All animal experiments were approved by the Animal Experimentation Committee of Wuhan University (Wuhan, China; approval no. WQ20210298). All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (25) and the ARRIVE 2.0 guidelines (26).

Male C57BL/6 mice aged 6–8 weeks were housed under specific pathogen-free conditions with a 12-h light/dark cycle and ad libitum access to food and water. Experimental and control mice underwent the same fasting protocol before cecal ligation and puncture (CLP) surgery. A total of two genetically modified mouse models were used for distinct experimental purposes. Global Erbin-knockout (Erbin^−/−) mice, obtained from Wuhan Xianran Biological Technology Co., Ltd., were used exclusively in the in vivo CLP model to evaluate the overall effect of Erbin deficiency on SLI. Macrophage-specific Erbin conditional-knockout mice (Erbin^{fl/fl/Lyz2-Cre}) were generated by crossing Erbin^fl/fl mice with Lyz2-Cre mice (both Cyagen Biosciences, Inc.). BMDMs isolated from Erbin^{fl/fl/Lyz2-Cre} mice and Erbin^fl/fl littermate controls were used exclusively for in vitro mechanistic experiments examining the macrophage-intrinsic role of Erbin in inflammatory signaling. Primers used to confirm the genotype can be found in Table I. A schematic of how the mice were used across experiments is presented in Fig. S1.

General anesthesia was induced by intraperitoneal administration of pentobarbital sodium (50 mg/kg). The mice were placed in the supine position, and a 1-cm midline abdominal incision was made to expose the cecum. The cecum was ligated with 3–0 silk and punctured with a 21-gauge needle. It was then returned to the abdominal cavity, and the incision was closed. Subsequently, 1 ml of prewarmed 0.9% saline was administered subcutaneously for fluid resuscitation. Sham-operated mice underwent the same surgical procedure without cecal ligation or puncture (27). Following sepsis induction, the mice were monitored at least twice daily. Humane endpoints included >20% loss of body weight, inability to eat or drink, persistent recumbency or markedly reduced responsiveness, severe respiratory distress or cyanosis, severe dehydration, seizures, paralysis, coma or other signs of a moribund condition. Animals reaching any of these endpoints were immediately euthanized.

Mice were euthanized by CO₂ inhalation in accordance with GB/T 39760-2021. CO₂ was delivered at a flow rate of 9 l/min, which corresponds to a volume displacement rate of 30% per minute for the 30-L euthanasia chamber, until respiration ceased, followed by cervical dislocation as a secondary method. Death was confirmed by the absence of respiration, heartbeat and pedal withdrawal reflex before tissue collection. Hepatic tissues were collected 24 h after CLP for histological evaluation. The right liver lobe was rinsed with PBS, fixed in 4% paraformaldehyde at 4°C for 24 h, embedded in paraffin, sectioned at 4 µm. The sections were then stained with hematoxylin at room temperature for 4 min and eosin at room temperature for 1 min. Histopathological changes were examined by light microscopy and digitally imaged. Liver injury was scored semi-quantitatively according to six parameters: Cytoplasmic pallor, vacuolization, nuclear condensation, nuclear fragmentation, chromatin dissolution and sinusoidal erythrocyte accumulation. Each parameter was graded from 0 to 3, yielding a total score of 0–18 per microscopic field (28).

After CLP surgery, mice were monitored every 12 h for 7 days, and survival was recorded throughout the observation period. Mice that failed to respond to external stimuli were considered dead. The sample size was comparable to those commonly used in survival analyses of CLP-induced sepsis models (29,30). Survival curves were generated using the Kaplan-Meier method and compared using the log-rank (Mantel-Cox) test in GraphPad Prism (version 10.6.1; Dotmatics).

Commercial assay kits were used to measure serum alanine aminotransferase (ALT; cat. no. C009-2-1; Nanjing Jiancheng Bioengineering Institute), aspartate aminotransferase (AST; cat. no. C010-2-1; Nanjing Jiancheng Bioengineering Institute) and lactate dehydrogenase (LDH; cat. no. A020-2-2; Nanjing Jiancheng Bioengineering Institute) levels according to the manufacturer's instructions.

Liver tissues were fixed in 10% neutral buffered formalin (NBF) at 4°C for 24 h, embedded in paraffin, sectioned at 4 µm thickness and then incubated overnight at 4°C with an anti-NLRP3 antibody (1:200; cat. no. GB114320-50; Wuhan Servicebio Technology Co., Ltd.). After three 5-min washes with PBS, the sections were incubated with the appropriate secondary antibody (1:200; cat. no. G1312; Wuhan Servicebio Technology Co., Ltd.) at 37°C for 30 min and developed using 3,3′-diaminobenzidine. The sections were counterstained with hematoxylin at room temperature for 4 min, dehydrated, mounted and examined by light microscopy. NLRP3-positive areas were quantified using ImageJ 1.54j software (National Institutes of Health).

Liver samples were fixed with 2.5% glutaraldehyde at 4°C for 1 h and then with 1% osmium tetroxide at room temperature for 2 h. After dehydration, the samples were infiltrated with epoxy resin at 37°C for 8 h and embedded in epoxy resin at 60°C for 48 h. Ultrathin sections (80 nm) were cut and stained with 2% uranyl acetate at room temperature for 10 min and then with lead citrate at room temperature for 5 min, followed by examination under a Hitachi H-600 transmission electron microscope (Hitachi, Ltd.). Abnormal mitochondria were identified as previously described (31).

Consistent with previous studies investigating macrophage-mediated inflammatory mechanisms in liver disease (32–34), BMDMs from macrophage-specific Erbin conditional-knockout mice (Erbin^fl/fl/^Lyz2-Cre) and Erbin^fl/fl littermate controls were used to examine the macrophage-intrinsic effects of Erbin deficiency. Following euthanasia, the femurs and tibiae were dissected, cleared of soft tissue and immersed in 75% ethanol for 5 min. Bone marrow was flushed with PBS, passed through a 70-µm cell strainer, and centrifuged at 300 × g for 10 min. Erythrocytes were lysed using ACK buffer (cat. no. R1013; Beijing Solarbio Science & Technology Co., Ltd.). The remaining cells were cultured in high-glucose DMEM (cat. no. 11995; Beijing Solarbio Science & Technology Co., Ltd.) supplemented with 10% FBS (Procell Life Science & Technology Co., Ltd.) and 25 ng/ml macrophage colony-stimulating factor (M-CSF; MedChemExpress) at 37°C in 5% CO₂ for 7 days to generate mature adherent BMDMs. To activate the inflammasome, BMDMs were primed with lipopolysaccharide (LPS; 1 µg/ml; MilliporeSigma) for 4 h and subsequently stimulated with ATP (5 mM; MedChemExpress) for 30 min at 37°C (35). For MitoQ experiments, BMDMs were pretreated with 1 µM MitoQ (HY-100116A; MedChemExpress; dissolved in 0.02% DMSO) for 2 h at 37°C before LPS/ATP stimulation. A concentration of 0.02% DMSO was used as the vehicle control in the corresponding groups.

For the cell counting kit (CCK)-8 assay, BMDMs were seeded in 96-well plates at 1×104 cells/well and subjected to the indicated treatments. CCK-8 reagent (10 µl; Biosharp Life Sciences) was then added to each well, followed by incubation at 37°C for 30 min, and cell viability was measured using a microplate reader.

For the lactate dehydrogenase (LDH) assay, culture supernatants were collected and centrifuged at 500 × g for 10 min at 4°C. LDH release was measured using a commercial kit (cat. no. A020-2-2; Beyotime Biotechnology) according to the manufacturer's instructions.

For Calcein-AM/propidium iodide (PI) staining, BMDMs were seeded in 24-well plates at 1×105 cells/well and subjected to LPS/ATP stimulation with or without MitoQ. Cells were stained using a Calcein-AM/PI kit (cat. no. BB-4126; Bebo Biotech). Briefly, cells were incubated with Calcein-AM working solution at 37°C in the dark for 30 min, washed twice with PBS, and then incubated with PI working solution under the same conditions for an additional 30 min. Following PBS washes, fluorescence images were acquired by microscopy, and cell death was quantified using ImageJ software.

At 24 h after CLP, liver sections were prepared for dihydroethidium (DHE) staining as previously described (36). Intracellular ROS in LPS/ATP-stimulated BMDMs was assessed using a commercial ROS assay kit (cat. no. BL3157A; Biosharp Life Sciences) according to the manufacturer's instructions. Briefly, cells were incubated with 10 µM DCFH-DA at 37°C for 30 min in the dark. Fluorescence images were acquired by microscopy, and fluorescence intensity was quantified using ImageJ software.

Commercial biochemical assay kits were used to measure malondialdehyde (MDA; cat. no. A003-1-2; Nanjing Jiancheng Bioengineering Institute) levels and superoxide dismutase (SOD; cat. no. A001-1-2; Nanjing Jiancheng Bioengineering Institute) activity in liver and BMDM protein homogenates according to the manufacturer's instructions.

mtROS production was assessed using MitoSOX Red (40778ES50; Shanghai Yeasen Biotechnology Co., Ltd.). BMDMs were seeded in 12-well plates, subjected to the indicated treatments, and incubated with 500 nmol/l MitoSOX Red in HBSS containing calcium and magnesium at 37°C for 30 min in the dark. After two washes with prewarmed PBS, fluorescence images were acquired by microscopy and analyzed using ImageJ software.

Mitochondrial membrane potential was assessed using a JC-1 Mitochondrial Membrane Potential Assay Kit (C2006; Beyotime Biotechnology) according to the manufacturer's instructions. BMDMs were seeded in six-well plates and subjected to the indicated treatments, followed by incubation with JC-1 working solution for 20 min at 37°C. The cells were then collected and analyzed by flow cytometry (Beckman Coulter, Inc.).

Liver tissue (100 mg) was rinsed with ice-cold PBS and homogenized on ice in 900 µl PBS (pH 7.4) containing 1× protease inhibitor cocktail (Roche Applied Science). Homogenates were centrifuged at 12,000 × g for 15 min at 4°C, and the supernatants were aliquoted and stored at −80°C until analysis. Cytokine concentrations were normalized to total protein concentrations determined using a BCA assay. TNF-α (cat. no. RK00027), IL-18 (cat. no. RK00104), IL-1β (cat. no. RK04878) and IL-6 (cat. no. RK00008) levels in liver homogenate supernatants (37) and BMDM culture supernatants were measured using ELISA kits (all from ABclonal Biotech Co., Ltd.) according to the manufacturer's instructions. Absorbance was measured at 450 nm using a microplate reader (BioTek; Agilent Technologies, Inc.).

Total RNA was extracted using TRIzol reagent (Beijing Solarbio Science & Technology Co., Ltd.), and reverse transcription was performed using ABScript III RT Mix (ABclonal Biotech Co., Ltd.) under the following conditions: 37°C for 2 min, 55°C for 15 min, 85°C for 5 min and then held at 4°C. qPCR was performed using 2X Universal SYBR Green Fast qPCR Mix (ABclonal Biotech Co., Ltd.) with the following thermocycling conditions: Initial denaturation at 95°C for 3 min, followed by 42 cycles of 95°C for 5 sec and 60°C for 30 sec. Relative gene expression was calculated using the 2^−ΔΔCq method (38), with β-actin as the internal control. Primer sequences are listed in Table II.

Whole-cell lysates were prepared in NP-40 lysis buffer (Beijing Solarbio Science & Technology Co., Ltd.) supplemented with freshly prepared protease inhibitor cocktail and centrifuged at 14,000 × g for 10 min at 4°C to remove insoluble material. For each sample 1 ml supernatant was collected, of which 100 µl was retained as the input and the remaining 900 µl was used for co-immunoprecipitation. For each co-immunoprecipitation reaction, 900 µl of lysate containing ~3 mg total protein was incubated overnight at 4°C with 4.5 µg anti-NLRP3 antibody (1.5 µg/mg protein; cat. no. A5652; ABclonal Biotech Co., Ltd.) or 4.5 µg species-matched normal IgG (1.5 µg/mg protein; cat. no. 30000-0-AP; Proteintech Group, Inc.) under gentle agitation. A volume of 30 µl Protein A/G agarose beads (MedChemExpress) were then added and incubated for an additional 2–4 h with rotation at 4°C. The beads were extensively washed three times with ice-cold immunoprecipitation buffer, and the immunoprecipitated proteins were eluted by heating in SDS-PAGE loading buffer at 95–100°C for 5–10 min. Eluted proteins were separated by SDS-PAGE and analyzed by western blotting. Normal IgG was used as the negative immunoprecipitation control.

Total proteins were extracted from liver tissues and BMDMs using RIPA lysis buffer (Beyotime Biotechnology) supplemented with protease inhibitors (BioSharp Life Sciences). Protein concentrations were determined using a BCA assay kit (Beyotime Biotechnology). A total of 20 µg protein per lane was separated on 8–12% SDS-PAGE gels and transferred to polyvinylidene difluoride membranes (MilliporeSigma). Membranes were blocked with 5% non-fat milk and incubated overnight at 4°C with primary antibodies against Erbin (1:1,000; cat. no. NBP2-56104; Novus Biologicals, LLC), thioredoxin-interacting protein (TXNIP; 1:1,000; 14715; Cell Signaling Technology, Inc.), NLRP3 (1:2,000; A5652; ABclonal Biotech Co., Ltd.), GSDMD (1:1,000; A10164; ABclonal Biotech Co., Ltd.), caspase-1 (1:500; AG-20B-0042; AdipoGen Life Sciences) and β-actin (1:10,000; AC006; ABclonal Biotech Co., Ltd.). Membranes were subsequently incubated with species-matched horseradish peroxidase-conjugated secondary antibodies (1:10,000; cat. nos. SA00001-1 and SA00001-2; Proteintech Group, Inc.). Protein bands were visualized using an enhanced chemiluminescence detection system (Tanon Science and Technology Co., Ltd.) and quantified using ImageJ software.

Publicly available transcriptomic datasets GSE294059 (39), GSE228541 (40) and GSE28750 (41) were obtained from the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) database and analyzed using R software (version 4.5.2; RStudio, Inc.). Samples were classified according to the phenotype information provided in GEO. For GSE294059, sepsis and non-septic control samples were included; for GSE228541, sepsis and healthy control samples were included; and for GSE28750, only sepsis and healthy control samples were retained, whereas post-surgical samples were excluded. Each dataset was analyzed independently without cross-dataset normalization or batch correction.

For the RNA-sequencing datasets GSE294059 and GSE228541, gene-level count matrices were imported into R and matched with the corresponding sample information using GEO accession identifiers. Gene identifiers were converted to official gene symbols using the AnnotationDbi (https://bioconductor.org/packages/AnnotationDbi/) and org.Hs.eg.db (https://bioconductor.org/packages/org.Hs.eg.db/) packages. ERBIN was identified using the official symbol ERBIN, its previous symbol ERBB2IP, and Entrez Gene ID 55914, whereas NLRP3 was identified using its official symbol and Entrez Gene ID 114548. Library sizes were normalized using the trimmed mean of M-values method implemented in the edgeR package (https://bioconductor.org/packages/edgeR/), and normalized expression was expressed as log2 counts per million (log2-CPM; prior.count=1). For the microarray dataset GSE28750, the GEO-provided RMA-normalized expression matrix was used. Probe sets were annotated using the hgu133plus2.db package (https://bioconductor.org/packages/hgu133plus2.db/.), and multiple probes mapping to the same target gene were averaged to obtain gene-level expression values. ERBIN expression was compared between sepsis and corresponding control groups using a two-sided Wilcoxon rank-sum test, with Welch's t-test additionally performed as a sensitivity analysis. Associations between ERBIN and NLRP3 expression were evaluated within the sepsis group using Spearman's rank correlation as the primary analysis, with Pearson's correlation additionally reported. A two-sided P<0.05 was considered to indicate a statistically significant difference.

Statistical analyses were performed using GraphPad Prism version 8.0.1 (Dotmatics). Data are presented as the mean ± SD unless otherwise indicated. Normality was assessed using the Shapiro-Wilk test before analysis. Comparisons between two unpaired groups with normally distributed data were performed using an unpaired Student's t-test, whereas multiple-group comparisons were performed using one-way ANOVA followed by Tukey's post hoc test. For experiments involving genotype and treatment as two independent factors, statistical differences were analyzed using ordinary two-way ANOVA, followed by Šídák's multiple-comparisons test for prespecified biologically relevant comparisons. Non-normally distributed data were analyzed using the Mann-Whitney U test. Survival curves were generated using the Kaplan-Meier method and compared using the log-rank (Mantel-Cox) test. P<0.05 was considered to indicate a statistically significant difference.

To investigate the role of Erbin in SLI, liver tissue and serum samples were collected 24 h after CLP. Histological analysis showed that CLP induced hepatocyte death, inflammatory cell infiltration, disruption of the hepatic vasculature and disorganization of hepatic cords. These pathological changes were more notable in Erbin^−/−+CLP mice than in WT+CLP mice (Fig. 1A and B). Consistently, serum AST, ALT and LDH levels were markedly higher in Erbin-deficient mice after CLP (Fig. 1C and D). The 7-day survival was lower in Erbin^−/−+CLP mice than in WT+CLP mice [20.0% (4/20) vs. 55.6% (10/18)], respectively; Fig. 1E). In addition, liver homogenates from Erbin^−/−+CLP mice contained significantly higher levels of the pro-inflammatory cytokines TNF-α, IL-1β, IL-18, and IL-6 than those from WT+CLP mice (Fig. 1F). Collectively, these findings indicate that Erbin deficiency exacerbates sepsis-induced liver injury.

To determine whether Erbin regulates pyroptosis-associated signaling during SLI, NLRP3, caspase-1 p20 and GSDMD-N expression was examined by western blotting. CLP markedly activated the NLRP3 inflammasome pathway, and this activation was further enhanced in Erbin^−/− mice (Fig. 2C and D). Consistently, immunohistochemical analysis showed increased hepatic NLRP3 expression after CLP, with a greater increase in Erbin-deficient mice (Fig. 2A and B). These findings indicate that Erbin deficiency promotes NLRP3 inflammasome activation during SLI.

Macrophage-hepatocyte crosstalk is increasingly recognized as a notable contributor to SLI progression (42). To investigate the macrophage-intrinsic role of Erbin, BMDMs were stimulated with LPS and ATP to activate the NLRP3 inflammasome and induce pyroptosis (43,44). Erbin deficiency markedly enhanced LPS/ATP-induced expression of NLRP3, caspase-1 p20, and GSDMD-N (Fig. 3A and B). CCK-8 assays, LDH release measurements and Calcein-AM/PI staining further showed reduced cell viability and increased cell death in Erbin-deficient BMDMs after LPS/ATP stimulation (Fig. 3C-F). In addition, LPS/ATP increased IL-6, IL-18, TNF-α and IL-1β mRNA expression, and these increases were further amplified by Erbin deficiency (Fig. 3G). Collectively, these findings indicate that Erbin negatively regulates pyroptotic and inflammatory responses in macrophages.

To evaluate the role of Erbin in mitochondrial injury and oxidative stress during SLI, mitochondrial ultrastructure and ROS production were examined in vivo and in vitro. Transmission electron microscopy revealed that CLP-induced mitochondrial abnormalities, including swelling, loss of cristae and matrix clearing, were more severe in Erbin^−/− mice than in WT mice (Fig. 4A and C). This mitochondrial damage was accompanied by increased hepatic ROS production, as indicated by DHE staining (Fig. 4B and D). Biochemical assays further showed higher MDA levels and lower SOD activity in Erbin^−/−+CLP mice than in WT+CLP mice (Fig. 4E and F). Consistent results were obtained in vitro, where Erbin-deficient BMDMs exhibited greater intracellular ROS accumulation after LPS/ATP stimulation (Fig. 4G and H), together with higher MDA levels and lower SOD activity (Fig. 4I and J).

Excessive ROS promotes the interaction of TXNIP with NLRP3 and thereby activates the inflammasome (45). Because Erbin deficiency exacerbated sepsis-induced oxidative stress, TXNIP expression and its interaction with NLRP3 were examined. TXNIP and NLRP3 protein levels were increased after CLP and were further elevated in Erbin^−/− mice (Fig. 5A and B). Similarly, LPS/ATP increased TXNIP and NLRP3 expression in BMDMs, with greater increases observed in Erbin-deficient cells (Fig. 5C and D). Co-immunoprecipitation further showed that LPS/ATP enhanced the TXNIP-NLRP3 interaction and that Erbin deficiency further strengthened this association (Fig. 5E).

To determine whether mtROS mediates the enhanced pyroptotic response associated with Erbin deficiency, MitoQ, a mitochondria-targeted antioxidant, was used to reduce mtROS. LPS/ATP stimulation increased mtROS levels in BMDMs, and this increase was further amplified by Erbin deficiency; MitoQ pretreatment markedly reduced mtROS accumulation in Erbin-deficient BMDMs (Fig. 6A and B). Mitochondrial membrane potential was subsequently assessed using JC-1 staining. LPS/ATP induced mitochondrial depolarization, which was exacerbated by Erbin deficiency, whereas MitoQ partially restored mitochondrial membrane potential in Erbin-deficient BMDMs (Fig. 6C and D). MitoQ did not alter Erbin expression (Fig. 6E and G), indicating that it acted downstream of Erbin. However, MitoQ partially reduced TXNIP, NLRP3 and GSDMD-N expression in Erbin-deficient BMDMs after LPS/ATP stimulation (Fig. 6F and H). Consistently, Calcein-AM/PI staining and cytokine assays showed that MitoQ attenuated LPS/ATP-induced cell death and inflammatory cytokine production (IL-1β, TNF-α, and IL-18) (Fig. 6I-K). Collectively, these results support a role for mtROS in Erbin deficiency-associated TXNIP/NLRP3 inflammasome activation and pyroptosis.

To further examine the clinical relevance of Erbin, three independent GEO datasets containing samples from patients with sepsis and healthy controls were analyzed. Erbin expression was significantly decreased in patients with sepsis in GSE294059, whereas no significant differences were observed in GSE228541 or GSE28750 (Fig. S2A, C and E). Next, the correlation between Erbin and NLRP3 expression in sepsis samples was examined. No significant correlation was detected in GSE294059 (Pearson r=0.499, P=0.392; Spearman ρ=0.400, P=0.505), GSE228541 (Spearman ρ=0.332, P=0.246) or GSE28750 (Pearson r=−0.431, P=0.213; Spearman ρ=−0.358, P=0.310) (Fig. S2B, D and F). These data indicate that Erbin expression is altered in at least one human sepsis cohort, although this change and its association with NLRP3 expression were not consistent across datasets.