ERK2- and p90Rsk2-dependent Pathways Regulate the CCAAT/Enhancer-binding Protein-β Interaction with Serum Response Factor*

The serum response element (SRE) of the c-fos promoter is a convergence point for mitogenic signaling pathways. Several transcription factors regulate SRE, including serum response factor (SRF), ternary complex factors, and CCAAT/enhancer-binding protein-β (C/EBPβ). C/EBPβ can interact with both SRF and the ternary complex factor family member Elk-1, but only in response to activated Ras. Transactivation of the SRE by C/EBPβ is also greatly stimulated by Ras. The Ras effectors that signal to C/EBPβ are unknown. In this report, we demonstrate that a consensus MAPK site in C/EBPβ is necessary for Ras stimulation of both C/EBPβ-SRF interaction and transactivation of the SRE by C/EBPβ. To dissect signaling pathways activated downstream of Ras, different Ras effector constructs were analyzed. We show that activated forms of Raf and phosphatidylinositol 3-kinase stimulate C/EBPβ-SRF interaction. We also show a novel selectivity for the MAPK family member ERK2, where dominant-negative ERK2, but not dominant-negative ERK1, blocks Ras stimulation of C/EBPβ-SRF interaction. In addition, recombinant C/EBPβ protein is phosphorylated by ERK2, but not by ERK1, in vitro. Finally, we demonstrate a requirement for p90Rsk2 in regulation of C/EBPβ-SRF interaction. These data show that multiple Ras effectors are required to regulate C/EBPβ and SRF association.

transcription factor that has a role in regulating the c-fos SRE (18,19). Our laboratory has previously shown that the activator isoform of C/EBP␤, p35-C/EBP␤ (also known as liver-enriched activator protein (LAP)), activates an SRE-driven reporter construct, whereas the repressor isoform, p20-C/EBP␤ (also known as liver-enriched inhibitory protein), represses serum stimulation of this reporter (19). We have also shown that p35-C/EBP␤ can interact with SRF and the TCF family member Elk-1 in vivo, but these interactions are dependent on the presence of activated Ras (20,21). It is unknown, however, which signaling pathways act downstream of Ras to stimulate p35-C/EBP␤ interaction with transcription factors at the SRE.
This report has analyzed the signaling pathways that are activated in response to Ras to stimulate the interaction of p35-C/EBP␤ with SRF and subsequently stimulate transactivation of the SRE. We show that a consensus MAPK site in C/EBP␤ is necessary for Ras-stimulated C/EBP␤-SRF interaction as well as C/EBP␤ transactivation of the SRE. Furthermore, we show that activated forms of Raf and PI3K can stimulate the C/EBP␤-SRF interaction. We also demonstrate that dominant-negative ERK2, but not ERK1, inhibits Ras stimulation of the C/EBP␤-SRF interaction. Recombinant C/EBP␤ is also selectively phosphorylated in vitro by ERK2, but not by ERK1. We also show a requirement for p90 Rsk2 in regulating the C/EBP␤-SRF interaction. We did not find a role for the Rho family of small GTPases in stimulation of the C/EBP␤-SRF interaction. These data strongly suggest that multiple Ras effectors are regulating C/EBP␤ and SRF to stimulate their interaction, resulting in enhanced transactivation of the SRE.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-NIH 3T3 fibroblasts (from American Type Culture Collection) were grown in Dulbecco's modified Eagle's medium with 10% calf serum (Colorado Serum Co.), 0.22% sodium bicarbonate, 25 units/ml penicillin G sodium, and 25 mg/ml streptomycin. NIH 3T3 transfections were performed using NovaFector (Venn Nova) or Trans-IT LT1 (PanVera) as described previously (21). Cell extracts of equivalent protein concentration were prepared and assayed for chloramphenicol acetyltransferase (CAT) reporter activity as previously described (22). For drug-treated samples, SB 202190 (10 M; Calbiochem) was added to the cells for 8 h prior to harvesting. An internal control plasmid to measure transfection efficiency could not be used because C/EBP␤ alters transcription of the control plasmid, thereby making the internal control invalid. The transfections were repeated multiple times to control for variability in transfection efficiency. In addition, different stocks of DNA were used for the transfections to control for the variability in transfection efficiency between DNA preparations.
pGAL4-SRFm103 was generated in a two-step protocol. First, pTL2/ SRFm103 (a gift of M. Greenberg, Harvard Medical School) (28) was digested with NotI and XbaI to release a 1.6-kilobase pair fragment. The fragment was inserted into similarly digested pcDNA3 to create pcDNA3-SRFm103. Second, pcDNA3-SRFm103 was digested with EcoRI to release a 1.6-kilobase pair fragment. This fragment was inserted into a similarly digested pGAL4 vector (CLONTECH) to generate pGAL4-SRFm103.
Recombinant Proteins-Histidine-tagged C/EBP␤ protein was produced in Escherichia coli by 80 mM isopropyl-␤-D-thiogalactopyranoside induction overnight at 37°C. Cells were harvested by centrifugation and resuspended in sonication buffer (50 mM sodium phosphate (pH 8), 300 mM NaCl, and protease/phosphatase inhibitors (1 g/ml aprotinin, 1 g/ml leupeptin, 1 mM phenylmethylsulfonyl chloride, and 0.1 M pepstatin)). Bacteria were lysed by lysozyme incubation, followed by three rounds of sonication at setting 2 and 30% duty. After clarification by centrifugation, the pellet was resuspended in equilibration buffer (6 M guanidine hydrochloride in 25 mM MES (pH 8) and protease/phosphatase inhibitors). The preparation was again clarified by centrifugation, and the supernatant was applied to a nickel-nitrilotriacetic acid affinity chromatography column (QIAGEN Inc.). The column was treated as described by the manufacturer for denaturing conditions. The protein was eluted from the column with 6 M guanidine hydrochloride in 25 mM MES (pH 8), 100 mM EDTA, and protease/phosphatase inhibitors. The protein was dialyzed to remove guanidine hydrochloride, with final dialysis in a solution containing 25 mM MES (pH 8), 5% glycerol, and 1 mM dithiothreitol.
Kinase Assays-Recombinant histidine-tagged p35-C/EBP␤ or myelin basic protein (MBP; Upstate Biotechnology, Inc.) was incubated with activated ERK1 (50 milliunits; Alexis Biochemicals) or ERK2 (20 milliunits; Upstate Biotechnology, Inc.) in the presence of 80 M ATP, 13 mM MgCl 2 , and 15 Ci of [␥-32 P]ATP. Reactions were carried out in a shaking 30°C incubator for 10 min. SDS sample buffer was added to the MBP kinase reaction and boiled for 5 min. For the p35-C/EBP␤ reactions, the tagged protein was immunoprecipitated with T7 tag antibody-agarose beads (Novagen). Briefly, the beads were mixed with 0.5 ml of immunoprecipitation buffer (10 mM Tris (pH 7.5), 50 mM NaCl, 1 mM EDTA, 0.25% IGEPAL CA-630 (Sigma), and protease/phosphatase inhibitors). The kinase reaction was added to the beads, and the beads were incubated with the protein for 2 h at 4°C. Beads were washed once with immunoprecipitation buffer before adding SDS sample buffer and boiled for 5 min. The samples were analyzed on an SDS-10% polyacrylamide gel, and phosphate incorporation was detected by autoradiography.
Immunoblotting-Whole cell extracts from NIH 3T3 cells containing equivalent protein concentration were analyzed by electrophoresis on an SDS-10% polyacrylamide gel. The gel was equilibrated in transfer buffer (33 mM Tris base, 192 mM glycine, and 20% methanol) for 15 min before transfer of proteins to Immobilon-P membrane (Millipore Corp.). After transfer, immunoblotting was performed as described previously (19), except a 1:4000 dilution of anti-SRF antibody (Santa Cruz Biotechnology) and a 1:8000 dilution of goat anti-rabbit secondary antibody (Roche Molecular Biochemicals) were used. The secondary antibody was detected using SuperSignal chemiluminescent substrate (Pierce).

RESULTS
An Intact MAPK Site in C/EBP␤ Is Necessary for Ras Stimulation of the C/EBP␤-SRF Interaction-Thr 235 (numbering for the human protein) of C/EBP␤, located in a consensus MAPK phosphorylation sequence, has been shown to be phosphorylated in vivo by activated Ras and also to be phosphoryl-ated by an MAPK preparation in vitro (30). Furthermore, mutation of Thr 235 to alanine inhibits Ras stimulation of the C/EBP␤-SRF interaction in the mammalian two-hybrid assay (20). To confirm that a proline-directed kinase is necessary for mediating Ras stimulation, we mutated Pro 189 (numbering for the rat protein) in the MAPK site (PGT 188 P 189 ) to glycine. If a proline-directed MAPK was targeting C/EBP␤ at this site, we hypothesized that this mutant would not be able to be phosphorylated and therefore be unresponsive to Ras. This hypothesis was tested in the mammalian two-hybrid assay.
The mammalian two-hybrid assay utilizes a CAT reporter gene upstream of five copies of the Gal4 DNA-binding site driven by the E1B minimal promoter (pG5CAT). A Gal4-SRF fusion protein has been previously described (20). SRF is therefore brought to the DNA by binding to the Gal4 sites. As shown in Fig. 1, when the reporter construct alone or the reporter and Gal4-SRF were transfected into NIH 3T3 cells, there was no increase in CAT activity in the presence or absence of an activated Ras construct (CMV-Ras.V12). SRF has a weak transactivation domain, as has been shown before (31), so no increase in transcription in the presence of SRF was observed. When an expression construct encoding wild-type p35-C/EBP␤ (CMV-LAP) was cotransfected with Gal4-SRF and the reporter into NIH 3T3 cells in the presence of an activated Ras construct, there was a large increase in CAT activity, with an average increase of 70-fold over basal levels. We previously showed that the transactivation domain of p35-C/EBP␤ is not stimulated by Ras in this assay (20). Thus, the stimulation by Ras reflects stimulation of the interaction between SRF and p35-C/EPB␤.
When an expression construct encoding the p35-C/ EBP␤(P189G) mutant was transfected with SRF and the reporter construct, there was a 2-fold increase in CAT activity in the absence of Ras and only a 5-fold increase when Ras was cotransfected ( Fig. 1). Therefore, the C/EBP␤ protein that contains a mutation in the MAPK site was no longer responsive to Ras in this assay, and Ras did not stimulate the interaction of the mutant with SRF. The p35-C/EBP␤(P189G) protein was expressed in vivo as detected by Western blotting, and it also bound the SRE in vitro in an electrophoretic mobility shift assay to the same extent as the wild-type protein (data not shown). Therefore, the mutation does not appear to affect protein stability or destroy the ability of the protein to fold cor-rectly as evidenced by DNA binding. We reason that the mutant was unable to respond to Ras in this assay due to the inability of the protein to be phosphorylated by the MAPKs.
Thr 235 of C/EBP␤ Is Necessary for Ras-stimulated Transactivation of the SRE-Due to the fact that mutation of the MAPK site in C/EBP␤ abolishes Ras stimulation of its interaction with SRF, we extended our studies to determine if the MAPK site was also necessary for transactivation of the SRE by C/EBP␤ in response to Ras. To address this question, we used a CAT reporter gene driven by one copy of a TCF mutant SRE upstream of the Rous sarcoma virus long terminal repeat minimal promoter (19). This reporter is unable to bind the TCF family members, which are known targets of Ras. Therefore, this oblates any effect of Ras that could be signaling through the TCF family members.
As shown in Fig. 2, when the TCFmut-CAT reporter construct was transfected together with an expression construct for human C/EBP␤ (NFIL6), there was a 7-fold increase in CAT activity in the absence of Ras. When activated Ras was also cotransfected, this stimulated the CAT activity to 19-fold. This was expected since Ras greatly stimulates the interaction between SRF and C/EBP␤, which would result in enhanced transactivation. Interestingly, the transactivation of a mutant C/EBP␤ construct containing a mutation of Thr 235 in the MAPK site to alanine (NFIL6(T235A)) was not stimulated by Ras. Therefore, mutating the MAPK phosphorylation site in C/EBP␤ abolished the ability of Ras to stimulate transactivation. These data also agree with a previous study showing that Ras does not stimulate the interaction of the NFIL6(T235A) mutant with SRF in the mammalian two-hybrid assay (20). Together, these data demonstrate that C/EBP␤ Thr 235 is required in mediating Ras stimulation of the C/EBP␤ interaction with SRF and transactivation of the SRE.
Analysis of Signaling Pathways Regulating the C/EBP␤-SRF Interaction-We next wanted to analyze the signaling pathways that are downstream of activated Ras in stimulating the interaction between C/EBP␤ and SRF. Ras has many known effectors, but the above data suggest that Ras is stimulating an MAPK pathway to target C/EBP␤. We tested several Ras effectors to determine if they could regulate the C/EBP␤-SRF interaction in the mammalian two-hybrid assay. Fig. 3A shows that when p35-C/EBP␤ was cotransfected with Gal4-SRF and the pG5CAT reporter construct into NIH 3T3 cells in the presence of an activated Ras construct, there was a large increase in CAT activity that was normalized to 100%. As also shown in Fig. 3A, activated constructs of Raf (Raf-CAAX) and PI3K (p110-CAAX), two known Ras effectors, stimulated the interaction of the two proteins. Raf was a stronger activator, stimulating to ϳ70% the level of Ras. PI3K also stimulated the interaction of the proteins, to ϳ35% of the level observed with Ras. There was no significant increase in CAT activity of the reporter alone or the reporter and Gal4-SRF in the presence of activated Raf or PI3K. In the presence of both PI3K and Raf, there was an additive effect on the stimulation over adding either effector alone. This suggests that the Raf and PI3K effectors are activating distinct pathways to stimulate the interaction between SRF and C/EBP␤.
Members of the Rho family of small GTPases are downstream of PI3K, and the SRE is a target of RhoA-, Cdc42-, and Rac1-dependent signaling cascades (15). Stimulation of SRE transactivation by the Rho family has been shown to be SRFdependent, possibly targeting an unknown accessory factor (15)(16)(17). We next tested members of the Rho family to determine if they regulated the interaction between SRF and p35-C/EBP␤ in the two-hybrid assay (Fig. 3B). Activated forms of RhoA (RhoA.V14), Cdc42 (Cdc42.V12), and Rac1 (Rac1(QL)) had no stimulatory effect on the p35-C/EBP␤-SRF interaction. Likewise, dominant-negative Rac1 (Rac1.N17) did not inhibit Ras stimulation. Rho family signaling to the SRE has been shown to be independent of the TCFs (15). These data show that p35-C/EBP␤ is also not a target of the TCF-independent Rho pathway.

MKP-1 Inhibits Ras Stimulation of the C/EBP␤-SRF
Interaction-Because the Ras signaling pathway appears to be working through an MAPK cascade, we tested whether inhibition of the MAPKs would inhibit Ras stimulation of the C/EBP␤-SRF interaction. To address this question, the mammalian two-hybrid assay was again used. As shown in Fig. 4, cotransfection of p35-C/EBP␤, Gal4-SRF, and the Gal4 reporter construct resulted in an average 3-fold increase in CAT activity in the absence of Ras, and Ras stimulated the C/EBP␤-SRF interaction to 95-fold over basal levels.
To determine if the MAPKs play a role in this Ras stimulation, we utilized an MKP-1 construct that can inactivate all three families of MAPKs (25). We tested this construct's ability to inhibit Ras stimulation of the interaction between SRF and p35-C/EBP␤. When MKP-1 was transfected with Gal4-SRF, p35-C/EBP␤, and activated Ras, MKP-1 blocked Ras stimulation from 95-to 18-fold (Fig. 4). These data show that inactivation of the MAPK family members inhibits Ras stimulation of the p35-C/EBP␤-SRF interaction and suggest that MAPKs are working downstream of Ras to target C/EBP␤. Dominant-negative ERK2, but Not ERK1, Blocks Ras Stimulation of the C/EBP␤-SRF Interaction-Since MKP-1 inhibits all three MAPK family members, the next step in our analysis was to inhibit specific MAPKs to determine what effect this would have on the interaction between SRF and p35-C/EBP␤. We therefore tested the requirement for the three MAPK family members in stimulation of the p35-C/EBP␤-SRF interaction by Ras.
We first tested the importance of the ERK family members by utilizing dominant-negative ERK1 and ERK2 constructs in the two-hybrid assay. When the pG5CAT reporter was transfected with Gal4-SRF and p35-C/EBP␤ into NIH 3T3 cells, there was a large stimulation in CAT activity in the presence of activated Ras to 47-fold over basal levels (Fig. 5A). When an expression vector encoding a kinase-inactive and dominantnegative (DN) ERK1 (32) was cotransfected with the above constructs, there was no significant inhibition by the DN ERK1 construct. Therefore, we did not observe an inhibition by DN ERK1, as we observed with MKP-1 shown above (Fig. 4). Interestingly, when DN ERK2 (32) was cotransfected with Gal4-

FIG. 3. Activated forms of Raf and PI3K, but not Rho family members, stimulate the p35-C/EBP␤-SRF interaction in vivo.
A, NIH 3T3 cells were transfected with 1 g of pG5CAT either alone or in the presence of 0.5 g of pGAL4-SRF, 0.5 g of CMV-LAP (encoding p35-C/EBP␤), 1 g of pEXV-Raf-CAAX, 1 g of pSG5-p110-CAAX, or 1 g of pCMV-Ras.V12 as indicated. B, NIH 3T3 cells were transfected with 1 g of pG5CAT either alone or in the presence of 0.5 g of pGAL4-SRF, 0.5 g of CMV-LAP (encoding p35-C/EBP␤), 1 g of pcDNA3-RhoA.V14, 1 g of pEFmyc/V 12 Cdc42hs, 1 g of pcDNA3-Rac1(QL), or 1 g of pCEV-N 17 Rac1 as indicated. Cells in both A and B were serum-deprived for 40 h prior to harvesting, and CAT activity was measured (22). Data are the average of four to five determinations; error bars represent S.E. SRF, p35-C/EBP␤, and Ras, there was a large inhibition of CAT activity, which decreased from 47-to 11-fold. This observed difference between the ERKs was surprising since ERK1 and ERK2 are rarely thought of as having distinct substrates.
Since the dominant-negative constructs provided us are for expression of untagged proteins, expression of the DN ERKs cannot be measured by Western blot analysis. To control for the possibility that the proteins are expressed at different levels, we utilized the fact that both ERK1 and ERK2 stimulate transcriptional activation of the TCF family member Elk-1 (11). We therefore tested the dominant-negative ERK constructs for their ability to inhibit Ras-stimulated Elk-1 transactivation of the SRE. As shown in Fig. 5B, when the wild-type SRE⅐CAT reporter was transfected with an Elk-1 expression construct, there was no increase in CAT activity in the absence of Ras, but Ras stimulated Elk-1 transactivation to 4.5-fold over basal levels. When a DN ERK1, DN ERK2, or an MKP-1 construct was cotransfected with the reporter, Elk-1, and activated Ras, there was a similar extent of inhibition of Ras stimulation by all three constructs, decreasing to about basal levels. These data show that both DN ERK1 and DN ERK2 could inhibit Ras stimulation of Elk-1 transactivation of the SRE to a similar level. The results in Fig. 5 suggest that the p35-C/EBP␤-SRF interaction is being regulated by an ERK2-dependent, but not an ERK1-dependent, pathway.
JNK and p38 Do Not Regulate the C/EBP␤ Interaction with SRF-In addition to testing the ERK family of MAPKs, we also tested the JNK and p38 MAPK family members to determine if they have a role in signaling downstream of Ras to regulate the p35-C/EBP␤-SRF interaction. Fig. 5C shows that neither dominant-negative JNK1 nor SB 202190, a drug inhibitor of p38 kinase, inhibited Ras stimulation of the p35-C/EBP␤-SRF interaction. These data suggest that the JNK and p38 kinases are not acting as downstream effectors of Ras in regulation of the p35-C/EBP␤-SRF interaction.
C/EBP␤ Is an in Vitro Substrate for ERK2, but Not for ERK1-Due to the differing effects of the DN ERK constructs observed in Fig. 5A, we determined whether C/EBP␤ 1) was a direct substrate of the ERKs and 2) would be differentially phosphorylated by purified ERK1 and ERK2 in vitro. Fig. 6 shows that when a bacterial p35-C/EBP␤ protein was incubated with activated ERK1 (lane 1) or ERK2 (lane 2) in the presence of [␥-32 P]ATP and Mg 2ϩ , the C/EBP␤ protein was only phosphorylated by ERK2. Both ERK1 and ERK2 phosphorylated a control substrate, MBP (lanes 3 and 4), showing that both kinases were active. These data show that p35-C/EBP␤ is selectively phosphorylated by ERK2, supporting the two-hybrid data and suggesting that ERK2 is targeting p35-C/EBP␤ to stimulate its interaction with SRF.
Dominant-negative p90 Rsk2 Inhibits Ras Stimulation of the C/EBP␤-SRF Interaction-To further analyze the signaling cascades that regulate the C/EBP␤-SRF interaction, we next tested the requirement for p90 Rsk2 in mediating Ras stimulation in the two-hybrid assay. This effector was an attractive candidate for several reasons. First, p90 Rsk2 is activated by the ERKs, and we have shown that ERK2 is necessary for Ras stimulation of the C/EBP␤-SRF interaction (Fig. 5A). Second, there is evidence that p90 Rsk2 is also activated by the PI3K pathway (10), and we also showed that activated PI3K can stimulate the C/EBP␤-SRF interaction (Fig. 3A). Third, p90 Rsk2 has previously been shown to phosphorylate C/EBP␤ (33) and SRF (28). Phosphorylation of C/EBP␤ by p90 Rsk2 is necessary for hepatocyte proliferation in response to transforming growth factor-␣ (33). The in vivo relevance of SRF phosphorylation by p90 Rsk2 (28) is unknown. Finally, experi-  6. p35-C/EBP␤ is phosphorylated by ERK2, but not by ERK1, in vitro. Histidine-tagged bacterial p35-C/EBP␤ protein (upper panel) was phosphorylated as described under "Experimental Procedures." The protein was then immunoprecipitated with T7 tag antibody-agarose beads, followed by analysis by SDS-polyacrylamide gel electrophoresis. MBP (lower panel) was used as a positive control and phosphorylated identically to p35-C/EBP␤. MBP was not immunoprecipitated, but rather was analyzed directly by SDS-polyacrylamide gel electrophoresis after the kinase reaction. ments using fibroblasts from RSK2 knockout mice strongly implicate RSK2 in regulation of the SRE (34). For these reasons, we tested a dominant-negative p90 Rsk2 construct in the two-hybrid assay to determine if p90 Rsk2 is necessary for Ras stimulation of the C/EBP␤-SRF interaction. Fig. 7A shows that when the pG5CAT reporter construct was cotransfected with Gal4-SRF and p35-C/EBP␤, there was a large increase in CAT activity to 15-fold, as we have observed before. Cotransfection of a dominant-negative p90 Rsk2 construct in which the critical lysine in the amino-terminal kinase domain is mutated to alanine (Rsk2(K100A)) inhibited Ras stimulation to an average of 3-fold over basal levels. Therefore, like ERK2 described above, p90 Rsk2 appears necessary for Ras stimulation of C/EBP␤-SRF interaction.
Since there is evidence that p90 Rsk2 can be a downstream effector of both the Raf/ERK and PI3K pathways, we next determined if the DN p90 Rsk2 construct would inhibit either Raf or PI3K stimulation of the p35-C/EBP␤-SRF interaction in the two-hybrid assay. Transfection of a DN p90 Rsk2 construct inhibited stimulation of the p35-C/EBP␤-SRF interaction by Raf (Fig. 7B), but not by PI3K (Fig. 7C). These data suggest that activation of p90 Rsk2 by Ras to regulate the p35-C/EBP␤ interaction with SRF is mainly through the Raf/ERK pathway.
The Consensus p90 Rsk2 Site in SRF, but Not p35-C/EBP␤, Is Necessary for Ras Stimulation of the Protein-Protein Interaction-As stated above, both SRF and C/EBP␤ have previously been shown to be phosphorylated by Rsk2. SRF is phosphorylated at Ser 103 (28), whereas C/EBP␤ has been shown to be phosphorylated at Ser 105 (33). We obtained constructs of both SRF and C/EBP␤ containing a mutation of the serine in the Rsk2 site to alanine. We tested these mutants in the twohybrid assay to determine if p90 Rsk2 was targeting both SRF and p35-C/EBP␤ to stimulate their interaction. As shown in Fig. 8A, cotransfection of the reporter construct, wild-type Gal4-SRF, and wild-type p35-C/EBP␤ resulted in a large increase in CAT activity in the presence of activated Ras, which was normalized to 100%. When the Gal4-SRFm103 construct, which contains the S103A substitution, was transfected in place of wild-type SRF, there was an inhibition of CAT activity to ϳ60% of the levels attained with the wild-type protein. Fig.  8B shows that there were equivalent amounts of the Gal4-SRF (lane 2) and Gal4-SRFm103 (lane 3) proteins expressed in the NIH 3T3 cells as determined by Western blot analysis. Thus, the decrease in stimulation observed in Fig. 8A was not due to a decrease in the expression of the SRFm103 protein. These data suggest that an intact Rsk2 site in SRF (Ser 103 ) is necessary for complete Ras stimulation of its interaction with p35- C/EBP␤ and that loss of the site results in an inhibition of the activation.
We next tested a C/EBP␤-Ala 105 construct (29), which contains the mutation of the serine residue in the Rsk2 site (Ser 105 ) to alanine, in the two-hybrid assay. Fig. 8C shows that mutation of the Rsk2 site in p35-C/EBP␤ had no significant effect on the ability of Ras to stimulate the mutant protein's interaction with SRF. Therefore, it appears that the p90 Rsk2 kinase is primarily targeting SRF, but not p35-C/EBP␤, to stimulate the interaction of the proteins. DISCUSSION In this report, we have analyzed the signaling pathways that regulate the interaction of the SRF and C/EBP␤ transcription factors. We have demonstrated that Ras signaling pathways are important in regulating the interaction of the two proteins and that Ras is targeting the MAPK site of C/EBP␤ to stimulate its transactivation of the SRE as well as its interaction with SRF in the two-hybrid assay. We have also shown that activated forms of both Raf and PI3K stimulate the interaction of C/EBP␤ and SRF, but that the Rho family of GTPases plays no role. Furthermore, dominant-negative constructs of p90 Rsk2 and ERK2, but not dominant-negative ERK1, inhibit stimulation of the C/EBP␤-SRF interaction by Ras in the mammalian two-hybrid assay. Importantly, C/EBP␤ is an in vitro substrate for ERK2, but not ERK1. Finally, p90 Rsk2 appears to target SRF, but not p35-C/EBP␤, as mutation of the Rsk2 site in SRF inhibits Ras stimulation of the interaction of the proteins.
The Rho family of GTPases has been shown to signal to the SRE by a TCF-independent mechanism (15). This pathway is dependent on SRF, and the DNA-binding domain of SRF was found to be the minimal domain necessary to mediate Rho signaling (16,17). Since C/EBP␤ interacts with the DNA-binding domain of SRF (20), it made an attractive candidate for the unknown target of the Rho pathway. However, in this report, we have shown that activated forms of RhoA, Rac1, and Cdc42 have no effect on the regulation of the C/EBP␤-SRF interaction. Therefore, the Rho family does not seem to be targeting C/EBP␤, at least in stimulating its interaction with SRF. This also agrees with a previous study showing that C/EBP␤ synergizes with Elk-1 in transactivating the SRE in response to Ras (21). Therefore, our studies indicate that C/EBP␤ is not in a suggested TCF-independent signaling pathway activated by the Rho family (15), but rather is working in concert with Elk-1 in response to Ras-dependent signals.
The fact that we saw specific effects of ERK2, in both the ability of dominant-negative ERK2 to inhibit Ras stimulation of the C/EBP␤-SRF interaction and the ability of ERK2 to phosphorylate C/EBP␤ in vitro, was surprising. ERK1 and ERK2 are thought to have overlapping substrates in vitro and in vivo; and therefore, differential phosphorylation by the two kinases was unexpected. Interestingly, a study by Hochholdinger et al. (35) shows that individually, dominant-negative forms of ERK1 and ERK2 inhibit induction of a c-fos promoter reporter construct. This implies that ERK1 and ERK2 are both necessary for c-fos induction. If the kinases were completely redundant, then inhibiting one kinase would not be expected to have an effect on gene induction. Other examples of differences between ERK1 and ERK2 have been reported: the ERK1 knockout mouse has a defect in thymocyte maturation (9); MEK partner-1 binds specifically to ERK1 and not ERK2 (36); v-Raf selectively activates ERK2 in Rat-6 fibroblasts (37); and the urokinase-type plasminogen activator promoter is specifically activated by ERK1 (38). Based on these examples, ERK1 and ERK2 are most likely not completely redundant in vivo. Our studies show another important difference between the ERKs: ERK2, but not ERK1, specifically phosphorylates C/EBP␤ in vitro and is necessary for Ras-stimulated interaction of C/EBP␤ with SRF in vivo.
These studies also suggest a requirement for p90 Rsk2 in Ras stimulation of the p35-C/EBP␤-SRF interaction in vivo. Rsk2(K100A) was a strong dominant-negative in the mammalian two-hybrid system, as effective as DN ERK2. This result was interesting due to the facts that both SRF (28) and p35-C/ EBP␤ (33) are known substrates of p90 Rsk2 , and p90 Rsk2 -dependent pathways have recently been shown to target the SRE (34). We further show that the Rsk2 site in SRF (Ser 103 ) is necessary for the complete Ras stimulation. Although SRF has previously been shown to be a substrate of p90 Rsk2 (28), an in vivo function for this phosphorylation is not known. We propose that a functional consequence of this phosphorylation is to promote interaction with p35-C/EBP␤ in response to growthpromoting signals. We did not observe any significant inhibition of Ras stimulation of the interaction between a C/EBP␤ Rsk2 site mutant (C/EBP␤-Ala 105 ) and SRF. Since the p90 Rsk2 site in C/EBP␤ is not required for Ras stimulation, this suggests that p90 Rsk2 is regulating the interaction of the proteins primarily through phosphorylation of SRF.
We also observed that DN p90 Rsk2 can inhibit stimulation of the p35-C/EBP␤-SRF interaction by Raf, but not by PI3K. Therefore, it appears that the activation of p90 Rsk2 by Ras in our system is through the Raf/ERK pathway. We did see a stimulation of the p35-C/EBP␤-SRF interaction by PI3K, but we currently do not know the kinase(s) targeting the proteins in vivo. We have studied many downstream effectors of PI3K, including Rac (Fig. 3B), Cdc42 (Fig. 3B), and protein kinase B (data not shown), but none of these proteins had a role in regulating the interaction of p35-C/EBP␤-SRF. We are currently examining other potential PI3K effectors in our studies.
We have not yet been able to observe a ternary complex between SRF and p35-C/EBP␤ at the SRE in vitro. A likely explanation for this lies in the complexity of the signaling pathways targeting p35-C/EBP␤. ERK2 and p90 Rsk2 are important downstream effectors of Ras in our studies, and ERK2 phosphorylates p35-C/EBP␤ (Fig. 6), whereas p90 Rsk2 phosphorylates SRF (28). We cannot exclude the possibility that additional phosphorylations of p35-C/EBP␤ and SRF, such as those mediated by downstream effectors of PI3K, are necessary to regulate the interaction of these two proteins. Furthermore, based on the study showing that p35-C/EBP␤ and Elk-1 synergize in Ras-stimulated transactivation of the SRE (21), Elk-1 may also need to be included to observe a quaternary complex of p35-C/EBP␤, SRF, and Elk-1 at the SRE. Therefore, due to the complexity of factors and the essential modifications necessary for interaction, further studies are needed to optimize conditions for a ternary or quaternary complex to be formed in vitro.
Overall, our studies suggest that the formation of a complex between SRF and C/EBP␤ at the SRE is a highly regulated event. Complex formation is dependent on signals from at least ERK2 and p90 Rsk2 and possibly other kinases. These studies, along with a previous study showing that C/EBP␤, Elk-1, and SRF are all necessary for maximal transactivation of the c-fos SRE (21), allow us to propose a stepwise model for SRE activation (Fig. 9). In a quiescent cell, the SRF and inactive Elk-1 proteins are constitutively bound at the SRE, based on in vivo footprinting analysis (6). Footprinting studies also indicate occupancy 3Ј of SRF, so either an inactive p35-C/EBP␤ isoform or the repressor p20-C/EBP␤ isoform could be present since both isoforms could interact with SRF in the absence of mitogenic signaling by Ras (20). However, in response to Ras-dependent signaling pathways, either ERK1 or ERK2 or both phosphorylate Elk-1, whereas p35-C/EBP␤ is phosphorylated by ERK2. p90 Rsk2 also phosphorylates SRF. Other kinases may also target these transcription factors as well, such as kinases downstream of PI3K. Thus, the convergence of these growth-promoting signals on factors at the SRE is necessary for a competent transcriptional complex to form and to allow maximal induction of c-fos.