Genomic instability in non–breast or ovarian malignancies of individuals with germline pathogenic variants in BRCA1/2

Abstract

Background

Individuals with germline pathogenic variants in BRCA1 or BRCA2 are at a high risk of breast and ovarian carcinomas with BRCA1/2 deficiency and homologous recombination deficiency that can be detected by analysis of genome-wide genomic instability features such as large-scale state transitions, telomeric allelic imbalances, and genomic loss of heterozygosity. Malignancies with homologous recombination deficiency are more sensitive to platinum-based therapies and poly(ADP-ribose) polymerase inhibitors. We investigated the fraction of non–breast or ovarian malignancies that have BRCA1/2 deficiency and genomic instability features.

Methods

The full tumor history of a large, historical, clinic-based, consecutive cohort of 2965 individuals with germline pathogenic variants in BRCA1/2 was retrieved from the Dutch nationwide pathology databank (Palga). In total, 169 non–breast or ovarian malignancies were collected and analyzed using targeted next-generation sequencing and shallow whole-genome sequencing to determine somatic second-hit alterations and genomic instabilities indicative of homologous recombination deficiency, respectively.

Results

BRCA1/2 deficiency was detected in 27% (21/79) and 23% (21/90) of 20 different types of non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1 and BRCA2, respectively. These malignancies had a higher genomic instability score than BRCA1- or BRCA2-proficient malignancies (P < .001 and P < .001, respectively).

Conclusions

BRCA1/2 deficiency and genomic instability features were found in 27% and 23% of a broad spectrum of non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1 and BRCA2, respectively. Evaluation of the effectiveness of poly(ADP-ribose) polymerase inhibitors in these individuals should be focused on tumors with a confirmed absence of a wild-type allele.

Individuals with hereditary breast and ovarian cancer have an increased risk of developing breast carcinomas (lifetime risk = 47%-66%) and ovarian carcinomas (lifetime risk = 35%-46%) (1). Hereditary breast and ovarian cancer is most often caused by an autosomal dominant inherited germline pathogenic variant in BRCA1 or BRCA2 (2,3). Individuals with germline pathogenic variants in BRCA1/2 are offered annual breast surveillance for early cancer detection or prophylactic bilateral mastectomy and salpingo-oophorectomy to reduce cancer risk (4,5). Furthermore, individuals with germline pathogenic variants in BRCA1/2 have increased risks to develop other cancers, including pancreatic (lifetime risk = 1%-5%) and prostate carcinomas (lifetime risk = 21%-60%) (6-9).

BRCA1 and BRCA2 are involved in the homologous recombination repair pathway, which repairs DNA double-strand breaks (10). Malignancies in patients with germline pathogenic variants in BRCA1/2 can have complete loss of functional BRCA1/2 through loss of the wild-type allele or by a somatic pathogenic variant in the wild-type allele. BRCA1/2 deficiency may result in homologous recombination deficiency. In the absence of homologous recombination, error-prone nonhomologous end joining is used to repair double-strand breaks (10). As a result, cells accumulate specific types of DNA aberrations that can be recognized by analysis of genome-wide genomic instability markers, such as large-scale state transitions, defined as deletions or duplications of at least 10 megabase pairs (Mb) in size per chromosome arm, genomic loss-of-heterozygosity (LOH) regions of at least 15 Mb in size but smaller than a complete chromosome, and telomeric allelic imbalances smaller than 2 Mb in size not crossing the centromere (11-14).

BRCA1/2 deficiency and the resulting homologous recombination deficiency are shown to be biomarkers for sensitivity to platinum-based therapies and poly(ADP-ribose) polymerase (PARP) inhibitors (15). Recent studies have shown that non–breast or ovarian malignancies with biallelic BRCA1/2 inactivation may benefit from platinum-based therapies and PARP inhibitor treatment (16,17). Little is known, however, about the frequency of BRCA1/2 deficiency and its effect on homologous recombination deficiency in non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1/2. Although these studies showed that most of these cancers develop independent of the BRCA1/2 germline pathogenic variant, only recently has a systematically collected small set of such cancers been published (14,18). We systematically investigated cancer incidence, BRCA1/2 deficiency, and genomic instability scores in non–breast or ovarian malignancies in a large and consecutive cohort of individuals with germline pathogenic variants in BRCA1/2 (N = 2965 individuals).

Methods

Study cohort

This study included 2965 individuals who were counseled at the Radboud University Medical Center (Nijmegen, the Netherlands) and were genetically diagnosed with a germline (likely) pathogenic variant in BRCA1/2 (index, n = 815; relatives, n = 2150; BRCA1, n = 1795; BRCA2, n = 1170). Complete tumor history for this cohort was retrospectively requested up to February 2021 through the Dutch nationwide pathology databank (Palga), which contains excerpts from histopathological reports from all pathology laboratories in the Netherlands since 1991 (19). Detailed information about the study cohort and patient selection can be found in the Supplementary Methods (available online). This study was performed in accordance with the standards of the Helsinki Declaration. Because deidentified pseudonymized data were used and no incidental findings regarding an individual could be made, individual informed consent was not required. The local ethical committee of the Radboudumc (Nijmegen, the Netherlands) approved this retrospective study under study No. CMO-2020-6887.

Pathological review and DNA isolation

Tumor material from formalin-fixed, paraffin-embedded blocks (n = 299) (Figure 1, A) was requested through Palga and the Dutch National Tissue Portal. In total, 252 tissue blocks were received for further analysis. Hematoxylin-eosin–stained slides were reviewed by pathologists (R.S.vdP., I.D.N., M.S.). Tumor subtype and tumor cell percentage were scored for each tissue block. In 49 cases, no or too little tumor material was available, and these cases were excluded from further analysis. For all remaining available tissue blocks (n = 203), genomic DNA was isolated from regions with at least 30% neoplastic cells, as described previously (20) (Figure 1, A; Supplementary Tables 1 and 2, available online). Using single-molecule molecular inversion probe sequencing, second-hit analysis could be performed for 169 malignancies, of which 158 were also analyzed for the number of copies of BRCA1 and BRCA2, and a genomic instability score using shallow whole-genome sequencing. The histology of available cancers from patients with multiple primary carcinomas were investigated for morphological subtype and copy number profiles to exclude metastases.

Figure 1.

Malignancies in individuals with a germline pathogenic variant in BRCA1 or BRCA2. A) Flowchart of the cohort and malignancy selection. B) Overview of individuals with or without malignancies, by gene and identification history. In the inner circle, the fraction of women and men is represented. Outer circles show that 1, 2, 3, 4, 5, 6, 7, or 8 developed malignancies in the fraction of individuals for breast carcinoma, ovarian carcinoma, prostate carcinoma, and other malignancies. A unique set of malignancies needed to occur in at least 2 individuals to be visualized. C) Standard incidence ratios for individuals with germline pathogenic variants in BRCA1 at 20-69 years of age. Cancer incidence in both the index individual and relatives with germline pathogenic variants in BRCA1 were included in the analysis and compared with the cancer incidence of the general Dutch population. No significantly increased incidence, significantly decreased incidence, or significantly increased incidence is shown by cancer type. Numbers are given by type of malignancy on the left of the figure. Malignancies with a potential selection bias because of breast or ovarian cancer surveillance measures are separated from other malignancies by a black line. Cancers that occurred in 5 or fewer individuals were grouped as “others.” D) Standard incidence ratios for individuals with germline pathogenic variants in BRCA2 at 20-69 years of age. Cancer incidence for both the index individual and relatives with germline pathogenic variants in BRCA2 were included in the analysis and compared with the cancer incidence of the general Dutch population. No significantly increased incidence or significantly increased incidence is shown by malignancy. Numbers are given by type of malignancy on the left of the figure. Malignancies with a potential selection bias because of breast or ovarian carcinoma surveillance measures are separated from other malignancies by a black line. Cancers that occurred in 5 or fewer individuals were grouped as “others.”

^aMetastases were removed for the standard incidence ratio analysis. Palga = Dutch nationwide pathology databank; N/A = not applicable.

Second-hit mutation analysis and shallow whole-genome sequencing

Somatic sequencing of genomic DNA to identify somatic aberrations in BRCA1/2 was performed using single-molecule molecular inversion probes, as described elsewhere (20). Further details of the definition of BRCA1/2 deficiency are provided in Supplementary Methods (available online). Inactivation of BRCA1/2 because of a second somatic hit or loss of the wild-type allele was defined as BRCA1/2 deficiency.

Libraries for shallow whole-genome sequencing were prepared using the Twist Library Preparation Kit (TwistBioscience, San Francisco, CA) and the Twist Universal Adapter System TruSeq Compatible Kit (TwistBioscience, San Francisco, CA) following the manufacturer’s instructions. Paired-end shallow whole-genome sequencing was performed on a NovaSeq 6000 system (Illumina, San Diego, CA), using 300 cycles to obtain an aimed coverage of approximately 1.5× per sample. Sequencing reads were aligned to GRCh38 using the Burrows-Wheeler Aligner. Subsequently, genome-wide copy number variations were called using ichorCNA, version 0.3.2, software (Supplementary Figure 1, available online). To determine genomic LOH we genotyped 1.3 × 10⁷ single-nucleotide variants (SNVs, formerly single-nucleotide polymorphisms) in each shallow whole-genome sequencing using bcftools, version 1.15, software and determined for SNVs covered by at least 5 reads the fraction of SNVs with a variant allele frequency between 0.35 and 0.65 in a bin-wise manner using regions 10 Mb in length. For each sample, the number of large-scale state transitions, telomeric allelic imbalances, and genomic LOHs were counted. Further details about copy number variation calling and determination of regions with genomic LOH are described in the Supplementary Methods (available online).

Statistical analysis

Standardized incidence ratios adjusted to age, sex, country, and birth cohort were calculated by tumor type. The cancer-specific incidence of individuals with either germline pathogenic variants in BRCA1 (n = 1795) or in BRCA2 (n = 1170) and the general population were compared (21,22). Detailed information can be found in the Supplementary Methods (available online).

For comparison of the genomic instability scores of BRCA1/2-deficient or BRCA1/2-proficient malignancies per affected gene, age, tumor cell percentage, or genomic ploidy, a Wilcoxon rank test or Fisher exact test was performed using R, version 3.6.2, software (R Foundation for Statistical Computing, Vienna, Austria). A Bonferroni adjustment was performed on P values to adjust for multiple testing. Adjusted P values were considered significant at less than .05.

Results

Malignancies in individuals with germline pathogenic variants in BRCA1/2

We selected a cohort of 2965 individuals (815 index patients) with germline pathogenic variants in BRCA1/2 from 979 families (median, 2 individuals per family [range = 1-48; interquartile range = 1-3]). In total, 1460 individuals (49% [1288/1460 (88%) women]) were diagnosed with at least 1 malignancy (Figure 1, B). Breast or ovarian carcinomas were the most frequently occurring types of malignancies, occurring in 85% (1235/1460) of individuals. In 349 (24%) individuals, non–breast or ovarian malignancies were identified, of which 124 (36%) also developed breast or ovarian carcinomas (Figure 1, A and B).

In individuals with germline pathogenic variants in BRCA1, the most frequently observed type of non–breast or ovarian malignancies were prostate carcinomas (5.4% [n = 24 men]), colorectal carcinomas (3.0% [n = 53 individuals]), and melanomas (1.5% [n = 27 individuals]) (Table 1). Individuals with germline pathogenic variants in BRCA2 were mostly diagnosed with prostate carcinomas (12.6% [n = 36 men]), colorectal carcinomas (2.7% [n = 32 individuals]), and lung carcinomas (2.3% [n = 27 individuals]) (Table 1).

Table 1.

Number of malignancies, median age, and BRCA1/2 deficiency, by sex and tumor tissue type, in individuals with a pathogenic germline variant in BRCA1/2

	BRCA1 (445 men, 1350 women)					BRCA2 (285 men, 885 women)
	Tumors in men, No.	Tumors in women, No.	Total No. of tumors	Age at diagnosis, median (range), y	BRCA1/2 deficiency/all tumors tested	Tumors in men, No.	Tumors in women, No.	Total No. of tumors	Age at diagnosis, median (range), y	BRCA1/2 deficiency/all tumors tested
Individuals with tumors
Men	N/Aa	N/A	97	67 (32-91)	8/32	N/A	N/A	122	64 (12-88)	13/48
Women	N/A	N/A	1092	47 (22-93)	13/47	N/A	N/A	635	50 (14-92)	8/42
Tumor types
Female breast	N/A	768	768	43 (23-93)	N/A	N/A	472	472	48 (23-92)	N/A
Ovarian	N/A	200	200	54 (31-82)	N/A	N/A	70	70	59 (26-87)	N/A
Male breast	0	N/A	0	N/A	N/A	17	N/A	17	57 (48-84)	N/A
Prostate	24	0	24	69 (57-84)	0/8	36	0	36	64.5 (50-80)	8/19
Pancreatic	2	5	7	67 (58-80)	1/2	5	6	11	56 (45-65)	3/3
Colorectal	24	29	53	67 (36-91)	8/19	18	14	32	59 (12-83)	1/16
Melanoma	6	21	27	45 (22-82)	1/10	4	15	19	50 (21-80)	1/9
Lung	5	12	17	63 (45-79)	0/3	11	16	27	64.0 (50-85)	2/9
Urinary tract	18	7	25	67 (56-84)	2/12	12	9	21	67 (42-88)	2/10
Endometrial	0	15	15	62 (51-77)	2/6	0	5	5	57 (48-65)	1/4
Lymphoma	6	7	13	60 (48-76)	1/3	5	5	10	63.5 (15-81)	0/5
Esophageal	4	4	8	56.5 (42-83)	2/5	2	6	8	61.5 (40-79)	1/3
Other	8	24	32	63.5 (37-87)	4/11	12	17	29	61 (14-83)	2/12

a

N/A = not applicable.

To identify which non–breast or ovarian malignancies were more frequent in our cohort compared with the general Dutch population, standardized incidence ratio analyses were performed. Individuals with germline pathogenic variants in BRCA1 showed an increased standardized incidence ratio for endometrial carcinomas (standardized incidence ratio = 3.1, 95% confidence interval [CI] = 1.9 to 4.8) (Figure 1, C).

Individuals with germline pathogenic variants in BRCA2 showed an increased standardized incidence ratio for male breast carcinomas (standardized incidence ratio = 210.3, 95% CI = 117.6 to 346.9), pancreatic carcinomas (standardized incidence ratio = 4.6, 95% CI = 2.3 to 8.2), prostate carcinomas (standardized incidence ratio = 3.2, 95% CI = 2.0 to 4.9), esophageal carcinomas (standardized incidence ratio = 3.5, 95% CI = 1.4 to 7.2), and urinary tract carcinomas (standardized incidence ratio = 2.4, 95% CI = 1.3 to 4.0) (Figure 1, D).

BRCA1/2 deficiency in malignancies of individuals with germline pathogenic variants in BRCA1/2

For the somatic mutation analysis, genomic DNA of sufficient quantity and quality for successful sequencing was available for 169 malignancies from 20 different tumor types; 79 (47%) malignancies developed in individuals with germline pathogenic variants in BRCA1, and 90 (53%) malignancies developed in individuals with germline pathogenic variants in BRCA2 (Figure 1, A; Supplementary Tables 1 and 2, available online). In individuals with germline pathogenic variants in BRCA1, BRCA1 deficiency was identified in 27% (21/79) of non–breast or ovarian malignancies (Figure 2, A left panel; Supplementary Figure 2, A left panel, available online; Table 1). In 6% (5/79) of malignancies, we could not establish whether a second hit was present, and these malignancies were termed inconclusive. The highest number of BRCA1-deficient malignancies was observed in colorectal carcinomas (8/18 unequivocal results [42%]). Overall, we observed loss of the BRCA1 wild-type allele in 25% (20/79) of malignancies with unequivocal results, whereas in 4% (3/79) of malignancies, the allele with the germline pathogenic variant was lost (P < .01) (Supplementary Tables 1 and 3, available online). BRCA1 deficiency was not statistically significant associated with patients who had multiple primary tumors (P > .99) or patients with breast or ovarian carcinomas (P > .99) (Supplementary Table 1, available online).

Figure 2.

BRCA1/2 deficiency and genomic instability score, by tumor type, in individuals with a pathogenic germline variant in BRCA1 or BRCA2. A, B) The left images show the frequency of BRCA status, by cancer tissue, in individuals with germline pathogenic variants in BRCA1 (A) or BRCA2 (B). Deficiency is based on targeted sequence analyses and the shallow whole-genome sequencing–based copy number analyses of the respective genes. Data were analyzed for BRCA1/2 deficiency, BRCA1/2 proficiency, and inconclusive results. The number of tested malignancies is given, by malignancy type, at the left of the figure. A, B) The right images show the genomic instability score for each malignancy in individuals with germline pathogenic variants in BRCA1 (A) or BRCA2 (B), by tumor type. Genomic instability scores are calculated by the sum of large-scale state transitions, telomeric allelic imbalances, and genomic loss of heterozygosity. BRCA status results show malignancies with BRCA1/2 deficiency, BRCA1/2 proficiency, and inconclusive results. “Others” include malignancies of soft tissue sarcomas, stomach, small bowel, kidney, larynx, oral cavity, testis, thyroid, and uterus, included in Supplementary Figure 2 (available online). C) Box plot of the genomic instability score, by BRCA status and gene. BRCA status results show malignancies with BRCA1/2 deficiency, BRCA1/2 proficiency, and inconclusive results. ns = not significant. *P < .05, **P < .01, ***P < .001. D) Box plot of the genomic instability score, by genomic ploidy and BRCA status. Genomic ploidy is diploid, triploid, or tetraploid. BRCA status results show malignancies with BRCA1/2 deficiency, BRCA1/2 proficiency, and inconclusive results. ns = not significant; *P < .05, **P < .01, ***P < .001.

In individuals with germline pathogenic variants in BRCA2, BRCA2 deficiency was identified in 23% (21/90) of non–breast or ovarian malignancies (Figure 2, B left panel; Supplementary Figure 2, B left panel, available online; Table 1). In 6% (5/90) of malignancies, it was unclear whether a second hit was present, and these malignancies were termed inconclusive. The highest number of BRCA2-deficient malignancies was observed in prostate carcinomas (8/19 [42%]). The BRCA2 wild-type allele was lost in 20% (18/90) of malignancies with unequivocal results, and the allele with the germline pathogenic variant was lost in 17% (15/90) of malignancies (P = 1) (Supplementary Tables 2 and 3, available online). BRCA2 deficiency was not significantly associated with patients who had multiple primary tumors (P = .78) or patients with breast or ovarian carcinomas (P > .99) (Supplementary Table 2, available online).

Genomic instability score in malignancies of individuals with a germline pathogenic variant in BRCA1/2

Next, we investigated large-scale state transitions, telomeric allelic imbalances, and genomic LOHs as genomic instability markers indicative of homologous recombination deficiency (Figure 3), which are associated with BRCA1/2 deficiency in breast and ovarian carcinomas (11-13). Overall, we were able to perform shallow whole-genome sequencing for 158 of the 169 malignancies previously successfully analyzed with single-molecule molecular inversion probe testing, of which 76 (48%) malignancies were from individuals with germline pathogenic variants in BRCA1 and 82 (52%) malignancies from individuals with germline pathogenic variants in BRCA2.

Figure 3.

Contribution of different types of genomic aberrations to the homologous recombination status of malignancies in individuals with germline pathogenic variants in BRCA1/2. A, B) The count of large-scale state transitions (rectangle), telomeric allelic imbalances (triangle), and genomic loss of heterozygosity (circle) for each malignancy in individuals with germline pathogenic variants in BRCA1 (A) or BRCA2 (B) were visualized and categorized by ovarian cancer controls, malignancies with BRCA1/2 deficiency, malignancies with BRCA1/2 proficiency, and malignancies with inconclusive somatic mutation results and sorted based on large-scale state transition scores. The result of the somatic mutation analysis and the tissue of origin for each malignancy is given below the figure. Somatic aberrations represent loss or inactivation of the wild-type allele, pathogenic somatic mutation, no second hit detected, loss of the mutant allele, or inconclusive results. Genomic ploidy represents diploid, triploid, or tetraploid genomic status. Malignancies analyzed were ovarian cancer controls; prostate, pancreatic, colorectal, lung, urinary tract, endometrial, and esophageal carcinomas; melanoma; lymphoma; and other malignancies (included in Supplementary Figure 2, available online).

For comparison, we included control cases of ovarian carcinomas with confirmed BRCA1/2 deficiency (Supplementary Tables 1 and 2, available online). The median genomic instability score for BRCA1- and BRCA2-deficient ovarian carcinomas was 51 (range = 38-64) and 66 (range = 36-85), respectively (Figure 2, A, right panel; Figure 2, B, right panel; Figure 3; and Supplementary Figure 2, A and B, right panel, available online). The median genomic instability score for BRCA1- and BRCA2-deficient non–breast or ovarian malignancies was 43 (range = 3-74) and 33 (range = 15-59), respectively (Figure 2, C; Supplementary Figure 3, A, available online). The genomic instability scores were significantly higher than those for BRCA1/2-proficient malignancies (BRCA1: median, 43 vs 10, P < .001; BRCA2: median, 33 vs 13, P < .001) (Figure 2, C; Supplementary Figure 3, A, available online). No significant difference was observed between the median genomic instability scores of BRCA1- and BRCA2-deficient malignancies (P = .92). Also, BRCA1- and BRCA2-proficient malignancies did not show a statistically significant difference between the median genomic instability scores (P > .99) (Figure 2, C). In addition, no statistically significant difference in genomic instability score was detected between malignancies with a low or high tumor cell percentages (tumor cell percentage ≤40%: median, 19 vs tumor cell percentage >40%: median, 22; P > .99) or between malignancies developed at a young/middle or later age (age <70 years: median, 22 vs age ≥70 years: median, 20; P > .99).

When we consider a genomic instability score equal to or above 42 as homologous recombination deficiency (23,24), homologous recombination deficiency was found in 63% (12/19) of BRCA1-deficient and 4% (2/51) of BRCA1-proficient malignancies (P < .001) and in 29% (5/17) of BRCA2-deficient and 3% (2/60) of BRCA2-proficient malignancies (P = .09) (Supplementary Tables 1 and 2, available online).

BRCA1/2-deficient malignancies do not present with higher genomic instability scores in malignancies with whole-genome duplications (tetraploid genomes) than in malignancies with diploid genomes (median, diploid: 39 vs median, tetraploid: 67; P = .40) (Figure 2, D; Supplementary Figure 3, B, available online). Malignancies with BRCA1/2 proficiency and whole-genome duplications showed a higher genomic instability score than those with diploid genomes (median, diploid: 10 vs median, tetraploid: 33; P < .01) (Figure 2, D; Supplementary Figure 3, B, available online).

Discussion

It is well established that individuals with germline pathogenic variants in BRCA1/2 who develop breast or ovarian carcinomas but also other malignancies can benefit from treatment with platinum-based therapies or PARP inhibitors (16,25) because of loss of the wild-type allele and subsequent homologous recombination deficiency in tumors (17). Here, we demonstrate that a substantial proportion (27% and 23%) of non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1/2 are BRCA1 and BRCA2 deficient, respectively. Moreover, these BRCA1/2-deficient malignancies presented with higher genomic instability scores than BRCA1/2-proficient malignancies, demonstrating that their homologous recombination repair mechanism is likely impaired. Our data suggest that on the basis of their genomic profile, one-quarter of non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1/2 may have increased sensitivity to platinum-based therapies or PARP inhibitors.

We found an increased standardized incidence ratio for endometrial carcinomas in individuals with germline pathogenic variants in BRCA1. Although this finding was similar to a previous Dutch study (26), in other studies, endometrial carcinoma was not associated with germline pathogenic variants in BRCA1 (6,7). BRCA1 deficiency in 2 out of 5 interpretable endometrium carcinomas could point toward an increased risk on the basis of the etiologic index (27), but these numbers are too small to draw conclusions. Based on the standardized incidence ratio, we could not confirm the suggested increased risk for colorectal, pancreatic, and gastric carcinomas (6,7). Even though for most non–breast or ovarian malignancies in our cohort the incidence does not seem to have increased, BRCA1 deficiency was observed in most types of non–breast or ovarian malignancies, and this frequency was higher than in cohorts of sporadic non–breast or ovarian malignancies, in which BRCA1 deficiency is present in less than 3% (14,28). Remarkably we did not observe an increased standardized incidence ratio for colorectal carcinomas, although BRCA1 deficiency was observed in nearly half of these cancers, which is indicative of a standardized incidence ratio of 2.25 according to the theory of the etiologic index (27). This standardized incidence ratio is not within our observed confidence interval (standardized incidence ratio = 1.4, 95% CI = 1.0 to 1.9). Altogether, an increased risk of colorectal cancer in individuals with a germline pathogenic variant in BRCA1 cannot be excluded.

In individuals with germline pathogenic variants in BRCA2, we detected an increased incidence of pancreatic, prostate, esophageal, and urinary tract carcinomas compared with the general population. Previous studies also indicated an increased risk for pancreatic, prostate, and esophageal carcinomas in individuals with germline pathogenic variants in BRCA2 (6,7). Two urinary tract carcinomas occurred synchronously or after a prostate cancer diagnosis and therefore may be screen-detected cancers, possibly leading to an overestimation of the standardized incidence ratio. Prostate cancer had the highest number of cases that could be evaluated for BRCA2 deficiency, and the associated etiologic index of 2.11 indeed also affirms an increased risk of prostate cancer (27). As with the BRCA1-deficient malignancies, malignancies with BRCA2 deficiency show higher genomic instability scores than malignancies with BRCA2 proficiency, which indicates an impaired homologous recombination repair pathway that can be treated with platinum-based therapies or PARP inhibitors (16,17,25). Overall, the frequency of BRCA1/2 deficiency in non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1 or BRCA2 was similar (both 27%), and this finding is comparable to a recent study in which a cohort of 45 non–breast or ovarian carcinomas and a cohort from The Cancer Genome Atlas of 73 non–breast or ovarian carcinomas presented with 20% and 32% of BRCA1/2 deficiency, respectively (18).

The genomic instability scores for BRCA1-deficient and BRCA2-deficient malignancies were not significantly different in our study, suggesting that the genomic scars included in the genomic instability score that are widely used for ovarian and breast cancer is informative for other BRCA1/2-affected tumor types, as well (11-13). Moreover, a recent study revealed that a normalization of the large-scale state transitions based on genomic ploidy status may be needed for greater accuracy (29). In our study, the presence of a whole-genome duplication event led to a significantly increased genomic instability score in BRCA1/2-proficient carcinomas, although in ichorCNA, the genomic instability scores were adapted based on genomic ploidy status. A recent study also measured homologous recombination deficiency using shallow whole-genome sequencing. This study focused on large genomic alterations, not taking telomeric allelic imbalance and genomic LOH into account, and found a 94% agreement with the MyChoice test (Myriad Genetics, Salt Lake City, UT) (30). Further studies are needed to investigate which genomic scars and which thresholds are optimal for therapy selection.

It has been described that the second hit resulting in BRCA1/2 deficiency is often LOH at the BRCA1/2 loci (31). In our study, we confirmed that BRCA1/2 deficiency is mainly caused by loss of the wild-type allele in BRCA1/2-deficient malignancies. We noticed that an increased variant allele frequency of the pathogenic variant and alterations in the variant allele frequency of neighboring SNVs may not be sufficient to define BRCA1/2 deficiency because this state may also be caused by duplication of the allele with the pathogenic variant without loss of the wild-type allele. Therefore, targeted sequencing should be combined with copy number variant analyses to reliably determine whether non–breast or ovarian malignancies are BRCA1/2 deficient. Interestingly, loss of the wild-type allele occurred 6.3 times more often than loss of the allele with the pathogenic variant in malignancies with germline pathogenic variants in BRCA1 and 1.2 times more often in malignancies with germline pathogenic variants in BRCA2. This observation suggests that although the loss of an allele may occur by chance, at least in cases of germline pathogenic variants in BRCA1, loss of the wild-type allele may give cells a selective advantage and thus may drive cancer development independent of tissue origin.

This study has a few limitations. We could not calculate the family-specific cancer burden because most families were small. Furthermore, our study likely underestimated the effect size of most standardized incidence ratios because malignancies were requested from Palga, which contains only excerpts from pathology reports submitted since 1991 (19) and we included only those malignancies confirmed by a pathology report. In addition, the date of death was unknown; hence, we used the last date of follow-up from Palga data to calculate the standardized incidence ratios. Therefore, not all years at risk were included, which may have led to a lower standardized incidence ratio. Primary malignancies were selected based on pathology reports and may have included unrecognized metastases. Based on our review of clinical history, histology, and molecular analyses, we tried to distinguish metastatic cancers from true primary malignancies. For example, analysis of tumor histology and copy number profiles from shallow whole-genome sequencing analysis revealed that 1 lung carcinoma and 2 endometrial carcinomas were not primary carcinomas, although they were initially classified as primary malignancies. Information about external risk factors was not available and therefore was not included in the standardized incidence ratio analysis. Furthermore, our genomic instability scores were probably underestimated because genomic LOH scores calculated from shallow whole-genome sequencing data with a coverage of 0.35× to 9.05× are less sensitive than analysis from a set of SNVs with much higher coverage than has been used in other studies (14,18,28,29,31).

In conclusion, individuals with germline pathogenic variants in BRCA1 or BRCA2 have different incidences of certain cancers. More than one-quarter of non–breast or ovarian malignancies in these individuals present with BRCA1/2 deficiency, which is a higher frequency than in malignancies in individuals without germline pathogenic variants in BRCA1/2 (14,28). In our study, we showed that BRCA1/2 deficiency and its association with genomic instability is a relatively common event in a broad spectrum of malignancies in individuals with germline pathogenic variants in BRCA1/2. Therefore, testing non–breast or ovarian malignancies in individuals with germline pathogenic variants in BRCA1/2 for the presence of BRCA1/2 deficiency or homologous recombination deficiency should be considered because the presence of homologous recombination deficiency may help determine treatment choices.

Data availability

Targeted sequencing data are available upon reasonable request from the European Genome-Phenome Archive (study identifier EGAS00001007258).

Author contributions

Lisa Elze, MSc (Data curation; Formal analysis; Investigation; Methodology; Project administration; Validation; Visualization; Writing—original draft); Rachel van der Post, PhD, MD (Data curation; Investigation; Methodology; Supervision; Validation; Writing—review & editing); Janet Vos, PhD (Data curation; Formal analysis; Investigation; Methodology); Arjen Mensenkamp, PhD (Investigation; Methodology; Validation); Samhita Pamidimarri Naga, MSc (Investigation); Juliet Hampstead, PhD (Investigation); Emma Vermeulen, BSc (Data curation; Formal analysis; Investigation; Project administration); Michiel Oorsprong, BSc (Data curation); Tom Hofste, BSc (Data curation); Michiel Simons, PhD, MD (Investigation); Iris Nagtegaal, PhD, MD (Investigation); Nicoline Hoogerbrugge, PhD, MD (Funding acquisition; Supervision); Richarda M. de Voer, PhD (Conceptualization; Methodology; Supervision; Validation; Visualization; Writing—original draft); Marjolijn Ligtenberg, PhD (Conceptualization; Funding acquisition; Methodology; Supervision; Validation; Visualization; Writing—review & editing).

Funding

Not applicable.

Conflicts of interest

M.J.L.L. received research funding from AstraZeneca, GlaxoSmithKline, Illumina, and Janssen Pharmaceuticals. None of this funding was related to this study, and funding was paid to the institution. A.R.M. received funds from AstraZeneca for contributions to sponsored quality assessment and variant interpretation of variants of uncertain significance in BRCA1 and BRCA2. This funding was not related to this study. All other authors declare that they have no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Prior presentations: National Hebon Congress, November 29, 2022, Utrecht, the Netherlands.

Acknowledgements

The authors would like to thank Neeltje Arts, Hicham Ouchene, Hanneke Volleberg-Gorissen, Annemiek Kastner-van Raaij, Erik Jansen, Daniel von Rhein, Franziska Bervoets-Metge and Snežana Hinić (Radboud university medical center, Nijmegen, the Netherlands) for expert assistance; the Genome Technology Center (Radboud university medical center, Nijmegen, the Netherlands) for library preparation and single-molecule molecular inversion probe sequencing support; and Palga for providing data for this study. We thank the following Dutch pathology laboratories for sharing their material: Amsterdam UMC, Amphia Ziekenhuis, Canisius Wilhelmina Ziekenhuis, Erasmus MC, Gelre Ziekenhuis, Isala Klinieken, Jeroen Bosch Ziekenhuis, Klinische Pathologie Deventer, Laboratorium Klinische Pathologie Midden Brabant, Meander Medisch Centrum, Medisch Centrum Alkmaar/Symbiant, Nederlands Kankerinstituut-Antoni van Leeuwenhoek Core Facility Molecular Pathology & Biobanking (institutional review board), Pathologisch Laboratorium voor Dordrecht, Reinier de Graaf Ziekenhuis, Rijnstate Ziekenhuis, Stichting Laboratorium Pathologie Oost Nederland, Stichting PAMM, Tergooiziekenhuis, UMC Groningen, UMC Utrecht, and VieCuri.

Author notes

Richarda M. de Voer and Marjolijn J. L. Ligtenberg contributed equally to this work.