SEARCH

SEARCH BY CITATION

Keywords:

  • deep vein thrombosis;
  • genetics;
  • pulmonary embolism;
  • thrombophlebitis;
  • venous thromboembolism

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Summary. Background: Venous thromboembolism (VTE) is highly heritable (estimated heritability [h2] = 0.62) and likely to be a result of multigenic action. Objective: To systematically test variation within genes encoding for important components of the anticoagulant, procoagulant, fibrinolytic and innate immunity pathways for an independent association with VTE. Methods: Non-Hispanic adults of European ancestry with objectively-diagnosed VTE, and age- and sex- matched controls, were genotyped for 13 031 single nucleotide polymorphisms (SNPs) within 764 genes. Analyses (n = 12 296 SNPs) were performed with plink using an additive genetic model and adjusted for age, sex, state of residence, and myocardial infarction or stroke. Results: Among 2927 individuals, one or more SNPs within ABO, F2, F5, F11, KLKB1, SELP and SCUBE1 were significantly associated with VTE, including factor (F) V Leiden, prothrombin G20210A, ABO non-O blood type, and a novel association with ABO rs2519093 (OR = 1.68, P-value = 8.08 × 10−16) that was independent of blood type. In stratified analyses, SNPs in the following genes were significantly associated with VTE: F5 and ABO among both genders and LY86 among women; F2, ABO and KLKB1 among FV Leiden non-carriers; F5, F11, KLKB1 and GFRA1 in those with ABO non-O blood type; and ABO, F5, F11, KLKB1, SCUBE1 and SELP among prothrombin G20210A non-carriers. The ABO rs2519093 population-attributable risk (PAR) exceeded that of FV Leiden and prothrombin G20210A, and the joint PAR of FV Leiden, prothrombin G20210A, ABO non-O and ABO rs2519093 was 0.40. Conclusions: Anticoagulant, procoagulant, fibrinolytic and innate immunity pathway genetic variation accounts for a large proportion of VTE among non-Hispanic adults of European ancestry.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Family and twin studies suggest that VTE is highly heritable (h2 = 0.62) and is likely to result from multigenic action as well as environmental exposures [1–3]. Several genetic disorders or mutations that affect the regulation of hemostasis have been associated with VTE in small case–control studies or observed in high VTE-risk pedigrees [4]. These include inherited deficiency of components of the anticoagulant pathway (e.g. deficiency of plasma antithrombin, protein C, protein S, protein Z/protein Z-dependent plasma protease inhibitor and tissue factor pathway inhibitor), mutations that upregulate (e.g. increased plasma concentrations of procoagulant factors, including fibrinogen [factor I], factor (F) II [prothrombin], FVIII, FIX, and FXI) or impair downregulation of the procoagulant pathway (e.g. FV Leiden), and mutations that downregulate fibrinolysis (e.g. FXIII and thrombin activable fibrinolysis inhibitor). Moreover, family and twin studies indicate that, among apparently healthy individuals, variation in plasma concentrations of procoagulant factors as well as markers of coagulation activation and fibrinolysis exhibit a high degree of heritability [1,5]. Finally, VTE has been associated with activation of the innate immunity pathway [6].

However, few studies have systematically tested variation within genes encoding for all known important components of these pathways for an independent association with VTE, and none estimated the joint population-attributable risk for each independent genetic VTE risk factor. To address these important gaps in knowledge, we performed a SNP-based candidate gene, case–control association study to test genetic variation within all known important components of the anticoagulant, procoagulant, fibrinolytic and innate immunity pathways for an independent association with VTE. Moreover, as such genetic variation may operate differently for VTE risk by gender and by other known genetic VTE risk factors, we performed stratified analyses by gender and by factor V Leiden, Prothrombin G20201A and ABO blood group non-O carrier status.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Study setting and population

We approached consecutive Mayo Clinic outpatients aged 18 years or older with objectively-diagnosed deep vein thrombosis (DVT) or pulmonary embolism (PE) who resided in the upper midwest United States and who were referred to the Mayo Clinic Special Coagulation Laboratory or Thrombophilia Center over the study period, 1994–2009, for study participation. We excluded patients with VTE related to active cancer, an indwelling central venous catheter, transvenous pacemaker or other mechanical cause of thrombosis, a lupus anticoagulant or other antiphospholipid antibodies, vasculitis or a vascular anomaly (e.g. Klippel-Trenaunay), other autoimmune disorders (including heparin-induced thrombocytopenia) or prior bone marrow or liver transplantation. A DVT or PE was categorized as objectively diagnosed when confirmed by venography, pulmonary angiography, compression venous duplex ultrasonography, ventilation/perfusion lung scan interpreted as high probability for PE, computed tomographic pulmonary angiography, magnetic resonance imaging or pathology examination of thrombus removed at surgery. We prospectively selected clinic-based controls from persons undergoing outpatient general medical examinations in 2004–2009 within the Mayo Clinic Divisions of General Internal Medicine and Primary Care Internal Medicine, Department of Internal Medicine, and general internal medicine practises that care for patients (> 10 000 per year) from the upper midwest United States. Additional controls were recruited from the Department of Family Medicine and the Mayo Clinic Sports Medicine Center. Controls were frequency matched on the age group (20–29, 30–39, 40–49, 50–59, 60–69, 70–79 years), sex, state of residence and myocardial infarction(MI)/stroke status distribution of the cases, and had no previous diagnosis of VTE or superficial vein thrombosis. Potential controls with active cancer, antiphospholipid antibody syndrome, rheumatologic or other autoimmune disorder, or prior bone marrow or liver transplant, were excluded.

For consenting cases and controls, we collected data by in-person questionnaire and medical record review on prior history of VTE and date(s) of VTE (for cases), other thrombotic events and dates of thrombosis (e.g. stroke, MI and peripheral artery thrombosis), current medications and prior exposures (and dates of exposures) that are VTE risk factors, including surgery, hospitalization for acute medical illness, trauma/fracture and neurological disease with leg paresis, and for women, oral contraceptives, obstetric history and hormone therapy. Cases and controls provided informed consent to use of a venous whole blood sample for leukocyte genomic DNA extraction, storage and use for research addressing the genetics of VTE. The study was approved by the Mayo Clinic Institutional Review Board.

Candidate gene selection

Candidate genes were selected from three electronic databases (i.e. the Kyoto Encyclopedia of Genes and Genomes [Complement & Coagulation Cascades]; the NHLBI Program for Genomic Applications; and the University of Washington FHCRC Variation Discovery Resource [Innate Immunity]) that annotate the anticoagulant, procoagulant, fibrinolytic and/or innate immunity pathways. In general, we focused on: platelet, monocyte, neutrophil and endothelial cell agonists, receptors, ligands, signal transduction and adhesion molecules, granule contents and effectors; plasma proteases (procoagulant, anticoagulant, fibrinolytic and complement [including cofactors and receptors]) and inhibitors (e.g. serine protease inhibitors); matrix metalloproteases; inflammatory cytokines and receptors (including leukotrienes and receptors); estrogen, progesterone and androgen receptors, co-regulators and enzymes related to estrogen metabolism; important enzymes for catechol, homocysteine, thromboxane A2 and prostacyclin biosynthesis and metabolism; and 3-Hydroxy-3-Methylglutaryl Coenzyme A (HMG-CoA) reductase.

SNP identification and selection

To identify SNPs for the custom 16 720 bead Illumina Infinium (14 612 SNPs) genotyping panel, we used genotypes from the genome-wide genotyping projects HapMap (http://www.hapmap.org) and Perlegen. Additionally we used genotypes from two gene resequencing programs: Seattle SNPs (http://pga.mbt.washington.edu/) and NIEHS SNPs (http://egp.gs.washington.edu/). To determine the HapMap and Perlegen SNPs for each of the 764 candidate genes, we picked SNPs 10 kb upstream and downstream of each gene. Our gene and SNP coordinates were based on NCBI build 35 and dbSNP build 125. If the gene had been resequenced in Seattle SNPs or NIEHS SNPs, we used genotypes from those sources as well. At the time of our SNP selection, Seattle SNPs had resequenced 205 of our candidate genes and NIEHS SNPs had resequenced 20.

A hierarchical approach was used for SNP selection. To select ld tagSNPs, we ran ldSelect [7] on each candidate gene for the Caucasian samples within each public genotype source (HapMap, Perlegen, Seattle and NIEHS). We used an r2 of 0.8 and a minor allele frequency (MAF) cut-off of 0.05 with the exception of one gene (GP9– Entrez gene id 2815) where we used a MAF cut-off of 0.01. To determine the best source of genotypes for each gene where a gene had been resequenced we took the source with the higher number of ld bins for the Caucasian samples after removing bins with no tag SNP meeting the minimum Illumina design score (design score = 0.4). If each source (e.g. HapMap, Seattle SNPs) had the same number of bins, we used HapMap as the best source because of its higher number of samples (60 unrelated Caucasian samples). HapMap was chosen as the best source for 626 genes, Seattle for 88 genes, Perlegen for 26 genes and NIEHS for six genes. Eighteen genes had no tagSNPs because no SNPs had a MAF ≥ 0.05 or met the minimum acceptable Illumina design score or were not mapped to the genome reference assembly. If possible, we selected additional tag SNPs when the bin was large; if there were ≥ 30 or ≥ 10 SNPs in an ld bin, we chose three and two tag SNPs per bin, respectively. After completing this process, 12 577 ld tagSNPs, representing 12 073 ld bins, were selected for genotyping. Four hundred and eighty-five bins were dropped due to low design scores. We next selected non-synonymous coding (nc)SNPs with a MAF ≥ 0.005, which met the minimum acceptable Illumina design score. For Perlegen and HapMap ncSNPs, we used the MAF for Caucasian samples. For ncSNPs not in those sources, we used the MAF for Caucasian samples in the Illumina annotation. This added 675 ncSNPs to the panel. Next, we added eight SNPs that had been identified in the literature and with collaborators. To test for population stratification, we included 557 ancestry informative markers [8]. Finally, to fill out the panel, we added 795 additional SNPs from our genes and with MAF ≥ 5%, resulting in a total of 14 612 SNPs. A list of the candidate genes and number of selected SNPs for each gene are provided in Table S1.

Genotyping and quality control

Leukocyte genomic DNA was extracted, quantified and diluted to the appropriate concentration for Illumina Infinium iSelect genotyping on all samples collected. Controls included 2% sample replicates and a CEPH (Centre du Polymorphisme Humain) trio for quality control. In addition, case and control DNA sample addresses were randomly assigned across both the 96-well plate as well as the 12-address iSelect BeadChip, insuring approximately equal numbers of case and control DNA samples by each strata to avoid potential plate and chip effects, respectively. Genotyping results from high-quality control DNA (SNP call rate ≥ 95%) were used to generate a cluster algorithm.

Statistical analyses

The primary outcome was VTE status, a binary measure. The covariates were age at interview or blood sample collection, sex, stroke and/or MI status, and state of residence (Table 1). To adjust for population stratification, we performed the multidimensional scaling (MDS) analysis option in plink v 1.07 (http://pngu.mgh.harvard.edu/purcell/plink/) to identify outliers in our population [9] using the ancestry informative markers. We tested for association between each SNP and VTE using unconditional logistic regression, adjusting for age, sex, stroke/MI status, and state of residence. The analyses were corrected for multiple comparisons using an extension of false discovery rates [10,11]. The false discovery rate is an analogue measure of the P-value that takes into account the number of statistical tests and estimates the expected proportion of false-positive tests incurred when a particular SNP is significant. All analyses were performed using plink v 1.07 [9]. Quantile-quantile (QQ) plots of observed −log10P-values for VTE association vs. the expected −log10P-values under the null hypothesis of no association were generated to display the potential significant associations [12], and to calculate the genomic inflation factor λ as a check for over-dispersion of the test statistics [13]. Penalized logistic regression models were used to determine possible interaction between the statistically significant SNPs [14].

Table 1.   Demographic and clinical characteristics by case–control status
CharacteristicCase n = 1488Control n = 1439P-value
Patient age, mean ± SD, years54.7 ± 16.355.5 ± 15.70.1796
Female, n (%)751 (50.5)754 (52.4)0.2970
Stroke or myocardial infarction, n (%)283 (19.0)149 (10.4)< 0.0001
State of residence, n (%)
 Minnesota619 (41.6)795 (55.2)< 0.0001
 Other states869 (58.4)644 (44.8)

Population attributable risk (PAR) was estimated for each genotype, which defines the percentage of the total risk for VTE due to genetic effect of that particular genotype [15]:

  • image

where p is the incidence of risk genotype associated with VTE among control subjects, and OR is the odds ratio associated with risk genotype. We used odds ratios from the additive genetic model adjusted for age at blood draw, gender, MI/stroke status, and state of residence. The group PAR was calculated as inline image on the basis of the individual PARi of each associated genotype assuming an additive genetic model and no multiplicative interaction among the genotypes (i.e. assuming independence between SNPs). The joint PAR was calculated asinline image, assuming multiple loci, inline image as the fraction of cases for each associated genotype, and inline image as the individual OR for each associated SNP or genotype calculated under the full logistic regression model [16]. The PAR was calculated, both unadjusted and adjusted, assuming an additive genetic model for the genotypes. The AttribRisk Splus function (glm function, binomial error) was used to estimate this PAR, unadjusted and adjusted for covariates using jackknife estimates for the standard error [17].

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Of the 3131 unique subjects recruited, 204 were excluded due to the following overlapping reasons: study exclusion criteria (n = 78), genotype issues (n = 40; 22 failed genotyping and 18 had a genotype call rate < 95%), mislabeled samples (n = 32), and non-European (i.e. African- or Asian-American) race (n = 20). Thirty-four subjects were removed due to relatedness using IBS clustering in plink [9]. Multidimensional scaling plots showed two subjects with race discrepancy (available in Fig. S1). After removal of these subjects, no evidence of population stratification was found. After these exclusions, a total of 2927 individuals (1488 VTE subjects [51%], 1439 controls; 51% women) were included in the analyses. The study population demographic and clinical characteristics by case status are presented in Table 1. Among the cases, the distribution of symptomatic VTE by event type was DVT only (n = 744; 50%), PE only (n = 390; 26.2%), and both DVT and PE (n = 354; 23.8%).

Of the 14 612 SNPs submitted to Illumina from 764 genes within the anticoagulant, procoagulant, fibrinolytic and innate immunity pathways (Table 2), 1585 SNPs (covering 1100 linkage disequilibrium [LD] bins) failed manufacture. Of the remaining 13 027 successfully manufactured SNPs (Illumina) and four SNPs on TaqMan, 735 SNPs were excluded due to poor performance (n = 554), a MAF < 0.005 (n = 127) or a call rate < 0.95 (n = 54), leaving 12 296 SNPs (covering 10 456 bins) for the association analysis. Prior to exclusion, concordance between the genotype results for FV Leiden (F5 rs6025) and data obtained clinically for cases was 100% (218/218) between positive cases and 99% (446/450) between negative cases.

Table 2.   Number of linkage disequilibrium bins after illumina infinium custom genotyping SNP selection, design, manufacture and assay by pathway
Pathway (number of genes)*After selectionAfter designAfter manufactureAfter QC% Total lost
  1. *Fifteen genes out of 764 did not have SNPs after quality control (QC).

Anticoagulant (16)23423221421010.3
Procoagulant (75)151614661354130014.2
Fibrinolytic (23)43541137335618.2
Innate immunity (635)10 37399649032859017.2

Using an additive genetic model and adjusting for age, gender, stroke/MI status and state of residence, and a false discovery rate (q-value) < 0.05, one or more SNPs within ABO, F2, F5, F11, KLKB1, SELP and SCUBE1 were significantly associated with VTE (Table 3 and Fig. 1). We confirmed the association between VTE and FV Leiden (F5 rs6025, OR = 3.40, P-value = 3.07 × 10−22), but after controlling for F5 rs6025, F5 rs6687813 was not associated with VTE. We also confirmed the association between VTE and ABO non-O blood type (ABO rs8176719, OR = 1.47, P-value = 5.68 × 10−12), and found a novel association with ABO rs2519093 (OR = 1.69, P-value = 8.08 × 10−16) that remained significant after controlling for non-O blood type (OR = 1.52, P-value = 1.35 × 10−6; Fig. 2). Finally, we found an association between VTE and prothrombin G20210A (F2 rs1799963, OR = 2.46, P-value = 1.69 × 10−6). An association analysis using a dominant genetic model gave similar results (data not shown).

Table 3.   Association* of candidate gene SNPs with venous thromboembolism
SNPGeneChromosomeMinor allele (MAF)nOdds ratio (95% CI)P-valueQ-value
  1. *Additive model; adjusted for age, gender, stroke/myocardial infarction and state of residence. Minor allele frequency (MAF). SNP is on KLKB1 gene but within 10 kb of F11. §SNP is on F11 gene but within 10 kb of KLKB1.

rs6025F51A (0.066)29273.40 (2.65–4.35)3.07E-223.77E-18
rs6687813F51A (0.103)29262.13 (1.78–2.56)4.66E-162.86E-12
rs2519093ABO9A (0.243)29071.68 (1.48–1.91)8.08E-163.31E-12
rs505922ABO9G (0.402)29261.49 (1.33–1.66)1.52E-124.68E-09
rs687289ABO9A (0.403)29241.48 (1.33–1.65)3.03E-127.46E-09
rs8176719ABO9G (0.419)29001.47 (1.32–1.64)5.68E-121.16E-08
rs643434ABO9A (0.421)29231.44 (1.30–1.61)3.39E-115.96E-08
rs630014ABO9A (0.421)29270.75 (0.67–0.84)2.67E-070.00041
rs3087505KLKB14A (0.099)29270.63 (0.52–0.75)4.34E-070.000593
rs660340ABO9A (0.408)29270.77 (0.69–0.85)1.13E-060.001389
rs659104ABO9A (0.408)29270.77 (0.69–0.85)1.28E-060.001425
rs1799963F211A (0.025)28912.46 (1.70–3.54)1.69E-060.001732
rs3917862SELP1G (0.075)29241.60 (1.31–1.97)6.13E-060.005562
rs4253399§F114C (0.409)29271.28 (1.15–1.43)6.33E-060.005562
rs4525F51G (0.245)29240.77 (0.68–0.87)2.34E-050.01917
rs4524F51G (0.245)29270.77 (0.68–0.87)2.51E-050.01932
rs10158595F51A (0.218)29270.76 (0.67–0.87)3.03E-050.02191
rs6032F51G (0.245)29160.77 (0.68–0.87)3.35E-050.02253
rs5759224SCUBE122G (0.113)28641.43 (1.21–1.70)3.48E-050.02253
rs2213867F51G (0.245)29260.78 (0.69–0.88)4.37E-050.02685
image

Figure 1.  Manhattan plot of association results between VTE and candidate gene SNPs by chromosome. The x-axis displays the chromosomes and the y-axis displays the –log10 P-values. The horizontal line represents the –log10 of the Bonferroni corrected P-value (−log[4.0E-06] = 5.40). The significant results (Bonferroni corrected P-value < 4.0E-06) are labeled by gene name and SNP rs number. The r2 value between the two F5 SNPs is 0. The r2 values between the ABO SNPs are as depicted in details in Fig. 2.

Download figure to PowerPoint

image

Figure 2.  Haploview linkage disequilibrium plot of ABO SNPs (n = 17). Blocks represent SNPs in high linkage disequilibrium. Greater color intensity corresponds to a higher level of linkage disequilibrium given by D’. The value inside each cell represents linkage disequilibrium given by r2. Blue rectangles outline the ABO blood group SNPs (homozygous deletion on rs8176719 determines O blood type). Green rectangles outline ABO SNPs significantly associated with VTE. ABO SNPs from HapMap build 36.3 are displayed above the Haploview plot.

Download figure to PowerPoint

In additional analyses stratifying by sex, FV Leiden (positive/negative for carrier of F5 rs6025 A allele), ABO blood type-O (yes/no based on homozygous deletion in ABO rs8176719 where the homozygous deletion results in ABO blood type-O) and prothrombin G20210A (positive/negative for carrier of F2 rs1799963 A allele), similar F5 and ABO SNPs were associated with VTE for both women and men, while F2 rs1799963 was no longer significant for either gender; SNPs within an additional gene (LY86) were significantly associated with VTE among women (Table S2). The odds of VTE appeared to be higher for FV Leiden among men. Among FV Leiden non-carriers, similar ABO and F2 SNPs, and an additional KLKB1 SNP, were associated with VTE (Table S3). Among FV Leiden carriers, SNPs within PRKCB1 and CD44 were marginally associated with VTE, possibly due to small sample size. Among persons with ABO blood type non-O (ABO rs8176719 G), SNPs within F5, F11 and KLKB1, and an additional gene (GFRA1), were associated with VTE (Table S4). Of note, the novel ABO rs2519093 remained significantly associated with VTE in this patient subset. Only the FV Leiden mutation was significantly associated with VTE among those with the ABO blood type-O (ABO rs8176719 G homozygous deletion). Among prothrombin G20210A non-carriers, SNPs within ABO, F5, F11, KLKB1, SCUBE1 and SELP were significantly associated with VTE (Table S5). The sample size of prothrombin G20210A carriers was insufficient for meaningful analysis. We also performed sex-chromosome analysis and no significant results were identified (data not shown). Q–Q plots of the −log10P-values for SNP associated with VTE under different analyses showed no evidence of over-dispersion in our samples (λ = 1.0; available in Fig. S2).

The individual, joint and group population attributable risks (PARs) were calculated for the risk genotypes for FV Leiden (F5 rs6025), prothrombin G20210A (F2 rs1799963), ABO blood type non-O (ABO rs8176719), and the novel ABO rs2519093 (Table 4). The unadjusted and adjusted individual and joint PAR values were very similar. The highest PAR value was from the ABO blood type non-O, followed by ABO rs2519093, FV Leiden, and prothrombin G20210A. The PAR values were very similar between the joint and group estimation methods when either ABO rs8176719 or ABO rs2519093 was included. When both ABO rs8176719 and ABO rs2519093 were included in the group PAR calculation, the method yielded an inflated value of 0.47 compared with the joint method (joint PAR = 0.40).

Table 4.   Attributable risk (AR) for significant single nucleotide polymorphisms (SNPs) assuming an additive genetic model for each SNP
Single nucleotide polymorphismUnadjustedAdjusted*
AR95% CIAR95% CI
  1. *Adjusted for age at blood sample collection, male gender, myocardial infarction/stroke status, and Minnesota residence. Assuming independence between SNPs (i.e. a non-multiplicative interaction between the genotypes).

Individual
 F5 rs6025 (factor V Leiden)0.13(0.11, 0.16)0.14(0.11, 0.16)
 F2 rs1799963 (prothrombin G20210A)0.04(0.02, 0.06)0.04(0.02, 0.06)
 ABO rs8176719 (ABO blood type non-O)0.28(0.20, 0.36)0.28(0.18, 0.38)
 ABO rs25190930.23(0.16, 0.29)0.25(0.16, 0.30)
Joint16
 F5 rs6025, F2 rs1799963, ABO rs81767190.40(0.33, 0.47)0.40(0.33, 0.48)
 F5 rs6025, F2 rs1799963, ABO rs25190930.35(0.29, 0.41)0.35(0.29, 0.40)
 F5 rs6025, F2 rs1799963, ABO rs8176719, ABO rs25190930.40(0.32, 0.47)0.40(0.33, 0.50)
Group
 F5 rs6025, F2 rs1799963, ABO rs8176719 0.35
 F5 rs6025, F2 rs1799963, ABO rs2519093 0.33
 F5 rs6025, F2 rs1799963, ABO rs8176719, ABO rs2519093 0.47

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

While we confirmed associations between FV Leiden (F5 rs6025), prothrombin G20210A (F2 rs1799963) and ABO blood type non-O (ABO rs8176719) and VTE [18–22], our most notable finding was a strong, independent association between VTE and a novel ABO SNP (rs2519093). Of further note, the odds ratio and significance level of ABO rs2519093 exceeded that of ABO blood type non-O blood group and approached that of FV Leiden (Table 3). Given the ABO rs2519093 false discovery rate (q = 3.3E-12) and the total number of SNPs significantly associated with VTE in this analysis (n = 20; Table 3), the estimated number of falsely positive SNPs would be 20 x (3.3E-12), indicating that the observed association of ABO rs2519093 with VTE is extremely unlikely to be falsely positive. Moreover, while the magnitude of VTE risk associated with ABO rs2519093 is relatively low (OR = 1.68), given the high minor allele frequency (MAF = 0.24), this SNP accounts for a higher proportion of VTE disease burden in this population than do the FV Leiden and prothrombin G20210A mutations. Indeed, the joint population-attributable risk for FV Leiden, prothrombin G20201A, ABO non-O blood group and ABO rs2519093 (PAR = 0.40; Table 4) suggests that these four SNPs account for a large proportion of VTE among non-Hispanic adults of European ancestry.

ABO blood group molecules are expressed on a wide variety of human tissues and are present on most epithelial and endothelial cells. In addition, FVIII and von Willebrand factor (VWF) both undergo extensive post-translational modification, including glycosylation by the ABO blood group-encoded glycosyltransferases [23]. Such glycosylation protects VWF from proteolysis by ADAMTS13 [24]. The ABO phenotype correlates with plasma levels of FVIII and VWF; individuals with type O blood group have about 25% lower plasma FVIII and VWF levels [21,25]. Increased plasma FVIII and VWF levels are associated with an increased risk of VTE [26,27], and non-O blood group type has also been associated with an increased VTE risk that is independent of FVIII [21]. ABO rs2519093 is a tag SNP located in intron 1 and is not in linkage disequilibrium with the ABO non-O blood type (ABO rs8176719) located in exon 6. This intronic region does not affect RNA splicing or harbour any suppressor RNA elements, nor is it close to any other ABO variants with known function, including variants in ABO exons 6 and 7 that have been variably associated with VTE risk [22,28–31]. In a genome-wide association study and subsequent replication studies, ABO rs657152, rs505922 and rs630014 were initially associated with VTE but not after adjusting for ABO blood group [32]. Further studies will be required to determine whether ABO rs2519093 is in linkage disequilibrium with other ABO functional variants.

Two tag SNPs located within 10 kb of KLKB1 (prekallikrein [PK]; rs3087505) and F11 (procoagulant FXI; rs4253399) were strongly associated with incident VTE, with very low false discovery rates but in opposite directions; the odds of VTE were decreased by about 35% and increased by about 30%, respectively, for KLKB1 rs3087505 and F11 rs4253399 (Table 3). PK, FXII and high-molecular-weight kininogen (HMWK) are components of the ‘contact activation’ procoagulant pathway, and together with FXI, FVIII and FIX constitute the intrinsic coagulation pathway. Animal model studies suggest that the intrinsic coagulation pathway is involved in thrombosis. FXII-, FXI-, FIX-, HMWK- and bradykinin B2 receptor-deficient mice have reduced or absent thrombus formation and/or stabilization at the site of vascular injury [33–38]. Increased FXI levels are associated with VTE [39], and genetic variants within F11 and KLKB1 have been associated with increased FXI levels [20,40] and VTE [19,40,41]. KLKB1 rs3087505 significantly protected against VTE among FV Leiden and among prothrombin G20210A non-carriers (Tables S3 and S5) and individuals with ABO blood type non-O (Table S4), while F11 rs4253399 increased VTE risk among prothrombin G20210A non-carriers and ABO blood type non-O individuals (Tables S4 and S5).

One tag SNP within SELP (rs3917862) was associated with a 1.6-fold increased risk of VTE at a false discovery rate q = 0.006 (Table 3). SELP encodes for P-selectin, an adhesion and signaling glycoprotein that is stored in platelet alpha-granules and endothelial cell Weibel-Palade bodies. P-selectin overexpression in mice leads to a procoagulant state [42]. Cell surface P-selectin expression and soluble P-selectin concentration are partly controlled by SELP genetic variation [43,44]. Increased soluble P-selectin is associated with incident and recurrent VTE [45,46]. SELP is located immediately upstream from F5, and associations between SELP haplotypes and VTE have been attributed to high linkage disequilibrium (LD) with FV Leiden [47]. While SELP rs3917862 was not in high LD with F5 rs6025 (FV Leiden; R2 = 0.12, D’ = 0.37), the association of rs3917862 with VTE was no longer significant (P = 0.3) after controlling for FV Leiden, possibly due to small sample size.

One tag SNP within SCUBE1 (rs5759224) was associated with VTE, both overall (OR = 1.43, P = 3.48E-05, q = 0.022; Table 3) and among prothrombin G20210A non-carriers (OR = 1.46, P = 1.95E-05, q = 0.016; Table S5). SCUBE1 encodes for signal peptide-CUB domain-EGF-related protein 1, a member of the epidermal growth factor superfamily synthesized within megakaryocytes, stored in platelet alpha granules and expressed on the platelet surface upon platelet stimulation [48]. SCUBE1 also is expressed on the cell surface of vascular endothelial cells [49]. Plasma SCUBE1 is significantly increased in acute coronary syndromes and acute ischemic stroke [50], but has not been studied in VTE.

Two tag SNPs within LY86 (rs1073897 and rs9328375) were significantly associated with VTE among women (Table S2); both SNPs were protective with highly significant P- and q-values. LY86 encodes for MD-1, a secretory molecule that binds and regulates the cell-surface expression of RP105, which is a homolog to toll-like receptor 4 (TLR4) [51]. The RP105/MD-1 complex is widely expressed on antigen-presenting cells and downregulates the proinflammatory response to the gram-negative bacterial cell wall endotoxin, lipopolysaccharide (LPS), by inhibiting LPS-induced TLR4 signaling [52]. VTE is associated with recent urinary tract infection (UTI) [53], and UTI is most commonly caused by gram-negative bacteria and much more frequent in women [54]. We hypothesize that genetic variation in MD-1 downregulates the individual inflammatory response to UTI caused by gram-negative bacteria, and protects against VTE via downregulation of the procoagulant response to inflammation [55].

Among FV Leiden (F5 rs6025) carriers, SNPs within PRKCB1, CD44 and ITPR1 were significantly associated with VTE but the false discovery rates were all ≥ 0.2 (Table S3), suggesting these associations may be falsely positive. While several GFRA1 SNPs were significantly associated with VTE among persons with ABO blood type non-O with false discovery rates ≤ 0.05 (Table S4), no clear mechanism for GFRA1 in the etiology of VTE is apparent. Thus, the association of these GFRA1 SNPs with VTE remains tentative until confirmed in replication studies.

Our study was limited to non-Hispanic Mayo Clinic outpatients of European ancestry who predominantly resided in the midwestern United States. Based on analyses of ancestry informative markers, our findings are unlikely to reflect confounding by population stratification. However, our findings may not be generalizeable to populations with different ancestry. The candidate genes for each pathway were selected for this study based on information available in 2005. Consequently, we may have missed important associations between VTE and sequence variation(s) within other genes subsequently identified as important components of one or more of these four pathways. While we have reported here the main effects of candidate gene SNPs on VTE risk, future analyses addressing the effects of gene-gene and gene-environment interactions will be important.

In summary, we have confirmed the association of FV Leiden, prothrombin G20210A and ABO blood type non-O with VTE, and identified a novel ABO intron 1 tag SNP that is strongly and independently associated with VTE; together, these SNPs account for 45% of VTE within this population. In addition, we identified additional novel SNPs within KLKB1, F11 and SELP, and within the novel genes SCUBE1 and LY86 that also appear to be strongly associated with VTE. These findings lend further support to the hypothesis that individual genetic variation in genes encoding for components of the procoagulant, anticoagulant, fibrinolytic and innate immunity pathways predispose to incident VTE.

Addendum

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

A. Heit designed and performed the research, collected and interpreted data, and wrote the manuscript. J. M. Cunningham performed the research, analyzed and interpreted data, and participated in manuscript preparation. T. M. Petterson performed the research, collected data, analyzed and interpreted data, performed statistical analyses, and participated in manuscript preparation. S. M. Armasu analyzed and interpreted data, performed statistical analyses, and participated in manuscript preparation. D. N. Rider contributed analytical tools, analyzed and interpreted data, and participated in manuscript preparation. M. de Andrade designed and performed the research, analyzed and interpreted data, performed statistical analyses, and participated in manuscript preparation.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Funded, in part, by grants from the National Institutes of Health (HL83141), US Public Health Service, and the Mayo Foundation. We thank C. E. Regnier, J. L. Alkhamis, L. M. Heimer, R. M. Weatherly, R. A. Mueller, A. Xue, R. A. Miller, S. Chiappa Windebank, C. A. Hilker, J. J. Larson, E. N. Jeavons and A. F. Beauseigneur for their excellent technical assistance. All were compensated as part of their regular duties.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Addendum
  8. Acknowledgements
  9. Disclosure of Conflict of Interests
  10. References
  11. Supporting Information

Figure S1. Comparative proteome network assembled by Ingenuity Pathways Analysis.

Table S1. List of primers used for real time PCR analysis.

FilenameFormatSizeDescription
JTH_4272_sm_Supplemental-Figures1-2.doc442KSupporting info item
JTH_4272_sm_Supplement-Table-1.xls162KSupporting info item
JTH_4272_sm_Supplemental-Table-2.doc63KSupporting info item
JTH_4272_sm_Supplemental-Table-3.doc66KSupporting info item
JTH_4272_sm_Supplemental-Table-4.doc49KSupporting info item
JTH_4272_sm_Supplemental-Table-5.doc63KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.