Unraveling the mechanism of tripterygium glycosides tablets-induced liver injury and the protective role of total glucosides of peony from immune-metabolic dysregulation to multi-cellular cascade
Yi Zhang, Zihe Ding, Xiaoyue Wang, Jiayun Shen, Jiayun Chen, Lin Chen, Weiheng Chen, Na Lin, Yanqiong Zhang
TL;DR
This study explores how a drug causes liver damage and how another drug helps prevent it by affecting multiple cells and metabolic processes in the liver.
Contribution
The study reveals a multi-cellular mechanism of liver injury and the protective role of a drug through immune-metabolic regulation.
Findings
TGT causes liver injury by disrupting iron-lipid homeostasis and triggering inflammatory cell responses.
TGP mitigates TGT toxicity by reversing immune cell polarization and reducing oxidative stress and lipid accumulation.
The iron-lipid axis is identified as a central driver of TGT-induced liver injury across multiple cell types.
Abstract
Tripterygium glycosides tablets (TGT) are effective against autoimmune diseases but cause significant drug-induced liver injury (DILI) that limits clinical use. While TGT disrupts hepatic iron-lipid homeostasis and co-administration with Total Glucosides of Peony (TGP) mitigates its toxicity, the multi-cellular mechanisms remain unclear. In the current study, using integrative single-cell RNA sequencing and pathological validation in controlled mouse models (TGT vs.Con and TGT + TGP vs. TGT.) we elucidated a pathogenic iron-lipid axis driving hepatotoxicity via cellular cascades. TGT initiated Kupffer cell M1 polarization, releasing pro-inflammatory cytokines (TNF-α and IL-1β) that recruited neutrophils and induced NETosis-mediated oxidative stress. Concurrently, hepatic endothelial cells developed iron overload with increased Hamp and decreased Slc40a1, alongside inflammatory damage,…
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- —National Natural Science Foundation of China
- —Beijing Natural Science Foundation
- —Scientific and technological innovation project of China Academy of Chinese Medical Sciences
Peer 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
TopicsNatural Compounds in Disease Treatment · Sphingolipid Metabolism and Signaling · Liver physiology and pathology
Background
Tripterygium glycosides tablet (TGT), a patented Chinese drug based on the Chinese herb Tripterygium wilfordii Hook. F [1], is widely used to treat various autoimmune diseases, including rheumatoid arthritis, nephrotic syndrome, psoriasis, and leprosy, with satisfactory clinical efficacy [2, 3]. Notably, TGT alleviates bone destruction and synovial hyperplasia in rheumatoid arthritis through multiple regulatory mechanisms, such as modulating iron and lipid metabolism homeostasis in the joints, reducing local inflammation, and inhibiting angiogenesis [4]. However, several clinical studies have reported a relatively high risk of drug-induced liver injury (DILI) from TGTs owing to their narrow and limited treatment window [3, 5, 6]. Of the 472 reported TGT-induced adverse reactions, 54 (11.44%) may result in abnormal liver function. Notably, among the 45 patients with severe adverse reactions, 11 (24.44%) experienced TGT-induced liver injury [7]. Consequently, our research group developed and issued the “Clinical Practice Guidelines for Tripterygium Glycosides/Tripterygium wilfordii Tablets for the Treatment of Rheumatoid Arthritis” (No. T/CACM 1337-2020), providing comprehensive recommendations for the rational use of TGTs in clinical settings [8].
In our previous study [9], we conducted a systematic network analysis by integrating multi-omics profiling data related to the clinical efficacy and DILI of TGT, using clinical samples along with its chemical and target profiling. We identified an association between iron-mediated metabolic homeostasis and both the clinical efficacy and toxicity of TGT in rheumatoid arthritis therapy, and revealed that TGT administration leads to “iron–lipid” disturbances in the liver. Total Glucosides of Peony (TGP), the bioactive component of Paeonia lactiflora Pallas, has been developed as a patent drug (GYZZ: H20055058) widely used for RA therapy. As the most commonly used companion drug with TGT in clinical practice, TGP significantly reduces the incidence of adverse reactions. Our previous studies using acute and chronic animal models confirmed that TGP enhances the efficacy and reduces the toxicity of TGT, and further clarified the mechanisms underlying their synergistic compatibility [10]. However, the mechanisms underlying TGT-induced hepatotoxicity and TGP-mediated detoxification at the level of cellular heterogeneity remain incompletely understood.
To address this problem, we established a mouse model of TGT-induced acute liver injury and, using single-cell RNA sequencing (scRNA-seq) to characterize the liver cell types involved in the “iron–lipid” disturbances that subsequently contribute to hepatotoxicity. Furthermore, we evaluated the therapeutic potential of TGP in reversing this injury, focusing on the underlying protective mechanisms from the perspective of cellular heterogeneity. Consequently, this study proceeds under the hypothesis that TGT-induced hepatotoxicity is primarily driven by the disruption of iron–lipid metabolic homeostasis, which propagates through a hierarchical multi-cellular cascade involving Kupffer cells, neutrophils, endothelial cells, hepatocytes, and adipocytes. We further hypothesize that TGP exerts its protective effect by intercepting this cascade at multiple cellular nodes, thereby restoring metabolic balance and mitigating liver injury. These findings provide a novel mechanistic basis for the safe clinical application of Tripterygium wilfordii derivatives within the framework of traditional Chinese medicine compatibility.
Materials and methods
Materials and reagents
TGTs were purchased from Zhejiang DND Pharmaceutical Co., Ltd. (Cat#Z44023753, Zhejiang, China). TGPs were purchased from Ningbo Liwah Pharmaceutical Co., Ltd (Cat#H20055058, Ningbo, China). Aspartate aminotransferase (AST), alanine aminotransferase (ALT), total cholesterol (TC) and triglyceride (TG) assay kits were obtained from Nanjing Jiancheng Biotech (Nanjing, China). Malondialdehyde (MDA) enzyme-linked immunosorbent assay (ELISA) kits were obtained from Shanghai Enzyme-linked Biotechnology (Shanghai, China). A reactive oxygen species (ROS) assay kit was purchased from Baiao Leibo Technology (Beijing, China). Tissue iron content assay kit, Hematoxylin and eosin (H&E), Prussian blue iron, and oil red O staining kits were purchased from Solarbio Biotech (Beijing, China). Anti-CitH3 (Cat# ab176842), anti-Ly6G (Cat# GB11229) and anti-Perilipin-2 (Cat# 15,294-1-AP) were purchased from Abcam (Cambridge, UK), Servicebio Biotech (Wuhan, China), and Proteintech (Wuhan, China), respectively.
Animal experiments
A total of 18 male KunMing mice, weighing 25 ± 2 g, which were used to construct the TGT-induced acute DILI animal model, were obtained from Beijing Vital River Laboratory Animal Technology Ltd, Beijing. The license number for the grant obtained for the experimental studies is SCXK (Jing) 2021-0006. 18 mice were randomly allocated to three groups and administered intragastrically with either phosphate-buffered saline (PBS; vehicle control), 300 mg/kg TGT (approximately 20-fold the clinical daily dose) or 450 mg/kg TGP (approximately twofold the clinical daily dose), pretreatment for 3 days followed by the administration of TGT 300 mg/kg for 24 h. The dosages were all selected based on our previous study [10], and mice were sacrificed after 24 h.
Biochemical and histological analysis
Liver function was assessed by measuring serum ALT and AST levels. Additionally, hepatic levels of ROS, MDA, TG, TC, and total iron content were quantified using corresponding kits following the manufacturer's instruction. Liver tissues were embedded in paraffin and cut into sections. Histological changes were assessed using H&E, Prussian blue iron, and oil red O staining, as described previously [11].
Organ/brain index analysis
The brain samples were obtained by separating the brain from the spinal cord on the posterior side. The intact brain was then carefully removed, weighed, placed in liquid nitrogen for quick-freezing, and preserved at −80 ℃. Thereafter, intact organs, including the thymus, spleen, kidney, and liver, were carefully separated. The collected organs were then weighed, and fixation and freezing fractionations were performed. The weight of the organs was expressed in grams (g) and the organ/brain index was calculated as the absolute weight of the organ/weight of the brain.
Immunofluorescence (IF) staining
Paraffin sections of liver tissues were dewaxed, dehydrated, permeabilized, and blocked with bovine serum albumin. Subsequently, the sections were washed with TBST, stained with 40,6-diamidino-2-phenylindole for 5 min, incubated with an anti-fading fixative. The liver tissue sections were then incubated with primary antibodies anti-F4/80 (1:500), anti-iNOS (1:100), anti-CD206 (1:100), anti-CitH3 (1:1000), anti-Ly6G (1:200), followed by incubation with fluorescent secondary antibodies Fluor 488 and 647, 1:200) and DAPI staining. Images were acquired using a microscope (Leica TCS SP8 SR, Germany).
Immunohistochemical (IHC) staining
To detect the expression patterns and subcellular localizations Perilipin-2 proteins in the liver tissues, immunohistochemical staining was performed according to the protocol described in our previous studies [10]. Immunoreactive quantification was performed using ImageJ by examining three different representative areas of the field.
Preparation of single-cell suspensions
Mouse liver tissues (n = 2) from the CON, TGT and TGT-TGP groups were collected, cut into small pieces, and enzymatically digested for approximately 30 min using a mouse liver tissue dissociation kit (130-105-807, Miltenyi Biotec). The separated cells were then filtered through a 70-mm filter and rinsed with 15 mL of Dulbecco’s Modified Eagle Medium (DMEM) until a single-cell suspension was obtained. Subsequently, the cells were centrifuged at 300×g for 10 min, collected, and resuspended in 1 mL of PBS. Red blood cells were eliminated using red blood cell lysis buffer (130-094-183, Miltenyi Biotec). Subsequently, the samples were washed twice with PBS to obtain a single-cell suspension.
ScRNA-seq
The prepared single-cell suspension of liver tissue was subjected to up-sequencing using the BD Rhapsody™ system to sequentially add cells and magnetic beads with cell barcodes to the microplate chip [12]. Pairing of single cells and magnetic beads in the same microtiter well was achieved via natural sedimentation, followed by the recovery of magnetic beads with mRNA using a magnetic field. Reverse transcription and PCR amplification were performed to construct single-cell transcriptome libraries for NGS high-throughput sequencing. The cell suspension quality control was performed based on the percentage of live cells (> 80%) and cell concentration (300–800 cells/μL). reverse transcription and random primer-mediated amplification were performed following single-cell mRNA capture. The amplified cDNAs were subjected to quality control and utilized for the construction of NGS libraries. Following quality control, high-throughput sequencing was performed in the PE150 mode using Illumina NovaSeq, with a recommended sequencing volume of 50 k reads/cell. The sequencing data were preprocessed using the BD Rhapsody Whole Transcriptome Assay Analysis Pipeline (v1.8) to generate single-cell expression matrices for downstream analyses.
Quality control and preprocessing of the dataset
Cluster analysis and visualization were performed using the Seurat analysis package [13, 14] by reading the gene expression matrix of each sample and converting it into a Seurat object. Cells with > 25% mitochondrial Unique Molecular Indentifie (UMI), < 500 UMI, or those containing < 200 genes were excluded. Data were log-normalized, scaled based on the UMI counts, and subjected to principal component analysis using the first 2000 highly variable features. Cluster resolution was set to 0.6, and the data were visualized using unified manifold approximation and projection (UMAP). The expression of specific genes in each cluster was visualized using feature plots, violin plots, and heat maps.
Cell clustering and type identification
Specific markers for each Seurat cluster were determined using the FindAllMarkers function of the Seurat analysis package in combination with the Wilcoxon test, with the screening criteria set to log2 fold change (FC) > 0.25 and min pct > 0.25. Cell-type annotation was performed using the R package SingleR (v1.4.1) [15] which is based on the computational framework for large-capacity transcriptomes, followed by checking and matching with the CellMarker 2.0 online database to complete the final annotation of cell types [16].
Differentially expressed genes (DEGs) and functional enrichment analysis
DEGs for different cell types were determined using the FindMarkers function of the Seurat with |log2FC|≥ 0.25 and adjusted P < 0.05. DEGs were visualized using the heatmaps and violin plots using R package pheatmap (v1.0.12) and MySeurat Wrappers (v0.1.0). For functional enrichment analysis, Gene Ontology (GO) analysis was performed using the clusterProfiler R package (v3.18.1) based on the upregulated and downregulated genes of DEGs among different groups. Pathway enrichment analysis using adjusted P < 0.05 was performed using the R package ggplot2 (v3.3.3). To quantitatively evaluate functional states across different cell populations, gene set activity scoring was performed using the R package AUCell (v 1.28.0).
Pseudotime analysis
After the cell-type annotation, single-cell proposed time-series analysis was performed independently for each cell type using Monocle2 and DDR-Tree reduction [17], following the standard parameters: initializing the Monocle object to utilize the expression matrices and metadata of the Seurat object, selecting the top marker genes of the Seurat clustering as the sorting genes, performing dimensionality reduction and eliminating batch effects to ensure accuracy, and finally visualizing the analysis results using trajectory graphs and heatmaps to determine changes in the cell state over time.
Statistical analysis
All data were expressed as the mean ± standard error of the mean from at least three biological replicates. GraphPad Prism (v8.0) was used for the statistical analyses. The statistical differences in animal experimental data between the two groups were analyzed using the ordinary one-way analysis of variance. Student's t-test was employed to analyze two-group differences, unless mentioned otherwise. Statistical significance was set at* P* < 0.05.
Results
TGT administration induces hepatotoxicity in mice
The toxic effects of TGTs on the liver tissues of the experimental mice were investigated by administering 300 mg/kg TGTs to male Kunming mice after 24 h, as described previously [18]. Compared with the CON group, TGT administration significantly increased the hepatic brain index (P < 0.001; Fig. 1A) and serum ALT (P < 0.001), AST (P < 0.001), and ROS (P < 0.001) levels in the experimental mice (Fig. 1B). In addition, H&E staining of the mouse liver tissues revealed pathological features, including hepatocyte injury, inflammation, bile duct injury, vascular injury, and many dilated capillary sinusoids with erythrocyte sludge, due to TGT-induced acute liver injury (Fig. 1C, D), indicating severely impaired liver function and the successful establishment of the TGT-induced DILI mouse model.Fig. 1. Toxicity characteristics, TGP-mediated attenuation, and single-cell transcriptome atlas of TGT-induced acute liver injury. A Organ indices in the control (CON), TGT-treated, and TGT-TGP co-administered groups; B Serum levels of ALT, AST, and ROS intensity in liver tissues in the three groups; C, D Liver histopathological changes (H&E staining) and corresponding pathological scores, showing TGT-induced hepatocyte injury, inflammation, vascular damage, and sinusoidal congestion, which are ameliorated by TGP; E UMAP clustering of single-cell transcriptomes in liver tissues from CON, TGT and TGT-TGP groups; F Violin plots showing expression of marker genes for each identified cell type; G Differentially expressed genes (DEGs) between CON and TGT groups and percentage changes of various cell types among CON, TGT and TGT-TGP groups; H UMAP clustering highlighting core cell types with significant differences from the three groups. (Data are expressed as the Mean ± SD. ^^P < 0.01, ^*^P < 0.001 vs CON, ^#^P < 0.05, ^##^P < 0.01, ^###^P < 0.001 vs TGT)
TGP co-administration attenuates TGT-induced hepatotoxicity
Building upon our prior evidence that TGP compatibility reduces TGT hepatotoxicity via autophagy regulation [10], we here validated the hepatoprotective effects of TGP in this model. Consistent with previous findings, TGP co-administration (450 mg/kg) significantly reversed TGT-induced elevations in serum ALT, AST, and ROS (P < 0.05, P < 0.001, P < 0.001 vs. TGT, respectively; Fig. 1B). Histologically, TGP ameliorated hepatocyte injury, vascular damage, and sinusoidal congestion (Fig. 1C, D), corroborating its role in mitigating acute liver injury.
Mouse liver cell atlas upon TGT administration
A total of 73,157 cells (26,781 in the control group, 13,916 in the TGT group and 32,460 in the TGT-TGP group shown in Additional file 1: Figure S1A) were analyzed after quality control, following single-cell sequencing of liver tissues from control, TGT-treated, TGT-TGP-treated mice. Subsequently, 13 cell types (Fig. 1E), including B cells (n = 11,670, expressing Cd79a), endothelial cells (n = 22,218, expressing Clec4g), Kupffer cells (n = 13,775, expressing Clec4f), monocytes (n = 3942, expressing Chil3), T cells (n = 3820, expressing Tcf7), NK cells (n = 4597, expressing Gzma), neutrophils (n = 3676, expressing S100a8), dendritic cells (n = 3523, expressing Ccr9), hepatocytes (n = 2165, expressing Cyp3a11), adipocytes (n = 1972, expressing Rspo3), progenitor cells (n = 1122, expressing Mki67), cholangiocytes (n = 438, expressing Sox9), and basophils (n = 239, expressing Ms4a2), were identified based on the corresponding marker genes [16, 19–21] (Fig. 1F). Additional file 1: Figure S1B shows the intragroup occupancy of each cell type. We systematically analyzed differentially expressed genes (DEGs) across all three experimental groups (CON, TGT, and TGT-TGP) for each of the 13 identified cell types. Compared with the CON group, TGT administration significantly altered the gene expression profiles of multiple cell types. Notably, among these, Kupffer cells, neutrophils, endothelial cells, hepatocytes, and adipocytes exhibited the most pronounced transcriptional changes. Importantly, comparative analysis of cell type proportions across the three groups revealed that TGP co-administration effectively restored the altered proportions of these five key cell populations toward normal levels (Fig. 1G). Consequently, these cell types were selected for subsequent in-depth cellular heterogeneity analysis to elucidate the protective mechanisms of TGP (Fig. 1H).
TGT Triggers Kupffer M1 polarization and inflammatory cascade via chemokine release while TGP reprograms M2 rescue
Kupffer cells, the primary macrophages in the liver, are crucial for maintaining tissue homeostasis and responding rapidly to liver injury [22]. Like other macrophages, Kupffer cells exist in three states—M0, M1, and M2—each with a distinct gene expression profile. Notably, M0 cells are in a resting state, M1 cells are pro-inflammatory and activated by external stimuli, and M2 cells have immunoregulatory and anti-inflammatory functions. M1 and M2 states are interconvertible under specific conditions, which allows macrophages to maintain tissue homeostasis and balance in response to environmental changes [23]. In the current study, polarized Kupffer cell types were defined based on their functional marker genes as Kupffer_M1 (n = 4762, CON = 461, TGT = 1903, TGT-TGP = 2398), Kupffer_M2 (n = 3872, CON = 1264, TGT = 460, TGT-TGP = 2148), and Kupffer_M0 (n = 5141, CON = 3232, TGT = 604, TGT-TGP = 1305). These quantitative data demonstrate that TGT administration markedly increased M1 Kupffer cells while decreasing M0 and M2 subsets; conversely, TGP co-administration reversed this trend by reducing M1 and promoting M2 expansion, indicating a shift from pro-inflammatory to reparative macrophage polarization.
In the CON group, most Kupffer cells were in the non-polarized M0 state, with only a few M1/M2 cells. However, significant M0 to M1 polarization was observed following TGT administration. Specifically, the combination of TGP significantly attenuated this polarization process. TGP administration effectively restored cellular homeostasis by decreasing the proportion of M1 Kupffer cells while concurrently promoting the expansion of the M2 subpopulation, suggesting a shift from an inflammatory to a more reparative and regulatory state. (Fig. 2A, B). These polarized Kupffer cell types were further characterized based on their functional marker genes (Fig. 2C). M1 Kupffer cells exhibited the highest DEGs, with the most upregulated genes, indicating that these cells may be more sensitive to TGT administration than other Kupffer cell types (Fig. 2D). Further functional enrichment analysis based on GO terms revealed that the upregulated genes were mainly associated with myeloid cell homeostasis and chemotaxis in M0 Kupffer cells; lipid storage, immune response, and infiltration in M1 Kupffer cells; and maintaining immune homeostasis and STAT3 signaling pathways in M2 Kupffer cells (Fig. 2E). Therefore, TGT may induce liver injury by affecting lipid storage, immune responses, and inflammatory infiltration in M1 polarized cells.Fig. 2TGT induces Kupffer cell M1 polarization and inflammatory cascade, while TGP promotes M2 polarization. A UMAP clustering of Kupffer cell polarized subtypes (M0, M1, M2) in CON, TGT and TGT-TGP groups; B Sankey plots showing the distribution of Kupffer cell subtypes, revealing TGT-induced M0-to-M1 polarization TGP administration effectively reversed this trend by decreasing the M1 proportion and promoting M2 subpopulation expansion; C Expression of functional marker genes in polarized Kupffer cell subtypes; D Differential gene expression profiles in polarized subtypes, with M1 Kupffer cells showing the highest number of upregulated genes; E GO-based functional enrichment analysis of upregulated genes in M0 (myeloid cell homeostasis, chemotaxis), M1 (lipid storage, immune response, infiltration), and M2 (immune homeostasis, STAT3 signaling) subtypes; F Pseudotime trajectory analysis of polarized subtypes, showing CON group cells transition to M0 and TGT group cells transition to M1 after Node 2; G Dynamic heatmap of differentially expressed genes (DEGs) during pseudotime trajectory, with clustered gene sets involved in M1 polarization; H Pseudotime trajectory analysis of inflammatory cytokines (Tnf, Il1b) and chemokines (Ccl2, Ccl4) showing dynamic upregulation in TGT group; I, J Dual immunofluorescence staining of iNOS (M1 marker) and CD206 (M2 marker) in Kupffer cells from CON, TGT, and TGT-TGP groups (^***^P < 0.001 vs. CON; ^###^P < 0.001 vs. TGT)
On further exploring the TGT-induced dynamic Kupffer cell M1 polarization, pseudotime trajectory analysis revealed that states 1 and 2 were the main stages of M1/M2 transitions. After Node 2, Kupffer cells in the CON and TGT groups transitioned to M0 and M1, respectively (Fig. 2F), suggesting that cells in both groups were in the same initial states but transitioned into resting M0 state in the CON group and the pro-inflammatory M1 state following TGT administration. Further analysis of the dynamic expression heatmap during the pseudotime trajectory identified key genes involved in M1/M2 polarization, and clustering of these genes based on their expression trends revealed four gene sets. GO analysis of the yellow cluster, which was significantly upregulated during M1 polarization, indicated functions related to immune cell differentiation, activation, chronic inflammation, and apoptosis (Fig. 2G), providing a detailed view of TGT-induced Kupffer cell M1 polarization. Following M1 polarization, Kupffer cells release chemokines and pro-inflammatory cytokines, which recruit other inflammatory cells to the liver. In the TGT group, the expression of genes encoding chemokines (Ccl2 and Ccl4) and cytokines (Tnf and Il1b) increased dynamically over time (Fig. 2H). In contrast, co-administration of TGP with TGT significantly counteracted this M1 polarization while promoting M2 polarization of Kupffer cells, as evidenced by a significant decrease in iNOS expression and a marked increase in CD206 expression compared with the TGT group (Fig. 2I, J).
TGT accelerates neutrophil temporal progression to NETosis-mediated tissue damage while TGP attenuates oxidative injury
Neutrophils are a part of the intrinsic immune response to tissue injury, with high migratory activity, aggregates at local inflammatory sites to remove dead or damaged cells, and exert their immune effects mainly by releasing antimicrobial peptides, ROS, and forming the neutrophil extracellular trapping networks (NETs) [24, 25]. We analyzed the role of neutrophils in the adaptive response to TGT-induced liver injury and TGP co-administration. Based on UMAP clustering with marker gene identification, neutrophils were classified into three clusters named Neu_1–3 (Fig. 3A). Based on marker gene expression and functional enrichment analysis, neutrophil subtypes were classified as: Neu_2 (primarily involved in pre-NET recruitment and transformation), Neu_1 (associated with mid-NET processes including extracellular matrix degradation and NET protein release), and Neu_3 (involved in late-NET ribosome generation and oxidative stress modulation) (Fig. 3B). Proportion analysis across the three groups revealed that TGT administration significantly increased the Neu_1 subpopulation while decreasing Neu_2 and Neu_3 compared with CON. Notably, TGP co-administration effectively reversed these altered proportions, restoring Neu_1 and Neu_2 toward baseline levels, suggesting that TGP mitigates TGT-induced neutrophil dysregulation and suppresses excessive NETosis (Additional file 1: Figure S2). Subsequently, DEG expression in each neutrophil subtype of the two groups was examined, and GO enrichment analysis of the upregulated DEGs was performed. The GO terms for each subtype further corroborated the occurrence of the three processes (Fig. 3C, D). Chronological trajectory analysis revealed that, compared with the CON group, the cells in the TGT group were biased towards the later stage, illustrating the progression from Neu_2 to Neu_1 to Neu_3 (Fig. 3E). Additionally, the expression of NET-related marker genes in the proposed timeline was significantly elevated in the TGT group over time (Fig. 3F). Therefore, TGT may lead to NET over formation, causing oxidative stress injury in liver tissues. Subsequently, IF staining was performed to localize neutrophils in mouse liver tissue, and the key NET formation markers, Ly6G (red) and CitH3 (green), were assessed (Fig. 3G, H). Notably, following TGT administration, red and green fluorescence indicating neutrophils and NET formation, respectively, significantly increased (both* P* < 0.001 vs. CON). Moreover, a web-like structure was observed between cells, suggesting that TGT may promote NET formation and exacerbate tissue damage. Importantly, the combination of TGT and TGP markedly attenuated these changes, as evidenced by reduced Ly6G and CitH3 fluorescence signals (both* P* < 0.001 vs. TGT), indicating effective suppression of NET overproduction (Fig. 3I).Fig. 3TGT accelerates neutrophil temporal progression to NETosis-mediated damage, while TGP attenuates oxidative injury. A UMAP clustering of neutrophil subtypes in CON, TGT and TGT-TGP groups; B Expression of marker genes associated with key stages of NET formation; C Differential gene expression profiles in neutrophil subtypes; D GO functional enrichment analysis of upregulated genes in each neutrophil subtype; E Chronological trajectory analysis showing TGT group cells biased toward later stages, progressing from Neu_2 to Neu_1 to Neu_3; F Temporal trajectory of NET-related marker gene expression, with significant upregulation in TGT group over time;** G, H** Dual immunofluorescence staining of liver tissues showing Ly6G (red, neutrophils) and CitH3 (green, NET formation); I Quantification and visualization of neutrophil localization and NET formation, with TGP co-administration reducing Ly6G and CitH3 signals (^***^P < 0.001 vs. CON; ^###^P < 0.001 vs. TGT)
TGT provokes endothelial subtype-specific iron overload and inflammatory ferroptosis while TGP ameliorates metabolic-immune perturbations
Hepatic endothelial cells are crucial for liver function and account for a significant proportion of liver cells. Damage to these cells is closely associated with liver dysfunction. Notably, a significant reduction in the proportion was observed following TGT administration, decreasing from the CON group to the TGT group (Fig. 4A), in line with the findings of previous studies [26]. In contrast, in the TGT-TGP co-treatment group, the number of endothelial cells was significantly increased, indicating that TGP co-administration effectively restores endothelial cell abundance and alleviates TGT-induced endothelial injury (Fig. 4A).Fig. 4TGT provokes endothelial subtype-specific iron overload and inflammatory responses, while TGP ameliorates metabolic-immune perturbations. A UMAP clustering of hepatic endothelial cell subtypes in CON, TGT and TGT-TGP groups; B Expression of marker genes distinguishing LVEC and LSEC subtypes; C Sankey plots illustrating the distribution of endothelial subclasses; D, E GO functional enrichment analysis of upregulated D and downregulated E DEGs in each subtype; F, G Linkage scores for iron metabolism F and inflammatory response G in endothelial subtypes, with TGT exerting greater effects on iron metabolism in LVECs and inflammation in LSECs; H, I H&E staining showing inflammatory infiltration of LSECs and LVECs, with TGP co-administration reducing infiltration (^^P < 0.01, ^*^P < 0.001 vs. CON; ^###^P < 0.001 vs. TGT)
Further, hepatic endothelial cells were divided into six clusters: liver vascular endothelial cells (LVECs; two subtypes) and hepatic sinusoidal endothelial cells (LSECs; four subtypes) (Fig. 4B). The percentage of LVECs significantly decreased following TGT administration, accompanied by dysregulation of iron metabolism genes (upregulated Hamp, downregulated Slc40a1) and enhanced inflammatory signaling. Notably, TGP co-administration not only restored the LVEC population to near-baseline levels but also partially reversed these iron metabolism and inflammatory gene expression changes, indicating its protective role in mitigating endothelial ferroptosis and inflammation (Fig. 4C). Subsequent GO-based functional enrichment analyses of each cell subtype revealed the distinct functions of DEGs in LVECs and LSECs. The upregulated genes in LSECs were primarily involved in biological processes associated with oxidative stress metabolism, mitochondrial apoptosis, and inflammatory apoptosis, whereas those in LVECs were significantly associated with iron ion transport, extracellular stimulus response, and autophagy (Fig. 4D). In contrast, the downregulated genes in LSECs were involved in immune regulation, endothelial cell migration, and iron transport, whereas those in LVECs were involved in cell number homeostasis and wound healing (Fig. 4E). Therefore, TGT may induce oxidative stress damage and disrupt iron metabolism in endothelial cells by upregulating relevant genes while also inhibiting metal ion transport, immune response, and cell migration. However, the toxic effects of TGT on LVECs and LSECs were heterogeneous.
The extent of the effect of TGT administration on different cell subtypes was evaluated using endothelial cell scores for two core segments: iron metabolism and inflammatory response. TGT significantly interfered with iron metabolism and exacerbated the inflammatory response in all endothelial cell subtypes (Fig. 4F, G). Notably, TGT exhibited greater sensitivity in regulating iron metabolism in LVECs and was more disruptive to LSECs in terms of the inflammatory response. Disturbances in iron metabolism and inflammation are significant features of cellular iron death. Evaluation of the core regulatory genes of iron death in endothelial cells revealed that TGT administration upregulated the expression of iron-modulating hormone (Hamp) and Alox5 and suppressed Slc40a1, which is consistent with the trend of iron death (Additional file 1: Figure S3). Notably, histological examination and pathological scoring of the liver tissues following TGT administration revealed significant inflammatory infiltration of LSECs and LVECs. Critically, TGP co-administration attenuated these pathological changes, significantly reducing inflammatory infiltration in both LSECs and LVECs and improving endothelial inflammatory scores (P < 0.001 vs. CON, P < 0.001 vs. TGT; Fig. 4H, I).
TGT provokes hepatocyte subclass-specific metabolic disruption while TGP alleviates iron overload
Hepatocytes represent a key target cell type in liver damage. Based on the marker genes, hepatocytes were classified into three subclasses (Fig. 5A): Hepa1-3, and the expression of the corresponding marker genes is shown in Fig. 5B. TGT administration significantly reduced the total number of hepatocytes, accompanied by a relative expansion of the Hepa2 subpopulation and a contraction of the Hepa1 subpopulation (Fig. 5C). Functionally, Hepa1 cells were primarily associated with fatty acid metabolism, whereas Hepa2 cells were enriched in iron transport and oxidative stress pathways. Importantly, TGP co-administration not only rescued the TGT-induced decline in total hepatocyte counts but also restored the proportional balance between Hepa1 and Hepa2 subsets, thereby normalizing both fatty acid metabolism and iron homeostasis (Fig. 5C). Notably, upregulated DEGs in Hepa2 cells were associated with the iron transport pathway (Fig. 5D), whereas downregulated DEGs in Hepa1 cells were primarily associated with fatty acid, arachidonic acid, and nitric oxide metabolism (Fig. 5E). Therefore, TGT may primarily affect hepatocytes via fatty acid metabolism, oxidative stress injury, and iron metabolism transporter pathways. Subsequent assessments of the linkage scores revealed different responses among the subclasses. TGT administration exerted marked effects on oxidative stress damage in Hepa2 cells (Fig. 5F), whereas opposing modulation trends on fatty acid metabolism were observed in Hepa1 and Hepa2 cells (Fig. 5G). In addition, TGT induced iron deposition in all cell subtypes (Fig. 5H). These results indicated the heterogeneous regulation of hepatocytes induced by TGT, leading to oxidative stress injury, fatty acid metabolism abnormalities, and iron deposition. This suggests that hepatocytes may be the primary target cells of TGT-induced lipid metabolism disorders. Moreover, Prussian blue staining revealed a significant increase in blue deposition in endothelial cells and hepatocytes following TGT administration compared with other areas. Consistently, TGP co-administration significantly reduced iron deposition in both hepatocytes and endothelial cells (P < 0.001 vs. CON; P < 0.001 vs. TGT; Fig. 5J, K), rescuing iron dysregulation and demonstrating its targeted mitigation of TGT-induced iron overload in these key cell types, thereby attenuating subsequent metabolic injury. These histopathological observations were further corroborated by the quantification of total hepatic iron content, which showed a significant elevation in the TGT group that was effectively reversed by TGP treatment (P < 0.001 vs. CON; P < 0.05 vs. TGT; Fig. 5I).Fig. 5TGT provokes hepatocyte subclass-specific metabolic disruption, while TGP alleviates iron overload. A UMAP clustering of hepatocyte subtypes in CON, TGT, and TGT-TGP groups; B Expression of marker genes in hepatocyte subtypes; C Distribution of hepatocyte subtypes in each group; D, E GO functional enrichment analysis of upregulated D and downregulated E differentially expressed genes (DEGs); F–H Linkage scores for oxidative stress damage (F), fatty acid metabolism (G), and iron metabolism (H) in hepatocyte subtypes, revealing heterogeneous responses to TGT; (I) Iron content in liver tissues in the three groups; (J, K) Prussian blue staining showing iron deposition in hepatic endothelial cells and hepatocytes, with TGP co-administration reducing deposition (^***^P < 0.001 vs. CON; ^#^P < 0.05, ^###^P < 0.001 vs. TGT)
TGT drives adipocyte hyperplasia and subtype-specific lipid peroxidation while TGP normalizes perilipin-2-mediated storage
Adipocytes are primarily located in the loose connective tissue of the liver. Under normal conditions, the liver fat content remains low, but the number of adipocytes distinctly increases along with disturbed lipid metabolism or lipid peroxidation in fatty liver condition [27]. Figure 6A illustrates the adipocyte distribution of adipocytes in the CON, TGT, and TGT-TGP groups. Based on the expression patterns of key marker genes and the corresponding results of functional enrichment analysis, four hepatic adipocyte subtypes were identified: Adip1 ~ 4 (Fig. 6B). Subpopulation analysis revealed that TGT administration markedly expanded the Adip1 subset, which was functionally enriched in lipid peroxidation, ROS generation, and lipid storage pathways, while concurrently depleting the Adip2 subset, which was associated with autophagy and lipid transport. Notably, TGP co-administration significantly attenuated the expansion of the pro-lipotoxic Adip1 subtype while rescuing the metabolically active Adip2 population, thereby restoring the functional balance of hepatic adipocytes and mitigating lipid accumulation (Fig. 6C).Fig. 6TGT drives adipocyte hyperplasia and subtype-specific lipid peroxidation, while TGP normalizes perilipin-2-mediated storage. A UMAP clustering of hepatic adipocyte subtypes in CON, TGT and TGT-TGP groups; B Bubble plots showing expression of marker genes identifying adipocyte subtypes; C Distribution of adipocyte subtypes; D Differential gene expression profiles of each adipocyte subtype between CON and TGT groups; E, F GO functional enrichment analysis of upregulated genes: Adip1 enriched in chemokine production, lipid metabolism regulation, ROS generation, and lipid storage (E); Adip2 enriched in autophagy, lipid transport, and fatty acid metabolism (F); G Gene set scores for lipid peroxidation and lipid transport; H, I Oil red O staining showing lipid droplet aggregation in liver tissues, with increased area percentage in TGT group, alleviated by TGP; J, K Immunohistochemical staining of perilipin-2 and its average optical density, showing TGT-induced upregulation and TGP-mediated normalization; (L–N) TG (L), TC (M) and MDA (N) levels of liver tissues in the three groups (^***^P < 0.001 vs. CON; ^#^P < 0.05, ^###^P < 0.001 vs. TGT)
Figure 6D shows the DEGs of each adipocyte subtype, with significantly different expression patterns of DEGs in the Adip1 and Adip2 subtypes. Further functional analysis revealed that the upregulated DEGs in Adip1 were mainly enriched in pathways involving chemokine production, lipid metabolism regulation, ROS generation, and lipid storage (Fig. 6E), whereas the upregulated DEGs in Adip2 were mainly involved in autophagy, autophagic vesicle generation, lipid transport, and fatty acid metabolism (Fig. 6F). Therefore, TGT administration induces lipid autophagy, along with lipid transport, storage, and oxidative catabolism regulation in adipocytes. Notably, the lipid peroxidation and lipid transport scores of both Adip1 and Adip2 subtypes reflected heterogeneous differences. Particularly, the Adip1 subtype primarily reflected the lipid peroxidation response following TGT administration, whereas the Adip2 subtype predominantly indicated the lipid metabolism disorder triggered by TGT administration (Fig. 6G).
To verify the significant increase in the number of adipocytes in the liver tissues observed using the single-cell data analysis, lipid droplet aggregation in the liver tissues following TGT administration was evaluated using oil red O staining. A notable increase in the area percentage of lipid droplets was observed in the TGT group (Fig. 6H), particularly around the endothelial cells (P < 0.001 vs. CON; Fig. 6I), in line with our previous observations using transmission electron microscopy [9]. Mechanistically, perilipin-2—a key regulator of lipid droplet biogenesis—was significantly upregulated in TGT-treated livers (P < 0.001 vs. CON; Fig. 6J, K), directly linking adipocyte expansion to aberrant lipid storage. Consistent with these findings, biochemical analysis revealed a substantial accumulation of triglycerides (TG) and total cholesterol (TC) in the TGT group, alongside significantly elevated levels of malondialdehyde (MDA), a marker of lipid peroxidation (P < 0.001 vs. CON; Fig. 6L–N). Crucially, TGP co-administration normalized perilipin-2 expression (P < 0.001 vs. TGT; Fig. 6J, K), concurrently reducing lipid droplet accumulation (P < 0.001 vs. TGT; Fig. 6H, I) and restoring TG, TC, and MDA to near-baseline levels (P < 0.05, P < 0.001 vs. TGT; Fig. 6L–N), thereby rescuing adipocyte metabolic function and mitigating oxidative damage.
TGT propagates multi-cellular cascade driving iron-lipid axis collapse while TGP resolves sequential toxicity
TGT propagates a multi-cellular toxicity cascade initiating with Kupffer cell M1 polarization and triggering CCL4/CCL6-mediated neutrophil recruitment, alongside NETosis-driven oxidative burst that releases peroxides and ROS. This cascade disseminates via hepatic vasculature to induce LVEC-specific iron overload characterized by Hamp upregulation and Slc40a1 downregulation, as well as inflammatory ferroptosis. It subsequently propagates iron-lipid dysmetabolism to LSECs and hepatocytes, as evidenced by Prussian blue-positive lipid peroxidation, ultimately culminating in adipocyte hyperplasia and perilipin-2-mediated lipid droplet accumulation—collectively driving iron-lipid axis collapse. Critically, TGP resolves this sequential toxicity by reprogramming Kupffer M2 polarization with decreased iNOS and increased CD206, suppressing NET formation with decreased Ly6G and CitH3, normalizing endothelial iron flux, alleviating hepatocyte oxidative stress, and restoring adipocyte lipid homeostasis with decreased perilipin-2 and TG, TC, MDA levels. This intercepts the cascade at each cellular tier to prevent metabolic collapse (Graphical abstract).
Discussion
DILI often exhibits cellular heterogeneity. Therefore, identifying toxic effector cell subpopulations and the underlying molecular mechanisms at the single-cell level has significant clinical implications for precise toxicity prevention, control, and reduction. In the present study, we established a mouse model of TGT-induced acute liver injury and comprehensively investigated the corresponding pathological manifestations and functional changes. We delineated the intricate landscape of liver tissues at the single-cell level in the control and TGT groups using scRNA-seq analysis. Our analysis revealed significant alterations in the abundance, composition, and gene expression profiles of five key effector cell types, including Kupffer cells, neutrophils, hepatic endothelial cells, hepatocytes, and adipocytes, following TGT administration. These cell populations were therefore selected for in-depth heterogeneity analysis to elucidate the mechanisms underlying TGT-induced liver injury.
Kupffer cells, as a distinct population of hepatic macrophages, play an important role in maintaining liver homeostasis as well as during acute and chronic liver injury [28]. In this study, we dynamically investigated the heterogeneity of Kupffer cells undergoing polarization following TGT-induced acute liver injury using single-cell sequencing analysis. TGT administration significantly induced M1 polarization in Kupffer cells. Notably, the enrichment of genes associated with lipid storage and immune infiltration in M1 Kupffer cells suggests that TGT-induced hepatotoxicity is driven by a coupling of metabolic dysregulation and inflammatory activation, a mechanism that extends beyond classical pro-inflammatory pathways. While previous studies have established that M1-polarized Kupffer cells promote liver injury and fibrosis [29–31], our single-cell resolution data further reveal that this polarization is accompanied by a distinct metabolic reprogramming, positioning Kupffer cells as central coordinators of the subsequent multi-cellular cascade. Thus, the TGT-induced shift toward M1 polarization likely represents a critical event driving immune-metabolic dysregulation in acute liver injury. Importantly, TGP co-administration appeared to re-establish macrophage functional balance by suppressing M1 Kupffer cells activation while favoring M2 Kupffer cells-associated reparative signaling.
Kupffer cells extend their dendrites into blood vessels and hepatic sinusoids and prevent the spread of peritoneal pathogens to other cells in liver tissues by recruiting neutrophils, wherein the release of chemokines is also crucial [32]. Notably, scRNA-seq and IF analyses of liver tissues in this study revealed that TGT markedly activated the recruitment and generation of NET in liver neutrophils, which may have been accompanied by an excessive release of ROS and myeloperoxidases (MPOs), leading to the localized oxidative stress injury [33, 34]. While neutrophil overactivation has been implicated in various liver diseases [35, 36], our findings provide a direct mechanistic link between Kupffer cell-initiated inflammation and neutrophil-mediated tissue damage in the context of TGT-induced hepatotoxicity. Specifically, the chemokines (Ccl2, Ccl4) released by M1-polarized Kupffer cells appear to drive neutrophil recruitment and subsequent NETosis, establishing a pro-inflammatory amplification loop. Importantly, TGP effectively suppressed NET formation and reduced oxidative injury, indicating that its protective mechanism extends beyond macrophage polarization to directly intercept neutrophil-driven inflammatory cascades. This positions TGP as a multi-node regulator of the immune cell network rather than a single-target intervention.
Under physiological conditions, hepatic endothelial cells with high endocytic features maintain hepatic blood pressure, and hematopoietic stem cells are in a quiescent state [37, 38], whereas the hepatic endothelial cell capillaries lose their hepatoprotective properties and promote angiogenesis and vasoconstriction. Therefore, these cells may be involved in the development and progression of various liver diseases, including acute or chronic liver injury, liver fibrosis, and hepatocellular carcinoma [39]. Our findings suggest that TGT-induced dysregulation of iron transport genes may trigger a state of ferroptosis within the endothelial cell, a process characterized by iron-dependent lipid peroxidation that compromises the integrity of the vascular barrier [40]. Importantly, TGT disturbed iron metabolism predominantly in LVECs while enhancing inflammatory signaling in LSECs, revealing a previously unrecognized heterogeneity in endothelial response to hepatotoxic stress. Such endothelial dysfunction likely amplifies the pro-inflammatory and pro-oxidative hepatic microenvironment established by Kupffer cells and neutrophils, thereby propagating injury to downstream parenchymal cells [41]. This aligns with the concept that endothelial cells serve as both targets and transducers of liver injury [42], translating immune-metabolic signals into vascular dysfunction and facilitating the spread of damage to hepatocytes and adipocytes.
Hepatocytes, as the primary functional liver cells, exhibited subclass-specific metabolic disruption that directly reflects the dual nature of the iron-lipid axis. Hepa1 cells, characterized by impaired fatty acid metabolism, appear to be the primary site of lipid dysregulation, while Hepa2 cells, displaying iron transport abnormalities and oxidative stress, represent the iron overload compartment. This functional segregation within the hepatocyte population suggests that TGT-induced hepatotoxicity involves parallel but distinct metabolic insults that converge to drive overall cell damage. The iron deposition confirmed by Prussian blue staining and hepatic iron content further validates hepatocytes as key targets of TGT-induced lipid peroxidation and metabolic imbalance. Critically, TGP reversed both aspects of hepatocyte dysfunction by reducing iron deposition while restoring metabolic homeostasis indicating that its protective mechanism operates at the intersection of iron and lipid metabolism, rather than targeting either pathway in isolation.
Adipocytes, involved in hepatic lipid storage, showed increased numbers (particularly Adip1 subtypes) and enhanced lipid peroxidation in TGT-treated livers, with upregulated perilipin-2 driving abnormal lipid droplet accumulation, in line with the oil red O staining results. The accumulation of TC, TG, and MDA confirms that lipid-driven oxidative stress represents the endpoint manifestation of the iron-lipid axis collapse, integrating upstream signals from Kupffer cell inflammation, neutrophil NETosis, endothelial ferroptosis, and hepatocyte metabolic disruption. Notably, TGP normalized perilipin-2 expression and restored adipocyte lipid homeostasis, demonstrating that its protective effect penetrates to the final stage of the cascade, effectively breaking the propagation chain from immune activation to metabolic dysfunction. This positions adipocytes not merely as passive recipients of injury but as active participants in the pathological process, and their recovery serves as a key indicator of successful multi-cellular intervention.
However, several limitations of this study warrant consideration. First, while this study provides a detailed mechanistic landscape of TGT-induced acute liver injury, the molecular mechanisms underlying chronic TGT exposure and its long-term hepatotoxic progression remain to be fully elucidated. Second, although TGP demonstrated significant protective effects at both systemic and single-cell levels, the optimal dose–response relationship and the pharmacokinetic interactions between TGT and TGP require further systematic investigation to ensure clinical safety and efficacy. Future studies will focus on chronic TGT-induced hepatotoxicity and the pharmacokinetic profiles of the TGT-TGP combination. Additionally, we plan to employ organ-on-a-chip technology to reconstruct the complex multicellular liver microenvironment in vitro, which will help clarify the specific cellular and molecular targets of TGT-induced liver injury and provide more comprehensive experimental evidence supporting the detoxification mechanism of the TGT-TGP combination.
Conclusions
This study delineates the heterogeneous cellular responses in liver tissues following TGT administration, validated by in vivo experiments. TGT induces hepatotoxicity through a multi-cellular cascade that M1 polarization of Kupffer cells initiates inflammation, followed by neutrophil NETosis-mediated oxidative stress, endothelial subtype-specific iron overload and inflammation, hepatocyte metabolic dysfunction, and adipocyte hyperplasia with abnormal lipid storage, collectively disrupting the iron-lipid axis. Critically, TGP effectively reverses these perturbations across all key cell types including reprogramming Kupffer cells from M1 to M2 polarization, suppressing neutrophil NETosis, normalizing endothelial iron flux, alleviating hepatocyte oxidative stress, and restoring adipocyte lipid homeostasis. By intercepting the toxicity cascade at multiple cellular nodes, TGP prevents the propagation of iron-lipid axis collapse. These findings not only identify specific cellular and molecular targets for TGT-induced liver injury but also provide a mechanistic framework for precision detoxification strategies and drug screening.
Supplementary Information
Additional file 1.