Potential conflict of interest: Nothing to report.
Address reprint requests to: Natalie Torok, M.D., Department of Gastroenterology and Hepatology, University of California Davis Medical Center, 4150 V Street, Suite 3500, Sacramento, CA 95817. E-mail: Natalie.Torok@ucdmc.ucdavis.edu; fax: 916-734-7908.
Advanced glycation endproducts (AGEs) accumulate in patients with diabetes, yet the link between AGEs and inflammatory and fibrogenic activity in nonalcoholic steatohepatitis (NASH) has not been explored. Tumor necrosis factor alpha (TNF-α)-converting enzyme (TACE) is at the center of inflammatory processes. Because the main natural regulator of TACE activity is the tissue inhibitor of metalloproteinase 3 (Timp3), we hypothesized that AGEs induce TACE through nicotinamide adenine dinucleotide phosphate reduced oxidase 2 (NOX2); and the down-regulation of Sirtuin 1 (Sirt1)/Timp3 pathways mediate fibrogenic activity in NASH. The role of NOX2, Sirt1, Timp3, and TACE was evaluated in choline-deficient L-amino acid defined (CDAA) or Western diet (WD)-fed wild-type (WT) and NOX2−/− mice. To restore Timp3, mice were injected with adenovirus (Ad)-Timp3. Sirt1 and Timp3 expressions were studied in livers from NASH patients, and we found that their levels were significantly lower than in healthy controls. In WT mice on the CDAA or WD, Sirt1 and Timp3 expressions were lower, whereas production of reactive oxidative species and TACE activity significantly increased with an increase in active TNF-α production as well as induction of fibrogenic transcripts. Ad-Timp3 injection resulted in a significant decline in TACE activity, procollagen α1 (I), alpha smooth muscle actin (α-SMA) and transforming growth factor beta (TGF-β) expression. NOX2−/− mice on the CDAA or WD had no significant change in Sirt1, Timp3, and TACE activity or the fibrosis markers assessed. In vitro, AGE exposure decreased Sirt1 and Timp3 in hepatic stellate cells by a NOX2-dependent pathway, and TACE was induced after exposure to AGEs. Conclusion: TACE activation is central to the pathogenesis of NASH and is mediated by AGEs through NOX2 induction and down-regulation of Sirt1/Timp3 pathways. (Hepatology 2013;58:1339–1348)
Diabetes mellitus (DM) is a major risk factor for disease progression with necroinflammation and fibrosis advancing to cirrhotic-stage nonalcoholic steatohepatitis (NASH).[1, 2] The factors implicated in this progression are poorly understood, specifically the effects of diabetes or insulin resistance (IR) on fibrogenesis. Advanced glycation endproducts (AGEs) are produced by a nonenzymatic glycation of serum proteins, and these modifications significantly influence the structure and function of key protein targets.[3, 4] AGEs are implicated in diabetic nephropathy, vascular complications, and retinopathy,[5, 6] but their role in inducing inflammatory or fibrogenic changes in the liver have not been adequately explored. Sirtuin 1 (Sirt1) belongs to the class III family of histone deacetylases, and its decreased activity was shown to be linked to the development of NASH. Accumulating evidence suggests that, in NASH, hepatic lipid metabolism pathways are affected by Sirt1 and low levels of Sirt1 are implicated in the development of steatosis in animals. Moreover, in the heterozygous Sirt1 knockout (KO) model, when chronically challenged with a 40% fat diet, mice became obese and insulin resistant, displaying increased serum cytokine levels and developing hepatomegaly. Hepatic metabolomic analyses revealed that Sirt1 heterozygous mice had elevated gluconeogenesis and oxidative stress (OS). Among the targets of Sirt1 is tissue inhibitor of metalloproteinases 3 (Timp3), a key regulator and inhibitor of the tumor necrosis factor alpha (TNF-α)-converting enzyme (TACE; also called A Metalloprotease and Disintegrin 17) activity. Timp3−/− mice have been described to develop vascular inflammation by increased TNF-α as well as hepatic steatosis.
In this study, we showed that exposure to AGEs leads to down-regulation of Sirt1 and Timp3 in hepatic stellate cells (HSCs) by activation of NADPH oxidase (nicotinamide adenine dinucleotide phosphate reduced oxidase) 2 (NOX2) and the release of reactive oxidative species (ROS). Accordingly, wild-type (WT) mice fed the choline-deficient L-amino acid defined (CDAA) or western diet (WD) showed decreasing Sirt1, Timp3 expression, and an increase in TACE activity, TNF-α production, and increased fibrogenic activity. Correction of the low TIMP3 levels by injection of adenoviral (Ad)-Timp3 into CDAA or WD-fed mice reduced TACE activity, TNF-α, expression of the receptor of advanced glycation endproducts (RAGE), and fibrogenic response. NOX2−/− mice on NASH diets did not develop increased TACE activity or fibrogenic response. In summary, these data suggest an important role of AGEs in NOX2-mediated induction of TACE, TNF-α, and fibrogenic activity during NASH progression.
Materials and Methods
Liver Biopsy Samples
Liver biopsy samples were obtained from the University of California Davis Cancer Center Biorepository (Sacramento, CA) funded by the National Cancer Institute (Bethesda, MD). Samples from 6 different patients and 6 normal livers were tested. All patients had IR or DM.
C57BL/6 mice (The Jackson Laboratory, Bar Harbor, ME) or B6.129S-Cybbtm1Din/J (NOX2−/−; Jackson Lab) were given choline-supplemented L-amino acid-defined (CSAA) or CDAA diet (Dyets Inc., Bethlehem, PA) for 10 weeks or control chow. Both diets contain higher calories, fat, and carbohydrates than standard chow. The fast-food (“Western”) diet contains 40% energy as fat (12% saturated fatty acids, 2% cholesterol; AIN-76, Western diet; TestDiet, St. Louis, MO) and is supplemented with high-fructose corn syrup (42 g/L). To test for the role of Timp3 in TACE regulation, Ad-Timp3 was injected on week 8 through the tail vein (2 × 107 plaque-forming units [pfu]/200 uL of phosphate-buffered saline [PBS]; Applied Biological Materials Inc., Richmond, British Columbia, Canada). As control, Ad-GFP (green fluorescent protein; 2 × 107 pfu/200 uL; Vector Biolabs, Philadelphia, PA) was used. In a group of mice, GdCl3 (10 mg/kg in saline; Sigma-Aldrich, St. Louis, MO) was injected intraperitoneally every other day throughout the experiment to inhibit macrophages in both models. Serum and liver tissue were collected and alanine aminotransferase (ALT) and bilirubin were tested. Tissue was processed for the further assays. Animals were housed in facilities approved by the National Institutes of Health (Bethesda, MD). All procedures were reviewed and approved by the animal welfare committee of the University of California Davis.
Primary HSCs were isolated either from C57BL/6 mice or Sprague-Dawley rats as described previously. Cells were cultured in medium 199/20% fetal bovine serum (FBS). Primary HSCs were used for experiments within the first 3 days after isolation. Cells were exposed to AGEs (50 μg/mL) or bovine serum albumin (BSA) in serum-free medium for 16 hours and/or transfected with small interfering RNAs (siRNAs).
For siRNA transfection, primary rat HSCs were cultured as described above for 1 day, then the medium was changed to Dulbecco's modified Eagle's medium, 0.5% FBS. HSCs were transfected with the siRNA to Sirt1 or Timp3 (Santa Cruz Biotechnology Inc., Santa Cruz, CA) or scrambled siRNA using the RiboJuice transfection reagent (EMD Chemicals Inc., Darmstadt, Germany), according to the manufacturer's instructions.
Exposure to AGEs Induces Sirt1 and Timp3 Down-Regulation by NOX2 in Primary HSCs
Activation of NOXs is one of the main sources of ROS in activated HSCs. We have previously shown that NOX2 is an important factor in HSC transdifferentiation. To study the role of AGEs in ROS production, primary cells were transfected with NOX2 or scrambled siRNA and incubated in serum-free conditions with glycolaldehyde (GA)-derived AGEs for 16 hours. Superoxide production was significantly increased in control (untransfected) and scrambled siRNA-transfected HSCs (Fig. 1A; 1.74- ± 0.13-fold; P < 0.05), whereas this was attenuated in NOX2 siRNA-transfected cells (Fig. 1A). To study the effects of AGEs/ROS on Sirt1 and Timp3 expression, real-time quantitative polymerase chain reaction (qPCR) was done on cells from the above-described experiment. Sirt1 and Timp3 were down-regulated in response to AGEs in a NOX2-dependent manner (Fig. 1B; 0.38- ± 0.14-fold; P < 0.05; 0.58- ± 0.10-fold; P < 0.05, respectively). To test whether Sirt1 directly targets Timp3 in primary HSCs, cells were transfected with Sirt1 siRNA, and Timp3 expression was found to be down-regulated in response to inhibition of Sirt1 (0.36- ± 0.16-fold; N = 3; P < 0.05; Fig. 1C).
AGEs Induce TACE Activity in HSCs
Because Timp3 is the main natural inhibitor of TACE activity, we next examined whether TACE could be induced by AGEs. This was studied by two different methods: fluorometry (Fig. 2A) and immunoprecipitation (IP) with western blotting to assess TACE tyrosine phosphorylation, which correlates with the active state of the enzyme (Fig. 2B). TACE activity was induced by AGEs in scrambled siRNA-transfected HSCs (1.48- ± 0.06-fold; P < 0.01; Fig. 2A), and this was significantly attenuated in NOX2 siRNA-transfected HSCs (1.10- ± 0.09-fold; P < 0.05). In addition, inhibiting Sirt1 by siRNA also resulted in an increase in TACE activity (1.43- ± 0.10-fold; P < 0.05), which was further increased by AGEs (2.29- ± 0.23-fold; P < 0.05). TACE was phosphorylated after exposure of HSCs to AGEs, indicating the activated state of the enzyme (Fig. 2B). To discern whether the other major cell types in the liver contribute to AGE-induced TACE activation, primary hepatocytes, HSCs, and Kupffer cells (KCs) were incubated either with BSA or AGEs. TACE activity of hepatocytes was very low and not induced by AGEs (Fig. 2C). Only HSCs responded with a significant increase (1.73- ± 0.14-fold; P < 0.05), and whereas the baseline activity of KCs was higher than that of hepatocytes; no induction was noted after incubation with AGEs.
Sirt1 and Timp3 Are Down-Regulated in Livers of Patients With NASH
To study Sirt1 and Timp3 in human livers with NASH, real-time qPCR was performed on liver biopsy samples from different patients with grade 2-3, stage 3-4 NASH, and healthy controls (HCs; Fig. 3). The patients represented here had either IR with fasting hyperglycemia or type II DM. Expression of both Sirt1 and Timp3 were significantly decreased in NASH livers, compared to normal HCs (expressed as fold over control: 0.47- ± 0.11-fold; P < 0.05; and 0.24- ± 0.07-fold; P < 0.01, respectively; N = 6; Fig. 3). Expression of RAGE was also significantly increased in patients with NASH (3.93- ± 0.2-fold; P < 0.05).
NOX2-Dependent ROS Production Is Increased in Two Different Diet Models of NASH and Is Involved in the Regulation of Sirt1 and Timp3
To recapitulate our in vitro findings, WT or NOX2−/− mice were fed either with CDAA, CSAA, or WD. The CDAA model was shown to recapitulate the features of human NASH with IR, inflammatory cell infiltration, hepatocyte death, and liver fibrosis,[19, 20] whereas mice on the CSAA diet mainly develop steatosis, but no necroinflammation or fibrosis. The advantage of using the CSAA/CDAA model is that we can correlate our findings to a model resulting in simple steatosis or progressive NASH. To confirm our findings, we also performed experiments in mice fed the WD supplemented with high fructose. This diet was shown to have higher fidelity to human pathophysiology, with mice exhibiting significant weight gain, metabolic syndrome, and diabetes and histologically developing ballooning, lipoapoptosis, necroinflammation, and fibrosis. Body weight has increased similarly both in WT and NOX2−/− mice on both NASH diets (Fig. 4A). WT mice on CDAA or WD had significantly increased ALT (P < 0.01), compared to mice on the CSAA or control chow diet. Mice on CDAA and WD had increased glucose levels (216.0 ± 36.2 mg/dL and 239.4 ± 12.8 mg/dL, respectively), and these diets were previously shown to induce impaired glucose tolerance or diabetes.[15, 16] AGEs were increased in both diet models in WT but not in the NOX2−/− mice (Supporting Fig. 1). The increase in bilirubin in WT CDAA mice may reflect impaired synthetic function, because these mice had more-severe steatohepatitis and fibrosis. In NOX2−/− mice, ALT showed an increasing trend on NASH diets, compared to baseline, albeit not significant, and below the level of liver injury observed in WT mice, whereas bilirubin had not changed. On histology, all mice on CDAA or WD displayed increased steatosis (Fig. 4B).
ROS production had significantly increased in WT mice on CDAA or WD (Fig. 4C,D; 1.57- ± 0.16-fold; P < 0.05; and 1.75- ± 0.01-fold; P < 0.05), whereas this was attenuated in NOX2−/− mice in both models (to 0.94- ± 0.11-fold and 0.89- ± 0.26-fold; P < 0.05). To correlate the data to humans with NASH, qPCR was performed in all experimental conditions (Fig. 4E,F). WT CDAA- or WD-fed mice showed a significant decrease in Sirt1 (to 0.10- ± 0.04-fold; P < 0.01; and 0.11- ± 0.031-fold; P < 0.05, respectively; N = 6) and Timp3 expression (to 0.21- ± 0.04-fold; P < 0.05; and 0.43- ± 0.01-fold; P < 0.05, respectively; N = 6). In contrast, in NOX2−/− mice, Sirt1 and Timp3 expression was not affected by CDAA or WD.
Correction of Low Timp3 Expression in WT Mice on CDAA or WD by Ad-Timp3 Ameliorates TACE and TNF-α Activity
To confirm the causal link between low Timp3 expression leading to an unopposed increase in TACE activity and consequent induction of TNF-α activity in vivo, we injected Ad-Timp3 or Ad-GFP into WT mice on CDAA or WD. Correcting Timp3 levels resulted in a significant decrease in TACE activity (Fig. 5A,B) and TNF-α activity (Fig. 5C,D) in both diet models. TACE and TNFα activity showed no significant changes in NOX2−/− mice on either diet. Because KCs also express NOX2, a group of mice in both models were also injected by GdCl3 to inhibit the function of these cells and elucidate the respective role of HSCs and KCs in AGEs-induced TACE activation in NASH. Macrophage inhibition did not change TACE activity in either diet model, but showed a decreasing trend for active TNF-α. Triglyceride (TG) content of the liver has significantly increased on both diets in both genotypes, and Ad-Timp3 did not affect TG content (Fig. 5E,F; P < 0.05; N = 6 for each group).
Interestingly, expression of RAGE showed an increase in livers of WT mice on the CDAA diet (2.34- ± 0.17-fold; P < 0.05; Supporting Fig. 2), and this was attenuated by Ad-Timp3 (0.86- ± 0.14; P < 0.05). No increase in RAGE was noted in NOX2−/− mice.
CDAA and WD Induce Fibrogenic Activity in WT Mice, Which Is Attenuated After Ad-Timp3 Injection and in NOX2−/− Mice
Correction of Timp3 may affect fibrogenesis. Therefore, procollagen α1(I), alpha smooth muscle actin (α-SMA), and transforming growth factor beta (TGF-β) were tested by real-time qPCR, and liver tissues were examined after Picro Sirius Red staining and hydoxyproline assay. Procollagen α1(I), α-SMA, and TGF-β expression were significantly elevated both in CDAA-fed (5.91- ± 1.85-fold; P < 0.01; 1.41- ± 0.07-fold; P < 0.01; and 3.37- ± 0.76-fold, respectively; P < 0.05; N = 6; Fig. 6A) and WD-fed mice (1.58- ± 0.12-fold; P < 0.05; 1.75- ± 0.001-fold; P < 0.05; and 1.51- ± 0.13-fold; P < 0.05). Expression of these fibrogenic transcripts significantly decreased after Ad-Timp3 injection in both mouse models: in CDAA-diet–fed mice procollagen α1(I) (to 1.54- ± 0.93-fold; P < 0.05); α-SMA (to 1.05- ± 0.31-fold; P < 0.05); and TGF-β (to 0.96- ± 0.1-fold; P < 0.05). In WD-fed mice after Ad-Timp3 transduction, procollagen α1(I) decreased (to 0.76- ± 0.1-fold; P < 0.05), as did α-SMA (to 0.62- ± 0.09; P < 0.05) and TGF-β (to 0.63- ± 0.1-fold; P < 0.05). These data were confirmed also by the Picro Sirius Red staining and morphometry (Fig. 6C-E) and hydroxyproline assay (Fig. 7). Injection of GdCl3 caused a decrease in expression of α-SMA and TGF-β in the CDAA model, but not in the WD. In NOX2−/− mice on the NASH diets, fibrosis was less pronounced with attenuated procollagen α1(I) and α-SMA expression as well as a significant decrease in Picro Sirius Red positive area (Figs. 6 and 7; P < 0.05) and hydroxyproline quantity (Fig. 7; P < 0.01).
IR and DM are major risk factors in the progression of NASH. Several important aspects of the pathogenesis are well defined, but how diabetes creates a proinflammatory and fibrogenic milieu is less clear. Here, we have shown that AGEs/NOX2/ROS-induced decrease in Sirt1 expression in HSCs culminates in an induction of TACE and, consequently, TNF-α activity. Loss of Sirt1 was shown to be associated with metabolic diseases, such as type 2 diabetes, atherosclerosis, and NASH. On the other hand, hepatic Sirt1 deficiency was described to induce IR, and an accumulation of ROS was also detected in the liver. Treating these mice with antioxidants reversed the phenotype, suggesting the key role of ROS in this process. Defining the source of ROS and analyzing the downstream effects of Sirt1 down-regulation are of great importance.
AGEs are formed at a highly accelerated rate during hyperglycemia by a nonenzymatic glycation of serum proteins. The process begins with the conversion of reversible Schiff base adducts to more stable, covalently bound products and, eventually, the irreversibly bound moieties known as AGEs are formed.[22, 23] AGE levels in serum or liver were found to be elevated in patients with steatohepatitis, compared to HCs or patients with simple steatosis. Interaction of AGEs with the associated cell-surface receptor, RAGE, has been linked to induction of OS in the liver, and it was postulated that activation of one of the NOXs was the source of ROS production. However, the specific NOX has not been identified. Here, we described the key role of NOX2 activation in AGE-induced ROS production and the down-regulation of Sirt1/Timp3 pathways. NOX2 is a phagocytic NADPH oxidase, and we and others have shown that it is highly expressed and enzymatically active in HSC during liver fibrosis.[18, 26] NOX2 directly induces HSC activation and production of collagen I by inducing promoter activity by H2O2, and NOX2−/− mice develop significantly less fibrosis in bile duct ligation and CCl4 models. AGEs have been described to induce ROS formation by NOXs in other systems,[27, 28] and the postulated mechanism could involve activation of protein kinase C (PKC)-α in the kidney or PKC-δ in neuronal tissue by AGEs. Whether or not in active HSCs other NOXs can elicit a similar response to AGEs could be further investigated in the future. Because p47phox KO HSCs had a decrease in ROS production after incubation with AGEs, it is possible that NOX1 also plays a role as a regulator of Sirt1, because p47phox is a common subunit to both NOX1 and 2. In our study, we demonstrated that, in NOX2−/− mice, there was a significant attenuation of fibrosis on both diets. Because KCs also express NOX2, we treated mice with gadolinium to inhibit their function. We found that this did not interfere with TACE activity in the liver. TACE activity of KCs was not induced by AGEs in vitro and therefore their profibrogenic role during NASH could be attributed to other, more dominant mechanisms, such as Toll-like receptor 4–mediated induction of HSCs. Steatosis was not affected by the lack of NOX2, corresponding to earlier data from the methionine- and choline-deficient model, nor was it influenced by correcting Timp3. This suggests that AGEs mainly induced oxidative and inflammatory pathways in these models. The role of Sirt1 down-regulation in HSCs during NASH progression has not been studied in detail. Because the main source of TACE activity in the liver are activated HSCs, and the natural regulator of TACE is Timp3, we sought to determine whether low Sirt1 could translate into a down-regulation of Timp3 and consequent induction of TACE. First, we confirmed that active HSCs are a potent source of TACE production in the liver and that exposure to AGEs directly induces pathways leading to TACE activation, whereas hepatocytes and KCs do not respond to AGE induction. Second, AGEs induced a NOX2-dependent down-regulation of Sirt1 and Timp3, and both of these transcripts were down-regulated in humans with NASH and also in mice on both diets. Correlating to this, TACE activity was increased by AGEs and also in WT mice on CDAA and WD, but not on the CSAA diet (steatosis). In contrast, no increase in TACE activity was noted in NOX2−/− mice on CDAA and WD, implying that NOX2-derived ROS plays a key a role in TACE activation. To confirm the role of Timp3 in NASH, we injected a group of mice on the CDAA diet with Ad-Timp3. This resulted in a decline in TACE and TNF-α activity as well as RAGE expression. In addition, Ad-Timp3 transduction also improved fibrosis with down-regulation of profibrogenic transcripts and lowering of hydroxyproline content. Because TACE targets multiple other cytokines and chemokines and their receptors, reducing its activity may have led to a decreased production of fibrogenic mediators. For instance, TACE is known to induce ectodomain shedding of fractalkine (CX3CL1); thus, decrease in fractalkine could have also contributed to the improvement in fibrogenic activity. Mice deficient in Timp3 demonstrated elevated levels of TNF-α and developed IR and hepatic steatosis, mediated by increased TACE activity.[10, 21]
In conclusion, we have demonstrated the central role of the Sirt1/Timp3/TACE cascade in AGE-induced proinflammatory and fibrogenic activity in NASH. Modulation of Timp3 or TACE activity could thus become a successful approach to halt disease progression in NASH.