BRCA2 is a clinically actionable gene implicated in breast and ovarian cancer predisposition that has become a high priority target for improving the classification of variants of unknown significance (VUS). Among all BRCA2 VUS, those causing partial/leaky splicing defects are the most challenging to classify because the minimal level of full-length (FL) transcripts required for normal function remains to be established. Here, we explored BRCA2 exon 3 (BRCA2e3) as a model for calibrating variant-induced spliceogenicity and estimating thresholds for BRCA2 haploinsufficiency. In silico predictions, minigene splicing assays, patients' RNA analyses, a mouse embryonic stem cell (mESC) complementation assay and retrieval of patient-related information were combined to determine the minimal requirement of FL BRCA2 transcripts. Of 100 BRCA2e3 variants tested in the minigene assay, 64 were found to be spliceogenic, causing mild to severe RNA defects. Splicing defects were also confirmed in patients' RNA when available. Analysis of a neutral leaky variant (c.231T>G) showed that a reduction of approximately 60% of FL BRCA2 transcripts from a mutant allele does not cause any increase in cancer risk. Moreover, data obtained from mESCs suggest that variants causing a decline in FL BRCA2 with approximately 30% of wild-type are not pathogenic, given that mESCs are fully viable and resistant to DNA-damaging agents in those conditions. In contrast, mESCs producing lower relative amounts of FL BRCA2 exhibited either null or hypomorphic phenotypes. Overall, our findings are likely to have broader implications on the interpretation of BRCA2 variants affecting the splicing pattern of other essential exons.
These findings demonstrate that BRCA2 tumor suppressor function tolerates substantial reduction in full-length transcripts, helping to determine the pathogenicity of BRCA2 leaky splicing variants, some of which may not increase cancer risk.
Since the identification of BRCA1 (MIM #113705) and BRCA2 (MIM #600185) as major hereditary breast and ovarian cancer (HBOC) genes, mutation screening has led to the discovery of 20,000 unique germline BRCA1 and BRCA2 variants (1). In female carriers of pathogenic BRCA1 and BRCA2 variants, the cumulative risk by 80 years of age of developing breast cancer was estimated at 72% and 69% and that of developing ovarian cancer at 44% and 17%, respectively (2). Heterozygous BRCA2 mutations also confer moderate risk to pancreatic and prostate cancers (3). Moreover, biallelic BRCA2 mutations are causative of Fanconi anemia (FA) where one allele is often hypomorphic (4). Today, one of the most important bottlenecks in HBOC diagnosis is the high fraction of variants of unknown significance (VUS) detected in the BRCA genes, estimated at approximately 50% of the detected variants (5–9). Carriers of VUS and their relatives cannot benefit from cancer-risk-reducing strategies, such as increased surveillance and prophylactic surgery, or from targeted therapies. VUS are generally typified by missense changes, but also include synonymous substitutions, small in-frame insertions and deletions as well as intronic variants, all of which can potentially affect pre-mRNA splicing (10). A significant number of BRCA VUS has been shown to induce RNA splicing defects (11). Splicing defects that lead to the exclusive production of aberrant transcripts carrying a premature termination codon (PTC) or of in-frame deletions disrupting known functional domain(s) (12) are regarded as pathogenic. In contrast, the biological consequences of variants that lead to partial/leaky splicing defects [i.e., that still produce a certain level of full-length (FL) reference transcripts] remain unknown. In this respect, variants affecting the in-frame alternatively spliced BRCA2 exon 3 (BRCA2e3) are particularly challenging to interpret.
BRCA2e3 encodes an essential bipartite region comprised of (i) a transactivation core that interacts with EMSY, and (ii) a PALB2-interaction domain that allows the recruitment of BRCA2 to the site of double stand breaks (DSB) to mediate homologous recombination (HR; Supplementary Fig. S1A; refs. 13, 14). Recently, it was demonstrated that variants that trigger total skipping of BRCA2e3 (leading to a 249-nucleotide in-frame deletion, Δ3; Supplementary Fig. S1B), such as c.316+5G>C, confer high-risk of developing BRCA2-associated cancers (12). In contrast, the biological impact of variants causing partial BRCA2e3 skipping remains unknown and their contribution to BRCA2 haploinsufficiency and pathogenicity is not yet established. Indeed, the minimal level of full-length transcripts that can provide sufficient BRCA2 function is still unknown. A low proportion of naturally occurring alternatively spliced transcripts lacking BRCA2e3 (Δ3∼1%–3%) has been detected in normal tissues, including blood, mammary glands, and prostate tissues (15–18). Furthermore, despite causing increased levels of Δ3 transcripts (estimated at ∼20% in blood-derived samples), c.68-7T>A was recently found to be a nonpathogenic variant as it was not associated with increased risk of breast cancer (posterior probability of 7.44 × 10−115; ref. 19). Here, we hypothesized that BRCA2e3 variants with partial effects on splicing, though stronger than c.68-7T>A, would confer minimal to severe reduction in BRCA2 function, depending on the severity of exon 3 skipping. These variants could thus be associated with varying degrees of cancer susceptibility.
We therefore investigated the impact of 100 variants on BRCA2e3 splicing by utilizing both in silico tools and functional assays. The biological consequences of seven noncoding variants (intronic or synonymous) that cause gradual increase of BRCA2e3 skipping were then assessed in a mouse embryonic stem cell (mESC)-based complementation assay. In addition, we collected patient-related data for these variants. We conclude that BRCA2 tumor suppressor activity may tolerate substantial reduction in the level of full-length transcripts, a finding that could contribute to improve BRCA variant classification guidelines.
Materials and Methods
To calibrate the potential spliceogenicity of BRCA2e3 variants, we started by focusing on 74 translationally silent/noncoding variants (intronic or synonymous), most of which were bioinformatically predicted to induce BRCA2e3 splicing defects (Supplementary Fig. S2; Supplementary Table S1). These included 61 noncoding variants that were expected to affect BRCA2e3 splicing by either weakening the strength of 3′ and 5′ splice sites (3′ss and 5′ss; n = 21, Supplementary Fig. S3A and S3B) or by disrupting putative splicing regulatory elements (SRE; n = 40, Supplementary Fig. S4). Bioinformatics predictions of variant-induced alterations of 3′ss and 5′ss strength were performed with MaxEntScan (MES) and SpliceSiteFinder-like (SSFL), interrogated by using either Alamut Batch v1.9 or Alamut Visual v2.11 integrated software tools (Interactive Biosoftware, http://www.interactive-biosoftware.com), as well as SPiCE (RRID: SCR_016603, https://sourceforge.net/projects/spicev2-1/) (20). Predictions of variant-induced SRE alterations were obtained with four SRE-predictors, namely: QUEPASA (21, 22) and HEXplorer (23), for which the ΔtESRseq and ΔHZEI scores, respectively, were calculated with the Alamut Batch prototype tool version 1.5.2 (ESRseq; http://www.interactive-biosoftware.com) as well as SPANR (24) and HAL (25), for which ΔΨ scores were retrieved from the corresponding online interfaces (http://tools.genes.toronto.edu and http://splicing.cs.washington.edu/SE, respectively). Variants were selected based on the “at least three” rule, that is, we considered that the most probable spliceogenic variants were those predicted as such by at least three SRE predictors according to thresholds described in ref. 26 and shown in Supplementary Fig. S4. In addition and without prior knowledge of splicing predictions, we also integrated 13 noncoding BRCA2 variants classified as VUS (or with discordant classifications in different databases), which had been detected in probands undergoing genetic testing in diagnostic laboratories from the French GGC-Unicancer network (Supplementary Fig. S2; Supplementary Table S1; Supplementary Note).
After the initial calibration experiments with the 74 noncoding variants, we also performed functional analyses with a set of missense variants (n = 24) identified in probands undergoing genetic counselling. Two pathogenic variants, c.92G>A (p.Trp31*) and c.145G>T (p.Glu49*), which were previously described as spliceogenic in minigene splicing assays (27, 28), were also retained as controls. Therefore, altogether, our collection included a total of 100 BRCA2e3 variants.
Minigene splicing assays
To evaluate the impact on splicing of the 100 BRCA2e3 variants, we performed functional assays based on the comparative analysis of the splicing pattern of wild-type (WT) and mutant reporter minigenes. Minigenes were prepared by using the pCAS2 vector backbone as described previously (26, 29) and explained in Supplementary Materials and Methods.
Analysis of the BRCA2 exon 3 splicing pattern in RNA samples from patients and control individuals
RNA samples from EBV-immortalized lymphoblastoid cell lines (LCL) or PAXgene-stabilized blood of healthy donors and patients were collected in collaboration with the GGC-Unicancer consortium (Supplementary Note). The splicing patterns of BRCA2 transcripts expressed in LCL or PAXgene samples were analyzed by semiquantitative fluorescent RT-PCR (26 and 40 cycles, respectively), as further described in Supplementary Materials and Methods, with primers mapping to BRCA2 exons 2 and 5 (Supplementary Table S2).
Allele-specific expression analysis
Allele-specific expression (ASE) of BRCA2e3-containing transcripts (FL) was measured by performing a quantitative primer extension assay (SNaPshot MultiplexKit, Applied Biosystems), as described previously (29), using specific primers targeting the variants of interest (Supplementary Table S2) and semiquantitative RT-PCR conditions further detailed in Supplementary Materials and Methods.
mESC-based complementation assay
To assess the functional impact of BRCA2e3 variants, we took advantage of a mESC-based assay (30, 31). Mouse ESC-expressing the human BRCA2 gene (hBRCA2) carrying specific variants were generated as described previously (30, 31) with a few modifications as detailed in Supplementary Materials and Methods. After Cre-mediated deletion of the conditional mouse Brca2 allele (mBrca2), we assessed cell survival and sensitivity to DNA-damaging agents, as described previously (30, 31). The splicing patterns of BRCA2e3 were analysed by semiquantitative fluorescent RT-PCR (24 cycles), as described for patient RNA.
Patient and family data
Genetic, clinical, familial, and tumoral data for patients carrying BRCA2e3 variants were collected in collaboration with the French GGC network and within the framework of the COsegregation of VARiants in the BRCA1/2 and PALB2 genes clinical trial (COVAR, Supplementary Note). The criteria for diagnostic variant screening of the BRCA2 gene were applied according to the current French GGC recommendations and the French National Cancer Institute (INCa) guidelines. Informed written consent was obtained from all patients.
Members of the French GGC network and of the COVAR (COsegregation of VARiants in the BRCA1/2 and PALB2 genes) group are listed in the Supplementary Note.
Identification of variants affecting BRCA2e3 splicing by using a minigene splicing assay
To get a glimpse of BRCA2e3 vulnerability to predicted splicing mutations and to calibrate the spliceogenicity of this exon, we started by evaluating the impact on splicing of 74 noncoding variants by performing a minigene assay (Fig. 1A). These variants, which map either to exon 3 or its flanking intronic regions, included 61 noncoding variants bioinformatically predicted to induce exon skipping (19 intronic, 2 synonymous but overlapping exon–intron junctions and 40 exonic synonymous variants; 26/61 being naturally occurring variants described in Human variation databases). We included 13 additional noncoding variants that were selected because they were either classified as VUS in BRCA-Share or had conflicting interpretations in other databases (4 intronic and 9 exonic synonymous; Supplementary Figs. S2–S4; Supplementary Table S1).
The minigene splicing assay, which was based on pCAS2-BRCA2e3-derived constructs (Supplementary Fig. S5), revealed that 52 out of the 74 variants (70%) altered the splicing pattern of BRCA2e3, whereas 22 of 74 remaining variants (30%) showed no major effect on splicing (Fig. 1B; Supplementary Table S1). More precisely, 51 variants induced BRCA2e3 skipping to different extents (6%–100% Δ3, i.e., 94%–0% FL) and 1 variant (c.68-1G>A) resulted in complete deletion of six nucleotides at the beginning of the exon, due to the creation of a new 3′ss (Supplementary Fig. S6). Because it causes a small in-frame deletion, the latter should now be considered a VUS instead of a likely pathogenic variant as currently indicated in ClinVar. We found that in silico analyses were generally useful for predicting variant-induced splicing alterations, but that not all predictions were correct (Supplementary Tables S1, S3–S5; Supplementary Figs. S7A, S7B, S8A–S8E and S9A–S9E). For instance, the naturally occurring c.231T>G synonymous VUS caused 31% Δ3 (69% FL), which is concordant with a previous experimental study (32), but disagrees with the SRE predictions (Supplementary Table S1). In sum, these data suggest that the regulation of BRCA2e3 splicing is very plastic and allowed us to initiate a calibration of the variant-induced spliceogenicity of this exon. Indeed, we observed a spectrum of splicing alterations varying from mild (corresponding to a level of exon inclusion lower than WT, i.e., <99%, but ≥92%, a threshold defined by the effect of the neutral c.68-7T>A variant) to drastic (exon inclusion ≤5%, defined by the impact of the pathogenic c.316+5G>C variant; Fig. 1B). On the basis of this comparative analysis, we infer that the noncoding variants in our dataset that show levels of exon inclusion in the minigene assay equal or superior to c.68-7T>A are neutral, whereas those that lead to exon inclusion levels equal or inferior to c.316+5G>C are pathogenic. Seven VUS causing intermediate levels of exon inclusion (92%>FL>5%) were retained for further analysis because they illustrate the continuum of variant-induced FL transcript loss observed in the minigene assay and could eventually help in defining a Δ3 threshold for pathogenicity (c.165C>T, c.231T>G, c.68-8_68-7delinsAA, c.102A>G, c.316+6T>C, c.316+6T>A and c.316+6T>G, in increasing order of severity; Fig. 1C).
Confirmation of variant-induced spliceogenicity in RNA samples from patients with HBOC
To assess the physiologic relevance of the splicing defects revealed by the minigene assay, we analyzed the splicing pattern of BRCA2e3 in human-derived samples, either LCL or PAXgene-stabilized blood, obtained from control individuals and from four patients carrying BRCA2e3 variants (c.231T>G, c.68-8_68-7delinsAA, c.102A>G, and c.316+6T>C). No biological samples were available for carriers of c.165C>T, c.316+6T>A, and c.316+6T>G. Besides LCL and PAXgene samples from healthy individuals, two additional controls were used in this experiment: (i) a LCL from a patient harboring an unequivocal pathogenic Alu insertion in the middle of exon 3 (c.156_157insAlu) known to cause total exon skipping (33) and LCLs from four carriers of the neutral variant c.68-7T>A, known to lead to mild exon skipping (19). The biallelic splicing patterns of BRCA2e3 in patient biological samples were analyzed by semiquantitative RT-PCR, a common approach used in clinical settings (34), and compared with those generated from equivalent RNA samples of control individuals (Fig. 2A and B; Supplementary Table S1). In LCL and PAXgene samples from healthy controls, we detected the natural alternative splicing pattern of BRCA2e3 with approximately 10% Δ3, whereas carriers of c.68-7T>A displayed approximately 25% Δ3. These splicing patterns are comparable with previous reports of approximately 3% Δ3 in LCL control samples and approximately 13% Δ3 in carriers of c.68-7T>A as determined by qRT-PCR (19). Importantly, heterozygous carriers of c.231T>G, c.68-8_68-7delinsAA, c.102A>G, and c.316+6T>C showed a decrease in the relative amount of FL transcripts due to an increase in Δ3 (75%–38% FL; i.e., 25%–62% Δ3). These relative levels of Δ3 were lower than that observed for c.156_157insAlu (FL = 30%, i.e., Δ3 = 70%), suggesting that these four variants are leaky not only in the minigene assay but also in patients' cells. Nevertheless, assuming a balanced biallelic expression of BRCA2 and taking into account that WT alleles seem to contribute to approximately 10% Δ3, it was expected that the LCL carrying the heterozygous c.156_157insAlu variant would display approximately 55% Δ3, instead of 70% as detected. It is thus possible that our RT-PCR conditions slightly overestimate the level of Δ3 possibly due to a PCR bias favoring the amplification of the smaller Δ3 RT-PCR products relative to FL. A similar impact on exon 3 splicing was observed by Colombo and colleagues but the effect was more exacerbated probably due to a higher number of PCR amplification cycles (19). Sequencing of the FL RT-PCR products of patients carrying the heterozygous c.231T>G variant showed the presence of both WT and mutant FL transcripts, with mutant FL transcripts appearing to be under-represented as compared with WT, further suggesting that this variation cause partial splicing defects (Supplementary Fig. S10). These results obtained from patient-derived RNA samples are in agreement with the minigene data. Indeed, we observed a good correlation (R² = 0.9601, Supplementary Fig. S11A) between the levels of BRCA2e3 skipping evaluated in the monoallelic minigene-based assay and those assessed in RNA samples derived from patients carrying the same variants at the heterozygous state. Moreover, to better evaluate the contribution of WT and mutant alleles to the production of FL transcripts, we took advantage of the quantitative nature of the SNaPshot assay, which allows the measurement of ASE, i.e., the relative contribution of each allele to the production of RT-PCR products containing BRCA2e3 (Fig. 2C). ASE analysis indicated that FL transcripts expressed from the mutant alleles were in fact present in cells of a patient carrying c.102A>G (53% of WT) and c.231T>G (56%, 62%, 65% or 108% of WT, depending on the patient). On the basis of the results of our minigene assays and RT-PCR analysis of patient RNA, we concluded that the observed allelic imbalances in FL transcripts are essentially due to variant-induced BRCA2e3 skipping and that the variants cause partial rather than total splicing defects. Surprisingly, we did not observe an allelic imbalance for one of the carriers of c.231T>G (108% of WT) despite an increase in Δ3 in patient biological material. Importantly, we found that this patient also harbors a pathogenic nonsense variation (BRCA2 c.6515C>A, p.Ser2172*) in trans of c.231T>G. Most likely, the PTC introduced by the nonsense variation leads to nonsense-mediated decay of the transcripts expressed from this allele to a level similar to that of c.231T>G-induced exon skipping. Of note, we observed a correlation between the levels of BRCA2e3 skipping evaluated in the minigene assay and the ASE results obtained with equivalent patient-derived RNA samples (R² = 0.9417; Supplementary Fig. S11B). We surmise from these results that the minigene assay is a good surrogate system to evaluate the spliceogenicity of BRCA2e3 variants.
Functional characterization of BRCA2e3 spliceogenic variants in a mESC-based assay
In an attempt to uncover a RNA-related threshold for BRCA2 haploinsufficiency, we evaluated the consequences of variant-induced BRCA2e3 skipping in BRCA2 function by performing a mESC-based complementation assay (Fig. 3A; Supplementary Fig. S12). Here, we tested seven variants: four intronic (c.68-8_68-7delinsAA, c.316+6T>A, c.316+6T>C, and c.316+6T>G) and three synonymous (c.102A>G, c.165C>T and c.231T>G). These variants were selected on the basis of four main reasons. First, they caused a gradual increase in Δ3 in the minigene assay. Second, they are all translationally silent, which avoids any potential confounding effect produced by coding changes. Third, they are clinically interesting given that all except one (c.316+6T>A) were identified in patients suspected of HBOC. And fourth, their pathogenicity is unknown. We also included c.68-7T>A and c.316+5G>C as controls because these spliceogenic variants have been unequivocally classified as neutral and pathogenic, respectively (12, 19).
We first tested the ability of the variants to rescue the cell lethality conferred by loss of BRCA2 after loxP/Cre-mediated deletion of the endogenous conditional mBrca2 allele (Fig. 3A; Table 1; Supplementary Fig. S13). The two loxP sites of the conditional mBrca2 are flanked by two halves of human HPRT minigene, which allow selection of recombinant clones in the presence of HAT. We observed that mESCs expressing the c.316+5G>C pathogenic variation were unable to form HAT-resistant colonies, indicating that BRCA2 function was severely impaired, a result that is in agreement with the deleterious nature of this variant (12). In contrast and as expected, the neutral variation c.68-7T>A (19) was able to fully complement the loss of endogenous mBrca2. Interestingly, the complementation phenotype of five of the seven VUS, namely c.165C>T, c.231T>G, c.102A>G, c.68-8_68-7delinsAA and c.316+6T>C, resembled that of the neutral variant. However, in the case of c.316+6T>A and c.316+6T>G (variants that had the most severe splicing defects in the minigene assay among the seven VUS of interest), we observed fewer HAT-resistant colonies as compared with WT-expressing cells, indicating that BRCA2 function was compromised in these cells, resulting in incomplete/poor complementation (Supplementary Fig. S13).
|.||.||.||.||Splicing (FL%)c .||.|
|Nucleotide .||Amino acid .||Complementationa .||Sensitivityb .||Detected .||% WT .||Classificationd .|
|.||.||.||.||Splicing (FL%)c .||.|
|Nucleotide .||Amino acid .||Complementationa .||Sensitivityb .||Detected .||% WT .||Classificationd .|
Abbreviations: n/a, not applicable; nd, not determined; p?, variant of unknown protein consequences (non-coding but it may affect RNA splicing) as per HGVS nomenclature.
aComplementation phenotypes of BRCA2e3 variants based on viability of Brca2ko/ko mESCs expressing either WT or mutant BRCA2 after deletion of the conditional allele (see also Fig. 3; Supplementary Fig. S13).
bSensitivity of HAT-resistant mESCs (expressing either WT or mutant hBRCA2) to six DNA-damaging agents (olaparib, methyl methanesulfonate, mitomycin C, cisplatin, camptothecin, and ionizing radiation). Survival was measured by XTT assay and then compared with that of mESCs expressing wild-type hBRCA2 (see also Supplementary Fig. S14).
cRelative quantification of splicing events in HAT-resistant mESCs expressing either WT or mutant hBRCA2 was evaluated by semiquantitative fluorescent RT-PCR, followed by capillary electrophoresis. Results represent the mean of three independent experiments. Values relative to %WT were extrapolated from RT-PCR results displayed in the previous column (FL%) and assuming a similar BRCA2 expression in the different mESCs clones.
dSuggested nucleotide variant classification based on the results of the mESC complementation assay.
Because BRCA2 has an important role in the repair of DSBs, loss of its function renders cells vulnerable to compounds that introduce toxic DNA lesions (35). Therefore, we next tested the sensitivity of the HAT-resistant mESC to six DNA-damaging agents (cisplatin, mitomycin C, methyl methanesulfonate, olaparib, camptothecin, and γ-irradiation) by cell survival measurements (Table 1; Supplementary Fig. S14). As expected, c.68-7T>A-expressing cells exhibited no difference in sensitivity to various DNA-damaging agents as compared with WT-expressing cells. Similarly, mESC carrying either c.165C>T, c.231T>G, c.102A>G, c.68-8_68-7delinsAA, or c.316+6T>C did not exhibit hypersensitivity to any of the DNA-damaging agents, strongly supporting that they are fully functional and are likely to be neutral variants. In contrast, c.316+6T>A and c.316+6T>G are associated with a severe hypersensitivity to various DNA-damaging agents as compared with WT or to c.68-7T>A-expressing cells, strongly indicating an impairment in BRCA2 function. We surmise that a reduction in FL transcripts down to 35% (as measured for the borderline c.316+6T>C variant in the minigene assay) is sufficient for BRCA2 function, whereas a level equal to or lower than 26% FL (as measured for c.316+6T>A in the minigene assay) leads to BRCA2 haploinsufficiency in mESCs. Moreover, a total loss-of-function was observed in mESCs when FL≤5% (as measured for c.316+5G>C in the minigene assay).
To validate the physiologic relevance of the mESC system for the evaluation of variant-induced BRCA2e3 splicing defects, we analyzed the splicing patterns of the different BRCA2 variants in viable mESCs by semiquantitative RT-PCR. The WT BRCA2 gene expressed in this system reproduced the alternative splicing of BRCA2e3 (Δ3 = 10%, Fig. 3B) typically detected in normal human cells. Importantly, we confirmed that the five VUS that were fully functional in the mESC assay had increased levels of Δ3 representing a decrease in FL transcripts down to 27% of the WT BRCA2 FL transcripts as extrapolated from the RT-PCR results. In contrast, c.316+6T>A and c.316+6T>G, which showed poor mESC-complementation, had even lower relative levels of FL transcripts (17% and 12%, respectively). These data suggests that the upper threshold for BRCA2 haploinsufficiency in the mESCs lies between approximately 17% and approximately 27% FL and that for total loss-of-function the cutoff lies below 12% FL. The results of the mESC-derived RNA analysis were fully consistent with those obtained in the minigene assay not only in terms of whether or not a variant caused aberrant splicing but also in terms of the relative severity of the splicing defects (R² = 0.9974; Supplementary Fig. S15). Moreover, the results obtained in mESCs agree with those obtained with patient-derived samples, which highlights the physiologic pertinence of the mESC model to evaluate the consequences on splicing of BRCA2e3 variants. A biological classification of the selected variants based on the mESC complementation results is shown on Table 1.
Association of leaky BRCA2 exon 3 variants with cancer risk
To gain insight into the clinical significance of leaky BRCA2e3 spliceogenic variants, we collected patient information for the seven VUS analyzed in the mESC assay (Supplementary Table S6). Only one variant (c.231G>T) could be classified as neutral according to ACMG (36) and ENIGMA (37) guidelines given its: (i) high frequency in the African population (0.66%), (ii) the description of an homozygous individual in the general population; and (iii) the cooccurrence in trans with pathogenic BRCA2 variants in HBOC patients without a FA phenotype. These data confirm that a decrease in FL transcripts from one BRCA2 allele at least down to approximately 60% of WT (as determined for c.231G>T in the ASE analysis) is not associated with HBOC. Multifactorial likelihood analyses based on clinical, familial, and tumoral data from patients carrying four of the seven VUS (c.68-8_68-7delinsAA, c.102A>G, c.316G>A, and c.316+6T>C) were inconclusive, in part, because few families were accessible for analysis and no cosegregation information was available.
Functional impact of BRCA2 missense VUS
Finally, we extended our minigene splicing analysis to 26 BRCA2e3 coding variants (Fig. 4A) including all missense changes reported in the BRCA-Share database (n = 23), as well as one FA-associated missense variant (c.316G>A) and two previously described spliceogenic nonsense variants (c.92G>A and c.145G>T) (27, 28), the latter used as additional BRCA2e3 spliceogenic controls. Of note, among the 24 missense variants, 21 were described as VUS (Supplementary Table S1). The minigene assay revealed that 12 out of the 26 variations tested (46%) altered the splicing pattern of BRCA2e3 relative to WT (Fig. 4B; Supplementary Table S1). More precisely, 11 variants increased BRCA2e3 skipping, including c.92G>A and c.145G>T as expected, and one variant (c.100G>A) caused two concomitant partial splicing defects, that is, skipping of BRCA2e3 and a 45-nucleotide in-frame deletion at the beginning of the exon caused by the creation of a new 3′ss (Supplementary Fig. S16). Given their positions within the exon, we suspect that c.316G>C and c.316G>A induce exon skipping by directly decreasing the strength of the 5′ss (Supplementary Fig. S7), whereas the remaining spliceogenic missense variants likely modify exonic SREs. Importantly, we confirmed that four of the variants (c.92G>A, c.145G>T, c.316G>A, and c.316G>C) are associated with a major relative increase of Δ3 in patients' RNA as compared with healthy controls (Fig. 4C). Moreover, ASE analysis revealed a reduction in FL transcripts expressed from the mutant alleles in these biological samples (50%–24% of WT, Fig. 4C and D). In contrast, the 14 remaining variants showed no effect on splicing in the minigene assay (Fig. 4B), including c.223G>C, a variant that is associated, in patient RNA, with a splicing pattern similar to that observed in control samples (Fig. 4C) and to an absence of in vivo allelic imbalance (Fig. 4D), which is consistent with the minigene results.
Our findings reiterate the importance of the minigene assay in assessing the impact on splicing of BRCA2e3 variants. Moreover, they pinpoint c.316G>A (p.Gly106Arg) as potentially leading to BRCA2 haploinsufficiency because although the splicing efficiency of this variant (both in the minigene assay and in patient RNA) is similar to that of the borderline c.316+6T>C noncoding variant, the c.316G>A transition: (i) produces a Gly>Arg missense change at a highly conserved amino acid position (Supplementary Fig. S17) in the residual BRCA2 FL transcripts, and (ii) has been identified in trans of BRCA2 c.2806_2809del [p.(Ala938Profs*21), paternal allele] in a FA patient, which is suggestive of a disease-causing hypomorphic variant.
To test this hypothesis, we then determined the functional consequences of c.316G>A (p.Gly106Arg) in the mESC assay. As shown in Table 1, c.316G>A lead to a drastic reduction of the number of HAT-resistant mESC colonies and a severe hypersensitivity to various DNA-damaging agents as compared with WT and to c.68-7T>A, strongly indicating an impairment of BRCA2 function. Importantly, the complementation phenotype and degree of sensitivity to DNA-damaging agents were similar to those observed for c.316+6T>G and c.316+6T>A (Table 1; Supplementary Figs. S13 and S14) revealing that indeed c.316G>A (p.Gly106Arg) causes BRCA2 haploinsufficiency but not total loss-of-function. Our results suggest that c.316G>A (Gly106Arg) is a hypomorphic allele that leads to BRCA2 haploinsufficiency because of a negative impact both on BRCA2e3 splicing and on FL protein function.
Determining the pathogenicity of leaky BRCA2 variants remains a major challenge in medical cancer genetics. To our knowledge, only one leaky BRCA2 variant (c.68-7T>A) has been thoroughly assessed for its implication in HBOC (19). This variant has been classified as neutral by multifactorial analysis while all remaining BRCA2 leaky variants linger as VUS until further evidence is collected (11, 19, 32). By using BRCA2e3 as a model system and complementary functional assays, we provide the first calibration of BRCA2e3 spliceogenicity and an estimation of BRCA2 mRNA-based haploinsufficiency, both of which contribute to the interpretation of leaky BRCA2 variants. In addition, our study identified a plethora of new spliceogenic BRCA2e3 variants.
To date, only 17 SNVs in/near BRCA2e3 (10 exonic and 7 intronic) have been reported as causing aberrant splicing, all shown to increase Δ3 (11, 12, 19, 27, 28, 32, 35, 38, 39). Our minigene results confirmed 15 of those initial findings (only 2 of the known spliceogenic variants being absent from our study) and uncovered 49 new splicing mutations (37 exonic, 10 intronic, and 2 at the exon–intron border) bringing the number of BRCA2e3 spliceogenic variants to a total of 66, most affecting potential SREs. Importantly, the vast majority of the detected splicing defects were leaky (54 of 64 spliceogenic variants), 45 being less drastic than the splicing anomaly caused by c.316+5G>C, but stronger than that observed for the neutral variant c.68-7T>A, which implied that they remained as of unknown significance. These observations suggest that variant-induced leaky splicing defects are prevalent in BRCA2e3 and that additional analyses are warranted for accurate assessment of their pathogenicity. Still, among the 49 newly identified spliceogenic variants, 4 caused major loss of BRCA2e3 in the minigene assay similar to what was observed for c.316+5G>C. These variants (c.316+1G>A, c.316+1G>C, c.316+3A>C, and c.316+4_316+6delinsCGA) can thus be classified as pathogenic without further evidence (12). Moreover, we could also immediately classify 34 noncoding variants as neutral because they either had no impact on splicing in the minigene assay or induced a minor decrease in FL transcripts (equal or milder than c.68-7T>A).
To evaluate the biological significance of leaky splicing defects caused by BRCA2e3 VUS, we next used a mESC-based assay known to reliably predict the pathogenicity of BRCA2 variants (30, 31, 35, 40). We found that this assay recapitulates the normal alternative splicing of BRCA2e3 as well as variant-induced partial splicing defects. Indeed, although BRCA2 Δ3 levels were somewhat variable in our complementary approaches probably due to different genomic environments, tissue-specific alternative splicing patterns and/or other experimental specificities, we demonstrated a high concordance in splicing phenotypes between the mESC assay and both minigene and patient RNA analysis. Moreover, data on mESC viability and sensitivity to DNA-damaging agents were concordant with the clinical classifications of c.68-7T>A and c.316+5G>C. By testing selected spliceogenic BRCA2e3 VUS in the mESC assay, we were able to redefine cut-off values, extrapolated from BRCA2e3 splicing efficiencies, meant for improving variant interpretation (Fig. 5A and B). These thresholds allowed us to classify the 100 variants in our dataset as follows: 59 as neutral, 11 as pathogenic, and 30 as VUS (Supplementary Table S1). Our data suggests that noncoding BRCA2e3 variants producing at least 27% FL transcripts in the mESC assay as compared with WT (or 35% in the minigene assay) can be considered neutral as they are expected to fully complement the loss of BRCA2 in mESCs as seen for c.316+6T>C. A more conservative approach would be to take into account, as threshold for neutrality, the fraction of variant-expressed FL transcripts observed for c.231T>G (a minimum of 68%, 70%, or ∼60% FL in mESC, minigene or patients' RNA assays, respectively, as compared with WT) given that this is clearly a nonpathogenic variant. Indeed, our results indicate, for the first time, that BRCA2 tumor suppressor activity tolerates a substantial reduction in FL expression from one allele independently of the synthesis of alternatively spliced functional transcripts. Interestingly, certain BRCA1 and BRCA2 variants responsible for drastic FL losses were recently reported as not being necessarily associated with high cancer risk or total loss-of-function (41, 42). However, in these cases, alternatively spliced transcripts leading to the production of potentially or partially functional isoforms (BRCA1 Δ9, 10, or BRCA2 Δ12, respectively) were generated, suggesting a rescue mechanism underlying the preserved BRCA function.
A window of uncertainty prevails for variants showing hypomorphic phenotypes in the mESC assay such as c.316+6T>A and c.316+6T>G that reduce FL transcripts to approximate;y15% WT in mESC (∼25% WT in the minigene assay). It is possible that these variants confer moderate risks of breast/ovarian cancer as shown for BRCA1 c.5096G>A/p.Arg1699Gln and BRCA2 c.9104A>C/p.Tyr3035Ser (43–45) or are associated with lower penetrance (45). Noncoding spliceogenic variants with incomplete penetrance and/or variable degrees of disease expressivity have already been described in other cancer predisposition genes such as RB1 (46), PMS2 (47), and VHL (48). Possibly such variability also exists within the spectra of BRCA2 variants.
Further studies will be essential for illuminating genotype–phenotype correlations of spliceogenic BRCA2 VUS identified as leaky in this and other studies. Given the rarity of most of these variants, informative studies will more likely stem from international collaborative efforts, such as those conducted by the ENIGMA consortium (49), allowing the collection of data on large number of patients and their families as well as to perform case–control genotyping analysis to accurately estimate cancer risks conferred by individual variants. Finally, because only minimal amounts of FL transcripts seem to be required for sufficient BRCA2 function, our study may pave the way for the development of new cancer prevention strategies based on RNA splicing correction in patients carrying BRCA2 splicing variants, according to the paradigm of spinal muscular atrophy (50).
Disclosure of Potential Conflicts of Interest
H. Tubeuf reports personal fees and nonfinancial support from Interactive Biosoftware [CIFRE PhD fellowship (#2015/0335)], grants from Groupement des Entreprises Françaises dans la Lutte contre le Cancer (Gefluc; Translational research grant), Fédération Hospitalo-Universitaire (FHU) Normandy Centre for Genomic and Personalized Medicine (NGP; translational research grant), European Union and Region Normandie (translational research grant), French National Cancer Institute and the Direction Générale de l'Offre des Soins (INCa/DGOS; translational research grant), European Regional Development Fund (ERDF; translational research grant); personal fees from EMBO (short-term fellowship #3436), Cancéropôle Nord-Ouest (short-term fellowship), and OpenHealth Institute during the conduct of the study. G. Castelain reports grants from French National Cancer Institute and the Direction Générale de l'Offre des Soins (INCa/DGOS), Groupement des Entreprises Françaises dans la Lutte contre le Cancer (Gefluc), and other from Fédération Hospitalo-Universitaire (FHU) Normandy Centre for Genomic and Personalized Medicine (NGP; financial support to Inserm U1245) and other from European Union and Region Normandie (financial support to Inserm U1245) during the conduct of the study; and H. Tubeuf was funded by a CIFRE PhD fellowship (#2015/0335) from the French Association Nationale de la Recherche et de la Technologie (ANRT) in the context of a public-private partnership between INSERM and Interactive Biosoftware. L. Meulemans reports grants from French National Cancer Institute and the Direction Générale de l'Offre des Soins (INCa/DGOS), grants from Groupement des Entreprises Françaises dans la Lutte contre le Cancer (Gefluc), other from Fédération Hospitalo-Universitaire (FHU) Normandy Centre for Genomic and Personalized Medicine (NGP; financial support to Inserm U1245), and other from European Union and Region Normandie (financial support to Inserm U1245) during the conduct of the study; and H. Tubeuf was funded by a CIFRE PhD fellowship (#2015/0335) from the French Association Nationale de la Recherche et de la Technologie (ANRT) in the context of a public-private partnership between INSERM and Interactive Biosoftware. P. Pujol reports grants and personal fees from Astrazeneca, Pfizer; personal fees from MSD and Exact Sciences outside the submitted work. F. Vaz reports other from Astrazeneca (collaborated with BRCA testing for observational study) outside the submitted work. C. Colas reports personal fees from AstraZeneca outside the submitted work. P. Gaildrat reports grants from French National Cancer Institute and the Direction Générale de l'Offre des Soins (INCa/DGOS), Groupement des Entreprises Françaises dans la Lutte contre le Cancer (Gefluc), other from Fédération Hospitalo-Universitaire (FHU) Normandy Centre for Genomic and Personalized Medicine (NGP; financial support to Inserm U1245), and other from European Union and Region Normandie (financial support to Inserm U1245) during the conduct of the study; and H. Tubeuf was funded by a CIFRE PhD fellowship (#2015/0335) from the French Association Nationale de la Recherche et de la Technologie (ANRT) in the context of a public-private partnership between INSERM and Interactive Biosoftware. A. Martins reports grants from French National Cancer Institute and the Direction Générale de l'Offre des Soins (INCa/DGOS), Groupement des Entreprises Françaises dans la Lutte contre le Cancer (Gefluc), other from Fédération Hospitalo-Universitaire (FHU) Normandy Centre for Genomic and Personalized Medicine (NGP; financial support to Inserm U1245), and other from European Union and Region Normandie (financial support to Inserm U1245) during the conduct of the study; and H. Tubeuf was funded by a CIFRE PhD fellowship (#2015/0335) from the French Association Nationale de la Recherche et de la Technologie (ANRT) in the context of a public-private partnership between INSERM and Interactive Biosoftware. No potential conflicts of interest were disclosed by the other authors.
H. Tubeuf: Conceptualization, software, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. S.M. Caputo: Resources, funding acquisition, investigation, writing-review and editing. T. Sullivan: Investigation. J. Rondeaux: Investigation. S. Krieger: Resources, writing-review and editing. V. Caux-Moncoutier: Resources. J. Hauchard: Investigation. G. Castelain: Investigation. A. Fiévet: Resources. L. Meulemans: Investigation. F. Révillion: Resources. M. Léoné: Resources. N. Boutry-Kryza: Resources. C. Delnatte: Resources. M. Guillaud-Bataille: Resources. L. Cleveland: Investigation. S. Reid: Investigation. E. Southon: Investigation. O. Soukarieh: Investigation. A. Drouet: Investigation. D. Di Giacomo: Investigation. M. Vezain: Investigation. F. Bonnet-Dorion: Resources. V. Bourdon: Resources. H. Larbre: Resources. D. Muller: Resources. P. Pujol: Resources. F. Vaz: Resources. S. Audebert-Bellanger: Resources. C. Colas: Resources. L. Venat-Bouvet: Resources. A.R. Solano: Resources. D. Stoppa-Lyonnet: Resources, funding acquisition. C. Houdayer: Resources, writing-review and editing. T. Frebourg: Resources, funding acquisition, writing-review and editing. P. Gaildrat: Conceptualization, supervision, funding acquisition, validation, investigation, writing-original draft, writing-review and editing. S.K. Sharan: Resources, supervision, funding acquisition, validation, investigation, writing-review and editing. A. Martins: Conceptualization, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing, study design and direction.
This work was funded by a translational research grant from the French National Cancer Institute and the Direction Générale de l'Offre des Soins (INCa/DGOS, AAP/CFB/CI), by the OpenHealth Institute, the Groupement des Entreprises Françaises dans la Lutte contre le Cancer (Gefluc), the Fédération Hospitalo-Universitaire (FHU) Normandy Centre for Genomic and Personalized Medicine (NGP), the European Union and Region Normandie as well as by the Intramural Research Program from the Center for Cancer Research of the NCI, NIH. Europe gets involved in Normandie with European Regional Development Fund (ERDF). H. Tubeuf was funded by a CIFRE PhD fellowship (#2015/0335) from the French Association Nationale de la Recherche et de la Technologie (ANRT) in the context of a public–private partnership between INSERM and Interactive Biosoftware and by two short-term fellowship from EMBO (#3436) and Cancéropôle Nord-Ouest. J. Hauchard and G. Castelain were sponsored by the French INCa. The COVAR study is supported by the French Ligue contre le cancer and INCA BCB (2013-1-BCB-01-ICH-1) and by the GENETICANCER and « Cercle de l'Olympe ». We thank Sabine Tourneur (Inserm U1245, Rouen, France) for technical assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.