To the best of our knowledge, this study is the first to examine the association between estrogen metabolism and signaling pathway genes, SULT1E1, COMT, and ESR1, and ischemic stroke of the young. The estrogen metabolism genetic polymorphism, COMT Val158Met, was significantly associated with risk of young ischemic stroke among females. Furthermore, we used multianalytic strategies to systemically examine the interaction among these genes. Using different analytic strategy, however, MDR and CART method showed consistent result that there was a strong gene-gene interaction between SULT1E1 -64G/A and COMT Val158Met on the risk of young ischemic stroke. Traditional multiple logistic regression results also showed that there was a significant interaction effect between SULT1E1 -64G/A and COMT Val158Met for development of young ischemic stroke.
Although SULT1E1, a gene encoding an estrogen-metabolizing enzyme, may contribute to individual differences in the biotransformation of this steroid hormone, the relationship between SULT1E1 -64G/A with ischemic stroke was not observed in our study. Owing to low allele frequencies of the three nonsynonymous SNPs among 23 polymorphisms of SULT1E1 identified by Adjei et al. 
, we selected -64G/A located in the promoter region which might influence estrogen sulfotransferase enzyme as the candidate SNPs in our study. We also found that subjects with SULT1E1 -64 G/A AA genotype had significantly lower serum estradiol level than G carriers among healthy controls in our study, especially for females (Table S1
). However, the controversial results concerning the association between SULT1E1 -64G/A polymorphism and cancers might be due to the uncertain function of this variant, which might be the reason for non-significant results found in this study 
COMT is an important enzyme in the degradation of both catecholamine and estrogens. A non-synonymous G to A base change, COMT Val158Met polymorphism, resulted in the reduction of COMT activity which may impair vascular health in several ways 
. Several clinical diseases such as preeclampsia 
, hypertension 
and heart disease 
have been reported to be associated with this SNP. In addition, growing evidence supports that 2-methoxyestradiol (2-ME), a natural estrogen metabolite produced by COMT, has a potent antiproliferative and antiangiogenic capacity 
and has direct involvement in redox-regulated signaling as a pro-oxidant 
, thus it could be a possible disease mechanism in the protection against atherosclerosis development. Therefore, these abundant studies support our findings that subjects with COMT Met allele had a significant higher risk of young ischemic stroke among females after 1000 permutation tests.
Estrogen influence multiple organ systems including cardiovascular, reproductive and skeletal systems by binding to specific estrogen receptors located within the nuclei of target cells 
. Numerous epidemiological and experimental studies indicate the protective roles of estrogens in many forms of cardiovascular and cerebrovascular diseases 
. However, most studies have focused on the association between ESR1 variants and cardiovascular disease with conflicting results 
, and the reason might owe to various study designs. Although our findings including genotype and haplotype analysis (Table S2
and Table S3
) reveal no statistically significant risk for ischemic stroke, a gene-environment interaction effect between ESR1 C-A haplotype and serum estradiol level on young ischemic stroke patients was observed (P for interaction
0.0348, Table S4). The possible mechanism might be that the transcription factor, ERα, interacts directly with specific promoter sequences comprising 15-bp inverted palindromes known as estrogen response elements (EREs) located in the regulatory region of target genes via binding of 17α-estradiol to their classical receptor ERα 53
In the present study, the power and the possibility of false positives must be considered. According to a relevant range of minor allele frequencies (22–38%), a post hoc
power calculation can reach to near 80% power to detect an effect size (OR) difference of 1.6 using Power and Sample Size Program (version 3.0.43) 54
. In addition, multiple testing is a major concern of this study. Genotype and allelic analysis for each of the 4 SNPs yielded no significant association with ischemic stroke of the young after the Bonferroni correction. The excessively conservative correction of the Bonferroni method might result in the decreased power; therefore, based on 10,000 random permutations, the association between risk of young ischemic stroke and COMT Val158Met among females remained significant.
A major strength of our study is that gene-gene interactions were consistently identified by both MDR and CART analysis. The results were also confirmed by logistic regression approach when controlling for confounding variables simultaneously. To improve the statistical power, the MDR method’s conversion from multiple to single variable resulted in efficient identification of potential gene-gene interactions in relatively small samples 55
. In addition, the MDR also reduces the chances of making type I errors as a result of multiple testing through cross validation and permutation testing procedure. The CART analysis is a nonparametric strategy, a decision tree-based data mining to identify specific combinations of genetic factors relating to disease, which requires no assumption of a genetic model. Recent researches have suggested that utilizing multiple complementary analytic approaches can increase statistical power to identify possible gene-gene interactions effectively 56
There were still some limitations in this study. First, the sample size is relatively small due to difficulty in enrollment of young ischemic stroke patients. Thus, further studies in larger populations are required to validate the findings. Second, we used a candidate approach to select SNPs focusing on the functional variants due to limited budget. However, with more advanced genome-wide association studies exploiting the genetic association study, we may have missed some signals that were not genotyped in the current study. Nevertheless, we cannot rule out the causal markers in the genes we studied. Finally, the menstrual status was acquired for some subjects when the estradiol level was measured. Therefore the misclassification may have occurred when we included females who were in the ovulation stage in the high estradiol level group as the reference. However, this misclassification is non-differential that might dilute the odds ratio and lead to the result toward the null.
In conclusion, these data indicate that COMT Val158Met polymorphism is significantly associated with ischemic stroke risk among females and suggest that gene-gene interaction effect of SULT1E1 -64G/A and COMT Val158Met polymorphisms play more important roles than the individual factors for the development of young ischemic stroke. Moreover, lower estradiol level could increase risk of young ischemic stroke for those who carried either COMT or SULT1E1 risk genotypes.