Transcription of Human ABO Histo-blood Group Genes Is Dependent upon Binding of Transcription Factor CBF/NF-Y to Minisatellite Sequence*

We have studied the transcriptional regulatory mechanism of the human histo-blood group ABO genes, and identified DNAcis-elements and trans-activating protein that control the expression of these genes which are important in blood transfusion and organ transplantation. We introduced the 5′-upstream sequence of ABO genes into the promoterless reporter vector and characterized the promoter activity of deletion constructs using transient transfection assays with gastric cancer cell line KATO III cells. The sequence just upstream of the transcription start site (cap site), and an enhancer element, which is located further upstream (between −3899 and −3618 base pairs (bp) from the transcription initiation site) and contains 4 tandem copies of a 43-bp repeat unit, were shown in gastric cancer cells to be responsible for the transcriptional activity of the ABO genes. DNA binding studies have demonstrated that a transcription factor, CBF/NF-Y, bound to the 43-bp repeat unit in the minisatellite. Functional importance of these CBF/NF-Y-binding sites in enhancer activity was confirmed by transfection experiments using reporter plasmids with mutated binding sites. Thus, transcriptional regulation of the human ABO genes is dependent upon binding of CBF/NF-Y to the minisatellite.

We have studied the transcriptional regulatory mechanism of the human histo-blood group ABO genes, and identified DNA cis-elements and trans-activating protein that control the expression of these genes which are important in blood transfusion and organ transplantation. We introduced the 5-upstream sequence of ABO genes into the promoterless reporter vector and characterized the promoter activity of deletion constructs using transient transfection assays with gastric cancer cell line KATO III cells. The sequence just upstream of the transcription start site (cap site), and an enhancer element, which is located further upstream (between ؊3899 and ؊3618 base pairs (bp) from the transcription initiation site) and contains 4 tandem copies of a 43-bp repeat unit, were shown in gastric cancer cells to be responsible for the transcriptional activity of the ABO genes. DNA binding studies have demonstrated that a transcription factor, CBF/NF-Y, bound to the 43-bp repeat unit in the minisatellite. Functional importance of these CBF/NF-Y-binding sites in enhancer activity was confirmed by transfection experiments using reporter plasmids with mutated binding sites. Thus, transcriptional regulation of the human ABO genes is dependent upon binding of CBF/NF-Y to the minisatellite.
Histo-blood group ABH(O) antigens, the major alloantigens in humans (1), have been characterized as defined trisaccharide determinants GalNAc␣133(Fuc␣132)Gal␤13 R, Gal␣133(Fuc␣132)Gal␤13 R, and disaccharide determinant Fuc␣132Gal␤13 R for A, B, and H, respectively (2,3). These structures represent the secondary gene products which are synthesized from the precursor H substrate by ␣133GalNAc (A transferase) and ␣133Gal transferase (B transferase), the primary gene products coded by the functional alleles at the ABO locus (4,5). Molecular genetic studies of the ABO genotypes have identified two critical single-base substitutions between A and B genes, the resultant 2-amino acid substitutions being responsible for the different donor nucleotide-sugar sub-strate specificity between A and B transferases. A single base deletion, which shifts the codon reading frame and abolishes the function of A transferase, has been identified in O allelic cDNAs (6,7).
ABH antigens are known to undergo drastic changes during development, differentiation, and maturation. Studies of these antigens in stratified squamous epithelia provided one of the clearest examples of differential expression during cell maturation (8). In non-keratinized stratified squamous epithelia, the immature cells in the basal layers are characterized by the expression of sialylated or unsubstituted precursor peripheral cores, while differentiated and mature cells in the upper layers sequentially express ␣132-fucosylated H structures, and A and B antigens depending on the ABO genotype of the individual. This sequential expression of carbohydrate antigens is associated with the differentiation pattern of the epithelium. An interesting question is how these changes are controlled during cell differentiation. Since keratinocytes are known to greatly change their gene expression during terminal cell differentiation (9), the switch-on of the ABO genes during the maturation may be governed by the same factor(s). To fill in the gap between the expression of the ABO genes and the appearance of the ABO phenotypes in the terminal differentiation of epithelial cells, it is essential to understand the transcriptional regulatory mechanism of the ABO genes. In addition to the normal cell differentiation process, the changes of ABH antigen expression have also been documented in abnormal processes such as tumorigenesis. Reduction or complete deletion of A/B antigen expression in primary lung, bladder, and colorectal carcinomas have been reported. This phenotypic change was well correlated with the invasive and metastatic potentials of the tumors, and with 5-or 10-year mortality rates of the patients (10,11). Expression of H/Le y , the A/B precursor, was inversely correlated with the survival rate of the patients with lung carcinoma after surgery (12). Disappearance of A/B antigens was ascribed to the absence of A or B transferase gene expression rather than the loss of these genes in human bladder tumors (13). Delineation of transcriptional regulation of the ABO genes, therefore, may provide clues as to the underlying mechanisms resulting in A/B antigen disappearance in cancer cells with invasive and metastatic potentials.
In an initial attempt to elucidate the molecular mechanisms controlling the expression of the human ABO genes, we have previously isolated five overlapping genomic clones to cover the entire coding sequence as well as some 5Ј-upstream and 3Јdownstream sequences (14). In this paper, we report the identification of a DNA cis-element and trans-activating protein which accounts for the expression of the ABO genes in cells from gastrointestinal origin.

MATERIALS AND METHODS
Cell Culture, Transfection, and Luciferase Assay-A human gastric cancer cell line, KATO III (JCRB0611), was grown in 45% RPMI 1640 with 45% minimum essential medium containing 10% fetal bovine serum and 0.5% penicillin-streptomycin. Cells were transfected by electroporation. Briefly, the cells were inoculated at 3 ϫ 10 5 /ml 16 -24 h prior to transfection. They were harvested by centrifugation at 500 ϫ g for 10 min at room temperature, washed twice with RPMI 1640 without fetal bovine serum and L-glutamine, and resuspended at 5 ϫ 10 6 /ml in RPMI 1640. An aliquot (0.8 ml) of cell suspension was placed into a 0.4-cm electroporation cuvette (Bio-Rad). Fourteen micrograms of supercoiled plasmid DNA was added to the cell suspension. Cells were incubated with DNA for 5 min at room temperature and then electroporated for 960 microfarads at 250 V. Immediately after transfection, the cuvettes were transferred to an ice bath and kept for 15 min. The cells were then transferred to 10 ml of RPMI 1640 plus minimal essential medium containing 10% fetal bovine serum and 2 mM L-glutamine, incubated for 48 h, and harvested for luciferase and ␤-galactosidase assays.
Cell lysis and luciferase assays were performed following the manufacturer's protocols using the Luciferase Assay System (Promega, Madison, WI). Light emission was measured in a model 1253 luminometer (BioOrbit), and the values were obtained in relative light units. The amount of cell lysate used for each luciferase assay was adjusted so that the observable light emission would fall within the linear range of titration curves (light emission observed versus amount of luciferase protein). Variations in transfection efficiency were corrected by normalizing the activities of ␤-galactosidase expressed from the co-transfected RSV LTR/␤-galactosidase plasmid DNA. ␤-Galactosidase activities were measured as described elsewhere (15). Relative activity of each construct in different experiments was obtained by arbitrarily assigning the activity of the pGL3-promoter vector containing the SV40 promoter to be 1.0.
Plasmids-A DNA fragment containing the 5Ј-upstream sequence of the human ABO gene was subcloned from the genomic clone HG-1 (14) into luciferase reporter vectors, pGL3-basic vector (Promega), pGL3promoter vector (Promega), pGL2-control vector (Promega), and pTKluc (16). The SmaI site of the pGL3-basic vector was converted to the EcoRI site to facilitate the subcloning of the EcoRI/NcoI genomic fragment into the polylinker site just upstream of the luciferase gene. This construct was then used for the generation of several progressive series of upstream end deletion constructs. Since an NcoI site is present at the translation initiation codon in both ABO and luciferase genes, the linkage between the upstream region and translated region of the ABO gene was conserved in the reporter plasmids containing the progressive series of upstream end deletions. The pGL3-promoter vector contains the SV40 viral promoter, while pTK-luc contains a thymidine kinase promoter. The pGL2-control vector contains both the SV40 promoter and enhancer. The SmaI site of the pGL3-promoter vector was converted to the EcoRI site to facilitate the subcloning of the genomic fragments or PCR 1 -amplified fragments into the polylinker site just upstream of the SV40 promoter. Nomenclature of the various ABO gene constructs is based on the nature of the inserted fragment. Letter symbols reflect the restriction enzyme cleavage sites used for the construction of these plasmids, while numerals indicate the end points of the primers used for the polymerase chain reaction (PCR). For example, EN construct contains the EcoRI-NcoI fragment (between Ϫ4661 and ϩ31), and Ϫ3899H construct contains the fragment bordered with PCR primer sequence starting at Ϫ3899 on one end and HindIII site on the other. All the DNA fragments were generated by either restriction endonuclease digestion or PCR. Construct Ϫ3899:Ϫ3618⌬/SN was produced by overlapping PCR mutagenesis. The genomic DNA fragment between the XmnI site at Ϫ3252 and the PstI site at Ϫ2371 was inserted downstream of the SV40 enhancer in the pGL2-control vector, in the same orientation as in the ABO genes. pGL3-control vector (Promega), containing the SV40 promoter and enhancer, was used in promoter assays.
The Ϫ3899 to Ϫ3618 fragment containing 4 tandem copies of a 43-bp repeat unit was introduced into the upstream of promoter sequence in some constructs ( (4)). Arrows to the right indicate that the fragment was inserted in the same orientation as it appears in the ABO gene; the fragment was cloned in the same orientation as the luciferase genes in the reporter plasmids are transcribed. The fragment (Ϫ3899:Ϫ3618) was PCR-amplified and cloned first into pCRII (Invitrogen) to generate construct pCRϪ3899:Ϫ3618. The plasmids Ϫ3899:Ϫ3618 (3)/SV and Ϫ3899:Ϫ3618 (3)/TK were prepared by digesting pCRϪ3899:Ϫ3618 with EcoRI, modifying both ends with SalI linker, and directionally ligating the fragment into the XhoI-digested pGL3-promoter vector and SalI-digested TK-luc, respectively. A fragment containing the Ϫ3899 to Ϫ3618 sequence was obtained by cleaving the plasmid Ϫ3899:Ϫ3618 (3)/SV with SmaI and BglII. One end of the SmaI/BglII fragment was modified to SalI site whereas the other end was blunted in either orientation. These fragments were then directionally ligated into the SalI/BamHI sites of pGL3-promoter vector and SN to construct four plasmids, SV/Ϫ3899:Ϫ3618 (3), SV/Ϫ3899: Ϫ3618 (4), SN/Ϫ3899:Ϫ3618 (3), and SN/Ϫ3899:Ϫ3618 (4). The plasmid Ϫ3899:Ϫ3618 (3)/SN was prepared by digesting Ϫ3899:Ϫ3618 (3)/SV with BglII, modifying the end to EcoRI site, cleaving with NheI, and directionally ligating the fragment with NheI-,EcoRI-digested SN. The double-stranded oligonucleotides, TR23 and mTR23, were inserted into the BamHI site of pT7T3U18 sequencing vector (Stratagene). To prepare the plasmids 4 ϫ 23/SV and 4 ϫ m23/SV, the EcoRI-XbaI fragments containing the above inserted oligonucleotide sequences were cleaved, and ligated to the EcoRI/BglII site upstream of the SV40 promoter in the modified pGL3-promoter vector.
Plasmid DNA was purified by applying alkaline lysed samples onto two successive CsCl-ethidium bromide gradients. Orientation, and the 5Ј and 3Ј boundaries of the insert of all the constructs used in this study were verified by detailed restriction enzyme mapping and DNA sequence analysis using Cycle Sequencing Kit with AmpliTaq DNA Polymerase, FS (Perkin-Elmer). For all the constructs containing PCRamplified fragments, sequencing was performed over the entire region of the amplified sequences.
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts and probes were prepared as previously reported (17,18). 1 ϫ 10 8 to 3 ϫ 10 8 cells were harvested, washed with phosphate-buffered saline, and incubated for 5 min in 5 ml of ice-cold buffer A (10 mM HEPES pH 7.9, 5 mM MgCl 2 , 10 mM NaCl, 0.3 M sucrose, 0.1 mM EGTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) containing 1 g per ml each of protease inhibitors, antipain, aprotinin, chymostatin, leupeptin, and pepstatin A. After centrifugation, the cells were resuspended in 1 ml of buffer A with protease inhibitors and then Dounce homogenized. The homogenate was microcentrifuged for 30 s. Nuclei were resuspended in 0.8 ml of buffer B (20 mM HEPES pH 7.9, 5 mM MgCl 2 , 300 mM KCl, 0.2 mM EGTA, 25% glycerol, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) with protease inhibitors, and gently rocked on a platform at 4°C for 30 min. After 30 min of microcentrifugation at 4°C, the supernatant was dialyzed against 50 ϫ volumes of buffer D (20 mM HEPES pH 7.9, 100 mM KCl, 0.2 mM EDTA, 20% glycerol, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) at 4°C overnight. After 30 min of microcentrifugation at 4°C, aliquots of the supernatant were frozen at Ϫ70°C, and the protein concentration was determined using a Bio-Rad protein assay kit.
We have employed two separate protocols for EMSAs. Results shown in Fig. 4, A and B, were obtained by the protocol previously described by Shirakawa et al. (18), using 4% polyacrylamide gel in 0.25 ϫ TAE buffer (6.67 mM Tris, pH 7.5, 3.3 mM sodium acetate, 1 mM EDTA). Binding reactions were performed in 15 l of a buffer (10 mM Tris-HCl, pH 7.5, 1 mM EDTA, 40 mM NaCl, 1 mM ␤-mercaptoethanol, 4% glycerol) containing 32 P-labeled probe (0.2 ng, 5,000 -20,000 cpm), 0.4 g of poly(dI-dC), and 5 g of nuclear extracts at room temperature for 15 min. The gels were electrophoresed at room temperature. The EMSAs performed in Fig. 5, A and B, followed the protocol described by Coustry et al. (19), using 4% polyacrylamide gel in 0.5 ϫ TBE buffer (45 mM Tris, 45 mM boric acid, 1 mM EDTA). Binding reactions were performed in 15 l of a buffer (25 mM HEPES pH 7.9, 75 mM KCl, 1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol, 10% glycerol) containing 32 P-labeled probe (0.2 ng, 5,000 -20,000 cpm), 0.4 g of poly(dI-dC), and 5 g of nuclear extracts at room temperature for 15 min. The gels were electrophoresed at room temperature. One hundred-fold molar excess of unlabeled competitors over the labeled probe were used for competition analyses except for those shown in Fig. 5A (see legend). For supershift experiments, 2 l of polyclonal rabbit anti-NF-Y A subunit antibody (Rockland, Gilbertsville, PA), 2 l of polyclonal rabbit anti-Egr-1 antibody (Santa Cruz Biotechnology) raised against a C-terminal Egr-1 peptide, or 2 l of normal rabbit serum was added to the nuclear extract, and preincubated for 15 min on ice prior to addition of radiolabeled probes.
The E␣-23 oligonucleotide sequence corresponds to nucleotides Ϫ66 to Ϫ45 in the major histocompatibility complex class II promoter (E␣promoter) (20). The NF-I oligonucleotide contains NF-I consensus sequence as described by Chodosh et al. (21). The underlines in the mTR23 oligonucleotide indicate the sites of specific mutations designed to abrogate the transcriptional factor CBF/NF-Y binding. We substituted the nucleotide sequences following Dorn et al. (20).

Identification of an Upstream Region Required for ABO
Blood Group Gene Transcription-We have previously isolated several genomic clones of the human ABO genes (14). A genomic clone, designated as HG-1, was shown by restriction enzyme analysis and Southern hybridization to contain the 5Ј-upstream sequence of the coding region. A 7.8-kb EcoRI/ EcoRI fragment, which encompassed 4.7 kb of the 5Ј-flanking region and the first exon as well as the first intron, was subcloned from this clone and sequenced. The complete nucleotide sequence of the 5Ј-flanking region has been deposited to Gene-Bank TM (accession number U22302). Two possible transcription initiation sites were mapped by the 5Ј-rapid amplification of cDNA ends technique using human pancreatic cDNA as a template. Sequence analysis of the proximal region to these sites demonstrates the presence of several GC boxes just upstream of these possible transcription initiation sites, whereas neither TATA nor CAAT boxes are found close to these sites (14).
We have recently identified a minisatellite in the 5Ј farther upstream region. This minisatellite sequence is composed of 4 tandem copies of a 43-bp repeat unit, and is located at positions Ϫ3843 to Ϫ3672 relative to the upstream transcription start site. Minisatellite sequences have been shown to serve as regulatory regions for cellular transcription in certain genes (22)(23)(24)(25)(26)(27). Nucleotide sequence between Ϫ3950 and Ϫ3470 of the human ABO gene is shown in Fig. 1. In this region around the minisatellite, several potential transcription factor-binding site motifs were identified. Therefore, it seemed important and interesting to assess the functional roles of this minisatellite sequence in the transcription of ABO genes.
As a means to examine the promoter/enhancer activity of the 5Ј-upstream sequences of the human ABO gene, we have employed reporter and transfection systems. We first obtained the EN construct by introducing the 4.7-kb EcoRI-NcoI genomic fragment 5Ј-flanking the coding sequence of the human ABO gene into the promoterless pGL3-basic vector at the upstream of the luciferase coding sequence (see Fig. 2). This plasmid was transiently transfected into the human gastric cancer cell line KATO III. This cell line was derived from the tumor of a blood group B individual. 2 We have observed large amounts of B antigens on its cell surface, as well as high activity of B transferase in cell extracts (1.28 nmol/mg/h). Therefore, we have chosen these cells as candidates for recipients of DNA transfection assuming that enough amounts of trans-activating factors are present to support the expression of the introduced reporter constructs. We have used the electroporation method to obtain the higher transfection efficiency with the KATO III cells. pGL3-promoter vector containing the SV40 promoter, which exhibits strong activity in mammalian cells, and pGL3basic vector without the promoter sequence were used as positive and negative controls, respectively. Forty-eight hours after transfection, the cells were harvested and the luciferase activities in cell extracts were determined. The promoter activity of EN construct was at least 20-fold higher than that of pGL3-basic vector, and 3-fold lower than that of pGL3-promoter vector, which demonstrated the promoter activity of the 5Ј-upstream region of the human ABO genes (Fig. 2).
To better locate the sequences essential for the transcription, 2 T. Kuroki, personal communication. we have constructed a series of the human ABO promoterluciferase chimeric plasmids in which 5Ј-flanking regions of the ABO gene of different lengths were fused with the coding sequence of the luciferase gene. The left panel of Fig. 2 depicts the 5Ј-flanking region of the human ABO gene and various restriction enzyme cleavage sites used to generate the fragments subcloned into the promoterless luciferase vector. Deletion of the human ABO gene upstream region from position Ϫ3950 to Ϫ3252, Ϫ2984, Ϫ2109, Ϫ1685, Ϫ1493, or Ϫ666 resulted in a large decrease in luciferase activity, demonstrating that the important sequences for the ABO gene transcription were contained within the deleted regions. The strongest decreasing effect appeared to be related to the deletion of the sequence between Ϫ3950 and Ϫ3252 containing 4 tandem copies of a 43-bp repeat unit. Surprisingly, however, deletion of the sequence from Ϫ3252 to either Ϫ202 or Ϫ117 resulted in an increase in the luciferase activity, suggesting that negative element(s) for the ABO gene transcription were present within this deleted region (Ϫ666 to Ϫ117). Furthermore, deletion of the sequence from Ϫ117 to Ϫ35 resulted in a large decrease in luciferase activity, showing the importance of promoter sequences immediately upstream of the transcription start site (cap site) in ABO gene expression. Sequence inspection of the region proximal to the cap site revealed potential binding sites for the regulatory protein Sp1. Taken together, two regions (Ϫ3950 to Ϫ3253 and Ϫ117 to ϩ31) appear to function as positive regulatory cis-elements for the transcription of ABO genes in KATO III cells, whereas negative elements seem to exist in the region between Ϫ3252 and Ϫ118. To further examine the promoter activity of the region between Ϫ3950 and ϩ31, an internal deletion downstream of the XmnI site at Ϫ3252 was introduced. The Ϫ3950 upstream terminus was chosen, because the maximum reporter activity was observed with the construct PN containing the region between the PflMI site at Ϫ3950 and the NcoI site at ϩ31 as shown in Fig. 2. An internal deletion located between the XmnI site at Ϫ3251 and the SacII site at Ϫ118 (construct PXm/SN) was constructed by introducing the PflMI/XmnI fragment in SN construct. Transfection of construct PXm/SN into KATO III cells showed a 21-fold higher luciferase activity than did that of construct PN (Fig. 3). This result with the internal deletion seemed consistent with those obtained with upstream end deletions confirming the presence of positive elements for transcription in the PflMI/XmnI fragment.
The sequence immediately upstream of the cap site (Ϫ666 to Ϫ117) seemed to be sufficient for the repression of the ABO cap site-proximal promoter activity in constructs XhN and KN. To examine whether additional regions are involved in the repression of the ABO cap site-proximal promoter activity in the construct XmN, the fragment between the XmnI site at Ϫ3252 and the PstI site at Ϫ2371 was inserted into downstream of the SV40 enhancer in construct SV/SV-XmPst. Addition of this fragment resulted in a loss of 30% in luciferase activity (data not shown). Thus, it is likely that other negative cis-element(s) are present in the region between the XmnI and the SacII sites.
We have obtained another result to support the presence of positive enhancer elements within the Ϫ3950 to Ϫ3252 sequence. When the PflMI-to-XmnI (PXm) fragment was separated from the ABO promoter-proximal sequence and linked directly to the SV40 promoter of pGL3-promoter vector (PXm/SV construct in Fig. 3), this construct expressed an 18fold increase in luciferase activity when compared with the original vector without this fragment. Several additional constructs were prepared from the PXm/SV plasmid and used for the further characterization of the ABO enhancer region. The sequence located between Ϫ3899 and Ϫ3618 appeared to be a significant functional component. Presence of additional sequence between Ϫ3617 and Ϫ3470 in construct Ϫ3899H/SV yielded a reduced activity, which implicated the presence of a negative regulatory element in the sequence. In agreement with this suggestion, a similar reduction was observed in the constructs PH/SV and P-3618/SV. The Ϫ3899 to Ϫ3618 sequence contains 4 tandem copies of a 43-bp repeat unit as shown in Fig. 1. Strikingly, deletion of the minisatellite in the construct Ϫ3899:Ϫ3618⌬/SV resulted in almost complete loss of the luciferase activity within a factor of 2 compared with the activity of pGL3-promoter vector. Thus, the minisatellite appeared to be crucial for the enhancer activity of the human ABO genes.
We have next examined whether the region containing the minisatellite possessed promoter-, position-, and/or orientation-independent enhancer activity. The Ϫ3899 to Ϫ3618 se-quences were inserted either immediately upstream of the various promoters in the same orientation as in the ABO gene, or downstream of the luciferase coding sequence in both the same and opposite orientations. The resulting plasmids were transfected into KATO III cells, and the reporter activity was analyzed. The results are shown in Tables I and II. Increased reporter activity was also observed with constructs containing the heterologous promoters (Table I). Furthermore, the Ϫ3899 to Ϫ3618 sequence was shown with both ABO gene and SV40 promoters to exert enhancer activity irrespective of its position and orientation (Table II). Constructs with the fragment inserted upstream of the promoters showed higher activity than the corresponding ones with the fragment inserted downstream of the luciferase gene. We suspect a distance effect for this difference in the activity. Taken together, we concluded that the Ϫ3899 to Ϫ3618 fragment containing minisatellite functioned in a promoter-, position-, and orientation-independent manner.
Identification of Protein-binding Sites within the Tandem  a The results are expressed as an average relative activity compared to that observed for the pGL3-Promoter vector. Standard deviations are indicated for a minimum of three repetitions. b The constructs as depicted above were transiently transfected into KATO III cells and assessed for functional importance; 10 g of luciferase reporter and 4 g of ␤-galactosidase control vector were used for each analysis.
c The Ϫ3899 to Ϫ3618 sequence was cloned upstream of the ABO gene promoter, thymidine kinase promoter, and SV40 promoter in the same orientation as it appears in the ABO gene. Arrow to the right indicates 5Ј-CCAAT-3Ј on upper strand.
Repeat DNA Sequence-We have performed electrophoretic mobility shift assays (EMSAs) to investigate the nature of specific DNA-protein interaction(s) within the minisatellite. The locations of gene fragments used either as radiolabeled EMSA probes or as specific unlabeled competitors are shown schematically in Fig. 1. The results shown in Fig. 4A demonstrate that oligonucleotide TR43, containing the entire repeat unit in the minisatellite, did bind a KATO III cell nuclear factor (lane 2), and that this binding was inhibited by nonlabeled competitor TR43 (lanes 3 and 4), suggesting that the formation of the DNA-protein complex was specific. This complex, however, was not affected by the addition of either unlabeled TR38 (lanes 5 and 6) or TR39 oligonucleotide (lanes 7 and 8) containing a deletion of 5 or 4 bp, respectively, from the repeat unit, indicating that TR43 bound with a KATO III cell nuclear factor through the 5Ј-half of the repeat unit. Actually, the DNAprotein complex was competed for by the addition of the TR23 oligonucleotide (lanes 9 and 10) which contained only the 5Јhalf of the repeat unit. When either TR38 or TR39 was radiolabeled as a probe, definite binding of nuclear proteins was not observed (lanes 11 and 12). These results also support that the 5Ј half-sequence is important for the binding.
To examine whether the 5Ј-half of the repeat unit is sufficient for binding with a nuclear factor, labeled TR23 oligonucleotide was used as an EMSA probe. Fig. 4B shows that this probe bound to a KATO III cell nuclear factor (lane 2). The complex formation was effectively inhibited by the addition of either the TR23 (lane 3) or the TR43 (lane 4) oligonucleotide as FIG. 4. The 43-bp repeat unit of the minisatellite 3.8 kb upstream from the transcription start site of the ABO blood group gene binds a nuclear factor. A, EMSA shows that the 43-bp sequence within the minisatellite binds a protein in nuclear extract prepared from KATO III cells. DNA binding reactions were performed with radiolabeled probes: TR43 probe (lanes 1-10), TR39 (lane 11), and TR38 (lane 12) in the absence or presence of either a 100-or 200-fold molar excess of competing unlabeled oligonucleotides as indicated. B, TR23 oligonucleotide is sufficient for binding of the nuclear protein to the repeat unit of the minisatellite. EMSAs were performed with nuclear extracts prepared from KATO III cells. DNA binding reactions were carried out using radiolabeled probe TR23 in the absence or presence of 100-fold molar excess of competing unlabeled oligonucleotides as indicated.

TABLE II
Effects of the ABO minisatellite on transcriptional activation when the minisatellite is inserted downstream of the luciferase gene The relative activities were calculated as described for Table I, Footnote a. Standard deviations are indicated for a minimum of three repetitions. Restriction enzyme sites in SN or pGL3-promoter vector are as follows: B, BamHI; S, SalI.
Notion for receptor construction: the Ϫ3899 to Ϫ3618 sequence was cloned into the BamHI/SalI site of the SN or pGL3-promoter vector downstream of the luciferase gene in both orientation. Arrow to the right indicates 5Ј-CCAAT-3Ј on upper strand. competitors, but not by the addition of the mutated oligonucleotide competitors TR38 (lane 5) and TR39 (lane 6). Therefore, the minisatellite likely binds to a KATO III cell-derived nuclear factor through the 5Ј-half of the repeat unit. Inspection of the 23-bp sequence revealed a near-consensus binding site for a heteromeric transcription factor CBF/NF-Y. This transcription factor consists of three subunits (A, B, and C subunits), and all of these three subunits have been shown to be present in the protein-DNA complex and required for DNA binding (20, 21, 28 -30). CBF/NF-Y Specifically Recognizes the 43-bp Repeat Unit of the Minisatellite-To examine whether CBF/NF-Y binds to the minisatellite, competition assays were carried out using unlabeled oligonucleotides containing binding sites for transcription factors. Fig. 5A shows that the TR23 probe bound to a KATO III cell nuclear factor. The complex was effectively competed for by the addition of either unlabeled TR23 oligonucleotide (lane 2) or oligonucleotide E␣-23 containing the strong CBF/NF-Y-binding site in the mouse MHC class II promoter (lane 3). Since the TR23 oligonucleotide contained a half-palindromic sequence TGG(C/A) of the NF-I-binding site (31, 32), we have examined oligonucleotide NF-I (lane 4) which contained the NF-I consensus sequence (21), however, no competition was observed. In addition, anti-NF-I antibody failed to cause a supershift of the DNA-protein complex (data not shown). To more precisely identify the binding location, the CCAAT site was mutated as reported for the mouse E␣ promoter (Fig. 1, CCCAT mutation). The TR23 oligonucleotide containing such a mutated CBF/NF-Y site (mTR23) did not compete for protein binding with the wild-type TR23 probe (lane 5). The mTR23 oligonucleotide did not reveal any binding activity with KATO III-derived nuclear extracts (lane 6), either. Since CBF/NF-Y has been well known to be a ubiquitous transcription factor, DNA binding was examined with the E␣-23 probe using oligonucleotides TR23, mTR23, and NF-I as competitors. Nuclear extracts from KATO III cells revealed a complex with the E␣-23 probe, which had the same relative migration as that observed with TR23 probe (compare lanes 1  and 7). Formation of this complex was inhibited by competition with unlabeled E␣-23 and TR23 oligonucleotides (lanes 8 and 10) but not with NF-I or mutated TR23 oligonucleotide (lanes 9 and 11). Therefore, we concluded that CBF/NF-Y in KATO III cell nuclear extracts bound to the minisatellite sequence at the 23-bp sequence. In Fig. 4A, we have shown that the nucleotide sequences in the TR38 or TR39 oligonucleotide were not sufficient to form DNA-protein complex similar to that obtained with TR43 oligonucleotide. Neither TR38 nor TR39 oligonucleotides contained all the nucleotide sequence necessary for the contact with CBF/NF-Y, which was reported by Chodosh (21). Furthermore, the motif flanking the 3Ј end of the CCAAT sequence was reported to be important for the high-affinity binding of CBF/NF-Y to the C1 element in the grp78/BiP promoter (33). Thus, it was unlikely that CBF/NF-Y could bind strongly to TR38 or TR39 oligonucleotides in the absence of an important nucleotide sequence. Since the E␣-23 competitor inhibited the binding more efficiently than the TR23 oligonucleotide, binding of CBF/NF-Y to the 5Ј-half of the minisatellite may be weaker than that to the mouse E␣ promoter CCAAT box.
To further investigate whether CBF/NF-Y binds to the repeat unit of the minisatellite, additional DNA binding studies were carried out using an anti-NF-Y A-subunit-specific antibody raised against an N-terminal NF-Y A-subunit peptide. The results of those experiments are shown in Fig. 5B. The anti-NF-Y A-subunit antibody supershifted the DNA-protein complexes between TR23 probe and the nuclear factor from KATO III cells (lane 4). In addition, the supershifted DNAprotein complex was also competed for by the E␣-23 oligo containing the strong CBF/NF-Y-binding site (lane 5). Furthermore, DNA binding experiments performed with E␣-23 probe, KATO III cell nuclear extracts, and anti-NF-Y antibody showed similar results to the ones obtained with TR23 probe (lanes  7-12). The supershifted DNA-protein complex was also competed for by the addition of TR23 oligonucleotide (lane 11). An  6 and 12). Radiolabeled probes used are TR23 oligonucleotide (lanes 1-6) and oligo E␣-23 (lanes [7][8][9][10][11][12]. Competition was performed with unlabeled E␣-23 oligo or TR23 oligo in a 100-fold molar excess. anti-Egr-1 antibody raised against a C-terminal Egr-1 peptide failed to supershift the DNA-protein complexes from KATO III nuclear extracts (lanes 6 and 12). Therefore, the minisatellite in the ABO blood group gene likely binds to CBF/NF-Y.
Mutation in CBF/NF-Y-binding Site of the Minisatellite Decreases Enhancer Activity-To ascertain the functional importance of CBF/NF-Y sites in the minisatellite, four copies of the 5Ј wild-type or mutated oligonucleotide were inserted upstream of the SV40 promoter. Transfection of the wild-type construct into KATO III cells showed an 8-fold increase in luciferase activity compared with that of the SV40 promoter vector (Fig. 6, construct 4ϫ23/SV). Mutation of the CBF/NF-Y-binding site, which abrogated CBF/NF-Y binding as shown in Fig. 5A, resulted in a drastic loss of the luciferase activity and approached within a factor of 2 the expression level of pGL3-promoter vector (construct 4ϫm23/SV). This result suggests that CBF/NF-Y sites are important for the enhancer function and that these sites in the minisatellite may play a key role in the human ABO blood group gene expression. The wild-type construct exhibited 8-fold reduction in relative luciferase activity when compared with Ϫ3899:Ϫ3618/SV. This may imply that the minisatellite and its flanking regions function to exert a synergistic effect, although another explanation may be that the decreased activity may reflect the altered pitch of the CBF/NF-Y-binding sites relative to the SV40 promoter. DISCUSSION In this study, we have identified sequences essential for the transcriptional control of the human ABO blood group genes. Transient transfection of KATO III cells with luciferase reporter constructs demonstrated the importance of the promoter sequence immediately upstream of the transcription start site (cap site) in regulating the ABO gene expression. In addition, the far-upstream sequences located between Ϫ3899 and Ϫ3618 was shown to be crucial for the ABO gene transcription. The region contains a minisatellite composed of 4 tandem copies of a 43-bp repeat unit which exhibits enhancer activity. EMSA analysis demonstrated that the transcription factor CBF/NF-Y binds the minisatellite. A mutant construct, 4ϫm23/SV, containing the mTR23 sequence with 1-bp substitution which abrogates the binding of this factor, showed almost complete loss of activity in KATO III cells. This finding demonstrates that the CBF/NF-Y-binding sites in the minisatellite have a functional role. Thus, transcription of the human ABO blood group gene may be regulated, at least in part, by the binding of CBF/NF-Y to the tandem copies of repeats in the minisatellite. Our finding that the minisatellite has enhancer activity offers additional clues as to how other repetitive sequences may play functional roles.
Minisatellites are highly repetitive DNA sequences typically found in mammalian genomes. They vary in length from a few base pairs to several thousands. They also vary in complexity from simple di-and trinucleotide repeats (microsatellites) to more complicated repetitive elements. Repetitive DNA, most notably trinucleotide repeats, have been implicated in several human diseases such as fragile X syndrome, myotonic dystrophy, and Kennedy's disease (34,35). Although the mechanisms by which repetitive DNA is associated with these diseases have been intensively studied in the past years, they are still unknown. However, the studies on the biological functions of minisatellites have brought new findings. For example, recent studies on trinucleotide repeats have underscored the importance of repetitive DNA in a variety of biological processes, ranging from recombination to generation of nucleosome positioning signals (36,37). Other observations on tandem repeats have demonstrated that some human minisatellites might serve as regulatory elements for cellular transcription. Indeed, the 28-bp repeat unit of a minisatellite 1-kb downstream from the human HRAS1 gene was shown to bind several members of the rel/NF-B family of transcriptional regulatory factors (24). This minisatellite was later reported to possess transcriptional regulatory activity that is dependent on the promoter and the cell lines used in the transient transfection experiments (25). A member of the myc/helix-loop-helix family closely related to upstream factor/major late transcription factor was shown to bind to the 50-bp repeat unit of a minisatellite within the D H -J H interval of the human immunoglobulin heavy chain locus. This minisatellite was demonstrated to compete for upstream factor/major late transcription factor, resulting in suppression of the activation of the adenovirus major late promoter (26). More recently, transcriptional function was recognized for a polymorphic minisatellite in the 5Ј-flanking insulin-linked polymorphic region of the human insulin gene. Numerous highaffinity binding sites for the transcription factor Pur-1 were identified in the minisatellite, and minisatellite-dependent enhancement in transcription was observed with both the native human insulin promoter and a heterologous promoter linked to the reporter gene (27). Moreover, transcriptional activity was dependent on the length and the number of repeats in the minisatellite, with the longer minisatellites (up to ϳ2.5 kb) being more active than the short ones (ϳ0.5-1 kb). Interestingly, a subgroup of short minisatellite seemed to be associated with susceptibility to type I diabetes (38). Our results, together with others, strongly indicate that minisatellites influence the behaviors of the nearby genes.
The results that transcriptional regulation of the human ABO gene is dependent upon the binding of transcription factor CBF/NF-Y to the minisatellite may provide clues as to how the expression of ABO genes is controlled during differentiation of epithelial cells. Although CBF/NF-Y is a well known constitutive and ubiquitous factor, a possibility remains that CBF/ NF-Y may be modulated during cellular differentiation. CBF/ NF-Y activity was shown to be serum dependent in IMR-90 diploid fibroblasts (39). It is of interest that the serum-induced enhancement of CBF/NF-Y binding to the distal CCAAT box of the human TK gene was mediated by the enhanced expression of NF-YA subunit, but not NF-YB. Recently, another example of enhanced CBF/NF-Y activity by elevated levels of NF-YA subunit has been reported in the heme-treated Friend leukemia cells and during the monocyte-to-macrophage differentia- tion process (40). The relevance of this enhanced binding of CBF/NF-Y to the CCAAT box has also been reported in the promoter region of the ferritin H gene in differentiated Caco-2 cells, a human colon cell line capable of differentiating to enterocyte-like cells (41). Therefore, it seems plausible to assume that the ABO gene expression is influenced by CBF/NF-Y binding activity through cell differentiation.
Various human cancers are known to frequently lose blood group A/B antigens. Recently, a complete correlation between the absence of the ABO mRNA transcript and the absence of blood group A/B antigens has been reported (13). The authors also showed that growth stimulation with the cholera toxin B or epidermal growth factor led to a total loss of the ABO mRNA in a urothelial cell line, suggesting the possible linkage between cell proliferation and down-regulation of ABO mRNA transcripts. This down-regulation may be due to both decreased transcription and increased breakdown of mRNA formed. Reduced levels of transcription can be caused several ways, including inhibition of an active transcription factor or appearance of repressors binding to the ABO gene sequence. Our results help delineate the mechanism of A/B antigen disappearance which correlates with invasive and metastatic potentials of cancer cells. Further investigation is required to elucidate the inhibitory mechanisms.