Polymorphisms in the gene encoding lipoprotein lipase in men with low HDL-C and coronary heart disease: the Veterans Affairs HDL Intervention Trial.

Our goal was to further define the role of LPL gene polymorphisms in coronary heart disease (CHD) risk. We determined the frequencies of three LPL polymorphisms (D9N, N291S, and S447X) in 899 men from the Veterans Affairs HDL Intervention Trial (VA-HIT), a study that examined the potential benefits of increasing HDL with gemfibrozil in men with established CHD and low high density lipoprotein cholesterol (HDL-C; < or =40 mg/dl), and compared them with those of men without CHD from the Framingham Offspring Study (FOS). In VA-HIT, genotype frequencies for LPL D9N, N291S, and S447X were 5.3, 4.5, and 13.0%, respectively. These values differed from those for men in FOS having an HDL-C of >40, who had corresponding values of 3.2% (P = 0.06), 1.5% (P < 0.01), and 18.2% (P < 0.01). On gemfibrozil, carriers of the LPL N9 allele in VA-HIT had lower levels of large LDL (-32%; P < 0.01) but higher levels of small, dense LDL (+59%; P < 0.003) than did noncarriers. Consequently, mean LDL particle diameter was smaller in LPL N9 carriers than in noncarriers (20.14 +/- 0.87 vs. 20.63 +/- 0.80 nm; P < 0.003). In men with low HDL-C and CHD: 1) the LPL N9 and S291 alleles are more frequent than in CHD-free men with normal HDL-C, whereas the X447 allele is less frequent, and 2) the LPL N9 allele is associated with the LDL subclass response to gemfibrozil.

S447X variant, which causes a deletion of the last two amino acids in the LPL protein, is associated with increased LPL activity (15).
The present study represents an extension of our earlier work with the Framingham Offspring Study (FOS) population. In FOS, the D9N and N291S alleles were associated with lower levels of high density lipoprotein cholesterol (HDL-C) and a trend toward increased triglycerides (16). In contrast, the S447X allele was associated with higher levels of HDL-C, decreased triglycerides, and reduced CHD risk in FOS males (17). To further explore the role of these LPL variants in CHD risk, we examined their associations with plasma lipid and lipoprotein levels, response to gemfibrozil therapy, and CHD end points in men participating in the Veterans Affairs HDL-C Intervention Trial (VA-HIT), a study designed to examine the potential benefits of increasing HDL-C in men with established CHD, who had low HDL-C as their primary lipid abnormality. Additionally, we have compared the genotypic data generated in VA-HIT with that generated in men without CHD from FOS, stratified according to HDL-C level.

Subjects
The rationale, design, and methods for VA-HIT have been described elsewhere in detail (18). Briefly, men were recruited at 20 Department of Veterans Affairs medical centers throughout the United States. Eligibility for the trial required a documented history of CHD, an age of Ͻ 74 years, an absence of coexisting conditions, an HDL-C level of р 40 mg/dl (1.0 mmol/l), an LDL-C of р 140 mg/dl (3.6 mmol/l), and a plasma triglyceride concentration of р 300 mg/dl (3.4 mmol/l). Information on age, alcohol consumption, smoking status, blood pressure, body mass index (BMI), and diabetes were available for all subjects enrolled in VA-HIT. However, informed consent for DNA analysis was only obtained from some of the subjects; thus, only these samples could be used in our genotyping analysis. Ninety-three percent of these subjects were white ( Table 1 ), with no differences noted in the race distribution among subjects within each genotype. Data used in our statistical analyses were obtained at baseline, with the exception of those used to examine the relationships between genotype and the plasma lipid response to gemfibrozil (1,200 mg/day), which were obtained at month 7.
The subjects included in the comparisons presented in Table  3 were men from FOS who were free of CHD, had an age of Ͼ 39 years, were not taking any medications known to influence plasma lipid levels, and had genotype information for at least two of the three LPL polymorphisms. FOS men were further stratified according to those having an HDL-C level of Ͻ 40 mg/dl (n ϭ 367) and those having an HDL-C level of у 40 mg/dl (n ϭ 577). Characteristics of the FOS subjects are as follows: 1 ) HDL-C у 40: age, 58 Ϯ 9 years; BMI, 27.8 Ϯ 4.1 kg/m 2 ; HDL-C, 52 Ϯ 11 mg/dl; LDL-C, 131 Ϯ 31 mg/dl; triglyceride, 110 Ϯ 61 mg/dl; 2 ) HDL-C Ͻ 40: age, 58 Ϯ 9 years; BMI, 29.4 Ϯ 4.5 kg/m 2 ; HDL-C, 33 Ϯ 5 mg/dl; LDL-C, 126 Ϯ 30 mg/dl; triglyceride, 195 Ϯ 297 mg/dl. The contrast between VA-HIT and FOS men with HDL-C у 40 provides a comparison of those with low HDL-C and CHD and those who have neither, whereas the contrast between VA-HIT and FOS men with HDL-C Ͻ 40 provides a comparison of those who have CHD versus those without CHD among men with low HDL-C.

Measurement of plasma lipid, lipoprotein, and apolipoprotein levels
Blood samples were collected from subjects after a 12-14 h fast into tubes containing 0.1% EDTA. Plasma was isolated and frozen for subsequent analysis of plasma lipid, lipoprotein, and apolipoprotein concentrations. Plasma total cholesterol (TC) and triglyceride concentrations were determined using enzymatic assays (19). Plasma HDL-C concentrations were measured after dextran sulfate-magnesium precipitation of apolipoprotein B (apoB)-containing lipoproteins (20), and HDL subfractions were separated by differential polyanion precipitation (21). LDL-C levels were calculated with the equation of Friedewald, Levy, and Fredrickson (22). ApoA-I and apoB levels in the plasma were measured with an immunoturbidimetric assay using reagents and calibrators from Incstar Corp. (Stillwater, MN) (23,24).

Analysis of lipoprotein subclass concentrations and particle size by NMR
The distribution of lipoprotein subclasses and particle size were determined by proton NMR spectroscopy, as previously described (25,26). The concentrations of six VLDL (V6-V1), three LDL (L3-L1), and five HDL (H5-H1) subclasses, listed from largest to smallest with respect to particle size (nm), are provided by this methodology. VLDL levels are expressed in units of triglyceride (mg/dl), whereas LDL and HDL subclass levels are expressed in units of cholesterol (mg/dl).

DNA analysis
Genomic DNA was extracted from whole blood samples using either QIAamp mini kits (Qiagen) or Generation Capture Column kits (Gentra Systems). Genotyping of the LPL N291S and S447X variants was carried out on an Applied Biosystems 7700 sequence detection system, using TaqMan probes for allelic discrimination. A detailed description of this methodology has previously been reported (27,28). For the N291S and S447X variants, PCR amplification was performed with the following primer pairs: 5 Ј -AGGGCTCTGCTTGAGTTGTAGAAA-3 Ј and 5 Ј -CTGAGAAC-GAGTCTTCAGGTACATTTT-3 Ј (N291S) and 5 Ј -AGGAAAGGC- ACCTGCGGTAT-3 Ј and 5 Ј -CAGGATGCCCAGTCAGCTTTA-3 Ј (S447X). For each genotype, PCR amplification was carried out in the presence of two probes, each of which had a different reporter dye attached to its 5 Ј end (FAM™ or VIC™) and a fluorescent quencher (TAMRA™) attached to its 3 Ј end. One probe was complementary to the wild-type DNA strand and the other to the DNA strand with the mutation. PCR was performed in a 10 l reaction volume, as previously described (27,28), using two steps of 50 Њ C for 2 min and 95 Њ C for 10 min, followed by 40 cycles of 95 Њ C for 15 s and 60 Њ C for 1 min. Genotyping of the D9N variant was performed manually, as previously described (29).

Statistical analyses
To evaluate the relationships between LPL genotype and plasma lipids and lipoproteins, we used analysis of covariance techniques, adjusting for the potential confounders of age, BMI, smoking status, hypertension, and diabetes. Chi-square tests were used for categorical measures, and either two-sample t -tests or Wilcoxon rank sums tests were used for continuous measures. SAS version 8 was used for all analyses. Analyses of time-to-event data were performed, and relative risks were calculated, using Cox proportional hazards models, with LPL genotype as the explanatory variable and adjustments made for the aforementioned confounders as well as treatment group.

Subject characteristics
To further investigate the frequency distribution and phenotypic effects of three common LPL variants, we analyzed 899 men in VA-HIT who had consented to DNA analysis. The demographic and biochemical characteristics of these subjects are presented in Table 1. This group of men had a mean age of 64 Ϯ 7 years and a mean BMI of 29 Ϯ 5 kg/m 2 , with normal levels of TC, LDL-C, and triglycerides. Because of the selection criteria of VA-HIT, the mean level of HDL-C in these men was low (32 Ϯ 5 mg/dl). This level resulted in a mean ratio of TC to HDL-C of 5.7, a value that is significantly higher than that defined as favorable (30).

Genotype distribution of LPL variants in VA-HIT
DNA analysis revealed that, among men in VA-HIT, 5.3% were carriers of the LPL N9 allele, 4.5% were carriers of the S291 allele, and 13.0% were carriers of the X447 allele ( Table 2). Comparatively, the meta-analysis of Wittrup, Tybjaerg-Hansen, and Nordestgaard (13) identified heterozygote frequencies of 2-4% for LPL D9N, 1-7% for LPL N291S, and 17-22% for LPL S447X. Based on these data, it appears that the LPL N9 allele occurs with increased frequency in the VA-HIT population, whereas the LPL X447 allele occurs with decreased frequency.

Comparison of LPL genotype distribution in VA-HIT versus FOS
To evaluate whether the LPL D9N, N291S, and S447X polymorphisms influence CHD risk via HDL, or possibly through other mechanisms, we compared the genotype distribution of each LPL variant in the VA-HIT population with age-matched men without CHD from FOS. In these analyses, men in FOS were classified as having an HDL-C level of Ͻ40 or у40 mg/dl. The contrast between VA-HIT and FOS men with HDL-C у 40 provided a comparison of those with low HDL-C and CHD and those who have neither, whereas the contrast between VA-HIT and FOS men with HDL-C of Ͻ40 provided a comparison of those who have CHD versus those without CHD among men with low HDL-C. As shown in Table 3, significant differences were observed between the VA-HIT and FOS populations in terms of LPL genotype distribution. The only statistically significant difference observed for the comparison of men in VA-HIT with men in FOS having low HDL-C was the increased frequency of the LPL D9N allele in VA-HIT (5.3% vs. 1.7%; P Ͻ 0.01). When men in VA-HIT were compared with men in FOS having an HDL-C level of у40, significant differences were seen in the percentage of carriers of the LPL N291S and S447X variants. Significantly more men in VA-HIT were carriers of the S291 allele (4.1% vs. 1.5%; P Ͻ 0.01), whereas significantly fewer were carriers of the X447 allele (13.0% vs. 18.2%; P Ͻ 0.01). Nearly identical differences were observed for the comparison of men in FOS with low HDL-C levels and men in FOS with normal HDL-C levels for both the LPL S291 (4.1% vs. 1.5%; P Ͻ 0.01) and X447 (11.8% vs. 18.2%; P Ͻ 0.01) alleles.

Association of LPL variants with plasma lipid and lipoprotein subclass concentrations at baseline
We have previously reported that male carriers of either the LPL N9 or S291 alleles in FOS had significantly lower levels of plasma HDL-C (Ϫ7.1%; P ϭ 0.03) than did noncarriers (16), whereas carriers of the LPL X447 allele had significantly higher levels of HDL-C (ϩ5.3%; P ϭ 0.008) and lower triglyceride levels (Ϫ14%; P ϭ 0.01) (17). In VA-HIT, we did not observe significant associations between baseline plasma lipid levels and LPL genotype (Table 4). Although HDL-C concentrations were lower in carriers of the D9N and N291S variants relative to noncarriers, neither of these differences was statistically significant (P ϭ 0.30 for both comparisons). As was the case in FOS, carriers of the LPL X447 allele in VA-HIT had relatively lower levels of triglycerides (153 Ϯ 66 vs. 164 Ϯ 69 mg/dl) and higher levels of HDL-C (32.7 Ϯ 6.3 vs. 31.7 Ϯ 5.1 mg/dl) than did noncarriers, but these differences were not statistically significant. Because VA-HIT consisted of a high number of men with diabetes and/or insulin resistance, we also examined our data for interactions between diabetes and insulin resistance and each LPL genotype in predicting plasma levels of HDL-C and triglycerides. Although no statistically significant (P Ͻ 0.05) interactions were observed, a few were close to statistical significance. Specifically, the P value for the interaction between LPL S447X and insulin resistance in predicting plasma HDL-C was 0.05, whereas that between LPL D9N and diabetes was 0.06.
To examine whether or not LPL genotypes were associated with lipoprotein subclass levels, we determined VLDL, LDL, and HDL subclass concentrations, as well as particle size, in a subgroup of men from VA-HIT using automated NMR spectroscopy. No statistically significant associations (P Ͻ 0.01) were observed between any of the LPL variants and concentrations of lipoprotein subclasses at baseline (data not shown).

Association of LPL variants with the plasma lipid and lipoprotein response to gemfibrozil
When on-trial concentrations of plasma lipids were compared in noncarriers versus carriers of either the LPL N9 or S291 allele, no statistically significant differences were observed (data not shown). This was likewise the case when data were expressed in terms of percentage change from baseline. In carriers of the LPL S447X variant, a relatively greater percentage reduction in plasma triglycerides versus placebo was observed in noncarriers compared with carriers (Ϫ27% vs. Ϫ20%; P Ͻ 0.04). However, this finding was most likely attributable to the higher plasma triglycerides seen in noncarriers at baseline, with on-trial concentrations of triglycerides being nearly identical in carriers (111 Ϯ 51 mg/dl) versus noncarriers (114 Ϯ 49 mg/dl) of the LPL S447X variant.
On-trial concentrations of plasma lipoprotein subclasses according to LPL genotype are provided in Table 5. The most striking differences in lipoprotein subclass concentrations in carriers versus noncarriers were seen among the LDL subclasses according to LPL D9N genotype. On gemfibrozil, carriers of the LPL N9 allele had significantly lower levels of large LDL (Ϫ32%) but significantly higher levels of small, dense LDL (ϩ59%) than did noncarriers. This is in contrast to baseline values, in which differences in LDL subclass levels were not observed among the LPL D9N genotypes. The substantial differences in concentrations of large and small LDL observed in N9 carriers versus noncarriers translated into a mean LDL particle diameter of 20.14 Ϯ 0.87 nm for carriers, which was significantly smaller than that of 20.63 Ϯ 0.80 nm for noncarriers (P Ͻ 0.003). In contrast, no statistically significant differences were observed in on-trial concentrations of lipoprotein subclasses for the LPL N291S and S447X genotypes.

DISCUSSION
HDL deficiency is the most common lipid abnormality observed among patients with premature CHD (31). It has been reported that more than 50% of the variation in HDL-C levels in humans is genetically determined (32), with gene products that influence the amount and nature of lipid contained within HDL particles having important effects on the metabolism of HDL and apoA-I. Included among these gene products is LPL. We have previously reported that genetic variation at the LPL locus was associated with plasma lipoprotein levels and CHD risk in men from the FOS (16,17). To further explore the role of common LPL variants in the modulation of CHD risk at the population level, we examined the relationships between three LPL polymorphisms, plasma lipoprotein levels, response to gemfibrozil therapy, and CHD end points in men from the VA-HIT. VA-HIT is distinct from other populations used previously to examine such associations in that it consists solely of men with preexisting CHD, low HDL-C, and normal LDL-C.
In VA-HIT, 5.3% of men were carriers of the LPL N9 allele, 4.5% were carriers of the LPL S291 allele, and 13.0% were carriers of the LPL X447 allele. Comparatively, the meta-analysis of Wittrup, Tybjaerg-Hansen, and Nordestgaard (13) identified heterozygote frequencies of 2-4% for LPL D9N, 1-7% for LPL N291S, and 17-22% for LPL S447X for white subjects. To further investigate the associations between these LPL genotypes and CHD risk, we compared men in VA-HIT with men without CHD from FOS, who were stratified according to HDL-C level (у40 or Ͻ40 mg/dl). The contrast between VA-HIT and FOS men with HDL-C у 40 provided a comparison of those with low HDL-C and CHD with those who have neither, whereas the contrast between VA-HIT and FOS men with HDL-C Ͻ 40 provided a comparison of those who have CHD versus those without CHD among men with low HDL-C. In contrast to what was seen in VA-HIT, these analyses revealed that the frequency of the LPL N9 allele was not significantly increased in the group of men from FOS having an HDL-C of Ͻ40 mg/dl compared with men from FOS having an HDL-C of у40. However, similar to what was seen in VA-HIT, the LPL S291 and X447 alleles were significantly increased and decreased, respectively, in men from FOS with low HDL-C relative to men in FOS with normal HDL-C. Taken together with the meta-analysis of Wittrup, Tybjaerg-Hansen, and Nordestgaard (13), our data suggest that the LPL N9 allele is associated with increased CHD risk, whereas the LPL S291 and X447 alleles may be more strongly associated with HDL-C phenotype than with CHD risk. In support of this concept, a number of studies have reported increased frequency of the mutant LPL N9 allele in patients with CHD (13,(33)(34)(35)(36)(37)(38).
The LPL N9, S291, and X447 alleles have been reasonably consistent in terms of their associations with plasma lipid concentrations (13). The N9 and S291 alleles are associated with lower levels of HDL-C and higher levels of triglycerides, whereas the converse is true for the X447 allele. The differential effects of these variants on plasma lipid levels may be related to their location within the LPL protein, with the D9N and N291S variants located in the N terminus and the S447X variant located in the C terminus (10). In the present study, we did not observe any statistically significant relationships between these common