The Microsomal Triglyceride Transfer Protein Gene-493T Variant Lowers Cholesterol But Increases the Risk of Coronary Heart Disease
Background— The microsomal triglyceride transfer protein (MTP) transfers lipids into apolipoprotein B-containing lipoproteins for secretion from liver, intestine, and heart. The T-variant of a functional polymorphism in the MTP promoter, MTP-493G/T, has been associated with reduced low-density lipoprotein cholesterol concentrations. We hypothesize that this polymorphism impacts on coronary heart disease (CHD) risk.
Methods and Results— The effect of the polymorphism was therefore tested in the West of Scotland Coronary Prevention Study biobank (580 cases and 1160 controls). MTP-493T carrier status was associated with significantly increased risk of CHD despite a small reduction in total cholesterol. Compared with the genotypic group with the lowest event rate (MTP-493GG, pravastatin treatment), the respective odds ratios (95% confidence interval) in the placebo group for CHD events were: GG, 1.23 (0.92 to 1.63); GT, 1.53 (1.12 to 2.08); and TT, 2.78 (1.53 to 5.05), suggestive of a gene-dose effect. The excess risk for CHD of the MTP-493T-variant was eliminated by pravastatin treatment. The Uppsala Longitudinal Study of Adult Men (ULSAM), which is a 20-year follow-up study of CHD, was used as an independent confirmatory database. These unexpected findings prompted the investigation of non-plasma lipid factors that could associate the MTP gene with CHD risk. In a limited number of subjects (n=18), heart muscle biopsies showed a MTP-493T genotype-specific depression of MTP mRNA expression.
Conclusion— The MTP-493T variant confers an increased risk of CHD that is unrelated to plasma lipids and lipoproteins, but eliminated by pravastatin treatment. A direct effect of the MTP polymorphism on myocardial lipid metabolism and vulnerability upon ischemic damage cannot be excluded.
Received July 17, 2004; de novo received October 14, 2004; revision received January 21, 2004; accepted February 6, 2004
Many studies of common polymorphisms in genes coding for proteins involved in lipid and lipoprotein metabolism have been reported, but few polymorphisms have shown a clinically significant effect on cardiovascular disease risk.
The microsomal triglyceride transfer protein (MTP) lipidates the growing apolipoprotein (apo) B polypeptide chain and thereby allows for the assembly and secretion process of lipoproteins.1,2 MTP is a heterodimer consisting of a unique 97-kDa subunit and the protein disulphide isomerase.3 The unique subunit confers the lipid transfer activity of the complex, whereas protein disulphide isomerase possesses the endoplasmic reticulum (ER) retention signal that is crucial for its intracellular localization.1 MTP is found in abundance on the luminal side of the ER and in the liver, intestine, and heart.4–6 Recent studies indicate that the actual concentration of MTP in the ER is a critical determinant of lipoprotein secretion.7 This seems to be true for the myocardium as well.8 Because the intracellular concentration of MTP is tightly controlled, any constitutive or induced alteration in MTP expression is likely to have an effect on secretion pattern of lipoproteins. Thus, functional genetic variants in the MTP gene may modulate MTP concentration or activity.
The MTP gene is polymorphic, with several genetic variants in linkage disequilibrium.9 The most studied promoter polymorphism is located 493 base pairs upstream from transcriptional start (MTP-493G/T), and the less-common T variant (allele frequency 0.25) has been associated with reduction of plasma LDL cholesterol levels.10 The genotype to LDL phenotype relationship has been confirmed in a larger cohort of similar ethnic background,9 but conflicting evidence has been presented from other cohorts. The Framingham investigators found no association between the MTP-493G/T polymorphism and lipid and lipoprotein concentrations,11 whereas the investigation of a large cohort of black men in the CARDIA study showed that MTP-493TT carrier status was associated with elevated apoB, LDL cholesterol, and plasma triglycerides.12 One possible explanation for this discrepancy could be that the phenotype of the MTP-493G/T polymorphism was strongly modulated by visceral obesity.13
Initial functional studies on the MTP-493G/T polymorphism, using minipromoter constructs, revealed increased promoter activity by the T-variant.10 Later functional studies on the MTP-164T/C polymorphism, which is in complete linkage disequilibrium with the MTP-493G/T polymorphism,9 revealed the opposite effect on transcriptional activity using similar reporter systems.14 In addition, there are several polymorphisms in the coding region,9,15 and their combined functional effects on a molecular level remain to be established.
The West of Scotland Coronary Prevention Study (WOSCOPS) was a 5-year, double-blind, randomized placebo-controlled trial of the efficacy of pravastatin for primary prevention of coronary heart disease (CHD).16 The biobank of the WOSCOPS consists of all of the 580 original cases and 1160 matched event-free control subjects. A second independent cohort (the Uppsala Longitudinal Study of Adult Men [ULSAM]17) was investigated for CHD event rate according to the MTP polymorphism. This is a prospective cohort study of men born 1920 to 1924 and living in Uppsala County, Sweden.
We hypothesized that the lowering of plasma LDL cholesterol seen in carriers of the MTP-493T variant would confer a reduced risk of CHD. The WOSCOPS and the ULSAM biobank cohorts were genotyped for the MTP polymorphism for this purpose, and the event rates in the various genotypic groups were compared. Because the results from both cohorts revealed a paradoxical increase in CHD risk in carriers of the MTP-493T variant, we also assessed a putative impact of the MTP polymorphism on MTP expression in human myocardial biopsies.
In the WOSCOPS, 6595 men (45 to 64 years of age) with a mean plasma cholesterol level of 7.0±0.6 mmol/L were randomly assigned to receive either pravastatin (40 mg/d) or placebo. The average follow-up period was 4.9 years. Medical records, electrocardiographic recordings, and the national death registry were used to determine the clinical end points.16 Each of the 580 men who had had a coronary event (nonfatal myocardial infarction, death from CHD, or a revascularization procedure) during the course of the study were matched for age (using 2-year age categories) and smoking status with 2 disease-free control subjects (n=1160) from the same cohort, as defined above.18 Subjects were categorized as either nonsmokers (those who had never smoked or had quit smoking) or smokers. In the present study, 295 case subjects had received pravastatin and 204 placebo, and 549 control subjects had received pravastatin and 568 placebo. All subjects provided written informed consent. The study was approved by the ethics committees of the University of Glasgow and the Karolinska Hospital, Stockholm, Sweden.
A second independent cohort (the ULSAM17) was investigated for CHD event rate according to the MTP polymorphism. This is a prospective cohort study of men born between 1920 and 1924 living in Uppsala County, Sweden. Risk factors for CHD were examined in 2322 men between 1970 and 1973, and a total of 103 cases with definitive myocardial infarction were recorded at the 20-year follow-up. All subjects gave informed consent, and the study was approved by the Uppsala University Ethics Committee.
An additional substudy was conducted to search for possible differential effects of the MTP-493G/T polymorphism on heart muscle mRNA expression. This small cohort of 19 patients undergoing coronary artery bypass grafting or valve replacement therapy has been described in detail previously.19 All subjects gave informed consent, and the study was approved by the Danish ethics committee system and the Karolinska Hospital Ethics Committee, Stockholm, Sweden.
Plasma total cholesterol, triglycerides, LDL cholesterol, and HDL cholesterol were measured twice during screening in the WOSCOPS, and the average of these measurements were used as the baseline level. Treatment effects were calculated by subtracting the baseline level from the treatment average. The personnel who performed the genotyping were blinded to the treatment status of the subjects.
Participants in the WOSCOPS underwent a baseline ECG recording to ensure that there was no evidence of overt CHD. This was based on significant ECG abnormalities, such as pathological Q waves (Minnesota Code 1) or marked ST- or T-wave abnormalities (Minnesota Code 4-1 and 5-1). Individuals with any of these codes were excluded from the study. Minor ECG abnormalities were defined as the presence of Minnesota codes 4-2, 4-3, 5-2, or 5-3. It should be noted that a code 4 cannot be reported in the absence of a code 5. All ECGs were processed centrally by a computer that automatically produced the Minnesota Codes.20
There are 7 common polymorphisms in the MTP gene.9 Analysis of haplotype structure of the MTP gene was evaluated to assess whether this had the potential of being more informative in relation to phenotypes. Data from a previously published study including 1064 alleles with complete information of the 7 polymorphisms were used to calculate the haplotype structure according to Stephens et al21 using the PhaseV2 program with a 1000 iteration setting.
Determination of the MTP-493G/T genotypes was performed by real-time sequencing using the Pyrosequencing equipment and protocol (Pyrosequencing AB, Uppsala, Sweden). DNA was amplified in a 50-μL reaction mix containing 100 ng of genomic DNA, 0.1 μmol/L of each primer, 3 mmol/L MgCl, 15 mmol/L Tris-HCl, pH 8.0, 50 mmol/L KCl, and 1 U AmpliTaqGold (Applied Biosystems). Downstream primers used were MTPNhe1 (5′-GCTAGCGCTGAT TTGCTCCAAC) and MTP493-U (5′-AGTTTCACACATAAGGACAATCATCTA). Upstream, biotin-labeled primer was MTP-4bio (5′-CCAGCTAGGAGTCACTGAGA). The amplification cycle started with activation of the AmpliTaqGold at 95°C for 7 minutes and was followed by denaturation at 94°C for 45 seconds. The cycling then proceeded according to the touch-down principle, denaturation at 94°C for 45 seconds, annealing at 64°C (7 cycles at this temperature), 60°C (7 cycles), 56°C (8 cycles), 54°C (8 cycles), 52°C (8 cycles), and 48°C (13 cycles) for 30 seconds, and elongation at 72°C for 1 minute. A final elongation at 72°C for 5 minutes ended the reaction. Forty microliters of the polymerase chain reaction (PCR) product were mixed with 15 μL streptavidin-coated magnetic beads (Dynabeads, Dynal A.S.) in a binding/washing buffer containing 5 mmol/L Tris-HCl, 1 mol/L NaCl, 0.5 mmol/L EDTA, and 0.05% Tween 20 (final concentrations in mix) at 65°C and 350 rpm for 30 minutes. The PCR fragments were denatured (and thus separated) by washing in 0.5 mol/L NaOH, and the magnetic beads with the immobilized biotinylated strands were washed in a buffer containing 20 mmol/L Tris-acetate and 5 mmol/L MgAc2. Sequencing primer MTP-493PSQ (5′-AACATTATTTTGAAGTGA TTGG) or MTP-493PSQ2 (5′-TATTTTGAAGTGATTGGT) was annealed to the fragments at 80°C for 2 minutes and then allowed to cool off to room temperature. The samples were then analyzed on the PSQ 96 Instrument (Pyrosequencing AB), and genotypes were analyzed using the PSQ SNP Software version 1.1 (Pyrosequencing AB).
Heart Muscle mRNA Quantification
Biopsies from the atrium of the heart were taken during coronary heart bypass grafting or valve replacement.19 The MTP mRNA content of each myocardial biopsy was determined in triplicate using real-time PCR, as described.19 Data are normalized for GAPDH mRNA content in the same specimen. Relevant biochemical determination of lipid content was not possible because of adipose tissue contamination of the atrial samples.
Allele frequencies were determined by gene counting. A χ2 test was used to compare the observed numbers of each genotype with those expected for a population in Hardy-Weinberg equilibrium. Distribution of continuous variables in the groups was expressed as mean±SD. Skewed variables were log transformed before statistical calculations. Concentrations of MTP mRNA were skewed, and statistical differences between groups were calculated using a nonparametric test (Mann-Whitney U test). ANOVA was used to determine any significant differences between genotypes. Multivariate conditional logistic-regression was used to assess the independent prognostic value of variables. Relative risk is presented as odds ratios (ORs) comparing genotype frequencies in patients versus controls with 95% confidence intervals (CIs).
Nine common haplotypes were detected (Table 1). The −493G/T, −164T/C, and I/T128 polymorphisms seemed to be in complete allelic association. To assess whether the predominant haplotype with the MTP-493T variant (haplotype 2) was more informative than the crude MTP-493T carrier status (essentially the sum of haplotypes 2, 4, 6, 7, and 9) in relation to the LDL phenotype, comparisons were made based on data from a previously published study.9 Homozygous carriers of haplotype 1 had LDL cholesterol concentrations of 3.55±0.86 mmol/L (n=190), whereas the concentration was reduced similarly between homozygous haplotype 2 carriers (3.11±0.51 mmol/L, n=9) and homozygous carriers of the MTP-493T variant (3.16±0.73 mmol/L, n=30). These results suggest that analysis of MTP haplotypes does not provide additional information over and above the MTP-493G/T polymorphism, and data presented are based on this polymorphism only.
Of the 580 cases and the 1160 control subjects in the WOSCOPS, MTP genotype was obtained in 498 cases and 1117 controls. In the remaining cases and controls, it was not obtained because of missing DNA (120 samples) or technical difficulties in determining genotype (5 samples). The distribution of genotypes between cases and controls is listed in Table 2. The MTP-493T variant had a frequency of 0.26 in the case group and 0.19 in the control group. The genotype frequencies showed Hardy-Weinberg equilibrium in both cases and controls.
In the ULSAM cohort, all 103 cases and 100 healthy control subjects were genotyped for the MTP-493G/T polymorphism. The MTP-493T variant showed a frequency of 0.27 in cases and 0.24 in controls. Genotype frequencies in the ULSAM cohort are shown in Table 2. Homozygosity for the MTP-493T variant seemed overrepresented among cases.
Association of MTP Genotype With Lipid and Lipoprotein Levels
Baseline total cholesterol in control subjects homozygous for the MTP-493T variant was slightly lower compared with carriers of the G, as follows: TT, 6.84±0.54; GT, 7.05±0.59; GG, 7.02±0.56 mmol/L (data previously published9). This genotype-phenotype association is in the same direction as previously reported, 9,10 albeit with a weaker genotype-phenotype effect. There was no significant association with HDL cholesterol or plasma triglycerides. The genotype-specific difference for total cholesterol was absent in the baseline sample of the case group.
Effect of Pravastatin Treatment According to MTP Genotype
The response to pravastatin treatment was not affected by MTP genotype in control subjects (Table 3). The lowering of total plasma and LDL cholesterol was 19% to 21% and 25% to 29%, respectively. There was a significant 6% to 8% increase of plasma HDL cholesterol. There was no significant difference in triglyceride-lowering effect of pravastatin between genotypes.
MTP Genotype and CHD Risk
In the WOSCOPS placebo-treated group, homozygous MTP-493T carriers had a significantly increased event rate compared with G carriers, as follows: ORGT, 1.24 (95% CI, 0.92 to 1.68); P=0.15; ORTT, 2.40 (95% CI, 1.32 to 4.38); P=0.004. However, this genotype-risk association was eliminated in the pravastatin-treated group, as follows: ORGT, 0.92 (95% CI, 0.66 to 1.30); P=0.65; ORTT, 0.70 (95% CI, 0.31 to 1.66); P=0.38. Using the pravastatin-treated group homozygous for the G variant as referent, the association between risk and MTP genotype became more prominent among the placebo-treated individuals, as follows: ORGG, 1.23; P=0.16; ORGT, 1.53; P=0.007; ORTT, 2.78, P<0.001 (Table 4). Of note, the heterozygous carriers were intermediate, suggestive of a gene-dose effect.
These genotype-risk associations were based on soft (definite myocardial infarction and revascularization) and hard (death from myocardial disease) end points combined. Restricting the analysis to hard end points only seemed to additionally increase the risk in carriers of the MTP-493T variant. Homozygous carriers of the MTP-493T variant in the placebo group showed an OR of 3.21(95% CI, 1.73 to 5.96; P<0.001) compared with pravastatin-treated homozygous carriers of the MTP-493G variant. Again, the MTP-493GT genotype was intermediate, with an odds ratio of 1.57 (95% CI, 1.13 to 2.17; P=0.007). Adjusting these ORs for other common CHD risk factors (age, systolic blood pressure, triglycerides, LDL and HDL levels, fibrinogen, white cell count, and C-reactive protein) additionally increased the OR to 4.04 (95% CI, 2.02 to 8.09; P<0.001) for individuals homozygous for the MTP-493T variant. Accordingly, the MTP genotype effect on cardiovascular risk seems to be independent of conventional and other cardiovascular risk factors.
The results were surprisingly strong and contradictory to the original hypothesis, and we sought confirmation in an independent cohort. In the ULSAM cohort, homozygous carriers of the MTP-493T variant show a markedly increased event rate in the case group, with an OR of 5.1 (95% CI, 1.4 to 18.3; P=0.005).
Association With Minor ECG Abnormalities at Baseline
Homozygosity for the MTP-493T variant showed a significant association with baseline ECG Minnesota code 4 and 5 abnormalities, with an OR of 1.68 (95% CI, 1.19 to 2.36; P=0.003) compared with homozygosity for the MTP-493G variant. Because this may be indicative of a genotype-specific subclinical disease present already at randomization, the results for CHD risk according to MTP genotype results were adjusted for such ECG abnormalities. This analysis did not alter the association between the MTP-493T variant and CHD risk. The OR for MTP-493TT adjusted for ECG abnormalities at baseline was 2.40 (95% CI, 1.31 to 4.38; P=0.005), and the corresponding OR for MTP-493GT was 1.23 (95% CI, 0.81 to 1.66; P=0.18), indicating that the ECG abnormalities were unrelated to preexisting disease and, instead, possibly showing an independent effect on myocardial function in relation to MTP genotype.
Association Between MTP Genotype and Heart Muscle MTP mRNA
The unexpected relationship between CHD risk and the MTP polymorphism prompted us to search for nonplasma lipid associations between heart disease and the MTP-493G/T genotype. It has been hypothesized that MTP expression in the heart relates to ischemic heart disease, and the relationship between the MTP promoter polymorphism and MTP expression in heart muscle was therefore investigated in a limited collection (n=19) of heart biopsies.19 There were 12 subjects homozygous for the MTP-493G variant, 5 heterozygotes, and 1 subject homozygous for the MTP-493T variant. One subject could not be genotyped because of missing DNA. The MTP mRNA content was significantly lower in the atria in carriers of the MTP-493T variant (n=6) compared with homozygous carriers of the G variant (n=12) (median [interquartile range], 17 arbitrary units [12 to 24] versus 63 arbitrary units [25 to 105]; P=0.049).
Elevated plasma concentration of LDL cholesterol is a well-established risk factor for CHD. The less-common T variant at position −493 in the promoter of the MTP gene has previously been associated with lowered plasma concentration of LDL cholesterol in healthy subjects.9,10 Paradoxically, the present study showed an increased CHD event rate among carriers of the MTP-493T variant despite the marginal genotype-specific lowering of total cholesterol. However, the increased CHD risk in MTP-493T carriers was completely eliminated by pravastatin treatment.
This study confirmed the cholesterol-lowering effect previously seen in individuals homozygous for the MTP-493T variant.9,10 The effect was seen despite the fact that subjects were included in the study on the basis of total cholesterol within a narrow range, a circumstance that is likely to limit the possibility to detect such genotype-phenotype relationships.
The underlying mechanism for the paradoxical finding in this study seems to be unrelated to plasma lipid and lipoprotein concentrations and remains to be resolved. It cannot be excluded that the MTP polymorphism is in linkage disequilibrium with a presently unidentified polymorphism strongly related to CHD risk, but if such a mechanism were to operate, it would have to be sensitive to pravastatin treatment.
MTP seems to take part in a recently described pathway for cardiac lipid metabolism, ie, secretion of apoB-containing lipoproteins from the heart. Absence of functional MTP in the heart leads to increased cardiac triglyceride stores in mice, which in turn is associated with heart failure and sudden cardiac death.8 In addition, Unger and Orci22 have proposed that overloading the myocardium with triglycerides causes lipotoxic heart disease. Nielsen et al19 have recently shown a direct relationship between MTP expression and triglyceride accumulation in human myocardial biopsies. Therefore, we speculate that dysfunctional lipid transport in the heart might increase the susceptibility of the myocardium to ischemic damage. The speculation could be supported by the observation that the MTP-493T variant was particularly strongly related to the hard end points. Such end points are likely to represent either extended myocardial lesions or a higher frequency of fatal arrhythmic events. In addition, the subtle MTP genotype-specific ECG changes could indicate a related myocardial dysfunction. The speculation is also supported by the differential effect on MTP mRNA concentrations by the MTP polymorphism in atrial tissue observed in the present study. Low myocardial MTP expression could attenuate an essential lipid export function from the heart.
The MTP-493G/T polymorphism is in complete allelic association with the −164T/C promoter polymorphism, which shows sequence homology to a sterol regulatory element (SRE) located in the same area (bp −171 to −164). The SRE binding protein, SREBP, downregulates the expression of MTP,23 and the C-variant of the MTP-164T/C polymorphism confers a higher degree of homology to the SRE compared with the common T-variant. An interaction with intracellular sterol synthesis induced by pravastatin could therefore possibly explain the protective effect of the drug in the present study. Additional studies are needed to examine these mechanisms.
A small contribution could arise also from the finding that the MTP-493G/T polymorphism induces subtle changes in lipid risk factors that were not quantified in this study. Lundahl et al24 have recently showed that individuals homozygous for the MTP-493T variant have a 2-fold increase in the chylomicron remnants in the postprandial state.
In summary, the T variant of the MTP-493G/T promoter polymorphism confers a plasma lipid-independent excess risk of CHD, which is eliminated by pravastatin treatment. We speculate that a genotype-specific MTP expression in the myocardium renders the heart particularly vulnerable to ischemic damage.
This study was supported by grants from the Swedish Heart-Lung Foundation, the Swedish Medical Research Council (12659), and the following foundations: Åke Wiberg, Fredrik and Ingrid Thuring, Professor Nanna Svartz, the Old Servants, and Karolinska Institutet. H. Ledmyr is a PhD student supported by the Swedish National Network for Cardiovascular Research. F. Karpe is a Wellcome Trust Senior Clinical Research Fellow.
Dr Packard has received research grants and honoraria from Bristol-Myers Squibb, sponsor of the WOSCOPS study.
Gordon DA, Jamil H, Sharp D, et al. Secretion of apolipoprotein B-containing lipoproteins from HeLa cells is dependent on expression of the microsomal triglyceride transfer protein and is regulated by lipid availability. Proc Natl Acad Sci U S A. 1994; 91: 7628–7632.
Leiper JM, Bayliss JD, Pease RJ, et al. Microsomal triglyceride transfer protein, the abetalipoproteinemia gene product, mediates the secretion of apolipoprotein B-containing lipoproteins from heterologous cells. J Biol Chem. 1994; 269: 21951–21954.
Wetterau JR, Combs KA, Spinner SN, et al. Protein disulfide isomerase is a component of the microsomal triglyceride transfer protein complex. J Biol Chem. 1990; 265: 9801–9807.
Nielsen LB, Veniant M, Boren J, et al. Genes for apolipoprotein B and microsomal triglyceride transfer protein are expressed in the heart: evidence that the heart has the capacity to synthesize and secrete lipoproteins. Circulation. 1998; 98: 13–16.
Leung GK, Veniant MM, Kim SK, et al. A deficiency of microsomal triglyceride transfer protein reduces apolipoprotein B secretion. J Biol Chem. 2000; 275: 7515–7520.
Bjorkegren J, Veniant M, Kim SK, et al. Lipoprotein secretion and triglyceride stores in the heart. J Biol Chem. 2001; 276: 38511–38517.
Ledmyr H, Karpe F, Lundahl B, et al. Variants of the microsomal triglyceride transfer protein gene are associated with plasma cholesterol levels and body mass index. J Lipid Res. 2002; 43: 51–58.
Karpe F, Lundahl B, Ehrenborg E, et al. A common functional polymorphism in the promoter region of the microsomal triglyceride transfer protein gene influences plasma LDL levels. Arterioscler Thromb Vasc Biol. 1998; 18: 756–761.
Juo SH, Han Z, Smith JD, et al. Common polymorphism in promoter of microsomal triglyceride transfer protein gene influences cholesterol, ApoB, and triglyceride levels in young African American men: results from the coronary artery risk development in young adults (CARDIA) study. Arterioscler Thromb Vasc Biol. 2000; 20: 1316–1322.
Ehrenborg E, Ledmyr H, Hamsten A, et al. Functional studies of a common promoter polymorphism in the microsomal triglyceride transfer protein (MTP) gene. Circulation. 1999; 100 (suppl II): II-1019. Abstract.
Nielsen LB, Perko M, Arendrup H, et al. Microsomal triglyceride transfer protein gene expression and triglyceride accumulation in hypoxic human hearts. Arterioscler Thromb Vasc Biol. 2002; 22: 1489–1494.
Sato R, Miyamoto W, Inoue J, et al. Sterol regulatory element-binding protein negatively regulates microsomal triglyceride transfer protein gene transcription. J Biol Chem. 1999; 274: 24714–24720.
Lundahl B, Hamsten A, Karpe F. Postprandial plasma ApoB-48 levels are influenced by a polymorphism in the promoter of the microsomal triglyceride transfer protein gene. Arterioscler Thromb Vasc Biol. 2002; 22: 289–293.