RAGE impairs murine diabetic atherosclerosis regression and implicates IRF7 in macrophage inflammation and cholesterol metabolism

of arches from Western —/— into —/— ( Ager , the gene encoding RAGE) versus WT diabetic recipient mice accelerated regression of atherosclerosis. RNA-sequencing experiments traced RAGE-dependent mechanisms principally to the recipient macrophages and linked RAGE to interferon signaling. Specifically, deletion of Ager in the regressing diabetic plaques downregulated interferon regulatory factor 7 ( Irf7 ) in macrophages. Immunohistochemistry studies colocalized IRF7 and macrophages in both murine and human atherosclerotic plaques. In bone marrow–derived macrophages (BMDMs), RAGE ligands upregulated expression of Irf7 , and in BMDMs immersed in a cholesterol-rich environment, knockdown of Irf7 triggered a switch from pro- to antiinflammatory gene expression and regulated a host of genes linked to cholesterol efflux and homeostasis. Collectively, this work adds a new dimension to the immunometabolic sphere of perturbations that impair regression of established diabetic atherosclerosis and suggests that targeting RAGE and IRF7 may facilitate vascular repair in diabetes. It was in these contexts that we hypothesized a role for the receptor for advanced glycation end-products (RAGE) in impaired regression of atherosclerosis. RAGE, a member of the immunoglobulin superfamily, is expressed on monocytes/macrophages and vascular cells and has been implicated in atherosclerosis progression in mice (20–24). We employed a model of transplantation of donor atherosclerotic aortic arches (25) from mice devoid of the LDL receptor ( Ldlr ) and fed a Western diet into normolipidemic mice to test the premise that Ager ( Ager , the gene encoding RAGE) contributes to macrophage dysfunction and impairs regression of atherosclerosis. We demonstrate that compared with WT recipient mice, deletion of Ager enhanced regression of atherosclerosis. RNA-sequencing of donor versus recipient diabetic macrophages after aortic arch transplantation suggested that Ager deletion downregulated interferon signaling, and, specifically, interferon regulatory factor 7 (IRF7). In macrophages, IRF7 regulated cholesterol metabolism and inflammation. Targeting RAGE and IRF7 might represent an important ther-apeutic opportunity to accelerate atherosclerosis regression. BMDMs imparts an antiatherogenic profile and downregulation of BMDM Irf7 reduced total cellular cholesterol in hyperlipidemic serum. Although global deletion of Irf7 reduced high fat diet–induced obesity, the proximate mechanisms were not elucidated The present body of work adds RAGE-de-pendent regulation of IRF7 to this network of molecules bridging immune and cholesterol metabolism functions in macrophages, which may contribute to impaired regression of atherosclerosis. Miao and colleagues exposed human THP1 monocytes to high glucose and showed that this upregulated IRF7 which was analogous to an IRF7-driven inflammatory network that was linked to the risk of type 1 diabetes High glucose upregulates ligands of RAGE. Mediating roles for AGEs were demonstrated in our studies by improved regression of diabetic atherosclerosis in transgenic mice overexpressing Glo1 , the enzyme that detoxifies pre-AGE species. It is plausible that non-AGE ligands of RAGE contribute to upregulation of Irf7 . Hyperlipidemic serum from Ldlr —/— mice contains TLR ligands, which also interact with RAGE. In vitro, CpG oligodeoxynucleotides and Complement C3a increased production IFN- α in WT but not Ager —/— BMDMs. In human peripheral blood mononuclear cells, treatment with C3a and CpG oligodeoxynucleotides increased IFN- α , which was reduced by RAGE antagonism of Irf7 (61, 62). These considerations place RAGE and IRF7 at critical nodes in the mediation of chronic inflammation. Here, we show that RAGE impairs regression of atherosclerosis through perpetuation of inflammation and through impairment of cholesterol metabolism, at least in part through IRF7. Interruption of RAGE/IRF7 may restore reparative mechanisms and accelerate regression of atherosclerosis, particularly in diabetes.


Introduction
Atherosclerosis, a leading cause of morbidity and mortality, begins in childhood. Fatty streaks are found in early life and may undergo progression, arrest, or regression during their natural history (1). Although pharmacological approaches have yielded success in mediating regression of established atherosclerotic plaques in humans (2), in people with diabetes, lipid-and glucose-lowering therapies are less effective (3). These observations are recapitulated in murine models of diabetic atherosclerosis regression (4)(5)(6)(7).
Despite advances in lipid-lowering therapies, people with diabetes continue to experience more limited cardiovascular benefits. In diabetes, hyperglycemia sustains inflammation and preempts vascular repair. We tested the hypothesis that the receptor for advanced glycation end-products (RAGE) contributes to these maladaptive processes. We report that transplantation of aortic arches from diabetic, Western diet-fed Ldlr -/mice into diabetic Ager -/-(Ager, the gene encoding RAGE) versus WT diabetic recipient mice accelerated regression of atherosclerosis. RNA-sequencing experiments traced RAGE-dependent mechanisms principally to the recipient macrophages and linked RAGE to interferon signaling. Specifically, deletion of Ager in the regressing diabetic plaques downregulated interferon regulatory factor 7 (Irf7) in macrophages. Immunohistochemistry studies colocalized IRF7 and macrophages in both murine and human atherosclerotic plaques. In bone marrow-derived macrophages (BMDMs), RAGE ligands upregulated expression of Irf7, and in BMDMs immersed in a cholesterol-rich environment, knockdown of Irf7 triggered a switch from pro-to antiinflammatory gene expression and regulated a host of genes linked to cholesterol efflux and homeostasis. Collectively, this work adds a new dimension to the immunometabolic sphere of perturbations that impair regression of established diabetic atherosclerosis and suggests that targeting RAGE and IRF7 may facilitate vascular repair in diabetes.
It was in these contexts that we hypothesized a role for the receptor for advanced glycation endproducts (RAGE) in impaired regression of atherosclerosis. RAGE, a member of the immunoglobulin superfamily, is expressed on monocytes/macrophages and vascular cells and has been implicated in atherosclerosis progression in mice (20)(21)(22)(23)(24). We employed a model of transplantation of donor atherosclerotic aortic arches (25) from mice devoid of the LDL receptor (Ldlr) and fed a Western diet into normolipidemic mice to test the premise that Ager (Ager, the gene encoding RAGE) contributes to macrophage dysfunction and impairs regression of atherosclerosis. We demonstrate that compared with WT recipient mice, deletion of Ager enhanced regression of atherosclerosis. RNA-sequencing of donor versus recipient diabetic macrophages after aortic arch transplantation suggested that Ager deletion downregulated interferon signaling, and, specifically, interferon regulatory factor 7 (IRF7). In macrophages, IRF7 regulated cholesterol metabolism and inflammation. Targeting RAGE and IRF7 might represent an important therapeutic opportunity to accelerate atherosclerosis regression.

RAGE is expressed in regressing atherosclerotic plaques in mice.
To test the premise that RAGE impairs regression of diabetic atherosclerosis, we rendered male Ldlr -/mice diabetic with streptozotocin at age 7 weeks. Hyperglycemic and control mice were placed on a Western diet (0.15% cholesterol for 16 weeks). At age 23 weeks, aortic arches from these donor mice were transplanted into male WT or homozygous Ager -/diabetic or nondiabetic mice (25). The grafted aortic arches were harvested at the indicated times after transplantation ( Figure 1A).
We verified that diabetes accelerated atherosclerosis progression in Western diet-fed Ldlr -/mice, as aortic arches from these mice served as the donors in these experiments. En face analysis of the aorta revealed a significant increase in atherosclerotic lesion area in diabetic versus nondiabetic aortas; P < 0.0001 ( Figure 1B). Immunofluorescence microscopy colocalized RAGE to CD68-expressing macrophages in the atherosclerotic plaques of these mice ( Figure 1C). When atherosclerotic aortic arches from diabetic Ldlr -/mice were transplanted into diabetic versus nondiabetic recipient WT mice, laser capture microdissection of lesional CD68-expressing macrophages 5 days posttransplantation revealed trends toward higher Ager mRNA transcripts by about 3-fold (Figure 1, D and E). We tested whether RAGE expression in the recipient mice contributed to impaired regression of atherosclerosis.
RAGE impairs regression of atherosclerosis. We first tested the role of RAGE in diabetes. Atherosclerotic aortic arches retrieved from diabetic donor Ldlr -/mice fed a Western diet were transplanted into diabetic WT or Ager -/mice. Five days later, H&E staining revealed that when the atherosclerotic aortic arches were transplanted into Ager -/diabetic mice, significantly lower lesion area was observed compared with that in the diabetic donor and diabetic WT recipient mice; P < 0.0001, and P < 0.001, respectively ( Figure 2A). There were no significant differences in plaque size when comparing donor aortic arches from diabetic Ldlr -/mice to the transplanted arches from diabetic WT recipient mice ( Figure 2A). With respect to macrophage content, compared with either the diabetic donor aortic arches retrieved from Ldlr -/mice or the transplanted aortic arches retrieved from WT diabetic recipient mice, the aortic arches retrieved from Ager -/mice revealed significantly reduced macrophage (CD68) content; P < 0.0001, and P < 0.01, respectively ( Figure 2B). There were no significant differences between the diabetic donor Ldlr -/aortic arches and those retrieved from the WT diabetic recipient mice ( Figure 2B). Oil red O staining (neutral lipid) revealed that compared with the donor diabetic Ldlr -/mice aortic arches, lipid content was significantly reduced in the diabetic recipient mice devoid of Ager but not the WT animals; P < 0.0001 ( Figure 2C). Lesional content of collagen, considered a beneficial stabilizing component in human atherosclerotic plaques, was assessed. Whereas there were no significant differences between the diabetic donor Ldlr -/mice aortic arches and those retrieved from the diabetic WT recipient mice, significantly higher Picrosirius red staining was observed in the diabetic Ager -/recipient mice; P < 0.0001 ( Figure 2D). Significantly higher collagen content was observed in the aortic arches of the diabetic recipient mice devoid of Ager versus the WT diabetic recipient mice; P < 0.0001 ( Figure 2D).
We analyzed metabolic data (Table 1). Ager -/mice recipients weighed significantly less than the donor Ldlr -/and the WT recipient mice. There were no significant differences in levels of blood glucose. Levels of total cholesterol were significantly lower in WT and Ager -/recipient groups versus the donor Ldlr -/mice; plasma cholesterol was significantly lower in the Ager -/versus the WT mice.
However, only in the donor Ldlr -/mice, not in the recipient mice (WT and Ager -/-), were significant correlations between atherosclerotic lesion area and plasma cholesterol levels observed ( Table 2). The levels of triglyceride were significantly higher in the Ldlr -/mice versus both recipient groups; no significant differences were observed between the recipient groups.
As RAGE ligands accumulate in nondiabetic atherosclerosis (20), we transplanted aortic arches from nondiabetic, Western diet-fed Ldlr -/mice into nondiabetic WT or Ager -/recipient mice and sacrificed the mice 5 days later. Compared with the nondiabetic donor Ldlr -/mice aortic arches and the aortic arches that were transplanted into the recipient WT nondiabetic mice, transplantation of atherosclerotic aortic arches into nondiabetic mice devoid of Ager resulted in significantly lower atherosclerotic lesion area (P < 0.0001, and P < 0.01, respectively); lesional macrophage content (CD68) (P < 0.0001); and Oil red O staining (P < 0.0001 and P < 0.001, respectively) and significantly higher collagen content; P < 0.0001) (Supplemental Figure 1, A-D; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.137289DS1). Ager -/recipient mice weighed less than the donor Ldlr -/mice. Levels of blood glucose were significantly lower in the Ager -/recipient mice versus the other 2 groups. Levels of total cholesterol were significantly lower in WT and Ager -/recipient mice versus the donor Ldlr -/mice but did not differ between the 2 recipient groups. There was no significant correlation between atherosclerotic lesion area and cholesterol levels in the recipient Ager -/mice, unlike that observed in the WT recipient mice (Supplemental Table 1). Levels of total triglycerides were significantly lower in both recipient groups versus the donor mice (Supplemental Table 1).
To test if the effects of Ager deletion on regression of diabetic atherosclerosis were sustained, recipient diabetic mice that had undergone transplantation of donor aortic arches from diabetic, Western diet-fed Ldlr -/mice were sacrificed 14 days posttransplantation. Atherosclerotic lesion area, macrophage CD68 content, and Oil red O staining were significantly lower and lesional content of collagen was significantly higher in the diabetic WT and Ager -/recipient mice compared with the diabetic donor Ldlr -/mice; P < 0.0001 (Supplemental Figure 2, A-D). All these parameters were significantly improved in the diabetic Ager -/versus the WT recipient mice; P < 0.05 (Supplemental Figure 2, A-D).
Effects of advanced glycation end-products, oxidative stress, Tg glyoxalase-1 expression, and deletion of Diaph1 on regression of diabetic atherosclerosis. We reasoned that the accumulation of RAGE ligands contributed to the observed differences between Ager-expressing and -deficient mice and probed for RAGE ligands, advanced glycation end-products (AGEs), in plasma and atherosclerotic lesions. AGEs may be generated by hyperglycemia, oxidative stress, and inflammation (20-24, 26, 27). Western blotting revealed that plasma AGE epitopes were significantly higher in diabetic versus nondiabetic WT mice; P < 0.001 ( Figure 3A). Aortic arches and atherosclerotic plaques of diabetic Ldlr -/mice revealed copious AGE immunoreactivity ( Figure 3B).
To determine if AGEs impaired regression of diabetic atherosclerosis, we modulated levels of the enzyme GLO1, which detoxifies pre-AGE intermediates and reduces AGE content (28). Tg Glo1 mice, which express murine Glo1, or WT controls were rendered diabetic (29). Five days posttransplantation, significantly lower mean lesion area was observed in the diabetic Tg Glo1 mice versus either the diabetic donor Ldlr -/mice or the diabetic WT recipient mice; P < 0.0001 (Figure 2A). CD68 staining revealed significant reduction in macrophage content in the Tg Glo1 diabetic mice versus the donor Ldlr -/mice or the WT recipient animals; P < 0.0001 ( Figure 2B). Neutral lipids were significantly lower in the aortas of Tg Glo1 In B, the secondary antibody-alone control is shown. Mean ± SEM; 1-way ANOVA with post hoc Tukey's test was used. **P < 0.01, ***P < 0.001, and ****P < 0.0001.
versus the diabetic donor Ldlr -/mice or the diabetic WT recipient mice; P < 0.0001 ( Figure 2C). Content of collagen was significantly higher in the diabetic Tg Glo1 recipient mice versus the diabetic donor Ldlr -/mice or the diabetic WT recipient mice; P < 0.0001 ( Figure 2D). The AGE content in the transplanted atherosclerotic plaques in the diabetic Tg Glo1 recipient mice was significantly lower than that in the aortic arches of the diabetic donor Ldlr -/or the diabetic WT recipient mice; P < 0.0001 ( Figure 3B). Per Table 1, compared with donor Ldlr -/mice, Tg Glo1 mice weighed significantly less. No differences in levels of blood glucose were observed. Levels of total cholesterol were significantly lower in the Tg Glo1 recipient mice versus the donor Ldlr -/mice and the WT recipient mice, and the total cholesterol levels in the Tg Glo1 mice correlated with atherosclerotic lesion area (Table 1 and Table 2). The levels of triglycerides were lower in the recipient Tg Glo1 and WT mice versus the donor Ldlr -/mice (Table 1). Collectively, these data indicate that the reduction of RAGE ligand AGEs accelerated regression in diabetic atherosclerotic plaques.
To identify if RAGE signaling impaired regression of diabetic atherosclerosis, we postulated a role for the cytoplasmic domain binding partner of RAGE, DIAPH1 (30). Deletion of macrophage Diaph1 reduces the upregulation of Egr1 mRNA and activity in hypoxia (31). We rendered mice devoid of Diaph1 and WT mice diabetic and probed regression of atherosclerosis. Five days after transplantation, compared with donor Ldlr -/mice and WT diabetic recipient mice, the recipient diabetic mice devoid of Diaph1 displayed significantly lower atherosclerotic lesion area, P < 0.0001; significantly lower CD68 macrophage content, P < 0.0001; significantly lower Oil red O staining, P < 0.01; and significantly higher collagen content; P < 0.0001 ( Figure 2, A-D). Compared with the diabetic donor Ldlr -/mice or diabetic WT recipient mice, the aortas of diabetic Diaph1 -/mice displayed significantly lower AGE content; P < 0.0001 ( Figure 3B). Dihydroethidium (DHE) staining, a marker for oxidative stress, was significantly lower in the Ager -/and Diaph1 -/mice compared with donor Ldlr -/mice; P < 0.0001 ( Figure 3C). Compared with the diabetic recipient WT mice, the regressing plaques of Ager -/or Diaph1 -/recipient diabetic mice displayed significantly lower DHE staining; P < 0.05 and P < 0.001, respectively. Per Table 1, the body weight of the Diaph1 -/mice was significantly lower than that of the donor Ldlr -/mice but was not different from the WT recipient mice. Levels of blood glucose did not differ. Levels of cholesterol in the recipient Diaph1 -/mice were significantly lower than those in the donor Ldlr -/mice or the WT recipient mice, but there was no correlation between atherosclerotic lesion area and cholesterol levels in Diaph1 -/mice. Levels of triglyceride were lower than those in the donor mice but Values represent mean ± SEM. Donor group received Western diet; all other groups received standard chow diet. A Donor versus recipient, P < 0.05. B WT recipient versus Ager -/recipient, P < 0.05. C WT recipient versus Tg Glo1 recipient, P < 0.05. D WT recipient versus Diaph1 -/recipient, P < 0.05.
did not differ from the levels in WT recipient mice (Table 1 and Table 2). Thus, the RAGE/DIAPH1 signaling axis contributed to impaired regression of diabetic atherosclerosis.

Effects of RAGE on macrophage properties in the regression of diabetic atherosclerosis.
To probe the mechanisms by which RAGE affected macrophage properties, we assessed macrophage proliferation and apoptosis in the regressing diabetic plaques. Compared with diabetic WT recipient mice, Ki67 + (proliferation) immunoreactivity in transplanted aortic arches was significantly lower in the CD68 + macrophages in the diabetic Ager -/recipient mice; P < 0.01 ( Figure 4A). TUNEL staining revealed a significant reduction in TUNEL + /CD68 + -expressing macrophages in the aortic arch lesions of the diabetic Ager -/recipient versus WT mice; P < 0.0001 ( Figure 4B).
To probe in vivo macrophage emigration, ( Figure 4C), diabetic Ldlr -/mice were fed a Western diet for 16 weeks. Eight days before retrieval of aortic arches for transplantation, fluorescently labeled, nondegradable beads were injected into the donor diabetic Ldlr -/mice via tail vein to label the circulating monocytes (32). The aortic arches of these donor mice were then transplanted into diabetic WT, Ager -/-, or Diaph1 -/mice, and the transplanted arches were harvested 5 days posttransplantation. There were no significant differences in fluorescently labeled bead content in the transplanted aortic arch lesions between the diabetic donor Ldlr -/mice and the diabetic WT recipient mice. The aortic arch lesional bead content was significantly lower in the diabetic Ager -/recipient mice and the diabetic Diaph1 -/mice compared with the diabetic donor Ldlr -/mice (P < 0.001, and P < 0.0001, respectively), when compared with the donor pretransplantation state ( Figure  4D). Thus, deletion of Ager or Diaph1 facilitated macrophage emigration from the regressing diabetic plaques.
in primary WT murine aortic endothelial cells (MAECs) (Supplemental Figure 3A) (33). WT MAECs were incubated with BMDMs from diabetic WT, Ager -/-, or Diaph1 -/mice. At 2, 24, and 48 hours of incubation, BMDMs devoid of Ager or of Diaph1 displayed significantly higher reverse transmigration compared with WT BMDMs in the coculture with WT MAECs; P < 0.05 ( Figure 4E). Thus, deletion of Ager or Diaph1 in BMDMs enhanced reverse endothelial transmigration and supported the conclusions of in vivo bead labeling macrophage studies ( Figure 4D).
Effect of RAGE on donor versus recipient-derived plaque macrophage content posttransplantation. We distinguished the source of the transplanted aortic arch atherosclerotic lesional macrophages, donor versus recipient, and probed their transcriptomes (Supplemental Figure 4A). Donor Ldlr -/mice (CD45.2) were rendered diabetic and fed a Western diet, and the aortic arches were transplanted into diabetic WT or Ager -/recipient mice (CD45.1) and were retrieved 3 days posttransplantation.
We determined the relative contribution of donor versus recipient macrophages in the transplanted aortic arches with respect to recipient genotype. By immunohistochemistry, 3 days after transplantation, in the WT diabetic recipient mice, the mean recipient CD45.1 macrophage lesional content was 73.1% ± 4.8%, and the mean donor CD45.2 macrophage lesional content was 26.9% ± 4.8%. In the diabetic Ager -/recipient mice, the mean CD45.1 macrophage lesional content was 87.7% ± 2.8%, and the mean CD45.2 macrophage lesional content was 12.3% ± 1.8%. These data indicated a significantly higher content of recipient CD45.1/CD68 + macrophages and a small but significantly lower content of donor CD45.2/CD68 + macrophages in the diabetic Ager -/versus WT diabetic recipient lesions; P < 0.01 (Figure 5A). By flow cytometry, although there were no significant differences in the percent recipient CD45.1 versus donor CD45.2 of the CD11B + /F4/80 + macrophage lesional content by recipient genotype, there were trends to higher recipient CD45.1/CD68 + macrophage content in the aortic arch lesions transplanted into the Ager -/versus the WT diabetic recipient mice ( Figure 5B). These findings, consistent with previous reports, indicated that the majority of the aortic arch lesional macrophages in regression are accounted for by newly recruited cells from the recipient mice (18).
RNA-sequencing implicates RAGE-dependent interferon signaling and IRF7 in impaired regression of diabetic atherosclerosis. We thus reasoned that the deletion of Ager affected key properties of macrophages that facilitated lesion regression. We performed a transcriptomic analysis of donor versus recipient lesional macrophages sorted from the aortas of recipient mice 3 days after transplantation. Our sorting strategy (CD11B + , F4/80 + CD45.1 or CD45.2) allowed us to examine the transcriptome of donor-versus recipient-derived aortic arch lesional macrophages (Supplemental Figure 4, A-C). Table 3 and Table 4 detail the experimental conditions and the nomenclature used in our RNA-sequencing studies and summarize the 4 major comparisons performed on the RNAsequencing data in which we consider the transcriptomes of the donor versus the recipient-derived macrophages.
We used Correlation Adjusted Mean Rank Gene Set Test (CAMERA) (35) to compare the number of differentially expressed Reactome gene sets in each comparison (Supplemental Table 2) using FDR 0.05. Table 5 displays the effect of Ager deletion versus the WT recipient genotype on the recipient CD45.1 + macrophages populating the transplanted aortic arch lesions (comparison 4, Supplemental Table 2). Four gene sets were identified, 3 of which related to the downregulation of the interferon signaling pathway (α, β, and γ), and the fourth gene set reflected an increase in glycolysis when comparing Ager -/versus the WT.
Subsequently, we analyzed differential gene expression. Supplemental Table 6 indicates the number of differentially expressed genes for each of the 4 comparisons using 2 criteria: first, a stringent cutoff with FDR ≤ 0.05, and second, a lenient cutoff with P ≤ 0.001 and FDR ≤ 0.2. No genes were differentially expressed in the donor macrophages (Supplemental Table 6, comparison 3); however, in the recipient macrophages 4 genes were differentially expressed by genotype using the stringent criterion, but 11 genes were differentially expressed by recipient genotype by the lenient cutoff ( Figure 5C and Supplemental Table 9). Consistent with the downregulation of the interferon signaling pathways, Irf7 was downregulated in the Ager -/versus WT diabetic recipient CD45.1 macrophages populating the lesions after transplantation of the atherosclerotic aortic arches.
We focused our attention to the interferon signaling pathway and to Irf7 (Figure 5C), which has not been previously linked to RAGE signaling to our knowledge.
IRF7 is expressed in mouse and human atherosclerosis, is regulated by RAGE ligands, and affects macrophage inflammation and cholesterol metabolism. IRF7 colocalized with CD68 in the WT and Ager -/recipient mice lesions, but the percentage of IRF7 and CD68 + (i.e., double positive) macrophages per lesion area was significantly lower in the Ager -/versus WT diabetic lesions; P < 0.001 ( Figure 6A). IRF7 colocalized with CD68 in human diabetic coronary artery atherosclerotic lesions ( Figure 6B, Supplemental Figure 5). Although IRF7 colocalizes with CD68 in nondiabetic atherosclerotic plaques (Supplemental Figure 6), this was significantly lower compared with that in the diabetic subjects; P < 0.05 ( Figure 6B).
We tested the relationship between RAGE and IRF7 in vitro. BMDMs isolated from Ager -/versus WT mice and exposed to serum (2%) from WT C57BL/6 mice displayed trends to lower Irf7 mRNA transcripts ( Figure 6C). Ager siRNA knockdown in WT murine BMDMs resulted in significant reduction in Irf7 mRNA transcripts compared with BMDMs exposed to scrambled siRNA; P < 0.05 ( Figure 6D).
To mimic the conditions relevant in atherosclerosis, we retrieved serum from Western diet-fed Ldlr -/mice (36), which is enriched in multiple damage-associated molecular patterns, many of which are RAGE ligands (37,38). BMDMs retrieved from Ager -/mice grown in this serum displayed significantly lower levels of Irf7 mRNA versus WT BMDMs; P < 0.05 ( Figure 6E). Upon Ager siRNA knockdown in BMDMs, modest reductions in Irf7 mRNA transcripts were observed compared with scrambled siRNA ( Figure 6F).
We assessed the effects of a prototypic AGE ligand of RAGE, CML-AGE (39) on the regulation of Irf7 using CML-AGE prepared from human serum albumin (39). WT BMDMs treated with CML-AGE (100 μg/mL) versus vehicle demonstrated a 5-fold increase in Irf7 mRNA; P < 0.001 ( Figure 6G). However, in BMDMs from Ager -/mice exposed to CML-AGE, there was no significant increase in Irf7 versus vehicle ( Figure 6G). Compared with WT BMDMs exposed to CML-AGE, the increase in Irf7 mRNA transcripts was significantly higher than that observed in CML-AGE-treated Ager -/-BMDMs; P < 0.001 ( Figure 6G). Thus, RAGE ligands directly increase Irf7 mRNA.
We tested if Ager knockdown would prevent CML-AGE-stimulated upregulation of Irf7 mRNA. In the presence of scrambled siRNA and CML-AGE, a 3-fold increase in Irf7 mRNA was observed compared with vehicle; P < 0.0001 ( Figure 6H). When BMDMs were treated with CML-AGE in the presence of Ager siRNA, there was no significant increase in Irf7 versus vehicle ( Figure 6H). Irf7 mRNA transcripts were significantly higher in CML-AGE-treated BMDMs in the presence of scrambled siRNA versus Ager siRNAs; P < 0.0001 ( Figure 6H).
Finally, we tested the levels of IFN-γ in the plasma of diabetic Ldlr -/mice donor mice and in the diabetic recipient WT or Ager -/mice. Compared with recipient WT mice, recipient mice devoid of Ager displayed a significant reduction in IFN-γ; P < 0.001 ( Figure 7D).

Discussion
Collectively, this work implicates RAGE in impaired regression of atherosclerosis in both diabetic and nondiabetic mice. Although previous studies underscored roles for RAGE in atherosclerosis progression (20)(21)(22)(23)(24), here, we demonstrate the maladaptive roles of AGEs, RAGE, and DIAPH1 in impairment of diabetic atherosclerosis regression.
Our studies suggested that RAGE exerted multiple effects on macrophage properties in the regression environment. First, macrophage proliferation, identified as an important contributor to atherosclerosis (40), was significantly reduced in the plaques of Ager -/versus WT diabetic recipient mice ( Figure 4A).
Second, TUNEL + macrophages were significantly lower in the absence of Ager, thereby suggesting that enhanced macrophage survival was associated with more functionally competent macrophages ( Figure  4B). Third, macrophage emigration from the regressing plaques was increased in an Ager -/environment. Furthermore, gene signatures in the reverse-transmigrated macrophages in an in vitro model suggested that Ager deletion reduced proinflammatory markers and upregulated antiinflammatory markers (18).
RNA-sequencing studies revealed that the recipient, not donor-derived, macrophages evoked the major RAGE-dependent pathways in regression ( Figure 4C), and RAGE-dependent modulation of Irf7 was pinpointed as a leading candidate gene. In mammals, the 9 IRF family members are transcription factors that regulate type 1 interferon antiviral responses and immune cell differentiation (41,42). The functions of these molecules are more complex, as IRFs are expressed in myeloid and lymphoid cells and in nonimmune cells, such as vascular cells (43)(44)(45)(46) and adipocytes (47,48).
Although IRF7 has not been tested in atherosclerosis, other members of the IRF family have been studied in murine atherosclerosis progression. Global deletion of Irf1 reduced foam cell formation and atherosclerosis in Apoe -/mice (49). Global deletion of Irf5 reduced lesion and necrotic core areas in mice devoid of Apoe (50). Dendritic cell deletion of Irf8 significantly reduced atherosclerosis in Ldlr -/mice, with reduced T cell priming, activation, and differentiation to T helper cells, with consequent marked reduction in B cell activation (51). In Apoe -/mice, deletion of Irf3 resulted in significant reduction in atherosclerosis; the major effects of IRF3 on atherosclerosis were traced to endothelial cells (52). Neutralization of B cell-activating factor (BAFF) in mice with deficiency of either Apoe or Ldlr increased atherosclerosis, despite depletion of mature B-2 cells. A function of BAFF receptor signaling was the repression of macrophage IRF7-dependent production of proatherogenic CXCL10, which was blocked in the presence of the antagonist (53).
Smooth muscle cell IRF7 (54) plays protective roles through reduction of proliferation consequent to arterial injury (55). Our data indicate that Irf7 knockdown in BMDMs attenuated proinflammatory effects, as shown by the significant reduction of Tnfa, Nos2, Il6, and Ccl2 and upregulated antiinflammatory markers, Il10 and Arg1. In addition, the knockdown of Irf7 in BMDMs exposed to hyperlipidemic conditions resulted in upregulation of cholesterol transporters and downregulation of Cd36, factors favoring efflux of cholesterol. Upregulation of Nr1h3 and Nr1h2 suggests that knockdown of Irf7 in   BMDMs imparts an antiatherogenic profile and downregulation of BMDM Irf7 reduced total cellular cholesterol in hyperlipidemic serum. Although global deletion of Irf7 reduced high fat diet-induced obesity, the proximate mechanisms were not elucidated (56). The present body of work adds RAGE-dependent regulation of IRF7 to this network of molecules bridging immune and cholesterol metabolism functions in macrophages, which may contribute to impaired regression of atherosclerosis. Miao and colleagues exposed human THP1 monocytes to high glucose and showed that this upregulated IRF7 (57), which was analogous to an IRF7-driven inflammatory network that was linked to the risk of type 1 diabetes (58). High glucose upregulates ligands of RAGE. Mediating roles for AGEs were demonstrated in our studies by improved regression of diabetic atherosclerosis in transgenic mice overexpressing Glo1, the enzyme that detoxifies pre-AGE species. It is plausible that non-AGE ligands of RAGE contribute to upregulation of Irf7. Hyperlipidemic serum from Ldlr -/mice contains TLR ligands, which also interact with RAGE. In vitro, CpG oligodeoxynucleotides and Complement C3a increased production of IFN-α in WT but not Ager -/-BMDMs. In human peripheral blood mononuclear cells, treatment with C3a and CpG oligodeoxynucleotides increased IFN-α, which was reduced by RAGE antagonism (59).
Finally, the regulation and biology of IRF7 bear similarities to those of RAGE. Akin to findings that RAGE ligands upregulate AGER, at least in part through NF-κB, and that RAGE activation perpetuates proinflammatory RAGE ligand generation (60), IRF7-mediated upregulation of IFNs stimulates IFN-mediated upregulation of Irf7 (61,62). These considerations place RAGE and IRF7 at critical nodes in the mediation of chronic inflammation. Here, we show that RAGE impairs regression of atherosclerosis through perpetuation of inflammation and through impairment of cholesterol metabolism, at least in part through IRF7. Interruption of RAGE/IRF7 may restore reparative mechanisms and accelerate regression of atherosclerosis, particularly in diabetes.

Methods
Animal studies and induction of diabetes. Male donor mice with atherosclerosis were Ldlr -/-. The recipient male mice, Ager -/mice (C57BL/6J) (63), Diaph1 -/mice (C57BL/6J) (30), and Tg mice expressing murine Glo1 (C57BL/6J) (29), were bred in-house, and WT C57BL/6J mice were purchased from The Jackson Laboratory. Mice were housed in a specific pathogen-free facility. Certain mice were rendered type 1 diabetic by intraperitoneal injections of streptozotocin (STZ; MilliporeSigma, S0130) at 7 weeks of age. STZ was freshly dissolved in citrate buffer (0.1 mol/L, pH 4.5) and administered at a dose of 55 mg/kg/d, once daily for 5 days. Blood glucose (>250 mg/dL) was measured (Free Style Freedom Lite glucometer) to ensure diabetes was induced. Nondiabetic animals were treated with equal volumes of citrate buffer. Donor Ldlr -/mice were fed a Western diet (Research Diets, Inc., D01061401Ci; 0.15% cholesterol) for 16 weeks. Recipient mice fed a chow diet were diabetic or control for 3 months before surgery. To test mechanisms of atherosclerosis regression, aortic arch transplantation was performed at 23 weeks of age (8,9,18,25,64,65). Buprenorphine was administered by subcutaneous injections every 12 hours for 3 days postsurgery. At sacrifice, whole blood was collected from the aorta. For serum isolation, whole blood was allowed to clot in BD Microtainer SST (365967) and isolated by centrifugation. For plasma isolation, whole blood was collected with EDTA (Thermo Fisher Scientific, BP2482) and isolated by centrifugation.
En face aorta staining. The full descending aorta was dissected, excised, and pinned (Thermo Fisher Scientific, NC9681411); fixed in 4% formaldehyde for 5 minutes; rinsed in 70% ethanol for 5 minutes; and stained in a working solution of 2.5 g of Sudan IV (Merck Eurolab), 250 mL of 70% ethanol, and 250 mL of acetone filtered through Whatman paper before use. Sudan IV solution was placed on top of the aorta for 10 minutes. This was rinsed twice in 80% ethanol for 3 minutes.
in Ager -/and WT BMDMs exposed to a hyperlipidemic environment for 48 hours (2% serum of Ldlr -/mice fed a Western diet). N = 4 mice/group. (F) Irf7 gene expression levels in WT BMDMs after 48 hours of Ager or scrambled (75 nM) in a hyperlipidemic environment. Knockdown of Ager by siRNA is shown. N = 6 mice/group. (G) Irf7 gene expression in WT and Ager -/-BMDMs grown in the presence of 1% serum of Ldlr -/mice fed Western diet for 48 hours, followed by an additional 16 hours in CML-AGE (100 μg/mL) versus vehicle. N = 4 mice/group. (H) Irf7 gene expression in WT BMDMs after 48 hours of Ager or scrambled (75 nM) in a normolipidemic environment in cells treated with CML-AGE (200 μg/mL) or vehicle for 16 hours. Knockdown of Ager is shown. N = 6 mice/group. Secondary antibody alone controls are shown in A and B. Mean ± SEM. Unpaired t test or Mann-Whitney U test was performed to assess the differences in A-F and H (for Ager endpoint) depending on the normality of data. One-way ANOVA with post hoc Holm-Šídák multiple comparisons test was used in G and H (for Irf7 endpoints). *P < 0.05, ***P < 0.001, and ****P < 0.0001.
Histological stains. The pretransplant donor and grafted aortic arches were removed after perfusion of cold PBS, embedded in optimal temperature cutting compound (OCT), and frozen. Murine aortic frozen sections (6 μm thick) were air-dried before the following procedures.
In situ apoptosis detection (TUNEL assay). The in situ detection of apoptotic plaque macrophages was performed using the Click-it Plus TUNEL Assay (Invitrogen, Thermo Fisher Scientific, C10619). Atherosclerotic plaque . Mean ± SEM. Unpaired t test or Mann-Whitney U test was performed to assess the difference in A-C depending on the normality of data. One-way ANOVA with post hoc Holm-Šídák multiple comparisons test was used in D. *P < 0.05, **P < 0.01, and ***P < 0.001. used to determine CD45.1 + and CD45.2 + / CD68 + , IRF7 + / CD68 + areas as well as to identify TUNEL + and Ki67 + /CD68 + in the plaques. All staining and immunohistochemistry and immunofluorescence analyses were performed by an observer blinded to the conditions. Blood and plasma parameters. Blood was harvested after a 6-hour fast. Total cholesterol (Infinity, Thermo Fisher Scientific, 948541) and triglyceride (Infinity, Thermo Fisher Scientific, TR22421) concentrations were measured.
Monocyte bead tracking. Ly6C lo monocytes were labeled in vivo by tail vein injection of 250 μL beads diluted 1:4 in PBS (Fluoresbrite YG microspheres 1 μm, 17154-10, Polysciences Inc.) into Ldlr -/donor mice before transplantation. (11,68,69). Quantification of the number of beads/cross section of the aortic arch at baseline and 5 days after arch transplantation into the diabetic recipients was performed.
Bone marrow-derived macrophages. Primary BMDMs were isolated from 12-week-old mice (70). Cut-end femora were spun for 6000 g for 1 minute for collection of bone marrow. After addition of medium, single-cell suspensions were plated in bone marrow-conditioned media (71) for 5 days and subjected to experimental conditions.
Laser capture microdissection of CD68-expressing cells. The aortic arch was cut in half, positioned vertically in OCT, and stored at -80°C for 24 hours under RNase-free conditions. Frozen sections (10 μm) were collected on Arcturus PEN membrane glass slides (Life Technologies, Thermo Fisher Scientific) and fixed with 70% ethanol and partially dehydrated in 70% ethanol followed by washing in 95% ethanol. Eosin Y (VWR Scientific) was used for staining, washed in 95% ethanol, and completely dehydrated in 100% ethanol and xylene (11). After air-drying, foam cells could be identified by light microscopy and were verified by CD68 staining of guiding slides. Atherosclerotic lesion CD68 + cells were laser-captured using Leica LMD6500 laser microdissection system.
Quantitative reverse transcription PCR experiments. BMDM total RNA was extracted using the RNeasy Plus Mini Kit (QIAGEN, 74136). RNA from laser capture-microdissected CD68 + cells was isolated by the Arcturus PicoPure RNA Isolation Kit (Life Technologies, Thermo Fisher Scientific, 12204-01). cDNA was synthesized using iScript cDNA Synthesis Kit (Bio-Rad, 1708891) and amplified using TaqMan assays (Supplemental Table 10) using a 7300 Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific).
Reverse transendothelial migration assay. Collagen gels were polymerized with collagen type I (BD 354231) before WT primary MAECs were isolated and seeded (72). Primary BMDMs (4 × 10 6 cells/mL) were added to confluent endothelial cultures for 2 hours to allow accumulation of macrophages in the subendothelial collagen. Cultures were rinsed to remove nonmigrated cells and incubation was continued. At intervals, cultures were rinsed to remove nonadherent macrophages that had accumulated in the apical compartment by reverse transmigration, and medium was replenished. Macrophages adherent to the apical surface of the endothelium were washed before evaluation of total DNA/well with Yo-Pro-1 (Molecular Probes, Thermo Fisher Scientific, Y3603) At each point, reverse transmigration was calculated as the percentage decrease in the number of macrophages beneath the endothelial monolayer compared with the number that initially accumulated below the endothelium at 2 hours baseline (33). At 48 hours, macrophages were eluted from the coculture with StemPro Accutase Cell Dissociation Reagent (Thermo Fisher Scientific, A1110501) (34). To digest collagen gels containing endothelial cell monolayer and non-reverse-transmigrated cells, 100 μL/well of collagenase D (2 mg/mL in DMEM) was added for 40 minutes at 37°C. The cell suspension was filtered through a cell strainer and subjected to reverse transcription PCR.
ELISA. Plasma samples were subjected to ELISA for detection of IFN-γ using the Mouse Interferon Gamma sandwich ELISA kit (LifeSpan Biosciences LS-F23028-1).
RNA-sequencing and bioinformatics. cDNA samples were prepared using the NuGEN Ovation RNA-Seq system v2 reagents and the Ovation Ultralow DR Multiplex system for the adapter ligation step. Libraries were amplified by 9 cycles. Single-end, 50-nucleotide reads were sequenced on an Illumina 2500 HiSeq using v4 chemistry. FASTQ files were trimmed of low-quality bases and adapters with Trimmomatic (75). RNA-seq reads were aligned to the mouse genome (mm10) using Bowtie2 version 2.1.0 (76) and TopHat2 version 2.0.9 (77). Reads per gene were counted with HTSeq (78). Data were deposited in the Gene Expression Omnibus database (79) (accession GSE142204). Differential expression was analyzed with weighted Limma-Voom (80). Genes were considered expressed if they had at least 1 count per million in at least 2 samples. Genes were considered significantly differentially expressed if the Benjamini-Hochberg (81) FDR ≤ 0.05. An additional, more lenient cutoff of P ≤ 0.001 and FDR ≤ 0.2 was used. All expressed genes were analyzed by Reactome pathway (82) gene lists using CAMERA (35). Hierarchical clustering (83,84) was performed with Cluster 3.0 (85) using distances based on a modified Pearson's correlation coefficient and average linkage clustering (83,84). Genes were mean centered and both genes and samples were clustered. The dendrogram and heatmap were plotted with Java TreeView (86).
Statistics. Analyses were performed using GraphPad Prism 8.2.0. Data are presented as mean ± SEM. Normality of the data was assessed using the Shapiro-Wilk normality test. A nonparametric test was performed when data did not follow a normal distribution. Independent 2-sample t tests (2 sided) were used to assess the difference between 2 groups of samples (Mann-Whitney U tests were used instead if normality was violated). For over 2 groups, 1-way ANOVA was used, and Tukey's or Holm-Šídák post hoc test for pairwise comparisons or comparisons of selected groups was performed, respectively. Kruskal-Wallis test with post hoc Dunn's test was performed instead if the normality test was not passed. Pearson's correlation coefficient was assessed to evaluate the associations between 2 variables. P < 0.05 denoted statistical significance.
Study approval. Animal experiments were performed according to the NIH Guide for the Care of Laboratory Animals (National Academies Press, 2011), and protocols were approved by the Institutional Animal Care and Use Committee at NYU Grossman School of Medicine. For human samples, the CVPath Institute was referred autopsy hearts from cases of unexpected sudden death from the Office of the Chief Medical Examiner of the State of Maryland. This study was approved by the CVPath Institutional Review Board as an exempt study (RP0063).

Author contributions
LS, RLD, LEG, EAF, RR, and AMS designed the research and edited the manuscript. LS, RLD, LEG, JL, GD, QL, KR, YV, TB, and MZD performed experiments and edited the manuscript. LS, RLD, LEG, HL, JH, and RAF performed data analyses and edited the manuscript. LG, DF, AVF, and RV provided key materials and edited the manuscript. AMS directed the research and the research team, wrote the original draft of the manuscript, and completed all the edits and final preparation of the manuscript.