Mitochondrial metabolic reprogramming by CD36 signaling drives macrophage inflammatory responses

Abstract

Rationale:

A hallmark of chronic inflammatory disorders is persistence of pro-inflammatory macrophages in diseased tissues. In atherosclerosis this is associated with dyslipidemia and oxidative stress, but mechanisms linking these phenomena to macrophage activation remain incompletely understood.

Objective:

To investigate mechanisms linking dyslipidemia, oxidative stress and macrophage activation through modulation of immunometabolism, and to explore therapeutic potential targeting specific metabolic pathways.

Methods and Results:

Using a combination of biochemical, immunological, and ex vivo cell metabolic studies, we report that CD36 mediates a mitochondrial metabolic switch from oxidative phosphorylation to superoxide production in response to its ligand, oxLDL. Mitochondrial-specific inhibition of superoxide inhibited oxLDL-induced NF-κB activation and inflammatory cytokine generation. RNAseq, flow cytometry, ³H-labeled palmitic acid uptake, lipidomic analysis, confocal and EM imaging, and functional energetics revealed that oxLDL upregulated effectors of long-chain fatty acid (FA) uptake and mitochondrial import, while downregulating FA oxidation and inhibiting ATP5A, an electron transport chain (ETC) component. The combined effect is long-chain FA accumulation, alteration of mitochondrial structure and function, repurposing of the ETC to superoxide production, and NF-κB activation. Apoe null mice challenged with high fat diet showed similar metabolic changes in circulating Ly6C⁺ monocytes and peritoneal macrophages, along with increased CD36 expression. Moreover, mitochondrial ROS was positively correlated with CD36 expression in aortic lesional macrophages.

Conclusions:

These findings reveal that oxLDL/CD36 signaling in macrophages links dys-regulated FA metabolism to oxidative stress from the mitochondria, which drives chronic inflammation. Thus, targeting to CD36 and its downstream effectors may serve as potential new strategies against chronic inflammatory diseases such as atherosclerosis.

INTRODUCTION

Atherosclerosis is the major underlying mechanism for many cardiovascular diseases and is the leading cause of death in the developed world ^{1, 2}. It is commonly associated with hyperlipidemia, oxidative stress and low-grade non-resolved chronic inflammation ^{3, 4}. Although atherosclerosis is a complicated disorder that involves various immune cell types, macrophages play a central role in all stages of the disease ⁵. For example, at early stages macrophages mediate uptake of oxidized LDL (oxLDL) which is converted from LDL in the setting of oxidative stress. Continuous macrophage oxLDL uptake leads to intracellular lipid accumulation and formation of foam cells, which become the major component of atherosclerotic plaque ⁶. At advanced stages, macrophage-derived foam cells undergo apoptosis contributing to formation of the atherosclerotic plaque necrotic core ⁷. Therefore, how the microenvironment elicits pro-atherogenic phenotypes in macrophages has been under intensive investigation over the past several decades.

CD36 is a scavenger receptor densely expressed on the macrophage surface and its roles in immunity, metabolism and atherogenesis are well recognized ⁸. We and others have reported that upon binding to oxLDL, macrophage CD36 recruits multiple transmembrane and intracellular protein partners into a signaling complex that mediates oxLDL uptake leading to foam cell formation ^{6, 9–11}. We also showed that oxLDL-CD36 signaling inhibits macrophage migration, which may contribute to the trapping of lipid-laden macrophages in the arterial intima ¹². Accordingly, deleting the cd36 gene in mice provides protection from high fat diet (HFD)-induced atherosclerosis ¹³. Pro-atherogenic signaling through CD36 is mediated by intracellular reactive oxygen species (ROS) ¹⁴. However, NADPH oxidase inhibitors do not fully block CD36 pro-atherogenic effects ¹² and deficiency in phagocyte NADPH oxidase fails to inhibit atherosclerosis in mice ¹⁵. These studies suggest that other intracellular ROS sources are involved downstream of CD36.

Mitochondrial function and bioenergetics have gained much recent attention in studies of chronic inflammatory diseases including atherosclerosis. Beyond their conventional role in generating energy, elegant studies have demonstrated that mitochondria play critical roles in immunity ¹⁶. Mitochondria not only produce signals for innate immune responses ^{17, 18}, but can be repurposed for ROS production to drive and sustain inflammatory status in macrophages ¹⁹. Moreover, mitochondrial dysfunction is associated with atherosclerosis based on studies in both animal models and human patients ²⁰. Despite accumulating evidence implicating mitochondria in the development of atherosclerosis, the molecular mechanisms linking mitochondria to chronic inflammation under atherogenic conditions are still not well understood. Here we hypothesize that mitochondria are important sources of intracellular ROS downstream of oxLDL/CD36 axis. We used mitochondria-targeted ROS probes and ROS inhibitors to demonstrate that CD36 mediates mitochondrial ROS production to drive inflammatory status in macrophages. Mechanistically, oxLDL-CD36 signaling re-programs fatty acid metabolism by up-regulating long-chain fatty acid (LCFA) trafficking into the mitochondria while down-regulating fatty acid oxidation (FAO). This ultimately leads to mitochondrial LCFA accumulation, which facilitates ROS generation. Using the atherosclerosis-prone apoe null mouse model, we demonstrated that HFD challenge induced mitochondrial ROS generation and up-regulated CD36 expression in circulating Ly6C⁺ monocytes. In addition, CD36 expression positively correlated with mitochondrial ROS levels in aortic lesional macrophages. Deletion of cd36 in apoe null mice attenuated HFD-induced mitochondrial ROS generation. These studies describe a novel CD36-dependent molecular pathway linking hyperlipidemia, oxidative stress, mitochondrial dysfunction, and chronic inflammation under atherogenic conditions.

METHODS

Detailed methods are available in the Online Data Supplement.

RESULTS

OxLDL stimulates CD36-dependent mitochondrial ROS production.

The sources of intracellular ROS generation in macrophages stimulated with oxLDL have not been completely defined. Using a cellular ROS indicator, 2’, 7’-dichlorodihydrofluorescein diacetate (H₂DCFDA), we showed that oxLDL stimulated cellular ROS production in a CD36-dependent manner (Figure 1A), consistent with previous studies from our lab and others ^{11, 12}. To test whether mitochondria contribute to oxLDL-induced ROS, we used a specific mitochondria-targeted superoxide fluorescent probe, MitoNeoD ²¹ and found that wild type (WT) murine macrophages treated with 20 μg/ml oxLDL showed a time-dependent increase in fluorescence compared to basal levels or to cells treated with native LDL, with significant increases seen as early as 30 minutes and a 3.4 fold increase at 24h (Figures 1B and C). Cd36 null cells showed less than 50% increase in fluorescence at 24h (Figure 1D). Similar results were seen in human monocyte-derived macrophages (Figure S1).

Figure 1. OxLDL Induces CD36-Dependent Mitochondrial ROS Production.

(A) Examples of histograms of carboxy-DCFDA fluorescence in WT or cd36 null peritoneal macrophages treated with 50μg/ml oxLDL for 1h. MFI was quantified and shown in the bar graph; n=3 per group. (B) Examples of histograms of MitoNeoD fluorescence in WT peritoneal macrophages treated with 20 μg/ml LDL or oxLDL for 24h. MFI are shown in the bar graph; n=4 per group. (C) WT peritoneal macrophages treated with 20 μg/ml oxLDL for indicated time and MitoNeoD MFI was quantified and shown in the bar graph; n=4 per group. (D) Examples of histograms of MitoNeoD fluorescence in WT or cd36 null peritoneal macrophages treated with 20 μg/ml oxLDL for 24h. MFI was quantified and shown in the bar graph; n=4 per group. ****p<0.0001 compared to control.

Mitochondrial ROS induced by oxLDL/CD36 signaling facilitate NF-κB activation and pro-inflammatory status.

As shown previously by Wang et al ²², we demonstrated by immunoblot with an antibody to phosphorylated serine 536 in the Rel subunit of NF-κB that 20 μg/ml oxLDL activated NF-κB in WT macrophages. Significantly lower levels of NF-κB activation, however, were seen in cd36 null macrophages (Figure 2A). To validate this finding we assessed mRNA levels of NF-κB target pro-inflammatory genes, including mcp1 ²³, il6 ²⁴, tnf-α ²⁵, and cxcl10 ²⁶ and found all four transcripts were significantly increased in macrophages from WT mice exposed to oxLDL, but not in cd36 null cells (Figure 2B). Interestingly, the anti-inflammatory gene, il10, was up-regulated by oxLDL in cd36 null but not WT cells (Figure 2B). We also measured cytokine protein levels by ELISA and in agreement with mRNA data, found that oxLDL induced less MCP-1, IL-6 and TNF-α in cd36 null macrophages compared to WT (Figure 2C). To demonstrate the role of mitochondrial ROS in this pathway, we showed that the mitochondrial-targeted anti-oxidant MitoTEMPO suppressed oxLDL-induced NF-κB activation (Figure 2D) as well as downstream expression of pro-inflammatory genes (Figure 2E) and proteins (Figure 2F).

Figure 2. OxLDL-Induced Mitochondrial ROS Facilitates NF-κB Activation and Pro-inflammatory Cytokine Production.

(A) Examples of Western blot images from WT and cd36 null macrophages treated with 20 μg/ml oxLDL for indicated time periods. NF-κB activation was assessed by immunoblot using an antibody against phospho-serine536 in NF-κB (p65). Membranes were stripped and re-probed with the antibody against NF-κB (p65). Blot images were quantified and expressed as percentage of control; n=3 per group. (B) WT and cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 8h. Total RNA was isolated and subjected to qRT-PCR for analysis of pro-inflammatory cytokine genes including mcp1, il6, tnf-α, and cxcl10. The anti-inflammatory cytokine il10 was also assessed n=4 per group. (C) WT and cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 8h. Then culture media were collected and subjected to ELISA assay for MCP-1, IL-6, and TNf-α; n=4 per group. (D) Examples of Western blot images from WT peritoneal macrophages treated with 20 μg/ml oxLDL for indicated time (lane 1-4), or cells pre-treated with 500μM MitoTEMPO for 1h before addition of 20 μg/ml oxLDL (lane 5-8), or cells pre-treated with 200μM etomoxir for 1h before addition of 20 μg/ml oxLDL (lane 9-12). Membranes were probed for phospho-serine536 in NF-κB, stripped and re-probed for NF-κB; n=3. (E) WT peritoneal macrophages were treated with 20μg/ml oxLDL for 4h or pre-treated with either 500μM MitoTEMPO or 200μM etomoxir for 1h before addition of 20μg/ml oxLDL for 4h. Total RNA was isolated and subjected to qRT-PCR for mcp1, il6, tnf-α, cxcl10, and il1b; n=4 per group. (F) WT peritoneal macrophages were treated with 20 μg/ml oxLDL for 24h or pre-treated with either 500μM MitoTEMPO or 200μM etomoxir for 1h before addition of 20 μg/ml oxLDL for 24h. Culture media were collected and subjected to ELISA assay for MCP-1, and IL-6; n=4 per group. Data are presented as mean ± SEM. **p<0.01, ***p<0.001, ****p<0.0001 compared to control conditions. ^##p<0.01, ^###p<0.001, ^####p<0.0001 compared to oxLDL treated alone conditions.

OxLDL induces a CD36-dependent metabolic switch from mitochondrial oxidative phosphorylation (OXPHOS) to glycolysis.

Mitochondrial ROS production is often associated with alteration in bioenergetics ¹⁹. We next investigated the effects of oxLDL on bioenergetics by treating WT murine peritoneal macrophages with increasing concentrations of oxLDL for 24h and then subjecting them to Seahorse extracellular flux analysis, which measures oxygen consumption rate (OCR) and extracellular acidification rate (ECAR). Figure 3A shows that oxLDL suppressed OCR, with ~50% suppression at 20 μg/ml and ~80% at 50 and 100 μg/ml, consistent with inhibition of mitochondrial OXPHOS. Native LDL at 20µg/ml had no effect on OCR and higher concentrations had only a minor effect (Figure S2A). Concomitant with suppression of OCR, oxLDL treatment induced elevation in ECAR, consistent with a metabolic switch to glycolysis (Figure 3B). Time course experiments showed that oxLDL-mediated reduction in OCR was apparent as early as 3h after exposure, while the increase in ECAR lagged behind (Figure 3C), suggesting that the switch to glycolysis was a compensatory response. To validate the ECAR studies we assessed macrophage glucose uptake by flow cytometry using a fluorescence tagged, non-metabolizable glucose analog, 2-NBD-deoxyglucose and found consistent 2-fold increase in uptake after oxLDL exposure (Figure 3D). To show human relevance, we also assessed mitochondrial function and glucose uptake in human monocyte-derived macrophages and found similar time and concentration dependent suppression of OCR and enhancement of ECAR and 2-NBD-deoxyglucose uptake by oxLDL (Figure S2B–D).

Figure 3. OxLDL-CD36 Signaling Induces a Metabolic Switch in Macrophages from Mitochondrial OXPHOS to Glycolysis.

(A/B) Left panels show representative OCR (A) and ECAR (B) curves from WT peritoneal macrophages treated with incremental concentrations of oxLDL for 24h. Right panels show quantified data from replicate studies; n=4 per group. Both OCR and ECAR values were normalized by protein content in each 96-well sample. (C) OCR and ECAR levels were quantified from WT peritoneal macrophages treated with 50 μg/ml oxLDL at timed points; n=4 per group. (D) Examples of histograms of 2-NBDG fluorescence in WT peritoneal macrophages treated with 50 μg/ml oxLDL for 3h. Mean fluorescence intensity (MFI) was quantified from 5 separate experiments and shown as fold change in the bar graph. (E) Examples of Western blot image of PDHK1 and β-actin (loading control) from WT peritoneal macrophages treated with 20 μg/ml oxLDL for 24h. Images were quantified, normalized to β-actin and expressed as percentage of control. n=3 per group. (F) OCR and ECAR were quantified and combined from WT or cd36 null peritoneal macrophages treated with 20 μg/ml oxLDL for 24h. n=4 per group; WT or cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 5h in the presence of 2-NBDG and intracellular 2-NBDG fluorescence signals were measured by flow cytometry; n=3 per group. (G) WT or cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for indicated time periods and cell surface levels of glucose transporter Glut1 were assessed by flow cytometry using PE-conjugated anti-Glut1 IgG; n=4 per group. (H) WT or cd36 null cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 24h and mitochondria membrane potential was assayed by TMRM fluorescence. MFI was quantified and shown in the bar graph; n=4 per group. (I) Examples of western blots with anti-FITC and anti-ATP5A antibodies in anti-ATP5A immunoprecipitates from WT peritoneal macrophages treated with indicated concentrations of oxLDL for 24hr and incubated with 5-IAF to tag cysteine free thiols. (J) Examples of histograms of 2-NBDG fluorescence in WT peritoneal macrophages treated with 20 μg/ml oxLDL for 5h or pre-treated with MitoTempo for 1h before addition of 20 μg/ml oxLDL for 5h; n=3 per group. *p<0.05, **p<0.01, ****p<0.0001 compared to control conditions.

An end product of glycolysis is pyruvate, which can enter the TCA cycle in mitochondria after conversion to acetyl-CoA by pyruvate dehydrogenase (pyruvate oxidation pathway). This enzyme is regulated by pyruvate dehydrogenase kinase 1 (PDHK1), which inhibits its activity by phosphorylation ²⁷. Figure 3E shows that oxLDL up-regulated PDHK1 protein levels by ~80%, suggesting that oxLDL may direct pyruvate away from mitochondria and towards conversion to lactate acid. To examine how the oxLDL-induced metabolic switch affects downstream metabolic products we measured cellular ATP, ADP, and NAD⁺ levels by high-performance liquid chromatography (HPLC) and found no significant differences after oxLDL treatment (Figure S3A). This suggests that although cells altered metabolic pathways for ATP production, they were able to control and maintain cellular energy homeostasis.

To investigate the role of CD36 in the oxLDL-induced macrophage metabolic switch, metabolic phenotypes of WT cells were compared to those from cd36 null animals after treatment with 20 μg/ml oxLDL for 24h. Figure 3F shows that oxLDL induced a significant 50% reduction in OCR and a 40-50% increase in ECAR and 2-NBD-deoxyglucose uptake in WT but not cd36 null macrophages. The increase in glucose uptake was associated with oxLDL/CD36 dependent up-regulated surface expression of the glucose transporter Glut1 (Figure 3G).

To understand why mitochondrial respiration was decreased, we assessed mitochondrial mass using MitoTracker Green and detected no differences after oxLDL treatment or in cd36 null macrophages compared to WT cells (Figure S3B). Despite a significant reduction in OXPHOS in response to oxLDL, mitochondrial membrane potential assessed by tetramethylrhodamine methyl ester (TMRM) was not decreased and in fact was slightly elevated (14%; Figure 3H). No differences were detected in cd36 null macrophages. Mitochondrial ATP synthase, the active component of ETC complex V uses a proton gradient across the inner mitochondria membrane generated by ETC complex I-IV for ATP production. Elevation of the membrane potential despite a reduction in OCR is consistent with inhibition of ATP synthase (complex V) activity. This hypothesis was supported by Seahorse data calculating OCR for ATP production (the difference between OCR levels before and after addition of oligomycin, the ATP synthase inhibitor). OxLDL suppressed mitochondrial ATP production in a dose-dependent manner (Figure S3C). ATP synthase inhibition was further supported by the observation that oligomycin but not the mitochondrial uncoupler FCCP or the ETC complex I inhibitor rotenone stimulated glycolysis in macrophages (Figure S3D). Mechanistically ATP synthase activity can be inhibited by oxidative post-translational modification of a critical cysteine within the α subunit (ATP5A) ²⁸. To assess whether ATP5A cysteine was oxidatively modified, we used a FITC-tagged 5-iodoacetamide probe for free cysteines ²⁹ and showed by immunoprecipitation followed by western blot (Figure 3I) that macrophage ATP5A was readily tagged by the probe for free cysteines, but after oxLDL treatment for 24h access to the probe was lost. However, this modification by oxLDL was not observed in macrophages from MCAT mice that express human catalase in mitochondria²², indicating the ATP5A cysteine modification was mediated by ROS (Figure S3E). Interestingly, oxLDL-induced reduction in OCR was also attenuated in MCAT macrophages (Figure S3F). Furthermore, MitoTEMPO suppressed oxLDL-induced glucose uptake (Figure 3J), suggesting ROS may modify and suppress the ATP synthase so that cells switch to glycolysis for energy.

RNA sequencing reveals that oxLDL treatment re-programs major metabolic pathways in macrophages.

To gain insights into potential mechanisms through which oxLDL alters macrophage function we performed RNAseq on WT cells treated with 20μg/ml oxLDL for 8h and found major changes in the transcriptome compared to untreated cells (Figure S4A). We also examined cd36 null cells treated with oxLDL. The Venn diagram in Figure 4A shows differentially expressed genes among 4 pair-wise comparisons using principle component (PC) analysis and shows that oxLDL had a strong effect in both genetic backgrounds, but that cd36 deletion resulted in loss of many gene responses and gain of others (Figure 4B). Consistent with qRT-PCR studies shown in Figure 2, the RNAseq analysis showed significant increases in pro-inflammatory cytokine and chemokine transcripts, most notably il1b, 1l12b, cxcl2, cscl3 and nos2. These studies also showed that major metabolic pathways were affected by oxLDL. Figure 4C shows that oxLDL down-regulated most of the genes involved in the TCA cycle suggesting that mitochondrial OXPHOS was suppressed not only in a post-translational manner (modification on ATP5A as shown in Figure 3I) but also at the level of transcription. In contrast to transcripts of genes in TCA cycle, most of the glycolytic pathway transcripts were unchanged or elevated (Figure S4B). Notably, genes that control the three rate-limiting steps of glycolysis including hk3, pfkl, and pkm were all up-regulated by oxLDL, consistent with the notion that cells up-regulated glycolysis to compensate for loss of mitochondrial OXPHOS. In contrast, genes in the pyruvate oxidation pathway were mostly down-regulated by oxLDL (Figure S4C), consistent with western blot data on PDHK1 (Figure 3E). Figure 4D shows that unlike TCA cycle transcripts, most of the transcripts encoding ETC components showed either no difference or were increased after oxLDL treatment. One exception was ETC complex II, which also participates in the TCA cycle as succinate dehydrogenase (Sdha-d in Figure 4C).

Figure 4. Next-Generation RNA Sequencing Reveals That OxLDL Leads to Re-programming of Major Metabolic Pathways in Macrophages.

WT peritoneal macrophages were treated with 20 μg/ml oxLDL for 8h (n=3 per group) and total RNA was isolated, purified and subjected to next-generation RNA sequencing. (A) Venn Diagram of the differentially expressed genes among four pair-wise comparisons shows that the treatment of oxLDL has a strong effect in both genetic backgrounds, a loss of many gene’s response for cd36 null, and a switch where new genes change expression for cd36 null that did not for WT. (B) PC analysis revealed a strong and consistent global gene expression change for both the treatment with oxLDL and for the cd36 null. The first PC was largely determined by treatment, while the second by cd36 null. (C) Expression of genes encoding metabolic enzymes involved in the TCA cycle. (D) Expression of genes encoding metabolic enzymes contributing to the electron transport chain. Gene expression values were expressed as log₂(FC) comparing between oxLDL and control conditions and shown in the heatmap. FC: fold change. (E) Expression of genes encoding antioxidant enzymes. Key enzymes emphasized in the text are highlighted in red rectangles. (F) Expression of genes encoding metabolic enzymes contributing to de novo fatty acid synthesis. (G) Expression of genes encoding metabolic enzymes contributing to the fatty acid oxidation. Gene expression values were converted by Z-score and shown in the heatmap.

Cellular redox status is determined by the balance between ROS production and the functions of anti-oxidant systems. Figure 4E shows that oxLDL up-regulated expression of many key anti-oxidant transcripts, including mitochondria-specific superoxide dismutase 2 (sod2), the peroxiredoxin (prdx) /thioredoxin (txn) /thioredoxin reductase (txnrd) system, and glutathione reductase (gsr). These data suggest that mitochondrial ROS production induced by oxLDL was unlikely due to down-regulation of the anti-oxidant systems. In fact, the anti-oxidant systems were up-regulated most likely in an attempt to counterbalance the oxidative stress.

Interestingly, RNAseq revealed major impact on transcripts associated with fatty acid metabolism. Figure 4F shows that genes for fatty acid synthesis (e.g. ATP citrate lyase (acly), fatty acid synthase (fasn), acetyl-CoA carboxylase alpha (acaca), and stearoyl-CoA desaturase 1 (scd1) were all down-regulated by oxLDL, while genes for converting fatty acids to fatty acyl CoA were mostly up-regulated. These data suggest that cells may sense intracellular fatty acid accumulation induced by oxLDL and reduce de novo fatty acid synthesis in response while attempting to convert free fatty acids into fatty acyl CoA, which can be either stored in triacylglycerides or transported into the mitochondria to be catabolized through FAO. However, almost all genes involved in FAO were down-regulated by oxLDL (Figure 4G).

OxLDL/CD36 signaling stimulates fatty acid trafficking into mitochondria, which facilitates mitochondria re-purposing and inflammatory activation.

Based on the RNAseq data, we hypothesized that oxLDL induces re-programming of fatty acid metabolism leading to the mitochondrial LCFA accumulation (Figure S5). To test this hypothesis, we assayed fatty acid (palmitate) uptake (Figure 5A) and showed that uptake was increased by 50% after oxLDL stimulation in WT macrophages, but only by 20% in cd36 null cells, consistent with the notion that oxLDL induces LCFA uptake mainly through CD36. We also found that oxLDL exposure upregulated expression of other genes involved in LCFA trafficking including fabp4 ³⁰, acsl1 ³¹, cpt1 and cpt2 ³² (Figure 5B). Western blot analysis showed a 63% increase in FABP4 and 2-fold increase in ACSL1 protein expression levels after oxLDL treatment, validating the mRNA data (Figure 5C).

Figure 5. OxLDL Stimulates Fatty Acid Trafficking into Mitochondria, Driving Re-purposing and Inflammation.

(A) WT or cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 24h. Cells were then incubated with 1μg/ml bodipy-palmitic acid for 5min. Fluorescence signals from bodipy-palmitic acid were assayed by flow cytometry and data were combined and shown in the bar graphs; n=3 per group. (B) WT peritoneal macrophages were treated with 20 μg/ml oxLDL for 24h. Total RNA was isolated and subjected to qRT-PCR for specific genes involved in FA uptake and intracellular trafficking, including cd36, fabp4, acsl1, cpt1a and cpt2; n=4 per group. (C) Examples of Western blots of FABP4, ACSL1 and β-actin (loading control) from WT peritoneal macrophages treated with culture medium or 20 μg/ml oxLDL for 24h. Images were quantified, normalized by β-actin and expressed as percentage of control; n=5 per group. (D) Examples of Western blot images from WT peritoneal macrophages treated with 20 μg/ml oxLDL for indicated times. ACC inactivation was detected by immunoblot using an antibody against ACC1/2 phospho-serine791. Membranes were stripped and re-probed with the antibody against ACC1/2. Blots of the mitochondrial specific isoform ACC2 were quantified and expressed as percentage of control; n=4 per group. (E) WT or cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 3h before addition of ³H-Palmitate and mitochondria fractions were isolated and ³H signals were quantified; n=5 per group. (F) WT peritoneal macrophages were treated with 20 μg/ml oxLDL for 8h. Total RNA was isolated and subjected to qRT-PCR for genes responsible for the initial step of FAO including acadm and acadl; n=4 per group. (G) WT or cd36 null peritoneal macrophages were treated with 20 μg/ml oxLDL for 24h and then subjected to FAO assay using the Seahorse XF system. The differences in OCR with or without etomoxir were calculated and considered basal FAO. The differences in OCR with or without etomoxir after addition of FCCP were calculated and considered maximum FAO. n=3 per group. (H) WT macrophage mitochondria were isolated and subjected to (QFAME) lipidomics analysis. All detected fatty acid species were quantified and shown in the bar graphs. n=3 per group. (I) WT peritoneal macrophages were treated with 20 μg/ml oxLDL for 5h after pre-treatment with 200μM etomoxir for 1h. Mitochondrial ROS levels were then assayed using MitoNeoD and MFI was quantified and shown in the bar graphs. Data are presented as percentage of mitochondrial ROS induction by oxLDL; n=3 per group. (J) WT peritoneal macrophages were treated with 20 μg/ml oxLDL for 5h in the presence of 2-NBDG after pre-treatment with 200μM etomoxir for 1h followed. Intracellular 2-NBDG fluorescence was measured by flow cytometry and MFI was quantified and expressed as percentage of control. n=4 per group. (K) WT or cd36 null peritoneal macrophages were pre-treated with different doses (50μM, 100μM and 200μM) of etomoxir for 1h followed by addition of 20 μg/ml oxLDL for 24h. Culture media were collected and subjected to ELISA assay for MCP-1, and IL-6; n=4 per group. *p<0.05, **p<0.01 compared to control conditions.

The mitochondrial outer membrane enzyme, acetyl-CoA carboxylase 2 (ACC2), catalyzes conversion of acetyl-CoA to malonyl-CoA which inhibits CPT1 and limits LCFA trafficking into the mitochondria (Figure S5) ³³. Phosphorylation and inactivation of ACC2 by AMP-activated protein kinase (AMPK) is an important regulatory mechanism for fatty acid mitochondrial trafficking and lipid homeostasis ³⁴. Figure 5D shows that oxLDL induced phosphorylation of ACC2, and correspondingly increased palmitic acid accumulation in mitochondria in WT but not in cd36 null macrophages (Figure 5E). We did not detect AMPK activation (Figure S6A), suggesting elevation in phospho-ACC2 may involve other unidentified mechanisms. Moreover, we found similar ACC phosphorylation dynamics in the cd36 null macrophages in response to oxLDL (Figure S6B) indicating that oxLDL inactivated ACC through a CD36-independent mechanism.

One potential explanation for increased fatty acid trafficking into mitochondria is to feed fatty acids into the TCA cycle through FAO. RNA seq data, however, indicated down-regulation of FAO pathways by oxLDL (Figure 4G). qRT-PCR of ACADM and ACADL, enzymes responsible for the initial steps of FAO, validated the RNAseq data, showing that both transcripts were decreased by half following oxLDL treatment (Figure 5F). We also measured FAO with a functional Seahorse XF assay (Figure 5G) and found that in WT macrophages the uncoupling agent FCCP significantly stimulated FAO (comparing the third bar with the first bar) under control conditions, indicating that WT macrophages can efficiently use the FAO pathway for ATP production when stressed. However, after oxLDL treatment, FAO was not stimulated by FCCP, consistent with the proposition that macrophages shut down mitochondrial respiration in response to oxLDL and switch to glycolysis for energy. Interestingly, in cd36 null macrophages, FAO was largely inhibited under either control or oxLDL conditions, indicating the essential role of CD36 in fatty acid metabolism and bioenergetics.

These data in sum suggest that oxLDL stimulates fatty acid mitochondrial trafficking and suppresses FAO. We therefore reasoned that these metabolic alterations would lead to mitochondria LCFA accumulation, which has been associated with local ROS production ³⁵. To test this hypothesis, we directly measured mitochondrial fatty acid species after oxLDL exposure by quantitative fatty acid methyl esters lipidomics analysis. Consistent with our hypothesis, we detected significant increases in three LCFA species, palmitic acid (C16:0), stearic acid (C18:0), and oleic acid (C18:1) (Figure 5H). The total mitochondrial FA level showed no change, while levels of short chain fatty acids (e.g. C11, C12, and C14) tended to decrease (Table S1). Moreover, we used etomoxir, a small molecular weight chemical that irreversibly binds and inhibits CPT1, which is required for mitochondrial import of LCFA ³⁶. In WT macrophages etomoxir reduced mitochondrial ROS induced by oxLDL by 40% (Figure 5H). This inhibitory effect of etomoxir appeared to be specific for CD36-mediated LCFA trafficking as etomoxir also blocked palmitic acid-induced mitochondrial ROS in WT macrophages, while it showed no effects in cd36 null macrophages (Figure S7). Consistently, oxLDL-stimulated glucose uptake was also attenuated by etomoxir (Figure 5I). Moreover, similar to the effects from MitoTEMPO, etomoxir blocked NF-κB activation (Figure 2D) as well as the downstream inflammatory gene expression (Figure 2E and Figure 5J). These inhibitory effects of etomoxir were less pronounced in cd36 null macrophages (Figure 5J).

OxLDL alters dynamics and morphology of the mitochondria network.

Many studies have demonstrated that mitochondria are highly dynamic networks and there is intrinsic connection between structural organization and respiratory status ^37–39. We thus conducted confocal fluorescence imaging studies to assess how oxLDL alters macrophage mitochondria morphology by overexpressing the fluorophore EYFP specifically targeting to mitochondria. Figure 6A shows a tubular mitochondria network in control cells, while oxLDL exposure led to a highly fragmented mitochondria network. Mitochondria morphology quantified in 3-D reconstructed confocal images using the specialized software, MitoGraph⁴⁰, revealed a decrease in PHI and three-way junctions (the parameters indicating interconnectivity), and an increase in free ends (indicator of fragmentation and fission events) (Figure 6B). Mitochondria fission/fusion events play a major role regulating their morphology³⁹. As the above data suggest that oxLDL may lead to dys-regulated fission events, we examined and observed a time-dependent induction in DRP1 expression (Figure 6C), a GTPase that mediates mitochondrial fission⁴¹. On the other hand, expression of mitofusin 1 (MFN1), a GTPase that mediates mitochondrial fusion, was not affected by oxLDL. Thus, our data suggest that oxLDL induced a progressive increase in mitochondria fission events.

Figure 6. OxLDL Alters Morphology and Dynamics of the Mitochondria Network.

(A) Examples of confocal images of WT bone marrow-derived macrophages expressing mitochondrial-targeting EYFP. Cells were either maintained in the culture media (control) or treated with 20 μg/ml oxLDL for 24h. Scale bar: 20μM. White rectangle area within the left image was magnified and shown on the right for both conditions. (B) Images from (A) were processed and 3D mitochondrial networks were reconstituted. PHI represents the relative size of the largest connected component to the total mitochondrial size. White arrows point to the three way junctions and white arrow heads to the free ends in the magnified views in (A). The above parameters were quantified by the software MitoGraph and shown in the bar graph. n=25~30 for each group. (C) Examples of Western blots of DRP1, MFN1 and β-actin (loading control) from WT peritoneal macrophages treated with 20 μg/ml oxLDL for indicated time. Images of DRP1 were quantified, normalized by β-actin and expressed as percentage of control; n=3 per group. *p<0.05 compared to control. (D) Examples of EM images of WT peritoneal macrophages treated with culture media (control) or with 50 μg/ml oxLDL for 24h were shown. Mitochondria with abnormal cristae structures were highlighted by black arrows. Scale bar: 500nm.

To further explore the mitochondria structural changes, we performed transmission electron microscopy and found that oxLDL treatment induced loss of mitochondrial cristae structures (Figure 6D; by black arrows). Since cristae harbor all ETC complexes essential for the OXPHOS, the structural defect is consistent with the notion that oxLDL significantly suppressed mitochondria OXPHOS (Figure 3A). Taken together, we demonstrate that oxLDL induces a significant morphological change in macrophage mitochondria network.

CD36 contributes to the mitochondrial metabolic switch under atherogenic conditions in vivo.

To test the physiological relevance of our in vitro findings, we assessed mitochondrial ROS levels in circulating white blood cells from apoe null mice over a 6-week period of HFD. Figure 7A shows a sustained 45-100% increase in levels of mitochondrial ROS beginning 1 week after initiating HFD. This was not observed in cells from apoe/cd36 double null mice. Interestingly, mitochondrial ROS induction was associated with significant increase in surface CD36 expression (Figure 7B). In a parallel study using HFD-fed WT mice we did not detect sustained elevation in WBC mitochondrial ROS (Figure S8A), consistent with the known resistance of this strain to diet-induced hyperlipidemia and atherosclerosis. We found, however, a significant increase in both mitochondrial ROS and CD36 expression at a single early time point (week 3; Figure S8A and S8B). Moreover, plasma pro-inflammatory cytokines including IL-6, TNF-α, and MCP-1 were all elevated at week 3 but returned to the levels seen in chow-fed mice. Taken together, these data suggest a potential link among WBC CD36 expression, mitochondrial ROS stimulation, and plasma pro-inflammatory cytokine levels.

Figure 7. CD36 Directs a Mitochondrial Metabolic Switch in Vivo under Atherogenic Conditions.

(A) WBC mitochondrial ROS levels were measured by flow cytometry in cells from apoe null mice (left panel) and apoe/cd36 double null mice fed chow or high fat diet (HFD) weekly for 6 weeks; n=5 per group. (B) CD36 surface expression levels were assessed in WBC from apoe null mice on chow or HFD weekly for 6 weeks; n=5 per group. (C) Examples of dot plots of WBC from apoe null mice on chow/HFD for 4 weeks. WBCs were stained with MitoNeoD for Mitochondrial ROS and FITC-conjugated anti-Ly6C. Percentage of MitoNeoD⁺Ly6C⁺ and MitoNeoD⁻Ly6C⁺ populations; n=5 per group. (D) apoe null WBCs were stained with PE-conjugated anti-Glut1 and FITC-conjugated anti-Ly6C. Percentage of Ly6C⁺Glut1⁺ population is shown; n=5 per group. (E) Aortic lesional F4/80⁺ cell population was stained with MitoNeoD and FITC-conjugated anti-CD36. A typical dot plot was shown and MitoNeoD MFI were compared between CD36⁺ and CD36⁻ populations. (F) At the conclusion of 6-weeks of HFD, apoe null mice or apoe/cd36 double null mice were sacrificed and peritoneal cells were isolated and analyzed by the Seahorse Mito Stress Test. Representative OCR curves are shown in the upper panels and individual OCR basal levels as well as spare capacity levels in the lower panels as dot plots. (G) Examples of images of anti-FITC and anti-ATP5A Western blots of macrophages from apoe null or apoe/cd36 double null mice collected after the 6-week HFD challenge. ATP5A cysteine modification was assayed by the same procedure as detailed in Figure 3I. *p<0.05, **p<0.01, ***p<0.001 compared to chow diet group.

The circulating Ly6C^high pro-inflammatory monocyte subpopulation is the predominant subpopulation that gives rise to plaque macrophages during atherogenesis in apoe null mice ^{42, 43}. Accordingly, we found that the MitoNeoD⁺ WBCs were predominantly Ly6C^high monocytes (Figure 7C). Furthermore, there was a significant increase in the MitoNeoD⁺Ly6C⁺ subpopulation (14.81%±1.90% in chow diet vs. 21.96%±1.95% in HFD) but not in the MitoNeoD^-Ly6C⁺ subpopulation (11.52%±0.49% in chow diet vs. 12.74%±0.36% in HFD). The continuous elevated mitochondrial ROS in Ly6C⁺ monocytes was accompanied by continuous elevated Glut1 expression (Figure 7D), consistent with our in vitro findings (Figure 3G).

At the conclusion of the 6-week diet challenge the mice were sacrificed and both aorta lesional cells and peritoneal mononuclear cells were harvested and analyzed. Figure 7E shows that, in the lesional F4/80⁺ macrophage populations, close to 3/4 of the cells were positive for both mitochondrial ROS and CD36 expression. Moreover, CD36 positive cells showed about 2 times more mitochondrial ROS signals compared to CD36 negative cells, which indicates a positive association between CD36 expression and mitochondrial ROS. In agreement with our ex vivo data, Figure 7F shows that HFD feeding led to significant reduction in basal OCR compared to chow feeding (0.77±0.10 vs 1.34±0.13 pmol/min/μg protein) as well as OCR spare capacity (1.93±0.42 vs 4.33±0.61 pmol/min/μg protein). Importantly, no significant differences on OCR or OCR spare capacity were detected in cells from HFD-fed apoe null mice that were also cd36 null (Figure 7F). Using the 5-IAF accessibility assay we also found significantly less free thiols in ATP5A immunoprecipitated from cells from HFD-fed apoe null mice, consistent with ROS-induced ATP synthase modification. This was not observed in cells from HFD-fed apoe/cd36 double null mice (Figure 7G) or cells from chow fed or HFD fed WT mice (Figure S8D).

DISCUSSION

Here, we report ex vivo and in vivo data showing that the macrophage scavenger receptor CD36 mediates mitochondrial metabolic reprogramming from oxidative phosphorylation to ROS production to drive chronic inflammation under atherogenic conditions. We further demonstrate that the underlying molecular mechanisms are due to its capacity to import LCFA leading to LCFA accumulation in the mitochondria.

Mitochondrial dysfunction is frequently observed in diseases associated with atherosclerosis in both animal models and human patients ²⁰. One study reported that mitochondria contributed to inflammatory phenotypes in macrophages by producing ROS, which subsequently activated NF-κB pathway ²². On the other hand, reduction in mitochondrial genes, especially those involved in respiration were observed during atherogenesis ^{44, 45}. Nevertheless, despite accumulating evidence implicating mitochondrial function in the development of atherosclerosis, the underlying molecular mechanisms remain not well understood.

In this study, we showed that the atherogenic ligand oxLDL induced a reduction in OXPHOS associated with production of mitochondrial ROS (Figures 1–3). RNA sequencing analysis revealed a unique pattern of transcripts regulating FA metabolism and bioenergetics induced by oxLDL (Figure 4), suggesting that FA metabolism underlies the mitochondrial functional switch. These findings were validated through a variety of methods including qRT-PCR, western blot, assays of functional metabolic flux and ³H-palmitic acid trafficking, all of which established a clear FA trafficking route (Figure S5) responsible for LCFA mitochondrial import. Meanwhile, down-regulation of the TCA cycle and FAO by oxLDL suggests that mitochondrial LCFA accumulation was the combined effect of both LCFA import and reduced ability to catabolize LCFA in the mitochondria. This notion was firmly validated by mitochondrial lipidomics analysis showing significant elevation of C16:0, C18:0, and C18:1 in cells exposed to oxLDL (Figure 5H). These findings are consistent with previous studies in human subjects showing that free FA elevation in plasma induced inflammation through ROS and NF-κB ⁴⁶ and abnormal FA metabolism was linked to mitochondrial dysfunction ⁴⁷. However, it is still not clear how LCFA mitochondrial accumulation leads to ROS production although one study suggested that ETC complex III and the electron transport flavoproteins and the associated oxidoreductase may be the sites of ROS production when LCFA accumulated locally ³⁵. Alternatively, LCFA may directly inhibit ETC complex V creating a condition for reverse electron transfer, which leads to ROS production ⁴⁸. A potential consequence of ROS production in mitochondria is post-translational modification of ETC proteins such as ATP5A (Figure 3I), which significantly reduces the efficiency of OXPHOS (Figure 3A). It is important to note that ATP5A or complex V may not be the only ETC complex affected by ROS as we observed reduction in both basal OCR and maximum OCR, which is uncoupled from complex V activity. Exploration of other ETC complexes that are influenced by ROS under oxLDL-treatment conditions would be interesting in future studies.

Our model of mitochondrial dysfunction is further supported by the observation of altered mitochondrial morphology (Figure 6). We detected a fragmented mitochondrial network induced by oxLDL (Figure 6A). Interestingly, this phenomenon was similar to that observed in cells treated by ETC inhibitors ³⁸, suggesting that oxLDL-altered mitochondrial morphology may be related to suppression of respiration. TEM images (Figure 6D) showing loss of cristae after oxLDL exposure provide another clue as to why mitochondrial respiration was suppressed by oxLDL, since intact cristae is essential for ETC and efficient OXPHOS. We suggest that LCFA accumulation in mitochondria may contribute to the cristae defect because these hydrophobic molecules may insert and disturb inner mitochondrial membrane organization.

Studies in the cancer field show that the FA trafficking property of CD36 in cancer stem cells has a significant impact on cell differentiation and maintenance of “stemness”, thus driving cancer progression ^{49, 50}. In macrophages, CD36-mediated FA trafficking has been implicated in M1/M2 polarization by providing fuel for FAO ⁵¹. Our in vivo data from diet-challenge experiments indicate a strong correlation among CD36 expression, mitochondrial ROS, and the level of inflammatory Ly6C^high monocytes in the circulation (Figure 7). We also detected a similar correlation between CD36 and mitochondrial ROS in lesional macrophages. As the Ly6C^high monocytes predominantly differentiate into pro-inflammatory lesional macrophages during atherogenesis, we propose that CD36 plays a major role in promoting Ly6C^high monocytosis as well as further macrophage differentiation through its FA trafficking function and subsequent mitochondria metabolic switch.

Our findings that deleting the cd36 gene or blocking FA mitochondrial import by etomoxir attenuated mitochondrial ROS as well as inflammatory activation (Figure 1, 5, and 7) may provide a potential strategy against chronic inflammation under atherogenic conditions. However, this will require more knowledge of how immune cells sense and handle intracellular FA imbalance among different organelles. Also, since mitochondrial ROS levels are tightly controlled by multiple layers of anti-oxidant enzymes ⁵², it is possible that FA accumulation may lead to inhibition of some of these enzymes to tip the balance for higher ROS levels. In summary, we demonstrate that oxLDL/CD36 axis couples FA metabolism with mitochondrial function to drive inflammation. These novel insights of a molecular mechanism linking dyslipidemia, oxidative stress, and chronic inflammation may pave the way for new strategies against atherosclerosis.

Novelty and Significance.

What is known?

• Atherosclerosis, a chronic inflammatory disease, is commonly associated with hyperlipidemia and oxidative stress and mitochondrial dysfunction.

• CD36, a long-chain fatty acid (LCFA) transporter, is also a receptor for oxidized LDL (oxLDL) and facilitates pro-atherogenic phenotypes in macrophages and contributes to atherosclerosis in vivo.

What new information does this article contribute?

• Oxidized LDL (oxLDL) up-regulates macrophage CD36 and its downstream fatty acid trafficking system and down-regulates fatty acid oxidation (FAO) as well as mitochondrial oxidative phosphorylation (OXPHOS), leading to mitochondrial long chain fatty acid (LCFA) accumulation.

• OxLDL-induced mitochondrial LCFA accumulation facilitates mitochondrial morphological change and ROS production, which subsequently promote NF-κB activation and production of pro-inflammatory cytokines.

• Mitochondrial ROS generation correlated with CD36 expression in lesional macrophages in a diet-induced atherosclerosis mice model.

Macrophages play a central role in atherosclerosis, a chronic inflammatory disease, which is often accompanied by hyperlipidemia, oxidative stress and mitochondrial dysfunction. Emerging evidence has indicated that metabolic reprogramming of mitochondria from OXPHOS to ROS production in macrophages can be a driving force for pro-inflammatory activation. In this study, we report that atherogenic oxLDL re-organizes fatty acid metabolism in macrophages in a CD36-dependent manner. By next generation RNA sequencing, radioisotope tracing, lipidomics analysis and metabolic functional assays, we found that oxLDL-treated macrophages actively transported LCFA into the mitochondria while suppressing FAO. Combined together, these resulted in LCFA accumulation, mitochondrial morphological change and ROS production, which drove inflammation. Our findings have provided a molecular mechanism linking extracellular hyperlipidemia with oxidative stress and chronic inflammation.

Non-standard Abbreviations and Acronyms:

ACC

acetyl-coA carboxylase

AMPK

AMP-activated protein kinase

ECAR

extracellular acidification rate

ETC

electron transport chain

Eto

etomoxir

FA

fatty acids

FAO

fatty acid oxidation

HFD

high fat diet

HPLC

high performance liquid chromatography

LCFA

long-chain fatty acids

OCR

oxygen consumption rate

oxLDL

oxidized LDL

OXPHOS

oxidative phosphorylation

PA

palmitic acid

PDHK1

pyruvate dehydrogenase kinase 1

ROS

reactive oxygen species

TMRM

tetramethylrhodamine, methyl ester