Granulocyte Colony-stimulating Factor Induces egr-1Up-regulation through Interaction of Serum Response Element-binding Proteins*

Granulocyte colony-stimulating factor (G-CSF) stimulates the proliferation and maturation of myeloid progenitor cells both in vitro and in vivo. We showed that G-CSF rapidly and transiently induces expression of egr-1 in the NFS60 myeloid cell line. Transient transfections of NFS60 cells with recombinant constructs containing various deletions of the humanegr-1 promoter identified the serum response element (SRE) between nucleotides (nt) –418 and –391 as a critical G-CSF-responsive sequence. The SRE (SRE-1) contains a CArG box, the binding site for the serum response factor (SRF), which is flanked at either side by an ETS protein binding site. We demonstrated that a single copy of the wild-type SRE-1 in the minimal promoter plasmid, pTE2, is sufficient to induce transcriptional activation in response to G-CSF and that both the ETS protein binding site and the CArG box are required for maximal transcriptional activation of the pTE2-SRE-1 construct. In electromobility shift assays using NFS60 nuclear extracts, we identified SRF and the ETS protein Fli-1 as proteins that bind the SRE-1. We also demonstrated through electrophoretic mobility shift assays, using an SRE-1 probe containing a CArG mutation, that Fli-1 binds the SRE-1 independently of SRF. Our data suggest that SRE-binding proteins potentially play a role in G-CSF-induced egr-1expression in myeloid cells.

Treatment of myeloid cells with G-CSF results in rapid and transient expression of egr-1 independently of protein synthesis. G-CSF induces egr-1 expression and granulocyte differentiation in 32Dcl3 cells (20) and also stimulates proliferation and egr-1 expression in human UT-7 epo cells overexpressing the wild-type G-CSF receptor (21). The expression of egr-1, like that of other immediate early genes, is governed by preexisting regulatory proteins that are posttranslationally modified and thus activated upon receptor stimulation.
The precise signaling events that mediate expression of egr-1 in response to G-CSF in proliferative responses have not been elucidated. Identification of this pathway may provide insights into possible mechanisms that lead to the development of leukemogenesis. It has been suggested that G-CSF selectively activates distinct early growth response genes through different Janus kinase-STAT proteins. For example, G-CSF stimulation of the early genes OSM, IRF-1, and egr-1 is dependent on STAT5 activation, whereas activation of c-fos is STAT5-independent (21). Although STAT5 protein expression is induced in response to G-CSF, that STAT5 DNA recognition element has not been identified in the murine egr-1 promoter (21).
Signaling pathways that control egr-1 expression and myeloid cell proliferation have been examined for granulocyte-macrophage colony-stimulating factor (GM-CSF) and IL-3 (22). The receptors for GM-CSF and IL-3, like G-CSF receptor, are also members of the cytokine receptor superfamily and regulate early myeloid development. GM-CSF-and IL-3-induced signals converge upon the cAMP response element in the egr-1 promoter in the factor-dependent human myeloid leukemic TF-1 cell line (22). egr-1 promoter sequences mediating G-CSF-induced egr-1 expression and myeloid proliferation, including the egr-1 promoter sites responsive to G-CSF, appear to be distinct from those involved in GM-CSF and IL-3 signaling, due to the different proteins activated by these proteins.
The goal of our study was to define G-CSF-responsive sequences of the egr-1 promoter and to identify interacting proteins in NFS60 cells. We show that egr-1 is rapidly and transiently expressed in NFS60 cells stimulated with G-CSF and that this activation occurs independently of protein synthesis. Transient transfections with recombinant egr-1 promoter constructs in NFS60 cells demonstrated that the CArG box and the ETS protein binding site (EBS) are required for maximal transcriptional activation of egr-1 in response to G-CSF. Electromobility gel shift assays (EMSAs) showed that SRF and Fli-1 bind the CArG and EBS between nucleotides (nt) -418 and -391 in the egr-1 promoter, respectively. Our experiments suggest that serum response element (SRE)-binding proteins may associate as a quaternary complex to maximally activate egr-1 transcription. Thus, signaling pathways activated by G-CSF may be distinct from those activated by GM-CSF or IL-3, suggesting a potential mechanism for specificity between growth factors that regulate myelopoiesis.
Northern Analysis-Cells were serum-and factor-starved for 18 h and then stimulated with G-CSF for 0, 30, 60, 90, and 120 min. Cells were also stimulated with 12-O-tetradecanoylphorbol-13-acetate (50 ng/ml) for 60 min (positive control) or diluent (0.02% BSA in phosphatebuffered saline). To demonstrate protein expression in the absence of protein synthesis, cells were stimulated with G-CSF (10 nM) or tetradecanoyl phorbol acetate (50 ng/ml) in combination with the protein synthesis inhibitor cycloheximide for 30 min. The cells were harvested, and total RNA was extracted by the Nonidet P-40 method (23). Twenty micrograms of extracted RNA from each sample was separated on a 1% formaldehyde gel and transferred to a nylon membrane. The membrane was hybridized with a [ 32 P]dCTP-labeled 1.3 EcoRI fragment of murine egr-1/tis8 or with a [ 32 P]dCTP-labeled 500-base pair actin fragment as control.
Construction of pTE2 Plasmids-The SRE from nt -418 to -391 of the human egr-1 promoter (SRE-1) was synthesized as a single element, with HindIII and XbaI sites created at the 5Ј and 3Ј ends, respectively (5Ј-AGCTTCGGA(CCGGAAT)G(CCATATAAGG)A(GCAGGAAG)GAT-CCT-3Ј). In addition, mutated SRE-1 were synthesized with the 5Ј and 3Ј HindIII and Transient Co-transfections and CAT Assays-Transient co-transfections of constructs into NFS60 cells were performed by electroporation (Bio-Rad) at 250 V, 960 F. Cells were serum-and growth factor-starved for 18 h in RPMI-0.5% BSA. Twenty million cells were transfected with 20 g of the specified egr-1 promoter/CAT construct or 20 g of the specified minimal promoter pTE2/CAT construct and 5 g of the pCMV-␤-galactosidase plasmid (pCMV-␤gal) (internal control). Transfected cells were resuspended in RPMI-0.5% BSA and stimulated with G-CSF (10 nM) or diluent control (0.02% BSA in phosphate-buffered saline) for 3 h. The cells were harvested; half the lysates were assayed for CAT activity and the other half for ␤gal activity. The CAT assay was used to measure egr-1 promoter activity and was performed as described previously (22). The amount of acetylated and unacetylated [ 14 C]chloramphenicol was determined by thin-layer chromatography and quantified by liquid scintillation counting. The ␤gal assay (Promega) was used as internal control for transfection efficiency.  -4), and the G-CSF-treated extracts produced five gel shift bands/complexes (G, [1][2][3][4][5]. C, the SRE-1 probe was incubated with 15 g of G-CSF-stimulated nuclear extracts in combination with a 200 M excess of unlabeled oligonucleotide (nonspecific (NS), SRE-1, or CArG sequence) or with SRF antibody or IgG control. Gel shifts were electrophoresed on a 4% SDS-polyacrylamide gel. The bands represent reactions that were run on the same gel. ng/l). After incubation periods, the samples were loaded onto a 4% polyacrylamide gel and run at 100 V in 0.4ϫ TBE. Gels were dried and exposed to film at Ϫ70°C.

NFS60 Cells Are Dependent on G-CSF for Proliferation-We
demonstrate that NFS60 cells are a growth factor-dependent myeloid cell line (data not shown). Furthermore, NFS60 cells proliferate in response to G-CSF in a dose-dependent manner (27,28). Therefore, NFS60 cells provide a good model in which to examine G-CSF-induced proliferative signals.
G-CSF Induces Rapid and Transient Expression of egr-1 in NFS60 Cells-Rapid and transient expression of egr-1 has been previously shown to occur in response to GM-CSF in myeloid leukemic TF-1 cells (22). To demonstrate egr-1 expression in NFS60 cells in response to G-CSF, Northern blot analysis was performed. Northern blot analysis with RNA from NFS60 cells stimulated with G-CSF for 0, 30, 60, 90, and 120 min demonstrated a rapid and transient induction of egr-1 (Fig. 1). Expression of egr-1 was not induced in diluent-treated cells (0 min). Accumulation of egr-1 RNA was observed within 30 min and was no longer observed at 60 min following G-CSF treatment. Stimulation of cells for 60 min with G-CSF or 12-Otetradecanoylphorbol-13-acetate in the presence of the protein synthesis inhibitor cycloheximide resulted in induction of egr-1. 12-O-tetradecanoylphorbol-13-acetate has been previously shown to induce egr-1 expression within 60 min of cell treatment (8). These results were confirmed in two independent experiments. Our results demonstrate that egr-1 expression occurs rapidly and transiently in cells stimulated with G-CSF and that induction of egr-1 occurs independently of protein synthesis.
We previously showed that the -56 nt region of egr-1 contains minimal activity that is equal to the pCAT empty vector (22). We have therefore used p-56 CAT as the vector control in our experiments. In response to G-CSF, the -600 and -480 nt egr-1 constructs demonstrated maximal stimulation (Fig. 2) at 9.0-fold activity relative to p-56 CAT control vector. Deletion of nucleotides between -480 and -387 resulted in a 6-fold decrease in transcriptional activation (p ϭ 0.0024; Fig. 2), and further deletion to nt -116 did not reduce this activity. In diluent-treated cells, the constructs had basal activities that were not statistically significantly different from that of the control vector (data not shown). All transfections were performed in duplicate or triplicate, representing an average of three to seven experiments. These results indicate that the region between -480 and -387 nt of the egr-1 promoter contains critical sequences required for maximal transcriptional activation of egr-1.
The region between nt -480 and -387 contains a single SRE, with a central CArG box flanked by EBSs, which we have called SRE-1 (Fig. 3A). Transfections with a construct containing -418 nt of the egr-1 promoter (Fig. 3A) showed similar activity levels with the p-480 nt construct (data not shown). These results suggested that the SRE, between nt -418 and -391 (SRE-1), may contain a critical transcription factor binding site regulating G-CSF-induced transcription of egr-1.
Nuclear Extracts from NFS60 Cells Contain Proteins That Recognize the SRE-1 Sequence-Previous studies have identi-fied SRF binding to the CArG box of SREs in the promoters of immediate early genes (29 -32). To determine whether endogenous SRE-binding proteins in nuclear extracts from NFS60 cells interact with the wild-type SRE-1 sequence between nt -418 and -391 of the egr-1 promoter (Fig. 3A), EMSAs were performed. Nuclear extracts prepared from diluent-or G-CSFtreated NFS60 cells were incubated with the SRE-1 oligonucleotide probe with an excess of unlabeled specific or nonspecific competitor and then analyzed on the same SDSpolyacrylamide gel. The diluent-and G-CSF-treated nuclear extracts produced four and five gel shift bands or complexes, designated D1-D4 and G1-G5, respectively (Fig. 3B, lanes 1  and 3). Band specificity was demonstrated by competition of the bands with a 200 M excess of specific unlabeled probe sequence (Fig. 3B, lanes 2 and 4) and by the lack of competition with an excess of unlabeled nonspecific sequence (Fig. 3B, lanes  1 and 3). The two slowest migrating bands in the diluenttreated extracts, D1 and D2 (Fig. 3B, lane 1), were only evident when an increased amount of protein was used, and band D1 migrated differently from band G1 in the G-CSF-stimulated extracts (Fig. 3B, lanes 1 and 3). These results suggest that bands/complexes 1-5 represent proteins that specifically bind the SRE-1 sequence and that potentially different or biochemically modified (i.e. phosphorylated) SRE-binding proteins bind to SRE-1 in G-CSF-treated compared with diluent-treated extracts.
SRF Binds the SRE-1 between Nucleotides -418 and -391 in the egr-1 Promoter-To determine whether SRF binds the SRE-1 in the egr-1 promoter, we performed EMSA supershift experiments with the SRF antibody. Addition of SRF antibody resulted in a supershift of bands 1 and 2 (Fig. 3C, lanes 4 -6), whereas addition of IgG control did not produce any change in the gel shift pattern (Fig. 3C, lane 7). The control probe con- taining the CArG sequence (Santa Cruz Biotechnology Inc.) resulted in a supershifted band that co-migrates with the supershifted band seen with the SRE-1 probe (Fig. 3C, lane 11). Furthermore, an excess of unlabeled CArG oligonucleotide alone specifically competed bands 1 and 2 (Fig. 3C, lane 3), suggesting that these bands are SRF-containing complexes. Addition of the SRF antibody in EMSAs using diluent-treated extracts also resulted in a supershifted band pattern (data not FIG. 5. Stimulation of pTE2 constructs containing the SRE-1 sequence. A, pTE2 oligonucleotide constructs containing one copy of the wild-type or mutant SRE-1 sequences (5Ј EBS mutant, L m R; 3Ј EBS mutant, LR m ; 5Ј and 3Ј EBS mutant, L m R m ; and CArG m/ BglII) were made. B, the pTE2 constructs (20 g) were individually transfected into NFS60 cells with CMV-␤gal (5 g) in NFS60 cells serum-and factorstarved for 18 h. Cells were then stimulated with diluent (phosphate-buffered saline with 0.02% BSA) or G-CSF (10 nM). The fold induction by G-CSF was determined as described under "Materials and Methods" and represents an average of 3-11 experiments performed in duplicate or triplicate. A significant increase in pTE2-SRE-1 activity (*) was observed as compared with the pTE2 empty vector (p Ͻ 0.001), and the pTE2 mutant constructs (L m R, p ϭ 0.0381; LR m , p ϭ 0.0022; L m R m , p ϭ 0.0056; and CArG m , p ϭ 0.0132). C, gel shift competition assays were performed to determine CArG mutations that inhibit SRF binding. An oligonucleotide SRE-1 sequence (0.1 g) was labeled with [␥-32 P]dATP and used as probe in EMSA experiments. The probe was incubated with 15 g of G-CSF-stimulated nuclear extracts and a 200 M excess of nonspecific (NS) oligonucleotide or an SRE-1 sequence with one of three different CArG mutations (GG deletion, GG to TT substitution, or BglII substitution). The arrow indicates the gel shift band (band 2) that is not competed by an excess of unlabeled CArG mutant oligonucleotide.
shown). These results were confirmed in two separate experiments. SRF binding to the SRE has been shown to be enhanced upon SRF phosphorylation (33,34), yet it has also been shown to bind the SRE constitutively (35)(36)(37)(38). Our data indicate that SRF binds the SRE-1, and the supershift band corresponds with the band observed with SRF binding to the CArG probe. Thus, these results indicate that SRF binds the CArG sequence of SRE-1 in G-CSF-treated nuclear extracts.
Fli-1 in NFS60 Nuclear Extracts Binds the EBS in SRE-1 of the egr-1 Promoter-A previous report has demonstrated that Fli-1 is one of the ETS proteins that recognizes the EBS of SRE-1 in the egr-1 promoter (39). To identify ETS proteins in NFS60 nuclear extracts that bind the EBS of SRE-1, EMSA experiments were performed using antibodies to various ETS proteins. Addition of Fli-1 antibody to G-CSF-stimulated extracts resulted in the disappearance of band 1 and the formation of two supershifted bands, 1A and 1B (Fig. 4, lanes 1-3). This suggests that band 1A represents a complex composed of both SRF and Fli-1 bound to the SRE-1 probe. Band 1B may represent a complex of SRF and Fli-1 with additional proteins. Addition of Fli-1 antibody in EMSAs using diluent-treated extracts resulted in a very weak supershift band pattern under certain conditions (data not shown). Addition of antibodies to other ETS proteins, including Elk-1, Sap1a/b, PU.1, Elf-1, ETS-1/2, ERG-1/2, and PEA3, did not produce a change in the gel shift pattern in either diluent-or G-CSF-treated extracts (data not shown). Our data suggest that in addition to SRF, Fli-1 also binds the SRE-1 in the egr-1 promoter in G-CSF-stimulated extracts.
The SRE-1 Sequence Is Sufficient to Induce Transcription in Response to G-CSF-Ternary complexes composed of SRF and ETS protein family members have been shown to regulate the expression of many immediate early gene SREs (32, 40 -47) and may also regulate G-CSF-induced egr-1 expression. To determine whether the SRE-1 sequence is sufficient to induce egr-1 transcriptional activation in response to G-CSF, NFS60 cells were transiently transfected with a construct containing a synthetic oligonucleotide representing a single human SRE-1 in the pTE2 vector that contains a heterologous TK promoter and the CAT gene. pTE2 constructs with mutations within the SRE-1 were also prepared (Fig. 5A). The mutant SRE-1 constructs included a 3-base substitution within the CArG core consensus binding sequence (CArG m /BglII) and a GG to TT substitution within the ETS consensus binding site (GGA) at the left (L m R), the right (LR m ), or both the left and right (L m R m ) EBSs of SRE-1.
Upon G-CSF treatment, the pTE2 SRE-1 wild-type construct demonstrated a 3.5-fold induction compared with the empty vector (p Ͻ 0.001; Fig. 5B). Upon G-CSF stimulation, the pTE2-CArG m/ BglII, -L m R, -LR m , and -L m R m constructs behaved similarly to vector control, but all demonstrated reduced induction levels as compared with pTE2 SRE-1 (p ϭ 0.0132, 0.0381, 0.0022, and 0.0056, respectively; Fig. 5B). In diluent-treated cells, the basal activity of the pTE2-CArG m/ BglII, -L m R, -LR m , and -L m R m constructs demonstrated percentage of acetylation values similar to the pTE2 empty vector, ranging from 1.5 to 0.82%, respectively (data not shown). These experiments were repeated 3-11 times and were performed in triplicate. The CArG box mutation (pTE2-CArG m/ BglII) also efficiently inhibited competition of the SRF/SRE gel shift band (gel shift band 2) in EMSA experiments (Fig. 5C, lanes 2 and 3). Mutations within the EBS core consensus sequence at the 5Ј (L m R), 3Ј (LR m ), or both 5Ј and 3Ј EBS (L m R m ) were previously shown to inhibit ETS protein binding (39). Our data suggest not only that the SRE-1 is sufficient to induce transcriptional activation in response to G-CSF but also that the CArG box and both the 5Ј and 3Ј EBSs are required for maximal stimulation of the pTE2 SRE-1 construct. Ternary complexes composed of SRF and ETS protein family members have been shown to regulate the expression of many early gene SREs (32, 40 -47), and they appear to regulate G-CSF-induced egr-1 expression.
Fli-1 Binds SRE-1 Independently of SRF-In vitro translated Fli-1 protein has previously been shown to bind the SRE-1 sequence in the murine egr-1 promoter independently of SRF (39). To determine whether Fli-1 in nuclear extracts from NFS60 cells requires SRF for binding to SRE-1, gel shifts were performed with nuclear extracts and a wild-type SRE-1 or a SRE-1 probe containing a mutation in the CArG box (CArG m/ BglII; Fig. 6). In the presence of nonspecific competitor, bands 3-5 were observed with nuclear extracts from G-CSF-stimulated cells (Fig. 6, lane 7). The absence of bands 1 and 2 indicates that SRF cannot bind the CArG mutant probe. Use of the unlabeled mutant probe sequence and the wild-type SRE sequence in excess resulted in competition of all gel shift bands, due to the presence of intact EBSs (Fig. 6, lanes 8 and 9). The addition of SRF antibody and IgG control did not result in a supershifted band, whereas the addition of Fli-1 antibody resulted in a band of slower mobility (Fig. 6, lanes 10 -12). Our data indicate that Fli-1 in nuclear extracts from myeloid cells can bind SRE-1 independently of SRF. Furthermore, although Fli-1 can bind SRE-1 independently of SRF, our transfection data suggest that both SRF-and ETS-binding protein(s) are required for maximal transcriptional activation of egr-1 in response to G-CSF.
Fli-1 Binds to the 5Ј EBS of SRE-1-ETS proteins have been demonstrated to be sequence-specific transcription factors. In vitro translated Fli-1 was shown to specifically bind the murine egr-1 at the 5Ј EBS of SRE-1 (39). To determine whether one or both EBSs are bound by endogenous Fli-1 in nuclear extracts from NFS60 cells, gel shifts using SRE-1 probes containing an intact CArG box and a mutation either in the 5Ј (L m R) or the 3Ј (LR m ) EBS were performed, using G-CSF-stimulated nuclear extracts (Fig. 7A). The L m R probe, in the presence of nonspecific competitor, failed to form band 1 (Fig. 7A, lane 5). The absence of band 1, which represents the Fli-1/SRF-probe com-FIG. 6. EMSA of nuclear proteins that bind a CArG mutant SRE-1 oligonucleotide. An SRE-1 labeled probe (0.1 g) either containing the wild-type or a 3-base mutation within the CArG box (CArG m/ BglII) was incubated with 15 g of G-CSF-stimulated nuclear extracts. A 200 M excess of unlabeled competitor sequences (nonspecific (NS), SRE-1, or CArG m ) or 2 g of SRF antibody, Fli-1 antibody, or IgG control was co-incubated with the samples. plex, occurs despite the presence of the wild-type 3Ј EBS. An excess of unlabeled L m R probe sequence as competitor resulted in competition of all bands (Fig. 7A, lane 6), and addition of SRF antibody, but not IgG control or Fli-1 antibody, resulted in a supershifted band (Fig. 7A, lanes 7-9). Therefore, our results suggest that Fli-1 binds the 5Ј EBS. The LR m mutant probe in the presence of nonspecific competitor formed gel shift bands 1-5 (Fig. 7A, lane 10). The presence of band 1 confirms that Fli-1 binds the 5Ј EBS. Addition of an excess of unlabeled LR m mutant probe sequence competed all the bands. Addition of SRF antibody resulted in the expected supershift pattern (Fig.  7A, lanes 12 and 14). However, despite the presence of an intact 5Ј EBS, there was no supershifted band with Fli-1 antibody (Fig. 7A, lane 13). The SRE-1 probe, upon the addition of SRF and Fli-1 antibody, resulted in supershifted bands (Fig. 7A,  lanes 2-4). These results were observed in two independent experiments. These results suggest that the 3Ј EBS may inhibit the Fli-1 antibody interaction with Fli-1.
To test our hypothesis that Fli-1 binds the 5Ј EBS of SRE-1, we performed EMSA experiments with the 5Ј EBS alone as the probe. Addition of Fli-1 antibody to G-CSF-treated nuclear extracts and the 5Ј EBS probe resulted in a supershifted band (Fig. 7B, lanes 9 -11), but the addition of Fli-1 antibody to diluent-treated extracts did not (Fig. 7B, lanes 3-5). We performed similar gel shift experiments using the 3Ј EBS as probe. In three independent experiments, addition of Fli-1 antibody to G-CSF-treated nuclear extracts and the 3Ј EBS probe resulted in either a weakly supershifted band or no supershifted band (Fig. 7C, lanes 4 -6). Thus, our data suggest that in G-CSFstimulated extracts, Fli-1 predominately binds the 5Ј EBS of SRE-1 and that SRF can bind independently of either EBS.

DISCUSSION
Growth factor induction of egr-1 expression is important for proliferative and differentiation responses in normal and leukemic myeloid cells. Studies indicating a role for Egr-1 in growth response have been reported (25, 48 -50). To elucidate the molecular events regulating myeloid cell proliferation, we examined the mechanism by which G-CSF-induced signals lead to an increase in egr-1 transcription. We identified an SRE complex bound by SRF and Fli-1, which mediates egr-1 induction in response to G-CSF. ETS proteins, in association with other DNA-binding proteins, have been previously suggested to play a role in hematopoiesis, including the regulation of genes involved in myeloid (22) and lymphoid cell development (51).
Mitogens and differentiating factors induce expression of various immediate early genes (egr-1, c-fos, c-jun, pip92, etc.) through the activation of SREs (32, 40 -46). The trans-activation of many early genes containing SREs is regulated by complexes composed of SRF and ETS protein family members (32, 40 -47). A ternary complex of SRF and Elk-1 or Sap1 bound to SRE has been shown to induce expression of both the c-fos (40 -42, 46) and egr-1 genes (47). ETS proteins generate selective transcriptional responses through their specific protein partnerships and their sequence-specific DNA binding (52). It has been shown that the in vitro-translated SRF and ETS proteins Elk-1, Sap1, Fli-1 and EWS-Fli-1 bind SRE-1 of the murine egr-1 promoter (39). Fli-1 and EWS-Fli-1 were demonstrated to exclusively bind the 5Ј EBS box of SRE-1, whereas Sap1 and Elk-1 bind both the 5Ј and 3Ј EBS (39). We have shown that endogenous SRF and Fli-1 proteins also bind the SRE-1 of the human egr-1 promoter, with Fli-1 predominately binding the 5Ј EBS (Fig. 7). The inability of Fli-1 antibody to produce a supershift with the LR m probe in our experiments may be due to a change in the DNA/protein conformation, which may inhibit the Fli-1 epitope from being exposed to the Fli-1 antibody. Furthermore, the 3Ј EBS as probe in EMSA experiments produced less intense gel shift bands as compared with the 5Ј EBS as probe (Fig. 7C). This may reflect less protein binding to the 3Ј EBS overall. Therefore, although Fli-1 may bind the 3Ј EBS, it does not appear to bind strongly, supporting our hypothesis that 5Ј EBS is the predominate Fli-1 binding site. We also did not identify SRE-1 binding by Elk-1 or Sap1 in EMSAs with nuclear extracts from NFS60 cells (data not shown).
The binding of ETS proteins to the SRE has been shown to occur in both an SRF-dependent and an SRF-independent manner. We demonstrated that SRF can bind an SRE-1 oligonucleotide probe in both unstimulated (data not shown) and G-CSF-stimulated nuclear extracts (Fig. 3B) and that that it can bind in the absence of either the 5Ј or 3Ј EBS (Fig. 7A). SRF binding to the SRE of the c-fos promoter is required for the recruitment of Elk-1 and Sap1 to the EBS sequences (53). However, binding of Elk-1 and Sap1 to the egr-1 promoter does not require prior assembly of an SRF/SRE binary complex (39). We have also demonstrated that endogenous Fli-1 protein does not require SRF for binding to the 5Ј EBS of egr-1 SRE, which is likely due to this sequence being a high affinity binding site for Fli-1.
Induction of egr-1 gene expression has been shown to be regulated by SREs during B-cell activation (54), myeloid cell proliferation (22), and monocyte (55) and preadipocyte (47) differentiation. In most studies, inducibility of egr-1 promoter activity has been localized to six SREs within -420 nt of the egr-1 promoter, with SRE-1 being the 5Ј-most element. Although most studies suggest that the individual SREs contribute equally to the induction of egr-1 expression (25,31,56,57), several studies have found regulatory roles for specific egr-1 SREs (54,55,58,59). Our data indicate that the 5Ј-most SRE (SRE-1), containing the central CArG box and flanking ETS boxes, is sufficient and necessary when isolated from the remainder of the egr-1 promoter sequences. However, experiments with constructs containing site-directed mutations of SRE-1 in the context of the -420 nt egr-1 promoter region indicated that SRE-1 was not required for maximal egr-1 expression (data not shown). Our results suggest that additional proteins recognizing SRE motifs 3Ј of SRE-1 between nt -369 and -326 or other elements may also associate with SREbinding proteins in response to upstream signals.
Transcriptional activation of other immediate early genes occurs through signaling cascades targeting SRF or its associated ETS proteins. SRF is phosphorylated in response to serum and growth factors, stress stimuli, and tumor-promoting agents (34,60,61). SRF phosphorylation has been demonstrated to increase SRF DNA binding affinity and to increase ternary complex formation with ETS proteins (33,34). However, several groups have reported that SRF constitutively binds the SRE, with no change in SRF binding patterns upon growth factor or serum stimulation (35)(36)(37)(38). The role of ternary complex factors Elk-1 and Sap1 as the direct targets of signaling cascades responsible for regulating early gene expression has been more extensively studied. Transcriptional activation of the c-fos promoter by growth factors and stress stimuli has been shown to be mediated through phosphorylation of Elk-1 and Sap1 by the mitogen-activated protein kinases extracellular signal-regulated kinase, c-Jun NH 2 -terminal kinase, and p38 (41,(62)(63)(64)(65)(66)(67)(68)(69). Recently, it was demonstrated that mitogen-activated protein kinase-independent kinases also phosphorylate Elk-1, thereby inducing SRF ternary complex trans-activation of the pip92 immediate early gene (32,44).
The results presented here suggest that SRF and Fli-1 and/or an unidentified ETS protein form a ternary or quaternary complex on SRE-1 of the egr-1 promoter, which is required for regulation of egr-1 transcription. We show that the egr-1 SRE-1 complex is bound by SRF, regardless of growth factor stimulation. This is consistent with constitutive binding of SRF to the c-fos promoter (35). Furthermore, we demonstrated that in supershift experiments, Fli-1 binds the SRE-1 very weakly in diluent-treated extracts and more strongly in G-CSF-stimulated extracts. This may reflect a posttranslational modification, such as phosphorylation, which increases the affinity of Fli-1 binding for SRE or which promotes Fli-1/SRF complexes to associate with other factors of the TFIID complex. We speculate that G-CSF induces changes in the electrophoretic mobility of SRF-containing complexes through the binding of a phosphorylated form of Fli-1 and/or SRF proteins. It has recently been demonstrated that calcium-dependent phosphorylation of Ets-1 inhibits Ets-1 DNA binding activity (70). However, phosphorylation of Fli-1 through the Ras/mitogenactivated protein kinase signaling pathway may potentially stimulate Fli-1 binding to SRE-1. The role of Fli-1 in G-CSF signal transduction is currently under investigation. In preliminary co-transfection experiments with a Fli-1 expression construct and pTE2-SRE-1, Fli-1 does not increase egr-1 transcription in NFS60 cells (data not shown). This may be due to rate-limiting kinases or to the fact that the pTE2-SRE-1 construct is maximally active at 3-fold stimulation.
We previously examined the signaling pathways that regulate myeloid cell proliferation in response to GM-CSF and IL-3, using the egr-1 gene as an end point (22). The receptors for GM-CSF and IL-3, like those for G-CSF, are members of the cytokine receptor superfamily and transduce signals through a common ␤ subunit. We reported previously (22) that phosphorylation of the cAMP response element-binding protein is critical for the induction of egr-1 in response to GM-CSF. In this study, we show that G-CSF activates egr-1 transcription through an SRE. Thus, our findings indicate that different signaling pathways converge on distinct promoter regions of egr-1, demonstrating a possible mechanism that determines specificity of myeloid growth factors and their actions on proliferation.