Loss of SOCS3 in the Liver Promotes Fibrosis by Enhancing STAT3

Oncogene (2006) 25, 2520–2530 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc ORIGINAL ARTICLE Loss of SOCS3 in the liver promotes ﬁbrosis by enhancing STAT3-mediated TGF-b1 production

H Ogata1,2, T Chinen1, T Yoshida1,3, I Kinjyo1, G Takaesu1, H Shiraishi1, M Iida2, T Kobayashi1 and A Yoshimura1

1Division of Molecular and Cellular Immunology, Medical Institute of Bioregulation, Fukuoka, Japan and 2Department of Medicine and Clinical Science, Graduate School of Medical Science, Kyushu University, Fukuoka, Japan

Recently, DNA methylation and reduced expression of the Introduction suppressor of the cytokine signaling-3 (SOCS3) gene in human hepatocellular carcinoma (HCC) patients have Most hepatocellular carcinomas (HCCs) develop as a been reported. However, the roles of SOCS3 in HCC result of liver cirrhosis and advanced fibrosis. IL-6 has development in vivo have not been clarified. Using RT– been implicated in the promotion of inflammation, PCR analysis and Western blotting, we confirmed that fibrosis, and carcinogenesis in the liver. This is because SOCS3 expression was reduced in HCC patients. high levels of IL-6 induce hepatic inflammation, However, reduced expression of SOCS3 occurred not collagen synthesis, tissue inhibitor of metalloprotei- only in HCC but also in nontumor regions, and this nase-1 (TIMP1) expression, and early transition of reduction was stronger as the fibrosis grade increased. stellate cells (also called Ito cells and lipocytes) into Furthermore, SOCS3 levels were inversely correlated with myofibroblasts in mice (Choi et al., 1994). Hyperactiva- signal transducers and activators of transcription-3 tion of the signal transducers and activators of (STAT3) activation as well as transforming growth factor transcription-3 (STAT3), which is a downstreamof (TGF)-b1 levels in the non-HCC region. To define the IL-6/gp130, has been reported in HCC (Yoshikawa molecular consequences of SOCS3 silencing/STAT3 et al., 2001). STAT3 is believed to contribute to cell hyperactivation and liver fibrosis, we examined liver- proliferation and antiapoptosis by inducing c-myc, specific SOCS3-deficient mice. We demonstrated that cyclin-D1, and Bcl-XL and, therefore, functions as an SOCS3 deletion in the liver resulted in hyperactivation of oncogene (Bromberg et al., 1999; Bromberg and STAT3 and promoted ConA-and chemical-induced liver Darnell, 2000). Recently, STAT3 was shown to induce fibrosis. The expression of TGF-b1, a mediator of fibrosis, the vascular endothelial growth factor (VEGF), which was enhanced by SOCS3 gene deletion, but suppressed by contributes to tumor-associated angiogenesis. Further- the overexpression of a dominant-negative STAT3 or more, STAT3 suppresses the tumor expression of SOCS3 both in vivo and in vitro. These data suggest that proinflammatory cytokines and chemokines, resulting TGF-b1 is a target gene of STAT3 and could be one of the in the reduced activation of antitumor immunity (Wang mechanisms for enhanced fibrosis in SOCS3-deficient et al., 2004). Therefore, STAT3 is deeply involved in mice. Thus, our present study provides a novel role of HCC development by multiple mechanisms (Wei et al., SOCS3 and STAT3 in HCC development: in addition to 2003). STAT3 is activated not only by IL-6 and related the previously characterized oncogenic potentials, STAT3 cytokines, but also by direct interaction with the enhances hepatic fibrosis through the upregulation of hepatitis C virus (HCV) core protein in the liver TGF-b1 expression, and SOCS3 prevents this process. (Yoshida et al., 2002). Oncogene (2006) 25, 2520–2530. doi:10.1038/sj.onc.1209281; Cytokines, including IL-6, activate the Janus kinase published online 13 February 2006 (JAK)/STAT signaling pathway, which plays a vital role in various conditions, including immune reaction, cell Keywords: SOCS; STAT; TGF-b1; liver fibrosis growth, and differentiation. Cytokine signaling is strictly regulated by suppressor of the cytokine signaling (SOCS) family proteins. Among them, SOCS1 is relatively specific to STAT1, and SOCS3 to STAT3 (Kinjyo et al., 2002; Croker et al., 2003; Yasukawa et al., Correspondence: Dr A Yoshimura, Division of Molecular and Cellular 2003). SOCS1 inhibits JAK activation through its N- Immunology, Medical Institute of Bioregulation, Kyushu University, terminal kinase inhibitory region (KIR) by direct 3-1-1 Maidashi, Higashi-ku, Fukuoka 812-8582, Japan. binding to the activation loop of JAKs, while SOCS3 E-mail: [email protected] inhibits JAK kinase by binding to the cytokine receptors 3Current address: Research Center for Innovative Cancer Therapy, Kurume University School of Medicine, Kurume 830-0011, Japan. through its SH2 domain. Furthermore, the SOCS Received 30 August 2005; revised 5 October 2005; accepted 25 October box, another domain in SOCS proteins, is able to 2005; published online 13 February 2006 interact with Elongins B and C and couples the SH2 SOCS3 regulates hepatic fibrosis and TGF-b1 H Ogata et al 2521 domain-binding proteins to the ubiquitin–proteasome Results pathway (Kamura et al., 1998; Kamizono et al., 2001). Therefore, SOCS1 and SOCS3 inhibit cytokine signaling Correlation between STAT3 activation, SOCS by suppressing JAK kinase activity and promoting the expression, and fibrosis in human HCC degradation of the activated cytokine receptor complex. We investigated the correlation between STAT3 activa- SOCS1 induction is restricted to interferons and some tion and the expression levels of SOCSs in HCC and other cytokines, while SOCS3 is induced by a variety of noncancerous tissues in human patients. As reported, we stimuli, such as interleukins, interferons, and TNFa confirmed that both SOCS1 and SOCS3 mRNA levels (Bode et al., 1999; Yasukawa et al., 2003; Kamio et al., were lower in the HCC region than in the nontumor 2004). However, since binding of the SH2 domain region (data not shown). However, we noticed that the of SOCS3 is relatively specific to the receptors for SOCS3 expression levels in the non-HCC region were STAT3-activating cytokines, such as IL-6, G-CSF, variable and often very low in samples with severe and leptin, the effect of SOCS3 may be relatively fibrosis. Therefore, we compared the correlation be- restricted to STAT3 (Croker et al., 2003, 2004; Mori tween fibrosis and STAT3 activation and SOCS3 et al., 2004). expression in non-HCC regions. We also examined the In connection with SOCS and HCC, the SOCS1 gene levels of TGF-b1, an essential mediator of fibrosis has been implicated as an antioncogene for HCC (Nakamura et al., 2000). The progression of liver development. Yoshikawa et al. (2001) reported aberrant fibrosis was classified into four stages, fromF1 to F4, methylation in the CpG island of SOCS1 that correlated as judged by histological examination (Bedossa and with its transcriptional silencing in HCC cell lines. In support of this, a recent experiment has shown that SOCS1 heterozygous mice are hypersensitive to diethyl- nitrosamine (DEN)-induced hepatocarcinogenesis (Yoshida et al., 2004). A simple explanation is that STAT3 hyperactivation is the result of reduced expression of SOCS1 by DNA methylation, which leads to hyperproliferation and antiapoptosis of hepatocarcino- ma cells. However, the effect of SOCS3 deletion on STAT3 hyperactivation has not been demonstrated in vivo, and, as we reported, STAT3 is mostly regulated by SOCS3 rather than SOCS1 (Kimura et al., 2004; Mori et al., 2004). Recently, reduced expression of SOCS3 and aberrant methylation in the CpG island of the SOCS3 locus in HCC patient samples and cell lines have been reported (Niwa et al., 2005). These researchers also reported that SOCS3 negatively regulates cell growth and cell motility by inhibiting STAT3 and FAK signalings in cultured HCC cell lines. However, the effect of SOCS3 gene silencing in the liver in HCC development and STAT3 activation has not been established in vivo. In this study, using RT–PCR and Western blotting, we confirmed that the expression of SOCS3 was reduced in the liver of human HCC patients. However, the reduction of SOCS3 expression was observed not only in HCC regions but also in non-HCC regions. SOCS3 expression levels were inversely correlated with STAT3 activation and fibrosis. Thus, we investigated the role of SOCS3 in liver fibrosis using liver-specific SOCS3-deficient mice. We demonstrated that STAT3 hyperactivation promoted chemical compound- and concanavalin-A (ConA)-induced liver fibrosis. Part of this mechanism is explained by the transcrip- Figure 1 Enhanced STAT3 activation and reduced expression tional induction of transforming growth factor levels of SOCS3 in human liver tissues. (a, b) Relative expression levels of the TGF-b1(a) and SOCS3 (b) mRNA in each fibrosis (F) (TGF)-b1, an essential mediator of fibrosis, through stage and HCC, as determined by real-time RT–PCR. N ¼ 5 for STAT3. These findings suggest that SOCS3 contributes each stage. (c, d) Western blot analysis of phosphorylated STAT3 to the suppression of HCC development by protecting (pY-STAT3) and SOCS3 in human liver lysates. Data are the liver cells against hepatic fibrosis mediated by representative of three separate experiments. In (d), STAT3 phosphorylation levels are quantified by densitometry and normal- TGF-b1. Our study also provides a basis for the ized to STAT3 levels. The data are shown as the mean7s.e.m. therapeutic potential of liver cirrhosis/HCC by STAT3 (number of each F stage ¼ 5), *Po0.05; **Po0.01; NS, no antagonists. significant difference.

Oncogene SOCS3 regulates hepatic fibrosis and TGF-b1 H Ogata et al 2522 Poynard, 1996). As shown in Figure 1a, the TGF-b1 gene deletion was specific to the liver and did not occur mRNA levels, determined by real-time RT–PCR, were in other tissues, including the brain, heart, kidney, fat, well correlated with the fibrosis grade. As expected, the and muscle (Figure 2c). The AdCre-mediated SOCS3 SOCS3 mRNA (Figure 1b) and protein (Figure 1c) deletion in the liver continued for more than 7 weeks levels, as determined by Western blotting, decreased as (Figure 2c). SOCS3 deletion in the liver did not occur the fibrosis grade increased. Furthermore, the activation when a control adenovirus carrying LacZ (AdLacZ) was of STAT3 increased as the fibrosis grade increased injected. In the AdCre-treated mice, the SOCS3 gene (Figure 1c and d). These data suggest that the SOCS3 was deleted in both the parenchymal and nonparench- levels were inversely correlated with liver fibrosis, TGF- ymal cells (Figure 2d). b1 expression, and STAT3 activation. SOCS3 deficiency in the liver promotes chemical Generation of liver-specific SOCS3-deficient mice compound-induced liver fibrosis To demonstrate that the reduction of SOCS3 gene Administration of dimethylnitrosamine (DMN) is a expression causes STAT3 activation and TGF-b1 well-established chronic liver fibrosis model in mice. An expression, we investigated experimental liver fibrosis intraperitoneal injection of a small amount of DMN has models using liver-specific SOCS3 knockout (KO) and been shown to induce fatty degeneration of hepatocytes, SOCS3 þ /À mice. Since SOCS3-deficient mice die as a activation and proliferation of hepatic stellate cells, result of placental defects during embryonic develop- infiltration by macrophages, and secretion of TGF-b1 ment (Roberts et al., 2001), a conditional KO approach fromvarious types of cells, which promotesfibrosis. has been used to determine the tissue-specific roles DMN was intraperitoneally (i.p.) injected into of SOCS3 (Yasukawa et al., 2003). To generate a SOCS3 þ / þ , SOCS3 þ /À, AdCre-, and AdLacZ-SOCS3fl/fl liver-specific deletion, SOCS3flox/flox(fl/fl) mice were intra- mice. The mortality rate and serum aspartate amino- venously (i.v.) injected with an adenovirus carrying transferase (AST) and alanine aminotransferase (ALT) Cre-recombinase (AdCre) (Figure 2a and b). SOCS3 levels in SOCS3 þ /À mice or AdCre-SOCS3fl/fl mice after

Figure 2 Deletion of SOCS3 in livers. (a) Wild-type SOCS3 locus, targeted SOCS3 locus in the SOCS3flox allele, and SOCS3 locus in liver fromSOCS3 flox/flox(fl/fl) mice. The positions of PCR primers A, B, and C are shown. (b) PCR analysis of genomic DNA from tails with primers A and B indicated in (a). The PCR product obtained with primers A and B from the wild-type SOCS3 locus is 280 bp (WT SOCS3), and that fromthe SOCS3 flox locus is 380 bp (SOCS3fl/fl). (c) SOCS3fl/fl mice were i.v. injected with 1 Â 109 pfu of AdCre or AdLacZ. PCR analysis of genomic DNA from various tissues 1 or 7 weeks after injecting AdCre was performed with primers A and C. The PCR products obtained with primers A and C from the nondeleted SOCS3flox locus are 1600 bp, and a band of 250 bp obtained with primers A and C indicates Cre-mediated deletion of SOCS3. Deletion of the SOCS3flox locus was only seen in the liver. (d) PCR analysis of the genomic DNA of liver parenchymal cells (P) and nonparenchymal cells (N) isolated from AdCre- or AdLacZ-infected mice.

Oncogene SOCS3 regulates hepatic fibrosis and TGF-b1 H Ogata et al 2523 DMN administration were higher than those in wild- examined STAT3 phosphorylation in ConA-treated type or control mice, suggesting that severe liver damage livers. As shown in Figure 4b, higher levels of occurred by SOCS3 haploinsufficiency or complete phosphorylated STAT3 were accumulated in the nucleus deletion (Figure 3a and b). DMN treatment resulted of both parenchymal and nonparenchymal cells in the in tissue remodeling with matrix deposition mainly in livers of AdCre-SOCS3fl/fl mice than in those of control the centrilobular area and portal tract, as assessed mice. The expression levels of TGF-b1 were also histologically. a-Smooth muscle actin (SMA) localized significantly higher in AdCre-SOCS3fl/fl mice than in mainly around the centrilobular area and the portal control mice (Figure 4e). SOCS3 protein and mRNA tract in the liver. Hepatic collagen deposition was levels were severely reduced in AdCre-SOCS3fl/fl mice stained with Sirius red and quantitated using digital (Figure 4b and e). These data indicate that reduced image analysis (Figure 3c and d). At 7 weeks after DMN expression of SOCS3 exacerbated not only DMN- treatment, SMA and Sirius red stainings were much induced but also ConA-induced liver fibrosis and led stronger in SOCS3 þ /À or AdCre-SOCS3fl/fl mice than in to hyperactivation of STAT3 and enhanced TGF-b1 control mice (Figure 3c). The histological fibrosis score expression. (Sirius red staining) as well as the amount of total collagen and serumhyaluronic acid levels were greater in SOCS3 þ /À mice and AdCre-SOCS3fl/fl mice than in Blockade of STAT3 suppressed DMN-induced liver injury wild-type and control mice (Figure 3c and d). The and fibrosis expression levels of TGF-b1 were also significantly Next, we examined the effect of the suppression of higher in SOCS3 þ /À mice and AdCre-SOCS3fl/fl mice STAT3 in TGF-b1 expression and liver fibrosis. SOCS3 than in control mice (Figure 3e). Hepatic SOCS3 and dominant-negative STAT3 (dnSTAT3) were over- mRNA levels in SOCS3 þ /À mice were about half of expressed in the liver using adenoviruses carrying those in SOCS3 þ / þ mice, and SOCS3 mRNA levels in SOCS3 (AdSOCS3) and dnSTAT3 (AddnSTAT3). An AdCre-SOCS3fl/fl mice were greatly reduced, while injection of AdSOCS3 and AddnSTAT3, but not SOCS1 mRNA levels were not significantly different AdLacZ, significantly increased the survival of mice among these mice (Figure 3e). These data indicate that treated with DMN (Figure 5a) when a dose of DMN a reduced expression of SOCS3 exacerbated DMN- higher than that used in the experiments shown in induced liver injury and fibrosis. Figure 3 was injected. SerumAST and ALT levels were Then, we examined STAT3 phosphorylation in also lower in mice injected with AdSOCS3 and DMN-treated livers. Phosphorylated STAT3 was accu- AddnSTAT3 than in mice injected with AdLacZ mulated in the nucleus of both parenchymal and (Po0.05) (Figure 5b). Furthermore, a histological nonparenchymal cells in DMN-treated livers analysis with Sirius red, phosphorylated STAT3 (pY- (Figure 3f). STAT3 phosphorylation was greater in STAT3), TGF-b1, and SMA staining, as well as with the SOCS3 þ /À and AdCre-SOCS3fl/fl mice than in control amount of total collagen, clearly demonstrated that mice, as confirmed by both immunohistochemistry and SOCS3 and dnSTAT3 overexpression strongly sup- immunoblotting (Figure 3f and g). On the other hand, pressed SMA expression (Figure 5c), ameliorated the STAT1 and ERK phosphorylation levels were not hepatic fibrosis (Figure 5d), and reduced collagen significantly different (Figure 3g). These data raise the deposition (Figure 5c and e). The expression levels of possibility that reduced expression of the SOCS3 gene TGF-b1, but not of SOCS1, were greatly suppressed by exacerbated DMN-induced liver fibrosis through hyper- the SOCS3 and dnSTAT3 adenoviruses, but not by the activation of STAT3, which induces the upregulation of LacZ control virus (Figure 5c and f). These data suggest fibrosis- and remodeling-related genes such as TGF-b1. that the blockade of STAT3 signaling suppressed DMN-induced chronic liver fibrosis and cirrhosis. SOCS3 deficiency in the liver promotes ConA-induced liver fibrosis STAT3 upregulates TGF-b1 expression in liver cells ConA-induced hepatitis also develops fibrosis after Our in vivo data suggest that TGF-b1 is upregulated by repeated liver injury, the mechanism of which is mainly activated STAT3. However, the indirect effect on TGF- based on the upregulation of TGF-b1 expression in the b1 levels by liver damage, which is modulated by liver. Since ConA activates NK and NKT cells in the STAT3, cannot be excluded. To define the relationship liver, ConA-induced fibrosis in mice has been used as a between STAT3 and TGF-b1, we examined TGF-b1 suitable model of viral hepatitis and fibrosis like human mRNA levels in cultured hepatic cells in response to HCV infection. At 4 weeks after ConA treatment, the IL-6. Since primary hepatocytes can be isolated in large serumALT levels were slightly elevated in AdCre- quantities and have the ability to produce a significant SOCS3fl/fl mice but not in control mice; however, the amount of TGF-b1 (Bissell et al., 1995), we used difference was not statistically significant. Therefore, primary hepatocytes first. IL-6 could be involved in ConA induced similar liver injury in both types of mice the induction of TGF-b1, since the expression of TGF- (Figure 4a). However, fibrosis levels assessed by SMA b1 was significantly reduced in IL-6-deficient mice and Sirius red staining were stronger in AdCre-SOCS3fl/fl (Natsume et al., 1999). Real-time PCR analysis indi- mice than in control mice (Figure 4b and c). The amount cated that the TGF-b1 mRNA levels increased in of total collagen was also higher in AdCre-SOCS3fl/fl response to IL-6. TGF-b1 induction in SOCS3-deficient mice than in control mice (Figure 4d). We then hepatocytes was 2–3 times higher than that in WT

Oncogene SOCS3 regulates hepatic ﬁbrosis and TGF-b1 H Ogata et al 2524 hepatocytes (Figure 6a). On the other hand, the over- IL-6-mediated TGF-b1 induction was not inhibited by expression of a dominant-negative STAT3 (dnSTAT3) cycloheximide treatment (Figure 6c), suggesting that suppressed IL-6-induced TGF-b1 expression (Figure 6b). activated STAT3 directly upregulates the transcription

Oncogene SOCS3 regulates hepatic ﬁbrosis and TGF-b1 H Ogata et al 2525

Figure 4 SOCS3 inhibits ConA-induced liver ﬁbrosis. (a) Serum ALT values were examined after the last ConA administration. (b) Sirius red staining ( Â 100) and immunostaining of pY-STAT3, TGF-b1, SMA, and SOCS3 of liver sections after 4 weeks of ConA treatment ( Â 200). (c) Quantiﬁcation of Sirius red staining of (b) using digital image analysis. (d) Total collagen content of the liver of mice. (e) Relative expression levels of the TGF-b1 and SOCS3 mRNAs analysed by real-time PCR. *Po0.05.

Figure 3 SOCS3 inhibits liver fibrosis. (a) Survival tests. The SOCS3 þ / þ , SOCS3 þ /À, and AdCre or AdlacZ adenovirus-injected SOCS3fl/fl mice were treated with DMN for 3 consecutive days per week. (b) The serum AST and ALT levels in mice were examined at the end point of this study (SOCS3 þ /À mice vs SOCS3 þ / þ mice, n ¼ 9; AdCre-SOCS3fl/fl mice vs AdLacZ-SOCS3fl/fl mice, n ¼ 7). (c) a- Smooth muscle actin (SMA) immunostaining ( Â 200) and Sirius red staining ( Â 100) of liver sections frommice7 weeks after DMN treatment. (d) Fibrosis scores determined by the METAVIR scoring system, Sirius red staining, total collagen content of the liver, and serum hyaluronic acid levels in the blood of mice at 7 weeks after DMN treatment. Sirius red staining was quantified with an NIH image. (e) Relative expression levels of the TGF-b1, SOCS3, and SOCS1 mRNAs analysed by real-time PCR. (f) Immunohistochemical detection of phosphorylated STAT3 in the liver sections of mice treated with DMN. (g) Western blot analysis of phosphorylated-STAT3 (pY-STAT3), STAT1 (pY-STAT1), and ERK1/2 (p-ERK) in whole liver lysates frommicetreated with DMN. The phosphorylation levels were quantified by densitometry (n ¼ 7). *Po0.05, **Po0.01.

Oncogene SOCS3 regulates hepatic ﬁbrosis and TGF-b1 H Ogata et al 2526

Figure 5 Overexpression of SOCS3 and dnSTAT3 inhibits liver damage and ﬁbrosis. (a) At 1 week after injection with adenoviral vectors, DMN was i.p. injected into mice for the survival test. (b–f) Mice were i.p. injected with DMN for 3 consecutive days per week and i.v. injected with the adenoviral vectors 4 weeks after the initial DMN administration. After an additional 3 weeks of DMN administration, the mice were killed and examined (n ¼ 7). (b) The serumAST and ALT levels were determinedat the end points. ( c) Sirius red staining ( Â 100) and immunostainings for pY-STAT3, TGF-b1, and SMA in liver specimens from these mice ( Â 200). (d) The surface area stained with Sirius red was quantitated using digital image analysis. (e) Total collagen content of the liver of mice treated with DMN for 7 weeks. (f) The relative expression levels of the TGF-b1 and SOCS1 mRNAs were analysed by real-time RT–PCR. The data are shown as the mean7s.e.m. *Po0.05, **Po0.01.

of TGF-b1 mRNA. Similarly, IL-6 upregulated TGF-b1 Discussion mRNA levels in the human hepatoma cell line, HepG2 (Figure 6d), and the human Ito cell line, LI90, and this HCC is believed to develop fromliver cirrhosis and upregulation was inhibited by dnSTAT3 or SOCS3 advanced ﬁbrosis. In this study, we have shown that overexpression (Figure 6e). Fromthese data, we propose SOCS3 and STAT3 reciprocally regulate liver damages, that TGF-b1 is a direct target of STAT3 and that ﬁbrosis, and TGF-b1 expression induced by chemicals SOCS3 negatively regulates TGF-b1 production by and ConA. Our data suggest that STAT3 directly suppressing STAT3. upregulates TGF-b1 expression at the transcriptional

Oncogene SOCS3 regulates hepatic fibrosis and TGF-b1 H Ogata et al 2527 levels. STAT3 is a well-characterized oncogene; it effect of TGF-b through the transcriptional induction of regulates gene expressions related to antiapoptosis, inhibitory Smad7 (Jenkins et al., 2005). Therefore, proliferation, neovascularization, and inflammation STAT3 positively regulates both TGF-b and Smad7 (Bromberg and Darnell, 2000; Wei et al., 2003; Wang et al., 2004). Our study adds a new target, TGF-b1,to the list of STAT3-regulated genes involved in carcinogenesis. We have identified the two potential STAT3- binding sites in the promoter region of TGF-b1 and demonstrated that STAT3 activates the promoter activity in vitro (Kinjyo et al., unpublished data). Therefore, TGF-b1 transcription seems to be directly regulated by STAT3. Furthermore, our study provides evidence for a unique feature of SOCS3 as an antioncogene: SOCS3 suppresses STAT3 activation and negatively regulates fibrosis in a precancerous stage in the liver. Our finding of a direct induction of TGF-b1 by active STAT3 may provide another mechanism of the con- tribution of STAT3 to carcinogenesis in addition to fibrosis (Factor et al., 1997; Bowman et al., 2000; Bissell, 2001). TGF-b1 produced from tumor cells may induce T-cell inactivation, which causes an escape of tumor cells fromimmunesurveillance (Gorelik and Flavell, 2001). In this context, STAT3 is shown to suppress the expression of inflammatory cytokines and chemokines from tumor cells that attract lymphocytes and leuko- cytes to tumor cells (Wang et al., 2004). Therefore, STAT3 activation may contribute to carcinogenesis by suppressing antitumor immunity through the upregulation of TGF-b1 and downregulation of inflammatory cytokines and chemokines. Interestingly, STAT3 also regulates intracellular TGF-b1 signaling. Tumor cells are frequently resistant to the growth-inhibitory effect of TGF-b. One mechan- ismis the reduced expression of the TGF- b type II receptor in tumor cells (Markowitz et al., 1995; Saeki et al., 2000). Indeed, the expression of the TGF-b type II receptor in the HCC region was greatly reduced compared with that in non-HCC region (Ogata et al., unpublished data). Another mechanism is the induction of Smad7, a negative-feedback regulator of TGF-b/ Smad signaling. Jenkins et al. (2005) recently reported that hyperactivation of STAT3 in mice carrying a mutant gp130, which lacked the SOCS3-binding site, promoted gastric adenomatous hyperplasia (Tebbutt et al., 2002). Recently, they found that hyperactivation of STAT3 desensitized gastric cells to the cytostatic

Figure 6 SOCS3 inhibits STAT3-mediated TGF-b1 expression. (a) Relative expression levels of the TGF-b1 and SOCS3 mRNAs analysed by real–time RT-PCR. Primary hepatocytes derived from AlbCre-SOCS3fl/fl and control SOCS3fl/fl mice were stimulated with IL-6 for the indicated periods. (b) Relative expression levels of the TGF-b1 mRNA in primary hepatocytes derived from wild-type mice i.v. injected with AddnSTAT3 or AdLacZ. (c) TGF-b1 mRNA levels in the primary hepatocytes from AlbCre-SOCS3fl/fl and control SOCS3fl/fl mice stimulated with IL-6 in the presence of cycloheximide (Chx) (10 mg/ml). (d) TGF-b1 mRNA levels in HepG2 cells at 12 h after stimulation with IFN-g (100 ng/ml) or IL-6 (100 ng/ml). (e) TGF-b1 mRNA levels in LI-90 cells infected with or without adenoviral vectors at MOI ¼ 100 or in the presence of cycloheximide. *Po0.05.

Oncogene SOCS3 regulates hepatic fibrosis and TGF-b1 H Ogata et al 2528 expression simultaneously. As a result, tumor cells with including lymphocytes, Kupffer cells, and stellate cells, constitutively activated STAT3 may be resistant to may be important for enhanced liver damage in SOCS3- TGF-b, even though these cells secrete a large amount deficient mice. Further study is necessary to define the of TGF-b. precise role of STAT3 in a different type of cells in Since the reduction of SOCS3 expression correlates the process of liver damage, fibrosis, and HCC deve- with the progression of fibrosis in the non-HCC region lopment. in humans, SOCS3 may play a suppressive role in The suppression of STAT3 activation in the early fibrosis and carcinogenesis in the liver. In many cases, stage of liver injury could be useful for preventing liver both SOCS1 and SOCS3 were simultaneously down- damage and fibrosis. We demonstrated that the adeno- regulated in HCC (Niwa et al., 2005). Like the SOCS1 viral gene transfer of SOCS3 or dominant-negative gene, DNA methylation of the SOCS3 gene is found in STAT3 efficiently reduced hepatitis and fibrosis (Figure 5). several cancers, including HCC (He et al., 2003; Niwa Recently, the administration of cell-penetrating forms of et al., 2005; Weber et al., 2005). However, the frequency SOCS3 (CP-SOCS3) into mice has been shown to of DNA methylation of the SOCS3 gene was not as high counteract enterotoxin B and lipopolysaccharide and as that of SOCS1 in HCC (33% for SOCS3 and 65% for ConA-induced liver damage and inflammation (Jo et al., SOCS1) (Niwa et al., 2005; Yoshimura et al., 2005). We 2005). Although these researchers did not examine found that the SOCS3 mRNA levels were lower in the fibrosis, since they used only acute inflammation HCC region than in non-HCC region in more than 90% models, a similar approach using cell-penetrating of HCC patients (Ogata et al., unpublished data). SOCS3 could be useful for preventing liver fibrosis. Therefore, another mechanism, in addition to DNA Similarly, suppression of STAT3 activity could be useful methylation, must be present for the reduction of for preventing not only viral hepatitis-induced liver SOCS3 gene expression in human HCC. fibrosis but also HCC development. The induction of It has not been identified what kind of cytokines are SOCS3 in the liver could also be effective, since SOCS3 involved in STAT3 activation in ConA- or DMN- expression can be induced in many ways, such as the treated mice or in human HCC. It was reported that elevation of cAMP (Boer et al., 2002) and the leptin, as well as IL-6, stimulates the activation of administration of statins (Huang et al., 2003). In STAT3 and increases collagen mRNA expression in conclusion, this study provides the basis for the hepatic stellate cells and subsequently accelerates liver therapeutic potential of liver fibrosis through the fibrogenesis (Saxena et al., 2002). Therefore, these STAT3 antagonists. STAT3-activating cytokines must be promoting factors of liver fibrosis and carcinogenesis. While the growth hormone (GH) has also been reported to increase Materials and methods SOCS3 mRNA expression and to reduce the severity of fibrosis associated with chronic intestinal inflamma- Patients, samples, and fibrosis score tion (Theiss et al., 2005). GH-mediated suppression of Liver samples, primary HCC samples, and adjacent noncan- fibrosis may be due to the inhibition of STAT3/TGF-b1 cerous tissues were obtained during partial hepatic resection production by upregulated SOCS3. surgery on patients at Kyushu University. All of themhad Then, which type of cells play a major role in SOCS3/ chronic hepatitis C, and all samples were collected with the STAT3-mediated TGF-b1 production and fibrosis? In informed consent of patients. The fibrosis scores (stages) of the livers of patients with chronic hepatitis C were evaluated by our mice models, adenoviral vectors infected both using the METAVIR scoring system(Bedossa and Poynard, parenchymal and nonparenchymal cells (Figure 2). It 1996). has been shown that Ito cells are a major source of TGF-b1; however, a recent study suggests that par- Generation of conditional SOCS3-deficient mice enchymal cells also produce significant amounts of SOCS3flox/flox (SOCS3fl/fl) mice were as described (Yasukawa TGF-b1 (Bissell et al., 1995). We confirmed that IL-6- et al., 2003). To delete SOCS3 in the liver, SOCS3fl/fl mice were mediated upregulation of TGF-b1 mRNA levels was i.v. injected with 1 Â 109 pfu of AdCre. In some experiments, suppressed by overexpression of dnSTAT3 in both Albumin-Cre (AlbCre) mice (Jakson Lab) expressing Cre primary hepatocytes and an Ito cell line (Figure 6). recombinase under the control of the mouse albumin gene Therefore, hyperactivation of STAT3 in both parench- regulatory region (Postic and Magnuson, 2000) were crossed ymal and nonparenchymal cells in SOCS3-deficient mice to SOCS3fl/fl mice and used to isolate SOCS3-deficient may contribute to enhanced production of TGF-b1. hepatocytes. Genotyping was carried out with primers as DMN-induced liver damage was also enhanced in liver- described (Yasukawa et al., 2003). specific SOCS3-deficient mice, but was suppressed in mice overexpressing SOCS3 or dnSTAT3 in the liver. Induction of fibrosis in mice We have not defined which type of cells would be For the liver fibrosis model, 10-week-old mice were adminis- important for liver damages in these models. Our tered DMN and ConA. For the fibrosis model, mice were i.p. fl/fl injected with DMN (3.0 mg/g body weight) for 3 consecutive preliminary data suggest that AlbCre-SOCS3 mice, days each week for 7 weeks. The administration of DMN was in which the SOCS3 gene was deleted only in adjusted weekly for body weight. After 2 days fromthe seventh parenchymal cells, were not sensitive and were rather series of injections, venous blood was collected, and the mice resistant to DMN-induced liver injury (Ogata et al., were killed. For ConA-induced fibrosis models, mice were i.v. unpublished data). Therefore, nonparenchymal cells, injected with ConA (10 mg/g body weight) weekly for 4 weeks.

Oncogene SOCS3 regulates hepatic fibrosis and TGF-b1 H Ogata et al 2529 At 1 day after the last injection, venous blood was collected, 1:100 dilution of anti-phospho-STAT3-specific antibodies and the mice were killed. All experiments on these mice were (anti-pY-STAT3; Cell Signaling) or a 1:100 dilution of approved by and performed in accordance with the Guidelines polyclonal anti-SOCS3 (Immuno-Biological Laboratories), of the Animal Ethics Committee of Kyushu University, anti-TGF-b1 (Promega), or anti-a-smooth muscle actin Fukuoka, Japan. (SMA) (Dako) and stained with the LSAB kit according to the manufacturer’s instructions (Dako). The samples were then lightly stained with hematoxylin and examined. Liver fibrosis Recombinant adenovirus was quantified with Sirius red (Polyscience Inc.) staining as Adenoviral vectors containing the genes for LacZ described (Yoshida et al., 2004). The sections were incubated (AdLacZ), Myc-tagged SOCS3 (AdSOCS3) and HA-tagged for 10 min in an aqueous solution of saturated picric Y705F dnSTAT3 (AddnSTAT3), and Cre (AdCre) were acid containing 0.1% Sirius red. Red-stained collagen fibers prepared by homologous recombination in HEK293 were quantitated by digital image analysis (Yoshida et al., cells as described previously (Shouda et al., 2001). For the in 2004). vivo gene transfer for the overexpression study, 200 mlof 5 Â 109 pfu/ml recombinant viruses in PBS was i.v. injected into C57BL/6 mice. Mice were i.p. injected with RNA extraction and real-time RT–PCR DMN (3.0 mg/g body weight) for 3 consecutive days per Total RNAs from mice livers and primary hepatocytes were week and i.v. injected with the adenoviral vectors at 4 weeks prepared using the TRIZOL reagent (Invitrogen). RT–PCR after the initial DMN administration. After an additional 3 was carried out using the one-step RT–PCR kit (Applied weeks of DMN administration, the mice were killed and Biosystems) according to the manufacturer’s instructions. examined. Quantitative real-time RT–PCR was monitored using the ABI PRISM 7700 (PE Applied Biosystems), and the results were analysed with the accompanying software, as described Isolation and culture of primary hepatocytes (Kimura et al., 2004). The SYBR Green PCR Master Mix was Primary hepatocytes were prepared using the two-step used for the detection of mouse SOCS1, SOCS3, and human TGF-b1. The primer pairs are described below; the primer sets collagenase perfusion method as described (Bissell et al., 0 0 0 1995). After isolation, the cells were washed twice in Williams’ were 5 -GAAGCCATCTTCACGCTG-3 and 5 -ACACTCA CTTCCGCACCTTC-30 (mouse SOCS1); 50-GGGTGGCA E medium with 10% fetal bovine serum, L-glutamine, 100 nM AAGAAAAGGAG-30 and 50-GTTGAGCGTCAAGACC dexamethasone, and 1 mM insulin. The cells were centrifuged CAGT-30 (mouse SOCS3); and 50-CAACAATTCCTGGC at 50 g for 1 min between washes. Cell viability, as estimated by 0 0 0 trypan blue exclusion, was routinely more than 97%. The cells GATACC-3 and 5 -GAACCCGTTGATGTCCACTT-3 were plated on collagen type I-coated plates at 3 Â 106 cells/ (human TGF-b1). The Taq Man PCR Master Mix was used t 100-mm dish and incubated for 10 h. For each experiment, for the detection of mouse TGF-b1 (Assays-on-Demand cells were stimulated with IL-6 (100 ng/ml, Calbiochem). Gene Expression Products, Applied Biosystems) and mouse Parenchymal cells (hepatocytes) and nonparenchymal cells glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (stellate cells, Kupffer cells, and sinusoidal endothelial cells) (Rodent GAPDH Control Reagents, Applied Biosystems), human-SOCS3, and b-actin. The data are shown as the were isolated separately fromliver, as described previously 7 (Bissell et al., 1995). mean s.e.m.

Statistical analysis Western blot analysis The results are expressed as the mean7s.e.m. Each piece of Liver tissues obtained frommicetreated with DMN and data was analysed using an unpaired two-tailed Student’s t-test patients with chronic hepatitis were immediately frozen in and analysis of variance (ANOVA). liquid nitrogen and stored at À801C. The whole liver of mice and hepatic resection specimens of human patients were prepared as described (Yoshida et al., 2002). The total cell extracts were resolved by SDS–PAGE, and membranes Abbreviations were immunoblotted with anti-phospho-STAT3 (Tyr705) (pY-STAT3), anti-phospho-STAT1 (Tyr701) (pY-STAT1), JAK, Janus kinase; STAT, signal transducers and activators of anti-phospho-ERK1/2 (Thr202/Tyr204) (p-ERK1/2) (Cell transcription; SOCS, suppressor of cytokine signaling; TGF, Signaling), anti-STAT3, anti-STAT1, anti-ERK2 (Santa transforming growth factor; AST, aspartate aminotransferase; Cruz), and polyclonal anti-SOCS3 (Immuno-Biological ALT, alanine aminotransferase. Laboratories). A macrophage cell line (RAW) stimulated with LPS for 3 h in vitro was used as a positive control. For Acknowledgements quantitative analysis, the density of the blots for phosphorylated STAT3 and total STAT3 was measured and expressed We thank Y Kawabata and T Yoshioka for their help, Mr T as the ratio of phosphorylated STAT3 to total STAT3. Kinoshita, Ms M Othsu, Ms Y Yamada, Mr M Sasaki, and In addition, the density of the blots for phosphorylated Ms E Fujimoto (Technical Support Center, Medical Institute STAT1, phosphorylated ERK1/2, total STAT1, and total of Bioregulation) for technical assistance, and Y Nishi for ERK2 was measured and expressed as the ratios of phos- manuscript preparation. This work was supported by special phorylated STAT1 and phosphorylated ERK1/2 to total grants-in-aid fromthe Ministry of Education, Science, STAT1 and total ERK2, respectively. Technology, Sports, and Culture of Japan, the Haraguchi Memorial Foundation, the Yamanouchi Foundation for Research on Metabolic Disorders, the Takeda Science Immunohistochemistry and Sirius red staining Foundation, the Mochida Memorial Foundation, the Kato Liver tissues were ﬁxed with 10% formalin, parafﬁn-em- Memorial Foundation, and the Uehara Memorial Founda- bedded, and sectioned. Slides were incubated with either a tion.

Oncogene SOCS3 regulates hepatic ﬁbrosis and TGF-b1 H Ogata et al 2530 References

Bedossa P, Poynard T. (1996). Hepatology 24: 289–293. Markowitz S, Wang J, Myeroff L, Parsons R, Sun L, Bissell DM. (2001). Exp Mol Med 33: 179–190. Lutterbaugh J et al. (1995). Science 268: 1336–1338. Bissell DM, Wang SS, Jarnagin WR, Roll FJ. (1995). J Clin Mori H, Hanada R, Hanada T, Aki D, Mashima R, Invest 96: 447–455. Nishinakamura H et al. (2004). Nat Med 10: 739–743. Bode JG, Nimmesgern A, Schmitz J, Schaper F, Schmitt M, Nakamura T, Sakata R, Ueno T, Sata M, Ueno H. (2000). Frisch W et al. (1999). FEBS Lett 463: 365–370. Hepatology 32: 247–255. Boer AK, Drayer AL, Rui H, Vellenga E. (2002). Blood 100: Natsume M, Tsuji H, Harada A, Akiyama M, Yano T, 467–473. Ishikura H et al. (1999). J Leukoc Biol 66: 601–608. Bowman T, Garcia R, Turkson J, Jove R. (2000). Oncogene 19: Niwa Y, Kanda H, Shikauchi Y, Saiura A, Matsubara K, 2474–2488. Kitagawa T et al. (2005). Oncogene 24: 6406–6417. Bromberg J, Darnell Jr JE. (2000). Oncogene 19: 2468–2473. Postic C, Magnuson MA. (2000). Genesis 26: 149–150. Bromberg JF, Wrzeszczynska MH, Devgan G, Zhao Y, Pestell Roberts AW, Robb L, Rakar S, Hartley L, Cluse L, RG, Albanese C et al. (1999). Cell 98: 295–303. Nicola NA et al. (2001). Proc Natl Acad Sci USA 98: Choi I, Kang HS, Yang Y, Pyun KH. (1994). Clin Exp 9324–9329. Immunol 95: 530–535. Saeki A, Tamura S, Ito N, Kiso S, Matsuda Y, Yabuuchi I Croker BA, Krebs DL, Zhang JG, Wormald S, Willson TA, et al. (2000). Cancer 88: 1025–1029. Stanley EG et al. (2003). Nat Immunol 4: 540–545. Saxena NK, Ikeda K, Rockey DC, Friedman SL, Anania FA. Croker BA, Metcalf D, Robb L, Wei W, Mifsud S, DiRago L (2002). Hepatology 35: 762–771. et al. (2004). Immunity 20: 153–165. Shouda T, Yoshida T, Hanada T, Wakioka T, Oishi M, Factor VM, Kao CY, Santoni-Rugiu E, Woitach JT, Jensen Miyoshi K et al. (2001). J Clin Invest 108: 1781–1788. MR, Thorgeirsson SS. (1997). Cancer Res 57: 2089–2095. Tebbutt NC, Giraud AS, Inglese M, Jenkins B, Waring P, Clay Gorelik L, Flavell RA. (2001). Nat Med 7: 1118–1122. FJ et al. (2002). Nat Med 8: 1089–1097. He B, You L, Uematsu K, Zang K, Xu Z, Lee AY et al. (2003). Theiss AL, Fuller CR, Simmons JG, Liu B, Sartor RB, Lund Proc Natl Acad Sci USA 100: 14133–14138. PK. (2005). Gastroenterology 129: 204–219. Huang KC, Chen CW, Chen JC, Lin WW. (2003). FEBS Lett Wang T, Niu G, Kortylewski M, Burdelya L, Shain K, Zhang 555: 385–389. S et al. (2004). Nat Med 10: 48–54. Jenkins BJ, Grail D, Nheu T, Najdovska M, Wang B, Waring Weber A, Hengge UR, Bardenheuer W, Tischoff I, Sommerer P et al. (2005). Nat Med 11: 845–852. F, Markwarth A et al. (2005). Oncogene 24: 6699–6708. Jo D, Liu D, Yao S, Collins RD, Hawiger J. (2005). Nat Med Wei D, Le X, Zheng L, Wang L, Frey JA, Gao AC et al. 11: 892–898. (2003). Oncogene 22: 319–329. Kamio M, Yoshida T, Ogata H, Douchi T, Nagata Y, Inoue Yasukawa H, Ohishi M, Mori H, Murakami M, Chinen T, M et al. (2004). Oncogene 23: 3107–3115. Aki D et al. (2003). Nat Immunol 4: 551–556. Kamizono S, Hanada T, Yasukawa H, Minoguchi S, Kato R, Yoshida T, Hanada T, Tokuhisa T, Kosai K, Sata M, Kohara Minoguchi M et al. (2001). J Biol Chem 276: 12530–12538. M et al. (2002). J Exp Med 196: 641–653. Kamura T, Sato S, Haque D, Liu L, Kaelin Jr WG, Conaway Yoshida T, Ogata H, Kamio M, Joo A, Shiraishi H, Tokunaga RC et al. (1998). Genes Dev 12: 3872–3881. Y et al. (2004). J Exp Med 199: 1701–1707. Kimura A, Kinjyo I, Matsumura Y, Mori H, Mashima R, Yoshikawa H, Matsubara K, Qian GS, Jackson P, Groopman Harada M et al. (2004). J Biol Chem 279: 6905–6910. JD, Manning JE et al. (2001). Nat Genet 28: 29–35. Kinjyo I, Hanada T, Inagaki-Ohara K, Mori H, Aki D, Ohishi Yoshimura A, Nishinakamura H, Matsumura Y, Hanada T. M et al. (2002). Immunity 17: 583–591. (2005). Arthritis Res Ther 7: 100–110.

Oncogene