Use of Large-Scale Genomic Datasets to Identify Additional Modifiers of Thrombosis Risk Among 26,436 Carriers of the Factor V Leiden and Prothrombin G20210A Mutations

Introduction

Individuals with the factor V Leiden (FVL, rs6025) and prothrombin G20210A (PTGM, rs1799963) mutations experience a significantly increased risk of venous thromboembolism (VTE). Nevertheless, most carriers never develop VTE. The identity of additional genetic modifiers in this population remains a major unanswered question in the field, and investigators have long hypothesized that genetic differences in both the rare and common variant space contribute to overall risk. To date, addressing this problem has proven extremely challenging due to the lack of sufficiently large cohorts with paired clinical annotation and genomic sequencing results. Deeply-phenotyped, population-scale genomic datasets offer a unique opportunity to overcome these limitations and provide novel insights on the determinants of thrombosis in individuals with these two commonly-encountered polymorphisms.

Methods

We utilized the UK Biobank (UKB), a population-scale biorepository with paired whole exome sequencing and phenotype data (N=454,787). Our primary analysis focused on the composite cohort of FVL (N= 21,153) and PTGM (N=10,072) variant carriers. To examine the common variant space, a genome-wide association study (GWAS) was conducted using imputed genome-wide data. To examine the rare-variant space with maximum power, we performed pathway-based collapsing analysis. Directionally consistent pro- and anti-coagulant gene sets were derived from the Kyoto Encyclopedia of Genes and Genomes (KEGG) coagulation cascade pathway definition (map04610). Firth’s logistic regression was then used to evaluate the collective impact of rare (MAF <0.1%) loss-of-function (LOF) variants in each of the two gene sets.

Results

The composite cohort (N=30,761) contained a total of 2,445 VTE cases and 28,316 controls. GWAS identified 330 variants in 7 loci that were significantly associated with the risk of VTE at genome-wide significance (P <5 x 10^-8), including 2 novel associations: NME7 (P=1.04 x 10^-12) and FOXK2 (P=9.85 x 10^-9). The top association result was rs8176719 in the ABO locus, which predicts group O blood type and was protective against VTE (OR=0.70, 95% CI: 0.65 – 0.74, P=1.9 x 10^-31). A secondary analysis restricted to FVL carriers demonstrated that 169 variants in 3 loci reached statistical significance, with the top result being rs687621 in the ABO locus (OR=1.48, 95% CI: 1.37 – 1.59, P=1.1 x 10^-26). Similarly, in PTGM carriers we identified 11 variants in 2 loci that reached genome-wide significance, including rs6025, the factor V Leiden variant (P=4.6 x 10^-9), and ABO (P=8.75x10^-9). Evaluating rare variants, the procoagulant gene set (N=20 genes) showed no association with VTE risk in the composite cohort. By contrast, rare LOF variants in known anticoagulant genes (N=11 genes) were collectively associated with increased VTE risk after adjusting for age, sex, and ancestry (OR=5.0, 95% CI: 2.19-10.5, P=0.0003). Within the anticoagulant gene set, rare LOF variant burden in PLAT (tissue plasminogen activator) reached nominal significance as a single gene (OR=12.2, 95% CI: 2.56-58.3, P=0.003).

Conclusion

We have shown for the first time that both known and novel common genetic risk factors for VTE continue to exert a significant effect in carriers of FVL and PTGM, particularly ABO blood type. Intriguingly, rare loss-of-function variants in known anticoagulant genes collectively enhanced thrombosis risk, with data suggesting an important role for the fibrinolytic system. Our work may help improve risk stratification of patients with the two most common inherited thrombophilias.