Binding of NF-Y Transcription Factor to One of the Cis-elements in the Myeloperoxidase Gene Promoter That Responds to Granulocyte Colony-stimulating Factor*

The expression of the myeloperoxidase (MPO) gene is restricted to cells of the myeloid cell lineage and is induced by granulocyte colony-stimulating factor (G-CSF). In this study, a series of deletion mutations was introduced in the promoter of the human MPO gene, which was then fused to the chloramphenicol acetyltransferase gene. The G-CSF-induced promoter activity was examined in mouse myeloid precursor FDC-P1 transformants that constitutively express the G-CSF receptor. A G-CSF-responsive element (GRE) in the MPO gene was found approximately 800 base pairs upstream from the transcription initiation site. When the 5′-flanking region of the human MPO gene contained this element, it yielded promoter activity in cells cultured with G-CSF but not in cells cultured with interleukin 3. Gel shift assays with the element showed that a specific nuclear factor(s) (NF/G-CSF) binds to the element. The NF/G-CSF was purified by affinity chromatography using an oligonucleotide of GRE. Protein sequence analysis of the purified NF/G-CSF indicated that NF/G-CSF is a ubiquitous transcription factor, NF-Y, which is composed of three subunits. The recombinant NF-Y was then shown to bind to GRE in a combination of the three subunits.

Neutrophils are generated in the bone marrow from pluripotent stem cells and differentiate over various intermediate stages. There is a series of well characterized myeloid cell lines whose maturation is blocked at defined stages of differentiation (1,2). Studies with these cell lines have indicated that granulopoiesis requires the sequential expression of a number of genes, which are regulated in tissue-and development-specific manners (3). Myeloperoxidase (MPO) 1 is a heme-containing glycoprotein present in azurophilic granules of polymorphonu-clear neutrophils (4) and plays an important role in an oxygendependent antimicrobial system of neutrophilic granulocytes. The MPO gene is expressed only during the late myeloblast and promyelocyte stages of myeloid cell development (5). The transcript is not found at earlier or later stages in neutrophilic development or in other cell types.
We and others have isolated human MPO cDNA and genomic DNA (6 -12). Using these cloned MPO genes, several DNase I hypersensitive sites that were associated with the changes of MPO gene expression during myeloid cell differentiation were assigned in the 5Ј-flanking region of the gene (13)(14)(15). Tissuespecific enhancer-like elements were also identified in the 5Јflanking region, which is located about 1 kb upstream from the translation start site of the murine MPO gene (16) and introns 7 and 9 of the human MPO gene (17).
Granulocyte colony-stimulating factor (G-CSF) is one of the colony-stimulating factors (18,19) that are required for proliferation and differentiation of progenitor cells. Specifically, when bone marrow cells or myeloid leukemia cells are incubated with G-CSF, G-CSF produces neutrophilic granulocytes accompanied with expression of neutrophil-specific genes including MPO and neutrophil elastase genes (3,20,21). Therefore, the elucidation of the G-CSF-induced MPO gene expression would provide insights into G-CSF-induced myeloid cell differentiation. Previously, we introduced a G-CSF receptor (G-CSFR) expression plasmid into an interleukin-3 (IL-3)-dependent murine myeloid precursor cell line FDC-P1, which normally does not express G-CSFR (22)(23)(24). Transformants grown in medium containing IL-3 behaved like myeloid precursor cells. However, when the transformants were cultured in G-CSF-containing medium, they started to express neutrophilspecific genes such as MPO and neutrophil elastase (23,24). These results indicated that the G-CSFR, but not the IL-3 receptor, transduces the differentiation signal in myeloid precursor cells and induces MPO gene expression. Previously, several attempts have been made to reconstitute the G-CSFinduced MPO gene expression using its promoter to study the differentiation signal from G-CSF receptor, but they have not been successful (16,25).
In this report, we localized one of the G-CSF-responsive cis-regulatory elements in the 5Ј-flanking region of the human MPO gene. The 5Ј-flanking region including this element functioned as a promoter in the cells cultured with G-CSF but not in the cells cultured with IL-3, reflecting the regulatory profile of MPO gene expression. Purification of NF/G-CSF and analysis of its amino acid sequence identified NF/G-CSF as a ubiquitous transcription factor, NF-Y.

EXPERIMENTAL PROCEDURES
Construction of the CAT Reporter Plasmid Carrying the MPO Promoter-A DNA fragment carrying the minimal promoter region of the thymidine kinase gene (from Ϫ32 to ϩ52) of the herpes simplex virus (26) was amplified by polymerase chain reaction (PCR) from pTK100CAT (27) and inserted into pSV0BCAT (28) to generate pTK-CAT. An approximately 4-kb AvaI-HindIII fragment carrying the 5Јflanking region of the human MPO gene (6) was inserted between the AvaI and HindIII sites of pTKCAT to generate pVPTKCAT. To prepare a series of 5Ј-deletion mutants of the MPO gene promoter, pVPTKCAT was digested with SmaI and KpnI, treated with exonuclease III at 37°C for 1-5 min, and self-ligated. The end points of the deletion were determined by DNA sequencing, and the plasmids carrying the appropriate length (0.75-3.0 kb) of the MPO gene promoter were designated as p⌬MPO-1, 2, 3, 4, 6, 7, and 8 ( Fig. 1). To construct p⌬MPO-5, pVPTKCAT was digested with SacI and religated.
To prepare the linker scanning mutants, a BamHI-HindIII fragment carrying the MPO promoter region of p⌬MPO-6 was subcloned in pBluescript II (designated as pB⌬MPO-4). Mutations in GRE-1 were introduced by recombinant PCR (29) using pB⌬MPO-4 as a template. Four sets of oligonucleotides (LS-1F and LS-1R, LS-2F and LS-2R, LS-3F and LS-3R, and LS-4F and LS-4R), in which the 20-nucleotide sequence at the 3Ј-end, corresponding to the target sequence, and the 10-nucleotide sequence at the 5Ј-end, made of ClaI linker sequence, were used as mutagen primers. T3 primer (AATTAACCCTCACTA-AAGGG) and MPO4 primer (CAGGCTGGTCTTGAACTC) were used as forward and reverse primers, respectively. In the primary PCR, the 5Ј-part of the promoter in pB⌬MPO-4 was amplified using T3 and LS-1R, -2R, -3R, or -4R primers, and the 3Ј-part was amplified with LS-1F, -2F, -3F, or -4F and MPO4 primers, respectively. After treatment with T4 DNA polymerase, each PCR product was mixed at 1:1, and a second PCR was performed using T3 and MPO4 primers. The products were digested with BglII and BstXI and ligated to p⌬MPO-6. The resulting plasmids were designated as pLS-1, pLS-2, pLS-3, and pLS-4, respectively (Fig. 6A).
Cell Culture and Transfection-The mouse myeloid cell line FD62M (23) was maintained in RPMI 1640 medium containing 10% fetal calf serum (Life Technologies, Inc.) and 45 units/ml mouse recombinant IL-3 (30). Human recombinant G-CSF was provided by Chugai Pharmaceutical Co. The reporter plasmids were introduced into FD62M cells by electroporation using a Gene Pulsar (Bio-Rad) as described previously (22). In brief, 5 ϫ 10 6 FD62M cells were electoporated with 40 g of the reporter plasmids linearized with NdeI and 1 g of BamHIdigested pBLIIhyg (31) at 350 V with a capacitance of 250 microfarads. The transformed cells were selected by culturing in RPMI medium containing 10% fetal calf serum, 45 units/ml IL-3, and 500 g/ml hygromycin B, and four independent transformants were pooled for further analysis.
CAT Assay-Stable transformants carrying the reporter plasmid were transferred to the medium containing either IL-3 (45 units/ml) or G-CSF (150 units/ml) and cultured for 2 days. The CAT assay using 2 g of protein was performed as described (28). After a 2-h incubation at 37°C, the products were separated by thin layer chromatography (Kieselgel 60; Merck). The amounts of radioactivity in the acetylated and nonacetylated forms of chloramphenicol were quantified using a Bio-Image Analyzer (BAS 2000, Fuji Photo Film Co).
Gel Shift Assay-Nuclear extracts were prepared according to Edgar et al. (32). In brief, 1 ϫ 10 6 cells were incubated in 400 l of 10 mM Hepes-KOH buffer (pH 7.9) containing 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride for 15 min on ice. Nonidet P-40 was added to the cell suspension at a final concentration of 0.625% (v/v), and cells were disrupted by vigorous mixing. After centrifugation, nuclei were resuspended in 50 l of 20 mM Hepes-KOH buffer (pH 7.9) containing 400 mM NaCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride and incubated at 4°C for 30 min. The mixture was spun at 10,000 ϫ g for 5 min, and the supernatant was recovered as nuclear extracts.
Five pairs of oligonucleotides for GRE (Fig. 3A) and one pair for NFY-1 (TGAAACATTTTTCTGATTGGTTAAAAGTTG and AGCACT-CAACTTTTAACCAATCAGAAAAATG) (33) were labeled with ␣-[ 32 P]dCTP and used as probes for the gel shift assay. The reaction mixture for the gel shift assay contained 20 mM Tris-HCl (pH 8.0), 100 mM KCl, 10% (v/v) glycerol, 2% (w/v) polyvinyl alcohol (Sigma), 2 g of poly(dI-dC) (Pharmacia Biotech Inc.), and 2-10 g of nuclear extract. After preincubation for 30 min on ice, 32 P-labeled probe DNA (60 fmol, about 20,000 cpm) was added to the mixture, and the binding reaction was allowed to proceed at 30°C for 30 min. The product was then resolved by electrophoresis on a 6% polyacrylamide gel as described (28).
Purification of NF/G-CSF-The latex beads carrying DNA fragments were prepared as described (34). In brief, two pairs of oligonucleotides (AAGAACCAGAGGCTGCCCATTGGGTGGCCCCC and GC-CACCCAATGGGCAGCCTCTGGTTCTTGGGGG for the wtGRE-latex and AAGAACCAGAGGCTGCCCCGGTTGTGGCCCCC and GCCA-CAACCGGGGCAGCCTCTGGTTCTTGGGGG for the mtGRE-latex) were synthesized. After annealing each pair of the oligonucleotides (25 g), they were phosphorylated and ligated as described (34). The oligomerized DNA fragments were digested with T4 DNA polymerase and mixed with 2.5 mg of the activated latex beads in 10 mM potassium phosphate buffer (pH 8.0). After incubation at 50°C for 24 h, the beads were washed twice with 2.5 M KCl. The unreacted active sites were then

FIG. 2. Effects of IL-3 and G-CSF on the promoter activity of the MPO gene.
The FD62M cells transformed with p⌬MPO-6, pTK-CAT, or pBOSCAT were cultured with G-CSF alone (lanes 3, 6, and 9), with IL-3 alone (lanes 1, 4, and 7), or with both of these factors (lanes 2, 5, and 8) for 48 h. The CAT assay was carried out using either 2 g (lanes 1-6) or 0.1 g (lanes 7-9) of cell extract. The reaction products were resolved by thin layer chromatography. The amounts of radioactivity were quantified using a BioImage Analyzer. Plus and minus indicate the presence and absence of G-CSF or IL-3 in the culture medium. Relative CAT activities are shown as percentages of that of p⌬MPO-6 in the presence of G-CSF (lane 3) after normalizing for the amount of the protein used for the assay. FD62M cells (2 ϫ 10 9 cells) were cultured in the presence of 150 units/ml human G-CSF for 72 h, and the nuclear extracts were prepared as described above. The nuclear extracts (about 65 mg of proteins) were incubated for 2 h at 4°C with 133 g/ml poly(dI-dC) and 50 mg of mtGRE-latex in 1 ml of the binding buffer (50 mM Tris-HCl (pH 7.9), 20% glycerol, 1 mM EDTA, 1 mM dithiothreitol, and 0.1% Nonidet P-40) containing 100 mM KCl. After removing the mtGRE-latex by centrifugation, 50 mg of the wtGRE-latex was added to the supernatant and incubated at 4°C for 2 h. The beads were washed with the binding buffer containing 100 mM KCl, and the bound proteins were eluted with the binding buffer containing 1 M KCl. The eluate was diluted with the binding buffer to reduce the concentration of KCl to approximately 100 mM, and the affinity purification using 50 mg of the mtGRE-latex and 50 mg of wtGRE-latex was repeated twice more. The purified NF/G-CSF was separated on a 16.5% polyacrylamide gel in a buffer consisting of 1 M Tris-HCl (pH 8.45) and 0.1% SDS (35). After electrophoresis, the proteins were transferred onto a polyvinylidene difluoride membrane (ProBlott; Applied Biosystems) and were stained with Coomassie Brilliant Blue R250. The stained band of about 19 kDa was subjected to sequencing analysis using a protein sequencer (Applied Biosystems protein sequencer, type 492).
Preparation of GST-NF-Y Fusion Proteins-Mouse cDNAs for NF-YA and NF-YB were isolated from mouse FD62M cells by PCR, using primers based on the published sequences (36). Their authenticity was confirmed by DNA sequence analysis. To amplify the mouse NF-YC cDNA from FD62M cells, the sequence for rat NF-YC cDNA (37) was used. The nucleotide sequence of mouse NF-YC in the coding region differed from that of rat NF-YC at 13 positions, while its amino acid sequence differed from that of rat NF-YC at two positions, i.e., Ser-9 of rat NF-YC was changed to Gly, and one Gln residue was inserted between Gln-296 and Leu-297 of rat NF-YC.
To produce the recombinant NF-YA, -YB, and -YC proteins, respective cDNAs were fused to the glutathione S-transferase (GST) gene of pGEX-4T (Pharmacia) and introduced into Escherichia coli AD202. The expression of the fusion proteins was induced by adding 0.1 mM isopropyl-␤-D-thiogalactopyranoside at 25°C. The cells were lysed by freezing and thawing or by sonication in phosphate-buffered saline containing 1 mM phenylmethylsulfonyl fluoride and 0.1% Nonidet P-40. GST-NF-YA protein was recovered as soluble protein, while the GST-NF-YB and -YC proteins formed inclusion bodies. GST-NF-YB and -YC were then solubilized with 8 M urea in buffer A (50 mM Tris-HCl (pH 7.9), 100 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10% glycerol) and dialyzed against buffer A. The fusion proteins were then purified by glutathione-Sepharose 4B (Pharmacia) as described (38).
Preparation of Antiserum and Western Blotting--To prepare the antiserum against NF-Y proteins, rabbits (New Zealand White, 10 weeks old) were immunized with 0.5 mg of the recombinant GST-NF-YA, -YB, and -YC proteins in complete Freund's adjuvant. Immunization with the GST-NF-YA, -YB, or -YC in incomplete Freund's adjuvant was repeated every 2 weeks for 6 weeks. At 1 week after the last immunization, rabbits were bled to prepare the antisera.
For Western blotting, the purified NF/G-CSF or the crude nuclear extract was heated for 5 min at 95°C in SDS sample buffer. The proteins were resolved by electrophoresis on a 4 -20% gradient polyacrylamide gel (Daiichi Chemicals, Co.), and blotted onto polyvinylidene difluoride membranes (Millipore Corp.). The membranes were incubated with anti-NF-YA, anti-NF-YB, or anti-NF-YC antibody and then with horseradish peroxidase-conjugated secondary antibody. The proteins recognized by the antibodies were visualized by the enhanced chemiluminescence system (Renaissance; NEN Life Science Products).

G-CSF-inducible and IL-3-repressive Elements in the Human
MPO Gene-We have previously established IL-3-dependent FDC-P1 cell transformants (FD62M) that constitutively express G-CSFR (23). These transformants do not express the MPO gene when they are cultured in medium containing IL-3. However, when the cells are cultured in G-CSF-containing medium, they abundantly express MPO mRNA (more than 50-fold induction compared with that in IL-3-containing medium) (23), as found with bone marrow cells (20,21). To examine the regulatory mechanisms behind MPO gene expression, an approximately 4-kb 5Ј-flanking region of the MPO gene was fused to the CAT gene carrying the minimal promoter of the thymidine kinase gene. The fusion plasmid (pVPTKCAT) was introduced into FD62M cells, and stable transformants carrying the reporter gene were maintained in IL-3-containing medium. The cells were then transferred to the G-CSF-containing medium and cultured for 48 h, and the CAT activities in the cell extracts were determined. As shown in Fig. 1, the cells carrying pVPTKCAT showed 200-fold more CAT activity than the cells carrying pTKCAT, suggesting that an enhancer element(s) is in FIG. 4. Binding of NF/G-CSF to GRE-1 and GRE-5. Nuclear extracts (6 g) from FD62M cells cultured with IL-3 were incubated with 32 P-labeled GRE-1 (lanes 1-5) or GRE-5 (lanes 6 -10) in the absence (Ϫ) (lanes 1 and 6) or in the presence of 10-fold (lanes 2, 4, 7, and 9) or 100-fold (lanes 3, 5, 8, and 10) amounts of unlabeled GRE-1 (lanes 2, 3, 9, and 10) or GRE-5 (lanes 4, 5, 7, and 8). The resulting complexes were resolved on a 6% polyacrylamide gel. The position of NF/G-CSF is indicated on the left. the 4-kb 5Ј-flanking region. However, in contrast to no detectable MPO mRNA in the IL-3-containing medium, the promoter showed an enhancer activity in the presence of IL-3 (Fig. 1, A  and B). To localize the enhancer element that responds to G-CSF but not IL-3, a series of 5Ј-deletion mutants was created in the promoter of the human MPO gene, and then their promoter activities were measured (Fig. 1A). Deletions up to Ϫ1023 (p⌬MPO-6) gradually reduced the CAT activity, and further deletion up to Ϫ928 (p⌬MPO-7), completely abolished the enhancing effect. On the other hand, when the cells were cultured with IL-3-containing medium, pVPTKCAT and its deletion mutants up to Ϫ1457 (p⌬MPO-5) showed some promoter activities, but its activity was not found with p⌬MPO-6 ( Fig. 1, A and B). These results indicate that one of the cisregulatory elements of the MPO gene, which is G-CSF-inducible but is not responsive to IL-3, is located in the region from Ϫ928 to Ϫ1023 base pairs in the 5Ј-flanking region of the MPO gene.
The G-CSF-induced MPO gene expression is inhibited by the co-stimulation of cells with IL-3 (23). To study whether this inhibitory effect is mediated by the cis-element identified above, the effects of IL-3 on the promoter activity in p⌬MPO-6 were examined. As shown in Fig. 2, the CAT activity in cells cultured with IL-3 was only 1.2% of that found in cells cultured with G-CSF. Moreover, when IL-3 was added to the G-CSFcontaining medium, it greatly inhibited the G-CSF-induced promoter activity of p⌬MPO-6. The inhibitory effect of IL-3 was specific to the MPO promoter; i.e. IL-3 had no effect on the reporter plasmid pBOSCAT, which carried the promoter of the human elongation factor 1␣ gene (39). These results suggest that a cis-regulatory element(s) within the 95-base pair fragment (Ϫ1023 to Ϫ928 base pair), located at about 0.8 kb upstream of the transcription start site, is responsible for the G-CSF-inducible and IL-3-repressive MPO gene expression.
Binding of Nuclear Factor to the Cis-regulatory Element-To localize the cis-regulatory element(s) and to identify the nuclear factors that bind to such elements, a set of oligonucleotides spanning the DNA sequence from Ϫ1023 to Ϫ928 of the human MPO promoter region was synthesized (Fig. 3A). Nuclear extracts were prepared from the FD62M cells, and the presence of nuclear factors binding to the DNA elements was examined by gel shift assay. As shown in Fig. 3B, the oligonucleotides GRE-1 and GRE-5 yielded a retarded band (lanes 1 and 5) when they were incubated with the nuclear extracts. The nuclear extracts prepared from cells cultured either with G-CSF or with IL-3 gave the retarded bands of the identical migration positions (lanes 6 and 7). The putative nuclear factor that bound to GRE-1 and GRE-5 was named NF/G-CSF.
The same mobility of NF/G-CSF with the GRE-1 and GRE-5 probe DNAs suggested that same nuclear factor(s) bind to GRE-1 and GRE-5. To examine this possibility, increasing amounts of unlabeled GRE-1 or GRE-5 were added to the gel shift assay as competitors. As shown in Fig. 4, the NF/G-CSF binding activity with GRE-1 could be competed not only by a 100-fold excess of unlabeled GRE-1 but also by GRE-5. Similarly, the nuclear binding activity with GRE-5 was competed by unlabeled GRE-5 as well as by GRE-1. These results suggested that the same nuclear factor(s) in the extracts bound to both GRE-1 and GRE-5. Both GRE-1 and GRE-5 consist of 30 nucleotides and carry a common DNA sequence motif (GCC(A/ C)ATTGGGT), suggesting that the nuclear factor(s) binding to GRE-1 and GRE-5 recognized this motif. To examine whether NF/G-CSF recognized this sequence, a series of oligonucleotides that carried point mutations in GRE-1 was prepared (Fig.  5A), and these oligonucleotides were then used as probe DNAs in a gel shift assay. As shown in Fig. 5B, the mutations in CCATTGGGT abolished the binding of NF/G-CSF to the DNA fragment, whereas the mutation (GRE-1-M7) downstream of this sequence only slightly affected the ability of NF/G-CSF to bind to the element, suggesting that NF/G-CSF mainly recognizes CCATTGGGT and weakly interacts with the downstream sequence.
To confirm the involvement of the GRE-1 and GRE-5 in the MPO promoter activity, mutations were introduced in the GRE-1 and GRE-5 of the p⌬MPO-6 construct, and their promoter activities were examined. As shown in Fig. 6B, the mutations in GRE-1 (constructs pLS-1 and pLS-2) (Fig. 6A) abolished the promoter activity of p⌬MPO-6. Mutations in GRE-5 (pLS-4) also affected the promoter activity, while the mutation (pLS-3) in the sequence between GRE-1 and GRE-5 did not have much effect on the promoter activity. These results suggested that both GRE-1 and GRE-5 are essential for the MPO gene expression.
Purification of NF/G-CSF and Its N-terminal Sequence-To purify NF/G-CSF, we prepared two sets of the oligonucleotides. One set of the oligonucleotides carried the sequence of GRE-1 (wtGRE), to which NF/G-CSF should bind. The other set of oligonucleotides (mtGRE) contained five point mutations in GRE-1 and would not bind NF/G-CSF. These oligonucleotides were oligomerized and covalently attached to the activated latex beads (34). Nuclear extracts were prepared from FD62M cells that had been cultured in the presence of G-CSF for 72 h. As shown in Fig. 7A, when the latex beads carrying the mutated GRE-1 (mtGRE-latex) were added to the extracts, most NF/G-CSF was recovered in the unbound fraction. The latex beads carrying the wild-type GRE (wtGRE-latex) were then added to the unbound fraction from mtGRE-latex. Most NF/G-CSF bound to the beads and was eluted with 1 M KCl. This cycle of affinity chromatography with the mtGRE-and wtGRE-latex beads gave rise to about 60-fold purification of NF/G-CSF with a yield of 23%. However, the eluates from the wtGRE-latex still contained many different proteins, as revealed by SDS-polyacrylamide gel electrophoresis analysis of the sample, followed by silver staining (Fig. 7B). Therefore, affinity chromatography with mtGRE-and wtGRE-latex beads was repeated twice more. Every time, most of the NF/G-CSF did not bind to the mtGRE-latex, but it bound to the wtGRE-latex beads (Fig. 7A). When the sample eluted from the third affinity chromatography of wtGRE-latex was analyzed by SDS-polyacrylamide gel electrophoresis, silver staining showed a prominent band of about 19 kDa with faint bands around 37 kDa (Fig. 7B). The overall purification of NF/G-CSF was about 10,000-fold from the crude nuclear extracts, and its yield was about 3%.
The major protein (19 kDa) in the purified NF/G-CSF was then subjected to protein sequence analysis after electrophoretic transfer onto a polyvinylidene difluoride membrane. Edman degradation yielded a sequence of the first 10 residues as YIGGSRYVIQ. A comparison of this sequence with all protein sequences in the data base of the National Center for Biotechnology Information using the BLAST algorithm (40) indicated that 9 of the 10 amino acids were identical to those of the mouse NF-YB (YIGGSHYVIQ, residues 21-30) (36). These results suggested that NF/G-CSF is the NF-Y transcription factor.
Identification of the 19-kDa Protein as One of the Subunits of NF-Y-To examine the possibility that NF/G-CSF is the NF-Y transcription factor, an oligonucleotide probe (NFY-1) carrying the consensus sequence ((G/A)CCAAT(C/G)AGG(C/A)) (41) for the NF-Y binding site was prepared. As shown in Fig. 8, the purified NF/G-CSF bound not only GRE-1 but also NFY-1 probe DNA. The complex formation between NF/G-CSF and GRE-1 was competed with an excess of unlabeled NFY-1. Similarly, the binding of NF/G-CSF to NFY-1 was competed not only by an excess of unlabeled NFY-1 but also by GRE-1. However, the mutated GRE-1 had only a slight inhibitory effect on the complex formation between the NF/G-CSF and NFY-1.
The purified mouse NF-Y transcription factor was reported to consist of two subunits, NF-YA and NF-YB (36), whereas the functional rat NF-Y is composed of three subunits, NF-YA (40 kDa), NF-YB (32 kDa), and NF-YC (40 kDa). Since the major protein in the purified NF/G-CSF carried the amino acid sequence of NF-YB, we first expected that the NF-YB subunit alone could bind GRE-1. However, the mouse recombinant NF-YB fusion protein with GST did not show a shifted band in a gel shift assay with the GRE-1 probe (Fig. 9). We therefore prepared the recombinant proteins for the other subunits (NF-YA and NF-YC). As shown in Fig. 9, when all subunits were added to the gel shift assay, it gave a shifted band with the GRE-1 probe, whereas all combinations of the two subunits were negative. Similar results were obtained with the recom- , and third (0.004 g, lane 9) chromatography. After electrophoresis for gel shift assay, the gel for lanes 1-3 was exposed to an x-ray film for 24 h, while the gel for lanes 4 -9 was exposed to an x-ray film for binant proteins in which the GST regions were removed by thrombin digestion (data not shown). These results indicated that, like rat NF-Y, the functional mouse NF-Y is a heterotrimer consisting of A, B, and C subunits. Moreover, these results suggested that the purified NF/G-CSF should contain not only the NF-YB but also the NF-YA and NF-YC subunits.
NF-Y Subunits in NF/G-CSF-To examine whether the purified NF/G-CSF contains the components of NF-Y, the antisera for NF-YA, NF-YB, and NF-YC were prepared, and the purified NF/G-CSF was analyzed by Western blotting. As shown in Fig.  10A, the purified NF/G-CSF showed bands of 39 and 36 kDa for NF-YA, 25 and 19 kDa for NF-YB, and 19 kDa for NF-YC, indicating that NF/G-CSF contains three subunits of NF-Y. However, the sizes of NF-YB and NF-YC in the purified preparation were smaller than those published for NF-Y subunits (32 kDa for NF-YB and 40 kDa for NF-YC) (36). Since proteinase activity is highly induced during granulocyte differentiation (16,25), it is possible that proteolysis of NF-Y components during purification has caused the apparent reduction of the NF-Y subunits. To examine this possibility, the crude nuclear extracts were prepared and analyzed by Western blotting. As shown in Fig. 10B, when the nuclear extracts prepared from the FD62M cells cultured in IL-3 were analyzed, all NF-Y subunits were close to the expected sizes. On the other hand, when the cells were cultured in the presence of G-CSF, the size of NF-YC subunits (19 kDa) was lower than expected (40 kDa). This degradation of NF-YC was partially inhibited by adding a high concentration of a protease inhibitor (10 mM of 4-(2-aminoethyl)benzenesulfonyl fluoride⅐HCl) to the nuclear extraction buffer (Fig. 10B). Although the size of NF-YB (32 kDa) in the crude extracts was comparable with that expected, it is possible that NF-YB is also degraded during purification to yield the 19-and 25-kDa proteins seen in the purified preparation (Fig. 10A). DISCUSSION IL-3 is a multicolony-stimulating factor that stimulates proliferation of the early progenitor cells of neutrophilic granulocytes, while G-CSF regulates the later stages of differentiation (18,19). IL-3 keeps the cells at the premature stage, whereas G-CSF induces their differentiation into neutrophils, accompanied by the expression of neutrophil-specific genes (3,20,21,23). The signals from the IL-3 receptor usually prevail over those from G-CSFR; i.e. IL-3 inhibits the G-CSF-induced terminal differentiation of neutrophilic granulocytes as well as the induction of neutrophil-specific genes (3,20,21,23). In this report, we have examined the G-CSF responsiveness of the human MPO gene promoter. When the ϳ4-kb 5Ј-flanking region of the human MPO gene was fused to a basal promoter element, it showed the enhancing activity not only in G-CSFbut also IL-3-containing medium. One possible explanation is as follows. A repressor(s) inhibiting the MPO gene expression in the IL-3-containing medium may be limiting. Since the exogenously induced gene usually integrates in the host chromosome as many copies, it is possible that such a repressor(s) may have been titrated out, as found in rat prolactin promoter (42). This may also explain the failure of several trials to localize the G-CSF-responsive element in the MPO gene by transient transfection (16,25). In any case, when this flanking region was truncated at Ϫ1023, only G-CSF and not IL-3 activated the promoter, suggesting that the region (Ϫ181 to Ϫ1023) of the MPO promoter is, at least in part, responsible for the regulated expression of MPO gene.  lanes 1 and 4) or presence of unlabeled NFY-1 (lanes 2, 3, 5, and 6), GRE-1 (lanes 7 and 8), or mutant GRE-1-M (lanes 9 and 10). The unlabeled oligonucleotide was added at a 10-(lanes 2, 5, 7, and 9) or 100-fold (lanes 3, 6, 8, and 10) excess of each probe. The resulting complexes were resolved on 6% polyacrylamide gel and visualized by autoradiography. FIG. 9. Binding of NF-Y to GRE-1. Recombinant NF-YA, -YB, and -YC proteins were produced in E. coli, as fusion proteins with GST, and purified by glutathione-Sepharose 4B (Pharmacia). Each recombinant protein (0.25 g) was mixed as indicated at the top and incubated with 32 P-labeled GRE-1 as a probe. The resulting complexes were resolved on 6% polyacrylamide gel and visualized by autoradiography. The GST-NF-Y⅐GRE-1 complex is indicated by an arrow. As a control, the purified NF/G-CSF was used for the gel shift assay (lane 9).  6 and 7), and anti-NF-YC (lanes 8 -10) as described above. I and G above the panels represent IL-3 and G-CSF, respectively. The sizes of standard proteins are shown in kDa on the right.
The GRE in the human MPO gene is indispensable for MPO gene expression, since its deletion or point mutations completely abolished the promoter activity. However, when this element, reiterated four times, was directly attached to the basal promoter carrying the TATA box of thymidine kinase gene, it did not enhance the promoter activity (data not shown). These results indicate that there are other cis-regulatory elements in the region of Ϫ927 to Ϫ181 of the 5Ј-flanking region of the human MPO gene. In fact, the deletion of about 100 base pairs in the proximal promoter region of the mouse MPO gene almost completely abolished the enhancer activity (16). These results suggest that several cis-elements cooperatively regulate the expression of the MPO gene. To fully understand the cisregulatory elements in the MPO gene expression, it will be necessary to examine linker scanning mutations in this region.
The DNA element recognized by NF/G-CSF contains a CCAAT box, to which a family of transcription factors (CCAATbinding proteins) are known to bind. Purification of the NF/G-CSF from FD62M cells and immunological analysis with anti-NF-Y serum indicated that NF/G-CSF is NF-Y, which is one of the CCAAT-binding proteins (36,38). NF-Y was originally identified as a heterodimer, since the purified NF-Y seemed to consist of two subunits (36,43). However, subsequent cloning of rat NF-Y cDNAs and reconstitution with the recombinant proteins revealed that the functional rat NF-Y consisted of three subunits (38). Here, we showed that functional mouse NF-Y is also composed of three subunits (NF-YA, NF-YB, and NF-YC) and that they are indispensable for binding to the GRE-1 as well as NFY-1 elements. This finding is consistent with the recent report that the yeast homologues of NF-YA and NF-YB (HAP-1 and HAP-2) require a third component (HAP-5) to bind the CCAAT element (44).
The CCAAT motif to which NF-Y binds is present in the upstream region of many other eukaryotic genes (33,(45)(46)(47). NF-Y works as a transcriptional activator in a reconstituted in vitro transcription system using the various promoters (46). Accordingly, the expression of NF-Y mRNA was detected ubiquitously (36). This is consistent with our observation that the DNA binding activity of NF/G-CSF was detected in all cell lines tested (data not shown). These results suggest that the G-CSFinduced and IL-3-repressive expression of the MPO gene is not just regulated by the binding of NF-Y to the GRE; it is possible that post-translational modification (i.e. phosphorylation) of NF-Y by G-CSF or IL-3 modulates the ability of NF-Y to transactivate the MPO promoter, as found in CCAAT enhancerbinding protein family members (48,49). Alternatively, G-CSF stimulation may activate another transcription factor(s), with which NF-Y cooperatively regulates the MPO gene expression. In this regard, it is noteworthy that direct or indirect interactions of NF-Y with other transcription factors have been suggested for several other promoters (41,50,51). Further characterization of the promoter region of the MPO gene and investigation into post-translational modification of NF-Y upon G-CSF or IL-3 stimulation will give further insights into the regulatory mechanism of MPO gene expression.