METTL14 downregulation in human and murine MAFLD

To assess METTL14 expression in MAFLD progression, we performed immunohistochemical analyses in tissue samples from patients with healthy liver and MAFLD. These studies revealed a robust downregulation of METTL14 in MAFLD liver tissue, characterized by macrovesicular steatosis (Fig. 1a, b). Next, to determine the clinical relevance of METTL14 in the pathogenesis of MAFLD, we investigated whether there were histological and transcriptomic differences between different METTL14 levels in the liver with the data from the GTEx project.8 We established groups of patients with high and low METTL14 expression using median METTL14 levels in 161 individuals. Patients with mild, moderate, or severe steatosis were more observed in the low-METTL14 group (Supplementary Fig. 1a, b). METTL14 expression was lowered in the group of patients with mild, moderate, or severe steatosis compared with patients with minimal or no steatosis (Supplementary Fig. 1c, d). Prompted by these findings, we expanded our analysis of METTL14 expression in multiple murine MAFLD models. First, we fed mice with a 60% high-fat diet (HFD) for 24 weeks (Fig. 1c), which is an excellent tool for studying the incremental steps of MAFLD progression.9 In line with previous reports, HFD-treated mice developed hepatic steatosis with ballooned hepatocytes. Immunohistochemical analysis confirmed a strong downregulation of METTL14 in the livers of MAFLD mice (Fig. 1d, e and Supplementary Fig. 1e), which is consistent with our findings in human MAFLD. We then generated another two diet-induced MAFLD models, including mice fed with a Western diet (WD) and an Amylin-Liver Nash (AMLN) diet (Supplementary Fig. 1f–h and Supplementary Fig. 1i–k).10,11 Interestingly, METTL14 expression was similarly decreased in the livers of WD-fed and AMLN-fed mice compared to their respective controls (Supplementary Fig. 1h, k).

Fig. 1

METTL14 is downregulated in the livers of mice with HFD and in human MAFLD. a Sections of human healthy liver (n = 5) and metabolic fat liver disease (MAFLD) (n = 5) were analyzed by immunohistochemical analysis for METTL14. Representative images are shown. b Relative quantitative analysis of METTL14 in normal human liver and metabolic fat liver tissues (n = 5). c Schematic diagram of the dietary feeding scheme. Six-week-old C57BL/6 wild-type mice were fed either a control diet (CON) or a 60% high-fat diet (HFD) for 24 weeks. d Mice were sacrificed, and liver sections were analyzed by immunohistochemical analysis for METTL14 expression (n = 3). e Relative quantitative analysis of METTL14 in liver tissues of mice fed the HFD and control diet (n = 3). f Comparison of the liver-to-body weight ratio of 8-week-old wild-type (WT, n = 10), heterozygous knockout mice (KO+/−, n = 7) and homozygous knockout mice (KO, n = 9). g, h Serum ALT (g) and AST (h) in wild-type (WT, n = 5), heterozygous knockout mice (KO+/−, n = 5) and homozygous knockout mice (KO, n = 3). i Schematic diagram of 6-week-old WT, KO+/− and KO mice fed a HFD for 16 weeks. KO mice fed a HFD for 4 weeks were randomly injected with AAV-8 overexpressing METTL14 (AAV-OV) or control vector (AAV-NC) through the caudal vein, while WT mice and KO+/− mice fed a HFD for 4 weeks were injected with AAV-NC. j–o Comparison of liver weight (j) and liver to body ratio (k), serum ALT (l) and AST (m), serum triglycerides (TG, n) and cholesterol (CHO, o) of WT (n = 4), KO+/− (n = 4), KO injected with empty AAV-8 vector (KO-NC, n = 5) and KO injected with AAV-8 vector overexpressing METTL14 (KO-OV, n = 5). p H&E (left) and Oil Red O (right) staining of liver sections from HFD-treated WT (n = 4), KO+/− (n = 4), KO-NC (n = 5) and KO-OV (n = 5) mice are shown. Data are represented as mean ± SEM. NS not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. The mouse image was created with BioRender.com

To study whether downregulation of METTL14 was indeed causally involved in the initiation and progression of MAFLD, we generated genetically modified mice with hepatocyte-specific depletion (Albumin-Cre) of METTL14 (heterozygous knockout: KO+/−; homozygous knockout: KO; wild type: WT). We generated a Mettl14 loxP floxed allele (Supplementary Fig. 2a, b) and verified the reduction of Mettl14 mRNA and protein abundance (Supplementary Fig. 2c–f). The liver weight and body weight of KO mice were markedly reduced at 3 months compared to those of KO+/− and WT mice (Supplementary Fig. 2g, h); however, the liver to body weight ratio increased (Fig. 1f). Histology assessments revealed hepatocyte swelling, disordered structures of hepatic plates as well as infiltration of lipid droplets and inflammatory cells in the livers of KO mice (Supplementary Fig. 2i). These results indicated that METTL14 played a vital role in body growth and liver tissue development to some extent. Additionally, ALT and AST levels were significantly increased in KO+/− and KO mice at 3 months, and the upregulation of ALT and AST levels in KO mice was much more obvious (Fig. 1g, h), revealing that hepatocyte-specific depletion of METTL14 leads to severe liver injuries.

Prompted by these findings, we utilized 6-week-old WT and Mettl14 knockout (KO+/−, KO) mice under HFD treatment for 16 weeks to further study the role of METTL14 in mediating MAFLD. WT and KO+/− mice were treated with adeno-associated virus-8 control vector (AAV-NC) via tail vein injection after four weeks of feeding, while KO mice were randomly injected with AAV-NC or AAV overexpressing METTL14 (AAV-OV) through the caudal vein and then fed a HFD for another 12 weeks (Fig. 1i). Compared with WT or KO+/− mice injected with AAV-NC (WT-NC or KO+/−-NC), KO mice injected with AAV-NC (KO-NC) were significantly aggravated from 16-week HFD-induced hepatic steatosis, as shown by increased liver weights, the ratio of liver to body weight, ALT, AST and blood lipid levels, such as triglyceride (TG) and cholesterol (CHO) (Fig. 1j–o). As expected, treatment of KO-NC mice with HFD resulted in more progressive steatosis, ballooning, and pronounced lobular inflammation compared with those of WT-NC and KO+/−-NC mice, which was also verified by MASH Activity Score (MAS) assessment and Oil red O staining (Fig. 1p and Supplementary Fig. 2j, k). In contrast, KO mice administered AAV-OV (KO-OV, AAV overexpressing METTL14) exhibited lower liver weight, the liver-to-body weight ratio, ALT and AST levels than the KO-NC group (Fig. 1j–m), as well as lower lipid infiltration in histological analysis of livers (Fig. 1p and Supplementary Fig. 2j, k), revealing that METTL14 restoration reduced lipid accumulation and abolished the progression of MAFLD. Collectively, these findings substantiated the causal role of METTL14 in the progression of MAFLD and hinted at potential therapeutic strategies.

GLS2 was downregulated in both KO mice and MAFLD mice and consequently promoted the oxidative stress microenvironment

To obtain a comprehensive explanation for the roles of METTL14 in MAFLD progression, we performed proteomic analysis of liver tissues from the Mettl14 KO (n = 6) and WT (n = 4) mice, as well as HFD-fed (n = 3) and control diet (CON)-fed (n = 3) mice (Fig. 2a and Supplementary Tables 1, 2), and then carried out differentially expressed protein (DEP) conjoint analysis among four groups of mice (Fig. 2b). Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) analyses employing these DEPs firmly revealed that METTL14 deficiency significantly impacted fatty acid metabolism and biosynthesis process (Fig. 2c). Subsequently, we analyzed gene expression patterns associated with lipid homeostasis in the liver after METTL14 deficiency employing lipid metabolic markers such as FA oxidation markers (CPT1A, ACOX1, ACAT2), de novo lipogenesis markers (FASN, SCD1, ACACA), TAG synthesis markers (GPAT3, DGAT1, MGAT1), and lipid uptake markers (CD36, LDLR, FABP7). Depletion of METTL14 was discovered to inhibit fatty acid oxidation, increase lipid synthesis and uptake, and consequently hasten lipid accumulation (Fig. 2d). To further investigate the effects of METTL14 deficiency on the metabolic microenvironment in mice, we performed a full-spectrum metabolomics analysis of the liver tissues from KO and WT mice (Supplementary Table 3). It was discovered that D-glutamine and D-glutamate metabolism were most significantly altered in KO mice among the metabolic pathways (Fig. 2e). Glutamine metabolism is principally regulated by two enzymes: glutaminase (GLS) and glutamine synthetase (GS). Glutaminase isoforms are tissue-specific: kidney-type glutaminase (GLS1) and liver-type glutaminase (GLS2).12 The proteomic analysis revealed that GS was marginally elevated in the KO mice (FC = 1.37), whereas it remained relatively unchanged in the HFD-fed mice (FC = 0.95) (Supplementary Fig. 3a, b). In addition, the expression of GS did not alter remarkably in the patients ranging from no steatosis, minimal, mild, moderate to severe steatosis (Supplementary Fig. 3c, d). However, the proteomic analysis also revealed that GLS2 was downregulated in both KO mice and HFD-fed mice (Fig. 2b). The expression of GLS2 protein was additionally observed, revealing a significant downregulation of GLS2 in the liver tissues of HFD mice and KO mice as compared to their respective controls (Fig. 2f–i), which was also in line with previous studies.13 It was also found a substantial downregulation of GLS2 in mice fed with a Western diet (Supplementary Fig. 1h). Nevertheless, GLS2 exhibited a decreased trend in the AMLN group, but there was no significant statistical disparity (Supplementary Fig. 1k).

Fig. 2

GLS2 was downregulated in both KO mice and HFD mice and consequently promoted the oxidative stress microenvironment. a Heat maps presenting differentially expressed proteins (DEPs) in HFD vs control diet-fed mice (n = 3) and KO vs WT mice (nKO= 4, nWT = 6), respectively. b Nine quadrant diagrams presenting the overlapping upregulated proteins and/or downregulated proteins among HFD and KO mice from proteomics sequencing (the dashed lines indicating the fold changes are at 1.5). c GO and KEGG analyses of differentially expressed proteins indicating lipid metabolic alterations among KO and WT mice. d The levels of specific markers related to lipid metabolism including fatty acid oxidation (CPT1A, ACOX1, ACAT2), de novo lipogenesis (FASN, SCD1, ACACA), TAG synthesis (GPAT3, DGAT1, MGAT1) and lipid uptake (CD36, FABP7, LDLR). e Overview of the pathway analysis based on metabolite alterations in KO mice from metabolomics analysis. f Western blot indicating the levels of GLS2 in 8-week-old WT and KO mice (top) and HFD-fed and control diet-fed mice (bottom). Representative images are shown (n = 4). g Relative quantitative analysis showing the levels of GLS2 in WT and KO mice (top) and HFD and control diet mice (bottom). h Immunohistochemical staining analysis of GLS2 in WT and KO mice, HFD-fed mice and control diet-fed mice (n = 4, scar bar = 50 μm). i Relative quantitative analysis of GLS2 in WT and KO mice (top) and HFD-fed and control diet-fed mice (bottom) (n = 4 respectively). j DCFH-DA was used to display the levels of intracellular ROS in primary cultured hepatocytes isolated from WT and KO mice. Representative images are shown(scar bar = 500 μm). k TAC levels showing total antioxidant capacity using ABTS methods in KO (n = 4) and WT (n = 5) mice. l ELISA showing the levels of 8-OHdG in KO and WT mice (n = 6 respectively). Data are represented as mean ± SEM. NS not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

It has been reported that limited expression of GLS2 in hepatocytes causes a decrease in antioxidant capacity and an increase in reactive oxygen species (ROS) levels in cells.14,15,16 We subsequently assessed the antioxidant capacity and ROS levels in liver tissues of KO mice. Compared to WT mice, the level of ROS in the hepatocytes of KO mice increased (Fig. 2j), and the antioxidant capacity decreased substantially (Fig. 2k). We also conducted an analysis on the levels of 8-hydroxy-2 deoxyguanosine (8-OHdG), a widely recognized biomarker for assessing oxidative stress and DNA damage.17 As anticipated, our findings revealed an upregulation of 8-OHdG in the liver tissues of KO mice compared with WT mice (Fig. 2l).

To determine whether GLS2 downregulation in Mettl14-deficient mice led to liver injury and oxidative stress milieu, we constructed and validated an GLS2-upregulation plasmid vector (Supplementary Fig. 4a–c), then generated serotype 8 AAVs overexpressing GLS2 and performed an in vivo rescue assay on KO mice (Supplementary Fig. 4d). We first validated the overexpression of GLS2 in the liver tissues of mice overexpressing GLS2 (KO-OE) and mice expressing control vector (KO-CTL) (Supplementary Fig. 4e, f). Although the morphology of the liver and the liver to body weight ratio did not differ significantly between KO-OE and KO-CTL mice (Supplementary Fig. 4g, h), KO-OE mice had substantially lower serum ALT levels (Supplementary Fig. 4i). Rescued GLS2 expression remarkably reduced the ROS levels and promoted the antioxidant capacity in the KO mice (Supplementary Fig. 4j, k). More importantly, fibrotic markers (Col1a1, Acta2, and Mmp2) were significantly reduced in GLS2 upregulated KO mice (Supplementary Fig. 4l), as confirmed by Sirius red staining and Masson staining (Supplementary Fig. 4m). Furthermore, immunofluorescence analysis of α-SMA revealed lower activation of hepatic stellate cells (HSCs) in the KO-OE mice (Supplementary Fig. 4m), indicating that GLS2 restoration could alleviate liver injury and fibrosis in Mettl14 deficient mice.

These results indicated that the reduced expression of METTL14/GLS2 protein might contribute to the development of an oxidative stress milieu, hence facilitating the progression of MAFLD.

METTL14 regulates GLS2 expression in an m6A-dependent manner via YTHDF1

To clarify the relationship between METTL14 and GLS2 expression, we first generated overexpression or interference with METTL14 expression in HUH7 and Hep3B cells in vitro, and discovered that there was no significant change in GLS2 mRNA levels, but GLS2 protein levels were consistently upregulated or downregulated (Fig. 3a and Supplementary Fig. 5a–d). To evaluate whether METTL14 could regulate the translation and degradation of GLS2 protein, we further treated cells with cycloheximide (CHX) to block protein translation or MG132 to inhibit proteasome activity. It was indicated that METTL14 interference or overexpression had no discernible effect on GLS2 protein levels in CHX-treated HUH7 cells (Supplementary Fig. 5e, f). MG132 could restore the level of GLS2 protein, however, the difference remained after interference or overexpression with METTL14 in HUH7 cells (Supplementary Fig. 5g, h). All these results indicated that METTL14 might regulate the protein translation rather than protein stability of GLS2.

Fig. 3

METTL14 regulates GLS2 expression in an m6A-dependent manner via YTHDF1. a Western blot showing METTL14 and GLS2 expression in HUH7 (left) and Hep3B (right) cells infected with METTL14 overexpression (OE) or shRNA lentivirus vector (SH). b, c Relative m6A enrichment in Gls2 mRNA in liver tissues of WT and KO mice (b), HFD and control diet mice (c) by MeRIP-qPCR. n = 4. d The image showing the position and sequences of three potential m6A-binding sites with very high confidence of GLS2 mRNA using online SRAMP database (https://www.cuilab.cn/sramp). e Relative luciferase activity of three mutant plasmids and their wild-type plasmids of GLS2 in HUH7 (left) and Hep3B (right) cells with knockdown of METTL14. f Relative luciferase activity of three mutant plasmids and their wild-type plasmids of GLS2 in HUH7 (left) and Hep3B (right) cells with overexpression of METTL14. g Western blot (left) and relative quantitative analysis (right) showing METTL14 and GLS2 expression in HUH7 cells infected with YTHDF1 overexpression or shRNA lentivirus vector. h Western blot (left) and relative quantitative analysis (right) showing METTL14 and GLS2 expression in Hep3B cells infected with YTHDF1 overexpression or shRNA lentivirus vector. i Relative luciferase activity of three mutant plasmids and their wild-type plasmids of GLS2 in HUH7 (left) and Hep3B (right) cells with knockdown of YTHDF1. j Relative luciferase activity of three mutant plasmids and their wild-type plasmids of GLS2 in HUH7 (left) and Hep3B (right) cells with overexpression of YTHDF1. k Western blot showing YTHDF1 and GLS2 expression in HUH7 (left) and Hep3B (right) cells transfected with YTHDF1 overexpression (oe-YTHDF1), YTHDF1 mutation (K395A & Y397A, Mut) and vector plasmid (Vector) respectively. l Relative luciferase activity of HUH7 (top) and Hep3B (bottom) cells transfected with YTHDF1 overexpression, YTHDF1 mutant (K395A and Y397A) and vector plasmids, respectively. m Western blot showing METTL14, YTHDF1 and GLS2 expression in HUH7 (left) cells and Hep3B cells (right) infected with YTHDF1 overexpression (oe-YTHDF1) and/or METTL14 shRNA (sh-METTL14) lentivirus vector as well as control vectors. Data are represented as mean ± SEM. NS not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

As METTL14 is a pivotal m6A methyltransferase, we then detected the m6A modification level of Gls2 mRNA in liver tissues of KO mice and HFD-fed mice, and found that the m6A modification level was significantly reduced in KO mice and HFD-fed mice as compared to their respective controls (Fig. 3b, c), suggesting that METTL14 might regulate GLS2 expression in an m6A-dependent manner. We then predicted the potential m6A binding sites of GLS2 mRNA and identified the top three sites using the SRAMP database (Supplementary Table 4).18 To explore the potential roles of m6A modification on GLS2 expression, firstly, we constructed three wild-type fragment reporters containing the above m6A sites named GLS2 WT1/2/3, or mutants named MT1/2/3 (A/GGAC to A/GGCC, Fig. 3d) after the firefly luciferase reporter gene. The dual-luciferase assay showed the luciferase activities were decreased in HUH7 and Hep3B cells interfering METTL14, but were restored after transfected with GLS2 m6A mutants (Fig. 3e). Conversely, the luciferase activities were elevated in METTL14-overexpressed HUH7 and Hep3B cells, but were attenuated after transfected with GLS2 MT1/2/3 (Fig. 3f).

It was well-acknowledged that YTHDF1 could recognize m6A methylated mRNA and promote the translation of targeted mRNA.19 In order to explore the potential involvement of YTHDF1 in the regulation of m6A modified GLS2 expression mediated by METTL14, we further detected the expression of GLS2 in HUH7 and Hep3B cells with YTHDF1 interference or overexpression. It was shown that GLS2 was consistently upregulated or downregulated in HUH7 and Hep3B cells with YTHDF1 overexpression or interference (Fig. 3g, h). The dual-luciferase assay showed the luciferase activities were decreased in YTHDF1-interfered HUH7 and Hep3B cells, but were restored after transfected with GLS2 m6A mutants (Fig. 3i). Conversely, the luciferase activities were increased in HUH7 and Hep3B cells overexpressing YTHDF1, but were attenuated after transfected with GLS2 m6A mutants (Fig. 3j). YTHDF1 is known to bind m6A sites via its m6A-binding pockets in the YTH domain; mutations in K395 and Y397 could eliminate the mRNA-binding capacity of YTHDF1.20,21 We then constructed YTHDF1-overexpressed plasmid (oe-YTHDF1-wt) as well as its mutant by introducing two-point mutations K395A and Y397A in YTH domain of YTHDF1 (oe-YTHDF1-mut) and transfected them into HUH7 and Hep3B cells, respectively (Supplementary Fig. 5i). Subsequently, RIP-PCR analysis by using an anti-YTHDF1 antibody revealed that GLS2 mRNA was immunoprecipitated effectively in HUH7 and Hep3B cells transfected with YTHDF1-wt, but the interaction between YTHDF1 mutants and GLS2 mRNA was drastically decreased (Supplementary Fig. 5j). GLS2 protein levels changed in the same way (Fig. 3k and Supplementary Fig. 5k). Moreover, the luciferase activities were increased in HUH7 and Hep3B cells co-transfected with GLS2 WT 1/2/3 plasmids and oe-YTHDF1-wt construct, but were attenuated when co-transfected with oe-YTHDF1-mut construct (Fig. 3l), suggesting the m6A-binding pocket were crucial for YTHDF1 to bind GLS2 mRNA. Furthermore, GLS2 expression was rescued when METTL14-interfered HUH7 and Hep3B cells were treated with YTHDF1 overexpression (Fig. 3m and Supplementary Fig. 5l).

Taken together, these results suggested that METTL14 regulated the protein translation of GLS2 mediated by the m6A reader YTHDF1 in an m6A-dependent manner.

Hepatocyte Mettl14 deficiency resulted in an increase in Cx3cr1 +Ccr2 + Mo-macs in liver tissue

Previous studies have shown that GLS2 functions as an antioxidant by catalyzing the hydrolysis of glutamine to produce glutamic acid and increasing reduced glutathione (GSH) levels.14,15 The low expression of GLS2 in hepatocytes leads to the accumulation of reactive oxygen species (ROS), which may lead to the release of damage-associated molecular pattern molecules (DAMPs).16 DAMPs are cell-derived molecules that respond to trauma, local ischemia, and tissue damage in the absence or presence of pathogenic infections, initiating and maintaining immunity.22 Macrophages are believed to play an important role in the initiation and spread of liver inflammation, as well as in the regulation of liver fibrosis.23 Considering that the oxidative stress microenvironment of liver tissue is the basic feature of MAFLD-related liver tissue and the key factor in fostering the progression of inflammation, non-parenchymal cells of one KO and one WT mouse liver were used for single-cell RNA sequencing (scRNA-seq).24,25 A total of 14374 individual non-parenchymal cells were included in the analysis at 18548 gene feature levels after quality control. Given the key role of macrophages in the progression of MAFLD, we focused on the composition changes of myeloid cells between KO and WT mice.23,26 To determine the expression features of various cell types, all the myeloid cells were classified into fifteen clusters and five specific cell types (Fig. 4a, b). We then displayed the marker genes and differentially expressed genes (DEGs) among each cell cluster from different cell types (Fig. 4c, d). The results of cell counting showed that KCs decreased significantly in KO mice (M01, M03, M04 and M07), while Mo-macs instead increased significantly (M05, M06 and M09) (Fig. 4e). M05 cluster comprises 12.6% of myeloid cells in KO mice and 5.6% in WT mice. We then performed differentially expressed gene analysis and presented the top 20 genes among M05 cluster (Supplementary Fig. 6a). GO and KEGG analysis further revealed that this cluster of Mo-macs was associated with ribosome activity and translation process (Supplementary Fig. 6b). Interestingly, we discovered that M06 and M09 clusters are characterized by high expression of Cx3cr1 and Ccr2 (Fig. 4c). To confirm this subset of cells, we labeled CX3CR1 and CCR2 double-positive cells in the liver tissues of KO and WT mice and discovered that CX3CR1+CCR2+ cells were markedly increased in the liver tissues of KO mice (Supplementary Fig. 7a). More importantly, CX3CR1+CCR2+ cells were consistently accumulated in the liver tissues of HFD mice when compared to those of control diet mice (Supplementary Fig. 7b), suggesting that these cells might play an important role in the progression of MAFLD.27,28,29 We also noticed that M09 and M15 Mo-macs also remarkably expressed specific markers of lipid-associated macrophages (LAMs) or scar-associated macrophages (SAMs) including Trem2, Gpnmb, Cd9, Cd63, Cx3cr1, Ccr2, and Fabp5, which have been documented to play pivotal roles in the progression of MAFLD.2,28,30,31,32,33 It was possible that M09 and M15 were a typical cluster of either SAMs or LAMs.

Fig. 4

Increased abundance of Cx3cr1+Ccr2+ macrophages in the liver of the KO group. a UMAP plot displaying the distributions of 6 Kupffer cell clusters, 4 Mo-mac clusters, 3 dendritic cell clusters, 1 neutrophil cluster and 1 other cluster. b The percentage of cell counts, cell clusters and cell types in the KO and WT groups. c Violin plots showing marker genes and markers of lipid associated macrophages (LAMs) or scar associated macrophages (SAMs) across each cluster. d DEGs for Kupffer cells, monocyte-derived macrophages, dendritic cells and neutrophil clusters. e Bar plot showing different proportions of each cluster in the KO and WT groups

To explore the roles of increased Cx3cr1+Ccr2+ Mo-macs in Mettl14 deficiency and MAFLD progression, we further analyzed the gene expression patterns of Cx3cr1+Ccr2+ Mo-macs. We displayed the top 20 differentially upregulated genes of Cx3cr1+Ccr2+ Mo-macs (Fig. 5a, b), which imply that these macrophages have the capacity to promote inflammatory progression and subsequent fibrosis. For example, pharmacological inhibition of CCR2+ monocyte recruitment efficiently ameliorates hepatic inflammation and fibrosis in patients with metabolic steatohepatitis.34 S100A4 derived from macrophages in the liver promotes liver fibrosis through hepatic stellate cell activation.35 Pathway analysis further revealed that Cx3cr1+Ccr2+ Mo-macs were enriched for inflammatory pathways and inflammation-activated key genes (Fig. 5c, d). When further analyzing the characteristic expression genes of M06 and M09 clusters, it was discovered that M09 highly expressed M1 macrophage-related markers (Il1b, Tlr2 and Cd86), whereas M06 expressed neither M1 nor M2 macrophage markers (Mrc1, Arg1 and Cd163). Strikingly, M06 expressed high levels of S100a4, S100a6, S100a10 and S100a11 (Fig. 5e, f). These findings imply that despite the fact that both the M06 and M09 subsets are Cx3cr1+Ccr2+ Mo-macs, their contributions to the progression of inflammation may be distinct.

Fig. 5

Cx3cr1+Ccr2+ Mo-macs in the late stage of the developmental trajectory highly expressed S100 family protein genes. a Volcano plot comparing DEGs for Cx3cr1+Ccr2+ Mo-macs. b Feature plots showing the top 20 DE genes. c KEGG pathway enrichment analysis of upregulated DE genes in Cx3cr1+Ccr2+ Mo-macs. d Dot plot presenting the expression strength of key KEGG pathway genes among Cx3cr1+Ccr2+, Cx3cr1-Ccr2-and Ccr2+ Mo-macs and Kupffer cells. e Violin plots displaying the expression levels of polarization state marker genes and S100a family protein genes. f Volcano plot comparing DEGs between M06 and M09. g Heatmap displaying the expression of selected marker genes in Mo-macs arranged along the pseudotime trajectory. h Scatter plots and fitting curves presenting the expression trend of S100a4, S100a6, S100a10 and S100a11

Furthermore, we visualized the transcriptional profile of Mo-macs and mapped them along pseudotime trajectories (Supplementary Fig. 8a). Our data implied that Cx3cr1+Ccr2+ macrophages were in the late stage of the developmental trajectory (Supplementary Fig. 8b). Moreover, the cells from KO mice were located in the pseudotime trajectories later than those from WT mice (Supplementary Fig. 8b). To gain insights into the differentiation process of Cx3cr1+Ccr2+ Mo-macs, we generated a branched heatmap to decipher the gene expression patterns of M06 and M09 on the dynamics of DEGs (Fig. 5g). These genes were then categorized into six groups according to their characterized patterns, and multiple S100a family members (S100a4, S100a6, S100a10 and S100a11) were evidently present at the end of the pseudotime trajectory among the M06 cluster (Fig. 5g, h). S100A proteins are calcium-binding proteins that are commonly dysregulated in various diseases, including MAFLD.36,37 HFD increased liver S100A11 expression, which may interact with HDAC6, disrupt its binding to FoxO1, release or enhance FoxO1 acetylation, and thus accelerate lipid accumulation and hepatic steatosis.37 A recent study showed that macrophage-derived S100A4 can promote liver fibrosis by activating HSCs in a CCl4-induced liver fibrosis model.35 Therefore, we had reason to believe that the high expression of S100A proteins in Cx3cr1+Ccr2+ Mo-macs might play a crucial role in activating HSCs in the progression of MAFLD.

S100A4 expression was activated by CX3CR1/MyD