Dihydrotanshinone I Targets PGAM1 to Induce SYVN1-Mediated Ubiquitination and Suppress Glycolysis in Hepatocellular Carcinoma.

1
Introduction
Hepatocellular carcinoma (HCC) is a highly heterogeneous malignancy and a leading cause of cancer globally. It ranks sixth in incidence and third in mortality among all cancers, with limited therapeutic options and a poor prognosis (Sung et al. 2021). Despite advances in surgical techniques and systemic therapies such as sorafenib and immunotherapy, HCC treatment remains challenging due to high recurrence rates, resistance to therapies, and limited long‐term survival benefits (Forner et al. 2018). These limitations underscore the urgent need for novel therapeutic approaches targeting specific molecular pathways involved in HCC progression.
Dihydrotanshinone I (DHT), a diterpene quinone derived from
Salvia miltiorrhiza
(Danshen), is a major lipophilic phenanthraquinone compound with diverse pharmacological activities (Dong et al. 2011). Traditionally used in Chinese medicine,
S. miltiorrhiza
has been applied in the treatment of cardiovascular, hepatic, and neurodegenerative diseases (Chen et al. 2019). Recent studies have highlighted DHT's wide‐ranging therapeutic potential, including anticancer, anti‐inflammatory (Yuan et al. 2019), cardioprotective (Wang, Wang, et al. 2020), and anti‐fibrotic effects (Ge et al. 2017). Several molecular targets of DHT have been proposed, including hypoxia‐inducible factor 1‐alpha (HIF‐1α) (Jiang et al. 2019), human antigen R (HuR) (Lal et al. 2017), reflecting its multi‐target characteristics.
In oncology, DHT has demonstrated potent anti‐tumor effects in various solid tumors, including gallbladder cancer (Li et al. 2024), ovarian cancer (Zhao et al. 2022), and breast cancer (Wang, Xu, et al. 2020). Its anti‐cancer mechanisms include inhibiting proliferation, inducing apoptosis and ferroptosis, suppressing metastasis, and overcoming chemoresistance (Yue et al. 2024). Mechanistically, DHT induces apoptosis, inhibits invasion (Nie et al. 2024), and enhances immunotherapy efficacy (Han et al. 2022), and has been reported to interact with estrogen receptor alpha (ESR1) and modulate the Src kinase signaling pathway (Nie et al. 2024; Jiang et al. 2022). Recent pharmacokinetic studies show that DHT has favorable properties, including efficient tissue distribution, high bioavailability, and metabolic stability, supporting its potential as a therapeutic agent for liver cancer (Wang, Yu, et al. 2020). Despite growing evidence for its antitumor activity, the role of DHT in tumor metabolic regulation—particularly in HCC—remains poorly understood. To date, the metabolic mechanisms underlying the therapeutic effects of DHT in HCC have not been systematically explored.
Enhanced aerobic glycolysis is a hallmark of HCC, supplying energy for the rapid proliferation of cancer cells and contributing to tumor aggressiveness and poor outcomes (Feng et al. 2020; Pan et al. 2024). Thus, targeting glycolysis is a key therapeutic strategy for HCC. However, resistance to glycolysis inhibitors due to metabolic reprogramming limits their clinical effectiveness (Xia et al. 2021). Therefore, developing novel drugs with improved efficacy that can overcome resistance is critical for advancing HCC treatment (Yang, Zhang, et al. 2023; Wang and Deng 2023).
Phosphoglycerate mutase 1 (PGAM1), a crucial enzyme in glycolysis, facilitates the conversion of 3‐phosphoglycerate (3‐PG) to 2‐phosphoglycerate (2‐PG) (Yang et al. 2022; Wang, Shu, et al. 2024). Beyond its metabolic role in glycolysis, PGAM1 has been implicated in regulating cancer cell proliferation, migration, and invasion through non‐metabolic pathways (Zhang et al. 2017; Luo et al. 2023). Targeting PGAM1, therefore, not only disrupts cancer metabolism but also impairs key signaling networks involved in tumor progression. PGAM1 is overexpressed in various cancers, including HCC, where its expression is closely linked to tumor malignancy and an unfavorable prognosis (Niu et al. 2024; Zheng et al. 2023). Inhibiting its expression or activity has been shown to reduce cell proliferation and impair tumor growth in multiple cancers, including HCC, prostate cancer, and non‐small pancreatic ductal adenocarcinoma (PDAC) (Huang et al. 2019; Wang et al. 2018). Due to its roles in both metabolic and non‐metabolic pathways, PGAM1 represents a dual target in cancer therapy (Zheng et al. 2023).
Synovial apoptosis inhibitor 1 (SYVN1), an E3 ubiquitin ligase, has emerged as a key regulator of protein stability in various cellular processes (Zhao et al. 2020). Initially identified for its role in endoplasmic reticulum‐associated degradation (ERAD), SYVN1 has been implicated in cancer progression through its ability to modulate the stability of multiple target proteins (Zhu et al. 2025). However, its potential role in regulating metabolic enzymes in HCC has not been explored.
In this study, we aimed to elucidate the anti‐tumor mechanism of DHT in HCC, focusing on its impact on PGAM1. Through proteomics, molecular docking, and both in vitro and in vivo experiments, we found that DHT promotes SYVN1‐mediated ubiquitination and proteasomal degradation of PGAM1, thereby disrupting glycolysis and suppressing tumor growth. These findings uncover a novel metabolic mechanism and support the therapeutic potential of DHT in liver cancer.

2
Results
2.1
DHT Suppresses HCC Cell Proliferation and Glycolytic Metabolism
To evaluate the therapeutic potential of DHT in HCC, we first assessed its cytotoxicity in human HCC cell lines (Huh7, PLC/PRF/5, and Hep3B) (Figure 1A). Treatment with DHT (0–8 μM) for 24, 48, and 72 h resulted in dose‐ and time‐dependent inhibition of cell proliferation, with IC50 values of 3.440, 2.119, and 1.978 μM for Huh7, PLC/PRF/5, and Hep3B cells, respectively (Figure S1A,B). At 5 μM, DHT reduced cell viability to approximately 20%, 15%, and 20% in Huh7, PLC/PRF/5, and Hep3B cells after 72 h. EDU staining revealed that 5 μM DHT treatment reduced DNA synthesis by approximately 50%, decreasing EDU‐positive cells from 50% to 30% (p < 0.001) (Figure 1B). Furthermore, colony formation assays demonstrated that DHT (2.5 μM) significantly decreased colony numbers from approximately 200–50 colonies (p < 0.001) compared to control groups (Figure 1C).
To understand the metabolic mechanisms underlying DHT's anti‐tumor effects, we performed glycolysis stress tests using the Seahorse analyzer. Real‐time measurements of extracellular acidification rate (ECAR) showed that DHT treatment (1, 2.5, and 5 μM) dose‐dependently decreased glycolytic activity (Figure 1D). Quantitative analysis revealed significant reductions in overall glycolysis from 70 to 25 mpH/min (64% reduction, p < 0.001), glycolytic capacity from 140 to 45 mpH/min (68% reduction, p < 0.001), and glycolytic reserve from 75 to 15 mpH/min (80% reduction, p < 0.001) at 5 μM DHT (Figure 1E–G). Even non‐glycolytic acidification showed a significant decrease from 35 to 25 mpH/min (p < 0.01) (Figure 1H). These results suggest that DHT effectively suppresses glycolysis in HCC cells, prompting us to further investigate its underlying molecular mechanisms.

2.2
Proteomic Analysis Identifies PGAM1 as a Key Glycolytic Target of DHT in HCC Cells
To elucidate the molecular mechanisms underlying DHT's anti‐glycolytic effects, we performed label‐free quantitative proteomic analysis on Huh7 and PLC/PRF/5 cells treated with 5 μM DHT for 24 h. Among the overlapping differentially expressed proteins, PGAM1 exhibited the greatest absolute fold change and consistent downregulation in both HCC cell lines, highlighting it as a potential key metabolic target of DHT. Principal component analysis (PCA) demonstrated clear segregation between control and DHT‐treated groups, indicating substantial DHT‐induced proteome alterations (Figure 2A). Volcano plot analysis identified X upregulated and Y downregulated proteins, with PGAM1 showing consistent downregulation in both cell lines (Figure 2B).
Hierarchical clustering analysis of glycolysis‐related proteins revealed a distinct downregulation pattern in DHT‐treated samples, particularly among key glycolytic enzymes including PGAM1 (Figure 2C). Gene ontology (GO) analysis of differentially expressed proteins highlighted significant enrichment in metabolic processes, with the top altered pathways including oxidative phosphorylation, aerobic respiration, and ATP metabolism (Figure 2D). KEGG pathway analysis further confirmed the enrichment of glycolysis/gluconeogenesis and carbon metabolism pathways (Figure 2E). Cross‐comparison of differentially expressed proteins between both cell lines identified PGAM1 as one of the most significantly downregulated proteins shared across both HCC cell lines, suggesting it might be a key mediator of DHT's anti‐glycolytic effects.

2.3
PGAM1 Is Upregulated in HCC and Associated With Glycolytic Metabolism
Through integrated analysis of TCGA and GTEx databases, we observed significant upregulation of PGAM1 expression in HCC tissues compared to normal liver tissues (Figure 3A). Consistently, tumor stemness index (mRNAsi), which reflects cancer stem cell‐like properties, was markedly elevated in HCC tissues (Figure 3B).
To delineate the cellular heterogeneity of PGAM1 expression within the tumor microenvironment, we performed single‐cell RNA sequencing analysis. t‐SNE visualization demonstrated distinct PGAM1 expression patterns across different cell populations, with predominant expression in malignant cells, followed by endothelial cells and fibroblasts (Figure 3C), suggesting potential involvement of PGAM1 in tumor–stromal interactions.
To explore the molecular mechanisms underlying PGAM1 function, we conducted GO and KEGG pathway analyses. GO analysis revealed significant enrichment in RNA splicing, unfolded protein binding, and ubiquitin protein ligase binding (Figure 3D, right panel). KEGG analysis highlighted enrichment in metabolic pathways, particularly glycolysis/gluconeogenesis and retinol metabolism (Figure 3D, left panel). Furthermore, correlation analysis demonstrated strong positive associations between PGAM1 and key glycolytic regulators, including HK1, HK2, MTOR, PKM, and AMPK subunits (PRKAA1, PRKAA2, PRKAB1, and PRKAG1) (Figure 3E). These findings suggest that PGAM1 may function as a critical metabolic regulator in HCC progression through modulation of glycolytic metabolism and related signaling pathways.

2.4
DHT Suppresses HCC Progression by Targeting PGAM1‐Mediated Glycolysis
To investigate the biological functions of PGAM1 in HCC, we established PGAM1 knockdown (KD‐PGAM1) (Figure S2A) and overexpression (OE‐PGAM1) (Figure S2B) cell lines. CCK8 assays revealed that OE‐PGAM1 significantly promoted cell proliferation compared to control cells (Figure 4A). Metabolic analysis showed enhanced glucose consumption and increased lactate and ATP production in OE‐PGAM1 cells, while these effects were reversed in KD‐PGAM1 cells (Figure 4B).
Based on the correlation between PGAM1 expression and tumor stemness (Figure 3B), we further examined its role in HCC metastasis. Wound healing and Transwell invasion assays demonstrated that OE‐PGAM1 significantly enhanced cell migration and invasion, whereas KD‐PGAM1 suppressed these aggressive behaviors (Figure 4C,D).
We next explored the mechanistic link between PGAM1 and glycolysis through DHT treatment. Western blot analysis showed that DHT dose‐dependently decreased the expression of PGAM1 and key glycolytic enzymes (HK1, HK2, PKM1, and PKM2) (Figure 4E). Consequently, DHT treatment reduced HK and PK activities, decreased lactate and ATP production, and increased glucose retention (Figure 4F). These results demonstrate that PGAM1 regulates HCC progression through glycolytic metabolism, while DHT exhibits anti‐tumor effects by targeting the PGAM1‐mediated glycolytic pathway.

2.5
Rescue of DHT‐Induced Anti‐Proliferative and Glycolysis Inhibition Effects in HCC by PGAM1 Overexpression
To investigate PGAM1's role in DHT‐mediated effects in HCC, we examined PGAM1 expression in PGAM1‐overexpressing Huh7 and PLC cell lines (Figure 5A,B). DHT treatment significantly reduced PGAM1 expression in these overexpressing cells to control levels. CCK8 assays revealed that PGAM1 overexpression enhanced proliferation, while DHT treatment suppressed it. Notably, DHT‐treated PGAM1‐overexpressing cells showed proliferation rates equivalent to controls and significantly higher than DHT‐only treated cells, indicating PGAM1 overexpression attenuated DHT's anti‐proliferative effects (Figure 5C,D).
Metabolic analysis revealed DHT increased extracellular glucose while reducing intracellular lactate and ATP, suggesting glycolysis inhibition. Conversely, PGAM1 overexpression decreased medium glucose levels while increasing lactate production and ATP generation, reflecting enhanced glycolysis. In OE‐PGAM1 + DHT cells, all metabolic parameters were intermediate between DHT‐only and OE‐PGAM1‐only groups, demonstrating that PGAM1 overexpression partially reversed DHT's anti‐glycolytic effects (Figure 5E,F).
In conclusion, PGAM1 overexpression counteracts DHT's anti‐glycolytic effects, confirming PGAM1's crucial role in DHT‐regulated glycolysis in HCC cells through multiple mechanisms.

2.6
DHT Directly Binds to PGAM1 With High Affinity and Stability
Molecular docking analysis revealed that DHT exhibited strong binding affinity to PGAM1, with a binding energy of −9.3 kcal/mol, lower than both the positive inhibitor (−8.6 kcal/mol) and co‐crystallized ligand (−9.2 kcal/mol) (Figure 6A). A 200 ns molecular dynamics simulation demonstrated stable complex formation, with RMSD and RMSF values remaining within 1 nm (Figures 6B and 5C). Additional stability analyses, including radius of gyration (Rg) and solvent‐accessible surface area (SASA), further confirmed the structural integrity of the PGAM1‐DHT complex (Figure S2C,D). Consistent hydrogen bond formation throughout the simulation also supported stable binding (Figure 6D).
Free energy analysis revealed favorable binding regions (shown in red), with the PGAM1‐DHT complex forming a well‐defined energy cluster (Figure 6E). Energy decomposition demonstrated that van der Waals forces and electrostatic interactions were the primary stabilizing factors, despite minor opposing effects from polar solvents (Figure 6F). Three key residues—VAL‐112, TRP‐115, and ARG‐116—contributed significantly to DHT binding, with interaction energies of −2.06, −2.07, and −2.03 kcal/mol, respectively (Figure S2F).
Surface plasmon resonance (SPR) experiments confirmed the direct interaction between DHT and PGAM1. DHT showed concentration‐dependent binding (15.6–1000 nM) with clear association and dissociation phases (Figure 6G), yielding a dissociation constant (KD) of 697 nM (Figure 6H).

2.7
DHT Promotes PGAM1 Degradation Through SYVN1‐Mediated Ubiquitination
To investigate the mechanism of DHT‐induced PGAM1 downregulation, we first analyzed PGAM1 mRNA levels in DHT‐treated HCC cells. No significant changes in PGAM1 mRNA expression were observed (Figure 7A), suggesting post‐transcriptional regulation. Treatment with the proteasome inhibitor MG‐132 significantly attenuated DHT‐induced PGAM1 protein reduction (Figure 7B,C). Cycloheximide (CHX) chase assays demonstrated that PGAM1 protein remained stable for 24 h under normal conditions, while ubiquitination assays revealed that DHT treatment markedly enhanced PGAM1 ubiquitination (Figure 7D,E). These results indicate that DHT promotes PGAM1 degradation through the ubiquitin‐proteasome pathway.
To identify the E3 ubiquitin ligase responsible for DHT‐induced PGAM1 degradation, we utilized the UbiBrowser database and identified five candidates with confidence scores > 0.8 (Figure S1A,B). Among these, SYVN1, an endoplasmic reticulum‐resident E3 ligase critical for protein quality control (Ji et al. 2021), showed the highest predicted binding affinity to PGAM1 (Figure S1C). Molecular docking analysis revealed specific binding interfaces between SYVN1 and PGAM1, with a calculated binding free energy of −35.8 kcal/mol (Figure 7G,D). Immunoprecipitation analysis in Figure 7F demonstrated that DHT treatment significantly enriched K48‐linked, but not K63‐linked, ubiquitin chains on PGAM1, confirming K48‐mediated proteasomal degradation of PGAM1. Co‐IP experiments using FLAG‐PGAM1 and Myc‐SYVN1 plasmids demonstrated that DHT enhanced their interaction (Figure 7H). Furthermore, SYVN1 knockdown significantly suppressed DHT‐induced PGAM1 ubiquitination (Figure 7I), establishing SYVN1 as the key E3 ligase mediating PGAM1 degradation.

2.8
DHT Exhibits Potent Antitumor Activity in HCC Models Without Observable Short‐Term Toxicity
To evaluate the therapeutic efficacy of DHT in HCC, we established both orthotopic and subcutaneous xenograft models. In the orthotopic model, C57BL/6 mice received hydrodynamic tail vein injection of AKT/NRAS plasmids to induce hepatocarcinogenesis. Following a one‐week acclimation period, mice were randomized into three groups: vehicle control, AKT/NRAS, and AKT/NRAS+DHT (20 mg/kg, i.p.) (Figure 8A). DHT treatment significantly attenuated AKT/NRAS‐induced hepatic tumor burden (Figure 8B), while maintaining stable body weight across all groups throughout the study period (Figure 8C).
We further validated DHT's antitumor activity using a subcutaneous xenograft model with four treatment arms: vehicle control, DHT (10 mg/kg), DHT (20 mg/kg), and sorafenib (30 mg/kg) as positive control (n = 5 per group). Gross examination of resected tumors revealed substantial growth inhibition in both DHT‐ and sorafenib‐treated groups (Figure 8D). Although high‐dose DHT (20 mg/kg) showed lower efficacy than sorafenib (30 mg/kg) at the end of treatment, it still resulted in significant tumor growth suppression compared to the control group (Figure 8E), which was corroborated by terminal tumor weight analysis (Figure 8F). Safety assessment through serum biochemistry, including hepatic (ALT, AST) and renal (CRE, BUN) function markers, showed no significant alterations across treatment groups, indicating favorable systemic tolerability of DHT (Figure 8G).
Histopathological examination revealed that DHT treatment profoundly altered tumor morphology, as demonstrated by H&E staining. Immunohistochemical analyses showed reduced Ki67‐positive proliferating cells and increased TUNEL‐positive apoptotic cells in DHT‐treated tumors, accompanied by a marked reduction in PGAM1 expression (Figure 8H). Western blot analysis of glycolytic enzymes demonstrated significant downregulation of HK1, HK2, PKM1, PKM2, and PGAM1 in DHT‐treated liver tissues compared to the AKT/NRAS group (Figure 8I). Collectively, these findings demonstrate that DHT effectively suppresses HCC progression through glycolysis inhibition, with no observable short‐term toxicity in vivo.

3
Discussion
HCC is one of the most prevalent and lethal cancers worldwide, with a poor prognosis due to high recurrence rates, late‐stage diagnosis, and resistance to current therapies (Sung et al. 2021; Luo et al. 2024). Despite advancements in targeted therapies and immunotherapy, treatment options for advanced HCC remain limited, highlighting the need for novel therapeutic strategies (Tang et al. 2020; Meng et al. 2023). A key feature of HCC, and many other cancers, is metabolic reprogramming, where tumor cells predominantly rely on glycolysis for energy production, even under aerobic conditions—a phenomenon known as the “Warburg effect” (Koppenol et al. 2011). This metabolic shift supports rapid tumor growth and survival in nutrient‐deprived environments, making glycolysis a promising therapeutic target (Pavlova et al. 2022).
PGAM1, a critical enzyme in glycolysis, is often upregulated in HCC and other cancers, where its overexpression is linked to poor prognosis and aggressive tumor behavior (Hitosugi et al. 2013; Zheng et al. 2023; Yang et al. 2022). Beyond its role in metabolism, PGAM1 also regulates processes such as cell migration and immune responses, enhancing its potential as a therapeutic target (McGrail et al. 2022; Zhang et al. 2017).
DHT, a diterpene quinone from
Salvia miltiorrhiza
, has shown promising antitumor effects across various cancer types, with relatively low toxicity, making it an attractive candidate for therapeutic development (Sun et al. 2022; Wu et al. 2023). However, the detailed mechanisms underlying DHT's action have not been fully explored, particularly from a proteomic standpoint.
In this study, we systematically investigated the antitumor effects of DHT in HCC, with a particular focus on its role in metabolic regulation. Our initial findings established DHT's antiproliferative effects through multiple experimental approaches, including CCK‐8, EDU staining, and colony formation assays, which collectively demonstrated dose‐ and time‐dependent inhibition of HCC cell growth. Importantly, our metabolic analyses using Seahorse technology revealed that DHT significantly suppressed glycolytic activity, as evidenced by decreased ECAR levels, suggesting that DHT's antiproliferative effects might be intrinsically linked to its metabolic modulation.
Although previous studies have predominantly focused on the cytotoxic effects of DHT, or its interactions with specific targets such as ESR1 and the Src kinase pathway (Nie et al. 2024; Huang et al. 2023), our study uniquely explored its impact on cellular metabolism, particularly on glycolysis. Importantly, we employed an unbiased, high‐throughput proteomic screening approach, providing a more comprehensive perspective that enabled the identification of PGAM1 as a critical target that may have been overlooked by previous hypothesis‐driven investigations. We discovered that DHT triggers a series of metabolic stress responses through glycolysis inhibition. Specifically, DHT treatment led to elevated ROS levels (Figure S1F) and decreased mitochondrial membrane potential (Figure S1B), ultimately resulting in cell cycle arrest and apoptosis. These observations align with previous findings: cells compensatorily upregulate mitochondrial oxidative phosphorylation following glycolysis inhibition (Li et al. 2019), leading to increased ROS production and oxidative stress (Yang, Wang, et al. 2023). Prior research has demonstrated that metabolic stress induced by glycolytic disruption can activate apoptotic pathways (Toshida et al. 2024). While the precise causal relationships between these events require further investigation in our system, the concurrent observation of elevated ROS levels and mitochondrial dysfunction provides strong evidence supporting the glycolysis‐inhibitory effects of DHT.
To systematically investigate this mechanism, we employed proteomic analysis, which identified PGAM1 as a key mediator of DHT's metabolic effects. This finding was particularly significant given PGAM1's established overexpression in HCC, as validated through TCGA database analysis.
The functional significance of PGAM1 in HCC was conclusively demonstrated through gain‐ and loss‐of‐function experiments. PGAM1 overexpression enhanced both cell proliferation and glycolytic metabolism, while its knockdown produced opposite effects, establishing PGAM1 as a critical regulator of HCC cell growth and metabolism. Mechanistically, we provided multiple lines of evidence supporting PGAM1 as a direct target of DHT. Molecular docking, dynamic simulation analysis, and SPR experiments (KD = 697 nM) collectively demonstrated specific and stable binding between DHT and PGAM1.
A particularly novel finding was our elucidation of DHT's mechanism in regulating PGAM1 protein levels. We discovered that DHT promotes PGAM1 degradation through the ubiquitin‐proteasome pathway, specifically by enhancing its interaction with the E3 ligase SYVN1. This mechanism was rigorously validated through a series of experiments including ubiquitination assays, proteasome inhibition, and co‐immunoprecipitation studies. Notably, SYVN1 has been implicated in processes such as immune escape and metastasis in HCC cells. The modulation of SYVN1 may influence the tumor's ability to evade immune surveillance and contribute to its metastatic potential (Ji et al. 2021; Xie et al. 2023). While the direct correlation between SYVN1 and patient survival requires further investigation, these observations suggest that SYVN1 could play a broader role in HCC biology beyond metabolic regulation. Our study highlights a potential mechanism by which DHT influences HCC progression through the modulation of SYVN1 and PGAM1, although further research is needed to fully understand the clinical implications of this pathway. Given the involvement of ROS and AMPK pathways in tanshinones as therapeutic agents (Luo et al. 2025; Yun et al. 2014), these pathways may also contribute to DHT's anti‐tumor effects, and their potential role in PGAM1 degradation warrants further investigation. In addition to SYVN1, bioinformatic analysis identified other potential E3 ligases such as SMURF2 and STUB1. Both have been implicated in regulating protein stability and cancer progression in various contexts (Liao et al. 2023; Zhang et al. 2023). Protein–protein docking analysis suggested that SYVN1 exhibits a relatively stronger predicted binding affinity with PGAM1 compared to these candidates (Figure S3C), supporting its prioritization for experimental validation. However, we acknowledge that the involvement of other ligases cannot be excluded, and further experimental work will be needed to clarify their potential contributions to PGAM1 regulation.
The therapeutic potential of DHT was further validated in both orthotopic and subcutaneous HCC mouse models. In the AKT/NRAS‐induced orthotopic model (Bell et al. 2007; Molina‐Sánchez et al. 2020), DHT significantly reduced tumor burden without apparent toxicity. Moreover, in the subcutaneous xenograft model, DHT demonstrated dose‐dependent antitumor effects, with the higher dose (20 mg/kg) showing slightly lower efficacy to the standard‐of‐care drug sorafenib (30 mg/kg). Importantly, comprehensive analysis of serum biochemical markers (ALT, AST, CRE, and BUN) revealed no significant alterations across treatment groups, suggesting a lack of overt short‐term toxicity following DHT administration. These preliminary safety data warrant further investigation, particularly in the context of long‐term or repeated dosing. While our study provides compelling evidence for DHT's therapeutic potential in HCC through PGAM1 targeting, several questions remain to be addressed. First, while we established SYVN1's role in DHT‐induced PGAM1 degradation, the specific ubiquitination sites remain to be identified (Tracz and Bialek 2021; Deng et al. 2020; Li and Song 2020). Although SYVN1's involvement was confirmed through knockdown and protein interaction assays, overexpression‐based validation was not conducted due to experimental constraints, and will be prioritized in future studies. Second, future studies should explore DHT's potential in combination therapy settings, particularly with current standard‐of‐care treatments, and investigate the detailed molecular mechanisms of PGAM1 ubiquitination, which could inform the development of PGAM1‐targeted therapeutics. Third, although we demonstrated PGAM1 downregulation in tumor tissues after DHT treatment using immunohistochemistry and Western blotting, we acknowledge that direct genetic validation—such as PGAM1 overexpression or knockdown in animal models—would provide stronger causal evidence. This limitation has been noted and will be addressed in future preclinical studies. Notably, previous studies have reported that DHT modulates ESR1, Nrf2, and Src signaling pathways. While these pathways were not the focus of the current study, they may potentially intersect with the PGAM1–SYVN1 axis. In particular, Nrf2 has been implicated in metabolic regulation and proteostasis (He et al. 2020; Buttari et al. 2025), raising the possibility that it may influence PGAM1 expression or modulate SYVN1 transcriptional activity. Further investigations are warranted to explore these mechanistic connections.
Comparing DHT with other known PGAM1 inhibitors reveals several compelling advantages. Current PGAM1 inhibitors include the covalent modifier MJE3 (Evans et al. 2005), the validated but low‐potency inhibitor PGMI‐004A (Hitosugi et al. 2012), the tea‐derived polyphenol EGCG (Liu et al. 2017) with its associated selectivity challenges, and the recently developed xanthone‐based compounds (Yousaf et al. 2023). Our data demonstrate that DHT exhibits superior binding affinity (KD = 697 nM) compared to conventional inhibitors like PGMI‐004A. More importantly, DHT employs a unique dual‐action strategy against PGAM1—beyond direct enzyme binding, it specifically accelerates PGAM1 protein degradation through SYVN1‐mediated ubiquitination, a regulatory mechanism not observed with existing inhibitors. This distinctive mode of action potentially addresses the limitations of current PGAM1‐targeting agents, offering a promising new approach for metabolic intervention in HCC.

4
Materials and Methods
4.1
Cell Cultivation and Transfection
The human HCC cell lines Huh7, PLC/PRF/5, and Hep3B were sourced from the Liver Cancer Institute at Fudan University, Shanghai. Cells were maintained in DMEM (Yeasen, 41401ES76) enriched with 10% fetal bovine serum (Yeasen, 40130ES76) at 37°C in an incubator with 5% CO2. Stable knockdown and overexpression of PGAM1 were achieved via lentiviral infection. The PGAM1 overexpression plasmid was constructed using a pLV backbone driven by a CMV promoter and was purchased from Shanghai Genechem Co. Ltd. Transient transfections were performed using Lipofectamine 3000 (Invitrogen) according to the manufacturer's protocol. For knockdown, shRNA plasmids targeting PGAM1 were also obtained from Genechem. Transduced cells were selected with 5 μg/mL puromycin. The shRNA sequences used for silencing PGAM1 were provided by the manufacturer.

4.2
Cell Counting Kit‐8 (CCK‐8)
HCC cell viability was evaluated using the CCK‐8 (New Cell & Molecular Biotech, C6005). HCC cells (6000 per well) were seeded in 96‐well plates, incubated overnight, and then treated with varying concentrations of DHT for 24, 48, or 72 h. After treatment, 10% CCK‐8 dissolved in DMEM was added to each well, incubated for 2 h, and the absorbance was measured at 450 nm using a Tecan Spark microplate reader (Tecan Group Ltd., Switzerland).

4.3
Plate Clone Formation Assay
HCC cells were seeded at 2000 cells per well in 6‐well plates to achieve uniform distribution. The DMEM was refreshed every three days for both the control and DHT‐treated groups. Cells were cultured for 14 days until visible colonies formed at the well bottoms. Colony formation was observed microscopically, and colonies in the blank group were defined as formed when containing ≥ 50 cells. After two PBS washes, the cells were fixed with 4% paraformaldehyde for 20 min, followed by staining with crystal violet (Beyotime, C0121‐100 mL) for 20 min. Wells were rinsed with water to remove excess dye, which was subsequently discarded. Colonies were imaged, observed using an Olympus IX73 inverted microscope, counted with ImageJ, and statistically analyzed.

4.4
Measurement of Glucose Consumption, Lactate, and ATP Production in the Cell Supernatant
HCC cells, such as Huh7 and PLC/PRF/5, in the logarithmic phase were seeded into 6 cm culture dishes at a density of 1 × 106 cells per dish. Cells were grouped by drug concentration, with suspensions collected and counted. The DMEM was then centrifuged, and the supernatant was saved for analysis. Cells were lysed using RIPA buffer (Beyotime, P0013B) on ice for 30 min, followed by centrifugation at 12,000 × g for 10 min at 4°C to collect the lysates. Glucose, lactate, and ATP assays were conducted following the manufacturer's protocols, using the corresponding kits: glucose (Solarbio, BC2505), lactate (Solarbio, BC2235), and ATP (Solarbio, BC0300). OD readings were measured using a microplate reader at 550 nm for glucose, 570 nm for lactate, and 340 nm for ATP with a Tecan Spark microplate reader (Tecan Group Ltd., Switzerland). Huh7 and PLC/PRF/5 cell lines were selected for these assays due to their relatively higher baseline glycolytic activity compared to Hep3B, which was therefore not included in this experiment.

4.5
Detection of Intracellular HK and PK Activity
HCC cells, including Huh7 and PLC/PRF/5 in the logarithmic phase, were seeded in 6 cm dishes. Cell lysates were prepared according to drug concentrations, using hexokinase (HK) and pyruvate kinase (PK) activity assay kits (Solarbio, BC0745 and BC0545). Enzyme activity was determined by monitoring the change in absorbance over time following the manufacturer's protocol. HK activity (nmol/min/104 cells) was calculated using the formula: 0.32 × ΔA, and PK activity (nmol/min/104 cells) was calculated using the formula: 1.29 × ΔA, where ΔA represents the change in absorbance per minute. All enzymatic activity values were normalized to the number of cells.

4.6
Seahorse Analysis
Sensor plates were prepared by adding 200 μL of sterile water to each well and incubating overnight at 37°C in a CO2 incubator. The next day, the water was replaced with 200 μL XF calibration solution and hydrated for 45–60 min at 37°C in a CO2‐free incubator. For the assay, Agilent DMEM phenol red‐free medium was supplemented with 100× pyruvate and glutamine. Huh7 cells were seeded at a density of 5 × 104 cells per well in a Seahorse XF 24‐well plate with growth medium and incubated overnight. Huh7 cells were selected for Seahorse analysis due to their stable and reproducible glycolytic performance observed in preliminary assays. Cells were treated with DMSO or DHT for 12 h. Drug master solutions included glucose (100 mM), oligomycin (100 μM), and 2‐deoxyglucose (2‐DG, 500 mM), with oligomycin used at a final concentration of 20 μM. On the day of the experiment, cells were washed twice with assay medium, leaving a final volume of 180 μL per well. The plate was equilibrated at 37°C in a CO2‐free incubator for 60 min. Drugs were injected as follows: 20 μL glucose (port A), 22 μL oligomycin (port B), and 25 μL 2‐DG (port C). The sensor plate was then assembled with the cell plate, and Seahorse XF analysis was performed.

4.7
5‐Ethynyl‐2′‐Deoxyuridine (EDU) Incorporation Assay
The EdU incorporation assay was conducted to assess cell proliferation by evaluating DNA synthesis within cells. The BeyoClick EdU Cell Proliferation Kit contained Alexa Fluor 594 (Beyotime, C0078S). After exposing the cells to DHT (0, 2.5, and 5 μM) for 24 h, they were incubated with EdU (100 μM) for 2 h. Huh7 cells were selected for this assay due to their stable proliferation performance under the treatment conditions, as observed in preliminary experiments. Cells were seeded at a density of 2 × 105 per well in 6‐well plates. The cells were then immobilized in 4% paraformaldehyde (PFA) for 30 min and rinsed again with 2 mg/mL glycine for 3 min. Following a 10‐min incubation in 0.2% Triton X‐100, the cells were incubated with the reaction buffer for 30 min, protected from light exposure. The cells underwent three washes with 0.5% Triton X‐100 and were stained using Hoechst 33342 (1:1000) for 30 min. Fluorescence images were acquired using an Olympus IX73 fluorescence microscope at 10× magnification. EdU‐positive cells were quantified using ImageJ software based on red fluorescence signal intensity.

4.8
Scratch Wound Assay
Huh7 cells were seeded at an initial confluency of approximately 50%–60% in 6‐well plates and cultured until they reached 70%–80% confluency prior to scratching. A 200 μL pipette tip was used to scratch the 50%–60% confluent monolayer of HCC cells in 6‐well plates, followed by PBS washing to remove cell debris. The cells were treated with DMEM containing 0, 2.5, and 5 μM DHT for 0, 12, and 24 h. Images were captured using an Olympus IX73 inverted microscope at 10× magnification, and wound closure distance was quantified using ImageJ software.

4.9
Apoptosis Analysis
Huh7 cells were used in this assay due to their stable proliferation and adherence properties, as well as for consistency across experimental procedures. Cells were seeded at approximately 50%–60% confluency and cultured until reaching 70%–80% confluency prior to treatment. By using an apoptosis detection kit (Solarbio, CA1020), apoptosis was assessed in cells exposed to DHT at concentrations of 0, 2.5, and 5 μM for 24 h, with DMSO‐treated cells serving as negative controls. Following treatment, the cells were collected and placed in binding buffer. Cells were then stained with 5 μL Annexin V‐FITC and propidium iodide (PI) in the dark for 20 min. Apoptosis was quantified using a Beckman CytoFLEX flow cytometer, and the results were analyzed with FlowJo X10.0.7r2 software.

4.10
Cell Cycle Analysis
Huh7 cells were used in this assay due to their stable proliferation and strong adherence, and to ensure consistency across multiple experimental procedures. Cells were seeded in 6‐well plates at an initial confluency of approximately 50%–60% and cultured until reaching 70%–80% confluency, then treated with DHT (0, 2.5, 5 μM) for 24 h. DMSO‐treated cells served as controls.
After treatment, the cells were collected and fixed in 70% ethanol at 4°C overnight. Fixed cells were washed and stained with PI solution in the dark for 30 min at room temperature.
Cell cycle distribution was analyzed using a Beckman CytoFLEX flow cytometer, and the results were analyzed using FlowJo X10.0.7r2 software.

4.11
ROS Analysis
ROS levels were assessed using a Solarbio kit (CA1410). Cells were cultured in 6‐well plates to 50%–60% confluency, treated with DHT (0, 2.5, 5 μM) for 24 h, and stained with 5 μM DCFH‐DA in the dark. After staining, cells were centrifuged, washed, and resuspended in 1 mL PBS for analysis.

4.12
Proteomic Sample Preparation
HCC cell lines PLC/PRF/5 and Huh7 were used for proteomic analysis. After treatment, the culture medium was discarded, and cells were rinsed twice with PBS. On ice, cells in 6‐well plates were lysed in 200 μL RIPA buffer containing 1% Phosphatase Inhibitor Cocktail I (MedChemExpress, HY‐K0021). After 5 min on ice, lysates were collected into EP tubes and centrifuged at 14,000 rcf, 4°C for 30 min. From the supernatant, 150 μL was transferred and mixed with 750 μL pre‐chilled acetone, followed by protein precipitation at −20°C for 4 h. The precipitate was resuspended in 100 μL of 150 mM ammonium bicarbonate (ABC) containing 8 M urea, and the protein concentration was determined using the BCA assay kit (Thermo Scientific, A55864). For digestion, 100 μg of total protein (~90 μL) was reduced with 10 μL of 100 mM DTT (Inalco Spa, Milano, Italy, 1758‐9030) at 56°C for 30 min, followed by alkylation with 11 μL of 200 mM IAA (Sigma‐Aldrich, I3750) in the dark at 37°C for 30 min. Proteins were digested with Trypsin (Beijing Life Proteomic, HLSTRY001C) at an enzyme‐to‐protein ratio of 1:50 and incubated at 37°C for 14–16 h. The reaction was quenched with 1.1 μL of 10% TFA, and the mixture was centrifuged at 14,000 rcf, room temperature, for 10 min. Peptide desalting was performed using Sep‐Pak C18 Vac Cartridges (Waters, WAT023590; 100 mg sorbent, 55–105 μm particle size, 1 cc cartridge). Columns were activated with 400 μL ACN, equilibrated with 800 μL of 0.1% TFA, and the sample was loaded. Columns were washed twice with 200 μL of 0.1% TFA, and peptides were eluted sequentially with 200 μL of 50% ACN in 0.1% TFA and 200 μL of 80% ACN in 0.1% TFA. The eluates were combined and freeze‐dried for 2.5 h at −50°C and 0.2 Pa under 1000 rpm using a vacuum freeze dryer (CHRIST, Germany, RVC 2‐25 CDplus). The dried peptides were reconstituted in 100 μL of 0.1% FA and quantified using the Peptide Quantification Kit (Thermo Scientific, 23275). Finally, 200 ng of peptides were injected into a timsTOF Pro mass spectrometer (Bruker, USA) operated in DIA mode for 1 h.

4.13
Proteomic Data Analysis
Proteomic data analysis and visualization were performed using the Wukong platform (https://www.omicsolution.org/wkomics/main/) and R software. All raw MS data were processed using DIA‐NN (version 1.8.1) for peptide and protein identification. The UniProt protein database (Proteome ID: UP000005640, Organism:
Homo sapiens
, Organism ID: 9606, downloaded on November 6, 2023) was used as the reference database for spectrum matching. The precursor m/z range was set to 300–1800, and the digestion enzyme was specified as Trypsin/P. The false discovery rate (FDR) for both peptide‐spectrum matches (PSMs) and protein identifications was set to < 1% (FDR < 0.01). The neural network classifier was set to double‐pass mode, and match between runs (MBR) was enabled, with cross‐run normalization turned off. After identification, missing values in the proteomic dataset were assessed, and variables with more than 50% missing values were excluded. Remaining missing values were imputed using the K‐nearest neighbor (KNN) algorithm. Data were then normalized, and differential expression analysis was performed using the limma package in R. PCA, volcano plots, heatmaps, and intersection plots were generated for data visualization. Functional enrichment analyses, including GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment, were conducted on differentially expressed proteins, and visualized using bubble plots.

4.14
Molecular Docking
4.14.1
Protein‐Ligand Docking
The 3D structure of PGAM1 (PDB ID: 7XB8) was obtained from the PDB. PyMOL 2.3.0 was used to inspect it for docking. The DHT (CID: 5316743) structure was retrieved from PubChem and optimized with the MMFF94 force field in OpenBabel for energy minimization. AutoDock Tools 1.5.6 was used to add hydrogen atoms, assign rotatable bonds, and save both the protein and small molecule structures as pdbqt files. The binding site for PGAM1 was defined based on the co‐crystallized ligand in the PDB structure, and the docking grid was set as follows: Center (X, Y, Z) = (37.2, 28.5, 30.5) and Size (X × Y × Z) = (20.0 × 20.0 × 20.0). Semi‐flexible docking was conducted using the Lamarckian genetic algorithm, with the exhaustiveness set to 25. AutoDock Vina 1.2.0 was used for molecular docking, producing the binding free energy and docking result files. The methodology was validated by re‐docking the co‐crystallized ligand to PGAM1, yielding an RMSD value below 2 Å, confirming successful validation.

4.14.2
Protein–Protein Docking
Protein–protein docking was performed using the GRAMM web server, a rigid docking method that identifies potential binding sites on protein surfaces without altering the shape of the ligand or receptor. The binding sites were predicted based on GRAMM's shape‐complementarity and electrostatic scoring function. Protein sequences of PGAM1 and SYVN1 were obtained from UniProtKB based on gene names and subsequently submitted to SWISS‐MODEL to predict the optimal protein structures. The resulting PDB files were uploaded to GRAMM for docking, and 10 docking models were generated. The top‐ranked model was selected according to GRAMM's internal scoring function, which reflects interface energy and geometric complementarity. The reported binding score for the best model was −35.8 kcal/mol, indicating strong interaction stability. Although GRAMM does not compute true thermodynamic binding free energy, the score is a relative interaction energy estimate based on its rigid‐body matching algorithm. A value lower than −4 kcal/mol generally suggests acceptable binding affinity, and more negative values represent stronger interactions. These docking results suggest a stable surface interaction between PGAM1 and SYVN1, supporting their potential biological relevance. Relevant output files, including receptor‐ligand interaction data (receptor‐ligand.res), are provided in the Supporting Information for transparency.

4.15
Molecular Dynamics Simulation Steps
Molecular dynamics simulations of the PGAM1 protein‐DHT complex were performed using Gromacs 2022, applying Amber14sb to the protein and Gaff2 to the ligand. The system was solvated with the SPC/E water model in a 1.2 nm periodic boundary box and neutralized with sodium and chloride ions via Monte Carlo ion placement. Energy minimization was achieved with 50,000 steps until forces were reduced below 1000 kJ/mol. Pre‐equilibration was performed under constant volume and temperature (310 K) for 50,000 steps, followed by another 50,000 steps under constant pressure (one atmosphere) and temperature. A 100 ns molecular dynamics simulation was then conducted, with structural data saved every 10 ps for analysis. Parameters such as RMSD, RMSF, radius of gyration, hydrogen bonds, and free energy distribution were evaluated, and the binding free energy was calculated using the MM/GBSA method.

4.16
Western Blot
Cells or liver tissue samples were lysed in RIPA buffer with a phosphatase inhibitor mix for 30 min, then total protein was extracted and quantified using a BCA kit (Beyotime, P0012). After denaturing the proteins by boiling, SDS‐PAGE was performed, followed by membrane transfer. The membranes were first incubated in 5% non‐fat milk for 1 h to block nonspecific binding, followed by overnight incubation with the primary antibody at 4°C for 16 h, followed by a 1‐h incubation with the secondary antibody. The membranes were then developed with NcmECL Ultra Reagent A/B (P10200), and images were captured after exposure. Western blot analyses of HK1, HK2, PKM1, and PKM2 were performed using Huh7 cells due to their stable and moderate basal expression of these enzymes, while other experiments such as co‐immunoprecipitation involved both Huh7 and PLC/PRF/5 cells. The antibodies used included HK1 (#2024), HK2 (#2867), PKM1 (#7067), PKM2 (#4053), SYVN1 (#14773), DYKDDDDK‐Tag (#14793), HA‐Tag (#3724), and HRP‐labeled goat anti‐rabbit IgG (#7074) from Cell Signaling Technology, as well as PGAM1 (A4170) from Abclonal, and MYC tag antibody (60003‐2‐Ig) from Proteintech.

4.17
Surface Plasmon Resonance (SPR)
SPR experiments were conducted using the Biacore 8K system (GE Healthcare, Piscataway, NJ, USA). Purified PGAM1 protein (20 μg/mL, pH 4.5) was immobilized onto a Series S Sensor Chip (GE Healthcare, Piscataway, NJ, USA) via standard amine‐coupling, achieving a final immobilization of approximately 15,000 response units (RU). A running buffer of PBS (BR100672, pH 7.2–7.4, Cytiva) containing 1% DMSO was employed for this immobilization process. Following this, a solution of DHT was prepared by diluting the stock in the same buffer. Eight different concentrations of DHT were then injected simultaneously at a flow rate of 20 μL/min for a 100‐s association phase at 25°C. Blank sensorgrams were subtracted to generate the final binding curves. Data were gathered using the Biacore 8K Manager software (GE Healthcare) and fitted using the 1:1 Langmuir binding model. The kinetic parameters were as follows: association rate constant (ka) = 3.96 × 104 M−1·s−1, dissociation rate constant (kd) = 2.86 × 10−2 s−1, and equilibrium dissociation constant (KD) = 7.22 × 10−7 M. The binding kinetics (ka, kd, KD) were calculated using the 1:1 Langmuir model. Full kinetic data are presented in Figure S2E.

4.18
Hydrodynamic Transfection and DHT Treatment
C57BL/6 mice (SiPeiFu, B203‐02), aged 4–5 weeks, received 2 mL of saline containing 16 μg pT3‐EF1α‐HA‐myr‐AKT, 16 μg pT3‐EF1α‐NRASV12, and 5 μg pCMV/SB via hydrodynamic injection. To evaluate therapeutic efficacy, 4 weeks after transfection, DHT (MedChemExpress, HY‐N0360, C18H14O3, 99.57%) was given intraperitoneally at 20 mg/kg/day, or solvent vehicle, once daily (n = 6 per group), with body weight tracked until endpoint. The solvent vehicle was composed of 40% PEG400, 10% DMSO, and 50% physiological saline.

4.19
Xenograft Tumor Model and Biochemical Analysis
A xenograft model was established by subcutaneous injection of HCC cells (Huh7, 5 × 106 cells per mouse) into mice (n = 5 per group). The mice were randomly allocated into four treatment groups: control, DHT (10 mg/kg), DHT (20 mg/kg), and sorafenib (30 mg/kg). DHT was administered via intraperitoneal (IP) injection at the indicated doses. Treatment was administered for 4 weeks. Tumor growth was monitored twice weekly using caliper measurements, with tumor volume calculated as 0.5 × width2 × length. At study endpoint, blood was collected via retro‐orbital sampling for serum analysis of liver function markers (ALT, AST) and kidney function indicators (CRE, BUN). Serum levels of SCR, BUN, AST, and ALT were measured using colorimetric assay kits according to the manufacturer's instructions. The following kits from Nanjing Jiancheng Bioengineering Institute were used: CRE (C011‐2‐1), BUN (C013‐2‐1), AST (C010‐2‐1), and ALT (C009‐2‐1).

4.20
Histological Analysis and Immunohistochemical Staining
Liver tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned at 5 μm thickness. Sections were stained with hematoxylin and eosin (H&E), KI67 (CST, #9129, 1:400), and PGAM1 (Abcam, A4170, 1:800), following standard immunohistochemistry protocols. Antigen retrieval was performed by heat‐induced epitope retrieval using citrate buffer (pH 6.0 or pH 9.0, depending on antibody) in a pressure cooker or microwave, and endogenous peroxidase activity was blocked using 3% hydrogen peroxide. Slides were blocked with 10% goat serum at 37°C for 20 min and incubated with primary antibodies overnight at 4°C. After washing, sections were incubated with HRP‐conjugated secondary antibody (goat anti‐rabbit IgG‐HRP, Abcam, ab205718, 1:2000) for 45 min at 37°C. DAB chromogen (Fuzhou Maixin, DAB‐4033, 1:20) was used for color development, followed by hematoxylin counterstaining. TUNEL staining was performed using the Roche In Situ Cell Death Detection Kit (Roche, 11684817910) following the manufacturer's enzymatic colorimetric protocol. The working solution was freshly prepared at a ratio of A:B:TBST = 1:100:100. After permeabilization and blocking, slides were incubated with the TUNEL reaction mixture for 60 min at 37°C in the dark. DAB was used for signal detection. All stained slides were observed and imaged using an Olympus CX31 bright‐field microscope at 100× and 400× magnifications.

4.21
Data Acquisition, Processing, Differential Analysis, Immune Correlation Analysis, Single‐Cell Analysis, and GO/KEGG Pathway Enrichment
This study downloaded STAR‐counts data and corresponding clinical information for HCC from the TCGA database (https://portal.gdc.cancer.gov) (Cancer Genome Atlas Research 2017), and extracted TPM format data. After log2 (TPM + 1) normalization, 647 samples with both RNA‐seq data and clinical information were selected, including 371 HCC samples with high PGAM1 expression and 276 normal tissue samples. Additionally, the study used GTEx data (version V8), with related datasets available for review on the GTEx portal (https://gtexportal.org/home/datasets) (The GTEx Consortium 2020). All statistical analyses were performed using R software (version 4.0.3). For single‐cell transcriptomic analysis, we downloaded the corresponding single‐cell data (.h5 format) and annotation information from the TISCH database (Zhang et al. 2020), specifically the GSE125449 dataset, which includes annotated single‐cell RNA‐seq data from HCC samples. The data (.h5 format) and accompanying cell‐type annotation files were processed using the R packages MAESTRO and Seurat. t‐SNE dimensionality reduction was applied for re‐clustering and visualization. Additionally, to reveal the functions and pathways of differentially expressed genes, we performed GO and KEGG pathway enrichment analyses. All enrichment analyses were conducted using the ClusterProfiler R package. All results were considered statistically significant when the p‐value was < 0.05.

4.22
Statistical Analysis
All data are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism 9. Comparisons between two groups were conducted using two‐tailed unpaired Student's t‐test. For comparisons involving multiple groups, one‐way analysis of variance (ANOVA) was utilized, followed by appropriate post hoc tests as indicated in the figure legends. P‐values less than 0.05 were considered statistically significant and are denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001.
This study followed the best practice guidelines in natural products pharmacological research as outlined by Heinrich et al. (2020) and Wang, Izzo, et al. (2024).

Author Contributions

Ru Xu: conceptualization, methodology, investigation, validation, visualization, writing – original draft. Jiawei Dai: software, resources, validation, methodology. Ruijie Gong: investigation, writing – original draft, formal analysis, data curation. Ruoxin Tu: conceptualization, validation, resources, visualization, methodology. Qiaozi Wang: methodology, validation, resources. Hongdan Zheng: methodology, validation, visualization. Li Zhou: validation, visualization, investigation. Shusheng Wang: validation, formal analysis, data curation. Jiabin Cai: resources, project administration, writing – review and editing. Haixiang Sun: supervision, resources, project administration, writing – review and editing. Pingting Gao: resources, supervision, project administration, writing – review and editing. Pengfei Gao: resources, writing – review and editing, project administration, formal analysis, conceptualization.

Ethics Statement
All experiments involving mice were conducted in accordance with the ethical guidelines and procedures approved by the Animal Welfare and Ethics Committee of the Department of Experimental Animal Science, Fudan University (approval number: 2024‐JSYY‐001). All experiments complied with the WMA Statement on animal use in biomedical research.

Conflicts of Interest
The authors declare no conflicts of interest.

Supporting information

Figure S1. DHT inhibits HCC cell viability, disrupts mitochondrial function, alters cell cycle progression, promotes apoptosis, and elevates ROS levels.

Figure S2. Validation of KD‐PGAM1 and OE‐PGAM1 cell lines and DHT–PGAM1 interactions in molecular dynamics simulations.

Figure S3. Prediction and comparative analysis of E3 ubiquitin ligases targeting PGAM1.