Transcriptional induction of collagenase-1 in differentiated monocyte-like (U937) cells is regulated by AP-1 and an upstream C/EBP-beta site.

In this report, we demonstrate that the AP-1 site and a distal promoter element regulate transcriptional induction of collagenase-1 during monocytic differentiation. Chloramphenicol acetyltransferase expression constructs containing regions of the human collagenase-1 promoter were stably or transiently transfected into U937 cells, and reporter activity was assessed at various times after the onset of phorbol 12-myristate 13-acetate (PMA)-mediated differentiation. Rapid and strong induction of promoter activity was lost in constructs with a mutant AP-1 element; however, at 16-96 h post-PMA, the mutant collagenase-1 promoter displayed AP-1 independent PMA-mediated transactivation. The AP-1 mutant constructs also showed delayed transcriptional activation in PMA-treated fibroblasts. Western and supershift analyses indicated that functional Jun and Fos proteins were present in nuclear extracts of PMA-differentiated U937 cells. Promoter deletion constructs demonstrated the potential role of distal promoter sequences in regulating collagenase-1 transcription. In particular, Western, supershift, and promoter deletion analyses suggested a role for CCAAT/enhancer-binding protein-beta (C/EBP-beta) binding site between -2010 and -1954 in regulating transcription of collagenase-1 in monocytic cells. Our findings suggest that distinct regulatory elements, acting somewhat independently of each other, control expression of collagenase-1. In addition, our data suggests that the rapid PMA-mediated induction of collagenase-1 transcription is controlled by a mechanism distinct from that regulating the sustained expression of this proteinase in activated macrophages.

We assessed the requirement of the AP-1 site and more distal promoter sequences to collagenase-1 gene activation during and subsequent to monocytic differentiation. We used PMAtreated U937 cells as an in vitro model because they mimic the differentiation of monocytes into macrophages (33) and because activation of collagenase-1 expression in these cells occurs strictly by a transcriptional mechanism (27,34). We report that collagenase-1 promoter activity is induced and maintained in the absence of a functional AP-1 site. We conclude that although the AP-1 site is required to mediate strong collagenase-1 transcription, other upstream elements, including a newly identified CCAAT/enhancer-binding protein-␤ (C/ EBP-␤) site, participate in achieving maximal and sustained PMA-mediated collagenase-1 transactivation in monocytic cells.

EXPERIMENTAL PROCEDURES
Cell Culture-U937 cells (35) were obtained from the American Type Culture Collection (CRL 1593) and maintained in RPMI 1640 medium (Life Technologies, Inc., Gaithersburg, MD) supplemented with 10% low endotoxin fetal calf serum (Life Technologies, Inc.), non-essential amino acids, L-glutamine, sodium pyruvate, 100 units/ml penicillin, and 100 g/ml streptomycin. For induction of cell differentiation, U937 cells were plated at 5 ϫ 10 5 cells/ml and exposed to 8 ϫ 10 Ϫ8 M PMA (Sigma). Human skin fibroblasts were grown in Dulbecco's modified Eagle's medium, 10% fetal calf serum (Life Technologies, Inc.) containing the same supplements listed above.
RNA and Transcription Analyses-Total RNA was isolated by the guanidinium phenol extraction method (36). Conditions for Northern hybridization and washes were as described (27). Blots were hybridized with a 2.2-kb human collagenase-1 (37), a 2.0-kb human c-fos (38), a 1.2-kb human c-jun (39), or a 1.3-kb rat glyceraldehyde-3-phosphate dehydrogeanse (40) random-primed, 32 P-labeled cDNA probe. Filters were washed and then visualized by autoradiography. Transcription rates of specific mRNAs were measured using 2.5 ϫ 10 7 isolated nuclei as described (41). Nascent RNA transcripts were isolated, and equivalent counts of 32 P-labeled RNA were hybridized to denatured, gelpurified cDNA inserts slotted on nitrocellulose. As an indicator of total transcription, a 2.5-kb pair human Alu repeat fragment derived from the ⑀-globin gene (42) was blotted as well.
Plasmids and Constructs- Fig. 1 shows maps of all collagenase-1 promoter constructs used in this study. pBLCAT2 contains the Ϫ105/ ϩ51 region of the herpes simplex virus thymidine kinase (TK) promoter fused to a CAT reporter gene (43). pAPCAT2a is derived from pBLCAT2, and contains a tandem triplet of the collagenase-1 AP-1 site subcloned 5Ј of the TK promoter. A plasmid containing the Ϫ2278/ϩ36 region of the human collagenase-1 promoter (44) was generously pro-vided by Dr. Steven Frisch (La Jolla Cancer Research Foundation, La Jolla, CA). To eliminate the possibility of transcriptional differences due to the vector backbone, all collagenase-1 promoter sequences were subcloned into pBLCAT2. The TK promoter of pBLCAT2 was removed during the synthesis of the collagenase-1 promoter deletion constructs (Fig. 1A). The p-2278CAT, p-2278MCAT, p-511CAT, p-179CAT, p-95CAT, and p-72CAT vectors were generated by PCR as described (26). Constructs p-2010CAT, p-1954CAT, p-1689CAT, p-1552CAT, p-1197CAT, and p-997CAT were made by digestion of p-2278CAT with BsaHI, HpaI, XmnI, EcoRV, BglII, or BamHI, respectively. The digestion products were blunted (when necessary) with the Klenow fragment of DNA polymerase, further digested with XhoI, and the appropriate blunt/XhoI fragment was subcloned into blunted-HindIII/XhoI digested pBLCAT2.
The internal deletion construct, p-1197⌬-997CAT, was created by cutting p-2278CAT with BglII and BamHI followed by ligation of the vector. Internal deletion construct p-2010⌬-1954CAT was created by recombinant, whole plasmid PCR (45) using the 5Ј (ATAgcatgcACCCT-GGAAGAGTCTCAT) and the 3Ј (CGCgcatgcCTATTAACTCACCCT-TGT) primers (deleted/mutated sequences in lowercase). The PCR product was digested with SphI and ligated. The resultant construct was digested with HindIII and BamHI, and the HindIII/BamHI fragment was subcloned into HindIII/BamHI cut p-2278CAT. All PCR was performed with either Vent TM or Deep Vent TM DNA polymerase (New England Biolabs, Beverly, MA) to minimize unwanted mutations. All newly created plasmids were sequenced to verify that only the desired alterations were introduced during PCR steps. Sequencing reactions were done using a Sequenase TM Kit (U. S. Biochemical Corp., Cleveland, OH).
Heterologous promoter constructs contain collagenase-1 promoter sequences upstream of Ϫ997 linked to the TK promoter. These were constructed by subcloning the BamHI/XhoI TK promoter fragment of pBLCAT2 into deletion constructs that had been digested with BamHI and XhoI. For example, p-2278CAT gives rise to p-2278TKCAT, in which Ϫ2278 is the most 5Ј and Ϫ997 (at the BamHI site) is the most 3Ј nucleotide of the collagenase-1 promoter fragment (Fig. 1B).
Stable U937 Transfectants-U937 cells (10 7 in 0.5 ml) were transfected with 5 g of linearized pRSV-Neo and 50 g of linearized pBLCAT2, pAPCAT2a, p-72CAT, p-511CAT, p-2278CAT, or p-2278MCAT. Cells were electroporated at 250 V and 600 microfarads in a 0.4-cm gap cuvette using a BTX 3000 electroporator (Biotechnologies and Experimental Research, Inc., San Diego, CA), placed on ice for 10 min, added to 9 ml of culture medium, centrifuged to pellet the cells, and plated in 10 ml of fresh medium. After 24 h, cells were shifted to medium supplemented with 400 g/ml Geneticin ® (Life Technologies, Inc.). After 2 weeks, G-418-resistant cells were subcloned by limiting dilution or maintained as a pooled population of clones in medium containing 200 g/ml Geneticin ® . To minimize insertion effects, two groups of stable clones, one consisting of 6 clones and the other of 12 FIG. 1. Human collagenase-1 promoter constructs. A, 5Ј deletion and internal deletion (Ϫ2010⌬-1954 and Ϫ1197⌬-997) constructs containing the collagenase-1 TATA box and transcription start site were made as described under "Experimental Procedures." The mutant AP-1 construct (p-2278MCAT) contains a SmaI restriction site in place of the wild-type AP-1 element. B, heterologous constructs contain upstream (Ϫ2278 to Ϫ997) regions of the human collagenase-1 promoter linked to the thymidine kinase (TK) promoter of pBLCAT2. pAPCAT2a contains 3 collagenase-1 AP-1 sites in tandem upstream of the thymidine kinase promoter of pBLCAT2. clones, were pooled. Southern hybridization with 32 P-labeled CAT cDNA was done on individual clones and demonstrated that incorporated DNA was roughly equivalent among clones (data not shown).
Transient Transfections-U937 cells were transfected by a modification of the DEAE-dextran method essentially as described (26,46). Human skin fibroblasts were transiently transfected by calcium phosphate precipitation. After transfection, cells were allowed to recover for 24 h prior to treatment with PMA. After recovery, cultures were divided equally and cells were plated in medium with or without PMA. After transfection, cells were given fresh medium and allowed to recover for 24 h prior to treatment with PMA. Hirt extraction (47) and Southern hybridization with 32 P-labeled CAT cDNA was done to determine transfection efficiency.
CAT Assays-At the indicated times, cells were harvested, washed, and lysed in 200 l of 250 mM Tris, pH 7.8, by freeze-thawing. Lysates were incubated at 65°C for 5 min and then cleared of debris by centrifugation. Equivalent amounts (25-100 g) of cleared lysate, normalized to total protein (Bradford protein assay; Bio-Rad), were assayed for CAT activity using acetyl-CoA (Sigma) and [ 14 C]chloramphenicol essentially as described (48). Reaction products were separated by thin layer chromatography and visualized by autoradiography. Results were quantified by cutting and counting the appropriate spots from the chromatography plate. Relative induction was obtained by dividing percent acetylation of treated versus untreated samples.
Electrophoretic Mobility Shift and Supershift Assays-Nuclear extracts were prepared by the method of Dignam et al. (49). The integrity of all nuclear extract preparations was assessed by determining the ability of proteins to bind a radiolabeled Oct-1 double-stranded oligomer (data not shown). Oct-1 protein binding was constitutive and, thus, served as an internal control. Only extracts without apparent protein degradation were used. For AP-1 studies, double-stranded oligomers containing either wild-type (GATCAAAGCATGAGTCAGACACCT) or mutant (GATCAAAGCAcccgggAGACACCT) human collagenase-1 promoter sequence were used as probes and competitors. For the binding analyses to the upstream region at Ϫ2010 to Ϫ1954, double-stranded oligomers containing collagenase-1 promoter sequence between Ϫ2013 to Ϫ1990 (TGACGTCTTAGGCAATTTCCTGTC), Ϫ1994 to Ϫ1968 (CT-GTCCAATCACAGATGGTCACATCAC), and Ϫ1970 to Ϫ1947 (CACAT-GCTGCTTTCCTGAGTTAAC) were used as probes (1, 2, and 3 in Fig.  10). Competition was done with oligomers 1, 2, or 3, wild-type (TGCA-GATTGCGCAATCTGCA) or mutant (TGCAGAgactagtcTCTGCA) C/EBP consensus oligomers (Santa Cruz Biotechnology, Santa Cruz, CA), or with a wild-type (AGTTGAGGGGACTTTCCCAGGC) NF-B consensus oligomer (Promega Corp., Madison, WI). Double-stranded oligomers were radiolabeled with [␥-32 P]ATP using T4 polynucleotide kinase or with [␣-32 P]dCTP using the Klenow fragment of DNA polymerase. Binding reactions and electrophoresis conditions were as described (26). Equivalent amounts of nuclear protein (5 g) and probe counts were used in all reactions.
Supershift reactions were identical to those described above, except 1 g of appropriate antibody was added to the binding reactions after addition of the labeled probes and reactions were incubated overnight at 4°C prior to electrophoresis. Western Analysis of Nuclear Proteins-Equivalent amounts of nuclear proteins were prepared for electrophoresis by adding 1 volume of 2 ϫ sample buffer and ␤-mercaptoethanol to 50 mM. Samples were boiled for 1 min and separated through a 10% SDS-polyacrylamide gel. Gels were equilibrated in 1 ϫ transfer buffer (10 mM CAPS, 10% methanol, pH 11.0) prior to transferring to polyvinylidene difluoride paper. After transfer, membranes were blotted according to the procedures suggested by Santa Cruz Biotechnology. Blots were developed using horseradish peroxidase-conjugated secondary antibodies and the enhanced chemiluminescence.

RESULTS
Kinetics of Collagenase-1 Expression-We reported that collagenase-1 transcription is induced in U937 cells at 16 -24 h after exposure to PMA (34). We used various assays to more carefully examine the kinetics of this differentiation-dependent induction. Collagenase-1 mRNA was not detected in untreated U937 cells ( Fig. 2A). By 12 h post-PMA, collagenase-1 mRNA was detected, increased by 24 h, and remained elevated at 48 h post-PMA. Nuclear run-off assays demonstrated that collagenase-1 transcription was detectable at 12 h of PMA differentiation and remained at a constant level thereafter (Fig. 2B). These observations demonstrate that the onset of collagenase-1 induction occurs earlier than reported previously (34). CAT activity conferred by the full-length collagenase-1 promoter construct (p-2278CAT) in transiently transfected, PMA-treated U937 cells paralleled the pattern of induction of the endogenous gene (Fig. 2C). Only slight background CAT activity was seen in untreated cells. By 4 h post-PMA, promoter activity was increased, and maximal and sustained levels of CAT activity were achieved by 8 h post-PMA. Consistent results were obtained in four separate experiments. A similar time course for induction of collagenase-1 promoter activity was observed in stable transformants (Fig. 3, Ϫ2278). Full induction of the wild-type collagenase-1 promoter was detected at 6 h after PMA treatment, and the levels were maintained for up to 96 h (Figs. 2C and 3). These data indicate that events necessary for maximal induction of collagenase-1 are activated within 4 -6 h post-PMA.
The AP-1 Site Is Not Necessary for Delayed Induction or Maintenance of Collagenase-1 Transcription-Stable transformants were created to determine if a functional AP-1 site is needed for collagenase-1 induction. The proximal AP-1 site in the Ϫ2278/ϩ36 promoter fragment was replaced with a SmaI recognition site (Fig. 1A). Gel shift analysis with labeled mutant oligomer demonstrated that the mutated AP-1 site does FIG. 2. Collagenase-1 transcription is induced in PMA-treated U937 cells. U937 cells were treated with 8 ϫ 10 Ϫ8 M PMA for 4 -48 h, then processed for: A, Northern hybridization; B, nuclear run-off analysis; or C, CAT assay. A, total RNA was isolated from U937 cells at the times indicated and 5 g were analyzed by blot hybridization with 32 P-labeled cDNAs for collagenase-1 (C'ase) and glyceraldehyde-3-phosphate dehydrogeanse (GAPDH) mRNAs. Autoradiography was for 18 h. B, nuclei were isolated from control U937 cells (ϪPMA) and from cells treated with PMA for 12, 24, or 48 h. Nascent transcripts were isolated, and equal amounts of 32 P-labeled pre-mRNA were hybridized to an Alu repeat sequence (Alu) or to full-length cDNAs for collagenase-1 (C'ase) and ␤-actin. The autoradiograms shown are of a run-off experiment with nuclei from control (Ϫ) and 48 h PMA-treated (ϩ) U937 cells. Autoradiography was for 6 days. Band intensity was quantified by densitometry. Background hybridization was subtracted from genespecific hybridization signal, and the data in the histogram are expressed relative to the signal for Alu. Five run-offs were done for the 0 and 48-h time points, and the results are shown as the mean Ϯ S.E. Transcription at 12 and 24 h post-PMA was assessed once. C, U937 cells were transiently transfected with p-2278CAT and treated with PMA 24 h later. Cells were harvested and lysed at 4 -48 h after the start of PMA exposure, and CAT activity was assessed using 50 g of cell lysates. The results shown are representative of four separate experiments. not bind nuclear proteins (data not shown). In two groups of pooled clones, mutation of the AP-1 site (p-2278MCAT) eliminated the rapid (i.e. by 4 -6 h) and strong transactivation observed with the wild-type construct (p-2278CAT) (Fig. 3). Between 0 and 8 h post-PMA, no CAT activity was detected in U937 cells stably transfected with the mutant AP-1 construct (data not shown). However, transcriptional induction of the mutant AP-1 collagenase-1 promoter was consistently detected at 16 h post-PMA (Fig. 3, Ϫ2278M). Although CAT activity expressed by Ϫ2278MCAT was much lower than that conferred by the wild-type promoter, the level of CAT activity was maintained for up to 96 h after PMA differentiation, similar to the sustained activity from the wild-type promoter (Fig. 3). Experiments with individual stable clones showed the same patterns of induction with both the wild-type and mutant promoters (data not shown). Southern hybridization demonstrated that incorporated DNA was roughly equivalent among stable lines (data not shown).
AP-1 Independent Induction of the Collagenase-1 Promoter Is Not Restricted to Monocytic Cells-PMA treatment stimulated activation of the wild-type collagenase-1 promoter in human skin fibroblasts (Fig. 4, Ϫ2278). Basal activity of the wild-type collagenase-1 promoter was high in these cells, likely due to constitutive c-Jun expression (data not shown), but at 8 and 24 h post-PMA, CAT activity increased. Mutation of the AP-1 site eliminated the high basal activity seen with the wild-type collagenase-1 promoter (Fig. 4, Ϫ2278M). However, similar to that observed in differentiated U937 cells, p-2278MCAT conferred transcriptional induction in transiently transfected fibroblasts at 8 and 24 h post-PMA (Fig. 4, Ϫ2278M). Thus, the collagenase-1 promoter can be transcriptionally induced by PMA in the absence of a functional proximal AP-1 element. Time matched controls incubated without PMA had the same level of CAT activity for p-2278CAT or p-2278MCAT as did the 0 h cells (data not shown).
c-Jun and c-Fos Are Present in Early and Late U937 Nuclear Extracts-The kinetics of c-fos and c-jun expression were assessed by Northern analysis (data not shown). In contrast to the delayed kinetics of collagenase-1 induction (Fig. 2) and in full agreement with data from others (50 -52), c-fos and c-jun transcripts were detected as early as 15 min post-PMA, peaked between 1 and 2 h after PMA addition, and were sustained at low levels over the next 48 h (data not shown). Because c-Fos protein expression may not correlate with expression of its mRNA in U937 cells (50) and because its subcellular localiza-tion is regulated (53), we used an immunoblotting assay to detect Fos family proteins in nuclear extracts from untreated and PMA-differentiated U937 cells. c-Fos protein was detected in both 1-and 24-h post-PMA nuclear extracts using pan-Fos or c-Fos-specific antibodies (Fig. 5). The upward shift in the c-Fos band seen in the 24-h extract may be due to increased protein phosphorylation (54). While the presence of c-Fos in nuclear extracts at 1 h post-PMA was anticipated, the clear abundance of c-Fos protein in the 24-h extract was not. A previous report indicated that c-Fos protein could not be detected in PMAdifferentiated U937 cells after 2 h of treatment (50). Because these authors immunoprecipitated metabolically-labeled protein from whole cell extracts, they may have underestimated c-Fos protein levels during periods of low c-Fos protein synthesis. We detected no FosB protein by immunoblotting nuclear extracts from untreated or PMA-treated U937 cells with a FosB-specific antibody (data not shown). Proteins distinct from c-Fos were detected by the pan-Fos antibody in the nuclear extract from untreated cells, but not in those from PMA-treated cells. The identity of these proteins is uncertain, but their sizes are consistent with the Fos-related proteins Fra-1 (29.4 kDa) and Fra-2 (35.2 kDa) which are expressed in U937 cells (55). Regardless of the identity of the bands in basal cell extracts, only c-Fos was detected in nuclear extracts of PMA-differentiated U937 cells (Fig. 5). The low molecular weight forms seen in the 1-and 24-h extracts with the c-Fos-specific antibody are probably c-Fos degradation products.
We also used the immunoblotting assay to detect Jun family proteins in nuclear extracts from untreated and PMA-differentiated U937 cells. Using a c-Jun-specific antibody, c-Jun pro- tein was detected in nuclear extracts of U937 cells treated with PMA for 1 or 24 h (Fig. 5). The high molecular weight bands detected in all extracts may be due to ubiquitination (56) or altered phosphorylation of c-Jun (57). Because c-Jun is not expressed in basal U937 cells, the low molecular mass bands between 30 and 20 kDa seen in all samples are likely nonspecific products. No additional bands were detected with the pan-Jun or JunB antibodies (data not shown).
Nuclear Proteins c-Fos and c-Jun Bind the Collagenase-1 AP-1 Site-Electrophoretic mobility shift and supershift assays were done to confirm the presence of active AP-1 binding Jun/ Fos dimers. Nuclear extracts from untreated cells did not support binding to a double-stranded oligomer containing the native collagenase-1 AP-1 site, where as extracts from 4, 8, and 24 h PMA-treated cells exhibited strong binding activity (Fig.  6). The binding activity was competed by excess unlabeled wild-type AP-1 oligomer (Fig. 6, 24C) but not by excess oligomer containing the mutated AP-1 site or by an unrelated sequence (data not shown). In addition, radiolabeled, doublestranded mutant AP-1 oligomers showed no binding to nuclear proteins (data not shown).
Supershift analysis demonstrated that JunD and c-Fos were present in the shifted complexes (Fig. 6). We consistently detected a weak supershifted complex with the c-Jun antibody and a very weak, if any, complex for JunB. Neither the relative amounts of shifted complexes nor their composition changed at any time after the onset of PMA differentiation. A FosB-specific antibody did not supershift complexes formed in extracts from PMA-treated cells (data not shown). The lack of other Fos family proteins agrees with our immunoblotting data (Fig. 5). Although we cannot definitively determine the identity of the Jun component, these results suggest that heterodimers of c-Fos and Jun family proteins contribute to maximal collagenase-1 transcriptional induction at early and late times of U937 differentiation.
Transient Transfection of Collagenase-1 Promoter Constructs in U937 Cells-Because the mutant AP-1 collagenase-1 promoter was induced by PMA, we assessed the influence of regions upstream of the proximal AP-1 site during collagenase-1 induction in U937 cells. Cells were transfected with the various promoter constructs, and CAT activity was assessed at 24 h after addition of PMA. Southern hybridization of Hirt extracted DNA indicated equivalent transfection efficiency among constructs and that the level of plasmid DNA remained constant up to 72 h after transfection (48 h post-PMA, data not shown). For presentation, we have divided the collagenase-1 promoter into upstream (Ϫ2278 to Ϫ511) and downstream (Ϫ511 to ϩ36) regions.
Relative to p-2278CAT (number 1), induction of construct p-511CAT (number 11) was reduced by 50% in response to PMA differentiation (Fig. 7), and p-511CAT had similar activity in PMA-treated stable transformants (data not shown). Deletion of sequences between Ϫ511 and Ϫ179 (p-179CAT, number 12) resulted in no further decrease in PMA responsiveness relative to p-511CAT (number 11). However, PMA responsiveness was further reduced when sequences between Ϫ179 and Ϫ95 were deleted (Ϫ95CAT, number 13). Weak, yet reproducible transcriptional induction was observed with the smallest AP-1 containing collagenase-1 promoter construct, p-72CAT (number 14). CAT activity from this construct was also stimulated to a similar degree in PMA-treated stably transfected U937 cells (data not shown). Thus, the AP-1 site, in the absence of upstream sequences, was sufficient for minimal PMA responsiveness. However, because the level of induction observed with p-72CAT (number 14) is extremely low compared with most other constructs, other upstream elements are needed for full activation of collagenase-1 transcription by PMA differentiation. Although mutation of the AP-1 site, in the context of the full-length promoter (p-2278MCAT, number 2) resulted in a loss of detectable PMA responsiveness in transiently transfected U937 cells, this construct was induced in stably transformed cells (Ϫ2278M; Fig. 3). Hirt extraction (47) verified that p-2278MCAT entered transiently transfected cells with the same efficiency as other constructs (data not shown). The seemingly contradictory transfection data with construct p-2278MCAT (Figs. 3 and 7) can be reconciled by the greater sensitivity inherent in the use of stable lines versus transient transfections.
To further characterize the upstream regions of the collagenase-1 promoter (Ϫ2278 to Ϫ997), we constructed a series of heterologous promoter constructs containing various fragments of the distal collagenase-1 promoter linked to the TK promoter of pBLCAT2 (Fig. 1B). CAT activity from the TK promoter of pBLCAT2 was not changed by PMA treatment of U937 cells (Figs. 7 and 8). Constructs p-2278TKCAT and p-2010TKCAT clearly responded to PMA (Fig. 8). Like construct p-2278MCAT, these constructs conferred PMA responsiveness in the absence of the proximal collagenase-1 AP-1 element. However, once sequences between Ϫ2010 and Ϫ1954 were deleted, the heterologous constructs (p-1954TKCAT to p-1197TKCAT) responded weakly or not at all to PMA treatment (Fig. 8).

C/EBP-␤ Is Present in Nuclear Extracts of U937 Cells and Interacts with Collagenase-1 Promoter Sequences between
Ϫ2010 and Ϫ1954 -The data with the mutant AP-1 construct (Fig. 3, Ϫ2278M) and the drop in PMA-mediated transactivation between p-2010TKCAT and p-1954TKCAT (Fig. 8) suggest the existence of functional AP-1-independent element between Ϫ2010 and Ϫ1954 of the collagenase-1 promoter. We inspected this region of the promoter for known transcription factor DNA-binding elements. A putative C/EBP-binding site (TTAG-GCAATT) and NF-B-like site (GGCAATTTCC) were identified between Ϫ2013 and Ϫ1990. Because the C/EBP family of transcription factors can regulate cellular differentiation (58), we looked for the presence of C/EBP proteins in U937 nuclear extracts. Immunoblotting of nuclear extracts with a pan-C/EBP antibody detected only one band of about 42 kDa (Fig. 9). Detection with a specific antibody verified that this band was C/EBP-␤ (Fig. 9). Furthermore, these analyses demonstrated that the relative abundance of C/EBP-␤ in U937 nuclear extracts increased with time of PMA treatment. The C/EBP-␤specific antibody detected a doublet in which the upper band may be the phosphorylated form of the lower band (59).
To determine if C/EBP-␤ could bind the sequences between Ϫ2013 and Ϫ1990 of the collagenase-1 promoter, we performed gel shift and supershift analyses with a double-stranded oligomer encompassing this region (Fig. 10). Gel shift analysis demonstrated that a nuclear factor in untreated and PMA-  7. Deletion analysis reveals that upstream promoter elements are needed to mediate full induction of collagenase-1 in response to PMAdifferentiation. A, shown are representative CAT assays for each deletion and mutant construct used. U937 cells were transiently transfected with various constructs (see Fig. 1), and half of the cells were treated with PMA for 24 h. CAT activity in lysates (50 g) was assessed for untreated (Ϫ) and PMA-treated (ϩ) cells. All procedures and manipulations were identical for each construct series. In this figure, each construct is numbered 1-16 corresponding to the number below the histograms. AP is pAPCAT2a, which contains 3 tandem repeats of the human collagenase-1 AP-1 site upstream of the thymidine kinase promoter, and BL is the parental plasmid pBLCAT2. B, for each CAT assay, the percent acetylation was determined by scintillation counting, and these data are shown in the upper histogram. The relative stimulation (fold change) of promoter construct activity in response to PMA is shown in the lower histogram. A fold change equal to 1.0 indicates no difference in CAT activity between untreated and PMA-treated samples. The results shown are the mean Ϯ S.E. of four to six determinations for each construct. differentiated cells bound this sequence (Fig. 10, left). In agreement with the immunoblotting data, the quantity of shifted probe increased with time after PMA treatment. To identify which site (C/EBP or NF-B-like) in this region bound the nuclear factors, we competed binding with wild-type (W) or mutant (M) C/EBP consensus oligomers or with a wild-type NF-B consensus oligomer. The wild-type C/EBP consensus oligomer competed nuclear factor binding to the collagenase-1 sequence probe, whereas mutant C/EBP (Fig. 10, left) or wildtype NF-B oligomers did not (data not shown). In addition, double-stranded oligomers to other regions of the sequences between Ϫ2010 and Ϫ1954 (probes 2 and 3 in Fig. 10) did not shift when incubated with nuclear extracts from control or PMA-treated cells (number 3, data not shown) or had only weak binding activity which was not modulated by PMA treatment (number 2, data not shown). Supershift analysis confirmed that C/EBP-␤ binds the collagenase-1 sequence between Ϫ2013 and Ϫ1990. Both pan-C/EBP and C/EBP-␤-specific antibodies caused a supershift of the bound probe, and the intensity of the super-shifted band increased after PMA treatment (Fig. 10, right). Thus, these results suggest that C/EBP-␤, but not NF-B, is the factor with increased activity in PMA-differentiated cells that binds the collagenase-1 promoter between Ϫ2013 and Ϫ1990.

DISCUSSION
Data presented here, as well as in other studies (27,34), show that once collagenase-1 production is induced in macrophages, enzyme expression remains active for days. In this report, we characterized regions of the collagenase-1 promoter which are involved in both activation and maintenance of collagenase-1 transcription during and subsequent to U937 differentiation. Our findings, in agreement with others (30,31), indicate that the proximal AP-1 element is necessary but not sufficient to confer maximal transcriptional activation of collagenase-1. We also report that mutation of the AP-1 element reduces and delays but does not eliminate maintained collagenase-1 promoter induction in U937 cells or fibroblasts. The sustained nature of the delayed, AP-1-independent induction of collagenase-1 suggests that the mediating factor(s) may be important in regulating maintained collagenase-1 expression by macrophages actively involved in tissue remodeling events associated with inflammation. As is discussed, C/EBP-␤ may mediate the AP-1-independent response and maximize AP-1dependent responses in differentiated monocytes.
Studies on the collagenase-1 promoter have concluded that the proximal AP-1 site is a key element necessary for rapid and full stimulation of gene transcription, regardless of the cell type or stimulus used. Typically, c-Jun/c-Fos heterodimers are believed to stimulate collagenase-1 transcription by binding the proximal AP-1 element after PMA treatment (60). Although Jun family homo-or heterodimers can bind the collagenase-1 AP-1 site to activate transcription, the affinity of such Jun dimers for the AP-1 site is far weaker than that of the corresponding Jun/c-Fos heterodimers (29,61). We find that c-Fos is the only Fos family protein expressed in U937 nuclei after PMA differentiation (Fig. 5) and that c-Fos is a component of the observed AP-1 binding complexes (Fig. 6). This observation strongly suggests a prominent role for c-Fos in regulating collagenase-1 transcription in monocytic cells, but the identity of the corresponding Jun family member is less apparent. Although our Western data indicate abundant c-Jun protein in PMA-differentiated U937 cells (Fig. 5), our supershift analyses demonstrated weak activity for c-Jun yet strong binding for JunD (Fig. 6). Angel and Karin (60) showed that expression of c-Jun, but not JunB or JunD, is necessary for collagenase-1 activation by PMA in various cell types. Therefore, although a role for JunD cannot be excluded, we suggest that c-Jun/c-Fos heterodimers are involved in induction of collagenase-1 expression in U937 cells and stimulated monocytes.
Comparable to findings in HeLa cells (30), but in contrast to those in fibroblasts (31), we found that the AP-1 element confers a minimal response to PMA in U937 cells (Fig. 7B). However, in association with the AP-1 site, the region between Ϫ179 and Ϫ95 of the collagenase-1 promoter was needed for strong PMA responsiveness in U937 cells (Fig. 7B). In fibroblasts, PMA-and oncogene-mediated transactivation of collagenase-1 is enhanced when a complete "12-O-tetradecanoylphorbol-13-acetate/oncogene responsive unit," polyoma enhancer A-binding protein-3, and AP-1 elements in tandem, is present in promoter constructs (32). In monocytic cells, the polyoma enhancer A-binding protein-3 site (Ϫ91 to Ϫ83) may not be critical for up-regulating collagenase-1 expression since only a small difference in promoter induction is seen between constructs p-95CAT and p-72CAT, with both responses being relatively weak (1.5-2-fold; Fig. 7B). Indeed, other studies have shown that the 12-O-tetradecanoylphorbol-13-acetate/oncogene responsive unit is not sufficient to maximally stimulate collagenase-1 transcription in response to PMA (30,31). In contrast to the weak induction of p-72CAT and p-95CAT, p-179CAT confers much of the PMA responsiveness (5-fold) observed in U937 cells (Fig. 7B). In the analogous region of the rabbit collagenase-1 promoter, a "TTCA" element (Ϫ105 to Ϫ100) and less characterized sequences at Ϫ182 to Ϫ161 are necessary to confer strong PMA responsiveness in fibroblasts (31,62). There is extensive homology (95-100% identical in the areas mentioned) between the human and rabbit promoters within this downstream region. This fact, together with the decreased PMA-mediated induction observed when sequences Ϫ179 to Ϫ95 are deleted (Fig. 7B) suggests the possibility that factors binding these sequences may play some role in activating collagenase-1 expression in monocytic cells. Nonetheless, although this downstream region conferred much of the PMA responsiveness, distal upstream promoter elements were needed to achieve a maximal response (Fig. 7B).
Consistent with the kinetics of c-Fos and c-Jun expression, the AP-1 element is needed for rapid and strong PMA-mediated induction of collagenase-1 transcription in U937 cells. However, our data indicate that, distal upstream elements are required to maximize AP-1-dependent induction and can induce and maintain collagenase-1 expression in an AP-1-independent manner. Most compelling is that the AP-1 mutant construct, p-2278MCAT, displayed delayed, yet maintained, PMA-mediated activation in U937 cells and fibroblasts (Figs. 3 and 4). While Buttice et al. (63) showed that mutation of the analogous AP-1 site in the related stromelysin-1 promoter did not fully eliminate PMA responsiveness in fibroblasts, to our knowledge this is the first report of AP-1 independent PMAmediated collagenase-1 promoter induction in any cell type. We detected no CAT activity in PMA-treated U937 cells transfected with either a Ϫ511 or Ϫ179 deletion construct containing the AP-1 mutation (26). Similarly, Jonat et al. (64) found no evidence of PMA-mediated induction of a Ϫ517/ϩ63 collagenase-1 promoter fragment containing an AP-1 mutation when assayed in HeLa cells. Thus, the delayed AP-1-independent response is likely controlled by a regulatory element(s) upstream of position Ϫ517.
While AP-1 factors are needed for the rapid and strong induction of collagenase-1 in U937 cells (Fig. 3), our data suggest that cooperation between distal upstream and downstream factors maintains gene expression over extended periods. We observed that collagenase-1 promoter activity was suppressed by sequences between Ϫ1954 and Ϫ179 (Fig. 7). This suppression was not seen with p-2278CAT, p-2010CAT, and p-1197⌬-997CAT, nor was it seen with smaller constructs (p-179CAT, p-95CAT, and p-72CAT). Furthermore, the p-2010⌬-1954CAT internal deletion construct had diminished FIG. 10. C/EBP-␤ binds to an upstream element of the collagenase-1 promoter. Nuclear extracts were isolated from U937 cells treated with PMA for 0, 4, 8, or 24 h and incubated with a radiolabeled, double-stranded oligomer containing the collagenase-1 C/EBP and NF-B-like sites (region 1 in schematic). Left panel, free probe (FP) was shifted (S) by proteins in nuclear extracts from control (0) U937 cells, and binding activity was increased in extracts from PMA-treated cells. Binding was competed by co-incubating extracts and probe with a 50-fold molar excess of unlabeled, wild-type C/EBP consensus (W). Binding was not decreased by co-incubation with a mutant C/EBP sequence (M) or with a wild-type NF-B consensus oligomer (not shown). Weak, nonmodulated, and no binding activity were detected with radiolabeled, double-stranded oligomers to regions 2 and 3, respectively (data not shown). Right panel, addition of either a pan-C/EBP or C/EBP-␤-specific antibody to the binding reactions resulted in a supershift (SS) of the shifted probe (S). Identical results were obtained with a second set of extracts.
transcriptional activity relative to p-2278CAT. Thus, we suggest that a suppressive element is located between Ϫ511 and Ϫ179 and that factors bound between Ϫ2010 and Ϫ1954 overcome this suppression to enhance AP-1 dependent responses. Imai et al. (65) have proposed a model in fibroblasts in which the collagenase-1 promoter is brought into an active conformation by the interplay of regulatory factors which bind elements between Ϫ1705 and Ϫ1595. Similarly, we speculate that factors bound between Ϫ2010 and Ϫ1954 might help maintain collagenase-1 expression in PMA-differentiated U937 cells by interacting with downstream factors to overcome the potential inhibitory effect of the intervening sequences.
Heterologous promoter constructs p-2278TKCAT and p-2010TKCAT were stimulated by PMA differentiation of U937 cells independent of any downstream collagenase-1 AP-1 site, with the majority of this response lost once sequences Ϫ2010 to Ϫ1954 were deleted (Fig. 8). The proserpine-binding site located at Ϫ1704 to Ϫ1689 (65) may function in monocytic cells because weak PMA responsiveness is lost when this sequence is deleted from heterologous promoter constructs (Fig. 8). However, this effect is minimal compared with the decreased PMA responsiveness caused by deletion of sequences between Ϫ2010 and Ϫ1954 from heterologous promoter constructs. Notably, no PMA responsiveness from heterologous constructs containing upstream collagenase-1 promoter regions was observed in HeLa cells (30). In the context of the wild-type collagenase-1 promoter, sequences between Ϫ2010 to Ϫ1954 were necessary to achieve maximal collagenase-1 promoter induction in U937 cells (Fig. 7B). In addition, deletion of this region caused decreased collagenase-1 promoter activity in untreated and PMAdifferentiated U937 cells (Fig. 7B). In contrast, deletion of Ϫ2010 to Ϫ1954 did not decrease basal collagenase-1 promoter activity when assayed in fibroblasts (65). We identified a putative C/EBP-binding site (Ϫ2006 to Ϫ1997) within this distal promoter region. Importantly, C/EBP-␤ protein capable of binding this collagenase-1 promoter site is present in untreated U937 nuclear extracts, and levels are increased in PMA-differentiated U937 cells (Figs. 9 and 10). Together, these data suggest that C/EBP-␤ may mediate, in part, the observed AP-1-independent activation and maintenance of collagenase-1 expression and specifically enhance AP-1 dependent responses in monocyte/macrophage cells.
The C/EBP family consists of five proteins (␣, ␤, ␦, ␥, and CRP-1) which are bZIP transcription factors that form homoand heterodimers to bind the consensus sequence (NT(T/ G)NNGNAA(T/G)) (66). Although C/EBP-␤ is found in many tissues (67), it seems to play a prominent role in activating and regulating gene expression in monocytes and macrophages (59,68,69). C/EBP-␤ expression is induced during later stages of monocytic, but not granulocytic differentiation (69), is constitutively low in monocyte/macrophages (68,70), and is strongly stimulated in macrophages by inflammatory mediators, such as lipopolysaccharide (71). Therefore, C/EBP-␤ may be a necessary factor for normal monocyte/macrophage development and function. Notably, C/EBP-␦, which forms heterodimers with C/EBP-␤ to synergistically activate transcription, is also strongly induced in monocytic cells by lipopolysaccharide (72), an agent that potently stimulates collagenase-1 expression in macrophages (27). In addition, C/EBP-␤ and C/EBP-␦ expression is stimulated or induced in various other tissues by inflammatory mediators such as interleukin-1, interleukin-6, and tumor necrosis factor-␣ (71,72). Thus, while our data suggest that C/EBP-␤ is involved in inducing collagenase-1 expression in monocytes, C/EBP-␤ may also be involved in activating collagenase-1 transcription in other cell types once C/EBP-␤ expression has been induced by inflammatory factors.