The ENTPD1 promoter polymorphism −860 A > G (rs3814159) is associated with increased gene transcription, protein expression, CD39/NTPDase1 enzymatic activity, and thromboembolism risk

Abstract

Ectonucleoside triphosphate diphosphohydrolase 1 (NTPDase1) degrades the purines ATP and ADP that are key regulators of inflammation and clotting. We hypothesized that NTPDase1 polymorphisms exist and that they regulate this pathway. We sequenced the ENTPD1 gene (encoding NTPDase1) in 216 subjects then assessed genotypes in 2 cohorts comprising 2213 humans to identify ENTPD1 polymorphisms associated with venous thromboembolism (VTE). The G allele of the intron 1 polymorphism rs3176891 was more common in VTE vs. controls (odds ratio 1.26–1.9); it did not affect RNA splicing, but it was in strong linkage disequilibrium with the G allele of the promoter polymorphism rs3814159, which increased transcriptional activity by 8-fold. Oligonucleotides containing the G allele of this promoter region bound nuclear extracts more avidly. Carriers of rs3176891 G had endothelial cells with increased NTPDase1 activity and protein expression, and had platelets with enhanced aggregation. Thus, the G allele of rs3176891 marks a haplotype associated with increased clotting and platelet aggregation attributable to a promoter variant associated with increased transcription, expression, and activity of NTPDase1. We term this gain-of-function phenotype observed with rs3814159 G “CD39 Denver.”—Maloney, J. P., Branchford, B. R., Brodsky, G. L., Cosmic, M. S., Calabrese, D. W., Aquilante, C. L., Maloney, K. W., Gonzalez, J. R., Zhang, W., Moreau, K. L., Wiggins, K. L., Smith, N. L., Broeckel, U., Di Paola, J. The ENTPD1 promoter polymorphism −860 A > G (rs3814159) is associated with increased gene transcription, protein expression, CD39/NTPDase1 enzymatic activity, and thromboembolism risk.

Purinergic signaling is a key pathway for the regulation of inflammation, lymphocyte trafficking, and clotting (1, 2). ADP release from platelet granules is a key amplification loop causing platelet aggregation and clotting as it binds and activates the purinergic receptors P2Y12 and P2Y1 on the platelet surface. Homeostatic mechanisms that degrade ADP and ATP are important counterbalances to clotting and other purinergic signaling pathways (3). The endothelial ectoenzyme nucleoside triphosphate diphosphohydrolase 1 (NTPDase1), also known as NTPDase1/CD39, is encoded by the gene ectonucleoside triphosphate diphosphohydrolase 1 (ENTPD1) and is the rate-limiting luminal enzyme for hydrolysis of ATP and ADP (4, 5). Not surprisingly, NTPDase1/CD39 is an important modulator of clotting, as ADP and ATP released by platelets are primarily degraded by NTPDase1/CD39 expressed on the luminal endothelial surface. NTPDase1/CD39-null mice generated by 2 research groups have variable phenotypes of clotting risk, while NTPDase1/CD39 transgenic mice that ubiquitously overexpress this enzyme are protected from clotting. Intravenous administration of a soluble recombinant NTPDase 1/CD39 enzyme has been shown to limit infarction in a murine cerebral thrombosis model (6–8). All of these mouse models support a critical role for NTPDase1/CD39 in the control of coagulation.

We postulated that any common genetic variants in NTPDase1/CD39 that affect its function and/or expression will influence human clotting risk. In particular, we postulated that venous thromboembolism (VTE) is likely to be associated with functional variants of NTPDase1/CD39. VTE is important as it is a common cause of cardiovascular death that can manifest as deep venous thrombosis (DVT) with or without pulmonary embolism (PE) (9). Of those subjects who develop VTE, 30 to 40% will have a recurrence in their lifetime, and 2 to 4% of subjects with PE will not experience resolution of their emboli and will go on to develop chronic thromboembolic disease with pulmonary hypertension (CTEPH) (10, 11). Family and twin studies suggest that genetics may explain up to 60% of VTE risk, but known genetic risk factors for VTE, such as factor V Leiden, comprise a minority of the inherited thrombophilias (12). The rationale for such a study is further supported by the prominent association of chromosome 10 (containing ENTPD1) with increased VTE risk (13). Studies of genetic variation in NTPDase1/CD39 as a disease risk factor have been reported in inflammatory bowel disease, AIDS, and diabetes (14–16). To date, no study has examined ENTPD1 to assess for variants with direct functional effects.

To test this hypothesis, we sequenced the coding and proximal regulatory regions of NTPDase1/CD39 in 108 consecutive inpatients with VTE and matched controls. We evaluated these groups for disparities in the frequencies of common polymorphisms in ENTPD1; those linked with VTE risk were further assessed for replication in a large independent case–control study. We then evaluated potential associations of an implicated polymorphism with transcriptional activity, endothelial protein expression, endothelial ADP/ATPase activity, surface expression of NTPDase1/CD39 on platelets, and for differences in platelet aggregation induced by purinergic agonists.

MATERIALS AND METHODS

Human subjects

Research was approved by the Medical College of Wisconsin, University of Washington, and University of Colorado institutional Review Boards, and written informed consent was obtained. Demographic data for the discovery and replication groups are listed in Table 1.

TABLE 1.

Demographics of human subject groups

Characteristic	Discovery group		Replication study		Platelet donor
	VTE	Control	VTE	Control
Cohort size	108	108	800	1197	26
Sex female (%)	52	52	100	100	38
Age, mean (yr)	56	33	65	68	38
Race/ethnicity (%)
White	76.9	75.9	99.6	94.1	76.9
African American	22.2	22.2	0	0	23.1
Asian/other	0.9	1.9	0.4	5.9	0
Family history (%)	13.8	0	NA	NA	0
Prior VTE (%)	35.2	0	0	0	0
Active malignancy (%)	22.2	0	NA	NA	0
Provoked status (%)	30.5	NA	NA	NA	NA

NA, not ascertained or not applicable. Provoked status (assessed in discovery cohort only) was defined as relation to active cancer, recent hospitalization, surgery, trauma, or oral contraceptive use in past 3 mo. All VTE were incident events.

Discovery cohort

A total of 108 consecutive inpatients were enrolled with an acute VTE identified during an admission (Froedtert Hospital, Milwaukee, WI, USA); 108 healthy controls were matched for self-reported race/ethnicity and gender. VTE diagnosis was confirmed by results of lower extremity ultrasound, lung ventilation–perfusion scans, or computed tomographic pulmonary angiography. Data were obtained for family history of VTE (one or more first-degree relative). Provoked VTE was defined as being related to surgery, trauma, oral contraceptive use, or hospitalization (>3 d) in the past 3 mo. In this small cohort, subjects were not split into PE or DVT groups. The 108 subjects per group provided detection powers of 99.6% if polymorphisms were present in 5% of subjects, and 66.2% if polymorphisms were present in 1% of subjects (88.6% for combined control and VTE groups), giving adequate power to find common [minor allele frequency (MAF) >5%] and uncommon (MAF 1–5%) polymorphisms. Rare was defined as MAF <1%.

Replication study

An independent population-based, case–control study of VTE cases and matched controls was obtained from the Heart and Vascular Health (HVH) study. Data were obtained from Group Health Cooperative, a large integrated health care system in Washington State. All incident inpatient and outpatient VTE cases in women aged 30 to 89 yr were identified from 1995 to 2009. Among cases, 51.8% had experienced a PE (with or without a DVT) and 48.3% had an isolated DVT. An objective imaging test was present and confirmatory in >97% of cases. Controls were randomly selected women without any history of VTE who were matched to HVH cases on design variables that included index year, hypertension diagnosis, and age. Subjects were predominantly white (96.3%). Blood was collected from women who were alive at the time of eligibility review and agreed to provide a blood sample. Single nucleotide polymorphism (SNP) data for rs3176891 and rs10748643 were measured previously as part of a larger study on genomewide variation and VTE risk (17). Genotyping was performed with Illumina 370CNV and Omni arrays (Illumina, San Diego, CA, USA) and imputed to the HapMap II sample of European ancestry (CEU). Genotyping personnel were blinded to case–control status.

Platelet donors

Healthy subjects were genotyped at rs3176891 to identify 13 AA and 13 GG subjects. Both men and women aged 18 to 55 were enrolled. A power analysis (using platelet aggregation data from recent projects) predicted that a number of 13 for each genotype would detect a 50% change in platelet aggregation between groups based on an expected sd of 10%, a power of 90%, and an α error of 5%. Subjects who were heterozygous for rs3176891 were not enrolled because of the nonfeasibility of functional assessments—45 subjects per group would be needed if effect size was half that of homozygotes. Exclusion criteria included use of tobacco, aspirin, or nonsteroidal antiinflammatory agents in the past 10 d; use of warfarin or P2Y12 antagonists; anemia (hemoglobin <12 g/dl); or prior VTE.

Materials

Chemicals, media, plastics, and reagents were obtained from Thermo Fisher Scientific (Waltham, MA, USA) unless otherwise stated. Technicians who performed functional assays were not aware of the assayed sample genotype.

PCR amplification of genomic DNA and DNA sequencing (discovery cohort)

Genomic DNA was isolated from blood (Gentra Puregene; Qiagen, Germantown, MD, USA) and resuspended at 10 ng/μl in Tris-EDTA buffer. Primer pairs were designed using DS Gene 2.5 (Accelrys, San Diego, CA, USA) with goals of amplifying all coding exons, 100 to 150 nt intron–exon boundaries, the first 200 nt of the 5′-UTR and proximal promoter, and 400 nt of the 3′-UTR. Exon boundaries conformed to that of Maliszewski et al. (18). Primers were designed using genomic ENTPD1 sequence (NM001098175.1). General PCR methods, dye terminator sequencing, and genotype calling were performed as published in Maloney et al. (19). Primers (Operon, Huntsville, AL, USA) and conditions are listed in Table 2.

TABLE 2.

Primer/probe sequences and conditions for PCR and pyrosequencing of ENTPD1

	Resequencing primer, 5′–3′
Region	Forward	Reverse	Program
Exon 1	TTACAACCTGGAAAAGGCTTC	CTTAAACTGCAAGGGGACTG	P(6–7)
Exon 2	TAAAGGACATGCTGCTTAGGGC	TTCACCCATGATGAAGAAGTG	P(54)
Exon 3	GCAAACAGCATGAAGAGC	CGATAACATAACAACAGCAGC	P(57)
Exon 4	TTGAACACTGAACAGACC	CTCTTATCCTCTTCCTCC	P(57)
Exon 5	CCCCATCTCTTCATTTATTCC	TCATCCATCCATCCATCC	P(6–7)
Exon 6	GTTTCTTTGCCTTAGGAAATCC	TTCTCAGCCTCCCAAGTAGC	P(57)
Exon 7	TGCCTGTTACACAAATCC	TGAGCTGAAGAACCAACC	P(6–7)
Exon 8	AAGCACAAGGGGAAAAAACC	AAACAGGCACAACAGCAGAGCT	P(6–7)
Exon 9	ACTACTTTTTTCCTGGAGGACC	GCAACTGAAAATGCAAGTAGC	P(54)
Exon 10	AACTCTTCTAACTCCTCCAACC	GATTCTTCTTTCAGCCAGC	P(57)
RNA splicing	AGGAGGAAAACAAAAGCTGC	AGATTCCAGGACCTTTAACCC	P(6–7)
Cloning
Promo	ATTAGGTACCCACCGTGCAAAGTAACAGAG	TAATCTCGAGAGGACAGATTGACTGAGGAG	P(6–7)
MCS	ATCTTCCATGGTGGCTTTA	GTGTTGGTTTTTTGTGTGAA	P(6–7)
Pyrosequencing	Biotin-TGCCCTTAGAGGGTTCTTTTC	GTATGGGAGAGATGTCCTCTTTGA
Sequencing	TCCTCTTTGATGCCAG
EMSA	[BIOTEG]CCTTTCAAAGGATTAACCCTTGTTTTGATTT	AAATCAAAACAAGGGTTAATCCTTTGAAAGG

P(6–7): lid 105°C constant; [95°C for 2:00, 95°C for 0:30, 56.7°C for 1:00, 55°C for 0:30] × 9 cycles; 40 cycles [95°C for 0:30, 48.7°C for 0:30, 68°C for 1:00]; 72°C for 10:00; 4°C forever. P(54): lid 100°C; [94°C for 1:30, 94°C for 0:30, 54°C for 0:30; 68°C for 1:30] × 29 cycles, 72°C for 3:00; 4°C forever; P(57) is same as P(54) except annealing is 57°C. For promoter cloning PGL4.23 primers (promo-), underlined nucleotides correspond to KpnI (forward) and XhoI (reverse) restriction sites. For EMSA oligonucleotides, promoter polymorphism rs3814159 alleles are underlined.

Genotyping (discovery cohort)

National Center for Biotechnology Information (NCBI) online genetic data indicated that an implicated SNP (rs3176891) in our study was in modest linkage disequilibrium (LD) with another SNP (rs10748643) previously linked with higher NTPDase1/CD39 expression in lymphocytes and inflammatory bowel disease (14). These SNP lie 64 kb apart at opposite ends of the first intron. The LD between the 2 was 0.784 by the Genome Variation Server (http://gvs.gs.washington.edu/GVS134/) using the HapMap CEU population. To test if rs10748643 and rs3176891 were in LD, rs10748643 was genotyped in the entire control group using a commercial assay (hCV27420559, TaqMan; Thermo Fisher Scientific). Standard reagents, PCR conditions, quality control, and allele calls were as published (19). The SNP rs11188513 was not genotyped because although it has been included in other studies as an ENTPD1 SNP, it is far from ENTPD1 and thus is intergenic (16). Genotypes for rs3176891 in platelet donors were determined by PCR pyrosequencing using manufacturer protocols and software (PSQ 96MA; Qiagen) (20). Primers are listed in Table 2.

ENTPD1 promoter constructs

NCBI online genetic data indicated that an implicated SNP (rs3176891) in our study was in strong LD with a promoter SNP (rs3814159), which was evaluated as a potential functional SNP within this haplotype. A 1.2 kb section of the ENTPD1 promoter with either the A or G allele of rs3814159 was PCR amplified from genomic DNA (Table 2). The primers used amplify from positions −1384 to −219 (distance from translational start). PCR products were purified and cut with XhoI and KpnI (New England Biolabs, Ipswich, MA, USA), then cloned into minimal promoter vector pGL4.23 [luc/minP] (Promega, Madison, WI, USA). Insert-containing clones were selected from Escherichia coli (Promega) and verified for the insert by cutting with BamH1; they then were miniprepped and purified. The insert sequence and orientation were verified by sequencing with primers (Table 2, MCS-forward, MCS-reverse). Vectors were maxiprepped, and purified DNA was diluted to 0.5 μg/μl in H₂O. Vectors were designated as pGL4.23[empty], pGL4.23[−860A], and pGL4.23[−860G].

HEK293 cell transfections

Transfections were performed using Fugene-HD (Promega) in human HEK293 cells grown at 37°C and 5% CO₂ to subconfluence. Cells were plated at 90,000 cells per well on 24-well plates. The next morning, cells were washed, 0.5 ml of OptiMEM medium (Thermo Fisher Scientific) was added, and cells were transfected with either pGL4.23[empty], pGL4.23[−860A], pGL4.23[−860G], no DNA, or a green fluorescent protein (GFP) expression vector (pmaxGFP; Lonza, Basel, Switzerland) at a 3:2 ratio (μl reagent to μg DNA) for each well. GFP vectors were added to the bottom well of each column to assess transfection efficiency, which was >90% on all attempts based on manual counting with laser microscopy (Axiovert 40 CFL; Carl Zeiss, Jena, Germany). At 24 h, cells were washed with PBS then either lysed or treated for 4 h with 250 μM of 8-bromo-cAMP (Tocris Bioscience, Bristol, United Kingdom), as this was previously reported to increase ENTPD1 expression (at promoter sites not included in this construct) (21). After incubation, cells were washed with PBS; then 0.1 ml of reporter lysis buffer (Promega) was added per well, and plates were placed at −70°C overnight. After mixing and centrifuging at 500 g, 20 μl of lysate supernatant per well was assayed with a luminometer (Turner Biosystems, Sunnyvale, CA, USA) and 100 μl luciferase reagent (Promega). Results were expressed as relative light units (RLUs). Transfection efficiencies of >90% (parallel GFP transfections) obviated the need for a normalizing Renilla vector, so RLUs were normalized to total protein (mg) for that well by Bradford assay (Thermo Fisher Scientific).

Electromobility shift assays

The online prediction software tools Alibaba 2.1 (http://gene-regulation.com/pub/programs/alibaba2/) and Promo (http://alggen.lsi.upc.edu/recerca/menu_recerca.html) were used to identify putative transcription factor sites in the region of the rs3814159 polymorphism and to predict genotype-related effects. Isolation and quantitation of human iliac vein endothelial cell (HIVEC) nuclear extracts and general electromobility shift assay (EMSA) techniques using genotype-specific biotin probes were as published (19). Single-stranded oligonucleotides 30 to 31 bp in length (Table 2, centered on the rs3814159 locus) were used to make double-stranded DNA probes. Recombinant human forkhead box P3 protein (FOXP3) protein (Abnova, Walnut, CA, USA) was utilized in some experiments to assess for binding to a putative contiguous FOXP3 site. Band densities from radiographs of transferred gels were analyzed with ImageJ software (Image Processing and Analysis in Java; National Institutes of Health, Bethesda, MD, USA; http://imagej.nih.gov/) to generate density values for statistical analysis.

Venous endothelial cell culture

Primary HIVECs (n = 16) were obtained from Coriell Repositories (Camden, NJ, USA). Cells were used in the passages 4 to 8 and grown on 1% gelatin-coated T-75 flasks in 21% O₂ and 5% CO₂ at 37°C. Growth media were as recommended by the supplier. Cells were split 1:2 when subconfluent. Genotypes for rs3176891 and rs10748643 were determined by DNA sequencing.

Evaluation for alternative RNA splicing

cDNA was made from HIVEC RNA and quantitative real-time PCR using published techniques (19), specific primers, and 25 μl of PCR Optimization PreMix J (Epicentre, Madison, WI, USA; final 50 μl volume) (Table 2). PCR products (expected 299 bp) were separated on agarose gels, and band size was inspected for genotype differences.

Determination of NTPDase1/CD39 activity

Activity was assessed in HIVECs based on ATP metabolism and inorganic phosphate (P_i) generation from breakdown of ATP and ADP (both at 1 and 25 μM). Activity was assessed up to 2 h in HIVECs on 24-well plates with 40,000 cells per well. The ATPase inhibitor POM-1 (Tocris Bioscience) caused substantial cell toxicity at the 0.01 to 10 μM doses imported from other studies; therefore, inhibitor studies were conducted using nonhydrolyzable nucleotide analogs (22).

ATP breakdown

After addition of exogenous 1 μM ATP in 0.5 ml of Ringer's buffer, the residual ATP at points up to 2 h was determined using the HSII kit (Roche, Basel, Switzerland). Fifty-microliter aliquots of supernatants were transferred into optical plates and light output was measured on a Glomax plate reader (Bio-Rad, Hercules, CA, USA). Calculation of unknowns was performed by standard curves and linear regression with GraphPad Prism 3.0 software (GraphPad Software, La Jolla, CA, USA). Control experiments indicated that no significant ATP breakdown occurred over 2 h in cell-free wells.

P_i generation

Medium was removed and cells were washed in Ringer's buffer. After incubation with 25 μM ATP and ADP in Ringer's, extracellular P_i release into supernatants was measured using a malachite green assay (Anaspec, Fremont, CA, USA).

Controls

Nonhydrolyzable analogs of ADP (25 μM α,β-methylene ADP; Sigma-Aldrich, St. Louis, MO, USA) and ATP (25 μM ATP-γS; Tocris Bioscience) were tested in parallel assays.

Quantitative immunofluorescence

Quantitative immunofluorescence (qIF) was adapted from published techniques where it was validated as being highly correlated with immunoblot data of isolated endothelial proteins (23). HIVECs were grown on coverslips, permeabilized, fixed, counterstained with DAPI, then stained with a monoclonal anti-NTPDase1 antibody (LS-C43407, 1 mg/ml at 1:25 dilution; Lifespan BioSciences, Seattle, WA, USA) and a secondary goat anti-mouse Alexa 488–labeled antibody (11001, 2 mg/ml at 1:200 dilution; Thermo Fisher Scientific). Expression was determined by fluorescent intensity for 10 HIVECs per genotype. HIVEC fluorescent intensity values were normalized to the mean fluorescent intensity for the same number of human umbilical vein endothelial cells (Lonza) similarly stained in parallel. Images were captured by a Photometrics CoolSnap ES2 camera (Roper, Tucson, AZ, USA) on a Nikon Eclipse 80I microscope (Nikon, Tokyo, Japan) with a Plan Fluor ×40 DIC M/N2 objective (Melville, NY, USA), and values in RLU were determined using NIS-Elements BR4.00.10 software (Nikon). All cells were stained in parallel for von Willebrand factor (VWF) to confirm their endothelial origin (FITC conjugated, 1:50 dilution; Lifespan BioSciences).

Assessment of CD39 protein content

HIVEC protein was isolated from cell lysates using a commercial kit (Active Motif, Carlsbad, CA, USA). Cytoplasmic CD39 protein in lysates was assessed using a commercial CD39 ELISA (Lifespan BioSciences) and was normalized to total protein as measured by Bradford assay (Thermo Fisher Scientific).

Platelet aggregation

Whole blood was drawn into vacuum tubes with 3.2% sodium citrate and spun for 15 min at 100 g to yield platelet-rich plasma (PRP). After PRP was removed, blood was centrifuged at 2400 g to yield platelet-poor plasma. Platelet aggregation was performed at 37°C by standard light transmission methods with a Chrono-Log 700 platform (Havertown, PA, USA) (24). PRP aliquots (adjusted with platelet-poor plasma to 250,000 cells/ml) were treated with serial concentrations of ADP (0, 1, 2.5, 3.75, 5, and 10 μm; Chrono-Log), the P2Y1-specific ADP analog MRS2365 (at 0.1, 1, 5, 10, and 100 μM; Tocris Bioscience), or vehicle (saline) for 7 min. Responses were measured as area under the curve with ImageJ software and normalized to a theoretical maximal area under the curve (% of potential maximal response). Only cells of AA and GG genotype were examined; it would have been futile to test AG if GG and AA were no different, moreover, a sample of 45 would have been needed for comparing AG and AA based on power calculations.

Flow cytometry of platelets

Blood was drawn into syringes containing 1/10 volume of 3.8% sodium citrate at room temperature. Acid-citrate-dextrose (ACD; 7.5 g citric acid, 12.5 g sodium citrate, and 10 g dextrose in 500 ml dH₂O) at 30°C was added, and blood was centrifuged at 200 g for 20 min. Supernatants (PRP) were combined and 12 μl prostacyclin (1 mg/ml in 50 mM Tris, pH 9.1) was added. PRP was centrifuged for 10 min at 1000 g. Pellets were resuspended in 1 ml modified Tyrode (mTyrode) buffer and 150 μl ACD. The volume was then increased with 25 ml of mTyrode and 3 ml ACD. Prostacyclin (12 μl) was added. The suspension was mixed and then centrifuged at 1000 g for 10 min. The pellets were resuspended with mTyrode (no ACD). Platelet numbers were determined using a Coulter counter, then diluted to 2 × 10⁴ cells/μl in mTyrode. Volumes (30 μl) were portioned into aliquots for antibody incubations, which were performed at room temperature for 5 min with the following primary antibodies at 10 μl per assay: CD39 (mAb-FITC 188-040, clone BU61; Ancell, Bayport, MN, USA), CD41 for costaining in all assays (559768, phycoerythrin-CY5; BD Biosciences Pharmingen, San Diego, CA, USA), and an isotype control antibody (Ancell). CD41 marks platelet signals (designated CD41⁺) but not debris or contaminating cells (designated CD41⁻). Cells were fixed for 10 min with 50 μl cold 2.5% paraformaldehyde; then 100 μl of fixed cells were diluted in 900 μl cold PBS for cytometry. Results are expressed in relative units of mean fluorescent intensity for all CD41⁺ cells.

Statistical analysis

Data were analyzed by GraphPad Prism 3.0 software except when noted. All analyses were 2 tailed, and P < 0.05 was considered significant. Comparison of group means was by Student's t test and Mann-Whitney U test. Values are expressed as means ± sem or sd. Odds ratios (ORs; by logistic regression) are expressed with 95% confidence intervals (CI). In the discovery cohort, significant differences in polymorphism frequencies between groups were analyzed by additive and dominant models with contingency tables (χ²) after Lewis (25). Hardy-Weinberg equilibrium (HWE) was determined as published in Guo and Thompson (26).

We anticipated that the discovery cohort would be predominantly white on the basis of hospital demographics; therefore, the a priori design was for analysis of both the entire discovery cohort (matched by gender and self-reported race to healthy controls) followed by a separate analysis of the white and African American subgroups. P values for intergroup differences in haplotype-tagged SNPs (htSNPs) are reported after a Bonferroni correction for multiple hypotheses (5, number of htSNPs). LD and haplotype analysis were performed in the discovery cohort; htSNP were determined pairwise by Haploview Tagger (Massachusetts Institute of Technology, Boston, MA, USA) but SNPs with a MAF of <5% were excluded. A CI rule was used to classify haplotype blocks (27). An r² value of 0.8 was the minimum LD for 2 polymorphisms to be defined as htSNPs. Adjustments for sex, race, and age were performed using logistic regression with SAS 9.4 software (SAS Institute, Cary, NC, USA).

For the replication study, analysis of the association between rs3176891 and rs10748643 and incident VTE in the HVH study was performed using logistic regression in Stata software (StataCorp, College Station, TX, USA). The primary replication analysis used an additive genetic model to characterize the association of rs3176891 with VTE. A secondary analysis was performed for rs3176891 and rs10748643 using a dominant model. Replication analyses were adjusted for study design variables and race.

RESULTS

Cohort characteristics

General demographics of the discovery and replication study participants are listed in Table 1.

Validation of previously reported ENTPD1 polymorphisms

Over 95% of individuals had interpretable sequence data in the analyzed regions. Eleven previously described ENTPD1 polymorphisms were identified in the discovery cohort; all were SNPs including the uncommon coding SNPs rs61731067 and rs3793744, and 9 intronic SNPs. Details of SNPs are listed in Supplemental Table S1. All were in HWE.

Novel ENTPD1 polymorphisms

Three novel ENTPD1 SNPs were detected and reported to NCBI; since then, all but Gly320Gly have been validated by others. These included a missense SNP (Val9Met) and a silent SNP (Gly320Gly), each seen in a single white control subject (Supplemental Table S1), and a SNP in the 3′-UTR (3UTR+44) in 2 African Americans with recurrent VTE (Fig. 1).

Figure 1.

Human gene (ENTPD1) encoding NTPDase1/CD39, including neighbors. All exons (rectangles) and introns (lines) are depicted. Exon sizes are relative to each other by width and intron sizes are relative to one another by width, but exons are expanded for clarity. On depicted chromosomal regions, direction of gene translation is indicated by arrows. Locations of SNPs of interest are depicted at bottom (haplotype-tagged SNPs are not shown). *Intron 1 SNP rs3176891 was associated with VTE and functional effects; SNPs to right of it are novel coding region SNPs found in this project. While not in sequenced regions of project, SNPs rs10748643 and rs3814159 are shown, as rs3814159 is promoter SNP in tight linkage disequilibrium with rs3176891 (r² = 0.935), while rs10748643, an intron 1 SNP, was previously associated with ENTPD1 gene expression and inflammatory bowel disease risk.

htSNPs for ENTPD1

Pairwise LD between all 14 SNPs identified by resequencing was examined in the entire discovery cohort (Supplemental Fig. S1). The presence of 2 haplotype blocks allowed for a set of 5 SNPs to be completely informative for common variation covering coding regions and exon–intron boundaries. htSNPs are listed in Supplemental Table S1. There were no apparent racial differences in allele frequencies of rs3176891 (as assessed between 78 whites and all 30 minority controls in each group; P = 0.17). A separate analysis of the white subjects did not reveal any difference in LD relationships based on race or ethnicity (data not shown). A SNP (rs10748643, Fig. 1) located outside of the sequenced domains was added to the project as described in Materials and Methods (14). In the discovery cohort, rs3176891 was not in tight LD with rs10748643 (r² = 0.72), though the G allele of rs3176891 was always seen with the rs10748643 G allele.

The ENTPD1 polymorphism rs3176891 is associated with VTE risk

Discovery cohort

rs3176891

The G allele was common, and its frequency was higher in the VTE group compared to the control group (allelic genetic model, 56 vs. 41%; Table 3). The association between rs3176891 and risk of VTE was significant with an OR of 1.88 (95% CI 1.08–3.25) in an additive allelic model (Table 3). This association reflects correction for multiple hypotheses and adjusting for race, age, and sex. Age was associated with VTE risk (Supplemental Table S2). All 4 subjects with CTEPH were carriers of the rs3176891 G allele (3 were GG homozygotes). This SNP was in HWE (P = 0.084 in all controls, 0.08 in non-Hispanic white controls). As controls were healthy with no history of cancer, adjustment for cancer was not possible. A subgroup analysis for African Americans also found that the rs3176891 G allele was associated with VTE (Supplemental Data). None of the other 4 htSNPs was associated with VTE risk.

TABLE 3.

Associations of ENTPD1 htSNPs with VTE in discovery and replication groups

SNP	VTE [n (%)]	Control [n (%)]	Genetic model
			Additive	Dominant
Discovery cohort
rs3176891
AA	24 (0.22)	33 (0.31)	OR 1.87	OR 1.73
AG	47 (0.44)	62 (0.57)	CI 1.08–3.25	CI 0.75–3.98
GG	37 (0.34)	13 (0.12)	P = 0.025*	P = 0.195
rs61731067
GG	95 (0.91)	98 (0.91)	OR 0.932	OR 0.96
GC	8 (0.08)	10 (0.09)	CI 0.36–2.42	CI 0.35–2.60
CC	1 (0.01)	0	P = 0.884	P = 0.93
rs3176886
AA	84 (0.78)	82 (0.76)	OR 1.26	OR 1.16
AG	20 (0.19)	20 (0.18)	CI 0.71–2.25	CI 0.60–2.27
GG	3 (0.03)	6 (0.06)	P = 0.43	P = 0.65
rs11188504
AA	56 (0.53)	40 (0.37)	OR 1.32	OR 1.87
AC	40 (0.37)	60 (0.56)	CI 0.77–2.28	CI 1.08–3.22
CC	11 (0.10)	8 (0.07)	P = 0.31	P = 0.12
rs3181123
CC	51 (0.47)	60 (0.56)	OR 1.31	OR 1.20
CT	48 (0.45)	41 (0.38)	CI 0.74–2.32	CI 0.60–2.39
TT	9 (0.08)	7 (0.06)	P = 0.35	P = 0.60
rs10748643
AA	21 (0.20)	30 (0.28)	OR 1.83	OR 1.96
AG	41 (0.38)	53 (0.50)	CI 1.1–3.06	CI 0.82–4.69
GG	45 (0.42)	24 (0.22)	P = 0.020*	P = 0.13

Replication study (Seattle HVH study)
rs3176891
AA	178 (0.22)	330 (0.28)	OR 1.08	OR 1.261
AG	417 (0.52)	552 (0.46)	CI 0.94–1.24	CI 1.01–1.59
GG	205 (0.26)	315 (0.26)	P = 0.26	P = 0.044*

Where numbers do not equal 108 in discovery group, few genotypes could not be determined. OR expressed as 95% CI. Analyses in discovery cohort were for entire group (all races). SNP appear in downward order based on 5′ to 3′ location, except for rs10748643. Primary analyses were additive genetic models. Data for secondary outcomes in discovery and replication studies are in Supplemental Tables S2, S3, and S4. All P values for discovery group reflect adjustment for multiple hypotheses (except for rs10748643, which was not htSNP), age, sex, and race. *Statistically significant, P < 0.05.

rs10748643

The G allele was common, and its frequency was higher in the VTE group compared to the control group (allelic genetic model, 61 vs. 47%; Table 3). In a secondary analysis, the association between rs10748643 and the risk of VTE was significant with an OR of 1.83 (95% CI 1.1–3.06) in an additive allelic model (Table 3). This association does not reflect correction for multiple hypotheses, as this SNP was not an htSNP, but it does reflect adjusting for race, age, and sex. Age was associated with VTE risk (Supplemental Table S2). This analysis was secondary because rs10748643 was not in the sequenced regions where the polymorphism discovery was performed. All 4 subjects with CTEPH were carriers of the rs10748643 G allele (3 were GG homozygotes). This SNP was also in HWE (P = 1.0 in all controls, 0.95 in non-Hispanic white controls). Because controls were healthy with no history of cancer, adjustment for cancer was not possible. A subgroup analysis for African Americans was not performed.

Replication study (Seattle HVH study)

rs3176891

In an a priori analysis using an additive model, rs3176891 was not associated with VTE. This was chosen as the most unbiased model, making no assumptions about the G allele’s biologic effect. However, in an a priori–defined secondary analysis using a dominant model, the G allele was associated with VTE (OR 1.26, 95% CI 1.01–1.59, Table 3) while adjusting for race and the study design variables age, index year, and hypertensive status. Risk estimates were similar in those who had a PE (with or without DVT) and those who had a DVT alone (Supplemental Tables S3 and S4). CTEPH and cancer data were not available. In this data set, rs3176891 was in HWE (among controls) and was in tight LD with rs10748643 (r² = 0.91).

rs3176891 genotype does not affect alternative RNA splicing

Real-time quantitative PCR products yielded a single band of the expected size using cDNA from HIVECs of both AA and AG genotypes (Supplemental Fig. S2).

Evaluation of LD partners for rs3176891 using international HapMap data

SNPs on chromosome 10 within 500 kb of rs3176891 that were in pairwise LD (r² > 0.80) were identified using the SNAP Web site for a SNP Annotation and Proxy Search (http://www.broadinstitute.org/mpg/snap/ldsearchpw.php). This analysis revealed 27 SNP that were in tight LD with rs3176891, and of these, no intronic SNPs were near splice junctions, and only a promoter SNP had an intriguing location, being in a proximal promoter location 860 nt 5′ of the translational start (rs3814159; r² = 0.935). Thus, rs3814159 seemed to be the most likely SNP to have functional effects in this haplotype. The G allele of rs3176891 is thus in tight LD with the G allele of rs3814159, forming a GG haplotype. In the discovery cohort, rs10748643 was not in tight LD with the intron 1 SNP rs3176891 (r² = 0.765). These SNPs are detailed in Supplemental Table S5. The SNP rs10748643 was also in LD with the promoter SNP rs3814159 (r² = 0.817); this pairwise LD was less than that seen for rs3176891-rs3814159.

Genotype of the promoter SNP rs3814159 affects binding of transcription factors to the ENTPD1 promoter

Multiple putative transcription factor sites were predicted in this region by online prediction software (Fig. 2A). In EMSA experiments, transcription factor–rich HIVEC nuclear extracts bound to oligonucleotides with the G allele more avidly (Fig. 2B). The only site present with the G allele but predicted as absent with the A allele was a degenerate p53 site (consensus RRRCATGYYY, where R is A or G, and Y is C or T; Fig. 2A). Sites present with the A allele but predicted as absent with the G allele were hepatocyte nuclear factor, TFII-I, and D-box binding PAR BZIP transcription factor (consensus AAGGATTAAC). Additionally, a reverse FOXP3 site beginning 4 nt 3′ of the rs3814159 polymorphism was predicted for both alleles. Of these, we could only identify literature implicating FOXP3 as a regulator of ENTPD1 transcription, so we performed EMSA experiments to evaluate for genotype-related changes in FOXP3 binding. There was no evidence of recombinant FOXP3 binding to this region, suggesting that this putative site is not an active FOXP3 site (Fig. 2B).

Figure 2.

A) Proximal promoter of ENTPD1 showing location of rs3814159 (−860A > C) in relation to translational start of gene, with predicted transcription factor sites for G and A alleles (alleles underlined). DNA sequence corresponds to oligonucleotides used in EMSA experiments. B) EMSA experiments testing binding of endothelial nuclear extracts and recombinant FOXP3 to oligonucleotide domain surrounding rs3814159, as shown in A. Nuclear extract–oligonucleotide complex formation (arrow) was more prominent with G allele (representative figure and analysis of 5 experiments shown) and was not seen with excess unlabeled probe (lane 3) or without nuclear extracts (lane 4). FOXP3 did not bind to oligonucleotides. C) Activity of luciferase reporter constructs containing proximal ENTPD1 promoter with either A or G allele of rs3814159. Reporter activity was assessed day after transfection of HEK293 cells and also 4 h after stimulation (+) with 250 μM 8-Br-cAMP. Luciferase activity was assessed in RLU and normalized to milligrams of total protein; data are presented as fold change in this ratio compared to unstimulated vector without ENTPD1 promoter insert (empty vector). *P < 0.05; **P < 0.001.

The G allele of the promoter SNP rs3814159 increases transcriptional activity of the ENTPD1 promoter

Utilizing luciferase reporter constructs containing the proximal ENTPD1 promoter and either the A or G allele of rs3814159, we found that the G allele was associated with an 8.1-fold increase in unstimulated transcriptional activity compared to the A allele in cultured human HEK293 cells (Fig. 2C). The rs3814159 SNP is located 860 nt 5′ of the translational start site and is thus termed A-860G. The −860A insert had a significant repressor effect on the existing minimal promoter of the empty reporter vector (pGL4.23luc2/minP), decreasing reporter activity by 21% (P = 0.03). Previously described functional SP1 and CREB sites located in ENTPD1 at −135 and −22,218 from the translational start, respectively, were purposefully excluded from the construct so as not play a role in the responses we assessed, and to test for the possibility of previously unappreciated cAMP response elements near rs3814159. Stimulation with the stable cAMP analog 8-Br-cAMP did not affect transcriptional activity for either allele, confirming that this proximal region of the ENTPD1 promoter is not cAMP responsive.¹⁹ The activity of the empty vector was significantly increased 1.9-fold by 8-Br-cAMP compared to PBS vehicle, suggesting a cryptic cAMP-binding site not appreciated by the manufacturer. Promoter activity in RLU was normalized to total protein content for each well.

ADP/ATPase activity is increased in HIVECs with the rs3176891 AG genotype

HIVECs with the rs3176891 heterozygous AG genotype displayed more metabolism of 1 μM exogenous ATP at all time points (Fig. 3A). These HIVECs also displayed more P_i generation at all time points (Fig. 3B). P_i generation and ATP metabolism were negligible in the presence of nonhydrolyzable analogs of ADP and ATP; P_i generation with nonhydrolyzable analogs was 6.5 ± 1.8 μM, mean ± sem, similar to the 6.8 ± 0.83 μM measured in cell-free wells at 2 h. We could not fully assess the effects of the rs10748643 genotype (discordant from rs3176891 in 4 cell lines), as its AA group had only 4 subjects. However, in 2 cell lines with the rs10748643 GG genotype, ATP metabolism and P_i generation values were at or below the mean values for rs3176891 AG heterozygotes (data not shown).

Figure 3.

NTPDase1/CD39 enzymatic activity in HIVECs based on genotype at rs3176891 as measured by ATP metabolism (A) and inorganic P_i generation from ATP/ADP (B); each was measured every 30 min up to 2 h. Cells of variant AG genotype (n = 10) had higher ATP metabolism and higher P_i generation values at all times compared to AA cells (n = 6). As negative controls, P_i generation and ATP metabolism were negligible when exogenous nonhydrolyzable nucleotides were used. Results represent 3 summed experiments for each cell line, grouped by rs3176891 genotype and shown as means ± sem. *P < 0.05.

Endothelial cell NTPDase1/CD39 surface protein expression is increased with the rs3176891 AG genotype

NTPDase1/CD39 surface protein expression in HIVECs was almost 2-fold higher in cells with the AG genotype compared to the AA genotype (Fig. 4A, B). Cytoplasmic CD39 protein levels in HIVECs (measured by ELISA) were also significantly higher in cells of the AG genotype (Fig. 4C). HIVECs with a GG genotype did not exist in the endothelial cell banks we used. All HIVEC lines demonstrated positive VWF staining, confirming the endothelial phenotype of these cells. Representative VWF staining is shown in Fig. 4A.

Figure 4.

Quantitative qIF staining for CD39 and cytoplasmic protein levels in HIVECs based on genotype at rs3176891. A) Characteristic staining of 2 cell lines based on genotype. Green signal is CD39 FITC staining, and blue signal is DAPI (nuclear stain). Characteristic staining for VWF (also using FITC-labeled antibody) is at right. B) Comparison of qIF group means by Student’s t test. Values in are expressed as in arbitrary RLU. C) Comparison of CD39 protein levels in HIVEC cytoplasm by genotype, normalized to total protein. None of the available cell lines had GG genotype. *P < 0.05.

Platelets of rs3176891 GG and AA genotypes have similar surface CD39

Unlike our experiments with endothelial cells, we were able to identify GG homozygotes for platelet studies by screening a larger number of blood donors. While CD39 is known to be minimally expressed on platelets, we tested its surface expression and found it was not significantly different between AA and GG genotype groups. Representative flow cytometry plots are included in Supplemental Fig. S3. There were also no differences in platelet number or platelet volumes between genotype groups (Supplemental Data).

Platelets of rs3176891 GG genotype demonstrate increased aggregation in response to a P2Y1-specific antagonist

GG platelets (vs. AA) had markedly increased reactivity to the P2Y1-specific agonist MRS2365 (a 190% increase at 5 μM; P < 0.05; Fig. 5A). The response to the nonselective purinergic agonist ADP did not differ between genotypes, though a trend for more aggregation in GG homozygotes was seen at the lowest dose (1 μM; Fig. 5B). Representative aggregation curves for each stimulus are included in Supplemental Figs. S4 and S5. Platelets from AG subjects were not studied, as we were primarily interested in demonstrating a phenotype difference with the most genetically different cells.

Figure 5.

Platelet aggregation based on rs3176891genotype in response to P2Y1-specific agonist (MRS2365) (A) and nonselective purinergic agonist (ADP) (B). Aggregation was increased by MRS2365 at doses of >1 μM in GG subjects, but ADP response did not vary with genotype. Results are expressed as means ± sem of area under curve (AUC) of individual assays. Comparison of group means is by unpaired Student’s t test; *P < 0.05, **P ≤ 0.01. C, D) Representative aggregation curves for 10 and 100 μM MRS2365 for AA (C) and GG (D) genotypes. Curves for ADP are in the Supplemental Data.

DISCUSSION

Using a candidate gene approach, we sequenced the coding and regulatory regions of the human gene for NTPDase1/CD39 (ENTPD1) to evaluate its common variants as risk factors for a phenotype known to be controlled by this enzyme: in vivo clotting. We chose VTE because it often occurs in younger subjects and those who often have few concomitant and potentially genetically confounding illnesses. This project ruled out any common ENTPD1 coding polymorphisms in these groups, but we found 2 rare missense variants. We found that the variant G allele of the ENTPD1 intron 1 SNP rs3176891 was modestly associated with VTE in 2 independent groups. The rs3176891 G allele was common, with an allele frequency of 51.7% in the VTE group of the replication study. Because we could not find direct effects of this intronic SNP on alternative RNA splicing, we then assessed publically available data to determine polymorphisms in LD with rs3176891 that were likely to have functional effects.

We found that the ENTPD1 promoter SNP rs3814159 was in strong LD with rs3176891. Transcriptional assays revealed that the rs3814159 G allele substantially increased promoter activity and was associated with increased binding of endothelial cell nuclear extracts to the region. While FOXP3 is a known regulator of ENTPD1 expression, we found no evidence that FOXP3 mediated the transcription effects at this locus. Allele-specific differences in binding of nuclear extracts to this region likely underlie the transcriptional differences we observed between alleles, but the specific transcriptional factors responsible remain to be identified. Consistent with this observed increase in transcriptional activity, carriers of the rs3176891 G allele had endothelial cells with increased NTPDase1 activity and protein expression. G allele carriers also had platelets with enhanced aggregation to a P2Y1 receptor agonist. Thus, using 2 human groups with a clotting phenotype, we elucidated an important genetic regulatory mechanism of NTPDase1/CD39 function that provides evidence for interindividual differences in activity of this enzyme. Such differences are important to the purinergic regulation of clotting and potentially relevant for the regulation of inflammation and regulatory T cell function (2, 28, 29). We suggest that the term “CD39 Denver” be used to more succinctly denote the gain-of-function phenotype of NTPDase1/CD39 due to the rs3814159 G allele and to recognize the region where this mechanism was described.

These data provide what is to our knowledge the first evidence of a functional promoter polymorphism in ENTPD1. Moreover, the genotype-specific effects of rs3814159 were not affected by cAMP stimulation, confirming the findings of others that the previously described augmentation of ENTPD1 transcription by cAMP occurs not in the proximal promoter of ENTPD1 but at upstream response elements (21). Such effects of cAMP in the intact gene could provide an important gene–environment interaction in the context of the rs3814159 G allele. We also tested the potential association of rs3176891 with the variant allele of the intron 1 SNP rs10748643; this SNP has no known direct functional effects and was not in the sequenced domains of our project, but it was previously linked with higher gene expression in leukocytes and with inflammatory bowel disease risk (14). In the discovery group and in 1000 Genomes Project data, these 2 SNPs were not in strong LD, but in the replication study they were, and rs10748643 was also associated with VTE in both study populations. It seems likely that neither of these intron 1 SNPs has direct functional effects, but rather that both serve as htSNPs (rs3176891 in particular) for the functional promoter variant rs3814159.

The role of platelets in VTE pathogenesis has been debated, but recent clinical trials demonstrated a benefit of the antiplatelet drug aspirin for primary VTE prevention after hip surgery and for secondary VTE prophylaxis (30–32). Our findings of a polymorphism linking VTE risk with increased platelet aggregation fit a paradigm where platelets play an important role in the pathogenesis of VTE (33). While the rs3176891G allele and its associated haplotype appear to be only a modest risk factor for VTE, they may be more important risk factors for VTE when they occur in the presence of other thrombophilias (e.g., factor V Leiden). On initial examination, a higher NTPDase1/CD39 activity should limit platelet aggregation by lowering ADP concentrations. In a model of PE tolerance, transgenic mice overexpressing NTPDase1/CD39 were less susceptible to death from PE (8). On the other hand, cooverexpression of NTPDase1/CD39 and P2Y1 in multiple cell lines has been shown to both increase ADP breakdown and up-regulate surface P2Y1 receptors, consistent with a classic ligand–receptor feedback loop (due to low ADP) (3).

In physiologic conditions, the expression of purinergic receptors in a given tissue should reflect ambient ADP and ATP levels and in turn local ATPase and ADPase activities. The increase in P2Y1-mediated platelet aggregation that we observed in rs3176891 GG homozygotes indicates P2Y1 receptor activation in platelets with this genotype, likely as a paracrine effect from altered luminal NTPDase1/CD39 activity. Such an increase in P2Y1 receptor activation on platelets would be predicted to increase clotting risk as ADP released from the granules of activated platelets will preferentially bind to platelet P2Y1 (dissociation constant, or K_D, in the nM range) before binding to endothelial CD39 (enzymatic Michaelis constant, or Km, in μM range) (34, 35). However, because platelet P2Y12 receptors are both more numerous and more potent mediators of platelet aggregation than platelet P2Y1 receptors, it seems likely that the increased NTPDase1 activity we observed with the rs3176891 G allele did not affect P2Y12 signaling, as ADP reactivity was not different between genotype groups. This also correlates with the modest risk for VTE we observed with the rs3176891 G allele, as P2Y1 signaling was affected while the dominant P2Y12-mediated platelet aggregation pathway was not affected. Further research will be needed to clarify the effects of the rs3176891G gain-of-function haplotype on P2Y12 and P2Y1 signaling in platelets.

To date, only a few large genomewide association studies in VTE populations have been published, confirming the accepted roles of common variants in the factor V, prothrombin, and ABO blood group genes (13, 17, 36–38). One study in a French population found that chromosome 10 was the second to third most important chromosome for VTE risk, contributing 4% of the overall risk (chromosome 1, containing the factor V and antithrombin loci, contributed only 3% of risk) (13). The implicated regions in that study were far away from ENTPD1, though the SNP array utilized included 2 ENTPD1 SNPs that are in LD with rs3176891 (rs4582902, rs10882675; Supplemental Table S5). Another genomewide association study utilized a SNP array that had one SNP in LD with rs3176891 (rs4582902), but it was not associated with VTE. Further evaluation of implicated ENTPD1 polymorphisms in large VTE populations, in other races, and in other thrombotic diseases will be important.

Our study has limitations. While our discovery group was large enough to find all common coding variants in ENTPD1 by resequencing, it was small for an association analysis. While such limitations are common in studies of disease risk (39), we addressed this by replication in a larger VTE study. Our replication study group was mostly women and also predominantly white; thus, future replication of these data will be important in other races including African Americans, who in the discovery cohort appeared to have a higher OR for VTE risk with the rs3176891 G allele compared to whites. While we confirmed genotype-related differences in CD39 protein levels with 2 assays, further investigation is needed to understand the transcriptional control mechanisms that lead to increased CD39 protein on the cell surface. Like other reports, we found that the cell surface expression of CD39 by immunocytochemistry paralleled its cytoplasmic expression (23). Replication of our findings in other diseases where NTPDase1/CD39 activity is postulated to play a major role (particularly diseases of inflammation) and study of endothelial cells from GG carriers are also needed.

In summary, we found that the ENTPD1 intron 1 SNP rs3176891 is associated with a modest but significant increase in VTE risk in 2 predominantly white populations. This association appears to be related to a functional LD partner—the promoter SNP rs3814159—whose G allele increases ENTPD1 transcriptional activity, thereby augmenting NTPDase1/CD39 protein levels and increasing ATP/ADPase activity (Fig. 6). This SNP deserves further evaluation as a risk factor for diseases where purinergic signaling is important.

Figure 6.

Diagram of how common genetic haplotype in ENTPD1 gene leads to gain-of-NTPDase1-function phenotype with downstream effects on purinergic pathways (CD39 Denver). Phenotype of increased clotting and P2Y1 receptor activation was shown in this project, but this genetic variation may also help explain interindividual differences in inflammation and immune regulation, given the critical role of NTPDase1/CD39 expression in these pathways.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

ACD

acid-citrate-dextrose

CI

confidence interval

CTEPH

chronic pulmonary thromboembolic disease with pulmonary hypertension

DVT

deep venous thrombosis

EMSA

electromobility shift assay

ENTPD1

ectonucleoside triphosphate diphosphohydrolase 1

FOXP3

forkhead box P3 protein

GFP

green fluorescent protein

HIVEC

human iliac vein endothelial cell

htSNP

haplotype-tagged single nucleotide polymorphism

HVH

Heart and Vascular Health

HWE

Hardy-Weinberg equilibrium

LD

linkage disequilibrium

MAF

minor allele frequency

mTyrode

modified Tyrode

NCBI

National Center for Biotechnology Information

NTPDase1

nucleoside triphosphate diphosphohydrolase 1

OR

odds ratio

PE

pulmonary embolism

P_i

inorganic phosphate

PRP

platelet-rich plasma

qIF

quantitative immunofluorescence

RLU

relative light unit

SNP

single nucleotide polymorphism

VTE

venous thromboembolism

VWF

von Willebrand factor

AUTHOR CONTRIBUTIONS

J. Maloney, B. Branchford, K. Maloney, K. Moreau, N. Smith, U. Broeckel, and J. Di Paola designed research; J. Maloney, D. Calabrese, W. Zhang, K. Wiggins, and N. Smith analyzed data; J. Maloney, B. Branchford, G. Brodsky, M. Cosmic, D. Calabrese, C. Aquilante, J. Gonzalez, K. Wiggins, and N. Smith performed research; J. Maloney, B. Branchford, C. Aquilante, K. Maloney, K. Wiggins, N. Smith, and J. Di Paola wrote the article.