Modulation of Smad2-mediated Signaling by Extracellular Signal-regulated Kinase*

Components of the transforming growth factor-β and mitogen-activated protein kinase pathways interact in controlling cell growth and differentiation. We show that phosphorylation of Smad2, a mediator of the activin/transforming growth factor-β signal, by activated extracellular signal-regulated kinase 1 (ERK1) increases the amount of Smad2 protein and leads to enhanced transcriptional activity. Epidermal growth factor increased phosphorylation of Smad2 in COS7 cells, and Smad2-dependent transcription in a mink lung epithelial cell line, L17, was enhanced by co-transfection of a constitutively active MEK1. In addition, transfection of Smad2 mutants lacking ERK sites resulted in reduced transcription, whereas mutants that mimicked ERK phosphorylation stimulated transcription. The amount of Smad2 protein was increased by transfection with a constitutively active MEK1 and reduced by co-transfection with the ERK phosphatase, HVH2. The elevation of Smad2 protein levels was because of increased half-life and resulted in increased complex formation with Smad4. A site of ERK-dependent phosphorylation on Smad2 was located to Thr8, a site that overlaps with the calmodulin binding region. We show that calmodulin inhibits Smad2 phosphorylation by ERK1, and overexpressing calmodulin, or stimulating calmodulin activity with ionomycin, reduces Smad2 levels. These findings suggest that the ERK pathway positively regulates Smad2 signaling by phosphorylating Smad2 and that negative regulation of Smad2 signaling by calmodulin is achieved in part by inhibiting this phosphorylation.

type II receptor serine kinases. Transphosphorylation of a type I receptor serine kinase, ALK4 for activin and ALK5 for TGF-␤, by the type II receptor serine kinase activates the type I receptor, which then phosphorylates the C-terminal Ser residues of Smad2 and -3. At present, no difference has been established between the intracellular signaling pathways of activin and TGF-␤. Once phosphorylated, Smad2 and -3 form heteromeric complexes with Smad4, followed by translocation of the complexes into the nucleus. There they modulate target gene transcription either by associating with various DNA binding partners or by inducing ubiquitin-mediated protein degradation (2)(3)(4)(5).
In addition to Smad activation by activin/TGF-␤-dependent phosphorylation (6), other interactions are also likely to regulate Smad signaling. We have revealed previously that calmodulin binds with Smad2 and acts as a Smad modulator (7). Scherer and Graff (8) have also suggested that calmodulin and extracellular signal-regulated kinase (ERK) may interact in their effects on modulating Smad signaling during Xenopus embryogenesis. Furthermore, cross-talk between the activin/ TGF-␤ pathway and the MAP kinase pathway is well known in various biological processes; a dominant negative ras mutant blocked mesoderm induction by activin, whereas a constitutively active ras mutant mimicked the inducing activity (9). In addition, TGF-␤ accelerated epithelial-fibroblastoid conversion of mammary epithelial cells in a Ras-dependent manner (10), and signaling by c-Jun N-terminal kinase (JNK) was necessary for TGF-␤-induced fibronectin synthesis in fibrosarcoma cells (11). Oncogenic Ras has also been shown to block the growth inhibitory effect of TGF-␤ (12). Furthermore, Dok-1, a rasGAPbinding protein that inhibits the Ras pathway (13), was required for activin-induced apoptosis in B-lineage cells (14).
The present study was performed to establish a mechanistic basis for the regulation of Smad signaling by these other major intracellular signaling pathways. Here we report that activated ERK phosphorylates Smad2 and that this in turn leads to enhanced transcription of a promoter containing an activin/ TGF-␤-responsive element. In addition, we have located one of the ERK phosphorylation sites on Smad2. That site lies within the primary calmodulin binding sequence, and we have therefore been led to examine the role of calmodulin in ERK-induced modulation of Smad2 function.
Cell Culture and DNA Transfection-Cell line L17, a derivative of the mink lung epithelial cell line (Mv1Lu) (24), obtained from by Dr. J. Massagué, and COS7 cells from ATCC were cultured as described previously (6,25). For transient transfection, cells in 24-or 6-well plates, or in 6-cm dishes, were transfected by the DEAE-dextran method.
Metabolic Labeling-For metabolic 32 P labeling of COS7 cells, cultures were transfected with HA-tagged Smad2. At 36 h after transfection the cells were transferred to phosphate-free medium supplemented with 1% dialyzed fetal bovine serum for 30 min, followed by labeling with 1.0 mCi/ml of [ 32 P]orthophosphate (ICN, Costa Mesa, CA) for 4 h. The MEK1 inhibitor PD98059 (50 M) was added to the indicated lanes 30 min prior to EGF addition. After EGF stimulation (50 ng/ml) for 15 min, they were rinsed three times with HEPES dissociation buffer (25 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM KCl, 0.7 mM Na 2 HPO 4 ). They were then lysed in TNE buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40) containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1% aprotinin) for 30 min at 4°C, in the presence of heat-killed Staphylococcus aureus (Pansorbin; Calbiochem, La Jolla, CA) preadsorbed with normal rabbit serum. The lysates were centrifuged at 10,000 ϫ g for 20 min at 4°C. The supernatants were immunoprecipitated overnight at 4°C with anti-HA antibody (12CA5; Roche Molecular Biochemicals) and incubated with protein A-agarose beads (Invitrogen) for 1 h at 4°C. Samples were then washed two times in TNE buffer, two times in 0.5 M LiCl, and two times in deionized water, followed by elution in SDS-PAGE sample buffer.
For pulse-chase experiment, COS7 cells were transfected with HAtagged Smad2 with or without MEK1*. 36 h after transfection, they were incubated for 30 min in Cys-and Met-free medium and then incubated in fresh medium of the same composition containing 0.2 mCi of [ 35 S]Cys and Met (PerkinElmer Life Sciences) for 30 min. Thereafter they were rinsed twice with HEPES dissociation buffer and once with chase medium (Dulbecco's modified Eagle's medium containing 100 g/ml Met and 100 g/ml Cys), followed by incubation in the chase medium. At the indicated times, cells were rinsed three times with HEPES dissociation buffer and lysed in RIPA buffer (10 mM sodium phosphate, pH 7.2, 150 mM NaCl, 5 mM EDTA, 1% Nonidet P-40, 0.1% SDS, 1% sodium deoxycholate) for 30 min at 4°C. Insoluble material was discarded after centrifugation at 600 ϫ g for 5 min at 4°C. Lysates were immunoprecipitated overnight at 4°C with anti-HA antibody (12CA5), and protein A-agarose beads were added for 1 h at 4°C. They were then washed three times in RIPA buffer, two times in 0.5 M LiCl, and two times in deionized water, followed by elution in SDS-PAGE sample buffer.
All samples were subjected to 10% SDS-PAGE. Gels of 32 P-labeled samples were blotted to polyvinylidene difluoride membranes prior to exposure to x-ray film. Gels of 35 S-labeled samples were soaked for 30 min in 1 M sodium salicylate prior to drying and exposure to x-ray film.
In Vitro Kinase Assay-Smad2 and Smad4 proteins were expressed as GST fusion proteins in Escherichia coli and extensively purified after GST cleavage by thrombin, as described previously (26). The proteins were Ͼ90% pure on SDS-PAGE. The purified Smad proteins had appropriate in vitro biological activities; TGF-␤ receptor complexes phosphorylated Smad2, and Smad4 protein bound to Smad-binding DNA element (26). GST-MEK1(ED), GST-ERK1, and GST-Elk1 (305-425) were expressed in E. coli and purified with glutathione-Sepharose beads (Roche Molecular Biochemicals). The proteins were eluted from the beads with 50 mM Tris-HCl, pH 8.0, containing 10 mM reduced glutathione, followed by change of elution buffer to 20 mM HEPES, pH 7.5, 50 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol by means of Centriprep-10 (Millipore, Bedford, MA).
Five ng/l of GST-MEK1(ED) and/or 5 ng/l of GST-ERK1 were incubated for 25 min at 30°C in kinase assay buffer (20 mM HEPES, pH 7.5, 10 mM MgCl 2 , 0.2 mM ATP, 1 mM dithiothreitol). Subsequently, 4 g of Smad protein or 3 g of synthetic Smad peptide was incubated for 25 min at 30°C with 5 l of the activated kinase (above), in a total of 30 l of kinase buffer and 0.5 Ci of [ 32 P]ATP. To examine the effect of calmodulin on Smad phosphorylation, 4 g of Smad2 or GST-Elk1 (305-425) was incubated for 10 min at 25°C with calmodulin in kinase buffer containing 0.1 mM CaCl 2 or 1 mM EGTA prior to in vitro kinase reaction with activated ERK1. The calmodulin was purified from bovine brain (27). On the basis of the dissociation constants of Smad2 and calmodulin (58 nM), 2 sufficient calmodulin was added to complex with either 50 or 99% of Smad2. The reaction was terminated by adding 7.5 l of SDS-PAGE sample buffer. The 32 P-labeled samples were visualized as described above, except for SDS-PAGE of synthetic peptides where Tricine buffer was used (28). For Western blotting, anti-Smad2 antibody (S-20; Santa Cruz Biotechnology, Santa Cruz, CA) or anti-Smad4 antibody (H-552; Santa Cruz Biotechnology) was used as primary antibody, and bands were visualized with ECL reagent (Amersham Biosciences).
Two-dimensional Tryptic Peptide Mapping and Phosphoamino Acid Analysis-Purified Smad2 was phosphorylated with [ 32 P]ATP in vitro by activated ERK1, separated by SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. Tryptic digestion, two-dimensional phosphopeptide mapping, and phosphoamino acid analyses were performed as described previously (29,30).
Mapping of Phosphorylation Sites-100 g of purified Smad2 was phosphorylated by activated ERK1 in the presence of [ 32 P]ATP, followed by digestion with TPCK-trypsin (29). The resulting peptides were separated by reverse phase HPLC using a C 18 column (4 -60% acetonitrile linear gradient for 60 min). Peptide fractions were collected every 0.375 min and counted for 32 P. The 32 P-labeled peptides were then subjected to amino acid sequencing.
Sequential Immunoprecipitation-Immunoblotting-COS7 cells were transfected with HA-tagged Smad2 and FLAG-tagged Smad4, with or without MEK1*. 48 h after transfection, cells were lysed in TNE containing 10% (v/v) glycerol and protease inhibitors. After 30 min on ice, cell debris was removed by centrifugation at 600 ϫ g for 5 min at 4°C, and the supernatant was immunoprecipitated overnight at 4°C with anti-HA antibody (12CA5) and incubated with protein A-agarose beads for 1 h at 4°C. The beads were washed four times with TNE buffer containing 10% (v/v) glycerol, followed by elution in SDS-PAGE sample buffer. The immunoprecipitates were subjected to Western blotting with anti-HA antibody (12CA5) or anti-FLAG antibody (M2; Sigma) as primary antibody, and the bands were visualized with ECL reagent (Amersham Biosciences).
GST Pull Down Assay- 35 S-Labeled Smad2 wild-type and mutants were translated in vitro with the TNT rabbit reticulocyte lysate kit (Promega, Madison, WI), and the GST and GST-fused Smad4 proteins were expressed in E. coli and purified with glutathione-Sepharose beads according to the manufacturer's protocol. GST pull down assays were performed as described previously (6).
Fluorescence Analysis-The interaction of Smad2 mutants with calmodulin was evaluated by fluorescence assay as described previously (6). Various concentrations of purified Smad2 (Ͼ90% pure) were mixed with 140 nM dansylated calmodulin (Sigma) in 20 mM HEPES, pH 7.5, 130 mM KCl in the presence of 0.1 mM CaCl 2 or 1 mM EGTA. Fluorescence was measured with a FluoreMax-2 (Instruments SA Inc., Edison, NJ) with excitation at 340 mm, and 5-and 10-nm slits for excitation and emission, respectively. Emission spectra were recorded between 430 and 570 nm, and the maximal fluorescence intensity was measured. The differences in maximum intensity with and without Smad2 were plotted against free Smad2 concentration. A one-site binding model was applied (y ϭ a ϫ x/(K d ϩ x), where x is the concentration of free Smad2, and y is the difference between the fluorescence intensity in the presence and absence of Smad2, and the dissociation constant was calculated by use of GraphPad PRISM (GraphPad Software, Inc., San Diego, CA). Free Smad2 concentration was estimated from the following equation: free Smad2 (nM) ϭ total Smad2 (nM) Ϫ total calmodulin (nM) ϫ ⌬F/⌬F ϱ where ⌬F and ⌬F ϱ are the difference in fluorescence intensity at a given Smad2 concentration and at the highest Smad2 concentration, respectively.

Smad2
Is Phosphorylated by Activated ERK1-To determine whether Smad2 is phosphorylated in response to activation of the MAP kinase pathway, we examined Smad2 phosphorylation in response to EGF stimulation in COS7 cells metabolically labeled with [ 32 P]orthophosphate. Incorporation of isotope into Smad2 increased rapidly in response to EGF stimulation (Fig. 1A, lanes 2 and 3), and pretreatment with MEK1 inhibitor PD98059 effectively blocked Smad2 phosphorylation (Fig. 1A,  lanes 4 and 5). These results were consistent with previous studies (31,32). Because EGF activates the MEK-ERK pathway, we checked whether this kinase phosphorylates Smad2 in vitro. 32 P-Labeled Smad2 was indeed detected upon co-incubation with ERK1 in the presence of activated MEK1 (Fig. 1B), suggesting that activated ERK1 was responsible for the phosphorylation of Smad2 in intact cells. 32 P-Labeled Smad4 was not detected following incubation with activated ERK1 (Fig.  1B). Tryptic peptide mapping of 32 P-labeled Smad2 revealed several spots that migrated toward the cathode (Fig. 1C, left  panel). This tryptic digestion pattern differs from that of Smad2 phosphorylated by TGF-␤ receptor complexes, because the latter migrates toward the anode (26,(33)(34)(35). Phosphoamino acid analysis of 32 P-labeled Smad2 showed that ERK1 phosphorylates Ser and Thr, but not Tyr, residues (Fig.  1C, right panel).
To identify the Smad2 sites phosphorylated by ERK1, 32 P-labeled Smad2 was digested with trypsin, and the resulting peptides were separated by HPLC ( Fig. 2A, upper). There were two peaks of radioactivity ( Fig. 2A, lower). Amino acid sequencing indicated that peak a was the peptide beginning 183 HIE-ILTϪ, and peak b was the N-terminal peptide. Kretzschmar et al. (32) have suggested that the linker region of Smad2 is phosphorylated by the Ras pathway, and they proposed that Thr 220 and Ser 245 , Ser 250 , and Ser 255 were possible phosphorylation sites. The phosphorylation of peak a peptide by ERK1 is consistent with that prediction. There is also an ERK site (PX(S/T)P) located near the N terminus of Smad2 (PF 8 TP). To establish whether it is this Thr that is phosphorylated by activated ERK1, an in vitro kinase assay was performed using synthetic Smad peptides as substrate. Smad2 (2-21) was phosphorylated by ERK1 in a manner dependent on activated MEK1, and the residue phosphorylated was a Thr (Fig. 2B). As there is only one Thr in this peptide, namely Thr 8 , this result confirms that the predicted ERK site is phosphorylated. Smad4 (78 -88) was not phosphorylated by activated ERK1, and Smad1 (4 -23) was only weakly phosphorylated; the latter has an 11 SP sequence that is phosphorylated by proline-directed protein kinases including ERK (36).
We next expressed a mutant form of Smad2, referred to as Smad2(VA), that has Thr 8 and Thr 220 changed to Val and Ser 245 , Ser 250 , and Ser 255 changed to Ala (see Fig. 4A), and examined its phosphorylation in response to EGF stimulation in intact cells and by activated ERK1 in vitro. 32 P labeling of Smad2(VA) did not increase in response to EGF treatment in COS7 cells (Fig.  2C) nor did incubation of purified Smad2(VA) with activated ERK1 give rise to phosphorylation (Fig. 2D). These data confirm that the residues mutated in Smad2(VA) include all the sites phosphorylated by ERK1, both in vivo and in vitro.
Phosphorylation of Smad2 by ERK Increases Its Transcriptional Activity-Next we explored the effect of ERK-dependent phosphorylation on the transcriptional activity of Smad2. For this purpose we used a reporter gene, FoxH3-dependent AR3lux, that contains three copies of an activin-responsive element from Mix.2, fused to luciferase (17). Transfection of type II (T␤RII) and type I (ALK5) TGF-␤ receptor into a mink lung epithelial cell line did not affect luciferase expression, whereas transfection of a constitutively active type I TGF-␤ receptor (ALK5(TD)) (21) increased luciferase expression 7.5-fold. Transfection of a constitutively active form of MEK1, MEK1(ED) (15), had little effect on transcriptional activity by itself. In addition, co-transfection of MEK1(ED) with T␤RII or ALK5 also did not affect luciferase expression. However, when it was co-transfected with ALK5(TD) there was a further increase in luciferase expression. This suggests that the MEK1-ERK pathway synergizes with the TGF-␤ pathway in promoting transcription of AR3. We next checked the effect of the MEK1-ERK pathway on Smad2-mediated AR3 transcription. Co-transfection of MEK1(ED) increased wild-type Smad2-induced luciferase activity (Fig. 3B). Expression of a constitutively active Smad2, Smad2(2E) that has the TGF-␤ receptor kinase phosphorylation sites Ser 465 and Ser 467 replaced by Glu (6), resulted in 15-fold elevation of basal luciferase expression. Co-transfection of MEK1(ED) caused a further increase in luciferase activity. These results suggest that the MEK1-ERK pathway promotes Smad2-mediated AR3 transcription and that modulation of Smad2 by the MEK1-ERK pathway is independent of regulation via TGF-␤.
To further explore the role of phosphorylation of Smad2 by ERK1 in signaling, we introduced mutations at possible sites of phosphorylation by ERK1 and by activin and TGF-␤ receptor complexes; both non-phosphorylatable (Thr 3 Val and Ser 3 Ala) and phosphorylation-mimicking (Thr and Ser 3 Asp or Glu) substitutions were made (see Refs. 34 and 35 and Fig. 4A). Transfection with wild-type Smad2 elevated basal expression of AR3-lux 24-fold (Fig. 4B). This increase was greater than that shown in Fig. 3B, presumably because of the higher pro- FIG. 1. Smad2 phosphorylation by activated ERK1. A, Smad2 phosphorylation in response to EGF stimulation. COS7 cells were transiently transfected with HA-tagged Smad2 without or with MEK1 inhibitor PD98059 prior to the addition of EGF. HA-Smad2 was recovered by immunoprecipitation from 32 P-labeled cells, followed by SDS-PAGE and autoradiography. B, in vitro phosphorylation of purified Smad2 and Smad4 by MEK1 and ERK1. Purified Smad2 or Smad4 were incubated with activated MEK1 and ERK1, or activated ERK1 alone, in the presence of [ 32 P]ATP, followed by SDS-PAGE and autoradiography. C, tryptic digest of ERK1-phoshorylated Smad2 resolved by two-dimensional peptide mapping (left) and phosphoamino acid analysis of 32 Plabeled Smad2 (right). portion of Smad2-encoding DNA in the transfection mixture. Although AR3 transcription was not affected by the T8V/T220V mutant and the T220V, 3SA mutant, the T8V/T220V, 3SA triple mutant greatly decreased expression. On the other hand, substitution of the possible ERK sites with Asp increased expression of the reporter.
Smad2(2A) (with C-terminal Ser 465 and Ser 467 replaced by Ala) elevated basal expression of luciferase 21-fold, whereas the expression induced by the T8V/T220V mutant of Smad2(2A) was limited to 12-fold. In addition, the T220V, 3SA mutant and the T8V/T220V, 3SA triple mutant of Smad2(2A) further decreased expression to 4-and 2-fold, respectively. In contrast, mutation of the same residues to Asp increased expression 42-to 47-fold. These results, that the Val/Ala mutants of the ERK sites decreased and that the Asp mutants increased AR3 transcription, were basically similar to those observed when the C-terminal sequence was wild-type, although the inhibitory effects of the T8V/T220V mutant and the T220V, 3SA mutant were distinct.
The inhibitory effect of non-phosphorylatable mutants of ERK sites was most striking in the presence of Ser-Glu-Met-Glu at the C terminus (2E). Smad2(2E) elevated basal luciferase expression 162-fold, whereas all three Val/Ala mutants of Smad2(2E) clearly decreased expression below 40-fold, indicating that AR3 transcriptional responses due to different Val/Ala mutants are qualitatively different with different C-terminal changes. Although there were no differences between the Asp mutants of Smad2(2E) and Smad2(2E) itself, this was no doubt because both gave maximal expression. Titration of the amount of transfected DNA suggested that Smad2(T8V/T220V, 3SD-2E) was more potent than Smad2(2E) (Fig. 4C). All these data point to the conclusion that phosphorylation of Smad2 by the

MEK1-ERK pathway promotes AR3 transcription and that phosphorylation by both activated ERK and activin/TGF-␤ receptor complexes is necessary for maximal Smad2 activation.
Smad2 Protein Is Stabilized by ERK Phosphorylation-To explore the mechanism by which phosphorylation of Smad2 by the MEK1-ERK pathway stimulates signaling, we tested the effect of activating the MEK1-ERK pathway on the amount of exogenously expressed Smad2 in COS7 cells. Western blot analysis revealed that the amount of Smad2 protein was increased by expression of another constitutively active MEK1, MEK1*, with amino acid residues 32-51 deleted and Ser 218 and Ser 222 replaced by Ala (15) (Fig. 5A). Co-transfection of an ERK phosphatase, HVH2 (16), with MEK1* resulted in a decrease in Smad2 protein (Fig. 5A). These results suggest that the MEK1-ERK pathway increases the amount of Smad2 protein. To determine whether the increase is because of increased protein stability, we performed pulse-chase experiments. These showed that the half-life of Smad2 was prolonged by the presence of MEK1* (Fig. 5B), suggesting that ERK phosphorylation stabilizes Smad2 protein. We then examined the level of Smad2 protein for various ERK phosphorylation site mutants. Fig. 5C reveals that protein levels were decreased in the phosphorylation-defective mutants and increased in the phosphorylationmimicking mutants. This tendency was also observed in the mutants with S465A/S467A or S465E/S467E at their C terminus (data not shown).
Because the formation of Smad2⅐Smad4 complexes is essential for Smad2-mediated signaling (6,(37)(38)(39), the association of Smad2 and Smad4 was examined by sequential immunoprecipitation and immunoblotting. Complexes of wild-type Smad2 with Smad4 were below detection limits in the absence of MEK1* (Fig. 6A, upper); co-transfection of MEK1* resulted in significant levels of Smad4 in the immunoprecipitates, suggesting that the formation of Smad2⅐Smad4 complexes is promoted by stimulation of the MEK1-ERK pathway. In contrast, even in the absence of MEK1*, Smad4 formed complexes with Smad2(T8D/T220D, 3SD), whereas in this case MEK1* cotransfection did not increase complex formation. The observed increases in Smad2⅐Smad4 complex formation correlate with increases in AR3 transcription (Figs. 3B and 4B).
To determine whether the increase in Smad2⅐Smad4 complex formation following activation of the MEK1-ERK pathway is because of an increase in the affinity of Smad2 for Smad4, their in vitro association was examined by GST-pull down assay. The results in Fig. 6B show that in vitro translated 35 S-labeled Smad2(T8D/T220D, 3SD) bound to GST-Smad4 beads with similar efficiency to 35 S-wild-type Smad2. Moreover, in agreement with our previous study (6), more Smad2(2E) than wild-type Smad2 bound to GST-Smad4. Smad2(T8D/T220D, 3SD-2E) also bound to GST-Smad4 more efficiently than did wild-type Smad2, but no difference was seen between Smad2(2E) and Smad2(T8D/T220D,3SD-2E). These data suggest that phosphorylation of Smad2 by the MEK1-ERK pathway does not directly affect its affinity for Smad4.
Calmodulin Blocks Smad2 Phosphorylation by ERK1 and Decreases Smad2 Protein-We have shown that calmodulin associates physically with Smad2 (7), and the primary calmodulin-binding site has been mapped to the N-terminal basic amphiphilic ␣-helix, 3 which overlaps with the ERK sites. It was therefore of interest to examine the effect of calmodulin on the phosphorylation of Smad2 by ERK1. Neither activated MEK1 nor activated ERK1 phosphorylated calmodulin (data not shown). Smad2 was incubated with activated ERK1 in the presence and absence of calmodulin. As shown in Fig. 7, A and B, calmodulin significantly inhibited Smad2 phosphorylation by ERK1, although it had no effect on phosphorylation of a well characterized target of ERK, Elk (40). The inhibitory effect was abolished when the reactions were carried out in the presence of EGTA to prevent calmodulin binding (7).
We next examined the effect of Smad2 phosphorylation on calmodulin binding. Because Smad2 could not be stoichiometrically phosphorylated by ERK1 in vitro, we purified various Smad2 mutants that have acidic amino acids substitutions at their phosphorylation sites (Fig. 4A). Interaction of these proteins with calmodulin was assayed fluorimetrically by binding to dansylated calmodulin. All the mutants bound with similar affinity (Fig. 8), suggesting that there is little effect of ERK phosphorylation on the binding of calmodulin to Smad2.
Because calmodulin blocks ERK phosphorylation of Smad2, and because phosphorylation results in stabilization, we expected that calmodulin would decrease Smad2 protein levels. As predicted, calmodulin overexpression decreased the amount of Smad2 protein in a dose-dependent manner (Fig. 9A). Treatment with ionomycin, an ionophore that activates calmodulin by stimulating calcium influx (41), also decreased Smad2 protein levels (Fig. 9B). We showed previously that overexpression of calmodulin inhibits activin/TGF-␤-induced transcription, whereas inhibiting calmodulin enhances it (7). In view of the present results, this inhibitory effect may be because of calmodulin binding to Smad2 and blocking its ERK-dependent phosphorylation, so reducing its stability and hence the overall level of Smad2 protein. DISCUSSION Our findings have revealed that Smad2 is phosphorylated by ERK1 and that this results in enhanced transcription of an activin/TGF-␤-responsive DNA element. They have also shown that the enhanced Smad2-mediated signaling is because of increased levels of Smad2 protein, as a result of stabilization of the protein, and that this in turn leads to increased formation of active signaling complexes with Smad4. These observations suggest that there is cross-talk between the activin/TGF-␤ and ERK signaling pathways at the level of Smad2 activation and imply that phosphorylation of Smad2 by both activated ERK1 and activin/TGF-␤ receptor complexes is essential for maximal transcriptional activity.
Previous studies have addressed potential cross-talk between the activin/TGF-␤ and MAP kinase signaling pathways, but no clear consensus has emerged with respect to the functional consequences of these interactions. For example, activation of MAP kinase negatively regulated activin/TGF-␤-induced cell growth inhibition and apoptosis (14,32), whereas activin/TGF-␤ signaling and MAP kinase pathways synergized during Xenopus mesoderm induction (9,42,43) and fibronectin synthesis in fibrosarcoma cells (11). In addition, oncogenic Ras and TGF-␤ collaborated in epithelial-fibroblastoid conversion and invasion of epithelial tumor cells (10). These results suggest that the functional relationship between activin/TGF-␤ signaling and MAP kinase pathway depends on specific cellular context. In fact, the ability of TGF-␤ to cause invasive growth was not inhibited by activation of ERK signaling, but the apoptotic effects of TGF-␤ was blocked in Madin-Darby canine kidney cells (44). A molecular explanation for the synergistic effects of ERK on Smad2 signaling revealed in this study may explain the relationship between activin/TGF-␤ signaling and the MAP kinase pathway during early Xenopus embryogenesis and during metastasis of epithelial tumor cells. This is also consistent with the recent results of Oft et al. (45) showing that both Ras and Smad2 signaling are essential for epithelial to mesenchymal transition and that metastasis is driven by sequential elevation of Ras and Smad2 expression.
In agreement with previous studies (31, 32), we found that Smad2 was phosphorylated in response to EGF stimulation. An in vitro kinase assay revealed that activated ERK1 could phosphorylate Smad2 but not Smad4, and tryptic peptide mapping yielded a different pattern from that induced by TGF-␤ receptor complexes (26,(33)(34)(35). Brown et al. (46) have also reported that transfection of constitutively active MEK kinase 1, a component of the JNK pathway, induces Smad2 phosphorylation outside the C-terminal Ser-Ser-Met-Ser site. Although it remains uncertain whether Smad2 phosphorylation by ERK1 and JNK occurs at the same sites, these results suggest that Smad2 phosphorylation is regulated not only by receptors for activin and TGF-␤ but also by the MAP kinase family.
Our data on the effects of ERK phosphorylation on Smad2dependent transcription are consistent with findings by others. de Caestecker et al. (31) showed that activation of receptor tyrosine kinases by EGF or hepatocyte growth factor activated Smad2-dependent gene expression. In addition, activated MEK kinase 1 enhanced Smad2-dependent gene transcription (46), and co-transfection of JNK with Smad3 and Smad4 increased basal expression of a CAGA 12 -lux reporter gene (46) whose CAGA elements bind complexes of Smad3 and Smad4 (47).
Furthermore activin and MAP kinase-coupled signals synergized during early embryogenesis in Xenopus (48,49). Because the MAP kinase family is large, with a complex and incompletely understood set of regulators and targets (50), we cannot state unequivocally which members of the MAP kinase family phosphorylate Smad2 in vivo. Nevertheless our data suggest that phosphorylation of Smad2 by activated ERK1 is necessary for maximum transcription of Smad2-dependent genes.
Ser residues (Ser 465 and Ser 467 ) at the C terminus of Smad2 are phosphorylated in response to TGF-␤ stimulation (34,35), and a Smad2 mutant with Ala substituted for these serines has been reported to associate stably with TGF-␤ receptor complexes and reduce transcription of a TGF-␤-responsive reporter gene, 3TP-lux. This finding was taken as indicating that the mutant protein acted in a dominant negative manner (33). However, in our hands Smad2(2A) and wild-type Smad2 had comparable transcriptional activity with AR3-lux as reporter gene, suggesting that the Smad2(2A) mutant does not have a dominant negative effect at least when AR3-lux is the reporter.
Kretzschmar et al. (32) have reported that Smad3 mutants with Ser 3 Ala or Thr 3 Val substitutions at four potential ERK sites in the linker region (corresponding to the residues changed in Smad2; see Fig. 4A) increased basal expression of AR3-lux in Ras transformed cells. This result was taken to suggest that phosphorylation of Smad3 by the Ras pathway inhibits Smad3-mediated signaling, a view that contrasts with our findings, as well as those of others (31,44,51). Kretzschmar et al. (32) interpreted the stimulatory effect of EGF or hepatocyte growth factor on transcription of a TGF-␤responsive reporter gene (31) as pointing to a general stimulation of transcription by the Ras pathway. However, the present study has shown that amino acid substitutions in Smad2 that either mimic or inhibit ERK1 phosphorylation increase and decrease transcription, respectively. Differences between Smad2 and Smad3 may be responsible for the different result obtained. Thus, although the two proteins are closely related structurally, and both are thought to be activin/TGF-␤ signal mediators, Smad3 has been observed to inhibit Smad2⅐Smad4mediated transcription of a reporter gene from the goosecoid promoter (52). It is also possible that differences between the cell types used in the experiments may account for the divergent observations.
Our data suggest that Smad2 that has been phosphorylated by ERK1 has higher transcriptional activity primarily because it has greater stability and forms a greater number of complexes with Smad4. The increase in complex formation is presumably a simple mass action effect, because a Smad2 substitution mimicking phosphorylation by activated ERK1 did not alter the affinity of the resulting protein for Smad4. Brown et al. (46) have shown similarly that stimulation of the JNK pathway enhances the Smad2-Smad4 interaction in endothelial cells. It is known that activation of the MAP kinase pathway affects protein stability; the half-life of c-Myc protein markedly increased in response to stimulation of the Ras pathway (53), and phosphorylation of Ser 62 by ERK (54) was required for the Ras-induced stabilization of c-Myc (55).
In previous work we showed that calmodulin interacts with Smad2 in a calcium-dependent manner and that the N-terminal helix of Smad2 is the primary calmodulin binding region (7). The present results show that phosphorylation of Smad2 by ERK1 is blocked by interaction with calmodulin. Scherer and Graff (8) have also shown that binding of calmodulin to Smad2 inhibits subsequent ERK2-dependent phosphorylation of Smad2. It is possible that binding of calmodulin changes the conformation of Smad2 such that it is less efficiently phosphorylated. Alternatively, it may sterically block the relevant phosphorylation sites.
Earlier we demonstrated that overexpression of calmodulin inhibited activin/TGF-␤-responsive gene transcription and that inhibition of calmodulin activity stimulated transcription (7). Calmodulin overexpression also blocked Smad2-dependent morphogenesis in Xenopus (8). Interestingly, Wicks et al. (56) have shown that inhibition of TGF-␤-mediated transcriptional activation can result from phosphorylation of Smad2 by calmodulin-dependent kinase II and consequent inhibition of its nuclear translocation. In the present study, overexpression of calmodulin, or ionomycin treatment, decreased the level of Smad2 protein in L17 cells. Hence the inhibition of activin/ TGF-␤ signaling by increased calmodulin activity may be at least partly because of reduced phosphorylation of Smad2 via the ERK pathway and its resulting instability. This, therefore, could be an additional mechanism by which calmodulin regulates Smad2-mediated signaling.
The present study examined the mechanistic basis for regulation of Smad2 signaling by the ERK pathway and by calmodulin binding. Our findings suggest that the ERK pathway positively regulates Smad2 signaling by phosphorylating Smad2 and that calmodulin negatively regulates Smad2 activation by inhibiting this phosphorylation. The rigorous control of Smad2 activity, a signal mediator for activin and TGF-␤, by these other major intracellular signaling pathways may explain some of the diverse biological activities of activin and TGF-␤ in development, normal physiology, and pathology.