based therapies for these patients could improve the current poor prognosis of AYA breast cancer patients. Notably, as shown in Figure 1A, six (46.2%) of the 13 HRD-high cases in the Japanese cohorts were negative for both BRCA1/2 germline mutation and TNBC; this fraction is much higher than among the other two cohorts (3/16 [18.8%] and 6/40 [15.0%], respectively; see Figure S1 in Additional file 2).
In routine oncological practice, these cases are considered not actively subjected to HRD examination. By contrast, our prediction model is simple and employs only four factors that can be assessed by gene panel tests and pathological examinations that qualify for insurance reimbursement from health care systems. In the near future, gene panel tests will be used routinely in the breast cancer clinic because new therapeutic regimens require testing for mutations in oncogenes such as PIK3CA and AKT1 [32, 33]. Thus, our prediction model will give physicians another option for usage of gene panel test data, i.e., select patients who may carry the HRD-high phenotype for whom it is worthwhile to perform HRD examination and/or to consider subsequent PARP inhibitor-and platinum agent-based therapies.
The HRD-high phenotype is associated with the therapeutic effects of PARP inhibitors and platinum agents in ovarian cancer [20]. In breast cancer, however, only germline mutations of the BRCA1/2 genes, rather than the HRD-high phenotype, have been proposed as strong predictive factors for such therapeutic effects [10]. Theoretically, HRD makes cancer cells susceptible to these agents irrespective of tumor type; therefore, the predictive power of the HRD-high phenotype for the efficacy of PARP inhibitory therapy is now being investigated in a clinical trial (NCT02401347) [19]. The prediction model proposed in our study would help to validate the significance of the HRD-high phenotype using real-world data accumulated routinely in the oncology clinic. Notably, distinguishing between germline and somatic BRCA1/2 mutations did not significantly affect the prediction power (Table S3B in Additional file  However, neither of these analyses yielded positive results, indicating that other unknown mechanisms caused HRD in these cases.
In this study, we established a prediction model for the HRD-high phenotype of AYA breast cancer patients based on gene panel tests and clinicopathological findings.
However, this study has a limitation, in that the numbers of AYA patients in all three cohorts is small, potentially leading to an overestimation of predictive power. Further studies that include a larger number of cases, as well as prospective studies, are needed to firmly conclude that a HRD-high phenotype of AYA breast cancer can be deduced based on genomic and pathological factors that are routinely examined in the oncology clinic.

Conclusions
The present prediction model provides a tool for identifying the AYA breast cancer patients who would benefit from PARP-and/or platinum-based therapies in the clinical setting.