The Comparative Role of Activator Protein 1 and Smad Factors in the Regulation of Timp-1 and MMP-1 Gene Expression by Transforming Growth Factor- (cid:1) 1*

, The balance between matrix metalloproteinases (MMPs) and their inhibitors, the tissue inhibitors of metalloproteinases (TIMPs), is pivotal in the remodeling of extracellular matrix. TGF- (cid:1) has profound effects on extracellular matrix homeostasis, in part via its ability to alter this balance at the level of gene expression. The intracellular signaling pathways by which TGF- (cid:1) mediates its actions include the Smad pathway, specific to the TGF- (cid:1) superfamily, but also, for example, mitogen-activated protein kinase pathways; furthermore, cross-talk between the Smads and other signaling pathways modifies the TGF- (cid:1) response. The reciprocal effect of TGF- (cid:1) on the expression of Timp-1 and MMP-1 supports its role in matrix anabolism, yet the mechanisms by which TGF- (cid:1) induces Timp-1 and represses induced MMP-1 have remained opaque. Here, we (i) investigate the mechanism(s) by which TGF- (cid:1) 1 induces expression of the Timp-1 gene and (ii) compare this with TGF- (cid:1) 1 repression of phorbol ester-induced MMP-1 expression. We report that the promoter-proximal activator

Timely breakdown and remodeling of the extracellular matrix (ECM) 1 is an essential process in development, morpho-genesis, and reproduction. ECM degradation is also associated with a variety of physiological and pathological processes such as joint destruction in the arthritides, wound healing, tumor metastasis, angiogenesis, and fibrosis (1). Pivotal to the turnover of ECM is the matrix metalloproteinase (MMP) family of enzymes; these enzymes have the capability, between them, of degrading the majority of the proteins that make up the ECM (2). The tissue inhibitors of metalloproteinases (TIMPs) protect ECM integrity by inhibiting MMPs (3). MMP-1, interstitial collagenase, is one of a subfamily of MMPs that can specifically degrade the collagen triple helix; hence, MMP-1 plays a central role in pathologies where collagen turnover is aberrant (4). As well as inhibiting most of the active MMPs, TIMP-1 is also reported to have diverse effects on cell growth and apoptosis (5).
Transforming growth factor-␤ (TGF-␤) is a multifunctional growth factor controlling cell growth and differentiation and has marked effects on ECM homeostasis (6). This includes the induction of ECM gene expression and generally suppression of MMPs and induction of TIMPs to give a "synthetic" phenotype (7). Hence, TGF-␤ is associated, for example, with fibrosis in a number of diseases (8). TGF-␤ has previously been shown to repress the expression of MMP-1, induced by a variety of stimuli, in a number of cell types (9,10). Conversely, TGF-␤ induces Timp-1 gene expression, often in synergy with other growth factors and cytokines (11,12).
TGF-␤ signals via transmembrane receptors to intracellular mediators of the Smad family. Smad 2 and Smad 3 are receptor-specific Smads that are phosphorylated on serine residues by the type I TGF-␤ receptor. Upon phosphorylation, these Smads form heteromeric complexes with a common mediator Smad 4 and can then be translocated to the nucleus, where they regulate gene expression either directly or in association with a number of co-activators and co-repressors. An inhibitory Smad 7 blocks this cascade to prevent TGF-␤-mediated alterations in gene expression (13). TGF-␤ can also signal through other pathways (e.g. mitogen-activated protein kinase pathways), although the mechanisms for activation of these pathways appear diverse (14,15). Cross-talk between the Smad signaling cascade and other pathways also adds complexity to the system (16 -18).
The proximal promoters of both MMP-1 and TIMP-1 genes contain an AP1 site, which, in each case, has been the focus for research on their regulation. In the murine Timp-1 gene, the AP1 site is located at Ϫ59 bp and has been shown to be impor-tant in both basal and inducible Timp-1 gene expression; mutation of this site in either the mouse or human gene results in greater than 90% reduction in expression of promoter-reporter constructs in transient transfection studies (19,20). In the human MMP-1 gene, the AP1 site is located at Ϫ72 bp and in both human and rabbit genes is critical for robust expression from promoter-reporter constructs (21,22). For each gene, with some dependence on promoter length used, induction of the gene by phorbol ester is still apparent when the promoterproximal AP1 site is mutated, although with much reduced basal expression (22). 2 Interestingly, the Timp-1 AP1 site is noncanonical (5Ј-TGAGTAA-3Ј), and this is conserved across all species sequenced, whereas the MMP-1 AP1 site is the consensus sequence (5Ј-TGAGTCA-3Ј). The Timp-1 AP1 site has recently been reported to bind a distinct nuclear factor compared with the consensus; this factor, ssT1, is a singlestranded DNA-binding protein of unknown identity or function (23).
Upstream of the AP1 site in the MMP-1 promoter is a putative TGF-␤-inhibitory element (TIE; Ϫ245 bp, 5Ј-GAATTG-GAGA-3Ј), first described in the rat MMP-3 promoter to mediate repression of epidermal growth factor-stimulated expression of this gene by TGF-␤ (24). Nuclear proteins from TGF-␤-treated fibroblasts, including c-Fos, were shown to bind to this sequence from the MMP-3 promoter. A recent study using the rabbit MMP-1 promoter transfected into rabbit synovial fibroblasts demonstrated that the MMP-1 TIE functions as a constitutive repressor of and an antagonist of phorbol-esterinduced MMP-1 gene expression (9).
We postulated that the ability of TGF-␤ to induce Timp-1 expression but suppress induced MMP-1 expression could be mediated at several possible levels (e.g. (i) differences in the AP1 motif, (ii) the presence of the TIE in the MMP-1 gene, and (iii) binding of Smad complexes to either gene). TGF-␤ regulation of many genes is dependent on AP1 motifs in their promoters (e.g. clusterin (25), PAI-1, and type I collagen (26)). Evidence for the first and third possibilities together comes from Zhang et al. (27), who reported that Smad 3 interacts directly with the MMP-1 AP1 motif to activate transcription in response to TGF-␤ and that Smad 3 and Smad 4 can activate TGF-␤-inducible transcription from this site in the absence of c-Jun and c-Fos. Smad 3 and c-Jun can bind to the MMP-1 AP1 site simultaneously, but footprinting suggested that Smad binding was at the 3Ј end of the AP1 motif (5Ј-GTCAGCC-3Ј), which is not identical in the Timp-1 gene (5Ј-GTAATGC-3Ј). Indeed, the authors suggest that different AP1 sites could have differing affinity for Smad 3, dependent on a few nucleotides flanking the site. Similarly, Smads are reported to bind directly to Jun family members (28), and Smad 3 is reported to potentiate the induction of gene expression by Jun family members in AP1-dependent promoters via protein-protein interaction (29). Yuan and Varga (30) document that TGF-␤ repression of IL-1-induced MMP-1 expression is Smad-mediated, although the cis-acting sequences through which this effect is mediated are not localized. Timp-1 has been reported as a Smad-responsive gene in dermal fibroblasts (31) using microarray and transient transfection technology.
Here, we investigate the mechanism(s) by which TGF-␤1 induces expression of the Timp-1 gene and compare this to the TGF-␤1 repression of PMA-induced MMP-1 expression. We report that the promoter proximal AP1 site is essential for the response of both Timp-1 and MMP-1 to TGF-␤ (induction and repression, respectively). c-Fos, JunD, and c-Jun are essential for the induction of Timp-1 gene expression by TGF-␤1, but these AP1 factors transactivate equally well from both the Timp-1 and MMP-1 AP1 sites. Smad-containing complexes do not interact with the Timp-1 AP1 site, and overexpression of Smads does not substitute or potentiate the induction of the gene by TGF-␤1; furthermore, Timp-1 is still induced by TGF-␤1 in Smad 2, Smad 3, or Smad 4 knockout cell lines, although to varying extents. In contrast, Smads do interact with the MMP-1 AP1 site and mediate the repression of induced MMP-1 gene expression by TGF-␤1.
The AP1 expression plasmids in pCMV were described by Harrison et al. (34) and were a kind gift of Dr. P. R. Dobner (University of Massachusetts). Smad expression plasmids in pCMV were described by Wicks et al. (18) and were from the laboratory of Dr. A. Chantry (University of East Anglia).
All mutagenesis was performed using the QuikChange method (Stratagene). All mutations were verified by sequencing.
p3TP-Lux is an artificial promoter consisting of the plasminogen activator inhibitor-1 TGF-␤-responsive promoter and three repeats of the MMP-1 AP1 site (35).
Transient Transfection-Cells were seeded in six-well plates at a density of 8850 cells/cm 2 and grown overnight in medium containing 10% fetal calf serum at 37°C in a 5% CO 2 atmosphere. Cells were transfected overnight in serum-containing medium with 1 g/well reporter plasmid using FuGene 6 (Roche Molecular Biochemicals) according to the manufacturer's instructions. The following day, cells were washed in Hanks' balanced salts solution and incubated in serum-free medium overnight. Cells were then stimulated with phorbol 12-myristate 13-acetate (PMA; 10 Ϫ7 M; Sigma) or TGF-␤1 (2 ng/ml; R&D Systems) or both together for varying times as shown, prior to harvest. Harvest and assay were according to the manufacturer's instructions (luciferase activity; Promega).
For co-transfection with expression constructs, an additional Յ1 g/well expression vector(s) was included, keeping the total DNA to 2 g/well using empty vector. For co-transfection with oligonucleotides, an additional 1 g/well of oligonucleotide (wild-type or mutant control) was included in the transfection.
Nuclear Extracts-Confluent cells at a density of ϳ2 ϫ 10 7 /150-mm dish were washed and incubated in medium containing 0.1% bovine serum albumin overnight. Cells were then treated with TGF-␤1 (2 ng/ml), PMA (10 Ϫ7 M), or both together for 3 h. Cells were scraped into ice-cold phosphate-buffered saline, pelleted at 500 ϫ g, and resuspended in 1 ml of phosphate-buffered saline, 0.1% Nonidet P-40 for ϳ30 s. After centrifugation at 13,000 ϫ g for 10 s, pellets were rinsed twice with phosphate-buffered saline, 0.1% Nonidet P-40 and then resuspended in 3 volumes of high salt buffer (25 mM HEPES, pH 7.8, 500 mM KCl, 0.5 mM MgSO 4 , 1 mM dithiothreitol) containing 1ϫ Complete protease inhibitors (Roche Molecular Biochemicals). Samples were in-cubated on ice for 20 min with occasional vortex and then centrifuged at 13,000 ϫ g for 2 min at 4°C. Supernatant was then divided into aliquots, frozen on dry ice, and stored at Ϫ80°C. Protein concentration in the nuclear extract was determined by Bradford assay (Bio-Rad) and was typically 2-5 g of protein/l.
Probe Labeling-Oligonucleotides for electrophoretic mobility shift assay (EMSA) were synthesized by MWG-Biotech.
. Double-stranded probes were labeled with [␣-32 P]dCTP using Klenow fill-in, whereas single-stranded probes were labeled with [␥-32 P]ATP using T4 polynucleotide kinase. Labeling reactions were followed by phenol/chloroform extraction and purification through a Sephadex G50 spin column.
Reverse Transcription-PCR-RNA was isolated from monolayer cultures using Trizol (Invitrogen). Quantitative reverse transcription-PCR was performed using the Applied Biosystems ABI Prism 7700 sequence detection system (TaqMan) as described (37).

RESULTS
Deletion Analysis of the Timp-1 Promoter-Prior to studying the response of the Timp-1 promoter to TGF-␤1, the transient transfection protocol was optimized to enable us to assay the response of the promoter-reporter constructs at early time points. Previous studies at the level of steady-state mRNA for Timp-1 have suggested that there is an early primary response to TGF-␤1, followed by a later secondary response, which may be mediated by or in conjunction with TGF-␤1-induced autocrine factors (12). Using luciferase as a reporter, transgene expression was robust at a 6-h time point, and this was therefore used throughout the studies of the Timp-1 promoter.
A Ϫ925/ϩ47 Timp-1 promoter construct reiterates the response of the endogenous Timp-1 gene (12) to PMA and TGF-␤1, demonstrating a significant induction by each factor alone and an augmented response to both factors together (Fig. 1). Deletion from the 5Ј end of this construct demonstrated that, whereas upstream sequences may impact on the level of induction (or indeed on basal expression), this pattern of expression is maintained in a Ϫ62/ϩ47 Timp-1 promoter construct; how- ever, induction is lost in a Ϫ50/ϩ47 construct in which an AP1 site at Ϫ59/Ϫ53 is absent. This construct loses TGF-␤1 inducibility and the synergism between TGF-␤1 and PMA, but it maintains a low level of PMA inducibility (ϳ1.5-fold). This demonstrates that the Ϫ59/Ϫ53 AP1 site is critical for TGF-␤1 induction of the Timp-1 gene. This is confirmed by an inactivating point mutation in this AP1 site in the context of the Ϫ223/ϩ47 construct, whereby PMA and TGF-␤1 induction are markedly reduced compared with wild-type (and the synergism between the two is lost) although not completely abolished. It should also be noted that the pattern of response of the Timp-1 promoter to TGF-␤1 and PMA is replicated in other cell lines (e.g. Swiss 3T3).
Protein Binding to the Timp-1 AP1 Site-EMSA and supershift analysis was used to probe protein binding to the Timp-1 AP1 site under PMA and TGF-␤1 stimulation (Fig. 2). At this 3-h time point, the AP1 site is bound in unstimulated nuclear extracts; upon TGF-␤1 or PMA stimulation, binding to this site increases, although the mobility of the complex remains unaltered. Treatment with both factors together increases binding further. Specificity of binding was ascertained by competition with cold self and mutant oligonucleotides (data not shown). Supershift/antibody blocking analysis using antibodies against members of the Jun and Fos family was performed to examine the components of the AP1-binding complex under each condition. Nuclear extracts from cells induced with PMA or PMA plus TGF-␤1 appear to contain all of the Fos and Jun family members assayed. Stimulation with TGF-␤1 alone gave a similar pattern of response except for the absence of a supershift with the anti-Fra1 antibody; this indicates that TGF-␤1 alone does not induce Fra1 binding to this oligonucleotide.
When these experiments were repeated using an oligonucleotide containing the MMP-1 consensus AP1 sequence, identical results were obtained (data not shown).

Overexpression of AP1 Factors Transactivates the Timp-1
Promoter-In order to assess the role of differing Fos and Jun family members in activating transcription from the Timp-1 promoter, co-transfection experiments were performed. The Ϫ95/ϩ47 Timp-1 luciferase construct was transiently transfected into C3H10T1/2 cells with combinations of expression constructs for c-Fos, Fra-1, Fra-2, FosB, c-Jun, JunD, and JunB; empty vector was used a control (Fig. 3). Expression and function of these factors was assessed by EMSA (data not shown). Alone, none of the Jun family members could transactivate the Timp-1 promoter, whereas c-Fos and, to a lesser extent, FosB, do transactivate, presumably in combination with endogenous Jun factors expressed in these cells. The combination of c-Fos with c-Jun or JunD gives the most potent induction of the Timp-1 promoter construct, ϳ4-fold above c-Fos alone and 8-fold above control.
Substitution of the wild-type AP1 (5Ј-TGAGTAA-3Ј) site in the Ϫ95/ϩ47 for the MMP-1 consensus AP1 site (5Ј-TGAGTCA-3Ј) (i.e. an A to C transversion) did not alter the pattern of response to these AP1 family members (data not shown). Furthermore, using a Ϫ125/ϩ60 MMP-1 luciferase construct, transiently transfected into C3H10T1/2 cells or Swiss 3T3 cells (in which PMA-induced MMP-1 expression either is not or is repressed by TGF-␤1, respectively; see below), with AP1 factors, gave the same pattern of response (data not shown).
Expression of Timp-1 in AP1 Knockout Cells-In order to verify the data gained from overexpression of AP1 factors, a Timp-1 promoter construct was transiently transfected into cell lines deleted for one of c-fos, c-jun, junD, or fra-1, and cells were stimulated with either PMA or TGF-␤1 (Fig. 4). In these experiments, a Ϫ925/ϩ47 Timp-1 luciferase plasmid was used, since many of the AP1 knockout cell lines transfect poorly, and this construct gives a higher level of expression. As a control vector, p3TP-Lux was used; this is an artificial construct consisting of three copies of the MMP-1 AP1 site and one copy of the TGF-␤1-responsive PAI-1 promoter. p3TP-Lux is known to be both AP1-and Smad-responsive. All constructs were cotransfected with pRSVCAT and CAT expression was used to normalize the data for transfection efficiency. It should be noted that the ϩ/ϩ cells in the c-Fos and c-Jun experiments are Swiss 3T3 cells, since cells from wild-type littermates of the knockouts were not available. Conversely, ϩ/ϩ cells in the JunD and Fra-1 experiments were from wild-type littermates. Variability in response of these "wild-type" cells to TGF-␤1 and PMA is apparent, and this reinforces the need for a PMA-and TGF-␤1-inducible control plasmid such as 3TP-Lux.
The data show that c-Fos, c-Jun, and JunD are each necessary, and Fra-1 is not necessary for TGF-␤1-induction of the Timp-1 gene, although PMA still induces expression in the absence of c-Fos. Interestingly, c-Jun is also necessary for induction of 3TP-Lux by either TGF-␤1 or PMA, and JunD is necessary for induction by PMA. These data underscore the importance of c-Fos, c-Jun, and JunD as AP1 factors that induce Timp-1 expression (as seen in overexpression studies above) and place them in the pathway by which TGF-␤1 induces Timp-1 gene expression.
EMSA analysis on overlapping oligonucleotides across the Ϫ62/ϩ47 region shows no obvious alterations in protein binding after TGF-␤1 or PMA stimulation other than on the AP1 site above (data not shown).
TGF-␤1 Represses PMA-induced Expression from the MMP-1 Promoter-It has been previously reported that TGF-␤1 represses the expression of MMP-1 when induced by a variety of factors (e.g. PMA, interleukin-1, and tumor necrosis factor-␣) (9,10). In order to reiterate this in our model system, a construct containing Ϫ517/ϩ60 of the human MMP-1 promoter in a luciferase vector was transiently transfected into both murine Swiss 3T3 cells and primary human skin fibroblasts. In both cases, PMA potently induces expression from this construct, and TGF-␤1 represses this induction although, at the doses used, not back to control levels (see Fig. 5, A and B). TGF-␤1 alone does not significantly repress basal expression of MMP-1. These data agree with data published in other cell lines, and we have confirmed that the same response is seen with either 8-or 24-h stimulation (data not shown). It should be noted that TGF-␤1 does not repress PMA-induced expression from the MMP-1 promoter in C3H10T1/2 cells.
A shorter MMP-1 promoter construct, Ϫ153/ϩ60, shows an identical pattern of response to Ϫ517/ϩ60. This suggests that the putative TIE at Ϫ245 is not involved in the repression of PMA-induced MMP-1 expression. A construct of Ϫ80/ϩ60, containing the proximal AP1 site at Ϫ72, gives very low levels of expression, but the TGF-␤1 repression of PMA-induced expression is still apparent; therefore, elements within the Ϫ80/ϩ60 region must be responsible for this effect of TGF-␤1 (Fig. 5A). Point Mutations in Ϫ517/ϩ60 Confirm the Role of the AP1 Site in TGF-␤1-mediated Repression of MMP-1-Since previous data suggest that in MMP-1 promoter constructs extending further 5Ј than Ϫ321 (18), the AP1 site at Ϫ72 may contribute less to expression of the transgene, functionally inactivating point mutations in both TIE and AP1 motifs were made in the context of Ϫ517/ϩ60. Fig. 5B shows that mutation of the TIE does not prevent TGF-␤1 repression of PMA-induced expression. However, the inactivating mutation in the AP1 motif, in the presence or absence of the TIE mutation, actually leads to a further induction of PMA-induced gene expression by TGF-␤1. Interestingly, the TIE mutation increases absolute levels of expression, whereas the AP1 mutation decreases absolute levels. Exchanging the consensus AP1 sequence in Ϫ517/ϩ60 for the Timp-1 AP1 sequence does not alter the pattern of expression; moreover, transient transfection of a Ϫ95/ϩ47 Timp-1 construct containing the MMP-1 AP1 site into C3H10T1/2 or Swiss 3T3 cells did not alter the pattern of TGF-␤1 and PMA induction seen in the same construct containing the wild-type Timp-1 AP1 motif (data not shown).
Overexpression of Smads 2, 3, and 4 Does Not Potentiate TGF-␤1 Induction of Timp-1 Promoter-In order to probe the role of the Smad signaling pathway in the response of the Timp-1 gene to TGF-␤1, expression vectors for Smads 2, 3, 4, and 7 were co-transfected into C3H10T1/2 with either the Ϫ95/ϩ47 Timp-1 promoter construct or the Smad-responsive 3TP-Lux, using empty vector as a control. Cells were then stimulated for 6 h with TGF-␤1. Fig. 6A shows that the 3TP- Lux construct behaves in a Smad-responsive manner as expected; TGF-␤1 induces expression of luciferase, and this is further induced by the addition of either Smad 2, 3, or 4 alone. Combinations of these Smads yield even higher levels of expression. The Smad dependence of this response is underlined by co-transfection of the inhibitory Smad 7, which potently blocks TGF-␤1 induction of 3TP-Lux. In comparison with this, TGF-␤1 induction of the Timp-1 promoter is not potentiated by Smad 2, 3, or 4 alone, with Smads 3 and 4 acting in a repressive fashion; combinations of Smads 2 and either 3 or 4 have no effect, whereas Smads 3 and 4 or Smads 2, 3, and 4 potently repress TGF-␤1 stimulation of the Timp-1 construct (Fig. 5B). Furthermore, Smad 7 does not repress TGF-␤1-stimulated Timp-1 expression. Together, these data suggest that the response of the Timp-1 gene to TGF-␤1 is not Smad-dependent.

Timp-1 Expression Is Induced by TGF-␤1 in Smad Knockout
Cell Lines-In our hands, Smad knockout cell lines and their wild-type partner lines proved difficult to transfect reproducibly. Hence, Smad 2, Smad 3, or Smad 4 knockout cells and their wild-type partners were stimulated with TGF-␤1, PMA, or both together, and expression of the endogenous Timp-1 gene was assessed by quantitative reverse transcription-PCR using the Taqman system. All three knockout cell lines retain some TGF-␤1 inducibility, although this is at a reduced level compared with wild types (data not shown). All three knockout cell lines also retain the synergism in Timp-1 induction with PMA and TGF-␤1 together. Interpretation of these data is clouded by the fact that wild-type cell lines show wide variation in their response to TGF-␤1 in both these experiments and, for example, in Fig. 4. A Smad-binding Oligonucleotide Blocks TGF-␤1-mediated Repression of the MMP-1 Promoter but Not Induction of the Timp-1 Promoter-In order to reinforce the overexpression data above, cells were co-transfected with the Ϫ517/ϩ60 MMP-1 promoter or Ϫ95/ϩ47 Timp-1 promoter construct with either the Smad-binding oligonucleotide, S4BE, containing three Smad-binding sites, or a mutant oligonucleotide S4BEmut, where the Smad-binding sites were functionally mutated. Binding of Smads to S4BE and the absence of binding to the S4BEmut were demonstrated using EMSA (see Fig. 8; data not shown). Fig. 7 shows that wild-type oligonucleotide does not alter the pattern of TGF-␤1 or PMA induction of the Timp-1 promoter. Conversely, co-transfection of S4BE blocks TGF-␤1mediated repression of PMA-induced MMP-1 expression, whereas the S4BEmut has no effect. Again, this suggests that whereas Smads do mediate TGF-␤1 repression of the MMP-1 gene, they are not involved in TGF-␤1 induction of the Timp-1 gene.
Smad Binding to the Timp-1 or MMP-1 AP1 Motifs-Using the S4BE oligonucleotide in EMSA with nuclear extracts from C3H10T1/2 cells, five DNA-protein complexes were apparent (Fig. 8). Supershift analysis with antibodies against Smad 2/3 or Smad 4 demonstrated that band A, the slowest migrating complex, contained all of these Smads. Indeed, competition with a 100-fold excess of cold S4BE shows that only band A is competed (data not shown). Mutation of the three 5Ј-CAGA-3Ј sequences in S4BE (to 5Ј-TACA-3Ј) leaves only band D intact (data not shown).
EMSA using oligonucleotides containing the Timp-1 or MMP-1 AP1 site and nuclear extracts from C3H10T1/2 or human skin fibroblasts shows an identical pattern of binding on FIG. 6. Co-transfection of p3TP-Lux, ؊95/؉47 Timp-1, or ؊517/ ؉60 MMP-1 reporter with Smad family expression constructs. A, and B, C3H10T1/2 cells were co-transfected with 1 g of Ϫ95/ϩ47 Timp-1 in pGL2-basic or p3TP-Lux in combination with 0.33 g of Smad family expression vectors or empty vector to total 2 g of DNA/ plate. Cells were serum-starved for 24 h and then stimulated with TGF-␤ (2 ng/ml). Cell lysates were harvested at t ϭ 6 h and assayed for luciferase activity. Data are expressed as -fold induction above the control (empty vector only), mean and S.E. (n ϭ 3). C, human skin fibroblasts were co-transfected with 1 g of Ϫ517/ϩ60 MMP-1 in pGL3basic or 3TP-Lux in combination with 1 g of Smad7 expression construct or empty vector. Cells were serum-starved for 24 h and then stimulated with TGF-␤ (2 ng/ml), PMA (10 Ϫ7 M), or both together. Cell lysates were harvested at t ϭ 8 h and assayed for luciferase activity. Data are expressed as -fold induction above the control (empty vector only), mean and S.E. (n ϭ 3).  (n ϭ 3). B, human skin fibroblasts were co-transfected with 1 g of Ϫ517/ϩ60 MMP-1 in pGL3-basic in combination with 1 g of either S4BE or the mutant oligonucleotide, which does not bind Smads (mS4BE). Cells were serum-starved for 24 h and then stimulated with TGF-␤ (2 ng/ml), PMA (10 Ϫ7 M), or both together. Cell lysates were harvested at t ϭ 8 h and assayed for luciferase activity. Data are expressed as -fold induction above the control (empty vector only), mean and S.E. (n ϭ 3). each oligonucleotide (as shown in Fig. 2). No supershift band could be seen in the additional presence of the anti-Smad 4 antibody (data not shown). However, using the labeled S4BE as probe, the MMP-1 AP1 oligonucleotide competes band A (the Smad-containing complex) at a 100-fold excess, whereas the Timp-1 AP1 oligonucleotide does not (Fig. 9). Intriguingly, the Timp-1 AP1 oligonucleotide competes band E, but because self-competition with cold S4BE shows this band as nonspecific, this finding is difficult to interpret (Fig. 9). DISCUSSION TGF-␤ has profound effects on extracellular matrix homeostasis, in part via its ability to alter the balance between proteinases and their inhibitors at the level of gene expression (7,8). The intracellular signaling pathways by which TGF-␤ mediates its actions are diverse. The Smad pathway, specific to the TGF-␤ family, is probably of prime importance, but many reports implicate other pathways (e.g. mitogen-activated protein kinase pathways (14,15)); furthermore, cross-talk between the Smads and other signaling pathways modifies the TGF-␤ response (16 -18). The reciprocal effect of TGF-␤ on the expression of TIMP-1 and MMP-1 initially described by Edwards et al. (11) supports its role in matrix anabolism. The mechanisms by which TGF-␤ induces Timp-1 yet represses induced MMP-1 have remained opaque; hence, the current study sought to address this, with a focus on Timp-1 gene expression.
The promoter requirements for TGF-␤-induction of the Timp-1 gene were ascertained using deletion mutants, demonstrating that the proximal (Ϫ59/Ϫ53) AP1 site plays a major role. Internal substitutions across Ϫ50 to ϩ47, leaving the AP1 site intact in the context of Ϫ223/ϩ47 Timp-1, failed to demonstrate further elements necessary for TGF-␤ induction (data not shown). This was reinforced by EMSA on overlapping oligonucleotides across Ϫ50 to ϩ47, which showed no altered protein-DNA interactions. EMSA on an oligonucleotide containing the AP1 sequence demonstrated an increase in binding upon either TGF-␤ or PMA stimulation with the only significant difference between the two being the induction of Fra-1containing complexes by PMA but not TGF-␤. Analysis of the contribution of Fos and Jun family members using both overexpression and cell lines from AP1 knockout mice revealed a requirement for c-Fos, c-Jun, and JunD in TGF-␤ induction of Timp-1, whereas c-Fos was not essential for PMA induction of the gene. Smad co-expression experiments show that the Timp-1 gene is not Smad-responsive; nor is TGF-␤-induction of the gene blocked by the inhibitory Smad 7. These facets are shown very clearly in the control plasmid 3TP-Lux. Finally, co-transfection of a Smad-binding oligonucleotide has no effect on TGF-␤ induction of the Timp-1 gene. From all of these data, it can be concluded that the induction of Timp-1 gene expression by TGF-␤1 is AP1-but not Smad-dependent. This firm conclusion must be tempered by the data coming from Smad knockout cells, where induction of Timp-1 is still evident, but its magnitude is reduced, compared with wild-type cells. However, it should be reiterated that the magnitude of response to TGF-␤1 varies among wild-type cell lines and also that the absence of Smads may have secondary consequences to cellular function that are separate from direct effects on Timp-1 expression.
A recent report by Verrecchia et al. (31) states that the Timp-1 gene is Smad-dependent, in disagreement with the majority of data in the current study. The following lines of evidence are presented: (i) TGF-␤ induces a greater than 2-fold induction of the gene within 30 min as assessed by cDNA microarray analysis, and this is not blocked by the protein synthesis inhibitor cycloheximide or the c-Jun N-terminal kinase inhibitor curcumin; (ii) a Timp-1 promoter construct driving CAT expression is induced by TGF-␤ at a 24-h time point, and this is blocked by dominant negative Smad 3 or by Smad 7; and (iii) co-expression of Smad 3 with the Timp-1 promoter construct mimics the effect of TGF-␤, and no promoter transactivation is seen in Smad 3 Ϫ/Ϫ cells. There are many differences in these data compared with the current study. (i) The microarray data in Verrecchia et al. (31) shows the TGF-␤induction of Timp-1 gene expression maximal at 30 -60 min, remaining at this level at 4 h. In our hands, in both murine fibroblasts and human fibroblasts (dermal or lung), TGF-␤ induction of Timp-1 becomes maximal at 12-24 h, as assessed by Northern blot (12). 2 (ii) The use of cycloheximide as a protein synthesis inhibitor in such studies is problematic, since it has been shown to augment the induction of immediate early genes such as c-fos and c-jun by growth factors (38) via the p38 mitogen-activated protein kinase pathway (39); indeed, Verrecchia et al. (31) state that cycloheximide treatment causes "a broad increase in gene expression . . . " Using emetine as a protein synthesis inhibitor, the induction of Timp-1 by TGF-␤ is dependent on new protein synthesis (data not shown). 2  Inhibitors of the mitogen-activated protein kinase pathways such as curcumin (c-Jun N-terminal kinase), U0126 (extracellular signal-regulated kinase), or SB202190 (p38) all block the TGF-␤ induction of Timp-1 to varying extents in the C3H10T1/2 cells used in the current study (data not shown). (iv) The promoter studies in Verrecchia et al. (31) use an undisclosed length of murine promoter driving CAT expression in human cells measured at a 24-h time point with no serum starvation prior to TGF-␤ treatment (Timp-1 is serum-responsive (40)); the studies above use constructs driving luciferase expression measured at a 6-h time point to avoid potential secondary responses of the cells to TGF-␤-induced growth factors. (v) Smad 3Ϫ/Ϫ cells, in our hands, were difficult to transfect; however, as discussed above, the endogenous Timp-1 gene is these cells is still induced by TGF-␤, albeit at a reduced level.
The dependence of Timp-1 gene expression on AP1 factors has been described in other systems and with other inducing agents (e.g. Botelho et al. (41) ascribe the induction of Timp-1 by oncostatin M to an induction of c-Fos and a change in the major AP1-binding complex from c-Jun/c-Fos to JunD/c-Fos in HepG2 cells; Smart et al. (42) demonstrate that JunD, Fra2, and FosB associate with the TIMP-1 AP1 site during hepatic stellate cell activation, of which JunD is functionally the most important. The current study indicates that c-Fos, Fra2, FosB, c-Jun, JunD, and JunB are all present in TGF-␤-induced cell nuclear extracts and can bind the Timp-1 AP1 sequence on EMSA (presuming the specificity of antibodies used in supershift experiments). An overall increase in AP1 binding is observed upon TGF-␤ treatment compared with control. Whereas c-Fos, c-Jun, and JunD were shown to be necessary for Timp-1 induction using knockout cell lines, JunB and Fra2 have not been assessed in this manner, since cells were not available. It is possible that members of the cAMP-response element-binding protein family of transcription factors, which can heterodimerize with AP1 factors, are also important in the response of Timp-1 to TGF-␤; indeed, ATF2 is a target of TGF-␤ signaling via both the Smad pathway and TAK1/p38 (43).
The AP1 motif in the Timp-1 promoter differs at a single base pair from the MMP-1 consensus (5Ј-TGAGTAA-3Ј compared with 5Ј-TGAGTCA-3Ј), with the only known consequence being binding of an unknown single-stranded DNA-binding protein to the former but not the latter (23). However, substitution of the wild-type site for the consensus in a Timp-1 promoter construct does not alter the response to TGF-␤; this suggests either that flanking sequence around the AP1 site is important, as suggested by Zhang et al. (27), and/or that interactions with other transcription factors binding at distal sites are involved.
TGF-␤ represses PMA-induced MMP-1 expression in human skin fibroblasts. Our data on Timp-1 expression were gained using the murine system, allowing us to use a well characterized cell line (C3H10T1/2) for TGF-␤ response, maintaining parity with our earlier studies and enabling us to exploit cell lines from knockout animals; however, the endogenous MMP-1 gene in mice is controlled very differently from that in humans (44). Hence, experiments using MMP-1 promoter constructs were performed both in human skin fibroblasts and in murine Swiss 3T3 cells, where the same pattern of response was observed; TGF-␤ is unable to repress PMA-induced expression from MMP-1 promoter constructs in C3H10T1/2 cells.
Deletion from a Ϫ517/ϩ60 MMP-1 promoter construct demonstrates that TGF-␤ repression is maintained, even down to Ϫ80/ϩ60; this suggests that the putative TIE at Ϫ245 is not necessary for this response. This is reinforced by a point mutation in Ϫ517/ϩ60, where mutation at the TIE has no functional consequence, but mutation in the AP1 site at Ϫ72 abrogates TGF-␤-mediated repression. A previous report (9) using the rabbit MMP-1 promoter conflicts with this, demonstrating that AP1 mutation has no effect, whereas TIE mutation abrogates TGF-␤ repression. In agreement, both groups demonstrate that AP1 mutation severely reduces the level of transgene expression seen and that TIE mutation increases basal expression. Potentially, this is a species difference (human versus rabbit) or a cell type difference (dermal fibroblasts versus synovial fibroblasts); the MMP-1 AP1 motif binds at least c-Fos, JunD, and Fra2 from PMA-stimulated rabbit synovial fibroblasts (22).
The importance of the AP1 motif in the TGF-␤ response of the MMP-1 gene was demonstrated by Mauviel et al. (45), whence TGF-␤ repressed MMP-1 expression in dermal fibroblasts via JunB-containing AP1 complexes, whereas TGF-␤ induced MMP-1 expression in epidermal keratinocytes via c-Jun-containing AP1 complexes.
The interaction of Smads with AP1 factors has recently been reported by several groups (27)(28)(29). Using EMSA, no supershift was detected using anti-Smad2/3 or anti-Smad4 antibodies on an AP1 shift; the nuclear extracts used show Smad binding to a canonical Smad-binding oligonucleotide (S4BE) using the same methodology. Intriguingly, however, the MMP-1 AP1containing oligonucleotide (but not the Timp-1 equivalent) does compete for binding of the Smad-containing complex to the S4BE. The functional relevance of this is shown in co-transfection studies, where either Smad 7 or the wild-type S4BE oligonucleotide relieve the TGF-␤ repression of PMA-induced MMP-1 expression. Hence, our data suggest that the TGF-␤mediated repression of PMA-induced MMP-1 expression is Smad-dependent and mediated through the promoter-proximal AP1 site.
In agreement with this, a recent report by Yuan and Varga (30) shows the Smad dependence of TGF-␤ repression of interleukin-1-induced MMP-1 expression. In this study, Smad3 overexpression mimics the TGF-␤-mediated repression of MMP-1, although this appears to be mediated via competition between NF-B and the coactivator p300, using a cis-acting element distinct from the AP1 site; this difference is presumably due to the use of IL-1 to stimulate MMP-1 compared with PMA used in the current study. The current study suggests that the MMP-1 AP1 sequence competes for Smad binding, and potentially, competition between Smads and AP1 for binding to this site could underlie TGF-␤-mediated repression. However, no evidence for this mode of action comes from the EMSA studies above, and the increase in AP1 binding upon TGF-␤ treatment suggests otherwise.
TGF-␤ induces expression of the MMP-13 gene, and this is at least partly via the AP1 site (5Ј-TGACTCA-3Ј) (46) and an increase in c-Fos, c-Jun, and JunD binding. Furthermore, the MMP-13 gene cannot be induced by either epidermal growth factor or platelet-derived growth factor in c-fos-deficient cells (47). The group of Kahari (48,49) has demonstrated that TGF-␤ induction of MMP-13 is dependent on both the p38 mitogen-activated protein kinase pathway and the Smad pathway.
In conclusion, the ability of TGF-␤ to induce Timp-1 and repress induced MMP-1 expression is dependent on a promoter-proximal AP1 motif in each case. Contrary to the repression of MMP-1, which is Smad-dependent, induction of Timp-1 does not involve the "classical" Smad pathway; however, data from Smad-deficient cell lines suggest some role for Smads that may involve indirect effects on the repertoire of transcription factors expressed by such cell lines. Future work will focus on dissecting the signaling pathways that link TGF-␤ to the Timp-1 gene.