TGF-β Signaling Is Often Attenuated during Hepatotumorigenesis, but Is Retained for the Malignancy of Hepatocellular Carcinoma Cells

The role of transforming growth factor-beta (TGF-β) signaling in hepatocarcinogenesis remains controversial. We aimed to reveal TGF-β signaling status in human and murine tissues of hepatocellular carcinoma (HCC) and the mechanisms that mediate TGF-β’s role in regulating HCC malignancy. Here, TGF-β pathway component expression and activation in human and murine HCC tissues were measured with quantitative RT-PCR and Western blotting assays. The role of TGF-β receptor and Smad signaling in the growth and survival of several HCC cell lines was determined with several in vitro and in vivo approaches. We found that TGF-β receptor II (TβRII) expression was downregulated in two different HCC patient cohorts. Consistently, Smad3 phosphorylation was also downregulated in HCC tissues in comparison to that in adjacent normal tissues. Interestingly, many HCC cell lines were sensitive to TGF-β and growth-inhibited by exogenous TGF-β. However, stable knockdown of TβRII inhibited cell growth on plastic and in soft agar, and induced apoptosis resulting in suppressed subcutaneous tumor growth and metastatic potential in vivo. Furthermore, knockdown of Smad4 also led to a significant inhibition of growth on plastic and in soft agar with concomitant increase of apoptosis, PTEN expression, and reduced nuclear accumulation of linker region-phosphorylated Smad3. Taken together, TGF-β signaling pathway plays a dichotomous role in hepatocellular carcinogenesis. It appears to suppress HCC development, but is retained for HCC cell survival and malignancy. Furthermore, Smad4 can mediate both growth inhibitory activity induced by exogenous TGF-β and the survival activity induced by autocrine TGF-β revealing a delicate selection of the two opposing activities of TGF-β during HCC evolution.


Introduction
Hepatocellular carcinoma (HCC), the major primary liver cancer, is the fifth most common cancer in men and women worldwide [1]. The mechanism underlying tumor initiation and progression of this disease is still not well understood, but partly due to deregulation of microenvironment homeostasis that involves the transforming growth factor b (TGF-b) signaling pathway [2,3].
TGF-b isoforms are polypeptide cytokines. They are secreted in latent forms, which need to be activated to interact with cell surface receptors. Active homodimeric TGF-b isoforms initiate signaling by binding to the type I (RI) and type II (RII) TGF-b receptors, which contain an intracellular kinase domain. The activation of the RI kinase by ligand binding to the RII leads to the activation of Smad2 and Smad3 transcription factors via phosphorylation at their C-termini. The interaction between TGF-b and its receptors has also been shown to activate Smadindependent signaling pathways including PI3K/AKT and MAP kinase pathways. Because of this diverse array of signaling pathways activated by TGF-b, the role of TGF-b signaling in regulating cellular functions is often complex and contextdependent. TGF-b signaling through Smads is a well-known tumor suppressive pathway as it inhibits cellular proliferation by stimulating the expression of cyclin-dependent kinase inhibitors, p15 and p21, and induces apoptosis via various mechanisms [4]. On the other hand, TGF-b signaling has also been shown to drive tumor progression. This has been attributed to the activation of the Smad-independent pathways by TGF-b signaling in some cases [5].
In the normal liver, TGF-b is produced by nonparenchymal cells including sinusoidal endothelial cells, Kupffer cells, and lipocytes, and acts as a paracrine suppressor on the proliferation of normal hepatocytes [6][7][8]. However, TGF-b expression is often upregulated in transformed hepatocytes. In fact, it has been reported that plasma TGF-b was increased in HCC patients, especially during angiogenesis of HCC, and could be regarded as a marker for HCC progression [9,10]. Most hepatocarcinoma cells are able to synthesize and secrete TGF-b continually by themselves. Interestingly, some reports showed low frequency of mutation of TGF-b receptor II (TbRII) and other TGF-b pathway genes in HCC, which are often found to be mutationally inactivated in other gastrointestinal cancers [11][12][13][14][15]. Thus, while TGF-b signaling is tumor-suppressive in various tissues, HCC cells often retain sensitivity to TGF-b and possess a functional autocrine TGF-b loop. However, the role of this autocrine TGF-b loop has not been well defined. A recent study showed that deletion of Tgfbr2 in the setting of p53 loss reduced the formation of liver tumors, suggesting that TGF-b signaling was playing a promoting role in HCC induced by the loss of p53 [16]. However, knockout of TGF-b signaling components in other mouse tissues have in general promoted oncogene-induced tumor progression [2].
Because of the controversy surrounding the role of TGF-b signaling in hepatocytes and HCC cells, we have carried out comprehensive analyses of TGF-b pathway component expression and activation in human and murine HCC tissues and human HCC cell lines. The results shown below indicate a dichotomous role of TGF-b/Smad pathway during hepatocarcinogenesis. While the attenuation of TGF-b receptor signaling through Smad appears needed for the development of HCC, the attenuation appears limited and may even be reversed during the tumor progression for the survival of HCC cells. Our study further demonstrates that while HCC cells are growth-inhibited by exogenous TGF-b, they require autocrine TGF-b signaling for survival and malignancy, both of which are dependent on Smad4. As such, our study suggests a delicate balance of the two opposing activities of TGF-b during HCC evolution.

Human and Mouse Tissue Samples
Human HCC and corresponding adjacent tissues were obtained from patients undergoing surgical resection or liver transplantation at the Organ Transplant Center of the University of Texas Health Science Center at San Antonio and at the First Affiliated Hospital of Nanjing Medical University (Table S1). All the patients gave written informed consent and the study was also approved by the the Institutional Review Boards at the University of Texas Health Science Center at San Antonio and the First Affiliated Hospital of Nanjing Medical University. Mouse normal liver, adjacent to HCC, and HCC tissues were collected from C3HeB/FeJ mice, which spontaneously develop HCC as described previously [17]. All animal experiments were conducted following appropriate guidelines. They were approved by the Institutional Animal Care and Use Committee and monitored by the Department of Laboratory Animal Resources at the University of Texas Health Science Center at San Antonio.

RNA Extraction, RT-PCR and Quantitative Real-time PCR
Total RNA was isolated from human tissues or HCC cell lines using Tri Reagent (Sigma-Aldrich, MO, USA) according to the manufacturer's instructions. The extracted RNA was dissolved in DEPC-treated ddH 2 O and subjected to DNAse I treatment (Fisher Scientific, IL, USA) to remove genomic DNA contamination. DNAse I-treated total RNA (2 mg) was reverse-transcribed into cDNA using ABI high-capacity cDNA Reverse Transcription Kit. Quantitative real-time PCR was performed using Power SYBR Green PCR Mix (ABI) in Applied Biosystems. All primers used in this study (Table S2) were designed by Primer Premier 5.0 and synthesized by Integrated DNA Technologies (Coralville, IA).

Luciferase Reporter Assay
Cells were plated at 1.5610 5 cell per well of a 12-well plate in triplicate 24 hours before transfection. pSBE4-Luc with repeated Smad binding elements and b-galactosidase expression plasmids were transiently co-transduced into cells by using LT-1 (Mirus Bio, Madison, WI) [20]. After 3 hours, the transfected cells were treated with/without 2 ng/ml TGF-b1 and/or 100 mM RI-KI. After additional 24 hours of incubation, cells were harvested and lysed as described previously [21]. Luciferase activity was assayed and normalized to b-galactosidase activity.

MTT Assay
To determine the growth of HCC cell lines, cells were seeded in 96-well plates at 2,000 cells/well in the presence or absence of different concentrations of TGF-b1. 50 ul 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (2 mg/ml in sterile PBS ) was added into each well at indicated time point and cells were incubated at 37uC for 2 hours assays. 100 ul DMSO was added into each well after the medium was removed, and the plate was gently shaken on a shaker for 10 minutes. The absorbance was measured at 595 nm with a Microplate Reader (Bio Tek Instrument, Winooski, VT).

Elisa Assay for TGF-b
Cells were plated at the same number in 24 well plates and fresh basic medium was changed when cells were exponentially growing at 70-80% confluence. After 24-hour incubation, the supernatants were collected, and cells were counted under microscope. The reading was normalized to cell number. This assay was performed by using the Duoset ELISA Development kit from R&D Systems (Minneapolis, MN) according to the manufacture's protocol.

Transfection and Determination of Knockdown of TGFBRII and Smad4
The TbRII shRNA, Smad4 shRNA and control shRNA in lentiviral vector pLK0

Soft and Hard Agarose Colony Formation Assay
Cells suspended in 0.5 ml of 0.4% or 1.2% low melting point agarose (Life Technologies, Carlsbad, CA, USA) with complete culture medium were gently seeded in 12-well plates coated with hour. (C) Levels of the TGFb isoforms in the media conditioned by four HCC cell lines were examined with the Duoset ELISA kit. Each data point represents mean 6SEM from three independent wells. (D) SBE-luciferase assay for TGF-b sensitivity in HCC cells. Cells were co-transfected with a TGFb-responsive promoter-luciferase reporter (pSBE4-Luc) and a b-galactosidase (b-gal) expression plasmid and then treated with or without 2 ng/ml TGF-b1 or 100 nM RI-KI for 24 hours. Each data point represents mean6SEM from three independent transfections. Luciferase units in cell lysates were normalized to b-gal activity. (E) MTT assay for the effect of indicated doses of TGFb1 on the proliferation of HCC cells after a 5-day treatment. Each data point represents mean6SEM from 4 independent wells. (F) soft-agar colony formation ability was assessed in HCC cells with or without 2 ng/ml TGFb1 treatment for 14 days. Colonies in each culture well were counted after staining. Each data point represents mean6SEM from three independent wells. **, p,0.01; ***, p,0.001; ''ns'' means ''no significance''. doi:10.1371/journal.pone.0063436.g002 0.8% agarose and 2 ng/ml TGF-b1 diluted in 250 ml medium was added on top when agarose was solidified as described previously [23]. Cell numbers were varied for different cell lines (SNU398, 4500 cells/well; SNU423, 6000 cells/well; HepG2, 4500 cells/ well; Sk-Hep-1, 4500 cells/well; Huh7, 4500 cells/well). The plates were incubated at 37uC in tissue culture incubator for the indicated days. Colonies were stained with p-iodonitrotetrazolium violet (Sigma-Aldrich, St. Louis, MO) and counted by eye.

Cell Death Detection ELISA
Cells were plated at 5610 5 cells/dish in 60 mm dishes and harvested after various treatments. Cell pellets were washed with cold 16PBS twice, lysed with 40 ml Apoptosis Lysis Buffer and cell death was assayed using Cell Death Detection ELISA PLUS (Roche Applied Science, Indianapolis, IN), which is based on the measurement of histone-complexed DNA fragments in the cytoplasm of apoptotic cells, according to the manufacture's protocol.

Annexin-V FITC Staining
Cells were plated at 5610 5 cells/dishes into 60 mm dishes. After reaching 70-80% confluence during exponential growth, cells were then starved for 48 hours when cells were exponentially growing at 70-80% confluence. Cells were harvested, washed with cold PBS and resuspended with binding buffer at a concentration of 2610 6 cell/ml. Cells were analyzed by using the ApoTarget TM Annexin-V FITC Apoptosis kit (Invitrogen, Grand Island, NY) according to the manufacture's protocol.

Animal Experiments
Male athymic nude mice (Harlan Sprague Dawley, Inc., Indianapolis, IN), at 4-5 weeks of age, were used for in vivo animal experiments. The animals were maintained under specific pathogen-free conditions at the University of Texas Health Science Center, San Antonio, Texas. All animal protocols were approved and monitored by the Institutional Animal Care and Use Committee.

In vivo Tumorigenicity and Metastasis Studies
To determine tumor growth in vivo, Sk-Hep-1/TbRIIshRNA/ Luc-GFP and control cells were harvested from subconfluent exponentially growing cultures and inoculated subcutaneously in the rear back hindquarters of 4-week-old male nude mice. Cells was inoculated injected on both sides of each mouse at 3610 6 cells per 0.1 ml sterile PBS. Growth of xenografts was determined by measuring the volume (V), which was calculated with V = (L6W 2 )60.5, where L is the length and W is the width of the xenograft measured with a caliper. Mice were put through bioluminescence imaging to identify metastasis every two weeks as described previously [24]. After the termination of experiment at about 8 weeks, tumors were resected from anesthetized mice. Tumors were flash-frozen in liquid nitrogen for RNA extraction.
To determine the metastatic potential of the cells, a tail vein injection assay was performed. Sk-Hep-1/TbRII shRNA/Luc-GFP and control cells were harvested as above and resuspended at 0.5610 6 cells/0.1 ml sterile PBS. Cells were inoculated into mice by intravenous injection through tail vein. Bioluminenscence imaging was performed to monitor metastasis burden.

Statistical Analysis
Differences between the means of control and experimental groups were analyzed by two-tailed Student's t-tests. One-way ANOVA was used for the comparison among more than two groups. All statistical calculations were performed using the GraphPad Prism 5.0 software. P,0.05 was considered as significant.

TGFb Signaling is Down-regulated in Human and Mouse HCCs
To investigate the status of TGF-b signaling pathway, we initially analyzed gene expression profile studies in Oncomine TM (Compendia Bioscience, Ann Arbor, MI) and found that in the report by Wurmbach and co-workers [25], TbRII, one of the TGF-b signaling pathway components that is frequently dysregulated in cancer [26], was significantly reduced in HCCs when compared to that in cirrhosis (Fig. 1A). To confirm this finding, we examined TbRII expression level by performing quantitative realtime RT-PCR analysis. Thirty-eight pairs of normal tissues adjacent to HCC and HCC tissues were obtained from HCC patients at the time of surgical resection. The characteristics of the HCCs and the patient information are summarized in Table S1. Similarly, TbRII transcript levels in tumor tissues were also found to be significantly decreased in comparison with those in the adjacent normal tissues (Fig. 1B). Thus, TGF-b signaling activity is likely downregulated during the development of HCC, suggesting a tumor suppressive role of TGF-b signaling pathway. Next, we measured the levels of phosphorylated Smad3 (P-Smad3) at its C-terminus as an indicator of TGF-b signaling activity in human and mouse HCCs by immunoblotting analysis. Interestingly, all six HCC had a modest to striking reduction of p-Smad3 in comparison to their adjacent normal tissues (Fig. 1C). The reduction of P-Smad3 in Patient 4 appears to be due to reduced total Smad3 level. Among five sets of mouse HCC tissues from C3HeB/FeJ male mice, which develop spontaneous HCC at a high frequency [17], we also observed a modest to striking decrease of p-Smad3 in the majority of HCCs (except Mouse 3) in comparison to the normal liver tissue and liver tissue adjacent to HCC (Fig. 1D). Thus, TGF-b signaling activity through the Smad pathway in both human and mouse HCCs appears downregulated.

TGF-b Signaling and Function in Human HCC Cell Lines
To further investigate the role of TGF-b signaling pathway in human HCCs, we evaluated expression of several TGF-b signaling pathway components including TbRI, TbRII, and Smad4 in five HCC cell lines which have shown different TGF-b responsive characteristics [27]. Among these five cell lines, only SNU398 cell showed impaired TGF-b signaling pathway with little expression of TbRII when compared with other HCC cells ( Fig. 2A). SNU423 cells also showed lower TbRI and TbRII expression whereas Sk-Hep-1, HepG2, and Huh7 cells showed higher expression ( Fig. 2A). Furthermore, we determined the response of these five cell lines to TGF-b1 or RI-KI in regulating the phosphorylation of Smad2 and Smad3 by Western blotting analysis. All showed increased P-Smad2 and P-Smad3 in response to TGF-b1 except the SNU398 cell line. RI-KI treatment reduced basal P-Smad2 and P-Smad3 in SNU423, Sk-Hep-1 and Huh7 cells (Fig. 2B) suggesting that these cells possess autocrine TGF-b signaling activity. This notion is consistent with our findings that HCC cells produce detectable levels of all three TGF-b isoforms in the media conditioned by the cells (Fig. 2C). With a TGF-b responsive promoter-luciferase reporter assay, we observed that TGF-b1 stimulated luciferase activity in SNU423, HepG2, Sk-Hep-1 and Huh7 cells, whereas RI-KI significantly attenuated the activity in these cells (Fig. 2D). In contrast, there is no effect of   Fig. 2E, TGF-b1 treatment induced various levels of growth inhibition in Huh7, HepG2, Sk-Hep-1, and SNU423 cells in a dose-dependent manner, but not in SNU398 cells. To evaluate the effect of TGF-b on in vitro tumorigenic ability of these HCC cell lines, we performed a soft-agar colony formation assay. Consistently, TGF-b1 attenuated colony formation ability of SNU423, HepG2, Sk-Hep-1 and Huh7 cells, but not SNU398 cell (Fig. 2F). Taken together, four of five HCC cell lines have an operational TGF-b/Smad signaling pathway and are growth inhibited by exogenous TGF-b1 to varying degrees in both two dimensional and three dimensional growth conditions.

Abrogation of TGF-b Signaling Pathway Inhibits HCC Cell Growth and Promotes Apoptosis
The above observations suggest that TbRII is a major target in the attenuation of TGF-b signaling activity during hepatocarcinogenesis and TGF-b treatment produced an apparent tumor suppressive activity in all HCC cell lines that are sensitive to TGFb. Interestingly, by analyzing the reported gene profiling data by Wurmbach and co-workers, TbRII expression was found to be increased in very advanced HCCs when compared to very early HCCs (Fig. 3A) [25]. Similar phenomenon was also observed in our 38 HCC tissue specimens. TbRII expression in HCCs of Edmondson grade III/IV was significantly increased in comparison to that in HCCs of Edmondson grade I/II (Fig. 3A). To determine the role of TbRII and consequently that of the TGF-b signaling pathway in regulating the malignant phenotypes of HCC cells, we knocked down TbRII in SNU423 and Sk-Hep-1 cells with the stable expression of a TbRII shRNA as described previously [22]. Results indicated that knockdown of TbRII, confirmed by Western blotting analysis and RT-PCR (Fig. 3B), reduced both basal and TGF-b1-induced P-Smad2 and P-Smad3 (Fig. 3C), as well as Smad-responsive promoter activity as reported by luciferase activity (Fig. 3D), suggesting that autocrine TGF-b signaling is also abrogated by TbRII knockdown. More interest-ingly, the constitutive abrogation of autocrine TGF-b resulted in a significant inhibition of the growth of both cell lines as measured with MTT assays (Fig. 3E), as well as a significant stimulation of apoptosis as detected with an apoptosis ELISA (Fig. 3F). These results indicate that the autocrine TGF-b signaling does not inhibit the proliferation, instead is necessary for the viability of these HCC cells.

Abrogation of TGF-b Signaling Attenuates the Anchorage-independent Colonogenic, Tumorigenic and Metastatic Potential of HCC Cells
We first determined the effect of TbRII knockdown on their in vitro tumorigenic potential with the anchorage-independent soft agar growth assay. We found that knockdown of TbRII significantly repressed the anchorage-independent growth ability of both SNU423 and Sk-Hep-1 cells when compared with their respective control cells (Fig. 4A). We also performed hard-agar colony formation assay, as it has been shown to predict the metastatic potential of tumor cells [23]. The results demonstrated that TbRII knockdown also reduced the colonogenic ability of both SNU423 and Sk-Hep-1 cells (Fig. 4B). These results suggest that TGF-b signaling is required for the maintenance of the malignancy of SNU423 and Sk-Hep-1 cells.
To confirm our in vitro findings, we next compared the growth of SNU423 and Sk-Hep-1 cells with or without TbRII knockdown in male athymic nude mice during a period of fifty-six days after inoculation of 3610 6 cells subcutaneously into rear hindquarters. Unfortunately, SNU423 cell did not form any tumors even after we inoculated a larger number of cells (5610 6 cells) mixed with matrigel, which has been shown to enhance the tumorigenicity of various transformed cells in nude mice [28,29]. In contrast, Sk-Hep-1 control and TbRII shRNA cells started to form noticeable tumors 24 days after inoculation (Fig. 4C). Interestingly, the mean tumor growth rate in the TbRII shRNA group was initially much slower than that in the control group. As a result, the mean tumor volume of the two groups became significantly different during the Annexin-V positive apoptotic cells as expressed with the percentage. (D) Immunoblotting analysis for indicated proteins in Sk-Hep-1 and Huh7 cells with or without Smad4 shRNA. (E) Cells were plated in a 6-well plate and treated with various concentrations of LY294002 for 24 hours when the culture was 70-80% confluence. Each data point represents mean 6SEM from three independent wells. (F) Immunoblotting analysis of P-Smad3L in the nuclei of Sk-Hep-1 and Huh7 with or without TbRII shRNA or Smad4 shRNA after treatment with TGF-b1 at 2 ng/ml for 1 hr. doi:10.1371/journal.pone.0063436.g006 late part of the experiment even though the mean tumor growth rates of the two groups eventually became similar (Fig. 4C). At the termination of the experiment, we harvested the tumors formed by the two cells and performed real-time RT-PCR to detect TbRII mRNA. As expected, TbRII shRNA tumors maintained the low expression of TbRII when compared with Sk-Hep-1/control tumors suggesting that the delayed increase of growth rate by TbRII shRNA tumors was not due to loss of TbRII knockdown (Fig. 4D).
Because the cell lines used for the in vivo experiments were stably transfected with a luciferase and GFP expression plasmid, we performed whole mouse bioluminescence imaging and also looked for GFP-expressing tumor cells in various visceral organs after they were excised from mice at the termination of the above experiment. No metastasis was observed with either imaging approach. Thus, to investigate how abrogation of TGF-b signaling may affect the in vivo metastatic potential of Sk-Hep-1 cells, we used an experimental metastasis model by inoculating the control and TbRII knockdown cells through tail vein. Metastasis induced by tumor cells was monitored by bioluminescence imaging every two weeks after inoculation. Consistent with the result from the hard-agar colony formation assay, the bioluminescence imaging taken four weeks after inoculation revealed that the knockdown of TbRII reduced the widespread dissemination of Sk-Hep-1 cell in nude mice (Fig. 4E). The incidence of significant metastasis in the Sk-Hep-1/control cell-inoculated mice was 100%, whereas in the Sk-Hep-1/TbRII shRNA cell-inoculated mice, it was only 20%.

Smad Pathway Mediates Growth Inhibition by Exogenous TGF-b
TGF-b-induced growth inhibition is known to be mediated by the Smad pathway. On the other hand, it has also been shown to stimulate carcinoma cell survival by signaling through Smadindependent pathways [29][30][31]. As such, we hypothesized that abrogation of Smad pathway by knocking down Smad4 should attenuate TGF-b's growth inhibitory activity while preserving the Smad-independent survival signaling of TGF-b, thus generating a different phenotype from that of the TbRII knockdown cells. As shown in Fig. 5A, expression of a Smad4 shRNA in Sk-Hep-1 and Huh7 cells reduced Smad4 protein levels in both cell lines. While the knockdown did not affect TGF-b-induced phosphorylation of Smad2 and Smad3, it led to a significant attenuation of TGF-binduced Smad-responsive promoter activity (Fig. 5B) suggesting that Smad4 knockdown significantly attenuated Smad2/3/4 activity. Consistently, both Sk-Hep-1 and Huh7 cells were less sensitive to exogenous TGF-b-induced growth inhibition when their Smad4 was knocked down (Fig. 5C). Because the cyclindependent kinase inhibitors, p15 and p21, are major effectors of TGF-b-induced cell cycle arrest [32;33], we compared their expression in the control and Smad4 knockdown cells. While p21 expression was not detected in the two HCC cell lines, immunoblotting analysis showed that TGF-b-induced p15 expression was noticeably attenuated in both Huh7 and Sk-Hep-1 cells with Smad4 knockdown (Fig. 5D). These data demonstrate Smad4 is necessary for the growth inhibitory function of TGF-b in the HCC cells.

Smad4 Supports HCC Cell Survival and Inhibits PTEN Expression
The confirmation of the growth inhibitory activity of the Smad pathway led us to expect that the Smad4 knockdown cells would grow faster and be more malignant than the control HCC cells. Thus, it was surprising to us that Smad4 knockdown in both Sk-Hep-1 and Huh7 cells led to significantly slower cell growth as detected with MTT assays (Fig. 6A) and significantly higher apoptosis as detected by both apoptosis ELISA (Fig. 6B) and Annexin-V staining assay (Fig. 6C). Thus, the knockdown of Smad4 produced identical phenotypes in the HCC cells as the knockdown of TbRII suggesting that the Smad pathway mediates both the growth inhibitory and cell survival activity of TGF-b signaling in the HCC cells. Because the tumor suppressor PTEN is downregulated in half of HCCs [34. and its expression is inhibited by both exogenous and autocrine TGF-b [19,35], we next examined whether Smad signaling supports HCC cell survival by inhibiting PTEN expression. Indeed, the knockdown of Smad4 in both Sk-Hep-1 and Huh7 cells led to increased PTEN expression with a concomitant reduction of the active phospho-AKT (Fig. 6D). These results suggest that Smad pathway mediates TGF-b-induced suppression of PTEN. The increased PTEN expression and the decreased active AKT level in the Smad4 knockdown HCC cells likely contributed to the increased apoptosis as the treatment with an inhibitor of PI3K, the activator of AKT, also induced apoptosis in HCC cells (Fig. 6E). In addition, we also measured the levels of phosphorylated Smad3 at its link region (P-Smad3L) as a function of TGF-b signaling abrogation because nuclear P-Smad3L has been shown to have tumor-promoting activity [36]. We found that the levels of P-Smad3L in cytosol were hardly detectable in the HCC cell lines. In contrast, its level in the nucleus of the Smad4 knockdown Huh7 cells was decreased in comparison with that of the control cells in the absence or presence of TGF-b treatment (Fig. 6F). Similar phenomenon was also observed in TbRII knockdown Sk-Hep-1 cells (Fig. 6F). To determine the effect of silencing Smad4 on the tumorigenicity of Sk-Hep-1 and Huh7 cells, anchorage-independent colony formation assay was performed. As in the case of the abrogation of TGFb signaling with the knockdown of TbRII, abrogation of Smad signaling in both Sk-Hep-1 and Huh7 cells also reduced their colony formation potential in both soft agar (Fig. 7A) and hard agar (Fig. 7B). Taken together, these results indicate that autocrine TGF-b/Smad signaling is indispensable for the survival and malignancy of HCC cells.

Discussion
TGF-b signaling through its cell surface receptors and Smad proteins has been demonstrated as a tumor suppressive pathway in various types of cells, particularly in gastrointestinal malignancies. Among various receptors and Smad proteins that mediate TGF-b signaling, TbRII and Smad4 are most widely inactivated via gene mutation [14]. Although mutational inactivation of TGF-b/Smad pathway components is relatively rare in HCC [11][12][13][14][15], other mechanisms that abrogate the tumor suppressive function of TGFb/Smad pathway have been reported. For example, TbRII expression was shown to be down regulated in HCC tissues in comparison with adjacent hepatic tissues [6,37]. Both HBV and HCV associated HCC tissues were shown to have reduced level of phosphorylation of Smad3 at its C-terminus, which mediates its growth inhibitory activity [38,39]. Consistent with these observations, our study also shows a significant downregulation of TbRII expression in HCC tissues and a widespread reduction of phospho-Smad3 at its C-terminus in both human and murine HCC tissues when compared to that in the adjacent hepatic tissues. Thus, our study further supports the tumor-suppressive concept of TGF-b pathway in hepatic tissue.
However, the controversy arises with respect to the role of TGFb signaling pathway in transformed HCC cells. Owing to the relatively low inactivating mutations of the TGF-b receptor and Smad genes, many HCC cell lines retain an operational TGF-b signaling pathway, which was shown to mediate the growth inhibitory activity on plastic and in soft agar by exogenous TGF-b in our study. Others have shown that treatment with exogenous TGF-b induced cellular senescence and growth inhibition in various HCC cell lines in vitro and peritumoral injection of TGFb inhibited the growth of tumors formed by Huh7 cells [40]. These observations appear to indicate that the tumor-suppressive activity of TGF-b is retained in various HCC cell lines. However, they do not address the role of autocrine TGF-b signaling in the control of malignant phenotypes of HCC cells. We sought to address this question by studying the effect of abrogation of autocrine TGF-b signaling via TbRII knockdown on the malignant properties of HCC cell lines. Our results indicate that the autocrine TGF-b signaling through its receptors is necessary for the survival and clonogenicity in suspension of both SNU423 and Sk-Hep-1 HCC cells that were used for TbRII knockdown experiments. The knockdown of TbRII also reduced the tumorigenic and metastatic properties of Sk-Hep-1 cells in vivo. Our observations are consistent with a recent report showing inhibition of hepatocarcinogenesis in hepatic specific knockout of p53 mice when Tgfbr2 was deleted in hepatic tissue [16]. Others have shown that both exogenous and autocrine TGF-b signaling stimulated the proliferation of HCC-M and HCC-T cell lines [41,42]. On the other hand, Senturk and co-workers reported that Hep3B-TR cells with deleted TGFBR2 gene were much more tumorigenic than its parental Hep3B cells [40]. However, because Hep3B-TR cell line was established after long-term growth inhibitory selection of Hep3B cells with TGF-b treatment as a TGF-b resistant cell line [43], it is not known whether other genetic alterations and gene expression profile changes that are independent of TbRII loss as reported by Zimonjic and co-workers also contributed to its malignancy.
TGF-b signaling has been shown to be mediated by both Smaddependent and Smad-independent pathways [2]. The latter includes MAPK and PI3K/AKT pathways, both of which have been shown to mediate mammary tumor cell survival and growth of TGF-b signaling [30,31]. Because the tumor suppressive activity of TGF-b signaling is believed to be mediated by the Smaddependent pathway and Smad4 plays a central role in TGF-binduced Smad transcriptional activity, we knocked down Smad4 in HCC cells to elucidate the pathway(s) that mediates autocrine TGF-b/TbRII-induced cell survival. While knockdown of Smad4 did attenuate Smad transcriptional activity, and the potency of growth inhibition and p15 induction by exogenous TGF-b, it was surprising to find that Smad4 was also necessary for the survival of both Sk-Hep-1 and Huh-7 cells suggesting that the autocrine TGF-b-induced cell survival is at least in part mediated by the Smad pathway in these model systems. Autocrine TGF-b-induced growth inhibition of HCC-M and HCC-T cell lines was shown to be associated with its suppression of the promoter activity of p15 implicating the inhibition of p15 expression as a mechanism of growth promotion by TGF-b [41]. In contrast, we did not observe this phenomenon in Sk-Hep-1 and Huh-7 cells after knocking down Smad4. On the other hand, we observed an increase of PTEN expression and a decrease of phosphorylated/activated AKT in the Smad4 knockdown cells suggesting that the reduced AKT activity may contribute to the increased apoptosis and the reduced growth potential on plastic and in soft agar. Furthermore, as reviewed by Dr Matsuzaki, the activation of MAP kinases by growth factors including TGF-b can lead to phosphorylation of Smad2 and Smad3 at their linker region and p-Smad3L is involved in oncogenic signaling when translocated into the nucleus with Smad4 [36]. Interestingly, we found that knockdown of either TbRII or Smad4 attenuated TGF-b-induced nuclear accumulation of p-Smad3L suggesting that the tumor-promoting activity of autocrine TGF-b is likely mediated in part by its stimulation of linker region phosphorylation of Smad3. Further studies are needed to test these hypotheses.
In summary, our study together with others indicates that TGFb signaling plays an important role in both suppression of HCC development and maintenance of malignant phenotypes of some HCCs. Its signaling strength appears to be finely tuned for its dichotomous actions during hepatocarcinogenesis. The Smad pathway appears to mediate the dual functions of TGF-b, likely in collaboration with different partners at different neoplastic stages. Further studies are needed to elucidate how Smad signaling network is altered during hepatocyte transformation to transduce a survival signal in HCC cells. Future studies will also determine whether HCC is uniquely suited for therapeutic intervention with novel TGF-b inhibitors since the Smad pathway in HCC cells is necessary for their survival rather than their growth inhibition.