Characterization of Inherited Differences in Transcription of the Human Integrin α2 Gene*

Inherited, single-base substitutions are found at only two positions, C− 52T and C− 92G, within the proximal 5′-regulatory region (within −1096 to +48) of the human integrin α2gene. We recently reported that the T− 52substitution results in decreased binding of transcription factor Sp1 to adjacent binding sites, decreased transcription of the α2 gene, and reduced densities of platelet α2β1. In this study, we identify an additional Sp1-binding site at position −107 to −99 and show that the adjacent dimorphic sequence C− 92G also influences the rate of gene transcription. In the erythroleukemia cell line Dami, transfected promoter-luciferase constructs bearing the G− 92 sequence exhibit roughly a 3-fold decrease in activity relative to the C− 92constructs. In transfected CHRF-288-11 megakaryocytic cells, the corresponding activity decreases by 5-fold. DNase I footprinting of the promoter region with Dami nuclear extracts showed a protected segment at −107 to −99 that can be deprotected by coincubation with molar excess of a consensus Sp1 oligonucleotide. Gel mobility shift assays and supershift assays with specific antibodies indicate that Sp1 binds to this region of the α2 gene promoter. Mutation of the Sp1 binding element within −107 to −99 in constructs containing either C− 92 or G− 92abolishes basal promoter activity and eliminates the binding of Sp1. The G− 92 sequence has a gene frequency of 0.15 in a typical Caucasian population, and the presence of this allele correlates with reduced densities of platelet α2β1. The combined substitution G− 92/T− 52 has an additive influence on gene transcription, resulting in an 8-fold decrease in transfected Dami cells or a 20-fold decrease in transfected CHRF-288-11 cells. In summary, the natural dimorphism C− 92G within the proximal 5′-regulatory region of the human integrin α2 gene contributes to the regulation of integrin α2β1 expression on megakaryocytes and blood platelets and must thereby modulate collagen-related platelet function in vivo.

Inherited, single-base substitutions are found at only two positions, C ؊52 T and C ؊92 G, within the proximal 5-regulatory region (within ؊1096 to ؉48) of the human integrin ␣ 2 gene. We recently reported that the T ؊52 substitution results in decreased binding of transcription factor Sp1 to adjacent binding sites, decreased transcription of the ␣ 2 gene, and reduced densities of platelet ␣ 2 ␤ 1 . In this study, we identify an additional Sp1-binding site at position ؊107 to ؊99 and show that the adjacent dimorphic sequence C ؊92 G also influences the rate of gene transcription. In the erythroleukemia cell line Dami, transfected promoter-luciferase constructs bearing the G ؊92 sequence exhibit roughly a 3-fold decrease in activity relative to the C ؊92 constructs. In transfected CHRF-288-11 megakaryocytic cells, the corresponding activity decreases by 5-fold. DNase I footprinting of the promoter region with Dami nuclear extracts showed a protected segment at ؊107 to ؊99 that can be deprotected by coincubation with molar excess of a consensus Sp1 oligonucleotide. Gel mobility shift assays and supershift assays with specific antibodies indicate that Sp1 binds to this region of the ␣ 2 gene promoter. Mutation of the Sp1 binding element within ؊107 to ؊99 in constructs containing either C ؊92 or G ؊92 abolishes basal promoter activity and eliminates the binding of Sp1. The G ؊92 sequence has a gene frequency of 0.15 in a typical Caucasian population, and the presence of this allele correlates with reduced densities of platelet ␣ 2 ␤ 1 . The combined substitution G ؊92 / T ؊52 has an additive influence on gene transcription, resulting in an 8-fold decrease in transfected Dami cells or a 20-fold decrease in transfected CHRF-288-11 cells. In summary, the natural dimorphism C ؊92 G within the proximal 5-regulatory region of the human integrin ␣ 2 gene contributes to the regulation of integrin ␣ 2 ␤ 1 expression on megakaryocytes and blood platelets and must thereby modulate collagen-related platelet function in vivo.
The integrin ␣ 2 ␤ 1 is a receptor for both laminin and collagen on most cell types but binds exclusively to collagen when expressed on megakaryocytes and blood platelets (1). A single copy of the ␣ 2 gene is present in the haploid genome, located on chromosome 5 (5q23-31) (2).
The expression of ␣ 2 is known to be regulated at the transcriptional level in megakaryocytes (8 -10), epithelial cells (11,12), and fibroblasts (13,14). Fig. 1 summarizes the major features of the proximal 5Ј-regulatory region of the human ␣ 2 gene. Zutter et al. (11) have defined a "core" promoter region within positions Ϫ92 to Ϫ30 relative to the transcription start site (ϩ1), and subsequent studies have determined that the proximal 961 bp 1 of the 5Ј-flanking region of ␣ 2 isolated from the K562 cell line direct cell type-specific suppressor and enhancer activity in cells of epithelial origin (8,11). Within the core promoter, two consensus Sp1/Sp3 binding elements are located on either side of the C Ϫ52 T dimorphism previously described by us (15). A second dimorphism, C Ϫ92 G, lies immediately downstream from the nonanucleotide sequence -GGGGCGGGG-with the general properties of another Sp1 and/or Sp3 binding element. Aside from this proximal 5Ј-regulatory region, additional distal sequences are thought to contain enhancer elements necessary for maximal transcription of the ␣ 2 gene in cells of megakaryocytic lineage (8,9). Partial sequences of the human ␣ 2 gene have been reported (3), and a complete reference gene sequence can be assembled from overlapping genomic clones deposited at GenBank TM (accession numbers AC008773, AC008966, AC016619, AF035968, AF062039, AF113511, NM002203, L24121, and U31518).
We recently provided evidence that allelic differences in receptor density, initially observed on blood platelets, correlate with the C Ϫ52 T dimorphism within the 5Ј-regulatory region of the ␣ 2 gene (15). This dimorphism influences transcription by altering the affinity of the flanking sequences for the transcription factors Sp1 and Sp3. Furthermore, the sequence T Ϫ52 , which attenuates Sp1/Sp3 binding and transcriptional activity, is in linkage disequilibrium with one of the alleles (A3) known to be associated with diminished expression of the integrin on platelets.
In the current study, we provide additional characterization of the 5Ј-regulatory region of this gene. First, we provide direct evidence that the nonanucleotide -GGGGCGGGG-within positions Ϫ107 and Ϫ99 is indeed a Sp1-binding site. In addition, we show that the second dimorphism, C Ϫ92 G, immediately downstream from this site also modulates gene transcription and acts synergistically with C Ϫ52 T to vary transcription rates by more than an order of magnitude in megakaryocytic cell lines, such as Dami or CHRF-288-11. These two sequence dimorphisms represent the only naturally occurring differences within the proximal 5Ј-regulatory region of the gene (position Ϫ1096 to ϩ48). These inherited sequence differences have a profound influence on expression levels of this gene in megakaryocytic cells.

MATERIALS AND METHODS
Cell and Tissue Samples-Platelets and mononuclear cells were isolated from citrated whole blood, as described previously (3) Genotyping-Donor genotype with respect to alleles A1, A2, and A3 was determined by BglII/AseI restriction fragment length polymorphism using blood mononuclear cell DNA, as described previously (15). Exon sequence numbering is based on the incomplete human ␣ 2 cDNA sequence reported by Takada and Hemler (17).
Quantitation of Platelet ␣ 2 ␤ 1 by Flow Cytometry-With appropriate informed consent, blood samples were collected from normal volunteers and processed in the GCRC. Six volumes of whole blood were mixed with 1 volume of acid-citrate dextrose National Institutes of Health formula A. An aliquot of whole blood was centrifuged over a cushion of Ficoll-Hypaque to isolate mononuclear cells that were preserved by freezing in Me 2 SO and served as a source of genomic DNA for genotyping. A second aliquot of whole blood was prepared for fluorescenceactivated cell-sorting analysis, as described previously (4,15,18). Fluorescence-activated cell-sorting analysis was initiated within 72 h after phlebotomy. Measurements were obtained using a Becton-Dickinson FACS Star Plus flow cytometer within the technical laboratory of the GCRC.
Sequence Analysis-DNA sequences were obtained using an Applied Biosystems ABI Prism Model 377 DNA Sequencer (PerkinElmer Life Sciences) by personnel in the DNA Core Laboratory of the Department of Molecular and Experimental Medicine, The Scripps Research Institute.
Statistical Analysis-Standard two-way analysis of variance was used to make comparisons of platelet ␣ 2 ␤ 1 density, as determined from flow cytometry binding measurements. In this regard, two murine monoclonal anti-␣ 2 ␤ 1 antibodies, 12F1 and 8C12, were used for the determination of ␣ 2 ␤ 1 density, and analyses were performed separately for each antibody. The main effects evaluated under the analysis of variance model were ␣ 2 genotype (three categories, denoted as 1/1, 1/2, or 2/2) and number of G Ϫ92 or T Ϫ52 substitutions (three categories: 0, 1, and Ͼ1). In addition, terms representing interaction between genotype and number of substitutions were incorporated into the analysis of variance models but were found not to be statistically significant at conventional ␣ level 0.05.
Gel Mobility Shift Analysis-Nuclear extracts were obtained from CHRF288-11 or Dami cells by isolation of nuclei, as described previously (19). An optimal number of cells (nominally, 0.5-1 ϫ 10 8 ) were washed twice in phosphate-buffered saline (centrifugation at 1800 ϫ g for 10 min) and then lysed in 0.5 ml of 0.5% (v/v) Nonidet P-40 in 25 mM HEPES, 50 mM KCl, pH 7.9, containing 1 mM phenylmethylsulfonyl fluoride, 100 M dithiothreitol, 10 g/ml leupeptin, and 20 g/ml aprotonin (lysis buffer). The nuclei were pelleted and rinsed once in lysis buffer without Nonidet P-40 (centrifugation at 10,000 ϫ g for 1 min). By vigorous micropipetting, nuclei were then physically disrupted in 25 mM HEPES, 500 mM KCl, pH 7.9, containing 10% (v/v) glycerol, 1 mM phenylmethylsulfonyl fluoride, 100 M dithiothreitol, 10 g/ml leupeptin, and 20 g/ml aprotonin (extraction buffer). When nuclei were sufficiently emulsified, the mixtures were centrifuged at 10,000 ϫ g for 5 min, and the supernatants were collected. The concentration of nuclear proteins in the supernatants was determined by the method of Bradford (20). Double-stranded DNA probes were labeled with [␣-32 P]dCTP using Klenow DNA polymerase (21). Ten g of nuclear protein were mixed with labeled DNA (5-10 ϫ 10 4 cpm) in 10 l of 25 mM HEPES, 50 mM KCl, 0.5 mM EDTA, pH 7.9, containing 10% (v/v) glycerol, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol (binding buffer) and incubated at ambient temperature for 1 h. In reactions using competitor DNA, a molar excess (as indicated) of the unlabeled competitor DNA fragment was used. In antibody-based supershift or inhibitor experiments, 3 g of anti-Sp1, anti-Sp3, anti-AP2, anti-nuclear factor-B, or anti-Egr-1 (each from Santa Cruz Biotechnology, Inc.) was preincubated with the nuclear extract for 15 min at ambient temperature before addition of the 32 P-labeled probe. The reaction products were separated by polyacrylamide gel electrophoresis using 4% acrylamide/bisacrylamide (37.5:1) in 0.045 M Tris borate and 0.001 M EDTA, pH 7.9 (0.5ϫ Tris borate-EDTA). Protein complexes were visualized by autoradiography.
Transfection Assays-The Dami and CHRF-288-11 cells were transfected by electroporation using a Cell-Porator (Life Technologies, Inc.). Approximately 1 ϫ 10 7 cells were transfected in 500 l of Iscove's modified medium containing 20 g of plasmid DNA and 20 g of pSV-␤-galactosidase DNA by electroporation at 300 V and 1180 microfarads.
Plasmid Constructions-The proximal 5Ј-regulatory region of the ␣ 2 gene corresponding to Ϫ244 through ϩ40 was amplified by polymerase chain reaction using genomic DNA from a donor homozygous for C Ϫ52 and C Ϫ92 as template. The oligonucleotide primers were designed to incorporate KpnI (forward) and BglII (reverse) restriction sites, and the product was inserted into the polylinker site upstream of the luciferase (LUC) reporter gene in the plasmid pGL2-enhancer (Promega, Madison, WI). The primer sequences were as follows (restriction sites are underlined): A2PROFORKpnI, 5Ј-GAGGGTACCAGGAAAGCCTG-CCA-3Ј; and A2PROREVBglII, 5Ј-GAGAGATCTAGAAGCTGTCCAG-AGGGC-3Ј.
DNase I Footprinting-The DNase I footprinting reactions were performed as described previously (22,23). Promoter DNA cloned in pGL2enhancer was digested with XmnI to produce a DNA probe representing the sequence segment from Ϫ248 to Ϫ14. The probe was end-labeled with [␥-32 P]ATP by T4 polynucleotide kinase reaction, digested with BspHI, and subsequently gel-purified (only one end is labeled). An aliquot of the labeled probe (20,000 cpm) was incubated with 0, 20, 40, or 80 g of nuclear extract in 50 l of 25 mM Tris-HCl (pH 8.0), 50 mM KCl, 6.25 mM MgCl 2 , 0.5 mM EDTA, 10% glycerol, 0.5 mM dithiothreitol, and 1 g of poly(dI-dC). The mixture was incubated for 20 min on ice and then incubated for 2 min at room temperature. RQ1 RNase-free DNase (Promega) (0.75 unit in 53 l of 5 mM CaCl 2 and 10 mM MgCl 2 ) was added. The mixture was incubated at room temperature for 2 min, and the reaction was stopped by the addition of 90 l of 1% SDS, 200 mM NaCl, and 30 mM EDTA plus 100 g/ml yeast RNA. The mixture was then extracted with phenol:chloroform, and DNA was precipitated with ethanol. Sequencing reactions of the DNA segment corresponding to the DNA probe used for footprinting were carried out by the dideoxynucleotide chain termination method of Sanger et al. (24) using the fmol sequencing kit (Promega) according to the manufacturer's instructions. The sequencing primer was 5Ј-GTTTCTGGGCAGCTCCTGCAGC-3Ј. Reaction products were analyzed using 6% denaturing (urea) polyacrylamide gel electrophoresis (QuickPoint; Invitrogen, Carlsbad CA) followed by autoradiography.

RESULTS
The 5Ј-regulatory Region Sequence-The naturally occurring dimorphism C Ϫ52 T is located precisely in the middle of tandem Sp1/Sp3 binding elements of the core promoter (15), whereas the C Ϫ92 G dimorphism lies closely downstream from the nonanucleotide -GGGGCGGGG-at Ϫ107/Ϫ99 (Fig. 1). Our findings, summarized below, indicate that the Ϫ107/Ϫ99 sequence represents an additional Sp1-binding site.
Identification of Transcription Factor-binding Sites Surrounding the Ϫ92 Dimorphism-To determine the identity of the transcription factors that might bind to the promoter sequence surrounding the dimorphic position at Ϫ92, DNase I footprinting analysis was performed with nuclear extracts from Dami, K562, or CHRF-288-11 cells and a promoter probe representing the region Ϫ246 to Ϫ14 on the plus strand. Comparable results were obtained with nuclear extracts from each of the three cell lines, and an example of the results obtained with Dami cell extracts is depicted in Fig. 2. The footprint spanned the promoter region from about Ϫ200 to roughly Ϫ30 on the plus strand. Incubation with Dami cell extracts showed protected segments at Ϫ107 to Ϫ99 and Ϫ61 to Ϫ43 (Fig. 2).
The Influence of the C3 G Substitution at Position Ϫ92 on the Binding of Nuclear Proteins-Gel mobility shift analyses were conducted with nuclear proteins from either Dami or CHRF288-11 cells using a 29-bp core promoter fragment (bp Ϫ111 to Ϫ82) that contains either Ϫ92C or Ϫ92G as probe (Fig.  3). In addition, the same segments synthesized with mutations in the upstream Sp1-like binding element to create Cm or Gm. In a representative gel mobility shift assay using a nuclear extract from Dami cells (Fig. 4) Additional gel mobility shift assays were conducted to examine whether the effect of the Ϫ92C or Ϫ92G substitution has an effect on the binding of Sp1 and Sp3 to the downstream elements at position Ϫ52. This was accomplished using the long probes depicted in Fig. 5. Note that these probes contain the Ϫ92C or Ϫ92G dimorphism and the two tandem Sp1/Sp3 binding elements that flank Ϫ52C, but not the Sp1 binding element located upstream at Ϫ107/Ϫ99. As shown in Fig. 6, the same four complexes are formed between Dami cell nuclear proteins and the Ϫ92C long probe (lane 2) or Ϫ92G long probe (lane 7): the slowest mobility complex (arrowhead) is given by binding of Sp1 and/or Sp3; the remaining three complexes are nonspecific (see below). In the absence of Dami cell nuclear proteins, no complexes are formed (lanes 1 and 6). Addition of either anti-Sp1 (lanes 3 and 8) or anti-Sp3 (lanes 4 and 9) results in a supershifted complex (asterisk) in the case of either probe. Coincubation with a 200-fold molar excess of the consensus Sp1-binding oligonucleotide (Sp1c) inhibits formation of the specific complex ( lanes 5 and 10), whereas the nonspecific complexes remain. Repeated attempts to generate specific complexes in gel mobility shift assays using even longer probes containing both the Ϫ52 Sp1-binding elements and the Sp1binding element upstream at Ϫ107/Ϫ99 consistently failed, most likely because of the sheer length of these longer oligonucleotide probe sequences.
Thus, the nuclear protein from Dami cells that forms a specific complex with either the Ϫ92C or the Ϫ92G probe is the Sp1 transcription factor, and the C/G substitution at Ϫ92 does not affect the binding of Sp1 to the downstream elements surrounding the Ϫ52 C/T dimorphism. Essentially identical results are obtained with nuclear proteins from the CHRF288-11 cell line (data not shown).
Reporter Assays-To investigate how decreased affinity of the G Ϫ92 and T Ϫ52 sequences might influence in situ transcription rates, we compared the relative activities of promoter-LUC FIG. 1. The proximal 5-regulatory region of the human ␣ 2 gene. The proximal 5Ј-regulatory region of the human ␣ 2 gene (from positions Ϫ961 to Ϫ1) (top) contains a core promoter region within positions Ϫ92 to Ϫ30 plus upstream enhancer and silencer elements. Within and immediately upstream from the core promoter (bottom) are two closely situated, confirmed Sp1/Sp3-binding sites and a third consensus Sp1 binding sequence (-GGGGCGGGG-). Based on a comparison of 62 human haplotypes, only two dimorphic sequences are present within the proximal 5Ј-regulatory region: namely, C Ϫ92 G and C Ϫ52 T. The former lies immediately downstream from the GGGGC-GGGG sequence; the latter is precisely in between the two confirmed Sp1/Sp3-binding sites.
constructs. The construct p␣ 2 244-LUC contains the consensus promoter sequence from Ϫ244 through ϩ40 (GenBank TM accession number AF062039) with C Ϫ92 and C Ϫ52 . This sequence, derived from a comparison of 62 haplotypes, is not identical to the published ␣ 2 promoter sequence (11) but can be considered the allele sequence most equivalent to that published sequence. The remaining plasmid constructs are as follows: p␣ 2 244⌬92G-LUC, which contains the replacement G at Ϫ92; p␣ 2 244⌬52T-LUC, which contains the replacement T at Ϫ52; and p␣ 2 244⌬92G⌬52T-LUC, which contains both replacements. Each construct was ligated into the vector pGL2-basic or pGL2enhancer (Promega). Identical results were obtained with either vector, so the cumulative results were averaged. Promoter activities were compared after transient transfection of Dami cells or CHRF-288-11 cells. Cotransfection with the vector pSV-␤-galactosidase (Promega) was used to normalize for transfection efficiency. The results of these assays are summarized in Fig. 7.
Dramatic differences were observed in both Dami cells (Fig.  7A) and CHRF-288-11 cells (Fig. 7B). In Dami cells (five separate experiments), the mutation ⌬92G results in a 3-fold decrease in promoter activity, ⌬52T also results in a 3-fold decrease in activity, and the combination ⌬92G⌬52T decreases activity by 8-fold. This finding is consistent with an additive effect of each substitution on gene transcription. In Dami cell transfectants, this translates into a decrease in reporter gene activity. Even more dramatic differences were obtained for CHRF288-11. In three experiments, ⌬92G caused roughly a 5-fold decrease in promoter activity, ⌬52T resulted in about a 4-fold decrease in activity, and the combination ⌬92G⌬52T decreased activity by 20-fold (Fig. 7). Again, the combined effect of the two substitutions was additive. In the case of either cell line, individual nonparametric tests of C versus T were statistically significant at the 0.05 level, using the Wilcoxon test.
Donor Differences in Platelet ␣ 2 ␤ 1 Expression-As summarized under "Introduction," there is a 5-fold range in platelet ␣ 2 ␤ 1 density among normal subjects. In this study, the surface  1, 5, 7, 9, 11, 13, and 15), G (lanes 2, 6, 8, 10 content of platelet ␣ 2 ␤ 1 in whole blood was measured by flow cytometry, using one of two murine monoclonal anti-␣ 2 ␤ 1 antibodies, 8C12 or 12F1. Identical findings were made with each antibody, and the results are depicted in Fig. 8. A total of 71 donors who expressed either allele A1 or allele A2 were subdivided by genotype into three categories (Fig. 8A), denoted 1/1 (homozygous for A1), 1/2 (heterozygous for A1 and A2), and 2/2 (homozygous for A2). Donors were also classified (Fig. 8B) as expressing none (0), one (1), or two or more (Ն2) of the G Ϫ92 or T Ϫ52 substitutions. For each donor, platelet ␣ 2 ␤ 1 density was determined, as reflected by relative binding of both the antibody 8C12 (E) and the antibody 12F1 (q). Two-way analyses of variance were performed with these relative binding data from each antibody to simultaneously evaluate the effects of ␣ 2 genotype and number of G Ϫ92 or T Ϫ52 substitutions on platelet ␣ 2 ␤ 1 density. Virtually identical findings were made with each antibody: in both cases, differences in densities between ␣ 2 genotypes (Fig. 8A) were highly statistically significant (with the 12F1 antibody, F2,62 ϭ 24.90, p Ͻ 0.001; with the 8C12 antibody, F2,62 ϭ 45.01, p Ͻ 0.001), as were differences in densities between the numbers of promoter substitutions (Fig.  8B) (with the 12F1 antibody, F2,62 ϭ 7.42, p ϭ 0.001; with the 8C12 antibody, F2,62 ϭ 4.87, p ϭ 0.011). Group means and standard errors of the mean for the relative binding data are depicted in Fig. 8. In the Fig. 8A, we note that the 1/1 genotypes have the highest ␣ 2 ␤ 1 densities, the 1/2 heterozygotes have intermediate ␣ 2 ␤ 1 densities, and the 2/2 homozygotes have the lowest ␣ 2 ␤ 1 densities, with little quantitative difference between the 12F1 and the 8C12 data. In Fig. 8B, we find an inverse relationship between the number of inherited substitutions within the promoter regions and the platelet density of ␣ 2 ␤ 1 , with the platelet integrin level of donors who express 2 or more of G Ϫ92 or T Ϫ52 consistently lower than that of donors who express a single or no substitutions at these positions.

DISCUSSION
Previous studies of Sp1 binding elements within the proximal promoter region of the human ␣ 2 gene have focused on two tandem elements at positions Ϫ61 to Ϫ53 and Ϫ51 to Ϫ43. These sites have been found to be essential for the transcriptional activity of this gene (9,12). In a recent report (15), we showed that the inherited dimorphism C Ϫ52 T significantly influences the binding of Sp1 to these flanking elements, such that the substitution T Ϫ52 results in decreased binding of Sp1 and decreased Sp1-dependent transcription in megakaryocytic cell lines. In this study, we identify an additional Sp1-binding site at positions Ϫ107 to Ϫ99 and show that the nearby naturally occurring base substitution C Ϫ92 3 G markedly impairs Sp1-dependent transcription of this gene. The C Ϫ92 G dimorphism cooperates with the previously defined C Ϫ52 T dimorphism to regulate the level of transcription of human ␣ 2 genes through mechanisms that undoubtedly involve the activity of both Sp1 and Sp3.
Direct participation of Sp1 and Sp3 at the Ϫ52 binding site was apparent in gel mobility shift assays because the substitution C3 T resulted in an obvious decrease in the binding of FIG. 5. Long oligonucleotide probes encompassing ؊92 and ؊52 for gel mobility shift assays. The 61-bp core promoter fragment (bp Ϫ98 to Ϫ38) that contains either Ϫ92C (C) or Ϫ92G (G) was used. The downstream Sp1-binding elements at Ϫ61 to Ϫ53 and Ϫ51 to Ϫ43 are underlined.
FIG. 6. Binding of nuclear proteins from Dami cells to long oligonucleotide probes. Radiolabeled long oligonucleotide probes containing either Ϫ92 C (lanes 1-5) or Ϫ92G (lanes 6 -10) were incubated alone (lanes 1 and 6) or with a nuclear protein extract from Dami cells, followed by no additions (lanes 2 and 7), addition of polyclonal anti-Sp1 antibody (lanes 3 and 8), addition of polyclonal anti-Sp3 antibody (lanes 4 and 9), or addition of a 200-fold molar excess of unlabeled consensus Sp1-binding oligonucleotide (Sp1c). An arrowhead to the left of the gel indicates the major Sp1-containing complex. Supershifted complexes are designated by an asterisk. were transfected with each of the promoter-LUC reporter gene constructs, and the relative luciferase activity (ordinate) was determined. The constructs examined contained the ␣ 2 5Ј-regulatory sequence extending from Ϫ244 through ϩ48 cloned into the minimal promoter construct, pGL2-basic. The construct ⌬92G⌬52T is indicated by G T, construct ⌬92G is indicated by G C, construct ⌬52T is represented as C T, and C C represents the reference sequence. The mean (solid bars) and one standard deviation are shown for each data group; in the case of Dami cells, n ϭ 5; and for CHRF-288-11 cells, n ϭ 3 both Sp1 and Sp3 (15). In the case of the Ϫ92 site, the substitution C3 G only partially impairs the ability of Sp1 to bind to the adjacent upstream binding site at Ϫ107/Ϫ99. At the same time, supershift assays failed to demonstrate any involvement of Sp3 at this site. Nonetheless, the same substitution markedly decreases transcriptional activity of promoter-LUC constructs in transfection assays. Our evidence for this conclusion is substantial. In transient transfection assays, promoter-LUC constructs that contain the C3 T substitution at nucleotide Ϫ52 exhibit 3-fold and 4-fold decreases in activity in Dami and CHRF-288-11 cells, respectively. In similar assays, the C3 G substitution at nucleotide Ϫ92 causes a 3-fold and a 5-fold decrease in activity in Dami and CHRF-288-11 cells, respectively. The combination of these substitutions has an additive effect, resulting in 8-fold and 20-fold decreases in transcription in either cell line. In a population of 71 normal donors, the expression of both T Ϫ52 and G Ϫ92 correlates with decreased expression of the integrin ␣ 2 ␤ 1 on platelets, acting synergistically with previously defined ␣ 2 allelic differences.
The substitution at the Ϫ92 position has only a weak direct effect on the binding of Sp1 to the upstream site at Ϫ107/Ϫ99, as evidenced by gel mobility shift assays, where one can see a reproducible but modest decrease in the intensity of complexes formed by the Ϫ92G probe relative to those formed by the Ϫ92C probe. On the other hand, the effect of this substitution on the rate of transcription in transfected cell lines is significant. The presence of three Sp1-binding sites within this limited region of the ␣ 2 gene 5Ј-regulatory region (positions Ϫ107 to Ϫ43) introduces a higher level complexity to Sp1-mediated effects on transcription. Thus, the substitution C Ϫ92 3 G may exert indirect effects on Sp1-dependent enhancement of transcription by disrupting the binding of other transcription regulators that might cooperate with Sp1 or bind directly to the promoter region. Previous studies have shown that multiple Sp1 sites within the same gene result in DNA looping when the Sp1 molecules bound to those sites interact with one another or with other proteins (25). This promotes interactions between promoters and distant regulatory elements and between different transcription regulators.
Sp1 and Sp3 share a highly conserved three-zinc finger DNAbinding domain and thus exhibit practically identical specificities and affinities for DNA-binding sites. On the other hand, the Sp1 and Sp3 domains responsible for protein-protein interactions are not conserved. Thus, each factor is differentially affected by the presence and composition of other protein regulators that may be recruited to the site of complex formation.
Consequently, Sp1 and Sp3 may function as positive or negative regulators, depending upon the nature of these regulatory proteins as well as the Sp1/Sp3 ratio within the nucleus under different conditions of cell growth and differentiation (10, 12, 26 -28). For example, Sp1 is a critical negative modulator of the megakaryocyte ␣ IIb gene (29) but a requisite positive modulator of the integrin ␣ 2 gene (9,12). Competition between Sp1 and Sp3 for identical binding elements is also a factor that contributes to gene regulation, and it has been proposed that Sp3 can repress transcriptional activation of the ␣ 2 gene by Sp1 (12). Our previous findings (15) confirmed that such a mechanism is probably involved in transcriptional regulation at the downstream Sp1/Sp3 sites surrounding the Ϫ52 dimorphism. The importance of changes in Sp1/Sp3 ratios and changes in GC box binding factors is further highlighted in studies of the differentiation of primary human keratinocytes (30) or rat lens epithelia explants (31).
In gel mobility shift assays, we observed that complexes between Dami cell nuclear proteins and either the 92C or 92G probes could be supershifted by polyclonal antibodies specific for Sp1, whereas polyclonal antibodies specific for Sp3 had no effect. Because both Sp1 and Sp3 are present in Dami cell nuclear extracts, this finding implies that there is a preferential binding of Sp1 to the upstream site at Ϫ107/Ϫ99. Whereas this result would not be a typical finding with the majority of Sp1/Sp3-binding sites, there is precedent for differential binding of Sp1 and Sp3. Two examples involve the human major intrinsic protein gene (25) and the cyclin D gene (32). A GC box at position Ϫ147/Ϫ152 within the major intrinsic protein gene interacts with purified Sp1 or Sp1 present in mouse lens nuclear extracts (25). A proximal CT box located at position Ϫ49/ Ϫ56 in the human major intrinsic protein gene and conserved in the mouse gene can react with purified Sp1 but cannot interact with Sp1 in lens nuclear extracts, reacting instead with Sp3 (25). This is a case in which auxiliary factors within the nuclear extract are likely modulating the binding of Sp1 without an effect on Sp3. On the other hand, Wang et al. (32) have observed an opposite effect, in which preferential binding of Sp1 to a segment of the cyclin D3 promoter was observed in the presence of ample levels of Sp3 by gel mobility shift assay. Such binding differences are always influenced by the ratio of Sp1 to Sp3 in the nucleus at a particular time, and additional studies to determine the effect of the relative levels of Sp1 and Sp3 on control of ␣ 2 gene transcription are warranted.
Expression differences in platelet integrin ␣ 2 ␤ 1 result from the additive effects of multiple regulatory elements within and FIG. 8. Relationship between ␣ 2 genotype and platelet ␣ 2 ␤ 1 density. The level of platelet ␣ 2 ␤ 1 was determined in freshly drawn whole blood by flow cytometry using murine monoclonal anti-␣ 2 ␤ 1 antibody 12F1 (closed symbols) or 8C12 (open symbols). Mean fluorescence intensity (MFI) is plotted on the ordinate (left ordinate, 12F1; right ordinate, 8C12). A, all donors expressed only alleles A1 or A2 and are subdivided (abscissa) with respect to expression of these alleles (1/1, homozygous A1/A1; 1/2, heterozygous for A1 and A2; 2/2, homozygous A2/A2). B, donors are classified into three groups (abscissa) relative to the number of inherited substitutions (G Ϫ92 or T Ϫ52 ) within the promoter region: none (0), one (1), or two or more (Ն2). Each data point represents the mean, and vertical bars represent the standard error of the mean for each group.
outside of the ITGA2 5Ј-regulatory region. The molecular basis for the expression differences associated with ITGA2 alleles (A1, A2, and A3) remains to be determined, but our results certainly indicate that it does not involve the proximal 5Јregulatory region. The molecular differences responsible for the distinction of alleles A1, A2, and A3 are not likely to be located within the 5Ј-regulatory region of the gene because the allele frequencies of the Ϫ52 and Ϫ92 dimorphisms are radically different from those of the intragenic dimorphisms that distinguish alleles A1, A2, and A3. In a "typical" Caucasian population, they are as follows: 0.85 for Ϫ92C, 0.15 for Ϫ92G, 0.65 for Ϫ52C, and 0.35 for Ϫ52T. On the other hand, the allelic frequencies of A1, A2, and A3 are 0.39, 0.53, and 0.08, respectively. Because we have found no further dimorphisms between position Ϫ52 and the beginning of intron 1, the molecular basis for expression difference between alleles A1, A2, and A3 must be found somewhere 3Ј to the beginning of intron 1. Our subsequent efforts will be directed toward the characterization of intragenic regulatory elements in the proximal portion of the gene, particularly within the unusually large 37-kilobase intron 1.
In the case of the human ␣ 2 gene, allelic differences have a profound effect on expression of the platelet integrin ␣ 2 ␤ 1 , and these expression differences can influence morbidity and risk for fatality in certain disease states, where the adhesive function of platelets is critical. Those alleles that are associated with low receptor density, alleles A2 and A3, are overrepresented in symptomatic patients with type 1 von Willebrand disease and may thus predispose these individuals to increased risk for bleeding (33). On the other hand, the inheritance of allele A1, associated with high receptor density, represents a risk factor for diabetic retinopathy (34) and for acute coronary disease (e.g. myocardial infarction) or stroke in younger patients (35,36). In this report, we show that additional molecular differences within the 5Ј-regulatory and promoter region of the ␣ 2 gene also contribute to transcriptional regulation of this gene. Combined with the underlying genetic basis for the existence of multiple alleles, the molecular differences described in this report help to explain the diversity of receptor densities among normal subjects and the absence of an apparent gene dosage effect. These findings will pave the way for a more thorough characterization of the mechanisms that control expression of this important receptor.