Structure and characterization of the human tissue inhibitor of metalloproteinases-2 gene.

We report here the characterization of the human tissue inhibitor of metalloproteinases-2 (TIMP-2) gene. The gene is 83 kilobase pairs (kb) long with exon-intron splicing sites located in preserved positions among the three members of the TIMP family. A 2.6-kb genomic DNA fragment flanking the 5′-end of the gene contains several regulatory elements including five Sp1, two AP-2, one AP-1, and three PEA-3 binding sites. Despite the presence of a complete AP-1 consensus at position −281, the promoter did not respond to 12-O-tetradecanoylphorbol-13-acetate treatment. However, 12-O-tetradecanoylphorbol-13-acetate response was generated by insertion of a similar AP-1 consensus at position −71, indicating the importance of the positioning of this motif. The promoter contains a typical CpG island; however, methylation of this island did not seem to influence gene expression. Analysis of the 3′-end of the gene revealed that the two mRNAs for TIMP-2 (1.2 and 3.8 kb) differ by the selection of their polyadenylation signal sites, but selection of these sites does not affect RNA stability. In summary, the TIMP-2 gene has several features observed in housekeeping genes, and differs significantly from TIMP-1 and TIMP-3 genes. These differences are likely to explain the specific roles that these inhibitors play in the regulation of matrix metalloproteinases.

Matrix metalloproteinases (MMP) 1 are a large family of secreted neutral endopeptidases with a broad spectrum of proteolytic activity for several components of the extracellular matrix (ECM). Among this family three groups have been well characterized based on their substrate specificity and include collagenases, gelatinases, and stromelysins (1). These enzymes have been implicated in physiological and pathological conditions associated with breakdown of the ECM such as trophoblastic implantation, embryonic development, angiogenesis, osteoarthritis and tumor invasion (2,3). The activity of these proteases in the ECM is regulated by specific inhibitors known as tissue inhibitors of metalloproteinases (TIMP). So far, the TIMP family consists of three members TIMP-1, -2, and -3, characterized in different species. The human TIMP-1 gene has been localized on the chromosome X (4), whereas the TIMP-2 and TIMP-3 genes have been assigned to chromosome 17 (5) and 22 (6,7), respectively. These inhibitors inhibit the proteolytic activity of activated MMP by forming tight (K i Յ 10 Ϫ9 M) 1:1 stoichiometric inhibitory complexes with the enzyme (8). Despite the fact that the three TIMP genes share a common inhibitory activity for all members of the MMP family, there is experimental evidence indicating that they have specific functions. For example, TIMP-1 and TIMP-2 form preferential complexes with pro-MMP-9 and pro-MMP-2, respectively (9, 10), and whereas TIMP-1 and TIMP-2 are present in a soluble form, TIMP-3 is associated with the ECM (11). TIMP-1 and TIMP-2 have been shown to promote the growth of erythroid precursor cells as well as of a variety of normal and malignant cells, suggesting a bifunctional role for these inhibitors (12)(13)(14).
Additional evidence supporting specific roles for these three inhibitors in vivo has been provided by experiments showing their differential expression in cells and in tissues and during development. For example, in adult mice, TIMP-1 is preferentially expressed in epithelial tissues, in cartilage, and in muscles (15); and during murine embryonic development, TIMP-1 is specifically expressed in developing bone, whereas TIMP-3 is preferentially found in developing epithelia, cartilage, and muscles (16). TIMP-1 and TIMP-3 are up-regulated by 12-Otetradecanoylphorbol-13-acetate (TPA) and TGF-␤, whereas TIMP-2 is down-regulated by both agents (10,11,17). In macrophages, lipopolysaccharides have been reported to downregulate TIMP-1 and up-regulate TIMP-2 (18). The promoters of the murine (19,20) and the human (21) TIMP-1 and the murine TIMP-3 (22) genes have been fully characterized and shown to contain a TPA-responsive element (TRE), consistent with their response to TPA, whereas the human TIMP-2 (23) and TIMP-3 (7) promoters have only been partially characterized. Altogether, these observations suggest that each individual member of the TIMP family has a specific physiological function.
These observations led us to examine the molecular basis for some of these differences, by characterizing the human TIMP-2 (hTIMP-2) gene. In this report, we provide evidence supporting a major role of this inhibitor in providing a stable basal level of inhibitory activity in tissues.

EXPERIMENTAL PROCEDURES
Isolation of TIMP-2 Genomic Clones-A human TIMP-2 cDNA probe containing the full (1,035-nt) sequence (24) was used to screen a human placenta library prepared in the cosmid pWE 15 vector (Stratagene, La Jolla, CA) and a human chromosome 17 library in the cosmid SuperCos I vector (originally provided by Dr. Larry Deaven at Los Alamos National Laboratory). The human chromosome 17 library was screened by hybridization to duplicate high density clone arrays on nylon filters as described elsewhere (25). For the human placenta library, a total of 3.6 ϫ 10 6 colony forming units were plated on nylon membranes (Hybond-N, Amersham Corp.) placed on 20 150-mm Luria Bertani broth (LB)/kanamycin (100 g/ml) agar plates. After colony growth, two replicates were made for each master membrane, and the bacteria were allowed to grow on the replica filters placed on fresh LB/kanamycin agar plates at 37°C. The replica membranes were then prepared for hybridization by placing them on a Whatman paper prewetted with 0.5 M NaOH for 30 s, followed by sequential washes in 1 M Tris-HCl (pH 7.6) and 1 M Tris-HCl (pH 7.6), 1.5 M NaCl. After UV cross-linking, using a Stratagene UV Crosslinker (model 1800), the filters were prehybridized for 2 h at 42°C in a solution of 0.8 M NaCl, 0.02 M PIPES (pH 6.5), 50% deionized formamide, 0.5% SDS containing sonicated salmon sperm DNA (100 g/ml) denatured by boiling for 10 min. Hybridization was carried out overnight at 42°C in the presence of 7 ϫ 10 6 cpm/filter of [␣-32 P]dCTP-radiolabeled TIMP-2 cDNA. After washing once for 30 min at 55°C with 3 ϫ standard sodium citrate (SSC), 0.1% SDS and once for 30 min at 55°C with 0.1 ϫ SSC, 0.1% SDS, the filters were autoradiographed on an X-Omat film (Eastman Kodak Co.) for 2 days at Ϫ80°C. Positive colonies matching in both replicate filters were picked from the master filter and inoculated in 100 l of LB/kanamycin and used for a secondary and a tertiary screening. This tertiary screening identified eight positive clones from the human placenta library and 12 positive clones from the human chromosome 17 library. The positive clones from the tertiary screening were then examined by Southern blot analysis, after digestion with EcoRI, using a 5Ј-end probe (a 2.6-kb PstI fragment containing exon 1), a 3Ј-end probe (a 1.4-kb EcoRI fragment containing the last exon) derived from a EMBL 3 library (23) and oligoprobes corresponding to TIMP-2 cDNA sequences extending from amino acid 21 to 27 (oligo YDC1: 5Ј-GCCAAAGCGGTCAGTGAGAAG-3Ј) and from amino acid 76 to 81 (oligo 1332: 5Ј-CTTTCCTCCAACGTC-CAG-3Ј). These oligonucleotides correspond to TIMP-2 cDNA sequences located in corresponding exons 2 and 3 for TIMP-1 (26). Seven among the eight positive clones derived from the human placenta library were found identical and hybridized with the 5Ј-end TIMP-2 probe. One was positive with the 3Ј-end probe. None of these eight clones hybridized with oligoprobes corresponding to putative exons 2 and 3. Among the 12 clones isolated from the chromosome 17 library, none hybridized with the 5Ј-end probe, but eight (including clone 27F6) hybridized with both oligoprobes YDC1 (putative exon 2) and 1332 (putative exon 3) and with the 3Ј-end probe. One clone (clone 67H4) hybridized with oligoprobe YDC1 only. One clone derived from the human placenta library and containing exon 1 (clone 1.3.3) and two clones derived from the chromosome 17 library (clones 67H4 and 27F6) were then selected to obtain the entire map of the human TIMP-2 gene.
Restriction Mapping of Genomic Clones-We used a modified method of Wahl et al. (27) to map the positions of the EcoRI sites in the TIMP-2 genomic clones. Genomic sequences from these positive clones including the flanking T3 and T7 promoter sequences were excised from either the pWE 15 vector or the SuperCos I vector by digestion with NotI. These genomic fragments were then subjected to partial digestion with EcoRI (1 g of DNA digested in the presence of one unit of enzyme) for 5, 10, 20, and 30 min at 37°C. The reaction was blocked by the addition of EDTA (final concentration 50 mM), and the samples containing digested DNA were electrophoresed in a 1% agarose gel prior to transfer onto a nylon membrane and to hybridization in the presence of [␥-32 P]dATP-labeled T3 and T7 oligonucleotides. Southern analysis and polymerase chain reaction using primers derived from different exon sequences of the hTIMP-2 cDNA were also performed to determine the size of some introns and to confirm the mapping of the hTIMP-2 gene.
Cell Culture-Human fibrosarcoma HT1080 cell line and mouse NIH3T3 fibroblasts were obtained from the American Type Culture Collection (Rockville, MD) and were cultured in 60-mm tissue culture plates in the presence of minimum essential medium containing 10% (v/v) fetal bovine serum, 2 mM L-glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin at 37°C under a 5% CO 2 atmosphere. Two human bladder carcinoma cell lines (EJ and RT4; Ref. 28) were cultured as described previously.
Analysis of the TIMP-2 Promoter Activity-All the plasmids used to carry out the expression of the hTIMP-2 promoter were constructed using standard recombinant DNA technology (29). A 2.6-kb PstI genomic fragment flanking the 5Ј-region of the hTIMP-2 gene and extending into the first exon was isolated, cloned into pBluescript (Stratagene, La Jolla, CA), and sequenced in both strands by the dideoxynucleotide method of Sanger et al. (30). From this fragment, a 2.3-kb PstI-NotI fragment extending from nt Ϫ2243 to nt ϩ34 (from the transcription initiation site, TIS) was subcloned in orientation upstream of a promoterless luciferase reporter gene in pGL2-basic vector (Promega, Madison, WI) to make pTIMP2-2243. This vector was then used to generate by digestion with exonuclease III (Erase-a-Base system, Promega) a series of subclones containing progressive unidirectional deletions from the 5Ј-end of the inserted TIMP-2 promoter. Seven deletion mutant clones were selected and sequenced on their 5Ј-end to identify the extent of the deletion (pTIMP2-2088, -1936, -1827, -1501, -1135, -585, and -276). These deletion constructs were then used in transient transfection assays in mouse NIH3T3 and human HT1080 cells. The reporter gene constructs were transfected using a standard calcium phosphate precipitation method (31). As control for transfection efficiency, the plasmid pSV 2 ␤-Galactosidase (Promega) was co-transfected with each construct. After 48 h, cells were harvested and lysed. Luciferase activity was measured as described by de Wet et al. (32), and ␤-galactosidase activity was carried out using the Galacto-light kit according to the instructions of the manufacturer (Tropix, Bedford, MA). The transfections were done in duplicate and carried out in at least three separate experiments using 10 g of test plasmid and 1.5 g of control plasmid for each 60-mm plate. Data were expressed as percentage of the relative activity (ratio of luciferase over ␤-galactosidase activities) obtained with the full-length promoter construct (pTIMP2-2243). When indicated, cells were treated with TPA using the following procedure. Twenty-four h after transfection, the culture medium was changed for medium supplemented with 0.5% fetal bovine serum (rather than 10%), and the cells were cultured for another 32 h. TPA (dissolved in ethanol) was then added to the culture medium at a final concentration of 100 ng/ml. After 16 h, cells were harvested and analyzed for luciferase and ␤-galactosidase expression as described above.
Mutations of the AP-1 Element in the TIMP-2 Promoter-Mutations in the hTIMP-2 promoter were achieved using the overlap extension method of Horton and Pease (33). We used an oligonucleotide corresponding to a sequence located between positions Ϫ702 and Ϫ678 as a sense oligonucleotide (5Ј-CAATGCCTCTGCTGCGATCCTACTG-3Ј) and an oligonucleotide corresponding to a sequence located in the pGL2basic vector as an antisense oligonucleotide (5Ј-CATCCTCTAGAG-GATAGAATGGCG-3Ј). Mutation in the TRE in the region Ϫ288 to Ϫ281 was achieved with two overlapping sense and antisense oligonucleotides corresponding to a sequence (AGAGGCAG) that substituted two thymidines in the TRE consensus (TGAGTCAG). This plasmid was designated pTIMP2-2243/AP-1*. Mutation that introduced a TRE between two thymidines in positions Ϫ71 and Ϫ72 was accomplished using overlapping oligonucleotides corresponding to the sequence (GAGTCA). This plasmid was designated pTIMP2-2243ϩAP-1.
DNase Footprinting-DNase footprinting assays were carried out using a BamHI-NarI fragment (nt positions Ϫ519 to Ϫ188) of the hTIMP-2 promoter encompassing the AP-1 binding consensus sequence (positions Ϫ288 to Ϫ281). This fragment was [␣-32 P]dCTP-end-labeled at the NarI site using Klenow enzyme and gel-isolated. This labeled fragment was then incubated with DNaseI (1 unit/l, diluted 1:3000 -1:500) at 22°C in the presence or absence of purified c-Jun homodimer protein (Promega). Fragments generated were analyzed by electrophoresis in a 7 M urea containing acrylamide (5%) gel. The gel was then examined by autoradiography at Ϫ80°C. Analysis of Methylation-The level of methylation of the cytosines within the human TIMP-2 promoter was determined using restriction endonucleases HpaII and MspI as described by Chandler et al. (34) in addition to other restriction endonucleases whose activity is affected by methylation of the cytosines in their restriction sites (FnuDII, NarI, and NotI). A BamHI-ApaI fragment (positions Ϫ520 to ϩ169) obtained by restriction digestion of the 2.6-kb PstI genomic fragment was used as probe for Southern blot analysis of genomic DNA.
Actinomycin D Treatment-Confluent cultures of HT1080 cells were treated with actinomycin D (final concentration 16 M) dissolved in dimethyl sulfoxide (Sigma). At indicated times RNA was extracted and processed for Northern blot analysis.
RNA Extraction and Northern Blot Analysis-Cytoplasmic RNA was isolated using the method of Chirgwin et al. (35). Poly (A) ϩ RNA was obtained using the Poly(A)tract ® mRNA isolation system (Promega). Cytoplasmic RNA samples (20 g) were electrophoresed on a formaldehyde-containing agarose (1%) gel and blotted onto a nylon membrane. Quantitative analysis of the mRNA was performed by measuring the intensity of the radioactive signals on a GS-250 Molecular Imager (Bio-Rad).

Molecular
Cloning and Mapping of the hTIMP-2 Gene-By screening two genomic libraries, we isolated three overlapping genomic clones, which contained the entire human TIMP-2 gene, spanned in approximately 83 kb (Fig. 1). The gene is composed of five exons separated by four introns of 54.8, 2.7, 9.1, and 1.7 kb. The exon-intron boundaries for these five exons have splicing sites located at positions corresponding to amino acids 17, 51, 87, and 129 (Table I). Sequence analysis of these exon-intron boundaries was consistent with conserved sequences found in splicing sites with a GT motif on the donor site and an AG motif on the acceptor site.
Characterization of the 5Ј-End of the hTIMP-2 Gene-The sequence of the 2.6-kb PstI genomic fragment containing the 5Ј-end of the hTIMP-2 gene and including part of the first exon is presented in Fig. 2. This region contains 2243 bp upstream of the major transcription initiation site (23) that includes a TATA-like motif (AATAAAA, Ref. 36) at positions Ϫ26 to Ϫ20, five consensus sequences for Sp1 (located at positions Ϫ1254 to Ϫ1246, Ϫ1222 to Ϫ1214, Ϫ493 to Ϫ485, Ϫ420 to Ϫ412, and Ϫ237 to Ϫ229), a complete consensus sequence for AP-1 (at positions Ϫ288 to Ϫ281), two binding motifs for AP-2 (at positions Ϫ277 to Ϫ270 and Ϫ213 to Ϫ206), and three PEA-3 binding sites (at positions Ϫ1657 to Ϫ1652, Ϫ842 to Ϫ837, and Ϫ726 to Ϫ721). In addition, some other consensus elements such as nuclear factor-1, NF-IL6, and myocyte-specific enhancer-binding factor-2 binding motifs were identified (37). Fur- thermore, the most proximal region of the hTIMP-2 promoter extending from nt ϩ1 to nt Ϫ300 has a G/C content of 76% and contains a typical CpG island (23). The promoter activity of 5Ј-end flanking region of the hTIMP-2 gene was examined by transient transfection assay in mouse NIH3T3 cells (Fig. 3). Deletion of the region extending from nucleotide Ϫ2243 to Ϫ276, which includes most of the binding consensus, did not significantly affect the promoter activity in NIH3T3 cells. Furthermore the activity of the largest promoter construct (pTIMP2-2243) was in the same range as the activity of the smallest promoter construct (pTIMP2-276), suggesting an absence of involvement of the various binding consensus elements in the basal expression of the gene and confirming, as previously reported (23), that the short Ϫ276 bp region encompassing a single Sp1 binding site and an AP-2 binding site contained all the elements required for basal expression.
Role of the AP-1 Element in Basal Expression and Response to TPA-The presence of a complete AP-1 binding consensus (TGAGTCAG) at position Ϫ288 to Ϫ281 in the hTIMP-2 gene suggested that this element could play a regulating role in the transcription of the gene. We first examined by DNaseI footprint analysis whether this consensus could bind the c-Jun homodimer protein (Fig. 4). The data showed the presence of a clear zone of protection, which appeared in a dose-dependent manner with the addition of c-Jun homodimer (Fig. 4, lanes  2-5). Thus, the AP-1 consensus was found to bind AP-1 in vitro.
To determine whether the AP-1 consensus was involved in basal expression of the hTIMP-2 gene, a mutation that replaced the TGAGTCAG consensus by a nonfunctional AGAGGCAG consensus (38) was generated (Fig. 5A). This mutation resulted in a 22 and 38% decrease in the basal expression of the reporter gene in HT1080 and NIH3T3 cells, respectively (Fig. 5B), suggesting some involvement of the AP-1 consensus in basal gene expression.
We then examined whether the presence of this consensus could affect the promoter activity after treatment with TPA in transient transfection assays. The data (Fig. 5C) indicated no significant changes in reporter gene expression after TPA treatment. Since in many TPA-responsive genes the TRE is found in close proximity to the TATA box (39), we postulated that the more distant position of the TRE in the hTIMP-2 gene may be responsible for the lack of response to TPA. To test this hypothesis, we inserted by mutagenesis, an AP-1 binding consensus TGAGTCAT between positions Ϫ72 and Ϫ71, in closer proximity of the TATA box. This mutation not only resulted in a 2-3-fold increase in basal expression of the hTIMP-2 promoter but was also associated with an additional 2-fold increase in activity after TPA treatment (Fig. 5C). Thus, the position of the AP-1 consensus in close proximity to the TIS seems to be an important factor influencing its activity, and the particular position of this consensus in the hTIMP-2 promoter may be responsible for a lack of up-regulatory function.
Characterization of the 3Ј-end of the hTIMP-2 Gene and RNA Stability-The characterization of the 3Ј-end of the hTIMP-2 gene brought novel information on the molecular basis for the presence of two TIMP-2 mRNAs of 1.2 and 3.8 kb as previously shown by us (23) and others (10). The size of the 1.2-kb mRNA is consistent with the positions of the TIS in the 5Ј-end of the gene and the polyadenylation signal in the 3Ј-UTR of exon 5 (1,069 nt). Two genomic DNA fragments located downstream of this site, one derived from the distal end of a 1.4-kb EcoRI fragment containing exon 5 (0.3-kb XhoI-EcoRI fragment) and one derived from a 2.8-kb EcoRI fragment located further downstream, were used as probes in Northern blot analysis (Fig. 6). The data show that both fragments failed to hybridize with the 1.2-kb mRNA but hybridized with the 3.8-kb mRNA. Furthermore, the 0.3-kb XhoI-EcoRI fragment weakly hybridized with an additional 1.7-kb mRNA. The data suggest that the hTIMP-2 mRNAs differ by the selection of their polyadenylation signal sites and indicate the presence of an additional mRNA of 1.7 kb in a small amount. Consistently, sequencing of the 2.8-kb EcoRI fragment revealed the presence of five polyadenylation signal consensus (AATAAA) located between 3.25 and 4.75 kb downstream of the TIS. To determine whether the length of the hTIMP-2 mRNA could influence RNA stability, we performed RNA analysis after treatment with actinomycin D in HT1080 cells (Fig. 7). These experiments indicated no significant difference in the half-life of the two hTIMP-2 mRNAs (32 and 26 h for the 1.2-and 3.8-kb TIMP-2 mRNAs, respectively), and showed that both mRNAs had a half-life longer than the human ␤-actin mRNA (20 h).
Regulation of the hTIMP-2 Gene by Methylation-The presence of a CpG island in the most proximal region of the hTIMP-2 promoter (Fig. 8A) led us to examine whether methylation of cytosines in the promoter could affect gene expression. For these analyses, we selected two human bladder carcinoma cell lines (EJ and RT4) because of the presence (EJ) or the absence (RT4) of TIMP-2 expression (Fig. 8B). The methylation level of the promoter region in these cells was first examined by Southern blot analysis of genomic DNA digested with PstI and HpaII or MspI and probed with a 0.7-kb BamHI-ApaI promoter fragment (Fig. 8, A and C). Whereas both HpaII and MspI digest 5Ј-CCGG-3Ј sequences, only MspI can cleave these restriction sites when the internal cytosine is methylated. The data revealed a higher degree of DNA cleavage with MspI than with HpaII, suggesting the presence of multiple  The methylation status at some specific G/C sequences was also determined using restriction enzymes NotI, NarI, and FnuDII, which all have a C/G motif in their restriction sequence (Fig. 8A). These data showed that the unique NotI site in the promoter is unmethylated, as indicated by the presence of a 2.3-kb band on the Southern blot (Fig. 8C). The presence of a 2.0-kb band after digestion with NarI also indicated that the two more distal NarI sites are methylated (uncleaved), whereas the one in position Ϫ191 is unmethylated, as shown by the presence of a 0.5-kb band on the Southern blot. The methylation status at the most proximal site (position ϩ306) could not be determined because of the small size of the cleaved fragment. Analysis of the FnuDII-generated fragments revealed the presence of a large 1.9-kb fragment, indicating that the most distal site (position Ϫ865) is methylated, whereas sites in most proximal positions are unmethylated, although, because of the high number of FnuDII sites in close proximity in that region, this analysis did not allow us to determine the methylation status of each of them. All of these analysis revealed no difference in the pattern of digestion between TIMP-2-expressing and nonexpressing cells. The data suggest, therefore, that although the methylation state of the various G/C sequences in the hTIMP-2 promoter varies, it is unlikely that it has any effect on gene expression. DISCUSSION We have isolated the entire hTIMP-2 gene and described its structural organization. Comparison of the structure of the hTIMP-2 gene with the published structures of the murine TIMP-1 gene (26) and the murine (40) and human (7, 41) TIMP-3 genes revealed both similarities and differences. As is the case in many genes that belong to a same family, we found that the exon-intron boundaries were preserved in the three members of TIMP family. In contrast to the TIMP-1 gene, which is contained within a 4.5-kb HindIII genomic fragment, the hTIMP-2 gene is much larger (spanning approximately 83 kb) and contains a first intron of 57 kb. The significance of the presence of such large intron, also observed in the mouse and the human TIMP-3 genes (7,40,41), is unclear, and whether this intron contains (as is the case in the TIMP-1 gene (26)) elements capable of enhancing gene expression is unknown. Another difference between the TIMP-1 and the TIMP-2 genes resides in the 5Ј-UTR, which contains in the case of TIMP-1 a short noncoding first intron (Fig. 9). Sequencing of the 2.6-kb PstI genomic fragment of the hTIMP-2 gene indicated no differences between the cDNA and the genomic sequences, confirming that the ATG codon was located within the first exon. Further comparison of the structure of the TIMP genes from different species should lead to a better understanding of their evolutionary relationship.
In this manuscript we have extended our first study of the hTIMP-2 promoter (23) to include a total of 2.3 kb of 5Ј-flanking sequences upstream from the major transcription initiation site. A comparison between this promoter region and similar regions of the murine TIMP-1 and TIMP-3 genes is shown in Fig. 9. The promoter region of the hTIMP-2 gene, like the murine TIMP-3 gene, has a higher G/C content (76 and 70% of G/C in the region extending from nt Ϫ300 to nt ϩ1 for hTIMP-2 and murine TIMP-3, respectively) than the corresponding region of the TIMP-1 gene (58% G/C content). Furthermore, the positions of some key transcription elements are different between the promoter region of the three members of the TIMP family. Whereas in the TIMP-1 promoter, the AP-1 binding consensus is proximal to the TIS (positions Ϫ59 to Ϫ53) and closely associated with a PEA-3 element, the AP-1 binding element in the hTIMP-2 gene is more distantly positioned from the TIS (positions Ϫ288 to Ϫ281), and three PEA-3 elements are located further upstream (position Ϫ721 and further upstream). In contrast, in the promoter region of the murine TIMP-3 gene there are six AP-1 binding consensus elements (located between positions Ϫ1950 and Ϫ611) with many PEA-3 elements variably dispersed in the promoter region.
The presence of an AP-1 binding site is a common feature of many genes up-regulated by TPA. Often this element is located in close proximity (Ϫ50 to Ϫ70 nt) of the TIS and is closely associated with one or several PEA-3 elements to form a TPAand oncogene-responsive unit (43-45) as seen in the promoters of many TPA-responsive genes including several MMP and TIMP-1 (19 -21, 38, 39, 42). We had previously demonstrated FIG. 8. Methylation status of the hTIMP-2 promoter. A comparison between cell-expressing and nonexpressing TIMP-2 mRNAs is shown. Genomic DNA (15 g) was isolated from cultures of human bladder carcinoma cell lines (EJ and RT4) and digested with PstI alone (Ϫ) or with PstI and HpaII, MspI, FnuDII, NarI, or NotI. Samples of digested DNA were electrophoresed, blotted to a nylon membrane, and hybridized to a BamHI-ApaI fragment extending from nt Ϫ520 to nt ϩ169 in the promoter region. A, map of the restriction sites for the enzymes used. The positions of the CpG and GpC sequences as well as the localization of the probe are also shown. B, cytoplasmic RNA was obtained from EJ and RT4 cells and analyzed by Northern blot using the hTIMP-2 cDNA and the human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as probes. C, autoradiography of the Southern blot. that a 715-bp-long promoter sequence of the hTIMP-2 gene containing the AP-1 consensus failed to respond to TPA (23). We now demonstrate that a longer construct containing three PEA-3 elements in addition to the AP-1 binding consensus also failed to respond and also provide evidence that the AP-1 consensus binds the c-Jun homodimer in vitro. We therefore postulated that the failure of this element to respond to TPA may be related to its position, far more distant (281 bp) from the transcription initiation site than in the case of most TPAinducible genes (39,42). By mutagenesis experiments, we demonstrated that insertion of a consensus for AP-1 at position Ϫ71 resulted in an increase in basal gene expression and also in an additional increase in expression after treatment with TPA, clearly confirming the importance of the position of the AP-1 consensus not only in basal gene expression but also in TPA inducibility. It is conceivable that such positioning prevents interaction of the Jun and Fos transactivation factors (AP-1) with the transcription initiation complex. We had previously shown that elimination of a 124-bp SmaI fragment of the hTIMP-2 promoter containing the AP-1 consensus resulted in a 2-fold increase in basal gene expression (23), suggesting the presence of inhibitory sequences in this region. Since the AP-1 consensus has been previously shown to repress gene expression in some cases (46,47), we examined whether the AP-1 consensus in the hTIMP-2 promoter could suppress expression. Our mutagenesis data clearly show that this is not the case, since mutation of the AP-1 consensus in its original position did not result in an increase of basal expression. Rather, a small but significant decrease in expression was observed, suggesting some involvement of the AP-1 site in basal expression.
The high G/C content of the hTIMP-2 promoter suggests that its activity could be controlled by methylation of the cytosine residues as shown in several G/C-rich promoters (48) including the murine TIMP-3 (22). Whereas Sun et al. (22) have shown that abnormal methylation of the mouse TIMP-3 promoter is responsible for lack of expression of the gene in neoplastic JB6 cells, our data suggest that methylation does not play a role in the regulation of TIMP-2 expression, since no differences were found in the restriction endonuclease digestion patterns between TIMP-2 expressing and nonexpressing cells.
The TIMP-2 gene has been shown to be down-regulated by lipopolysaccharides in macrophages (18) and up-regulated by cAMP in HT1080 cells (49). Interestingly, our analysis identifies several potential target sequences for these agents. AP-2 binding sites have been shown to mediate cAMP response in many genes, and two of these sequences were found in the hTIMP-2 promoter. Lipopolysaccharides have been shown to affect gene regulation via interleukin-6 and NF-IL6 consensus elements (37), and two NF-IL6 elements were identified in the hTIMP-2 promoter. Whether these elements are responsible for the reported effect of cAMP and lipopolysaccharides in TIMP-2 remains to be determined.
Analysis of the 3Ј-end of the hTIMP-2 gene brought some important information on transcription of the hTIMP-2 gene. Two mRNA species of 1.2 and 3.8 kb coding for the TIMP-2 gene have been previously reported, but the molecular basis for these differences has not been previously examined. Although alternative splicing has been suggested (50), the absence of isoforms of TIMP-2 made this possibility unlikely. Our data show that the difference in size of the mRNAs of hTIMP-2 gene is the result of the use of different polyadenylation signals within the 3Ј-end of the gene. The biological relevance of this observation is not yet clear; however, our data show that these differences in the polyadenylation signal site do not affect RNA stability.
In summary, our analysis of the structure of the hTIMP-2 gene has pointed to several features suggesting that TIMP-2 provides a stable basal level of antimetalloproteinase activity in tissues. As our understanding of the role of the various TIMP in controlling MMP activity during tissue remodeling continues to improve, the significance of differences between the promoter of TIMP-2 and TIMP-1 and TIMP-3 discussed here will become more obvious.