Alcohol Promotes Hepatocellular Carcinoma Metastasis Through SIRT7-Dependent EMT Regulation

Long-term alcohol use is a conrmed risk factor of liver cancer tumorigenesis and metastasis. Multiple mechanisms responsible for alcohol related tumorigenesis have been proposed, including toxic reactive metabolite production, oxidative stress and fat accumulation which trigger hepatocyte cell death and inammation. However, mechanisms underlying alcohol-mediated liver cancer metastasis remain largely unknown. SIRT7 expression pattern and its association with HCC metastasis were investigated by bioinformatic analysis and veried by western blot and immunochemistry in HCC tissues. The biological consequences of overexpression and knockdown of SIRT7 in HCC metastasis were studied in vitro and in vivo. qRT-PCR, immunouorescence assay, ChIP assay were utilized to assess the effects of SIRT7 on E-cadherin expression. Effects of alcohol on SIRT7 expression were evaluated by qRT-PCR, immunouorescence and inhibitor treatments and pulmonary metastasis model mice fed with Lieber Decarli alcohol diet were used to clarify the mechanisms by which SIRT7 facilitated alcohol mediated HCC metastasis. Our ndings revealed that high SIRT7 levels are associated with metastasis in human HCC. SIRT7 regulates HCC EMT and suppresses E-Cadherin transcription. Our data also suggested that alcohol upregulates SIRT7 expression in both human primary hepatocytes and alcohol-fed mice liver in CYP2E1-dependent mechanism. Reducing SIRT7 activity signicantly prevented alcohol-mediated liver cancer metastasis in vivo. The current ndings present a novel mechanism that controls alcohol-mediated HCC metastasis and reveals SIRT7 as a pivotal regulatory factor in regulating EMT and determining metastasis in human HCC. Therapeutic strategies that inhibit SIRT7 may offer novel options for the treatment of HCC.


Abstract
Background Long-term alcohol use is a con rmed risk factor of liver cancer tumorigenesis and metastasis. Multiple mechanisms responsible for alcohol related tumorigenesis have been proposed, including toxic reactive metabolite production, oxidative stress and fat accumulation which trigger hepatocyte cell death and in ammation. However, mechanisms underlying alcohol-mediated liver cancer metastasis remain largely unknown.
Methods SIRT7 expression pattern and its association with HCC metastasis were investigated by bioinformatic analysis and veri ed by western blot and immunochemistry in HCC tissues. The biological consequences of overexpression and knockdown of SIRT7 in HCC metastasis were studied in vitro and in vivo. qRT-PCR, immuno uorescence assay, ChIP assay were utilized to assess the effects of SIRT7 on E-cadherin expression. Effects of alcohol on SIRT7 expression were evaluated by qRT-PCR, immuno uorescence and inhibitor treatments and pulmonary metastasis model mice fed with Lieber Decarli alcohol diet were used to clarify the mechanisms by which SIRT7 facilitated alcohol mediated HCC metastasis.

Results
SIRT7 is a critical factor in promoting liver cancer metastasis. SIRT7 expression is closely associated with disease stage and high SIRT7 predicts worse overall and disease-free survival. Overexpression of SIRT7 promotes HCC cell migration and EMT while knockdown of SIRT7 showed opposite effects.
Mechanistically, we found that SIRT7 suppresses E-Cadherin expression through promoter binding and H3K18 deacetylation. Most importantly, we identi ed that alcohol upregulates SIRT7 expression in hepatocyte both in vitro and in vivo. Reducing SIRT7 activity completely abolished alcohol-mediated liver cancer metastasis in vivo. Conclusion SIRT7-dependent EMT regulation is a pivotal regulatory mechanism of alcohol-mediated HCC metastasis and reveals previously unidenti ed roles of SIRT7 in promoting EMT and metastasis in human HCC.
Therapeutic strategies that inhibit SIRT7 may offer novel options for the treatment of HCC. Background Hepatocellular carcinoma (HCC), the most common pathological form of primary liver cancer, accounts for the fth leading cause of cancer-related human death worldwide and its incidence still increasing (1).
Despite great efforts that have been made during the last decades to improve diagnosis and treatment for intrahepatic lesions, advanced HCC often has a poor prognosis with a ve-year survival rate less than 18% mainly due to drug resistance, tumor metastasis and recurrence (1,2). While extensive research has identi ed a number of molecular biomarkers, cellular networks and signaling pathways affected in liver cancer, mechanisms underlying tumor metastasis are poorly understood (2).
Alcohol is a con rmed risk factor of liver cancer and long-term alcohol use has been linked to an increased risk of liver cancer tumorigenesis as well as distant metastasis (3)(4)(5)(6). Multiple mechanisms are responsible for alcohol related tumorigenesis, including production of toxic reactive metabolites, oxidative stress and fat accumulation in hepatocytes. All of those pathological changes result in hepatocyte death, in ammation, brosis deposition and ultimately liver cancer (4,7,8). Emerging evidence has suggested that long term alcohol use causes cancer stem cell activation and epithelialmesenchymal transition (EMT) which facilitates liver cancer cell metastasis (6,9). However, factors and mechanisms underlying alcohol-mediated liver cancer metastasis remaining largely unknown.
Epithelial mesenchymal transition (EMT) is the most common feature when advanced cancer initiates metastasis (10,11). In the case of HCC, EMT is utilized by liver cancer cells to initiate metastatic spread.
After EMT, liver epithelial cells lose cell-to-cell adhesion and become motile mesenchymal cells, which allows for the migration from the primary tumor site to establish distant metastasis (12). A critical molecular feature responsible for liver cancer cell EMT is the downregulation of E-cadherin. E-cadherin is a cell adhesion molecule present in most normal epithelial cell plasma membranes but it is frequently repressed or degraded in various type of human cancer cells (13). Mechanisms underlying the functional loss of E-cadherin in cancer cells include posttranslational loss of protein function, transcriptional silencing due to promoter hypermethylation, and the activation of several transcription repressors of Ecadherin, such as Snail, Slug, Sip1 and Ets (13,14). Previous studies have shown that alcohol was able to promote liver cancer EMT and metastasis but the factors and mechanisms underlying alcohol mediated EMT remain largely unknown (6, 9). SIRT7 is a family member of the silent information regulator (Sir2) proteins that are described as NAD +dependent class III histone deacetylases (HDAC3). Compared with other family members like SIRT1 and SIRT6, the enzymatic activity and functions of SIRT7 are poorly understood. SIRT7 is the only Sir2 family member that is predominantly localized in the nucleus where it regulates RNA polymerase I transcription by acting as an H3K18 deacetylase (15). Besides H3K18, SIRT7 has also been reported to target several non-histone proteins, including p53 (16), GABP-β (17), FOXO3 (18) and U3-55k (19) for deacetylation, and has been implicated in multiple cellular functions including hepatic lipid metabolism, mitochondrial homeostasis and adipogenesis. Emerging evidence has also implicated SIRT7 in various aspects of cancer biology (20)(21)(22). H3K18 deacetylation by SIRT7 is important for maintaining the fundamental properties of the cancer cell phenotype (20). Knockdown of SIRT7 in uences cell cycle control and causes an increased proportion of cancer cells that remain in the G1/S phase (15,20,21). In epithelial prostate carcinomas, high SIRT7 levels are associated with metastatic disease and poor prognosis (22).
In HCC, SIRT7 expression is also upregulated in a large cohort of HCC patents (21) and we have identi ed elevated SIRT7 expression is associated with chemosensitivity by regulating TP53 activity (23).
In the present study, we further demonstrated that SIRT7 as a critical regulator of liver cancer metastasis.
Our ndings revealed that high SIRT7 levels are associated with metastasis in human HCC. SIRT7 regulates HCC EMT and suppresses E-Cadherin transcription. Our data also suggested that alcohol upregulates SIRT7 expression in both human primary hepatocytes and alcohol-fed mice liver in CYP2E1dependent mechanism. Reducing SIRT7 activity signi cantly prevented alcohol-mediated liver cancer metastasis in vivo. The current ndings present a novel mechanism that controls alcohol-mediated HCC metastasis and reveals SIRT7 as a pivotal regulatory factor in regulating EMT and determining metastasis in human HCC. Therapeutic strategies that inhibit SIRT7 may offer novel options for the treatment of HCC.

Materials And Methods
Cell culture, plasmids and transfection

Animal model
Male NSG mice (4 weeks of age) were purchased form The Jackson Laboratory (Bar Harbor, ME) and Gempharmatech (Nanjing, China). Mice were housed in a temperature-controlled, pathogen-free environment with 12-hour light-dark cycles. All animal handing procedures were approved by the Institutional Animal Care and Use Committees at The University of Kansas Medical Center (Kansas City, KS) and Hunan Normal University School of Medicine (Changsha, China).
Mice received a single tail vein injection of 1x10 6 cells suspended in 100 μL DMEM and tumor formation was determined in lung 4 weeks after injection by using Hematoxylin &Eosin (H&E) staining. Alcohol feeding procedures were previously described (25). One week after cell injection, mice were initially fed the control Lieber-DeCarli diet ad libitum for 5 days to acclimatize them to a liquid diet. Then mice were subsequently allowed free access to the ethanol Lieber-DeCarli Diet containing 5% (vol/vol) ethanol for 2 weeks, and control-fed groups were pair-fed with an isocaloric control diet. All mice were sacri ced 4 weeks after injection and lungs were collected for determining tumor formation. Liver sections from CYP2E1 -/and control mice were kindly provided by Dr. Laura Nagy as previously described (26).

Human specimens and immunohistochemistry
De-identified human liver specimens from liver explants were obtained from The University of Kansas Medical Center, The First A liated Hospital of Chongqing Medical University, and The A liated Hospital of Hunan Normal University (People's Hospital of Hunan Province). Written informed consent was obtained from all patients and all studies using human tissue samples were approved by the Human Subjects Committee of the University of Kansas Medical Center, Chongqing Medical University and Hunan Normal University School of Medicine.
Immunohistochemistry was performed as previously described (23,25). After depara nization and rehydration, antigen retrieval was achieved by heating in a pressure cooker for 5 min in 10mM of sodium citrate (pH6). Peroxidase activity was blocked by incubation in 3% H2O2 for 10 minutes. Sections were rinsed three time in PBS/PBS-T (0.1% Tween-20) and incubated in Dako Protein Block (Dako, Agilent Technologies, Santa Clara, CA) for 10 minutes. After removal of blocking solution, slides were placed into a humidi ed chamber and incubated overnight with primary antibodies in blocking buffer (4% normal goat serum in PBS) and incubated over night at 4 °C. After washing, slides were covered with SignalStain Boost IHC Detection Reagent (Cell Signaling Technologies, Boston, MA) for 30 min at room temperature. After washing two times with PBS-T, the Substrate-Chromgen Solution (VECTOR NovaRED, Substrate Kit, Vector Laboratories, Burlingame, CA) was applied, slides were incubated 5-10 min and counterstained with Hemtoxylin. Images were acquired using a Nikon Eclipse 80i microscope (Nikon Americas Inc., Melville, NY).

Isolation of human and mouse primary hepatocyte
Primary human hepatocytes were freshly isolated from liver resections as previously described (24,27) by the Cell Isolation Core of the Department of Pharmacology at the University of Kansas Medical Center. All human tissues were obtained with informed consent from each patient, according to ethical and institutional guidelines. The study was approved by the Institutional Review Board at the University of Kansas Medical Center.
Mouse primary hepatocyte were isolated by using a multi-step collagenase procedure (27). In brief, the liver was perfused with calcium-free solution and then digested with a collagenase (Sigma-Aldrich) perfusion. Dispersed cells were released from the isolated liver, hepatocytes were collected by 50×g centrifugation and then seeded on collagen coated plates and allowed to attach in a humidi ed 37 °C, 5% CO 2 incubator for 12 h and treated with 50mM ethanol for indicated times Quantitative polymerase chain reaction (qRCR) Total RNA was isolated from cells using the TRIzol reagent (Thermo Fisher Scienti c, Inc.), followed by cDNA synthesis using an RNA reverse transcription kit (Applied Biosystems; Thermo Fisher Scienti c, Inc.). Subsequently, a CFX96 real-time system (Bio-Rad, Hercules, CA, USA) was used to perform qPCR.

Western blot analysis
Total cell lysates prepared from cells were used for detection. Brie y, cells were washed twice with chilled PBS and then cell lysis was performed using the RIPA buffer (50 mM Tris, pH 7.5, 150 mM sodium chloride, 1% NP-40, 0.2% SDS, 0.5% sodium deoxycholate, 0.1 mM EDTA, and 1% protease and phosphatase inhibitors (Sigma-Aldrich). After centrifugation at 16,000 g for 15 min, supernatants were collected. Cell lysates (25 μg) were separated on a 10% SDS-PAGE and transferred to polyvinylidene di uoride membranes (Immobilon-P membranes; EMD Millipore, Billerica, MA, USA). Membranes were blocked with blocking buffer (5% skim milk, 0.1% Tween-20 in PBS) for 1 h at room temperature. Following incubation with primary antibodies (1:1000) overnight at 4°C, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (Thermo Fisher Scienti c, Inc. Waltham, MA, USA). Signals were detected using the ECL Western Blotting Detection system (Thermo Fisher Scienti c, Inc).

Immuno uorescence
For indirect immuno uorescence, cells grown on coverslips were xed with 4% paraformaldehyde at room temperature for 5 min and 0.2% Triton X-100 was used for cell permeation. The coverslips were inverted on 40-μl droplets of the blocking buffer (4% goat serum) and incubated at room temperature for 45 min to prevent non-specific binding. Subsequently, cells were incubated with primary antibodies for 1 h at room temperature. Coverslips were washed with PBS, followed by incubation for 1 h at room temperature in the dark with Alexa Fluor-conjugated secondary antibodies (1:5,000; Molecular Probes; Thermo Fisher Scienti c, Inc.). Additionally, Dihydrochloride (DAPI) was added for 5 min at room temperature to stain nuclear DNA. Images were acquired using a Nikon Eclipse microscope (Nikon Corporation, Tokyo, Japan).
Wound healing assay 2×10 4 cells were plated in 24-well plates. When cells reached 100% con uence, sterile pipette tips (100 μl) were used to scratch the wound. Cell motility was assessed by measuring the movement of cells into the scratched wound after 24h incubation.

Migration assay
The 24-well transwell plates (0.4 μm pore size, Corning, Tewsbury, MA) were used to measure the migration ability of the cells. HCC cells (2×10 3 ) in 200 μl FBS-free media were added into the upper chamber, whereas 600 μl of the media with 10% FBS was added into the bottom chambers. Cells were xed with methanol after a 48h incubation. Cells left on the membrane in the top chamber side were wiped off with a cotton swab, whereas the cells that migrated through the membrane were stained with crystal violet. The number of migratory cells in 10 random elds was counted under a microscope.

Statistical Analysis
Data are presented as mean ± sem. Statistical signi cance between groups was calculated by using oneway ANOVA followed by Turkey's test. Statistical signi cance between two groups was calculated by 2tailed unpaired Student's t-test. Variance between groups met the assumptions of the appropriate test. Unless otherwise stated, a P-value of <0.05 was considered statistically signi cant. The Kaplan-Meier method was used to estimate the survival rates for SIRT7 expression. Equivalences of the survival curves were tested by log-rank statistics.

Results
Aberrantly increased SIRT7 expression was associated with HCC metastasis and poorer survival. To con rm clinical signi cance of SIRT7 in HCC patients, We rst analyzed RNA sequencing data collected from The Cancer Genome Altas (TCGA) public database. The results showed that SIRT7 expression was increased in HCC tissues and positively correlated with disease progression except stage IV due to limited sample size (Fig. 1A). The Kaplan-Meier method (using the log-rank test) also suggested that lower SIRT7 expression was associated with increased overall survival (Fig. 1B) and disease-free survival (Fig. 1C) compared with those of patients with higher levels of SIRT7. To further evaluate whether SIRT7 contributed to HCC metastasis, we collected HCC tumor samples from patients who underwent surgical resection or transplantation and examined SIRT7 expression by IHC and western blot. Consistent with our previous observations, we found that SIRT7 expression in tumors showing evidence of vascular invasion was signi cantly higher than in tumors that lacked this more aggressive, metastasis-promoting feature (Fig. 1D). IHC staining results indicated that in both normal and cirrhotic liver sections, SIRT7 staining was undetectable. However, SIRT7 showed positive staining in HCC and strong nuclear staining in metastatic HCC tissue (Fig. 1E).
SIRT7 is critical for HCC metastasis both in vitro and in vivo. To further con rm the role of SIRT7 in HCC metastasis, we knocked down SIRT7 in HCC cells using siRNA ( Fig. 2A) and evaluated cell migration by using a wound healing and migration assay. Knockdown of SIRT7 signi cantly impaired the cell migration capability in Huh7 and HepG2 cells compared with those of controls ( Fig. 2B-D). We further stably knocked down SIRT7 in SK-Hep1 cells with two different lentiviruses (shSIRT7#1 and shS IRT7#2) and evaluated whether SIRT7 regulates HCC metastasis in vivo using a xenograft model of pulmonary metastasis (Fig. 2E-G). We found that knockdown of SIRT7 markedly decreased lung metastasis compared with controls (shTRC, Fig. 2F and G). In contrast, overexpression of SIRT7 in HepG2 cells resulted in a signi cant increase of lung metastasis compared with formed by control cells (UT, p<0.001, Fig. 2H and I).
SIRT7 promotes EMT in HCC. To explore whether SIRT7 promoted HCC cell migration and metastasis by regulating EMT of HCC, we examined EMT associated markers in HCC cell lines using western blotting analysis. As expected, signi cantly downregulated E-cadherin and upregulated vimentin were observed in Huh7 cells overexpressed SIRT7 compared with those transfected with empty vectors (EV). Knockdown SIRT7 showed the opposite effects (Fig. 3A). Immuno uorescence showed similar results of SIRT7 knockdown and demonstrated increased E-cadherin and pan-claudin and decreased vimentin levels (Fig. 3B). We further performed western blot and IHC staining to compare expression of SIRT7, E-cadherin and vimentin in HCC samples with or without metastasis (Fig. 3C-E). Higher SIRT7 levels were present in metastatic HCC samples compared with non-metastatic HCC tissue. SIRT7 expression patterns were negatively correlated with E-cadherin expression level ( Fig. 3C and D). Spearman correlation analysis showed a strong negative correlation between SIRT7 and E-cadherin (R=-0.566, p<0.05). We also analyzed RNA sequencing data collected from the TCGA and the results also showed negative correlation between SIRT7 and E-cadherin (R=-0.1287, p<0.008) (Fig. 3F). Kaplan-Meier method suggested that lower E-cadherin expression was associated with decreased overall survival (Fig. 3G). Most importantly, high expression of SIRT7 and low E-cadherin was associated with the lowest survival probability (Fig. 3H). SIRT7 induces H3K18 deacetylation and suppresses E-cadherin via a FOXO3-dependent mechanism. Downregulation of E-cadherin is pivotal process of EMT in cancer cells and our data indicated that SIRT7 overexpression reduced E-cadherin expression. We thus sought to determine the mechanisms of SIRT7 suppression of E-cadherin. Overexpression of SIRT7 signi cantly decreased transcriptional activity of the E-cadherin promoter (Fig. 4A), indicating that SIRT7 suppresses E-cadherin at a transcriptional level. To assess whether SIRT7 binds to the E-cadherin promoter we performed ChIP assays. As shown in Fig. 4B, SIRT7 showed positive binding to E-cadherin promoter compared with IgG and this binding was signi cantly abolished in cells treated with siSIRT7. SIRT7 is involved in deacetylating H3K18 and this is responsible for gene silencing and maintenance of oncogenic transformation in human cancer cells. W thus next examined whether SIRT7 suppresses E-cadherin through similar mechanisms and performed ChIP assays measuring H3K18 acetylation levels at the E-cadherin promoter region. The results show that knockdown of SIRT7 resulted in signi cantly elevated H3K18 but not H3K56 acetylation level (Fig. 4C). This demonstrates that SIRT7 is responsible for H3K18 deacetylation at the E-cadherin promoter.
We further sought to determine how SIRT7 is recruited to the E-cadherin promoter since SIRT7 lacks known sequence-speci c DNA binding domains (20). We analyzed E-cadherin promoter regions and identi ed FOXO3, ELK4 binding sites and E-box within this region. We thus made a series of mutants and performed luciferase assays to investigate whether those sites are required for SIRT7-mediated Ecadherin repression (Fig. 4D). The results showed that overexpression of SIRT7 suppressed transcriptional activity of E-cadherin. Mutations of E-box or ELK4 showed no effect on SIRT7-dependent repression. However, mutants of either FOXO3 binding site completely abolished SIRT7-mediated suppression (Fig. 4E). To further address whether FOXO3 is required for SIRT7-mediated E-cadherin repression, we knocked down FOXO3 (Fig. 4E) and performed luciferase assays in those cells. As expected, in the absence of FOXO3, SIRT7 no longer suppressed transcriptional activity of E-cadherin compared with empty vectors (Fig. 4F). We further investigated expression of FOXO3 and SIRT7 patterns in determining disease progression in human liver cancer and Kaplan-Meier method suggested that high FOXO3 expression were associated with decreased overall survival (Fig. 4G) and disease-free survival (Fig. 4H) compared with those of patients with lower levels of FOXO3. Most importantly, high expressions of SIRT7 and FOXO3 was associated with the lowest survival probability while low expressions of SIRT7 and FOXO3 were associated with the highest survival probability compared with other expression patterns, respectively (Fig. 3H).
Elevated SIRT7 expression in hepatocytes upon ethanol exposure. Alcohol is well known to promote HCC metastasis (9, 28) and our results indicated that SIRT7 is also critical for HCC metastasis. We thus evaluated whether alcohol promotes HCC metastasis through SIRT7. To test this hypothesis, we rst determined whether ethanol treatment modulates SIRT7 expression in HCC cells. Consistent with a previous report (29), ethanol treatment of HCC cells resulted in downregulation of SIRT1 expression, but SIRT7 expression was signi cantly upregulated in both SK-Hep1 and Hep3B cells (Fig. 5A). Elevated mRNA levels of SIRT7 were also observed in SK-Hep1 cells starting from 24h after alcohol exposure (Fig. 5B), indicating ethanol treatment regulates SIRT7 at a transcriptional level. In primary human and mouse hepatocytes, ethanol treatment resulted in increased SIRT7 expression and concomitantly decreased E-cadherin expression (Fig. 5C). We further examined SIRT7 expression in alcohol fed mouse liver and the results showed that alcohol feeding increased SIRT7, and decreased E-cadherin expression compared with a control diet (Fig. 5D). To address whether alcohol feeding increased SIRT7 expression in hepatocytes, we isolated hepatocytes from mouse liver and analyzed protein levels by western blot, the results indicated that alcohol feeding increases SIRT7 expression but suppressed E-cadherin expression in hepatocytes (Fig. 5E).
Cyp2E1-dependent oxidative stress is responsible for alcohol mediated SIRT7 induction. To determine whether alcohol dehydrogenase and cytochrome P450-dependent oxidizing systems were responsible for SIRT7 induction, we rst treated cells with acetaldehyde (ALD) or ethanol in the presence of inhibitors of either aldehyde dehydrogenase (ALDH) alcohol dehydrogenase (ADH) or Cyp2E1 and measured SIRT7 mRNA levels (Fig. 6A). We found that ethanol increased SIRT7 mRNA level almost 3-fold but ALD treatment did not change SIRT7 mRNA level. The Cyp2E1 inhibitor (Cyp2E1 inhi) alizarin completely abolished ethanol induced SIRT7 elevation, while neither the alcohol dehydrogenase inhibitor (ADH inhi) fomepizole nor the aldehyde dehydrogenase inhibitor (ALDH inhi) daidzin showed any effects (Fig. 6A). We further con rmed these ndings by using immuno uorescence (IF) and similar results were observed (Fig. 6B). Next, we measured Cyp2E1 expression after ethanol treatment and found signi cant induction of Cyp2E1 after 48h of ethanol treatment both in primary human hepatocyte and HepG2 cells (Fig. 6C).
To further con rm whether Cyp2E1 was required for ethanol induced SIRT7 elevation, we performed IHC staining in liver sections from wild type (WT) or CyP2E1 −/− mice fed with control or alcohol diet (Fig. 6D). In WT mice, alcohol feeding resulted in a signi cant increase of SIRT7 staining which predominantly localized in nuclei. This increase was not seen with control diet. Consistent with our in vitro data, decreased E-cadherin levels were also observed in the liver after alcohol feeding compared with control diet (Fig. 6D). In contrast, in Cyp2E1 −/− mice, there were no obvious changes of SIRT7 staining after alcohol feeding compared with control diet and the decrease of E-cadherin levels reduced when compared with WT mice after alcohol (Fig. 6D). Since Cyp2E1 has been suggested as a major contributor to ethanol-induced oxidant stress (8), we thus investigated whether reactive oxygen species (ROS) were responsible for ethanol induced SIRT7 elevation by using N-acetly-cysteine (NAC) as an ROS inhibitor and 4-hydroxynonenal (4-HNE) as an inducer of oxidative stress (Fig. 6E and F). NAC treatment signi cantly decreased while 4-HNE further enhanced ethanol mediated SIRT7 elevation at mRNA levels ( Fig. 6E and  F). More importantly, we found that 4-HNE treatment itself was su cient to increase SIRT7 level in HepG2 cells (Fig. 6G).
Knockdown SIRT7 abolished alcohol mediated HCC metastasis in vivo. The above studies demonstrate that alcohol promotes SIRT7 which is critical for HCC metastasis, suggesting that SIRT7 may play a role in alcohol mediated HCC metastasis. To test this, we rst compared SIRT7 and E-cadherin expression in clinical samples (Table. 1) from HCC patients with and without a history of alcohol consumption (Fig. 7A and B). Consistent with our previous observations, SIRT7 expression in alcohol patients was signi cantly higher than in non-alcohol patients (Fig. 7A, Table 1). IHC staining results indicated mild SIRT7 staining in non-alcohol patients but in alcohol patients, SIRT7 showed strong nuclear and cytosolic staining may due to post translational modi cations caused by alcohol (30) (Fig. 7B). In contrast, E-cadherin expression was signi cantly decreased in alcohol-associated tumors compared with those from non-alcohol patients ( Fig. 7A and B). These results suggest that alcohol consumption modulates SIRT7 and E-cadherin expression in human HCC. To further con rm the role of SIRT7 in alcohol mediated HCC metastasis, we used an established pulmonary metastasis model and fed mice with control diet or alcohol. Due to the high mortality of control mice after 5 weeks alcohol feeding, we fed our mice with alcohol for shorter period of time (2 weeks) one week after tail vein injection (Fig. 7C-E). We found that 2 weeks of alcohol feeding markedly increased lung metastasis compared with those of control diet, evidenced by surface tumor formation (Fig. 7C), IHC staining (Fig. 7D) and tumor number (Fig. 7E). This effect was completely abolished by knockdown of SIRT7 in SK-Hep1 cells with lentiviruses (shSIRT7#2, p<0.01). These data clearly indicate that SIRT7 plays a crucial role in alcohol mediated HCC metastasis in vivo.

Discussion
Alcohol has long been recognized as a risk factor for liver cancer and long-term alcohol use has been linked to an increased risk of both primary liver tumorigenesis as well as distant metastasis (3-6). Multiple mechanisms seem to be related to alcohol-induced liver cancer tumorigenesis including production of toxic reactive metabolites, oxidative stress and fat accumulation in hepatocyte. These factors all trigger hepatocyte death and promote in ammation (3,31). Evidence suggests that alcohol use is associated with cancer stem cell activation and EMT, and these facilitate cancer metastasis (6,9).
Nonetheless, additional factors and mechanisms that are responsible for alcohol-mediated liver cancer metastasis remain elusive. In the present study, we have demonstrated that SIRT7 is a pivotal factor which is responsible for alcohol-mediated HCC metastasis. Alcohol upregulates SIRT7 in human HCC and this promotes EMT and increases the risk of metastasis. Knockdown SIRT7 completely abolished alcohol-mediated HCC metastasis in vivo. Our data thus provides evidence that SIRT7 induction by CYP2E1-dependent oxidative stress is a molecular mechanism responsible for alcohol-mediated HCC metastasis.
SIRT7 is the most recently identi ed mammalian sirtuin and growing evidence suggests that its activity is important for human cancer (21,(32)(33)(34). High SIRT7 expression is associated with an aggressive cancer phenotype, distant metastasis and poor patient survival in diverse cancers (20,23,35,36). In particular, SIRT7 is important for regulating cancer cell EMT and promoting metastasis, and SIRT7 inactivation reverses metastatic phenotypes in both epithelial and mesenchymal tumors (22,37,38). Consistent with these ndings, we observed that SIRT7 expression is associated with disease grade in human HCC and high SIRT7 expression showed signi cant correlation with poor overall and a reduced disease free survival rate. We also demonstrated that SIRT7 is a crucial EMT regulator in HCC and that it promotes cancer cell metastasis. Together with previous work, our ndings further highlight the importance of SIRT7 in regulating EMT and maintaining the metastatic phenotype in human HCC.
EMT is the most common feature of advanced cancers, and it initiates metastatic spread (10,11). In the case of liver cancer, downregulation of E-cadherin expression is the critical molecular feature responsible for cancer cell EMT (39). While promoter hypermethylation and activation of transcriptional repressors are reported to suppresses E-cadherin (13,14), our data further revealed that SIRT7 selectively deacetylates H3K18 at the E-cadherin promoter and acts as an E-cadherin repressor in human HCC. We demonstrated that this repression requires FOXO3, a well characterized transcription factor responsible for antioxidant responses and longevity (40). In the absence of FOXO3, SIRT7 no longer suppress E-cadherin transcription. As E-cadherin is a well-documented biomarker of liver cancer prognosis (39,41), our data suggested that SIRT7 may cooperate with FOXO3 in determining liver cancer progression. Consistent with this, we observed that patients with both high SIRT7 and FOXO3 expression showed the worst survival rate when compared with other expression patterns.
SIRT7 is frequently upregulated in various type of human cancer (42). Studies have been mainly focused on the functional roles and therapeutic options by targeting SIRT7 (23,35,36,43) while mechanisms underlying its regulation have been less well studied. Here we provide evidence that alcohol has the ability to upregulate SIRT7 in both normal liver and human HCC. We further revealed that this upregulation requires CYP2E1 which constitutes the major component of the ethanol oxidizing system.
The catalytic activity of the CYP2E1 enzyme results in the generation of ROS which are involved in the initiation and perpetuation of alcohol-associated liver disease. ROS also play a role in signal transduction and cellular physiology to prevent cellular damage (44). Our data indicated that oxidative stress signaling is crucial in alcohol induced SIRT7 upregulation as an anti-oxidant (NAC) prevented and an ROS-induced reactive aldehyde (4-HNE) upregulated SIRT7 mRNA level. Our ndings thus provide evidence that ROS serves as a critical factor responsible for alcohol induced SIRT7 upregulation in human HCC. However, even though ROS is a main product of CYP2E1 mediated ethanol oxidation, how ROS activate SIRT7 transcription and whether ROS is primarily responsible for alcohol induced SIRT7 upregulation requires further investigation.
In summary, our data illustrate molecular mechanisms responsible for alcohol-mediated HCC metastasis and demonstrate a crucial role of SIRT7 in alcohol-mediated HCC. Our ndings thus highlight the importance of SIRT7 in HCC progression and provide a useful target for the development of mechanismbased cancer therapeutic strategies. Conclusions SIRT7-dependent EMT regulation is a pivotal regulatory mechanism of alcohol-mediated HCC metastasis.
Therapeutic strategies that inhibit SIRT7 may offer novel options for the treatment of HCC.

AVAILABILITY OF DATA AND MATERIALS
All data generated or analyzed during this study are included in this published article.

COMPETING INTERESTS
The authors declare that they have no nancial con icts of interest relevant to this study.  Tables   Table 1 Correlative analysis of alcohol consumption with SIRT7 levels or clinicopathological features by using a nonparametric Fisher's Exact Test.  injected into NSG mice via tail vein injection, tumor formation was determined in lung 4 weeks after injection by using H&E staining and quanti ed data shown in G. (H-I) HepG2 cells were transfected with SIRT7 for 24 h and then 1x106 cells were injected into NSG mice via tail vein injection, tumor formation was determined in lung 4 weeks after injection by using H&E staining in lung tissues and quanti ed data shown in I. Graphs show mean ± SEM of at least three independent experiments. **P<0.01, ***P<0.001, Student's t-test.