Results

Association of LPL SNPs with lipid concentrations
We sequentially discarded 6 SNPs with MAF < 0.05, 2 SNPs with call rate < 95%, and 1 SNP with HWE p-value < 0.002, leaving 19 SNPs for the association study at stage 1. Our results show that SNPs located in the LPL gene, including the 3' flanking region, were strongly associated with HDLC and TG levels. The linear regression, adjusted for age, gender, and BMI, using combined data from the two study stages, revealed that 15 SNPs were associated with HDLC and TG levels (Table 2). The strongest associations were between rs10503669 and both HDLC (combined p = 3.6 × 10-22; 2.2-mg/dL increase per minor T allele) and TG (combined p = 3 × 10-15; 13.6-mg/dL decrease per minor T allele) in additive models. However, no strong association was detected between the SNP and either LDLC or TCHL levels (Supplementary Fig. 1).
For LPL SNPs, we tested 2 intronic SNPs and 2 SNPs located in the 3' untranslated region (UTR) from stage 1. Of these, two SNPs in the 3' UTR (rs11570892 and rs3200218) were not significantly associated with any lipid concentration (Supplementary Table 1). Two intronic SNPs, rs263 and rs271, were significantly associated to higher HDLC levels (p = 1.2 × 10-10 and p = 2.5 × 10-11; 1.2- and 1.3-mg/dL increase per minor A and T allele, respectively) and lower TG levels (p = 1.7 × 10-6 and p = 7.2 × 10-7; 8.8- and 9-mg/dL decrease per minor A and T allele, respectively) (Table 2). In stage 2, 7 intronic SNPs and 2 SNPs that were located in the 3' UTR were tested. It was found that rs263 and rs271 were also strongly associated with increased HDLC (p = 5.5 × 10-6 and p = 5.1 × 10-6, respectively; both showed a 1.5-mg/dL increase per minor A and T allele) (Table 2) and decreased TG levels (p = 3.3 × 10-3 and p = 2.2 × 10-3; 8.6- and 8.7-mg/dL decrease per minor A and T allele, respectively). Additionally, 4 intronic SNPs (rs253, rs326, rs327, and rs12679834) were significantly associated with increased HDLC (p = 4.2 × 10-3, p = 1.5 × 10-9, p = 1.6 × 10-9 and p = 3.1 × 10-14; increase of 0.8, 2.1, 2.1, and 3.3 mg/dL per minor allele, respectively) and decreased TG levels (p = 3.6 × 10-3, p = 9.1 × 10-7, p = 5.9 × 10-7 and p = 1.3 × 10-6; increase of 7.4, 12.2, 12.3, and 13.7 mg/dL per minor allele, respectively) (Supplementary Table 2). The analysis using combined data from the two stages showed that rs263 and rs271 were also significantly associated with HDLC (p = 1.7 × 10-11 and p = 5.7 × 10-12, respectively; increase in both cases of 1.2 mg/dL per minor A and T allele) and TG levels (p = 6.8 × 10-7 and p = 2.7 × 10-7; decrease of 8.1, 8.3 mg/dL per minor A and T allele, respectively) (Table 2).

Association of LPL HTs with lipid concentration
Two blocks, including SNPs located in the LPL gene, were identified (Supplementary Fig. 2), and the association of HT with lipid profiles for the two blocks was examined. The analysis using combined data from the two stages showed that block 2, which includes SNPs located in the LPL gene, including the 3' flanking region, was the region most strongly associated with lipid concentrations (Table 3). HT3 in block 2 was strongly associated with increased HDLC levels (p = 2.86 × 10-22; 2.25-mg/dL increase per HT3) and reduced TG levels (p = 9.09 × 10-15; 13.7-mg/dL decrease per HT3), whereas HT1 was strongly associated with reduced HDLC levels (p = 1.1 × 10-14; 1.2-mg/dL decrease per HT1) and higher TG levels (p = 3.05 × 10-12; 8.6-mg/dL increase per HT1). There was no significant association between HT and either LDLC or TCHL.
The analysis using imputed data also showed a strong association between HTs and lipid levels in the 2 blocks, including LPL SNPs (Supplementary Table 3). HT1 in block 2 was significantly associated with a reduced HDLC level (p = 4.2 × 10-15; 1.3-mg/dL decrease per HT1) and an increased level of TG (p = 1.9 × 10-8; 8-mg/dL increase per HT1), whereas HT3 was significantly associated with increased HDLC (p = 7.8 × 10-16; 2-mg/dL increase per HT3) and LDLC levels (p = 5.1 × 10-3; 2.3-mg/dL increases per HT3) and decreased TG levels (p = 5.6 × 10-13; 15.1-mg/dL decrease per HT3).

Effect of the interaction between polymorphisms located in the LPL gene and lifestyle on lipid concentration
We tested LPL SNPs and HTs from stage 1. All significant associations were found in the results of the dominant model for the minor allele. The association of minor alleles for 3 SNPs (rs263, rs271, rs328) with higher HDLC levels also depended on energy intake (Supplementary Fig. 3). Of these, the minor alleles of 2 SNPs (rs263, rs271) were associated with lower TG levels, also depending on energy intake. The association of the minor alleles of two SNPs (rs263, rs271) with higher HDLC levels also depended on fat intake (Fig. 1). The minor alleles of 2 SNPs (rs263, rs271) were also associated with higher HDLC levels and lower TG levels through an interaction with cigarette smoking and with lower TG levels through an interaction with alcohol consumption (Fig. 2, Supplementary Fig. 4). Interaction effects of HT1 and 2 in block 1, which were formed by two SNP (rs271, rs263) in strong LD (r2 = 0.99), with lifestyle factors on HDLC and TG levels were found to be almost same with those of rs271 and rs263 (data not shown). In block 2, HT3 was associated with higher HDLC levels and lower TG levels and depended on an interaction with energy intake (Fig. 3, Supplementary Fig. 3). HT4 in block 2 was associated with lower TG levels through an interaction with cigarette smoking (Supplementary Fig. 4). For the interaction between LPL polymorphisms and physical activity, there was no significant association with any lipid levels.