Genipin induces liver injury by disrupting the CPS1-regulated urea cycle and redox homeostasis

Abstract

Genipin (GP) is the key metabolite of geniposide (GE), the primary active component of the traditional Chinese medicine Gardeniae Fructus. Its hepatotoxicity poses a potential risk to safe clinical administration, while its core pathophysiological mechanism remains unclear. This study systematically elucidated the molecular mechanisms of GP-induced hepatotoxicity through integrated multi-omics and functional validation. GP treatment induced significant liver damage and oxidative stress in rats, characterized by elevated serum transaminases (alanine aminotransferase (ALT) increased by approximately 63.5-fold; aspartate aminotransferase (AST) by 18.6-fold), an approximately 1.8-fold decrease in superoxide dismutase (SOD) activity, and an approximately 10-fold accumulation of malondialdehyde (MDA) (all P < 0.001). Metabolomic and proteomic analyses revealed that GP binds to carbamoyl phosphate synthetase 1 (CPS1), triggering urea cycle dysfunction marked by citrulline depletion and significant downregulation of its downstream product fumarate. In vitro experiments confirmed that disrupting CPS1 impairs the urea cycle and induces oxidative stress, while CPS1 overexpression or exogenous fumarate supplementation significantly reversed oxidative stress. Clinical sample analysis revealed similar urea cycle metabolic disturbances, characterized by citrulline depletion and reduced fumarate levels, in plasma from patients with liver injury due to inappropriate use of Gardeniae Fructus-containing formulas, confirming clinical relevance. This study reveals a novel hepatotoxic mechanism whereby GP targets CPS1 to disrupt the urea cycle, reducing fumarate production and impairing nuclear factor erythroid 2-related factor 2 (Nrf2)-mediated antioxidant defense, thereby inducing oxidative stress. These findings provide a scientific basis for improving the safety evaluation and rational clinical use of Gardeniae Fructus-containing preparations.

Graphical abstract

Highlights

• •Genipin covalently binds CPS1, a novel urea cycle target, disrupting ammonia metabolism and inducing liver injury.
• •CPS1 inhibition by genipin depletes fumarate, impairing Keap1-Nrf2-ARE defense and causing redox imbalance.
• •DILI patients show urea cycle disturbances mirroring animal models, highlighting citrulline ratios as biomarkers.

1. Introduction

As the central hub of the human metabolic network, the liver is particularly vulnerable to damage from exogenous compounds because of its vital physiological functions and first-pass metabolism [1,2]. Drug-induced liver injury (DILI) is a global public health challenge [3,4]. Additionally, herb-induced liver injury has garnered increasing attention, especially with the rising use of herbal products and dietary supplements worldwide [5]. The widespread misconception that “natural equals safe” among the public further exacerbates potential risks. With the growing global popularity of herbal products, the incidence of DILI has shown an upward trend internationally. This situation underscores the urgency of prioritizing DILI as a key public health issue for its prevention and management [6,7].

Gardeniae Fructus, the dried ripe fruit of Gardenia jasminoides Ellis from the Rubiaceae family, is widely used in traditional Chinese medicine as a therapeutic agent and dietary supplement, particularly valued for its significant applications in the treatment of inflammatory and metabolic diseases [8,9]. Currently, nearly a thousand formulations containing Gardeniae Fructus as the primary component are extensively applied in clinical practice and daily health management. Its major bioactive constituent, geniposide (GE), exhibits multiple pharmacological effects including anti-inflammatory and hypoglycemic properties [10]. With the increasing clinical application of GE, its potential hepatotoxicity has gradually become a concern. Long-term or high-dose administration can induce liver injury, which has emerged as one of the critical constraints limiting its safe clinical use [[11], [12], [13]]. In-depth studies have revealed that GE itself exhibits relatively low toxicity, and its hepatotoxicity primarily derives from its in vivo metabolite, genipin (GP). Although GP possesses multiple pharmacological activities such as anti-inflammatory and antioxidant effects and is widely utilized as a natural biological cross-linking agent, recent studies have confirmed that, as a metabolite of GE, its toxic effects are substantially more potent than those of the parent compound [14,15]. Animal experiments have demonstrated that GP, as a highly reactive molecule, can induce hepatocyte injury through inhibition of cytochrome P450 enzyme systems, induction of mitochondrial dysfunction, and oxidative stress [16]. Structural analysis reveals that GP contains a hemiacetal group, which can be metabolically converted into electrophilic intermediates bearing α, β-unsaturated carbonyl moieties (e.g., diosbulbin B) [17,18]. These intermediates readily form covalent bonds with nucleophilic groups such as sulfhydryl and amino groups of intracellular proteins, and this intermolecular cross-linking effect may constitute an important molecular basis underlying its hepatotoxicity [19].

Despite the aforementioned understanding, the specific pathways, key targets, and interrelationships among various mechanisms underlying GP-induced liver injury have yet to be systematically elucidated. To address this, the present study employed a multi-level integrative strategy to systematically investigate the molecular mechanisms underlying GP-induced liver injury. First, untargeted metabolomics was utilized to characterize the global metabolic changes, combined with targeted metabolomics for a quantitative analysis of key metabolites. Further, proteomics was employed to screen differentially expressed proteins, and potential targets were identified from these data. During the functional validation phase, the role of candidate proteins in GP-mediated cytotoxicity was elucidated through genetic manipulation in cell models. Concurrently, molecular biology techniques were employed to investigate the regulatory mechanisms of related enzymes on redox homeostasis. Subsequently, exogenous supplementation with fumarate, a downstream metabolite of the urea cycle, was performed to observe its reversal effect on the oxidative stress phenotype. Finally, by analyzing the plasma metabolic profiles of patients with DILI, the correlation between these mechanisms and clinical DILI was validated. This study revealed key metabolic pathway nodes in GP-induced liver injury and identified the metabolic enzyme carbamoyl phosphate synthetase 1 (CPS1) as its covalent binding target. These findings provide evidence for establishing the urea cycle metabolite citrulline as a potential biomarker for DILI and systematically elucidate a novel mechanism underlying the effects of GP toxicity. The experimental procedure is illustrated in Fig. 1.

Fig. 1.

Flowchart of the whole experiment. Unpaired Student's t-tests were performed to compare the differences between two groups. Data are presented as mean ± standard deviation. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. ALT: alanine aminotransferase; AST: aspartate aminotransferase; H&E: hematoxylin and eosin; TUNEL: terminal deoxynucleotidyl transferase dUTP nick end labeling; DHE: dihydroethidium; GP: genipin; Glu: glutamate; Cit: citrulline; Arg: arginine; Orn: ornithine; Asp: aspartate; Gln: glutamine; ASA: argininosuccinic acid; Fum: fumarate; Suc: succinate; CPS1: carbamoyl phosphate synthetase 1; mRNA: messenger RNA; Keap1: Kelch-like ECH-associated protein 1; Nrf2: nuclear factor erythroid 2-related factor 2; DILI: drug-induced liver injury.

2. Materials and methods

2.1. Chemicals and materials

Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) assay kits were purchased from Jiancheng Bioengineering Institute (Nanjing, Jiangsu, China). Superoxide dismutase (SOD) and malondialdehyde (MDA) assay kits were purchased from Beyotime Biotechnology (Shanghai, China). Urethane (20%) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Carboxymethylcellulose sodium (CMC-Na) and heparin were purchased from Sangon Biotech (Shanghai, China). Methanol (mass spectrometry (MS) grade) and formic acid were purchased from Thermo Fisher Scientific Inc. (Waltham, MA, USA). Acetonitrile (MS grade) was purchased from Merck (Darmstadt, Germany). 4% paraformaldehyde was purchased from Titan Technology Co., Ltd. (Shanghai, China). Phosphate-buffered saline (PBS) was purchased from Beijing Solarbio Science & Technology Co., Ltd. (Beijing, China). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay kit and dihydroethidium (DHE) were purchased from Beyotime Biotechnology (Shanghai, China). Ethanol, xylene, and neutral balsam were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). H3B-120, polybrene (hexadimethrine bromide), and puromycin were purchased from Adamas Life (Shanghai, China). GP was purchased from Chengdu Aifa Biotechnology Co., Ltd. (Chengdu, Sichuan, China). N-Acetyl-L-lysine (NAL) was purchased from MedChemExpress (Shanghai, China). Anti-fade mounting medium, 4',6-diamidino-2-phenylindole (DAPI), and glacial acetic acid were purchased from Wuhan Baiqiandu Biotechnology Co., Ltd. (Wuhan, Hubei, China). High-purity reference standards including arginine, argininosuccinate, aspartate, citrulline, fumarate, glutamine, glutamate, succinate, and ornithine were purchased from Titan Technology Co., Ltd. (Shanghai, China). 4-Chloro-L-phenylalanine (4-Cl-phe) was purchased from MedChemExpress (Shanghai, China). Minimum essential medium (MEM) was purchased from Wuhan Pricella Life Science & Technology Co., Ltd. (Wuhan, Hubei, China). Fetal bovine serum (FBS) was purchased from EK Bioscience Co., Ltd. (Suzhou, Jiangsu, China). The small interfering RNA targeting Cps1 (siCps1) and that for small interfering RNA negative control (siNC) were purchased from GenePharma Co., Ltd. (Shanghai, China). Lipofectamine 3000 transfection reagent (Lipo3000) was purchased from Thermo Fisher Scientific Inc. (Shanghai, China). The lentiviral transfection experiment utilized the Cps1 overexpression plasmid and empty vector control provided by Shanghai GeneChem Co., Ltd. (Shanghai, China). MolPure TRIeasy™ Plus Total RNA, Reverse Transcription and Hieff qPCR SYBR Green Master Mix kits were purchased from Yeasen Biotechnology (Shanghai) Co., Ltd. (Shanghai, China). Anti-CPS1 (1:500, ab129076) antibody was purchased from Abcam (Shanghai, China). Anti-nuclear factor erythroid 2-related factor 2 (Nrf2) (1:2000, 33123-1-AP) antibody was purchased from Proteintech Group, Inc.(Wuhan, Hubei, China). Anti-β-actin (1:5000, 4970) antibody was purchased from Cell Signaling Technology (Shanghai, China). Annexin V-fluorescein isothiocyanate/propidium iodide (FITC/PI) apoptosis detection kit was purchased from MultiSciences Biotech Co., Ltd. (Hangzhou, Zhejiang, China).

2.2. Animal treatment

Male Sprague-Dawley (SD) rats (6−8 weeks of age, 200 ± 20 g) were purchased from Hangzhou Ziyuan Laboratory Animal Technology Co., Ltd. (Hangzhou, China; SCXK [Zhe] 2024–0004). All SD rats were housed in a controlled environment with constant temperature (25 ± 2 °C), humidity (55% ± 5%), and a 12 h light/dark cycle (7:00−19:00), with free access to food and water. All procedures performed in this study complied with the WMA Statement on animal use in biomedical research and the Guidelines for the Care and Use of Laboratory Animals. All animal experimental procedures complied with laboratory animal ethics guidelines and were approved by the Biomedical Ethics Committee of Naval Medical University, Shanghai, China.

2.3. Human samples

This study enrolled a total of 25 patients with liver injury and 21 healthy volunteers. The research protocol was approved by the Ethics Review Committee of the Second Affiliated Hospital of Naval Medical University (Shanghai, China) (Approval No.: 2023SL006). All plasma sample collection procedures were conducted in accordance with standardized protocols after participants had signed written informed consent forms and explicitly expressed voluntary participation. The inclusion/exclusion criteria for patients with liver injury, the selection criteria for healthy volunteers, and their demographic baseline characteristics were shown in the Section S1 in the Supplementary data and Table S1.

2.4. Biochemical and oxidative stress indicators index testing

Serum samples were thawed at room temperature, and ALT and AST levels in serum were measured using ALT and AST detection kits according to the instructions of the manufacturer. Liver tissue samples were homogenized, and SOD activity and MDA levels in liver tissue were measured using SOD and MDA detection kits according to the manufacturer's instructions.

2.5. H&E staining

The SD rats were anesthetized by intraperitoneal injection of 20% urethane (0.4 mL/100 g). Following cardiac perfusion with physiological saline, liver tissues were excised and immediately fixed in 4% paraformaldehyde. The fixed tissues were then dehydrated through a graded ethanol series, cleared in xylene, and embedded in paraffin. After deparaffinization and rehydration, the sections were subjected to H&E staining, followed by gradient ethanol dehydration, xylene clearing, and mounting with neutral balsam. The histopathological morphology of the liver tissues was observed under a light microscope (Nikon Corporation, Tokyo, Japan).

2.6. TUNEL staining

After dewaxing and hydration, the SD rat liver tissue sections underwent sequential treatments including Proteinase K digestion, post-fixation, and equilibration buffer incubation. Subsequently, TdT incubation buffer was added for TUNEL labeling at 37 °C with light protection for 1 h. Following the reaction, the sections were washed with PBS and counterstained with PBS for mounting. Finally, visualization was carried out using fluorescence microscopy (Nikon Corporation, Tokyo, Japan). Micrographs were acquired and TUNEL-positive cells were quantified using ImageJ software.

2.7. DHE staining

Frozen sections were cut at a thickness of 8 μm using a cryostat (Leica Bio-systems, Shanghai, China) and mounted onto glass slides. The sections were washed briefly with PBS to remove the embedding medium, then incubated with DHE in a light-protected humidified chamber at 37 °C for 30 min. After washing three times with PBS (5 min each), to remove unbound probes. Nuclei were stained with DAPI at room temperature for 5 min, followed by mounting with anti-fade mounting medium. The liver tissue sections were observed and micrograph images were acquired using a fluorescence microscope (Nikon Corporation, Tokyo, Japan). The fluorescence absorbance values were calculated using ImageJ software as an indicator of intracellular reactive oxygen species levels.

2.8. Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR)

Total messenger RNA (mRNA) was extracted from liver tissues using the MolPure TRIeasy™ Plus Total RNA Kit. Following the instructions of the manufacturer, RNA was reverse transcribed into complementary DNA (cDNA) using the reverse transcription kit. Subsequently, mRNA levels were detected by qRT-PCR using Hieff qPCR SYBR Green Master Mix on a QuantStudio 1 Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA, USA). Using glyceraldehyde-3-phosphate dehydrogenase as the internal reference, the expression of target genes in each sample group was normalized, and the relative expression levels of target genes were calculated using the 2^−ΔΔCt method. The sequences of primers used in qRT-PCR reactions were shown in Table S2.

2.9. Western blot

Protein extraction, quantification, preparation, and Western blot experiments were performed as described previously [20]. To quantitatively determine the expression level of the target protein, the grayscale intensity values of the target protein bands and the internal reference β-actin bands were measured using ImageJ software. All data were normalized based on the expression level of β-actin, and the relative expression of the target proteins were calculated.

2.10. Flow cytometry

Cell apoptosis was detected using the annexin V-FITC/PI Apoptosis Detection Kit according to the instructions of the manufacturer. In brief, cells were collected and centrifuged at 1000 rpm for 5 min at 4 °C. The cell pellets were washed twice with ice-cold PBS to remove residual medium and serum that may interfere with Annexin V binding. Subsequently, 4 μL of annexin V-FITC and 2.5 μL of PI were added to the cell suspension, followed by incubation at room temperature in the dark. After staining, the samples were analyzed by flow cytometry (POWCLIN, Hangzhou, China). The apoptosis rate was determined by quantifying the proportion of annexin V-positive cell populations.

2.11. Cell culture, treatment and transfection

BRL-3A cells were cultured at 37 °C with 5% CO₂ in medium containing 10% FBS and 1% penicillin-streptomycin. For the experiment, cells were seeded in six-well plates at a density of 4 × 10⁵ cells per well. After 24 h of culture, cells were washed twice with PBS and treated as follows: the blank control group was replaced with 2 mL of fresh complete medium; the inhibitor group received an equal volume of medium containing H3B-120; and the GP group was treated with medium containing 100 μM GP. The cell transfection protocol was as follows: cells were plated one day prior to transfection, and transfection was performed when cell confluence reached approximately 70%. The mixture of siRNA and Lipo 3000 was added to BRL-3A cells. To evaluate the gene silencing effect, cell samples were collected at 24 h and 48 h post-transfection, and their mRNA and protein expression levels were detected respectively for subsequent experiments. The siCps1 and siNC were shown in Table S3.

2.12. Lentiviral transfection

To establish a stable cell line with Cps1 overexpression, lentiviral vectors were employed. The experiment utilized a three-plasmid system (overexpression plasmid:psPAX2:pMD2.G = 4:3:1, m/m/m) for lentiviral packaging in 293T cells using lipo 3000 transfection reagent. The packaged lentiviral supernatant was centrifuged, filtered, and concentrated before being used to infect pre-seeded BRL-3A cells, with polybrene added to enhance infection efficiency. After 48 h of transfection, the cells were subjected to pressure selection by replacing the medium with complete medium containing puromycin until the negative control cells died completely. Efficient expression of the reporter gene in the experimental group cells was confirmed via fluorescence microscopy, thereby obtaining a cell line that stably overexpresses Cps1 that was used for subsequent experiments.

2.13. Untargeted metabolomics analysis based on ultra-high performance liquid chromatography-quadrupole/time-of-flight MS (UHPLC-Q/TOF-MS)

Approximately 20 mg of lyophilized liver tissue homogenate powder was weighed out to extract metabolites using 150 μL of saline, 600 μL of protein precipitant (acetonitrile:methanol = 2:1, v/v), and 30 μL of internal standard solution (10 μg/mL 4-Cl-phe methanol solution). The collected supernatant was dried under nitrogen gas and reconstituted with 100 μL of 50% (v/v) acetonitrile. The solution was centrifuged and the collected supernatant was used for MS analysis. The chromatographic column used was Waters AtlantisTM® T₃ (2.1 mm × 100 mm, 3 μm, Waters Corporation, Milford, MA, USA). The column temperature was set at 40 °C, with a mobile phase consisting of phase A (0.1% formic acid aqueous solution) and phase B (0.1% formic acid acetonitrile solution) at a flow rate of 0.4 mL/min. The gradient elution program was as follows: 0–2 min, 5%–10% B; 2–10 min, 10%–60% B; 10–15 min, 60%–65% B; 15–20 min, 65%–95% B; 20–23 min, 95% B; and 23–23.1 min, 95%–5% B. The injection volume of the sample solution was 2 μL. MS analysis was performed using electrospray ionization (ESI) in both positive and negative polarity modes, with a spray voltage of 80 V, ion source temperature of 325 °C, nebulizer gas flow rate of 11 L/min, nebulizer gas temperature of 350 °C, and nebulizer gas pressure of 45 psig. The scan range was set at m/z 50–1700. Metabolites with a variable importance in projection (VIP) > 1.3 and P < 0.05 were defined as differential metabolites between the two groups. Subsequently, these candidate differential metabolites were further identified by comparing them with the Human Metabolome Database (HMDB) and Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

We collected patient blood in anticoagulant tubes, centrifuged blood samples to obtain plasma, immediately froze and stored them in a −80 °C freezer for later use. Prior to analysis, the plasma samples were thawed at room temperature, and four times the sample volume of methanol and 10 μL of methanol solution containing 50 μg/mL 2,6-dihydroxybenzoic acid were added to precipitate proteins. The collected supernatant was dried under nitrogen gas, reconstituted with 200 μL of acetonitrile, centrifuged, and used for mass spectrometric analysis. The chromatographic column and mobile phase were the same as stated above. The gradient elution program was set as follows: 0–3 min, 5%–10% B; 3–15 min, 10%–65% B; 15–20 min, 65%–95% B; 20–20.1 min, 95% B; and 20.1–25 min, 95%–5% B. The injection volume was 3 μL. Ionization was performed using a heated ESI ion source, with data acquisition in positive ion mode. The capillary voltage was set at 380 V, capillary temperature at 320 °C, sheath gas flow rate at 30 arb, auxiliary nitrogen gas flow rate at 10 arb, and probe heater temperature at 300 °C. The scanning range was m/z 50–1000.

2.14. High-performance liquid chromatography-tandem MS (HPLC-MS/MS)-based targeted metabolomics analysis

Approximately 10 mg of lyophilized liver tissue homogenate powder was used to extract metabolites using 200 μL of saline, 800 μL of methanol solution, and 5 μL of internal standard solution (50 μg/mL 4-chloro-phenylalanine methanol solution). The collected supernatant was dried under nitrogen, and the residue was reconstituted with 1 mL of 50% (v/v) acetonitrile aqueous solution containing 0.2% (v/v) formic acid. After centrifugation, the supernatant was used for mass spectrometric analysis by employing a Shimadzu LC-40D XS system equipped with a Shimadzu 8045 triple quadrupole mass spectrometer (Shimadzu, Kyoto, Japan). An Inertsil HILIC column (GL Sciences Inc., Tokyo, Japan) was used with a column temperature of 40 °C and an injection volume of 1 μL. The mobile phase consisted of 0.2% formic acid in water (A) and 0.2% formic acid in acetonitrile (B) with a flow rate of 0.3 mL/min. The gradient elution program was set as follows: 0–5.1 min, 10%–30% A; 5.1–10 min, 30%–35% A; 10–10.2 min, 35%–95% A; 10.2–12.2 min, 95% A; 12.2–12.5 min, 95%–10% A; and 12.5–15 min, 10% A. Mass spectrometric analysis was performed using ESI in multiple reaction monitoring mode. The desolvation line temperature and heating block temperature were maintained at 250 and 400 °C, respectively. Other parameters included: nebulizer gas flow rate of 3 L/min, drying gas flow rate of 10 L/min, heating gas flow rate of 10 L/min, interface temperature of 300 °C, and desolvation temperature of 526 °C.

The pretreatment of SD rat serum was consistent with that of liver tissue. The mobile phase consisted of 0.35% formic acid aqueous solution (A) and 0.35% formic acid acetonitrile solution (B). The specific elution method (except for fumaric acid) was the same as above. For fumaric acid detection, an ACQUITY UPLC® T₃ column (2.1 mm × 100 mm, 1.8 μm; Waters Corporation, Milford, MA, USA) was employed. The gradient elution program was as follows: 0–3 min, 5% B; 3–3.1 min, 5%–23% B; 3.1–6 min, 23%–30% B; 6–6.2 min, 30%–95% B; 6.2–8 min, 95% B; 8–8.2 min, 95%–5% B; and 8.2–10 min, 5% B. The MS conditions were the same as stated above. After washing the SD rat hepatocyte samples with PBS, trypsin was added for digestion. The digestion was terminated with complete medium, followed by pipetting to achieve thorough mixing. The mixture was then transferred to a 1.5 mL microfuge tube. A 10 μL aliquot of the suspension was used for cell counting using an automated cell counter (MONWEI, Suzhou, Jiangsu, China), and the cell count was recorded. Subsequent processing was the same as stated before. The gradient elution program was as follows: 0–1 min, 10% A; 1–1.1 min, 10%–25% A; 1.1–4 min, 25% A; 4–4.2 min, 25%–40% A; 4.2–6 min, 40% A; 6–6.2 min, 40%–95% A; 6.2–8 min, 95% A; 8–8.2 min, 95%–10% A; and 8.2–10.5 min, 10% A. Other chromatographic and MS conditions were the same as stated above.

2.15. Shotgun proteomics

GP and NAL solutions were diluted with potassium phosphate buffer (pH 7.4, 0.1 mol/L) to prepare solutions A and B, with both having final concentrations at 5.0 mmol/L. Subsequently, the two solutions were pre-incubated at 37 °C for 2 min. A 200 mg of liver tissue sample was weighed and homogenized with 1.0 mL phosphate buffer (pH 7.4). After centrifugation of the homogenate at 9000 g for 10 min, 300 μL of the supernatant was collected and thoroughly mixed with 4 vol of glacial acetic acid followed by centrifugation at 10,000 g for 5 min to collect the resulting precipitate, which was then subjected to trypsin digestion sequentially. Subsequent analysis was performed using the previously established two-dimensional (2D) strong cation-exchange-nano-liquid chromatography-Q/TOF-MS (2D SCX-nano-LC-Q-TOF-MS) method [21]. In parallel, control samples without GP treatment were prepared as the experimental control group.

2.16. Data processing and statistical analysis

Statistical analysis and graph plotting were conducted using GraphPad Prism software version 8.0.2 (GraphPad software). The data were first tested for normality and homogeneity of variance. Unpaired Student's t-tests were performed to compare the differences between two groups. Three or more groups were compared using one-way analysis of variance (ANOVA) with post hoc Dunnett's. All experimental data were presented as mean ± standard deviation, with P < 0.05 considered to be statistically significant.

3. Results

3.1. GP-induced acute liver injury in SD rats

In this study, an acute liver injury model was established by administering GP via oral gavage to SD rats at a dose of 100 mg/kg body weight for three consecutive days (Fig. 2A). This dosage was determined based on preliminary dose-finding experiments (50, 100, and 200 mg/kg GP), and the detailed screening procedure and results were shown in Supplementary data (Fig. S1). Compared with the control group, rats treated with GP for 3 days exhibited loss of appetite, significant weight reduction (249.2 ± 7.4 vs. 217.2 ± 5.5 g, P < 0.001) (Fig. 2B), and blue-green feces. Histopathological examination through H&E staining demonstrated marked differences between the two groups. The control group displayed normal liver tissue structure with well-arranged hepatic cords, uniform hepatocytes, and no inflammatory cell infiltration or abnormal proliferation. In contrast, the liver tissues of SD rats in the GP administration group exhibited characteristics of acute liver injury, including disordered hepatic cord arrangement, complete destruction of the hepatic lobule radial structure, extensive ballooning degeneration of hepatocytes in necrotic areas, enhanced eosinophilic cytoplasm, and the presence of pyknosis or karyolysis (P < 0.001) (Fig. 2C). Serum biochemical analysis revealed significantly elevated levels of AST and ALT in the GP-treated group (ALT: 12.5 ± 2.3 vs. 772.3 ± 29.7 U/L, P < 0.001; AST: 20.0 ± 6.3 vs. 353.9 ± 75.45 U/L, P < 0.001) (Fig. 2D). In liver tissues, the activity of the antioxidant enzyme SOD was significantly decreased, while the content of MDA, a lipid peroxidation product, was markedly increased (SOD: 46.0 ± 3.6 vs. 25.7 ± 6.1 U/mg protein, P < 0.001; MDA: 0.2 ± 0.1 vs. 2.0 ± 1.1 nmol/mg protein, P < 0.001) (Fig. 2E). To investigate whether GP induces apoptosis, we performed TUNEL staining analysis. Representative fluorescence images revealed extensive aggregation of red fluorescent signals in the liver tissues of the GP-treated group, indicating widespread apoptosis. Quantitative analysis using ImageJ software confirmed that the apoptosis rate in the GP group was significantly higher than that in the control group (16.5 ± 2.9 vs. 78.2 ± 6.1, P < 0.001) (Fig. 2F). Additionally, this study evaluated the effect of GP on oxidative stress. Compared with the control group, obvious accumulation of yellow fluorescent signals was observed in the liver tissues of the GP group, suggesting elevated superoxide levels. Statistical analysis of fluorescence intensity further confirmed that GP treatment significantly increased hepatic superoxide production (1.0 ± 0.2 vs. 11.3 ± 0.7, P < 0.001) (Fig. 2G), demonstrating its induction of significant oxidative stress. The above results indicate that administration of GP for three days induced significant liver injury in SD rats.

Fig. 2.

Development and assessment of an acute liver injury rat model. (A) Animal grouping and dosing regimen. (B) Changes in body weight after three days of genipin (GP) administration (n = 6). (C) Hematoxylin and eosin (H&E) staining results of liver tissue in GP-induced liver injury rats (n = 3): representative images (left) and statistical analysis of histopathological scores (right). (D) Serum biochemical indices in GP-induced liver injury rats (n = 6). (E) Detection of oxidative stress markers in GP-induced liver injury rats liver tissues (n = 6). (F, G) Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining and semi-quantitative analysis (F) and dihydroethidium (DHE) staining and semi-quantitative analysis (G) of liver tissue in GP-induced liver injury rats (n = 3). Comparisons between the control and GP groups were performed using Student's t-test. Data are presented as mean ± standard deviation. ^∗∗∗P < 0.001. SD: Sprague-Dawley; ALT: alanine aminotransferase; AST: aspartate aminotransferase; SOD: superoxide dismutase; MDA: malondialdehyde; DAPI: 4′,6-diamidino-2-phenylindole.

3.2. GP-induced acute liver injury is associated with dysregulation of hepatic core amino acids and antioxidant metabolism

To investigate the mechanism of GP-induced acute liver injury, we conducted untargeted metabolomics analysis of SD rat liver tissues using UHPLC-Q/TOF-MS. Initially, principal component analysis and supervised orthogonal partial least squares-discriminant analysis analyses were performed on liver tissue metabolites from each group, with results shown in supplementary data (Fig. S2A). The quality control samples clustered tightly, while a significant separation trend was observed between the control and model groups, indicating stable instrument conditions during analysis and that GP treatment induced notable metabolic profile alterations in the liver. A volcano plot also revealed distinct metabolite composition differences among groups (Fig. 3A). Subsequently, the Soft Independent Modelling of Class Analogy online platform was used to obtain discriminant metabolite information such as retention time and characteristic ions based on the criteria of VIP ≥1.3 and P ≤ 0.05. Metabolite identification was performed by comparing with open-source web resources such as the Metabolite and Chemical Entity, HMDB, and KEGG databases, as well as matching with secondary spectra. The results revealed 74 differential metabolites between the Ct and GP groups, and the content changes of differential metabolites was shown in supplementary data (Fig. S2B and Table S4). To further elucidate the metabolic significance of these metabolites, the HMDB IDs of the 74 differential metabolites were imported into the MetaboAnalyst website (https://www.metaboanalyst.ca/) for metabolic pathway analysis. Significantly altered metabolic pathways (P < 0.05 and impact > 0.1) were selected as potentially disturbed pathways. A total of 25 metabolic pathways were identified in liver tissue, with arginine biosynthesis, alanine, aspartate and glutamate metabolism, and glutathione metabolism being the core pathways regulated by GP (Fig. 3B). The content changes of differential metabolites in these core pathways were shown in Fig. 3C.

Fig. 3.

Metabolic features of genipin (GP)-induced liver injury in rats versus drug-induced liver injury (DILI) in patients. (A) Volcano plot of differential metabolites. (B) Enrichment analysis of differentially expressed metabolites. (C) Heatmap of differential metabolites in key pathways. (D) Relative peak areas of key urea cycle metabolites. (E) Activity levels of key urea cycle enzymes in liver tissue from rats with GP-induced liver injury. (F) Heatmap of mean relative peak areas of urea cycle metabolites and activities of related enzymes in serum from GP-induced liver injury rats (up) and in plasma from patients with DILI (down). (G) Comparison of citrulline (Cit) relative peak area, and Cit/glutamine (Gln), and Cit/ornithine (Orn) ratios in liver tissue and serum from GP-induced liver injury rats (left) and in plasma from DILI patients (right). (H) Correlation analysis of Cit, Cit/Gln, and Cit/Orn with biochemical and oxidative stress indicators in GP-induced liver injury rats, and correlation analysis of Cit with biochemical indicators in the plasma of patients with DILI. Unpaired Student's t-tests were performed to compare the differences between two groups. Data are presented as mean ± standard deviation. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. TCA: tricarboxylic acid cycle; Glu: glutamate; Arg: arginine; Asp: aspartate; ASA: argininosuccinic acid; Fum: fumarate; Suc: succinate; GLS2: glutaminase 2; CPS1: carbamoyl phosphate synthetase 1; ARG: arginase; OTC: ornithine transcarbamylase; ASL: argininosuccinate lyase; ASS1: argininosuccinate synthase 1; SOD: superoxide dismutase; ALT: alanine aminotransferase; AST: Asp aminotransferase; MDA: malondialdehyde.

3.3. The urea cycle metabolism in SD rat models of acute liver injury exhibits significant disturbances

Based on the differentially regulated metabolites and core perturbed pathways identified through untargeted metabolomics analysis, abnormal arginine biosynthesis, as well as alanine, aspartate, and glutamate metabolism pathways were found to be key metabolic characteristics of GP-induced liver injury. The core substrates and products of these pathways participate in the regulatory network of the urea cycle, suggesting that a disturbed urea cycle may be one of the critical mechanisms underlying GP-induced acute liver injury. Building upon these findings, this study employed HPLC-MS/MS-based targeted metabolomics analysis to quantitatively assess the relative levels of key urea cycle metabolites in SD rat liver tissues and serum. The results (Fig. 3D) showed that in SD rat liver tissues, compared with the blank control group, the relative contents of glutamate, aspartate, and fumarate were significantly reduced in the GP group, while citrulline showed a decreasing trend but without statistical significance. Additionally, the relative arginine content was reduced, and succinate exhibited an increasing trend without a statistically significant difference. These results indicate that GP can disrupt the homeostasis of urea cycle metabolites in SD rat liver. Further analysis of metabolite ratios (Fig. 3E) revealed that the ratios of glutamate/glutamine, citrulline/ornithine, ornithine/arginine, citrulline/glutamine, arginine/argininosuccinate, and fumarate/argininosuccinate were significantly reduced, while the argininosuccinate/citrulline ratio was significantly increased. These ratio changes indicate that the activities of urea cycle-related enzymes such as glutaminase 2 (GLS2), CPS1, ornithine transcarbamylase (OTC), and argininosuccinate lyase (ASL) were significantly affected by GP treatment. Consistent with the liver tissue results, the relative serum levels of glutamate, aspartate, fumarate, and arginine in GP-treated SD rats were also significantly reduced, while citrulline showed a significant decreasing trend (Figs. 3F and S2C). Unlike liver tissue, the serum succinate content in GP group SD rats exhibited a decreasing trend (without reaching statistical significance), which differed from the altered ornithine trend in liver tissue. As a core intermediate product of the urea cycle, citrulline levels were reduced in liver tissues of SD rats with liver injury, SD rat serum, and human plasma samples. Importantly, both citrulline/ornithine and citrulline/glutamine ratios showed consistent downward trends (Fig. 3G). Our findings indicate that GP may affect the activity or expression levels of key enzymes (such as CPS1, GLS2, etc.) in the upstream and downstream arms of the citrulline pathway, thereby interfering with urea cycle homeostasis.

Our correlation analysis results (Fig. 3H) showed that the SD rat serum citrulline content, the citrulline/ornithine and citrulline/glutamine ratios were negatively correlated with ALT and AST activities. In recent years, researchers have gradually linked the metabolic changes of urea cycle disorder-related biomarkers with the mechanisms of oxidative stress. Previous studies have confirmed that alterations in citrulline levels can induce oxidative stress in the body, leading to dysfunction in both enzymatic and non-enzymatic antioxidant systems. Our present correlation analysis results in SD rats demonstrated that serum citrulline content, citrulline/ornithine and citrulline/glutamine ratios were positively correlated with SOD activity but negatively correlated with 3,4-methylenedioxyamphetamine content. The results indicate that the dysregulation of key enzymes in the citrulline upstream and downstream pathways during GP-induced liver injury may serve as a critical link mediating hepatic oxidative stress imbalance.

To determine whether fumarate supplementation could counteract GP-induced hepatotoxicity, rats were administered exogenous fumarate (100 mg/kg, i.g.) following GP treatment. Compared with the GP group, rats receiving fumarate intervention exhibited significantly reduced serum levels of ALT and AST (ALT: 12.5 ± 2.3 vs. 772.3 ± 29.7 vs. 28.0 ± 7.2 U/L, P < 0.001; AST: 20.0 ± 6.3 vs. 353.9 ± 75.45 vs. 83.0 ± 11.8 U/L, P < 0.001) (Fig. S3A), indicating amelioration of hepatocellular injury. In liver tissues, fumarate intervention significantly restored SOD activity and decreased MDA content compared with the GP group (SOD: 46.0 ± 3.6 vs. 25.7 ± 6.1 vs. 41.8 ± 3.2 U/mg protein, P < 0.001; MDA: 0.2 ± 0.1 vs. 2.0 ± 1.1 vs. 0.3 ± 0.1 nmol/mg protein, P < 0.001) (Fig. S3B), suggesting reversal of GP-induced oxidative stress. Histopathological examination through H&E staining revealed the GP group displayed characteristic features of acute liver injury including disordered hepatic cord arrangement, extensive ballooning degeneration, and inflammatory infiltration; fumarate treatment notably alleviated these pathological alterations, with preserved hepatic architecture and reduced necrotic areas (17.64 ± 2.7 vs. 3.1 ± 0.6, P < 0.001) (Fig. S3C).

3.4. GP disrupts urea cycle homeostasis by binding to CPS1

Studies have shown that GP contains hemiacetal groups that can be metabolized into highly reactive cis-enedione intermediates such as cis-enediol and γ-keto-enal, all of which possess an α,β-unsaturated carbonyl structure. This electrophilic structure readily forms stable covalent bonds with nucleophilic groups of intracellular proteins and free amino acids, affecting the activity, expression, and localization of target proteins. Based on the perturbation characteristics of the urea cycle pathway revealed by the aforementioned metabolomics analysis, this study hypothesized that GP may bind to key proteins in the urea cycle, inducing urea cycle dysfunction. Therefore, we further investigated the specific molecular mechanisms of GP-induced liver injury through shotgun proteomics of GP-characteristic adducts, aiming to identify its core targets that inhibit the urea cycle. In this study, GP was first co-incubated with NAL, and the products were analyzed by UHPLC-Q/TOF-MS. The GP-NAL adduct was successfully detected, and its potential structure was proposed based on MS/MS spectra (Figs. 4A and S4A), confirming its potential for covalent protein binding. However, the extremely low stoichiometry of covalent protein modifications in complex biological samples poses significant challenges for characterization. To rapidly and sensitively identify potential protein targets modified by GP, we first completely digested liver protein samples into free amino acids and peptides using pronase E and α-chymotrypsin. The digested products were then analyzed by 2D SCX-nano-LC-Q-TOF-MS, and a Mascot database search identified at least 17 proteins covalently modified by the reactive lysine metabolite (RM) of GP (Fig. S4B and Table S5). These modifications were further verified by accurate MS and characteristic MS/MS fragment ions.

Fig. 4.

Role and mechanism of carbamoyl-phosphate synthase 1 (CPS1) in genipin (GP)-induced hepatotoxicity. (A) Characterization of the potential GP-N-acetyl-L-cysteine (GP-NAL) tandem mass spectrometry (MS/MS) spectrum. (B) MS/MS spectrum of the GP reactive metabolite (RM)-modified peptide in protein CPS1. (C, D) Relative Cps1 messenger RNA (mRNA) expression (C) and Western blot image of the relative expression of CPS1 protein (D). (E, F) RelativeCps1 mRNA expression (E) and Western blot image of the relative expression of CPS1 protein CPS. (F) in siCps1-treated cells. (G, H) Relative Cps1 mRNA expression (G) and Western blot image of the relative expression of CPS1 protein in OE-Cps1 BRL-3A cells (H). (I) Flow cytometry analysis of apoptosis rates in BRL-3A cell following Cps1 knockdown and overexpression. Unpaired Student's t-tests were performed to compare the differences between two groups. Data are presented as mean ± standard deviation. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. siNC: small interfering RNA negative control; siCps1: small interfering RNA targeting Cps1; OE-NC: overexpression negative control; UL: upper left; UR: upper right; LL: lower left; LR: lower right; PI: propidium iodide; V-FITC: V-fluorescein isothiocyanate.

Among GP-modified proteins identified, we successfully screened out CPS1, a key enzyme in the upstream citrulline pathway of the urea cycle, and notably, no other key urea cycle enzymes were identified among the modified proteins, suggesting that GP may covalently bind with CPS1 with high selectivity, thereby interfering with the normal operation of the urea cycle and ultimately inducing its dysfunction (Fig. S5). Specifically, the quasi-molecular ion [M+4H]⁴⁺ detected at m/z 812.1592 could be matched to the 519–535 peptide segment (sequence: LFGDKLNEINEKIAPSFAVESMEDALK) of carbamoyl-phosphate synthase [ammonia], mitochondrial (CPSM_RAT). Compared with the unmodified form of this peptide segment, its mass increased by 307.0262 Da, indicating that the lysine (K) residue in the peptide underwent GP modification (Fig. 4B).

Therefore, we focused on the key metabolic enzyme CPS1 in the urea cycle pathway and conducted quantitative detection at both the mRNA and protein levels. The results showed that compared with the blank control group, both the mRNA and protein levels of CPS1 in the liver of GP-treated SD rats were significantly reduced (1.0 ± 0.1 vs. 0.4 ± 0.1, P < 0.001) (Figs. 4C and D). Based on these findings, and to clarify the cascade effects of CPS1 downregulation on the urea cycle pathway, we investigated the expression of its key upstream and downstream metabolic enzymes. The results revealed that core enzymes such as Gls2 and Otc exhibited feedback downregulation (Gls2: 1.0 ± 0.3 vs. 0.2 ± 0.1, P < 0.001; Otc: 1.1 ± 0.5 vs. 0.3 ± 0.1, P < 0.01) (Fig. S6). In summary, GP may inhibit the catalytic activity of CPS1 through covalent binding with it, triggering a cascade reaction in the urea cycle metabolic network, disrupting the homeostasis of the urea cycle, and ultimately leading to the occurrence of liver injury.

3.5. CPS1 is the core target mediating GP-induced liver injury

3.5.1. CPS1 expression level regulates the apoptotic phenotype

Previous studies have confirmed that GP may induce urea cycle disorder through covalent binding to CPS1. Based on the evidence of urea cycle dysfunction provided by metabolomics, and the proteomic finding that GP binds to CPS1, we ultimately identified CPS1 as the central target and performed in-depth mechanistic validation. We conducted gene interference-based targeted metabolomics analysis at the cellular level to examine the impact of CPS1 dysfunction on the homeostasis of urea cycle metabolites. First, qRT-PCR and immunoblotting were carried out to assess the knockdown efficiency of our gene interference strategy. Our qRT-PCR results showed that in BRL-3A cells, siCps1 reduced Cps1 mRNA levels by 73.25% (1.0 ± 0.5 vs. 0.3 ± 0.05, P < 0.05) (Figs. 4E and S7A). Western blot results further demonstrated that siCps1 could serve as an effective gene interference sequence for subsequent functional studies (1.0 ± 0.3 vs. 0.6 ± 0.3, P < 0.01) (Figs. 4F and S7B). The above data indicated that the designed siRNA effectively knocked down Cps1 expression at both the mRNA and protein levels. In the overexpression model, qRT-PCR and immunoblotting results confirmed that an exogenous Cps1 gene fragment had been successfully integrated via lentivirus mediation into the genome of BRL-3A cells, achieving high expression of CPS1 (1.1 ± 0.3 vs. 10 ± 1.2, 1.0 ± 0.2 vs. 3.1 ± 0.6, P < 0.001) (Figs. 4G and H). Next, we conducted functional experiments to evaluate the effects of Cps1 knockdown and overexpression on cellular behavior. Flow cytometry analysis revealed that Cps1 knockdown increased the apoptosis rate of BRL-3A cells by approximately 2.4-fold (16.0 ± 0.6 vs. 37.7 ± 4.7, P < 0.01), while Cps1 overexpression significantly decreased the apoptosis rate (21.6 ± 2.1 vs. 17.3 ± 1.4, P < 0.05) (Fig. 4I).

3.5.2. CPS1 expression interferes with the functioning of the urea cycle

The targeted metabolomics approach results (Fig. 5A) demonstrated that after genetic intervention, enzyme activity inhibitor treatment, and GP intervention, the alterations in amino acid levels of the urea cycle in BRL-3A cells displayed, in the most part, a trend that was consistent with that observed in the urea cycle at the animal level; namely, the levels of core intermediate metabolites were reduced significantly, indicating a reduced efficiency in urea cycle operation. Finally, to confirm the core mediating role of CPS1 in regulation of the urea cycle by GP, we carried out Cps1 overexpression experiments in BRL-3A cells. Specifically, by transfecting Cps1 overexpression vectors to significantly elevate intracellular CPS1 protein levels, followed by GP intervention, we detected changes in the levels of key urea cycle metabolites through HPLC-MS/MS targeted metabolomics. Our experimental results demonstrated that compared with the negative control group, cells overexpressing Cps1 followed by GP treatment showed no significant reduction in the levels of key urea cycle metabolites, indicating that the inhibitory effect of GP on the urea cycle was effectively reversed by Cps1 overexpression.

Fig. 5.

Effects of carbamoyl-phosphate synthase 1 (CPS1) interference on the urea cycle and oxidative stress. (A) Relative levels of urea cycle metabolites and activities of related enzymes (n = 6). (B) Relative messenger RNA (mRNA) expression of Kelch-like ECH-associated protein 1 (Keap1), nuclear factor erythroid 2-related factor 2 (Nrf2), heme oxygenase-1 (Ho-1), glutathione peroxidase 4 (Gpx4), and NAD(P)H:quinone oxidoreductase 1 (Nqo1) (n = 3). (C) Western blot image of the relative expression of Nrf2 protein (n = 3) in the four CPS1-interfered groups. Unpaired Student's t-tests were performed to compare the differences between two groups. Data are presented as mean ± standard deviation. ^∗P < 0.05, ^∗∗P < 0.01, and ^∗∗∗P < 0.001. Cit: citrulline; Gln: glutamine; Glu: glutamate; Fum: fumarate; ASA: argininosuccinic acid; Orn: ornithine; Asp: aspartate; Suc: succinate; Arg: arginine; siNC: small interfering RNA negative control; GP: genipin; siCps1: small interfering RNA targeting Cps1; OE-NC: overexpression negative control; OE-Cps1 : overexpression Cps1.

3.5.3. CPS1 expression mediates the oxidative stress process

Previous studies have reported a close association between metabolite homeostasis imbalance caused by urea cycle disorders and oxidative stress responses in the body, with both disturbances participating synergistically in the regulation of pathological processes that induce liver injury. Based on the altered pathology observations and correlation analysis results in this study, SD rats with GP-induced acute liver injury exhibited significant oxidative stress imbalance, and a clear dynamic correlation was observed between oxidative stress indicators and key metabolites of the urea cycle (such as citrulline) during liver injury progression. Based on the results of previous studies and our current experimental evidence, we proposed the following hypothesis: GP may interfere with the urea cycle by targeting CPS1, thereby inducing oxidative damage in hepatocytes and ultimately leading to liver injury. To verify this hypothesis, we measured the expression levels of oxidative stress-related genes using qRT-PCR. As shown in our results (Fig. 5B), compared with the negative control group and blank group, the transcriptional levels of kelch-like ECH-associated protein 1 (Keap1) and Nrf2 were significantly upregulated in the siCps1 group, H3B-120 group, and GP group, while the transcriptional levels of heme oxygenase 1 (Ho-1), NAD(P)H:quinone oxidoreductase 1 (Nqo1), and glutathione peroxidase 4 (Gpx4) were significantly downregulated. In the GP administration following Cps1 overexpression (OE-Cps1 + GP) group, the transcriptional level of Gpx4 was significantly upregulated compared with the negative control group; the transcriptional levels of Keap1, Nrf2, Ho-1, and Nqo1 also showed an upward trend but without statistical significance. Our immunoblotting data were consistent with the gene expression changes (Fig. 5C). The above results confirmed that CPS1, as a core regulatory target in GP-induced liver injury, interfered with the working of the urea cycle by binding to GP, disrupting the redox homeostasis of the organism, and thereby inducing liver injury.

3.6. Patients with DILI exhibit disorders in urea cycle metabolism

Preliminary animal experiments revealed that urea cycle disturbance may be the core regulatory mechanism of GP-induced acute liver injury. To validate the clinical relevance of this mechanism, we compared the relative plasma levels of key urea cycle metabolites between patients with DILI and healthy volunteers. Compared with healthy volunteers, the relative plasma levels of glutamine, citrulline, ornithine, and arginine were significantly reduced in patients with DILI, while those of aspartate and glutamate were significantly increased. An analysis of metabolite ratios showed that the plasma level trends of citrulline/ornithine, ornithine/arginine, and citrulline/glutamine ratios in patients with DILI were consistent with those in the serum of GP-treated rats, while the glutamate/glutamine ratio exhibited an opposite trend (Fig. 3F). Our data confirmed that patients with DILI also exhibited disturbances in urea cycle metabolites and related enzyme activities, which were partially consistent with the results of the GP-induced SD rat model. As a core intermediate product of the urea cycle, citrulline levels were decreased in liver tissues and serum of SD rats with liver injury, as well as in human plasma samples. Both citrulline/ornithine and citrulline/glutamine ratios showed consistent downward trends (Figs. 3G and S8). These findings suggest that GP may affect the activity or expression of key enzymes (such as CPS1 and GLS2) in the citrulline upstream and downstream pathways, disrupting urea cycle homeostasis. Our correlation analysis revealed (Fig. 3H) that citrulline levels in human plasma were negatively correlated with ALT and AST activities. Our current results suggest that citrulline has the potential to serve as a specific biomarker for DILI.

4. Discussion

This study establishes, for the first time, the core molecular mechanism underlying GP-induced hepatotoxicity. GP, the key metabolite of GE, irreversibly inhibits the urea cycle rate-limiting enzyme CPS1 through covalent modification and downregulates its expression, leading to urea cycle blockade, citrulline depletion, and reduced downstream fumarate production. As an endogenous signaling molecule, the decreased abundance of fumarate attenuates its activation of the Keap1-Nrf2 antioxidant pathway, resulting in transcriptional downregulation of antioxidant enzymes such as HO-1 and NQO1, thereby inducing oxidative damage. Consequently, the CPS1-fumarate-Nrf2 regulatory axis was identified, linking amino acid metabolic disruption to impaired antioxidant defense.

This study employed a rat model of acute liver injury induced by GP (100 mg/kg for three consecutive days). This dose is equivalent to an adult daily intake of approximately 12.23 g of raw Gardeniae Fructus, slightly exceeding the pharmacopoeia-recommended dose (6−10 g) [22]. This exposure level simulates clinical scenarios of high exposure resulting from irrational drug use, such as the combined use of multiple preparations containing Gardeniae Fructus or other GE-containing Chinese medicines, or failure to follow compatibility principles for toxicity reduction. It is noteworthy that although Gardeniae Fructus is classified as a “medicine and food homology” substance, the daily dietary intake level is far below that used in this model; therefore, safety assessment must consider the exposure level comprehensively.

Metabolomics revealed consistent metabolic profile characteristics in both liver tissue and serum of GP-treated rats, manifested as significantly decreased citrulline levels and reduced citrulline/ornithine and citrulline/glutamine ratios. Mechanistically, GP inhibits CPS1, reducing carbamoyl phosphate production and hindering the conversion of ornithine to citrulline, leading to citrulline depletion [23]. The decreased citrulline/ornithine ratio reflects CPS1 dysfunction and relative ornithine accumulation; the decreased citrulline/glutamine ratio results from compensatory glutamine catabolism triggered by CPS1 inhibition [24,25]. These metabolic features establish the central role of urea cycle dysfunction in GP-induced hepatotoxicity.

Proteomic analysis identified 17 proteins covalently modified by GP, including CPS1, betaine-homocysteine S-methyltransferase 1 (BHMT1), high mobility group box 1 (HMGB1), and apolipoprotein E (APOE), reflecting its broad electrophilic reactivity. Notably, among these modified proteins, no other key urea cycle enzymes were identified besides CPS1, suggesting a high degree of selectivity for targeting CPS1 within this pathway. However, although other urea cycle enzymes (e.g., GLS2 and OTC) were not directly modified, their activities and expression were also significantly reduced−likely a consequence of metabolic feedback or transcriptional dysregulation triggered by CPS1 inactivation. This suggests a dual mechanism of GP-induced urea cycle inhibition: direct covalent inhibition of the rate-limiting enzyme CPS1, and a subsequent dysfunctional cascade that indirectly suppresses other urea cycle enzymes. Combined with the finding that GP downregulates CPS1 mRNA and protein expression, this further confirms the central role of CPS1 in GP-induced urea cycle dysregulation. Given the complexity of GP's reactivity, it is plausible that modifications of the additional proteins (e.g., BHMT1 and HMGB1) contribute to the overall toxic effects and may synergize with CPS1 dysfunction to exacerbate liver injury [26,27]. However, based on the prominent role of CPS1 in mitochondrial ammonia detoxification and its direct association with oxidative stress observed in our study, we propose CPS1 as a primary target whose modification triggers a cascade of events leading to hepatotoxicity. Nevertheless, the specific transcriptional regulatory mechanisms underlying CPS1 downregulation remain unclear and will be an important direction for future research.

The hypocitrullinemia phenotype induced by GP closely resembles congenital urea cycle disorders caused by Cps1/Otc gene mutations and is also comparable to secondary urea cycle dysfunction induced by drugs such as valproic acid or pathological conditions [[28], [29], [30], [31]]. However, unlike genetic defects or reversible inhibition by drugs like valproic acid, GP irreversibly inhibits CPS1 through covalent modification. This irreversibility is crucial for the specificity and severity of its toxicity, positioning GP as a unique tool for studying acquired urea cycle disorders driven by covalent modification.

In vitro experiments showed that the metabolic profile of GP-treated BRL-3A cells was highly consistent with that of the CPS1 knockdown or enzyme inhibition groups, confirming its hepatocyte-autonomous regulatory effect. CPS1 overexpression partially reversed the urea cycle disruption induced by GP, clearly identifying CPS1 as the core target of GP-induced hepatic metabolic disorders. To verify the regulatory role of CPS1 on the Keap1-Nrf2 pathway, CPS1 overexpression combined with GP treatment was employed. Results showed that CPS1 overexpression reversed the GP-induced transcriptional downregulation of Nrf2 downstream antioxidant enzymes (HO-1, NQO1, and GPX4). Combined with the significant disruption of this pathway in CPS1 dysfunction models, this confirms that CPS1 is a critical node regulating the Keap1-Nrf2 pathway, and its functional integrity is essential for maintaining hepatocyte antioxidant defense.

Based on existing research, fumarate, as an intermediate product of the urea cycle, can covalently modify cysteine residues (Cys-151 and Cys-288) of the Keap1 protein, inducing conformational changes that promote Nrf2 nuclear translocation and transcription of downstream antioxidant enzymes (HO-1, NQO1, and GPX4) [[32], [33], [34]]. This study found that GP, by inhibiting CPS1, reduces endogenous fumarate production, attenuates its modifying effect on Keap1, hinders Nrf2 activation, diminishes antioxidant capacity, and ultimately induces oxidative damage in hepatocytes. Exogenous fumarate intervention significantly decreased serum transaminase levels, ameliorated liver histopathology, reversed oxidative stress indicators, and restored Nrf2 nuclear translocation in the GP rat model, confirming that CPS1 influences Keap1-Nrf2 pathway activity by regulating fumarate abundance. This provides direct evidence for the “metabolism-antioxidant” crosstalk. This mechanism parallels the action of the clinical drug dimethyl fumarate, which activates the Nrf2 pathway through exogenous modification of Keap1, a principle utilized in treating psoriasis and multiple sclerosis. Conversely, this study reveals the pathological manifestation of endogenous fumarate deficiency, namely weakened Nrf2 pathway function [[35], [36], [37]]. Notably, in fumarate hydratase-deficient tumors, aberrant accumulation of fumarate can overactivate Nrf2; together with the pathway inhibition observed in fumarate deficiency, this illustrates the concentration-dependent regulatory pattern of this signaling axis: its abundance must be maintained within a physiological range, too low impairs antioxidant defense while too high may promote tumor progression [[38], [39], [40], [41]].

In the plasma metabolic profile analysis of patients with DILI, the altered patterns of urea cycle metabolites were generally consistent with those observed in the rat model, particularly characterized by reduced citrulline levels and decreased citrulline/ornithine and citrulline/glutamine ratios. Citrulline is primarily synthesized in the mitochondria of hepatocytes and is rapidly effluxed via amino acid transporters, resulting in transiently low levels in liver tissue that are difficult to capture stably [[42], [43], [44]]. However, in the blood, due to circulatory accumulation, more pronounced differences are observed, making serum indicators more sensitive for reflecting intrahepatic urea cycle function. It is noteworthy that the clinical sample test results in this study showed considerable dispersion, suggesting the need for larger sample sizes and refined inclusion criteria in subsequent studies to more accurately assess the stability and diagnostic efficacy of these metabolic indicators. Nevertheless, these results still validate, at the clinical level, the urea cycle disorder mechanism suggested by animal models and indicate the potential of citrulline and related ratios as specific biomarkers for DILI.

It must be emphasized that the DILI patients included in this study all resulted from the irrational use of compound preparations containing Gardeniae Fructus. Compared to other classical models of DILI, liver injury caused by different etiologies exhibits significantly distinct metabolic profiles: acetaminophen overdose primarily acts through its toxic metabolite N-acetyl-4-benzoquinoneimine (NAPQI) depleting glutathione, with the core early metabolic feature being drastic glutathione depletion and accumulation of related metabolites [45,46]; Anti-tuberculosis drugs (e.g., isoniazid) are associated with the accumulation of toxic metabolites mediated by N-acetyltransferase 2 genotype, with metabolic disturbances often involving hydrazine metabolite accumulation, vitamin B6 metabolic interference, and carnitine metabolism abnormalities [47,48]; In contrast, GP-induced liver injury originates from urea cycle dysfunction, with citrulline depletion and decreased related ratios (citrulline/ornithine, citrulline/glutamine) as its core metabolic signature. Although amino acid metabolism abnormalities can also occur in acetaminophen and anti-tuberculosis drug models, urea cycle dysfunction is not their early and core change, nor is citrulline depletion a primary metabolic feature. This etiology-specific pattern of metabolic perturbation suggests that metabolomics technology holds promise for development into a molecular diagnostic tool for DILI subtyping. By identifying distinct metabolic profiles, it can achieve etiological classification, which has significant implications for guiding the clinical translation of citrulline and related ratios as biomarkers based on etiological stratification.

Liver injuries driven by different mechanisms require different treatment strategies. For instance, supplementation with exogenous fumarate effectively reverses the fumarate deficiency resulting from CPS1 inhibition and the subsequent weakening of downstream antioxidant defense in the GP model; N-acetylcysteine supplementation is the standard therapy for acetaminophen overdose, targeting the core mechanism of glutathione depletion by its toxic metabolite NAPQI [49]; Future strategies for anti-tuberculosis DILI may need to explore carnitine supplementation or interventions targeting mitochondrial energy metabolism [50]. This mechanism-oriented therapeutic concept represents the future direction for the clinical management of DILI. Based on this, subsequent phases of this research will systematically collect clinical samples from DILI cases of different etiologies, combined with corresponding animal models, to delve into the dynamic evolution patterns of key urea cycle metabolites under different initial injury mechanisms, aiming to construct an etiological stratification model for DILI based on metabolic profiling.

5. Conclusions

This study systematically elucidated, through integrated in vivo, in vitro, and clinical approaches, the core molecular mechanism of GP-induced hepatotoxicity and revealed a functional link between the urea cycle and hepatocellular antioxidant defense. GP was found to bind to and inhibit CPS1, causing urea cycle disruption and subsequently impairing redox homeostasis, establishing CPS1 as the central regulatory target of GP-mediated oxidative stress. Clinical validation confirmed these urea cycle abnormalities in patients with DILI from Gardenia Fructus-containing preparations. This positions CPS1 as a key toxicity target and its associated metabolic signature as a potential warning biomarker, while suggesting that exploring attenuating herb combinations or interventions targeting the fumarate/Nrf2 axis could support the safe clinical application of Gardenia Fructus preparations.

CRediT authorship contribution statement

Hongzhan Xu: Writing – original draft, Visualization, Methodology, Formal analysis, Data curation. Xinhui Huang: Writing – original draft, Validation, Supervision, Methodology, Formal analysis. Yujia Zhang: Writing – original draft, Visualization, Validation, Data curation. Shilin Gong: Validation, Methodology, Data curation. Yusha Luo: Validation, Methodology. Yichao Fang: Methodology. Xiaolan Hu: Validation, Data curation. Miaomiao Luo: Validation, Data curation. Yujia Zhai: Resources, Investigation. Fangyuan Gao: Investigation, Methodology. Xinglong Chen: Resources, Investigation. Rongping Zhang: Supervision, Resources. Longshan Zhao: Resources, Supervision. Yong Wang: Investigation, Resources. Jian-Lin Wu: Writing – review & editing, Resources. Xin Zhang: Supervision, Resources, Methodology. Huiqing Liang: Writing – review & editing, Methodology, Resources. Jun Wen: Writing – review & editing, Supervision, Methodology, Conceptualization. Tingting Zhou: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Conceptualization.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors utilized AI-assisted technologies solely to enhance the readability and language of the manuscript. Following the use of these tools, the authors carefully reviewed and revised the content as necessary and assume full responsibility for the final published work.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: