The ETS Domain Transcription Factor Elk-1 Regulates the Expression of Its Partner Protein, SRF*

The ternary complex factors (TCF) are a subfamily of ETS domain transcription factors that bind and activate serum response elements (SREs) in the promoters of target genes in a ternary complex with a second transcription factor, serum response factor (SRF). Here, we have identified the SRF gene as a target for the TCFs, thereby providing a positive feedback loop whereby TCF activation leads to the enhancement of the expression of its partner protein SRF. The binding of the TCF Elk-1 to the SRF promoter and subsequent regulation of SRF expression occurs in a ternary complex-dependent manner. Our data therefore reveal that SRF is an important target for the ERK and Rho signaling pathways that converge on a ternary TCF-SRF complex at the SRE on the SRF promoter.

The ternary complex factors (TCF) are a subfamily of ETS domain transcription factors that bind and activate serum response elements (SREs) in the promoters of target genes in a ternary complex with a second transcription factor, serum response factor (SRF). Here, we have identified the SRF gene as a target for the TCFs, thereby providing a positive feedback loop whereby TCF activation leads to the enhancement of the expression of its partner protein SRF. The binding of the TCF Elk-1 to the SRF promoter and subsequent regulation of SRF expression occurs in a ternary complex-dependent manner. Our data therefore reveal that SRF is an important target for the ERK and Rho signaling pathways that converge on a ternary TCF-SRF complex at the SRE on the SRF promoter.
Elk-1, SAP-1, and SAP-2/ERP/Net comprise the ternary complex factor (TCF) 1 subfamily of ETS-domain transcription factors (reviewed in Refs. 1 and 2). These proteins form ternary complexes on target promoters together with the MADS-box protein, serum response factor (SRF). Both protein-DNA and protein-protein interactions with SRF are required to form ternary complexes at the promoters of target genes like c-fos. The conserved B-box region of the TCFs plays a pivotal role in mediating these protein-protein interactions (3)(4)(5)(6)(7)(8). The TCFs can be phosphorylated within their transcription activation domains by members of all three of the major mitogen-activated protein kinase pathways present in mammals: ERK, JNK, and p38 (reviewed in Refs. 1, 2, and 9). In the case of Elk-1, this phosphorylation leads to the enhancement of its transactivation properties both by the recruitment and activation of the coactivator proteins Sur-2 and p300/CBP (10,11) and also the loss of corepressor complexes containing histone deacetylase 2 (12). Thus, Elk-1 provides an adaptor protein for SRF that can link it to the mitogen-activated protein kinase signaling pathways.
In addition to binding to the TCFs, SRF has recently been shown to be able to bind to members of the MAL/myocardin family of coactivator proteins (13)(14)(15)(16)(17)(18). In the case of MAL, this interaction permits linkage of the MAL-SRF complex to the Rho signaling pathway and subsequent activation of SRF-dependent gene expression (13,16). The SRF gene itself is thought to be one target gene for the MAL/myocardin-SRF complex. The interaction of the TCFs and MAL/myocardin proteins with SRF is mutually exclusive, where the Elk-1 B-box inhibits MAL/myocardin recruitment by SRF (19,20). This suggests either the existence of two different classes of SRF target genes (21) or the possibility of sequential (13) or mutually exclusive interactions of different coregulatory proteins with the same SRF target genes (20).
To date the number of TCF target genes identified is limited and comprises well studied immediate-early genes such as c-fos and egr-1 (reviewed in Refs. 1, 2, and 9). On these latter targets, SRF appears to aid TCF recruitment, but the reciprocal situation appears to operate on other genes such as nur77 where the TCFs recruit SRF to the promoter (22,23). It is currently unclear if the TCFs act in an SRF-independent manner although evidence has been gathered to suggest such a role on genes encoding tumor necrosis factor-␣ (24) and 9E3/cCAF (25).
Because of the possibility of functional redundancy among the TCFs, suggested by the minimal phenotypes obtained in mouse knock-out studies (23,26,27), we developed a cell line encoding an inducible repressive form of Elk-1 (Elk-1 fused to the engrailed repression domain; Elk-En) to probe the potential role of TCFs in regulating gene expression (28). The induction of Elk-En caused apoptosis and one key target gene identified was the anti-apoptotic gene Mcl-1. Importantly, the activity of Elk-En was B-box dependent demonstrating that it functions in a SRF-dependent manner. Here, we have extended these studies and show that the gene encoding the TCF partner protein SRF is also a target for the TCF Elk-1. Elk-1 works through a ternary complex with SRF on the SRF promoter, thereby providing a link to the ERK signaling pathway in addition to the well characterized link to the Rho pathway that acts through SRF on its own promoter.
Transient transfection experiments were carried out using Polyfect transfection reagent in 12-well plates (Qiagen). Luciferase assays were carried out using the dual light reporter gene assay system (Tropix) as described previously, using pCH110 as an internal control (28).
Western Blot Analysis-Western blotting was carried out using Supersignal West Dura Extended Duration Substrate (Pierce) and the following primary antibodies: anti-Elk-1 (Santa Cruz), anti-SRF (Santa Cruz), and anti-M2 FLAG antibody (Sigma). Data were visualized using a Bio-Rad Fluor-S MultiImager. Quantification of proteins was carried out using Quantity One software (Bio-Rad).
Gel Retardation Assays-Gel retardation assays were carried out with 32 P-labeled 116-base pair fragments of the mouse SRF promoter generated by PCR on templates pSRF-Luc(WT) and pSRF-Luc(mETS-103) (primers; ADS1251, GCAGCGAGTTCGGTATGTC, and ADS1252, CCGCTCCTTATATGGCGAGC) as described previously (32). Core SRF was produced as a glutathione S-transferase-tagged protein in bacteria (6), C-terminal-truncated Elk-1 was produced by in vitro transcription and translation using a TnT TM -coupled reticulocyte lysate system. 35 S-Labeled proteins were analyzed by electrophoresis through a 0.1% SDS-12% polyacrylamide gel, before visualization and quantification using a phosphorimager and Quantity One software (Bio-Rad). Protein-DNA complexes were analyzed on nondenaturing 5% polyacrylamide gels cast in 0.5ϫ Tris borate-EDTA and visualized by autoradiography and phosphorimaging.
Northern and Microarray Analysis-Total RNA extracted using the RNeasy kit (Qiagen) and Northern analysis was carried out as described previously (28). Probes were made by random priming (Roche) using templates derived from a human SRF and Elk-1 full-length cDNAs.
Induction of Apoptosis and Hoechst Staining-Apoptosis was scored by counting disrupted nuclei revealed by Hoechst stain (Sigma) as described previously (28). Etoposide (0.2 mM) was added to 293 cells for 48 h to induce apoptosis.

Repressive Elk-1 Constructs Down-regulate the Expression of SRF-
To study the effects of inhibiting TCF-mediated gene expression, and thereby uncover novel TCF target genes, we created a 293-derived stable cell line that inducibly expresses a fusion of Elk-1 with the powerful Engrailed (En) repression domain (EcR293(Elk-En)#1.3) (28). Upon induction of its expression with PA, this fusion protein is expected to be recruited to all TCF-dependent promoters and hence ablate the expression of genes normally controlled by TCFs. To identify potential target genes, we used Affymetrix microarray analysis to compare the mRNA expression profiles of unstimulated EcR293-(Elk-En)#1.3 cells with the same cells stimulated with PA for 6 h to induce the expression of Elk-En. One of the downregulated genes identified was SRF (1.28-fold reduced). Northern analysis confirmed the microarray data, demonstrating a clear reduction in SRF mRNA following PA induction, which coincides with the induction of the expression of Elk-En (Fig.  1A). Similarly, SRF is also down-regulated at the protein level following loss of the SRF message (Fig. 1B). We also tested whether the prior induction of Elk-En could block the activation of SRF expression by serum (FCS) stimulation. In comparison to the uninduced cells, pretreatment of cells with PA to induce Elk-En expression led to a reduction in SRF induction by serum ( Fig. 2A). Thus, Elk-1 appears to be able to regulate the transcription of the gene encoding its partner protein SRF.

Elk-En Represses SRF Expression in a B-box-dependent
Manner-Elk-1 is thought to act primarily through ternary complexes in conjunction with SRF on SREs (reviewed in Refs. 1 and 2). However, it is possible that overexpression of Elk-En fusions could result in the repression of genes usually regulated by different ETS domain transcription factors in an SRFindependent manner by virtue of their overlapping DNA binding specificities. We therefore created a control cell line (EcR293(Elk-En(L158P))#L8) that contains an Elk-En derivative that has a point mutation in its B-box region that blocks its ability to bind to SRF and hence be recruited to DNA in an SRF-dependent manner (28). This fusion protein should still inhibit transcription through high affinity binding sites. Control experiments demonstrated that this was the case in reporter assays and that the expression levels of the wild-type and L158P version of Elk-En were similar (28).
The prior induction of Elk-En(L158P) expression had little effect on serum-stimulated SRF expression (Fig. 2B). This contrasts with the induction of WT Elk-En, which caused a clear reduction in serum-inducible SRF expression ( Fig. 2A). We also examined the protein levels of SRF in EcR293(Elk-En(L158P))#L8 cells following induction of Elk-En(L158P), in contrast to the reductions in SRF levels observed upon induction of wild-type Elk-En (Fig. 1B), no change in SRF expression was observed, even after 63 h of induction with PA (Fig. 2C).
To demonstrate that Elk-En could act directly on the SRF promoter, we carried out transient transfection assays with increasing amounts of Elk-En and a SRF promoter-driven luciferase reporter construct (Fig. 3A). Wild-type Elk-En efficiently repressed the activity of the SRF promoter. In contrast, the mutant derivative, Elk-En(L158P), was unable to repress the activity of the SRF promoter. This differential effect was not because of differences in expression of the Elk-En fusions (Fig. 3B). Although Elk-En(L158P) is unable to affect the activity SRF promoter, it is possible that other ETS domain proteins might regulate this promoter. Indeed, overexpression of a different ETS domain protein, PEA3, also caused upregulation of this promoter (data not shown). To establish the specificity of ETS domain protein action, we determined whether PEA3 could compete with Elk-En fusions in regulating the SRF promoter. However, PEA3 was unable to compete with Elk-En (Fig. 3C), suggesting that Elk-1 is an important regu-lator of this promoter. Collectively, these data therefore demonstrate that Elk-1-mediated regulation of the SRF promoter takes place in a B-box-dependent manner, suggesting that it is recruited by SRF into a ternary, promoter-bound complex.
Elk-1 Recruitment to the SRF Promoter Is Mediated by an Ets Binding Site-The SRF promoter contains two CArG boxes that are bound by SRF and are important for signal-mediated activation of SRF expression (33,34). In addition, there is a functionally important ets binding motif located upstream from these binding sites (29) (Fig. 4A). We therefore tested whether Elk-1 can bind to this module in vitro using a DNA fragment encompassing the upstream CArG box and the ets site (Fig.  4A). In the absence of SRF, Elk-1 was unable to bind to the SRF promoter (Fig. 4B, lane 3). However, in the presence of SRF, the binding of Elk-1 was strongly stimulated and a ternary complex was formed on the promoter (Fig. 4B, lane 4). This binding was dependent on the integrity of the B-box region of Elk-1 as Elk-1(L158P) was unable to bind to the promoter (Fig. 4B, lane

5)
, demonstrating the importance of interactions with SRF for its recruitment. Next, we tested the requirement for the ets motif for recruitment of Elk-1 to the SRF promoter. Mutation of the ets motif reduced the ability of Elk-1 to bind to the SRF promoter in a complex with SRF (Fig. 4B, lane 9). Thus, both protein-DNA contacts and protein-protein interactions with SRF are essential for the efficient recruitment of Elk-1 to the SRF promoter To extend these observations in vivo, we investigated the ability of Elk-1 derivatives to regulate an SRF promoter-reporter construct that contains a mutation in the ets motif (Fig.  5A). The wild-type SRF promoter was repressed by Elk-En and activated by a constitutively active Elk-VP16 fusion protein (Fig. 5B). In contrast, the activity of the mutated promoter, SRF(ets-mut), was unaffected by either fusion protein (Fig.  5B). We also investigated the ability of wild-type Elk-1 to activate the SRF promoter in the presence of PMA, which stimulates the mitogen-activated protein kinase pathway (Fig.  5C). Elk-1 activated the wild-type SRF promoter in a dose-dependent manner. In contrast, its ability to activate the mutant SRF promoter was severely reduced. Collectively, these data demonstrate the importance of the ets motif within the SRF promoter for the binding of Elk-1 and the subsequent ability of Elk-1 to regulate its expression.
Elk-1 Binds to the SRF Promoter in Vivo-The above results indicate that Elk-1 can bind to the SRF promoter in vitro and regulate its activity in vivo. To demonstrate that endogenous Elk-1 can occupy the SRF promoter in vivo, we carried out a chromatin immunoprecipitation experiment in HeLa cells. In the presence of Elk-1 antibodies, the SRF promoter was precipitated from formaldehyde cross-linked total cell lysates (Fig.   6B, top panel, lanes 5 and 6). In contrast, control antibodies precipitated background levels of the SRF promoter (Fig. 6B,  lanes 3 and 4) and the Elk-1 antibody was unable to precipitate an intronic fragment of the SRF gene above background levels (Fig. 6B, lower panel). The association of Elk-1 with the SRF promoter was enhanced irrespective of the presence of PMA stimulation (Fig. 6B, lanes 5 and 6). Similarly, Elk-1 could be detected at the promoter of the well characterized target gene, egr-1, both in the presence and absence of PMA stimulation. Thus, on the SRF and egr-1 promoters, it appears that ERK pathway-dependent activation of Elk-1 does not enhance promoter occupancy, unlike previous in vitro studies that demon- strate phosphorylation-enhanced recruitment of Elk-1 to the c-fos SRE (35,36). These data therefore demonstrate that endogenous Elk-1 binds to the SRF promoter in vivo, thereby suggesting that the regulatory effects of Elk-1 on SRF expression are direct.

Regulation of SRF Expression by the ERK and Rho
Pathways-Previously, SRF has been identified as a target gene of the Rho signaling pathway (21), and this regulation is mediated by coactivator proteins from the MAL family that act through SRF bound to the promoter (Fig. 7A). The binding of MAL is mutually exclusive with TCF binding to SRF (19), suggesting that regulation by TCFs and MAL might also be mutually exclusive. However, it is also possible that the ERK pathway, through the TCFs, can contribute to activation of the SRF promoter. This would be important physiologically, as under different conditions, signaling through either the Rho or ERK pathway alone or simultaneous activation of several pathways might occur.
To investigate the convergence of these two pathways on the SRF promoter, we first examined the expression of SRF in response to activating the ERK pathway with PMA or the ERK and Rho pathways by serum (FCS). The ERK and Rho pathway inhibitors, U0126 and LB, respectively, were used to identify the individual contributions of these pathways to SRF induction. SRF expression was increased by PMA stimulation, and this increase was selectively blocked with U0126, demonstrating that this activation was ERK pathway dependent (Fig. 7B,  lanes 2-4). Similarly, FCS stimulated SRF expression, and this was inhibited by both LB and, to a lesser extent, U0126 (Fig.  7B, lanes 5-7). Similar observations have been made in NIH-3T3 cells (21). Thus, the ERK pathway can activate SRF expression and contribute to its expression following serum induction.
We also tested whether the ERK and Rho pathways contributed directly to the regulation of the SRF promoter using reporter gene assays. First, we overexpressed Elk-1 to sensitize the SRF promoter to ERK pathway-mediated stimulation. Under these conditions, both PMA and FCS treatment led to an increase in the activity of a SRF promoter-driven luciferase reporter (Fig. 7C). The inhibition of the ERK pathway reduced the activation of the SRF promoter by both treatments. However, LB was less effective in reducing the activation of the SRF promoter by FCS, presumably because of Elk-1 competing with MAL recruitment. To probe this possibility further, we tested the response of the SRF promoter to activation of the Rho pathway by jasplakinolide in the presence and absence of exogenous Elk-1 (Fig. 7D). The addition of Elk-1 inhibited the activation of the SRF promoter by jasplakinolide. This was not a general inhibitory effect as the promoter now became more responsive to activation by PMA. Thus, the level of Elk-1 in the cell can determine whether the ERK or Rho pathway can activate the SRF promoter.
Conversely, we asked whether overexpression of a constitutively active form of MAL, MAL⌬N (16), could interfere with Elk-1-mediated SRF promoter regulation. We transfected 293 cells with either the WT or mutant (ets-mut) versions of the SRF-luciferase reporter construct in the presence of a low amount of Elk-En to dampen down the activity of the promoter. A dose-dependent increase in the activity of the WT promoter was observed upon transfection of increasing amounts of MAL⌬N expression plasmid (Fig. 7E). However, in contrast, the ets-mut promoter was activated to maximal levels at the lowest concentration of MAL⌬N expressed, with no further increases seen at higher levels of MAL⌬N (Fig. 7E). Thus, the absence of the ets motif within the SRF promoter makes it more sensitive to activation by MAL⌬N, thereby demonstrating that occupancy of this ets motif in vivo can inhibit activation via MAL.
Finally, we compared the ability of wild-type Elk-1 and an alternative ETS domain transcription factor, PEA3, to affect the activity of MAL⌬N on the SRF promoter in serum-starved cells. Experiments were carried out under serum-starved conditions where the ERK pathway and hence the ETS domain protein targets are not activated. Elk-1 inhibited the action of MAL⌬N in a dose-dependent manner, but PEA3 did not, and instead weakly potentiated the activity of MAL⌬N on the SRF promoter (Fig. 7F). This is consistent with a model whereby Elk-1 binds through SRF interactions, thereby inhibiting MAL recruitment, but that other ETS domain proteins like PEA3 may in some circumstances function in an SRF-independent manner to activate the SRF promoter.
Collectively, these data corroborate previous observations that indicate that TCF and MAL/myocardin recruitment by SRF is mutually exclusive (19,20). However, importantly, they demonstrate a clear role for the ERK pathway in regulating SRF expression and suggest how multiple signaling inputs might lead to up-regulation of the SRF promoter (see "Discussion").

DISCUSSION
The TCF transcription factors play an important role in transducing extracellular signals into a nuclear response by acting as targets for the mitogen-activated protein kinase signaling pathways (reviewed in Refs. 1, 2, and 9). To fully understand the physiological role of the TCFs, it is important to gain a fuller insight into the spectra of target genes that they regulate. To date, the focus has been primarily on the immediate-early genes such as c-fos and egr-1 (reviewed in Refs. 1, 2, and 9). Here we have identified SRF as a direct TCF target gene. This provides an elegant positive feedback loop whereby the TCFs can regulate the expression of the TCF partner protein SRF.
We initially identified SRF as a TCF target gene by demonstrating down-regulation of SRF expression in cell lines expressing the repressive Elk-En fusion protein (Fig. 1). By using a combination of reporter gene and chromatin immunoprecipitation analyses, we have shown that the TCFs can directly affect the SRF promoter and that endogenously expressed Elk-1 can be found on this promoter in vivo (Figs. 3, 5, and 6). It is currently not known whether TCFs other than Elk-1 can participate in the regulation of SRF. Indeed, as overexpressed Elk-En can potentially block regulation by all the TCFs because of their structural similarity, it will be important to probe whether the other TCFs can work in the same way. It remains possible that other ETS domain proteins might act on the SRF promoter through the ets site. Indeed, overexpression of PEA3, a target of the ERK pathway (37), can activate the SRF promoter in an ets site-dependent manner (data not shown). However, PEA3 is unable to compete with Elk-1 for promoter occupancy (Fig. 3C). Indeed, we show that in cells containing Elk-1, the SRF promoter is occupied by Elk-1 (Fig.  6). It remains possible that in other cell types, where low levels of Elk-1 and high levels of ETS domain proteins such as PEA3 exist, that different ETS domain proteins contribute to SRF regulation. However, the interplay with the Rho/MAL pathway would likely differ, as interference with SRF-dependent MAL recruitment would not occur (Fig. 7F).
Elk-1 appears to act to couple signals generated by the ERK pathway to the activation of the SRF promoter (Fig. 7), whereas the coactivator MAL has previously been shown to couple Rho pathway signals to this promoter (21). Signaling through the ERK pathway is of importance where signaling does not activate the Rho pathway, here induced using the mitogen PMA, and as a potential contributing pathway to more complex signaling triggers such as serum. Importantly, we show that the ability of Elk-1 to regulate SRF expression is dependent on the formation of a ternary complex with DNAbound SRF (Figs. 2, 4, and 5). Previous studies have shown that signaling via the Rho pathway also functions through SRF bound to its own promoter, although in this case, the MAL/ myocardin family protein MAL represents the coregulatory partner that links to this pathway (13,16). We demonstrate that a functional antagonism can occur depending on the relative levels of MAL and the TCFs within the cell (Fig. 7), which is consistent with previous data demonstrating that Elk-1 can inhibit MAL recruitment in a B-box-dependent manner to SRFregulated promoters (19,20). Thus depending on the relative levels of the TCF and MAL/myocardin family proteins, the relative contributions of these two classes of coregulators might differ. Alternatively, there may be a role for both coregulators in permitting convergence of the ERK and Rho pathways through SRF bound to promoters. Our data suggest that the latter scenario may well exist at the SRF promoter, as inhibitor studies demonstrate a contribution of the ERK pathway to mitogenic stimulation of SRF expression (Fig. 7B) that is consistent with results from a previous study (21). In addition, it has been proposed that the TCFs might contribute to the early induction phase of other genes like c-fos in response to mitogens, whereas the MAL/myocardin family contributes to the later phases (13). Furthermore, recent studies demonstrate that promoters in several smooth muscle genes can be regulated through SRF by either the TCFs or MAL/myocardin family members depending on the signaling pathways that are active (20). Thus our data add further weight to an emerging model that SRF can form a platform that can differentially recruit coregulatory transcription factors and permit selective or combinatorial responses to different signaling pathways depending on the target promoter.
Finally, our data further suggest an important role for SRF expression in cell survival. Previously, SRF has been shown to be a target for degradation in response to apoptotic pathways (38,39) and also to regulate the expression of anti-apoptotic proteins like Bcl-2 (40). Recently we showed that the induction of Elk-En causes apoptosis and that one key anti-apoptotic target gene was the antiapoptotic gene Mcl-1 (28). As SRF expression is also down-regulated upon Elk-En induction, then this too might play a role in triggering apoptosis. Indeed, the down-regulation of SRF transcript levels is a natural process observed upon apoptotic induction caused by etoposide treatment. 2 However, although we could partially rescue apoptosis induced by Elk-En (data not shown), we have been unable to differentiate whether this was because of replacement of the down-regulated endogenous SRF, or merely titration of the Elk-En away from the promoters of antiapoptotic genes. Thus, the TCFs can potentially promote cell survival through the regulation of multiple genes and provide an important link to the prosurvival effects of the ERK pathway.