Protein S Heerlen mutation heterozygosity is associated with venous thrombosis risk

Hereditary Protein S (PS) deficiency is a rare coagulation disorder associated with an increased risk of venous thrombosis (VT). The PS Heerlen (PSH) mutation is a rare S501P mutation that was initially considered to be a neutral polymorphism. However, it has been later shown that PSH has a reduced half-life in vivo which may explain the association of PSH heterozygosity with mildly reduced levels of plasma free PS (FPS). Whether the risk of VT is increased in PSH carriers remains unknown. We analyzed the association of PSH (rs121918472 A/G) with VT in 4,173 VT patients and 5,970 healthy individuals from four independent case-control studies. Quantitative determination of FPS levels was performed in a subsample of 1257 VT patients. In the investigated populations, the AG genotype was associated with an increased VT risk of 6.57 [4.06–10.64] (p = 1.73 10−14). In VT patients in whom PS deficiency was excluded, plasma FPS levels were significantly lower in individuals with PSH when compared to those without [72 + 13 vs 91 + 21 UI/dL; p = 1.86 10−6, mean + SD for PSH carriers (n = 21) or controls (n = 1236) respectively]. We provide strong evidence that the rare PSH variant is associated with VT in unselected individuals.

Scientific RepoRts | 7:45507 | DOI: 10.1038/srep45507 Dutch patients with unexplained VT (0.52% vs 0.68% respectively) 6 . Apart from this study with poorly described patients, the impact of PS Heerlen on VT has only been studied in families either with PS deficiency or with other combined genetic or acquired risk factors. In these situations, PS Heerlen was thought to increase VT susceptibility 10 .
Although PS Heerlen is more frequent than any other PROS1 mutation, it is still not considered a common variant. Consequently, previous genome-wide association studies (GWASs) which mainly interrogate common genetic variations (i.e. frequencies higher than 1%) were not able to identify the mutation 11 .
In the present study, we aimed to systematically determine if there was an association between PS Heerlen and VT by combining four French case-control DNA samples totaling 4,173 patients and 5,970 healthy individuals.

Results
Genotype distributions of the rs121918472 variant within the four case-control studies are shown in Table 1. No homozygous carrier of the rare rs121918472-G allele was observed. When samples were combined, the AG genotype was more frequent in VT patients than in healthy controls, 2.2% vs 0.4% respectively, and was associated with an increased VT risk (Odds Ratio OR) of 6.57 [4.06-10.64] (p = 1.73 10 −14 ). The association was significantly heterogeneous across studies (p < 10 −3 ) due to strong associations in the EDITH cohort (2.4% vs 0.4%), the MARTHA12 cohort (2.5% vs 0.6%), the EOVT/MARTHA cohort (2.4% vs 0%) samples, but not in FARIVE study samples (1.1% vs 1.8%). Note that the observed AG genotype frequency was higher in FARIVE controls with hypertension (2.9%, n = 246) than in those without (0.9%, n = 343) in whom the frequency was similar to the other cohort controls. Excluding the FARIVE study from the combined analysis increased the significance of the association between rs121918472 and VT (OR = 14.58 [7.42-28.65], p = 7.66 10 −15 ).
We also examined whether the rs121918472-G allele frequency varied in specific VT subgroups, such as in patients with provoked or unprovoked VT, patients with deep vein thrombosis or pulmonary embolism, and patients with F5 or F2 G20210A mutations. We did not observe preferential associations in any specific strata.
Interestingly, the rs121918472 variant was not identified as a VT-associated polymorphism in the two most recent VT GWAS projects by INVENT 11 and 23 and Me consortia 12 . In the INVENT GWAS, the Heerlen mutation imputation quality was extremely low, r 2 = 0.15 11 , which is far lower than the gold standard of r 2 > 0.8 for imputation quality, and the increased risk associated with the imputed rs121918472-G allele was OR = 1.46 [0.28-7.58], p = 0.655. Because 865 MARTHA/EOVT participants examined in the current study also took part in the INVENT study, we were able to calculate the Spearman correlation between the true rs121918472 genotype as provided by Illumina Exome array results, and the imputed dose-derived from GWAS arrays. The resulting low correlation of ρ = 0.25 illustrated how imputation techniques are not reliable for inferring rare variants such as rs121918472.
To provide additional evidence implicating rs121918472 in increased VT risk, we assessed its association with plasma FPS levels in the MARTHA study samples. Even if by design FPS deficiency was excluded in this study, plasma FPS levels were significantly lower in individuals with PS Heerlen (p = 1.86 10 −6 ) compared to those without (mean ± standard deviation: 72 ± 13 vs 91 ± 21 UI/dL, respectively) ( Fig. 1).

Discussion
In the present study we systematically analyzed four French case-control collections totalling 4,173 patients and 5,970 healthy individuals and provide strong evidence that the rare PS Heerlen variant is associated with VT.
In the investigated populations, the AG genotype was significantly more frequent in VT patients than in healthy controls, 2.2% vs 0.4%, respectively, and was also associated with an increased VT risk of 6.57. This  13 . In the MARTHA study, even if FPS deficiency was excluded by design, plasma FPS levels were significantly lower in individuals with PS Heerlen compared to those without. Assessing whether the disease effect is completely mediated through a direct effect on decreased FPS levels would require further investigations, including epidemiological studies where both the PSH mutation and FPS levels are measured in case-control samples. However, due to the extremely low frequency of the PSH variation in well matched control population, it will be nearly impossible to conduct mediation analysis to properly assess whether the effect of the mutation on disease risk is completely due to its effect on lowering FPS levels. However, in an additional population based study, only very low levels of FPS (< 1/10 th percentile or < 33 UI/dl) were associated with increased VT risk 1,14 , suggesting that the effect of PS Heerlen on VT risk might thus be related to mechanisms other than solely half-life reduction. However, recombinant PS Heerlen does not differ in either binding to, or dissociation to or from, immobilized C4BP in vitro, nor does it show any differences in the inactivation of FVa or FVIIIa 15 . PS is also a cofactor for TFPI, and accelerates the inhibition of activated factor X (FXa). The PS Heerlen mutation is located within the SHBG-like domain (Val243-Ser635) critical for TFPI binding and enhancement 16 . Thus, one might hypothesize that PS Heerlen could affect thrombosis risk by impairing TFPI function. The recent GWAS on VT has enabled the identification of several new VT susceptibility variants. Very recently, the INVENT consortium gathered 7,507 VT subjects and 52,632 controls and identified two new loci (SLC44A2 and TSPAN15) and robustly replicated six previously identified loci 11 . PS Heerlen was not among the identified disease polymorphisms, likely because this GWAS relied on imputation techniques that were not optimal for inferring rare variants 17 . The exome chip used in the present study was enriched by rare coding variants predicted to be causal. The results presented herein as well as those from the recently published paper on key hemostatic traits 18 emphasize the promise of this new chip in identifying low-frequency variants associated with VT.
In conclusion, we provide evidence that PS Heerlen is associated with increased VT risk in individuals of the general population. Further investigations are required to better understand the surprisingly high prevalence of PS Heerlen mutation in apparently healthy individuals from the FARIVE study. Functional studies investigating the potential influence of PS Heerlen on VT risk would also be of benefit. These results also raise the question of whether PS Heerlen should always be genotyped irrespective of plasma PS levels in thrombophilia screening.

Methods
The present work utilized four case-control sample cohorts (EDITH, FARIVE, EOVT/MARTHA) 11 where VT events (pulmonary embolism or deep vein thrombosis) were objectively diagnosed 11 . Detailed descriptions of the four case-control sample groups are included in the supplementary text.
In the FARIVE and EDITH studies, the PS Heerlen variation was genotyped by High Resolution Melting (HRM). All EOVT and MARTHA patients were typed using the Illumina Human Exome Beadchip 12v1-2-A array and compared to control participants from the Paris Prospective Study 19 typed with the Human Exome Beadchip 12v1-A chip (Illumina, Inc., Sand Diego, CA). Both of these genotyping arrays were designed to assess coding variants across the whole genome. Data analysis revealed 239,569 single nucleotide polymorphisms common to both arrays, including PS Heerlen rs121918472. Individuals with less than 98% genotyping success or high heterozygosity were excluded from the analyses, as were individuals demonstrating previously unknown close relatedness. The latter were separately assessed via pairwise clustering of identity by state distance (IBS) and multi-dimensional scaling (MDS) using PLINK software 20 . Genetic outlier analysis was performed using IBS statistics and principal component analysis to detect individuals of non-European origin, who were then removed from the analysis. This led to a selected cohort of 2,630 VT patients and 3,414 controls typed with Illumina exome chips for final association analysis.
The overall rs121918472 genotype call rate was 0.98, with 0.97 in the FARIVE and EDITH studies typed with HRM, and 0.99 in individuals typed with the Illumina Human Exome chips.
In MARTHA, fasting blood was drawn for FPS measurement. Quantitative FPS determination was performed by enzyme-linked immunosorbent assay (ELISA) using the Asserachrom FPS assay (Diagnostica Stago). PS deficiency was defined as FPS plasma antigen levels < 55 IU/dL −1 using the specific ELISA assay. FPS data were available in 1,257 MARTHA individuals.
Genotype distribution of the rs121918472 was tested for deviation from Hardy-Weinberg (HW) equilibrium via the HW exact test. Association of rs121918472 with VT was analyzed via the Cochran-Armitage trend test and by logistic regression analysis adjusting for age and sex. Association of rs121918472 with FPS was assessed using linear regression analysis. Analyses were first performed separately in each study, and were then performed on all samples combined and further adjusted for the study group.
Informed consent was obtained from all participants, and the study met all institutional ethics requirement. The procedures employed were reviewed and approved by the Assistance Publique des Hopitaux de Marseille institutional review committee.