SCN5A Mutation Type and a Genetic Risk Score Associate Variably With Brugada Syndrome Phenotype in SCN5A Families

Supplemental Digital Content is available in the text.


Classification of variant pathogenicity and subsets
All variants were curated using CardioClassifier (including paralogue annotation). 2 Variants were annotated further by including published functional data to determine variant class using ACMG criteria. 3 Only pathogenic and likely pathogenic variants were included as mutations for this study (see Supplemental Table 1). Frameshift or splice site variants were defined as 'loss-of-function' causing haploinsufficiency.

SNP selection
Three single nucleotide polymorphisms (SNPs) have been associated previously with BrS phenotype in Caucasian and Japanese cohorts at a genome-wide level of significance: rs11708996 (SCN5A), rs10428132 (SCN10A), and rs9388451 (HEY2) (see Supplemental Table 2 for general population frequencies). 4 The genes tagged by rs10428132 and rs9388451 have undergone functional evaluation previously. [5][6][7][8] rs11708996 and rs9388451 were genotyped in all individuals. In families harboring SCN5A mutations other than SCN5A-E1784K, rs6800541 formed part of a multi-SNP assay previously carried out in this cohort and was found to be in complete linkage disequilibrium with rs10428132 (r 2 = 1). It was therefore used as a proxy SNP.
SNP genotyping SNP genotyping in SCN5A-E1784K families was carried out using the PCR-based Kompetitive Allele Specific PCR (KASP TM ) genotyping assay provided by LGC, UK. 9 The KASP TM assays were designed using the Kraken TM software system and validated through LGC's assay validation pipeline. 10  or where the genotype for any of the three SNPs was missing.

Statistical analyses
Categorical variables are described as count and percentage and numerical variables as mean ± standard deviation (SD) or median and interquartile range (Q1-Q3: IQR). Categorical variables were compared using Chi-squared and Fisher exact tests where appropriate and numerical variables were compared using ANOVA. For SNP effects, an additive genetic model was assumed. The BrS-GRS was calculated by summing the number of risk alleles for the three included SNPs that each person carried (range 0-6 alleles). The weighted BrS-GRS was calculated by multiplying the individual risk allele counts with their respective associated effects size (β). The risk alleles and associated effect sizes were assigned based on literature. 5 The weighted BrS-GRS was also tested, but this did not outperform the non-weighted BrS-GRS (data not shown). The assumption of a linear relationship between the BrS-GRS and the natural logarithm of the odds ratio was tested by comparing the fit of the linear model with that of a more flexible model with restricted cubic splines. No significant deviation from linearity was found. Next to a model with the BrS-GRS as a numerical predictor, we also tested the association with a dichotomized predictor (BrS-GRS  4 yes/no) based on the distribution of risk alleles seen in the overall cohort. All comparisons between BrS phenotype-negative and -positive individuals were carried out using generalized linear models and corrected for sex and age. Generalized estimation equations (GEE), with an independence correlation structure and robust sandwich variance estimations were used to correct for relatedness (geepack, version 1.2-1). 11 The BrS-GRS thus developed was then examined in the total population and subsets of family members harboring SCN5A mutations: loss-offunction causing haploinsufficiency; SCN5A-E1784K; and other missense mutations.
SCN5A negative relatives were then studied separately. Interaction terms were added to the models to check for differences in mutation/GRS effects in genotype negative vs. positive and among genotype positive subgroups. Haplotype analyses were not feasible due to small sample sizes.
NA= not applicable.