S-adenosylmethionine blocks collagen I production by preventing transforming growth factor-beta induction of the COL1A2 promoter.

To study the anti-fibrogenic mechanisms of S-adenosylmethionine (AdoMet), transgenic mice harboring the -17 kb to +54 bp of the collagen alpha2 (I) promoter (COL1A2) cloned upstream from the beta-gal reporter gene were injected with carbon tetrachloride (CCl4) to induce fibrosis and coadministered either AdoMet or saline. Control groups received AdoMet or mineral oil. AdoMet lowered the pathology in CCl4-treated mice as shown by transaminase levels, hematoxylin and eosin, Masson's trichrome staining, and collagen I expression. beta-Galactosidase activity indicated activation of the COL1A2 promoter in stellate cells from CCl4-treated mice and repression of such activation by AdoMet. Lipid peroxidation, transforming growth factor-beta (TGFbeta) expression, and decreases in glutathione levels were prevented by AdoMet. Incubation of primary stellate cells with AdoMet down-regulated basal and TGFbeta-induced collagen I and alpha-smooth muscle actin proteins. AdoMet metabolites down-regulated collagen I protein and mRNA levels. AdoMet repressed basal and TGFbeta-induced reporter activity in stellate cells transfected with COL1A2 promoter deletion constructs. AdoMet blocked TGFbeta induction of the -378 bp region of the COL1A2 promoter and prevented the phosphorylation of extracellular signal-regulated kinase 1/2 and the binding of Sp1 to the TGFbeta-responsive element. These observations unveil a novel mechanism by which AdoMet could ameliorate liver fibrosis.

polyamine biosynthesis in liver, and the precursor of reduced glutathione (GSH) through its conversion to cysteine by means of the transsulfuration pathway (1). Under normal conditions, most of the AdoMet generated is used in transmethylation reactions, in which methyl groups are added to compounds and AdoMet is converted to S-adenosylhomocysteine (SAH) (2).
In alcoholic liver disease, many of the enzymatic steps in methionine metabolism are affected (3). In the intragastric infusion model of ethanol feeding, hepatic levels of methionine, AdoMet, and DNA methylation decrease by ϳ40% (4). The hepatic methionine level depends on the stage of liver injury in rats fed ethanol (5). Reduced AdoMet levels and methylation can affect gene expression, membrane fluidity, and GSH levels in liver (2,6,7). Rats with decreased hepatic levels of AdoMet are predisposed to liver injury caused by lipopolysaccharide, and this effect is prevented by exogenous AdoMet treatment (8).
AdoMet has been used increasingly for the treatment of liver diseases, although the protective mechanisms remain unclear and are likely to be multiple. Impaired mitochondrial uptake of GSH has been postulated to be an important pathogenic factor in alcoholic liver injury. AdoMet administration restores GSH uptake into mitochondria in rats fed ethanol for 4 weeks by preventing changes in mitochondrial membrane fluidity (7). AdoMet inhibits tumor necrosis factor-␣ (TNF␣) release from macrophages (8), and it has also been used to prevent the development of hepatocarcinogenesis (9) and to increase survival in patients with alcoholic liver disease (10).
Excessive collagen I accumulation is the histopathologic hallmark of liver fibrosis. Central to the development and progression of fibrosis are cytokines that are normally involved in matrix remodeling; among them, transforming growth factor-␤ (TGF␤) enhances collagen I production and inhibits the synthesis of proteolytic enzymes that catalyze extracellular matrix degradation while enhancing the expression of protease inhibitors (11,12). As a result, TGF␤ is believed to play a critical role in liver fibrosis. There are no data as to whether administration of AdoMet may affect levels of TGF␤ itself and/or the actions of TGF␤ on collagen I expression and fibrosis.
The current study focused on analyzing potential mechanisms by which administration of AdoMet could decrease collagen I deposition under fibrogenic stimuli. Treatment with AdoMet was found to have a repressive effect on the COL1A2 promoter both in vivo and in vitro. We now propose that these antifibrogenic effects of AdoMet appear to be mediated, at least in part, by lowering TGF␤ levels and by inhibiting TGF␤ binding to the COL1A2 promoter through a Sp1-pERK1/2 coupled mechanism.

EXPERIMENTAL PROCEDURES
Most reagents unless otherwise stated were purchased from Sigma. Protein concentration was determined by the method of Lowry et al. (13) using the DC-20 protein assay kit (Bio-Rad).
Animal Study Design-Transgenic mice harboring the Ϫ17 kb to ϩ54 bp of the proximal promoter of the mouse COL1A2 gene cloned up-stream from the Escherichia coli ␤-gal reporter gene (LacZ) were used. These transgenic mice were obtained from Dr. Benoit de Crombrugghe (Department of Molecular Genetics, University of Texas, M. D. Anderson Cancer Center, Houston) (14 -18). Mice were bred in our institution and received humane care in compliance with the guidelines of the National Institutes of Health and the Animal Care Committee of Mount Sinai School of Medicine. Chronic liver injury was induced by injecting intraperitoneally 5 ml of CCl 4 /kg of body weight (25% v/v in mineral oil) three times a week for 4 weeks. AdoMet was administered intraperitoneally at a dose of 10 mg/kg body weight every day for 4 weeks and was always given 2 h before CCl 4 . Control groups were injected with mineral oil or mineral oil plus AdoMet alone. Mice were maintained at 22°C with a 12 h light/dark cycle, were fed standard chow diet, had free access to water, and were sacrificed under pentobarbital anesthesia 48 h after the last dose of CCl 4 .
Serum Transaminases-Blood was collected from the retro-orbital vein under anesthesia, centrifuged at 3,000 rpm for 3 min, and serum was separated to assay for alanine aminotransferase (ALT) and aspartate aminotransferase (AST) using kits from Sigma (Infinity ALT and AST).
Liver Histology and Immunohistochemistry-Liver samples were fixed in 10% buffered formalin and embedded in paraffin. Five-m sections were dehydrated and stained with hematoxylin and eosin (H&E) or Masson's trichrome and evaluated by a liver pathologist who was blinded from the experimental conditions. The Sirius red/fast green staining was carried out as according to López-De León and Rojkind (19) as described previously (20). Immunofluorescence was carried out using a rabbit IgG fraction to ␤-gal (ICN Biomedical), the reporter protein for the COL1A2 promoter activation, and a rabbit IgG to ␣-smooth muscle actin (␣-Sma), a specific marker for stellate cell activation. Alexa Fluor 488 dye-conjugated goat anti-rabbit antibody and Alexa Fluor 568 dye-conjugated goat anti-mouse antibody (Molecular Probes) were used for immunofluorescent staining of ␤-gal (corresponding to the active COL1A2 promoter) and ␣-Sma, respectively, and colocalization by overlaying both stainings using Adobe Photoshop computer software. TGF␤ immunostaining was carried out using a TGF␤ monoclonal antibody (Sigma) and a ImmunoCruz staining system.
␤-Gal Staining and Quantification-Samples were fixed in 10% formalin for 1 h, rinsed, and permeabilized in 0.1 M, pH 7.3, sodium phosphate, 2 mM MgCl 2 , 0.1% sodium deoxycholate, and 0.2% Nonidet P-40, incubated overnight with 1 mg/ml 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside plus 20 mM potassium ferri-and ferrocyanide in rinsing solution at 37°C. Samples were washed in 1 ϫ phosphatebuffered saline three times, followed by paraffin embedding and sectioning. Finally, 12-m liver slices were deparaffinized, rehydrated, and counterstained with nuclear fast red for 1 min, rinsed, and rehydrated. Each ␤-gal staining was also performed using frozen sections from the same tissues that were embedded in optimal cutting temperature compound and immediately frozen using methylbutane on dry ice. Proteins were extracted from individual livers, and ␤-gal activity was measured by means of a chemiluminescent reporter assay (Galactolight Plus, Tropix). ␤-Gal activity was expressed as units/mg of protein.
GSH Levels, Lipid Peroxidation, and CYP2E1 Activity-50 g of liver was homogenized in 5% trichloroacetic acid at a ratio of 1:10 (w/v) and centrifuged for 5 min at 8,000 rpm and 4°C. GSH levels were determined in the protein free extract by the enzymatic method of Tietze (24). Levels of malondialdehyde were measured in liver homogenates using a lipid peroxidation colorimetric assay kit (Calbiochem). Catalytic activity of CYP2E1 was determined as the rate of production of pnitrocatechol from p-nitrophenol (25).
Northern Blot Analysis-Northern blot analysis was performed as described previously (21,26) using cDNA probes for COL1A1 (provided by Dr. Scott L. Friedman, Mount Sinai School of Medicine) and COL1A2 (provided by Dr. Francesco Ramírez, New York Hospital for Special Surgery) and a cDNA for GAPDH, a housekeeping gene, from the ATCC.
Cell Culture and Transfection Experiments-All cell culture experiments were carried out using primary hepatic stellate cells (HSC) isolated by in situ liver perfusion and Histodenz gradient centrifugation (27). HSC were cultured in Dulbecco's modified Eagle's medium without serum from the beginning of each treatment. The doses of AdoMet, MTA, SAH, and methionine were either 10 or 30 M. TGF␤ was used at 10 ng/ml. Reporter DNA constructs containing upstream sequences of the human COL1A2 promoter linked to the chloramphenicol acetyltransferase (CAT) gene were provided by Dr. Francesco Ramírez (New York Hospital for Special Surgery) (28). In these constructs, human COL1A2 sequences span from Ϫ3500 to ϩ58 bp (Ϫ3500COL1A2-CAT), from Ϫ772 to ϩ58 bp (Ϫ772COL1A2-CAT), and from Ϫ378 to ϩ58 bp (Ϫ378COL1A2-CAT) (28). Parallel transfection of the corresponding empty vector pEMBL8-CAT at equivalent concentrations were performed. The total amount of plasmid DNA was equalized using pBluescript SKϪ (Stratagene). Cells were plated at a density of 10 5 /well in 6-well plates. Complexes containing FuGENE 6 (Roche Applied Science) plus plasmid DNA were prepared according to the manufacturer's instructions with a final concentration of plasmid DNA for each of the chimeric COL1A2 DNA constructs of 1 g/ml. Parallel cotransfection with 25 ng/ml of the control pRL-null (Promega) containing the cDNA encoding for Renilla luciferase was performed to normalize for transfection efficiency. Cells were incubated in the presence of the transfection mix for 24 h, after which the media were replaced and the cells treated with 30 M AdoMet for 24 h. Samples for the CAT activity were collected, and the reaction was run using a kit from Promega as described previously (29).
TGF␤-TGF␤ was measured by enzyme-linked immunosorbent assay using a kit from BIOSOURCE.
Gel Mobility Shift Assays-Nuclear extracts were prepared according to the method of Dignam et al. (30). For the gel mobility shift assays, a probe was generated by digestion with BglII and BstXI (Ϫ378 to Ϫ183 bp) of the Ϫ378COL1A2-CAT plasmid, which contains the TGF␤-responsive element (11,12,31,32). This probe was end labeled with [␥-32 P]ATP and T4 kinase (Promega). Binding reactions in a total volume of 10 l contained 5 g of the nuclear protein extract from HSC, 1 l of 10 ϫ binding buffer, 1 g/l poly[d(I-C)], 150 mM KCl, 0.05 mM ZnCl 2 , and 20,000 cpm of labeled probe at room temperature for 30 min. For competition studies, 200-fold of cold probe was added along with labeled probe. PAGE (6%) was performed at 150 V for 2 h in 0.25 ϫ TBE buffer, the gels were dried under vacuum and exposed overnight using Kodak films.
Southwestern Blot Analysis-Southwestern blot analysis was performed by the method of Singh et al. (33). The membrane containing the immunoprecipitated pERK1/2 bound to the immunoblotted Sp1 was blocked in 10% nonfat dry milk in TNE buffer (50 mM Tris, pH 7.5, 40 mM NaCl, 1 mM EDTA). DNA binding was carried out for 3 h with TNE buffer containing 5 g/ml poly[d(I-C)] and 2 ϫ 10 5 cpm/ml [␥-32 P]dCTP multiprime-labeled BglII and BstXI-digested fragment of the Ϫ378COL1A2-CAT plasmid containing the TGF␤-responsive element. Membranes were washed three times for 5 min each with TNE at room temperature and exposed in the PhosphorImager screen.
Statistical Analysis-Analyses of variance were performed for all experiments except for those in Fig. 4B, which were analyzed by an unpaired Student's t test. Values are expressed as the means Ϯ S.E. and are the average values of three and six experiments for the in vitro and for the in vivo studies, respectively.

Induction of Liver Fibrosis in Mouse COL1A2 Promoter
Transgenic Mice and Protection by AdoMet-Administration of CCl 4 increased the liver to body weight ratio ϳ15% over the non-CCl 4 -treated mice with no amelioration by AdoMet. Hepatocellular damage and protection by AdoMet were assayed by measuring serum levels of transaminases and by H&E staining. The CCl 4 treatment elevated ALT and AST serum levels 7and 3.5-fold, respectively. AdoMet lowered the elevated ALT levels in the CCl 4 -treated mice to control levels (Fig. 1A). H&E staining in the CCl 4 -treated mice revealed the presence of Mallory bodies, lymphocyte infiltration, centrilobular steatosis, and perivenular and pericellular fibrosis; AdoMet prevented these changes and minimized the pathology (Fig. 1B). Similarly, Masson's trichrome staining showed less endogenous total collagen deposition in mice treated with CCl 4 plus AdoMet than in the CCl 4 -treated mice (Fig. 1C). Quantitative analysis of total collagen in Sirius red-and fast green-stained liver sections showed a 2.5-fold increase by CCl 4 , which was prevented by AdoMet treatment (Fig. 1E). Collagen I expression assessed by Western blot analysis was elevated 3-fold by the CCl 4 treatment and reduced by about 50% by coadministration of AdoMet (Fig. 1D).
AdoMet Prevents the Activation of the Mouse COL1A2 Promoter in Vivo-Transgenic mice harboring the Ϫ17 kb to ϩ54 bp of the proximal promoter of the mouse COL1A2 gene cloned upstream from the E. coli ␤-gal LacZ reporter gene have been shown to be valuable in studies of activation of the COL1A2 gene in vivo (14,16,18). CCl 4 activated the COL1A2 promoter as shown by the ␤-gal-positive blue staining ( Fig. 2A). This activation was repressed in mice cotreated with AdoMet ( Fig.  2A). Quantification of the ␤-gal activity by chemiluminescence showed an approximate 27-fold increase in ␤-gal activity by CCl 4 which was reduced to a 5-fold increase in the presence of AdoMet (Fig. 2B). Immunofluorescence analysis of mice injected with CCl 4 revealed colocalization of ␣-Sma, a marker for stellate cell activation, and ␤-gal-positive cells, indicating that the activation of the COL1A2 promoter occurred in HSC (Fig. 2D).
AdoMet increased total GSH about 17% in control mice. When CCl 4 was injected, a 15% decrease in GSH levels was observed compared with the mineral oil-injected mice; however, cotreatment with AdoMet restored GSH to above control levels (Fig. 3A). AdoMet decreased lipid peroxidation by-products generated under CCl 4 treatment to the levels found in control mice (Fig. 3B). Immunostaining for 4-hydroxynonenal showed an extensive but diffuse pattern of positive staining, indicating development of oxidant stress in all liver cells (not shown). CYP2E1 expression and activity were lower in the CCl 4 -treated mice compared with the control mice because CCl 4 is known to lower CYP2E1 levels via radical inactivation and lipid peroxidation (34,35). AdoMet further decreased CYP2E1 activity and content in the presence of CCl 4 (Fig. 3D); this would decrease the potential for reactive oxygen species (ROS) production.
AdoMet Down-regulates Collagen I in Primary HSC-We analyzed the effect of AdoMet on the expression of collagen I protein in primary HSC. Neither cell viability (95%) nor cell proliferation assessed by [methyl-3 H]thymidine incorporation was affected by AdoMet treatment (data not shown). HSC were incubated for 1, 3, 5, and 7 days in the presence or absence of 10 or 30 M AdoMet added daily. Western blot analysis revealed a time-dependent up-regulation of collagen I production (Fig. 4A). AdoMet was very effective in lowering collagen I protein levels in cultured HSC. Part of the generated collagen I is usually secreted into the medium. AdoMet lowered basal intracellular as well as secreted collagen I protein (Fig. 4, A and B). The ability of AdoMet to prevent HSC activation in culture is shown by the 60% decrease in ␣-Sma levels at 7 days (Fig. 4B). Metabolites of AdoMet, such as SAH and MTA, and methionine, a precursor of AdoMet, were used at 30 M and found to decrease collagen I protein and COL1A1 and COL1A2 mRNA levels in HSC (Fig. 4, C and  D); however, the effects mediated by methionine were lower compared with those of the other treatments perhaps because of the extremely low expression of methionine adenosyltransferase 1A in HSC (36).
Transient transfection experiments with chimeric constructs harboring progressive 5Ј-deletions of the human COL1A2 promoter linked to the CAT reporter gene (see scheme in Fig. 5) were performed to identify the promoter regions required for the AdoMet-mediated repressive effect on collagen I production. Primary HSC were transfected with the constructs described in Refs. 28 and 29 or with the parental empty vector pEMBL8-CAT. As shown in Fig. 5, the basal acetylation of chloramphenicol in HSC transfected with the Ϫ3500COL1A2-CAT and the Ϫ378COL1A2-CAT plasmids was similar. On the other hand, the activity of the Ϫ772COL1A2-CAT vector was significantly lower. The Ϫ772 to Ϫ378 region of the human or mouse COL1A2 gene contains a silencer element (28,29,31). Addition of 30 M AdoMet for 24 h after transfection decreased the basal activity of both the Ϫ3500COL1A2-CAT and the Ϫ378COL1A2-CAT reporter vectors. It is interesting that AdoMet lowered CAT activity in cells transfected with the Ϫ378COL1A2-CAT construct, which contains the Ϫ378 to ϩ58 bp region that is essential for increased basal COL1A2 expression and responsiveness to oxidative stress and cytokines such as TNF␣ and TGF␤ (11,12,21,29,32,37).
Is TGF␤ Involved in the Effects of AdoMet on Collagen I?-TGF␤, one of the most important profibrogenic cytokines, can be secreted by HSC as well as by Kupffer cells. Immunostaining as well as Western blot analysis of liver samples from the COL1A2 promoter transgenic mice showed strong expression of TGF␤ under CCl 4 treatment, which was blunted by AdoMet cotreatment (Fig. 6, A and B). Future studies will evaluate whether in vivo AdoMet per se may be inhibiting TGF␤ production by Kupffer cells or by HSC or both, or whether it may induce its degradation. TGF␤ levels were evaluated in the culture medium of primary HSC incubated in the presence or absence of AdoMet. There was only a small increase in TGF␤ levels with time in culture and a modest decrease in the HSC incubated with AdoMet (Fig. 6C); these modest effects on TGF␤ levels are not consistent with the strong repression of collagen I protein and mRNA levels by AdoMet (Fig. 4). These results suggest that the down-regulation of TGF␤ production under CCl 4 plus AdoMet treatment in vivo (Fig. 6, A and B) is likely happening in Kupffer cells.
Experiments were next carried out to assess whether AdoMet could prevent the mechanism(s) involved in the TGF␤mediated stimulation of collagen I expression in HSC. Western blot analysis of HSC lysates showed a clear induction of collagen I expression by TGF␤ treatment compared with control cells, and this effect was totally blunted by coadministration of AdoMet 24 h prior to the TGF␤ addition (Fig. 6D). These results imply that AdoMet can affect collagen I protein likely by modulating TGF␤ levels and actions.
To address whether AdoMet could prevent the TGF␤-mediated transactivation of the human COL1A2 promoter, primary HSC were transfected with the Ϫ378COL1A2-CAT construct, which contains the TGF␤-responsive element (Ϫ313 to Ϫ250 bp from the transcription start site, Fig. 7B), and treated 24 h later with AdoMet. AdoMet blunted the TGF␤ transactivation of the COL1A2 promoter (Fig. 7A). Furthermore, gel mobility shift assays using the Ϫ378 to Ϫ183 bp region of the human COL1A2 promoter, containing the TGF␤-responsive element as a probe, confirmed the increased presence of bound proteins after TGF␤ treatment and showed that AdoMet administration prevented the formation of these TGF␤-stimulated DNA complexes (Fig. 7C). These complexes were competed using the cold probe containing the TGF␤-responsive element (Fig. 7C,  Comp lane).  Ramirez and collaborators (11,12,32,38) have located the TGF␤-responsive element of the human COL1A2 gene between nucleotides 313 and 250, relative to the transcription start site. The TGF␤-responsive element sequence consists of two nearly juxtaposed footprints (Boxes 3A and B), the most distal of which represents the 3Ј half of a larger footprinted area (Box A, Fig. 7B). Gel mobility shift assays documented increased binding to the TGF␤-responsive element of nuclear proteins from TGF␤-treated cells compared with those from control cells (Fig. 7C). In the gel mobility shift assays, the slowest migrating band (top) corresponds to Sp1 bound to the 3A region, the middle band is C/EBP bound to the 5A region, and the fastest migrating band (bottom) are Smad3 and Smad4 bound to the B region (11,12,32,38). The B region contains possible artificial binding sites for AP1 and NF-B when using short oligonucleotide probes. This same laboratory has demonstrated that C/EBPs are not essential for the TGF␤-mediated activation of the COL1A2 promoter, but they are for the TNF␣ repression of the COL1A2 promoter (11,12,32,38). We therefore evaluated whether AdoMet modulates the levels of these proteins that bind to the TGF␤-responsive element of the COL1A2 promoter. Levels of Sp1, Smad3, and Smad4, and C/EBP␤ and C/EBP␦ were similar in HSC treated with TGF␤ alone or cotreated with TGF␤ plus AdoMet; therefore, AdoMet does not alter the availability of these transcription factors (Fig. 7D).
TGF␤ induced the phosphorylation of p38 and ERK1/2 as early as 30 min. AdoMet prevented the phosphorylation of ERK1/2 but not that of p38 (Fig. 8A). Neither phosphatidylinositol 3-kinase nor phosphorylated Akt was affected by TGF␤ or AdoMet treatment (data not shown). Addition of an inhibitor of ERK1/2 phosphorylation, PD98059, prevented the increase in collagen I by TGF␤ (Fig. 8B), indicating that activated ERK1/2 is necessary for the induction of collagen I production by TGF␤. Phosphorylated ERK1/2 (pERK1/2) was immunoprecipitated and immunoblotted for either Sp1 or Smad3 and Smad4 to determine whether AdoMet prevention of phosphorylation of ERK1/2 could block its binding to either Sp1 or Smad3 or Smad4 transcription factors and the further transactivation of the COL1A2 promoter. pERK1/2 coimmunoprecipitated with Sp1 ( Fig. 8C) but not with Smad3 or Smad4 (not shown). This Sp1-pERK1/2 interaction was increased by TGF␤ treatment of the HSC, and AdoMet completely blunted the TGF␤ stimulation (Fig. 8C). Southwestern analysis demonstrated that pERK1/2 bound to Sp1 was able to bind the TGF␤-responsive element, and this binding was prevented by the addition of AdoMet (Fig. 8D).

DISCUSSION
During the course of liver fibrosis, AdoMet availability becomes compromised. AdoMet administration protects against liver damage elicited by a number of hepatotoxins, including ethanol, CCl 4 , galactosamine, acetaminophen, bile acids, or the administration of a choline-deficient diet (39 -41). The beneficial effect of AdoMet against CCl 4 -induced liver fibrosis has been attributed to protecting AdoMet synthase from oxidation of a cysteine residue in the ATP binding site of the enzyme (42), to reconstitution of the reduced GSH pool (3, 7, 42, 43), lower lipid peroxidation as a result of AdoMet antioxidant actions (44), and to the capacity of preserving plasma membrane Na ϩ / K ϩ -ATPase. AdoMet administration also seems to restore levels of antioxidants such as vitamin E, which can inhibit stimulation of collagen synthesis in fibroblasts by decreasing lipid peroxidation (45), but other potential protective mechanisms are now emerging.
Several studies of transcriptional control of the murine COL1A2 suggest that there are two regions of promoter sequences, a proximal promoter between Ϫ378 and ϩ58 bp, which is active in transfection assays and responsive to cytokines and ROS (46), and an upstream enhancer whose functional sequences are located between Ϫ17 and Ϫ15.5 kb (18). A critical element is located at Ϫ378 bp from the transcription start site because it contains a TNF␣, TGF␤, and ROS-sensitive region all of which can modulate collagen I levels (12,29,37). HSC are also activated by lipid peroxidation products such as malondialdehyde and 4-hydroxy-2,3-nonenal (47,48), which also stimulate collagen I synthesis (49,50). Up-regulation of TGF␤ and COL1A1 and COL1A2 genes occurs in HSC during CCl 4 -induced fibrosis (51). Freshly isolated HSC produce little collagen I protein; but when cultured, they gradually produce collagen I while acquiring a phenotype resembling myofibroblasts. Such a phenotypic change of cultured HSC is thus considered a useful system to evaluate the mechanism underlying the process of hepatic fibrosis. With these two models in mind as possible working tools, we set out to analyze the hypothesis that AdoMet could prevent the fibrogenic response by lowering TGF␤ levels and by repressing the COL1A2 promoter activation, and that a potential inhibition of the binding of TGF␤-sensitive transcription factors to the COL1A2 promoter may play a role in such regulation. As a first approach, transgenic mice harboring the fulllength mouse COL1A2 promoter upstream sequence linked to the ␤-gal reporter gene were used. Liver fibrosis was induced by injection of CCl 4 . Injury was observed by increased levels of transaminases, H&E staining, endogenous total collagen deposition, and elevated TGF␤ expression as has also been observed by others (52,53). Coadministration of AdoMet ameliorated these effects. Induction of the COL1A2 promoter was observed in HSC from the CCl 4 -treated transgenic mice, and AdoMet blocked such activation as demonstrated by ␤-gal staining, activity, and colocalization with ␣-Sma, a marker for activation of HSC.
Several potential mechanisms are likely to contribute to the protective actions of AdoMet. Although the CCl 4 -induced decrease in hepatic GSH and increase in lipid peroxidation were modest, AdoMet reduced lipid peroxidation and increased GSH FIG. 7. AdoMet (SAM) prevents TGF␤ binding to the human COL1A2 promoter in primary HSC. A, HSC were transfected with the Ϫ378COL1A2-CAT construct containing the TGF␤-responsive element (RE), and 24 h later the cells were incubated in the presence or absence of 10 ng/ml TGF␤ for an additional 24 h, and the CAT reaction was run. Numbers under the blots reflect AU of densitometry indicating acetylated chloramphenicol (AcC). fff, p Ͻ 0.001 for AdoMet versus control; ***, p Ͻ 0.001 for TGF␤ versus control; and OEOEOE, p Ͻ 0.001 for AdoMet ϩ TGF␤ versus TGF␤. SAM, S-adenosylmethionine. B, scheme of the TGF␤-responsive element in the human COL1A2 promoter. C, gel mobility shift with nuclear extracts from HSC under no treatment (Ϫ), with 10 ng/ml TGF␤, or with 30 M AdoMet plus 10 ng/ml TGF␤, using a radiolabeled probe containing the TGF␤-responsive element in the human COL1A2 promoter. In the last lane, the protein-DNA complexes were competed with a 200-fold of cold probe (Comp). D, Western blot analysis for the expression of Sp1, C/EBP␤, and C/EBP␦, and Smad3 and Smad4 in nuclear extracts from cells treated with TGF␤ or with TGF␤ plus AdoMet. Numbers under the blots reflect AU of densitometry for each protein corrected by the ␤-tubulin signal as a control for loading (n ϭ 2). content to above control values; these most likely reflect antioxidant actions of AdoMet and replenishment of GSH. Even though AdoMet alone had no effect, AdoMet plus CCl 4 downregulated the activity and protein expression of CYP2E1 over the decrease produced by CCl 4 alone. This further lowering of CYP2E1, a potential source of oxidative stress, may participate in the protective actions of AdoMet because of the sensitivity of the COL1A2 promoter to ROS and to ROS-activated cytokines (e.g. TNF␣ and TGF␤) (11,12,21,32,47,49). Importantly, AdoMet lowered the CCl 4 -induced elevation of TGF␤ to control levels. Thus, it is likely that several factors combine to promote the protective actions of AdoMet in vivo.
As a second approach and to gain mechanistic insight, HSC cultured in the presence or absence of AdoMet were studied.
There was a time-and dose-dependent down-regulation of intracellular and secreted collagen I and ␣-Sma proteins in the presence of AdoMet. Furthermore, two metabolites of AdoMet, SAH and MTA, were able to replicate these same effects at the protein and mRNA level. Methionine, an AdoMet precursor, also lowered collagen I protein and mRNA although less effectively, probably because of the minimal methionine adenosyltransferase 1A activity in HSC (36). Because SAH and MTA, unlike AdoMet, are not methylating agents, unless they generate AdoMet again, it is likely that the down-regulation of collagen I protein and mRNA by AdoMet may not be the result of gene silencing via methylation.
To assess whether the down-regulation of collagen I could be caused, at least in part, by repression of the collagen I promoter FIG. 8. AdoMet blocks ERK1/2 phosphorylation in TGF␤-treated HSC and lowers Sp1 binding to the COL1A2 promoter. A, primary HSC were isolated, placed in culture for 3 days, and treated with 30 M AdoMet 24 h prior to the addition of 10 ng/ml TGF␤. Samples were collected at 30, 60, and 120 min or 24 h. The figure depicts a Western blot analysis for collagen I, phosphorylated and total ERK1/2, and p38 under TGF␤ and TGF␤ plus AdoMet treatment. Numbers under the blots refer to AU of densitometry for collagen I and the pp38 to total p38 ratio or the pERK1/2 to total ERK1/2 ratio. SAM, S-adenosylmethionine. B, PD98059, an inhibitor of ERK1/2 phosphorylation, prevents the TGF␤-mediated increase in collagen I protein (n ϭ 1). C, pERK1/2 forms a complex with Sp1. pERK1/2 was immunoprecipitated and immunoblotted for Sp1 (n ϭ 1). D, Southwestern analysis of the pERK1/2-Sp1 complex bound to the probe containing the TGF␤-responsive element (RE) from the COL1A2 promoter. In all cases numbers under the blots refer to AU of densitometry with the control not treated HSC assigned a value of 1 (n ϭ 1).
(COL1A1 and COL1A2 mRNA levels were decreased by AdoMet), HSC were transfected with a series of deletion constructs for the human COL1A2 promoter. There was repression of the basal promoter activity in the presence of AdoMet. The Ϫ378 to ϩ58 bp of the human COL1A2 promoter contains a TGF␤ (Ϫ313 to Ϫ250), TNF␣, and a ROS-responsive element, and it is critical for basal and induced responsiveness (29,54,55). The in vitro experiments were not carried out under CCl 4 treatment because of the lack of cytochrome P450s, which carry out CCl 4 metabolism in HSC.
Based on the low TGF␤ expression in livers of mice injected with CCl 4 and cotreated with AdoMet, we measured the concentration of TGF␤ secreted to the culture medium by HSC treated with AdoMet and found only a slight decrease compared with control values. Therefore, it is unlikely that AdoMet inhibition of TGF␤ production by HSC plays a major role in the down-regulation of collagen I production by isolated HSC. The possibility that AdoMet could prevent the mechanisms and signaling pathways by which TGF␤ increases collagen I protein levels as a protective mechanism was then evaluated. When HSC were preincubated with AdoMet 24 h prior to the addition of TGF␤, collagen I expression was lowered to levels found in control HSC not treated with TGF␤, indicating that AdoMet prevents the stimulatory actions of TGF␤ on collagen I production.
A reporter assay using the Ϫ378COL1A2-CAT promoter region transfected into HSC and then treated with AdoMet plus TGF␤ showed repression of the basal and TGF␤-stimulated CAT activity. Furthermore, binding assays using the TGF␤responsive element as a probe indicated that administration of AdoMet could prevent TGF␤-sensitive proteins from binding to this site in the human COL1A2 promoter as confirmed by competition studies with cold probe containing the TGF␤-responsive element. Sp1, C/EBP␤, and C/EBP␦, and Smad3 and Smad4 availability was similar under TGF␤ and TGF␤ plus AdoMet treatments; thus, AdoMet did not lower the content of these transcription factors, essential for up-regulating the COL1A2 gene.
When the potential signaling pathways for the TGF␤-mediated effects on collagen I expression were evaluated (e.g. ERK, p38, phosphatidylinositol 3-kinase) (56), an increase in pERK1/2 and pp38 was observed in the presence of TGF␤. AdoMet pretreatment prevented the phosphorylation of ERK1/2 but not of p38. An inhibitor of ERK1/2 activation blocked TGF␤ stimulation of collagen I production, indicating that ERK1/2 was essential in the mechanism by which TGF␤ activates the COL1A2 promoter. We then analyzed whether pERK1/2 may be essential for the binding of critical transcription factors (e.g. Sp1 and Smad3 and Smad4) to the COL1A2 promoter. Immunoprecipitated pERK1/2 was immunoblotted for Sp1 and Smad3 or Smad4. Sp1 but not Smad3 or Smad4 (the latter not shown) was able to bind pERK1/2, and this was prevented by AdoMet. To address whether the complex of pERK1/2-Sp1 is essential for the TGF␤ effects on the transactivation of the COL1A2 promoter and whether AdoMet could prevent such effects, we performed a Southwestern analysis using the complex pERK1/2-Sp1 and a probe containing the TGF␤-responsive element and found that the pERK1/2-Sp1 complex binds the TGF␤-responsive element, and such binding was lowered by the addition of AdoMet.
These results suggest, therefore, that an important protective mechanism of AdoMet against liver fibrosis may be to lower TGF␤ levels (future studies will evaluate the effect of AdoMet and its metabolites on TGF␤ production by Kupffer cells) and the activation of collagen I production, and especially to repress the activation of the COL1A2 gene by preventing TGF␤ effects on its responsive site in the COL1A2 promoter.
Whether AdoMet further modifies the recruitment, affinity, and binding of other cofactors repressing the COL1A2 gene transcription is also conceivable. In addition, although these results demonstrated that AdoMet protected from a fibrogenic stimulus via decreasing TGF␤ levels and the interaction of TGF␤ with the COL1A2 promoter, the possibility of post-translational modifications on collagen I protein, secretion, and extracellular degradation affecting the net accumulation of collagen I in the space of Disse is not ruled out.