Epipolymorphisms within lipoprotein genes contribute independently to plasma lipid levels in familial hypercholesterolemia

Abstract

Gene polymorphisms associated so far with plasma lipid concentrations explain only a fraction of their heritability, which can reach up to 60%. Recent studies suggest that epigenetic modifications (DNA methylation) could contribute to explain part of this missing heritability. We therefore assessed whether the DNA methylation of key lipoprotein metabolism genes is associated with high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C) and triglyceride levels in patients with familial hypercholesterolemia (FH). Untreated FH patients (61 men and 37 women) were recruited for the measurement of blood DNA methylation levels at the ABCG1, LIPC, PLTP and SCARB1 gene loci using bisulfite pyrosequencing. ABCG1, LIPC and PLTP DNA methylation was significantly associated with HDL-C, LDL-C and triglyceride levels in a sex-specific manner (all P < 0.05). FH subjects with previous history of coronary artery disease (CAD) had higher LIPC DNA methylation levels compared with FH subjects without CAD (P = 0.02). Sex-specific multivariable linear regression models showed that new and previously reported epipolymorphisms (ABCG1-CpGC3, LIPC-CpGA2, mean PLTP-CpGC, LPL-CpGA3, CETP-CpGA2, and CETP-CpGB2) significantly contribute to variations in plasma lipid levels (all P < 0.001 in men and P < 0.02 in women), independently of traditional predictors such as age, waist circumference, blood pressure, fasting plasma lipids and glucose levels. These results suggest that epigenetic perturbations of key lipoprotein metabolism genes are associated with plasma lipid levels, contribute to the interindividual variability and might partially explain the missing heritability of plasma lipid levels, at least in FH.

Keywords: :

• plasma lipid levels
• coronary artery disease
• missing heritability
• candidate genes
• DNA methylation

Introduction

High low-density lipoprotein cholesterol (LDL-C) and triglyceride (TG) levels, as well as low high-density lipoprotein cholesterol (HDL-C) concentrations, are strong independent risk factors for cardiovascular disease (CVD).^1,2 Several primary (heritable) and secondary (environmental) factors are recognized determinants of plasma lipid concentrations. According to heritability studies, inheritance could account for up to 60% of the interindividual variability in plasma lipid concentration.^3,4 However, genetic association studies have so far identified only a few well validated genes that account for a small proportion of the heritability estimates (~20 to 25%).^5,6 Therefore, approximately 35 to 40% of plasma lipid level heritability cannot be explained by the current knowledge, a concept called missing heritability.

Previous epigenetic studies have suggested that DNA methylation, the most stable epigenetic mark, could contribute to explain the missing heritability of most complex traits, including plasma lipid levels.⁷ DNA methylation is a non-traditional heritable factor involved in gene transcription regulation, especially by disrupting the binding of transcription factors.⁸ It implicates the addition of a methyl group by a DNA methyltransferase at the position 5′ of the cytosine pyrimidine ring, most exclusively when cytosines are located upstream of guanines (sequence called CpG dinucleotides).⁹ CpG dinucleotides are often clustered in CpG islands (CpGi), which are commonly located in regulatory regions of many genes.¹⁰ Some groups have reported that genome-wide and gene specific DNA methylation levels are associated with plasma lipid concentrations and CVD risk.^11-15 More recently, our group reported that DNA methylation levels at the ATP-binding cassette transporter A1 (ABCA1), cholesteryl ester transfer protein (CETP) and lipoprotein lipase (LPL) gene promoter loci are especially associated with variations in HDL size and cholesterol contents, as well as previous history of coronary artery diseases (CAD) in familial hypercholesterolemia (FH).^16,17 However, the association between DNA methylation levels of other key lipoprotein metabolism genes and plasma lipid concentrations remains to be determined.

In the current study, we analyzed the epigenetic profile at four additional genes implicated in lipoprotein metabolism: ATP-binding cassette transporter G1 (ABCG1; 21q22.3), hepatic lipase (LIPC; 15q21.3), phospholipid transfer protein (PLTP; 20q13.12), and scavenger receptor B1 (SCARB1; 12q24.31). ABCG1 encodes a transmembrane regulator of cholesterol and phospholipid transport, which mediates cholesterol efflux from human macrophages to mature HDL particles. ABCG1 might then be involved in foam cell and atherosclerotic plaque formation.¹⁸ Hepatic lipase (HL) is a protein anchored on the endothelial surface specific to the liver where it catalyzes the hydrolysis of lipoproteins’ TG content.¹⁹ HL plays an important role in the regulation of plasma lipid concentrations, considering that LIPC polymorphisms were previously associated with dyslipidemia and premature atherosclerosis.^19-21 PLTP is secreted in circulation where it mediates the transfer of phospholipids from very low-density lipoprotein (VLDL) and LDL to HDL particles.²² Scavenger receptor B1 (SRBI) is a plasma membrane receptor that catalyzes the bidirectional transfer of cholesterol between most of the peripheral cells and HDL particles. SRBI plays a fundamental role in the reverse cholesterol transport, since it is the major molecular pathway mediating selective cholesterol uptake by the liver. Moreover, common polymorphisms within ABCG1, LIPC, PLTP, and SCARB1 genes have been repeatedly and strongly associated with plasma lipid concentration variability.^5,6,23

The aim of the current study was thus to assess whether DNA methylation at key genes involved in lipoprotein metabolism is associated with plasma lipid concentrations and previous history of CAD in FH, a human dyslipidemia model thoroughly studied in CVD research. We focused our analysis primarily on the ABCG1, LIPC, PLTP, and SCARB1 genes. We also examined whether epipolymorphisms within these genes and others that our group has already studied (ABCA1, CETP, and LPL)^16,17 might independently contribute to HDL-C, LDL-C, and TG concentration variability in FH.

Results

DNA methylation analyses and CpG selection in the subsample of 28 FH men

DNA methylation variation at the ABCG1, LIPC, PLTP and SCARB1 gene loci (Fig. 1) was first assessed in a subsample of finely phenotyped men selected from our FH cohort (n = 28). This allowed us to identify the loci showing DNA methylation variability and that were the most promising to find associations with plasma lipids. According to these initial results, ABCG1-CpGC3, LIPC-CpGA2, and mean PLTP-CpGC were retained to be analyzed in the complete data set (Table S1). None of these CpGs was found significantly associated with leucocyte cellular heterogeneity (Table S2). DNA methylation levels at the SCARB1 gene locus were not associated with plasma lipid profile and either showed very low interindividual variability or were associated with blood leucocyte heterogeneity (Tables S1 and S2). Based on all these subsequent observations, analyses of the association between plasma lipid profile and epipolymorphisms within the ABCG1, LIPC and PLTP genes (respectively ABCG1-CpGC3, LIPC-CpGA2, and mean PLTP-CpGC) were undertaken in the complete FH sample.

Figure 1.ABCG1, LIPC, PLTP and SCARB1 gene loci. Three different regions (named from [A‒C]) were epigenotyped at the ABCG1, LIPC and PLTP gene loci and four (named from [A‒D]) at the SCARB1 gene locus. These epigenotyped regions are located at the gene promoter region and near or in a CpG island (represented by thin dashed rectangles). Overall 16 CpG dinucleotides were epigenotyped at the ABCG1 gene locus, 11 at the LIPC gene locus, 9 at the PLTP gene promoter locus and 20 at the SCARB1 gene locus.

Associations between ABCG1, LIPC and PLTP DNA methylation and plasma lipid levels

Table 1 shows the characteristics of subjects from the full sample according to sex. Overall, men and women showed differences in anthropometric parameters and plasma lipid profile. Men (n = 61) were older, had lower HDL-C and LDL-C levels and had a larger waist circumference as compared with women (n = 37). Moreover, we observed significant sex differences in DNA methylation for ABCG1 and PLTP (P < 0.05). Therefore, the analyses between plasma lipid profile and epipolymorphisms were conducted in men and women separately. In men, a higher ABCG1-CpGC3 DNA methylation was associated with age (r = 0.39; P = 0.01), a larger waist circumference (r = 0.34; P = 0.008) and lower total cholesterol (r = -0.46; P < 0.001) and LDL-C concentrations (r = -0.43; P = 0.001). In women, a higher ABCG1-CpGC3 DNA methylation was associated with lower TG concentrations (r = 0.39; P = 0.02). LIPC-CpGA2 DNA methylation was negatively correlated with HDL-C (r = -0.28; P = 0.03) and TG concentrations (r = -0.27; P = 0.04) in men, whereas a trend for a positive association was observed with TG concentrations in women (r = 0.28; P = 0.10). Lower mean PLTP-CpGC DNA methylation levels were associated with higher HDL-C concentrations in men (r = -0.26; P = 0.05) (Fig. 2; Table S3). Results remained unchanged after adjustment for these potential confounding factors: age, waist circumference, HDL-C and TG levels.

Table 1. Characteristics of all subjects according to sex (n = 98)

Characteristics	Men (n = 61) mean ± s.e.	Women (n = 37) mean ± s.e.	P value
Age (years)	46.3 ± 1.7	36.9 ± 2.0	0.001
CAD (%)	32.8	18.9	0.14
Waist circumference (cm)	95.1 ± 1.6	78.2 ± 2.0	< 0.001
BMI (kg/m^2)	27.3 ± 0.5	24.7 ± 0.8	0.008
Total cholesterol (mmol/L)	7.90 ± 0.32	9.16 ± 0.25	0.002
LDL-C (mmol/L)	6.07 ± 0.30	7.07 ± 0.22	0.009
HDL-C (mmol/L)	1.00 ± 0.03	1.22 ± 0.07	0.002
TG (mmol/L)^a	1.50 ± 0.06	1.64 ± 0.06	0.44
Systolic blood pressure (mm de Hg)	126.4 ± 2.1	113.2 ± 2.9	< 0.001
Fasting glucose (mmol/L)	5.21 ± 0.07	4.92 ± 0.14	0.07
DNA methylation level (%)
ABCG1-CpGC3	60.8 ± 0.6	58.3 ± 0.6	0.007
LIPC-CpGA2	84.8 ± 0.3	85.0 ± 0.3	0.51
PLTP-CpGC1	23.0 ± 0.4	23.0 ± 0.4	0.99
PLTP-CpGC2	46.8 ± 0.6	48.4 ± 0.4	0.025
Mean PLTP-CpGC	34.9 ± 0.5	35.7 ± 0.4	0.19

ABCG1, ATP-binding cassette transporter G1; BMI, body mass index; CAD, previous personal history of coronary artery disease; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; LIPC, hepatic lipase; PLTP, phospholipid transfer protein; TG, triglyceride. Statistically significant results are shown in bold. ^aGeometric mean and P value obtained after log10-transformation of the variable.

Figure 2. Partial Pearson’s correlation between DNA methylation levels and plasma lipid profile in FH men and women. (A) LIPC-CpGA2 and HDL-C levels in men (n = 61). (B) Mean PLTP-CpGC and HDL-C levels in men (n = 61). (C) ABCG1-CpGC3 and LDL-C levels in men (n = 61). (D) LIPC-CpGA2 and fasting triglyceridemia in men (n = 61). (E) ABCG1-CpGC3 and fasting triglyceridemia in women (n = 37); and F) LIPC-CpGA2 and fasting triglyceridemia in women (n = 37). All correlation coefficients (r) and P values showed were adjusted for age, waist circumference, and fasting triglyceridemia or HDL-C levels. *Correlation coefficients (r) and P values reported for the associations with fasting triglyceridemia were obtained after log₁₀-transformation.

Next, we investigated which factors among traditional and epigenetic predictors contribute to plasma lipid concentration variability. As expected, traditional risk factors (age, waist circumference, blood pressure, fasting glucose and plasma lipid levels) contributed significantly to the variability in plasma lipid concentrations (Table 2). More interestingly, Table 2 shows that ABCG1, LIPC and PLTP DNA methylation levels contribute independently to plasma lipid level variability in men and women. When previously reported epipolymorphisms found to be associated with plasma lipid levels were added into the statistical models (mean ABCA1-CpGA, CETP-CpGA2, CETP-CpGB2 and LPL-CpGA3 DNA methylation levels), the amount of variance explained was further improved for all lipids tested (R²_ajd were then between 16.8% and 49.9%), except for TG levels in women, which remained unchanged at 41.8% (Table 3). The independent association between epipolymorphisms and plasma lipid levels was confirmed by the analysis of the residual scores. Table 4 shows that epipolymorphisms contribute significantly to the variability of residual scores of plasma lipid levels corrected for traditional predictors. Indeed, epipolymorphisms within key genes implicated in lipoprotein metabolism independently explained up to 15.8% of the HDL-C, 24.0% of the LDL-C and 17.4% of the TG concentration variability in FH subjects (Table 4).

Table 2. Multivariable linear regression analysis of traditional and epigenetic predictors (ABCG1, LIPC and PLTP DNA methylation) of HDL-C, LDL-C and TG levels in FH men and women

	Men (n = 61)			Women (n = 37)
Predictors	HDL-C	LDL-C	TG^a	HDL-C	LDL-C	TG^a
Age (years)	NS	NS	NS	NS	NS	NS
Waist circumference (cm)	β = -0.19 P = 0.09	NS	β = 0.21 P = 0.06	NS	NS	β = 0.48 P < 0.001
Systolic blood pressure (mm de Hg)	NS	NS	NS	NS	β = 0.28 P = 0.09	NS
Glycemia (mmol/L)	β = 0.24 P = 0.02	NS	NS	NS	NS	NS
HDL-C (mmol/L)	-	β = -0.34 P = 0.002	β = -0.52 P < 0.001	-	β = -0.27 P = 0.10	β = -0.25 P = 0.06
LDL-C(mmol/L)	β = -0.30 P = 0.006	-	NS	NS	-	NS
TG (mmol/L)^a	β = -0.40 P = 0.001	NS	-	β = -0.37 P = 0.02	NS	-
ABCG1-CpGC3 DNA methylation (%)	NS	β = -0.49 P < 0.001	NS	NS	NS	β = 0.31 P = 0.02
LIPC-CpGA2 DNA methylation (%)	β = -0.22 P = 0.03	NS	β = -0.23 P = 0.04	NS	NS	β = 0.21 P = 0.10
Mean PLTP-CpGC DNA methylation (%)	β = -0.27 P = 0.01	NS	NS	NS	NS	NS
R^2 (%)	49.5	35.3	33.6	13.9	15.6	48.3
Adjusted R^2 (%)	43.8 P < 0.001	33.0 P < 0.001	30.1 P < 0.001	11.4 P = 0.02	10.7 P = 0.05	41.8 P < 0.001

ABCG1, ATP-binding cassette transporter G1; CETP; cholesteryl ester transfer protein; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; LIPC, hepatic lipase; LPL, lipoprotein lipase; PLTP, phospholipid transfer protein; TG, triglyceride. ^aβ coefficients and p-values obtained after log₁₀-transformation of the variable. NS = P > 0.10; only variables with a P ≤ 0.10 were included in the regression model.

Table 3. Multivariable linear regression analysis of epigenetic and traditional predictors of HDL-C, LDL-C and TG levels in FH men and women

	Men (n = 61)			Women (n = 37)
Predictors	HDL-C	LDL-C	TG^a	HDL-C	LDL-C	TG^a
Age (years)	NS	NS	NS	NS	NS	NS
Waist circumference (cm)	β = -0.24 P = 0.02	NS	β = 0.19 P = 0.08	NS	NS	β = 0.48 P < 0.001
Systolic blood pressure (mm de Hg)	NS	NS	NS	NS	β = 0.26 P = 0.09	NS
Glycemia (mmol/L)	β = 0.23 P = 0.02	NS	NS	NS	NS	NS
HDL-C (mmol/L)	-	β = -0.35 P = 0.001	β = -0.49 P < 0.001	-	NS	β = -0.25 P = 0.06
LDL-C(mmol/L)	β = -0.30 P = 0.004	-	NS	NS	-	NS
TG (mmol/L)^a	β = -0.33 P = 0.003	NS	-	β = -0.35 P = 0.03	NS	-
ABCG1-CpGC3 DNA methylation (%)	NS	β = -0.40 P < 0.001	NS	NS	β = -0.26 P = 0.10	β = 0.31 P = 0.02
LIPC-CpGA2 DNA methylation (%)	β = -0.17 P = 0.10	NS	β = -0.25 P = 0.03	NS	NS	β = 0.21 P = 0.10
Mean PLTP-CpGC DNA methylation (%)	β = -0.22 P = 0.03	NS	NS	NS	NS	NS
CETP-CpGA2 DNA methylation (%)	β = -0.28 P = 0.009	NS	NS	NS	β = -0.34 P = 0.03	NS
CETP-CpGB2 DNA methylation (%)	NS	β = -0.26 P = 0.02	NS	NS	NS	NS
LPL-CpGA3 DNA methylation (%)	NS	NS	β = -0.20 P = 0.07	β = 0.28 P = 0.08	NS	NS
R^2 (%)	55.8	41.2	37.5	21.4	23.8	48.3
Adjusted R^2 (%)	49.9 P < 0.001	38.2 P < 0.001	33.0 P < 0.001	16.8 P = 0.02	16.9 P = 0.02	41.8 P < 0.001

ABCG1, ATP-binding cassette transporter G1; CETP; cholesteryl ester transfer protein; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; LIPC, hepatic lipase; LPL, lipoprotein lipase; PLTP, phospholipid transfer protein; TG, triglyceride. ^aβ coefficients and P values obtained after log₁₀-transformation of the variable. NS, P > 0.10; only variables with a P ≤ 0.10 were included in the regression model.

Table 4. Multivariable linear regression analyses of epigenetic predictors of residual scores of HDL-C, LDL-C and triglyceride (TG) levels in FH men and women

	Men (n = 61)			Women (n = 37)
Predictors	Residual HDL-C^a	Residual LDL-C^b	Residual TG^c	Residual HDL-C^d	Residual LDL-C^e	Residual TG^c
ABCG1 DNA methylation (%)	NS	β = -0.34 P = 0.007	NS	NS	β = -0.29 P = 0.08	β = 0.37 P = 0.02
CETP DNA methylation (%)	β = -0.30 P = 0.009	β = -0.29 P = 0.02	NS	NS	β = -0.31 P = 0.05	NS
LIPC DNA methylation (%)	NS	NS	β = -0.29 P = 0.02	NS	NS	β = 0.25 P = 0.10
LPL DNA methylation (%)	NS	NS	β = -0.24 P = 0.06	β = 0.29 P = 0.08	NS	NS
PLTP DNA methylation (%)	β = -0.24 P = 0.06	NS	NS	NS	NS	NS
R^2 (%)	18.7	26.5	12.4	8.7	15.3	22.0
Adjusted R^2 (%)	15.8 P = 0.003	24.0 P < 0.001	9.3 P = 0.02	6.1 P = 0.08	10.3 P = 0.06	17.4 P = 0.02

ABCG1, ATP-binding cassette transporter G1; CETP; cholesteryl ester transfer protein; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; LIPC, hepatic lipase; LPL, lipoprotein lipase; PLTP, phospholipid transfer protein; TG, triglyceride. NS = P > 0.10; only variables with a P ≤ 0.10 were included in the regression model. Residual scores of plasma lipid levels were obtained by using unstandardized analysis of residuals computed by linear regressions which included traditional risk factors that contribute significantly to the plasma lipid level: ^aAge, waist circumference, TG and fasting glucose level; ^bAge and HDL level; ^cWaist circumference and HDL level; ^dTG level; ^eSystolic blood pressure and HDL level.

Association between LIPC DNA methylation and previous history of CAD

Afterward, we examined whether the ABCG1, LIPC and PLTP epigenetic profile might be associated with a previous history of CAD. Table 5 shows subjects’ characteristics according to CAD status in the two groups matched for age, gender, BMI and total cholesterol concentrations. FH patients with a prior history of CAD had higher LIPC-CpGA2 DNA methylation levels compared with those without CAD (85.2% vs. 83.8%; P = 0.02). This association remained significant even after consideration for age, sex and plasma lipid profile (P = 0.04). No significant difference was observed between a prior history of CAD, ABCG1-CpGC3 and mean PLTP-CpGC DNA methylation.

Table 5. Characteristics of subjects according to CAD status (n = 44)

Characteristics	Non-CAD (n = 22) mean ± s.e.	CAD (n = 22) mean ± s.e.	P value
Age (years)	49.1 ± 2.6	50.0 ± 2.5	0.81
Sex (M/W)	19/3	19/3	1
Waist circumference (cm)	90.2 ± 2.5	93.4 ± 2.6	0.38
BMI (kg/m^2)	26.1 ± 0.7	26.9 ± 0.7	0.44
Total cholesterol (mmol/L)	8.61 ± 0.61	7.91 ± 0.53	0.40
LDL-C (mmol/L)	6.79 ± 0.59	6.15 ± 0.51	0.41
HDL-C (mmol/L)	1.14 ± 0.07	1.00 ± 0.05	0.10
TG (mmol/L)^a	1.41 ± 0.15	1.58 ± 0.18	0.44
Systolic blood pressure (mm de Hg)	127.4 ± 4.0	126.8 ± 3.1	0.92
Fasting glucose (mmol/L)	5.28 ± 0.12	5.18 ± 0.14	0.59
DNA methylation level (%)
ABCG1-CpGC3	60.3 ± 0.9	60.5 ± 1.2	0.91
LIPC-CpGA2	83.8 ± 0.4	85.2 ± 0.4	0.02
Mean PLTP-CpGC	34.9 ± 0.7	35.0 ± 0.8	0.94

ABCG1, ATP-binding cassette transporter G1; BMI, body mass index; CAD, coronary artery disease; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; LIPC, hepatic lipase; PLTP, phospholipid transfer protein; TG, triglyceride. The two groups were matched for age, gender, BMI and total cholesterol concentrations. Statistically significant results are shown in bold. ^aGeometric mean and P value obtained after log10-transformation of the variable.

Association between DNA methylation and mRNA levels measured for ABCG1 and PLTP genes

Total leucocyte RNA was available for the relative quantification of ABCG1 and PLTP mRNA levels in the subsample of 28 FH men in order to evaluate the potential functional impact of the ABCG1 and PLTP DNA methylation variability. Interestingly, higher ABCG1-CpGC3 DNA methylation levels were associated with lower ABCG1 mRNA levels in leucocytes (r = -0.45; P = 0.02) (Table S1). No significant association was observed between mean PLTP-CpGC DNA methylation and leucocyte PLTP mRNA levels. However, DNA methylation levels at other CpG dinucleotides within the PLTP gene locus (PLTP-CpGA2 and PLTP-CpGB4) were modestly correlated with PLTP mRNA levels (Table S1). Potential binding sites for transcription factors were identified proximal to the epigenotyped regions. Specifically, binding sites for SP2/EKLF (erythroid transcription factor), upstream stimulatory factor (USF) and STAT5 transcription factors were observed respectively close to or within the ABCG1-CpGC3, LIPC-CpGA2, and PLTP-CpGC2 (Fig. S1).

Discussion

To the best of our knowledge, this study is the first to report associations between ABCG1, PLTP, and LIPC DNA methylation levels in leucocytes and the plasma lipid profile. Moreover, we showed that the combination of these and other epipolymorphisms previously identified within key lipoprotein metabolism genes contribute independently of traditional risk factors to interindividual variations in HDL-C, LDL-C, and TG concentrations. Again, the current results support the hypothesis that epigenetic changes within lipid metabolism genes might account for plasma lipid profile variability and could therefore be a molecular mechanism for dyslipidemias.

DNA methylation was previously shown to be partially inherited. Epigenetic modifications might therefore contribute to partly explain the missing heritability observed for many complex traits, including dyslipidemias.⁷ Indeed, the mean DNA methylation heritability was estimated at 18% in twin studies, whereas it might be as high as 48% at some gene specific loci according to family studies.^15,24 The current results, as well as previous studies performed by our group^16,17,25 and others,^14,15,24 suggest that epipolymorphisms account for plasma lipid level interindividual variability and could thus be a key molecular mechanism to explain the missing heritability of dyslipidemias and CVD.

The transmembrane transporter ABCG1 is a main player in cholesterol efflux from macrophage to mature HDL. Polymorphisms within the ABCG1 gene have been previously associated with HDL-C levels and CAD risk.^23,26,27 Our results suggest that a higher ABCG1 DNA methylation, associated with gene expression repression in leucocytes (potentially through transcriptional activators EKLF and/or SP2 binding disruption), is associated with lower LDL-C in men and higher TG levels in women, whereas no conclusive association was observed with HDL-C levels. Interestingly, prior GWAS and candidate gene studies reported that ABCG1 polymorphisms are associated with TG and total cholesterol levels.^6,28 Moreover, a previous study by Olivier et al. suggests that ABCG1 function is required for optimal bioavailability and activity of LPL produced by human macrophages.²⁹ Briefly, they showed that the silencing of ABCG1 expression in macrophages led to the retention of LPL at the cell surface, thus inhibiting its activity in TG-rich lipoprotein metabolism, and consequently decreasing the accumulation of lipids in macrophages. Indeed, an increase in macrophage LPL activity has previously been associated with an acute lipid accumulation in macrophages, which promotes the formation of macrophage foam cells and increases the risk for atherosclerosis.^30,31 These observations combined with ours suggest that ABCG1 might not only be involved in reverse cholesterol transport, but also in the metabolism of TG-rich lipoprotein particles by macrophages and the formation of foam cells. Further studies will be needed to confirm the possible impact of epigenetic modifications in the ABCG-mediated lipid accumulation in macrophages.

PLTP is essential for the transfer of excess surface lipids from TG-rich lipoproteins to HDL particles. PLTP-mediated phospholipid transport among HDL particles is also known to be associated with HDL particle size and lipid composition.³² Moreover, PLTP polymorphisms have been previously associated with HDL particle phenotype and TG levels.^5,6,23,33 Concordant with this, we observed that a higher PLTP DNA methylation was associated with smaller HDL particles, as well as lower concentrations of HDL-phospholipid and HDL-C in men. Therefore, our observations are also consistent with the known role of DNA methylation in gene expression repression. These results strongly suggest that PLTP-mediated lipid transport is compromised when PLTP DNA methylation increases.

HL hydrolyzes TG and phospholipids in all lipoproteins, but is predominantly implicated in the removal of TG from VLDL, intermediate density lipoprotein and postprandial TG-rich HDL particles.²¹ Indeed, several genome-wide and gene candidate studies have shown that polymorphisms in the HL gene (LIPC) are associated with fasting TG levels and HDL particle phenotype.^5,6,33-35 Accordingly, we report here that a higher LIPC DNA methylation is associated with lower HDL particle size, HDL-phospholipid, HDL-C and fasting TG levels in men, whereas a higher LIPC DNA methylation in women is associated with higher fasting TG levels. Interestingly, we also showed that LIPC DNA methylation was slightly higher in CAD patients compared with patients without a previous history of CAD. Previous studies showed that even a small DNA methylation change might induce substantial differences in gene-expression levels over time and predispose to late-onset disease such as CVD.^36,37 The role of HL in CAD and other CVD is still controversial. For instance, a previous study observed a significant association between LIPC polymorphisms and CVD,^38-40 while others did not.^41-43 Moreover, both increased and decreased HL activity have been associated with CAD.²¹ However, it was suggested that in patients with isolated hypercholesterolemia, such as FH subjects, an increase in HL activity might be favorable as evidenced by its negative association with CAD risk, possibly because HL increases the reverse cholesterol transport in this specific population.²¹ Based on our results, a higher LIPC DNA methylation might be associated with an increased CAD risk in FH. This observation is concordant with the expected functional impact of DNA methylation on gene repression. However, this hypothesis will have to be assessed in further studies, as the LIPC gene is not expressed in leucocytes.⁴⁴ Interestingly, we observed that DNA methylation variability at the LIPC-CpGA2 is located within a potential binging site for the transcription factor USF (previously reported as the -307/-312 E-box). Prior studies made by Van Deursen et al. showed that LIPC gene promoter activity is increased by USF and that this upregulation is mainly mediated through the -307/-312 E-box.^45,46 Moreover, a recent study showed that DNA methylation of the USF binding site abolishes the liaison to DNA.⁴⁷ We therefore suggest that DNA methylation at this specific locus might interfere with the recruitment of USF to the LIPC gene promoter and decrease gene expression.

In order to decrease the impact of possible confounding factors such as genetic and environmental heterogeneity, the recruitment of FH subjects was restricted to those carrying the low-density lipoprotein receptor (LDLR) p.W66R mutation, which strengthens our study. Phenotyping of all subjects performed by well-trained research health professionals and experienced lipidologists, analysis of DNA methylation at most key lipoprotein metabolism genes, as well as the use of a robust and reproducible technology for the DNA methylation quantification (i.e., pyrosequencing of bisulfite treated DNA), are among the other strengths of this study. One limitation of this study relates to the impossibility to infer the causal relationship observed between DNA methylation and plasma lipid profile. DNA methylation could both regulate blood lipid concentrations and/or be a marker of plasma lipid profile variability, considering that a patient’s methylome can also be affected by lipoprotein themselves.^48,49 Moreover, previous studies advised that the interpretation of whole blood DNA methylation should be performed with great caution since differentially methylated CpG might result from the heterogeneity in leucocyte cell types.⁵⁰ In order to avoid the potential impact of cellular heterogeneity on data interpretation, we only focused on CpG dinucleotides showing no association with cell blood count. Finally, these results will have to be validated in other FH and non-FH populations before strong conclusions can be drawn.

To conclude, we described for the first time that epipolymorphisms at the ABCG1, LIPC and PLTP gene loci are associated with variations in plasma lipid concentrations, and that hypermethylation of the LIPC gene is associated with previous history of CAD in FH. Moreover, we showed that DNA methylation at five key lipoprotein metabolism gene loci (ABCG1, CETP, LIPC, LPL, and PLTP) contributed independently of well-known traditional predictors to the plasma lipid level variability in a sex-specific manner in FH. Again, the identification of these epipolymorphisms raises very exciting new hypotheses in lipidology and cardiology research. Further larger epigenetic studies integrating genetics and transcriptomics are needed and will contribute to a better understanding of the molecular bases of dyslipidemia and the pathophysiological processes associated with the development of CAD and other cardiovascular phenotypes.

Materials and Methods

Sample and clinical data

This study included 98 FH patients (61 men and 37 women) from the Saguenay-Lac-Saint-Jean (SLSJ) region. All participants were French-Canadians carrying the lowLDLR p.W66R gene founder mutation (rs121908025). These participants were screened and genotyped at the Chicoutimi Hospital Lipid Clinic or the ECOGENE-21 Clinical Research Center between August 1994 and June 2010. All subjects gave their informed consent to participate in studies on genetic determinants of CAD combining genome-wide scans and candidate gene strategies and were assigned a code that systematically de-identifies all clinical data.⁵¹ This project received the approval of the Chicoutimi Hospital Ethics Committee in accordance with the Helsinki Declaration.

A registered research nurse and a physician specialized in lipidology performed a complete anthropometric and metabolic profiling of subjects. The acquisition process of all clinical and biochemical characteristics has been described previously.^16,17,25 FH subjects were all normoglycemic, naïve of treatment and with no previous history of diabetes, uncontrolled thyroid, inflammatory and pituitary diseases.

Blood samples were obtained after a 12 h overnight fast from the antecubical vein into vacutainer tubes containing EDTA. Plasma cholesterol and TG concentrations were enzymatically measured on a CX7 Analyzer (Beckman). Cholesterol levels were measured in plasma and HDL particles contained in the infranatant after precipitation of VLDL and LDL particles (d < 1.006 g/ml) with heparin (2300 unit/ml) and MnCl2 (1 M). Then, plasma LDL-C levels were estimated using the Friedewald formula.⁵²

Because no blood sample was available for the study’s needs (fine phenotyping, blood cell counts and mRNA measurements), 28 FH men from our initial sample were selected for an additional clinical evaluation, as previously described.¹⁷ Briefly, only male subjects were recruited to decrease the confounding effects of sexual hormones. Additional exclusion criteria were also applied: weight instability (± 2 kg) 3 mo prior to the study, severe hypertriglyceridemia (TG > 10 mmol/L), alcohol consumption > 2 drinks/day and the existence of diabetes, uncontrolled thyroid, inflammatory and pituitary diseases. In this subsample of 28 FH men, a fine phenotyping of HDL particles isolated by ultracentrifugation was performed (mean HDL particle size, HDL-TG, HDL-apolipoprotein A1 (apoA1) and HDL-phospholipid levels), as previously described.²⁵ The relative counts (%) of neutrophils, lymphocytes, monocytes, eosinophils and basophils were also determined on an impedance technology hematology counter (COULTER LH 780 hematology system; Beckman Coulter).

Nucleic acid extraction

DNA was purified from whole blood samples with the Gentra Puregen Blood Kit (Qiagen) as previously described.²⁵ Isolation and purification of intracellular RNA from whole blood stabilized in PAXgene Blood RNA Tubes (Qiagen) was completed with the PAXgene Blood RNA kit without DNase treatment (Qiagen). RNA samples were only available for the subsample of 28 FH men. RNA quality was assessed with the Agilent 2100 Bioanalyser RNA Nano Chips (Agilent Technologies). On average, RNA showed good quality (mean RNA integrity number = 8.3 ± 0.6).

DNA methylation measurement

The gold standard pyrosequencing technology is an accurate and quantitative sequencing assay. It was used to determine base-specific cytosine methylation levels at the ABCG1, LIPC, PLTP, and SCARB1genes. Pyrosequencing assays combine sodium bisulfite DNA conversion chemistry (EpiTech Bisulfite Kits; Qiagen), PCR amplification (PyroMark PCR Kit, Qiagen) and sequencing by synthesis assay (PyroMark Gold Q24 Reagents; Qiagen) of the target sequence. Sodium bisulfite preferentially deaminates unmethylated cytosines to thymines (after PCR amplification), whereas methyl-cytosines remain unmodified. Primers were selected using the PyroMark Assay Design software v.2.0.1.15. The PCR and pyrosequencing primers, as well as the primary and bisulfite-treated DNA sequences of both loci, are described in Table S4. DNA methylation levels were measured at genomic regions more likely to be associated with functional impacts on gene expression regulation, i.e., the gene promoter locus as well as regions near (CpGi shores) or within CpGi. Overall, 16 CpG dinucleotides were epigenotyped at the ABCG1 locus, 11 at the LIPC locus, 9 at the PLTP locus and 20 at the SCARB1 locus. The genomic context of the epigenotyped regions is shown in Figure 1.

Leucocyte DNA methylation at each locus was first analyzed in the subsample of 28 FH men in order to identify loci showing DNA methylation level variability and CpG dinucleotides potentially associated with plasma lipid concentrations. When DNA methylation levels were found to be generally well correlated among the CpG dinucleotides from a same region (r > 0.7), a mean DNA methylation level for these CpGs was computed and used in the subsequent statistical analysis. Otherwise, the CpG dinucleotides were analyzed separately. Overall, four criteria were applied to identify CpG dinucleotides to be further analyzed: (1) a mean DNA methylation level between 10 and 90%; (2) variability in DNA methylation levels (s_d > 3%); (3) no significant association between DNA methylation and the relative (%) leucocyte count (|r| < 0.38 and P > 0.05); and (4) at least one significant correlation between DNA methylation and plasma lipid levels (|r| > 0.39 and P < 0.05; after consideration for potential confounders, such as age and waist circumference). ABCG1-CpGC3, LIPC-CPGA2 and mean PLTP-CpGC met all four criteria and were therefore further analyzed in the whole sample of 98 FH subjects (Tables S1 and S2). DNA methylation levels at the ABCA1 (mean ABCA1-CpGA locus), CETP (CETP-CpGA2 and CETP-CpGB2) and LPL (LPL-CpGA3) gene promoter loci were previously determined in the whole sample and already met all four criteria.^16,17

Relative leucocyte mRNA level measurement

ABCG1, PLTP and SCARB1 leucocyte mRNA levels were measured in the subsample of 28 FH men. cDNA was generated from total RNA (250 ng) using random primers provided with the High Capacity cDNA Archive Kit (Life Technologies). Equals amounts of cDNA (2 µL) were run in duplicate and amplified in a 20 µL reaction containing 10 µL of 2× Universal PCR Master Mix (Life Technologies). Primers and TaqMan probes were obtained from Life Technologies (ABCG1: Hs00245154_m1; PLTP: Hs00272126_m1 and SCARB1: Hs00969821_m1). Each sample was calibrated to the GAPDH housekeeping gene (reference gene; GAPDH: Hs99999905_m1).^53,54 All TaqMan Gene expression assays were performed with probes covering an exon-exon junction thus avoiding the possible detection of genomic DNA. Relative quantification estimations were performed using a 7500Fast Real-Time PCR System as recommended by the manufacturer. The amplification efficiency of all qPCR reactions was determined with a series of serial dilution of a RNA sample pool. The slope computed on the qPCR standard curve was used to calculate the efficiency of the qPCR reactions (efficiency = -1+10^(-1/slope)). The amplification efficiency obtained for all qPCR reactions fell between 97% and 103%. ABCG1/GAPDH, PLTP/GAPDH and SCARB1/GAPDH Ct ratio (1/x) values were used in the analyses.

Identification of potential binding motifs of transcription factors

The MatInspector program from the Genomatix software was used to identify potential transcription factor binding motifs within the epigenotyped loci of the ABCG1, LIPC and PLTP genes.⁵⁵ The primary sequences of the ABCG1-CpGC, LIPC-CpGA and PLTP-CpGC loci were used as the input sequence to determine specific regions where DNA methylation might interfere with transcription factor binding and therefore influence its expression regulation. Only transcription factor binding sites with a matrix similarity score above 0.90 were considered.

Statistical analyses

The normal distribution of all variables was assessed using a Kolmogorov-Smirnov test. Fasting TG and HDL-TG levels did not follow normal distributions and were log₁₀-transformed. The association between DNA methylation, blood leucocyte cellular heterogeneity (relative cell count [%]), relative gene expression and plasma lipid profile was measured using a partial Pearson’s correlation controlling for potential confounding factors (age, waist circumference, and plasma lipid levels). Categorical variables were compared using a Pearson’s χ²-statistic, whereas group differences for continuous variables were compared using a Student t test and general linear models adjusted for potential confounding factors (age, sex, waist circumference, blood pressure and glycemia). Significant differences in DNA methylation according to CAD status were assessed in a subsample of FH subjects matched for age, gender, body mass index (BMI) and total cholesterol concentration (n = 44; 22 subjects with and 22 without previous history of CAD). Results were considered statistically significant when P-values were < 0.05 (two-sided).

Stepwise multivariable linear regression was used to identify predictors of HDL-C, LDL-C and TG levels in men and women among traditional predictors (age, waist circumference, systolic blood pressure, glycemia and plasma lipid profile) and the epipolymorphisms identified (mean ABCA1-CpGA, ABCG1-CpGC3, CETP-CpGA2, CETP-CpGB2, LIPC-CpGA2, LPL-CpGA3, mean PLTP-CpGC). Standardized coefficients (β coefficients) of the significant modulators are reported. Only variables with a P ≤ 0.10 were included in the regression model. The coefficient of determination (adjusted R²) is reported as the estimation of the variance explained by the statistical model. Multivariable regression results were confirmed using residual scores of HDL-C, LDL-C and TG levels corrected for traditional factors by using an unstandardized analysis of residuals computed by linear regression. These regression models included all traditional risk factors that contribute significantly to plasma lipid concentrations (age, waist circumference, systolic blood pressure, glycemia and/or plasma lipid profile). All statistical analyses were performed with the IBM SPSS Statistics 20 software (release 20.0.0, SPSS, Chicago, Il, USA).

Abbreviations:
ABCA1	=	ATP-binding cassette transporter A1
ABCG1	=	ATP-binding cassette transporter G1
apoA1	=	apolipoprotein A1
apoB	=	apolipoprotein B
AU	=	arbitrary unit
BMI	=	body mass index
CAD	=	coronary artery disease
CETP	=	cholesteryl ester transfer protein
CpGi	=	CpG island
CVD	=	cardiovascular disease
EKLF	=	erythroid transcription factor
FH	=	familial hypercholesterolemia
HDL-C	=	high-density lipoprotein cholesterol
HL	=	hepatic lipase
LDL-C	=	low-density lipoprotein cholesterol
LDLR	=	low-density lipoprotein receptor
LIPC	=	hepatic lipase gene
LPL	=	lipoprotein lipase
PL	=	phospholipids
PLTP	=	phospholipid transfer protein
SCARB1	=	Scavenger receptor B1 gene
SLSJ	=	Saguenay-Lac-Saint-Jean
SRBI	=	Scavenger receptor B1
TG	=	triglyceride
USF	=	upstream stimulatory factor
VLDL	=	very low-density lipoprotein cholesterol

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

Authors are thankful to all participants and the staff of the ECOGENE-21 Laboratory and Clinical Research Center. We particularly acknowledge the contribution of Sébastien Claveau (MSc), Nadia Mior, Denise Morin, Jeannine Landry (RN), Chantale Aubut (RN) and Céline Bélanger for their dedicated work. Simon-Pierre Guay was recipient of a doctoral research award from the Canadian Institutes for Health and Research (CIHR). During this study, Dr Daniel Gaudet held the Canada Research Chair in preventive genetics and community genomics. Luigi Bouchard is a junior research scholar from the Fonds de la recherche en santé du Québec (FRQS) and a member of the FRQS-funded Centre de recherche clinique Étienne-Le Bel. This project was supported by ECOGENE-21, the Canada Institutes for Health and Research (CIHR) team in community genetics (grant #CTP-82941), the Fondation des maladies du coeur du Québec, the Fonds de la recherche en santé du Québec (FRQ-S) and the Banting Research Foundation.

10.4161/epi.27981