Different Domains of the Transcription Factor ELF3 Are Required in a Promoter-specific Manner and Multiple Domains Control Its Binding to DNA*

Elf3 is an epithelially restricted member of the ETS transcription factor family, which is involved in a wide range of normal cellular processes. Elf3 is also aberrantly expressed in several cancers, including breast cancer. To better understand the molecular mechanisms by which Elf3 regulates these processes, we created a large series of Elf3 mutant proteins with specific domains deleted or targeted by point mutations. The modified forms of Elf3 were used to analyze the contribution of each domain to DNA binding and the activation of gene expression. Our work demonstrates that three regions of Elf3, in addition to its DNA binding domain (ETS domain), influence Elf3 binding to DNA, including the transactivation domain that behaves as an autoinhibitory domain. Interestingly, disruption of the transactivation domain relieves the autoinhibition of Elf3 and enhances Elf3 binding to DNA. On the basis of these studies, we suggest a model for autoinhibition of Elf3 involving intramolecular interactions. Importantly, this model is consistent with our finding that the N-terminal region of Elf3, which contains the transactivation domain, interacts with its C terminus, which contains the ETS domain. In parallel studies, we demonstrate that residues flanking the N- and C-terminal sides of the ETS domain of Elf3 are crucial for its binding to DNA. Our studies also show that an AT-hook domain, as well as the serine- and aspartic acid-rich domain but not the pointed domain, is necessary for Elf3 activation of promoter activity. Unexpectedly, we determined that one of the AT-hook domains is required in a promoter-specific manner.

The ETS transcription factor family is comprised of nearly 30 members. They are characterized by a highly conserved DNA binding domain (DBD), 3 known as the ETS domain. During the past several years, interest in one of its family members, Elf3, has grown considerably. Elf3 (also known as Ese-1, ESX, ERT, and jen) belongs to the Elf subfamily of ETS proteins whose expression is restricted to epithelial tissues (1)(2)(3)(4). Gene knockout studies have shown that Elf3 plays essential roles during development, in particular the development of the gut. Thirty percent of Elf3 null fetuses die in utero, and the remaining 70% die shortly after birth. These mice exhibit severe alterations in the cellular architecture of the small intestine (5). Defects in the development and function of other tissues, especially those needed later in life, are likely to be masked by the early demise of these mice.
Other studies have directly implicated Elf3 in the normal physiology of the breast, as well as in breast cancer (3, 4, 6 -10). In the normal breast, Elf3 is believed to be expressed in a subset of pluripotent ductal epithelial cells, which are retained after involution of the breast (8). Given that pluripotent ductal epithelial cells are likely targets of carcinogenesis, it is not surprising that 40% of ductal carcinoma in situ samples display high levels of Elf3 expression (4,8). Furthermore, increased expression of Elf3 correlates closely with elevated expression of HER2/ neu in ductal carcinoma in situ (4). Interestingly, the disruption of Elf3 activity leads to significant decreases in HER2 expression, which strongly suggests that Elf3 influences HER2 expression (6). Although it is unclear whether the effect of Elf3 on HER2 is direct or indirect, Elf3 has been shown to directly regulate other promoters. This has been shown most clearly in the case of the type II TGF-␤-receptor (T␤R-II) promoter. Elf3 synergistically activates the T␤R-II promoter by binding to two overlapping ets sites located just downstream of the major transcription start site of the gene (11)(12)(13). Elf3 not only binds to the promoter of T␤R-II in vivo, both over-and underexpression studies have directly implicated Elf3 in the regulation of the endogenous T␤R-II gene (11,14,15).
Elf3 has also been implicated in the expression of at least 10 other genes, including the collagenase-1 gene (MMP-1, interstitial collagenase, and collagenase type I) (7), a gene important in extracellular matrix remodeling and tumorigenesis (16). The collagenase-1 promoter contains closely spaced AP-1 and ets sites, which allow cooperation between transcription factors (17,18). The cooperative action of Elf3 with other transcription factors has also been observed in the case of the nitric-oxide synthetase (NOS2) gene, where Elf3 cooperates with NF-B, by binding to adjacent ets and B sites (19). These and other studies demonstrate that Elf3 works cooperatively with other transcription factors to achieve specificity and to regulate genes involved in inflammation, differentiation, tumorigenesis, and metastasis (6, 7, 14, 19 -25).
Although there is growing evidence that Elf3 regulates a number of important genes that are involved in a wide range of cellular activities, Elf3 structure and function have not been studied in-depth. The work described in this study focuses on the domains of Elf3 required for DNA binding and the activation of gene transcription. Based in part on sequence analysis, Elf3 contains five defined domains (Fig. 1). The N terminus contains a pointed domain (PNT), which is located in residues 63-127 (1). This domain has been implicated in protein-protein interaction for other ETS proteins (26,27). Thus far, the PNT domain in Elf3 has not been assigned a functional role. The transactivation domain (TAD), which has been studied more extensively, is located in residues 129 -159 (6,28). Unlike TADs of many other transcription factors, the TAD of Elf3 appears to contain an ␣-helix (6), and point mutations in this ␣-helix disrupt its ability to activate promoter activity as well as its interaction with the coactivator Med23 (CRSP3, Drip130, and Sur2) and the TATA box-binding protein (TBP) (6,28). Elf3 also contains a serine-and aspartic acid-rich (SAR) domain located in residues 189 -229 (4,29), which is unique. This domain appears to influence cellular transformation when Elf3 is localized to the cytoplasm, but a nuclear function for this domain has not been identified (29). Unlike other ETS proteins, Elf3 appears to contain two AT-hook domains located within residues 236 -267, which have not been examined in detail. AT-hook domains were initially identified in high mobility group (HMG) family members, where they are believed to be involved in nonsequence-specific binding to AT-rich regions of DNA, as well as protein-protein interactions (30,31). This region of Elf3 also contains a bipartite nuclear localization signal (NLS) (29,32). Finally, the ETS domain, which is required for DNA binding, is located in residues 272-354 (1). The ETS domain of Elf3 has not been characterized, but it is expected to be structurally similar to those of other ETS family members (1).
Earlier studies argue that Elf3, like eight other ETS proteins, possesses an autoinhibitory domain (AID), which limits its binding to DNA (11,(33)(34)(35)(36)(37)(38). Autoinhibition is an important regulatory mechanism used to control biological processes. Many proteins are held in an autoinhibited state in vivo until the correct physiological conditions are met (39). For transcription factors, a common target of autoinhibition is DNA binding, and there is growing evidence that many transcription factors possess AIDs. Because the ETS domains of most, if not all, ETS proteins are likely to be very similar structurally and bind to similar DNA sequences in vitro, specificity is believed to be achieved, at least in part, through autoinhibition (39). Autoinhibition limits promiscuous binding of these transcription factors to gene regulatory regions in vivo in the absence of partner proteins or signaling events.
Currently, two different models have been proposed to explain how autoinhibition affects DNA binding of ETS family members. The model for Ets1-like autoinhibition posits that regions flanking its ETS domain fold together to form an inhibitory module, which interacts with the ETS domain and, as a result, lowers its affinity for DNA (40,41). In vivo, Ets1 autoinhibition is believed to be relieved when a partner protein interacts with either the autoinhibitory module or the ETS domain itself (42)(43)(44)(45)(46)(47)(48)(49). In contrast, a different model for autoinhibition has been proposed for Elk1 (50,51), which involves three separate regions (TAD, B-box, and R-domain) acting in a collective manner to regulate Elk1 activity through autoinhibition (50,51). The primary AID affecting DNA binding is within the TAD of Elk1, and phosphorylation of the TAD affects the domain-domain interaction within Elk1 for release of autoinhibition (50). Although previous studies argue that Elf3 contains an AID, it is unclear what region(s) of the protein is involved in limiting its binding to DNA (11) and how autoinhibition is regulated mechanistically.
To identify the regions of Elf3 that affect autoinhibition, as well as the regions that affect its activation of transcription, we created and tested an extensive series of Elf3 mutant proteins with specific domains deleted or targeted by point mutations. We determined that the PNT domain does not appear to be needed for Elf3 to activate the promoter of either the T␤R-II or collagenase-1 gene, whereas the SAR domain is needed for both promoters. Interestingly, we show that one of the AT-hook domains is required to activate transcription in a promoterspecific manner. We also determined that several regions outside the ETS domain influence Elf3 binding to DNA both in vitro and in vivo. On the basis of our findings, we present a model of autoinhibition that implicates intramolecular interactions in the creation of the autoinhibited state. Moreover, we propose that the release of autoinhibition is facilitated by interactions with a coactivator, which disrupts intramolecular interactions within Elf3 and increases the affinity of the ETS domain for DNA.

EXPERIMENTAL PROCEDURES
Materials-Dulbecco's modified Eagle's medium was obtained from Invitrogen. Fetal bovine serum was purchased from HyClone Laboratories (Logan, UT). Unless indicated otherwise, all chemicals were obtained from Sigma.
Cell Culture-As described previously, Dulbecco's modified Eagle's medium containing 10% fetal bovine serum was used in the culture of each cell line, including F9 embryonal carcinoma (EC) (11,52) and 293T (11). F9 EC cells were cultured on gelatin-coated tissue culture plastic and induced to differentiate by a 72-h treatment with 5 M retinoic acid, as described previously (11,52).
We utilized site-directed mutagenesis to create several point mutations and one deletion mutation of five amino acids (FLAG-Elf3⌬246 -250) within FLAG-Elf3 or GFP-Elf3 by the "QuickChange" method, as described previously (11). The primers, template, and restriction enzyme sites utilized to obtain several new expression plasmids are listed in Table 1. FLAG-Elf3⌬N173 was created using the primer pair 5Ј TTGAGGATCCCGCCAGGCCAGTCCCTACTAC 3Ј and 5Ј CGAGCTCTAGCATTTAGGTGACAC 3Ј. The primers amplified a PCR fragment containing residues 174 -371. Restriction sites for BamHI (underlined) and XbaI were used to directionally insert the fragment into the pcDNA3.0 FLAG-NLS vector, the same plasmid backbone of FLAG-Elf3. FLAG-Elf3⌬173-241 was created by utilizing PCR to amplify the entire FLAG-Elf3 plasmid except residues 173-241. We utilized primers that, upon digesting with the appropriate restriction enzyme and religation, resulted in the incorporation of the amino acid sequence TGDD between residues 172 and 242. FLAG-Elf3⌬235-270 was created by a similar method. PCR was used to amplify the entire plasmid except residues 235-270, and upon cutting with the appropriate restriction enzyme and religation to form a circular plasmid, the amino acid sequence TGTG was incorporated between residues 234 and 271. We used a similar method to create FLAG-Elf3⌬237-241, FLAG-Elf3⌬237-241⌬246 -250, FLAG-Elf3⌬235-260, FLAG-Elf3⌬129 -159, FLAG-Elf3⌬60 -127, and FLAG-Elf3⌬355-362 by using the restriction enzyme XhoI and the primer pairs and plasmid templates that are listed in Table 2. FLAG-Elf3⌬C235 was created in making FLAG-Elf3⌬235-270 by a frameshift mutation that inserted five amino acids, TGDDR, and a stop codon.
FLAG-Elf3⌬60-127, FLAG-Elf3⌬235-260, FLAG-Elf3⌬C363, FLAG-Elf3⌬355-362, FLAG-Elf3⌬C354, FLAG-Elf3⌬173-241, FLAG-Elf3⌬235-270, FLAG-Elf3⌬N270, and FLAG-Elf3⌬N173 were utilized to make their GFP fusion counterparts. By digesting with BamHI and XbaI, the protein coding sequence of the Elf3 mutant was isolated from the pcDNA3.0 FLAG/NLS expression plasmid and ligated into the GFP plasmid vector (Clontech). The coding sequence was in-frame with the GFP coding region. The sequences of these Elf3 expression constructs, as well as the ones mentioned previously, were verified by DNA sequence analysis at the Genomic Core Research Facility (University of Nebraska, Lincoln) or the Molecular Biology Core Facility of the Eppley Institute (Omaha, NE).
Transfection and Activities of P/R Gene Constructs-F9-differentiated cells were transiently transfected by the calcium phosphate precipitation method, as described previously (11). In addition to the indicated amount of P/R gene construct, the cells were transfected with 1 g of the internal control plasmid, pCMV-␤-gal, to normalize for any differences in transfection efficiency or general effects on transcription. Chloramphenicol acetyltransferase (CAT) and ␤-galactosidase activities were determined 72 h after transfection of F9-differentiated cells, as described previously (55). Data shown are means Ϯ S.D. for duplicate samples.
Transfection and Expression of GFP Fusion Proteins-F9-differentiated cells were transiently transfected with expression vectors for GFP-Elf3 or mutant proteins using Lipofectamine 2000 (Invitrogen) as described previously (53). This transfection protocol was utilized because transfection efficiencies are very similar from plate to plate (53). Expression of the fluorescent proteins was determined after 24 h by using a FACSCalibur flow cytometer (BD Biosciences). The data were collected   (60), and 1FLI (61), respectively).
Extract Preparation and Western Blotting-293T cells were transiently transfected by calcium phosphate precipitation with 20 g of the relevant expression plasmid. Forty eight hours after transfection, the cells were harvested, and nuclear extracts were prepared using the Pierce NE-PER TM nuclear and cytoplasmic extraction kit, as described in the manufacturer's protocol (Pierce). Extracts were supplemented with protease and phosphatase inhibitors and stored at Ϫ80°C. Western blot analysis was performed using the anti-FLAG M2 antibody against blots in which 10 l of each nuclear extract was separated by SDS-PAGE. Proteins were detected using the enhanced chemifluorescence (ECF) kit (GE Healthcare) and scanned on a Storm PhosphorImager (GE Healthcare). Quantitation was performed using the ImageQuant 5.0 analysis software (GE Healthcare).
Electrophoretic Mobility Shift Assays-To use equal amounts of each FLAG tagged Elf3 protein, the volume of each nuclear extract added was adjusted based on Western blot analysis. Annealed double-stranded oligodeoxynucleotide (ODN) probes (see below) were labeled by a fill-in reaction using the Klenow fragment of DNA polymerase I (New England Biolabs, Beverly, MA). Nuclear extracts were incubated for 20 min at room temperature in binding buffer containing 10 mM Tris, pH 7.5, 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, 0.8% polyvinyl alcohol, 10% bovine serum albumin, ϳ40,000 cpm of probe, and 0.02 g/l of mutETS-annealed doublestranded ODN. The mutETS ODN was utilized in place of traditional nonspecific competitors to reduce nonspecific binding and to allow better visualization of protein-DNA complexes. In supershift assays, the monoclonal antibody to the FLAG epitope (M2) was preincubated with the nuclear extract for 40 min on ice before addition of the labeled probe. DNA-protein complexes were resolved on a 4% nondenaturing polyacrylamide gel in 0.5ϫ Tris/glycine/EDTA buffer as described previously (52). The gels were dried and visualized by PhosphorImager (GE Healthcare). ODN probes corresponding to ϩ3 to ϩ34 of the T␤R-II gene were as follows: wild type, 5Ј TGGCG-AGGAGTTTCCTGTTTCCCTCTCGGCGC 3Ј and its complement; mutETS, 5Ј TGGCGAGGAGTTTGAATTCACCCT-CTCGGCGC 3Ј and its complement (underline indicates the two overlapping ets sites, and italicized letters indicate the mutated sequence). In the case of the wild-type probe, the sequence TAGC was added to the 5Ј end of each ODN for use in radiolabeling.
Chromatin Immunoprecipitation (ChIP) Assay-ChIP assays with 293T cells were carried out as described previously (11). Briefly, 24 h after seeding, cells were cotransfected by the cal-cium phosphate precipitation method with the mT␤R-II Ϫ108/ ϩ56(B/A) P/R construct and FLAG tagged expression vectors for Elf3 proteins. Approximately 42 h after transfection, proteins were cross-linked with formaldehyde, and then nuclei were isolated and lysed, and chromatin was sheared by sonication. The supernatant was cleared by protein G-agarose (preblocked with salmon sperm DNA and bovine serum albumin; Upstate, Charlottesville, VA) for 1 h at 4°C. The sample was then split into aliquots for incubation with the M2 antibody (FLAG-specific; Sigma) or the Gal4 DBD antibody (nonspecific; Santa Cruz Biotechnologies, Santa Cruz, CA) overnight at 4°C. Protein complexes were collected with protein G-agarose (Upstate) and washed extensively. Complexes were eluted, decross-linked, and treated with RNase A and proteinase K, and the DNA was purified using the Gene Clean Turbo kit (Qbiogene, Irvine, CA). DNA enrichment was determined by quantitative real time PCR analysis using a Cepheid Smart Cycler Version 2.0c Detection System, and SYBR Green Master Mix (SuperArray, Frederick, MD). Relative occupancy values were calculated by normalizing the amount of specific DNA (mT␤R-II Ϫ108/ϩ56(B/A) P/R) immunoprecipitated to the amount of nonspecific DNA (GAPDH promoter) immunoprecipitated and evaluating enrichment of the specific antibody, M2, as compared with the nonspecific antibody, Gal4. mT␤R-II Ϫ108/ϩ56(B/A) P/R-specific primers were as published previously (11), and GAPDH primer pairs were supplied in the EZ ChIP kit from Upstate Biotechnology, Inc. The sequence is as follows: forward 5Ј TACTAGCGGTTTTACGGGCG 3Ј and reverse 5Ј TCGAACAGGAGGAGCAGAGAGCGA 3Ј.
FLAG Pulldown Assay-293T cells were cotransfected with 10 g of an expression vector for a FLAG tagged protein, 5 g for a GFP fusion protein, and 10 g of the coactivator expression plasmid as indicated. Vector DNA was added, if needed, to equal 25 g of total DNA. Twenty four hours after transfection cells were frozen on liquid nitrogen and solubilized in a cell lysis solution containing 50 mM Hepes, pH 7.8, 1% Triton X-100, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 mM NaF, 1 mM vanadate, 30 mM NaPP i , 25 mM benzamidine, and 20 g/ml aprotinin. Insoluble matter was removed by centrifugation for 10 min at 14,000 rpm. Input fractions were removed for analysis of protein expression. Then FLAG proteins were immunoabsorbed by EZ-Red M2 affinity gel overnight at 4°C. The affinity gel was then washed three times with RIPA buffer and one time with a solution containing 10 mM Hepes, pH 7.8, 1.5 mM MgCl 2 , 0.1 M NaCl, 0.05% Triton X-100. M2-specific complexes were eluted using 150 ng/l of 3ϫ FLAG peptide in the previous solution at 4°C. Western blot analysis was utilized to detect the amount of GFP-tagged protein coimmunoprecipitated by FLAG-Elf3⌬C235. The M2 and GFP (sc-8334; Santa Cruz Biotechnology) antibody were used at a 1:2000 dilution in 3% bovine serum albumin in TBST. Proteins were detected using the enhanced chemifluorescence (ECF) kit (GE Healthcare) and scanned on a Storm PhosphorImager (GE Healthcare). Quantitation was performed using the ImageQuant 5.0 analysis software (GE Healthcare).

The ETS Domain of Elf3 Is Similar to the ETS Domain of Ets1-
Transcription factors are composed of modular domains that assist with biological functions, including transactivation and DNA binding. Elf3 contains five domains defined by homology or functional studies (Fig. 1A). To examine the contribution of each Elf3 domain to the overall function of the protein, we created a large number of Elf3 mutant proteins. Initially, we addressed the function of the ETS domain. This domain is highly conserved within the ETS transcription factor family and has been well characterized structurally for other family members but not for Elf3. To examine the structure of the ETS domain of Elf3, we generated a homology model of its ETS domain using PDB structural files corresponding to the ETS domains of GABP␣, Sap-1, Ets1, Pdef, and Elf5 as templates. Pdef and Elf5 ETS domains share 42 and 66% sequence similarity with the ETS domain of Elf3, respectively. The resulting molecular model of Elf3 contained three ␣-helices, four ␤-sheets, and a turn connecting helix 2 and 3 ( Fig.  1B, dark gray). This structure corresponds to the winged helix-loop-helix structure of the typical ETS domain, with the third helix as the DNA recognition helix. The molecular model of the ETS domain of Elf3 when overlaid with the crystal structure of the ETS domain of Ets1 ( Fig. 1B, white) demonstrated high similarity (the root mean square deviation between the backbone atoms was equal to 0.46 Å). This suggests that although the primary sequences of the ETS domains of Elf3 and Ets1 share only a 35% similarity, the tertiary structure of the ETS domain is highly conserved.
Hence, it appears the tertiary structure of the ETS domain is more highly conserved evolutionarily than its amino acid sequence would suggest.
To test our molecular model for the ETS domain of Elf3, proline substitutions were introduced to disrupt the predicted ␣-helices. We hypothesized that these mutations should disrupt the secondary structure of the ETS domain and disrupt its activity. A proline was inserted in the putative helix 1 (Leu 275 -Leu 283 ) in place of isoleucine 279 by site-directed Elf3 also contains a putative NLS from residues 237-267 and putative nuclear export signal from residues 102-112. B, molecular model of the Elf3 ETS domain (dark gray) is overlaid with the Ets1 ETS domain structure (white) that is based on crystallographic data. The model contains three ␣-helices, four ␤-sheets, and a turn connecting helix 2 and 3. This is the typical description of the ETS domain. The ␣-helices of the Elf3 ETS domain are predicted to be located in residues Leu 275 -Leu 283 (helix 1), Ser 308 -Lys 318 (helix 2), and Tyr 326 -Tyr 337 (helix 3). C, F9-differentiated cells were transiently transfected with 15 g of the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct. Where indicated, the cells were also transfected with 0.5 or 1 g of an N-terminal GFP fusion expression vector for Elf3, Elf3I279P, Elf3A312P, Elf3K320E, or Elf3R331P. Activities were assayed and normalized as described under "Experimental Procedures." The promoter activity of each construct is calculated relative to the CAT activity observed with mT␤R-II Ϫ108/ϩ56(B/A) alone, which was set to 1. This experiment was performed three times, and similar results were obtained. D, top, diagram of FLAG-Elf3 deletion constructs. Elf3 represents the full-length protein (aa 1-371) with the FLAG tag and SV40 NLS N-terminal to the coding region. Residues 1-270 were deleted from Elf3 to create Elf3⌬N270. Bottom, equal amounts of the FLAG-Elf3⌬N270 protein indicated above the lane were studied by EMSA with a radiolabeled ODN probe containing the wild-type ets-binding sites from ϩ3 to ϩ34 T␤R-II gene promoter. Mock lane contains nuclear extract from 293T cells not transfected with an expression vector. Supershift analysis (indicated above the lane with a ϩ sign) was performed by incubating the binding reaction for 40 min at 4°C with the anti-FLAG antibody M2 before adding the probe. The EMSA was analyzed as described under "Experimental Procedures." This experiment was repeated three times, and similar results were obtained. The position of the free probe (F) is indicated. E, ChIP was performed on 293T cells transiently cotransfected with the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct and the indicated expression plasmid. The M2 antibody against the FLAG epitope was utilized for immunoprecipitation. Fold enrichment measured by quantitative real time PCR of the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct and normalized to enrichment of the endogenous human GAPDH promoter is shown. FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5 JOURNAL OF BIOLOGICAL CHEMISTRY 3031 mutagenesis. Likewise, the putative helix 2 (Ser 308 -Lys 318 ) and the putative recognition helix, helix 3 (Tyr 326 -Tyr 337 ), were disrupted by substituting proline for alanine 312 and arginine 331, respectively. We initially assessed the effect of these mutations on the ability of Elf3 to stimulate the P/R gene construct mT␤R-II Ϫ108/ϩ56(B/A) transiently transfected into F9-differentiated cells. We utilized F9-differentiated cells, which are derived from F9 EC cells after treatment with retinoic acid, because these cells have been characterized extensively and used to examine gene regulatory mechanisms (62,63). Notably, expression of the T␤R-II gene, including its regulation by Elf3, has been characterized extensively in the F9-differentiated cells (11,12). Furthermore, previous studies have shown that the T␤R-II gene and the Elf3 gene are up-regulated upon differentiation of F9 EC cells (11,52). To test the ETS domain mutants, we used GFP-Elf3 fusion proteins, which we have shown previously stimulates mT␤R-II Ϫ108/ϩ56(B/A) in a manner similar to FLAG-Elf3 (12). The use of the fusion proteins demonstrated that these particular ETS domain mutations did not affect nuclear localization (data not shown) or expression levels (data not shown). In the presence of increasing amounts of the parent GFP-Elf3 protein, the activity of the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct increased 4 -6-fold over the basal activity of the P/R gene construct (Fig. 1C). As predicted, all three helix mutations (GFP-Elf3I279P, GFP-Elf3A312P, and GFP-Elf3R331P) resulted in a complete loss of Elf3 stimulation of mT␤R-II Ϫ108/ϩ56(B/A). In contrast, other non-helix breaking mutations within the turn between helix 2 and 3 (K320E) and within the second and third helix (S308A and S330A, respectively) did not greatly affect the ability of Elf3 to stimulate mT␤R-II Ϫ108/ ϩ56(B/A) ( Fig. 1C and data not shown). This suggested that the loss of activity is because of a disruption of the ability of the ETS domain to bind to DNA as a result of structural abnormalities.

Functional Domains of ELF3
To more directly test the ability of these mutations to affect DNA binding, the same point mutations were made within a mammalian expression vector of an N-terminally truncated form of Elf3 (Elf3⌬N270) (Fig. 1D, top). Earlier studies demonstrated that Elf3 contains an AID, which limits its binding to DNA in vitro, because Elf3⌬N270, which contains only the ETS domain and the 17-amino acid C-terminal tail of Elf3, displays increased affinity for DNA in vitro (11). In light of the limited binding of the full-length protein, we utilized the truncated protein Elf3⌬N270, which forms two specific complexes because of binding to both mT␤R-II ets sites, to examine how the mutations affected DNA binding independent of the possible influence of an AID. EMSA using equal amounts of Elf3⌬N270 with or without the ETS domain mutations detected a loss of binding with mutants Elf3⌬N270I279P, Elf3⌬N270A312P, and Elf3⌬N270R331P to the T␤R-II probe but not with the Elf3⌬N270K320E mutant (Fig. 1D). We validated the results for the I279P mutant by ChIP. Specifically, we demonstrated that the mutant protein, I279P, in contrast to its wild-type counterpart (Elf3⌬N270), exhibited little or no association with the T␤R-II promoter in vivo (Fig. 1E). These findings support the premise that the predicted ␣-helical regions, which would be disrupted by prolines, are necessary for the binding of Elf3 to DNA. Moreover, these findings also suggest that the winged helix-loop-helix structure of the ETS domain is conserved within Elf3 even though the sequence homology between Ets1 and Elf3 is relatively low.
Requirement for Regions Flanking the ETS Domain-Regions flanking the defined ETS domain have also been reported to affect the ability of ETS proteins to bind to DNA. For example, removal of the C-terminal tail of Ets1 or Ets2, which flanks the C-terminal side of the ETS domain, results in enhanced binding to DNA in vitro (33). To test the possibility that removing the C-terminal tail of Elf3 would increase binding to DNA and thus increase its ability to activate transcription, we inserted a stop codon at amino acid 355 and created Elf3⌬C354 (Fig. 2A). However, when increasing amounts of Elf3⌬C354 and wild-type Elf3 were compared, Elf3⌬C354 failed to stimulate mT␤R-II Ϫ108/ ϩ56(B/A) (Fig. 2B). Using a GFP-Elf3⌬C354 fusion protein, we A, diagram of FLAG-Elf3 deletion constructs. Elf3 represents the full-length protein (aa 1-371). Residues 355-362, 363-371, or 354 -371 were deleted from Elf3 to create Elf3⌬355-362, Elf3⌬C363, or Elf3⌬C354, respectively. B, F9-differentiated cells were transiently transfected with 15 g of the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct. Where indicated, the cells were also transfected with expression vectors for Elf3 or Elf3⌬C354 at the amounts shown. Activities were assayed and normalized as described under "Experimental Procedures." The promoter activity of each construct is calculated relative to the CAT activity observed with mT␤R-II Ϫ108/ϩ56(B/A) alone, which was set to 1. This experiment was performed two times, and similar results were obtained. C, F9-differentiated cells were transiently transfected and assayed as in B, except where indicated, and the cells were transfected with 0.5 or 1 g of an N-terminal GFP fusion expression vector for Elf3, Elf3⌬355-362, or Elf3⌬C363. This experiment was performed three times, and similar results were obtained. determined the Elf3⌬C354 protein was present in the nuclear compartment and that it was expressed at levels similar to wildtype Elf3 (data not shown). There are two possible explanations for the lack of activity of Elf3⌬C354 as follows: 1) removal of the C-terminal tail eliminates part of the protein needed for transactivation; 2) removal of the C-terminal tail resulted in a protein unable to bind DNA. We tested these possibilities by determining whether Elf3⌬C354 acts as a dominant negative. In this regard, a protein that is able to bind to the ets sites, but is unable to activate transcription, will compete with wild-type Elf3 for binding to the T␤R-II promoter and interfere with its ability to activate the promoter. We have shown a similar dominant negative activity for an N-terminally truncated Elf3 protein (Elf3⌬N270) (12). For this purpose, we transfected F9-differentiated cells with increasing amounts of mutant Elf3⌬C354 in the presence of a fixed amount of the wild-type Elf3. Elf3⌬C354 did not reduce the activation of mT␤R-II Ϫ108/ϩ56(B/A) by wild-type Elf3 (Fig. 2B). This argues that removal of the C-terminal 17 amino acids of Elf3 disrupts its ability to bind to DNA in vivo.
To elucidate which residues of the C-terminal tail were critical for maintaining the ability of Elf3 to bind to DNA, we created two smaller deletions, Elf3⌬355-362 and Elf3⌬C363 ( Fig.  2A). Removal of the last eight amino acids of Elf3 (Elf3⌬C363) reduced the ability of Elf3 to activate transcription from the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct by 30 -40% (Fig. 2C). However, when adjusted for GFP-Elf3⌬C363 expression, which was ϳ20 -30% lower than wild-type GFP-Elf3 (data not shown), the activity of Elf3⌬C363 appears to be only about 20% lower than Elf3. In contrast, Elf3⌬355-362, which lacks the 8 amino acids adjacent to the ETS domain, failed to stimulate mT␤R-II Ϫ108/ϩ56(B/A) (Fig. 2C). Furthermore, Elf3⌬355-362, like Elf3⌬C354, did not act as a dominant negative, was localized in the nucleus (data not shown), and was expressed at levels similar to wild-type Elf3 (data not shown). We also examined the ability of Elf3⌬355-363 to occupy the T␤R-II promoter in vivo by ChIP, and we found that its presence is greatly reduced when compared with wild-type Elf3 (Fig. 1E). Taken together these findings suggest that Elf3⌬355-362 lacks amino acids essential for the activity of the ETS domain. Hence, amino acids directly flanking the C terminus of the ETS domain of Elf3 appear to be critical for binding to DNA.
We also tested whether residues flanking the N-terminal side of the ETS domain were necessary for maintaining binding to DNA. Recent studies indicated that this region contains a bipartite NLS (29,32). Indeed, we confirmed by fluorescent microscopy using GFP fusion Elf3 proteins that removal of residues 235-260 or 235-270 resulted in a protein that was primarily located within the cytoplasm (data not shown). To overcome the loss of nuclear localization, we created the mutations ⌬235-260 and ⌬235-270 within a mammalian Elf3 expression vector that contained the SV40 NLS and a FLAG tag at the N terminus of Elf3 (FLAG-Elf3⌬235-260 or FLAG-Elf3⌬235-270) (Fig. 3A). By using Western blot analysis, we confirmed that the resulting proteins were located within nuclear extracts at similar levels (data not shown). FLAG-Elf3⌬235-260 displayed 70 -80% of wild-type FLAG-Elf3 activity when tested with mT␤R-II Ϫ108/ϩ56(B/A) (Fig. 3B). Unexpectedly, FLAG-Elf3⌬235-270 retained only 30% of the activity of wild-type Elf3 (Fig. 3B). This raised the question of whether removal of residues 235-270, like the removal of 355-362, disrupts the ability of Elf3 to bind to DNA. To address this question, we utilized the FLAG-Elf3⌬235-270 expression construct and tested its ability to disrupt wild-type Elf3 stimulation of mT␤R-II Ϫ108/ ϩ56(B/A) and act as a dominant negative. However, when FLAG-Elf3⌬235-270 was added in addition to Elf3, it did not reduce the ability of wild-type Elf3 to stimulate mT␤R-II Ϫ108/ ϩ56(B/A) (Fig. 3C). Thus, FLAG-Elf3⌬235-270 does not appear to act as a dominant negative. These results suggest that removal of residues 235-270, similar to removal of residues 355-362, influences the ETS domain of Elf3 and results in a A, diagram of FLAG-Elf3 deletion constructs. Elf3 represents the full-length protein (aa 1-371). Residues 235-260 or 235-270 were deleted from Elf3 to create Elf3 expression constructs FLAG-Elf3⌬235-260 or FLAG-Elf3⌬235-270, respectively. Each of the constructs also contained an SV40 NLS. B and C, F9-differentiated cells were transiently transfected with 15 g of the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct. Where indicated, cells were also transfected with 0.5 or 1 g of FLAG-Elf3, FLAG-Elf3⌬235-260, or FLAG-Elf3⌬235-270 expression constructs. Activities were assayed and normalized as described under "Experimental Procedures." The promoter activity of each construct is calculated relative to the CAT activity observed with m T␤R-II Ϫ108/ϩ56(B/A) alone, which was set to 1. Components of this experiment were performed three times, and similar results were obtained. C, this experiment was performed two times, and similar results were obtained. FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5

JOURNAL OF BIOLOGICAL CHEMISTRY 3033
protein that exhibits reduced binding to DNA in vivo. Collectively, these findings argue that residues (aa 261-270 and aa 354 -362) that flank the canonical ETS domain (aa 272-354) are necessary for a functional ETS domain.
The SAR Domain but Not the PNT Domain Is Needed for Elf3 Activity-In view of the importance of the regions flanking the ETS domain, we also examined the contribution of the N-terminal domains to the activity of Elf3. The TAD (aa 129 -159) of Elf3 has been extensively characterized for its ability to stimulate promoter activity (6,28). However, the contribution of the PNT and SAR domains to the transcriptional activity of Elf3 has not been examined. Therefore, we created Elf3 expression vectors in which the PNT (Elf3⌬60 -127) or SAR (Elf3⌬173-241) domains were removed (Fig. 4A). We then examined the contribution of each domain to the ability of Elf3 to stimulate the T␤R-II promoter. Removal of the PNT domain did not affect the ability of Elf3 to stimulate the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct (Fig. 4B). This is not entirely unexpected given that the residues in Ets1 required for recognition and phosphorylation by ERK are not conserved in Elf3 (1). Interestingly, the effect of this mutation may be slightly underestimated, because the expression of this protein was ϳ20 -30% lower than wildtype Elf3 (Fig. 4C). In contrast to the PNT domain, removal of the SAR domain did not affect the expression level of Elf3 (Fig.  4C), but it significantly decreased the ability of Elf3 to stimulate transcription, ϳ30 -40% (p Ͻ 0.05) upon comparing multiple experiments (Fig. 4B). Taken together, these findings argue that Elf3 does not require the PNT domain to stimulate the promoter of the T␤R-II gene, whereas the SAR domain exerts a subtle, but significant, effect on Elf3 transactivation.
The Collagenase-1 Promoter Requires Different Elf3 Domains-To determine whether the findings described above with mT␤R-II Ϫ108/ϩ56(B/A) apply to another Elf3-responsive promoter, we examined the effects of the Elf3 mutants on the human collagenase-1 gene promoter. The collagenase-1 gene, like the T␤R-II gene, is up-regulated upon differentiation of the F9 EC cells with retinoic acid (64). The collagenase-1 promoter contains a single ets site located at Ϫ85 to Ϫ88 relative to the transcription start site, which responds to ETS family members (17,65). In HeLa cells, Elf3 has been shown to increase transcription of a CAT reporter gene that was driven by the Ϫ516/ϩ63 region of the collagenase-1 gene promoter (7). In F9-differentiated cells, Elf3 stimulated the collagenase-1 P/R construct as much 15-20-fold (Fig. 5). As with the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct, removal of the PNT domain (Elf3⌬60 -127) did not affect the stimulation of the collagenase-1 P/R construct by Elf3 (Fig. 5). Similarly, removal of the residues flanking the ETS domain of Elf3 (aa 355-362, 235-270, or 354 -371) resulted in a loss of stimulation of the collagenase-1 P/R construct (data not shown). In addition, Elf3⌬173-241 exhibited a decrease in the ability to stimulate the collagenase-1 promoter when compared with wild type (Fig. 5). Interestingly, Elf3⌬173-241 is 60 -70% less active with the collagenase-1 promoter but only 30 -40% less active with the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct (Fig. 4B). These findings suggest that the SAR domain may play a more important role in the activation of the collagenase-1 promoter.
Intriguingly, removal of residues 235-260 from Elf3 resulted in a more dramatic decrease with the collagenase-1 promoter than with the mT␤R-II promoter. FLAG-Elf3⌬235-260, which contains an exogenous NLS, retains 70% of the wild-type Elf3 activity with the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct (Fig.  6A). However, nearly all activity is lost when FLAG-Elf3⌬235-260 was tested with the collagenase-1 P/R construct (Fig. 6A). Our findings argue that residues 235-260 are essential for activation of the collagenase-1 promoter but not for the mT␤R-II promoter. When comparing the two promoters, it was noted that the collagenase-1 promoter contains a single ets site in close proximity to an AP-1 site; however the mT␤R-II promoter were deleted from Elf3 to create Elf3⌬60 -127 or Elf3⌬173-241, respectively. B, F9-differentiated cells were transiently transfected with 15 g of the mT␤R-II Ϫ108/ϩ56(B/A) P/R construct. Where indicated, the cells were also transfected with 0.5 or 1 g of an N-terminal GFP fusion Elf3, Elf3⌬60 -127, or Elf3⌬173-241 expression vector. Activities were assayed and normalized as described under "Experimental Procedures." The promoter activity of each construct is calculated relative to the CAT activity observed with mT␤R-II Ϫ108/ϩ56(B/A) alone, which was set to 1. This experiment was performed four times, and similar results were obtained. Analysis of the data by the Student's t test indicated that the activity of Elf3⌬173-241 was significantly less than the activity of Elf3 with a p value of less than 0.05. C, F9-differentiated cells were transiently transfected with 1.5 g of the GFP fusion expression vector indicated below the bar using Lipofectamine 2000, and GFP fluorescence was measured by flow cytometry, as described under "Experimental Procedures." Level of expression is based on the intensity of GFP per cell. Data shown are means Ϯ S.D. for quadruplicate samples from one representative experiment. This experiment was performed three times, and similar results were obtained. possesses two overlapping ets sites. Given this difference, we hypothesized that residues 235-260 are needed in cases where a single ets site is present, although this region is not necessary for regulatory regions that contain two adjacent ets sites. To test this hypothesis, we tested the mT␤R-II P/R construct in which one of the ets sites was removed. Specifically, we utilized the mT␤R-II Ϫ108/ϩ56(B/x) P/R construct in which the nonconsensus A site is mutated (12). mT␤R-II Ϫ108/ϩ56(B/x), which retains the consensus B site, was ϳ40% less active than the wild-type mT␤R-II Ϫ108/ϩ56(B/A) P/R construct (12). In this study, we directly compared the ability of FLAG-Elf3 and FLAG-Elf3⌬235-260, which both contain an exogenous NLS, to stimulate mT␤R-II Ϫ108/ϩ56(B/A) and mT␤R-II Ϫ108/ ϩ56(B/x). As shown earlier (Fig. 3B), FLAG-Elf3⌬235-260 displayed ϳ60 -70% of wild-type FLAG-Elf3 activity on the wildtype mT␤R-II Ϫ108/ϩ56(B/A) P/R construct (Fig. 6B). However, FLAG-Elf3⌬235-260 displayed only about 25% of wild-type FLAG-Elf3 activity when tested with the mutant mT␤R-II Ϫ108/ϩ56(B/x) P/R construct (Fig. 6B). Taken together, these experiments argue that residues 235-260 are important for stimulation of promoters that contain only a single ets site.
Promoter-specific Effects of an AT-hook Domain of Elf3-Elf3 contains two AT-hook motifs from residues 237-241 and 246 -250, which are identical to the HMG-1 and HMG-I/Y protein AT-hooks, respectively (1). This protein motif is unique to Elf3 within the ETS transcription factor family (1), but it is found within other transcription factors, mainly HMG proteins. To further elucidate the reason why Elf3⌬235-260 exhibits reduced activity with the collagenase-1 gene P/R construct, we deleted each of the AT-hook domains with five amino acid deletions either alone or in combination (FLAG-Elf3⌬237-241, FLAG-Elf3⌬246 -250, or FLAG-Elf3⌬237-241⌬246 -250) (Fig. 7A). These domains are not necessary for Elf3 stimulation of mT␤R-II Ϫ108/ϩ56(B/A) (data not shown and Fig. 7B). Moreover, FLAG-Elf3⌬237-241, which removed the first AThook, stimulated the collagenase-1 P/R construct to levels similar to the wild-type FLAG-Elf3 (Fig. 7C). In contrast, FLAG-Elf3⌬246 -250 exhibited only about 20 -30% of the activity of wild-type FLAG-Elf3 (Fig. 7C). These findings argue that the second AT-hook plays a significant role in allowing Elf3 to stimulate the collagenase-1 promoter.
Dual Roles of TAD-Multiple regions of Elf3 affect binding to DNA. Earlier in this study, we presented evidence that the ETS domain, as well as the regions directly flanking the ETS domain, are necessary for Elf3 to bind to DNA. Furthermore, previous studies suggested that Elf3 also contains an AID, which negatively affects the ability of Elf3 to bind to DNA in vitro (11,12). AIDs are not unusual in the ETS transcription factor family, and the best described are found in Elk1 and Ets1. To further characterize the AID of Elf3, which alters its in vitro binding to FIGURE 5. Elf3⌬173-241 does not stimulate the human collagenase-1 promoter. F9-differentiated cells were transiently transfected with 5 g of the collagenase-1 Ϫ517/ϩ63 P/R construct. Where indicated, the cells were also transfected with 1 or 3 g of Elf3, Elf3⌬60 -127, or Elf3⌬173-241 expression vector. Each of the expression vectors for Elf3 contained an exogenous SV40 NLS to ensure entry into the nucleus. Activities were assayed and normalized as described under "Experimental Procedures." The promoter activity of each construct is calculated relative to the CAT activity observed with collagenase-1 Ϫ517/ϩ63 alone, which was set to 1. These experiments were performed three times, and similar results were obtained.

FIGURE 6. Residues 235-260 of Elf3 are needed for activation of the collagenase-1 promoter and the T␤R-II promoter with one ets-binding site.
A, F9-differentiated cells were transiently transfected with 15 g of mT␤R-II Ϫ108/ϩ56(B/A) (black bars) or 5 g of the collagenase-1 Ϫ517/ϩ63 (white bars) P/R constructs. Where indicated, the cells were also transfected with 1 or 3 g of FLAG-Elf3 or FLAG-Elf3⌬235-260 expression vector. The promoter activity of each construct is calculated relative to the CAT activity observed with mT␤R-II Ϫ108/ϩ56(B/A) or collagenase-1 Ϫ517/ϩ63 alone, each of which was set to 1. B, F9-differentiated cells were transiently transfected with 15 g of the mT␤R-II Ϫ108/ϩ56(B/A) (black bars) or mT␤R-II Ϫ108/ϩ56(B/x) (white bars) P/R construct. Where indicated, the cells were transfected with 0.5 or 2 g of FLAG-Elf3 or FLAG-Elf3⌬235-260 expression vector. The promoter activity of each construct is calculated relative to the CAT activity observed with mT␤R-II Ϫ108/ϩ56(B/A) or mT␤R-II Ϫ108/ϩ56(B/x) alone, each of which was set to 1. A and B, activities were assayed and normalized as described under "Experimental Procedures." These experiments were performed three times, and similar results were obtained. All of the expression vectors contain an exogenous SV40 NLS to ensure entry into the nucleus. FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5 DNA, we utilized a FLAG tagged Elf3 mammalian expression construct to express full-length Elf3 in 293T cells. We prepared nuclear extracts from these cells and utilized these extracts to test the binding of Elf3 to a double-stranded ODN probe containing the ϩ3 to ϩ34 region of the T␤R-II promoter. This region contains two overlapping ets-binding sites to which Elf3 had been shown to bind in vivo (11). However, full-length Elf3 binds poorly to a DNA probe containing these sites in vitro in an EMSA (Fig. 8B) (11). Remarkably, upon addition of the anti-FLAG antibody, which recognizes the FLAG tag at the N terminus of Elf3, a significant increase in binding was observed (Fig. 8B). This was specific for Elf3 and the M2 antibody, because an increase in FLAG-Elf3 binding did not occur in the presence of a nonspecific IgG antibody or with mock-trans-fected extracts (data not shown). This phenomenon has also been described for PEA3 and its AID (38) and argues that Elf3 does not bind well to DNA in vitro and that the M2 antibody can stabilize its binding in vitro, possibly by artificially dimerizing or reinforcing intrinsic dimerization of Elf3 and/or creating an Where indicated, the cells were also transfected with 1 or 3 g of Elf3 and Elf3⌬246 -250 expression vectors. The promoter activity of each construct is calculated relative to the CAT activity observed with mT␤R-II Ϫ108/ ϩ56(B/A) alone, which was set to 1. C, F9-differentiated cells were transiently transfected with 5 g of the collagenase-1 Ϫ517/ϩ63 P/R. Where indicated, the cells were also transfected with 1 or 3 g of Elf3, Elf3⌬236 -260, Elf3⌬237-241, Elf3⌬237-241⌬246 -250, or Elf3⌬246 -250 expression vector. The promoter activity of each construct is calculated relative to the CAT activity observed with collagenase-1 Ϫ517/ϩ63 alone, which was set to 1. B and C, activities were assayed and normalized as described under "Experimental Procedures." These experiments were performed three times, and similar results were obtained. All of the expression vectors contain an exogenous SV40 NLS to ensure entry into the nucleus. FIGURE 8. The in vitro autoinhibitory activity of Elf3 is linked to its TAD by mutations within the ␣-helix of the TAD. A, diagram of FLAG-Elf3 deletion constructs. Elf3 represents the full-length protein (aa 1-371) with the FLAG tag N-terminal to the coding region. Residues 129 -159, 1-173, 1-233, or 1-270 were deleted from Elf3 to create expression constructs Elf3⌬129 -159, Elf3⌬N173, Elf3⌬N233, or Elf3⌬N270, respectively. B and C, equal amounts of the FLAG-Elf3 protein indicated above the lane were studied by EMSA with a radiolabeled ODN probe containing the wild-type ets-binding sites from ϩ3 to ϩ34 T␤R-II gene promoter. Mock lanes contain nuclear extract from 293T cells not transfected with an Elf3 expression vector. Supershift analysis (indicated above the lane with a ϩ sign) was performed by incubating the binding reaction for 40 min at 4°C with the anti-FLAG antibody M2 before adding the probe. The EMSA was analyzed as described under "Experimental Procedures." Components of this experiment were repeated three times, and similar results were obtained. The position of the free probe (F) is indicated. artificial release of autoinhibition, which may lead to increased stability of DNA binding.

Functional Domains of ELF3
To elucidate the location of the AID of Elf3, we used 293T cells to express several Elf3 mutant proteins with specific domains deleted. Western blot analysis was used to determine the relative amounts of Elf3 in the nuclear extracts. This enabled us to use equal amounts of Elf3 in our EMSA studies. As shown previously, removal of residues 1-233 or 1-270 from Elf3 enhanced its binding to DNA (Fig. 8B) (11). Furthermore, a smaller deletion of residues 1-173 also displayed enhanced binding compared with the ability of full-length Elf3 to bind to DNA (Elf3⌬N173; Fig. 8B). Interestingly, Elf3⌬N173, like Elf3⌬N270 but not Elf3⌬N233, forms two DNA-protein complexes (Fig. 8B), suggesting that Elf3⌬N173 and Elf3⌬N270 adopt a conformation that allows them to bind to both ets sites in the T␤R-II probe in vitro. The faster migrating complex in this EMSA is believed to contain only one Elf3 molecule (binary complex), whereas the slower migrating complex is believed to be a ternary complex of two molecules of Elf3 bound to the DNA probe. These findings also suggested that autoinhibition could be relieved by removal of the N terminus of Elf3, which contains the TAD and PNT domains.
Next, we examined which Elf3 domain within residues 1-173 led to increased DNA binding upon its removal. For this purpose, we examined the effect of removing the PNT domain (Elf3⌬60 -127) or the TAD (Elf3⌬129 -159). Deleting the PNT domain did not increase Elf3 binding to DNA in vitro (data not shown). However, removal of the TAD caused a prominent increase in binding to DNA when compared with wild-type Elf3 (Fig. 8B). This suggests that removal of the TAD disrupts autoinhibition and led us to hypothesize that the action of the AID is linked with promoter activation via the TAD. Previous studies have reported that the TAD of Elf3 has potent activity (11,28) and also contains a conserved ␣-helix from residues Ser 136 to Glu 144 (6,28). This NMR-predicted ␣-helix (6) is unusual for TADs, because they are typically unstructured regions. However, it is not completely unprecedented given that the ETS subfamily PEA3 is also reported to contain ␣-helical regions within their transactivation regions (66). To test whether the ability of Elf3 to activate transcription is necessary for the autoinhibition seen in vitro, we created point mutations within the TAD previously shown to affect function. As predicted, point mutations within the TAD, which are known to disrupt the ability of Elf3 to activate transcription (6,11,28), significantly enhanced Elf3 binding to DNA in vitro (Fig. 8C). Specifically, mutation of leucine 142 or tryptophan 137 to proline resulted in an increase in binding (Fig. 8C). In contrast, mutation of aspartic acid 133 to alanine did not significantly alter Elf3 binding to DNA (data not shown). Given that the D133A mutation affects transactivation, but not binding to DNA in vitro, it appears that at some level the AID and the TAD are distinct. Furthermore, only mutations predicted to alter the conformation of the ␣-helix (L142P and W137P) allowed binding to DNA in vitro, whereas mutations not expected to alter the ␣-helix (S136A, D133A; data not shown) did not increase binding in vitro. Finally, we also confirmed that Elf3, Elf3⌬129 -159, and Elf3L142P made by in vitro transcription/translation exhibited a similar increase in binding upon mutation of the TAD (data not shown). This argues that the autoinhibition is intrinsic to Elf3 and not an artifact of the nuclear extract preparation.
The N-terminal Region of Elf3 Interacts with Its C Terminus-The term autoinhibition suggests that another domain of Elf3 is involved in repressing its ability to bind to DNA. To address whether this is because of an intramolecular interaction within Elf3, we performed coimmunoprecipitation studies with the N-terminal portion of the Elf3 protein, FLAG-Elf3⌬C235, and the C-terminal portion containing the ETS domain and its 17-amino acid C-terminal tail fused to GFP, GFP-Elf3⌬N270 (Fig. 9A). We determined that the GFP-Elf3⌬N270 protein was pulled down when FLAG-Elf3⌬C235 was immunoprecipitated with the M2 antibody (Fig. 9B). This argues that the N terminus of Elf3 is able to interact with its C terminus.
These data, combined with our knowledge that the TAD is involved in autoinhibition, transactivation, and interaction with the coactivator Med23, led us to test the hypothesis that the interaction of the coactivator Med23 with the TAD of Elf3 in its N-terminal region could influence the ability of its N terminus to interact with its C terminus. For this purpose, 293T cells were cotransfected with an expression construct for fulllength Med23 or a fragment of Med23, which interacts with the A, diagram of the Elf3 expression constructs created for coimmunoprecipitation studies. B, 293T cells were transfected with 5 g of the indicated GFP-based expression vectors and 10 g of the FLAG or coactivator expression plasmids. After 24 h, extracts were made and exposed to M2-anti-FLAG affinity agarose overnight. The agarose was washed extensively, and FLAG proteins were then eluted with 150 ng/l 3ϫ FLAG peptide and analyzed for GFP and FLAG protein content. Input samples were taken before addition of the M2-anti-FLAG affinity agarose for analysis of protein expression and normalization of GFP-Elf3⌬N270 content present before immunoprecipitation (IP). FEBRUARY 2, 2007 • VOLUME 282 • NUMBER 5 TAD of Elf3 (6), along with expression vectors for FLAG-Elf3⌬C235 and GFP-Elf3⌬N270. Importantly, the presence of Med23 or Med23-(352-625) reduced the ability of GFP-Elf3⌬N270 to be coimmunoprecipitated with FLAG-Elf3⌬C235 when the M2 antibody was used (Fig. 9B). After normalization to the amount of FLAG-Elf3⌬C235 immunoprecipitated and the expression of GFP-Elf3⌬N270 (input), we determined that there was a 40 -60% decrease in interaction between GFP-Elf3⌬N270 and FLAG-Elf3⌬C235 in the presence of the coactivator as compared with the amount immunoprecipitated in the absence of the coactivator. This suggests that the coactivator interaction with the TAD can alter the interaction between the N terminus and C terminus. Below, we suggest that this may serve as a mechanism to relieve autoinhibition in vivo.

DISCUSSION
Previous studies demonstrated that Elf3 influences a wide range of biological functions (6, 7, 14, 19 -25). In the work described here, we examined how the five domains of Elf3 regulate its ability to bind to DNA and activate transcription. The SAR domain contributes to Elf3 transcriptional activity; however, the mechanism by which this happens is unclear. The PNT domain (Elf3⌬60 -127) and the last eight amino acids of the C-terminal tail (Elf3⌬C363), however, do not contribute to Elf3 activation of the promoter activity. Intriguingly, removal of the second AT-hook domain differentially affects the ability of Elf3 to stimulate the collagenase-1 and T␤R-II promoters. We also demonstrate that regions flanking the ETS domain are essential for maintaining transcriptional activity of Elf3. Removal of these regions disturbs the ability of Elf3 to bind to DNA, therefore affecting transcription. It is also evident that a domain distant from the ETS domain affects DNA binding. Specifically, our data argue that the TAD region is responsible, at least in part, for the AID of Elf3. The TAD is also needed for transcription activity arguing that the two functions may be related, as discussed below.
Promoter-specific Effects of the AT-hook Domain-Elf3 appears to be a unique protein in the ETS family, because it contains two AT-hook domains that are similar to the AT-hook domains of HMG-1 and HMG-I(Y), respectively (1). AT-hook domains are typically the DBDs of HMG proteins (30). The HMG proteins usually contain more than one AT-hook domain, which they utilize to bind in the minor groove of ATrich DNA and/or to interact with other proteins (31). However, few studies have directly addressed the role(s) of the AT-hook domains of Elf3 (19,21). The two AT-hook domains of Elf3 are located N-terminal to the ETS domain from residues 237-241 and 246 -250. Our studies demonstrate the second AT-hook domain of Elf3, which is conserved between mouse and human, plays an important role in activating transcription but in a promoter-dependent manner. Importantly, Elf3⌬246 -250, which removes the second AT-hook domain, is significantly less active with the collagenase-1 promoter than with the T␤R-II promoter.
We examined two likely explanations for the disparity in the function of this AT-hook domain between the collagenase-1 promoter and the T␤R-II promoter. The first possibility is that DNA-protein interactions between an AT-rich site directly downstream of the ets site and the AT-hook are necessary for Elf3 stimulation of the collagenase-1 promoter. However, this is unlikely because the AT-hook domain is also needed for maximal Elf3 activation of the T␤R-II (B/x) promoter, which contains only one ets site (data not shown), and an AT-rich region was not created when the second ets site was destroyed in the T␤R-II (B/A) P/R construct. Moreover, when we tested this possibility further by mutating the AT-rich base pairs downstream of the ets site of the collagenase-1 promoter, the ability of the modified promoter to respond to Elf-3 was not affected, nor did it alter the relative difference in the response to Elf-3 and Elf3⌬246 -250 (data not shown). This argues that the ATrich sequence is unlikely to be responsible for the promoterspecific effects of the AT-hook domain. Therefore, we examined a second possibility whereby AP-1 and Elf3 interact via the AT-hook domain to stimulate the collagenase-1 promoter. In this regard, the AP-1 site and the ets site of the collagenase-1 promoter are known to cooperate (17,18). Moreover, the HMG-I(Y) protein interacts with NF-B, SRF, and AP-1 through its AT-hook domain (31,(67)(68)(69). We tested the connection between the AT-hook domain of Elf3 and AP-1 site in the collagenase-1 promoter by mutating the AP-1 site and examining whether mutation of this site removed the requirement for the AT-hook domain. Although removal of the AP-1 site reduced the activity of the collagenase promoter, it did not eliminate the requirement for the AT-hook domain of Elf3 (data not shown). These results suggest that another, as yet unrecognized, mechanism is responsible for the requirement for the AT-hook domain for stimulation of the collagenase-1 promoter but not the T␤R-II promoter.
Requirement for Areas Flanking the ETS Domain-Our studies show that residues outside of the typical 80 -85 amino acids of the ETS domain are necessary for Elf3 to bind to DNA. Specifically, we show that removal of residues 260 -270 or 355-362 disrupts binding to DNA. Removal of these residues may influence the overall structure of the ETS domain, thereby disrupting protein-DNA interactions within the ETS domain. Alternatively, the regions flanking the ETS domain may themselves make direct contacts with DNA. We favor the former possibility, because other ETS proteins, namely of the Ets1 and PEA3 subfamilies, contain regions flanking their ETS domains that affect its structure (33,40,41). In these cases, the regions flanking the ETS domain are believed to stabilize a conformation that has a lower affinity for DNA (40,41). However, in the case of Elf3, the regions that directly flank the ETS domain may be necessary to maintain the overall structure of the ETS domain and allow Elf3 to bind to DNA once autoinhibition is relieved or disrupted (see below). Future structural studies will be necessary to determine how these regions contribute to the structure of the ETS domain of Elf3 and its binding to DNA.
Model of Elf3 Autoinhibition-Previous studies suggested that Elf3 contains an important AID that affects its binding to DNA in vitro (11). Autoinhibition of transcription factors is often demonstrated by the inability of the protein to bind to DNA in vitro (39,41). In vivo, autoinhibition is believed to prevent promiscuous DNA binding and adds another layer of control over gene regulation (39). Many proteins are reported to contain an AID. Germane to this discussion, many transcription factors contain AIDs that affect their ability to bind DNA (39), including nine ETS transcription factor family members (11,(33)(34)(35)(36)(37)(38). Thus far, two models have been proposed to explain autoinhibition of ETS proteins. The first, the Ets1based model, involves regions directly flanking the ETS domain inhibiting DNA binding and direct interaction with other activators within these regions or the ETS domain itself to release autoinhibition (39). The second, the Elk1-based model, involves distant intramolecular interactions to create autoinhibition and release occurs by post-translational modification (50).
Based on three key findings described in this study, we propose a novel model for autoinhibition that differs from the models proposed for Ets1 and Elk1. First, we show that mutation within the ␣-helix of the TAD increases Elf3 binding to DNA, which argues that its TAD is responsible, at least in part, for Elf3 autoinhibition. Second, we show that the N terminus of Elf3 interacts with its C terminus. Third, we demonstrate that the coactivator Med23, which interacts with the TAD of Elf3 (6), can reduce the interaction between the N and C terminus of Elf3. On the basis of these findings, we propose a model for Elf3 autoinhibition (Fig. 10). Specifically, we propose that Elf3 is held in a conformation that is unable to interact with DNA, because of an interaction between the N terminus and the C terminus, possibly directly through the ETS domain and the TAD (Fig. 10A, left). This would be a reversible state in which the autoinhibited state would predominate, but a limited amount could bind to DNA (Fig 10B, left). In the presence of a coactivator, such as Med23, which interacts with the TAD ␣-helix, the interaction between the N and C terminus of Elf3 would be weakened or altered shifting the conformation of Elf3 to the DNA binding form. Therefore, in the presence of the coactivator the DNA binding form would predominate ( Fig  10A, right). As a result, the interaction of the coactivator with Elf3 would release autoinhibition and simultaneously promote DNA binding and transcription.
There are two possible spatial locations in which coactivator interaction and release of Elf3 autoinhibition may occur. 1) The coactivator may interact with Elf3 in the absence of DNA, which stabilizes the DNA binding form, and this coactivator-Elf3 complex then binds to gene regulatory regions to affect gene transcription (Fig 10B, middle). Alternatively, if the coactivator is present at the promoter prior to Elf3, it could stabilize the transient binding of Elf3 to gene regulatory regions (Fig 10B, right).
This model is similar, in part, to the autoinhibition model described for the ETS transcription factor Elk1. Elk1 is reported to contain three AIDs as follows: the TAD, the R-domain, and the B-box (50,51). The main AID of Elk1, which affects DNA binding, is located within the TAD (50). The TAD of Elk1 is reported to bind directly to the ETS domain, limiting interaction with DNA until the TAD is phosphorylated (50). In this case, phosphorylation triggers the release of autoinhibition in Elk1, whereas we propose the triggering mechanism(s) for release of autoinhibition in Elf3 is an interaction with a coactivator. Importantly, several reports suggest that protein-protein interactions with cofactors or coactivators can result in a release of autoinhibition in vivo (40, 43-45, 59, 70). For example, TBP is reported to increase DNA binding of p53 independent of the interaction of TBP with DNA (70). Furthermore, several transcription factors display increased affinity for DNA in vitro upon the removal of their TADs, including Sox11, IRF-3, and Ets2 (12,71,72). In combination, these studies suggest that other transcription factors may contain an AID that coordinates transcriptional activation and binding to DNA. Furthermore, these studies suggest the interaction of coactivators/cofactors, and the TADs of transcription factors may be important for increasing transcription factor binding in vivo. Future studies will be needed to determine whether this is a common mechanism utilized to enhance the binding of many transcription factors in vivo.
In summary, our work provides a comprehensive examination of the domains utilized by Elf3 to regulate DNA binding and gene expression. Furthermore, we propose a novel model by which transcription and DNA binding may be intimately connected. Future investigation into the autoinhibition may find that multiple mechanisms are used to relieve autoinhibition of individual ETS factors. Moreover, our model for the release of autoinhibition may be applicable to a wide range of transcription factors. However, further work will be needed to test this possibility. Finally, it will be important to identify the cellular cues that regulate the transcriptional activation and/or autoinhibition of Elf3. In the future, this information could prove useful for manipulating Elf3 in disease states, such as cancer. A, Elf3 protein can exist in an autoinhibited form that interacts weakly with DNA. Upon interaction with a coactivator (Co-A), Elf3 is converted to a DNA binding form, and its affinity for ets sites increases significantly. This model depicts a direct interaction between the TAD and the ETS domain of Elf3. However, it remains to be determined whether the TAD and the ETS domain interact directly. B, left, in the absence of the coactivator, Elf3 cannot stably bind to DNA, and transcription is not activated. Middle, Elf3 and the coactivator interact in the absence of DNA. This complex exhibits a higher affinity for DNA than Elf3 in an autoinhibited form. Right, Elf3 binds unstably to the promoter region unless the coactivator (CoA) is present to stabilize Elf3 in the DNA binding form.