Identification of an upstream regulatory region essential for cell type-specific transcription of the pro-alpha 2(V) collagen gene (COL5A2).

The transcriptional features of the human alpha 2(V) collagen gene (COL5A2) were examined by transfection experiments coupled to various DNA binding assays. This approach identified an upstream region essential for the cell type-specific expression of the COL5A2 promoter. Within this region are two nuclear factor-binding sites, FP-A and FP-B, responsible for the formation of distinct DNA-protein complexes. Mutations introduced across each of the two binding sites eliminated the formation of the cognate complex and decreased promoter activity by about 3-fold (FP-A) and 40-fold (FP-B) in transfection experiments. Competition experiments using recognition sequences for known transcription factors exhibiting some similarity to the FP-A- and FP-B-binding sites failed to inhibit COL5A2/protein interactions. Thus, COL5A2 expression appears to be under the positive control of a short regulatory sequence likely to harbor two novel nuclear factor-binding sites.


Identification of an Upstream Regulatory Region Essential for Cell Type-specific Transcription of the pro-a2(V) Collagen Gene (COL5A2)"
(Received for publication, April 1, 1992)

From the Brookdale Center for Molecular Biology, Mt. Sinai School of Medicine, New York, New York 10029
The transcriptional features of the human a2(V) collagen gene (COL5A2) were examined by transfection experiments coupled to various DNA binding assays. This approach identified an upstream region essential for the cell type-specific expression of the COL5A2 promoter. Within this region are two nuclear factorbinding sites, FP-A and FP-B, responsible for the formation of distinct DNA-protein complexes.

Mutations introduced across each of the two binding sites eliminated the formation of the cognate complex and decreased promoter activity by about %fold (FP-A) and IO-fold (FP-B) in transfection experiments. Competition experiments using recognition sequences for known transcription factors exhibiting some similarity to the FP-A-and FP-B-binding sites failed to inhibit COLBAZ/protein interactions. Thus, COL5A2 expression appears to be under the positive control of a short regulatory sequence likely to harbor two novel nuclear factor-binding sites.
Fibril-forming collagens represent a structurally related group of molecules within the larger family of collagen proteins. The fibrillar class of collagens includes nine distinct polypeptides that associate into five different trimers (types I, 11, 111, V, and X I ) (1,2). Based on their relative levels of expression, these collagens are broadly divided into "major" and "minor" types (1, 2).
While the pathophysiology of the major fibrillar collagens (types I, 11, and 111) is somewhat understood, very little is known about the function and pathogenesis of the minor fibrillar collagens (types V and X I ) (1, 2). Some evidence suggests that types V and X I may play a role in fibrillogenesis by regulating the growth and orientation of type I and I1 collagen fibrils in noncartilaginous and cartilaginous tissues, respectively (3, 4). Accordingly, modulation of the relative proportions of major and minor collagen types may be responsible for different tissue architectures during development and in the adult organism. Conversely, altered production of minor fibrillar collagens may play some role in connective tissue pathology. For example, type V collagen production is elevated in inflammatory and fibrotic condi-* This work was supported by Grant AR-38648 from the National Institutes of Health. This is article number 98 from the Brookdale Center for Molecular Biology at the Mt. Sinai School of Medicine. The costs of Publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Characterization of the cis-acting DNA sequences and trans-acting nuclear factors that modulate correct spatial/ temporal patterns of collagen gene expression is an important step toward understanding connective tissue pathophysiology. Several investigations have recently indicated that proper expression of collagen genes is mediated by distinct arrangements of regulatory sequences often, but not always, residing both upstream and downstream of the transcription initiation site (for recent reviews, see Refs. 11 and 12). In our own work, we have previously analyzed some of the regulatory cis-acting elements present in the human type I collagen genes (13, 14).
Here we extend these studies to COL5A2,' the gene coding for the a 2 chain of human type V collagen.
The transcriptional features of the COL5A2 promoter were analyzed by transfecting various chimeric plasmids into normal and transformed cell lines and by performing a variety of DNA binding assays. As a result, we identified a 52-bp long regulatory region that is essential for the cell type-specific expression of COL5A2.

Cell Culture and Transfection Experiments-Human cells used in
these studies were the fibrosarcoma line HT-1080 (15), the rhabdomyosarcoma line A-204 (16), Jurkat cells (171, and primary fetal fibroblasts (CF-37) prepared as previously described (13). Cells were grown and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum.
Plasmid DNA was transfected into cells using the calcium phosphate procedure (18) or the electroporation technique (19). Conditions for the DNA preparation, cell transfection, and analysis of plasmid transcriptional activity have been previously detailed (13). Irrespective of the method employed and the cell line used, transfection efficiencies, calculated according to the activity of a standard amount of pSV2CAT (13), were consistently comparable.
Construction of Chimeric Plusmids-Isolation and characterization of the genomic clones used for the generation of the plasmids utilized in this study have been described elsewhere (43). Location of the major start site of transcription and numbering of the upstream sequence are as for Greenspan et al. (20).
Various portions of the COL5A2 promoter, sharing the same 3' end point, were inserted into the polylinker sequence of plasmid pBLCAT3 (21) upstream of the chloramphenicol acetyltransferase (CAT) gene. To generate the -2.3COL5A2/CAT construct, a 3.1-kb pUC18 (20). The DNA of this subclone was then amplified by the EcoRIISacI genomic fragment containing exon 1 was subcloned into polymerase chain reaction (PCR) technique (22) using an oligonucleotide primer corresponding to the 5' sequence of the genomic subclone The abbreviations used are: COLlA2, the gene coding for pro-a2(U collagen; COL2A5, the gene coding for pro-cu2(V) collagen; AP-1, activator protein-1; CAT, chloramphenicol acetyltransferase; CTF, CCAAT-binding transcription factor; NF-1, nuclear factor-1; PCR, polymerase chain reaction; kb, kilobase(s); bp, base pair(s).

25390
Regulation of pro-a2(V) Collagen Transcription and a primer located immediately upstream to the ATG codon (nucleotide position +160) (20). To maintain the correct orientation of the PCR product in the expression vector, the two primers contained HindIII and BamHI recognition sequences, respectively. The 5' deletions constructs -1.7COL5A2/CAT, -1.lCOL5AZ/CAT, and -0.66COL5A2/CAT were derived from circularization of the -2.3COL5A2/CAT DNA cleaved at unique restriction sites. Constructs with 5' deletions beginning at nucleotides -296 (-0.3COL5A2/CAT) and -99 (-O.lCOL5AS/CAT) were generated by PCR amplifications of the 3.1-kb EcoRIISacI genomic subclone using oligonucleotides specific for these upstream sequences and the aforementioned exon 1 primer. Internal deletions were obtained by linking 5' deletion constructs to appropriate genomic fragments derived by exonuclease III/mung bean nuclease treatment (23) of the -2.3COL5A2/CAT and -0.3 COL5AZ/CAT plasmids DNA. Generation of the -0.15COL5A2/CAT construct was achieved by bluntend subcloning of a PCR product (-150 to -100) into the -0.lCOL5A2/CAT plasmid. Relative orientation of the PCR insert was determined by DNA sequencing. Likewise, relative orientation of the -150 to -100 PCR product, subcloned into the unique SmaI site of the -772COLlA2 promoter (13), was ascertained by sequencing the relevant area of the resultant plasmid. Mutant forms of the sequences corresponding to the two upstream elements that are the subject of this study were generated by PCR amplification of cloned DNA using as primers mutated and wild-type oligonucleotides (22). The resultant molecules were purified and used for DNA binding assays, as well as for generating mutant plasmids for transfection experiments.
All PCR-derived constructs were fully sequenced and compared to the parental genomic subclones. Both DNA strands were sequenced by the dideoxynucleotide chain termination technique using universal primers or synthetic oligonucleotides (24). In some of the sequencing experiments, progressively overlapping deletions were generated with the exonuclease III/mung bean nuclease method (23). Conditions of PCR amplification were described previously (25).
Nuclear Extracts-Crude nuclear extracts were prepared at 4 "C according to the protocol of Morris et al. (26) with minor modifications. Cells were harvested, washed in phosphate-buffered saline, resuspended in 5-10 volumes of buffer I (0.32 M sucrose, 0.01 M Tris, pH 8.0, 5 mM MgC12, 1 mM EDTA, 1 mM spermidine, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride), and homogenized with 10-20 strokes in a Dounce homogenizer. Nuclei were centrifuged twice at 1,000 X g for 5 min in the same buffer, resuspended 10 volumes of buffer 11 (0.1 mM KCI, 20 mM HEPES, pH 7.9, 1 mM dithiothreitol, 0.1 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride), extracted with 0.4 M (NH4)2S04 for 30 min, and centrifuged at 35,000 rpm for 60 min in a 70 Ti rotor to pellet the chromatin. The supernatant was collected and the proteins were precipitated by the addition of (NH4),SO4 at a final concentration of 0.25 g/ml supernatant. Proteins were pelleted by centrifugation at 15,000 rpm for 15 min in the Sorvall HB-4 rotor. Pellets were resuspended in 1-2 ml of buffer I1 containing 20% glycerol and dialyzed against the same buffer until the conductivity of the extract reached 0.1-0.15 M KCI. Nuclear extract, cleared by 10 min of centrifugation in a microcentrifuge, was stored in aliquots at -80 "C. The protein concentration of the extract was 5-10 mg/ml as determined by the method of Bradford (27) using a commercial protein assay kit (Bio-Rad).
Probes for DNAIProtein Binding Assays-Fragments containing COL5A2 promoter sequences used as probes or competitors in DNA/ protein binding assays were derived from the -0.3COL5A2/CAT plasmid. To generate the 456-bp probe (-296 to +160), plasmid DNA was cleaved with HindIII, end-labeled with either Klenow DNA polymerase or T4 polynucleotide kinase, digested with BamHI, and purified by polyacrylamide gel electrophoresis (28). The same treatment was applied for preparing the 73-bp probe (-152 to -79) which was generated by digesting the -0.3COL5A2/CAT construct with MluI and FokI.
Oligonucleotides corresponding to the footprinted regions of the COL5A2 promoter extend from nucleotides -91 to -115 (FP-A) and from nucleotides -114 to -150 (FP-B) (17). Their DNA sequences along with those of other oligonucleotides used in the DNA binding experiments are listed in Table I. Oligonucleotides were synthesized with an Applied Biosystems model 380 synthesizer and purified in a denaturing polyacrylamide gel electrophoresis (28). One strand was end-labeled by using T4 polynucleotide kinase, after which it was annealed to the complementary strand and purified from singlestranded material by polyacrylamide gel electrophoresis in 0.5 x TBE (1 x TBE, 0.089 M Tris borate, 2 mM EDTA) (28). Computer search for sequence homology to known nuclear factor-binding sites was carried out using the program MacVector (International Biotechnologies Inc., New Haven, CT).
DNase Z Footprinting-DNase I footprinting was performed using a modification (29) of the technique described by Galas and Schmitz (30). About 10,000 cpm of end-labeled DNA were incubated at 4 "C for 30 min with various amounts of nuclear extract in the presence of 1 pg of poly(d1-dC) in a 25-pI reaction volume containing 0.01 M Tris, pH 7.5, 0.08 M NaCI, 4% glycerol, 0.01 M 8-mercaptoethanol, and 1 mM EDTA. After incubation, samples were brought to 5 mM MgCl, and 2.5 mM CaCI2 and digested with 0.1-1 unit of DNase I (Boehringer Mannheim) for 60 s at room temperature. Reactions were terminated by the addition of 3 pl of 500 mM EDTA, 1 p1 of 5 pg/pl tRNA, and 0.1 volume of 3 M sodium acetate, extracted with an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1) and ethanol-precipitated. The samples were resuspended in denaturing buffer (10 mM NaOH, 80% formamide, 0.1% xylene cyano], 0.1% bromphenol blue), heated at 90 "C for 2 min, and electrophoresed through 5 or 8% polyacrylamide-7 M urea sequencing gels (191 acrylamide/bisacrylamide ratio) in 1 X TBE buffer. Gels were exposed to x-ray film overnight at -80 "C with an intensifying screen.
Gel Mobility-Shift Assay-Binding reactions were performed by incubation of 2-5 pg of nuclear extract with end-labeled probes using the conditions described above. After incubation, samples were layered onto a polyacrylamide gel (301 acrylamide/bisacrylamide ratio) in 0.25 X TBE buffer, electrophoresed at 25 mA for 1-2 h at 4 "C, dried, and autoradiographed (29). Competition experiments were performed using 100-fold molar excess of competitor DNA molecules.
DNA Methylation Interference-The end-labeled double-stranded FP-A or FP-B oligonucleotides were treated with 0.5% dimethyl sulfate for 3 min as previously described (31). Methylated oligonucleotides were used in a standard binding reaction with nuclear extracts and electrophoresed as described above. Wet gels were exposed overnight at 4 "C with an intensifying screen. Protein complexes and the unbound DNA were excised from the gel and electroeluted into dialysis bags. Samples were purified by phenol/chloroform extraction, ethanol precipitation, and passage through a NACS ion exchange resin prepac column (BRL Life Technologies Inc.). G + A cleavage of the eluted methylated DNA and subsequent electrophoresis (15% polyacrylamide-7 M urea) followed standard protocols (28).

RESULTS
Cell Type-specific Expression of the COWA2 Promoter-In order to elucidate the transcriptional features of the COL5A2 promoter, a series of chimeric constructs containing different lengths of human sequence linked to the CAT gene were analyzed after transfection of cultured cells. In these initial functional assays, two different tumor lines were used as the recipient cells. The first is the human rhabdomyosarcoma line A-204 which produces significant amounts of type V collagen (16). The second is the human fibrosarcoma line HT-1080 which does not synthesize any of the known fibrillar collagen types (15).
The transcriptional activity of a COL5A2/CAT plasmid containing 2.3 kb of upstream sequence was first compared in the two tumor lines. High levels of CAT expression were obtained with A-204 cells, whereas HT-1080 extracts exhibited only basal CAT activity (Fig. M). This suggested that the -2.3COL5A2 promoter contains cis-acting DNA elements capable of sustaining high and cell type-specific gene expression.
Second, the effects of progressive 5' deletions on the transcription activity of the COL5A2 promoter were tested in A-204 and HT-1080 cells (Fig. M). Removal of nearly half of the COL5A2 promoter sequence resulted in a 2-fold increase in CAT expression. This enhancement may conceivably result from the elimination of negative cis-acting elementb) located between -2.3 and -1.0 kb. Likewise, the presence of both negative and positive cis-acting sequences may be inferred by comparing the transcriptional activity of plasmids -0.66COL5A2/CAT and -0.3COL5A2/CAT (Fig. M). The most dramatic effect on COLSA2-driven CAT expression was, however, observed with the removal of nucleotides -296 to -100, in that the resultant plasmid -0.lCOL5A2/CAT directs only background levels of CAT activity in A-204 cells (Fig.   1A).
Since 5' deletion of promoter sequence could result in readthrough transcription originating from the upstream vector region, CAT activity was also determined using -2.3COL5A21 CAT constructs harboring three different internal deletions between nucleotides -660 and -99 (Fig. 1R). These experiments confirmed the contribution of the -296 to -100 region to the transcriptional activity of the COL5A2 promoter (Fig.   1R). Finally, none of the deletion mutants displayed substantial changes in background CAT activity when assayed in the typeV collagen-nonproducingcell line HT-1080 (Fig. 1). From these functional assays we concluded that cis-acting elements capable of directing high and cell type-specific transcription from the COL5A2 promoter are located between nucleotides -296 and -100.
Identification of Nuclear Factor-binding Sites-To identify DNA/protein binding sites within the transcriptionally active -0.3COL5A2 promoter, the 456-bp region extending from nucleot,ide -296 to immediately 5' of the ATG codon (+160) ( Fig. 2A ) was subjected to DNase I footprinting analysis using A-204 nuclear extracts. This resulted in the protection of at least four distinct regions that extend from about nucleotides -290 to -90 (Fig. 3). In the present study, we focused on the characterization of the two most proximal footprints ( A and R in Fig. 3).
T o confirm the footprints and delineate more clearly their boundaries, a DNase I protection assay wras performed using a shorter probe that spans from nucleotides -152 to -79 (Fig.  2 A ) . Binding of A-204 nuclear proteins to this 73-bp long probe was challenged by molar excess of the same sequence and an unrelated segment of the COL5A2 promoter (Fig. 4). This documented the binding specificity of the footprints which were more accurately mapped from nucleotides -97 to

Regulation of pro-n2(V) Collagen Transcription
-113, FP-A, and from nucleotides -116 to -147, FP-R (Fig.  2R). These footprinted regions of the COL5A2 promoter were also protected by HT-1080 nuclear extracts, as well as by nuclear extracts from other cell lines that do not produce type V collagen (data not shown).
The footprinting data were then verified by the gel mobility shift assay using A-204 nuclear extracts incubated with the same 73-hp probe. This gave rise to two clear sets of retarded DNA/protein complexes, termed I and I1 (Fig. 5A). In addition, a less intense band migrating slower than complex I was observed occasionally (see lanes I of Figs. 5 and 8). The nature and specificity of this larger complex were not investigated further in the present study.
The specificity of complexes I and I1 was demonstrated by competing the interaction between the 73-bp probe and nuclear proteins with unlabeled homologous and unrelated sequences (Fig. 5A). Moreover, competition with FP-A or FP-B sequences correlated the slower and faster moving complexes with nuclear factors binding to FP-A and FP-B, respectively (Fig. SA). This conclusion was independently corroborated by an experiment in which the labeled FP-A and FP-B oligonucleotides were used separately as probes in the gel mobility shift assay (Fig. 5, C and D). In each of these two assays, specificity of binding was documented by competition with unlabeled FP-A and FP-R oligonucleotides and an unrelated sequence as well (Fig. 5, C and D).
Finally, a methylation interference experiment was performed for each of the two retarded complexes to determine which bases of the COL5A2 elements participate in binding t o nuclear proteins (Fig. 6). Points of contact were determined with clarity only for G residues (Fig. 6). In the FP-A element, four contacts were mapped between nucleotides -105 and -113, whereas two Gs complementary to the C residues of a  experiments, respectively. Sequence around the contact points (methylated G') are shown on the l~f t of the autoradiograms (see Fig. 4 ) .
duplicated ATCA motif were protected by the factor binding to FP-R (Fig. 2R). Structural Analysis of the FP-A and FP-R Elemmts-Inspection of the sequences around the protein contact points of the COL5A2 elements revealed some level of homology to the consensus recognition sequences of two distinct families of transcription factors. Retween nucleotides -105 and -109 of the FP-A noncoding strand is a DNA motif (GCCAAC) closely resembling a CTF/NF-1 binding site (GCCAAT) (32). while a potential AP-1 binding site (TGAC/GTCA) (32) is located between nucleotides -124 and -118 of the FP-R coding strand (TCAATCA) (Fig. 4).
Several oligonucleotides containing high affinity CTF/NF-1 binding sites (33) and the AP-1 recognition sequence of the human collagenase gene (34) were tested for the ability to compete for the formation of the two DNA/protein complexes (Table I). Since CCAAT proteins have been categorized on the basis of their affinity to different DNA binding sites (33). four distinct CTF/NF-1 oligomers were used in these competition experiments (Table I). Aside from the high affinity CP-1, CP-2, and NF-1 binding sites, a modified version of the adenovirus NF-1 sequence was also used. This sequence. previously shown to compete effectively NF-1 binding to the COLlA2 promoter (10,351, displays a DNA motif (GCCAAG) identical with that of FP-A (Table I). None of these potential competitor molecules affected binding of the 7.7-bp probe to A-204 nuclear proteins (Fig. 5R). I t should be noted that in this figure only the result of the modified adenovirus NF-1 oligonucleotide is shown as a representative example of all CTF/NF-1 competitions. In conclusion, these data strongly suggest that FP-A and FP-B may represent novel nuclear factor-binding sites.
On the basis of the methylation interference data, five nucleotide substitutions were introduced across the contact points of FP-A and FP-R (Fig. 2R). These mutant sequences were then assayed for their ahility to inhibit the interaction between A-204 nuclear extracts and wild-t-ype Z -h p probe. Neither of the two mutant sequences, added alone or in comhinat.ion, was capable of affecting binding of the factor to the cognate wild-t-ype sequence (Fig. 7). There was also no discernible difference in complex formation when a combi-Regulation of pro-n2(V) Collagen Transcription  nation of wild-type and mutant sequences was present in the same competitor molecule (Fig. 7). Thus, the structural integrity of the two COL5A2 sequences appears to be essential for in vitro binding to nuclear proteins.

Functional Analysis of the FP-A and FP-R Elements-In
the subsequent set of experiments, the functional features of the FP-A and FP-B elements were characterized. To this end, several chimeric constructs were engineered, and their CAT activity was ascertained after transfection of A-204 and HT-1080 cells. First, the transcriptionally inactive -0.lCOL5A2 promoter was extended by adding the 52 bp (from nucleotides -100 to -152) that encompass the FP-A and FP-B elements (Fig. 2, A and R ) . As a result, two -0.15COL5A2/CAT plasmids with the additional 52 hp oriented in either direction were generated (Fig. 8). Transfection experiments documented the orientation-dependent contribution of the 52-hp region to transcription of the COL5A2 promoter (Fig. 8). More importantly, these assays narrowed down the 5' boundary of the minimal COL5A2 promoter sequence to nucleotide -152.
In the third construct, the 52-bp region of COL5A2 was inserted upstream of the heterologous COLlA2 promoter (Fig.  8). This -0.7COLlAB/CAT construct has been shown previously to display cell-type specificity, for it is transcriptionally active in fibroblasts and inactive in lymphoblasts (13). This conclusion was now confirmed and extended, since this plasmid was also inactive in A-204 and HT-1080 cells, neither of which produces type I collagen (Fig. 8). In contrast, the COLlA2/COL5A2 mosaic promoter directs substantial CAT activity when transfected into A-204 but not into HT-1080 cells (Fig. 8). Thus, although nuclear proteins of both A-204 and HT-1080 nuclei hind in vitro to the FP-A and FP-R sequences, only those from the former are capable of forming transcriptionally active complexes. Lastly, the effects of the FP-A and FP-R mutations on the activity of the minimal promoter were assayed functionally by DNA transfection of A-204 cells. This revealed that -O.lFiCOLFiA2/CAT plasmids containing the mutated FP-A and FP-R elements display a 3-and 40-fold drop in CAT activity, respectively, compared to the wild-type minimal promoter construct (Fig. 9 ) . Collectively, these results demonstrate that optimal activity of the minimal COLFiA2 promoter transiently expressed in A-204 cells depends on the structural integrity of each of these two positively acting upstream elements.
COL5A2 Regulation in Nontransformpd Cells-Having established the transcriptional features of the minimal COL5A2 promoter in a tumor cell line, we re-examined them in primary fetal fihrohlasts (CF-37). To this end, CF-47 and .Jurkat cells were transfected with plasmids -0.3COL5A2/CAT, -0.15COL5A2/CAT, and -0.lCOL5AZ/CAT, as well as with a -0.3 COL5A2 promoter construct that harhors an internal deletion of the (FP-A, FP-R)-containing region (Fig. 10). Consistent with the endogenous levels of COL5A2 expression, the transcriptional activity of the chimeric constructs was significantly less in fihroblasts than rhahdomvosarcoma cells (Figs. 1, 8, and 10). There were also some differences in the relative levels of expression of individual constructs transfected into the two cell lines. T o he precise, -0.4COL5A2/ CAT and -0.ICOL5A2/CAT exhibited lower and higher transcriptional activities in fibroblasts compared to A-204 cells, respectively ( Figs. 1 and 10). Conceivably, these differences may reflect the intrinsic diversity of the remulatory programs that modulate COL5A2 expression in the two cell lines. Regardless of this point and consistent with the A-204 functional data, elimination of the 52-hp region suhstantially reduced transcription from the COL5A2 promoter in nontransformed fibrohlasts as well (Fig. 10).
Rased on these results, the (FP-A, FP-R)-containing prohe was subjected to the gel mobility shift assays after incubation with nuclear extracts from CF-47 and .Jurkat cells. This resulted in the formation of two complexes seemingly identical to those previously seen with A-204 nuclear proteins and, like them, specifically competed by FP-A and FP-R unlabeled oligonucleotides (Fig. 1 L4 ). A difference was, however, not,ed in the nature of the complexes formed between the FP-R element (complex 11) and factors present in nuclear extracts of Jurkat and CF-37 cells (compare lanm 1 and 2 wit.h 4 and ,nucleotides ( Fig. 11R). Altogether, these binding assays confirmed the data obtained with A-204 nuclear extracts and reiterated that the COL5A2 cis-acting elements hind in uitro t o proteins present in nuclei of both type V-producing and -nonproducing cells.

DISCUSSION
The major objective of this study was to identify and characterize the shortest DNA sequence capable of directing high and cell type-specific transcription from the human COL5A2 promoter. This goal was pursued by correlating and integrating data obtained from cell transfection experiments and in uitro DNA binding assays. As a result, the 5' boundary of the minimal COL5A2 promoter was located 152 bp upstream of the start site of transcription. Previously described features of this promoter (20, 43) include the absence of a CCAAT motif and the presence of a TATA-like element (TATTTA) located between residues -32 and -27 ( Fig. 2.4).
Cell type-specific transcription of the minimal promoter is under the positive control of two distinct, nearly overlapping regulatory sequences, termed FP-A (-115 to -99) and FP-R (-149 to -118). The specificity of DNA/protein interactions at these two sites was extensively documented by competition experiments which utilized homologous, unrelated, and mutant sequences. Collectively, these experiments suggest that the two COL5A2 elements are likely to he novel nuclear protein binding sites. Rased on the availahle evidence, we cannot, however, exclude that either or both of the transacting factors binding to FP-A and FP-R may represent an alternative form of an already characterized nuclear protein or a structurally related hut geneticallv distinct product.
Preliminary analysis of the ionic requirements for DNA hinding indicates that formation of complexes I and I1 is not inhibited by high concentrations of EDTA nor is it improved by the addition of zinc ions. In contrast, hinding of the two complexes differs somewhat from each other with respect to temperature and optimal M$+ concentration. Complex I dissociates a t lower M$+ concentration and lower temperature than complex 11. The distinct behavior of the two complexes under these experimental conditions adds support to the Regulation of pro-a2(V) Collagen Transcription 25395 notion that the FP-A and FP-B sequences are recognized by different transcription factors.
It is of interest to note that the corresponding 52-bp region of the mouse promoter differs from the human sequence for only three nucleotides, none of which is contacted by nuclear factors (43). This phylogenetic conservation of sequence indirectly substantiates the importance of this upstream regulatory segment for COL5A2 expression. Finally, our DNA binding experiments left unresolved whether or not the occupancy of the two COL5A2 sites by transcription factors is mutually exclusive. This is because of the lack of reproducibility and specificity of the third, slower migrating DNAprotein complex seen occasionally in the DNA binding assays. Additional information about the transcriptional features of the minimal promoter was obtained in functional assays that utilized various manipulations of the 52-bp sequence and different recipient cells. From these experiments, several conclusions can be drawn. First, the formation of transcriptionally active complexes depends greatly on the steric conformation of the 52-bp region, in that this DNA element is substantially more active when inserted into the CAT-plasmid in its natural orientation.
Second, transcriptionally active complexes can be formed irrespective of the surrounding sequences, as shown by the behavior of the COLlA2/COL5AZ mosaic promoter. Third, the contribution to transcriptional activity of the FP-B element is significantly greater than FP-A, since a combination of sequences in which only the integrity of the former sequence is preserved still confers substantial activity to the COL5A2 promoter. Lastly and in contrast to type V-nonproducing cells, nuclear proteins from type Vproducing cells that bind to the 52-bp region appear to be in a transcriptionally active form. The conclusion is obviously based on the assumption that in uitro binding of FP-A and FP-B is to the same proteins in different nuclear extracts. This seems not to be the case at least for the DNA-protein complex that is formed in uitro between the FP-B sequence of COL5A2 and proteins present in Jurkat nuclei (Complex 11). Indeed, from the available evidence, one might even argue that, in addition to being the major contributor to COL5A2 transcription in type V-producing cells, FP-B is also capable of binding to negative trans-acting factors that repress COL5A2 expression in type V-nonproducing cells.
Such a redundancy of binding and duality of effects are both well documented phenomena (36). Similarly, there are in the literature numerous examples of putative gene-specific transcription factors that are expressed more broadly than the target genes (36). Relevant to the collagens, a case in point is the recently characterized CCAAT-binding factor, CBF (37). Despite its postulated specificity for coordinating transcription of the two type I collagen genes, the pattern of expression of this heteromeric activator is not confined to tissues of mesenchymal origin (38, 39).
Additional characterization of the trans-acting factors binding to the two &-acting elements of the minimal COL5A2 promoter will eventually clarify the nature of these regulatory interactions in type V-producing and -nonproducing cells. This will also enable us to detail the interplays between these two regulatory elements as well as between them and addi-tional control elements, such as those mapped further upstream of the 52-bp region.