Ski-interacting Protein Interacts with Smad Proteins to Augment Transforming Growth Factor-β-dependent Transcription*

Transforming growth factor-β (TGF-β) signaling requires the action of Smad proteins in association with other DNA-binding factors and coactivator and corepressor proteins to modulate target gene transcription. Smad2 and Smad3 both associate with the c-Ski and Sno oncoproteins to repress transcription of Smad target genes via recruitment of a nuclear corepressor complex. Ski-interacting protein (SKIP), a nuclear hormone receptor coactivator, was examined as a possible modulator of transcriptional regulation of the TGF-β-responsive promoter from the plasminogen activator inhibitor gene-1. SKIP augmented TGF-β-dependent transactivation in contrast to Ski/Sno-dependent repression of this reporter. SKIP interacted with Smad2 and Smad3 proteins in vivo in yeast and in mammalian cells through a region of SKIP between amino acids 201–333. In vitro, deletion of the Mad homology domain 2 (MH2) domain of Smad3 abrogated SKIP binding, like Ski/Sno, but the MH2 domain of Smad3 alone was not sufficient for protein-protein interaction. Overexpression of SKIP partially overcame Ski/Sno-dependent repression, whereas Ski/Sno overexpression attenuated SKIP augmentation of TGF-β-dependent transcription. Our results suggest a potential mechanism for transcriptional control of TGF-β signaling that involves the opposing and competitive actions of SKIP and Smad MH2-interacting factors, such as Ski and/or Sno. Thus, SKIP appears to modulate both TGF-β and nuclear hormone receptor signaling pathways.

Transforming growth factor-␤ (TGF-␤) signaling requires the action of Smad proteins in association with other DNA-binding factors and coactivator and corepressor proteins to modulate target gene transcription. Smad2 and Smad3 both associate with the c-Ski and Sno oncoproteins to repress transcription of Smad target genes via recruitment of a nuclear corepressor complex. Ski-interacting protein (SKIP), a nuclear hormone receptor coactivator, was examined as a possible modulator of transcriptional regulation of the TGF-␤-responsive promoter from the plasminogen activator inhibitor gene-1. SKIP augmented TGF-␤-dependent transactivation in contrast to Ski/Sno-dependent repression of this reporter. SKIP interacted with Smad2 and Smad3 proteins in vivo in yeast and in mammalian cells through a region of SKIP between amino acids 201-333. In vitro, deletion of the Mad homology domain 2 (MH2) domain of Smad3 abrogated SKIP binding, like Ski/Sno, but the MH2 domain of Smad3 alone was not sufficient for protein-protein interaction. Overexpression of SKIP partially overcame Ski/Sno-dependent repression, whereas Ski/Sno overexpression attenuated SKIP augmentation of TGF-␤-dependent transcription. Our results suggest a potential mechanism for transcriptional control of TGF-␤ signaling that involves the opposing and competitive actions of SKIP and Smad MH2-interacting factors, such as Ski and/or Sno. Thus, SKIP appears to modulate both TGF-␤ and nuclear hormone receptor signaling pathways.
Recently it has been shown that Smad proteins also interact with other nuclear factors such as c-Ski and the Ski-related novel (Sno) protein and nuclear hormone receptors, including the vitamin D receptor (VDR) to modulate TGF-␤ signaling (5)(6)(7)(8)(9)(10). Ski and Sno are involved in oncogenic transformation and enhancement of muscle differentiation by blocking TGF-␤ signaling (11)(12)(13)(14). The mechanism of Ski/Sno repression of TGF-␤ signaling appears to involve an interaction with a complex consisting of the nuclear corepressor (N-CoR) and a histone deacetylase enzyme (15,16). N-CoR, and its related corepressor silencing mediator for retinoic acid and thyroid receptors (SMRT), interact with a wide variety of other nuclear factors to mediate transcriptional repression (17)(18)(19). Interestingly, the Ski-interacting protein (SKIP) was initially identified in a two hybrid screen using v-Ski as a bait and was later independently identified as a VDR-and CBF1-interacting factor (20 -22). Thus, the recent observation that SKIP modulates CBF1 and Notch-dependent signaling suggests that SKIP may play a role in the regulation of a number of different and distinct cellular signaling pathways (23).
As Ski and Sno can modulate TGF-␤-dependent signaling, it was of interest to determine whether SKIP could also modulate the TGF-␤-signaling pathway through interaction with the Smad proteins. In this study, in contrast to Ski-and Snomediated repression, SKIP augmented TGF-␤-dependent transcription. A region of SKIP, aa 201-333, appeared to be required for the Smad interaction. SKIP interacted in vitro and in vivo with Smad3 and partially counteracted Ski-and Snodependent repression, while Ski/Sno attenuated SKIP transactivation of TGF-␤ signaling. These results suggest that SKIP may play an opposite role to Ski and Sno in the control of TGF-␤-dependent transcription.

MATERIALS AND METHODS
Plasmid Constructs-SKIP wild-type cDNA was PCR cloned with the forward primer (5Ј-GGG AAT TCC CGG GGT CTA GAA CCA CCA TGG CGC TCA CCA GCT TTT TA-3Ј) and reverse primer (5Ј-GCG GGA TCC CTA TTC CTT CCT CCT CTT-3Ј). The PCR product was ligated into pGEM-T Easy plasmid (Promega, Madison, WI) from which an EcoRI/ BamH1 insert was excised and subcloned into a modified GAL4AD pACTII plasmid (pACTIIb) and the vector pSG5 (Stratagene, La Jolla, CA). The pACTIIb plasmid was created by replacing the BglII polylinker fragment of pACTII (CLONTECH, Palo Alto, CA) with the double-stranded oligonucleotide: 5Ј-GAT CTG TGA ATT CCC GGG GAT CCG TCG ACC TA-3Ј. The GAL4 DBD wild-type Smad-pBridge yeast two hybrid constructs were made by EcoRI/XhoI digestion of Smad2-, Smad3-, and Smad4-pcDNA3 plasmids (3) and MH2 (aa 400 -425 of hSmad3)-pBridge by PCR and subcloning of Smad cDNAs into the EcoRI/SalI sites of pBridge (CLONTECH). The GST-MH2 construct was made by cloning cDNA of PCR product into the EcoRI/XhoI sites of modified pGEX-4T2 plasmid (Amersham Pharmacia Biotech). The SKIP deletion constructs (aa 1-200, aa 1-333, aa 201-536, and aa 334 -536) were made by PCR and cloned into the EcoRI/BglII site of pACTIIb and the XbaI/BamH1 site of pCGN. The Sno cDNA was amplified by PCR using the forward primer 5Ј-GCA ATC TAG AGA AAG CCC ACA AGC AAA TTT CCC-3Ј and reverse primer 5Ј-GCA AGG ATC CCT ATT TTC CAT TTC CAT TTT TG-3Ј and the PCR product ligated into the XbaI/BamHI site of pCGN. The GST-SKIP construct was made by PCR and cloned into the EcoRI and SalI sites of pGEX-KG (24). All PCR primer sequences not listed are available on request. All constructs were sequenced by automated fluorescent sequencing and confirmed to be in frame and correct. The wild-type SKIP-pCGN, c-Ski-pMT2, GST-Smad3, Smad2-, Smad3-, and Smad4-pcDNA3 constructs and the 3TP-Lux reporter have been described previously (3,20,(25)(26)(27).
Yeast Two-hybrid Analysis-Yeast transformation was performed using a lithium acetate transformation kit (BIO101, Vista, CA). Wildtype SKIP-pACTII plasmid encoding the SKIP-GAL4-AD fusion protein was transformed into the Y187 yeast strain and Smad2-, Smad3-, Smad4-, and MH2-pBridge encoding the Smad-GAL4 DBD fusion proteins were transformed into the opposite yeast mating strain, CG1945. To co-express the two different fusion proteins, yeast matings were performed (CLONTECH Yeast Handbook, PT3024-1). Yeast ligand experiments were performed as described previously (28). ␤-Galactosidase activity in protein lysates was measured with the Tropix Galactolight chemiluminescence assay (PerkinElmer Life Sciences) using the Berthold LB953 luminometer (Berthold, Bad Wildbad, Germany) and expressed as relative light units. All results are shown as mean Ϯ S.E. of at least three different yeast colonies, from at least two experiments, and corrected for protein concentration (Bio-Rad protein assay).
Cell Culture and Reporter Assays-COS-1 African green monkey kidney cells were grown in Dulbecco's modified Eagle's medium with 5% fetal calf serum at 37 o in 5% CO 2 . Cells were plated the day before transfection in 24-well plates at a cell density of 2 ϫ 10 4 cells/well. Transfections was performed with FuGENE-6 transfection reagent (Roche Molecular Biochemicals) as per the manufacturer's instructions using 1.5 l of FuGENE with 0.75 g of total DNA/well. Transfected cells were left in FuGENE transfection reagent for 16 -20 h, then treated with TGF-␤ (Sigma) or vehicle (4 mM HCl and 1 mg/ml bovine serum albumin) in 2% charcoal-stripped medium for 16 -24 h. The medium was then removed, and cells were lysed with 2ϫ Promega lysis buffer. Luciferase assays were performed in triplicate with the firefly luciferase assay kit (Promega) and measured with a luminometer (Berthold).
Glutathione-Sepharose Binding Assays-Expression of appropriately sized GST-SKIP and GST-Smad3 wild-type or mutant fusion proteins was confirmed by SDS-PAGE. GST-binding assays were performed in triplicate with equal amounts of 35 S-labeled SKIP, VDR, or luciferase as a negative control (28). Bound proteins were resolved on 10% SDS-PAGE gels and subjected to autoradiography. In vitro translation and transcription was performed according to the manufacturer's instructions (Promega) with [ 35 S]methionine (Amersham Pharmacia Biotech).
Far Western and Immunoblot Analysis-Far Western analysis and preparation of nuclear extracts were essentially as described previously (5,28). COS1 nuclear extracts overexpressing Smad3 were run on 10% SDS-PAGE and electroblotted onto a polyvinylidene difluoride membrane (Millipore, Bedford, MA). Proteins were denatured with 6 M guanidine hydrochloride and renatured by the stepwise dilution of guanidine hydrochloride. The Smad3 membrane was then blocked and hybridized overnight at 4°C with 20 g of COS1 nuclear extracts containing HA-SKIP. The filter was rinsed three times in HYB (20 mM Hepes-KOH, pH 7.4, 75 mM KCl, 0.1 mM EDTA, 2.5 mM MgCl 2 , 1% nonfat milk, 1 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol) and then probed with an anti-HA antibody (Roche Molecular Biochemicals), which detected HA-SKIP, followed by probing with a antimouse-HRP secondary antibody (Santa Cruz) prior to ECL chemiluminescent detection (Amersham Pharmacia Biotech) and autoradiography.

RESULTS
SKIP Augments TGF-␤-dependent Transcription-As Ski and Sno interact directly with Smad proteins (Smad2 and Smad3) to repress TGF-␤-dependent transcription, the effects of SKIP on the 3TP-lux TGF-␤-responsive reporter construct (27) were tested (Fig. 1). In COS1 cells this reporter responded to TGF-␤ with a 4-fold increase in reporter activity, consistent with these cells expressing endogenous Smad proteins (30). Smad3 alone, or Smad2 and Smad4 together (but neither alone) augmented both basal (2-fold) and TGF-␤ responses (10-fold) of this reporter activity. Smad3 co-transfection with Smad4 led to a 6-fold increase in basal and a 30-fold increase in ligand-dependent reporter activity. This augmentation was similar to that of SKIP alone on ligand-dependent reporter activity (Fig. 1). An interaction between SKIP and Smads was suggested in co-transfection studies, with the -fold increase of basal activity progressively increasing when SKIP was cotransfected with Smad2 (8-fold), Smad2 and 4 (20-fold), Smad3 (53-fold), and Smad3 and Smad4 (164-fold). The comparable increases in TGF-␤ induced activity were 39-, 116-, 96-, and 323-fold, respectively. These data are consistent with a functional interaction primarily occurring between SKIP and Smad3, with or without exogenous Smad4.
Mapping of SKIP-Smad Interaction Domains in Yeast and Mammalian Cells-SKIP interaction with Smad proteins was investigated by yeast two-hybrid interaction analysis. SKIP interacted with both Smad2 and Smad3 (Fig. 1B). Smad4 induced a high level of reporter activity, which was unaltered by co-expression of SKIP. However, Smad4, as expected, interacted strongly with v-Ski-GAL4-AD in yeast (data not shown). Domains of SKIP required for Smad interaction were examined using deletion constructs of SKIP (Fig. 1B). The C-terminally deleted aa 1-333 and N-terminally deleted aa 201-536 SKIP mutants had interaction with Smad2 comparable to that of wild-type SKIP. The interaction between these two mutants and Smad3 were about 50% and 25% of wild-type SKIP, respectively. No interaction of the SKIP N-terminal (aa 1-200) or C-terminal (aa 334 -536) domain with Smad2 or Smad3 was observed. Thus, these results suggest that the region of SKIP between aa 201 and 333 interacts with Smad2 and Smad3.
The same SKIP deletion constructs were tested with the 3TP-lux reporter in the COS1 mammalian cell line (Fig. 1C). Co-expression of wild-type SKIP (aa 1-536) with Smad3 caused a synergistic 3-fold increase in reporter activity above SKIP or Smad3 alone. The N-terminal domain of SKIP (aa 1-200) had no effect on reporter activity, while the other SKIP constructs had comparable transactivation to that of wild-type SKIP. Western blot analysis of these deletion constructs showed comparable expression with wild-type, except for the aa 1-200 construct, which, despite its lack of transactivation, was 2-3 times more highly expressed (data not shown). Thus, these transfection data were consistent with the yeast interaction data and suggest that expression of the aa 201-333 region of SKIP with Smad3 is sufficient for near maximal transactivation of the 3TP-lux reporter. Surprisingly, the aa 334 -536 SKIP construct was able to activate the 3TP-lux reporter with Smad3, even though no interaction was observed with Smad3 in yeast. This suggests that an additional C-terminal domain may also be transcriptionally functional and possibly recruits other Smad-interacting co-factors present in mammalian cells, but not yeast.
SKIP Interaction with Smad2 and Smad3 in Vitro-The potential direct physical interaction between the Smad proteins and SKIP was explored using a GST "pull-down" assay. GST-SKIP bound both Smad2 and Smad3 ( Fig. 2A). In comparison there was minimal, if any, binding of Smad2 or Smad3 to GST-0 and no binding of luciferase to GST-SKIP.
To determine which domains of Smad3 may be involved in SKIP interaction, a GST-Smad3 binding assay was performed with 35 S labeled in vitro translated SKIP (Fig. 2B). GST-wild-type Smad3 bound SKIP and the positive control VDR (9). Deletion of the MH1 domain of Smad3 (aa 199 -427) had no effect on SKIP binding, but, as expected, VDR binding was abolished. Both SKIP and VDR binding was lost when both the MH1 and MH2 domains of Smad3 were deleted (GST-Smad3 aa 199 -405). However, no binding of SKIP was observed to a GST-MH2 construct, which expressed only the last 26 aa of hSmad3. This result was further supported by a lack of interaction between SKIP-GAL4-AD and a MH2-GAL4-DBD construct containing the same C-terminal 26 aa of hSmad3 in vivo in yeast (data not shown). These results indicate that, although deletion of the C-terminal MH2 domain abrogates SKIP binding, the MH2 domain alone is not sufficient for SKIP interaction.
To further support the existence of a direct protein-protein interaction in vitro, a Far Western assay was performed using mammalian cell nuclear extracts overexpressing HA-SKIP (upper panel) or Smad3 (middle panel) (Fig. 2C). In the Far Western analysis (lower panel), Smad3 detected by Western analy- sis co-localized with HA-SKIP detected by using an anti-HA antibody, but not with the negative empty vector control extracts. These results together with the GST-binding studies thus strongly support the existence of a protein-protein interaction in vitro between SKIP and Smad3.
Ski and Sno Competitively Inhibit SKIP-dependent Activation-The Smad3 transcriptional repressors, Ski and its related protein, Sno, are known to bind to the MH2 domain of Smad3 (26). Since SKIP also interacts with Ski and Sno, we tested whether SKIP modulates Ski/Sno repression of Smad3dependent transcription. As shown above, SKIP increased basal and TGF-␤-dependent transactivation, particularly in the presence of Smad3 (Fig. 3). Both Ski and Sno attenuated this SKIP-dependent transactivation by about 80% and 40%, respectively (Fig. 3). SKIP transactivation, either alone or with Smad3, was repressed in a dose-dependent manner by cotransfection with Ski. Sno had a similar but weaker effect (Fig.  4). These data suggest that SKIP and Ski/Sno may act as counteracting regulators of the TGF-␤ transcriptional response.
As Ski/Sno interact with both Smad3 and SKIP, one alternative possibility other than a competitive interaction between these proteins is that they form a ternary complex. To address this question, a gel shift analysis was performed (Fig. 5). Using the PE-2 probe from the PAI-1 promoter, which binds a Smad3/4 heterodimer (29) (Fig. 5, lane 2), we showed that, with addition of increasing amounts of SKIP nuclear extracts, there was augmentation of binding of a higher molecular weight complex, which presumably contained SKIP and Smad3/4 (lanes 3-6). The addition of Sno nuclear extracts also led to increased Smad3/4 binding with a similar mobility shift (lanes 9 and 10 and lanes 12 and 13). This complex was specific as it was abrogated by addition of cold probe (lane 11). However, in the presence of both SKIP and increasing Sno, although we observed increased intensity of the upper complex, no further supershift and hence no ternary complex was observed (lanes 7-10). Similar results were obtained using Ski-overexpressing nuclear extracts (data not shown). DISCUSSION The Ski and Sno oncoproteins have been shown to negatively modulate TGF-␤ signaling through an interaction with a N-CoR repressor complex (15). As SKIP, a nuclear hormone receptor-interacting cofactor, also associates with both Ski and Sno, this study was undertaken to determine the potential role of SKIP in TGF-␤ signaling. In these studies SKIP augmented TGF-␤Ϫdependent transcription and exhibited a direct interaction with Smad proteins. This SKIP-Smad interaction was apparent both in vitro and in vivo, as demonstrated by GST pull-down assays, Far Western analysis, and yeast two hybrid protein-protein studies. The region between aa 201 and 333 within SKIP appeared to act as the Smad-interacting domain, while, SKIP, like Ski and Sno, interacted with the MH2 domain of Smad3. Moreover, Ski and Sno attenuated SKIP transactivation, while SKIP partially counteracted Ski-and Sno-mediated transcriptional repression. The C-terminal MH2 domain of Smad2 and Smad3 has been reported to be a key region involved in multiple protein-protein interactions, including those with the coregulators CBP/p300 and the Smad repressors Ski and Sno (1). The N-terminal MH1 domain of the Smads confers only low affinity DNA binding to a consensus SBE (1). Although natural TGF-␤-responsive promoters contain functional clusters of SBEs, other DNA-binding factors, such as FAST-1, TFE3, and AP-1, as well as non-DNA binding factors through protein-protein interaction with the C-terminal MH-2 domain, are involved in determining the specificity and direction of Smad target gene action (29,31,32). As such, SKIP appears to play a role in augmentation of TGF-␤-specific Smad transcriptional activity via an interaction with the MH2 domain of Smad3. Our data also suggest, as SKIP was unable to interact with an isolated MH2 domain (last 26 aa) of Smad3 in GST binding assays and yeast two-hybrid studies, that other regions within the Smad proteins possibly within the context of the whole Smad protein may also modulate Smad-SKIP interaction.
Although SKIP was able to interact with Smad2 and Smad3 in yeast, SKIP co-transfection with Smad3 with or without Smad4, led to the greatest increases in reporter activity in mammalian cells, presumably because the 3TP-lux reporter is Smad3-selective (27). Thus, as SKIP interacted with Smad2 in vivo and in vitro, it is also possible that SKIP in mammalian cells may be able to modulate TGF-␤ signaling through Smad2 in certain situations (30). Furthermore, in the transient transfections we observed that Smad3 augmented basal reporter activity, as previously described with this promoter (5), but this activity was further increased by SKIP. Further studies will be required to address the specific reasons for this effect of SKIP.
The deletional analysis of SKIP in yeast and mammalian cells suggests that the aa 201-333 region of SKIP is required for Smad interactions in vivo. However, some functional differences were observed between yeast and mammalian cells. Specifically, although the C-terminal SKIP construct (aa 334 -536) did not interact with Smad2 or Smad3 in yeast, its transactivation activity in mammalian cells was comparable to wildtype SKIP. A C-terminal transactivation domain of SKIP that functions in mammalian cells, distinct from the Smad interaction domain, is consistent with the domain C-terminal to aa 437 of murine SKIP (NcoA-62) being involved in vitamin D-dependent transactivation (21).
As SKIP interacts with Ski/Sno and Smad3 and, in turn, Ski/Sno interact with Smad3, to address the possibility that these proteins form a ternary complex, we performed a gel shift analysis using the PE2 probe from the PAI-1 promoter, as used in the transient transfections. The EMSA clearly showed that both SKIP and Ski/Sno alone formed a slightly higher migrating complex with Smad3/4. However, we did not observe the formation of a ternary complex in the presence of all three proteins. Thus, these data are consistent with the transient transfection results, which suggests competition occurring between SKIP and Ski/Sno for Smad3 transactivation, but cannot exclude the presence of a ternary complex forming between these proteins.
In this study SKIP acted as a coactivator of TGF-␤-dependent transcription. SKIP similarly acts as a coactivator of nuclear hormone receptor-dependent transcription, but also as a repressor of Notch signaling through its interaction with SMRT and associated histone deacetylase enzyme proteins (21,23). These divergent effects of SKIP may depend on interaction of SKIP with other, possibly cell-specific nuclear factors. For example, SKIP converts CBF1 from a transcriptional repressor to activator through switching its interaction between the corepressor SMRT and Notch 1C (23). In our study SKIP and Ski/Sno modulated each other's opposing transcriptional activities, raising the intriguing possibility that the relative cellular expression of SKIP versus Ski or Sno may play a regulatory role on TGF-␤-dependent transcription and hence its effects on cell growth and differentiation. Interestingly, the Smad-interacting domain of SKIP (aa 201-333) appears to be also involved in interaction with Ski and Sno. 2 These results and those showing that SKIP, like Ski/Sno, interacted with the MH2 domain of Smad3 suggest that the opposing transcriptional effects of SKIP and Ski/Sno may involve competition for Smad3 binding between SKIP and c-Ski/Sno, and/or other Smad3 MH2-interacting factors, such as with CBP/p300 (7,33,34). Thus, the modulatory effects of SKIP through the MH2 domain potentially increase the complexity and diversity of Smad-dependent transcriptional effects. Furthermore, as SKIP and Ski/Sno interact with each other and also with the related corepressors N-CoR/SMRT, an additional mechanism could involve SKIPmediated derepression (1,15,20,23). This may possibly occur via SKIP sequestration of corepressors such as SMRT or N-CoR from the Ski/Sno repressor complex, a mechanism similar to that suggested for Hoxc-8 and Smad1 (35). Whatever the molecular mechanism of SKIP action, it is nevertheless clear that SKIP plays a role in modulation of this important cellular and signaling pathway.
In summary, our results support a model in which SKIP positively modulates TGF-␤-dependent-transcription and potentially competes with other MH2-interacting factors, such as c-Ski and Sno, to determine the transcriptional outcome of a TGF-␤-responsive target gene. This suggests a potential role for SKIP in the regulation of TGF-␤ effects on cell growth and differentiation.