Multiple and Essential Sp1 Binding Sites in the Promoter for Transforming Growth Factor-β Type I Receptor*

Maximal gene expression driven by the promoter for the transforming growth factor β type I receptor (TGF-βRI) occurs with a 1.0-kilobase pair fragment immediately upstream of exon 1. This region lacks a typical TATA box but contains CCAAT boxes, multiple Sp1, and PEBP2/CBFα binding sites among other possiblecis-acting elements. Alterations within two CCAAT box sequences do not mitigate reporter gene expression driven by the basal promoter, and no nuclear factor binds to oligonucleotides encompassing these sites. In contrast, other deletions or site-specific mutations reveal an essential Sp1 site in the basal promoter and several dispersed upstream Sp1 sites that contribute to maximal reporter gene expression. The proportions of transcription factors Sp1 and Sp3, and their ratios of binding to consensus elements, are maintained in bone cells at different stages of differentiation. Finally, nuclear factor that binds to PEBP2/CBFα-related cis-acting elements in the basal promoter sequence also occurs in osteoblasts. Our studies reveal that constitutive expression of TGF-βRI may be determined by constitutive nuclear factor binding to Sp1 sites, whereas other elements may account for the variations in TGF-βRI levels that parallel changes in bone cell differentiation or activity.

Transforming growth factor-␤ (TGF-␤) 1 receptors occur on most cells, and a functional TGF-␤ type I receptor (TGF-␤RI) is required for all known TGF-␤-dependent effects. In some situations its activity is controlled by complex interactions with other cell surface components (1)(2)(3). However, in contrast to TGF-␤RII and the cell surface proteoglycan also termed TGF-␤RIII or betaglycan, expression of TGF-␤RI is maintained on differentiated bone cells (4). For these reasons, and because little is known about the molecular control of TGF-␤RI expression, we cloned the rat TGF-␤RI promoter and characterized several of its functional aspects in cultures of primary and continuous skeletal and nonskeletal cells derived from fetal rats. The rat TGF-␤RI promoter lacks a typical TATA box, but initiates transcription at multiple sites within a 220-bp span upstream of the initial methionine codon in differentiated bone cells. The 3Ј-terminal 300-bp sequence encompassing this region contains a GC-rich CpG island, seven consensus Sp1 binding sites, and two CCAAT boxes. Transfection studies using different fragments of TGF-␤RI promoter cloned upstream of the reporter gene luciferase demonstrated maximal activity by a 1.0-kb fragment that encompassed these and other possible cis-acting elements. Importantly, several dispersed elements appeared to cooperate for maximal reporter gene expression in osteoblast-enriched cultures (5). Coincident with this work, the human TGF-␤RI promoter was cloned, and its sequence reveals a similar organization with identically spaced CCAAT box motifs (6).
These features suggested that the TGF-␤RI gene is driven by a constitutively active promoter that maintains expression of TGF-␤RI in many cells. Nevertheless, this promoter is partly unusual to the extent that other promoters organized in a similar way tend to lack CCAAT box sequences. Imposed on this are our previous observations that the proportions of TGF-␤RI mRNA and protein may vary with the osteoblast phenotype and that its levels are rapidly controlled by certain stimulatory and inhibitory bone growth regulators (4,7). 2 Initial TGF-␤RI promoter activity studies substantiate that osteoblast-related variations in steady state mRNA levels are controlled at least in part at the level of gene transcription (4,5). 3 Therefore, the widespread expression of TGF-␤RI, driven by a constitutively active promoter, may in some instances be regulated by other cis-acting regulatory elements.
In the present study we investigated in more detail sequences within the TGF-␤RI promoter that are required for maximal and basal activity. We examined the importance of two CCAAT boxes and various consensus and putative binding sites for Sp1 transcription factor family members that occur in this region and identified the presence of PEBP2/CBF␣ binding sites. Our results identify that some of elements are not used under basal conditions, some appear to be essential components of constitutive TGF-␤RI gene expression, and yet others may help to determine phenotype-dependent TGF-␤RI expression by differentiated bone cells.

EXPERIMENTAL PROCEDURES
Cell Cultures-Using procedures approved by Yale Animal Care and Use Committee, parietal bones from 22-day-old Harlan Sprague Dawley rat fetuses (Charles River Breeding Laboratories) were dissected free of sutures and digested for five 20-min intervals with collagenase. The first digestion releases less differentiated periosteal cells, and the last three digestions are enriched with cells with differentiated osteoblast characteristics. Primary cultures were plated at 5 ϫ 10 3 cells/cm 2 in Dulbecco's modified Eagle's medium containing 20 mM HEPES (pH 7.2), 100 g/ml ascorbic acid, penicillin and streptomycin, and 10% fetal bovine serum. Cultures reach confluence (5-6 ϫ 10 4 cells/cm 2 ) within 6 -7 days. Proliferating cultures were collected at 75% confluence. Every 3-4 days, confluent cultures were re-fed the same medium except that ascorbic acid and serum were reduced by half. Differentiated cultures were collected 1 week after confluence. Mineralizing cultures were supplemented with 3 mM ␤-glycerol phosphate and collected 2 weeks after confluence. Mineralized nodules were only observed in population 3-5, were evident 3-4 days after adding ␤-glycerol phosphate, and accumulated throughout 2 weeks of incubation (8,9). Clonal rat osteosarcoma-derived osteoblast-like ROS 17/2.8 cultures (obtained from Dr. Gideon Rodan; Merck Sharp and Dohme Research Laboratories, West Point, PA) and first passage skin fibroblasts from rat fetuses used to isolate bone cells were cultured and treated by similar procedures (4).
Plasmids-Constructs pEN1.0, pEXH0.9, pSN0.8, pAN0.4, pAX0.2, pAS0.2, pXN0.1, and pSN0.1 containing fragments of the rat TGF-␤RI promoter cloned upstream of the reporter gene luciferase were described previously (5). Since its publication, we noted several sequence compressions in the CpG island when we confirmed oligonucleotide sequences for nuclear factor binding studies. A corrected sequence for accession number U48401 has been submitted to GenBank. To produce the deletion mutant pSX3, an oligonucleotide primer (pMR2) corresponding to nucleotides Ϫ240 to Ϫ221 of the rat TGF-␤RI promoter was prepared to include an XhoI site (underlined) (5Ј-CCTGGCGGAGCTC-GAGCGGCCCTGGACTTCTGC-3Ј). Twenty-five PCR cycles were performed at 94°C for 30 s (denaturation), 58°C for 1 min (annealing), and 72°C for 2 min (extension) with pMR2, reverse Sp6 sequencing primer (5Ј-GATTTAGGTGACACTATAG-3Ј), pGEM-ES1.1 (comprising 1.1 kb of TGF-␤RI promoter sequence; Ref. 5) as template, all four dNTPs, and 1 unit of Taq DNA polymerase (Boehringer Mannheim). The 0.73-kb PCR product was gel-purified (Qiagen), and the fragment released with XhoI and SacI was used to replace the corresponding wild type sequence within pSN0.8. Other mutation and deletion constructs were prepared by analogous PCR procedures. Constructs pCAAT1, with a mutation in the forward CCAAT box at nucleotides Ϫ124 to Ϫ120, and pAN0.42, incorporating the substitutions in probe XN22 as shown in Fig. 9, were prepared with pAN0.4 DNA as template. Insert for pCAAT1 was produced with forward RVprimer3 (5Ј-CTAGCAAAAT-AGGCTGTCCC-3Ј), corresponding to vector DNA upstream of polylinker, and reverse primer PMR1 [5Ј-TAACTGCTCGAGGGGCG-CACGAAGCTTCTGCCCGGCCTCC-3Ј), corresponding to nucleotides Ϫ136 to Ϫ98 of the TGF-␤RI promoter, with XhoI site (underlined) and substitutions (double underlined) as shown in Fig. 4. The 0.25-kb product was released with SacI and XhoI and inserted into the corresponding region in wild type pAN0.4. Insert for pAN0.42 was produced with forward primer (5Ј-AGCGAGGCCGCGGCGGCGGCGGGGACCTGG-GGCGAGGAGA-3Ј), corresponding to nucleotides Ϫ86 to Ϫ47 with substitutions (double underlined) at nucleotides Ϫ60 to Ϫ61, and backward primer GLprimer2 [5Ј-CTTTATGTTTTTGGCGTCTTCCA-3Ј], corresponding to vector DNA downstream of the polylinker NcoI site. The 0.14-kb product was digested with SacII and NcoI (see Fig. 4), and the 54-bp downstream fragment was inserted into the corresponding region in wild type pAN0.4. Constructs pCAAT2 (wild type) and pCAAT2 (with a 3-base pair substitution in the backward CCAAT box at nucleotides Ϫ216 to Ϫ220) were prepared with TGF-␤RI promoter DNA from nucleotides Ϫ700 to Ϫ203, obtained by digestion with SacI and ApaI, and cloned into pBluescript II KS as template. Insert for pCAAT2 was produced with a forward primer that added a downstream SacI site (underlined) at the 5Ј end, immediately upstream of promoter DNA nucleotides Ϫ238 to Ϫ217 (5Ј-CGGGAGCTCAGAAGTCCAGGGC-CGCTCATTG-3Ј), and backward primer T3 (5Ј-ATTAACCCTCACTA-AAG-3Ј), corresponding to vector DNA. Insert for pCAAT2 was produced with a similar forward primer, also including a 3-base substitution in the CCAAT motif (double underlined) plus 9 bases at the 3Ј end (5Ј-CGGGAGCTCAGAAGTCCAGGGCCGCTCGAGGGCCGC-CCAG-3Ј) and backward primer T3. The products were digested with SacI and ApaI and inserted into pSN0.8 (5) previously digested with the same restriction enzymes. Plasmid constructs were purified with the Wizard Maxiprep Kit (Promega). DNA of all constructs was verified by sequencing.
Transfections-Cultures at 50 -60% confluence were rinsed in serum-free medium and exposed to 1-1.5 g of plasmid construct per 4.5 cm 2 culture with 0.5% Lipofectin (Life Technologies, Inc.) for 3 h. Cells were re-fed medium supplemented with 5% fetal bovine serum and cultured 48 -72 h to reach confluence. Cultures were rinsed with phosphate-buffered saline and extracted with cell lysis buffer (Promega). Nuclei were cleared by centrifugation at 12,000 ϫ g for 5 min. A commercial kit was used to measure luciferase activity (Promega) in supernatants and corrected for protein by the Bradford method (10).
Nuclear Extracts-Cultures were rinsed with phosphate-buffered saline containing the phosphatase inhibitors sodium orthovanadate (1 mM) and sodium fluoride (10 mM) on ice. Cells were scraped into the buffer and collected by centrifugation. Nuclear extracts were prepared by the method of Lee et al. (11,12) with minor modifications. Briefly, cells were lysed in hypotonic buffer (10 mM HEPES (pH 7.4), 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol) supplemented with phosphatase inhibitors, protease inhibitors, phenylmethylsulfonyl fluoride (0.5 mM), pepstatin A (1 g/ml), leupeptin (2 g/ml), and aprotinin (2 g/ml), and 1% Triton X-100. Nuclei were pelleted and resuspended in hypertonic buffer containing 0.42 M NaCl, 0.2 mM Na 2 EDTA, 25% glycerol, and the phosphatase and protease inhibitors described above. Soluble proteins released by 30-min incubations on ice were collected by centrifugation at 12,000 ϫ g for 5 min, and the supernatant was aliquoted, and corrected for protein content (10), and stored at Ϫ75°C.
Electrophoretic Mobility Shift Assays-Double-stranded oligonucleotide probes were annealed by heating to 95°C and cooling to 25°C in 10 mM Tris-Cl (pH 8.0), 1 mM EDTA, 5 mM MgCl 2 over a period of 1 h. Probes were end-labeled to 1-3 ϫ 10 5 cpm/ng DNA with [␣-32 P]dCTP and Klenow fragment of Escherichia coli DNA polymerase I at 25°C for 25 min and gel-purified. Nuclear extracts (5-7 g protein) were incubated in binding buffer (25 mM HEPES (pH 7.5), 80 mM KCl, 2 mM dithiothreitol, 0.5 mg/ml bovine serum albumin, 12.5% glycerol) containing 62.5 g/ml of poly(dI/dC) (Sigma) on ice for 10 min, and then supplemented with 3 ϫ 10 4 cpm (0.1-0.2 ng) of 32 P-labeled oligonucleotide probe for 30 min in a total volume of 20 l. In competitive binding studies, unlabeled native or mutated oligonucleotides were added just before 32 P-labeled probe. Probes used in this study are shown in Table  I and Fig. 9. To assess transcription factor immunologically in gel shift analyses, 0.1-1.0 g of rabbit polyclonal anti-Sp1 or anti-Sp3 antibody or rabbit IgG (Santa Cruz) was incubated with nuclear extract in binding buffer on ice for 60 min before adding 32 P-labeled probe. Nuclear protein-DNA complexes were resolved on 5% nondenaturing polyacrylamide gels in 0.5 ϫ TBE buffer (90 mM Tris borate (pH 8.3), 2 mM TABLE I Oligonucleotides used in gel electrophoretic mobility shift assays Positions of oligonucleotides are by reference to rat TGF-␤RI promoter sequence (5). Shift refers to slower migration through polyacrylamide gel: ϩ, presence; Ϫ, absence. Factor refers to gel shift complexes identified with transcription factor specific antisera. Nucleotide substitutions in XN1 that differ from the wild type sequence of region XN1 the rat TGF-␤RI promoter are shown in boldface and underlined.
EDTA) by electrophoresis for 2.5 h at 20°C with 130 V. Gels were dried and analyzed by autoradiography.
Statistical Analysis-Data were analyzed in multiple samples after multiple determinations and where appropriate are expressed as means Ϯ S.E. In experiments comparing more than one variable or group, statistical differences were assessed by analysis of variance with limits set by Dunnet. In experiments where a single group was compared with control, analysis defaulted to Student's t test. Comparisons were performed with a commercial statistical software package (Sig-maStat®). Differences among groups were considered significant when p values were Ͻ0.05.

Multiple Sp1
Binding Sites within Active Elements of the TGF-␤RI Promoter-By sequence analysis, many transcription factor binding sites occur within a CpG island at the 3Ј-terminal 300-bp region of the rat TGF-␤RI promoter. Consistent with other similarly organized promoters, transcription initiates at several sites within a ϳ200-bp span immediately upstream of exon 1 (5). There are seven consensus Sp1 binding sites, including three GC boxes, and at least nine other potential Sp1 binding sites with 80 -90% sequence homology to consensus GC boxes within the 0.9-kb region upstream of the initiator methionine codon at position ϩ22 to ϩ25 (Ref. 5; see Table II). As shown in Fig. 1, deleting various spans that include Sp1 binding sites from the maximally active 1.0-kb region termed EN (flanked by EcoRI and NcoI restriction sites) significantly limited reporter expression. As in earlier studies, pSN0.7, pAN0.4, pXN0.1, and pSN0.1, derived from pEN1.0 by truncation from the 5Ј end, caused incremental decreases in reporter gene expression. However, even the short fragments contained in pXN0.1 and pSN0.1 maintained a low but significant level of promoter activity by comparison to the promoterless pGL3-Basic vector. Deletions from the 3Ј end (pEXH0.8, pAX0.2, and pAS0.2) also limited reporter expression. With pAX0.2 and pAS0.2, reporter gene expression was consistently below the activity driven by pGL3-Basic. These findings indicated that sequences in region XN (flanked by XhoI and NcoI restriction sites) in the TGF-␤RI promoter are essential for basal promoter function, although other sequences upstream of the XhoI site appear necessary for maximal promoter activity. Many cis-acting Sp1 binding elements occur within regions EN, AN (flanked by ApaI and NcoI restriction sites), and XN of the promoter (see Fig. 1 and Table II). Region AX (flanked by ApaI and XhoI restriction sites) contains four consensus Sp1 binding sites, including one GC box. When the Sp1 binding sites in this region were deleted internally in construct pSX3, promoter activity was severely limited. Therefore, sequences encompassing various Sp1 sites appear to be important components of the TGF-␤RI promoter, and several may be required for maximal activity.
Two CCAAT Boxes in the Rat TGF-␤RI Promoter Are Inactive in Bone Cells-Because promoter construct pAN0.4 maintains strong activity, we used oligonucleotide probes spanning each of the several clusters of cis-acting elements in this area in FIG. 1. Cooperative effect among Sp1 binding sites in the rat TGF-␤RI promoter. A, positions of various restriction endonuclease cleavage sites used to generate fragments of the rat TGF-␤ promoter. B, putative nuclear factor binding sites identified by sequence analysis. The black arrow at position ϩ1 indicates the most downstream transcription initiation site (5). C, DNA encoding the portions of the TGF-␤RI promoter indicated by gray bars was ligated upstream of the reporter gene luciferase in transfection vector pGL3-Basic and co-transfected with a reporter construct encoding ␤-galactosidase (5) in primary osteoblast-enriched cultures from fetal rat bone. Data are shown as relative luciferase activity (by comparison to pGL3-Basic vector), corrected for protein content. ␤-Galactosidase activity never varied by more than 6% (S.E.) within an experiment. Data are results from 2 to 14 separate overlapping studies with 6 -44 replicate cultures per condition. By analysis of variance, all other constructs are significantly lower in activity than pEN1.0.

TABLE II
Locations of GC boxes, Sp1 binding sites, and potential Sp1 binding sites in the rat TGF-␤RI promoter sequence Positions of oligonucleotides are by reference to rat TGF-␤RI promoter sequence (5). Nucleotides that differ from various DNA binding sequences previously defined as consensus GC boxes ((G/T)(G/A)GG(C/ A)G(G/T)(G/A)(G/A)(C/T)) or Sp1 binding sites (GGGCGG) are in boldface and underlined.
Ϫ58 to Ϫ49 80 electrophoretic mobility shift assays. As shown in Fig. 2 and Table I, probes SA1, SX1, AX2, AX3, AX5, and AX6 possess several putative Sp1 binding sites. While most growth factor receptor gene promoters so far identified lack TATA and CCAAT boxes, two CCAAT boxes occur in this GC-rich region of the TGF-␤RI promoter. One is at Ϫ216 to Ϫ220 in the backward orientation within SX1, and the other is at Ϫ124 to Ϫ120 in the forward orientation within AX3. When oligonucleotide probes were 32 P-labeled and combined with nuclear extract from primary osteoblast-enriched bone cell cultures, no DNAprotein complexes occurred with probes SX1 or AX3 (Fig. 3A). Probe AX6, encompassing but slightly larger than AX3, was also examined with nuclear extract from these cells, from dermal fibroblasts, from undifferentiated periosteal bone cells, and from the highly differentiated osteosarcoma-derived osteoblast-like cell line, ROS17/2.8. Analogous to results with 32 P-AX3, 32 P-AX6 never bound nuclear factor from osteoblast-enriched cultures (Fig. 3B) or any other cell type examined so far (data not shown). PCR primers with substitutions in the CCAAT box regions were used to create mutated reporter plasmid constructs. As shown in Fig. 4, pCAAT1 (with alterations in the forward CCAAT box) and pCAAT2 (with alterations in the backward CCAAT box) each promoted reporter gene expression equivalent to the parental constructs. Although less overall TGF-␤RI promoter activity occurs with undifferentiated bone cells (5), similar results occurred in these cells and the osteoblast-enriched bone cell cultures transfected with pCAAT1. Consequently, neither CCAAT box motif binds nuclear factor nor are they essential for basal promoter activity in bone cells. Functional Sp1 Binding Sites in the TGF-␤RI Promoter-Primary osteoblast-enriched cell cultures are derived from normal tissue and appear to be controlled in appropriate physiological ways. Because they express high levels of TGF-␤RI mRNA, protein, and promoter activity (4,5,7,9), and protein that associates with Sp1 binding sites, many subsequent studies were performed with extracts from this culture model. Probes SA1, AX2, and AX5 all formed slowly migrating radiolabeled complexes identical to those with 32 P-SP1, containing a consensus Sp1 binding site, and were effectively reduced by unlabeled consensus oligonucleotide SP1 (Fig. 3B). With oligonucleotide AX5 as a representative site to characterize Sp1 binding further, a portion of band S1 supershifted to an even more slowly migrating complex with anti-Sp1 antiserum, while the remainder of band S1 and all of band S2 were insensitive to any amount of anti-Sp1 antiserum that we examined (see below). Parallel results occurred with 32 P-labeled oligonucleotide probes SA1, AX2, AX5, and SP1 (data not shown). Since unlabeled oligonucleotide SP1 displaced bands S1 and S2 completely, inefficient binding by anti-Sp1 antibody may account in part for this difference. Alternatively, other Sp1-like nuclear proteins might also bind these probes. Two Sp1 family members, Sp2 and Sp3, bind analogous DNA elements with similar affinities (13). Sp3 is comparable in M r to Sp1 and could account for complexes in bands S1 and S2 that are resistant to anti-Sp1 antibody. Addition of 0.1 or 1 g of anti-Sp1 antibody each supershifted band S1 to a similar extent, and 0.1 or 1.0 g of anti-Sp3 antibody depleted band S2 and in part band S1. Simultaneous use of both antibodies eliminated nearly all 32 Pprobe from bands S1 and S2. The more obvious effect at band S1 with both antibodies suggests that loss of Sp3 (band S2) when only anti-Sp3 antibody is used may make more 32 P-probe available for binding by Sp1. No gel shift, supershift, or depletion occurred with normal rabbit IgG (Fig. 5). Band S1 therefore appears to contain Sp1 (by supershift) and Sp3 (by antibody depletion), whereas band S2 contains predominantly Sp3. The fractional amount of nucleotide binding to Sp1 and Sp3 was similar with each 32 P-labeled probe examined. Partitioning of Sp1 and Sp3 in these ways has been noted previously (14 -17) and may relate in part to multiple Sp3 isoforms arising from differential initiation codon utilization. 4 Binding to Sp1-like sites occurred with oligonucleotide probes XN1 (from Ϫ108 to Ϫ71) and XN2 (from Ϫ70 to Ϫ46) of region XN and nuclear extract from osteoblast-enriched cultures (Fig. 6A). 32 P-XN2 formed nuclear factor complexes similar to those with probes SA1, AX2, AX5, or SP1, whereas binding to 32 P-XN1 occurred in several complexes. As shown in Fig. 6B, after a longer exposure to film, band S1, while minimal, was detected with 32 P-XN1, was specifically reduced by excess unlabeled probe SP1 (Fig. 6B), and supershifted with anti-Sp1 antibody (data not shown). Band S2 was not detected with 32 P-XN1, consistent with the faint binding in band S1 that normally accounts for the majority of nuclear factor binding by Sp1-like sites. Also, unlabeled SXN1.2 in which the Sp1 binding site was eliminated did not reduce 32 P-XN1 in band S1 (  , and ROS 17/2.8 cultures (ROS) was combined with 32 P-AX5, 32 P-SX1, or 32 P-AX3 as indicated. B, nuclear protein (7 g) from fetal rat osteoblast-enriched cultures was combined with 32 P-labeled oligonucleotides SA1, SX1, AX5, AX2, AX6, AX3, or SP1, without or with 50-fold molar excess of unlabeled SP1 as indicated. Oligonucleotide sequences are shown in Table I and positions are shown in Fig. 2. Protein-DNA complexes were resolved on 5% nondenaturing polyacrylamide gels and visualized by autoradiography. 6B). No nuclear factor complexes formed with 32 P-XN3.
The various Sp1 sites in regions SA, AX, and XN were also compared by competitive binding studies where 25-and 50-fold excess unlabeled oligonucleotides inhibited nuclear factor binding to 32 P-SP1 to different extents. Consistent with direct binding studies (Fig. 3), oligonucleotides SA1, AX5, and XN2 competed with very high affinity, and SX1 and XN1 did not. Because the same Sp1 sites occur in SX1 and AX5, upstream sequences may restrict Sp1 binding to this region in some instances. Probe AX2, which contains both an Sp1 binding site and a GC box, inhibited Sp1 binding to 32 P-SP1 less efficiently than SA1, AX5, and XN2 (Fig. 7). Although AX2 clearly binds Sp1 (Fig. 3), the proximity of the two Sp1 sites or flanking sequences may cause less avid binding to Sp1 than other sequence configurations. Analogous to the low to negligible amounts of Sp1 binding to 32 P-XN1 and 32 P-XN3 seen in Fig. 6, oligonucleotide XN1 inhibited 32 P-SP1 binding only weakly (Fig. 7), and XN3 had no effect (data not shown).
Oligonucleotide XN2 contains four potential, overlapping Sp1 binding sites (designated as Sp1 binding domains 2 to 5 in Fig. 8) and effectively competed with 32 P-SP1. Because Sp1related complexes were not competed by and did not form with oligonucleotide SXN1.2 (Fig. 6B, and other data not shown), binding domains 2 and 3 in oligonucleotide XN2 are unlikely to be active, whereas domain 4 may require more 3Ј sequence. To assess whether domain 4 or 5 or both are functional, probes XN21, XN22, and XN23, derived from XN2 by specific nucleotide substitutions shown in Fig. 8, were examined. Although small decreases in binding occurred with XN21 and XN23, only XN22 could not form nuclear factor complex (Fig. 9A), indicating site 4 as the most active site in XN2. Consistent with this, when the nucleotide substitutions found in XN22 were introduced into the reporter construct pAN0.4 to create pAN0.42, reporter activity fell to the level of pGL3Basic (Fig. 9B), revealing that this specific Sp1 binding site is essential for basal gene expression from the TGF-␤RI promoter.
Ratios of Sp1 to Sp3 Do Not Change Significantly during Bone Cell Differentiation-TGF-␤RI is constitutively expressed in many tissues, although its level may vary with cell pheno-  Table I, and positions are shown in Fig. 2. Protein-DNA complexes were analyzed by 5% native polyacrylamide gel electrophoresis and autoradiography. S1 and S2 refer to complexes that are distinguished by anti-Sp1 and anti-Sp3 specific IgGs.
FIG. 6. Nuclear factor binding to regions downstream of ؊0.1 kb in the rat TGF-␤RI promoter. A, nuclear protein (7 g) from fetal rat osteoblast-enriched cultures was combined with oligonucleotides 32 P-oligonucleotides XN1, XN2, or XN3 from the TGF-␤RI promoter. B, nuclear protein was combined with 32 P-XN1 and the following additions: no addition (Ϫ); 100-fold molar excess unlabeled oligonucleotides XN1, XN1, SXN1.2, PEBP2/CBF, or SP1 as indicated. Oligonucleotide sequences are shown in Table I and positions are shown in Fig. 2. Protein-DNA complexes were resolved on 5% nondenaturing polyacrylamide gels and visualized by autoradiography. S1 and S2 refer to a complexes reactive with anti-Sp1-specific IgG, C refers to a complex consistent with binding to PEBP2/CBF␣ transcription factor(s), and U refers to complexes containing presently uncharacterized nuclear protein.
type and differentiation status (1,4). Using several fetal rat skin-and bone-derived cultures, we found consistent increases in the relative amounts of cell surface TGF-␤RI protein, mRNA, and promoter activity in parallel with expression of the osteoblast phenotype (4,5). These variations could not be accounted for by the amounts of Sp1 or Sp3 in nuclear extracts from these cultures. In contrast to TGF-␤RI protein, mRNA and promoter activity profiles, Sp1 and Sp3, were more abundant in fetal rat fibroblasts and osteoblast-like ROS 17/2.8 osteosarcoma-derived cells relative to primary bone-derived cell cultures, by both gel mobility shift and immunoblot analyses (Figs. 3B and 10). Furthermore, the ratio of band S1 (containing Sp1 and Sp3) to band S2 (containing Sp3) was similar in each cell type when the nuclear extracts were assessed by gel mobility shift assay without or with anti-Sp1 antibody (Fig. 10A). Nuclear extracts from proliferating, differentiating, or mineralizing osteoblast-enriched cell cultures (18) also showed analogous ratios between bands S1 and S2 and isoform distribution patterns with anti-Sp1 and anti-Sp3 antisera, although at later stages of culturing the nuclear factor profiles became more complex (Fig. 11).
Other Possible Transcription Factor Binding Sites in the TGF-␤RI Promoter-Sequence analysis also revealed other possible cis-acting elements within region AX. As described earlier, upstream oligonucleotides SA1, AX5, and AX2 bound nuclear protein from fetal rat bone cells and dermal fibroblasts essentially in complexes consistent with binding to Sp1 sites. Several possible cis-acting sequences reside downstream of these sites. Of these, a binding site for hepatocyte nuclear factor (HNF)-5 occurs downstream of nucleotide Ϫ21 and has not yet been investigated. No nuclear factor complexes formed with oligonucleotide XN3, which spans nucleotides Ϫ48 to Ϫ20 (Fig. 6A). In addition to the Sp1 binding sites in XN1 and XN2, XN1 contains a potential cis-acting element for transcription factors of the PEBP2/CBF␣ family (19 -21). As shown in Fig.  6B, 32 P-XN1 binding within complexes designated as band C was inhibited by unlabeled oligonucleotides XN1 (intact) and SXN1.2 (where the Sp1 binding site in XN1 was eliminated). Binding to band C was also reduced by a probe containing a PEBP2/CBF␣ consensus sequence (19), but not by XN1, where the PEBP2/CBF␣ binding site in XN1 contained three nucleotide substitutions. Other complexes designated as band U also formed with 32 P-XN1. They were essentially eliminated by unlabeled XN1 but varied slightly in intensity in the presence of the other site specific, truncated, or mutated XN1derived probes that we tested. Formation of these complexes may in some instances depend on the presence of other nuclear factors. In other cases they may become more evident when other binding sites found in 32 P-XN1 are eliminated and fewer nuclear factors are therefore competing for probe. We have not yet identified the proteins that elicit the U bands.

DISCUSSION
The rat TGF-␤RI promoter contains a variety of cis-acting elements that could contribute to constitutive or conditional expression. By transfecting reporter gene constructs into osteoblast-enriched cultures, we previously defined regions within the TGF-␤RI promoter that are associated with maximal and basal activity. To understand TGF-␤RI gene expression in more detail, we have now defined certain important transcription factor binding sites that may control basal promoter activity in many cells.
Similar to various growth factor receptor promoters, the TGF-␤RI promoter lacks TATA box sequence but contains a GC-enriched so-called CpG island (22,23) with many transcription factor Sp1 binding sites. Analogous to the human TGF-␤RI promoter (6), CCAAT box-like sequences that occur in this region make them unlike promoters for many other growth factors or growth factor receptors (24 -32). Nonetheless, oligonucleotides spanning the CCAAT box sites do not bind detectable levels of nuclear protein, and reporter gene expression was not reduced when these sites were disrupted. Therefore, flanking sequences or the association of other transcription elements in nearby areas may limit the contribution of the CCAAT boxes to TGF-␤RI expression under the conditions that we have examined so far.
Unlike genes controlled by CCAAT box and TATA box elements, rat TGF-␤RI mRNA transcription initiates from multiple locations, characteristic of a constitutively expressed gene controlled by an Sp1-dependent promoter (5,33). Although FIG. 7. Relative affinities for Sp1 in various regions of the rat TGF-␤RI promoter. Nuclear protein (5 g) from osteoblast-enriched cultures was combined with 32 P-SP1 oligonucleotide and no addition (none), or with 25-or 50-fold molar excess (xs) of unlabeled oligonucleotides SP1, SA1, SX1, AX5, AX2, XN1, or XN2, as indicated. Oligonucleotide sequences are shown in Table I  deletions that included various Sp1 binding sites invariably limited TGF-␤RI promoter activity, some of these elements, clustered within a 0.3-kb sequence at the 3Ј end of the promoter, appeared more essential than others. By gel shift and immunodetection assays, we determined that these regions associated to equivalent extents with either Sp1 or with the closely related transcription factor Sp3, also present in the rat cell nuclear extracts that we examined. This result is consistent with the similar structural features, conservation of DNA binding domains, and similar abilities of both Sp1 family members factors to recognize specific cis-acting elements with identical affinities (13-17, 34, 35). Unlike Sp1, Sp3 may reduce Sp1-dependent gene expression (34). We detected two com-plexes reactive with antibody specific for Sp3 by gel shift analysis consistent with earlier reports (14 -16). Furthermore, whereas TGF-␤RI and its mRNA levels vary by relation to other TGF-␤ receptors on various bone-and skin-derived cells, we did not find changes in the amounts of Sp1 or the ratio of Sp1 to Sp3 that could account for those differences in the various cell types, or in osteoblast-enriched cultures at various stages of differentiation. Consequently, in the basal state, constitutively low levels of TGF-␤RI expression may be tempered by the presence of both Sp1 and Sp3.
In other situations the proportions of Sp1 and Sp3 may change, or other cellular proteins may modify their function. For example, Sp1 forms heteromeric complexes with several cellular proteins. p107, a member of the retinoblastoma family of proteins, binds Sp1 and represses Sp1-dependent transcription, whereas retinoblastoma itself has been reported to interact with Sp1 and Sp3. Furthermore, p107 and retinoblastoma may complex with E2F, with cyclins, and cyclin-dependent kinases. Sp1 also interacts with the RelA subunit of transcription factor NF-B, and the cellular protein YY1 (36 -40). Therefore, several conditions may arise that could account for variations in Sp1 activity, and its ability to drive TGF-␤RI gene expression during the cell cycle, in various cell lineages, or in cell phenotype development.
Overall, our studies support a crucial role for several sequences throughout the TGF-␤RI promoter that contain Sp1 binding sites. Deletion of 5Ј upstream sequences, reducing the promoter region to 0.7 kb, decreases its activity by 40 -60%. Elimination of either another 0.4 kb from the 5Ј end or an internal downstream sequence further suppresses promoter function. However, elimination of 0.1 kb of sequence from the 3Ј end, a region that itself directs only moderate reporter gene expression, potently suppresses the activity of longer promoter fragments that still retain multiple Sp1 binding sites. Therefore, several regions can contribute to optimal TGF-␤RI gene expression, although sequence information within the 0.1-kb 3Ј span is essential for basal promoter activity. Using several overlapping oligonucleotides spanning this region, we located a specific sequence where substitution by two nucleotides completely eliminated Sp1 binding and suppressed reporter gene expression from a minimal promoter fragment. These results FIG. 9. Localization of an essential Sp1 binding domain in region XN2 of the rat TGF-␤RI promoter. A, nuclear protein (5 g) from osteoblast-enriched cultures was combined with 32 P-labeled oligonucleotides XN2, XN21, XN22, or XN23. Oligonucleotide sequences are shown in Table I, and positions are shown in Fig. 2 10. Relative Sp1 levels in various fetal rat-derived cell cultures. A, nuclear protein (5 g) derived from fetal rat dermal fibroblasts (RDF), less differentiated periosteal bone cells (PERIOS), osteoblast-enriched cultures (OB), and ROS 17/2.8 cultures (ROS) was combined with 32 P-AX5 without (Ϫ) or with 0.5 g anti-Sp1 IgG (ϩ). When present, IgG was included for 1 h before the addition of 32 P-AX5. Protein-DNA complexes were resolved on 5% nondenaturing polyacrylamide gels and visualized by autoradiography. Oligonucleotide sequences are shown in Table I, and positions are shown in Fig. 2. B, nuclear protein (30 g) from RDF, PERIOS, OB, and ROS cultures was resolved on 8% SDS-polyacrylamide gel, transferred to membrane, and blotted with anti-Sp1 IgG. The immunoreactive 105-and 95-kDa Sp1 bands are indicated.
FIG. 11. Distribution of Sp1 and Sp3 during osteoblast differentiation in vitro. A nuclear protein (5 g) from proliferating (P), differentiating (D), and mineralizing (M) osteoblast-enriched cultures was combined with 32 P-XN2 without (Ϫ) or with (ϩ) 0.5 g of anti-Sp1 or anti-Sp3 IgG, as indicated. When present, IgGs were included for 1 h before the addition of 32 P-XN2. Oligonucleotide sequences are shown in Table I and positions are shown in Fig. 2. Protein-DNA complexes were resolved on 5% nondenaturing polyacrylamide gels and visualized by autoradiography. Identical binding ratios were obtained with other upstream Sp1 reactive oligonucleotide probes.
confirm the importance of multiple Sp1 sites throughout the TGF-␤RI promoter and establish that one downstream site at position Ϫ63 to Ϫ54, ϳ90% homologous to consensus Sp1 binding sites, contributes heavily to basal promoter activity. This finding is analogous results with the TGF-␣ promoter, where several related but nonconsensus Sp1 binding sites are also required for optimal promoter activity (41). It is difficult to compare our results directly with those for the human TGF-␤RI promoter. Even the longest construct used to assess the human promoter region reached upstream only as far as 0.7 kb and, most importantly, did not contain the 3Ј 109-bp sequence where we detect an essential Sp1 site (6). Thus, at least two regulatory sites, including the important downstream Sp1 binding site (numbered Ϫ80 to Ϫ71 in the human promoter), have not yet been assessed for their effect on gene expression driven by the human TGF-␤RI promoter.
Within the 3Ј-terminal 0.1-kb region of the TGF-␤RI promoter, we also found a related binding site for members of the PEBP2/CBF␣ transcription factor family. PEBP2/CBF␣ family members (also termed polyoma virus enhancer binding protein 2, or PEBP2␣; and acute myelogenous leukemia factors; Refs. 19 -21) were previously identified in nuclear extracts from differentiated osteoblasts and found to have a critical role in the expression of the osteoblast-related protein, osteocalcin (18,42,43). Our studies with the TGF-␤RI have now identified a new target for PEBP2/CBF␣ activity beyond the virus-infected or immunological tissues where their identity was first established (19). The presence of a PEBP2/CBF␣-related element in this downstream region of the TGF-␤RI promoter may account in part for the high level of TGF-␤RI mRNA and protein expression and promoter activity by differentiated osteoblasts (4), imposed upon the constitutive levels regulated by Sp1 and other basal elements. We are continuing to characterize this and several other even more potent PEBP2/CBF␣ binding sites further upstream (see Fig. 1) to identify the osteoblast-enriched PEBP2/CBF␣ family members that bind to these sequences and to examine variations in PEBP2/CBF␣ expression during osteoblast differentiation. 2,3 In addition to the Sp1 and PEBP2/CBF␣-related complexes, others designated as band U also form with an oligonucleotide from this important 3Ј-terminal control region. This sequence contains elements for two other transcriptional regulators, HNF-5 and heat shock protein 70 (Hsp70). By relative migration, the slower migrating band presently seems inconsistent with a complex containing the M r of transcription factor HNF-5. It also seems unlikely to be accounted for by Hsp70 because of the basal growth conditions of our studies and the presence of another possible Hsp70 site in oligonucleotide SX1 that does not exhibit the same complex. However, it may represent a complex containing a basal transcription factor of the TFII family. TFII-related proteins are commonly involved in the expression of many genes transcribed by polymerase II, although the TGF-␤RI promoter lacks a TATA box where these agents customarily bind (5,6). Nevertheless, our studies demonstrate that basal gene expression from the TGF-␤RI promoter relies heavily on several Sp1 binding sites. One of these cis-acting elements, which occurs far downstream, appears essential for optimal TGF-␤RI promoter activity. However, the effectiveness of these sites may be modified by other negative or positive transcription regulators whose expression may vary with cell phenotype or with other extracellular circumstances.
These and other differences may account for changes in TGF-␤RI levels and therefore sensitivity to this important growth regulator during development, differentiation, or hormonal control in skeletal tissue (4,44).