Transcriptional Induction of Salt-inducible Kinase 1 by Transforming Growth Factor β Leads to Negative Regulation of Type I Receptor Signaling in Cooperation with the Smurf2 Ubiquitin Ligase*

Background: The control of TGFβ signaling depends on many not well understood regulators. Results: TGFβ transcriptionally induces SIK1, which cooperates with the ubiquitin ligase Smurf2 to negatively regulate the signaling output. Conclusion: Transcriptional induction of SIK1 controls TGFβ signaling together with Smurf2 and Smad7. Significance: The molecular interplay between SIK1 and Smurf2 provides new means for controlling TGFβ signaling. Transforming growth factor β (TGFβ) regulates many physiological processes and requires control mechanisms to safeguard proper and timely action. We have previously described how negative regulation of TGFβ signaling is controlled by the serine/threonine kinase salt-inducible kinase 1 (SIK1). SIK1 forms complexes with the TGFβ type I receptor and with the inhibitory Smad7 and down-regulates the type I receptor. We now demonstrate that TGFβ induces SIK1 levels via a direct transcriptional mechanism that implicates the Smad proteins, and we have mapped a putative enhancer element on the SIK1 gene. We provide evidence that the ubiquitin ligase Smurf2 forms complexes and functionally cooperates with SIK1. Both the kinase activity of SIK1 and the ubiquitin ligase activity of Smurf2 are important for proper type I receptor turnover. We also show that knockdown of endogenous SIK1 and Smurf2 enhances physiological signaling by TGFβ that leads to epithelial growth arrest. In conclusion, TGFβ induces expression of Smad7, Smurf2, and SIK1, the products of which physically and functionally interlink to control the activity of this pathway.

Transforming growth factor ␤ (TGF␤) regulates many physiological processes and requires control mechanisms to safeguard proper and timely action. We have previously described how negative regulation of TGF␤ signaling is controlled by the serine/threonine kinase salt-inducible kinase 1 (SIK1). SIK1 forms complexes with the TGF␤ type I receptor and with the inhibitory Smad7 and down-regulates the type I receptor. We now demonstrate that TGF␤ induces SIK1 levels via a direct transcriptional mechanism that implicates the Smad proteins, and we have mapped a putative enhancer element on the SIK1 gene. We provide evidence that the ubiquitin ligase Smurf2 forms complexes and functionally cooperates with SIK1. Both the kinase activity of SIK1 and the ubiquitin ligase activity of Smurf2 are important for proper type I receptor turnover. We also show that knockdown of endogenous SIK1 and Smurf2 enhances physiological signaling by TGF␤ that leads to epithelial growth arrest. In conclusion, TGF␤ induces expression of Smad7, Smurf2, and SIK1, the products of which physically and functionally interlink to control the activity of this pathway.
TGF␤ signaling initiates when the extracellular dimeric TGF␤ ligand associates with serine/threonine kinase receptors type II (T␤RII) 5 and type I (T␤RI), also known as activin receptor-like kinase 5 (ALK5) (1). T␤RII trans-phosphorylates T␤RI, which in turn phosphorylates receptor-regulated Smads (R-Smads, Smad2, and Smad3). R-Smad phosphorylation is necessary for their association with Smad4, accumulation in the nucleus, and cooperation with transcriptional complexes to regulate gene expression (2). TGF␤ signaling is regulated by various mechanisms that operate outside the cell, at the cell membrane, in the cytoplasm, or in the nucleus (3). Intracellular regulation of the TGF␤ signaling network relies to a large extent on the time point of signal transduction and on the cell compartment where various post-translational modifications of signaling proteins occur (4). In this manner, the multifaceted actions of TGF␤ are regulated during embryonic development, adult organ homeostasis, and disease.
Among the negative regulators of TGF␤ signaling are the inhibitory Smad7 and the E3 ubiquitin ligase Smurf2, both operating within a negative feedback mechanism and controlling the strength and duration of signal transduction (3,4). The Smad7 and Smurf2 genes are immediate-early TGF␤-inducible genes (5,6). Smad7 binds directly to ALK5, leading to competitive inhibition of Smad2 and Smad3 phosphorylation by the receptor (5,7). Smad7 also binds directly to Smurf2 and its homolog Smurf1, thus leading to ALK5 ubiquitination and down-regulation (8,9). In addition, Smurf1 and Smurf2 ubiquitinate and regulate the stability of Smad proteins (10), the mitotic checkpoint protein Mad2 that controls proper spindle assembly during cell division (11), and also of the serine/threonine kinase MEKK2, which is required for the differentiation of bone cells (12). Furthermore, Smurf1 and Smurf2 ubiquitinate the small GTPases RhoA and Rap1B, the actin-binding protein talin, and the planar cell polarity protein Prickle, thus regulating epithelial and neuronal cell polarity, contractility of the cytoskeleton, and amoeboid cell migration (13)(14)(15)(16)(17). Such molecular functions may explain the role of Smurf1 and Smurf2 in the process of breast cancer cell invasiveness and metastasis (18,19).
We have previously identified a new gene target of TGF␤ signaling, the salt-inducible kinase 1 (SIK1, hereby abbreviated as SIK), which encodes a serine/threonine kinase of the AMPactivated protein kinase (AMPK) family (20). SIK has a modular structure with an N-terminal kinase domain and a middle ubiquitin-associated domain, which is followed by a long C-terminal sequence (21). SIK expression is induced during cardiogenesis and skeletal muscle differentiation (22). SIK function is required for cardiomyocyte differentiation, where it affects the expression of the cell cycle inhibitor p57 (23), and for skeletal myogenesis, where SIK phosphorylates class II histone deacetylases (24). SIK is also induced in adrenal glands, leading to steroidogenesis (25) and regulation of sodium transport (26). Another important pathway under the control of SIK activity is the regulation of the cAMP-responsive element binding protein (CREB) (27). SIK directly phosphorylates and inactivates the transducer of regulated CREB activity (TORC), a critical transcriptional co-activator of CREB, and in this manner SIK represses CREB function. The same mechanism, when catalyzed by the SIK isoform SIK2 that phosphorylates TORC2, leads to recruitment of the COP1 signalosome regulator that mediates TORC2 ubiquitination and degradation (28). This specific mechanism appears to be defective in diabetes, resulting in TORC2 stabilization and enhancement of the gluconeogenic gene expression program. We have demonstrated that SIK also inhibits TGF␤ signaling by inducing T␤RI/ALK5 receptor down-regulation (29). Regulation of TGF␤ signaling by SIK is compatible with independent reports on the Caenorhabditis elegans ortholog of SIK, KIN-29, that regulates chemosensory neuronal signaling and body size, processes dependent on TGF␤/Smad signaling (30,31).
In this study we have explored the mechanisms by which TGF␤ regulates SIK expression and achieves ALK5 down-regulation. Our findings clarify how the SIK gene is regulated by Smads and place SIK in close association and functional interaction with the ubiquitin ligase Smurf2.
Adenoviruses expressing human SIK epitope tagged with FLAG at its N terminus and control virus expression bacterial lacZ protein were previously described by us and were amplified, titrated, and propagated as described before (29).
Promoter Cloning and Luciferase Reporter Constructs-The human SIK promoter-enhancer sequences were amplified from genomic DNA isolated from human HaCaT keratinocytes using primers mapping upstream and downstream of the transcriptional start site (TSS) and upstream and downstream of the investigated putative enhancer element of the human gene. For the amplification of the promoter fragment the primers used were: forward (5Ј-GAGCTCATCCTCGTTTCTCCG-3Ј) and reverse (5Ј-GAGCTCGGGTGCCTACTGCT-3Ј). For the amplification of the enhancer fragment the primers used were forward (5Ј-GGATCCCATGAGGAGAGCAGGC-3Ј) and reverse (5Ј-GTCGACGAGGCTGCCTGGAGAC-3Ј). The amplified sequences were cloned into vector pGL4.12 (Pro-mega Corp., Madison, WI) in two steps. (a) The PCR-amplified genomic DNA fragments were blunt end-ligated into the pGL4.12 vector after cutting with EcoRV. (b) The subcloned promoter and enhancer fragments were removed from the first recombinant plasmids with SacI (promoter) or SalI/BamHI (enhancer) and religated to pGL4.12, producing pGL4.12-hSIKP (carrying the human SIK promoter only) and pGL4.12-hSIKPE (carrying the SIK promoter and enhancer), the latter cloned downstream of the luciferase cDNA sequence, aiming at mimicking the endogenous SIK gene organization (Fig. 3E). The cloned promoter fragment corresponds to 1214 bp spanning Ϫ1151 to ϩ63 bp relative to the TSS of the SIK gene. The cloned enhancer fragment corresponds to 451 bp spanning ϩ14,423 to ϩ14,874 bp relative to the TSS, the sequences located in the 3Ј intergenic region downstream from the SIK gene (Fig. 3E). All SIK gene base pair coordinates are given based on the ENSEMBL GRCh37 version of the human genome. pEGFP-N3 (Takara Bio Europe/Clontech, France) was used for normalization of promoter assays.
Real-time RT-PCR-HaCaT cells were treated and/or transfected as indicated in the figures before extraction of RNA using either RNeasy (Qiagen NORDIC, Sollentuna, Sweden) or the nucleic acid extractor NorDiag Arrow (CE, IVD) using the manufacturer's kit and protocol (NorDiag AB, Hägersten, Sweden). For the cycloheximide and actinomycin D experiments, the HaCaT cell culture was treated with 50 M cycloheximide, 2 g/ml actinomycin D (ActD), or the corresponding volume of vehicle (DMSO), which were added to the cells 1 h before the respective TGF␤ stimulus per time point.
Promoter Reporter Assays-The human SIK promoter-enhancer constructs were co-transfected with reporter plasmid pEGFP-N3 for normalization and Smad expression vectors as described in the figures in HEK-293T cells. The enhanced luciferase assay kit from BD Pharmingen was used. Normalized luciferase activity data are plotted in bar graphs representing the mean Ϯ S.D. from triplicate samples.
Chromatin Immunoprecipitation (ChIP)-HaCaT cells were cultured in 10-cm plates to 80% confluence, and 1 plate was used per immunoprecipitation. Cells were fixed with 1% formaldehyde for 10 min at room temperature with swirling. Excess aldehyde was quenched with glycine, which was added to a final concentration of 0.125 M, and the incubation was continued for an additional 5 min. Cells were washed twice with ice-cold phosphate-buffered saline and harvested, and their pellets were resuspended in 1 ml of sodium dodecyl sulfate (SDS) lysis buffer (50 mM Tris-HCl (pH 8.1), 1% SDS, 10 mM EDTA, protease inhibitors (Complete EDTA-free protease inhibitors from Roche Diagnostics)). Samples were sonicated 3 times for 30 s each time (output H) at intervals of 30 s with a Diagenode Bioruptor sonicator.
Immunoblotting and Immunoprecipitation Assays-SDS-PAGE, immunoblot, and co-immunoprecipitation analysis was as described (20,34). Protein G-Sepharose was purchased from GE Healthcare and Dynabeads protein A from Invitrogen. For the endogenous co-immunoprecipitation experiment of Smad7, Smurf2, and ALK5 after anti-SIK immunoprecipitation, anti-FLAG (M5) IgG was used as the negative control (see Fig. 4C).
Immunofluorescence and Confocal Microscopy-Approximately 70% confluent-transfected Mv1Lu monolayers were analyzed by immunofluorescence 24 h post-transfection as described (29). Nuclei were counterstained with 4Ј,6Ј-diamidino-2-phenylindole (DAPI) or propidium iodide. A Zeiss Axiovert 200M confocal microscope equipped with LSM 510 laser was used with the Zeiss 63ϫ/0.75 objective lens and photographing at ambient temperature in the presence of immersion oil. For the endogenous immunofluorescence experiments, MDA-MB-231 cells were cultured on standard 8-well glass plates before fixation, and photomicrographs were obtained by a Zeiss Axioplan 2 microscope with a Hammamatsu C4742-95 digital camera using the Zeiss Plan-neofluar 100ϫ/01.4 Iris objective lens. For fluorescence microscopy of live NMuMG-Fucci cells growing on a culture dish, a Zeiss Axiovert 40 CFL with an AxioCam MRc digital camera was used using the Zeiss Plan-neofluar 10ϫ/0.3 objective lens. Primary images were acquired with the camera's Volocity (MDA-MB-231 assays), QED Camera Plug-in v.1.1.6 (QED Imaging Inc.) (Mv1Lu assays) and AxioVision v4.8.2.0 (NMuMG-Fucci) software. Image memory content was reduced, and brightness-contrast was adjusted using Adobe Photoshop CS3 Extended.
Live Cell Cycle Analysis Assay-NMuMG-Fucci cells were transiently transfected with siRNAs twice followed by a transient adenoviral infection with control Ad-LacZ or Ad-SIK viruses as described above. Forty-eight hours after the first siRNA transfection and 24 h after the adenoviral infection and second siRNA transfection, cells were stimulated with vehicle or TGF␤1 for 56 h before fluorescence microscopy and image acquisition as explained above. The red and green fluorescent cells were counted using the ImageJ software (rsbweb.nih.gov) and are expressed as percent of red or green cells relative to the total number of cells counted. For each independent condition, two photomicrographs were captured by the microscope's camera, and two separate fields of 500 cells were counted. The numbers of cells were averaged among the two fields and the two independent photomicrographs to calculate the percentage of red or green cells.

RESULTS
Comparative Analysis of SIK and Smad7 mRNA Induction by TGF␤-We have previously shown that SIK down-regulates the T␤RI/ALK5 in cooperation with Smad7 (29). Interestingly, both SIK and Smad7 are up-regulated relatively early after TGF␤ stimulation, and their mRNAs showed a roughly 7-fold peak induction after 1 h of TGF␤ stimulation in HaCaT cells (Fig. 1, A and B). The transcriptional induction of SIK mRNA has previously been seen in mouse mammary epithelial NMuMG cells (36), human breast cancer MDA-MB-468 cells (20), and in independent studies of transcriptomic responses to TGF␤ (37,38). We confirmed such previous reports beyond the HaCaT cell system by measuring induction of SIK and Smad7 mRNAs by TGF␤ after 1 h of stimulation of human breast epithelial cells MCF10A (MI), their Ras-transformed derivatives (MII), and tumorigenic clones derived from the latter (MIII and MIV) and in the human metastatic breast cancer cell line MDA-MB-231 (supplemental Fig. 1, A and B).
The increase on SIK and Smad7 mRNA levels depended on RNA polymerase II activity as ActD reduced the levels of each mRNA to background, and TGF␤ was unable to exert any effect in the presence of ActD (Fig. 1, A and B). As control, ActD was shown to minimally or not at all affect expression of the 5 S rRNA that is transcribed by RNA polymerase III (supplemental Fig. 1C). TGF␤ signaling did not appreciably alter 5 S rRNA levels as expected. These data confirm that TGF␤ does not act by stabilizing the mRNA of SIK or Smad7.
Next, we compared the regulation of SIK and Smad7 mRNA levels over longer times of TGF␤ stimulation (0 -24 h) and examined whether both genes are direct targets of TGF␤/Smad signaling (Fig. 1, C-F). Both SIK and Smad7 mRNAs rapidly reached peak levels after 1-2 h stimulation, and their expression remained elevated above basal levels throughout the 24 h period (Fig. 1, C and D).
Treating HaCaT cells with cycloheximide before stimulation to block de novo protein synthesis did not inhibit SIK and Smad7 mRNA induction by TGF␤ (Fig. 1, E and F). This indicates that both SIK and Smad7 are direct target genes of TGF␤ signaling. Interestingly, however, the long term pattern of mRNA expression changed with the addition of cycloheximide. Both SIK and Smad7 mRNAs accumulated slower, reaching their peak levels at 4 h (Fig. 1, E and F). Also notable is that when de novo protein synthesis was blocked, neither of the two genes could uphold its plateau levels, as the peak of expression was followed by a slow but steady decline toward basal levels during the 24-h time course.
In summary, the rapid mRNA accumulation of SIK and Smad7 by TGF␤ is independent of de novo protein synthesis (Fig. 1). This strongly suggests that both genes are direct targets of the TGF␤ signaling pathway, and their regulation can be verified in all mouse and human cell models examined so far.
Smad-dependent Transcriptional Regulation of SIK in Response to TGF␤-The direct effect of TGF␤ on SIK gene expression (Fig. 1) suggested that Smad signaling might be responsible for this regulation. Knockdown of each Smad protein of the TGF␤ pathway in HaCaT cells showed that Smad2 and Smad3 as well as Smad4 contribute to the up-regulation of SIK and Smad7 mRNAs (Fig. 2, A and B). We confirmed the efficiency of Smad knockdown by measuring their respective mRNA level (Fig. 2, C-E) and the corresponding protein level (Fig. 2, F and G). It is worth noting that the Smad3 siRNA pool used was less efficient compared with the Smad2 and Smad4 siRNA pools. Despite this, the knockdown of Smad3 had a strong impact on SIK and Smad7 gene expression (Fig. 2, A and  B). Thus, quantitatively Smad3 and Smad4 have a larger impact on the induction of SIK and Smad7 mRNA than Smad2. The contribution of each Smad to the induction of SIK and Smad7 mRNA seems to be similar between these two genes, suggesting that the same organization of Smad complexes might regulate their enhancers/promoters.
We then verified the presence of Smad complexes on a SIK enhancer region residing downstream of the 3Ј end of the SIK gene (Fig. 3). This region was previously identified in a genomewide screen for TGF␤-induced Smad2/3 binding using ChIP- chip analysis in HaCaT cells (35). Immunoprecipitation of Smad2/Smad3 revealed recruitment of these Smads to the SIK enhancer under control conditions, and this recruitment was enhanced ϳ2.7-fold after TGF␤ stimulation (Fig. 3A). This recruitment was specific to the SIK enhancer as shown by a lack of Smad2/3 recruitment to the control HBB gene sequence (Fig.  3B). Smad2/3 binding to the SIK enhancer exhibited similar pattern to those of the well characterized promoters of the PAI-1 and Smad7 genes (Fig. 3, C and D). A similar experiment with a Smad4-specific antibody confirmed the above result (supplemental Fig. 2). A clear TGF␤-dependent enrichment of SIK enhancer chromatin was measured in the Smad4 immunocomplexes (supplemental Fig. 2A), similar to the PAI-1 promoter enrichment (supplemental Fig. 2C) and unlike the negative control HBB sequence (supplemental Fig. 2B).
To confirm the endogenous ChIP data on the role of the SIK enhancer, we cloned the human SIK promoter and enhancer into a luciferase reporter construct (Fig. 3E) and performed promoter activation experiments in transfected HEK-293T cells. As predicted from the Smad-binding ChIP assays, the upstream promoter could be activated weakly by TGF␤ and/or co-transfection of Smad3 and Smad4 (Fig. 3, F and G). However, a luciferase construct that carried both the SIK promoter and the enhancer cloned downstream from the luciferase cDNA showed higher basal activity (Fig. 3F) and responded better to TGF␤ and/or Smad3/Smad4 co-transfection (Fig. 3, F and G).
In conclusion, this analysis demonstrates that TGF␤ signaling sends the Smad complex to the 3Ј intergenic enhancer of the SIK gene. This mechanism of gene regulation can only be partially recapitulated in vitro after cloning the enhancer element downstream of the basic promoter unit of this gene, which implies that additional regulatory sequences mediating TGF␤ responses may exist on the SIK gene.
Smurf2 Forms Complexes with SIK and Smad7-We have previously investigated the cooperation between SIK and Smad7 in down-regulating the T␤RI/ALK5 receptor (29); however, the mechanism whereby SIK works together with Smad7 has not been clear. A possible link could be the E3 ubiquitin ligase Smurf2, which is known to be induced upon TGF␤ stimulation, interacts with ALK5 and Smad7, and targets ALK5 receptors for ubiquitination (8,9). We verified that Smurf2 mRNA is rapidly but weakly up-regulated by TGF␤ signaling in HaCaT cells even in the presence of cycloheximide (Fig. 4A). Furthermore, the weak but reproducible induction of Smurf2 mRNA by TGF␤ could also be verified  in three (MI, MII, MIV) of the five breast cancer cell lines tested (supplemental Fig. 3A). Interaction between Smad7 and SIK has been shown previously (29). Here we further investigated a possible complex for-mation between Smad7, Smurf2, and SIK. First, we immunoprecipitated FLAG-Smad7 and confirmed that it could bind both 6Myc-SIK and catalytically inactive Myc-Smurf2 (C716G), which was used to avoid strong degradation of the   proteins in the complex (Fig. 4B). However, immunoblotting with the Myc antibody suggested a weaker binding of SIK to Smad7 compared with binding of Smad7 to Smurf2. Next, we also tested whether Smurf2 could interact with both SIK and Smad7. Indeed, co-immunoprecipitation experiments with catalytically inactive Smurf2 (C716G), 6Myc-SIK, and FLAG-Smad7 showed that they interacted also when Smurf2 was immunoprecipitated (supplemental Fig. 3B). Using our homemade anti-SIK antibody and HaCaT cell extracts, we immunoprecipitated endogenous SIK from cells stimulated with vehicle or TGF␤ (Fig. 4C). In the absence of TGF␤ stimulation we observed a complex with endogenous Smad7 and ALK5. After TGF␤ stimulation, the complex between SIK, Smad7, and ALK5 was again visible, and it now had also incorporated Smurf2. This experiment also demonstrated the induction of endogenous protein levels of SIK, Smad7, and Smurf2 by TGF␤ and the concomitant down-regulation of the ALK5 receptor (Fig. 4C). We then examined the role of the kinase activity of SIK on formation of protein complexes between Smad7, Smurf2, and SIK (Fig. 4D). Smad7 and SIK interacted irrespective of the kinase activity of SIK (Fig. 4D, lanes 4 and 7). The protein com-plex between SIK and Smad7 was weakly enhanced by the presence of wild-type Smurf2 (Fig. 4D, lane 5); however, the addition of catalytically inactive Smurf2(C716G) dramatically enhanced the complex between the three proteins (Fig. 4D, lane  6) at equal expression levels of wild-type and mutant Smurf2 (Fig. 4D). Surprisingly, when the same co-immunoprecipitation experiment was repeated with the SIK(K56R) mutant instead of wild-type kinase, the ability of Smurf2(C716G) to promote an enhanced SIK/Smad7/Smurf2 complex was reduced (Fig. 4D,  lanes 8 and 9). This suggests that the catalytic activity of SIK has an impact on formation of the complex between these three proteins. Lack of strongly enhanced protein complex accumulation by the wild-type Smurf2 is most likely due to the rapid dissociation or degradation caused by the recruitment of active Smurf2 into this complex, making it difficult to visualize the dynamics of this protein complex. Overall, these biochemical experiments suggested that SIK, Smad7, and Smurf2 are induced by TGF␤ signaling and can engage with each other in mutual complexes.
SIK Cooperates with Smurf2 and Smad7 to Down-regulate T␤RI/ALK5 Receptor-The interaction data suggested that the two enzymes, SIK and Smurf2, cooperate or depend on each Quantitative real-time RT-PCR analysis measuring Smurf2 mRNA levels with or without cycloheximide pretreatment for 1 h before stimulation with 5 ng/ml TGF-␤ for 1, 2, 4, 8, or 24 h is shown. The data are presented as in Fig. 1. B, co-precipitation of wild-type 6Myc-SIK and catalytically inactive Myc-Smurf2(CG) after immunoprecipitation of FLAG-Smad7 is shown. Total cell lysate (TCL) controls are shown. An asterisk indicates a nonspecific protein band. Note that the immunoprecipitation and total cell lysate proteins have been resolved on two different gels as illustrated by the size markers. C, shown is co-precipitation of endogenous Smad7, Smurf2, and ALK5 after immunoprecipitation (IP) of endogenous SIK in the absence (Ϫ) or presence (ϩ) of TGF␤1 stimulation for 16 h. Immunoprecipitation with an unrelated immunoglobulin (IgG Ctrl) served as negative control. Total cell lysate controls are also shown, and GAPDH serves as the protein loading control. D, shown is co-precipitation of wild-type or kinase-dead (K56R) 6Myc-SIK and wild-type or catalytically inactive Myc-Smurf2(CG) after immunoprecipitation of wild-type FLAG-Smad7. Note that the two top immunoblots represent anti-Myc blots at two different exposure times. The top, long exposure shows the 6Myc-SIK, and the second, short exposure shows the Myc-Smurf2. Lane numbers are duplicated at the top and bottom of the immunoblots for convenience.
other during TGF␤ receptor down-regulation. We tested this possibility by co-expressing wild-type SIK and Smurf2 (Fig. 5A). Increasing levels of SIK led to down-regulation of the constitutively active (CA) ALK5 receptor and Smurf2 combined with wild-type SIK had the same effect. Combining Smad7 with SIK had a similar effect on receptor down-regulation, but when we co-expressed SIK, Smurf2, and Smad7, we then observed essentially complete loss of the receptor (Fig. 5A). In a similar experiment where the level of Smurf2 was increased, we observed the same cooperation between SIK and Smurf2 on ALK5 receptor down-regulation (supplemental Fig. 4A). SIK also led to a significant down-regulation of Smurf2, and the presence of Smad7 enhanced this effect.
We examined the role of the kinase activity of SIK for T␤RI/ ALK5 receptor down-regulation and its cooperation with Smurf2 (Fig. 5B). Catalytically inactive SIK(K56R) blocked the effects of Smurf2 on receptor down-regulation (Fig. 5B). This implies that Smurf2 down-regulates the TGF␤ receptor more effectively when it cooperates with a catalytically active SIK.
Smurf2 resides in the nucleus and moves together with Smad7 to the cytoplasm in response to TGF␤ to reach the ALK5 receptor (9). SIK shuttles between the cytoplasm and the nucleus, and its localization can be regulated by steroids or by the 14-3-3 adaptor protein (25,39). Using immunofluorescence experiments, we investigated localization of SIK, Smurf2, and ALK5. In transfected, TGF␤-sensitive Mv1Lu cells, SIK showed both nuclear and cytoplasmic distribution, as expected but additionally localized close to the plasma membrane in pronounced punctated clusters (Fig. 5C). Co-localization of SIK, CA-ALK5, and Smurf2 was observed in these peripheral clusters (Fig. 5C, insets), with no obvious co-localization in the more diffuse pattern scattered in the rest of the cell body. Thus, SIK, Smurf2, and ALK5 may be able to form complexes in cytoplasmic regions proximal to the plasma membrane. We also attempted to perform the co-localization experiments at the endogenous level. We were hampered from succeeding in this aim as all of our antibodies that gave positive and specific results with endogenous proteins are raised in rabbits and thus prohibit us from performing double or triple immunofluorescence experiments. Despite this, single antibody experiments verified TGF␤-induced expression and distribution of endogenous SIK, Smad7, and Smurf2 in the nucleus and cytoplasm of HaCaT (not shown) and MDA-MB-231 cells (supplemental Fig. 4B). In conclusion, the evidence so far supports the existence of protein complexes between SIK, Smad7, and Smurf2 that could initiate the process of receptor turnover.
Functional Cooperation of SIK and Smurf2 in Regulation of Endogenous TGF␤ Signaling-To further investigate the effects of endogenous SIK and Smurf2 on TGF␤ signaling, we performed siRNA knockdowns and subsequently analyzed established cellular responses to TGF␤. The efficiency of silencing of SIK, Smurf2, or both was significant at both mRNA and protein levels (supplemental Fig. 5). We first analyzed the mRNA levels of well established target genes of TGF␤, such as p21, Smad7, Gadd45␤, fibronectin, and PAI-1 (Fig. 6A, and supplemental Fig. 6). When SIK, Smurf2, or both were silenced, all these target genes showed enhanced magnitude of response to TGF␤ (Fig. 6A, supplemental Fig. 6). We did not observe any genespecific differences in terms of the effect of silencing of SIK or Smurf2 on this set of genes, which is compatible with a role of SIK and Smurf2 at the receptor level. Overall, the observed enhancement in gene responses was similar after SIK or Smurf2 silencing (Fig. 6A, supplemental Fig. 6). Interestingly, during simultaneous silencing of both SIK and Smurf2, we observed similar gene responses as after single silencing (Fig. 6A, supplemental Fig. 6). The latter result is compatible with the model that SIK and Smurf2 participate in the same linear pathway or possibly act as part of one and the same functional protein complex as suggested by the biochemical evidence.
One of the hallmarks of biological TGF␤ responses in epithelial cells is the cell cycle arrest at the early G 1 phase mediated by  APRIL 13, 2012 • VOLUME 287 • NUMBER 16 transcriptional induction of cell cycle inhibitors such as p21 (40). We, therefore, examined the impact of SIK on TGF␤mediated epithelial cell growth arrest using the well established model of mouse mammary epithelial NMuMG cells. We employed a stable clone of NMuMG that expresses two fluorescent proteins providing the cell with a fluorescent ubiquitination-based cell cycle indication (Fucci) (33). In this system the green fluorescent protein mAG fused to geminin marks cells in S/M/G 2 phases, whereas the red fluorescent protein mKO2 fused to Cdt1 marks cells in G 1 /G 0 phases. The cells were either transiently transfected with siRNA against endogenous SIK or with an adenoviral vector expressing SIK (Fig. 6B). TGF␤ clearly induced cell cycle arrest in control cells as it shifted the percentage of cells in G 1 /G 0 from 25 to 80% (Fig. 6C). SIK overexpression via the adenoviral vector had minor effects on the basal level of cycling cells but had a strong negative effect on the response to TGF␤, reducing the cell cycle arrested cells from 80 to 58%. Conversely, silencing the endogenous SIK led again to minimal basal effects, but essentially 98% of the cells in multiple cultures became arrested in G 1 /G 0 (Fig. 6, B and C). The above data collectively demonstrate that SIK mediates a significant negative regulatory effect that appears to be specific to TGF␤, and for SIK to elicit this function, complementation with additional inhibitor proteins of the TGF␤ pathway is required.

DISCUSSION
SIK has been shown to be a negative regulator of TGF␤ receptor signaling (29). Here we provide insights into the mechanisms of (i) TGF␤-induced SIK gene expression and (ii) TGF␤ type I receptor down-regulation by SIK (Fig. 7). We describe that (a) the SIK gene is a direct target of TGF␤/Smad signaling, and its protein product participates in protein complexes to down-regulate ALK5, (b) the kinase activity of SIK and the ubiquitin ligase activity of Smurf2 affect the dynamics of protein complexes with Smad7, and both are required for optimal ALK5 down-regulation, and (c) the regulation of receptor levels has an immediate impact on physiological signaling by TGF␤, and this includes several genes and the cytostatic response.
SIK and Smad7 exhibit both direct (cycloheximide-insensitive/actinomycin D-sensitive) early transcriptional peaks and prolonged profiles of sustained expression that are indirect (cycloheximide-sensitive/actinomycin D-sensitive) (Fig. 1). It is possible that newly synthesized Smads are required for the expression of SIK and Smad7 over long periods of time. More likely, sustained SIK and Smad7 induction by TGF␤ requires the synthesis of additional transcriptional cofactors or regulators of mRNA processing, as has been previously established for other genes responding to TGF␤ signaling (40). At this stage we have not yet identified putative regulatory proteins whose synthesis is required for the sustained SIK or Smad7 expression.
The regulation of SIK gene expression by TGF␤ was previously found to be dependent on Smad4, based on genome-wide studies in the Smad4-deficient breast cancer cell line MDA-MB-468 (20). Here we established that all three Smads of the TGF␤ pathway, Smad2, Smad3, and Smad4, contribute to the accumulation of SIK and Smad7 mRNA in response to TGF␤ (Fig. 2). This is an important finding, as previous genome-wide   Fig. 1. Asterisks indicate significant differences determined by Student's t test with significance at p Ͻ 0.01. B, shown is live fluorescence microscopy of mammary epithelial NMuMG-Fucci cells transiently transfected with control scrambled siRNA followed by infection with control Ad-LacZ virus (control, upper panels), transiently transfected with specific siRNA targeting SIK (siSIK, bottom panels) or transiently infected with an adenoviral vector expressing SIK (middle panels). The cells were treated with vehicle (no TGF␤) or 5 ng/ml TGF␤1 for 56 h before photography. A bar indicates 20 m. C, shown is quantitative analysis of the cell cycle from experiments such as that shown in panel B. The number of red cells (G 1 /G 0 phase) were counted in duplicate photos from two independent experiments from each experimental condition and are plotted as % relative to the total number of cells. Statistical significance between conditions is shown with a asterisk that indicates p Ͻ 0.05. attempts to define the contribution of each one of the TGF␤ pathway Smads to target gene expression have not delivered clear conclusions due to differences in the technical platforms used or differences in the cell models used (35,(41)(42)(43). Finally, ChIP assays at the endogenous level and cloned promoter assays in transfected cells established that at least one Smadsensitive genomic region resides in the 3Ј direction of the SIK gene (Fig. 3). This enhancer element seems to account for a significant part of the SIK gene response to TGF␤; however, the experimental evidence suggests the existence of additional TGF␤-responding regulatory sequences on this gene.
The mechanism by which SIK acts on T␤RI/ALK5 clearly involves the adaptor protein Smad7 (29) and the E3 ubiquitin ligase Smurf2 (Figs. 4 and 5). Because Smurf2 ubiquitination activity has been linked to the regulation of Smad protein stability and function (4), it is possible that SIK might also regulate Smad protein function and turnover in addition to the regulation of the TGF␤ type I receptor.
An important event during the cooperative action of these proteins toward T␤RI/ALK5 seems to be phosphorylation by SIK, as its kinase activity is critical for receptor down-regulation. At this stage we do not know the substrate(s) of the SIK kinase in the T␤RI-Smad7-Smurf2 complex. One possibility is that formation and function of this protein complex requires phosphorylation by SIK. Alternatively, SIK-mediated phosphorylation might promote receptor trafficking to lysosomes/ proteasomes for degradation. The latter model is compatible with our unpublished evidence, which does not support a role of SIK in inducing ALK5 ubiquitination. Rather, SIK recognizes ubiquitinated Smad7 (or other proteins) via its ubiquitin-associated domain and localizes in proteasome-rich locations (29). Future work in the direction of understanding the role of SIKmediated phosphorylation during TGF␤ receptor internalization and degradation is warranted.
Altering the levels of SIK in epithelial cells clearly showed an impact on cell cycle regulation by TGF␤ (Fig. 6). However, we failed to observe the effects on the cell cycle in the absence of TGF␤ signaling (Fig. 6C), suggesting that SIK may not play an important functional role in regulating the cell cycle but, rather, acts as a regulator of other pathways, such as TGF␤, that feed into the control of cell division.
In summary, SIK provides molecular means for multifunctional regulation of the TGF␤ receptor-Smad7 complex (Fig. 7). More work into the mechanistic details of SIK-mediated TGF␤ receptor regulation could uncover novel targets for therapeutic intervention against the TGF␤ pathway. FIGURE 7. The role of SIK during TGF␤ signaling. Shown is a graphic presentation of the TGF␤ receptor-Smad pathway leading to transcriptional induction of the SIK gene in the nucleus and formation of a protein complex involving SIK, Smad7, Smurf2, and T␤RI/ALK5. The hetero-tetrameric T␤RII/ T␤RI receptor complex is shown bound to extracellular TGF␤. A circled P indicates established phosphorylation events. The individual Smad2 and Smad3 proteins are shown as a single Smad2/3 molecule for simplicity. On the DNA double helix, the promoter and enhancer sequences are indicated with distinct coloration, the promoter is shown occupied by the RNA polymerase II (Pol II), and the TSS of the gene is illustrated as a green arrow pointing in the direction of transcription.