The BRCA2 c.68‐7T > A variant is not pathogenic: A model for clinical calibration of spliceogenicity

Abstract Although the spliceogenic nature of the BRCA2 c.68‐7T > A variant has been demonstrated, its association with cancer risk remains controversial. In this study, we accurately quantified by real‐time PCR and digital PCR (dPCR), the BRCA2 isoforms retaining or missing exon 3. In addition, the combined odds ratio for causality of the variant was estimated using genetic and clinical data, and its associated cancer risk was estimated by case‐control analysis in 83,636 individuals. Co‐occurrence in trans with pathogenic BRCA2 variants was assessed in 5,382 families. Exon 3 exclusion rate was 4.5‐fold higher in variant carriers (13%) than controls (3%), indicating an exclusion rate for the c.68‐7T > A allele of approximately 20%. The posterior probability of pathogenicity was 7.44 × 10−115. There was neither evidence for increased risk of breast cancer (OR 1.03; 95% CI 0.86–1.24) nor for a deleterious effect of the variant when co‐occurring with pathogenic variants. Our data provide for the first time robust evidence of the nonpathogenicity of the BRCA2 c.68‐7T > A. Genetic and quantitative transcript analyses together inform the threshold for the ratio between functional and altered BRCA2 isoforms compatible with normal cell function. These findings might be exploited to assess the relevance for cancer risk of other BRCA2 spliceogenic variants.

(OR 1.03; 95% CI 0.86-1.24) nor for a deleterious effect of the variant when co-occurring with pathogenic variants. Our data provide for the first time robust evidence of the nonpathogenicity of the BRCA2 c.68-7T > A. Genetic and quantitative transcript analyses together inform the threshold for the ratio between functional and altered BRCA2 isoforms compatible with normal cell function. These findings might be exploited to assess the relevance for cancer risk of other BRCA2 spliceogenic variants. Carriers of germline pathogenic variants in these genes are at high risk of developing breast and ovarian cancers, and BRCA1/2 gene testing has become a widely used procedure in the clinical management of families suspected of hereditary susceptibility to these malignancies. The individuals within these families, identified as at-risk based on their genetic profile, may benefit from risk-reduction options. However, the usefulness of genetic testing relies on the ability to ascertain the pathogenic nature of the identified genetic variants, which is not necessarily straightforward for small in-frame deletions and insertions, variants in regulatory sequences, missense, synonymous and intronic changes, and variants introducing premature proteintruncating codons at the 3 ′ end of the coding sequence.
The Evidence-based Network for the Interpretation of Germline Mutant Alleles (ENIGMA) has developed and documented criteria aimed at determining the clinical significance of sequence variants in BRCA genes (https://www.enigmaconsortium.org). The classification, based on a five-class system (Plon et al., 2008), is intended to differentiate high risk variants (risk equivalent to that of protein-truncating pathogenic variants), including pathogenic and likely pathogenic variants (class 5 and 4, respectively), from variants with low or no risk, including not pathogenic and likely not pathogenic variants (class 1 and 2, respectively). Variants for which clinical significance is unclear are placed in class 3 and are referred to as variants of uncertain significance (VUSs).
One controversial variant in BRCA2 is c.68-7T > A, which lies upstream of the acceptor splice site of exon 3. This variant (rs81002830) has been reported in several populations worldwide with an allelic frequency ranging from 0.02% in East Asians to 0.5% in non-Finnish Europeans (Lek et al., 2016). Several authors have reported c.68-7T > A being spliceogenic, that is, able to alter normal premRNA splicing. In particular, using semiquantitative approaches, it has been documented that the variant leads to an increase of the naturally occurring transcripts lacking exon 3 (∆3) ( Thery et al., 2011;Vreeswijk et al., 2009). A competitive quantitative PCR (qPCR) analysis estimated that the proportion of the ∆3 transcript compared to full length was approximately 25% in variant samples versus 4% in normal samples (Muller et al., 2011). More recently, segregation analyses in two families indicated that the variant did not segregate in the affected branches (Santos et al., 2014).
Although a few of the above studies tentatively classified the variant as benign or likely benign, they do not provide robust genetic evidence to justify this conclusion. Conversely, a recent article asserted that the variant was associated with breast cancer, based on a relatively limited case control association study in the Norwegian population (Møller & Hovig, 2017 In the present study, we combined genetic approaches, including a large multicentre case-control study and segregation analysis in a sizable number of families, with qualitative and quantitative analyses of the transcripts, and Mitomycin C growth inhibition test. Our findings provide a robust classification of the BRCA2 c.68-7T > A variant with respect to its effect on cancer risk, and add evidence that splicing pattern alterations do not necessarily lead to pathogenicity.

Nomenclature
The nucleotide numbering was based on the reference BRCA2 complementary deoxyribonucleic acid (cDNA) sequence NM_000059.3. For the purposes of the study, we defined as ▼3 all BRCA2 isoforms retaining exon 3 and as ∆3 all BRCA2 isoforms missing exon 3, irrespective of additional alternative splicing events.

Cell lines
Epstein-Barr virus (EBV)-immortalized human lymphoblastoid cell lines (LCLs) were obtained as previously described (Colombo et al., 2013). In this analysis 18 LCLs were considered, including six LCLs obtained from women carrying the BRCA2 c.68-7T > A variant and 12 LCLs obtained from healthy female blood donors, recruited at the Isti- (c.1380dupA). Only BRCA2 exon 3 was sequenced in the LCLs from normal controls and no pathogenic variants or VUS were observed.

Cytoplasmic RNA isolation and first strand cDNA synthesis
Cytoplasmic RNA was isolated from fresh LCLs using the Cytoplas-

Capillary electrophoresis analysis
Multiplex fluorescently-labeled PCRs were performed with primers located upstream and downstream of exon 3, to simultaneously amplify both ▼3 and ∆3 transcripts, followed by CE analysis. A beta-2microglobulin (B2M; MIM# 109700) cDNA fragment of 377 bp was coamplified to normalize CE peaks and allow comparison between cases and controls. The sequences of the primers are listed in Supp. Table S1. PCR amplifications were performed in 20 l reaction volume containing 2 l of cDNA solution under end-point conditions. Cycling conditions were as follows: 95 • C for 7 min, followed by 35 cycles at 95 • C for 30 ′′ , 58 • C for 30 ′′ and 72 • C for 30 ′′ . A final 7 min elongation step was performed at 72 • C. The fluorescent amplification products were run on an ABI 3130 Genetic Analyzer (Applied Biosystems). GeneScan TM 500 ROX TM dye size standard was used as internal size-standard and size calling was performed with GeneMapper software v4.0 (Applied Biosystems).

Assessment of allelic expression of ▼3 and ∆3 transcripts
The allelic origin of the ▼3 and ∆3 transcripts were ascertained by amplification and sequencing of the region containing the common c.-26G > A variant (rs1799943) in the 5 ′ -UTR of BRCA2. PCR reactions were performed as described above. The forward primer was designed to anneal to a region upstream of c.-26G > A and the reverse primers to sequences in exon 3 and across the exon2-exon4 junction, specific of the ▼3 and ∆3 transcripts, respectively (Supp . Table S1).
Sequencing conditions were as previously described (Colombo et al., 2013).

Quantitative PCR analysis
Specific quantitative assays were designed to capture the expression levels of the ▼3 and Δ3 transcripts. The primer sets (Supp . Table S1) were validated with end-point PCR reactions, and the specificity of the amplification products were confirmed by sequencing.
The qPCR analysis were performed on the Eco real-time PCR system (Illumina) using SYBR R Green I dye chemistry (KAPA SYBR R FAST qPCR Kit, Kapa Biosystems). All reactions were carried out in a final volume of 10 l containing 1 l of cDNA and 200 nM of GUSB and ▼3 transcript specific primers, and 300 nM of Δ3 transcript specific primers. The efficiency of qPCR assays was evaluated based on a relative standard curve, using threefold serial dilutions starting from pooled control cDNAs in triplicate. The thermal profile was the same for all assays (95 • C for 3 min, followed by 40 cycles of 95 • C for 3 sec and 62 • C for 20 sec). The melting curve analysis was performed according to default conditions (95 • C for 15 sec, 55 • C for 15 sec and 95 • C for 15 sec). All samples from both cases and controls were individually analyzed in triplicate, and the corresponding average values were considered. No template controls and no-RT controls were included in the analysis. The data, obtained in the form of quantification cycle (Cq), were normalized using the beta-glucuronidase gene (GUSB) (de Brouwer, van Bokhoven, & Kremer, 2006). The obtained values were used to compute, in both normal and mutated samples, BRCA2 exon 3 exclusion rate, that is, the percentage of BRCA2 mRNA isoforms missing exon 3 over the total amount of BRCA2 transcripts, as follows: The distribution of transcript levels in control and mutant LCLs was calculated by normalization to that of pooled control cDNAs (reference sample) using the ∆ΔCq method (Livak & Schmittgen, 2001).
Statistical analysis was performed using GraphPad Prism software (version 5.02). The significance of the results was established using the F test.
To detect ▼3 transcripts, we used a 2 ′ -chloro-7 ′ phenyl-1,4-dichloro-6-carboxy-fluorescein labeled (VIC-labeled) predesigned TaqMan assay (Applied Biosystems, Hs00609076) specific for the exon 3-4 junction (5-AATTAGACTTAG-GAAGGAATGTTCC-3 ′ ). All relative quantification experiments were performed combining Δ3 and ▼3 assays in individual chips. dPCR chips were analyzed in the QuantStudio 3D Analysis Suit Cloud software v2.0 (Applied Biosystem, Foster City, CA), defining FAM as target. Default settings were used in all cases. After reviewing automatic assessment of the chip quality by the software, only green flag chips (data meet all quality thresholds, review of the analysis result not required) and yellow flag chips (data meet all quality thresholds, but manual inspection is recommended) were considered for further analyses. We used the target/total ratio, FAM/(FAM+VIC), calculated by the software as a proxy for BRCA2 exon 3 exclusion rate. Different amounts of each sample were individually tested in 20K chips, but only data from the chip with the highest precision (according to the analysis software) was included in the expression analysis shown in Figure 3.

Genotyping and sample sets
Direct genotyping of BRCA2 c.68-7T > A was conducted as part of the Collaborative Oncological Gene-environment Study (COGS) detailed elsewhere (Michailidou et al., 2013). This study included genotype results from breast cancer cases and controls participat- were available from at least one relative. All study participants had been previously enrolled into national or regional studies under ethically approved protocols.

Statistical methods
The association of the BRCA2 c.68-7T > A variant with breast cancer risk was evaluated in BCAC using logistic regression models, as previously detailed .
In addition, multifactorial likelihood analysis was conducted as detailed in the Supp. Text. In brief, odds for causality were calculated based on carrier frequency and ages at diagnosis/interview in cases and controls, as previously described (Goldgar et al., 2004).
Pathology likelihood ratios (LRs) were applied as indicated in Spurdle et al., (2014). The segregation scores, pathology LRs and casecontrol LRs are mutually independent and were combined to derive a combined odds for causality as described previously (Goldgar et al., 2004;Goldgar et al., 2008). Prior probability of pathogenicity was

Mitomycin C (MMC) growth inhibition test and statistical analyses
A total of 3 × 10 6 cells/ml were seeded in triplicate in 25 ml flasks and grown for 72 hr in the absence or in the presence of MMC (Sigma-Aldrich) at a final concentration of 170 ng/ml. Percentage of viable cells was determined using trypan blue dye exclusion assay, following the manufacture's instruction (Sigma-Aldrich). Statistical differences in cell viability after exposure to MMC compared to controls were determined by two-tailed Student t-test using GraphPad Prism software. and three cases). Each transcript was selectively amplified in separate reactions and sequenced. Even considering that transcript quantification by sequencing analysis is not entirely accurate, it was apparent that, while in normal samples the levels of the Δ3 transcripts originating from the two alleles were comparable, in carriers the contribution of the variant allele was higher than that of the wild-type allele. In addition, it was also observed that in carriers the variant allele retained the ability to synthesize the ▼3 transcripts. A representative example is shown in Figure 1B.
Subsequently, an independent dPCR-based quantitative analysis was performed to measure BRCA2 exon 3 exclusion rates directly in the same sample set. After excluding two outliers, we found that the exclusion rate in LCLs carrying the variant allele (15.5%; range 14.4%-17.2%) was 4.2-fold higher than in normal LCLs (3.7%; range 3.0%-4.5%; p < 10 −4 ) ( Fig. 3 and Supp. Figure S1). The odds for causality based on carrier frequency and ages at diagnosis/interview in these cases and controls was 9.44 × 10 −93 . There was also strong evidence against causality from segregation analysis (6.39 × 10 −9 ) and breast tumor pathology (2.40 × 10 −14 ). Considering all data together, and assigning prior probability of 0.34 based on splicing prediction, the posterior probability of pathogenicity was calculated to be 7.44 × 10 −115 (see Supp. Text for further details).

Co-occurrence of the c.68-7T > A with pathogenic variants
Overall 15 Table S2). None of these cases was included in the RNA analyses described above.

Carriers of bi-allelic BRCA2 inactivating variants are affected with
Fanconi Anemia (FA), complementation group D1. FA is characterized by congenital defects, including anatomical abnormalities, congenital disabilities and increased risk of cancer, most often acute myelogenous leukemia (Auerbach, 2009). In addition, the cells of FA patients exhibit hypersensitivity to DNA interstrand cross-links (ICLs) caused by agents such as mitomycin C (MMC) (Godthelp et al., 2006).

DISCUSSION
In the present study, we analyzed the BRCA2 c.68-7T > A variant, located in the proximity of the acceptor site of exon 3, in order to establish its clinical relevance and association with breast cancer risk.
Moreover, we found that in LCLs of carriers of the variant the exon 3 exclusion rate (i.e., the relative amount of BRCA2 ∆3 transcripts) was approximately 4-to 5-fold higher than in LCLs of controls and the total amount of ▼3 transcripts in carriers was approximately 50% compared to controls. The latter finding would seem to contradict the observation that the variant allele maintains the ability to express a transcript coding for a normal (full-length) protein. The apparent discrepancy may be explained comparing the overall expression of BRCA2 transcripts in cases and controls. In fact, summing up in each sample the amount of ▼3 and Δ3 transcripts assessed by qPCR, and setting as 1 the average expression of BRCA2 mRNA observed in our cohort, we observed notable inter-individual variability (ranging from 0.43 to 1.50), with many control samples clustering above the average (Supp. Figure S3). Hence, it is very much possible that the strong reduction in the amount of ▼3 transcripts observed in carriers simply reflects random inter-individual variability in BRCA2 gene expression levels.
Although the above findings were confirmed using two complementary assays (qPCR and dPCR), it must be noted that the outcomes of transcript quantification analyses may be influenced by the nature of examined biological material. Therefore, the magnitude of changes in transcript ratio associated with the c.68-7A > T should be verified also in samples other than LCLs.
The pathogenic implication of BRCA2 exon 3 deletion has been long debated. Exon 3 is 249-bp long and its deletion does not alter the open reading frame. In addition, the ∆3 isoform has been described as one of the major naturally occurring alternatively splicing events in BRCA2 (Fackenthal et al., 2016). However, the predicted protein product is expected to be lacking important functional activities. In fact, this exon codes for BRCA2 amino acids 23 to 106, including the C-terminal portion of a primary transactivating domain (PAR, amino-acid residues 18-60) and an auxiliary transactivating domain (AAR, residues 60-105) (Milner, Ponder, Hughes-Davies, Seltmann, & Kouzarides, 1997), whose activity may be regulated by phosphorylation (Milner, Fuks, Hughes-Davies, & Kouzarides, 2000). Interestingly, the region spanning residues 21-39 mediates the interaction with PALB2, a nuclear protein that promotes the stable intranuclear localization and accumulation of BRCA2, making possible its DNA recombinational repair and checkpoint functions, eliciting tumor suppression (Oliver, Swift, Lord, Ashworth, & Pearl, 2009;Xia et al., 2006). Moreover, the PALB2binding site directly overlaps that of EMSY, another nuclear protein that has endogenous transcriptional repressor activity (Hughes-Davies et al., 2003).
Several BRCA2 alterations causing the complete loss of exon 3 and the exclusive synthesis of ∆3 transcripts have been ascertained, including c.316 + 5G > C (Bonnet et al., 2008), c.316 + 3delA and c.68-925_316 + 2889del (Muller et al., 2011) and c.156_157insAlu, a variant reported as a founder Portuguese mutation (Peixoto et al., 2009 The difference between the findings from our much larger casecontrol study and that of Møller & Hovig, (2017) need for caution when utilizing case-control data for clinical interpretation of rare variants, such that significant differences in frequency can nonetheless be unreliable due to random error and bias arising from small sample size, incomplete matching of cases and controls, and when considering familial cases, co-occurrence of (other) risk-related genetic factors as acknowledged by the authors themselves.
Different hypotheses, not necessarily mutually exclusive, can be proposed to explain the lack of pathogenicity of c.68-7T > A despite it being spliceogenic. First, the reduction in full-length BRCA2 mRNAs in variant carriers compared to normal controls, which was not statistically significant, might not be enough to affect cellular tumor suppressor ability. Second, the ∆3 transcripts are predicted to lead to the synthesis of an unstable and nonfunctional protein product and, therefore, unlikely to interfere with the activity of the normal protein due to the loss of the PALB2 interaction domain, whose binding stabilizes the BRCA2 protein (Xia et al., 2006). Assuming that in the examined samples, the overall BRCA2 expression level from both alleles is similar, and that in carrier samples the accompanying normal alleles contribute on average an exclusion rate of approximately 3% as assessed by our quantitative analyses, we estimated, based on an average cumulative exclusion rate of both alleles in variant carriers of 13%, that the average exclusion rate (x) for the c.68-7T > A allele is close to 23% [(x% + 3%)/2 = 13%.]. Therefore, the present study strongly suggests that BRCA2 spliceogenic alleles demonstrating up to approximately 20% exon 3 exclusion rates are not associated with high or even moderate risk of cancer.
The classification of variants based on mRNA splicing data alone is problematic for spliceogenic variants that lead to equivocal or "leaky" transcript profiles. The quantitative in vitro transcript and genetic analyses conducted for BRCA2 c.68-7T > C provide important data to inform the threshold for ratio between functionally proficient and altered BRCA2 isoforms compatible with normal cell function. These findings might facilitate the future classification of rare spliceogenic variants whose relevance for cancer risk cannot easily be ascertained through multifactorial likelihood analyses.