The data of scRNA-seq was collected from the study by Guo et al. (11). Mice were injected with 0.5 mL/kg CCl4 twice a week (CCl4: oil = 1: 4), and data for 3, 6, and 10 weeks were employed. The available fastq data of the ATAC-seq dataset GSE233080 was downloaded from the European Nucleotide Archive (ENA) database. RNA-seq datasets GSE222576 and GSE207855 for chronic stimulation of CCl4; GSE144600 for acute CCl4 stimulation; thioacetamide (TAA), choline-deficient diet with 0.5% DL-ethionine supplement (CDE) damage scRNA-seq dataset GSE200366 (12); and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) damage scRNA-seq dataset GSE148339 were obtained from the Gene Expression Omnibus (GEO) database (13). GSE233080 contains 0-day control and CCl4-injured mice (injected with 2.5 mL/kg of 10% CCl4 twice a week) (14). GSE222576 includes controls and mice injected with CCl4 for 2, 4, 6, 8, and 10 weeks, with an injection condition of 1.6 g/kg twice a week (CCl4: oil = 1: 3) (15). Mice in GSE207855 were injected with 0.5 mL/kg three times a week for 6 weeks (CCl4: oil=1:6) (16). The data of CCl4-treated and control mice in GSE144600 was used, and mice were injected with 5 mL/kg of 10% CCl4 (17). Notably, variations in CCl4 administration protocols exist across datasets. To address this, we standardized units during data processing. Importantly, the cumulative CCl4 dosage correlates with injury severity, whereas differences in dosing schedules or dilution ratios did not confound our assessment of liver injury progression in this study.

Fastp (v0.23.4) was used to process ATAC-seq data, which was then compared and deduplicated through BWA (v0.7.18) and SamBamba (v1.0.1), then sorted by Samtools (v1.20). MACs2 (v2.2.9.1) software was applied for peak calling, and finally, the DiffBind (v3.12.0) package, with the edgeR method and default parameters, was utilized for differential analysis to obtain the peaks of difference between different groups. The above analysis used the default parameters. Visualization was conducted with the software DeepTools (v3.5.5).

We performed quality control (QC) on scRNA-seq data using the Seurat (v5.0.1) package. The screening criteria were nFeature_ RNA > 508, percent.mt < 10, nCount_RNA < 7000, nFeature_RNA < 4000, and nCount_RNA > 2000. The threshold values for these screening criteria were derived from the original literature (11). The decontX (v1.0.0) package was used to remove free RNA, and the NormalizeData function was utilized for standardization. The number of hypervariable genes was set to 3000. In ScaleData, mitochondrial proportion and nCount_RNA was regressed out. The harmony (v1.2.0) package was employed to remove batch effects between samples. 20 principal components (PCs) were used for dimensionality reduction clustering, with a resolution of 0.3. The cells were annotated using markers from the original research in the dataset (11). Seurat and ggplot2 (v3.4.4) packages were applied for visualization.

DEGs in RNA-seq data were identified using the limma (v3.58.1) package (18), while DEGs in scRNA-seq data were analyzed by FindMarkers function of the Seurat package. GSEA was performed through the clusterProfiler (v4.10.0) package (19). The clusterProfiler package was also utilized to analyze DEGs in both RNA-seq and scRNA-seq data, and pathways with significant normalized enrichment score (NES) in RNA-seq and scRNA-seq data were screened (p < 0.05). Visualization was conducted using the enrichplot (v1.22.0) package.

The related genes involved in fatty acid (FA) and glucose metabolism were identified by consulting existing literature, and the differential expression of these genes was analyzed among different groups of hepatocytes using scRNA-seq data. The analysis was conducted using the Seurat package, and visualization was performed through the ggplot2 package.

Previous studies have demonstrated that the transcription factor Zhx2 effects acute liver injury through two primary mechanisms: First, Zhx2 binds to the conserved motif in the 5’-UTR of ETC genes and occupies the ETC gene promoter, directly suppressing ETC gene transcriptional activity; Second, Zhx2 enhances the transcriptional activation of FBXW7, which in turn reduces the protein level of PGC-1α and consequently limits mitochondrial biogenesis (20). Building on these findings, our study further explored the role of Zhx2 in the setting of chronic liver injury, revealing novel mechanistic insights into the mechanisms by which Zhx2 dynamically regulates both metabolic adaptation and fibrotic remodeling during sustained hepatotoxic stress. We analyzed the expression changes of Zhx2 and ETC-related genes in scRNA-seq data. Additionally, we explored and compared the expression changes of Zhx2 and ETC-related genes in scRNA-seq datasets of chronic liver injury induced by other drugs, with our previous analysis.

We identified TFs that may regulate Zhx2 by intersecting ATAC-seq data, DEGs and TFs obtained from the AnimalTFDB database. These TFs were searched in the Cistrome database to examine if they have a peak at Zhx2 locus. The results were visualized using deeptools software.

ScRNA-seq hepatocyte data were analyzed using the ReactomeGSA (v.1.16.1) package to investigate comparative changes in function over different periods, and the results were visualized as heat maps.

The AML-12 mouse hepatocyte cell line was obtained from the American Type Culture Collection (CRL-2254, ATCC, USA) and were cultured in DMEM/F-12 medium (1:1) (HyClone, Thermo Scientific, MA, USA) supplemented with 10% fetal bovine serum (HyClone, Thermo Scientific, MA, USA), 1% ITS Media Supplement (Cat. No. C0341, Beyotime, China) and 40 ng/ml dexamethasone (Cat. No. ST1254, Beyotime, China). Cells were grown in a 37 °C humidified atmosphere containing 5% CO2. AML-12 cells were treated with 10 mM CCl4 (Cat. No. 56-23-5, Macklin, China) or solvent for 24 h, 48 h and 72 h, respectively (16, 17).

AML-12 cells or liver tissue were collected and put into pre-cooled normal saline followed by ultrasonic trituration and centrifugation at 3000 rpm for 10 min to obtain supernatant. Acetyl-CoA in the supernatant was measured by ELISA Kits (Cat. No. ADS-0449M1, Jiangsu Sumeike Biological Technology) according to the manufacturer’s instructions.

Total RNA was extracted from AML-12 cells or liver tissue using the TRIzol (Cat. No. DP424, Tiangen Biochemical Technology) reagent. cDNA was synthesized using the RevertAid First Strand cDNA Synthesis Kit (Cat. No. K1622, Thermo, USA). Subsequently, 50 ng of each cDNA was amplified as a template, and qPCR was performed using a StepOne Plus device (Art. No. 272008342, Life Technologies, USA) with a SYBR^® Premix Ex TaqTM II kit (Cat. No. RR820A, Takara, Japan). All samples were run in triplicate, and β-actin was used as an internal control. Data were quantified using the 2^−ΔΔCt relative quantitative method, normalized to β-actin expression, and expressed as the ratio of Zbtb20 or Zhx2 to β-actin mRNA levels. Primers were designed as follows:

Wild-type C57BL/6 mice (male, 6–8 weeks) were purchased from Vital River Laboratories (Beijing, China) and were housed in a clean, temperature-controlled environment with a 12-hour cycle of light and darkness, with direct access to regular laboratory food and water. The mice were injected intraperitoneally with 20% CCl4 or corn oil (Cat. No. 032-17016, BioDee, China) twice a week at a dose of 2 mg/kg (18). The mice were euthanized by anesthesia overdose at weeks 3/6/10 of modeling. Blood samples and liver tissue were collected for further analysis. The experimental protocols of this work followed the guidelines and regulations of our Ethics Committee and were approved by the Committee of Animal Use for Research and Teaching at The Air Force Medical Center.

To evaluate the hepatic injury at the enzymatic level, serum AST and ALT levels were measured by using an AST (Cat. No. C010-2-1, Nanjing Jincheng, China) or ALT Assay Kit (Cat. No. C009-2-1, Nanjing Jincheng, China) based on manufacturer’s instructions.

Serum samples were collected by centrifugation of whole blood samples at 3000 rpm for 15 min at 4 °C and stored at −80 °C. Liver tissues were fixed in 4% Paraformaldehyde for 72 h, processed histologically, embedded in paraffin blocks, cut to 4 μm sections and stained with H&E or Picrosirius red staining. Individual scores were calculated as the sum of steatosis (0-3), ballooning (0-2) and lobular inflammation (0-3). Sirius red positive area and lipid droplet area were calculated using a BZ-X800 microscope (Keyence, Osaka, Japan) at 20× magnification.

Whole-cell or tissue lysates were separated by SDS-PAGE on denaturing 10% or 12% gels and transferred to polyvinylidene fluoride membranes (Cat. No. ISEQ00010, Millipore, USA) (21). Blots were blocked in 5% milk for 1 h at room temperature and then separately incubated at 4 °C overnight with the following specific primary antibodies: anti-Zbtb20 (Cat. No. sc-515370, Santa Cruz Biotechnology, USA), anti-Zhx2 (Cat. No. ab205532, Abcam, UK) and anti-β-actin (Cat. No. 66009-1-Ig, Proteintech, China). Anti-Zbtb20 and anti-Zhx2 primary antibodies were diluted 1:1000, and anti-β-actin primary antibody was diluted 1:10000. After rinsing, blots were incubated in HRP-conjugated anti-rabbit or anti-mouse (Cat. No. A0239 and A0216, Beyotime, China) secondary antibodies for 1 h at room temperature. Enhanced chemiluminescence (Cat. No. 34094, Thermo, USA) was used to detect immunoreactive bands.

SPSS Statistics 26.0 (IBM, USA) and Prism 8.0 (GraphPad, USA) software was used to analyze the data. All data are shown as Means ± SD. Student’s t-test was used to analyze differences between two groups. One-way ANOVA was used to compare three or more groups. Significant differences are expressed in the figures and figure legends as *P < 0.05. All cell experiments were repeated using three biological replicates. Six mice were used for all animal experiments.

We compared the open peaks between the CCl4-treated group and control groups using ATAC-seq data. The results showed that the chromatin accessibility of hepatocytes treated with CCl4 was significantly higher than that of both the Dimethyl sulfoxide (DMSO)-treated natural recovery group and the untreated control groups ( Figures 1A, B ). Statistical analysis of genes with peak variation multiples revealed that between the CCl4 and control groups, there were 8,573 genes with peak variation multiples greater than 0 (CCl4-Control-H) and 137 genes with peak variation multiples less than 0 (CCl4-Control-L), indicating that the number of genes with increased peaks was much higher than those with decreased peaks. A similar trend was observed between the CCl4 and DMSO groups. However, the number of genes with a peak change multiple greater than 0 and less than 0 was almost equal and less than 40, showing no obvious peak changes between the DMSO and control groups ( Figure 1C ). These results demonstrated that compared to healthy hepatocytes, the chromatin accessibility of hepatocytes treated with CCl4 is significantly increased. Considering the high correlation between chromatin accessibility and histone modifications, we speculate that histone-related regulatory pathways may also be enhanced.

After processing, the scRNA-seq data from samples subjected to chronic CCl4 stimulation were annotated into five distinct cell subpopulations: hepatocytes, HSCs, Kupffer cells, liver sinusoidal endothelial cells (LSECs), and portal fibroblasts ( Figure 2A ). Given the overwhelming proportion of hepatocytes, these cells were excluded from the histogram depicting cell proportions to better highlight the relative changes in the proportions of the other cell subpopulations. The results of cell proportion indicate that with the prolonging of treatment time, the proportion of hepatocytes and HSCs in the control and stimulation groups showed a slight decrease and increase, respectively ( Figures 2A, B ). Usually, CCl4 mice model was used to study liver fibrosis (characterized by a large increase of HSCs) (22), but the cell proportion in these samples of scRNA-seq did not show such a significant change. Additionally, considering the potential impact of using data from fibrotic samples on the results, analysis of HSCs using fibrosis markers Col1a1 and Col1a2 demonstrated a progressive worsening of liver fibrosis over time ( Supplementary Figure S1 ), indicating that the samples had not yet developed fibrosis. Therefore, we considered this dataset could be used for our research on chronic liver injury.

Based on the analysis results of ATAC-seq, we first examined the GO enrichment results of histone-related pathways such as histone acetylation, histone deacetylation, histone methylation, and histone demethylation in scRNA-seq and RNA-seq datasets GSE222576 and GSE207855. In the scRNA-seq data (CCl4-3/6/10-0.1), with the extension of injury time, the NES related to histone acetylation and histone methylation gradually increased, among which the NES related to histone acetylation increased relatively more. In the high concentration group of normal transcription group (CCl4-2/4/6/8/10-0.4), the NES of histone acetylation and methylation pathways showed a trend of increasing first and then decreasing, with the highest value at 4 weeks. To avoid the influence of different sample sources in scRNA-seq and RNA-seq datasets, we also employed a low-concentration RNA-seq dataset (CCl4-6-0.071). The results revealed that the NES of acetylation in the low-concentration CCl4 injection group was higher than that of methylation at 6 weeks, but lower than that in the scRNA-seq data at 6 weeks (CCl4-6-0.1). We also explored common RNA-seq data (CCl4-one-0.5) from a high-concentration acute stimulus, and the results showed that the NES of histone acetylation and methylation pathways was relatively high and enriched ( Figure 2C ).

In order to further analyze the changes of chromatin opening state during hepatocyte injury, we mainly paid attention to “gobp_positive_regulation_of_histone_acetylation” and “gobp_histone_methylation” in enrichment analysis. We investigated the differential expression of 13 core genes associated with the “gobp_positive_regulation_of_histone_acetylation” using the RNA-seq dataset. Among these, only the gene App showed significant differential expression with a logFC value greater than 0. “gobp_histone_methylation” included 55 core genes, but differential expression analysis revealed no significantly DEGs.

We also analyzed the differential gene expression in hepatocytes using scRNA-seq data. Among the 13 core genes enriched in the “gobp_positive_regulation_of_histone_acetylation”, nine genes (App, Auts2, Baz1b, Brca1, Dek, Ep300, Ercc6, Sf3b1, Smarca5) were significantly up-regulated in the CCl4-3-0.1 group (p_val_adj < 0.05, log2FC > 0). In the CCl4-6-0.1 group, four genes (App, Auts2, Ddx21, Ercc6) showed significant differential expression (p_val_adj < 0.05), with Ddx21 and Ercc6 being significantly down-regulated (Log2FC < 0). In the CCl4-10-0.1 group, six genes (App, Auts2, Baz1b, Ddx21, Mybbp1a, Sf3b1) were significantly differentially expressed, with only Sf3b1 being significantly down-regulated ( Table 1 ). Based on these results, we observed that genes involved in histone acetylation exhibited significant changes at different stages of liver damage caused by low-concentration drugs. At 3 weeks, Auts2 had the most significant p-value (3.24E-149), and Sf3b1 had the highest (pct.1 = 0.588) proportion of cells. At 6 weeks, Auts2 again had the most significant p-value (1.79E-31), while at 10 weeks, Mybbp1a had the most significant p-value (1.54E-66) and there was a large change in the proportion of Mybbp1a with a lower proportion of Sf3b1 (pct.1 = 0.464) ( Table 1 ). Additionally, we analyzed the differential expression of 55 genes related to methylation in scRNA-seq data. The results showed that Kansl1 had the most significant difference at 3 weeks and the largest proportion of cells, Xbp1 had the most significant p-value at 6 weeks with a larger proportion of Gcgr, and Setd7 had the most significant difference at 10 weeks with a larger proportion of Kansl1 ( Supplementary Table S1 ).

Furthermore, the expression of histone acetylation writing and erasing genes gradually decreased from high to low as chronic injury progressed. However, unlike at 3 and 6 weeks, the overall expression of acetylation erasure genes was lower than that of writing genes at 10 weeks. This indicated that acetylation in the drug stimulation group was higher than in the control group at 10 weeks, consistent with the increased activity of the histone acetylation pathway observed in our previous enrichment analysis results ( Figure 2D ). In conclusion, in CCl4-induced chronic liver injury cells, histone acetylation-related pathways were more active and hepatocyte chromatin accessibility increased.

GSEA was performed on hepatocytes under different treatment conditions. The results indicated that the functions of hepatocytes are predominantly associated with cell division, immunity, and inflammation. In the scRNA-seq data (CCl4-3/6/10-0.1), we observed a gradual increase in the NES of inflammation-, immunity-, and mitosis-related pathways with the progression of the injection period. Notably, the NES of the mitosis-related pathway demonstrated a significant elevation at 10 weeks, indicating an elevated level of activity. Additionally, it was evident that the relevance of the pathways associated with inflammation, immunity, and proliferation exhibited a heightened significance when the injection concentration of CCl4 increased (CCl4-2/4/6/8/10-0.4) compared to when the injection concentration was low (CCl4-3/6/10-0.1). The results of low-concentration RNA-seq data (CCl4-6-0.071) showed that the enrichment results of low-concentration CCl4 injection at 6 weeks still indicated that the NES of the division-related pathway was high, and the NES of the immune-related pathway was relatively low, consistent with previous findings. The results of RNA-seq data from high-concentration acute stimulation (CCl4-one-0.5) revealed that the NES of the mitotic pathway was higher, similar to that observed in CCl4-10-0.1 ( Figure 3A ).

We subsequently examined the top 10 enriched results in GSEA of the RNA-seq groups, based on GO enrichment. In the acute group CCl4-one-0.5, the NES associated with chromosome separation was elevated, whereas the immune-related score was lower. In the CCl4-2-0.4 and CCl4-4-0.4 groups, the NESs related to immunity were high, but those associated with chromosome segregation were low. The CCl4-6-0.4 group exhibited increased NES related to chromosome segregation, while the CCl4-8-0.4 and CCl4-10-0.4 groups displayed relatively elevated immune-associated NESs. Furthermore, the chromosome-related NES was higher in the CCl4-6-0.071 group ( Figure 3B ).

GO enrichment analysis of DEGs in RNA-seq revealed that the main enriched functions of DEGs were associated with immunity, cell division, and metabolism. In the acute CCl4-one-0.5 group, chromosome separation and cell cycle regulation were the primary terms. The CCl4-2-0.4 group showed a stronger correlation with metabolic processes, while immune system function had a relatively weaker association. In the CCl4-4-0.4 group, immune correlation enrichment increased, whereas metabolic correlation enrichment decreased. In the CCl4-6-0.4 group, metabolism, immunity, chromosome segregation, and cell cycle all appeared as significant terms. In the CCl4-8-0.4 group, the main functions were metabolism and stimulation, with few terms related to immunity. In the CCl4-10-0.4 group, metabolism was the main function, with minor roles for stimulation and immunity. In the CCl4-6-0.071 group, chromosome separation and cell cycle regulation were the primary terms, with some involvement of immune reactions. Based on GO enrichment results, metabolism is required in the process of chromosome separation and immune responses ( Supplementary Figure S2A ).

Subsequently, we examined the top 10 results of GSEA on scRNA-seq data. From the GSEA results, the NES for mitochondrial electron transfer and ribosome small subunit assembly at 3 weeks was less than 0. The regulation of membrane potential and synaptic signal transmission reached their highest levels at 6 weeks, with NES decreasing at 10 weeks. In pathways related to chromosome segregation, the NES was greater than 0 and significant at 3 weeks, low and insignificant at 6 weeks, and the highest and significant at 10 weeks ( Figure 3C ). We then performed GO enrichment analysis on the DEGs from scRNA-seq data, revealing that the top 5 terms primarily focused on cell division and metabolism ( Supplementary Figure S2B ). We employed the cell proliferation marker Mki67 to analyze proliferation in cell subpopulations, finding that proliferation occurred exclusively in hepatocytes. An examination of various stages of hepatocyte injury revealed that Mki67 expression was higher at 3 weeks, lower at 6 weeks, and exhibited an upward trend at 10 weeks. These findings were consistent with the trajectory of change observed in the NES of chromosome segregation-related signaling pathways ( Figure 3D ).

Combining the above results, the open genes appear to be primarily located in those associated with mitotic proliferation and metabolism of hepatocytes. The metabolic process plays a crucial role in CCl4-induced chronic liver injury, particularly focusing on FA metabolism. This is evidenced by the enrichment results from both overall liver tissues and hepatocytes specifically.

Based on the common pathways of FA metabolism, we first explored whether the capacity for external FA absorption was enhanced. We investigated the expression levels of Got2 and Slc27a5, two key genes involved in FA uptake. At 3 weeks, the expressions of both Got2 and Slc27a5 were significantly lower in the injury group compared to controls. By 6 weeks, the expression of Got2 in the injury group had significantly decreased, while by 10 weeks, the expression of Slc27a5 in the injury group had notably reduced ( Figure 4A ). These results suggest that the liver’s ability to absorb external FAs diminishes during CCl4-induced chronic injury.

To investigate whether the capacity for internal de novo FA synthesis is augmented, we assessed the expression levels of key genes involved in this pathway: Acly, Acaca, Acacb, Fasn, Srebf1 and Mlxipl. Our findings revealed that at 3 weeks, the expression levels of Acaca, Acacb, and Mlxipl were notably elevated in the injured group, suggesting an enhanced capability of hepatocytes to convert acetyl-CoA into malonyl-CoA through carboxylation. By 6 weeks, with the exception of Acacb, the expression of the remaining genes in the injured group diminished, implying a reduced FA synthetic potential. Remarkably, at 10 weeks, aside from Mlxipl, there was a substantial upregulation in the expression of other genes within the injured cohort, highlighting that the peak period for de novo FA synthesis occurs at this time point ( Figure 4A ).

To examine the alterations in FA activation capabilities within cells, we conducted an analysis of gene expression for key enzymes involved in this process: Acsl1, Acsl3, Acsl5, Acot11, and Acot13. At the 3-week, a significant decrease was observed in the expression of Acsl5 within the injured group. Conversely, there was a marked upregulation in the expression of both Acsl1 and Acsl3. By week 6, Acsl1 and Acsl3 in the injury group exhibited an elevated expression. At the 10-week, a notable increase in the expression of both Acsl3 and Acsl5 was observed in the injury group. Throughout this chronic injury timeline, the expression patterns of Acot11 and Acot13 remained constant, implying no significant fluctuations in the conversion of acyl-CoA back to free FAs. Notably, the reduced expression of Acsl5 at 3 weeks in the injured group could potentially be interpreted as a compensatory mechanism in response to the heightened expression of Acsl1. Collectively, these results underscore the robust capacity for FA transformation into fatty acyl-CoA throughout the entire observation period ( Figure 4A ).

In the process of FA oxidation, carnitine palmitoyl transferase I (Cpt1a) play a pivotal role by acting on long-chain acyl-CoA, converting it into acylcarnitine, and facilitating its transport into the mitochondria. Our study revealed that the expression level of Cpt1a was elevated in the injury group at both 3 and 6 weeks, but exhibited a decrease by week 10. The expression of the carnitine/acylcarnitine translocase Slc25a20 was found to be notably increased in the injury group at 3 and 6 weeks, followed by a reduction at week 10. Cpt2, which is essential for the conversion of acylcarnitine back to acyl-CoA within the mitochondrial matrix to initiate β-oxidation, showed no significant difference compared to the control group at 3 and 6 weeks; however, its expression was lower in the injury group by week 10. These findings suggest that the capacity for FA oxidation transportation is robust at 3 and 6 weeks under chronic CCl4 stimulation but diminishes by week 10 ( Figure 4B ).

Furthermore, we investigated the expression profiles of long-chain, medium-chain, and short-chain FA dehydrogenases, which are primarily involved in β-oxidation. Results suggest that the capacity for FA β-oxidation is most pronounced during the initial and intermediate stages, peaking at 6 weeks, followed by a decline at 10 weeks. This temporal pattern corresponds with the observed fluctuations in mitochondrial transport efficiency related to FA oxidation ( Figure 4B ).

Overall, our findings indicate that the hepatocytes’ capacity to absorb external FAs diminished across all stages following CCl4-induced chronic injury, while their ability to synthesize FAs exhibited a slight increase at 3 weeks, followed by a decrease at 6 weeks and a significant enhancement by 10 weeks. Acetyl-CoA generation through FA oxidation persisted throughout all stages examined.

Based on the aforementioned analysis, we observed that the expression levels of Srebf1 and Mlxipl were diminished at 6 weeks, indicating a reduced capacity for lipid production and carbohydrate metabolism. Further investigation was conducted on the expression of genes involved in glycolysis, glycogen synthesis, decomposition, and TCA cycle. At 3 weeks, no significant difference was detected in the expression of Gck. However, notable variations were evident in the expressions of Pygl, G6pc, Pfkfb1, and Pklr, which were significantly elevated in the injury group. This suggests an enhancement in glycogen hydrolysis and glucose conversion efficiency, as well as an increased capability to convert phosphoenolpyruvate into pyruvate (regulated by Pklr). Conversely, the expressions of Gys2, Gpil, Aldob, Tpi1, Gapdh, Pgk1 and Pgam1 were substantially decreased in the injury group, implying inhibited glycolysis and reduced glycogen synthesis. By 6 weeks, there was a significant downregulation of Gck expression in the injury group, with most glycolysis-related genes exhibiting lower expression levels; however, glycogen synthesis potential appeared to strengthen while hydrolysis weakened. At 10 weeks, genes associated with glycogen hydrolysis and glycolysis were highly expressed, except for those related to glycogen synthesis, which indicated a weakening ability. These findings collectively suggest that glycolysis became more prominent at 10 weeks ( Figure 4C ).

Following glycolysis, glucose can be further metabolized into FAs through the TCA cycle via de novo lipogenesis (23). Consequently, we investigated the expression of genes associated with the TCA cycle ( Figure 4D ). At 3 weeks, reactions within the TCA cycle were enhanced, as evidenced by increased expression levels of Dlst, Ogdh, Idh, Aco, Cs, Did and Dlat. Conversely, the expression of succinate dehydrogenase (Sdh), which is involved in the conversion of succinate to fumarate, was decreased, along with a reduction in the expression of downstream enzymes Fh and Mdh, indicating diminished activity in subsequent reactions. The expression of the succinate transporter Sucl family remained unchanged, suggesting potential impacts on mitochondrial ETC function. By 6 weeks, the TCA cycle exhibited a suppressed state, whereas at 10 weeks, it was not significantly inhibited.

In summary, the capacity for carbohydrate utilization in hepatocytes chronically damaged by CCl4 progressively increased over time, potentially correlating with enhanced de novo FA synthesis observed at 10 weeks post-injury.

Based on the aforementioned results, we further employed GSEA to examine the expression profiles of genes associated with the ETC. Our findings confirmed that the ETC was indeed inhibited at 3 weeks, as evidenced by reduced expression levels of ETC-related genes in the injured group. However, at 6 and 10 weeks, the expressions of these genes gradually recovered over time within the injured group ( Figure 5B ).

When integrated with our analysis of glucose metabolism, it was observed that the TCA cycle was enhanced up until the succinate stage but subsequently decelerated. Concurrently, both glycolysis and the ETC were found to be suppressed, suggesting the presence of alternative metabolic substrates supporting the TCA cycle. To investigate this possibility, we analyzed amino acid metabolism and discovered that genes involved in branched-chain amino acid metabolism and glutamine metabolism were significantly upregulated in the injured group at 3 weeks ( Table 2 ). These findings imply that the TCA cycle may be sustained, at least partially, through metabolites derived from amino acid catabolism.

We identified Zhx2 as a key regulatory gene in the upstream regulation of the ETC based on literature review. Our analysis revealed that its expression was elevated at 3 weeks. Further exploration using the Cistrome database demonstrated that, within the Chip-seq of H3K27ac in hepatocytes, the peak intensity at the Zhx2 promoter region was notably higher, suggesting a propensity for acetylation. However, at both 6 and 10 weeks, no significant differences in Zhx2 expression were observed ( Figures 5A, B ).

These findings collectively suggest that during the early stages of chronic CCl4-induced liver injury, the mitochondrial ETC is inhibited. As the injury progresses and recovery mechanisms engage, the functionality of the ETC and oxidative phosphorylation gradually approach a near-normal state.

In order to compare with other chronic liver injuries, we examined the expression levels of ETC-related genes and Zhx2 in liver injuries induced by TCDD, TAA and CDE. The results revealed that in chronic injuries caused by both TCDD and TAA, Zhx2 expression was diminished in the injured groups, whereas the expression of ETC-related genes was elevated. In the data of chronic injury induced by CDE, no significant difference was observed in Zhx2 expression, and the expression patterns of different ETC-related genes exhibited inconsistencies ( Figures 5C, D ). Collectively, these data suggest that while Zhx2 expression varies under the influence of different drugs, there exists a consistent negative correlation between the expression of Zhx2 and that of ETC-related genes, aligning with our previous findings.

The heat map of the top 30 pathways revealed a significant difference between the CCl4-3w and the control group. In contrast, both the CCl4-6w and CCl4-10w groups exhibited functional profiles closer to that of the control group, indicating a recovery trend over time ( Figure 6A ). Principal Component Analysis (PCA) further corroborated these findings; PC1 demonstrated that CCl4-3w was markedly distinct from the other groups in terms of function. By CCl4-6w, the function state began to normalize, with CCl4-10w showing even closer resemblance to normal function ( Figure 6B ). These results collectively support the hypothesis that metabolism in chronically injured hepatocytes transitions from an initial inhibitory phase towards a state of metabolic balance as recovery progresses.

A significant difference was observed in the expression of Zhx2 at 3 weeks post-treatment, yet no substantial differences were noted at 6 and 10 weeks ( Figure 6D ). Subsequent analysis of ATAC-seq data revealed that the peak value of Zhx2 (CCl4_1/2.bigwig) following CCl4 administration was notably higher compared to the control group, suggesting an enhancement in chromatin accessibility. However, its expression exhibited a gradual decline with increasing duration of injections ( Figure 6C ).

The increase in Zhx2 chromatin accessibility, coupled with the subsequent inhibition of its expression over time, suggested the potential involvement of multiple TFs interacting with Zhx2. To investigate this further, we compared scRNA-seq data between CCl4-6w and CCl4-3w, as well as between CCl4-10w and CCl4-3w. The differential expression analysis was conducted with a significance threshold set at padj < 0.05 and logFC > 0. We then intersected the resulting DEGs with those identified in the control versus CCl4 comparison to refine our list. Finally, we cross-referenced this refined list with a database of known TFs to identify ten candidate TFs: Zbtb20, Tef, Mafb, Rxra, Egr1, Xbp1, Ppara, Ncor2, Zbtb7a, and Zbtb7c. We subsequently examined the Cistrome database to determine if these TFs had binding sites to Zhx2.

Among the TFs investigated, Zbtb20, Xbp1, and Zbtb7a exhibited peaks at the TSS of Zhx2, with Zbtb20 showing the highest peak. The mouse dataset was utilized for Xbp1, while human data were used for the other two TFs. Additionally, Mafb, Rxra, Ppara, and Ncor2 displayed peaks within the transcription region of Zhx2, with Mafb demonstrating the highest peak. For these four TFs, mouse data were employed. The human data for Egr1 indicated an inconspicuous peak value at Zhx2. Furthermore, Cistrome lacked data for Tef and Zbtb7c. In summary, Zbtb20, Mafb, Rxra, Xbp1, Ppara, Ncor2, and Zbtb7a are likely regulators of Zhx2 expression. The binding sites of Zbtb20, Xbp1, and Zbtb7a were located at the TSS of Zhx2, while the binding sites of Mafb, Rxra, Ppara, and Ncor2 were situated within the transcription region ( Supplementary Figure S3 ).

We observed that Zbtb20 and Rxra exhibited relatively high peak values, suggesting a significant inhibitory effect on transcription. The expression of Zbtb20 was consistently up-regulated in the CCl4-treated injury group, with the up-regulation trend intensifying over time. Conversely, the expression of Rxra initially decreased in the injured group following CCl4 treatment but gradually recovered as time progressed ( Figure 6D ). By integrating the ATAC-seq results with Zhx2 expression data, we hypothesize that Zbtb20 may act as a crucial upstream transcriptional repressor, binding to the TSS of Zhx2, which is negatively correlated with the expression pattern of Zhx2.

To validate our bioinformatic results, changes in acetyl-CoA, Zbtb20 and Zhx2 were first detected after CCl4 treatment of AML12 cells; as shown in Figures 7A, B , CCl4 treatment significantly increased their levels in a time-dependent manner. Subsequently, the chronic liver injury mouse models were established by intraperitoneal injection of CCl4. The degree of liver tissue structure abnormality increased with the increase of CCl4 treatment time; steatosis, the number of hepatocytes with intracytoplasmic vacuoles and edema increased ( Figure 7C ). Pathological indices of the liver such as collagen proportionate area ( Figure 7D ), AST and ALT ( Figure 7E ) also worsened with increasing CCl4-treatment time. We detected changes in acetyl-CoA, Zbtb20 and Zhx2 levels in the chronic liver injury mouse models. Similar to the cellular results, acetyl-CoA ( Figure 7F ), and Zbtb20 ( Figures 7G, H ) continued to rise with increasing treatment time. However, Zhx2 ( Figures 7G, H ) showed an increase in expression only in the week 3, and the expression returned to the normal level after the week 6, consisting with our bioinformatic analysis.