Leveraging Population-Scale Multiomic Datasets to Elucidate the Risk of Thrombosis Associated with Protein S (PROS1) Deficiency

Introduction

Despite the clinical importance of protein S (PS), effect size estimates for the risk of venous thromboembolism (VTE) associated with PS deficiency remain strikingly divergent. Prior investigations have largely relied on plasma PS levels without genetic confirmation, and the largest study to date (N=5,317) found little or no association between plasma PS levels and risk of a first episode of VTE (Pintao, et al. Blood 2013;122(18): 3210-3219). By contrast, others have reported that low plasma PS levels moderately increase the risk of VTE. Most studies were limited by small sample size, heterogenous case and control groups, failure to adjust for potential confounders, and substantial ascertainment bias (i.e. reliance on families with high penetrance of VTE). To overcome these limitations, we utilized population-scale next-generation sequencing data to assess the clinical impact of inherited PS deficiency across over 600,000 individuals.

Methods

We utilized two biorepositories with paired genetic and phenotypic data: the UK Biobank (UKB, N=426,436) and the NIH All of Us program (AoU, N=204,006). Rare (MAF <0.1%) germline variants in protein S (PROS1) were each assigned an in silico functional impact score (FIS) between 0.0 and 1.0, with higher scores reflecting greater predicted likelihood that a variant results in protein loss of function. An FIS of 1.0 denotes mutations that cause stop-gained (nonsense) mutations, essential splice site disruptions, or frameshifts. Multivariate Firth’s logistic regression adjusting for age, sex, and ancestry was used to assess the risk of thrombosis in participants with PROS1 loss-of-function variants (FIS ≥0.7 or 1.0) relative to non-carriers. Circulating total PS levels were measured in a subset of UKB participants (N=44,864) using plasma proteomics (Olink® Explore 3072) and reported as linearized normalized protein expression (L-NPX) units.

Results

In the UKB, we identified 961 carriers of a PROS1 variant with FIS ≥0.7 (99.8% heterozygous) and 39 carriers of a variant with FIS=1.0 (100.0% heterozygous). Plasma PS was significantly decreased in FIS=1.0 (mean PS=51.2%, P=0.0003) but not FIS ≥0.7 (mean PS=97.1%, P=0.21) variant carriers. The presence of PROS1 variants (FIS ≥0.7) was significantly associated with VTE (OR=1.98, 95% CI: 1.55-2.48, P=1.95x10^-7). Participants with FIS=1.0 variants experienced further elevation in VTE risk (OR=14.0, 95% CI: 6.98-27.14, P=9.09x10^-11). We confirmed the accuracy of clinical annotation in the UKB by reproducing known VTE associations, including age, BMI, C-reactive protein, and smoking. Our findings independently were replicated in the AoU dataset (OR=2.99, 95% CI: 2.13-4.08, P=7.54x10^-9 for FIS ≥0.7; OR=13.8, 95% CI: 6.66-28.16, P=8.1x10^-11 for FIS=1.0). Sensitivity analysis showed that the VTE risk conferred by PROS1 was independent of carrier status for the Factor V Leiden and prothrombin G20210A mutations. Additionally, PROS1 variant carriers had significantly higher rates of incident VTE (log-rank P=0.0006). By contrast, PROS1 loss-of-function did not significantly increase the risk of any form of arterial thrombosis, including myocardial infarction, stroke, and peripheral arterial disease.

Conclusions

We have conducted the largest study to date of protein S deficiency. True genetically-defined quantitative (Type I) PS deficiency is rare but more strongly linked to VTE than previously appreciated, with effect size estimates resembling those typically seen in thrombophilias such as Type I antithrombin deficiency. Missense variants in PROS1 that collectively have little effect on circulating protein levels (Type II deficiency) are more common and modestly increase VTE risk. Our data may help resolve longstanding uncertainty around the clinical implications of PS deficiency.