Lipoprotein-associated phospholipase A₂ (Lp-PLA₂) has emerged as a potential therapeutic target in coronary heart disease (CHD) and phase III clinical trials are underway. Supporting evidence includes apparent atherogenic biochemical properties; Lp-PLA₂ cleaves oxidized phosphatidylcholine on modified low-density lipoprotein (LDL), producing inflammatory lysophosphatidylcholine and oxidized nonesterified fatty acids (1- 2). In addition, enzymatic expression of Lp-PLA₂ is up-regulated in human atherosclerosis (3), whereas circulating levels are associated with incident CHD (4). Promising proof-of-principle pre-clinical and clinical trials have been carried out (5- 7). However, whether Lp-PLA₂ is causal and whether its inhibition will prevent CHD events remain undetermined.

Data for Lp-PLA₂ in human atherosclerosis remain indirect and confounded by species differences in physiology and actions. Lp-PLA₂ circulates in blood bound to lipoproteins that modulate its actions. In rodents, Lp-PLA₂ is carried mostly on high-density lipoprotein particles, whereas in humans, the enzyme is bound to LDL particles. Thus, confounding may be particularly marked for plasma Lp-PLA₂ relative to other inflammatory markers, as regulation of atherogenic lipoproteins is a major influence on circulating Lp-PLA₂ levels and activity (8). Indeed, whether circulating Lp-PLA₂ is associated with CHD beyond a complete assessment of atherogenic lipoproteins remains uncertain (9).

Arterial Lp-PLA₂ biosynthesis by macrophages and foam cells, rather than circulating levels or activity, may determine its atherogenicity (10). Lp-PLA₂ expression within the necrotic core and surrounding macrophages of vulnerable and ruptured plaques, compared with less-advanced lesions, is increased (11), suggesting a potential role in promoting plaque instability. The extent to which human Lp-PLA₂ is regulated in circulation by systemic inflammation, however, versus locally controlled in arterial macrophage-foam cells is uncertain. Further, lesion biosynthesis is difficult to measure in humans, limiting our ability to monitor Lp-PLA₂ activity in disease-relevant tissue and to assess vascular efficacy of pharmacological inhibition.

In this report, we examined inflammatory regulation of circulating Lp-PLA₂ during experimental endotoxemia in humans, probed the source of Lp-PLA₂ in human leukocytes under inflammatory conditions, and determined the relationship of genetic variation in phospholipase A₂, group VII (PLA2G7), the gene encoding Lp-PLA₂, to coronary artery calcification (CAC) as well as plasma levels of Lp-PLA₂ mass and activity. We found that, unlike blood tumor necrosis factor (TNF)-alpha and C-reactive protein (CRP), circulating Lp-PLA₂ did not increase during the acute phase response in humans; that inflammatory macrophages and foam cells, but not circulating or ex vivo monocytes, are primary leukocyte sources of Lp-PLA₂; and that common genetic variation in PLA2G7 is associated with subclinical coronary atherosclerosis. These data link Lp-PLA₂ to atherosclerosis in humans while providing a human physiological context for the difficulty in using circulating Lp-PLA₂ as a biomarker of disease or pharmacological efficacy in atherosclerosis.

Healthy volunteers on no medications and no significant medical history (n = 32, 50% women; mean age 25.7 ± 3.90 years) were studied as described previously (12- 13) and in the Online Appendix. Serial blood samples were collected before and after intravenous bolus infusion of 3 ng/kg U.S. standard reference endotoxin and were prepared for plasma, whole-blood ribonucleic acid (RNA) and monocyte RNA (12).

The PennCAC (Penn Coronary Artery Calcification) resource included European-ancestry subjects recruited to 3 separate studies at the University of Pennsylvania: the SIRCA (Study of Inherited Risk of Coronary Atherosclerosis) (n = 799), the PDHS (Penn Diabetes Heart Study) (n = 782), and the PAMSyN (Philadelphia Area Metabolic Syndrome Network) (n = 480). These studies are described in detail previously (14- 15) and in the Online Appendix. In each study, subjects with clinical atherosclerotic cardiovascular disease were excluded. PLA2G7 single nucleotide polymorphisms (SNPs) were genotyped in all 3 studies. Plasma Lp-PLA₂ mass and activity data were available in SIRCA and PDHS. Global CAC scores were determined by electron beam tomography (Imatron, San Francisco, California) according to the method of Agatston et al. (16). For all human studies described, the University of Pennsylvania Institutional Review Board approved each study, and written informed consent was provided by all participants.

Human monocyte isolation, macrophage (M1 and M2 phenotype) differentiation (17), and “foam cell” preparation were performed as described elsewhere (12) and in the Online Appendix. Experiments were performed in batches using freshly isolated monocytes, macrophages, and foam cells derived from the same human volunteer.

Plasma and cell-media levels of Lp-PLA₂ mass and activity, TNF-alpha, and CRP, as well as lipid and biochemical markers were measured as described elsewhere (13- 14,18) and in the Online Appendix.

Whole-blood, isolated circulating monocyte, and human cultured monocyte, macrophage, and foam cell messenger ribonucleic acid (mRNA) were subjected to quantitative polymerase chain reaction (PCR) using primers and probes (Applied Biosystems 7300 Real-Time PCR System, Foster City, California) as previously described (12) for measurement of Lp-PLA₂, TNF-alpha, and beta-actin mRNA (Online Appendix). The relative quantitation 2^-(ΔΔCt) method was used to determine fold-change from baseline (19). Exploratory expression quantitative trait locus (eQTL) analysis is described in the Online Appendix.

As described previously (20) and in the Online Appendix, PennCAC participants were genotyped using the ITMAT Broad Care CARDIOVASCULAR DISEASE candidate gene array, which surveys ∼50,000 SNPs in ∼2,000 candidate genes (21). SNP data for PLA2G7 (n = 19) and CRP (n = 16) were selected for the present analysis.

The effect of endotoxemia on plasma Lp-PLA₂ mass and activity, TNF-alpha, and CRP, as well as whole blood and monocyte mRNA was tested by repeated-measures analysis of variance. Analysis of variance was applied also to in vitro cell data. Post hoc t tests were used to compare specific time points and treatments. We observed heterogeneity of variance in several variables following lipopolysaccharide (LPS) challenge, which was to be expected given the known variation in responses to endotoxin. We tested for homogeneity of variance using Levene test, and in cases where the assumption of homogeneity of variance was violated, we confirmed whether the group differences were significant using Tamhane post-hoc test.

In PennCAC, CAC scores were transformed by the natural log after adding 1, [Ln(CAC + 1)], to correct for skewed distribution. This variable was used as the outcome in a linear regression model, with PLA2G7 and CRP SNPs, adjusting for age, sex, and age–sex interaction. For linear regression analysis of SNP associations with plasma proteins, Lp-PLA₂ mass and activity were normally distributed and, therefore, used as outcomes, whereas CRP was log-transformed. The linear regression model included adjustments for age, sex, and smoking. Analysis used PLINK (version 1.06, Shaun Purcell, Boston, Massachusetts). Analyses of CAC and plasma proteins were performed separately in each sample and then subjected to meta-analysis. Meta-analysis applied a weighted Z-score method using METAL (Center for Statistical Genetics, University of Michigan, Ann Arbor, Michigan) (22- 23) as we applied it in (24) as described in the Online Appendix. In analysis of SNP data, we corrected for the number of independent tests within each gene (10 tests for 19 PLA2G7 SNPs, unadjusted p value threshold of 0.005, and 15 tests for 15 CRP SNPs, unadjusted p value threshold of 0.0033) using the method of Nyholt (25).

As we described previously (13,26), endotoxemia produced an acute, febrile illness associated with a marked, transient induction of plasma TNF-alpha (p < 0.001), followed by a delayed ∼100-fold induction of plasma CRP at 24 h (p < 0.001) (Figure 1A). In contrast, plasma Lp-PLA₂ mass and activity did not increase following LPS (Figure 1B). Indeed, levels of Lp-PLA₂ mass tended to decline (by 18% at 6 h, p < 0.01). The mRNA response to LPS in whole blood for TNF-alpha (Figure 1C) and Lp-PLA₂ (Figure 1D), as well as in circulating monocytes for TNF-alpha (Figure 1E) and Lp-PLA₂ (Figure 1F), was similar to that of plasma proteins. The mRNA levels of Lp-PLA₂ in circulating monocytes were low but detectable (baseline cycle threshold [CTs] ∼30, varying from CTs of 28 to 32 post-LPS).

Lp-PLA₂ mRNA levels were low (CTs ∼30) in freshly isolated human monocytes but increased markedly (CTs ∼20) following 6 days of differentiation to mature macrophages (p < 0.0001) (Figure 2A) and increased modestly during further polarization to M1 (p < 0.0001) but not M2 macrophages (Figure 2B). Lp-PLA₂ protein mass also was induced during differentiation to macrophages, with increases in both the cell-associated protein (p < 0.0001) and the secreted protein (p = 0.0004) (Figure 2C). Following loading of human monocyte-derived macrophages with acetylated LDL-cholesterol for 48 h, cholesterol ester (128 vs. 0.6 μg cholesterol ester /mg protein) and total cholesterol (422 vs. 316 μg cholesterol/mg protein) were significantly higher in loaded versus unloaded cells, which is consistent with findings for in vivo foam cells (27). Lp-PLA₂ mRNA levels were significantly greater in foam cells than in mature macrophages (p < 0.01) (Figure 3A). Similarly, cell-associated (p = 0.05) and secreted (p = 0.008) Lp-PLA₂ protein levels were higher in foam cells than in macrophages (Figure 3B). There was no Lp-PLA₂ protein detectable in the media or acetylated LDL used to treat cells. Overall, these data are consistent with lack of in vivo increase in plasma or monocyte levels of Lp-PLA₂ during the acute phase and suggest that, in human atherosclerosis, Lp-PLA₂ may be generated by macrophages and foam cells rather than by circulating leukocytes.

Exploratory interrogation of PLA2G7 SNP eQTLs for Lp-PLA₂ mRNA expression in publicly available data revealed nominal associations of several SNPs in the PLA2G7 region with exon probe levels in peripheral blood mononuclear cells (best p = 0.0059, rs12181971) and brain (best p = 0.008, rs12195701) (28), as well as skin (best p = 0.021, rs16874962), fat (best p = 0.019, rs16874962), and lymphoblastoid cells (best p = 0.037, rs7745519) (MuTHER [Multiple Tissue Human Expression Resource] twin2 study (29)). These modest associations, however, were not significant after correction for multiple testing (see the Online Appendix). Macrophage and foam cell expression datasets were not available for testing a more atherosclerosis-relevant cell type.

Individually in SIRCA or PDHS samples, there were no significant associations between PLA2G7 SNPs and Lp-PLA₂ mass or activity. In the combined meta-analysis, only 1 SNP (rs1805017) had nominal association with Lp-PLA₂ mass (p = 0.02; p = 0.2 after Bonferroni correction) (Table 1A). As a positive control, we performed similar analysis of plasma CRP using common CRP SNPs. In contrast to findings for PLA2G7-Lp-PLA₂, there were significant associations between numerous CRP SNPs and plasma CRP in the SIRCA and PDHS samples and in the overall meta-analysis (Table 1B). Of 16 CRP SNPs, 9 had nominal (p < 0.05) associations with CRP levels and 8 of these SNPs had significant associations after Bonferroni correction.

Association of PLA2G7 SNPs with CAC was assessed initially in SIRCA and PAMSyN together with follow-up in PDHS. Multiple PLA2G7 SNPs had nominal associations with CAC in SIRCA and PAMSyN (11 with p < 0.05; lowest p < 0.0001 for rs1421378). Replication signals in PDHS were modest (strongest rs10948300, p = 0.02) likely due to limited power; however, in PDHS, 16 of 19 SNPs had effects in the same direction as in SIRCA or PAMSyN (chi squared: 8.9, p = 0.003). Meta-analysis of the combined sample found several SNP associations with CAC (rs9349373, p = 0.002; rs2216465, p = 0.002; rs12195701, p = 0.004) that were significant after Bonferroni correction (Table 2A). Including plasma Lp-PLA₂ mass or activity in the model did not attenuate the association between PLA2G7 SNPs and CAC. These findings support recent associations of variation in PLA2G7 with CHD (30- 32).

As an expected negative control (14,33- 34), we examined CRP variant associations with CAC in the same sample and found minimal signal, with 1 SNP having nominal association in SIRCA and PAMSyN (rs3093068, p = 0.04); however, there were no associations in PDHS nor in combined meta-analysis.

We provide novel insight into the pathophysiology of Lp-PLA₂ in humans. First, we show that unlike TNF-alpha and CRP, circulating Lp-PLA₂ does not increase during experimental endotoxemia and, therefore, does not contribute to human acute phase response. Second, we found that inflammatory macrophages and foam cells, but not circulating monocytes or cultured primary monocytes, generate significant Lp-PLA₂. This is consistent with the concept that the majority of Lp-PLA₂ in atherosclerotic plaque is derived from local biosynthesis by inflammatory macrophage and foam cells rather than from circulating leukocytes. Third, we found that common variants in PLA2G7 are associated with CAC but had limited relation to circulating Lp-PLA₂ mass or activity. This supports an atherogenic role for PLA2G7-Lp-PLA₂ in humans that may be independent of circulating Lp-PLA₂ mass or activity.

We demonstrate that Lp-PLA₂ is not an acute phase protein in humans. This is in contrast to rodent models in which LPS challenge was shown to induce a rapid increase in plasma and tissue levels of Lp-PLA₂ (35). This provides further evidence of fundamental differences between humans and rodents in the physiology and action of Lp-PLA₂ (36). Lack of induction in blood and circulating monocytes by endotoxemia in vivo also suggests limited, if any, role for circulating leukocyte production of Lp-PLA₂ in atherosclerosis. In contrast, marked in vitro up-regulation in macrophages and foam cells is consistent with a specific role for local vascular production of Lp-PLA₂ in atherosclerosis. Although it is possible that local macrophage Lp-PLA₂ production in plaque may contribute to a portion of circulating Lp-PLA₂, it is unlikely to render circulating levels useful as independent biomarkers of Lp-PLA₂ actions in atherosclerosis because published data show that circulating Lp-PLA₂ mass and activity do not correlate with plaque Lp-PLA₂ in patients undergoing elective carotid endarterectomy (37) and because there is substantial confounding of plasma Lp-PLA₂ by circulating lipoproteins regardless of tissue source. Overall, these data suggest that levels of Lp-PLA₂ mRNA and protein in blood may be poor surrogates of PLA2G7 actions in arterial plaque.

We found that Lp-PLA₂ expression was markedly increased during the differentiation of monocytes to macrophages, and further induced in vitro in “foam cell–like macrophages. This is consistent with constitutive expression and activity in inflammatory macrophages (38) and foam cells in atherosclerosis. Indeed, Lp-PLA₂ expression is increased in atherosclerotic lesions in humans (10). In this environment, secreted Lp-PLA₂ can hydrolyze oxidized phospholipids and fatty acids on atherogenic lipoproteins, generating reactive lipid mediators thought to promote plaque instability. Inhibition of Lp-PLA₂ suppressed oxidized-LDL–induced macrophage apoptosis (39), a feature of inflammatory plaque. Further, in a porcine model of complex atherosclerosis, suppression of Lp-PLA₂ retarded atherosclerosis progression and decreased plaque inflammation, necrosis, and fibrous cap erosion (7). Compared with placebo, short-term Lp-PLA₂ inhibition in humans also reduced several markers of plaque inflammation in carotid lesions examined ex vivo (5- 6). Overall, these data provide indirect evidence for atherogenic actions of Lp-PLA₂ in vascular lesions. Indeed, Lp-PLA₂ inhibition is currently being tested in large phase III clinical trials of CHD in high-risk patients (NCT01000727).

Several epidemiological studies revealed an association of higher plasma Lp-PLA₂ mass and activity levels with risk of CHD (9,40- 42). Meta-analyses support a modest CHD relationship independent of traditional risk factors and plasma CRP (40,43- 44). Published studies, however, may underestimate the degree of confounding because of incomplete measurement and control for all atherogenic lipoproteins (9). In circulation, Lp-PLA₂ associates with apolipoprotein B and high-density lipoproteins, with the majority found on LDL particles. Because Lp-PLA₂ protein and activity are closely linked to circulating apolipoprotein B lipoproteins (36,45), it is not surprising that genetic factors (e.g., APOC1, PSRC1, ZNF259) that regulate plasma apolipoprotein B lipoproteins are also associated with plasma Lp-PLA₂ (46). Parenthetically, we found modest association of lipid-related genes (e.g., LRP2, LPL, APOA2) with plasma Lp-PLA₂ likely reflecting this indirect post-translational influence (6). Interpretation of studies of plasma Lp-PLA₂ in CHD is challenging partly because circulating lipoproteins may grossly confound the association of plasma Lp-PLA₂ with CHD (8) and furthermore because lesion macrophage production may be more relevant to the disease than circulating protein.

While we failed to detect significant association between plasma Lp-PLA₂ and common SNPs in PLA2G7, the same PLA2G7 variants were associated with CAC within our study samples. Our preliminary exploration also revealed only nominal associations of PLA2G7 SNPs with Lp-PLA₂ mRNA levels in multiple cells and tissues. These eQTL findings should be interpreted cautiously because of limited power, relatively low levels of Lp-PLA₂ expression in tested cells, and (unlike CRP) well-characterized cis-acting SNPs for PLA2G7 are lacking. Further, appropriately powered studies are needed to determine whether PLA2G7 SNPs are related to expression of Lp-PLA₂ in inflammatory macrophages and foam cells, sources that may be most relevant to atherosclerosis. However, our data suggest caution in using circulating leukocyte Lp-PLA₂ mRNA levels as surrogates for effects of PLA2G7 variation on arterial pathology. Overall, our findings support the concept that PLA2G7 may relate to atherosclerosis independent of circulating Lp-PLA₂ mRNA and protein.

Published studies of PLA2G7 in CHD are conflicting. In a meta-analysis of individuals of European ancestry, PLA2G7 SNPs did not associate with risk of CHD (n ∼5,000) (42), although there were relationships between Lp-PLA₂ activity and CHD and between PLA2G7 SNPs and Lp-PLA₂ activity. However in a meta-analysis of over 13,000 Asians, a common nonsynonymous PLA2G7 SNP showed evidence of association with CHD (31). Additional nonsynonymous SNPs have been associated with carotid plaque in Japanese (32) and recently a loss-of-function variant in PLA2G7 was shown to protect against CHD in Koreans (30). Due to the absence in Caucasian samples of the functional PLA2G7 SNP found in Asians (rs76863441 or V279F), we were not able to evaluate the effect of this functional variant in our samples. However, common variation in PLA2G7 is well covered on the ITMAT Broad Care array platform (tag SNP-coverage r² >0.8 for alleles with minor allele frequency ≥2% in the gene ± 5 kilobases) (21). Therefore, we are confident that we achieved excellent coverage of common variation in this gene region in Caucasians. Whereas ethnic difference in the presence of allelic variation may exist, most published data suggest a relationship of PLA2G7 with clinical CHD supporting our CAC findings.

Findings for CRP in our samples are consistent with published data and contrast with that observed for PLA2G7-Lp-PLA₂. Thus, even though a number of SNPs in CRP had strong associations with circulating CRP levels, there was no relationship between these same SNPs and CAC. These data are in line with hallmark Mendelian randomization studies of clinical CHD outcomes (33- 34) and support a model of confounding or reverse causation for CRP associations with CAC and CHD.

First, our studies are correlative and do not define causality. We have not studied loss-of-function or gain-of-function variants in PLA2G7 for their relation to CAC or CHD and, therefore, cannot infer Lp-PLA₂ directional actions in atherosclerosis. However, expression data in inflammatory macrophages and foam cells coupled to preliminary studies of Lp-PLA₂ inhibition in human atherosclerosis support an atherogenic role for human PLA2G7. Second, recent studies have shown stronger associations of PLA2G7 with circulating Lp-PLA₂ measures than were found in our sample. This may relate to our smaller sample size, heterogeneity in the SIRCA and PDHS study samples, or differences in Lp-PLA₂ assays used across studies. The PLA2G7-Lp-PLA₂ system, however, may be a poor target for Mendelian randomization studies for several reasons including: heterogeneous environmental and genetic influences on circulating levels; PLA2G7 actions in atherosclerosis are likely to be independent of circulating Lp-PLA₂; and well-characterized cis-acting SNPs to use as instrumental variables for PLA2G7 are lacking. Finally, although not a direct measure of coronary atherosclerosis, studies show that CAC provides a quantitative estimate of coronary atherosclerosis (47) and is a useful predictor of CHD events (48).

We have demonstrated that Lp-PLA₂, in contrast to CRP, is not an acute phase protein in humans. Lp-PLA₂ has limited expression in circulating leukocytes or unstimulated monocytes ex vivo, but it is induced during differentiation to macrophages and in foam cells. Thus, robust biomarkers of Lp-PLA₂ action in atherosclerosis and of its pharmacological modulation in vascular tissues are lacking. Common variation in PLA2G7, but not in CRP, is related to the burden of CAC, suggesting that PLA2G7 may indeed modulate human atherosclerosis. Our data provide support for the atherogenicity of Lp-PLA₂ in humans while highlighting the challenges in using plasma Lp-PLA₂ as a biomarker of CHD and in determining drug-dosing and therapeutic efficacy in atherosclerosis.

For supplementary methods, please see the online version of this paper.