RAGE Deletion Confers Renoprotection by Reducing Responsiveness to Transforming Growth Factor-β and Increasing Resistance to Apoptosis

The receptor of advanced glycation end products (RAGE) is a cell surface multiligand receptor (1,2) and a member of the immunoglobulin superfamily; it plays a role in tissue repair early in disease (3) but also has been linked to chronic proinflammatory gene activation in the pathogenesis of several diseases including Alzheimer disease, rheumatoid arthritis, cancer, and diabetes complications (4). In the kidney, advanced glycation end products (AGEs) binding to RAGE results in increased expression of growth factors and cytokines such as transforming growth factor (TGF)-β, connective tissue growth factor, vascular endothelial growth factor, and MCP-1, ultimately leading to glomerular and tubulointerstitial injury (5).

Transgenic overexpression of RAGE worsens kidney disease in both nondiabetic and diabetic mice (6), whereas administration of soluble RAGE (7) or RAGE-neutralizing antibodies (8) confers protection against complications, similar to that observed in RAGE^−/− mice (7,9–12). Specifically targeting RAGE as a therapeutic treatment remains controversial because RAGE plays a key role in both innate and adaptive immune responses (4,13,14), lung homeostasis, bone metabolism, the immune system, and the nervous system (15). A better understanding of the role of RAGE in physiological processes—particularly in the kidney—is required because RAGE-directed therapeutic strategies are being considered for diabetic nephropathy (DN).

This study revealed an important homeostatic role for RAGE in the normal kidney and particularly in mesangial cells (MCs). RAGE deletion resulted in increased macrophage infiltration in the diabetic kidney, despite improved renal function, and decreased mesangial expansion. Gene profiling experiments demonstrated that RAGE^−/− MCs were less responsive to the actions of TGF-β, resulting in the downregulation of several signaling pathways, despite increases in the baseline expression of fibrotic and inflammatory markers. Restoring RAGE expression in RAGE^−/− MCs by adenoviral delivery returned the expression of fibrotic, inflammatory, and cell survival genes to control levels. These observations contribute to our understanding of the renoprotection conferred by RAGE deletion in response to diabetes.

MCs were isolated from the kidneys of C57BL6/J mice (16) and RAGE^−/− mice on a C57BL6/J background (17) through glomeruli collected after sieving (106-, 180-, and 75-µm pore size, respectively) and culturing MC outgrowths in DMEM (20% FCS), as previously described for rats (9,18). MCs were identified as positive for desmin, Thy-1.1, and α-smooth muscle actin (SMA) expression; as negative for von Willebrand factor; and by morphological evaluation (19). MCs were cultured up to 10 passages in DMEM (20% FCS). For experiments, cells were seeded at 8 × 10⁴ cells/well in six-well plates. Fresh DMEM (2% FCS) was added after 24 h, with or without TGF-β1 (R&D systems), in normal glucose (5.5 mmol/L) or high glucose (HG; 25 mmol/L), and cells were incubated another 3 days. Unless otherwise specified, in vitro studies were conducted in HG medium.

To study proliferation, MCs were seeded at 2.5 × 10⁴ cells/well and the medium was replaced with DMEM (5% FCS) after 24 h. Cells were counted with a Tali Image-Based Cytometer (Life Technologies). To determine viability, cells were seeded at 10 × 10⁴ cells/well, and the next day cells were starved in DMEM (no FCS) for 24 h. Cell death was assessed using the Tali Viability Kit (1:100; Life Technologies). To study apoptosis, cells were seeded at 10 × 10⁴ cells/well on collagen I–coated cover slips. The next day cells were treated with staurosporine (12.5 nmol/L; Sigma-Aldrich) for 24 h to induce apoptosis before immunostaining for active caspase-3 (1:500; Neuromics).

Primary antibodies were collagen I (1:5,000; Abcam), α-SMA (1:2,000; Dako), phospho-smad3 (1:1,000; Invitrogen), smad3 (1:5,000; OriGene), phospho–extracellular signal–regulated kinase (ERK) (1:500; Cell Signaling), ERK (1:1,000; Cell Signaling Technology), S100A4 (1:1,000; Cell Signaling Technology), P53 (1:500; Abcam) and β-actin (1:10,000; Abcam). Secondary antibodies were goat antimouse antibody (1:20,000; Dako) or goat antirabbit horseradish peroxidase–conjugated antibody (1:10,000; Dako).

Human full-length (FL) RAGE (FL-RAGE) and endogenous secretory RAGE (ES-RAGE) cDNA (gifts from Prof. H. Yamamoto [20]) were subcloned into AdEasy-1 (Ad-FL-RAGE and Ad-ES-RAGE) (9). pAdTrack-CMV without RAGE was used as an empty vector control. Virus was packed using HEK293 cells. MCs were seeded at 6 × 10⁴ cells/well and grown to 60% confluence in DMEM (10% FCS). The medium was then replaced with DMEM (2% FCS) containing Ad-FL-RAGE, Ad-ES-RAGE vector, or empty vector (2.85 × 10⁸ optical particle units). After 3 days cells were harvested to analyze gene expression. Transfection efficiency was 80–90% on the basis of green fluorescent protein fluorescence.

Cells were seeded at 6 × 10⁴ cells/well in 12-well plates in DMEM (10% FCS) without antibiotics. Medium was replaced with DMEM (2% FCS) and cells were transfected with small interfering RNA (siRNA)—5 nmol/L for RAGE or 10 nmol/L for S100A4—using RNAiMAX (Invitrogen); siRNA-negative controls were used at the same concentrations. Cells were harvested 24 h after transfection. RAGE siRNA was transfected on two consecutive days, and cells were harvested 48 h after the first transfection.

Total RNA (2 μg) extracted from the left kidney cortex or MCs was used to synthesize cDNA, as previously described (21). Gene expression was analyzed by real-time PCR using a TaqMan system (ABI Prism 7500; PerkinElmer) and normalized to 18S rRNA (18S rRNA TaqMan Control Reagent kit; PerkinElmer). Results were expressed relative to control (untreated) cells, which were arbitrarily assigned a value of 1, with four to six replicates per group.

RNA was quantified using a Qubit fluorometer and integrity was assayed using a MultiNA Bioanalyzer (Shimadzu, Kyoto, Japan). An NEBNext Poly(A) mRNA Magnetic Isolation Module was used to enrich mRNA from 1 μg of total RNA. Following elution of mRNA, we used the NEBNext Ultra Directional RNA Library Prep Kit (Illumina) to generate barcoded libraries. Libraries were validated with the MultiNA Bioanalyzer (Shimadzu), pooled to equimolar ratios, and sequenced at the Australian Genome Research Facility (Melbourne) using an Illumina HiSeq2500 system with version 4 single-end flow cell and reagents for 60 cycles.

Single-end mRNA sequencing reads underwent quality trimming to remove low-quality bases from the 3′ end of reads using the FASTX-Toolkit (version 0.0.14), applying a Phred quality threshold of 20 and minimum 20-nucleotide read length. STAR version 3.2.0.1 (22), with the default settings, was used to align reads to the mouse genome (Mus_musculus.GRCm38.dna.toplevel.fa; downloaded from Ensembl, www.ensembl.org/info/data/ftp/). We used Ensembl version 77 gene annotations (Mus_musculus.GRCm38.77.gtf). Exon-mapped reads were counted using featureCounts version 1.4.2 (23), with a mapping quality threshold of 10. Genes with fewer than 10 reads per sample on average were excluded from downstream analysis. Multidimensional scaling was analyzed using the cmdscale function in R software. Differential gene expression was statistically analyzed using edgeR software version 0.20 with the default settings (24). Factorial edgeR analysis was performed to identify genes and pathways whose differential regulation in response to TGF-β treatment is dependent on the presence of RAGE. To facilitate pathway analysis, mouse gene identifiers were mapped to human gene names using a mouse-human homolog relationship table downloaded from Ensembl BioMart (www.ensembl.org/biomart). Pathways were analyzed using Gene Set Enrichment Analysis (GSEA)-P software version 2.2.1.0 using the unweighted “classic” scoring scheme (25). Gene sets for pathway analysis were downloaded from MSigDB (software.broadinstitute.org/gsea/msigdb). Heatmaps were generated in base R.

Male mice were injected with streptozotocin (55 mg/kg i.p. daily for five consecutive days; diabetes group) (26) or sodium citrate buffer (control group). Plasma glucose levels (Accutrend; Boehringer Mannheim Biochemica, Mannheim, Germany) were determined after 10 days, and mice were overtly diabetic (plasma glucose >15 mmol/L; >95% of mice). The diet (AIN-93G; Specialty Feeds, Glen Forrest, Perth, Australia) and water were provided ad libitum. Mice (n = 8/group) were housed under specific pathogen-free conditions with exposure to a 12-h light/12-h dark cycle for 24 weeks. All animal studies were performed in accordance with the guidelines of the Alfred Medical Research & Education Precinct Animal Ethics Committee and the National Health and Medical Research Council of Australia. At the end of the study, kidneys and plasma were collected for analysis, and glycated hemoglobin was determined by high-performance liquid chromatography (HPLC) (27). Albumin excretion rate (AER) was determined by ELISA (Bethyl Laboratories, Montgomery, TX) (28), and creatinine in urine and plasma was determined by HPLC (Agilent HP1100 system; Hewlett Packard, Böblingen, Germany), as recommended previously (29). The glomerular sclerotic index (GSI) was assessed using a semiquantitative method and point-counting technique (30). Tubulointerstitial area (TIA) was assessed using a point-counting technique (31).

Sections (4 μm thick) were stained for collagen IV (1:600, rabbit polyclonal; Abcam). Briefly, sections were dewaxed, hydrated, and quenched with 3% H₂O₂ in Tris-buffered saline (TBS) and digested with 0.4% pepsin (Sigma-Aldrich) in 0.01 mol/L HCl at 37°C. Sections were blocked with 0.5% milk/TBS and incubated with the primary antibody at 4°C for 24 h. After avidin/biotin blocking, biotinylated antirabbit IgG (1:500) was applied as the secondary antibody at room temperature (RT) for 10 min.

For CD68 and CD169 immunostaining, frozen sections (5 μmol/L) were air-dried and fixed in 2% paraformaldehyde-lysine-periodate for 10 min. Sections were stained for CD68⁺ (1:500; BD Biosciences) leukocytes overnight at 4°C and for CD169 (1:200; Serotec) leukocytes at RT for 90 min. After washing, sections were incubated in 0.6% H₂O₂/TBS at RT for 20 min, followed by avidin/biotin blocking. Sections were incubated with biotinylated rabbit antirat IgG (1:200) at RT for 1 h. Finally, horseradish peroxidase–conjugated streptavidin (VECTASTAIN Elite ABC staining kit; Vector Laboratories) was added and sections incubated at RT for 30 min. Peroxidase conjugates were visualized using 3,3′-diaminobenzidine tetrahydrochloride (Sigma-Aldrich) in 0.08% H₂O₂/TBS.

Sections were finally counterstained with Mayer hematoxylin, dehydrated, and mounted. Sections were examined under a light microscope (Olympus BX-50; Olympus Optical, Tokyo, Japan) and digitized with a high-resolution camera. Collagen IV–positive staining in the cortex (20 cross sections per animal; magnification ×200) was quantitated using Image-Pro Plus 6.0 (Media Cybernetics, Bethesda, MD) (32). CD68⁺ and CD169⁺ leukocytes were counted in the glomerulus (15 hilar glomerular tuft cross sections per animal; magnification ×400). All assessments were performed in a blinded manner. Five or six kidneys were investigated in each group.

Renal cortex (50 mg) was homogenized (Polytron PT-MR2100) in extraction buffer to isolate cytosol and membranous fractions, as previously described (33). Total protein was determined by the BCA Protein Assay Kit (Pierce Chemical Company). MCP-1 levels in renal cortical cytosolic fractions were determined by sandwich ELISA (R&D Systems). Renal cytosolic fractions were concentrated using Microcon filters (Ultracel YM membrane, 10 kDa cut-off; Millipore Corporation). TGF-β1 ELISA was run on acid-treated cortical cytosolic fractions to measure total TGF-β1 protein levels (TGF-β1 Emax ImmunoAssay kit; Promega Corporation). Fibronectin was measured in renal cortical fractions by ELISA (Mouse Fibronectin PicoKine ELISA Kit; Boster Biological Technology).

Kidney sections were stained with periodic acid Schiff (PAS) to quantitate glomerulosclerosis. The degree of glomerulosclerosis, which was defined as thickening of the glomerular basement membrane and mesangial expansion, was evaluated with a semiquantitative method, as described previously (30, 32). Following Masson trichrome staining, tubulointerstitial collagen deposition was assessed in 20 fields (×40 magnification) using Image-Pro Plus 6.0 software (Media Cybernetics), as previously described (11). TIA was evaluated in PAS-stained sections by point-counting, as previously described (30). Mesangial area was analyzed (as a percentage of glomerular area) from digital images of glomeruli (15–20 glomeruli per kidney per animal) using Image-Pro Plus, as described previously (11).

Statistical analyses were performed using GraphPad Prism 6.0 (GraphPad Software, San Diego, CA). Values for experimental groups are shown as means, with bars indicating the SEM (Tables 1 and 3), unless otherwise stated. One-way ANOVA with Tukey posttest analysis or two-way ANOVA with Bonferroni posttest analysis were used to determine statistical significance. Where appropriate, we performed a two-tailed t test. P < 0.05 was considered statistically significant. Differential gene expression was determined by mRNA sequencing with edgeR (24), and differential pathway regulation was determined with GSEA-P (25). The Benjamini-Hochberg method was used to correct for the false discovery rate (FDR) in results from edgeR and GSEA-P (34). FDR-adjusted P values ≤0.05 were considered significant.

Gene expression, extracellular matrix protein levels, and macrophage infiltration were examined in the cortex of kidneys from nondiabetic and STZ-induced diabetic wild-type (WT) and RAGE^−/− mice at 24 weeks of diabetes. Plasma glucose and glycated hemoglobin were similarly elevated in both diabetic WT and diabetic RAGE^−/− mice (Fig. 1A and B). Urinary AER (Fig. 1C), creatinine clearance (Fig. 1D), glomerular damage, and GSI (Fig. 1E) all were elevated in diabetic WT mice but were significantly reduced in diabetic RAGE^−/− mice. The TIA was elevated in the kidneys of diabetic WT and RAGE^−/− mice (Fig. 1E). Although renal collagen IV staining was further increased in the cortex of diabetic RAGE^−/− mice, glomerular collagen IV levels were reduced compared with those from diabetic WT mice (Fig. 1F). Furthermore, glomerular expansion, which was increased in diabetic WT mice, was absent in the kidneys of diabetic RAGE^−/− mice (Fig. 1G). Similarly, genetic deletion of RAGE prevented increased expression of fibronectin and TGF-β1 protein induced by diabetes in the kidneys of diabetic WT mice (Fig. 1H and I).

The renoprotective effect of RAGE deletion was observed despite the elevated renal expression of COL1A1, α-SMA, MCP-1 (CCL2), TGFBR1, C5AR1, COL4A3, fibronectin (FN1), and Ki67 (MKI67) in the diabetic WT and diabetic RAGE^−/− mice (Fig. 2A–H). Renal MCP-1 protein levels were also elevated in kidneys from diabetic RAGE^−/− mice (Fig. 2I). The macrophage markers CD68 and CD169 were elevated in RAGE^−/− mouse kidneys and increased further in diabetic RAGE^−/− mice (Fig. 2J and K); however, the glomerular CD169⁺-to-CD68⁺ ratio, a measure of macrophage activation, was similar in both diabetic WT and diabetic RAGE^−/− mice (Fig. 2L).

Because renal function, GSI, mesangial expansion, and collagen IV expression were improved in the glomeruli of RAGE^−/− mice despite overall increased expression levels of fibrotic and inflammatory markers, we decided to study further the renoprotective effect of RAGE deletion, focusing on the glomerulus, specifically in MCs. RNA sequencing experiments were performed on primary mouse WT and RAGE^−/− MCs grown in the presence or absence of TGF-β. Multidimensional scaling analysis demonstrated distinct and reproducible differences among the four sample groups (Fig. 3A). The smear plot in Fig. 3B demonstrates significantly altered expression of 7,842 genes (red dots) following TGF-β treatment of WT MCs. Only 5,223 genes were significantly altered in by TGF-β in RAGE^−/− MCs (Fig. 3C, red dots). A factorial analysis investigating the effect of RAGE deletion on TGF-β-mediated gene regulation demonstrated that 5,246 genes were RAGE dependent (Fig. 3D). In this analysis, red dots in the positive region show a greater fold induction in RAGE^−/− MCs with TGF-β treatment, whereas those in the negative region demonstrate a reduced fold change in RAGE^−/− MCs in response to TGF-β. This is more clearly demonstrated in the scatterplot of fold changes with TGF-β treatment in WT cells compared with RAGE^−/− MCs (Fig. 3E). Relatively few genes show discordant fold changes.

The GSEA analysis of the reactome gene pathways are shown in Fig. 3F. Pathways such as protein metabolism, cholesterol biosynthesis, immune system adaptation, extracellular matrix organization, class I MHC–mediated antigen presentation, collagen formation, axon guidance, apoptosis, carbohydrate metabolism, and translation are significantly downregulated in RAGE^−/− MCs. On the other hand, only the generic transcription pathway is significantly upregulated in RAGE^−/− MCs. For example, cholesterol biosynthesis and collagen formation gene sets exhibit striking downregulation (Fig. 3G). Downregulation of these pathways is consistent with their induction by diminished TGF-β in RAGE^−/− MCs. A heatmap of cholesterol biosynthesis genes shows that TGF-β exposure causes upregulation of these genes in WT MCs, whereas these genes are downregulated in RAGE^−/− MCs (Fig. 3H). Similarly, collagen synthesis genes are downregulated in RAGE^−/− MCs; COL8A2 is one of the most significant differential expression events (FDR-adjusted P = 2.64E−39; Fig. 3I).

RT-PCR analysis of gene expression revealed upregulation of several fibrotic and inflammatory genes, transcription factors, and cell survival genes in RAGE^−/− MCs under baseline conditions. COL4A1, COL4A3, PKCα and PKCβ, and DIAPH1 (which is known to interact with RAGE [35]) were significantly downregulated (Table 1). The RAGE ligand S100A4 was upregulated 22-fold in RAGE^−/− MCs.

HG had no effect on baseline gene expression in WT MCs; however, baseline expression of COL1A1 and MCP-1 was significantly elevated in RAGE^−/− MCs under HG conditions (Table 2). TGF-β increased the expression of COL1A1, αSMA, PAI-1, fibronectin, and COL4A3 in WT MCs (Table 2). By contrast, the fold change in expression of these genes in response to TGF-β was attenuated in RAGE^−/− MCs. α-SMA protein was increased in RAGE^−/− MCs (Fig. 4A) compared with WT MCs. While TGF-β further increased α-SMA in WT MCs, it did not alter α-SMA levels in RAGE^−/− MCs. MCP-1 protein was also elevated in control RAGE^−/− MCs but was significantly reduced by TGF-β (Fig. 4B), a pattern similar to that seen with respect to MCP-1 RNA levels (Table 2). TGF-β increases fibronectin protein levels in both WT and RAGE^−/− MCs, but the effect was attenuated in the RAGE^−/− MCs, a pattern similar to that seen with respect to fibronectin RNA levels (Table 2). Collectively, these data demonstrate that RAGE^−/− MCs are less responsive to TGF-β than WT MCs, confirming the trends observed upon review of the RNA sequencing data (see Fig. 3).

Cell growth rate was significantly increased in RAGE^−/− MCs compared with WT MCs (Fig. 5A). Propidium iodide staining demonstrated fewer dead (positive staining) RAGE^−/− MCs over time than in WT MCs following FCS depletion (Fig. 5B). Staining for active caspase-3 following staurosporine treatment revealed fewer apoptotic nuclei in RAGE^−/− MCs than in WT MCs (Fig. 5C and D). Based on these data, RAGE^−/− MCs seemed to acquire a more proliferative and an antiapoptotic phenotype.

Smad3 and ERK phosphorylation were both significantly increased in RAGE^−/− MCs compared with WT MCs, indicating increased activity of the TGF-β and ERK pathways, respectively (Fig. 6A and B). S100A4, a ligand known to signal through RAGE and Toll-like receptors (TLRs), was significantly upregulated in RAGE^−/− cells (Fig. 6C). The level of p53 was also significantly reduced in RAGE^−/− MCs (Fig. 6D), consistent with the antiapoptotic phenotype of these cells. S100A4 knockdown by siRNA increased p53 protein levels (Fig. 6E) and decreased expression of α-SMA, MCP-1, TGFBR1, COL4A3, PAI-1, and Ki67 in RAGE^−/− MCs (Fig. 6F). These data support the view that RAGE deletion promotes the expression of cell survival genes and increases the baseline activity of profibrotic and inflammatory pathways.

Based on the data presented above, we hypothesized that RAGE may play a role in inflammatory, fibrotic, and cell survival processes under normal physiological conditions. RAGE knockdown by siRNA in WT MCs increased the expression of profibrotic and inflammatory genes (COL1A1, α-SMA, MCP-1, and TGFBR1; Supplementary Fig. 1). It is interesting to note that restoring the expression of FL-RAGE in RAGE^−/− MCs decreased expression of COL1A1, α-SMA, MCP-1, TGFBR1, PAI-1, and S100A4 to levels observed in WT MCs (Table 3). These data clearly demonstrate anti-inflammatory and antifibrotic roles for FL-RAGE in MCs.

It is also interesting that ES-RAGE, a novel splice variant of RAGE, was equally as effective as FL-RAGE in restoring the expression of most genes in RAGE^−/− MCs (Table 3). We explored this effect further and observed that HMGB1 increased the expression of COL1A1, α-SMA, NFκB-P65, TGFBR1, PAI-1, and COL4A3 in RAGE^−/− MCs (Supplementary Fig. 2). However, when this experiment was repeated with ES-RAGE adenovirus–infected RAGE^−/− MCs, HMGB1 had little effect on the expression of these genes.

By contrast, overexpression of FL-RAGE in WT MCs increased the expression of COL1A1, αSMA, MCP-1, TGFBR1, PAI-1, fibronectin, COL4A3, and S100A4 (Table 3). ES-RAGE decreased the expression of MCP-1, PAI-1, and S100A4 (Table 3) in these cells. Taken together, these data reveal a complex role for FL-RAGE in kidneys, where RAGE negatively regulates inflammatory and fibrotic processes under normal physiological conditions but also promotes these pathological processes when RAGE expression is increased.

In this study we examined the effect of genetic deletion of RAGE on gene expression in, and the structure and function of, diabetic kidneys. We observed the resolution of hyperfiltration, an early renal functional change seen in both human and experimental diabetes; less albuminuria; and less glomerular structural injury in diabetic RAGE^−/− mice compared with diabetic WT mice. These occurred despite the increased expression of a range of well-characterized fibrotic and inflammatory markers in the kidney cortexes of diabetic RAGE^−/− mice. Furthermore, increased macrophage infiltration was also observed in diabetic RAGE^−/− mouse kidneys, but it is important to note that this was not accompanied by increased renal injury. It is interesting that a previous study demonstrated the requirement for RAGE expression to mediate HMGB1 activation of mitogen-activated protein kinase and stress-activated protein kinase/c-Jun N-terminal kinase pathways in macrophages, but not to reduce TNF-α, interleukin-1β, and interleukin-6 levels (36). It was also demonstrated that diabetic WT mice reconstituted with RAGE^−/− bone marrow had improved renal function and reduced renal injury compared with diabetic WT mice reconstituted with WT bone marrow (37). These studies, along with our data, confirm that RAGE expression on hemopoietically derived immune cells contributes to the pathological changes seen in DN.

RNA sequencing of MCs derived from both WT and RAGE^−/− mice revealed gene expression differences involving several signaling pathways relevant to DN. In particular, the effect of TGF-β on gene expression was attenuated in RAGE^−/− MCs. RAGE is known to cooperate with TGF-β (5,38), the angiotensin II type 1 receptor subtype (38–41), and TLR signaling (41) pathways, resulting in the amplification of fibrotic and inflammatory responses (38–41). Our data are consistent with those from these studies, which demonstrate that renoprotection in RAGE^−/− MCs is due in part to the reduced responsiveness to growth factor stimulation of these injurious pathways in DN.

Despite the reduced responsiveness to TGF-β, expression of profibrotic and inflammatory genes was increased in RAGE^−/− MCs compared with WT MCs. Compensatory signaling via TGF-β, TLRs, and immune system pathways may account for this, as we observed elevated expression of S100A4, TGFBR1, TLR4, and C3aR1 (Table 1), as well as activated SMAD3 and ERK, which are downstream mediators of these pathways (Fig. 6). Many ligands mediate activation of TLR signaling, including S100A4, a member of the family of S100 calcium-binding proteins, which acts through TLR4 to activate MyD88, the transcription factor that drives nuclear factor-κB and the tyrosine kinases ERK1/2 and p38 expression in mononuclear cells (42). Liliensiek et al. (17) observed in the peritoneum the same proinflammatory pattern that we observed in the kidneys of RAGE^−/− mice in our study, although they were studying a different disease context.

The prosurvival phenotype observed in RAGE^−/− MCs may also contribute to renoprotection in diabetic RAGE^−/− mice. Glomerular cells are lost through apoptosis in experimental DN (43), and apoptosis of MCs correlates with worsening albuminuria (44). Patients with mild diabetic renal injury display a sixfold increase in apoptotic glomerular cells, which is associated with a loss of kidney function (45).

RAGE restoration experiments in RAGE^−/− MCs suggest a protective role for RAGE in the normal kidney against fibrotic and inflammatory processes. Delivery of FL-RAGE or ES-RAGE reversed the upregulation of profibrotic and inflammatory genes (Table 3). The reverse experiment, with knockdown of RAGE in WT MCs, complemented these observations and resulted in increased expression of fibrotic and inflammatory genes (Supplementary Fig. 1). These data support a homeostatic role for RAGE in normal kidney cells. RAGE plays a protective role under normal physiological conditions, in which it regulates the inflammatory response to a variety of different ligands, including AGEs. However, sustained activation of RAGE results in inflammation and promotes the microvascular and macrovascular complications of diabetes (46–48). Our RNA sequencing data indicate that RAGE enhances the TGF-β-mediated activation signaling pathways, including pathways related to protein metabolism, cholesterol biosynthesis, immune system adaptation, extracellular matrix organization, and collagen formation (Fig. 3). Consistent with our findings, Englert et al. (49) demonstrated that RAGE^−/− mice developed more severe lung fibrosis than WT controls in a model of pulmonary fibrosis.

RAGE overexpression highlighted the complex role of RAGE in normal physiology. FL-RAGE delivery to WT MCs increased the expression of profibrotic and inflammatory genes, whereas ES-RAGE decreased the expression of some of these genes (Table 3). Our data results are consistent with those of previous reports of a pathological role for FL-RAGE when overexpressed (6) and the beneficial effect of ES-RAGE (7), the decoy receptor, in diabetes.

RAGE^−/− MCs were resistant to apoptosis induced by staurosporine, through activation of caspase-3 (50), and by serum starvation (Fig. 5), potentially through upregulation of S100A4. Overexpression of S100A4 is observed in cancer (51–53) and directly interacts with and results in the degradation of nuclear p53 (53). Consistent with these reports, decreased p53 protein levels were observed in RAGE^−/− MCs (Fig. 6D), and knockdown of S100A4 resulted in increased p53 (Fig. 6E). The importance of S100A4 in fibrosis in other tissues has been previously been reported (52). Our findings suggest that upregulation of S100A4 contributes to the prosurvival and profibrotic phenotype of RAGE^−/− MCs.

RAGE plays a major role in the development and progression of DN and is highly expressed in response to various RAGE ligands that are elevated in diabetes (54). Our findings identify a key physiological role for RAGE in the kidney by regulating the expression of profibrotic and inflammatory genes and by influencing cell survival under normal conditions. The direct targeting of FL-RAGE may therefore not be an ideal therapeutic approach; rather, strategies that target the AGE-RAGE interaction either by directly inhibiting ligands such as HMGB1 and AGEs, or by using a decoy receptor such as sRAGE, may be preferable.

Funding. S.H. was supported by the Manpei Suzuki Diabetes Foundation. This study was also supported by the Diabetes Australia Research Trust. M.E.C. is supported by the National Health and Medical Research Council. P.K. was supported by the Vera Dalgleish Phillips Fellowship and the National Health and Medical Research Council. This study was also supported in part by the Victorian Government Operational Infrastructure Support Program.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.H. researched data and wrote the manuscript. K.S., M.Z., W.T., M.M., A.D.M., E.B., M.C., B.H., S.P., B.W., G.H., R.P., A.E.-O., M.C.T., and P.K. researched data. J.F., M.E.C., and P.K. reviewed and edited the manuscript. P.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.