Thrombomodulin gene polymorphisms or haplotypes as potential risk factors for venous thromboembolism: a population-based case–control study

Authors


John A. Heit, Hematology Research, Stabile 610, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, USA.
Tel.: (507) 284 4634; fax: (507) 266 9302; e-mail: heit.john@mayo.edu

Abstract

Summary.  Dysfunction of the protein C anticoagulant system is associated with venous thromboembolism (VTE) and thrombomodulin (TM) is a critical cofactor within the protein C system. The aim of this study was to test the hypotheses that polymorphisms or haplotypes within the TM gene are common risk factors for VTE. We screened the TM putative promoter, exon and 3′-untranslated region for sequence variations in a random sample (n = 266) of consecutive idiopathic, objectively confirmed non-Olmsted County VTE patients referred to the Mayo Clinic. We then genotyped a sample of Olmsted County, MN residents with a first lifetime, objectively confirmed VTE in the 25-year period, 1966–90 (n = 223), and a sample of Olmsted County residents without VTE (n = 237) for polymorphisms either discovered in the screening population or previously published, and tested for an association of VTE with TM genotype or haplotypes using unconditional logistic regression and generalized linear models, respectively. We also genotyped these Olmsted County cases and controls at 20 ‘null’ genetic maker loci and tested for population admixture. Nine novel and three previously described mutations were identified in the screening population. Mutations within the TM promoter, EGF1-5, serine/threonine-rich, transmembrane, and cytoplasm regions were absent or uncommon. TM845G→A (Ala25Thr; lectin region), TM2136T→C (Ala455Val; EGF6 region), TM2470C deletion (3′-untranslated region), and 4363A→G (3′-flanking region) were more common, but were not associated with VTE by genotype or haplotype. Null genetic marker allele frequencies did not differ significantly among cases and controls. We conclude that polymorphisms or haplotypes within the TM gene are not common risk factors for incident VTE.

Introduction

Family-based studies have established that venous thromboembolism is, at least in part, an inherited disease, with estimated heritabilities of approximately 60%[1,2]. The mode of venous thromboembolism inheritance is probably complex [2]. Moreover, family-based and twin studies have established that over 25 plasma hemostasis-related analytes (traits) both correlate with thrombosis and are heritable [3–7].

Thrombomodulin is a transmembrane protein that is constitutively expressed on the luminal surface of vascular endothelial cells [8]. Thrombomodulin has at least three independent anticoagulant activities that, if altered, could predispose to venous thromboembolism: thrombomodulin catalyzes thrombin activation of protein C to activated protein C (APC); thrombomodulin binds and alters thrombin substrate specificity such that thrombin-mediated clotting of fibrinogen and activation of platelets and procoagulant factors V, VIII, XI, XIII are inhibited; and thrombomodulin catalyzes the inhibition of thrombin by antithrombin. Thrombomodulin consists of a large N-terminal extracellular region, a single transmembrane segment, and a short cytoplasmic tail [9,10]. The extracellular region is comprised of an N-terminal lectin-like domain followed by six tandem repeats of epidermal growth factor (EGF)-like domains, and a glycosylated (chondroitin sulfate) serine/threonine-rich domain. Each thrombomodulin anticoagulant activity depends on specific thrombomodulin structures [11,12]. The thrombin-binding region has been localized to the fifth and sixth EGF-like domains, while the fourth EGF-like domain is required for protein C binding to the thrombin–thrombomodulin complex. The serine/threonine-rich spacer region is required for both thrombin binding and thrombomodulin cofactor activity for membrane-associated thrombomodulin. The chondroitin sulfate may stabilize thrombin binding to thrombomodulin, possibly by interacting with the thrombin apolar region.

Animal model data suggest that thrombomodulin dysfunction or deficiency is associated with a prothrombotic disorder. Transgenic mice with a thrombomodulin mutation corresponding to human TMGlu387Pro have a prothrombotic disorder [13]. This amino acid change is located between the interdomain loop of the 4th and 5th EGF-like domains and abolishes the ability of soluble thrombomodulin to catalyze in vitro thrombin activation of protein C to APC. Mice with thrombomodulin deficiency limited to the vascular endothelium die shortly after birth as a result of a consumptive coagulopathy that can be prevented by warfarin anticoagulation [14].

Based on the important antithrombotic role of thrombomodulin, we hypothesized that polymorphisms within the thrombomodulin gene that alter thrombomodulin expression and/or impair anticoagulant function could predispose to venous thromboembolism. To test this hypothesis, we screened a large population of unrelated patients with idiopathic, objectively confirmed, deep vein thrombosis or pulmonary embolism to identify mutations within the thrombomodulin gene putative promoter, the 5′-flanking region, the exon, the 3′-untranslated region, and a part of the 3′-flanking region. We then performed a population-based case–control study to test discovered or published polymorphisms and their haplotypes for an association with venous thromboembolism. We also genotyped our cases and controls for several ‘null’ genetic markers to achieve genomic control of population stratification and admixture.

Methods

Study populations

We first wished to screen a sufficient number of DNA samples from a population of unrelated patients with idiopathic venous thromboembolism to be confident that we identified important mutations within the thrombomodulin gene among patients with venous thromboembolism. A sample size of 227 provides 90% power to detect a mutation that has an attributable risk of 1% and a relative risk of > 2 for both dominant and recessive models [15]. To further insure adequate power, we screened 266 patients randomly sampled from 645 consecutive non-Olmsted County residents who were referred to either the Mayo Clinic Special Coagulation Laboratory or the Mayo Clinic Thrombophilia Center for a clinical suspicion of an underlying thrombophilia and who had idiopathic, objectively confirmed, deep vein thrombosis or pulmonary embolism. Beginning in 1996, these patients were approached for consent to collect an extra blood sample for leukocyte genomic DNA extraction, storage and utilization for subsequent studies of thrombophilia; over 95% consented.

Associations with venous thromboembolism were assessed in a population-based case–control study. The cases were selected from the inception cohort of 1269 Olmsted County, Minnesota residents with an objectively confirmed, first-lifetime, deep vein thrombosis or pulmonary embolism during the 25-year period, 1966–90 [16]. A venous thromboembolism was categorized as objectively confirmed when symptoms and signs of deep vein thrombosis or pulmonary embolism were present and venography or pulmonary angiography, computed tomography scan, magnetic resonance imaging, impedance plethysmography, compression duplex ultrasonography, or a lung scan was positive, or pathology examination of material removed at surgery or autopsy was interpreted as venous thrombosis. Because of the possibility of ascertainment bias, we excluded from this study the 489 patients whose venous thromboembolism was discovered at autopsy. Of the 780 cases not discovered at autopsy, 478 were deceased. We approached the remaining 302 cases that were alive and available to provide a blood sample for DNA extraction, and obtained DNA from 223 (74%). Using an enumeration of the Olmsted County population provided by the Rochester Epidemiology Project [17], we previously identified the Olmsted County resident without venous thromboembolism who most closed matched each case on age, sex and calendar year (date of venous thromboembolism for cases; date of closest medical visit for matched controls) [18]. Of the 879 previously identified controls, 366 were deceased. Sampling to obtain DNA from the remaining 513 controls was the same as for the cases, and we obtained DNA from 237 (46%) of them. It was not possible to maintain the individual matching because DNA was not always available from each case–control pair. All patients were non-Hispanic, White Americans in keeping with the racial composition of the region. The study protocol was approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards.

Measurements

Either denaturing high-pressure liquid chromatography (DHPLC, Transgenomic, Omaha, NE) with fluorescence detection of 6-carboxyfluorescein (FAM)-labeled primers [19] or direct fluorescence-based automated sequencing were used to screen stored leukocyte genomic DNA for mutations within the thrombomodulin gene putative promoter region, the 5′-flanking region, the exon, the 3′-untranslated region, and a part of the 3′-flanking region. Both upstream and downstream DNA sequencing were performed to characterize the exact sequence variation detected by DHPLC, and to screen genomic DNA regions that were unsuitable for DHPLC analysis. Genotyping of cases and controls was performed using restriction fragment length polymorphism, real-time fluorescence polymerase chain reaction (LightCycler, Roche Molecular Biochemicals, Mannheim), or DHPLC. Appropriate wild-type, heterozygous mutant, and no-DNA controls were included in each assay.

Because we first sought to identify mutations with clinical relevance for stratifying patient risk for incident venous thromboembolism, we restricted our initial analyses to polymorphisms (e.g. mutations with geqslant R: gt-or-equal, slanted 1% carrier frequency) identified in the screening population. We first concentrated on missense (non-synonymous) polymorphisms [20], and categorized such polymorphisms into classes of increasing chemical dissimilarity between encoded amino acids (e.g. conservative, moderately conservative, moderately radical, radical and ‘stop’ or nonsense mutations) according to the Grantham scale [20–22]. To further assess the potential importance of these polymorphisms on overall protein function, we assessed the degree of amino acid evolutionary conservation across species [20]. Disease-causing mutations are significantly more likely to occur at amino acid residues that are perfectly preserved in evolution [23,24]. We then studied polymorphisms affecting 5′ or 3′ regulatory regions. Finally, we performed haplotype analyses because certain haplotypes may be in linkage dysequilibrium with 5′ or 3′ regulatory sequences not included in our DNA mutation screening strategy [25], and because combinations of missense mutations may be more deleterious to protein function than any single mutation. For haplotype analyses, we genotyped our cases and controls for other thrombomodulin mutations previously reported as possibly being associated with thromboembolism [26–35].

Statistical Analyses

Thrombomodulin polymorphisms identified in the screening population were tested in the Olmsted County cases and controls for an association with venous thromboembolism using unconditional logistic regression. Odds ratios for each mutation are presented both unadjusted and adjusted for age, sex and calendar year. We also tested for an association between venous thromboembolism case status and haplotypes of characterized thrombomodulin mutations using an SPLUS routine developed at the Mayo Clinic [36]. The method uses generalized linear models for exponential family data and handles unrelated individuals without requiring that the entire gene be sequenced. The method assumes the marker genotypes are in Hardy–Weinberg equilibrium, which was tested in the controls for each polymorphism. All three-way combinations and all two-way combinations of the four most common thrombomodulin genotypes were examined.

In genetic association studies, a potential concern is false-positive or false-negative associations because of population stratification. To test for population stratification, we genotyped all cases and controls for a set of 20 ‘null’ genetic markers selected to be independent (i.e. not linked) of each other and the candidate gene under investigation, and not suspected to be genetic risk factors for venous thromboembolism [37–39]. We selected these markers so that, on average, they had the same range of allele frequencies as the thrombomodulin polymorphisms. We also calculated pairwise associations among the null markers to provide additional empiric information on the potential for population stratification. Since the null markers were selected to be independent of each other, any observed association is indicative of population stratification.

Results

Screening the thrombomodulin gene for sequence variation

The PCR primer location, sequence and thrombomodulin gene regions screened for sequence variation are shown in Table 1. Among our screening population of non-Olmsted County patients with idiopathic venous thromboembolism, we identified nine novel mutations and confirmed three previously described mutations; one nucleotide insertion was found in the 3′-flanking region that was present in all patients screened (Table 2). Two novel mutations within the thrombomodulin lectin domain [TM898G→A (Met42Ile), TM1309C→T (Phe179Phe)] were rare or synonymous. Three patients were heterozygous carriers for the previously described TM845G→A (Ala25Thr) lectin domain mutation [28]. No patients were carriers for other previously reported lectin domain mutations [TM847G→C (Ala25Ala) [28], TM954G→C (Glu61Ala) [28], TM2000G→C (Pro136Pro) [30]]. A novel 1614 A→T (Asn281Ile) mutation within the second EGF-like domain was rare, affected a thrombomodulin region (EGF2) without known function, and asparagine-281 was not evolutionarily conserved. No patients were carriers for the previously reported 1926C→G (Arg385Asp) mutation within the fourth EGF-like domain [30]. A novel TM2151G→A (Thr460Ile) mutation, as well as the previously reported TM2136T→C (Ala455Val) mutation [41], were found within the sixth EGF-like domain, an important region for thrombin binding and activation of protein C [11]. The TM2151G→A mutation was rare and threonine-460 was not evolutionarily conserved. Within the serine/threonine-rich domain, two patients were heterozygous carriers for the previously described TM2174G→T (Asp468Tyr) mutation [34,35], and we found two additional novel mutations [TM2215G→A (Pro481Pro) and TM2226C→T (Ser485Phe)]. No patients were carriers for other previously reported mutations [TM2201C→T (Pro477Ser) [28] and TM2220C→T (Pro483Leu) [28,30]] within this domain. No mutations were found within the transmembrane or cytoplasmic domains, nor did any of the patients carry the previously reported 2404G→C, 2407T-insertion frame-shift mutation within the cytoplasmic domain [41,42]. A novel promoter region mutation (TM124C→T) was rare and did not affect a known transcription regulatory site [43]. One novel 5′-region mutation (TM661G→C) was also rare and was located downstream from the reported transcription initiation site [44]. No patients were carriers for other previously reported promoter region mutations (e.g. TM417C→A, TM517G→A, TM540–41GG→AT) [26]. One novel mutation (2470C deletion) affected the thrombomodulin 3′-untranslated region and two additional novel mutations were found in the 3′-flanking region (TM4363A→G and TM4438T→C).

Table 1.   Denaturing high pressure liquid chromatography (DHPLC) or automated fluorescence sequencing polymerase chain reaction primer locations and sequences
Primer location*Primer sequence (5′ to 3′)Method
  1. *Numbering according to Shirai et al. [56] (GenBank accession no. D00210). F = forward primer; R = reverse primer; 6FAM = 6-carboxyfluorescein.

TM66.F .6FAM6FAM-AAAGTGCCCTTGGTGCCGATCADHPLC
TM600.RCAGACACTTCTTGCCGCTG 
TM479.F .6FAM6FAM-CTTCCTTCCTTTTCCCGAACDHPLC
TM943.RCAGTAGCAAGGAAATGACATCG 
TM852.F. 6FAM6FAM-TCCTCAATGCCAGTCAGATCDHPLC
TM1395.RGCGGTGCACATTAGCTGTAA 
TM1211.F. 6FAM6FAM-TTCCTCTGCGAGTTCCACTTDHPLC
TM1609.RAACGCAGAAGTGCTCGCAGA 
TM1570.FGACGCAGTCCTGCAACGASequencing
TM2160.RTCACAGTCGGTGCCAATGT 
TM2096.FGGTACCTTCGAGTGCATCTSequencing
TM2526.RGCTAAGGTGCTTTGGTAGCAAAGCTG 
TM3961.FAAAGGCTAGGTACACAGCTCSequencing
TM4574.RGCCCCAGGAAGGGTTTATTT 
Table 2.   Genotype and carrier frequency for mutations discovered within the thrombomodulin gene among patients with idiopathic venous thromboembolism (n = 266)
Mutation*Amino acid changeThrombomodulin regionHomozygous (n)Heterozygous (n)Carrier frequency (%)
  • *Numbering according to Shirai et al. 1988 [56] (GenBank accession no. D00210). Numbering in parentheses is according to nucleotide 1 = first nucleotide (A) of first signal peptide amino acid (M).

  • Novel mutation.

  • EGF, epidermal growth factor; UTR, untranslated region.

124(− 595)C→TPromoter010.4
661(− 58)G→CPromoter020.8
845(127)G→AA25TLectin031.1
1309(591)C→TF179FLectin010.4
1614(896)A→TN281IEGF2010.4
2136(1418)C→TA455VEGF6179140.6
2151(1433)C→TT460IEGF6010.4
2174(1456)G→TD468YSer/Thr-rich020.8
2470(1752)C Del3′ UTR031.1
4248–9(3530–31)C Insn3′ UTR2660100
4363(3645)A→G3′ UTR1011145.5
4438(3720)T→C3′ UTR020.8

Demographic and baseline clinical characteristics among Olmsted County venous thromboembolism cases and Olmsted County controls

In univariate analyses of demographic and baseline characteristics among Olmsted County venous thromboembolism cases and controls, the median age was lower while the sex distribution was similar (Table 3). Body mass index and oral contraceptive use were higher and congestive heart failure was lower in cases, while current or prior tobacco smoking was similar for cases and controls. Other characteristics previously identified as risk factors for venous thromboembolism [18] were also risk factors among our sample of Olmsted County cases.

Table 3.   Univariate logistic analyses of demographic and baseline characteristics among Olmsted County residents with a first lifetime venous thromboembolism diagnosed in 1966–90 compared to community controls
CharacteristicCasesControlsOR (95% CI)P-value
  • *

    Fisher's exact test.

  • BMI, body mass index; CHF, congestive heart failure.

Age, mean (SD)48.9 (16.60)54.9 (15.85)0.98 (0.97, 0.99)< 0.001
Female, n (%)102 (45.7)116 (49.0)0.88 (0.61, 1.27)0.49
BMI, mean (SD; kg/m2)27.12 (5.28)26.13 (4.27)1.05 (1.01, 1.09)0.03
Event year, mean (SD)1983.5 (5.7)1985.1 (4.1)0.94 (0.90, 0.97)< 0.001
Smoking (ever), n (%)99 (44.4)96 (40.5)1.17 (0.81, 1.70)0.40
Trauma, n (%)46 (20.6)1 (0.4)61.28 (8.38, 448.22)< 0.001
Surgery, n (%)77 (34.5)5 (2.1)24.46 (9.67, 61.85)< 0.001
Hospital/nursing home confinement, n (%)102 (45.7)8 (3.4)24.13 (11.37, 51.21)< 0.001
Active malignancy*, n (%)10 (4.5)0< 0.001
Prior superficial vein thrombosis, n (%)14 (6.3)3 (1.3)5.22 (1.48, 18.43)0.01
Neurologic disease, n (%)9 (4.0)2 (0.8)4.94 (1.06, 23.13)0.04
Varicose veins, n (%)37 (16.6)20 (8.4)2.25 (1.26, 4.02)0.006
CHF, n (%)12 (5.4)24 (10.1)0.51 (0.25, 1.04)0.06
Pregnancy/postpartum, n (%)13 (12.8)3 (2.6)5.50 (1.52, 19.90)0.009
Oral contraceptives, n (%)13 (12.8)6 (5.2)2.68 (0.98, 7.33)0.06

To assess how representative our sample of Olmsted County cases was of the entire Olmsted County venous thromboembolism incidence cohort, we compared the demographic and baseline characteristics among cases with a DNA sample to those without a DNA sample (data not shown). The differences were mostly the result of exclusion of cases discovered at autopsy. However, cases without DNA more often had chronic heart, lung, or neurologic disease or active cancer, and thus were less likely to be alive at follow-up (and available to provide a DNA sample). With the exception of cancer and neurologic disease with extremity paresis, none of these characteristics was an independent risk factor for venous thromboembolism [18], and malignancy-associated venous thromboembolism is unlikely to be inherited. Thus, with the potential exception of neurologic disease, our DNA samples are from those cases that are most likely to have an inherited cause for venous thromboembolism.

Association of thrombomodulin gene polymorphisms with venous thromboembolism

Using our sample of Olmsted County incident venous thromboembolism cases and Olmsted County controls, four polymorphisms (TM845G→A, TM2136C→T, TM2470C deletion, and TM4363A→G) were tested for an association with venous thromboembolism (Table 4). For cases and controls combined, the TM845G→A carrier frequency was approximately 1%. None of the cases or controls were homozygous carriers. The odds ratio (1.07; 95% CI: 0.15, 7.86) for an association of this mutation with venous thromboembolism was not significantly different from 1.0 (P-value = 0.94, adjusting for age, sex, and event year). The TM2136C→T carrier frequency was 35.6% for cases and 33.1% controls. The odds ratio for an association of this mutation with venous thromboembolism was 1.09 (95% CI: 0.73, 1.62; adjusted P-value = 0.67). Only six of the 428 patients were heterozygous for the TM2470C deletion. Four were controls (1.8%) and two were cases (1%). None were homozygous for the TM2470C deletion. The odds ratio, 0.56 (95% CI: 0.10, 3.11), was not significantly different from 1.0 (adjusted P-value =0.51). Finally, about one-third of cases and controls were heterozygous for TM4363A→G, and 6–7% were homozygous. The odds ratio for an association of this mutation with venous thromboembolism was 0.83 (95% CI: 0.56, 1.22; adjusted P-value = 0.34)

Table 4. Thrombomodulin gene mutation genotype distribution in venous thromboembolism cases vs. controls
MutationGroupWild-type n (%)Heterozygous n (%)Homozygous n (%)Total nOR*95% CIP-value
  • *

    Adjusted for age, gender, and calendar year of event

845(127)G→ACase206 (99.0)2 (1.0)02081.070.15, 7.860.94
Control218 (99.1)2 (1.0)0220   
2136(1418)C→TCase143 (64.4)71 (32.0)8 (3.6)2221.090.73, 1.620.67
Control158 (66.9)76 (32.2)2 (0.8)236   
2470(1752)C DelCase206 (99.0)2 (1.0)02080.560.10, 3.110.51
Control216 (98.2)4 (1.8)0220   
4363(3645)A→GCase135 (61.4)71 (32.3)14 (6.4)2200.830.56, 1.220.34
Control134 (57.8)81 (34.9)17 (7.3)232   

Haplotype analyses

We determined the haplotype frequencies for 10 thrombomodulin mutations (TM417C→A, TM517G→A, TM540–41GG→AT, TM845G→A, TM954G→C, 2130C→T, TM2136C→T, TM2174G→T, TM2220C→T, TM2470C Del, TM4363A→G) among Olmsted County venous thromboembolism cases and controls. Because only one case or control was a carrier for six of the thrombomodulin mutations, we analyzed the haplotypes of the remaining four thrombomodulin mutations (TM845G→A, TM2136C→T, TM2470C Del, TM4363A→G). For the 423 cases and controls that were genotyped for all four of the thrombomodulin mutations, the global score statistic was 4.37 (6 d.f., P = 0.63, simulated P-value = 0.70), indicating that no haplotype was significantly associated with venous thromboembolism. The most common haplotype, with an overall frequency of 58.1%, was wild-type at all four loci [TM845G, TM2136C, TM2470C (no deletion), and TM4363A]. The carrier frequency for this haplotype was the same in cases (58.4%) and controls (57.8%; P = 0.87). Haplotypes for all three-way and two-way combinations of the four thrombomodulin mutations were also tested and none was associated with venous thromboembolism (data not shown).

Null genetic markers

Testing at the 20 null genetic marker loci showed no allele frequency differences between cases and controls (for each locus using 1 d.f. χ2 test, Table 5). The global χ2 test also indicated no allele frequency differences between cases and controls (20 d.f. χ2, P-value = 0.83) [37].

Table 5.   Null genetic markers
GeneSNP ID no.Normal/ variantAmino acid changeMinor allele freq.ChromosomeP-value
  • *

    Personal communication from Idaho Technology, Salt Lake City, UT.

  • †1 df χ2 test.

BDNF (brain-derived neurotrophic factor)rs1048221G/TR127L0.011p131.00
GHR (growth hormone receptor)rs6184C/AP392T0.015p13-p120.97
PRDM4 (PR domain containing transcription factor)MID 128indel 3 bpintron0.4312q23-q24.10.18
Similar to ser/thr kinase 35MID 152indel 9 bpintron0.2220p130.36
CGA (glycoprotein hormones, α-polypeptide)rs6155A/GR5R0.056q12-q210.64
DBCCR1 (deleted/bladder c. chromosome region 1)rs28453A/GS691S0.099q32-q330.70
FLJ20307 (may bind nucleic acids)rs948615A/CT158N0.1318q230.52
OR1F1 (olfactory receptor 1F1)rs1834026T/CF75S0.4116p13.30.06
ADD1 (α-adducin)rs4961G/TG460W0.204p16.30.58
CYP7A1 (7 α-hydroxylase)IDT*C/A5′ UTR0.408q11-q120.62
NCAM2 (neural cell adhesion molecule 2)rs232518T/CL186P0.3421q21.10.97
BH (dopamine β-hydroxylase)rs77905A/GT207T0.509q340.62
RDH8 (retinol dehydrogenase 8, all trans)rs747574C/TL267L0.019p13.2-p13.31.00
MPHOSPH10 (M-phase phosphoprotein 10)rs6574G/AE634K0.272p120.83
EXTL3 [exostoses (multiple)-like 3]rs240951A/GP409P0.288q210.82
LCCP (Leman coiled-coil protein)rs881712C/TV93V0.2121q22.20.06
CBR3 (carbonyl reductase 3)rs88172C/TV93V0.4421q22.20.13
TOM1 (target of myb1) lysosome traffickingrs743810T/GG429G0.2822q13.10.72
FLJ20080rs918291A/CN264T0.02p141.00
CAV3 (caveolin 3)rs1008642C/TN33N0.303p250.76

Discussion

Several mutations within the thrombomodulin gene that predict a change in protein structure have been reported in small numbers of venous thromboembolism patients (Ala25Thr, Gly61Ala, Ala455Val, Asp468Tyr, Pro477Ser, Pro483Leu) [28,32–34,42,45–47] but few have been systematically studied for an association with venous thromboembolism. Several reported mutations within the putative thrombomodulin promoter region (TM417C→A, TM517G→A, TM540–41GG→AT) are in close proximity to consensus sequences for transcription control elements [26]. However, reports of the effect of these mutations on thrombomodulin transcription are contradictory [27,48].

To address these issues, we performed a population-based case–control study to test thrombomodulin polymorphisms and haplotypes for an association with venous thromboembolism. We reviewed all public databases for reported single nucleotide polymorphisms. However, because we sought to identify mutations with clinical relevance for stratifying patient risk for incident venous thromboembolism, we first wished to determine the carrier frequency of these mutations among a population of patients with the disease of interest. Therefore, we used very sensitive methods (DHPLC [49,50] or direct sequencing) to screen for sequence variation within the thrombomodulin gene in a sufficiently large sample of patients with idiopathic, objectively confirmed, venous thromboembolism to have adequate power to detect clinically important mutations [15].

Although we identified a large number of unique mutations, most were uncommon and did not affect a known transcription element, protein domain with known function, or elements known to be responsible for RNA processing. Since the inception of our study, an additional 5′-thrombomodulin promoter sequence has been reported and one study identified eight novel mutations in this region [51]. However, none of the three common or potentially important mutations were associated with venous thromboembolism.

We chose to test TM845G→A, TM2136T→C, TM2470C deletion, and TM4363A→G for an association with venous thromboembolism because they were relatively common and affected regions that were of potential functional or regulatory significance [29,42,43,48,52], were previously associated with thrombosis [53], or were evolutionarily conserved. In a population-based sample of objectively confirmed venous thromboembolism cases and community controls of comparable age and sex, none of these mutations was individually associated with venous thromboembolism, nor were any of the haplotypes. We conclude that individual mutations or haplotypes within the thrombomodulin gene are not common risk factors for venous thromboembolism.

It is important to address the potential limitations of our study. Because Olmsted County was 98% Caucasian over the 25-year period, 1966–90, the number of ethnic minority cases was insufficient to exclude a possible association between thrombomodulin gene mutations or haplotypes and venous thromboembolism in these populations [53]. Additional studies for such an association among minority populations are warranted. Our study only addressed genomic mutation. Thus, we cannot exclude changes in thrombomodulin co-translational or post-translational modification or transport that alter anticoagulant function as potential risk factors for venous thromboembolism. Finally, we cannot exclude a possible association between venous thromboembolism and a combination of multiple rare mutations within important thrombomodulin functional domains, or multiple mutations within the components of the protein C system (e.g. thrombomodulin and protein C, thrombin, thrombin-activatable fibrinolysis inhibitor).

In summary, polymorphisms or haplotypes within the thrombomodulin gene are not common risk factors for incident venous thromboembolism.

Acknowledgements

The study was funded, in part, by grants from the National Institutes of Health (HL 60279, HL66216, and AR 30582), and the Centers for Disease Control and Prevention (TS102), US Public Health Service, and by Mayo Foundation. The authors thank the staff of the Mayo Clinic Special Coagulation Laboratory and the Mayo Clinic Thrombophilia Center for assistance in patient recruitment; Christy Allred, Eric Aschenbrenner, Brian Dukek, Jennifer Eichmann, Troy Lokken, and Hong Ye for laboratory assistance; John Hermans and Brian Lahr for assistance with programming; Sara Farmer for assistance with data analysis; and Ann Beauseigneur for secretarial support.

Ancillary