Cloning and Analysis of the Thrombopoietin-induced Megakaryocyte-specific Glycoprotein VI Promoter and Its Regulation by GATA-1, Fli-1, and Sp1*

The exposure of collagen fibers at sites of vascular injury results in the adherence of platelets and their subsequent activation. The platelet collagen receptor glycoprotein (GP)1 VI plays a crucial role in platelet activation and thrombus formation and decreased levels or defective GPVI may lead to excessive bleeding. In addition, elevated levels of collagen receptors may predispose individuals to coronary heart disease or strokes. GPVI expression is restricted to platelets and their precursor cell, the megakaryocyte. In this study we investigate the regulation of GPVI expression and show that thrombopoietin induces its expression in the megakaryocytic cell line UT-7/TPO. A 5′-region flanking the transcription start point of the GPVI gene was cloned (−694 to +29) and we report that this putative GPVIpromoter bestows megakaryocye-specific expression. Deletion analyses and site-directed mutagenesis identified Sp1227, GATA177, and Ets48 sites as essential forGPVI expression. We show that transcription factors GATA-1, Fli-1, and Sp1 can bind to and activate this promoter. Finally, GPVI mRNA was detected only in megakaryocytic cell lines expressing both Fli-1 and GATA-1, and we show that overexpression of Fli-1 in a stable cell line (which expresses endogenous GATA-1 and Sp1) results in expression of the endogenous GPVI gene.

Megakaryopoiesis is the process by which hematopoietic stem cells in the bone marrow differentiate into mature megakaryocytes (MKs), 1 which in turn release platelets into the bloodstream. The molecular mechanisms controlling MK development and platelet production are of considerable interest for a number of reasons. First, the identification of tran-scription factors controlling MK-specific gene expression aid our understanding of blood cell lineage commitment. Second, MKs possess unique features such as polyploid nuclei, secretory granules, and proplatelets, the study of which contributes to our understanding of the mechanisms controlling cell cycle and organelle formation. Finally, the genetic basis of disease states such as bleeding disorders and leukemias may be elucidated from the study of platelet function and normal blood cell development.
Thrombopoietin (TPO) is the major humoral regulator of megakaryopoiesis, and TPO-responsive genes play a major role in regulating MK development and platelet production. The binding of TPO to its receptor c-Mpl on the cell surface activates a number of distinct intracellular signaling cascades (Ref. 1 and references within), resulting in the proliferation of MK progenitors and the expression of a number of cell-specific genes associated with MK differentiation such as GPIIb and GPIX (2,3).
One approach to understanding the molecular basis of megakaryopoiesis is to identify genes that increase in expression following TPO stimulation. Because of the rarity of MKs in bone marrow, their isolation and subsequent culturing in vitro is not trivial. However, a number of cell lines with MK features exist and these have provided useful insights into megakaryopoiesis and the regulation of platelet-specific genes in particular. We have used the cytokine-responsive MK cell line UT-7/ TPO and the technique of representational difference analysis (RDA) to isolate TPO-induced genes. This technique has the advantage over cDNA arrays as novel genes can potentially be identified and it is possible that MK-specific genes are underrepresented in data bases of expressed sequences and also on commercial cDNA arrays because of the low numbers of MKs. In this way, we have identified the MK and platelet-specific collagen receptor GPVI as one of the genes up-regulated in UT-7/TPO cells in response to TPO. Although GPVI has been the focus of several recent reports, the transcriptional regulation of this clinically relevant gene has not been determined. Therefore, we set about cloning the GPVI promoter and have identified the key transcription factor binding sites necessary for MK-specific expression, which include GATA, Ets, and Sp1. We demonstrate using gel shift analysis that GATA-1, Fli-1, and Sp1 can bind to these sites within the GPVI promoter and that their overexpression in non-MK cells can activate a GPVIluciferase reporter in transient transfection assays. Moreover, Northern analysis of a variety of erythro-megakaryocytic cell lines revealed GPVI expression only in those cell lines expressing both Fli-1 and GATA-1. Finally, we show that overexpression of Fli-1 in K562 cells, which express Sp1 and GATA-1 but normally lack GPVI and Fli-1, results in the expression of the endogenous GPVI gene.
RNA Isolation and cDNA Synthesis-Total RNA was isolated from 5 ϫ 10 7 cells by guanidinium isothiocyanate lysis (TRIzol, Invitrogen) as described by the manufacturer. Polyadenylated (poly(A) ϩ ) RNA was subsequently isolated using biotinylated oligo(dT) primers and streptavidin coupled to magnetic beads (PolyATtract TM ; Promega). The Ribo-Clone cDNA synthesis system (Promega) was then used to convert 2-g aliquots of poly(A) ϩ RNA into cDNA.
RDA of TPO-induced UT-7/TPO Cells after Cytokine Starvation-UT-7/TPO cells were grown in IMDM plus 1% FCS for 16 h (TPO "starved") and then TPO was added to 100 ng/ml for 48 h (TPO "induced"). RNA from TPO-starved and TPO-induced UT-7/TPO cells were then isolated and converted to cDNA (see above). RDA of these cDNAs was as previously described (6 -8) using the TPO-induced cDNA as "tester" and TPO-starved cDNA as "driver." After three rounds of hybridization and PCR, discrete DNA bands could be visualized on 1.2% agarose gels. These DNA bands were excised from the gels, cloned into the BamHI site of pBluescript-KSϩ (Stratagene), and sequenced. The identities of the cloned cDNAs were revealed by comparison of the sequences to those in the GenBank TM data base using BLAST algorithms.
Probing RDA cDNA Arrays with TPO-starved and TPO-induced cDNAs-A number of the pBluescript-KSϩ/cDNA clones isolated using the RDA described above were blotted in duplicate onto Hybond N ϩ nylon membrane (Amersham Biosciences) using a slot-blot apparatus. Actin and GAPDH cDNA fragments were included as controls. Equivalent amounts of TPO-starved and TPO-induced UT-7/TPO cDNAs were radiolabeled by PCR and the incorporation of [␣-32 P]dCTP as described previously (8), and these probes were purified through a Sephadex G-50 column (Amersham Biosciences) and hybridized to the filters using ExpressHyb (Clontech) as described by the manufacturer. The signals from the ␤-actin and GAPDH genes were compared with ensure that cDNA probes with similar specific activities had been used.
Northern Blotting-As indicated in the figure legends, 1 to 3 g of poly(A) ϩ RNA were subjected to electrophoresis on 1% agarose gels containing formaldehyde and MOPS (9) and transferred to Hybond N ϩ nylon membrane (Amersham Bioscience). Membranes were then hybridized overnight at 68°C using ExpressHyb (Clontech) with radiolabeled cDNA probes. Membranes were then washed as described in the ExpressHyb protocol.
Construction of GPVI Luciferase Reporter Constructs-The genomic sequence adjacent to the human GPVI exon 1 was found in a Gen-Bank TM working draft sequence of chromosome 19 and has been deposited at GenBank TM AF521646. Primers were designed corresponding to sequences overlapping the GPVI translation start site at ϩ29 ( Fig. 2A, nt ϩ22 to ϩ44) and also a sequence some 0.7 kb upstream ( Fig. 2A, nt Ϫ701 to Ϫ679). This putative promoter region was amplified by PCR using human genomic DNA as template and a DNA band of the expected size was isolated. Restriction sites for KpnI (adjacent to GPVI nt Ϫ694) and BglII (adjacent to GPVI nt ϩ29) endonucleases were incorporated using a second set of PCR primers and the 0.7-kb PCR fragment as template, enabling cloning into the promoterless luciferase reporter vector pGL3-basic (Promega). This construct was confirmed by sequencing and named GPVI-694. The GPVI promoter deletions (see diagram in Fig. 2, B and D) were generated by PCR using GPVI-694 as template, the BglII primer as above, and various forward primers with KpnI sites incorporated adjacent to the GPVI site as indicated by the construct name (e.g. in GPVI-315, the KpnI site is adjacent to nt Ϫ315, Fig. 2A) and each of these PCR fragments were cloned into pGL3-basic and sequenced. The name of the GPVI constructs is indicative of the GPVI promoter region cloned into pGL3-basic, e.g. GPVI-118 includes GPVI nt Ϫ118 to ϩ29, GPVIdel-605/Ϫ462 includes GPVI nt Ϫ695 to ϩ29 with a deletion of nt Ϫ605 to Ϫ462 (see diagram in Fig. 2B). Plasmids GPVI-315Smut227, GPVI-315Gmut177, and GPVI-315Emut48 were generated by the mutation of Sp1 227 (GGGGCCTGGT to GGGTACTGGT), GATA 177 (AGATAA to CGCTTA), and Ets 48 (CAG-GATGT to CAGTCTGT) binding sites within GPVI-315, respectively, using the QuikChange site-directed mutagenesis kit (Stratagene).
Transient Transfections-Expression plasmids for Sp1 (human Sp1 cloned into pEF-Bos) and GATA-1 (murine GATA-1 cloned into pRc/ CMV vector, Invitrogen) were a gift from Dr. Merlin Crossley, (Department of Biochemistry, University of Sydney, Sydney, Australia). The Fli-1 expression vector (human Fli-1 cloned in pcDNA3, Invitrogen) is described elsewhere. 2 The GPVI promoter-luciferase reporter constructs are described above. All plasmids used in transfections were purified using a plasmid purification kit (Qiagen) and at least 2 different batches of each GPVI-reporter construct were tested. The total amount of plasmid DNA in each transfection was kept constant by adding pcDNA3 as required.
For UT-7/TPO cell transfection, exponentially growing cells were washed with IMDM, seeded at 5 ϫ 10 5 cells/ml in IMDM (no FBS or TPO; 0.8 ml/well) in six-well plates, and transfected with 1.2-1.5 g of plasmid DNA using LipofectAMINE reagent (Invitrogen). After 4 h, an equal volume of IMDM supplemented with 20% FBS and 20 ng/ml TPO was added to each well. The cells were incubated for 48 h at 37°C in a CO 2 incubator and luciferase activities were assayed according to the manufacturer's instructions (luciferase assay system, Promega).
HeLa cells were seeded at 1 ϫ 10 5 /ml Dulbecco's modified Eagle's medium plus 10% FBS in six-well plates (2 ml/well) and incubated for 24 h at 37°C in a CO 2 incubator. Cells were then washed with Dulbecco's modified Eagle's medium and transfected with the indicated plasmids using LipofectAMINE reagent. Luciferase activities were assayed after 48 h according to the manufacturer's instructions. Each experiment was carried out at least 3 times with duplicate or triplicate samples.

Identification of Differentially Expressed Genes in UT-7/ TPO Cells following Stimulation with Thrombopoietin-UT-7/
TPO cells resemble mature megakaryocytes and require TPO for growth and survival (4). To identify TPO-regulated genes, we have used the technique of RDA to isolate cDNAs corresponding to genes that are expressed at higher levels in TPOstimulated UT-7/TPO cells than in cells starved of this cytokine. A selection of the cDNAs isolated in this way, including GPVI, Stat5b, and SOCS-3, are shown in Fig. 1A. cDNAs corresponding to the housekeeping genes ␤-actin and GAPDH were included on the blots as controls and the intensity of these bands in the left and right panels are approximately equal indicating that cDNA probes of similar specific activities had been used. In comparison, the intensity of the bands in the right panel is generally greater than those on the left, suggesting that the majority of the cDNAs isolated did represent up-regulated genes. These bands of differing intensity and the isolation of Pim-1, which had been reported previously to be induced by TPO in another MK cell line FD-TPO (11), were encouraging and suggested that our RDA had identified differentially expressed genes following TPO induction. The identification of GPVI is of particular interest as its expression is restricted to megakaryocytes and platelets (12,13), and as such this gene was subject to further analysis.
To confirm that GPVI expression is regulated by TPO, GPVI mRNA levels in cytokine-starved and TPO-stimulated UT-7/ TPO cells were assessed by Northern analysis. Fig. 1B shows the time course of GPVI induction by TPO. There is minimal expression of GPVI mRNA at t ϭ 0 and 2 h following the addition of TPO. However, by 4 h GPVI is markedly up-regulated. Two GPVI transcripts are detected and GPVI mRNA levels remain elevated to 120 h following TPO exposure. By comparing the position of these bands relative to the ribosomal RNA bands, we estimate that the major GPVI transcript is ϳ2 kb and the less abundant, higher molecular weight transcript is ϳ3.6 kb (data not shown). As there have been no reports on the regulation of GPVI expression, we set about cloning the GPVI promoter and identifying the important regulatory elements and trans-acting factors crucial for MK-specific expression of the GPVI gene.
Cloning of the GPVI Gene 5Ј-Flanking Region and Identification of 2 Major Regulatory Regions-A search of the Gen-Bank TM high throughput genomic sequences resulted in the identification of genomic sequences upstream of the GPVI transcriptional start site (ϩ1 as defined in Ref. 14; Fig. 2A) and the sequence of a 819-bp region has been submitted to GenBank TM (AF521646). To determine whether this sequence included the GPVI promoter region, a fragment extending from Ϫ694 to ϩ29 was amplified by PCR from human genomic DNA and cloned into a promoterless luciferase construct, creating the reporter plasmid GPVI-694. Transient transfection of the megakaryocyte-like cell line UT-7/TPO with this plasmid resulted in luciferase levels some 30-fold higher than the promoterless control plasmid pGL3basic, and more than 2 times higher than the SV40 promoter-driven construct (Fig. 2B, compare columns 1, 2, and 9). Importantly, when this GPVI-694 plasmid was transfected into the nonmegakaryocytic HeLa cell line, luciferase levels were less than 2% of that of the SV40-luciferase construct, and only slightly higher than the promoterless control (Fig. 2C). Together, these results suggest that the Ϫ694 to ϩ29 region includes the cis-regulatory regions required for MK-specific expression of GPVI.
To roughly map the region(s) necessary for high level expression of GPVI in UT-7/TPO cells, an initial series of GPVI promoter constructs were created bearing large deletions from the 5Ј-end of the promoter. As can be seen in Fig. 2B, deletion of the region from Ϫ694 to Ϫ462 reduced luciferase levels by ϳ30% (compare rows 2 and 3), but further deletion from Ϫ462 to Ϫ315 resulted in luciferase levels comparable with the fulllength construct GPVI-694 suggesting positive and negative regulatory regions, respectively. However, the deletion of the region from Ϫ315 to Ϫ118 (GPVI-118) drastically reduced luciferase levels to around 10% of the GPVI-694 or GPVI-315 activities (compare column 5 to columns 2 and 4).
The removal of upstream sites in the 5Ј-deletion series described above may mask potential effects of binding motifs further downstream. For example, because expression of the GPVI-118 construct was only 10% of the full-length promoter, the effect of further deletions may not be observed. Therefore, we created a second deletion series in which segments of the GPVI promoter were removed, leaving upstream elements intact (Fig. 2B, rows 6 -8). The deletion from Ϫ605 to Ϫ462 (GPVIdelϪ605/Ϫ462) resulted in a modest (ϳ30%) decrease in promoter activity (compare row 6 to row 2). The second deletion (GPVIdelϪ315/Ϫ118), removing a 193-nt sequence between Ϫ315 and Ϫ118 but otherwise leaving the remaining 0.5-kb promoter region intact, resulted in a dramatic (94%) decrease in luciferase expression to levels similar to the GPVI-118 construct (compare row 7 to rows 2 and 5). Finally, deletion of a region proximal to the transcription start site, from Ϫ118 to Ϫ27, also affected GPVI promoter activity, reducing luciferase levels to 20% of the full-length construct GPVI-694 (compare row 8 to row 2).
Previous studies with several MK-specific promoters have identified negative regulatory regions, which when deleted result in expression of promoter constructs in nonmegakaryocytic cell lines (15,16). To determine whether GPVI is similarly regulated, we tested the deletion constructs in the nonhematopoietic HeLa cell line. None of the GPVI constructs were expressed at significant levels ( Fig. 2C) suggesting that no such regions exist in the GPVI promoter. The two regions, Ϫ315 to Ϫ118 and Ϫ118 to Ϫ27, which, when deleted, caused significantly reduced (16-and 5-fold, respectively) luciferase expression compared with GPVI-694 in UT-7/TPO cells were then analyzed in more detail.
Fine Mapping of the Ϫ315 to Ϫ118 GPVI Region-A second series of GPVI 5Ј-deletion mutants were generated to further define the region(s) necessary for maximal expression. As luciferase levels driven by the GPVI-315 construct were comparable with that of the longer GPVI-694 plasmid, all future transfection experiments show luciferase levels relative to this shorter construct. Fig. 2D shows the results of a typical experiment when these constructs were transfected into UT-7/TPO cells. The deletions from Ϫ315 to Ϫ235 had only a modest effect on the promoter activity (compare rows 2-4 to row 1). However, the deletion of a region from Ϫ235 to Ϫ210 within the GPVI promoter, including a potential Sp1 binding site (Sp1 227 ), reduced the activity of the promoter by ϳ60% (compare row 5 to row 4). Luciferase levels were further reduced (3-fold, row 6 compared with row 5) with the deletion of sequences from Ϫ210 to Ϫ159. This region contains a consensus GATA site at Ϫ177 (Fig. 2A, GATA 177 ). The similar luciferase activities of GPVI-159 and GPVI-118 suggest that removing an additional 41 nucleotides between Ϫ159 and Ϫ118 had no apparent effect on residual GPVI promoter activity.
Gel Mobility Shift Assays Identify Proteins in UT-7/TPO Nuclear Extracts Binding to Sp1, GATA, and Ets Sites within the GPVI Promoter-Our deletion analyses using transient transfection assays had identified three regulatory regions within the GPVI promoter: Ϫ235 to Ϫ210, Ϫ210 to Ϫ159, and Ϫ118 to Ϫ27. Presumably, transcription factors present in UT-7/TPO cells could bind to sites within these regions and modulate the transcriptional activity of the GPVI promoter. The sequence of these regions were analyzed for consensus binding sites using Transfac (bioinformatics.weizmann.ac.il/ transfac) and Motif (motif.genome.ad.jp) data bases. A number of potential sites were identified including Sp1 227 , GATA 177 , and Ets 48 sites. Therefore, we performed gel mobility shift experiments with various probes derived from these regions to determine whether protein factors expressed in UT-7/TPO cells could bind to these sites. Fig. 3A shows the EMSA analysis of the first of these regions: the probe S227 corresponds to the region Ϫ231 to Ϫ209 within the GPVI promoter fragment, encompassing the potential Sp1 binding site at Ϫ227 (see Fig. 2A). Incubation of S227 with nuclear extracts from UT-7/TPO cells resulted in the formation of two distinct protein complexes (Fig. 3A, lane 1). One of these (labeled ns) was a nonspecific complex as it could be competed away using nonrelated sequences (data not shown). The second lower mobility complex (labeled s1) is shifted by an Sp1 monoclonal antibody (lane 2). Moreover, mutation of the Sp1 227 site (Smut227, lane 3) prevents the formation of this s1 complex. Therefore, it appears that Sp1 is both present and able to bind the Sp1 227 site within the GPVI promoter in UT-7/TPO cells.
The second region we analyzed, Ϫ210 to Ϫ159, included a potential GATA binding site at Ϫ177 (GATA 177 ). GATA-1 is highly expressed in UT-7/TPO cells (see Fig. 5A), and GATA sites have been shown to play an important role in the regulation of a number of MK-specific genes. Thus, we first used recombinant GATA-1-NC protein (truncated GATA-1 corresponding to the DNA-binding domain 2 ) to determine whether GATA-1 could bind to the GPVI promoter fragment G177 (from Ϫ195 to Ϫ163, Fig. 2A). Fig. 3B shows that GATA-1-NC does bind to G177 (lane 1) and that mutation of the GATA 177 site prevents GATA-1-NC binding to Gmut177 (lane 2). Next, we determined whether UT-7/TPO nuclear extracts contained proteins that interact with G177, and in particular, to the GATA 177 site, and these results are also shown in Fig. 3B, lanes 3-6. A major complex, labeled g1, could be competed away with a GATA consensus oligonucleotide (lane 4), whereas the addition of an excess of either Gmut177 (the probe with mutated GATA 177 site, lane 5), or a nonrelated sequence (lane 6), had no effect on the formation of this complex. There is also a reduction in the intensity of a faster migrating complex labeled ns in two of the samples (lanes 4 and 6 compared with lanes 3 and 5).
However, as an oligonucleotide containing no GATA binding site (Non-Rel, lane 6) competed as effectively as the GATA consensus oligonucleotide, we concluded that this complex was not GATA site-specific. Therefore, we have shown that a GATA site-binding complex is present in UT-7/TPO cells and that it is able to bind to the G177 fragment within the Ϫ210 to Ϫ159 region of the GPVI promoter.
The third promoter region, Ϫ122 to Ϫ27, included an Ets  lanes 1 and 2) purified from E. coli, or UT-7/TPO nuclear extracts (UT-7/TPO NE, lanes 3 and 4), incubated with GPVI Ϫ69 to Ϫ32 promoter fragments, E48 and Emut48 (mutated Ets site). The two bands of interest in lane 3 are labeled e1 and e2. motif at Ϫ48 (Ets 48 ). Ets factors have also been implicated in the regulation of many MK-specific promoters. As the Ets factor Fli-1 is present in both MK and platelets (17), we tested whether recombinant Fli-1 protein could bind to a fragment, E48 (see Fig. 2A), corresponding to Ϫ69 to Ϫ32 of the GPVI promoter and encompassing the potential Fli-1/Ets binding site. Fig. 3C shows the results of these gel retardation experiments: recombinant Fli-1-Ets protein (truncated Fli-1 including the DNA-binding Ets domain) does bind to the GPVI E48 probe (lane 1) and mutation of the Ets 48 site disrupts Fli-1 binding (lane 2). This figure also shows the complexes formed when nuclear extracts from UT-7/TPO cells are incubated with either the E48 or Emut48 probes (lanes 3 and 4, respectively). Two low mobility complexes in lane 3, labeled e1 and e2, are of particular interest as their formation requires an intact Ets 48 site. Although the intensity of a faster migrating complex binding to the Emut48 probe was also decreased slightly (lane 4 labeled ns), this complex was not Ets site-specific as when a related probe was used (overlapping E48, Ets site mutated) the intensity of this complex increased with respect to the wildtype sequence (data not shown). Together, these EMSA results show that proteins present in UT-7/TPO nuclear extracts bind to the Ets 48 site within the Ϫ122 to Ϫ27 GPVI promoter region and that the Ets factor Fli-1 can bind to this site.
Functional Importance of the Sp1 227 , GATA 177 , and Ets 48 Sites-To assess the functional importance of the Sp1 227 , GATA 177 , and Ets 48 sites identified in the EMSA analysis above, we used site-directed mutagenesis to alter the sites within the GPVI-315 luciferase reporter construct to those corresponding to the mutated probes used in the EMSA studies. The effect of each of these mutations on GPVI promoter activity in transiently transfected UT-7/TPO cells are shown in Fig. 4A. Mutation of the Ets 48 site resulted in luciferase levels ϳ5-fold lower than the wild type GPVI-315 construct (compare column 3 to column 4), indicating that the Ets 48 site plays an important role in GPVI promoter activity. The GATA 177 mutation (GPVI-315Gmut177) also had a dramatic effect on GPVI promoter activity as luciferase levels were reduced 16-fold compared with the wild type promoter (compare column 2 to column 4). Finally, mutation of the Sp1 227 site (GPVI-315Smut227) decreased luciferase levels by 40% (compare column 1 to column 4).
GATA-1, Fli-1, and Sp1 trans-activate the GPVI-315 Construct in Nonmegakaryocytic Cell Lines-Because of the importance of the Sp1 227 , GATA 177 , and Ets 48 sites for GPVI promoter activity, and the presence of Sp1, GATA-1, and Fli-1 in MKs, we decided to determine whether these factors could trans-activate the GPVI promoter in transient transfection assays. Neither GATA-1 nor Fli-1 are expressed in HeLa cells and luciferase levels driven by the GPVI-315 construct are barely higher than the promoterless control (Fig. 2C, rows 4 and 9). However, co-transfection of a Fli-1 expression plasmid with the GPVI-315 reporter construct resulted in a dose-dependent increase in luciferase levels, at least 9-fold above basal levels (GPVI-315 alone; Fig. 4B, columns 2-4, compared with column  1). Interestingly, the GATA-1 expression plasmid alone had only a modest effect on GPVI-315 activity: luciferase levels were 2-3-fold higher than GPVI-315 alone (Fig. 4B, columns  5-7 compared with column 2). This was somewhat surprising as the GATA 177 mutation reduced GPVI promoter activity in UT-7/TPO cells by greater than 90% (Fig. 4A, column 2 compared with column 4). This suggested that activation of the GPVI promoter by GATA-1 might require an additional factor(s) not present in HeLa cells. In another study, we have found that GATA-1 and Fli-1 can physically interact and their association is involved in the synergistic activation of two other MK-specific promoters, GPIX and GPIb␣. 2 Therefore, we cotransfected HeLa cells with the GPVI-315 reporter construct and both GATA-1 and Fli-1 expression vectors (Fig. 4B, columns 8 -10). The resulting luciferase levels were greater than additive suggesting that GATA-1 and Fli-1 may synergize to regulate GPVI expression, perhaps through their physical interaction or the recruitment of additional factors (see "Discussion").
The GPVI-315 construct was also cotransfected with an Sp1 expression vector. As can be seen in Fig. 4C, Sp1 had a dramatic effect on GPVI-315 expression and resulted in luciferase levels 30-fold higher than the reporter alone. Thus, GATA-1, Fli-1, and Sp1 can activate the MK-specific GPVI promoter.
Expression of GPVI, GATA-1, and Fli-1 in Cell Lines with Erythroid and Megakaryocytic Characteristics-To determine whether GPVI expression correlates with the endogenous expression of GATA-1, Fli-1, and Sp1, we performed Northern blot hybridization of mRNAs from four erythro-megakaryocytic cell lines: Dami, K562, UT-7/TPO, and UT-7/EPO Mpl. Sp1 is ubiquitously expressed so was not included in this analysis. As can be seen in Fig. 5A, GATA-1 is expressed in all four cell lines. However, Fli-1 mRNA can only be detected in Dami and UT-7/TPO cells (lanes 1 and 3, respectively), and it is these same two cell lines that express GPVI. The similarity in GAPDH levels indicates that equivalent amounts of poly(A) ϩ RNA were loaded in each lane. Therefore, all three of these factors are expressed in cell lines expressing GPVI, further supporting a role for Sp1, GATA-1, and Fli-1 in GPVI regulation.
Overexpression of Fli-1 in K562 Cells Results in Expression of the Endogenous GPVI Gene-Although GATA-1 and Fli-1 had been shown to regulate GPVI-315 in transient assays in a non-MK cell line (Fig. 4B), chromosomal context and cellular environment play essential roles in gene regulation. Therefore, we assessed the importance of GATA-1 and Fli-1 on the expression of the endogenous GPVI. The cell line K562, which has some MK features but is generally thought to represent erythroid cells, expresses GATA-1 but expresses little or no Fli-1 or GPVI (Fig. 5A, lane 2). We have established the clonal cell lines K562-GFP and K562-Fli by transfecting K562 cells with either the control vector pIRES 2 -EGFP or the Fli-1 expression vector pIRES 2 -EGFP/Fli-1, respectively. 2 Stably transfected cells were then selected that were G418-resistant and expressed high levels of green fluorescent protein (GFP) as determined by fluorescence-activated cell sorter analysis (data not shown).
Overexpression of Fli-1 in the K562-Fli cell line was confirmed by Northern blotting (Fig. 5B, lane 2). Both of the cell lines chosen for Northern analysis exhibited similar levels of GFP and GFP expression was maintained even after extensive cell passaging.
The levels of GATA-1 and GPVI mRNAs in the cell lines K562-GFP and K562-Fli were then analyzed by Northern hybridization (Fig. 5B). The similar intensity of GAPDH bands indicate equal amounts of mRNAs were loaded. Overexpression of Fli-1 did not appear to effect GATA-1 expression as similar mRNA levels are seen in both cell lines. Importantly, although no GPVI mRNA was detected in the K562-GFP cell line (lane 1), GPVI was expressed in K562-Fli cells (lane 2). Therefore, overexpression of Fli-1 in K562 cells resulted in expression of the endogenous GPVI gene, providing further evidence that Fli-1 in combination with GATA-1 and Sp1 may play a role in GPVI regulation.
Expression of GATA-1, Fli-1, and Sp1 in UT-7/TPO Cells Stimulated with Thrombopoietin-Our experiments to date have shown that TPO stimulation of UT-7/TPO cells results in the increased expression of GPVI; GATA, Ets, and Sp1 sites in the promoter play an important role in GPVI expression in UT-7/TPO cells maintained in TPO; and that GATA-1, Fli-1, and Sp1 can bind to the GPVI promoter. Therefore, we were interested in determining whether the increase in GPVI expression in TPO-stimulated UT-7/TPO cells is because of the TPO-induced expression of GATA-1, Fli-1, or Sp1. As can be seen in Fig. 5A, lane 3, GPVI mRNA can readily be detected in UT-7/TPO cells maintained in TPO. However, following 16 h starvation of TPO, GPVI mRNA levels are negligible (Figs. 1B and 6, "0" time point) but increase to near maximal levels within 4 h of TPO stimulation. Fig. 6 shows the levels of GATA-1, Fli-1, and Sp1 mRNAs in TPO-starved and -stimulated UT-7/TPO cells. The level of expression of all three of these factors are maintained throughout the time course of this experiment. Therefore, although our experiments have demonstrated a critical role for GATA-1, Fli-1, and Sp1 in regulating constitutive expression of the GPVI promoter, the role that these factors play in TPO-induced expression of GPVI is unclear (see "Discussion").

DISCUSSION
Thrombopoietin supports hematopoietic stem cell survival, stimulates progenitors committed to various lineages, causes megakaryocyte progenitors to proliferate, and induces the expression of surface proteins necessary for platelet function. To identify TPO-responsive genes, we used the RDA technique and this resulted in the cloning of a number of different cDNAs. Several of these correspond to genes encoding proteins involved in signaling, such as the hematopoietic-specific G protein G␣16 (18), Stat5b, the Stat5 target genes SOCS-3 and Pim-1 (19), and CalDAG-GEF-1 (Map4K). Genes encoding enzymes such as phosphoenolpyruvate carboxykinase, which is involved in gluconeogenesis, and cystathionine ␤ synthase were also identified. Heterogeneous nuclear ribonucleoprotein D (or AUF1) was also cloned and this gene encodes a protein that binds to short lived mRNAs and is involved in regulating expression of genes such as c-myc, c-jun, and c-fos (20). Although preliminary experiments suggested that the expression of these genes did increase following TPO stimulation of UT-7/TPO cells (Fig. 1A), their regulation is beyond the scope of this investigation and they were not pursued further. However, the isolation of GPVI and Pim-1 by RDA attest to the success of this technique as both GPVI and Pim-1 are indeed regulated by TPO (this report and Ref. 11).
GPVI, a collagen receptor specifically expressed on the surface of MKs and platelets, plays an important role in limiting blood loss. It is noncovalently associated with the Fc receptor ␥ chain, the signaling subunit of the complex. Upon injury of the vessel wall, collagen fibers are exposed that are highly thrombogenic. Initially, the platelets adhere to the collagen via integrin ␣ 2 ␤ 1 , and this is followed by platelet activation involving GPVI. Defective collagen-induced responses in a patient lacking GPVI testify to the importance of GPVI in this process (21). In addition, the level of collagen receptors on platelets may predispose individuals to either excessive bleeding (low collagen receptor levels) or coronary heart disease and strokes (high levels), and genetic screening of patients with a family history of these diseases for GPVI polymorphisms could aid the planning of treatment strategies (22). It is also possible that GPVI will prove to be a therapeutic target as mice treated with an anti-GPVI antibody prior to collagen and adrenalin infusion were protected from lethal thrombus formation (23).
Because of the clinical importance of GPVI and its restricted expression to MKs and platelets, we chose to examine the regulation of this gene. Northern blot analysis confirmed that the abundance of GPVI mRNA increased in UT-7/TPO cells stimulated with TPO. Two transcripts were detected, a major 2-kb mRNA and a less abundant message of ϳ3.6 kb. These two transcripts may represent alternatively spliced mRNAs, the utilization of an alternative promoter, or differences in the 3Ј-untranslated region. Three GPVI isoforms and differences in the 3Ј-regions have been discussed in another study (14). GPVI levels have previously been shown to be higher in mature MKs than in immature cells (24) and to increase in megakaryoblastic cell lines HEL and CMK induced to differentiate with phorbol 12-myristate 13-acetate (13). However, the factors controlling the regulation of GPVI expression have not been investigated.
Several groups reported the cloning of the GPVI gene from mice and humans and have shown that its expression is restricted to embryonic liver, megakaryocytes, and platelets (12,14,25). The major transcription start site has been mapped to an adenosine 29 nucleotides upstream of the coding region (14). In the present study, we have cloned a 694-bp region 5Ј-flanking the GPVI gene and have demonstrated that it includes the regulatory regions necessary for MK-specific expression. Our promoter deletion analysis has resulted in the identification of three regions (Ϫ235 to Ϫ210, Ϫ210 to Ϫ159, and Ϫ118 to Ϫ27), which are important for the positive regulation of GPVI expression in UT-7/TPO cells. The binding of proteins present in UT-7/TPO nuclear extracts to Sp1 227 , GATA 177 , and Ets 48 sites within these three regions was determined by EMSA. As Sp1, GATA-1, and Fli-1 are expressed in MKs and platelets, we used an Sp1 antibody and recombinant GATA-1 and Fli-1 proteins to demonstrate that these proteins can bind to the GPVI Sp1 227 , GATA 177 , and Ets 48 motifs, respectively. The functional importance of these three sites was demonstrated by transient trans-fection of UT-7/TPO cells with GPVI-reporter constructs; the targeted disruption of any one of the sites reduced GPVI promoter activity. Mutation of GATA 177 decreased the GPVI promoter activity in UT-7/TPO cells by ϳ94%. The disruption of Sp1 227 and Ets 48 sites decreased the GPVI promoter activity by 40 and 80%, respectively. In addition, co-transfection of Sp1, GATA-1, and Fli-1 expression plasmids with the GPVI-reporter construct increased luciferase levels between 2-and 30-fold. We found that GPVI is expressed only in the cell lines that express all three of these factors. Furthermore, the overexpression of Fli-1 in K562 cells (which express endogenous GATA-1 and Sp1) results in the expression of the endogenous GPVI, whereas, GPVI mRNA was not detected in K562 cells in the absence of Fli-1. Taken together, these data strongly suggest that Fli-1, GATA-1, and Sp1 play a role in regulating GPVI expression. However, the role if any, of these three factors in the induction of GPVI expression by TPO is unclear as their mRNA levels are constant throughout the time course of TPO induction. It is possible that these factors are subject to posttranslational modifications or that additional factors that are regulated by TPO may bind to Fli-1, GATA-1, and/or Sp1 and augment their activity. The precedent for this is that although the TPO-responsive element within another MK-specific promoter, GPIX, has been mapped to an Ets site, and Fli-1 had been shown to bind to this site, a difference in Fli-1-binding activity before and after TPO induction could not be demonstrated (3). Future work could include the mapping of TPOresponsive elements within the GPVI promoter.
The analysis of cis-acting regulatory sequences in megakaryocyte-specific promoters has advanced our understanding of the control of differentiation and maturation in this lineage. A common theme to emerge from these studies is the presence of functional GATA and Ets elements in most MK-expressed genes (26 -29). An excellent example of the importance of GATA sites comes from the study of a patient suffering from Bernard-Soulier syndrome, a congenital bleeding disorder caused by insufficient expression of the GPIb-IX-V complex on MK and platelets (30). The cause was a mutation mapped to a GATA site in the GPIb␤ promoter, and the corresponding mutation in a GPIb␤-reporter construct was shown to reduce expression by 84%.
GATA-1 is the founding member of a family of DNA-binding zinc finger proteins that play crucial roles in cell development (for a review, see Ref. 31). Although initial studies focused on the importance of GATA-1 in the erythroid lineage, such as its involvement in the regulation of the globin genes (32), the selective knockout of GATA-1 in the megakaryocytes of mice revealed that GATA-1 also plays a key role in megakaryocyte maturation and platelet production (33). In addition, a number of GATA-1 mutations have now been identified in humans suffering from X-linked thrombocytopenia (34,35). However, the control of gene expression in megakaryocytes is not achieved through GATA-1 alone. Both cell-restricted and ubiquitous factors act in combination to enable the flexibility necessary for cell-and temporal-specific gene expression. Some of these factors, such as Fli-1, bind to distinct sequences in the regulatory regions of MK genes (17), whereas others such as FOG-1 do not appear to contact DNA directly but modulate gene expression through their interaction with other factors (36,37).
Fli-1, a member of the Ets family of winged helix-turn-helix proteins, has been implicated in the normal development of megakaryocytes. Inactivation of the Fli-1 gene in mice results in embryonic lethality because of hemorrhaging caused by vascular development abnormalities, but these mice also were found to have small, undifferentiated MK progenitors (38,39). The cooperative action of GATA-1 and Fli-1 during megakaryo-poiesis is supported by the observed similarities between the GATA-1-null and Fli-1-null megakaryocyte abnormalities, the presence of functional GATA and Ets sites in the majority of MK gene promoters, and the demonstration that these two factors can directly interact. 2 The GPVI promoter lacks a TATA box (14). Sp1 has frequently been shown to be involved in enhancing transcription of TATA-less promoters (40) and these include the ␣ 2 integrin and GPIIb gene promoters expressed in MKs (16,41). The ␣ 2 integrin, together with the ␤ 1 subunit, form another MK collagen receptor. A reduction in ␣ 2 ␤ 1 receptors on MKs and platelets is associated with two single-base polymorphisms, C Ϫ52 T and C Ϫ92 G, found in the general population (42). The T Ϫ52 and G Ϫ92 substitutions have been shown to decrease binding of Sp1 to two adjacent sites and to reduce ␣ 2 gene transcription (42,43). Interestingly, there appears to be a correlation between ␣ 2 ␤ 1 and GPVI content (22) and the involvement of Sp1 in the expression of both ␣ 2 and GPVI genes was also suggestive that these two receptors may be coordinately regulated.
Recently, we have found that Fli-1 can bind to GATA-1 and together these factors synergistically activate the expression of two other MK genes GPIX and GPIb␣. 2 Sp1 is known to physically interact with GATA-1 (44). Therefore, it is possible that Sp1, GATA-1, and Fli-1 binding to the Sp1, GATA, and Ets sites within the GPVI promoter are able interact directly with each other, and these interactions may be important for the correct regulation of the GPVI promoter. For example, physical interaction between GATA-1 and Fli-1 may be important for the transcriptional synergy we observe between GATA-1 and Fli-1 on the GPVI promoter. MK-specific expression of GPVI may then be achieved through the presence of a distinct combination of cis-acting sites, the overlapping expression profiles of Sp1, GATA-1, and Fli-1 factors, and the recruitment of multiprotein complexes involved in transcriptional activation. A better understanding of the regulation of GPVI expression adds to our knowledge of the molecular basis of MK maturation and should aid investigations into bleeding disorders and the role of collagen receptor levels as a potential risk factor for strokes and coronary heart disease. In turn, this knowledge may be exploited in the development of treatment strategies for these diseases.