Macrophage RAGE activation is proinflammatory in NASH

Abstract

Intrahepatic macrophages in nonalcoholic steatohepatitis (NASH) are heterogenous and include proinflammatory recruited monocyte-derived macrophages. The receptor for advanced glycation endproducts (RAGE) is expressed on macrophages and can be activated by damage associated molecular patterns (DAMPs) upregulated in NASH, yet the role of macrophage-specific RAGE signaling in NASH is unclear. Therefore, we hypothesized that RAGE-expressing macrophages are proinflammatory and mediate liver inflammation in NASH. Compared with healthy controls, RAGE expression was increased in liver biopsies from patients with NASH. In a high-fat, -fructose, and -cholesterol-induced (FFC)-induced murine model of NASH, RAGE expression was increased, specifically on recruited macrophages. FFC mice that received a pharmacological inhibitor of RAGE (TTP488), and myeloid-specific RAGE KO mice (RAGE-MKO) had attenuated liver injury associated with a reduced accumulation of RAGE+ recruited macrophages. Transcriptomics analysis suggested that pathways of macrophage and T cell activation were upregulated by FFC diet, inhibited by TTP488 treatment, and reduced in RAGE-MKO mice. Correspondingly, the secretome of ligand-stimulated BM-derived macrophages from RAGE-MKO mice had an attenuated capacity to activate CD8+ T cells. Our data implicate RAGE as what we propose to be a novel and potentially targetable mediator of the proinflammatory signaling of recruited macrophages in NASH.

Keywords: Drug therapy; Hepatitis; Hepatology; Inflammation; Macrophages.

Figure 1. RAGE expression is upregulated in human and murine NASH.

(A) Relative mRNA expression of Ager in FFC- compared with chow-fed mice (n = 4 each), P < 0.05. (B) MCD- (n = 3) compared with MCS-fed mice (n = 5), P < 0.05. (C) Representative images of RAGE IHC from FFC (n = 6) and chow (n = 3) mice. Scale bar: 50 μm. (D) Quantification of C; P < 0.05. (E) Fold-change of mRNA expression of AGER in whole-liver RNA-Seq of biopsies obtained from patients with advanced NASH on the y axis (n = 12) relative to 2 groups on the x axis — normal obese (n = 9, P < 0.05) and simple steatosis (n = 10, P < 0.05). (F) Representative images of RAGE IHC from patients with NASH (n = 5) and healthy patients (n = 3). Scale bar: 50 μm. (G) Quantification from F; P < 0.05. Mann-Whitney U test was used for statistical analyses,

Figure 2. RAGE elevation in murine NASH is enriched on recruited macrophages.

(A) Representative co-IF images of liver cryosections from FFC mice (n = 6) compared with chow mice (n = 5), stained for DAPI, F4/80, and RAGE. A negative control is shown. Scale bar: 50 μm. (B) Schematic depicting experimental design for isolating IHLs for flow cytometry from FFC mice. (C) Quantification of RAGE⁺ among F4/80^–CD45⁺ (nonmacrophages [NonMacs]) and F4/80⁺CD45⁺ (macrophages [Macs]) cells in livers from chow and FFC mice (n = 4 chow, n = 5 FFC, P < 0.01). Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons. (D) Representative flow cytometry gating demonstrating RAGE expression, depicted as a heatmap of RAGE expression intensity, among recruited (CD11b^hiF4/80^int) and resident macrophages (CD11b^intF4/80^hi) in chow and FFC mice. (E) Quantification of RAGE⁺ recruited macrophages in chow (n = 5) and FFC (n = 6) mice; P < 0.05. (F) Quantification of RAGE⁺ resident macrophages in chow (n = 5) and FFC (n = 6) mice; P < 0.05. Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 3. Pharmacological inhibition of RAGE attenuates diet induced NASH.

(A) Schematic depicting experimental design for pharmacological inhibition of RAGE. (B) Representative images of H&E-stained livers from vehicle- or TTP488-treated chow (n = 7 and n = 8, respectively) and FFC mice (n = 11 and n = 10, respectively) demonstrating inflammatory foci (inset showing region of interest at same magnification). Scale bar: 50 μm. (C) Distribution of inflammation and steatosis subscores of the NAFLD activity score. (D) Comparison of serum ALT levels standardized to vehicle-treated chow mice (Chow-vehicle n = 7, Chow-TTP n = 8, FFC vehicle n = 10, FFC-TTP n = 9; P < 0.05). (E) Comparison of body mass (Chow-vehicle n = 7, Chow-TTP n = 8, FFC vehicle n = 10, FFC-TTP n = 9; P < 0.001) and relative liver mass (Chow-vehicle n = 7, Chow-TTP n = 8, FFC vehicle n = 10, FFC-TTP n = 9; P < 0.001). (F) Liver triglyceride content (Chow-vehicle n = 4, Chow-TTP n = 4, FFC vehicle n = 5, FFC-TTP n = 5; P < 0.05). Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 4. TTP488 attenuates accumulation of RAGE⁺ recruited macrophages.

(A) Quantification of CD45⁺ cells in livers from vehicle- or TTP488-treated chow and FFC mice (Chow-vehicle n = 3, Chow-TTP n = 4, FFC vehicle n = 5, FFC-TTP n = 7; P < 0.05). (B) Total macrophages (P < 0.05). (C) Resident macrophages (CD11b^intF4/80^hi) (P = 0.06). (D) Recruited (CD11b^hi F4/80^int) macrophages (P < 0.05). (E) Representative flow cytometry gating depicting recruited (CD11b^hiF4/80^int) and resident macrophages (CD11b^intF4/80^hi) macrophages. (F) Quantification of RAGE⁺ recruited macrophages (P < 0.05). (G) Quantification of RAGE⁺ resident macrophages (P < 0.05). Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 5. TTP488 attenuates diet-induced changes in the intrahepatic immune transcriptome.

(A) Principal component analysis plots comparing transcriptome of IHLs from chow and FFC mice (n = 3 each). (B) Vehicle- and TTP488-treated FFC mice (n = 3 each). (C) Bubble plot of IPA canonical pathways significantly upregulated (orange) and downregulated (blue) with FFC diet (n = 3 each). (D) TTP488 treatment (n = 3 each). (E) Venn diagram depicting overlap of significantly different mRNA transcripts common to the experimental comparisons of diet and treatment. (F) Representative IPA canonical pathways that were significantly differentially regulated by FFC diet and TTP488 treatment.

Figure 6. Mass cytometry identifies subsets of RAGE-expressing macrophage influenced by TTP488.

(A) Histograms depicting changes in RAGE expression on macrophages (CD45⁺F4/80⁺) with FFC diet and TTP488 treatment (n = 4 each). (B) viSNE plots depicting expression of representative surface-markers. (C) Quantification of RAGE⁺CCR2⁺ (n = 4 each; P < 0.05) and RAGE⁺CCR2^– macrophages (n = 4 each; P = 0.7). Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 7. Unbiased hierarchical clustering of macrophages reveals distinct RAGE-enriched subsets.

(A) CITRUS hierarchical tree with shaded groups of clusters that are differentially abundant between the experimental groups. The numbered groups (denoted as 1–3) represent clusters with RAGE enrichment. (B–E) Differential expression of representative surface markers colored based on intensity of expression are depicted in CCR2 (B), MHCII (C), RAGE (D), and TIM4 (E).

Figure 8. RAGE activation on macrophages is associated with induction of IRFs.

(A) Histograms depicting expression of IRF5 and IRF7 on macrophages (CD45⁺F4/80⁺) assessed by CyTOF with FFC diet and TTP488 treatment. The arrows denote an increase in RAGE count and intensity among FFC vehicle compared with chow and a decrease in FFC TTP488. (B) Representative CyTOF biplots depicting coexpression of RAGE and IRF5 and IRF7 among recruited macrophages (CD45⁺/F4/80⁺/MHCII⁺TIM4^– cells); (C) Quantification of abundance of IRF5⁺ and IRF7⁺ recruited macrophages (CD45⁺/F4/80⁺/MHCII⁺TIM4^– cells) (n = 4 each; P < 0.05). Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 9. RAGE-expressing T cell subsets are unchanged by RAGE inhibition.

(A and B) Flow cytometry based quantification of CD4⁺ T cells (Chow n = 3 each, FFC n = 4 each; P = 0.07) (A) and CD8⁺ T cells (Chow n = 3 each, FFC n = 4 each; P < 0.05) (B). (C–E) CyTOF-based quantification of percentage of RAGE⁺ among CD8⁺ T cells (Chow n = 3 each, FFC n = 4 each; P = 0.2) (C); RAGE⁺CD8⁺ T cells (Chow n = 3 each, FFC n = 4 each; P = 0.4) (D); and PD1⁺CD8⁺ T cells (Chow n = 3 each, FFC n = 4 each; P < 0.05) (E). Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 10. RAGE MKO attenuates NASH.

(A) Schematic depicting experimental design for RAGE MKO. (B) Representative images of H&E-stained livers from the chow and FFC WT and MKO mice (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 5, FFC-MKO n = 6). Scale bar: 50 μm. (C) Distribution of inflammation and steatosis subscores of the NAFLD activity score. (D) Serum ALT (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 5, FFC-MKO n = 6; P < 0.05). (E) Body mass (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 5, FFC-MKO n = 6; P < 0.01). (F) Relative liver mass (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 5, FFC-MKO n = 6; P < 0.01). Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 11. RAGE MKO reduces RAGE expression.

(A) Representative images of RAGE IHC from chow and FFC WT and MKO mice. Scale bar: 50 μm. (B) Quantification of A (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 5, FFC-MKO n = 6; P < 0.05). (C) Quantification of CD45⁺ cells by flow cytometry (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 4, FFC-MKO n = 3; P = 0.7). (D) Quantification of resident macrophages (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 4, FFC-MKO n = 3; P = 0.9). (E) Quantification of recruited macrophages (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 4, FFC-MKO n = 3, P = 0.2). (F) Quantification of RAGE⁺ recruited macrophages (Chow-WT n = 3, Chow-MKO n = 3, FFC-WT n = 4, FFC-MKO n = 3; P < 0.05). Cell numbers are presented as cell count per gram of liver. Mann-Whitney U test was used for statistical analyses, and P values were adjusted by the Benjamini-Hochberg method for multiple comparisons.

Figure 12. Macrophage RAGE signaling mediates proinflammatory crosstalk with CD8⁺ T cells.

(A) Schematic depicting experimental design for demonstrating in vitro crosstalk between macrophages and CD8⁺ T cells. BM-derived macrophages (BMDM) isolated from WT or MKO mice were stimulated ex vivo with recombinant S100A11, a known RAGE agonist and the supernatant used to stimulate CD8⁺ T cells. Release of IFN-γ was measured by ELISA. (B) Quantification of IFN-γ release from CD8⁺ T cells measured by ELISA, normalized to untreated CD8⁺ T cells. The horizontal axis shows source of supernatant from treated BMDMs (n = 3 each; P < 0.05). Mann-Whitney U test was used for statistical analyses. (C) Proposed model depicting the proinflammatory role of RAGE⁺ macrophage–T cell crosstalk in NASH. Herein, we propose that RAGE upregulation occurs in human and murine NASH, specifically on recruited macrophages. RAGE activation on recruited macrophages, potentially by DAMPs, leads to upregulation of IRFs, which may mediate proinflammatory crosstalk with T cells. Interruption of RAGE with a pharmacological inhibitor or with RAGE MKO is sufficient to ameliorate NASH.