Dorsal Root Ganglia Neuron-specific Promoter Activity of the Rabbit β-Galactoside α1,2-Fucosyltransferase Gene*

The rabbit H-blood type α1,2-fucosyltransferase (RFT-I), gene and its biosynthetic products, H antigens (Fucα1,2Galβ), are abundantly expressed in a subset of dorsal root ganglia (DRG) neurons. To investigate the regulatory mechanisms for the RFT-I gene expression, we determined the genomic structure and promoter activity of this gene. PCR amplification of the 5′ cDNA end analysis revealed two transcriptional start sites, 498 and 82 nucleotides upstream of the translational initiation codon, the latter site yielding a major 3.1-kb transcript specifically expressed in DRG, as revealed by Northern blotting. Promoter analysis of the 5′-flanking region of the RFT-I gene using a luciferase gene reporter system demonstrated strong promoter activity in PC12 cells, which express the rat H-type α1,2-fucosyltransferase gene, and Neuro2a mouse neuroblastoma cells. Deletion analysis revealed the 704-base pair minimal promoter region flanking the translational initiation codon, for which two distinct promoter activities were detected and differentially used in PC12 and Neuro2a cells. The minimal promoter region contained a GC-rich domain (GC content 80%), in which a Sp1 binding sequence and a GSG-like nerve growth factor-responsive element were found, but lacked TATA- and CAAT-boxes. Promoter analysis with a primary culture of DRG neurons demonstrated that the minimal promoter region of the RFT-I gene was sufficient for the expression of a reporter gene in DRG neurons. We conclude that the TATA-less GC-rich minimal promoter region of the RFT-I gene controls DRG small neuron-specific expression of the RFT-I gene.

The rabbit H-blood type ␣1,2-fucosyltransferase (RFT-I), gene and its biosynthetic products, H antigens (Fuc␣1,2Gal␤), are abundantly expressed in a subset of dorsal root ganglia (DRG) neurons. To investigate the regulatory mechanisms for the RFT-I gene expression, we determined the genomic structure and promoter activity of this gene. PCR amplification of the 5 cDNA end analysis revealed two transcriptional start sites, 498 and 82 nucleotides upstream of the translational initiation codon, the latter site yielding a major 3.1-kb transcript specifically expressed in DRG, as revealed by Northern blotting. Promoter analysis of the 5-flanking region of the RFT-I gene using a luciferase gene reporter system demonstrated strong promoter activity in PC12 cells, which express the rat H-type ␣1,2-fucosyltransferase gene, and Neuro2a mouse neuroblastoma cells. Deletion analysis revealed the 704-base pair minimal promoter region flanking the translational initiation codon, for which two distinct promoter activities were detected and differentially used in PC12 and Neuro2a cells. The minimal promoter region contained a GC-rich domain (GC content 80%), in which a Sp1 binding sequence and a GSG-like nerve growth factor-responsive element were found, but lacked TATA-and CAAT-boxes. Promoter analysis with a primary culture of DRG neurons demonstrated that the minimal promoter region of the RFT-I gene was sufficient for the expression of a reporter gene in DRG neurons. We conclude that the TATA-less GC-rich minimal promoter region of the RFT-I gene controls DRG small neuron-specific expression of the RFT-I gene.
The H-blood type antigens (Fuc␣1,2Gal␤) are synthesized by GDP-L-fucose:␤-D-galactoside 2-␣-L-fucosyltransferase (␣1,2-FT) 1 (for a review, see Ref. 1). The expression of H determinants is strictly regulated temporally and spatially during vertebrate development (2)(3)(4). The H antigens are rarely detected in adult nervous tissues of human and other mammals but are present on a subset of neurons. Analyses with Ulex europaeus agglutinin 1 lectin, which binds to type 2 H (Fuc␣1,2Gal␤1,4GlcNAc) determinants, and anti-H antibodies revealed that the expression of H antigens was restricted to olfactory bulb and cochlear hair cells in rats (5,6) and to primary sensory neurons and their axons in human and other primates (7)(8)(9). Most of the H-positive axons of primary sensory neurons were unmyelinated and thought to be C-fibers that mediate nociceptive or thermoceptive inputs or both.
We have extended the analysis of the expression of H determinants in the mammalian nervous system using anti-fucosyl GM1 antibodies and U. europaeus agglutinin 1 lectin (10 -13). In the human and rabbit nervous systems, H antigens are abundantly expressed in dorsal root ganglia (DRG), which consist of several types of primary sensory neurons. The antifucosyl GM1 antibodies and U. europaeus agglutinin 1 lectin recognized a subpopulation of neurons in DRG and the dorsal horn of the spinal cord. The anti-fucosyl GM1 antibodies also bound to the satellite cells surrounding the fucosyl GM1-positive neurons (10,12). In addition, in rabbits, the anti-fucosyl GM1 antibodies bound to the axons and the myelin of the small myelinated fibers in the dorsal root and the large neurons in the ventral horn (13). In rabbit DRG, fucosyl GM1 is readily detectable immunohistochemically on embryonic day 25, followed by the appearance of U. europaeus agglutinin 1 lectinreactive antigens postnatally (12,13).
The expression of H antigens in human and rabbit DRG neurons seems to be under similar control, although little is known about the molecular basis of their regulated expression. To investigate the mechanisms underlying the regulation of the biosynthesis of H antigens in DRG neurons, we recently cloned three types of rabbit ␣1,2-FT gene, i.e. one H-type and two Se-type genes, as judged from the results of kinetic studies (14,15). Analysis of the expression of these genes revealed that all three ␣1,2-FT genes were expressed in DRG of late embryonic rabbits but that only the H-type ␣1,2-FT gene, the RFT-I gene, was expressed postnatally (16). In situ hybridization demonstrated that abundant RFT-I mRNA was present in adult rabbit DRG neurons of small diameter. We have shown that the RFT-I gene specifies the mRNAs of a major 3.  1 The abbreviations used are: ␣1,2-FT, GDP-L-fucose:␤-D-galactoside 2-␣-L-fucosyltransferase; RFT-I, rabbit H-blood type ␣1,2-FT; DRG, dorsal root ganglia; Se-type, secretor-type; PCR, polymerase chain reaction; RACE, rapid amplification of cDNA ends; PCR, polymerase chain reaction; CHO, Chinese hamster ovary; NGF, nerve growth factor; bp, base pair(s); kb, kilobase pair(s); CMV, cytomegalovirus; NGFI, nerve growth factor-induced gene. The nomenclature for gangliosides and glycolipids follows the system of Svennerholm (38). neuron-specific promoter activity. In this study, we determine the genomic structure of the RFT-I gene and the promoter activity of the 5Ј-flanking region using a primary culture of DRG neurons.

PCR Rapid Amplification of 5Ј and 3Ј cDNA Ends (RACE)-Poly(A)-
rich RNAs were extracted from adult rabbit DRG by the guanidinium isothiocyanate method and purified with Oligotex-dT30 (Takara-Shuzo, Japan). Amplification of the 5Ј-end of the RFT-I cDNA was performed essentially according to the procedure of Frothman et al. (17). cDNA was synthesized by reverse transcription (Superscript II; Life Technologies, Inc.) of 5 g of rabbit DRG poly(A)-rich RNA using primer H4B3, 5Ј-AAGCAAGAAGGCCAGACAGAGCTG-3Ј, which is complementary to nucleotides ϩ45 to ϩ22 (taking the translational initiation site as ϩ1) of the RFT-I gene. The excess primer and deoxynucleotide were removed by passage of the cDNA through a MicroSpin S-400 column (Amersham Pharmacia Biotech). The cDNA was A-tailed with 0.6 units of terminal deoxynucleotidyltransferase (Boehringer Mannheim), using 0.05 mM dATP. Two consecutive PCRs were performed with two nested sets of primers; for the first PCR, the forward primer was NotI-(dT) 18 (Amersham Pharmacia Biotech), and the reverse primer was H4B3; for the second PCR, the forward primer was as above but without the T-tail, 5Ј-AACTGGAAGAATTCGCGGCCGCAGGAA-3Ј, and the reverse primer was H4B4 (5Ј-AGAGCTGCCGGCGGCTCGGAGGCCA-CAT-3Ј; complementary to nucleotides ϩ28 to ϩ1). The cDNA was amplified for 40 cycles of a step program (95°C, 30 s; 55°C, 30 s; and 72°C, 30 s). The amplification products were subcloned into pBluescript II SK(ϩ) (Stratagene) and then sequenced.
Northern Blot Analysis-We previously reported two RFT-I mRNA transcripts, 4.2 and 3.1 kb, the latter of which is major and specific to DRG (16). To determine the transcription pattern of the DRG-specific mRNA isoform, we performed Northern blot analysis using three probes located in the 5Ј-flanking region relative to the translation initiation site. Probes A (Ϫ481 to Ϫ302), B (Ϫ294 to Ϫ55), and C (Ϫ75 to ϩ29) were labeled with [␣-32 P]dCTP (NEN Life Science Products), using a pair of synthetic oligonucleotide primers specific to each probe and Klenow fragment (Amersham Pharmacia Biotech). Total RNAs (3 g) from adult rabbit brain and DRG were fractionated on a denaturing formaldehyde-agarose gel (1.0%) and then transferred to a nylon membrane (Nytran; Schleicher & Schuell). Three membranes prepared from a gel were hybridized with probe A, B, or C.
Cell Culture and Promoter Activity Analysis-Neuro2a and Chinese hamster ovary (CHO) cells were seeded at 5 ϫ 10 4 cells/35-mm diameter dish in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum 24 h prior to transfection, respectively. PC12 cells were seeded in the same manner in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 5% horse serum. The luciferase plasmid (2 g) used as a reporter and the pSR-␤-gal plasmid (0.2 g) used as an internal control for transfection efficiency were transfected into the cells by means of Lipofectoamine (Life Technologies, Inc.). As negative and positive controls, the pPGBII and pBII-SV40 plasmids, respectively, were also transfected.
Plasmid pH4-SK3.6, which contains 5Ј-untranslated region, the entire coding region, and 3Ј-untranslated region with a poly(A) signal, was linearized with HindIII, combined with pTK-Hyg (CLONTECH), and then transfected into Neuro2a and PC12 cells. After culturing for 72 h, the cells were selected with 0.3 g/ml of hygromycin (Life Technologies, Inc.) and subsequently subcloned.
Cultures of DRG neurons and cerebellar granule cells were obtained as follows. DRG were excised from P14 rabbits, trypsinized, treated with 0.01% DNase I and then 10 g/ml of collagenase I (Sigma), and finally mechanically dissociated. The cell suspension was plated onto collagen S (Boehringer Mannheim)-coated coverslips in Dulbecco's modified Eagle's medium/F-12 medium supplemented with 10% fetal calf serum and 100 ng/ml of 2.5 S nerve growth factor (NGF; Sigma). Dorsal roots were also excised and treated as above to obtain fibroblasts and Schwann cells, but without DRG neurons. Cerebella excised from P5 rabbits were trypsinized and then mechanically dissociated. The cell suspension was plated onto poly-L-lysine (Sigma)-coated coverslips at 2-5 ϫ 10 5 cells/cm 2 in minimum essential medium with 26 mM potassium supplemented with 5% horse serum. After 48 h of plating, the luciferase plasmid (0.5 g) and the pSR-␤-gal plasmid (0.05 g) were transfected into the cells by means of FuGene 6 (Boehringer Mannheim). To verify the viability of cells and actual transfection of the plasmids into DRG neurons or cerebellar granule cells, pCMV-EGFP (0.5 g) was also transfected in another coverslip culture.
After 48 h of transfection, the cells were washed three times with PBS and then lysed with cell lysis buffer (PG␤-50; Toyo-ink, Japan). Luciferase activity was measured using a PicaGene Luciferase assay system (Toyo-ink) and a Luminescencer AB-2000 (Atto, Japan). Light activity measurements were performed in duplicate, averaged, and then normalized relative to ␤-galactosidase activity in order to correct for the transfection efficiency. ␤-Galactosidase activity was measured using a Luminescent ␤-galactosidase detection kit (CLONTECH).
Electrophoretic Mobility Shift Assay-Nuclear extracts from Neuro2a and PC12 cells with and without treatment with NGF (100 ng/ml) for 90 min were prepared by the method of Masuda et al. (18) Briefly, 5 ϫ 10 7 cells were collected by centrifugation and then sequentially resuspended with buffer A (10 mM HEPES, pH 7.6, 15 mM KCl, 2 mM MgCl 2 , 0.1 mM EDTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 10 g/ml of leupeptin), buffer A containing 0.2% Nonidet P-40, and buffer A containing 0.2 M sucrose and centrifugation at 800 ϫ g. The cell pellet was then resuspended with buffer D (50 mM HEPES, pH 7.9, 400 mM KCl, 10% glycerol, 0.1 mM EDTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 10 g/ml leupeptin) and centrifuged. Protein concentration of supernatant was determined by a Bio-Rad protein assay.
Binding assays were performed with a labeled probe in the presence of 2 g of poly(dI-dC)⅐poly(dI-dC) (Amersham Pharmacia Biotech) and 0.4 footprinting units of recombinant human Sp1 (Promega) or 0.5 g of nuclear extracts. Binding reactions were carried out for 30 min on ice in 10 mM HEPES-KOH (pH 7.8), 50 mM KCl, 5 mM MgCl 2 , 1 mM EDTA, 10% glycerol, 25 mM dithiothreitol, 3.5 mM phenylmethylsulfonyl fluoride, 5 mM sodium orthovanadate, 10 g/ml aprotinin, and 10 g/ml leupeptin. Competitor fragments of a 20-fold excess amount or anti-Egr-1 polyclonal antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were added where indicated. After incubation, the samples were loaded onto a 4% polyacrylamide gel in 0.5 ϫ TBE. The gel was run in the cold at 200 V and dried, and then the radioactivity was detected with a BAS 2000 image analyzer (Fuji Film, Japan).

Determination of the Transcriptional Start Site of the RFT-I
Gene-We previously showed that two RFT-I mRNA transcripts, 4.2 and 3.1 kb, were expressed in rabbit nervous tissues (16). The 4.2-kb transcript was broadly found in the central and peripheral nervous tissues examined but in low amounts, whereas the 3.1-kb transcript was abundantly and specifically observed in DRG neurons of small diameter. To isolate the 5Ј-end of the RFT-I cDNA, we performed RACE-PCR using poly(A) RNA from adult rabbit DRG, and found two transcriptional start sites, at positions Ϫ498 and Ϫ82 (Figs. 1 and 2). The RACE-PCR results revealed that there are three types of mRNA transcripts; one transcript (3.1-kb mRNA in Fig. 1) started at position Ϫ82 and contained no intron, the second one (3.3-kb mRNA in Fig. 1) started at position Ϫ98 and skipped the intron of nucleotides Ϫ265 to Ϫ3, yielding an additional 3.3-kb transcript, and the last one (4.2-kb mRNA in Fig. 1) also started at position Ϫ498 but without splicing out the 3.3-kb mRNA intron sequence. We also performed 3Ј RACE-PCR and found a poly(A) signal at position ϩ2579 without an intron.
To analyze the differential use of transcriptional start sites in rabbit nervous tissues, we next performed Northern analysis using three kinds of probes (probes A, B, and C in Fig. 1). Probe A, recognizing 3.3-and 4.2-kb mRNAs, gave weak signals at the corresponding bands in both DRG and brain RNA (Fig. 3). Probe B gave weak signals at 4.2-kb in DRG and brain RNA. Probe C hybridized to a 3.1-kb transcript in DRG RNA strongly enough to be detected on short exposure. These results showed that the 3.1-kb mRNA of the RFT-I gene was abundantly and specifically expressed in DRG and that the 3.3-and 4.2-kb mRNAs were broadly found in DRG and other nervous tissues but in low amounts.
Promoter Activity Analysis-To determine the RFT-I gene promoter activity, we used three types of cells: PC12 rat pheochromocytoma cells expressing rat H-blood type ␣1,2-FT, a counterpart of RFT-I; Neuro2a mouse neuroblastoma cells originating from the neural crest but not expressing ␣1,2-FT; and nonneuronal CHO cells. A series of reporter plasmids containing progressive deletions of the 5Ј-flanking region of the RFT-I gene fused to the promoterless luciferase gene were mixed with an internal control plasmid, pSR-␤-Gal, carrying a ␤-galacto-sidase gene under the control of the SR␣ promoter, and then transfected into cells. The luciferase activity due to each luciferase reporter plasmid was normalized as to the ␤-galactosidase activity. The promoter activity was calculated relative to the SV40 promoter activity taken as 100%.
pBH4-BP3.0, containing the 3.0-kb 5Ј-flanking sequence of the RFT-I gene (Ϫ2970 to Ϫ4), showed high levels of promoter activity when expressed in PC12 and Neuro2a cells but not in CHO cells (Fig. 4). The pBH4-SP0.7 reporter plasmid also showed high promoter activity in PC12 and Neuro2a cells, although the activity was lower than that of pBH4-BP3.0. The 0.7-kb 5Ј-flanking region was divided into three portions, and each fragment was ligated into a pPGBII plasmid. The pBH4-SSm0.2 plasmid, containing the 0.2-kb sequence (Ϫ707 to Ϫ539), showed a substantial level of promoter activity when expressed in PC12 but not in Neuro2a cells, whereas the pBH4-NP0.3 plasmid, containing the 0.3-kb sequence (Ϫ294 to Ϫ4), showed a level of promoter activity comparable with pBH4-SP0.7 in Neuro2a cells but not in PC12 cells. pBH4-Sma0.3 showed little promoter activity in both types of cells. These results suggested that the promoter activity was differentially regulated in PC12 and Neuro2a cells.
To determine the effect of the differential promoter activity on the transcriptional pattern, plasmid pH4-SK3.6, containing the 5Ј-and 3Ј-flanking regions and the entire coding region of the RFT-I gene, was stably transfected into Neuro2a and PC12 cells. The transcriptional pattern was analyzed by reverse transcription-PCR using three pairs of primers; the H4A5 and H4B4 primers could discriminate the presence of a 3.3-kb mRNA, the H4A5 and H4B5 (5Ј-GGAGGAGGTCTGGGAAAA-GAGGCG-3Ј) primers could detect a 4.2-kb mRNA, and the H4A4 and H4B4 primers could detect 3.1-and 4.2-kb mRNAs but could not discriminate them. All pairs of primers gave positive bands when RNA from Neuro2a cell-derived stable transfectants with pH4-SK3.6 was analyzed (data not shown), indicating the presence of 3.3-and 4.2-kb transcripts of the RFT-I gene in these cells, although this result did not exclude the possibility of the presence of a 3.1-kb transcript. On the contrary, a positive band was amplified only by the H4A4 and H4B4 primers in PC12 cell-derived stable transfectants with pH4-SK3.6 (data not shown), showing the presence of a 3.1-kb transcript of the RFT-I gene.
Effects of NGF and Mutations on Promoter Activity-A data base search for possible binding of transcription factors revealed several Sp1 binding sites and a N-Myc binding site in the region of nucleotides Ϫ707 to Ϫ4. Among possible Sp1 binding sites, one (nucleotides Ϫ650 to Ϫ638) showed the highest homology to the consensus sequence and overlapped a GSG (GCGGGGCG)-like motif (nucleotides Ϫ645 to Ϫ637). To determine whether or not these elements were functional, we constructed plasmids pBH4-SacB8 (containing nucleotides Ϫ707 to Ϫ626) and pBH4-SacB8m (the same region with mutations) and transfected them into PC12 cells. PC12 cells transfected with pBH4-SacB8 and treated with 100 ng/ml of NGF for 48 h showed 2-fold higher luciferase activity than those without NGF (Fig. 5A). The mutations in the Sp1 and GSG overlapping domain decreased the luciferase activity and abolished the effect of NGF treatment.
We then examined whether or not a possible N-Myc binding site (nucleotides Ϫ244 to Ϫ233) found within the pBH4-NP0.3 construct could be demonstrated functionally. We constructed plasmids pBH4-A6Pst (containing nucleotides Ϫ264 to Ϫ4) and pBH4-A6mPst (the same region with mutations), and transfected them into Neuro2a cells. The pBH4-A6Pst and pBH4-A6mPst constructs showed comparable promoter activities when transfected into Neuro2a cells (Fig. 5B), suggesting that the N-Myc was not a major factor for the activity.
Electrophoretic Mobility Shift Assay-To confirm the actual binding of Sp1 to the putative Sp1 binding site between nucleotides Ϫ650 and Ϫ638, we performed an electrophoretic mobility shift assay. In the mobility shift experiments involving the DNA fragments of nucleotides Ϫ707 to Ϫ626, recombinant Sp1 bound to the DNA fragments (Fig. 6A, lane 2). The shifted band completely disappeared in the presence of the nonlabeled spe- To determine the transcriptional factors that regulate the specific promoter activity of this region in PC12 but not in Neuro2a cells, we next performed the assay using nuclear protein extracts of Neuro2a and PC12 cells with and without NGF treatment. When nuclear protein extracts of Neuro2a and PC12 cells were used, the labeled DNA fragment of nucleotides Ϫ655 to Ϫ626 (A8B8) appeared as two shifted bands with apparently the same electromobility (Fig. 6B, lanes 2-4, bands  a and c). An additional band (band b) was observed when using a nuclear extract of PC12 cells treated with NGF for 90 min, and this shifted band disappeared upon the addition of the anti-Egr-1 polyclonal antibodies (Fig. 6B, lane 7). All shifted bands completely disappeared in the presence of the nonlabeled specific competitor (A8B8, lane 5) but were not abolished by the mutant competitor (A8mB8m, the same region of the DNA fragments with mutations; lane 6). No shifted band was found when using the labeled DNA fragment of nucleotide Ϫ655 to Ϫ626 with mutations (A8mB8m) and nuclear protein extracts of Neuro2a and PC12 cells.
Promoter Activity Analysis of Cultures DRG Neurons-DRG neurons were cultured and transfected with reporter plasmids containing several lengths of the 5Ј-flanking region of the RFT-I gene. The pBH4-SP0.7 construct showed the highest level of promoter activity when expressed in DRG neurons (Fig.  7), which was consistent with the results for PC12 and Neuro2a cells. However, the pBH4-SSP1.3, pBH4-SSm0.2, and pBH4-NP0.3 constructs showed relatively low promoter activities. The DRG culture contained neurons, Schwann cells, and fibroblasts, and all of them were transfected with plasmids and expressed the reporter gene, as revealed by the transfection of pCMV-EGFP and observation by fluorescence microscopy (data not shown). To determine the contributions of Schwann cells and fibroblasts to the promoter activity, we then used a dorsal root culture, which contained Schwann cells and fibroblasts but no neurons. All of the constructs showed a relatively low level of promoter activity (Fig. 7), suggesting that the high promoter activity of the pBH4-SP0.7 construct could be due to the expression of the reporter gene in DRG neurons. Next, we determined whether or not the promoter activity of the pBH4-SP0.7 construct was specific to DRG neurons. The pBH4-SP0.7 and other constructs showed lower promoter activity when expressed in cerebellar granule cells, another type of neuron, than in DRG neurons (Fig. 7). DISCUSSION We previously showed that the expression of RFT-I is strictly regulated spatially and temporary in the rabbit nervous system and abundant in adult DRG neurons of small diameter (16). In this study, we determined the genomic organization and promoter activity that regulates the DRG neuron-specific expression of the RFT-I gene. Our results demonstrated that the RFT-I gene used two transcriptional start sites yielding three types of mRNA and that the minimal promoter region flanking the translational initiation codon of the RFT-I gene was sufficient for DRG neuron-specific expression of the gene. DRG neuron-specific promoter activity has not been detected so far, and analysis of the proteins binding to the minimal promoter region will facilitate understanding of the differentiation of DRG neurons.
The RFT-I gene specified three types of mRNA; a 3.1-kb transcript starting at position Ϫ82 was abundantly and exclusively expressed in DRG neurons, and 3.3-and 4.2-kb transcripts both starting at position Ϫ498 were broadly found in the rabbit nervous system but in low amounts. Recently, evidence of multiple transcriptional start sites in such glycosyltransferase genes as the rat and human ␣2,6-sialyltransferase genes (19 -21), murine ␤1,4-galactosyltransferase gene (22), human ␤1,6-N-acetylglucosaminyltransferase V gene (23), human ␤1,4-N-acetylglucosaminyltransferase gene (24), and human ␣1,3and ␣1,2-fucosyltransferase genes (25,26) has been accumulated. Similar to our results for the RFT-I gene, the murine ␤1,4-galactosyltransferase gene uses at least two transcriptional start sites yielding 4.1-and 3.9-kb mRNAs; the former is ubiquitously expressed in all tissues, and the latter is abundantly expressed in lactating mammary glands (22). The differential use of multiple transcriptional start sites and associated promoters of the glycosyltransferase gene could regulate the tissue-and stage-specific expression of glycosyltransferase genes and subsequent glycosylation patterns.
The distribution of H antigens in the nervous system is quite similar in human and rabbit; in both species, small neurons of DRG are positive for both U. europaeus agglutinin 1 lectin staining and anti-fucosyl GM1 antibody immunostaining (8,11,12). The human H-blood type ␣1,2-FT FUT1 gene is known to control the H antigens on erythrocytes, whereas the expression of the FUT1 gene in human nervous tissues remains undetermined. The H antigens on DRG small neurons in humans seem to be synthesized by FUT1, because FUT1 preferentially uses type 2 precursor glycochain as an acceptor to yield type 2 H, which the U. europaeus agglutinin 1 lectin recognizes, rather than FUT2 (27)(28)(29). The exon-intron organization and splicing patterns of FUT1, which are different from those of the RFT-I gene, have been determined in human bone marrow cells, HEL human erythroleukemic cells, and MCAS human ovarian cancer cells (26). It would be interesting to analyze the transcriptional pattern of FUT1 in human DRG neurons and to compare it with that of the RFT-I gene.
In the present study, we detected the promoter activity in the 5Ј-flanking region (nucleotides Ϫ707 to Ϫ4) of the RFT-I gene when reporter plasmids were transfected into PC12 and Neuro2a cells. PC12 (rat pheochromocytoma) and Neuro2a (murine neuroblastoma) cells share a common developmental origin, the neural crest, with DRG neurons. Although the species of PC12 and Neuro2a cells and rabbit DRG neurons are all different, common mechanisms and highly conserved factors might function to control the expression of H-type ␣1,2-FT or other neural crest-originating neuron-specific genes. Indeed, FIG. 3. Northern blotting of RFT-I. The differential expression of three types of RFT-I mRNA was analyzed using three probes. Probe A hybridized to 4.2-and 3.3-kb transcripts in adult rabbit DRG and brain, probe B hybridized to a 4.2-kb transcript in both tissues, and probe C hybridized to a 3.1-kb transcript in DRG only. The results for probe C were obtained with a shorter exposure (overnight) than for probe A or B (three overnight exposures). the pBH4-SP0.7 construct containing the region of nucleotides Ϫ707 to Ϫ4 showed high promoter activity in PC12 and Neuro2a cells. In this region, at least two promoter domains are thought to exist; the pBH4-SSm0.2 (nucleotides Ϫ707 to Ϫ539) construct showed the promoter activity when expressed in PC12 cells, and the pBH4-NP0.3 (nucleotides Ϫ294 to Ϫ4) construct showed the promoter activity when expressed in Neuro2a cells.
The region of nucleotides Ϫ707 to Ϫ539 included a GC-rich domain of 80% GC content, in which several Sp1 binding sites were found in a data base search for transcription factors. One of the Sp1 binding sites showing the highest homology with the consensus sequence overlapped a GSG-like (8/9 consensus) NGF-responsive element (30,31). Promoter analysis involving pBH4-SSm0.2 with and without mutations demonstrated the reduction of the promoter activity and the abolition of NGF responsiveness upon the insertion of mutations, suggesting that these elements are functional. An electrophoretic mobility shift assay involving recombinant Sp1 and the labeled DNA fragment (nucleotides Ϫ707 to Ϫ626) with and without muta- Each construct and the pSR-␤-gal plasmid were transfected into PC12, Neuro2a, or CHO cells. Luciferase activity was normalized as to ␤-galactosidase activity, and the relative promoter activity was calculated relative to SV40 early promoter activity taken as 100%. Each experiment was performed in duplicate, and the results are the averages of four experiments.

FIG. 5. Effects of NGF and mutations on the promoter activity.
A, PC12 cells were transfected with the pBH4-SSm0.2, pBH4-SacB8, and pBH4-SacB8m constructs and then cultured with and without 100 ng/ml of NGF for 48 h. The wild type (pBH4-SSm0.2 and pBH4-SacB8) and mutational (pBH4-SacB8m) sequences of the Sp1 binding site and the overlapping GSG-like (8/9 consensus) elements are shown. B, Neuro2a cells were transfected with the pBH4-A6Pst and pBH4-A6mPst constructs, and then cultured for 48 h. The wild type (pBH4-A6Pst) and mutational (pBH4-A6mPst) sequences of the inverted N-Myc binding site are shown. Luciferase activity was normalized relative to ␤-galactosidase activity, and the relative promoter activities were calculated with SV40 early promoter activity taken as 100%. Each experiment was performed in duplicate, and the data are the means Ϯ S.E. for three experiments.
tions confirmed the actual binding of Sp1 to the putative Sp1 binding site of nucleotides Ϫ650 to Ϫ638. The partial disappearance of signals upon the addition of an excess amount of unlabeled DNA fragments with mutations could be due to the weak binding of Sp1 to another Sp1 binding site (nucleotides Ϫ658 to Ϫ646) exhibiting lower homology with the consensus sequence.
The GSG element is a consensus binding motif recognized by members of the early response gene family, such as NGFI-A (32)/Egr-1 (33), Krox-20 (34), Wilms' tumor gene product (30), and NGFI-C (31). Among them, NGFI-A/Egr-1 and NGFI-C are rapidly and temporally induced in PC12 cells by NGF stimu-lation. In the present study, an electrophoretic mobility shift assay involving a nuclear extract of PC12 cells stimulated with NGF and anti-Egr-1 antibodies revealed that NGFI-A/Egr-1, at least, actually binds to the Sp1 and GSG-like overlapping element. It is still possible that other NGF-responsive transcriptional factors weakly bind to this element and enhance the promoter activity of this region. Sp1 is a ubiquitous transcriptional factor that could recruit transcription initiation complexes to the initiation site in TATA-less promoters (35) and may require other co-factors for the tissue-specific expression of genes. As for the RFT-I gene promoter, NGFI-A/Egr-1 and possibly other members of the NGF-inducible gene family may be responsible for the PC12 cell-specific promoter activity in the region of nucleotides Ϫ707 to Ϫ539.
The promoter region of nucleotides Ϫ707 to Ϫ539 seemed to control the expression of a 3.1-kb DRG-specific transcript, because PC12 cell-derived stable transfectants with pH4-SK3.6 expressed a transcript starting at position Ϫ82. On the other hand, Neuro2a cell-derived stable transfectants with pH4-SK3.6 expressed mRNAs starting at position Ϫ498 corresponding to 3.3-and 4.2-kb transcripts, although reverse transcription-PCR analysis of these transfectants did not exclude the presence of a transcript starting at position Ϫ82. The promoter region of nucleotides Ϫ294 to Ϫ4 might control the expression of the 3.3-and 4.2-kb mRNAs of the RFT-I gene. A N-Myc binding site found in this region was not thought to be functional, because the pBH4-A6Pst and pBH4-A6mPst (containing mutations at the N-Myc binding site) constructs showed comparable promoter activity when transfected into Neuro2a cells.
High promoter activity was detected in the region of nucleotides Ϫ707 to Ϫ4, when the pBH4-SP0.7 construct was transfected into a primary culture of DRG neurons. This activity was thought to be due to the expression of the reporter gene in DRG neurons, because the same construct showed far lower promoter activity in a primary dorsal root culture. The difference between DRG and dorsal root cultures was the presence or absence of neurons. Moreover, the pBH4-SP0.7 construct showed low promoter activity in a primary cerebellum culture, which contained another type of neuron, cerebellar granule cells. This observation supports the notion that the region of nucleotides Ϫ707 to Ϫ4 of the RFT-I gene involves the pro-FIG. 6. Electrophoretic mobility shift assay. A, the labeled probes of the DNA fragments from Ϫ707 to Ϫ626 without (SacB8, lanes 1-4) or with (SacB8m, lanes 5-8) mutations in the Sp1 binding site were incubated with 0.4 footprinting units of recombinant human Sp1 either alone (lanes 2 and 6) or with the nonlabeled wild-type (lanes 3 and 7) and mutant (lanes 4 and 8) competitors in a 20-fold molar excess. The arrow indicates the binding of recombinant Sp1 to the labeled SacB8 probe, which completely disappeared in the presence of the nonlabeled SacB8 DNA fragment and partially disappeared in the presence of the nonlabeled SacB8m DNA fragment. B, the labeled probes of the DNA fragments from Ϫ655 to Ϫ626 without (A8B8, lanes 1-7) or with (A8mB8m, lanes 8 -11) mutations in the Sp1 and GSG overlapping domain were incubated with 0.5 g of nuclear extract from Neuro2a (lanes 2 and 9) and PC12 cells with (lanes 4 -7 and 11) and without (lanes 3 and 10) NGF treatment. The nonlabeled wild-type (lane 5) and mutant (lane 6) competitors in a 20-fold molar excess or anti-Egr-1 polyclonal antibodies (lane 7) were also added.
FIG. 7. RFT-I gene promoter activity in DRG neurons. Primary cultures of DRG, DR, and cerebellar tissue were transfected with reporter plasmids, pBH4-SSP1.3, pBH4-SP0.7, pBH4-SSm0.2, and pBH4-NP0.3 and then cultured for 48 h. The viability of cells and actual transfection of the reporter plasmids were verified by transfection of the pCMV-EGFP construct to another primary culture preparation and observation by fluorescence microscopy. Luciferase activity was normalized relative to ␤-galactosidase activity, and the relative promoter activity was calculated with SV40 early promoter activity taken as 100%. Each experiment was performed in duplicate, and the results are the averages of three experiments. moter sequences sufficient for the DRG neuron-specific gene expression.
The pBH4-SSm0.2 and pBH4-NP0.3 constructs showed relatively low promoter activity as compared with the pBH4-SP0.7 construct, when transfected into a primary culture of DRG neurons. DRG neurons are known to express trk (neurotrophin receptor family genes) and to be dependent on neurotrophin-mediated signaling (36,37). Most of the small neurons of DRG bearing H antigens on their surface express TrkA, which binds to NGF, as in the case of PC12 cells. Despite the fact that the DRG neurons were cultured with 100 ng/ml NGF, the pBH4-SSm0.2 construct showed a low level of promoter activity in DRG neurons comparable with that in dorsal roots and cerebellar granule cells. Some other factors binding to somewhere within the region of nucleotides Ϫ538 to Ϫ4 might be necessary for sufficient promoter activity. It is still possible that the Sp1 binding site and the overlapping GSG-like element could act in a stage-specific manner during the late embryonic period when DRG neurons are critically dependent on NGF for their survival.