Binding of THZif-1, a MAZ-like zinc finger protein to the nuclease-hypersensitive element in the promoter region of the c-MYC protooncogene.

A detailed analysis is reported of the binding of the zinc finger protein THZif-1 to the nuclease-hypersensitive element (NHE) in the promoter region of the c-MYC gene using the electrophoretic mobility shift assay and a series of mutants of a fusion protein composed of glutathione S-transferase and THZif-1. The THZif-1 protein bound specifically to the single-stranded (ss) pyrimidine-rich DNA of the NHE (ss c-myc NHE-C) with an apparent dissociation constant (Kd (app)) of 0.077 μM. By contrast, no binding to the single-stranded purine-rich DNA of the NHE (ss c-myc NHE-G) was detected. Moreover, the binding affinity of THZif-1 protein was 2-fold higher for the single-stranded 5-methyl-2′-deoxycytidine derivative of NHE (ss c-myc NHE-me5C) than for the unmethylated NHE. In the case of the binding of THZif-1 to methylated double-stranded (ds) NHE (ds c-myc NHE-me5CG), no significant binding to the DNA was observed. The decrease in binding to DNA of THZif-1 was significant in the case of mutated ds c-myc NHE, in which more than two sites of deoxycytidine residues were methylated. However, the binding affinity of THZif-1 protein for methylated and for unmethylated triple-helical DNA of the NHE was almost identical. Moreover, the domain of the THZif-1 protein that made the major contribution to binding to ss c-myc NHE-C or ss c-myc NHE-me5C corresponded to the amino-terminal second zinc finger motif. Taken together, the results indicate that the THZif-1 protein exhibits preferential DNA-binding activity with ss c-myc NHE-C, ds c-myc NHE-CG, and ts c-myc NHE but not with ss c-myc NHE-G and ds c-myc NHE-me5CG in vitro.

c-myc gene is a member of a family of genes with basic, helixloop-helix, and leucine zipper domains, all of which encode sequence-specific DNA-binding proteins (5). The c-Myc protein forms a heterodimer with the Max protein that is required for the oncogenic activity of c-myc. The Myc-Max complex recognizes the core sequence CACGTG (5).
Evidence for complementary, but not fully redundant, activities among members of the Myc family (c-, N-, and L-myc) comes from studies of phenotypes of Myc-deficient mice. Although inactivation of c-myc and N-myc results in death of embryos by mid-gestation (6 -9), a phenomenon that demonstrates that each is required for normal development, the survival of the mutant embryos to such a late stage of embryogenesis suggests that members of the c-myc family of genes might have overlapping functions at early, but not later, stages of development. At mid-gestation in the mouse, enhanced expression of c-myc is correlated with active proliferation, and downregulation is associated with mitotic arrest and the onset of differentiation (10). Furthermore, in most experiments with cultured cells and transgenic mice, forced expression of c-myc prevents withdrawal from the cell cycle and inhibits differentiation (1)(2)(3)(4)(5), indicating that down-regulation of c-myc is required for mitotic arrest and terminal differentiation in the various cell lineages.
It has also been suggested that transcription of c-MYC might be regulated at the level of template structure (17)(18)(19)(20). A colinear triplex formed between a site-specific oligodeoxynucleotide and c-MYC duplex DNA at Ϫ150 base pairs (bp 1 ; counted from the P1 initiation site) can repress transcription of c-MYC from the P2 promoter in vitro (18 -21). This result suggests that the region around Ϫ150 bp of the c-MYC promoter, which is rich in purine/pyrimidine sequences and is also hypersensitive to nuclease (we named it a nuclease-hypersensitive element III1 (NHE; 3, 12, 14, 17, 19 -26), is important for the transcription of the c-MYC gene. This region is known as a putative intra-or intermolecular triple helix-forming element (18 -20, 27) that might serve as a target of a specific transcription factor(s) needed for P2-directed expression. Previous studies on triplex formation by polypurine/polypyrimidine regions in gene promoters demonstrated the ability of a pyrimidine-rich third strand to form Hoogsteen hydrogen bonds with the purine-rich acceptor strand in the major groove of the target duplex (17). The third strand in pyrimidine-purine-pyrimidine triple helices adapts a parallel orientation relative to the polypurine strand of the duplex (28). The stability of a pyrimidine-purine-pyrimidine triplex is dependent on nonphysiological acidic conditions required by the C ϩ GC triplexes that are responsible for sequence specificity (29 -31). In contrast, previous studies demonstrated triplex formation by a purine-rich or mixed purine/ pyrimidine third strand that involved GGC and AAT or TAT triads at physiological pH (32)(33)(34). These triple helices arise through reverse Hoogsteen bonding, which results in the antiparallel orientation of the third strand (32,35). Consistent with this view, it has been reported that this region of the c-MYC protooncogene exerts a modest stimulatory effect on P2-directed transcription and plays a much more important role in utilization of the P1 promoter (15,36,37). Moreover, in a recent report, Arcinas and Boxer (38) demonstrated that the nucleasehypersensitive sites III1 and III2 in the 5Ј end-flanking region of the c-MYC gene exhibit a decrease in sensitivity to DNase I during differentiation of HL60 cells. Thus, the changes in DNase I-hypersensitive sites might be correlated with structural changes in the NHE. Moreover, in vertebrate DNA, the cytosine residues in the dinucleotide sequence CpG and homopurine/homopyrimidine stretches in the flanking regions of genes are often methylated. Such CpG dinucleotides and homopurine/homopyrimidine stretches are often found clustered within so-called CpG islands or homopurine/homopyrimidine regions (29,30,39,40). The presence of unmethylated cytidine residues is indicative of the presence of active genes. The 5Ј ends of most housekeeping genes are, for example, located within unmethylated CpG islands or unmethylated homopurine/homopyrimidine regions (39 -41). It was reported that the methylation of cytosine strongly enhanced the stability of triplex DNA (42) and that 5-methyl-2Ј-deoxycytidine in singlestranded DNA directed the methylation of CpG sites on the same strand (43). The NHE in the c-MYC promoter region can also be considered in these terms. It seems plausible that transcription factors that are involved in the control of the expression of c-MYC during differentiation might bind near these DNase I-hypersensitive sites or that some modification of the NHE DNA might occur during the differentiation of HL60 cells.
Davis et al. (20) reported the association of a ribonucleoprotein (RNP), which appeared to involve RNA-DNA hybridization, with the NHE sequence of c-myc. Postel and co-workers (21,44) demonstrated the binding to c-MYC NHE of the protein, PuF, which they later cloned as PuF/NM23-H2/NDPK-3, a putative suppressor of tumor metastasis (44). The nucleasesensitive element protein (NSEP-1) (45) and the heterogeneous nuclear (hn) RNA protein K, which exhibit overlapping but distinct single-stranded DNA-binding, double-stranded DNAbinding, and RNA-binding specificities, have also been shown to interact with the NHE sequence in the human c-MYC promoter (46 -48). Zinc finger proteins, such as SP1 (49), and cellular nucleic acid-binding protein (CNBP) (50), were also shown to bind to the NHE-III1 site of the human c-MYC gene. Distinct cis-acting elements, designated MEla2, E2F, and MEla1, are required for the optimal initiation of transcription from the P2 promoter (52). The DNase I-hypersensitive site III2 is situated almost exactly in the center of the sequence between the P1 and P2 promoters of the c-MYC gene (20 -25). The zinc finger protein ZF87/Maz (or Pur-1) has been reported to bind to one of these sites, MEla1, in addition to binding to the NHE-III1 site (37,(52)(53)(54)(55). Another zinc finger protein, CCCTCbinding factor or CTCF (51,56), has also been reported to make contact with sequences in GC-rich core regions between the P1 and P2 promoters.
We showed previously that synthesis of the transcription factor THZif-1, a MAZ-like zinc finger protein, can be induced in the nuclei of HL60 cells that have been transformed with antisense c-MYC, and that it serves as a repressor of the endogenous transcription of the c-MYC protooncogene via binding to the NHE element (57)(58)(59)(60). This THZif-1 factor is a zinc finger protein with strong binding affinity for the triple-helical conformation of the NHE of the c-MYC promoter, as well as for the single-stranded and double-stranded NHE (60). In this report, we compared the relative binding activities of THZif-1 protein to the methylated and unmethylated forms of c-myc NHE. Moreover, we identified the region of THZif-1 that binds to the NHEs using deletion and substitution mutants of the zinc finger motifs. The amino-terminal second zinc finger of the THZif-1 protein is required for binding to the methylated or unmethylated, single-stranded pyrimidine-rich NHE. Our results imply a significant role for THZif-1 in the regulated function of the NHE of the c-MYC gene during the growth and differentiation of HL60 cells.

MATERIALS AND METHODS
Cell Culture, Transfection, and Assay of CAT Activity-HL60 cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum (Life Technologies, Inc.) and 60 g/ml kanamycin monosulfate (Sigma). Transfection for long term expression was carried out as described previously (61,62). Assays of chloramphenicol acetyltransferase (CAT) activity were performed as described elsewhere (61)(62)(63). Each thin-layer chromatography plate was exposed to RX film (Fuji, Tokyo, Japan). The extent of conversion of chloramphenicol to its acetylated form was determined with a Bio-Imaging analyzer (model BAS 2000; Fuji). ␤-Galactosidase activity was assayed as described by Kimura et al. (61). The ratio of CAT activity to that of ␤-galactosidase was used for normalization of results.
PCR-Conditions for all PCRs described herein were the same, with the exception that concentrations of Mg 2ϩ ions were varied from 2.0 to 2.5 mM. The conditions for PCR were an initial denaturation for 2 min with subsequent denaturation for 1 min at 94°C, annealing for 1 min at 56°C, and extension for 3 min at 72°C.
Purification of Fusion Proteins-Derivatives of pGEX-THZif-1 were introduced into Escherichia coli AD202. All fusion proteins were induced with 1.0 mM isopropyl-␤-D-thiogalactopyranoside when the absorption of 595 nm (A 595 ) of the culture had reached 0.6 -0.8. After induction for 3 to 4 h at 30°C, bacteria were harvested by centrifugation at 4,000 ϫ g for 10 min. Bacterial pellets were resuspended in 15 ml of phosphate-buffered saline (PBS) with 1 mM phenylmethylsulfonyl fluoride, 2% Nonidet P-40, and 2% Tween 20, and then cells were lysed by treatment with 200 g/ml lysozyme at room temperature. Bacterial DNA was eliminated by treatment with 15 g/ml DNase I in the presence of 20 mM MgCl 2 and 2 mM MnCl 2 for an additional 5 min at room temperature. The lysates were sonicated three times for 20 s each with a UR-20P Sonicator (Tomy Seiko Co., Tokyo, Japan) at 40% output power. Lysates were clarified by centrifugation at 12,000 ϫ g for 20 min at 4°C. The supernatants were incubated with 300 l of packed glutathione-agarose beads (Pharmacia) for at least 2 h at 4°C. Beads were then washed in a column with PBS until A 280 was less than 0.002, and the protein was eluted with 10 mM glutathione (Sigma) in 50 mM Tris (pH 8.0). The concentrations of protein were estimated with a protein assay kit from Bio-Rad with bovine serum albumin as the standard.
Preparation of Single-stranded, Double-stranded. and Triplestranded DNA Probes-DNA probes for electrophoretic mobility shift assays (EMSA) were prepared, as described by Boles and Hogan (17), by incubating trace amounts of the 32 P-radiolabeled purine (or pyrimidine) strand with increasing concentrations of the pyrimidine-rich or purinerich strand of c-myc NHE in 10 mM Tris (pH 7.8) and 5 mM MgCl 2 at 4°C for 60 min. Samples were subjected to electrophoresis through polyacrylamide gels prepared with 89 mM Tris (pH 7.4), 89 mM boric acid, and 5 mM MgCl 2 . Gels were then dried and autoradiographed. Each single-stranded, double-stranded, and triple-stranded DNA probe was cut out from gels, eluted in 10 mM Tris (pH 7.8), and 5 mM MgCl 2 .
The Electrophoretic Mobility Shift Assay (EMSA)-The EMSA was performed basically as described by Durland et al. (34). The 32 P-radiolabeled oligodeoxynucleotide probe was incubated in a reaction buffer (50 l) that consisted of 10 mM Tris (pH 7.8), 5 mM MgCl 2 , 1 mM spermidine, 10% sucrose, 25 M ZnCl 2 , 5 mM dithiothreitol, and 1 g poly(dI-dC) (Pharmacia). The GST-THZif-1 fusion protein was added to the reaction buffer at the indicated amount, and the mixture was incubated for 60 min at 4°C. In competition experiments, indicated amounts of unlabeled oligomers were added, and incubation was continued for a further 60 min at 4°C. Samples were subjected to electrophoresis at 4°C at a constant current of 10 mA on polyacrylamide gels (4 or 12%) in 50 mM Tris-based buffer as described above. Gels were then dried and autoradiographed. Similarly, the rabbit polyclonal THZif-1 specific antiserum 2 and preimmune serum were used for the supershifting experiment of EMSA as described elsewhere (65). Effect of proteinase K digestion on DNA-protein complex with the THZif-1 protein and c-myc NHE was examined as described by Fang and Cech (66).
Preparation of Cell Extracts and DNA-binding Assay-After washing with phosphate-buffered saline (PBS(Ϫ)), cells were harvested and lysed in 100 l (per 0.1 g of cells, wet weight) of lysis buffer that contained 400 mM KCl, 10 mM Hepes (pH 8.0), 15% (w/v) glycerol, 1 mM dithiothreitol, 0.1 mM phenylmethanesulfonyl fluoride (Sigma), and 2.5 g/ml each leupeptin, aprotinin, and pepstatin (all from Sigma). After centrifugation at 40,000 ϫ g at 4°C for 15 min, each supernatant was divided into aliquots and stored at Ϫ70°C as described elsewhere (65). Cell extracts were incubated on ice with 2 g of poly(dI-dC) (Pharmacia) in 20 l of binding buffer (10 mM Hepes (pH 8.0), 4 mM MgCl 2 , 1 mM dithiothreitol, and 15% glycerol). After 10 min, the 32 P-labeled probe (c-myc NHE-CG) and competitor DNAs were added, and the incubation was continued for 20 min. The reaction mixture was subjected to electrophoresis on a 5% polyacrylamide gel in 0.5 ϫ TBE (44.5 mM Tris, 44.5 mM boric acid, and 1.25 mM EDTA). The gel was dried and exposed to x-ray film at Ϫ80°C. Monoclonal antibodies specific for SP-1 and for c-JUN (Santa Cruz Biotech, Santa Cruz, CA) and polyclonal antiserum specific for ZF87/Maz 2 were used for the "supershifting" experiment.

THZif-1 Is a Maz-like Zinc-Finger Protein-
We reported previously that the NHE sequence in the promoter region of the c-MYC protooncogene is the target site of DNA-binding proteins that are induced in the antisense c-MYC transformed cell line AM93-4-12 (57)(58)(59). We purified a DNA-binding protein specific for this NHE from a nuclear extract of AM93-4-12 cells and isolated a recombinant cDNA clone (for THZif-1, triple helix-binding zinc-finger protein-1) of the corresponding gene (60). The deduced THZif-1 is a polypeptide of 253 amino acids with a molecular mass of 27,830 kDa. Nucleotide sequencing of the cDNA for THZif-1 demonstrated the extremely high degree of similarity to the Myc-associated zinc finger protein ZF87/ Maz and to another zinc finger protein, Pur-1, that binds to purine-rich sequences (98.2 and 96.9% homology, respectively, at the amino acid level; [53][54][55]. It appears that THZif-1 is a member of the ZF87/Maz (or Pur-1) family. However, the DNA sequence that encodes the second zinc finger motif is different from that of the gene for ZF87/Maz (33.3% homology at the amino acid level). Therefore, THZif-1 seems to be a protein that is similar to but distinct from ZF87/Maz. This conclusion is reflected by the difference in size of the corresponding mRNAs (1.6 versus 2.6 kilobase pairs) and the distribution of THZif-1 mRNA in human tissue. Furthermore, a major difference is that the THZif-1 protein regulates the expression of the c-MYC gene in a negative manner, whereas the ZF87/MAZ protein up-regulates the expression of this gene (54,55). However, we cannot rule out the possibility that THZif-1 might be a product of a transcript that is differentially spliced after transcription from the same genomic sequence as that encoding ZF87/Maz.
THZif-1 Represses the NHE-mediated Expression of the c-MYC Gene-In an attempt to examine the intrinsic activity of c-myc NHE during the differentiation of HL60 cells, we compared the CAT activities of c-MYC promoter-CAT fusion constructs with and without NHE (pMycRNCAT versus pMycRN⌬55CAT; Fig. 1A). A significant reduction in the promoter activity of the c-MYC gene was observed 2 days after the short of treatment with 12-O-tetradecanoyl-phorbol-13-acetate (TPA; lanes 9 -12). This reduction was similar to that in the case of pMycCAT2 (lanes 1-4). However, the deletion reporter construct of c-myc NHE did not show such a repression (lanes 5-8). Thus, the intrinsic activity of the c-myc NHE reflects the fact that the c-myc NHE is one of the major regulatory elements in the expression of the c-MYC gene during the differentiation of HL60 cells. Therefore, the characterization of factors that bind to c-myc NHE is important if we are to understand the mechanism of repression of the c-MYC promoter during the differentiation of HL60 cells (52)(53)(54)(55)(56)(57)(58)(59).
To determine whether the THZif-1 protein was responsible for the negative regulation of transcription of the c-MYC gene, we analyzed the effect of the THZif-1 protein on the promoter activity of c-MYC gene by introducing CAT reporter fusion constructs of c-MYC with or without NHE and an expression plasmid pTHZif-1 or its mutant plasmid pTHZif-1m of second zinc-finger motif, into HL60 cells. We also constructed a promoter for the gene for thymidine kinase (tk) from herpes simplex virus-1 with or without the NHE fused to the gene for CAT and examined CAT activities (61,64). These plasmids were introduced into HL60 cells and isolated each stable clone. We assayed the reporter activities of these clones. As shown in Fig. 1B, CAT activity of the c-MYC promoter-CAT with NHE (pMycRNCAT) was repressed from 8-to 9-fold by THZif-1, whereas that of c-MYC promoter-CAT without NHE (pMycRN⌬55CAT) was not (lanes 1, 2, 4, and 5). Moreover this repression activity was not detected in the case of THZif-1 mutant construct of second zinc finger motif (lanes 3 and 6). These results indicate that at least the second zinc finger motif of THZif-1 protein is corresponded to the regulated expression of c-MYC-CAT reporter gene in HL60 cells. Similarly, we also observed a reduction in CAT activity in the case of the tk promoter-CAT with NHE (ptk-NHECAT) (lanes 7, 8, 10, and 11). We observed that the CAT activity of pMycRNCAT was reduced in the differentiated HL60 cells (7-fold reduction; lanes 13 and 14) and, however, the CAT activity of pMycRN⌬55CAT was not affected (lanes 16 and 17). We detected the similar reduction of ptk-NHECAT in the differentiated HL60 cells. It is clear that this repression was also detected at a significant level in U937 cells but not in F9 or P19 embryonic carcinoma cells (data not shown). These variations in CAT activity might reflect variations in endogenous levels of THZif-1 protein in the  (57), pMycRNCAT (58), pMycRN⌬55CAT (58) or pSV00CAT, 3 g of pSV2neo (61), and 5 g of pCH110 (61) were used to transfect HL60 cells to obtain the respective stably transformed clones. These clones were treated with TPA (Sigma) at 10 Ϫ6 M as described elsewhere (57)(58)(59), and respective cell lysates prepared 24, 48, and 120 h after treatment with TPA were analyzed for CAT activity. The percent conversion of chloramphenicol to its acetylated form is indicated. Results typical of one of five independent assays are shown. B, the second zinc finger motif of pTHZif-1 is responsible for the c-myc NHE-mediated repression of transcription of the c-MYC gene. Five g of pMycRNCAT (58), pMycRN⌬55CAT (58), ptkCAT (64), or ptk-NHE-CAT (64), 3 g of pSV2neo (61), and 5 g of pCH110 (61) were used to cotransfect HL60 cells together with (lanes 1, 4, 7, 10, 13, 16, 19, and 22) or without (lanes 2, 5,8,11,14,17,20, and 23) the pCMV-THZif-1 expression plasmid (5 g) or with (lanes 3, 6, 9, 12, 15, 18, 21, and 24) the plasmid that encoded the protein with a mutant second zinc finger, pTHZif-1m (5 g). After incubation with G418 (500 g/ml) for 4 weeks, the respective stably transformed clones were isolated. Cell lysates were prepared and analyzed for CAT activity as described under "Materials and Methods." The percent conversion of chloramphenicol to its acetylated form is indicated. Results typical of one of five independent assays are shown. C, effects of THZif-1-specific antibodies on the DNAprotein complex in the DNA-binding assay. Nuclear extracts (4 g) from TPA-treated differentiated HL60 cells were incubated on ice for 30 min without (lane 2) or with 2 g of affinity-purified THZif respective cells (data not shown). These data suggest that at least THZif-1 protein, one of the transcription factors that bind to the NHE, is involved in the NHE-mediated negative regulation of the transcription of the c-MYC gene.
THZif-1 Protein Is Included in the Protein-DNA Complex with c-myc NHE-To assess the possibility that the THZif-1 protein binds to the c-myc NHE during the differentiation of HL60 cells, we performed a DNA-binding assay with oligodeoxynucleotides that corresponded to ds c-myc NHE-CG as a probe (Fig. 1C). The binding of nuclear proteins to c-myc NHE-CG was detected as a sequence-specific protein-DNA complex in extracts of differentiated HL60 cells (lane 2). The incubation of extracts with antiserum specific for THZif-1 resulted in the supershifting of the DNA-protein complex (lane 3). The supershifted band disappeared in the presence of the GST-THZif-1 fusion protein (lane 4). Moreover, control preimmune serum did not cause any change in the complex (lane 5). Furthermore and to our surprise, the antiserum specific for SP-1 also shifted the DNA-protein complex (lane 6). By contrast, antiserum specific for c-JUN and ZF87/Maz did not affect the binding to DNA (lanes 7 and 8). However, we did not detect any involvement of the THZif-1 protein in the formation of the DNA-protein complex with c-myc NHE when we used nuclear extracts of undifferentiated HL60 cells. This observation is consistent with previous data (58,59), and these observations, taken together, indicate that the specific shifts observed in our DNA-binding assays were due to the THZif-1 protein or a closely related protein.
To address the question of whether the negative expression of c-MYC by the THZif-1 protein is primarily due to the binding of THZif-1 to the NHE sequence, we carried out a DNA-binding study using the single-stranded (ss), double-stranded (ds), and triple-stranded (ts) forms of the NHE of the c-MYC gene as DNA probes. The NHE of the c-MYC gene consists of pyrimidine-rich sequences in one strand that seem to be sensitive to de novo methylation and purine-rich sequences in the other strand which remain unmethylated. Therefore, it seems highly likely that the expression of the c-MYC gene is regulated by the methylation of the pyrimidine-rich strand of the NHE. Thus, we also attempted to study the effect of methylation of the cytosine residues in the NHE of the c-MYC promoter on the DNA-binding activity of the THZif-1 protein.
Binding to Single Strand Unmethylated or Methylated NHE in Vitro-To characterize the DNA-binding properties of THZif-1, we performed an EMSA study in the presence of Mg 2ϩ ions (5 mM), using the recombinant GST-THZif-1 fusion protein and a single-stranded (ss) NHE oligodeoxynucleotide as a DNA probe. To circumvent the insolubility of recombinant intact THZif-1 protein in E. coli cells, we used the fusion construct with pGEX to produce a soluble protein for our assays. As shown Fig. 2A, the THZif-1 protein had the capacity to bind to the pyrimidine-rich strand of NHE (ss c-myc NHE-C). The GST protein itself did not bind to any of the ss, ds, or ts c-myc NHE oligodeoxynucleotides used in this study (data not shown). It was also clear that a shifted DNA-protein complex appeared when the concentration of THZif-1 protein was increased (position B1, lane 2). When the dose of THZif-1 protein was increased further, a new shifted band of a DNA-protein complex, which seemed to be the complex of DNA with a dimer of the THZif-1 protein, appeared (position B2, lanes 3 and 4). This result is consistent with the previous results that the dimer of THZif-1 protein had higher DNA-binding activity than that of the monomeric form of THZif-1 protein (data not shown). Moreover, we were surprised that the binding of THZif-1 to ss methylated NHE-C (ss c-myc NHE-me 5 C) resulted in 2-fold higher affinity than that to ss unmethylated NHE-C at all positions (compare lanes 2-4 with lanes 6 -8; Table II). In contrast, the THZif-1 protein did not bind to the purine-rich strand of NHE (ss c-myc NHE-G) (lanes 10 -12). When the reaction products were treated with proteinase K, the shifted bands were lost in the case of both the methylated and the unmethylated DNA probe (data not shown). In addition, the antiserum raised against THZif-1 caused supershifting of the DNA-protein complex (Fig. 2B). By contrast, the preimmune serum had no such activity (Fig. 2B). Thus, we concluded that these protease-sensitive bands really represented the complexes between the THZif-1 protein and its target DNA.
Sequence Specificity of DNA Binding-Under the same conditions as the EMSA for which results are shown in Fig. 2, we performed a competition experiment using various single strand unmethylated oligodeoxynucleotides that corresponded to the known CT-elements reported by other investigators (45). As shown in Fig. 3A, the non-radiolabeled oligodeoxynucleotides that corresponded to the pyrimidine rich-strand of c-myc NHE (ss c-myc NHE-C) competed for binding to the THZif-1 protein (lane 3). In addition, the methylated oligodeoxynucleotide that corresponded to the pyrimidine-rich strand of c-myc NHE (methylated to yield the me 5 C derivative of the pyrimidine-rich strand of c-myc NHE, ss c-myc NHE-me 5 C) as a competitor gave a similar result (lane 4). We mutated the oligodeoxynucleotides by changing pyrimidine-rich sequences of the NHE to purine residues and analyzed their effects. Mutated oligodeoxynucleotides, such as ss c-myc NHE m1C and m2C, did not compete for binding to THZif-1 with radiola- beled ss c-myc NHE-C. The oligodeoxynucleotides corresponding to other NHEs, included K-ras and EGFR (K-ras NHE-C; EGFR NHE-C), clearly competed for binding to THZif-1 with ss c-myc NHE-C (lanes 7 and 8). Similarly, we also examined competition with the radiolabeled methylated pyrimidine-rich strand of DNA (ss c-myc NHE-me 5 C) as a probe. As shown in Fig. 3B, the competition pattern obtained with each inhibitor was almost the same as in the case with the DNA probe of ss c-myc NHE-C, with the exception of patterns obtained with K-ras and EGFR NHE-Cs. The competition by the oligodeoxynucleotides that corresponded to K-ras and EGFR NHEs in the formation of DNA-protein complexes with ss c-myc NHEme 5 C as DNA probe was weak as compared with that with the ss c-myc NHE-C as probe DNA (lanes 7 and 8 in Fig. 3, A and  B). The competition with the methylated forms of K-ras and EGFR NHEs was complete when each homologous DNA was used as probe (data not shown). Thus, the DNA-binding specificity of the THZif-1 protein to the NHEs of K-ras or EGFR might be slightly different from the binding to the NHE of c-MYC gene. These results imply that the sequence specificity of the binding to DNA of THZif-1 protein is restricted to singlestranded pyrimidine-rich sequences of c-myc NHE, irrespective of the presence or absence of methylation. The neu-1, neu-2, and Y-box NHEs, as well as poly(dC) [12][13][14][15][16][17][18] and poly(dT) 12-18 , did not compete in the binding of GST-THZif-1 protein to ss c-myc NHE-C (45; data not shown). Thus, it appears that the THZif-1 protein binds preferentially to the pyrimidine-rich ss DNA of the c-myc NHE (ss c-myc NHE-C) in a sequence-specific manner.
Binding to Double-stranded Unmethylated or Methylated cmyc NHE-We prepared unmethylated DNA probes from the double-stranded (ds) NHE of the c-MYC gene, whose pyrimidine-rich strand or the purine-rich strand had been radiolabeled. Each probe was isolated from a nondenaturing gel as described under "Materials and Methods." The DNA-binding activity of THZif-1 protein was analyzed by the EMSA. As shown in Fig. 4A, a shifted band (band B) appeared, and the formation of the DNA-protein complex was dose-dependent in the case of ds c-myc NHE-CG, of which pyrimidine-rich strand was radiolabeled (lanes 1-4). However no shifted bands were detected with the tested range of concentrations of THZif-1 protein in the case of ds c-myc NHE-me 5 CG, of which all cytosine residues were methylated (lanes [5][6][7][8]. When the ds c-myc NHE-CG, of which the purine-rich strand had been radiolabeled, was used, the similar shifted band (band B) and the formation of the DNA-THZif-1 protein complex appeared (Table I). Thus we conclude that the THZif-1 protein has an ability to bind the ds c-myc NHE-CG.
We next examined the sequence specificity of the binding of THZif-1 protein to ds c-myc NHE-CG. As shown in Fig. 4B, oligodeoxynucleotides with the same sequence as ds c-myc NHE-CG competed for binding of THZif-1 protein (lanes 4 and 5) while the mutant oligodeoxynucleotides of ds c-myc NHE-CG did not have such inhibitory activity. It is noteworthy that ds NHE-CG of K-ras competed for the binding of THZif-1 protein with ds c-myc NHE-CG, but this was not the case for ds NHE-CG of EGFR (lanes 6 and 7). The differential binding of THZif-1 to the ds NHEs might have been due to the higher degree of similarity between the sequence of the NHE-CG of the K-ras gene and that of the c-MYC gene than is the case for EGFR NHE-CG. Antiserum against THZif-1 caused supershifting of the DNA-protein complex, an indication that the THZif-1 protein is included in this DNA-protein complex (Fig. 4C). Therefore, the sequence specificity of the binding to DNA of the THZif-1 protein was more restricted in the case of the ds NHE-CG of the c-MYC gene, even though the binding affinity for the ds NHE was lower than for the ss c-myc NHE-C (Table  II). The specificity of the binding of the THZif-1 protein to the ds NHE-CG might be slightly different from that of ss c-myc NHE-C.
We then examined the DNA-binding activity of the THZif-1 protein to various forms of double-stranded c-myc NHE-CG, in which the deoxycytidine residues were methylated at different positions (Fig. 5, Table I). To our surprise, the ds c-myc NHE probes, in which deoxycytidine residues at positions 3 or 33 were methylated, bound THZif-1 (Fig. 5). However, methylation at additional positions of NHE resulted in the absence of specific DNA-binding activity of THZif-1 (Fig. 5). These results are summarized in Table I. These data indicate that the effect of methylation of deoxycytidine residues at at least two positions, namely 3 and 6 (or 29 and 33), in c-myc NHE resulted in a decrease in the DNA-binding activity of the THZif-1 protein (Table I and Fig. 5).
We compared the K d values of the THZif-1 protein for its binding to various ss, ds, or ts forms of the c-myc NHE (Table  II). It was clear that the THZif-1 protein had lower affinity for ds c-myc NHE-CG. The binding avidity of THZif-1 protein for ds c-myc NHE-CG was about 50-fold lower than that for ss c-myc NHE-me 5 C (all positions were methylated). Therefore, the methylation of the ds NHE of the c-MYC gene might be critical to the regulation of the affinity of the DNA for the THZif-1 protein, at least under our assay conditions. Potential Triple-helical Structure of the NHE of the c-MYC Gene-We examined the structural differences associated with the formation of the triple-helical structure CGG with respect to the effect of methylation of the pyrimidine-rich strand of c-myc NHE (CGG versus me 5 CGG). We radiolabeled the oligodeoxynucleotides that corresponded to the pyrimidine-rich strand and the methylated me 5 C-derivatives of NHE-C and tested the capacity for triplex formation at a given concentration of the pyrimidine-rich strand, after addition of increasing amounts of the oligodeoxynucleotides that corresponded to the purine-rich strand of the NHE-G in the presence of 5 mM Mg 2ϩ ions (Fig. 6). In both cases (Fig. 6, A and B) slowly migrating bands corresponding to double-stranded and triple-stranded NHEs appeared, and that formation was dependent on the concentration of the unlabeled purine-rich oligodeoxynucleotides that corresponded to NHE-G. However, in the absence of Mg 2ϩ ions, we did not detect the triple-helical NHE on gels, and we only found the shifted band corresponded to the ds NHE (Fig. 6, C and D). Then we radiolabeled the purine-rich NHE-G of the c-MYC gene, and the pyrimidine-rich strand of the NHE-C or the methylated pyrimidine-rich NHE-me 5 C was then added to the reaction mixture at a given concentration of the radiolabeled purine-rich NHE-G, in an attempt to produce a triplex such as CGC ϩ or CGC me at both acidic and neutral pH. We detected only the ds DNA-protein complex and no triple-helical DNA-protein complex, no matter what the acidic or neutral pH (data not shown). Thus, we concluded that the   NHE of the c-MYC gene was able to form a Mg 2ϩ -dependent triplex, such as CGG and me 5 CGG, irrespective of the methylation of NHE-C. There was no difference in the extent of the migration of the triple complex that consisted of unmethylated NHE or methylated NHE of c-MYC gene on a gel (Fig. 6). Moreover, we did not detect any significant difference in K d values for binding of THZif-1 protein to the ts DNA probe when the probe was in an unmethylated or a methylated form (Table II).
Domains of the THZif-1 Protein That Bind to Unmethylated and Methylated ss-NHEs-In view of observations that the binding to DNA of the THZif-1 protein is specific for ss c-myc NHE-C and for the methylated derivative of the NHE of the c-MYC gene (ss c-myc NHE-me 5 C), we next attempted to identify the domain(s) of the THZif-1 protein required for the specific binding to the DNA. We constructed a set of serially deleted mutant GST-fusion proteins with removal of each zinc finger motif of THZif-1 from either the amino terminus or the carboxyl terminus, and we examined their binding to the ss pyrimidine-rich NHE-C in both its methylated and its unmethylated form (Fig. 7). We compared the radioactivity of the shifted bands of DNA-protein complexes relative to the total radioactivity of the input DNA probe and calculated the relative DNA-binding efficiency based on the binding avidity of clone 1 as a control, which encoded the intact recombinant GST-fusion protein with five zinc fingers and for which the results are indicated as 100%. With ss c-myc NHE-C as probe, the mutant construct without the fifth finger motif in the carboxyl-terminal region had about half the DNA-binding activity of the intact GST-THZif-1 fusion protein (see Fig. 7, lines  2 and 3). The mutant without the carboxyl-terminal four fingers exhibited an approximately 5-fold reduction in DNA-binding activity, as compared with that of clone 5 (Fig. 7, lines 5 and  6). We next examined a series of constructs with deletions from the amino terminus of the THZif-1 protein. Loss of the second amino-terminal zinc finger resulted in a significant reduction in the binding to ss c-myc NHE-C (Fig. 7, lines 9 and 10). In the case of amino-terminal deletions, we did not detect any contribution of the fifth finger motif of the THZif-1 protein to DNAbinding activity. Thus, the second amino-terminal zinc finger motif seemed to play a crucial role in the binding to ss c-myc NHE-C. The methylated NHE-me 5 C was used as the DNA probe in the same assays, and slightly different specificities of binding to DNA were obtained with various mutant proteins. The carboxyl-terminal deletion of three zinc fingers caused a 50% reduction in binding to DNA (Fig. 7, lines 4 and 5), and the deletion of four carboxyl-terminal zinc fingers resulted in a 4-fold reduction in DNA-binding activity as compared with that of clone 5 (lines 5 and 6). The results with the amino-terminally deleted GST-THZif-1 protein yielded a similar conclusion (lines 9 -11). The second and third zinc fingers of the THZif-1 protein, counted from the amino terminus, and in particular the second finger motif, were important for binding of THZif-1 protein to ss c-myc NHE-me 5 C. Therefore, it appears that the second amino-terminal zinc finger is very important for binding of the THZif-1 protein to the methylated and unmethylated ss c-myc NHE-C of the c-MYC gene in vitro. Moreover, the contributions of other zinc finger motifs, such as the fifth zinc finger, to the binding to ss c-myc NHE-C and the third zinc finger to the binding to ss c-myc NHE-me 5 C were smaller. The deletion mutant without the second zinc finger and the mutation of the second zinc finger motif confirmed the significant reduction in FIG. 5. Binding of the THZif-1 protein to various 5-methyl-2deoxycytidine derivatives of ds c-myc NHE-CG. Eight g of the recombinant fusion protein GST-THZif-1 were incubated with the indicated 5-methyl-2Ј-deoxycytidine derivatives with methylation at various positions in c-myc NHE. The reaction mixtures conformed to radiolabeled c-myc NHE-G and unlabeled c-myc NHE-C oligodeoxynucleotides and were incubated for 60 min at 4°C. The resultant DNA-protein complexes were resolved in nondenaturing 5% polyacrylamide gels. In some cases, the unlabeled double-stranded oligodeoxynucleotides, as competitors, were incubated at 100-fold molar excess with GST-THZif-1 protein (4 g) after 10 min at 4°C, and then the 32 P-radiolabeled 5-methyl-2Ј-deoxycytidine derivatives of ds c-myc NHE-CG were added as DNA probes, and mixtures were incubated for another 50 min at 4°C. Lanes 1 and 2, without GST-THZif-1 protein; lanes 3 and 4, 4 and 8 g of GST-THZif-1 protein, respectively; lane 5, ds c-myc NHE-CG, and lane 6, ds c-myc NHE-m1CG. Lane 1, the DNA probe was radiolabeled ss c-myc NHE-G; lane 2, the DNA probe was a 5-methyl-2Ј-deoxycytidine derivative of ds c-myc NHE-CG. B(ds), DNAprotein complex. F(ss), the free DNA probe, namely ss c-myc NHE-G; F(ds), the free DNA probe, namely a 5-methyl-2Ј-deoxycytidine derivative with methylation at appropriate positions in ds c-myc NHE-CG. As defined in Table I the extent of the binding to the ss NHE-C DNA in both its methylated and unmethylated forms (Fig. 7, lines 14 and 15). The similar results were obtained in the case of the binding of the THZif-1 protein to the ds c-myc NHE-CG and to the methylated and unmethylated ts c-myc NHE-CG (data not shown).

Regulation of Transcription of the c-MYC Gene by
THZif-1-We isolated a cDNA clone for a protein that binds to CACC-TCC repeated elements, designated THZif-1, which is a member of the ZF87/Maz (Pur-1) family (53)(54)(55). The introduction of the pTHZif-1 plasmid into HL60 cells resulted in the negative regulation of the expression of the human c-MYC gene (see Fig.  1B). The THZif-1 protein with a mutated second zinc finger did not reduce the transcriptional expression of the c-MYC gene (Fig. 1B). Thus, we can conclude that the THZif-1 protein might negatively regulate the promoter activity of the human c-MYC gene. By contrast, ZF87/Maz (Pur-1) has a positive effect on the transcription of the c-MYC gene (54,55). Thus it is clear that THZif-1 is similar but distinct from ZF87/Maz (Pur-1). We observed the phenotypic conversion of HL60 cells that has been transformed with the pTHZif-1 expression vector into normal macrophages, as estimated by the expression of surface markers MAC-1 and CD 14 and the activities of nonspecific esterase and myeloperoxidase, in addition to the regulated expression of G-CSFR and fms mRNAs. By contrast, the introduction of the pTHZif-1m mutant construct did not induce the appearance of the differentiation markers of HL60 cells. 2 Our results indicate that the THZif-1 protein might be a key regulator of the differentiation of the HL60 cells to normal macrophages.
The gel shift assay using nuclear extracts of differentiated HL60 cells demonstrated that the THZif-1 protein is included in protein-DNA complexes with c-myc NHE-GC. As shown in Fig. 1C, the antiserum specific for THZif-1 protein shifted the band of the protein-DNA complex, whereas the antisera specific for c-JUN and ZF87/Maz did not shift the complex of protein and DNA. However, the THZif-1 protein bound weakly to the c-myc NHE when we used extracts of undifferentiated HL60 cells (58,59, data not shown). Thus, it is clear that the THZif-1 protein can bind to the NHE of the human c-MYC gene in differentiated HL60 cells. We next examined the binding of the recombinant THZif-1 protein to the NHE sequence to determine the DNA-binding specificity and to characterize the effect of the methylation of the cytidine residues in c-myc NHE on the binding to DNA of the THZif-1 protein.
Effect of Methylation of NHE on Binding of GST-THZif-1-We found that the recombinant GST-THZif-1 protein bound to the single-stranded pyrimidine-rich NHE (ss c-myc NHE-C) of the c-MYC gene in a dose-dependent manner (lanes 1-4 in Fig. 2A). By contrast, binding of the THZif-1 protein to the single-stranded purine-rich c-myc NHE-G was not detected (lanes 9 -12 in Fig. 2A). Thus, the DNA-binding specificity of the THZif-1 protein was restricted to the pyrimidine-rich single-stranded NHE of c-MYC gene. Moreover the binding avidity of GST-THZif-1 protein for ss c-myc NHE-me 5 C was about twice that for unmethylated ss NHE-C (lanes 5-8 in Fig. 2A). According to the migration of DNA-protein complexes, the slowest migrating bands seemed to correspond to dimeric proteins in a complex with DNA (see Band B2; lanes 3, 4, 7, and 8).
The binding specificity of THZif-1 protein for both methylated and unmethylated ss NHE-C was almost identical. The competition experiment yielded the same results (Fig. 3, A and B).
We next tested the differential requirements for zinc finger motifs in the THZif-1 protein for binding to methylated or unmethylated c-myc NHE-C. According to the results of the DNA-binding study with methylated and unmethylated NHE-C and the deletion mutants of the GST-THZif-1 fusion protein, the second zinc finger motif of the THZif-1 protein plays a central role in binding to the ss c-myc NHE-C (see Fig.  7). It is consistent with the results of transcriptional repression of THZif-1 on the expression of human c-MYC gene.
The DNA-binding activity of THZif-1 protein with methylated and unmethylated ds NHE of the c-MYC gene was examined. To our surprise, the THZif-1 protein did not bind to the methylated deoxycytidine derivatives at all positions of NHEme 5 CG (see Fig. 4A). However, the exact position of 5-methyl-2Ј-deoxycytidine residues of c-myc NHE in vivo has not been known. Thus, we introduced methyl residues into every cluster of deoxycytidine residues in c-myc NHE and examined the binding of THZif-1 to these methylated sequences. We also radiolabeled the purine-rich c-myc NHE and allowed it to hybridize with the pyrimidine-rich strand to form ds c-myc NHE as a DNA probe. We detected DNA-binding activity of THZif-1 similar to that to ds c-myc NHE-CG (Fig. 5), in which the pyrimidine-rich strand of NHE had been radiolabeled (Fig. 4). Thus, we concluded that the THZif-1 protein binds the ds c-myc NHE-CG DNA. The THZif-1 protein did not bind to c-myc NHE-CG, when at least two deoxycytidine residues in the c-myc NHE had been methylated (see Table I). Therefore, our studies in vitro indicate that the THZif-1 protein has differential DNA-binding activities that depend on the extent of methylation of c-myc NHE-CG.
We present here two tentative models that might explain how the THZif-1 protein recognizes the single-stranded NHE-C FIG. 7. Mapping study of the domains of the GST-THZif-1 protein required for binding to methylated or unmethylated ss c-myc NHE-C. A diagrammatic representation of the GST-THZif-1 fusion protein and the derivatives used in the present study is shown. The full-length cDNA for THZif-1 and the mutant DNA fragments were amplified by PCR as described under "Materials and Methods," and the nucleotide sequences were verified by nucleotide sequencing. The THZif-1 DNA sequence corresponded to the entire coding region of THZif-1. The deletion mutants of this region were generated, and the various pGEX-THZif-1 plasmids were constructed as described under "Materials and Methods." The binding of the protein encoded by clone 1 (GST-THZif-1) to ss c-myc NHE-C or ss c-myc NHE-me 5 C was taken arbitrarily as 100% in each case. The amino terminus and the carboxyl terminus of THZif-1 are indicated by a circle and a square, respectively. and the triplex of NHE of the c-MYC gene. In model A (see Fig.  8A), the double-stranded DNA of NHE (ds c-myc NHE-GC) unwinds and then the THZif-1 protein binds to the c-myc NHE-C strand as either a monomer or dimer. In the case of binding to the methylated strand of c-myc NHE-me 5 C, the THZif-1 protein binds mostly as a dimer to the me 5 C strand of c-myc NHE. This feature is consistent with data that a dimer of the DNA-binding protein functions as a repressor of expression of the c-MYC gene (58,59). In model B (see Fig. 8B), the THZif-1 protein has affinity for an intermolecular triplex of NHE, which is dependent on Mg 2ϩ ions and is formed by reverse Hoogsteen bonding. At this stage, the pyrimidine-rich strand of NHE might be unwound, and the ss DNA-binding protein for the pyrimidine-rich strand of NHE included THZif-1, which bound the pyrimidine-rich strand of NHE-C, to stabilize the molecular structure of the protein-DNA complexes. Of course, we cannot rule out the possibility of another model that explains these results. However, we have no direct information about the tertiary structure and the extent of methylation of c-myc NHE in vivo and about the transcription factors that play a major role in regulating the transcription of the c-MYC gene in particular types of cells.
Previous studies demonstrated the formation of an intermolecular triplex by a purine-rich third strand that involves the GGC triad at physiological pH (21,(32)(33)(34). Such triple helices arise through reverse Hoogsteen bonding, with the resultant antiparallel orientation of the third strand (32,35). We examined the effects of the methylation of the pyrimidine-rich strand of c-myc NHE on the formation of intermolecular triplexes of the CGG and me 5 CGG types. Under our experimental conditions, no differences in the formation of a triplex were observed between methylated and unmethylated pyrimidinerich strands of NHE (see Fig. 6). Thus, we can conclude there is no major difference in the capacity for intermolecular formation of a triplex of c-myc NHE in the presence of Mg 2ϩ ions between the methylated and the unmethylated pyrimidine-rich strand of NHE (Fig. 6). In fact, the THZif-1 protein bound to the triplex NHE in a similar manner irrespective of whether NHE was in the methylated or unmethylated form (see Table II). The apparent dissociation constant of THZif-1 protein was almost the same for c-myc NHE-CGG and c-myc NHE-me 5 CGG (K d ϭ 0.89 versus 0.80 M). It has been known that the presence of 5-methylcytosine in place of cytosine in the third strand of C ϩ GC triple helices increases the apparent pK for triplex formation and permits formation of detectable triplex up to neutral pH (67). Firulli et al. (19,20,27) reported that the DNA sequences of NHEs that form intramolecular triple helices seem to be positive indicators of promoter strength. Transcriptionally active, stable triplexes included C ϩ GC triplets in an intramolecular complex that induced formation of a singlestranded portion and the parallel orientation of polypyrimidine triplex-forming strands. Firulli et al. (21,(32)(33)(34) suggested that one or more of the proteins known to bind to the NHE might function to stabilize such a structure at neutral pH. However, we did not detect the intramolecular formation of a triplex, such as C ϩ GC, under our experimental conditions in vitro, a result that is consistent with those in other reports.
Possible Role for Binding of THZif-1 to c-myc NHE-The capacity of the c-myc NHE to form a triplex to inhibit transcription of the c-MYC gene was first demonstrated by the addition of an oligodeoxynucleotide targeted to the NHE (18). Subsequently, several studies have shown that the transcription of c-MYC is repressed in human cervical carcinoma and ovarian carcinoma cells (21,68), indicating an inhibitory effect on cell proliferation. Previously, our studies indicated that the c-myc NHE might involve the negative regulation of c-MYC gene by introduction of antisense c-MYC gene in HL60 cells. We also detected the similar negative regulation of c-myc NHE during the cell differentiation of HL60 cells (Fig. 1A). Although the mechanism for inhibition of transcription has not been elucidated, it has been suggested that triplex formation prevents the binding of essential regulatory factors (69). Similarly, triplex-induced inhibition of the binding of nuclear factors to an enhancer-promoter element had been demonstrated in other systems, for example in the case of the NFB that binds to the interleukin-2␣ promoter (70), a factor that binds to the Her-NEU promoter (71), and SP-1 that binds to the human Ha-RAS promoter (72,73,74). These studies demonstrated in vitro and in vivo the gene-specific repression of transcription that accompanied formation of triplex targeted to the nuclear proteinbinding sites. However, we found in the present study that the THZif-1 protein has the ability to bind to the triple helix of c-myc NHE. Moreover, the binding avidity of the THZif-1 protein to ss, ds, and ts c-myc NHEs was well correlated with the negative regulation of transcription of the c-MYC gene (60, data not shown) since the THZif-1 protein with a mutant second zinc finger did not have the DNA-binding activity and did not repress transcription of the c-MYC gene (Figs. 1B and 7) (60).
Previous deletion studies of the c-MYC promoter clearly showed that the c-myc NHE is a positive element (3,15,20) that is required for transcription from the c-MYC P1 promoter and that augments expression from P2. Several different candidates have been proposed as factor(s) that actually mediated activation of c-myc NHE, including PuF/NM23-H-2/NDPK-3 (44), NSEP-1 (45), hnRNP K (46 -48), and SP-1 (49). The hnRNP K can bind to either single-stranded or supercoiled target sequences on the "top" strand of the c-myc NHE (46 -48). When binding to a torsionally strained duplex, hnRNP K exposes sites on the "bottom" strand for binding with cellular nucleic acid-binding protein (47)(48)(49)(50). Thus, factors and processes that facilitate binding of hnRNP K will also augment the activity of CNBP (47,50). By contrast, SP-1 (37) antagonizes the DNA-binding interactions of both hnRNP K (Fig. 2) and CNBP, with the effect being profound in the latter case. THZif-1 protein might bind to the pyrimidine-rich strand of c-myc NHE in vivo, as in the case of hnRNP K (Fig. 2), in addition to its DNA-binding activities with the ds and ts c-myc NHEs (Figs. 4 and 5; Ref. 60). Thus, THZif-1 might behave in a similar DNA-binding manner to hnRNP K. However, the biological functions of these proteins are conflicting (46 -48). The configuration of the c-myc NHE should be governed by the intrinsic equilibrium between single, double (75,76), and, probably, triple strands, with constraints imposed by the degree of torsional stress and the relative amounts and affinities of the relevant factors for the DNA.
It has been reported that SP-1 (49), CNBP (50), or ZF87/Maz (52)(53)(54)(55), which are all zinc-finger proteins, in addition to hnRNP K (46 -48), NSEP-1 (45), and PuF/NM23-H-2/NDPK-3 (44) bind to the same sequences of c-myc NHE. Thus, it is possible that the restricted binding of the THZif-1 protein to the c-myc NHE might prevent the binding of factors that are essential for transcription of the c-MYC gene. It might displace such factors within the c-myc NHE sequences or it might modify the tertiary structure of the NHE DNA to inhibit the basal transcription of the c-MYC gene. Arcinas and Boxer (38) reported that the nuclease-hypersensitive site III1 in the NHE of the c-MYC gene exhibits decreased sensitivity to DNase I during differentiation of HL60 cells. It is possible that the THZif-1 protein might bind to the c-myc NHE and prevent digestion by DNase I.
Other relevant cis-acting regions are the GC-rich sequences at the ME1a1 site in the promoter of the c-MYC gene. This site has been shown to enhance the initiation of transcription of the c-MYC gene in vitro (68) and to be essential for initiation at P2 in vivo (26). Moreover, the E2F and the ZF87/Maz (Pur-1) proteins both bind to two of the three distinct elements within the mouse c-MYC promoter that are required for transcription (52,53). The novel triplex-forming sites encompass half of the E2F site and all of the sites of the canonical cis elements of ME1a1. However, repression of transcription of the c-MYC gene might additionally require the melting of the target c-myc NHE or the formation of a triplex among P1 and P2 and the NHE sites. In such a case, the THZif-1 protein might play a role in changing the conformation of the DNA.
It is clear from our DNA-binding study in vitro that the THZif-1 protein has twice the affinity for ss methylated NHEme 5 C as it does for NHE-C and that it has the capacity to bind to the triple-helical conformation of c-myc NHE, irrespective of methylation of the NHE-C or the absence of methylation. By contrast, the THZif-1 protein does not bind to the methylated ds c-myc NHE-me 5 CG. Such differential binding of the THZif-1 protein suggests that the protein might be one of the key regulators of the altered conformation and topology of the DNA in the chromatin, with resultant changes in the rate of transcription of the c-MYC gene in vivo. Further studies are clearly required if we are to understand the molecular basis of the conformation and the extent of the methylation of c-myc NHE and the combinational interactions between the THZif-1 protein, other DNA-binding proteins and the NHE sequence during the differentiation of HL60 cells.