Soluble P-Selectin, SELP Polymorphisms, and Atherosclerotic Risk in European-American and African-African Young Adults
The Coronary Artery Risk Development in Young Adults (CARDIA) Study
Objective— To characterize the genetic and clinical correlates of soluble P-selectin, and the relationship of P-selectin to atherosclerotic risk, in young European-American (EA) and African-American (AA) adults.
Methods and Results— We assessed the interrelationships between 25 common SELP polymorphisms, soluble P-selectin, and atherosclerotic risk in 1222 EA and 1072 AA from the longitudinal population-based CARDIA study. Male sex, smoking, blood pressure, and metabolic status were strong cross-sectional correlates of soluble P-selectin among CARDIA subjects aged 33 to 45 years, explaining 13% of the variance. Among EAs, higher soluble P-selectin predicted carotid intima-media thickness (IMT) measured 5 years later, even after accounting for traditional risk factors. Common SELP nucleotide sequence variants explained 11% and 5% of the interindividual variation in soluble P-selectin levels in EAs and AAs, respectively. Four distinct variants contributed to P-selectin phenotype in EAs, including a polymorphism of the 5′ SELP haplotype block associated with carotid IMT. Half of the phenotypic variation attributable to SELP in EAs could be explained by the Thr715Pro polymorphism, whereas Val599Leu was more strongly associated with soluble P-selectin among AAs.
Conclusions— Common SELP polymorphisms were associated with soluble P-selectin and carotid IMT in young adults, but the patterns of association differed between EAs and AAs. These results support the role of P-selectin in the preclinical stages of atherosclerosis.
P-selectin (CD62P) is a cell adhesion receptor expressed mainly at the surface of platelets and endothelial cells and plays an important role in the early stages of atherosclerosis.1 A soluble form of P-selectin lacking the transmembrane domain circulates in plasma. Increased soluble P-selectin levels are associated with various cardiovascular risk factors and occurrence or severity of acute and chronic cardiovascular disorders, primarily among older adults.1–6⇓⇓⇓⇓⇓ In some7 but not all8 cohorts of apparently healthy individuals, higher soluble P-selectin levels predicted future clinical cardiovascular events. Cross-sectional studies suggest soluble9 or platelet10 P-selectin is significantly correlated with subclinical atherosclerosis, even after accounting for known risk factors. Prospective data on soluble P-selectin as a predictor of early atherosclerotic disease, however, have not been previously reported.
The heritability of circulating P-selectin levels has been estimated between 45% and 70%.4,5,11⇓⇓ The Thr715Pro polymorphism (rs6136) of the P-selectin gene (SELP) has been associated with lower soluble P-selectin levels, accounting for ≈10% to 20% of the variation in healthy European-American (EA) populations.12–16⇓⇓⇓⇓ Other SELP polymorphisms, including Val599Leu (rs6133), appear to account for some of the residual variance in soluble P-selectin among EAs.5,12,16⇓⇓ The Thr715Pro polymorphism is rare in people of African descent,14,15⇓ but little is known about determinants of soluble P-selectin among African-Americans (AA) or allelic heterogeneity of other SELP polymorphisms between population groups.
Using data from a large well-characterized population of young EA and AA adults recruited as part of the longitudinal Coronary Artery Risk Development in Young Adults (CARDIA) study, we assessed (1) cross-sectional relationships among multiple common SELP polymorphisms, soluble P-selectin levels, and atherosclerotic risk factors; and (2) the ability of soluble P-selectin and SELP genotype to predict future carotid atherosclerosis.
Study Participants, Soluble P-Selectin Levels, SELP Genotyping, and Carotid Ultrasound
CARDIA study recruitment and data collection procedures are described under Supplemental Methods (available online at http://atvb.ahajournals.org). As part of the Young Adult Longitudinal Trends in Antioxidants (YALTA) study, an ancillary study to CARDIA, soluble P-selectin levels were measured and genotyping for the Thr715Pro polymorphism performed on 2950 participants who attended the year 15 examination (age range 33 to 45 years). Soluble P-selectin levels were measured by enzyme-linked immunosorbent assay (R & D systems Inc; Cat No: BBC 6) using a 1:5 dilution of plasma at the University of Minnesota. Genotype data for 24 additional SELP gene SNPs were available on a subset of 2294 of these YALTA study participants through the Inflammation Genomics and Atherosclerosis project (IGAP), another CARDIA ancillary study. SELP polymorphism discovery, SNP selection, and genotyping are described under Supplemental Methods. Carotid intimal-medial wall thickness (IMT) was determined by B-mode ultrasound at the CARDIA year 20 examination on 1911 eligible participants, as described under Supplemental Methods.
Multiple linear regression was used to determine the cross-sectional associations of cardiovascular risk factors and other participant characteristics with year 15 soluble P-selectin as the dependent variable, adjusted for other covariates. Linear regression coefficients indicate the mg/L change in soluble P-selectin per indicated unit (SD for continuous variables or category for discrete variables). Multiple linear regression models were similarly used to determine the association between quartile of soluble P-selectin measured at year 15 and mean standardized IMT measured at year 20 as the dependent variable, adjusted for other year 15 participant characteristics and risk factors.
Hardy–Weinberg equilibrium and pair-wise linkage disequilibrium statistics (r2 and D′) were calculated using JLIN.17 Percentage of African ancestry among self-identified CARDIA was estimated from a genome-wide set of ancestry informative markers, as previously described.18 Associations between soluble P-selectin and individual SNPs were assessed using multiple linear regression models, separately for EAs and AAs. For each SNP, genotypes were coded as 0, 1, or 2 copies of the minor allele, assuming an additive genetic model. Regression models were adjusted for age, sex, clinic, education, smoking status, BMI, systolic blood pressure, LDL-cholesterol, log (triglycerides), fasting glucose (and % African ancestry in AAs), all of which were cross-sectional correlates of soluble P-selectin. In EA, we performed a second regression model additionally adjusting for the effect of Thr715Pro (rs6136). The covariate-adjusted SNP-specific effect size was estimated from the regression coefficients as the amount of change in plasma P-selectin (mg/L) associated with each additional copy of the minor allele in comparison with the baseline common homozygote. In sensitivity analyses, we ran additional regression models that included separate terms for the heterozygous and uncommon homozygote genotype groups; the results confirmed that additive models are a reasonable assumption for P-selectin genotype-phenotype associations. In the 3 instances in which homozygosity for the minor allele was uncommon (<10 participants per genotype) within a particular population sample, there was no evidence that the results from additive models were skewed by the small size of the rare homozygous genotype. To account for testing of multiple correlated SNPs, we adjusted the experiment-wide significance threshold to keep the type I error rate at 5% by using permutation testing.19 After 1000 permutations, we took the fifth percentile of these minimum probability values as the new multiple-testing corrected threshold for the probability values obtained with the original data. Haplotype association analyses were performed within regions of strong linkage disequilibrium using a generalized linear-regression framework that incorporates haplotype phase uncertainty, as implemented in the haplo.stats package for R.20 Once a “global” haplotype association was identified, we performed additional analyses to further localize haplotype and individual SNP effects by constraining regression parameters across various haplotypes and testing nested models using the conditional haplotype-based likelihood ratio test method of Purcell et al.21 Multiple linear regression models were used to determine the association between individual SNPs and mean standardized IMT measured at year 20 as the dependent variable, adjusted for age, sex, and other atherosclerotic risk factors at the year 20 examination (hypertension, diabetes, smoking, BMI, and LDL- and HDL-cholesterol).
Cross-Sectional Correlates of Soluble P-Selectin at Year 15
Study participant characteristics are summarized in supplemental Table I. Higher soluble P-selectin levels were associated with male sex, smoking, BMI, blood pressure, LDL cholesterol, triglycerides, and fasting glucose (age, sex, race, and clinic-adjusted probability values all <0.001). When all variables were entered into a multivariable regression model, sex, smoking status, blood pressure, LDL cholesterol, triglycerides, and fasting glucose remained associated with soluble P-selectin in the CARDIA cohort (Table 1). When stratified by self-reported ethnicity, the association with LDL cholesterol levels appeared to be confined to EAs (probability value for ethnicity interaction=0.02). Together, the proportion of variance in soluble P-selectin levels explained by race, sex, smoking, and other atherosclerotic risk factors was 13% in the study sample overall, and dropping race as a covariate from each model, 15% in EAs and 10% in AAs.
Among self-identified AAs, each SD increase in % African ancestry was significantly associated with lower (b=−1.09; 95%CI −0.41 to −1.77; P=0.002) age-, sex-, clinic-, and education-adjusted soluble P-selectin levels. Additional covariate adjustment for smoking, blood pressure, BMI, lipids, and glucose had little effect on the association between African ancestry and soluble P-selectin in AAs (b=−1.03, 95%CI −0.36 to −1.71; P=0.002). In contrast, on adjustment for the effect of SELP genotypes, the magnitude of the African ancestry–soluble P-selectin association in AAs was greatly attenuated (b=−0.17, 95%CI −0.96 to +0.62) and no longer statistically significant (P=0.67).
SELP Polymorphism Linkage Disequilibrium
Genotype distributions of the 25 SELP polymorphisms were in Hardy-Weinberg equilibrium in both the EA and AA CARDIA cohorts. On the basis of pair-wise D′ measurements between SNPs, the SELP gene region could be divided into 2 segments showing little evidence of historical recombination: SNPs 1 through 13 (haplotype block 1) and SNPs 16 through 25 (haplotype block 2) (supplemental Figures I and II). The selection of multiple tagSNPs from some of the larger LD bins during our SNP selection procedure accounted for the high correlation between several of the typed SNPs.
Association Between SELP Polymorphisms and Soluble P-Selectin at Year 15
Overall, SELP genotype was significantly associated with soluble P-selectin levels in both EAs and AAs (global permutation probability values 0.001 and 0.005, respectively). When SELP SNP genotypes were combined together with atherosclerotic risk factors and other participant characteristics, the proportion of variance explained in P-selectin levels was 20%. The minor allele frequencies and specific pattern of allelic association differed between EA and AA (Table 2). Ten of the 25 SNPs tested were associated with soluble P-selectin levels in EAs, whereas fewer SNPs (5 of 25) were associated with soluble P-selectin levels in AAs. Overall, SELP multi-locus genotype explained 11% of the interindividual phenotypic variation in EAs and 5% of the variance in AAs.
The Thr715Pro polymorphism (rs6136) showed the largest association with soluble P-selectin levels in EAs, explaining 7% of the phenotypic variance. When adjusted for bio-demographic factors, each additional copy of the minor Pro715 allele was associated with 6.63 (95%CI 4.96 to 7.27) mg/L lower soluble P-selectin levels, which is equivalent to an approximately 15% reduction in the mean EA population P-selectin level per copy of the Pro715 allele. The Pro715 variant was considerably less common among AAs (allele frequency 2%).
In EAs, a number of SNPs located within the 5′ haplotype block 1 of SELP were significantly associated with soluble P-selectin. These associations were only slightly attenuated on additional adjustment for the effects of the Thr715Pro polymorphism (Table 2). On conditional haplotype testing (Supplemental Table II), H1.2, H1.5, and H1.6 were all homogeneously associated with decreased P-selectin levels by comparison to H1.1 (P=0.0002), whereas no difference was observed between H1.1 and H1.3 (P=0.45). Moreover, after adjusting for the SNP12 effect, block 1 haplotypes were no longer associated with P-selectin (P=0.45 and P=0.22 in EAs and AAs respectively; supplemental Table II). These results putatively localize the observed block 1 haplotype association to rs2235302. There was no significant block 1 SNP or haplotype association with P-selectin among AAs.
In addition to rs6136 or Thr715Pro (SNP23 which uniquely tags haplotype H2.5), several SNPs located within haplotype block 2 were independently associated with lower P-selectin (Table 2 and Supplemental Table III). SNP21 (rs6133 or Val599Leu), which tags H2.2 in EAs and AAs and H2.6 in AAs, was associated with lower soluble P-selectin in both populations. In AAs, adjustment for the individual effect of SNP21 (or any of the SNPs in strong LD) abolished the global haplotype block 2 - soluble P-selectin association (supplemental Table III). Haplotype H2.3 (containing the minor alleles of SNP 18, SNP20, and SNP25) was associated with higher P-selectin in EAs, even after adjusting for the effects of H2.2 and H2.5 (conditional likelihood ratio test probability value=0.01).
Conditional analysis of EA global SELP haplotypes constructed from SNP12 (rs2235302), SNP21 (rs6133), SNP23 (rs6136), and SNP25 (rs3917854), controlling for any remaining haplotype effects present in both block 1 and block 2, confirmed the independent effects of each of the 4 SELP variants: conditional probability values for SNP12, SNP21, SNP23, and SNP25 were 2×10−5, 0.004, 6×10−9, and 0.009, respectively.
SELP Genotype, Year 15 Soluble P-Selectin, and Year 20 Carotid IMT
When adjusted for age, sex, and recruitment site, year 15 soluble P-selectin was associated with a graded increase in carotid IMT measured at year 20; the association was stronger among EA than AA (Table 3). When additionally adjusted for other cardiovascular risk factors, the soluble P-selectin-IMT association was attenuated, but remained significant among EA. Thr715Pro and several 5′ block 1 haplotype SNPs were nominally associated with carotid IMT (P<0.05) among EAs, but not among AAs (supplemental Table IV). The IMT associations were consistent with the direction of the SNP- soluble P-selectin associations, and the SELP genotype -IMT associations were independent of other vascular risk factors at the year 20 examination. The 5′ block SNP4 (rs2236866) and SNP12 (rs2235302 IMT) associations reached the experiment-wide significance threshold in EAs (P<0.0036). Comparing EA subjects in the upper 20% distribution of IMT to those in the lower 20% distribution of IMT, the frequency of the rs2235302 G allele was 58% versus 48% (covariate-adjusted odds ratio per each additional G allele=1.53, 95%CI 1.04 to 2.26; P=0.03). Conditional analysis of EA haplotype block 1 and also global SELP haplotypes constructed from SNP12 (rs2235302), SNP21 (rs6133), SNP23 (rs6136), and SNP25 (rs3917854) confirmed that the 5′ block 1 variant SNP12 (rs2235302) accounted for the majority of the SELP association with IMT and suggested the association between SNP4 and IMT was most likely attributable to LD with SNP12 (data not shown).
In the large population-based CARDIA cohort, atherosclerotic risk factors were highly correlated with soluble P-selectin concentration measured at the year 15 examination. Year 15 soluble P-selectin predicted carotid IMT at year 20 among EAs, even after accounting for traditional risk factors. Multiple common SELP nucleotide sequence variants were independently associated with soluble P-selectin, but the magnitude and allelic patterns of association differed between populations. Thr715Pro had the strongest influence on soluble P-selectin in EAs, whereas Val599Leu was more strongly associated among AAs. An additional polymorphism (rs2235302) located within the 5′ region of SELP was associated with both soluble P-selectin and carotid IMT among EA.
The cross-sectional association of soluble P-selectin with various cardiovascular risk factors is consistent with previous reports.3–6⇓⇓⇓ In the combined Framingham Offspring (EA) and Omni minority cohorts,5 the proportion of interindividual variability in soluble P-selectin explained by clinical factors was 10%, similar to the 13% explained among CARDIA participants. Circulating P-selectin concentration has been reported to be lower in individuals of African origin relative to Europeans and South Asians living in England.6 In the Framingham and Omni cohorts, soluble P-selectin was higher among ethnic minority participants (comprised of African-, Hispanic- and Asian-Americans) than among EA participants.5 In CARDIA, self-reported ethnicity was not strongly correlated with soluble P-selectin. Regional environmental or genetic factors, including recent admixture, might account for some of the discrepant findings among studies. Notably, we did find that higher African ancestral proportions were associated with lower P-selectin levels among self-identified AA participants from CARDIA; the ancestry association persisted after adjustment for various atherosclerotic risk factors, some of which also tend to differ by ethnicity. It is interesting to note that adjustment for SELP genotype greatly attenuated the ancestry-soluble P-selectin association. These results suggest that the allelic heterogeneity between Africans and Europeans at the SELP locus may account at least in part for the association of soluble P-selectin with genetic ancestry.
By using SNP discovery data from SeattleSNPs we were able to assess in a relatively comprehensive manner the major patterns of polymorphism across the SELP gene in European and African populations. Our data from CARDIA are consistent with a recent report from Framingham that the SELP locus explains ≈13% of the interindividual variation in soluble P-selectin measured in plasma.5 A higher proportion of 25% variance explained was reported by Marteau et al,16 who measured serum P-selectin levels. Thr715Pro polymorphism, which predicts a conformational change near the proteolytic cleavage site of the membrane-bound form of P-selectin,5,16⇓ accounts for the majority of the influence of the SELP gene region among EAs. The lower amount of phenotypic variance explained by SELP genotype among AAs from CARDIA (5%) may reflect the lower frequency of the Pro715 allele among African descent populations,14,15⇓ or lower linkage disequilibrium coverage among AAs (50%) than EAs (80%) in the current study (attributable to the greater SELP nucleotide diversity among Africans relative to Europeans and issues related to SNP assay design).
The Leu599 allele (rs6133) is more common among AA than EA and was associated with lower soluble P-selectin in both populations. Our results confirm recent findings from the Framingham cohort5 and the STANISLAS family study16 that Val599Leu is associated with lower soluble P-selectin independently of Thr715Pro. Val599Leu is located near the transmembrane domain of P-selectin in a region that may have a functional role in P-selectin/leukocyte interaction,22 and the amino acid substitution is predicted to be functionally deleterious.23 In addition, we identified a new block 2 association with soluble P-selectin in EAs (haplotype 2.3 which is tagged by several intronic SNPs including rs3917854) that was independent of either Val599Leu or Thr715Pro.
The association of the 5′ block 1 SNP (rs2235302) with both soluble P-selectin and carotid IMT, along with the ability of soluble P-selectin measured during young adulthood to predict carotid IMT, support experimental evidence from animal models of the importance of P-selectin in the early stages of atherogenesis.24–27⇓⇓⇓ The 5′ SELP SNP (rs2235302) is in strong linkage disequilibrium with 2 other typed noncoding SNPs tagging haplotype 1.1 (including rs2236866, rs2235302), as well as with a −2123 C/G promoter polymorphism (rs1800807) reported to be associated with higher soluble P-selectin.12 The −2123 C/G promoter polymorphism is located within a putative transcription factor binding site for c-Ets-1.28 However, the precise functional sites responsible for the difference in P-selectin levels associated with SELP 5′ haplotype block remains to be determined. The reason for the stronger association observed between the 5′ SELP polymorphisms and IMT relative to the Thr715Pro association (which was more strongly associated with soluble P-selectin) is not immediately apparent. One possible explanation is that the 5′ variants have a more pronounced effect on cellular P-selectin expression, whereas Thr715Pro or Val599Leu may preferentially affect the amount of the soluble form released into plasma. Further assessment of the relationship between SELP genotype, soluble P-selectin, and cellular P-selectin expression, and the mechanism by which the soluble form is derived the cellular form, will be required to resolve this question.
In summary, clinical factors and multiple common SELP polymorphisms are associated with soluble P-selectin, albeit with differences between populations in the specific patterns of genotype-phenotype association. Because the genetic and clinical factors studied accounted for only ≈20% the interindividual differences in circulating P-selectin, a substantial amount of the phenotypic variance still remains to be explained. Studies that assess the role of less common SELP gene polymorphisms, additional variants outside the SELP gene, and including diverse ethnic groups, along with functional studies, may help to characterize additional factors that influence soluble P-selectin. Further elucidation of the role of P-selectin genotype and phenotype in occurrence of subclinical atherosclerosis in humans ultimately may have important implications for identifying younger individuals at high risk for future cardiovascular events,28,29⇓ as well as developing antiatherosclerotic therapies targeted at cellular adhesion molecules.26
Sources of Funding
This work was supported by the Young Adult Longitudinal Trends in Antioxidants (YALTA) Study, an ancillary study to CARDIA 1RO1-HL53560–01A1 from the National Heart, Lung, and Blood Institute (to D.R.J. and M.D.G), Inflammatory Genomics and Atherosclerosis Prevention ancillary CARDIA grant HL71017 (to D.S.S. and A.P.R.), by CARDIA contracts N01-HC-95095, N01-HC-48047, N01-HC-48048, N01-HC-48049, N01-HC-48050, N01-HC-45134, and N01-HC-05187 from NHLBI, and by NHLBI Program for Genomic Applications grants HL66682 and HL66642 (to D.A.N. and M.J.R.).
Original received April 1, 2008; final version accepted May 22, 2008.
- ↵Blann AD, Nadar SK, Lip GY. The adhesion molecule P-selectin and cardiovascular disease. Eur Heart J. 2003; 24: 2166–2179.
- ↵Ridker PM, Buring JE, Rifai N. Soluble P-selectin and the risk of future cardiovascular events. Circulation. 2001; 103: 491–495.
- ↵Chironi G, Dosquet C, Del-Pino M, Denarie N, Megnien JL, Drouet L, Bal Dit Sollier C, Levenson J, Simon A. Relationship of circulating biomarkers of inflammation and hemostasis with preclinical atherosclerotic burden in nonsmoking hypercholesterolemic men. Am J Hypertens. 2006; 19: 1025–1031.
- ↵Koyama H, Maeno T, Fukumoto S, Shoji T, Yamane T, Yokoyama H, Emoto M, Shoji T, Tahara H, Inaba M, Hino M, Shioi A, Miki T, Nishizawa Y. Platelet P-selectin expression is associated with atherosclerotic wall thickness in carotid artery in humans. Circulation. 2003; 108: 524–529.
- ↵Barbaux SC, Blankenberg S, Rupprecht HJ, Francomme C, Bickel C, Hafner G, Nicaud V, Meyer J, Cambien F, Tiret L. Association between P-selectin gene polymorphisms and soluble P-selectin levels and their relation to coronary artery disease. Arterioscler Thromb Vasc Biol. 2001; 21: 1668–1673.
- ↵Volcik KA, Ballantyne CM, Coresh J, Folsom AR, Wu KK, Boerwinkle E. P-selectin Thr715Pro polymorphism predicts P-selectin levels but not risk of incident coronary heart disease or ischemic stroke in a cohort of 14595 participants: the Atherosclerosis Risk in Communities Study. Atherosclerosis. 2006; 186: 74–79.
- ↵Purcell S, Daly MJ, Sham PC. WHAP: haplotype-based association analysis. Bioinformatics. 2007; 23: 255–256.
- ↵Ruchaud-Sparagano MH, Malaud E, Gayet O, Chignier E, Buckland R, McGregor JL. Mapping the epitope of a functional P-selectin monoclonal antibody (LYP20) to a short complement-like repeat (SCR 4) domain: use of human-mouse chimaera and homologue-replacement mutagenesis. Biochem J. 1998; 332: 309–314.
- ↵Ng P, and Henikoff S. Accounting for Human Polymorphisms Predicted to Affect Protein Function. Genome Research. 2002; 12: 436–446.
- ↵Manka D, Collins RG, Ley K, Beaudet AL, Sarembock IJ. Absence of p-selectin, but not intercellular adhesion molecule-1, attenuates neointimal growth after arterial injury in apolipoprotein e-deficient mice. Circulation. 2001; 103: 1000–1005.
- ↵Phillips JW, Barringhaus KG, Sanders JM, Hesselbacher SE, Czarnik AC, Manka D, Vestweber D, Ley K, Sarembock IJ. Single injection of P-selectin or P-selectin glycoprotein ligand-1 monoclonal antibody blocks neointima formation after arterial injury in apolipoprotein E-deficient mice. Circulation. 2003; 107: 2244–2249.
- ↵Smyth SS, Reis ED, Zhang W, Fallon JT, Gordon RE, Coller BS. Beta(3)-integrin-deficient mice but not P-selectin-deficient mice develop intimal hyperplasia after vascular injury: correlation with leukocyte recruitment to adherent platelets 1 hour after injury. Circulation. 2001; 103: 2501–2507.
- ↵Herrmann S, Ricard S, Nicaud V, Mallet C, Evans A, Ruidavets J, Arveiler D, Luc G, Cambien F. The P-selectin gene is highly polymorphic: reduced frequency of the Pro715 allele carriers in patients with myocardial infarction. Hum Mol Genet. 1998; 7: 1277–1284.
- ↵Tregouet DA, Barbaux S, Escolano S, Tahri N, Golmard JL, Tiret L, Cambien F. Specific haplotypes of the P-selectin gene are associated with myocardial infarction. Hum Mol Genet. 2002; 11: 2015–2023.