Identification and validation of key PANoptosis-related genes via integrative machine learning and single-cell sequencing in AILI
Shuo Li, Chenhui He, Meng Wang, Binli Ran, Guanghui Li, Guangzhong Dong, Heng Zhang, Chaopeng Mei, Yaodong Song, Qiwen Yu, Sanyang Chen, Changju Zhu, Jiye Li
TL;DR
This study identifies Cdkn1a and Pdk1 as key genes linked to PANoptosis in acetaminophen-induced liver injury, offering new insights into liver disease mechanisms and potential biomarkers.
Contribution
The novel contribution is the identification of Cdkn1a and Pdk1 as PANoptosis-related biomarkers in AILI using integrative machine learning and single-cell sequencing.
Findings
Cdkn1a and Pdk1 are key PANoptosis-related genes with high diagnostic potential for AILI.
Pdk1 knockdown in vivo worsens AILI by promoting PANoptosis.
Immune cell interactions are linked to Cdkn1a and Pdk1 expression patterns.
Abstract
Acetaminophen (APAP) overdose is a leading cause of drug-induced liver injury and acute liver failure. PANoptosis, a recently defined form of programmed cell death, is closely linked to immune regulation; however, its role in APAP-induced liver injury (AILI) remains unclear. Here, we aimed to identify PANoptosis-related biomarkers and elucidate their functions in AILI. By integrating bulk RNA-seq, weighted gene co-expression network analysis, machine learning, and single-cell RNA-seq, we identified Cdkn1a and Pdk1 as key PANoptosis-related genes with high diagnostic potential. Immune infiltration analyses revealed significant associations between these genes and multiple immune cell populations. Single-cell analysis demonstrated cell-type-specific expression patterns and enhanced signaling between hepatocytes and macrophages, as well as between T cells and neutrophils. Experimental…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsInflammasome and immune disorders · Lysosomal Storage Disorders Research · Genomics and Rare Diseases
Introduction
Acetaminophen (APAP) is widely used as a drug to relieve pain and lower fever.1 At therapeutic levels, APAP demonstrates both safety and efficacy. However, due to abuse or accidental overdose, APAP is recognized as the most common cause of drug-induced hepatic toxicity.2 Mechanistic insights into APAP-induced liver injury (AILI) remain incomplete, and effective systemic therapies are limited. Thus, it is imperative to elucidate the regulatory mechanisms underlying AILI, as well as to identify novel treatment targets.
Under therapeutic conditions, the majority of APAP is converted into harmless metabolites through glucuronide and sulfate conjugation, with only a small fraction (5%–10%) oxidized via cytochrome P450 enzymes, generating toxic NAPQI within the liver. Following overdose, however, the accumulation of NAPQI depletes glutathione (GSH) in both the cytosol and mitochondria, resulting in the formation of APAP-protein adducts.3 These toxic metabolites can trigger inflammation and hepatocyte death.4 Immune responses have been shown to be critical in regulating the initiation and resolution of AILI. Multiple immune subsets, including macrophages and neutrophils, migrate into the liver during APAP intoxication and play dual roles in exacerbating injury or promoting tissue repair.5 For example, Kupffer cells (liver-resident macrophages) initially produce inflammatory cytokines to amplify sterile inflammation, but subsequently switch to a reparative phenotype through efferocytosis of dead cells and secretion of anti-inflammatory mediators.6 Similarly, neutrophil infiltration has been linked to collateral tissue damage via reactive oxygen species (ROS) and protease release, while regulatory T cells (Tregs) may attenuate inflammation by suppressing excessive immune activation.5^,^7 Understanding the dynamic crosstalk between immune cells and the hepatic injury microenvironment plays a pivotal role in clarifying the mechanisms underlying AILI.
Innate immune system recognizes and responds to danger signals though releasing pro-inflammatory mediators and initiating programmed cell death (PCD) signaling cascades.8 Complex cross-linked mechanisms among these cell death types have been recently identified. PANoptosis is a newly described inflammatory PCD that exhibits features of apoptosis, pyroptosis, and/or necroptosis, yet none of these processes alone can fully explain the phenomenon.9 Regulated by the PANoptosome, PANoptosis is crucial in a range of pathological contexts, such as malignancies, viral infections, and metabolic dysfunction-associated steatohepatitis (MASH).10^,^11^,^12^,^13 However, research on PANoptosis in AILI remains limited, and integrative analyses directly focusing on this question are still lacking.
In this study, we revealed a close correlation between PANoptosis and AILI, and also identified critical PANoptosis-associated genes together with their possible links to immune cell infiltration. In addition, we investigated pathways potentially regulated by key PANoptosis-related genes using single-cell analysis. We aimed to clarify the involvement of PANoptosis in AILI and discover novel biomarkers using an integrated strategy that combined bioinformatics analyses, single-cell RNA sequencing, and experimental confirmation.
Results
Screening for differential gene expression in AILI and functional enrichment analysis
Our overall experimental flow-chart is shown in Figure 1. We first obtained four bulk RNA sequencing datasets ([GSE51969](GSE51969), [GSE205201](GSE205201), [GSE160732](GSE160732), and [GSE166868](GSE166868)) associated with AILI in mice from the GEO database. We acquired and normalized the merged expression profile of the four datasets, which included 35 disease and 25 control samples from 9 to 24 h after APAP treatment (Figures 2A and S1A). Analysis with the R package limma identified 293 differentially expressed genes (DEGs) in AILI versus control samples, of which 189 were downregulated and 104 were upregulated. To visualize the clustering of DEG expression, volcano maps and heatmaps were constructed with the pheatmap and ggplot2 modules in R (Figures 2B and 2C). GSEA and KEGG results showed enrichment of these genes in pivotal pathways, including drug metabolism by cytochrome P450, chemokine signaling, TNF signaling, and oxytocin signaling (Figures 2D and 2E). GO analysis revealed significant enrichment of these genes in processes involving toxic substance response, inflammation and innate immunity regulation, oxidative stress, chemokine signaling, and immune receptor activity (Figure 2F). These findings indicate that immune response is the core driver of AILI.Figure 1. Flowchart of the researchFigure 2Screening for differentially expressed genes (DEGs) in AILI and functional enrichment analysis(A and B) Volcano plot of DEGs between AILI and control samples.(C) Heatmap of gene expression patterns between AILI and control samples.(D) GSEA enrichment of DEGs.(E) KEGG pathway enrichment analysis of DEGs.(F) GO enrichment analysis of DEGs (containing subsections of biological processes, cellular components, and molecular functions).
Identification of module genes associated with AILI via WGCNA
We examined gene expression networks and modules to uncover the key genes associated with AILI though WGCNA. Outlier samples were removed following hierarchical clustering of all data (Figure 3A). During the WGCNA procedure, the soft threshold β was set to ten (Figure 3B). We specified a minimum of 50 genes per module and set the MEDissThres to 0.3, which resulted in the identification of seven co-expression modules (Figure 3C). Correlation analysis revealed that the gray modules were not significantly associated with AILI and were therefore excluded. As shown in Figure 3D, the turquoise (cor = 0.56, p = 3e-06) and brown (cor = 0.33, p = 0.01) modules exhibited strong positive correlations with AILI. Furthermore, gene significance of members in the two key modules showed strong correlations with module membership (r = 0.61, FDR = 1.8e-160; r = 0.51, FDR = 6.2e-53, respectively) (Figures 3E and 3F). To further characterize the biological functions of the turquoise and brown modules in AILI, GO and KEGG enrichment analyses were performed. Genes within these modules were significantly enriched in pathways related to inflammation, oxidative stress, drug metabolism, and immune responses (Figures S1B and S1C).These findings suggest that the turquoise and brown modules genes may be functionally relevant to AILI.Figure 3. Identification of module genes associated with AILI via WGCNA(A) Sample clustering tree.(B) Soft threshold power and mean connectivity of WGCNA.(C) Cluster dendrogram.(D) Heatmap depicting the relationships between modules and disease traits, specifically in AILI and control samples.(E) Scatterplot between gene significance and module membership in the turquoise module.(F) Scatterplot between gene significance and module membership in the brown module.
Identifying PANoptosis-associated DEGs in AILI and functional annotations
We obtained 1,766 PANoptosis-associated genes from GeneCards database, comprising 1,671 genes linked to apoptosis, 57 to pyroptosis, and 18 to necroptosis, using a relevance score threshold greater than three. Human-mouse homologous conversion of 1,746 PANoptosis-associated genes yielded 1,422 genes. On this basis, we performed an overlap analysis of 293 DEGs, 2,349 WGCNA-associated modular genes, and 1,422 PANoptosis-associated genes and identified 32 hub genes. The overlaps among DEGs, PANoptosis-related genes, and turquoise and brown module genes are illustrated using Venn diagrams (Figure 4A). A PPI network of the 32 hub genes was constructed through the STRING database and displayed in Cytoscape (Figure 4B). A heatmap was generated (Figure 4C), and functional enrichment of the 32 hub genes was conducted for GO terms and KEGG pathways. KEGG pathway analysis showed that PANoptosis-related genes were significantly enriched in several signaling cascades, including PI3K-Akt, IL-17, p53, JAK-STAT, ErbB, AGE-RAGE in diabetic complications, HIF-1, AMPK, Toll-like receptor, and oxytocin pathways (Figure 4D). As shown in Figures 4E and 4F, the significant GO terms among BPs, CCs, and MFs include control of the inflammatory response, ROS metabolic processes, cytokine activity, cytokine-mediated signaling pathways, oxidoreductase complexes, NADPH oxidase complexes and apoptotic signaling. Meanwhile, ssGSEA was performed based on the 32 hub genes, and the top 200 genes most strongly correlated with this gene set were selected for subsequent KEGG and GO enrichment analyses. The top 30 correlated genes are shown in Figure S2A. Enrichment analysis of this expanded gene set (Figures S2B and S2C) revealed pathways that were highly consistent with those identified using the original 32 hub genes, including cellular response to toxic substance, regulation of inflammatory response, and regulation of innate immune response. Collectively, these findings imply that the interaction between PANoptosis-associated genes and inflammation-related pathways is central to AILI.Figure 4. Identification of PANoptosis-related DEGs in AILI and functional annotations(A) Venn diagram showing the intersection of DEGs, module genes, and PANoptosis-associated genes. The overlapping genes were identified as key genes.(B) PPI network constructed via Cytoscape.(C) The 32 key genes identified in AILI are displayed in a heatmap.(D) Chord diagram showing ten significantly enriched signaling pathways in KEGG and the distribution of key genes in each pathway.(E) Bar plot of GO (containing subsections of biological processes, cellular components, and molecular functions) and KEGG terms with significant enrichment.(F) GO enrichment analysis (the top five significant results of BP, CC, and MF enrichment analysis).
Identification of potential biomarkers using diverse machine learning strategies
To predict diagnostic biomarkers of AILI, variable selection and model building were performed using multiple machine learning approaches. [GSE205201](GSE205201) was used as the training cohort, while two independent merged cohorts (validation cohort 1: [GSE51969](GSE51969), [GSE160732](GSE160732), and [GSE166868](GSE166868); validation cohort 2: [GSE173595](GSE173595) and [GSE243679](GSE243679)) were used for external validation. The performance of each algorithm combination was examined in the training cohort through 10-fold cross-validation, and the area under the curve (AUC) was subsequently obtained. The AUC rankings of all algorithm combinations are presented in Figure 5A. Interestingly, we found that the generalized linear model (GLMBoost) + SVM was ranked first. Although the AUC reached 1.00 in the training cohort, the model maintained high discriminative performance in two independent merged validation cohorts, with AUCs of 0.987 and 0.992, respectively. This analysis revealed four key PANoptosis-related genes (Cdkn1a, Ccnd1, Pdk1, and Prodh). In the training cohort, Cdkn1a and Ccnd1 expression was significantly upregulated, whereas Pdk1 and Prodh expression was significantly downregulated in AILI, as shown in Figures 5B–5E. The validation cohort 1 also showed the same trend (Figures 5F–5I). To evaluate the accuracy of these candidate genes in diagnosing AILI, we conducted ROC curve analysis on the training cohort (Figures 5J–5M) and validation cohort 1(Figures 5N–5Q) separately. We observed that Cdkn1a, Ccnd1, Pdk1, and Prodh were effective in distinguishing AILI from control samples, with each AUC >0.88, thereby suggesting high diagnostic potential. Gene expression and ROC analyses in validation cohort 2 are shown in Figures S3A–S3H. Nomograms presented the important predictive value of these four genes, which was further confirmed by calibration and DCA plots (Figures S4A–S4C). Thus, these four PANoptosis-related genes may serve as reliable biomarkers for AILI.Figure 5. Discovery of candidate biomarkers via multiple machine learning methods(A) Multiple combinations of prediction models using 10-fold cross-validation with a ranked AUC index. Training cohort consisted of [GSE205201](GSE205201). Test 1 and test 2 represent two independent merged validation cohorts. Test 1 includes [GSE51969](GSE51969), [GSE160732](GSE160732), and [GSE166868](GSE166868), whereas test 2 includes [GSE173595](GSE173595) and [GSE243679](GSE243679).(B–E) Expression values of four genes shared between AILI and control samples in the training cohort.(F–I) Expression values of four genes shared between AILI and control samples in the validation cohort 1.(J–M) ROC curve of the four shared genes in the training cohort for AILI.(N–Q) ROC curve of the four shared genes in the validation cohort 1 for AILI.
Immune infiltration landscape and associations between PANoptosis-related genes and immune cells
Using the CIBERSORT algorithm, we evaluated immune cell composition and microenvironmental differences in model versus normal groups. Subsequent comparative analysis revealed distinct changes in certain immune cell subsets across the two groups (Figure 6A). In the AILI group, neutrophils, CD8^+^ T cells, Tfh cells, and M1 macrophages were enriched, whereas multiple CD4^+^ T cell subsets (total, naive, and activated memory) together with resting NK cells were reduced (Figure 6B). To further elucidate the functional implications of these differences and the potential interplay among immune cells, correlation analysis was performed. Our research revealed that neutrophils, M1 macrophages, and Tfh cells were all negatively correlated with various CD4^+^ T cell subsets (total, naive, and activated memory) and with resting NK cells (Figure 6C). These correlations highlight the intricate immune dynamics that might contribute to the pathophysiology of AILI, thus emphasizing the need for further exploration into how these specific immune cell alterations impact broader immunoinflammatory processes in AILI.Figure 6. Immune infiltration analysis and correlation analysis between key PANoptosis-related genes and infiltrating immune cells(A) Differences in the relative abundance of 22 infiltrated immune cells between the AILI and control samples.(B) Violin plots indicating the differences in immune cell infiltration between the AILI and control samples. The results revealed that 7/22 immune cells substantially differed between the AILI and control samples.(C) The results of immune cell correlation analysis.(D) Correlations between hub genes and immune cells.
In addition, we analyzed the associations of the four signature genes and infiltrating immune cells (Figure 6D). Among these genes, Ccnd1 exhibited a positive association with resting CD4^+^ memory T cells but an inverse correlation with resting NK cells. Cdkn1a correlated positively with neutrophils and M1 macrophages, whereas negative associations were observed with activated and naive CD4^+^ T cells as well as resting NK cells. In contrast, Pdk1 correlated positively with activated CD4^+^ memory T cells but negatively with neutrophils (Figures 7A–7M). Prodh was not significantly correlated with these immune cells.Figure 7. Immune infiltration analysis and associations with diagnostic signatures(A–M) Association between diagnostic signatures and infiltrating immune cells.
Through the above results, we describe the changes in the overall immune landscape caused by AILI. We propose that key PANoptosis-related genes may be critically involved in this process.
Exploring expression patterns and roles of key PANoptosis-related genes using single-cell sequencing
Next, Cdkn1a, Ccnd1, Pdk1, and Prodh1 were investigated in single-cell sequencing data from mice AILI model ([GSE255834](GSE255834)) to further confirm their expression patterns. A total of 46,304 cells were identified from AILI samples through single-cell analysis with “Seurat”. These cells were classified into nine populations according to marker genes (Figure 8A). We clearly observed a shift in cell composition after APAP challenge, characterized by reduced T cells and increased proportions of macrophages and neutrophils in the AILI cohort relative to controls (Figure 8B). The exact distributions of various subsets in control and AILI groups are also displayed in the UMAP graphic (Figures S5A and S5B). The marker genes were visualized via a bubble chart (Figure 8C). Here, examination of single-cell sequencing data showed that only Cdkn1a and Pdk1 exhibited expression patterns consistent with previous bulk RNA sequencing data; therefore, these two genes were selected for further study.Figure 8. Expression and functional exploration of the key genes in single-cell sequencing data(A) The UMAP plot shows the total sample composition, tissue sources, and cell subtypes.(B) The stacked graph shows the proportion of each type of cell in the control group and the AILI group.(C) Bubble plots of marker gene expression demonstrating the accuracy of the cell annotations.(D) Bubble chart showing the expression of Cdkn1a and Pdk1 in various cells in the control group.(E) Bubble chart showing the expression of Cdkn1a and Pdk1 in various cells in the AILI group.(F) Circle plot and heatmap showing the cell communication weights and numbers of all cell subtypes.(G–J) Receptor‒ligand communication weights between AILI and control samples.
Subsequent analysis focused on expression of the two genes in different cell types. We found that, in the AILI groups, Cdkn1a was upregulated in hepatocytes, endothelial cells, and macrophages, while Pdk1 was downregulated in hepatocytes and T cells when compared with controls (Figures 8D and 8E). We then quantified the intensity and frequency of intercellular communication for all cell types, revealing changes in the AILI cohort relative to controls (Figure 8F). According to CellChat analysis, interactions between T cells and neutrophils exhibited a notable enhancement in both strength and frequency. A similar result was observed between hepatocytes and macrophages. Interactions between macrophages and hepatocytes were mediated mainly through the C3-(Itgam + Itgb2) and C3-C3ar1 pathways. Interactions between T cells and neutrophils were mediated mainly through the Ccl5-Ccr1 pathway (Figures 8G–8J). Hepatocytes, identified as major sender cells in AILI, showed high expression of the ligand C3, which mediates dual signaling to macrophages through two receptor complexes, Itgam/Itgb2 (forming the Mac-1 integrin) and C3ar1. Binding of C3 to Mac-1 activates macrophage phagocytic activity and cytokine production, facilitating the clearance of necrotic cell debris, whereas C3-C3ar1 signaling promotes macrophage chemotaxis and M1 polarization, thereby strengthening their pro-inflammatory response.14^,^15 In addition, the Ccl5-Ccr1 axis mediated communication from T cells to neutrophils. CD4^+^ T cells showed high expression of Ccl5, which binds to Ccr1 on neutrophils and drives their recruitment into the liver parenchyma, contributing to hepatocellular injury and amplifying inflammation in AILI.16^,^17
Thus, our results indicate that Cdkn1a and Pdk1 may participate in AILI through modulation of interactions between hepatocytes and immune cells.
Experimental confirmation of critical PANoptosis-related genes in animal model
The AILI animal model was employed to validate the findings of our bioinformatics analysis. Hematoxylin and eosin (H&E) staining showed markedly enlarged necrotic regions in liver sections of AILI mice relative to controls (Figure 9A). RT-qPCR and western blot analyses showed Cdkn1a upregulation, whereas Pdk1 was downregulated (Figures 9B and 9G). To clarify the relationship between hepatocellular injury severity and key gene expression, we examined the correlation between serum ALT/AST levels and hepatic Cdkn1a and Pdk1 mRNA expression. Cdkn1a expression was positively correlated with both ALT and AST levels, whereas Pdk1 expression was negatively correlated with ALT levels (Figures 9C–9F). In addition, western blot was performed to examine PANoptosis marker proteins in vivo. We detected significantly high levels of P-MLKL, N-GSDMD, and cleaved caspase 3, indicating pronounced activation of the PANoptosis signaling pathway in AILI (Figure 9G). Consistent results were obtained by P21 and PDK1 immunohistochemistry (Figure 9H). These results strongly indicate that Cdkn1a and Pdk1, as PANoptosis-associated genes, may function as regulators of AILI in vivo.Figure 9. Validation of key PANoptosis-related genes in animal models(A) H&E staining of liver tissues from WT and AILI mice (scale bars, 100 μm; n = 5).(B) mRNA expression of Cdkn1a and Pdk1 by RT-qPCR.(C–F) Correlation analyses between hepatic Cdkn1a and Pdk1 mRNA levels and serum ALT and AST levels at 24 h after AILI.(G) Detection and statistical analysis of key PANoptosis-related gene and marker protein expression in liver tissues from WT and AILI mice.(H) Immunohistochemical staining for P21 and PDK1 in liver tissues from WT and AILI mice (scale bars, 50 μm; n = 5). All the data are presented as the means ± SDs. One-way ANOVA with Tukey’s test and a two-tailed Student’s t test were used for statistical analysis. Spearman’s rank correlation was used to assess the associations between relative mRNA expression levels of Cdkn1a and Pdk1 and serum ALT and AST levels. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.
Pdk1 knockdown exacerbates AILI by promoting PANoptosis in vivo
To further investigate the role of Pdk1 in AILI, Pdk1 expression was knocked down by tail vein injection of an AAV-shRNA vector. The experimental design is summarized in Figure 10A, and efficient Pdk1 knockdown was confirmed by western blot analysis (Figure 10B). H&E staining showed that, compared with controls, Pdk1 knockdown exhibited a significantly larger hepatic necrotic area following APAP administration (Figure 10C). Consistent with these histological findings, Pdk1 knockdown markedly increased serum levels of liver injury markers (ALT and AST) as well as multiple proinflammatory cytokines, including TNF-α, IL-1β, and IL-6 (Figures 10D–10F). Together, these results indicate that Pdk1 deficiency exacerbates AILI.Figure 10Pdk1 knockdown exacerbates AILI by promoting PANoptosis in vivoMice were assigned to four groups: control, sh-NC, APAP, and sh-Pdk1 + APAP.(A) Schematic illustration of the experimental design.(B) Western blot analysis confirming efficient knockdown of PDK1 protein in liver tissues.(C) Hematoxylin and eosin (H&E) staining of liver sections and quantification of hepatic necrotic areas. Scale bars, 100 μm; n = 5.(D–E) Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels.(F) Serum levels of TNF-α, IL-1β, and IL-6 were measured by ELISA.(G) Western blot analysis and quantification of PANoptosis marker proteins in liver tissues.(H–K) Representative immunofluorescence staining of liver sections showing albumin (ALB, green) and PANoptosis marker proteins (ZBP1, p-MLKL, cleaved caspase-1, and cleaved caspase-3; red). Scale bars, 20 μm; n = 5. Data are presented as mean ± SD. One-way ANOVA with Tukey’s test and a two-tailed Student’s t test were used for statistical analysis. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.
To determine whether Pdk1 modulates APAP-induced PANoptosis, the expression of PANoptosis marker proteins was examined by western blot. Compared with the control group, Pdk1 knockdown showed increased expression of ZBP1 (a key mediator of PANoptosis), cleaved caspase-3 (apoptosis), cleaved caspase-1 and N-GSDMD (pyroptosis), and P-MLKL (necroptosis) (Figure 10G). Consistently, immunofluorescence staining further demonstrated significantly elevated expression of ZBP1, cleaved caspase-3, cleaved caspase-1, and P-MLKL in Pdk1 knockdown group compared with controls (Figures 10H–10K). Collectively, these results demonstrate that Pdk1 knockdown exacerbates AILI by promoting PANoptosis.
Discussion
Drug-induced liver injury (DILI), induced by various widely used medications, represents a worldwide health concern. The leading cause of DILI is APAP overdose.18 Excessive APAP intake causes acute liver toxicity or even hepatic failure via mechanisms such as the immune response, mitochondrial dysfunction, oxidative stress, and the formation of metabolic protein adducts.3^,^19 Earlier studies indicated that AILI could activate multiple hepatocyte death pathway, including pyroptosis, apoptosis, autophagy, and necroptosis.20^,^21^,^22 PANoptos is considered a distinct type of regulated cell death that integrates pyroptotic, apoptotic, and necroptotic signaling, and is closely linked to inflammation.23^,^24 The involvement of PANoptosis in AILI has yet to be fully clarified. Therefore, we employed public databases to explore the roles of PANoptosis-associated genes as well as immune infiltration in AILI.
Here, we systematically analyzed multiple transcriptomic datasets of AILI and control liver tissues and detected 293 DEGs. Next, using PANoptosis genes derived from GeneCards together with WGCNA analysis, we identified 32 AILI-related genes. Subsequent functional enrichment highlighted biological processes such as programmed inflammatory response, ROS metabolism, cytokine activity and signaling, oxidoreductase and NADPH oxidase complexes, as well as apoptotic signaling regulation. At the pathway level, enrichment analysis further revealed PI3K-Akt, IL-17, p53, JAK-STAT, ErbB, AGE-RAGE (diabetic complications), HIF-1, AMPK, Toll-like receptor, and oxytocin signaling as the most significantly involved. These results emphasize the central role of PANoptosis-associated genes in modulating immune and inflammatory responses in AILI.
By employing multiple machine learning models, we revealed the four genes most critical for AILI, of which Cdkn1a and Ccnd1 were upregulated and Pdk1 and Prodh were downregulated. Analysis of ROC curves showed AUC values above 0.88 for all these genes, suggesting excellent diagnostic efficacy. P21 is a central regulator of cell cycle control and a key mediator of the p53-p21-RB signaling pathway. Upon activation by stimuli such as DNA damage, p53 induces p21 expression, which inhibits RB phosphorylation and triggers cell-cycle arrest.25 PDK1 (PDPK1) is a critical regulatory kinase in the PI3K/AKT/mTOR signaling pathway that directly phosphorylates AKT at T308 and cooperates with mTORC2-mediated S473 phosphorylation to achieve full AKT activation, thereby regulating cell survival, proliferation, growth, and metabolism.26 CCND1 is a core regulator of the G1-S phase transition and exists as two functionally distinct isoforms: CCND1a, which promotes cell proliferation through CDK4/6-mediated RB phosphorylation, and CCND1b, which fails to efficiently phosphorylate RB and is associated with cell-cycle arrest or apoptosis.27 PRODH/POX is a key enzyme in proline metabolism, and its primary biological function is to regulate downstream signaling by generating ROS in the mitochondrial matrix or intermembrane space.28 Collectively, these four genes are involved in key biological processes, including cell death, cell cycle regulation, oxidative stress, and inflammatory responses, which are closely associated with the core mechanisms of AILI. According to the results of the single-cell data analysis, only Cdkn1a and Pdk1 showed the same trend as in the transcriptome sequencing analysis. Hence, our results suggest that Cdkn1a and Pdk1 may represent novel candidate biomarkers for AILI, warranting further investigation.
Cyclin-dependent kinase inhibitor 1A (p21) is encoded by Cdkn1a.25^,^29 PANoptosis-associated genes such as Cdkn1a, Gadd45b have been reported to be essential for the pathogenesis of aortic dissection.30 In diabetic wounds, activation of the p53-p21-RB axis can promote PANoptosis.31 According to the literature, Cdkn1a exacerbated AILI by influencing pathways linked to oxidative stress and inflammatory signaling.32^,^33 On the basis of our current results, we speculated that Cdkn1a is involved in AILI through its regulation of PANoptosis. PDK1 functions as a central upstream kinase that regulates the activation of several AGC family members, including AKT, SGK, RSK, PLK, and S6K.34 Evidence indicated that, in the context of intervertebral disc degeneration, HIF-1α exerted a protective effect on nucleus pulposus cells against oxidative insult through sustaining mitochondrial homeostasis as well as glycolytic activity via Pdk1.35 In recent years, gain-of-function mutations in Pdk1 have been shown to drive prostate tumor growth though the PDK1-AKT axis activation.36 However, Pdk1’s contribution to AILI is still not fully understood. Our findings indicated that Pdk1 was closely associated with AILI, highlighting the need for further studies on its exact function and mechanisms.
Immune and inflammatory processes are central to AILI.37 APAP overdose triggers immune cell infiltration into the liver parenchyma, notably involving macrophages, T cells, and neutrophils.38^,^39 PANoptosis is also strongly associated with immune effects.40 Through immune infiltration analysis, we gained insight into the alterations of immune cell populations between APAP-treated and control groups. Alterations in several immune subsets—including CD8^+^ T cells, M1 macrophages, and neutrophils—were found significantly associated with AILI. These results corroborate previously published data.41^,^42 Moreover, immune-related analysis further indicated that Cdkn1a and Pdk1 could affect a variety of immune cells. Our data suggested that elevated Cdkn1a was positively associated with M1 macrophages and neutrophils, while showing inverse correlations with activated memory and naive CD4^+^ T cells. Pdk1 displayed a positive association with activated CD4^+^ memory T cells but an inverse relationship with neutrophils. According to the literature, the p21-activated secretory phenotype of MEFs can attract macrophages and further promote M1 polarization.43 p21 mediates the proinflammatory reprogramming of phagocytic macrophages by repressing SIRPα transcription in murine models of T-ALL.44 PDK1 kinase is essential for conventional T cell differentiation and Treg survival by modulating redox homeostasis.45 Activation of PDK1 leads to PKA phosphorylation, which in turn phosphorylates and inhibits RhoA, a critical regulator of neutrophil phagocytosis.46 Taken together, the data imply that Cdkn1a and Pdk1 contribute significantly to AILI, possibly via regulation of the immune subsets described above.
We further mapped hub gene expression patterns using the scRNA-seq dataset. Cdkn1a was expressed mainly in hepatocytes, endothelial cells and macrophages, whereas Pdk1 was expressed predominantly in hepatocytes and T cells. CellChat is an important analytical method in single-cell research and is used to observe the interactions between cells.47 CellChat analysis demonstrated significantly elevated C3-C3ar1 and C3-(Itgam + Itgb2) communication frequencies between hepatocytes and macrophages in APAP-treated samples compared with controls. The interactions between T cells and neutrophils via Ccl5-Ccr1 showed also a significant increase. Interestingly, the involvement of these immune cells in AILI was also confirmed by our immune infiltration analysis described above. Previous studies have shown that APAP triggered C3/C3aR signaling, enhancing STAT3 and c-Fos expression and phosphorylation in Kupffer cells, thereby causing intrahepatic hemorrhage and eventual hepatocyte necrosis.15 The Ccl5-Ccr1 pathways is also of vital importance in the immune response and inflammatory regulation, characterized by immune cell recruitment (e.g., T cells), secretion of mediators such as TNF-α and CCL3, and the promotion of phagocytosis.48^,^49 Collectively, these findings indicate that Cdkn1a may aggravate AILI progression through activation of C3-C3ar1 and C3-(Itgam + Itgb2) signaling between macrophages and hepatocytes; Pdk1 may alleviate AILI by regulating the Ccl5-Ccr1 pathway between T cells and neutrophils.
Finally, experiments were performed to verify these analyses. We set two time points, 12 and 24 h, to establish the APAP-induced hepatotoxicity mouse model. H&E staining verified the effective construction of the model. The expression of p21 and PDK1 was further validated by immunohistochemistry and western blot in vivo. We also examined several PANoptosis marker proteins by western blot, and our results demonstrated that PANoptosis was strongly activated in APAP-induced hepatotoxicity. Moreover, the expression levels of Cdkn1a and Pdk1 were significantly correlated with liver injury severity in AILI. While Cdkn1a (p21) has been previously reported in AILI,33 the contribution of Pdk1 remains unclear; therefore, we selected Pdk1 for further functional validation. Pdk1 knockdown markedly exacerbated AILI, as evidenced by a larger hepatic necrotic area, higher serum ALT and AST levels, and increased production of proinflammatory cytokines. In addition, western blot and immunofluorescence analyses provided direct evidence that Pdk1 knockdown markedly aggravates PANoptosis in AILI. Collectively, these results indicate that the findings of our bioinformatic analysis are reliable and meaningful.
In conclusion, this work integrated single-cell sequencing with machine learning to characterize the significance of PANoptosis-associated genes in AILI together with their influence on the immune landscape. Our findings reveal that PANoptosis-related genes are highly associated with AILI. Moreover, the key PANoptosis-related genes Cdkn1a and Pdk1 are likely to serve as critical regulators of the immune response in APAP-induced hepatotoxicity. The expression levels of Cdkn1a and Pdk1 are significantly correlated with the severity of AILI, and Pdk1 knockdown exacerbates AILI by promoting PANoptosis in vivo. Overall, the study highlights new perspectives on AILI pathogenesis, with Cdkn1a and Pdk1 emerging as promising diagnostic and therapeutic candidates.
Limitations of the study
Some limitations of this work should be acknowledged. First, although multiple murine AILI datasets were included for integrated analysis, no suitable human datasets were available for inclusion. Second, the mechanisms underlying the crosstalk between hepatocytes and immune cells in AILI are complex and remain to be fully elucidated. Despite these limitations, the present study provides a novel perspective by characterizing immune cell infiltration and outlining the landscape of hepatocyte-immune cell interactions in AILI in the context of PANoptosis. Moreover, Cdkn1a and Pdk1 were identified as key PANoptosis-related genes with strong diagnostic potential. Their expression levels were significantly correlated with AILI severity, and in vivo knockdown of Pdk1 exacerbated AILI by promoting PANoptosis, thereby providing preliminary functional support for our bioinformatics-based findings. Future studies will focus on elucidating the molecular mechanisms by which Cdkn1a and Pdk1 regulate PANoptosis in AILI and evaluating their translational potential.
Resource availability
Lead contact
Further details and resource inquiries should be directed to and will be fulfilled by the lead contact, Jiye Li ([email protected]).
Materials availability
No new materials were generated in this study.
Data and code availability
- •Only publicly accessible datasets were analyzed in this work, and their details provided in the key resources table.
- •The research did not involve the creation of original code.
- •The raw data supporting this study are available at Mendeley Data: https://doi.org/10.17632/mg6rvwh8bs.1. Further inquiries related to data reanalysis should be directed to the lead contact.
Acknowledgments
This study was funded by the 10.13039/501100001809National Natural Science Foundation of China (82300688); the Henan Provinciale Medical Science and Technology Research Project (LHGJ20230232); the “Three 100” Program of the Henan Academy of Medical Sciences (HNMOT2024044); and the Outstanding Young Scientists Fund of the Henan Provincial Natural Science Foundation (252300421110). The funding agencies had no role in the study design, data interpretation, manuscript preparation, or publication decision. We thank Yongxing Yu for generously sharing his experience and code.
Author contributions
S.L. and C.H. analyzed data, performed experiments, and prepared the manuscript. M.W., B.R., G.L., G.D., H.Z., C.M., Y.S., Q.Y., and S.C. collected data, performed experiments, reviewed and revised the draft. C.Z. and J.L. conceived the study, oversaw the experiments, and verified the accuracy of the draft. All authors reviewed and approved the final manuscript.
Declaration of interests
The authors state that there are no competing interests.
STAR★Methods
Key resources table
REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesp21 Polyclonal antibodyProteintechCat No.28248-1-AP; RRID: AB_2881097PDK1 Polyclonal antibodyProteintechCat No.18262-1-AP; RRID: AB_10598310MLKL Monoclonal antibodyProteintechCat No.66675-1-lg; RRID: AB_2882029ZBP1 antibodySantaCat No. sc-271483; RRID: AB_10650130Cleaved-Caspase1 antibodyAffinityCat No. AF4022; RRID: AB_2845464Cleaved-Caspase3 antibodyProteintechCat No. 25128-1-AP; RRID: AB_3073913Caspase1 antibodyProteintechCat No. 22915-1-AP; RRID: AB_2876874Phospho-MLKL (S345) Rabbit Monoclonal AntibodyHUABIOCat No.ET1705-51; RRID: AB_3070579Caspase-3 AntibodyCell Signaling TechnologyCat No.9662; RRID: AB_331439β-Actin (13E5) Rabbit mAbCell Signaling TechnologyCat No.4970; RRID: AB_2223172Anti-GSDMD antibodyAbcamCat No. ab219800; RRID: AB_2888940Chemicals, peptides, recombinant proteins and reagentAcetaminophenMCECAS No.: HY-66005TRIzolInvitrogenCat No.15596018HI Script III RT SuperMixVazymeCat No. R323-01SYBR Green PCR master mixVazymeCat No. Q711-02RIPA lysis bufferBeyotimeCat No. P0013BCritical commercial assaysAlanine aminotransferase Assay KitNanjing Jiancheng Bioengineering InstituteCat No.C009-2-1Aspartate aminotransferase Assay KitNanjing Jiancheng Bioengineering InstituteCat No. C010-2-1Mouse IL-1 beta ELISA KitProteintechCat No. KE10147Mouse IL-6 ELISA KitProteintechCat No. KE10007Mouse TNF-alpha ELISA KitProteintechCat No. KE10002BCA kitSolarbioCat No. PC0020Deposited dataRNA-seqGEO database[GSE51995](GSE51995)RNA-seqGEO database[GSE205201](GSE205201)RNA-seqGEO database[GSE160732](GSE160732)RNA-seqGEO database[GSE166868](GSE166868)Single-Cell RNA-seqGEO database[GSE255834](GSE255834)Experimental models: Organisms/strainsC57BL/6J miceBeijing Vital River Laboratory Animal Technology Co., Ltd.N/AOligonucleotidesCdkn1a forward: GCAGAATAAAAGGTGCCACAGGTsingke BiotechN/ACdkn1a reverse: AAAGTTCCACCGTTCTCGGGTsingke BiotechN/APdk1 forward: GGATTACTTTATAGACCGGGTCAGTsingke BiotechN/APdk1 reverse: GTACGGATGGGGTCCTGAGATsingke BiotechN/Aβ-actin forward: TGAGCTGCGTTTTACACCCTTsingke BiotechN/Aβ-actin reverse: CGCCTTCACCGTTCCAGTTTTsingke BiotechN/ASoftware and algorithmsImageJImageJ Softwarehttps://imagej.net/Graphpad Prism 9GraphPad Softwarehttps://www.graphpad.com/R StudioThe R foundationhttps://cran.r-project.org/Adobe IllustratorAdobehttps://www.adobe.com/Cytoscape (v3.9.0)Cytoscape Softwarehttps://cytoscape.org/
Experimental model and study participant details
Animal experiments
Male C57BL/6J mice (6–8 weeks, 20 ± 2 g) were supplied by Beijing Vital River Laboratory Animal Technology Co., Ltd. and maintained in SPF facilities with free access to food and water. All experimental protocols were approved by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University (No. 2022-KY-1402-001). Following a 12-h fasting period, mice were randomly divided to control or AILI groups. AILI was triggered through intraperitoneal injection of acetaminophen (300 mg/kg), while the control group received an equal volume of saline. Animals were sacrificed at 12 and 24 h post-injection, and liver samples were collected for subsequent experiments.
Method details
Construction of adeno-associated viral vectors
Adeno-associated viral (AAV) vectors were designed, validated, and synthesized by OBiO (Shanghai, China), including both the inhibitory constructs and their corresponding negative controls. The AAV interference vector used in this study was GL3009 pAAV-TBG-ZsGreen1-WPRE, driven by the hepatocyte-specific TBG promoter. The target sequence for shRNA construction was CAATTAGAATGCTACTCAA (19 bp). The resulting recombinant virus was pAAV-TBG-ZsGreen1-miR30-shRNA (Pdk1)-WPRE, and the corresponding negative control was Y23847 pAAV-TBG-ZsGreen1-miR30-shRNA (NC)-WPRE. Each mouse received a tail-vein injection of either sh-NC or sh-Pdk1 AAV at a dose of 2 × 10^11^ viral genomes (vg), and the AILI model was established four weeks after AAV administration.
Data acquisition and preprocessing
All genomic data are publicly accessible via the NCBI Gene Expression Omnibus (GEO). Four mouse liver tissue datasets ([GSE51969](GSE51969), [GSE205201](GSE205201), [GSE160732](GSE160732), and [GSE166868](GSE166868)) containing 25 normal samples (control group) and 35 AILI samples (experimental group) were selected. All samples analyzed in this study were mouse liver tissues collected 9-24 hours after intraperitoneal injection of acetaminophen (APAP) at a dose of 300 mg/kg. The meta-cohort comprised four publicly available datasets: Dataset 1 ([GSE51969](GSE51969)), including 5 control samples and 5 APAP-treated samples; Dataset 2 ([GSE205201](GSE205201)), including 10 control samples and 10 APAP-treated samples; Dataset 3 ([GSE160732](GSE160732)), including 5 control samples and 10 APAP-treated samples; and Dataset 4 ([GSE166868](GSE166868)), including 5 control samples and 10 APAP-treated samples. Similarly, we acquired a dataset ([GSE255834](GSE255834)) from single-cell RNA sequencing. For diagnostic feature screening and prediction assessment in AILI, we extracted PANoptosis-related genes from the GeneCards database.
Differentially expressed gene identification and functional enrichment
We combined the [GSE51969](GSE51969), [GSE205201](GSE205201), [GSE160732](GSE160732), and [GSE166868](GSE166868) datasets into a meta-cohort (merged dataset), which included 35 AILI samples and 25 normal controls. Raw expression data were normalized using the “normalizeBetweenArrays” function from the limma package. Batch effects arising from different datasets were corrected using the ComBat algorithm implemented in the sva package, with dataset origin specified as the batch variable while preserving biological group information (AILI vs. control). Principal component analysis (PCA) and uniform manifold approximation and projection (UMAP) were applied to evaluate sample distribution and to assess the effectiveness of batch correction. Differential expression analysis between AILI and control samples was performed using limma on the batch-corrected merged dataset, with adjusted P < 0.05 and |log2FC| > 0.75 as cutoffs. Visualization of DEGs was conducted using volcano plots and heatmaps generated with the ggplot2 and pheatmap packages. To explore the biological functions of DEGs, enrichment analyses including GSEA, KEGG, and GO were performed using the clusterProfiler package.
Weighted correlation network analysis (WGCNA)
We applied the R package WGCNA to identify biologically meaningful coexpression modules and to assess their associations with AILI. To construct a scale-free network, multiple soft-thresholding powers (β) were evaluated, and a soft threshold of 10 was selected. Module detection was performed using the dynamic tree cut algorithm, with a minimum module size of 50 genes. Modules with high similarity were subsequently merged using dynamic hybrid tree cutting with a threshold of 0.05. Module–trait relationships were evaluated by Pearson correlation analysis between module eigengenes (MEs) and sample traits. Statistical significance was assessed using Student’s t-tests based on the correlation coefficients, and multiple testing was controlled by applying false discovery rate (FDR) correction.
Identification of PANoptosis-related DEGs and enrichment analysis
DEGs and key module genes were overlapped with PANoptosis-related genes using a Venn diagram, yielding 32 core candidates associated with AILI. To investigate the functional roles of the 32 core genes in AILI, enrichment analyses were carried out with clusterProfiler (version 3.14.3) using GO and KEGG datasets.
Building the protein–protein interaction (PPI) network
The interactions of PANoptosis-related genes were characterized using STRING database. For this study, Mus musculus was chosen as the reference organism. Subsequently, Cytoscape (v3.9.0) was employed to construct and visualize PPI networks.
Integrative machine learning–based screening of diagnostic genes
For constructing highly predictive diagnostic models, the [GSE205201](GSE205201) cohort was divided into a training set, and the combined data were used as a validation set. Ten machine-learning algorithms—lasso, ridge, elastic net (ENET), random forest (RF), stepwise generalized linear model (StepGLM), gradient boosting machine (GBM), support vector machine (SVM), XGBoost, GLMBoost, and naïve Bayes—were implemented using the training dataset. Gene signature selection and model building were performed using tenfold cross-validation, and model performance was subsequently evaluated in both the training and validation sets.
To ensure fair comparison across models, training and validation datasets were harmonized prior to analysis, and features were standardized within each dataset to prevent information leakage. Regularized regression models (lasso, ridge, and ENET) were trained using cross-validation–based selection of regularization parameters. Tree-based and boosting models, including RF, GBM, XGBoost, and GLMBoost, were trained using established cross-validation or stopping-criteria–based strategies. StepGLM was constructed using information criterion–based stepwise selection, whereas SVM and naïve Bayes classifiers were implemented with default parameter settings.
Model performance was assessed using the area under the receiver operating characteristic curve (AUC), with 95% confidence intervals calculated for diagnostic accuracy. Comparative performance across models was summarized using AUC heatmaps. Based on integrated model performance, four prioritized diagnostic candidates—Cdkn1a, Ccnd1, Pdk1, and Prodh—were identified. The expression levels of these genes in the training and validation cohorts were visualized using box plots. All analysis code used in this study is available in a publicly accessible GitHub repository (https://github.com/mikelu1997/code).
Immune infiltration analysis
Immune infiltration was analyzed with CIBERSORT, which calculated the composition of 22 immune cell populations per sample based on the LM22 signature (100 permutations). Bar plots were generated to show cell distributions, and correlation heatmaps were produced with corrplot to display intercellular relationships. Comparisons between control and AILI groups, as well as correlations of PANoptosis-related genes with immune cell subsets, were visualized using ggplot2.
Retrieval and processing of single-cell transcriptome sequencing data
Single-cell RNA-seq data were retrieved from GEO (accession: [GSE255834](GSE255834)). Seurat (v4.4.1) was employed for scRNA-seq data processing. Cells with mitochondrial gene expression <30% combined with UMI counts between 300–7000 were filtered and retained, yielding 3000 highly variable genes for subsequent steps. Differential expression analysis across clusters was performed with “FindAllMarkers” and Wilcoxon test. Cell populations were annotated using canonical marker genes. Finally, we performed cell interaction analyses via CellChat.
scRNA-seq data were filtered to retain high-quality cells (genes detected in ≥3 cells, ≥200 genes per cell, and mitochondrial gene ratio <10%) and normalized using Seurat’s NormalizeData function. Cell clusters identified from the scRNA-seq data were defined as communication senders and receivers, and the average gene expression of each cluster was used to construct the communication matrix. A CellChat object was generated using createCellChat with CellChatDB.mouse as the ligand–receptor (LR) reference database. Communication probabilities of LR interactions were calculated using computeCommunProb (raw.use = FALSE, nboot = 1000). Significant LR interactions were retained using filterCommunication. Downstream visualizations were generated using CellChat’s visualization functions. Only LR interactions with a communication probability > 0.05 and bootstrap P < 0.05 were included.
Histological analysis
Liver samples were fixed in 10% neutral-buffered formalin to preserve morphology, embedded into paraffin, and sectioned at a slice thickness of 4 μm. Hematoxylin–eosin (HE) staining was performed to visualize hepatic architecture and assess hepatocellular necrosis. For immunohistochemical (IHC) analysis, paraffin-embedded sections were processed and treated with primary antibodies targeting P21 and PDK1, followed by secondary antibody detection and chromogenic development.
RT‒qPCR
TRIzol reagent was employed for total RNA extraction. Following quality control, cDNA synthesis was performed with HI Script III RT SuperMix. qPCR amplification utilized SYBR Green PCR Master Mix, with β-actin as the reference gene. Amplification settings included 95 °C for 30 s, then 40 cycles at 95 °C for 10 s and 60 °C for 30 s.
Western blot analysis
Western blot analysis was conducted to detect target proteins and PANoptosis markers. Total proteins from liver tissues were extracted using RIPA lysis buffer, and the protein concentration was quantified with a BCA assay kit. Equal amounts of protein were separated by SDS–PAGE and subsequently incubated overnight at 4 °C with specific primary antibodies. Following secondary antibody incubation, bands were visualized using an imaging system.
Immunofluorescence staining
Liver tissues were fixed in 4% paraformaldehyde at room temperature for 30 min, permeabilized with 0.1% Triton X-100 for 5 min, and blocked with 5% bovine serum albumin at room temperature for 30 min. Primary antibodies against Cleaved Caspase-1, phosphorylated MLKL (p-MLKL), Cleaved Caspase-3, albumin (ALB), and ZBP1 were incubated with the samples at 4 °C overnight, followed by incubation with fluorescently conjugated secondary antibodies for 1 h in the dark. Nuclei were counterstained with DAPI for 10 min, and fluorescent images were acquired using a Nikon A1R confocal laser scanning microscope.
Measurement of liver function
Blood samples were collected and centrifuged at 3,000 × g for 5 min, and the supernatant was harvested as serum for subsequent analysis. Serum levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were measured using commercially available assay kits according to the manufacturer’s instructions.
Enzyme-linked immunosorbent assay (ELISA)
Serum concentrations of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α) were quantified using commercially available ELISA kits according to the manufacturer’s instructions.
Quantification and statistical analysis
Statistical analysis
The data were processed using SPSS 21.0 and GraphPad Prism 9. Group comparisons were carried out through Student’s t-test or one-way ANOVA followed by Dunnett’s multiple comparison test. Nonparametric tests were employed for data deviating from normality. Spearman’s rank correlation was used to assess the associations between relative mRNA expression levels of Cdkn1a and Pdk1 and serum ALT and AST levels. All statistical procedures are described in the figure legends, with results considered significant when p < 0.05.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Chidiac A.S.Buckley N.A.Noghrehchi F.Cairns R.Paracetamol (acetaminophen) overdose and hepatotoxicity: mechanism, treatment, prevention measures, and estimates of burden of disease Expert Opin. Drug Metab. Toxicol.19202329731710.1080/17425255.2023.222395937436926 · doi ↗ · pubmed ↗
- 2Mohan S.Bloom J.Kerester S.Hoffman R.S.Su M.K.An international survey of the treatment of massive paracetamol overdose in 2023 Clin. Toxicol.61202396897310.1080/15563650.2023.228692238112311 · doi ↗ · pubmed ↗
- 3Jaeschke H.Ramachandran A.Acetaminophen Hepatotoxicity: Paradigm for Understanding Mechanisms of Drug-Induced Liver Injury Annu. Rev. Pathol.19202445347810.1146/annurev-pathmechdis-051122-09401638265880 PMC 11131139 · doi ↗ · pubmed ↗
- 4Zhang Q.H.Jin L.M.Lin M.S.Wang M.X.Cui Y.Q.Ye J.X.Xiong Y.Q.Luo W.Zhu W.W.Liang G.FNDC 5/Irisin exacerbates APAP-induced acute liver injury through activating JNK/NF-kappa B and inflammatory response Acta Pharmacol. Sin.4620251946195710.1038/s 41401-025-01509-740016523 PMC 12205071 · doi ↗ · pubmed ↗
- 5Qian Y.Zhao J.Wu H.Kong X.Innate immune regulation in inflammation resolution and liver regeneration in drug-induced liver injury Arch. Toxicol.99202511512610.1007/s 00204-024-03886-039395921 · doi ↗ · pubmed ↗
- 6Yan M.Huo Y.Yin S.Hu H.Mechanisms of acetaminophen-induced liver injury and its implications for therapeutic interventions Redox Biol.17201827428310.1016/j.redox.2018.04.01929753208 PMC 6006912 · doi ↗ · pubmed ↗
- 7Yang T.Wang H.Wang X.Li J.Jiang L.The Dual Role of Innate Immune Response in Acetaminophen-Induced Liver Injury Biology 112022105710.3390/biology 11071057 PMC 931269936101435 · doi ↗ · pubmed ↗
- 8Lee S.Karki R.Wang Y.Nguyen L.N.Kalathur R.C.Kanneganti T.D.AIM 2 forms a complex with pyrin and ZBP 1 to drive PA Noptosis and host defence Nature 597202141541910.1038/s 41586-021-03875-834471287 PMC 8603942 · doi ↗ · pubmed ↗