Repression of Transcriptional Enhancer Factor-1 and Activator Protein-1-dependent Enhancer Activity by Vascular Actin Single-stranded DNA Binding Factor 2*

Transcriptional repression of the murine vascular smooth muscle (cid:97) -actin gene in fibroblasts results from the interaction of two sequence-specific single-stranded DNA binding activities (VACssBF1 and VACssBF2) with opposite strands of an essential transcriptional en- hancer factor-1 (TEF-1) element (Sun, S., Stoflet, E. S., Cogan, J. G., Strauch, A. R., and Getz, M. J. (1995) Mol. Cell. Biol. 15, 2429–2436). Here, we identify a sequence element located within a protein-coding exon of the gene that bears structural similarity with the TEF-1 enhancer. This includes a 30-base pair region of purine- pyrimidine asymmetry encompassing a perfect 6-base pair GGAATG TEF-1 recognition motif. Unlike the en- hancer, however, the exon sequence exhibits no TEF-1 binding activity nor does the pyrimidine-rich strand bind VACssBF1. However, VACssBF2 interacts equally well with the purine-rich strand of both the enhancer and the exon sequence. To test the ability of VACssBF2 to independently repress transcription, the exon se- quence was placed upstream of a deletionally activated promoter containing an intact TEF-1 binding site. The exon sequence repressed promoter activity, whereas a mutant deficient in VACssBF2 binding did not. More-

The cell-and tissue-specific transcription of eukaryotic genes is regulated by cis-acting elements that are generally localized outside of the RNA/protein coding region of the gene. These regulatory elements, known commonly as enhancers or silencers, are DNA sequences that influence gene transcription in either a positive or negative fashion by virtue of their ability to bind specific proteins that, in turn, regulate the assembly of a basal transcriptional complex. Although these DNA elements are commonly found upstream of the transcription start site, cis-acting silencing elements have also been localized within introns and the protein-coding sequences of a number of genes (1)(2)(3). In most cases, the fidelity of enhancer or silencer function is strongly correlated with the sequence-specific binding of a protein(s) to a region of double-stranded DNA (dsDNA) 1 encompassing the enhancer or silencer motif (4,5). However, recent studies suggest that proteins that bind to singlestranded DNA (ssDNA) in a sequence-specific manner may also participate in the regulation of gene transcription. Although the molecular mechanism(s) by which ssDNA-binding proteins affect transcription remains largely speculative, these proteins are often associated with transcriptional silencing elements (6 -11).
Recent studies have shown that repression of vascular smooth muscle (VSM) ␣-actin promoter activity in both BC3H1 myoblasts and AKR-2B fibroblasts can be linked to the interaction of two sequence-specific ssDNA-binding activities within a 30-bp polypurine-polypyrimidine tract residing in the 5Ј promoter (12,13). These activities, designated VSM ␣-actin singlestrand binding factor 1 and 2 (VACssBF1 and VACssBF2) bind to opposite strands of the DNA helix within a region of the promoter (Ϫ194 to Ϫ165) spanning the binding site for a dsDNA-binding protein closely related with, and possibly identical to, transcriptional enhancer factor 1 (TEF-1) (13), an SV40 enhancer, and muscle-specific M-CAT motif-binding transcription factor (14 -20). Promoter activation, site-directed mutagenesis, and DNA-binding studies have led to the hypothesis that VACssBFs function by stabilizing a local single-stranded DNA conformation within the 30-bp promoter element (PE) which precludes TEF-1 binding and hence transcriptional activation (12). Interestingly, amino acids 44 -53 of VSM ␣-actin are encoded by a DNA sequence that is structurally similar to the 5Ј PE. This intragenic sequence element or coding element (CE) lies immediately adjacent to the 3Ј end of intron 2 and is positionally conserved in all vertebrate actin genes (21). The coding element (CE) and promoter element (PE) each possess a core GGAATG consensus TEF-1 binding motif centered within an asymmetric polypurine-polypyrimidine tract.
The present study illustrates that VACssBF2, but not VACssBF1, interacts with the CE in a sequence-and single strand-specific manner. The identification of this promoterindependent VACssBF2 binding site permitted a more rigorous * This work was supported by the NHLBI Grant HL54281 and by the Mayo Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶  1 The abbreviations used are: dsDNA, double-stranded DNA; ssDNA, single-stranded DNA; VSM, vascular smooth muscle; VACssBF, VSM ␣-actin single-strand binding factor; TEF-1, transcriptional enhancer factor-1; PE, promoter element; CE, coding element; CAT, chloramphenicol acetyltransferase; EMSA, electrophoretic mobility shift assay; mu, mutant; wt, wild type; AP-1, activator protein-1; PVDF,polyvinylidene difluoride; bp, base pair. evaluation of the ssDNA-binding properties, subunit composition, and functional activity of VACssBF2. VACssBF2 is shown to consist of multiple distinct polypeptide species of differing apparent molecular weights and ssDNA binding specificity. Functional analysis of the VACssBF2-CE interaction conducted within the context of homologous and heterologous promoters suggests that VACssBF2 can repress enhancer-dependent transcription independently of both VACssBF1 and the dsDNA-binding protein governing enhancer activity.

EXPERIMENTAL PROCEDURES
Cell Culture, Transient Transfection, and Reporter Gene Assay-Mouse embryo-derived AKR-2B fibroblasts were maintained in culture and transfected as described previously (22,23). Transfected cells were allowed to recover in McCoys 5A medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT) for 18 -24 h. The cells were then washed twice and rendered quiescent by incubating an additional 48 h in serum-free MCDB402 medium (JRH Biosciences, Lenexa, KS). Quiescent cells were stimulated for 6 h with MCDB402 medium and 20% fetal bovine serum or for 4 h with the same medium supplemented with 10 g/ml cycloheximide (Sigma). Serumand cycloheximide-stimulated cells were washed twice and subjected to a 2-h washout in medium lacking cycloheximide to allow for the recovery of CAT protein synthesis (24). Cells were harvested, and cellular extracts were prepared by repeated freeze-thawing. Protein concentration was determined by dye-binding assay (Bio-Rad) using bovine serum albumin as a standard. Chloramphenicol acetyltransferase (CAT) reporter protein was measured in cell lysates using an immunoassay kit (Boehringer Mannheim).
Construction of VSM ␣-Actin Promoter/Coding Element Reporter Plasmids-The deletionally activated VSM ␣-actin promoter construct pC3VSMP4 (VSMP4) (25) was modified by insertion of a synthetic copy of a 33-base pair segment of DNA encoding the intron 2/exon 3 splice site and amino acids 44 -53 of VSM ␣-actin (Table I) both 3Ј and 5Ј of the transcription start site. Complementary oligonucleotides possessing the appropriate sticky ends as well as an internal BclI restriction site (underlined) of the form 5Ј-AGCTTTGATCAXnG-3Ј and 5Ј-TCGAC-XRnTGATCAA-3Ј or 5Ј-GATCCTGATCAXn-3Ј and 5Ј-TCGATGATCA-XRnG-3Ј where Xn is 5Ј-CAGGGAGTAATGGTTGGAATGGGCCAAA-AAGAC-3Ј(wt), XRn is 5Ј-GTCTTTTTGGCCCATTCCAACCATTACTC-CCTG-3Ј (wt) or Xn is 5Ј-CAttGAGTAcTGGTTttccgtGGCCAAccAGAC-3Ј(mu2), XRn is 5Ј-GTCTggTTGGCCacggaaAACCAgTACTCaaTG-3Ј(mu2) were synthesized on an Applied Biosystems model 394 DNA/ RNA synthesizer (Foster City, CA) and gel-filtered over a NAP-25 column (Pharmacia Biotech, Inc.) in distilled water. Oligonucleotides were phosphorylated, annealed, and ligated into either HindIII and SalI (5Ј construct) or BamHI and XhoI (3Ј construct) cut, and phosphatase-treated VSMP4 to create P4/CE(5Ј) and P4/CE(3Ј) constructs, respectively. The CE(3Ј) plasmids were constructed such that the VSM ␣-actin coding sequence element and the CAT coding sequence are in reading frame. The CE was similarly cloned into a heterologous tissue factor promoter construct (AP1TF60CAT) upstream of two AP-1 elements (23) using synthetic oligonucleotides possessing 5Ј HindIII and 3Ј XbaI sticky ends, to create TF/CE(5Ј). Plasmid constructs were transformed into Escherichia coli HB101 cells using standard CaCl 2 techniques. All plasmid DNA used in transfection experiments was purified by double cesium chloride gradient centrifugation. The fidelity of all cloned sequences described in this paper was confirmed by automated dideoxy DNA sequencing.
Primer Extension-Transcriptional activity of promoter constructs was assessed by analysis of correctly initiated CAT transcripts via primer extension using a CAT coding sequence-specific primer (24). AKR-2B cells were cotransfected (in quadruplicate) as described above with 10 g of VSM ␣-actin promoter plasmid along with 5 g of p99␤Ac-CAT as an internal control (22). Following 6 h stimulation with 20% fetal bovine serum and 10 g/ml cycloheximide, total RNA was harvested using Trizol reagent (Life Technologies, Inc.). Poly(A) ϩ mRNA was subsequently isolated using a PolyATtract mRNA isolation system III kit (Promega, Madison, WI). Lyophilized poly(A) ϩ mRNA (1.0 g) was resuspended in 12 l of nuclease-free water containing 3 ng (ϳ250,000 cpm) of 32 P-end-labeled primer complementary to nucleotides 29 -61 of the CAT coding sequence (24). The primer was annealed to its complementary mRNA by sequential 10-min incubations at 70, 65, 60, 55, and 4°C. The reaction mixtures were then supplemented with cDNA synthesis buffer components (7 l) to a final concentration of 50 mM Tris-HCl (pH 8.3), 75 mM KCl, 3 mM MgCl 2 , 10 mM dithio-threitol, 500 M dNTPs. The primer was then extended to the transcriptional initiation site by addition of 200 units of reverse transcriptase (1 l) (Superscript RT, Life Technologies, Inc.) and incubation for 1 h at 42°C. Extension products were resolved by denaturing electrophoresis on a 0.75-mm thick 10% acrylamide (29:1, acrylamide:bisacrylamide), 7 M urea gel in TBE (50 mM Tris, 50 mM borate, 1 mM EDTA (pH 8.3)). Following electrophoresis, the gels were wrapped in plastic and exposed to x-ray film (XAR-5, Eastman Kodak) at Ϫ70°C in a Kodak X-Omat film cassette with intensifying screens for 18 -36 h.
Oligonucleotide Gel Shift Probes-Wild-type and mutant oligonucleotides (30 base) corresponding to either the promoter element (PE, nucleotide Ϫ194 to Ϫ165) or coding element (CE, G/nucleotide ϩ201 to ϩ229) (Fig. 1, Table I) were synthesized and purified as described above.
Preparation of DNA-binding Protein Extracts-Whole cell protein extract of rapidly growing AKR-2B fibroblasts was prepared by hypertonic lysis as described previously (13). Cytosolic and nuclear protein extracts were prepared from rapidly growing AKR-2B and NIH-3T3 fibroblasts according to the mini-extraction method of deGroot and co-workers (26). Protein concentration was determined by dye-binding assay (Bio-Rad) using bovine serum albumin as a standard.
Molecular Weight Determination of Nucleic Acid-binding Proteins-Apparent molecular weights of proteins bound to [␥-32 P]ATP end-labeled ssDNA oligonucleotide in EMSAs were determined by UV crosslinking and subsequent SDS-polyacrylamide gel electrophoresis as described previously (13).
Southwestern Blotting-An equivalent amount of protein from whole cell, cytosolic, and nuclear extracts was resolved by denaturing electrophoresis on 10% SDS-polyacrylamide gels according to the method of Laemmli (27) using a Bio-Rad Mini-PROTEAN II apparatus at 150 V for 90 min. Separated proteins were then electrotransferred to rehydrated PVDF membranes (Millipore, Bedford, MA) at 4°C for 90 min (300 mA) in transfer buffer containing 25 mM Tris, 192 mM glycine (pH 8.3), and 20% methanol. After transfer, the membranes were blocked with 5% Carnation nonfat dry milk in TNE buffer (25 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1 mM EDTA) overnight at 4°C. All subsequent steps were performed at room temperature. Prior to addition of probe, the PVDF membranes were preequilibrated in binding buffer (1% nonfat dry milk-TNE) for 30 min. 32 P-End-labeled ssDNA oligonucleotide probe, 2 ϫ 10 6 cpm in 5 ml of binding buffer containing 10 g/ml

VACssBF2-induced Suppression of Enhancer Function
poly(dI-dC), was then added, and the membranes were incubated for 3 h with continuous agitation. The membranes were washed three times, 10 min each with TNE containing 0.05% Tween 20 and then one final time with TNE alone. After air drying, the membranes were wrapped in plastic, placed in a x-ray cassette, and autoradiographed at Ϫ70°C for 1-3 days as described above. Apparent molecular weights were calculated from a plot of log molecular weight versus relative mobility of pre-stained molecular weight standards (Bio-Rad) using a linear least squares regression fit.

Promoter Element and Coding Sequence Element
Homology-In previous studies transcriptional regulation of the murine VSM ␣-actin gene in fibroblasts and myoblasts was shown to be mediated, in part, by both positive and negative elements residing in a 30-bp asymmetric polypurine-polypyrimidine tract within the 5Ј promoter (12,13,22,25). Promoter activation and DNA-binding experiments suggested a model in which repression of this TEF-1-dependent enhancer element is maintained by disruption of base pairing via the interaction of two sequence-specific ssDNA binding proteins, VACssBF1 and VACssBF2, that bind opposite stands of the DNA helix encompassing the TEF-1 recognition sequence (12,13). Interestingly, a search of the cDNA sequence (28) revealed that a protein coding exon of the gene contains another, perfect, consensus TEF-1 recognition motif (GGAATG) centered within a purinerich tract. As shown in Fig. 1, this protein-coding element or CE which encodes the intron 2/exon 3 splice site and amino acids 44 -53 (29) also exhibits additional nucleotide sequence similarity (Fig. 1, bold letters) to the negatively regulated TEF-1 enhancer, termed the promoter element (PE). It is noteworthy that the intragenic positioning of the CE at the intron 2/exon 3 boundary is conserved in all isoforms of mammalian actin genes characterized thus far (21). In addition, the amino acid sequence encoded by the CE is identical among the different actin isoforms, although the nucleotide sequence differs due to multiple third position codon changes (Table II  ( 28, 30 -34)).
Interaction of a Subset of Single-stranded PE-binding Proteins with the CE-Direct and competitive electrophoretic mobility shift assays (EMSAs) with wild-type and mutant 30 base oligonucleotides (Table I) and AKR-2B cellular extract were performed to delineate the protein-binding properties of the CE in comparison to the PE. The double-stranded form of the PE has been previously shown to interact with a protein that is related with, and likely identical to, TEF-1 (13). Despite the presence of a GGAATG motif in both elements, only the doublestranded form of the PE exhibited detectable affinity for TEF-1 (data not shown). Similar results were obtained on examination of VACssBF1 binding affinity. As shown in Fig. 2, the pyrimidine-rich, noncoding strand of the CE (CE-MCATss) did not compete for binding of VACssBF1 to 32 P-PE-MCATss (lanes 6 -10), whereas competition was readily observed by the homologous PE-MCATss sequence (lanes [1][2][3][4][5]. Direct binding EMSAs also indicated that VACssBF1 had little or no detectable affinity for 32 P-CE-MCATss (data not shown).
In marked contrast to both TEF-1 and VACssBF1, the previously described, electrophoretically distinct band-shifted components of VACssBF2 (12) were clearly resolved when either the PE (PE-PrMss) or CE (CE-PrMss) sense strands were used as probes (Fig. 3). The ability of excess, unlabeled CE-PrMss and PE-PrMss to independently compete for binding of VACssBF2 to 32 P-CE-PrMss (Fig. 4, lanes 2-6 and 14 -18) confirmed that these ssDNA probes likely interact with the identical protein(s). Importantly, as illustrated by the competitive properties of CE-PrMmu2 (Fig. 4, lanes 8 -12), mutation of nucleotides that are positionally conserved in both the PE and CE (Fig. 1, bold letters) significantly reduces the affinity of the CE for the more rapidly migrating VACssBF2 doublet. The binding of the more slowly migrating VACssBF2 component was not affected by these mutations implying that this species has a weaker affinity for ssDNA or, alternatively, a different sequence specificity (see "Discussion") than that of the VACssBF2 doublet. These data are consistent with previous mutagenesis experiments conducted with the homologous PE sense strand which indicated that binding activity of the VACssBF2 doublet is impaired by select purine to pyrimidine transversions introduced either within the GGAATG motif or toward the 5Ј and 3Ј ends of the ssDNA (12).
Subunit Composition and Molecular Weight Estimation of Polypeptide Components of VACssBF2-Binding site mutagenesis studies presented above (Fig. 4) as well as those reported previously (12) indicate that the effect of tested mutations within either the PE or CE is exclusively restricted to the more rapidly migrating VACssBF2 doublet, implying that VACssBF2 is not a homogeneous ssDNA-binding entity. Therefore, an analysis of the individual VACssBF2 band shift species was undertaken. Single-stranded DNA-protein complexes were first resolved in nondenaturing EMSA gels following incubation of AKR-2B whole cell protein extract with a coding strandspecific oligonucleotide probe ( 32 P-PE-PrM). The protein-DNA complexes were irreversibly cross-linked by exposure of the wet gels to 254 nm UV light and then excised and subjected to denaturing electrophoresis in SDS-polyacrylamide gels. Fig. 5 illustrates that VACssBF2 consists of multiple distinct polypeptide species. Relative apparent molecular weights of the individual species were calculated after correcting for the contribution of the DNA component (35). The two bands of the rapidly migrating doublet were independently excised following UV cross-linking and analyzed separately. As shown in Fig.   FIG. 1. DNA sequence elements within the promoter and coding 2 and 3), the individually cross-linked components of the VACssBF2 doublet migrated somewhat differently with corrected apparent molecular weights of 52,000 and 49,000, respectively. The more slowly migrating component (Fig. 5, lane 1) resolved as two components, a major M r ϳ 84,000 species and a faint M r ϳ 30,000 species. Given the relative difference in intensity of the two bands, the M r ϳ 30,000 species is likely only a minor component of the more slowly migrating VACssBF2 complex.

(lanes
Southwestern blotting experiments were also performed with 30 base CE ssDNA probes to validate the conclusion that VACssBF2 consists of multiple distinct polypeptide species. As shown in Fig. 6, the profile of the Southwestern blot probed with the wild-type CE oligonucleotide (CE-PrMss) is similar in appearance to previously described EMSAs. In particular, the CE ssDNA-binding reactivity of AKR-2B-derived cellular protein resolved into at least two distinct components, a more rapidly migrating doublet (M r ϳ 51,000/48,000) and a more slowly migrating species (M r ϳ 121,000). The specificity of these ssDNA-protein complexes was indicated by the lack of signal generated in a parallel blot probed with a mutant CE oligonucleotide (Fig. 6, CE-PrMmu2). A number of other bands (M r ϳ133,000, 39,000, 33,000/32,000 doublet, 25,000) were also discernible in the CE-PrMss blot (Fig. 6, left panel), most notably in the lane containing AKR-2B nuclear extract (lane 3). Since these bands were absent or diminished in the parallel CE-PrMmu2 blot, they may represent different post-translationally modified forms and/or degradation products of VACssBF2 that are not detected by band shift analysis. Nuclear extracts from cultured NIH-3T3 fibroblasts also exhibited VACssBF2 ssDNA binding activity (Fig. 6, lane 4). It is also noteworthy that the overall intensity of signal relative to whole cell extracts (Fig. 6, lane 1) was enhanced in nuclear extracts (lane 3) and diminished in cytosolic extracts (lane 2) of AKR-2B fibroblasts. Because an equivalent amount of protein from each extract was applied to the gel, these data suggest that VACssBF2 is enriched in the nucleus. Results of comparative band shift assays conducted with whole cell, nuclear, and cytosolic extracts of AKR-2B fibroblasts were consistent with this conclusion (data not shown).
Repression of Enhancer-dependent Transcription by VAC-ssBF2-Previous deletion-mapping studies demonstrated that a mouse VSM ␣-actin promoter truncated to position Ϫ224 (VSMP3, Fig. 7) is transcriptionally silent in both AKR-2B fibroblasts and undifferentiated BC3H1 myoblasts (22,25). Deletion to position Ϫ195 (the 5Ј end of the PE) yielded a promoter that was still fully repressed in AKR-2B fibroblasts but only partially repressed in subconfluent myoblasts (25). Further truncation of the PE to position Ϫ191 produced a serum-inducible promoter (VSMP4, Fig. 7) that was transcriptionally active in both cell types (22,25). Subsequent studies revealed that a purine-rich, TEF-1-binding element located between Ϫ181 and Ϫ176 is essential for the activity of VSMP4 (13) and that ssDNA binding activities (VACssBF1 and

VACssBF2-induced Suppression of Enhancer Function
VACssBF2) that interact with opposite strands of the TEF-1 enhancer (i.e. the PE) are, at least, partially responsible for the repression of VSMP3 in both fibroblasts and myoblasts (12,13).
The discovery that the CE element binds VACssBF2 but not VACssBF1 nor TEF-1 allowed us to examine the functional activity of VACssBF2 independently by assessing the effect of the CE on promoter function. A synthetic copy of the CE was inserted 5Ј of the TEF-1 enhancer binding site in VSMP4 (13) to create P4/CE(5Ј) (Fig. 7). This chimeric construct was tested in parallel transfections with VSMP4 and VSMP3 for CAT protein expression in quiescent and serum-stimulated AKR-2B fibroblasts. As shown in Fig. 7, the activity of P4/CE(5Ј) was repressed approximately 5-10-fold relative to its parent construct, VSMP4. This level of repression is similar, although not identical, to that of VSMP3, which was repressed 9 -16-fold relative to VSMP4 in quiescent and serum-stimulated transfectants. These data suggest that the CE can partially substitute for the intact PE in suppressing the activity of the promoter when positioned 5Ј of the TEF-1 binding site.
Because promoter activity was also suppressed in P4/CE(5Ј) transfectants superinduced with serum and cycloheximide (Fig. 7), we next tested whether or not the effect of the CE was mediated at the transcriptional level. The above constructs were transiently transfected into AKR-2B fibroblasts together with a ␤-actin promoter construct (p99␤Ac-CAT, (22)) as a transfection efficiency control. Following superinduction with serum and cycloheximide, poly(A) ϩ mRNA was isolated. Promoter activity was evaluated by assaying the levels of correctly initiated CAT RNA transcripts by primer extension analyses using a CAT coding sequence-specific primer. As shown in Fig.  8, while positive control transcripts from the ␤-actin promoter were equally expressed in all cotransfectants, the levels of VSM ␣-actin promoter-driven transcripts varied. This pattern was also evident using shorter autoradiographic exposure times. In particular, VSMP3 was transcriptionally silent while VSMP4 was transcriptionally active (Fig. 8, compare lanes 2 and 3). Incorporation of the CE into the 5Ј position of VSMP4 significantly impaired the ability of the promoter to drive transcription of the CAT reporter gene (Fig. 8, compare lanes 3 and 4). However, the CE had no substantive effect when positioned 3Ј of the transcription start site at a distance from the TEF-1 binding site (Fig. 8, lane 6) suggesting that the repressive function of the CE is position-dependent. Importantly, a promoter construct possessing a 5Ј mutant CE defective in VACssBF2 binding, P4/CEmu2(5Ј), exhibited transcriptional activity comparable with that of VSMP4 (Fig. 8, compare lanes  3 and 5, and Fig. 9). These data suggest that VACssBF2 can function in the absence of VACssBF1 to transcriptionally re-  1-3) and NIH-3T3 fibroblasts (lane 4) were resolved on 10% SDS-polyacrylamide gels and electrotransferred to PVDF membranes. Blots were probed with either wild-type (CE-PrMss) or mutant (CE-PrMmu2) 32 P-labeled ssDNA oligonucleotides corresponding to the VSM ␣-actin CE coding strand. Bands demonstrating an electrophoretic mobility and nucleic acid-binding specificity consistent with that of VACssBF2 are indicated by solid arrows.

FIG. 4. Effect of wild-type and mutant oligonucleotides on
VACssBF2 binding to CE coding strand DNA. 32 P-Labeled ss oligonucleotide corresponding to the coding strand of the CE (CE-PrMss) was incubated with AKR-2B whole cell extract protein (3.6 g) in the presence of the indicated molar excess of either homologous oligonucleotide (CE-PrMss, lanes 2-6), mutant CE coding strand oligonucleotide (CE-PrMmu2, lanes 8 -12), or the PE wild-type coding strand oligonucleotide (PE-PrMss, lanes 14 -18). Protein-DNA complexes were resolved by EMSA. FP, free probe.

VACssBF2-induced Suppression of Enhancer Function
press a nearby TEF-1 enhancer in fibroblasts.
To test whether VACssBF2-induced repression was unique to the VSM ␣-actin TEF-1-dependent enhancer element, the CE was also incorporated into a heterologous tissue factor promoter construct immediately upstream of two activator protein-1 (AP-1) dsDNA-binding elements (Fig. 9). This construct was chosen because these two AP-1 elements had been previously shown to be necessary and sufficient for serum inducibility of the basal tissue factor promoter in AKR-2B fibroblasts (23). The heterologous tissue factor/VSM ␣-actin CE construct, TF/CE(5Ј), was tested in parallel transfections with its parent construct, TFAP1, as well as VSMP3, VSMP4, P4/CE(5Ј), and P4/CEmu2(5Ј) for CAT protein expression in serum and cycloheximide-stimulated fibroblasts. As shown in Fig. 9, the promoter activity of both TF/CE(5Ј) and P4/CE(5Ј) was only 28 -35% of that exhibited by the parent constructs, TFAP1 and VSMP4. For comparison, the transcriptionally silent control promoter, VSMP3, was found to have about 8% activity, whereas P4/CEmu2(5Ј) displayed 84% of the activity of VSMP4 (Fig. 9). Primer extension analyses of p99␤Ac-CAT cotransfectants confirmed that the differences observed between TFAP1 and TF/CE(5Ј) were transcription-dependent and could not be accounted for by differences in transfection efficiency (data not shown). Thus, the proportionate reduction in promoter activity displayed by both TF/CE(5Ј) and P4/CE(5Ј) suggests that VACssBF2-induced repression of enhancer function is independent of the enhancer motif involved. DISCUSSION A recent study suggested a model in which two ssDNAbinding factors, designated VACssBF1 and VACssBF2, function to repress transcription of the mouse VSM ␣-actin gene by disrupting base pairing within an essential TEF-1 enhancer element (12). Although these activities bind to opposite strands of the DNA helix encompassing the TEF-1 recognition motif, we could not determine whether both or only one of these factors were required to repress transcription. In this study we provide evidence that VACssBF2 can function autonomously (i.e. in the absence of an opposing VACssBF1 binding site) to transcriptionally repress the activity of two distinct promoters in an enhancer-independent manner. These results were based upon the initial identification of a DNA element within the protein-coding region of the gene that bears intriguing sequence similarity to the negatively regulated enhancer element FIG. 7. Effect of the CE on VSM ␣-actin promoter activity in quiescent and serum-stimulated AKR-2B fibroblasts. Upper panel, a synthetic copy of the CE was cloned into a transcriptionally active VSM ␣-actin promoter-CAT reporter construct, VSMP4, 5Ј of the TEF-1 binding site to create P4/CE(5Ј). VSMP3, which possesses an intact PE, is a transcriptionally repressed promoter construct (25). VSMP4, a deletion mutant of VSMP3 possessing a truncated PE, is a transcriptionally active promoter construct (25). Lower panel, VSM ␣-actin promoter-CAT reporter constructs were tested in parallel transfections for CAT protein expression in quiescent, serum, and serum and cycloheximide-stimulated AKR-2B fibroblasts as described under "Experimental Procedures." Cellular extracts from duplicate transfectants were assayed for CAT protein by immunoassay.
FIG. 8. Transcriptional activity of chimeric VSM ␣-actin promoter/CE constructs. VSM ␣-actin promoter-CAT constructs were tested in parallel transfections for CAT mRNA expression in AKR-2B fibroblasts. Transfectional efficiency was controlled by cotransfecting a ␤-actin promoter-CAT construct (p99␤Ac-CAT). Correctly initiated CAT mRNA transcripts were assayed by primer extension analyses of poly(A) ϩ RNA pooled from four independent transfections as described under "Experimental Procedures." Arrows indicate the position of the predicted extension product from p99␤Ac-CAT (␤Ac, 129 nucleotides), from VSMP3, VSMP4, P4/CEwt(5Ј), and P4/CEmu2(5Ј) (␣Ac, 168 nucleotides), and from P4/CEwt(3Ј) (␣Ac/CE, 207 nucleotides). 32 P-Labeled markers (M, MspI digest of pBR322 DNA) are shown for reference. Primer extension of RNA harvested from mock-transfected cells (mock) is shown as a control for the fidelity of the CAT mRNA primer. located in the proximal promoter (Fig. 1). This sequence element, termed the coding element (CE), contains a 6-bp GGAATG TEF-1 recognition motif positioned within a purinerich sequence and is thus structurally analogous to the promoter element (PE). Comparative DNA binding studies revealed that while TEF-1 and VACssBF1 binding are restricted to the PE (Fig. 2), VACssBF2 interacts with the purine-rich coding strand of the CE (Fig. 3) in a sequence-dependent manner (Fig. 4). Because the noncoding strands of both the PE and CE exhibit a high pyrimidine content (83 and 77%, respectively), the determinants of VACssBF1 binding must also include specifically ordered sets of bases rather than a simple high ratio of pyrimidines. Our results also demonstrate that TEF-1 binding is strongly influenced by sequences that flank the central GGAATG motif.
Several recent reports identifying a silencing element within the protein-coding sequence of the osteocalcin gene (1,2) prompted us to investigate the possibility that the CE may function as an intragenic, transcriptional silencer. Although the CE mimicked the PE when placed 5Ј of a TEF-1 binding site in a deletionally activated promoter (P4/CE(5Ј), Fig. 7), it did not function as a classic silencing element. Analyses of CAT reporter gene mRNA were consistent with the interpretation that the CE suppressed TEF-1-dependent transcriptional activity when adjacent to and upstream of the TEF-1 binding site but not when positioned 3Ј to the transcription start site at a distance from the enhancer element (Fig. 8). Moreover, mutagenesis experiments established a correlative relationship between the ability of VACssBF2 to interact with ssDNA and the maintenance of transcriptional repression (Figs. 8 and 9).
While we cannot formally preclude some residual VACssBF1 binding to the truncated PE in P4/CE(5Ј), VACssBF1 binding to either strand of the two AP-1 elements in TFAP1 is highly unlikely given the lack of sequence homology between AP-1 and TEF-1 enhancer elements (13,23). Therefore, the ability of the CE to also repress the activity of a heterologous, AP-1-dependent tissue factor promoter ( Fig. 9) suggests that VACssBF2 can disrupt enhancer function by a mechanism that is independent of both VACssBF1 and the double-stranded enhancer-binding protein affected.
In a previous survey of selected TEF-1-dependent enhancers, VACssBF1 binding was found to be largely restricted to the VSM ␣-actin PE, but VACssBF2 readily interacted with a broader spectrum of ssDNA sequences including those of the cardiac troponin T MCAT motif and the SV40 GT IIC, SphI and SphII enhanson motifs (12). The ability of VACssBF2 to interact with enhancer elements present in diverse promoters, coupled with an intrinsic ability to disrupt enhancer function, greatly expands the regulatory potential of this ssDNA-binding factor. Thus, there is a strong likelihood that VACssBF2 participates in the regulation of a larger number of genes than is currently appreciated.
The chimeric promoter activation studies presented here provide important clues as to the functional role of VACssBF2. Given the ssDNA binding properties of VACssBF2 and its ability to function in the absence of VACssBF1, we postulate that VACssBF2 may interfere with enhancer activity by inducing a localized ssDNA conformation that diminishes proximal enhancer protein binding affinity. Precedence for a mechanism of this type exists in the literature. Several other ssDNA- FIG. 9. Effect of the VSM ␣-actin CE on the activity of an AP-1-dependent promoter. Upper panel, a synthetic copy of the CE was cloned upstream of two AP-1 binding sites within a heterologous tissue factor promoter construct (TFAP1) to create TF/CE(5Ј). VSMP3, VSMP4, P4/CE(5Ј), and P4/CEmu2(5Ј) are shown to highlight the relative positions of the TATA box and operative enhancer elements within the VSM ␣-actin and tissue factor promoters. Lower panel, the above constructs were tested in parallel transfections for CAT protein expression in serum and cycloheximide-stimulated AKR-2B fibroblasts as described under "Experimental Procedures." Cellular extracts were assayed for CAT protein by immunoassay. Data were normalized to CAT protein measured in the parent constructs VSMP4 and TFAP1 (defined as 100%). Values shown are the mean Ϯ S.E. of four independent transfection experiments.

VACssBF2-induced Suppression of Enhancer Function
binding proteins, including YB-1 (36) and the far-upstream element binding protein of c-myc (37)(38)(39), appear to influence transcription, in part, by altering DNA structure within specific promoter elements. However, because VACssBF2 does not appear to be enhancer type-selective, we cannot exclude the possibility that VACssBF2 does not disrupt enhancer protein dsDNA binding directly but rather functions to disrupt interactions between such proteins and components of the basal transcription complex. VACssBF1 binding within the context of the native promoter may be required to augment the function of VACssBF2 under some conditions or to increase specificity for the VSM ␣-actin TEF-1 enhancer element.
The issue of whether these ssDNA-binding proteins induce a localized ssDNA conformation or stabilize a preexisting, non B-DNA structure will ultimately require the purification, cloning, and characterization of these factors. A review of the literature reveals that VACssBF1 and VACssBF2 are not unique in their affinity for ssDNA encompassing a TEF-1 motif. Another ssDNA binding factor, designated muscle factor 3 (MF3), has been previously reported to interact with the noncoding strand of the muscle-specific TEF-1/M-CAT sequence element from the chicken skeletal ␣-actin gene (40). However, MF3 differs from either VACssBF1 or VACssBF2 in that it also interacts with dsDNA comprising the TEF-1/M-CAT motif (40). VACssBF1 and VACssBF2 have no detectable binding affinity for the double-stranded form of the VSM ␣-actin TEF-1-binding sequence (12). Moreover, while VACssBF1 and VACssBF2 ssDNA binding activity is restricted to tissues that are enriched in smooth muscle (13), MF3 is expressed in both skeletal muscle and liver (40).
Structural studies have revealed that while VACssBF1 is a single, M r ϳ 52,000, polypeptide, 2 VACssBF2 is more complex. UV cross-linking and SDS-polyacrylamide gel electrophoresis analyses indicated that VACssBF2 consists of multiple polypeptide components of differing apparent molecular weights (Fig. 5). Results of Southwestern blots confirmed that VACssBF2 consists of at least three distinct electrophoretic species, a M r ϳ 51,000/48,000 doublet and a larger molecular weight component (Fig. 6). The estimated apparent molecular weight of the more slowly migrating component varied somewhat depending upon the analytic technique employed (M r ϳ84,000 by UV cross-linking and M r ϳ 121,000 by Southwestern blotting). The inherent nonideality in electrophoretic mobility of a ssDNA-protein complex (UV cross-linking) versus protein alone (Southwestern blot) may account for this difference. Results of preliminary VACssBF2-RNA binding studies suggest that the more slowly migrating VACssBF2 species preferentially interacts with the mRNA counterpart of the CE, whereas the VACssBF2 doublet binds exclusively to ssDNA. 3 These data reinforce the conclusion that the M r ϳ 121,000 component is distinct from the VACssBF2 doublet and also suggest that this species may also play a role in post-transcriptional regulation.
Using a DNA-binding site screening methodology, VAC-ssBF1 has been tentatively identified on the basis of its molecular size, binding site specificity, and antigenic determinants as the murine homolog of YB-1, a member of the Y-box family of proteins. 2 Owing to the recent study that reported that YB-1 promotes the formation of single-stranded regions within the major histocompatibility DRA promoter and hence the transcriptional repression of the class II major histocompatibility genes (36), VACssBF1, in concert with VACssBF2, may play a similar role in the context of the VSM ␣-actin promoter. Analogous cloning strategies are now underway in an attempt to characterize the VACssBF2 ssDNA binding activities. Structural information provided by amino acid sequence determination will almost certainly provide more definitive insights into their molecular roles in transcriptional regulation.