Introduction

Elevated triglycerides (TG), total cholesterol (TC), low density lipoprotein (LDL-C) and apolipoprotein B (apoB), together with decreased levels of high density lipoprotein (HDL-C), and apolipoprotein AI (apoAI), are associated with an increased risk of CHD (1–4). Genetic and environmental factors are believed to regulate both metabolic and cellular function, and the concentration of lipoproteins and apolipoproteins (4,5). The heritability estimates of lipoproteins and apolipoproteins from twin and family studies are in the range of 40–80% (6,7), suggesting a considerable genetic contribution, and discovery of the genes that contribute to these changes may lead to a better understanding of these processes.

Two of the most well-studied common genotypes that determine plasma lipid levels are the two single-base changes in the coding region of the APOE gene that create the ε2, ε3 and ε4 alleles (8,9). ApoE circulates in the blood as a component of a number of lipoprotein particles, including chylomicrons and their remnants; TG-rich very low density lipoprotein (VLDL) particles and their remnants, and HDL (10). ApoE is a ligand for receptors involved in the clearance of these lipoproteins, particularly in the liver (10), and the three common alleles of APOE code for three isoforms which vary in their affinity for binding to the LDL-receptor (11). The APOE ε2 allele has reduced binding affinity, causing lower efficiency of clearance of VLDL and chylomicron remnants from the blood to the liver, thus slowing the uptake of postprandial lipoprotein particles compared with carriers of the ε3 and ε4 alleles (11,12). APOE ε4 and ε3 bind with approximately equal affinity to lipoprotein receptors while ε2 binds with less than 2% affinity (13). As would be expected from the relationship between APOE genotypes and lipid levels, meta analyses also has clearly demonstrated a strong relationship with risk of CHD (14), and compared with subjects with the genotype ε3/ε3, the odds ratio for coronary disease was 0.80 (95% confidence interval [CI]: 0.70–0.90) in ε2 carriers and 1.06 (95% CI: 0.99–1.13) in ε4 carriers.

The APOE gene is located on the long arm of chromosome 19q12–13.2 within a cluster of other genes coding for apolipoprotein CI (APOC1), apolipopro-tein CIV (APOC4) and apolipoprotein CII (APOC2). ApoCI is a constituent of VLDL and HDL, inhibiting the binding of VLDL to LDL-related proteins (15,16), as well as the apoE-mediated binding of VLDL to the LDL receptor (17)—leading to reduced levels of LDL-C (18). ApoCII is an essential cofactor of lipoprotein lipase (LPL), the rate-limiting enzyme for the hydrolysis and removal of triglycerides in chylomicrons and VLDL (19), and individuals who are completely deficient in apoCII have chylomicronaemia and grossly elevated plasma TG levels, although their parents, who have half normal plasma levels of apoCII, have normal TG levels (20). The role of apoCIV in human lipoprotein metabolism is unclear, mainly due to the relatively low apoCIV gene expression in the liver and the low levels of the apoCIV protein detectable in plasma (21), where it is associated mainly with VLDL particles. However, mice transgenic for APOC4 show elevated levels of TG (22,23) and in one study, common missense variants (L96RL, rs5167) were associated with differences in plasma TG levels in women (23).

Despite the fact that each protein encoded by the APOE/C1/C4/C2 gene cluster has a clear or potential role in plasma lipid metabolism, the association between variation in the entire cluster and plasma lipid concentration is understudied. Similarly, while the association between the APOE variants and CHD risk has been well studied and confirmed by meta-analysis (14), the association between the other cluster gene variants and CHD is less clear. The primary purpose of this study is to evaluate the contribution of common SNPs and haplotypes in the APOE/C1/C4/C2 gene cluster to plasma lipid concentrations in a large group of healthy middle-aged men. A secondary aim was to determine to what extent SNPs and haplotypes, associated with differences in plasma lipid concentrations, were contributing to the ε2 low- ering and ε4 raising effect on prospective risk of CHD.

Materials and Methods

Study Subjects and Data Collection

The prospective Second Northwick Park Heart Study (NPHSII), commenced in 1989, and includes 3,052 middle-aged men (49–64 years) recruited from nine general medical practices in the UK. Participants were free of unstable angina, myocardial infarction, evidence of silent infarcts, coronary surgery, anticoagulant drugs (including aspirin), cerebrovascular disease, malignancy and any condition or disease preventing the attainment of written, informed consent or long-term follow-up. Informed consent was received from all study subjects. Information on lifestyle habits, height, weight, blood pressure and a number of blood biomarkers was recorded at baseline and on subsequent prospective follow-ups. Details of recruitment, measurements, follow-up and definitions of disease are described elsewhere (24–26). Lipids, TC and TG concentrations were measured with automated enzyme procedures, while serum apoAI and apoB concentrations were measured by immuno-turbidometry. CHD events taken as end points were fatal (sudden or not) and nonfatal MI, based on World Health Organization criteria (27), plus coronary artery surgery and silent MI on the follow-up ECG or coronary revascularization procedures and sudden unexplained death. Clinical information for each event was assembled by inquiries through the participating practices, hospitals attended and, for fatal events, coroners ′ offices. This information was collated and submitted to an independent assessor who assigned qualifying events to the appropriate category. A DNA sample was obtained from 2,767 of the men from a blood sample taken at the time of recruitment. Interviews and repeat measurements were conducted annually for surviving participants.

Genetic Analysis

Nine SNPs were genotyped using an Illumina GoldenGate candidate gene chip (Illumina Inc., San Diego, CA, USA) (28). APOE: rs405509 (promoter), rs439401 (2 kb); APOC1: rs4420638 (200bp), rs4800770 (5 kb), rs7259004 (10kb); APOC4: rs12691089 (exon 2), rs5167 (exon 3); APOC2: rs5127 (3’URT) rs10413089 (3 kb). Tagging SNPs (tSNPs) for the APOE/C1/C4/C2 genes cluster were selected with Tagger (developed by scientists at Broad Institute, Cambridge, MA, USA) (29) using the CEU panel of HapMap (see International HapMap Project [https://doi.org/hapmap.ncbi.nlm.nih.gov]), applying an r2 threshold of 0.8 and a minor allele frequency threshold of 0.04. The tSNPs were optimized for the Illumina platform by preferentially selecting the SNP with the highest genotyping success rate in each block as a tag and reevaluating the r2 in the sample. Additionally, two SNPs, APOE: rs429358 (exon 4), rs7412 (exon 4), which dectect the E2/3/4 protein isoforms were genotyped by polymerase chain reaction (PCR) and restriction enzyme digestion as described previously (30,31).

Statistical Analysis

Hardy-Weinberg equilibrium (HWE) was assessed with a chi-square test. Pairwise linkage-disequilibrium (LD) between the SNPs was calculated from the genotype data measured as D ′ and r2. Haplotype frequencies were calculated using the PHASE algorithm (32 –34). Missing SNP genotypes were imputed by using the haplotypic reconstruction (35). As for both baseline (measured at the start of the study) and “average ” (the mean of all the available annual measurements for each individual in the five-year follow-up period), lipids levels were used (25,26). The lipid concentrations were transformed logarithmically to reduce skewness in their distributions. The Student t test and chi-square test were used to compare a number of phenotypes between those with and those without a CHD event. General linear model analysis was performed to identify associations between the SNPs or haplotypes and plasma lipid levels using indicator variables. The common homozygote genotype of each SNP and the most frequent haplo-type were used as a reference. Stepwise regression analysis with Akaike ′s information criterion included all SNPs genotyped and was performed to identify the independent determinants of plasma lipid levels and those with the strongest contribution to overall variance under an additive model. All linear models included adjustment for age and recruiting center (clinics). Since associations were formally examined for 10 SNPs and five lipid traits, some adjustment of the significance level is needed to account for multiple comparisons. However, there are high correlations between pairs of the phenotypes, and many of the SNPs also showed a degree of LD, therefore the Bonferroni correction would be too conservative (36 –38). In addition, type I errors are random, and patterns in results that confirm previous reports should be given more weight than isolated results with a single low P value (36,38). Also, correction for multiple comparisons largely increases the likelihood of type II errors and important differences are considered nonsignificant (38). To take this into account, for individual tests of association, rather than applying a correction for multiple testing at global significance level of 5% (P < 0.05), we defined statistical significance as less than 1% (P < 0.01). Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (95% CI) describing the association between SNPs, haplotypes and risk of CHD events.

All supplementary materials are available online at https://doi.org/www.molmed.org.

Results

A total of 2,767 men were included in the analysis. The mean (± SD) age was 56.1 ± 3.44 and the mean (± SD) follow-up time was 12.2 ± 3.53 years. During the follow-up period (overall 33,271 person years), 275 subjects experienced a CHD event. Table 1 shows the baseline charac- teristic of the subjects who remained event free and those who developed CHD over 15 years. CHD patients were older and characterized by higher body mass index (BMI); systolic blood pressure (SBP) and diastolic blood pressure (DBP); concentration of TC, TG, LDL-C and apoB; and lower HDL-C. In addition, the CHD group included a higher proportion of smokers and diabetics than the non-CHD group.

Table 1 Mean baseline characteristics (SD) of event-free subjects and those who developed CHD over 15 years.
Full size table

Allele Frequencies and Pairwise LD Structure at the APOE/C1/C4/C2 Gene Cluster

All samples were genotyped for 11 SNPs, 9 using the Illumina platform and 2 by PCR and restriction enzyme digestion. The frequency of the rare allele for the rs12691089 SNP was 0.002%, so it was not considered further. All polymorphisms genotyped were in Hardy-Wein-berg equilibrium. The LD structure, expressed as D′ and r2, is shown in Figure 1. Although most of the SNPs throughout the APOE/C1/C4/C2 gene cluster did not show statistically significant evidence of LD, weak LD (D′ ≤ 0.05 and r2 ≤ 0.05) was detected between SNPs rs7259004, rs429358, and rs4420638, SNPs rs5167, rs405509, and rs439401; SNPs rs5127, rs4420638 and rs4803770; SNPs rs10413089, rs429358, rs439401, and rs4420638. Moderate LD (D′ < 0.50 and r2 ≤ 0.50) was detected between SNPs rs4420638 and rs429358, and SNPs rs5127 and rs5167.

Figure 1
figure 1

Pairwise linkage disequilibrium structure represent as D’ (different color intensities) and r2 values (numbers) in 10 SNPs. The rs numbers and the relative physical distance between the SNPs are shown above (genes are the larger rectangular boxes). The color gradient indicates relative level of LD from black complete to white no LD.

Full size image

Association between APOE/C1/C4/C2 Gene SNPs and Plasma Lipid Levels

The genotype frequencies and uni-variate analysis for each of the 10 SNPs with plasma lipid traits are presented in Table 2. As expected, compared with APOE ε3 homozygotes, carriage of the APOE ε4 allele was associated with higher plasma LDL-C; apoB and TG concentration and of the ε2 allele with lower LDL-C and apoB (but higher TG). The rare alleles of the SNPs rs405509, rs4420638 and rs4803770 were associated with higher plasma LDL-C, and apoB concentration compared with the noncarriers, while the rare alleles of rs7259004 were associated with lower plasma LDL-C, apoB concentration and higher apoAI concentration compared with noncarriers. Carriers of the rare allele of rs4803770 had lower plasma HDL-C concentration, and carriers of rs5127 and rs5167 had higher plasma HDL-C and apoAI concentration.

Table 2 Univariate analysis between APOE/C1/C2/C4 genotypes and lipid traits.
Full size table

To determine the independent effect of the cluster SNPs, stepwise analysis with the Akaike′s information criterion was carried out as shown in Table 3. For all traits, the presence of APOE polymorphisms was the main predictor of plasma lipid concentrations after adjustment for age, recruiting center, BMI and smoking. Smoking had a modest effect on all traits (0.1–0.6% of the residual variance), with significant effects on apoB, TG, HDL-C and apoAI. The common APOE polymorphisms explained 3.4% of the residual variance in LDL-C and apoB concentration, and 0.1% and 0.4% of the residual variance in TG, HDL-C and apoAI concentration, respectively. By comparison, the other genotypes explained an additional 1.2% for LDL-C, 1.3% for apoB, 0.05% for TG, but an additional 1.4% for HDL-C and 1.2% for apoAI plasma concentration.

Table 3 Multivariate regression with stepwise analysis with Akaike′s information criterion (AIC) of the independent contributions of genotype and covariates to lipid traits.
Full size table

Association between APOE/C1/C4/C2 Gene Cluster Haplotypes and Plasma Lipid Levels

Initially, haplotypes were inferred using all 10 SNPs as shown in supplementary Table S2. Since each gene in the cluster might be characterized by a different haplotype, in turn associated with a phenotype linked to the gene function, the haplotypes were re-estimated for each phenotype using only the SNPs which showed statistically significant association with the trait in the univariate analysis.

SNPs rs429358, rs7412 (APOE ε2/ε3/ε4), rs439041 (APOE), rs4420638 and rs4803770 (APOC1), and rs7259004 (between APOC1 and APOC4) were included in the haplotype estimation for the LDL-C and apoB traits. The ten inferred haplotypes that had a relative frequency of more than 1% (which included 97.3% of the inferred haplotypes), were tested for their association with LDL-C and the results are shown in Table 4. The single haplotype bearing the APOE 84 variant (A3 frequency 10.9%) was associated with higher levels of LDL-C and apoB and the two haplotypes bearing the APOE ε2 variants frequency 4.4% and 4.0% (designated A5 and A7) were significantly associated with lower levels of LDL-C and apoB. No other haplotypes were significantly associated with effects on these or other traits. The haplotypes explained 2.2% (P < 0.001) and 2.3% (P < 0.001) of the residual variance in LDL-C and apoB concentration, respectively.

Table 4 Common haplotypes estimation using different SNPs from the APOE/C1/C2/C4gene cluster and their association with LDL-C, apoB, HDL-C, and apoAI.
Full size table

For the HDL-C and apoAI traits, SNPs rs429358, rs7412 (APOE ε2/ε3/ε4), rs4803770 and rs7259004 (APOC1), and rs5167 (APOC2) were included in the haplotype estimation. Eight haplotypes with a relative frequency of more than 2% (which included 92.6% of the inferred haplotypes), were analyzed. Compared to the most common haplo-type, two haplotypes (B3 and B8), which both carry the rare allele of rs5167 (frequency 11.8% and 2.8%, respectively), were associated with significantly higher HDL-C and apoAI concentration (Table 4). Compared to noncarriers, carriers of this haplotype had, respectively, 4.0%, 6.6% higher levels of HDL-C and 2.4%, 4.1% higher levels of apoAI. Conversely, hap-lotype B4 (an ε4-containing haplotype) was associated with lower HDL-C and apoAI concentration. Compared to non-carriers, carriers of this haplotype had, respectively, 4.3% and 3.2% lower levels of HDL-C and apoAI. The haplotypes explained 0.8% (P ε 0.001) of the residual variance in HDL-C and apoAI concentration.

Association between APOE/C1/C4/C2 Cluster SNPs and Haplotypes with CHD

The association between the APOE/C1/C4/C2 SNPs and risk of CHD is presented in Table 5. Compared to the APOE ε3/ε3 genotype, carriers of the APOE 82 allele had lower risk of CHD with a HR of 0.63 (95% CI: 0.42–0.95) after adjustment for nonlipid classical risk factors, an effect unchanged after adjustment for lipid risk factors (HR = 0.65, 95% CI: 0.43–0.98). In addition there was evidence of a protective effect associated with the rs5127 SNP; where compared with noncarriers, subjects with one or more rare allele had a HR of 0.71 (95% CI: 0.55–0.91), an effect which was relatively unchanged after adjustment for nonlipid and lipid risk factors (HR = 0.72, 95% CI: 0.56–0.93). By contrast, carriers of the rs4803770 rare allele had a higher risk of CHD (HR = 1.42, 95% CI: 1.09–1.85) compared with noncarriers, with a similar effect after adjustment (HR = 1.36, 95% CI: 1.04–1.78). There were no significant associations between any other SNPs and risk of CHD.

Table 5 Cox proportional hazards models for the association between APOE/C1/C4/C2genotypes and risk for CHD event.
Full size table

Haplotype estimation with CHD events was determined using the common APOE SNPs rs429358, rs7412, and rs4803770 (APOC1) and rs5127 (APOC2), present in Table 6. Eight haplotypes with a relative frequency of more than 1% (which included 99.0% of the inferred haplotypes), were identified. The APOE ε2 variant occurs on two haplotypes both associated with (nonsignificantly) lower CHD risk. The rare allele of rs5127 occurs on four different haplotypes and in each case was associated with a lower CHD risk than the comparable haplotype lacking this allele (for example, C3 versus C2, C5 versus C4, C8 versus C6, and C7 versus C1), but because of the low frequency, none of these effects were statistically significant. The APOE ε4 variant occurs on two haplotypes, with the haplotype lacking the rare allele of rs5127 showing a nonsignificantly higher risk. The rs4803770 rare allele also occurs on two haplotypes with again the haplo-type C2 lacking the common allele of rs5127 showing a significantly higher risk compared with the common C1 haplotype. The consequence of this is that carriers and homozygotes of haplotype C2 show a significantly higher risk of CHD (HR = 1.37, 95% CI: 1.15–1.63, and HR = 1.63, 95% CI: 1.17–2.26) compared with noncarriers, and this estimate was relatively unchanged after adjustment for both nonlipid and lipid risk factors (HR = 1.29, 95% CI: 1.08–1.54, and HR = 1.54, 95% CI: 1.11–2.15). In addition, carriers and homozygotes of haplotype C4 show a significantly higher risk of CHD (HR = 1.32, 95% CI: 1.07–1.62) compared with noncarriers, and, once more, this estimation was relatively unchanged after adjustment for age, clinic, BMI, diabetes, total cholesterol, smoking and systolic blood pressure (HR = 1.38, 95% CI: 1.12–1.70). The analysis was repeated using the THESIAS program and similar results were obtained (not shown).

Table 6 Cox proportional hazards models between inferred common haplotypes, using different SNPs from the APOE/C1/C2/C4gene cluster, and CHD events. The reference level is noncarriers of the haplotypes.
Full size table

Discussion

In this paper we have used a tagging SNPs (tSNPs) approach to capture essentially all of the common genetic variation at the APOE/C1/C4/C2 gene cluster, allowing a comprehensive analysis of the genetic architecture of this locus and its relationship with plasma lipid levels and prospective CHD risk to be carried out. The major findings are that the APOE δ2/δ3/δ4 SNPs have the largest effect on LDL-C and apoB levels, with common SNPs elsewhere in the gene cluster increasing the proportion of sample variance explained by roughly 25%. By con- trast, other SNPs, particularly one in APOC4, have a proportionally larger effect on the locus variance of HDL-C and apoAI levels, explaining roughly 90% compared with roughly 10% for APOE. Finally, we present suggestive evidence that the well-known effects on CHD risk associated with APOE ε2/ε3/ε4 SNPs are modified by variants elsewhere in the cluster, particularly in APOC1 and APOC2.

SNPs, Haplotypes and Lipid Levels

Of the 10 SNPs examined in APOE/C1/C4/C2 gene cluster, 5 SNPs, 2 in APOE and 3 in APOC1, had significant effects on plasma LDL-C and apoB levels. Only the common APOE polymorphisms were associated with modest but statistically significant effects on plasma TG levels in univariate analysis, but four SNPs—two in APOC1, one in APOC4 and one in APOC2—had significant effects on the plasma HDL-C, and three of these had significant effects on apoAI levels. Multivariate analysis suggested that the strongest associations between common genetic variations in this gene cluster are with the APOE ε2 and ε4 alle-les, with their effects mainly being to influence LDL-C and apoB concentration. APOE genotype was associated with about a threefold greater effect on LDL-C and apoB concentration than all other SNPs in the gene cluster.

To fully explore the information provided by the SNPs analysis, a haplotype analysis was performed. In general, the genotyped SNPs in the APOE/C1/C4/C2 gene cluster did not show strong evi- dence of LD, since, apart from the ε2/ε3/ε4 SNPs, they were chosen to be tagging SNPs. By use of all 10 SNPs, more than 28 haplotypes were observed with a frequency of greater than 0.5%, but, of these, only some of the SNPs were required to explain the genetic architecture at this locus. Ten haplotypes using seven SNPs were inferred with a frequency greater than 0.5%, representing 92.0% of the subjects, for the analysis of the LDL-C and apoB phenotypes. Compared to the common haplotype H1, only haplotypes with APOE ε2 and ε4 al-leles had significant effects on the plasma lipid levels. For the HDL-C and apoAI phenotypes, eight haplotypes were inferred with a frequency greater than 0.5%, representing 92.6% of the subjects. Compared to the common haplotype H1, three haplotypes had a significant independent association with HDL-C and apoAI concentration, with one ε4-carrying haplotype having lower levels, and one common ε3-carrying and one rare ε2-carrying haplotype having higher levels and with both of these haplotypes including the rare allele of the APOC4 SNP rs5167.

It is possible that these more rare haplo-type effects are due to chance, and confirmation of these effects in larger samples is required. This SNP has been associated previously with elevated TG levels in women (but not men) (23) and this gender difference may explain the lack of association with TG levels in the all male NPHSII samples. To attempt to address this in part, we have carried out an in sil-ico replication of this SNP effect by analyzing the association in a previously published study (39) with a focus on 3,403 middle-aged (39–63 years) men from the Whitehall II study (WH-II). As shown in supplementary Table S4, the minor allele was associated with significantly higher levels of HDL-C (+2.1%) and apoAI (+1.1%) concentration, with no significant effect on LDL-C, apoB or TG levels. There were no significant effects of this SNP with any lipid traits in the smaller group of women (n = 1,202). It thus appears that the modest HDL and ApoAI raising effect associated with this SNP is of interest.

These data confirm previous studies that have shown that the APOE ε2 and ε4 alleles account for a significant proportion of lipid and apolipoprotein variability, with estimates between 3% and 6% of the LDL-C variation (40,41). While other studies (28,42,43) also have identified that there are other important variants in the APOE gene or in other nearby genes that influence variation in LDL-C, the analysis here indicates that the effects are small at best. With a sample of 2,767, the study has the power (α = 0.05, β = 0.80) to detect an effect on LDL-C of 0.15, 0.12, 0.10 mml/L with SNPs with a MAF of 10%, 20% or 30%, respectively. However while the ε2/ε3/ε4 SNPs are having a small effect on HDL-C and apoAI levels, a single SNP in APOC4, rs5167, is having a significant effect on these traits.

Mechanism of SNP Effects

The mechanism of the effect of the APOE ε2/ε3/ε4 SNPs on LDL-C and apoB levels is well documented and is due to the impact of the resulting amino-acid changes causing changes in the affinity of the lipoprotein particles which carry these “ligand” apoproteins for LDL receptor binding (11). The number of apoE molecules per particle is thought to be in the range of 2–4 (10,44), with in vitro data suggesting that two apoE molecules may act together to improve binding (44). The lipoprotein particles in the blood of an individual carrying ε2 or ε4 will therefore be a spectrum of “ε3-only-containing” and “ε3-and-ε2-containing” or “ε3-and-ε4-containing” particles, but a particle with two ε3 molecules and one ε2 (or e4) may be able to be taken up as effectively as an “ε3-only” particle. Therefore, the amount of APOE gene transcription and thus mRNA and protein being made in the liver and secreted may also have an influence on lipoprotein metabolism. This is supported by studies that have examined the association between APOE promoter variants and plasma lipid levels, and, in particular, the rs405509 -219 G < T variant (45,46). It has been shown that the -219T allele is associated with decreased transcriptional activity (45) and plasma apoE concentration (47,48), and, as expected, from the known role of apoE in lipoprotein clearance with moderately increased LDL-C concentrations in at least some studies (49). Several studies have also suggested -219T allele carriers have an increased risk of myo-cardial infarction (47) and premature CHD (50) after adjustment for lipid traits. In this study, the rare allele of rs405509 (that is, -219T) was associated with significantly higher LDL-C and apoB levels in univariate analysis but not in multivariate analysis. By inspection of the haplotype data, the rare allele occurs on three independent haplotypes and only on the ε4-carrying haplotype is it associated with a significant LDL and apoB raising effect. This suggests that the effect, at least on lipid traits may not be statistically independent.

The other relevant potentially functional SNP s5167 is a T < G missense SNP located in exon 3 of gene APOC4 which changes Leucine at position 96 to Arginine. In both this and the WHII study, the 96R allele was associated with significantly higher levels of HDL-C and apoAI concentration. While this SNP may be in LD with a functional SNP elsewhere in the gene cluster, for example, determining expression of one or more of the apoproteins, the amino acid change from a neutral to a charged amino acid is likely to affect function. The computer prediction program POLYPHEN predicts that this change is likely to be damaging. As shown in Supplementary Figure S1, the amino acid at position 96 is an argi-nine in mouse, rat, rabbit and cow, and thus the rare allele restores the common sequence. It may thus be that the amino acid-change results in an altered relative affinity of the apoCIV for VLDL versus HDL particles, which affects the rate of lipolysis of VLDL by LPL. However the SNP was not significantly associated with effects on TG levels in either study, so it may be influencing HDL and apoAI levels by a more direct mechanism. Since at this time the exact role of apoCIV in lipoprotein metabolism is unclear, further speculation is unwarranted.

SNPs, Haplotypes and CHD Risk

We confirm here the well-reported effects of the common APOE variants on CHD risk (14), but apart from the reported higher risk associated with rs405509 (47,49), there is little data for other SNPs in the APOE/C1/C4/C2 gene cluster or for the gene cluster haplotypes. The APOE ε2 allele and the APOC2 SNP rs5127 were both independently associated with roughly 30% lower prospective CHD risk, effects which remained after adjustment for all other risk factors. By contrast, the APOC1 SNP rs4803770 was associated with 40% higher prospective CHD risk after adjustment for all other risk factors. None of the other SNPs had a significant effect on prospective CHD risk. We were not able to confirm previous reports of an association between rs405509 and higher risk of CHD, but the study is underpowered to detect the reported effect of 1.29 (47).

We used haplotype analysis to explore whether the effect of the common APOE genotypes were modified by other SNPs. The APOE ε2 allele was found on two haplotypes, one of which also carried the APOC2 rs5127 rare allele, and both of which were associated with low CHD risk. There was suggestive evidence that the protective effect of ε2 was enhanced by the APOC2 rs5127 rare allele, with this haplotype being associated with the lowest risk, although its rarity means the effect was not statistically significant. The APOE ε3 allele occurred on four common haplotypes and one of these, haplotype C2, which also carries the minor allele of rs4803770 and the major allele of rs5127, was associated with a significantly higher (25%) prospective CHD risk after adjustment for all other risk factors. The consequence of this was that individuals who were homozygous for this haplotype, who represented 5.7% of these subjects, had a 63% higher risk while individuals heterozygote for C2 (40% of the subjects) had ~37% higher risk compared with noncarriers of C2, with these effects being only slightly attenuated by adjustment for CHD risk factors. The APOE ε4 allele was found on two haplotypes, but, interestingly, only one of these, haplotype C4, which also carried the risk alleles of rs4803770 and rs5127, was associated with higher risk of CHD. The combined group of homozygotes and those heterozygote for the C4 haplotype had a 38% higher risk compared with noncarri-ers of C4. Although for all these haplo-types, the risk effect could be at least partly due to the associated effects on plasma lipid levels; based on statistical arguments, this seems unlikely.

The mechanism of the SNP and haplo-type effects in the APOE/C1/C4/C2 gene cluster with CHD risk effect is unclear. The SNP 5127 is a G > T change located in the 3′UTR of APOC2 and, by inspection, the T allele creates a new NKB4 binding site. Whether or not this is functional is currently unknown. The SNP rs4803770 is located 4.5 kb downstream of APOC1 (3′ end), and the minor allele was associated with a significant raising effect of ~5.3% on LDL-C and apoB concentrations, but its impact on CHD was maintained after adjustment for lipid levels. In silico, this variant does not occur within a recognized transcription factor binding site. For both of these sites, it is likely that they are markers by LD for functional sites affecting transcription of one or more of the apoproteins in the cluster.

Limitations

A major limitation of this study is the relatively small number of CHD events, and, clearly, the association of SNPs and haplotypes with CHD risk needs to be interpreted with caution before confirmation in larger studies. However, the prospective nature of the NPHSII study reduces the possibility of the result being confounded by population stratification (51). The study also would be strengthened by measures of plasma apoE and of the other apoproteins in the cluster, since this would allow determination of to what extent the effects seen on lipid traits and CHD risk are being mediated through differences in levels of circulating proteins. As well as having a role in lipoprotein metabolism, apoE has antiox-idant properties (47), and measures of antioxidant potential may also be of value.

Disclosure

The authors declare that they have no competing interests as defined by Molecular Medicine, or other interests that might be perceived to influence the results and discussion reported in this paper.