Enhancing variant classification in hereditary breast and ovarian cancer predisposition genes through splicing analysis and new tumour evidence

Abstract

[eng] The clinical management of hereditary breast and ovarian cancer (HBOC) families relies on identifying germline pathogenic variants in cancer susceptibility genes, such as BRCA1 and BRCA2. However, genetic testing often yields an inconclusive result due to detection of only neutral variants or to the identification of variants of unknown significance (VUS), whose effects on gene function and associated cancer risk are unknown. While carriers of pathogenic variants benefit from options available for early cancer detection, preventive measures and targeted therapies, VUSs carriers are generally managed without regard to the result of genetic testing, thus limiting the options for these patients. The primary aim of this thesis is to increase the number of patients receiving a genetic diagnosis of HBOC by improving the accuracy of classifying germline genetic variants. To achieve this, we focused on improving the detection of potentially splicing-altering variants in poorly understood deep intronic regions (Part 1), evaluating the relevance of mRNA splicing experimental validation for variant classification (Part 2) and exploring the potential use of tumour information in the classification of germline variants (Part 3). Deep intronic regions may contain variants that affect RNA splicing by introducing “pseudoexons” or other changes, which could potentially explain cases where the genetic cause of HBOC remains unknown. Current computational in silico tools to predict spliceogenic variants leading to pseudoexons have limited efficiency. In Part 1 of this thesis, we have assessed the performance of the SpliceAI tool combined with ESRseq scores to identify spliceogenic deep intronic variants by affecting cryptic sites or splicing regulatory elements (SREs) using literature and experimental datasets. Our results with 233 published deep intronic variants showed that SpliceAI, with a 0.05 threshold, predicts spliceogenic deep intronic variants affecting cryptic

splice sites, but is less effective in detecting those affecting SREs. Next, we characterised the SREs profiles using ESRseq, showing that pseudoexons are significantly enriched in SRE-enhancers compared to adjacent intronic regions. This indicates that intronic regions with a high potential to be included as pseudoexons can be systematically identified throughout the HBOC genes. Although the combination of SpliceAI with ESRseq scores (considering ∆ESRseq and SRE landscape) showed higher sensitivity, the global performance did not improve because of the higher number of false positives. Both tools combined were tested in a tumour RNA dataset with 207 intronic variants disrupting splicing, showing a sensitivity of 86%. Following the pipeline, five spliceogenic deep intronic variants were experimentally identified from 33 variants in HBOC genes. Overall, our results provided a framework to detect deep intronic variants disrupting splicing. A possible pathogenic mechanism for a subset of VUS is the disruption of normal mRNA splicing. However, experimental validation is generally restricted to variants affecting dinucleotide splice site positions. Consequently, variants outside these regions are not consistently evaluated across laboratories, leading to conflicting and uncertain interpretations. Additionally, variants leading to in- frame isoforms affecting clinically relevant domains are challenging to classify due to the limited knowledge about in-frame encoded proteins. One clear example of both issues are the variants located in the in-frame exon 18 (legacy number) of BRCA1. The results of a comprehensive study conducted within the ENIGMA consortium to identify spliceogenic variants mapping to this exon and characterize the severity of their impact are presented in Part 2 of this thesis. Patient RNA, splicing reporter minigene, and mouse embryonic stem cell (mESC)-based functional assays were combined to determine the impact of 166

variants on splicing and estimate the tolerability functional range of exon 18 skipping (∆18). Splicing, functional and clinical data were integrated for variant interpretation according to the recently ClinGen-approved BRCA1/BRCA2- specified ACMG/AMP (American College of Medical Genetics and Genomics/Association for Molecular-Pathology) classification guidelines. Our results demonstrated that exon ∆18 is not a rescue event and that per-allele levels up to 34%, combined with at least 59% of the full-length (FL) isoform, are tolerable. Furthermore, we show that applying the recently developed BRCA1/2 specifications significantly reduced the rate of VUS in this exon from 45.7% to 8.5%, being experimental mRNA testing essential in solving uncertainty for approximately 10% of variants Lastly, despite the existence of multiple types of evidence for classifying germline variants, additional information is still required to classify many of them. Accordingly, we explored the potential use of tumour homologous recombination repair (HRR) status by RAD51 foci immunofluorescence detection (RAD51 test) as a predictor of pathogenicity or benignity for BRCA1/2 and PALB2 germline variants. HRR status by RAD51 test was successfully assessed in a total of 148 primary untreated breast and ovarian tumours, from Vall d’Hebron Hospital, ENIGMA and EraPerMed RAD51predict collaborators. Samples were split into three cohorts depending on patients’ germline status: pathogenic (91), benign (47) and VUS (8). Our results demonstrated that the RAD51 test could distinguish pathogenic from benign variants with high sensitivity (91.5%) and specificity (70.2%) (AUC 0.91). Likelihood ratios (LR) were calculated for the association of RAD51 test outcomes (HRR deficiency-HRD or proficiency-HRP), with BRCA1, BRCA2 and

PALB2 variant pathogenicity considering patients' clinical features. Then LR estimates were aligned to ACMG/AMP code strengths. Overall, RAD51-HRD status provided moderate evidence for pathogenicity if detected in primary breast tumours diagnosed before age 50, whereas RAD51-HRP provided informative moderate evidence against pathogenicity if detected in primary breast tumour independently of age at diagnosis. Lastly, we applied our recommendations to classify the eight VUS and about 38% (3/8) of variants were classified as likely benign, thus demonstrating the utility of the RAD51 test in variant classification.