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

1Metabolic dysfunction-associated steatohepatitis (MASH), formerly known as nonalcoholic steatohepatitis (NASH), is a major risk factor for cirrhosis and hepatocellular carcinoma (HCC) and the second leading cause of liver transplants in the United States. Activation and recruitment of macrophages are an early response to hepatocyte injury that play an important role in MASH with fibrosis development. Although the molecular pathways in macrophages that promote MASH and fibrosis have been investigated, little is known about the pathways that restrain macrophage activation within the disease microenvironment.

2Triggering receptors expressed on myeloid cells (TREM) are cell surface receptors that participate in diverse cellular processes. MASH is associated with the emergence of lipid-associated macrophages (LAM) that highly express Trem2 along with Cd9, Gpnmb, and Spp1 and are equivalent to the scar-associated macrophages (SAM) in human MASH. The functional role of hepatic LAM/SAMs, whether they promote or protect against MASH, remains unresolved. LAMs/SAMs increase with MASH severity, localize around steatotic hepatocytes (LAM) and fibrotic scars (SAM), activate hepatic stellate cells (HSC), the major fibrogenic cells in the liver, and promote TGF-β1-induced collagen deposition, suggesting that these TREM2hi macrophages may be pathogenic. However, recent reports indicate that the TREM2 molecule itself plays a protective role in liver damage, steatosis, and MASH.

3Single-cell analyses have deepened our understanding of macrophage heterogeneity during MASH progression, but the composition of macrophage subpopulations during MASH regression remains obscure. For example, it is not known whether a distinct restorative macrophage subpopulation emerges during MASH regression. While TREM2 expression increases in LAM/SAM during MASH progression, the fate and the functional role of these TREM2+ macrophages during MASH regression are unknown. Furthermore, the identity and hallmarks of the key macrophage subpopulations that facilitate MASH and fibrosis resolution when the damage-evoking agent is removed (such as Western diet) are unknown.

4Most of the existing knowledge regarding liver fibrosis resolution is derived from mouse models of chemical (CCl4 or acetaminophen)-induced fibrosis and regression. While these models offer insights into non-MASH-related liver fibrosis, they do not mimic the key pathophysiological aspects of human MASH patients such as obesity, steatosis, insulin resistance, global metabolic syndrome, and emergence of LAMs.

5Here, we used two preclinical mouse models of MASH that resemble human pathobiology to elucidate macrophage heterogeneity in MASH progression and regression. We determine the dominant macrophage subtype during the various phases of the disease (healthy, MASH, and regression); identify the key macrophage subtype that facilitates MASH/fibrosis resolution; and address the mechanistic role of TREM2 in MASH/fibrosis resolution. Using single-cell RNAseq (scRNAseq), we identified four distinct macrophage clusters that expand during MASH and determined their fate during MASH regression. We further determined how the key cellular pathways enriched in these clusters are affected during regression. We demonstrate that the regression-associated macrophage (RAM) subpopulations that maintain LAM/SAM gene signatures, including high Trem2 expression, are the key restorative macrophages that orchestrate MASH-fibrosis resolution. During MASH regression, absence of TREM2+ macrophages not only prevented efficient collagen resorption but also affected elimination of hepatic steatosis and HSC inactivation, indicating their significance in metabolic coordination with other cell types. Mechanistically, TREM2 regulates multiple restorative pathways, including increased collagenase activity, phagocytosis, improved lipid handling, thereby, conferring a protective (during MASH progression) as well as restorative roles (during resolution phases) to the LAM/SAMs.

6Mice with the Alms1 mutation (Foz/Foz mice) were kindly provided by Dr. Geoffrey C. Farrell (Australian National University Medical School) and have been further characterized by our laboratory. Trem2−/− mice were from the Jackson Laboratory (Strain# 027197). Foz::Trem2−/− mice were generated by crossing Foz/Foz with Trem2−/− mice. All mice were on a C57BL/6J background, and littermates were used for experiments. Six- to eight-week-old (time point: 0 wk) males were placed on a Western diet (WD) (AIN-76A; Test Diet, St. Louis, MO, containing 40% fat, 15% protein, 44% carbohydrates based on caloric content, and 0.2% cholesterol) or on a standard chow diet (12% fat, 23% protein, and 65% carbohydrates) for up to 24 wk. For the regression study, Foz + WD and Foz::Trem2−/− + WD 12 wk mice were switched to the chow diet for an additional 8 to 12 wk. MASH was induced in WT and Trem2−/− males by feeding them WD along with 30% (w/v) fructose in drinking water (FrWD) for up to 24 wk. All animals were maintained in accordance with NIH guidelines and approved by the University of California San Diego Institutional Animal Care and Use Committee (IACUC, Protocol #S07022).

7Data are shown as means ± SEM. Group differences were compared using ANOVA and Student’s t test. Significance was set at P < 0.05 unless otherwise mentioned. GraphPad Prism was used for all nonsequencing analyses.

8Additional methods are described in SI Appendix.

9When fed a Western diet (WD), Foz/Foz mice (Foz+WD) progress to MASH-fibrosis within 8 to 12 wk and to cirrhosis and HCC by 24 wk. The Foz+WD liver gene signature strongly correlates with that of human MASH. TREM2 is one of the most highly up-regulated genes in both Foz and human MASH livers (Fig. 1A). Previous studies have reported that TREM2 is expressed in recruited monocytes, liver-resident Kupffer cells, hepatic stellate cells (HSC), and endothelial cells. However, analyses of two publicly available scRNAseq datasets indicate that macrophages are the major cell types that express TREM2 in the MASH livers (SI Appendix, Fig. S1.1A). In fibrotic human MASH as well as in Foz/Foz livers, TREM2+ macrophages are located next to collagen scars (Fig. 1 B and C). This spatial association has led to their SAM designation. TREM2+ macrophages are also classified as LAM that are found to be associated with hepatic crown-like structures (hCLS) surrounding hypertrophic lipid-laden hepatocytes (SI Appendix, Fig. S1.1B) (discussed more in Fig. 5).

10Although TREM2 has been shown to be up-regulated in advanced MASH livers, at what stage TREM2 is induced during the disease progression and what event (such as steatosis, fibrosis, or hepatocyte death) triggers TREM2 upregulation are not clear. A time-course analysis of Trem2 expression in WT and Foz/Foz livers demonstrated the onset of Trem2 upregulation in Foz + WD livers at 8 wk post-WD feeding (Fig. 1D), which coincides with the onset of liver injury. Trem2 continued to increase as MASH and fibrosis progressed. Trem2 expression remained low in WT + WD 12 wk mice, which only had MAFL (Fig. 1D), indicating that Trem2 expression is associated with hepatocyte death and liver inflammation during the progression of MAFL to MASH.

11Recent reports indicate that TREM2 is protective during MASH development. To model MASH regression and evaluate macrophage heterogeneity in livers undergoing MASH/fibrosis resolution as well as to evaluate the mechanistic role of TREM2+ macrophages in the various phases of the disease, we modeled MASH progression and regression using two different preclinical mouse models. To assess the functional role of TREM2 in the development of MASH-fibrosis in our models, we generated Foz::Trem2−/− mice, fed them WD, and compared them to Foz + WD mice (Fig. 1E). The absence of TREM2 did not affect body weight, liver weight (SI Appendix, Fig. S1.1C), liver injury (SI Appendix, Fig. S1.1D), or liver steatosis (Fig. 1F). However, Foz::Trem2−/− mice developed more fibrosis (Fig. 1 F–H) and inflammation (Fig. 1 I and J).

12Foz/Foz mice advance to cirrhosis after 24 wk on WD23. At later time points (20 to 24 wk WD), the differences in fibrosis between the two groups diminished since the disease scores reached a point of saturation (SI Appendix, Fig. S1.1E). While the TREM2 slowed MASH development, the long-term continuation of WD-induced liver injury negated the TREM2-mediated protective mechanisms. This also indicates that TREM2 signaling axis alone is insufficient to provide complete protection against long-term MASH development. Therefore, when using Foz/Foz mice to study TREM2 signaling, we evaluated mice at 8 to 12 wk post-WD feeding.

13Using a second model of MASH, we assessed whether Trem2−/− by itself (independent of the Alms1 mutation found in Foz/Foz mice) produced the same phenotype (SI Appendix, Fig. S1.1F). When fed WD+30% fructose in drinking water (FrWD), both WT and Trem2−/− littermates had similar gains in body and liver weight liver injury and hepatic steatosis (SI Appendix, Fig. S1.1 G–I). However, deletion of Trem2 resulted in more severe MASH and fibrosis (SI Appendix, Fig. S1.1 I and J), consistent with the Foz/Foz model.

14Moreover, we found that Trem2−/− mice, but not their WT littermates, on FrWD progressed spontaneously from MASH-fibrosis to HCC (SI Appendix, Fig. S1.2 A–D), indicating that TREM2 is protective in MASH-associated HCC. The NT area showed extensive fibrosis (SI Appendix, Fig. S1.2B). About 60% of Trem2−/− mice developed an average of 5 tumors/liver with an average diameter of 2.5 mm (SI Appendix, Fig. S1.2C).

15Endotoxemia due to gut-barrier disruption and translocation of bacterial products (e.g., LPS) to the liver is a hallmark of MASH. It remains unclear whether the aggravated hepatic inflammation in Trem2−/− mice is due to increased gut permeability. In both MASH models, the systemic levels of LPS and LPS binding protein (LBP) remained unaltered irrespective of the presence or absence of TREM2 (Fig. 1K and SI Appendix, Fig. S1.2E). Conversely, in the absence of TREM2, MASH livers had higher levels of activated Caspase 1 (SI Appendix, Fig. S1.2F), indicating that Trem2−/− MASH livers have enhanced NLRP3 inflammasome activation, a key activator of liver inflammation in MASH. Indeed, consistent with our in vivo data, studies with WT and Trem2−/− BMDMs have reported an inherent hyperinflammatory phenotype of Trem2−/− BMDM in response to LPS as well as an increased NLRP3 activation in the context of bacterial infection and neuro-inflammation.

16Although Trem2−/−BMDMs have increased inflammatory responses, the specific hepatic chemokines/cytokines that are modulated by TREM2 signaling in MASH are unknown. We therefore designed a panel of 12 frequently studied cytokines/chemokines in MASH and subjected the liver lysates (from both preclinical models) to Luminex analyses. Both models had a significant increase in the levels of KC, MCP1, MIP2, IL1α, IL33, C1qR1, and S100A8 in the absence of TREM2, while the levels of TNFα, IL6, PDGFB, S100A9, and CRP were not significantly affected by TREM2 (SI Appendix, Fig. S1.2G).

17MASH in Foz+WD 12 wk mice regresses when the WD is switched to a standard chow diet for an additional 8 to 12 wk (mimicking lifestyle modification in humans). MASH regression is characterized by downregulation of profibrotic and inflammatory genes, HSC inactivation, and resolution of steatosis and fibrosis.

18Although macrophages are pivotal in both the progression and regression of MASH, the characteristics of restorative macrophages that orchestrate fibrosis resolution during MASH regression have not been identified. To evaluate whether TREM2+ macrophages can modulate regression of MASH-fibrosis, we subjected Foz + WD 12 wk and Foz::Trem2−/− + WD 12 wk MASH mice to a regression protocol by switching the WD back to a chow diet (Fig. 2A). To model the kinetics of MASH/fibrosis resolution, we harvested tissues at 4 and 8 wk post initiation of regression. A separate cohort of age-matched mice continued WD for 8 wk (20 wk on WD total).

19The WD to chow switch started to improve MASH within 4 wk with continued resolution through 8 wk. Strikingly, while Foz/Foz mice successfully eliminated >50% of hepatic fat deposits by 8 wk of WD to chow switch, Foz::Trem2−/− mice failed to eliminate hepatic steatosis (Fig. 2 B and D) and had higher liver weights compared to Foz/Foz regression mice (SI Appendix, Fig. S2A). Moreover, resolution of fibrosis in Foz::Trem2−/− mice was significantly impaired compared to the Foz/Foz mice (Fig. 2 C–F). Foz/Foz mice and Foz::Trem2−/− had equivalent steatosis and fibrosis when the mice continued on WD till 20 wk (Fig. 2 C–F), since the disease phenotype catches up by this time (as discussed in SI Appendix, Fig. S1.1E).

20Foz::Trem2−/− + WD 12 wk mice had more fibrosis than Foz/Foz mice (Fig. 1F) when they were subjected to regression. Therefore, to evaluate the effective reduction in collagen levels following the dietary switch, the SR score of each regression group was normalized to the corresponding MASH SR score (at 12 wk WD just before the switch) as previously reported for CCl4 fibrosis resolution studies. The resulting linear regression plot (Fig. 2E) shows a significantly greater fibrosis reduction in Foz/Foz mice (slope: −0.087) compared to Foz::Trem2−/− (slope: −0.039). Similarly, percent fibrosis resolution (normalized to 12 wk MASH scores) also indicates significantly greater fibrosis reduction in Foz/Foz mice (70% resolution at 8 wk) compared to Foz::Trem2−/− mice (30% resolution at 8 wk) undergoing regression (Fig. 2F).

21Absence of TREM2 slowed the resolution of liver inflammation, as indicated by higher residual Il1α and Il1β gene expression (Fig. 2G) and higher NLRP3 activation (Cleaved Caspase 1 levels) (Fig. 2H). Absence of TREM2 also impaired effective HSC inactivation indicated by higher residual levels of αSMA (activated HSC marker) and desmin (total HSC marker) in Foz::Trem2−/− regressed mice compared to their Foz/Foz counterparts (Fig. 2H and SI Appendix, Fig. S2B). Interestingly, even though Foz::Trem2−/− mice had greater residual liver collagen deposits, the gene expression of Col1α1 and Col3α1 were equally suppressed during regression in Foz/Foz and Foz::Trem2−/− mice (Fig. 2I). Therefore, the higher amount of residual collagen in Foz::Trem2−/− mice (Fig. 2 D−F) most likely reflects a defect in collagen degradation.

22We analyzed macrophage heterogeneity in Foz + WD 12 wk MASH livers and the fate of these MASH-associated macrophage subpopulations during MASH regression. We further elucidate the key molecular pathways by which TREM2 exerts its protective function. scRNAseq was performed on immune cells isolated from healthy (Foz + WD 0 wk), MASH (Foz + WD 12 wk), and regression mouse livers (SI Appendix, Fig. S3 A and B). From the 18 different immune cell clusters (SI Appendix, Fig. S3 A and B and Dataset S1), we focused our analysis on 6 monocyte/macrophage clusters (clusters 0, 1, 2, 5, and 15) (Fig. 3A). The identity of each macrophage subcluster was assigned based on previously published cell-type identity marker genes (Fig. 3 B and C and SI Appendix, Fig. S3C).

23As expected, chow→WD switch changed the macrophage transcriptome (Fig. 3 A and B). In contrast, WD→chow (regression) did not induce dramatic changes in macrophage heterogeneity, and no new subcluster emerged during MASH regression compared to MASH (Fig. 3 A and B). Regression macrophage subclusters (red) aligned with MASH macrophage subclusters (Blue) (Fig. 3 A and B).

24Clusters 0 and 8 were represented by macrophages from the healthy livers and were enriched with Timd4+embryo-derived Kupffer cells (EmKCs) (clusters 0, 8: Timd4+, Vsig4+, Clec1b+, Clec4f+, Ccr2−) (Fig. 3 A−C and SI Appendix, Fig. S3C). In MASH livers the Timd4+EmKCs were depleted, as reported, and were replaced by four distinct Timd4-clusters (clusters 5, 15, 1, 2) that were derived from circulating monocytes, also known as recruited macrophages (RM) (Fig. 3 A−C and SI Appendix, Fig. S3C).

25Cluster 5 (hereafter referred as Ly6chi RM) had a characteristic gene expression of previously described Ly6c-high macrophages that were Ly6c2hi, Chil3hi, Ccr2hi, Lyz2hi, Fn1hi, whereas cluster 15 (hereafter referred as Mac1) highly expressed Itgax (Cd11c) and Clec4a1 along with Fcrl5, Fabp4, Ighm, Cd300e, Fcgr4 (Fig. 3 A–C and SI Appendix, Fig. S3C). A defining feature of Cluster 2 was high Spp1 (osteopontin) and Gpnmb expression (Fig. 3 A–C and SI Appendix, Fig. S3C), similar to the recently described hepatic LAM subpopulation. All three clusters (clusters 2, 5, and 15) lack expression of EmKC markers (Vsig4, Clec4f, Clec1b, and C1q). Cluster 1 highly expressed Adgre1 (F4/80), along with several other EmKC markers Vsig, Clec4f, Clec1b, C1q even though they remain Timd4-, Cd63int, Cd9int, Gpnmbint (Fig. 3 A–C and SI Appendix, Fig. S3C). Thus, cluster 1 represents macrophages that are similar to but distinct from the EmKCs known as monocyte-derived Kupffer cells (MoKC), which occupy the EmKC niche during MASH.

26Trem2 is highly expressed in both LAM and MoKC subpopulations (Fig. 3C). Besides Trem2, both LAM and MoKC subpopulation express Gpnmb, Cd9, and Fabp5 (albeit higher expression in LAM compared to MoKC) (SI Appendix, Fig. S3 C–E), that are characteristics of SAM described in humans. This indicates that SAM might be a subset of LAM and/or MoKC.

27We analyzed the scRNAseq data to evaluate the relative makeup of each macrophage cluster in healthy control (Foz + WD 0 wk), during MASH progression (Foz + WD 12 wk), and MASH regression (Fig. 3D). EmKCs (clusters 0 and 8), the major macrophage subtype in the healthy livers, are replaced by the four monocyte-derived macrophage populations [MoKC, LAM, Ly6chi-RM, and Mac1] during MASH (Fig. 3D). Among these, MoKC (Cluster 1) was the most dominant macrophage subpopulation (Fig. 3D). During MASH regression, the total macrophage population decreases, as expected, because of improvement in overall liver health due to the removal of the disease-triggering WD (Fig. 3 D and E). In regression livers, while no new transcriptionally distinct macrophage subpopulation emerged (compared to MASH), the relative macrophage cluster composition changed. During MASH regression, MoKC population shrunk (Fig. 3 D and E), while the LAM population (Cluster 2) remained constant (Fig. 3E), making it the major macrophage subpopulation during regression (Fig. 3 D and E). Although MoKC still expressed Trem2 during regression, its expression was reduced compared to that in MASH (Fig. 3F). The LAMs, however, maintained high Trem2 expression during both MASH and regression (Fig. 3F). EmKC populations (clusters 0 and 8) that were mostly lost during MASH did not return (albeit a slight recovery) during the 8 wk of regression (Fig. 3 D and E).

28The macrophages in MASH livers are more inflammatory and profibrogenic compared to the macrophages residing in a healthy liver (SI Appendix, Fig. S4A). To identify the key biological differences among macrophage population that exist within the same MASH liver but are distinguished based on Trem2 expression levels, we first computed the Pearson correlation of all genes with Trem2 across all macrophage clusters (clusters 0, 1, 2, 5, 8, and 15). These correlation values were then used in a gene set enrichment analysis (GSEA) to identify pathways enriched with genes that are positively or negatively correlated with Trem2 expression (Fig. 4A). We found that the majority of significantly enriched “Reactome pathways” (rectangular boxes, Fig. 4A) can be broadly categorized into six major meta-pathways (Red/Blue circles, Fig. 4A). Genes positively correlated with high Trem2 expression in MASH-associated macrophages are enriched for meta-pathways such as phagocytosis, anti-fibrosis, and improved lipid/lipoprotein handling. On the other hand, MASH-associated macrophages expressing low levels of Trem2 were enriched for meta-pathways that correspond to inflammation, enhanced cytokine signaling, and cell death (apoptosis, pyroptosis, and regulated necrosis) (Fig. 4A and Datasets S2 and S3).

29Trem2hi macrophages are enriched in antifibrotic pathways, involving ECM degradation, and downregulation of TGF-β and related fibrotic signaling. However, conditioned media from TREM2+ SAM were reported to be more fibrogenic compared to their TREM2− non-scar-associated counterparts. We hypothesized that the absence of TREM2 in SAM would make them even more fibrogenic. To test this, we isolated macrophages from WT and Trem2−/− MASH mice and used the conditioned media to stimulate primary HSC isolated from healthy mice. Conditioned media from Trem2−/− MASH macrophages activated primary HSC more than the conditioned media from WT MASH-associated macrophages, as evidenced by increased Col1α1, Col3α1, and αSMA gene expression in HSC (Fig. 4 B and C). Thus, TREM2 dampens the fibrotic effect of macrophages within the MASH environment.

30To elucidate the functional role of different macrophage subpopulations in MASH progression and how these functions are altered during regression, we performed pathway enrichment analysis on the differentially regulated genes within the macrophage subpopulations during MASH and compared them with MASH regression. Within the MASH macrophage populations, the MoKC and LAM were relatively less inflammatory and enriched in protective pathways during MASH progression (Fig. 4D), consistent with the observation that MoKC and LAMs highly express Trem2 (Fig. 3C), and Trem2hi macrophages positively correlate these protective pathways (Fig. 4A). During regression, both MoKC and LAM continued to highly express disease-resolving pathways such as efficient phagocytosis, superior lipid handling, ECM degradation, and anti-fibrotic pathways (Fig. 4D). On the other hand, Ly6chi-RM and Mac1 continued to be enriched in inflammatory pathways even during the regression phase (Fig. 4D).

31Differential gene expression analysis revealed that only 25 genes changed significantly during regression in the LAM (SI Appendix, Fig. S4 B and C) and 75 genes in MoKC (SI Appendix, Fig. S4 B and D), suggesting that regression did not induce a profound shift in the transcriptome signature of individual macrophage subclusters. Since new macrophage subpopulations emerge during MASH that are maintained during regression, it is not feasible to perform head-to-head cluster-based comparisons with healthy macrophages (which are primarily Kupffer cells that get depleted during MASH). Gene expression analyses of all macrophage subclusters in healthy, MASH, and regression groups indicate that most genes dysregulated during MASH progression do not return to baseline levels during regression (SI Appendix, Fig. S4E). We further dissected the status of these genes across various macrophage subtypes during MASH regression (thereby identifying the regression-specific macrophage clusters enriched with these genes). Even though the profibrotic/inflammatory gene signature remained high during regression, the profibrotic/inflammatory genes were primarily expressed in the Ly6ghi macrophage subcluster (SI Appendix, Fig. S4 E, Right panel). On the other hand, the reparative genes (such as anti-fibrotic and phagocytic) were predominantly expressed in the MoKC and LAM subsets. Therefore, at 8 wk regression, the relative sizes of various macrophage subpopulations were altered more substantially (compared to MASH) than the gene expression. It is possible that the myeloid population needs longer time to return to homeostasis.

32MASH is associated with the emergence of hepatic LAM. LAMs highly express Gpnmb, Cd9, and Trem2 and form distinctive hCLS around the steatotic regions within the liver (SI Appendix, Fig. S1.1B). The functional role of LAMs and hCLS in MASH is debated. Even though the inflammation and fibrosis is exacerbated in Trem2−/− mice, the impact of Trem2 on evolution and functionality of hepatic LAMs has not been evaluated.

33Immunohistochemistry reveals that GPNMB+ LAMs form fewer hCLS in the absence of Trem2 (Fig. 5A and SI Appendix, Fig. S5A). Gpnmb expression is also reduced in the absence of Trem2 (Fig. 5B). Intriguingly, despite increased inflammatory cell infiltration in Trem2−/− MASH livers (as indicated by abundance of Cd11b+ cells) (Fig. 5C and SI Appendix, Fig. S5B), there is a paradoxical decrease in GPNMB+ LAM in these mice (Fig. 5 A and B and SI Appendix, Fig. S5A).

34GPNMB+ hCLS were also less abundant in the Trem2−/− regression mice (Fig. 5 D and E), despite having a significant residual lipid load (Foz::Trem2−/− regression vs. Foz/Foz 8 wk regression, Fig. 2 B and D) and more residual infiltrating macrophages (Fig. 5F). Overall, our data indicate that MASH progression and regression in the absence of TREM2 are associated with a lower number of LAMs/hCLS along with an increased residual collagen burden.

35To further investigate how absence of TREM2 influences the overall macrophage heterogeneity during MASH progression and regression, we performed flow cytometry analyses of primary liver macrophages from Foz/Foz and Foz::Trem2−/− mice (Fig. 5 G–I and SI Appendix, Fig. S5 C–K). We observed that the abundance of Cd9+ (Fig. 5G) and Clec4f+ (Fig. 5H) macrophages (equivalent of LAM and MoKC, respectively, that have beneficial roles) was significantly lower in Foz::Trem2−/− livers compared to their WT counterparts in both MASH and regression. On the other hand, the inflammatory Ly6c+ macrophages (Fig. 5I) were more abundant in the absence of TREM2.

36While the restorative macrophages that play key roles in fibrosis resolution have been studied in the context of chemical-induced fibrosis regression, their identity during MASH resolution is unknown. During regression of CCl4-induced liver fibrosis, a new Ly6clo restorative macrophage population emerges that play a pivotal role in orchestrating fibrosis resolution, in part through the expression of matrix-degrading metalloproteinases (MMPs). We interrogated the gene signatures of CCl4 regression-associated restorative macrophages in our MASH and regression livers. Both MoKC and LAM are Ly6clo macrophages (Fig. 3C), that emerge in MASH livers and express several of these gene signatures, including genes for ECM degradation (Mmp9, Mmp12, Mmp13, Mmp14, Plau), lipid metabolism (Cd36, Ctsb, Hexa, Lipa, Lpl, Lrp1), and phagocytosis (Axl, Gpnmb, Mertk) (Fig. 6A). To identify additional markers of restorative macrophages, we evaluated the relative expression levels of all Trem2-correlated genes that are significantly altered among the various macrophage subpopulations during MASH vs regression. The resulting heat-map reveals several genes that are specifically up-regulated in the LAM and MoKC subpopulations (SI Appendix, Fig. S6 A–D, compare MASH vs. Reg for various macrophage subpopulations).

37Interestingly, the expression of these restorative macrophage-associated genes is either sustained or further increased during MASH resolution (Fig. 6A and SI Appendix, Fig. S6 A–D). This imparts a disease-protective phenotype to the LAM and MoKCs even in livers undergoing active MASH that is further sustained as a reparative function during MASH regression when the dietary trigger is removed. Overall, our data suggest that TREM2+ LAM (along with MoKC) are the restorative macrophage populations that facilitate MASH regression.

38To confirm whether the restorative gene signatures of the MASH-associated macrophages and RAMs rely on Trem2 expression (irrespective of the disease dynamics), BMDMs from healthy WT and Trem2−/− mice were cultured under conditions mimicking tissue repair and analyzed using Nanostring nCounter (SI Appendix, Fig. S6E). Under both Il-4 (M2a) and Il10+TGFβ (M2c) stimulation, Trem2−/− BMDMs exhibited decreased expression of most genes associated with the restorative macrophage signature.

39As discussed above (Fig. 2I), the collagen gene expression was equally down-regulated in both Foz and Foz::Trem2−/− mice during regression. However, the residual fibrosis post WD→chow was significantly higher in Foz::Trem2−/− compared to the Foz livers (Fig. 2B) suggesting a defect in collagen degradation. Indeed, TREM2lo macrophages had relatively lower expression of collagenase genes (Figs. 4A and 6A). To investigate whether the presence of Trem2 can directly affect the collagenolytic activity of the macrophages, we performed collagenase assays with BMDMs isolated from WT and Trem2−/− mice (Fig. 6 B and C). Compared to WT, Trem2−/− BMDMs were less efficient in collagen degradation (Fig. 6C). Next, we evaluated whether the same defect exists in TREM2 deficient macrophages residing within MASH and regression livers. Macrophages isolated from Foz/Foz and Foz::Trem2−/− livers undergoing either active MASH (WD 12 wk) or regression (4 wk post WD to chow switch) were evaluated for their ability to degrade collagen (Fig. 6D). During both MASH progression and regression phases, TREM2+ cells had much higher collagenolytic activities compared to TREM2− (Fig. 6D). Importantly, while the collagenolytic activities of TREM2+ cells increased significantly during MASH regression (compared to MASH), the collagen degradation abilities of TREM2- cells remained low (Fig. 6D). Overall, our scRNAseq data together with in vivo and in vitro biochemical assays demonstrate that TREM2+ macrophages in both MoKC and LAM compartments restrict fibrosis development during MASH and facilitate fibrosis resorption during regression.

40The role of TREM2 in the resorption of the fibrotic scars was also evaluated in a CCl4-induced fibrosis model (Fig. 6E). Similar to our findings in MASH, CCl4 regression was associated with efficient suppression of fibrogenic genes in both WT and Trem2−/− livers (Fig. 6F). However, the resorption of existing collagen was significantly attenuated in the absence of Trem2 (Fig. 6 G and H), supporting our findings that TREM2+ macrophages are needed for efficient resolution of fibrosis.

41Macrophages play a central role in orchestrating MASH progression and resolution. While macrophage heterogeneity during MASH progression is well established, the fate of the various MASH-associated macrophage populations during regression is not known. Moreover, the key RAM subpopulation that drives MASH-fibrosis resolution remains obscure. This study defined the RAM heterogeneity in mouse livers undergoing MASH regression and compared their gene signature to the macrophages residing in active MASH livers. Furthermore, we demonstrate that TREM2+ macrophages are required for efficient MASH regression and fibrosis resorption.

42Macrophage heterogeneity in regression livers resembles the active MASH. While no new macrophage subpopulation emerged during MASH regression, the dynamics of macrophage subpopulation changed compared to active MASH. As reported, we also observed that the Kupffer cells (Clusters 0 and 8) that constitute the major macrophage population in healthy livers were lost (more than 90%) and replaced by bone marrow–derived macrophages during MASH progression (Fig. 3). Two predominant TREM2hi macrophage subpopulations (MoKC and LAM) and two TREM2lo macrophage subpopulations (Ly6chi-RM and Mac1) emerged within the active MASH livers from circulating monocytes. These monocyte-derived macrophage subpopulations continue to populate the livers undergoing MASH regression.

43The Kupffer cell population that was lost during MASH progression did not return to normal levels in regressed livers. It remains unknown whether the original yolk-sac-derived Kupffer cells ever return to original numbers and the consequence of this change in repeated liver injuries. In contrast, MASH-associated activated HSCs return to an inactive state during regression, that closely resembles the quiescent HSCs but is distinct enough to cluster as a separate population.

44While MoKC are the major macrophage subpopulation during MASH progression, they decrease during regression with a modest decrease in Trem2 expression. However, despite a significant decrease in the hepatic lipid load during regression, the hepatic LAM population does not shrink. Rather, LAMs become the major macrophage subpopulation during regression and maintains Trem2 expression. A similar phenomenon was observed in adipose tissue, where adipose tissue LAMs increased during obesity, persisted after weight loss, and continued to express Trem2. Although LAM/SAM have been widely described as a major macrophage population during MASH, their role in MASH pathology remains poorly understood. Our study provides evidence that TREM2+ LAMs and MoKCs restrain MASH and fibrosis development during active disease phases, and LAMs become the major restorative macrophage population aiding disease resolution during regression.

45TREM2+ hepatic LAMs are transcriptionally similar to adipose tissue LAMs, emerge during MASH, and are recruited to hCLS8. The role of hepatic LAM and hCLS in MASH as well as adipose tissue LAM and CLS during obesity remains unresolved. It is important to understand whether the immune cells associated with CLS (in both adipose and hepatic tissues) are protective or harmful. Ccr2−/− mice, that lack hepatic LAMs, have exacerbated MASH. We demonstrate that absence of Trem2 also reduced the emergence of LAM and hCLS, both during progression and regression, and this reduction is associated with poor disease outcome. Our data indicate that TREM2+ LAMs (the major macrophage subpopulation during regression, enriched in beneficial pathways) along with TREM2+ MoKCs play pivotal roles in MASH and fibrosis resolution. A recent study indicates that macrophage-derived Osteopontin (a hallmark of hepatic LAM) is protective in MASH. Importantly, while MASH regression led to equivalent suppression of collagen mRNA expression, the degradation of existing fibrotic scar is impaired in the absence of TREM2. The explanation is revealed in our RNAseq data as well as biochemical assays demonstrating that TREM2+ macrophages have higher anti-fibrotic gene expression as well as collagen degradation activity than TREM2− macrophages.

46Whether the absence of TREM2 influences whole-body glucose metabolism and obesity remains unclear. A recent study using the same Trem2−/− mouse strain (exon-2 deleted, Jax Strain# 027197) used in our studies revealed that absence of TREM2 does not cause whole-body metabolic abnormalities. While some studies reported increased body and liver weight in Trem2−/− mice on HFD or WD, the Winn et al. study and ours did not observe such differences. Whether these observed differences are due to the way Trem2−/− mice were generated [exon 2 deletion vs. exon 3 to 4 deletion] or due to other lab-independent factors remains unknown. Importantly, however, myeloid-cell-specific deletion of Trem2 (Trem2ΔMye, generated by crossing Trem2-Floxed mice with Lyz2-Cre) did not affect gain of liver and body weight or hepatic steatosis (similar to our observation).

47The absence of TREM2 inhibited the regression of hepatic steatosis as well as HSC inactivation. How the absence of TREM2 in macrophage modulates HSC inactivation as well as hepatic lipid metabolism during regression needs further investigation. A recent study indicates that the absence of TREM2 changes the exosome miRNA profile that alters the macrophage–hepatocyte metabolic coordination in MAFL and sepsis.

48Macrophages in MASH livers, in general, are fibrogenic and inflammatory compared to healthy livers. However, scRNAseq indicated that despite residing in the same MASH livers, TREM2hi macrophages were less inflammatory and more anti-fibrotic than TREM2lo macrophages. TREM2hi macrophages in both MASH and regression livers were enriched in protective pathways such as phagocytosis, ECM degradation, and superior lipid handling. TREM2lo macrophages, on the other hand, expressed disease-worsening pathways, such as inflammatory and cell death pathways. Comparing individual macrophage subpopulations from MASH progression vs. regression reveals that the macrophage subpopulations that were enriched in beneficial pathways during regression were already present in active MASH livers. The WD to chow switch allowed these beneficial macrophage subpopulations to decrease hepatic inflammation, resorb the fibrotic scars, and restore liver homeostasis, and TREM2 is in the hub of controlling these beneficial pathways.

49TREM2+ macrophages were shown to be efficient in efferocytosis and removal of damaged hepatocytes. We have elucidated additional mechanisms by which TREM2+ macrophages restrain MASH pathology, as well as facilitate MASH with fibrosis resolution triggered by diet. Our studies demonstrate that TREM2 attenuates inflammasome activation and tissue inflammation while promoting phagocytosis, ECM degradation, and lipid metabolism. Inflammasome activation is a hallmark of MASH livers, and we found that TREM2 restrains NLRP3 inflammasome activation. Absence of TREM2 not only leads to pronounced Caspase 1 cleavage and activation but the resolution of this pathway is also affected during MASH regression.

50Thus, our studies indicate that TREM2 agonists might be beneficial for MASH/fibrosis therapy and might be used as an adjuvant to facilitate MASH and fibrosis regression in patients undergoing lifestyle modification, weight loss, or bariatric surgery.