Cell culture. The human hepatoma cell line HepG2 (ATCC) and HAOECs (ATCC; PCS-100-011) were obtained. Huh7 cells and HUVECs were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). GSTZ1-KO HepG2 cells were generated by lentiCRISPR system as described previously (22). A recombinant adenovirus expressing GSTZ1 (AdGSTZ1) was generated using the AdEasy system in Huh7 cells, which endogenously express low levels of GSTZ1, and were infected with AdGSTZ1 to establish the GSTZ1-OE cell model. An analogous adenovirus expressing green fluorescent protein was used as control. Cells were maintained in MEM (for HepG2; Hyclone) or DMEM (for Huh7 and HUVEC; Hyclone) supplemented with 10% FBS (Natocor) and 1% penicillin/streptomycin (Hyclone) in a 5% CO2 incubator at 37°C. For the hypoxic cultures, cells were cultured in a hypoxia chamber flushed with a humidified gas mixture of 1% O2, 5% CO2, and 95% N2 at 37°C for 12 hours.

Patient specimens. A total of 42 HCC and paired adjacent noncancerous tissues were obtained from The Second Affiliated Hospital of Chongqing Medical University between 2015 and 2021. Each sample was frozen immediately after surgery and stored in liquid nitrogen for later use. IHC staining was performed in a tissue microarray of HCC patient samples purchased from Shanghai Ming Yi Biotech (catalog LVC1805).

TCGA data analysis. GSTZ1, VEGFA, and HIF-1α mRNA expression profiles were obtained from TCGA LIHC data set. The Kaplan-Meier survival curves were generated with the survminer R package (Version 3.6.3).

RNA-Seq analysis. GSTZ1-KO HepG2 cells and parental cells were cultured under hypoxic conditions for 12 hours. Then, total RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. RNA-Seq was performed at the Shanghai Novel Bio-Pharm Technology Co., Ltd, in Shanghai, China. Briefly, strand-specific RNA-Seq libraries were prepared using the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs). Six samples were sequenced on a HiSeq 4000 sequencing platform (Illumina). A bioinformatics pipeline was used to analyze differential gene expression data between GSTZ1-KO and parental cells. The RNA-Seq data were submitted to the Gene Expression Omnibus database under the accession number GSE192760.

IHC and immunofluorescence staining. GSTZ1, VEGFA, HIF-1α, PD-L1, PD-1, and CD31 were detected by IHC as previously described (50). In brief, liver tissue samples from paraffin-embedded human or mouse tumors were incubated at 4°C overnight with dilutions of the indicated primary antibodies (anti-VEGFA, ab1316, Abcam; anti–HIF-1α, ab51608, Abcam; anti–PD-L1, 13684, Cell Signaling Technology; anti–mPD-1, AF1021, R&D Systems, Bio-Techne; anti-CD31, 77699, Cell Signaling Technology). The slides were then incubated with a secondary anti-rabbit or anti-mouse IgG (ZSGB-BIO) and visualized using 3,3′-diaminobenzidine (ZSGB-BIO). Stained slides were scanned using the Motic Easy Scanner. Images were acquired using DSA assistant Lite (Motic VM V1 Viewer 2.0).

For tissue immunofluorescence staining, freshly prepared frozen liver sections were blocked with 5% goat serum for 1 hour at 25°C, then incubated with anti-CD31 (at 1:200 dilution in PBS, ab222783, Abcam) overnight at 4°C. Next, bound primary antibody was detected using an Alexa Fluor 488–conjugated goat anti-mouse IgG (A-10680, Invitrogen), and the nuclei were stained with DAPI (Roche, 1:200). The samples were then mounted with Fluoromount-G (Southern Biotech), and immunofluorescence images were acquired using a confocal microscope (FV3000, Olympus). The images were further processed using ImageJ 1.49v (NIH) or OlyVIA VS200 (Olympus).

siRNA and plasmid DNA transfection. HepG2 cells plated in collagen-coated 100 mm dishes at a density of 1 × 106 cells were incubated overnight in complete MEM. Cells were then washed once with MEM, then incubated with 4 mL of Opti-MEM (Gibco) containing transfection mixture: 25 μL of Oligofectamine (Invitrogen) and 25 μL of 10 μM siRNAs targeting HIF-1α in 600 μL of Opti-MEM. At 48 hours posttransfection, cells were harvested and assayed for immunoblotting. For plasmid DNA transfection, Huh7 cells were transfected with 5 μg of GSTZ1 plasmid using Lipofectamine 3000 (Invitrogen) following the manufacturer’s protocol. At 36 hours posttransfection cells were cultured under hypoxic conditions for another 12 hours before harvest.

qRT-PCR. Total RNA was extracted from cell lines and frozen tumor specimens using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Purified RNA samples were reverse-transcribed into cDNA using the PrimeScript RT Reagent Kit with gDNA Eraser (RR047A, Takara). Quantitative real-time PCR analysis of target genes was performed using the SYBR Green qPCR Master Mix (Bio-Rad) with specific primers (Supplemental Table 1). The qPCR reaction was performed in 10 μL of reaction mixture: 2 μL of cDNA, 0.5 μL each of 10 μM forward and reverse primers, 5 μL of iTaq Universal SYBR Green Supermix (Bio-Rad), and 2 μL of nuclease-free water. The reaction conditions were as follows: 1 cycle at 9°C for 30 seconds, followed by 35 cycles of amplification at 95°C for 10 seconds, at 62°C for 30 seconds, and then at 72°C for 30 seconds. The relative gene expression was calculated using the ΔΔCT method using β-actin as the reference gene for normalization. Each sample was analyzed in triplicate.

Western blotting. All procedures were performed as we previously reported (50). Cytoplasmic and nuclear proteins were extracted using the Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime) in accordance with the manufacturer’s instructions. The steps are roughly as follows: cells were scraped off with cell scraper, washed with PBS, and collected by centrifugation at 100g for 2 minutes at 4°C. Then, according to the manufacturer’s instructions, cytoplasmic protein extraction reagents A containing PMSF and cytoplasmic protein extraction reagents B were added successively, and the supernatant absorbed after vortex centrifugation was the cytoplasmic protein. The nuclear protein extraction reagent containing PMSF was added to the precipitation, and the supernatant absorbed from the precipitation after vortex centrifugation was nuclear protein. Antibodies were used as follows: anti-GSTZ1 (GTX106109, GeneTex), anti-VEGFA (ab1316, Abcam), anti–HIF-1α (36169, Cell Signaling Technology), anti–Hydroxy–HIF-1α (3434, Cell Signaling Technology), anti–β-actin (TA-09, ZSGB-BIO), anti-GAPDH (AG019, Beyotime), anti-PHD2 (4835, Cell Signaling Technology), anti-MMP2 (GTX133806, GeneTex), and anti-MMP9 (S1241, Bioworld). The protein samples were separated by 10% SDS-PAGE and transferred to PVDF membranes (MilliporeSigma). The immunoblots then were probed with the indicated antibodies. Finally, protein bands were visualized with ultrasensitive chemiluminescence substrate kits (Biosharp).

ELISA. Cells were cultured under hypoxic conditions for 12 hours, and cell supernatants were collected for VEGFA detection. The amount of VEGFA in the CM was determined using the Human VEGF Quantikine ELISA Kit (Proteintech) by measuring absorbance values at 450 nm in accordance with the manufacturer’s instructions.

Wound scratch assay. HUVECs and HAOECs (10,000 cells/well) were cultured with CM from hypoxia-treated GSTZ1-KO HepG2 cells or GSTZ1-OE Huh7 cells until 90% confluent, followed by scratching using WoundMaker (Essen BioScience) to create wounds. After 24 hours, cells migrating at the front of the wound were photographed by the IncuCyte ZOOM Live-Cell Imaging system (Essen BioScience).

Cell proliferation and Transwell assays. HUVECs and HAOECs were seeded in 96-well plates at 2,000 cells/well, then cultured with CM from hypoxia-treated GSTZ1-KO HepG2 or GSTZ1-OE Huh7 cells. The plates were scanned by the IncuCyte ZOOM Live-Cell Imaging system (Essen BioScience), and phase-contrast images were acquired at 0, 24, 36, 48, and 60 hours. The final number of cells was equal to the number of cells seeded at day 0 divided by the area of cells seeded at day 0 and multiplied by the final area of cells. For the migration assays, 1 × 104 HUVECs and HAOECs were seeded onto the upper chamber of a Transwell insert with serum-free DMEM medium, while the underside chamber of the Transwell insert was filled with CM derived from hypoxia-treated GSTZ1-KO HepG2 cells or GSTZ1-OE Huh7 cells. After 24 hours, the inserts were fixed with 0.5% crystal violet. Images were acquired under a microscope (IX73, Olympus).

PHD2 activity assay. For exogenous PHD2 enzyme activity, His-tagged pET28a PHD2 constructs were expressed in Escherichia coli BL21 (DE3), and their expression was induced with 1.0 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). We purified cell lysates to obtain PHD2 protein and used it for enzyme activity assays. For endogenous PHD2 enzyme activity, we prepared cells in a T75 flask and added Reagent A, which homogenized the cells. A typical assay mixture containing 20 mM MES buffer pH 6.5, 2 mM ATP, 2 mM CoA, 10 mM MgCl2, and 1 mM succinate was incubated time dependently (for 0, 2, 4, 6, and 8 minutes) at 37°C. The assay mixture (100 μL) was then added to 100 μL of acidic colorimetric solution to stop the reaction, and the mixture was incubated at room temperature for 5 minutes. The enzymatic activity was measured by absorbance values at 340 nm according to the manufacturer’s instructions (AmyJet Scientific).

Orthotopic HCC model in mice. BALB/c nude mice (4 to 6 weeks old, male, 20–25 g) were used to construct the syngeneic orthotopic liver cancer models. Briefly, 1 × 106 HepG2 cells with or without GSTZ1 knockout were suspended in a 40 μL PBS/Matrigel (356234, BD Biosciences) mixture (1:1 v/v ratio) and implanted into the liver lobe. Four weeks later, tumor-bearing mice were intraperitoneally injected with 2-ME2 (MCE) at a dose of 50 mg/kg/d (twice a week) for 4 weeks. At 8 weeks postimplantation, mice were sacrificed, and liver tissues were harvested for histological examination.

Endothelial cell tube formation assay. Briefly, a 96-well plate coated with 100 μL of 1 mg/mL Matrigel (BD Biosciences) was incubated at 37°C for 30 minutes to polymerize. Then, 1 × 104 HUVECs and HAOECs were seeded into each well of a precoated 96-well plate and incubated with CM. After 6 hours, capillary-like tubes were photographed (original magnification, 40×) from 4 randomly chosen fields, then analyzed with Image ProPlus 8.0 software.

Aortic ring sprouting assay. Aortas were excised from 8-week-old C57BL/6 mice. Aortic rings were embedded in 150 μL Matrigel (BD Biosciences) in a 24-well culture plate. CM was added to the wells in a final volume of 200 μL culture medium. The aortic rings were incubated at 37°C for 6 days with medium replaced with fresh medium every other day. On day 6, the microvessel sprouting was photographed and scored from 0 (least positive) to 5 (most positive) in a double-blinded manner; 3 independent experiments were carried out with 5 rings per group in each assay. Representative micrographs were shown.

Co-IP. Huh7 and HepG2 cells were plated into 10 cm dishes until 80% confluent. SA (50 μg/mL) was added to the medium for 24 hours. Cells were lysed with cell lysis buffer (Beyotime) containing a protease inhibitor cocktail (Roche). Protein concentrations were measured using a BCA protein assay (Dingguo). Equal protein samples were incubated with anti–HIF-1α (36169, Cell Signaling Technology) or anti-PHD2 (4835, Cell Signaling Technology) or anti-IgG (5946, Cell Signaling Technology) overnight at 4°C, followed by incubation with Protein G Agarose Beads (MilliporeSigma) at 4°C for 4 hours. The beads were lightly washed twice in cell lysis buffer, followed by 1 wash in PBS containing 0.1% Tween 20. The bound antigen was finally eluted and prepared for Western blotting analysis.

Metabolite detection and analysis. For cell samples, GSTZ1-KO HepG2 and parental cells were washed twice with PBS, and metabolites were extracted with 400 μL cold methanol and acetonitrile (1:1, v/v), followed by the addition of the 2-chloro-d-phenylalanine internal standard. The mixture was centrifuged for 20 minutes at 14,000g at 4°C, and then 10 μL of the supernatant was injected into an Agilent 1290 Infinity II LC System coupled to an Agilent 6495c mass spectrometer. The concentrations of metabolites were detected using LC-MS/MS analysis.

For tissue samples, tissue was homogenized with an internal standard (2-chloro-d-phenylalanine) using a tissue homogenizer, followed by metabolite extraction with a mixture of methanol/water (1:1, v/v). The samples were centrifuged for 20 minutes at 4°C and 14,000g to pellet insoluble material, and supernatants were transferred to clean tubes.

PHD2 expression and purification. His-tagged pET28a-PHD2 constructs were expressed in E. coli BL21 (DE3), and expression was induced with 0.2 mM IPTG. Cells were resuspended in a buffer containing 20 mM Tris (pH 8.0), 300 mM NaCl, and 10 mM imidazole, then lysed using sonication for 30 minutes. Cell lysates were purified using a Ni-NTA affinity column (GE Healthcare, now Cytiva). PHD2 was eluted in 20 mL of elution buffer (20 mM Tris-HCl pH 8.0, 300 mM NaCl, 400 mM imidazole).

In vitro hydroxylation assay. Prepared GSTZ1-KO HepG2 cells and control group cells were treated with or without NTBC in a 15 cm cell culture dish, and purified HIF-1α–ODD (530 to 826 residues) were purchased from Abcam (ab48734). The cellular extract was prepared in 1 mL of reaction buffer (20 mM HEPES pH 7.5, 5 mM KCl, and 1.5 mM MgCl2) followed by 2 freeze/thaw cycles in an ethanol/dry ice bath. The lysates were passed 8 times through a 21-gauge needle, followed by 2 passages through a 26-gauge needle before centrifugation (17,000g at 4°C for 30 minutes). The supernatants were aliquoted and stored at −80°C. The hydroxylation assay was performed by incubating 10 mM HIF-1α–ODD with 25 μL of GSTZ1-KO HepG2 cell extract in the reaction buffer for 15 minutes at 37°C. The reaction was stopped by the addition of SDS loading buffer, and the proteins were separated using SDS-PAGE electrophoresis. Hydroxylation was measured using an HIF prolyl hydroxylation–specific antibody (3434T, Cell Signaling Technology). DMOG was used as a negative control in this assay as an inhibitor of PHD2 activity. Measurements of DMOG of HIF-1α hydroxylation were performed in a similar manner, except that the lysate was preincubated with the compounds for 10 minutes at 37°C before the addition of the HIF-1α–ODD protein.

DARTS. DARTS was conducted to identify the potential targets of SA. Briefly, 50 × 106 cells were lysed in M-PER (78501, Thermo Fisher Scientific) with a protease inhibitor cocktail and phosphatase inhibitor cocktail. TNC buffer (50 mM Tris-HCl pH 8.0, 50 mM NaCl, and 10 mM CaCl2) was added to the lysate, and the protein concentration was determined using BCA assay (Beyotime). Protein concentration was adjusted to 4 μg/μL cells, and the lysates were incubated with varying concentrations of either SA or DMSO (vehicle) for 1 hour at room temperature and digested with Pronase (1:2,000 for PHD2, TopScience) for 30 minutes at room temperature. The digestion was stopped by adding a protease inhibitor cocktail, and the samples were immediately placed on ice. Subsequently, Western blotting was used to determine whether PHD2 was a direct target of SA. GAPDH was used as a negative control.

CETSA. CETSA was performed to determine the direct binding between SA and PHD2 in cells. Briefly, 4 × 106 HepG2 and Huh7 cells were pretreated with 300 μM SA for 12 hours before being subjected to the CETSA protocol. Cells were chilled on ice, washed with PBS buffer plus protease inhibitor cocktail, and then subjected to 3 freeze/thaw cycles using dry ice and Thermal Cycler to lyse cells. Then, the cells were centrifuged at 20,000g for 20 minutes at 4°C to separate between the lysate and cell debris with aggregated proteins. Next, the supernatant was transferred into 200 μL PCR tubes; all samples were heat shocked in a Bio-Rad T100 Thermal Cycler at the indicated temperature for 3 minutes to denature proteins, then immediately cooled down at room temperature for 3 minutes. Finally, the supernatant was boiled with 4× loading buffer for Western blotting. The bands were quantified using the Gel-Pro analyzer software (http://www.gelanalyzer.com/?i=1) and plotted with 3 biological replicates.

SPR. Affinity analysis was performed using a Biacore X100 instrument (GE Healthcare Life Sciences, now Cytiva). His-PHD2 protein (Sino Biological, 11084-H08H) was directly immobilized on the sensor chip NTA. Then, 20 μg/mL of PHD2 in immobilization buffer was injected into the Fc2 sample channel at a flow rate of 10 μL/min. The metabolite SA or α-KG was serially diluted with the running buffer to obtain concentrations of 400, 200, 100, 50, 25, and 12.5 nM, respectively. Different concentrations of metabolite were then injected into the Fc2-Fc1 channels at a flow rate of 30 μL/min, with a contact time of 120 seconds, followed by a dissociation time of 400 seconds. After each cycle of interaction analysis and analyte injection, the association and dissociation processes were all handled in the running buffer. Data analysis was performed on the Biacore X100 computer, using the Biacore X100 evaluation software.

Flow cytometry. For flow cytometry analysis of in vitro mouse immune cells, the mouse T cells were incubated with 1× Golgiplug (BD Biosciences) for 6 hours. Immune cells were stained with surface antibodies in PBS for 30 minutes as described previously (51). Fixation and permeabilization processes were carried out with fixation buffer (BD Biosciences) according to the manufacturer’s protocols. The immune cells were then stained with antibodies in PBS to detect proteins. The following antibodies were used for staining: Brilliant Violet 510 anti-mouse/human CD11b (BioLegend, catalog 101245), FITC anti-mouse F4/80 antibody (BioLegend, catalog 123107), PerCP/Cyanine5.5 anti-mouse CD4 antibody (BioLegend, catalog 100540), APC anti-mouse CD206 antibody (BioLegend, catalog 141708), Brilliant Violet 421 anti-mouse CD86 antibody (BioLegend, catalog 105032), and PE anti-mouse FoxP3 antibody (Elabscience, catalog E-AB-F1238D). The flow cytometry was run using FACSCelesta flow cytometer (BD Biosciences), and the results were analyzed with FlowJo V10.7.1.

Gstz1–/– mouse study. Heterozygous C57-Gstz1tm1Jmfc/Cnbc mice (EM: 04481) were obtained from the European Mouse Mutant Archive and were crossed to breed WT and Gstz1–/– mice. For the DEN and CCl4 -induced mouse HCC model, WT and Gstz1–/– mice (at 2 weeks of age) were administered an intraperitoneal injection of DEN (MilliporeSigma) at a dose of 75 mg/kg. At 3 weeks of age, the mice were administrated with 10% CCl4 (Macklin) intraperitoneally at a dose of 2 mL/kg twice a week for 12 weeks and then received an intraperitoneal injection of DEN at a dose of 50 mg/kg. At 20 weeks of age, Gstz1–/– mice were intraperitoneally administered 2-ME2 (50 mg/kg, once per week, S1233, Selleck), anti–PD-L1 (100 μg/kg, once per week), both, or vehicle (5% DMSO) (n = 6 per group) for 4 weeks. Mice in the NTBC treatment group received continuous NTBC treatment (8 mg/L, S5325, Selleck) through their drinking water until sacrifice (52). Mice were sacrificed at 24 weeks of age, and liver tissues were harvested for histological examination.

Statistics. Graphical representation and statistical analyses included 2-tailed unpaired Student’s t test, 1-way ANOVA with Tukey’s test, Pearson’s correlation coefficient, Gehan-Breslow-Wilcoxon test, and 2-way ANOVA with Bonferroni’s test, calculated using GraphPad Prism 8. P values less than 0.05 were deemed significant. The experiments were not randomized, except that the mice were randomly grouped before treatments. Samples were allocated to their experimental groups according to their predetermined type, and allocation was not blinded during the experiments and outcome assessment. Statistical information is otherwise provided in the figure legends.

Study approval. Primary HCC tissue samples and paired adjacent normal tissue samples were obtained from The Second Affiliated Hospital of Chongqing Medical University between 2018 and 2022, with approval from the Institutional Review Board of Chongqing Medical University. Written informed consent in accordance with a protocol approved by The Second Affiliated Hospital of Chongqing Medical University (Chongqing, China) was obtained from all patients. All animal experiments were performed under the guidelines of the institutional Animal Care and Use Committee at Chongqing Medical University. All animal procedures were also approved by the Research Ethics Committee of Chongqing Medical University.

Data availability. All data needed to evaluate the conclusions in the paper are present in the paper or the supplemental materials. Expression profile data analyzed in this study were deposited in Gene Expression Omnibus at GSE192760. See complete unedited blots in the supplemental material. Values for all data points are available in the Supporting Data Values file.