Nrf2 Mutagenic Activation Drives Hepatocarcinogenesis

Hepatocellular carcinoma (HCC), which comprises approximately 90% of cases of liver malignancy, is the third leading cause of cancer-related mortality worldwide (1) and is associated with diverse etiologic factors, such as hepatitis B and C viral infection, chronic alcohol intake, and exposure to a variety of dietary and environmental chemical carcinogens (2). Recent progress in genomic analyses allows identification of several critical driver mutations and activating signaling pathways involved in HCC development (3, 4). Somatic mutations in nuclear factor-erythroid 2-related factor 2 (Nrf2, also known as Nfe2l2) were also noticed through whole-genome sequencing of HCC patients (1, 3).

Nrf2 was first characterized as a transcriptional activator that binds to a tandem repeat sequence named NF-E2 sequence (5′-TGCTGAG/CTCAT/C-3′; ref. 5). Subsequently, Itoh and colleagues recognized the similarity between the NF-E2 binding sequence and the antioxidant response element sequence (ARE, 5′-GCNNNG/CTCA-3′; N: A/T/C/G), which is identified within the regulatory regions of many antioxidant and phase II carcinogen–detoxifying enzymes (6). Hence, it has been widely accepted that activating Nrf2 protects cells from the DNA-damaging effects of reactive oxygen species (ROS) and electrophilic metabolites of carcinogens. Moreover, use of Nrf2 activators has been considered a beneficial therapeutic approach in the management of disorders associated with the accumulation of oxidative stress, such as diabetic complications (7), cardiovascular diseases (8), neurodegenerative disorders (9), and cancer (10). Nevertheless, it is now speculated that Nrf2 signaling can also be hijacked by some cancer cells as an adaptive mechanism that restricts ROS and electrophilic burdens within the tumor microenvironment, allowing the cancer cell survival (11). Alternatively, it has also been suggested that Nrf2 may direct reprogramming of intermediary metabolism to support cancer cell growth (12).

A 605-amino acid protein Nrf2 is composed of seven highly conserved domains, named Nrf2-erythroid cell-derived protein with Cap ‘n’ Collar homology (ECH) domains, Neh 1-7 (13). Among them, two units in Neh2 drive interaction of an Nrf2 molecule with two molecules of Klech-like ECH-associated protein 1 (Keap1), a substrate adaptor of the Cul3-based E3 ligase machinery: the high-affinity ETGE (DxETGE) motif and the weaker binding DLG (LxxQDxDLG) motif to regulate Nrf2 stability. Once bound to Keap1, Nrf2 is ubiquitylated and subsequently degraded by proteasomes (13). Somatic inactivating gene mutations in Keap1 (14) or epigenetic mechanisms, such as hypermethylation of the Keap1 promoter suppressing Keap1 expression (15), might prevent Nrf2 degradation, leading to an enhanced accumulation of Nrf2 in the nucleus. In addition, somatic mutations of Nrf2 in the region responsible for the Keap1 binding also underlie Nrf2 activation (16). Recently, a growing body of evidence has revealed that Keap1/Nrf2 signaling is commonly impaired in human HCC, which is attributable to genetic alterations in either Keap1 (17) or Nrf2 (3, 4). These mutations were all found to be of somatic origin and account for the accumulation and subsequent elevated transcriptional activity of Nrf2. In addition, somatic mutations in Nrf2 were noticed as an early molecular event in an experimental model of HCC in rodents (18). However, the biological significance of these changes in the context of Nrf2 involvement in hepatocellular carcinogenesis remained overlooked. Here, by utilizing Nrf2 knockout (KO) mice, we demonstrate that Nrf2 is a critical driving force for diethylnitrosamine (DEN)-induced hepatocellular carcinogenesis.

Wild-type (WT) and Nrf2 KO mice on a C57BL6/129SV mixed background generated by the laboratory of Yuet Wai Kan as described previously were used in this study (19). Transgenic (Tg) mice carrying the core ARE sequence coupled to the human placental alkaline phosphatase (hPAP) gene (ARE-hPAP mice) were generously gifted by Dr. Jeffrey A. Johnson (University of Wisconsin-Madison, Madison, WI; ref. 20). In brief, a core sequence ARE carried by 51-bp fragment of rat NAD(P)H:quinone oxidoreductase 1 (Nqo1) was inserted into a TATA-Inr minimal promoter of the heat-stable hPAP reporter construct. This construct was thereafter injected into fertilized oocytes that were then implanted into uteri of surrogate mothers. Resulting pups were identified by genotyping. Mice were maintained in a 12-hour dark–light cycle with the light cycle occurring from 7:00 am to 7:00 pm and given ad libitum access to food and water. All animal experiments were approved by the Seoul National University Animal Care and Use committee (approval number: SNU 20140624-2).

For genotyping, clipped mouse toes were incubated with a genotyping buffer containing 1% SDS, 0.1 mol/L NaCl, 100 mmol/L EDTA, 50 mmol/L Tris (pH 8.0), and 1 μg/μL proteinase K at 55°C overnight. On the next day, 5 mmol/L NaCl was added to remove cell debris. DNA was precipitated by incubating in cold 2-propanol at −20°C for 2 hours, and then washed with cold 70 % ethanol. Six-hundred nanograms of DNA was used as a template for the PCR following a standard procedure. PCR primer sequences employed are presented in Supplementary Table S1. Genomic Nrf2 (gNrf2) was also used as an internal control in genotyping ARE-hPAP mice. PCR products were separated on 2% agarose gel and visualized under Imagequant LAS 4000 after stained with SYBR.

At day 15 of age, male mice of each genotype (WT, Nrf2 KO, and ARE-hPAP mice) were given a single dose of 25 mg/kg DEN prepared in PBS (pH 7.4) by intraperitoneal injection, following the protocol described previously (21). Tumor burden was evaluated at indicated times during the course of experiments. Visible nodules (≥0.5 mm diameter) were counted (21).

Fifteen-day-old WT and Nrf2 KO mice were given a dose of 25 mg/kg DEN prepared in PBS (pH 7.4) by intraperitoneal injection. Livers were isolated at indicated times and subjected to subsequent experiments.

Mice were perfused with PBS (pH 7.4). Livers were isolated and fixed in 4% paraformaldehyde (PFA)/PBS overnight. After paraffin embedding and serial sectioning (5 μm), hematoxylin and eosin (H&E) staining was performed.

Paraffin-embedded tissue sections were deparaffinized by soaking consecutively in xylene followed by a series of gradient ethanol (100 % × 2 times, 90 %, 80 %, and 70 %) and fixed again in 4% PFA/PBS (pH 7.4), then incubated in PBS at 70°C. Slides were thereafter washed in AP buffer (100 mmol/L Tris-HCl pH 9.5, 100 mmol/L NaCl, 50 mmol/L MgCl₂, 0.01% sodiumdeoxycholate, and 0.02% Nonidet-P40), and treated with 0.34 mg/mL nitroblue tetrazolium salt (NBT; Sigma) and 0.17 mg/mL 5-bromo-4-chloro-3-indolyl phosphate, toluidinium salt (BCIP; Sigma) prepared in AP buffer for 18 hours at 37°C. Afterwards, slides were mounted with coverslips, and observed under a light microscope (Leica DM5000B; Leica Microsystems, Inc.).

At the indicated times, mice were perfused with PBS (pH 7.4), and livers were then isolated and fixed in 4% PFA/PBS overnight. After dehydration through a series of gradient sucrose (10% and 30% sucrose in PBS), the livers were immersed in optimal cutting temperature (OCT) embedding compound (Tissue-Tek) and subsequently sectioned at 5-μm thickness on a microtome (HM525; Thermo Scientific). Cryosections were washed briefly with PBS (pH 7.4) and then the nonspecific binding was blocked with a buffer containing 0.3% Triton X-100 and 5% (v/v) donkey serum (Sigma) in PBS before being incubated with primary antibodies prepared in the blocking buffer at 4°C in a humidified chamber overnight. On the next day, the slides were treated with Alexa 488 and 546 secondary antibodies (Molecular Probes, Invitrogen), followed by nuclear counterstaining with 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes, Invitrogen). Afterwards, images were obtained under a confocal microscopy (LSM 700; Carl Zeiss).

Fifteen-day-old WT and Nrf2 KO mice were given an intraperitoneal dose of DEN (25 mg/kg) prepared in PBS. A single intraperitoneal dose of 50 mg/kg bromodeoxyuridine (BrdUrd; Sigma) was administered to DEN-injected mice 2 hours before mice were sacrificed. Livers were isolated at 48 hours after DEN injection. Paraffin-embedded liver sections were deparaffinized and subsequently denatured in a mild hydrochloric acid solution before stained with an antibody specific for BrdUrd. Signals were developed using the ABC staining kit and 3,3′-diaminobenzidine (DAB; Vector Laboratories). The sections were counterstained with hematoxylin, mounted with coverslips, and visualized under a light microscope (Leica DM5000B; Leica Microsystems, Inc.).

Antibodies used in this study and their sources were GAPDH (sc-365062; Santa Cruz Biotechnology; 1:2,000), α-Tubulin (MU-121UC; Biogenex; 1:2,000), β-actin (A2066; Sigma; 1:2,000), Lamin B (sc-6216; Santa Cruz Biotechnology; 1:2,000), Nrf2 (sc-722; Santa Cruz Biotechnology; 1:1,000 for Western blotting and 1:200 for immunostaining), Nqo1 (Ab28947; Abcam; 1:2,000), cyclin D1 (#2926; Cell Signaling Technology; 1:2,000), proliferating cell nuclear antigen (PCNA; #2586; Cell Signaling Technology; 1:500), phosphogluconate dehydrogenase (Pgd; sc-398977; Santa Cruz Biotechnology; 1:2,000), malic enzyme (Me) 1 (sc-100569; Santa Cruz Biotechnology; 1:2,000), cytochrome P450 2E1 (Cyp2e1; HPA009128; Sigma; 1:1,000 for Western blotting and 1:200 for immunostaining), and BrdUrd (ab1893; Abcam; 1:200).

Liver tissue was lysed in a lysis buffer [50 mmol/L Tris-Cl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA (pH 8.0), 1% NP-40, 0.25% deoxycholate, protease cocktail tablets]. Cytosolic and nuclear fractions used in some experiments were prepared following the method described previously (22). Protein concentrations were measured using Pierce BCA Protein Assay Kit (Thermo Scientific). The lysates were then analyzed to examine the expression levels of indicated proteins by immunoblotting in accordance with the procedure described elsewhere (22).

Total RNA was isolated from liver tissue using TRIzol (Invitrogen). RNA was then used to synthesize complementary DNA (cDNA) and further analyzed by using RealHelix qPCR kit (Nanohelix) with Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems) following the protocol described elsewhere (22). Gapdh was included as an internal control. The primers used in quantitative (q) RT-PCR assay are listed in Supplementary Table S2.

HEK293T cells were obtained from ATCC in 2010. The cells were cultured in DMEM (Genedepot) supplemented with 10 % v/v FBS and 100 U/mL penicillin and 100 μg/mL streptomycin (Gibco, Invitrogen) in a humidified atmosphere of 95% O₂ and 5% CO₂ in an incubator at 37°C and used within ten passages. Mycoplasma test was routinely performed using a BioMycoX Mycoplasma PCR Detection Kit (CellSafe) to ensure that cells were negative for Mycoplasma contamination. As HEK293T cells are solely used as an expression system and not for cancer-related biological studies, they were not authenticated. The cells were transfected with indicated plasmids using transfection reagent polyethylenimine (PEI; Sigma) following manufacturer's instructions. Twenty-four hours after transfection, the cells were harvested for subsequent assays.

Total RNA was extracted from vehicle- or DEN-treated liver tissue and converted to cDNA using a reverse transcriptase following the standard procedure. The cDNA region harboring exon 2 of murine Nrf2 gene was amplified using PCR with HiPi Taq polymerase (Elpisbio). The PCR products were purified by QIAquick PCR Purification Kit (Qiagen), and sequenced by ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems). Primers for sequencing are listed in Supplementary Table S3. Mutations are defined and given in accordance to the numbering of amino acid in murine Nrf2 protein sequence (NCBI CCDS Database; gene number: NC_000068.7).

Plasmid of pcDNA3.1-EGFP-C4-human NRF2 was obtained from Addgene (plasmid #21549, generated by Dr. Yue Xiong's laboratory). For the murine Nrf2 expression vector, plasmid of murine Nrf2 (AAV-MCS8-Nrf2) was purchased from Addgene (plasmid #67636, generated by Dr. Connie Cepko's laboratory). Various constructs of Nrf2 bearing point mutations were generated using Muta-Direct Site Directed Mutagenesis Kit (iNtRON). A stop codon ATG was introduced to the murine Nrf2 construct to halt the expression of Nrf2. Primer sequences used appear in Supplementary Table S4. All plasmid clones were verified by DNA sequencing.

The statistical significance of differences between two groups was evaluated based on two-tailed Student t test. Statistical significance was accepted at P < 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ^#, P < 0.0001. Data are presented as mean ± SEM. Data analyses were performed using GraphPad Prism (Version 6).

To investigate the role of Nrf2 in liver tumorigenesis, we compared the development of HCC in WT and Nrf2 KO mice. Their genotype was verified by PCR (Supplementary Fig. S1A and S1B). Male mice of each genotype were given a single intraperitoneal injection of DEN at day 15 after birth following the protocol described previously (21). After 9 months, we noticed that all the WT mice treated with DEN developed hepatomas, whereas no tumors were visible in the livers of Nrf2 KO mice with the same treatment (Fig. 1A). We subsequently examined the liver sections of DEN-treated WT and Nrf2 KO mice by H&E staining. WT mice showed multiple lesions with normal hepatocytes interspaced among the tumor cells, while the livers from Nrf2 KO mice showed a conserved normal hexagonal structure (Fig. 1B). We further confirmed this result in another set of mouse livers collected at 14 months after the DEN injection, the duration reported to cause an advanced stage of HCC in the DEN-induced murine HCC model (23). Again, the Nrf2 KO mice had no visible tumors in livers even at 14 months after the DEN injection, while all of the WT mice showed severe tumor burden (Fig. 1C). H&E staining of DEN-treated liver sections revealed hepatic lesions in the WT mice, but a normal hexagonal liver structure in the Nrf2 KO mice without tumors (Fig. 1D). Collectively, these results indicate that deletion of Nrf2 confers resistance to DEN-induced hepatocarcinogenesis in mice.

DEN undergoes initial metabolic conversion to a reactive α-hydroxydiethylnitrosamine (hydroxylated DEN) by hepatic cytochrome P450 (CYP)-dependent monooxygenases, followed by the formation of an ethyldiazonium ion intermediate that gives rise to DNA adducts (24). Among CYPs involved in the bioactivation of DEN, cytochrome P450 2E1 (Cyp2e1) has been considered to be the primary catalyst (24, 25). Accordingly, we compared the expression levels of Cyp2e1 in 15-day-old WT and Nrf2 KO mice at 6 and 48 hours after DEN injection. The results show that the hepatic protein levels of Cyp2e1 were not significantly different between WT and Nrf2 KO mice (Fig. 2A). As illustrated in Fig. 2B, immunohistologic analysis also revealed that WT and Nrf2 KO mice displayed a similar zonation of Cyp2e1 in the centrilobular region of the liver where the distribution of Cyp2e1 is mainly confined (26).

The magnitude of the hepatocarcinogenicity of DEN could be determined by its bioavailability in the liver as well, which is in part affected by the degree of blood supply to the hepatocytes. A vascular anomaly, such as portosystemic shunt, would decrease the amount of chemicals carried to the liver (27). To ascertain whether or not any congenital malformations possibly present in the hepatic vasculature of Nrf2 KO mice, we used could affect the delivery of DEN to the liver, we perfused the livers of 15-day-old mice with a 0.4% solution of Trypan blue. The perfusion showed that Trypan blue perfused, through the hepatic portal vein, penetrated the liver vasculature of Nrf2 KO mice as well as WT mice (Supplementary Fig. S2A). In addition, none of the Nrf2 KO mice examined displayed dysplasia in the regions surrounding the portal triad (Supplementary Fig. S2B). These findings suggest that the genetic disruption of Nrf2 does not provoke noticeable hepatic vascular abnormalities in the mice used in our study.

To understand the mechanisms by which Nrf2 drives tumor progression in the liver of DEN-treated mice, we first determined the level of Nrf2 expressed in hepatomas. As shown in Fig. 3A, Nrf2 was detected at elevated levels within the DEN-induced liver tumors as compared with that in the normal liver tissue of vehicle-treated WT mice. Furthermore, we noticed that nuclear accumulation of Nrf2 was increased markedly in the tumors (Fig. 3B). In addition, immunohistologic analysis showed an increase in the levels of Nrf2 in the hepatomas (Fig. 3C). The overexpression and increased accumulation of Nrf2 in the nucleus were found to be functional as indicated by an increased expression of its principal target protein Nqo1 (Fig. 3D) and its mRNA transcript (Fig. 3E). To more precisely examine the transcriptional activity of Nrf2 in the hepatomas, ARE-hPAP Tg reporter mice were utilized (20). These mice carry a construct of a reporter gene hPAP containing a core sequence ARE, which is expressed under the control of Nrf2, and hence its expression could specifically represent the transcriptional activity of Nrf2 (Fig. 3F, top). Liver sections from DEN-treated ARE-hPAP mice displayed strong hPAP activity, reflecting the enhanced transcriptional activation of Nrf2 (Fig. 3F, bottom). Together, these results indicate that Nrf2 is functionally overactivated in the hepatomas of DEN-treated mouse livers.

Cell proliferation is a fundamental process required for cancer progression (28). Examination of liver sections of mice at 9 months after DEN injection revealed that proliferating cells stained with PCNA were heavily detected in the hepatomas from WT mice but rarely in adjacent normal tissues of WT mice or the livers of Nrf2 KO mice (Fig. 4A). Similarly, Western blot analysis showed an increase in the protein level of cyclin D1 (Fig. 4B) in WT livers but not in the livers of DEN-treated Nrf2 KO mice. These observations suggest that Nrf2 is likely to drive DEN-induced hepatocarcinogenesis by stimulating hepatocyte proliferation. In an attempt to corroborate this speculation, we examined the short-term responses elicited by the DEN administration to WT and Nrf2 KO mice. It has been well established that hepatocytes are quiescent in the normal liver, but acquire a remarkable capacity to proliferate in response to liver damage triggered by carcinogens or inflammatory insults. Not only does this mode of proliferation of surviving hepatocytes compensate for cell death to allow the maintenance of liver mass, but it also transmits genetic alterations to daughter cells in the favor of liver neoplastic progression, followed by dysplasia and later HCC development (29). Consistently, it has been firmly believed that compensatory hepatocyte proliferation is required for DEN-induced carcinogenesis (21, 30). We therefore assessed hepatic proliferation in WT and Nrf2 KO mice at the age of 15 days following DEN treatment for 48 hours. BrdUrd pulse-chase experiment and PCNA staining demonstrated the dramatically reduced compensatory proliferative response in the livers of Nrf2 KO mice compared with WT mice (Fig. 4C and D). Collectively, our current study suggests that Nrf2 is necessary for cancer progression, presumably through stimulation of compensatory proliferation, which subsequently enables hepatoma formation following DEN injection in mice.

For growth and proliferation, cancer cells require not only energy but also the building blocks including nucleotides that are synthesized from glucose-derived precursors formed in the pentose phosphate pathway (PPP; ref. 31). Thus, it is not surprising to note that PPP plays an indispensable role in cancer progression and that PPP-related enzymes are frequently overexpressed in HCC (32, 33). In line with this notion, suppressing these enzymes reportedly hampered HCC development (31–33). Notably, the expression of some key PPP-related enzymes is known to be Nrf2-dependent (12). Thus, we assessed the expression levels of the transporter of glucose uptake and representative PPP enzymes in the livers of WT and Nrf2 KO mice that had been treated with DEN. We first noticed that the level of glucose transporter type 1 gene (Glut1) was markedly elevated in the livers of WT mice but not in those of Nrf2 KO littermates given DEN (Supplementary Fig. S3). The qRT-PCR assay revealed that the hepatic expression levels of PPP metabolic genes encoding glucose-6-phosphate dehydrogenase (G6pd), phosphogluconate dehydrogenase (Pgd), transaldolase (Taldo), transketolase (Tkt), malic enzyme 1 (Me1), isocitrate dehydrogenase 1 (Idh1), phosphoribosyl pyrophosphate amidotransferase (Ppat), and methylenetetrahydrofolate dehydrogenase 2 (Mthfd2) were markedly reduced in the DEN-treated Nrf2 KO mice compared with those in WT animals subjected to the same treatment (Fig. 5A). Western blot analysis also confirmed that there was a profound downregulation of Pgd and Me1 in the livers of Nrf2 KO mice treated with DEN (Fig. 5B).

Mutations in the Keap1-binding domain of Nrf2 might prevent its degradation, which arises from Keap1-binding in the cytoplasm. This will result in enhanced nuclear translocation of Nrf2 and subsequently its transcriptional activity by binding to ARE (16). Sequencing analysis of Nrf2 in human HCC and experimentally induced rat HCC has revealed that Nrf2 is commonly mutated in the region encoding the domain interacting with Keap1 (3, 18). In agreement with this notion, we also observed that Nrf2 was mutated in DEN-induced HCC in mice, particularly in pairs of codons encoding amino acid residues 29 (80%) and 32 (100%) of its DLG motif (Fig. 6A). The Nrf2 levels in cells transfected with constructs expressing either murine or human Nrf2, which had been engineered to carry a single mutation in the codon encoding the amino acid residue 29, 32, or 80, were higher compared with those in cells harboring WT Nrf2 (Fig. 6B and C). These findings suggest that mutations in the DLG motif of Nrf2 may account for its overactivation in the DEN-induced HCC. Together, our data demonstrate that the Nrf2 activity in the hepatomas of mice given DEN was enhanced, likely due to mutations, including those in the Keap1-binding domain-coding region of Nrf2.

Nrf2, the master regulator of cellular redox status, is now being put at the heart of debate over its dual roles as a tumor suppressor and an oncogenic factor (34). Previous studies have demonstrated that deficiency of Nrf2 renders animals more vulnerable to various types of carcinogens (35, 36), while activating Nrf2 may reduce tumor burden (37). However, utilization of Nrf2 KO mice allowed us to demonstrate the unexpected tumorigenic potential of Nrf2 in the DEN-induced murine HCC model. Our findings are in agreement with the recently reported observations that activation of Nrf2 promotes the progression of lung (38), skin (39), and pancreatic cancers (11, 40).

Many different types of cancers are known to express elaborated levels of ROS (41). It is crucial for cancer cells to maintain redox homeostasis as the excessive oxidative stress can also be detrimental to their survival (41). This strategy could be achieved by programmed expression of antioxidant enzymes that are mainly under the control of Nrf2. In line with this speculation, the progression of pancreatic cancer was promoted by K-Ras, B-Raf, and Myc oncogenes via the Nrf2-triggered antioxidant programs (11). Our current study also demonstrates that the expression levels of antioxidant genes including Nqo1, heme oxygenase 1 (Hmox1), and glutamate-cysteine ligase catalytic subunit (Gclc), whose transcriptional expression is under the tight control of Nrf2 (10), were elevated in the livers of WT mice but not in Nrf2 KO counterparts treated with DEN (Supplementary Fig. S4).

Redox status of cancer cells can also be modulated by the reduced form of nicotinamide adenine dinucleotide phosphate (NADPH), a reducing equivalent generated in the PPP (31). We noticed markedly elevated transcription levels of ARE-regulated PPP enzymes in the DEN-treated WT livers, but they were strongly suppressed in the livers of Nrf2 KO mice given DEN. In addition to generating NADPH, the principal function of PPP is to produce phosphopentoses and ribonucleotides required for cell growth and proliferation (31). Therefore, the disparity in the extent of these metabolic enzymes in the livers of WT and Nrf2 KO mice treated with DEN might account for the difference in the tumor burden between these two genotypes. In agreement with our observation, Nrf2 Tg mice developed more tumors than their control littermates in a chemically induced skin carcinogenesis model due to the overexpression of ARE-mediated metabolic enzymes that favored the tumor cell growth (39). These findings suggest that the oncogenic function of Nrf2 is achieved, at least in part, through metabolic reprogramming toward anabolic glucose metabolism via the PPP.

It is noteworthy to mention that in another study, Nrf2-mutant mice rather had sporadic congenital intrahepatic shunt, which evoked an altered expression of Cyp2e1 and consequently reduced hepatotoxicity of acetaminophen whose bioactivation is also Cyp2e1-dependent (42). These findings may not be compatible with our current study. Although it is challenging to compare these two studies, we anticipate that the discrepancy might presumably be due to variations in strategies utilized to generate Nrf2-mutant mice. Nrf2 gene in the former study was disrupted using a positive-negative selection targeting vector whose lacZ cassette substitutes for part of exon 5 of Nrf2. This results in the deletion of 280 amino acids in carboxyl terminus of Nrf2, generating mutant Nrf2 with remaining N-terminal 301 residues linked to LacZ (6, 42). In contrast, we used mice with part of exon 4 and all of exon 5 of Nrf2 replaced with a reporter gene lacZ in the targeting vector. The replacement depletes the carboxyl 457 amino acids of Nrf2 and effectively nullifies the functions of Nrf2 gene (19).

As Nrf2 appears to serve as an oncogenic factor in hepatocarcinogenesis, deletion of Nrf2 might alleviate liver cancer formation. However, Nrf2 is required for maintenance of physiologic redox homeostasis. Therefore, harnessing but not completely deleting Nrf2 might be a better strategy for cancer therapy. Moreover, Nrf2 overactivation often occurs as a consequence of its mutations by environmental carcinogens rather than amplification of its normal gene, so it might be a more fundamental preventive option to avoid genetoxic insults damaging Nrf2 and/or its regulator Keap1. Large-scale whole-exome sequencing and whole-genome sequencing have identified somatic mutations in Nrf2 in 3.7% and 6.4% of patients with HCC, respectively (3, 4). Notably, Nrf2 mutations were heavily clustered within the region coding for the Keap1-binding DLG and ETGE motifs (3, 4), which endows Nrf2 with gain-of-function activity (16). Another study in an attempt to classify genes driving HCC by exome sequencing, however, spotted mutations exclusively in Keap1 but not Nrf2, which was associated with decreased expression level of Keap1 (17). The contradiction between these studies is obscure and somehow dependent on variations in samples collected. Nevertheless, all these genetic alterations share one thing in common, that is, they all result in the overactivation of Nrf2. Therefore, correcting these mutations may hamper the Nrf2-dependent cancer cell growth and tumor progression. In this context, it is interesting to note that replenishment of the tumor suppressor gene adenomatous polyposis coli (APC), which is frequently mutated and inactivated in colorectal cancer, helps promote differentiation of colon cancer cells and reverse colon cancer in mice (43). A recently developed ready-to-used genome editing technology CRISPR-Cas9 (clustered regularly interspaced short palindromic repeats/ CRISPR-associated protein 9) may offer a new therapeutic approach for intervention of liver cancer caused by “gain-of-function” mutations in Nrf2 (44–46). On the one hand, CRISPR-Cas9 would restrain the accumulation of Nrf2 by precisely allowing the replacement of hotspot mutations in Nrf2 with the corresponding WT nucleotides. On the other hand, inserting deactivating mutations to DNA-binding domain coding region of Nrf2 (Neh1 domain) also could attenuate its transcriptional activity regardless of a large amount of Nrf2 accumulated in the nucleus. The application of this genome editing tool would be extended to correct the deactivating mutations in Keap1, allowing restoration of its function to sequester the transcription factor Nrf2 in the cytoplasm.

It is also noticeable that mutations in Nrf2 are related to shorter survival time in patients with HCC (Supplementary Fig. S5; Supplementary Dataset S1). Therefore, testing and/or profiling Nrf2 mutations would be recommended as part of every single clinical practice guideline for liver cancer to categorize patients into groups with or without genetic alterations in Nrf2 and subsequently determine a precision therapy for associated HCC patients.

In summary, our findings reveal that Nrf2 serves as a key driver in hepatocarcinogenesis. In this disorder, Nrf2 gains its oncogenic functions to enhance the expression of genes involved in the uptake and redistribution of glucose into PPP to support cancer cell growth as illustrated in Fig. 7.

No potential conflicts of interest were disclosed.

Conception and design: H.K.C. Ngo, D.-H. Kim, Y.-N. Cha, H.-K. Na, Y.-J. Surh

Development of methodology: H.K.C. Ngo, Y.-N. Cha, H.-K. Na

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.K.C. Ngo, D.-H. Kim

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.K.C. Ngo, Y.-N. Cha, H.-K. Na, Y.-J. Surh

Writing, review, and/or revision of the manuscript: H.K.C. Ngo, D.-H. Kim, Y.-N. Cha, Y.-J. Surh

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.K.C. Ngo, D.-H. Kim

Study supervision: Y.-N. Cha, Y.-J. Surh

The authors thank Dr. Hoang Le (Seoul National University) for discussion, comments, and reading the manuscript.

This work was supported by the Global Core Research Center (GCRC) grant (no. 2011-0030001 to Y.J. Surh) from the National Research Foundation, Republic of Korea.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.