Regulatory function of delta/YY-1 on the locus control region-like sequence of mouse glycophorin gene in erythroleukemia cells.

The far upstream region (−1.2-0.9 kilobase pairs) of the mouse glycophorin gene contains the locus control region (LCR)-like region, which acts as an erythroid-specific enhancer dependent on chromosomal integration in murine erythroleukemia (MEL) cells. In the present study, we demonstrated that this region binds six nuclear factors. The binding of GATA-1 to corresponding sites did not show any change before or after induction with dimethyl sulfoxide, but the binding of Spi-1/PU.l and an unidentified factor called glycophorin regulatory element binding factor (GRBF) showed a change during induction. While binding activity of Spi-l/PU.l dropped soon after induction, the GRBF activity increased after induction when expression of the glycophorin gene began. After identification of the consensus binding site of GRBF, we cloned cDNA for that factor by Southwestern method, and it was identified as a previously reported transcription factor, delta, a murine form of YY-l which is a versatile transcription factor. Mutation analysis in the delta/YY-1 binding site within the LCR-like region indicated that delta/YY-1 acts as a regulatory protein in combination with the E-box-binding protein that binds to the neighboring sequence.

The far upstream region (؊1.2-0.9 kilobase pairs) of the mouse glycophorin gene contains the locus control region (LCR)-like region, which acts as an erythroidspecific enhancer dependent on chromosomal integration in murine erythroleukemia (MEL) cells. In the present study, we demonstrated that this region binds six nuclear factors. The binding of GATA-1 to corresponding sites did not show any change before or after induction with dimethyl sulfoxide, but the binding of Spi-1/ PU.l and an unidentified factor called glycophorin regulatory element binding factor (GRBF) showed a change during induction. While binding activity of Spil/PU.l dropped soon after induction, the GRBF activity increased after induction when expression of the glycophorin gene began. After identification of the consensus binding site of GRBF, we cloned cDNA for that factor by Southwestern method, and it was identified as a previously reported transcription factor, delta, a murine form of YY-l which is a versatile transcription factor.

Mutation analysis in the delta/YY-1 binding site within the LCR-like region indicated that delta/YY-1 acts as a regulatory protein in combination with the E-box-binding protein that binds to the neighboring sequence.
Mouse glycophorin is a red blood cell membrane protein and a good erythroid-specific marker protein such as hemoglobin. To understand the mechanisms underlying the stage-and tissue-specific expression of erythroid-specific genes, we cloned a mouse glycophorin gene which is homologous to human glycophorin A or B (1) and examined its expression in murine erythroleukemia (MEL) 1 cell differentiation. Upon treatment with dimethyl sulfoxide (Me 2 SO), MEL cells undergo terminal differentiation into red blood cells and induce a number of different erythroid-specific genes (2). Glycophorin gene expression precedes that of globin gene in MEL cell differentiation (3). By nucleotide sequence analysis of the 5Ј upstream region of the mouse glycophorin gene (4) and its transient transcription assay (5), we have shown that the erythroid-specific promoter element consisted of the basal promoter containing GGTGGG and GATA motifs, and the distal elements containing binding motifs for two erythroid-specific nuclear factors GATA-1 and NF-E2. In addition, we found a locus control region (LCR)-like region that was reported in the human ␤-globin gene cluster. The LCR-like region located far upstream of the mouse glycophorin gene seemed to act as an enhancer of the erythroidspecific promoter in the stable transfection assay, but not in the transient transfection assay (5). The role of LCR of the human ␤-globin gene cluster located far upstream of ⑀-globin gene in the regulation of developmental switching and erythroid-specific expression of ␤-like globin genes was demonstrated by analysis using transgenic mice and MEL cells (6 -10). The major regulatory element upstream of the human ␣-globin gene cluster has similar LCR activity (11). The LCR of the human ␤-globin genes includes two tandemly arranged NF-E2/AP-1binding motifs and GATA-1-binding motif. The LCR-like region in the mouse glycophorin gene has a structure similar to globin LCR. In this study we made a detailed analysis of the nuclear factors that bound to the mouse glycophorin LCR-like region in MEL cells before and after Me 2 SO induction.

MATERIALS AND METHODS
Cell Culture and Preparation of Nuclear Extracts-Culture and induction of MEL cells were performed as described previously (12). For preparation of the nuclear extracts (13), the cells were cultivated for 2 days with or without 1.4% Me 2 SO. Cells were washed twice in phosphate-buffered saline and lysed in 10 mM HEPES, pH 7.9, 10 mM EDTA-0.5 mM dithiothreitol, 0.5% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride for 10 min on ice. After centrifugation at 2,000 rpm for 5 min, the nuclear pellet was washed twice with the same buffer and resuspended in the nuclear buffer, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 400 mM NaCl. The elution of proteins was performed on ice for 30 min, and the nuclear extract was recovered by centrifugation at 10,000 rpm for 10 min.
Oligonucleotides-The DNA sequences of synthetic oligonucleotides used for probes and competitors in the gel shift assays and UV crosslinking analysis (both strands were shown from 5Ј to 3Ј) were: GA, The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
DNase I Footprinting and Electrophoretic Mobility Shift Assay (EMSA)-For DNase I footprinting assay, a 284-bp DNA fragment from Ϫ1185 to Ϫ901 bp of the glycophorin gene containing the LCR-like region was 32 P-labeled according to standard procedures. DNase I footprinting reaction mixtures (50 l) contained 36 mM HEPES-KOH, pH 7.9, 57 mM KCl, 9 mM MgCl 2 , 0.1% Nonidet P-40, 10% glycerol, 1 mM dithiothreitol, 10,000 cpm of the DNA fragment, 50 ng of poly(dI-dC)⅐poly(dI-dC) and nuclear extracts (4 g) (14). After incubation on ice for 15 min, various concentrations of DNase I were added. Following incubation on ice, 6 volumes of solution containing 20 mM EDTA, 185 M NaCl, 1% SDS, 0.8 g of proteinase K were added to stop the reaction. After incubation at 42°C for 30 min, DNAs were extracted by phenol-chloroform and ethanol-precipitated. The samples were heated to 95°C for 3 min in a sequencing dye solution and separated in an 8% sequencing gel. Gels were fixed and dried prior to autoradiography. EMSA was performed as described by Sigh et al. (15) with minor modifications. For probe preparation, oligonucleotides synthesized by a DNA synthesizer were annealed and 5Ј-end labeled with [ 32 P]ATP (Amersham Corp.) and T4 polynucleotide kinase (Toyobo Co., Ltd., Japan) according to standard procedures. The labeled DNA fragment (10,000 cpm) was incubated in a final volume containing 10 mM Tris-HCl (pH 7.5), 50 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 5% glycerol, 3 g of poly(dI-dC)⅐poly(dI-dC), and 4 g of nuclear extract proteins. Following 15 min of incubation on ice, electrophoresis was performed on a 5% polyacrylamide gel in 1 ϫ Tris borate-EDTA. Gels were dried prior to autoradiography.
UV Cross-linking Analysis-The binding reaction mixtures were UVirradiated in the presence or absence of competitor oligonucleotides by UV Stratalinker (Stratagene), and the cross-linked protein-DNA complexes were separated on SDS-10% polyacrylamide gels.
Isolation of a GRBF Clone from cDNA Library-The gtll cDNA library was constructed from cDNA made from poly(A) ϩ RNA isolated from the uninduced MEL cells by oligo(dT) priming. We initially screened this cDNA library by the method of Vinsion et al. (10) with three ligated copies of a double-stranded synthetic oligonucleotide e which was the GRBF recognition sequence site determined by the gel shift analysis. Purified phage plaques that gave positive signals with the oligonucleotide e and gave negative signals with the oligonucleotide Mb were selected as positive clones. An EcoRI fragment containing the cDNA sequence was excised from the selected gtll phage and inserted into the EcoRI site of transcriptional vector BSK(ϩ).
Construction of Plasmids Containing Mutations in the LCR-like Region of the Mouse Glycophorin Gene-The 5Ј-flanking sequence (1.3 kb) containing the LCR-like region and the transcription start site of the mouse glycophorin gene or its site-directed mutants (⌬delta/YY-1; -CCATTT-to -CCAGGT-) (⌬E-box; -CACGTG-to -CGATCG-) and the 0.9-kb sequence deleting the LCR-like region (⌬LCR) were subcloned and linked to luciferase gene.
Stable Transfection Assay-Cells were transfected by a modification of the recommended protocol of the synthetic, cationic lipopolyamine molecule (Transfectam reagent, Promega). MEL cells were split 3 h before transfection and replated onto poly-L-lysine (Sigma)-coated 60-mm diameter dishes at 5 ϫ 10 5 cells/dish (4 -8 dishes for each test plasmid) in ES-fetal bovine serum. For transfection, cells were washed three times with ES and then were incubated with 5 g of supercoiled plasmid DNA and Transfectam reagent, 25 g in opti-MEM I (Life Technologies, Inc.) (1.5 ml) for 12 h at 37°C. After the opti-MEM I was removed, the cells were incubated for 24 h at 37°C in ES-fetal bovine serum (6 ml/plate) to express the selection marker, aminoglycoside phosphotransferase. For the selection, G418 (Geneticin, Sigma) was added to the culture medium at 400 g/ml. After 9 days of incubation, scores of G418-resistant colonies were observed. To avoid clonal variation and positional effect, the cells were resuspended and subcultured in G418-containing ES-fetal bovine serum for 7 more days. Four to eight independent transfected populations were established for each test plasmid (wild-type, ⌬delta/YY-1, ⌬E-box, and ⌬LCR). Each population was split, and one half was cultured in ES-fetal bovine serum while the other half was induced to differentiate by addition of 1.4% Me 2 SO for 4 days. Luciferase assay (16) was performed at 4 days according to the protocol of the luciferase assay kit (Toyo Ink Co. Tokyo). Six days after induction of differentiation, benzidine staining was done to estimate the extent of differentiation.
Northern Blot Hybridization-To measure levels of endogenous gly-cophorin mRNA, total RNAs from MEL cells were prepared by guanidinium-thiocyanate-phenol-chloroform procedure. Ten g of RNA from each sample were electrophoresed in a denatured 1% gel agarose gel, transferred to a nylon membrane, and hybridized with a mouse glycophorin cDNA probe (1) radiolabeled with a random primer extension kit (DuPont) in the hybridization buffer (50% formamide, 5 ϫ SSC, 1 ϫ Denhardt's reagent, 20 mM NaHPO 3 , pH 6.5, 100 mg/ml salmon sperm DNA, 10% dextran sulfate, 0.1% SDS). After overnight incubation at 42°C, the membrane was washed in 2 ϫ SSC, 0.1% SDS for 10 min at room temperature and in 0.1 ϫ SSC, 0.1% SDS for 30 min at 60°C, then autoradiographied. sensitivity are shown in Fig. 2. Analysis of the sequences within the footprint regions revealed four putative erythroidspecific binding sites; two GATA motifs and two AP-1/NF-E2like motifs in the region assigned as II in Fig. 2, A and C. Sequence of the general form (T/A)GATA(A/G) binds the erythroid-specific transcription factor GATA-1 (4,(17)(18)(19). Of two potential GATA-1 binding sites referred to a nonconsensus site (AGATAC) and a consensus site (TGATAG), the footprinting revealed a protection in the non consensus site, but did with no difference between the induced and uninduced MEL cells. Two AP-1/NF-E2-like consensus motifs were found upstream of the two GATA elements, as reported for the erythroid-specific promoter of the human prophobillinogen deaminase gene (20) and for the LCR of the human ␤-globin gene (3). These motifs are not completely fitted to the NF-E2 consensus motif (T/C)GCT-GA(G/C)TCA(C/T) which Andrews et al. (21) recently reported. Between the two motifs, footprinting revealed the hypersensitive site only in the 5Ј side of the upstream AP-1/NF-E2-like motifs (Fig. 2C).

Analysis of the Mouse Glycophorin LCR-like Region by
The Spi-1/PU.1 consensus motif (PU box) (22), GAGGAA resides in the central portion of the LCR-like region. The PU box in the noncoding strand was protected in the uninduced extract (DMSO (Ϫ) in Fig. 2B), but was not in the induced extract (DMSO (ϩ) in Fig. 2B). In addition, footprinting revealed that a hypersensitive site (an arrowhead in Fig. 2A, III) in the coding strand at the 3Ј-side of the PU box was more apparent in the uninduced cells than in the induced cells. The feature of the nuclear factor binding of the PU box was similar to that observed in the PU box within the intervening sequence of the mouse ␤-globin gene (23). The Spi-1/PU.1 activity reduced after induction of MEL cell differentiation.
The E-box (CACGTG) motif which is the binding site for basic helix-loop-helix protein (24,25) was found in the downstream of the Spi-1/PU.1 element. While footprinting revealed the hypersensitive site and the protection in the noncoding strand of the region containing E-box element (see Fig. 2B), more apparent protection was seen in the coding strand in the induced extract than the uninduced one ( Fig. 2A). This apparent difference indicated induction of nuclear factors that bind the E-box and/or the neighboring element after MEL cell differentiation. From these results we tentatively concluded that the LCR-like region contains at least six binding sites, in two of which the protein binding is changed after MEL cell differentiation (Fig. 1).
Characterization of the Nuclear Factors by EMSA-To identify which individual proteins interact with each of the binding sites, many oligonucleotides were used as probes and competitors in EMSA; in addition to the GATA-1 site, two binding sites in particular in which protein binding is changed between the induced and uninduced MEL cells were extensively studied.
LG-oligo (Ϫ1104 to Ϫ1090) of the LCR-like region containing a GATA motif showed the retarded bands (one major slow and one minor fast migrating bands) and they were competed with excess amounts of either GA-oligo containing a consensus GATA-1 site or LG-oligo, but not with unrelated LP-oligo, indicating that the GATA-motif within the LCR-like region can bind GATA-1.
LP-oligo (Ϫ1071 to Ϫ1044) containing a PU.1 site within the LCR-like region showed three retarded bands which could be competed to the same extent with excess amounts of either PU-oligo containing the Spi-1/PU.1 consensus binding site or LP-oligo, but not with unrelated LN-oligo (Fig. 3B), indicating that the PU.1 site within the LCR-like region can bind PU.1. PU.1 binding activities were titrated with PU-oligo (Fig. 3C). Two major retarded bands (shown by arrowheads) were detected in the MEL cell extract as well as in the NS-1 cell extract, in which PU.1 activity was reported (22). MEL cell extracts contained more abundant PU.1 activity than NS-1 cell extracts. Binding with the induced MEL cell extract showed a five times weaker band than with the uninduced one at the same position. This result is consistent with that observed in DNase I footprinting assay.
Oligonucleotide V (Ϫ982 to Ϫ947) which contains the E-box gave rise to three retarded bands in the MEL extract (Fig. 3D). The top band competed with Hm which includes the E-box consensus site (CACGTG) (Fig. 3E). The second band (arrowhead in the figure) was competed with Vm which possesses a mutation in the E-box consensus site and thus cannot bind E-box binding proteins. The third band did not competed with E, thus this complex may be nonspecific binding. Interestingly, the proportion of the second band increased after induction (approximately 3-fold) and was highest 2 days after induction when glycophorin gene expression began (26) (Fig. 3D). Thus, the second band may contain a key factor for glycophorin gene expression and we called this factor GRBF.
Determination of the Consensus Binding Sequence of GRBF-In order to determine the consensus binding sequence of GRBF, we performed competitor scanning experiments by adding five different oligonucleotides (Fig. 4A). Under these conditions, efficient competition of the second band was only achieved with oligonucleotide e, while the top band competed with c and d, both of which contain the E-box consensus sequence (Fig. 4A). The competition experiments using mutated e oligonucleotides (Ma, Mb, and Mc) indicated that consensus binding sequence of this factor requires the TTT sequence since Mb and Mc did not bind (Fig. 4B). The EMSA with mutated e oligonucleotides (Md, Me, and Mf) indicated that CCA is required for the binding (Fig. 4C). In the gel shift assay, oligonu-cleotide e showed a single band that was inducible in the Me 2 SO-treated cells, and Me oligonucleotide may be an upmutant because the band in the same migration was stronger than that of e oligonucleotide, while Md was a weak mutant (Fig. 4C). Mutated Mf oligonucleotide showed a strong band whose migration was different from that of oligonulceotide e and did not change during induction of MEL cells, thus Mf might bind a different factor. Combining these results, we concluded that this inducible DNA-binding factor called GRBF could bind to CCATTT as the most probable consensus sequence. EMSA with e oligonucleotide in the MEL cell extracts showed that GRBF was induced by the addition of Me 2 SO and was maximum on day 2 (Fig. 4D). The radiolabeled oligonucleotide (PU) that is the consensus PU.1 binding site was incubated with the induced (ϩ) and uninduced (Ϫ) MEL cell nuclear extracts. For a positive control for the PU.1 signal, nuclear extracts of NS-1 cells were used. D, for identification of the E-box binding activity, probe V was incubated with the induced (ϩ) and uninduced (Ϫ) MEL cell nuclear extracts. E, competition of the nuclear factors that bind to oligonucleotide V in the MEL cell nuclear extracts with mutated oligonucleotides competitors. Hm is an E-box binding element and Vm is mutated only in the E-box element of V. Pds is the negative control. F, sequences of competitor oligonucleotides. The sequences show only one strand. Numbers correspond to sequence numbers in Fig. 1 and each consensus sequence is marked as boldface letters. In the Vm-oligo, mutations within the E-box are underlined. Estimation of the Molecular Weight of GRBF-To determine the molecular weight of GRBF, we performed UV cross-linking analysis (Fig. 5). A mixture of 32 P-labeled probe e and MEL cell nuclear extracts was irradiated with short wavelength UV light for 5 min. Then the mixture was resolved by SDS-polyacrylamide gel electrophoresis and analyzed by autoradiography. To demonstrate the sequence-specific nature of the protein, we added a 500-fold excess of cold competitors to the binding reaction mixture. We observed four bands without competitor, although two of them with higher molecular weight were not diminished by competitor F. Moreover, the band in the lowest molecular weight (arrowhead) was diminished by competitor e but not by b which cannot bind GRBF. This indicates that the lowest band is the complex formed with GRBF. As the size of this band was 70 kDa, the molecular mass of this protein is expected to be about 60 kDa.
Isolation of cDNA for GRBF by Southwestern Cloning-Identification of the consensus binding sequence of GRBF made it possible to isolate its cDNA clone by Southwestern cloning method using the expression library constructed in gt11 vector from the mRNA of MEL cells. By screening with the labeled oligonucleotides containing the CCATTT consensus sequence for GRBF we obtained clones, and the specificity of binding of these clones was confirmed with mutant oligonucleotides including the up mutant, Me and the weak mutant Md (data not shown).
The nucleotide sequence determination of the longest cDNA clone indicated that the longest open reading frame completely matched with the known DNA binding factor, delta (27), a murine form of YY-1 (28), also called NF-E1 (29) or UCRBP (30). The molecular size of delta is about 65 kDa by SDSpolyacrylamide gel electrophoresis and resembles that of GRBF, which our UV cross-linking analysis identified as about 60 kDa (Fig. 5).
Functional Analysis of Delta/YY-1 Binding on LCR-like Activity-To confirm the binding of delta/YY-1 to the CCATTT within the LCR-like region of the mouse glycophorin gene, the antibody was made against bacterially produced delta/YY-1 protein in rabbits and added to the gel shift assay (Fig. 6). The gel shift band corresponding to delta/YY-1 binding of the MEL cell extract disappeared by addition of the anti-delta/YY-1 antibody, while anti-c-Myc antibody showed no effect. The result indicated that delta/YY-1 is really bound to the LCR-like region.
Then we examined the functional involvement of delta/ YY-1 on the LCR-like activity of the mouse glycophorin gene by stably transfecting the mutants in the delta/YY-1 binding site and the neighboring E-box binding site into MEL cells (Fig. 7). The 5Ј-flanking sequence (1.3 kb) containing the LCR-like region and the transcription start site of the mouse glycophorin gene and the 0.9-kb sequence deleting the LCRlike region (⌬LCR) were subcloned and linked to the luciferase gene. For introducing a mutation in the delta/YY-1 binding site, CCATTT was changed to CCAGGT which could not bind to delta/YY-1 as shown in Fig. 4 (Mb mutant) and for introducing a mutation in the E-box binding site, CACGTG was completely changed to CGATCG. The plasmid DNAs were introduced into MEL cells by a lipopolyamine method to obtain the stable transformants. To minimize the variation of each transformant, four to eight independent neomycin-resistant transformants for each plasmid were pooled, and the transcriptional activities in the induced and uninduced cells were measured by luciferase assays at 4 days of culture. To confirm that the cells selected after transfection did not change their properties, levels of endogenous glycophorin mRNAs before and after induction in each pool of transformants were measured by Northern blotting (Fig. 7D). The result showed that induction of the endogenous glycophorin gene was not altered in these transformants. Induction of differentiation monitored by benzidine staining in each transformant also showed that the transfection did not affect extent of differentiation in these transformants (data not shown). When the luciferase activities were measured at 4 days of culture, the deletion in the LCR-like region (⌬LCR) reduced its transcriptional activity to 50% of that of the 1.3-kb sequence containing the LCR-like region (Ϫ) (Fig. 7B). The mutation in the delta/YY-1 binding site (⌬delta/YY-1 in Fig. 7A) itself reduced to 60% that of the original activity. On the other hand, mutation in the neighboring E-box biding site (⌬E-box) resulted in complete loss of LCR-like activity. Activity of the mutations in both sites resumed to a level similar to that of the deletion in the delta/YY-1 binding site alone. To confirm the change in the transcriptional activities in these transformants, the time course of luciferase activities after  6. Effect of anti-delta/YY-1 antibody on the nuclear factor binding. A, MEL cell nuclear extracts were incubated with 32 P-labeled oligonucleotide F (TGTCTCCATTTCCACA) containing the delta/YY-1 binding motif (Ϫ) and competed by increasing amounts (ϩ, 50-fold excess; ϩϩ, 500-fold excess) of either F oligonucleotide or the related oligonucleotide GT (GACTCCACCCT). Arrowhead shows the specific band. B, rabbit antisera were obtained by injecting the MBP-delta/YY-1 fusion protein produced in Escherichia coli and IgG was purified by affinity matrices. Rabbit anti-c-Myc antibody was obtained from Santa-Cruz Biotechnology Inc. Increasing amounts of anti-delta/YY-1 (ϩ, 1 g; ϩϩ, 10 g) or anti-c-Myc (ϩ, 1 g; ϩϩ, 10 g) were added to the reaction mixtures of the MEL cell nuclear extracts and the radiolabeled oligonucleotide F. induction was measured (Fig. 7C). The result showed that luciferase activities were dependent on induction.

DISCUSSION
The LCR-like region found in the far upstream region (Ϫ1.2 to Ϫ0.9 kb) of the mouse glycophorin gene has been shown to contain several erythroid-specific nuclear factor binding sites as observed in the human ␤-globin gene. Deletion analysis of the 5Ј far upstream region of this gene conferred high levels of expression in a stable transfection assay, but, interestingly, when tested in transient expression assays in MEL cells, it showed no stimulation of transcriptional activity (5) as shown in the LCR of ␤-globin gene (31). Thus, the far upstream region of the mouse glycophorin gene was clearly distinguished from general enhancers, which were originally defined in transient transcription assays (16). It is likely that this region may activate the glycophorin gene on the chromatin structure like the LCR of the ␤-globin gene (7,31,32).
To understanding the mechanism of the differentiation-specific gene regulation by the glycophorin LCR-like region, it is important to investigate the nuclear proteins that bind to it before and after induction of MEL cell differentiation. Our DNase I footprinting and EMSA consistently showed that several nuclear factors bound to the LCR-like region (Figs. 1-4). We showed the presence of binding sites for the erythroidspecific transcription factor GATA-1, and they did not show any change in their binding activities before or after induction of MEL cells. Although the hypersensitive site was observed in the 5Ј side of the upstream AP1/NF-E2-like motifs in the two AP1/NF-E2-like motifs within this region (Fig. 2C), suggesting the nuclear factor binding, we could not observed clear gel shift bands with these DNA sequences. 2 The co-localization of these binding sites in the glycophorin LCR-like region was very similar to the globin gene LCR (3,6,16,21,33,34). Of interest is the apparent difference in the binding activity of GATA-1 to the GATA site within the LCR-like region and the promoter region of the glycophorin gene during MEL cell differentiation. It is likely that GATA-1 binding is affected by the neighboring sequence (5).
In contrast to these elements, two binding sites for Spi-1/ PU.1 and GRBF showed a change. Spi-1 is an oncogene activated during leukemogenesis induced by Friend virus (35) and is identified as PU.1, which is a transcription factor and one of the protein products of the ets protooncogene family (22,36,37). The binding activity of PU.1 dropped after induction as shown in Figs. 2 and 3, and these results are consistent with our previous report that levels of PU.1 mRNA dropped rapidly when these cells were treated with Me 2 SO (26). We speculate that Spi-1 itself acts as a negative factor for transcription and/or that its binding to the element may interfere with the binding of other nuclear factors to it. The globin gene seems not to contain the PU.1 binding site within the LCR-like region; however, its binding site was found in the 3Ј-enhancer of the 2 Y. Nemoto, unpublished observation.  mouse ␤-globin gene (23). Thus, it is possible that PU.1 acts as a negative factor for erythroid-specific gene expression, as it has a regulatory function in other hematopoietic cells such as T cells (22).
Another factor with binding activity that changes after induction is GRBF. We concluded by gel shift competition analysis that the consensus sequence for GRBF is CCATTT. The GRBF activity increased after induction of MEL cells when the transcription of glycophorin began to rise (1,26). By screening with the CCATTT consensus sequence, we isolated the cDNA clone for GRBF and surprisingly, found that GRBF was identical to delta, a murine form of YY-1 (28) which is a versatile transcription factor that has a zinc finger DNA binding domain. The band shift with oligonucleotide e in the MEL cell extract was retarded with the addition of rabbit anti-delta/ YY-1 antibody. Delta/YY-1 is involved in negative and positive regulation of transcription of many genes (38). For example, two YY-1 binding sites at Ϫ60 and at ϩ1 of the adeno-associated virus P5 promoter mediate repression of transcription by YY-1, an effect which is relieved by E1A (28), and delta activates transcription of mouse ribosomal protein genes that lack a TATA box from binding sites upstream of the transcription start site (27). Its binding sites were found in many genes (27-30, 39 -41); those found in adeno-associated virus (28), in the upstream conserved region of MuLV (30), and in one of c-myc upstream elements (40) are CCATTT, which is completely matched with that of GRBF, while those found in immunoglobulin 3Ј gene enhancer (29) and in ribosomal protein rpL30 gene (27) are CCATCT. By comparison with the functional core of the globin LCR-HS3, we found that the well conserved element FP-6 (33) has CCTTT, which resembles to a consensus core sequence for delta/YY-1. Our preliminary gel shift assay with the FP-6 sequence probe demonstrated that delta/YY-1 can bind to this element (data not shown). This result suggests that delta/YY-1 may play an important role in regulating LCR-like activity in erythroid differentiation. This possibility was examined by stable transfection of the point mutants in the GRBF binding site and the neighboring E-box binding site within the LCR-like region of the mouse glycophorin gene. The stable assays showed that a point mutation in the delta/YY-1 binding site reduced its LCR activity to 60% of original activity. More interestingly, a mutation in the E-box binding site completely lost the LCR-like activity. Based on these results, the following models may be plausible. 1) YY-1 is a transcriptional activator which inhibits transcription in the absence of E-box binding protein.
2) E-box binding protein is a transcriptional activator which is nonfunctional without YY-1, and YY-1 by itself acts as an inhibitor. In either model, it is clear that delta/YY-1 regulates the LCR-like activity in combination with the E-box binding protein.
It is interesting to learn how the activity of delta/YY-1 can be modulated during induction of MEL cells. The nucleotide sequence adjacent to the delta/YY-1 binding site consisted of E-box binding site whose binding activity did not change before or after induction (Fig. 3B). It is important to identify the E-box-binding protein in MEL cells. Since the E-box found in the LCR-like region is CACGTG which is the consensus binding site for Max/Max and Myc/Max (42,43), it would be of interest to know whether c-Myc could bind to this element, because overexpressed c-Myc interferes with induction of differentia-tion of MEL cells (12), and the association of c-Myc with YY-1 inhibits YY-1 activity (38).