Stromelysin-3 induction and interstitial collagenase repression by retinoic acid. Therapeutical implication of receptor-selective retinoids dissociating transactivation and AP-1-mediated transrepression.

Human stromelysin-3 and interstitial collagenase are matrix metalloproteinases whose expression by stromal cells in several types of carcinomas has been associated with cancer progression. We compared here the regulation of the expression of both proteinases by retinoids in human fibroblasts. Physiological concentrations of retinoic acid were found to simultaneously induce stromelysin-3 and repress interstitial collagenase. In both cases, the involvement of a transcriptional mechanism was supported by run-on assays. Furthermore, in transient transfection experiments, the activity of the stromelysin-3 promoter was induced by retinoic acid through endogenous receptors acting on a DR1 retinoic acid-responsive element. The ligand-dependent activation of the receptors was also investigated by using selective synthetic retinoids, and we demonstrated that retinoic acid-retinoid X receptor heterodimers were the most potent functional units controlling both stromelysin-3 induction and interstitial collagenase repression. However, specific retinoids dissociating the transactivation and the AP-1-mediated transrepression functions of the receptors were found to repress interstitial collagenase without inducing stromelysin-3. These findings indicate that such retinoids may represent efficient inhibitors of matrix metalloproteinase expression in the treatment of human carcinomas.

Human stromelysin-3 and interstitial collagenase are matrix metalloproteinases whose expression by stromal cells in several types of carcinomas has been associated with cancer progression. We compared here the regulation of the expression of both proteinases by retinoids in human fibroblasts. Physiological concentrations of retinoic acid were found to simultaneously induce stromelysin-3 and repress interstitial collagenase. In both cases, the involvement of a transcriptional mechanism was supported by run-on assays. Furthermore, in transient transfection experiments, the activity of the stromelysin-3 promoter was induced by retinoic acid through endogenous receptors acting on a DR1 retinoic acid-responsive element. The ligand-dependent activation of the receptors was also investigated by using selective synthetic retinoids, and we demonstrated that retinoic acid-retinoid X receptor heterodimers were the most potent functional units controlling both stromelysin-3 induction and interstitial collagenase repression. However, specific retinoids dissociating the transactivation and the AP-1-mediated transrepression functions of the receptors were found to repress interstitial collagenase without inducing stromelysin-3. These findings indicate that such retinoids may represent efficient inhibitors of matrix metalloproteinase expression in the treatment of human carcinomas.
Stromelysin-3, based on sequence homologies and its domain organization, belongs to the matrix metalloproteinase (MMP) 1 family consisting of extracellular proteinases that are implicated in a variety of tissue remodeling processes. Stromelysin-3 expression has been associated with cutaneous wound healing (1), mammary gland involution (2), cycling endometrium (3), embryonic development (4), and metamorphosis (5), where its expression was predominantly found in cells of mesodermal origin. In human carcinomas, stromelysin-3 was the first MMP identified as being expressed by stromal cells (6,7). Although human stromelysin-3 appears to be unable to degrade any major component of the extracellular matrix (8,9) and exhibits unusual activation properties (10,11), its role in cancer progression is supported by high expression levels, which are predictive of a poor clinical outcome (12,13). Furthermore, we have demonstrated that stromelysin-3 facilitates the tumor take of cancer cells in nude mice (14). Following the identification of stromelysin-3, a number of other MMPs have also been found to be expressed by stromal cells of human carcinomas (15), indicating that the stromal cell production of MMPs represents a significant contribution to the overall proteolytic activities in malignant tumors (Refs. 15 and 16 and references therein). Despite the observation that most stromal MMPs are expressed by fibroblastic cells, no regulatory sequence that could account for this cell-specific expression pattern has yet been identified in the promoter of the corresponding genes.
While stromelysin-3 expression, like other MMPs, can be induced in human fibroblasts by agents such as phorbol ester (TPA) or growth factors (6,17), very little is known about the mechanisms regulating its expression. We have recently isolated the stromelysin-3 gene (18) and shown that its proximal promoter differs from those of other MMPs by the absence of a consensus AP-1 (c-Jun/c-Fos) binding site and the presence of a retinoic acid-responsive element (RARE) of the DR1 type. This RARE can be transactivated by retinoid receptors (RARs/ RXRs) in a ligand-dependent manner in COS-1 cells. In contrast, AP-1 binding sites were found to play a crucial role in controlling both the activation of other MMP gene promoters in response to growth factors and cytokines (19,20) and their inhibition by retinoic acid (RA) (20 -22). Gene transcription studies have shown that while RARs and RXRs can induce transcriptional activation through specific DNA binding sites, they can also interact indirectly with AP-1 through transcriptional mediators to repress gene transcription (23)(24)(25). In agreement with these findings, inhibition of base-line and TPAinduced RNA expression by RA has been reported for interstitial collagenase (20) and stromelysin-1 (21).
Retinoid effects are achieved through two classes of liganddependent transactivators, the retinoic acid receptors (RAR␣, * This work was supported by funds from the Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, the Centre Hospitalier Universitaire Régional, the Bristol-Myers Squibb Pharmaceutical Research Institute, the BIOMED 2 (contract BMH4CT96 -0017) and BIOTECH 2 (contract ERBBIO4CT96 -0464) programs, the Association pour la Recherche sur le Cancer, the Ligue Nationale Française contre le Cancer and the Comité du Haut-Rhin, the Fondation pour la Recherche Médicale Française, the Fondation de France, and a grant from the Fondation Jeantet (to P. Chambon). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: MMP, matrix metalloproteinase; CAT, chloramphenicol acetyltransferase; DR1, directly repeated consensus hexameric half-sites (5Ј-PuG(G/T)TCA-3Ј) separated by one base pair; DR5, directly repeated consensus hexameric half-sites (5Ј-PuG(G/T)-TCA-3Ј) separated by five base pairs; RA, retinoic acid; 9C-RA, 9-cisretinoic acid; t-RA, all-trans-retinoic acid; RARE, retinoic acid-responsive element; RAR, retinoic acid receptor; RXR, retinoid X receptor; -␤, and -␥ and their isoforms) and retinoid X receptors (RXR␣, -␤, and -␥ and their isoforms), which are members of the nuclear receptor superfamily. RARs bind and are activated by t-RA and 9C-RA, whereas RXRs bind and are activated by 9C-RA (26 -28). Retinoids are known to regulate cell proliferation and differentiation, and they are regarded as agents that may be used to prevent or suppress human cancer (29 -31). In addition, experimental and clinical studies suggest that retinoids may also be therapeutically useful in preventing connective tissue degradation caused by MMP overproduction in arthritis (32,33). However, despite extensive knowledge of RA action at the molecular level, only a few RA target genes have been identified.
In the present study, we investigate further the regulation of stromelysin-3 gene expression by RA in cells of mesodermal origin. We have found that while nanomolar concentrations of RA can induce the expression of both stromelysin-3 RNA and protein in human fibroblasts, they prevent the expression of interstitial collagenase. The involvement of a transcriptional control in RA action is supported by run-on analyses, showing that the elongation of stromelysin-3 nuclear RNA was upregulated and that of interstitial collagenase was down-regulated by RA. Furthermore, using stromelysin-3 promoter-based plasmid constructs, we showed that the stromelysin-3 promoter can be activated in cells of mesodermal origin exposed to RA, without the addition of exogenous RARs/RXRs. The observation that both induction of the stromelysin-3 gene and repression of the interstitial collagenase gene were optimally achieved by combining selective RAR agonists with a pan-RXR retinoid indicates that combination of receptors of the RAR and RXR types are required for optimal transcriptional regulation of these genes.

MATERIALS AND METHODS
Ligands-t-RA was purchased from Sigma, and 9C-RA was provided by P. F. Sorter, J. F. Grippo, and A. A. Levin (Hoffmann-La Roche, Nutley, NJ). CD666 (34) was donated by B. Shroot (Centre International de Recherches Dermatologiques Galderma, Valbonne, France). Am80 (35) was provided by K. Shudo (University of Tokyo). BMS649 (36) (originally known as SR11237) was provided by the Bristol-Myers-Squibb Pharmaceutical Research Institute (Buffalo, NY). The last retinoid BMS753 (37,38), which is a pure RAR␣ agonist, was also provided by the Bristol-Myers-Squibb Pharmaceutical Research Institute, where it is available to academic investigators upon request.
Cell Culture-Human fibroblasts (HFL1, CCL 153) and rhabdomyosarcoma tumor cell line (RD, CCL 136) were obtained from the American Tissue Culture Collection (Rockville, MD) and maintained in monolayer culture in Dulbecco's modified Eagle's medium with or without 5% calf serum. Retinoids t-RA, 9C-RA, BMS753, BMS649, Am80, and CD666 were dissolved in ethanol and added at desired concentrations for the time periods indicated in the figure legends.
Nuclear Run-on Transcription Assays-Control cells and cells treated with 9C-RA (1 M) for 1-3 days were washed twice with ice-cold phosphate-buffered saline, harvested, and centrifuged at 1300 ϫ g at 4°C for 5 min. The pellet was resuspended in 4 ml of lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.5% (v/v) Nonidet P-40), incubated for 5 min on ice, and centrifuged at 1300 ϫ g at 4°C for 5 min. This procedure was repeated twice. The final pellet containing the nuclei was resuspended in storage buffer consisting of 50 mM Tris-HCl, pH 8.3, 5 mM MgCl 2 , 0.1 mM EDTA, 40% (v/v) glycerol, and aliquots of 2 ϫ 10 7 nuclei were stored at Ϫ80°C before use. In vivo initiated RNA transcripts from these 2 ϫ 10 7 nuclei aliquots were elongated in vitro for 30 min at 30°C in the presence of 200 Ci of [␣-32 P]UTP in a final volume of 200 l containing 1 mg/ml heparin, 0.6% (v/v) sarkosyl, 0.4 mM concentrations of ATP, CTP, and GTP, 2.5 mM dithiothreitol, 0.15 mM phenylmethylsulfonyl fluoride, 350 mM (NH 4 ) 2 SO 4 . The reaction was stopped by the addition of DNase I-RNase free (800 units) in the presence of 1.8 mM CaCl 2 for 10 min at 30°C, followed by protein digestion with proteinase K (100 g/ml) in 50 mM Tris-HCl, pH 7.4, 20 mM EDTA, 1% SDS and incubation (45-90 min) at 42°C until clear samples were obtained. RNA extraction was then performed with phenol/chloroform (1:1, v/v), and the organic phase was further extracted with 10 mM Tris-HCl, pH 7.4, 5 mM EDTA, 1% SDS. Pooled aqueous phases were finally extracted with chloroform, and RNA precipitation was carried out at 4°C for 15 min after the addition of 1 volume of 20% trichloroacetic acid in the presence of 20 g of tRNA as a carrier. RNA pellets were washed 3 times in 5% trichloroacetic acid and once with 80% ethanol. Dried pellets were then dissolved in hybridization buffer (as described above) to a final specific activity of 5 ϫ 10 6 cpm/ml and hybridized to cDNAs corresponding to human stromelysin-3 (ZIV in Ref. 18), human interstitial collagenase (41), 36B4 (42), and the pBluescript II SKϩ plasmid. These DNAs were denatured in the presence of 0.3 N NaOH and immobilized onto Hybond nylon membranes (Amersham) by using a slot blot apparatus. Prehybridization at 42°C for 18 h and hybridization to in vitro 32 P-labeled elongated RNAs at 42°C for 3 days were carried out in the same hybridization buffer. Filters were subjected to various washing conditions as follows: twice in 2 ϫ SSC, 1% SDS for 15 min at 22°C; twice in 0.1 ϫ SSC, 0.1% SDS for 15 min at 52°C; once in 2 ϫ SSC, in the presence of RNase A (10 g/ml) for 15 min at 37°C; twice in 2 ϫ SSC, 1% SDS for 15 min at 22°C; and finally, once in 0.1 ϫ SSC, 0.1% SDS for 15 min at 52°C. Signal quantification was carried out as described for Northern blot analysis.
Cell Transfection and CAT Assay-Human RD rhabdomyosarcoma cells were transiently transfected by the calcium phosphate procedure as described previously (18), except that the total amount of DNA transfected in each 10-cm diameter culture dish was made up to 20 g with pBluescribe plasmid DNA. For a 4-day treatment with RA, cells were first exposed to 1 M 9C-RA for 2 days before transfection, whereas for a 2-day 9C-RA treatment, cells were directly transfected at 4 h after plating. In both cases, cells were incubated in the presence of 1 M 9C-RA for 2 days after transfection. The ␤-galactosidase expression vector pCH110 (Pharmacia) was used as a internal control to normalize for transfection efficiency. Cell extracts containing 4 units of ␤-galactosidase activity were used for chloramphenicol acetyltransferase (CAT) assays, and the reaction products were separated by thin layer chromatography and visualized by autoradiography. Signal quantification was performed as described for Northern blot analysis.

Stimulation of Stromelysin-3 and Inhibition of Interstitial Collagenase RNA Expressions by Retinoic Acid in Fibroblasts-
Having previously identified a RARE that conferred ST3 promoter inducibility in COS-1 cells in the presence of RA and its receptors (18), we decided to evaluate whether stromelysin-3 gene expression was also regulated by RA in human fibroblasts. Time course and dose response experiments were performed, and expression of the stromelysin-3 gene was compared with that of interstitial collagenase by Northern blot analysis in HFL1 fibroblasts exposed to 9C-RA in the presence of 5% calf serum (Fig. 1).
As shown in Fig. 1A, in the presence of 1 M 9C-RA, stromelysin-3 RNA levels progressively increased from day 1 to day 4, with a 20-fold increase measured after 4 days of incubation. In contrast, the levels of interstitial collagenase RNA remained constant when fibroblasts were exposed to 9C-RA for 1 day and rapidly decreased to almost undetectable levels after 2 days of treatment. Dose response experiments were conducted after incubation during 3.5 days with 9C-RA concentrations ranging from 0.1 nM to 1 M (Fig. 1B). The effect of 9C-RA was dosedependent for both genes. Nevertheless, the repression of interstitial collagenase expression was much more sensitive to 9C-RA treatment than the induction of the ST3 gene. Indeed, the half-maximal values for stromelysin-3 induction (EC 50 ) and interstitial collagenase repression (IC 50 ) differed by a factor of about 100 (Fig. 1B; EC 50 ϳ 10 nM and IC 50 ϳ 0.1 nM). Similar results were obtained by using t-RA instead of 9C-RA or when the experiments were carried out in serum-free conditions (data not shown). However, in the latter case, the interstitial collagenase base line was much lower (Fig. 2), hampering analysis of its repression by RA isomers.
Our results showing a significant induction of ST3 RNA levels in HFL1 fibroblasts in the presence of RA are in apparent contradiction with the observation recently made by Anderson et al. (48). Using the same fibroblasts as models, they found that TPA-mediated induction of stromelysin-3 RNA was inhibited by RA. As shown on Fig. 2, we could reproduce this inhibition. However, this effect was only observed for a 10 M 9C-RA concentration, which by far exceeds the RA concentrations usually found in physiological conditions. Furthermore, we would like to point out that when using this 10 M 9C-RA concentration, we could not obtain any repression of interstitial collagenase RNA expression, whereas this repression was observed at lower concentrations (Fig. 2), as previously noted by others (23).
Induction of Stromelysin-3 Protein Synthesis and Secretion by Retinoic Acid in Fibroblasts-To find out whether stromelysin-3 protein synthesis and/or secretion were also increased by RA treatment, conditioned media from HFL1 fibroblasts were analyzed by Western blot (Fig. 3). In serum-free conditions, only low levels of the mature stromelysin-3 form were detected at about 47 kDa. However, when fibroblasts were exposed for 3 days to 1 M of either 9C-RA or t-RA, high levels of this form were detected together with additional protein species. The highest molecular weight form corresponds to the stromelysin-3 proform, which is known to be converted by furin or furin-like enzymes into the mature form (10,11), which in turn can be processed further into another low molecular weight species (Fig. 3 and Ref. 14).
Transcriptional Control of Stromelysin-3 and Interstitial Collagenase Genes by Retinoic Acid in Fibroblasts-To determine whether a transcriptional mechanism was involved in controlling the levels of stromelysin-3 and interstitial collagenase RNAs by RA, we analyzed the nuclear RNAs of both MMPs by using run-on assays performed on nuclei isolated from HFL1 fibroblasts after they had been treated for 1-3 days with 1 M 9C-RA. Radiolabeled RNAs resulting from nascent nuclear RNA transcripts elongated in vitro were hybridized to cDNAs cloned into the pBluescript II SKϩ plasmid and corresponding to interstitial collagenase, stromelysin-3, 36B4, or the plasmid alone as a control for nonspecific hybridization. The results presented on Fig. 4 show that both MMP genes are constitutively transcribed in HFL1 fibroblasts. After 3 days in the presence of 9C-RA, interstitial collagenase transcription was no longer detectable. On the other hand, 9C-RA was found to increase the rate of stromelysin-3 gene transcription by 2-fold, thereby reaching levels similar to those observed for the 36B4 gene, whose expression is not affected by 9C-RA ( Fig. 4 and Ref. 37). Shorter exposure times of HFL1 fibroblasts to 9C-RA (1 or 2 days) led to either no increase or a very little increase in stromelysin-3 gene transcription (data not shown).

Activation of the Human Stromelysin-3 Gene Promoter by Retinoic Acid via Endogenous Retinoid Receptors in Rhab-
domyosarcoma Cells-HFL1 fibroblasts, like other nonimmortalized human diploid fibroblasts, are difficult to use for promoter studies in transient transfection experiments. Therefore, we looked for an established cell line expressing the stromelysin-3 gene that would be easier to use for transfection studies. Since the stromelysin-3 gene is only weakly expressed in human fibrosarcoma cell lines such as HT-1080 and cannot be induced by TPA in these cells (17), we screened several human cell lines of mesodermal origin based on their ability to respond to TPA and RA. We thus identified a rhabdomyosarcoma tumor cell line (RD) that exhibits a stromelysin-3 expression pattern very similar to that of HFL1 fibroblasts. In particular, basal levels of stromelysin-3 RNA expression as well as its induction by 9C-RA, which is maximal after 4 days of incubation, were found to be similar in both cell types (Fig. 5). However, we observed that these RD rhabdomyosarcoma cells do not express the interstitial collagenase gene (Fig. 5), even upon exposure to TPA (data not shown).
To further evaluate whether a transcriptional regulation was involved in the induction of stromelysin-3 gene expression by RA, we analyzed stromelysin-3 promoter activity in RD rhabdomyosarcoma cells exposed to 9C-RA for 4 days. RD rhabdomyosarcoma cells that had been preincubated with 9C-RA for 2 days were transiently transfected by a CAT reporter gene driven by various lengths of stromelysin-3 promoter and further exposed to 9C-RA for an additional period of 2 days, before measurement of CAT activities ( Fig. 6 and data not shown). Upon the addition of 9C-RA, the activities of all three stromelysin-3 promoter constructs containing the DR1-RARE (0.45-, 1.47-, and 3.40-ST3-CAT) were induced 2.8 Ϯ 0.5-, 3.2 Ϯ 0.6-, and 3.3 Ϯ 0.5-fold (n ϭ 3), respectively. Conversely, the absence of the DR1-RARE in the 0.29ST3-CAT and the 3.40ST3-⌬DR1 constructs reduced 9C-RA inducibility to 1.2 Ϯ 0.1-and 1.6 Ϯ 0.1-fold (n ϭ 3), respectively. A similar remaining activation by RA was previously observed for the 0.29ST3-CAT construct when transfected into COS-1 cells (18). This may be attributed to the presence of several widely spaced half-RARE motifs (PuG(G/T)TCA) present in this promoter region that have been shown to activate transcription in the presence of RA (49). The activation by 9C-RA was also tested on the RAR␤2 promoter, which contains a RARE of the DR5 type (47), and on the isolated DR1 element inserted upstream of the herpes simplex virus thymidine kinase promoter. The activity of these two constructs was induced 3.1 Ϯ 0.2-and 4.9 Ϯ 0.7-fold (n ϭ 3) by 9C-RA, respectively, thus to levels comparable with those observed for ST3 constructs. However, the transactivation of the DR1-tk-CAT construct was weaker (1.4 Ϯ 0.3-fold, n ϭ 2) when RD rhabdomyosarcoma cells were exposed to 9C-RA for only 2 instead of 4 days, thereby suggesting that some of the regulatory factors implicated in this activation are not constitutively expressed in these cells. Also, it should be noted that these experiments were performed without the cotransfection of any retinoid receptor, indicating that the observed effects were mediated through endogenous receptors.
Retinoic Acid Receptor Expressions in Fibroblasts-To determine the respective contribution of RARs and RXRs in mediating stromelysin-3 induction and interstitial collagenase repression by RA, their expression in HFL1 fibroblasts was first analyzed by Northern blot. We found that untreated fibroblasts cultured in serum-free conditions expressed similar levels of RAR␣, RAR␥, and RXR␣ RNAs (Fig. 7), with steady state levels relatively constant over the time of culture (not shown). No expression was detected, however, in untreated cells for RAR␤, RXR␤, and RXR␥ RNAs, even when up to 30 g of total RNA were loaded for analysis ( Fig. 7 and data not shown). These results are consistent with recent studies that have shown that RAR␣, RAR␥, and RXR␣ are the predominant receptors expressed in human skin (50) as well as in various human cell lines (51)(52)(53). The expression of RXR␣ was only slightly increased (less than 2-fold) in cells treated with either 9C-RA or t-RA, whereas RAR␣ and RAR␥ levels remained unaffected. In   FIG. 3. Induction of stromelysin-3 synthesis and secretion by RA in HFL1 fibroblasts. Culture media conditioned by HFL1 fibroblasts exposed to either 1 M 9C-RA or 1 M t-RA in serum-free conditions for 3 days were collected prior to being concentrated 100-fold by ammonium sulfate precipitation. Western blot analysis was performed by using monoclonal antibody 5ST-4C10 raised against the stromelysin-3 catalytic domain and enhanced chemiluminescence, as described under "Materials and Methods." Molecular mass markers (kDa) are indicated on the left.  5. Comparative expression of stromelysin-3 and interstitial collagenase RNAs in response to 9C-RA in HFL1 fibroblasts and RD rhabdomyosarcoma cells. Total RNA (10 g) from HFL1 and RD rhabdomyosarcoma cells grown in 5% calf serum and treated with 1 M 9C-RA for the indicated times was analyzed by Northern blot for ST3, interstitial collagenase (Int. Col.), and 36B4 expression, as described under "Materials and Methods." contrast, RAR␤ RNA levels increased from undetectable to high levels in cells exposed to either of the RA isomers (Fig. 7). Similar results were obtained by using fibroblasts cultured in 5% calf serum, although RAR␤ was induced to a lower extent than in serum-free conditions (data not shown).
Synergistic Activation of Stromelysin-3 Gene and Repression of Interstitial Collagenase Gene by Specific Synthetic Retinoids-We then investigated whether the ligand-dependent activation of both RARs and RXRs was required for inducing or repressing the expression of stromelysin-3 and interstitial collagenase genes, respectively. We used the synthetic ligands Am80 (35) and CD666 (34), which at appropriate concentrations selectively activate RAR␣ and RAR␥, respectively (37), BMS753, a pure RAR␣ agonist (38), and BMS649, a pan-RXR agonist (36). The expression levels of stromelysin-3 and interstitial collagenase RNAs in HFL1 fibroblasts cultured in 5% calf serum were evaluated after 3.5 days of culture in the presence of these synthetic retinoids and compared with those observed in the presence of 9C-RA and t-RA.
Using these retinoids individually at low and/or selective concentrations, either no induction or a very weak induction of stromelysin-3 was detected (Fig. 8, A and B), while interstitial collagenase expression was reduced by at least 50% (Fig. 8, C  and D). At higher concentrations (Ͼ10 nM), when Am80 and CD666 lose their specificity and act as pan-RAR agonists (37), higher levels of stromelysin-3 RNA were observed, while interstitial collagenase expression was repressed further. Interestingly, very little stromelysin-3 gene induction was noted with the pure RAR␣ agonist BMS753 and the pan-RXR agonist BMS649, even when these retinoids were used at a 1 M con-centration. In marked contrast, the combination of either Am80 (100 nM and 1 M) or CD666 (100 nM) with the pan-RXR ligand BMS649 (1 M) resulted in a synergistic induction of the stromelysin-3 gene, reaching expression levels close to those observed with the natural ligands. A synergistic effect was also observed when the BMS753 and BMS649 ligands were combined, although the expression levels of stromelysin-3 RNA did not exceed 50% of those observed in the presence of the natural ligands. However, any of these combinations was found to fully repress interstitial collagenase gene expression. We note that stromelysin-1 gene expression was similarly repressed in HFL1 fibroblasts, 2 suggesting that the retinoids used here may efficiently repress the expression of any AP-1-regulated MMP.
Taken together, our observations indicate that while the selective activation of RAR␣, RAR␥, or RXRs substantially represses interstitial collagenase gene expression, the combination of RARs and RXRs is required for optimal stromelysin-3 gene induction and for full repression of interstitial collagenase. DISCUSSION We have previously shown that the stromelysin-3 gene promoter differed from most other MMP promoters by the absence of a functional AP-1 binding site and the presence of a RARE in its proximal region. In the present study, we have further investigated the regulation of stromelysin-3 gene expression by RA and compared this regulation with that of interstitial collagenase, another MMP. Stromelysin-3 and interstitial collagenase are both predominantly expressed by stromal cells of human carcinomas (15), and their high expression levels were found to be associated with a poor clinical outcome in some carcinomas (12,13,54). Considering that retinoids by themselves or when associated with other drugs such as tamoxifen, are regarded as potential new anticancer agents (29 -31), it is important to elucidate the mechanisms by which the expression of MMPs implicated in cancer progression is regulated by RA. We demonstrate here that both natural RA isomers, 9C-RA and t-RA, strongly induce stromelysin-3 RNA and protein expression and simultaneously repress interstitial collagenase expression in human fibroblasts. In addition, we show that both genes are controlled by RA through a transcriptional mechanism, and we provide evidence indicating that RAR⅐RXR heterodimers are the functional units required for an optimal control of these genes by RA.
AP-1 and retinoid receptors are regarded as effectors of opposite pathways of cell proliferation and differentiation, and 2 P. Anglard, unpublished results. they mutually antagonize each other at the level of transactivation and DNA binding (22)(23)(24)(25). Indeed, MMP genes containing an AP-1 binding site in a conserved position in their promoter or other genes like those for tumor growth factor-␤1 (55) and interleukin-6 (56) are TPA-inducible, while their expression is inhibited by RA. Knowing that, reciprocally, AP-1 can inhibit transactivation by RARs and RXRs, the observation that the stromelysin-3 gene is induced by both TPA and RA in a given cell type is quite unexpected and represents an unusual example of a gene up-regulated by both agents.
We have found that physiological concentrations of RA efficiently induce both the expression of the stromelysin-3 gene and the repression of the interstitial collagenase gene in HFL1 human fibroblasts, the latter being observed at RA concentrations lower than those necessary for stromelysin-3 induction. Interestingly, the IC 50 for interstitial collagenase and the EC 50 for stromelysin-3 that we found were very similar to the values recently reported by Chen et al. (25) in promoter studies. These authors have shown that the repression of AP-1-induced tran-scription from the interstitial collagenase promoter was about 100 times more sensitive than the transactivation of a RAREtk-CAT construct in the presence of RA. Since these observations suggested that the regulation of both genes by RA may be achieved through a transcriptional mechanism, we further evaluated this possibility by measuring the transcriptional rate of both genes in HFL1 fibroblasts in run-on assays. In the presence of 9C-RA, we observed a complete inhibition of interstitial collagenase transcription, which is likely to result from an RAR/AP-1 interaction, since this has been previously documented (24,57). On the other hand, a 2-fold increase in the stromelysin-3 gene transcriptional rate was found when HFL1 fibroblasts were exposed to 9C-RA for 3 days. No clear transcriptional activation could be detected for shorter exposure times. Although it is difficult to determine whether this 2-fold increase can fully account for the 20-fold increase in stromelysin-3 RNA levels observed after 4 days of RA treatment, we note that run-on studies with other RA-inducible genes containing a RARE in their promoter exhibited similar profiles.
FIG. 8. Comparative expression of stromelysin-3 and interstitial collagenase genes in HFL1 fibroblasts exposed to selective retinoids. The effects of natural and synthetic retinoids on ST3 gene induction (A and B) and interstitial collagenase (Int. Col.) gene repression (C and D) were analyzed by Northern blot as described under "Materials and Methods." HFL1 fibroblasts cultured in 5% calf serum were treated for 3.5 days with RAR␣-(Am80 and BMS753; panels A and C) or RAR␥-(CD666; panels B and D) specific synthetic retinoids and/or the pan-RXR agonist BMS649, used at the indicated concentrations. The two natural isomers 9C-RA and t-RA were used at 1 M. Relative levels of RNA transcripts, as evaluated by using a bioimaging analyzer (BAS 2000; Fuji Ltd.) for each blot after normalization with 36B4 RNA levels, are presented in the histograms. The induction of stromelysin-3 RNA is expressed relative to that observed in the presence of 9C-RA, while the repression of interstitial collagenase RNA is expressed relative to its level in untreated cells.
Thus, the RAR␤ and the laminin B1 RNAs were found to be induced at high levels by RA in F9 cells, while no increase or only a moderate increase in transcriptional rates could be detected for these genes by nuclear run-on assays (58,59). In all instances, the contribution of a transcriptional mechanism in stromelysin-3 gene induction is further supported by our findings showing that 9C-RA induces stromelysin-3 promoter activity in RD rhabdomyosarcoma cells. By analyzing various lengths of this promoter in transient transfection experiments, we observed a 3-fold induction of stromelysin-3 promoter activity in the presence of 9C-RA, which was strongly reduced in the constructs lacking the DR1-RARE. Interestingly, this transactivation was observed without the addition of retinoid receptors, indicating that the DR1-RARE was activated by functional endogenous retinoid receptors in these cells.
Previous studies have shown that while all RARs could potentially mediate the induction of RA targets genes, the involvement of a given receptor was dependent on many parameters including promoter context and cell type (37,47). When the expression of RARs and RXRs was evaluated in HFL1 fibroblasts, we found that RAR␣, RAR␥, and RXR␣ RNAs were constitutively expressed at high levels. In contrast, we could not detect any RNA for RXR␤ and RXR␥, while that for RAR␤ was strongly induced from barely detectable levels in untreated fibroblasts to high levels in the presence of 9C-RA or t-RA. Similar observations have been made in human dermal (60) and lung (61) fibroblasts.
Since these observations suggested that specific retinoid receptors could be involved in the regulation of stromelysin-3 and interstitial collagenase expression by RA in HFL1 fibroblasts, we tested the expression of both genes in the presence of selective retinoids. We found that these retinoids, when used individually at concentrations at which they selectively activate a given RAR (Am80, CD666, BMS753) or all three RXRs (BMS649), led to very weak, if any, stromelysin-3 gene induction, whereas they repress interstitial collagenase expression by at least 50%. In marked contrast, a clear induction of the stromelysin-3 gene was observed when any of the selective RAR ligands was used in combination with the BMS649 RXRspecific ligand. We note, however, that the combination BMS753-BMS649 (RAR␣⅐RXRs) was less efficient than the other combinations. Since stronger inductions were observed by combining the BMS649 RXR agonist with either Am80 or CD666 at concentrations at which they both promiscuously activate all three RARs, it is reasonable to believe that either RAR␤ and/or RAR␥ could interact with RXRs for an optimal stromelysin-3 induction. In keeping with the fact that RXR␣ seems to be the major RXR expressed in fibroblasts, we can conclude from our results that the two heterodimers RAR␤⅐RXR␣ and/or RAR␥⅐RXR␣ are likely to represent the functional units required to induce the expression of the stromelysin-3 gene at physiological RA concentrations. This possibility is also consistent with in vitro studies that have shown that heterodimers bind to RARE much more efficiently than the respective homodimers (27,28). In this respect, it is noteworthy that the activation of a single RAR or RXR was sufficient to substantially repress interstitial collagenase expression in HFL1 fibroblasts but that the activation of both partners of heterodimers was necessary for a full repression.
In summary, while transcription studies have demonstrated that RA regulates the expression of target genes by either activating RAREs or repressing AP-1 activity, we have looked at the regulation of two genes that belong to the MMP family and shown that they are differentially regulated by RA in human fibroblasts. Indeed, we have shown that physiological concentrations of RA induce stromelysin-3 expression but re-press interstitial collagenase expression. Compared with the repression of interstitial collagenase, stromelysin-3 gene induction relies on more restricted conditions, based on a lower sensitivity to both natural and synthetic retinoids, and on a more restricted receptor requirement involving RAR⅐RXR heterodimers. In contrast, a substantial transcriptional repression of interstitial collagenase is achieved by retinoids activating only one type of receptors, although the involvement of RAR⅐RXR heterodimers is required for a full repression. Taking into account the ongoing efforts in designing potent MMP inhibitors to inhibit cancer progression, the finding that dissociating retinoids such as RXR-selective ligands can prevent the expression of some MMPs by an AP-1 transrepression mechanism without inducing stromelysin-3 gene expression may be of interest for therapeutic applications.