Transcriptional Regulation of the Human β-1,4-Galactosyltransferase V Gene in Cancer Cells

β-1,4-Galactosyltransferase (β-1,4-GalT) V is a constitutively expressed enzyme that can effectively galactosylate the GlcNAcβ1→6Man group of the highly branched N-glycans that are characteristic of tumor cells. Upon malignant transformation of cells, the expression of the β-1,4-GalT V gene increases in accordance with the increase in the amounts of highly branched N-glycans. Lectin blot analysis showed that the galactosylation of highly branched N-glycans is inhibited significantly in SH-SY5Y human neuroblastoma cells by the transfection of the antisense β-1,4-GalT V cDNA, indicating the biological importance of the β-1,4-GalT V for the functions of highly branched N-glycans. We cloned the 2.3-kb 5′-flanking region of the human β-1,4-GalT V gene, and we identified the region –116/–18 relative to the transcription start site as that having promoter activity. The region was found to contain several putative binding sites for transcription factors, including AP2, AP4, N-Myc, Sp1, and upstream stimulatory factor. Electrophoretic mobility shift assay showed that Sp1 binds to nucleotide positions –81/–69 of the promoter region. Mutations induced in the Sp1-binding site showed that the promoter activity of the β-1,4-GalT V gene is impaired completely in cancer cells. In contrast, the promoter activity increased significantly by the transfection of the Sp1 cDNA into A549 human lung carcinoma cells. Mithramycin A, which inhibits the binding of Sp1 to its binding site, reduced the promoter activation and expression of the β-1,4-GalT V gene in A549 cells. These results indicate that Sp1 plays an essential role in the transcriptional activity of the β-1,4-GalT V gene in cancer cells.


Takeshi Sato ‡ and Kiyoshi Furukawa
We cloned the 2.3-kb 5-flanking region of the human ␤-1,4-GalT V gene, and we identified the region ؊116/؊18 relative to the transcription start site as that having promoter activity. The region was found to contain several putative binding sites for transcription factors, including AP2, AP4, N-Myc, Sp1, and upstream stimulatory factor. Electrophoretic mobility shift assay showed that Sp1 binds to nucleotide positions ؊81/؊69 of the promoter region. Mutations induced in the Sp1-binding site showed that the promoter activity of the ␤-1,4-GalT V gene is impaired completely in cancer cells. In contrast, the promoter activity increased significantly by the transfection of the Sp1 cDNA into A549 human lung carcinoma cells. Mithramycin A, which inhibits the binding of Sp1 to its binding site, reduced the promoter activation and expression of the ␤-1,4-GalT V gene in A549 cells. These results indicate that Sp1 plays an essential role in the transcriptional activity of the ␤-1,4-GalT V gene in cancer cells.
One of the most prominent transformation-associated changes in the sugar chains of glycoproteins is an increase in the large N-glycans of cell surface glycoproteins (reviewed in Ref. 1). This was discovered by comparing the gel filtration patterns of glycopeptides obtained by Pronase digestion of met-abolically labeled normal and malignant cell glycoproteins (2)(3)(4). Detailed structural studies of N-glycans isolated from baby hamster kidney cells and polyoma-or Rous sarcoma virustransformed baby hamster kidney cells showed that the increase in the GlcNAc␤136 branch attached to the Man␣136Man arm of the trimannosyl cores is the structural basis for this phenomenon (5,6). This was further confirmed by studies on a number of malignant cell lines transformed with different agents (7)(8)(9).
Three transformed cell lines, MT1, MTPy, and MTAg, established from mouse NIH3T3 cells by transfection with the SV40 or polyoma virus early gene segments (10), showed marked differences in their tumorigenic and metastatic potentials when transplanted subcutaneously or intravenously into athymic mice. Structural analysis of the N-glycans of these cells revealed that only about 20% of the glycoproteins from 3T3 and MT1 cells have highly branched N-glycans with the Gal␤134GlcNAc␤136(Gal␤134GlcNAc␤132)Man branch compared with 31 and 39% of the glycoproteins from MTPy and MTAg cells, respectively, indicating a proportionality between increased highly branched N-glycans and tumorigenic and metastatic potentials (11). The increased expression of the highly branched N-glycans also correlates with the tumor forming activities of various other transformed cells (7,9). The increased cell surface binding of fluorescein-labeled leukophytohemagglutinin (L-PHA), 1 which specifically binds to highly branched N-glycans with the Gal␤134GlcNAc␤136-(Gal␤134GlcNAc␤132)Man branch (12), was found to correlate with the metastatic potentials of mouse mammary carcinoma cells, mouse lymphoma cells, transformed rat fibroblasts, and human breast cancers (13,14). Thus the results of several studies established that an increase in highly branched Nglycans is actually related to the in vivo tumor forming and metastatic potentials of transformed cells.
In accordance with the structural studies, the specific activity of UDP-GlcNAc:Man N-acetylglucosaminyltransferase (Gl-cNAcT) V, which synthesizes the GlcNAc␤136 branch, has been shown to be elevated 2-3-fold in transformed cells (15,16). In contrast, no significant changes in the specific activities of other glycosyltransferases involved in the biosynthesis of N-glycans have been observed in transformed cells (15,17). Because the Gal␤134GlcNAc outer chains form the basis for the expression of a variety of carbohydrate antigens, whether or not the gene expression of UDP-Gal:GlcNAc ␤-1,4-galactosyltransferases (␤-1,4-GalT) I-VI, most of which are involved in the biosynthesis of N-glycans (18,19), is changed by malignant transformation was investigated using NIH3T3 and MTAg cells as described above. Northern blot analysis revealed that the ␤-1,4-GalT V transcript increases 2-3-fold and the ␤-1,4-GalT II transcript decreases to one-tenth, whereas those of other ␤-1,4-GalTs remain constant upon malignant transformation (17). Similar results were obtained in several human cancer cell lines (20), indicating that the expression pattern of ␤-1,4-GalT genes also changes upon malignant transformation of cells despite little apparent change in the enzymatic activity. Because our preliminary study suggested that the ␤-1,4-GalT V can effectively galactosylate the GlcNAc␤136 branch, which is synthesized by GlcNAcT V (21), it is important to elucidate the biological significance of the ␤-1,4-GalT V and the mechanism by which ␤-1,4-GalT V gene is regulated in cancer cells. For these purposes, we showed the biological importance of the galactosylation of highly branched N-glycans by the ␤-1,4-GalT V and then isolated the promoter region of the human ␤-1,4-GalT V gene and examined the cis-elements and trans-acting factors that regulate ␤-1,4-GalT V gene expression.
Lectin Blot Analysis of Antisense ␤1,4-GalT V cDNA-transfected Cells-SH-SY5Y cells (1 ϫ 10 5 cells) were transfected with 2 g of the pcDNA3.1 (Invitrogen) as a control or the pcDNA3.1 containing a partial sequence of human ␤-1,4-GalT V cDNA (Ϫ5 to ϩ500 relative to the initiation codon) in an antisense orientation and the FuGENE6 transfection reagent (Roche Applied Science). Cells were cultured for 72 h, and then the plasmid-transfected cells were selected by culturing in a medium containing geneticin (500 g/ml G418 sulfate, Sigma) for 2 weeks. Membrane glycoprotein samples were prepared from the mockand antisense ␤-1,4-GalT V cDNA-transfected cells, subjected to SDS-PAGE using a Mini Protean II Electrophoresis Cell (Bio-Rad), and then transferred to polyvinylidene difluoride filters using a Mini Trans-Blot Electrophoretic Transfer Cell (Bio-Rad). Western blot analysis using HRP-conjugated RCA-I and L-PHA was performed by the method as described previously (22).
Isolation, Southern Blot Analysis, and DNA Sequencing of Human Genomic DNA Clones-A human placenta genomic library in FIX II (Stratagene, La Jolla, CA) was screened using the 480-bp EcoRI-ApaI fragment of human ␤-1,4-GalT V cDNA (21) as a probe. The probe was random-labeled with [␣-32 P]dCTP using a "Ready to Go TM " DNA labeling kit (Amersham Biosciences). Positive clones were obtained by four successive screenings. Phage DNAs were purified and digested with several restriction endonucleases, separated on a 1% agarose gel, and transferred onto a GeneScreen nylon membrane (PerkinElmer Life Sciences). The membranes were hybridized with the biotinylated 100-bp PCR fragment containing the initiation codon of the ␤-1,4-GalT V gene as a probe according to the manufacturer's instructions for the random primer biotinlabeling kit (PerkinElmer Life Sciences). The hybridization-positive 1.5-kb NdeI fragment or the 1.3-kb BamHI fragment was subcloned into the NdeI site of pGEM-T Easy vector (Promega, Madison, WI) to generate pGEM/Nde or into the BamHI site of pBluescript II KS vector (Stratagene) to generate pBlue/Bam. The nucleotide sequences of the DNA fragments were determined by the dideoxynucleotide chain termination method (23) by using an Auto Read Sequencing kit (Amersham Biosciences). To identify the putative binding sites of transcription factors, the 2-kb 5Ј-flanking region of the human ␤-1,4-GalT V gene was analyzed by the MatInspector program (24).
RNA Ligase-mediated Rapid Amplification of the 5Ј cDNA End (RLM-RACE)-RLM-RACE analysis was performed to map the transcription start site using a GeneRacer kit (Invitrogen) according to the manufacturer's instructions. In brief, the total RNA preparation (3 g) from SH-SY5Y cells was treated with calf intestinal phosphatase to dephosphorylate non-mRNAs or truncated mRNAs, and then treated with tobacco acid pyrophosphatase to remove the 5Ј-cap structure from the full-length mRNAs. The GeneRacer RNA Oligo (5Ј-CGACUGGAG-CACGAGGACACUGACAUGGACUGAAGGAGUAGAAA-3Ј) was ligated to the 5Ј-ends of the decapped mRNAs. Subsequently, the cDNAs were synthesized with random primers and SuperScript III reverse transcriptase. Two antisense oligonucleotide primers corresponding to the coding region of the ␤-1,4-GalT V cDNA as shown in Fig. 4A were synthesized and used for the PCR. The primary PCR was conducted using the synthesized cDNAs as templates, the antisense gene-specific primer TS138 (5Ј-GGAGTCTTTCAGGGCAGGTATGGTT-3Ј; complementary to nucleotides ϩ355/ϩ331 relative to the initiation codon), and the GeneRacer 5Ј-Primer supplied with the kit under the following conditions: 98°C, 10 s; 65°C, 30 s; and 72°C, 1 min; 40 times. The secondary PCR was conducted using the primary PCR products as templates, the antisense gene-specific primer TS137 (5Ј-TGCCGGGC-GCCACATAGACGAAGTA-3Ј; complementary to nucleotides ϩ112/ϩ88 relative to the initiation codon), and the GeneRacer 5Ј-Nested Primer supplied with the kit under the following conditions: 98°C, 10 s; 65°C, 30 s; and 72°C, 1 min; 25 times. The primary and secondary PCR products were subjected to electrophoresis in a 2% agarose gel, stained with ethidium bromide, and visualized under a UV lamp as described previously (20). As a final product, a DNA fragment comprising ϳ300 bp was obtained in secondary PCR and subsequently cloned into pGEM-T Easy vector and sequenced.
Reporter Plasmid Constructions-To assay the promoter activity, a variety of the 5Ј-flanking regions of the ␤-1,4-GalT V gene that differed in length was inserted into the firefly luciferase reporter vector, pGL3-Basic (Promega), which contained no eukaryotic promoter or enhancer element. The strategy for cloning of the fragments of the ␤-1,4-GalT V gene promoter into a pGL3-Basic vector was as follows, with the numbers indicating the nucleotide positions relative to the transcription start site. 1) For pGL(Ϫ2099/ϩ170), the 1.0-kb BamHI-NotI fragment was excised from pBlue/Bam (Fig. 2C) and cloned into the BamHI-NotI sites of pGEM/Nde (Fig. 2B) to generate pGEM/Nde-Not (Fig. 2D). Similarly, the 2.3-kb SacI-NotI fragment was excised from pGEM/Nde-Not and subcloned into the SacI-NotI sites of the pBluescript II KS vector to generate pBlue/Nde-Not. The 2.3-kb SacI-XhoI fragment was excised from pBlue/Nde-Not and subcloned into the SacI-XhoI sites of the pGL3-Basic vector. In order to minimize the effect of sequences derived from the pBluescript II KS vector or pGEM-T Easy vector, the resultant plasmids were digested with NotI and SpeI to remove the multiple cloning sites in the vectors. The ends were blunted with T4 DNA polymerase and then self-ligated with T4 DNA ligase. 2) For pGL(Ϫ1121/ϩ170), pGL(Ϫ2099/ϩ170) was digested with KpnI and Sse8387I. The ends were blunted and then self-ligated. 3) For pGL(Ϫ2099/Ϫ835), the KpnI-BamHI fragment was excised from pGL(Ϫ2099/ϩ170) and subcloned into the KpnI-BglII sites of the pGL3-Basic vector. 4) For pGL(Ϫ834/ϩ170), the BamHI-HindIII fragment was excised from pGL(Ϫ2099/ϩ170) and subcloned into the BglII-HindIII sites of the pGL3-Basic vector. 5) For pGL(Ϫ551/ϩ170), pGL(Ϫ2099/ϩ170) was digested with NdeI, and the ends were selfligated. 6) For pGL(Ϫ313/ϩ170), pGL(Ϫ2099/ϩ170) was digested with SacI and PstI. The ends were blunted and then self-ligated. 7) For pGL(Ϫ116/ϩ170), pGL(Ϫ2099/ϩ170) was digested with SacI and BssHII. The ends were blunted and then self-ligated. 8) For pGL(ϩ23/ ϩ170), the NaeI-HindIII fragment was excised from pGL(Ϫ116/ϩ170) and then subcloned into the SmaI-HindIII sites of pGL3-Basic vector. 9) For pGL(Ϫ2099/Ϫ117), pGL(Ϫ2099/ϩ170) was digested with BssHII and HindIII. The ends were blunted and then self-ligated. 10) For pGL(Ϫ116/ϩ22), pGL(Ϫ116/ϩ170) was digested with KpnI and NaeI and subcloned into the KpnI-SmaI sites of the pGL3-Basic vector. 11) For pGL(Ϫ116/Ϫ18), pGL(Ϫ116/ϩ170) was digested with SfiI and HindIII. The ends were blunted and then self-ligated. 12) For pGL(Ϫ17/ ϩ22), pGL(Ϫ116/ϩ22) was digested with KpnI and SfiI. The ends were blunted and then self-ligated. 13) For pGL(Ϫ116/ϩ170)-Sp1 mutation 1 and pGL(Ϫ116/ϩ170)-Sp1 mutation 2, both plasmids were constructed using pGL(Ϫ116/ϩ170) as a template with a GeneEditor in vitro sitedirected mutagenesis system (Promega) according to the manufacturer's instructions. The mutations in the Sp1-binding site are indicated by an underline at nucleotide positions Ϫ81/Ϫ69 (Sp1 mutation 1, GGC-CCCAATCCC and Sp1 mutation 2, GGCCCCGTTTCCC, instead of the wild type, GGCCCCGCCTCCC). The correct orientation and sequences of all plasmid constructs were verified by sequence analysis. The unaltered plasmid, pGL3-Basic, was used as a promoterless control, and the plasmid, pRL-TK (Promega) containing the Renilla luciferase gene driven by the herpes simplex virus thymidine kinase (TK) promoter, was used as a normalization control to correct for variable transfection efficiencies. pGL3-SV40 (Promega) contained the firefly luciferase gene driven by the SV40 promoter as a positive control.
Transfection and Luciferase Assay-One day prior to transfection, the cell lines (1 ϫ 10 5 cells each) were seeded in 35-mm tissue culture dishes. Cells were transfected with 1 g of the reporter plasmid, 0.1 g of pRL-TK, and the FuGENE6 transfection reagent (Roche Applied Science). Cells were harvested 48 h after transfection, lysed in 200 l of lysis buffer, and subjected to freeze-thaw lysis. The firefly or Renilla luciferase activity in 10 l of cell lysate was determined with a Dual-Luciferase Reporter Assay System (Promega) by Luminescence PSN AB-2200 (Atto Instruments, Tokyo, Japan). Firefly luciferase activities were normalized to Renilla luciferase activities except in the case of CMV-Sp1, which significantly stimulated the TK promoter, which was co-transfected. In the latter case, the firefly luciferase activities were normalized to the protein content of each sample as described previously (25,26). The results show the mean values of three experiments with standard errors.
Electrophoretic Mobility Shift Assay (EMS Assay)-Nuclear extract was prepared from SH-SY5Y cells with NE-PER nuclear and cytoplasmic extraction reagents (Pierce), containing multiple protease inhibitors, such as benzamidine, aprotinin, leupeptin, and phenylmethylsulfonyl fluoride, according to the manufacturer's instructions. Synthesized oligonucleotides were 3Ј-end-labeled with biotin-N4-CTP and terminal deoxynucleotidyltransferase according to the instructions for the Biotin 3Ј-end DNA labeling kit (Pierce), and then the biotinylated complementary oligonucleotides were annealed to generate the double-stranded oligonucleotides as probes. The sequences of the upper strands of the oligonucleotides used are listed in Table I. EMS assays were performed according to the instructions for the LightShift Chemiluminescent EMS assay kit (Pierce). In brief, the binding reaction was performed by preincubating 4 g of nuclear extract with 50 ng of poly(dI-dC) in buffer solution containing 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 5 mM MgCl 2 , and 1 mM dithiothreitol for 10 min at room temperature. Approximately 30 -50 fmol of probe was added, and the reaction mixtures were incubated for 20 min at room temperature. Subsequently, the samples were separated from the free probes by electrophoresis in a 6% nondenaturing polyacrylamide gel using 45 mM Tris borate buffer (pH 8.5) containing 1.25 mM EDTA as a running buffer. The samples in the gel were electrophoretically transferred to a Hybond N ϩ membrane (Amersham Biosciences). To detect DNA-protein complexes, the membranes were incubated with streptavidin-conjugated horseradish peroxidase and then visualized with LightShift Luminol/ Enhancer solution and LightShift Stable Peroxide solution according to the manufacturer's instructions. For competition experiments, unlabeled Sp1-consensus oligonucleotides were added in a 100-fold molar excess prior to the addition of the biotinylated probes. To identify the transcription factor comprising the DNA-protein complexes by the supershift assay, the nuclear extract was incubated in the binding buffer for 60 min at 4°C with anti-human Sp1 antibody or anti-human Sp3 antibody prior to the addition of the biotinylated probes.
Reverse Transcription-PCR Analysis-Total RNA preparations were obtained from mithramycin A-treated and -untreated A549 cells using Sepasol RNA I total RNA isolation reagent (Nacalai Tesque, Kyoto, Japan). RT-PCR analysis was conducted using the cDNAs as templates and oligonucleotide primers specific to the ␤-1,4-GalT V gene. The PCR products were analyzed by agarose gel electrophoresis as described previously (20).

Decreased Galactosylation of Highly Branched N-Glycans by
Transfection of Antisense ␤-1,4-GalT V cDNA into SH-SY5Y Cells-To elucidate the biological significance of the ␤-1,4-GalT V in cancer cells, the antisense ␤-1,4-GalT V cDNA was transfected into SH-SY5Y cells. RT-PCR analysis using oligonucleotide primers specific to the ␤-1,4-GalT V gene showed that the expression level of the ␤-1,4-GalT V transcript decreases 20 -30% by the transfection of the antisense ␤-1,4-GalT V cDNA into SH-SY5Y cells when compared with that of the control cells (data not shown). Under these conditions, the constant expression levels of the glyceraldehyde-3-phosphate dehydrogenase (G3PDH) gene were obtained in both cells (data not shown). Membrane glycoprotein samples were prepared from the mock-and antisense ␤-1,4-GalT V cDNA-transfected cells. They were subjected to SDS-PAGE, and proteins were transferred electrophoretically to polyvinylidene difluoride filters. When the filters were stained with Coomassie Brilliant Blue, both samples contained similar protein components (Fig. 1, CBB). In order to examine whether or not the galactosylation changes in SH-SY5Y cells by the transfection of the antisense ␤-1,4-GalT V cDNA, lectin blot analysis was performed using HRP-conjugated RCA-I and L-PHA. The filters were initially subjected to mild acid treatment to remove sialic acids prior to incubation with lectins. When the filters were incubated with RCA-I, which interacts with oligosaccharides terminating with the Gal␤134GlcNAc group (27) Fig. 9A. A549 cells. 2 Therefore, the ␤-1,4-GalT V is considered to be very important in cancer cells for expressing the highly branched N-glycans that are involved in abnormal cell growth and metastasis (11,13).
Isolation of the 5Ј-Promoter Region of Human ␤-1,4-GalT V Gene-We identified in the GenBank TM a human genomic clone (RP5-1063B2) derived from 20q13.1-13.2 containing the human ␤-1,4-GalT V cDNA sequence reported by us (DDBJ/ GenBank TM /EMBL Data Bank, accession number AB004550). The genomic sequence completely matched the coding region, ϩ114/ϩ1,167, relative to the initiation codon of the ␤-1,4-GalT V gene reported, but lacked the 5Ј-flanking region and the coding region upstream from ϩ113. To isolate the 5Ј-flanking region of the ␤-1,4-GalT V gene, a human placenta genomic library was screened by the plaque hybridization method using the 480-bp fragment containing the initiation codon of the ␤-1,4-GalT V cDNA as a probe, and 10 genomic clones were obtained. In order to identify the genomic clones that contained the 5Ј-promoter region, Southern blot analysis was performed using a 100-bp fragment containing the coding region ϩ1/ϩ100 relative to the initiation codon as a probe. The hybridizationpositive 1.5-kb NdeI fragment or 1.3-kb BamHI fragment was obtained and subcloned into the NdeI site of a pGEM-T Easy vector to generate pGEM/Nde (Fig. 2B) or into the BamHI site of a pBluescript II KS vector to generate pBlue/Bam (Fig. 2C). Nucleotide sequence analysis showed that the BamHI fragment contained a 1.0-kb 5Ј-flanking region, the first exon and intron, and the NdeI fragment contained a 1.2-kb 5Ј-flanking region upstream from the BamHI site and a 0.3-kb BamHI/ NdeI overlapping region ( Fig. 2A).
Genomic Structure of the ␤-1,4-GalT V Gene-We also found in the GenBank TM another human genomic clone, RP5-1041C10 from chromosome 20, containing the coding region ϩ57/ϩ115 relative to the initiation codon of the ␤-1,4-GalT V gene. The alignment of the human ␤-1,4-GalT V cDNA, the genomic clone reported in the present study, and two genomic clones in the GenBank TM allowed us to deduce the exon-intron organization of the gene, which is shown in Fig. 3A. Although the human ␤-1,4-GalT I, II, III, and IV genes consist of six exons and five introns (28 -30), the human ␤-1,4-GalT V gene consists of nine exons and eight introns. All the exon-intron boundaries were found to adhere to the consensus sequence (Fig. 3B). This exon-intron structure is highly conserved in the mouse. 3 A similar genomic organization has been observed for the human ␤-1,4-GalT VI gene (31), which was also suggested by the similarity between human ␤-1,4-GalTs V and VI at the amino acid levels (32) and by a closer position of human ␤-1,4-GalT VI to that of human ␤-1,4-GalT V rather to other ␤-1,4-GalTs in the phylogenetic tree (33).
Mapping of the Transcription Start Site-To determine the position of the transcription start site, RLM-RACE analysis was performed using two sets of oligonucleotide primers, 5Ј-Primer and TS138, and 5Ј-Nested Primer and TS137, respectively (Fig. 4A). As shown in Fig. 4B, the primary PCR using 5Ј-Primer and TS138 produced a smeared product (Fig. 4B, lane 1), whereas the secondary PCR using 5Ј-Nested Primer and TS137 produced a 300-bp product (Fig. 4B, lane 2). The product was extracted from the agarose gel and cloned into pGEM-T Easy vector for sequencing. The results showed that RNA Oligo was linked to an adenine residue at nucleotide position 188 bp upstream from the initiation codon (Fig. 4C), indicating that the transcription of the ␤-1,4-GalT V gene starts at this position. Prolonging the extension time in the primary and secondary PCRs did not amplify any larger PCR products (data not shown). Therefore, the transcription start site of the ␤-1,4-GalT V gene in SH-SY5Y cells is located 188 bp upstream from the initiation codon. The immediate upstream region from the transcription start site lacks canonical TATA and CAAT boxes. However, five GC-rich sequences, including the Sp1-binding sites, are found in the regions Ϫ82/Ϫ69, Ϫ62/ Ϫ49, Ϫ56/Ϫ43, Ϫ50/Ϫ37, and Ϫ16/Ϫ3 (Fig. 5).
Functional Analysis of the ␤-1,4-GalT V Gene Promoter-To analyze the activity of the promoter in the ␤-1,4-GalT V gene, a construct containing the full-length promoter, pGL(Ϫ2099/ ϩ170), was constructed by inserting the 2.3-kb NdeI-NotI genomic fragment (Fig. 2D) containing the promoter set upstream of the firefly luciferase cDNA into the pGL3-Basic vector (Fig. 6). Upon transient transfection into several cancer cell lines, pGL(Ϫ2099/ϩ170) exhibited significant luciferase activity as compared with that of the pGL3-Basic vector (data not shown). The luciferase assay showed that the ␤-1,4-GalT V gene promoter is activated mostly in SH-SY5Y cells among the 2 T. Sato and K. Furukawa, unpublished data. 3 S. Hayakawa, T. Sato, and K. Furukawa, unpublished data. cell lines examined (Fig. 7). Therefore, SH-SY5Y cells were used to identify the promoter region of the ␤-1,4-GalT V gene. To characterize the cis-elements of the ␤-1,4-GalT V gene promoter, eight additional reporter plasmids containing the promoter in variable lengths, as shown in Fig. 6, were constructed and transfected into SH-SY5Y cells, and the promoter activities were determined. The results showed that pGL(Ϫ116/ϩ170), containing the sequence downstream to nucleotide position Ϫ116, retains relatively strong promoter activity, whereas the promoter activity is significantly reduced for pGL(ϩ23/ϩ170), which lacks the region upstream from nucleotide position ϩ22, indicating the presence of important positive regulatory elements between nucleotide positions Ϫ116 and ϩ23. When pGL(Ϫ2099/Ϫ117) was transfected into SH-SY5Y cells, the luciferase activity was almost at the background level (Fig. 8), indicating that no regulatory element is included between nucleotide positions Ϫ2099 and Ϫ117. For further characterization, the region between nucleotide positions Ϫ116 and ϩ23 was divided into two regions by digestion with SfiI, and pGL(Ϫ116/Ϫ18) and pGL(Ϫ17/ϩ22) were constructed. The luciferase assay showed significant promoter activity in the region between nucleotide positions Ϫ116 and Ϫ18, which con-tains one putative binding site each for AP2, upstream stimulatory factor (USF), N-Myc, and AP4 and four putative binding sites for Sp1 as predicted by the MatInspector program, but no activity was detected in the region between nucleotide positions Ϫ17 and ϩ22, which contains one putative Sp1-binding site (Fig. 9, A and B). Similar results were obtained in other cancer cell lines (data not shown). These results indicate that the region between nucleotide positions Ϫ116 and Ϫ18 is responsible for promoter activation in cancer cells.
Identification of Transcription Factors Bound to the ␤-1,4-GalT V Gene Promoter-To determine the transcription factors that bind to the promoter of the ␤-1,4-GalT V gene, the EMS assay was conducted using three oligonucleotide probes (Fig.  9A) covering the putative binding sites for all transcription factors found in the promoter. Probe A (Ϫ102/Ϫ77) failed to form any complex with the nuclear extract of SH-SY5Y cells (data not shown). In contrast, probe B (Ϫ83/Ϫ58) formed DNAprotein complexes with nuclear extract (Fig. 10, lanes 2, 3 and  6), whereas no complex was formed with probe B in the absence of nuclear extract (Fig. 10, lane 1). Probe B contains one GC box as shown in Fig. 9A. The formation of a major DNA-protein complex was markedly reduced by incubation with the Sp1 mutation probe, which contains mutations in the Sp1-binding site (Fig. 10, lane 4), or excess amounts of unlabeled Sp1consensus oligonucleotides, which contain the Sp1-binding site (Fig. 10, lane 5). Moreover, a major DNA-protein complex was detected as a supershifted band by incubation with anti-Sp1 antibody but not anti-Sp3 antibody (Fig. 10, lane 7, and data not shown). A similar supershifted band with anti-Sp1 antibody was observed when probe B was incubated with nuclear extracts prepared from A549 and HepG2 cells, which express significant levels of Sp1 and Sp3 (34,35), but no supershifted band was detected with anti-Sp3 antibody (data not shown). Mutations in the N-Myc-or USF-binding site did not inhibit the formation of a DNA-protein complex with probe B (data not shown), suggesting that these two factors are not involved in the formation of the DNA-protein complex. These results indicate that probe B binds to Sp1. However, some of the minor bands were not shifted with anti-Sp1 antibody, suggesting that some other transcription factors may have occupied or hindered the Sp1-binding site of probe B. Probe C (Ϫ62/Ϫ36) formed at least three DNA-protein complexes with nuclear extract of SH-SY5Y cells (Fig. 10, lane 8). Because probe C contains three Sp1-binding sites, a supershift assay was performed in the presence of anti-Sp1 antibody. However, no shift was observed (data not shown), suggesting that Sp1 does not bind to probe C. This is also supported by the fact that excess amounts of unlabeled Sp1-consensus oligonucleotides cannot compete with probe C. Moreover, probes containing mutations in the Sp1-binding sites of probe C formed DNA-protein complexes with the nuclear extracts (data not shown), suggesting that Sp1 is not involved in the formation of DNA-protein complexes with probe C. The transcription factors involved in the formation of DNA-protein complexes with probe C remain to be determined.
Mutational Analysis of Sp1-binding Site-The presence of GC-rich sequences within the 5Ј-flanking region is a feature of TATA-less genes, and the expression of such genes is regulated by Sp1 (reviewed in Refs. 36 and 37). Therefore, whether or not the most distal Sp1-binding site at nucleotide positions Ϫ81/ Ϫ69 is involved in the transcriptional activation of the promoter of the ␤-1,4-GalT V gene was investigated. Two different mutations were introduced into the most distal Sp1-binding site of pGL(Ϫ116/ϩ170) (Fig. 11A), and the luciferase activity was determined. The results showed that Sp1 fails to bind to the mutated probes as determined by EMS assay (data not shown). When construct pGL(Ϫ116/ϩ170), containing Sp1 mutation 1, or pGL(Ϫ116/ϩ170), containing Sp1 mutation 2 (see Fig. 11A), was transfected into SH-SY5Y or A549 cells, the luciferase activities decreased more than 95% as compared with cells transfected with the wild type plasmid (Fig. 11B), suggesting that the Sp1-binding site at nucleotide positions Ϫ81/Ϫ69 is essential for the activity of the ␤-1,4-GalT V gene promoter and that other sites for transcription factors may contribute, at most, 5% to promoter activation of the ␤-1,4-GalT V gene. Similar results were obtained in several other cancer cell lines (data not shown). Therefore, the factors bound to probe C are not essential for the promoter activity of the ␤-1,4-GalT V gene in cancer cells. These results indicate that the Sp1-binding site at nucleotide positions Ϫ81/Ϫ69 of the ␤-1,4-GalT V gene plays an essential role in promoter activity in cancer cells.
Activation of the Human ␤-1,4-GalT V Gene by Sp1-To investigate the role of Sp1 in the promoter activation of the ␤-1,4-GalT V gene, pGL(Ϫ116/ϩ170) was transiently co-transfected into A549 cells with CMV-Sp1. After transfection, the cells were cultured for 48 h, and luciferase activity was assayed using cell extracts. The results showed that the ectopic co-expression of Sp1 stimulates the promoter activation of the ␤-1,4-GalT V gene by 5.5-7.5-fold (Fig. 12A). Mithramycin A binds to the GC box in DNA, thereby inhibiting the binding of Sp1 to its binding site (38,39). The treatment of A549 cells with mithramycin A resulted in reduced promoter activity of the ␤-1,4-GalT V gene (Fig. 12B). In order to compare the expression levels of the ␤-1,4-GalT V gene in mithramycin A-treated and -untreated A549 cells, RT-PCR analysis was performed using oligonucleotide primers specific to the ␤-1,4-GalT V gene. The results showed that the expression levels of the ␤-1,4-GalT V gene decrease dramatically upon treatment of A549 cells with mithramycin A as compared with untreated cells (Fig.  12C, ␤-1,4-GalT V), whereas the expression levels of the G3PDH gene in the cells remain constant (Fig. 12C, G3PDH). These results strongly suggest that the expression of the ␤-1,4-GalT V gene is regulated by Sp1 and that Sp1 plays an essential role in the promoter activation of the ␤-1,4-GalT V gene in cancer cells.

DISCUSSION
The ␤-1,4-GalT V gene is expressed at high levels in most human tissues (21), and its expression increases upon malig- nant transformation of cells (17,20). Because the reduction in the expression level of the ␤-1,4-GalT V gene in SH-SY5Y cells resulted in the decreased galactosylation of highly branched N-glycans, the ␤-1,4-GalT V is considered to be involved in the expression of highly branched N-glycans, which are involved in abnormal growth and metastasis of cancer cells (11,13). Recently, the transfection of the antisense ␤-1,4-GalT V cDNA into some cancer cell lines has showed the suppression of tumor development in experimental animals, 4 indicating the particular importance of the ␤-1,4-GalT V for tumor biology. However, the mechanism by which ␤-1,4-GalT V gene expression is regulated remains unknown. In the present study, we cloned and characterized the promoter region of the ␤-1,4-GalT V gene and found that the GC-rich promoter lacks canonical TATA and CCAAT boxes. The lack of TATA and CCAAT boxes appears to be common to mammalian glycosyltransferase genes including human ␤-1,2-GlcNAcT I (40), human ␤-1,4-GlcNAcT III (41), human ␣-2,6-sialyltransferase I (42), mouse ␤-1,6-GlcNAcT (43), mouse ␣-2,8-sialyltransferase II (polysialic acid synthase) (44), and mouse glucuronyltransferase (45). In the promoter region of the ␤-1,4-GalT V gene, however, four Sp1-, one N-Myc-, one USF-, one AP2-, and one AP4-binding sites were identified. We have demonstrated that the promoter region of the ␤-1,4-GalT V gene binds to Sp1 and that mutations in the Sp1-binding site at nucleotide positions Ϫ81/Ϫ69 significantly impair its promoter activity. Moreover, the treatment of A549 cells with mithramycin A, which inhibits the binding of Sp1 to its binding site (38,39), reduced the promoter activity. In contrast, the promoter activity increased dramatically by transfection of the Sp1 cDNA into A549 cells. These results indicate that Sp1 plays an essential role in regulating the promoter activation of the ␤-1,4-GalT V gene in cancer cells.
The Sp1 was originally identified as a transcription factor that binds to the GC box and activates the transcription of viral and cellular genes. To date, five proteins have been identified in the Sp family, designated Sp1, Sp2, Sp3, Sp4, and Sp5 (36,37,46). Sp2 does not bind to the classical Sp1 GC box but to a GT-rich element in the promoter region of the T-cell antigen receptor V␣ gene (47), whereas Sp3, Sp4, and Sp5 bind to the same consensus DNA site with affinities similar to that of Sp1 (46, 48 -50). Sp1 and Sp3 are ubiquitously expressed, whereas Sp4 is expressed predominantly in brain (48,51). Sp5, whose cDNA was isolated by screening the genes, is expressed differentially during mouse gastrulation and exhibits a remarkably dynamic expression pattern throughout early development (46). Although Sp1 and Sp3 can bind to the GC box (49,52), Sp3 was originally found to suppress Sp1-mediated activation by binding to the same site, thereby preventing Sp1 binding and activation (49). However, whether Sp3 acts as an activator or a suppressor of Sp1-mediated activation depends on the cellular conditions (53). The targeted disruption of the Sp1 gene in mouse results in growth retardation and the early death of embryos (54), indicating that Sp1 is essential for embryogenesis. Furthermore, Sp1 can activate the transcription of a number of viral and cellular genes, including structural proteins, metabolic enzymes, cell cycle regulators, transcription factors, growth factors, and surface receptors (36,37). Some of the genes activated by Sp1 are closely associated with tumor angiogenesis, invasion, and metastasis such as laminin-␥1 chain (55), matrix metalloproteinase-2 (56), thymidine phosphorylase (also known as platelet-derived endothelial cell growth factor) (57), protease-activated receptor-1 (58), and vascular endothelial growth factor (59,60). These reports suggest that the expression of Sp1 is essential for the malignant phenotypes seen in cancer cells.
Abnormal Sp1 expression and activation have been observed in human hepatocellular carcinoma (55), gastric carcinoma (61,62), pancreatic adenocarcinoma (60), and mouse epidermal tumors (63). Elevated Sp1 expression has also been shown to be correlated with malignancy and reduced survival of patients with gastric cancer (62). Transfection with an Sp1-decoy oligodeoxynucleotide, in which the consensus sequence for Sp1 binding is maintained, suppresses the growth and invasion of A549 cells and U251 human glioma cells, showing that an abnormal expression of Sp1 is associated with malignant phenotypes of cancer cells (64). In human hepatocellular carcinoma and gastric carcinoma, the expression of the ␤-1,4-GalT V gene is increased when compared with normal counterparts, 2 sug-4 T. Sato and K. Furukawa, manuscript in preparation.  Fig. 6, and the luciferase activity was determined 48 h after transfection. Transfection efficiency was adjusted by co-transfection with pRL-TK and parallel transfections with pGL3-SV40 and pGL3-Basic, used as positive and negative control, respectively. The promoter activity of pGL3-SV40 was taken as 100%. Three experiments were conducted, and the data are shown as the mean values with standard errors.
gesting that the expression of the ␤-1,4-GalT V gene is activated by abnormal Sp1 expression in these carcinomas. The activation of genes by Sp1 is also enhanced if multiple Sp1binding sites are present (65,66). In the case of the promoter region of the ␤-1,4-GalT V gene, probe C, corresponding to the promoter region Ϫ62/Ϫ36, contains three Sp1-binding sites, whereas the EMS assay showed that Sp1 cannot bind to this probe in experiments using nuclear extracts of SH-SY5Y cells. However, probe C has the ability to bind Sp1 because excess amounts of unlabeled probe C can fully compete with biotinylated Sp1-consensus oligonucleotides as probes. 2 The reason that Sp1 cannot bind to promoter region Ϫ62/Ϫ36 when nuclear extract of SH-SY5Y cells is used may be because of the binding of other transcription factors prior to Sp1. Under some pathological conditions or with the aid of some other transcription factors that are associated with Sp1, Sp1 may bind to promoter region Ϫ62/Ϫ36 and activate the promoter of the ␤-1,4-GalT V gene in cancer cells.
The increased amount of highly branched oligosaccharides is brought about by an elevation in the activity of the GlcNAcT V (11,13,14,67). The expression of the GlcNAcT V gene has been shown to be regulated by the Ets family of transcription factors, including Ets-1 and Ets-2 in cancer cells (68 -70). Our previous study showed that the expression level of the ␤-1,4-GalT V gene, but not other ␤-1,4-GalT genes, is highly correlated with that of the GlcNAcT V gene in several human cancer cell lines (20). Moreover, the 5Ј-flanking region of the human ␤-1,4-GalT V gene contains five Ets-1-binding sites. Therefore, it is considered that the ␤-1,4-GalT V gene is also regulated by Ets-1 in cancer cells. Our preliminary study showed that the promoter activity and expression of the ␤-1,4-GalT V gene increased 2-4-fold in the cells transfected with the ets-1 cDNA. 4 One of the Ets-1-binding sites at nucleotide positions Ϫ76/Ϫ67 was found to overlap the Sp1-binding site at nucleotide positions Ϫ81/Ϫ69, and Ets-1 does not bind to this site as revealed by the EMS assay using anti-Ets-1 antibody or the Ets-1-consensus  1-8). The arrowhead indicates a supershifted DNA-protein complex with anti-Sp1 antibody (lane 7). The asterisks indicate the nonspecific binding of probe B since these two bands were still observed at similar positions when competition experiments were conducted using excess amounts of unlabeled probe B (lanes 4 and 5). Three experiments were conducted, and representative results are shown.
oligonucleotides as a competitor. 4 However, four Ets-1-binding sites are present in the upstream and downstream positions of the promoter region of the ␤-1,4-GalT V gene, and they may interact with Ets-1 to regulate the gene expression. Because Sp1 and Ets-1 have been shown to be co-precipitated by anti-Sp1 antibody and cooperatively to regulate the gene expression of Fas ligand (71), the expression of the ␤-1,4-GalT V gene may be regulated by a similar mechanism. In the present study, however, we clearly demonstrated that the ␤-1,4-GalT V gene is regulated by Sp1 in cancer cells. It is of interest to investigate how these two transcription factors are involved in the expression of the ␤-1,4-GalT V gene in cancer cells.
The transcriptional regulation of mammalian ␤-1,4-GalT I has been well studied; the ␤-1,4-GalT I gene specifies two transcripts of 4.1 and 3.9 kb in somatic cells (72,73). The 5Ј-flanking region of the mouse ␤-1,4-GalT I gene is unusual in that three transcription start sites are present. In mouse somatic tissues, the start site for the 4.1-kb transcript is used predominantly, and the expression from this start site is controlled by a promoter containing multiple Sp1-binding sites (74,75). The only exception to this pattern is found in the mammary gland during lactation, where there is a switch to the preferential use of the start site for the 3.9-kb transcript, and expression from this start site is controlled by the promoter regulated by lactating mammary gland-restricted transcription factors (74,75). The most distal transcription start site is used exclusively during the late stages of spermatogenesis (76,77). However, which transcription start site in the promoter region of the ␤-1,4-GalT I gene is used in cancer cells is unknown. Because no significant change in the expression of the ␤-1,4-GalT I gene is observed upon malignant transformation of cells (17), the expression of the ␤-1,4-GalT I gene could remain unchanged upon malignant transformation. In the case of the ␤-1,4-GalT II gene, the expression level decreases dramatically upon malignant transformation (17), and it would be of interest to investigate the mechanism for the transcriptional regulation of the ␤-1,4-GalT II gene in cancer cells and how the ␤-1,4- The cells were harvested 24 h after the addition of mithramycin A and subjected to luciferase assay. The luciferase activity was normalized to the protein content of each sample, and the promoter activity of pGL(Ϫ116/ϩ170) was taken as 100%. Bars represent the mean values with standard errors of the results of three experiments. C, comparison of the expression levels of the ␤-1,4-GalT V gene between the mithramycin A-treated and -untreated A549 cells. RT-PCR was carried out with total RNA (1 g) from cells and oligonucleotide primers specific to the ␤-1,4-GalT V and G3PDH genes. The PCR products were visualized in a 2% agarose gel stained with ethidium bromide. The analysis was performed three times, and identical results were obtained each time.
GalTs II and V are involved in the changes in N-glycan biosynthesis characteristic of cancer cells.
In summary, this is the first report showing the transcriptional regulation of the ␤-1,4-GalT V gene by Sp1 in cancer cells. By regulating the expression of Sp1 in cancer cells, the expression of the ␤-1,4-GalT V gene can be reduced, and consequently the galactosylation pattern of the highly branched N-glycans characteristic of cancer cells can be modified, which may lead to the suppression of tumor growth and metastasis.