Transcriptional Regulation of Cell-specific Expression of the Human Cystathionine (cid:1) -Synthase Gene by Differential Binding of Sp1/Sp3 to the (cid:2) 1b Promoter*

,

Cystathionine ␤-synthase (CBS 1 ; EC 4.2.1.22), a pyridoxal 5Ј-phosphate-dependent enzyme involved in the transsulfuration pathway, catalyzes the condensation of L-serine and Lhomocysteine to form cystathionine, an intermediate step in the synthesis of cysteine. The human CBS gene product is a 63,000-Da polypeptide. The tetrameric protein is catalytically dependent on both heme and pyridoxal phosphate and is allosterically regulated by S-adenosylmethionine (1)(2)(3)(4). The human CBS gene spans over 30 kilobases and consists of 23 exons ranging in size from 42 to 209 base pairs (5). The CBS polypeptide is encoded by exons 1-14 and 16. The human CBS gene encodes multiple mRNAs differing in their 5Ј-untranslated regions, resulting from the use of five alternative noncoding exons (designated Ϫ1a to Ϫ1e) and a constant exon 0. Transcripts containing exons Ϫ1a and Ϫ1b appear to be the most abundant and are found in an assortment of adult and fetal tissues (6). By contrast, use of exons Ϫ1c, Ϫ1d, and Ϫ1e appears to be rare. There are at least two GC-rich TATA-less promoters upstream of exons Ϫ1a and Ϫ1b, containing numerous putative transcription elements (Sp1 (specificity protein 1), AP1, AP2, etc.) (5,6). In our recent study of the CBS Ϫ1b minimal promoter (mapping between positions Ϫ3792 and Ϫ3667), we demonstrated important transactivating roles for Sp1, Sp3, NF-Y, and USF-1 (7).
The CBS gene has been localized to human chromosome 21 (e.g. 21q22. 3), and its overexpression has been suggested to be linked to certain of the phenotypic features of Down syndrome (DS). Elevated CBS expression in DS results in low plasma homocysteine compared with non-DS individuals and has been suggested to contribute to decreased atherosclerosis in DS patients (8). In our own studies of CBS and DS, we also found striking increases (ϳ12-fold) in CBS transcripts in myeloblasts from DS children with AML compared with non-DS myeloblasts (9). Interestingly, this elevated CBS expression was associated with increased in vitro sensitivities to cytosine arabinoside (Ara-C) and generation of Ara-C triphosphate, likely due to downstream effects of CBS on endogenous folate and nucleotide pools (9 -12). This may explain, in part, the remarkably high event-free survival rates (70 -100%) and low relapse rates (Ͻ15%) of DS children with AML compared with non-DS children treated with Ara-C-based chemotherapy protocols. Increased CBS transcripts in DS over non-DS myeloblasts in-volve those transcribed from the CBS Ϫ1b promoter (13) and may arise from differences in mechanisms of transcriptional control. In this study, we significantly extend our earlier studies of CBS promoter structure and function by identifying the critical cis-elements and transcription factors in the CBS Ϫ1b promoter upstream region.
Cell Culture-The human HT1080 fibrosarcoma and HepG2 hepatocellular carcinoma cell lines were obtained from American Type Culture Collection (Manassas, VA). Drosophila SL2 cells were provided by Dr. Bonnie Sloane (Wayne State University, Detroit, MI). The HT1080 cell line was maintained in RPMI 1640 medium containing 10% heatinactivated iron-supplemented calf serum (Hyclone Laboratories), 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin in a humidified atmosphere at 37°C in the presence of 5% CO 2 and 95% air. The HepG2 cell line was maintained in minimal essential medium with 10% fetal bovine serum and the antibiotics. SL2 cells were maintained in Schneider's insect medium supplemented with 10% fetal bovine serum and 2 mM glutamine plus antibiotics at 25°C.
Northern Blot Analysis and Assay of CBS mRNA Turnover-Total RNAs were isolated from HT1080 and HepG2 cells using Trizol reagent (Life Technologies, Inc.). Total RNA from each cell line (20 g) was electrophoresed on 0.9% agarose gel containing 2.2 M formaldehyde and 1ϫ MOPS and capillary-transferred to GeneScreen Plus membrane (PerkinElmer Life Sciences). The membrane was baked at 80°C for 1.5 h in vacuo, prehybridized, and hybridized with [␣-32 P]dCTP-labeled CBS cDNA (labeled by random priming). The membrane was washed to a final stringency of 0.1ϫ SSC and 0.1% SDS at 42°C and exposed to film. Densitometry was performed on a Molecular Dynamics Storm 860 fluorescence and radioactivity imaging system with ImageQuant software.
For assays of CBS transcript turnover, HT1080 and HepG2 cells in replicate 150-cm dishes were treated with 10 g/ml actinomycin D when they were 70% confluent. Cells were harvested every 2 h over 12 h, and total RNA was isolated using Trizol reagent. Changes in CBS transcripts were followed by Northern blotting.

Construction of Luciferase Plasmids and Site-directed Mutagenesis-
The full-length CBS Ϫ1b promoter-reporter gene construct pCBSbϪ 4046/Ϫ3565, the basal CBS Ϫ1b promoter construct pCBSbϪ3792/ Ϫ3565, and the CAAT box-mutated basal CBS Ϫ1b promoter construct were generated as previously described (7).
A series of luciferase plasmids harboring nucleotide substitutions that alter or abolish consensus sequences for transcription factor binding were also constructed by site-directed mutagenesis using an overlap extension PCR protocol (14). Two separate PCR fragments for each half of a final hybrid product were generated employing mutagenesis primers (see Table I) and outside primers PROB/F2 and PROB/R2 (7). The two products were mixed, and a second PCR was performed using the two outside primers. The resulting products were blunt-ended and subcloned into the SmaI site of pGL3-Basic. All deletion and sitedirected mutagenesis constructs were sequenced to confirm the intended mutations.
Transient Transfections and Luciferase Assay-CBS-luciferase reporter gene constructs or promoterless vector pGL3-Basic (1 and 4 g for HepG2 and HT1080 cells, respectively) were cotransfected with 12.5 ng of pRLSV40 into ϳ50% confluent HT1080 and HepG2 cells using Lipofectin reagent (Life Technologies, Inc.) in accordance with the manufacturer's protocols. Lipofectin treatments were for 24 h; and after an additional 48 h of incubation in complete medium, cells were harvested, and lysates were prepared. Firefly luciferase activities were assayed with a dual-luciferase reporter assay system (Promega) in a Turner TD20/20 luminometer and normalized to Renilla luciferase activity.
Drosophila SL2 cells were cotransfected with 1 g of the CBS-luciferase reporter gene constructs and 25-200 ng of the Sp1 (pPacSp1) or NF-Y (pPacNF-YA, pPacNF-YB, and pPacNF-YC) cDNA constructs using Fugene TM 6 reagent (Roche Molecular Biochemicals) according to the manufacturer's recommendations. Cells were harvested after 48 h for luciferase assays using a single-luciferase assay system (Promega). Luciferase activities were normalized to cellular protein, measured by the Bio-Rad protein assay system.
Preparation of Nuclear Extracts and DNase I Footprint Analysis-Nuclear extracts from HT1080 and HepG2 cells were prepared by standard methods (15), with slight modifications. Final protein concentrations were determined by the Bio-Rad protein assay. DNase I footprinting was performed using the core footprinting kit (Promega) according to the manufacturer's recommendations. 50 -100 g of nuclear proteins from HT1080 and HepG2 cells were used for each reaction. The DNA sample was analyzed on an 8% polyacrylamide sequencing gel. Corresponding sequencing reactions were performed by PCR using a Thermo Sequenase radiolabeled terminator cycle sequencing kit (United States Biochemical Corp.).
Western Blot Analysis-Nuclear protein was isolated from HT1080 and HepG2 cells as described above. 50-g aliquots of each nuclear protein were fractionated on a 7.5% polyacrylamide gel with SDS and electroblotted onto a polyvinylidene difluoride membrane. The blot was blocked overnight at room temperature in TTBS (Tween/Tris-based saline with 0.1% Tween 20 (pH 7.5)) containing 1% fat-free dried milk powder and was then incubated with anti-Sp1 or anti-Sp3 antibody (diluted 1:10,000 and 1:2000, respectively) in TTBS containing 0.5% fat-free dried milk powder for 2 h at room temperature. The blot was washed with TTBS, incubated with a secondary antibody (horseradish peroxidase-conjugated goat anti-rabbit IgG diluted 1:5000 in TTBS FIG. 1. Northern blot analysis of CBS transcripts in HT1080 and HepG2 cells. Total RNAs (20 g) from HT1080 and HepG2 cells were electrophoresed on an agarose gel, transferred to GeneScreen Plus membrane, and hybridized to a 32 P-labeled CBS cDNA probe. The 2.9-kilobase band corresponds to the major CBS transcript in these two cell lines. Ethidium bromide-stained 28 S and 18 S RNAs were used as references to permit normalization. containing 0.5% milk powder) for 1 h at room temperature, and detected by Lumi-Light Western blotting substrate (Roche Molecular Biochemicals).
Phosphorylation and Dephosphorylation of Nuclear Extracts-HT1080 nuclear extracts (100 g) were incubated with or without 40 units of protein kinase A catalytic subunit (Sigma) in 50 mM Tris (pH 7.4), 10 mM MgCl 2 , and 10 M ATP at 30°C for 1 h in a total volume of 50 l. 10 g of the treated nuclear extract were used for EMSA as described above. Nuclear extracts (100 g) from HepG2 cells were incubated with or without 20 units of calf intestinal alkaline phosphatase (Promega) for 1 h at 37°C in a total volume of 30 l. The reaction was stopped by the addition of a mixture of phosphatase inhibitors to final concentrations of 10 nM NaF, 10 nM okadaic acid, and 10 nM Na 3 MoO 3 . 10 g of the treated nuclear extract were used for EMSA.

CBS Gene Expression and mRNA Turnover in HepG2 and
HT1080 Cells-The levels of CBS transcripts were compared between human HT1080 fibrosarcoma and HepG2 hepatocellular carcinoma cell lines on Northern blots. For both lines, a single 2.9-kilobase transcript was detected; however, the levels were ϳ10-fold higher in HepG2 compared with HT1080 cells (Fig. 1).
This disparity in the levels of CBS transcripts could conceivably reflect differences in mRNA stabilities between the lines. To assess this possibility, cells were treated with 10 g/ml actinomycin D, and changes in the levels of CBS transcripts were followed over 12 h on Northern blots. Under these conditions, CBS turnover was virtually identical (half-lives of ϳ10 h for both HT1080 and HepG2 cells) (data not shown). This suggests that the difference in the levels of CBS transcripts between the lines likely arises at the transcriptional level. Thus, the HT1080 and HepG2 cell lines appeared to be appropriate models for a systematic study of the transcriptional regulation of cell-specific expression of the human CBS gene.
Since two promoters were previously identified in the human CBS gene (GenBank TM /EBI Data Bank accession number AF042836) (4), designated Ϫ1a (positions Ϫ4467 to Ϫ4093) and Ϫ1b (positions Ϫ4055 to Ϫ3576), it was essential to establish whether the differential gene expression between the HT1080 and HepG2 lines on Northern blots was due to differences in promoter usage. To test this possibility, a 5Ј-RACE assay was performed. Essentially identical results for HT1080 and HepG2 cells were obtained. For both lines, all the transcripts contained 5Ј-untranslated regions with exon Ϫ1b sequence. Four of the five 5Ј-RACE subclones sequenced from HT1080 cells were identical and began at position Ϫ3637, or 28 base pairs downstream from the reported 5Ј-end of exon Ϫ1b. One clone contained 58 additional nucleotides upstream from this position (position Ϫ3695) (data not shown). For the HepG2 cells, all four 5Ј-RACE clones were identical and began at position Ϫ3634 (7). For all of the 5Ј-RACE clones from both lines, there were no differences from the published GenBank TM /EBI Data Bank sequence for human CBS (accession number AF042836). These results establish that both HT1080 and HepG2 cells exclusively use the Ϫ1b promoter and suggest that transcription of CBS Ϫ1b transcripts initiates at multiple sites.
In Vitro Analysis of the Upstream Region of the CBS Ϫ1b Promoter-In our previous study (7), we localized the basal CBS Ϫ1b promoter to positions Ϫ3792 to Ϫ3666 (relative to ATG) and identified Sp1/Sp3, NF-Y, and USF-1 as important transactivating factors (Fig. 2, lower panel shows a schematic of the CBS Ϫ1b promoter region). Our results also suggested an important transcriptional role for the CBS Ϫ1b promoter region upstream of position Ϫ3792 since progressive 5Ј-deletions from the full-length promoter construct (positions Ϫ4046 to Ϫ3565) in pGL3-Basic resulted in a 70% loss of luciferase activity in HepG2 cells, accompanying deletions from positions Ϫ4046 to Ϫ3792 (7). However, upon 3Ј-deletion of the basal CBS Ϫ1b promoter region from positions Ϫ3565 to Ϫ3792, promoter activity was completely lost. This shows that the region between positions Ϫ4046 and Ϫ3792 is by itself incapable of driving luciferase activity.
To localize the major regulatory elements in the CBS Ϫ1b promoter upstream region, DNase I footprint analysis was performed using two overlapping probes (positions Ϫ4046 to Ϫ3792 and positions Ϫ3938 to Ϫ3659) and nuclear extracts prepared from HT1080 and HepG2 cells. Four clearly protected regions were demarcated with nuclear extracts from HepG2 cells (Fig. 2, upper panel, designated A, B, C, and D) that, from data base analysis, contained numerous transcription elements, including GC boxes (designated GCe, GCf, and GCg), a GT box (GTd), MZF-1 (MZF-1a and MZF-1b), and Ikaros-2 (IK2) (Fig. 2, lower panel). Conversely, protected regions with equivalent amounts of nuclear extracts from HT1080 cells were far less obvious (Fig. 2, upper panel).
Differential Binding of Sp1/Sp3 to the CBS Ϫ1b Promoter in HT1080 and HepG2 Cells-To characterize transcription factors that interact with the binding sites in the upstream region of the CBS Ϫ1b promoter, EMSAs and supershift assays were performed. Synthetic FPA (labeled a in Table I), FPB (labeled c), FPC (labeled f), and FPD (labeled h) oligonucleotides (containing wild-type CBS Ϫ1b promoter sequence) were designed from the protected regions on DNase footprints (Fig. 2) and database analysis of potential transcription factor-binding sequences. Double-stranded FPA, FPB, FPC, and FPD oligonucleotides were labeled with [␥-32 P]ATP and incubated with nuclear extracts prepared from HT1080 and HepG2 cells. In HepG2 cells, 11 major DNA-protein complexes were detected (numbered 1-11 in Fig. 3, A-D, lanes 4, 12, 26, and 36) that were effectively competed by a 100-fold excess of unlabeled nucleotides (lanes 5, 13, 27, and 37), establishing specificity. The identities of the bound transcription factors were assessed by competitions with unlabeled competitor oligonucleotides (Table I). Complexes 1, 3, 5-8, 10, and 11 were nearly completely abolished by competition with a commercial Sp1 consensus oligonucleotide (labeled k Table I; Fig. 3, A-D, lanes 6, 14, 28, and 38). With the exception of complex 11, these were all supershifted by antibodies to Sp1 and/or Sp3 (lanes 8, 9, 21, 22, 31, 32, 44, and 45). Complex 2 was partially competed by an Sp1 consensus oligonucleotide (lane 6) and partially supershifted by anti-Sp1 antibody (lane 8). For anti-Sp3 antibody, the loss of signal is consistent with an effect on a DNA-Sp3 protein complex. Thus, complex 2 appears to represent two DNA-protein complexes migrating together, only one of which involves Sp1 and Sp3.
These results establish that DNA-protein complexes 1, 3, 5-8, and 10 and part of complex 2 involve Sp1 and/or Sp3. Complexes 4 and 9 appear to be MZF-1-like proteins. The identity of the additional factor comprising complex 2 is unclear. Since complex 11 could be completely competed by an   Table I. For the supershift assays, specific antibodies to Sp1 (Sp1-ab) and Sp3 (Sp3-ab) were added to the reaction mixtures and incubated for 30 min prior to electrophoretically separating the DNA-protein complexes. The major DNA-protein complexes are numbered 1-11, and the supershifted complexes for Sp1 (Sp1-SS) and Sp3 (Sp3-SS) are also noted. In E are shown the results from a gel shift assay of the Ϫ3724/Ϫ3682 and Ϫ3766/Ϫ3725 oligonucleotides comprising the basal CBS Ϫ1b promoter region (7) using nuclear extracts prepared from HepG2 and HT1080 cells. The major complexes, including Sp1/Sp3, NF-Y, NF-1, and USF-1, are indicated. NS designates a nonspecific complex. Sp1 consensus oligonucleotide, but not supershifted by antibody to Sp1 or Sp3, this might be another member of the Sp family, such as Sp4. The DNA-protein complexes identified in gel shift assays in the FPA, FPB, FPC, and FPD regions of the CBS Ϫ1b promoter with HepG2 nuclear extracts are summarized in Table II.
In contrast to HepG2 cells, the levels of the DNA-protein complexes in HT1080 nuclear extracts were nearly undetectable (Fig. 3, A-D, lanes 2, 11, 24, and 34). Gel shift assays were also performed with probes including the CBS Ϫ1b minimal promoter (CBS Ϫ1b Ϫ3729/Ϫ3682 and Ϫ3766/Ϫ3725 oligonucleotides) (7) and nuclear extracts from HT1080 and HepG2 cells (Fig. 3E). Although Sp1 and Sp3 did not bind appreciably to either the upstream or the minimal promoter regions in extracts prepared from HT1080 cells (lanes 47 and 52), binding of NF-Y, USF-1, and NF-1 was detected to the same extent as in HepG2 cells (i.e. compare lanes 47 and 49 and lanes 52 and  54).
Functional Analysis of Transcription Factor-binding Sites by Site-directed Mutagenesis-To further confirm the functional significance of the DNA-protein complexes detected by gel shift and supershift analyses in Fig. 3, the Sp1/Sp3 and MZF-1 consensus elements in the CBS Ϫ1b promoter upstream region were mutated individually using the mutant oligonucleotides in Table I, and the mutant CBS-reporter gene constructs were transiently transfected into HT1080 and HepG2 cells. Luciferase activities of the mutant constructs were compared with that of the wild-type full-length promoter construct pCBSbϪ4040/Ϫ3565 (Fig. 4).
Qualitatively similar results were obtained with HT1080 and HepG2 cells despite striking differences in promoter activities. Although 4-fold more reporter plasmid (with a constant 4 g of plasmid) was used for the HT1080 transient transfections, maximal luciferase activity (normalized to Renilla luciferase activity) was still ϳ6-fold less than in the HepG2 cells (Fig. 4). Since cotransfections with the pRLSV40 plasmid resulted in nearly identical levels of Renilla luciferase activity in both HT1080 and HepG2 cells, this disparity was not due to differences in transfection efficiencies between the lines.
The effects of mutations of the assorted cis-elements implicated by EMSAs were variable despite similar losses of factor binding in gel shift assays. Thus, mutation of GCg (FPA-GCg mt) (Table I) resulted in an ϳ30% loss of CBS Ϫ1b promoter activity in both lines. Greater losses of CBS Ϫ1b promoter activity accompanied mutation of the MZF-1b (FPD-MZF-1b mt; decreases of 59 and 45% for HT1080 and HepG2 cells, respectively) and GTd (FPD-GTd mt; decreases of 73 and 55% for HT1080 and HepG2 cells, respectively) elements. Mutation of the MZF-1a site (FPB-MZF-1a mt) only slightly suppressed the promoter activity (ϳ15% in both lines). Interestingly, mutation of GCf (FPB-GCf mt) was accompanied by a potent loss of promoter activity (ϳ90% in both lines). These results confirm an important transactivating role for the GCf element in the CBS Ϫ1b promoter and, to a lesser extent, for other GC/GT boxes (GCg, GCe, and GTd) and the MZF-1 elements.
Collectively, these results demonstrate that an assortment of transcription factors, including Sp1, Sp3, and MZF-1-like proteins, can bind to the CBS Ϫ1b promoter upstream region, resulting in transactivation. The differential binding of Sp1/ Sp3 to the CBS Ϫ1b promoter for HepG2 versus HT1080 cells may explain, in part, the differences in the levels of CBS transcripts and CBS Ϫ1b promoter activity between the lines.
Cell-specific Transactivation of the CBS Ϫ1b Promoter by MZF-1-The role of MZF-1 in the regulation of the CBS Ϫ1b promoter was assessed through transient cotransfections of the CBS Ϫ1b promoter-reporter gene construct pCBSbϪ4046/ Ϫ3565 and an MZF-1 cDNA expression vector (pcDNA3-MZF-1) in HT1080 and HepG2 cells. Cotransfections with pcDNA3-MZF-1 resulted in a stimulation of CBS Ϫ1b promoter activity in HT1080 cells (1.7-and 2.3-fold at 50 and 100 ng, respectively). Conversely, the effect of overexpressing MZF-1 on CBS Ϫ1b promoter activity in HepG2 cells was insignificant (Fig. 5). These studies suggest that MZF-1 binds to the CBS Ϫ1b promoter and transactivates promoter activity and that this occurs in a cell-specific manner.  Table I. Luciferase activities of the mutant constructs were normalized to Renilla luciferase activity and compared with that of the wild-type Ϫ4046/Ϫ3565 construct. Data are presented as the means Ϯ S.E. from three independent experiments performed in duplicate.  Fig. 3  (A-D).

Functional Analysis of the GC and GT Boxes in the Upstream
b A second unidentified transcription factor was detected along with Sp1/Sp3. c This complex was identified as an Sp family member since it could be competed by Sp1 consensus oligonucleotide; however, it could not be supershifted by either anti-Sp1 or Sp3 antibody. Its identity is unclear. sophila SL2 cells, which provide a null background for the Sp family of transcription factors and NF-Y (16,17). The pCBSbϪ4046/Ϫ3565 reporter gene construct including the fulllength CBS Ϫ1b promoter was cotransfected with expression vectors for Sp1 (pPacSp1) (16) and/or NF-Y (pPacNF-Y), each under control of the Drosophila-specific promoter. Parallel transfections were performed with mutant CBS Ϫ1b promoter constructs in which each of the GC/GT box elements was mutated (Table I).
In cotransfections with pPacSp1, CBS Ϫ1b promoter activity for the wild-type construct was stimulated (Fig. 6A). Activity for the mutant GTd construct was nearly the same as for the wild-type construct. Interestingly, mutation of the GCg and GCe boxes effected an activation (2.0-and 1.3-fold, respectively) of luciferase activity in these experiments, suggesting a trans-repressive role for these GC box elements. As with the HT1080 and HepG2 cells, mutation of the GCf box potently suppressed CBS Ϫ1b promoter activity in SL2 cells (80%) (Fig.  6A).
Sp1 transactivation of the full-length CBS Ϫ1b promoter (positions Ϫ4046 to Ϫ3565) exceeded that of the basal promoter (positions Ϫ3792 to Ϫ3565) by ϳ5-fold (Fig. 6B). When both Sp1 and NF-Y were cotransfected together with the full-length promoter construct, a striking synergism was observed (ϳ7fold over that in cotransfections with Sp1 alone) (Fig. 6B). This synergistic activation far exceeded that with the basal promoter construct (2.3-fold) (Fig. 6B) (7). Thus, deletion of the CBS Ϫ1b promoter upstream region significantly decreases transactivation by Sp1 alone and the extent of synergistic transactivation by Sp1 in combination with NF-Y.
These results further suggest important transcriptional roles for the Sp family of factors via their binding to the upstream GC or GT box elements in the CBS Ϫ1b promoter. Furthermore, they demonstrate that the transactivating effects of Sp factors and NF-Y are highly synergistic.
Differential Binding of Sp1/Sp3 to the CBS Ϫ1b Promoter Is Not Due to Differences in Levels of Sp1/Sp3 or to Specific Cofactor(s)-Since Sp1 and Sp3 levels were nearly identical in HT1080 and HepG2 cells (Fig. 7), other explanations are necessary to explain the differences in Sp factor binding to the CBS Ϫ1b promoter upstream region in EMSAs and, potentially, transcriptional activity between the lines. This could reflect the presence of an Sp1/Sp3 binding inhibitor in HT1080 cells or an activator in HepG2 cells. To test these possibilities, 10 g of a nuclear extract from HepG2 cells were mixed with different amounts (from 2.5 to 10 g) of nuclear extract from HT1080 cells, and Sp1/Sp3 binding was analyzed in gel shift assays with 32 P-labeled FPD probe. As shown in Fig. 8, binding of Sp1 and Sp3 in the HepG2 nuclear extract to 32 P-labeled FPD was unaffected by the addition of increasing amounts of HT1080 nuclear extract. Thus, the presence of an Sp1/Sp3 binding inhibitor or activator is unlikely.
Phosphorylation of Sp1 Modulates Its Binding to the CBS Ϫ1b Promoter in Vitro-Sp1 phosphorylation has been reported to increase trans-factor binding to GC box elements (18) and to facilitate promoter activation for a number of genes (18 -20). A similar mechanism may be operative for CBS. Consistent with this possibility, treatment of nuclear extracts from HT1080 cells with protein kinase A catalytic subunit prior to EMSA resulted in an increase in Sp1/Sp3 binding to the 32 Plabeled FPD oligonucleotide by EMSA (Fig. 9A, lane 4). Likewise, pretreatment of HepG2 nuclear extracts with calf intestinal phosphatase significantly decreased Sp1/Sp3 binding to the FPD probe (Fig. 9B, lane 12). Thus, the differential Sp1/Sp3 binding to the CBS Ϫ1b promoter in HT1080 and HepG2 cells appears to be at least partially due to differences in the phosphorylation status of Sp1/Sp3 between these two lines. The lack of a complete response to protein kinase A treatment may reflect a requirement for additional post-translational modifications (phosphorylation by other protein kinases or glycosylation) of Sp1/Sp3 or other mechanisms altogether. DISCUSSION CBS plays a very important role in human disease. Deficiencies of the enzyme result in the genetic disorders homocystinuria and hyperhomocystinemia (40,41), whereas increased expression of the CBS gene may contribute to some of the phenotypic features of DS (9). Our previous studies suggested an important role for variations in CBS gene expression as a determinant of the enhanced Ara-C sensitivities of DS myeloblasts and CBS-transfected leukemia cell line models (9, 12). A wide range of CBS gene expression has been described among assorted human tissues (6). Furthermore, multiple transcripts have been reported for CBS (6). Five distinct 5Јnoncoding exons have been described, the most frequent termed Ϫ1a and Ϫ1b, each encoded by its own unique GC-rich TATA-less promoter (5). In our previous study, we defined a minimal transcriptional region (positions Ϫ3792 to Ϫ3667) of the CBS Ϫ1b promoter by 5Ј-and 3Ј-deletions and transient transfections in HepG2 cells, a cell line characterized by high levels of CBS transcription exclusively from the Ϫ1b promoter (7). Included in this 125-base pair CBS Ϫ1b minimal promoter region are three GC boxes (termed GCa, GCb, and GCc), an inverted CAAT box, and an E box. In gel shift and supershift assays, binding of Sp1 and Sp3 to the GC box elements, USF-1 to the E box, and both NF-Y and an NF-1-like factor to the CAAT box could be demonstrated. Synergism between Sp1/Sp3 and NF-Y in CBS transactivation was also observed (7). The functional significance of the region upstream of the CBS Ϫ1b minimal promoter (positions Ϫ3792 to Ϫ3565) was suggested by 5Ј-deletion analysis (7).
In this study, we used DNase I footprint analysis of the Ϫ4046/Ϫ3792 fragment, immediately upstream of the basal promoter region, to identify protected regions in HT1080 and HepG2 cells, designated FPA (positions Ϫ3992 to Ϫ3970), FPB (positions Ϫ3955 to Ϫ3935), FPC (positions Ϫ3874 to Ϫ3849), and FPD (positions Ϫ3813 to Ϫ3790). Four potential GC or GT box elements, two MZF-1 sites, and an IK2-binding site were identified by data base analysis. In gel shift and supershift assays with a HepG2 nuclear extract, binding of Sp1 and Sp3 to the GC and GT box elements (designated GCg, GCf, GCe, and GTd) and binding of MZF-1-like proteins to MZF-1-binding sites in FPB and FPD (designated MZF-1a and MZF-1b, respectively) were demonstrated. Interestingly, in nuclear extracts prepared from HT1080 cells, Sp1 and/or Sp3 DNA-protein complexes were detected in only trace amounts despite comparable binding of CAAT-binding factors (NF-Y and NF-1) and USF-1 and nearly identical levels of Sp1 and Sp3 proteins on Western blots.
The functional roles of all these upstream elements in CBS Ϫ1b promoter activity were further evaluated by mutating these sequences and transfecting the mutant constructs into HT1080 and HepG2 cells. These mutations all effected changes in promoter activities; however, this was somewhat variable and was strongly dependent on the expression model. Thus, mutations of GCg, GCe, GTd, and the two MZF-1-binding sites (MZF-1a and MZF-1b) individually resulted in losses (15-73%) of CBS Ϫ1b promoter activity in HT1080 and HepG2 cells. Mutation of GCf resulted in a striking loss of promoter activity (90%) in both lines (Fig. 4), suggesting a critical transcriptional role for this GC box element. FIG. 7. Western blot analysis of Sp1 and Sp3 in HT1080 and HepG2 nuclear extracts. 50 g of nuclear extract protein from HT1080 and HepG2 cells were fractionated on a 7.5% polyacrylamide gel with SDS and electroblotted onto a polyvinylidene difluoride membrane. Immunoreactive Sp1 and Sp3 proteins were detected with anti-Sp1 and anti-Sp3 antibodies, respectively, and Lumi-Light Western blotting substrate.
FIG. 8. Combinations of HT1080 and HepG2 nuclear extracts do not alter the binding of Sp1 and Sp3 to the CBS Ϫ1b promoter. 10 g of nuclear extracts from HepG2 cells were preincubated with different amount of nuclear extracts (NE) from HT1080 cells (2.5-10.0 g) on ice for 1 h under the gel shift assay binding conditions. The 32 P-labeled FPD probe was added; and following incubation on ice for 30 min, the DNA-protein complexes were electrophoresed. Data are shown for the free probe (no extract; lane 1), DNA-protein complexes in the absence of competitors (lanes 2 and 4 for HT1080 and HepG2 cells, respectively), competition with a 100-fold molar excess of unlabeled FPD oligonucleotide (lanes 3 and 5), and DNA-protein complexes at a constant amount of HepG2 extract with increasing amounts of HT1080 nuclear extract (lanes 6 -9). The specific DNA-protein complexes are numbered as described in the legend to Fig. 3D. NS refers to a nonspecific DNA-protein complex.

FIG. 9. Phosphorylation of Sp1 modulates its binding activity in vitro. Nuclear extracts from HT1080 cells (A) and HepG2 cells (B)
were pretreated with protein kinase A (PKA) catalytic subunit and calf intestinal alkaline phosphatase (CIP). The 32 P-labeled FPD probe was added; and following incubation on ice for 30 min, the DNA-protein complexes were electrophoresed as described under "Materials and Methods." 10 g of treated nuclear extracts (NE) were used for the gel shift assays. Lanes 5 and 10 show competition with a 100-fold molar excess of unlabeled FPD oligonucleotides; lanes 6 and 11 show competition with a 100-fold excess of commercial Sp1 oligonucleotide. Specific DNA-protein complexes are numbered as described in the legend to Fig.  3D. NS indicates a nonspecific DNA-protein complex.
In Drosophila SL2 cells, the effects of these mutations ranged from moderate (GTd mt, ϳ15%) to potent (GCf, ϳ80%) inhibition to a strong activation response (GCg, ϳ200%). Sp1 and Sp3 have been previously reported to function as either transcriptional activators or repressors, depending on the cell and promoter context (21)(22)(23)(24). Interactions between NF-Y binding to the CAAT box and Sp1/Sp3 binding to the GC boxes were indicated since in cotransfections of Sp1 and NF-Y with the full-length construct, a potent synergistic transactivation of the CBS Ϫ1b promoter was detected (7-fold), exceeding that obtained with the basal CBS Ϫ1b promoter (2.3-fold) (7). There is ample precedent for interactions between Sp1 and NF-Y in the functional transactivation of a number of genes (30 -39).
MZF-1 is a myeloid zinc finger protein that is essential for granulopoiesis (25). MZF-1 contains 13 zinc finger domains, divided into two groups (26). The amino-terminal group has four zinc fingers, and the carboxyl-terminal group has nine zinc fingers. MZF-1 is a bifunctional transcription factor, capable of repressing transcription in non-hematopoietic cells and activating transcription in cells of hematopoietic origin (25). In our gel shift assays, we demonstrated binding of MZF-1-like proteins to the MZF-1a-and MZF-1b-binding elements in the CBS Ϫ1b promoter upstream region. Whereas mutagenesis of the MZF-1 sites resulted in ϳ15-59% losses of promoter activity in both HepG2 and HT1080 cells, only HT1080 cells showed CBS Ϫ1b transactivation in cotransfections with MZF-1 cDNA and CBS Ϫ1b promoter-reporter gene constructs, suggesting a cellspecific response.
Our finding that significantly decreased CBS transcripts and Ϫ1b promoter activity in HT1080 cells were accompanied by very low levels of transcription factor binding compared with HepG2 cells in gel shift assays was of particular interest since similar differences in CBS expression were previously seen in AML specimens, including those from children with DS (9). Furthermore, levels of CBS were implicated as a critical determinant of sensitivities to Ara-C for both AML and CBS-transfected acute lymphocytic leukemia cells (9,12). Since there were no differences in Sp factor levels between HT1080 and HepG2 cells, we performed mixing experiments to explore the possibility that these results might reflect a requirement for an unknown "cofactor" present in the HepG2 cells, but missing in HT1080 cells or, alternatively, the presence of an inhibitor in HT1080 cells. However, there were no effects on the levels of Sp factor binding upon mixing HT1080 and HepG2 nuclear extracts.
Sp1 is post-translationally modified by glycosylation and phosphorylation (27). The significance of Sp1 phosphorylation was suggested by reports that dephosphorylated Sp1 added to in vitro transcription extracts becomes rapidly phosphorylated in a manner that correlates with function (27). However, in other studies, phosphorylated Sp1 bound DNA with reduced affinity (28,29). A similar regulation would presumably occur for Sp3 (42). Our experiments with protein kinase A and calf intestinal alkaline phosphatase treatments of nuclear extracts strongly suggest that the differential binding of Sp1/Sp3 to the CBS Ϫ1b promoter in HT1080 and HepG2 cells is at least partially due to the different phosphorylation status of these factors in the two cell lines. The lack of a complete response upon protein kinase A treatment in our experiments may reflect phosphorylation effects independent of protein kinase A, a role for Sp factor glycosylation (43), or separate mechanisms altogether. Experiments are underway to explore these possibilities.
Studies of CBS gene expression should eventually provide insights into the bases for variations in CBS levels between DS and non-DS myeloblasts and for the remarkable responsiveness of DS/AML patients treated with chemotherapy including Ara-C. Likewise, they could potentially clarify the link between chromosome 21 and the increased risk of leukemia in DS children or the relationships between CBS and certain phenotypic features of DS. Finally, our findings should promote a better understanding of the molecular bases for variations in CBS gene expression in assorted human diseases, including homocystinuria, hyperhomocystinemia, and atherosclerosis.