Expression of Gfi (growth factor-independence)-1B, a Gfi-1-related transcriptional repressor, is restricted to erythroid lineage cells and is essential for erythropoiesis. We have determined the transcription start site of the human Gfi -1B gene and located its first non-coding exon ∼7.82 kb upstream of the first coding exon. The genomic sequence preceding this first non-coding exon has been identified to be its erythroid-specific promoter region in K562 cells. Using gel-shift and chromatin immunoprecipitation (ChIP) assays, we have demonstrated that NF-Y and GATA-1 directly participate in transcriptional activation of the Gfi -1B gene in K562 cells. Ectopic expression of GATA-1 markedly stimulates the activity of the Gfi-1B promoter in a non-erythroid cell line U937. Interestingly, our results have indicated that this GATA-1-mediated trans-activation is dependent on NF-Y binding to the CCAAT site. Here we conclude that functional cooperation between GATA-1 and NF-Y contributes to erythroid-specific transcriptional activation of Gfi-1B promoter.
Received May 3, 2004; Revised and Accepted July 6, 2004
Gfi (growth factor-independence)-1 is a cellular proto-oncogene that was initially identified as a target of provirus integration in retrovirus-induced T-cell-lymphoma line selected for IL-2 independence in culture ( 1 ) and in primary retrovirus-induced lymphoma, where it collaborates with c- myc and pim -1 in lymphomagenesis ( 2 – 4 ). Gfi-1B, which encodes a Gfi-1-related zinc finger protein, was identified by low stringency hybridization screening with a partial Gfi -1 cDNA probe ( 5 ). These two proteins exhibit transcriptional repressor function with a high degree of homology in the SNAG repressor and zinc finger domains, and their DNA binding site has been defined to be TAAATCAC(A/T)GCA ( 5 – 7 ). Despite their similarity in domain function and organization, expression of Gfi-1B is confined to erythroid lineage cells and megakaryocytes in human ( 8 ), whereas Gfi-1 is more abundant in the lymphopoietic thymus ( 9 – 13 ). Overexpression of Gfi-1B in T cells leads to defects in T cell lineage commitment ( 14 ), indicating that Gfi-1 and Gfi-1B play distinct functions in hematopoiesis. Expression of Gfi-1B is now known to be essential in blood cell development, since targeted gene disruption of Gfi -1B in mice results in embryonic lethality due to a profound defect in erythroid and megakaryocyte development ( 15 ). A parallel study has shown that enforced expression of Gfi-1B in CD34+ cells isolated from human umbilical cord blood samples results in EPO-independent erythroblast formation ( 16 ), further indicating the direct function of Gfi-1B in erythroid development.
While Gfi-1B plays an important role in the biology of the hematopoietic system and erythroid progenitor cells ( 15 , 16 ), its gene regulation remains unknown due to the unavailability of its gene promoter sequence. Understanding the mechanism governing Gfi-1B expression in erythroid cells will give us the important information on how Gfi-1B is regulated in the hematopoietic system during blood cell development. In this study, using the oligo-capping method ( 17 , 18 ) we determined the sequence of the 5′-untranslated region (5′-UTR) of Gfi-1B mRNA prepared from K562 cells, in which the expressed level of Gfi-1B mRNA is elevated ( 8 ). We mapped this sequence onto the human genome and found that the transcription start site of the Gfi -1B gene is 7.82 kb upstream of the first coding exon containing the translation initiation ATG codon. Using RNase protection assays, we confirmed that transcripts from K562 cells contain the 5′-UTR sequence covering the non-coding exon in conjunction with the first coding exon.
We demonstrated that the 5′-adjacent sequences spanning from −145 relative to this transcription start site functioned as the promoter region and conferred cell-type-specific transcriptional activity of the Gfi -1B gene in K562 cells. We further showed that GATA-1 is an essential and direct activator for the Gfi-1B promoter in K562 cells. In this study, it is interesting to note that the transcription factor NF-Y binding to the CCAAT site within the Gfi-1B promoter plays a critical role in GATA-1 responsiveness. In summary, we have identified this novel promoter and provided evidence that cooperation between GATA-1 and NF-Y is necessary for erythroid-specific transcriptional activation of the Gfi-1B promoter.
MATERIALS AND METHODS
U937, K562 and D2 cells were maintained in RPMI 1640 (Invitrogen Life Technologies) supplemented with 10% heat-inactivated fetal bovine serum (HyClone), 2 mM l -glutamine, 100 U/ml of penicillin G and 100 U/ml of streptomycin. NIH 3T3 fibroblasts were grown in DMEM containing 10% calf serum.
5′ End oligo-capping of total RNA
Oligo-capping was performed as described previously ( 17 ). In brief, 100 μg of total RNA from K562 was treated with 2.5 U of bacterial alkaline phosphatase (BAP) (Takara), followed by phenol/chloroform extraction and subsequently treated with tobacco acid pyrophosphatase (TAP) (Wako, Japan). The TAP-treated RNAs were then subjected to an overnight ligation reaction with 0.4 μg of a 30 nt long oligoribonucleotide tag (5′-AGCAUCGAGUCGGCCUUGUUGGCCUACUGG-3′) using T4 RNA ligase (Takara). Following extraction with phenol/chloroform and ethanol precipitation, the resulting oligo-capped RNA was used for the first strand cDNA synthesis by reverse transcription.
5′ End amplification of Gfi-1B cDNA
The 5′ end region of Gfi -1B cDNA was amplified by incubating the first strand cDNA made from oligo-capped RNA using an XL PCR kit (Perkin-Elmer) with the oligo-cap-specific primer (5′-AGCATCGAGTCGGCCTTGTTG-3′) and one of the Gfi -1B gene-specific primers of 28 nt in length (5′-GAAGTCCAAGGCGGGCTCAGTGCTGGGC-3′). The second nested PCR was carried out by using the first-round PCR product as template and a pair of primers, the same oligo-cap primer used for the first-round PCR, and another 28 nt long Gfi -1B gene-specific primer (5′-CGGGCTGGTGGTAGGTGTGAGCCTTCT-3′) located 5′ downstream of the primer used for the first-round PCR. PCR products were inserted into pGEM-T Easy vector (Promega). DNA sequences of the PCR inserts were determined by the dideoxy termination method. The plasmid containing this sequence was then designated as a 5′ oligo-capping +1/+207 plasmid.
Amplification of 3′ end of Gfi-1B cDNA
The 3′ rapid amplification of cDNA ends was performed according to the manufacturer's protocol using a SMART TM RACE cDNA Amplification Kit (Clontech) using a Gfi -1B gene-specific primer (5′-GATCAAGCTTGTCCTTAGCACTCTATTC-3′). PCR products were inserted into pGEM-T Easy vector (Promega). Their DNA sequences were determined by the dideoxy termination method.
Preparation of nuclear extracts and gel-shift analysis
Nuclear extracts were prepared from K562 and U937 cells as previously described ( 19 ). The gel-shift reaction each contained 0.025 pmol of radiolabeled probe, 0.5 μg of poly(dI–dC) and 10 μg of nuclear protein in a buffer of 50 mM NaCl, 30 mM KCl, 10 mM Tris–HCl, pH 7.5, 1 mM EDTA, 1 mM DTT and 10% glycerol. Competitive DNA oligomer, GATA-1 (N6), NF-YA (H-209) or NF-YB antibody (C-20) (Santa Cruz, CA) were added to the nuclear extract for 30 min on ice prior to the DNA binding reaction. After the DNA binding reaction at room temperature for 25 min, samples were analyzed by electrophoresis at 150 V for 2.5 h through non-denaturing 4% polyacrylamide gels. Gels were then dried for autoradiography.
RNase protection assays
A genomic DNA fragment obtained from PCR covering the −87/+133 region was inserted into the pGEM-T Easy vector. The resultant plasmid was digested by BspEI and SphI to obtain a DNA fragment, which was then inserted to the BspEI/SphI site of the 5′ oligo-capping cDNA plasmid encompassing the +1/+207 sequence. In this way, the +1/+123 sequence was replaced by the −87/+123 sequence after insertion to generate a plasmid containing the −87/+207 region of the hGfi-1B gene sequence. The −87/+207 and +1/+207 plasmids were linearized by SphI prior to transcription for the generation of probes A and B, respectively. The plasmids containing +53/+248 of the human Gfi-1B cDNA and pTRI-β-actin-human (Ambion) control template were linearized by HindIII and NcoI, respectively. All the transcription reaction mixtures contain SP6 RNA polymerase (Promega) and [α- 32 P]CTP for the synthesis of Gfi-1B and β-actin riboprobes. Total RNAs or mRNA were isolated and hybridized to the riboprobes at 42°C overnight. The RNase protection assay (RPA) was performed according to the manufacturer's protocol using an RPA kit (RPA III; Ambion, Austin, TX). The protected RNA products were separated on a 6% denaturing polacrylamide gel followed by autoradiography.
Plasmid constructs and site-directed mutagenesis
The DNA fragment containing −1000/+19, −325/+19 or −145/+19 sequence corresponding to the 5′-flanking of Gfi -1B genomic DNA was amplified using the primers pair A + B, C + D or D + E, (A, 5′-GATGC GCTAGC CAAATGGGAAGAAATGCC-3′; B, 5′-GCGC CTCGAG ATAGATACTTCTCCTTTTTG-3′; C, 5′-AGCTA CTCGAG TTTTATAAGTTAGAG-3′; D, 5′-GAGCT AAGCTT AGATACTTCTCCTTTTTGC-3′; E, 5′-GATC CTCGAG CCTGGAAAGTTTTGATAAGC-3′; underline indicates the NheI, XhoI and HindIII digested sites). The amplified DNA fragments were accordingly digested with NheI + XhoI or XhoI + HindIII for the subsequent insertion at the site upstream of the luciferase reporter vector, pGL2-Basic (Promega) in a sense direction to obtain pGL2-hG (−1000/+19), pGL2-hG (−325/+19), pGL2-hG (−145/+19) plasmids. The pGL2-hG (−115/+19) plasmid was constructed by digesting pGL2-hG (−325/+19) with SacI followed by self-ligation. The GATA-1 mutant, pCMV-GATA-1 (C258G) carrying point mutations in the C-terminal zinc finger of GATA-1, was created using PCR-directed mutagenesis. Briefly, the mutated GATA-1 cDNAs were subcloned in-frame into the BamHI site of pCMV-Tag 2B vector (Stratagene). The resulting construct pCMV-GATA-1 (C258G) carried a mutation of the cysteine residue (TGC) to glycine ( G GC) at the amino acid 258 in the carboxyl zinc finger domain. Mutant types of pGL2-hG (−145/+19) were generated using Quick-Change Site-directed Mutagenesis Kit (Stratagene) by specific mutated primers. The following nucleotide changes were made: the GATA binding site at −132/−127 (G1) GATAAG to CT T CTA , the −96/−92 CCAAT to G C T AT, the +13/+16 GATA site (G2) to CT T C . The underlined letter indicates the mutated nucleotide. All constructs were verified by dideoxy termination sequencing. The NF-YA DNA-binding mutant ( 20 ) was a gift from Mantovani, R. (Dipartimento di Scienze Biomolecolari e Biotechnologiche, Università di Milano, Via Celoria 26, 20133 Milano, Italy).
Chromatin immunoprecipitation was performed as described ( 21 , 22 ). K562 and U937 cells (1 × 10 7 ) were fixed with formaldehyde at a final concentration of 1% for 20 min at room temperature. Glycine was added at a final concentration of 125 mM to quench crosslinks. Cells were then lysed with RIPA buffer [10 mM Tris–HCl, pH 7.8, 140 mM NaCl, 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, 1 mM PMSF and protease inhibitors (Sigma)]. Chromatin was sheared by sonication with the Ultraschallprozesser UP50H Sonifier to obtain the size of DNA fragments <1 kb. The solution was clarified by centrifugation at 10 000 r.p.m. at 4°C for 10 min. After pre-adsorbing the solution with protein A Sepharose beads (Amersham) for 1 h at 4°C, the pre-cleared extracts were immunoprecipitated with 2 μg of anti-GATA-1 (N6) or anti-NF-YA (H-209) antibody (Santa Cruz, CA) on ice for 3 h, followed by incubation with protein A beads on a rotating wheel at 4°C for 1 h. After extensive washing, the immunocomplex beads were eluted with 150 μl of 1% SDS in 50 mM NaHCO 3 solution for 15 min at room temperature. After heating at 65°C overnight, DNA was purified and resuspended in 30 μl 1× TE buffer, in which 2 μl of immunoprecipitated DNA was applied to PCR with the following thermal cycling program: 94°C for 5 min, 30 cycles of 30 s at 94°C, 30 s at 52°C and 1 min at 72°C, followed by a 5 min extension time at 72°C. The plasmid pGL2-hG (−325/+19) was used as a positive control for the PCR reaction. Of the products, 30% were separated by 10% PAGE and visualized by ethidium bromide staining. The primer sequences were specific primers C + D described in the section on plasmid construction for hGfi-1B; sense primer 5′-CAGGCCCCTGTCGGGGTCAG-3′ and antisense primer 5′-GATGCCGTTTGAGGTGGAGTTTTAG-3′ for the HS2 region of β-globin control locus; sense primer 5′-CCTCACCCTGTGGAGCCACA-3′ and antisense primer 5′-GAGGTTGCTAGTGAACACAGTTGTG-3′ for the β-globin promoter region; sense primer 5′-GGATTCCTCCCACGAGGGGGCGGGCT-3′ and antisense primer 5′-AGCCCCTGGTTCCCGCGCCGACCGCT-3′ for the human thymidine kinase (hTK) gene. The lengths of PCR products were 345 bp for hGfi-1B gene, 250 bp for the HS2 region of β-globin control locus, 147 bp for the β-globin promoter region and 140 bp for hTK gene.
Transient transfection and luciferase assays
We used electroporation for K562 and U937 cell transfection; 5 × 10 6 cells were suspended in 0.3 ml of RPMI 1640 medium containing 15 μg of reporter plasmid, 15 μg of each expression vector or control empty vectors and 2 μg of an internal control plasmid (pCMV-β-Gal).
Cells were electroporated by Gene Pulser (Bio-Rad) at 300 mV for K562, or 250 mV for U937. NIH3T3 cells plated on 35 mm dishes were transiently transfected with a mixture containing 12 μg of Lipofectamine (Invitrogen Life Technologies) and 1 μg reporter plasmid, 1 μg of each expression vector or control vector and 0.1 μg of pCMV-β-Gal. After transfection for 24 h, cells were washed and lysed in a reporter lysis buffer [0.5 M HEPES (pH 7.8), 0.2% Triton X-100, 1 mM CaCl 2 , 1 mM MgCl 2 ], and 50 μl of the cell lysate was mixed with 50 μl of luciferase assay buffer (Packard). The luciferase activity was measured with a luminescence counter (Packard).
Identification of novel 5′-untranslated region in the human Gfi -1B gene
Because our initial investigation indicated that the region upstream of the initial ATG codon of the Gfi -1B gene does not possess the promoter activity, we then used the oligo-capping strategy ( 17 , 18 ) to precisely identify the transcription start site of the Gfi -1B gene. In principle, this method prevents all uncapped RNA from ligation, and generates capping RNA ligated with oligoribonucleotide for RT–PCR reaction. To know the 3′-untranslated region (3′-UTR) sequence of Gfi -1B gene, we used the 3′ RACE method for RT–PCR. The DNA sequences from the RT–PCR products of 5′ and 3′ end amplifications were assembled and full-length sequence of human Gfi -1B cDNA (accession no. AY428733). Blast search of the human genome using the +1/+133 sequence of the human Gfi -1B mRNA as the query sequence, we found that this sequence is present in a contig (accession no. 000393) located at chromosome 9q34. The genomic sequence confirms that the exon 1 of the Gfi -1B gene is situated 7.82 kb upstream of the exon 2 where the translation initiation site ATG is located. The intron/exon splice junction of exons 1 and 2 in the genomic region contains the consensus sequences for splicing. To verify the existence of the transcript containing this 5′-UTR, we used the probe of +53/+248 sequence encompassing exons 1 and 2 regions for RNase protection assay. As expected, 196 nt of this probe was protected by the RNA from K562 and another erythroblastoma cell line D2 ( Figure 1B ), suggesting the presence of this novel 5′-UTR in the Gfi-1B RNA transcript.
Since there is a Gfi-1B mRNA sequence 2.479 kb long deposited in the NCBI database (accession no. BC043371), whose 5′ end is situated at −674 nt upstream of the start site defined here, we further tested whether the oligo-capping strategy does give a true transcription initiation site by RNase protection assays using RNA from K562 cells hybridized with different riboprobes with or without a promoter sequence. To this end, we constructed a plasmid, in which a genomic DNA fragment covering the 5′ upstream and exon 1 region (−87/+133) was fused to exon 2 (+134/+207), to generate probe A containing the −87/+207 sequence. Probe B was generated from a plasmid of the 5′ oligo-capping cDNA clone encompassing the +1/+207 sequence. If the transcript start sites were much longer than what we defined here, the size of the major protected fragment detected after hybridization with probe A should be ∼294 nt. However, it turned out that the size of the major protected fragments detected after hybridization with various amounts of RNA isolated from K562 cells were essentially the same 207 nt in the parallel experiments using these two probes ( Figure 1C ). This result clearly demonstrated that the region we defined here represents a true transcription start site of the Gfi-1B gene in K562 cells.
Identification of the human Gfi-1B promoter
The genomic DNA sequences preceding this exon 1 region were analyzed by MatInspector. The result showed that this proximal promoter sequence lacks a TATA-like element but contains an initiator element (INR) site at the −2/+5 from the start site. Putative Sp1, CCAAT and GATA binding sites were also found in this region ( Figure 2A ). To examine the promoter function of the region upstream of the transcription start site, we then generated four reporter plasmids by introducing −1000/+19, −325/+19, −145/+19 or −115/+19 region upstream of the luciferase gene of pGL2-Basic. These plasmids were transiently transfected to K562 and U937 cells, the latter as a non-erythroid control, and luciferase activity was determined in cell lysates. In K562 cells the promoter-driven luciferase activities of −1000/+19, −325/+19 and −145/+19 promoter were about 15-fold higher than that of pGL2-Basic; a significant reduction of luciferase activity was seen for the reporter of −115/+19 promoter, indicating that the cis -element present in the sequence spanning from −145 to −116 plays an important role in transcriptional activation. In contrast, in U937 cells there was little difference in the luciferase activities expressed from these reporter plasmids and pGL2-Basic ( Figure 2B ). These data suggest that the −145/+19 region contains the elements necessary for the cell-type-specific expression of Gfi-1B in K562 cells.
To investigate the functional contribution of the GATA binding sites and the CCAAT box to this promoter activity, these sites were inactivated by mutagenesis in the −145/+19 promoter construct, and the expression was analyzed in K562 cells ( Figure 2C ). Disruption of either the −132 GATA site or the −96 CCAAT box reduced the promoter activity to 40 and 20%, respectively, relative to wild type. We found no further reduction when both sites were mutated, at least suggesting that transcription factors binding to these two sites do not activate this promoter additively. We also examined the contribution of another putative GATA binding site located at +13/+16 to the promoter activity. It appeared that mutagenesis of the +13 GATA site did not cause a significant effect on the promoter activity of the −145/+19 region. According to these results, we concluded that the −132 GATA and the −96 CCAAT sites are critical cis -elements controlling the Gfi-1B promoter activity in K562 cells.
Since GATA-1 is an erythroid-specific transcriptional activator ( 23 – 26 ), we then determined whether GATA-1 could bind at the −132 GATA site by gel-shift assays using the radiolabeled double-stranded oligonucleotide containing −146/−116 sequence of the Gfi-1B promoter. Incubation of this probe with nuclear extract of K562 cells gave rise to the formation of DNA–protein complexes I and II. These two complexes were abolished by an excessive amount of unlabeled oligomer ( Figure 3 ), but not by the Sp1 or Ap1 oligomer (data not shown), indicating their specificity to the −146/−116 sequence. One retarded complex seen in the assay was indicated to be non-specific, since different kinds of oligomers could compete for its formation (data not shown). Complex II formation, but not complex I, could be inhibited by the specific antibody against GATA-1 and competed by an excessive amount of unlabeled consensus GATA-1 or −146/−116 oligomer, but not by the mutated consensus GATA-1 or −146/−116 oligomer each containing GATA to CT TA deviation (the underlined letters indicate the mutated nucleotide) ( Figure 4 ). Complex I could be competed by the −146/−116 oligomer containing the −132 GATA mutation, but not by the GATA-1 consensus oligomer regardless of GATA mutation. This indicated that complex I formation is not a result of protein binding to the GATA site. When using the nuclear extract of U937 cells for the assay, a little amount of complex I was formed with this probe, while complex II was not seen. Here, we deduced that complex II contains GATA-1 bound to the −132 GATA site, and complex I is devoid of GATA-1.
GATA-1 activates the Gfi-1B promoter
Next, we ectopically expressed GATA-1 with Gfi-1B reporter plasmid in U937 cells, which lack endogenous GATA-1 ( 24 ), to examine the role of GATA-1 in transcriptional activation of the Gfi-1B promoter. We found that enforced expression of GATA-1, but not GATA-2 or GATA-3, increased luciferase activity from the Gfi-1B promoter more than 100-fold ( Figure 4A ). Expression of the dominant negative form of GATA-1 (C258G), which is defective in DNA binding ( 27 , 28 ), did not stimulate the promoter activity. In this experiment, we used the −325/+19 reporter construct to rule out the possibility that −325/+146 contains a responsive element to GATA-2 or GATA-3 expression. A similar result could also be obtained in the transfection experiment using NIH3T3 fibroblasts ( Figure 4B ), where the ectopic expression of GATA-1, GATA-2 and GATA-3 was well detected ( Figure 4C ). Thus, GATA-1 is a potent activator of the Gfi-1B promoter.
GATA-1 binding to the Gfi-1B promoter is necessary for Gfi-1B expression in K562 cells
To know whether GATA-1 directly binds to the Gfi-1B promoter in the cellular environment of K562 cells, we then performed ChIP assay using specific antibody against GATA-1 to immunoprecipitate formaldehyde-fixed chromatin. Endogenous Gfi-1B promoter DNA was amplified by PCR of immunoprecipitated chromatin. In this experiment, we used the HS2 region of β-globin control locus and human thymidine kinase (hTK) gene promoter as the positive and negative controls, respectively, since the GATA-1 binding site is present in HS2 region of β-globin ( 21 ) but not in the hTK promoter ( 29 , 30 ). As shown in Figure 5A , in K562 cells GATA-1 binds to the Gfi-1B promoter as well as to the HS2 region of β-globin locus, but not to the hTK promoter region. As expected, the ChIP experiment by GATA-1 antibody is not able to detect these DNA sequences in U937 cells. This experimental result let us to conclude that GATA-1 directly interacts with the Gfi-1B promoter region in K562 cells. To further test whether the function of GATA-1 is necessary for endogenous expression of Gfi-1B RNA transcript in K562 cells, we transfected cells with the expression vector of the GATA-1 (C258G) mutant and analyzed the Gfi-1B RNA transcript level by RNase protection assay. Because GATA-1 (C258G) retains its interacting ability with the other co-activator, sequestration of co-activator of GATA-1 by expressing this mutant is expected to disturb the transcriptional function of endogenous GATA-1 ( 27 , 28 ). Our results demonstrated that the expressed level of endogenous Gfi-1B transcript in K562 cells was significantly decreased in the cells overexpressing GATA-1 (C258G) mutant ( Figure 5B ). Taken together, we concluded that GATA-1 is a necessary transcription activator for the Gfi-1B promoter by binding at the −132 GATA site in K562 cells.
The CCAAT site is necessary for GATA-1-mediated transcriptional activation of Gfi-1B promoter
Because ectopic expression of GATA-1 activates the Gfi-1B promoter activity significantly in U937 cells, we then defined the cis -element contributing to GATA-1 responsiveness by transfecting U937 cells with the −145/+19 reporter plasmid carrying various mutations as indicated together with the expression vector of GATA-1. Mutation at the −132 GATA site caused ∼50% reduction of the promoter activity relative to the wild type. Of note, mutation at the −96 CCAAT site abolished 76% of promoter activity and mutation of both sites further decreased the GATA-1 activation effect ( Figure 6 ). Given that the effect of the −96 CCAAT site mutation was greater than what was observed for the mutation at the −132 GATA site, it is evident that the −96 CCAAT box also acts as a critical cis -element for GATA-1-mediated trans-activation.
NF-Y binds to the Gfi-1B promoter and cooperates with GATA-1 for trans-activation
To know which transcriptional activator binds to the −96 CCAAT box, we performed gel-shift assay using double-stranded −125/−86 oligonucleotides. After incubating this probe with the nuclear extract of K562 cells, we detected three specific DNA–protein complexes, which were competed by the unlabeled wild-type oligomer, but not by the oligomer with a mutation converting ATTGG to AT A G C . The formation of these three complexes were abolished when unlabeled consensus NF-Y oligonucleotide was included in the assay, and the complexes were supershifted by the NF-YB antibody and abolished by the NF-YA antibody. These data indicate that NF-Y is present in complexes I, II and III formed in this assay ( Figure 7A ). In agreement with the ubiquitous nature of NF-Y ( 31 , 32 ), we found that the extract of U937 cells also exhibited similar DNA-binding pattern to this labeled probe, except for the complex III. To examine whether GATA-1 could also bind to this region, unlabeled oligomer of consensus GATA-1 binding sequence was used for the competition assay. It appeared that the GATA oligomer could not compete for the complex formation, suggesting that GATA-1 does not bind to this CCAAT site. We then further performed ChIP assay to confirm the binding of NF-Y to the Gfi-1B promoter in chromatin ( Figure 7B ). Because the NF-Y binding site is present in human TK promoter ( 29 , 30 ) and it has been shown that a little amount of NF-Y is able to bind to the β-globin promoter in K562 cells ( 33 , 34 ), here we used these two promoters as the positive and negative controls for this analysis, respectively. The results showed that NF-Y binds to the Gfi-1B promoter in K562 and U937 cells as well. Since Gfi-1B promoter activity is very low in U937 cells, this result reflected the fact that binding of NF-Y, a ubiquitous CCAAT binding protein, to −96/−92 is insufficient to activate this promoter in the cells lacking GATA-1.
To test the functionality of NF-Y in trans-activating the Gfi-1B promoter, the expression vector of a dominant negative form of NF-YA, which is defective in DNA binding but still heterotrimerizes with NF-YB and NF-YC ( 20 ), together with the Gfi-1B, hTK or SV40 promoter reporter were transfected into K562 cells for promoter activity assay. Clearly, both Gfi-1B and hTK promoter activities were markedly decreased by co-expression of dominant active form of NF-YA, while SV40 promoter activity was increased by the same co-transfection experiment. This indicates that the DNA binding ability of NF-Y is specifically necessary for trans-activating the Gfi-1B promoter ( Figure 7C ). Furthermore, we showed that the extent of GATA-1-mediated transcriptional activation of Gfi-1B promoter in U937 cells was significantly reduced by co-expressing the dominant negative form of NF-YA ( Figure 7D ), indicating that functional cooperation between these two transcriptional activators is required for activation of the Gfi-1B promoter. In summary, these results point out that there is a mutual requirement between NF-Y and GATA-1 in transcriptional activation of the Gfi-1B promoter.
Gfi-1B has been shown to play a crucial role in erythropoiesis and megakaryopoiesis ( 8 , 15 , 16 ). Therefore, a detailed structural and functional analysis of Gfi -1B gene is a prerequisite to unravel its regulation mechanism during development of blood cells. Here we analyzed the 5′-UTR of the Gfi -1B gene for the first time and located the 133 bp of its 5′-UTR to ∼7.82 kb upstream of the first coding exon. The finding of this first exon region let us determine whether its upstream region is the proximal promoter of human Gfi -1B gene. Results from the functional reporter assays verified that the transcription of the Gfi -1B gene in K562 cells is driven by this proximal region, which is a TATA-less promoter spanning from −145 to the transcription start site. We further provided several evidences to demonstrate that cell-type-specific expression of Gfi-1B is dependent on GATA-1-mediated transcription, which plays a dominant role in erythroid/megakaryocytic differentiation ( 35 – 37 ). First, the promoter activity of the Gfi -1B gene in U937 cells could be increased significantly by enforced expression of GATA-1. Second, ChIP assay indicated the binding of GATA-1 in the Gfi-1B promoter region of chromatin from K562 cells. Third, expression of a mutant type of GATA-1 (C258G) in K562 cells diminished endogenous expression of Gfi-1B RNA transcript in K562 cells.
In this study, we located the functional GATA-1 site at −132/−129 within the promoter −145/+19 region of Gfi -1B gene, which contributes a great part to the promoter activity. It is interesting to observe that mutation at the −96 CCAAT site caused a reduction of GATA-1-mediated activation of Gfi-1B promoter to an extent greater than mutation of GATA site. Thus, in addition to the −132 GATA site, the CCAAT box is also involved in GATA-1-mediated transcription. One possibility explaining this result is that GATA-1 could directly bind to the CCAAT box to trans-activate this promoter. However, inclusion of an excess amount of GATA-1 consensus oligonucleotide in gel-shift assay did not ablate complex formation with the labeled oligonucleotide covering this CCAAT box, thus excluding the possibility of direct binding of GATA-1 to this site. Our experimental results from both gel-shift and ChIP assays indicate that NF-Y binds to this CCAAT box on the Gfi-1B promoter region in K562 and U937 cells. Because Gfi-1B promoter activity is very low in U937 cells, it is obvious that the association of NF-Y is insufficient to activate this promoter. Since expression of the dominant negative form of NF-YA effectively decreased GATA-1-mediated activation effect on the Gfi-1B promoter in U937 cells, apparently there is a cooperative relationship between GATA-1 and NF-Y in trans-activation of the Gfi-1B promoter. Following these clues, we have speculated that GATA-1 may interact with NF-Y to enhance the NF-Y-mediated transcription for the Gfi-1B promoter. However, we were unable to detect a reproducible interaction between GATA-1 and NF-Y in the nuclear extract of K562 cells by different approaches. Nor did we find that GATA-1 is present in the DNA–protein complex formed with the CCAAT box sequence. Therefore, it is unlikely that GATA-1-responsiveness via the −96 CCAAT box is due to the interaction between GATA-1 and NF-Y. It should be mentioned that the in vivo footprint pattern around this NF-Y binding site was more extensively protected in K562 cells than in U937 cells (data not shown). Perhaps, the presence of GATA-1 in K562 cells affects the interaction context between NF-Y and this promoter. Therefore, it is possible that GATA-1 stimulates the recruitment of the co-activator that is necessary for the trans-activation function of NF-Y. Another possibility is that NF-Y binding pre-sets the Gfi-1B promoter for GATA-1-mediated activation of transcription. The mechanism responsible for the cooperation between NF-Y and GATA-1 remains to be investigated.
The involvement of both NF-Y and GATA-1 in gene transcription has been demonstrated in the γ-globin and FcγRIIA promoters. In the case of γ-globin promoter, it has been suggested that they bind to the CCAAT region of this promoter in a competitive manner, because disruption of NF-Y binding facilitates the recruitment of GATA-1 to the γ-globin promoter ( 33 , 38 ). In the study of FcγRIIA gene promoter, it has been shown that NF-Y binding to the CCAAT box together with GATA-1 or GATA-2 binding to the GATA site result in activation of this promoter in an additive manner ( 39 ). Distinct from these two erythroid-specific promoters, here our data exemplify one situation that GATA-1 and NF-Y are mutually required for trans-activation of an erythroid-specific promoter.
The present study has defined the location and the sequence of the Gfi-1B promoter, which is located far away from its first coding exon. The detailed gene structure of Gfi-1B obtained from this study will provide useful information for future investigation on the regulatory mechanism of its transcriptional and post-transcriptional control. Through this investigation, we proved that erythroid-specific expression of Gfi-1B is attributed to GATA-1-mediated transcription, which explained the previous report showing that expression of Gfi-1B is well correlated with the expressed level of GATA-1 in human peripheral blood CD34+ cells during differentiation from erythroid-blasts forming units to mature erythroblasts ( 16 ). Studies of gene knockout mice have clearly indicated that Gfi-1B and GATA-1 are transcription factors essential for development of the closely related erythroid and megakaryocytic lineage ( 15 , 35 ). Given the fact that expression of Gfi-1B does not stimulate GATA-1 expression ( 16 ), our data suggest that Gfi-1B is one of downstream targets of GATA-1 in this developmental lineage. In addition, this work provided an example of a natural erythroid-specific promoter, whose activation requires the GATA-1-mediated transcription involving cooperation with NF-Y binding to CCAAT site.
We are grateful to Dr J.-Y. Lin for the suggestion on the oligo-capping method and Dr R. Mantovani for the gift of the dominant negative NF-YA plasmid. This research is supported by grant NSC91-3112-B-002-012 and NSC93-2752-B-002-006-PAE from the National Science Council, and a grant EDU89-B-FA01-1-4 from Education Missionary, Taiwan, Republic of China.
Graduate Institute of Biochemistry and Molecular Biology, College of Medicine, National Taiwan University, No. 1 Jen Ai Road 1st Section, Taipei, Taiwan, Republic of China, 1Institute of Biomedical Sciences, Academia Sinica, Taipei 115, Taiwan, Republic of China and 2Laboratory of Genome Structure Analysis, Human Genome Center, the Institute of Medical Science, University of Tokyo, 4-6-1 Shirokanedai, Minatoku, Tokyo 108-8639, Japan