Transforming Growth Factor + Stimulates a 2 ( I ) Collagen Gene Expression through a cis-Acting Element That Contains an Spl-binding Site *

Transforming growth factor-pl (TGF-P) is a strong and rapid inducer of several genes coding for extracellular matrix components, such as type I collagen. We report here that TGF-P stimulates transcription of the human (u2(I) collagen gene (COLlA2) promoter by increasing the affinity of an Spl-containing protein complex for its cognate DNA-binding site. Cell transfection experiments  mapped  the  TGF-P-responsive lement (TbRE) of the COLlA2 promoter to a 131 bp region that contains at least two cis-acting elements. Insertion of the TbRE upstream of the thymidine kinase promoter conferred TGF-f.3 inducibility to the otherwise unresponsive thymidine kinase promoter. Footprinting assays revealed that the TbRE contains two neighboring proteinbound sequences, termed Box 3A and Box B. Within Box 3A is an Spl recognition sequence whose structural integrity is required for nuclear protein binding in vitro, and for promoter inducibility in uiuo. Gel mobility shift assays documented increased binding to the TbRE of nuclear proteins from TGF-P-treated cells compared with those from TGF-P-untreated fibroblasts. There was, however, no binding increase with Box 3A alone or with an Spl oligonucleotide. Thus, the results strongly suggest a functional interaction between Spl and other components of the TbRE complex in mediating TGF-/3 stimulation of COLlA2 gene expression.

fibrotic states (3,4). In this respect, one of the most extensively studied effector-target relationships is the one between transforming growth factor-pl (henceforth referred to as TGF-P) and type I collagen.
TGF-P is a member of a group of related proteins which promote, inhibit, and modulate a wide variety of cellular activities and morphogenetic processes, including ECM elaboration (5). TGF-P influences both ends of the dynamic balance that is normally maintained between synthesis and degradation of the ECM (6). This is achieved by coordinating the expression of several functionally distinct but biologically related gene products (7). TGF-P stimulates the synthesis of most of the structural ECM components, as well as the expression of the cognate cell surface receptors. It also inhibits the production of proteolytic enzymes that catalyze ECM degradation, while enhancing the expression of protease inhibitors. As a result of these multifaceted effects, TGF-P is believed to play a critical role in a variety of developmental programs and physiopathologic processes, such as wound healing, tissue repair, and fibrosis (8-10).
Part of the TGF-P action on gene expression is exerted at the transcriptional level. The amount of information regarding the factors and mechanisms that mediate the transcriptional regulation of TGF-P is still somewhat limited (11). For example, the c-fos gene product (Fos) and nuclear factor-1 (NF-1) have been implicated in the inhibition of the rat transin gene and in the stimulation of the mouse a2(I) collagen gene (Colla2), respectively (12,13). Transin down-regulation by TGF-P occurs through the induction of Fos and its subsequent binding to DNA, presumably as a component of a multimeric protein complex (12). The Fos-containing protein complex binds to a 10-bp cis-acting element, termed TIE, which lacks a bona fide AP-1 site (12). In contrast to transin gene inhibition, the mechanism underlying TGF-P stimulation of the mouse Colla2 gene through the CTFiNF-1-binding site is unknown (13). In this report, we demonstrate that the TGF-P-responsive element (TbRE) of the human COLlA2 promoter maps to a different cis-acting element that contains a binding site for specificity protein 1 (Spl). We also present evidence suggesting how TGF-P action is eventually translated into transcriptional stimulation of collagen gene expression.

MATERIALS AND METHODS
Cell Culture and Dansfection Experiments-Primary human fetal skin fibroblasts (CF-37) were grown and maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (HyClone, Logan, UT). Preparation and transfection of plasmid DNA into cells chain, respectively; CAT, chloramphenicol acetyltransferase; CTF, CAA'IT-binding transcription factor; NF-1, nuclear factors-1; PCR, polymerase chain reaction; Spl, specificity protein l; TGF-p, transforming growth factor-p; TIE, TGF-/3 inhibitory element; TNF-a, tumor necrosis factor-a; TbRE, TGF-p-responsive element; bp, base pairb). using the calcium phosphate precipitation method were performed as previously described (14). Five h after transfection, fibroblasts were subjected to a 15% glycerol shock for 105 s and then placed in medium containing 0.1% fetal calf serum. Two h later, human platelet-derived TGF-P (Collaborative Biomedical Products, Bedford, MA) was added at the final concentration of 2 ng/ml. Cells were harvested 48 h after transfection, lysed, and subjected to luciferase and chloramphenicol acetyltransferase (CAT) assays according to published protocols (15,16). Transcriptional activity of each of the chimeric constructs was normalized against cotransfected pSV2CAT and pTKGH plasmids, respectively (15,17). Measurement of human growth hormone concentration in the medium was assessed according to manufacturer's recommendations using a commercial kit (Nichols Institute, Los Angeles, CA). The Mann-Whitney U test was used to determine the statistical value of the functional data.
Chimeric Plasmids-Some of the deletion plasmids are luciferase gene constructs derived from previously described COLlA2KAT plasmids (14). Additional sequence manipulations followed standard cloning techniques (18) or employed the polymerase chain reaction (PCR) amplification of cloned DNA using appropriate combinations of synthetic primer oligonucleotides (19). Integrity of wild-type and mutant sequences was verified by DNA sequencing of each construct (18). In the mutant constructs, the wild-type COLlA2 sequence was substituted by unrelated polylinker sequences (mutants M2, M3, M4, and M5 in Fig.  1 B ) after joining together subclones of synthetic double-stranded oligonucleotides. Aside from pSV2CAT and pTKGH, other vectors used in the engineering of the chimeric constructs were the pBLCAT2 plasmid and the luciferase vector pXP[ll (20,21). Fragments excised from the upstream sequence of the COLlA2 promoter and synthetic oligonucleotides were cloned into the polylinker sequence of the two plasmids (18).
DNA Binding Assays-For the preparation of the nuclear extracts from untreated and TGF-P-treated fibroblasts, cells were placed in medium containing 0.1% fetal calf serum 24 h prior to TGF-P treatment. After incubation without and with 2 ng/ml of TGF-P for the indicated length of time, cells were harvested and nuclear extracts were prepared at 4 "C following our previously described modification (22) of the Morris et al. (23) protocol. DNase I footprinting and gel mobility shift assays were performed as previously described (22). Sole exception was the binding condition for the CTF/NF-1 and Boxes A and 5A probes, which utilized 150 m~ KC1 in place of 80 m M NaC1, and required the presence of 5 m M MgC1,. In experiments testing the ionic requirements of DNAprotein binding, the binding buffer was adjusted to 5 m~ EDTA or to 20 p~ ZnC1,. The sequences of the four high affinity CTFNF-1 oligonucleotides used in the competition assays have been already described (22). Additional competing oligonucleotides were 5'-GATCGAACTGACCGC-CCGCGGCCCGT-3' (AP-2) and 5'-TGGGCGGAGTTAGGGGCGGGAT-GGGCGGAGTTAGGGGCGGGA-3' (Spl). Boundaries of the COLlA2 oligonucleotides used in the DNA binding assays are from -330 to -286 (Box A), from -330 to -297 (Box 5A), from -313 to -286 (Box 3A) and from -313 to the BstXI site at -183 (TbRE) (Fig. 1). Mutations in the Spl recognition sequence of Box 3A (sequence MI in Fig. 1 B ) were created by synthesizing complementary oligonucleotides covering the -313 to -281 sequence (Fig. 1B). These were subsequently purified and sequenced prior to be used in DNAbinding assays (18). Antisera against NF-1 and AP-2 were generously provided by Drs. Burrow (Cleveland Clinic Foundation, Cleveland, OH) and Williams (Yale University, New Haven, CT), respectively. The recombinant NF-1 protein was a gift of Dr. Gronostajski (J. Hopkins University, Baltimore, MD). Spl antiserum and recombinant Spl protein were purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and Promega Corp. (Madison, WI), respectively. Identical amount of nuclear extracts from untreated and TGF-P-treated CF-37 fibroblasts were used in the gel mobility shift assays that analyzed the time course effect of TGF-P treatment.

RESULTS
Mapping the TGF-P-responsive Sequence-We have previously shown that the human COLlA2 promoter sequence extending from nucleotides +58 to -3500 can replicate the -4-fold enhancement of endogenous gene expression in response t o TGF-P (24). To locate the DNA segment that mediates most of this stimulation, a series of plasmids were constructed containing progressive 5' deletions of the COLlA2 upstream sequence linked to the luciferase reporter gene. The resulting COL1A2I LUC constructs were transfected into human dermal fibroblasts (CF-37) subsequently cultured with or without TGF-p. Initial tests showed that removal of the sequence between nucleotides -3500 and -378 did not reduce the level of stimulation by TGF-P (Fig. IA). In contrast, TGF-P failed to stimulate luciferase gene expression of either the -183COLlA2/LUC or the -108COLlA2ILUC plasmids (Fig. lA). The data therefore suggested that response of COLlA2 gene expression to TGF-p is mediated by cis-acting DNA element(s1 located between nucleotides -378 and -183 of the collagen promoter. This conclusion is in agreement with previous results of similar experiments using the promoter of the mouse Colla2 gene (13).
In the next set of experiments, we employed the DNase I footprinting assay to identify sites of DNA-protein interaction within the -378 to -183 promoter segment. This revealed two distinct areas of nuclease protection (Fig. 2). The proximal footprint (Box B ) extends from nucleotides -271 to -255 and is separated by 14 bp from the distal footprint (Box A), which extends from nucleotides -330 to -286 (Fig. 1B). Box A corresponds to the mouse sequence where an NF-1 binding activity associated with TGF-P inducibility has been previously located (13).
In light of the DNase I footprinting data, we evaluated the functional contribution of each of the two boxes to TGF-P responsiveness. Initial assays were performed using -378COLlA2/LUC constructs containing internal deletions of Box A (AACOLIAZILUC) or Box B (ABCOLlA2ILUC) (Fig. lA). Deletion of the Box A containing region reduced the ability of AACOLlA2ILUC to respond to TGF-0 as efficiently as the positive control, plasmid -378COLlA2/LUC (Fig. IA). A similar reduction in TGF-p inducibility has been previously observed with a comparable mouse Colla2 promoter construct (13). Deletion of the Box B containing region brought TGF-P responsiveness of ABCOLlA2/LUC to the same level as the negative control, plasmid -183COLlA2/LUC (Fig. LA). There is currently no available information about the possible partici- pation in the TGF-P response of the mouse counterpart of Box B (13).
In order to provide additional functional evidence, subsequent assays were performed using a promoter unresponsive to TGF-P. One, two, and four copies of Box Aor Box B, and one and two copies of both boxes were inserted upstream of the thymidine kinase (TK) promoter in a CAT expression vector. Resultant plasmids were then analyzed for TGF-P responsiveness ( Table I). The data showed that either Box A (-330 to -286, see  (Table I). In contrast, plasmids with the BglII-BstXI promoter fragment containing both boxes (Fig. lA) exhibited -&fold induction in CAT activity in response to TGF-P (Table I).
The functional tests concurred in suggesting that most of the TGF-P-mediated stimulation of COLlA2 transcription is correlated with the interaction of at least two cis-acting elements located between nucleotides -378 and -183. They also revealed some interesting differences on how Box B influences basal transcription of the COLlA2 and thymidine kinase promoters. Unlike the former, inclusion of Box B within the thymidine kinase promoter context increased transcription -10-fold and in a copy-dependent manner (Table I). This finding is identical to that of Karsenty et al. (25) in the mouse gene. These authors previously showed that the mouse counterpart of Box B acts like an enhancer when out of its natural sequence context. It is also consistent with the established concept that artificial juxtaposition of heterologous promoter sequences often results in the activation of cryptic DNA-protein-binding sites, and/or in the creation of new ones. Along these lines, inclusion of Box A in the Box B-thymidine kinase promoter construct restored the natural interaction between the two COLlA2 elements; as a result, it abrogated the enhancing effect of Box B, while conferring TGF-P inducibility to an otherwise unresponsive promoter (Table I).
Box A Contains an Spl-binding Site-In the next set of experiments, we employed the gel retardation assay to examine the nuclear factors that bind to the COLlA2 sequence. Incuba- Based on the Mann-Whitney U test, they are significantly higher than the ratio value of pBLCAT2 ( p < 0.01). tion of nuclear extract from untreated fibroblasts with radiolabeled Box A oligonucleotide (-330 to -286) yielded several retarded bands (Fig. 3 A , lane I ) . Competition experiments documented the specificity of these DNA-protein complexes. They in fact showed the disappearance of the retarded bands when the unlabeled Box A oligonucleotide was added in 100fold molar excess ( Fig. 3 A , lane 2 ) .
The complexity of the binding pattern suggested to us that more than one nuclear factor might be interacting with Box A.
To test this hypothesis, we used oligonucleotides covering the 5' (-330 to -297) and 3' (-313 to -286) halves of Box A (Fig. lB) either as competitors or as probes for nuclear protein binding. Incubation of the nuclear extract with radiolabeled Box Ain the presence of 100-fold molar excess of unlabeled 5' or 3' oligonucleotides eliminated the faster (complex C2) or the slower (complex C1) migrating bands (Fig. 3 A , lanes 3 and 4), respectively. In the converse experiment, incubation of the nuclear extract with radiolabeled 5' or 3' oligonucleotides gave rise to complex C2 or complex C1 (Fig. 3A, lanes 5 and 6), respectively.
The presence of a high affinity Spl-binding site prompted us to explore the possibility that this transcription factor, or a related protein, may interact with Box 3A. Toward this end, we first examined the ionic and structural requirements responsible for proper formation of complex C1. We found an increase in the intensity of complex C 1 with the addition of Zn2' and a decrease with EDTA (Fig. 3B, lanes 2 and 3 with lane 1). The strong dependence of complex C1 formation on the cation concentration of the binding buffer is in agreement with the established requirements for Spl binding (26). Furthermore, we found that nucleotide substitutions known to interfere with Spl binding (27,28) affected protein interaction with Box 3A as well. To be precise, a 33-bp oligonucleotide (-313 to -281) containing mutations in the duplicated Spl sites of Box 3A (-305 to -291, sequence MI in Fig. lB) failed to compete protein binding to Box 3A (Fig. 3B, compare lane 5 with lanes 4 and  6).In addition, the same mutated sequence failed to interact with CF-37 nuclear proteins (Fig. 3B, compare lane 8 with lane   7). Three additional lines of evidence confirmed these results, thus strengthening the notion that Spl binds to the human COLlA2 promoter element. First, an oligonucleotide containing a high affinity Spl recognition sequence competed binding of complex C1 to Box 3A (Fig. 4A). Second, incubation of a nuclear extract with Spl antiserum decreased substantially the formation of complex C1 (Fig. 4B). Third, recombinant Spl protein bound effectively to the Box A probe (Fig. 4 0 . In contrast to the above results, the same experimental approach demonstrated that NF-1 does not interact with the se- quence of human Box A. First, unlabeled oligonucleotides containing different high afinity CTF/NF-1-binding sites failed to compete the formation of either complex C1 or complex C2 (Fig.  5A ). Conversely, an excess of unlabeled Box A oligonucleotide did not affect protein binding to a wild-type CTF/NF-1 recognition sequence (Fig. 5A). Second, incubation of CF-37 nuclear extract with NF-1 antiserum altered the protein complex formed with the CTFNF-1 sequence, but not that formed with the Box A probe (Fig. 5B). Third, recombinant NF-1 protein interacted with the cognate recognition sequence but failed to bind to the Box A probe (  Fig. 6A). It should be noted that mutation M2 deliberately eliminated the Spl-binding site of Box 3A, as well as the cryptic Spl sequence located in the segment between Boxes 3A and B (Fig. lB 1. Because of the necessity of mutating both sites (27,28), we also tested the possible consequence of altering only the composition of the intervening segment (sequence M3 in Fig. ll?). The results showed that construct M3 exhibits the same level of TGF-/3 responsiveness as the wild-type plasmid (Fig. 6A). Hence, the different behavior of the M2 and M3 mutant plasmids implies that optimal response to TGF-p depends on the structural integrity of the Spl-binding site of Box 3A. Deletion of the sequence immediately upstream of Box 3A had no effect on COLlA2 promoter inducibility, since the seemingly high degree of inducibility of construct A5A was not statistically different from that of plasmid WT (Fig. 6A). Once more, this finding excluded the participation of complex C2 in mediating TGF-p stimulation. On the other hand, the relatively higher basal activity of A5A compared with WT suggested that Box 5A might conceivably represent a negatively cis-acting element (Fig. 6A).
The functional assays therefore defined the shortest promoter sequence that mediates most of the TGF-p-elicited upregulation of the human COLlA2 promoter. This sequence, This showed that TbRE confers the same degree of TGF-p inducibility as the larger BglII-BstXI fragment to the thymidine kinase promoter (compare results of Fig. 6B and Table I). The single asterisk indicates that the ratio value is statistically lower than that of plasmid WT (Mann-Whitney U test,p < 0.01), while the double asterisk signifies that the ratio value is statistically higher than that of plasmid pBLCAT2 (Mann-Whitney U test,p < 0.05).

S p l and Collagen Stimulation by TGF-p
Consistent with the postulated negative activity of Box SA, we also noted that basal transcription of the TbRE-containing thymidine kinase construct is nearly 3-fold higher than that of the thymidine kinase plasmid containing Boxes A and B (compare results of Fig. 6B and Table I). TGF-/3 Increases the Binding Affinity of the TbRE Protein Complex-It has been reported that the steady-state level of a2(I) collagen mRNA increases severalfold during the first few hours of TGF-/3 treatment in a cyclohexamide-independent manner (29). In our initial DNase footprinting assays, we noted that nuclear extracts from TGF-/3-treated cells consistently yielded clearer data (in terms of the extent and intensity of the protected areas) than nuclear extracts from untreated cells (Fig. 2). This was a novel finding, for such a change had not been previously observed in the mouse study (13). To investigate this important point further, we used the more sensitive gel retardation assay and nuclear extracts prepared a t different time points during the first 6 h of TGF-/3 treatment.
Numerous retarded bands were visualized with the 131-bp TbRE probe using an untreated nuclear extract (Fig. 7A, lane  1). Competition with unlabeled Spl oligonucleotide and treatment with Spl antiserum identified the bands corresponding to the Spl complex (Fig. 7, lanes 5-7). Concomitant with TGF-/3 treatment, and thus parallel to endogenous gene stimulation, we observed a gradual increase in the binding affinity of the TbRE protein complex, including Spl (Fig. 7A, lanes 2 4 ) . Surprisingly, however, TGF-/3 treatment had no effect on nuclear protein binding to the Box 3A or to the Spl probe (Fig. 7, B and   C). The results therefore indicate that the TGF-/3-elicited increase of Spl binding can only occur when this transcription factor is part of a larger protein complex. In turn, this observation raises the intriguing possibility that Spl-specific coactivators might be involved in mediating this effect.
Preliminary Analysis of Box B-The above results implicated the sequence extending downstream of Box 3A, and thus including the Box B footprint, in participating in the formation of  (lanes 1 ) and after 1  h (lanes 2 ) , 3 h (lanes 3 ) and 6 h (lanes 4  the TbRE complex. In order to evaluate more rigorously the role of Box B, we introduced five nucleotide substitutions in the highly conserved 5' half of Box B (sequence M5 in Fig. lB).
Transfection of the resulting M5 plasmid revealed that the mutation causes a significant reduction in TGF-/3 responsiveness, compared with the reference WT plasmid (Fig. 6A).
In the second set of experiments, we attempted to evaluate nuclear factor binding by the gel mobility shift assay using a Box B-containing oligonucleotide (-271 to -235). However, we soon realized that cryptic sites lying outside of the Box B footprint (-271 to -255) were preferentially bound in the absence of the Box 3Asequence (data not shown). We strongly believe that these are the same elements responsible for the enhancer-like activity of Box B in the thymidine kinase promoter constructs. Because of this experimental problem, we used the alternative approach of competing nuclear protein binding to the TbRE. Accordingly, we challenged binding of proteins from untreated nuclear extract with increasing molar excess of the Box 3A and Box B oligonucleotides, as well as the control TbRE sequence. The results of this gel mobility shift assay revealed that the three oligonucleotides affect the TbRE-bound complex differently. First, the TbRE oligonucleotide interferes with binding of Spl to Box 3A more efficiently than Box 3A itself (Fig. 8). Incidentally, the same thing was also observed for the faster moving band of the TbRE-bound complex (band ZZ in Fig. 8). This finding conceivably indicates that the postulated Spl coactivators may also participate in stabilizing the TbRE-bound complex. Second, we found that Box B begins to compete bands I and I1 only a t 400-fold molar excess (Fig. 8). We believe that the result is consistent with the hypothesis of a complex-stabilizing interaction between Boxes 3A and B, as well as with the notion of low affinity protein binding to Box B. The latter postulate is also supported by the artifactual utilization of stronger protein binding sites when the Box B is used by itself in the gel mobility shift assay.
Although the validity of this hypothesis must await experimental confirmation, the results nevertheless suggest that TGF-P exerts its stimulatory effect on COLlA2 transcription through a multimeric protein complex. Aside from Spl, or an Spl-related factor, this multimeric complex is likely to include proteins that bind to Box B and/or interact with Spl. Finally, the TbRE-bound complex responds to the TGF-/3 stimulation by increasing its binding affinity.

DISCUSSION
It is becoming increasingly evident that TGF-P plays a critical role in modulating type I collagen production in both physiologic and pathologic states. For example, the pattern of type I collagen gene expression during mouse embryogenesis has been correlated with extracellular localization of the TGF-P protein (30). Similarly, TGF-P has been reported to be present in the same foci of activated fibroblasts which are responsible for most of type I collagen deposition in idiopathic pulmonary fibrosis (31). Finally, TGF-P has been shown to cause an immediate and persistent increase in steady-state type I collagen mRNAs in cultured cells (29). This response is independent of protein synthesis, sensitive to actinomycin D, and mostly related to elevated gene transcription (24, 29).
Stimulation of mouse Colla2 gene expression was originally associated with an NF-1 binding activity (13). However, that study failed to detect a difference between nuclear extracts prepared from untreated and from TGF-P treated cells. As a result, Rossi et al. (13) left unresolved the issue of how NF-1 mediates the stimulating effect of TGF-P. The work presented in this report differs from the mouse data in two respects. First, it convincingly excludes NF-1 and implicates Spl in mediating TGF-p stimulation. Because of the lack of relevant mouse data, we cannot presently reconcile this discrepancy. Second, and more importantly, our results provide the first mechanistic insight into how TGF-P action is translated in transcriptional stimulation of gene expression.
Functional assays identified the sequence responsible for most of the effect of TGF-P in the activation of the human COLlA2 promoter. This sequence, termed TbRE, is 131-bp long and contains two neighboring cis-acting elements, Box 3A and Box B. It should be noted that other cis-acting elements not identified by our DNA-binding assays may lie within this 131-bp promoter fragment. Aside from the negative consequences of the TbRE mutations on TGF-/3 inducibility, the specificity of the transcriptional response was further corroborated by the ability of the COLlA2 sequence to confer TGF-P inducibility to the otherwise unresponsive thymidine kinase promoter. Three lines of evidence suggested that the ubiquitous activator S p l or a related protein, binds to Box 3A. First, protein binding was competed by an oligonucleotide containing a high affinity S p l consensus recognition sequence. Furthermore, nucleotide substitutions in the Spl-binding sites of Box 3A prevented protein binding in vitro and abrogated TGF-P inducibility of the COLlA2 promoter in vivo. Second, formation of the Box 3A-protein complex showed a strict dependence on the concentration of Zn'+ and EDTA in the binding buffer.
Third, recombinant S p l protein interacted with the Box 3A sequence and, conversely, S p l antiserum reduced nuclear protein binding to Box 3A.
Additional assays revealed that TGF-P action is translated into a gradual increase in the binding affinity of the TbRE protein complex. Such an increase includes the Spl protein, but only when its cognate recognition sequence is part of the TbRE. It is presently unclear what is the contribution of the sequence downstream of Box 3A to the binding increment of the TbRE complex. Work in progress is addressing this important question. Based on the available data, we nevertheless propose that TGF-P stimulates collagen gene expression by changing the binding affinity of a pre-existing multimeric complex that contains Spl. We also postulate that TGF-P modifies Spl-specific coactivators. This hypothesis is supported by recent work showing that Spl-promoted transcriptional activation requires the participation of tightly associated co-activators (32,33). The proposed model for TGF-P action on collagen gene expression can be further extended to include TNF-a. A parallel study has in fact indicated that this cytokine utilizes the same transcriptional complex to inhibit COLlA2 gene expression.' Hence, it is possible that the antagonistic stimuli of TNF-a and TGF-P may be transcriptionally elaborated by changing the identity of the Spl-interacting proteinb), or the nature of their modification(s) or by a combination of both mechanisms.
The work of Kerr et al. (12) has previously delineated a different set of regulatory elements mediating TGF-P inhibition of gene expression. These authors have shown that TGF-p acts at an upstream site in the promoter of the transin gene by dominantly repressing its expression. The inhibition occurs through the stimulation of the TIE-binding complex, a Foscontaining multimeric repressor. Interestingly, TIE and TIElike sequences are found in the promoters of several other metalloproteinase genes which are also inhibited by TGF-p (12). Our study is the first to provide a detailed characterization of the factors and mechanisms implicated in TGF-/3 stimulation of gene expression. Along these lines, it is of interest to note that a sequence homologous to the TbRE is located immediately upstream of the second major start site of transcription of the TGF-p gene (34). This sequence is apparently responsible for mediating the stimulation of the TGF-P gene by its own product and by phorbol ester (34). We also note that a sequence closely resembling a potential Spl-binding site lies within the recently described TGF-P-responsive element of the rat al(1) collagen gene (35). It is therefore tempting to speculate that coordinated stimulation of the type I collagen genes, as well as feedback regulation of TGF-P gene expression might be all promoted by Spl-containing protein complexes. In this respect, an analogy can be drawn between the collagen and the a-actin gene, whose regulation is under the control of interacting protein complexes that include the tissue-specific MyoDl protein and the ubiquitous factor S p l (36).