In the present two-stage GWAS, we confirmed that previously identified locus ABO was associated with serum ALP levels in a Chinese population. When comparing the present study with previously published findings, ethnic differences were observed across populations. In addition, a significant interaction between overweight and obesity and ABO rs651007 on serum ALP was found. The effects of rs651007 on serum ALP levels were attenuated in overweight or obese individuals than in normal weight subjects.
The association between ABO locus and serum ALP has been reported in other populations [15–17]. The present study was consistent with previous findings that serum ALP levels are much higher in individuals with blood type B or O ; the lead SNP rs651007 on ABO locus, which is in high LD with previously reported rs495828 (r2 = 0.916), accounted for 2.15% of the total variance of serum ALP in our population, similar with that reported in earlier studies (2.0% for rs657152 in Europeans and 3.79% for rs495828 in Japanese) [16, 17]. The coding product of human ABO gene is a glycosyltransferase, of which catalytic activity could facilitate the transfer of carbohytrates to H antigen to form the antigenic structure of the ABO blood groups . Current evidence found that human ABO blood groups or genetic variations on ABO gene were associated with the risk of various diseases, such as pancreatic cancer, gastric cancer, falciparum malaria, venous thromboembolism, myocardial infarction and type 2 diabetes [22–28]. Genetic variations on ABO gene were also found to be related with levels of serum E-selectin, sICAM, RBC, Hb, vWF, and CEA [16, 25, 29]. In healthy fasted individuals, approximately 90% of serum ALP originate from liver/bone/kidney, nearly 10% from intestine and in some cases 1% from placenta . Most of the intestinal ALP is attached to ABO antigens on the surface of erythrocytes by a glycosyl-phosphatidylinositol anchor. Erythrocytes of blood type A bind to almost all intestinal ALP, while erythrocytes of blood type B/O bind to a much lesser degree, therefore results in a more prevalent presence of intestinal ALP in serum in individuals with blood type B/O [19, 30]. In this sense, the differences in the binding capacity between intestinal ALP and erythrocytes among individuals with different blood groups might be a potential mechanism underlying the association between ALP and ABO locus identified in GWAS. The observation that association between rs651007 and serum ALP dramatically reduced after ABO blood group adjustment also lend support to this deduction, indicating that ABO blood groups may act as the driving force behind this association.
Differences were observed among studies when comparing our findings with previously published results and parts of the reason might be due to the ethnic heterogeneity, such as varied effect allele frequencies and unique LD structures. For example, the MAF of rs16856332 are 0.39, 0.06 and 0.07 in Europeans, Chinese and Japanese, respectively; the MAF of rs9467160 on GPLD1 locus is 0.21 in European but is only 0.03 in Chinese Han, and even lower, 0.003 in Japanese; the scenario is the same for rs7267979 on ABHD12 that its effect allele frequency is 0.57 in European but is only 0.07 in Chinese. On the contrary, rs281377 has a MAF of 0.43 in European but the frequency of this effective allele is 0.86 in Chinese. Meanwhile, not all ABO SNPs identified in Europeans are significant in Chinese, which may partially due to the differences in LD structure, e.g. the LD relation between rs8176720 and rs514708 on ABO gene is weak in Asian populations, but get stronger in Europeans (r2 = 0.38, 0.36 and 0.67 in the present, Japanese and European populations). Unique population structures, including different modifier genes, gene-gene or gene-environment interactions, different lifestyles or environment exposures can also be explanations for the heterogeneity across populations . Notably, although it showed no evidence for heterogeneity across studies for SNPs on ABCB11, PPP1R3B, C9orf125, JMJD1C, REEP3, ST3GAL4, and ABHD12, SNPs on ALPL, GPLD1, PMFBP1 and FUT2 represented considerable heterogeneity among studies. The lack of replication of many loci may also due to different sample selection criteria, different study designs and statistical analysis methods, as well as the relatively small sample size of the GWAS stage of the present study. Studies with larger sample size, different ethnic sources and multi-center cooperations are needed to explain the ethnic differences better.
Previous study found that serum ALP was higher in obese patients . The interaction between rs651007 and overweight and obesity on ALP levels found in the present study may be one of the explanations. The effect of rs651007 on serum ALP was attenuated in overweight or obese individuals compared with normal weight subjects, resulting in higher ALP levels in overweight or obese individuals. eQTL analysis  showed that rs651007 could trans-regulate the expression of TNFRSF1A (Effect = 0.403, P = 1.30 × 10-5, LOD = 4.126); meanwhile, rs651007 was observed associated to higher serum TNF-R2 levels in a previous study , demonstrating that rs651007 may be involved in the regulation of TNF receptors. In addition, TNF signaling is believed playing an important role in obesity since TNFα mRNA or protein was found overexpressed in obese subjects in both experimental and epidemiological studies. Researchers also found that TNF signaling plays an important role in the insulin resistance of obesity [34, 35]. Moreover, TNF-α has been demostrated to positively regulate ALP levels in various types of cells [36–39]. Taken together, all evidence suggests that the interaction between rs651007 and overweight and obesity on serum ALP levels may act through the regulation of TNF system. Further studies are needed to uncover the real mechanism of the interaction.
Our study are not only consistent with findings of previous GWA studies [15–17], but also in line with early reports that serum ALP levels varied in individuals with different ABO blood types [2, 4, 18], However, it is undeniable that the relatively small sample size of our GWAS stage limited us to detect SNPs with minor effects and SNPs with low MAF, which calls for further large sample size studies and consortium cooperation.