Abstract

   The development of nonalcoholic fatty liver disease (NAFLD) is
   associated with increased reactive oxygen species (ROS) production.
   Previous observations on the contradictory roles of general control
   nonderepressible 2 (GCN2) in regulating the hepatic redox state under
   different nutritional conditions prompted an investigation of the
   underlying mechanism by which GCN2 regulates ROS homeostasis. In the
   present study, GCN2 was found to interact with NRF2 and decrease NRF2
   expression in a KEAP1-dependent manner. Activation of GCN2 by
   halofuginone treatment or leucine deprivation decreased NRF2 expression
   in hepatocytes by increasing GSK-3β activity. In response to oxidative
   stress, GCN2 repressed NRF2 transcriptional activity. Knockdown of
   hepatic GCN2 by tail vein injection of an AAV8-shGcn2 vector attenuated
   hepatic steatosis and oxidative stress in leptin-deficient (ob/ob) mice
   in an NRF2-dependent manner. Inhibition of GCN2 by GCN2iB also
   ameliorated hepatic steatosis and oxidative stress in both ob/ob mice
   and high fat diet-fed mice, which was associated with significant
   changes in lipid and amino acid metabolic pathways. Untargeted
   metabolomics analysis revealed that GCN2iB decreased fatty acid and
   sphingomyelin levels but increased aliphatic amino acid and
   phosphatidylcholine levels in fatty livers. Collectively, our results
   provided the first direct evidence that GCN2 is a novel regulator of
   NRF2 and that specific GCN2 inhibitors might be potential drugs for
   NAFLD therapy.

   Keywords: GCN2, NRF2, KEAP1, Hepatic steatosis, Oxidative stress

Graphical abstract

   [35]Image 1
   [36]Open in a new tab

Highlights

     * •
       GCN2 interacts with NRF2 and decreases NRF2 expression in a
       KEAP1-dependent manner.
     * •
       Activation of GCN2 by amino acid starvation decreases NRF2
       expression by increasing GSK-3β activity.
     * •
       GCN2 knockdown attenuates hepatic steatosis and oxidative stress in
       an NRF2-dependent manner.
     * •
       Inhibition of GCN2 alleviates hepatic steatosis and oxidative
       stress in obese mice.

1. Introduction

   As an important metabolic organ, the liver is often impaired or even
   pathologically damaged by overnutrition, obesity and metabolic syndrome
   [[37]1]. Lipotoxicity is one of the most frequent causes of liver
   abnormalities [[38]2,[39]3]. Moreover, the liver can also be injured by
   chemical toxins, viruses and pathogens [[40]4,[41]5]. Many of these
   types of disorders are associated with increased reactive oxygen
   species (ROS) production, and oxidative stress is one of the most
   important mechanisms for multiple hepatic diseases, including alcoholic
   or nonalcoholic fatty liver disease (NAFLD), hepatitis C virus
   infection, steatohepatitis and cirrhosis [[42]1,[43][6], [44][7],
   [45][8]].

   Nuclear factor erythroid 2-related factor 2 (NRF2) play an important
   role in maintaining cellular redox homeostasis. Under normal
   conditions, NRF2 is negatively regulated by Kelch-like ECH-associated
   protein 1 (KEAP1) through the connection between NRF2 and Cullin-3
   (Cul3), which forms a ubiquitin E3 ligase complex [[46]9,[47]10]. In
   response to oxidants, NRF2 is released from the complex and
   translocates to the nucleus to form a heterodimer with the small MAF
   protein, which then activates the transcription of antioxidant and
   detoxification genes that contain antioxidant response element (ARE)
   sequences in their promoters [[48]11]. NRF2 also affects hepatic lipid
   metabolism. Genetic deletion of Nrf2 profoundly increases the
   susceptibility of mice to nonalcoholic steatohepatitis (NASH) and
   cirrhosis upon consumption of a high-fat diet (HFD) [[49]12] or
   methionine- and choline-deficient (MCD) diet [[50]13], whereas genetic
   or pharmacologic activation of NRF2 decreases their sensitivity to NASH
   [[51]14,[52]15].

   As an amino acid sensor, general control nonderepressible 2 (GCN2) is
   activated by amino acid starvation and phosphorylates eukaryotic
   initiation factor 2α (eIF2α) to maintain amino acid homeostasis by
   attenuating global mRNA translation and inducing selective stimulation
   of the expression of amino acid biosynthetic genes [[53]16,[54]17].
   GCN2 is highly expressed in the liver and is involved in the regulation
   of hepatic lipid metabolism and redox state under different nutritional
   conditions. In response to both dietary and pharmaceutical amino acid
   deprivation, Gcn2^−/− mice display severe hepatic steatosis and
   increased oxidative stress [[55]18,[56]19]. When mice are fed a HFD,
   GCN2 deficiency significantly attenuates liver dysfunction, insulin
   resistance, hepatic steatosis and oxidative stress in HFD-fed mice
   [[57]20].

   There is evidence that GCN2 may affect NRF2 pathway. For example, In
   high glucose-stimulated ARPE-19 cells, GCN2 depletion reduced cell
   apoptosis and enhanced activation of the NRF2 pathway [[58]21].
   Halofuginone (HF), a GCN2 agonist, represses NRF2 expression in
   NRF2-addicted cancer cells [[59]22]. However, the mechanism by which
   GCN2 regulates the NRF2 pathway remains unclear. In the present study,
   we found that GCN2 overexpression decreased NRF2 expression via
   protein-protein interactions and that GCN2 activation decreased NRF2
   signaling by activating glycogen synthase kinase-3β (GSK-3β). Moreover,
   knockdown of GCN2 improved hepatic steatosis, insulin resistance and
   oxidative stress in leptin-deficient (ob/ob) mice in an NRF2-dependent
   manner. Finally, we showed that the GCN2-specific inhibitor GCN2iB has
   potential therapeutic effects on NAFLD.

2. Results

   GCN2 represses NRF2 expression. To investigate whether GCN2 regulates
   the NRF2 pathway, we transfected 293T cells with Myc epitope-tagged
   GCN2 (GCN2-Myc) and GFP-tagged NRF2 (NRF2-GFP) expression constructs.
   Forced expression of GCN2-Myc reduced the abundance of NRF2-GFP in a
   dose-dependent manner ([60]Fig. 1A). The reduction in NRF2-GFP
   expression induced by GCN2 overexpression was significantly attenuated
   by treatment with 2 μM GCN2iB (a GCN2-specific inhibitor) or MG132 (a
   proteasome inhibitor) ([61]Fig. 1B). In HepG2 cells, forced expression
   of GCN2-Myc also decreased the expression of endogenous NRF2 in a
   dose-dependent manner ([62]Fig. 1C). In contrast, stable knockdown of
   GCN2 by transfection of a GCN2-specific shRNA lentiviral vector
   (shGCN2) resulted in an approximate 2-fold increase in NRF2 expression
   ([63]Fig. 1D).

Fig. 1.

   [64]Fig. 1
   [65]Open in a new tab

   GCN2 negatively regulates NRF2 expression. (A) 293T cells transfected
   with a NRF2-GFP expression vector and various amounts (1 × , 2 × and
   3 × ) of GCN2-Myc vector were lysed and subjected to Western blot
   analysis with antibodies against Myc, GFP or β-tubulin (loading
   control). (B) 293T cells transfected with NRF2-GFP and GCN2-Myc (2 × )
   vectors were treated with 2 μM GCN2iB or MG132 for 6 h and then lysed
   and subjected to Western blot analysis. (C) HepG2 cells were
   transfected with various amounts (1 × , 2 × and 3 × ) of GCN2-Myc
   vector for 24 h. Cells were lysed and subjected to Western blot
   analysis. (D) HepG2 cells were stably transfected with shRNA lentiviral
   vectors targeting a scrambled sequence (shScr) or GCN2 (shGCN2), and
   cell lysates were examined by Western blot analysis. (E) HepG2 cells
   were treated with 0–200 nM halofuginone (HF) for 24 h, and cell lysates
   were then subjected to Western blot analysis. (F) HepG2 cells were
   cotreated with 100 nM HF and 2 μM chloroquine or MG132 for 24 h, and
   the cell lysates were then subjected to Western blot analysis. (G)
   HepG2 cells treated with 0–5 μM GCN2iB were lysed and subjected to
   Western blot analysis. (H) After administration of cycloheximide (CHX),
   control and GCN2iB-treated (2 μM) HepG2 cells were lysed at different
   times, and NRF2 expression in the cell lysates was determined by
   Western blot analysis. (I) Mice were treated with 3 mg/kg HF for 3
   days, and liver lysates were examined by Western blot analysis. (J)
   Mice were fed a control (nutritionally complete amino acid, Leu (+)) or
   leucine-deficient (Leu (−)) diet for 7 days. Liver lysates were then
   examined by Western blot analysis. N=3, values represent the
   mean ± SEM; * indicates p<0.05; ** indicates p<0.01.

   HF has been shown to activate GCN2 by increasing the levels of
   uncharged prolyl-tRNA [[66]23]. Consistently, HF treatment increased
   the phosphorylation of eIF2α in HepG2 and NCTC 1469 cells ([67]Fig.
   S1). Interestingly, HF decreased NRF2, heme oxygenase-1 (HO-1) and
   NAD(P)H quinone dehydrogenase 1 (NQO-1) expression in HepG2 and NCTC
   1469 cells in a dose-dependent manner ([68]Fig. 1E, [69]Fig. S2). The
   HF-mediated decrease in NRF2 expression was attenuated by MG132, but
   not by chloroquine (an inhibitor of autophagy), indicating that HF
   decreases NRF2 expression via the proteasome pathway ([70]Fig. 1F). In
   293T cells, HF treatment significantly decreased ARE luciferase
   activity in control and tertiary butylhydroquinone (tBHQ)-treated 293T
   cells. In contrast, ARE luciferase activity was significantly increased
   by GCN2iB ([71]Fig. S3).

   We next treated cells with GCN2iB to examine the effect of GCN2
   inhibition on NRF2 expression. In HepG2 and NCTC 1469 cells, GCN2iB
   increased NRF2, HO-1 and NQO-1 expression in a dose-dependent manner
   ([72]Fig. 1G, [73]Fig. S4). To determine whether GCN2iB affects NRF2
   protein stability, we treated HepG2 cells with 50 μg/mL cycloheximide
   (CHX) to inhibit protein synthesis. CHX treatment decreased NRF2
   expression in a time-dependent manner, but this decrease in NRF2
   expression was significantly attenuated by GCN2iB ([74]Fig. 1H).

   In the livers of C57BL/6 mice, HF treatment (3 mg/kg, 3 days) decreased
   the protein expression of NRF2, HO-1 and NQO-1 ([75]Fig. 1I). Compared
   to mice fed regular chow (nutritionally complete amino acids, Leu(+)),
   the hepatic protein expression of NRF2, HO-1 and NQO-1 was decreased in
   mice fed a leucine-deficient (Leu(−)) diet for 7 days, but the
   decreases in hepatic NRF2, HO-1 and NQO-1 expression caused by the Leu
   (−) diet were diminished in Gcn2^−/− mice ([76]Fig. 1J).

   GCN2 overexpression decreases NRF2 expression in a protein-protein
   interaction- and KEAP1-dependent manner. Analysis with the STRING
   online database suggested a strong functional association between GCN2
   (EIF2AK4) and NRF2 (NFE2L2) ([77]Fig. S5). There are five conserved
   folded domains in the protein structure of GCN2, including an
   N-terminal RWD domain, a pseudokinase domain, a catalytically active
   kinase domain, a ‘HisRS-like’ domain and a C-terminal domain (CTD)
   [[78]24]. To investigate the interaction between GCN2 and NRF2, the
   following four truncated GCN2 constructs were generated: N-terminal
   truncation (GCN2-N) construct, encoding AA 1–500 of GCN2; middle
   truncation (GCN2-M) construct, encoding AA 500–1100 of GCN2; C-terminal
   truncation (GCN2-C) construct, encoding AA 1010–1649 of GCN2; and
   GCN2-1100aa truncation (GCN2-1100aa), containing the regions of GCN2-N
   and GCN2-M truncations ([79]Fig. 2A). We then transfected 293T cells
   with the NRF2-GFP and GCN2-Myc constructs or GCN2 truncations and
   performed coimmunoprecipitation (co-IP) analysis with anti-Myc magnetic
   beads. Western blot analysis showed that NRF2-GFP was only detected in
   the GCN2-Myc or GCN2-C-myc immunoprecipitates ([80]Fig. 2B). In HepG2
   cells expressing GCN2-C-Myc constructs, endogenous NRF2 could also be
   immunoprecipitated by anti-Myc magnetic beads ([81]Fig. S6a). The
   interaction between NRF2 and the GCN2-C truncation construct was
   further confirmed by reciprocal immunoprecipitation assay in 293T cells
   expressing NRF2-Myc and GCN2-C-GFP constructs ([82]Fig. S6b). Moreover,
   forced expression of GCN2-C-Myc caused an approximately decrease in the
   abundance of NRF2-GFP, whereas forced GCN2-N-Myc expression had no
   obvious effect ([83]Fig. 2C). Although forced GCN2-M-Myc expression
   also reduced NRF2 expression, the lowering effect was weaker than that
   of the GCN2-C-Myc construct ([84]Fig. 2C). Confocal microscopy images
   showed that when 293T cells were cotransfected with GCN2 or GCN2-C
   vectors, the NRF2-GFP fluorescence intensity was dramatically reduced
   and the remaining NRF2 was mainly located in the nucleus ([85]Fig. 2D).

Fig. 2.

   [86]Fig. 2
   [87]Open in a new tab

   GCN2 interacts with NRF2 and decreases its expression via
   protein-protein interactions. (A) Protein domain diagrams of GCN2 and
   its truncation constructs. (B) Lysates of 293T cells transfected with
   NRF2-GFP plus different GCN2 truncation constructs were subjected to
   immunoprecipitation (IP) with anti-Myc magnetic beads, and the
   resulting immunoprecipitates (IPs) as well as cell lysates used for IP
   (Input) were subjected to Western blot analysis with antibodies against
   the indicated tags. (C) Lysates of 293T cells transfected with NRF2-GFP
   plus different GCN2 truncation constructs were subjected to Western
   blot analysis. (D) 293T cells were cotransfected with NRF2-GFP and an
   empty vector, mCherry-tagged GCN2 or GCN2 truncation constructs, and
   cells were then examined by confocal microscopy. Scale bar=20 μm. (E)
   Protein domain diagrams of NRF2 and its truncation constructs. (F)
   Lysates of 293T cells transfected with GCN2-GFP plus different NRF2
   truncations were subjected to IP analysis as in (B). (G) Lysates of
   293T cells transfected with GCN2-GFP plus different NRF2 truncation
   constructs were subjected to Western blot analysis. (H) 293T cells
   expressing NRF2-GFP were cotransfected with GCN2-Myc and/or single
   guide RNA (sgRNA) targeting the KEAP1 expression construct plus the
   SpCas9 expression vector for 24 h. Cells were then lysed and subjected
   to Western blot analysis. (I) KEAP1^−/− HepG2 cells were transfected
   with various amounts (1 × , 2 × and 3 × ) of GCN2-Myc vector for 24 h.
   Cells were then lysed and subjected to Western blot analysis. (J)
   Lysates of 293T cells cotransfected with KEAP1-GFP plus GCN2-Myc were
   subjected to IP analysis. (K) Lysates from WT and KEAP1^−/− HepG2 cells
   were subjected to IP analysis with anti-NRF2 antibody or IgG.
   Independent experiments were performed at least 3 times. In Figure C
   and G, N=3. Values represent the mean ± SEM; * indicates p<0.05; **
   indicates p<0.01.

   NRF2 is a short half-life protein, and its stability is mainly
   regulated by KEAP1, which interacts with the Neh2 domain of NRF2 by
   binding to ETGE and DLG motifs [[88]10]. To determine the GCN2-binding
   region in NRF2, the following four NRF2 truncation constructs were
   generated: NRF2-Δ33aa construct, in which the KEAP1-binding motif DLG
   was deleted; NRF2-Δ100aa construct, in which the total Neh2 domain was
   deleted; N-terminal truncation (NRF2-N) construct, encoding AA 1–350 of
   NRF2; and C-terminal truncation (NRF2-C) construct, encoding AA 300–605
   of NRF2 ([89]Fig. 2E). Co-IP analysis revealed that GCN2-GFP was only
   detected in the immunoprecipitates of NRF2-Myc and NRF2-N-Myc ([90]Fig.
   2F), indicating that the DLG motif in the Neh2 domain of NRF2 is
   required for the interaction between GCN2 and NRF2. The interaction
   between GCN2 and NRF2-N truncation was further confirmed by reciprocal
   immunoprecipitation assay in 293T cells expressing GCN2-Myc and
   NRF2-N-GFP constructs ([91]Fig. S6c). Moreover, overexpression of GCN2
   significantly decreased NRF2-N-Myc expression but had no obvious effect
   on the expression of other truncations ([92]Fig. 2G). Confocal
   microscopy images showed that NRF2-N was mainly located in the
   cytoplasm and NRF-C-GFP was in the nucleus. The fluorescence intensity
   of NRF2-N-GFP was reduced, whereas the fluorescence intensity of
   NRF2-C-GFP was unaffected by cotransfection with the GCN2-mCherry
   vector ([93]Fig. S7).

   To determine whether KEAP1 is involved in GCN2-mediated NRF2
   downregulation, 293T cells were transfected with a KEAP1-specific
   single-guide RNA (sgRNA) expression construct plus SpCas9 expression
   vector (Lenti-U6-sp-KEAP1-gRNA-EFFS-spCas9-P2A-Puro), which decreased
   KEAP1 expression by approximately 90%. Depletion of KEAP1 significantly
   attenuated the GCN2-induced reduction in NRF2 expression ([94]Fig. 2H).
   We next generated KEAP1^−/− HepG2 cells via stable transfection with
   the Lent-U6-sp-KEAP1-gRNA-EFFS-spCas9-P2A-Puro lentiviral vector.
   Similarly, the effect of GCN2 overexpression on NRF2 reduction was
   almost diminished in KEAP1^−/− HepG2 cells ([95]Fig. 2I).
   Interestingly, co-IP analysis with anti-Myc magnetic beads indicated
   that KEAP1-GFP was detected in the GCN2-Myc immunoprecipitates,
   indicating that GCN2 also interacts with KEAP1 ([96]Fig. 2J). However,
   co-IP analysis with anti-NRF2 or IgG showed that endogenous NRF2
   interacted with GCN2 in both wild-type (WT) and KEAP1^−/− HepG2 cells
   ([97]Fig. 2K), suggesting that the interaction between NRF2 and GCN2 is
   independent of KEAP1.

   GCN2 activation decreases NRF2 expression via a GSK-3β-dependent
   pathway. To determine whether protein interaction is required for GCN2
   activation-mediated NRF2 reduction, 293T cells were transfected with
   different NRF2 truncations and then treated with 50 nM HF for 24 h. HF
   treatment resulted in significant decreases in the expression of
   NRF2-Δ100aa and NRF2-C but had no obvious effect on NRF2-N expression
   ([98]Fig. 3A). To determine whether KEAP1 is involved in HF-induced
   downregulation of NRF2, WT and KEAP1^−/− HepG2 cells were treated with
   100 nM HF for 24 h. After treatment, there were similar decreases in
   NRF2 expression in WT and KEAP1^−/− cells ([99]Fig. 3B). These results
   suggested that neither the interaction between GCN2 and NRF2 nor KEAP1
   is required for HF-induced NRF2 reduction.

Fig. 3.

   [100]Fig. 3
   [101]Open in a new tab

   GCN2 activity affects NRF2 expression by regulating the phosphorylation
   of GSK-3β. (A) 293T cells transfected with different NRF2 truncation
   constructs were treated with 50 nM HF for 24 h, and cell lysates were
   then examined by Western blot analysis. (B) WT and KEAP1^−/− HepG2
   cells were treated with 100 nM HF for 24 h, and cell lysates were then
   subjected to Western blot analysis. (C) HepG2 cells were treated with
   0–200 nM HF for 24 h. Cell lysates were then subjected to Western blot
   analysis. (D) HepG2 cells treated with 0–5 μM GCN2iB were lysed and
   subjected to Western blot analysis. (E) HepG2 cells were cotreated with
   100 nM HF and 2 μM GCN2iB or CHIR for 24 h, and the cell lysates were
   then examined by Western blot analysis. (F) Diagram shows the serine
   residues in the Neh6 domain of human NRF2 that are predicted to be
   phosphorylated by GSK-3β. 293T cells transfected with WT or mutated
   NRF2 constructs were treated with 50 nM HF for 24 h, and the cell
   lysates were then examined by Western blot analysis. (G) Mice were fed
   a Leu (+) or Leu (−) diet for 7 days, and liver lysates were then
   examined by Western blot analysis. (H) WT mice fed a Leu (−) diet were
   administered CHIR (5 mg/kg). Liver lysates were examined by Western
   blot analysis. N=3, values represent the mean ± SEM; * indicates
   p<0.05; ** indicates p<0.01; *** indicates p<0.001.

   In addition to KEAP1, dimeric β-transducin repeat-containing protein
   (β-TrCP) also controls NRF2 stability by binding to the Neh6 domain of
   NRF2, and the binding of β-TrCP to NRF2 is controlled by GSK-3β
   [[102][25], [103][26], [104][27]]. Therefore, we examined the effect of
   HF treatment on GSK-3β phosphorylation. In HepG2 and NCTC 1469 cells,
   HF treatment decreased the phosphorylation of GSK-3β at serine 9 in a
   dose-dependent manner ([105]Fig. 3C, [106]Fig. S8). On the other hand,
   GCN2iB increased p-GSK-3β^ser9 levels in HepG2 cells in a
   dose-dependent manner ([107]Fig. 3D). As a serine/threonine kinase, it
   is unlikely that GCN2 directly induces dephosphorylation of GSK-3β.
   Therefore, we investigated whether GCN2-inactivated kinases regulate
   GSK-3β phosphorylation. It has been reported that GCN2 sustains
   mammalian target of rapamycin (mTOR)/70-kDa ribosomal S6 kinase 1
   (p70S6K1) suppression upon amino acid deprivation [[108]28] and that
   p70S6K1 inhibits GSK-3β activity via phosphorylation at serine 9
   [[109]29]. Thus, we examined the phosphorylation of p70S6K1 at
   threonine 389 in HepG2 cells. As expected, HF treatment decreased the
   p-p70S6K1^T389 levels in HepG2 cells, but GCN2iB increased the
   p-p70S6K1^T389 levels in HepG2 cells ([110]Fig. 3C and D). In HepG2
   cells, the HF-induced decreases in NRF2 expression and the
   phosphorylation of GSK-3β were attenuated by GCN2iB or the CHIR-99021
   (CHIR) GSK-3β-specific inhibitor ([111]Fig. 3E).

   GSK-3β has been reported to decrease NRF2 expression by phosphorylating
   groups of serine residues in the Neh6 region of NRF2, including S347,
   S356, S361 and S383 [[112][25], [113][26], [114][27]]. To determine
   whether these serine residues are required for HF-induced NRF2
   reduction, several of the serine residues were mutated as alanine
   residues, including S347A, S347A/S356A, S347A/S356A/S361A,
   S347A/S356A/S383A and S347A/S356A/S361A/S383A. Western blot analysis
   showed that HF treatment resulted in a similar degree of reduction in
   WT-NRF2 and S347A expression but caused less reduction in the
   expression of other mutations ([115]Fig. 3F), indicating that mutation
   in these serine residues decreased the sensitivity of NRF2 to HF
   treatment.

   Interestingly, mutation of serine residues did not affect the GCN2
   overexpression-induced decrease in NRF2 expression ([116]Fig. S9).
   However, overexpression of the GCN2-M truncation did not decrease the
   expression of the S347A/S356A/S361A/S383A mutation (NRF2-4S/4A) in 293T
   cells. Moreover, the reduction in NRF2 expression caused by
   overexpression of GCN2-M truncation was blocked by GCN2iB and CHIR
   ([117]Fig. S10).

   The Leu (−) diet decreased the phosphorylation of GSK-3β and p70S6K1 in
   the livers of WT mice but significantly increased the p-GSK-3β and
   p-p70S6K1 levels in the livers of Gcn2^−/− mice ([118]Fig. 3G).
   Inhibition of GSK-3β by CHIR increased NRF2 expression and GSK-3β
   phosphorylation in the livers of mice fed a Leu (−) diet ([119]Fig.
   3H), indicating that Leu (−) diet-induced activation of hepatic GSK-3β
   decreases NRF2 expression.

   GCN2 repressed NRF2 transcriptional activity under stress conditions.
   To investigate the effect of GCN2 on NRF2 transcriptional activity,
   shGCN2 or shRNA lentiviral vector targeting a scrambled sequence
   (shScr) stably transfected HepG2 cells were treated with
   100 μM H[2]O[2]for 24 h. Knockdown of GCN2 attenuated the
   phosphorylation of eIF2α in H[2]O[2]-treated cells ([120]Fig. S11).
   After H[2]O[2] treatment, NRF2 was translocated from the cytoplasm to
   the nucleus, and the mRNA levels of NRF2 downstream target genes (GCLM,
   GSTM1, GSTA1 and GPX1) were significantly increased. H[2]O[2]-induced
   NRF2 nuclear translocation and the upregulation of NRF2 downstream
   target genes were further promoted by GCN2 knockdown ([121]Fig. 4A and
   B). Knockdown of GCN2 increased the viability of H[2]O[2]-treated cells
   ([122]Fig. 4C). When NRF2 was depleted by an NRF2-specific shRNA
   lentiviral vector (shNRF2), the protective effect of GCN2 knockdown on
   cell viability loss was diminished ([123]Fig. 4D). In control HepG2
   cells, GCN2 knockdown did not affect intracellular ROS levels. When
   cells were treated with 200 μM H[2]O[2], the increase in intracellular
   ROS levels was significantly attenuated by GCN2 knockdown ([124]Fig.
   4E). When NRF2 was depleted, GCN2 knockdown still decreased
   intracellular ROS levels in H[2]O[2]-treated cells ([125]Fig. 4F). We
   also found that GCN2iB increased, whereas HF decreased, HO-1 and NQO-1
   expression in H[2]O[2]-treated cells, ([126]Fig. 4G). Moreover, GCN2iB
   attenuated, while HF exacerbated, the H[2]O[2]-induced loss of cell
   viability ([127]Fig. 4H). Together, these data suggested that GCN2
   negatively regulated NRF2 transcriptional activity under oxidative
   stress conditions, thereby rendered cells more vulnerable to
   H[2]O[2]-induced cell death.

Fig. 4.

   [128]Fig. 4
   [129]Open in a new tab

   Gcn2 inhibits NRF2 transcriptional activity under stress conditions.
   (A, B) HepG2 cells were stably transfected with shScr or shGCN2. After
   incubation with 100 μM H[2]O[2] for 24 h, NRF2 in the cytoplasm and
   nucleus was examined by Western blot analysis (A), and the mRNA levels
   of NRF2 downstream target genes were measured by quantitative real-time
   PCR (B). (C) HepG2 cells were treated with 0–200 μM H[2]O[2] for 24 h,
   and cell viability was measured by MTT assay. (D) HepG2 cells were
   transfected with shRNA lentiviral vector targeting NRF2 (shNRF2). After
   incubation with 0–100 μM H[2]O[2] for 24 h, cell viability was
   measured. (E) The intracellular levels of reactive oxygen species (ROS)
   in control or 200 μM H[2]O[2]-treated cells were measured. (F) HepG2
   cells transfected with shNRF2 were treated with 50 μM H[2]O[2] for
   24 h, and the intracellular ROS levels were measured. (G) Lysates of
   HepG2 cells treated with 50 μM H[2]O[2] plus GCN2iB (2 μM) or HF
   (50 nM) were examined by Western blot analysis. (H) After treatment
   with H[2]O[2] and GCN2iB (2 μM) or HF (50 nM) for 24 h, cell viability
   was measured. In Figure A, B, E-G, N=3; in Figure C, D and H, N=8.
   Values are the mean ± SEM; * indicates p<0.05; ** indicates p<0.01.

   Knockdown of Gcn2 alleviates insulin resistance, hepatic steatosis and
   oxidative stress in ob/ob mice. To investigate whether depletion of
   Gcn2 alleviates NAFLD in an NRF2-dependent manner, we depleted Gcn2
   and/or Nrf2 from the livers of ob/ob mice via tail vein injection of
   the adeno-associated virus serotype 8 (AAV8)-shGcn2 and/or AAV8-shNrf2
   vectors. Mice that received AAV8-GFP injection were used as controls.
   Knockdown of Gcn2 decreased serum aspartate aminotransferase (AST),
   aminotransferase (ALT), nonesterified fatty acid (NEFA) and insulin
   levels but increased serum high-density lipoprotein cholesterol (HDL-C)
   levels. Nrf2 knockdown increased serum AST, ALT, NEFA and insulin
   levels but decreased serum HDL-C levels. Moreover, the effect of Gcn2
   knockdown on these serum biochemical indexes was diminished when Nrf2
   was depleted ([130]Fig. 5A–E). The results of the oral glucose
   tolerance test (OGTT) and insulin tolerance test (ITT) revealed that
   GCN2 knockdown increased glucose excretion and insulin sensitivity as
   indicated by the decreased area under the curve (AUC) values ([131]Fig.
   5F–G). However, the GCN2 knockdown-induced improvements in glucose
   tolerance and insulin sensitivity were abolished by NRF2 knockdown.
   Hematoxylin and eosin (H&E), oil red O and DHE staining showed that
   Gcn2 knockdown ameliorated liver injury, steatosis and superoxide
   production in ob/ob mice ([132]Fig. 5H). Gcn2 knockdown decreased
   triglyceride (TG), total cholesterol (TC), NEFA, 3′-nitrotyrosine
   (3′-NT) and 4-hydroxynonenal (4-HNE) levels but increased the ratio of
   reduced glutathione (GSH)-to-oxidized glutathione (GSSG) in the livers
   of ob/ob mice ([133]Fig. 5I-N). Knockdown of Nrf2 not only caused more
   liver injury, steatosis and oxidative stress in the livers of ob/ob
   mice, but also abolished the Gcn2 knockdown-mediated decreases in liver
   TG, TC, NEFA, 3′-NT and 4-HNE levels as well as increase in GSH/GSSG
   ratio ([134]Fig. 5H-N). Western blot analysis demonstrated that
   AAV8-shGcn2 decreased hepatic GCN2 expression by approximately 60%.
   Furthermore, Gcn2 knockdown increased the protein expression of NRF2,
   HO-1, NQO-1, KEAP1 and p-GSK-3β. In contrast, Nrf2 knockdown resulted
   in significant decreases in the protein expression of NRF2, HO-1, NQO-1
   and p-GSK-3β (Fig. 5O). Knockdown of Nrf2 did not affect the Gcn2
   knockdown-induced increases in the protein expression of NRF2, HO-1 and
   p-GSK-3β, but it abolished the changes in the protein expression of
   KEAP1 and NQO-1 (Fig. 5O).

Fig. 5.

   [135]Fig. 5
   [136]Open in a new tab

   GCN2 knockdown ameliorates hepatic steatosis and oxidative stress in
   ob/ob mice in an NRF2-dependent manner. (A–E) Ob/ob mice were treated
   with AAV8-GFP, AAV8-shGcn2, AAV8-shNrf2 or AAV8-shNrf2 & AAV8-shGcn2
   via tail vein injection. At 4 weeks after injection, the mice were
   sacrificed, and serum aspartate transaminase (AST) (A), alanine
   transaminase (ALT) (B), nonesterified fatty acid (NEFA) (C), insulin
   (D) and high-density lipoprotein cholesterol (HDL-C) (E) levels were
   measured. (F, G) Oral glucose tolerance tests (OGTTs) (F) and insulin
   tolerance tests (ITTs) (G) were performed on GFP-, shGcn2 shNrf2-and
   shNrf2 & shGcn2-treated ob/ob mice, and the corresponding area under
   the curve (AUC) values for blood glucose levels in each group were
   calculated. (H) Representative liver sections were stained with
   hematoxylin and eosin (H&E), oil red O and dihydroethidium (DHE). Scale
   bar = 100 μm. (I–N) Liver triglyceride (TG) (I), total cholesterol (TC)
   (J), NEFA (K), 3′-nitrotyrosine (3′-NT) (L) and 4-hydroxynonenal
   (4-HNE) (M) levels as well as the ratio of reduced glutathione to
   oxidized glutathione (GSH/GSSG) (N) were measured. (O) Liver lysates
   were examined by Western blot analysis. In Figure A–N, N=5. In Figure
   O, N=3. Values represent the mean ± SEM; * indicates p<0.05; **
   indicates p<0.01. (For interpretation of the references to colour in