RAGE Suppresses ABCG1-Mediated Macrophage Cholesterol Efflux in Diabetes

Reagents

The following materials were purchased: apolipoprotein A1 (apoA1), HDL, and acetylated LDL (Biomedical Technologies, Inc.); fatty acid–free BSA (Equitech-Bio); T 0901317 (Tocris Bioscience); human plasma LDL (Sigma-Aldrich); assays for measurements of HDL, total cholesterol, and triglycerides (Wako Diagnostics); rosiglitazone (Sigma-Aldrich); and U0126 and PD98509 (Cell Signaling).

Animal Studies

Male homozygous Ager^−/− (RAGE) mice (C57BL/6 Ager^−/− mice backcrossed >12 generations into C57BL/6J [The Jackson Laboratory]) and littermate Ager-expressing control mice were used. Male mice (wild type [WT] and devoid of Ager) were rendered diabetic with streptozotocin (STZ) (Sigma-Aldrich). For high-fat diet (HFD) and low-fat diet (LFD) experiments, mice at the age of 6–8 weeks were fed HFD with 60% of calories from lard (D12492; Research Diets, Inc.) or LFD with 13% of calories from fat (5053, PicoLab Rodent Diet 20; LabDiet) for at least 3 months. Ldlr^−/− mice devoid of or expressing Ager were rendered diabetic at age 7 weeks and then placed on a Western diet (0.15% cholesterol) (D01061401C; Research Diets, Inc.) for 15 weeks followed by harvest of the aortic arches at age 22 weeks. All animal procedures were approved by the Institutional Animal Care and Use Committees of Columbia University and New York University (NYU) and performed in accordance with the National Institutes of Health animal care guidelines.

Cell Culture

Human THP-1 peripheral blood monocytic cells, L929 cells, and human embryonic kidney (HEK) 293T cells were obtained from ATCC and cultured per the manufacturer’s instructions. THP-1 cells were used as suspension cells throughout this study.

Isolation of Murine Macrophages

For primary murine bone marrow–derived macrophages (BMDMs), bone marrow was retrieved after kill from bilateral femora, and BMDMs were cultured as previously described (16) and used on day 7 of incubation. BMDMs retrieved from nondiabetic mice were subsequently grown in 5.5 mmol/L d-glucose, and BMDMs retrieved from diabetic mice were grown in 25 mmol/L d-glucose to mimic their condition of origin. In other studies, BMDMs from nondiabetic mice were cultured in 5.5 or 25 mmol/L d-glucose for 7 days postisolation and before studies. Peritoneal macrophages were isolated as previously described without the use of thioglycollate (17).

RNA Isolation and Quantitative Real-Time PCR

Quantitative real-time PCR was performed on isolated RNA using TaqMan Fast Universal Master Mix (2X) with premade primer sets (Life Technologies) (Supplementary Table 1).

RNA Interference Silencing

Small interfering RNA (siRNA) to reduce levels of AGER (ID# 110857; sense [5′–3′: CGGCUGGUGUUCCAAUAAtt] and antisense [5′–3′: UUAUUGGGAACACCAGCCGtg]) and scramble negative control siRNA (ID# AM435) were purchased from Life Technologies.

Overexpression of Peroxisome Proliferator–Activated Receptor-γ and ABCG1 in THP-1 Cells

Human peroxisome proliferator–activated receptor-γ (PPARG) cDNA cloned into pCMV-SPORT6 plasmid was obtained from Thermo Scientific, and human ABCG1 cDNA cloned into pCMV6-AC-GFP plasmid was purchased from OriGene.

Cholesterol Efflux Assays

Primary murine BMDMs or THP-1 cells were labeled with 5 μCi/mL ³H-cholesterol (Perkin Elmer) in the presence of acetylated LDL (50 mg/L) and fatty acid–free BSA (1%) for 24 h. After washing the cells and achieving equilibrium for 1 h with fatty acid–free BSA (1%), cells were treated with either apoA1 or HDL. After 6 h of incubation at 37°C, the supernatant was collected. Cells were washed and lysed; aliquots of both supernatant and medium were measured in a multipurpose scintillation counter (LS6500; Beckman Coulter). The percent cholesterol efflux was calculated by dividing the radioactivity in the supernatant by the sum of the radioactivity measured in the supernatant and the cell lysates.

Western Blot Analysis

The following antibodies were used: rabbit anti-ABCG1 (NB400-132, 1:500; Novus Biologicals); rabbit anti-ABCA1 (NB400-105, 1:500; Novus Biologicals); mouse anti-tubulin (clone KMX-1, catalog MAB3408, 1:500; Millipore); mouse anti-β-actin (A1978, 1:4,000; Santa Cruz Biotechnology); rabbit anti-PPARG (ab41928, 1:1,000; Abcam); rabbit anti-RAGE (GTX23611, 1:1,000; GeneTex); rabbit anti-apoA1 (ab20453, 1:1,000; Abcam); rabbit anti-p44/42 mitogen-activated protein kinase (MAPK) (Erk1/2) (9102, 1:1,000; Cell Signaling); and mouse anti-phospho-p44/42 MAPK (Erk1/2) (9106, 1:1,000; Cell Signaling). After incubation with the indicated secondary antibodies, protein signals were visualized with an enhanced chemiluminescence detection system (Thermo Scientific) using the Odyssey Fc Dual-Mode Imaging system (LI-COR) or IRDye 680RD goat anti-mouse (1:25,000) (LI-COR) and IRDye 800CW goat anti-rabbit (1:10,000; LI-COR), and protein signals were visualized using the Odyssey Infrared Imaging System Model 9120 (LI-COR).

Luciferase Reporter Assays

Human ABCG1 promoter A region spanning −1,104 to +37 base pairs (upstream of exon 1) cDNA was generated following published methods by Uehara et al. (18) and cloned into pCR-script vector. Subsequently, this cDNA for human ABCG1 promoter A region (−1,104/+37) was cloned into the pGL3-basic vector using restriction sites KpnI and HindIII. Before transfection, HEK293T cells were grown to 50–80% confluence on 12-well plates in DMEM supplemented with 10% FBS + 1% penicillin-streptomycin. Cells were transiently transfected with 1 μg mouse Ager cDNA in pcDNA3 vector on day 1, and then on day 2, cells were transiently transfected with 1 μg pGL3 luciferase firefly vector (Promega) containing human ABCG1 promoter A or human ABCA1 promoter cDNA (gift from Antonio Castrillo, Instituto de Investigaciones Biomédicas “Alberto Sols”), and 1 ng pRL-TK vector Renilla (Promega) (to normalize transfections) using Lipofectamine 2000 (Invitrogen). Luciferase activities were measured using the Dual-Luciferase Reporter Assay System (Promega) and were divided by Renilla activity to obtain a normalized value (in relative luciferase units). ABCG1 deletion mutants were prepared as described by Uehara et al. Site-directed mutations to PPAR response element (PPRE) 1 (sense) and PPRE2 (antisense) sites (single mutations and a double mutation) on the human ABCG1 promoter were constructed using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies).

Bioinformatics

The −413 to −303 region of the ABCG1 promoter was compared with the high-quality vertebrate subset of Transfac Professional version 2013.2 (19) using the Match program (20) set to minimize false-positive results. Binding sites of 19 unique transcription factors were found. Among these were binding sites for Pparg (V$PPARG_02) (21,22). There are two V$PPARG_02 sites, both of which were at −330, one in the sense direction and the other in the antisense direction. The most conserved regions of the site core sequence were used to design single–base pair substitutions to interfere with binding. In the following text, the core recognition element is capitalized and the replacement nucleotide is indicated in boldface. On the sense strand, the following substitutions were made:

• S1: −325G→C

• gagatGGGTAgattttcctactt

• gagatCGGTAgattttcctactt

For antisense, the following substitutions were made:

• A1: −313C→G

• gagatgggtagatTTTCCtactt

• gagatgggtagatTTTCGtactt

Electrophoretic Mobility Shift Assays

Nuclear protein extracts were prepared from THP-1 cells, and electrophoretic mobility shift assays (EMSAs) were performed using the Odyssey EMSA Buffer Kit (LI-COR) and 10X orange loading dye (LI-COR) according to the manufacturer’s instructions. Gels were visualized using a LI-COR Odyssey imager. The PPARG DNA probe sequences used were as follows:

• Sense strand: 5′-/5IRD700/GAGATGGGTAGATTTTCCTACTT-3′

• Antisense strand: 5′-/5IRD700/CTCTACCCATCTAAAAGGATGAA-3′

HDL Cholesterol, Total Cholesterol, and Triglyceride Measurements

HDL cholesterol and total cholesterol concentrations were measured using colorimetric assays (Wako Diagnostics); triglyceride levels were measured enzymatically using a colorimetric assay (L-Type TG M Enzyme Color A; Wako Diagnostics).

Chromatin Immunoprecipitation

Chromatin immunoprecipitation (ChIP) assays were performed using the ChIP-IT Express enzymatic kit (Active Motif). ChIP-enriched DNA (2 μL) was used for quantitative real-time PCR for the ABCG1 promoter using SYBR Green reagents and the following primers: forward primer, 5′-TTTGCCGTAATTGTTTTCAATG-3′; reverse primer, 5′-GCAGGGTTACTAAAGGGCAGT-3′.

Laser Capture Microdissection

Aortic arches from the indicated Ldlr^−/− or Ldlr^−/−/Ager^−/− mice were subjected to laser capture microdissection (LCM) of CD68-expressing cells (detected by rat anti-mouse CD68 clone FA-11 (MCA1957; AbD Serotec) using a Leica LMD6500 Laser Microdissection System and quantitative real-time PCR for measuring Abca1, Abcg1, and Pparg mRNA transcripts.

Oil Red O Staining

Oil red O staining of atherosclerotic lesions was performed according to the manufacturer’s instructions (AMTS Inc.) (23).

Macrophage RCT

Male WT C57BL/6 or littermate Ager^−/− mice were rendered diabetic with STZ for 2 months, and bone marrow was collected. BMDMs were prepared and radiolabeled with 5 μCi/mL ³H-cholesterol and acetylated LDL (50 μg/mL) for 24 h. BMDMs were washed and injected into WT C57BL/6 nondiabetic mice. On the day of injection, recipient mice were housed individually with unlimited access to food and water. ³H-cholesterol/acetylated LDL–loaded BMDMs (2.5–3.5 × 10⁶/mL) were injected intraperitoneally. Blood was collected and plasma subjected to counting in a β-counter. Feces were collected continuously for 64 h. At 64 h, mice were humanely killed, and liver tissue was collected for further analyses. Results were expressed as a percentage of the counts per minute (cpm) injected as previously described (24).

Statistics

All data are reported as mean ± SEM unless otherwise stated (n is noted in figure legends). Data were analyzed with the Student t test or by ANOVA followed by two-tailed distribution t test. P ≤ 0.05 was considered statistically significant. Cholesterol efflux and RCT were log-transformed before analysis to ensure approximate normality and were analyzed in R. Experiments at each time point were analyzed separately to characterize the effect of each genotype/disease state combination on RCT at that time. Additionally, each genotype/disease state combination experiment was analyzed separately at different times to characterize the effect of time on RCT for that combination. Both the Fligner-Killeen test (25) and visual inspection showed heteroscedasticity, so the Welch test for the comparison of several means (26) was used instead of standard ANOVA to estimate significance. Significance of comparisons was estimated using the Westfall method (27,28) as implemented in multcomp (29), with heteroscedasticity accounted for by a Sandwich estimator (30) as implemented in Sandwich (31). The figures display means and SEs back-transformed to percentages. Supplementary Table 2 presents the Welch test P value, which is the probability that at least one of the conditions is not the same as the other. Supplementary Table 2 also gives the probability of (Ager^−/− diabetic mice [DM] − Ager^−/− nondiabetic mice [NDM]) − (WT DM − WT NDM), differing from 0 by chance (i.e., the probability that the effect of diabetes on RCT in Ager^−/− mice differs from the effect of diabetes in WT mice). By simple algebraic manipulation, this expression is also equal to (Ager^−/− DM – WT DM) – (Ager^−/− NDM − WT NDM), which is the effect of deletion of Ager in diabetic mice relative to the effect on nondiabetic mice. This later expression shows whether the comparison of the effect of deletion of Ager in diabetic mice differs from the corresponding effect in nondiabetic mice.

RAGE Impairs Macrophage Cholesterol Efflux to ApoA1 and HDL in Diabetes

Primary BMDMs were retrieved from WT male C57BL/6 mice after 2 months of hyperglycemia induced by STZ. Controls were age- and sex-matched nondiabetic mice treated with intraperitoneal injections of the vehicle for STZ citrate buffer. BMDMs were cultured in concentrations of glucose consistent with the glycemic state of the mice from which they were retrieved. WT BMDMs retrieved from diabetic animals displayed significantly reduced cholesterol efflux to apoA1 and HDL versus BMDMs from nondiabetic WT mice (P < 0.01 and P < 0.001, respectively) (Fig. 1A and B and Supplementary Table 2). In the nondiabetic state, although deletion of Ager imparted no significant effect on cholesterol efflux to apoA1 versus WT BMDMs, significantly higher cholesterol efflux to HDL was observed (P < 0.0001) (Fig. 1A and B and Supplementary Table 2). In the diabetic state, deletion of Ager resulted in significantly higher cholesterol efflux to both apoA1 and HDL compared with WT BMDMs (P < 0.001) (Fig. 1A and B and Supplementary Table 2).

Because it has been reported that mRNA transcript and protein levels of Abca1 and Abcg1, principal regulators of cholesterol efflux to apoA1 and HDL, respectively, are reduced in murine and human diabetes (9,10,32,33), we tested the role of RAGE. The mRNA transcript and protein levels of Abca1 were lower after 2 months of hyperglycemia in WT BMDMs (Fig. 1C and D). In the nondiabetic state, Abca1 mRNA transcript levels did not differ significantly between WT and Ager^−/− BMDMs, but ABCA1 protein levels were significantly higher in Ager^−/− versus WT BMDMs. In the diabetic state, higher levels of Abca1 mRNA transcripts and protein were observed between Ager^−/− and WT BMDMs (Fig. 1C and D).

Diabetes resulted in a significant reduction in Abcg1 mRNA transcripts and ABCG1 protein in WT BMDMs versus the nondiabetic state (Fig. 1E and F). In both nondiabetic and diabetic BMDMs, deletion of Ager resulted in significantly higher levels of Abcg1 mRNA transcripts and ABCG1 protein than in WT BMDMs (Fig. 1E). These data indicate that diabetes exerted significant effects on the downregulation of ABCA1 and ABCG1 transporters through RAGE.

To directly test the effects of glucose, we incubated nondiabetic mice BMDMs from WT mice in high glucose (25 mmol/L) for 7 days versus normal glucose levels (5.5 mmol/L) and determined Abca1 and Abcg1 mRNA levels. No significant differences were observed in mRNA levels in nondiabetic WT BMDMs cultured for 7 days in 5.5 or 25 mmol/L d-glucose (Supplementary Fig. 1A and B).

Additionally, we retrieved primary peritoneal macrophages (without thioglycollate treatment) and prepared mRNA immediately upon retrieval. Compared with WT peritoneal macrophages from nondiabetic mice, WT diabetic peritoneal macrophages displayed significantly lower levels of Abca1 and Abcg1 mRNA transcripts. In the nondiabetic state, a significant increase in Abcg1 mRNA transcripts was observed in Ager^−/− versus WT peritoneal macrophages, and in the diabetic state, Ager^−/− peritoneal macrophages displayed significantly higher levels of both Abca1 and Abcg1 mRNA transcripts (Supplementary Fig. 1C and D). These results are completely analogous to those observed in primary murine BMDMs (Fig. 1C and E).

AGE-RAGE Interaction Suppresses Cholesterol Efflux to ApoA1 and HDL in Human THP-1 Cells

To ensure that the results applied to cells that normally express RAGE and are of human origin, we treated THP-1 cells with carboxymethyllysine (CML)-AGE, which was previously shown to be a specific AGE ligand of RAGE that accumulates in human atherosclerotic lesions (34). Compared with vehicle/scramble siRNA treatment, CML-AGE resulted in significantly reduced cholesterol efflux to apoA1 and HDL (P < 0.0001) (Fig. 2A and B). RNA interference (RNAi) knockdown of AGER in CML-AGE–treated THP-1 cells resulted in significantly higher cholesterol efflux to apoA1 (P < 0.01) and HDL (P < 0.0001) compared with scramble siRNA/CML treatment (Fig. 2A and B).

Treatment of THP-1 cells with CML-AGE in the presence of scramble siRNA resulted in significant suppression of ABCA1 and ABCG1 mRNA transcripts (P < 0.01) (Fig. 2C and D). RNAi knockdown of AGER resulted in significantly higher levels of ABCA1 and ABCG1 mRNA transcripts versus scramble siRNA/CML-AGE treatment (P < 0.001 and P < 0.01, respectively) (Fig. 2C and D).

At the protein level, Western blotting revealed a trend to lower levels of ABCA1 protein in CML-AGE–treated THP-1 cells; in the presence of RNAi knockdown of AGER, ABCA1 protein levels were similar to those in non-CML-AGE–treated cells (Fig. 2E). In the case of ABCG1, treatment with CML-AGE resulted in a significant suppression of ABCG1 protein levels versus vehicle (P < 0.01) (Fig. 2F). RNAi knockdown of AGER in the presence of CML-AGE resulted in significantly higher levels of ABCG1 protein compared with CML-AGE treatment (P < 0.0001) (Fig. 2F). Thus, in human RAGE-expressing THP-1 cells, RAGE ligand CML-AGE reduces cholesterol efflux to apoA1 and HDL at least in part through regulation of ABCA1 and ABCG1.

RAGE Mediates Impaired RCT in Diabetic Macrophages

Based on in vitro findings that RAGE contributes to suppression of macrophage cholesterol efflux to apoA1 and HDL, particularly in diabetes, we tested the effects of RAGE on macrophage RCT in vivo. WT and Ager^−/− mice were rendered hyperglycemic with STZ or treated with the control citrate buffer (nondiabetes) for 2 months. BMDMs were isolated, loaded with radiolabeled cholesterol, and injected into WT nondiabetic mice. As shown in Fig. 3A and B, in WT mice, diabetes resulted in reduced macrophage RCT to plasma at 43 and 64 h postinjection of WT BMDMs (P < 0.001, and P < 0.01, respectively). In nondiabetic BMDMs, no statistically significant effects of Ager deletion were noted at 43 or 64 h postinjection (Fig. 3A and B). In diabetic BMDMs, however, deletion of Ager resulted in significantly higher RCT to plasma at 43 and 64 h postinjection (P < 0.001 and P < 0.01, respectively) (Fig. 3A and B and Supplementary Table 2).

Diabetes resulted in a trend to reduced macrophage RCT to the liver versus the nondiabetic state (Fig. 3C and Supplementary Table 2). Although deletion of Ager had no significant effect on macrophage RCT to the liver in the nondiabetic state, significantly higher RCT in diabetic BMDMs devoid of Ager was noted compared with diabetic WT BMDMs (P < 0.01) (Fig. 3C and Supplementary Table 2). In feces of WT mice, diabetes resulted in significantly reduced macrophage RCT (P < 0.01). Although deletion of Ager in nondiabetic BMDMs resulted in modest but significantly lower RCT to feces versus WT nondiabetic BMDMs, trends to higher RCT to feces were observed in diabetic BMDMs devoid of Ager compared with diabetic WT BMDMs (Fig. 3D and Supplementary Table 2).

We measured plasma levels of HDL and found that diabetes resulted in significantly lower levels of HDL in WT mice than in age-matched WT nondiabetic mice (35.1 ± 0.8 vs. 41.7 ± 2.3, P < 0.01) (Fig. 3E). In the nondiabetic state, levels of HDL did not differ by Ager genotype (41.7 ± 2.3 vs. 42.3 ± 1.4 in WT nondiabetic vs. Ager^−/− nondiabetic mice, respectively; P not significant) (Fig. 3E). In contrast, diabetic Ager^−/− mice displayed significantly higher levels of HDL versus age-matched WT diabetic mice (47.2 ± 0.4 vs. 35.1 ± 0.8, P < 0.0001) (Fig. 3E). Levels of total cholesterol were tested (Supplementary Table 3). In the nondiabetic state, no significant effects of Ager genotype were seen. Total levels of cholesterol were higher in diabetic versus nondiabetic WT mice (120.7 ± 4.0 vs. 96.4 ± 3.2, P < 0.001). Diabetic Ager^−/− mice displayed significantly lower levels of total cholesterol than their diabetic WT counterparts (95.4 ± 2.9 vs. 120.7 ± 4.0, P < 0.001) (Supplementary Table 3).

On the basis of these findings of higher levels of plasma HDL in diabetic mice devoid of Ager, we tested mRNA transcript and protein levels of Apoa1 and Abca1 in mouse liver tissue because apoA1 and ABCA1 are critical for HDL production (35). Ager genotype had no effect on mRNA transcript levels of Apoa1 (Supplementary Fig. 2A) or Abca1 (Supplementary Fig. 2B) in nondiabetic mice. However, diabetes resulted in a significant reduction in Apoa1 and Abca1 mRNA transcripts versus the nondiabetic state in WT mice (P < 0.05). Levels of Apoa1 and Abca1 mRNA transcripts were lower in diabetic Ager^−/− versus diabetic WT liver tissue (P < 0.001 and P < 0.05, respectively) (Supplementary Fig. 2A and B). Deletion of Ager resulted in increased apoA1 protein levels in nondiabetic mouse livers (P < 0.01), with trends to increased protein levels of apoA1 in diabetic mouse liver devoid of Ager versus WT (Supplementary Fig. 2C). No difference was observed in ABCA1 protein levels between nondiabetic Ager^−/− and WT liver tissue, whereas protein levels were significantly higher in diabetic Ager^−/− liver than in diabetic WT (P < 0.05) and nondiabetic Ager^−/− (P < 0.05) livers (Supplementary Fig. 2D). These findings suggest that the higher levels of HDL in mice devoid of Ager, particularly in the diabetic state, are likely accounted for by significantly higher levels of ABCA1 protein in the liver.

AGE-RAGE Interaction and Molecular Regulation of ABCG1

To define the molecular mechanisms accounting for these findings, human ABCA1 and ABCG1 promoter reporter luciferase constructs were transfected into AGER-transfected HEK293T cells (HEK293T cells do not natively express RAGE); treatment of these cells with CML-AGE resulted in significant suppression of both ABCA1 (P < 0.05) (Supplementary Fig. 3A) and ABCG1 (P < 0.001) (Fig. 4A) promoter luciferase activity. We focused on the effects of RAGE on molecular regulation of ABCG1 because the greatest reduction in transcription of the two transporters by CML-AGE/RAGE in macrophages was on ABCG1 and because macrophage cholesterol efflux to HDL was most potently affected by RAGE. Supporting the RAGE dependence of these findings on CML-AGE–mediated suppression of ABCG1 promoter luciferase activity in the absence of AGER transfection in HEK293T cells, CML-AGE exerted no suppressive effects on ABCG1 promoter luciferase activity (Supplementary Fig. 3B). Furthermore, the effect of CML-AGE treatment on suppression of the ABCG1 promoter luciferase activity in AGER-expressing HEK293T cells was prevented by incubation with sRAGE, the extracellular ligand-binding domain of RAGE (P < 0.0001) (Fig. 4A).

A series of deletion mutants of the human ABCG1 promoter (1,104/+37) was prepared as described by Uehara et al. (18). CML-AGE treatment resulted in significant suppression of promoter luciferase activity in full-length and deletion mutants A, B, and C (Fig. 4B). However, in deletion mutants D, E, F, and G, there were no suppressive effects of CML-AGE, thereby implicating elements within C but not D in the suppressive effects of CML-AGE on promoter reporter luciferase activity. As shown in Supplementary Table 4, multiple putative transcription factor binding sites within −413 to −304 were identified. Of note, NRLH3 (liver X receptor [LXR]-α) putative binding sites were not identified within this region (36). Consistent with this finding, diabetes exerted no significant effect on mRNA transcript levels of Nr1h3 (Lxr-α) or Nrlh2 (Lxr-β) compared with nondiabetic BMDMs (Supplementary Fig. 4A and B). Additionally, CML-AGE exerted no suppressive effect on LXR agonist T0901317–mediated increase in ABCA1 (Supplementary Fig. 5A) or ABCG1 (Supplementary Fig. 5B) promoter reporter luciferase activity, suggesting a mechanism of action independent of LXRs.

Given the known potent effects of PPARG on macrophage properties (37–39), we determined whether putative PPREs within the human ABCG1 promoter were linked to the suppressive effects of CML-AGE. Although LXR putative binding sites were not observed in the region of interest (−413 to −304) within the ABCG1 promoter, we identified at least two putative PPREs in this region. First, to directly assess the binding of RAGE-expressing THP-1 cell nuclear extract to the PPRE1- and PPRE2-containing binding elements on the −413 to −304 promoter region, we performed EMSAs using a consensus probe incorporating both PPRE sites. As shown in Fig. 4C, CML-AGE reduced nuclear protein binding to the ABCG1 promoter elements containing the two indicated PPREs in a manner prevented by sRAGE. Of note, the supershift experiments demonstrated the specific bands of interest by virtue of shifting (reduced intensity) in the presence of anti-PPARG IgG (Fig. 4C).

Second, we performed site-directed mutagenesis of PPRE sense (gagatGGGTAgattttcctactt) and PPRE antisense (gagatgggtagatTTTCCtactt) and tested promoter reporter luciferase activity in CML-AGE versus vehicle-treated RAGE-expressing HEK293T cells. As shown in Fig. 4D, mutation of PPRE antisense (mutation 2) and both PPRE antisense and PPRE sense (mutation 2 and 1) prevented the suppressive effects of CML-AGE on the promoter reporter luciferase activity in the −413 to −304 region. Taken together, these results demonstrate that CML-AGE suppresses ABCG1 promoter activity in a manner that can be prevented by sRAGE and that CML-AGE mediates suppressive effects on the −413 to −304 region on the ABCG1 promoter in a manner involving the antisense (TTTCC) core motif recognized by PPARG.

Finally, we performed ChIP assays to directly study the effects of CML-AGE on PPARG transcription factor recruitment to the −413 to −304 ABCG1 promoter region. A robust signal of ABCG1 mRNA transcription by quantitative real-time PCR was noted in THP-1 under basal conditions; when THP-1 cells were treated with CML-AGE, this signal was significantly reduced by >50% (P < 0.05) (Fig. 4E). Treatment with sRAGE significantly rescued the suppressive effects of CML-AGE on ABCG1 mRNA transcription levels in the ChIP assay (P < 0.001) (Fig. 4E). Thus, CML-AGE suppresses PPARG recruitment to the ABCG1 promoter locus in the promoter region −413 to −304 in a RAGE-dependent manner. Consistent with the principal effects of CML-AGE/RAGE on PPARG-dependent mechanisms, treatment of CML-AGE–treated THP-1 cells with the PPARG agonist rosiglitazone (10 μmol/L) partially, but significantly, rescued the CML-AGE–mediated suppressive effects on ABCG1 mRNA transcripts (P < 0.01) (Supplementary Fig. 5C).

CML-AGE/RAGE and Modulation of Levels of PPARG in BMDMs and RAGE-Expressing THP-1 Cells

These findings prompted us to test whether CML-AGE affected levels of PPARG. First, WT BMDMs treated with CML-AGE displayed a reduction in Pparg mRNA (Fig. 5A) and PPARG protein (Fig. 5B) compared with vehicle-treated WT cells (P < 0.01 and P = 0.05, respectively). In contrast, when Ager^−/− BMDMs were treated with CML-AGE, no reduction in Pparg mRNA or PPARG protein levels was noted (Fig. 5A and B). Similarly, in CML-AGE–treated THP-1 cells, treatment with siRNA-AGER versus siRNA-scramble prevented the effects of CML-AGE on reducing levels of PPARG mRNA and PPARG protein (P < 0.05 and P < 0.01, respectively) (Fig. 5C and D).

To examine the molecular mechanisms by which CML-AGE suppresses PPARG binding to the ABCG1 promoter, we assessed the MAPK pathway because it has been previously demonstrated to be an important regulator of PPARG expression (40). Treatment of THP-1 cells with CML-AGE resulted in increased phospho/total ERK versus treatment with vehicle (P < 0.05) (Fig. 5E). Pretreatment of THP-1 cells with the selective MAPK inhibitor U0126 or PD98059 prevented the effects of CML-AGE on suppression of PPARG mRNA versus vehicle (Fig. 5F). Thus, the molecular mechanism by which CML-AGE suppresses PPARG is at least partly through the MAPK pathway.

Rescue of CML-AGE Modulation on THP-1 Cell Cholesterol Efflux and Expression of ABCG1 and Modulation of Pparg mRNA in Atherosclerotic Plaques in Diabetic LDLr^−/− Mice

To definitively link the suppressive effects of CML-AGE on cholesterol efflux to HDL to regulation of ABCG1, we overexpressed human ABCG1 in THP-1 cells and treated the cells with CML-AGE versus vehicle. CML-AGE treatment resulted in reduced cholesterol efflux to HDL (P < 0.01) (Fig. 6A), which was prevented by ABCG1 overexpression compared with CML-AGE treatment alone (P < 0.0001) (Fig. 6A). To unequivocally link PPARG reduction to the suppression of ABCG1 transcription, we overexpressed human PPARG in THP-1 cells to determine whether this rescued the suppressive effects of CML-AGE on ABCG1 and ABCA1 mRNA transcript levels. Although CML-AGE resulted in reduced ABCG1 and ABCA1 mRNA transcripts versus vehicle in THP-1 cells (P < 0.001 and P < 0.0001, respectively) (Fig. 6B and C), transient transfection with PPARG cDNA prevented the suppressive effects of CML-AGE because levels of ABCG1 were about fivefold higher and levels of ABCA1 were about fourfold higher in PPARG-overexpressed cells versus CML-AGE treatment alone (P < 0.0001 in each case) (Fig. 6B and C).

Next, to test whether deletion of Ager modulates lesional macrophage levels of Abca1, Abcg1, and Pparg in diabetic atherosclerosis (12), mice devoid of Ldlr, either expressing or devoid of Ager, were rendered diabetic with STZ. Mice were fed a Western diet for 15 weeks, which has been shown to induce insulin resistance (41). The accumulation of fats, lipids, and CD68-expressing macrophages was significantly reduced in atherosclerotic plaques from diabetic Ldlr^−/− mice devoid of Ager versus Ldlr^−/− mice expressing Ager (Fig. 7A). LCM CD68-expressing macrophages from the aortic plaques of diabetic Ldlr^−/− mice devoid of Ager displayed higher levels of Abca1 mRNA transcripts (P = 0.05) and significantly higher levels of Abcg1 and Pparg mRNA transcripts than macrophages retrieved from atherosclerotic plaques of diabetic Ager-expressing Ldlr^−/− mice (P < 0.05) (Fig. 7B). The beneficial effects of Ager deletion were independent of the levels of glucose, total cholesterol, or total triglycerides; levels of plasma HDL were higher in Ldlr^−/− mice than in Ager-expressing mice devoid of Ldlr (P < 0.05) (Table 1).

Finally, to complement the findings in type 1 diabetes, we performed additional experiments to test the role for RAGE in the regulation of Abca1, Abcg1, and Pparg in BMDMs from mice fed an HFD versus an LFD for 3 months. Compared with LFD, HFD feeding in WT mice resulted in significant suppression of Abca1 and significant, but modestly higher levels of Abcg1 and Pparg mRNA transcripts (Supplementary Fig. 7A–C). In BMDMs of HFD-fed Ager^−/− mice, mRNA transcript levels of Abca1, Abcg1, and Pparg were significantly higher than those observed in WT HFD-fed mice (Supplementary Fig. 7A–C).