ABSTRACT
Sp3 is an ubiquitously expressed transcription factor, closely related to Sp1 but, unlike Sp1, it often functions as a transcriptional repressor. In this study we investigated the role of Sp3 in regulating the transcription of the human [alpha]2(I) collagen gene. We show that Sp1 and Sp3 specifically bind to three of the previously characterized cis-elements in this promoter, including two positive cis-elements between -303 and -271 and -128 and -123, and a repressor site between -164 and -159, but do not bind to the fourth cis-element bound by CBF. Functional analyses of Sp3 and Sp1 in Drosophila cells indicate that each protein transactivates the human [alpha]2(I) collagen promoter with equal potency and, when tested together, have an additive effect on the promoter activity. Furthermore, in vitro transcription assays demonstrate that both Sp1 and Sp3 are capable of supporting transcription from the collagen promoter independently of each other. However, when activities of both Sp1 and Sp3 are blocked with specific antibodies, in vitro transcription from this promoter is almost completely abolished. The results of this study demonstrate that Sp3 is as potent an activator of the human [alpha]2(I) collagen promoter as is Sp1 and that a transcriptional activity of the human [alpha]2(I) promoter is dependent on both proteins.
Collagen type I, the most abundant mammalian collagen, consists of two [alpha]1(I) chains and one [alpha]2(I) chain, which are coordinately expressed (1 ,2 ). Excessive deposition of type I collagen is characteristic of many fibrotic disorders (3 ) and most likely results from transcriptional activation of collagen genes in response to cytokines and other factors present in the prefibrotic/inflammatory lesions.
Sp1 is a well characterized transcription factor that is an important regulator of a variety of cellular and viral promoters (4 ). Recently, three Sp1-related proteins (Sp2, Sp3 and Sp4) have been characterized (5 ,6 ). Sp1 and Sp3 proteins are ubiquitously expressed at high levels in almost all tissues and mammalian cell lines, whereas Sp4 expression is restricted to certain cell types of the brain (5 ). Sp1, Sp3 and Sp4 have highly conserved DNA- binding domains and bind to DNA with similar specificity and affinity, whereas Sp2 has different binding specificities (6 ). It has been suggested that Sp3 is an inhibitory member of the family which works by repressing Sp1-mediated transcriptional activation (7 -9 ). Furthermore, it was proposed that this repression involves protein-protein interaction between Sp3 and components of the general transcription complex (10 ). However, in selected promoters such as sis/PDGF-B or p21Cip1/WAF1 promoters, both Sp1 and Sp3 can independently or additively activate transcription (11 ,12 ).
Previous studies have indicated that Sp1 plays an important role in regulating the activation of both human and mouse collagen type I genes (13 -16 ). Specifically, it has been shown that constitutive activity of the human [alpha]2(I) collagen promoter is mediated by Sp1 which binds to the three GC-boxes located between bp -303 and -271 (13 ,14 ), and to the TCCTCC motif located between -133 and -119 (Ihn et al., submitted). Two additional response elements which have also been characterized in the human [alpha]2(I) promoter include a repressor site between bp -173 and -155, and a CBF binding site between -101 and -72 (17 ,18 ). Binding proteins interacting with a repressor site between -173 and -155 have not been characterized in the human promoter.
This study was undertaken to further characterize the role of the members of the Sp1 family in regulating the human [alpha]2(I) collagen promoter. We report here that, in addition to Sp1, Sp3 interacts with several response elements in the human [alpha]2(I) collagen promoter. More importantly, we demonstrate that Sp3 is a potent transactivator of the human collagen promoter.
Human fibroblasts were obtained from foreskins of healthy newborns, following institutional approval and informed consent. Primary explant cultures were established in 25 cm2 culture flasks in DMEM supplemented with 10% fetal calf serum (FCS), 2 mM l-glutamine and 50 [mu]g/ml amphotericin. Fibroblast cultures independently isolated from different individuals were maintained as monolayers at 37oC in 90% air, 10% CO2, and studied between the third and sixth subpassages. Drosophila Schneider line 2 (SL2) cells were obtained from Dr Hsu (Medical University of SC) and propagated as previously described (19 ).
Drosophila Schneider cells were plated in 100-mm dishes at a density of 107 per dish in M3 medium supplemented with 10% FCS and 1 mM glutamine. The following day the cells were transfected using the calcium phosphate technique as previously described (14 ) with 10 [mu]g of various promoter-chloramphenicol acetyltransferase constructs and variable amounts of pPacSp0, pPacSp1 (19 ) and pPacSp3 and pPacUSp3 (kindly provided by G. Suske) expression plasmids. With pPacUSp3 only the long Sp3 protein is made in Drosophila cells due to the presence of ultrabithorax leader sequence at the 5' end (G. Suske, personal communication). Cells were incubated for 48 h, then harvested in 0.25 M Tris-HCl, pH 8, and fractured by freeze-thawing. Extracts were heated to 65oC for 5 min, and cell debris was pelleted for 10 min in a microfuge. Protein contents in supernatants were equalized using the Bio-Rad reagent and incubated with butyryl-CoA and [14C]chloramphenicol for 90 min at 37oC, assay conditions predetermined to be within the linear range of chloramphenicol acetyltransferase activity for these samples. Butyrated chloramphenicol was extracted using an organic solvent (2:1 mixture of tetramethylpentadecane and xylene) and quantitated by scintillation counting. Each experiment was performed in duplicate.
Nuclear extracts were prepared according to the method of Andrews and Faller (20 ). Briefly, confluent cells from five 150-mm dishes were washed with phosphate-buffered saline and scraped into 1 ml of cold buffer A (10 mM HEPES-KOH pH 7.9 at 4oC, 1.5 mM MgCl2, 10 mM KCl, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate). The cells were allowed to swell on ice for 10 min and then vortexed for 10 s. After centrifugation for 3 min the supernatant was discarded. The pellet was resuspended in 80 [mu]l of cold buffer C (20 mM HEPES-KOH pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 1 mM dithiotreitol, 0.7 mM phenylmethylsulfonyl fluoride, 1 mM sodium orthovanadate) and incubated on ice for 20 min for high-salt extraction. Cellular debris was removed by centrifugation for 2 min at 4oC and the supernatant fraction was stored at -80oC until use. The protein concentration of the extracts was determined using the Bio-Rad reagent.
Radioactive probes (132mer) were generated by polymerase chain reaction using [[gamma]-32P]ATP end-labeled primers or direct end labelling of oligonucleotide probes. For the DNA mobility shift assay, the binding reaction was carried out for 30 min in 20 [mu]l of binding buffer containing 10 000 c.p.m. of labeled probe, 2 [mu]g of poly(dI-dC)[middot]poly(dI-dC), and nuclear extracts containing 5 [mu]g of protein. Where indicated, specific antibodies were included in the reaction mixture. Polyclonal anti-Sp1, anti-Sp2, anti-Sp3 and anti-Sp4 antibodies were purchased from Santa Cruz, and preimmune rabbit IgG was purchased from Sigma. Specificity of the anti-Sp1 and anti-Sp3 antibodies was tested using blocking peptides purchased from Santa Cruz.Separation of free radiolabeled DNA from DNA-protein complexes was carried out on a 5% non-denaturing polyacrylamide gel in a 0.5* Tris borate electrophoresis buffer at 200 V at 4oC. Autoradiography was performed by overnight exposure to Kodak X-OMAT XAR2 film with intensifying screens at -80oC.
The reaction mixture for in vitro transcription contained 50 [mu]g nuclear extract, 1 [mu]g of template DNA [-772 COL1A2/CAT (14 )], 20 mM HEPES (pH 7.9), 6 mM MgCl2, 100 mM KCl, 0.2 mM EDTA, 0.5 mM DTT, 12% glycerol, 600 [mu]M of each of ATP, GTP, CTP and UTP in a final volume of 25 [mu]l. Nuclear extracts were preincubated with anti-Sp1 and/or anti-Sp3 antibodies (Santa Cruz) for 1 h at 4oC. The NTPs were added only after all other ingredients were preincubated for 15 min at 30oC, and then reactions allowed to proceed for 1 h at 30oC. Reactions were terminated by adding 175 [mu]l of stop solution (0.3 M Tris-HCl, pH 7.4, 0.3 M sodium acetate, 0.5% SDS, 2 mM EDTA, 3 [mu]g/ml tRNA). To detect the newly synthesized transcripts, antisense oligonucleotide primers corresponding to a sequence in the CAT gene were generated. These primers were end-labeled with T4 polynucleotide kinase and [[gamma]-32P]ATP, hybridized to in vitro transcription products and extended using avian myeloblastosis virus reverse transcriptase as previously described (11 ). The primer extended products were analyzed on 8% polyacrylamide-7 M urea gel. Gels were dried and autoradiographed at -80oC with an intensifying screen. The densities of bands were measured using phosphorimager scanner.
Several functional cis-regulatory elements located within the -303 and -34 region in the human [alpha]2(I) collagen promoter have been characterized (13 ,14 ,17 ). One of the cis-elements located between -303 and -271 has been shown to interact with a factor immunologically related to Sp1 (13 ,14 ). Since other Sp1-related proteins have similar binding specificity, we sought to determine whether other members of the Sp1 family also interact with this promoter region. Gel shift assays were performed with a collagen promoter fragment from -313 to -183 and nuclear extracts from human fibroblasts in the absence or presence of antibodies to various members of the Sp1 family. Four DNA-protein complexes were detected (Fig. 1 A). The same complex formation pattern has been shown before and all four complexes can be competed off by unlabeled specific competitor oligonucleotide, but not by random sequences (13 ). In agreement with previous data (13 ,14 ), addition of the anti-Sp1 antibody caused complex 2 to be supershifted (Fig. 1 A, lane 3). Addition of the anti-Sp3 antibody prevented formation of complex 3, but not complex 2 (lane 5). When anti-Sp1 and anti-Sp3 antibodies were added simultaneously formation of complexes 1, 2 and 3 was affected (lane 7). These results suggest that Sp1 and Sp3 bind independently to form complexes 2 and 3, respectively; but bind cooperatively and/or simultaneously to form complex 1. Anti-Sp2 and anti-Sp4 antibodies did not affect protein binding to this promoter region (lanes 4 and 6). Interestingly, formation of complex 4 was partially (~50%) inhibited by the addition of the anti-Sp3 antibody (lanes 5 and 7), suggesting that Sp3 also contributes to the formation of this DNA-protein complex. It is possible that different Sp3 isoforms contribute to formation of complexes 3 and 4. Several isoforms of Sp3 with molecular weights of 97 and 58-60 kDa were present in the nuclear extracts from human fibroblasts (Fig. 2 , lane 1). Presence of different isoforms has been reported in other cell types (7 ,11 ).
To test the functional role of Sp1 and Sp3 in the regulation of the human [alpha]2(I) collagen promoter we utilized Drosophila SL cells (Schneider cells) which express very low levels of endogenous Sp1-like activity and thus provide a convenient system to study transcriptional regulation by Sp1 proteins (19 ). Using these cells we performed cotransfection assays of the human -353 [alpha]2(I) collagen promoter fragment linked to the CAT reporter gene (14 ) with the Sp1 (pPacSp1) andthe two Sp3 (pPacSp3 and pPacUSp3) expression vectors. As shown in Figure 2 , after transfection into Drosophila cells a smaller Sp3 isoform is predominantly expressed from pPacSp3, whereas a longer isoform is predominantly expressed from pPacUSp3. Both Sp1 and long Sp3 transactivate the [alpha]2(I) collagen promoter in a dose-dependent manner with similar potencies (Fig. 3 A). Stimulation plateaued at ~200 ng of cotransfected Sp1 or Sp3 expression vectors. On the other hand, the small Sp3 isoform transactivates the collagen promoter with lesser potency than the long isoform (Fig. 3 B). At present, it is not clear whether this is due to its lower expression levels in Drosophila cells (Fig. 2 , lane 2) or to different biological properties between the two isoforms. In agreement with previous reports (7 ,21 ), the early SV40 promoter was exclusively activated by Sp1, but not by Sp3 in our system (Fig. 3 C). This establishes that the stimulatory effects of Sp3 on the collagen promoter are not the result of non-specific modulation of transcription apparatus due to high-level overexpression. We have shown that Sp1 and long Sp3 can each stimulate the collagen promoter activity by a maximum of 30-fold (Fig. 3 A). The stimulatory functions are not mutually exclusive, however. When added together in the collagen promoter activity assay, the inductive activity can be further increased to over 80-fold (Fig. 3 D), suggesting an additive effect of Sp1 and Sp3 in transactivating the human [alpha]2(I) collagen promoter.
We have previously constructed single and double substitution mutations within the four response elements in the -353 to -34 region of the [alpha]2(I) collagen promoter and characterized their effects on activity of the collagen promoter in human fibroblasts (17 ). To determine whether binding of Sp1 or Sp3 to the repressor site (Fig. 1 B, lanes 4-6) has functional consequences and to further analyze the contribution of Sp1 and Sp3 to the constitutive activity of the collagen promoter mediated by the positive cis-response elements, we cotransfected Sp1 or Sp3 expression vectors with various promoter mutants carrying single and double substitution mutations (Fig. 4 ). When compared to the -323 bp wild type promoter construct, mutating either the GC-boxes or the TCCTCC site resulted in an ~50% decrease of the promoter activity. Mutations in either repressor site (the TCCCCC motif) or the CBF-binding site had no effect on the transactivation potency of Sp1 or Sp3. Simultaneous mutations of the GC-boxes and the TCCTCC motif virtually eliminated stimulatory effects of either Sp1 or Sp3, whereas double mutations of either the GC-boxes, or the TCCTCC motif along with the repressor site decreased the stimulatory effects of Sp1 and Sp3 by 50%, similarly to single mutants in these activator sites. We have also investigated whether the additive effect of Sp1 and Sp3 can be observed with constructs mutated in either all three GC-boxes or the TCCTCC site. As shown in Figure 4 , the additive effect is also observed with both mutated constructs, suggesting that a single Sp1/Sp3 binding site (TCCTCC) is sufficent to observe this effect.
Transient transfection assays performed in Drosophila cells strongly suggested that both Sp1 and Sp3 contribute similarly to the constitutive activity of the human [alpha]2(I) collagen promoter. Similar assay is not feasible in human fibroblasts since these cells express Sp1 and Sp3. However, this constitutive Sp1 and Sp3 expression can be utilized in an in vitro transcription assay in which activity of the total nuclear extracts can be compared with activity of the extracts depleted for Sp1 and Sp3 by addition of the specific antibodies. As shown in Figure 6 , addition of either anti-Sp1 or anti-Sp3 antibodies had modest inhibitory effects (32 and 25%, respectively) on the levels of in vitro transcription (compare lanes 2 and 3 with lane 1). However, simultaneous addition of both antibodies almost completely abolished in vitro transcription (90% inhibition). Addition of preimmune serum had no effects on transcription (lane 5). These results demonstrate that both Sp1 and Sp3 can independently support [alpha]2(I) collagen transcription in human fibroblasts and that both factors are needed to achieve the maximal inductive activity.
This study demonstrates that Sp1 and Sp3 proteins interact with at least three previously characterized response elements within the -323 and -34 bp region of the human [alpha]2(I) collagen promoter. These include a positive cis-acting element between -303 and -271 which consists of three GC-boxes (14 ), a second positive cis-acting element between -128 and -123 that contains the TCCTCC motif, and a negative cis-acting element between -164 and -159 containing the TCCCCC motif (17 ). Whereas Sp1 and Sp3 appear to be dominant transcription factors interacting with the proximal region of the human [alpha]2(I) collagen promoter, there are also additional binding proteins that interact with this promoter region. One of these is the ubiquitous transcription factor CBF that binds to the -80 region of both the mouse and human [alpha]2(I) collagen promoters (17 ,18 ,23 ). At present, the nature of the additional binding proteins interacting with the human promoter is unknown. The complex 2 interacting with the TCCCCC motif may be either the c-Krox or BFCOL1, which have been characterized as binding factors interacting with the corresponding cis-regulatory element in the mouse [alpha]2(I) collagen promoter (24 ,25 ). However, further studies with the human promoter are required to confirm this possibility. Taken together, the human [alpha]2(I) collagen promoter contains multiple Sp1/Sp3 binding sites, which are probably necessary to ensure high levels of expression of this abundant protein. The role of other transcription factors interacting with the different promoter elements is probably to provide as was postulated for c-Krox tissue specificity (24 ).
Functional analyses of Sp1 and Sp3 in Drosophila SL cells indicate that either protein can transactivate the human [alpha]2(I) collagen promoter with potency equal to the other (Fig. 3 A). More importantly, combined functions of Sp1 and Sp3 can activate the collagen promoter to a level that is 2-fold higher than the maximal activation levels achievable by Sp1 or Sp3 alone (Fig. 3 C).Significantly, a single Sp1/Sp3 binding site is sufficient to observe the additive effect of Sp1 and Sp3. Formation of multimeric complexes was previously observed using purified Sp1 protein (26 ). It is possible that Sp1 and Sp3 are capable of forming hetero-multimeric protein complexes, and that the activation domain formed in the Sp1-Sp3 complex is more potent than those in either Sp1 or Sp3 homo-multimer. These results, although obtained from assays performed in Drosophila cells, appear to be valid, since comparable observations were made in the in vitro transcription assays using nuclear extracts from human fibroblasts (Fig. 6 ). While depletion of endogenous Sp1 or Sp3 by cognate antibodies can decrease the promoter activity by ~30%, simultaneous depletion of Sp1 and Sp3 can completely abolish the promoter activity, again demonstrating the cooperative nature of the Sp1 and Sp3 functions. The cooperative roles of Sp1 and Sp3 are further supported by the formation of high molecular weight complex between the collagen promoter fragment and both Sp1 and Sp3 (e.g. complex 1 in Fig. 1 A). Recent studies of the mouse [alpha]2(I) collagen promoter suggest that Sp1 may play only a minor role in the regulation of that promoter (21 ). However, at present, a full comparison of the role of Sp1 family members in regulation of the human and mouse promoters cannot be made because the role of Sp3 in regulation of the mouse collagen promoter is not known.
We have previously shown that three positive cis-acting elements of the human [alpha]2(I) collagen promoter, i.e., the GC-boxes, the TCCTCC motif and the CBF binding site, contribute equally to the constitutive activity of this promoter in human fibroblasts (17 ). Furthermore, our previous data suggested that a repressor which binds to a promoter region located between the GC-boxes and the TCCTCC motif interferes with the activation of the promoter via these two positive cis-elements, but activation by the third positive response element that binds CBF seemed to be unaffected by this repressor (17 ). The current study corroborates the results of the previous analyses. Here we provide evidence that Sp1 and Sp3 proteins bind to the GC-boxes, the TCCTCC motif and to the repressor site, but do not bind to the CBF response element. How binding of Sp1 and Sp3 to the repressor site interferes with the function of the activator sites will require further studies. The possibility that Sp1 or Sp3 plays an active role in repressing transcription when binding to the repressor site is not likely since mutating the repressor site does not cause an increase in the transcription level inducible by Sp1 or Sp3 (Fig. 4 ). One possible explanation is that the repressor site acts as a `molecular sink' to take away Sp1 or Sp3 from the activator sites. The role of other transcription factors interacting with these sites may be to facilitate binding of Sp1 and Sp3 to the particular response element. Purification and characterization of the other transcription factors interacting with the human [alpha]2(I) collagen promoter will be necessary to completely understand the regulation of the collagen gene in human fibroblasts.
In conclusion, this study demonstrates that, in contrast to many other genes which are inhibited by Sp3 (7 -9 ), the human [alpha]2(I) collagen gene is stimulated by this transcription factor. Since Sp1 and Sp3 proteins are present in all tissues tested so far (5 ) and are probably involved in regulation of the majority, if not all, of the genes the different function of Sp3 in the context of different promoters may provide one way to differentially regulate gene expression by these ubiquitous transcription factors.
We thank Dr Guntram Suske for providing pPacSp3 and pPacUSp3 expression vectors and Dr Robert Tjian for providing pPac0 and pPacSp1 expression vectors. We thank Drs Tien Hsu and Edwin Smith for critically reading the manuscript. This work was supported by National Institutes of Health grant AR42334 and by the RGK Foundation.
*To whom correspondence should be addressed. Tel: +1 803 792 8453; Fax +1 803 792 7121; Email: trojanme@musc.edu
REFERENCES