S. Verschuere, P. Coucke, O. M. Vanakker
Center for Medical Genetics Ghent, Ghent University Hospital
Objectives: Pseudoxanthoma elasticum (PXE) is an autosomal recessive ectopic mineralization disorder affecting the skin, eyes and cardiovascular system. There is high interindividual and familial phenotypic variability, without solid genotype-phenotype correlations. PXE is caused by loss-of-function variants in ABCC6, with more than 400 ABCC6 variants reported in literature and variant databases such as NCBI ClinVar. Inconsistencies in variant interpretation methods raised the question whether all these variants are truly disease-causing. Especially for missense substitutions - accounting for more than half of reported variants - it is challenging to predict the impact on protein function. Correct variant interpretation is essential for diagnosis, management and counseling of patients and their relatives - particularly as it has become clear that heterozygous carriers may also have a partial phenotype. We therefore re-evaluated all ABCC6 missense variants in a reproducible manner and classified them according to pathogenicity.
Methodology: We evaluated all ABCC6 missense variants from ClinVar, from literature and novel variants from in-house patient screenings. In total, 234 variants were analysed using the numerical score-based variant classification system Sherloc (Nykamp et al. 2017). Clinical (allele frequency, case reports, segregation data) and functional evidence (experimental data, effect on protein, in silico data) were scored with benign (B) or pathogenic (P) points using hierarchical decision trees. Classification was based on the total score: benign (score ≥ 5B), likely benign (3B ≤ score < 5B), variants of uncertain significance (VUS) (score < 3B or < 4P), likely pathogenic (4P ≤ score < 5P) and pathogenic (score ≥ 5P). To distinguish between different types of VUS we created additional subclasses: those with a benign tendency (2B ≤ score < 3B), those that are truly uncertain (score < 2B and < 3P) and those with a pathogenic tendency (3P ≤ score < 4P).
Findings: Comprehensive classification revealed that 74% of ABCC6 missense variants are of uncertain significance. (Likely) pathogenic variants account for 12% and 10% respectively, almost all of which are located in the protein’s functional domains. (Likely) benign variants are the least represented (1% and 3% respectively). VUS classification was further refined into those that lean towards likely benign (1% of VUS), those that are truly uncertain (69% of VUS) and those that lean towards likely pathogenic (30% of VUS) based on their total score. Taking the latter also into account, 44% of ABCC6 variants is (likely) pathogenic when considering all available population, clinical, experimental and in silico data. This is significantly different from ClinVar, where 87% of ABCC6 missense variants are said to be pathogenic and only 11% are VUS.
Significance: Evaluation of ABCC6 missense variants using the most recent and comprehensive criteria reveals an overestimation of (likely) pathogenic variants in literature and databases. Data that are missing for many variants are experimental validation and - surprisingly- segregation data. Our results underline that variant classification should be done systematically and with caution, as it has important consequences for patients and carriers identified via familial or expanded carrier screening. The high number of VUS confirms the need for functional testing to prove or refute their causality before returning them to patients.