ATF7IP inhibits Sorafenib-induced ferroptosis in hepatocellular carcinoma cells by inhibiting CYB5R2 transcription and stabilizing PARK7 protein

Abstract

Ferroptosis, an iron-dependent form of programmed cell death, arises from the accumulation of lipid peroxides at toxic levels. Sorafenib, a first-line treatment for advanced hepatocellular carcinoma, shows limited clinical efficacy due to drug resistance. However, the mechanisms underlying Sorafenib resistance, especially related to ferroptosis, remain poorly understood. In this study, we identify activating transcription factor 7-interacting protein (ATF7IP) as a key inhibitor of ferroptosis. ATF7IP depletion promotes Sorafenib-induced ferroptosis, resulting in decreased cell viability, reduced cellular glutathione (GSH) levels, increased lipid peroxidation, and altered mitochondrial crista structure. Notably, ATF7IP knockdown shows cooperative effects with Sorafenib in inhibiting hepatocellular carcinoma growth in mice. Mechanistically, ATF7IP interacts with SET domain bifurcated histone lysine methyltransferase 1 (SETDB1) to epigenetically silence the transcription of cytochrome b5 reductase 2 (CYB5R2), thereby reducing cellular Fe²⁺ levels. Meanwhile, ATF7IP stabilizes the antioxidant sensor Parkinsonism-associated deglycase (PARK7) protein which preserves the transsulfuration pathway to produce GSH, also leading to the inhibition of Sorafenib-induced ferroptosis. In conclusion, our findings identify ATF7IP as a critical ferroptosis inhibitor and represent ATF7IP as a novel therapeutic target for Sorafenib-based combination therapies of hepatocellular carcinoma.

Graphical abstract

1. Introduction

Ferroptosis, unlike other forms of cell death, is a specialized process characterized by morphological changes such as smaller mitochondria with increased mitochondrial membrane density, and reduction or vanishment of mitochondrial cristae [1]. It is driven by iron-dependent accumulation of lipid peroxides [1]. Therefore, excessive intracellular accumulation of Fe²⁺, which triggers the Fenton reaction [2], and abnormal accumulation of polyunsaturated fatty acids [3], can induce ferroptosis. While cellular antioxidant systems, such as System Xc⁻ [4], the transsulfuration pathway [5], and Coenzyme Q10 [6], prevent the onset of ferroptosis. It has been demonstrated that dysregulated ferroptosis leads to various human diseases, including cancer [7]. Notably, ferroptosis holds the potential to overcome cancer cell resistance to conventional therapies [8,9], and the identification of FDA-approved drugs as ferroptosis inducers further highlights this possibility [10]. Therefore, gaining a better understanding of the underlying mechanisms regulating ferroptosis is critical for the development of novel therapies for cancers.

Hepatocellular carcinoma (HCC) is one of the leading causes of cancer-related mortality worldwide [11]. Patients diagnosed with advanced-stage hepatocellular carcinoma which is not amenable to surgical resection often have poor prognoses [12,13]. Sorafenib, a molecular-targeted therapy with proven survival benefits in hepatocellular carcinoma, has been shown to induce ferroptosis [14,15]. Mechanistically, Sorafenib triggers ferroptosis through inhibiting the cystine/glutamate transporter SLC7A11 by enhancing the binding of Beclin-1 to SLC7A11, reducing the synthesis of intracellular antioxidant GSH [16]. However, its clinical efficacy is limited, due to both primary and acquired resistance [17]. Enhancing Sorafenib-induced ferroptosis thus can improve the therapeutic efficacy of Sorafenib in hepatocellular carcinoma patients.

Our previous research identified several epigenetic factors contributing to Sorafenib resistance in hepatocellular carcinoma cells using an epigenetic factors-targeted CRISPR/Cas9 library screening [18]. Among these, activating transcription factor 7-interacting protein (ATF7IP) emerged as a prominent candidate, ranking first in our analysis. However, whether it inhibits Sorafenib-induced ferroptosis and the underlying mechanisms were unclear. ATF7IP is a multifunctional protein that primarily regulates methylation of histone H3 at lysine 9 (H3K9) and promotes heterochromatin formation through its interaction with SET domain bifurcated histone lysine methyltransferase 1 (SETDB1) [19,20]. As a binding partner of SETDB1, ATF7IP facilitates its nuclear localization and increases its monoubiquitination, enhancing its enzymatic activity [21].

In this study, we found that ATF7IP inhibits Sorafenib-induced ferroptosis in cells and mice. Mechanically, ATF7IP together with SETDB1 represses the transcription of cytochrome b5 reductase 2 (CYB5R2), an enzyme involved in Fe²⁺ production [22,23]. On the other hand, ATF7IP interacts with Parkinsonism-associated deglycase (PARK7), a key antioxidant protein [24,25]. This interaction stabilizes PARK7 and restores intracellular GSH levels via the transsulfuration pathway, counteracting the suppressive effect of Sorafenib on the cellular antioxidant system. Collectively, our findings identify ATF7IP as a novel ferroptosis inhibitor, presenting it as a therapeutic target for Sorafenib-based combination therapies of hepatocellular carcinoma.

2. Results

2.1. ATF7IP deficiency sensitizes hepatocellular carcinoma cells to Sorafenib-induced ferroptosis

Sorafenib, a first-line treatment for hepatocellular carcinoma, induces ferroptosis in cancer cells. However, its usage is often limited by the development of resistance in hepatocellular carcinoma patients. In our previous study, a CRISPR/Cas9 library targeting epigenetic factors was used to identify potential contributors to Sorafenib resistance in hepatocellular carcinoma cells [18]. Our screening identified ATF7IP as a potential Sorafenib resistance gene which ranked first [18]. To validate the findings from our CRISPR/Cas9 knockout library screening, we first generated stable ATF7IP knockdown cells by infecting hepatocellular carcinoma cells with lentiviruses expressing ATF7IP shRNAs (Fig. S1A). MTT assays revealed that ATF7IP knockdown significantly reduced the 50 % inhibitory concentration (IC50) of Sorafenib on HepG2, Hep3B, and Huh7 cells (Fig. 1A and S1B). To further evaluate the impact of ATF7IP knockdown on Sorafenib-induced cell death, we performed Calcein-AM/PI double-staining assays, which showed that ATF7IP knockdown increased the percentage of dead cells following Sorafenib treatment (Fig. 1B and S1C). Given that Sorafenib induces ferroptosis in hepatocellular carcinoma cells, we performed ferroptosis-related assays in control and ATF7IP knockdown cells to determine if ATF7IP knockdown influences Sorafenib-induced ferroptosis. Ferroptosis, driven by iron-dependent phospholipid peroxidation, is accompanied by elevation in lipid reactive oxygen species (ROS) and reduction in cellular GSH [26,27]. Morphologically, ferroptotic cells exhibit condensed mitochondrial membrane densities, reduced mitochondrial volume, and diminished mitochondrial cristae [28]. Our results showed that ATF7IP knockdown decreased cellular GSH levels (Fig. 1C and S1D) and increased cellular lipid ROS levels after Sorafenib treatment (Fig. 1D and S1E). Furthermore, 4-Hydroxynonenal (4-HNE), a terminal product of lipid oxidation that plays a crucial role in ferroptosis [29,30] was also examined by liquid chromatography-tandem mass spectrometry, revealing that ATF7IP knockdown elevated intracellular 4-HNE levels upon Sorafenib treatment (Fig. 1E). Using transmission electron microscopy, we observed that the number of mitochondrial cristae decreased in ATF7IP knockdown cells following Sorafenib treatment (Fig. 1F).

Fig. 1.

ATF7IP deficiency sensitizes hepatocellular carcinoma cells to Sorafenib-induced ferroptosis. A. MTT assay was performed to calculate the IC50 of Sorafenib in HepG2 and Hep3B cells with ATF7IP knockdown or not. Data are mean ± SEM for n = 3; ∗p < 0.05, ∗∗p < 0.01 (one-way ANOVA followed by Dunnett's multiple comparisons). B. Control or ATF7IP knockdown HepG2 and Hep3B cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Dead and alive cells were labeled by Calcein-AM/PI staining, with quantitative analysis of cell death percentage on the right. Scale bar: 50 μm. C-D. Control or ATF7IP knockdown HepG2 and Hep3B cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Cellular GSH (C) and cellular lipid ROS (D) were detected. E. Control or ATF7IP knockdown HepG2 cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Relative levels of 4-HNE were analyzed by liquid chromatography-tandem mass spectrometry. Data are mean ± SD for n = 4; ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons). F. Mitochondria crista was observed with a transmission electron microscope (Scale bar: 1 μm). G-H. Control or ATF7IP knockdown HepG2 cells were treated with DMSO, Sorafenib (10 μM), or Sorafenib together with DFO (3 μM) or Fer-1 (5 μM) for 48 h. MTT assay was performed (G) and cellular lipid ROS was measured (H). For figures B-D and G-H, data are mean ± SD for n = 3; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons).

To further validate that ATF7IP suppression enhances Sorafenib-induced ferroptosis, specific ferroptosis inhibitors deferoxamine (DFO) and ferrostatin-1 (Fer-1) were added. DFO chelates ferrous iron, preventing iron-mediated Fenton reactions [31], while Fer-1 scavenges alkoxyl radicals generated from lipid peroxides [32]. Both inhibitors effectively reversed the phenotype changes induced by ATF7IP depletion upon Sorafenib treatment, as confirmed by MTT assays and lipid ROS analysis (Fig. 1G and H). Moreover, to further elucidate the role of ATF7IP in inhibiting ferroptosis, we treated control and ATF7IP knockdown hepatocellular carcinoma cells with Erastin which is the first discovered inducer of ferroptosis through inhibiting the cystine/glutamate transporter SLC7A11 [33]. Our results showed that ATF7IP knockdown enhanced ferroptosis induced by Erastin, as evidenced by decreased cell viability, and elevated cellular lipid ROS levels (Fig. S1F and S1G) compared to control cells. Collectively, these results underscore the crucial role of ATF7IP in inhibiting Sorafenib-induced ferroptosis in hepatocellular carcinoma cells.

2.2. Depletion of SETDB1 promotes Sorafenib-induced ferroptosis

SETDB1 is a lysine methyltransferase responsible for methylating histone H3 at lysine 9 (H3K9), playing a critical role in H3K9 trimethylation (H3K9me3)-mediated silencing of genes and retrotransposons [34,35]. As a cofactor of SETDB1, ATF7IP facilitates its nuclear localization and increases its enzymatic activity. Loss of ATF7IP was reported to result in a reduction in H3K9me3 levels and derepression of SETDB1-regulated genes, which is similar to the effects induced by SETDB1 inactivation or depletion [21,36,37]. ATF7IP depletion also led to a reduction of SETDB1 protein and H3K9me3 levels in hepatocellular carcinoma cells (Fig. 2A). Given the close functional relationship between SETDB1 and ATF7IP, we further investigated the role of SETDB1 in Sorafenib-induced ferroptosis. We first generated SETDB1 knockdown HepG2 cell lines and confirmed the knockdown efficiency via Western blotting (Fig. 2B). Using MTT assays, we showed that SETDB1 knockdown significantly reduced the IC50 of Sorafenib on HepG2 cells (Fig. 2C). Calcein-AM/PI double-staining assays further revealed an increased percentage of dead cells after SETDB1 knockdown upon Sorafenib treatment (Fig. 2D). Moreover, significant reduction in GSH (Fig. 2E) and increase in lipid ROS levels (Fig. 2F and G) were indicated. In addition, a decrease in mitochondrial cristae number was observed through transmission electron microscopy (Fig. 2H) in SETDB1 knockdown cells after Sorafenib treatment, indicating that reduced SETDB1 expression enhanced Sorafenib-induced ferroptosis. Moreover, the phenotype changes induced by SETDB1 depletion under Sorafenib treatment were effectively reversed by DFO and Fer-1, as confirmed through MTT assays and lipid ROS analysis (Fig. 2I and J). Meanwhile, results from MTT and lipid ROS assays demonstrated that SETDB1 depletion similarly enhanced ferroptosis induced by Erastin (Fig. S2A and S2B). In conclusion, these findings reveal that SETDB1 also suppresses Sorafenib-induced ferroptosis in hepatocellular carcinoma cells.

Fig. 2.

Depletion of SETDB1 promotes Sorafenib-induced ferroptosis. A. Western blotting was conducted to examine the expression of ATF7IP, SETDB1 and H3K9me3 in control and ATF7IP knockdown HepG2 cells. B. Cell lysate of control and SETDB1 knockdown HepG2 cells were obtained and subjected to Western blotting. C. MTT assay was performed to calculate the IC50 of Sorafenib in HepG2 cells with SETDB1 knockdown or not. Data are mean ± SEM for n = 3; ∗∗p < 0.01 (Student's t-test). D. Control or SETDB1 knockdown HepG2 cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Dead and alive cells were labeled by Calcein-AM/PI staining, with quantitative analysis of cell death percentage on the right. Scale bar: 50 μm. E-F. Control or SETDB1 knockdown HepG2 cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Cellular GSH (E) and cellular lipid ROS (F) were detected. G. Control or SETDB1 knockdown HepG2 cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Relative levels of 4-HNE were analyzed by liquid chromatography-tandem mass spectrometry. Data are mean ± SD for n = 4; ns, p > 0.05, ∗∗p < 0.01 (two-way ANOVA followed by Tukey's test for multiple comparisons). H. Mitochondria crista was observed with a transmission electron microscope (Scale bar: 1 μm). I-J. Control or SETDB1 knockdown HepG2 cells were treated with DMSO, Sorafenib (10 μM), or Sorafenib together with DFO (3 μM) or Fer-1 (5 μM) for 48 h. MTT assay was performed (I) and cellular lipid ROS was measured (J). For figures D-F and I-J, data are mean ± SD for n = 3; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons).

2.3. ATF7IP-SETDB1 complex inhibits Sorafenib-induced ferroptosis by suppressing CYB5R2 transcription

To elucidate the precise mechanism by which the ATF7IP-SETDB1 complex inhibits Sorafenib-induced ferroptosis, RNA-seq was performed in HepG2 cells expressing control shRNAs, ATF7IP shRNAs, and SETDB1 shRNAs, respectively. RNA-seq analysis revealed 1,611 differentially expressed genes in ATF7IP knockdown cells, with 704 genes upregulated, and 991 differentially expressed genes in SETDB1 knockdown cells, of which 604 genes were upregulated (Fig. 3A). We conducted a joint analysis for the upregulated gene sets in SETDB1 knockdown cells and ATF7IP knockdown cells, and identified 240 overlapping genes (Fig. 3B). Gene Ontology analysis of these overlapping genes revealed significant enrichment in the “unsaturated fatty acid biosynthesis process” pathway (Fig. 3C). Ferroptosis is driven by excessive lipid peroxidation [38], and polyunsaturated fatty acids are the most susceptible lipids to peroxidation during ferroptosis [39]. Therefore, we picked genes within this pathway, including ALOX15 and ALOX15B, for further validation. In addition, potential target genes involved in the regulation of ferroptosis, including CYB5R2 [22], PLTP [40], and TP73 [41], were also incorporated into our analysis. RT-qPCR assays identified CYB5R2 as the gene with the most pronounced expression changes among the selected candidates after ATF7IP and SETDB1 knockdown (Fig. 3D). Furthermore, to detect whether ATF7IP and SETDB1 directly bind to the promoter of CYB5R2, ChIP-quantitative PCR (ChIP-qPCR) assays were performed in HepG2 cells using ATF7IP and SETDB1 antibodies, respectively. The results validated the direct binding of ATF7IP and SETDB1 to the promoter of CYB5R2 (Fig. 3E). Moreover, ATF7IP and SETDB1 depletion decreased the enrichment of H3K9me3 on the promoter of CYB5R2 (Fig. 3F).

Fig. 3.

ATF7IP-SETDB1 complex suppresses the transcription of CYB5R2 via catalyzing H3K9me3 on its promoter. A. HepG2 cells stably expressing control shRNAs, ATF7IP shRNAs, or SETDB1 shRNAs were cultured with 10 μM Sorafenib for 48 h, and RNA was extracted for RNA sequencing. Volcano plot shows differentially expressed genes in ATF7IP knockdown cells or SETDB1 knockdown cells compared with control cells. Genes meeting the criteria of padj < 0.05 and fold change ≥ 2 are represented as red dots for upregulated genes and blue dots for downregulated genes. B. The Venn diagram shows the overlap of upregulated genes in ATF7IP and SETDB1 knockdown cells. C. The overlapped gene set in (B) was subjected to Gene Ontology Biological Process analysis. D. Control and ATF7IP knockdown HepG2 cells or SETDB1 knockdown HepG2 cells were treated with DMSO and Sorafenib (10 μM) for 48 h. RT-qPCR was performed to detect the mRNA expression levels of the indicated genes. Data are mean ± SD for n = 3; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (One-way ANOVA followed by Dunnett's multiple comparisons for ATF7IP knockdown group and Student's t-test for SETDB1 knockdown group). E. ChIP assays were performed in HepG2 cells using anti-ATF7IP, anti-SETDB1, or IgG, and then qPCR assays were used to detect their enrichment with primer pairs targeting the promoter of CYB5R2. Data are mean ± SD for n = 3; ∗∗∗∗p < 0.0001 (Student's t-test). F. ChIP-qPCR was used to analyze the enrichment of H3K9me3 in the promoter of CYB5R2 in control and ATF7IP knockdown HepG2 cells or SETDB1 knockdown HepG2 cells. Data are mean ± SD for n = 3; ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons).

CYB5R2 belongs to the cytochrome b5 reductase family that plays a key role in maintaining redox balance and providing Fe²⁺ within cells [22,23,42]. To prove that the upregulation of CYB5R2 mediates the promotion of ferroptosis caused by ATF7IP and SETDB1 knockdown, CYB5R2 was knocked down using its specific siRNAs in ATF7IP or SETDB1 knockdown cells, and the knockdown efficiency was determined by Western blotting (Fig. 4A). MTT and Calcein-AM/PI double-staining assays showed that diminished CYB5R2 expression attenuated the reduction in cell viability and the increase in cell death (Fig. 4B and C) in ATF7IP and SETDB1 knockdown cells under Sorafenib treatment. Further ferroptosis-related experiments revealed that CYB5R2 knockdown inhibited ferroptosis mediated by ATF7IP or SETDB1 knockdown under Sorafenib treatment, as evidenced by increased cellular GSH and decreased cellular lipid ROS levels (Fig. 4D and E). As stated above, CYB5R2 is a source of Fe²⁺ within cells, we further examined cellular Fe²⁺ level using the FerroOrange dye. The findings indicated that depletion of ATF7IP and SETDB1 led to an increase in cellular Fe²⁺ levels after Sorafenib treatment, which was mitigated by CYB5R2 knockdown (Fig. 4F). These experiments collectively demonstrate that ATF7IP and SETDB1 suppress CYB5R2 transcription by catalyzing H3K9me3 on its promoter region, thereby inhibiting the generation of intracellular Fe²⁺ and consequently suppressing Sorafenib-induced ferroptosis.

Fig. 4.

ATF7IP-SETDB1 complex inhibits Sorafenib-induced ferroptosis through downregulating CYB5R2 transcription. A. Control cells, ATF7IP knockdown HepG2 cells or SETDB1 knockdown HepG2 cells were transfected with control siRNAs or CYB5R2 siRNAs, and then cell lysate was subjected to Western blotting. B. Indicated cells were treated with DMSO or 10 μM Sorafenib for 48 h, and the cell viability was monitored with MTT assay. C. Calcein-AM/PI staining was performed in the indicated cells after the treatment with DMSO or Sorafenib (10 μM) for 48 h. Scale bar: 100 μm. D-E. Cellular GSH (D) and cellular lipid ROS (E) were assessed in the indicated cells treated with DMSO or Sorafenib (10 μM) for 48 h. F. Intracellular Fe²⁺ was detected by the probe of FerroOrange under a fluorescent microscope (red, FerroOrange-stained Fe²⁺). Quantitative analysis of fluorescence intensity was performed using ImageJ. Scale bar: 5 μm. For figures B–F, data are mean ± SD for n = 3; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons).

2.4. ATF7IP interacts with PARK7 and promotes its stabilization

As shown above, the elevated level of Sorafenib-induced ferroptosis promoted by ATF7IP depletion is higher than that promoted by SETDB1 knockdown (Fig. 4D and E), which prompts us to propose the hypothesis that ATF7IP also suppresses ferroptosis by interacting with other proteins. To test this, we applied Flag affinity purification and mass spectrometry to identify proteins that potentially interact with ATF7IP in vivo. The lysates of HEK-293FT cells expressing Flag-ATF7IP were prepared and subjected to Flag affinity purification. The eluates were resolved on SDS-PAGE and silver-stained (Fig. 5A). Mass spectrometric analysis of the resolved protein bands showed that besides SETDB1, PARK7 was also co-purified with ATF7IP (Supplementary Table 2). PARK7 is a well-known antioxidative protein that can protect cells against oxidative stress and stabilize cellular GSH through the transsulfuration pathway, participating in ferroptosis regulation [24,25,43].

Fig. 5.

ATF7IP interacts with PARK7 and promotes its stabilization. A. Cell lysates from HEK-293FT cells transfected with Flag-ATF7IP or control vectors were immunoprecipitated using anti-Flag affinity gel and bound proteins were eluted with Flag peptides. Samples were resolved by SDS-PAGE and silver-stained followed by mass spectrometry analysis. B. Endogenous co-immunoprecipitation was carried out in HepG2 cells using anti-ATF7IP or anti-PARK7, followed by Western blotting. C. The schematic representations of ATF7IP full-length and truncations. D. Whole-cell lysates from HEK-293FT cells transfected with control vectors, Flag-ATF7IP-1, Flag-ATF7IP-2, or Flag-ATF7IP-3 expression vectors were subjected to immunoprecipitation using anti-Flag, followed by immunoblotting with the indicated antibodies. E. The cytoplasmic and nuclear fractions of HepG2 cells treated with DMSO or Sorafenib (10 μM) were obtained and subjected to Western blotting. α-Tubulin was used as a cytoplasm marker, and Lamin B1 and H3 were used as the nucleus markers. F. Western blotting was used to detect the protein levels of PARK7 in control and ATF7IP knockdown HepG2 cells with DMSO or Sorafenib treatment (10 μM, 48 h). G. The half-lives of PARK7 protein were examined in control and ATF7IP knockout cells using CHX pulse-chase assays. H. Control and ATF7IP knockout HepG2 cells were treated with Sorafenib (10 μM) for 48 h, followed by treatment with DMSO or MG132 (10 μM) for an additional 4 h. Then the protein level of ATF7IP and PARK7 was detected by Western blotting.

To validate affinity purification results, we performed endogenous co-immunoprecipitation (co-IP) experiments in HepG2 cells, confirming a robust association between ATF7IP and PARK7 (Fig. 5B). It is reported that ATF7IP contains a C-terminal Fibronectin type-III domain and a defined 562–817 amino acid stretch responsible for SETDB1 engagement [44,45]. To illustrate the interaction details, three expression plasmids of ATF7IP truncations were constructed (Fig. 5C). Co-immunoprecipitation assays were then performed using anti-Flag in HEK-293FT cells expressing Flag (vector) or Flag-ATF7IP truncations respectively. The results demonstrated that both PARK7 and SETDB1 interact with the amino acids 521–941 of ATF7IP (Fig. 5D).

To determine the spatial feasibility of the interaction between ATF7IP and PARK7, we conducted subcellular fractionation assays to determine their localizations. The results showed that PARK7 was predominantly located in the cytoplasm, as reported previously [46], and part of ATF7IP was distributed in cytoplasm (Fig. 5E). The observed interaction between ATF7IP and PARK7 raised the question of whether ATF7IP regulates the stability of PARK7 protein. To test this hypothesis, we initially examined PARK7 mRNA and protein expression changes following ATF7IP knockdown. Our results demonstrated that, upon ATF7IP depletion, PARK7 protein levels were significantly reduced under Sorafenib treatment, while its mRNA levels remained unchanged (Fig. 5F and S3A). Then, we performed cycloheximide (CHX) pulse-chase assay to examine the effects of ATF7IP knockdown on the half-life of PARK7 protein. The results indicated that the half-life of PARK7 protein was significantly shortened after ATF7IP knockdown under Sorafenib treatment (Fig. 5G). Additionally, the decreased protein level of PARK7 in ATF7IP knockdown HepG2 cells was rescued by proteasome-specific inhibitor MG132, demonstrating that ATF7IP knockdown led to PARK7 degradation via the ubiquitin-proteasome pathway (Fig. 5H). Since PARK7 typically exists as a homodimer under physiological conditions, with the dimer containing an active site essential for its role in ferroptosis [43], we further detected whether ATF7IP affects the homodimer formation ability of PARK7 using disuccinimidyl suberate (DSS) cross-linking experiments. The results showed the presence of PARK7 dimers in both control and ATF7IP knockdown cells, indicating that ATF7IP did not interfere with PARK7 dimerization (Fig. S3B). Collectively, these findings indicate that ATF7IP interacts with PARK7 and promotes its stability in cells.

2.5. PARK7 and the transsulfuration pathway play a critical role in ATF7IP-suppressed Sorafenib-induced ferroptosis

To further elucidate whether the downregulation of PARK7 protein mediates the promotion of ferroptosis caused by ATF7IP knockdown, we overexpressed PARK7 protein in ATF7IP knockdown cells (Fig. 6A), and subsequently measured cell viability, cell death, cellular GSH, and cellular lipid ROS. The results revealed that PARK7 overexpression inhibited ATF7IP knockdown–promoted ferroptosis, as evidenced by increased cell viability, decreased percentages of cell death, promoted GSH level, and decreased cellular lipid ROS level in ATF7IP-depleted hepatocellular carcinoma cells upon Sorafenib treatment (Fig. 6B–E). In contrast, because SETDB1 does not interact with PARK7 (Fig. 5B and S4A), overexpression of PARK7 in SETDB1-depleted cells did not affect Sorafenib-induced ferroptosis (Fig. S4B–F), further indicating that ATF7IP collaborated with PARK7 independently of SETDB1. These findings demonstrate that PARK7 plays a key role in ATF7IP-suppressed Sorafenib-induced Ferroptosis.

Fig. 6.

ATF7IP suppresses Sorafenib-induced ferroptosis depending on PARK7 and the transsulfuration pathway. A. Western blotting was used to detect the protein levels of PARK7 and ATF7IP in control cells and ATF7IP knockdown HepG2 cells transfected with HA-PARK7 or control vectors. B-E. The indicated cells were treated with DMSO or Sorafenib (10 μM) for 48 h. Cell viability was monitored with MTT assay (B). Calcein-AM/PI staining was performed (C, Scale bar: 50 μm). Reduced GSH levels (D) and cellular lipid ROS (E) were detected. F–I. Control or ATF7IP knockdown HepG2 cells were treated with DMSO, Sorafenib (10 μM), or Sorafenib together with Hcy (20 μM) for 48 h. MTT assays (F), Calcein-AM/PI staining (G, Scale bar: 50 μm), reduced GSH analysis (H), and cellular lipid ROS analysis (I) were performed. For figures B–I, data are mean ± SD for n = 3; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons).

Building on previous reports that reduced PARK7 levels inhibit GSH synthesis by suppressing homocysteine (Hcy) formation, a key factor for the transsulfuration pathway [25], we further investigated the role of the transsulfuration pathway in ATF7IP-suppressed ferroptosis. After supplementing homocysteine in ATF7IP knockdown cell lines, the results of MTT and Calcein-AM/PI double-staining assays demonstrated that the reductions in cell viability and the increase in cell death caused by ATF7IP knockdown under Sorafenib treatment were reversed by homocysteine supplementation (Fig. 6F and G). Notably, exogenous homocysteine protected cells from ferroptosis triggered by ATF7IP silencing under Sorafenib treatment, as evidenced by the restoration of GSH levels and the reduction of lipid ROS levels (Fig. 6H and I). Overall, these results elucidate that PARK7 as well as the transsulfuration pathway plays a key role in ATF7IP-inhibited ferroptosis in hepatocellular carcinoma cells.

2.6. ATF7IP knockdown sensitizes hepatocellular carcinoma to Sorafenib treatment in mice

We have demonstrated that ATF7IP and SETDB1 depletion promotes the sensitivity of hepatocellular carcinoma cells to Sorafenib-induced ferroptosis in cells. A subcutaneous tumor model in nude mice was then applied to investigate the in vivo effect of ATF7IP and SETDB1 knockdown on hepatocellular carcinoma response to Sorafenib. Control, ATF7IP or SETDB1 knockdown HepG2 cells were subcutaneously injected into the right dorsal region of nude mice. Once the tumor volume reached approximately 100 mm³, mice in the control and knockdown groups were treated with either DMSO or Sorafenib (20 mg/kg/day) by oral gavage. The growth of the implanted tumors was measured. After 40 days, the mice were euthanized, and the tumors were excised, weighed, and photographed (Fig. 7A). Our results showed a significant suppression of tumor growth in mice receiving SETDB1 and ATF7IP knockdown tumors under Sorafenib treatment, as evident from the reduced tumor volume and weight in the SETDB1 and ATF7IP depleted groups (Fig. 7B–D). The knockdown of SETDB1 and ATF7IP expression in the xenograft was confirmed by immunofluorescence staining of SETDB1 and ATF7IP in tumor frozen sections (Fig. S5A and B), along with the H3K9me3 level reduced in SETDB1 and ATF7IP knockdown tumor tissues (Fig. 7E). Immunohistochemistry staining of Ki-67 in frozen sections from all six groups indicated that SETDB1 and ATF7IP knockdown inhibited hepatocellular carcinoma malignancy after Sorafenib treatment (Fig. 7F and G).

Fig. 7.

Knockdown of ATF7IP or SETDB1 sensitizes hepatocellular carcinoma to Sorafenib treatment in vivo. A. Schematic diagram of animal experiment procedure. B. Female athymic nude mice were subcutaneously transplanted with control, SETDB1 or ATF7IP knockdown HepG2 cells in the right dorsal region. When the tumor volume reached approximately 100 mm³, mice received oral gavage of vehicle (DMSO) or Sorafenib (20 mg/kg/day) for 40 days. Xenograft tumors were excised and photographed after the sacrifice of the mice. C. Tumor volumes were monitored at the indicated time. Data are mean ± SD for n = 5; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 (two-way ANOVA followed by Tukey's test for multiple comparisons). D. Final tumor weights were recorded at sacrifice. Data are mean ± SD for n = 5; ns, p > 0.05, ∗p < 0.05, ∗∗p < 0.01 (two-way ANOVA followed by Tukey's test for multiple comparisons). E. Immunofluorescence staining for H3K9me3 was performed in subcutaneous tumor tissues, and fluorescence intensity was quantitatively analyzed using ImageJ. Scale bar: 20 μm. F. Immunohistochemistry staining for Ki-67 and CYB5R2 was performed in subcutaneous tumor tissues. Scale bar: 50 μm. G-H. Quantitative evaluation of immunohistochemistry staining in figure F. Data are mean ± SD for n = 3; ns, p > 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 (two-way ANOVA followed by Tukey's test for multiple comparisons). I. Immunohistochemistry staining for PARK7 was performed in control or ATF7IP knockdown subcutaneous tumor tissues. Scale bar: 50 μm. J. Quantitative evaluation of immunohistochemistry staining in figure I. Data are mean ± SD for n = 3; ns, p > 0.05, ∗∗∗p < 0.001 (two-way ANOVA followed by Tukey's test for multiple comparisons).

Consistent with the results at the cellular level, the protein level of CYBR2 was increased in SETDB1 and ATF7IP knockdown tumor tissues (Fig. 7F and H), and the protein level of PARK7 was decreased in ATF7IP knockdown tumor tissues after Sorafenib treatment (Fig. 7I and J). Altogether, these results provide evidence that ATF7IP depletion increases CYBR2 level and decreases PARK7 stability, leading to enhanced Sorafenib sensitivity of hepatocellular carcinoma cells in vivo.

3. Discussion

Recent studies have identified ferroptosis as a promising therapeutic strategy for overcoming chemoresistance in cancer [47]. In this study, we identify ATF7IP as a suppressor of Sorafenib-induced ferroptosis in vitro and in vivo. In terms of molecular mechanism, we discover that, on the one hand, ATF7IP together with SETDB1 suppresses the transcription of CYB5R2 through catalyzing H3K9me3 on its promoter, reducing cellular Fe²⁺ accumulation; on the other hand, ATF7IP interacts and stabilizes PARK7 to activate the transsulfuration pathway to produce GSH. This discovery highlights the potential of ATF7IP as a target in Sorafenib-based treatment of hepatocellular carcinoma.

Current research on ATF7IP primarily focuses on its role as a regulator of SETDB1, which is involved in histone methylation and heterochromatin formation. SETDB1 promotes tumorigenesis by suppressing critical tumor suppressor genes in cancers [48]. Additionally, SETDB1 functions as an epigenetic checkpoint in immune priming within tumor cells by inhibiting toxic T-cell responses. This is achieved by repressing the transcription of TE-derived regulatory elements, immunostimulatory genes, and TE-encoded retroviral antigens [49]. Similarly, ATF7IP also suppresses tumor antigen expression and antitumor immunity [50]. Our findings in this study support the functional synergy between ATF7IP and SETDB1, where both proteins jointly reduce Fe²⁺ levels through transcriptional suppression of CYB5R2, thereby inhibiting Sorafenib-induced ferroptosis.

Although ATF7IP and SETDB1 are closely related, their functions do not fully overlap. Studies have shown divergent outcomes upon depletion of Setdb1 or Atf7ip in CD4⁺ T cells: Setdb1 deficiency impedes thymocyte development [51] but enhances Th1 priming [52], whereas Atf7ip deficiency in CD4⁺ T cells impairs Th17 differentiation [53]. In our study, we observed that the elevated level of Sorafenib-induced ferroptosis promoted by ATF7IP depletion is higher than that promoted by SETDB1 knockdown. Our further investigation identifies PARK7 as a specific interacting partner of ATF7IP, which helps explain the observation. PARK7, first identified in 1997 as an oncogene with RAS co-transformation activity [54], was later linked to familial early-onset Parkinson's disease in 2001, where its loss of function was recognized as a pathogenic factor [55]. PARK7 is involved in a variety of cellular processes, with its primary role in cellular antioxidant defense [56]. By undergoing oxidation at the Cys106 residue to form –SO₂^- or –SO₃^- groups, PARK7 can directly neutralize ROS and protect cells from oxidative stress-induced damage [57]. PARK7 also facilitates the transsulfuration pathway, through which the production of cysteine limits lipid ROS accumulation and consequently suppresses ferroptosis [25]. Our study demonstrates that ATF7IP interacts with PARK7 and promotes its stability in cells. PARK7 as well as the transsulfuration pathway plays a key role in ATF7IP-inhibited ferroptosis in hepatocellular carcinoma cells.

In addition to its role in Parkinson's disease, PARK7 also regulates calcium ion transport channels, participating in Alzheimer's disease [58]. Specifically, iron accumulation in the brain is strongly associated with various neurodegenerative diseases in middle-aged and elderly individuals, including stroke, Alzheimer's disease, and Parkinson's disease [59]. Localized or widespread iron accumulation can exacerbate the progression of these diseases, impairing cognitive and motor functions [60]. Ferroptosis inhibitor α-lipoic acid can mitigate Alzheimer's disease by inhibiting ferroptosis [61], while Cu^II (atsm) neutralizes lipid ROS, thereby preventing neuronal ferroptosis and providing therapeutic potential for Parkinson's disease [62]. Given the distinctive role of PARK7 in neurological disorders and the stabilization of PARK7 by ATF7IP, we hypothesize that ATF7IP may not only influence ferroptosis in hepatocellular carcinoma but also contribute to the pathogenesis of neurological diseases. Furthermore, a study suggested that ATF7IP may serve as a potential diagnostic marker for Alzheimer's disease [63]. The function of ATF7IP in the pathogenesis of neurological diseases is worth investigating in the future.

In summary, our findings enhance the understanding of the mechanisms underlying ferroptosis and provide a foundation for the development of targeted therapies against hepatocellular carcinoma, particularly those with Sorafenib resistance.

4. Materials and methods

4.1. Cells and reagents

HepG2, Hep3B, Huh7, and HEK-293FT cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 10 % FBS (Biological Industries, Beit HaEmek, ISRAEL) at 37 °C in a humidified atmosphere with 5 % CO₂. Cell lines were authenticated by examining their morphology and growth characteristics. All cells were regularly tested using the mycoplasma detection kit (D101, Vazamy, Nanjing, China). The primary antibodies against ATF7IP (A10446), SETDB1 (A6145), PARK7 (A19097), β-actin (AC026), Flag (AE061), and HA (AE008) were obtained from Abclonal Technology Co., Ltd. (Wuhan, Hubei, China). The antibodies against H3K9me3 (ab8898), H3 (ab1791), and α-Tubulin (ab7291) were obtained from Abcam Ltd. (Cambridge, MA, USA). The primary antibody against Ki-67 (D2H10) was purchased from Cell Signaling Technology Inc. (Danvers, MA, USA). The antibody against Lamin B1 (RLM3060) was purchased from ImmunoWay Biotechnology Company (Suzhou, Jiangsu, China). Horseradish peroxidase-conjugated secondary antibodies (5220-0341 and 5220-0336) were purchased from SeraCare Life Sciences Inc. (KPL, MA, USA). DL-Homocysteine (HY–W040821), Sorafenib (HY-10201), Erastin (HY-15763), RSL3 (HY-100218A), and DFO (HY-D0903) were obtained from MedChemExpress LLC. (NJ, USA). Ferrostatin-1 (T6500) was purchased from Target Molecule Corp. (Boston, USA). Suberic acid bis (N-hydroxysuccinimide ester) (S1885) was obtained from Merck KGaA Co. (Darmstadt, Germany).

4.2. Transfection and lentivirus infection

Small interfering RNA (siRNA) sequences targeting ATF7IP and SETDB1 were constructed into PLKO.1 lentiviral vector. The full‐length coding sequences of ATF7IP and PARK7 were cloned into pCDH-CMV-MCS-EF1-Puro lentiviral vector to express ATF7IP and PARK7 proteins. All primers used were listed in Supplementary Table 1. These constructs were transfected into HEK-293FT cells together with the packaging plasmids (psPAX2 and pMD2.G) using PEI (23966; Polyscience, Warrington, PA, USA) to produce lentiviruses. The supernatant of HEK-293FT was collected and filtered using a 0.45 μm pore size filter at 24 h and 48 h after transfection and applied to infect HepG2, Hep3B, or Huh7 cells. Puromycin (2 μg/mL) was added to select stable knockdown or overexpression cell lines. Small interfering RNA (siRNA) targeting CYB5R2 (CAGAGGCUUUGUGGACCUAAUTT) and negative control siRNA (UUCUCCGAACGUGUCACGUTT) were transfected into cells using Lipofectamine RNAiMAX (13778150; Thermo Fisher, Waltham, MA, USA).

4.3. Cell viability and IC50 measurement

Cell viability was typically assessed using 3‐(4, 5‐dimethylthiazol‐2‐yl)‐2, 5‐diphenyl‐2H‐tetrazolium bromide (MTT) (M8180; Solarbio, Beijing, China) assay. Briefly, cells (2000 per well) were seeded into 96-well plates and incubated for 24, 48, and 72 h. The medium was replaced with a MTT solution (0.5 mg/mL MTT in serum‐free DMEM) and incubated for 4 h at 37 °C. Then, 110 μL dimethyl sulfoxide (DMSO) was added to dissolve the formazan product for 10 min, and the absorbance was spectrophotometrically measured at 490 nm using a microplate reader. To analyze the IC50 of Sorafenib in different cell lines, cells (3000 per well) were seeded into 96-well plates and gradient doses of Sorafenib were added to the cells. After incubation for 48 h, cell viability was assessed utilizing MTT assays and IC50 was calculated.

4.4. Calcein-AM/propidium iodide (PI) double staining

About 1 × 10⁵ cells were seeded in a 6-cm dish. After treatment with DMSO or drugs for 48 h, cells were digested with 0.25 % trypsin, collected, and rinsed twice with PBS. Then the cells were stained using a Calcein-AM/PI double staining kit (C542, Dojindo, Kumamoto, Japan). The green fluorescence-labeled living cells and red fluorescence-labeled dead cells were observed simultaneously under a fluorescence microscope, and the proportions of dead and living cells were calculated.

4.5. Reduced glutathione (GSH) measurement

About 2 × 10⁶ cells were seeded in a 10-cm dish. After treatment with either DMSO or indicated drugs, cells were collected and counted. Intracellular levels of reduced GSH were detected using the Reduced GSH assay kit (BC1175; Solarbio) according to the manufacturer's instructions. The concentration of reduced GSH was determined using a standard curve and normalized to the cell number.

4.6. Lipid peroxidation measurement

Approximately 1 × 10⁵ cells were seeded in a 6-cm dish. After treatment with the indicated drugs for 48 h, the cells were digested using trypsin, collected, and rinsed twice in PBS. The cell pellet was suspended in PBS containing BODIPY 581/591 C11 (D3861; Thermo Fisher, Massachusetts, USA) at a final concentration of 5 μM. After incubation for 30 min at 37 °C in the dark, the cells were washed twice with PBS and resuspended in 500 μL PBS. Finally, lipid peroxidation levels were measured using a flow cytometer. Data were collected and analyzed using FlowJo Version 10 software.

4.7. Liquid chromatography-tandem mass spectrometry analysis

Approximately 1 × 10⁷ cells were scraped into ice-cold PBS and centrifuged at 1000 rpm for 5 min. After discarding the PBS, the cell pellet was resuspended in 400 μL 75 % methanol and 1 μL 1 mg/mL 4-HNE-d3 (332101; Ann Arbor, MI, USA). The cells were then sonicated for 10 s, followed by centrifugation at 12,000 rpm for 10 min at 4 °C. The supernatant was collected and mixed with 1 mL Methyl tert-butyl ether (MTBE). The mixture was vortexed for 10 min, and the precipitate was used for subsequent BCA quantification. After vortexing, 200 μL triple-distilled water was added to each sample tube, and the samples were incubated on ice for 10 min. The mixture was then centrifuged at 12,000 g for 10 min at 4 °C to separate the upper organic phase from the lower aqueous phase. The upper organic phase was transferred to a new tube and evaporated to dryness. 500 μL 75 % methanol solution was then added and vortexed for 5 min. Subsequently, 500 μL 2,4-dinitrophenylhydrazine (2 mg/ml, prepared in 60 % ethanol) was added to each tube, and the samples were derivatized in an oven at 60 °C for 2 h. After derivatization, the liquid was transferred to a 15 mL centrifuge tube, and 1 mL ethyl acetate and 700 μL triple-distilled water were added. The mixture was vortexed for 5 min, followed by centrifugation at 4000 g for 10 min at 4 °C to collect the upper organic phase. The extraction was repeated with an additional 1 mL ethyl acetate and 700 μL triple-distilled water. The products from two extractions were combined in the same tube, nitrogen evaporated, and reconstituted in 100 μL ethanol. After vigorous mixing, samples were filtered using a 0.22 μm centrifuge tube before they were introduced into the mass spectrometer. The details of the subsequent mass spectrometry analysis have been described previously [64]. The mass spectrometry data were divided by the BCA quantification values and normalized by the control group treated with DMSO.

4.8. Intracellular Fe²⁺ measurement

Cells were seeded at a density of 2000 cells per well in 96-well plates and cultured overnight. After 48 h of treatment with indicated drugs or DMSO, cells were rinsed twice with serum-free medium and incubated with serum-free medium containing 1 μM FerroOrange (F374; Dojindo) at 37 °C for 30 min. Fluorescence images were captured using a fluorescence microscope. Fe²⁺ levels were quantified using ImageJ software.

4.9. RNA isolation and quantitative real-time RT-PCR (RT-qPCR)

Total RNA was extracted from cells using TRIzol (P118; GenStar, Beijing, China), and reverse-transcribed into cDNA using StarScript III All-in-one RT Mix with gDNA Remover (A230-10; GeneStar). RT-qPCR was performed using RealStar Power SYBR qPCR Mix (A311; GenStar). The relative mRNA expression was calculated using the 2^−ΔΔCt method. The primers used were listed in Supplementary Table 1.

4.10. RNA-seq and data analyses

HepG2 cells stably expressing sh-control, shATF7IP or shSETDB1 were treated with 10 μM Sorafenib for 48 h and lysed using TRIzol reagent. The mRNA was extracted and subjected to RNA-seq by BGI Genomics Co., Ltd. (Beijing, China). Following high-throughput sequencing, reads containing adapters, poly-N, and low-quality sequences were removed. Trimmed reads were mapped to the human reference genome (UCSC hg38) using Hisat2 (v2.2.1). Samtools (v1.6) was used to sort and set the index of mapped results and then transfer them from sam to bam format. Gene counts and transcript counts were calculated using StringTie (v2.2.1) and the PrepDE.py script. Differentially expressed genes (DEGs) were identified using DESeq2 (R package v1.34.0).

4.11. Chromatin-immunoprecipitation (ChIP)

Cells were cross-linked with 15 mL 1 % formaldehyde solution for 10 min at room temperature and quenched with 1 mL 2 M glycine (final concentration: 0.125 M) for 5 min. After rinsed twice with PBS, cells were collected in SDS buffer (100 mM NaCl, 50 mM Tris-HCl, pH 8.1, 5 mM EDTA, and 10 % SDS) containing protease inhibitors (B14002; Bimake, Houston, Texas, USA) and centrifugated for 6 min at 1200 rpm. Cell pellets were resuspended in ice-cold IP buffer (100 mM NaCl, 66.67 mM Tris-HCl, pH 8.0, 5 mM EDTA, 0.33 % SDS, and 1.67 % Triton X-100) and sonicated with Bioruptor (Diagenode, Denville, NJ, USA). After centrifugation at 13,000 rpm for 15 min at 4 °C, primary antibodies were added to the supernatant and incubated overnight at 4 °C. Protein A/G magnetic beads (B23201; Bimake) were added and incubated for another 2 h at 4 °C. Beads were rinsed three times with washing buffer Ⅰ (1 % Triton X-100, 0.1 % SDS, 150 mM NaCl, 2 mM EDTA, and 20 mM Tris-HCl, pH 8.0), and once with washing buffer Ⅱ (1 % Triton X-100, 0.1 % SDS, 500 mM NaCl, 2 mM EDTA, and 20 mM Tris-HCl, pH 8.0), followed by reverse crosslinking in de-crosslinking buffer (1 % SDS, 0.1 M NaHCO₃) for 4 h at 65 °C. DNA was extracted and subjected to qPCR. The ChIP-qPCR primers used were listed in Supplementary Table 1.

4.12. Flag affinity purification

HEK-293FT cells expressing empty vector or Flag-ATF7IP were lysed in cell lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 0.5 % Triton X-100) supplemented with protease inhibitors. After centrifugation at 12,000 rpm at 4 °C for 20 min, the supernatant was incubated with anti-Flag M2 gel (A2220; Sigma-Aldrich, St. Louis, MO, USA) at 4 °C overnight. After thorough washes, the bound proteins were eluted using Flag peptide and subjected to SDS-PAGE which was then subjected to silver staining using a PAGE Gel Silver Staining Kit (G7210; Solarbio). Unique protein bands identified by silver staining in the Flag-ATF7IP overexpression group were excised from the gel and analyzed by liquid chromatography-mass spectrometry (LC-MS).

4.13. Co-immunoprecipitation

About 4 × 10⁶ cells were washed twice with ice-cold PBS and harvested in cell lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 0.5 % NP-40) supplemented with protease inhibitors at 4 °C for 30 min. After centrifugation at 12,000 rpm for 10 min, the supernatant was incubated overnight at 4 °C with specific primary antibodies or normal IgG. Protein A/G magnetic beads were incubated with the protein–antibody complex at 4 °C for another 2 h. After washing three times with washing buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 0.1 % NP-40), the proteins were eluted by denaturation with 1 × loading buffer at 95 °C for 30 min. The supernatant was then subjected to Western blotting.

4.14. Subcellular fractionation

Approximately 1 × 10⁶ cells were resuspended in 100 μL buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 0.15 % NP-40, 1 mM dithiothreitol, and protease inhibitor cocktail) on ice for 10 min, and nuclei were collected by centrifugation (1 min, 12,000 rpm, 4 °C). The supernatant was further clarified by centrifugation (1 min, 12,000 rpm, 4 °C) and collected as the cytosolic fraction. The nuclei were further washed once in 150 μL buffer B (20 mM HEPES, pH 7.9, 0.4 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5 % NP-40, 1 mM dithiothreitol, and protease inhibitor cocktail). The nuclear fraction was then isolated by centrifugation (30 min, 12,000 rpm, 4 °C), and collected in the precipitate. The cytosolic and nuclear fractions were subsequently analyzed by Western blotting.

4.15. DSS-mediated cross-linking assays

After washing twice with ice-cold PBS (pH 8.0), approximately 5 × 10⁴ cells were detached and incubated at room temperature for 10 min in a 2 mM DSS crosslinking buffer (with no reducing agents added). The reaction was then quenched with 1 M Tris-HCl buffer (pH 7.4, final concentration: 2 mM) for 10 min at room temperature. Then, 5 × SDS-PAGE loading buffer containing β-mercaptoethanol was added. The samples were boiled at 95 °C for 30 min and subsequently subjected to SDS-PAGE and Western blotting.

4.16. Animal experiments

1 × 10⁶ HepG2 cells stably expressing ATF7IP shRNAs, SETDB1 shRNAs, or control shRNAs were suspended in 50 μL of PBS, mixed with 50 μL of Matrigel, and subcutaneously injected into female athymic nude mice (BALB/c, Charles River Laboratories, Wilmington, MA; 5–6 weeks old; 5 mice per group). Once the tumor volume reached approximately 100 mm³, the mice were administered either vehicle (DMSO) or Sorafenib (20 mg/kg/day) by oral gavage. Tumor volume was measured using a Vernier caliper every 5 days and calculated using formula V = π/6 × length × width². The mice were euthanized 40 days after injection. The tumors were then removed and photographed. All animal handling and procedures were approved by the Institutional Animal Care and Use Committees of Tianjin Medical University.

4.17. Immunohistochemistry analysis

Briefly, xenograft tumors retrieved from mice were embedded in Optimal Cutting Temperature (OCT) Compound (#4583; SAKURA Finetek, Torrance, CA, USA), followed by rapid freezing at −40 °C. Subsequently, the frozen tissues were sectioned into consecutive 8 μm slices. Prior to immunostaining, the tissue sections were treated with 3 % hydrogen peroxide for 10 min in darkness to quench endogenous peroxidase activity, and non-specific signals were blocked with PBST (PBS with 0.1 % Triton X-100) containing 10 % goat serum for an hour. The slices were incubated with primary antibodies overnight at 4 °C, followed by incubation with secondary antibodies conjugated to horseradish peroxidase (HRP) at room temperature for another hour. Color development was achieved using a diaminobenzidine (DAB) substrate kit (ZLI-9017; ZSGB-BIO, Beijing, China), and the slides were counterstained with hematoxylin (G1120; Solarbio). Images were captured using a microscope.

4.18. Statistical analysis

All experimental data are presented as mean ± SD or mean ± SEM from at least three independent experiments. Statistical analyses were performed using GraphPad Prism (version 8.0). All data were subjected to normal distribution testing (Shapiro-Wilk test). For two individual groups, unpaired Student's t-test was applied, and one-way ANOVA followed by Dunnett's test was used between multiple groups. For the experiments with drug treatment, two-way ANOVA followed by Tukey's test for multiple comparisons was used to calculate the p-value. A p-value of < 0.05 was considered statistically significant.

CRediT authorship contribution statement

Yijie Su: Writing – original draft, Visualization, Validation, Investigation, Formal analysis. Sirui Huang: Investigation. Yang Duan: Investigation, Funding acquisition. Liang Zhang: Investigation. Shengyun Feng: Investigation. Yingge Lv: Investigation. Bei Lan: Validation, Data curation. Chenghao Xuan: Writing – review & editing, Writing – original draft, Supervision, Project administration, Funding acquisition, Conceptualization.

Declaration of competing interest

Authors declare that they have no competing interests.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

The raw and processed high-throughput sequencing data (RNA-seq) were deposited in the Gene Expression Omnibus (GEO) database under accession number GSE294091.