Study cohort

We previously reported the prevalence of germline variants among Japanese patients with BRCA1/BRCA2-wildtype HBOC syndrome and a strong family history [7]. In that study, we noted that the PGV-carrying status of HBOC-causative genes was complex between index cases (n = 13) and their BRs (n = 34), as well as between affected and unaffected BRs [7]. The present study was undertaken to better understand this complexity. We expanded the cohort to 684 family members (682 BRs and 2 spouses) of 281 index cases, of which germline samples were obtained, regardless of the proband genotype (Fig. 1 and Table S1A). Degree of relatedness and gender were significantly associated with the rate of subject participation (Table S1B).

We conducted exome sequencing for samples from all 281 probands and 522 BRs. The remaining samples from 160 BRs and 2 spouses were sequenced with target panels covering all exons of 30 previously known HBOC-causative genes (Fig. 1 and Table S2; target genes are described in the Supplementary Methods). The clinical features of the BRs and spouses who received genetic analyses are described in Table 1.

Table 1 Characteristics of blood relatives (BRs) and spouses who received a genetic analysisDetection of PGVs in BRs of PGV-negative probands

Variant interpretations were used to ascertain PGVs on probands and BRs/spouses. Overall, 17 PGVs (11 genes) and 37 PGVs (13 genes) were identified for 17 probands and 36 BRs, respectively (Fig. 1 and Table S3). Two spouses were negative for PGV. Representative pedigree charts with PGV information are shown in Figures S1 and S2, with additional information provided in the Supplementary Document. Many families exhibited complexity in concordance with PGV-carrying status between probands and BRs, and between affected and unaffected BRs, as noted previously [7].

The prevalence of PGVs on HBOC-causative genes was similar between probands and BRs (6.0% and 5.3%; Fig. 2A). However, in some cases, PGVs found for BRs were absent in the probands (Fig. 2B, left panel). BRCA1/BRCA2 wild-type patients were selected as the index cases; therefore, all probands were negative for PGVs of BRCA1/BRCA2 genes (Fig. 2B, proband in left panel) [7]. However, BRCA1 and BRCA2 PGV carriers were identified among BRs (Fig. 2B, BRs in left panel). Similarly, we detected mutations in TP53, NF1, and MSH6 genes in the BRs, which were also absent among the probands (Fig. 2B left panel). After stratifying the mutated genes in terms of risk (see also “Gene selection and risk assignment” in Supplementary Methods), we found a significant difference in the presence of PGVs within high-risk genes in BRs as compared with probands (Fisher’s exact test, p = 0.0104; odds ratio [OR] = 0.000 [95% confidence interval, 0.000–0.5489]; Fig. 2B right panel). This significant enrichment was further confirmed against the metadata from a non-cancer East-Asian population (Tables S4 and S5).

Fig. 2

Differences in mutated genes between probands and BRs. A Frequency of PGV carriers among probands and BRs. Numbers of PGVs are shown. B Prevalence of risk class of mutated genes between probands and BRs. Shown are the number of PGVs per gene for probands and BRs (left panel) and the frequency of risk class for probands and BRs (right panel). Frequency is based on the number of cases. Note that all 281 probands had wild-type BRCA1/2 alleles due to the study design. The p value was computed using Fisher’s exact test for PGV carriers of high-risk genes. BR, blood relatives; PGV, pathogenic germline variant; OR, odds ratio; and 95% CI; 95% confidence interval

Discordance of PGVs between probands and BRs

We examined concordance/discordance rates for the presence of PGVs between probands and BRs for various genes (Fig. 3A; Table S3). Concordance was observed for PGVs in PALB2, CHEK2, RAD51C, RAD51D, BLM, BRIP1, FANCM, and MRE11 among 19 BRs from 14 families (Fig. 3A; Table S3). Two types of discordance were noted: (1) PGVs identified in BRs but absent in the probands, and (2) different PGV(s) identified in the BR(s) from those in the proband. In the first instance, we identified mutations in BRCA1, BRCA2, TP53, NF1, RAD51D, MSH6, or FANCM genes in 14 BRs from 11 families but no PGVs in the probands (Fig. 3A). In the second case, 2 BRs from 2 different families (Fig. 3A) had different PGVs from that of their respective probands; this was as described previously [4, 7, 8]. Specifically, in one family, the BR had a discordant PGV, whereas, in the other family, the BR had two PGVs, one concordant and one discordant (Fig. 3A; Table S3; Supplementary Document). Specifically, one PGV on TP53 (c. 743G>A [p.R248Q]) was identified as a de novo mutation, as confirmed by trio analysis (Figure S2 and Table S3; A0867; daughter of A0350 and A0866). In a different family, another TP53 variant (c. 713G>A [p.C238Y]) was found (BR, A0815) and assumed to be de novo, because the BR’s mother (A0235; the proband) and sister (A0812) lacked the variant, and her father had no previous history of cancer (nor was he subjected to testing in this instance; Figure S2 and Table S3). Overall, discordant PGVs were not rare among BRs of BRCA1/BRCA2 wild-type HBOC patients.

Fig. 3

Risk class of mutated genes and concordance of PGV detection in BRs. A Concordance of PGVs between probands and BRs. The number in the square indicates the specific BR with concordant or discordant detection of PGVs. Concordance is defined by the same PGV detection between the proband and BR; discordance is defined based on either the detection of a different PGV between the proband and BR, or detection of a PGV in the BR but not in the proband. B PGVs in the BRs of probands with or without PGVs. Frequency of cases is indicated. The p value was computed by Fisher exact test. C Frequency of risk class of mutated genes detected in the BRs of probands with or without PGV. Frequency is based on the number of cases. The p value was computed using Fisher’s exact tests. D Frequency of risk class of mutated genes in BRs of concordant (conc.) or discordant (disc.) PGV detection with respect to the proband. Frequency is based on the number of PGVs. BR, blood relative; PGV, pathogenic germline variant; OR, odds ratio; and 95%CI, 95% confidence interval

Enrichment of high-risk genes in discordant PGVs

BRs of PGV-positive probands contained a higher number of mutations on genes with insufficient evidence of risk (p = 0.0334; OR = 7.200 [1.250–36.29]). In contrast, significantly fewer PGVs were found among the BRs of PGV-negative probands than among those of PGV-positive probands (p < 0.0001; OR = 36.33 [16.89–78.16]). However, PGVs were still detected in 14 BRs of PGV-negative probands (Fig. 3B), with a significant presence of high-risk genes (p = 0.0008; OR = 0.05556 [0.01079–0.3066]) (Fig. 3C). Indeed, by stratifying the BRs in terms of PGV concordance or discordance, we observed enrichment of high-risk genes among the discordant PGVs (Fig. 3D). These findings indicate a clear association between high-risk genes and discordant PGV detection.

Impact of positive cancer history and route of PGV inheritance

Given that we initially selected index cases with a strong family history of breast or ovarian cancer [7], we considered that there could be enrichment of PGV carriers among BRs with a positive cancer history. However, binary comparisons revealed no significant difference in prevalence (Fig. 4A), risk class (Fig. 4B), or genotype concordance (Fig. 4C) of PGVs between cancer-affected and cancer-unaffected BRs. These observations indicate a negligible impact of positive cancer history on the PGV-carrying status of BRs.

Fig. 4

Prevalence, risk class, and concordance of PGVs in BRs with (affected) or without (unaffected) cancer history. A Prevalence of PGVs in BRs with (affected) or without (unaffected) cancer history. Frequency is based on the number of cases. B Frequency of risk class of mutated genes in BRs with (affected) or without (unaffected) cancer history. Frequency is based on the number of cases. C Concordance of PGV detection in BRs with (affected) or without (unaffected) cancer history. Frequency is based on the number of PGVs. BR, blood relative; PGV, pathogenic germline variant

We also considered the relevance of the pattern of PGV inheritance in a family, and sought to determine whether maternal or paternal transmission could be estimated based on the status of breast or ovarian cancer history among BRs (Figure S3). In terms of cancer history, PGVs were estimated as having been inherited maternally and paternally in 16 and 4 families, respectively. For 8 families, insufficient information prohibited an estimation of the route of inheritance (Figure S3). We then compared these estimates with genetic analyses. The genetic analyses revealed maternal and paternal transmissions in 8 and 4 families, indicating that only 6 maternal and 2 paternal transmissions were correctly estimated (Figure S3). In addition, as mentioned above, trio analysis revealed a de novo PGV on TP53 (Figure S2 and S3). Together, these observations show that it is difficult to estimate the route of PGV inheritance from phenotypic status alone.