Selectin haplotypes and the risk of venous thrombosis: influence of linkage disequilibrium with the factor V Leiden mutation


Rogier M. Bertina, Department of Thrombosis and Hemostasis, Einthoven Laboratory for Experimental Vascular Medicine, Albinusdreef 2, 2333 ZA, PO Box 9600, 2300 RC, Leiden, the Netherlands.
Tel.: +31 71 526 1893; fax: +31 71 526 6755; e-mail:


Summary. Background: Selectins (E-, L- and P-selectin) and their most important counter-receptor P-selectin glycoprotein ligand (SELPLG) facilitate the interaction of platelets, leukocytes and endothelial cells at inflammatory sites. Selectin polymorphisms/haplotypes have been associated with cardiovascular disease. Objectives: We investigated the association between haplotypes (H) of these four genes and deep venous thrombosis (DVT) risk. We additionally explored the effect of linkage disequilibrium (LD) with the nearby Factor V Leiden mutation (FVL). Furthermore, interactions between SELPLG polymorphisms and selectin polymorphisms were investigated. Patients/methods: Leiden Thrombophilia Study (LETS) subjects were genotyped for 24 polymorphisms by TaqMan or PCR–RFLP, detecting all common haplotypes in four blocks. P-selectin was analyzed in two blocks, upstream (SELPup) and downstream (SELPdown) of the recombination hotspot. Results: In E- and L-selectin, none of the haplotypes was associated with DVT risk. In SELPup, H2-carriers had a 1.3-fold increased risk (95% CI, 1.0–1.7), whereas H4-carriers had a 1.4-fold decreased risk (95% CI, 0.5–1.0). In SELPdown, H2-carriers had a 1.3-fold increased risk (95% CI, 1.0–1.7). Because of LD with FVL, we subsequently excluded all FVL-carriers and all risks disappeared. Mutual adjustment within a logistic regression model resulted in disappearance of the risks for the SELP haplotypes, whereas FVL risk remained. Conclusions: After adjustment for LD with FVL, none of the selectin haplotypes was associated with DVT risk, showing that the increased risks of the selectin haplotypes were a reflection of the effect of FVL on thrombosis risk.


The initial attachment and subsequent movement of leukocytes and platelets to vascular surfaces is in part mediated by selectins, a family of three vascular cell adhesion molecules (CD62) [1,2]. This family of type I membrane proteins includes E-, L- and P-selectin, encoded by SELE, SELL and SELP, respectively. Recent data support a role of selectins and their ligands in hemostasis and thrombosis [3,4]. In mouse models, P- and E-selectin seem to be critical for the recruitment of leukocytes to venous thrombi and overexpression of P-selectin in mice induces a pro-coagulant state [5,6]. Celi et al. [7] reported that P-selectin can induce tissue factor (TF) expression in human monocytes. Furthermore, there seems to be a major role for P-selectin and P-selectin glycoprotein ligand (SELPLG) in the accumulation of TF in the thrombus and the subsequent generation of fibrin via recruitment of TF bearing microparticles and leukocytes [4]. In addition, elevated plasma levels of soluble P-selectin have been shown to be associated with increased venous thrombosis risk [8–10].

P-selectin is stored in platelet alpha-granules and in Weibel–Palade bodies of endothelial cells and is translocated to the surface when these cells are activated [11]. P-selectin and the endothelial E-selectin are essential for leukocyte attachment to, and movement on, the monolayer of vascular endothelial cells [12], and both play a role in the accumulation of inflammatory cells and fibrin generation after venous thrombosis [5,13]. L-selectin, which is constitutively expressed at the leukocyte cell surface, is critical for the interaction of lymphocyte homing to the lymphatic organs. In addition, it triggers the adhesion of circulating leukocytes on activated endothelial cells, thereby amplifying inflammatory reactions [14]. In these processes, the best defined and biologically most important counterreceptor on leukocytes and platelets is SELPLG [15].

The genes encoding the selectins are localized in a cluster of 160 kb on the long arm of chromosome 1, 1q24-q25, closely linked to the gene for coagulation factor V. SELE contains 14 exons and spans over 11 kb of DNA, while SELL is larger, extending over 20 kb, but consisting of just nine exons. SELP spans > 50 kb and contains 17 exons, most of which encode structurally distinct domains. SELP is located just upstream of the gene for coagulation factor V. All genes are transcribed in the same direction. The SELPLG gene maps on chromosome 12q24, spans approximately 13 kb, and contains two exons and one intron, with the complete coding sequence residing in exon 2.

All four genes contain many single nucleotide polymorphisms (SNPs) and there are recombination hotspots present within the SELP and SELPLG genes [16,17]. The majority of studies on polymorphisms or haplotypes of the selectin genes focused on arterial disease and did not include the entire gene cluster [18–24]. The present study aimed at investigating the association between haplotypes of the three genes of the selectin cluster and their main counter-receptor and the risk of deep venous thrombosis (DVT). As the gene for coagulation factor V (F5) is located directly downstream of SELP, and harbors the most important genetic risk factor for DVT, the factor V Leiden mutation (FVL) [25], we additionally explored the degree of linkage between polymorphisms in the selectin genes and the FVL polymorphism and adjusted risk estimates for this linkage.


Study population

The design of the Leiden Thrombophilia Study (LETS) has been described in detail previously [26,27]. In short, 474 consecutive patients with an objectively confirmed first episode of DVT and 474 controls, frequency matched for sex and age, were included. Patients were all younger than 70 years and individuals with malignancies were excluded. Acquaintances and partners of patients without a history of both DVT and cancer served as controls. Mean age for patients and controls was 45 years (range: 15–69 for patients, 15–72 for controls). Both groups consist of 202 (42.6%) men and 272 (57.4%) women. Venous blood was collected into 0.1 volume of 0.106 mol L−1 trisodium citrate. Plasma was prepared by centrifugation for 10 min at 2000 × g at room temperature and stored at −70 °C. High molecular weight DNA was isolated from leukocytes by standard methods and stored at −20 °C. DNA samples were available from 471 patients and 471 controls.

Genetic analysis

SeattleSNPs re-sequenced the genes of the selectin cluster and SELPLG in 23 individuals of European–American descent [16]. For each gene, haplotypes for the 46 chromosomes were reconstructed from the unphased SNP genotype data, using the software phase 2.0 [28]. We used these data to select 24 SNPs (Table 1) to cover the 23 common haplotype groups of the genes. Because of a recombination hotspot in SELP intron 8, we analyzed the regions upstream (SELPup) and downstream (SELPdown) from the recombination hotspot separately. Genotyping for SELE SNP 135A>G was performed by PCR–RFLP. All other SNPs were genotyped using the 5′ nuclease/TaqMan assay. Polymerase chain reactions with fluorescent allele-specific oligonucleotide probes (Assay-by-Design/Assay-on-Demand, Applied Biosystems, Foster City, CA, USA) were performed on a PTC-225 thermal cycler (Biozym, Hessisch Oldendorf, Germany) and fluorescence endpoint reading for allelic discrimination was done on an ABI 7900 HT instrument (Applied Biosystems). Information on primer and probe sequences and restriction enzyme used is available upon request. The FVL mutation [1691G>A (rs6025)] was genotyped previously [29].

Table 1. SELE, SELL, SELP and SELPLG polymorphisms, rs numbers and minor allele frequencies in patients and controls
Gene (GenBank accession number)SNP*rs numberPatients (= 471)Controls (= 471)
  1. Minor alleles shown in bold and underlined.

  2. *SNP numbering according to SeattleSNPs (

SELE (AF540378)135A>Grs39173890.4160.401
SELL (AY233976)4318C>Trs49872840.1300.122
SELP (AF542391)
SELPLG (AY331789)7436C>Trs49642690.4620.442

Statistical analysis

In healthy controls, Hardy–Weinberg equilibrium for each SNP was tested by chi-squared analysis. We used the software program tagsnps (V2) [30], which minimizes the uncertainty in predicting common haplotypes for individuals with unphased genotype data, to calculate the number of haplotypes for each gene, the frequency of the haplotypes and the statistic R2h, the squared correlation between true haplotype dosage (0, 1 or 2 copies of a haplotype) and the haplotype dosage predicted by tagsnps [31]. A high R2h indicates that a haplotype could be assigned with high certainty. We assigned haplotypes individually to the patient and control subjects. An R2h > 0.95 and overall haplotype frequency ≥ 1% were used as criteria for assigning haplotypes of SELE, SELL and SELP (Fig. 1A). Because a recombination hotspot covers part of the SELPLG gene, haplotype construction for this gene was less accurate. Only one of the 11 set SELPLG haplotypes had an R2h > 0.95 and frequency ≥ 1%. Consequently, no haplotypes were constructed for SELPLG, and the SNPs were analyzed separately.

Figure 1.

 Haplotype structure of the selectin genes. (A) Haplotype groups and typed SNPs (bold and circled) of SELE, SELL and SELP (upstream and downstream of the recombination hotspot) upstream of the factor V gene. SNP numbering according to SeattleSNPs ( F, haplotype allele frequency in LETS population (tagsnps results); R2h, squared correlation between true and predicted haplotype dosage. SELE H5 tagging-SNP 3689A>C encodes Ser128Arg, SELPup H2 tagging-SNP SNP 20732G>A encodes Ser290Asn, SELPdown H5 tagging-SNP 36279G>T encodes Val599Leu and SELPdown H6 tagging-SNP 37674A>C encodes Thr715Pro. (B) haploview LD plot of E-, L- and P-selectin (upstream and downstream of the recombination hotspot), together with the factor V Leiden mutation. LD (number shown in squares) ranges from 0 (D′ = 0) to 100 (D′ = 1). The color of the squares indicates the LOD-score, with white color for LOD < 2 and gray color for LOD ≥ 2 (the higher the LOD-score, the darker the gray color).

To investigate whether SNPs and haplotypes of the selectin genes were associated with DVT risk, odds ratios (ORs) and 95% CI were calculated.

The degree of linkage disequilibrium (D′) between the SNPs of the selectin genes and FVL was estimated using the software program haploview [32]. This analysis indicated that significant linkage existed between some selectin SNPs and FVL. To adjust for the effect on risk due to this linkage, risks were calculated after exclusion of all FVL carriers. Furthermore, FVL and the selectin haplotypes were adjusted for each other in one logistic regression model.

The program tagsnps (V2) was used to calculate the frequencies of haplotypes formed by coding SNPs of SELPup and SELPdown. Subsequently haplotypic ORs were calculated with the most common haplotype as reference category.

Interactions between coding SNPs of SELPup and SELPdown were analyzed using tagsnps (V2), which estimates frequency-based haplotypic ORs. Interactions between SELPLG SNPs and SNPs of SELE, SELL and SELP were tested under a dominant model within the logistic regression model.



The re-sequencing data from SeattleSNPs were used to select 24 SNPs to cover all common haplotype groups of the four selected genes (Fig. 1A). Table 1 shows the minor allele frequencies of the 24 genotyped SNPs in 471 patients and 471 controls. Only for SELP SNPs 20732G>A and 28972G>A there was a small deviation from Hardy–Weinberg equilibrium (= 0.037 and 0.027, respectively). DVT risks for the separate SNPs of SELE, SELL and SELP are shown in Tables S1, S2 and S3 in Supplementary material.

Linkage disequilibrium

SELE, SELL and SELP are located in a large cluster on chromosome 1q. We estimated the degree of LD between the SNPs of these genes in our study population (Fig. 1B). haploview analysis showed that SELE and SELL are present in a single haplotype block with a high degree of LD (D′ ranges from 0.42 to 1.00), indicating that recombination in this region is rare. In SELP a clear recombination hotspot is present, dividing the gene into two parts. Within each part, the degree of LD is high. In SELPup, D′ ranges from 0.83 to 0.94, in SELPdown, D′ ranges from 0.65 to 1.00. In addition, SELPup forms a haplotype block together with SELE and SELL, although the degree of LD is somewhat lower (D′ ranges from 0.00 to 0.90).


Based on our genotyping data in 942 subjects, tagsnps analysis resulted in nine SELE haplotypes, 13 SELL haplotypes, 12 haplotypes for SELPup, 18 haplotypes for SELPdown and 11 haplotypes for SELPLG. Implementing the criteria used for individually assigning haplotypes (R2h > 0.95 and frequency ≥ 1%) resulted in assignment of five SELE haplotypes, six SELL haplotypes, five haplotypes for SELPup and six haplotypes for SELPdown (Fig. 1A). As SELE haplotype 1 (H1) had a frequency < 1%, it was excluded from further analysis. Because of the criteria used, we did not assign haplotypes to seven patients and 15 controls for SELE, to 19 patients and 19 controls for SELL, to 22 patients and 22 controls for SELPup and to 50 patients and 58 controls for SELPdown. These individuals were excluded from further haplotype analysis for these genes or regions.

In Table 2, crude ORs and 95% CIs for the haplotypes of SELE, SELL and SELPup and SELPdown are shown. In SELE and SELL none of the haplotypes was associated with DVT risk. In SELPup, H2-carriers (H2Hx + H2H2) had a slight increase in risk (OR, 1.3; 95% CI, 1.0–1.7), whereas H4-carriers had a slightly decreased risk (OR, 0.7; 95% CI, 0.5–1.0). These effects already had been detected in the single SNP analyses (Table S3, SNPs 20732G>A and 14668A>G). In SELPdown, H2-carriers had a slight increase in risk (OR, 1.3; 95% CI, 1.0–1.7), whereas H6-carriers had a slight decrease in risk (OR, 0.7; 95% CI, 0.5–1.1), although this protection was not significant.

Table 2.   Thrombosis risk for the haplotype groups of SELE, SELL and the upstream and downstream part of SELP
HaplotypeSELESELLSELP upstreamSELP downstream
Patients (n = 464)Controls (n = 456)OR95% CIPatients (n = 452)Controls (n = 452)OR95% CIPatients (n = 449)Controls (n = 449)OR95% CIPatients (n = 421)Controls (n = 413)OR95% CI
Haplotype 1
  1. All odds ratios were calculated with HxHx as the reference category (OR = 1); Hx indicates all haplotypes but the one given; NA, not applicable.

 H1HxNANANA 65770.80.6–1.215916610.7–1.388940.90.7–1.3
 H1H1NANANA 460.60.2–2.334251.40.8–2.4951.70.6–5.3
H1Hx/H1H1NANANA 69830.80.6–1.119319110.8–1.3979910.7–1.3
Frequency H1    0.0810.098  0.2530.241  0.1260.126  
Haplotype 2
Frequency H20.4090.390  0.4080.433  0.2120.177  0.3420.303  
Haplotype 3
Frequency H30.3320.355  0.2710.241  0.3910.423  0.0870.085  
Haplotype 4
Frequency H40.1100.102  0.0320.031  0.0910.119  0.2430.255  
Haplotype 5
Frequency H50.1310.126  0.1310.124  0.0530.040  0.1090.127  
Haplotype 6
 H6Hx16200.80.4–1.5646210.7–1.5NANANA 43550.70.5–1.1
 H6H602321.50.3–9.1NANANA 360.50.1–1.9
H6Hx/H6H616220.70.4–1.467641.10.7–1.5NANANA 46610.70.5–1.1
Frequency H60.0170.026  0.0770.073      0.0580.081  

Linkage disequilibrium with FVL

Since F5 is located only 2.4 kb downstream of SELP and contains the most important risk factor for DVT, FVL, we explored the degree of linkage between SNPs in the selectin genes and FVL. Although F5 is located downstream from SELP and separated from SELP by a recombination hotspot in F5, we observed significant linkage between FVL and SNPs in all three selectin genes (Fig. 1B). Because of this high degree of linkage with FVL, a risk association of a selectin haplotype may reflect the effect of FVL on thrombosis risk (OR = 8.0, 95% CI, 4.5–14.2 in LETS).

To account for this LD effect, we first excluded all FVL carriers from the analyses; 14 heterozygous controls (control frequency 1.5%), 84 heterozygous patients and eight homozygous patients (patient frequency 10.6%). We found that the odds ratios for SELPupH4 and SELPdownH6 slightly increased (OR = 0.8, 95% CI, 0.6–1.1 and OR = 0.8, 95% CI, 0.5–1.3, respectively) due to linkage of the common allele of this haplotype to FVL. The risks for SELPupH2 and SELPdownH2 completely disappeared (OR = 1.0, 95% CI, 0.8–1.4 and OR = 1.0, 95% CI, 0.7–1.3, respectively), showing that the earlier observed risks were a reflection of the effect of FVL on thrombosis risk. The FVL allele was found in all SELPup and SELPdown haplotypes (see also Fig. 1B), but with the highest frequencies in the H2 haplotypes. For SELPdownH2 carriers, 6.5% of the controls carried FVL, and 32.7% of the patients. For SELPupH2 carriers, 5.3% of the controls carried FVL, and 30.7% of the patients. Mutual adjustment of FVL and either SELPup or SELPdown haplotypes in one logistic regression model gave the same results for the SELP haplotypes as after exclusion of all FVL carriers (i.e. disappearance of the thrombosis risk), whereas the risk of FVL remained (data not shown).


As haplotype assignment for SELPLG was not accurate, the SNPs were analyzed separately. Table 3 shows the DVT risk for SELPLG SNPs. None of the SELPLG SNPs was significantly associated with thrombosis risk.

Table 3.   Thrombosis risk for SELPLG SNPs
SNPGenotypePatients (= 471)Controls (= 471)OR 95% CI
  1. All odds ratios were calculated with homozygous wild-type as the reference category (OR = 1).

CT + TT3433251.20.9–1.5
CT + TT2582611.00.8–1.3
AG + GG1501451.10.8–1.4
TA + AA2903110.80.6–1.1


As haplotypes were generated independently for an upstream and downstream element of SELP, we need to consider the possibility of interaction between SNPs in these two regions, especially where it concerns the coding SNPs S290N (20732G>A, SELPupH2), N562D, V599L (36279G>T, SELPdownH5) and T715P (37674A>C, SELPdownH6) [33]. Therefore, we used tagsnps (V2) to set haplotypes using these four SNPs and we calculated frequency-based haplotypic ORs. Instead of SNP N562D, we used SNP 36060C>A for these calculations, because according to SeattleSNPs, SNPs 36060C>A and N562D are in LD. There were eight frequency-based haplotypes constructed from the genotypes of the four coding SNPs in SELP. In our study population, the combination of the rare alleles of SNPs 20732G>A and 36060C>A, haplotype AAGA, was associated with a 1.6-fold increase in DVT risk (Table S4). After exclusion of FVL carriers the risk for this haplotype disappeared (OR, 0.9; 95% CI, 0.6–1.3).

AS SELPLG is the main counterreceptor for each of the three selectin genes, we investigated possible interactions between SNPs of SELPLG and SELE, SELL and SELP. No interactions between SNPs of the different genes were found, because none of the ORs exceeded the risk of the SNPs themselves (data not shown).


We investigated the effect of haplotypes of SELE, SELL, SELP upstream and downstream of the recombination hotspot and SNPs in SELPLG on DVT risk. In SELE and SELL, none of the haplotypes was associated with risk. In SELPup, H2-carriers had a 1.3-fold increased risk, whereas H4-carriers had a 1.4-fold decreased risk. In SELPdown, H2-carriers had a 1.3-fold increased risk, whereas H6-carriers had a 1.4-fold decreased risk. After adjustment for linkage with FVL, the crude risk associations of the selectin haplotypes disappeared, whereas the effect of FVL on DVT risk remained. This shows that the risk associations found for the selectin haplotypes were caused by linkage to FVL.

In the risk analyses, we used subjects with genotype HxHx (Hx: all other haplotypes but the one being analyzed) as reference group. Using subjects homozygous for the most common haplotype of the haplotype block as reference group, thereby avoiding the use of different reference groups, did not result in different odds ratios, but gave wider confidence intervals, because the reference groups were smaller (data not shown).

Several polymorphisms and haplotypes of selectin genes have been studied in cardiovascular disease. SELPupH2, in this study associated with a 1.3-fold increased DVT risk, is tagged by SNP 20732G>A. This polymorphism is located in exon 7 and confers a serine to asparagine change at position 290. The haplotype carrying the rare allele of this polymorphism increased the risk of myocardial infarction 2- to 3-fold in populations in Belfast and France [21].

SELP polymorphism 37674A>C is located in exon 13 and confers a threonine to proline change in position 715. SELP 37674C has been found to be protective against myocardial infarction [18,21,34], and to be associated with low levels of soluble P-selectin [19,20,35]. In our study this polymorphism tagged SELPdownH6 (Fig. 1A), which was associated with a slight decrease in risk. Very recently, Ay et al. [36] reported on the effect of SELP haplotypes on the risk of recurrent venous thrombosis. This study reports a lower frequency of 715 Pro carriers among patients with recurrent thrombosis (17/116) than among healthy subjects (28/129). In our study only patients with a first venous thrombotic event were included.

Volcik et al. [33] recently showed that certain combinations of P-selectin S290N (20732G>A, SELPupH2), D562N, V599L (36279G>T, SELPdownH5) and T715P (37674A>C, SELPdownH6) modulate the risk of CHD. They showed that haplotype NNVP was associated with an approximately 2-fold increased risk of CHD. In our study of venous thrombosis, this haplotype was associated with a slight decrease in risk, whereas haplotype AAGA (NDVT in Volcik et al.) increased the risk 1.6-fold. After adjustment for FVL in this haplotype, the risk disappeared.

Another polymorphism of interest is the Ser128Arg in SELE, corresponding to polymorphism 3689A>C in our study. This polymorphism has been found to be functional, as it alters ligand affinity [37], enhances tethering of myeloid cells [38], and regulates leukocyte endothelial interaction in vitro [39]. Additionally, it is associated with enhanced endotoxin-triggered, tissue factor-mediated coagulation in humans [40]. The Arg-allele has been associated with atherosclerosis [41–43], myocardial infarction [39] and restenosis after angioplasty [44]. Homozygosity for the Arg-allele was associated with increased risk of recurrent venous thromboembolism [45]. In contrast, in our study of first events of DVT, homozygous carriers of SELE H5, tagged by the rare allele of this polymorphism, had a non-significant reduction in risk (OR, 0.7; 95% CI, 0.3–1.9), indicating that SELE H5 might be protective against a first event of DVT.

In conclusion, after adjustment for LD with FVL, none of the selectin haplotypes was associated with VT risk, showing that the increased risks of the selectin haplotypes were a reflection of the effect of FVL on thrombosis risk.


This study was financially supported by grant 912-02-036 from the Netherlands Organization for Scientific Research (NWO). The LETS study was supported by grant 89-063 from the Netherlands Heart Foundation.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.