Dual Roles for Lipolysis and Oxidation in Peroxisome Proliferation-Activator Receptor Responses to Electronegative Low Density Lipoprotein*

Low density lipoprotein (LDL) exists in various forms that possess unique characteristics, including particle content and metabolism. One circulating subfraction, electronegative LDL (LDL(–)), which is increased in familial hypercholesterolemia and diabetes, is implicated in accelerated atherosclerosis. Cellular responses to LDL(–) remain poorly described. Here we demonstrate that LDL(–) increases tumor necrosis factor α (TNFα)–induced inflammatory responses through NFκB and AP-1 activation with corresponding increases in vascular cell adhesion molecule-1 (VCAM1) expression. LDL receptor overexpression increased these effects. In contrast, exposing LDL(–) to the key lipolytic enzyme lipoprotein lipase (LPL) reversed these responses, inhibiting VCAM1 below levels seen with TNFα alone. LPL is known to act on lipoproteins to generate endogenous peroxisomal proliferator-activated receptor α (PPARα) ligand, thus limiting inflammation. These responses varied according to the lipoprotein substrate triglyceride content (very low density lipoprotein ≫ LDL > high density lipoprotein). The PPARα activation seen with LDL, however, was disproportionately high. We show here that MUT LDL activates PPARα to an extent proportional to its LDL(–) content. As compared with LDL(–) alone, LPL-treated LDL(–) increased PPARα activation 20-fold in either cell-based transfection or radioligand displacement assays. LPL-treated LDL(–) suppressed NFκB and AP-1 activation, increasing expression of the PPARα target gene IκBα, although only in the genetic presence of PPARα and with intact LPL hydrolysis. Mass spectrometry reveals that LPL-treatment of either LDL or LDL(–) releases hydroxy-octadecadienoic acids (HODEs), potent PPARα activators. These findings suggest LPL-mediated PPARα activation as an alternative catabolic pathway that may limit inflammatory responses to LDL(–).

Extensive data links low density lipoprotein (LDL) 1 to atherosclerosis (1,2). This occurs in part through the induction of early atherogenic inflammatory responses, including the expression of adhesion molecules like VCAM1 by endothelial cells (ECs) (3,4). Consistent with this, increased dietary cholesterol rapidly induces atherosclerosis in animal models, with changes in VCAM1 expression seen within 2 weeks (3). LDL, a major carrier of cholesterol, circulates in several forms in vivo (5,6). Most LDL pathogenicity becomes manifest after LDL oxidation (1). For example, oxidized LDL (oxLDL), but not native LDL, induces endothelial VCAM1 expression in the presence of TNF␣ (7).
Despite some similarities, native LDL, oxLDL, and LDL(Ϫ) have distinct characteristics that likely determine their biologic effects (14). Most fundamentally, these particles have unique compositional profiles. For example, LDL(Ϫ) contains fewer lipid peroxidation products than oxLDL, but more than native LDL (9,14). These forms of LDL are also cleared from the circulation in different ways, potentially contributing to their unique roles in atherosclerosis. In contrast to both native LDL and LDL(Ϫ), which are taken up through the LDL receptor (LDLR) (15), oxLDL is removed after binding to scavenger or Fc receptors (1). Hydrolytic pathways for LDL particles also differ. For example, lipoprotein lipase (LPL), the predominant enzyme in triglyceride-rich lipoprotein (TRL) metabolism, can hydrolyze mildly oxidized LDL forms like LDL(Ϫ) (16), although this ability may be limited as suggested by the higher content of both triglycerides (8) and the LPL inhibitor Apo CIII (17) in LDL(Ϫ) as compared with native LDL fractions.
Recently, we reported LPL enzymatic action as a mechanism for generating endogenous peroxisome proliferator-activated receptor ␣ (PPAR␣) ligands (18). This LPL/PPAR␣ pathway replicated synthetic PPAR␣ agonist effects, e.g. decreasing cytokine-induced VCAM1 expression (18) in vitro and in vivo. The PPAR␣ activation through LPL varied depending on the lipoprotein substrate; it was greatest with VLDL, less with LDL, and minimal with HDL, a series corresponding to triglyceride content (18). Although this pattern and the absolute requirement for intact enzymatic catalysis for LPL-mediated PPAR␣ activation suggests fatty acid release accounted for the responses seen, LPL-treated LDL activated PPAR␣ to a disproportionate extent. This suggested LPL might release different PPAR␣ mediators from LDL as compared with VLDL. If so, this indicates that the transcriptional responses to LDL might vary depending on LPL action, LDL particle composition, or its mechanism of uptake.
To pursue this hypothesis, we tested cellular responses, including PPAR␣ activation, to different forms of native and oxidized LDL with and without LPL treatment. In the presence of TNF␣, LDL(Ϫ) uptake by the LDLR induced VCAM1 through NFB and AP-1 activation, a previously unreported pathogenic LDL(Ϫ) effect. In contrast, LPL treatment of LDL(Ϫ) reversed this response, decreasing VCAM1 expression in a PPAR␣-dependent manner. Further studies reveal that LPL hydrolysis of LDL(Ϫ) generated oxidized linoleic acid (hydroxy-octadecadienoic acid, or HODE) in concentrations likely to account for the PPAR␣ activation and the subsequent antiinflammatory effects.

EXPERIMENTAL PROCEDURES
Reagents-All reagents were purchased from Sigma-Aldrich unless otherwise indicated. All media were obtained from BioWhittaker (Walkersville, Maryland) and contained fungizone/penicillin/streptomycin. Human and murine TNF␣ were purchased from R&D Systems (Minneapolis, Minnesota). Fenofibric acid was a generous gift from Laboratories Fournier (Daix, France).
LDL Isolation-Lipoproteins were isolated using gradient ultracentrifugation of human plasma pooled from at least six healthy donors (9). Plasma of healthy donors was subjected to hemoglobin-mediated oxidation as described (13). The specific form of LDL that is formed by this type of oxidation is referred to as HB LDL. Anion exchange chromatography was utilized to purify the LDL(Ϫ) fraction and measure LDL(Ϫ) content in LDL modified during hemoglobin-mediated plasma oxidation (9).
RNA Analysis-Total cell RNA was isolated using RNeasy kit (Qiagen, Valencia, California), separated in 1% agarose gel, and transferred to Hydrobond membrane (Amersham Biosciences). Northern blotting was performed using cDNA probes obtained from American Type Culture Collection (Manassas, VA).
Enzyme-linked Immunosorbent Assay (ELISA)-ELISA was performed in 96-well plates on confluent human umbilical vein endothelial cell (HUVEC) monolayers (19). Treated cells were kept on ice for 10 min, washed with cold phosphate-buffered saline, incubated with human VCAM1 monoclonal antibodies (gift from Dr. M. Gimbrone), and visualized using alkaline phosphate secondary antibody (450 nm).
Flow Cytometry-Flow cytometry was performed using confluent mouse ECs obtained from heart of PPAR␣ ϩ/ϩ and PPAR␣ Ϫ/Ϫ mice. Cells were washed in phosphate-buffered saline, harvested by trypsinization, and incubated (1 h at 4°C) with fluorescein isothiocyanate-conjugated anti-mouse VCAM1 antibody (BD PharMingen). The EC culture purity was examined using anti-mouse phosphatidylethanolamine-conjugated platelet endothelial cell adhesion molecule 1 (BD PharMingen). Subsequently, washed cells were analyzed in a BD Biosciences FACScan TM flow cytometer using CELLQuest TM software. At least 20,000 viable cells per condition were analyzed.
HPLC-MS-A stock standard solution containing hydroperoxy-octadecadienoic (HpODE) and hydroxy-octadecadienoic acids was prepared by diluting and combining solutions of standard mixtures obtained from Cayman Chemical (Ann Arbor, MI). Samples were extracted by the Folch method, reconstituted in isopropanol, and filtered (0.2-m nylon). The samples and stock standard were serially diluted with 50:50 acetonitrile/H 2 O.
The separation was performed using a Waters 2690 LC with photodiode array (Waters 996) and time-of-flight mass spectrometry (TOF-MS) detection (Micromass LCT). The column used was a Luna C18 (2) 50 ϫ 2 mm, 5 m from Phenomenex (Torrance, CA). Mobile phases for the isocratic separation were 50% A, 4 mM ammonium acetate (Sigma-Aldrich) in H 2 O, and 50% B, acetonitrile (Sigma-Aldrich) flowing at 0.5 ml/min. The separation was performed at 30°C with a total run time of 5 min. UV absorption was acquired from 200 -400 nm. MS was performed using electrospray ionization operating in negative ionization mode. The ionization parameters were as follows: capillary voltage, 3200V; sample cone, 37V; extraction cone, 4V; desolvation temperature, 300°C; source temperature, 120°C; and ion scanning m/z range, 100 -1000. Extracted ion chromatograms were constructed for HODE and HpODE using m/z values of 295 and 293, respectively. When the signalto-noise ratio was sufficient, samples were quantified using an external calibration curve. When the signal-to-noise ratio was too low, only semi-quantitative estimates were made.
Scintillation Proximity Assay (SPA)-A scintillation proximity assay was carried out using human full-length cDNA for PPAR␣, PPAR␦, and PPAR␥ 2 that were subcloned into the pGEX-KT expression vector (22). The 3 H 2 -labeled known synthetic PPAR agonists used were nTZD3 and nTZD4 with relative K d values as follows: nTZD3, PPAR␥ 2.5 nM PPAR␣ 5 nM; and nTZD4, PPAR␦ 1 nM (22). Results are expressed as percent inhibition with a calculated inhibitory constant (K is ).

LPL Decreases LDL(Ϫ)-mediated VCAM-1 Induction in a PPAR␣-dependent Manner-
We compared the effect of native LDL and LDL(Ϫ) on TNF␣-induced VCAM1 expression in human ECs. As expected, TNF␣ induced VCAM1 protein in ECs (8-fold, set as 100% induction) on ELISA (Fig. 1A). Although concomitant treatment with native LDL led to only a modest further increase in VCAM1 (20%), LDL(Ϫ) augmented VCAM1 levels by 70% relative to TNF␣ alone (14-fold as compared with basal untreated levels). The presence of LPL inhibited LDL(Ϫ)mediated VCAM1 induction in a dose-dependent fashion (Fig.  1A). Similar effects were evident on Northern blotting (data not shown) and activation of the human VCAM1 promoter (Fig.  1B). Although LDL(Ϫ) treatment induced the VCAM1 promoter 120 -180% in bovine aortic EC transfections, this same stimulation in the presence of LPL repressed this response by 60 -80% as compared, in both cases, with TNF␣ alone (Fig. 1B). These effects of LPL/LDL(Ϫ) on TNF␣-induced VCAM1 expression equaled those seen with synthetic PPAR␣ ligands (Wy14163 or fenofibric acid, Fig. 1B).
To test whether treatment of LDL(Ϫ) reduced VCAM1 expression in a PPAR␣-dependent manner, VCAM1 responses were examined using FACS analysis of PPAR␣ ϩ/ϩ and PPAR␣ Ϫ/Ϫ microvascular ECs (Fig. 1C). Both TNF␣ and TNF␣/ LDL(Ϫ) stimulation of PPAR␣ ϩ/ϩ ECs markedly increased VCAM1 content on the EC surface. LPL-treated LDL(Ϫ) repressed this TNF␣ induction in the presence, but not the genetic absence, of PPAR␣ with effects replicating those seen with the synthetic PPAR␣ agonist WY14643. These data suggest that LPL action on LDL(Ϫ) limits inflammation through a PPAR␣ mechanism.
LPL Treatment of LDL(Ϫ) Decreases NFB Binding-We and others have shown that PPAR␣ ligands decrease cytokinemediated VCAM1 expression through effects on NFB signaling (19,23). Mechanisms involved reportedly include direct PPAR␣ interaction with p65 and PPAR␣-dependent expression of IB␣, sequestering the inactive p65/p50 complex in the cytoplasm (24). To explore how LPL negatively regulates LDL(Ϫ)/ TNF␣ induction of the VCAM1 promoter, we performed an EMSA of EC nuclear extracts using NFB and PPRE binding sites. As expected, TNF␣ induced NFB binding, a response augmented in the presence of LDL(Ϫ) (Fig. 2A). In the presence of LPL, however, LDL(Ϫ) significantly decreased NFB binding; Wy14163 had similar effects. This decrease in NFB binding was paralleled by increased IB␣ levels from the very same ECs (Fig. 2B). Similarly, the increase in AP-1 binding induced by LDL(Ϫ)/TNF␣ was decreased after LPL treatment (data not shown). In parallel with decreased NFB binding, both LPL/ LDL(Ϫ) and WY14643 markedly increased binding to a canonical PPRE (Fig. 2C). This response was much greater than that seen with LDL(Ϫ), TNF␣, or their combination (Fig. 2C). PPRE binding involved PPAR␣ as indicated by the supershift in the presence of PPAR␣ but not the PPAR␥ antibody. These effects appear to be due to increases in PPAR␣ activators and not PPAR␣ itself, given the lack of significant differences in nuclear PPAR␣ protein levels after treating cells with LDL(Ϫ), LPL, or their combination (Fig. 2D). Together, these data suggest that LPL treatment of LDL(Ϫ) exerts its effects through direct interaction of LDL-derived components with NFB and AP-1 in a manner similar to that of synthetic PPAR␣ ligands.
LPL Lipolysis of LDL(Ϫ) Generates PPAR␣ Ligands-We further examined PPAR␣ ligand generation as a result of LPL treatment of LDL(Ϫ). The yeast Gal4/PPAR LBD hybrid assay is classically used to screen for PPAR␣ ligand formation (25). LDL(Ϫ) stimulation led to a modest 3-fold PPAR␣ LBD activation; in the presence of LPL and LDL(Ϫ), PPAR␣ activation increased 30-fold (Fig. 3A). In addition to its catalytic activity, LPL can also promote the uptake of lipoproteins like LDL through non-enzymatic bridging of lipoproteins to receptors like the LDLR (26). We investigated the contribution of LPL bridging to the effects reported above. Transient transfection of the LDLR into EC did not alter PPAR␣ LBD activation by LDL(Ϫ)/LPL. In contrast, LDL(Ϫ) treatment of LDLR-transfected ECs markedly increased VCAM1 promoter activity (Fig. 3B).
LPL/LDL(Ϫ) mediated PPAR␣-LBD activation required intact LPL catalysis, as evident by the concentration-dependent inhibition of PPAR␣ responses in the presence of the synthetic lipase inhibitor tetrahydrolipstatin or the natural LPL inhibitor ApoCIII (Fig. 3C). Similar repression was seen with an antibody raised against LPL, but not an antibody to the LDLR receptor (data not shown). Furthermore, repeating these PPAR␣ LBD experiments in bovine EC transfected with a catalytically inactive LPL point mutant failed to activate PPAR␣ by LDL(Ϫ) above levels induced by LDL(Ϫ) alone (Fig. 3D).
The generation of direct PPAR␣ ligands by LPL treatment of LDL(Ϫ) was tested further in cell-free PPAR radioligand displacement assays. LPL treatment of LDL(Ϫ) decreased the EC 50 of PPAR␣ activation 20-fold as compared with nontreated LDL(Ϫ) (Fig. 4A). This LPL treatment may have produced greater amounts of the ligands already present in LDL(Ϫ), converted less potent ligands to more potent forms, or both. In contrast to our prior findings with native triglyceriderich lipoprotein (18), LPL/LDL(Ϫ) did not preferentially activate any one PPAR isoform, at least in direct ligand displacement and LBD assays (Fig. 4A).
We next sought to examine how these responses varied depending on the nature of LDL particles, especially in LDL with high LDL(Ϫ) content. In the presence or absence of LPL, LDL(Ϫ) had a much more potent concentration-dependent effect on PPAR␣-LBD activation (Fig. 4B). One fundamental characteristic of LDL(Ϫ) is its higher relative proportion of oxidized lipids as compared with native LDL (14). To examine the contribution of oxidation on PPAR␣ responses, we utilized standard in vitro techniques to increase the proportion of LDL(Ϫ) present in human LDL samples, employing the mild oxidative stress method in which different plasma samples are exposed to hemoglobin under anaerobic conditions (13). The LDL isolated from such plasma has a higher LDL(Ϫ) content, although TG concentrations remain unchanged (13). Using the

FIG. 2. LPL treated LDL(؊) inhibits NFB activation and increases IB␣ expression while inducing PPRE binding.
Confluent human ECs were pretreated with LPL (200 units/ml, 4 h) in the presence or absence of LDL(Ϫ) (10 g/ml) and then stimulated with human TNF␣ (10 ng/ml, 1.5 h). Fenofibric acid (Feno; 100 M) pre-treatment was used for comparison. Nuclear extracts were analyzed by EMSA using 5 g of nuclear extracts and 100 ng of radiolabeled NFB (A), AP-1 (data not shown), and PPRE (C) sequences (Santa Cruz Biotechnology). Supershift analysis was performed to confirm the identity of PPAR␣ and PPAR␥ as well as p65 and p50 (data not shown) using specific antibodies. Cytosolic fractions (25 g protein) of the same cells were analyzed for IB␣ expression (B), whereas nuclear fractions (25 g protein) were studied for PPAR␣ expression (D) using Western blotting. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) expression confirmed equal loading. Data shown are representative of one experiment from three with similar results. same total LDL (10 g) for all treatments, we found that a direct linear relationship existed between PPAR␣-LBD activation and the proportion of LDL(Ϫ) in the absence of LPL. This was evident for both unoxidized (Fig.4B, white circles) and oxidized (Fig. 4B, black circles) plasma samples (Pearson, p Ͻ 0.001, Fig. 4C). Thus, increased LDL(Ϫ) proportions substantially affect PPAR␣ activation in the absence of LPL. In contrast, LPL treatment of these LDL(Ϫ) samples (Fig. 4C, squares) markedly increased PPAR␣ LBD activation regardless of the LDL(Ϫ) content. Taken together, this suggests that degree of LDL oxidation and the extent of LPL-mediated hydrolysis are separate variables influencing PPAR␣ activation. Of note, although LPL-treatment increased the PPAR␣ activation seen with all lipoproteins tested, the greatest induction occurred with LDL(Ϫ) as a substrate. Although TG is the principal LPL substrate, the robust PPAR␣-activation seen with LPL treatment of LDL(Ϫ) as opposed to TG-rich VLDL suggests generation of a potent PPAR␣ ligand.
To identify this PPAR␣-activating component in LDL(Ϫ), hydrolytic products of LPL-treated LDL(Ϫ) were extracted and separated using HPLC-MS. The major lipophylic compounds in LDL(Ϫ), either with or without LPL treatment, were HpODE and HODE acids (Fig. 5A). Even though HODE was present at a negligible amount in untreated LDL(Ϫ), after LPL treatment HODE amounts were ϳ2.5 g per 100 g of LDL(Ϫ). Importantly, LPL treatment of native LDL also increased HODE and HpODE, albeit to a much smaller extent. The identity of compounds was confirmed using MS (HODE and HpODE, Fig. 5, B and C, respectively).
Prior work established HODE as a PPAR␥ activator (27,28), although its affinity may be relatively low (29). Recently, PPAR␣ LBD activation by HODE has been reported (30). To evaluate the extent of PPAR␣ activation by HODE and HpODE, cell-free displacement and cell-based LBD assays were performed. Two common forms of HODE, 9-and 13-HODE, were all effective PPAR␣ ligands. The EC 50 for 13-HODE equaled 10 M, whereas the EC 50 for 9-HODE was 2.6 M in PPAR␣ LBD displacement assays (data not shown). These results suggest that the combined action of hydrolysis and oxidation increases the proportion of HODE to levels sufficient for PPAR␣ activation. DISCUSSION These studies provide evidence that LPL acts on LDL(Ϫ) to generate PPAR␣ ligands, thus countering the endothelial adhesion molecule expression induced by LDL(Ϫ) alone. These data highlight the differing transcriptional responses to LDL that depend on the mechanism of lipoprotein uptake, namely through the LDLR as opposed to hydrolysis by LPL (Fig. 6). Enhanced LDL(Ϫ) uptake through LDLR overexpression further increases VCAM-1 expression and promoter activity through TNF␣-mediated NFB and AP-1 activation. In con- trast, LPL hydrolysis of LDL(Ϫ) limits these responses by generating PPAR␣ ligands, thereby increasing IB␣ expression with subsequent NFB inhibition. Interestingly, LPL hydrolysis of LDL(Ϫ) activates PPAR␣ to a greater extent than either LDL or VLDL. These efficient PPAR␣ responses were proportional to the level of LDL oxidation, particularly to the HODE content of the lipoprotein particles. Thus, both hydrolysis and oxidation appear to influence PPAR␣ activation by LDL(Ϫ), suggesting that responses to this pro-inflammatory, proatherogenic particle may depend on its catabolism. LDL(Ϫ) is a unique lipoprotein (14). It is the only mildly oxidized LDL subfraction in the circulation that undergoes both LPL hydrolysis and LDLR uptake. One recent report suggests that mild oxidation increases LDL susceptibility to LPL-mediated hydrolysis (16), potentially amplifying the responses seen here with LDL(Ϫ). In contrast, oxLDL, which is more extensively oxidized, does not bind to LPL and is taken up via scavenger and Fc receptors in an LPL-independent manner (1,16). Even though oxLDL contains several reported PPAR␣ activators, including oxidized phospholipids (30, 31) and nonesterified fatty acids (such as HODE) (30), PPAR␣ activation in vitro by oxLDL does not decrease inflammation as is charac-teristically seen with synthetic PPAR␣ agonists. OxLDL reportedly increases VCAM1 expression in the presence of TNF␣ (7). This discrepancy may be due to high levels of other lipid peroxidation and decomposition products that may have potent inflammatory and cytotoxic effects, e.g. oxysterols, hydroxynonenal, or lysophosphatidylcholine (32). This last product is a known potent NFB activator (33) that may overcome PPAR␣ activation. Alternatively, responses may vary depending on the pathway of lipoprotein uptake. The effects of oxLDL on PPAR␣ may depend on cytoplasmic phospholipase A 2 action (30); this is not likely relevant to the results presented here, because LPL neither acts on nor binds to oxLDL (16). In our studies, LDLR overexpression significantly increased LDL(Ϫ) activation of the VCAM1 promoter but not PPAR␣ LBD activation. In contrast, LDL(Ϫ) hydrolysis significantly decreased VCAM1 expression in a PPAR␣-dependent manner, with responses equivalent to synthetic PPAR␣ agonists (Fig. 1). Inhibiting LPL hydrolytic activity through expression of a catalytically inactive LPL mutant, the presence of either natural (ApoCIII) or synthetic (tetrahydrolipstatin) LPL inhibitors, or co-stimulation with an LPL antibody prevented LDL(Ϫ)/LPL-mediated PPAR␣ activation (Fig. 3). Together, these findings imply that intact LPL hydrolysis is re- FIG. 4. LPL hydrolysis of LDL generates PPAR ligands in a manner dependent on oxidation. A, lipolysis markedly increases the generation of PPAR ligands from LDL(Ϫ). The most potent response was seen with PPAR␣ and PPAR␦, with less displacement of PPAR␥ ligand. Direct radiolabeled ligand displacement (scintillation proximity assay) PPAR assays were performed using LDL(Ϫ) hydrolyzed with LPL (200 units/ml, 2 h, 37°C under argon). Competition curves were generated across a range of LDL(Ϫ) concentrations (0.003-10 g protein/ml) by incubating the reaction mixture with specific radiolabeled PPAR activators, i.e. 5 nM nTZD3 for GST-hPPAR␣ or GST-hPPAR␥, or 2.5 nM nTZD4 for GST-hPPAR␦ (22,37). Data from duplicate determinations were plotted and K is values (IC50) were obtained from the dose-response curves. The white bars indicate the higher concentration (K is ) of non-hydrolyzed LDL(Ϫ) required for PPAR ligand displacement; black bars reveal the more potent displacement for the same LDL(Ϫ) after LPL hydrolysis. is shown. VLDL, LDL, LDL(Ϫ), and HDL were isolated from human blood, and the plateau of maximum PPAR␣ activation was determined from concentration-dependent curves as shown in panel A. Responses were also compared with HB LDL, an LDL species enriched in LDL(Ϫ) generated in vitro by exposing plasma to hemoglobin (10 M) under anaerobic conditions before lipoprotein isolation (13). oxLDL was obtained after exposing LDL to CuSO 4 (10 M, 24 h), which completely oxidizes polyunsaturated fatty acids (32). quired for PPAR␣ ligand release from LDL(Ϫ) while also supporting the concept that differential lipoprotein catabolism and uptake can influence distal transcriptional responses.
Several prior studies have found that oxidation plays a role in PPAR ligand generation, with certain oxidized molecules, e.g. oxidized phospholipids and certain fatty acids, having greater effects than their native forms (27, 29 -31). To examine this issue, the amount of LDL(Ϫ) in a given LDL sample was increased in a controlled fashion using hemoglobin plasma oxidation in vitro ( HB LDL). Greater PPAR␣ activation occurred with HB LDL stimulation than with native LDL, suggesting that oxidation changes the ligands present in these lipoproteins. In fact, the proportion of LDL(Ϫ) in these samples correlated with PPAR␣ activation in a linear fashion (Fig. 4C). The oxidized molecules likely responsible for these effects have been identified as 9-and 13-HODE, both established PPAR␥ (27) and PPAR␣ activators (30). Such HODEs are released during LDL hydrolysis of LDL(Ϫ) (Fig. 5A). In fact, the PPAR␣ activation seen with LPL-treated LDL(Ϫ) equaled that seen after direct HODE stimulation. The larger proportion of HODE in LDL(Ϫ) as compared with native LDL may account for the greater PPAR␣ activation seen either before or after LPL treatment. Moreover, hydrolysis of native LDL also increases its HODE content, potentially contributing to its LPL-mediated PPAR␣ activation. The possibility that the HODEs liberated from LDL and LDL(Ϫ) undergo further changes intracellularly cannot be excluded. The greater PPAR␣ activation seen after HODE treatment of cell-based LBD assays as compared with cell-free direct displacement assays supports such a notion. Although HODEs seem the most likely candidate for explaining the LDL(Ϫ) responses seen, PPAR␣ activators other than HODEs may be generated by LPL or augmented in its presence.
Previously, the formation of the natural PPAR␣ ligands  6. Transcriptional responses to LDL depend on both the LDL species and its mechanism of uptake. Different LDL species generate unique cellular responses depending on specific mechanisms of metabolism and uptake. Oxidized LDL, taken up by scavenger receptors (SR), promotes inflammatory responses through NFB and AP-1 activation. Here we demonstrate that whereas LDL(Ϫ) uptake through the LDLR induces VCAM1 expression through activation of NFB and AP-1, LDL(Ϫ) hydrolysis by LPL has the opposite effect, decreasing VCAM1 expression by generating PPAR␣ ligands and suppressing NFB and AP-1 responses (7,38,39). These LPL effects on LDL(Ϫ), which are independent of the LDLR, may be especially relevant given the presence of LDL(Ϫ) in the circulation. HETE and LTB 4 during acute inflammatory events was suggested as a feedback loop terminating inflammation and decreasing pro-inflammatory eicosanoids via ␤-oxidation (34). The data presented here suggest another endogenous mechanism through PPAR␣ that may limit inflammatory responses, namely LPL catalysis of LDL(Ϫ). Increased concentrations of LDL(Ϫ) have been associated with pro-inflammatory conditions such as familial hyperlipidemia, diabetes, and hemodialysis, with levels up to 20% of total LDL found in some subjects (10,13). Interestingly, many of these clinical conditions are also characterized by dysfunctional LPL (35,36). One recent study found that only the LDL content of the LPL inhibitor apolipoprotein CIII independently predicted cardiovascular events, whereas total LDL concentration did not (35). Our results suggest that the combination of increased circulating pro-inflammatory LDL(Ϫ) and ineffective LPL action may be a particularly deleterious combination, with the responses to LDL(Ϫ) compounded by decreased PPAR␣ ligand generation. Our prior observations suggested that LPL mediates PPAR␣ ligand generation under physiologic conditions (18); these findings extend lipolytic PPAR activation to limiting pathologic responses through LDL(Ϫ). Such data further support the concept of circulating lipoproteins as a reservoir for PPAR ligands and lipolysis as a means of accessing them.