Lipid-associated macrophages' promotion of fibrosis resolution during MASH regression requires TREM2

Abstract

While macrophage heterogeneity during metabolic dysfunction-associated steatohepatitis (MASH) has been described, the fate of these macrophages during MASH regression is poorly understood. Comparing macrophage heterogeneity during MASH progression vs regression, we identified specific macrophage subpopulations that are critical for MASH/fibrosis resolution. We elucidated the restorative pathways and gene signatures that define regression-associated macrophages and establish the importance of TREM2⁺ macrophages during MASH regression. Liver-resident Kupffer cells are lost during MASH and are replaced by four distinct monocyte-derived macrophage subpopulations. Trem2 is expressed in two macrophage subpopulations: i) monocyte-derived macrophages occupying the Kupffer cell niche (MoKC) and ii) lipid-associated macrophages (LAM). In regression livers, no new transcriptionally distinct macrophage subpopulation emerged. However, the relative macrophage composition changed during regression compared to MASH. While MoKC was the major macrophage subpopulation during MASH, they decreased during regression. LAM was the dominant macrophage subtype during MASH regression and maintained Trem2 expression. Both MoKC and LAM were enriched in disease-resolving pathways. Absence of TREM2 restricted the emergence of LAMs and formation of hepatic crown-like structures. TREM2⁺ macrophages are functionally important not only for restricting MASH-fibrosis progression but also for effective regression of inflammation and fibrosis. TREM2⁺ macrophages are superior collagen degraders. Lack of TREM2⁺ macrophages also prevented elimination of hepatic steatosis and inactivation of HSC during regression, indicating their significance in metabolic coordination with other cell types in the liver. TREM2 imparts this protective effect through multifactorial mechanisms, including improved phagocytosis, lipid handling, and collagen degradation.

Keywords: Trem2; fibrosis; lipid associated macrophages (LAM); macrophage; steatohepatitis.

Fig. 1.

TREM2 in MASH. (A) Volcano plot of liver RNAseq demonstrating up-regulated (orange) and down-regulated (Green) genes common to mice and human MASH vs. healthy obese (see SI Appendix for methods). (B and C) Formalin-fixed paraffin-embedded (FFPE) human (B) and Alms1^−/− (Foz/Foz) mice (C) liver sections were costained with anti-type 1 collagen (red) and TREM2 (green) antibodies (Scale bar, 30 µm for human and 10 µm for mouse sections) with corresponding quantifications. (D) qRT-PCR analysis of Trem2 in WT and Foz/Foz livers. (E) Experimental design. Six- to eight-week-old Foz/Foz and Foz::Trem2^−/− (T2^−/− Foz) mice were placed on WD for 8 to 24 wk. (F) Representative hematoxylin and eosin (H&E) and Sirius red (SR)–stained mouse liver sections and corresponding ImageJ quantifications. (Scale bar, 200 µm.) (G–J) Total liver RNA was subjected to qRT-PCR for (G) Col1α1, (H) Timp1, (I) Mcp1, and (J) Cd11b, gene expression. (K) Plasma lipopolysaccharide (LPS) and lipopolysaccharide-binding protein (LBP) levels were determined using ELISA. Data are expressed as mean ± SEM; one-way ANOVA and T-test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 2.

Absence of TREM2 prevents effective MASH and fibrosis resolution. (A) Experimental design: 6- to 8-wk-old Foz/Foz and Foz::Trem2^−/− were fed WD for 12 wk to induce MASH/fibrosis. One cohort of mice were then switched to a chow diet for an additional 4 to 8 wk to model MASH regression. Another cohort continued WD as age-matched controls (MASH progression 20 wk). Three distinct time-points were analyzed: 1) MASH 12 wk (MASH-fibrosis immediately before regression), 2) MASH progression 20 wk (MASH-fibrosis in age-matched controls that continued WD), 3) MASH regression (4 and 8 wk regression). (B and C) Representative H&E (B) and SR-stained (C) mouse liver sections (Scale bar, 200 µm) and (D) ImageJ quantifications. (E) Linear regression analysis of SR quantification in Foz/Foz and Foz::Trem2^−/− mice after 4 and 8 wk regression normalized to their respective 12 wk MASH SR scores. A greater negative slope in the Foz/Foz group indicates an increased rate of SR-reduction in these mice. (F) The extent of fibrosis resolution during MASH regression was quantified by normalizing SR-positive areas of Foz/Foz and Foz::Trem2^−/− mice undergoing regression with the corresponding SR-positive areas at 12-wk MASH (refer to Fig. 1F). This was done to normalize for the higher SR scores of Foz::Trem2^−/− + WD mice compared to Foz+WD mice at 12 wk, the time when the mice were put on regression. (G) Total liver RNA was analyzed by qRT-PCR for the expression of indicated inflammatory genes. (H) Liver lysates from regression and age-matched control mice were subjected to immunoblot analysis with the indicated antibodies. (I) Total liver RNA was analyzed by qRT-PCR for the expression of indicated genes. Data are expressed as mean ± SEM; one-way ANOVA; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 3.

Macrophage heterogeneity during MASH progression and regression. (A–F) Total nonparenchymal cells (NPC) isolated from healthy (Foz + WD 0 wk), MASH (Foz + WD 12 wk), and MASH-regression (Foz + WD 12 wk + Chow 8 wk) mice were subjected to scRNAseq. This analysis focuses on cells of macrophage and monocyte origin. Other NPCs, such as T-cells, B-cells, and dendritic cells (listed in SI Appendix, Fig. S3A), were excluded from this analysis. (A) UMAPs highlighting Monocyte-macrophage clusters, including cells from healthy, MASH, and regression. Individual UMAPs from each condition are shown in the Left panel and the merged UMAP on the Right. (B) Cluster composition plot showing the relative proportion of cells from healthy, MASH, and regression across all the clusters. (C) Heatmap showing relative average expression of cell-type marker genes (Y axis) with their cluster identity (X axis) for monocyte-macrophage subclusters. (D and E) Cluster composition plot showing the relative makeup (D), and cell counts (E), of each cluster in healthy control (Foz + WD 0 wk) and during MASH (Foz + WD 12 wk) and regression. (F) Violin plot showing Trem2 expression in the MoKC and LAM during MASH progression and regression.

Fig. 4.

Pathways affected by TREM2 in MASH and regression-associated macrophages. (A) A cartoon summarizing the meta-pathways (red/blue circles) and the corresponding GSEA pathways (boxes) which are positively (red) or negatively (blue) correlated with Trem2 expression in macrophages during MASH progression (see SI Appendix for methods). (B) Experimental design. Conditioned media (CM) from liver macrophages isolated from 24 wk FrWD-fed WT and Trem2^−/− mice were added to HSC isolated from chow-fed WT mice. (C) Total RNA isolated from the HSC 24 h post CM incubation was subjected to qRT-PCR to identify gene expression changes in fibrogenic markers. Gene expression is normalized to HPRT as well as to the number of live macrophages that produced the corresponding CM. (D) Heatmap showing relative mean expression of the genes belonging to Trem2 correlated pathways, of clusters 1, 2, 5, and 15 during MASH progression (MASH) and MASH regression (Reg) (see SI Appendix for methods). Data are expressed as mean ± SEM; T-test; *P < 0.05, **P < 0.01.

Fig. 5.

Absence of TREM2 suppresses emergence of LAMs during MASH progression and regression. (A–C) MASH progression. (A) Representative images of GPNMB (red) (DAPI = blue) (Scale bar, 200 µm) stained FFPE liver sections from Foz/Foz and Foz::Trem ^−/− mice undergoing MASH progression (12 wk). Corresponding quantification of GPNMB⁺ hCLS in each randomly selected high-power field (hpf) on the Right. (B) qRT-PCR analysis of total liver RNA for Gpnmb expression. (C) FFPE sections from these mice were also stained with anti-Cd11b antibody. (Scale bar, 200 µm.) Representative images are shown with corresponding quantifications. (D–F) MASH regression. (D) Representative images of GPNMB (red) (DAPI = blue) (Scale bar, 200 µm) stained FFPE liver sections and quantification. (E) qRT-PCR analysis of total liver RNA for Gpnmb expression. (F) Representative Cd11b stained images of FFPE liver sections with quantification. (G–I) Flow cytometry of liver NPC isolated from Foz/Foz and Foz::Trem2^−/− mice undergoing MASH progression and regression for LAM (CD9⁺), MoKC (Clec4f⁺), and Ly6C^hi macrophage subpopulations. Bar diagrams show the fold change in the proportion of (G) F4/80⁺CD9⁺, (H) F4/80⁺Clec4f⁺, and (I) Cd11b⁺Ly6c⁺ cells in the indicated groups compared to the Foz + WD 12 wk MASH group. Data are expressed as mean ± SEM; one-way ANOVA; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 6.

TREM2⁺ macrophages have high collagenolytic activities. (A) scRNAseq data from MASH and regression macrophage subpopulations (clusters 1, 2, 5, and 15) was analyzed as described in Figs. 3 and 4. Heat map showing relative expression of indicated genes belonging to reparative pathways (lipid handling, phagocytosis, and anti-fibrotic), among the various macrophage subpopulations during MASH and regression (Reg). These genes were selected based on a previous report (20), for their association with restorative macrophage phenotype during regression of CCl₄-induced liver fibrosis. Only statistically significant genes (P < 0.05) were plotted. (B–D) Biochemical analysis of collagen degradation. (B) The fluorescent signal in intact fluorescein-labeled DQ^TM type-1 collagen remains quenched. Upon substrate hydrolysis by collagenases, the fluorescence signal increases which is used to measure enzymatic activity. (C) WT and Trem2^−/− BMDMs were subjected to DQ^TM collagen assay. The measured fluorescence intensity was normalized to the cell count and represented as fold change compared to WT BMDM. (D) Foz/Foz and Foz::Trem2^−/− mice were subjected to MASH (12 wk WD) and regression (4 wk). Macrophages isolated from the livers of indicated groups and were subjected to the DQ^TM collagen assay. The fluorescence intensity was normalized to the protein content of the cell lysates and plotted. (E–H) WT and Trem2^−/− mice were administered CCl₄ for 6 wk, followed by 10 d recovery. (E) Experimental design. (F) Total liver RNA was subjected to qRT-PCR for Col1α1 and Timp1 genes, normalized to Hprt, and plotted as fold change compared to the WT + CCl₄ group. (G) Representative SR-stained mouse liver sections (Scale bar, 200 µm) and ImageJ quantifications. (H) The extent of fibrosis resolution during regression of CCl₄-induced liver fibrosis was calculated by normalizing SR-positive areas of WT and Trem2^−/− mice undergoing regression with the corresponding SR-positive areas at 6-wk CCl₄ dosing. Data are expressed as mean ± SEM; one-way ANOVA; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.