Evaluation of the contribution of germline variants in BRCA1 and BRCA2 to uveal and cutaneous melanoma

Abstract

Germline mutations of BRCA1 and BRCA2 predispose individuals to a high risk of breast and ovarian cancer, and elevated risk of other cancers, including those of the pancreas and prostate. BRCA2 mutation carriers may have increased risk of uveal melanoma (UM) and cutaneous melanoma (CM), but associations with these cancers in BRCA1 mutation carriers have been mixed. Here we further assessed whether UM and CM are associated with BRCA1 or BRCA2 by assessing the presence, segregation and reported/predicted pathogenicity of rare germline mutations (variant allele frequency (VAF) <0.01) in families with multiple members affected by these cancers.

Whole-genome or exome sequencing was performed on 160 CM and/or UM families from Australia, the Netherlands, Denmark, and Sweden. Between one and five cases were sequenced from each family, totalling 307 individuals. Sanger sequencing was performed to validate BRCA1 and BRCA2 germline variants and to assess carrier status in other available family members.

A nonsense and a frameshift mutation were identified in BRCA1, both resulting in premature truncation of the protein (the first at p.Q516 and the second at codon 91, after the introduction of 7 amino acids due to a frameshift deletion). These variants co-segregated with CM in individuals who consented for testing and were present in individuals with pancreatic, prostate and breast cancer in the respective families. Additionally, 33 rare missense mutations (VAF ranging from 0.00782 to 0.000001 in the aggregated ExAC data) were identified in 34 families. Examining the previously reported evidence of functional consequence of these variants revealed all had been classified as either benign or of unknown consequence.

Seeking further evidence of an association between BRCA1 variants and melanoma, we examined two whole genome/exome sequenced collections of sporadic CM patients (total N = 763). We identified one individual with a deleterious BRCA1 variant, however, this allele was lost (with the wild-type allele remaining) in the corresponding CM, indicating that defective BRCA1 was not a driver of tumorigenesis in this instance.

While this is the first time deleterious BRCA1 mutations have been described in high-density CM families, we conclude that there is an insufficient burden of evidence to state that the increased familial CM or UM susceptibility is due to these variants. Additionally, in conjunction with other studies, we conclude that the previously described association between BRCA2 mutations and UM susceptibility represents a rare source of increased risk.

Introduction

Although rare, melanoma of the uveal tract (the iris, ciliary body and choroid) is the most common intraocular malignancy in adults, with an incidence of 5.6 cases per million, per year in the USA SEER (Surveillance, Epidemiology and End Result) database [1]. Several somatic mutations have been associated with uveal melanoma (UM) development and progression, including GNAQ [2], GNA11 [3], BAP1 [4], SF3B1 [5], PLCB4 [6], CYSTLR2 [7], and EIF1AX [8]. UM also has a heritable component, with germline variants in BAP1 [9,10], and rarely CDKN2A [11,12], being associated with predisposition.

Cutaneous melanoma (CM) accounts for ~4% of skin cancers, but ~75-80% of skin cancer deaths [13]. CM results from the malignant transformation of melanocytes, the pigment-producing cells responsible for skin, hair and eye colour. CM have highly aberrant genomes, with driver mutations mainly affecting the mitogen-activated protein kinase pathway and telomere maintenance [14]. As with UM, susceptibility to CM is sometimes heritable due to single-gene defects, with germline mutations in CDKN2A, CDK4, BAP1, MITF, TERT, POT1, ACD and TERF2IP contributing to CM development in high-density melanoma families [15].

The cellular origin for UM and CM are similar, but melanoblasts migrate to the epithelia in CM, while in UM, they migrate to mesodermic tissues. Indeed, the coexistence of UM and CM in some patients suggests a similar predisposition to both melanoma subtypes [16], exemplified by predisposition to CM and UM conferred by germline BAP1 and CDKN2A mutations. There are significant differences between UM and CM in environmental risks and associations with other tumour types. For example: unlike CM, UM has not been linked to ultra-violet radiation (UVR) exposure; and UM has been associated with an increased incidence of breast adenocarcinoma [17] and as a second malignancy among long-term survivors of ovarian cancer [18], which is not the case with CM [19].

Germline mutations of BRCA1 and BRCA2 predispose individuals to a high risk of breast and ovarian cancer [20-22]. Both BRCA1 and BRCA2 are crucial for the process of homologous recombination repair, largely involving the repair of DNA lesions that stall DNA replication forks and/or cause DNA double-strand breaks. In cohorts of BRCA1 or BRCA2 mutation carriers, increased risks of other cancers have been observed, including those of the pancreas and prostate. Both UM [23-28] and CM were found at significantly increased frequency in BRCA2 mutation carriers [27,29,30], but only a trend towards increased risk of CM in BRCA1 mutation carriers was found [20,25,26,28,31,32]. One screen for germline variants in BRCA1 or BRCA2 in 82 individuals with both a breast cancer and CM diagnosis revealed 2 pathogenic BRCA1 and 2 pathogenic BRCA2 mutations [33]. Some screens of UM patients have revealed a small number of deleterious truncating mutations in BRCA2 [34,35], but this was not seen in all cohorts [36].

We sought to assess the role that novel or rare (VAF<0.01) germline BRCA1 and BRCA2 mutations play in melanoma susceptibility in a large, well characterised, cohort of families and individuals with CM and/or UM. Potential pathogenicity of the mutations was analysed with respect to co-segregation with melanoma, bioinformatic prediction of protein effect and evidence from functional studies.

Materials and Methods

Samples Used for Whole-Genome/Exome Sequence Analysis

All participants gave written informed consent for participation and were wild-type for germline CDKN2A, CDK4 and BAP1 mutations. In the Australian cohorts, a family history of melanoma is considered to be present when three or more close relatives have been diagnosed with either CM or UM; in the European cohorts, this is classified as two or more close relatives. Whole genome (WGS) or exome (WES) sequencing was performed on CM or CM/UM families or on individuals with UM, resulting in between 1 and 5 melanoma cases being sequenced per family. Queensland samples were selected from those ascertained as part of the Q-MEGA project [37], or were recruited following referral from specialist clinicians and the Queensland Ocular Oncology Service (QOOS). Q-MEGA is a population-based study investigating the link between genetics and environment in melanoma development and consists of four study samples: The Queensland Study of Childhood Melanoma (n=101); The study of Melanoma in Adolescents (n=298); The Study of Men over 50 (n=178); and the Queensland Familial Melanoma Project (QFMP; n=1897) [38]. All Queensland participants with a family history of CM or CM/UM or individuals with UM provided detailed information on personal and family cancer history; additional family members (both cancer affected and unaffected) were invited to participate. The number of sequenced families and individuals with UM from Queensland are 118, totalling 233 affected participants. Thirty-two Danish individuals from seventeen families were ascertained through the Danish Project of Hereditary Malignant Melanoma. Seventeen samples from ten families were ascertained from the Australian Melanoma Family Study [39]. Sixteen patients were ascertained from eight families at the Oncogenetic Clinic at Skåne University Hospital, Sweden. The Karolinska Institute, Sweden, ascertained nine individuals from seven families. In total, 307 individuals from 160 families are included in this study (Supplementary Table 1).

Ethical approval

Ethics approval was granted by: the Committee of Biomedical Research Ethics of the Capital Region of Denmark, and the Human Research Ethics Committees of the QIMR Berghofer Medical Research Institute, Lund University, the University of Sydney, and the Karolinska Institute.

Next-Generation Sequencing

WGS or WES was performed on 298 individuals from 160 high density CM, CM/UM families, or UM cases from Australia, Denmark, and Sweden, to examine if germline variants in BRCA1 or BRCA2 predispose to CM or UM (Supplementary Table 1). Sequencing was performed by Macrogen (South Korea) on the Illumina HiSeq 2000 platform, with mean coverage of 60 to 96 X. Using the Burrows-Wheeler Aligner, the sequence output was mapped to the UCSC human genome reference build 19 [40]. SNPs were detected using bcftools and SAMtools mpileup with disabled BAQ computation [41], in/del were detected with pindel [42] and both were annotated to dbSNP144, including ExAc population frequencies [43], by ANNOVAR [44]. Variants were filtered for stringency using quality score (>70) and alternate read counts (>2 and >20% of all reads at a given position). Sanger sequencing was used to confirm the variants found by next-generation sequencing and assess co-segregation in other family members. Primers are listed in Supplementary Table 2.

Results and Discussion

With a population frequency cut-off of <0.01, 10 variants in BRCA1 were identified in 13 families (Table 1) and 24 variants in BRCA2 were identified in 30 families (Table 2).

Table 1: Rare BRCA1 mutations identified in cutaneous and/or uveal family members.

Variants indicated in bold font are present in all CM affected family members available for testing.

Family ID	Amino acid change	Genomic position	dbSNPID	Population frequency	Number affected with variant / Total affected tested	Number of breast cancer, ovarian and prostate cancer patients genotyped in family*
BRCA1: Cutaneous Melanoma Family
Aus1	p.T1726S	17:41203095	rs80357324	0.0000065	1 / 3	0, 0, 0
Aus2	p.S1465I	17:41226488	rs1800744	0.00215	4 / 5	1 (het), 0, 1 (WT)
Aus3	p.S1465I	17:41226488	rs1800744	0.00215	1 / 2	1 (bilat, hom), 0, 0
Den1	p.S1465I	17:41226488	rs1800744	0.00215	3 / 3	0, 0, 0
Aus4	p.R1300G	17:41243509	rs28897689	0.00398	2 / 3	0, 0, 0
Aus5	p.R1300G	17:41243509	rs28897689	0.00398	2 / 2	0, 0, 0
Aus6	p.N1189K	17:41243840	rs28897687	0.00024	1 / 4	0, 0, 0
Aus7	p.N1189K	17:41243840	rs28897687	0.00024	3 / 4	0, 0, 2 (het; WT)
Aus8	Frameshift†	17:41244865	n/a	0	5 / 6	1 (het), 0, 0
Aus9	p.T779K	17:41245071	rs28897683	0.00018	3 / 7	1 (WT), 0, 0
Swe1	p.Q516X	17:41245861	rs80356898	0.000038	2 / 2	0, 0, 0
BRCA1: Cutaneous and Uveal Melanoma Family
Aus10	p.D167G	17:41247892	rs55680408	0.00002	2 / 3	1 (het), 0, 0
Aus10	Splice†	17:41247941	rs80358033	0.00001	2 / 3	1 (het), 0, 0
BRCA1: Uveal Melanoma Family
Den2	p.E1172D	17:41243891	rs80356876	0.00002	1 / 2	0, 0, 0

^*Column order: breast, ovarian, prostate cancer. WT=wild-type; het=heterozygous for variant; hom=homozygous for variant; bilat=bilateral cancer

^†The frameshift variant is: c.2450-2451 TT deletion. The splice variant is: c.594-2A>C

Table 2: Rare BRCA2 mutations identified in cutaneous and/or uveal family members.

Variants indicated in bold font are present in all CM affected family members available for testing.

Family ID	Amino acid change	Genomic position	dbSNPID	Population frequency	Number affected with variant / Total affected tested	Number of breast cancer, ovarian and prostate cancer patients genotyped in family*
BRCA2: Cutaneous Melanoma Family
Aus11	p.Y42C	13:32893271	rs4987046	0.00169	4 / 6	0, 0, 2 (het; het)
Aus12	p.F81I	13:32893387	rs80358507	0.0000188	1 / 2	0, 0, 1 (het)
Aus6	p.S384F	13:32906766	rs41293475	0.00068	2 / 4	0, 0, 0
Den3	p.S489C	13:32907081	rs587782535	0.0000517	2 / 2	0, 0, 0
Aus13	p.D935H	13:32911295	rs28897716	0.00005	1 / 9	0, 0, 2 (WT; WT)
Aus14a	p.D935N	13:32911295	rs28897716	0.00005	2 / 3	0, 0, 1 (het)
Aus15	p.E1382 del	13:32912638	rs80359432	0.0000712	1 / 1	0, 0, 0
Aus16	p.D1420Y	13:32912750	rs28897727	0.00671	1 / 6	0, 0, 2 (WT; WT)
Swe2	p.D1420Y	13:32912750	rs28897727	0.00671	1 / 3	0, 0, 0
Swe2	p.I1705V	13:32913605	rs80358737	0.0000517	1 / 3	0, 0, 0
Aus17	p.L1908P	13:32914215	rs80358797	0.0000065	1 / 1	1 (het), 0, 0
Aus18	p.A2671S	13:32937350	rs786201976	0	3 / 4	0, 0, 0
Aus19	p.A2717S	13:32937488	rs28897747	0.00116	1 / 1	0, 0, 0
Aus20	p.V2728I	13:32937521	rs28897749	0.00203	1 / 3	1, 0, 0
Aus21	p.V2728I	13:32937521	rs28897749	0.00203	2 / 7	1 (bilat, WT), 0, 0
Aus22	p.K2950N	13:32953549	rs28897754	0.00068	1 / 1	0, 0, 0
Aus6	p.A2951T	13:32953550	rs11571769	0.00782	2 / 4	0, 0, 0
Aus4	p.A2951T	13:32953550	rs11571769	0.00782	3 / 3	0, 0, 0
Aus23	p.A2951T	13:32953550	rs11571769	0.00782	1 / 5	2 (bilat, WT; WT), 0, 0
Aus24	p.A2951T	13:32953550	rs11571769	0.00782	2 / 2	1 (het), 0, 0
Aus25	p.T3013I	13:32953971	rs28897755	0.00023	3 / 5	0, 0, 3 (het; het; WT)
Aus26	p.K3326X	13:32972626	rs11571833	0.00701	2 / 5	0, 0, 0
Aus21	p.K3326X	13:32972626	rs11571833	0.00701	1 / 4	1 (bilat, WT), 0, 0
BRCA2: Cutaneous and Uveal Melanoma Family
Aus27	p.A75P	13:32893369	rs28897701	0.00016	3 / 3	0, 0, 0
Aus28	p.S384F	13:32906766	rs41293475	0.00068	1 / 3	1 (WT), 0, 0
Aus29	p.I505T	13:32907129	rs28897708	0.00072	1 / 7	0, 0, 1 (WT)
Aus30	p.R2034C	13:32914592	rs1799954	0.00323	1 / 2	0, 0, 0
Aus14b	p.E2856A	13:32945172	rs11571747	0.00088	3 / 4	0, 0, 1 (het)
Aus27	p.K3326X	13:32972626	rs11571833	0.00701	1 / 3	0, 0, 0
BRCA2: Uveal Melanoma Family
Aus31	p.T598A	13:32907407	rs28897710	0.00227	1 / 1	0, 0, 0
Aus32	p.S617P	13:32907464	rs587782871	0.0000065	1 / 1	0, 0, 0
Aus33	p.D1420Y	13:32912750	rs28897727	0.00671	1 / 1	0, 0, 0
Swe3	p.S2247G	13:32915231	rs80358896	0.0000194	1 / 2	0, 0, 0
Den2	p.E2856A	13:32945172	rs11571747	0.00088	2 / 2	0, 0, 0

^*Column order: breast, ovarian, prostate cancer. WT=wild-type; het=heterozygous for variant; bilat=bilateral cancer

In BRCA1, a deleterious novel mutation (i.e. not present in the ExAC aggregated population, no dbSNP ID and not present in the LOVD BRCA1 database) was identified: a c.2450-2451 TT deletion resulting in a frameshift and introduction of seven amino acids (894: T-K-S-K-S-H-F :900) before a stop at codon 901 (Figure 1A). This variant prematurely terminates the protein by 962 amino acids. The protein product would lack several important functional domains, including two BRCA1 C-terminal (BRCT) domains, which are integral signalling modules in the DNA damage response, and serine residues that are important phosphorylation targets by proteins involved in the control of DNA damage response and cell cycle control (e.g. ATM, ATR, CHK2 and cyclin E). This Australian family was independently referred for genetic counselling/testing by their primary care physician, due to the high incidence of breast cancer. This mutation was present in 5/6 available family members who had CM; other cancers in carriers in this family included: pancreatic cancer (age 53 years), renal transitional cell carcinoma (TCC) (age 75 years), prostate cancer (first individual, age 80 years, second individual, age 59 years) and colorectal cancer (age 63 years); Figure 2. A non-mutation carrier had early onset testicular cancer (age 32 years) and prostate cancer (family recollection, age unknown). Several other cancers are present in family members who did not donate DNA for this project, so their mutation status is unknown; these include meningioma, prostate cancer, acoustic schwannoma and breast cancer (data not shown).

Figure 1:

The structure of the BRCA1 (A) and BRCA2 (B) proteins, with the location of identified mutations indicated.

Figure 2: Pedigrees for the families with loss of function mutations in BRCA1.

The individuals available for segregation analysis using Sanger sequencing and the result of mutation screening are indicated on each family pedigree.

A) Family Aus8, frameshift mutation (c.2450-2451 TT deletion), resulting in the introduction of a stop codon at p.901.

B) Family Swe1, with a p.Q516X nonsense mutation

WT = wildtype for the mutation and M = mutation present. Upper ages on pedigree are ages at which indicated cancers are diagnosed; lower ages are the age at last follow-up confirming unaffected status.

A second deleterious variant was identified in BRCA1, a c.C1546T mutation (rs80356898), leading to a premature stop at p.Q516X; this is classified as pathogenic by ClinVar (Supplementary Table 3). This mutation was not present in the LOVD BRCA1 database (Supplementary Table 4). The protein truncates 1347 amino acids prematurely, resulting in significant loss of many functional domains, including RAD50, RAD51, MSH2, ATM, PALB2 and BACH1 binding domains, as well as serine residues that are important phosphorylation targets of proteins involved in the control of DNA damage response and cell cycle control. The p.Q516X mutation was present in both available Swedish individuals who had CM; no further information was available regarding other cancers in this family (Figure 2). Family members have been referred for genetic counselling/testing.

This is the first time BRCA1 mutations have been described in families that were specifically ascertained due to a history of CM. While the deleterious variant does not fully segregate in the larger Australian family, it should be noted that phenocopies are common in Australian familial melanoma cohorts due to high ambient UVR exposure. Previously, studies have assessed which cancers are present in BRCA1 cohorts collected due to a family history of breast and/or ovarian cancer; these have not revealed a robust link with CM [20,25,26,28,31,32]. A significant indicator of the importance of BRCA1 and BRCA2 in breast and ovarian cancers has been the second hit to these genes in the tumours of germline mutation carriers [45]. Unfortunately, no tumour tissue was available from mutation carriers in either of the families described here, to allow this assessment. We have therefore undertaken an examination of the sporadic CM cohorts of the Australian Melanoma Genome Project (AMGP, n=284) [46] and The Cancer Genome Atlas (TCGA, n=469) [14]. No deleterious variants in BRCA1 or BRCA2 were identified in the TCGA cohort. In the AMGP cohort, one deleterious germline BRCA1 mutation was identified (c.69_79delGTGTCCCATCT, rs80357696; ClinVar ID: 55676); however, in the tumour sample this allele was lost (with the remaining allele being the wild-type), suggesting that this tumour was not driven by BRCA1 loss.

Despite more robust association having been described between UM and deleterious BRCA1 or BRCA2 [23-28] mutations in breast and/or ovarian family cohorts, we did not observe any deleterious BRCA2 variants in the germline of 47 UM cases from 42 families assessed by WES/WGS. These observations are in agreement with a previous screen of 385 UM patients [36], but not with two other studies [34,35], assessing 62 and 143 UM cases for deleterious BRCA2 mutations, respectively. These latter two studies identified 3 deleterious BRCA2 mutations, and 4 carriers of the Ashkenazi population-specific BRCA2 c.6174delT variant in an Israeli Jewish cohort, respectively. Together, these data suggest that BRCA1 or BRCA2 mutations are not a common cause of UM susceptibility, but are associated with increased risk in some carriers.

A large number of missense variants with an aggregated ExAC [43] population frequency of <0.01, were identified in the CM/UM patients examined in this study (Figure 1A and 1B; Table 1 and Table 2. While 6 of the 33 variants, across nine families, were present in all family members affected with melanoma, all 33 variants were classified as either benign, or of unknown clinical significance, by ClinVar (Supplementary Table 3), which is corroborated by functional analyses compiled by LOVD, when the mutation is present in the database (n=22/33; Supplementary Table 4). The recent investigation of BRCA1 missense variants in the RING (exons 2 – 5) and BRCT (exons 15 – 23) domains by saturation genome editing only contained a single of our variants (p.T1726S), which was classified as benign in this study; it is classified as unknown clinical significance in ClinVar (Supplementary Table 3). In silico prediction of effect of the missense variants on protein function was often not in concordance with these clinical/functional analyses (Supplementary Table 5), demonstrating the importance of this type of further analysis of mutation effect.

Statistical analysis of these missense variants not classified as pathogenic by ClinVar/LOVD (n=33) did not reveal any over-representation in the total CM/UM cohort compared to the control population frequencies for any of the variants observed (Benjamini Hochberg corrected p>0.05; data not shown).

Of potential note, however, is the BRCA1 p.S1465I variant, which has an ExAC population frequency of 0.002 and we observed it in 3 families: 1) in 4/5 CM individuals; 2) in 3/3 CM individuals and 3) in 1/2 CM individuals. In the latter family, one member (who has not developed melanoma by >80 years) was homozygous for this variant and had developed bilateral breast cancer in their 50s. The BRCA1 p.S1465l variant was predicted as damaging by 2 of the 4 in silico prediction algorithms employed (Supplementary Table 5). A number of functional experiments have been performed by different research groups to assess the potential deleterious effects of this variant, as summarised in LOVD (Supplementary Table 4) and reviewed by an expert panel in ClinVar, concluding that this variant is benign, with a neutral effect on BRCA1 function (Supplementary Table 3).

Finally, while several individuals carried multiple variants (in BRCA1: p.D167G and an intronic A/G putative splicing variant (rs80358033); and in BRCA2: (i) p.A75P and p.K3326X; (ii) p.V2728I and p.K3326X; (iii) p.S384F and p.A2951T; Table 1 and Table 2), not all melanoma-affected individuals within a family carry both variants. The contribution of a combination of variants to melanoma susceptibility is therefore likely minimal.

Conclusion

Despite the roles of BRCA1 and BRCA2 in the DNA repair pathway and control of the cell cycle [47] and potential association with UM development via BAP1, we did not find a compelling role for either gene as a major cause of CM or UM susceptibility in this study. The identified missense variants have all been previously described and none are considered to be pathogenic using a variety of assessment methods. No association between BRCA1 or BRCA2 and UM was indicated in the current assessment of 47 UM cases. While we did observe deleterious truncating mutations in BRCA1 in two CM families, we are unable to corroborate a role for these variants in melanoma development, as no tumour samples from carriers were available. Therefore, we sought to examine if deleterious truncating variants in BRCA1 or BRCA2 were present in CM cohorts of the AMGP or TCGA, and if they were associated with a subsequent somatic second hit in the melanoma. In the one instance where a BRCA1 germline frameshift variant was found, no second hit was present in the tumour. Taken together with previous studies, we conclude that BRCA1 and BRCA2 loss of function mutations are not a common risk factor for CM or UM.