Experimental Nonalcoholic Steatohepatitis and Liver Fibrosis Are Ameliorated by Pharmacologic Activation of Nrf2 (NF-E2 p45-Related Factor 2)

Background & Aims Nonalcoholic steatohepatitis (NASH) is associated with oxidative stress. We surmised that pharmacologic activation of NF-E2 p45-related factor 2 (Nrf2) using the acetylenic tricyclic bis(cyano enone) TBE-31 would suppress NASH because Nrf2 is a transcriptional master regulator of intracellular redox homeostasis. Methods Nrf2+/+ and Nrf2-/- C57BL/6 mice were fed a high-fat plus fructose (HFFr) or regular chow diet for 16 weeks or 30 weeks, and then treated for the final 6 weeks, while still being fed the same HFFr or regular chow diets, with either TBE-31 or dimethyl sulfoxide vehicle control. Measures of whole-body glucose homeostasis, histologic assessment of liver, and biochemical and molecular measurements of steatosis, endoplasmic reticulum (ER) stress, inflammation, apoptosis, fibrosis, and oxidative stress were performed in livers from these animals. Results TBE-31 treatment reversed insulin resistance in HFFr-fed wild-type mice, but not in HFFr-fed Nrf2-null mice. TBE-31 treatment of HFFr-fed wild-type mice substantially decreased liver steatosis and expression of lipid synthesis genes, while increasing hepatic expression of fatty acid oxidation and lipoprotein assembly genes. Also, TBE-31 treatment decreased ER stress, expression of inflammation genes, and markers of apoptosis, fibrosis, and oxidative stress in the livers of HFFr-fed wild-type mice. By comparison, TBE-31 did not decrease steatosis, ER stress, lipogenesis, inflammation, fibrosis, or oxidative stress in livers of HFFr-fed Nrf2-null mice. Conclusions Pharmacologic activation of Nrf2 in mice that had already been rendered obese and insulin resistant reversed insulin resistance, suppressed hepatic steatosis, and mitigated against NASH and liver fibrosis, effects that we principally attribute to inhibition of ER, inflammatory, and oxidative stress.

N onalcoholic fatty liver disease (NAFLD) is associated with type-2 diabetes mellitus, insulin resistance, and obesity, as well as chronic overconsumption of an energy-dense diet containing high-fat (HF) food and sweetened beverages that contain fructose. 1 It comprises a spectrum of phenotypes ranging from simple steatosis to nonalcoholic steatohepatitis (NASH). 2 NAFLD is a major health concern because between 20% and 40% of adults who consume a western-style diet have NAFLD, of which approximately 15% suffer NASH. In some individuals, NASH progresses to cirrhosis and hepatocellular carcinoma. 3 The development of NASH entails the presence of insulin resistance, and increases in de novo lipogenesis, inflammation, and oxidative stress. 4,5 The relationship between simple steatosis and NASH can be viewed as a dynamic one, with steatosis representing successful adaptation to metabolic stress, and NASH reflecting failure on the part of mitochondria to adapt adequately to an increased metabolic burden, which in turn leads to increased mitochondrial production of reactive oxygen species. 6,7 Although NASH is principally a disease of hepatocytes, the gut and adipose tissue also contribute to hepatic insulin resistance and inflammation. 4 The appearance of insulin resistance, de novo lipogenesis, inflammation, and oxidative stress during the development of NASH seems intertwined and each is linked to endoplasmic reticulum (ER) dysfunction. 4,8 Aberrant protein folding within the ER, which represents ER stress, stimulates the unfolded protein response (UPR) through activation of 3 pathways, controlled by inositol requiring kinase-1a (IRE1a), activating transcription factor-6 (ATF6), and PRK-like ER kinase (PERK), and together these initiate an adaptive program that serves to restore proteostasis. 9 However, if ER stress persists in the liver for a prolonged period, it produces pathophysiological changes associated with NASH. For example, chronic ER stress can result in insulin resistance through stimulation of a pathway downstream of IRE1a that leads to increased serine phosphorylation of insulin receptor substrate-1. 10 Also, persistent activation of the UPR stimulates hepatic de novo lipogenesis via all 3 arms of the UPR, causing increased activity of sterol regulatory element-binding protein-1c (SREBP-1c, encoded by SREBF1), carbohydrate-responsive element-binding protein (ChREBP, encoded by MLXIPL), the spliced variant of X-box binding protein-1 (XBP1s), peroxisome proliferatoractivated receptor gamma (PPARg), CCAAT/enhancerbinding protein (C/EBP) a and C/EBPb, [11][12][13] and in turn increased lipogenesis exacerbates insulin resistance. Moreover, chronic ER stress triggers inflammation via IRE1a, resulting in stimulation of IkB kinase (IKK) and c-jun N-terminal protein kinase 1 (JNK1) that activate nuclear factor-kB (NF-kB) and activator protein 1 (AP-1), respectively, 14 events that are also linked to insulin resistance. 4,8 It may also stimulate inflammation via PERK by attenuating translation of the inhibitor of NF-kB (IkB), which results in a relative excess of NF-kB. 15,16 Lastly, chronic ER stress can initiate oxidative stress by augmenting both oxidoreductin-1 activity and release of Ca 2þ from the ER, which in turn heighten mitochondrial production of reactive oxygen species, 17 and so sensitizes the innate immune system to proinflammatory stimuli. 18 Thus, chronic ER stress promotes insulin resistance, lipogenesis, inflammation, and oxidative stress.
In the present study, we tested whether pharmacologic activation of Nrf2 suppresses NASH and have examined if this can happen after disease is manifest because this scenario reflects the clinical situation. We therefore stimulated NASH by feeding mice chronically with a HF diet along with fructose-containing drinking water (called HFFr diet). After the high-calorie regimen had been in place for 24 weeks (Study 1), and mice exhibited impaired glucose tolerance and hyperinsulinemia, we used the potent Nrf2 activator TBE-31 32,33 (Figure 1) to test if it could improve insulin sensitivity and glucose homeostasis, and mitigate NASH in the liver. Lastly, we used Nrf2 -/mice (Study 2) to demonstrate that the beneficial effects of TBE-31 in this setting require the presence of Nrf2.

Chemicals
Unless otherwise stated, these were from Sigma-Aldrich (St. Louis, MO). The acetylenic tricyclic bis(cyano enone) compound TBE-31 was synthesized as described previously. 32,34 Animals Male C57BL/6 mice were purchased from Charles River (Elphinstone, Tranent, Scotland, UK) and allowed to acclimatize for 2 weeks before being fed specific diets. Nrf2 -/mice, 35 backcrossed over at least 7 generations onto a C57BL/6 background, were also studied. All animal care protocols and experimental procedures were performed in accordance with the Animal Scientific Procedures Act (1986) and with approval of the local University of Dundee Animal Ethics Committee. From 8 to 10 weeks of age, the mice were provided ad libitum either regular chow (RC; containing 7.5% fat by energy) or a HF diet (containing 60% fat by energy), purchased from SDS Ltd (Witham, Essex, UK). Animals were killed by cervical dislocation or by terminal anesthesia using isofluorane. On sacrifice, livers were removed immediately and portions rapidly snap-frozen in liquid N 2 (and subsequently stored at -80 C) or fixed in formalin as described elsewhere. 20 All our experiments were performed in mice on a C57BL/6 background because this strain is widely used to study obesity and insulin resistance. 36 It is, however, well recognized that a proportion of these mice do not become obese when fed a HF diet, for a variety of reasons, including failure to become diabetic and increased metabolic adaptation resulting in an inability to saturate the trichloroacetic acid cycle and mitochondrial oxidative metabolism. 37,38 Consistent with this previously documented characteristic of the C57BL/6 mouse, we found a small number of animals did not become obese when fed the HF55Fr/HF30Fr diet, and these were segregated from the others as nonresponders, and the remaining obese mice were randomized into 2 different treatment groups (see Results section).
To ensure rapid development of NASH in wild-type C57BL/6 mice, the HF diet was initially provided along with drinking water containing 55% (wt/vol) fructose (ie, the HF55Fr diet), which was used to prime Nrf2 þ/þ mice for NASH before attenuating the rate of disease development by placing them on the HF diet along with drinking water containing 30% (wt/vol) fructose (ie, the standard HF30Fr diet). The sequential provision of HF55Fr and HF30Fr diets, given only to Nrf2 þ/þ C57BL/6 mice, is described in the text as the HF55Fr/HF30Fr diet. For dietary challenge of Nrf2 -/mice, only the standard HF30Fr diet was provided because they are more sensitive to NASH (when placed on a HF diet) 24 : because of this innate sensitivity, we thought it unnecessary to prime Nrf2 -/mice by placing them on a HF55Fr diet.
To test whether pharmacologic activation of Nrf2 inhibits the progression of NASH, 2 experimental protocols were adopted, with different objectives. In both cases, TBE-31 was administered (at 5 nmol/g body weight) after insulin resistance had been established, and while the animals were still receiving the HF30Fr diet: the objective of Study 1 was to test whether activation of Nrf2 by TBE-31 could reverse insulin resistance and suppress NAFLD once disease was established; the objective of Study 2 was to test whether the ability of TBE-31 to reverse insulin resistance and suppress NAFLD requires the presence of Nrf2. In Study 1, Nrf2 þ/þ mice were initially primed for 15 weeks with the HF55Fr diet, followed by a further 9 weeks on the standard HF30Fr diet before treatment with TBE-31, or dimethyl sulfoxide (DMSO) vehicle control, 3 times a week for a total of 6 weeks by oral gavage while still being fed the standard HF30Fr diet (Figure 2A). In Study 2, Nrf2 -/and Nrf2 þ/þ mice were provided with the standard HF30Fr diet for 10 weeks, before being treated with TBE-31, or DMSO, for a further 6 weeks while continuing on the same diet ( Figure 2B). In both Study 1 and Study 2, mice were weighed weekly throughout the experiments and were killed 20-24 hours after receiving the final dose of TBE-31 or In group (i), Nrf2 þ/þ C57BL/6 mice were first primed over a period of 15 weeks for NASH by feeding a HF55Fr diet before being transferred to the standard HF30Fr diet at the beginning of Week 16. In group (ii), an equal number of age-matched Nrf2 þ/þ mice were fed an RC diet along with unadulterated drinking water throughout. After being placed for 24 weeks on either of these 2 dietary regimens, mice in group (i) and group (ii) were treated with either TBE-31 (5 nmol/g body weight) or DMSO vehicle control, by oral gavage 3 times/week for a total of 6 weeks, while still being provided with the same HF30Fr diet or RC diet. Glucose homeostasis was monitored in all mice by ITT, GTT, and PTT at the times indicated. (B, Study 2). In group (i), Nrf2 þ/þ and Nrf2 -/-C57BL/6 mice, of 8-10 weeks of age, were fed the standard HF30Fr diet for 10 weeks before being treated with either TBE-31 or DMSO for a total of 6 weeks while being maintained on the same diet. In group (ii), Nrf2 þ/þ and Nrf2 -/mice were fed the RC diet for 10 weeks, with no fructose in the drinking water, before being treated with either TBE-31 or DMSO for a further 6 weeks while being maintained on the same diet. Glucose homeostasis was monitored in all mice by ITT and PTT at the times indicated. DMSO by gavage. On sacrifice, blood was collected from the mice and their livers removed immediately for histologic, biochemical, and molecular analyses.

Physiological and Clinical Chemistry Measurements
The glucose tolerance test (GTT), insulin tolerance test (ITT), and pyruvate tolerance test (PTT) were carried out by intraperitoneal administration of standard doses of glucose, insulin, or pyruvate. 20,24 For Study 1, an ITT and GTT were first performed on the mice 1.5À2.5 weeks immediately preceding treatment with TBE-31, and were repeated when the mice had been treated for 4À5 weeks with TBE-31 (during Weeks 28 and 29 of Study 1). Subsequently, a single PTT was performed on these mice during Week 30. For Study 2, an ITT and PTT were first performed on mice 5À10 days immediately preceding treatment with TBE-31, and were repeated when they had been treated for 4À5 weeks with TBE-31 (during Weeks 14 and 15 of Study 2). Blood samples were collected via the tail vein or via cardiac puncture. Blood glucose, triglycerides, cholesterol, and plasma leptin and insulin were measured as described previously. 24 Fasting blood glucose and insulin levels were measured after diet and drinking water had been withdrawn for 5 hours. Plasma alanine aminotransferase activity was measured using commercial kits. 24

Histology
Mouse liver samples were fixed in 10% neutral buffered formaldehyde and processed for hematoxylin and eosin (H&E) and van Gieson staining by standard methods. The severity of liver disease in mice was evaluated histologically on H&E-stained sections using the NAFLD activity score (NAS) 39 in a blinded fashion, in which the pathologists were unaware of the diet, treatment group, or genotype of the mice.

Biochemical Analyses
Frozen mouse livers (about 100 mg each) were pulverized under liquid nitrogen using a pestle and mortar. The ground material was resuspended in ice-cold RIPA lysis buffer to which had been added protease and phosphatase inhibitors (Roche) before homogenization. Whole-cell lysates used for Western blot analyses represented the supernatant fraction (15,000 Â g, 15 min at 4 C) obtained from the ground hepatic extracts. For the Nfkb and Srebp-1c Western blots, nuclear and cytoplasmic fraction extracts were prepared from frozen liver using the Pierce NE-PER kit (ThermoScientific Life Science Research Products). Protein concentrations were determined using bicinchoninic acid, and assays for NAD(P)H:quinone oxidoreductase-1 (Nqo1) enzyme activity toward menadione, total glutathione, reduced glutathione (GSH), oxidized glutathione (GSSG), and malondialdehyde were performed as described previously. 20,40 Gene Expression Profiling Total RNA was extracted from frozen mouse liver using the RNeasy kit (Qiagen, Hilden, Germany), and cDNA prepared using the Omniscript kit (Qiagen) according to the manufacturer's instructions. The relative abundance of hepatic mRNA species was measured against actin as an internal control by TaqMan real-time polymerase chain reaction (PCR; Applied Biosystems Prism model 7700) using commercial primer and probe sets (Table 1), all purchased from Life Technologies.

Western Blotting
The electrophoretic resolution of proteins was performed using precast NuPAGE Bis-Tris gels from Invitrogen (ThermoFisher Scientific). Following transfer to polyvinylidene difluoride membranes, the immunoblots were developed with an electrochemiluminescence-based system (Millipore) according to the manufacturer's protocols. Film was used to allow visualization of cross-reacting proteins; several exposures were tested before choosing the exposure that permitted detection of bands in all lanes. Quantification of the immunoblot data was conducted by measuring the band intensities using ImageJ software, which is freely available. In Study 1, immunoblotting was performed on liver samples prepared from 6 different mice for each of the diet and treatment groups. In Study 2, immunoblotting was performed on liver samples prepared from 4 different mice for each of the genotypes, diet, and treatment groups.

Statistics
Comparisons between the biochemical and molecular biology results from the experimental groups were made using Student t test or Mann-Whitney test. Data for GTT, ITT, and PTT were analyzed by 1-or 2-way analysis of variance with Bonferroni postcorrection. When appropriate, repeated analysis of variance was performed with time as subject factor. Results presented are means ± SEM. On all occasions, data from RC-fed Nrf2 þ/þ mice that had been treated with DMSO vehicle control were used as the reference point, with significant increases being indicated by asterisk symbols and decreases associated with TBE-31 treatment indicated by dollar symbols, depending on the degree of significance (corrected P values .05 considered statistically significant); in addition, significant differences between data from HFFr-fed mice and other data sets are indicated by a horizontal bar that is placed above the relevant groups. Comparisons between the histology NAFLD activity scores were made using the Kruskal-Wallis H test.

TBE-31 Improves Insulin Sensitivity and Diminishes Histologic Evidence of Nonalcoholic Steatohepatitis and Cirrhosis
In Study 1, NASH was produced in wild-type C57BL/6 mice using the HF55Fr/HF30Fr diet. After 24 weeks on this regimen (Figure 2A), Nrf2 þ/þ mice were approximately 2.3fold heavier than age-matched RC-fed animals. At this time point, 20 of the 25 HF55Fr/HF30Fr-fed mice were obese, having gained 21.0À30.3 g weight over the 24-week period, whereas the remaining 5 gained 12.1À16.6 g weight and were judged to be nonobese ( Figure 3A). The 5 nonobese HF55Fr/HF30Fr-fed mice, which had not gained more weight than the RC-fed mice over the same period, were excluded from further study because they were considered to be nonresponders 37,38 ; interestingly, ITT analyses revealed these nonresponders cleared glucose from the blood more effectively than did the remaining 20 obese mice, suggesting they were sensitive to insulin ( Figure 3B). Among the remaining animals, fasting blood glucose concentrations were greater in HF55Fr/HF30Fr-fed mice than in RC-fed mice ( Figure 3C).
ITT and GTT analyses showed the 20 obese Nrf2 þ/þ HF55Fr/HF30Fr-fed mice exhibited impaired glucose clearance and were less responsive to insulin (measured at 22-23 weeks) than the RC-fed mice ( Figure 4A and B). Following 9 weeks on the standard HF30Fr diet (ie, starting at the beginning of Week 25 of the HF55Fr/HF30Fr regimen), the obese Nrf2 þ/þ mice were randomly assigned to 2 groups that were either treated with TBE-31 or DMSO vehicle control, for 6 weeks, while continuing to be fed the HF30Fr diet. A second ITT and GTT, performed in Weeks 4 and 5, respectively, after TBE-31 administration commenced, showed that TBE-31-treated mice cleared glucose more effectively than vehicle control HF55Fr/ HF30Fr-fed mice ( Figure 4C and D). PTT also revealed that TBE-31 suppressed hepatic gluconeogenesis Acly
( Figure 4E). The improvement in glucose homeostasis in HF55Fr/HF30Fr-fed mice was associated with a decrease in weight gain ( Figure 5A), but because this was not accompanied by a significant decrease in plasma leptin ( Figure 5B) it is unlikely to be associated with differences in hunger. Interestingly, TBE-31 treatment decreased fasted plasma insulin and cholesterol levels, and plasma alanine aminotransferase activity, in HF55Fr/HF30Fr-fed mice ( Figure 5C-E).
Once the mice were killed, we first examined whether the 6-week treatment with TBE-31 had resulted in activation of Nrf2. Immunoblotting of liver extracts revealed that the abundance of Nrf2 protein was increased approximately 2-fold by the HFFr diet, and although TBE-31 treatment did not produce a significant further increase in Nrf2 abundance a trend was nevertheless apparent ( Figure 6A). Most importantly, examination of Nqo1 catalytic activity, an enzyme that is a prototypic marker of Nrf2 transactivation activity, revealed a 1.6-fold increase in livers of TBE-31-treated RC-fed mice and a 1.8-fold increase in TBE-31-treated HF55Fr/HF30Fr-fed mice, each relative to their respective DMSO-treated controls ( Figure 6B). H&E staining revealed that livers of Nrf2 þ/þ mice fed the highcalorie diet for 30 weeks, and given DMSO vehicle over the final 6 weeks, exhibited marked steatosis, inflammation, and hepatocyte ballooning ( Figure 6C), which combined to give an average NAS of 4.9 ( Figure 6D). By comparison, the number of hepatic steatotic vesicles and extent of inflammation and ballooning was diminished in HF55Fr/HF30Frfed mice administered TBE-31 over the final 6 weeks, giving an average NAS of 2.9. Individual scores for fat, inflammation, fibrosis, and hepatocyte ballooning revealed that TBE-31 decreased substantially hepatocyte ballooning (P < .05) and fibrosis (P < .05) ( Table 2). Van Gieson staining also revealed TBE-31 substantially diminished fibrosis in livers of HF55Fr/HF30Fr-fed mice ( Figure 6E).

TBE-31 Attenuates Expression of Lipid Metabolism Genes in Livers of HF55Fr/HF30Fr-Fed Mice
TBE-31 treatment decreased triglyceride and cholesterol levels in livers of Nrf2 þ/þ mice fed a HF55Fr/HF30Fr diet ( Figure 7A and B). It also decreased expression of adipose differentiation-related protein (Adrp), a marker for the presence of lipid droplets, 41 in livers of the high-calorie diet fed Nrf2 þ/þ mice ( Figure 7C), which is consistent with the steatosis observed on H&E staining of liver sections.
Because Nrf2 positively controls lipid catabolism genes, 19 we explored whether these were induced by TBE-31 treatment and found TBE-31 increased hepatic mRNA Figure 3. Nrf2 D/D mice become obese when fed the HFFr diet, and this is associated with hyperglycemia. Nrf2 þ/þ mice were fed either the RC or the HF55Fr/HF30Fr (HFFr) diet for 24 weeks. Before treatment with TBE-31 or DMSO control, the physiological effect of these dietary regimens was assessed. (A) Weight gain of individual mice over the 24-week period on the RC or HFFr diets is presented. The encircled mice, shown at the bottom of the HFFr plot, were excluded from the study on the basis that they failed to become obese. [36][37][38] (B) A comparison of insulin sensitivity (ITT at 22 weeks) of the 5 nonobese encircled unresponsive HFFr-fed mice (triangles) with that of the 20 obese responsive HFFrfed mice (squares). Results are means ± SEM (n ¼ 5 or 20 mice, for nonobese and obese groups, respectively), and significant decreases in blood glucose in the nonobese unresponsive mice compared with the obese responsive mice are indicated by: $ P < .05; $$ P < .01; $$$ P < .001. (C) The fasting blood glucose levels of Nrf2 þ/þ mice fed RC or HFFr diets. Results are means ± SEM (n ¼ 8-12 mice per group). Significant increases in fasting blood glucose, relative to that in RC-fed Nrf2 þ/þ mice, are indicated by: **P < .01. Results are means ± SEM. Significant increases in results, relative to those in livers from RC-fed Nrf2 þ/þ mice, are indicated by: *P < .05, **P < .01, ***P < .001. Significant decreases in results as a consequence of treatment with TBE-31, relative to HFFr-fed Nrf2 þ/þ mice, are indicated by: $ P < .05, $$$ P < .001. ALT, alanine aminotransferase. for acetyl-CoA oxidase 2 (Acox2), carboxylesterase 1g (Ces1g), and acetyl-CoA thioesterase 7 (Acot7) ( Figure 8A). Changes in fatty acid oxidation genes were also observed, with the HF55Fr/HF30Fr diet increasing mRNA for peroxisome proliferator-activated receptor a (Ppara), and decreasing carnitine palmitoyltransferase 1a (Cpt1a) and short-chain acyl-CoA dehydrogenase (Scad), with TBE-31 intervention raising levels of all 3 of the mRNA species ( Figure 8B).
We considered whether decreased hepatic steatosis in HF55Fr/HF30Fr-fed mice treated with TBE-31 might involve diminished de novo lipogenesis. Among transcription factors that control lipid metabolism, mRNAs for Srebp-1c and Chrebp (encoded by Srebf1 and Mlxipl, Figure 6. TBE-31 treatment increases hepatic Nrf2 activity and improves liver histology in HFFr-fed Nrf2 D/D mice. On completion of the Study 1 protocol, Nrf2 þ/þ mice were killed and livers removed. (A) A representative immunoblot for Nrf2 protein in liver extracts from RC-fed or HFFr-fed mice treated with DMSO or TBE-31 (left side), with densitometric scans of blots (right side) (n ¼ 6 biologic replicates). (B) Nqo1 catalytic activity in hepatic extracts from RC-fed and HFFr-fed mice (n ¼ 8-12 mice per group). (C) Representative images for H&E staining of liver sections from RC-and HFFr-fed Nrf2 þ/þ mice treated with DMSO or TBE-31 (scale bars ¼ 100 mm). (D) The NAFLD activity score 39 was calculated (n ¼ 8-12 mice per group): note, on the y-axis the score includes negative values because livers from RC-fed Nrf2 þ/þ mice gave NAFLD activity scores of essentially zero. (E) Representative images for van Gieson staining of liver sections from Nrf2 þ/þ mice after 30 weeks RC or HF55Fr/HF30Fr feeding, followed by 6 weeks DMSO or TBE-31 treatment (scale bars ¼ 100 mm). White bars, DMSO treated; black bars, TBE-31 treated. Results are means ± SEM. Significant increases in Nrf2 protein, Nqo1 activity, or NAFLD activity score, relative to that in livers from RC-fed Nrf2 þ/þ mice, are indicated by: *P < .05; ***P < .001. Significant decreases in NAFLD activity score upon treatment with TBE-31, relative to HFFr-fed Nrf2 þ/þ mice, are indicated by: $ P < .05. respectively) and the liver X receptor a (Lxra, also called Nr1h3) and Xbp1s, were increased by the HF55Fr/HF30Fr diet, and treatment of mice fed this high-calorie diet with TBE-31 decreased their expression ( Figure 8C). We also found that mRNA for the lipogenic transcription factors Pparg and C/ebpa were induced approximately 4.0-fold by the HF55Fr/HF30Fr diet but this was not altered significantly by treatment with TBE-31 (data not shown). By contrast, mRNA for C/ebpb was not significantly changed by either the HF5Fr/HF30Fr diet or by TBE-31 (data not Liver ballooning was significantly higher in mice fed the HFFr diet when compared with their RC-fed counterparts (Kruskal-Wallis H test; P < .0001). Ballooning in livers of mice fed the HFFr diet and treated with DMSO was significantly higher than in livers of HFFr-fed mice treated with TBE-31 (Kruskal-Wallis H test; P < .05). e Liver fibrosis (this is not included in the NAS calculation) was significantly higher in mice fed the HFFr diet when compared with their RC-fed counterparts (Kruskal-Wallis H test; P < .0001). Fibrosis in livers of mice fed the HFFr diet and treated with DMSO was significantly higher than in livers of HFFr-fed mice treated with TBE-31 (Kruskal-Wallis H test; P < .05). shown). Importantly, immunoblotting showed higher levels of Srebp-1c protein in nuclear fractions from livers of HF55Fr/HF30Fr-fed mice than in their RC-fed counterparts, and this was decreased by TBE-31 treatment ( Figure 8D). Because Srebp-1c and Xbp1s have been linked to the UPR, 12,42 their apparently coordinated decrease in expression following treatment with TBE-31 suggests that ER stress might be attenuated (see below).
Because LXRa is regulated independently of ER stress, 43 we explored whether levels of mRNA for the farnesoid X receptor (Fxr) and small heterodimer partner (Shp) were altered by the diet, or by TBE-31 treatment, because they have both been implicated in repression of Lxra. 44 Quantitative reverse transcriptase PCR (qRT-PCR) revealed that neither Fxr nor Shp mRNA levels were increased by the HF55Fr/HF30Fr diet, and neither was Data are means ± SEM. Significant increases in gene expression or protein abundance, relative to that in livers from RC-fed Nrf2 +/+ mice, are indicated by: *P < .05; **P < .01; ***P < .001. Significant decreases in gene expression or protein abundance, relative to that in livers from RC-fed Nrf2 +/+ mice, are indicated by: $ P < .05; $$ P < .05.
induced by TBE-31 treatment (data not shown). Similar analysis of mRNA for retinoid X receptor alpha (Rxra), which forms a heterodimer with Lxra, revealed that its expression was also not affected by either diet or TBE-31 (data not shown).
To assess whether a decrease in lipid import or an increase in lipid export contributes to the reduction in hepatic steatosis affected by TBE-31, we measured mRNA for cluster of differentiation 36 (Cd36), which contributes to lipid import, and microsomal triglyceride transfer protein (Mttp) and apolipoprotein B (ApoB), which contribute to lipid export. This revealed that the HF55Fr/HF30Fr diet increased Cd36 mRNA levels but that TBE-31 did not alter it ( Figure 9C). By contrast, although the HF55Fr/HF30Fr diet suppressed Mttp and ApoB mRNA levels, TBE-31 increased their abundance ( Figure 9C). Collectively, these results suggest that TBE-31 probably influences lipid transport to just a limited degree.

Treatment With TBE-31 Suppresses Endoplasmic Reticulum Stress in Livers of HF55Fr/HF30Fr-Fed Mice
Because TBE-31 decreased expression of Srebf1 and Xbp1s, we next explored whether TBE-31 antagonizes steatosis by suppressing ER stress. Initially, we examined the abundance of Bip and Pdi because their expression is increased on activation of the UPR. 45 Immunoblotting showed the HF55Fr/HF30Fr diet increased Bip levels in liver of Nrf2 þ/þ mice, and that treatment with TBE-31 attenuated the increase ( Figure 10A); this was not, however, apparent for Pdi. Further immunoblotting indicated that all 3 arms of the UPR were activated in livers Figure 10. TBE-31 suppresses ER stress in livers of Nrf2 +/+ mice fed a HFFr diet. Livers from Nrf2 +/+ mice in Study 1 were examined for changes in proteins and genes engaged in the UPR. (A) Representative immunoblots of Bip and Pdi (with actin as loading control) in hepatic extracts from RC-fed and HF55Fr/HF30Fr (HFFr)-fed mice treated with DMSO or TBE-31, along with densitometric scans of blots (n ¼ 6 biologic replicates). (B) Representative immunoblots of Atf6 p50 and p90, p-Ire1a, Xbp1s and Xbp1u, p-eIf2a, and Atf4 (with actin as loading control) along with densitometric scans of blots as indicated (n ¼ 6 biologic replicates). (C) qRT-PCR for Perk, Atf4, and Chop (n ¼ 8-12 mice per group). In all cases, white bars represent DMSO and black bars represent TBE-31. Data are means ± SEM. Significant changes are indicated: * ,$ P < .05; ** ,$$ P < .01; ***P < .001.
of HF55Fr/HF30Fr-fed mice when compared with RC-fed mice, and the abundance of these markers was decreased by TBE-31 treatment: evidence for activation of Ire1a was deduced from increases in phospho-Ire1a, for activation of Atf6 by increases in Atf6-p50, and for activation of Perk by increases in phospho-eIf2a ( Figure 10B). Hepatic levels of mRNA for Atf4 (downstream of Perk) and C/ebp homologous protein (Chop) (downstream of Perk) were also increased by the high-calorie diet and attenuated by TBE-31 ( Figure 10C). Together, these data indicate that consumption of the high-calorie diet stimulates ER stress and this can be mitigated by TBE-31. Thus, alleviation of ER stress by TBE-31 likely contributes to suppression of hepatic steatosis by blunting induction of Srebf1 and Xbp1s.
Immunoblotting revealed marked increases in nuclear levels of Nfkb p65, p52, and p50 proteins in livers from HF55Fr/HF30Fr-fed mice, which were attenuated by TBE-31 ( Figure 11B). Remarkably, the dramatic increases in hepatic nuclear Nfkb p65, p52, and p50 protein in mice fed the high-calorie diet, and the attenuation by TBE-31 treatment, were accompanied by correspondingly large decreases in Ikba protein and large increases in Ikka/b phosphorylation ( Figure 11B). The HF55Fr/HF30Fr diet also increased Jnk phosphorylation, which was suppressed by TBE-31 ( Figure 11C). Taken together with the data in Figure 10, these results suggest that ER stress in the livers of HF55Fr/HF30Fr-fed mice may stimulate Nfkbdirected transcription by decreasing translation of Ikba (downstream of Perk) and by increasing Ikk activity (downstream of Ire1a). Moreover, the ability of TBE-31 to suppress ER stress in livers of HF55Fr/HF30Fr-fed mice seems to be reflected in a diminution of diet-stimulated increases in nuclear Nfkb protein and Ikk phosphorylation, coupled with enhanced levels of Ikb protein.

TBE-31 Treatment of Mice Fed an HF55Fr/ HF30Fr Diet Decreases Oxidative Stress
Consistent with the notion that NASH is accompanied by oxidative stress, increases in malondialdehyde and oxidized protein were observed in livers of HF55Fr/HF30Fr-fed mice, and these increases were attenuated by TBE-31 treatment ( Figure 13A and B). Moreover, the high-calorie diet decreased hepatic levels of reduced glutathione (GSH) relative to that of oxidized glutathione (GSSG), which is indicative of a more oxidized intracellular environment, with the relative abundance of GSH increased by TBE-31 treatment ( Figure 13C). We also assessed whether the HF55Fr/HF30 diet and TBE-31 treatment affected expression of Nrf2-target genes that contribute to antioxidant defenses. This revealed that the high-calorie diet increased hepatic mRNA for glutamate-cysteine ligase catalytic (Gclc) and modifier (Gclm) subunits, heme oxygenase-1 (Hmox1) and solute carrier family 7 member 11 (Slc7a11), but not Nqo1, glutathione S-transferase Mu-1 (Gstm1), glutathione peroxidase-2 (Gpx2), thioredoxin-1 (Txn1) or thioredoxin reductase-1 (Txnrd1), catalase (Cat), or peroxiredoxin 6 (Prdx6). However, the abundance of all these mRNA species was increased by TBE-31 ( Figure 13C), findings that suggest TBE-31 helps restore normal redox homeostasis in livers of HF55Fr/HF30Fr-fed mice by inducing Nrf2-regulated antioxidant genes.

TBE-31 Requires Nrf2 to Increase Insulin Sensitivity and Ameliorate Adverse Liver Histology in HF30Fr-Fed Mice
Study 2 was instigated to determine whether Nrf2 is essential for TBE-31 to improve glucose use in obese mice and to mitigate NASH. When placed on the standard HF30Fr diet, Nrf2 þ/þ mice rapidly gained weight before treatment with TBE-31 or DMSO ( Figure 14A); the Nrf2 -/mice also gained substantial weight when placed on the HF30Fr diet, relative to RC-fed Nrf2 -/mice, but a wellrecognized feature of the mutant mice is that they are thinner than wild-type mice. Subsequently, however, when HF30Fr-fed Nrf2 þ/þ mice were treated with TBE-31 they gained less weight than Nrf2 þ/þ mice treated with DMSO. This diminution of weight gain in HF30Fr-fed wild-type mice on treatment with TBE-31 was not observed in HF30Fr-fed Nrf2-null mice. In both wild-type and Nrf2-null mice, the weight gain caused by the HF30Fr diet was accompanied by increases in plasma leptin levels, but these increases in leptin were not attenuated by TBE-31 treatment (data not shown). Both wild-type and Nrf2-null mice that had been placed on the HF30Fr diet for 9 weeks exhibited impaired PTT and ITT. Although treatment of HF30Fr-fed wild-type mice for 5 weeks with TBE-31 reduced glucose production (determined by PTT) ( Figure 14B) and improved insulin sensitivity (determined by ITT) (Figure 14C), treatment of HF30Fr-fed Nrf2 -/mice had substantially diminished beneficial effects.
H&E staining of liver sections revealed that consumption of the HF30Fr diet for 16 weeks stimulated greater steatosis and inflammation in Nrf2 -/mice than in their wild-type counterparts ( Figure 15A), and that TBE-31 treatment of mice on the HF30Fr diet decreased steatosis and inflammation in wild-type livers, but not in Nrf2 -/livers ( Figure 15B). Histologic evaluation of steatosis, inflammation, and hepatocyte ballooning showed that the average NAS in livers of Nrf2 þ/þ mice was reduced from 2.2 to 1.0 by TBE-31 ( Figure 15C, Table 3). By contrast, the average NAS in livers of Nrf2 -/mice, estimated to be 3.1, was not reduced by TBE-31. Examination of the H&E-stained sections revealed significantly greater fibrosis in Nrf2-null livers than wild-type livers. Similarly, van Gieson staining revealed fibrosis only in livers of Nrf2 -/mice fed the HF30Fr diet ( Figure 15D).

TBE-31 Requires Nrf2 to Suppress Steatosis, Endoplasmic Reticulum Stress, Inflammation, and Oxidative Stress in Livers From HF30Fr-Fed Mice
As expected, immunoblotting showed the 6-week TBE-31 treatment increased abundance of Nrf2 protein in livers of wild-type mice; by contrast, Nrf2 protein was not detected in livers of Nrf2-null mice under basal conditions or following TBE-31 treatment ( Figure 16A). These Nrf2 immunoblotting data closely mirrored levels of Nqo1 enzyme activity in the same livers ( Figure 16B). Collectively, these findings confirm that classic Nrf2-target genes are downregulated and cannot be induced by TBE-31 in livers of Nrf2 -/mice.
Although TBE-31 treatment of HF30Fr-fed Nrf2 þ/þ mice decreased total liver triglycerides and cholesterol, this was not observed in HF30Fr-fed Nrf2-null mice ( Figure 16C, D). Similarly, TBE-31 did not decrease expression of the lipid droplet marker protein Adrp in livers of HF30Fr-fed Nrf2null mice ( Figure 16E). Although TBE-31 suppressed the increase in expression of Srebf1, Lxra, and Xbp1s in livers of Nrf2 þ/þ mice fed the HF30Fr diet, it was unable to do so in livers of HF30Fr-fed Nrf2-null mice ( Figure 17A), suggesting that the ability of TBE-31 to suppress expression of these transcription factors is mediated by Nrf2. Consistent with the qRT-PCR data for Srebf1, immunoblotting revealed TBE-31 blunted the increase in nuclear Srebp-1c protein in livers of HF30Fr-fed Nrf2 þ/þ mice, but did not do so in livers of HF30Fr-fed Nrf2 -/mice ( Figure 17B). We also noted that the decrease in mRNA for Lxra in livers of HF30Fr-fed mice treated with TBE-31 was not associated with an obvious reciprocal increase in mRNA for Fxr and Shp in either Nrf2 þ/þ or Nrf2 -/genotypes (data not shown), suggesting that the Nrf2-dependent suppression of Lxra does not involve Fxr or Shp. Surprisingly, TBE-31 decreased mRNA for Chrebp, encoded by Mlxipl, in livers of both HF30Fr-fed Nrf2 þ/þ and Nrf2 -/mice ( Figure 17A), a result that indicates it can antagonize Chrebp through a mechanism that does not involve Nrf2. Consistent with the qRT-PCR results for Srebf1, Lxra, and Xbp1s, TBE-31 did not downregulate mRNA for the fatty acid and triglyceride synthesis enzymes Acaca, Fasn, and Dgat2 in livers of HF30Fr-fed Nrf2 -/mice ( Figure 17C). Although the HF30Fr diet increased mRNA for the lipid importer Cd36 in livers of Nrf2 þ/þ and Nrf2 -/mice, TBE-31 had no additional effect on Cd36 expression in livers of either Nrf2 þ/þ or Nrf2 -/mice ( Figure 17D). Lastly, although TBE-31 increased mRNA levels for Mttp and ApoB, which are involved lipoprotein assembly, in livers of Nrf2 þ/þ mice, it did not increase mRNA for these proteins in livers of Nrf2 -/mice ( Figure 17D).
Immunoblotting suggested that at least 2 arms of the UPR were activated in livers of RC-fed Nrf2 -/mice when compared with livers of RC-fed Nrf2 þ/þ mice, as evidenced by increases in Xbp1s (downstream of Ire1a), and Atf4 (downstream of Perk) (Figure 18). This also revealed livers from HF30Fr-fed Nrf2 þ/þ mice contained increased levels of the ER marker proteins p-Ire1a, Xbp1s, p58 IPK , p-eIf2a, and Atf4 that was attenuated by treatment with TBE-31. By contrast, TBE-31 did not decrease these ER marker proteins in livers from HF30Fr-fed Nrf2 -/mice.
Immunoblotting revealed that the HF30Fr diet stimulated nuclear accumulation of Nfkb p65 and p50 in livers of Nrf2 þ/þ and Nrf2 -/mice, and that although TBE-31 could attenuate this in livers of wild-type mice it was unable to do so in Nrf2 -/livers ( Figure 19A). qRT-PCR of mRNAs for Cox2 and Nos2 indicated that the HF30Fr diet stimulated expression of proinflammatory genes in the livers of Nrf2 þ/þ and Nrf2 -/mice but, whereas TBE-31 blunted this increase in wild-type mice, TBE-31 was unable to do so in the mutant mice ( Figure 19B).
As anticipated, the inability of TBE-31 to suppress NASH in Nrf2 -/mice was associated with higher levels of hepatic malondialdehyde, oxidized protein, and GSSG than in Nrf2 þ/þ mice ( Figure 19C-E) suggesting that TBE-31 did not suppress oxidative stress caused by the HF30Fr diet in Nrf2 -/mice.

Discussion
We have examined whether pharmacologic activation of Nrf2 using the acetylenic tricyclic bis(cyano enone) compound TBE-31 can suppress NASH in mice in which diet-stimulated metabolic disease had already been established, reflecting the situation encountered in clinical practice. The experiments described herein reveal that treatment of mice with TBE-31 conferred multiple metabolic benefits insofar as it increased whole-body insulin Data are means ± SEM. Significant differences are represented as: * ,$ P < .05; ** ,$$ P < .01; *** ,$$$ P < .001. Figure 14. TBE-31 fails to improve insulin sensitivity in HFFr-fed Nrf2 -/mice. During Study 2, physiological end-points and glucose homeostasis were examined in wild-type and Nrf2-null mice fed an RC or HF30Fr (HFFr) diet (n ¼ 6-8 mice per group). (A) Body weight gain of mice up until intervention at end of Week 10 (left, vertical striped bars, RC; diagonal striped bars, HFFr) and following (Weeks 11-16) of treatment (right). (B) Glucose production (pyruvate tolerance) with AUC in Nrf2 þ/þ (squares) and Nrf2 -/-(triangles) mice after 10 weeks HF30Fr diet, followed by 5 weeks treatment with DMSO (white squares and triangles) or TBE-31 (black squares and triangles). (C) Insulin sensitivity (% change in blood glucose) in Nrf2 þ/þ (squares) and Nrf2 -/-(triangles) mice after 10 weeks HF30Fr diet followed by 4 weeks with DMSO (white squares and triangles) or TBE-31 (black squares and triangles). White bars, DMSO; black bars, TBE-31. Data are means ± SEM. Significant changes: *P < .05; ** ,$$ P < .01; *** ,$$$ P < .001. D-F, Scale bars ¼ 100 mm. AUC, area under the curve, sensitivity and improved glucose homeostasis (better glucose disposal and reduced gluconeogenesis). Moreover, within the liver, TBE-31 markedly diminished steatosis and inflammation (ie, NASH, as evidenced by the NAS calculations) and fibrosis (which is not included in NAS calculations). Most importantly, our data show that TBE-31 requires the presence of Nrf2 to produce these beneficial pleiotropic effects.

Influence of Nrf2 Activation on Glucose Homeostasis
It is known that Nrf2 -/mice are more sensitive than Nrf2 þ/þ mice to streptozotocin-induced diabetes, and that pharmacologic activation of Nrf2 with sulforaphane (SFN) can inhibit the development of diabetes, 46,47 but it is not known whether activation of Nrf2 in wild-type mice after Results are means ± SEM. Significant increases in NAFLD activity score, relative to that in livers from RC-fed DMSO-treated Nrf2 þ/þ mice, are indicated by: ***P < .001. The significant decrease in NAFLD activity score resulting from treatment with TBE-31, relative to HFFr-fed DMSO-treated Nrf2 þ/þ mice, is denoted by: $ P < .05. diabetes has been induced by physiological stimuli (ie, chronic overnutrition) can mitigate disease. Uniquely, in the present study, we have demonstrated that pharmacologic activation of Nrf2 can reverse type 2 diabetes by performing metabolic analyses before and after 4À5 weeks therapeutic intervention with TBE-31 in adult Nrf2 þ/þ mice fed a HFFr diet, thereby providing clear evidence that the acetylenic tricyclic bis(cyano enone) can improve glucose disposal and insulin sensitivity. Our finding that TBE-31 treatment reverses insulin resistance in mice that have been fed chronically an HFFr diet is consistent with a report that treatment of 6-week-old diabetic db/db mice with the Nrf2 activator CDDO-Im for 10 weeks increased glucose clearance and lowered plasma insulin levels during an oral GTT. 31 However, our results are at variance with a recent report that 4week treatment of adult C57BL/6J mice that had received a 60% HF diet for 10 weeks with the Nrf2 activator SFN improved glucose tolerance during an intraperitoneal GTT but did not improve insulin sensitivity. 48 In the present investigation, changes in glucose homeostasis affected by TBE-31 were monitored in Study 1 using ITT and GTT, and in Study 2 using ITT and PTT. None of these provide a reference method for defining insulin resistance, because this is only achieved using hyperinsulinemic euglycemic clamps. Therefore, further work is required to determine whether the difference in the ability of TBE-31 and SFN to improve insulin sensitivity is caused by methodological factors (eg, the type of diet, duration of feeding, potency, or pharmacokinetic properties of Nrf2 activators).
The fact that Nrf2 -/mice fed a HFFr diet became insulin resistant (as assessed by ITT) is surprising because we have previously reported that the knockout mice fed an HF diet exhibit better insulin sensitivity than HF-fed Nrf2 þ/þ mice. 24 This outcome, along with worsened NASH, displayed by HFFr-fed Nrf2 -/mice, suggests Nrf2 plays a previously unrecognized role in protection against fructose-driven metabolic disorder. The reasons for this are not known. Possibly overconsumption of fructose may cause insulin resistance by increasing hepatic diacylglycerol levels and activating protein kinase Cε. 49 Clearly, this issue warrants further investigation.

Activation of Nrf2 Attenuates Hepatic Lipid Metabolism
In the present study, Nrf2-mediated inhibition of HFFrstimulated liver steatosis by TBE-31 involves positive and negative mechanisms. On the one hand, TBE-31 increased, in a Nrf2-dependent manner, the abundance of mRNA for proteins that limit lipid accumulation in the liver, such as those involved in fatty acid oxidation (eg, Acox2, Ces1g, Cpt1a, and Scad) and export of triglycerides (eg, Mttp and ApoB). On the other hand, TBE-31 decreased, in a Nrf2dependent manner, mRNA for Srebp-1c, Xbp1s, and Lxra, transcription factors that orchestrate de novo lipogenesis and have been implicated in hepatic steatosis. 49,50 Among the previously mentioned lipogenic transcription factors, Srebp-1c and Xbp1s are integrated into the UPR and are activated by ER stress. 42 Because evidence suggests Nrf2 marshals the expression of genes that collectively antagonize ER stress, 24 we propose that activation of Nrf2 by TBE-31 blunts ER stress caused by the HFFr diet, and this in turn decreases the activity of Srebp-1c and Xbp1s, along with expression of their target genes (eg, Acaca, Acly, Dgat2, Fasn, and Scd1). Our hypothesis that TBE-31 protects Ballooning was seen only in the liver of 1 Nrf2 -/mouse that was fed the HFFr diet. e Livers from mice fed the HFFr diet showed more fibrosis (this is not included in NAS calculation) when compared with mice fed the RC diet in both genotypes. Nrf2 þ/þ mice fed the HFFr diet exhibited significantly less hepatic fibrosis when treated with TBE-31 than did Nrf2 þ/þ mice treated with DMSO (Kruskal-Wallis H test; P < .05). Fibrosis in livers of Nrf2 -/mice fed the HFFr diet was significantly higher than in livers of their wild-type counterparts (Kruskal-Wallis H test; P < .05). TBE-31 treatment had no effect on liver fibrosis in Nrf2 -/mice. against excessive hepatic steatosis, at least in part, by blunting ER stress, could be likened to protection against the toxic effect of the ER stressor tunicamycin by the Nrf2 activator 3H-1,2-dithiole-3-thione, 51 or protection conferred by knockdown of Keap1 against the ER stress-mediated apoptotic effects of alkylating agents. 52 Besides downregulating Srebp-1c and Xbp1s, we found pharmacologic activation of Nrf2 is associated with a decrease in abundance of mRNA for Lxra. However, by contrast with Srebp-1c and Xbp1s, Lxra is not activated by ER stress, 43 and so some other mechanism must be involved. It is possible that the ability of Nrf2 to Livers were collected from Nrf2 +/+ and Nrf2 -/mice at the end of Study 2, and proteins and genes involved in the UPR were examined. Representative immunoblots of p-Ire1a, Xbp1s and Xbp1u, p58 IPK , p-eIf2a and Atf4 proteins, along with actin as a loading control, are shown at the top. Plots of densitometric scans from the blots are shown at the bottom (n ¼ 4 biologic replicates). White bars, DMSO treated; black bars, TBE-31 treated. Results are means ± SEM. Significant increases relative to that found in livers of RCfed Nrf2 +/+ mice are indicated by: *P < .05; **P < .01; ***P < .001. Significant decreases relative to HF30Fr-fed Nrf2 +/+ mice are indicated: $ P < .05; $$ P < .01. Significant increases relative to that found in livers of RC-fed DMSO-treated Nrf2 +/+ mice are indicated by: *P < .05; **P < .01; ***P < .001. Significant decreases relative to HF30Fr-fed DMSO-treated Nrf2 +/+ mice are indicated: $ P < .05; $$ P < .01. downregulate the expression of Lxra is a consequence of Nrf2 decreasing the levels of endogenous oxysteroids: this reasoning is based on the fact that Lxra is a nuclear receptor that is activated by oxysteroid ligands, and that the human LXRa gene autoregulates itself because it contains an LXR response element (LXRE) in its promoter region 53,54 ; it is, however, unclear if this mechanism is relevant to the mouse because there is controversy about whether the putative LXRE in the promoter region of mouse Lxra is functional. 55,56 An additional possibility is that activation of Nrf2 downregulates expression of Lxra because it decreases the hepatic levels of thyroid hormone: this suggestion is based on the report that triiodothyronine induces expression of Lxra in mouse liver via the thyroid hormone receptor. 56 Another explanation is that activation of Nrf2 antagonizes Lxra activity, and hence its expression, by facilitating inhibitory phosphorylation of Lxra at one of several Thr residues by adenosine monophosphate-activated protein kinase. 57 An alternative possibility proposed by Kay et al 44 is that activation of Nrf2 downregulates Lxra expression because it leads to induction and deacetylation of Fxr that results in induction of Shp protein, which in turn blunts expression of the Lxra gene by forming a nonproductive heterodimer with Lxra. However, this hypothesis was based on experiments using SFN to activate Nrf2, and we think this is unlikely in the case of TBE-31 because although expression of Shp was lower in livers of Nrf2 -/mice than those of Nrf2 þ/þ mice, the expression of Fxr did not differ in livers from either genotype, nor was the mRNA for Shp or Fxr inducible by TBE-31 in livers of Nrf2 þ/þ mice. Clearly, Figure 19. (See previous page). TBE-31 fails to suppress inflammation and oxidative stress in HFFr-fed Nrf2 -/mouse liver. The extent of inflammation and oxidative stress in livers of Nrf2 þ/þ and Nrf2 -/mice at the end of Study 2 was examined. (A) Representative immunoblots of hepatic nuclear levels of Nfkb p65 and Nfkb p50 (using proliferating cell nuclear antigen as a loading control), along with densitometric scans (n ¼ 4 biologic replicates), from Nrf2 þ/þ and Nrf2 -/mice fed RC or HFFr diets and treated with DMSO or TBE-31. (B) qRT-PCR for the Nfkb-target genes Cox2 and Nos2 from the same livers (6-8 mice per group). (C) Malondialedehyde levels in livers from Nrf2 þ/þ and Nrf2 -/mice (6-8 mice per group). (D) Oxidized protein levels in livers of Nrf2 þ/þ and Nrf2 -/mice shown as a representative Oxyblot, with densitometric quantification below: lanes 1, 3, 5, and 7 negative controls; 2, RC-fed DMSO; 4, RC-fed TBE-31; 6, HF30Fr-fed DMSO; 8, HF30Fr-fed TBE-31 (n ¼ 4). (E) Ratio of GSH to GSSG (6-8 mice per group). In all cases, white bars represent DMSO and black bars represent TBE-31. Data are means ± SEM. Significant differences are signified: * ,$ P < .05; ** ,$$ P < .01; *** ,$$$ P < .001. further work is required to determine the mechanisms by which Nrf2 represses Lxra.
It remains unclear how TBE-31 antagonizes the expression of Chrebp. Most importantly, we found TBE-31 attenuated the expression of Chrebp in livers of Nrf2 -/mice, indicating that the acetylenic tricyclic bis(cyano enone) exerts Nrf2-independent effects. This action of TBE-31 possibly reflects changes in insulin sensitivity because the abundance of Chrebp mRNA responds dramatically to changes in insulin levels in human differentiated preadipocytes. 58 Further work is required to address this issue.

Influence of Nrf2 Activation on Hepatic Inflammation and Nonalcoholic Steatohepatitis
We found Nrf2-mediated suppression of liver inflammation during the development of NASH involves blunting of activation of AP-1 and Nfkb. It is unclear in our murine model whether ER stress or oxidative stress is more influential in stimulating activation of AP-1 and Nfkb by the HFFr diet, but because both forms of stress are interrelated 59,60 it is plausible that pharmacologic activation of Nrf2 by TBE-31 diminishes activation of AP-1 and Nfkb by countering ER and oxidative stress. TBE-31 decreased Il-1b mRNA levels, in line with the proposal that Nrf2 inhibits transcription of the Il-1b gene by binding to the promoter and preventing recruitment of RNA pol II. 61 Therefore, it is likely TBE-31 increases the effectiveness of this mode of repression of inflammation. Furthermore, TBE-31 likely inhibits recruitment of monocytes and neutrophils into the liver because it suppressed mRNA levels for Mcp1, and for elastase and myeloperoxidase.
The demonstration that activation of Nrf2 by TBE-31 ameliorates NASH is consistent with several earlier reports in which pharmacologic activators of Nrf2 including baicalein, 62 SFN, 63 or Ezetimibe, 64 or genetic activation of Nrf2 by expression of a hypomorphic Keap1 allele 65 or hepatocyte-specific knockout of Keap1, 66 inhibit NASH in rodents caused by an MCD diet or an HF diet. In all these previous experiments, however, the inducing agents were coadministered with the MCD/HF diet, or the genetic manipulation was constitutive, and therefore they do not address the question of whether Nrf2 can ameliorate established disease. Thus, our present findings add substantially to those in the literature by showing TBE-31Àtreatment can reverse insulin resistance and, by implication, also NASH.

Influence of Nrf2 Activation on Liver Fibrosis
Our discovery that TBE-31 can inhibit liver fibrosis in HFFr-fed mice with NASH in an Nrf2-dependent manner is particularly noteworthy. Specifically, we found that histologic evidence of fibrosis and expression of Col1a1 and a-Sma were substantially diminished in HFFr-fed wild-type mice treated with TBE-31, when compared with DMSO-treated mice. These results are in good agreement with a previous study in rats in which the Nrf2 activators oltipraz and NK-252 attenuated progression of NASH-related fibrosis stimulated by a choline-deficient L-amino acid-defined diet, 67 but whether or not the antifibrosis effects of oltipraz and NK-252 were mediated by Nrf2 was not addressed.

Influence of Endoplasmic Reticulum Stress and Cirrhosis on Nrf2
Wu et al 68 have demonstrated that during end-stage liver cirrhosis in mice, hepatic Nrf2 is suppressed as a consequence of activation of the Ire1a-Xbp1 arm of the UPR, thereby exacerbating oxidative stress and promoting further disease. Specifically, these workers reported that experimental liver cirrhosis, caused by CCl 4 , resulted in increases in Xbp1s and its downstream target Hrd1, which in turn stimulated ubiquitylation and degradation of Nrf2, thereby decreasing expression of Nqo1 and Gclm. In the present study, we discovered chronic feeding of Nrf2 þ/þ mice with the HFFr diet stimulated ER stress in the liver, with clear evidence of activation of the Xbp1 arm ( Figure 10A and B). However, HFFr-feeding had a mixed effect on expression of hepatic Nrf2-target genes insofar as the diet increased expression of Gclc, Gclm, Hmox1, and Slc7a11, while decreasing the expression of Txnrd1, Cat, and Prdx6 ( Figure 13D). Importantly, we found that treatment of HFFrfed mice with TBE-31 increased the expression of all these Nrf2-target genes. Collectively, our findings suggest that stimulation of ER stress by chronic consumption of an HFFr diet is not sufficient to cause Hrd1-mediated suppression of Nrf2 that cannot be overcome by administration of TBE-31. Further work is required to define at what stage during the progression of NASH through to cirrhosis is Nrf2 suppressed by Hrd1, and what additional environmental and/or biochemical factors, besides just ER stress, are required.

Conclusions
This investigation shows TBE-31 provides an effective therapeutic strategy to treat NASH through the ability of Nrf2 to direct changes in gene expression that antagonize lipogenesis, ER stress, inflammation, oxidative stress, and fibrosis (summarized in Figure 20). Work is now required to establish whether pharmacologic activation of Nrf2 in the human can improve metabolic syndrome, NASH, and liver fibrosis. Such clinical studies are entirely feasible because Nrf2 activators have already been demonstrated to improve diabetes-related biomarkers. 48