Background

Dyslipidemia (defined by elevated levels of fasting and post-prandial plasma triglyceride-rich lipoproteins, abnormally low high-density lipoprotein (HDL) levels and elevated low-density lipoprotein (LDL) concentrations) is associated with atherosclerosis and coronary heart disease (CHD) [1, 2]. Coronary artery disease is a major cause of morbidity and mortality in both industrialized countries and developing countries, such as Algeria [3, 4]. It has been estimated that there were 7.3 million deaths worldwide from ischemic heart disease in 2008 (12.4% of all mortality). Furthermore, it is predicted that CHD will still be the leading cause of death in 2020 [5, 6].

Commonly studied intermediate traits associated with CHD include plasma levels of cholesterol (total cholesterol, LDL-cholesterol and HDL-cholesterol), body mass index (BMI) and blood pressure. These traits are influenced by a combination of genetic [7] and environmental factors (such as diet, alcohol and physical activity) [8–10].

The results of several meta-analyses have firmly established that the apolipoprotein E (APOE) epsilon polymorphism (defined by the rs7412 and rs429358 single nucleotide polymorphisms (SNPs)) is a genetic risk factor for CHD [11–14]. Indeed, APOE has an important role in the metabolism of lipoproteins and is a ligand for LDL-cholesterol and APOE receptors [15]. The epsilon polymorphism in the APOE gene leads to the generation of APOE2, APOE3 and APOE4 isoforms, which are coded by three codominant alleles (designated as ϵ2, ϵ3 and ϵ4). The three isoforms differ by an amino acid substitution at position 112 or position 158 in the 299-amino-acid peptide chain [16]. The isoforms interact differently with specific lipoprotein receptors and thus influence plasma cholesterol concentrations [17]. The APOE ϵ4 allele is associated with higher total and LDL-cholesterol levels and a higher risk of CHD, whereas the APOE ϵ2 allele is associated with the opposite (i.e. protective) effects in Caucasian populations [18–21].

Furthermore, two other APOE polymorphisms have been found to display associations with various metabolic traits. Firstly, the rs439401 SNP was associated with higher plasma triglyceride and lower plasma HDL-cholesterol concentrations in a meta-analysis of genome-wide association studies (GWAS) in 16 European cohorts [22]. Secondly, the rs4420638 SNP is reportedly associated with lower plasma HDL-cholesterol levels, higher total cholesterol and LDL-cholesterol levels and higher total cholesterol/HDL-cholesterol and LDL-cholesterol/HDL-cholesterol ratios [14, 23–25].

Very few studies have investigated putative associations between the APOE epsilon polymorphism and plasma lipid levels in North African populations. Indeed, only two studies (in Moroccan and Tunisian populations) reported that the APOE ϵ4 allele is associated with higher plasma concentrations of total cholesterol and LDL-cholesterol, whereas the APOE ϵ2 allele shows the opposite association [26–28]. However, no data for the rs439401 and rs4420638 polymorphisms in these populations are available.

To the best of our knowledge, the relationship between APOE polymorphisms and plasma lipid and lipoprotein concentrations in an Algerian population has never previously been studied. We therefore decided to assess the relationships between APOE epsilon, rs439401 and rs4420638 polymorphisms and plasma lipid concentrations in a population sample from the city of Oran in north-west Algeria, the ISOR study.

Results

Genotype and allele distributions

The allele and genotype distributions of the APOE polymorphisms are presented in Table 1. There was no evidence of significant deviation from Hardy-Weinberg equilibrium in any distributions.

Table 1 Genotype distributions of the APOE polymorphisms in the ISOR study
Full size table

Linkage disequilibrium

We evaluated the linkage disequilibrium (LD) between the APOE epsilon (rs429358 and rs7412), rs439401 and rs4420638 polymorphisms (Figure 1). The rs439401 and rs4420638 SNPs were not in LD with the APOE epsilon polymorphism (r2<0.16). So the analysis of all polymorphisms was investigated.

Figure 1
figure 1

Relative position in the APOE locus and LD values (D’ in black and white, r2values) of the SNPs. SNPs with the highest level of D’ are shown in black and those with the lowest level of D’ are shown in white. r2 values are indicated in the squares.

Full size image

Association studies

The various associations between the three APOE polymorphisms and the anthropometric phenotypes (weight, waist and hip circumferences and BMI), biochemical phenotypes (glucose, insulin, fasting plasma triglyceride, total cholesterol, HDL-cholesterol and LDL-cholesterol levels) and clinical phenotypes (SBP and DBP) were assessed (Tables 2–3).

Table 2 Association between the APOE epsilon polymorphism and anthropometric, biochemical and clinical parameters in the ISOR study
Full size table
Table 3 Association between the APOE rs439401 and rs4420638 SNPs and anthropometric, biochemical and clinical parameters in the ISOR study
Full size table

The APOE epsilon polymorphism

No significant association could be detected when comparing ϵ2 allele carriers with ϵ3 allele carriers (Table 2). In contrast, ϵ4 allele carriers had significantly higher mean plasma triglyceride (p=0.0002), mean plasma total cholesterol (p=0.009), mean LDL-cholesterol (p=0.003) levels and LDL-cholesterol/HDL-cholesterol ratio (p=0.001) than ϵ3 allele carriers did (Table 2).

The rs4420638 and rs439401 polymorphisms

No significant associations were detected for the rs4420638 SNP in the ISOR study (Table 3). In contrast, the T allele of rs439401 was significantly associated with lower plasma triglyceride levels (p=0.005) (Table 3). This association disappeared after further adjustment for the APOE epsilon polymorphism (p=0.25).

Gene-environment analyses

As gender, menopausal status in women and lipid-lowering drugs use may significantly impact plasma lipid levels, we took into account these confounders and investigated associations between the APOE polymorphisms and plasma lipids, in men and women separately and in the non-menopausal women group. All previously described associations were replicated in each group, meaning that gender and menopausal status had no notable influence on the present associations (data not shown).

Haplotype analysis

We explored the haplotype effects of the APOE epsilon (rs429358 and rs7412) and the rs439401 and rs4420638 SNPs on plasma LDL-cholesterol levels. First, we selected the most informative haplotype configuration. The best haplotype model included the APOE epsilon and the rs4420638 polymorphisms (p=0.002, with 4 d.f.). We then performed haplotype analysis (using Thesias [29]) for the APOE epsilon and rs4420638. Five haplotypes were inferred (ϵ3A, ϵ3G, ϵ4A, ϵ4G and ϵ2A) (Table 4). The test for an overall haplotype effect was significant (p= 0.002).

Table 4 Effect of the APOE haplotypes on plasma LDL-cholesterol levels in the ISOR study
Full size table

When compared with the most frequent (reference) ϵ3A haplotype, the ϵ4A haplotype was associated with higher LDL-cholesterol levels (Δ [95% CI]=+0.35 [0.18 ─ 0.51] mmol/L, p=0.0001), whereas the ϵ2A haplotype was associated with lower LDL-cholesterol levels (Δ [95% CI]=-0.20 [-0.41 ─ -0.01] mmol/L, p=0.05), confirming the respectively deleterious and protective impacts of the ϵ4 and ϵ2 alleles. The ϵ3G haplotype was associated with similar LDL-cholesterol levels (Δ [95% confidence interval (CI)]=0.02 [-0.17 ─ +0.21] mmol/L, p=0.74) compared with carriers of the ϵ3A haplotype; this result suggests that the rs4420638 G allele does not have an effect in an ϵ3 background. Interestingly and contrarily to what we observed for the ϵ4A haplotype, the ϵ4G haplotype was not associated with higher LDL-cholesterol levels, relative to the reference ϵ3A haplotype (Δ [95% CI]=-0.05 [-0.27 ─ -0.17] mmol/L, p=0.81). Hence, the G allele of the rs4420638 SNP may have counterbalanced the deleterious effect of the ϵ4 allele.

Similar haplotype results were obtained for plasma total cholesterol levels (data not shown).

Discussion

The importance of APOE in lipid and lipoprotein metabolism is well established. However the impact of APOE polymorphisms has never been investigated in an Algerian population. So in this study, we assessed the relationship between three APOE polymorphisms (epsilon, rs439401 and rs4420638) and metabolic trait variations in a population from Oran, Algeria, the ISOR study. To the best of our knowledge, this was the first study to characterize associations of the three above-mentioned APOE gene polymorphisms with anthropometric, biochemical and clinical parameters in an Algerian population.

The frequency of the ϵ4 allele in Europe parallels the incidence of CHD and other diseases [11, 30, 31]. It follows a north-to-south gradient and ranges from more than 0.22 in Finland and Greenland [32–34] to less than 0.07 in Greece and Italy [35–37]. In our study, the ϵ4 allele frequency (0.10) was similar to values observed in other North African populations (around 0.10 in Moroccan populations [26, 38] and around 0.08 in Tunisian populations [27, 28]).

The minor allele frequency of the rs439401 SNP (located within the APOE/APOC1 cluster) was 0.38 in the ISOR study. Similar frequencies have been reported in the literature and range from 0.33 to 0.36 [14, 39].

For the rs4420638 SNP (situated lying 14 kb downstream of the APOE locus in the adjacent APOC1 gene), the minor allele frequency (0.11) was markedly lower than the value observed in European samples [14, 24, 40].

In the ISOR study, the ϵ2 allele was not significantly associated with plasma lipid variations (0.06≤p≤0.18). This is probably due to too few of ϵ2 carriers (n=67). Only the ϵ4 allele showed an association with higher levels of total cholesterol and LDL-cholesterol. Furthermore, the ϵ4 allele was associated with higher concentrations of triglycerides, as previously observed by Kofler et al. in a study performed in the United Kingdom [41]. The APOE ϵ4 isoform (which binds preferentially to VLDLs) can interfere with plasma lipase activity or with the triglyceride removal system; this results in delayed lipolysis or clearance of plasma triglycerides in subjects bearing the ϵ4/ϵ3 phenotype [18]. Therefore, our results are consistent with previous studies [42–45]. In contrast with previous studies in multi-ethnic populations [41, 46–48] but consistently with other studies in Tunisian subjects [27, 28], we did not detect significant associations between the APOE epsilon polymorphism and HDL-cholesterol levels in the ISOR sample. Other studies conducted in Arab population samples are therefore needed to conclude.

Despite the deleterious effect of the ϵ4 allele on lipid parameters observed in the ISOR study, the lipid concentrations did not reach abnormal threshold values. A case–control study on the risk of myocardial infarction conducted in Oran in 2001 also showed that plasma concentrations of all lipid parameters were systematically lower in Algerian subjects than in Irish and French individuals [49].

We found significant association between the rs439401 SNP and triglyceride concentrations. This association has previously been observed in European GWAS of blood lipid levels [14, 22]. However, this association disappeared after further adjustment for the APOE epsilon polymorphism, suggesting that the effect of rs439401 was overshadowed by the epsilon polymorphism.

In the ISOR study, the rs4420638 SNP was not significantly associated with any metabolic traits, contrary to what has been previously described [50]. These discrepancies could be explained by differences in the LD structure between rs4420638 and the APOE epsilon polymorphism in the Algerian and European populations (D’=0 and +16 with rs7412 and rs429358, respectively in the ISOR study versus D’=+66 and -69 with rs7412 and rs429358, respectively in the EUR panel of the 1000 Genomes database (December 2012 release, http://browser.1000genomes.org)). Therefore the effect of the rs4420638 SNP could be mainly due to the epsilon polymorphism in European samples but be independent in the ISOR study. This result should be investigated in other North-African populations.

Haplotype analysis was performed to better understand the information provided by the individual SNP analysis. This analysis showed that regarding plasma LDL-cholesterol (or total cholesterol) levels, the rs4420638 polymorphism could counterbalance the deleterious effect of the epsilon 4 allele.

Conclusion

We showed that the impact of the APOE ϵ4 allele on the plasma lipid profile is similar in an Algerian population to what is known in other countries. It is noteworthy that the rs4420638 G allele could counterbalance the deleterious effect of the APOE ϵ4 allele on LDL-cholesterol levels in a North African population. Replications in larger studies are required.

Methods

Subjects

The ISOR (InSulino-résistance à ORan) study

The ISOR study was performed between 2007 and 2009. The study’s objectives and procedures were approved by the independent ethics committee at the Algerian National Agency for the Development of Health Research. The ISOR study was a population-based, cross-sectional study of a representative sample of 787 subjects (378 men and 409 women, aged between 30 and 64) recruited within the city of Oran. Subjects were selected at random from social security rolls. All subjects consented freely to participation in the study. A questionnaire on lifestyle (physical activity, tobacco use and alcohol intake), personal and family medical histories, current medication and socio-economic and educational levels was completed during a face-to-face interview. Anthropometric data were also recorded.

The level of physical activity was defined in quartiles as “none”, “low”, “medium” and “high” after summing exercise scores for sporting activities, walking, housework and physical activity at work.

In terms of tobacco use, participants were categorized as either smokers (i.e. individuals reporting at least one cigarette per day) or non-smokers. In the study questionnaire, subjects were asked to report their weekly consumption of wine, beer, and spirits. As there were very few drinkers (n=25, 3.2%), this variable was not taken into account in the final analysis.

The anthropometric measurements included height, body weight, waist circumference and hip circumference. Height and weight were measured while the subject was barefoot and lightly dressed. The BMI was calculated according to the Quetelet equation. Systolic and diastolic blood pressure values (SBP and DBP, respectively) were measured on the right arm with the subject in the sitting position, using a standard mercury sphygmomanometer. Measurements were made before and after completion of the questionnaire, with an interval of at least 10 minutes. The mean value of the blood pressure readings was considered in the final analysis.

After a 12 h overnight fast, two 15 ml blood samples were collected for each subject (in a disodium EDTA tube for subsequent DNA analysis and in a heparin tube for clinical chemistry tests).

A multichannel analyzer and dedicated kits (Humastar®, HUMAN Diagnostics, Wiesbaden, Germany) were used for the colorimetric, enzymatic measurement of cholesterol (kit: monotest cholesterol with cholesterol esterase, cholesterol oxidase and peroxidase), triglycerides (kit: peridochrom triglyceride with glycerol phosphate oxidase and peroxidase) and glucose (kit: glucose, glucose oxidase and peroxidase). Plasma LDL-cholesterol levels were calculated according to the Friedwald equation. High-density lipoprotein cholesterol levels were measured after sodium phosphotungstate/magnesium chloride precipitation of chylomicrons and VLDL and LDL-cholesterol and then centrifugation. Plasma insulin levels were measured in a microparticle enzyme immune assay running on an AxSYM analyzer (Abbott Laboratories, Abbott Park, Illinois, USA).

Genomic DNA was extracted from white blood cells by using the Stratagene® kit (Agilent Technologies, Les Ulis, France), according to the manufacturer’s protocol.

Genotyping

The APOE SNPs (rs429358, rs7412, rs439401 and rs4420638) were genotyped using KASPar technology (KBioscience, Hoddesdon, UK) with the following probes: rs429358: [GACATGGAGGACGTG[C/T]GCGGCCGCCTGGTGC], rs7412: [GATGACCTGCAGAAG[C/T]GCCTGGCAGTGTACC], rs439401: [GCCGGCACTCTCTTC[C/T]CCTCCCACCCCCTCA], rs4420638: [TGCTACACTTTTCCT[A/G]GTGTGGTCTACCCGA]. The genotyping success rates ranged from 93% to 96%.

Statistical analyses

Statistical analyses were performed with SAS 9.1 software (SAS Institute Inc., Cary, NC, USA). The Hardy-Weinberg equilibrium was tested using a χ2 test with one degree of freedom (d.f.).

Intergroup comparisons of means were performed with (i) a general linear model comparing ϵ2 carriers (ϵ2ϵ2 and ϵ2ϵ3 subjects) or ϵ4 carriers (ϵ4ϵ4 and ϵ4ϵ3 subjects) with ϵ3 carriers (ϵ3ϵ3 homozygotes) for the APOE epsilon polymorphism and (ii) an additive model for the rs439401 and rs4420638 polymorphisms. Subjects with the ϵ2ϵ4 genotype (n=3) were excluded from the analyses because of the possible opposing biological effects of the ϵ2 and ϵ4 alleles.

Data on triglycerides, glucose and insulin levels were log-transformed to obtain normal distributions. Estimated means were subsequently back–transformed for presentation in the tables.

For anthropometric variables, the confounding variables were age, gender, smoking status and physical activity. For biological and biochemical variables, the confounding variables were age, gender, BMI, smoking status and physical activity. For the rs439401 and rs4420638 polymorphisms, data were further adjusted for the APOE epsilon polymorphism coded in three genotypes as follows: ϵ2 carriers (ϵ2ϵ2 and ϵ2ϵ3 subjects); ϵ4 carriers (ϵ4ϵ4 and ϵ4ϵ3 subjects) and ϵ3 carriers (ϵ3ϵ3 subjects).

After Bonferroni correction, only associations with an uncorrected p value below 0.017 were considered to be statistically significant (i.e. 0.05 divided by the number of polymorphisms considered).

General linear models were used to investigate potential interactions by adding an interaction ((gender or BMI) x polymorphism) term.

Linkage disequilibrium figure and values were calculated with Haploview 4.2 (http://www.broadinstitute.org/scientific-community/science/programs/medical-and-population-genetics/haploview/haploview).

Haplotype frequencies derived from all studied polymorphisms were estimated independently of phenotype. A two-step haplotype analysis was performed. Firstly, in order to reduce the haplotype dimension and select the most informative, parsimonious haplotype configuration for the prediction of phenotypic variability, we applied the maximum likelihood model to all the possible 1 to k-loci combinations of polymorphisms that could be derived from the 4 APOE SNPs with the GridHaplo software [51]. Akaike’s information criterion (AIC) was calculated for each model (including a model with no polymorphisms) [52]. All AIC values were rescaled by subtracting the smallest AIC value obtained for the whole set of models. According to a rule derived by extensive Monte Carlo simulation, all models with a rescaled AIC ≤2 can be considered to be “equivalent” to the model with the lowest min AIC. The most parsimonious of the latter (corresponding to the minimal haplotype configuration) was selected. Secondly, haplotype analyses were performed using the Thesias software package (http://ecgene.net/genecanvas) [29]. Haplotype analyses were adjusted for age, gender, BMI, smoking status and physical activity level.