Human liver samples were collected from 39 patients with MASH (39 frozen tissues for mRNA expression detection and 6 paraffin-embedded slides for protein expression detection), and 31 patients without hepatic steatosis (patients who presented with hepatic hemangioma at the First Affiliated Hospital of Nanjing Medical University) were used as controls (Supplementary Table 1). Individuals in the control group had no history of diabetes, alcohol use, viral hepatitis, or other liver diseases. All liver specimens were assessed independently by two experienced pathologists blinded to the NAFLD activity score (NAS) clinical data, which is defined as the sum of steatosis, inflammation, and hepatocyte ballooning scores. Patients with an NAS ≥5 were considered likely to have MASH. The clinical characteristics of the patients with MASH are listed in Supplementary Table 1. The study was approved by the Institutional Ethics Committee of the First Affiliated Hospital of Nanjing Medical University (approval number: 2020-SRFA-288). Informed consent for tissue analysis was obtained before liver biopsy or surgery. All experiments were performed in accordance with government policies and the Declaration of Helsinki and Istanbul.

Wild-type (WT), FloxP-Atg16l1 (Atg16l1^fl/fl), Lyz2-Cre Atg16l1-knockout (Atg16l1^ΔMϕ), Lyz2-Cre Atg16l1 and Lyz2-Cre Tmem173-double knockout (Atg16l1^ΔMϕ Tmem173^ΔMϕ), and myeloid-specific Atg16l1-overexpressing-knockin (Atg16l1^OE) male mice, 6–8 weeks old, on a C57BL/6 background were used in the experiments (Supplementary Table 2). Mice were randomly assigned to receive a normal chow diet (NCD), a high-fat and high-cholesterol diet (HFHCD) (Research Diets, Inc., New Brunswick, CA, USA) for 16 weeks (n=6–8 per group), or a methionine- and choline-deficient (MCD) diet (Research Diets, Inc.) for 6 weeks (n=6–8 per group). Male wild-type C57BL/6 mice (6–8 weeks old) were administered the ATG16L1 enhancer peretinoin (0.03% for 16 consecutive weeks or 6 consecutive weeks) in the diet [17], or vehicle (saline) for 16 weeks of HFHCD feeding or for 6 weeks of MCD feeding (n=6–8 per group). All the mice were housed under specific pathogen-free, ventilated, thermostatic conditions with a 12-hours light/dark cycle at 24°C. Food intake was measured over one week using racks in a regular cage that were weighed every day, and daily food intake was calculated as kcal/day within the indicated period. The mice were fasted at the end of the experiments. Finally, serum and tissues were harvested as described in our previous study [18]. The livers and adipose tissues were rapidly excised and weighed. All animal studies were performed according to the guidelines of the Institutional Animal Use and the Animal Experimentation Ethics Committee of The First Affiliated Hospital of Nanjing Medical University.

Primary hepatocytes were isolated from male Atg16l1^fl/fl or WT mice as previously described [19]. The liver macrophage pellet was resuspended and cultured. Cells were treated with a mixture of palmitic acid (PA; 0.5 mM; SigmaAldrich) and oleic acid (OA; 1.0 mM; Sigma-Aldrich) coupled with bovine serum albumin (BSA; Sigma-Aldrich) for the indicated periods. BSA was used as a vehicle control. Bone marrow cells were isolated from the femurs and tibias of male Atg16l1^ΔMϕ, Atg16l1^OE, and Atg16l1^fl/fl mice. The bone marrow was flushed with DMEM loaded into a 1 mL syringe. Bone marrow cells were cultured at a concentration of 3×10⁶/mL in DMEM supplemented with 10% FBS and 20% L929 conditioned medium to differentiate them into bone marrow-derived macrophages (BMDMs). On day 7, all adherent cells differentiated into mature macrophages.

For coculture studies, primary hepatocytes from male Atg16l1^fl/fl mice and LPS-primed (10 ng/mL of LPS for 4 hours) macrophages [20] from the bone marrow of male Atg16l1^ΔMϕ, Atg16l1^OE, or Atg16l1^fl/fl mice were seeded in a coculture chamber (Corning Inc., Corning, NY, USA). The coculture system was supplemented with palmitic acid and oleic acid media. Mouse TNF-α-neutralizing mAb (10 ng/mL) (CST, Danvers, MA, USA), MR16-1 (12.5 ng/mL) (Chugai Pharmaceutical, Shizuoka, Japan), IL-1RA (12.5 ng/mL) (MedChemExpress, Princeton, NJ, USA), TNF-α (10 ng/ mL), IL-6 (12.5 ng/mL), or IL-1β (12.5 ng/mL) (Novus Biologicals, Littleton, CO, USA) was added to the coculture system for 24 hours, and the hepatocytes were subsequently analyzed for fat accumulation and inflammatory response indices. LPS-primed (10 ng/mL LPS for 4 hours) BMDMs were incubated with hepatocyte-CM for 48 hours and subjected to RNA sequencing analysis. Primary mouse HSCs were isolated as previously described [21]. Briefly, primary hepatic stellate cells (HSCs) were isolated from the livers of mice. After portal vein intubation, the livers were perfused in situ with Ca²⁺-free Hank’s balanced saline solution (HBSS) at 37°C for 15 minutes and subsequently perfused with the solution containing 0.05% collagenase and Ca²⁺ for 15 minutes at a flow rate of 10 mL/min. The perfused livers were minced, filtered through a 70 M cell strainer (BD Bioscience), and centrifuged at 50 × g for 3 minutes. The supernatant was further centrifuged at 500 × g for 10 minutes, resuspended in Ficoll plus Percoll (1:10, GE Healthcare), and centrifuged at 1,400 × g for 17 minutes. The HSCs were collected from the interface. Primary mouse HSCs were incubated with BMDM-CM in the absence or presence of TGF-β1 (8 ng/mL for 24 hours) and analyzed for HSC activation status.

H&E-stained liver sections were assessed by two experienced liver pathologists who were blinded to experimental grouping, and each liver specimen was evaluated for the presence or absence of MASH according to histopathology and for the NAFLD activity score (NAS), defined as the sum of steatosis, inflammation, and hepatocyte ballooning scores. Patients with an NAS of ≥5 were considered likely to have MASH [3,22]. For oil red O staining, primary hepatocytes and frozen liver sections (10 μm) were rinsed with 60% isopropanol and stained with oil red O solution (Sigma‒Aldrich) for 15 minutes. Afterward, the sections were rinsed again with 60% isopropanol, and the nuclei were stained with hematoxylin before microscopic analysis.

The serum levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were determined using an automated chemical analyzer (Olympus, Tokyo, Japan). Hepatic triglyceride (TG) levels were determined with a Wako E-test triglyceride kit (Wako Pure Chemical Industries, Osaka, Japan). Cellular TG synthesis was determined using standard methods [23]. Briefly, cellular lipids were extracted with hexane/isopropanol (3:2). Lipids were then dried with nitrogen gas, redissolved in chloroform, and resolved by thin-layer chromatography using successive solvent systems containing chloroform, acetone, methanol, acetic acid, and water at a volumetric ratio of 10:4:2:2:1, as well as hexane, methanol, and acetic acid at a volumetric ratio of 80:20:1. Phosphor images were obtained with a Storm Gel and Blot Imaging System (GE Healthcare).

The rates of fatty acid β-oxidation were determined using a modification of a previously used method [24] that measured the rate of carbon dioxide production. Carbon dioxide trapped for 1 hour at 37°C from stimulated cells was released onto filter paper soaked in 100 mM sodium hydroxide. The rate of β-oxidation was calculated as the amount of trapped carbon dioxide in relative units produced per mg protein per hour.

Murine liver and adipose tissue immunohistochemistry was performed to detect F4/80 expression in paraffin-embedded liver biopsies using an anti-F4/80 antibody (#70076, CST). Immunohistochemistry of the murine liver was performed to detect GSDMD expression in paraffin-embedded human liver biopsies using an anti-GSDMD antibody (#AF4012, Affinity). In brief, liver sections were preblocked for 10 minutes after deparaffinization. The slides were then heated in an autoclave with sodium citrate for antigen retrieval, covered in 1% hydrogen peroxide to eliminate endogenous peroxidase activity, and blocked with 2% goat serum. The slides were then incubated with primary antibodies at 4°C overnight. Biotinylated goat anti-rabbit IgG (Vector, Burlingame, CA, USA) was used as the secondary antibody, followed by incubation with immunoperoxidase (ABC Kit; Vector) according to the manufacturer’s protocol. ATG16L1 and CD68 expression in human liver tissues were determined by immunofluorescence with an anti-rabbit ATG16L1 mAb (#8089, CST) and an anti-mouse CD68 mAb (ab955, Abcam), followed by incubation with secondary goat anti-rabbit IgG (ab150088, Abcam) or goat anti-mouse IgG (ab150117, Abcam). The cells were fixed using 3% paraformaldehyde and subsequently blocked and incubated with primary and corresponding Cy5- and Texas Red-conjugated secondary antibodies. Lipid droplets (LDs) were stained by incubating cells with BODIPY 493/503 (D2191, Invitrogen) for 30 minutes. The cells were then fixed and processed for immunofluorescence as previously described [25]. STING, α-SMA, LAMP1, LC3B, and GM130 expression in cells was determined by immunofluorescence using an anti-rabbit STING pAb (PA5-116052, Invitrogen), an anti-rabbit α-SMA mAb (#19245, CST), an anti-rabbit LAMP1 mAb (#99437, CST), an anti-rabbit LC3B mAb (#43566, CST), and an anti-mouse GM130 mAb (sc-55591, Santa Cruz Biotechnology), followed by incubation with secondary goat anti-rabbit IgG (ab150088, Abcam) or anti-mouse IgG (ab150117, Abcam). DAPI was used to stain the nuclei. The slides were washed twice with PBS and observed via confocal microscopy (Zeiss, Oberkochen, Germany) according to the manufacturer’s protocol.

A c-Jun adenoviral vector was constructed by GenePharma (Shanghai, China) to overexpress c-Jun in primary macrophages isolated from bone marrow. Targeted cells (2×10⁵) were infected with 1×10⁶ adenovirus-transducing units in the presence of 1 mg/mL polybrene. Total RNA or proteins were prepared 48 hours after transfection. Negative control adenoviral plasmid vectors were generated and designated “Ad-Con.” Transfection of cells was performed using Lipofectamine 3000 Transfection Reagent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s instructions, and the transfection efficiency was validated by qPCR.

The STING inhibitor C-176, the STING activator DMXAA, the JNK activator anisomycin, the JNK inhibitor SP600125, the autophagy inhibitor 3-methyladenine (3-MA), the lipid peroxidation end product 4-hydroxynonenal (4-HNE), the β-oxidation inhibitor etomoxir, the ATG16L1 enhancer peretinoin were purchased from MedChemExpress, and the lysosomal acid lipase (LIPA) inhibitor lalistat was purchased from Enamine. All mice were euthanized for further analysis after being fed a HFHCD for 16 weeks or an MCD for 6 weeks.

Total mRNA was isolated from liver samples or cells using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and reverse transcribed using a high-capacity cDNA reverse transcription kit (Roche, Indianapolis, IN, USA) according to the manufacturer’s protocols. The mRNA expression levels were quantified by quantitative PCR using SYBR Green (Roche). The expression level of each cDNA relative to that of the β-actin endogenous control was determined using the 2^-ΔΔCt method. Quantitative real-time PCR was repeated three times for each sample. The primers used in our study are listed in Supplementary Table 3.

DNA was extracted from 200 μL of cell culture supernatant from untreated and treated primary murine hepatocytes using a QIAmp DNA Mini Kit (QIAGEN, Duesseldorf, Germany) according to the manufacturer’s instructions. Real-time polymerase chain reaction (qPCR) was performed for the quantification of mtDNA. mtDNA was quantified using mouse mt-ATP6 primers/TaqMan 5' FAM-3' MGB probes (Bio-Rad, Hercules, CA, USA). The mitochondrial lysates of primary hepatocytes were used to isolate mtDNA using a mitochondrial DNA isolation kit (Abcam) according to the manufacturer’s instructions. Total DNA was isolated using a DNeasy Blood and Tissue Kit (QIAGEN) following the manufacturer’s instructions. RT‒qPCR analysis was performed using a sequence detection system (ABI Prism 7000; Applied Biosystems, Foster City, CA, USA) with a SYBR Green 1-step kit (Invitrogen). To evaluate the possibility of BMDMs engulfing mtDNA released from hepatocytes, mtDNA isolated from primary murine hepatocytes was tagged with Cy5-dCTP (Amersham Cy5-dCTP, GE Healthcare) using PCR following the manufacturer’s protocol (BioPrime DNA Labeling system, Life Technologies). The Cy5-labeled DNA PCR product was cleaned using a DNeasy Mini Spin Column (QIAGEN) before the DNA concentration was measured using a Nanodrop (Thermo Scientific).

Whole-cell lysates were obtained as previously described [18]. Proteins in tissues or cells were extracted with ice-cold lysis buffer (0.5% sodium deoxycholate, 1% Triton X-100, 10% glycerol, 0.1% SDS, 137 mM sodium chloride, 20 mM Tris, and pH 7.4). Proteins (20 g) were separated by 12% SDS‒PAGE and transferred to PVDF nitrocellulose membranes. Western blot analysis was performed using the following antibodies: anti-ATG16L1 (#8089, CST), antiβ-actin (#4970, CST), anti-α-SMA (#19245, CST), anti-collagen I (#72026, CST), anti-TIMP1 (#8946, CST), anti-pp65 (Ser536) (ab76302, Abcam), anti-p65 (66535-1-Ig, Proteintech), anti-p-IkBα (#2859, CST), anti-IkBα (10268-1- AP, Proteintech), anti-TNF-α (60291-1-Ig, Proteintech), anti-IL-6 (66146-1-Ig, Proteintech), anti-cleaved IL-1β (#63124, CST), anti-pro-IL-1β (#12242, CST), anti-cGAS (#79978, CST), anti-STING (#13647, CST), anti-p-TBK1 (#5483, CST), anti-TBK1 (#38066, CST), anti-p-IRF3 (#37829, CST), anti-IRF3 (ab68481, Abcam), anti-NLRP3 (#15101, CST), anti-cleaved caspase-1 (#89332, CST), anti-procaspase-1 (#24232, CST), anti-p-JNK (#9255, CST), anti-JNK (#9252, CST), anti-c-Jun (#9165, CST), anti-c-Fos (#74620, CST), anti-LC3B (#43566, CST), anti-p62 (#5114, CST), anti-GSDMD (#AF4012, Affinity), anti-rabbit IgG (#7074, CST), and anti-mouse IgG (#7076, CST). Image Lab software (National Institutes of Health) was used to quantify the protein expression, and β-actin was the control.

STING protein was expressed using a TNT Quick Coupled Transcription/Translation System (Promega) according to the manufacturer’s instructions [26]. STING protein was incubated with 100 μM FeSO4/25 mM ascorbate and 25 mM H₂O₂ in 100 mM KCl, 100 mM MgCl₂, and 50 mM HEPES (pH 7.2), and the reaction proceeded for 5 hours [27]. To remove excess low-molecular-weight compounds, proteins were concentrated with an Amicon Ultra 10 kDa filter (Millipore). STING carbonylation was evaluated using OxyBlot technology.

BMDMs were incubated with 100 μM palmitic acid probes (Invitrogen C10265) for 4 hours at 37°C. The cells were washed twice in PBS and then lysed on ice in 100 μL of 50 mM Tris, pH 8.0, containing 0.4% SDS and protease inhibitors, and the lysate was incubated with 1 mM CuSO4, 100 μM TBTA ligands, 100 μM biotin-alkyne, and 1 mM tris(2- carboxyethyl) phosphine for 1 hour at 25°C. After precipitation, protein complexes were enriched by adding streptavidin beads for 3 hours. Samples were washed three times with PBS, and SDS loading buffer was used to elute proteins at 95°C for 20 minutes, after which SDS–PAGE was used to separate the proteins.

An in vitro pull-down assay was performed to detect the binding of c-Jun and c-Fos to Tgfb1. BMDMs were lysed in NP-40 lysis buffer. Lysates were incubated with biotin-cJun or c-Fos (BioLog Life Science Institute) at 4°C for 4 hours and then incubated with streptavidin beads at 4°C for 4 hours. The beads were washed four times with lysis buffer and analyzed for Tgfb1 binding by immunoblotting.

The ChIP assay was performed using a Simplechip Enzymatic Chromatin IP Kit (Cell Signaling Technology, Billerica, MA, USA) according to the manufacturer’s instructions. Briefly, the cells were cross-linked with 37% formaldehyde, and the nuclei were subsequently extracted. After fragmentation by sonication and enzymatic digestion, the DNA‒protein complex was subjected to immunoprecipitation with 8 µg of anti-c-Jun antibody (Cell Signaling Technology) or rabbit IgG as a control. Purified ChIP DNA was amplified by real-time quantitative PCR. The primers for the AP-1 region of the Tgfb1 promoter were as follows: forward, 5’-GAAGGGGAGAGATGGCTCCACTGGG-3’; reverse, 5’-CTCCTCCTCATGGACTTGCTTT-3’.

A Tgfb1 promoter luciferase reporter plasmid (Tgfb1 luciferase) was constructed using a pGL3 luciferase vector (Promega, WI) according to the manufacturer’s instructions. BMDMs were transfected with the pGL3-Tgfb1- luciferase vector. After transfection for 6 h, the cells were washed and transfected with Ad-c-Jun or Ad-Con. After 48 h, the cells were lysed with passive lysis buffer, and transcriptional activity was measured using a luciferase assay system (Promega) according to the manufacturer’s instructions.

VO₂, VCO₂, the respiratory exchange ratio (RER), and locomotor activity were assessed using an eight-chamber Oxymax system (Columbus Instruments) [3]. Mice were placed in the chambers at 23°C with free access to food and water and acclimated for more than 50 minutes before measurement. Energy expenditure (EE) was calculated as (3.815+1.232×RER)×VO₂/lean mass.

Primary hepatocytes were immersed in fixative (2.5% glutaraldehyde) and stored overnight at 4°C. The samples were rinsed in the same buffer and postfixed for 1 h in 1% osmium tetroxide and 1% potassium ferrocyanide in 0.1 M cacodylate buffer to enhance membrane staining. The cells were then rinsed in distilled water, dehydrated in acetone at a low temperature to preserve lipids, and embedded in epoxy resin. Contrasted ultrathin sections (70 nm) were analyzed under a transmission electron microscope (HITACHI, Tokyo, Japan).

LPS-primed Atg16l1^fl/fl or Atg16l1^ΔMϕ BMDMs were incubated with Atg16l1^fl/fl primary hepatocyte-conditioned media (CM) for 48 h, and Atg16l1^fl/fl or Atg16l1^ΔMϕ BMDMs were subsequently subjected to RNA sequencing analysis (n=3/group). Total RNA was extracted from cells using TRIzol reagent (Invitrogen). cDNA libraries were constructed for each pooled RNA sample using the NEBNext^® Ultra^TM RNA Library Prep Kit for Illumina^® (NEB, Ipswich, MA, USA) according to the manufacturer’s instructions. Pathway analysis was used to determine the pathways significantly associated with DEGs according to the KEGG database.

The serum and media levels of the cytokines TNF-α, IL-6, IL-1β (Thermo Fisher Scientific), and 4-HNE (MyBioSource) were determined using ELISA kits according to the manufacturer’s instructions.

All the data were analyzed using unpaired Student’s t tests or one-way ANOVA accompanied by Bonferroni post hoc t tests. Statistical significance was set at P≤0.05 (twotailed). Statistical analysis was performed using GraphPad Prism Version 7.0 software (GraphPad Software, San Diego, CA, USA).

To further explore the role of ATG16L1 expression in MASH progression, we incubated LPS-primed Atg16l1^fl/fl or Atg16l1^ΔMϕ bone marrow-derived macrophages (BMDMs) with Atg16l1^fl/fl primary hepatocyte CM for 48 hours (Fig. 2A). Proinflammatory factors were examined, and the results showed that the gene expression and secretion of proinflammatory cytokines were markedly upregulated (Supplementary Fig. 3A, B). By assessing macrophage glycolysis and mitochondrial metabolism, we found that ATG16L1 expression could affect immune responses to MASH by regulating macrophage glycan metabolism. We measured glycolytic and mitochondrial metabolic activity using a Seahorse XFe96 analyzer. The results showed that the level of glycolysis in the Atg16l1^ΔMϕ BMDMs was markedly greater than that in the Atg16l1^fl/fl BMDMs stimulated with CM+LPS (Supplementary Fig. 3C). Moreover, the mitochondrial metabolism level in the Atg16l1^ΔMϕ BMDMs was significantly lower than that in the Atg16l1^fl/fl BMDMs stimulated with CM+LPS (Supplementary Fig. 3D). Next, the treated Atg16l1^fl/fl or Atg16l1^ΔMϕ BMDMs (shown in Fig. 2A) were subjected to RNA sequencing analysis. In total, we identified 5,888 DEGs (Fig. 2B), including 3076 genes with upregulated expression and 2812 genes with downregulated expression (Atg16l1^ΔMϕ vs. Atg16l1^fl/fl). The top 19 uniquely upregulated and downregulated genes included the IFN-I–dependent genes Oas1g, Oas3, Irgm1, and Irf7, which are known to play a role in innate immunity and the antiviral response (Fig. 2C). Using the STRING database [30], we performed an interaction analysis of the 19 abovementioned genes, which revealed a densely connected network of known IFN-stimulated genes including the Oas1g, Irf7, Oas2, and Isg15 genes (Fig. 2D). Moreover, qRT‒PCR analysis of these 19 upregulated IFN-stimulated genes confirmed the above findings (Fig. 2E). Next, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis according to genotype and treatment, which revealed significant enrichment of upregulated genes related to NF-κB signaling and the phagosome signaling pathway (Fig. 2F). The upregulation of the proinflammatory NF-κB signaling pathway was validated by observation of increased protein expression levels of p-p65 (Ser536), p-IkBα, TNF-α, IL-6, and IL-1β (Fig. 2G). We analyzed the upregulated phagosome signaling pathway and found 65 significantly upregulated phagosome-related genes (Fig. 2H). The increased expression levels of three representative phagosome-related genes, Marco, Mrs1, and Cd14, in activated Atg16l1^ΔMϕ BMDMs further confirmed this result (Fig. 2I). Moreover, RNA sequencing and mRNA expression data indicated that the expression of the transcription factor EB (Tfeb), a master transcriptional regulator of autophagy and lysosomes [31], was significantly lower in activated Atg16l1^ΔMϕ BMDMs than in activated Atg16l1^fl/fl BMDMs (Fig. 2J, K). Further analysis of the KEGG pathway enrichment analysis results also revealed significant enrichment of downregulated genes related to lysosomes (Fig. 2L). Next, we analyzed the downregulated lysosome signaling pathway and identified 46 significantly downregulated lysosome-related genes (Fig. 2M). The decreased expression levels of three representative lysosome-related genes, Hexa, Ctsb, and Gla, in activated Atg16l1^ΔMϕ BMDMs further validated this result (Fig. 2N). To clarify the above results, we constructed myeloid-specific Atg16l1 gene knock-in mice (Atg16l1^OE). The overall expression levels of ATG16L1 in liver tissues or hepatocytes from Atg16l1^OE and Atg16l1^fl/fl mice showed no significant difference. BMDMs isolated from Atg16l1^OE mice showed significantly higher levels of ATG16L1 than those isolated from Atg16l1^fl/fl mice (Supplementary Fig. 3E). We incubated LPS-primed Atg16l1^fl/fl or Atg16l1^OE BMDMs with primary hepatocyte CM from Atg16l1^fl/fl mice for 48 h (Supplementary Fig. 3F). The expression and secretion of proinflammatory factors were significantly decreased (Supplementary Fig. 3G, H). The level of glycolysis in the Atg16l1^OE BMDMs was markedly lower than that in the Atg16l1^fl/fl BMDMs stimulated with CM+LPS (Supplementary Fig. 3I), whereas the level of mitochondrial metabolism in the Atg16l1^OE BMDMs was significantly greater than that in the Atg16l1^fl/fl BMDMs stimulated with CM+LPS (Supplementary Fig. 3J). Moreover, the expression levels of IFN-related genes were markedly lower in activated Atg16l1^OE BMDMs than in activated Atg16l1^fl/fl BMDMs (Supplementary Fig. 3K). Furthermore, the expression levels of proinflammatory proteins in the NF-κB signaling pathway were also lower in activated Atg16l1^OE BMDMs than in activated Atg16l1^fl/fl BMDMs (Supplementary Fig. 3L).

A previous study demonstrated that cyclic GMP-AMP synthase (cGAS) synthesizes the dinucleotide secondary messenger 2’3’-cGAMP, which then activates STING, an endoplasmic reticulum-resident adaptor, inducing its trafficking to the endoplasmic reticulum–Golgi intermediate complex (ERGIC), where TANK-binding kinase 1 (TBK1) phosphorylates the IFN-I-inducing transcription factor interferon regulatory factor 3 (IRF3) [32]. Meanwhile, activated IRF3 promotes NLRP3 inflammasome expression [33]. Our results demonstrated that the cGAS/STING signaling pathway was markedly activated in LPS-primed Atg16l1^ΔMϕ BMDMs compared with Atg16l1^fl/fl BMDMs cultured in CM from Atg16l1^fl/fl primary hepatocytes for 48 hours (Fig. 3A–C).

Our previous study revealed that mitochondrial DNA (mtDNA) released from pyroptotic hepatocytes contributes to STING activation in macrophages [34]. In this study, we also investigated hepatocyte pyroptosis in liver tissues from MASH patients and found that hepatocyte pyroptosis was significantly greater in liver tissues from MASH patients than in those from normal controls, which was confirmed by the increase in Gasdermin D (GSDMD) activation (Supplementary Fig. 4A, B). Furthermore, we examined GSDMD expression in murine MASH model mice and found that GSDMD was significantly activated in hepatocytes isolated from MASH mice, which was reflected by increased levels of GSDMD-N fragments and increased hepatocyte death (Supplementary Fig. 4C, D). As damage-associated molecular pattern molecules (DAMPs) play an important role in the inflammatory response [35], we examined the levels of DAMPs in hepatocytes isolated from MASH mice. The levels of ATP and mtDNA in the supernatants of pyroptotic hepatocytes isolated from MASH mice were significantly greater than those in the supernatants of control hepatocytes (Supplementary Fig. 4E, F). Next, we isolated mtDNA from pyroptotic hepatocytes (Supplementary Fig. 4G) and stimulated WT BMDMs with isolated mtDNA tagged with Cy5. mtDNA-Cy5 was engulfed by BMDMs at 12 h postcoculture (Supplementary Fig. 4H). Furthermore, CM from palmitic acid- and oleic acid (PAOA)-stimulated hepatocytes was also used to coculture LPS-primed BMDMs with or without the mtDNA scavenger ethidium bromide (EtBr) (Supplementary Fig. 4I). We found that the expression of proinflammatory genes (Tnfa, Il6, Il1b), Ifnb1, and Tgfb1 was significantly decreased in activated BMDMs with mtDNA depletion (Supplementary Fig. 4J–L). Next, mtDNA was isolated from pyroptotic hepatocytes and used to stimulate Atg16l1^fl/fl, Atg16l1^ΔMϕ or Atg16l1^OE BMDMs. mtDNACy5 was more engulfed by the Atg16l1^ΔMϕ BMDMs and less engulfed by the Atg16l1^OE BMDMs at 12 hours postcoculture with mtDNA-Cy5 than by the Atg16l1^fl/fl BMDMs (Supplementary Fig. 4M).

To explore the key role of mtDNA in regulating the ATG16L1-mediated macrophage immune response, we cocultured BMDMs from Atg16l1^fl/fl or Atg16l1^ΔMϕ mice with mtDNA isolated from pyroptotic hepatocytes (Fig. 3D). The results suggested that the cGAS/STING signaling pathway was markedly activated (Fig. 3E–G) in LPS-primed Atg16l1^ΔMϕ BMDMs compared with Atg16l1^fl/fl BMDMs stimulated with mtDNA. Next, we treated activated Atg16l1^ΔMϕ BMDMs with the STING inhibitor C-176 to clarify the important role of STING signaling in ATG16L1-mediated macrophage inflammation. The results showed that the downstream signaling activation of STING was markedly decreased (Fig. 3H–J). Subsequently, we treated Atg16l1^ΔMϕ Tmem173^ΔMϕ BMDMs with mtDNA to clarify the important role of STING signaling in ATG16L1-mediated macrophage inflammation. STING signaling and inflammation were significantly suppressed in the Atg16l1^ΔMϕ Tmem173^ΔMϕ BMDMs compared with the Atg16l1^ΔMϕ BMDMs (Fig. 3K, L). Furthermore, we cocultured Atg16l1^OE BMDMs with CM from PAOA-stimulated hepatocytes or mtDNA from pyroptotic hepatocytes and found that the STING signaling pathway was markedly inhibited in the Atg16l1^OE BMDMs compared with the Atg16l1^fl/fl BMDMs (Supplementary Fig. 5A–E). Next, we treated activated Atg16l1^OE BMDMs with the STING activator DMXAA. The results indicated that STING signaling was upregulated in the opposite manner in the Atg16l1^OE BMDMs treated with DMXAA (Supplementary Fig. 5F–H). To explore the role of STING in MASH, we constructed MASH model Atg16l1^ΔMϕ Tmem173^ΔMϕ mice. As a result, attenuated MASH was found in Atg16l1^ΔMϕ Tmem173^ΔMϕ mice compared with that in Atg16l1^ΔMϕ mice (Supplementary Fig. 6A–F). Furthermore, we also assessed the body weight and energy expenditure of HFHCD-fed Atg16l1^ΔMϕ Tmem173^ΔMϕ mice compared to Atg16l1^ΔMϕ mice. The results indicated that HFHCD-fed Atg16l1^ΔMϕ Tmem173^ΔMϕ mice exhibited decreased body weight compared to Atg16l1^ΔMϕ mice, along with increased energy expenditure (Supplementary Fig. 6G, H).

Next, we determined whether ATG16L1-mediated macrophage inflammation affects lipid metabolism in hepatocytes. We assessed the ability of macrophage-derived TNF-α, IL-6, and IL-1β to induce steatohepatitis and the effectiveness of neutralizing antibodies against these cytokines against steatohepatitis. We used a coculture system of LPS-primed Atg16l1^ΔMϕ, Atg16l1^OE, or Atg16l1^fl/fl macrophages with primary Atg16l1^fl/fl hepatocytes (Fig. 4A). The neutralizing antibodies against TNF-α, IL-6, or IL-1β, including anti-TNF-α, MR16-1, and IL-1RA, decreased hepatocyte lipid accumulation and inflammation (Fig. 4B, C) in the Atg16l1^fl/fl hepatocytes cocultured with LPS-primed Atg16l1^ΔMϕ macrophages, whereas administration of TNF-α, IL-6, or IL1β increased lipid deposition (Supplementary Fig. 7A) and inflammatory responses (Supplementary Fig. 7B) in the Atg16l1^fl/fl hepatocytes cocultured with LPS-primed Atg16l1^OE macrophages.

A previous study showed that JNK1 activation in macrophages could promote TGF-β1 expression and affect HSC activation [36]. The activation of hepatic JNK1, c-Jun, and AP-1 signaling occurs in parallel with the development of steatohepatitis in MASH mice [37]. We determined the phosphorylation levels of JNK, c-Jun, and c-Fos and found that JNK, c-Jun, and c-Fos phosphorylation was significantly greater in Atg16l1^ΔMϕ macrophages than in Atg16l1^fl/fl macrophages (Fig. 4D). The expression of Tgfb1 in Atg16l1^ΔMϕ macrophages was also greater than that in Atg16l1^fl/fl macrophages (Fig. 4E). We performed sequence alignment analysis of the Tgfb1 gene and identified an AP-1 binding site between amino acids -1382 and -1376 in the Tgfb1 promoter region (Fig. 4F). Based on this information, we performed a DNA pull-down assay to determine whether LPS treatment might alter the association of AP-1 with the Tgfb1 promoter. We observed that LPS treatment increased the binding of c-Jun and c-Fos to the AP-1 binding site in the Tgfb1 promoter region (Fig. 4F). Moreover, a chromatin immunoprecipitation (ChIP)-qPCR assay revealed the enrichment of c-Jun at the Tgfb1 promoter region in response to LPS treatment (Fig. 4G). Next, mutation of the AP-1 binding sites to which c-Jun binds in the Tgfb1 promoter region abrogated the increase in Tgfb1 promoter activity triggered by c-Jun overexpression (Fig. 4H). We treated activated Atg16l1^ΔMϕ BMDMs with the STING inhibitor C-176 and found that the protein expression levels of phosphorylated JNK, c-Jun, and c-Fos and the mRNA expression level of Tgfb1 were significantly decreased (Fig. 4I, J), whereas the protein expression levels of phosphorylated JNK, c-Jun, and c-Fos and the mRNA expression level of Tgfb1 were significantly increased when C-176-treated Atg16l1^ΔMϕ BMDMs were stimulated with the JNK activator anisomycin (Fig. 4K, L). To confirm the above results, we determined the protein expression levels of phosphorylated JNK, c-Jun, and c-Fos and the mRNA expression level of Tgfb1 in Atg16l1^OE BMDMs. The results showed that Atg16l1 overexpression markedly reduced the expression levels of phosphorylated JNK, c-Jun, and c-Fos and the mRNA expression of Tgfb1 (Supplementary Fig. 7C, D). Next, we treated activated Atg16l1^OE BMDMs with the STING activator DMXAA and found that the protein levels of phosphorylated JNK, c-Jun, and c-Fos significantly increased (Supplementary Fig. 7E), whereas the protein expression levels of phosphorylated JNK, c-Jun, and c-Fos significantly decreased when DMXAA-treated Atg16l1^ΔMϕ BMDMs were stimulated with the JNK inhibitor SP600125 (Supplementary Fig. 7F). To validate whether macrophage ATG16L1 expression affects HSC activation, we incubated primary mouse HSCs in the absence or presence of TGF-β1 (8 ng/mL) with CM from Atg16l1^ΔMϕ, Atg16l1^OE or Atg16l1^fl/fl BMDMs treated with primary hepatocyte CM from Atg16l1^fl/fl mice for 24 hours (Fig. 4M). α-SMA immunofluorescence and Acta2, Col1a1, and Timp1 gene expression in HSCs indicated that macrophage Atg16l1 knockout markedly promoted the activation of HSCs, while Atg16l1 overexpression significantly suppressed HSC activation (Fig. 4N–Q).

Next, we explored the specific mechanisms underlying the ATG16L1-mediated regulation of the STING signaling pathway in MASH. In yeast and mammalian cells, ATG16L1 is responsible for proper subcellular localization of the autophagy machinery [38,39]. The decreased protein expression level of LC3B and increased protein expression level of p62 in the Atg16l1^ΔMϕ BMDMs suggested that Atg16l1 knockout markedly suppressed autophagy (Fig. 5A), while the increased protein expression level of LC3B and decreased protein expression level of p62 in the Atg16l1^OE BMDMs suggested that Atg16l1 overexpression markedly promoted autophagy (Supplementary Fig. 8A). We found that the inhibition of lipophagy led to increased lipid accumulation, as evidenced by the elevated TG (Fig. 5B), BODIPY 493/503 (Fig. 5C), oil red O (Supplementary Fig. 8B), and lipid droplet levels in BMDMs, as visualized by electron microscopy (Supplementary Fig. 8C). Previous studies indicated that lipophagy, i.e., the autophagy of lipid droplets (LDs) [40], plays an important role in regulating cellular lipid metabolism [41]. Thus, we hypothesized that ATG16L1 promotes lipophagy and thereby decreases cellular lipid accumulation. Autophagy is characterized by the formation of autophagosomes, which deliver cytoplasmic material to lysosomes [42]. To confirm that lysosomes regulate intracellular lipid levels, the effect of lysosomal hydrolysis on lipid stores was examined. Dual immunofluorescence studies revealed increased colocalization of LDs with lysosomeassociated membrane protein 1 (LAMP1) in LPS-primed and CM-treated Atg16l1^fl/fl BMDMs; Atg16l1 deletion inhibited lysosomal hydrolysis of lipid stores, whereas Atg16l1 overexpression promoted lysosomal hydrolysis of lipid stores (Fig. 5D). Furthermore, LD colocalization with the autophagosome marker LC3 demonstrated a direct association between LDs and autophagosomes, supporting a constitutive function for autophagy in regulating LDs (Fig. 5E). Electron microscopy was used to verify LD degradation by autophagic lysosomes (Supplementary Fig. 8D). A previous study indicated that the inhibition of autophagy decreases TG β-oxidation [38]. Our study also demonstrated that Atg16l1 knockout decreased macrophage β-oxidation, whereas Atg16l1 overexpression increased β-oxidation (Fig. 5F). A recent study demonstrated that lipid peroxidation promoted the production of 4-hydroxynonenal (4-HNE), which is the major end product of lipid peroxidation. Increased 4-HNE production enhanced STING carbonylation, resulting in the inhibition of STING trafficking from the endoplasmic reticulum to the Golgi apparatus and the suppression of STING activation [43]. We also examined cellular 4-HNE levels and revealed that Atg16l1 knockout decreased 4-HNE production, whereas Atg16l1 overexpression increased 4-HNE production (Fig. 5G). Another study showed that peroxisomal β-oxidation acted as a sensor for intracellular fatty acid levels and regulated lipolysis [44]. Our study demonstrated that the presence of 4-HNE suppressed the mRNA expression of Ifnb1 (Fig. 5H, I). Moreover, the inhibition of lipophagy and β-oxidation decreased the production of 4-HNE and increased the mRNA expression of Ifnb1 (Fig. 5J, K). In addition, 4-HNE induced STING carbonylation (Fig. 5L and Supplementary Fig. 8E), Atg16l1 knockout inhibited STING carbonylation, and Atg16l1 overexpression promoted STING carbonylation (Fig. 5M). Study has indicated that the palmitoylation of STING was crucial for the IFN I response in macrophages, but it didn’t affect the translocation of STING [45]. Our results indicated that 4-HNE blocked STING palmitoylation (Fig. 5N), Atg16l1 knockout promoted STING palmitoylation, and Atg16l1 overexpression inhibited STING palmitoylation (Fig. 5O). The trafficking of STING from the ER to the Golgi apparatus is a crucial step for STING activation and subsequent IRF3 activation [32]. To investigate STING trafficking from the ER to the Golgi apparatus in the context of differential ATG16L1 expression, we costained for STING and the Golgi marker GM130 in LPS-primed Atg16l1^fl/fl, LPS-primed and CM-treated Atg16l1^fl/fl, LPS-primed and CM-treated Atg16l1^ΔMϕ, and LPS-primed and CM-treated Atg16l1^OE BMDMs (Fig. 5P). The results showed that macrophage ATG16L1 depletion promoted STING localization to the Golgi apparatus, whereas ATG16L1 overexpression inhibited STING localization to the Golgi apparatus. Furthermore, compared to Atg16l1^fl/fl MASH mice, liver tissues of Atg16l1^ΔMϕ MASH mice exhibited impaired autophagy and activation of the STING signaling pathway, indicated by upregulation of protein expression levels of P62, cGAS, Pal-STING, P-TBK1, and P-IRF3, along with downregulation of protein expression levels of LC3B and Carbonyl-STING (Supplementary Fig. 8F).

Next, Atg16l1^OE mice and Atg16l1^fl/fl mice were bred and fed an HFHCD for 16 weeks or an MCD for 6 weeks. Macrophage Atg16l1 overexpression decreased proinflammatory gene expression (Fig. 6A) and macrophage infiltration (Fig. 6B). Histopathology of liver sections demonstrated that Atg16l1^OE mice had improved liver histology and less steatosis and inflammation than Atg16l1^fl/fl mice (Fig. 6C). Accordingly, the levels of serum ALT and hepatic TG were significantly lower in the Atg16l1^OE mice than in the Atg16l1^fl/fl mice (Fig. 6D). Moreover, Atg16l1^OE mice exhibited significantly mitigated liver fibrosis, as indicated by decreased Sirius red and α-SMA staining (Fig. 6C); decreased expression levels of the profibrotic genes Acta2, Col1a1 and Timp1; and decreased protein expression levels of α-SMA, collagen-I, and TIMP-1 (Fig. 6E, F). Moreover, compared to Atg16l1^fl/fl mice, liver tissues of Atg16l1^OE mice exhibited enhanced autophagy and suppression of the STING signaling pathway, indicated by decreased protein expression levels of P62, cGAS, Pal-STING, P-TBK1, and P-IRF3, along with increased protein expression levels of LC3B and Carbonyl-STING (Supplementary Fig. 8G). The weights of tissues, including livers, BAT, eWAT and ingWAT, of the Atg16l1^OE mice fed an HFHCD were lower than those of the Atg16l1^fl/fl mice fed the same diet (Fig. 6G, H).

The body weights and fat masses of the NCD-fed mice were not significantly different between the two genotypes, but we observed a marked decrease in weight gain driven by fat mass in the HFHCD-fed Atg16l1^OE mice (Supplementary Fig. 9A). Furthermore, there was no significant difference between the amount of food consumed by HFHCD-fed Atg16l1^OE mice and that consumed by Atg16l1^fl/fl mice (Supplementary Fig. 9B). Similarly, there was no significant difference between the amount of food consumed by the NCD-fed Atg16l1^OE mice and that consumed by the Atg16l1^fl/fl mice (Supplementary Fig. 9C). Nevertheless, the rates of VO₂, VCO₂, and EE in the Atg16l1^OE mice were significantly greater than those in the Atg16l1^fl/fl mice (Supplementary Fig. 9D-F, 9H-J, 9L). In contrast, locomotor activity did not significantly differ between the Atg16l1^OE and Atg16l1^fl/fl mice (Supplementary Fig. 9G, K). Collectively, these results suggest that Atg16l1 overexpression in macrophages ameliorates steatohepatitis in mice.

Finally, we explored the effects of pharmacologically induced ATG16L1 overexpression on steatohepatitis progression. An ATG16L1 enhancer (peretinoin) was used to treat activated WT BMDMs stimulated with or without DMXAA. The results showed that increased ATG16L1 protein expression levels inhibited STING signaling pathway activation, decreased the expression of phosphorylated JNK, cJun, and c-Fos, and suppressed macrophage inflammation and Tgfb1 expression (Fig. 7A–F). Moreover, peretinoin was added to the diets of WT mice fed an HFHCD for 16 consecutive weeks or fed an MCD for 6 consecutive weeks (Fig. 7G). Peretinoin treatment increased ATG16L1 protein expression levels in liver tissues from MASH mice (Fig. 7H). Histopathology of liver sections revealed that compared with untreated mice fed the same diet, WT MASH mice treated with peretinoin had improved liver histology with markedly reduced steatosis, inflammation, and fibrosis markers (Fig. 7I). Accordingly, the levels of serum ALT and hepatic TG were significantly lower in WT MASH mice treated with peretinoin than in untreated mice fed the same diet (Fig. 7J). Similar results were found regarding the gene and protein expression levels of profibrotic markers (Fig. 7K, L). Moreover, proinflammatory gene expression and macrophage infiltration were significantly lower in WT MASH mice treated with peretinoin than in untreated mice fed the same diet (Fig. 7M, N). Furthermore, compared with untreated mice fed the same diet, HFHCD-fed WT mice treated with peretinoin exhibited a significant decrease in visceral adipose tissue inflammation (Fig. 7O).

To further validate the role of macrophage ATG16L1 in MASH, we also treated Atg16l1^ΔMϕ mice and Atg16l1^fl/fl mice with peretinoin in the MCD-induced MASH model. Histopathology of liver sections revealed that compared with untreated group, the MASH phenotype in Atg16l1^ΔMϕ mice treated with peretinoin showed slight improvement, with mild reductions in markers of steatosis, inflammation, and fibrosis. However, it was more severe than in Atg16l1^fl/fl mice treated with peretinoin (Supplementary Fig. 10A). We speculate that peretinoin might also partially by promoting ATG16L1 expression in hepatocytes and other cell types to ameliorate MASH, which deserved further exploration in our future study. Accordingly, serum ALT and liver TG levels in Atg16l1^ΔMϕ mice treated with peretinoin were lower than untreated Atg16l1^ΔMϕ mice, but higher than peretinoin-treated Atg16l1^fl/fl mice (Supplementary Fig. 10B). Additionally, profibrotic and proinflammatory genes also exhibited similar outcomes (Supplementary Fig. 10C, D). These results demonstrate that pharmacological enhancement of ATG16L1 protein expression levels alleviates steatohepatitis and fibrosis in mice.