Free Cholesterol Accumulation in Macrophage Membranes Activates Toll-Like Receptors and p38 Mitogen-Activated Protein Kinase and Induces Cathepsin K

Clinical complications of atherosclerosis usually result from the formation of thrombus on a ruptured or eroded atherosclerotic plaque.¹ Whereas early fatty streak formation involves lipid accumulation in macrophage foam cells in arteries, the growth of plaques and their complications are thought to involve a modified inflammatory response comprising both innate and acquired immune systems.² Nonetheless, LDL cholesterol lowering remains the cornerstone for the treatment of established atherosclerosis and accumulation of cholesterol and oxidized lipids likely continue to drive the inflammatory response even in advanced plaques. The molecular mechanisms linking cellular lipid accumulation and inflammatory gene expression remain poorly understood.

Murine models have been useful for analyzing the role of different genes in plaque development,³ and recently there has been intense interest in the use of Apoe^−/− mice to monitor plaque complications such as intraplaque hemorrhage, or apparent rupture involving the brachiocephalic artery.⁴ We observed a high frequency of medial degradation and atherothrombosis associated with lesions in the proximal aorta of Apoe^−/−, Npc1^−/− mice,⁵ and the incidence was higher in mice in the BALB/c compared with C57BL6 genetic background. The Npc1 mutation results in prominent accumulation of unesterified cholesterol in late endosomes⁶ and peritoneal macrophages from chow-fed Apoe^−/−, Npc1^−/− mice displayed marked accumulation of free cholesterol (FC) and, to a lesser extent, cholesteryl ester.⁵ Using microarrays, we discovered that Apoe^−/−, Npc1^−/− peritoneal macrophages have increased expression of cathepsin K (Ctsk), as well as several matrix metalloproteinase (MMP) genes. Cathepsin and MMPs have important roles in plaque development, medial breakdown and aneurysm formation.^7,8 Cathepsin K protein (CATK) is a potent elastase and Ctsk^−/−, Apoe^−/− mice show reduced lesion area, increased intimal collagen accumulation and decreased medial elastin fiber disruption compared to Apoe^−/− controls,^9,10 suggesting an important role of Ctsk in plaque growth and complications. The present study was undertaken to elucidate the mechanisms linking macrophage cholesterol accumulation to increased expression of Ctsk and various Mmps.

Materials and Methods

Reagents

Aggregated (Agg)LDL and moderately oxidized (Ox)LDL,^11,12 and cyclodextrin–cholesterol complex (CD-chol)¹³ were prepared as described in the online data supplement, available at http://circres.ahajournals.org.

Mice

BALB/cNctr-Npc1^m1N/+/J, C3H/HeJ, Tlr4^del (strain C57BL/10ScNJ), Tlr4^+/+ (strain C57BL/10ScSnJ), Tlr3^−/− (strain 129S1/Sv/C57BL/6), Tlr3^+/+ (strain B6129SF1/J) mice were purchased from The Jackson Laboratory (Bar Harbor, Me). Apoe^−/−, Npc1^−/− mice on BALB/cJ background were bred as described.⁵ Macrophage p38-deficient mice,¹⁴Myd88/Trif^−/− mice in the C57BL/6J background (kindly provided by Dr Shizuo Akira¹⁵), and Mitf^mi/mi mice¹⁶ have been described (online data supplement).

Cell Culture

Peritoneal and bone marrow–derived macrophages were cultured as described.¹⁷ Human THP-1 and mouse RAW264.7 cells were purchased (American Type Culture Collection). Spleen-derived macrophages were derived from myeloid precursor cells from spleen (see the online data supplement).

RNA Analysis

Total RNA was isolated using the RNeasy Mini kit (Qiagen, Valencia, Calif). Real-time quantitative PCR assays were performed as described.⁵

Small Interfering RNA Knockdown

RANK and scrambled small interfering (si)RNAs were purchased from Applied Biosystems; Toll-like receptor (TLR)3, -4, -7, and -8 were from Invitrogen. siRNA was transfected into macrophages with Oligofectamine (Invitrogen).

Immunofluorescent Staining of Aortic Cross-Sections

Paraffin-embedded sections of the proximal aorta were immunostained with affinity-purified rabbit polyclonal CATK antibody or mouse monoclonal α-actin antibody as described¹⁸ (see the online data supplement).

Molecular Imaging

Ex vivo fluorescence reflectance imaging was performed using Image Station 4000MM Pro and Molecular Imaging Software (Kodak) as described.¹⁹

Promoter Luciferase Assay

Human CTSK promoter (−928 to +2 bp) was cloned into a pGL3 vector (Promega) and mutated as described.²⁰CTSK promoter-luciferase and TK-Renilla luciferase were transfected into RAW264.7 cells with Lipofectamine 2000 (Invitrogen). Cells were collected and assayed 24 hours after loading using the Dual-Luciferase Reporter assay system and a TD20/20 luminometer (Promega).

Chromatin Immunoprecipitation Assay

Chromatin immunoprecipitation (ChIP) assays were performed as described (see the online data supplement).¹⁶

Results

We previously found increased Ctsk expression and secretion in Npc1^−/− macrophages and increased elastase activity in aortic homogenates of Apoe^−/−, Npc1^−/− mice.⁵ To confirm in vivo relevance, we used an affinity-purified antibody¹⁸ and found strong CATK immunostaining in the macrophage-rich regions of atherosclerotic plaques in Apoe^−/−, Npc1^−/− mice (Figure 1A, white arrows), but not in Apoe^−/− controls. To confirm an increase in CATK activity, we performed near-infrared fluorescence (NIRF) imaging using peptide substrates that fluoresce following cleavage by cysteinyl proteinases. This procedure revealed a marked increase in signal over the proximal aortic region of Apoe^−/−, Npc1^−/− mice compared to Apoe^−/− controls (Figure 1B and 1C).

• Download figure
• Download PowerPoint

Cholesterol Loading of Endosomes and Plasma Membrane Induces Ctsk Expression

Npc1^−/− macrophages have a defect in the trafficking of cholesterol from late endosomes to the plasma membrane and endoplasmic reticulum (ER) and accumulate FC in late endosomes.²¹ We used different methods of cholesterol loading in concanavalin A (ConA)-elicited Npc1^+/+ peritoneal macrophages to determine whether increased Ctsk expression resulted from decreased trafficking of cholesterol to the ER, or from cholesterol accumulation in late endosomes. When Npc1^−/− cells take up LDL or acetylated (Ac)LDL, there is defective suppression of LDL receptors and HMG-CoA reductase; however, these SREBP target genes are effectively repressed by 25-OH cholesterol (Figure I, A, in the online data supplement).²² 25-OH-chol treatment of Npc1^+/+ or Npc1^−/− macrophages (Figure 2A and 2B) did not reduce Ctsk expression levels suggesting that induction does not result from defective SREBP function. Liver X receptor target genes are poorly regulated in Npc1^−/− cells following cholesterol loading of the endosomal system.²³ However, LXR activation with compound T0901317 (T0) in macrophages did not affect Ctsk expression. AcLDL loading of Npc1^+/+ macrophages, which results in uptake into lysosomes and re-esterification of lipoprotein-derived cholesterol in the ER,²⁴ had no effect on Ctsk expression. However, FC loading of Npc1^+/+ macrophages by treatment with AcLDL+ACAT inhibitor 58035 resulted in a 2-fold induction of Ctsk mRNA. Addition of U18666A, which causes a block in the exit of cholesterol from late endosomes and thus mimics many aspects of the Npc1^−/− phenotype,²⁵ caused a further increase in Ctsk in FC-loaded cells, indicating that effects of FC loading reflect endosomal rather than ER cholesterol accumulation. Because AcLDL+58035 treatment results in FC loading of the ER and induces the unfolded protein response,¹⁴ we also tested the effects of unfolded protein response inducers thapsigargin and tunicamycin to ensure that FC-induced ER stress was not involved in Ctsk induction. These treatments did not increase Ctsk expression (shown for tunicamycin, Figure 2A) although CHOP protein was induced (supplemental Figure I, B). These experiments suggested that Ctsk induction results from FC loading of late endosomes rather than altered sterol-induced ER regulatory events.

• Download figure
• Download PowerPoint

Prolonged loading of human THP-1 macrophages with aggregated LDL or OxLDL causes accumulation of FC in late endosomes/lysosomes resembling the phenotype of Npc1^−/− cells.^11,12 This effect is not observed in mouse peritoneal macrophages.¹¹ Loading of human THP-1 macrophages with aggregated LDL or OxLDL led to a 2- to 3-fold increase in Ctsk expression (Figure 2C). These findings further support the hypothesis that cholesterol loading of late endosomes leads to Ctsk induction and show relevance in human macrophages.

Because the Npc1 mutation results in defective trafficking of both cholesterol and glycolipids,⁶ and aggregated LDL and OxLDL are complex preparations containing many different bioactive lipids, we tested whether cholesterol itself could lead to induction of Ctsk. CD-chol loading of Npc1^+/+ macrophages resulted in a robust, dose-related induction of Ctsk (Figure 2A and 2D). Ctsk was also induced in CD-chol–loaded THP-1 macrophages (Figure 2C) but not 293 cells (data not shown). CD-chol loading similarly induced Ctsk in the presence of 58035. Thus, Ctsk induction can be directly related to the effects of FC loading. Because CD-chol loading enriches the plasma membrane with FC,²¹ whereas AcLDL+U18666A treatment loads late endosomes, it appears that cholesterol enrichment of either endosomal or plasma membranes can induce Ctsk.

A Role of p38 Mitogen-Activated Protein Kinase in the Induction of Ctsk

We repeated microarray experiments using Affymetrix arrays with more complete coverage of the genome. This revealed additional genes upregulated in Npc1^−/− cells. We confirmed the upregulation of newly discovered genes and Ctsk with real-time PCR in peritoneal macrophages elicited in various ways and in bone marrow–derived cells. Several genes upregulated in Npc1^−/− cells are known p38 targets: Ctsk, S100a8 and Mmp8 (supplemental Table). Activation of p38 was confirmed by detection of increased phospho-p38 levels in Npc1^−/− macrophages (Figure 3A). Elevated phospho-p38 levels were also observed in OxLDL loaded THP-1 cells (Figure 3B) and CD-chol–loaded mouse peritoneal macrophages (Figure 3C). In a time-course study, CD-chol loading led to an early peak (≈0.5 hour), a variable decrease and then a prolonged late-phase (>8 hours) of phospho-p38 induction. Interestingly, AcLDL+U18666A loading of macrophages caused predominantly late-phase p38 phosphorylation (Figure 3D), possibly attributable to a delay in hydrolysis of AcLDL CE.

• Download figure
• Download PowerPoint

We next assessed the role of p38 in Npc1^−/−- and cholesterol-induced Ctsk expression. Treatment of Npc1^−/− cells with two different p38 inhibitors reduced Ctsk expression (Figure 4A). In contrast, treatment with c-Jun NH₂-terminal kinase (JNK) or nuclear factor (NF)-κB inhibitors increased Ctsk mRNA levels (Figure 4A and supplemental Figure II), whereas ERK inhibitors had no effect (data not shown). Prolonged treatment with the p38 inhibitor SB202190 reduced Ctsk mRNA in Npc1^−/− cells by more than 50% (Figure 4B). As shown by Western blotting in macrophage cell lysates, Npc1^−/−- or CD-chol–induced CATK protein levels were also greatly reduced by the p38 inhibitors (supplemental Figure III). Mmp8, S100a8, and Mmp14 mRNA levels were all increased in Npc1^−/− cells, and reduced by treatment with the p38 inhibitors (Figure 4C). The cells treated with inhibitors appeared healthy by microscopy and the p38 inhibitors did not affect p38 phosphorylation as expected.

• Download figure
• Download PowerPoint

To confirm a role of p38 in Ctsk induction, we used p38α-deficient macrophages obtained from p38^flox/flox X LysMCre mice. Basal levels of Ctsk were reduced in these macrophages, and the effects of FC loading (AcLDL+58035 or CD-chol) on Ctsk expression were abolished (Figure 4D). CD-chol loading also induced the expression of Mmp8 and S100a8 and these effects were abolished in p38^−/− macrophages (Figure 4E). To determine the specificity of these p38-dependent responses, we measured transcript levels of other sterol-regulated genes (Figure 4E). Apart from a modest reduction of Abca1 mRNA in p38^−/− macrophages, also previously observed in tumor necrosis factor–treated macrophages,¹⁷ the cholesterol-mediated responses of these genes were not affected by p38 deficiency. These findings indicate that a novel signaling pathway can be initiated by cholesterol accumulation in plasma or endosomal membranes, leading to activation of p38 and induction of Ctsk, Mmp8, S100a8, and Mmp14.

Cholesterol-Mediated Induction of Ctsk Requires Microphthalmia Transcription Factor

Ctsk is highly induced during macrophage differentiation into multinucleated osteoclasts. Microphthalmia transcription factor (MITF) and NFATc1 are transcription factors necessary for osteoclast differentiation, and are transcriptionally induced and regulated by p38-mediated phosphorylation.²⁶ We measured the expression of Mitf, Nfatc1, and Ctsk following CD-chol loading over time and related these changes to p38 phosphorylation (Figure 5A). Nfatc1 and cathepsin S (Ctss) were not induced by cholesterol loading. There was an early peak of Mitf induction. The induction of Ctsk was initiated after Mitf induction and during the late-phase p38 phosphorylation. Addition of p38 inhibitor following 7 hours of CD-chol loading resulted in substantial inhibition of Ctsk induction (data not shown), suggesting that prolonged p38 activation was required for Ctsk induction. Interestingly, time course studies showed that NF-κB and JNK activation accompanied the early phase of p38 activation (<5 hours) but were not sustained during the late phase (data not shown). Thus, NF-κB and JNK repression of Ctsk (Figure 4A) may explain the delay in Ctsk induction by CD-chol.

• Download figure
• Download PowerPoint

To assess the role of MITF in Ctsk induction, we carried out cholesterol loading of splenic macrophages from mice carrying a dominant negative mutation of Mitf (mi/mi). The mi/mi mutation abolished the response to CD-chol loading (Figure 5B). Three E boxes in the human CTSK promoter are responsible for MITF transactivation during osteoclast differentiation.²⁰ Transfection of a human CTSK promoter–luciferase construct into RAW macrophages resulted in significant induction of gene expression in response to CD-chol loading, and mutations in any of the 3 E-box elements abolished the induction (Figure 5C). These experiments indicate an essential role for MITF in cholesterol-mediated CTSK induction and suggest that regulation occurs via binding of E-box elements in the promoter.²⁰ This was confirmed by ChIP analysis,¹⁶ showing a modest increase in binding of MITF, no change in binding of PU.1, and a marked increase in binding of phospho-p38 to the Ctsk promoter in mouse splenic macrophages (Figure 5D; note binding of p38 was only detected after CD-chol loading). These findings are similar to previous studies in which receptor activator of NF-κB ligand (RANKL) was shown to induce Ctsk expression via p38-mediated phosphorylation of MITF on the Ctsk promoter.¹⁶

RANK or Small GTPase Rab Do Not Mediate Ctsk Induction

We next carried out experiments to elucidate the signaling pathways acting upstream of p38/MITF-mediated induction of Ctsk. RANKL and CSF-1 are cytokines that commit myeloid precursor cells to the osteoclast lineage.²⁷ Another secreted protein, osteoprotegerin (OPG), competes with RANK for RANKL binding and blocks RANK/RANKL signaling in cells. We treated bone marrow–derived macrophages from Npc^+/+ and Npc1^−/− mice with OPG to assess the potential role of RANK/RANKL signaling in cholesterol-mediated Ctsk induction. Whereas OPG abolished Ctsk induction by exogenous RANKL and CSF-1 in both Npc1^+/+ and Npc1^−/− cells, there was no effect of OPG in untreated Npc1^−/− cells (supplemental Figure IV, A). Also, OPG treatment did not affect CD-chol–induced Ctsk expression in Npc1^+/+ macrophages, and injection of OPG into the peritoneal cavity of ConA-injected Npc1^−/− mice did not affect Ctsk induction (data not shown). Moreover, a 70% decrease in RANK mRNA by siRNA knockdown did not reduce Ctsk mRNA (supplemental Figure IV, B). These experiments suggested that the RANK-RANKL signaling pathway is unlikely to be involved in Npc1^−/−- or CD-chol–induced Ctsk expression. Although Rabs have been suggested to rescue the lipid trafficking defect in Npc1^−/− cells,²⁸ overexpression of either wild-type or dominant negative forms of rab7, rab9, and recycling endosome–associated rab11 did not affect Ctsk promoter activity in CD-chol–loaded RAW macrophages (data not shown). We also assessed the possible involvements of autocrine or paracrine factors in the induction of Ctsk with a media transfer experiment. However, levels of Ctsk in bone marrow–derived cells from either Npc1^+/+ or Npc1^−/− mice were not affected by conditioned media (supplemental Figure V).

Toll-Like Receptor Signaling Mediates Cholesterol-Induced Ctsk Expression

Multiple TLRs can activate p38 signaling.²⁹ MyD88 and TRIF are adaptor proteins that link TLR activation to downstream mitogen-activated protein (MAP) kinase signaling.²⁹ To evaluate the role of TLRs in cholesterol-mediated Ctsk induction, we assessed Ctsk expression in bone marrow–derived Myd88^−/−, Trif^−/− macrophages and wild-type controls. The expression of Ctsk and other p38 targets was reduced in untreated Myd88^−/−, Trif^−/− cells compared to controls, and the cholesterol-mediated induction was abolished (Figure 6A). In addition, late-phase p38 phosphorylation was abolished in Myd88^−/−, Trif^−/− macrophages (Figure 6B). These data indicate that TLRs or interleukin (IL)-1 receptor initiate cholesterol-induced signaling leading to sustained p38 phosphorylation and Ctsk induction. IL-1 (5 ng/mL) treatment did not induce Ctsk in mouse peritoneal macrophages. To determine whether TLR activation was sufficient for Ctsk induction, we treated mouse peritoneal macrophages with activators of various TLRs (Figure 6C). TLR3 [poly(I:C), 2.5 μg/mL], TLR4 (lipid A, the active component of lipopolysaccharide, 100 ng/mL) and TLR7 (gardiquimod, 10 μg/mL) ligands increased Ctsk mRNA by 2- to 4-fold, whereas TLR2 (PGN) and TLR9 (CpG) ligands did not. Lipid A and poly(I:C) had additive effects on Ctsk mRNA induction (Figure 6D). Whereas lipid A did not have any additional effect on CD-chol–induced Ctsk expression, the combination of poly(I:C) and CD-chol synergistically increased Ctsk mRNA (Figure 6D). Importantly, the CD-chol preparation contained <0.125 endotoxin units/mL. Moreover, CD-chol showed greater induction of Ctsk than lipid A, whereas the induction of inflammatory genes such as IL-6 or tumor necrosis factor by CD-chol was only approximately one-tenth of that observed with lipid A treatment, indicating that LPS contamination could not explain the Ctsk response.

• Download figure
• Download PowerPoint

To assess the role of various TLRs in Ctsk induction in Npc1^−/− macrophages, we used siRNA knockdowns (Figure 7A). As determined by real-time PCR, the knockdowns were effective (63% to 79%; legend of Figure 7) and specific for each TLR. Knockdown of TLR3 had the largest effect on Ctsk expression. However, knockdown of all TLRs localized to the endosomal system (TLR3, 7 and 8)³⁰ as well as TLR4 led to some reduction in Ctsk, indicating activation of signaling via multiple TLR family members in Npc1^−/− cells.

• Download figure
• Download PowerPoint

To determine whether TLR4 is necessary for cholesterol-induced Ctsk expression, we loaded peritoneal macrophages from Tlr4^del mice with CD-chol or AcLDL+U18666A. The response to lipid A was abolished, whereas the response to CD-chol loading was significantly but only partially reduced by the Tlr4 mutation (32% to 44% in different experiments) (Figure 7B). Although baseline Ctsk expression was higher in Tlr4^del macrophages (possibly reflecting decreased NF-κB and JNK expression), AcLDL+U18666A resulted in no further increase in Ctsk expression, suggesting that TLR4 was also involved in the response to endosomal cholesterol loading. Similar results were obtained using macrophages from C3H/HeJ mice that carry a spontaneous mutation in Tlr4 (data not shown). CD-chol induction of Ctsk was unchanged but poly (I:C) and AcLDL+U18666A-mediated expression were abolished in Tlr3^−/− macrophages (Figure 7C). These data suggest that Ctsk induction by CD-chol loading of plasma membrane depends partially on TLR4, whereas AcLDL+U18666A loading of the endosomal compartment leads to TLR3 and TLR4 signaling.

Discussion

Our studies have elucidated a novel signaling pathway in macrophages, initiated by FC accumulation in plasma or endosomal membranes and leading to activation of TLRs, notably TLR3 and -4, and sustained phosphorylation of p38 MAP kinase. Several p38 targets that were induced by cholesterol loading, such as Ctsk,^9,10 various Mmps,⁸ and S100a8,³¹ participate in atherogenesis or its complications, suggesting the relevance of this signaling pathway to the mechanisms of accelerated, complicated atherosclerosis of Apoe^−/−, Npc1^−/− mice.⁵

Although the Npc1 mutation is rare, our findings may be relevant to the common forms of chronic atherosclerosis that involve overloading of macrophages with FC. For example, chronic loading of human macrophages with AggLDL or OxLDL, a situation that is likely to be physiologically relevant to human atherogenesis,³² also leads to accumulation of unesterified cholesterol in late endosomes,^11,12 resulting in p38 activation and Ctsk induction (Figure 3A and 3B). Moreover, the enrichment of plasma membrane cholesterol using CD-chol led to marked activation of p38 signaling and Ctsk induction. This may simulate the effects of cellular membrane enrichment occurring as a result of contact between cells and atherogenic plasma lipoproteins with a high cholesterol/phospholipid ratio.³² Although CD-chol loading also leads to increased sterol in the endocytic recycling compartment,³³ reversal of CD-chol induction of Ctsk by mutant forms of ABCA1 that are localized to plasma membrane³⁴ suggest that the plasmalemma is the relevant compartment (M. Ishibashi, A.R. Tall, unpublished studies, 2007).

Previous studies have indicated an important role of TLRs in atherogenesis, providing a link between hyperlipidemia and expression of a variety of inflammatory and chemokine genes.³⁵ Deficiency of Myd88 in Apoe^−/− mice led to a major reduction in atherosclerosis and inflammatory gene expression. Similarly, deficiency of TLR4 in Apoe^−/− mice led to reduced atherosclerosis but the effect was much smaller than that of MyD88 deficiency, suggesting involvement of different TLRs upstream of MyD88. TLR2 has also been implicated in atherogenesis.³⁵ Interestingly, our studies indicate an important role of TLR3 in the response to endosomal cholesterol loading with AcLDL+U18666A as well as the Npc1^−/− mutation. In fact, multiple TLRs that reside in endosomes appeared to be involved in the induction of Ctsk in Npc1^−/− cells, ie, TLR3, 7 and 8, as well as TLR4. Endosomes play a critical role in sorting and silencing signaling receptors such as TLRs. Inhibition of the endosomal pathway blocks postendosomal cholesterol sorting and increases TLR4 signaling³⁶ and mislocalization of late endosomes results in the sustained activation of MAP kinases.³⁷ In Npc1^−/− fibroblasts, TLR4 signaling fails to shut off, and active TLR4 accumulates in endosomes and increases cytokine secretion.³⁸

Our studies suggest spontaneous activation of TLR signaling by cholesterol loading of endosomes, but indicate that cholesterol accumulation may also enhance the responses to exogenous TLR ligands (Figure 6D). A variety of endogenous ligands of TLR4 as well as LPS may be involved in atherogenesis.³⁹ Although TLR3 or 7 have not yet been implicated in atherogenesis, necrotic cells release endogenous RNA and heat shock protein-70 that stimulate TLR3 and 7.^40,41 Poly(I:C) and endosomal cholesterol loading synergistically induced Ctsk (Figure 6D), suggesting that macrophage phagocytosis of virally infected apoptotic cells may also result in a similar synergy between RNA-mediated and cholesterol-induced responses.

In summary, our study provides new insights into the mechanisms linking macrophage cholesterol accumulation and inflammatory responses mediated by p38 activation. Our findings add to the growing evidence for involvement of TLRs and implicate TLR3, as well as other TLRs, in mediating signaling from cholesterol-loaded endosomes. There is increasing evidence for involvement of MMPs and cathepsins in aneurysm formation and possibly plaque destabilization.^7,8 Although there is likely to be redundancy among individual MMPs or cathepsins, our studies suggest a common underlying mechanism involving p38 activation in plaque complications.

Footnotes