trans-Retinoic Acid Blocks Platelet-derived Growth Factor-BB-induced Expression of the Murine Monocyte Chemoattractant-1 Gene by Blocking the Assembly of a Promoter Proximal Sp1 Binding Site*

Proper regulation of the CC chemokine MCP-1 (monocyte chemoattractantprotein-1) is important for normal inflammatory responses. MCP-1 is regulated by a wide variety of agents, including platelet-derived growth factor-BB (PDGF-BB) and tumor necrosis factor-α (TNF). Using both in vivo and in vitro assays, the elements required for expression between these two cytokines were compared. In vivo genomic footprinting showed that PDGF-BB induction occurred through the occupancy of the proximal regulatory region, and unlike TNF induction, no changes in the NF-κB binding, distal regulatory region occurred. Treatment of cells with trans-retinoic acid, an inhibitor of PDGF-BB activity, resulted in a 50% reduction in PDGF-BB-mediated induction and a concomitant block in the assembly of the proximal regulatory region.trans-Retinoic acid had minimal effect on TNF induction or promoter occupancy. An inhibitor of histone deacetylation was found to stimulate expression of MCP-1 in a manner that correlated with increased accessibility to the proximal regulatory region. These results show that the mechanisms of PDGF-BB and TNF activation ofMCP-1 are distinct, although they both require the proximal regulatory region Sp1 binding site. The results also suggest that part of the mechanism used by both of these cytokines involves a process that regulates transcription factor access to the regulatory regions.

Toward this goal, we and others have identified several regulatory regions that are important for MCP-1 expression (11,(23)(24)(25)(26). Using in vivo genomic footprinting (IVGF) protocols to map regions of factor occupancy and changes in factor occupancy that result from TNF induction in vivo, a distal and a proximal regulatory region separated by 2.1 kilobase pairs of DNA were described (11). The distal regulatory region contains four elements: site A, B1, B2, and HS. Site A, the most 5Ј of these sites, was constitutively occupied in vivo and required for maximum induction of the MCP-1 gene by TNF (11). The protected and hypersensitive banding pattern that defined site A did not change in response to TNF. Two NF-B binding sites, B1 and B2, surrounded a region (HS) that became hypersensitive to dimethylsulfate treatment upon TNF stimulation. The two B sites and the HS were required for TNF induction and could function as an independent TNF-responsive element (11,23). The proximal region contains multiple elements as well and was required for stress-induced expression of MCP-1 in cells derived from human, mouse, and rat tissues (11,16,27). During TNF induction, three elements within this region become occupied: a GC box, site B, and B3. The GC box is essential for TNF-mediated induction of MCP-1 (11,23). Mutagenesis of site B and B3 revealed that these sites were not essential for TNF induction (23). A functional interferon-␥ activated site (13,14) has been described in the human MCP-1 gene, which overlaps the B3 site of the murine MCP-1 gene. A potential AP-1 binding site located just 5Ј to the GC box has also been described, but a role for this site and AP-1 in this system is not clear (16,27,28).
The MCP-1/JE gene was identified originally because of its induction by the cytokine PDGF-BB (12,29). Using heterologous expression vectors and BALB/3T3 cells, PDGF-BB induction was mapped to a 240-base pair region encompassing the MCP-1 distal regulatory region (25,26), suggesting that similar elements may control TNF and PDGF-BB regulation. In this report, we have examined whether similar mechanisms and elements controlled TNF and PDGF-BB induction of MCP-1. Three experimental approaches were employed: RNA expression analysis, IVGF, and in vitro DNA-protein binding assays. In contrast, to some previously reported results (25,26), we found that PDGF-BB stimulates MCP-1 through the proximal regulatory region and that the distal B elements were not involved. PDGF-BB induction of MCP-1 directed the assembly of the proximal region, including the GC box region. Both Sp1 and Sp3 were found to interact with this sequence in vitro. trans-Retinoic acid (TRA), an agent that inhibits PDGF-BB activity, was found to block Sp1/Sp3 assembly in vivo. With the exception of a slight reduction in total mRNA, TNF induction was unaffected by TRA treatment, including the occupancy of the proximal regulatory region. Treatment of cells with the histone deacetylase inhibitor, trichostatin A (TSA), resulted in an immediate but moderate increase in MCP-1 mRNA that correlated with an induced occupancy of the proximal regulatory region. These data divide the various stimuli of MCP-1 induction into distinct groups and show that the mechanism(s) by which PDGF-BB and TNF induce MCP-1 are distinct but are likely to share some elements. Additionally, our data suggest that both cytokines increase the accessibility of the DNA to transcription factor binding and assembly.

MATERIALS AND METHODS
Cells and Reagents-NIH3T3 and BALB/3T3 clone A31 murine fibroblasts were obtained from the American Type Culture Collection. Embryonic fibroblast cell lines containing targeted disruptions of the NF-B subunits p50 and p65 were provided by Drs. A. Hoffmann and D. Baltimore (Massachusetts Institute of Technology, Cambridge, MA). Unless otherwise indicated, all cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% bovine calf serum (Hyclone, Inc., Logan, UT), 1 mM glutamine, and antibiotics. PDGF-BB (Roche Molecular Biochemicals) and human recombinant TNF-␣ (Genzyme, Inc., Cambridge MA) were used at 30 ng/ml and 500 units/ml, respectively, for the time indicated. TRA (Sigma) was prepared as a 10 mM stock in Me 2 SO and stored at Ϫ20°C.
RNA Analysis-Cells were split and grown to approximately 85% confluency. As described by others (25,26), serum was reduced prior to the addition of PDGF-BB and TRA; this resulted in consistent results and responses to PDGF-BB and TRA. BALB/3T3 cells were transferred to media containing 0.5% serum for 24 h prior to the addition of TNF, PDGF-BB, TRA (10 M), or TSA (50 ng/ml) for the time indicated in the experiment. These concentrations were determined empirically to produce the maximal effect. Total RNA was isolated from cells using the Nonidet P-40 lysis method as described previously (11). 12 g of RNA were separated on denaturing formaldehyde gels, blotted, and probed with random primed labeled cDNA encoding MCP-1 or glyceraldehyde dehydrogenase (GAPDH) following standard procedures as described previously (11,23). Northern blots were recorded on film as well as processed by PhosphorImager analysis.
In Vivo Genomic Footprinting-IVGF assays were done exactly as described (11,23). IVGF assays of the coding strand of the distal regulatory region and the noncoding strand for the proximal regulatory region are shown because these are the most informative strands. In all cases tested, in vivo footprints obtained from the other strands produced the appropriate pattern (11). Only BALB/3T3 cells were used for these experiments because NIH3T3 cells have sequence polymorphisms in one of their MCP-1 promoter alleles that prevent the reading of the sequence pattern (11,23).
NIH3T3 cells were transiently transfected by electroporation (Bio-Rad) with the indicated CAT reporter constructions as described previously (11,23). BALB/3T3 cells were not used for transient transfections because of their poor transfection efficiency (11,23,26). As a control for transfection efficiency between samples, all transfections contained 1 g of pSV 2 AlkPhos, an alkaline phosphatase reporter vector. At 28 h post transfection, PDGF-BB was added. At 36 h, the cultures were collected and assayed for CAT protein by enzyme-linked immunosorbent assay (Roche Molecular Biochemicals) and for alkaline phosphatase activity using a kit from Bio-Rad. Assays were normalized to their alkaline phosphatase reporter activity. The average of three transfections are shown.
EMSAs-Nuclear extracts were prepared from twenty 10-cm plates according to the procedures previously described (30 -32). DNA binding reactions were performed in a solution containing 4 g of nuclear extract, 0.6 g of poly(dI⅐dC):poly(dI⅐dC), 250 ng of denatured sonicated salmon sperm DNA, 15 mM HEPES (pH 7.9), 10% glycerol, 50 mM KCl, 0.12 mM EDTA, 5 g of bovine serum albumin, 12 mM dithiothreitol, and 5 mM MgCl 2 for 30 min on ice. The sequence of the coding strand of MCP-1 GC box specific probe used in the EMSAs was 5Ј-GCACCCT-GCCTGACTCCACCCCCCTGGCTTACAA. Competitor DNA for consensus Sp1 and AP-1 sites were 5Ј-ATTCGATCGGGGCGGGGCGAG and 5Ј-CGCTTGATGACTCAGCCGGAA (Santa Cruz Biotechnology, Valencia, CA), respectively, where the underlined bases represent the consensus sites. Nonspecific DNA competitor encoded Site A of the MCP-1 gene: 5Ј-AGAACTGCTTGGCTGCAGGCCCAGCATCTGGAGC-TCACATT. Sp1 and Sp3 antibody supershift assays were carried out by adding 1 l of antibody (Santa Cruz Biotechnology) to the reaction mixture 5 min prior to the addition of the labeled probe. Samples were separated by electrophoresis in a 5% polyacrylamide gel (49:1::acrylamide:bis) at 4°C with recirculating buffer and analyzed by autoradiography.

RESULTS
Mechanistic studies of MCP-1 gene in human, mouse, and rat demonstrated that MCP-1 expression is transcriptionally regulated through a distal regulatory region and a proximal regulatory region (11,23,24,26,33). Systematic analysis of the distal regulatory region identified an NF-B-dependent enhancer that was required for TNF induction of the MCP-1 gene in human and mouse cells (23,34). Although the distal element could enhance the activity of a heterologous promoter in response to TNF, the proximal GC box was essential for activity in the context of the MCP-1 gene (23). Additionally, inhibitors of translation induced MCP-1 RNA expression through a mechanism that involved the proximal but not the distal-regulatory elements (11). Thus, it appeared as if the proximal region might be capable of independent activity. Previous analysis of the PDGF-BB induction of the MCP-1 gene suggested that the distal regulatory region was required for PDGF-BB mediated induction (25,26). Because PDGF-BB signals through a pathway distinct from TNF (35,36), experiments were conducted to determine whether each of these cytokines used the same cisacting control elements to induce the expression of the MCP-1 gene.
PDGF-BB Induces MCP-1 through Changes in Occupancy of the Proximal Regulatory Region-To compare the induction of MCP-1 mRNA by TNF to that by PDGF-BB, murine BALB/3T3 fibroblasts were treated with either cytokine for 4 h. RNA was isolated and probed for the expression of the MCP-1 and GAPDH genes using a Northern blot assay. Both PDGF-BB and TNF were able to stimulate the induction of the MCP-1 gene in BALB/3T3 cells by 14-and 20-fold over untreated cells, respectively (Fig. 1A). To map PDGF-BB-inducible elements and compare them to those used during TNF induction, IVGF of the proximal (Fig. 1B) and distal (Fig. 1C) regulatory regions in both treated and untreated control cells was carried out. IVGF provides a snapshot of sites that are protected or altered in vivo from dimethylsulfate attack during the short treatment time (2 min). In vivo footprints and hypersensitive sites have been shown to correspond to sites of transcription factor occupancy (11,(37)(38)(39). In vitro prepared DNA was processed to provide a no protein/control pattern (Fig. 1, B and C, lane V). Following PDGF-BB (1 h) or TNF (30 min) treatment, full occupancy of the proximal region site B and the GC box occurred (Fig. 1B). In the distal regulatory region, site A was fully occupied in control cells and was unchanged during the PDGF-BB or TNF treatment. The occupancy of site A was previously shown to be unaltered during TNF treatment (11) and most likely does not change during induction of MCP-1 under any conditions. In contrast to TNF-treated cells, no occupancy was detected in or around the sequences that encode the two NF-B binding sites in cells treated with PDGF-BB.
These data suggest that 1) the distal B elements play no role in PDGF-BB induction and 2) the proximal regulatory region alone may contain elements required for PDGF-BB regulation.
TRA was shown previously to inhibit induction of MCP-1 in osteoblasts (19). To determine whether this effect was through the proximal or distal region, cells were pretreated with TRA and then with TNF or PDGF-BB and analyzed for expression of MCP-1 by Northern blot (Fig. 1A). Changes influenced by TRA in the in vivo occupancy of the MCP-1 regulatory regions were monitored by IVGF (Fig. 1, B and C). TRA treatment had no effect on the basal level of MCP-1 and did not affect the occupancy of site A in the distal region. TRA only slightly inhibited TNF-induced MCP-1 expression, reducing the induction from 20-to 17-fold (Fig. 1A). However, TRA did inhibit the PDGF-BB-induced induction of MCP-1 mRNA by almost 50% (Fig.  1A). A concomitant block in the PDGF-BB-induced occupancy of the proximal GC box and site B (Fig. 1, B and C) was observed, suggesting that TRA has a direct effect on the ability of PDGF-BB-activated transcription factors to assemble on the MCP-1 proximal promoter. A slight decrease in the delineation of the TNF-induced footprint in the proximal region was also observed in the TNF/TRA-treated samples (Fig. 1B, lane 3).
Thus, it appears as if TNF induction can bypass the TRA inhibition of the assembly of the proximal regulatory region. It is possible that the TNF activation of NF-B and assembly of the distal regulatory region may allow stable complexes to form between proximal and distal regulatory region factors that make the proximal region resistant to the effects of TRA action.

NF-B Is Not Required for PDGF-BB-mediated Induction of MCP-1-Previous regulatory region analyses and in vitro gel
shift assays suggested that a region encompassing the distal regulatory sequences used two NF-B sites and another factor to regulate PDGF-BB induction of MCP-1 in murine fibroblasts (25,26). Because these results are in conflict with the data presented above, experiments were carried out to assess the involvement of the major NF-B transcription factor proteins, p50 and p65, in PDGF-BB induction of MCP-1. Murine embryonic fibroblasts derived from p50 and p65 homozygous knockout animals were assessed for their ability to induce MCP-1 in response to PDGF-BB. Northern blot analysis showed that MCP-1 was induced by PDGF-BB equally well in p50 Ϫ/Ϫ and p65 Ϫ/Ϫ cells ( Fig. 2A). PDGF-BB induction in these cells is comparable with that seen in BALB/3T3 cells and NIH3T3 cells. Our previous analysis of these cell lines showed that p65 but not p50 was necessary for TNF induction of MCP-1 (23). To determine whether induction of cells by PDGF resulted in NF-B binding to the distal regulatory region, an EMSA was performed using the B-2 sequence as a probe (Fig. 2B). Extracts from BALB/3T3 cells treated with TNF showed strong induction of NF-B binding to this site, whereas PDGF treatment shows only weak binding of a different set of complexes, none of which are supershifted by an anti-NF-B p65 specific antiserum. These results are consistent with the IVGF data presented in Fig. 1 and demonstrate that the predominant cellular NF-B subunits are not required for PDGF-BB induction of MCP-1. These data indicate that the proximal regulatory region responds to PDGF-BB signaling in a manner that is independent of NF-B and the distal regulatory region.
The GC Box in the Proximal Regulatory Region Is Required for PDGF-BB Induction of MCP-1-To determine whether the proximal region could respond to PDGF-BB independently of the distal sequences, a series of transfections were carried out using plasmids containing just the proximal region or the distal and proximal sequences combined (Fig. 3). The results of these experiments showed that constructions containing either the proximal regulatory region alone or both the proximal and distal regulatory regions were induced by PDGF-BB to similar levels. Therefore, the proximal regulatory region alone was sufficient to regulate MCP-1 expression by PDGF-BB. A GC box, an AP-1 site, an interferon-␥-activated site (B3), and site B were identified in the proximal regulatory region (11,13,23).
To determine the role of the GC box, substitutions were intro- duced that altered all the nucleotides within the site. This mutation ablated PDGF-BB-induced expression in a transient transfection assay (Fig. 3). A similar observation was made for TNF-induced expression (23). The above analysis also suggests that sequences between the distal and proximal regulatory regions are not required for MCP-1 induction by PDGF-BB because no significant differences were observed between the three constructions containing 2.6, 2.4, or 0.3 kilobase pairs of 5Ј-flanking DNA (Fig. 3). Thus, the proximal region is necessary and does not require the distal regulatory region for PDGF-BB-mediated induction of MCP-1. Moreover, both TNFand PDGF-BB-mediated induction of MCP-1 requires the proximal GC box.
Both Sp1 and Sp3 Bind to the GC Box-To identify the factors that bind to the GC box and to determine whether there were differences between PDGF-BB-and TNF-induced cells, EMSAs were carried out using a probe encompassing the GC box. The GC box of MCP-1 is juxtaposed downstream of a site that has a 1-base pair mismatch from a consensus AP-1 binding site sequence. Despite the sequence similarity to an AP-1 site, no in vivo binding to this site was observed in either TNFtreated (11) or PDGF-BB-treated cells (Fig. 1B). Nuclear extracts prepared from untreated cells, PDGF-BB-induced cells, and TNF-induced cells were assayed for their activity on the MCP-1 GC box probe, which contains both the GC box and the AP-1 site. Similar protein-DNA complexes were detected among all the extracts tested. Three specific bands were detected with nuclear extracts from either the control, TNFtreated, or PDGF-BB-treated cells (Fig. 4A, bands a-c). Specific (MCP-1 GC box and consensus Sp1 DNA) and nonspecific DNA competition assays showed that these bands are specific for the probe (Fig. 4). The addition of Sp1-or Sp3-specific antibodies to the DNA binding reactions resulted in supershifted complexes, indicating that Sp1 is part of the DNA protein complex in band c and that Sp3 is contained in DNA   11-15). C, nuclear extracts were prepared from BALB/3T3 cells treated with PDGF-BB as above.
protein complexes in bands a and b (Fig. 4).
Several reports have suggested that AP-1 was involved in MCP-1 expression (16,19,28,40,41) and functioned through the proximal region AP-1 site. Because PDGF-BB can induce AP-1 activity, the protein DNA complexes formed on the GC box probe were assayed for c-Fos or c-Jun by EMSA supershift analysis. Extracts prepared from PDGF-BB-induced cells did not show binding of AP-1 to this site, and no supershift with c-Jun or c-Fos antiserum was observed (Fig. 4). Additional experiments using an AP-1 consensus probe showed that strong AP-1 activity was indeed induced by PDGF-BB and that this activity could be competed by consensus AP-1 DNA (Fig.  5). The activity observed with the MCP-1 GC/AP-1 box DNA was not competed by the AP-1 consensus competitor DNA. Thus, these data suggest that AP-1 is most likely not interacting with this site and that this element functions primarily as an Sp1 and/or Sp3 binding site.
To determine whether TRA affected the ability of Sp1 or Sp3 in nuclear extracts to recognize its target, nuclear extracts were prepared and analyzed from TRA-treated cells and TRA/ PDGF-BB-treated cells. No obvious changes in Sp1 or Sp3 DNA binding activity were observed in EMSAs using extracts prepared from control, PDGF-BB-, TRA/PDGF-BB-, or TNFtreated cells were detected (Fig. 4). Thus, both Sp1 and Sp3 are present in the nucleus of cells and can bind to DNA in vitro. This suggests that TRA inhibition does not function at a level that directly prevents the factors from binding DNA but rather at the level of assembly or accessibility to the MCP-1 DNA in vivo.
Activation of MCP-1 by PDGF-BB Is Not Restricted by Chromatin Structure-Because the IVGF assays showed that the GC box was unoccupied in the control cells and that occupancy of the GC box correlated directly with the MCP-1 induction by PDGF-BB, it is possible that PDGF-BB treatment opens the chromatin structure and allows Sp1/Sp3 access to the GC box, subsequently activating transcription. Acetylation and deacetylation of histone N-terminal lysine residues have been found to be associated with activation and repression of gene expression, respectively (42)(43)(44). If alteration of the chromatin structure is a mechanism involved in PDGF-BB induction of the MCP-1, TSA, a histone deacetylase inhibitor (45)(46)(47), may also be able to induce MCP-1 expression. To investigate this possibility, BALB/3T3 cells were treated with TSA with or without subsequent PDGF-BB treatment and analyzed for the expression of MCP-1 mRNA (Fig. 6). The addition of TSA alone to cell cultures resulted in a moderate 11-fold induction of MCP-1 mRNA at 8 h (Fig. 6). IVGF showed that as early as after 30 min of TSA treatment, the GC box became occupied but the distal regulatory region B sites remained unoccupied (Fig.  7). After 22 h of TSA treatment, a time point when the effects of TSA should be maximal, a footprint over the proximal region was still observed; however, its intensity was diminished significantly. The reduction in footprint correlated with the reduction in MCP-1 mRNA at 16 h (Fig. 6). PDGF-BB was able to up-regulate the TSA induction of MCP-1 at both 8 and 16 h, 15and 12-fold, respectively (Fig. 6), suggesting that PDGF-BB provides additional signals required for maximal expression. One such signal might be the recruitment of a complex with histone acetylase activity to the promoter.
To determine whether TRA could block the effects of TSA, cells were pretreated with TRA for 24 h, then treated with TSA, and examined. TRA pretreatment resulted in an inhibition of MCP-1 induction and occupancy of the GC box (Figs. 6 and 7). TRA was also able to block the combined effects of PDGF-BB and TSA on the induction of MCP-1 at 16 h and to a lesser extent at 8 h. These results imply that PDGF-BB-induced expression is controlled by a process that actively regulates the accessibility and assembly of the proximal regulatory region. DISCUSSION MCP-1 functions as a critical cytokine during inflammatory responses. Basal expression of MCP-1 is normally low, and the gene is highly inducible in many cell types (17, 48 -50). Using both in vivo and in vitro approaches, we characterized the transcriptional mechanism of PDGF-BB induction of MCP-1 and compared it to that for TNF. Although TNF requires both the distal and the proximal regulatory regions for induction, PDGF-BB requires only the proximal region. The assembly of the proximal region as assayed by in vivo footprinting is identical. Additionally, the pathway that PDGF-BB uses is sensitive to trans-retinoic acid through a mechanism that blocks assembly of the MCP-1 proximal regulatory region.
Agents that can induce MCP-1 can be divided into at least three groups. Group I comprises agents that cause stress on the cellular machinery and include cycloheximide (11), mechanical  2, 4, 7, and 9. Anti-c-Jun antiserum was added to the indicated reactions to determine whether complexes contained AP-1 (lanes 5 and 10). A supershift is shown by the arrow labeled with an asterisk. The gray bars indicate which probe was used in each experiment. stress (16), and possibly TSA. These agents stimulate the assembly of the proximal regulatory region, with a subsequent moderate level of expression. It is possible that these agents lead to the opening of chromatin and therefore accessibility to the DNA by the already present transcription factors, such as Sp1. Group II agents also use only the proximal regulatory region. These include PDGF-BB (33), transforming growth factor-␤ (40), and interferon-␥ (13,14). Like PDGF-BB, interferon-␥ stimulates the assembly of factors at the proximal region and does not require the distal regulatory region to stimulate MCP-1 expression (13). The level of induction by these agents is higher than those in group I. A 7-nucleotide motif in the 3Јuntranslated region of the MCP-1 gene was found to play a significant role in PDGF-BB-induced expression of MCP-1 (51). Perhaps the use of this sequence motif explains part of the difference between the overall levels of mRNA induced by the stimulants of groups I and II. This may also explain the lower levels of MCP-1 induction seen with PDGF-BB in the transient expression assays as compared with Northern blot assay because our constructions do not contain this motif. The induction of MCP-1 in human cells by interferon-␥ has been shown to use a STAT-like binding element that is homologous to the B-3 site (13,14). Clear occupancy of the B-3 site was not observed for PDGF-BB-treated cells (data not shown), suggesting some differences between some of the stimuli of this group. However, in all cases, the GC box appears to be critical. Because the levels of expression from group II simulators are higher than those of group I, we suggest that this group may activate new transcription factors, modify existing factors, or open chromatin in a broader context than that of group I. Support for modification of transcription factors in this system was ob-served in cells treated with inhibitors of protein kinase A and TNF (11). This combination of treatments led to the complete assembly of the MCP-1 regulatory regions but no transcription, suggesting the requirement for factor modification. TNF, IL-1, and other factors that activate gene expression through NF-B signal pathways require both the proximal and distal regulatory regions and comprise group III. In group III, the distal regulatory region coordinates NF-B binding and transcriptional activation (23,24,34,41). Group III stimulants of MCP-1 expression require the GC box (23).
Role of NF-B in PDGF-BB Induction of MCP-1-Previous studies suggested that PDGF-BB regulates MCP-1 expression through the NF-B sites of the distal regulatory region (25,26). Thus it was surprising to find that only the proximal GC box was occupied in IVGF following PDGF-BB stimulation. Although the sources of PDGF-BB and cell culture conditions may result in the different conclusions, our results were supported by several approaches. First, following PDGF-BB treatment IVGF detected the occupancy in the proximal regulatory region but not the distal regulatory region. Second, MCP-1 was induced by PDGF-BB equally well in p50 Ϫ/Ϫ and p65 Ϫ/Ϫ cells. Because the activity of the distal NF-B-enhancer is dependent on p65 (23,34), this observation suggests that PDGF-BB induction of MCP-1 is independent of the distal B sites. IVGF of both control and PDGF-BB treated p50 Ϫ/Ϫ and p65 Ϫ/Ϫ cells also showed that the distal regulatory region was not occupied after PDGF-BB treatment (data not shown). Third, transient transfections of MCP-1 promoter constructions showed no differences between constructions containing only the proximal region and those containing both the proximal and distal regions. Last, treatment of cells with TRA for 22 h resulted in an inhibition of PDGF-BB induction and a loss of occupancy of the proximal regulatory region, including the GC box. In contrast, treatment of cells with TRA and TNF resulted in no change in expression or occupancy of the proximal GC box, suggesting significant differences in the mechanisms of induction between these two cytokines. Thus, the distal regulatory region was not required for PDGF-BB-induced transcription in this expression system.
Role of Sp Family Members-To characterize the proximal regulatory region, a series of EMSAs was performed. Both Sp1 and Sp3 were found to bind with high affinity to the proximal GC box. Preliminary results using Drosophila cells, NF-B and Sp1 or Sp3 expression vectors, and an MCP-1 reporter suggest that both Sp1 and Sp3 can stimulate expression. 2 Thus, it is possible that in mammalian cells, both factors are functional. It is intriguing to speculate that the differences in expression between the three groups may be in part determined by which Sp family member binds to the proximal GC box in vivo. The development of in vivo chromatin immunoprecipitation assays (52,53) may allow this question to be addressed directly .
Role of AP-1-Juxtaposed on the 5Ј side of the GC box is an element that has an 1-base pair mismatch from a consensus AP-1 site. Despite this high degree of homology, AP-1 binding activity was not detected either by IVGF or by EMSA in the current or our previous study (11). The data in the literature are conflicting on the issue of whether AP-1 is involved in MCP-1 expression. Several reports have suggested that this is a functional AP-1 site (19,40). We have observed that mutation of the AP-1 site results in a decrease TNF-inducible MCP-1 expression (23). Curcumin, an anti-c-Jun drug, was shown to inhibit MCP-1 induction (28). However, given the lack of an in vivo footprint and in vitro DNA binding activity of the murine AP-1 site for AP-1, it is possible that TRA may inhibit MCP-1 2  expression indirectly by competing with a trans-activator (54,55), modifying the Sp1/Sp3 proteins in some novel manner, or affecting the expression of one of the MCP-1-specific transcription factors. It is also possible that a role for AP-1 is solely tissue-specific and not important in the murine fibroblast cell lines that we have examined. Recently, a report analyzed the induction of MCP-1 in c-Fos Ϫ/Ϫ targeted cell lines and found it to be normal (56), lending support to the latter statement. This does not rule out a role for c-Jun homodimers. If c-Jun is involved in this system, perhaps its involvement is through protein-protein interactions that do not involve it binding to DNA. Although this would be a novel role for c-Jun, it would explain the data.
Chromatin and Factor Assembly-Cooperative binding between different transcription factors to adjacent cis-elements can stabilize the binding of each factor and facilitate the formation of a stable transcription complex (57,58). Unlike the MnSOD gene, whose multiple GC boxes were occupied in untreated cells (59), the single Sp1/Sp3 binding site in the MCP-1 promoter was not occupied without induction. The lack of detectable occupancy at the GC box may be due to kinetically unfavorable binding of Sp1 to a single site or inhibition of binding by chromatin structure. To examine a possible role in alterations in chromatin structure, TSA, a histone deacetylase inhibitor, was used. TSA was found to induce MCP-1 at early time points and to allow factor occupancy of the proximal region within 30 min of treatment. The exact mechanism of TSA action is not known; however, it is known that TSAinduced hyperacetylation of 50% of histones requires approximately 8 h (60). If TSA randomly inhibits nucleosome deacetylation, then at early time points it is unlikely that MCP-1 would be induced by remodeling of chromatin structure in a manner that is dependent on TSA. However, it is possible that for the MCP-1 gene to be in the resting state, the nucleosomes surrounding the MCP-1 gene may be actively deacetylated, allowing TSA to interrupt this process. If this is the case, then induction of MCP-1 may require the recruitment of a coactivation complex containing histone acetyltransferase activity. The inhibition of PDGF-BB induction by TRA may be explained by this recruitment of such a complex. The nuclear hormone RAR/ RXR receptors induced by TRA have been shown to bind the PCAF coactivation complex (61). If these coactivation complexes are limiting, then it is possible that TRA treatment prevents the PDGF-BB-mediated coactivation complex from accessing the MCP-1 promoter. This scenario is also consistent with the dominant effect of TRA on TSA, because TSA has no effect on histone acetyltransferase activity complexes. Because TNF induction is not affected by TRA treatment, we suggest that TNF activation either uses a distinct coactivator complex from both the PDGF-BB pathway and/or RAR/RXR receptors or that it has a higher affinity for such complexes. Thus, the data presented here suggest that although PDGF-BB and TNF regulation of the MCP-1 gene use distinct mechanisms, both pathways are likely to involve a complex interplay between histone acetylase and deacetylases.