Report

Transcriptional Repression of Atherogenic Inflammation: Modulation by PPARδ

See allHide authors and affiliations

Science  17 Oct 2003:
Vol. 302, Issue 5644, pp. 453-457
DOI: 10.1126/science.1087344

This article has a correction. Please see:

Abstract

The formation of an atherosclerotic lesion is mediated by lipid-laden macrophages (foam cells), which also establish chronic inflammation associated with lesion progression. The peroxisome proliferator-activated receptor (PPAR) γ promotes lipid uptake and efflux in these atherogenic cells. In contrast, we found that the closely related receptor PPARδ controls the inflammatory status of the macrophage. Deletion of PPARδ from foam cells increased the availability of inflammatory suppressors, which in turn reduced atherosclerotic lesion area by more than 50%. We propose an unconventional ligand-dependent transcriptional pathway in which PPARδ controls an inflammatory switch through its association and disassociation with transcriptional repressors. PPARδ and its ligands may thus serve as therapeutic targets to attenuate inflammation and slow the progression of atherosclerosis.

Lipid-accumulating macrophages, or foam cells, are the major component of the atherogenic lesion. Loss of either of the nuclear receptors PPARγ or liver X receptor (LXR) α in the macrophage disables lipid export and accelerates progression of the atherosclerotic lesion (13), establishing a transcriptional basis for lipid homeostasis within coronary arteries (4, 5).

Synthetic ligands of the nuclear receptor PPARδ (also called PPARβ) modulate lipid transport in monocytic cell lines through regulated expression of the scavenger receptor CD36 and the efflux pump ABCA1 (6, 7). By analogy with PPARγ, we hypothesized that loss of PPARδ in macrophages may also disturb the dynamic balance of lipid uptake and efflux, thereby aggravating disease progression. Because mice lacking the low-density lipoprotein receptor (LDLR–/– mice) readily develop atherosclerosis when fed a high-fat diet, the contribution of PPARδ to cardiovascular disease can be evaluated in this mouse model by transplantation with PPARδ–/– bone marrow.

We divided γ-irradiated LDLR–/– mice (C57/BL6 background) into three groups and then performed bone marrow transplants from wild-type C57/BL6, wild-type SV129, or PPARδ–/– mice [high percentage chimera, SV129 background, >95% by fluorescence-activated cell sorting (8)]. The mice were then placed on an atherogenic diet for 8 weeks and quantitatively analyzed for lesion progression. Plasma total cholesterol levels were unchanged between the PPARδ–/– bone marrow transplant (PPARδ–/– BMT) mice and the two control BMT groups before or after the diet (fig. S1A). Similarly, lipid profiling revealed no difference in levels of very low density lipoprotein (VLDL), LDL, or high-density lipoprotein (HDL) between groups (8). However, analyses of the aortic valve sections revealed a drastic reduction in fatty streak in PPARδ–/– BMT mice (Fig. 1, A and B). The average lesion area was 62% and 51% less in PPARδ BMT mice than in the control BMT groups, respectively (Fig. 1C), making this one of the most striking protective effects among the nuclear receptor family described.

Fig. 1.

PPARδ–/– bone marrow transplant suppresses atherosclerosis in the LDLR–/– mouse model. (A and B) Representative Oil Red O–stained sections of aortic valve from each group (magnification, 40×). Wild-type (WT) BMT mice (A) show more advanced, larger lesions than do PPARδ–/– BMT mice (B). (C) The average lesion area was similar between C57/BL6 and SV129 control groups (P < 0.08, Mann-Whitney U test), whereas the PPARδ–/– group lesion area was significantly reduced (P < 0.0003). Lesion area means ± SD: C57/BL6, 113,347 ± 23,094 μm2 (n = 8); SV129, 88,490 ± 27,585 μm2 (n = 10); PPARδ–/–, 43,609 ± 14,723 μm2 (n = 11). (D) PPARδ does not regulate genes involved in lipid uptake and efflux in the macrophage. Bone marrow–derived macrophages were treated with vehicle, PPARδ synthetic ligand GW501516 (GW, 0.1 μM), or LXR ligand 22-ox (5 μM) for 24 hours. Total RNA was harvested and analyzed for gene expression on Northern blots. ABCA1, an LXR target gene, was induced by 22-ox in both WT and PPARδ–/– cells. Actin probe is included as a loading control. (E) PPARδ–/– cells exhibit a normal cholesterol efflux rate. SEMs from triplicate experiments are shown. (F) Lipids were extracted from macrophages with or without ox-LDL loading (25 μg/ml, 24 hours). Cholesterol contents were determined with an enzymatic analysis kit.

To determine whether the reduced lesion area in PPARδ–/– BMT mice was due to an intrinsic inability of the macrophage to accumulate lipid, we examined macrophages derived from wild-type and PPARδ–/– bone marrow. Wild-type and PPARδ–/– cells expressed similar levels of scavenger receptors CD36 and SR-A and of endosomal cholesterol carrier HE1 (also called NPC2). Neither the expression of the ABCA1 transporter, which mediates reverse HDL transport, nor its activation by the LXR ligand 22-hydroxycholesterol (22-ox) was affected in the PPARδ–/– cells (Fig. 1D). In contrast, ligand-induced activation of the direct target gene, ADRP (adipose differentiation-related protein) (9), was abolished in the PPARδ–/– macrophage. Interestingly, the basal level of ADRP was elevated in the null cells (Figs. 1D and 2A), suggesting a repression function for the unliganded receptor. In addition, cholesterol efflux mediated by Apo AI, the major cholesterol acceptor in HDL, was comparable between wild-type and PPARδ–/– cells in the absence or presence of the PPARδ synthetic agonist GW501516 (Fig. 1E). As a positive control, the LXR-specific ligands 22-ox or synthetic compound T0901317 (8) increased the efflux rate. Similarly, internalization of ox-LDL— which promotes macrophage inflammatory response, lipid accumulation, and lesion progression—induced cholesterol ester accumulation independent of PPARδ or GW501516 (Fig. 1F) (fig. S1B). Collectively, these results suggest that PPARδ does not appear to appreciably modulate either cholesterol uptake or efflux, and thus the observed lesion reduction in PPARδ–/– BMT mice may not be a direct consequence of defects in cholesterol homeostasis.

Fig. 2.

PPARδ–/– macrophages express lower levels of proinflammatory genes. (A and B) Northern blot analyses of peritoneal macrophages. Cells were treated with (+) or without (–) GW501516 (GW, 0.1 μM) orox-LDL (25 μg/ml) for 24 hours. Actin served as a loading control. (C) Northern blot analysis of RNAs from control RAW cells [expressing yellow fluorescent protein (YFP)] or cells stably expressing PPARδ. Overexpression of PPARδ increases expression of proinflammatory genes, which could be reversed by ligand treatment. (D and E) Magnified (250×) images of aortic valve sections from WT and PPARδ–/– BMT mice, respectively, stained with an antibody to mouse MCP-1 protein. The intensity of the red color corresponds to the expression level. Macrophages were identified by staining with MOMA-2, a cell surface marker (8).

In addition to accumulating lipid, macrophages are attracted to sites of tissue injury such as atherosclerotic lesions, where they elicit a chronic inflammatory response. To examine the inflammatory status of PPARδ–/– macrophages, we examined the expression profile of proinflammatory genes in peritoneal macrophages (elicited by inflammatory agent thioglycollate). Northern blot analysis revealed that levels of monocyte chemoattractant protein MCP-1, cytokine IL-1β (interleukin-1β), and matrix metalloproteinase MMP-9 were down-regulated in the PPARδ–/– macrophage with or without ox-LDL stimulation (Fig. 2, A and B) (fig. S2A). Gene targeting studies of MCP-1 or its receptor CCR2 have shown that loss of either gene results in a 50 to 80% lesion reduction (1013). In addition, both IL-1β and MMP-9 have been directly linked to lesion progression (14). However, the expression of other proinflammatory genes [such as those for tumor necrosis factor (TNF) α and IκB kinase (IKK) β] and of the transcriptional repressor BCL-6, which suppresses the production of multiple cytokines and chemokines including MCP-1, MCP-3, and macrophage inflammatory protein (MIP) 1β (15, 16), was similar between wild-type and PPARδ–/– macrophages. A mouse macrophage cell line (RAW 264.7) stably overexpressing PPARδ expressed higher levels of MCP-1, IL-1β, and MMP-9 relative to control cells (Fig. 2C). However, as seen in the peritoneal macrophage, TNFα, IKKβ, and BCL-6 expression remained unchanged. Overexpression of PPARγ in these cells had no effect on these target genes (8), consistent with the competence of PPARγ–/– macrophages to respond to inflammatory stimuli (17). Immunohistochemical staining of the aortic valve section confirmed that PPARδ–/– macrophages express a lower level of MCP-1 relative to wild-type macrophages (Fig. 2, D and E). These results suggest that MCP-1, IL-1β, and MMP-9 are critical PPARδ targets that control macrophage inflammation and are associated with lesion reduction in the PPARδ–/– BMT mice.

Although results from both PPARδ–/– and PPARδ-overexpressing macrophages suggest a proinflammatory role for PPARδ, treatment of cells with GW501516 suppressed the expression of MCP-1 and IL-1β in a receptor-dependent manner, indicating that activation of PPARδ is anti-inflammatory (Fig. 2, A and C). Moreover, MCP-1 expression in response to cytokine M-CSF (macrophage colony-stimulating factor) was compromised in PPARδ null cells (as was also true for IL-1β, fig. S3A), and the addition of ligand in wild-type macrophages also reduced expression to levels similar to those of PPARδ–/– cells (Fig. 3A). We also found that overexpression of PPARδ (mediated by adenovirus) increased MCP-1 levels in the primary macrophage in a dose-dependent manner; this was completely reversed by ligand (Fig. 3B), which suggests that the amount of unliganded receptor determines MCP-1 expression.

Fig. 3.

PPARδ exerts both pro- and anti-inflammatory activity. (A) A compromised response to M-CSF stimulation in PPARδ–/– macrophages. Bone marrow–derived macrophages were withdrawn from differentiation medium for 1 day and restimulated with M-CSF (10 ng/ml) in the absence (–) orpresence (+) of ligand GW501516 (GW) for 24 hours and analyzed for MCP-1 expression by Northern blots. (B) The pro- and anti-inflammatory activity of apo- and holo-PPARδ. WT macrophages were transiently infected with adenovirus expressing YFP control (CT) or PPARδ. GW501516 (GW) was added 2 days after infection and RNAs were collected for Northern analyses. Viral-PPARδ transcript migrates at a lower position because it contains only the coding region. Endogenous PPARδ transcript and actin (8) were used for the loading control. (C) Loss of ligand-mediated anti-inflammatory activity in the PPARδΔAF2 mutant. RNAs from RAW cells stably expressing PPARδ or PPARδΔAF2 mutant in the absence (–) orpresence (+) of ligand (GW) were analyzed for candidate gene expression by Northern blot. (D) Analyses of the PPARδ sequestering effect on mouse MCP-1 promoter in RAW cells. RAW cells were transfected with a luciferase reporter containing a 2.8-kb MCP-1 promoter and expression vectors encoding either individual GAL4-PPAR LBD fusion protein or full-length PPARδΔAF2 together with the heterodimer partner retinoid X receptor (RXR), along with a β-galactosidase internal control. The concentration for PPARα ligand Wy14,643 was 30 μM and PPARγ ligand rosiglitazone (BRL) was 1 μM. Vehicle, dimethyl sulfoxide control; RLU, relative luciferase unit.

We could not identify a functional PPARδ response element within a 2.8-kb 5′ regulatory region of the MCP-1 promoter, implying an indirect mechanism. As one plausible explanation, apo-PPARδ could function as an MCP-1 activator by sequestering a negative regulator. Ligand activation could release the regulator, which would then repress MCP-1 expression. To test this hypothesis, we examined how expression of a mutant PPARδ receptor (PPARδΔAF2) that is deficient in the ligand-induced corepressor release affects MCP-1 expression. In principle, this mutant should bind the repressor but not release it in the presence of ligand. As expected, PPARδΔAF2 is a potent inducer of MCP-1 but no longer shows ligand-induced suppression when expressed in RAW cells (Fig. 3C). As a control, GW501516 induced ADRP expression in cells expressing PPARδ but not PPARδΔAF2. A survey of chemokine expression revealed that MCP-3 and MIP-1β are also regulated by PPARδ in a similar manner (Fig. 3, B and C). To determine whether this sequestering activity of PPARδ is independent of DNA binding, we transfected RAW cells to express a fusion protein of the GAL4-PPARδ ligand-binding domain (LBD) and a luciferase reporter containing the MCP-1 promoter. PPARδLBD and PPARδΔAF2, but not PPARαLBD or PPARγLBD, increased reporter activity (Fig. 3D). Addition of ligand suppressed the effect of only PPARδLBD. Similar results were obtained with IL-1β promoter (fig. S3B). These data suggest that ligand activation can trigger an exchange between the corepressor and coactivator association with PPARδ.

Because we observed an inverse correlation in regulating MCP-1, MCP-3, and MIP-1β expression between PPARδ and the transcriptional repressor BCL-6 (16), we examined whether PPARδ may act on BCL-6 either by direct binding or by corepressor competition. Immunoprecipitation of transfected 293 cells (human kidney epithelial cells) expressing epitope-tagged BCL-6 and PPARδ indicated that the proteins interacted in a complex that was disrupted by PPARδ ligand, whereas PPARδΔAF2 maintained interaction with BCL-6 regardless of the ligand status (Fig. 4A). This suggests that BCL-6 and PPARδ associate when PPARδ is in its repressive mode. In contrast, PPARα and PPARγ displayed almost no affinity for BCL-6. Deletion and truncation mutants of BCL-6 were generated to map the BCL-6–PPARδ interaction domain in glutathione S-transferase (GST) pull-down assays. Immobilized GST or GST-PPARδ fusion proteins were incubated with [35S]methionine-labeled BCL-6 truncation mutants and tested for their interactions. The BCL-6 zinc finger domain (amino acids 501 to 706) was both necessary and sufficient for binding to PPARδ (Fig. 4B). Because class II histone deacetylase (HDAC) shares the same interaction domain in BCL-6 (18, 19), we examined whether PPARδ association is competitive or compatible. GST-PPARδ associated with BCL-6 alone as well as in the presence of BCL-6–interacting proteins SMRT and HDAC7 (Fig. 4C), suggesting a potential extensive cross talk between these transcription factors. In addition, BCL-6 overexpression reversed the induction of MCP-1 promoter activity by apo-PPARδ in transfected cells. However, it did not affect PPARδ regulation of the ADRP promoter (Fig. 4, D and E). Thus, BCL-6 may be a key molecular target of PPARδ in regulating macrophage chemokine and cytokine release.

Fig. 4.

(A) Unliganded PPARδ interacts with BCL-6. Coimmunoprecipitation was performed in 293 cells transfected with PPARs (FLAG-tagged) and BCL-6 [hemagglutinin (HA)– and FLAG-tagged] in the absence (–) or presence (+) of ligand. Agarose beads conjugated with antibody to HA were used for immunoprecipitation. Protein interaction was analyzed by Western blots detected with an antibody to FLAG. IP, immunoprecipitation products; Input, 10% of cell lysate used for IP, showing expression levels of PPARs. (B and C) GST pull-down assays. BCL-6, BCL-6 truncation mutations, HDAC7, and SMRT (all in vitro translated and [35S]methionine labeled) were tested for interactions with GST-PPARδ (bacterially expressed). Interaction complex was captured by glutathione-sepharose beads and resolved on an 8% SDS gel. GST alone was included as a negative control. Numbers in (B) correspond to amino acid residues of the human BCL-6 protein (19). Residues 501 to 706 contain the C-terminal zinc finger motif. Input: 10% of the [35S]methionine-labeled product. (D and E) Transient transfection experiments. RAW cells were transfected with a luciferase reporter containing either a MCP-1 or ADRP promoter (9) and expression vectors encoding PPARδ and BCL-6. Overexpression of BCL-6 reversed the induction of MCP-1 promoter activity by the unliganded PPARδ (D) but had no effect on the up-regulation of ADRP promoter by the liganded receptor (E). (F) Model for the differential regulatory network by PPARδ and PPARγ in the macrophage. A PPARγ-LXRα regulatory cascade was previously shown to regulate lipid homeostasis (1). Here, an unconventional transcriptional pathway is proposed in which PPARδ controls inflammation status by its association (proinflammatory) and disassociation (anti-inflammatory) with transcriptional repressors such as BCL-6. These observations suggest a division of labor within the macrophage, where PPARγ and PPARδ provide distinct points of transcriptional control of genes regulating cholesterol homeostasis and inflammation, respectively.

Macrophages perform two important functions, lipid uptake and inflammation, in response to lipid insults at the vessel wall (20, 21). Insight into how members of the PPAR subfamily, which serve as lipid sensors, coordinate with each other to deal with such insults may assist the development of drugs for coronary artery diseases. We propose an unconventional ligand-dependent transcriptional pathway in which PPARδ controls an inflammatory switch by virtue of its association (proinflammatory) and disassociation (anti-inflammatory) with transcriptional repressors (Fig. 4F). This pathway may become particularly important under inflammatory conditions when the corepressor pool is limited, as it has been shown that the inflammatory stimulus IL-1β causes nuclear export of a corepressor complex (22). However, given the complex nature of the atherosclerotic lesion (e.g., the composition of lipids and the involvement of various cytokines, chemokines, and cell types), further investigation is required to explore the contribution of PPARδ/BCL-6 and other potential signaling pathways to the dynamic inflammatory status of the macrophage. Nonetheless, our findings suggest a consideration of PPARδ modulators as a means to attenuate inflammation and slow the progression of atherosclerosis through a bona fide therapeutic pathway.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1087344/DC1

Materials and Methods

Figs. S1 to S3

References and Notes

References and Notes

View Abstract

Navigate This Article