Binding of upstream stimulatory factor to an E-box in the 3'-flanking region stimulates alpha1(I) collagen gene transcription.

Since several lines of evidence implicate the 3′-flanking region in regulating α1(I) collagen gene transcription, we analyzed 12.4-kilobase pairs of 3′-flanking sequence of the murine α1(I) collagen gene for transcriptional elements. A region of the 3′-flanking region stimulated expression of the heterologous β-globin gene promoter in an enhancer trap plasmid and of the α1(I) collagen gene promoter in a collagen-luciferase reporter gene construct when located 3′ to the luciferase reporter gene. DNase I footprinting analysis demonstrated the presence of three regions where DNA binding proteins specifically interact within this 3′-stimulatory region. Inspection of the DNA sequence revealed a consensus E-box, a binding site for basic helix-loop-helix proteins, in one of the protein binding sites. Mobility shift assays demonstrated that upstream stimulatory factors (USF) USF-1 and USF-2 bind to this E-box. Mutating the E-box in the context of the 3′-flanking region confirmed that it contributes to the enhancement of transcriptional activity of the α1(I) collagen gene promoter. Mutations in all three protein binding sites abolished transcriptional activation by the 3′-flanking region, suggesting a complex interaction among the trans-acting factors in enhancing transcriptional activity. Thus, a region of the 3′-flanking region of the α1(I) collagen gene stimulates transcription of the α1(I) collagen gene promoter, and USF-1 and USF-2 contribute to this transcriptional stimulation.

Type I collagen, the most abundant protein in vertebrates, has diverse biological functions. It promotes cell migration, differentiation, and tissue morphogenesis during development. Additionally, it provides tensile strength to connective tissues such as bone, tendons, and skin, and forms a supporting framework of connective tissues in all major internal organs and the vascular system. Type I collagen is also the major protein produced during repair of tissue injuries and wound healing. Excess deposition of type I collagen occurs in fibrogenic dis-eases, such as hepatic fibrosis (1), pulmonary fibrosis (2), primary systemic sclerosis (3), and eosinophilic myalgia syndrome (4). Type I collagen is the product of two genes, the ␣1(I) and the ␣2(I) collagen genes, whose products form a heterotrimeric protein composed of two ␣1(I) and one ␣2(I) polypeptide chains. Although located on different chromosomes, both genes are generally coordinately regulated in a developmental and tissue-specific manner. Expression of the type I collagen genes is active in many cell types and under various physiological conditions, and their regulation is accordingly complex (5)(6)(7)(8). Type I collagen gene expression is also modulated by agents such as cytokines and chemical or viral transformation.
Transcriptional regulatory elements have been previously identified in the 5Ј-flanking region, the promoter region, and the first introns of both type I collagen genes in several species (9 -19). While sequences in the minimal promoter, within 220 bp 1 upstream of the start site of transcription, appear to be sufficient for the basal activity and partial tissue specificity of the ␣1(I) collagen gene promoter in transient transfection assays, the precise function of distal 5Ј-flanking sequences and the first intron is less clear. Moreover, several lines of evidence suggest that, in addition to these regulatory elements, sequences in the 3Ј-flanking region may contribute to transcriptional regulation of the ␣1(I) collagen gene. First, the 5Ј-regulatory elements were not always sufficient for precisely regulated, tissue-specific, high level expression of the gene when tested in transient transfection experiments or in transgenic mice (10, 20 -23). Second, in the transgenic HucII mouse strain, a single copy of the human ␣1(I) collagen gene, which included 1.6 kb of 5Ј-flanking region, the entire structural gene, and 20 kb of 3Ј-flanking sequence, was expressed as efficiently as the endogenous collagen gene in an appropriate tissuespecific manner (24) and was induced appropriately during hepatic fibrogenesis (7). Finally, the human ␣1(I) collagen gene contains several DNase I-hypersensitive sites located immediately 3Ј of the structural gene (25), which are often indicative of regulatory elements. Therefore, we initiated a systematic analysis for regulatory elements located within the 3Ј-flanking region of the murine ␣1(I) collagen gene. We located a segment of the 3Ј-flanking region which was found to enhance expression of the heterologous ␤-globin gene promoter as well as the endogenous ␣1(I) collagen gene minimal promoter in NIH 3T3 fibroblast cells. DNase I footprinting analysis demonstrated the location of three sites of DNA-protein interactions within this transcriptional stimulatory region. One of the binding sites contained a consensus E-box, to which both USF-1 and USF-2, two basic * This study was supported in part by a University Research Council grant from the University of North Carolina at Chapel Hill (to R. A. R.) and National Institutes of Health Grants DK34987 (to R. A. R. and D. A. B.), GM41084, CA50528 (to D. A. B.), and AR41909 (to M. B.). 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  helix-loop-helix proteins (bHLH), were found to bind. When the E-box was mutated within the context of surrounding wild-type 3Ј-flanking sequence, reduced levels of reporter gene activity were observed. However, when all three cis-acting elements were mutated a complete loss of transcriptional stimulatory activity was obtained. These results demonstrate that transcription of the ␣1(I) collagen gene is stimulated by a region within the 3Ј-flanking region of the gene and that USF-1 and USF-2 participate in the stimulation of transcriptional activity of the ␣1(I) collagen gene promoter.

MATERIALS AND METHODS
Plasmid Construction and DNA Sequencing-Fragments of the genomic clones pCE4 and pCE5 (kindly provided by K. Harbers) which contain the 3Ј-flanking sequences of the murine ␣1(I) collagen gene (26) (Fig. 1) were cloned into the unique SphI site of the enhancer trap plasmid p␤e Ϫ (27). The entire ␣1(I) collagen sequence present in pCE5 was sequenced. This fragment contains the last 12 codons of the ␣1(I) collagen gene. As a reference, the first base following the translational stop codon is designated as ϩ1. The complete genomic insert in pCE5 (4.6 kb) was cloned into p␤e Ϫ , creating plasmid p␤CE5, and a series of deletions shown in Fig. 1B were created by digestion with Bal31 exonuclease utilizing unique PstI and HindIII sites in the plasmid. The plasmid p␤COL-(ϩ1899 -4597) was created by digesting p␤CE5 with PstI, gel-purifying the large DNA fragment, and religating using T4 DNA ligase. A partial HincII digestion of the pCE4 genomic insert produced two fragments (5.4 and 2.4 kb) which were individually cloned into the SphI site of p␤e Ϫ . The reporter gene pGLCOL3 was constructed by ligating the ␣1(I) collagen promoter (Ϫ220 to ϩ116) into the BglII and HindIII sites of the luciferase reporter gene pGL2-Basic (Promega, Madison, WI). Reporter genes pCOL-(ϩ1899 -4597) and pCOL(ϩ4597 to 1899) were constructed by ligating the 2698-bp 3Ј ␣1(I) collagen fragment from plasmid p␤COL-(ϩ1899 -4597) ( Fig. 1) into the blunted SalI site of pGLCOL3, which is located at the 3Ј end of the luciferase gene, in both orientations. Plasmid pCOL-(ϩ1899 -4597) contains the 3Ј-flanking region in the 5Ј to 3Ј orientation, with respect to the direction of transcription from the collagen promoter, while in plasmid pCOL-(ϩ4597 to 1899) the 3Ј-flanking sequence is positioned 3Ј to 5Ј, with respect to the direction of transcription from the collagen promoter. Plasmids pCOL-(ϩ3590 -4597) and pCOL-(ϩ4597 to 3590) were created by digesting p␤COL-(ϩ1899 -4597) with HindIII and PstI, blunting the ends of the isolated 1-kb fragment, and ligating the fragment into pGLCOL3 which was digested with BamHI and the ends blunted. This places the ␣1(I) collagen 3Ј-flanking sequence 3Ј with respect to the luciferase gene. Plasmids pCOL-(ϩ4090 -4597) and pCOL-(ϩ4597 to 4090) were constructed by generating a PCR product using primers that hybridized to positions ϩ4090 to ϩ4113 (5Ј-ATC CGG ATC CGT AAC CTA AAG ATG GTG GGT TTT C-3Ј) and positions ϩ4597 to ϩ4577 (5Ј-ATC CGG ATC CGA ATT CCC ACT AGT GCG GGG G-3Ј) of the ␣1(I) collagen 3Ј-flanking sequence. The PCR reaction contained 10 mM Tris-HCl, pH 8.9, 50 mM KCl, 2.5 mM MgCl 2 , 2 M dNTPs, 1 M each primer, 100 ng of pCE5 as the template, and 2.5 units Taq DNA polymerase (Boehringer Mannheim). The reaction was cycled as follows: 94°C for 1 min, 60°C for 1 min, 72°C for 1 min for 30 cycles, an extension incubation at 72°C for 10 min, followed by incubation at 4°C, using a GeneAmp PCR System 9600 (Perkin-Elmer). The PCR fragment was digested with BamHI, using the BamHI recognition site designed in the primers, and cloned into the BamHI site of pGLCOL3 in both orientations. To construct plasmids containing the 3Ј-footprinted regions, oligonucleotides were synthesized and complementary strands annealed. The annealed oligonucleotides were ligated together and the ligation reaction subsequently electrophoresed in polyacrylamide gels. Bands representing three and five copies of the ligated oligonucleotides were excised and eluted from the gel. The oligonucleotides were then ligated into the BamHI site of pGLCOL3. The oligonucleotides used for the three footprinted regions were as follows, al. (28) was used. Briefly, two oligonucleotides were synthesized which contained the 6-bp mutation (underlined) (primer 1: 5Ј-CAC CCC GCA GCG GCT GTT GTA CAG CAT GGG CTG GTA TGT GCT-3Ј; primer 2: 5Ј-AGC ACA TAC CAG CCC ATG CTG TAC AAC AGC CGC TGC GGG GTG-3Ј). Two additional primers were synthesized which flanked the 5Ј end (primer 3: 5Ј-ATC CGG ATC CTG CGC CTG AAA ATC TAT ACA TAT AC-3Ј) and 3Ј end (primer 4: 5Ј-ATC CGG ATC CGA ATT CCC ACT AGT GCG GGG G-3Ј) of the ␣1(I) collagen 3Ј-flanking sequence of the pCOL-(ϩ3590 -4597) insert. Two PCR reactions were performed using the conditions as described above, one using primers 2 with 3 and the second PCR reaction with primers 1 and 4. An aliquot of each PCR reaction was combined in a second round PCR reaction along with primers 3 and 4 using the same PCR conditions described above. The product of this second PCR reaction was then digested with BamHI and subsequently cloned into the BamHI site of pGLCOL3.
To mutate all three of the footprinted regions in the 3Ј-flanking region while in the context of plasmid pCOL-(ϩ3590 -4597), a similar mutagenesis approach, as described above, was utilized. The plasmid p3ЈFP1M was used as a template in PCR reactions using primers designed to mutate 3ЈFP2 (primer 1: 5Ј-TGA ACC CAA GCC CTC CTT TCT ATC AGT GCT GCC GCT GCC TTA AAT ACA GAT GCC-3Ј; primer 2: 5Ј-GGC ATC TGT ATT TAA GGC AGC GGC AGC ACT GAT AGA AAG GAG GGC TTG GGT TCA-3Ј; the mutated nucleotides, ϩ3797-3809, are underlined) along with the 5Ј-flanking primer (primer 3 above) and 3Ј-flanking primer (primer 4 above). After the second PCR reaction the 3ЈFP1 and 3ЈFP2 mutated fragment was used as a template in PCR reactions to generate mutations in 3ЈFP3 as described above using 3ЈFP3 M primers (primer 1: 5Ј-TTG GAA TCC AAG TCC CTT TGG AAG AAA GTT CGG AGC ATG GTC ACT CCT GG-3Ј; primer 2: 5Ј-CCA GGA GTG ACC ATG CTC CGA ACT TTC TTC CAA AGG GAC TTG GAT TCC AA-3Ј; the mutated nucleotides, ϩ4050 -4059, are underlined) along with the 5Ј-flanking primer (primer 3 above) and 3Јflanking primer (primer 4 above). After the second PCR reaction the PCR product was digested with BamHI, gel-purified, and cloned into pGLCOL3, digested with BamHI, creating p3ЈFP1-2-3M.
DNA sequencing was performed by the dideoxy method using the Sequenase version 2.0 kit (Boehinger Mannheim) according to the manufacturer's recommended protocol to confirm the presence of the mutations in the respective footprinted regions.
Transfections and Reporter Gene Assays-NIH 3T3 fibroblast cells and HeLa cells were cultured in 150-mm plates with Dulbecco's minimum essential medium (Life Technologies, Inc.) supplemented with 10% calf serum and grown in a 5% CO 2 , 95% air atmosphere at 37°C. Transfections using the ␤-globin reporter gene were performed using the calcium phosphate precipitation method as described (13,19). In addition to 20 g of the p␤e Ϫ based reporter gene constructs, each transfection contained 20 g of the p␣ reporter gene (27) which served as an internal control for transfection efficiencies. Cells were treated with 75 M chloroquine during the transfection, and after 4 h the transfection mixture was removed, and the cells were shocked for 1 min using 10% glycerol. The cells were harvested 36 -48 h after transfection, and total RNA was prepared by the acid-phenol method (29). Radiolabeled antisense ␤-globin and ␣-globin RNA probes were generated by SP6 and T7 RNA polymerase, respectively. T7 transcription of p␣ creates a 244-nucleotide transcript of which 131 nucleotides are protected (30). SP6 transcription of p␤e Ϫ creates a 500-nucleotide transcript of which 350 nucleotides are protected (27). Total RNA samples were analyzed by RNase protection assay as described (31), and ␣and ␤-globin transcripts were quantitated by scanning with an image analyzer or direct counting of the bands. Transient transfections using the luciferase reporter gene plasmids were performed using Lipo-fectAMINE (Life Technologies, Inc.). NIH 3T3 cells were seeded into 6-well dishes at a density of 9 ϫ 10 4 cells per well. The day after seeding 0.5 g of luciferase reporter plasmid, 0.5 g of pRSV-␤gal and 1.1 g of carrier DNA, pUC19, was added to the cells using 11 g of Lipo-fectAMINE reagent per well following the recommended protocol of the manufacturer. Liposomes were incubated with the cells for 8 h. The RSV-␤gal reporter gene plasmid was co-transfected to normalize for transfection efficiencies. Luciferase and ␤-galactosidase reporter gene assays were performed as described previously 36 -48 h after transfection (13,19).
DNase I Footprinting and Mobility Shift Assays-DNase I footprinting analysis and mobility shift assays were performed as described previously (13). Nuclear proteins were obtained using the method described by Schreiber et al. (32). Supershift assays were performed, as described previously (13), using USF-1 and USF-2 polyclonal antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). The oligonucleotides used in the mobility shift assays representing 3ЈFP1 were top strand: 5Ј-GAT CCG CTG TCA CGT GGC ATG GGC TGA-3Ј; bottom strand: 5Ј-GAT CTC AGC CCA TGC CAC GTG ACA GCG-3Ј. The oligonucleotides that contained a mutation within the consensus E-box, 3ЈFP1M were top strand: 5Ј-GAT CCG CTG TTG TAC AGC ATG GGC TGA-3Ј;  bottom strand: 5Ј-GAT CTC AGC CCA TGC TGT ACA ACA GCG-3Ј (the  mutated nucleotides are underlined).
Oligonucleotides-Oligonucleotides were synthesized using a Cyclone Plus Oligonucleotide Synthesizer (Milligen, Novato, CA) and were gel-purified after electrophoresis in 10% polyacrylamide gels.
DNA Sequence-The DNA sequence described in this paper has been deposited into the GenBank data base.

The 3Ј-Flanking Region of the ␣1(I) Collagen Gene Contains a Region That Stimulates Transcription in an Enhancer Trap
Plasmid-To investigate the presence of transcriptional regulatory elements located in the 3Ј-flanking region of the ␣1(I) collagen gene, we cloned several fragments representing 12.4 kb of this region into the enhancer trap plasmid, p␤e Ϫ (27). This plasmid contains the human ␤-globin promoter and structural gene with a unique SphI cloning site located 2.2 kb downstream (or 3.3 kb upstream) of the ␤-globin transcriptional start site. The recombinant plasmids were transiently transfected into NIH 3T3 cells along with a plasmid containing the human ␣-globin gene, to serve as an internal control for transfection efficiency. The relative levels of ␤and ␣-globin mRNA were assessed by RNase protection assays using total RNA from transfected cells. Insertion of the entire 4.6 kb genomic fragment of pCE5 into the enhancer trap plasmid stimulated the expression of the ␤-globin gene 1.7-fold compared with expression of p␤e Ϫ alone, whereas ␣1(I) collagen gene fragments derived from the pCE4 plasmid did not exhibit any stimulatory effects on ␤-globin expression ( Fig. 1A and summarized in Fig. 1B). Therefore, we concentrated our efforts on ␣1(I) collagen 3Ј-flanking sequences in pCE5 to further localize the position of the enhancing activity using deletional analysis. Fig. 1B summarizes the results of two to six transfection experiments carried out with independently purified plasmid preparations. The smallest construct containing enhancing activity was p␤COL-(ϩ2643-4597) which reproducibly stimulated ␤-globin expression 2.8-fold as compared to p␤e Ϫ . This level of stimulation by the ␣1(I) collagen 3Ј fragment is comparable with the stimulation of p␤e Ϫ by the strong SV40 early gene enhancer in NIH 3T3 cells (27). To eliminate the possibility that the expression of the ␣-globin gene interferes with or affects the ␤-globin expression, the ␤-globin constructs were also transfected alone in a separate experiment with identical results (data not shown).
To determine whether this enhancing activity is cell typespecific, transient transfections were performed using the same reporter gene constructs in HeLa cells, which express no or low levels of type I collagen (33). None of the ␣1(I) collagen 3Јflanking genomic sequences stimulated ␤-globin gene expression in HeLa cells (Fig. 1C). In fact, the ␣1(I) collagen derived fragments in plasmids p␤COL-(ϩ1899 -4597) and p␤COL-(ϩ2643-4597) had a slight inhibitory effect on ␤-globin gene expression relative to the control p␤e Ϫ plasmid in these cells.
The ␣1(I) Collagen Gene 3Ј-Flanking Region Stimulates Expression of the ␣1(I) Collagen Gene Promoter-Since a segment of the ␣1(I) collagen gene 3Ј-flanking region stimulated expression from the heterologous ␤-globin promoter, we wanted to determine if this region could also stimulate expression from the homologous ␣1(I) collagen gene promoter. Therefore, we cloned the ␣1(I) collagen genomic insert from the plasmid p␤COL-(ϩ1899 -4597) in both orientations, behind the luciferase reporter gene in pGLCOL3, in which the luciferase gene is driven by the ␣1(I) collagen minimal promoter (Ϫ220 to ϩ116), creating the plasmids pCOL-(ϩ1899 -4597) and pCOL-(ϩ4597 to 1899) ( Fig. 2A). When these plasmids were transiently transfected into NIH 3T3 cells, and transfection efficiencies were normalized by co-transfection with pRSV-␤gal, a significant increase in transcriptional activity of the ␣1(I) collagen promoter was observed when the 3Ј-flanking sequence was in the 5Ј to 3Ј orientation. However, when positioned in the opposite orientation (3Ј to 5Ј) an inhibitory effect was observed (Fig. 2B) indicating the presence of an inhibitory element. To further localize the stimulatory element within the ␣1(I) collagen 3Јflanking region, the shorter collagen sequence in p␤COL-(ϩ3590 -4597) was inserted behind the luciferase reporter gene in the pGLCOL3 plasmid, in both orientations, creating the plasmids pCOL-(ϩ3590 -4597) and pCOL-(ϩ4597 to 3590) ( Fig. 2A). Transient transfections with these plasmid constructs into NIH 3T3 cells also demonstrated a stimulatory effect of this region on the ␣1(I) collagen gene promoter when FIG. 1. The 3-flanking region of the ␣1(I) collagen gene contains a region that enhances transcription from a heterologous promoter. A, NIH 3T3 cells were transiently transfected by the calcium-phosphate method and expression levels transfected the plasmids determined by RNase protection assays. The plasmid, p␣, was used as an internal reference to normalize for transfection efficiency. Several independent plasmid preparations gave the same results of ␤-globin transcriptional activation. B, summary of expression of p␤e Ϫ -derived plasmids in NIH 3T3 cells. The first line shows the 3Ј region of ␣1(I) collagen with stop codon, polyadenylation sites, the last exon (stippled box), and selected restriction sites. The 4.6-and 7.8-kb fragments contained in pCE5 and pCE4 plasmids, respectively, are shown in the second line. The fragments used in transfection assays and the levels of ␤-globin RNA expressed relative to the level produced by the enhancerless p␤e Ϫ plasmid and corrected for transfection efficiency by normalizing for ␣-globin mRNA levels are shown. C, HeLa cells were transiently transfected by the calcium-phosphate method with the indicated plasmids, and RNase protection assays were performed as with NIH 3T3 cells. positioned in either orientation (Fig. 2B). However, further deleting 3Ј-flanking sequence to nucleotides ϩ4090 to ϩ4597 (plasmids pCOL-(ϩ4090 -4597) and pCOL-(ϩ4597 to 4090)) eliminated the stimulatory effect. Together, these data indicate that a transcriptional stimulatory element located in the ␣1(I) collagen gene 3Ј-flanking region is positioned between nucleotides ϩ3590 to ϩ4089.
DNA Binding Proteins Specifically Interact at Three Locations within the ␣1(I) Collagen Gene 3Ј-Flanking Region-To determine the locations of DNA binding proteins in this stimulatory element in the 3Ј-flanking region of the ␣1(I) collagen gene, we performed DNase I footprinting analysis using NIH 3T3 nuclear extracts. Three sites of specific DNA-protein interactions were located and called 3ЈFP1, 3ЈFP2, and 3ЈFP3 (Fig.  3A, left, middle, and right, respectively; and summarized in Fig. 3B). To determine the effect that each of the footprinted regions exerts on expression of the ␣1(I) collagen promoter region, we cloned one copy, in both orientations, as well as three and five copies of the oligonucleotides representing each footprint into the BamHI site of pGLCOL3. We investigated single and multiple copies of the binding sites in order to optimize their effects on transcriptional activation. The plasmid constructs were transiently transfected into NIH 3T3 cells and transfections normalized to the activity of the co-transfected pRSV-␤gal plasmid. Although no stimulatory effect on transcriptional activity of the ␣1(I) collagen gene minimal promoter was observed when one copy of 3ЈFP1 was present in either orientation, the 3ЈFP1 sequence stimulated expression nearly 4-fold when three copies were present and approximately 6-fold when five copies were present (Fig. 4). On the other hand, 3ЈFP2 slightly stimulated expression of the ␣1(I) collagen gene promoter with only one copy present but not with five copies (Fig. 4). 3ЈFP3 slightly stimulated expression of the parental pGLCOL3 plasmid when either a single copy or with three copies of the footprinted regions were positioned 3Ј of the luciferase reporter gene (Fig. 4).

USF-1 and USF-2 Interact at a Consensus E-box Binding Site Located in the ␣1(I) Collagen 3Ј-Flanking Region-Since
3ЈFP1 stimulated expression of the ␣1(I) collagen gene minimal promoter, we wished to identify the protein(s) that interacts with this cis-acting element. Mobility shift assays were performed with a radiolabeled 3ЈFP1 oligonucleotide using NIH 3T3 nuclear extracts. A DNA-protein complex was formed in the mobility shift assay (Fig. 5, lane 2) which was specifically competed with cold unlabeled 3ЈFP1 oligonucleotide (Fig. 5,  lane 3). Examination of the nucleotide sequence of 3ЈFP1 demonstrated a consensus E-box (CANNTG), a known binding site for bHLH proteins (34), located at nucleotide ϩ3695 to ϩ3700 in the ␣1(I) collagen 3Ј-flanking sequence. To determine if the consensus E-box is involved in protein binding to 3ЈFP1, we used a mutated 3ЈFP1 oligonucleotide in which the six nucleotides representing the consensus E-box were mutated in the context of surrounding wild-type sequence. This mutated oligonucleotide failed to compete for protein binding (Fig. 5, lane  4) demonstrating that the E-box is required for protein binding to the 3ЈFP1 footprint sequence. The nucleotide sequence of this E-box matches the USF-1/USF-2 consensus binding site (CACGTG) (35). Therefore, we used antibodies directed against USF-1 or USF-2 in the mobility shift assay to test if these two bHLH proteins interact with 3ЈFP1. Addition of either the USF-1 or the USF-2 antibody in the binding reaction supershifted the DNA-protein complex (Fig. 6, lanes 4 and 5), with the USF-1 antibody supershifting the complex to a greater degree than the USF-2 antibody. When both antibodies were included in the binding reaction (Fig. 6, lane 6), the entire complex was eliminated indicating that both USF-1 and USF-2 interact with 3ЈFP1 as either homodimers or heterodimers. Additionally, binding to 3ЈFP1 requires Mg 2ϩ for binding with maximal binding activity requiring at least 5 mM MgCl 2 (data not shown). This is in agreement with a previous report demonstrating that decreased Mg 2ϩ concentration reduces binding affinity of USF-1 to DNA (36). Binding activity of USF-1/USF-2 to 3ЈFP1 was not salt-dependent as binding occurs over a wide range of salt concentrations (26 to 200 mM NaCl) in the binding reaction (data not shown).
USF Participates in the Activation of the ␣1(I) Collagen Gene Promoter-To determine the functional role of USF interaction with this 3Ј E-box in the stimulatory effect of the 3Ј-flanking region on the ␣1(I) collagen gene promoter, we generated a mutated reporter gene construct in which only the CACGTG (nucleotides ϩ3695 to ϩ3700) E-box site was mutated in pCOL-(ϩ3590 -4597) (see Fig. 2A), creating p3ЈFP1M. When p3ЈFP1M was transiently transfected into NIH 3T3 cells and transfection efficiencies normalized to the co-transfected pRSV-␤gal plasmid, a significant decrease in reporter gene activity was observed compared with the wild-type parental pCOL-(ϩ3590 -4597) plasmid (Fig. 7). This indicates that at least part of the stimulatory properties of this region on the ␣1(I) collagen gene promoter are due to USF-1/USF-2 binding to the E-box. To assess if the other footprinted regions, 3ЈFP2 and 3ЈFP3, participate in the stimulatory properties in the pCOL-(ϩ3590 -4597) plasmid, we created plasmid p3ЈFP1-2-3M which contained mutations in all three footprinted regions (nucleotides ϩ3695-3700; nucleotides ϩ3797-3809; nucleotides ϩ4050 -4059). When this plasmid was transiently transfected into NIH 3T3 cells, a complete inhibition of tran-scriptional activation was observed and, in fact, resulted in a reduction of expression compared with the parental pGLCOL3 plasmid (Fig. 7). This result supports data presented in Fig. 2B demonstrating an inhibitory element is present as shown in plasmid pCOL-(ϩ4597 to 1899). Taken together, these data indicate that USF-1/USF-2 is involved in the stimulatory properties of the 3Ј-flanking region on the ␣1(I) collagen gene promoter; however, 3ЈFP2 and 3ЈFP3 also cooperate in stimulating transcription of the ␣1(I) collagen gene promoter.

DISCUSSION
The genes encoding the ␣1 and ␣2 polypeptide chains of type I collagen are regulated in a developmental, inducible, and tissue-specific manner. In previous studies of the human, murine, and rat ␣1(I) collagen genes using either transient transfections into cultured cells or transgenic animals, cis-regulatory elements have been identified in the 5Ј-flanking regions, the promoter region, and first introns of these genes (9 -19). The precise functions of many of these cis-acting elements remain to be elucidated, and some conflicting results have been reported, perhaps reflecting the use of different reporter gene constructs and different cell types. Most studies support the notion that the 5Ј region of the ␣1(I) collagen gene contains a strong proximal promoter which exhibits at least partial tissue specificity and which is modulated by additional regulatory elements.
Several lines of evidence suggested that the 3Ј-flanking region of the ␣1(I) collagen gene contains transcriptional regulatory elements. Therefore, we initiated a systematic analysis of the 3Ј-flanking region to locate important regulatory elements. Our analysis of the 3Ј-flanking region of the ␣1(I) collagen gene has located a fragment (between nucleotides ϩ1899 to ϩ4597 Each of the 3Ј-footprinted regions were cloned behind the luciferase gene in the pGLCOL3 plasmid with either one copy, in both orientations, three, or five copies. Each construct was transiently transfected into NIH 3T3 cells and luciferase activity measured and normalized to the activity of the co-transfected pRSV-␤gal plasmid. Data represent the average of four transfections. Those plasmids in which the activity was significantly different compared to pGLCOL3, using the Wilcoxan signed ranks statistical analysis, are designated by an asterisk; p Ͻ 0.05. Error bars represent S.E. downstream of the translational termination site of the ␣1(I) collagen gene) that enhances transcription of the heterologous ␤-globin gene promoter in NIH 3T3 cells, a cell line that expresses moderate levels of type I collagen. Although the level of enhancement is not dramatic, it is comparable to the enhancement of the same reporter gene construct by the SV40 enhancer in the same cells (27). This fragment also stimulates transcription driven by the ␣1(I) collagen promoter in an orientation-dependent manner. Deletional analysis of the 3Ј-stimulatory region narrowed the enhancing region to within a 500-bp segment. DNase I footprinting assays demonstrated the presence of three regions that specifically interact with DNA binding proteins. Only one of the footprinted regions was capable of stimulating expression of the ␣1(I) collagen promoter when multiple copies were inserted 3Ј of the luciferase reporter gene. Further analysis of this protein binding site demonstrated that both USF-1 and USF-2 interact with the nucleotide sequence of a consensus E-box, the binding site for basic helix-loop-helix (bHLH) proteins. A functional role of this site in stimulating expression of the ␣1(I) collagen gene promoter was assessed by mutating only the nucleotides comprising the consensus E-box while in the context of the wild-type pCOL-(ϩ3590 -4597) collagen containing sequence. The mutated collagen-luciferase reporter gene construct containing the mutated E-box had decreased transcriptional activity compared with the wild-type E-box. However, the transcriptional stimulatory properties of the 3Ј-flanking region was completely lost only when all three protein binding sites were mutated. This was surprising since 3ЈFP2 and 3ЈFP3 alone did not stimulate transcription (Fig. 4). These data suggest a complex interaction between the transacting factors interacting with the 3Ј-flanking region. In addition, when all three cis-acting elements were mutated, pro-moter activity was reduced compared with to the level of the ␣1(I) promoter-driven plasmid pGLCOL3. This indicates that a negative regulatory element is present between nucleotides ϩ3590 and 4597. This finding is supported in the construct pCOL-(ϩ4597 to 1899) where an inhibitory effect was observed on transcriptional activity (see Fig. 2).
The USF-1 binding site in the endogenous ␣1(I) collagen gene is located over 20 kb downstream from the promoter. Regulatory elements have also been demonstrated in the 3Јflanking region of several other genes, including the mouse cytosolic glutathione peroxidase gene (37), the human keratin 19 gene (38), the murine c-fos protooncogene (39), the human angiotensinogen gene (40), and the human tyrosine hydroxylase gene (41). One could speculate that looping of the chromatin would place the two regions in close enough proximity to one another for appropriate interactions to occur.
Both USF-1 and USF-2 are ubiquitously expressed, although their relative abundance varies in different cell types (42). In fact, USF was originally identified and characterized in HeLa cells (43)(44)(45). Thus, it was surprising that the 3Ј region of the ␣1(I) collagen gene enhanced transcription in NIH 3T3 cells but not in HeLa cells. Perhaps differences in the USF proteins, transcriptional co-factors, or other DNA binding proteins allow for this difference in transcriptional activity between the two cell types. The expression of both USF-1 and USF-2 genes results in multiply spliced messages producing variations in the amino terminus of the protein (42). This, therefore, may give rise to proteins with different transcriptional stimulating FIG. 5. Protein binding to 3FP1 requires an intact consensus E-box. The mobility shift assay was performed using a radiolabeled probe representing 3ЈFP1 and 10 g of NIH 3T3 nuclear extract. A specific DNA-protein complex is formed with the wild-type (wt) 3ЈFP1 oligonucleotide used in the mobility shift assay (lane 2), which is efficiently competed with wild-type unlabeled oligonucleotide (lane 3). A mutant (mut) 3ЈFP1 oligonucleotide, in which the consensus E-box is mutated, fails to compete for protein binding (lane 4). Lane 1 is the probe incubated without nuclear extract. A 100-fold molar excess of unlabeled oligonucleotides was used in the competition reactions.
FIG. 6. Both USF-1 and USF-2 bind to 3FP1. A mobility shift assay was performed using 3ЈFP1 as a radiolabeled probe and incubated with 10 g of NIH 3T3 nuclear extract (lanes 2-6). Lane 1 is the probe incubated without nuclear extract. A 100-fold molar excess of unlabeled 3ЈFP1 oligonucleotide was used as a competitor in the binding reaction (lane 3). Supershift studies were performed using either a polyclonal USF-1 antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in the binding reaction (lanes 4 and 6) or a polyclonal USF-2 antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in the binding reaction (lanes 5 and 6).
activities since the amino-terminal portion of the proteins contain the trans-activating domain. Since the USF family members homodimerize and heterodimerize with each other, but do not heterodimerize with other bHLH proteins (46,47), various ratios of multiple USF-1 and USF-2 proteins will generate complex sets of USF members with potentially different transcriptional activating capacities. Additionally, transcriptional co-factors may be absent in the HeLa cells that are necessary for USF to stimulate the ␣1(I) collagen gene promoter. USF does not appear to trans-activate effectively in highly purified in vitro systems (48) suggesting a role for additional interacting proteins. It is believed that USF-1 affects transcription by interacting with the multisubunit TFIID complex (43,49) and specifically with the TFIID subunit TAF II 55 (50). Additional factor(s) also appear to be required for USF to stimulate transcription, one of which appears to be PC5, a novel co-factor which has been demonstrated to mediate transcriptional activation by USF-1 (51).
USF contributes to the regulation of multiple genes, some of which are expressed in a tissue-specific or inducible manner. The human ␤-globin locus control region (52), the murine type II␤ regulatory subunit gene of cyclic adenosine 3Ј,5Ј-monophosphate-dependent protein kinase gene (53), the murine p53 gene (54), the human cell cycle-dependent cyclin B1 gene (55), the chicken ␣A-crystallin gene (56), the rat cardiac ventricular myosin light chain 2 gene (57), and the rat preprotachykinin-A gene (58) are regulated in part by USF. USF has also been shown to be involved in the TGF-␤1 responsive element in the human plasminogen activator inhibitor gene (59) and in the glucose responsive element in the pyruvate kinase gene (60). Our study demonstrates that the ubiquitous USFs contribute to the moderate high transcriptional rate of the ␣1(I) collagen gene in NIH 3T3 cells.
The present study, in conjunction with previous studies, demonstrates that the ␣1(I) collagen gene is regulated by a complex array of cis-acting elements in the 5Ј-flanking region, the promoter, the first intron, and 3Ј-flanking region. Like USF-1 and USF-2, which interact with the 3Ј-regulatory element identified in this study, most of the other factors involved in transcriptional regulation of the ␣1(I) collagen gene are ubiquitous factors. For example, the proximal promoter has been shown to interact with SP1, NF-I, CBF, and cKrox (13,61,62), none of which is present exclusively in collagen-producing cells. Several elements, however, have been shown to contribute to cell-specific expression of the ␣1(I) collagen promoter. These include a TGF-␤ response element located approximately 1.6 kb upstream of the transcriptional start site (63), a sequence required for expression in bone (64) and elements that modulate expression in dermal fibroblasts, osteoblasts, and odontoblasts, and in tendon and fascia fibroblasts (65). However, even without these elements the ␣1(I) collagen promoter exhibits a remarkable degree of tissue specificity (10,65,66). Taken together, these observations suggest that the various regulatory elements located in the 5Ј-flanking region, the first intron, the 3Ј-flanking region, and possibly additional sites so far unidentified act in concert to provide appropriate tissue specificity and high levels of activity of the ␣1(I) collagen gene promoter and that the correct spatial arrangement of the various elements is required for appropriate activity.