The Dominant Role of Sp1 in Regulating the Cystathionine β-Synthase –1a and –1b Promoters Facilitates Potential Tissue-specific Regulation by Kruppel-like Factors*

Cystathionine β-synthase (CBS) catalyzes the condensation of serine with homocysteine to form cystathionine and occupies a crucial regulatory position between the methionine cycle and transsulfuration. The human cystathionine β-synthase gene promoters –1a and –1b are expressed in a limited number of tissues and are coordinately regulated with proliferation through a redox-sensitive mechanism. Site-directed mutagenesis, DNase I footprinting and deletion analysis of 5276 bp of 5′ proximal –1b flanking sequence revealed that this region does not confer tissue-specific expression and that 210 bp of proximal sequence is sufficient for maximal promoter activity. As little as 32 bp of the –1b proximal promoter region is capable of driving transcription in HepG2 cells, and this activity is entirely dependent upon the presence of a single overlapping Sp1/Egr1 binding site. Co-transfection studies in Drosophila SL2 cells indicated that both promoters are transactivated by Sp1 and Sp3 but only the –1b promoter is subject to a site-specific synergistic regulatory interaction between Sp1 and Sp3. Sp1-deficient fibroblasts expressing both Sp3 and NF-Y were negative for CBS activity. Transfection of these cells with a mammalian Sp1 expression construct induced high levels of CBS activity indicating that Sp1 has a critical and indispensable role in the regulation of cystathionine β-synthase. Sp1 binding to both CBS promoters is sensitive to proliferation status and is negatively regulated by Kruppel-like factors in co-transfection experiments suggesting a possible mechanism for the tissue specific regulation of cystathionine β-synthase.

is the most common cause of homocystinuria, an inherited autosomal recessive metabolic disease, which if untreated, causes skeletal abnormalities, dislocated optic lenses, mental retardation, and a dramatically increased incidence of vascular disorders particularly thromboembolic disease (1). Moderate elevation of plasma tHcy has been identified as a major risk factor for Alzheimer's disease (2,3), neural tube defects (4), and cardiovascular disease (5). Because of the key role of CBS in regulating plasma tHcy levels, there is a need for a greater understanding of the mechanisms and principles that govern its regulation.
Previously, we have determined the complete genomic sequence of human CBS (6) and mapped the transcriptional start sites of five human CBS mRNA isoforms, designated CBS Ϫ1a, Ϫ1b, Ϫ1c, Ϫ1d, and Ϫ1e, respectively (7). Of these, we found that isoforms Ϫ1a and Ϫ1b form the vast majority of transcripts, whereas isoforms Ϫ1c, Ϫ1d, and Ϫ1e are relatively rare. We identified two promoter regions upstream of exons Ϫ1a and Ϫ1b and found that the Ϫ1b promoter has ϳ10-fold greater promoter activity than that of the Ϫ1a promoter in both HepG2 and COS7 cells (6). The CBS promoter regions constitute an interesting paradox in that they have multiple sites of transcriptional initiation, are GC-rich and lack the classic TATA box sequence. Yet, despite having all of the usual characteristics of a ubiquitously expressed "housekeeping" gene, CBS is expressed in a highly tissue-specific manner with a wide range of tissues completely negative for CBS activity. The temporal and spatial expression patterns of CBS appear to be developmentally regulated (8 -10). Recently, we have shown that both human and yeast CBS are coordinately regulated with proliferation and that the human CBS Ϫ1b promoter is serum-and fibroblast growth factor-responsive and is downregulated by growth arrest due to nutrient depletion or the action of differentiation-inducing reagents (10,11). Another group has recently investigated the CBS Ϫ1b promoter and defined the minimal promoter region as a 125-bp region of sequence proximal to exon Ϫ1b. These investigators also found that the Ϫ1b promoter is activated by a synergistic interaction between NF-Y and either Sp1 or Sp3 (12). CBS expression levels in HepG2 and HT1080 cells were directly correlated with differences in the level of Sp1/Sp3 binding to the CBS Ϫ1b promoter region in these cells possibly as a consequence of phosphorylation by protein kinase A. This finding was proposed as a possible mechanism for the tissue-specific regulation of CBS (13).
In this report, we examine ϳ5 kbp of CBS 5Ј-flanking sequence for the ability to modulate CBS expression and to confer tissue-specific expression. We demonstrate that as little as 32 bp of the Ϫ1b proximal promoter region is capable of driving transcription in HepG2 cells and that a single Sp1/Egr1 binding site is essential for this activity. Both the CBS Ϫ1b and Ϫ1a promoters are regulated in a proliferation-sensitive manner by Sp1 and that, although this transcription factor is involved in a previously unreported synergistic interaction with Sp3 in specifically regulating the Ϫ1b promoter, neither Sp3 nor NF-Y are able to substitute for Sp1 in vivo. Additionally, we present evidence to show that members of the KLF family of transcription factors have the potential to play a role in the tissuespecific regulation of the CBS gene by acting to block transcriptional activation of the CBS promoters by Sp1.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Unless otherwise stated, all chemicals were obtained from Sigma (St. Louis, MO). Reagents for enhanced chemiluminescence (ECL) were obtained from Amersham Biosciences (Piscataway, NJ). Restriction and modifying enzymes were purchased from New England Biolabs (Beverly, MA). All reagents for mammalian tissue culture were obtained from Life Sciences Technologies (Rockville, MD) except for fetal calf serum, which was obtained from HyClone (Logan, UT). Synthetic oligonucleotides were obtained from Integrated DNA Technologies (Coralville, IA).
Media and Mammalian Cell Culture-HepG2 hepatocellular carcinoma cells were obtained from the American Type Culture Collection (Manassas, VA) and were cultured as described previously (10). Drosophila SL2 (Schneider) cells were a generous gift from Dr. Joan Hooper (Department of Cellular and Structural Biology, University of Colorado Health Sciences Center). SL2 cells were maintained in Schneider's insect medium supplemented in the dark at room temperature with heat-inactivated 10% fetal bovine serum and 2 mM glutamine plus 100 units of penicillin and 100 g of streptomycin/ml. Sp1 Ϫ/Ϫ fibroblasts were a generous gift from Dr. Jeremy Boss (Emory University) and were passaged as described previously (14).
Plasmids, Transfections, and Luciferase Assays-The sequence and construction of the CBS Ϫ1b promoter luciferase reporter construct pCBS47 and the CBS Ϫ1a promoter luciferase reporter construct pCBS37 has been described previously (6). A series of extended CBS Ϫ1b plasmid reporter constructs containing progressively larger regions of 5Ј-flanking sequence were generated by cleaving pCBS474 with SphI and either XbaI, SacI, BglII, BamHI, or XhoI and subsequently sub-cloning in the adjacent SphI-XbaI (pCBS478 Ϫ8236 to Ϫ3576), SphI-SacI (pCBS479 Ϫ7241 to Ϫ3576), SphI-BglII (pCBS4780 Ϫ5189 to Ϫ3576), SphI-BamHI (pCBS Ϫ4518 to Ϫ3576), or XhoI-SphI restriction fragments of 5Ј-flanking DNA that were isolated from a previously described P1 cosmid containing the CBS genomic DNA (6). The largest reporter construct CBS476 (Ϫ8943 to Ϫ3576), was generated by subcloning the EcoRI-XbaI fragment from the P1 cosmid into pZERO-2 (Invitrogen). This region was then excised from this vector as a KpnI-XbaI fragment and ligated into pCBS478, which had been cut with the same enzymes. In this manner we generated a range of larger CBS promoter luciferase constructs with up to 5276 kbp of 5Ј-flanking sequence. To investigate the minimal sequence region of the Ϫ1b promoter capable of driving transcription, we made progressive deletion derivatives of our base construct pCBS47. Two constructs, pCBS474 (Ϫ3877 to Ϫ3576) and pCB475 (Ϫ3792 to Ϫ3576) were made by end filling and subsequent blunt-end ligating after digesting pCBS47 with KpnI and SphI or ApaI, respectively. The pCBS473 construct (Ϫ3737 to Ϫ3576) was generated by subcloning of a PCR fragment amplified from a pCBS 47 template using sense (5Ј-CGTAGGTACCTGCTCTGGCAC-GAGACAT-3Ј) and antisense (5Ј-TACGAAGCTTCTGGACGGATACA-TGGAAAAGAGG-3Ј) primers. Nucleotides shown in boldface represent KpnI (sense) and HindIII (antisense) sites introduced to facilitate directional cloning of the amplification product. The sequence of this clone was verified by DNA sequencing. The pCBS472 (Ϫ3699 to Ϫ3576) and pCBS471 (Ϫ3679 to Ϫ3576) were generated by sub-cloning double-stranded synthetic oligonucleotides containing the desired sequences. This strategy was also used to generate mutant forms of pCBS472 with the CCAAT sequence mutated to CGTTG (pCBS472-1) or the Sp1 core CGCC sequence mutated to CTTC (pCBS472-2). All of the pCBS47 Ϫ1b reporter construct derivatives tested had a common 3Ј end that included the transcription start sites defined previously upstream of exon Ϫ1b (7).
The pPacNF-YA, pPacNF-YB, and pPacNF-YC plasmid constructs were kindly provided by Dr. T. F. Osborne (University of California, Irvine, CA). The mammalian expression construct D4 NF-YA-m29 expressing the dominant negative NF-YA29 subunit, capable of forming an inactive NF-Y complex by trimerization with NF-YB and NF-YC, was obtained from Dr. Roberto Mantovani (University of Milan, Italy).
The Sp1 mammalian and Drosophila expression constructs pAct-Sp1, pPacUSp3, pRc/CMV/Sp3, and pCGN-Sp1 and their relevant empty parent vectors that were used as negative controls were all provided by Dr. Paul Gardner (Brudnick Neuropsychiatric Research Institute, Worcester, MA) and were used with the kind permission of Drs. Edward Seto (University of South Florida, FL), Thomas Shenk (Princeton University, NJ), and Guntram Suske (Philipps-University, Marburg, Germany). The LKLF mammalian expression construct pBKCMVwtLKLF was a generous gift from Dr. Jerry Lingrel (University of Cincinnati College of Medicine, OH) and the KLF8 and BKLF expression constructs pMT2.BKLF, pMT2.KLF8, and pPac.BKLF were kindly provided by Dr. Merlin Crossley (University of Sydney, New South Wales, Australia). Transient transfections and subsequent luciferase reporter assays in actively proliferating cells were performed as described previously (10).
Analysis of CBS Expression-CBS enzyme activity was determined by a previously described radioisotope assay using [ 14 C]serine as the labeled substrate (15). One unit of activity is defined as the amount of CBS that catalyzes the formation of 1 mol of cystathionine in 1 h at 37°C. SDS Western blots were performed after running 10-to 75-g samples of total protein on 9% polyacrylamide gels. Transfer, primary, and secondary antibody incubation conditions and subsequent detection using ECL chemiluminescent reagents were performed as described previously (16).
Nuclear Extracts and EMSA-Nuclear extracts were prepared from actively proliferating HepG2 cells (60% confluent) as described previously (17). Electrophoretic mobility shift assays (EMSAs) were performed using a lightshift chemiluminescent EMSA kit according to the manufacturer's standard protocol. The minimally active Ϫ1b promoter region was generated for EMSA by synthesizing oligonucleotides representing the proximal 32-bp probe of sequence.
The oligonucleotide 5Ј-GCCAATCACCGCCCGCCTGCTCCCCTCG-CCGT-3Ј and its reverse complement, 5Ј-ACGGCGAGGGGAGCAGGC-GGGCGGTGATTGGC-3Ј, were labeled using a biotin 3Ј-end labeling kit according to the manufacturer's instructions (Pierce, Rockford, IL). Standard EMSA reactions contained 20 g of cell extract and 60 fmol of biotin end-labeled DNA in a 20-l volume binding reaction in the presence of 2.5% glycerol, 5 mM MgCl 2 , 50 ng/l of poly(dI⅐dC), and 0.05% Nonidet P-40. EMSA reactions were incubated at room temperature for 30 min, terminated by adding 2 l of 10ϫ loading buffer (0.2% (w/v) bromphenol blue and 0.2% xylene cyanol containing 10% (v/v) glycerol), and then separated on 6% polyacrylamide gels pre-run at 100 V for 30 min in 0.5ϫ TAE buffer (0.02 M Tris-acetate, 0.001 M EDTA) and subsequently run at 150 V for 4 h after sample loading. Reaction products were then transferred to a Biodyne B membrane (Pierce) by capillary transfer and fixed by UV cross-linking. The biotin-labeled reaction products were then visualized by incubation with streptavidin horseradish peroxidase conjugate and subsequent incubation with ECL chemiluminescent reagents.
DNase I Footprinting-The SphI-HindIII DNA fragment bearing 210 bp of Ϫ1b proximal sequence was excised from pCBS474 as an SphI-HindIII fragment into the high copy number vector pUC19. This region was then excised as a BamHI-HindIII fragment, gel-purified, 5Ј-endlabeled with 32 P on either the top (at the BamH1 site) or bottom strand (at the HindIII site), and used as a probe in DNase I footprinting as described previously (18). Approximately 4 ϫ 10 4 cpm of labeled probe was incubated with either 50 or 75 g of crude nuclear extract derived from actively proliferating (ϳ60% confluence) HepG2 cells for 10 min at room temperature in DNase I digestion buffer (100 mM NaCl, 50 mM Tris-HCl (pH 8.0), 3 mM MgCl 2 , 0.15 mM spermine, and 0.5 mM spermidine). The reaction mixture was then subjected to digestion with DNase I (20 g/ml) for 90 s at room temperature. Reaction products were run on a 6% polyacrylamide/8 M urea sequencing gel with control probe that had been incubated with DNase I in the absence of nuclear extract and a guanine sequencing ladder derived from the fragment for sequence positioning purposes. After electrophoresis, the gel was dried and visualized by autoradiography.

Defining the CBS Ϫ1b
Basal Promoter-In our first study of the CBS Ϫ1b promoter we examined the 388-bp region between the end of exon Ϫ1a and the start of exon Ϫ1b for promoter activity (6). Subsequent reports regarding the CBS basal promoter activity have been essentially confined to this 481-bp region (12,13,19). However, it is not unusual for sequence elements several kbp upstream of the transcriptional start sites to contain sequence elements that significantly modulate basal activity or confer tissue-specific expression. To fully define the CBS Ϫ1b basal promoter and to search for both enhancer and repressor sequence elements, we examined a series of CBS Ϫ1b promoter reporter constructs with up to 5276 kbp of 5Ј-flanking sequence in promoter assays after transient transfection into HepG2 cells. We used our previously reported 479-bp CBS Ϫ1b luciferase reporter construct pCBS47 (Ϫ4055 to Ϫ3576) as a reference for comparative purposes. The results of these experiments ( Fig. 1) indicated that the presence of an additional 4787 kbp of flanking sequence did not lead to any significant change in promoter activity relative to the original pCBS47 CBS Ϫ1b base construct. Deletion of an additional 259 bp from the pCBS47 construct did not adversely affect promoter activity. Further deletion of the CBS Ϫ1b promoter region was accompanied by significant reductions in the resultant promoter activity allowing us to define the 210 bp proximal to exon Ϫ1b as the basal promoter region. The observation that deletion of the region between the SphI and ApaI sites is accompanied by an approximate 50% reduction in promoter activity indicates that the 84-bp region between Ϫ3877 and Ϫ3792 contains sequence elements that contribute significantly to basal expression of the CBS Ϫ1b promoter.
As Little As 32 bp of Proximal CBS Ϫ1b Promoter Sequence Is Capable of Driving Transcription in HepG2 Cells-To define the minimal portion of the CBS Ϫ1b promoter sequence capable of driving transcription, we generated a series of progressive 5Ј deletion derivatives of pCBS47 and assayed them for promoter activity after transient transfection into HepG2 cells (Fig. 1). We were able to delete down from 82 to 32 bp (Ϫ3699 to Ϫ3576) before losing all detectable promoter activity. This definition of the minimal promoter differs significantly (4-fold smaller) from the previously published version, and, consequently, we examined the function of this region in detail. The pGL-enhancer vector used to make our pCBS472 expression construct contains an SV40 enhancer sequence located downstream of lucϩ and the poly (A) signal. To assess the possible contribution of this downstream enhancer element to the sensitivity of our promoter assays, the Ϫ1b minimal promoter region was subcloned into pGL-basic, which lacks this SV40 enhancer sequence but is otherwise identical. When this 32-bp minimal promoter region was examined in the absence of the enhancer sequence in HepG2 cells it was found to have essentially identical promoter activity as was observed previously.
Analysis of the minimal promoter sequence ( Fig. 2A) shows one candidate Sp1 binding site that overlaps with a binding site for the zinc-finger transcription factor early growth-response protein (Egr1), one CCAAT box, and putative binding sites for Ap2, MZF1, IK2, and Gfi1. Ap2 is unlikely to contribute to the promoter activity of this construct, because it is not expressed in HepG2 cells (20) and has previously been shown to downregulate CBS when heterologously expressed in these cells (10). The tissue distribution of Gfi1, IK2, and MZF1 (21) (22-24) effectively precludes any role for these transcription factors in the regulation of CBS in tissues such as liver, kidney, pancreas, and brain, which are the main sites of transsulfuration in humans (8). Consequently, the contribution of these transcription factors to CBS Ϫ1b promoter activity was not investigated. To assess the relative importance of the CCAAT box and the Sp1 site, we generated derivatives of the pCBS472 construct where these binding sites had been mutated. The effect of these mutations was examined by measuring their relative promoter activities in HepG2 cells (Fig. 2B). Clone names and nucleotide positions relative to the CBS ATG contained within each respective reporter construct are given on the y-axis. Numbers in parentheses define the amount of base pairs of Ϫ1b proximal sequence. Transient transfections into HepG2 cells and subsequent luciferase-based promoter reporter assays were performed as described previously (10). All of the promoter constructs used are derivatives of our previously described CBS Ϫ1b promoter construct pCBS47 (6) and contain a common 3Ј-end. The negative control pGL3 designates the results for the vector without insert. Data were calculated as normalized luciferase activity by comparison to a co-transfected Renilla luciferase control construct pRL-CMV and are shown as percentage of the promoter activity of the base construct pCBS47 (388 bp of proximal sequence), which is arbitrarily set at 100% for comparative purposes. Data shown represent the mean and S.E. of at least three independent experiments performed in triplicate.

FIG. 2. Deletion analysis reveals the ؊1b minimal promoter region.
A, the sequence of the CBS Ϫ1b minimal promoter region defined by deletion analysis is shown in italics. Core transcription factor binding sites shown in boldface above and below the Ϫ1b sequence were detected using the MatInspector version 2.2 program with the TRANSFAC 5.0 matrices with a core similarity value of 1.0 and a matrix similarity value of 0.9 (25). B, site-directed mutagenesis of the Ϫ1b minimal active region reveals a key Sp1/Egr site that is essential for promoter activity. To investigate the relative contribution of the Sp1 and CCAAT sites to promoter activity we performed transient transfection of the Ϫ1b minimal promoter reporter construct pCBS472 and its mutated derivatives pCBS472-1 (CCAAT sequence mutated to CGTTG) and pCBS472-2 (Sp1 core CGCC sequence mutated to CTTC) into HepG2 cells. B shows the level of promoter activity observed for each construct expressed as a percentage of the wild type pCBS472 construct for comparative purposes. Values shown represent the mean and S.E. of at least two independent experiments performed in triplicate. Transient transfections and luciferase assays were performed as described previously (10).
Mutagenesis of the CAAT box (pCBS472-1) did not affect the level of basal promoter activity in HepG2 cells indicating that this sequence does not play a crucial role in directing transcription from the Ϫ1b promoter in these cells. Mutagenesis of the Sp1/Egr1 site in the pCBS472-2 construct lead to the complete abolition of promoter activity indicating that the promoter activity of the CBS Ϫ1b minimal promoter sequence is dependent upon the Sp1-Egr1 binding site.
DNA Footprinting Reveals Key Areas of the CBS Ϫ1b Minimal Promoter Region-DNase I footprinting analysis was performed using double-stranded fragments corresponding to nucleotides Ϫ3877 to Ϫ3786. This SphI-HindIII restriction fragment from pCBS474 contains 210 bp of Ϫ1b proximal sequence and was used as a probe in DNase I footprinting using a HepG2 cell extract generated as described under "Experimental Procedures" (Fig. 3, A and B). Scoring and interpretation of protected regions was performed blind by two independent observers with no prior knowledge of the CBS promoter or the results of our deletion and mutagenesis study described above. Three distinct regions designated Fp1 (Ϫ3808 to Ϫ3787 on the top strand and Ϫ3810 to Ϫ3783 on the bottom strand), Fp2 (Ϫ3715 to Ϫ3681 on the bottom strand), and Fp3 (Ϫ3664 to Ϫ3639 on the bottom strand) were found to be either partially or totally protected from DNase I digestion. Examination of the sequence contained within the protected Fp1 region revealed two adjacent Sp1 binding sites separated by a MZF1 site. Interestingly, these binding sites are present at the 3Ј end of the 85-bp region that was shown in the previous section to contribute significantly to basal transcription indicating that these Sp1 sites are of considerable importance for maximal promoter activity. The vast majority of the Fp1 region was also found to be protected in a DNase I footprinting analysis of the Ϫ1b promoter region performed previously where it was described as region D (13).
The second protected region (Fp2) was not previously reported to be protected (13) and contains the Sp1/Egr1 overlapping binding site that was shown to be essential for the minimal promoter function in the previous section and serves as an independent confirmation of the functional importance of this binding site. This protection is relatively weak, but it is part of the nature of the DNase I footprinting methods that the relative binding affinities of different sites afford different levels of protection from the DNase I treatment. The lower degree of protection of this site is presumably due to a relatively reduced level of Sp1 binding that is entirely consistent with the level of promoter activity that this site was shown to confer in the deletion analysis described above. Interestingly, the inverted CCAAT box-NF-Y binding site 40 bp upstream of the Ϫ1b minimal promoter sequence that has previously been identified as playing a key role in regulating the CBS Ϫ1b promoter (12, 13) was not protected. We have independently found that NF-Y can weakly transactivate the CBS Ϫ1b promoter (but not Ϫ1a) in SL2 cells. 2 However, this site is close to a DNase I-hypersensitive site, which is typically indicative of conformational changes induced by protein binding and often occurs at sites where only a small fraction of the probe molecules are bound by protein. This finding suggests that NF-Y may bind to this sequence but is not present at a sufficiently high concentration in our cell extract to confer complete protection.
A third protected fragment (Fp3) was located within the Ϫ1b exon itself. This region was not investigated in previous DNase I footprinting experiments (13). Examination of this sequence using the MatInspector version 2.2 program with the TRANS-FAC 5.0 matrices (25) revealed binding sites for NF1 and MZF1. The ability of NF1 to transactivate the CBS Ϫ1b promoter was previously investigated, but the results gained were inconclusive (12) and clearly merit further investigation.  3. A, DNase I protection footprinting analysis of the Ϫ1b basal promoter region. A DNA fragment of the CBS Ϫ1b promoter encompassing nucleotides Ϫ3877 to Ϫ3576 was 5Ј-end-labeled with 32 P on either the top (at the BamHI site in the vector polylinker) or bottom strand (at the HindIII site) and used as a probe in DNase I footprinting as described previously (18). Approximately 4 ϫ 10 4 cpm of labeled probe was incubated with either 50 or 75 g of crude nuclear extract derived from actively proliferating (ϳ60% confluence) HepG2 cells for 10 min at room temperature. Reactions mixtures were subjected to digestion with DNase I (20 g/ml) for 90 s at room temperature. The reaction products were then run on a 6% polyacrylamide sequencing gel with control probe (lanes marked C) that had been incubated with DNase I in the absence of nuclear extract and a guanine sequencing ladder (lanes marked G) derived from the fragment for sequence positioning purposes. The asterisks denote the positions of DNase I-hypersensitive sites within the CBS Ϫ1b promoter region. B, three distinct DNA regions designated Fp1, Fp2, and Fp3, were found to be either partially or completely protected from DNase I digestion. The protected sites identified on both the top and bottom strands of the labeled probes used are given as shaded regions along the DNA sequence of the Ϫ1b promoter sequence. The lowercase letter "a" refers to the top strand, whereas "b" refers to those protections identified on the bottom strand.
Both the CBS Ϫ1a and Ϫ1b Promoters Are Transactivated by Sp1 and Sp3 in Drosophila SL2 Cells-Both the human CBS promoters are GC-rich and contain a large number of candidate Sp1/Sp3 sites. To investigate the role of these transcription factors in regulating CBS expression, we transfected pCBS47 (Ϫ1b) or pCBS37 (Ϫ1a) both alone and in combination with an expression construct for either Sp1 (pActSp1 (26)) or Sp3 (pPacUSp3 (27)) into Drosophila SL2 cells (Fig. 4A) that are deficient in many mammalian transcription factors, including the Sp1 family (28,29). To confirm that Sp1 and Sp3 expression is involved in transactivation of the target DNA, we also transfected cells with the empty pAct and pPacU vectors without either the Sp1 or Sp3 coding sequences, respectively. When either pCBS47 (Ϫ1b) or pCBS37 (Ϫ1a) were transfected alone or with these empty vector controls, no promoter activity was observed. When either of the Ϫ1a and Ϫ1b promoter reporters were co-transfected with either Sp1 or Sp3 expression constructs, we observed promoter activity ϳ5and 12-fold over background, respectively, indicating that these transcription factors are capable of transactivating both of the human CBS major promoters.
SP1 and SP3 Interact Synergistically to Up-regulate the CBS Ϫ1b Promoter but Not the Ϫ1a Promoter-Simultaneous cotransfection of SL2 cells with pCBS47, Sp1, and Sp3 dramatically increased the observed promoter activity of the Ϫ1b promoter to between 70-and 100-fold over background (Fig. 4A). Interestingly, the scale of this synergistic up-regulation was found to be influenced by growth status and was significantly reduced when the luciferase assays were performed in late confluence (data not shown). Synergistic Sp1 and Sp3 activation was not observed for the CBS Ϫ1a promoter, indicating that the observed interaction between Sp1 and Sp3 is not common to all functional Sp1 binding sequences in the CBS promoter regions.
Interestingly, although Sp1 and Sp3 are capable of transactivating all of these CBS Ϫ1b constructs, the synergistic interaction of Sp1 and Sp3 was only observed for pCBS47 and pCBS474. This finding indicates that not all of the functional Sp1 sites in the Ϫ1b promoter are equivalent and that Sp1 and Sp3 interact synergistically at a site within the 84-bp region that was shown by our deletion analysis to contribute significantly to basal promoter activity. Our DNA footprinting analysis showed only two protected overlapping Sp1 sites in this region, and it is likely that one or both of these sequences are the site of the Sp1/Sp3 synergistic interaction. Similar cotransfection experiments in SL2 cells by another group were limited to only one reporter construct (Ϫ3792 to Ϫ3565) that lacked this critical region and thus the synergistic interaction between Sp1 and Sp3 described here was missed (12). Further experiments are currently in progress in our laboratory to investigate if these Sp1 sites function as a focal operator site for modulating CBS expression levels.
Endogenous Sp1/Sp3 Levels in HepG2 Cells Preclude Further Activation by Exogenously Added Sp1/Sp3-To further investigate the regulation of the CBS Ϫ1b promoter by Sp1 and Sp3, we co-transfected proliferating HepG2 cells with the CBS Ϫ1b reporter construct pCBS47 alone and in combination with mammalian expression constructs for either Sp1 (pCGN-Sp1) or Sp3 (pRc/CMV/Sp3). Additionally, to assess the contribution of NF-Y to basal CBS expression in a mammalian host, we co-transfected HepG2 cells with pCBS47 and a mammalian expression construct for the NF-Y dominant negative mutant NF-YA29. The results of these experiments (Fig. 4B) showed very modest activation of the Ϫ1b promoter as a consequence of overexpression of Sp1 or Sp3 indicating that endogenous levels of Sp1 and Sp3 are sufficient to fully activate the CBS Ϫ1b promoter under the conditions tested. Co-expression of the NF-YA29 dominant negative mutant form of NF-Y lead to an approximate 30% decrease in basal activity further indicating that NF-Y plays a role in the regulation of CBS. However, the scale of the decrease suggests that NF-Y is not indispensable for CBS expression.
The Role of Sp1 Appears to Be Dominant in Vivo-In an attempt to further investigate the relative roles of Sp1, Sp3, and NF-Y in regulating the expression of CBS, we assayed a Ϫ1b basal promoter construct pCBS47 (Ϫ4055 to Ϫ3576), the Ϫ1a basal promoter construct pCBS37 (Ϫ4467 to Ϫ4093), or the Ϫ1b minimal promoter construct pCBS472 (Ϫ3699 to Ϫ3576) were transfected into Drosophila SL2 cells either alone or in combination with the pAct-Sp1 and/or pPacUSp3 expression constructs. Transient transfections and luciferase assays were performed as described previously (10). Luciferase values were normalized to total protein concentrations. The -fold induction was calculated relative to the normalized luciferase activity obtained by transfecting the appropriate empty vector control. B, the effect of exogenously added Sp1, Sp3, and the dominant negative mutant NF-YA29 upon CBS Ϫ1b promoter activity in HepG2 cells. The luciferase promoter reporter construct pCBS47 (Ϫ1b) was transiently transfected into HepG2 cells either alone or in combination with mammalian expression constructs for Sp1, Sp3, or the NF-Y dominant mutant NF-YA29. Values shown represent the mean and S.E. of at least two independent experiments performed in triplicate. Data were calculated as normalized luciferase activities by comparison to a co-transfected Renilla luciferase control construct (pRL-CMV) and are given as a percentage of the promoter activity of pCBS47. Repetition of this experiment with the CBS Ϫ1a promoter construct pCBS37 with exogenously added Sp1 and Sp3 showed essentially identical results (data not shown).
previously described Sp1-deficient fibroblast line (14) for CBS activity. We could detect no CBS activity or protein in this cell line indicating that Sp1 has a critical and indispensable role in the regulation of CBS. Transfection of this cell line with the mammalian Sp1 expression construct pCGN-Sp1 caused the induction of high levels of CBS activity (60 milliunits/mg of protein) (Fig. 5). Subsequent Western blotting of cell extracts of the Sp1-deficient fibroblasts indicated that the ubiquitous transcription factors Sp3 and NF-Y are present in this cell line (data not shown) but are clearly unable to drive CBS expression in the absence of Sp1. These findings indicate that the synergistic co-regulation of CBS by Sp1, Sp3, and possibly NF-Y may involve an interaction between these transcription factors prior to binding to the CBS promoter and that, without this interaction, neither Sp3 nor NF-Y are capable of activating the CBS promoter in vivo.
Sp1 Is Necessary and Sufficient for Growth-specific Regulation of the CBS Ϫ1b Promoter-Sp1 binding has previously been shown to be sensitive to proliferation status and intracellular redox (30 -32). We have previously shown that growth arrest and oxidative stress (in the form of exogenously added hydrogen peroxide) can negatively influence transcription of the CBS Ϫ1b basal promoter in HepG2 cells (10). To investigate the mechanisms that underlie these responses, we investigated the influence of growth status upon the ability of Sp1 to bind to the 32-bp Ϫ1b minimal promoter region using EMSAs. Because the Sp1 binding sites in the Ϫ1b promoter has the potential to bind up to 40 members of the Sp1/KLF family, we simplified the interpretation of our experiments by using cell extracts derived from Drosophila SL2 cells that had been transfected with the Sp1 expression construct pAct-Sp1. This approach allowed us to be absolutely certain that we were analyzing the binding of Sp1 specifically and not another related transcription factor or complex thereof. The ability of Sp1 to bind the CBS Ϫ1b promoter was assessed using Sp1-expressing SL2 cell extracts derived from actively proliferating SL2 cells expressing Sp1 and identical cells that had growth-arrested at confluence due to nutrient depletion. It can be seen that the ability of Sp1 to bind the CBS Ϫ1b minimal promoter, although not abolished, is significantly reduced concomitant with growth arrest at confluence (Fig. 6). Western blotting of these cell extracts indicated that they contained essentially identical levels of Sp1 protein (data not shown) thus reinforcing the point that it is not necessarily the intracellular concentration of Sp1 per se that dictates CBS expression levels but rather it is the degree of Sp1 binding that is important. Taken together, these results indicate that the dominant role of Sp1 in regulating the CBS promoters is instrumental in the growth-specific and differentiation status-sensitive regulation of CBS expression.
The CBS 5Ј-Flanking Sequence Does Not Confer Tissue-specific Expression-Although clearly regulated by the ubiqui-tously expressed transcription factors Sp1, Sp3, and NF-Y, CBS is expressed in a restricted number of tissues with many cell types completely negative for transsulfuration. A number of GC-rich TATA-less promoters regulated by Sp1 and Sp3 have previously been shown to contain sequence elements that confer tissue-specific expression (33,34). Additionally, the CBS 5Ј-flanking sequence contains numerous binding sites for delta EF1 and HNF-3 beta, which have previously been shown to regulate gene expression in a tissue-specific manner (35,36). Although our investigation of the Ϫ1b basal promoter indicated that the proximal 5 kbp of 5Ј-flanking sequence neither significantly enhances nor represses CBS Ϫ1b promoter activity in HepG2 cells, it is possible that this region contains sequence elements that act to repress promoter activity in CBS-deficient tissues. To investigate this possibility, we examined the relative expression levels of our series of CBS Ϫ1b reporter constructs in a number of CBS-negative cell lines. Previous work has shown that CBS is not expressed in the lung at any stage of development or in adult tissues (8,9). We used the previously described E10 cell line as a model system for a CBS-deficient tissue. This cell line was originally established from normal lung explants and has alveolar type II cell features such as lamellar bodies and surfactant apoprotein immunoreactivity at early passage (37). To assess the suitability of this cell line, we assayed E10 cells for CBS activity. We could detect absolutely no CBS activity in this cell line, and Western blots could find no evidence of CBS protein (data not shown). We individually transfected batches of E10 cells with our full range of CBS Ϫ1b reporter constructs to see if any portion of the Ϫ1b 5Ј-flanking sequence confers tissue-specific expression. All of the CBS promoter constructs were active in E10 cells and showed a level of promoter activity essentially identical to that observed in HepG2 cells (Fig. 1) indicating that the 5276 kbp of flanking sequence does not contain any repressor sequence elements that confer tissue specificity upon CBS expression. These findings were confirmed in two other previously identified CBSnegative cell lines, a23 (38) and Buffalo rat liver cells (39). Taken together, these findings indicate that the complete absence of CBS in numerous tissues is not likely to be due to FIG. 5. CBS enzyme activity is absent in Sp1-deficient fibroblasts but can be induced by exogenously added Sp1. Western blotting of 100-g samples of Sp1-deficient fibroblast cell lysates was performed as described previously (16). Lanes: 1, CBS-negative a23 cells (negative control); 2, 10 ng of purified CBS 63-kDa protein (positive control); 3, Sp1-deficient fibroblasts transfected with pCGN empty vector (CBS activity, 0 milliunits/mg of protein); 4, Sp1-deficient fibroblasts after transient transfection with the pCGN-Sp1 mammalian expression construct (CBS activity, 60 milliunits/mg).
FIG. 6. The influence of growth status upon Sp1 binding to the CBS ؊1b promoter. Electrophoretic mobility shift assays (EMSAs) were performed to assess the influence of growth status upon Sp1 binding to the CBS Ϫ1b minimal promoter. EMSAs were performed using a biotin-labeled double-stranded DNA fragment spanning the 32 bp of sequence proximal to exon Ϫ1b as described under "Experimental Procedures." All experiments were performed using cell extracts derived from Drosophila SL2 cells either alone (lanes labeled C) or transfected with the Sp1 expression construct pACT-Sp1. Experiments used 20-g samples of nuclear extracts isolated as described previously (17) from both proliferating (60% confluence) and quiescent Drosophila SL2 cells expressing the Sp1 transcription factor. The figure shown is representative of three independent experiments. NP, no probe negative control; C, SL2 cells without Sp1; ECP, excess cold probe control. This binding specificity control shows the effect of competition with 100-fold excess of unlabeled probe; P, nuclear extract derived from proliferating SL2 cells expressing Sp1; Q, nuclear extract derived from quiescent SL2 cells expressing Sp1. differences in the relative levels or phosphorylation status of the ubiquitous transcription factors Sp1 and Sp3 as was suggested previously (12,13).
The Human CBS Promoters Can Be Negatively Regulated by Kruppel-like Factors-A number of possible mechanisms that could allow tissue-specific regulation of an Sp1-driven promoter have been investigated in our laboratory. During the course of this investigation, we observed that co-transfection of mammalian expression constructs for several KLF members of the Sp1 family of transcription factors act to repress activity of the CBS Ϫ1b and Ϫ1a promoters in the CBS positive hepatoma cell line HepG2 (Fig. 7A). This regulatory interaction was shown to be reproducible in another hepatoma cell line FAO (data not shown). In contrast to Sp1 and Sp3, these KLF transcription factors are not ubiquitously expressed and have a relatively limited tissue distribution. The observation that lung Kruppel-like factor (LKLF), basic Kruppel-like factor (BKLF), and KLF8 are capable of almost completely abolishing expression of CBS in both HepG2 and FAO cells suggest a novel mechanism by which the Sp1-dependent expression of both CBS promoters can be regulated in a tissue-specific fashion.
BKLF Acts to Repress the CBS Ϫ1b Promoter by Inhibiting Sp1 Activation-Some negatively acting KLF transcription factors have been reported to function by binding to GC box sequence elements with high affinity and thus act to competitively displace activating transcription factors such as Sp1 from their cognate binding sites (40). Alternatively it has been reported that KLF factors can actively bind Sp1 itself and thus prevent it from binding a given promoter (41). In mammalian cells, as many as 40 members of the Sp1/KLF family and splice variants thereof may be expressed at any given time depending upon the growth and differentiation status of the cell (42). This complexity of possible interactions can obscure efforts to determine specific regulatory mechanisms for a given promoter. To investigate the mechanism by which the KLF factors may be repressing expression of the CBS promoters, we used co-transfection studies in SL2 cells, which provide an Sp1/KLF null background (Fig. 7B). When the CBS Ϫ1b base expression construct pCBS47 (1 g) was transfected into SL2 cells alone, no promoter activity was detected. Co-transfection of this construct with either pACT-Sp (1 g) or pPACUSp3 (1 g) leads to transactivation of the Ϫ1b promoter as described above. However, when pCBS47 and either of the Sp1/3 expression constructs was accompanied by the BKLF Drosophila expression construct pPac. BKLF, the transactivation of the CBS Ϫ1b promoter by either Sp1 or Sp3 was almost completely abolished. Essentially identical results were observed when this experiment was repeated using the CBS Ϫ1a promoter reporter construct pCBS37 (data not shown). The implication of these results is that the BKLF-based inhibition of the CBS Ϫ1b and Ϫ1a promoters in HepG2 cells proceeds by inhibiting activation by Sp1. Taken together; these results indicate that KLF inhibition of CBS expression functions by blocking activation of the CBS promoters by Sp1. The expression of many of the various KLF family members is highly tissue-specific and constitutes a possible mechanism by which the CBS promoter can be regulated by the ubiquitous transcription factor Sp1 in a tissuespecific manner.
Heterologous Expression of LKLF Abolishes CBS Activity in HepG2 Cells-To investigate the possible negative regulation of CBS by KLF family members in a more natural chromosomal setting, we examined CBS activity levels in HepG2 cells in the presence and absence of heterologously expressed LKLF. HepG2 cells (60% confluent) were transfected with the mammalian LKLF expression construct pBKCMVwtLKLF. Cells were allowed to grow for a further 24 h post-transfection and were then harvested for CBS enzyme assay as described previously (15). As a control, this experiment was repeated using HepG2 cells (60% confluent), which were individually transfected with empty vector. These control cells showed normal levels of CBS activity (15 milliunits/mg of total protein, S.E. ϭ 0.473, n ϭ 3), whereas expression of LKLF effectively abolished CBS activity (0.066 milliunit/mg of total protein, S.E. ϭ 0.033, n ϭ 3). These results further illustrate the potential of negative regulation by multiple members of the KLF family of transcription factors to contribute to the tissue-specific expression of the CBS gene in mammals.

DISCUSSION
Binding of the Sp1 transcription factor has been shown previously to be serum-inducible (43) and to mediate proliferationdependent and redox-sensitive regulation of numerous promoters (44 -47, 48). Additionally, the binding activity of Sp1 has been shown to be down-regulated upon growth arrest and terminal differentiation of the liver (30,49). Consequently, the dominant role of Sp1 in regulating CBS expression is in perfect agreement with all of our previously published observations regarding the coordinate regulation of this gene with prolifer- Ϫ1a and Ϫ1b promoter reporter constructs pCBS37 and pCBS47 were transiently transfected into HepG2 as described under "Experimental Procedures" either alone or in combination with mammalian expression constructs for LKLF, KLF8, or BKLF. Luciferase activities were normalized to a co-transfected Renilla luciferase control construct pRL-CMV. Values shown are the mean and S.E. of two independent experiments performed in triplicate and are given as a percentage of the promoter activities of pCBS37 or pCBS47 alone for comparative purposes. B, BKLF represses the CBS promoters by inhibiting activation by Sp1. The pCBS47 Ϫ1b promoter construct was transfected into Drosophila SL2 cells either alone or in combination with the pAct-Sp1 and/or pPacUSp3 or pPac.BKLF expression constructs. Values shown represent the average and S.E. of -fold induction compared with that of pCBS47 alone and were derived from two independent experiments performed in triplicate. ation, differentiation, and redox stress (10).
Our findings with regard to the CBS Ϫ1b minimal promoter region and the Fp2 region that is protected therein in our DNase I footprinting analysis differ significantly from previously published investigations (12,13). Previously, another group defined the minimal promoter region as the proximal 125-bp region, and they did not observe protection of the Fp2 region from DNase I digestion in HepG2 cell extracts. Interestingly, the 125-bp region defined by this group to be the minimal promoter, which we both agree is perfectly capable of driving transcription, does not contain any protected regions in their footprinting analysis. It is hard to imagine that this minimal promoter functions without some transcription factor binding, indicating that some more subtle interactions such as our Fp2 region were lost in their analysis.
We have recently shown that CBS promoter activity is critically linked to growth state (10), and the role of proliferation status in regulating CBS promoter activity offers a possible explanation for both of these discrepancies. The previously published work did not provide any details of the degree of confluence the HepG2 cells were at when harvested for luciferase assays or for making nuclear extracts. The failure of this group to detect promoter activity in constructs with less than 125 bp of Ϫ1b proximal sequence or protection of the Fp2 region could easily be due to harvesting cells at a sub-optimal level of confluence where promoter activity is dramatically reduced.
This work is the first report of the site-specific synergistic interaction between Sp1 and Sp3 in regulating the CBS Ϫ1b promoter. Previously, another group examined the effect of co-transfecting Sp1 and Sp3 upon the expression of a 125 bp (Ϫ3792 to Ϫ3667) promoter fragment in SL2 cells but found no evidence of a synergistic interaction within this sequence (12). This finding is consistent with our observation that only constructs containing the tandem Sp1 binding sites immediately upstream of this region are capable of synergistic up-regulation. Although there are numerous functional Sp1 sites in the Ϫ1b promoter that contribute to basal activity, they are clearly not all equivalent in their relative contribution.
In our SL2 co-transfection experiments, and those of others, it was observed that Sp3 or to a lesser extent, NF-Y could substitute for Sp1 in activating the CBS Ϫ1b promoter. The failure of these transcription factors to substitute for Sp1 in the Sp1-deficient fibroblasts cells is therefore very interesting particularly, because transfection of this cell line with an Sp1 expression construct was sufficient to induce relatively high levels of CBS expression. Co-transfection studies along with EMSA and DNase I footprinting are very useful and widely used tools for the functional analysis of promoter regions but a potential problem associated with these approaches is that they use cloned fragments that have been produced in Escherichia coli. As a result, these DNA fragments lack the cellspecific methylation, histone acetylation, and general chromosomal architecture that is present in vivo. Additionally, the interpretation of binding experiments using artificially high concentrations of overexpressed transcription factors is complicated when analyzing the role of different transcription factors that bind to very similar cognate recognition sequences such as Sp1 and Sp3. The ability of Sp3 to substitute for Sp1 in SL2 cell co-transfection experiments but not in the Sp1-deficient fibroblasts highlights the need for in vivo correlates to support conclusions derived from these kinds of studies.
One model to explain our findings is that Sp1 and Sp3 form a ternary complex with each other prior to binding the CBS Ϫ1b promoter region. Alternatively, because Sp1 binding has previously been shown to be capable of inducing conformational changes in DNA (50,51), it is conceivable that one of the roles of Sp1 is to induce conformational changes in the Ϫ1b promoter region that facilitate concomitant binding of Sp3 with a resultant synergistic increase in CBS promoter activity when cellular growth conditions require. Either of these models would explain why Sp3 is unable to substitute for Sp1 in the Sp1deficient fibroblasts and why heterologous expression of Sp1 in this line is accompanied by an induction of CBS.
We have confirmed the previous observation that NF-Y plays a role in regulating the CBS Ϫ1b promoter (12,13). However, our observation that transfection with the NF-Y dominant negative mutant NF-YA29 results in only a 30% decrease in CBS promoter activity in HepG2 cells and that NF-Y is unable to drive transcription in the absence of Sp1 indicate that the role of NF-Y in regulating CBS is not dominant. The relative importance of NF-Y in promoters varies considerably. Relatively simple TATA-less promoters have been shown to have an absolute requirement for an intact CCAAT box for function. Other stronger promoters have been shown to be less dependent. It is clear from the work presented in this report that the CBS Ϫ1b promoter is not dependent upon NF-Y. This finding is consistent with previous reports that NF-Y by itself is a relatively weak activator of transcription and that its main function is to enhance transactivation from adjacent sequence elements or to enhance the binding of other transcription factors (52).
Previous work has shown that the level of CBS expression in different cell types is directly associated with the degree of binding by Sp1/Sp3, which is possibly influenced by phosphorylation (13). This finding is consistent with our observation that CBS regulatory factors such as proliferation status and intracellular redox clearly influence the degree of binding of Sp1 to the CBS Ϫ1b promoter. These authors proposed that the relative ratios of Sp1 and Sp3 and their phosphorylation by protein kinase A could be a mechanism for the tissue-specific regulation of CBS. However, CBS is not simply expressed at low levels in some tissues and high levels in others. It is completely absent from a wide range of tissues all of which express Sp1 and protein kinase A. Consequently, it is highly unlikely that differences in either the phosphorylation status or relative abundance of two ubiquitously expressed positively acting transcription factors could be responsible for the complete absence of CBS from specific tissues. In this context, the finding that other members of the Sp/KLF family such as BKLF, LKLF, and KLF8 are capable of repressing both CBS promoters by competing for Sp1 binding sites offers a novel insight on how the CBS gene might be regulated in a tissuespecific fashion. In contrast to Sp1 and Sp3, these genes are not ubiquitously expressed. Interestingly, LKLF has a tissue distribution that is essentially restricted to CBS-negative tissues (53). Perhaps most revealing is the relative expression patterns of LKLF and CBS in resting and proliferating lymphocytes. In quiescent lymphocytes, LKLF is readily detectable and CBS is completely absent. When these cells are induced to proliferate by mitogenic stimuli, expression of LKLF is rapidly extinguished and CBS is induced (54,55).
However, although our KLF co-transfection experiments and the expression patterns of some members of the KLF family are consistent with a role in regulating CBS, several lines of evidence indicate that the suppression of CBS expression may be more complicated. First, none of the KLF expression constructs was capable of completely abolishing CBS expression in the co-transfection experiments. Characterization of the regulatory interactions of KLF family members with other transcription factors is still at a very early stage, so it is conceivable that regulators like LKLF could require the presence of other transcription factors and/or regulatory processes such as methyla-tion to completely block CBS expression in a cell-specific manner. A more fundamental problem arises when one considers the obvious activity of CBS Ϫ1a and Ϫ1b promoter expression constructs after transfection into the lung-derived CBS-negative E10 cell line. Whatever mechanisms are repressing expression of the chromosomal CBS promoters in these cells, they are clearly not repressing the expression of the CBS Ϫ1a and Ϫ1b promoter constructs.
One obvious difference between the silenced chromosomal promoters and the plasmid-based promoters is the absence of the normal chromosomal architecture, including chromatin, histones, and methylation. Our work in this report has shown with Sp1 and Sp3 that this can significantly impact the behavior of transcription factors when compared with plasmid-based transfection experiments. In this report we have shown that overexpression of LKLF in HepG2 cells is capable of downregulating CBS expression in a natural chromosomal context, and it is thus conceivable that differences in the architecture of the CBS promoter in the chromosomal context impact upon the binding affinities of KLF family members and/or other accessory proteins. Alternatively, it remains possible that a sequence element outside of the 5 kbp of 5Ј-flanking sequence that we investigated is required to confer tissue-specific expression. Further investigations are currently in progress in our laboratory to characterize the mechanisms that control tissuespecific regulation of CBS.