Thrombomodulin Promoter Mutations, Venous Thrombosis, and Varicose Veins
Abstract—We analyzed the distal promoter region of the thrombomodulin (TM) gene (nucleotides −300 to −2052) in subjects from the Paris Thrombosis Study (PATHROS), a French case-control study of venous thrombosis, to identify polymorphisms that might modify TM gene expression. Eight novel mutations were found in the 40 DNA samples initially screened. Two of these mutations (−1748G/C and −1208/−1209 del TT) were frequent. One rare transition (−1166G/A) might have functional consequences owing to its position. These 3 mutations were screened for in the entire study population of 327 patients and 398 controls. None of the 3 was significantly associated with thrombosis. Interestingly, the −1208/−1209 TT deletion was associated with varicose veins in the patients. This mutation was in tight linkage disequilibrium with the +1418 C/T change in the coding sequence, a known polymorphism that predicts an Ala 455 Val substitution in the sixth epidermal growth factor–like TM module, a domain previously implicated in the proliferative functions of TM. This linkage suggests that the Ala 455 Val mutation may promote changes in these functions and thus be involved in varicose vein formation.
- Received July 2, 2000.
- Accepted October 27, 2000.
Identification of novel genetic risk factors for venous thromboembolism has led thrombophilia to be considered a multigene disorder. However, thrombotic events in many thrombophilic families remain largely unexplained. The natural anticoagulant protein C/protein S (PC/PS) system contributes largely to the regulation of hemostasis. Thrombin binds to thrombomodulin (TM), an endothelial cell surface glycoprotein receptor, in a stoichiometric 1:1 complex that rapidly activates PC.1 In the presence of its cofactor PS, activated PC degrades the clotting cofactors activated factors Va (FVa) and VIIIa (FVIIIa), thereby downregulating the coagulation cascade. The thrombin-TM complex is also involved in the physiological regulation of fibrinolysis by activating the recently discovered thrombin-activated fibrinolysis inhibitor.2 Genetic defects in components of the PC/PS system are risk factors for venous thrombosis. This is the case of PC and PS gene mutations, as well as the FV Leiden mutation (which modifies the activated PC cleavage site at position 506 of FVa3 ) and the prothrombin gene 20210G/A mutation associated with high levels of circulating prothrombin.4
The TM gene is located on chromosome 20. It spans 6.1 kb and contains no introns. The sequences of the adjacent 5′-(2200 nucleotides) and 3′-flanking regions have been determined.5 6 7 8 Nine point mutations of the coding sequence (nucleotides 127G/A, 129G/C, 236G/C, 543G/A, 1418C/T, 1456G/T, 1483C/T, 1502C/T, and 1689 ins T) have been found in patients with venous or arterial thrombosis,9 10 11 but no clear association with thrombosis has been demonstrated. The promoter region contains several positive and negative regulatory elements for constitutive and modulated expression.8 12 13 14 15 16 Ohlin et al9 recently studied the proximal promoter up to nucleotide −287 (numbering according to Yu et al12 ). No mutations were found in >300 patients with thromboembolism. Three mutations (nucleotides −133C/A, −33G/A, and −9 to −10 GG/AT) were found by Ireland et al17 in 1, 1, and 3 subjects, respectively, among 104 patients with myocardial infarction. The −33G/A mutation was also found in 1 control. Interestingly, both −33G/A mutations were found in Asians, pointing to a polymorphic site in this population. This idea was recently confirmed by Li et al,18 who found a 15% frequency of this mutation in Han Chinese subjects of Mongolian origin. Moreover, Li et al also found evidence that the G−33A mutation was significantly associated with coronary artery disease. In a recent case-control study of venous thromboembolism involving >200 patients and 300 healthy subjects of similar age and sex living in Paris (the Paris Thrombosis Study [PATHROS]), we also found the −33G/A mutation, but no other sequence variations of the proximal promoter were identified. The −33G/A mutation, albeit rare, was slightly more frequent in the patients (0.97%) than in the controls (0.25%). However, transient transfection experiments with unstimulated cultured endothelial cells failed to show a clear association between the −33G/A transition and gene expression.19
The distal promoter of the TM gene (nucleotides −300 to −2052) has not yet been screened in patients with thromboembolism. This region contains a silencer element,15 a putative shear stress–responsive element (SSRE),20 and 4 retinoic acid response elements that could modulate TM promoter activity.8 In the present study, we screened the entire TM promoter region (up to nucleotide −2052) in the PATHROS population for novel polymorphisms that could modify TM gene expression.
Subjects, Blood Sampling, and DNA Preparation
The study population consisted of 327 patients <61 years who had had at least 1 objectively diagnosed episode of deep venous thrombosis and/or pulmonary embolism and 398 controls of similar age and sex. The PATHROS study design has been described elsewhere.19 Varicose veins of the lower limbs were diagnosed by physical examination. As expected, known genetic risk factors for thrombosis, the FV R506Q mutation and the prothrombin gene 20210G/A mutation, and oral contraceptive use (an acquired risk factor for thrombosis) were more frequent in the cases than in the controls (Table 1⇓). Venous blood (9 vol) was collected onto 0.129 mol/L trisodium citrate (1 vol), and DNA was isolated by the method described by Miller et al21 and stored at 4°C. Plasma was stored at −30°C until analysis.
Coagulation activation markers (prothrombin fragment 1+2 [F1+2] and D-dimer) and soluble TM were assayed in plasma from control subjects by using the Enzygnost F1+2 (Behring), Asserachrom D-Di (Diagnostica Stago), and Asserachrom TM (Diagnostica Stago) kits, respectively.
For a mutation with an allele frequency of 5%, the likelihood of detecting at least 1 case of this mutation in a sample of 40 subjects is 95%. We therefore initially sequenced the TM promoter regions of the first 40 patients enrolled in PATHROS to establish the spectrum of “frequent” polymorphisms. We then used restriction site analysis to study the entire PATHROS population to determine whether venous thromboembolism was associated with the 3 novel mutations identified in the 40 subjects, namely, 2 frequent polymorphisms and 1 rare mutation at an interesting location.
Genomic DNA Studies
Direct Sequencing of the TM Promoter Region
The TM promoter region (up to nucleotide −2052) was screened for sequence variations by sequencing the 4 overlapping fragments: 101B/102B (nucleotides −2073 to −1465), PTM9/302 (nucleotides −1598 to −447), 301/302 (nucleotides −1049 to −466), and 401/402 (nucleotides −526 to +55) after polymerase chain reaction (PCR) amplification. The PCR mixtures contained 500 ng of genomic DNA, 200 μmol/L dNTPs (Pharmacia Biotech), 30 pmol of each primer, 1× PCR buffer 1 (10 mmol/L Tris-HCl, 20 mmol/L KCl, pH 8.3) or 1× PCR buffer 2 [166 mmol/L (NH4)2SO4, 666 mmol/L Tris-HCl, 67 μmol/L disodium EDTA, 100 mmol/L mercapto-2-ethanol, and 2 μg/mL bovine serum albumin) and 1 U of Taq polymerase (Super Taq, ATGC Biotechnologie) in a final volume of 100 μL. The sequences of the primers and the specific PCR conditions used to amplify each fragment are given in Table 2⇓.
Automated cycle sequencing of double-stranded DNA was performed with the ABI prism dye terminator cycle sequencing ready reaction kit (Perkin Elmer) according to the manufacturer’s instructions. The primers used for sequencing were the same as those used for PCR, except for the PTM9/302 and 301/302 fragments. The primers used to sequence the PTM9/302 fragment were 201 and 202B (Table 2⇑). For 301/302, the reverse sequencing primer was 302, and the forward sequencing primer was changed to 301B (5′ -1034 CGAGCAAGTGGCGTTTCTATG -1014 3′). The sequencing profiles were analyzed on an ABI prism 310 or 377 apparatus (Perkin Elmer).
Restriction Site Analysis
The novel TM promoter mutations identified by sequencing were verified by restriction site analysis, and the same method was then used to screen for these mutations in the entire study population. The sequences of the primers used for amplification are reported in Table 2⇑, together with the specific PCR conditions. The restriction enzymes used and the restriction profiles associated with these mutations are shown in Table 3⇓. For 2 mutations (−1084C/T and −1078/−1079 del CC), restriction sites were created in the amplified fragments by using modified primers (pTM StuI and pTM XcmI).
The G−33A mutation was screened for as previously described.19 The nucleotide 1418 C/T common dimorphism in the TM gene, responsible for the Ala 455 Val mutation,9 was also screened for by digestion after amplification of the region of interest with a primer modified to introduce a cleavage site for AflIII in the presence of a T at position 1418 (Table 2⇑).
Data were analyzed by using sas software (SAS Institute Inc). The frequencies of oral contraceptive use and of the FV Leiden and prothrombin gene G20210A mutations were compared between the cases and controls by using the χ2 test. The χ2 test was also used to test whether the genotype distributions in cases and controls were in Hardy-Weinberg equilibrium. Allele frequencies were deduced from genotype frequencies. Standard disequilibrium coefficients (Δ) were calculated by using likelihood methods.
Characteristics of cases heterozygous for the TM del TT mutation and those homozygous for the wild-type TM allele were compared by using Student’s t test for age and mean follow-up since the first venous thrombosis, whereas the χ2 test was used to compare the percentage of women, oral contraceptive use, type of venous thrombosis, and the presence of varicose veins, postthrombotic syndrome, and known genetic risk factors for thrombosis (FV Leiden, prothrombin gene G20210A mutation, and PC or PS deficiency). The few subjects homozygous for the TM del TT mutation were excluded from these analyses.
Logistic regression analysis was used to determine whether venous thrombosis, varicose veins, and a postthrombotic syndrome were related to the TM gene mutations, with and without adjustment for age and sex. Because of their small number, subjects homozygous for the mutations were combined with heterozygotes for these analyses. The effect of sex, age, and type of venous thrombosis was evaluated by repeating the same analyses in each group separately (men versus women, young versus old, spontaneous versus nonspontaneous venous thrombosis, and first versus recurrent venous thrombosis). ANOVA with adjustment for age, sex, and oral contraceptive use was used to evaluate the mean concentrations of soluble TM, F1+2, and D-dimer by genotype of TM in controls. Differences with a probability value <0.05 were considered statistically significant.
Comparison of the patients’ TM promoter sequences with the sequence published by Dittman et al8 showed 5 differences present in all of the sequences, namely, 2 nucleotide changes (the G at position −827 upstream of the transcription start site proposed by Yu et al12 was replaced by a T, and the G at position −753 was replaced by a C) and 3 insertions of a G at positions −735, −730, and −701. These insertions thus change the numbering upstream of nucleotide −701. For clarity, nucleotides will hereafter be numbered as proposed by Yu et al, followed by the new numbering system in brackets and italics.
The 40 DNA samples initially screened contained 7 substitutions in the TM promoter, namely, the previously identified nucleotide −33 G/A mutation and 6 novel substitutions located at positions −1848 [−1851] (C/G), −1799 [−1802] (G/C), −1748 [−1751] (G/C), −1166 [−1169] (G/A), −1084 [−1087] (C/T), and −797 [−800] (C/G). We also found 2 deletions of 2 nucleotides each at positions −1208/−1209 ([−1211/−1212], del TT) and −1078/−1079 ([−1081/−1082], del CC) (Table 3⇑). The sequence variations at positions −1748 [−1751] and −1208/−1209 [−1211/−1212] were frequent: the first was found in the heterozygous state in 10 patients and in the homozygous state in 3 patients, whereas the second was found in the heterozygous state in 11 patients. These variations were then screened for in the entire study population (Table 4⇓).
The other variations were rare, being found in the heterozygous state in 1 patient each. These rare variations did not modify or create consensus binding sequences for transcription factors and were unlikely to have functional consequences, except for the −1166 [−1169] G/A mutation, which modified the core binding sequence (GACGACC) of a putative SSRE.20 For this reason, the only rare novel mutation screened for in the entire study population was −1166G/A (Table 4⇑).
The 2 frequent sequence variations, −1748 [−1751] G/C and −1208/−1209 [−1211/−1212] del TT, and the −1166 [−1169] G-to-A transition were in Hardy-Weinberg equilibrium in both the controls and the cases. The mutated alleles were not more frequent in the patients than in the controls: the odds ratios (followed by their 95% confidence intervals) adjusted for age and sex were 0.90 (0.66 to 1.22), 0.92 (0.63 to 1.35), and 0.78 (0.25 to 2.43) for the −1748 [−1751] G/C, −1208/−1209 [−1211/−1212] del TT, and −1166 [−1169] G-to-A mutations, respectively, suggesting that these mutations are not risk factors for thrombosis.
We then investigated whether these mutations influenced coagulation activation status by measuring activation markers (F1+2 and D-dimer) in controls. We found significantly higher D-dimer levels in subjects with the −1748 [−1751] G/C mutation (Table 5⇓). Nineteen percent of the patients investigated had varicose veins. As expected, varicose veins were significantly associated with thrombosis (P=0.004) and the postthrombotic syndrome (P<0.001). Because varicose veins are characterized by disorganization of vessel wall structure, which might be associated with altered TM expression, we sought an association between varicose veins and the 3 mutations. Interestingly, patients with the del TT allele were more likely to have varicose veins of the lower limbs than were patients with the wild-type allele (33% vs 17%, P=0.007). The relative risk of varicose veins associated with this mutation was 2.21 (1.11 to 4.38) after adjustment for age and sex. The other mutations had no significant effect. Patients with the del TT allele did not differ significantly from other patients with regard to demographic data, clinical characteristics, or known genetic risk factors for venous thrombosis (FV Leiden, prothrombin gene G20210A mutation, and PC or PS deficiency). The del TT mutation did not increase the risk of recurrent thrombosis or the postthrombotic syndrome in patients with varicose veins.
To elucidate the association between the del TT mutation and varicose veins, we sought disequilibrium linkage with another mutation that truly modulates TM function. We therefore sequenced DNA from the 53 patients heterozygous for the del TT mutation to identify variations in the cAMP-regulating element located in the 3′-untranslated region of the TM gene.16 No mutations were found. We then screened the same subjects for the Ala 455 Val mutation in the DNA coding sequence, which was also a good candidate, being frequent9 and located in a region of the TM protein putatively involved in proliferative functions.22 Fifty-one (96%) of the 53 patients heterozygous for the del TT mutation were also heterozygous for the Ala 455 Val mutation, while the other 2 patients were homozygous for Ala 455 Val. These results suggested a tight linkage disequilibrium of the del TT mutation and the Ala 455 Val mutation. We then genotyped the entire study population for this mutation (Table 6⇓) and confirmed the tight linkage of the 2 mutations in cases (Δ=0.983) and controls (Δ=0.895) (P<0.01). No effect of the del TT mutation on TM expression was found when soluble TM levels in control subjects were analyzed according to genotype.
We found 8 novel sequence variations in the TM promoter of patients with venous thromboembolism. Six of the novel sequence variations were rare: −1848 [−1851] C/G, −1799 [−1802] G/C, −1166 [−1169] G/A, −1084 [−1087] C/T, −1078/−1079 [−1081/−1082] del CC, and −797 [−800] C/G. Among them, the −1166 [−1169] G/A mutation modified the core binding sequence (GAGACC) of a putative SSRE.20 Recent studies have shown that blood flow and the associated fluid shear stress cause morphological and functional changes in endothelial cells, including up- or downregulation of TM expression,23 24 25 which could involve this SSRE. By impairing this action, the mutation might alter endothelial cell anticoagulant functions and create a thrombotic tendency. However, we found no relationship between this mutation and the risk of thrombosis in the entire PATHROS population (327 patients and 398 controls).
The other 2 novel mutations, −1748 [−1751] G/C and −1208/−1209 [−1211/−1212) del TT, were frequent. The −1748 [−1751] G/C mutation was located in a promoter region shown by Tazawa et al15 to be responsible for positive transcription in transient transfection assays. Moreover, this mutation created a putative Sp1-like responsive element (C-1748TCCGCGTC) with only 1 mismatch relative to the Sp1 binding consensus sequence (C/AC/TCCGC/AT/CT/CC/A). We found a significant association between the −1748 [−1751] G/C mutation and higher plasma D-dimer levels, but no association with the risk of venous thromboembolism.
The del TT mutation did not affect a known regulating element or create a binding sequence for known transcription factors. The del TT allele did not significantly influence levels of coagulation activation markers (F1+2 or D-dimer) in controls, suggesting no direct influence of the del TT allele on thrombin generation. This concept was supported by the absence of any association between this mutation and the risk of thrombosis.
Interestingly, we found a significant association between the del TT allele and varicose veins in the cases of the PATHROS study. No such relationship was found with the other sequence variations. Because the PATHROS study was not designed to identify risk factors for varicose veins, we have no information on the distribution of del TT in controls according to their history of varicose veins.
The role of heredity in the development of varicose veins of the lower limbs has been raised many times in the literature,26 27 but no genetic risk factor had been firmly identified. Varicose veins are characterized by a disorganized state of the vessel wall structure, and blood stasis is involved in their development.28 29 Blood stasis provokes ischemia, thereby decreasing oxygen availability to the tissues. The effect of oxygen privation on endothelial function could be the starting point for a cascade of events leading to endothelial cell activation, profound changes in subendothelial structures with leukocyte infiltration, and proliferation of smooth muscle cells and qualitative and quantitative changes in their functions.
It is not clear how a mutation in the TM promoter could intervene in varicose vein formation. However, we know that TM is expressed on endothelial cells,30 smooth muscle cells,31 and leukocytes32 (the functions of all 3 cell types are modified in varicose veins) and that hypoxia can quantitatively modulate TM expression31 33 34 by mechanisms involving cAMP.35 36 Low intracellular cAMP levels downregulate TM expression at the surface of endothelial cells,37 and a functional cAMP-responsive element has been located in the 3′-untranslated region of the human TM gene.16 TM has been suggested to regulate cell functions, such as proliferation, through mechanisms independent of those regulating hemostasis. An antiproliferative effect of TM has been demonstrated in tumor cells from patients with malignant melanoma, and it has been suggested that the lectin and cytoplasmic domains of TM could be involved in this effect.38 Conversely, TM expression might promote atherosclerosis through mitogenic activity on vascular smooth muscle cells, the epidermal growth factor–like domain of the protein possibly having a juxtacrine mechanism.22 Indeed, this domain has been found to be mitogenic for a fibroblastic cell line.39
The del TT mutation could have a direct or indirect effect on the level of TM expression at the surface of cells activated by hypoxia. It has been suggested that in healthy subjects, the soluble TM level might reflect the quantity of TM expressed on the endothelial surface.40 There is also some evidence to support this hypothesis from a family study of the mutation 1689 ins T.10 The del TT mutation was not associated with soluble TM levels in our control population. A direct effect of this mutation on the level of TM expression seems doubtful, because it does not create or modify a known regulating element. The del TT mutation might rather act in tight linkage disequilibrium with another mutation with functional consequences. Because this second mutation might be located in the cAMP-responsive element of the 3′-untranslated region,16 we sequenced this region in all of the patients heterozygous for the del TT mutation but found no sequence variations. The second mutation could also be located in the epidermal growth factor or lectin domain of the protein, which have been implicated in cell proliferation. One polymorphism, a nucleotide 1418C/T change predicting an Ala 455 Val substitution, has been identified in the sixth epidermal growth factor–like module. The frequency of the Val allele was ≈20% (16% to 26%) in subjects from the Netherlands, Sweden, and North America.9 In our study population, the frequency of the Val allele was 13% in cases and 18% in controls. It has been suggested that the C/T dimorphism could be neutral with respect to venous thrombophilia and that it might be involved in the pathogenesis of myocardial infarction, although the latter hypothesis remains to be confirmed. The observed association between the Val allele and the del TT allele should now be explored for its possible involvement in the development of varicose veins.
- ↵Esmon CT. The roles of protein C and thrombomodulin in the regulation of blood coagulation. J Biol Chem. 1989;264:4743–4746.
- ↵Poort SR, Rosendaal FR, Reitsma PH, Bertina RM. A common genetic variation in the 3′-untranslated region of the prothrombin gene is associated with elevated plasma prothrombin levels and an increase in venous thrombosis. Blood. 1996;88:3698–3703.
- ↵Jackman RW, Beeler DL, Fritze L, Soff G, Rosenberg RD. Human thrombomodulin gene is intron depleted: nucleic acid sequences of the cDNA and gene predict protein structure and suggest sites of regulatory control. Proc Natl Acad Sci U S A. 1987;84:6425–6429.
- ↵Shirai T, Shiojiri S, Ito H, Yamamoto S, Kusumoto H, Deyashiki Y, Maruyama I, Suzuki K. Gene structure of human thrombomodulin, a cofactor for thrombin-catalyzed activation of protein C. J Biochem. 1988;103:281–285.
- ↵Dittman WA, Nelson SC, Greer PK, Horton ET, Palomba ML, McCachren SS. Characterization of thrombomodulin expression in response to retinoic acid and identification of a retinoic acid response element in the human thrombomodulin gene. J Biol Chem. 1994;269:16925–16932.
- ↵Kunz G, Ireland H, Stubbs P, Kahan M, Coulton GC, Lane DA. Identification and characterization of a thrombomodulin gene mutation coding for an elongated protein with reduced expression in a kindred with myocardial infarction. Blood. 2000;95:569–576.
- ↵Yu K, Morioka H, Fritze LM, Beeler DL, Jackman RW, Rosenberg RD. Transcriptional regulation of the thrombomodulin gene. J Biol Chem. 1992;267:23237–23247.
- ↵Von der Ahe D, Nischan C, Kunz C, Otte J, Knies U, Oderwald H, Wasylyk B. Ets transcription factor binding site is required for positive and TNF-α-induced negative promoter regulation. Nucleic Acids Res. 1993;21:5636–5643.
- ↵Conway EM, Liu L, Nowakowski B, Steiner-Mosonyi M, Jackman RW. Heat shock of vascular endothelial cells induces an up-regulatory transcriptional response of the thrombomodulin gene that is delayed in onset and does not attenuate. J Biol Chem. 1994;239:22804–22810.
- ↵Tazawa R, Hirosawa S, Suzuki K, Hirokawa K, Aoki N. Functional characterization of the 5′-regulatory region of the human thrombomodulin gene. J Biochem. 1993;113:600–606.
- ↵Ireland H, Kunz G, Kyriakoulis K, Stubbs PJ, Lane DA. Thrombomodulin gene mutations associated with myocardial infarction. Circulation. 1997;96:15–18.
- ↵Le Flem L, Picard V, Emmerich J, Gandrille S, Fiessinger JN, Aiach M, Alhenc-Gelas M. Mutations in promoter region of thrombomodulin and venous thromboembolic disease. Arterioscler Thromb Vasc Biol. 1999;19:1098–1104.
- ↵Miller SA, Dykes DD, Polesky HF. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res. 1988;16:1215.
- ↵Tohda G, Oida K, Okada Y, Kosaka S, Okada E, Takahashi S, Ishii H, Miyamori I. Expression of thrombomodulin in atherosclerotic lesions and mitogenic activity of recombinant thrombomodulin in vascular smooth muscle cells. Arterioscler Thromb Vasc Biol. 1998;18:1861–1869.
- ↵Malek AM, Jackman R, Rosenberg RD, Izumo S. Endothelial expression of thrombomodulin is reversibly regulated by fluid shear stress. Circ Res. 1994;74:852–859.
- ↵Gosling M, Golledge J, Turner RJ, Powell JT. Arterial flow conditions downregulate thrombomodulin on saphenous vein endothelium. Circulation. 1999;99:1047–1053.
- ↵Vanhoutte PM, Corcaud S, de Montrion C. Venous disease: from pathophysiology to quality of life. Angiology. 1997;48:559–567.
- ↵Michiels C, Arnould T, Remacle J. Hypoxia-induced activation of endothelial cells as possible cause of venous diseases: a hypothesis. Angiology. 1993;44:639–646.
- ↵Michiels C, Arnould T, Thibaut-Vercruyssen R, Bouaziz N, Janssens D, Remacle J. Perfused human saphenous veins for the study of the origin of varicose veins: role of endothelium and hypoxia. Int Angiol. 1997;16:135–141.
- ↵Maruyama I, Bell CE, Majerus PW. Thrombomodulin is found on endothelium of arteries, veins, capillaries, and lymphatics and on syncytiotrophoblast of human placenta. J Cell Biol. 1985;101:363–371.
- ↵Soff G, Jackman R, Rosenberg R. Expression of thrombomodulin by smooth muscle cells in culture: different effects of tumor necrosis factor and cyclic adenosine monophosphate on thrombomodulin expression by endothelial cells and smooth muscle cells in culture. Blood. 1991;775:515–518.
- ↵Conway EM, Nowakowski B, Steiner-Mosonyi M. Human neutrophils synthesize thrombomodulin that does not promote thrombin dependent protein C activation. Blood. 1992;80:1254–1263.
- ↵Ogawa S, Gerlach H, Esposito C, Pasagian-Macaulay A, Brett J, Stern D. Hypoxia modulates the barrier and coagulant function of cultured bovine endothelium. J Clin Invest. 1990;85:1090–1098.
- ↵Ogawa S, Koga S, Karakurum M, Brett J, Morrow B, Morris A, Bilezikian JP, Joseph-Silverstein J, Stern D. Hypoxia-induced increased permeability of endothelial monolayers occurs through lowering of cellular AMPc levels. Am J Physiol. 1992;262:C546–C554.
- ↵Dufourcq P, Seigneur M, Pruvost A, Dumain P, Belloc F, Amiral J, Boisseau MR. Membrane thrombomodulin levels are decreased during hypoxia and restored by cAMP and IBMX. Thromb Res. 1994;77:305–310.
- ↵Zhang Y, Weiler-Guettler H, Chen J, Wilhelm O, Deng Y, Qiu F, Nakagawa K, Klevesath M, Wilhelm S, Böhrer H, Nakagawa M, Graeff H, Martin E, Stern DM, Rosenberg RD, Ziegler R, Nawroth PP. Thrombomodulin modulates growth of tumor cells independent of its anticoagulant activity. J Clin Invest. 1998;101:1301–1309.
- ↵Hamada H, Ishii H, Sakyo K, Horie S, Nishiki K, Kazama M. The epidermal growth factor-like domain of recombinant human thrombomodulin exhibits mitogenic activity for Swiss 3T3 cells. Blood. 1995;86:225–233.
- ↵Salomaa V, Matei C, Aleksic N, Sansores-Garcia L, Folsom AR, Juneja H, Chambless LE, Wu KK. Soluble thrombomodulin as a predictor of incident coronary heart disease and symptomless carotid artery atherosclerosis in the Atherosclerosis Risk in Communities (ARIC) Study: a case-cohort study. Lancet. 1999;353:1729–1734.