A Sequence-selective Single-strand DNA-binding Protein Regulates Basal Transcription of the Murine Tissue Inhibitor of Metalloproteinases-1 (Timp-1) Gene*

Tissue inhibitor of metalloproteinases-1 (TIMP-1) is important in maintaining the extracellular proteolytic balance during tissue remodeling processes. To allow homeostatic tissue turnover, the murine Timp-1 gene is expressed by most cells at a low basal level, and during acute remodeling its transcription is activated by a variety of stimuli. A non-consensus AP-1-binding site (5′-TGAGTAA-3′) that is conserved in mammalian timp-1 genes is a critical element in basal and serum-stimulated transcription. We show here that each strand of this unusual AP-1 site binds a distinct single-stranded DNA-binding protein, although neither strand from a perfect consensus AP-1 site from the human collagenase gene shows similar binding. One of the single-strand binding factors, which we term ssT1, binds to a second upstream Timp-1 region between nucleotides −115 and −100. Deletion analysis demonstrated that this region is important in basal but not serum-inducible transcription. The ssT1 factor was 52–54 kDa by UV cross-linking of electrophoretic mobility shift assays and Southwestern blot analysis. Its binding to DNA shows sequence selectivity rather than specificity, with 5′-CT/ATTN(4–6)ATC-3′ as a favored motif. Multiple ssT1-like activities were found in nuclear extracts from mouse fibroblasts and rat liver and testis, suggesting that these factors may regulate basal Timp-1 transcription in a tissue-specific fashion.

The TIMP family includes four members, designated TIMP-1, -2, -3, and -4, each of which can inactivate the active forms of MMPs by formation of a tight 1:1 non-covalent complex (3). Although the TIMPs are largely interchangeable in their MMPinhibitory abilities, it has become clear that they have certain properties that are specific to particular family members, supporting the idea that they have unique as well as shared physiological roles. For instance, compared with TIMP-2, -3, and -4, TIMP-1 is a poor inhibitor of the membrane-type MMP group of MMPs, several of which function as cell surface activators of progelatinase-A (MMP-2 (4 -7)). TIMP-1 also stands out in its ability to form a complex with pro-gelatinase-B (MMP-9) (8), through which it regulates the rate of pro-MMP-9 activation (9,10). Finally, TIMPs show highly individual patterns of expression in vivo and in vitro (2,11), indicating that the use of a particular TIMP may be advantageous in certain tissue remodeling situations.
Of the four timp genes, timp-1 and -3 are both highly inducible in diverse cell types in vitro, in response to stimuli such as serum, phorbol ester, and transforming growth factor-␤ (12)(13)(14), whereas timp-2 expression remains largely constitutive under most stimulatory conditions (15). In the developing mouse embryo, strong expression of Timp-1 is restricted to sites where extensive tissue remodeling is occurring, particularly at sites of osteogenesis in the limbs, digits, ribs, vertebrae, and skull (16,17). Tight spatio-temporal control of expression is also seen in the adult ovary and uterus during pregnancy under the influence of paracrine and endocrine factors (18). However, many tissues express Timp-1 at a low basal level that can only be detected by sensitive techniques such as ribonuclease protection assay, possible reflecting a tissue maintenance function (16).
Previous work from several laboratories has shown that expression of Timp-1 is controlled principally at the level of transcription (19,20) which therefore dictates that a detailed analysis of its transcriptional regulation be performed. The timp-1 genes from mouse, rat, and human have been cloned and their promoter and regulatory regions analyzed (19,(21)(22)(23)(24). In several characteristics Timp-1 resembles a housekeeping gene, including the absence of a canonical TATA box, the presence of Sp1 sites, and multiple transcription start sites (25)(26)(27)(28). Additionally, the promoter proximal region contains several cis-acting regulatory elements that have been shown to be important for conveying transcriptional induction in response to stimuli. Most notable is a highly conserved region that has an AP1-(activator protein-1) binding motif in close proximity to a PEA3/Ets domain, an arrangement found in many other AP1-responsive genes (29). Several laboratories have demonstrated the importance of both of these regions, but clearly, the AP1 element is the principal site for conferring stimulatory responses following serum stimulation in vitro (27)(28)(29). Interestingly, the AP1-binding site is a non-consensus form (5Ј-TGAGTAA-3Ј) that differs by a single base from the consensus AP1-binding motif (5Ј-TGAGTCA-3Ј) found in most AP-1-responsive genes, including inducible MMPs such as human collagenase (mmp-1), stromelysin-1 (mmp-3), and gelatinase-B (mmp-9) (30,31). We previously showed by electrophoretic mobility shift assay that in addition to the ability of the Timp-1 Ϫ59/Ϫ53 AP-1 site to bind complexes containing Fos and Jun proteins, other proteins could bind that showed no affinity for the consensus collagenase AP-1 sequence (otherwise known as the 12-O-tetradecanoylphorbol-13-acetate-responsive element) (28).
We therefore undertook a more detailed investigation of the characteristics of the Timp-1 AP-1 site, in conjunction with a search for further important regulatory elements in the promoter proximal zone of the gene. We report here that singlestrand versions of the Timp-1 AP-1 site bind distinct nuclear factors and that the consensus AP-1-binding site does not share this ability. Furthermore, similar single-strand DNA binding activity is also seen at a site further upstream, at Ϫ115/Ϫ100. These interactions affect the basal expression of Timp-1, rather than its induction in response to serum stimulation. We speculate that these promoter-interacting single-stranded DNAbinding proteins may be significant factors in the dual housekeeping/inducible behavior of Timp-1 and other genes.
Transient Transfection Analysis of mTIMP-1 Luciferase and CAT Reporter Constructs-All transient transfections were performed by the Chen and Okayama (32) procedure essentially as described previously (28). Briefly, cells were plated at a density of 5 ϫ 10 5 cells/ml and grown overnight in DMEM/F12 containing 10% FBS. To transfect the cells, 20 g of plasmid was brought up to 900 l of final volume in TE, followed by the addition of 100 l of CaCl 2 (2.5 M). The mixture was then added dropwise to 1 ml of 2ϫ BES-buffered saline (50 mM BES, 280 mM NaCl, 1.5 mM Na 2 HPO 4 ) with constant vortexing, which enables the formation of DNA-calcium phosphate precipitates. Precipitation proceeded for exactly 20 min, and 1 ml of the mixture was added dropwise to each of two plates of cells. The cells were incubated for 18 h at 37°C in an atmosphere of 97% (v/v) air, 3% (v/v) CO 2 . After the transfection, the medium was exchanged for fresh DMEM/F12 containing 10% FBS medium, and the culture dishes were returned to 95% (v/v) air, 5% CO 2 for 8 h. The medium was then exchanged for serum-free DMEM/F12 overnight. The cells were stimulated for 24 h with serum as described above or given fresh serum-free media for basal conditions. Following stimulation, cell extracts were collected by harvesting with reporter lysis buffer as per manufacturer's instructions (Promega).
Luciferase reporter assays were performed using the luciferase assay system according to manufacturer's instruction (Promega, Madison, WI). Briefly, 20 l of cell lysate was mixed with 100 l of luciferase assay reagent, and then light emission was immediately measured on a luminometer (TD-20, Promega, Madison WI). Luciferase transfections were standardized to CAT activity by cotransfection with a cytomegalovirus promoter-driven CAT construct (CMV-CAT).
CAT assays were performed using standard methodology previously described (28). First, 20 g of protein extract was incubated for 10 min at 37°C with [ 14 C]chloramphenicol (0.0002 Ci) in Tris-HCl (pH 7.5), in a final volume of 100 l. Next, 20 l of 4 mM acetyl coenzyme A was added and incubated for 1 h at 37°C. The reaction was terminated by extraction with 900 l of ethyl acetate. The sample was then lyophilized, resuspended in 20 l of ethyl acetate, and spotted onto a thin layer chromatography plate. Chromatography was performed in an atmosphere containing 95% chloroform, 5% methanol. Finally, the plates are exposed onto x-ray film (Eastman Kodak Co.).
All transfections were standardized with the Hirt's assay for DNA input as described (33). Nuclear pellets (remaining from the cell lysate extraction) were lysed in 0.5 ml of Hirt's solution (0.6% SDS, 10 mM Tris-HCl (pH 7.5), 10 mM EDTA) for 10 min at room temperature. This was followed by the addition of 60 l of 5 M NaCl and 20 l of proteinase K (10 mg/ml), which was then incubated for 3 h at 37°C. After the protease digestion, the mixture was centrifuged for 5 min at 12,000 rpm, and the supernatant was retained and extracted with 1:1 phenol: chloroform. The samples were then prepared for slot-blot transfer by mixing 50 l of supernatant with 16.7 l of 1 M HCl for 5 min to denature, followed by the addition of 117 l of neutralization solution (0.9 M NaOH, 2.25 M NaCl) for 15 min. The volume was increased to 510 l with buffer (1.4 M Tris-HCl (pH 7.0), 1.5 M NaCl) and then loaded onto a slot-blot manifold for transfer onto duralon-UV membranes (Stratagene, La Jolla, CA). Following transfer, the samples were fixed by UV cross-linking, hybridized to specific probes (pBLCAT3 or pGL2basic), and exposed to x-ray film (Kodak). Intensity of signal (measured densitometrically) was then used to standardize the reporter expression to amount of input plasmid.
The Ϫ223/ϩ47 CAT constructs (coll.AP1, mutAP1, and mutPEA3) were created using a two-step PCR mutagenesis strategy utilizing the WT Ϫ223/ϩ47 construct (in pBluescript) described previously (28) as a template. The first step involves two separate reactions, using sense and antisense primers incorporating specific mutations from the wildtype sequence (Table I) and either T3 or T7 primers. In this manner, T3 and the appropriate antisense primer amplified the upstream half of the fragment (containing the mutation) and T7 and the corresponding sense primer amplified the downstream half. The products were then mixed and used as a template for the second round of PCR, using T3 and T7 only as primers. Full-length constructs will only be amplified if the upstream and downstream halves anneal, thereby generating the mutant constructs. Following PCR, the fragments were digested with HindIII/Sau3A and cloned into pBLCAT3. The constructs were verified by sequencing.
The Ϫ115/ϩ47 constructs (WT, coll.AP1 and mutAP1) were produced by PCR using the templates described above for mutant and coll.AP1. In the case where the ssT1 was mutated, the primer contained specific sequence changes to amplify a mutant product (Mut1; see Table I). The PCR fragments were then restricted with HindIII/BamHI and cloned into linearized pBLCAT3.
Oligonucleotides-Oligonucleotides for electromobility shift assays (EMSAs) and Southwestern blotting were synthesized by the UCDNA synthesis laboratory (University of Calgary, Canada) (Tables I and II). If annealing of the DNA was required to produce double-stranded probes, equimolar amounts of sense and antisense strands were diluted to 1 g/l in 10 mM Tris-HCl (pH 8.0), 100 mM NaCl and heated to 95°C. To facilitate annealing, the oligonucleotides were allowed to cool slowly to 4°C.
Probes-Single-stranded probes were generated by adding 60 ng of Nuclear Extract Preparation-A total of 2 ϫ 10 7 10T1/2 cells (either unstimulated or serum-stimulated for 30 min or 3 h) were scraped in 1 ml of phosphate-buffered saline, 0.1% Nonidet P-40. After a brief centrifugation (10 s, 12,000 rpm), the cells were decanted and rinsed in phosphate-buffered saline, 0.1% Nonidet P-40. Cells were visually inspected for lysis, centrifuged briefly again, and the supernatant aspirated. The nuclear pellet was then resuspended in 3 volumes of high salt buffer (25 mM HEPES (pH 7.8), 500 mM KCl, 0.5 mM MgSO 4 , and 1 mM DTT) with protease inhibitors (0.1 mg/ml leupeptin, 0.1 mg/ml phenylmethylsulfonyl fluoride, 0.1 mg/ml aprotinin) and extracted on ice for 20 min. Extracts were centrifuged at 12,000 rpm for 2 min (4°C), and the supernatant was transferred to a new tube. The protein concentration of the nuclear extract was determined by Bradford assay using the Bio-Rad assay kit (Bio-Rad) following the manufacturer's directions.
Electrophoretic Mobility Shift Assay-Radiolabeled probe (40,000 cpm) was incubated with 4 g of 10T1/2 cell nuclear extract in 1ϫ binding buffer (10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 0.5 mM DTT, 5 mM MgCl 2 , and 5% (v/v) glycerol) in the presence of nonspecific competitors (0.5 g of Sau3A cut pBluescript plasmid and 0.5 g of poly(dI-dC)⅐(dI-dC) (Amersham Pharmacia Biotech)), with a final reaction volume of 20 l. The binding reaction was electrophoresed on a 7% polyacrylamide gel (38:2 acrylamide:bisacrylamide) in 1ϫ TBE buffer (90 mM Tris-HCl, 90 mM boric acid, 2 mM EDTA) which had been prerun for at least 30 min at 4°C. The gels were run for 1.5 h at 10 V/cm 2 at 4°C to retain the stability of the protein-DNA complexes. Completed gels were dried onto Whatman 3M paper (Bio-Rad, model 583 gel dryer) and exposed to x-ray film (Kodak Biomax).
Southwestern Blotting-Nuclear extracts (30 g) were loaded onto a 10% SDS-polyacrylamide gel electrophoresis mini-gel (Bio-Rad) in an equal volume of loading buffer (5% SDS, 5 mM Tris-HCl (pH 6.8), 200 mM DTT, 20% (v/v) glycerol, 0.25% bromphenol blue), using a lane of pre-stained molecular weight markers as a standard (Life Technologies, Inc.). The samples were then transferred to nitrocellulose (Bio-Rad transblot transfer medium) using the mini-trans-blot electrophoretic transfer cell (Bio-Rad) as per manufacturer's directions. The blot was blocked for 1 h at 4°C in Blotto (50 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM EDTA, 1 mM DTT, 5% skim milk powder) with gentle agitation, followed by soaking overnight at 4°C in binding buffer (25 mM NaCl, 10 mM Tris-HCl (pH 7.5), 10 mM MgCl 2 , 5 mM EDTA, 1 mM DTT). The next day, the samples were soaked in binding buffer with single-stranded, end-labeled probe (100,000 cpm/ml) for 6 h to overnight at 4°C. Following the probing, the blots were rinsed 4 times in binding buffer for 8 min at 4°C and air-dried. The washed blots were then exposed to Biomax x-ray film (Kodak).

The Unusual Ϫ59/Ϫ53 AP1-binding Site Is the Key Regulatory Element in the Timp-1 Promoter Proximal Region-We
showed previously that the 5Ј-TGAGTAA-3Ј motif at Ϫ59/Ϫ53 in the Timp-1 promoter was able to bind AP-1 and activate transcription from a linked promoter (28). Subsequent studies have confirmed the importance of this site in both the human and mouse promoters (29,34,35). However, we found that this sequence was able to associate with additional proteins besides Fos/Jun AP-1, which were unable to bind to a consensus collagenase AP-1 site (5Ј-TGAGTCA-3Ј), and we speculated that this unusual AP1-binding site might convey additional properties to the Timp-1 promoter (28). To test the involvement of the Ϫ59/Ϫ53 AP-1 site in promoter activity, we incorporated point mutations that either convert the site to a consensus AP1 motif (coll.AP1), or render it non-functional (mut AP1). In addition we introduced an inactivating mutation into the neighboring PEA3/Ets-binding site ( Fig. 1). Transient transfection assays were carried out using CAT reporter plasmids carrying these alternative versions of the Ϫ223/ϩ47 region of the Timp-1 promoter. The mutAP1 construct confirmed that a functional AP1-binding motif is essential for maximal basal and seruminducible gene expression. In contrast, mutation of the PEA3/ Ets site led to a more modest reduction of activity, to about 60% of the maximum observed with the wild-type promoter. Fig. 1 also shows that the unusual Timp-1 AP1 and the consensus AP1 sites have equivalent abilities to drive both basal and serum-stimulated expression of CAT in the context of the Ϫ223/ϩ47 region.
A Single-stranded Binding Protein Interacts with the Timp-1 Promoter AP-1 Region-We examined in more detail the nature of the nuclear factors that are able to interact with the Ϫ59/ Ϫ53 AP1 and the collagenase AP1 sites. We performed electrophoretic mobility shift assays (EMSA) using an annealed, PNK-labeled double-stranded probe corresponding to sequences from Ϫ63 to Ϫ49 (Fig. 2). It is important to note that this procedure may produce a mixed population of single-and double-stranded probes if one of the oligonucleotides is present in slight excess over its partner. Two complexes were resolved using 7% polyacrylamide gels. A slow migrating complex la- beled A was competed by both unlabeled Ϫ63/Ϫ49 Timp-1 and collagenase AP1 sequences (lanes 3 and 4). Band A was also supershifted when anti-Fos or anti-Jun antibodies were added to the binding reactions (data not shown). These data identify shift A as AP1 bound to the Ϫ63/Ϫ49 probe, confirming our previous data and other reports (28,29). However, in addition to AP1 binding activity, Ϫ63/Ϫ49 also demonstrated additional specific binding properties (band B) that were not competed by the collagenase AP1 (lane 4). Unlabeled single-stranded Ϫ63/ Ϫ49 sequences were effective competitors of complex B formation (lanes 7 and 8). The AP1 shift was not competed by singlestranded Ϫ63/Ϫ49 oligonucleotides, and single-stranded collagenase AP1 cold competitors (top or bottom strands) were unable to compete complexes A or B. These data demonstrate that complex B represents the interaction of a nuclear factor that is not AP1 with the individual strands of the Ϫ63/Ϫ49 region and that this interaction is specific since it could not be competed by a closely related perfect consensus AP1 motif.
To examine whether the complex B shift was attributable to one of the oligonucleotides that comprise the Ϫ63/Ϫ49 probe (i.e. the top or bottom strands), we carried out EMSAs using each labeled oligonucleotide separately, as well as corresponding sequences from the collagenase AP1 (Fig. 2B). The samples were electrophoresed longer for greater resolution of the bands. Both Timp-1 Ϫ63/Ϫ49 strands were able to complex with nuclear factors, generating bands with slightly different mobilities that are labeled as B and BЈ for the top and bottom strand probes, respectively. The corresponding single strands of the collagenase AP1 sequence did not interact with anything in the nuclear extracts. We were unable to carry out competition assays to determine whether both top and bottom Ϫ69/Ϫ43 strands interact with the same protein, since an excess of the complementary strand generated double-stranded DNA, which does not bind the single strand-specific nuclear factor, as will be shown later in Fig. 5. However, other data to be discussed argue that the B and BЈ complexes involve different nuclear factors.
A Positive Regulatory Element That Affects Basal Timp-1 Transcription Is Located between Ϫ125/Ϫ95-Whereas the above studies were ongoing, we also aimed to refine the analysis of regulatory elements in the promoter proximal region, since our previous studies had shown that additional positive acting elements must reside within the Ϫ223/ϩ47 promoter used in Fig. 1 (28). Constructs with alternative 5Ј-end points were generated by PCR and subcloned into the luciferase reporter plasmid, pGL2-basic, and then used in transient trans- fection analyses (Fig. 3). The shortest construct (Ϫ95/ϩ47) showed an approximate 4-fold reduction in both basal and serum-stimulated expression compared with the longest (Ϫ223/ϩ47). The principal positive regulatory region was located between Ϫ125/Ϫ95, as its inclusion resulted in a 3-fold increase in luciferase expression in both basal and seruminduced conditions. Since the fold increase from basal to serumstimulated conditions remained essentially unchanged with inclusion of sequences upstream from Ϫ95, we conclude that the primary effect of these sequences is at the level of basal (unstimulated) expression. A minor but consistent increase in serum-stimulated expression from sequences between Ϫ223/ Ϫ195 suggests the actions of an additional inducible transcription factor in this region, but these have not been followed up at this time.
A Single-stranded DNA-binding Protein Interacts with the Timp-1 Promoter at Ϫ115/Ϫ100 -To investigate the nature of the nuclear factors able to interact with the Ϫ125/Ϫ95 region, a series of probes were made to obtain a detailed EMSA analysis (Fig. 4A). Protein-DNA interactions were observed with the Ϫ115/Ϫ100 probe in addition to the full-length Ϫ125/Ϫ95 probe (lanes 2 and 4). However, no shift was detected for probes corresponding to either the Ϫ123/Ϫ110 or Ϫ108/Ϫ92 regions (lanes 3 and 5). Three bands were seen with the Ϫ125/Ϫ95 probe, although only a single shift was seen for the Ϫ115/Ϫ100, suggesting that the Ϫ115/Ϫ100 region contains a critical binding site for a nuclear factor, but flanking sequences may also influence the nature of the complex that is formed. With this preparation of PNK-labeled Ϫ63/Ϫ49 probe, we observed three bands migrating in the position of the B/BЈ complexes seen with individual top and bottom Ϫ63/Ϫ49 single-stranded probes in Fig. 2B. The number and intensities of these bands labeled B/BЈ were somewhat variable with different experiments and nuclear extracts (see also Fig. 5A). The data shown in Fig. 4 were obtained with a nuclear extract from C3H 10T1/2 cells stimulated for 30 min with serum, but there were no qualitative or quantitative changes in band patterns with extracts from unstimulated or 3-h serum-stimulated cells.
Since the experiment in Fig. 4A was performed with annealed, PNK-labeled, oligonucleotides that may contain both double-stranded and single-stranded forms (as in Fig. 2), we also carried out an analysis with Klenow-labeled doublestranded probes. The Klenow-labeled, double-stranded Ϫ115/ Ϫ100 and Ϫ125/Ϫ95 probes were greatly impaired in their ability to interact with DNA-binding proteins compared with the PNK-labeled probes (Fig. 4B; compare lanes 2 with 7, and 4 with 8). This reduction paralleled the diminished representation of the B/BЈ shifts seen when the Klenow-labeled Ϫ63/49 Timp-1 AP1 probe (Fig. 4B, lane 9) was compared with the same probe labeled with PNK (Fig. 4A, lane 1). These data suggested that as is the case for the B/BЈ shifts with the Ϫ63/Ϫ49 probes, a single-stranded DNA-binding protein may interact with the Ϫ115/Ϫ100 mTIMP-1 promoter region.
Both Ϫ115/Ϫ100 and Ϫ63/Ϫ49 Promoter Regions Interact with the Same Single-stranded DNA-binding Protein-We carried out EMSA with single-stranded PNK-labeled probes from the Ϫ125/Ϫ95 region (Fig. 5). Single-stranded probes corresponding to Ϫ125/Ϫ95, Ϫ115/Ϫ100, and Ϫ63/Ϫ49 all show complex patterns of protein interactions (Fig. 5A). Strong shifts were obtained with both the Ϫ125/Ϫ95-top and Ϫ115/Ϫ100-top oligonucleotides but not with corresponding bottom strands (lanes 1-4). No shifts were seen with either top or bottom strands from the Ϫ123/Ϫ110 and Ϫ108/Ϫ92 sequences (data not shown). In contrast, both top and bottom Ϫ63/Ϫ49 strands were able to interact with nuclear proteins, giving B/BЈ shifts as in Fig. 2. Since all of the probes used had the same additional nucleotides added at the 5Ј-ends to generate restriction enzyme cloning sites (5Ј-AGCTT-for the "top" strand probes and 5Ј-GATCC-for the "bottom"), we also generated and tested additional single-stranded probes lacking these sequences. These probes gave the same results as those shown in Fig. 5 (data not shown), confirming that the sequence-specific DNA binding that we had observed was attributable to the indicated Timp-1 sequences alone.
In order to determine the specificity of the protein-DNA interactions the Ϫ115/Ϫ100-top probe was used for EMSA in competition with various unlabeled oligonucleotides (Fig. 5B). The top strand of Ϫ115/Ϫ100 was competed effectively by Ϫ115/Ϫ100-top and Ϫ125/Ϫ95-top (lanes 10 and 12), demonstrating specificity of the complex. Surprisingly, competition between the Ϫ63/Ϫ49-bottom strand and the Ϫ115/Ϫ100-top strand was also seen (lane 15), although the top strand Ϫ63/ Ϫ49 did not compete (lane 14). This competition from the

FIG. 2. A single-stranded binding protein interacts with the
Timp-1 AP1-binding site. A, electrophoretic mobility shift assays were performed on C3H10T1/2 nuclear extracts using an annealed PNK-labeled probe that contains the Timp-1 AP1-binding site. End labeling in this way generates both double-stranded and singlestranded probe. Lane 1, no nuclear extract; lane 2, nuclear extract without added competitor DNA; lanes 3-11, extract plus the indicated single-or double-stranded competitors. In lane 2, two complexes were seen, labeled A and B. As discussed under "Results," complex A corresponds to AP1 and complex B to the interaction of single-stranded Ϫ63/Ϫ49 oligonucleotides with a nuclear factor. A further indication of single-stranded DNA-protein interactions was shown by labeling single-stranded probes for EMSA (B). Top or bottom strands of both Ϫ63/ Ϫ49 and the corresponding strands from the collagenase consensus AP1-binding site were labeled and used for EMSA with nuclear extracts from 10T1/2 cells, showing two distinct bands, B and BЈ. These bands were resolved more compared with A by longer electrophoresis.
Ϫ63/Ϫ49-bottom strand implies that the BЈ shift obtained with the Ϫ63/Ϫ49 probe likely involves the same protein that interacts with the Ϫ115/Ϫ100 region or a closely related protein.
Moreover, it indicates that the B and BЈ shifts seen with the Ϫ63/Ϫ49-top and -bottom strands involve different nuclear factors.
The bottom strands of either Ϫ125/Ϫ95 or Ϫ115/Ϫ100 were also effective "competitors" of the Ϫ115/Ϫ100-top shift, but this is misleading. The bottom strand would anneal to the probe to form duplex DNA, which we had established in Fig. 4 has relatively poor binding to the single-stranded DNA binding factor. Such annealing is seen in a slower migration of the free probe at the bottom of the EMSA gel in Fig. 5B in the lanes with either Ϫ125/Ϫ95-bottom or Ϫ115/Ϫ100-bottom as cold competitor. Neither the Ϫ63/Ϫ49-top nor either of the coll.AP1 strands competed for the same nuclear factors that bind to the Ϫ115/Ϫ100-top probe (lanes 14, 16, and 17).
These data establish that a nuclear single-stranded DNA binding factor, hereafter called ssT1, that is unrelated to AP1 interacts with both the Ϫ63/Ϫ49 AP1-bottom and the Ϫ115/ Ϫ100-top sequences.
The Single-stranded DNA-binding Protein Shows Relaxed Specificity-The sequence of Ϫ115/Ϫ100-top is 5Ј-agcttATCT-TTGGGTTTATC-3Ј and Ϫ63/Ϫ49-bottom is 5Ј-gatccGCAT-TACTCATCCA3-Ј. There is minimal similarity between these sequences, the two commonalities being 5Ј-C(T/A)TT motif and a 5Ј-ATC sequence toward the 3Ј-side. This prompted us to analyze further the specificity of the sequences in which the protein binds. Mutations were produced in which the first 8 bases of the Ϫ115/Ϫ100-top sequence were replaced with the first 8 bases of the consensus coll.AP1-top sequence that was shown earlier to not bind to ssT1 (2B) (mutant 1). Similarly, the last 8 bases of the Ϫ115/Ϫ100 were replaced with the last bases of the coll.AP1-top (mutant 2). Both single-stranded mutants were labeled and used as probes for EMSA (6B-left) and also used as cold competitors against the wild-type Ϫ115/Ϫ100-top probe (6B-right). When end-labeled and used to probe nuclear extracts, both mutants were impaired in their interaction with ssT1 compared with the wild-type sequence. However, disruption of the upstream 7 bases (mut1) had a far greater effect, greatly reducing complex formation compared with mut2. This observation is mimicked by cold competition experiments, where both mut1 and mut2 were not as effective a competitor as wild-type sequences, with mut2 being a slightly better competitor compared with mut1. These results show that both halves of the Ϫ115/Ϫ100 sequence contribute to protein-DNA interaction, with a greater contribution of the upstream half.
We also tested competitor oligonucleotides in which 3 bases at a time in the Ϫ115/Ϫ100-top sequence were varied. Quantification of the abilities of these DNAs to compete with the wild-type sequence in EMSA is shown in Fig. 6C. None of these mutations completely eliminated competition ability, although disruption of the 5Ј-CTT at Ϫ113/Ϫ111 as well as the 5Ј-ATC at Ϫ102/Ϫ100 were the most deleterious. These data suggest that both of these sequences may contribute to the ssT1 showing sequence selectivity for a consensus 5Ј-CA/TTTN 4 -6 ATC motif. Deletions spanning Ϫ223 and Ϫ95 were produced by the polymerase chain reaction, and the resulting fragments were cloned into pGL2-basic luciferase reporter plasmid (Promega). Following transfection of the deletion constructs, nuclear extracts were prepared from C3H10T1/2 cells cultured in serum-free media (unstimulated) and cells that were serumstimulated for 24 h. Luciferase reporter activity for each trial was standardized for sample-sample variation by measuring the input DNA using the Hirt's assay (as described under "Methods and Materials"). The graph shows relative activity, with the largest value displayed by the Ϫ123/ϩ47 construct in serum-stimulated cells, designated as 1.
The data also indicate that there are likely many sequences that will interact with the ssT1 nuclear factor.
The ssT1 Nuclear Factor Is a Positive Regulator of the Ϫ63/ Ϫ49 Region of the Timp-1 Promoter-To analyze further the role of ssT1 in basal promoter activity, we assessed the impact of mutation of the ssT1 site in the context of Ϫ115/ϩ47 constructs with either a normal Ϫ63/Ϫ49 AP1 site or with the site replaced by a consensus coll.AP1 motif or mutant AP1 as in Fig.  1. Since both the mutant AP1 and the coll.AP1 sequences do not interact with ssT1, this allowed us to discriminate the contributions of the Ϫ115/Ϫ100 and Ϫ63/Ϫ49 ssT1 sites in basal promoter activity. Incorporation of the mut1 sequence used in Fig. 6 into the Ϫ115/ϩ47 reporter caused a 15-20% reduction in promoter activity (Fig. 7). However, this mutation had greater impact in the reporter carrying the coll.AP1 site, reducing expression by 40%. Elimination of AP1 binding ability had a dramatic effect on promoter activity, as we had seen previously with the Ϫ223/47 constructs used in Fig. 1, with the mutant AP1 construct yielding only 20% of the expression seen with the Ϫ115/ϩ47 wild-type Timp-1 promoter region. However, this low basal level could be further reduced to approximately 10% that of the wild-type promoter by inclusion of the mut1 sequence at Ϫ115/Ϫ100. These data argue that both ssT1-binding sites at Ϫ63/Ϫ49 and Ϫ115/Ϫ100 contribute to the basal activity of the Timp-1 promoter. We suspect that the effect of mutating the ssT1 site is not as severe as deleting it (compare Fig. 7 with Fig. 3) because the mut1 mutation still retains some ability to interact with ssT1 (Fig. 6).
Both Regions, Ϫ115/Ϫ100-Top and Ϫ63/Ϫ49-Bottom, Interact with a Protein of Approximate Molecular Mass of 50 -55 kDa-To characterize further the ssT1 factor, EMSA binding reactions with the Ϫ115/Ϫ100-top probe were UV cross-linked and then separated on a 10% SDS-polyacrylamide electrophoresis gel (Fig. 8A). A complex pattern of bands was observed migrating between 40 and 65 kDa; however, two major bands at approximately 50 and 55 kDa were detected. Specificity of these interactions is demonstrated by competition from excess unlabeled Ϫ115/Ϫ100-top oligonucleotides. The complexes were not competed by cold coll.AP1 single strands, but they were competed by Ϫ63/Ϫ49-bottom as expected (data not shown).
As further characterization of the ssT1 factor, Southwestern blots were performed (Fig. 8B). The Ϫ115/Ϫ100-top and Ϫ63/ Ϫ49-bottom oligonucleotides both interact with proteins from nuclear extracts of 10T1/2 cells and rat liver, and different binding activities were revealed. All of the bands labeled 1-4 were specific since they were eliminated by adding excess unlabeled oligonucleotide competitor to the binding solution (data not shown). The Ϫ115/Ϫ100-top strand interacted strongly with a protein of approximately 54 kDa in nuclei from mouse 10T1/2 fibroblasts, identified as band 2 (Fig. 8B). There are, however, additional weaker interactions with bands 1 and 3 at 90 and 32 kDa, respectively. When using the Ϫ63/Ϫ49-bottom probe with 10T1/2 nuclear extracts, we observed a similar banding pattern; however, there is an approximately equal distribution of signal between bands 1 and 3, with an additional band 4 at 22 kDa being detected. Band 2 co-migrated for both Ϫ63/Ϫ49 and Ϫ115/Ϫ100 probes, which likely accounts for the cross-competitions we have observed. We suggest that the band-2 54-kDa nuclear binding activity identified by this Southwestern analysis is the ssT1 factor. At this time, we do not know the identity of any of the bands. It is possible that the lower molecular weight bands (labeled 3 and 4 in Fig. 8B) are breakdown products of the higher molecular weight species. The band 2 signal is the only significant binding activity detected in liver nuclear extracts with either the Ϫ115/Ϫ100-top or the Ϫ63/Ϫ49-bottom probes; in the case of the weak signal with the Ϫ63/Ϫ49-bottom this resolved into a doublet. Both probes interacted with proteins of approximately 54 kDa in liver nuclear extracts; however, Ϫ115/Ϫ100-top gave a much stronger response. Alternatively, only Ϫ115/Ϫ100-top interacted with nuclear factors from rat testes, giving a series of 4 -5 bands of approximately 58, 54, 48, 40, and 36 kDa, which may represent a distinct family of testis-specific single-stranded binding proteins or different isoforms of the ssT1 factor. DISCUSSION The mechanisms involved in the tissue-specific and stimulus-responsive transcription of Timp-1 are not fully understood. Several groups have now demonstrated that the AP1binding site in the promoter of mammalian timp-1 genes is of critical importance in serum-inducible transcription in fibroblastic cells (27)(28)(29). We show here that this site, which differs from a consensus AP1-binding motif by a single base (5Ј-TGAG-TAA-3Ј), confers additional protein binding properties on single-stranded versions of the sequence covering Ϫ63/Ϫ49 of Timp-1. Our data indicate that the top and bottom strands of the sequence bind distinct nuclear single-stranded DNA binding factors. Double-stranded Ϫ63/Ϫ49 AP1 probes either do not bind these factors or bind them very inefficiently. Likewise single-or double-strand versions of a consensus collagenase AP1 site with core motif 5Ј-TGAGTCA-3Ј do not bind either factor. The bottom strand of the sequence binds a 54-kDa protein that we have termed ssT1. The ssT1 factor binds to a second site in the Timp-1 promoter between Ϫ115/Ϫ100, deletion of which results in a 3-fold decrease of both basal and serum-stimulated transcription from the promoter. These data argue that ssT1 may be involved in maintaining efficient transcription of Timp-1 in unstimulated basal conditions in mouse fibroblasts, which in turn may affect the overall level of gene activity that can be attained following stimulation.
The binding of ssT1 to ssDNA shows clear sequence prefer- Nuclear extracts from C3H10T1/2 cells that had been serum-stimulated for 30 min were used for assays with probes corresponding to Ϫ63/Ϫ49, Ϫ125/Ϫ95, Ϫ123/Ϫ110, Ϫ115/Ϫ100, and Ϫ108/Ϫ92 of the Timp-1 promoter and the consensus AP1 site (coll.AP1). The "B" complex seen previously from the Ϫ63/Ϫ49 probe in Fig. 2 migrated as several bands that are indicated by the bracket in A. The B bands for the Ϫ63/Ϫ49 probe were greatly diminished when the Klenow-labeled probe (B) was used. This was also the case for co-migrating bands seen with PNKlabeled Ϫ125/Ϫ95 and Ϫ115/Ϫ100 (compare A and B).
ences. Although the Ϫ63/Ϫ49 AP1-bottom strand and the Ϫ115/Ϫ100-top strand were able to cross-compete effectively for binding of the ssT1 factor, and they both detected a 54-kDa protein by Southwestern blot analysis, the Ϫ63/Ϫ49 sequence bound strongly to other proteins at 90, 32, and 22 kDa that were only weakly bound (if at all) by the further upstream site. Thus ssT1 may be one of a family of factors, each of which may prefer particular sequence motifs. The distinction between ssT1 and the factor that binds the Ϫ63/Ϫ49 AP1-top strand was shown by both competition data and UV cross-linking studies analogous to those of Fig. 7A, which revealed bands at approximately 35, 41, and 50 kDa (data not shown).
Comparison of the Ϫ63/49 AP1-bottom and the Ϫ115/Ϫ100top sequences suggests a possible consensus motif for ssT1 binding as 5-CA/TTTN 4 -6 ATC-3Ј. Within this motif the 5Јsequences may be the most critical for several reasons. First, the underlined T residue indicates the distinguishing difference between the unusual Timp-1 AP1 site and the consensus collagenase AP1. Second, mutational analysis involving fusing either half of the Ϫ115/Ϫ100 sequence to the inactive collagenase AP1 site indicated the loss of the first half containing the 5Ј-CTTT was somewhat more deleterious (Fig. 6B), and this was supported by additional mutations involving triplet replacements through the sequence (Fig. 6C). Third, this sequence is most conserved between mouse and rat (5Ј-CTTT-GGGTTTATC-3Ј versus 5Ј-CTTTGGGCTCAGC, respectively (24)). However, these mutational studies also show that multiple sequences participate since disruption of the 5Ј-CTTT still allowed some binding. Furthermore, either half of the Ϫ115/ Ϫ100-top sequence alone was insufficient to confer ssT1 binding, as shown by the failure of the Ϫ123/Ϫ110-top and the Ϫ108/Ϫ92-top sequences to bind. Thus sequences around the motif may also contribute to binding preferences, and as a consequence we prefer to term the binding of ssT1 "sequenceselective" rather than sequence-specific.
The DNA-ssT1 interaction data from EMSA studies complement the functional analysis of promoter activity from transient transfection studies of the various deletions of the Timp-1 promoter. Loss of the Ϫ125/Ϫ95 region lowered basal activity from the promoter without a profound effect on the fold induc-tion following serum stimulation. Mutation of the ssT1 site located at either Ϫ115/Ϫ100 or Ϫ63/Ϫ49 are both associated with a decrease in promoter activity under basal conditions. Likewise, ssT1 was present at equal levels in nuclei isolated from unstimulated and serum-stimulated mouse fibroblasts. This supports the idea that ssT1 may function to maintain the housekeeping level of Timp-1 promoter activity.
The involvement of sequence-selective DNA-binding proteins in transcriptional regulation has been documented for other genes. The muscle factor 3 (MF3) single-stranded binding activity (36) binds to three individual sequences that show few significant regions of identity as follows: the cARG motif of muscle regulatory element (CC(A/T) 6 GG), the E box of creatine kinase (TCAGGCAGCAGGTGTTGGGGG), and MCAT (CAT-TCCT), which is found in many muscle gene promoters. However, the relevance of these interactions remains unknown.
At present, we can only speculate about the function of the ssT1-DNA interaction. A number of genes are regulated in part by interactions with single-stranded DNA binding activity through a number of different mechanisms. Control of the adipsin gene bears similarity to what we have found for Timp-1, as it is regulated by two factors each specific for singlestranded DNA, with little double-stranded DNA binding activity (37). One of the two single-stranded DNA-binding proteins is expressed in a differentiation-dependent fashion and is thought to play a role in establishment or maintenance of the differentiated state. A 40-base pair regulatory region upstream of the gelatinase-A (MMP-2) promoter is involved in high level expression of the gene in glomerular mesangial cells (38). It has recently been shown that this site binds transcription factors AP2 and YB-1 (39), with YB-1 showing preferential binding to the isolated single strands of the response element (38).
Other identified single-stranded DNA-binding proteins provide additional possible modes of action. The ssDNA-BP, DNA binding stimulatory factor interacts with purified estrogen receptor, enabling it to bind to its response element (40), which implicates transcription factor recruitment as a mechanism of transcriptional activation. Such recruitment is also seen for the A␣ core protein, which is involved in regulation of the A␣ fibrinogen gene. The A␣ core protein, which is related to the FIG. 5. Two regions of the Timp-1 promoter, ؊115/؊100-top, and ؊63/ ؊49-bottom cross-compete for the same single-stranded DNA binding factor. Single-stranded probes corresponding to the Ϫ125/Ϫ95, Ϫ115/Ϫ100, Ϫ63/Ϫ49, and collagenase AP1-binding site were used for EMSA analysis with C3H10T1/2 cell nuclear extracts (A). Complexes were formed with the Ϫ125/ Ϫ95 top and Ϫ115/Ϫ100-top and both strands of the Ϫ63/Ϫ49 region. B, the Ϫ115/Ϫ100-top strand was labeled and used in EMSA with the indicated unlabeled competitor oligonucleotides (100ϫ molar excess). mitochondrial ssDNA-BP, P16 (41), has been shown in overexpression studies to be involved in interleukin-6-induced transcription, possibly through recruitment of STAT signaling molecules (42). Work on the rat timp-1 promoter identified the sequences between the AP1 and PEA3/Ets sites (which are precisely conserved in the mouse promoter, corresponding to Ϫ53/Ϫ45) as an oncostatin-M/interleukin-6-responsive element that binds STAT3 (24). It will be interesting to determine if ssT1 is in any way involved in STAT3 recruitment.
Another mechanism of activity for ssDNA-BP activity is through a direct recruitment of RNA polymerase (RNAP), as shown by the coliphage protein N 4 SSB, which activates 70, and does not bind to double-stranded DNA (43). It has been demonstrated that the protein interacts directly with the RNAP BЈ subunit, which has relevance to eukaryotic transcription because the region of interaction is conserved in the largest subunit of the eukaryotic RNAP II. There are several possible functions of interaction with RNAP subunits. The protein could act as a tether to link the RNAP to a promoter (44,45). Alternatively, subsequent steps of RNAP function could be targeted (43). Another mechanism of action, as demonstrated by the EcoSSB single-stranded binding protein is the enhancement of transcription by establishing adequate DNA secondary structure (46). Finally, a single-stranded DNA-BP might be involved in establishment of single-stranded regions at sites of transcription. Such activity might serve to enable open DNA complex formation during initiation by the RNAP or, alternatively, to maintain an open state in order to relieve torsional stresses during the act of transcription itself (47,48). It is possible that ssT1 induces some conformational change in DNA once bound, since this may explain why the EMSA complex of ssT1 with the Ϫ125/Ϫ95-top probe migrates faster than the corresponding FIG. 6. The single-stranded DNAbinding protein displays sequence selectivity. The specificity of interaction of the nuclear factors with the Ϫ115/ Ϫ100-top probe was characterized by EMSA using various competitor oligonucleotides (listed in A). B, fusion oligonucleotides were made combining the 5Ј-half of the collagenase AP1-binding site to the 3Ј-half of the Ϫ115/Ϫ100-top binding sequence (mut1) or the 5Ј-half of Ϫ115/ Ϫ100-top to the 3Ј-half of collagenase AP1 (mut2). These mutants were used for EMSA analysis both as probes with C3H10T1/2 nuclear extracts and as cold competitors against Ϫ115/Ϫ100-top probes (B). Both mut1 and mut2 oligonucleotides were less efficient at complex formation than the WT Ϫ115/Ϫ100-top. Cold competitor oligonucleotides were used at 5ϫ, 25ϫ, and 125ϫ molar excesses. Both mut1 and mut2 oligonucleotides competed for binding to the factor and the overall competitor effectiveness was WT Ͼ mut2 Ͼ mut1. C, another set of mutants was designed such that the regions of mutation were more refined. Purine/pyrimidine switches were constructed within the Ϫ115/Ϫ100-top sequence corresponding to the numerical name of the oligonucleotide (for example, bases Ϫ115 and Ϫ114 were mutated in the Ϫ115/Ϫ114 oligonucleotide). This mutant series was used as cold competitors in EMSAs, and the resulting shifted bands were measured densitometrically using NIH Image TM . The resulting plots shows level of competition relative to the wild-type oligonucleotide sequence.
complex with the Ϫ115/Ϫ100 oligonucleotide, despite the larger size of the former.
The inability of ssT1 to bind to double-stranded DNA appears to rule out a function in denaturation of the promoter region. However, the possibility remains that the proteins may stabilize a single-stranded conformation. Other possibilities exist for preinitiation complex recruitment and binding being aided by the ssDNA-BP. Such a function is seen for PC4, which is a positive transcription cofactor (49 -51). PC4 interacts with both free and bound VP16 activation domains and also with TFIIA. However, there is no interaction with TATA-binding protein. Similar to ssT1, PC4 interacts with single-stranded DNA (52).
Conversion of the unusual timp-1 AP1 motif to a consensus collagenase AP1 site did not affect basal or serum-inducible expression from the promoter in transient transfection assays with a reporter driven by the Ϫ223/ϩ47 promoter. This may be due to compensation via the presence of the Ϫ115/Ϫ100 ssT1binding site. However, it is clear that the Ϫ223/ϩ47 Timp-1 region does not recapitulate the full range of expression of the endogenous gene; for instance, expression from these constructs is unresponsive to phorbol ester stimulation. 2 Thus the Ϫ63/Ϫ49 AP-1-bottom ssT1 binding may be more important in the context of other regulatory elements upstream or downstream from the promoter. It is also possible that the full effects of ssT1 function will only be evident with a chromatin template, rather than the naked DNA format generated in the transient transfections. Another potential area of involvement of ssT1-like factors is tissue-specific expression. Only one cell line (fibroblastic) and two tissue types (liver and testes) were analyzed by Southwestern blot, yet significant differences in binding activities were seen. The 54-kDa ssT1 that is recognized by both the Ϫ115/Ϫ100-top and the Ϫ63/Ϫ49 AP1-bottom probes is present in fibroblasts and liver, but the testis pattern was more complex. It will be important to determine the relationships between these various binding activities.
In conclusion, we have identified a 54-kDa nuclear singlestranded binding activity that is involved in establishing basal expression of Timp-1 through interaction with at least two regions of the promoter. The ssT1 factor and related molecules may be involved in the regulation of a number of genes. Therefore, efforts at purifying the protein are presently under way. FIG. 7. ssT1 interactions at both promoter regions (؊115/؊100 and ؊63/؊49) positively affect reporter activity. Mutations of the ssT1 site at Ϫ115/Ϫ100 and the Ϫ63/Ϫ49 AP1 site were introduced into the Ϫ115/ϩ47 Timp-1 promoter region. The Timp-1 AP1 motif (WT) was converted to a consensus collagenase or mutant AP1 site (coll.AP1 and mutAP1, respectively, as in Fig. 1) and placed in the context of wild-type or mutant ssT1 (mut1, Fig. 6). Relative basal CAT activities following transient transfection of C3H10T1/2 cells are displayed. Mutation of the Ϫ115/Ϫ100 ssT1 site reduced basal expression relative to the wild-type promoter, and this reduction was exacerbated with either a canonical coll.AP1 site or mutated AP1 at Ϫ63/Ϫ49. Samples from an EMSA binding reaction were UV cross-linked and then separated on a 7% SDS-polyacrylamide electrophoresis gel (A). The Ϫ115/Ϫ100 was used as a probe and was run without competitors or against 100-fold molar excess of cold Ϫ115/Ϫ100-top or Ϫ115/Ϫ100bottom. As a control, the reaction was run without UV cross-linking. Additionally, Southwestern blots were performed on nuclear extracts from C3H10T1/2 cells, rat liver, and rat testes (B). Blots were probed with Ϫ115/Ϫ100-top and then stripped and probed with Ϫ63/Ϫ49bottom. Four primary bands were seen, labeled 1-4.