Genome-Wide Association Study Evaluating Lipoprotein-Associated Phospholipase A2 Mass and Activity at Baseline and After Rosuvastatin TherapyClinical Perspective
Background—Lipoprotein-associated phospholipase A2 (Lp-PLA2) is a proinflammatory enzyme bound to low-density lipoprotein cholesterol and other circulating lipoproteins. Two measures of Lp-PLA2, mass and activity, are associated with increased cardiovascular risk. Data are sparse regarding genetic determinants of Lp-PLA2 mass and activity, and no prior data are available addressing genetic determinants of statin-induced changes for this proinflammatory biomarker.
Methods and Results—We performed a genome-wide association study of Lp-PLA2 mass and activity at baseline and after 12 months of rosuvastatin therapy (20 mg/d) among 6851 participants of European ancestry from the Justification for Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER) and performed replication in a meta-analysis of 13 664 participants from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium. Novel associations were identified and replicated at MS4A4E and TMEM49 for baseline Lp-PLA2 activity with genome-wide significant joint P values (P=2.0×10−11 and P=2.9×10−9, respectively). In addition, genome-wide associations (P<5×10−8) were identified and replicated for baseline Lp-PLA2 mass at CETP and for Lp-PLA2 activity at the APOC1-APOE and PLA2G7 loci. Among 2673 statin-allocated participants, both Lp-PLA2 mass and activity were reduced by >30% and low-density lipoprotein cholesterol by 50% after 12 months of statin therapy (P<0.001 for both). Variants in ABCG2 and LPA were associated with change in statin-induced Lp-PLA2 activity at genome-wide significance but were substantially attenuated after adjustment for statin-induced changes in lipid levels.
Conclusions—Genome-wide significant associations at MS4A4E and TMEM49 may reflect novel influences on circulating levels of Lp-PLA2 activity. In addition, genome-wide significant associations with rosuvastatin-induced change in Lp-PLA2 activity were observed in ABCG2 and LPA, likely because of their impact on statin-induced low-density lipoprotein cholesterol lowering.
- genome-wide association study
- lipoprotein-associated phospholipase A2
- statin therapy
Lipoprotein-associated phospholipase A2 (Lp-PLA2) is a proinflammatory enzyme bound to plasma lipoproteins in circulation (≈70%–80% to low-density lipoprotein cholesterol [LDL-C] and the remaining to high-density lipoprotein cholesterol [HDL-C] and very LDL-C). Epidemiological studies have associated both higher concentrations of Lp-PLA2 and elevated Lp-PLA2 enzyme activity with greater risk of developing atherosclerosis and cardiovascular disease (CVD), independent from risk associated with circulating lipid levels.1 Additional lines of evidence support a causal role for Lp-PLA2 in CVD and coronary artery disease (CAD): higher expression of Lp-PLA2 has been observed in unstable and ruptured lesions,2,3 slower progression of advanced atherosclerotic lesions has been observed in swine after pharmacological inhibition of Lp-PLA2 activity,4 and a rare loss of function mutation in the gene encoding Lp-PLA2 (PLA2G7) was associated with lower risk of developing CAD among South Koreans,5 although other studies of this mutation in other populations, Asian6,7 and non-Asian,7,8 have not always yielded consistent associations with CAD.
Clinical Perspective on p 685
The heritability of plasma Lp-PLA2 mass is estimated to be between 0.21 and 0.37, and the heritability of Lp-PLA2 enzymatic activity is estimated to be between 0.28 and 0.41.9–11 Previous investigations of genetic determinants of plasma Lp-PLA2 focused on the gene that encodes Lp-PLA2 (PLA2G7) and genes encoding other inflammatory molecules.8,12,13 A genome-wide association study (GWAS) of circulating Lp-PLA2 was performed among 6688 participants of the Framingham Heart Study (FHS).14 Variants in 5 genetic loci were associated with Lp-PLA2 at a genome-wide significance: PLA2G7 for Lp-PLA2 mass and APOC1-APOE, SORT1, SCARB1, and APOA1-APOA5 for Lp-PLA2 activity. More recently, a genome-wide meta-analysis of 13 664 participants from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE, which includes the FHS) Consortium identified genome-wide associations for the same variants, in addition to CETP for Lp-PLA2 mass and PLA2G7 and LDLR for Lp-PLA2 activity.15 Variants in the loci identified through the FHS GWAS and the CHARGE meta-analysis are known to be involved in lipid metabolism and have been observed to strongly influence circulating levels of LDL-C and HDL-C.
Agents that reduce Lp-PLA2 activity are currently under investigation for reduction of cardiovascular risk.16,17 In addition, statins reduce Lp-PLA2 activity.18 However, little is known about genetic influences on the response of Lp-PLA2 concentration or enzymatic activity to statin therapy. Therefore, we conducted a GWAS among 6851 Justification for Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER) participants who were randomly allocated to 20 mg/day of rosuvastatin or placebo for 12 months to identify genetic determinants of baseline Lp-PLA2 mass and activityand genetic determinants of the response of Lp-PLA2 mass and activity to statin therapy.
The study population for this analysis was derived from JUPITER trial (NCT00239681). JUPITER is a multinational, randomized, placebo-controlled evaluation of rosuvastatin (20 mg/d) conducted among men and women free of CVD and type 2 diabetes mellitus with moderate-to-low LDL-C levels (<130 mg/dL) and elevated C-reactive protein (CRP) levels (≥2 mg/L) at baseline.19 Blood samples were obtained at time of randomization and after 1 year of treatment. All trial participants had Lp-PLA2 mass measured by immunoassay (PLAC assay) and Lp-PLA2 activity measured by calorimetry (CAM kit), performed at diaDexus (San Francisco, CA); LDL-C, HDL-C, and triglycerides were measured in a core laboratory facility.
Each study participant underwent genotyping for a total of 1 006 348 single-nucleotide polymorphisms (SNPs) using the Omni 1M Quad platform and GenomeStudio software v1.6.2 (both Illumina, San Diego, CA) by the manufacturer. SNPs with poor clustering metrics for parameters, such as ABrMean (intensity), cluster separation, Hardy-Weinberg Equilibrium, and call frequency, were visually inspected and either annotated, removed, or manually edited. Markers were retained for the final data if the updated clusters met quality standards and the genotyping was successful in >98.5% of the samples. Multidimensional scaling procedures implemented in PLINK20 were used for verification of self-reported European ancestry. Sub-European population stratification was estimated using EIGENSTRAT.21 First-degree relatives were identified by identity by state clustering in PLINK and excluded.
The investigation of the genetics of baseline Lp-PLA2 mass and activity was limited to 6851 JUPITER participants who (1) provided informed consent for genetic analyses, (2) provided baseline blood samples that successfully underwent Lp-PLA2 measurement, (3) did not self-report nontrial statin use, and (4) were of European ancestry so that confounding by population stratification could be minimized a priori. Both placebo- and rosuvastatin-allocated participants were included in the analysis of baseline levels. Analysis of change in Lp-PLA2 mass and activity was limited to 5329 JUPITER participants who (1) provided 1-year blood samples, (2) were free of diabetes mellitus at follow-up, and (3) were compliant with study medication based on pill counts.
Genetic associations were evaluated for 6 outcomes in this study: (1) baseline Lp-PLA2 mass, (2) baseline Lp-PLA2 activity, (3) absolute change in Lp-PLA2 mass, (4) percentage change in Lp-PLA2 mass, (5) absolute change in Lp-PLA2 activity, and (6) percentage change in Lp-PLA2 activity. Absolute change was calculated as the difference between the 12-month and baseline value, and percentage change was calculated from absolute change divided by the baseline value; analyses of change variables were stratified by statin allocation. For each outcome, estimates of genetic effects were obtained from analysis of these measures, and P values for association were obtained from analysis of these measures that were transformed using inverse-quantile normalization procedures to decrease the influence of extreme outliers.22
Our main objectives were to perform a GWAS of Lp-PLA2 mass and activity measures in JUPITER using 796 174 genotyped SNPs with minor allele frequencies >1%, Hardy-Weinberg equilibrium P>10–8, and successful genotyping in >95% of the samples. rs7412, encoding the e2 allele of APOE, failed Hardy-Weinberg equilibrium (P=1.1×10−9) but was retained because of its impact on LDL-C levels. Examination of raw genotypes for rs7412 in 1400 participants (see p19 online-only Data Supplement Materials)23 indicates good discrimination between genotype clusters and suggests that the deviation of rs7412 from Hardy-Weinberg equilibrium was not caused by genotyping error. We suggest ascertainment bias as a potential alternative explanation; enrollment criteria for JUPITER required LDL-C <130 mg/dL, leading to oversampling of the e2 allele that is associated with lower levels of LDL-C. A conventional threshold of α=5×10−8 was set to determine genome-wide significance. Linear regressions assuming an additive mode of inheritance (minor allele count: 0, 1, and 2) were performed in PLINK to analyze all Lp-PLA2 measures. Primary models were adjusted for age, sex, body mass index (BMI), region, current smoking status, and subpopulation stratification measures to gain power by reducing variance as a result of known clinical influences on Lp-PLA2 measures, to avoid finding genetic variants acting on Lp-PLA2 through BMI or smoking, and to avoid confounding by subpopulation stratification. To assess whether genetic effects on Lp-PLA2 were mediated through effects on plasma lipoproteins, we examined models simultaneously adjusting for baseline LDL-C, HDL-C, and triglycerides or simultaneously adjusting for change in these biomarkers. In addition, we performed regression models individually adjusted for baseline, absolute change, or percentage change (depending on the outcome under analysis) in LDL-C, HDL-C, triglycerides, apolipoprotein (apo) B, LDL-C/apoB ratio, HDL-C/LDL-C ratio, and CRP to further explore genetic mechanisms influencing Lp-PLA2 concentration and enzymatic activity. For any associations attaining genome-wide significance in analysis of absolute change or percentage change among statin-allocated participants, we performed interaction analyses combining the statin- and placebo-allocated participants and by adding an SNP-by-statin interaction term to the regression model on untransformed absolute or percentage change in Lp-PLA2 trimmed for outliers.
Sensitivity analyses were performed to ensure the robustness of our genome-wide association results; all genome-wide effects observed in this study were examined for effect of exclusion of BMI and smoking status and examined for interactions with age, sex, BMI, and smoking status.
The top results from the analysis of baseline Lp-PLA2 mass and activity were evaluated in a meta-analysis from the CHARGE Consortium of Lp-PLA2 mass and activity15; if association results for the index SNP at a particular locus were not available, we chose the next most significant SNP in linkage disequilibrium (LD), with the index SNP as a proxy for replication in CHARGE. The CHARGE sample consists of 5 studies totaling up to 13 664 participants of European ancestry. As previously described,15 Lp-PLA2 mass and activity were natural log-transformed before analysis. A minimal model was adjusted for age and sex, and a multivariable model was adjusted for cardiovascular risk factors, including type 2 diabetes mellitus, lipid-lowering medication, antihypertensive treatment, aspirin intake ≥3 per week, current smoking, hormone replacement therapy, BMI, systolic and diastolic blood pressure, waist circumference, prevalent CVD, LDL-C, HDL-C, and triglycerides. Combined P values were calculated by standard sample size–weighted meta-analysis in METAL24; the regression model selected for replication was comparable with the discovery model. Individual CHARGE Consortium cohort characteristics and Lp-PLA2 assay information are available from Grallert et al.15 The threshold for replication was set to 3.6×10−3 (α=0.05/14) to account for testing 5 loci meeting genome-wide significance (P<5×10−8) and 9 loci meeting sub–genome-wide significance (5×10−6>P>5×10−8).
The population characteristics of the 6851 JUPITER participants who consented for genetic research and had baseline Lp-PLA2 mass or activity available for analysis are available in online-only Data Supplement Table I. Of these, 5329 (2673 randomly allocated to rosuvastatin and 2656 to placebo) were compliant with statin or placebo therapy and had successful measurement of Lp-PLA2 mass or activity on blood samples provided at 12 months after randomization. No differences in major cardiovascular risk factors were observed between the placebo- and rosuvastatin-allocated participants at baseline. After 12 months on statin therapy, the median reduction in Lp-PLA2 mass was 107.4 ng/mL (−36% [interquartile range, −44% to −25%]), Lp-PLA2 activity was 64.8 nmol/min per mL (−35% [interquartile range, −41% to −27%]), and LDL-C was 54.0 mg/dL (−52% [interquartile range, −60% to −41%]).
SNPs Associated With Baseline Lp-PLA2 Mass
Genome-wide analysis of Lp-PLA2 mass among 6851 JUPITER participants at baseline identified 2 loci—CETP (rs3764261) and GCKR (rs1260326)—that met genome-wide significance thresholds (P<5×10−8) (Figure 1A and Table 1). Quantile-quantile analysis of association results excluding SNPs within 1Mb of the 2 genome-wide signals showed residual evidence of excess SNPs with smaller P values than expected under the null hypothesis (online-only Data Supplement Figures IA and II). Therefore, a threshold of P<5×10−6 was chosen to indicate sub–genome-wide significance, indicating that some associations were below genome-wide significance. Three potential candidate loci—APOB, TBL2, and GPR180—were identified at sub–genome-wide significance (P<5×10−6; Table 1). To investigate whether these associations were independent of lipid levels, primary regression models were simultaneously adjusted for baseline LDL-C, HDL-C, and triglycerides. Lipid adjustment attenuated the association between SNP rs3764261 at the CETP locus and Lp-PLA2 mass, but remained genome-wide significant (P=3.2×10−8; Table 1), and the attenuation can be largely attributed to adjustment for HDL-C (online-only Data Supplement Table II). All other associations were attenuated by lipid adjustment and were no longer sub–genome-wide significant (P>5×10−6 for all; Table 1). Adjustment for baseline CRP levels, a second inflammatory biomarker, did not influence observed genetic associations with Lp-PLA2 mass (online-only Data Supplement Table II).
The most significantly associated variant at each genome-wide or sub–genome-wide locus was evaluated for independent replication with Lp-PLA2 mass in the CHARGE Consortium meta-analysis (Table 1). Only the association for rs3764261 at the CETP locus was replicated (CHARGE P value, PCHARGE=6.5×10−8; joint P value, Pjoint=7.2×10−21); this locus was previously identified by the CHARGE Consortium meta-analysis (rs247616, PCHARGE=2.5×10−8; r2 with rs3764261=1.0).15
SNPs Associated With Baseline Lp-PLA2 Activity
Genome-wide analysis of baseline Lp-PLA2 activity identified 2 genome-wide significant loci at APOE (rs7412) and MS4A4E (rs600550) (Figure 1B and Table 1; locus association plot for MS4A4E in Figure 2A). Excluding SNPs within 1 Mb of the 2 genome-wide significant loci, quantile-quantile analyses of baseline Lp-PLA2 activity revealed an excess of small P values at sub–genome-wide significance (P<5×10−6; online-only Data Supplement Figures IB and III) and identified 6 potential additional candidate loci for baseline Lp-PLA2 activity: PLA2G7 (that encodes for the Lp-PLA2 enzyme), SORT1, PPARG, SCARB1, and LIPC that have been previously associated with lipid metabolism; and TMEM49 (Table 1). Additional adjustment for baseline LDL-C, HDL-C, and triglycerides had a minor effect on most of these associations (Figure 1C, Table 1, and online-only Data Supplement Table II). After adjustment, the association at the APOE locus was attenuated but remained genome-wide significant (P=3.4×10−40), the association for the MS4A4E locus remained genome-wide significant (P=1.1×10−11), the PLA2G7 locus attained genome-wide significance (P=2.7×10−9), and the TMEM49 locus was close to genome-wide significance (P=8.5×10−8; locus association plot in Figure 2B). Adjustment for baseline CRP levels did not affect these observed genetic associations with Lp-PLA2 activity, nor did adjustment for LDL-C/apoB ratio (surrogate for mean LDL particle size) or HDL-C/LDL-C ratio (online-only Data Supplement Table II).
The most significantly associated variant at each genome-wide and sub–genome-wide locus was evaluated for independent replication with Lp-PLA2 activity by the CHARGE Consortium. The index SNP at the APOC1-APOE locus rs7412 was not available in the CHARGE meta-analysis; therefore, the next most significant SNP at the same locus, rs445925 (P=1.3×10−53), was selected for replication in CHARGE (r2=0.87 with rs7412 in JUPITER). The associations of all 3 genome-wide loci with Lp-PLA2 activity were replicated by the CHARGE meta-analysis (PLA2G7, MS4A4E, and the APOC1-APOE gene complex; P<3.6×10−3 for all; Table 1). The sub–genome-wide associations of the SORT1, SCARB1, and TMEM49 loci were replicated in the CHARGE meta-analysis (P<3.6×10−3 for all; Table 1). The associations at SORT1, PLA2G7, SCARB1, and the APOC1-APOE complex were initially identified by the FHS analysis14 and the CHARGE meta-analysis.15 The 2 loci (MS4A4E and TMEM49) solely identified in JUPITER were both replicated in association with Lp-PLA2 activity by the CHARGE Consortium and attained genome-wide significance in a joint meta-analysis (MS4A4E: PCHARGE=9.3×10−5, Pjoint=2.0×10−11; TMEM49: PCHARGE=7.7×10−4, Pjoint=2.8×10−9; Table 1).
Replication of CHARGE Consortium Lp-PLA2 Mass and Activity Meta-Analysis
Eight variants at 7 loci (SORT1, PLA2G7, APOA1-A5 gene complex, SCARB1, CETP, LDLR, APOC1-APOE gene complex; listed in online-only Data Supplement Table III) were associated with Lp-PLA2 mass or activity at genome-wide significance in a meta-analysis of up to 13 664 participants from the CHARGE Consortium.15 We examined these SNPs in association with Lp-PLA2 mass and activity in 6851 participants in JUPITER. Using the Bonferroni method to correct for multiple hypothesis testing, we replicated the association for 1 of 2 Lp-PLA2 mass loci at CETP (P=6.8×10−17 [α=0.025=0.05/2 Lp-PLA2 mass loci]) and all 6 of the Lp-PLA2 activity loci at SORT1, APOA1-A5 gene complex, SCARB1, LDLR, and APOC1-APOE gene complex (P=1.1×10−6, 3.3×10−5, 8.8×10−7, 1.5×10−3, and 9.2×10−21, respectively [α=8.3×10−3=0.05/6 Lp-PLA2 activity loci]; online-only Data Supplement Table III).
SNPs Associated With Statin-Induced Change in Lp-PLA2 Mass
In genome-wide analyses of genetic determinants of rosuvastatin-induced change in Lp-PLA2 mass, no locus reached genome-wide significance for association with either statin-induced absolute change or percentage change in Lp-PLA2 mass (online-only Data Supplement Figures IV and V). Quantile-quantile plot analysis of association results for both measures of change in Lp-PLA2 mass did not show evidence of excess SNPs with small P values that may not have reached genome-wide significance. No genome-wide significant associations were detected in analysis of absolute or percentage change in Lp-PLA2 mass among the placebo-allocated arm (data not shown).
SNPs Associated With Statin-Induced Change in Lp-PLA2 Activity
In contrast, the LPA and the ABCG2 loci were identified in a genome-wide scan for SNPs associated with rosuvastatin-induced percentage change in Lp-PLA2 activity at a genome-wide level (Figure 1D, Table 2, and online-only Data Supplement Figure ID). The lead SNP at the LPA locus was rs10455872 (P=1.8×10−16) and was associated with a mean increase of 4.8% (SE, 0.7%) in Lp-PLA2 activity per minor allele. The next strongest association was in the ABCG2 gene for rs2199936 (P=1.6×10−10), with a mean per-allele decrease of 2.8% (SE, 0.5%) in Lp-PLA2 activity. Neither of these 2 genome-wide associations was observed among placebo-allocated participants (P=0.81 and 0.28 for rs10455872 and rs2199936, respectively). P values for SNP-by-statin interaction analysis of percentage change in Lp-PLA2 activity were significant for both rs10455872 and rs2199936 (Pinteraction=8.0×10−7 and 1.2×10−3, respectively).
In quantile-quantile analysis of these association results after excluding SNPs within 1 Mb of genome-wide significant loci, an excess of small P values was observed (online-only Data Supplement Figure VI). Sub–genome-wide associations were observed at EDG7, C14orf177, NOX5, AP3B2, and RNF213 for percentage change in Lp-PLA2 activity (Table 2). None of these 5 loci was previously linked to rosuvastatin pharmacodynamics or lipid metabolism.
Most of the genome-wide associations for statin-induced percentage change in Lp-PLA2 activity were attenuated in analyses simultaneously adjusting for change in lipid levels after 12 months of statin therapy (Table 2) and could be largely attributed to adjustment for change in LDL-C (online-only Data Supplement Table II). However, the sub–genome-wide association of rs11638815 in AP3B2 with change in Lp-PLA2 activity was not affected by adjustment for change in lipid levels, CRP, LDL-C/apoB ratio, or HDL-C/LDL-C ratio after 12 months of statin therapy (Table 2 and online-only Data Supplement Table II).
No genome-wide significant associations were observed in analysis of statin-induced absolute change in Lp-PLA2 activity (Figure 1E and online-only Data Supplement Figure IE). No genome-wide significant associations were detected in analysis of absolute or percentage change in Lp-PLA2 activity among the placebo-allocated arm (data not shown).
All genome-wide effects observed in this study were robust to regression models that did not include BMI and smoking status (online-only Data Supplement Table IV). None of the genome-wide associations reported here showed evidence of interaction with age, sex, BMI, or smoking status after correction for multiple testing (online-only Data Supplement Table V).
In this evaluation of circulating plasma levels of Lp-PLA2 among 6851 JUPITER participants of European ancestry, we observed a novel genome-wide and sub–genome-wide association at the MS4A4E locus (P=3.0×10−9) and TMEM49 locus (P=8.5×10−8), respectively, for baseline Lp-PLA2 activity. Associations at both loci were replicated in a meta-analysis of ≈13 000 participants (P<3.6×10−3), and joint P values attained genome-wide significance (Pjoint<5×10−8). Additional genome-wide significant associations identified for baseline Lp-PLA2 measures replicated loci identified in prior genetic analyses.14,15 In analysis of change in Lp-PLA2 activity after 12 months of statin therapy, we identified 2 novel genome-wide loci, ABCG2 and LPA, with differential effects on lowering Lp-PLA2 activity in those randomly allocated to rosuvastatin versus placebo, as well as 5 additional loci (EDG7, C14orf177, NOX5, AP3B2, and RNF213) at sub–genome-wide significance. Except for AP3B2, all genetic associations for statin-induced change in Lp-PLA2 activity were substantially attenuated after simultaneous adjustment for statin-induced change in LDL-C, HDL-C, and triglycerides.
Our study, in addition to the genome-wide meta-analysis performed by the CHARGE Consortium,15 gives insight into the disparate genetic influences on Lp-PLA2 mass and activity. Differential associations may not be surprising, given modest correlation between these measures in JUPITER (Spearman’s ρ=0.24). Among JUPITER participants, SNPs at some loci (APOB, MS4A4E, SCARB1, and TMEM49) had at least nominal association with both mass and activity, whereas SNPs at other loci were exclusively associated with only 1 Lp-LPA2 measure (online-only Data Supplement Table VI). The differences may reflect different biological influences on Lp-PLA2, and, consistent with this hypothesis, differential associations of Lp-PLA2 mass and activity with vascular events have also been observed, although only in some studies.25 However, current debate over the accuracy of the mass assay26,27 limits strong conclusions about the sources of the differential associations. We were not able to examine genetic associations with vascular events in the sample for this analysis because of power constraints (44 and 72 events in the rosuvastatin- and placebo-allocated arms, respectively).
In plasma, Lp-PLA2 is bound to LDL particles and to a lesser extent also to HDL and VLDL particles, but the degree to which the genetic associations with baseline Lp-PLA2 mass and activity are mediated by lipoprotein concentration is not clear. The genetic effect of the CETP and GCKR loci on baseline Lp-PLA2 mass was attenuated but nevertheless remained genome-wide significant in models simultaneously adjusting for lipid levels, as well as after adjusting for apoB (surrogate for mean LDL particle number), LDL-C/apoB ratio (surrogate for mean LDL particle size), and HDL-C/LDL-C ratio. The same is true for the association of the APOC1-APOE locus and Lp-PLA2 activity. Our results support the notion of pleiotropic genetic effects on lipid and lipid-associated biomarkers and also suggest that some of these genetic effects on Lp-PLA2 may not be mediated entirely through lipoprotein composition or concentration.
In contrast to the above examples, the genome-wide significant association of MS4A4E with baseline Lp-PLA2 activity was not affected by lipoprotein adjustment. MS4A4E encodes for a subunit of a membrane-spanning protein and is located near the MS4A gene cluster on chromosome 11; therefore, we cannot rule out the possibility that the association is mediated through another gene in this region. The apparent independence of this association from lipoprotein levels is consistent with the lack of reported genome-wide associations with lipid fractions at the MS4A cluster.28 Variants near MS4A gene cluster have been identified as influencing risk of developing Alzheimer disease,29 but the gene function is uncertain.
TMEM49, 1 of the 2 loci along with PLA2G7 attaining genome-wide significance after adjustment for lipoproteins, has not been previously associated with any aspect of lipid metabolism. Also known as VMP1, TMEM49 encodes a stress-induced transmembrane protein that induces the formation of vacuoles and promotes apoptosis. Overexpression of this gene has been observed in rodent models with experimentally induced acute pancreatitis,30 but TMEM49 has not been associated with Lp-PLA2 activity or lipoproteins.
Among statin-allocated participants, the effect of ABCG2 on change in Lp-PLA2 activity was markedly attenuated with the inclusion of 12-month change in lipid measures, especially by adjustment for 12-month change in LDL-C or apoB. ABCG2 is a known hepatic transporter of rosuvastatin,31 and this ABCG2 variant was previously associated with statin-induced LDL-C reduction at genome-wide significance in JUPITER with the same direction of effect (−5.1% [SE, 0.9%] per minor allele; P=1.7×10−15).23 Therefore, the attenuated association can be explained by the high correlation observed between statin-induced Lp-PLA2 activity reduction and LDL-C/apoB reduction and by the physical interaction of Lp-PLA2 with LDL particles containing both apoB and LDL-C. A modest sub–genome-wide association remained after adjustment for 12-month change in lipid levels between LPA and change in Lp-PLA2 activity on statin therapy. LPA is located at chromosome 6q26 and encodes apo(a). Similar to the association of ABCG2 and statin-induced change in LDL-C in JUPITER, this variant in LPA was also associated with statin-induced LDL-C change in this cohort (6.8% [SE, 1.2%] per minor allele; P=5.0×10−15)23 and has been previously associated with baseline LDL-C levels,14 HDL-C levels,14 and higher risk of cardiovascular events.32 However, the residual association between the LPA variant and change in Lp-PLA2 activity may potentially be a result of an interaction among lipoprotein(a), Lp-PLA2, and oxidized LDL.33,34
In contrast, the sub–genome-wide association of statin-induced change in Lp-PLA2 activity with AP3B2 was not affected by adjustment for statin-induced change lipid levels. AP3B2 is a vesicle-coat protein complex involved in protein sorting expressed in neuronal cells. Variants in this gene have not been associated with the pharmacodynamics of statins or lipoprotein metabolism.
Although there is no doubt that genetics influence Lp-PLA2 concentration and activity, the causality of Lp-PLA2 in CVD is still under debate. Primary results from trials of the Lp-PLA2 activity inhibitor darapladib16,17 have not been published yet, and formal Mendelian randomization studies have not been performed. Several quasi-Mendelian randomization analyses examining variants in PLA2G7 known to affect Lp-PLA2 mass and activity and coronary heart disease have been published5,7,8; however, the results have not been consistent. In addition to the results of genome-wide meta-analyses for Lp-PLA2 mass and activity, the CHARGE Consortium reported a null association of variants in the PLA2G7 gene with prevalent coronary heart disease/CAD from CARDIoGRAM15 (a large consortium of coronary heart disease/CAD cases and controls). However, Grallert et al15 suggest that null finding could partially be attributed to low power (from a combination of modest genetic effect of PLA2G7 on Lp-PLA2 mass and activity and modest influence of Lp-PLA2 mass and activity on coronary heart disease/CAD), despite a sample size of >80 000 individuals in CARDIoGRAM.
Several limitations of this study warrant discussion. First, Lp-PLA2 mass and activity are correlated with LDL-C, and the effect of statins on LDL-C lowering may confound the effect of statins on reduction in Lp-PLA2 mass and activity. However, adjustment for lipoproteins revealed a novel genetic association specific to baseline Lp-PLA2 activity at the TMEM49 locus and a potential novel candidate locus for statin-induced change in Lp-PLA2 activity at AP3B2. Second, participants were enrolled on the basis of low LDL-C levels, which may have affected the distribution of baseline Lp-PLA2 mass and activity and possibly affected the power to detect variants associated with higher levels of mass and activity. The truncated distribution of LDL-C in our sample may have mitigated the previously reported association of the PLA2G7 loci with Lp-PLA2 mass15 and may have also affected the replication efforts for the GCKR–Lp-PLA2 mass association. Third, with respect to the null results from the genome-wide analysis of statin-induced change in Lp-PLA2 mass, it is possible that our study was not adequately powered to detect genetic effects if they were relatively weak, due to rare variants, or the result of environmental interactions.
Nevertheless, our study has several strengths. The current investigation is the largest single-study genetic evaluation of baseline Lp-PLA2 mass and activity and the only genetic study of statin-induced change in Lp-PLA2 mass and activity to date. All measures of Lp-PLA2 mass were determined in a single experiment using a single assay, and all assays were performed by the manufacturer, further ensuring assay accuracy.
The genetic associations of MS4A4E and TMEM49 suggest that some biological pathways underlying Lp-PLA2 activity may be distinct from pathways for lipoprotein metabolism. In contrast, the genetic associations with statin-induced reduction of Lp-PLA2 activity could be largely attributed to statin-induced LDL-C/apoB reduction. It will be interesting to learn how these findings may be related to the outcome of ongoing clinical trials of darapladib for the impact of Lp-PLA2 activity reduction on vascular events among patients already treated with statin therapy.
We thank Jean MacFadyen, Lynda Rose, the JUPITER study participants, and the >1000 physicians worldwide for their personal time, effort, and commitment to JUPITER.
Sources of Funding
This work was supported by research funds from AstraZeneca to Drs Ridker and Chasman.
Drs Barratt and Nyberg are employees of AstraZeneca. Dr Ridker reports being listed as a coinventor on patents related to the use of inflammatory biomarkers in cardiovascular disease, which are held by the Brigham and Women’s Hospital and have been licensed by Siemens and AstraZeneca. Dr Ridker receives funding from the National Heart, Lung, and Blood Institute, the National Cancer Institute, AstraZeneca, Sanofi-Aventis, and Novartis. The other authors report no conflicts.
↵* On behalf of the CHARGE Inflammation Working Group.
↵† These authors contributed equally to this work.
The online-only Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.112.963314/-/DC1.
- Received March 14, 2012.
- Accepted October 12, 2012.
- © 2012 American Heart Association, Inc.
- Thompson A,
- Gao P,
- Orfei L,
- Watson S,
- Di Angelantonio E,
- Kaptoge S,
- et al
- Häkkinen T,
- Luoma JS,
- Hiltunen MO,
- Macphee CH,
- Milliner KJ,
- Patel L,
- et al
- Kolodgie FD,
- Burke AP,
- Skorija KS,
- Ladich E,
- Kutys R,
- Makuria AT,
- et al
- Casas JP,
- Ninio E,
- Panayiotou A,
- Palmen J,
- Cooper JA,
- Ricketts SL,
- et al
- Guerra R,
- Zhao B,
- Mooser V,
- Stafforini D,
- Johnston JM,
- Cohen JC
- Ninio E,
- Tregouet D,
- Carrier JL,
- Stengel D,
- Bickel C,
- Perret C,
- et al
- Hoffmann MM,
- Winkler K,
- Renner W,
- Winkelmann BR,
- Seelhorst U,
- Wellnitz B,
- et al
- Grallert H,
- Dupuis J,
- Bis JC,
- Dehghan A,
- Barbalic M,
- Baumert J,
- et al
- 16.↵The stabilization of atherosclerotic plaque by initation of darapladib therapy trial (STABILITY). http://clinicaltrials.gov/ct2/show/nct00799903, Accessed August 24, 2012.
- 17.↵The stabilization of plaques using darapladib-thromboysis in myocardial infarction 52 Trial (SOLID-TIMI). http://clinicaltrials.gov/ct2/show/NCT01000727. Accessed August 24, 2012.
- Chasman DI,
- Giulianini F,
- MacFadyen J,
- Barratt BJ,
- Nyberg F,
- Ridker PM
- Willer CJ,
- Li Y,
- Abecasis GR
- Mallat Z,
- Lambeau G,
- Tedgui A
- McConnell JP,
- Jaffe AS
- Stein EA
- Hollingworth P,
- Harold D,
- Sims R,
- Gerrish A,
- Lambert JC,
- Carrasquillo MM,
- et al
- Dusetti NJ,
- Jiang Y,
- Vaccaro MI,
- Tomasini R,
- Azizi Samir A,
- Calvo EL,
- et al
- Kitamura S,
- Maeda K,
- Wang Y,
- Sugiyama Y
- Blencowe C,
- Hermetter A,
- Kostner GM,
- Deigner HP
Lipoprotein-associated phospholipase A2 (Lp-PLA2) is a proinflammatory enzyme associated with cardiovascular disease. Lp-PLA2 is primarily bound to low-density lipoprotein (LDL) particles in serum, but its association with cardiovascular disease is independent of risk conferred by LDL cholesterol (LDL-C). Although statins reduce Lp-PLA2 levels concomitant with their effects on LDL-C, direct inhibitors of Lp-PLA2 are also alternative agents for cardiovascular disease prevention in clinical trials. To better understand the biology of Lp-PLA2 and its response to statin therapy, we evaluated genetic variation across the genome for effects on Lp-PLA2 mass and activity at baseline and after 12 months of rosuvastatin therapy (20 mg/d) or placebo among 6851 participants of Justification for Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER), a randomized trial for primary prevention of cardiovascular disease. Lp-PLA2 activity at baseline was influenced by genetic variation at the novel loci MS4A4E and TMEM49, independent of association with LDL-C, high-density lipoprotein cholesterol, and triglycerides. In addition, effects on baseline measures of Lp-PLA2 mass and activity were confirmed at the previously recognized loci CETP, APOC1–APOE, and PLA2G7. Among 2673 rosuvastatin-allocated participants in JUPITER, only the ABCG2 and LPA loci were associated with rosuvastatin response of Lp-PLA2 activity. However, these effects were explained by concordant effects at these loci on LDL-C response to rosuvastatin. These results suggest that genetic determinants of circulating Lp-PLA2 at baseline may differ from those of LDL-C and provide evidence that genetic effects on statin-induced changes in Lp-PLA2 activity are dependent on the effects of these loci on LDL-C reduction.