Reversible Smad-Dependent Signaling between Tumor Suppression and Oncogenesis Free

Smads are central mediators of signals from the receptors for transforming growth factor-β (TGF-β) superfamily members to the nucleus (1). Smads are modular proteins with conserved Mad-homology 1 (MH1), intermediate linker, and MH2 domains (2). The catalytically active TβRI phosphorylates the COOH-terminal serine residues of receptor-activated Smads (3), which are Smad2 and the highly related protein Smad3. The linker domain can undergo regulatory phosphorylation by other kinases, including mitogen-activated protein kinases such as extracellular signal-regulated kinase (Erk), c-Jun NH₂-terminal kinase (JNK) and p38, or cyclin-dependent kinases (4–7). Phosphorylated Smad2 and Smad3 rapidly oligomerize with Smad4, forming functional trimeric protein complexes (2). Although monomeric Smad proteins constantly shuttle in and out of the nucleus, formation of the activated Smad complex favors their nuclear accumulation (2). In the nucleus, the Smad complex binds directly to DNA and associates with a plethora of transcription factors, coactivators, or corepressors, leading to transcriptional induction or repression of the target genes (3).

TGF-β can inhibit epithelial cell growth (8), acting as a tumor suppressor. As currently understood, loss of sensitivity to growth inhibition by TGF-β in most cancer cells is not synonymous with complete shutdown of all TGF-β signaling. Instead, cancer cells gain advantage by selective reduction of the tumor-suppressive activity of TGF-β together with augmentation of its oncogenic activity (9, 10). Such a state of altered TGF-β responsiveness is observed in Ras-transformed cells, which typically exhibit a limited growth-inhibitory response to TGF-β, while acquiring ability to invade tissues and form metastases (11). One, therefore, needs to distinguish between the tumor-suppressive function of TGF-β in epithelial cells and its tumor-promoting role in Ras-transformed cells. In the present study, we sought to better understand molecular mechanisms regulating responses to TGF-β in these cells.

Domain-specific antibodies against phosphorylated Smad2 and Smad3. Nine polyclonal anti–phospho-Smad3 and anti–phospho-Smad2 sera were raised against the phosphorylated linker and COOH-terminal regions of Smad2 and Smad3 by immunization of rabbits with synthetic peptides (Supplementary Fig. S1). Relevant antisera were affinity purified using the phosphorylated peptides as previously described (6).

Constructs. H-Ras^V12 was inserted into a pRX retroviral vector together with blasticidin deaminase gene. Smad3WT, Smad3EPSM, and Smad3(3SA) were inserted into another retroviral vector, pQCXIN (Clontech), containing a puromycin resistance marker. Integrity of the constructs was confirmed by sequencing.

Retroviral infection, expression of Ras^V12, and mutant Smad3. An ecotropic virus packaging cell line, BOSC23, was seeded at 1 × 10⁶ cells per 100-mm dish. Twenty-four hours later, the cells were subjected to transfection with LipofectAMINE (Invitrogen) and 3 μg of the indicated constructs for 4 h. After culture of the cells for 48 h, the filtrated culture supernatants were used for infection.

Parental RGM1 cells (1 × 10⁵ per a six-well plate) were infected with a retroviral solution carrying H-Ras^V12 in the presence of 8 μg/mL polyblene (Sigma) for 6 h, and then were incubated in 10% FCS/DMEM. After trypsinizing and replating the cells at a 1:10 dilution on the 3rd day, selection was initiated with 1 μg/mL blasticidin (Funacoshi), continuing for several weeks until blasticidin-resistant Ras-transformed cells (RGMRas) emerged.

RGMRas cells were infected additively with other retroviral solutions carrying Smad3WT, Smad3EPSM, or Smad3(3SA). Selection was initiated with 3 μg/mL puromycin (Nacalai Tesque), continuing for several weeks until puromycin-resistant RGMRas cells expressing Smad3 derivatives emerged. Five colonies per each infection were cloned and were subjected to assay or passage.

Immunoprecipitation and immunoblotting. Immunoprecipitation and immunoblotting of endogenous Smad2 and Smad3 were done as previously described (6).

Immunoblots of total cell lysates also were analyzed using 0.5 μg/mL anti-Ras antibody (BD Biosciences), 5 μg/mL anti–β-actin (Sigma), 3 μg/mL anti–phosphorylated JNK1/2 antibody (Promega), 0.1 μg/mL anti-JNK1/2 antibody (Cell Signaling), 1 μg/mL anti–c-Myc antibody (Santa Cruz Biotechnology), 0.1 μg/mL anti–plasminogen activator inhibitor-1 (PAI-1) antibody (BD Bioscience), 1 μg/mL anti–matrix metalloproteinase 1 (MMP1) antibody (Chemicon International), 1 μg/mL anti-MMP2 antibody (Chemicon International), and 0.5 μg/mL anti-MMP9 antibody (Daiichi Fine Chemicals).

In vitro kinase assay. Bacterial expression and purification of glutathione S-transferase (GST)-Smad2 and GST-Smad3 were carried out according to the manufacturer's instructions (GE Healthcare). Endogenous kinases were isolated from the protein extracts using anti-pJNK1/2 antibody (Promega). Immunocomplexes, collected with protein G-Sepharose, were suspended in kinase assay buffer supplemented with 100 μmol/L ATP and bacterially expressed GST-Smad2 or GST-Smad3. Assays were carried out as described previously (5). Degrees of phosphorylation of Smad2/3 were monitored by immunoblotting using each domain-specific phospho-Smad2/3 antibody.

Immunofluorescence. The subcellular localization of Smads was determined as previously described (5). To block binding of anti-pSmad3C antibody to phosphorylated domains in Smad2, anti-pSmad3C antibody was adsorbed with 1 μg/mL COOH-terminally phosphorylated Smad2 peptide.

[³H]thymidine incorporation. DNA synthesis was measured by incorporation of 1 μCi/mL [³H]thymidine (GE Healthcare) into 5% trichloroacetic acid–precipitable material after 4-h pulse as described previously (5).

Matrigel invasion assay. Membranes with 8-μm pores covered with Matrigel (BD Biosciences) on the upper surface were coated with type I collagen on the lower side. Infiltrating cells were counted in five regions selected at random as described previously (5).

As lack of antibodies able to selectively distinguish phosphorylation sites in Smad2 and Smad3 has impeded investigation of the role of each phosphorylation domain in TGF-β signaling, we generated nine antibodies directed at various phosphorylation sites (Fig. 1A), and then verified that anti–phospho-Smad2/3L antibody would react only with specific phosphorylated domains in the linker regions (Supplementary Fig. S1).

Initially, we investigated TGF-β signaling in well-characterized cells (RGM1), which were isolated from normal rat gastric epithelial cells and were sensitive to growth inhibition by TGF-β (5). Previous work showed that both TGF-β and hepatocyte growth factor (HGF) were physiologic activators of the JNK pathway, which was shown to have important implications for Smad2 and Smad3 signaling (12, 13). In support of this notion, TGF-β or HGF treatment caused inducible JNK phosphorylation in RGM 1 cells (Fig. 1B). Subsequently, the signal stimulated linker phosphorylation of Smad2 and Smad3 (Fig. 1C). The JNK inhibitor SP600125 inhibited TGF-β– or HGF-dependent linker phosphorylation in vivo. In addition, JNK activated by TGF-β or HGF signals could directly phosphorylate Smad2 and Smad3 at linker regions in vitro (Fig. 1D). We conclude from these findings that the JNK/pSmad2/3L pathway can be activated in response to TGF-β or HGF signal in the immortalized epithelial cells.

We previously reported that TGF-β signaling converted Smad2 and Smad3 into distinct phosphoisoforms: COOH-terminally phosphorylated Smad2/3 (pSmad2/3C) and linker phosphorylated Smad2/3 (pSmad2/3L; ref. 14). Translocation of Smad2 and Smad3 into the nucleus upon COOH-terminal phosphorylation by TβRI is a central event in TGF-β signal transduction (3). To gain additional insight into the significance of linker phosphorylation, we examined intracellular localization of each Smad2/3 phosphoisoform in RGM1 cells in response to TGF-β or HGF (Fig. 2A). As expected, most pSmad2C and pSmad3C were located in RGM1 cell nuclei after TGF-β treatment. In contrast, exposure to excess HGF did not lead to nuclear accumulation of pSmad2C or pSmad3C. Although weak pSmad2L (Ser²⁴⁵ and Ser²⁵⁰) and pSmad3L (Ser²⁰⁴) staining remained in the cytoplasm of RGM1 cells after TGF-β or HGF treatment, these treatments led to nuclear translocation of pSmad2/3L (Thr), pSmad2L (Ser²⁵⁵), and pSmad3L (Ser²⁰⁸ and Ser²¹³). Likewise, Smad2/3 phosphorylation by activated JNK has been shown to facilitate nuclear accumulation of Smad2 and Smad3 (12, 13). Taken together, the various results show that the linker phosphorylation can allow the Smad2/3 phosphoisoforms to translocate into nuclei via the activated JNK pathway, irrespective of COOH-terminal phosphorylation.

Ras participates importantly in human carcinogenesis; mutational activation of Ras is frequent in human cancer, and facilitates tumor invasion and metastasis. To investigate whether excessively active Ras altered TGF-β signaling in immortalized epithelial cells, hyperactive Ras was expressed in RGM1 cells by retroviral infection using a vector carrying H-Ras^V12, which had a simple amino acid replacement of a glycine residue by valine. This substitution represents the critical change in conversion of the proto-oncogene to an active oncogene (15). Ras-transformed cells (RGMRas) selected by exposure to blasticidin (Fig. 1B) had a fibroblast phenotype, resisted growth inhibition by TGF-β, and showed increased invasiveness.

Hyperactive Ras resulted in sustained JNK activation (16). Similarly to the JNK phosphorylation profile (Fig. 1B), the linker regions of Smad2 and Smad3 were constitutively phosphorylated in RGMRas cells (Fig. 1C). SP600125 inhibited the linker phosphorylation in vivo. In addition, JNK activated by Ras-mediated signal could directly phosphorylate Smad2 and Smad3 at linker regions in vitro (Fig. 1D). These results indicate that the linker regions are constitutively phosphorylated via the Ras/JNK pathway. Although nuclear translocation of pSmad2L (Ser²⁴⁵ and Ser²⁵⁰) and pSmad3L (Ser²⁰⁴) required TGF-β addition in RGMRas cells, pSmad2/3L (Thr), pSmad2L (Ser²⁵⁵), and pSmad3L (Ser²⁰⁸ and Ser²¹³) were located in nuclei without exposure to exogenous TGF-β and HGF (Fig. 2A). In contrast, neither basal nor TGF-β–dependent Smad3 phosphorylation at the COOH-terminal region was demonstrable (Fig. 1C). Impaired Smad3 phosphorylation at the COOH-terminal region was not a result of TβRI inactivation because Ras-transformation did not interfere with TβRI-mediated Smad2 phosphorylation at the COOH-terminal region. On the other hand, RGMRas cells showed restoration of TGF-β–dependent Smad3 phosphorylation at the COOH-terminal region upon treatment with SP600125. Taken together, the results indicate that a high degree of JNK-dependent Smad3 phosphorylation at the linker region in Ras-transformed cells indirectly suppresses Smad3 phosphorylation at the COOH-terminal region.

A JNK inhibitor can block alternative Smad-independent signaling pathway in the nucleus (17). To inactivate Smad2/3L selectively, a Smad3 mutant lacking four phosphorylation sites in the linker region (Erk/prodirected kinase site mutant; Smad3EPSM) was expressed in an additive manner in RGMRas cells by retroviral infection using another vector, carrying a puromycin resistance gene. RGMRas cells additively expressing wild-type Smad3 (Smad3WT) or Smad3(3SA), in which three COOH-terminal serine residues phosphorylated by TβRI were changed to alanine, were used as controls. Although endogenous Smad2 and Smad3 were phosphorylated constitutively at linker regions in RGMRas cells expressing Smad3WT and Smad3(3SA), their linker phosphorylation dramatically reduced in RGMRas cells expressing Smad3EPSM (Fig. 2B). Blockade of the linker phosphorylation by highly expressed Smad3EPSM did not affect Smad-independent signaling pathway because RGMRas cells retained a high degree of c-Jun and ATF2 phosphorylation, despite highly expressed Smad3EPSM (Fig. 2C). These results suggest that high Smad3EPSM expression specifically blocks linker phosphorylation of endogenous Smad2 as well as Smad3 in a dominant-negative manner.

We further investigated oncogenic Smad signaling in RGMRas cells. Tumor cell invasion, the first step toward metastasis, requires complex interactions including recognition and attachment of tumor cells to extracellular matrix (ECM)–binding sites, proteolytic dissociation of the ECM, and tumor cell migration within the surrounding tissue. In particular, because degradation of the ECM is conspicuous, enzymes with a proteolytic effect on the ECM such as MMP-1, MMP-2, and MMP-9 have been investigated (18). In addition, PAI-1 facilitates cell migration and invasion by enhancing cell adhesion. Reflecting the pSmad2/3L profile in RGMRas cells (Fig. 1C), PAI-1, and the MMPs involved in cell invasion were constitutively up-regulated (Fig. 2D). Although expression of Smad3WT and Smad3(3SA) in RGMRas cells did not affect amounts of PAI-1 or the MMPs, expression of Smad3EPSM notably reduced these invasion-related proteins in RGMRas cells. Moreover, RGMRas cells expressing Smad3EPSM showed less capacity to invade in a chamber assay than the cells expressing Smad3WT or Smad3(3SA) (Fig. 3A and B). The results suggest that pSmad2/3L–mediated signaling maintains overexpression of PAI-1 and the MMPs that promotes malignant behavior in Ras-transformed cells.

Consistent with restoration of Smad3 phosphorylation at the COOH-terminal region upon treatment to RGMRas cells with SP600125 (Fig. 1C), RGMRas cells expressing Smad3EPSM showed TGF-β–dependent Smad3 phosphorylation at the COOH-terminal region (Fig. 2B). This restoration could be explained in terms of remobilization of the Smad3 molecule from the nucleus to the cytoplasm where it would have access to membrane-anchored TβRI. In sum, hyperactive Ras drastically alters TGF-β signaling through the JNK pathway, increasing basal nuclear pSmad2/3L activity while shutting down TGF-β–dependent pSmad3C available for action in the nuclei. This could account for a lack of TGF-β–dependent Smad3C phosphorylation in cell nuclei of sporadic human colorectal cancer (19) and hepatocellular carcinoma (20).

TGF-β inhibits cell growth via Smad-mediated transcriptional regulation of critical regulators of the cell cycle (1). The first direct transcriptional target of the TGF-β pathway that explains how this cytokine inhibits proliferation of epithelial cells is c-Myc (8), the expression of which in parental RGM1 cells was repressed by TGF-β (Fig. 2D). Aberrant expression of c-Myc in RGMRas cells (Fig. 2D) might contribute to resistance to the growth suppression in response to TGF-β (Fig. 3C and D), because c-Myc actively represses expression of critical cell cycle regulatory genes like p15^Ink4B and p21^Cip1 (2). Similarly to restoration of TGF-β–dependent Smad3 phosphorylation at the COOH-terminal region in RGMRas cells expressing Smad3EPSM (Fig. 2B), the cells exhibited TGF-β–dependent inhibition of c-Myc expression (Fig. 2D) and [³H]thymidine incorporation (Fig. 3C and D) comparable with findings in parental RGM1 cells.

In this study, we showed that TGF-β transmitted a signal through TβRI-dependent pSmad3C, participating in the cytostatic response by repressing transcriptional activity of c-Myc gene. On the other hand, Ras-activated JNK/pSmad2/3L signaling alone was able to provide oncogenic potential to the epithelial cells via up-regulation of c-Myc, PAI-1, MMP-1, MMP-2, and MMP-9, resulting in strongly enhanced tumor growth and invasion. Taken together, domain-specific phosphorylation of Smad2 and Smad3 is a key determinant regulating transcriptional activation of several target genes, ultimately selecting either tumor suppression or oncogenesis (Fig. 4).

Oncogenic Ras has been reported to activate Erk1/2, which directly phosphorylate the linker regions of Smad2 and Smad3 (4), with consequent blockage of all Smad signaling including Smad-dependent transcriptional activities of several target genes. However, our current results showed that hyperactive Ras constitutively activated the JNK pathway (Fig. 1B), leading to sustained linker phosphorylation of Smad2 and Smad3 (Fig. 1C and D), their nuclear translocation (Fig. 2A), and expression of PAI-1 and MMPs (Fig. 2D). Accordingly, exogenous TGF-β and HGF were unable to additionally enhance the JNK/pSmad2/3L–mediated invasive capacity of the Ras-transformed cells (Fig. 3A and B). In support of this notion, selective blockade of the linker phosphorylation in the Ras-transformed cells by a mutant Smad3 lacking JNK-dependent phosphorylation sites (Fig. 2B) resulted in minimal expression of PAI-1 and MMPs (Fig. 2D), and consequent disappearance of invasion by the cells (Fig. 3A and B). Taking the findings together, we conclude that oncogenic TGF-β signaling results from the functional collaboration of Ras and Smad rather than from Ras-mediated inhibition of the Smad pathway.

Deepening molecular understanding of signaling pathways closely associated with changes in human tissues during carcinogenesis has spurred and guided efforts to develop new molecularly targeted therapeutics for human cancer. The intrinsic value of target evaluation in model systems ultimately lies in the extent to which these systems accurately represent characteristics of human disease. In this respect, we have reported that TGF-β signaling conferred a selective advantage upon tumor cells by shifting from a tumor-suppressive TβRI/pSmad3C pathway to an oncogenic JNK/pSmad3L pathway during sporadic human colorectal carcinogenesis (19), an observation extended to hepatic carcinogenesis (20).

Our current results showed reversibility of Smad-dependent signaling between tumor suppression and oncogenesis (Fig. 4). By using genetic as well as pharmacologic approaches, we showed that blockade of linker phosphorylation abolished oncogenic properties in Ras-transformed cells and restored the pSmad3C-mediated tumor-suppressive function present in parental epithelial cells. A key therapeutic aim in cancer would be restoration of the lost tumor-suppressive function observed in normal epithelial cells, together with disruption of fundamental signaling pathways that enable tumors to grow and invade. Accordingly, we have reason to hope that specific inhibition of the JNK/pSmad3L pathway can suppress progression of human cancer by a shift from oncogenesis to tumor suppression. In evaluating effectiveness of targeted therapies for human cancer, pSmad2/3L and pSmad3C should serve as useful biological markers that measure patient responses.

Grant support: the Ministry of Education, Science, and Culture of Japan (K. Matsuzaki and K. Okazaki).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. R. Derynck (University of California at San Francisco, San Francisco, CA) for providing us with cDNAs encoding human Smad2 and Smad3; C. Kitano for assistance in constructing ecotropic retrovirus; and N. Ohira for assistance with immunoblotting.