Nrf2 drives hepatocellular carcinoma progression through acetyl-CoA-mediated metabolic and epigenetic regulatory networks

Abstract

Correlations between the oxidative stress response and metabolic reprogramming have been observed during malignant tumor formation; however, the detailed mechanism remains elusive. The transcription factor Nrf2, a master regulator of the oxidative stress response, mediates metabolic reprogramming in multiple cancers. In a mouse model of hepatocellular carcinoma (HCC), through metabolic profiling, genome-wide gene expression, and chromatin structure analyses, we present new evidence showing that in addition to altering antioxidative stress response signaling, Nrf2 ablation impairs multiple metabolic pathways to reduce the generation of acetyl-CoA and suppress histone acetylation in tumors, but not in tumor-adjacent normal tissue. Nrf2 ablation and dysregulated histone acetylation impair transcription complex assembly on downstream target antioxidant and metabolic regulatory genes for expression regulation. Mechanistic studies indicate that the regulatory function of Nrf2 is low glucose dependent, the effect of which is demolished under energy refeeding. Together, our results implicate an unexpected effect of Nrf2 on acetyl-CoA generation, in addition to its classic antioxidative stress response regulatory activity, integrates metabolic and epigenetic programs to drive HCC progression.

Implications:

This study highlights that Nrf2 integrates metabolic and epigenetic regulatory networks to dictate tumor progression and that Nrf2 targeting is therapeutically exploitable in HCC treatment.

Introduction

Metabolic reprogramming in cancer cells is crucial for increased macromolecular biosynthesis, cell growth, proliferation, and stress response and is now considered a hallmark of cancer cells (1). Mounting evidence has revealed that metabolic pathways, including glycolysis, glycogenolysis, tricarboxylic acid (TCA) cycle, pentose phosphate pathway (PPP), and fatty acid metabolism are abnormally activated to support tumorigenesis. Specifically, metabolic intermediates such as fumarate, succinate, and alpha-ketoglutarate (αKG) are important cofactors for αKG-dependent dioxygenases in mediating epigenetic histone methylation, while acetyl-CoA, the connector of glycolysis, the TCA cycle, and fatty acid metabolism, has been demonstrated to modulate histone acetylation (2-4). Therefore, aberrant metabolic reprogramming not only alters cellular energy generation, stress response, and cellular proliferation but also actively participates in gene regulation (5, 6). However, sufficient evidence is still required to demonstrate the coupling between metabolic and epigenetic regulatory programs in driving tumor development.

Hepatocellular carcinoma (HCC) is the most common type of primary liver cancer and the third leading cause of cancer-related death worldwide. Besides chronic infections by hepatitis B and C viruses, many risk factors, including alcohol use and nonalcoholic fatty liver disease (NAFLD), are well defined in HCC (7). Notably, NAFLD, a hepatic manifestation of metabolic syndrome, is a rapidly growing medical condition, and obesity is becoming an epidemic worldwide. Recently, the integration of multi-omics data, including genomics, transcriptomics, proteomics, and metabolomics, has promoted the identification of factors associated with disease incidence, prognosis, and potential drug targets (8, 9). NFE2L2 (NRF2) and its interactor KEAP1 are significantly mutated in 5% and 8% of HCC, respectively, and are considered to be tumor drivers (10). Additionally, aberrant NRF2 activation by p62/SQSTM1 induces and maintains tumor initiation and exacerbates HCC malignancy (11, 12). Importantly, NRF2 is a master transcriptional regulator of the cytoprotective response against electrophiles and reactive oxygen species (ROS). During tumor development, increased metabolic output and changing environmental conditions result in high oxidative stress, whereas NRF2 activation detoxifies ROS and promotes tumorigenesis (13). Furthermore, NRF2 has been demonstrated to redirect glucose and glutamine into anabolic pathways to mediate metabolic reprogramming, such as PPP for the synthesis of NADPH, a reducing equivalent necessary for the reduction of glutathione, and redox cycling (11, 14). NRF2 also participates in the regulation of amino acid metabolism to provide cancer cells with nitrogen for the synthesis of nucleotides and nonessential amino acids (13, 15). Notably, during the NRF2-mediated metabolic reprogramming, aberrant levels of metabolites such as fumarate, succinate, and αKG are produced to affect tumor cell growth (11, 13, 14). Recently, we also showed that ablation of the NRF2 downstream target NQO1 triggered simultaneous inhibition of the PI3K/Akt and MAPK/ERK pathways, suppressed the expression of glycolysis and glutaminolysis genes, and blocked metabolic adaptation in liver cancer (16). However, the effect of NRF2 on energy generation and production of these metabolites remains largely unknown.

Here, we investigated the causal function of Nrf2 in oxidative stress, metabolic energy generation, and epigenetic modifications in HCC. We found that Nrf2 ablation reduced acetyl-CoA availability by impairing glycolysis, glycogenolysis, and fatty acid metabolism. This reduction of acetyl-CoA limits both TCA cycle efficacy for metabolic energy production and epigenetic histone acetylation for gene expression. In vitro Nrf2 suppression demonstrated that the impact of Nrf2 on acetyl-CoA availability is dependent on low glucose content. Therefore, our findings highlight the role of Nrf2 in mediating an integrated metabolic and epigenetic program to promote HCC.

Materials and Methods

Cell culture.

The HCC human cell lines HepG2 (HB-8065), Hep3B (HB-8064), HuH7 (PTA-4583), and SNU449 (CRL-2234) obtained from ATCC and mouse Hepa1-6 (DT81 subline) (17) are maintained in Dulbecco’s modified Eagle’s medium (DMEM) (10-013-CV, Corning) supplemented with 10% fetal bovine serum (SH30396, HyClone) and 2 mM glutamine. The cells were tested negative for mycoplasma contamination in a PCR-based analysis using the following primers: forward, GTGGGGAGCAAA(C/T)AGGATTAGA, and reverse, GGCATGATGATTTGACGTC(A/G)T. HCC cells were reinitiated from cryopreserved stocks every 3 to 6 months.

HCC cells were cultured depending on the experimental design. For hypoxia condition, the cells were cultured in a hypoxic chamber (Coy Laboratory Products, Grass Lake, MI) with 1% O₂. For nutrient-poor and refeeding culture conditions, cells were cultured with glucose free DMEM (D5030, Sigma), 10% dialyzed FBS (SH30079, HyClone) supplemented with the indicated concentrations of glucose (G5146, Sigma), sodium pyruvate (SH30239, Hyclone), sodium acetate (S8625, Sigma), 2-deoxy glucose (2DG) (D8375, Sigma), respectively.

Mice.

Nrf2 gene knockout mice (Nrf2^−/−) were obtained (18). A single dose of the hepatic carcinogen diethylnitrosamine (DEN) (25 mg/kg body weight) was intraperitoneally injected into 14-day-old male mice to initiate tumor formation. At 36 weeks post-DEN treatment, mice were sacrificed with body weight, liver weight, numbers of tumors per liver and maximal tumor size determined. The animal handling and experimental procedures were approved by the Institutional Animal Care and Use Committee of Augusta University.

Mouse xenografts.

Control shRNA (shCtl) and shNRF2 HuH7 cells (5 x 10⁶ cells per mouse) were injected subcutaneously into the right dorsal flanks of 5–6 weeks old immunodeficient NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) mice (Jackson Laboratory, Bar Harbor, ME), 5 mice each cell line. Tumors were measured by luminescence imaging (IVIS Spectrum, PerkinElmer, USA) 18 days after xenograft.

Human samples.

Formalin-fixed, paraffin-embedded (FFPE) HCC tissue samples were obtained and de-identified from the Georgia Cancer Center Biorepository at Augusta University (Augusta, GA); snap-frozen normal and HCC tissues were from the Cooperative Human Tissue Network (Charlottesville, VA); and HCC tissue microarrays were from US Biolab (Rockville, MD). Written informed consent was obtained from all patients by Georgia Cancer Center Biorepository with the protocol approved by the IRB Committee of Augusta University.

Gene expression analyses.

Total RNA from mouse liver tumors was extracted with PureLink^™ RNA Mini Kit (12183020, Thermo Fisher Scientific), subjected to library preparation using Illumina TrueSeq RNA kit V2 and sequenced on HiSeq2500 platform (50 bp pair-end reads). The trimmed reads were aligned to the mouse genome mm10 using STAR v2.6.1 aligner and an expression count matrix was generated from the aligned reads using HTSeq. Normalization and differential expression analysis were performed on the raw count matrix using DESeq2 v2_1.6.3 in the R/Bioconductor package. Gene Set Enrichment Analysis (GSEA) was performed using GSEA V4.0.3. Differentially expressed genes using DESeq2 v2_1.6.3 with q-value < 0.05 were considered to be significantly differentially expressed and uploaded to Ingenuity Pathway Analysis v.49309495 (IPA, QIAGEN) to be analyzed using the “Core Pathway Analysis”.

ChIP-seq and ChIP-PCR.

Liver tumors were collected for the ChIP-seq experiment using ChIP-IT High Sensitivity Kit (ab185913, Active Motif) with adaptations. Chromatins from two different liver tumors for each genotype of mice were sonicated to 200-1000 bp and immunoprecipitated with ChIP-grade AcH3K27 antibody (Millipore). The ChIP DNA library was prepared using TruSeq ChIP Sample Preparation kit (Illumina) and sequenced at 75bp single reads with an average depth of 35 million reads on a NextSeq 500 Illumina machine. ChIP sequencing reads were mapped to the reference mouse genome (mm9) with Bowtie2 and analyzed by MACS2 for the peak calling.

ChIP-PCR was performed with antibodies against TBP (SC-273x, Santa Cruz Biotechnology), CBP (SC-369x, Santa Cruz Biotechnology), RNA polymerase II (05-623, Sigma) with gene-specific primers (Supplemental Table S1). Normal rabbit IgG (I8410, Sigma) was used as a non-specific antibody control. qPCR was performed on immunoprecipitated DNA using SYBR Green Supermix (172-5272, Biorad).

Gas chromatography–mass spectrometry (GC-MS) analysis.

Approximately 50 mg of snap-frozen mouse liver tumors was determined by GC-MS analysis as previously reported (19). Metabolites were extracted with cold 90% methanol and an internal standard d4-succinic acid (Sigma) was added for every 25 mg of tissue. The samples were homogenized, and supernatant was collected by centrifugation and another internal standard, d27-myristic acid, was added. All samples were then dried en vacuo and resuspended in pyridine containing 40 mg/ml O-methoxylamine hydrochloride (MOX) before GC-MS analysis on an Agilent Zorbax DB-5MS column. Metabolites were identified and their peak area was recorded using MassHunter Quant and analyzed with MassHunter software (Agilent). Statistical analysis was performed using MetaboAnalystR.

Immunoblotting analysis.

Whole-cell extract was prepared as previously described (20). For histone extract preparation, tumor tissues or cultured cells were homogenized in Triton extraction buffer (TEB) (PBS containing 0.5% Triton X-100, 2 mM phenylmethylsulfonyl fluoride) for nuclei isolation and nuclear extracts were prepared by 0.2 N HCl acid extraction followed by neutralization with 1/10 volume of 2 M NaOH. Protein concentrations of the whole-cell and histone extracts were determined using the Bradford assay.

Primary antibodies:

β-actin (A5316), histone H3 (06-755), AcH3 (06-599), AcH3K4 (07-539), AcH3K9 (06-942), HDAC1 (05-100) and ACC1 (04-322) were from Sigma; NRF2 (ab62352), GCLC (ab41463), NQO1 (ab80588) and AcH3K27 (ab4729) were from Abcam; ME1 (A3956), ACAT1 (A5335), ACO2 (A3716), ACOX1 (A8091), ALDOB (A3728), CD36 (A5792), DLD (A5403), GLUT2/SLC2A2 (A12307), GPI (A6916), HADH (A1076), PDHA (A13687), PKLR (A12084) and SUCLG2 (A8976) were from Abclonal; HDAC2 (SC-7899), HDAC3 (SC-11417), Sirt1 (SC-15404) and CBP (SC-369X) were from Santa Cruz Biotechnology; acetyl lysine (9441) and FASN (3180) were from Cell Signaling Technology. Secondary horseradish peroxidase–conjugated goat anti-rabbit or goat anti-mouse antibodies (Santa Cruz Biotechnology) were used.

Acetyl-CoA and CoASH assays.

Mouse liver tumor or HepG2 cells were homogenized, deproteinized and determined for the acetyl-CoA and CoASH contents with acetyl Coenzyme A Assay Kit (MAK039, Sigma) and Coenzyme A Assay Kit (ECOA-100, BioAssay) according to the manufacturers’ instructions. Nuclear levels of acetyl-CoA and CoASH were similarly determined immediately after nuclei isolation. The levels of acetyl-CoA and CoASH were normalized to protein content.

HAT/HDAC activity assay.

Nuclei were isolated from liver tumors or HepG2 cells as reported previously (21) and resuspended in HAT and SIRT1 Assay buffer provided in the HAT Inhibitor Screening Assay and SIRT1 Direct Fluorescent Screening Assay Kits (10006515/10010401, Cayman Chemical). HAT and HDAC activities were determined according to the manufacturer’s instructions.

To determine the effect of the supplement of acetyl-CoA and CoASH on HAT activity, liver tumor nuclei were treated with the indicated concentrations of acetyl-CoA (A2056, Sigma) and CoASH (C4282, Sigma) for 1 h at 37°C. After treatment, nuclei were pelleted by centrifugation with histone acetylation determined by immunoblotting analysis.

Statistical analysis.

Statistical tests were justified for all figures. All the samples represent biological replicates. Data are presented as the mean ± SD. Student’s t-test and Pearson’s correlation coefficient (Pearson’s r) were used as appropriate, with the corresponding two-tailed significance (p value) determined. Statistical analysis was performed using the GraphPad Prism 9 software environment (https://www.graphpad.com/). Values of p < 0.05 were considered statistically significant (*p < 0.05, **p < 0.01, and ***p < 0.001).

Data and material availability:

All data needed to evaluate the conclusions in the paper are presented in the paper and/or Supplementary Materials. The RNA-seq and ChIP-seq data were deposited in the Gene Expression Omnibus database (GSE183712 and GSE227068).

Additional methods and materials are provided in the Supplementary Information section.

Results

NRF2 promotes hepatocarcinogenesis

Recent findings suggest that NRF2, a master regulator of oxidative stress response, shunts cancer development through antioxidative response regulation and metabolic reprogramming (14). To explore the NRF2 function in HCC in more detail, we determined its expression in resected HCC tissues from human patients. Compared to the adjacent non-tumor tissue, significantly increased expression of NRF2 was detected in tumor sections (Fig. 1A). The levels of the antioxidant protein NQO1 and the critical metabolic enzyme ME1 were also elevated in tumors compared to peritumoral areas (Fig. 1A). Notably, antioxidant proteins and metabolic enzymes have been previously demonstrated to be highly expressed and correlated with advanced clinical stage and poor HCC prognosis (11, 22, 23).

Fig. 1. NRF2 promotes hepatocarcinogenesis.

(A) Immunoblots of NRF2, NQO1, and ME1 in the whole cell extracts of liver tumor (T) and adjacent normal tissues (NT) in human patient HCC samples. β-Actin is a loading control. (B) Serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels in 36-week post DEN-treated male wild type (Nrf2^+/+) and Nrf2 knockout (Nrf2^−/−) mice. Data represent mean ± SD (n=6 mice per group). (C) Representative images of livers from Nrf2^+/+ and Nrf2^−/− mice at an advanced stage of disease. Quantification of liver tumor gross weight, numbers, multiplicity, and maximal sizes are presented. Data represent mean ± SD (n=36-38 mice per group). (D) Representative staining of H&E, Nrf2, Sirius Red and Ki67 in Nrf2^+/+ and Nrf2^−/− mice liver tumors. Insert showing the percentages of Sirius Red area and Ki67-positive cells in the liver tumors. Data represent mean ± SD (n=6 mice per group). * p < 0.05; ** p < 0.01; ***p < 0.001. Scale bar, 50 μm.

To investigate the specific role of NRF2 in HCC, 14-day-old Nrf2 knockout (Nrf2^−/−) (18) and wild-type (Nrf2^+/+) male mice were induced liver tumor with an injection of the hepatic carcinogen diethylnitrosamine (DEN), which incorporates chronic liver injury, inflammation, and DNA damage, thus mimicking the key pathological and histological features of human HCC. At 36-week post-DEN treatment, Nrf2^−/− mice displayed lower serum alanine aminotransferase (ALT) activity, whereas aspartate transaminase (AST) activity was similar to that in Nrf2^+/+ mice (Fig. 1B), suggesting that Nrf2^−/− mice suffered milder hepatic dysfunction than Nrf2^+/+ mice did. Indeed, tumor incidence and size were significantly reduced in Nrf2^−/− mice compared with those in Nrf2^+/+ mice (Fig. C). Histological analyses of the liver tumors revealed significantly reduced levels of fibrosis and proliferation after Nrf2 ablation (Fig. 1D. Together, these results suggest that NRF2 plays an important role in the development of aggressive liver tumors in both humans and mice.

Nrf2 regulates hepatic energy generation through key metabolic processes

Since the liver is the main organ in energy generation, we investigated whether Nrf2 ablation inhibits liver tumor cell growth by regulating energy generation. To assess this, we determined ADP and ATP levels in the liver tumors. Compared to Nrf2^+/+ mice, Nrf2^−/− mice had lower levels of ATP, but comparable levels of ADP (Fig. 2A), indicating that Nrf2 ablation leads to a shortage of energy supply in liver tumors.

Fig. 2. Nrf2 ablation impairs multiple metabolic pathways of hepatic energy generation.

(A) Levels of high energy nucleotides ATP and ADP in liver tumors of Nrf2^+/+ and Nrf2^−/− mice (n=10 mice each genotype) (*p < 0.05). (B) Top selected GSEA results of Nrf2^+/+ and Nrf2^−/− liver tumors 36-week post-DEN treatment as determined by RNA-seq analysis (n=3 mice). (C) GSEA results showing that ROS, glycolysis, fatty acid metabolism and oxidative phosphorylation related pathways in liver tumors were significantly impaired in Nrf2^−/− mice, compared to that of Nrf2^+/+ mice at 36-week post-DEN treatment (n=3 mice). (D) Representative immunoblot analyses of proteins involved in antioxidative, energy generation pathways for oxidative stress response, glycolysis, fatty acid metabolism, and the TCA cycle in whole tissue extracts of Nrf2^+/+ and Nrf2^−/− mouse liver tumors. Transcription factor TATA-box binding protein (Tbp) and β-Actin are loading controls. Data represent mean ± SD (n=6 mice) (*p < 0.05). (E-F) Nrf2^−/− mice have lower glucose levels in peripheral blood and liver tumors (E) but have higher glycogen levels (F) in liver tumor tissues than Nrf2^+/+ mice. Data represent mean ± SD (n=12 mice) (*p < 0.05). (G) Representative Oil Red O staining of Nrf2^+/+ and Nrf2^−/− mouse liver tumors. Broken red lines depict the tumor (T) border. Nuclei were counterstained with Hematoxylin QS. Insert showing the percentages of Oil Red O staining area (n=6 mice). Scale bar, 50 μm. (H-J) Nrf2 ablation reduced the levels of triglyceride (H) and the levels of intermetabolites isocitrate, fumarate and malate in the TCA cycle (I), and NADH, NAD+ and NADH/NAD+ ratio (J) in liver tumors of Nrf2^+/+ and Nrf2^−/− mice. The levels of the TCA cycle intermetabolites were determined by GC-MS analysis. Data represent mean ± SD (n=6 mice each genotype). *p < 0.05; ** p < 0.01; ***p < 0.001. Scale bar, 50 μm.

To identify the genes affected by the oxidative stress response and energy generation, we performed whole-genome RNA sequencing analysis of Nrf2^+/+ and Nrf2^−/− mouse liver tumors at 36-week post-DEN treatment. As expected, the transcription levels of genes whose products are involved in oxidative stress response were among those most affected (Fig. 2B-D and Supplementary Fig. S1A), suggesting a role of Nrf2 in the regulation of ROS response. Interestingly, gene signatures corresponding to glycolysis, fatty acid metabolism, oxidative phosphorylation, and amino acid metabolism were among the highest-ranking regulated pathways (Fig. 2B-C). The gene and protein expression levels for those participating in glucose transport, glycolysis, and glycogenolysis were decreased after Nrf2 ablation (Fig. 2D and Supplementary Fig. S1B). In alignment, the glucose levels in both blood and liver tumors were significantly reduced (Fig. 2E) while the levels of glycogen were increased in Nrf2^−/− mouse liver tumors, as compared to that in Nrf2^+/+ mice (Fig. 2F).

Significant differences in the expression of genes whose products are involved in fatty acid metabolism (fatty acid uptake, accumulation, transport, and β-oxidation) were detected between Nrf2^−/− and Nrf2^+/+ mouse liver tumors (Fig. 2C-D and Supplementary Fig. S1C). Histological examination of liver tumors showed that Nrf2 ablation diminished lipid levels in liver tumors (Fig. 2G). Consistently, we detected a significant reduction in triglyceride levels in Nrf2^−/− mouse liver tumors (Fig. 2H). Additionally, although the expression of most TCA cycle enzymes was at comparable levels in Nrf2^−/− and Nrf2^+/+ mouse liver tumors, reduced expression of mitochondrial aconitase (Aco2), dihydrolipoamide dehydrogenase (Dld), and succinyl-CoA ligase (Suclg2) was observed after Nrf2 ablation (Fig. 2C-D and Supplementary Fig. S1D). Dld mediates the irreversible conversion of αKG to succinyl-CoA, thus its decreased expression suggests that the efficacy of the TCA cycle may be reduced. To assess this, we measured TCA cycle intermetabolites by GC-MS, which showed that isocitrate, fumarate, and malate levels were decreased after Nrf2 ablation (Fig. 2I). In addition, compared to Nrf2^+/+ mice, Nrf2^−/− mice showed reduced levels of NADH, while NAD+ levels were similar, which resulted in a reduced NADH/NAD+ ratio (Fig. 2J). Thus, these data indicate that Nrf2 regulates glycolysis, glycogenolysis, fatty acid metabolism, and the TCA cycle in energy generation for liver tumor growth.

NRF2 expression correlates with histone acetylation in liver tumors

We next determined the mechanism by which NRF2 regulates the signaling pathways involved in energy generation. In paradigm NRF2, as a transcriptional activator, binds to the antioxidant response element (ARE) for gene regulation (24). Interestingly, recent evidence supports that aberrant levels of acetyl-CoA as cofactors contribute to epigenetic histone acetylation in solid tumors (21, 25, 26). Given that Nrf2 ablation impairs energy generation and acetyl-CoA connects the metabolic pathways of glycolysis, the TCA cycle, and fatty acid metabolism, we hypothesized that Nrf2 ablation could cause aberrant levels of acetyl-CoA and thus affect histone acetylation. To test this possibility, we determined the levels of acetylated histone H3 (AcH3), acetylated histone H3 lysine 9 (AcH3K9), and AcH3K27 in liver tumors after Nrf2 ablation. Compared to Nrf2^+/+ mice, Nrf2^−/− mice had significantly reduced levels of AcH3, AcH3K9, and AcH3K27 (Fig. 3A-B). We then tested whether NRF2 affects histone acetylation in tumor-adjacent normal liver tissues. Surprisingly, no differences in histone acetylation were observed between the Nrf2^−/− and Nrf2^+/+ mice (Supplemental Fig. S2A). Thus, Nrf2 ablation affects histone acetylation in liver tumors but not in adjacent non-tumor liver tissues.

Fig. 3. NRF2 expression correlates with histone acetylation through acetyl-CoA regulation in liver tumors.

(A) Immunoblotting determined the histone lysine acetylation of AcH3, AcH3K9, and AcH3K27 in the liver tumor nuclear extracts of Nrf2^+/+ and Nrf2^−/− mice. Quantitation represents the ratio of acetylated lysine signals to loading control histone H3. Data represent mean ± SD (n=3 mice each genotype). (B) Representative images and quantification of AcH3K27 levels detected by immunohistochemistry in paired liver tumors (T) and adjacent non-tumor normal tissues (NT) of Nrf2^+/+ and Nrf2^−/− mice. Broken red lines depict the tumor (T) border. Scale bar, 200 μm. (C) Representative images of NRF2 and AcH3K27 expression detected by immunohistochemistry in human normal liver tissue and patient liver tumors. Nuclei were counterstained with Hematoxylin QS. Ten random fields of each tissue section were selected for semi-quantitative scoring in a blinded manner (detailed in Supplemental methods) and given a combined score for the percentage of positive cells and intensity of staining for both NRF2 and AcH3K27. The Pearson’s correlation coefficient (Pearson’s r) between NRF2 and AcH3K27 is determined for a total of 81 tumor samples from patients with HCC. (r = 0.5239, p < 0.0001) (n=81). Scale bar, 50 μm. (D) NRF2 and AcH3K27 staining intensity in (C) and their correlation with tumor grade in HCC tumor microarrays. The staining was performed in monoplicate. Expression levels of NRF2 and AcH3K27 in normal liver tissue were shown as controls (n = 81). *p <0.05, statistical significance in NRF2 or AcH3K27 levels over range of tumor grades (one-way ANOVA test). Scale bar, 200 μm.

Notably, significant correlations have been reported between histone acetylation and cancer prognosis (27-29). To explore the clinical relevance of NRF2 and histone acetylation, we assessed their abundance in a panel of liver tumor patient samples. Each tumor sample was scored in a blinded manner and assigned a combined score for the percentage of positive cells and intensity of staining for both NRF2 and AcH3K27 with the expression of NRF2 detected in both cytoplasm and nuclear. A statistically significant positive correlation was observed between NRF2, AcH3K27 (Fig. 3C) and tumor grade (Fig. 3D).

NRF2 regulates acetyl-CoA availability in HCC

Histone acetylation is affected by the expression of histone acetyltransferases (HATs) and deacetylases (HDACs), which are frequently deregulated in malignant conditions (30). To determine whether Nrf2 ablation affects the expression of these histone modifiers, RNA-seq data were re-analyzed. No differences in the expression of known HATs or HDACs were detected between Nrf2^+/+ and Nrf2^−/− liver tumors (Supplementary Fig. S3A-B). Protein analyses of the selected HAT and HDACs also did not detect their different expression in Nrf2^−/− and Nrf2^+/+ mouse liver tumors (Supplementary Fig. S3C), suggesting that histone acetylation affected by Nrf2 ablation is irrelevant to the expression change of HATs or HDACs.

Importantly, histone acetylation can be regulated by the levels of acetyl-CoA and NAD+, which are substrates for HATs and sirtuins, respectively. To determine whether a dysregulated level of acetyl-CoA or NAD+ contributes to histone acetylation changes after Nrf2 ablation, we assayed HAT and SIRT1 activities in mouse liver tumors. Although there were no differences in SIRT1 activity between the nuclear extracts of Nrf2^+/+ and Nrf2^−/− mouse liver tumors, we detected reduced HAT activity in the liver tumor extracts of Nrf2^−/− mice compared to that in Nrf2^+/+ mice (Fig. 4A-B). Thus, the histone acetylation changes after Nrf2 ablation are less likely to occur through sirtuin/NAD+-mediated deacetylation, but due to reduced HAT activity.

Fig. 4. NRF2 regulates histone acetylation through acetyl-CoA.

(A-B) Relative activities of HDACs (A) and HATs (B) in the nuclear extracts of Nrf2^+/+ and Nrf2^−/− mouse liver tumors at 36-week of age. Data represent mean ± SD (n=6 mice each genotype). (C) Levels of acetyl-CoA, CoASH and ratio of acetyl-CoA:CoASH in liver tumors of Nrf2^+/+ and Nrf2^−/− mice at 36-week post-DEN treatment of age. Data represent mean ± SD (n=6 mice each genotype). (D) Representative immunoblot of histone AcH3 and AcH3K27 levels upon incubation of isolated nuclei of Nrf2^+/+ and Nrf2^−/− mouse liver tumors with defined concentrations of acetyl-CoA. Experiments were performed in duplicate in tumors from three different mice of each genotype. Quantitation represents the ratio of acetylated histone to total histone H3 in three independent experiments with no acetyl-CoA addition in Nrf2^+/+ nuclei set at 1. Data represent mean ± SD (n=3 Nrf2^+/+ mice, n=4 Nrf2^−/− mice). *p < 0.05.

Since there were no differences in the expression of HATs in Nrf2^+/+ and Nrf2^−/− mouse liver tumors, the reduced HAT activity may originate from a short supply of the HAT substrate acetyl-CoA. To test this possibility, we measured acetyl-CoA and reduced CoA (CoASH) levels. The level of acetyl-CoA in Nrf2^+/+ mouse liver tumors was 2-fold of that in Nrf2^−/− mice while CoASH was at a comparable level, resulting in a decreased ratio of acetyl-CoA: CoASH in Nrf2^−/− liver tumors (Fig. 4C). We also determined the levels of acetyl-CoA in tumor-adjacent normal liver tissues. In agreement with the absence of any effect on histone acetylation (Supplementary Fig.S2A), we did not observe differences in the levels of acetyl-CoA or CoASH between them (Supplementary Fig. S2B).

Whether the reduction of acetyl-CoA in Nrf2^−/− mouse liver tumors affects histone acetylation was assessed by treating mouse liver tumor nuclei with defined concentrations of acetyl-CoA. The addition of acetyl-CoA increased histone acetylation levels in Nrf2^+/+ mouse liver tumor nuclei (Fig. 4D). Notably, at low concentrations of acetyl-CoA supplementation up to 5 μM, the level of histone acetylation in Nrf2^−/− liver tumor nuclei extract was significantly lower than that in Nrf2^+/+ mouse liver tumors; however, histone acetylation was increased and reached comparable levels to that of Nrf2^+/+ mouse liver tumor nuclei when supplemented with high concentrations of acetyl-CoA (Fig. 4D). Together, these data indicate that Nrf2 ablation reduces histone acetylation by restricting the availability of acetyl-CoA and that acetyl-CoA supplementation can rescue histone acetylation.

Nrf2 ablation regulates target gene expression through histone acetylation

The significance of the reduction in histone acetylation after Nrf2 ablation in affecting chromatin structure in mouse liver tumors was assessed by ChIP-seq. We focused on AcH3K27, an active enhancer marker that is enriched in the transcriptional start site (TSS) for gene upregulation (31). AcH3K27 has been previously demonstrated to be correlated with poor HCC differentiation and prognosis (29), and our data support this correlation (Fig. 3D). Notably, AcH3K27 is an important target for the treatment of HCC, especially in chronic hepatitis C virus infections(32, 33). Analysis of the enrichment of AcH3K27 around promoters, defined as TSS ± 3kb, revealed an overall decrease in AcH3K27 levels after Nrf2 ablation (Fig. 5A). Among the representative genes involved in the oxidative stress response (Nqo1 and Cat), glycolysis (Slc2a1, Gck, and Eno1), TCA cycle (Aco2 and Suclg2), and fatty acid biosynthesis (Fasn and Scd1), we detected significantly reduced association of AcH3K27 in Nrf2^−/− mouse liver tumors (Fig. 5B-E); however in housekeeping β-actin and TATA-box binding protein gene loci, no changes on AcH3K27 association were detected (Fig. 5F).

Fig. 5. Nrf2 ablation affected the formation of transcription complexes on metabolic related genes.

(A) ChIP-seq analysis of the enrichment of AcH3K27 around gene promoters (± 3kb) in liver tumors of Nrf2^+/+ and Nrf2^−/− mice 36 weeks after DEN injection. (B-F) ChIP-seq data tracks for AcH3K27 enrichment at the representative antioxidant (Nqo1 and Cat), glucose transport and glycolysis (Slc2a1, Gck and Eno1), the TCA cycle (Suclg2), and fatty acid metabolism (Fasn and Scd1), or housekeeping β-actin (Actb) and TATA-box binding protein (Tbp) gene loci in the liver tumors of Nrf2^+/+ and Nrf2^−/− mice at 36-week of age. Blue arrows indicate the transcription start site (TSS) and gene orientation of each gene locus. The data are representative of biological replicate samples. (G-J) ChIP-PCR analysis of the association of AcH3K27 (G), Tbp (H), Pol II (I) and Cbp (J) to the promoter regions of antioxidant (Nqo1 and Cat), glycolysis (Gck, Gpi1, Aldoa, Eno1 and Aco1), the TCA cycle (Suclg2 and Aco2), pyruvate dehydrogenase complex (Pdha1), fatty acid metabolism (Cd36 and Fasn), and housekeeping actin and Tbp loci in the liver tumors of Nrf2^+/+ and Nrf2^−/− mice after normalization to the input. Rabbit normal IgG was used as an antibody control (K). (L) The relative mRNA transcript levels of indicated genes were determined by qRT-PCR in the liver tumors of Nrf2^+/+ and Nrf2^−/− mice with their levels in Nrf2^+/+ mice set at 1 after normalization to β-actin. Data represent mean ± SD of three biological replicates (n=6 mice for each genotype). *, p < 0.05.

To further investigate whether the global reduction in histone acetylation affects NRF2 downstream target gene loci chromatin structure, we determined the assembly of transcription complexes for the association of AcH3K27, general transcription factor TATA-binding protein (Tbp) and RNA polymerase II (Pol II) after Nrf2 ablation by ChIP-PCR. Consistent with the reduced AcH3K27 association in Nrf2 target genes, significant decreases in the association of Tbp, and Pol II with the antioxidant (Nqo1, Cat, Gstt1, Gclc, and Gclm), glycolysis, fatty acid metabolism, and oxidative phosphorylation-related genes (Aldoa, Gck, Eno1, Pgk, Me1, Aco2, Suclg2, Pcx, Gpi, Pdha1, and Cd36) but not in housekeeping actin and Tbp gene loci, were observed in liver tumors of Nrf2^−/− mice, compared to those of Nrf2^+/+ mice (Fig. 5G-K). In alignment with the reduced transcription complex assembly, we also detected their decreased expression in liver tumors of Nrf2^−/− mice (Fig. 5L). Together, these observations suggest that the repressive chromatin structures after Nrf2 ablation and histone deacetylation at these gene loci reduce gene expression.

To determine the mechanism of Nrf2 ablation on histone acetylation on those gene loci specifically, we investigated the protein-protein interaction between Nrf2 and CREB-binding protein (CBP), a highly homologous histone acetyltransferase and detected their interaction in DT81 (17), a Hepa1-6 derived mouse HCC cell line (Supplementary Fig. S4), in agreement with previous report (34). To further determine whether the binding of CBP is affected in those gene loci with different histone acetylation after Nrf2 ablation, we performed ChIP analysis. As expected, the binding of CBP was significantly impaired in Nrf2^−/− than in Nrf2^+/+ mouse liver tumors (Fig 5J), suggesting an Nrf2-dependent association of CBP to gene loci for histone acetylation modification. Together, these data suggest that Nrf2, by recruiting histone acetyltransferase CBP, can modify gene loci specific histone acetylation.

NRF2-mediated histone acetylation is of low glucose dependency

To further investigate the mechanism by which NRF2 regulates histone acetylation through acetyl-CoA, we generated two different NRF2-silenced HepG2 cell lines (shNRF2-1 and shNRF2-2), in which shNRF2 reduced colony formation and cellular proliferation compared to shRNA controls (shCtl) (Fig. 6A). The expression of NRF2 typical downstream antioxidant proteins NQO1 and GCLC was reduced after NRF2 silencing (Fig. 6B); however, an unexpected observation was that shNRF2 did not affect the expression of ACO2 or SUCLG2 (Fig. 6B), which is inconsistent with the aforementioned in vivo findings in mouse liver tumors (Fig. 3D). In addition, shNRF2 also did not affect the levels of histone acetylation or acetyl-CoA (Supplementary Fig. S5).

Fig. 6. NRF2-mediated histone acetylation through acetyl-CoA is of low glucose dependency.

(A) The clonogenic assays for shRNA control (shCtl) and two different shNRF2 (shNRF2-1 and shNRF2-2) for HCC cells (HepG2, HuH7, Hep3B, and SNU449). The same amount of cells were grown for 12 (HuH7, Hep3B and SNU449) or 21 (HepG2) days, fixed, and stained with 0.5% crystal violet. The relative colony formation for shCtl (purple), shNRF2-1 (red), and shNRF2-2 (blue) for each HCC cell line is indicated by the circle size on the right. (B) Representative immunoblot of NRF2 and downstream target proteins (NQO1, ACO2, SUCLG2) in shCtl and shNRF2 HCC cell lines. β-actin was used as a loading control. (C) Glucose concentration in liver tumors and adjacent normal liver tissues. (D) Cellular acetyl-CoA levels in shNRF2 HCC cells cultured with 0.1mM or 1mM glucose. (E) Representative immunoblot of histone AcH3, AcH3K9 and AcH3K27 in nuclear extracts of different HCC cells cultured under 0.1mM or 1mM glucose. (F) Quantitative RT-PCR showing the relative mRNA levels of classic NRF2 downstream target NQO1 and metabolism genes ACO2 and SUCLG2 in low glucose cultured HCC cells. (G) ChIP-PCR analysis of the association of AcH3K27 to the NQO1, ACO2, and SUCLG2 gene loci in HepG2 shCtl and shNRF2 cells after normalization to the input. Rabbit normal IgG (right) was used as an antibody control. (H) NRF2 silencing dramatically suppressed the tumor growth of HuH7 xenografts in NSG mice. (I) Acetyl-CoA, CoASH, and the ratio of Acetyl CoA:CoASH in shCtl and shNRF2 HuH7 xenografts. (J) Immunoblot of histone AcH3, AcH3K9, and AcH3K27 in nuclear extracts of shCtl and shNRF2 HuH7 xenografts. Data represent mean ± SD (n=3-6). *p < 0.05; ** p < 0.01.

To exclude the cellular context effect of shNRF2 on HepG2 cells, we used additional HCC cell lines (HuH7, Hep3B, and SNU449). Similar observations were made that shNRF2 did not affect the expression of ACO2 and SUCLG2 (Fig. 6A-B). Thus, there is an unappreciated correlation between NRF2 expression, acetyl-CoA generation, histone acetylation, and gene regulation as detected in vivo in mouse liver tumors, but not in vitro in HCC cell lines under the tested conditions.

To investigate the discrepancy in NRF2 silencing on acetyl-CoA and histone acetylation in vivo versus in vitro, we considered the dependency of the cellular environment on NRF2 function. Recent studies have shown that hypoxia is critical for histone methylation and acetylation in tumor cells (35, 36). To assess whether hypoxia is involved in NRF2-mediated histone acetylation, we cultured HepG2 cells with 1% O₂ and 20% O₂. However, no differences in histone acetylation were detected between normoxic and hypoxic treatments after NRF2 silencing (Supplementary Fig. S6). Thus, NRF2-mediated histone acetylation is unlikely relevant to oxygen availability.

Importantly, histone acetylation is correlated with glucose concentration, and cells cultured at high glucose concentrations have high levels of histone acetylation (37). Notably, the glucose concentration in mouse liver tumors is approximately one-tenth of that in non-tumor livers (Fig. 6C). In addition, Nrf2^−/− mice showed significantly reduced glucose levels in both blood and liver tumors (Fig. 2E); therefore, we investigated whether glucose availability is involved in NRF2-mediated acetyl-CoA production and histone acetylation by treating HCC cell lines with different concentrations of glucose. Interestingly, shNRF2 significantly decreased the levels of acetyl-CoA when cultured in 0.1 mM glucose, but not at higher concentrations (Fig. 6D). Similarly, shNRF2 did not affect AcH3 or AcH3K27 levels at higher glucose concentrations (1 mM); however, at low glucose concentration (0.1 mM), it reduced AcH3 and AcH3K27 levels (Fig. 6E).

Whether shNRF2 under low-glucose culture conditions affects gene expression, as detected in vivo in mouse liver tumors, was determined for ACO2 and SUCLG2. In contrast to no effect on the expression of ACO2 and SUCLG2 when cultured in high glucose conditions, shNRF2 reduced their expression and the classic NRF2 downstream target NQO1 when cells were grown under low glucose conditions (Fig. 6F). In agreement with the reduced histone acetylation and gene expression, we further observed that shNRF2 reduced the association of histone acetylation (AcH3K27) in the ACO2 and SUCLG2 gene loci (Fig. 6G).

To further demonstrate a glucose dependent effect of NRF in affecting liver tumor cell growth, we cultured shCtl and shNRF2 HepG2 cells in various concentrations of glucose (0.1 mM and 1 mM), supplemented with dialyzed fetal bovine serum. Notably, low glucose culture and shNRF2 significantly reduced cell growth but barely affected cellular viability, compared to controls (Supplementary Fig. S7A-B). Concomitantly, low glucose was found to promote NRF2 expression in a dose- and time-dependent manner (Supplementary Fig. S7C-D), as well as its nuclear translocation (Supplementary Fig. S7E). These observations support a glucose concentration dependent effect on the expression of NRF2 in regulating liver tumor cells growth.

To confirm that shNRF2 in HCC cell lines decreases acetyl-CoA production and histone acetylation in vivo, we engrafted shNRF2 HuH7 cells into NSG mice. Tumor expansion of shNRF2 HuH7 xenografts was significantly inhibited (Fig. 6H). In agreement with the observations that Nrf2 ablation in vivo and silencing in vitro in low-glucose culture conditions reduced histone acetylation (Figs. 3 and 6E), shNRF2 in HuH7 xenografts showed significantly reduced levels of acetyl-CoA and histone acetylation (Fig. 6I-J).

We next determined the effects of Nrf2 suppression on histone acetylation in vivo. Mice at 32-week post-DEN treatment were chronically treated with Brusatol or ML385, with the former selectively enhances ubiquitination and degradation of Nrf2 (38) and the later interrupts Nrf2 expression and its DNA binding activity (39). After 4 weeks of treatment, we determined the serum ALT and AST activities. There is a higher ratio of AST/ALT with Brusatol and ML385 treatments, compared to vehicle treatment (Supplementary Fig. S8A), suggesting that both treatments presented protective effects against liver damage.

Treatment of Brusatol and ML385 significantly decreased the expression of Nrf2 downstream targets antioxidant protein and metabolic enzyme genes (Supplementary Fig. S8B), supporting their Nrf2 suppressing effects. In agreement with the afore finding that Nrf2 ablation decreased acetyl CoA levels, Brusatol and ML385 significantly reduced liver tumor acetyl CoA levels compared to vehicle treatment (Supplementary Fig. S8C). Subsequent immunoblotting analyses showed that both AcH3 and AcH3K27 levels were reduced after Brusatol and ML385 treatment, compared to vehicle control (Supplementary Fig. S8D). Together, these data indicate that Nrf2 suppression reduces the levels of acetyl CoA and histone acetylation in both human and mouse models of HCC, and this regulation is of low glucose dependency.

Energy refeeding ablates NRF2-mediated histone acetylation

To demonstrate that NRF2 function in histone acetylation is sensitive to glucose concentration, we determined whether NRF2 affects histone acetylation after glucose refeeding. High concentrations (1 mM) of glucose refeeding were found to rescue histone acetylation and acetyl-CoA generation in both shCtl and shNRF2 cells; however, when glucose was refed at a low concentration (0.1 mM), neither histone acetylation nor acetyl-CoA generation was fully rescued in shNRF2 cells (Fig. 7A-B).

Fig. 7. Energy refeeding ablates NRF2-mediated histone acetylation.

(A) Immunoblot (top) and quantification (bottom) of histone AcH3 and AcH3K27 levels in glucose-deprived shNRF2 HepG2 cells after refed with different concentrations of glucose (0.1 mM, 1mM or 10 mM) for 24 hours. (B) Cellular levels of acetyl-CoA, CoASH, and the ratio of acetyl-CoA: CoASH in shNRF2 HepG2 cells treated in (A). (C) Immunoblot of histone AcH3 and AcH3K27 levels in the glucose-deprived shNRF2 HepG2 cells after refeeding with low, medium, or high levels of pyruvate (1mM, 5mM or 20 mM), acetate (0.1 mM, 1 mM or 2.5 mM), 2DG (0.1 mM, 1 mM or 2.5 mM) or Glucose+2DG (1 mM glucose plus 0.1 mM, 1 mM or 2.5 mM 2DG) for 24 hours. Histone levels in shCtl and shNRF2 cells before treatment were included as controls. (D) Cellular levels of acetyl-CoA, CoASH, and the ratio of acetyl-CoA: CoASH in shNRF2 HepG2 cells treated in (C). Data represent mean ± SD (n=3 independent experiments). *p < 0.05.

To further determine whether energy generation is essential for histone acetylation mediated by NRF2, shCtl and shNRF2 HepG2 cells were treated with metabolites or their analogs that contributed to acetyl-CoA production, including pyruvate, acetate, and 2-deoxy glucose (2DG). Supplementation with pyruvate and acetate was found to rescue histone acetylation (Fig. 7C-D) and acetyl-CoA generation (Fig. 7E). In contrast, 2DG did not rescue histone acetylation or acetyl-CoA generation; conversely, it suppressed the glucose-induced rescue of histone acetylation and acetyl-CoA production (Fig. 7C-E).

Next, we determined whether the affected histone acetylation after energy refeeding contributes to gene regulation. Both pyruvate and acetate treatments significantly increased the expression of glycolysis and fatty acid metabolism genes, which were downregulated by shNRF2. In contrast, 2DG treatment did not induce gene expression; instead, it suppressed the induction effect of glucose (Supplementary Fig. S9). Thus, NRF2 regulates histone acetylation through acetyl-CoA and energy refeeding abolishes such regulatory effects, supporting a glucose concentration-dependent manner of NRF2 in chromatin modification and gene regulation.

Discussion

Emerging evidence shows that metabolism and histone modifications are inextricably linked. In this study, we report that NRF2 integrates metabolic and epigenetic regulatory programs to promote HCC in line with its classic antioxidative stress response activity. Our metabolic profiling analyses showed that Nrf2 ablation strongly suppressed glycolysis and fatty acid metabolism, resulting in a low fuel of acetyl-CoA with less effective TCA cycle activity for energy generation and epigenetic histone acetylation to inhibit tumor growth.

Here, we found that acetyl-CoA levels were significantly reduced in the absence of Nrf2 in HCC. Nrf2 and its partner Keap1 have been demonstrated to regulate glucose metabolism, acetyl CoA production and subsequently protein acetylation in a variety of stressed situations including both physiological and malignant conditions (40-43). Nrf2 has also been shown to recruit histone acetylation modifiers such as CBP and the mediator complex to affect histone acetylation in target gene loci for expression regulation (44, 45). Notably, acetyl-CoA could modulate histone and non-histone protein acetylation to induce tumor cell growth (37, 46, 47). These findings inspired our hypothesis that downregulation of acetyl-CoA in the absence of Nrf2 affects histone acetylation. Consistent with these findings, Nrf2 ablation reduced acetyl-CoA levels and histone acetylation. However, when validating the findings in vitro in HCC cells, NRF2 silencing did not affect histone acetylation. Meanwhile, in non-tumor mouse liver tissue, we could not detect any change in histone acetylation after Nrf2 ablation, as in liver tumors, implicating the elusive function of Nrf2. In fact, other than acetyl-CoA, metabolic intermediates serve as cofactors and substrates for epigenetic “writers and erasers” to influence epigenetics. The TCA cycle intermediates αKG, succinate, and fumarate influence DNA and histone methylation by regulating the activity of αKG-dependent dioxygenases (48). Other metabolites, such as D-2-hydroxyglutarate and L-2-hydroxyglutarate are competitors for αKG in regulating the activity of αKG-dependent dioxygenases for gene regulation(49).

Certainly, the cellular environment of solid tumors is quite heterogeneous and characterized by some levels of hypoxia and nutrient-poor conditions. Hypoxia modifies regulatory programs, such as activating hypoxia-inducible factors and inducing changes in chromatin for histone methylation and acetylation to turn on gene expression for cancer development (35, 36); whereas deficiency of nutrients such as glutamine and glucose alters metabolic and epigenetic regulatory pathways to drive tumor development (50, 51). However, commercial nutrient-rich media were formulated for in vitro cell cultures to maximize cell proliferation and are distinct from in vivo conditions. Considering the striking differences in oxygen and nutrient availability that could affect metabolic activities and dependencies, we cultured HepG2 cells under hypoxic and nutrient-poor conditions with a low supply of glucose. Although we did not detect whether hypoxia contributed to histone acetylation when NRF2 was silenced, we mimicked the effect of NRF2 suppression on reducing histone acetylation under low-glucose culture conditions. Previously, glucose starvation was found to reduce intracellular acetyl-CoA, decrease amounts of bulk H3 acetylation, redistribute H3K9ac occupancy and thus compromise cell proliferation (52). Akt-dependent metabolic reprogramming was demonstrated to dynamically regulate glucose availability for acetyl-CoA generation which affects histone acetylation in glioblastoma tumor cell (21). Here, we demonstrate that NRF2 can modulate histone acetylation for gene regulation under nutrient-poor conditions, and that NRF2 ablation reduces the ability of tumor cells to adapt to varying nutrient challenges. Our and those previous findings supported the availability of glucose in tumor cellular environment dependent manner plays an important role in histone acetylation modification to mediate gene regulation.

Notably, NRF2 activation promotes the proliferation of hepatoma and other types of tumor cells by regulating glutathione synthesis (11, 13, 53). Hepatocyte-specific NRF2 activation due to p62 accumulation or KEAP1 inhibition leads to hepatomegaly associated with enhanced glycogenosis, steatosis, cell cycle arrest, and hyperplasia via upregulation of AKT signaling, rather than being dependent on the effect of NRF2 through cytoprotective antioxidant signaling or metabolic regulatory mechanisms (54). In addition, NRF2 promotes the self‐renewal of cancer stem cells (12), activates cancer metastatic programs (55, 56), antagonizes tumor-promoting inflammation (54, 57), and upregulates genomic damage repair to facilitate tumor proliferation and survival (58, 59). Therefore, NRF2 mediates multiple signals, including the ROS stress response, and metabolic and epigenetic regulatory pathways in the regulation of HCC. In this regard, NRF2 affects all hallmarks of cancer (60), and the intervention of NRF2-regulated pathways is a pivotal strategy in HCC management.