Introduction

The high-density lipoprotein cholesterol (HDL-C) level is inversely correlated with the development of atherosclerosis and is inversely related to the incidence of coronary artery disease (Castelli et al. 1986) and ischemic stroke in the elderly (Sacco et al. 2001). The HDL-C level has been shown to be affected by both genetic and environmental factors, including obesity, smoking, and alcohol consumption. Among genetic factors, the apolipoprotein A1 (ApoA1) (Groenendijk et al. 2001a,b) and ApoE genotypes (Lefevre et al. 1997; Katsuya et al. 2002) are well known to influence the HDL-C level.

Genetic mutations in the ATP-binding cassette transporter A1 (ABCA1) gene have been shown to cause Tangier disease (TD) (Bodzioch et al. 1999; Brooks-Wilson et al. 1999; Rust et al. 1999) and familial HDL deficiency (Marcil et al. 1999). ABCA1 regulates cellular cholesterol efflux and facilitates lipid binding to ApoA1 (Wang and Tall 2003). Patients with TD show characteristic HDL deficiency, defective apolipoprotein-mediated phospholipid and cholesterol efflux from cells, and the accumulation of macrophage foam cells in various tissues, including arteries (Clifton-Bligh et al. 1972). Recent epidemiological studies have reported that ABCA1 polymorphisms were associated with the HDL-C level (Wang et al. 2000; Clee et al. 2001; Lutucuta et al. 2001; Harada et al. 2003). However, few of these findings have been replicated, and there are inconsistencies among previous association studies. Accordingly, the associations between ABCA1 variants and HDL-C are still controversial (Singaraja et al. 2003). One possible reason for these differences may be that the sample sizes in these studies were relatively small and lacked statistical power. Thus, to evaluate the effect of polymorphisms in ABCA1 on the HDL-C level, we conducted an association study using a large cohort (the Suita population, n=1,880), representing the general population in Japan.

Materials and methods

Subjects

The Suita population

The selection criteria and design of the Suita Study have been described previously (Mannami et al. 1997; Shioji et al. 2004). The genotypes were determined in 1,880 consecutive subjects who visited the National Cardiovascular Center between April 2002 and February 2003 (867 male subjects, 1,013 female subjects).

The hypertension group

The hypertension (HTN) group consisted of 743 Japanese subjects (422 men and 321 women), aged 18–91 years [65.2±0.4 (mean±SEM)], who were enrolled in the Division of Hypertension and Nephrology at the National Cardiovascular Center between May 2001 and April 2003.

The myocardial infarction group

The selection criteria and design of the myocardial infarction (MI) group have been described previously (Takagi et al. 2002). This group consisted of 706 patients with MI (598 men and 108 women, aged 61.3±0.4 years) who were enrolled in the Division of Cardiology at National Cardiovascular Center between May 2001 and April 2003. In the present study, we investigated only males (n=598).

Written informed consent was obtained from every subject after a full explanation of the study, which was approved by the Ethics Committee of the National Cardiovascular Center and by the Committee on Genetic Analysis and Genetic Therapy of the National Cardiovascular Center.

DNA studies

The regions of the promoter and exon 1 in ABCA1 were sequenced for polymorphisms in 24 subjects (Fig. 1). The primer sequences are available on request. For exonic regions (Fig. 2), we selected nine SNPs for genotyping from the public database (JSNP, http://snp.ims.u-tokyo.ac.jp) (Iida et al. 2001; Hirakawa et al. 2002). Well-known common variants, ABCA1 R219K and I823M, were also selected (Wang et al. 2000; Clee et al. 2001; Harada et al. 2003).

Fig. 1
figure 1

Nucleotide sequence of the 5′-flanking region and exon 1 of ABCA1. The nucleotide sequence in italics indicates exon 1

Full size image
Fig. 2
figure 2

Schema of ABCA1 and the position of the determined polymorphisms.Grayboxes indicate exonic regions

Full size image

The preliminary study revealed that JST-IMS005607 had the greatest effect on the HDL-C level among seven SNPs on the ApoA1 region, including the promoter region (up to −3Kb). Thus, we selected JST-IMS005607 for adjusting HDL-C. The genotyping of ApoE was performed according to a previous report (Katsuya et al. 2002). ApoE polymorphisms were categorized into three genotypes: E2 (ε2/ε2+ε2/ε3+ε2/ε4 subjects), E3 (ε3/ε3 subjects), E4 (ε3/ε4+ε4/ε4 subjects) (Lefevre et al. 1997). All polymorphisms were determined by the TaqMan System.

Statistical analysis

Values are expressed as mean±standard error of the mean (SEM). For triglyceride values, a logarithmic transformation was applied for the statistical test, but untransformed values are shown in the Tables 1 and 2. All statistical analyses were performed with the JMP statistical package (SAS Institute). Values of P<0.05 were considered to indicate statistical significance. Multiple linear regression and multiple logistic analyses were performed with other covariates. The residual HDL-C level was calculated by adjusting for sex, age, and body-mass index (BMI), smoking (cigarettes/day) and consumption of alcohol (ethanol, ml/week). For analyses of the effects of the ABCA1 genotype (in the Suita population), the residual HDL-C level was calculated by adjusting not only for the above five factors, but also for the ApoA1 (JST-IMS005603), and the ApoE (E2, E3, and E4) genotypes. Differences in numerical data among the groups were evaluated by one-way analysis of variance (ANOVA). Hardy-Weinberg equilibrium was calculated by a chi-square test (Table 3). To measure linkage disequilibrium (LD) between SNPs, D′ and r 2 values were analyzed using the SNPAlyze statistical package (Dynacom).

Fig. 3A, B
figure 3

Residual HDL cholesterol levels among the ABCA1 G(−273)C, C(−297)T, and G(−790)A genotypes. A Residual HDL cholesterol levels adjusted for sex, age, body-mass index, smoking, and alcohol consumption. B Residual HDL cholesterol levels adjusted for sex, age, body-mass index, smoking, alcohol consumption, ApoE genotype, and ApoA1 genotype (JST-IMS005603)

Full size image

Results

Polymorphisms of the 5′-flanking region and exon 1 of the ABCA1 gene

We found 14 polymorphisms in the promoter region, 1 polymorphism in exon 1 (5′-untranslated region), and 2 polymorphisms in intron 1 (Fig. 1).

LD was evaluated by calculating r 2 values (Table 1). We regarded r 2>0.5 as tight linkage. The minor allele frequency of the T(−1423)C and G52A polymorphisms was low (4% each), and these SNPs were neglected in further analyses. The frequencies of T(10), T(9), and T(8) were 4, 92, and 4%, respectively, in the (−980)T(10)/T(9)/T(8) polymorphism, and this polymorphism was also neglected because this is not suitable for TaqMan genotyping. Accordingly, we selected three polymorphisms, G(−790)A, C(−297)T, and G(−273)C, for the following association study.

Table 1 Linkage disequilibrium between SNPs in the 5′-flanking region and exon 1 of the ABCA1 gene. I/D#1 GTTTTGTTTT(−752)(−), I/D#2T GGGG(−226)(−)
Full size table

Association study of ApoA1 and ApoE

To observe the effect of ABCA1 polymorphisms on the HDL-C level more clearly, the HDL-C level should be adjusted by various well-known influential factors.

The ApoA1 IMS-JST005603 polymorphism was associated with the levels of HDL-C and triglyceride [HDL-C: TT 1.54±0.001 mmol/l, TC 1.59±0.02, CC 1.68±0.04, P=0.0002 (residual); triglyceride: TT 1.26±0.03 mmol/l, TC 1.15±0.04, CC 0.95±0.09, P<0.0001 (residual)]. IMS-JST005603 corresponds to the HaeIII (C317T) polymorphism described in a previous paper (Groenendijk et al. 2001b).

The ApoE polymorphism was also strongly associated with the levels of total cholesterol and HDL-C [total cholesterol: E2 5.13±0.06 mmol/l, E3 5.37±0.02, E4 5.41±0.05, P=0.0002 (residual); HDL-C: E2 1s.67±0.03 mmol/l, E3 1.56±0.01, E4 1.52±0.02, P<0.0001 (residual)].

Accordingly, we evaluated the effect of the ABCA1 polymorphisms on the HDL-C level adjusted for the ApoA1 IMS-JST005603 and ApoE polymorphisms in addition to standard factors, including sex, age, BMI, smoking, and consumption of alcohol.

Association study of ABCA1 (Suita population)

The association between the G(−273)C polymorphism and the lipid level in the Suita population is presented in Table 2. The genotype frequency of the G(−273)C polymorphism in the Suita population was not deviated from the Hardy-Weinberg equilibrium. The HDL-C level adjusted for age, sex, BMI, smoking, and consumption of alcohol was significantly associated with the G(−273)C polymorphism (P=0.0148). The G(−273)C polymorphism was even more tightly associated with the HDL-C level when adjusted for the ApoE and ApoA1 (IMS-JST005603) genotypes in addition to the standard factors (P=0.0074). The C(−297)T polymorphism was also associated with the HDL-C level (P=0.0455 adjusted for age, sex, BMI, smoking, and consumption of alcohol; P=0.0195 when also adjusted for the ApoE and ApoA1 genotypes). The effect of the C(−297)T polymorphism on the HDL-C level may be, at least in part, explained by its linkage with the G(−273)C polymorphism (r 2=0.46667, D′ value=1, P<0.0001). G(−790)A was not associated with the lipid levels. Among the polymorphisms selected from JSNPs, including R219K and I823M, only the IMS-JST071749 polymorphism was associated with the HDL-C level (P=0.0060 adjusted for age, sex, BMI, smoking, and consumption of alcohol; P=0.0093 when also adjusted for the ApoE and ApoA1 (IMS-JST005603) genotypes). The R219K and I823M polymorphisms were not associated with the HDL-C level [P=0.3877 (R219K) and P=0.2286 (I823M) adjusted for age, sex, BMI, smoking and consumption of alcohol; P=0.1926 (R219K) and P=0.1209 (I823M) when also adjusted for the ApoE and ApoA1 genotypes].

Table 2 Lipid levels in the ABCA1 G(−273)C genotypes (Suita population). Subjects who were receiving anti-hyperlipidemic medication were excluded. Values are mean±SEM. P-values calculated by ANOVA
Full size table

Association study of ABCA1 (HTN group)

To reconfirm the association between the G(−273)C, C(−297)T, and IMS-JST071749 polymorphisms and the HDL-C level, we determined the genotypes in the HTN group. As shown in Table 3, the G(−273)C polymorphism was associated with the residual HDL-C level (P=0.0310). The genotype frequency of the G(−273)C polymorphism in the HTN group was in accordance with Hardy-Weinberg equilibrium and did not differ from that of the Suita population (P=0.2953). The C(−297)T (P=0.1829) and IMS-JST071749 (P=0.4130) polymorphisms were not associated with the residual HDL-C level. Thus, a positive association was observed between G(−273)C and the HDL-C level in two groups: the Suita population and the HTN group.

Table 3 Lipid levels in the ABCA1 G(−273)C genotypes (hypertension group). Values are mean±SEM. P-values calculated by ANOVA
Full size table

Association between ABCA1 G(−273)C and incidence of MI

We next evaluated whether the ABCA1 G(−273)C polymorphism was associated with the incidence of MI. The HDL-C level in the male MI group (1.09±0.01, P<0.0001) was significantly lower than that in the male Suita subjects (1.44±0.02). The effects of this genotype on the HDL-C level were not observed in this group, probably because a substantial proportion of this group had dyslipidemia and had been treated with hypolipidemic drugs.

No significant association was observed between the ABCA1 G(−273)C polymorphism and the incidence of MI [the MI group: GG n=212 (38.6%), GC n=289 (45.2%), CCn=130 (16.2%); the Suita population: GGn=309 (35.5%), GCn=362 (48.3%), CCn=130 (16.2%), P=0.4443].

Discussion

In the present study, we evaluated the effects of polymorphisms in ABCA1 on the HDL-C level using a large cohort representing the general population in Japan (the Suita Study). To evaluate the genetic influence of ABCA1 polymorphisms on HDL-C level, the HDL-C level was adjusted not only for standard factors but also for other important genetic factors including the ApoA1 and ApoE polymorphisms. Moreover, we reconfirmed the effects of ABCA1 G(−273)C polymorphism on HDL-C in the HTN group. We next investigated the association between the ABCA1 G(−273)C and the incidence of MI, but did not observe any association.

The present study is distinguished by three main features: (1) an association study using a large cohort study (the Suita population), (2) taking into account of the influence of the ApoA1 and ApoE polymorphisms, and (3) a confirmation of the association using another set of subjects (the HTN group).

We found that three SNPs were associated with the HDL-C level in 14 SNPs of the ABCA1 gene in the Suita population. However, if we applied Bonferroni’s correction for multiple tests, three SNPs might not be considered significantly associated with the HDL-C level [G(−273)C, P=0.1036: C(−297)T, P=0.273: IMS-JST071749, P=0.1302, P values are corrected by multiplying with 14 (14 SNPs)]. Thus, we verified this positive association in another set of subjects (the HTN group). This association study revealed that G(−273)C, but not C(−297)T or IMS-JST071749, was associated with the HDL-C level. Thus, it is highly likely that ABCA1 G(−273)C was truly associated with the HDL-C level.

Since the ABCA1 G(−273)C polymorphism is in the promoter region, it is likely that this polymorphism may alter the expression level of ABCA1. However, this polymorphic site had no consensus sequence for transcriptional factors. The TGGGG(−226)(−) insertion-deletion polymorphism, which is one of the polymorphisms in LD with the G(−273)C polymorphism (r 2=0.46667), was in the middle of the consensus sequence of the ZNF202 binding site (GnT repeat)(Porsch-Ozcurumez et al. 2001). The insertion allele, which mainly corresponds to the (−273)C allele, should disrupt this binding site and may be associated with higher transcriptional activity of the ABCA1 gene, which may lead to higher HDL cholesterol levels. However, the C(−297)T polymorphism, which was in more tight LD with the TGGGG(−226)(−) insertion-deletion polymorphism, appeared to have less effect on the HDL cholesterol level than the G(−273)C polymorphism. It remains to be determined whether this discrepancy merely reflects a statistical error or if the G(−273)C polymorphism might have additional functional significance. A more detailed promoter analysis will be needed to determine which polymorphisms are functionally important.

The present study revealed that the ABCA1I823M polymorphism was not associated with the HDL-C level, inconsistent with a previous report (Harada et al. 2003). This discrepancy may be due to the study design, since a small-scale association study has relatively weak statistical power. In the present study, the sample power was 0.77 for the distribution, sample size, frequencies of the alleles, and α value (0.05, two-tailed). The sample size in the previous study (n=410) does not seem to be sufficient to give adequate statistical power. Moreover, the frequency of the I823 allele in the previous study (allele frequency 0.492) was different from that in the Suita population (0.36) and JSNP information (0.38). Thus, the subjects in the previous study did not seem to be representative of the general Japanese population, as noted by Harada et al. (2003).

Recently, the polymorphisms in the promoter region of ABCA1, which corresponds to C(−559)T in the present study and seems to be in tight linkage with G(−273)C (r 2=1, D′-value=1), was found to be modestly, but not significantly (P=0.09), associated with the HDL-C level using LCAS subjects (Lutucuta et al. 2001). The effect of the ABCA1 G(−273)C polymorphism on the HDL-C level was significant, but still relatively weak (r 2=0.0050). Accordingly, the sample size (n=372) in the previous study (Lutucuta et al. 2001) seems to have been too small to detect the effect of polymorphisms on the HDL-C level clearly.

While the ABCA1 G(−273)C polymorphism was associated with HDL-C level, it was not found to be associated with the incidence of MI. The ApoE polymorphism (E2, E3, and E4) had the greatest influence on the HDL-C level among the three polymorphisms, ABCA1 G(−273)C (r 2=0.0050), ApoA1 JST-IMS005603 (0.0100), and ApoE (0.0118). However, the ApoE polymorphism was only weakly associated with the incidence of MI (P=0.0840). Thus, ABCA1 G(−273)C may have too weak an influence on the HDL-C level to alter the incidence of MI through a reduction of the HDL-C level. More large numbers of MI subjects might be necessary to detect the influence of the ABCA1 G(−273)C polymorphism on MI incidence.

In summary, the present study provides the first evidence that the common ABCA1 G(−273)C polymorphism in the promoter region is significantly associated with the level of HDL cholesterol in the Japanese.