TAK1 MAPK Kinase Kinase Mediates Transforming Growth Factor-β Signaling by Targeting SnoN Oncoprotein for Degradation*

Transforming growth factor-β (TGF-β) regulates a variety of physiologic processes through essential intracellular mediators Smads. The SnoN oncoprotein is an inhibitor of TGF-β signaling. SnoN recruits transcriptional repressor complex to block Smad-dependent transcriptional activation of TGF-β-responsive genes. Following TGF-β stimulation, SnoN is rapidly degraded, thereby allowing the activation of TGF-β target genes. Here, we report the role of TAK1 as a SnoN protein kinase. TAK1 interacted with and phosphorylated SnoN, and this phosphorylation regulated the stability of SnoN. Inactivation of TAK1 prevented TGF-β-induced SnoN degradation and impaired induction of the TGF-β-responsive genes. These data suggest that TAK1 modulates TGF-β-dependent cellular responses by targeting SnoN for degradation.

Transforming growth factor-␤ (TGF-␤) 2 is a multifunctional cytokine involved in the regulation of proliferation, differentiation, migration, and survival of many different cell types (1,2). TGF-␤ ligand binds to and activates Ser/Thr kinase receptors (3). This leads to the phosphorylation and activation of the receptor-regulated Smad family (R-Smad), Smad2 and Smad3 (4). Phosphorylated R-Smad forms a functional complex with the co-mediator Smad (Co-Smad), Smad4, and this complex accumulates in the nucleus and modulates expression of the TGF-␤-responsive genes such as plasminogen activator inhibitor type-1 (PAI-1) (1,5,6). The nuclear Smads complex is maintained in an inactive state via its association with Ski family oncoproteins, Ski and SnoN (7,8). By binding to Smads, Ski and SnoN recruit transcriptional repressor complexes such as N-CoR/SMRT and mSin3A to TGF-␤ target promoters and thereby repress transcription of TGF-␤-responsive genes (7,9). Upon TGF-␤ stimulation, SnoN is immediately down-regulated via the ubiquitin-proteasome pathway induced by anaphase-promoting complex (APC) or Smurf2 E3 ubiquitinprotein isopeptide ligases (10 -12). Degradation of SnoN initially allows the Smad heteromeric complex to activate TGF-␤ target genes (13). However, longer TGF-␤ treatment leads to higher expression via transcriptional activation of the SnoN gene (14). This functions as a negative feedback circuit to limit the effects of TGF-␤. Importantly, overexpression of SnoN results in the loss of TGF-␤-induced growth arrest of the cells, suggesting a potential mechanism for SnoN-mediated oncogenesis (8,14).
TGF-␤-activated kinase 1 (TAK1) is a member of mitogenactivated protein kinase kinase kinase (MAPKKK) and functions as a signaling intermediate in several intracellular signaling pathways including the TGF-␤ and interleukin-1 pathways (15)(16)(17)(18). TAK1 is catalytically activated by TGF-␤ stimulation (15) and plays an essential role in TGF-␤-induced p38 activation (17). TAK1 has also been implicated in several TGF-␤induced biological processes including apoptosis (19) and vascular development (20). However little is known about how TAK1 mediates TGF-␤ signaling. In this study, we found that TAK1 interacts with SnoN and targets it for degradation. The TAK1 regulation of SnoN may participate in TGF-␤-induced cellular responses.
Cell Culture, Transfection, and Virus Infection-293 cells, HaCaT cells, and HeLa S3 cells were cultured in Dulbecco's modified Eagle's medium plus 10% fetal calf serum or bovine growth serum (HyClone). Transfection of 293 cells was carried out using the calcium phosphate precipitation method. Stable transfections of HaCaT cells and HeLa S3 cells were carried out using TransFast TM (Promega). The retrovirus for expression of HA-SnoN full-length and HA-SonN mutant full-length were generated and infected into HaCaT cells according to the manufacturer's instruction. Stable cell line selection was achieved using G418, hygromycin B, or puromycin.
Yeast Two-hybrid Screening-Plasmid pGBD-C-TAK1(K63W) was used as bait to screen a mouse B cell library (in pGAD) (21). The bait plasmid and the library cDNAs were co-transformed into the yeast strain PJ69-4A using the lithium acetate method. Yeast cells were plated on selective medium plates and allowed to grow at 30°C. Positive colonies were then restreaked on selective medium plates. Plasmid DNA was rescued from positive colonies that grew on selective medium plates and subject to further sequence analysis.
Immunoprecipitation and Immunoblotting-Whole cell extracts were prepared in lysis buffer (20 mM HEPES, pH 7.4, 150 mM NaCl, 12.5 mM ␤-glycerophosphate, 1.5 mM MgCl 2 , 2 mM EGTA, 10 mM NaF, 2 mM dithiothreitol, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 100 units/ml aprotinin, 0.5% Triton X-100). Proteins from cell lysates were immunoprecipitated with 1 g of various antibodies and 15 l of protein G-Sepharose (Amersham Biosciences). The immune com-plexes were washed three times with wash buffer containing 20 mM HEPES (pH 7.4), 500 mM NaCl, and 10 mM MgCl 2 and once with rinse buffer containing 20 mM HEPES (pH 7.4), 150 mM NaCl, and 10 mM MgCl 2 and suspended in 30 l of rinse buffer. For immunoblotting, the immunoprecipitates or cell lysates were resolved on SDS-PAGE and transferred to Hybond-P membranes (Amersham Biosciences). The membranes were immunoblotted with various antibodies, and bound antibodies were visualized with horseradish peroxidase-conjugated antibodies against mouse or rabbit IgG using the ECL Western blotting system (Amersham Biosciences).
Cellular Fractionation-To isolate the nuclear and the cytoplasm fractions, cells in 10-cm dishes were treated with TGF-␤ (5 ng/ml) and then lysed with 500 l of hypotonic buffer A (50 mM HEPES (pH 7.4), 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 100 units/ml aprotinin) containing 0.1% Nonidet P-40 and homogenized in a Dounce homogenizer (30 strokes). Lysates were then centrifuged at 4,000 ϫ g for 4 min. The supernatant was centrifuged at 12,500 ϫ g for 4 min to obtain the cytosolic fraction. This pellet was resuspended in hypotonic buffer B (buffer A containing 1.7 M sucrose) and then centrifuged at 15,000 ϫ g for 30 min. The pellet (nuclear fraction) was resuspended in 0.5% Triton X-100 lysis buffer and sonicated. All steps were performed on ice or at 4°C. Protein concentrations were determined using the micro BCA protein assay kit (Pierce). The purity of the cytosolic and nuclear fractions was assessed by immunoblotting of IB (a cytosolic marker) and lamin B (a nuclear marker).

RESULTS AND DISCUSSION
To study the role of TAK1 in TGF-␤ signaling, we screened for TAK1-binding proteins using the yeast two-hybrid system. A kinase-inactive mutant of TAK1, TAK1(K63W), was used as the bait to screen a mouse B cell cDNA library. From a total of 2 ϫ 10 6 transformants, 28 clones were identified as potential interactors. Sequence analysis revealed that one of the positive clones encoded SnoN2 (Fig. 1A). SnoN undergoes alternative splicing, creating four splicing isoform, SnoN, SnoN2, SnoI, and SnoA (7,24). The N terminus of SnoN, from 1 to 366 amino acids, is identical among the four isoforms. The SnoN2 sequence is completely identical with that of SnoN except for the C-terminal 46 amino acids (Fig. 1B). As SnoN2 is the less abundantly expressed isoform in human cells (24), we focused on the interaction between TAK1 and SnoN. Interaction of TAK1 with SnoN was further confirmed in mammalian cell by co-immunoprecipitation assays. HA epitope-tagged TAK1 and FLAG-tagged SnoN were transiently co-expressed in 293 human embryonic kidney cells. The cell extracts were immunoprecipitated with anti-FLAG antibody and followed by immunoblotting analysis (Fig. 1C). When FLAG-SnoN was immunoprecipitated, both HA-TAK1 and HA-TAK1(K63W) were co-immunoprecipitated.
To establish the connection between TAK1 and SnoN, we next determined the subcellular localizations of TAK1 and SnoN. We performed biochemical fractionation with human keratinocyte HaCaT cells. TGF-␤-stimulated and unstimulated HaCaT cells were fractionated into the nuclear and the cytosolic extracts. Endogenous SnoN was localized only in the nucleus but was degraded upon TGF-␤ stimulation (Fig. 2, left panel). Endogenous TAK1 was localized primarily in the cytoplasm but also was detected in the nuclear fractions. Upon TGF-␤ stimulation, the amount of nuclear TAK1 was increased. The fractions were reasonably pure, as determined by the presence of IB only in the cytosolic fraction and not in the nuclear fraction. Conversely, a nuclear protein, lamin B, was detected only in the nuclear fraction but not the cytosolic fraction (supplemental Fig. S1). These data raised the possibility that TAK1 is co-localized with SnoN in the nucleus. To examine the interaction between TAK1 and SnoN in the nucleus, endogenous TAK1 was immunoprecipitated following fractionation (Fig. 2, right  panel). SnoN was found to be associated with TAK1 in the nuclear fraction independently of TGF-␤ stimulation, but no association could be detected in the cytosolic fraction. Thus, TGF-␤ stimulation induces the TAK1 accumulation in the nucleus, and this nuclear TAK1 interacts with SnoN.
The observed association between TAK1 and SnoN suggested that TAK1 is involved in the TGF-␤-dependent degradation of SnoN. To test the possibility, we employed siRNA to reduce the levels of endogenous TAK1 (25). We generated two independent HeLa S3 cell lines stably expressing TAK1 siRNA. Expression of TAK1 in each clone was determined by immunoblotting. Expression of TAK1 siRNA greatly reduced the amount of endogenous TAK1 but not affect ␤-catenin (Fig. 3A). Both TAK1 knockdown cells and parent cells were then treated with TGF-␤ and subjected to biochemical fractionation. The nuclear fraction was further subjected to immunoblotting with anti-SnoN, anti-TAK1, and anti-Smad2/3. As shown in Fig. 3B, TGF-␤ induced the degradation of SnoN in HeLa S3 cells, but degradation was impaired in TAK1 knockdown cells. In contrast, TGF-␤-induced nuclear accumulation of Smad2/3 was observed to be normal in cells expressing TAK1 siRNA.
To assess the influence of TAK1 knockdown on the TGF-␤dependent biological events, we examined expression of a TGF-␤-responsive gene, PAI-1. PAI-1 participates in wound healing processes. The mRNA levels of PAI-1 were increased at around 30 -60 min, which occurred subsequent to SnoN degradation and Smad accumulation. The accumulation of PAI-1 mRNA in response to TGF-␤ was impaired in the TAK1 siRNA-expressing cells (Fig. 3C). These results suggest that TAK1 is involved in TGF-␤-dependent biological processes.
SnoN is ubiquitinated and degraded upon TGF-␤ stimulation (11,13). Our finding raised the possibility that TAK1-dependent phosphorylation of SnoN may induce SnoN ubiquitination and degradation. To investigate the relationship between SnoN phosphorylation and ubiquitination, we asked  whether TAK1 induces SnoN ubiquitination. 293 cells were transfected with HA-tagged SnoN or SnoN-(1-366), TAK1, TAB1, and FLAG-ubiquitin. We immunoprecipitated SnoN followed by immunoblotting for ubiquitin (anti-FLAG) (Fig.  5A). The full-length SnoN was ubiquitinated to some extent in the absence of TAK1, and the level of ubiquitination was increased by co-expression of TAK1 ϩ TAB1. However, the kinase-inactive TAK1(K63W) could not increase the ubiquitination. This suggests that TAK1-dependent phosphorylation of SnoN can trigger its ubiquitination. To verify the role of TAK1-dependent phosphorylation, we used the SnoN mutant lacking the phosphorylation sites SnoN(1-366 AAA). 293 cells were transfected with FLAG-tagged SnoN-(1-366), TAK1, TAB1, and HA-tagged ubiquitin. SnoN-(1-366) was immunoprecipitated with anti-FLAG antibody, and ubiquitinated SnoN-(1-366) was detected by immunoblotting with anti-HA antibody (Fig. 5B). Co-expression of TAK1 and TAB1 resulted in a marked increase in the ubiquitination of SnoN-(1-366). In contrast, SnoN(1-366 AAA) mutant, which lacks major phosphorylation sties, showed almost no ubiquitination under the same conditions. To further investigate the effect of TAK1 phosphorylation on the full-length SnoN, we generated a mutant full-length SnoN carrying the mutation at Ser-115, Ser-117, and/or Thr-119 (SnoN AAA). Although the basal level of ubiquitination was unchanged in the mutant SnoN (SnoN AAA), The TAK1 ϩ TAB1-induced increase of ubiquitination was abrogated in SnoN AAA. These results suggest that phosphorylation of SnoN at Ser-115, Ser-117, and/or Thr-119 by TAK1 is important for SnoN ubiquitination.
To investigate whether phosphorylation of SnoN was required for TGF-␤-induced degradation, we generated HaCaT cells stably expressing SnoN-(1-366), and SnoN-(1-366 AAA) mutant. Whereas SnoN-(1-366) was rapidly decreased in response to TGF-␤ stimulation, the SnoN-(1-366 AAA) levels decreased slowly compared with the wild type (Fig.  6B). Finally, we tested the TGF-␤-induced degradation of the full-length SnoN. We generated HaCaT cells stably expressing SnoN wild type and AAA mutant and examined their levels upon TGF-␤ treatment (Fig. 6C). We used two independent stable clones that express SnoN or SnoN AAA at different levels. SnoN AAA decreased slowly compared with the wild type SnoN in both clones. These results suggest that TAK1-dependent phosphorylation is important for the TGF-␤-induced degradation of SnoN.
SnoN represses TGF-␤ signaling by recruiting transcriptional repressors to Smad complex (3,14). We next examined whether the phosphorylation of SnoN modulates its transcriptional activity. Transient transfection experiments were performed in 293 cells with a transcriptional reporter, 3TP-lux, which contains TGF-␤-responsive elements of the PAI-1 promoter region (29). At 48 h after transfection, cell lysates were prepared, and luciferase activities were measured (supplemental Fig. S2). The constitutively active form of TGF-␤ receptor ALK5(TD) was sufficient to induce the expression of a TGF-␤responsive gene in 293 cells, and overexpressed full-length SnoN as well as SnoN-(1-366) suppressed the TGF-␤-dependent transcription as reported previously (3,14). SnoN-(1-366 AAA and 8A) mutants could also reduce the TGF-␤-dependent transcription, suggesting that the mutation does not affect binding to Smads or to transcriptional co-repressors. SnoN-(1-366) mutants may be a more potent inhibitor compared with SnoN-(1-366), because it is stable upon TGF-␤ stimulation. However, in the transiently transfection experiments, we could not detect the difference between SnoN-(1-366) and SnoN-(1-366 AAA and 8A), which is likely because they were highly expressed and ALK5(TD) could not effectively reduce the amount of SnoN-(1-366). Collectively, our results suggest that SnoN-(1-366 AAA and 8A) can bind to and inhibit Smads but is resistant to TAK1-mediated degradation. Therefore, TAK1 is likely to inhibit SnoN by modulating SnoN stability.
In this report, we have determined the role of TAK1 MAP-KKK in TGF-␤. Previous works had shown that TAK1 is activated by TGF-␤ (15) and that SnoN, an inhibitor of Smads, is degraded upon TGF-␤ stimulation (13). This study links these two observations and suggests that TAK1 contributes to the induction of TGF-␤-responsive genes by inducing the degradation of SnoN (Fig. 7). SnoN has been shown to be the important negative regulator of TGF-␤ signaling via its interaction with Smad proteins (7,8). Upon TGF-␤ stimulation, SnoN is rapidly degraded by ubiquitin-dependent proteasome pathway (13,14). Two ubiquitin ligases are reported as SnoN ubiquitin ligases. One is anaphase-promoting complex, which induces the ubiquitination of SnoN on Lys-440, Lys-446, and Lys-449 and its consequent degradation in a Smad3-dependent manner (11,12). Another ubiquitin ligase is Smurf2, which is recruited to SnoN by Smad2, resulting in the ubiquitination and degradation of SnoN (10). SnoN-(1-366), which lacks sites ubiquitinated by APC or Smurf2, is neither ubiquitinated nor degraded in Ba/F3 pro-B cells (14). However, we show that TGF-␤ induces degradation of SnoN-(1-366) in human keratinocyte HaCaT cells in a manner dependent on TAK1-induced phosphorylation and ubiquitination. Mutation of TAK1-dependent phosphorylation sites on SnoN-(1-366) blocked TGF-␤dependent degradation. Moreover, when endogenous TAK1 was inactivated by a small molecule TAK1 inhibitor in keratinocyte HaCaT cells or by siRNA-mediated knockdown in epithelial-like HeLa S3 cells, degradation of endogenous SnoN was impaired. Collectively, these results suggest that TAK1 phosphorylation of SnoN is required for its ubiquitination and degradation in some epithelial cells. These results further suggest that several different pathways induce SnoN degradation, depending on the cell type.
Phosphorylation-induced degradation of proteins is a widely used mechanism by which protein levels can be modulated rap- idly. We have found that TAK1 phosphorylates SnoN-(1-366) at several threonine and serine residues and that the SnoN mutant, which lacks its phosphorylation sites, did not undergo ubiquitination or degradation. We should note that the mutant SnoN is still capable of inhibiting TGF-␤-induced transcription (supplemental Fig. S2). This indicates that the mutations at the phosphorylation sites do not interfere with interaction of SnoN with Smads or with transcriptional co-repressors. TAK1 is likely to inhibit SnoN solely by modulating SnoN stability. TAK1 is the first kinase demonstrated to phosphorylate SnoN and target it for ubiquitin-dependent proteasomal degradation.