Cis eQTLs were over-represented among expression traits that were also affected by gender, age, BMI, smoking and CRP, with odds ratios as high as 3.24 for cigarette smoking ( Table 5 ). This suggests that some genes are more responsive than others to the influence of multiple factors. For expression traits that were simultaneously associated with cis eSNPs and risk factors (n = 465), we determined the joint effects of the two sources of variability on expression level. For this purpose, each eQTL was modelled as a function of the best cis-acting eSNP, the associated risk factor and the interaction between the two. When several risk factors were associated with the same eQTL, each of them was tested separately for interaction with the corresponding SNP. The best FDR-corrected P-values for interaction (File S7) were 0.014 (ISCU expression, gender and rs4830487) and the second one was 0.042 (HIST1H2AE expression, smoking and rs16891378). This first genome-wide exploration of interaction between cis eSNPs and risk factors on gene expression therefore suggests that the two sources of variability mostly act additively on expression. This
