Hepatocyte‐specific sirtuin 6 deletion predisposes to nonalcoholic steatohepatitis by up‐regulation of Bach1, an Nrf2 repressor

ABSTRACT

Sirtuin (Sirt)6 has been implicated in negative regulation of inflammation and lipid metabolism, although its function in the progression from simple steatosis to nonalcoholic steatohepatitis (NASH) remains to be defined. To explore the role of hepatocyte Sirt6 in NASH development, we generated hepatocyte‐specific Sirt6‐knockout (KO) mice that were fed a high‐fat and high‐fructose (HFHF) diet for 16 wk. HFHF‐fed KO mice had increased hepatic steatosis and inflammation and aggravated glucose intolerance and insulin resistance compared with wild‐type mice. HFHF‐induced liver fibrosis and oxidative stress and related gene expression were significantly elevated in KO mice. In the livers of KO mice, nuclear factor erythroid 2‐related factor 2 (Nrf2) was down‐regulated; conversely, BTB domain and CNC homolog 1 (Bach1), a nuclear repressor of Nrf2, were up‐regulated. We discovered that Sirt6, which interacts with Bach1 under basal condition, induces its detachment from the antioxidant response element (ARE) region of heme oxygenase 1 promoter. Furthermore, we found that Sirt6 promotes Nrf2 binding to ARE in response to oxidative stimuli, which leads to the expression of phase II/antioxidant enzymes. Finally, we showed that HFHF‐induced steatosis, inflammation, and fibrosis were ameliorated by adenoviral Sirt6 overexpression. Sirt6 may be a useful therapeutic target for amelioration of NASH by curbing inflammation and oxidative stress.—Ka, S.‐O, Bang, I. H., Bae, E. J., Park, B.‐H. Hepatocyte‐specific sirtuin 6 deletion predisposes to nonalcoholic steatohepatitis by up‐regulation of Bach1, an Nrf2 repressor. FASEB J. 31, 3999–4010 (2017). www.fasebj.org—Ka, Sun‐O, Bang, In Hyuk, Bae, Eun Ju, Park, Byung‐Hyun Hepatocyte‐specific sirtuin 6 deletion predisposes to nonalcoholic steatohepatitis by up‐regulation of Bach1, an Nrf2 repressor. FASEB J. 31, 3999–4010 (2017)

ABBREVIATIONS

AdLacZ

adenovirus expressing β‐galactosidase

AdSirt6

adenovirus expressing sirtuin 6

ALT

alanine aminotransferase

ARE

antioxidant response element

AST

aspartate aminotransferase

Bach1

BTB and CNC homolog 1

ChIP

chromatin immunoprecipitation

co‐IP

coimmunoprecipitation

DHE

dihydroethidium

GSH

glutathione

HFD

high‐fat diet

HFHF

high‐fat and high‐fructose

HO‐1

heme oxygenase 1

Keap1

Kelch‐like ECH‐associated protein 1

KO

knockout

MCP‐1

monocyte chemoattractant protein‐1

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

NCD

normal chow diet

Nrf2

nuclear factor erythroid 2–related factor 2

qPCR

quantitative PCR

ROS

reactive oxygen species

SFN

sulforaphane

Sirt

sirtuin

shRNA

short hairpin RNA

siRNA

small interfering RNA

SOD

superoxide dismutase

TG

triglyceride

WT

wild type

Nonalcoholic fatty liver disease (NAFLD) is a condition with symptoms ranging from simple triglyceride (TG) accumulation in the liver (steatosis) to nonalcoholic steatohepatitis (NASH) (1). The latter is characterized by steatosis, inflammation, and fibrosis. Although simple steatosis alone is relatively benign, NASH greatly increases the risk of cirrhosis and even hepatocellular carcinoma (2, 3). Several factors have been shown to be associated with NASH. These include age, obesity, insulin resistance, dyslipidemia, and hypertension (4–6). Although the molecular mechanism underlying disease progression from simple steatosis to NASH has not been fully elucidated, accumulating data suggest that altered redox balance due to oxidative stress plays a crucial role in the pathogenesis of NAFLD, NASH, and fibrosis (7). Because patients over 50 yr of age are more likely to have steatosis and NASH (5, 8), there could be a mechanistic link between NASH and the aging process. Sirtuins are good examples of such molecules, among which sirtuin (Sirt)1 has been paid great attention because it increases expression of antioxidant proteins and decreases inflammation through direct deacetylation of related proteins (9). Indeed, specific ablation of hepatic Sirt1 in mice decreases fatty acid β‐oxidation, increases expression of lipogenic enzymes and reactive oxygen species (ROS) production, and ultimately leads to aggravation of hepatic steatosis and inflammation in response to high‐fat diet (HFD) feeding (10, 11). Conversely, overexpression or activation of Sirt1 protects against NAFLD/NASH in animals and humans (12, 13).

Sirt6, a nuclear form of sirtuin, preferentially deacetylates histone H3 lysine 9 (H3K9) and lysine 56 (H3K56), and represses the transcriptional activities of several transcription factors such as NF‐κB, c‐JUN, and hypoxia‐inducible factor‐1α (14). For example, Sirt6 controls inflammation by deacetylating H3K9 on the promoters of NF‐κB target genes to decrease promoter occupancy by the p65 subunit of NF‐κB (14). Accordingly, overexpression of Sirt6 suppressed collagen‐induced arthritis in mice (15) and cytokine production in vitro (16), whereas the lack of Sirt6 caused the chronic liver inflammation (17) and impaired wound healing (18). In addition to histone H3, Sirt6 directly deacetylates the C‐terminal binding protein interacting protein (19), acetyltransferase GCN5 (20), forkhead box protein O1 (21), and GATA binding protein 3 (22).

To date, unlike Sirt1, the role of Sirt6 in NASH has not been reported. To delineate the role of Sirt6 in the development of NASH, we investigated the effect of hepatic Sirt6 deletion in mice after feeding a high‐fat and high‐fructose (HFHF) diet for 16 wk. We also investigated the potential benefit of Sirt6 overexpression on inflammatory and fibrotic responses during HFHF feeding to mice. After observing significant effects in these mice models, we could conclude that Sirt6 is a critical molecule for the reversal of oxidative stress and disease progression to NASH. In particular, Sirt6 activates a transcription factor, nuclear factor erythroid 2‐related factor 2 (Nrf2), to increase phase II/antioxidant enzymes, and this finding is accompanied by inhibition of BTB domain and CNC homolog 1 (Bach1), a functional repressor of Nrf2.

MATERIALS AND METHODS

Animals

Sirt6^flox/flox mice (B6;129‐Sirt6^tm1Ygu/J) and albumin‐Cre mice [B6. Cg‐Tg(alb‐Cre)21Mgn/J] were obtained from The Jackson Laboratory (Bar Harbor, ME, USA). Sirt6^flox/flox and homozygous albumin‐Cre mice were crossed to obtain hepatocyte‐specific Sirt6‐knockout (KO) mice. KO mice and age‐matched wild‐type (WT) littermates older than 4 wk were fed ad libitum either a standard laboratory normal chow diet (NCD) or a 60% fat and 30% fructose HFHF diet (Research Diet, New Brunswick, NJ, USA) for 16 wk. Oral glucose tolerance tests (1 g/kg of body weight) and insulin tolerance tests (0.75 U/kg of body weight) were performed after 16 and 6h of food withdrawal, respectively. All experimental procedures were approved by the Institutional Animal Care and Use Committee of Chonbuk National University (CBNU‐2015‐0085).

Human tissues

Human liver tissues were obtained from Chonbuk National University Hospital Biobank. Informed consent was obtained from the patients, and the study was approved by the Institutional Review Board of Chonbuk National University (T14‐18).

Histology

Tissues were removed and immediately placed in fixative (10% formalin solution in 0.1 M PBS). Histologic sections (4 μm) were cut from formalin‐fixed, paraffin‐embedded tissue blocks. Tissue sections were stained with hematoxylin‐eosin under standard conditions. Staining of fibrillar collagen with Sirius red (saturated picric acid containing 0.1% DirectRed 80) was performed on paraffin sections. Stained sections were quantified by iSolution DT 36 software (Carl Zeiss, Oberkochen, Germany), and results were expressed as percentage of area. Immunohistochemical staining was performed using the Dako Envision system (Dako, Carpinteria, CA, USA). After deparaffinization, tissue sections were immunostained with antibodies against anti‐perilipin2 (Progen Biotechnik, Heidelberg, Germany) and anti‐F4/80 or anti‐Sirt6 (Abcam, Cambridge, United Kingdom). Peroxidase activity was detected with 3‐amino‐9‐ethyl carbazole. Liver inflammation in liver biopsies was graded using a modified histologic activity index (23). Briefly, liver inflammation was defined as high grade if mice had more than 4 foci in a ×200 field (inflammation score 3) or 2–4 foci in a ×200 field (inflammation score 2). Inflammation was defined as low grade if mice had no foci (inflammation score 0) or fewer than 2 foci in a ×200 field (inflammation score 1). For each animal, 3–5 areas in 4 different sections each were analyzed. The data were first averaged per section and then per animal. Superoxide anion (O₂⁻) levels were quantified in frozen sections (10 μm) with the oxidative fluorescent dye dihydroethidium (DHE) (Molecular Probes, Eugene, OR, USA).

Preparation of recombinant adenovirus

An adenovirus expressing Sirt6 (AdSirt6) or β‐galactosidase (AdLacZ) was prepared as previously described (15).

Biochemical analysis

TNF‐α (Thermo Fisher Scientific, Waltham, MA, USA), insulin (Millipore, Billerica, MA, USA), and monocyte chemoattractant protein‐1 (MCP‐1) (Peprotech, Rocky Hill, NJ, USA) were measured using specific ELISA kits. Plasma levels of total cholesterol, TG, nonesterified fatty acid, aspartate aminotransferase (AST), alanine aminotransferase (ALT) (Asan Pharmaceutical, Seoul, South Korea), and lactate dehydrogenase (Biovision, Milpitas, CA, USA) were measured using commercially available kits. To quantify liver TG, liver tissues were homogenized and extracted in a mixture of chloroform, methanol, and distilled water (2/1/1 ratio). The TG concentration was measured with a TG Assay Kit (Asan Pharmaceutical) and expressed as milligrams of TG per 100 mg of liver tissue.

Isolation of primary hepatocytes and cell culture

Primary hepatocytes were prepared from 6‐ to 8‐wk‐old KO mice and WT littermates by perfusion with collagenase type IV (Sigma‐Aldrich, St. Louis, MO, USA) as previously described (24). To express exogenous proteins for promoter assay, HepG2 cells obtained from American Type Culture Collection (Manassas, VA, USA) were transfected with 2 μg of pFlag or pFlag‐Sirt6 with 0.5 μg of antioxidant response element (ARE)‐promoter and 20 ng of pRL‐TK (Promega, Madison, WI, USA) using the Lipofectamin 2000 (Thermo Fisher Scientific). Luciferase activity was measured using a Dual Luciferase Reporter assay (Promega) by Lumat LB 9507 (Berthold, Bad Wildbad, Germany).

Knockdown of Sirt6 in HepG2 cells

HepG2 cells were transduced with lentiviral particles containing either control short hairpin RNA (shRNA) (sc‐108080; Santa Cruz Biotechnology, Dallas, TX, USA) or Sirt6 shRNA (sc‐63023‐V; Santa Cruz Biotechnology) in the presence of 8 μg/ml polybrene (Sigma‐Aldrich). For selection of cells with Sirt6 shRNA, cells were grown in a medium with 5 μg/ml puromycin (Sigma‐Aldrich). Isolated individual colonies were kept for up to 3 wk, and Sirt6 knockdown was verified in each colony using Western blotting.

RNA interference

Duplexes of small interfering RNA (siRNA) targeting mouse Bach1 mRNA (target sequences (si‐Bach1‐1: 5′‐GAACAUUA‐CUCUUCCAGAA‐3′ for sense and 5′‐UUCUGGAAGA‐GUAAUGUUC‐3′ for antisense; si‐Bach1‐2: 5′‐GCAGAUGAC‐UGAUAAAUGU‐3′ for sense and 5′‐ACAUUUAUCAGU‐CAUCUGC‐3′ for antisense) and a negative control (scrambled sequence) were purchased from Bioneer (Daejeon, South Korea). HepG2 cells (2 × 10⁶) were transfected with 10 nM of siRNA oligonucleotides using Lipofectamine 2000 (Thermo Fisher Scientific). siRNA‐transfected cells were incubated for 24 h at 37°C before drug treatment.

Chromatin immunoprecipitation assay

HepG2 cells were cross‐linked by incubating cells in 1% formaldehyde for 10 min at room temperature. Cross‐linking was stopped by 5 min of incubation with 125 mM glycine. Cells were lysed with cytosolic lysis buffer [5 mM KOH (pH 8.0), 85 mM KCl, 0.5% NP‐40 with protease inhibitor cocktail (Calbiochem, San Diego, CA, USA), and 1 mM PMSF] on ice. Lysates were centrifuged, and pellets were resuspended in nuclei lysis buffer [50 mM Tris (pH 8.1), 10 mM EDTA, 1% SDS, and protease inhibitor cocktail] on ice. Chromatin was sonicated to 300‐ to 1000‐bp fragments at 4°C. Samples were diluted to 5‐fold in chromatin immunoprecipitation (ChIP) buffer [0.01% SDS, 1.1% Triton X‐100, 1.2 mM EDTA, 16.7 mM Tris (pH 8.1), 167 mM NaCl, and protease inhibitor cocktail]. Samples were precleared with protein A/G agarose beads and salmon sperm DNA for 30 min at 4°C. After centrifugation, each sample was immunoprecipitated with 5 μg of anti‐Nrf2, anti‐Bach1 (Santa Cruz Biotechnology), anti‐Sirt6, H3, or nonspecific IgG (Cell Signaling Technology, Beverly, MA, USA) overnight at 4°C and then with protein G agarose at 4°C for 2 h. Beads were washed sequentially with low‐salt wash buffer, high‐salt wash buffer, and LiCl wash buffer and twice with TE buffer. DNA‐protein complexes were eluted from protein G agarose beads with elution buffer (1% SDS in 0.1 M NaHCO₃) on a 1200 rpm vortex mixer for 30 min. Cross‐linking was reversed at 65°C for 6 h with RNase and 0.3 M NaCl, and 1.25 ml 100% ethanol was added. The solution was stored overnight at −80°C. DNA was purified using QiaQuick spin columns (Qiagen, Hilden, Germany). Primers used for ChIP‐quantitative PCR (qPCR) analyses were heme oxygenase 1 (HO‐1) promoter primer (forward: 5′‐AGGATAAGGGACACCGGTCAC‐3′;reverse:3′‐CAGAGTTT‐CCGCATACAACCA‐5′), HO‐1 enhancer 1 (forward: 5′‐TGAAG‐TTAAAGCCGTTCCGG‐3′; reverse: 3′‐AGCGGCTGGAATGCT‐GAGT‐5′), HO‐1 enhancer 2 (forward: 5′‐GGGCTAGCATGC‐GAAGTGAG‐3′; reverse: 3′‐AGACTCCGCCCTAAGGGTTC‐5′), and nonspecific binding site from GAPDH primer (forward: 5′‐AATGAAGGGGTCATTGATGG‐3′; reverse: 3′‐AAGGT‐GAAGGTCGGAGTCAA‐5′).

Western blot and coimmunoprecipitation

Liver homogenates or cell lysates (20 μg) were separated by 10% SDS‐PAGE and transferred to PVDF membranes. After blocking with 5% skim milk, the blot was probed with primary antibodies against Sirt6, H3K9 (Cell Signaling Technology), H3K56 (Active Motif, Carlsbad, CA, USA), H3K18, MafK (Abcam), Nrf2, Kelch‐like ECH‐associated protein 1 (Keap1), lamin B, Bach1 (Santa Cruz Biotechnology), HSP90, HO‐1, NQO1 (Enzo Life Sciences, Plymouth Meeting, PA, USA), or β‐actin (Sigma‐Aldrich). For coimmunoprecipitation (co‐IP), 500 μg of nuclear protein precleared with protein G agarose was incubated with anti‐Sirt6, anti‐Nrf2, or anti‐Bach1 overnight at 4°C and then with protein G agarose at 4°C for 2 h. Blots were probed with primary antibody against Sirt6, Nrf2, MafK, or Bach1, and signals were detected with a Las‐4000 imager (GE Healthcare Life Science, Pittsburgh, PA, USA).

RNA isolation and real‐time RT‐qPCR

Total RNA was extracted from frozen liver tissue using an RNA Iso kit (TaKaRa, Tokyo, Japan). First‐strand cDNA was generated using the random hexamer primer provided in the first‐strand cDNA synthesis kit (Applied Biosystems, Foster City, CA, USA). Specific primers for each gene (Table 1) were designed using qPrimerDepot (http://mouseprimerdepot.nci.nih.gov). qPCR reactions comprised a final volume of 10 μl, containing 10 ng of reverse‐transcribed total RNA, 200 nM of forward and reverse primers and PCR master mixture. qPCR was performed in 384‐well plates using an ABI Prism 7900HT Sequence Detection System (Applied Biosystems).

Statistical analysis

Data are expressed as means ± sem. Statistical comparisons were made using 1‐way ANOVA followed by Fisher’s post hoc analysis. The significance of differences between groups was determined using an unpaired Student’s t test. A value of P < 0.05 was considered significant.

RESULTS

Sirt6 expression is decreased in murine models of NASH and human fibrotic liver tissues

To determine if Sirt6 expression is changed by nutrient or inflammation status, we first analyzed hepatic Sirt6 expression in NAFLD and NASH models in mice. After feeding an HFD or HFHF diet for 16 wk to C57BL/6 mice, we found that hepatic Sirt6 levels in these mice were markedly decreased compared with levels in mice fed an NCD (Fig. 1A). Likewise, Sirt6 levels in liver tissues of patients with liver fibrosis were significantly decreased compared with levels in liver tissue of healthy subjects (Fig. 1B).

Hepatocyte‐specific Sirt6 deletion accelerates hepatic steatosis and insulin resistance

To gain insight into the influence of Sirt6 suppression on NASH development, we generated hepatocyte‐specific Sirt6‐KO mice. Western blotting and immunohistochemistry results confirmed complete deletion of Sirt6 in the livers of KO mice (Fig. 2A, B). KO mice and their WT littermates were fed either NCD or HFHF diet for 16 wk. On an NCD, there were no genotype differences in body weight gain, food and water intake, and liver weight, with the exception of increased hepatic TG content in KO mice (Fig. 2C, D and Supplemental Fig. S1). When fed an HFHF diet, hepatic steatosis was enhanced in KO mice, as identified by gross examination of the liver and increases in liver mass, liver weight/body weight, and hepatic and plasma levels of TG and cholesterol (Fig. 2D and Supplemental Fig. S1). Immunohistochemistry for hepatic lipid droplets with antibody against perilipin confirmed the results (Fig. 2E).

Table 1. Sequences and accession numbers for primers used in real‐time RT‐PCR

	Primer, 5′‐3′
Gene	Forward	Reverse	Accession no.
α‐SMA	ACCAACTGGGACGACATGGAA	TGTCAGCAGTGTCGGATGCTC	NM_007392
Col1a1	TAGGCCATTGTGTATGCAGC	ACATGTTCAGCTTTGTGGACC	NM_007742
Pdgfb	GAAGATCATCAAAGGAGCGG	CCTTCCTCTCTGCTGCTACC	NM_011057
Tgfb1	GTGTGGAGCAACATGTGGAACTCTA	TTGGTTCAGCCACTGCCGTA	NM_011577
Timp1	AGGTGGTCTCGTTGATTCGT	GTAAGGCCTGTAGCTGTGCC	NM_011593
Fgf21	CTCCAGCAGCAGTTCTCTGA	CCTGGGTGTCAAAGCCTCTA	NM_020013
Nrf2	TGCCTCCAAAGGATGTCAAT	CCTCTGCTGCAAGTAGCCTC	NM_010902
HO‐1	TCAAGGCCTCAGACAAATCC	ACAACCAGTGAGTGGAGCCT	NM_010442
Nqo1	AATGGGCCAGTACAATCAGG	CCAGCCCTAAGGATCTCTCC	NM_008706
Gclc	GTCTCAAGAACATCGCCTCC	CTGCACATCTACCACGCAGT	NM_010295
Sod1	AAAATGAGGTCCTGCACTGG	ACCATCCACTTCGAGCAGAA	NM_011434
Keap1	GGCAGTGTGACAGGTTGAAG	GATCGGCTGCACTGAACTG	NM_016679
Rps	AATGAACCGAAGCACACCATAG	ATCAGAGAGTTGACCGCAGTTG	NM_012052
Saa1	GCGAGCCTACACTGACATGA	TTTTCTCAGCAGCCCAGACT	NM_009117
Icam1	AACAGTTCACCTGCACGGAC	GTCACCGTTGTGATCCCTG	NM_010493
Il1b	GGTCAAAGGTTTGGAAGCAG	TGTGAAATGCCACCTTTTGA	NM_008361
Tnfa	AGGGTCTGGGCCATAGAACT	CCACCACGCTCTTCTGTCTAC	NM_013693
F4/80	TTTCCTCGCCTGCTTCTTC	CCCCGTCTCTGTATTCAACC	NM_010130
Cd11c	CACTCAGTGACTGCCCAAAA	CCTCAAGACAGGACATCGCT	NM_021334
Ccl2	ATTGGGATCATCTTGCTGGT	CCTGCTGTTCACAGTTGCC	NM_011333
Ccr2	AGCACATGTGGTGAATCCAA	TGCCATCATAAAGGAGCCA	NM_009915
Cxcl10	GACGGTCCGCTGCAACTG	GCTTCCCTATGGCCCTCATT	NM_021274
Ccl5	CCACTTCTTCTCTGGGTTGG	GTGCCCACGTCAAGGAGTAT	NM_013653

We next measured blood glucose and insulin levels in mice. Basal concentrations of blood glucose and glucose‐stimulated insulin release were elevated in KO mice fed an HFHF diet (Fig. 3A, B). Consistently, glucose tolerance test and insulin tolerance test results indicated that HFHF‐fed KO mice developed glucose intolerance and insulin resistance vs. WT mice (Fig. 3C, D). Altogether, these data validate that loss of Sirt6 in liver aggravates hepatic steatosis and systemic insulin resistance.

Hepatocyte‐specific Sirt6 deletion exacerbates liver inflammation

To examine morphologic liver changes, we performed a histologic analysis. Hematoxylin and eosin staining revealed that HFHF feeding induced macrovesicular hepatic steatosis in WT mice, and this change was aggravated in KO mice. Immunohistochemistry for F4/80 showed that KO mice had greater accumulation of F4/80‐positive cells in the liver compared with WT mice under both NCD and HFHF diet (Fig. 4A). Biochemical parameters of liver injury, such as plasma levels of ALT and AST and lactate dehydrogenase activity, correlated well with the degree of inflammation (Fig. 4B). qPCR results also confirmed the increased accumulation of macrophages and inflammation in HFHF‐fed KO mouse liver, with increased expression of macrophage marker genes (Cd11c and Ccr2), cytokines/chemokines genes (Tnf‐α, Il‐1β, Ccl2, Saa1, Cxcl10, and Ccl5), and adhesion molecule (Icam‐1) (Fig. 4C). ELISA results showed that plasma TNF‐α and CCL‐2/MCP‐1 were increased in HFHF‐fed KO mice compared with WT mice (Fig. 4D).

Figure 1

Sirt6 expression in NASH. A) Hepatic expression of Sirt6 protein in murine models of NAFLD and NASH (n = 3). B) Sirt6 protein levels in human liver specimens from healthy subjects and patients with fibrosis (n = 4 or 5). Values are expressed as means ± sem. *P < 0.05, **P < 0.01 vs. NCD‐fed mice or healthy subjects.

Figure 2

Effect of Sirt6 deficiency on hepatic fat accumulation in mice fed an HFHF diet. A, B) Generation of hepatocyte‐specific Sirt6‐KO mice was confirmed by Western blotting (A) and immunohistochemistry (B). C) Body weight was monitored during 16 wk of NCD or HFHF feeding (n = 10). D) Photographs of representative mouse liver, liver weight, and TG amount were obtained after 16 wk of diet feeding (n = 8). E) Perilipin immunostaining for hepatic lipid droplets from WT and KO mice. Scale bars, 250 μm. Values are means ± sem. *P < 0.05, **P < 0.01 vs. HFHF‐fed WT; ^#P < 0.05, ^##P < 0.01 vs. NCD‐fed WT.

An HFHF diet elicits an exaggerated oxidative stress response and liver fibrosis in livers of Sirt6‐KO mice

NAFLD/NASH is associated with increased oxidative stress, and superoxide anion is the primary species of oxidative stress (25). To explore superoxide anion production in liver tissues with NASH, fresh liver sections were DHE stained for in situ production of superoxide anion. Fluorescence microscopy examination showed a marked increase in DHE fluorescence in Sirt6‐deleted livers in comparison with WT livers (Fig. 5A). Accordingly, superoxide dismutase (SOD) activity and glutathione (GSH) content were significantly reduced in Sirt6‐deleted liver tissues (Fig. 5B).

Sirius red staining was performed to quantify the accumulation of collagen fibers. The results showed that Sirius red–positive areas were significantly increased in HFHF‐fed KO mice (Fig. 5C). Expression of mRNAs for fibrogenic genes [α‐smooth muscle antigen (α‐SMA), collagen type 1A (Col1A), platelet‐derived growth factor‐β (Pdfgb), TGF‐β1 (Tgfb1), and tissue inhibitor of metalloproteinases 1 (Timp1)] was significantly increased in the livers of HFHF‐fed KO mice compared with WT mice (Fig. 5D).

One key regulator of cellular defense against oxidative stress is the transcription factor Nrf2, which binds to the ARE motif in DNA and promotes expression of genes whose protein products detoxify reactive oxidants (26). Alterations in ROS accumulation and liver fibrosis between genotypes led us to examine the expression of Nrf2 and related proteins. Western blotting shown in Fig. 5E indicated that Nrf2 levels in whole‐liver lysates and nuclear extracts were drastically suppressed in KO vs. WT mice in both the NCD and HFHF groups. In sharp contrast to Nrf2, the level of Bach1, a competitive Nrf2 repressor, was markedly increased in whole liver lysates of KO mice under both NCD and HFHF diet. Nuclear levels of Bach1 were also increased in KO mice regardless of diet but to a milder degree. The whole liver and nuclear expression of MafK, a heterodimerization partner of Nrf2, was decreased in KO mice, whereas the expression of Keap1, a cytoplasmic inhibitor of Nrf2, did not differ between genotypes. Consistent with the reduction in protein expression, the mRNA levels of Nrf2 and its target genes (e.g., HO‐1, Nqo1, Gclc, and Sod1) were decreased in KO mice (Fig. 5F).

Sirt6 activates the Nrf2‐ARE pathway by stimulating Nrf2 recruitment and by inhibiting Bach1 recruitment to ARE sites in the HO‐1 promoter

We next investigated the functional consequence of Nrf2 down‐regulation and the mechanism by which Nrf2 is controlled by Sirt6. Primary hepatocytes were isolated from WT and KO mice, and the functional identity of Sirt6 KO was confirmed by hyperacetylation of H3K9, H3K56, and H3K18 (Supplemental Fig. S2A). Consistent with expression in liver tissues, expression of Nrf2 was decreased, whereas expression of Bach1 was increased in KO hepatocytes (Supplemental Fig. S2B). When these cells were treated with the Nrf2 activator sulforaphane (SFN), nuclear translocation of Nrf2 was significantly decreased, whereas Bach1 level was increased in KO cells (Fig. 6A). Consistently, SFN stimulation of expression of Nrf2 target genes such as HO‐1 and NQO‐1 was also markedly suppressed in KO cells (Supplemental Fig. S2C, D). On the other hand, adenoviral overexpression of Sirt6 in HepG2 cells resulted in increased Nrf2 and decreased Bach1 in nuclear extracts (Supplemental Fig. S2E). Unlike in KO cells, Sirt6 overexpression was not able to change the nuclear level of MafK, suggesting the presence of a differential regulatory pathway for MafK. In addition, in HepG2 cells with Sirt6 knockdown by shRNA transfection, both tertiary butylhydroquinone‐ and SFN‐stimulated ARE promoter activity were repressed compared with control cells. On the contrary, overexpression of Sirt6, but not deacetylase inactive Sirt6‐H133Y (mtSirt6), increased basal‐ and SFN‐stimulated ARE promoter activity (Fig. 6B). These findings confirm the role of Sirt6, dependent on its deacetylase enzymatic activity, in the activation of the Nrf2‐ARE signaling pathway. Next, to determine whether binding of Nrf2 and Bach1 to ARE sites was changed by Sirt6 deletion or reexpression, we performed ChIP assays using HO‐1 promoter (promoter, enhancer 1, and enhancer 2). In the presence of SFN, Nrf2 binding to promoter, enhancer 1, and enhancer 2 sites was reduced in Sirt6‐knockdown cells relative to control cells but was rescued by reexpression of Sirt6 (Fig. 6C). In contrast, Bach1 binding to these sites was increased in Sirt6‐knockdown cells but decreased by reexpression of Sirt6. Sirt6 ChIP results showed that Sirt6 binds to HO‐1 promoter sites. These findings clearly indicate that Sirt6 switches from Bach1 to Nrf2 in the ARE region of the HO‐1 promoter.

Figure 3

Evaluation of systemic glucose and insulin tolerance. A, B) Six‐hour fasting blood glucose levels (A) and plasma insulin levels before and 10‐min after glucose (1 g/kg) injection (B) in 16‐h unfed mice (n = 8). C, D) Blood glucose concentrations during intraperitoneal glucose tolerance tests (C) and insulin tolerance tests (D) in 16‐ or 6‐h unfed mice, respectively (n = 8). Areas under the curve (AUC) were compared. Values are means ± sem. *P < 0.05, **P < 0.01 vs. HFHF‐fed WT; ^#P < 0.05 vs. NCD‐fed WT.

Co‐IP analysis revealed that SFN treatment induced Sirt6 binding to Nrf2 and MafK in WT hepatocytes but not in KO hepatocytes (Fig. 6D). On the contrary, Sirt6 basally interacted with Bach1, but this binding was reduced by SFN treatment in WT cells and without change in KO cells. The SFN‐stimulated Nrf2‐MafK interaction was resumed by reexpression of Sirt6 in KO cells (Fig. 6E), indicating that Sirt6 promotes Nrf2‐MafK binding while reducing Bach1‐MafK interaction. Altogether, these data suggest that Sirt6 favors the shift of the competition equilibrium between Bach1 and Nrf2 for small Maf proteins toward Nrf2. To prove whether Sirt6 control of ARE‐promoter activation requires Bach1, we conducted ARE‐luciferase assay in Bach1‐knockdown cells. Sirt6 deletion– or Sirt6 overexpression–induced changes of ARE‐luciferase activity was abolished in Bach1‐knockdown cells (Fig. 6F). Taken together, these data suggest that under the oxidative stress or electrophile stimuli, Sirt6 promotes Nrf2 binding to MafK, inducing the departure of Bach1 from Maf dimers for the expression of phase II/antioxidant genes.

Figure 4

Effect of Sirt6 deficiency on hepatic inflammation. A) After 16 wk of NCD or HFHF feeding, retrieved liver tissues were subjected to hematoxylin and eosin (H&E) staining or F4/80 immunostaining. The numbers of F4/80⁺ macrophages were counted and expressed as a percentage of hepatocyte numbers. Inflammation index was determined as described in Materials and Methods (n = 5). Scale bars, 250 μm. B) Plasma levels of ALT, AST, and lactate dehydrogenase (LDH) were measured by colorimetric detection kit (n = 8). C) mRNA expression of inflammation‐related genes was determined by qPCR (n = 8). D) Plasma levels of TNF‐α and CCL2 were measured by ELISA (n = 8). Values are means ± sem. *P < 0.05, **P < 0.01 vs. HFHF‐fed WT; ^#P < 0.05, ^##P < 0.01 vs. NCD‐fed WT.

Adenoviral Sirt6 overexpression prevents NASH progression in mice

Finally, we used Sirt6 adenovirus to validate the requirement of Sirt6 for the prevention of NASH development in vivo. C57BL/6 mice were fed an HFHF diet for 8 wk and injected twice with AdSirt6 or AdLacZ at 2‐wk intervals, as depicted in Supplemental Fig. S3A. Four weeks after the initial injection of adenovirus, metabolic phenotypes were compared. Liver TG content was significantly decreased by Sirt6 overexpression, although body weight and liver weight were not changed (Fig. 7A). Staining of liver sections with antibody against F4/80, Sirius red, or DHE revealed that HFHF‐induced hepatic inflammation, fibrosis, and ROS production were significantly attenuated in mice with Sirt6 overexpression (Fig. 7B, C). These results were further confirmed by qPCR results, which showed that mRNA expression of proinflammatory and profibrogenic genes was suppressed in AdSirt6‐injected mice compared with AdLacZ‐injected mice (Supplemental Fig. S3B, C). Liver SOD activity was reduced and GSH content was increased by AdSirt6 injection (Supplemental Fig. S3D). Collectively, these results confirm the importance of Sirt6 in preventing the progression of hepatic steatosis, fibrosis, and inflammation in mice.

DISCUSSION

In this study using an HFHF‐diet‐induced murine NASH model, we show that Sirt6 plays an important role in preventing the progression from simple steatosis to NASH under metabolic stress by regulation of the Nrf2‐ROS pathway. In particular, Bach1, a nuclear repressor of Nrf2, is negatively regulated by Sirt6, resulting in Nrf2 activation and the subsequent expression of phase II/antioxidant enzymes. Thus, loss of Sirt6 in hepatocytes leads to severe liver inflammation, oxidative stress, and liver fibrosis in HFHF‐fed mice, whereas adenoviral Sirt6 overexpression in mice attenuated NASH progress. In addition, Sirt6‐KO mice displayed enhanced hepatic fat accumulation under HFHF feeding, but this phenotype was reversed by adenoviral Sirt6 overexpression, highlighting Sirt6 as a dual player to curb 2‐hits of NASH development (27). Moreover, we show that hepatic expression of Sirt6 was repressed in patients with fibrosis, consistent with previous reports (28, 29). Thus, we suggest that these alterations could accelerate the progression of hepatic steatosis to NASH and fibrosis, proposing Sirt6 as a novel determinant in NAFLD/NASH.

Figure 5

Effects of Sirt6 deficiency on hepatic oxidative stress and fibrosis. A) After NCD or HFHF feeding for 16 wk, DHE staining was performed on liver cryosections to determine superoxide anion levels. Representative fluorescence microscopy images are shown. B) SOD activity and GSH content in liver tissues were determined by commercial assay kits (n = 8). C) Sirius red staining was performed on liver paraffin sections, and pericellular fibrosis was determined (n = 5). Bar graph indicates quantitative analysis of computed imaging of areas that were Sirius red positive. D) Expression levels of fibrogenesis‐related genes were determined by qPCR (n = 8). E) Western blotting for Nrf2, Bach1, MafK, and Keap1 in whole‐liver lysates, nuclear extracts (NE), or cytosolic extracts (CE). F) mRNA expression for Nrf2 and its target genes (n = 7). Scale bars, 250 μm. Values are means ± sem. *P < 0.05, **P < 0.01 vs. HFHF‐fed WT; ^#P < 0.05, ^##P < 0.01 vs. NCD‐fed WT.

We report here that Sirt6 deletion in the liver also aggravated HFHF‐diet‐induced insulin resistance and fatty liver formation, confirming previous reports. Aged mice with liver‐specific Sirt6 deletion displayed fatty liver formation under NCD with increased glycolysis, TG synthesis, and reduced β‐oxidation, and the suggested mechanism was that Sirt6 deacetylates H3K9 in the promoter of the genes involved in fat metabolism to negatively regulate the transcription (30). Likewise, Sirt6 global transgenic mice were protected from developing high‐caloric‐diet–induced type 2 diabetes mellitus (31, 32). In their studies, enhancement of insulin sensitivity after Sirt6 overexpression was prominent in liver compared with fat tissue, supporting our current findings that Sirt6 in hepatocytes plays a pivotal role in the control of systemic insulin resistance as well as hepatic steatosis. In the current study, the expression level of β‐oxidation genes such as peroxisome proliferator‐activated receptor a and carnitine palmitoyltransferase 1 was reduced in HFHF‐fed Sirt6‐KO mice, whereas the level of de novo lipogenesis genes was not altered (data not shown); therefore, increased hepatic fat accumulation in Sirt6‐KO mice may be due to the reduction of fatty acid oxidation.

Figure 6

Effects of Sirt6 deficiency on the recruitment of Bach1 and Nrf2 to ARE. A) Primary hepatocytes obtained from WT or KO mice were treated with 10 μM SFN, and Western blotting was performed for Nrf2, Bach1, and MafK in the nuclear extracts (n = 3). B) ARE promoter assay in HepG2 cells treated with 10 μM SFN or 30 μM tertiary butylhydroquinone (tBHQ) after transfection with shRNA targeting nonspecific sequence (sh‐Cont) or Sirt6 (sh‐Sirt6) (left panel; n = 3) and after transfection with vectors of Flag‐Sirt6 or Flag‐mtSirt6 (right panel; n = 3). *P < 0.05, **P < 0.01 vs. sh‐Cont or Flag+SFN; ^#P < 0.05, ^##P < 0.01 vs. sh‐Cont+VEH or Flag+VEH; ^$P < 0.05 vs. Flag‐Sirt6+SFN. C) Effect of Sirt6 knockdown or its reexpression on the occupancy of ARE sites of HO‐1 by Nrf2, Bach1, or Sirt6 was evaluated by ChIP assay in HepG2 cells with stable Sirt6 knockdown or overexpression (n = 3). *P < 0.05, **P < 0.01 vs. sh‐Sirt6; ^#P < 0.05, ^##P < 0.01 vs. sh‐Cont or H3. D, E) Primary hepatocytes were treated with 10 μM SFN for 3 h, and co‐IP assay was performed (n = 3). E) For reexpression of Sirt6 in Sirt6‐KO cells, KO hepatocytes were infected with AdSirt6 or AdLacZ (n = 3). F) ARE promoter activity after Bach1 deletion in HepG2 cells. Stable Sirt6‐KO (sh‐Sirt6) or Sirt6 overexpressed (Flag‐Sirt6) HepG2 cells were transfected with 2 different siRNA sequences targeting Bach1 (si‐Bach1 1, si‐Bach1 2) and incubated with SFN for 24 h for determination of ARE luciferase activity (n = 3). Gene knockdown or overexpression was confirmed by Western blotting. VEH, vehicle. ^##P < 0.01 vs. sh‐Cont or Flag.

The key question is the mechanism through which hepatic Sirt6 deficiency promotes the progression of steatosis to NASH. Our results show that oxidative stress in the liver was exaggerated in Sirt6‐KO mice fed an HFHF diet, with marked decreases in hepatic SOD and glutathione levels. Regulation of many antioxidant enzymes depends on the transcription factor Nrf2, and thus livers of Nrf2‐null mice are more susceptible to oxidative stress (33, 34). Accordingly, an inverse relationship between Nrf2 activity and NASH development has been reported (35, 36). Because Sirt6 activates the Nrf2 signaling pathway but its expression is suppressed in NASH, the impairment of Sirt6 activation of Nrf2‐dependent antioxidant and cellular defense response may contribute to NASH development. In the same way, the inability to activate Nrf2 may render Sirt6‐KO mice more susceptible to the HFHF diet–induced hepatocellular injury. All of these observations support the functional importance of Sirt6‐ and Nrf2‐mediated attenuation of oxidative stress in NASH development.

Figure 7

Effects of Sirt6 overexpression on NASH development. A) C57BL/6 mice were fed an NCD or HFHF diet for 8 wk and intravenously injected twice in a 2‐wk interval with 1 × 10⁹ pfu of either AdLacZ or AdSirt6. Body weight, liver weight, and liver TG content were measured 4 wk after initial virus injection (n = 5). B) Liver tissues retrieved 4 wk after initial virus injection were subjected to H&E staining, F4/80 immunostaining, Sirius red, or DHE staining. Scale bars, 250 μm. C) F4/80⁺ macrophages were counted, and pericellular fibrosis (Sirius red) was determined (n = 5). Values are means ± sem. *P < 0.05, **P < 0.01 vs. HFHF‐AdLacZ; ^##P < 0.01 vs. NCD. D) Proposed scheme for the role of Sirt6 in NASH development. Pol II indicates RNA polymerase II.

Under physiologic conditions, Nrf2 primarily exists in the cytoplasm linked with its cytoplasmic repressor, Keap1 (37). In response to oxidative stress, Nrf2 dissociates from Keap1, translocates into the nucleus, and binds to cis‐elements (ARE sites) as a heterodimer with other transcription factors, such as small Mafs (MafK, MafF, and MafG) or c‐Jun (38). Bach1 forms heterodimers with small Mafs and competes with Nrf2 for binding to ARE sites. Thus, Bach1 switches off Nrf2‐mediated antioxidant gene expression and acts as a functional inhibitor of Nrf2. Based on this background, we explored the possibility of Sirt6 regulation of Nrf2 through repression of Bach1. Deficiency or overexpression of Sirt6 in the liver or hepatocytes led to the following changes: 1) Sirt6‐KO hepatocytes had an increased level of nuclear Bach1 compared with WT, and, on the contrary, Sirt6 overexpression decreased it; 2) Sirt6 deletion led to reduction of Nrf2 binding to ARE sites in the promoter region of HO‐1 while increasing Bach1 binding, and these changes were reversed by Sirt6 reexpression; 3) SFN stimuli induced the switch of the MafK binding partner from Bach1 to Nrf2 in complex with Sirt6 in WT hepatocytes, but in KO hepatocytes Bach1‐MafK interaction persisted despite stimuli; 4) Sirt6 deficiency– or Sirt6 overexpression–induced alterations of ARE promoter activity were abolished by Bach1 deletion in hepatocytes. Altogether, these results suggest that Sirt6 promotes the displacement of Bach1 from MafK and enhances the nuclear import of Nrf2, which leads to increased Nrf2‐MafK interaction and transcriptional activation of phase II/antioxidant genes, thereby protecting against ROS‐mediated NASH pathologies (Fig. 7D). A previous report has shown that HO‐1, but not NQO1, is regulated by Bach1, although both genes contain ARE sites in their promoter regions, suggesting that the contribution of Bach1 could be different among ARE‐dependent gene regulation (39). In our study, most of the Nrf2‐ARE pathway–dependent phase II/antioxidant genes are suppressed in Sirt6‐deficient liver and hepatocytes, and therefore we suggest that mechanisms in addition to Bach1 regulation may be involved in Sirt6 promotion of phase II/antioxidant enzymes. Consistent with this notion, a recent report showed that Sirt6 interacts with the RNA polymerase II complex and positively regulates Nrf2‐mediated HO‐1 expression in human mesenchymal stem cells (40). Despite incomplete understanding of Sirt6′s action mechanism, this report and our current study identify Sirt6 as a positive regulator of Nrf2 antioxidant pathway.

We also observed that hepatocyte Sirt6 deletion promoted accumulation or activation of F4/80⁺ macrophages in the livers of HFHF‐fed Sirt6‐KO mice. Consistently, the expression and secretion of numerous inflammatory cytokines such as TNF‐α, IL‐β, and MCP‐1 was increased in HFHF‐fed Sirt6‐KO mice. Jiang et al. (41) found a secretion of TNF‐α by removing the fatty acyl modification on K19 and K20 of TNF‐α. Given that TNF‐α is mainly produced by Kupffer cells in the liver (42), fatty deacylation of TNF‐α by Sirt6 in hepatocytes and concomitant secretion may be negligible. Instead, the enhanced inflammation status in liver tissues of HFHF‐fed Sirt6‐KO mice, which are largely mediated by Kupffer cells and infiltrated macrophages, might be the source of TNF‐α.

In summary, this study provides a novel function of Sirt6 as a transcriptional regulator of Nrf2. Sirt6‐mediated, ARE‐dependent gene transcription occurred through the dynamic changes of protein‐protein interactions among Sirt6, Nrf2, Bach1, and small Mafs and the consequent modification of their binding to the ARE. Further work is required to elucidate how Sirt6 influences the transcriptional suppressor activity of Bach1 in hepatocytes. We suggest that nutritional or pharmacological modulation of Sirt6 could be therapeutic for treating patients with NAFLD.

ACKNOWLEDGMENTS

This work was supported by the Basic Science Research Program (Grants 2015R1D1A1A01058502, 2017R1A2B2005730, and 2017R1A2B4008593), and by the Medical Research Center Program (Grant 2008‐0062279) through the Korean National Research Foundation (Ministry of Science, Information and Communication Technology and Future Planning). The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

S.‐O. Ka and I. H. Bang performed the experiments and analyzed the data; E. J. Bae and B.‐H. Park designed the experiments, interpreted the data, and wrote the manuscript; and all authors reviewed the manuscript.