An Sp1-binding Silencer Element Is a Critical Negative Regulator of the Megakaryocyte-specific αIIb Gene*

The Sp1 family of transcription factors are often involved in the regulated expression of TATA-less genes, frequently enhancing gene transcription. In this paper, we demonstrate that an Sp1-binding element inhibits the expression of the megakaryocyte-specific αIIb gene in all cell lines tested and that this inhibition is actively overcome only in megakaryocyte-like cell lines. We had noted previously in primary megakaryocytes that a 50-base pair (bp) deletion from −150 to −101 bp in the rat αIIb promoter region resulted in increased expression. We now show that deletion of this region markedly increased expression in both megakaryocytic and non-megakaryocytic cell lines, eliminating the tissue specificity of the αIIb promoter. Electrophoretic mobility shift assays (EMSA) defined a single complex, which bound to a −145 to −125 bp subregion. Point mutations within this region, localized the critical point of binding around bases −136/−135, and expression studies showed that introduction of the −136/−135 mutation into the rat αIIb promoter had a comparable result to that seen with the 50-bp deletion. EMSA studies with the homologous human αIIb promoter region gave an identical migrating band. Southwestern blots of HeLa nuclear proteins with both the rat −145 to −125 DNA and its human homologue bound to a single ∼110-kDa protein, the known molecular weight of Sp1. Confirmation that this region of the αIIb gene promoter bound Sp1 was accomplished using EMSA studies with an Sp1 consensus probe, anti-Sp1 and -Sp3 antibodies, and recombinant Sp1 protein. Further support for the role of Sp1 in the silencing of the αIIb promoter was obtained using a Gal4 binding site substitution for the silencer region of αIIb and co-expression of near full-length Sp1/Gal4 fusion protein expression vectors. Ectopic reinsertion of the −150 to −101 bp region, back into the −150 to −101 bp deleted promoter, enhanced rather than decreased expression, suggesting that Sp1’s inhibitory role at −136/−135 depends on its local interactions. In summary, we believe that we have identified a cross-species, non-consensus Sp1-binding site that binds Sp1 and that acts as a silencer of αIIb expression in many cell lines. A model is presented as to how this Sp1-binding silencer element contributes to the megakaryocyte-specific expression of αIIb gene.

The Sp1 family of transcription factors are often involved in the regulated expression of TATA-less genes, frequently enhancing gene transcription. In this paper, we demonstrate that an Sp1-binding element inhibits the expression of the megakaryocyte-specific ␣ IIb gene in all cell lines tested and that this inhibition is actively overcome only in megakaryocyte-like cell lines. We had noted previously in primary megakaryocytes that a 50base pair (bp) deletion from ؊150 to ؊101 bp in the rat ␣ IIb promoter region resulted in increased expression. We now show that deletion of this region markedly increased expression in both megakaryocytic and nonmegakaryocytic cell lines, eliminating the tissue specificity of the ␣ IIb promoter. Electrophoretic mobility shift assays (EMSA) defined a single complex, which bound to a ؊145 to ؊125 bp subregion. Point mutations within this region, localized the critical point of binding around bases ؊136/؊135, and expression studies showed that introduction of the ؊136/؊135 mutation into the rat ␣ IIb promoter had a comparable result to that seen with the 50-bp deletion. EMSA studies with the homologous human ␣ IIb promoter region gave an identical migrating band. Southwestern blots of HeLa nuclear proteins with both the rat ؊145 to ؊125 DNA and its human homologue bound to a single ϳ110-kDa protein, the known molecular weight of Sp1. Confirmation that this region of the ␣ IIb gene promoter bound Sp1 was accomplished using EMSA studies with an Sp1 consensus probe, anti-Sp1 and -Sp3 antibodies, and recombinant Sp1 protein.
Further support for the role of Sp1 in the silencing of the ␣ IIb promoter was obtained using a Gal4 binding site substitution for the silencer region of ␣ IIb and co-expression of near full-length Sp1/Gal4 fusion protein expression vectors. Ectopic reinsertion of the ؊150 to ؊101 bp region, back into the ؊150 to ؊101 bp deleted promoter, enhanced rather than decreased expression, suggesting that Sp1's inhibitory role at ؊136/؊135 depends on its local interactions. In summary, we believe that we have identified a cross-species, non-consensus Sp1-binding site that binds Sp1 and that acts as a silencer of ␣ IIb expression in many cell lines. A model is presented as to how this Sp1-binding silencer element contributes to the megakaryocyte-specific expression of ␣ IIb gene.
Platelets have a central role in thrombus formation. These anuclear cytoplasmic fragments are derived from bone marrow megakaryocytes and are highly differentiated (1). One of the specialized features found on the platelet membrane is the ␣ IIb /␤ 3 (also known as glycoprotein IIb/IIIa or CD41b) integrin receptor (2). This receptor is densely packed on the platelet surface. Following platelet activation, this receptor binds fibrinogen and plays an important role in platelet aggregation (2). Normally, ␣ IIb /␤ 3 is only found on developing megakaryocytes and circulating platelets. This is due to the tissue-specific nature of the ␣ IIb subunit. While ␤ 3 is expressed in several different cell types (3), the ␣ IIb gene is normally limited in its expression to megakaryocytes (4). We have shown previously that the ␣ IIb gene is a TATA-less gene comprising 30 exons, extending over an 18-kilobase pair region of the long arm of chromosome 17 (5,6).
The ␣ IIb gene is presently the best studied megakaryocytespecific gene. Better understanding of the regulation of ␣ IIb transcription in a lineage-restricted fashion allows us to learn more about the molecular mechanisms controlling hematopoietic differentiation. It may also allows us to develop new approaches for regulating gene expression in developing megakaryocytes and modulating platelet thrombogenic tendency. Other investigators defined four important elements in the 5Ј-flanking region of the human ␣ IIb gene that promote tissue-specific expression: two pairs of GATA-and Ets-binding sites, one proximal to the transcriptional start site and one more distal (7)(8)(9)(10). Deletion or mutation of any of these GATAor Ets-binding sites had a 2-3-fold effect on the level of reporter gene expression in transient expression studies in megakaryocytic cell lines. Studying the rat ␣ IIb promoter in a primary rat marrow system, we found the same four regulatory regions, but demonstrated a significant quantitative difference in promoter activity; the distal GATA-binding site at Ϫ454 bp 1 (GATA 454 ) had a 50-fold effect on expression and was essential for observing any promoter activity (11). We have since shown that these quantitative differences in promoter strength were not speciesspecific differences in the promoter region, but due to differences in the cell systems studied. When the same rat ␣ IIb promoter constructs were restudied in a megakaryocytic cell line (HEL), the results seen were indistinguishable from the human data (12).
In addition, we used the same rat primary marrow transient expression system to define a GA-rich Sp1-binding site at Ϫ14 bp (Sp1 14 ) (13). We showed that the complex bound to this site interacted with the complex bound at the proximal Ets-binding site at Ϫ35 bp (Ets 35 ). The Ets 35 site appears to tether Sp1 to its binding site, and we proposed that this tethered Sp1 is important in mooring the transcriptional initiation complex to the TATA-less ␣ IIb gene.
While carrying out these studies, we noted that, when a series of 50-bp deletions were made in a 912-bp ␣ IIb 5Ј-flanking region reporter construct, leaving the GATA 454 intact and beginning the deletions at Ϫ450 bp, all of the constructs had decreased activity compared with the wild-type construct, except for one construct that expressed twice as well as the wild type. We proposed that either this region between Ϫ150 bp and Ϫ101 bp contained a silencer element or the increase in expression was due to architectural disruption of the promoter region by the deletion (11). Since then, two other groups have defined silencer elements near this region in the human ␣ IIb promoter region (14,15). The common element between these two studies suggests that there is a region at Ϫ120 to Ϫ116 bp in the human ␣ IIb 5Ј-flanking region 5Ј-ATGAG-3Ј (corresponding to the rat Ϫ113 to Ϫ109 bp region) that binds a silencer complex. In this paper, we demonstrate a different site ϳ30 bp further upstream as being important in silencing ␣ IIb expression. We show that an Sp1-binding site that is conserved across species is the center of this silencer element and that this site appears to bind Sp1 (16). Mutation of this site leads to high levels of expression in both megakaryocytic and non-megakaryocytic cell lines, and therefore eliminates the tissue specificity of the ␣ IIb promoter. This site does not appear to bind to Sp3, a known negative regulator in the Sp1 family (17)(18)(19). Thus, it appears that the silencer domain of the ␣ IIb gene involves an increasingly recognized role of Sp1 as a negative regulator. We show that ectopic reintroduction of the silencer domain into the silencer-deleted ␣ IIb promoter enhanced rather than decreased expression, suggesting that Sp1 silencing effect may depend on local interactions with other bound nuclear proteins. A model is presented as to how the silencer element may function in ␣ IIb expression.
Plasmid Construction-Both 912 and 453 base pairs of rat ␣ IIb 5Јflanking region were PCR-amplified from a 2.2-kilobase pair SstI fragment of the rat 5Ј-␣ IIb gene that was isolated from a partial Sau3A rat genomic library (5) and subcloned into single-stranded M13mp18 (11). The sense primers for PCR were designed according to published sequence (5) with a BamHI site flanking their 5Ј ends. The antisense primer 5Ј-AAGCTTCTTCCTTCTCCCCAAATGT-3Ј, includes the untranslated region of rat ␣ IIb gene and a HindIII site (underlined). PCR products were cut with both BamHI and HindIII, and subcloned into BglII/HindIII-digested promoter-less luciferase reporter vector pGL3basic (Promega Corp., Madison, WI). Ϫ150 to Ϫ101 bp deletion, and CC 3 AA substitutions at Ϫ140/Ϫ141 bp, Ϫ135/Ϫ136 bp, and Ϫ130/Ϫ131 bp, were then created by overlapping PCR (11) using 912bp-pGL3 or 453bp-pGL3 construct as templates. All the PCR-based constructs were sequenced to exclude any PCR-induced mutations.
A single Gal4 binding site 5ЈGGGAGTACTGCCTCCGA-3Ј (21) was substituted for the rat ␣ IIb promoter sequence between Ϫ145 bp and Ϫ125 bp in the 453bp-pGL3 construct using a similar overlapping PCR strategy. The Gal4/Sp1 expression vectors pSG424 (no Sp1 fusion), pSG-Sp1WT (containing Sp1 amino acid residues 83-778, which include domains A-D of Sp1), pSG-Sp1N (containing domains A-C of Sp1), pSG-Sp1A&B (containing Sp1 domain A and B), and pSG-Sp1A (containing Sp1 domain A) were generously provided by Robert Tjian (University of California, Berkeley, CA). All these vectors contain the Gal4 binding domain, amino acids 1-147 of the Gal4 protein, in addition to the various domains of the Sp1 transcription factor (22) and are driven by an SV40 promoter.
We have ectopically reinserted the Ϫ150 to Ϫ101 bp region back into the 453bp⌬-pGL3 (⌬ refers to Ϫ150 to Ϫ101 bp deletion) after digestion of the vector with SacI, whose restriction site is immediately upstream of the rat ␣ IIb promoter region. The complementary sense and antisense oligonucleotides covering the sequence of Ϫ150 to Ϫ101 bp region were synthesized by Integrated DNA Technologies (Coralville, IA) with overhanging SacI ends. Annealed double-stranded DNA was T4 ligated into the cut 453bp⌬-pGL3 vector. The orientation of the inserted 50-bp region and the copy number of this insert were determined by sequence analysis.
pGL3-basic vector was used as a negative control in transfection assays. The SV40-pGL3 vector (Promega Corp.) containing both SV40 promoter and enhancer sequence, was included as a positive control, since it results in strong expression of luciferase gene in many types of mammalian cells. The pCMV␤ vector (CLONTECH), which contains the ␤-galactosidase gene driven by the CMV promoter, was used as an internal standard for transfection efficiency.
Transfection and Reporter Gene Assays-For CHRF cell transfection, cells were seeded at 0.3 ϫ 10 6 /ml. After growing for 2 days, cells were collected and resuspended in electrophoresis buffer (30.8 mmol/liter NaCl, 120.7 mmol/liter KCl, 8.1 mmol/liter Na 2 HPO 4 , 1.46 mmol/liter KH 2 PO 4 , 5 mmol/liter MgCl 2 ) at a concentration of 30 ϫ 10 6 cells/ml. Thirty micrograms of assay plasmid DNA and 20 g of pCMV␤ were added to 0.5 ml of cells in 0.4-cm electroporation cuvettes. After a 15-min incubation on ice, these cells were electroporated by Cell-Porator (Life Technologies, Inc.) at 230 V and 800 millifarads. Cells were then allowed to recover on ice for 10 min and at room temperature for 15 min. After washing with complete growth medium, cells were resuspended in 3 ml of growth medium and grown in six-well tissue culture plates for 24 h before being assayed for luciferase activity. HEL, HeLa, NIH 3T3, and HL-60 cells were transfected using LipofectAMINE™ reagent from Life Technologies, Inc. Briefly, 2.5 g of assay plasmid DNA and 0.5 g of internal control DNA pCMV␤, were incubated with 10 l of LipofectAMINE™ reagent in Opti-MEM I reduced serum medium (Life Technologies, Inc.) for 30 min. The formed DNA-liposome complexes were added to either exponentially growing HEL and HL-60 cells with 1.5 ϫ 10 6 cells/sample or 60 -70% confluent HeLa and NIH 3T3 cells grown in six-well tissue culture plates. All cells were washed with and then suspended in Opti-MEM medium before the addition of DNA-liposome mixture. A 5-h incubation at 37°C in a CO 2 incubator followed. After being washed with the appropriate complete growth medium, cells were grown in 3 ml (HEL and HL-60) or 2 ml (HeLa and NIH 3T3) of the same medium in six-well tissue culture plates for 48 h before the reporter gene assays. For Gal4/Sp1 cotransfection assays in HeLa cells, 2 g of the 453Gal4-pGL3 vector (having the Gal4 binding site substituted into the Ϫ145 to Ϫ125 region of the 453-bp rat ␣ IIb promoter), 1 g of the various Gal4/Sp1 expression vectors, and 0.5 g of pCMV␤, were cotransfected to each well of cells using 12.5 l of LipofectAMINE™ reagent.
Cells in each sample were collected, washed twice with phosphatebuffered saline, and lysed in 100 l of lysis buffer (1% Triton X-100, 50 mM Tris, pH 7.6, 1 mM dithiothreitol). For the luciferase assay, 20 l of cell lysate were mixed with 350 l of Reaction Mix (100 mM Tris, pH 7.8, 15 mM MgSO 4 ) and 10 l of 100 mM ATP in a luminometer cuvette. The relative light units for each sample were immediately measured in a luminometer that automatically pumps into each cuvette 100 l of 1 mM D-luciferin (Analytical Luminescence Labs) solution. Ten seconds measuring time was used.
To account for variable transfection efficiency among different samples, the luciferase activity for each sample were normalized by its ␤-galactosidase counts. For ␤-galactosidase activity, the Galacto-Light Plus™ chemiluminescent reporter assay kit (TROPIX, Inc., Bedford, MA) was used. Briefly, 20 l of cell lysate, diluted in 200 l of Reaction Buffer Diluent, were incubated at 48°C for 1 h to inactivate endogenous ␤-galactosidase activity in cell extracts. The Galacton Plus™ substrate was diluted 100-fold with Reaction Buffer Diluent to make Reaction Buffer. 200 l of Reaction Buffer were added into 20 l of the diluted and heat-inactivated cell lysate in a luminometer cuvette. After incubation at room temperature for 1 h, 300 l of Accelerator were manually injected into each cuvette, and the light output was immediately counted by luminometer using a 5-s measuring time.
Electrophoretic Mobility Shift Assays (EMSA)-Nuclear extracts were prepared as described previously (11) from HeLa, CHRF, and K562 cells with or without PMA induction. The single-stranded oligonucleotides were synthesized by Integrated DNA Technologies (Coralville, IA), and the complementary sense and antisense strands were then annealed into double-stranded DNAs used for EMSA. The Sp1 consensus binding sequence was purchased from Promega Corp. These double-stranded DNAs were end-labeled by T 4 polynucleotide kinase and [␥-32 P]ATP. Then, 0.1-0.2 ng of probes (ϳ2 ϫ 10 5 cpm) were incubated with 5 g of nuclear extract on ice for 20 min in a 20 l binding reaction that contained 18 mM HEPES, pH 7.8, 40 mM KCl, 4 mM MgCl 2 , 0.5 mM dithiothreitol, 3 g of poly(dI-dC), 2 g of bovine serum albumin, and 10% (v/v) glycerol. For competition studies, prior to the addition of radioactive probes, unlabeled competitors were added to binding reaction and incubated with nuclear extract for 10 min on ice. Samples were then electrophoresed at 4°C on a 4% (v/v) polyacrylamide, non-denaturing gel in 0.5ϫ Tris-boric acid-electrophoresis (TBE) buffer. Rabbit anti-human Sp1, Sp3, and YY1 polyclonal antibodies were purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). For gel supershift assays, 0.5-2 l (1 g/l) of antibody were added to the binding reaction and incubated on ice for additional 15 min before loading. A total of 0.5 footprint units of recombinant human Sp1 protein (Promega Corp.) was used instead of nuclear extracts in EMSA. Southwestern Blotting Analysis-Thirty micrograms of HeLa nuclear extract was electrophoresed on a 6% sodium dodecyl sulfate (SDS)polyacrylamide gel with prestained protein molecular weight standards (14,300 -200,000 molecular range) from Life Technologies, Inc. The proteins were then electroblotted to nitrocellulose filters (Schleicher & Schuell BA85, 0.45 mm), and renatured on the filters by serial dilution from 6 M guanidine hydrochloride to ZЈ buffer (25 mM Hepes-KOH, pH 7.6, 12.5 mM MgCl 2 , 20% glycerol, 0.1% Nonidet P-40, 0.1 M KCl, 10 M ZnSO 4 , 1 mM dithiothreitol). The membrane was then incubated for 30 min in blocking solution, containing 3% nonfat dried milk in ZЈ buffer, and probed with 32 P-end-labeled double-stranded rat Ϫ145 to Ϫ125 bp sequence, its human homologue, and M2 mutant in binding buffer (ZЈ buffer containing 0.25% nonfat dried milk) for 1 h at room temperature. The filter was then washed three times with ZЈ buffer for a total time of 15 min and exposed to autoradiographic film. A parallel gel containing the same amount of HeLa nuclear extract was also stained with Coomassie Brilliant Blue.

Functional Studies of the Potential Silencer Element in Both
Megakaryocytic and Non-megakaryocytic Cell Lines-Previous studies have shown that Ϫ150 to Ϫ101 bp region upstream of the transcriptional start site of rat ␣ IIb gene, when deleted, caused a 2-fold increase in the reporter gene, human growth hormone expression in a rat primary marrow transient expression system (23). To confirm this finding in other expression systems and to examine the tissue-specific nature of this deletion on expression, we shuttled the 453-bp and 912-bp 5Јflanking region of the rat ␣ IIb gene into the pGL3-basic vector, which contains a luciferase reporter gene. The Ϫ150 to Ϫ101 bp region was then deleted from the wild-type constructs. The important positive regulator GATA 454 element is present in 912-bp constructs but not in 453-bp constructs. Transient transfection studies of all these constructs were carried out both in the ␣ IIb -expressing CHRF and HEL cells, and in the ␣ IIb -non-expressing cell lines, epithelioid HeLa, fibroblastic NIH 3T3, and myeloid HL60 cells.
In the megakaryocyte-like cell lines, CHRF and HEL cells (Fig. 1, top), we found that the 912-bp construct expressed well (1.8-and 4.5-fold, respectively, higher than the SV40-driven pGL3 control), while the 453-bp construct, lacking the GATA 454 element, expressed poorly (0.3-and 0.2-fold of the same control). The Ϫ150 to Ϫ101 bp deletion (⌬) in the 912-bp promoter construct led to a 2.0-and 1.7-fold further increase in expression, respectively, much as in the published rat primary marrow transient expression studies (11). Surprisingly, in the 453-bp construct, which lacks the GATA 454 site, the same deletion had a 5.1-and a 25.9-fold increase in expression, respectively, resulting in levels comparable to the 912-bp wild-type construct (1.4-and 4.9-fold of the positive control, respectively).
These data, in agreement with our published data using primary megakaryocytes, show that the GATA 454 site needs to be present to see expression in megakaryocyte-like cell lines. In addition, they show that the Ϫ150 to Ϫ101 bp deletion enhances expression in these cells irrespective of whether the GATA 454 site is intact.
In the three non-megakaryocytic cell lines tested, neither the 912-nor the 453-bp wild-type construct was expressed (Fig. 1,  bottom). However, the corresponding constructs with the Ϫ150 to Ϫ101 bp deletion resulted in levels comparable to the SV40-pGL3 control (varying from 0.2-to 2.6-fold in the three cell lines tested). It appears that the removal of Ϫ150 to Ϫ101 bp region eliminated the tissue specificity of ␣ IIb promoter and induced high level of expression in ␣ IIb -non-expressing cells of different origins. GATA 454 does not seem to be functionally important in these cells, since both 912-and 453-bp constructs behaved in an identical fashion in expression.
These data suggest a ubiquitous silencer element being present in the Ϫ150 to Ϫ101 bp region. The deletion of this region by itself led to significant, non-tissue-specific, promoter activity from the 5Ј-flanking region of the ␣ IIb gene. It seems that only with the addition of sequences further upstream than 453 bp of the 5Ј-flanking region of the ␣ IIb gene can the silencing effect of this region be overcome in a megakaryocyte-specific fashion. Our previous studies suggest that this additional element includes the GATA 454 site (11).
DNA-Protein Interaction in the Ϫ150 to Ϫ101 bp Region by EMSA-To examine if there are any nuclear proteins binding to the Ϫ150 to Ϫ101 bp region, EMSA was carried out using nuclear extracts prepared from CHRF, K562 (another megakaryocyte-like cell line that may represent a less differentiated megakaryocyte-stage; Ref. 14) with or without PMA induction, and HeLa cells. The results were very similar with all three cell lines. Because the silencer effect was seen with non-megakaryocytic as well as megakaryocytic cell lines, the data presented below focus on the HeLa mobility shift studies. As seen in Fig. 2A, three initial double-stranded DNA probes were made, two overlapping the Ϫ150 to Ϫ101 bp region, and one corresponding to the human silencer region (14,15) spanning the rat sequence from Ϫ114 to Ϫ88 bp (5). The rat Ϫ114 to Ϫ88 bp region is the homologue of human Ϫ124 to Ϫ98 bp sequence, used by Fong et al. (14) as a probe in EMSA. The rat Ϫ114 to Ϫ88 bp double-stranded DNA did not complex with nuclear extracts from all three cell lines tested (data not shown). The other two double-stranded DNAs were from Ϫ150 to Ϫ116 bp and from Ϫ135 to Ϫ101 bp, and overlap by 15 bp.
Only the Ϫ150 to Ϫ116 bp DNA formed a complex (Fig. 2B), and this complex could be cold competed away by non-radiolabeled Ϫ150 to Ϫ101 DNA (lanes 3 and 4). The complex could not be competed away by the Ϫ135 to Ϫ101 bp region (lane 5) or by an unrelated Ets sequence (lane 6), indicating specific binding to the Ϫ150 to Ϫ116 bp region. The Ϫ135 to Ϫ101 bp double-stranded probe, containing the homologous human silencer region, failed to give rise to any significant complex formation (Fig. 2C). For the Ϫ135 to Ϫ101 bp double-stranded probe, varying salt conditions (from 20 to 100 mM), Mg 2ϩ concentration (from 0 to 10 mM, Fig. 2C), amount of nonspecific competitor (from 1 to 3 g of poly(dI-dC) per reaction), and buffering capacity (from 5 mM to 25 mM HEPES, pH 7.8) did not lead to a detectable band.
To further localize the binding site of the complex, we tested a double-stranded DNA probe that spanned from Ϫ150 to Ϫ126 bp ( Fig. 2A). This region formed a complex similar to that for the Ϫ150 to Ϫ116 bp probe, and these two probes can cross-FIG. 2. Initial EMSA of HeLa nuclear extracts with two overlapping probes for the ؊150 to ؊101 bp region. A, a comparative analysis of the ␣ IIb promoter sequences in this region between the available rat, human, and mouse sequences (5,67) is shown at the top with numbers based on the rat sequence. The human silencer core element (15) is shown as a dark gray area between rat Ϫ113 and Ϫ109 bp. The three introduced CC 3 AA mutations in the rat sequence are also shown at the top as gray ovals. The subregions used as probes in this and other EMSA figures are indicated at the bottom. B, EMSA studies with the Ϫ150 to Ϫ116 probe. The arrow indicates the band of interest. Cold competition was done at both a 100-fold and a 500-fold molar excess for the Ϫ150 to Ϫ116 DNA and a 500-fold for the other cold competitors. C, EMSA studies of the Ϫ135 to Ϫ101 region with varying magnesium concentrations indicated. compete (data not shown). In contrast, a probe from Ϫ140 to Ϫ116 bp (Fig. AA) did not form a complex, nor could it compete away the band seen with the Ϫ150 to Ϫ126 bp probe (data not shown). We then further shortened the probe to Ϫ145 to Ϫ125 bp ( Fig. 2A). We found that the Ϫ145 to Ϫ125 bp probe formed a similar mobility complex as the Ϫ150 to Ϫ116 bp and Ϫ150 to Ϫ126 bp probes (Fig. 3, lanes 2, 8, and 10, respectively) and that it can be cold competed away by itself and by the other two DNAs (Fig. 3, lanes 3-6).
Therefore, it appears by EMSA that we can identify a single complex in nuclear extracts from a number of different cell lines that bind specifically to the Ϫ150 to Ϫ101 bp region of interest. We further localized this DNA binding to a subregion between Ϫ145 and Ϫ125 bp upstream of the rat ␣ IIb transcriptional start site.
Further Characterization of the Protein Binding Sequence in the Ϫ145 to Ϫ125 bp Region-To identify the base pairs that are important for protein binding in the Ϫ145 to Ϫ125 bp region, three CC 3 AA substitutions were made at Ϫ140/Ϫ139 bp (M1), Ϫ136/Ϫ135 bp (M2), and Ϫ131/Ϫ130 bp (M3) as shown in Fig. 2A. These mutated double-stranded DNAs were used in EMSA as cold competitors against wild-type Ϫ145 to Ϫ125 bp binding (Fig. 4, lanes 4 -6) as well as probes to assay their own binding activity (Fig. 4, lanes 7-12). It is of interest that these findings are consistent with the degree of sequence conservation at these three sites ( Fig. 2A); M2 is conserved among all three species, while M1 is a CT rather than a CC sequence in the human ␣ IIb promoter region. M3, while conserved between rat and human, is deleted in the mouse promoter region. Thus, the site that is most highly conserved across species appears to be the crucial point of complex binding. Expression Studies with the Three CC 3 AA Mutations-We then asked whether the EMSA findings would be reflected in expression studies. We introduced the three CC 3 AA mutations into the wild-type rat 453bp-pGL3 construct, and tested these new constructs in HeLa, NIH 3T3, and HL60 cells. All three cell lines gave similar results (Fig. 4B). In agreement with the EMSA data, M1 had a modest increase in expression (3-16% of the 453⌬ construct), M2 had a significant increase (40 -80% of the 453⌬ construct), and M3 had the same low level of expression as the wild-type 453 construct (0.12-1.2% of the 453⌬ construct). Thus, elimination of binding of a single complex to this region appears to correlate with lost of silencing of the ␣ IIb promoter.
EMSA with the Human Homologue of the Rat Ϫ145 to Ϫ125 bp Region-To examine whether this site was species-specific or more universally applicable, we then tested whether the human homologue to this region, which has 7/21 nucleotide substitutions, would also bind to the same complex in EMSA studies. As can be seen in Fig. 5A (lanes 2-6), this human homologue formed a similar size complex that can be effectively competed away by itself and by the rat Ϫ145 to Ϫ125 bp double-stranded DNA. Conversely, the human homologue was equally effective at competing away binding to the rat Ϫ145 to Ϫ125 bp probe (Fig. 5A, lanes 8 -12). M1, M2, and M3 cold competition studies (Fig. 5B, lanes 7-12) were identical to those shown in Fig. 4A for the rat wild-type probe. These findings suggest that the same complex formed with both the human and rat sequence in this region.
Sp1-related Protein(s) Are Involved in the Binding of Ϫ145 to Ϫ125 Region-In an attempt to identify the silencer-binding protein, we searched the Transfac data base by Transcription Element Search Software (TESS), 2 looking for known transcription factors that have consensus binding sequence homologous to our silencer region around bases Ϫ136/Ϫ135. The search suggested MAZ (Myc-associated zinc-finger protein) (24 -26), Yi (27), and NF-1 (28) as potential candidates. Mela1 (24 -26) and YiMT3 (27), known MAZ-and Yi-binding elements, respectively, and the NF-1 consensus binding sequence (Santa Cruz Biotechnology, Inc.), were used as cold competitors in EMSA for the binding to the Ϫ145 to Ϫ125 bp region using HeLa nuclear extract. While YiMT3 and NF-1 consensus did not show any effect on the binding, Mela1 did cold compete to the same extent as unlabeled wild-type Ϫ145 to Ϫ125 bp DNA, suggesting that a MAZ-related protein might be involved (data not shown).
To extend our characterization of the complex, a Southwestern blotting analysis was performed using HeLa nuclear extract to determine the molecular weight(s) of the proteins that bind to the Ϫ145 to Ϫ125 bp region. A single band at ϳ110 kDa (Fig. 6, lanes 3 and 4, respectively) was recognized by both the rat Ϫ145 to Ϫ125 bp probe and its human homologue, while no protein appears to bind to the M2 mutant probe (Fig. 6, lane 5). MAZ is a 58-kDa protein, so that Southwestern blotting did not appear to support a role for this protein in binding to the silencer region. Nevertheless, when ME1a1 was used as a probe, it bound to several bands including an ϳ58-kDa band 2 TESS software is available via the World Wide Web (http://www.cbil.upenn.edu//cbil-home/index.html).
FIG. 3. EMSA of HeLa nuclear extracts, further localizing the region of interest to the ؊145 to ؊125 region upstream of the rat ␣ IIb transcriptional start site. EMSA studies with the Ϫ145 to Ϫ125 probe and similar studies with two longer probes. The arrow indicates the band of interest. Cold competition was done at both a 50-fold and a 200-fold molar excess for the Ϫ145 to Ϫ125 probe and at a 200-fold molar excess for the other DNAs. In this figure and other EMSA figures in this paper, inconsistent bands of faster mobility were seen in lanes that contain nuclear extract. Specific cold competitors, while readily competing away the band of interest, did not consistently compete away these bands. We believe that these bands represent nonspecific binding. and an ϳ110-kDa band (Fig. 6, lane 2). It is known that ME1a1 binds Sp1 protein in addition to MAZ (29,30), and that Sp1 has molecular mass of ϳ105 kDa (31,32).
We, therefore, examined whether Sp1 was involved in the complex binding to the silencer element. In Fig. 7A, supershift studies were performed to see if an anti-Sp1 antibody could recognize the complex bound to the rat Ϫ145 to Ϫ125 bp probe. The addition of rabbit anti-human polyclonal Sp1 antibody (Santa Cruz Biotechnology, Inc.) supershifted the protein complex (lane 3). A control rabbit anti-YY1 antibody (Santa Cruz Biotechnology, Inc.) did not cause any shift in the band (lane 4).
(Not shown is that on another EMSA gel, using a consensus YY1 DNA probe, the YY1 antibody resulted in a supershift band, while the rat Ϫ145 to Ϫ125 bp complex was not supershifted.) In Fig. 7A (lanes 13 and 14), the human homologue probe was used. The Sp1 antibody again formed a supershifted band.
Also shown in Fig. 7A, an Sp1 consensus binding sequence that contains a classic GC box Sp1 binding site from the SV40 promoter is a strong cold competitor for the complex binding to the rat Ϫ145 to Ϫ125 bp probe (lane 6). When this Sp1 consensus sequence was labeled, it formed a similar protein complex with HeLa nuclear extract as seen with the rat Ϫ145 to Ϫ125 probe (lanes 8 and 2, respectively). Sp1 antibody also super-shifted the consensus complex (lane 9), while the control YY1 antibody had no effect (lane 10). Although not a strong cold competitor in comparison with Sp1 consensus sequence, the rat Ϫ145 to Ϫ125 did compete with Sp1 binding (lanes 11 and 12). Not shown is that the M2 substituted rat probe did not compete at all against the Sp1 consensus probe. Fig. 7B shows that recombinant human Sp1 protein forms a complex with the rat Ϫ145 to Ϫ125 bp probe with the same mobility as seen with the HeLa nuclear extract (lane 10 versus lane 2). As predicted, M2 does not form a complex with the recombinant Sp1 protein (lanes 11 and 12).
We also tested whether the Ϫ145 to Ϫ125 bp region can bind Sp3, an Sp1 family member with the same binding sequence as Sp1, but which has been shown to act as a negative regulator of other genes (33)(34)(35). We used varying amounts of anti-Sp1 and anti-Sp3 specific antibodies. As can be seen in Fig. 7B (lanes  3-5), increasing the amount of anti-Sp1 antibody from 0.5 to 2 g per sample virtually eliminated the initial band. In lanes 6 -8, increasing the amount of anti-Sp3 had no effect on the intensity of the original band and did not supershift the band. In lane 9, the addition of both 1 g anti-Sp1 and 1 g of anti-Sp3 was no better than the result when 1 g of anti-Sp1 alone was used in lane 4.
Therefore, it appears that the complex that binds to both the rat Ϫ145 to Ϫ125 bp region and its human homologue contains Sp1. Furthermore, the M2 mutation that eliminated complex formation in this region and that released the ␣ IIb promoter from silencing also prevented Sp1 binding to this region. Fig. 7B, lane 12 Sp3 is less likely to be involved in the Ϫ145/Ϫ125 binding.

Further Confirmation of the Role of Sp1 in the Silencing of ␣ IIb Gene Expression by Gal4/Sp1 Cotransfection Studies-To
further confirm that Sp1 is important in ␣ IIb silencing, we substituted a Gal4 binding sequence from nucleotides Ϫ145 to Ϫ125 in the 453-bp ␣ IIb promoter region. As expected, this substitution in the silencer region (453Gal4-pGL3) markedly increased expression of the 453bp-pGL3, leading to levels of expression comparable to the SV40-driven positive control (40% of SV40-pGL3) and nearly twice as high as 453M2-pGL3 (Fig. 8A). More importantly, when this construct is coexpressed with a series of Sp1/Gal4 fusion proteins, containing the DNA binding domain of the yeast transcription factor Gal4 (21), silencing activity is seen with the two longest fusion proteins, pSG-Sp1WT and pSG-Sp1N (see "Materials and Methods") (Fig. 8B). While pSG-Sp1A and pSG-Sp1A&B increased the expression of 453Gal4-pGL3 2.3-and 1.7-fold, respectively, pSG-Sp1N and pSG-Sp1WT both decreased its expression to 50% and 60%, respectively, compared with the pSG424 control (Gal4 only vector). In addition to providing further support for our findings that Sp1 is involved in silencing, these data suggest that near full-length Sp1 is necessary to achieve silencing of the ␣ promoter and that other constructs that may not be able to interact in a specific fashion with regulatory domains in this promoter actually lead to a further increase in expression.
Ectopic Reinsertion of the Ϫ150 to Ϫ101 bp Region Immediately Upstream of the 453bp⌬-pGL3 Construct and Its Effect on Expression in HeLa Cells-To examine whether or not the Sp1-binding silencer element functions in a position-and orientation-independent fashion, we deleted the Ϫ150 to Ϫ101 bp region from 453 bp of rat ␣ IIb promoter and reintroduced this 50-bp region back into the same promoter, but upstream from the 453-bp region. A single copy was inserted in two different constructs, one in forward and one in reverse orientation. These constructs were transfected into HeLa cells to examine the relocation effect on expression. As we have shown above, the 453⌬-pGL3 construct gave rise to a significant level of expression. However, the reintroduction of this 50-bp region into a different position in either forward or reverse orientation did not suppress expression. Instead, these constructs resulted in a 1.7 Ϯ 0.23-fold and 2.4 Ϯ 0.33-fold further increase in expression, respectively, when compared with the 453⌬-pGL3 construct. Therefore, it appears that the inhibitory activity of this silencer element is dependent upon its physical location in the ␣ IIb promoter. DISCUSSION We have found that the previously described effect of deleting the Ϫ150 to Ϫ101 region upstream of the transcriptional start site of the rat ␣ IIb gene (11) was not due to a physical disruption of the promoter region, but rather due to the lost of a cross-species conserved, Sp1-binding site. This site centers on nucleotides Ϫ136/Ϫ135, and we have named it the Sp1 135 site. Complex binding to this site silences ␣ IIb promoter-driven expression in both megakaryocytic and non-megakaryocytic cell lines.
Our data also suggest that this silencer site binds Sp1, a member of a multigene family of zinc finger transcription factors (36,37). Sp1 is ubiquitously expressed in all tissues, and its ubiquity is consistent with our finding that Sp1 135 complex suppresses ␣ IIb expression in all tested cell lines. Sp1 binds to GC boxes and similar motifs (38 -41), but also binds to a number of non-consensus sequences as well (42)(43)(44). The Ϫ145 to Ϫ125 bp region is a CT-rich region that is consistent with such non-consensus Sp1-binding sites.
Other Sp1 family members include Sp2, Sp3, and Sp4 (36, 37). Sp2 binds a GT box in the promoter of a T-cell receptor gene and seems to have divergent nucleotide recognition se- quence (36), while Sp3 (36,37) and Sp4 (37) have binding specificity and affinity similar to those of Sp1.
Sp3 also appears to be ubiquitously expressed, while Sp4 appears to be limited in vivo to neural tissues (36). The zinc finger DNA-binding domains and glutamine-and serine/threonine-rich activation domains are highly conserved between Sp1, Sp3, and Sp4 (17,(45)(46)(47). However, while Sp1 and Sp4 appear to mostly promote transcription, Sp3 has often been shown to be a negative regulator of expression (17,33,47). The DNA-binding domains of Sp1 and Sp3 appear to be functionally interchangeable, but the activation domain of Sp3 is not functional when chimerically linked to the Sp1 DNA binding domain (17). This suggests that the negative regulatory activity of Sp3 may be due to its competition with Sp1 for a common binding site. Indeed, Sp3 suppression of Sp1-mediated transcriptional activation has been described in many genes, including both basal and Tat-activated expression of the human immunodeficiency virus promoter (33). Differential expression of Sp1 and Sp3 in different tissues and altered Sp1/Sp3 ratio during cell differentiation and transformation have been shown to be responsible for the regulation of several epithelial-specific promoters (44,48). Therefore, Sp3 binding would have provided an explanation for the involvement of an Sp1-binding element in negative regulation of gene transcription.
We have examined whether or not Sp3 is involved in Sp1 135 binding by EMSA studies using an anti-Sp3-specific antibody. Neither supershift nor blocking was observed. Furthermore, the anti-Sp1-specific antibody can supershift virtually all of the complex (Fig. 7B, lane 5), suggesting that the bound protein is Sp1. However, it is still possible that in the intact cell, the Sp1 135 site, in the context of the entire proximal ␣ IIb promoter region, binds Sp3. Perhaps the silencer elements proposed by others just upstream or downstream of this site (14,15), interact with the Sp1 135 site to allow it to bind specifically to Sp3 and inhibit transcription.
Several recent reports (49 -51) also suggest that a negative regulatory protein, whose DNA binding site overlaps a Sp1 site, may competitively interfere with Sp1 binding and inhibit transcription. Since our mobility gel shift studies and Southwestern analysis only detected one single protein complex at the Sp1 135 site in all cell lines tested, it is less likely that multiple proteins interact with this site. Activation by Sp1 can also be repressed FIG. 7. EMSA studies with antibodies to Sp1 and recombinant Sp1 protein. A, labeled rat Ϫ145 to Ϫ125 bp, its human homologue, and an Sp1 consensus probe was used with and without the addition of an anti-Sp1 antibody (1 g/lane). The rat and Sp1 consensus probes were also studied using an anti-YY1 (66) (control) antibody (1 g/lane). Cold cross-competition studies between the rat probe and the Sp1 consensus probes were done at 200-fold molar excess. B, EMSA studies of the rat Ϫ145 to Ϫ125 bp wild-type and M2 probes. The wild-type probe was studied using anti-Sp1 antibody (lanes 3-5, 0.5 g, 1.0 g, and 2.0 g, respectively) and anti-Sp3 antibody (lanes 6 -8, 0.5 g, 1.0 g, and 2.0 g, respectively) with HeLa nuclear extracts. In lane 9, 1.0 g of anti-Sp1 and 1.0 g of anti-Sp3 antibody were used together. Both probes were also used with recombinant Sp1 protein (0.5 footprint units/lane) indicated by an asterisk in lanes 10 and 12. by the formation of inactive (non-DNA-binding) complexes between Sp1 and other nuclear proteins such as Sp1-I and p107 (52,53). However, interference with Sp1 activation, an established mechanism by which gene transcription can be altered, obviously cannot explain the activity of Sp1 135 site in ␣ IIb promoter, because the silencing function of this site correlates directly rather than inversely with its binding of Sp1-related proteins.
Therefore, our study raises the possibility that Sp1 may itself decrease transcription when bound to certain Sp1 elements. Several recent reports have suggested a similar negative regulatory role for bound Sp1 (53)(54)(55). How this occurs is unclear, but one mechanism may involve its interactions with other nuclear elements. Sp1 has been shown previously to interact with other nuclear factors such as GATA-1 and Ets proteins (57)(58)(59)(60). Sp1 is also described to be a tethering factor to recruit the transcription initiation complex to TATA-less promoters by physically interacting with components of general transcriptional machinery (61,62).
Interestingly, we have shown previously that the Sp1 14 site is a positive regulator of ␣ IIb expression (13). The complex at this site appears to interact with the complex bound to the Ets 35 site. It is suggested that the Sp1 14 complex promotes transcription by tethering the transcriptional initiation complex to this TATA-less promoter. Our present finding is that another Sp1-binding site is a negative regulator of ␣ IIb expression. Therefore, depending on which site Sp1 is bound to, this nuclear factor appears to have both a positive and negative role in the regulation of ␣ IIb expression. A similar dual effect of Sp1 has been described in the proximal promoter of human adenine nucleotide translocase 2 (ANT2) gene (56). For the TATA-containing ANT2 gene, the more proximal Sp1 site (from Ϫ7 bp to Ϫ2 bp) partially inhibits transcription, probably by disrupting the recruitment and assembly of transcriptional initiation complex. A more distal site, containing two adjacent Sp1-binding sites at Ϫ87 bp to Ϫ58 bp, activates expression.
Given the fact that the effect of Sp1 on transcriptional activity is context-dependent, it is not surprising that only specific Gal4/Sp1 fusion protein shown in Fig. 8 inhibit expression, and that other fusion constructs actually increase expression. The context dependence of Sp1 binding in this region is also consistent with what was seen following ectopic reinsertion of the Ϫ150 to Ϫ101 bp region, which did not return silencer activity, but rather resulted in increased expression. These finding support the idea that the complex bound to the Sp1 135 site interacts with other bound nuclear factors in the ␣ IIb promoter region. Physical disruption of the structural organization between Sp1 and these other regulators, therefore, may alter the regulatory activity of Sp1 135 in ␣ IIb transcription.
Silencer elements have been proposed in the regulated expression of several other megakaryocyte-specific genes. A silencer element was identified for the rat platelet factor 4 (PF4) promoter (23). However, whether this region actually contains a silencer element was questioned in studies of the human PF4 promoter, which suggest that the homologous region is actually a strong promoter of PF4 expression (63). The ␣ 2 integrin gene is also TATA-less, and its regulated expression in megakaryocytic cell lines has also been suggested to involve a silencer region (64). In addition to a core promoter (Ϫ92 bp to the transcriptional start site) that is active both in megakaryocytic and non-megakaryocytic cells, a silencer element (Ϫ92 to Ϫ351 bp) region was defined that showed tissue specificity. It is inactive in non-hematopoietic cells but active in megakaryocytic cells. Located further upstream is a strong megakaryocyte-specific enhancer that overcomes the silencer effect and restores megakaryocyte-specific expression of the ␣ 2 gene. The tissue specificity of the ␣ 2 gene silencer contrasts with the Sp1 135 site, which appears to be an active silencer of ␣ IIb expression in all cell lines tested. However, the two genes share a common mechanism in that the silencer element is overcome by a distal megakaryocyte-specific enhancer, leading to tissuespecific expression.
Two studies of the human ␣ IIb gene have defined a silencer domain (14,15). The exact site(s) of the involved silencer element varied between the two studies. EMSA studies by Fang and Santoro (14) suggested that there may be two sites involved in the silencer effect seen in the human ␣ IIb promoter, Ϫ198 to Ϫ178 bp and Ϫ124 to Ϫ99 bp. Their studies focused on K562 cells and suggested that PMA induction markedly increased ␣ IIb expression and led to a inverse decrease in EMSA complex formation at both sites. Preliminary data that was not shown by the authors suggested that the protein that bound to the silencer element had a molecular mass of ϳ30 kDa. Prandini et al. (15) also detected a silencer element that was in part consistent with the above studies, suggesting that there were two silencer elements at Ϫ120 to Ϫ116 bp and at Ϫ102 to Ϫ93 bp, whose mutation increased promoter activity of the human promoter region 4-and 8-fold, respectively, and 20-fold in combination. No EMSA studies were done, but DNase I footprinting showed a protected region between Ϫ120 bp and Ϫ116 bp. Thus, these two studies suggest a common silencer site centered at human Ϫ120 to Ϫ116 bp with the sequence 5Ј-AT-FIG. 8. Sp1/Gal4 chimera expression studies demonstrating the role of Sp1 in the ␣ IIb silencer. Expression studies in HeLa cells using luciferasebased ␣ IIb promoters corrected for ␤-galactosidase expression. In A, the expression of the various ␣ IIb promoter vectors are shown relative to the SV40 promoterdriven pGL3 control. In B, the cells were co-transfected with a Gal4 or Gal4/Sp1 expression vector plus the 453Gal4-pGL3 vector. Relative expression of the various chimeric vectors were compared with the pSG424 control containing the Gal4 binding domain alone. Each condition was studied five separate times, with two repeats per study.
Our studies suggest another site as being critical for the observed silencing, ϳ30 bp upstream from the proposed human silencer site. Prandini et al. (15) suggested that the sequence of 5Ј-ATGAG-3Ј, which is found in the 5Ј-flanking region of several platelet-specific promoters is a common silencer element for all of these genes. However, this region is not conserved in the rat and mouse ␣ IIb promoter regions, being 5Ј-GTGTG-3Ј in the rat and 5Ј-G-ACG-3Ј in the mouse, thereby having 2/5 and 4/5 mismatches ( Fig. 2A). Indeed, the rat homologue (probe Ϫ114 to Ϫ88 in Fig. 2A) of the human ␣ IIb silencer region did not form a complex in EMSA studies with HeLa cell nuclear extract. We focused our analysis on the Ϫ145 to Ϫ125 bp region that did form a complex. It turned out that this region has a well localized site that binds Sp1 in both the rat and human sequences, and that the human homologue of this region is protected on the DNase I footprinting studies by Prandini et al. (15). It may be that silencing of the ␣ IIb gene is more complex than any of these studies suggest. Interactions between a number of sites may be necessary to achieve ␣ IIb gene silencing. Perturbations of any of these sites may then have the same effect of relieving the silencer effect.
In summary, we present two important new findings: 1) that there is an important Sp1 135 silencer domain involved in the regulation of ␣ IIb gene expression, and 2) that despite the presence of the Sp1 135 silencer domain in megakaryocytes, sequences further 5Ј to the Ϫ453 bp can overcome this silencer effect in the developing megakaryocytes. Our previous studies suggest that this upstream megakaryocyte-specific element involves the GATA 454 site. Thus, we present in Fig. 9 a simplified model of how the ␣ IIb gene is regulated in agreement with our data. In non-megakaryocytic tissues, a silencer complex forms around the Sp1 135 site. This complex interacts with the proximal promoter/initiation complex of this TATA-less gene to prevent ␣ IIb expression. In megakaryocytes, this silencer complex still interacts and inhibits expression, but a second more distal complex involving the GATA 454 site forms, and this complex can overcome the silencer effect. The fact that the silencer domain still interacts with the proximal promoter/initiation complex is based on the consistent doubling in expression seen in transient expression studies of both primary rat marrow cells (11) and megakaryocyte-like cell lines. Since GATA proteins are not uniquely expressed in megakaryocytes, the critical GATA 454 element is not likely to be solely responsible for the megakaryocyte-specific nature of the upstream enhancer complex. This complex may, in addition, involve the Ets ele-ment around Ϫ512 bp. It has been proposed by others (15,65) that, while GATA and Ets proteins are not individually megakaryocyte-specific, they may act in concert to confer tissue specificity. In fact, physically paired GATA-and Ets-binding sites have been found in most megakaryocyte-specific gene promoters so far characterized (11,(65)(66)(67).
One caveat with these studies on the Sp1 135 silencer domain and with studies by us and others on other regulatory elements 5Ј to the ␣ IIb gene is that they are all based on analysis of relatively short stretches of the 5Ј-flanking region of this gene. Whether or not a ubiquitous silencer element will be demonstrated in studies with the intact gene and whether the upstream enhancer elements maintains their role as a dominant, tissue-specific regulator remain to be tested. It is possible that a targeted mutation of the Sp1 135 site in transgenic mice may have a number of different outcomes, varying from having no effect on ␣ IIb expression to having an effect only in megakaryocytes to having a wider affect of expression on other hematological or non-hematological lineages. FIG. 9. Model of megakaryocytespecific expression of ␣ IIb . The model shows the GATA 454 complex and Sp1 135 complex and the proximal promoter/transcriptional initiation complex. We propose that in all tissues of the body, an Sp1based complex forms at the Sp1 135 site. The Sp1 135 complex, by interacting with the proximal promoter/transcriptional initiation complex, inhibits transcription. The GATA 454 complex is formed in a megakaryocyte-specific fashion. In these cells, this complex overcomes the Sp1 135 effect and allows transcription to occur. We also propose that, in megakaryocytes, the Sp1 135 site still interacts with the proximal promoter/transcriptional initiation complex, so that its removal leads to a further increase in expression.