The transforming activity of Ski and SnoN is dependent on their ability to repress the activity of Smad proteins.

The regulation of cell growth and differentiation by transforming growth factor-beta (TGF-beta) is mediated by the Smad proteins. In the nucleus, the Smad proteins are negatively regulated by two closely related nuclear proto-oncoproteins, Ski and SnoN. When overexpressed, Ski and SnoN induce oncogenic transformation of chicken embryo fibroblasts. However, the mechanism of transformation by Ski and SnoN has not been defined. We have previously reported that Ski and SnoN interact directly with Smad2, Smad3, and Smad4 and repress their ability to activate TGF-beta target genes through multiple mechanisms. Because Smad proteins are tumor suppressors, we hypothesized that the ability of Ski and SnoN to inactivate Smad function may be responsible for their transforming activity. Here, we show that the receptor regulated Smad proteins (Smad2 and Smad3) and common mediator Smad (Smad4) bind to different regions in Ski and SnoN. Mutation of both regions, but not each region alone, markedly impaired the ability of Ski and SnoN to repress TGF-beta-induced transcriptional activation and cell cycle arrest. Moreover, when expressed in chicken embryo fibroblasts, mutant Ski or SnoN defective in binding to the Smad proteins failed to induce oncogenic transformation. These results suggest that the ability of Ski and SnoN to repress the growth inhibitory function of the Smad proteins is required for their transforming activity. This may account for the resistance to TGF-beta-induced growth arrest in some human cancer cell lines that express high levels of Ski or SnoN.

the molecule shows little homology among the family members. The N-terminal homology region is necessary and sufficient for the known biological activities of Ski and SnoN (6,7). Compared with c-Ski, v-Ski is truncated at the carboxyl terminus (5,8). However, this truncation does not appear to be responsible for the activation of ski as an oncogene because overexpression of wild-type c-Ski results in oncogenic transformation of chicken and quail embryo fibroblasts (9). Thus, the transforming activity of Ski is likely due to overexpression, not truncation, of the c-Ski protein. Consistent with this notion, an elevated level of c-Ski or c-SnoN has been detected in many human tumor cell lines derived from neuroblastoma, melanoma, breast cancer, and carcinomas of the stomach, chorion, thyroid, and epidermoid (1,10,11). In addition to up-regulation of ski expression, mislocalization of Ski may also contribute to malignant progression. Ski was found to be present in the nucleus in cells derived from normal skin or early-stage tumors, but localized in the cytoplasm in highly malignant melanoma cells (11). However, the mechanism of transformation by Ski and SnoN has not been defined.
Ski and SnoN are incorporated into the histone deacetylase-1 complex through binding to the nuclear hormone receptor corepressor N-CoR and mSin3A and mediate transcriptional repression of the thyroid hormone receptor, Mad, and pRb (12,13). Ski has also been shown to interact with the retinoic acid receptor in a ligand-independent manner, as well as with pRb and another transcription factor, Skip (14,15). However, the role of these interactions in transformation by Ski or SnoN remains to be determined.
We (17,18,21) and others (16,19,20) have recently shown that Ski and SnoN interact with the Smad proteins to negatively regulate transforming growth factor-␤ (TGF-␤) 1 or bone morphogenic protein signaling. Smad proteins are critical components of the TGF-␤ signaling pathways (22,23). In the absence of TGF-␤, the two highly homologous R-Smad proteins (Smad2 and Smad3) are distributed mostly in the cytoplasm (24 -26). Upon ligand binding, the activated type I TGF-␤ receptor kinase phosphorylates the R-Smad proteins, allowing them to translocate into the nucleus (25)(26)(27)(28)(29)(30)(31) and to form heteromeric complexes with Smad4 (32)(33)(34)(35). In the nucleus, the Smad complexes interact with various cellular partners and participate in diverse downstream activities. The Smad proteins can bind to the TGF-␤-responsive promoter DNA either directly or in conjunction with other sequence-specific DNAbinding proteins (36). Through the C-terminal Mad homology-2 (MH2) domains, the Smad proteins interact with general or promoter-specific transcriptional coactivators and corepressors to regulate the transcription of various TGF-␤ target genes (36). The Smad proteins play a central role in mediating the growth inhibitory response of TGF-␤ by activating the expression of cyclin-dependent kinase inhibitors such as p21 CIP1 and p15 INK4B (37)(38)(39)(40). Because of this, the Smad proteins are considered to be important tumor suppressors. Inactivation of the Smad proteins either by deletion or by mutation has been found to accompany the malignant progression of many human cancer cells (41). In addition, Smad proteins can also be inactivated through interaction with other proteins (16 -20, 42, 43).
Interaction of Ski or SnoN with the Smad proteins results in disruption of an active heteromeric Smad complex (16,44), displacement of the transcriptional coactivator p300/CBP from the Smad proteins (44), and recruitment of the nuclear hormone receptor corepressor N-CoR (17,18). Through these mechanisms, Ski and SnoN repress the ability of the Smad proteins to mediate TGF-␤-induced cell cycle arrest. Because the Smad proteins are tumor suppressors, we hypothesized that this ability of Ski and SnoN to inactivate Smad function may be responsible for their transforming activity. To test whether the interaction of Ski/SnoN with the Smad proteins is indeed responsible for the transcriptional repression of the Smad proteins and for the oncogenic activity of Ski and SnoN, we carried out biochemical and structural analyses to map the Smad-binding sites in Ski and SnoN. We recently solved the crystal structure of a Ski fragment bound to the Smad4 MH2 domain (44). Based on this structure and on our mutational analysis, we have identified the amino acid residues in Ski and SnoN that mediate interaction with the Smad proteins. This has allowed us to determine whether Ski/SnoN mutants defective in binding to the Smad proteins can still repress Smad function and induce oncogenic transformation of chicken embryo fibroblasts.

EXPERIMENTAL PROCEDURES
Cells, Antisera, and Constructs-Hep3B, a human hepatoma cell line (American Type Culture Collection), was maintained in minimal essential medium supplemented with 10% fetal bovine serum. 293T and Phoenix-Eco cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Ba/F3, a pro-B cell line, was maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum and 10% WEHI cell-conditioned medium as a source of interleukin-3 (45). Primary cultures of chicken embryo fibroblasts (CEFs) were prepared from 10-day-old embryos and cultured as described (46,47). CEFs were maintained in Dulbecco's modified Eagle's medium/nutrient mixture F-10 supplemented with 2% tryptose phosphate broth, 1% chicken serum, and 1% bovine calf serum.
Antiserum against N-CoR (sc-1609) was purchased from Santa Cruz Biotechnology. Anti-FLAG monoclonal antibody M2 was purchased from Sigma. Alexa Fluor® 488 goat anti-mouse IgG (H ϩ L) was purchased from Molecular Probes, Inc.
Transfection and Retroviral Infection-293T and Hep3B cells were transiently transfected using the LipofectAMINE Plus protocol (Invitrogen). To generate stable Ba/F3 cell lines overexpressing Ski or SnoN, FLAG-tagged wild-type (WT) and mutant Ski and SnoN in the pBABEpuro or pMX-IRES-GFP retroviral vector were transfected into Phoenix-Eco packaging cells to generate retroviruses. 48 h after transfection, 2.8 ml of viral supernatant was collected and added to 5 ϫ 10 5 Ba/F3 cells in the presence of 6 g/ml Polybrene. Following centrifugation at 1000 rpm for 2 h at 37°C, the cells were resuspended in 2 ml of RPMI 1640 complete medium and co-cultivated with the transfected Phoenix-Eco packaging cells. 24 h later, the Ba/F3 cells were removed from the Phoenix-Eco cells and cultured in fresh RPMI 1640 complete medium for an additional 24 h. The infected cells were selected in RPMI 1640 medium containing 2 g/ml puromycin (for pBABEpuro) or by fluorescence-activated cell sorting based on expression of the green fluorescence protein (for pMX-IRES-GFP) (Sigma).
Immunoprecipitation and Western Blotting-FLAG-and HA-tagged proteins were isolated from transfected 293T cell lysates by immunoprecipitation with anti-FLAG antibody-agarose, followed by elution with the FLAG peptide and analysis by Western blotting as described previously (18). Endogenous N-CoR was isolated from 293T cell lysates by immunoprecipitation with anti-N-CoR antibody.
Growth Inhibition and Transcription Reporter Assays-For the growth inhibition assay, 2 ϫ 10 4 Ba/F3 cells were incubated with various concentrations of TGF-␤1 for 4 days. The cells were then counted and compared with unstimulated cells to determine the percent growth inhibition (45).
For the transcription reporter assay, a total of 2.5 g of DNA were transfected into Hep3B cells. To examine TGF-␤-induced transcriptional activation, Hep3B cells were transfected with 0.5 g of p3TP-lux or pLuc800 (a luciferase reporter driven by the 800-bp natural promoter region of the PAI-1 gene) (49) and 2 g of SnoN or with various concentrations of Ski together with 0.5 g of p3TP-lux. Luciferase activity was measured 16 h after stimulation with 50 pM TGF-␤1.
Glutathione S-Transferase (GST) Pull-down Assay-Recombinant WT or mutant Ski was expressed in and purified from Escherichia coli as GST fusion proteins. To test whether these GST-Ski proteins bind to Smad3, 1.5 g of GST-Ski immobilized on glutathione-Sepharose (Amersham Biosciences) was blocked at 4°C for 30 min with bacterial cell lysates, followed by 0.2% bovine serum albumin in GST binding buffer (20 mM HEPES, 10% glycerol, 0.2 mM EDTA, 1 mM dithiothreitol, 100 mM NaCl, 0.5% Nonidet P-40, and 0.25 mM KCl) (21). The immobilized GST-Ski protein was then incubated with 293T cell lysate expressing FLAG-Smad3 for 1 h in GST binding buffer at 4°C. After extensive washes, Smad3 associated with GST-Ski was eluted with glutathione and detected by immunoblotting with anti-FLAG monoclonal antibody.
Soft Agar Colony Assay-10 6 CEFs were incubated with 10 g of DNA (FLAG-Ski or FLAG-SnoN in RCAS.BP) in the presence of 30 g/ml Polybrene. 6 h later, cells were shocked with 25% Me 2 SO for 3.5 min. The transfected CEFs were passaged for 3 weeks, and the percentage of infected cells at this stage was Ͼ70%. For the soft agar colony assay, 4 ml of normal growth medium containing 0.66% agar was poured into a p50 dish to form the bottom layer (50). 10 4 cells were then suspended in 2 ml of medium containing 0.44% agar and overlaid on the hardened bottom layer. 4 ml of fresh medium containing 0.44% agar was added to the dish every week. After 3 weeks of incubation, colonies were visualized by staining with 0.1 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (Sigma) and scanned on a Hewlett-Packard ScanJet J300C to visualize colonies.
Immunofluorescence-To monitor the expression and localization of the introduced WT and mutant Ski and SnoN proteins, CEFs growing on glass coverslips were fixed with 4% paraformaldehyde. The FLAG-Ski and FLAG-SnoN proteins were detected by staining with anti-FLAG monoclonal antibody, followed by Alexa Fluor® 488-conjugated goat anti-mouse antibodies. Nuclei were detected by 4,6-diamidino-2phenylindole staining. Positively stained cells were scored for each transfected construct and used to calculate the efficiency of transfection/infection.
Pulse-Chase Assay-293T cells growing in p50 dishes were washed with 2 ml of Dulbecco's modified Eagle's medium lacking Met and Cys and pulsed with 0.2 mCi/ml 35 S-express for 30 min. The cells were then chased in Dulbecco's modified Eagle's complete medium for various periods of time.

Mapping of Smad2-, Smad3-, and Smad4-binding Sites in
Ski and SnoN-A series of deletion mutants of Ski were constructed to identify the Smad3-binding sites in Ski (Fig. 1A, left panel). GST fusions of WT and mutant forms of Ski were immobilized on glutathione-Sepharose and used to precipitate FLAG-tagged Smad3 from transfected cell lysates. Ski fragments containing either residues 1-241 (Fig. 1A, right panels, lane 5) or 241-441 (lane 9) interacted with Smad3, suggesting that, at least in vitro, there are two Smad3-binding sites in Ski. Because Ski-(1-241) associated with Smad3 with a much higher affinity than Ski-(241-441), the major binding site is located between residues 1 and 241. Further deletion analysis indicated that residues 16 -23 are required for binding to Smad3 since fragment 16 -728 bound to Smad3 to a similar extent as WT Ski, but fragment 24 -728 interacted with Smad3 with a much lower affinity (lanes 1-3 and 8). Furthermore, mutations that changed residues 16 -23 to alanine either abolished (16 -19A) or greatly impaired (20 -23A) this interaction (lanes 6 and 7).
The minor Smad3-binding site has been previously mapped to the region between residues 241 and 323 (17). Further analysis using deletion and point mutants of Ski showed that residues 316 -319 are required for this weak interaction (data not shown). Mutant Ski-(241-441) with these residues changed to alanine no longer bound to Smad3 in vitro (Fig. 1A, right panels, lane 10). This minor binding site does not appear to be sufficient to mediate the Ski-Smad interaction or to repress TGF-␤ signaling in vivo. Mutant Ski or SnoN lacking the major Smad3-binding site but still retaining this minor binding site did not associate with Smad2 or Smad3 in vivo (Fig. 1, B and C, right panels), nor was this minor binding site alone able to mediate repression of TGF-␤-induced transcriptional activation (data not shown). Therefore, we focused on the major Smad3-binding site in all subsequent binding and functional assays.
To examine whether residues 16 -19 in Ski are also required for binding to the Smad proteins in vivo, we mutated these residues to alanine (mS2/3) (Fig. 1B, left panel). Epitope-tagged WT or mS2/3 mutant Ski was cotransfected with HA-tagged Smad proteins, and their interactions were analyzed by coimmunoprecipitation. As shown in Fig. 1B (right panels), mutation of residues 16 -19 disrupted the interaction of Ski with both Smad2 and Smad3 (but not Smad4) in vivo, indicating that Smad2 and Smad3 bind to the same residues in Ski. This is not surprising since Smad2 and Smad3 are 91% identical in amino acid sequence. Furthermore, these residues are well conserved in SnoN because mutation of the corresponding residues in SnoN (residues 85-88, mS2/3) also disrupted binding of SnoN to Smad2 or Smad3 in vivo (Fig. 1C, right panels).
The Smad4-binding site in Ski and SnoN was determined by mutagenesis and by structural studies. We have previously shown that His 222 and Glu 223 in Ski are important for binding to Smad4 (21). In the crystal structure of a Ski fragment bound to the Smad4 MH2 domain, both residues contribute to the extensive interface between Smad4 and Ski, with Glu 223 generating an intermolecular hydrogen bond and His 222 stabilizing the Ski structure (44). In addition, several amino acid residues, including Trp 274 , Thr 271 , and Cys 272 , also make direct contact with Smad4 (44). Mutation of these residues either individually (mS4 W ) or in combination (mS4 HE ) abolished binding of Ski to Smad4 in vivo in a co-immunoprecipitation assay (Fig. 1B,  right panels) (44). These residues are well conserved in SnoN (1). Mutation of His 266 , Glu 267 , or Trp 318 greatly impaired binding of SnoN to Smad4 (Fig. 1C, right panels).
Taken together, our results indicate that the R-Smad proteins (Smad2 or Smad3) and Co-Smad (Smad4) interact with different regions in Ski or SnoN. Amino acid residues 16 -19 in Ski or residues 85-88 in SnoN are required for interaction with Smad2 or Smad3, whereas multiple residues, including His 222 , Glu 223 , and Trp 274 in Ski (or the corresponding residues in SnoN), are critical for interaction with Smad4. Mutation of both regions completely abolished the association of Ski or SnoN with the Smad proteins in vivo (mS3S4 W and mS3S4 HE ) (Fig. 1, B and C, right panels).
Interaction between Ski/SnoN and the Smad Proteins Is Required for Repression of TGF-␤-induced Transcriptional Activation-We (17,18) and others (16,19,20) have shown previously that Ski and SnoN can be recruited to the SBEs through interaction with the Smad proteins and repress transcriptional activation of TGF-␤ target genes through multiple mechanisms. In an electrophoretic mobility shift assay, WT Ski or SnoN immunoprecipitated from cells cotransfected with Smad4 ( Fig. 2A) or Smad3 (data not shown) formed a complex with the SBE that could be supershifted with antibodies directed against the Smad protein or epitope tags of Ski or SnoN ( Fig.  2A, lanes 2-4 and 8 -10) (16 -20). No such complex was detected with mutant forms of Ski or SnoN defective in binding to the Smad proteins (mS3S4 HE ) (lanes 5-7 and 11-13), indicating that the interaction of Ski or SnoN with the Smad proteins is necessary for binding to the SBE.
When cotransfected with the TGF-␤-responsive reporter constructs, WT Ski or SnoN readily repressed TGF-␤-induced transcriptional activation of either the natural PAI-1 promoter or p3TP-lux, but the mS3S4 mutants failed to do so (Fig. 2,  B-D). Interestingly, mutation of either the R-Smad-or Smad4binding site alone did not significantly affect the repressive activity of SnoN (Fig. 2, C-D) and Ski (data not shown), suggesting that they can inactivate the activity of the heteromeric Smad complex through binding to either Smad protein. This is consistent with our previous structural analysis showing that binding of Ski to either Smad2 or Smad4 results in disruption of the heteromeric Smad complex (44). Only after both binding sites are mutated is the repression by Ski or SnoN abolished.
To confirm that the lack of repression by mS3S4 mutant Ski or SnoN is not due to a destabilization of the structure of Ski or SnoN by the mutation, we measured the half-lives of mutant Ski and SnoN and examined their ability to interact with the corepressor N-CoR. Because misfolded proteins are usually degraded rapidly, we hypothesized that if mutation of Smad-binding sites affects the folding of Ski and SnoN, the mS3S4 mutants should be less stable than WT Ski and SnoN. Their interaction with N-CoR may also be affected by these mutations. However, in a pulse-chase assay (Fig. 3A), mS3S4 mutant Ski and SnoN were as stable as their WT counterparts. In fact, the SnoN mutant was even more stable in the presence of Smad3 (Fig. 3A) or TGF-␤1 (data not shown) than WT SnoN, presumably because it does not bind to Smad3 and cannot be degraded as a result of Smad3-dependent polyubiquitination (38). In a co-immunoprecipitation assay, mS3S4 mutant Ski or SnoN bound to endogenous N-CoR in a manner indistinguishable from that of WT Ski or SnoN (Fig. 3B). However, whereas WT Ski or SnoN recruited N-CoR to the Smad proteins (Fig. 3C,  lanes 3 and 6), the mS3S4 mutants failed to do so (lanes 2 and 5), most likely because of their inability to interact with the Smad proteins. Thus, the mS3S4 mutation did not disrupt the folding of Ski or SnoN. The lack of repressive activity of these mutants is due to their inability to be recruited to the TGF-␤responsive promoter element by the Smad proteins and a fail-

FIG. 2. Interaction of Ski or SnoN with the Smad proteins is required for repression of transcriptional activation.
A, interaction with the Smad proteins is required for the binding of Ski or SnoN to the SBE. Smad4 in complex with FLAG-tagged Ski or SnoN was isolated from transfected 293T cells by immunoprecipitation with anti-FLAG monoclonal antibody, followed by elution with the FLAG peptide (lanes 2-13). As a positive control, recombinant Smad4 was isolated from E. coli as a GST fusion protein (lane 1). These protein complexes were then incubated with 32 P-labeled SBE oligonucleotide. The resulting protein-DNA complexes were resolved by 4% nondenaturing PAGE. The proteins used in the electrophoresis mobility shift assay reactions were as follows: purified Smad4, 0.2 g; and WT Ski-or SnoN-bound Smad4, 0.2 g. Lane 1, GST-Smad4; lanes 2-4, FLAG-tagged SnoN-Smad4 complex purified from cotransfected 293T cells; lanes 5-7, FLAG-tagged mS3S4 mutant SnoN-Smad4 complex; lanes 8 -10, FLAGtagged Ski-Smad4 complex; lanes 11-13, FLAG-tagged mS3S4 mutant Ski-Smad4 complex. B-D, repression of transcription by the Ski and SnoN proteins. Hep3B cells were transfected with two different concentrations of ski together with 0.5 g of p3TP-lux (B) or with 2 g of WT or mutant snoN and 0.5 g of pLuc800 (C) or p3TP-lux (D). Luciferase activity was measured 48 h later. ure to recruit N-CoR and the associated transcription repressors to the Smad proteins. In addition, our previous structural studies indicated that the mS3S4 mutant cannot disrupt the active conformation of a heteromeric Smad complex, nor does it block binding of Smad3 to the coactivator p300/CBP (44). Taken together, these results indicate that interaction between Ski/SnoN and the Smad proteins is indeed necessary for the repression of TGF-␤ target gene expression.

Interaction of Ski and SnoN with the Smad Proteins Is Required to Block TGF-␤-induced Growth Inhibition-Ski and
SnoN are oncoproteins that, when overexpressed, induce oncogenic transformation of chicken and quail embryo fibroblasts (9,52,53). Additionally, high levels of Ski or SnoN have been detected in many human cancer cells (1, 10, 11). We speculated that the ability of Ski and SnoN to interact with the Smad proteins and to repress the growth inhibitory activity of TGF-␤ may contribute to the transforming activity of Ski and SnoN. In support of this hypothesis, overexpression of Ski or SnoN blocks TGF-␤-induced growth inhibition (17,18). To confirm that this activity indeed requires the interaction with the Smad proteins, the mutant forms of Ski or SnoN deficient in binding to the Smad proteins (mS2/3, mS4, and mS3S4) were introduced stably into Ba/F3 pro-B cells and examined for their ability to block TGF-␤-induced cell cycle arrest. We have shown previously that WT SnoN is rapidly degraded upon stimulation with TGF-␤1 and blocks TGF-␤1-induced growth inhibition only modestly (17, 18, 54 -56). We therefore used a truncated form of SnoN, SnoN-(1-366), which cannot be degraded by TGF-␤ and consequently is more potent than full-length SnoN in repression of TGF-␤ signaling (55). When expressed at similar levels, the mutant forms of Ski lacking one of the Smadbinding sites retained the ability to block TGF-␤-induced growth arrest (Fig. 4A) (data not shown), whereas mutants lacking both Smad-binding sites did not (Fig. 4B). Similarly, mS3S4 mutant SnoN lacking both Smad-binding sites exhibited a markedly impaired ability to block TGF-␤-induced cell cycle arrest (Fig. 4C). This indicates that interaction of Ski and SnoN with the Smad proteins is indeed required for antagonism of TGF-␤-induced growth inhibition.

Interaction of Ski and SnoN with the Smad Proteins Is Critical for Their Transforming Activity-Previous studies have indicated that overexpression of Ski or SnoN in CEFs induces
anchorage-independent growth as measured by the soft agar colony assay (9,52). To investigate whether this activity depends on the ability of Ski or SnoN to antagonize the growth inhibitory pathway of TGF-␤, FLAG-tagged WT or mutant Ski or SnoN was cloned into a replication-competent avian retroviral vector, and the constructs were transfected into primary CEFs. The extent of infection of the CEFs and the expression of these Ski and SnoN proteins were monitored by immunofluorescent staining (Fig. 5D) and by immunoblotting with anti-FLAG monoclonal antibody. 3 weeks after the initial transfection, ϳ70% of the cells were infected for each construct (data not shown). Both WT and mutant Ski or SnoN proteins were expressed at similar levels in the nucleus (Fig. 5, A and D) (data not shown). In growth inhibition assays, WT (but not mutant) Ski blocked TGF-␤-induced cell cycle arrest of CEFs (data not shown), similar to that observed in Ba/F3 cells (Fig.  4). In soft agar colony assays, whereas CEFs expressing WT Ski, SnoN, or SnoN-(1-366) grew readily in soft agar, those expressing the mutant Ski or SnoN proteins showed a greatly reduced ability to form soft agar colonies (Fig. 5, A-C), suggesting that the transforming activity of Ski and SnoN is dependent on their ability to interact with the Smad proteins. Interestingly, the truncation Ski-(24 -441) (human v-Ski equivalent) readily transformed CEFs even though it interacted only with Smad4, but not with the R-Smad proteins (data not shown). Mutation of the Smad4-binding site in this molecule markedly impaired its transforming activity (data not shown), suggesting that the Ski-Smad4 interaction plays a crucial role in the transforming activity of Ski. Taken together, our results indicate that the ability of Ski and SnoN to interact with the Smad proteins and to antagonize TGF-␤-induced growth arrest is responsible for their transforming activity.  1 and 4) or with WT (lanes 3 and 6) or mutant (lanes 2 and 5) SnoN or Ski and isolated by immunoprecipitation with anti-HA antibody. Endogenous N-CoR associated with HA-Smad4 was detected by Western blotting with anti-N-CoR antibody (upper panels). As controls, the anti-HA immunoprecipitates were blotted with anti-FLAG monoclonal antibody for the associated SnoN or Ski proteins (second panels) or with anti-HA antibody for Smad4 (third panels). Total cell lysates were blotted with anti-N-CoR antibody or with anti-FLAG monoclonal antibody to control for the expression of N-CoR and SnoN or Ski protein (fourth and lower panels).
been reported in pigmented avian melanocytes and bone marrow-derived multipotent progenitor cells (11,14). High levels of Ski or SnoN expression have been found in human melanoma; breast cancer; esophageal cancer; and carcinomas of the vulva, stomach, and lung (1,10,11,57). However, the mechanism by which Ski and SnoN induce transformation is unclear.
Recently, we and others have shown that Ski and SnoN can interact with the Smad proteins and repress their ability to activate TGF-␤ signaling (16 -21). To examine whether this is responsible for the transforming activity of Ski or SnoN, we first determined the amino acid residues in Ski or SnoN required for interaction with the Smad proteins. Ski and SnoN were found to bind to the R-Smad proteins (Smad2 and Smad3) and Co-Smad (Smad4) through different regions. The presence of two Smad-binding sites suggests that Ski or SnoN can either interact with one of the Smad proteins individually (35,54,55) or bind to a heterodimer of R-Smad and Smad4, depending on the absence or presence of ligand and the status of the signaling pathway. Binding of Ski or SnoN to R-Smad or Smad4 individually may result in recruitment of Smad-associated cellular proteins to Ski or SnoN, allowing cross-talk with other intracellular signaling pathways or modulation of the activity or expression of Ski or SnoN. For example, binding of SnoN to the R-Smad proteins independent of Smad4 results in recruitment of two ubiquitin ligases, Smurf2 and the anaphasepromoting complex, leading to the degradation of SnoN (54,55). On the other hand, binding of Ski or SnoN to a heteromeric Smad complex results in the disruption of such a functional complex since Ski and SnoN were found to compete with the R-Smad proteins for binding to a common region in Smad4 (44). Although the disrupted Smad complexes remain bound to Ski, these complexes are not in an active conformation to interact with transcriptional coactivators such as CBP to activate TGF-␤ target genes (44). Based on this model, binding of Ski or SnoN to one of the Smad molecules, either R-Smad or Smad4, is sufficient for the disruption of the Smad complex and subsequent repression of TGF-␤ signaling. Indeed, we found that mutation of one of the Smad-binding sites in Ski or SnoN did not affect the ability of Ski or SnoN to repress the transactivation activity of the Smad proteins significantly. Only when both binding sites were abolished was the repression of Smad function impaired. Consistent with this, although v-Ski lacks the R-Smad-binding site due to a truncation of the first 27 amino acid residues from c-Ski, it still represses Smad function and induces potent transformation of CEFs, probably because it can still interact with the Smad complex and inactivate it through binding to Smad4.
The residues in Ski that mediate Smad4 binding are well FIG. 5. Interaction of Ski with the Smad proteins is critical for the transforming activity of Ski and SnoN. FLAG-tagged WT or mutant Ski or SnoN was introduced into CEFs by transfection. 3 weeks after the initial transfection, a soft agar colony assay was performed as described under "Experimental Procedures." The soft agar plates were stained with 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide and scanned, and the Ski plates are shown in A. The expression levels of WT or mutant Ski in these cells were determined by Western blotting with anti-FLAG monoclonal antibody, shown on the right. Shown is B is a microscopic view of representative regions from the 3-(4,5-dimethylthiazol-2-yl)-2,5diphenyltetrazolium bromide-stained soft agar plates expressing various SnoN constructs. The number of soft agar colonies in a representative 2.5-cm 2 area was quantified and is summarized in C. In D, FLAG-tagged Ski and SnoN expressed in CEFs were visualized by immunofluorescent staining with anti-FLAG monoclonal antibody (left panels) to control for the localization of Ski or SnoN. 4,6-Diamidino-2-phenylindole (DAPI) staining was performed to show the nuclei (right panels). conserved in all Ski family members, including v-Ski, c-Ski, and c-SnoN and its isoforms (SnoN2, SnoI, and SnoA), and the region required for R-Smad binding is present in all but v-Ski. A careful comparison of the amino acid sequences within the two Smad-binding sites did not reveal any extensive homology between the two regions. These Smad-binding sites do not shown any obvious similarity to the PPNK-containing motif found to mediate Smad2-binding in FAST1 and in the Milk family of proteins or to the Smad-binding domains of SARA, type I TGF-␤ receptor, and Smurf2 (58 -62). Consistent with these observations, different domains or residues in Smad2 may mediate binding to these proteins. For example, a basic amino acid residue stretch in the L3 loop of Smad2 interacts with activated type I TGF-␤ receptor, whereas the PY motif in the linker region of Smad2 recognizes the WW domain in Smurf2 (54). Thus, Smad2 and Smad3 may recognize multiple sequence motifs through different amino acid residues. Interestingly, the R-Smad-binding site is not 100% conserved between Ski and SnoN. This partial difference may contribute to the different affinity of Ski and SnoN for the bone morphogenic protein Smad proteins (17,18). A thorough understanding of the Ski-Smad and SnoN-Smad protein interactions will require a detailed analysis of the three-dimensional crystal structures of these complexes.
Because Smad proteins are important tumor suppressors, we hypothesized that the ability of Ski and SnoN to inactivate Smad function may be responsible for their transforming activity. In this study, we employed CEFs as a model system to investigate the transforming activity of Ski and SnoN because Ski was originally identified in CEFs as a viral oncogene by virtue of its ability to transform CEFs. Unlike many mammalian fibroblast cell lines that proliferate in response to TGF-␤, CEFs, like many epithelial cells, undergo growth arrest in the presence of TGF-␤ (data not shown). Thus, overexpression of Ski or SnoN blocks TGF-␤-induced growth arrest (data not shown), and this may be responsible for the transformation of CEFs. Indeed, we have shown here that the interaction of Ski and SnoN with Smad proteins is required for the antagonism of TGF-␤ signaling and, more importantly, for their transforming activity. Mutant Ski and SnoN lacking the Smad-binding sites are defective in repression of TGF-␤-induced cell cycle arrest and fail to induce anchorage-independent growth of CEFs. Extrapolating from this, the high level expression of Ski or SnoN in some human cancer cells may be responsible for the resistance of these cancer cells to TGF-␤-induced growth arrest, a key step in the malignant progression of mammalian tumor cells.
TGF-␤ signaling pathways are considered to be both tumor suppressor pathways and promoters of tumor progression and invasion. TGF-␤1, TGF-␤ receptors, and the Smad proteins are expressed in virtually all tissues and cell types. Activation of TGF-␤1 and TGF-␤ signaling in vivo can be regulated at both the levels of extracellular ligand activation and intracellular signal transduction. These highly regulated processes regulate the differentiation or proliferation state of a given cell type or tissue. In normal cells and at early stages of tumorigenesis, activation of TGF-␤ and Smad proteins inhibits cell growth. Perturbation of this growth inhibitory pathway by activation or overexpression of oncogenes such as ski and snoN results in a diminished growth inhibitory response, leading to rapid tumor growth and clonal expansion and permitting the accumulation of additional mutations and tumor progression. As tumor cells lose their ability to be inhibited by TGF-␤ and progress to a more malignant stage, stimulation by TGF-␤ causes these cells to undergo epithelial-to-mesenchymal transdifferentiation, leading to increased tumor metastasis and invasion. Thus, the activity of Ski and SnoN may function to promote the switch of the tumor cell responses to TGF-␤ from growth inhibition to accelerated malignant progression.