Lenvatinib triggers an EGR1-ZNF768-SLC7A11 adaptive response to limit ferroptosis-mediated therapeutic efficacy in hepatocellular carcinoma

Human tissue specimens and patient cohorts

A cohort of 113 paired HCC and adjacent non-tumorous tissues was obtained from patients who underwent curative resection at Tianjin Medical University Cancer Institute and Hospital (TMUCIH). Lenvatinib combined with anti-PD-1 therapy has largely replaced lenvatinib monotherapy as the standard first-line regimen for advanced HCC in our institution. A dedicated lenvatinib monotherapy cohort of sufficient size was therefore not available for this analysis. The lenvatinib-based combination therapy cohort was used as a proxy, given that the resistance mechanism identified in this study is driven specifically by lenvatinib-induced AKT inhibition and operates independently of the anti-PD-1 component. Formalin-fixed paraffin-embedded (FFPE) tumor tissues were collected from HCC patients who received lenvatinib-based therapy at TMUCIH for the lenvatinib treatment cohort. Patient outcomes were evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 with patients classified as non-progressive disease (NPD), which included complete response, partial response, or stable disease, or progressive disease (PD). All human sample collection and use were approved by the Institutional Review Board of TMUCIH (Approval No. Ek2023059) and written informed consent was obtained from all patients.

Bioinformatic analysis of public datasets

A multi-step bioinformatic analysis was performed to identify dysregulated ZNF genes. First, differential expression analysis of bulk RNA-seq data from three independent Gene Expression Omnibus (GEO) datasets (GSE135631, GSE94660, and GSE56545) was performed to compare tumor and adjacent non-tumor tissues¹⁷. Genes with a log2 fold change (logFC > 0.5) and a false discovery rate (FDR) < 0.01 were considered significantly upregulated. Second, prognostic ZNF genes were identified from The Cancer Genome Atlas-Liver Hepatocellular Carcinoma (TCGA-LIHC) cohort using survival analysis with a hazard ratio (HR) > 1.5 and an FDR < 0.01 as the significance threshold. The JASPAR database¹⁸ was used to predict transcription factor binding motifs within promoter regions of interest.

Single-cell RNA-sequencing (scRNA-seq) analysis

ScRNA-seq datasets from human and murine HCC (GSE149614, GSE181515, and GSE237823) were acquired from the GEO. Raw data were processed using standard pipelines for quality control, normalization, and clustering^19,20. Malignant hepatocytes were identified based on established markers. Differentially expressed genes (DEGs) between ZNF768-positive and -negative hepatocytes were identified. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed on these DEGs to uncover associated biological pathways.

Cell culture and reagents

The human HCC cell line, Hep 3B2.1-7 (Hep3B), the human embryonic kidney cell line, 293T, and the murine HCC cell line, Hepa1-6, were obtained from the American Type Culture Collection [ATCC] (Manassas, VA, USA). The human HCC cell lines, HCCLM3, SNU387, Li-7 and Huh7, were purchased from Servicebio (Wuhan, China). Hep3B, Huh7, HCCLM3, 293T, and Hepa1-6 cells were cultured in high-glucose Dulbecco’s Modified Eagle’s Medium [DMEM] (C3113-0500; Vivacell, Shanghai, China). Li-7 and SNU387 cells were cultured in RPMI-1640 medium (C3010-0500; Vivacell). All media were supplemented with 10% fetal bovine serum [FBS] (C04001-500; Vivacell) and 1% penicillin/streptomycin solution (A3010-100 mL; Metacell, Tianjin, China). All cell lines were maintained at 37°C in a humidified atmosphere with 5% CO₂, authenticated by short tandem repeat (STR) profiling, and routinely tested for mycoplasma contamination. The reagents used in this study are listed in Table S1.

Plasmid constructs and stable cell line generation

shRNA sequences targeting human ZNF768 and murine Zfp768 were cloned into the pLsiCTRL-Hygro vector for gene knockdown. Human cDNAs for ZNF768 and SLC7A11 were cloned into lentiviral expression vectors (pLV3-CMV-3×FLAG-Puro and pCDH-CMV-EF1a-Puro, respectively) for overexpression. The lentiviral expression vector, pLV3-CMV-EGR1 (human)-Myc-mCherry-Puro (P50355), was purchased from Miaoling Plasmid (Wuhan, China). Lentivirus was produced in 293T cells by co-transfecting the lentiviral expression vector with packaging plasmids (pCMV-dR8.2 and pCMV-VSV-G). Target cells were infected with viral supernatant (MOI 5–10) in the presence of 10 μg/mL of polybrene (H8761; Solarbio, Beijing, China). Stable cell lines were selected and maintained in medium containing puromycin (2 μg/mL, G4017; Servicebio) or hygromycin B (200 μg/mL, 60224ES03; Yeasen, Shanghai, China). Sequences for RNA interference used in this study are listed in Table S2.

Measurement of ferroptosis markers

Intracellular ROS were measured using the dichlorodihydrofluorescein diacetate (DCFH-DA) probe (S0033S; Beyotime, Shanghai, China). Lipid peroxidation was assessed qualitatively and quantitatively using the C11-BODIPY 581/591 probe (S0043S; Beyotime) via fluorescence microscopy and flow cytometry. Malondialdehyde (MDA) levels were quantified using a colorimetric assay kit (ab118970; Abcam, Cambridge, UK). The ratio of GSH-to-oxidized glutathione disulfide (GSSG) was determined using a specific kit (S0053; Beyotime). Intracellular ferrous iron (Fe²⁺) levels were measured using the Cell Ferrous Iron Colorimetric Assay kit (E-BC-K881-M; Elabscience, Wuhan, China). The NADPH:NADP⁺ ratio was determined using an Enhanced NADP⁺/NADPH Assay kit with WST-8 (S0180S; Beyotime). All assays were performed according to the manufacturers’ instructions with final values normalized to total protein concentration, as appropriate.

RNA extraction and quantitative real-time PCR (RT-qPCR)

Total RNA was extracted from cells using Freezol^® Reagent (R711; Vazyme, Nanjing, China). Reverse transcription was performed with the HiScript IV Ultra RT SuperMix (R433; Vazyme). RNA extraction and RT-qPCR was performed on a QuantStudio 5 system (Thermo Fisher Scientific, Waltham, MA, USA) using Taq Pro Universal SYBR Master Mix (Q712; Vazyme). Gene expression was normalized to ACTB as an internal control and relative quantification was calculated using the 2^−ΔΔCt method. The primers used in this study are listed in Table S3.

Western blotting

Cells were lysed in boiling buffer supplemented with 1 mM PMSF (P0100; Solarbio) and a protease inhibitor cocktail (IKM1010; Solarbio). Protein concentration was determined using a nanodrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Equal amounts of protein (20–30 μg) were separated by SDS-PAGE, transferred to 0.22-μm PVDF membranes (E802; Vazyme), and blocked with 5% non-fat milk. The membranes were incubated with primary antibodies overnight at 4°C, followed by incubation with HRP-conjugated secondary antibodies. Immunoreactive bands were visualized using an enhanced chemiluminescence (ECL) system (Boluteng Biotechnology, Guangzhou, China). The antibodies used in this study are listed in Table S4.

Cell proliferation and colony formation assays

Cells were seeded in 96-well plates (1,000–2,000 cells/well) for proliferation assays. Cell viability was measured at 0, 24, 48, 72, and 96 h using Cell Counting Kit-8 [CCK-8] (C0005; TargetMol, Shanghai, China) according to the manufacturer’s protocol with absorbance read at 450 nm. Cells were seeded at low density (500–1,000 cells/well) in 12-well plates and cultured for 10–14 d for colony formation. Colonies were fixed with 4% paraformaldehyde (PFA), stained with 0.1% crystal violet, and counted. Cells (2,000 cells/well) were treated with erastin (5 μM) or vehicle with medium refreshed every 3 d for drug-challenged colony formation assays.

5-ethynyl-2′-deoxyuridine (EdU) incorporation assay

DNA synthesis was assessed using a EdU Incorporation Assay kit (G1603; Servicebio). Cells were incubated with 50 μM EdU for 2 h, fixed with 4% PFA, and permeabilized with 0.3% Triton X-100 (T8204, Solarbio). Incorporated EdU was labeled via a Click-iT reaction with an Alexa Fluor 488-azide conjugate (contained within the EdU Incorporation Assay kit). Nuclei were counterstained with Hoechst 33342 (contained within the EdU Incorporation Assay kit). The percentage of EdU-positive cells was quantified from images captured on a fluorescence microscope (Mshot, Guangzhou, China).

Drug cytotoxicity assay

Cells were seeded in 96-well plates (5,000–8,000 cells/well) to determine drug sensitivity. After 24 h, cells were treated with serial dilutions of lenvatinib, erastin, or RSL3 for 48 h. Cell viability was assessed using the CCK-8 kit. The IC₅₀ was calculated by fitting the dose-response curves using non-linear regression in GraphPad Prism (GraphPad Software, San Diego, CA, USA).

Transmission electron microscopy

Cells were harvested and fixed with 2.5% glutaraldehyde (P1126; Solarbio) at 4°C overnight. After washing, samples were post-fixed with 1% osmium tetroxide, dehydrated through a graded ethanol series, and embedded in epoxy resin. Ultrathin sections (70 nm) were cut, stained with uranyl acetate and lead citrate, and examined under a transmission electron microscope [TEM] (HT7700; Hitachi, Tokyo, Japan) to observe mitochondrial ultrastructure.

Luciferase reporter assay

Human SLC7A11 or ZNF768 promoter regions (−2,000 to +100 bp relative to the transcription start site) were cloned into the pmir GLO Dual-Luciferase vector (E1330; Promega, Madison, WI, USA). Truncated and site-directed mutant constructs were generated via overlap extension PCR. Cells were co-transfected with the reporter plasmid, a Renilla luciferase control vector, and an expression plasmid for the relevant transcription factor (ZNF768 or EGR1) or corresponding control vector. Firefly and Renilla luciferase activities were measured using the Dual-Lumi™ II kit (RG089S; Beyotime) after 48 h and the ratio of Firefly-to-Renilla activity was calculated.

Chromatin immunoprecipitation (ChIP)-PCR

ChIP assays were performed using the BeyoChIP™ kit (P2083S; Beyotime). Briefly, cells were cross-linked with 1% formaldehyde and chromatin was sheared to an average size of 200–500 bp by enzymatic digestion and sonication. The chromatin was then immunoprecipitated with antibodies against ZNF768, EGR1, or control IgG. Co-precipitated DNA was purified and enrichment of specific promoter regions was quantified by PCR.

Flow cytometry

After peroxidation induction (10 μM erastin for 24 h), cells (2 × 10⁶/mL) were stained with 5 μM C11-BODIPY 581/591 (S0043S; Beyotime) in 1 × phosphate buffered saline [PBS] (C1511-500 mL; Metacell, Tianjin, China) at 37°C for 30 min in the dark. Then, cells were washed with PBS/0.1% bovine serum albumin [BSA] (A8020; Solarbio) and analyzed immediately on a flow cytometer (Beckman Coulter, Brea, CA, USA). Fluorescence was detected in FITC (530/30 nm, oxidized) and PE-Texas Red (615/20 nm, reduced) channels.

RNA-sequencing and analysis

Total RNA was extracted from ZNF768-knockdown and control Hep3B cells, and libraries were prepared by Novogene Biotech Company (Beijing, China). Sequencing was performed on an Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA). Raw reads were filtered for quality using fastp. Clean reads were aligned to the human reference genome, GRCh38, using HISAT2. Differential gene expression analysis was performed using the DESeq2 R package. Genes with an adjusted FDR P-value ≤ 0.05 and an absolute logFC ≥ 1 were considered differentially expressed. KEGG pathway analysis was performed on the resulting gene list.

Non-targeted metabolomics

Metabolites were extracted from ZNF768-knockdown and control Hep3B cell pellets using a pre-chilled solvent mixture of methanol, acetonitrile, and water [2:2:1 (v/v/v)]. The supernatant was collected for analysis after sonication and incubation. UHPLC-MS/MS analysis was performed on a UPLC-ESI-Q-Orbitrap-MS system. Metabolite identification was achieved by matching accurate mass and MS/MS fragmentation patterns against public and in-house databases. Differential metabolites were identified using a combination of VIP scores >1.0 from the OPLS-DA model and a P-value < 0.05 from a Student’s t-test. Pathway analysis was performed using the KEGG database.

Immunohistochemistry (IHC)

Paraffin-embedded tissue sections were deparaffinized, rehydrated, and subjected to heat-induced antigen retrieval. Endogenous peroxidase activity was blocked with 3% H₂O₂. Sections were incubated with primary antibodies overnight at 4°C after blocking with 3% BSA. A universal HRP-polymer detection system (PV-6000; ZSGB-BIO, Beijing, China) and DAB chromogen (ZLI-9018; ZSGB-BIO) were used for visualization, followed by hematoxylin counterstaining. Staining was scored independently by two pathologists based on intensity [0 (negative), 1 (weak), 2 (moderate), and 3 (strong)] and the percentage of positive cells: [0 (<5%), 1 (5%–25%), 2 (25%–50%), 3 (50%–75%), 4 (>75%)]. The final IHC score (0–12) was calculated by multiplying the intensity and percentage scores. Score of 0–5 were considered low expression and scores of 6–12 considered high expression for statistical analyses.

Animal studies

The sample size (n = 6 per group) was determined based on our prior experience with subcutaneous tumor models using the same cell lines. A group size of 6 was estimated to provide >80% power to detect a 40% difference in mean tumor weight at a significance level of 0.05, assuming a coefficient of variation of 20%–25%, which is typical for this model system. For the initial tumorigenesis study, 5 × 10⁶ Hepa1-6 cells suspended in 100 μL of PBS (stably expressing shZNF768 or ZNF768-overexpression constructs and controls) were injected subcutaneously into the flanks of 5-week-old male C57BL/6 mice (n = 6/group; GemPharmatech, Nanjing, China). For the in vivo rescue experiment, 1 × 10⁷ Hep3B cells suspended in 100 μL of PBS with 50% Matrigel (ZNF768-KD or ZNF768-KD with SLC7A11 rescue, G4130; Servicebio) were injected into BALB/c nude mice (n = 6/group; GemPharmatech, Nanjing, China). Tumors were established in BALB/c nude mice using Hepa1-6 cells (5 × 10⁶ cells per tumor) for the therapeutic efficacy study. Mice were randomized into four groups (n = 6/group) to receive the following when tumors reached 50–100 mm³: (1) vehicle (0.5% sodium carboxymethyl cellulose solution, IR9037; Solarbio) + adeno-associated virus serotype 8 (AAV8)-shScramble; (2) lenvatinib (10 mg/kg/day) + AAV8-Scramble; (3) vehicle + AAV8-shZNF768; or (4) lenvatinib + AAV8-shZNF768. AAV vectors (2 × 10¹¹ vg/injection) were administered via intratumoral injection immediately after grouping. The virus tools were packaged by BrainVTA Co., Ltd. (Wuhan, China). In brief, for ZNF768 knockdown, AAV-pan with a TBG promoter driving shZNF768 (5′-GGGTATGAACCGCAGAACTCT-3′) was constructed, a sequence targeting murine Zfp768 that is distinct from the shRNA sequences used for human ZNF768 knockdown in the in vitro experiments, ensuring species specificity throughout the in vivo study. AAV-pan with a TBG promoter driving scramble (5′-CCTAAGGTTAAGTCGCCCTCG-3′) served as a control. At the same time, lenvatinib and an equal volume of the vehicle was administered daily by oral gavage until the end of the experiment. Tumor volume [V = (Length × Width²)/2] and body weight were monitored every 2–3 d. Tumors were excised, weighed, and processed for IHC and MDA analysis at the experimental endpoint. Mouse blood, obtained via eyeball extraction at the end of the experiment, was left at room temperature for 2 h before centrifugation at 2,000 × g for 15 min at 4°C. The supernatant was then collected. Serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), and creatinine (CRE) levels were measured using a Roche Cobas E701 analyzer (Roche Diagnostics, Basel, Switzerland) according to the manufacturer’s instructions. Murine blood assays were performed in the Department of Laboratory Medicine at Tianjin Medical University Cancer Institute and Hospital (Tianjin, China). All animal experiments were approved by the Tianjin Medical University Institutional Animal Care and Use Committee and performed in accordance with institutional guidelines (Approval No. AE2023020).

Statistical analysis

Student’s t-test or the Mann–Whitney U test was used as appropriate for two-group comparisons. Multiple group comparisons were performed using one-way ANOVA followed by Tukey’s post-hoc test and tumor growth curves were analyzed using two-way ANOVA. Data are presented as the mean ± standard deviation (SD) or mean ± standard error of the mean (SEM), as indicated in the figure legends. Western blot intensities were quantified via densitometry using ImageJ and normalized to loading controls. Drug IC₅₀ values were calculated using four-parameter non-linear regression. Clinical analyses included Kaplan–Meier survival estimation with log-rank testing, Spearman’s rank correlation for IHC scoring relationships, chi-square tests for treatment response associations, and receiver operating characteristic (ROC) curve analysis with the Youden index for cut-off determination. Multivariate logistic regression, incorporating ZNF768 expression, tumor stage, and baseline AFP, was used to evaluate independent predictive values. All statistical analyses were performed using GraphPad Prism 9.0 and R 4.2.0. A two-tailed P < 0.05 was considered statistically significant.

Integrative multi-omics analyses identify ZNF768 as a transcriptional driver linking aggressive HCC to poor survival

Advanced-stage HCC is characterized by aggressive phenotypes and poor response to systemic therapies, like lenvatinib, suggesting a common molecular basis for malignant progression and therapeutic tolerance. To identify transcriptionally dysregulated ZNF family members underlying these processes, a tripartite bioinformatics strategy integrating bulk transcriptomics, scRNA-seq, and clinical survival analysis was implemented (Figure 1A). Differential expression analysis across 3 GEO bulk RNA-seq datasets (GSE135631, GSE94660, and GSE56545) identified 308 consistently upregulated ZNF genes in tumors versus adjacent tissues (logFC > 0.5; FDR < 0.01) and designated as bulk DEG-ZNFs. scRNA-seq profiling of advanced HCC specimens subsequently revealed 23 ZNF transcripts significantly enriched in malignant hepatocytes from advanced-stage tumors compared to early-stage counterparts (scRNA advanced ZNFs). Survival analysis of TCGA-LIHC cohort further identified 58 prognostic ZNF genes [HR > 1.5, FDR < 0.01 (prognosis-ZNFs)] in which elevated expression correlated with reduced overall survival. The intersection of these sets yielded two candidates (ZNF768 and ZNF207; Figure 1B). ZNF768 was prioritized for functional characterization due to complete absence from the published HCC literature, contrasting with the ZNF207’s established roles in other malignancies^21,22.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

ZNF768 is overexpressed in HCC and correlates with aggressive phenotypes and poor prognosis. (A) Schematic workflow of the integrated bioinformatics strategy used to identify prognostically significant ZNF family members in HCC. (B) Venn diagram showing the intersection of upregulated ZNFs from three bulk RNA-seq datasets (bulk DEG-ZNFs), progression-associated ZNFs from scRNA-seq (scRNA advanced ZNFs), and prognostically significant ZNFs from TCGA survival analysis (prognosis-ZNFs), identifying ZNF768 and ZNF207 as candidates. (C–E) ZNF768 mRNA expression is significantly elevated in tumor versus adjacent non-tumor tissues in three independent GEO datasets: GSE135631; GSE56545; and GSE94660. (F) UMAP plot from scRNA-seq analysis showing ZNF768 enrichment in malignant hepatocytes. Violin plots show the distribution; bars show the mean values. (G, H) Representative IHC images and quantification of ZNF768 protein expression in a cohort of 113 paired HCC and adjacent tissues. (I) Kaplan–Meier analysis of overall survival in the 113-patient HCC cohort, stratified by high versus low ZNF768 expression (IHC score). (J, K) Representative IHC images of serial sections and Spearman correlation analysis demonstrating a positive correlation between ZNF768 and Ki67 expression. Data in (H) were analyzed using the Mann–Whitney U test. Survival curve P-value was determined using the log-rank test. The scale bars in (G, J) represent 100 μm and 200 μm respectively. **P < 0.01, ***P < 0.001, ****P < 0.0001. FDR, false discovery rate; GEO, Gene Expression Omnibus; HCC, hepatocellular carcinoma; IHC, immunohistochemistry; scRNA-seq, single-cell RNA-sequencing; TCGA, The Cancer Genome Atlas.

Validation confirmed ZNF768 dysregulation across all datasets. Bulk RNA-seq showed significant tumor overexpression (Figure 1C–E), while scRNA-seq revealed enrichment in malignant hepatocytes from advanced/progressive disease states (Figure 1F). Western blot verified elevated protein levels (Figure S1A). IHC analysis of 113 HCC specimens demonstrated higher tumor staining intensity (median IHC score: 6.0 vs. 3.0 in adjacent tissues, P < 0.0001; Figure 1G, H). High ZNF768 expression (IHC score ≥ 2+) predicted worse overall survival (median OS: 27.0 vs. 61.0 months, HR = 2.22, P = 0.0004; Figure 1I), a pattern replicated in TCGA analyses across multiple endpoints (Figure S1B–E). Serial section IHC revealed a strong positive correlation between ZNF768 and the proliferation marker, Ki67 (Figure 1J, K), suggesting ZNF768 promotes tumor cell proliferation. Together, these multi-omics and clinicopathologic analyses established ZNF768 as a previously unrecognized biomarker linking overexpression to advanced HCC pathology and poor outcomes.

ZNF768 drives malignant proliferation and tumor growth in vitro and in vivo

Building on the clinical association between ZNF768 overexpression and aggressive HCC phenotypes, the functional role was investigated in malignant progression. Western blot analysis revealed heterogeneous ZNF768 expression across 7 HCC cell lines. Hep3B and Huh7 had the highest levels, while HCCLM3 and Li-7 exhibited minimal expression (Figure S2A). ZNF768 was stably knocked down in Hep3B/Huh7 cells and overexpressed in HCCLM3/Li-7 cells using 2 independent shRNAs (Figures 2A, B and S2B, C). ZNF768 knockdown significantly reduced proliferation in high-expressing cells, as evidenced by decreased viability in CCK-8 assays (Figure 2C). Conversely, overexpression enhanced proliferation in low-expressing cells (Figure 2D). Colony formation assays further confirmed these results. Specifically, knockdown suppressed colony growth, while overexpression increased colony growth (Figure 2E, F). EdU incorporation assays demonstrated that ZNF768 silencing impaired DNA synthesis (Figures 2G, H and S2D, E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

ZNF768 drives malignant proliferation and tumor growth in vitro and in vivo. (A, B) Western blot analysis confirming stable ZNF768 KD in Hep3B and Huh7 cells and OE in HCCLM3 and Li-7 cells. (C, D) CCK-8 proliferation assays showing that ZNF768 KD inhibited, while OE enhanced, the viability of HCC cells over 96 h. (E, F) Colony formation assays demonstrating that ZNF768 KD suppressed, while OE increased, the clonogenic capacity of HCC cells. (G, H) EdU incorporation assays showing impaired DNA synthesis in ZNF768 KD cells and enhanced synthesis in ZNF768 OE cells (scale bar = 50 μm). (I–K) ZNF768 KD in murine Hepa1-6 cells significantly impaired tumor growth in a syngeneic subcutaneous tumor model, as shown by tumor growth curves, final tumor volumes, and tumor weights (n = 6 per group). (L–N) ZNF768 OE in Hepa1-6 cells accelerated tumor growth in the same model (n = 6 per group). Data are presented as the mean ± SD from three independent experiments or mean ± SEM for in vivo data. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by Student’s t-test or ANOVA. EdU, 5-ethynyl-2′-deoxyuridine; HCC, hepatocellular carcinoma; KD, knockdown; OE, overexpression; SD, standard deviation; SEM, standard error of the mean.

To extend these findings in vivo, stable ZNF768-knockdown and -overexpressing Hepa1-6 cells were generated for syngeneic transplantation (Figure S2F, G). Subcutaneous tumors showed that knockdown significantly slowed tumor growth, reduced volumes, and decreased weights (Figure 2I–K). Conversely, overexpression accelerated tumorigenesis, yielding larger, heavier tumors (Figure 2L–N). IHC analysis linked this growth to proliferation. ZNF768 knockdown reduced Ki67 staining intensity, while overexpression increased Ki67 staining intensity (Figure S2H, I). In summary, these results established ZNF768 as a functional oncogene that drives HCC malignancy in vitro and in vivo.

Transcriptional and metabolic profiling identifies ZNF768 as a regulator of ferroptosis

How cancer cells overcome oxidative stress during rapid proliferation was investigated to elucidate mechanisms underlying ZNF768-driven aggressive tumor growth. Transcriptomic correlation analysis was performed across three HCC datasets using integrated multi-omics approaches (scRNA-seq, bulk transcriptomics, and metabolomics): human GSE149614; murine DEN/CCl4-induced GSE181515; and genetically engineered GSE237823 (Figure 3A). Hepatocyte-focused analysis revealed DEGs between ZNF768⁺ and ZNF768⁻ cells. KEGG enrichment of these DEGs consistently showed significant pathway enrichment for cell proliferation and ferroptosis modulation (Figure 3B–D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Transcriptional and metabolic profiling identifies ZNF768 as a regulator of ferroptosis. (A) Schematic of the workflow for identifying ZNF768-associated pathways by integrating three independent scRNA-seq datasets from human and murine HCC models. (B–D) KEGG enrichment analysis of DEGs between ZNF768-positive and ZNF768-negative hepatocytes across the three datasets, consistently identifying enrichment in pathways related to ferroptosis, proliferation, and motility. (E) Heatmap of ferroptosis-related genes from RNA-sequencing of Hep3B cells with stable ZNF768 KD (shZNF768) versus control (Scramble), showing downregulation of ferroptosis suppressors and upregulation of drivers. (F) KEGG pathway analysis of differentially expressed genes from the RNA-seq experiment with ferroptosis ranked as a top enriched pathway. (G) Metabolomic pathway impact analysis in ZNF768-KD Hep3B cells, identifying glutathione metabolism as the most significantly altered pathway. DEGs, differentially expressed genes; HCC, hepatocellular carcinoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; KD, knockdown; scRNA-seq, single-cell RNA-sequencing.

Functional validation via RNA-seq in ZNF768-knockdown Hep3B cells showed upregulated ferroptosis drivers (e.g., ACSL4 and NCOA4)^23,24 along with downregulated suppressors (e.g., SLC7A11 and FTH1; Figure 3E)^25,26. KEGG analysis confirmed ferroptosis as the second most enriched pathway (rich factor = 0.24, FDR = 0.011; Figure 3F), which aligned with single-cell data. Given the metabolic basis of ferroptosis, non-targeted metabolomics in knockdown cells revealed pronounced ferroptosis-pathway enrichment (Figure S3A). Pathway impact analysis identified GSH metabolism as the most altered pathway, exhibiting the highest fold change and significance (FDR < 0.001; Figure 3G). Together, these multi-omics layers established ZNF768 as a transcriptional master regulator of ferroptosis homeostasis in HCC.

ZNF768 maintains GSH redox homeostasis to confer protection against ferroptosis

The role of ZNF768 in GSH-dependent ferroptosis suppression was validated following multi-omics analyses identifying ZNF768 as a transcriptional regulator of ferroptosis. The reduced GSH:oxidized GSSG ratio was measured as a key redox indicator to assess the impact of ZNF768 on GSH metabolism. ZNF768 knockdown significantly reduced the GSH:GSSG ratio in Hep3B and Huh7 cells (Figure 4A), indicating an oxidative shift. In addition, ZNF768 depletion decreased the NADPH:NADP⁺ ratio (Figure 4B), which is essential for sustaining reduced GSH pools.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

ZNF768 maintains glutathione homeostasis and confers protection against ferroptosis. (A, B) Relative ratios of GSH-to-GSSG and NADPH-to-NADP⁺ were measured in ZNF768 KD and control Hep3B and Huh7 cells. (C) Intracellular ROS levels were assessed by DCFH-DA staining in indicated cells (scale bar = 50 μm). (D) Representative fluorescence microscopy images show lipid peroxidation levels in ZNF768 KD and control cells (scale bar = 20 μm). (E) Lipid peroxidation was measured by flow cytometry using the C11-BODIPY 581/591 probe in indicated cells. (F, G) Quantification of cellular MDA and intracellular Fe²⁺ levels in ZNF768 KD and control cells. (H) Cell viability of ZNF768 KD and control cells following 48-h treatment with the indicated doses of the ferroptosis inducer, erastin. (I) Rescue experiment in ZNF768-knockdown cells treated with erastin (10 μM) in the presence or absence of 2 μM Fer-1, 100 nM Lip-1, 20 μM Z-VAD, or 10 μM Nec-1s. Cell viability was assessed by CCK-8. (J) Representative TEM images showing mitochondrial ultrastructure in ZNF768 KD and control cells treated with 10 μM erastin, highlighting mitochondrial shrinkage and cristae dissolution (arrows) upon ZNF768 loss (scale bar = 1 or 2 μm). Data are presented as the mean ± SD from three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, and ns by Student’s t-test or ANOVA. Fer-1, ferrostatin-1; KD, knockdown; Lip-1, liproxstatin-1; MDA, malondialdehyde; Nec-1s, necrostatin-1s; ns, non-significant; SD, standard deviation; TEM, transmission electron microscope; Z-VAD, Z-VAD-FMK.

Whether ZNF768 loss induces ferroptosis hallmarks was examined next. Intracellular ROS levels substantially increased in ZNF768-knockdown cells using the 2′,7′-DCFH-DA probe (Figures 4C and S4A). Lipid peroxidation, a ferroptosis indicator, was assessed via C11-BODIPY staining. Microscopy revealed enhanced green fluorescence in ZNF768-deficient cells (Figures 4D and S4B), while flow cytometry quantitatively confirmed increased lipid peroxidation through a right-shifted fluorescence histogram (Figure 4E). Elevated MDA levels (Figure 4F) and concurrent accumulation of Fe²⁺ (Figure 4G) further corroborated these findings. Together, ZNF768 deficiency induces a pro-ferroptotic state characterized by ROS accumulation, iron overload, and lipid peroxidation.

Cells were treated with inducers (erastin and RSL3) to evaluate ferroptosis susceptibility. ZNF768-knockdown cells exhibited dose-dependent viability reduction upon treatment with erastin (Figure 4H) and RSL3 (Figure S4C). Consistently, the long-term proliferative capacity was also severely compromised under erastin challenge (Figure S4D). Rescue experiments were performed using distinct cell death inhibitors to confirm that the observed cell death was specifically driven by ferroptosis. The erastin-induced cell death in ZNF768-deficient cells was significantly reversed by the ferroptosis inhibitors, Fer-1 and Lip-1, whereas the apoptosis inhibitor, Z-VAD, and the necroptosis inhibitor, Nec-1s, failed to confer protection (Figure 4I). Moreover, the iron chelator, deferoxamine (DFO), rescued cell viability to a degree comparable to Fer-1 and Lip-1, confirming that this cell death is iron-dependent (Figure S4E). Having confirmed the biochemical nature of ferroptosis, the ultrastructural features were further examined. TEM of erastin-treated cells revealed pronounced mitochondrial shrinkage and cristae dissolution (Figure 4J), characteristic of ferroptosis. Thus, ZNF768 sustains GSH redox balance and reducing power, preventing iron-dependent lipid peroxidation and ferroptotic death.

ZNF768 directly binds and transactivates the SLC7A11 promoter

To investigate the mechanism of heightened ferroptosis sensitivity after ZNF768 knockdown, we performed RNA-seq analysis and identified SLC7A11 as one of the most significantly downregulated transcripts (Figure 5A). This correlation was further validated in public HCC datasets (Figure S5A, B). This regulatory axis was robustly confirmed. ZNF768 knockdown reduced, whereas overexpression elevated SLC7A11 mRNA and SLC7A11 protein levels across multiple cell lines (Figures 5B–D and S5C–G). To determine whether the reduction in SLC7A11 upon ZNF768 knockdown was mediated through NRF2, total NRF2 protein levels and nuclear translocation in ZNF768-depleted Hep3B and Huh7 cells were assessed. Western blot analysis showed no significant change in total NRF2 protein following ZNF768 knockdown (Figure S5H). Nuclear-cytoplasmic fractionation confirmed that NRF2 nuclear accumulation was similarly unaffected (Figure S5I). These results indicated that ZNF768 protein regulates SLC7A11 transcription through a direct and NRF2-independent mechanism. Moreover, these results suggested transcriptional regulation, which was verified by luciferase reporter assays demonstrating direct modulation of SLC7A11 promoter activity by ZNF768 (Figure S5J, K). To map the regulatory element, three potential ZNF768 binding sites (BS1, BS2, and BS3) were predicted within the promoter using JASPAR (Figure 5E, F)¹⁸. Functional analysis revealed that simultaneous mutation of all three sites abolished ZNF768-mediated promoter activation (Figure 5G). Furthermore, promoter truncation constructs (Figure 5H) identified the BS1-containing region as essential for activation (Figure 5I). Accordingly, mutating BS1 alone fully abrogated ZNF768-driven transactivation (Figure 5J, K). Finally, ChIP-PCR assays confirmed direct binding of ZNF768 to the SLC7A11 promoter with strongest enrichment at the functionally critical BS1 site (Figure 5L). To directly link lenvatinib treatment to ZNF768-driven SLC7A11 transcription, ChIP-qPCR was performed in lenvatinib-treated and vehicle-treated Hep3B cells. Lenvatinib treatment significantly enhanced ZNF768 occupancy at the BS1 site of the SLC7A11 promoter compared to vehicle control, which was consistent with the lenvatinib-induced increase in ZNF768 protein levels shown in Figure 7A (Figure S5L).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

ZNF768 directly binds and activates the SLC7A11 promoter. (A) Volcano plot of RNA-seq data from ZNF768 KD Hep3B cells vs. control, identifying SLC7A11 as a top downregulated gene. (B–D) Validation of ZNF768-mediated regulation of SLC7A11 at the mRNA and SLC7A11 protein levels in ZNF768 KD (Hep3B and Huh7) and OE (HCCLM3 and Li-7) cell lines. (E) The ZNF768 binding motif predicted by the JASPAR database. (F) Schematic of the human SLC7A11 promoter showing three putative ZNF768 binding sites (BS1, BS2, and BS3). (G) Luciferase reporter assay in 293T cells showing that mutation of all three binding sites abrogates ZNF768-mediated activation of the SLC7A11 promoter. (H-K) Schematics and corresponding luciferase assays using truncated and site-specific mutant constructs of the SLC7A11 promoter, identifying BS1 as the essential site for ZNF768-mediated transactivation. (L) ChIP-PCR confirming direct ZNF768 binding at the BS1 site. (M) Representative IHC images of serial sections from HCC clinical specimens showing concordant high ZNF768/SLC7A11 expression with low 4-HNE staining and vice versa. (N, O) Spearman correlation analysis confirming a significant positive correlation between ZNF768 and SLC7A11 and a negative correlation between ZNF768 and 4-HNE in HCC tissues. Data are presented as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, and ns by Student’s t-test or one-way ANOVA. ChIP-PCR, chromatin immunoprecipitation-PCR; HCC, hepatocellular carcinoma; IHC, immunohistochemistry; KD, knockdown; ns, non-significant; OE, overexpression; SD, standard deviation.

Clinically, consecutive sections of HCC specimens showed high ZNF768 expression correlated with elevated SLC7A11 and reduced ferroptosis marker 4-hydroxynonenal (4-HNE). Conversely, low ZNF768 expression corresponded to diminished SLC7A11 and enhanced 4-HNE staining (Figure 5M). Moreover, histochemical scoring confirmed a significant positive correlation between ZNF768 and SLC7A11 (Figure 5N), as well as a negative correlation with 4-HNE levels (Figure 5O). Together, ZNF768 suppresses ferroptosis in HCC by directly binding the BS1 site to activate SLC7A11 transcription.

SLC7A11 is the critical downstream effector mediating ZNF768-dependent ferroptosis defense

Rescue experiments were performed in Hep3B and Huh7 cells to determine whether ZNF768 promotes tumors via its transcriptional target SLC7A11. Western blot confirmed ZNF768 knockdown and concurrent SLC7A11 overexpression (Figures 6A and S6A). Re-expressing SLC7A11 reversed ferroptotic hallmarks and restored the diminished GSH:GSSG ratio (Figure 6B) and reduced intracellular iron and MDA accumulation (Figure 6C, D). Furthermore, C11-BODIPY staining showed that SLC7A11 overexpression attenuated elevated lipid ROS (Figures 6E, F and S6B). Moreover, TEM revealed that SLC7A11 re-expression mitigated ferroptosis-associated mitochondrial damage, including shrunken membranes and reduced cristae (Figure 6G).

Functionally, this molecular rescue restored ferroptosis defense capacity. When challenging cells with erastin, a system Xc⁻ inhibitor²⁷, ZNF768 depletion increased erastin sensitivity [reduced half-maximal inhibitory concentration (IC₅₀)], in which SLC7A11 overexpression was fully reversed (Figure 6H). Similarly, the antioxidant, N-acetylcysteine (NAC), rescued impaired cell viability and colony formation in erastin-treated, ZNF768-knockdown cells (Figure 6I, J). In vivo validation using a Hep3B xenograft model (control, shZNF768-1, and shZNF768-1 + SLC7A11 groups) demonstrated that ZNF768 knockdown suppressed tumor growth, an effect entirely rescued by SLC7A11 re-expression (Figure 6K–M). Together, these results established SLC7A11 as the key downstream effector for ZNF768-mediated ferroptosis suppression and HCC progression.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 6

SLC7A11 is the key downstream effector of ZNF768-mediated ferroptosis defense. (A) Western blot confirming SLC7A11 re-expression in ZNF768 KD Hep3B and Huh7 cells. (B–D) Restoration of SLC7A11 in ZNF768-depleted cells rescued the diminished GSH:GSSG ratio and normalized intracellular Fe²⁺ and MDA levels. (E, F) Representative images and flow cytometry analysis of C11-BODIPY staining showing that SLC7A11 re-expression abrogated the increase in lipid ROS caused by ZNF768 loss (scale bar in Figure 6E = 20 μm). (G) TEM imaging demonstrating that SLC7A11 restoration ameliorated the mitochondrial damage observed in ZNF768-KD cells (scale bar = 1 or 2 μm). (H) SLC7A11 OE reversed the increased sensitivity to erastin in ZNF768-depleted cells, as shown by IC₅₀ values. (I, J) Treatment with the antioxidant, NAC, rescued the impaired viability and colony formation ability of ZNF768-KD cells under erastin-induced stress. (K–M) In a subcutaneous xenograft model using ZNF768-KD Hep3B cells, tumor growth suppression caused by ZNF768 KD was completely reversed by SLC7A11 re-expression, as shown by tumor growth curves, images of excised tumors, and tumor weights (n = 6 per group). Data are presented as the mean ± SD or mean ± SEM. *P < 0.05, ***P < 0.001, and ****P < 0.0001, and ns by Student’s t-test or ANOVA. KD, knockdown; MDA, malondialdehyde; ns, non-significant; OE, overexpression; SD, standard deviation; SEM, standard error of the mean; TEM, transmission electron microscope.

Lenvatinib triggers an adaptive feedback loop involving the AKT-EGR1-ZNF768 axis to limit ferroptosis

Lenvatinib treatment increased ZNF768 and SLC7A11 expression in a dose-dependent manner in Hep3B and Huh7 cells (Figures 7A and S7A). Notably, upregulation of both proteins was already detectable at concentrations of 1–5 μM, which approximate the estimated intratumoral drug levels achievable at standard clinical doses. Higher concentrations (up to 20 μM) were included to capture the full dose-response relationship and are consistent with the concentrations used in published in vitro lenvatinib studies. Time-course experiments at 10 μM further confirmed sustained induction of both proteins over 24 h (Figures 7B and S7B). By integrating transcriptome sequencing data from lenvatinib-stimulated cells with JASPAR database predictions¹⁸, EGR1 was identified as the most significantly upregulated transcription factor potentially regulating ZNF768 (Figure 7C, D). Lenvatinib indeed induced time-dependent EGR1 protein elevation in both cell lines (Figures 7E and S7C). EGR1 overexpression subsequently increased ZNF768 mRNA and SLC7A11 protein levels, whereas EGR1 silencing reduced ZNF768 mRNA and SLC7A11 protein levels, confirming EGR1-mediated ZNF768 induction (Figures 7F, G and S7D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7

Lenvatinib treatment triggers an adaptive feedback loop that upregulates ZNF768. (A, B) Western blot analysis showing dose- and time-dependent upregulation of ZNF768 and SLC7A11 in Hep3B and Huh7 cells treated with lenvatinib. (C, D) Integrated analysis of transcriptome data and JASPAR database predictions identifying EGR1 as a key transcription factor mediating lenvatinib-induced ZNF768 expression, TFs. (E) Lenvatinib treatment induced time-dependent upregulation of EGR1 protein. (F, G) EGR1 OE increased, while siRNA-mediated EGR1 silencing decreased ZNF768 mRNA and ZNF768 protein levels. (H, I) Luciferase assays showing that EGR1 OE enhanced, while silencing reduced ZNF768 promoter activity. (J, K) Treatment with lenvatinib (10 μM) activated the ZNF768 promoter, and this effect was abrogated by EGR1 silencing. (L) Schematic of the ZNF768 promoter showing three putative EGR1 binding motifs. (M) ChIP-PCR confirming direct binding of EGR1 to all three sites on the ZNF768 promoter. (N, O) Lenvatinib or the AKT inhibitor, MK2206-2HCl, suppressed AKT phosphorylation and increased EGR1 and ZNF768 protein levels. (P) EGR1 silencing reversed the lenvatinib-induced increase in ZNF768 protein. (Q, R) Silencing of EGR1 or ZNF768 enhanced lenvatinib-induced lipid peroxidation and cell death, effects that were rescued by re-expressing ZNF768 in EGR1-silenced cells (scale bar = 20 μm). Data are presented as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, and ns by Student’s t-test or ANOVA. ChIP-PCR, chromatin immunoprecipitation-PCR; ns, non-significant; OE, overexpression; SD, standard deviation.; TFs, transcription factors.

Whether EGR1 directly activates ZNF768 transcription was determined next. Luciferase assays showed that EGR1 overexpression enhanced ZNF768 promoter activity, while knockdown suppressed ZNF768 promoter activity (Figure 7H, I). Notably, lenvatinib-induced promoter activation was abolished by EGR1 silencing, demonstrating the essential role of EGR1 (Figure 7J, K). Bioinformatics analysis revealed three potential EGR1 binding motifs on the ZNF768 promoter (Figure 7L), and ChIP-PCR confirmed EGR1 binding to all sites in Hep3B cells (Figure 7M). Consistent with reports that the PI3K/AKT pathway negatively regulates EGR1^28,29, lenvatinib was shown to inhibit AKT phosphorylation, which coincided with EGR1 upregulation (Figures 7N and S7E). Importantly, pharmacologic AKT inhibition with MK2206 was sufficient to mimic the lenvatinib effect by increasing EGR1 and ZNF768 levels (Figures 7O and S7F), suggesting that AKT inhibition is a critical mediator of this process. Consistently, EGR1 silencing reversed lenvatinib-induced ZNF768 upregulation (Figures 7P and S7G).

Finally, the functional impact of EGR1-ZNF768 in lenvatinib-induced ferroptosis was assessed. Silencing EGR1 or ZNF768 intensified lenvatinib-induced lipid peroxidation in Hep3B cells. ZNF768 re-expression rescued this effect in EGR1-silenced cells, identifying ZNF768 as the EGR1 key ferroptosis suppressor (Figures 7Q and S7H). Similarly, EGR1 or ZNF768 silencing reduced cell viability under lenvatinib treatment and ZNF768 restoration reversed viability loss in EGR1-knockdown cells (Figure 7R). These results demonstrated that HCC cells survive lenvatinib-induced ferroptosis via an AKT-EGR1-ZNF768 pathway that sustains cell viability. Consistent with this regulatory cascade, EGR1 overexpression increased SLC7A11 mRNA and SLC7A11 protein levels, while EGR1 silencing reduced SLC7A11 mRNA and SLC7A11 protein levels (Figure S8A, B), providing direct evidence that EGR1 drives SLC7A11 expression through ZNF768.

Disrupting the ZNF768 adaptive shield synergizes with lenvatinib to induce ferroptosis and suppress HCC growth

Additional analyses in the lenvatinib-based therapy cohort were performed to further evaluate the clinical relevance of ZNF768 expression in predicting treatment outcome. ZNF768 IHC scores were significantly higher in patients with PD compared to NPD (Figure 8A, B). Multivariate logistic regression analysis demonstrated that high ZNF768 expression remained an independent predictor of disease progression after adjusting for tumor stage and baseline AFP levels (odds ratio: 9.28; 95% confidence interval: 2.28–49.36, P < 0.01, Figure S9A). ROC curve analysis identified an optimal ZNF768 expression cut-off with an area under the curve of 0.661 for predicting disease progression (Figure S9B). HCC patients receiving lenvatinib-based therapy were analyzed to establish clinical relevance. IHC revealed that higher ZNF768 expression correlated with poorer outcomes because patients with PD exhibited significantly elevated ZNF768 levels compared to NPD (Figure 8A, B). Based on this correlation, we hypothesized that targeting ZNF768 could potentiate lenvatinib efficacy. Therefore, we tested this in vivo using a Hepa1-6 subcutaneous tumor model in BALB/c nude mice. An AAV8 vector delivered shZNF768 intratumorally (Figure 8C). While lenvatinib monotherapy showed modest tumor growth inhibition, combining lenvatinib with ZNF768 knockdown profoundly suppressed tumor growth (Figure 8D, E). This synergistic effect was further confirmed by significantly reduced tumor weights in the combination group (Figure 8F).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8

Targeting the ZNF768 adaptive shield synergizes with lenvatinib to suppress HCC growth. (A, B) Representative IHC staining and quantitative analysis showing significantly higher ZNF768 expression in tumors from HCC patients with PD compared to NPD following lenvatinib-based therapy. (C) Schematic of the in vivo therapeutic experiment timeline using a Hepa1-6 subcutaneous xenograft model in BALB/c nude mice (n = 6 per group). The AAV8-shZNF768 vector used in this experiment carries an shRNA sequence targeting murine Zfp768 and was verified for species specificity prior to use. (D–F) Combination of lenvatinib and intratumoral AAV-mediated ZNF768 KD resulted in profound tumor growth suppression compared to treatment alone, as shown by tumor growth curves, images of excised tumors, and endpoint tumor weights. (G) MDA levels were significantly elevated in tumor tissues from the combination therapy group, indicating robust in vivo induction of ferroptosis. (H–K) Measurement of serum CRE, ALT, and AST levels, along with mouse body weights, demonstrating no significant toxicity from the combination therapy. Data are presented as the mean ± SD or mean ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001 by ANOVA. KD, knockdown; MDA, malondialdehyde; NPD, non-progressive disease; PD, progressive disease; SD, standard deviation; SEM, standard error of the mean.

p-AKT, EGR1, ZNF768, and SLC7A11 protein levels were assessed across the four treatment groups in vitro and in tumor tissues to validate the signaling sequence underlying the combination therapy. Lenvatinib suppressed p-AKT and elevated EGR1 independent of ZNF768 status, confirming that the upstream adaptive signal remained active. SLC7A11 protein levels were substantially reduced in the combination group alone, demonstrating that ZNF768 knockdown effectively blocked the downstream antioxidant response. Concordant results were obtained from tumor tissue lysates (Figure S9C, D). Lipid peroxidation in tumor tissues was assessed to confirm the enhanced therapeutic effect involved ferroptosis. Consistent with in vitro data, the combination group showed significantly elevated levels of the ferroptosis marker MDA (Figure 8G). IHC corroborated this finding, revealing stronger staining for another ferroptosis marker, 4-HNE, following ZNF768 knockdown. This analysis also reconfirmed the inverse regulatory relationship between ZNF768 and SLC7A11 in vivo (Figure S9E). The combination therapy was well-tolerated with no significant changes observed in overall body weight (Figure 8K), kidney function markers (CRE; Figure 8H), or liver function markers (ALT and AST; Figure 8I, J) across groups, indicating a favorable safety profile. Together, the findings illustrate the following central mechanism: lenvatinib triggers EGR1-dependent ZNF768 upregulation, which transcriptionally activates SLC7A11 to promote glutathione synthesis and confer protection against ferroptosis. Targeted AAV-mediated ZNF768 silencing disrupts this adaptive defense loop, thereby sensitizing HCC cells to lenvatinib-induced ferroptosis and potentiating therapeutic efficacy (Figure 9).

• Download figure
• Open in new tab
• Download powerpoint

Figure 9

Schematic model of the EGR1-ZNF768-SLC7A11 adaptive axis and the therapeutic implications in HCC. (1) Lenvatinib treatment promotes EGR1 protein accumulation in HCC cells. (2) Upon nuclear translocation, EGR-1 binds to the ZNF768 promoter, driving transcription. (3) In turn, the elevated ZNF768 directly activates the SLC7A11 promoter, (4) leading to increased SLC7A11 expression at the plasma membrane and enhancing the import of extracellular cystine. (5) This cystine fuels the synthesis of GSH inside the cell, and when catalyzed by GPX4, serves as an electron donor to detoxify membrane lipid peroxides, thereby inhibiting ferroptosis and building an adaptive antioxidant defense against lenvatinib. (6) To overcome this resistance, AAV-mediated silencing of ZNF768 mRNA disrupts this adaptive feedback loop, preventing SLC7A11 upregulation and GSH accumulation, and restoring ferroptosis sensitivity. AAV, adeno-associated virus; EGR1, early growth response protein 1; GPX4, glutathione peroxidase 4; GSH, glutathione; HCC, hepatocellular carcinoma; mRNA, messenger RNA; SLC7A11, solute carrier family 7 member 11; ZNF768, zinc finger protein 768. Figure created with BioRender.