Characterization of the human thrombopoietin gene promoter. A possible role of an Ets transcription factor, E4TF1/GABP.

Thrombopoietin (TPO), the ligand for c-Mpl, is a cytokine that regulates megakaryocyte growth and development. We have cloned the 5′-flanking region of the human TPO gene and analyzed its promoter activity. The human TPO gene promoter lacks a TATA box and directs transcription initiation at multiple sites over a 50-nucleotide region. Transient expression in a human liver cell line (PLC) of promoter fragment-luciferase reporter gene constructs containing a series of 5′-truncated sequences or site-directed mutations identified a sequence 5′-ACTTCCG-3′ from −69 to −63 as a positive cis-acting element for high level expression of TPO gene. This sequence contains a core motif (C/A)GGA(A/T) for Ets family proteins in the noncoding strand. Gel mobility shift assays performed with nuclear protein from PLC cells identified a DNA binding protein(s) specific for the element. Anti-E4TF1-60(GABPα) or anti-E4TF1-53/47(GABPβ) antibodies supershifted the complex in gel shift assay. Furthermore, co-expression of E4TF1-60 and E4TF1-53/47 squelched TPO gene expression in PLC and HepG2 cells. It is concluded that Ets family transcription factor E4TF1(GABPα/β), an ubiquitously expressed protein, is required for high level expression of the TPO gene in liver.

Thrombopoietin (TPO) 1 binds to and activates the c-Mpl protein, a member of the hematopoietic cytokine receptor family, and regulates proliferation and maturation of megakaryocytic cells (1)(2)(3)(4). Human TPO cDNA has recently been cloned and shown to consist of 353 amino acids containing a 21-amino acid signal peptide (1,2). The mature form of human TPO has two domains: an amino-terminal half of 153 amino acids with homology to erythropoietin and an unique 179-amino acid carboxyl-terminal domain that contains multiple potential N-gly-cosylation sites (1,2). Analysis of truncated forms of TPO showed that the amino-terminal region is sufficient to fully stimulate c-Mpl (1,2).
Analysis of circulating TPO levels in thrombocytopenic animals revealed that TPO is rapidly destroyed when the platelet counts are high, but its turnover becomes slower when the platelet levels are low, raising the possibility that the TPO level is regulated by platelet mass (5). More recently, Fiedler et al. (6) reported that the level of TPO gene expression remains constant, but the level of TPO in circulation was determined by the rate of Mpl-mediated uptake and destruction of TPO by platelets in both normal and mutant mice (c-mplϪ/Ϫ). On the other hand, TPO mRNA levels in bone marrow cells were inversely related to platelet count while hepatic or renal TPO message were unaltered (7). Bone marrow cells rather than liver or kidney were considered to be a more physiologically relevant source of TPO (7). However, Northern blot and reverse transcriptase-PCR analysis indicated that TPO was expressed mainly in liver and also slightly in many tissues such as kidney, smooth muscle, spleen, and bone marrow. Thus, further work is required to assess the significance of TPO gene expression in various tissues, including liver, for regulating the circulating level of TPO.
The human TPO gene locus has been mapped to chromosome 3q26 or q27, a site where a number of chromosomal abnormalities associated with elevated platelet counts and increased bone marrow megakaryocytes were reported in patients with acute nonlymphocytic leukemia (8 -11). The structure of the TPO gene has been reported by four groups (12)(13)(14)(15). Three of them reported that it consists of 6 exons and 5 introns (12,13,15), while Chang et al. (14) found 7 exons and 6 introns. Furthermore, these reports suggest different transcription initiation sites. Thus, the structure of 5Ј-flanking region of the human TPO gene, especially that of the promoter region, is still controversial.
In this article, we report the structure and functional characterization of the human TPO gene promoter. Our data demonstrate that the human TPO gene has multiple transcription initiation sites and lacks a putative TATA box near the initiation sites. We identified a strong cis-acting ets motif, ACT-TCCG, at Ϫ69 to Ϫ63 from the first initiation site for an optimal expression of the human TPO gene. Furthermore, E4TF1/GABP/NRF-2 (16 -18), an ubiquitously expressed Ets transcription factor, was shown to be involved in the stimulated expression of the gene. The functional significance of the element and its trans-acting factor in TPO gene expression is discussed.

MATERIALS AND METHODS
Cell Lines-The two human hepatocarcinoma cells, PLC and HepG2, and the human cervical carcinoma cell HeLa cell line used in this study were originally obtained from the American Type Culture Collection. * This work was supported in part by grants from the Ministry of Education, Science, Sports and Culture of Japan (to S. K. and N. H.) and the Ministry of Health and Welfare of Japan (to N. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Isolation of Human TPO Genomic Clone and Sequencing of the 5Ј-Flanking Region-A human TPO cDNA from 651 to 1070 nucleotides (2) was amplified by reverse transcriptase-PCR using a total RNA of human adult liver with a sense primer (5Ј-AATGCCATCTTCCT-GAGCTT-3Ј) and an antisense primer (5Ј-CCTGTGTCTGATGTTC-CTGA-3Ј), and used as a probe. Approximately 1 ϫ 10 5 plaques of the human genomic library in the EMBL3 vector, a gift from Dr. Fukumaki in Kyushu University, were screened using an ECL labeling and detection kit (Amersham Corp.). One positive clone containing the 5Ј-flanking region of TPO gene was obtained, subcloned into the pUC19 plasmid, and sequenced by the dideoxy sequence method (19) using a SQ-5500 DNA sequencer (Hitachi).
Northern Blot Analysis-Poly(A) ϩ RNAs from PLC, HepG2, and HeLa cells were obtained using a micro mRNA purification kit (Pharmacia Biotech Inc.). 10 g of poly(A) ϩ RNA were separated on a 1% formaldehyde-agarose gel and transferred to a Hybond N ϩ membrane (Amersham Corp.). The membrane was hybridized with a TPO cDNA probe in 5 ϫ SSC, 1% SDS, 50% formamide at 42°C overnight. The membrane was washed twice with 1 ϫ SSC, 0.1% SDS at room temperature for 15 min, twice with 0.1 ϫ SSC, 0.1% SDS at 65°C for 5 min and then subjected to autoradiography at Ϫ80°C for 24 h. The cDNA fragment from 651 to 1070 was radiolabeled with [␣-32 P]dCTP (6000 Ci/mmol), using a megaprime labeling kit (Amersham Corp.), and was used as a probe.
Determination of the Transcription Initiation Sites-Poly(A) ϩ RNAs from human adult liver, PLC cells, and HepG2 cells were analyzed to determine the transcription initiation sites using a rapid amplification of 5Ј-cDNA end (5Ј-RACE) method and an RNase protection assay. The 5Ј-RACE analysis was performed using a Marathon cDNA amplification kit (CLONTECH). After the first-strand cDNA was synthesized by incubating 1 g of mRNA with an oligo(dT) primer and Moloney murine leukemia virus reverse transcriptase at 45°C for 60 min, the secondstrand cDNA was synthesized with a mixture of Escherichia coli DNA polymerase I, RNase H, and E. coli DNA ligase (20). After blunt-ending with T4 DNA polymerase, the double-stranded cDNA was ligated to the Marathon cDNA adaptor. A PCR reaction was then performed with a gene-specific internal antisense primer T1 (5Ј-TGTGAAGGACATGG-GAGTCA-3Ј), which was complementary to nucleotides 329 -348 of the TPO cDNA in exon 3 (1,13), and the Marathon adaptor sense primer, AP1 (5Ј-CCATCCTAATACGACTCACTATAGGGC-3Ј), followed by nested PCR using the nested gene-specific antisense primer, T2 (5Ј-TTGCACTTCTGGGCAGAGTA-3Ј), which was complementary to nucleotides 109 -128 of the cDNA in exon 2 (1,13), and the nested adaptor primer, AP2 (5Ј-ACTCACTATAGGGCTCGAGCGGC-3Ј). The amplified products were then subcloned into the pGEM-T vector (Promega), and a total of 31 independent clones were sequenced. For the RNase protection assay, the genomic fragment from Ϫ239 to ϩ100 was amplified by PCR and subcloned into the plasmid pGEM4Z (Promega). The plasmid was linearized by EcoRI digestion, and the antisense cRNA probe was generated with T7 RNA polymerase using a mixture of 1 mM each ATP, CTP, and UTP, and 10 M [␣-32 P]GTP (3000 Ci/mmol). After 1 g of cRNA product and 20 g of poly(A) ϩ RNAs were mixed, heated at 80°C for 10 min, and hybridized in 30 l of hybridization buffer of 40 mM PIPES, pH 6.4, 400 mM NaCl, 1 mM EDTA, and 80% formamide at 45°C overnight, the mixture was digested by RNase A (10 g/ml) and RNase T1 (0.5 g/ml, Sigma) at 30°C for 30 min. The protected RNA fragments were recovered, resolved on a 6% sequencing gel, and subjected to autoradiography at Ϫ80°C for 24 h.
Construction of the Reporter Gene Plasmids-Human TPO promoterluciferase reporter plasmids were constructed by cloning various lengths of the 5Ј-flanking sequence of the TPO gene into the multicloning sites of pLUC-Basic (Promega). The plasmid pUC19-T5500, containing a 5.5-kb fragment of the 5Ј-flanking region, was used as a template for PCR to generate fragments containing from Ϫ1318 to ϩ100, from Ϫ107 to ϩ100, from Ϫ88 to ϩ100, and from Ϫ58 to ϩ100 of the TPO gene, respectively. Briefly, the required fragments of the human TPO gene were amplified by using primer sets containing KpnI and BglII sites at each end, and the resulting fragments were subcloned into KpnI and BglII sites of the pLUC-Basic to obtain pLUC-T1318, pLUC-T107, pLUC-T88, and pLUC-T58, respectively. pLUC-T39 containing Ϫ39 to ϩ100 fragment was obtained by subcloning SacI/BglII fragment of pUC19-T5500 into the SacI/BglII sites of pLUC-Basic. For the plasmids obtained by using PCR, the sequence of each clone was confirmed. The plasmid pLUC-T735 was obtained after self-ligation of the large fragment generated by digestion of pLUC-T1318 with SmaI and PvuII, and pLUC-T397 was obtained by self-ligation of the large fragment after pLUC-T1318 was digested with SmaI and AvrII and treated by the Klenow fragment to make the AvrII site blunt-ended. pLUC-T158 was constructed by subcloning a 258-bp fragment of BamHI and HindIII digestion of pLUC-T1318 into BglII/HindIII sites of pLUC-Basic. Plasmid pLUC-T4063 was constructed as follows. First, a 5500-bp HindIII/ EcoRI fragment from pUC19-T5500 was subcloned into the HindIII and EcoRI sites of pLUC-Basic, and the resulting pLUC-T5500 was completely digested by EcoRI and then partially digested by AvrII to obtain a 8800-bp fragment. pLUC-T735 described above was also digested with EcoRI and AvrII to obtain a 800-bp fragment. Finally, both 8800-and 800-bp fragments were ligated to produce pLUC-T4063.
Construction of the Reporter Plasmids Harboring the Site-directed Mutations-Plasmids containing the mutations, as defined in the figure legends, were constructed using the plasmid pLUC-T158 as the matrix by an overlap extension PCR protocol (21). Briefly, two separate PCR products, one for each half of the hybrid product, were generated with either an antisense-or a sense-mutated oligonucleotide and one outside primer. The two products were mixed, and a second PCR was then performed using the two outside primers. The product was digested with BamHI and HindIII and ligated into BglII/HindIII sites of pLUC-Basic. The inserts were sequenced to confirm the intended mutations.
Transient Expression of Human TPO-Luciferase Gene and Assay of Luciferase and ␤-Galactosidase Activities-All the plasmids of pLUC-TPO fusion genes and pCI-␤-galactosidase gene that was used for internal control of transfection were purified using the QIAGEN plasmid purification kit (QIAGEN, Inc.) and at least two different batches of plasmids for each construct were tested for transfections. Cells were plated on a 35-mm dish at a density adjusted so that they reached 70 -80% confluence (approximately 3 ϫ 10 5 cells) prior to transfection. 4 g of each plasmid DNA in 100 l of 0.3 M NaCl and 20 g of cationic liposome (Promega) in 100 l of H 2 O were mixed and added to the cells. After 60 min of incubation at 37°C, the transfection medium was replaced by 2 ml of complete medium containing 10% fetal calf serum. After an additional 40-h culture, the cells were lysed by 150 l of Picagene reporter lysis buffer LC␤ (Toyoinki) and centrifuged at 10,000 ϫ g for 10 min to remove cell debris. The supernatants were assayed for luciferase and ␤-galactosidase activities using the luciferase assay and ␤-galactosidase assay systems (Promega), respectively. The luciferase activities were normalized to those of ␤-galactosidase and expressed as the percentage of that of a plasmid as described in the figure legends.
Preparation of Nuclear Extracts, in Vitro Translated Proteins, and Gel Mobility Shift Assay-PLC and HeLa cell nuclear extracts were prepared from 1 ϫ 10 8 cells according to the method of Dignam et al. (22). In vitro translated proteins were prepared by using 1 g of pET3d plasmids containing each cDNA for E4TF1-60, E4TF1-53, or E4TF1-47 (23) in the TNT T7-coupled reticulocyte lysate system (Promega). For gel mobility shift assay, oligonucleotides were labeled with [␥-32 P]ATP (6000 Ci/mmol) and T4 polynucleotide kinase. Approximately 1 ng of the labeled oligonucleotide (10000 cpm) was mixed with 2 l (5 g of protein) of nuclear extract or 1 l of in vitro translated proteins in a total of 20 l of binding buffer (15 mM HEPES, pH 7.9, 50 mM KCl, 15 mM NaCl, 4 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 10 units/ml Trasylol) containing 1 g of poly(dI⅐dC). After incubation at room temperature for 20 min, the reaction mixture was separated on a 5% nondenaturing polyacrylamide gel with 0.5 ϫ TBE buffer (40 mM Tris borate, 1 mM EDTA). The gel was dried and subjected to autoradiography at Ϫ80°C for 24 h. For supershift experiments, 1 l of control antibody or 0.5 l of polyclonal rabbit antibody against either E4TF1-60 or E4TF1-53/47 was added to the reaction mixture 20 min after the addition of probe, and the mixture was further incubated for 20 min at room temperature.

RESULTS
Cloning and Sequencing of the 5Ј-Flanking Region of the Human TPO Gene-The human TPO genomic clone EMBL3-T1 was isolated as described under "Materials and Methods." The clone contained a 15-kb insert containing 5 kb of the 5Ј-flanking region (Fig. 1A). Southern blot analysis of human genomic DNA with a DNA fragment from this clone revealed that restriction fragments obtained by HindIII and BamHI digests were identical in size to those in the clone, indicating that no major DNA rearrangements occurred during isolation of the clone (data not shown). The 4876 nucleotides of the 5Ј-flanking region and the first exon were sequenced on both strands, and a portion of the sequence that encompasses transcription initiation site(s) is shown in Fig. 1B (see also Fig. 2).
Northern Blot Analysis-Northern blot analysis of 5 g of poly(A) ϩ RNA from PLC, HepG2, and HeLa cells showed the presence of a 1.8-kb single band in both PLC and HepG2 mRNAs, but no band was detectable in HeLa mRNA, even after a 3-fold longer exposure (data not shown).
Determination of the Transcription Initiation Site-To determine the transcription initiation sites of the human TPO gene, both 5Ј-RACE and RNase protection assays were performed as described under "Materials and Methods." As shown in Fig. 2, A and C, 5Ј-RACE assay of human adult liver cDNA revealed a total of six initiation sites from ϩ1 to ϩ51 nucleotides relative to the first initiation site. Among 31 clones analyzed, five clones began at ϩ1, one clone at ϩ8, two clones at ϩ11, 14 clones at ϩ24, seven clones at ϩ25, and the other two clones at ϩ51. Thus, the initiation sites are clustered between at ϩ1 and ϩ51, with the major sites at ϩ24 and ϩ25. RNase protection assays were also performed using a cRNA probe from Ϫ239 to ϩ100 nucleotides of the human TPO gene. The results showed three major RNase-protected bands at ϩ24, ϩ25, and ϩ51 (Fig. 2, B and C), and their positions matched those of the most prominent sites determined by 5Ј-RACE assay. Similar results were obtained when mRNAs from HepG2 or PLC cells were analyzed, indicating that the same transcription initiation sites as in adult liver are employed in these two cell lines (data not shown). A putative TATA box was not observed in any promoter region of the multi-initiation sites (see Fig. 1B). Chang et al. (14) recently reported the existence of exon 0 at a more 5Ј-upstream region than shown in Fig. 1. If that were the case, a protected band at Ϫ189 must be detected in our RNase protection assay, since their exon 2, which is from Ϫ189 to ϩ114, should be protected by our cRNA probe of Ϫ239 to ϩ100.
However, as shown in Fig. 2B, no such protected band at the Ϫ189 position (indicated by an arrow) was observed. Thus, the significance of exon 0 described by Chang et al. could not be assessed in our assay.
Transient Expression Analysis of the 5Ј-Flanking Region-To localize the region essential for human TPO gene expression, a series of 5Ј-to 3Ј-deletion mutants were generated and transiently expressed in PLC cells (Fig. 3A). Since PLC cells expressed the TPO gene and employed the same transcription initiation sites as adult human liver (Fig. 2), we used the PLC cell line to assay the promoter activity. Furthermore, it was demonstrated that the transcription of the reporter gene also initiated from several start sites that were essentially the same as in endogenous TPO gene (data not shown). As shown in Fig.  3B, all the plasmids containing 5Ј deletions of various lengths from Ϫ4063 to Ϫ88 promoted high level expression of luciferase activity that was approximately 14 -25 times higher than the background of pLUC-Basic. This indicated the motifs observed in the region from Ϫ4063 to Ϫ88 had no significant regulatory effect, while pLUC-T158 reproducibly showed a 1.3-2-fold higher activity than the other constructs. This might suggest the existence of an inhibitory element(s) from Ϫ397 to Ϫ158 and an enhancing element(s) from Ϫ158 to Ϫ107. In contrast, a further deletion to Ϫ58 resulted in an extreme reduction of activity to about 13% of that of pLUC-T88. The results clearly indicate that a positive regulatory element(s) is located from Ϫ88 to Ϫ58, and this region is essential for optimal transcription of the TPO gene. The pLUC-T39 plasmid still showed as much activity as pLUC-T58, 2.6 times higher than pLUC-Basic. We next transfected the deletion mutants into HeLa cells, which do not express the TPO gene. Fig. 3C shows that HeLa cells did express the pLUC-T58 reporter gene to an equivalent level to that in PLC cells, indicating that the 3Јsequence from Ϫ58 functions as a minimal promoter in both PLC and HeLa cells. However, the proximal cis-element(s) from Ϫ88 to Ϫ58 increased the expression only 1.8-fold in HeLa cells compared with a significant activation in PLC cells (6.2fold). This result suggests that the promoter of the TPO gene can function in HeLa cells but that the cis-element from Ϫ88 to Ϫ58 supports tissue-specific activation only in PLC cells.
Mutational Analysis of the Proximal Cis-regulatory Region of the TPO Gene-We next introduced clustered mutations into the plasmid pLUC-T158. As shown in Fig. 4, the activities of mutants, M5, 6, 8, 9, and 10, all showed 75-80% reduction as compared with that of pLUC-T158, and M3, 7, and 11 exhibited small deviations or changes from the control, while M4 showed a slightly enhanced activity (50%). The M5 mutations introduced into the longer 5Ј sequence, pLUC-T1318-M5 and pLUC-T4063-M5, also reduced the activities. The data clearly indicate that the sequence ACTTCCG from Ϫ69 to Ϫ63 functions as a major positive cis-element for the efficient expression of the TPO gene. It was also noted that ACTTCCG contains a core sequence of ets motif (C/A)GGA(A/T) in the noncoding strand (24,25). Other mutations, M1, M2, and M14, that were directed toward a GGGTG (CACCC) element at Ϫ123 to Ϫ119 (26), a GGGAAG (ets-like) element at Ϫ85 to Ϫ80 (24,25), and a ATTGG(CCAAT) element at Ϫ26 to Ϫ22 (27), respectively, exhibited almost comparable activity to that of the wild type. It is indicated that these elements do not play significant regulatory roles in TPO gene expression.
Gel Mobility Shift Assay-The nuclear protein factor(s) specific for binding to the cis-element from Ϫ69 to Ϫ63 was analyzed by gel mobility shift assay using wild type or mutated oligonucleotides (Fig. 5A). As shown in Fig. 5B, lane 2, the nuclear proteins from PLC cells produced three bands, a prominent band I and two faint complexes, II and III, of higher mobilities. Band I was shown to contain two bands of close mobilities on short exposure of the gel. The specificity of these bands for the sequence was shown by a competition experiment in which a 50 times molar excess of cold specific oligonucleotide completely abolished the formation of band III (lane 3) while partially preventing the formation of bands I and II. In contrast, the same molar excess of mutated oligonucleotides failed to compete (lanes 6 -8). Among a series of mutated oligonucleotides, M3, 4, 11, and 13, which had little effect on expression, efficiently competed with the wild type sequence (lanes 4, 5, 9, and 10), while M11 competed less (lane 9). In contrast, M5, 9, and 10, which markedly reduced expression, had little effect on complex formation (lanes 6 -8). HeLa cell nuclear proteins also yielded three bands that showed similar mobilities and sequence specificities to those in PLC cells (lanes 11-19), indicating that the factor(s) is also present in cells that do not express TPO. Since the cis-element of the TPO gene contains a core sequence, -CGGAA-, recognized by Ets domains, oligonucleotides from the stromelysin gene promoter (Ϫ209/Ϫ187, STROM Ϫ209/Ϫ187) (24, 28) and cytochrome oxidase subunit IV gene promoter (ϩ13/ϩ36, RCO4 ϩ13/ϩ36) (25,29) corresponding to DNA-binding sites for Ets-2 and E4TF1/GABP, respectively, were tested for competition. Both STROM and RCO4 sequences strongly inhibited formation of bands I, II, and III, while inhibition by RCO4 was more marked than by STROM under conditions where the control sequence had no effect (Fig. 6). Taken together, these data indicate that (i) the DNA-binding protein(s) that recognizes the cis-element of the TPO gene is present in liver as well as HeLa cells and (ii) shares DNA sequence specificity with E4TF1/GABP or Ets-2, putative Ets family proteins (18, 24, 25, 28 -31).

E4TF1 Binds to the ets Consensus Site of the TPO Gene
Promoter-The effect of antibodies against two subunits of E4TF1, E4TF1-60, or E4TF1-53/47, on complex formation was investigated. Fig. 7A showed that an antibody against E4TF1-60 sharply reduced the formation of bands I, II, and III, and generated a supershifted complex (shown by an arrow) in both PLC and HeLa cell nuclear extracts (lanes 2 and 6). Anti-albumin antibody, used as a control, had no effect on complex formation under these conditions (lanes 4 and 8). In addition, anti-E4TF1-60 antibody yielded no band when incubated with the probe in the absence of nuclear extract (lane 9). Thus, the effect of anti-E4TF1-60 antibody was specific, and bands I, II, and III all contained E4TF1-60 protein as a component. Anti-E4TF1-53/47 antibody, in contrast, slightly reduced the intensity of band I with generation of a band of lower mobility (indicated by an arrowhead in lanes 3 and 7). The intensities of bands II and III remained almost unchanged (lanes 3 and 7). As shown in lane 10, incubation of the antibody FIG. 2. Identification of the transcription initiation sites of the human TPO gene. A, 5Ј-RACE analysis of mRNA from human adult liver. The products obtained by 5Ј-RACE reaction were subcloned into pGEM-T vector, and a total of 31 clones were sequenced as described under "Materials and Methods." Sequences of fusion with adaptor of four independent clones were shown. Arrows indicate the positions of the 5Ј ends of the cDNA obtained. B, RNase protection analysis. 20 g of poly(A) ϩ RNAs from human adult liver, HepG2 cells, PLC cells, or yeast tRNA were hybridized with a cRNA probe spanning the nucleotides from Ϫ239 to ϩ100. After treatment with RNase, the protected products were separated on a 6% sequencing gel. Arrows at ϩ24, ϩ25, and ϩ51 indicate the protected products. Arrow at Ϫ189 represents the position of expected band when the TPO gene transcription initiated at exon 0 of Chang et al. (14). C, summary of the transcription initiation sites. Arrowheads and arrows indicate the positions of the initiation sites determined by 5Ј-RACE and RNase protection analysis, respectively. The number of clones obtained by 5Ј-RACE analysis were also shown below each arrowhead.
with the probe formed no band, indicating that the antibody did not react with the probe DNA. These results suggested that the band generated by anti-53/47 antibody in lanes 3 and 7 represented a specific supershifted complex of band I containing E4TF1-53/47 protein. This view was supported by the experiment shown in Fig. 7B. In this experiment, we tested in vitro translated E4TF1-60, E4TF1-53, and E4TF1-47 for binding to the ets motif of the TPO gene promoter (Fig. 7B). E4TF1-60 produced a complex c (lane 2), while neither E4TF1-53 nor E4TF1-47 alone bound the oligonucleotide W (lanes 3 and 4). In contrast, when mixtures of E4TF1-60 and E4TF1-53, or E4TF1-47 were tested, a prominent band a or b of lower mobilities was produced along with abolishment or reduction of band c (lanes 5-7). The mobilities of complex c, or a and b, were similar or very close to that of band III or I produced by PLC nuclear proteins, respectively (compare lanes 2, 5, 6, and 7 with lane 8). From these data, it is evident that bands III or I formed by nuclear extract represented the complex by monomeric E4TF1-60 or heteromer of E4TF1-60 and E4TF1-53/47, respectively. Band II could not be reproduced by in vitro translated subunits of E4TF1, suggesting that it might represent a complex of DNA and E4TF1-60 that has been modified in vivo.
Effects of Overexpression of E4TF1-60 and -53-Since our results indicated that the ets motif of the human TPO gene promoter is recognized by E4TF1/GABP/NRF-2, the effect of overexpression of two subunits of E4TF1, E4TF1-60 or -53 (32), on TPO promoter activity was investigated. Transfection of increasing amounts of plasmid encoding E4TF1-60, from 0.1 to 2.0 g, into PLC cells suppressed TPO gene promoter activity in a dose-dependent manner. In contrast, transfection of 0.1 or 0.5 g of plasmid for E4TF1-53 enhanced TPO promoter activity 1.1-or 1.4-fold, respectively, while higher amounts of the plasmid suppressed the activity. Co-transfection of both plasmids for E4TF1-60 and -53 suppressed the activity as did E4TF1. These data suggest that the subunits of E4TF1 can sequester some factor or factors needed for transactivation of the human TPO gene and are consistent with the possibility that both E4TF1 subunits are involved in transactivation of human the TPO gene, and the amount of E4TF1-53 might be limiting in cells compared with that of E4TF1-60. Similar results were obtained using HepG2 cells (white columns in Fig.  8). Squelching by overexpressed Ets protein has also been observed in insulin-mediated prolactin gene expression (33). DISCUSSION The present analysis of the 5Ј-flanking region of the human TPO gene has clarified several features of the mechanisms that regulate its expression. Our analysis of the transcription initiation site of the TPO gene in human adult liver revealed multiple initiation sites with the majority at ϩ24 and ϩ25, rather than at Ϫ20 or ϩ48 as previously reported. The sites were essentially identical in both 5Ј-RACE and RNase protec- In B or C, 2 g each of chimeric TPO-luciferase vector and 2 g of PCI-␤-galactosidase vector were transfected into PLC cells (B) or HeLa cells (C), respectively, and the luciferase activity was measured as described under "Materials and Methods." After normalization to ␤-galactosidase activity, the relative luciferase activity of each construct was expressed as a percentage of that of pLUC-T158. The data represent the mean of three independent experiments with a S.E. bar. tion assays and were found to be also utilized in the liver cell lines PLC and HepG2. More recently, Chang et al. (14) reported the existence of one more exon (exon 0) from Ϫ1498 to Ϫ1397 with a transcription initiation site at Ϫ1498. Although their initiation site could not be detected in our RNase protection analysis, reverse transcriptase-PCR of liver mRNA, performed using primers flanking the region from exon 0 and exon 3, did amplify a band of the expected length. But that amount was significantly lower than that of the product amplified using exon 1 primers (data not shown). Thus, exon 0 detected in the study of Chang et al. might be utilized in the adult liver as one of several initiation sites but considerably less frequently than the other sites reported in this study. Alternatively, it might reflect different mechanisms of TPO gene expression during liver development, because we used human adult liver whereas they used fetal liver. However, this possibility awaits further investigation.
The analysis of 5Ј-flanking regions of the TPO gene clearly demonstrated that the 7-bp sequence -ACTTCCG-from Ϫ69 to Ϫ63, was essential for the high level expression of human TPO gene in liver cells. The sequence contains a core (C/A)GGA(A/T) motif for an Ets family of transcription factors in its noncoding strand. It has been reported that Ets proteins recognize a purine-rich (C/A)GGA(A/T) motif in the middle of 10-bp nucleotides, and the flanking sequences are considered to be also involved in determining sequence specificity of each Ets protein (23-25, 34, 35). Since mutations in the flanking region of the ets consensus sequence of TPO gene, M3, 4, and 11, also exerted a marginal effect, that is slight reduction in M3 and M4 and slight stimulation in M5, respectively, a total 10-bp sequence might be involved in the specific binding of a factor to the human TPO ets sequence.
The present study further demonstrated that a specific DNA binding protein(s) for ACTTCCG motif of the human TPO gene is highly likely to be E4TF1/GABP. This is supported by the following results. (i) The DNA-protein complex detected in gel shift assays was effectively abolished by an ets sequence for E4TF1/GABP from an RCO4 gene promoter that matches 8 out of 10 nucleotides of that of the TPO gene (24,25,29). Furthermore, there is a complete match of sequence observed between that of the human TPO gene and of the MCO5b site D, another site for E4TF1/GABP/NRF-2 binding (29). (ii) Antibodies against subunits of E4TF1, E4TF1-60 or E4TF1-53/47, specifically supershifted the complex. (iii) The combination of in vitro translated E4TF1-60 and E4TF1-53/47 produced a complex with essentially the same mobility as that produced by a PLC nuclear extract. (iv) Overexpression of E4TF1-60 and -53/47 squelched TPO promoter activity in PLC and HepG2 cells.  Fig. 4 and used as competitors. The essential sequence of the cis-element is underlined. B, interactions of nuclear proteins with the cis-element were assayed as described under "Materials and Methods." The radiolabeled W oligonucleotide was incubated with 5 g of bovine serum albumin (lane 1), 5 g of PLC cell nuclear extracts (lanes 2-10), or 5 g of HeLa cell nuclear extracts (lanes [11][12][13][14][15][16][17][18][19] in the binding reactions in the absence (Ϫ) or presence of a 50-fold molar excess of the oligonucleotides depicted on the top of each lane. Bands I, II, and III represent specific complexes. E4TF1/NRF-2 has been cloned from HeLa cells and shown to bind to ets elements in the cytochrome c oxidase subunit IV gene (16) or adenovirus early region 4 gene (17). GABP, a rat homologue of E4TF1/NRF-2, binds to an ets element in the rat HSV IE gene (18). E4TF1 contains three subunits, E4TF1-60, -53, and -47, among which E4TF1-60 specifically binds to DNA through the Ets domain in its carboxyl terminus (16,17). E4TF1-53 and -47, homologues of Drosophila Notch and Caenorhabditis elegans Glp-1 and Lin-1, respectively, associate with E4TF1-60 through four tandemly repeated Notchankyrin motifs in the amino terminus and mediate transcriptional activation through the COOH terminus (16,17). The factor has been proposed to maximally activate the promoter by binding to two or more ets sites in a tetrameric form (17,18). A construct containing one more ets sequence just upstream of the wild type ets motif of the TPO gene exhibited enhanced expression (data not shown), consistent with the idea that two ets sequence synergistically activate the TPO gene. We speculate that the human TPO gene is unique in containing only one ets element and is thus incapable of forming a more competent DNA-protein complex. This might explain why the level of TPO expression is low, even in the liver.
Although the present study clearly indicated that an ubiquitously expressed E4TF1/GABP is involved in stimulated expression of TPO gene in liver, it remains to be defined why the TPO gene is not expressed in HeLa cells that contained an FIG. 6. Competition gel mobility shift assay of the human TPO cis-element by putative ets motifs. A, sequences of oligonucleotides used in this study are shown. The W oligonucleotide is the same as in Fig. 5. The STROM Ϫ209/Ϫ187 and the RCO4 ϩ13/ϩ36 contain the Ets-1/2 binding site of rat stromelysin gene promoter (24,28), and the GABP/NRF2-binding site of rat cytochrome c oxidase subunit IV gene promoter (25,29), respectively. The control contains the sequence from Ϫ133 to Ϫ109 of human TPO gene promoter. B, the radiolabeled oligonucleotide W was incubated with 5 g of PLC cell nuclear extracts in the absence (Ϫ) or presence of 25-or 50-fold molar excess of the oligonucleotides depicted on the top of each lane and analyzed as described under "Materials and Methods." Bands I, II, and III were as in Fig. 5.   FIG. 7. E4TF1-60 and E4TF1-53/47 bind to the human TPO cis-element. A, the radiolabeled oligonucleotide W was incubated with 5 g of PLC cell nuclear extracts (lanes 1-4), or 5 g of HeLa cell nuclear extracts (lanes 5-8) in a binding mixture containing antibody against E4TF1-60 (lanes 2 and 6), E4TF1-53/47 (lanes 3 and 7), or control antibody against human albumin (lanes 4 and 8). Lanes 9,10,11,and indicated by a, b, and c, respectively. NS indicates nonspecific binding. Bands I, II, and III are as in Fig. 5 for the PLC nuclear extract. equivalent level of E4TF1/GABP to that in liver (Figs. 5B and 7A). We speculate that E4TF1/GABP is necessary but not sufficient for liver-specific expression of the TPO gene, and there might be a liver-specific mechanism of gene activation, since the element transactivated the TPO gene in liver more strongly than in HeLa cells (Fig. 3C). These could include (i) liverspecific post-translational modification of E4TF1/GABP protein, although in our in vitro footprinting experiment to determine ets site occupancy, no apparent differences were observed between liver and HeLa cells (data not shown); (ii) the presence of liver-specific mediator(s) or co-activator(s) that intervenes between E4TF1/GABP and general transcription machinery. This might be one of the likely mechanism since the TPO reporter gene containing ets motif could not be transactivated in transient expression study of HeLa cells; (iii) the difference of higher order chromatin structure of the TPO gene that is open in liver but closed in HeLa cells for the factor(s) to be recruited into the gene promoter. Alternatively, there still remains the possibility that other liver-specific as yet undefined Ets transcription factor(s) that is closely related to but different from E4TF1/GABP is involved in the liver-specific expression of TPO gene.
Finally, our results clearly demonstrated that the ets motif ACTTCCG at Ϫ69 to Ϫ63, which is recognized by E4TF1/ GABP, an ubiquitously expressed Ets protein, is essential for the high level expression of the human TPO gene in liver cells. This work provides us the first clues toward understanding the molecular mechanism of expression of the human TPO gene. More research into this question, especially the functional role of E4TF1/GABP in liver-stimulated expression and regulation of circulating TPO level in vivo, is necessary, and such an investigation is now in progress.