ERM transactivation is up-regulated by the repression of DNA binding after the PKA phosphorylation of a consensus site at the edge of the ETS domain.

The final step of the transduction pathway is the activation of gene transcription, which is driven by kinase cascades leading to changes in the activity of many transcription factors. Among these latter, PEA3/E1AF, ER81/ETV1, and ERM, members of the well conserved PEA3 group from the Ets family are involved in these processes. We show here that protein kinase A (PKA) increases the transcriptional activity of human ERM and human ETV1, through a Ser residue situated at the edge of the ETS DNA-binding domain. PKA phosphorylation does not directly affect the ERM transactivation domains but does affect DNA binding activity. Unphosphorylated wild-type ERM bound DNA avidly, whereas after PKA phosphorylation it did so very weakly. Interestingly, S367A mutation significantly reduced the ERM-mediated transcription in the presence of the kinase, and the DNA binding of this mutant, although similar to that of unphosphorylated wild-type protein, was insensitive to PKA treatment. Mutations, which may mimic a phosphorylated serine, converted ERM from an efficient DNA-binding protein to a poor DNA binding one, with inefficiency of PKA phosphorylation. The present data clearly demonstrate a close correlation between the capacity of PKA to increase the transactivation of ERM and the drastic down-regulation of the binding of the ETS domain to the targeted DNA. What we thus demonstrate here is a relatively rare transcription activation mechanism through a decrease in DNA binding, probably by the shift of a non-active form of an Ets protein to a PKA-phosphorylated active one, which should be in a conformation permitting a transactivation domain to be active.

The regulation of gene expression by specific signal transduction pathways is tightly connected to cell phenotype. Several molecules involved in intracellular signaling are encoded by oncogenes, which directly link their potential aberrant expression to cell transformation or altered proliferation. The final step of the transduction pathway is the activation of nuclear transcription factors. For example, the cyclic-AMP-and calcium-regulated nuclear factor is activated through the protein kinase A (PKA) 1 pathway via phosphorylation. The differential phosphorylation of transcription factors by signal transduction pathways such as the mitogen-activated protein kinase (MAPK) plays a crucial role in the regulation of gene expression. This is the case for the c-fos gene expression, which is regulated by the binding to DNA of the Elk/TCF factor after phosphorylation through the MAPK pathway. The activation of MAPK cascades leads to changes in the activity of many Ets factors such as Elk/TCF (for reviews, see Refs. 1 and 2).
The Ets family of transcription factors, which includes more than 30 members from sponges to humans (for reviews, see Refs. 2 and 3), has been involved in both tumorigenesis and a number of developmental processes. They all contain the ETS domain (4), a domain of 85 amino acids structured as a winged helix-turn-helix structure and responsible for DNA inding to the specific core sequence GGA(A/T) (for review, see Refs. 2). These factors can be subclassified primarily because of the amino acid conservation in their ETS domains and also because of the conservation of other domains generally characterized as transactivating.
In the case of the PEA3 group, which is made up of three members, PEA3 (also called E1AF in the human or ETV4) (5,6), ER81 (also called ETV1 in the human) (7)(8)(9), and ERM (also called ETV5) (10 -12), these factors are more than 95% identical in the ETS domain and more than 85% in the NH 2 -terminal 32 residue acidic domain, and almost 50% identical in the final 61 residues corresponding to the carboxyl-terminal tail of the proteins (carboxyl-terminal domain) (for review, see Ref. 13). The transactivating activity of these factors is because of the two conserved acidic and carboxyl-terminal domains (11, 14 -16).
Although the putative target genes of these three transcription factors are multiple, their most frequently studied role concerns their involvement in breast cancer metastasis (17,18), probably by activating the transcription of matrix metalloproteinases (19,20). Concerning their expression in the adult, although pea3/e1af and er81/etv1 display more restricted expression patterns (5,(7)(8)(9)(10)12), erm has been characterized as a more ubiquitously expressed gene with its highest expression in the brain (10). Recently, it has been demonstrated that ERM is induced by interleukin-12 through a Stat4-dependent pathway in Th1 and could play a highly specific role by inducing Th1 differentiation in the mouse (21).
The transcription capacities of mouse and zebrafish PEA3, mouse and human ER81/ETV1, and human ERM are increased by components of the MAPK cascades, Ras, Raf-1, MEK, and the MAPK ERK-1 and ERK-2; this suggests that these factors may contribute to the nuclear response to cell stimulation and Ras-induced cell transformation (15,16,22,23). Interestingly, PKA is also able to increase the transcriptional activity of human ERM (15), human ETV1 (24), and zebrafish PEA3 (23) probably through phosphorylation since we have previously demonstrated that ERM can be phosphorylated in vitro by active PKA (15). In this study we show that, in contrast to human ERM and ETV1, PKA is not able to significantly increase the transcriptional activity of mouse PEA3. In fact, this transcription factor lacks a classical PKA phosphorylation site present at the beginning of the ETS domain of ERM and ETV1; this domain is phosphorylated by PKA, and is necessary for the activation of ERM by the kinase. We further show that the increase in ERM transcriptional activity after PKA phosphorylation is closely correlated with a drastic reduction in the DNA binding of the transcription factor. These results indicate that the phosphorylation of ERM by PKA is involved in ERMmediated transcription and suggest that the activation of ERM is probably related to conformational changes.

EXPERIMENTAL PROCEDURES
Cell Lines-Rabbit kidney epithelial-like RK13 cells (ATCC number CCL 37), HeLa and COS-7 cells were maintained at 37°C in a humidified atmosphere of 5% CO 2 in Dulbecco's medium supplemented with 10% fetal calf serum (Invitrogen).
Plasmids-The 3xE74-tk-Luc reporter plasmid and the overexpression vectors for the catalytic subunit of PKA have been described previously (15,25). The plasmid Ϫ517/ϩ63 Coll-Luc containing the indicated region of the human collagenase 1 promoter sequence linked to a luciferase gene (20) has been described.
The full-length region of human ERM (10), human ETV1 (8), and mouse PEA3 (5) were amplified by PCR, digested, and subcloned in the EcoRI and BglII sites of the pSV vector, an expression vector which provides an amino-terminal hemagglutinin tag (14), to produce pSV-ERM, pSV-ETV1, and pSV-PEA3, respectively. The point mutations of ERM were generated by the double PCR method: oligonucleotide 5Ј-AGAGGCGAGGTNNNCTTCAGCTGT-3Ј where NNN corresponded to GCC, GAT, and GAG and its antisense counterpart, generated a specific mutation of Ser 367 into alanine, aspartic acid, and glutamic acid, respectively. The final PCR fragments were subcloned as SmaI-BstEII fragments into the SmaI and BstEII sites of pSV-ERM yielding pSV-ERM S367A, pSV-ERM S367D, and pSV-ERM S367E. The ERM mutant containing a deletion of amino acids 226 -250 was constructed by the double PCR method using oligonucleotides 5Ј-CAGAGACAACT-GTCTGAACCT-3Ј (nucleotides 849 -870) and 5Ј-TTTCAGAGACAACT-GTCTGAACCTATTGTC-3Ј (nucleotides 922-951); the underlined nucleotides indicate changes in nucleic acid bases to optimize overlapping in the second round of PCR. The PCR fragment was cloned as a BglII-SmaI fragment into the BglII-SmaI sites of the pSV-ERM and pSV-ERM S367A to generate pSV-ERM ⌬ and pSV-ERM ⌬ S367A, respectively. The full-length region of ERM wild-type and ERM S367A was also subcloned in the EcoRI and BamHI sites of the pEFIN 3 vector to produce pEFIN-ERM and pEFIN-ERM S367A, respectively.
GAL4-ERM fusions were made by cloning parts of the human ERM coding region in-frame with the DNA-binding domain of yeast GAL4-(1-147) in pGAP (26). The 5xGAL4 Luc plasmid containing five GAL4binding sites upstream of the luciferase gene has been described earlier (14). All the mutations and PCR products were verified by dideoxy sequencing.
Transactivation Analyses-RK13 cells grown in 6-well plates were transfected in triplicate using the PEI Exgen 500 procedure (Euromedex, France) with a total of 300 ng of DNA, including 50 ng of reporter plasmid, 50 ng of pSV expression vectors and, when mentioned, 150 ng of PKA expression vector. Luciferase activity was determined 24 -48 h after transfection and normalized to transfection efficiency by assaying for ␤-galactosidase activity expressed from 50 ng of co-transfected Rous sarcoma virus-␤-galactosidase plasmid. Typically, the results are given as a mean Ϯ S.E. of three independent experiments. HeLa cells were transfected using the same procedure except that 100 ng of pEFIN 3 expression plasmid (Euroscreen, Belgium) was used with a total of 350 ng of DNA. COS-7 cells were transfected using the same procedure with a total of 1 g of DNA including 150 ng of reporter plasmid, 300 ng of pSV expression vectors and, when mentioned, 150 ng of a pSV-PKA expression vector.
Labeling of Transfected Cells and Immunoprecipitation-The cells were allowed to grow for 40 h before the labeling experiments were performed. Transiently transfected cells were placed into Met/Cys-free or phosphate-free Dulbecco's modified Eagle's medium for 2 h. 100 Ci of [ 35 S]Met/Cys per ml or 1 mCi of [ 32 P]orthophosphate per ml were added to the medium and left for 3 h. At the end of the labeling period the cells were washed with phosphate-buffered saline and lysed in an ice-cold radioimmunoprecipitation assay (RIPA) buffer containing 1 mM phenylmethylsulfoxyl fluoride, 1% (v/v) aprotinin, 2 mM Na 3 VO 4 , 40 mM ␤-glycerophosphate, and 25 mM pyrophosphate. The cells debris was centrifuged and the supernatant subjected to immunoprecipitation using anti-ERM antibody (15,17) and protein A-Sepharose CL-4B (Amersham Bioscience, Inc.) overnight at 4°C. After washing and elution, a second immunoprecipitation step was performed with the anti-hemagglutinin antibody as described previously (17). The immunocomplexes were then eluted in Laemmli sample buffer and analyzed by SDS-PAGE and autoradiographed.
Protein Expression-GST-ERM fusion proteins were expressed in bacteria and purified, essentially as described (17), using glutathione-Sepharose beads (Amersham Bioscience, Inc.). The concentrations of GST fusion proteins were estimated by SDS-PAGE analysis. In vitro transcription-translation reactions were performed using the TNT-coupled reticulocyte lysate system (Promega) with 1 g of DNA template and [ 35 S]methionine. One l of the reaction products was separated by electrophoresis followed by autoradiography.
In Vitro Phosphorylation and Dephosphorylation-The in vitro phosphorylation by PKA was performed for 1 h at 30°C using the catalytic subunit of PKA from a bovine heart (Sigma) in a reaction buffer containing 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 12 mM MgCl 2 , and 1 mM dithiothreitol in the presence of GST-ERM fusion proteins or the products of in vitro translation reactions. Assays were carried out in the presence of either non-radiolabeled ATP (EMSA) or [␥Ϫ 32 P]ATP. The specific inhibition of the PKA activity was performed by adding 1 g of the peptide inhibitor PKI (Sigma) to the reaction buffer. The dephosphorylation of in vitro translated ERM with bacterial alkaline phosphatase was carried out in 20 mM Tris-HCl, pH 7.5, 12 mM MgCl 2 , 100 mM NaCl for 1 h at 30°C.
Electrophoretic Mobility Shift Assay (EMSA)-The effect of phosphorylation and dephosphorylation on DNA binding by ERM protein was tested by first treating the ERM proteins as described above. A portion of the reaction was mixed with 1 ng of the 32 P-labeled E74 probe (sense strand: 5Ј-GAGCTGAATAACCGGAAGTAACTCAT-3Ј) or the 32 P-labeled Coll1 probe (sense strand: 5Ј-CTAATCAAGAGGATGTTATA-AAGC-3Ј) in the presence of 25 mM Hepes, pH 7.9, 50 mM KCl, 5 mM MgCl 2 , 0.5 mM EDTA, 10% glycerol, 1 mM dithiothreitol, and 1 g of poly(dI-dC). The mixture was incubated for 1 h at room temperature and loaded onto a 6% polyacrylamide gel. The gel was run at 4°C in a 0.5 ϫ Tris borate EDTA buffer at 180 V. EMSA were also performed using nuclear extracts made by standard procedures (27). The specificity of the DNA-protein interaction was established by supershift assays using a goat polyclonal antibody generated against the carboxyl-terminal 20 amino acids of human ERM (Santa Cruz Biotechnology, Santa Cruz, CA).
Western Blot Analysis-COS-7 cells were lysed in SDS sample buffer and total cell lysates were subjected to SDS-PAGE. Proteins were transferred to polyvinylidene difluoride membranes (Millipore), and the blots were probed with anti-ERM antibodies and horseradish peroxidase-conjugated secondary antibodies. Immune complexes were visualized by enhanced chemiluminescence.

RESULTS
Activation of the PEA3 Group Members by PKA-Although human ERM and ETV1 are targeted by the PKA signaling pathway (15,24), no information is available concerning the transactivation of mammalian PEA3 through PKA activation.
We therefore began by comparing the effect of the PKA pathway on the transactivation capacities of human ERM, human ETV1, and mouse PEA3 in transient transfection experiments. Based on previous experiments showing that the E74 reporter plasmid is not activated by PKA in RK13 cells (15,24), we used this system to compare the activation of the PEA3 group members by the kinase. In the absence of a PKA expression vector none of the three proteins was found to stimulate the transactivation of the E74 reporter plasmid (Fig. 1). In contrast, the expression of these transcription factors led to a 30 -70% transcription repression when compared with the basal level. The co-transfection of a PKA expression vector did not increase the basal transactivation on the E74 reporter. However, as previously reported, the expression of PKA significantly increased transactivation when ETV1 or ERM were coexpressed. Concerning PEA3, the coexpression of PKA only abolished the repression on the reporter since no significant increase in transactivation was observed as compared with the basal level. These data clearly show that of the three mammalian PEA3 group members, only ERM and ETV1 are targeted by the PKA pathway on the E74 reporter.
Mapping of the Major ERM Phosphorylation Sites-As previously described (15), ERM is phosphorylated in vitro by purified PKA. The examination of the ERM amino acid sequence for phosphorylation consensus sites by known kinases revealed a single consensus PKA phosphorylation site (RRGS 367 ) in a region located at the beginning of the DNA-binding ETS domain. Furthermore, the amino acid sequence of the region including this site is well conserved among the PEA3 group members of different species. Interestingly, only mouse and human PEA3 contain an alanine substitution at the putative serine phosphoacceptor site (RRGA, Fig. 2A). Since PKA is also able to putatively phosphorylate other sites present in the protein (RXXS or RXS (28)), we first determined the regions of ERM that are phosphorylated in vitro by this kinase. We thus expressed human ERM (aa 12-510) and mutants lacking portions of the NH 2 or COOH termini as GST fusion proteins in Escherichia coli (Fig. 3A). Partially purified proteins were then tested for their ability to serve as substrates for purified PKA (Fig. 3B).
The results indicate that phosphorylation occurred at the COOH-terminal end of ERM since ERM-(354 -510), which con-tains the consensus PKA phosphorylation site, was an excellent substrate for the kinase (Fig. 3C). However, the proteins of the 363 and 298 residues obtained by the COOH-terminal truncation of ERM were also efficiently phosphorylated by the kinase, whereas further COOH-terminal truncation of ERM to a protein with 226 residues led to a severe reduction in phosphorylation. Accordingly, when ERM-(199 -283) was used as a substrate, there was efficient phosphorylation of the fusion protein, and only the GST moiety was not phosphorylated (data not shown). Thus, by deletion analysis, the PKA phosphorylation domains of ERM were located both at the COOH-terminal region including the consensus PKA phosphorylation site (Ser 367 ), and at a central region between residues 226 and 298. This latter domain contains two contiguous potential sites for PKA phosphorylation (Arg-Pro-Ser 242 -Tyr-His-Arg-Gln-Met- Ser 248 ), the first being specific to human and mouse ERM and the second shared by ERM and ETV1 (Fig. 2B).
To specify the PKA phosphorylation sites, mutations within the mapped regions were generated in the GST-ERM protein, i.e. GST-ERM S367A, in which the serine indicated was mutated into an alanine, GST-ERM ⌬, in which the 226 -250 region was deleted and GST-ERM ⌬ S367A, which combined the two mutations. When compared with GST-ERM, a significant decrease in 32  Mutation of Ser 367 Phosphorylation Site Reduces ERM Transcriptional Capacity by PKA-To analyze the functional role of phosphoacceptor sites in the activity of ERM protein we assessed whether the mutated proteins were still transcriptionally active in the presence of PKA. In the absence of the PKA expression vector, the co-transfection of RK13 cells with the E74 reporter and plasmids expressing either wild-type ERM or mutant ERMs (ERM S367A, ERM ⌬, and ERM ⌬ S367A) resulted in similar levels of luciferase activity, which were slightly lower than the basal level (Fig. 5A). In contrast, when the catalytic subunit of PKA was coexpressed, wild-type ERM gave a 7.4-fold transcription activation in relation to ERM alone, which was very similar to that seen with ERM ⌬ (7.7-fold activation). However, a significantly lower level of transcription activation was observed when Ser 367 was mutated into alanine, i.e. in cells expressing ERM S367A (3.2-fold activation) and ERM ⌬ S367A (2.3-fold activation). Under the conditions used for the transcriptional assays it was very difficult to determine the expression level of the wild-type and mutant ERM proteins. Nevertheless, an immunoprecipitation analysis of 35 S-radiolabeled RK13 cells co-transfected with higher concentrations of the ERM effectors demonstrated that the Ser 367 mutation did not affect the relative expression level of the mutated ERM protein compared with the expression level of the wild-type protein in either the absence or presence of coexpressed PKA (Fig. 5B). Taken together, these data indicate that whereas mutating Ser 367 to alanine results in a protein with a reduced transcriptional enhancement capacity by PKA, the PKA phosphorylation site located in the central part of the protein does not affect ERM in response to the kinase. Furthermore, immunoprecipitation assays were also performed on cotransfected cells labeled with [ 32 P]orthophosphate to investi- gate in intact cells whether the expression of PKA alters ERM phosphorylation. As shown in Fig. 5C, the incorporation of [ 32 P]orthophosphate into immunoprecipitated wild-type ERM or ERM S367A was observed in cells in the absence of PKA. Concerning wild-type ERM this incorporation was, however, increased in cells co-transfected with the PKA catalytic subunit as compared with cells in the absence of active kinase. In contrast, the phosphorylation of ERM S367A was relatively unaffected in the cells expressing the kinase. These results therefore strongly suggest that ERM is the subject of PKA phosphorylation in intact cells and that the activation of ERM transcriptional activity in the presence of PKA is associated with the phosphorylation of Ser 367 .
The PKA-dependent activation of ERM was also tested on another Ets-responsive reporter construct, the human collagenase 1 (Coll1 (20)). Neither wild-type ERM nor ERM S367A was able to activate the promoter in RK13 cells and coexpression of PKA resulted in induction of promoter activity by wildtype ERM (about 10-fold) whereas ERM S367A was less efficient in activating the promoter under these conditions (about 5-fold, Fig. 6). Thus, while Ets proteins are described as weak activators of transcription, ERM was unable to activate the E74 and Coll1 promoters in the absence of PKA in RK13 cells. However, since the transcriptional activity of ERM has been reported to behave differently depending of the cell type, we consequently examined the transcriptional activation of ERM on the E74 and Coll1 promoters in another cell line, i.e. HeLa cells. In contrast to the results obtained in RK13 cells, expression of wild-type ERM induced an increase in the activity of the promoters used. While the luciferase activity was slightly above background level for the E74 promoter (about 2-fold activation), activation of the Coll1 promoter was significantly higher (about 7-fold, Fig. 6). This activation, absent in RK13 cells, was not linked to the endogenous level of PKA in HeLa cells since the induction effect of exogenous ERM was unaffected when the transfected cells were treated with the PKA inhibitor H89 (data not shown). Accordingly, similar ERM-dependent transactivation was obtained with wild-type and S367A-mutated molecules. When PKA was expressed, transcription in the absence of exogenous ERM was not significantly modified indicating that endogenous Ets proteins do not affect the activity of the promoters in the presence of the kinase. However, PKA expression led to a marked transactivation increase by the wild-type ERM (about 15-and 12.5-fold activation on the E74 and Coll1 reporters, respectively) and the magnitude of this activation was lower for the ERM S367A mutant protein (about 4-and 5.5-fold activation on the E74 and Coll1 reporters, respectively). These observations are therefore consistent with a PKA-dependent phosphorylation effect on ERM transcriptional activity in two different reporter systems, and argue that the ERM residue Ser 367 is a PKA target in cells.
PKA Does Not Directly Affect the ERM Transactivation Domains-ERM was shown to contain two transactivation domains: the NH 2 -terminal part of the protein including the acidic domain (amino acids 1-72), and the last 61 residues of the protein (14). We had previously reported that the deletion of the COOH-terminal transactivation domain did not affect the PKA-dependent activation of ERM. In contrast, the deletion of the NH 2 -terminal transactivation domain nearly eliminated the transcriptional activation of the protein in the presence of PKA (15). We therefore investigated whether the ERM NH 2 -terminal transactivation domain was capable of responding to PKA when fused to a heterologous DNA-binding domain, i.e. that of the yeast protein GAL4. We consequently fused full-size ERM and various Ct-truncated mutants to the GAL4 DNA-binding domain and tested the transcriptional activity in the context of a luciferase reporter construct driven by GAL4binding sites. A marked increase in basal transcription in the RK13 cells was observed with the GAL4-(1-72) and GAL4-(1-122) constructs. However, the expression of PKA exhibited no significant effect on transactivation (Fig. 7A). This agrees with the fact that no PKA phosphorylation site has been identified in the NH 2 -terminal region. As reported before, a dramatic reduction of activation was observed from GAL4-(1-181) to GAL4-(1-370), a phenomenon probably because of the presence of a negative regulatory domain in the region from residues 166 to 326 (11). Again, the coexpression of PKA did not affect the activation potential of these fusion proteins even in the case of GAL4-(1-370), which contains the Ser 367 phosphorylation site. Finally, GAL4-(1-510), which is transcriptionally inactive, was also unresponsive to PKA (Fig. 7A). However, when this fusion protein of the GAL4 DNA-binding domain and full-length ERM was assayed with the Coll1 reporter, it was active and a PKAmediated activation was observed. Furthermore, mutation of Ser 367 into Ala in the fusion protein resulted in a large decrease in transcription activity only when PKA was expressed in the transfected cells (Fig. 7B). It thus seems likely that the ERM NH 2 -terminal transactivation domain is not directly targeted by PKA, and that the kinase is unable to activate ERM in the context of GAL4 fusion proteins when tested with a GAL4binding site-driven reporter.
PKA Phosphorylation Affects the DNA Binding Activity of ERM-Because of the fact that the functional PKA phosphorylation site of ERM is situated at the edge of the ETS domain, we wondered whether PKA phosphorylation alters the capacity of ERM to interact with DNA. For this study we used gel mobility shift assays with the full-length GST-ERM protein, which was first subjected to different kinase reactions and then used for DNA-binding assays. As in the absence of PKA (Fig.  8A, lanes 1-2), in the presence of PKA but in the absence of ATP (lane 3), ERM was able to interact with an E74 oligonucleotide probe. However, the binding of ERM was nearly eliminated (lane 4) upon the addition of ATP and PKA, and addition of the PKA inhibitor PKI to the phosphorylation reaction eliminated the effect of PKA on the DNA binding activity of the wild-type ERM (lane 5). It must be noted that a remaining DNA binding was observed for the PKA-phosphorylated ERM when a higher amount of protein was used (Fig. 8B). In the two remaining ERM-DNA complexes detected, the faster migrating complex most probably corresponded to the remaining unphosphorylated ERM. On the other hand, the slowly migrating complex was specifically observed after the PKA phosphorylation, so indicating that phosphorylation may have induced a conformational change in ERM.
These results thus strongly suggest that phosphorylation causes the large decrease in DNA binding activity. However, it was not possible to deduce from our in vitro analysis whether phosphorylation at Ser 367 was solely responsible for the change in DNA binding activity since in vitro, PKA phosphorylates another site in the central part of ERM. For this reason we examined whether phosphorylation by PKA was able to alter the capacity of ERM S367A to interact with DNA. As shown in Fig. 8A the incubation of GST-ERM S367A with PKA and ATP did not impair the DNA binding activity of the mutated protein as compared with the other corresponding conditions used (compare lane 9 to lanes 6 -8 and 10). In agreement with this observation, whereas phosphorylation of GST-ERM ⌬ eliminated its DNA binding activity, the DNA-binding capacity of GST-ERM ⌬ S367A remained unaffected by PKA treatment (data not shown). These results indicate that the consensus PKA-dependent phosphorylation site located at the beginning of the DNA-binding domain plays a crucial role in modulating ERM DNA binding activity on an E74 probe. Furthermore, as illustrated in Fig. 8C for a collagenase 1 probe, similar results were obtained with other Ets-binding sites.
To ensure that the variation observed in DNA binding was not because of the GST portion of the fusion proteins we also performed experiments with ERM synthesized in rabbit reticulocyte lysate. As for the GST fusion protein, preincubation with PKA and ATP nearly eliminated the DNA binding activity of wild-type ERM (Fig. 8D, compare lanes 1 and 3), while the binding of ERM S367A was apparently unchanged (compare lanes 2 and 4). It must, however, be noted that we always observed a lower degree of DNA binding in the case of wild-type ERM when compared with mutated ERM. This difference in DNA binding of the reticulocyte proteins was not because of less wild-type ERM because the amounts of protein used were equalized through normalization to [ 35 S]methionine incorporation (data not shown). Since this difference was not detected with bacterially produced proteins, this finding suggests that DNA binding might be modulated by Ser 367 phosphorylation in reticulocytes. To test this hypothesis the two reticulocyte proteins were treated with alkaline phosphatase prior to gel shift assays. As illustrated in Fig. 8E, whereas dephosphorylation treatment did not significantly alter binding by the serine to alanine mutant (compare lanes 2 and 4), it enhanced the DNA binding of the wild-type protein (compare lanes 1 and 3), which appeared similar to that of the mutant protein. This observation strongly suggests that the S367A mutation does not interfere with DNA binding, and that the phosphorylation of Ser 367 by a protein kinase present in reticulocytes is responsible for a slight inhibition of the DNA binding in reticulocyte-translated wild-type ERM.
EMSAs were finally performed to determine whether PKA regulated binding of ERM in transfected cells. Comparison was made using equal amounts of nuclear extracts derived from COS-7 cells transfected with expression vectors encoding either wild-type ERM or ERM S367A alone or together with a PKA expression plasmid (Fig. 9). In comparison to control cells (lane 1, Fig. 9A), a band was detected in nuclear extracts of cells transfected with wild-type ERM or ERM S367A alone (lanes 2 and 3) and these complexes bound to the E74 probe were supershifted by the addition of an ERM antibody (Fig. 9B). However, when PKA was co-expressed in transfected cells, the binding of wild-type ERM was significantly reduced (lane 5, Fig. 9, A and B) whereas that of ERM S367A was unaffected (lane 6, Fig. 9, A and B). This result was not because of less wild-type ERM in cells co-transfected with PKA since immunoblot analysis revealed equivalent levels of protein expression for wild-type ERM and ERM S367A in PKA-transfected cells (Fig. 9C). Taken together, all the results are consistent with the notion that PKA-mediated phosphorylation of Ser 367 acts to negatively regulate the DNA binding activity of ERM.
DNA Binding Capacity of the PEA3 Group Members Phosphorylated by PKA-The above results indicate that Ser 367 phosphorylation by PKA changes the DNA binding capacity of ERM. Since the site for PKA phosphorylation identified in ERM was present at the same location in human ETV1 but absent in human PEA3 ( Fig. 2A), we evaluated the possible effects of PKA phosphorylation on the DNA binding of these other two PEA3 group members. To this end full-length ERM, ETV1, and PEA3 were synthesized in rabbit reticulocyte lysate, treated with or without purified PKA and subjected to gel shift assays. ETV1 and PEA3 bound to the E74 probe after treatment without PKA (Fig. 10A, left-hand panel) as did ERM. It must, however, be observed that PEA3 bound to DNA more avidly than ERM or ETV1. This difference was not because of more PEA3, since the experiment was performed with roughly equal amounts of the proteins (Fig. 10B). Furthermore, a significant difference in DNA binding was still observed after alkaline phosphatase treatment of the three reticulocyte-translated proteins, so indicating that the lower binding capacity of ERM and ETV1 as compared with PEA3 cannot be explained by the phosphorylation of these two proteins in reticulocytes (data not shown). After in vitro PKA treatment, similar results were obtained for ERM and ETV1, i.e. phosphorylation by the kinase eliminated the DNA binding. On the other hand, the DNA binding of PEA3 following PKA treatment was similar to that of the untreated protein (Fig. 10A, right-hand panel). It thus appears that ERM and ETV1, both of which are transcriptionally activated by PKA, show altered properties in DNA binding following in vitro PKA phosphorylation whereas PEA3, which is not significantly transcriptionally activated by PKA, is not affected in its DNA binding after PKA treatment.
Effect of Mutation of Ser 367 into Acidic Residues on ERM Properties-Since the phosphorylation of Ser 367 influences the capacity of ERM DNA binding activity, we employed site-directed mutagenesis to change this phosphoacceptor site into either aspartic acid or glutamic acid, which may mimic a phosphorylated serine. The mutated cDNAs encoding full-length ERM S367D and ERM S367E were transcribed and translated in vitro, and the mutant proteins were examined for DNA binding. In contrast to the result obtained for ERM S367A (Fig.  11A, lane 2), we found that the substitution of serine for glutamic acid converted ERM from an efficient DNA-binding protein into a poor DNA binding one (compare lanes 1 and 3). Similar data were obtained with a mutated protein created by FIG. 8. The effect of the in vitro phosphorylation by PKA of ERM on DNA binding activity. A, bacterially produced GST-ERM-(12-510) and GST-ERM-(12-510) S367A were treated in the presence (ϩ) or absence (Ϫ) of purified PKA with or without ATP or the PKA inhibitor PKI, and tested in gel mobility shift assays using an E74 double strand oligonucleotide probe. B, GST-ERM-(12-510) was treated and analyzed as in A except that the experiment was performed with double the amount of proteins. C, bacterially produced GST-ERM-(12-510) and GST-ERM-(12-510) S367A were treated in the presence (ϩ) or absence (Ϫ) of purified PKA and tested in gel mobility shift assays using a collagenase 1 double strand oligonucleotide probe. D and E, full-size ERM and its S367A mutant were produced in reticulocyte lysate, incubated in the presence (ϩ) or absence (Ϫ) of PKA (D) or alkaline phosphatase (E) and tested in gel mobility shift assays using an E74 probe. inserting aspartic acid in place of serine (data not shown), and the variation observed in binding between wild-type and mutant proteins was not because of differences in protein amounts (Fig. 11B). When the proteins were submitted to PKA phosphorylation before the gel shift assays, no significant change in DNA binding was observed for the two mutated proteins ERM S367A and ERM S367E, while the binding of wild-type ERM was nearly eliminated (Fig. 11A). This confirms that mutating Ser 367 into non-phosphorylatable residues results in mutated proteins unresponsive to PKA in terms of DNA-binding capacity and clearly shows that inserting acidic residues in place of Ser 367 decreases ERM DNA binding, but not to the same extent as after the PKA-dependent phosphorylation of the wild-type protein.
We also tested whether the substitution of serine for glutamic acid in ERM led to changes in the transcriptional activation of the E74 reporter in RK13 and HeLa cells. In the absence of the PKA expression vector, the expression of ERM S367E led to a weak activation of transcription as compared with wild-type ERM (Fig. 11C). Further activation was observed in cells co-transfected with the PKA expression vector. However, the degree of activation in the presence of PKA was lower than that obtained for wild-type ERM, but greater than that observed for ERM S367A (compare Fig. 11C and Fig. 6). Similar results were also obtained with ERM S367D (data not shown). These results indicate that: 1) the substitution of serine for glutamic or aspartic acid leads to an increase in transcriptional activity for the mutated proteins in the absence of PKA, but is not sufficient to mimic serine phosphorylation; and 2) the variations in ERM-mediated E74 activation by PKA are closely correlated with changes in DNA binding activity. However, as illustrated in Fig. 11C, mutation of Tyr 419 to Pro in the ETS domain of ERM, which abolishes DNA binding (14), generated a protein unable to activate transcription in the presence or the absence of PKA. This indicates that ERM binding is a prerequisite for transcriptional activation by PKA. DISCUSSION In the present investigation, we confirm that ERM is targeted by the PKA pathway, and we show that of the two other known mammalian members of the PEA3 group of Ets-related transcription factors, i.e. ETV1 and PEA3, only ETV1 is significantly activated transcriptionally by PKA. Our results indicate that Ser 367 plays a crucial role in the transcriptional activation of ERM by PKA inasmuch as its mutation into nonphosphorylatable alanine significantly reduces ERM-mediated transcription in the presence of the kinase. This residue, which is in a favorable consensus sequence, is phosphorylated by PKA in vitro as well as in vivo and, interestingly, the corresponding region in mouse ETV1 is also highly phosphorylated by PKA in vitro (16). In fact, this region of the protein is well conserved in the three PEA3 group members. However, while the RRGS 367 motif of this region is conserved in ERM and ER81/ETV1, a substitution of serine for alanine is found in mouse and human PEA3, thus strongly suggesting that the presence of this site is necessary for the transcriptional activation of PEA3 group members by PKA. In support of this conclusion there is the recent observation that zebrafish PEA3, which contains this consensus site along with the serine residue, is activated transcriptionally in the presence of PKA (23).
We show here that the phosphorylation of ERM by PKA FIG. 10. The effect of the in vitro phosphorylation by PKA of PEA3 group members on DNA binding activity. A, full-length ERM, ETV1, and PEA3 were produced in reticulocyte lysate, incubated in the absence (Ϫ) or presence (ϩ) of PKA and tested in mobility shift assays using an E74 probe. B, the electrophoretic analysis of radiolabeled proteins used in A. Note that the in vitro translated proteins give a two-band pattern because of internal methionine initiation: the slowly migrating bands are the hemagglutinin-tagged full-length proteins, while the faster migrating bands represent non-tagged full-length proteins.
FIG. 11. The effects of Ser 367 for glutamic acid substitution on ERM properties. A, pSV plasmids encoding wild-type ERM, ERM S367A, or ERM S367E were transcribed and translated in vitro. The reticulocyte proteins were then incubated in the presence or absence of PKA and used in gel retardation assays with an E74 probe. B, the electrophoretic analysis of radiolabeled proteins used in A. C, the expression plasmids encoding wild-type ERM, ERM S367E, or ERM Y419P were co-transfected in RK13 and HeLa cells with the reporter plasmid 3xE74-tk80-Luc with or without a PKA expression plasmid as described in the legend to Fig. 6. greatly decreases the amount of protein bound to Ets-binding sites. This effect is due specifically to the phosphorylation of Ser 367 since no change in DNA binding activity is observed after the phosphorylation of ERM S367A. Since the phosphorylation site is located near the DNA-binding domain, we envisaged that phosphorylation could interfere with DNA binding by electrostatic repulsion, or by inhibiting contacts between the protein and the DNA by steric hindrance. To examine these possibilities, we replaced Ser 367 with either aspartic or glutamic acid in an attempt to mimic phosphorylation in the ERM molecule. Both mutant proteins had the same DNA-binding capacity and were insensitive to PKA action, so confirming the role of Ser 367 in the change of ERM DNA binding activity upon PKA phosphorylation. However, both ERM mutants exhibited lower DNA binding activity as compared with the wild-type non-phosphorylated protein, but the effect was not as pronounced as that of the phosphorylation of Ser 367 in the wildtype protein. This indicates that not only phosphorylation can act by electrostatic repulsion and/or steric hindrance, but also that other mechanisms may be key determinants. Particularly, structural differences between phosphorylated and non-phosphorylated ERM may be critical for DNA binding. Indeed, we have shown that the phosphorylation of ERM by purified PKA alters the mobility of the remaining ERM-DNA complexes, so suggesting that the DNA-binding conformations before and after phosphorylation are distinct. Furthermore, conformational rearrangements have been shown to be related to the regulation of DNA binding in Ets proteins. A more extensively studied example is Ets-1, whose DNA binding activity is inhibited by a module consisting of two ␣-helices located NH 2 -terminal to the ETS domain, and a single COOH-terminal ␣-helix. A conformational change involving the unfolding of one of the NH 2 -terminal helices is associated with the relief of this intramolecular inhibition (29,30). This conformation is thought to be stabilized through protein interactions with inhibitory regions (29,31). Furthermore, it has been shown that the phosphorylation of Ets-1 in the NH 2 -terminal inhibitory region results in the inhibition of Ets-1 binding to DNA (32)(33)(34), favoring a shift of the conformational equilibrium toward a folded state (35). It has also been proposed that the phosphorylation of Elk-1 induces a conformational change in its structure, which affects its ability to bind DNA (36 -38).
While the ability of PKA-phosphorylated ERM to bind DNA in vitro is decreased, PKA stimulates ERM-dependent transcription in transient transfected cells. In fact, the ability of ERM to bind DNA is correlated with its transactivation activity. First, ERM and ETV1 avidly bound the E74 probe, although the proteins did not or modestly transactivate the E74 reporter, while an important decrease in DNA binding for the PKA-phosphorylated proteins is associated with PKA-mediated transactivation in transfected cells. In contrast, PEA3, which also avidly binds the probe but is unresponsive to PKA in terms of DNA binding, is not significantly transcriptionally activated in the presence of PKA. Second, the mutation of Ser 367 into alanine eliminates the apparent reduction in DNA binding following PKA phosphorylation and leads to a marked decrease of ERM transcriptional activation in the presence of PKA. Finally, the mutation of Ser 367 into aspartic acid or glutamic acid leads to the formation of an ERM protein with a moderate reduction in DNA binding and which is transcriptionally more active than the wild-type protein. Interestingly, this latter observation has been made on transfected cells in the absence of PKA, thus indicating that the decrease in DNA binding and the increase in transcriptional activity are intimately related. However, the interaction of ERM with DNA is essential since a mutant version of ERM which does not bind DNA because of a single mutation in its ETS-domain (ERM Y419P (14)) is transcriptionally inert in the absence or presence of PKA.
Although often described (25,39), increased transcriptional activation does not necessarily correlate with enhanced transcription factor binding to a promoter. In a recent work (40) on the transcription factor B-Myb, it has been shown that mutation of a single phosphorylation site enhanced binding to a Myb-binding sequence but decreased B-Myb transactivation potential. Furthermore, there is growing evidence that high affinity binding does not always lead to transactivation. Thus, some NF-B DNA-binding sites have been found to be very poorly activated by NF-B proteins (41) and high affinity binding of GATA-1 to a reporter gene does not necessarily induce transactivation (42). Concerning Ets proteins, it has recently been described that some natural promoter sites do not correspond to high consensus sites derived from in vitro studies of isolated Ets proteins (43). Moreover it has been observed that the Ets protein Erg activates the collagenase 1 promoter while any binding of the Erg protein is detected to the promoter, whereas the protein binds to the stromelysin I promoter but does not activate this Ets-regulated gene (44). A major factor that plays a role in Ets protein-DNA recognition is cooperative protein-protein interactions between Ets proteins and other factors that can alter the affinity of the Ets proteins for DNA (2). Thus, considering the ability of transcriptional regulators to adapt their structures to the particular DNA sequence that they recognize (42,45,46), it is possible that an Ets protein can bind to some Ets-binding sites without engaging protein-protein contacts or appropriate interactions with protein partners. This would induce an inability or a poor capacity to activate transcription. We therefore propose that phosphorylation of the transcription factor may affect its activity by altering its ability to interact with DNA and other proteins. In this model, the events which lead to decreased ERM DNA-binding capacity would result in conformational changes of the ERM molecule favoring its association with proteins involved in transcription. As suggested above, it is likely that decreased DNA binding of phosphorylated ERM is at least partially because of structural change. Furthermore, it cannot be excluded that a change in the DNA binding activity of phosphorylated ERM might lead to an altered conformation of the ERM-DNA complex affecting the protein interaction capacities of the bound protein. This would explain why the GAL4/ERM fusion protein is unresponsive to PKA when assayed with the GAL4-binding site-driven reporter while it is activated by the kinase when tested on an Etsbinding site reporter. Thus, enhanced protein-protein interactions could stabilize the interaction of the protein with DNA to overcome the decrease in DNA binding of the PKA-phosphorylated protein and favor the formation of a multiprotein complex to activate transcription.
A comparison in transfected cells of the properties of wild-type ERM with those of mutant versions carrying Ser 367 for alanine substitutions has shown that this site is necessary for mediating a full transcriptional response with respect to PKA. However, mutating this site is insufficient to fully eliminate the effect observed since residual activation is still detected in the mutated proteins as compared with the wild-type protein in the absence of PKA. This residual activation cannot be attributed to the phosphorylation of the other PKA site identified in the central part of the protein since mutant ERM carrying an Ser 367 for alanine substitution and the deletion of the second PKA-targeted region is still weakly activated transcriptionally when the kinase is expressed. An explanation would be that the residual transactivation effect is mediated by a PKA-activated kinase. Indeed, PKA may activate the MAP kinase ERK via a Ras-independent path-way (47), and ERK regulates several downstream signaling events, including the function of the ERM transcription factor (15). However, the inhibitor PD 98059, previously shown to selectively inhibit the ERK1/ERK2 MAPK pathway (48), did not reduce the PKA-induced ERM transcriptional activity (data not shown). Another possibility is that PKA may activate protein partners acting in concert with the PEA3 group members to activate transcription. This would explain why PEA3 which does not contain the PKA site is weakly activated by PKA. It must also be mentioned that the PKA-mediated activation of the ERM S367A protein occurs similarly in both RK13 and HeLa cells but that this effect is more pronounced on the Coll1 promoter as compared with the E74 one (about 50 and 25% of wild-type ERM activation, respectively). An activation mediated indirectly through other proteins might thus be promoter-dependent and the activation level might reflect differences in the ability of ERM to interact with these proteins depending of the binding site it recognizes. Thus these observations suggest a model in which PKA phosphorylation might not only increase the interaction of ERM with protein partners but also activate these partners to obtain a full transcriptional response.
At the present time, little is known about the protein partners of ERM. We have shown that this protein is able to physically interact with TBP, TAF II 40 and 60 (26) as well as the androgen receptor (20). However, one interesting candidate is the important transcriptional coactivator CBP/p300. Indeed, phosphorylation of transcription factors including cyclic-AMPand calcium-regulated nuclear factor (49), p53 (50), NF-B (51), and SMAD (52) increases their interactions with CBP/p300 and Ets proteins recruit this coactivator acting as a bridging molecule between DNA-binding factors and basal transcription machinery, so resulting in transcriptional activation (53)(54)(55)(56). Of particular interest is that ER81/ETV1, which is activated by PKA, interacts with CBP/p300 to regulate gene transcription. Moreover the complex ER81⅐p300 is able to recruit an unidentified kinase that can phosphorylate ER81 on serine residues distinct from the PKA-phosphorylation site described in this study and negatively regulate ER81-dependent transcription (57). Thus it is tempting to speculate that phosphorylation of ER81 by different kinases might have opposite effects on the recruitment of CBP/p300 to control ER81-dependent transcription.
In conclusion, we provide evidence that the Ets transcription factor ERM and the closely related factor ETV1 are direct targets of PKA, and that activation of PKA signaling is sufficient to activate them via an unexpected mechanism, i.e. the decrease in DNA binding. To our knowledge, this is the first example of PKA-dependent phosphorylation regulating the activity of a member of the Ets family. In addition to regulation at an artificial ETS-binding site, E74 site, the PKA-mediated activation of ERM occurs at another Ets-binding site located in the collagenase 1 promoter, thus suggesting a potential regulatory mechanism for ERM functions at different promoters via the PKA-signaling pathway. Furthermore, it should be noted that, in contrast to ERM and ER81/ETV1, PEA3/E1AF the third PEA3 group member is not directly activated by PKA. Since the three PEA3 group members display similar DNA binding specificities (23) and have been reported to be coexpressed (17,58,59), it is conceivable that specific promoters may therefore be targeted by different PEA3 group members depending on the kinases that phosphorylate them.