We report the results of a splicing assay for a number of variants, some with aberrations and some with normal results. In both cases, the results are of significant clinical utility.
All the detected abnormal transcripts resulted in premature stop codons. Transcripts containing premature termination codons are degraded by nonsense-mediated mRNA decay (NMD), unless they are localized less than 50 nucleotides upstream of the position of the terminal intron [10, 11]. We concluded that the variants that gave rise to only transcripts with premature stop codons are pathogenic as they are all appear in parts of the gene where we expect them to be degraded by NMD.
We characterized two novel pathogenic splicing variants, BRCA1 c.5407–10G > A and c.4484 + 3A > C. BRCA1 c.5407–10G > A had been reported as VUS in the ClinVar database (a database on the relationships between human variations and phenotypes) and listed once in the genome Aggregation Database (gnomAD) (> 246,000 alleles). BRCA1 c.4484 + 3A > C had not been reported before to our knowledge, neither in publications nor in databases. We classified these two variants as pathogenic as both led to complete splicing aberrations which altered the reading frame leading to premature stop codon; BRCA1 c.5407–10G > A through retention of eight intronic bases, and BRCA1 c.4484 + 3A > C thorough skipping of exon 13 (the same transcript as for BRCA1 c.4484G > A). In these two cases the mRNA results directly affected the clinical management of the patient. In addition, their relatives were given the opportunity to undergo predictive genetics testing, giving carriers the possibility to opt for disease screening and/or prevention.
For the other two novel variants with abnormal splicing, we were unfortunately not able to draw any conclusions about pathogenicity. BRCA1 c.5332 + 4A > G had been reported in ClinVar, three times as a variant of unknown significance and one registration as likely pathogen by CIMBA, but without provided evidence for the conclusion. BRCA1 c.5332 + 4A > G was listed four times in gnomAD (> 246,000 alleles). BRCA2 c.8754 + 5G > C led to retention of 46 bases of the 5′-end of intron 21 in the mutant transcript. This variant has not been reported before to our knowledge, neither in publications, nor in databases, but the same transcript has previously been found for another pathogenic variant in the close vicinity, namely BRCA2 c.8754 + 3 [12].
Five of the variants with aberrant splicing had been published before. The methods used for splicing assay were slightly different, but we detected the same transcripts and came to the same conclusions regarding pathogenicity for these variants. The two known pathogenic BRCA1 variants, c.4484G > A and c.4675G > A, had both been published with functional studies in patient RNA in 2012 [6, 13]. They have since been reported by different laboratories in ClinVar and in different populations. BRCA1 c.4484G > A leading to skipping of exon 13 (previously known as exon 14) has been reported in patients in France [14], Australia [15] and Brasil [16]. BRCA1 c.4675G > A leading to loss of the last 11bases of exon 14 (previously known as exon 15) has been reported in patients in Germany [13] and Lithuania [17]. The variant BRCA2 c.631 + 4A > G was found in a Danish breast and ovarian cancer family and analyzed with a combination of minigene and RT-PCR with the conclusion of a disease causing mutation [18]. The variant BRCA2 c.8331 + 2 T > C has been reported several times in ClinVar and is assumed to be pathogenic because of the localization in the canonical donor splice site. However, it was only in 2017, that the result of a functional assay was published, a mingene assay that demonstrated 87% exon 18 skipping [19]. We confirmed a complete pathogenic exon 18 skipping in a patient sample.
BRCA2 c.7992 T > A led to a partial skipping of exon 18, which is a naturally occurring minor alternative transcript [20]. BRCA2 c.7992 T>A increased the amount of the alternative transcript lacking exon 18. This variant remains a variant of unknown significance, consistent with the conclusion drawn by Fraile-Bethencourt et al. [19]. They quantified the ratio between the transcripts produced by this variant in a minigene assay and found 69% of the normal transcript and 31% of the transcript lacking exon 18.
Sixteen variants showed normal splicing. Twelve of them were demonstrated to have normal transcripts produced by both alleles. Of these, seven were intronic variants and one was a synonymous exonic variant. We classified these eight variants as benign based on the results of the splicing assay. To our knowledge, we are the first to report the results of a splicing assay for four of these, including BRCA1 c.594-20A > G, c.4676-8C > G and c.4987-4 T > G and BRCA2 c.7436-4A > G. These variants are rare (reported in gnomAD once, eighteen, two and four times, respectively).
Four of the variants with normal splicing pattern were exonic variants leading to amino acid substitutions, which have to be assessed separately regarding a potential effect on protein function, but we can at least rule out major splicing aberrations. Three of these exonic variants (BRCA1 c.441G > C p.(Leu147Phe) and BRCA2 c.7006C > T p.(Arg2336Cys) and c.9257G > C p.(Gly3086Ala)) had not been analyzed in a splicing assay before, as far as we know.
Three BRCA1 variants c.81-13C > G, c.81-14C > G and c.4676-8C > G were predicted to attenuate the acceptor splice site. For two of these, c.81-13C > G and c.4676-8C > G, we and others, demonstrated normal splicing. For BRCA1 c.81-14C > G, we detected only normal transcripts, but were unfortunately not able to demonstrate biallelic expression and could therefore not draw a firm conclusion. BRCA1 c.81–13 lies in the polypyrimidine tract. Even though the substitution of a C with a G interferes with the polypyrimidine stretch, there are still 16 pyrimidines in a row, so we assume that the substitution does not affect the qualities of the tract. The same argument would apply for BRCA1 c.81-14C > G. For the variant BRCA1 c.4676-8C > G, the pyrimidne tract is shorter, but closer to the intron-exon boundery, which could be an explenation for preserved normal splicing [21].
The fact that several of the variants, both the pathogenic and the normal ones, are very rare, illustrate the need for functional studies for variant classification in these cases.
When it comes to selecting variants for splicing assays, ideally one should analyze all variants, as any variant could potentially affect splicing [5]. There is no efficient way to do this yet. Sensitivity of splicing prediction tools is highest in the vicinity of consensus splice sites, whereas predictions of variant effect on splicing enhancers, silencers and branchpoint have poor specificity [6]. Bioinformatic predictions must be confirmed by functional mRNA analysis.
Overall, there was a good concordance between splicing predictions and results of mRNA analyses. All variants that altered transcription were predicted to do so. However, in line with others, we demonstrated that even if the bioinformatic tools are successful in predicting reduction in a splice site, they may not be able to discriminate between activation of a new site versus exon skipping [8]. For the variants that maintained a normal splicing pattern, a few were predicted to alter splicing. These were located in the polypyrimidine tract. It is known that the polypyrimidine tract is quite variable from gene to gene and intron to intron. The function of the polypyrimidine tract is dependent in a number of factors, including total length of the tract, number of consecutive pyrimidines, distance to the intron-exon boundary, distance to the branch point and strength of the branch point [21, 22]. It seems that the variation in all these factors might make it difficult to predict whether a single change in the polypyrimidine tract results in altered splicing or not, by the bioinformatic tools, giving us another example of the benefit of splicing assays.
Minigene assays are valuable in assessing splicing effects of a variant, especially if there is no available patient mRNA [19, 23]. An advantage with this method is the investigation of the expression of one allele at a time. However, as the transcription takes place in a different environment than a human cell, and discrepancies in results from minigenes and patient samples have been reported, confirmation in patient RNA is recommended if possible. We confirmed the result of minigene reported by others in patient RNA for a couple of the variants. A disadvantage with the patient RNA method is the need for a normal variant in the patient sample, to be able to discriminate between the alleles in the distinction between partial or complete splicing aberration, and the assessment of biallelic expression in normal transcripts.
As the splicing process could be differently regulated in different tissues, a sample from breast or ovarian tissue would be ideal for the study of splicing variants in the breast cancer genes. This is usually not available in the clinical setting. However, ENIGMA concluded that BRCA1 alternative splicing is similar in blood and breast, supporting the clinical relevance of blood based in vitro splicing assays [24].
The methods used here were shown to work well and help in the functional assessment of several variants. However, there are numbers of potential splicing variants that we are missing because of lack of efficient ways of identifying these, such as variants affecting splicing enhancers and silencers. These and other regulatory variants will probably be better characterized in the near future.