The characteristics of study subjects are summarized in . A total of 33,671 subjects were available for the meta-analysis of sICAM-1, 4,921 for sP-selectin, and 2,860 for sE-selectin.
Summary of participating studies for the meta-analysis
The meta-analysis showed that sICAM-1 levels were 4.6% (95% CI 3.4–5.8%) lower in heterozygotes and 7.2% (4.7–9.7%) lower in minor allele homozygotes, than in major allele homozygotes (p=7.3×10−14
, ). Similarly, an allele dose dependent association was observed for sP-selectin, with heterozygotes and minor allele homozygotes having 11.5% (7.2–15.8%) and 18.6% (9.1–28.1%), respectively, lower levels than in major allele homozygotes (p=1.7×10−7
, ). An allele dose dependent association also was seen for sE-selectin whose level was 25.6% (19.0–32.2%) lower in heterozygotes and 43.3% (36.9–49.3%) lower in minor allele homozygotes, than in major allele homozygotes (p=2.1×10−14
, ). Standardized effect size was larger for sE-selectin than for sICAM-1 and sP-selectin (Supplemental Figures S1 to S3
). We noted heterogeneity (Supplemental Table 1
) which a meta-regression analysis indicated was not attributed to differences among individual studies in age, sex, type of subjects (population-based or diabetics), number of subjects (n>1000 or <1000), type of blood sample used (plasma or serum) or which SNP studied, although the meta-regression analysis had low power due to the relatively small numbers of individual studies. There was no evidence of publication bias. We observed correlations between sICAM-1, sP-selectin and sE-selectin levels (Supplemental Table 2
Figure 1 Weighted mean difference by genotype in soluble intercellular adhesion molecule-1 (sICAM-1), soluble P-selectin (sP-selectin), and soluble E-selectin (sE-selectin) levels. Data shown are weighted mean difference ± 95% confidence interval in circulating (more ...)
SNP rs507666 is located within the ABO
gene, and SNP rs579459 and rs651007 are in its proximity. The ABO
gene encodes a glycosyltransferase that transfers sugar residues to the H antigen and determines the ABO blood group.22
Group A has two subtypes, i.e. A1 and A2, respectively. It has been shown that the A1 subtype has over 30-fold higher transferase activity than the A2 subtype.23
The A1 allele is perfectly tagged by the minor allele of SNP rs507666.11
SNP rs507666 is in near perfect LD (r2
=0.96) with rs579459 and rs651007. Thus, the associations of these SNPs with sICAM-1, sP-selectin and sE-selectin levels may represent an effect of the ABO group A1 subtype. It has been suggested that the increased glycosyltransferase activity in individuals carrying the A1 allele might have an effect on the shedding, clearance or secretion of adhesion molecules, thereby influencing their levels in the circulation.11;12
Adhesion molecules are crucial to platelet leukocyte interaction and leukocyte migration into the vessel wall and thus important players in the atherosclerosis process underlying CHD.2;24
In a number of previous studies increased CHD risk has been associated with high sICAM-1, sP-selectin and sE-selectin levels.3;5;6
Unexpectedly, variants at the ABO
locus conferring elevated CHD risk15;16;25
, like the minor allele of SNP rs57945916
, were associated with decreased levels of soluble adhesion molecules in our meta-analysis. One possible explanation for this seeming paradox may be that soluble adhesion molecules, although elevated in the case of endothelial dysfunction, actually compete with leukocyte adhesion to the endothelium (competition to cell surface adhesion molecules). Another possibility may be that the lower levels of soluble adhesion molecules might arise because of lower shedding of ectodomains, potentially leaving higher levels of intact cell surface adhesion molecules to recruit leukocytes to the blood vessel wall. To date, it is not known whether elevated levels of soluble adhesion molecules in vascular high-risk patients represent an epiphenomenon of vessel wall pathology, a true risk factor or a counter-regulatory per se protective mechanism as indicated by preliminary experimental data16
. Experimental studies are required to further elaborate the pathophysiological role of soluble adhesion molecules and to clarify whether the prominent alterations in sICAM-1, sP-selectin and sE-selectin observed in this study are relevant to the recently discovered association between ABO
SNPs and CHD risk.
Some limitations to our study warrant mentioning. First, the mechanism underlying the association of SNPs at the ABO locus with sICAM-1, sP-selectin and sE-selectin levels has remained unclear. Second, since SNP rs579459 is in strong LD with a number of other SNPs at this locus, it remains unknown which SNP is the causal variant. Third, since this study was conducted in individuals of European ancestry, the findings may not be generalizable to other races/ethnicity.
In conclusion, our study provides compelling evidence of an allele dose dependent association of variation at the ABO locus with circulating sICAM-1, sP-selectin and sE-selectin levels. These results contribute to the knowledge of genetic influences on these adhesion molecules which play important roles in many inflammatory diseases.