Genomic Cloning and Expression of Three Murine UDP-galactose: β-N-Acetylglucosamine β1,3-Galactosyltransferase Genes*

Based on the detection of expressed sequence tags that are similar to known galactosyltransferase sequences, we have isolated three novel UDP-galactose:β-N-acetylglucosamine β1,3-galactosyltransferase (β3GalT) genes from a mouse genomic library. The three genes, named β3GalT-I, -II, and -III, encode type II transmembrane proteins of 326, 422, and 331 amino acids, respectively. The three proteins constitute a distinct subfamily as they do not share any sequence identity with other eucaryotic galactosyltransferases. Also, the entire protein-coding region of the three β3GalT genes was contained in a single exon, which contrasts with the genomic organization of the β1,4- and α1,3-galactosyltransferase genes. The three β3GalT genes were mainly expressed in brain tissue. The expression of the full-length murine genes as recombinant baculoviruses in insect cells revealed that the β3GalT enzymes share the same acceptor specificity for β-linked GlcNAc, although they differ in their K m for this acceptor and the donor UDP-Gal. The identification of β3GalT genes emphasizes the structural diversity present in the galactosyltransferase gene family.

The structural diversity of glycoconjugates results from the combined action of glycosidases and glycosyltransferases that are differentially expressed throughout tissues. Until recently, each carbohydrate linkage was believed to be catalyzed by one single glycosyltransferase. Deduced from this assumption, it was estimated that ϳ250 genes were required for shaping the glycoconjugate repertoire of vertebrates (1). This count is constantly corrected to a higher mark as many glycosyltransferase activities are found to be encoded by multiple related genes. This apparent redundancy likely reflects a selective preference of glycosyltransferase isozymes for related although distinct acceptor structures, as observed for ␣1,3-fucosyltransferase (2) and GalNAc ␣2,6-sialyltransferase (3)(4)(5) isozymes.
Various strategies have been applied to identify genes related to known glycosyltransferase genes. The use of low stringency hybridization was only successful to detect very closely related genes, as was the case for ␣1,3-fucosyltransferases (6,7). Often, the comparison of related glycosyltransferase genes unveils conserved regions that correspond to residues essential for catalytic activity. Such comparisons performed on sialyltransferase genes led to the detection of a stretch of conserved sequences called the sialyl motif (8). By designing degenerate primers according to the sialyl motif, it has been possible to isolate additional sialyltransferase genes by polymerase chain reaction (PCR) 1 (5,8,9). The recent availability of "single-pass" cDNA sequences, or expressed sequenced tags (ESTs), represents a tremendous amount of information that can be analyzed to retrieve sequences sharing limited but significant sequence similarity. This approach is particularly well suited to detect similarity at the level of the protein sequence, which would be difficult to isolate by DNA-based retrieval techniques such as cross-hybridization and PCR.
We have applied the EST screening procedure to identify genes encoding ␤1,3-galactosyltransferases, which refers to an activity yet uncharacterized at the genetic level. To this end, we have probed the EST division of the GenBank™/EMBL Data Bank with the sequence of a ␤-galactosyltransferase protein that was believed to encode a ␤1,3-galactosyltransferase (␤3GalT) (10). This procedure led to the identification of novel open reading frames (ORFs) that were similar to the query ␤-galactosyltransferase, but distinct from the previously described ␣1,3-galactosyltransferase (11,12) and ␤1,4-galactosyltransferase (13) genes. We have used probes designed after these ORFs to isolate the corresponding genes from a mouse genomic library. We have cloned three genes coding for proteins of 326, 422, and 331 amino acids. The three proteins are type II transmembrane proteins containing a single transmembrane domain of 17-19 amino acids. The sequence identity between the three proteins was the highest in the luminal domain, where it ranged from 35 to 51%. Several conserved motifs were detected in the catalytic domain. None of these motifs were present in ␣1,3and ␤1,4-galactosyltransferases, although one motif was identified in some bacterial galactosyltransferases. The expression of the three murine genes as recombinant baculoviruses in Sf 9 insect cells confirmed them as encoding UDP-galactose:␤-N-acetylglucosamine ␤1,3-galactosyltransferase enzymes. The genomic organization of the murine ␤3GalT-I, -II, and -III genes differed from that of ␤1,4galactosyltransferase (14) and ␣1,3-galactosyltransferase (15) genes. The entire protein-coding region of the murine ␤3GalT genes was contained in a single exon. Northern blot analysis showed that the three genes were mainly expressed in brain tissue and also at lower levels in other tissues.

EXPERIMENTAL PROCEDURES
Materials-The mouse 129/SvJ genomic library, Duralose-UV nitrocellulose membranes, and the Bluescript SKII plasmid were purchased from Stratagene. The Vent polymerase, restriction enzymes, and DNAmodifying enzymes were obtained from New England Biolabs Inc. The AmpliTaq FS Dye Terminator Cycle Sequencing kit, the CentriSep columns, and the ABI 310 Genetic Analyzer were from Perkin-Elmer/ Applied Biosystems. The baculovirus expression system Bac-to-Bac and Grace's medium were obtained from Life Technologies, Inc. Carbohydrate acceptors and donors were purchased from Sigma. Asialo-ovine submaxillary mucin was a gift from Dr. Robert L. Hill (16). Gal␤1,4GlcNAc-␤-pNP was generated enzymatically by the action of recombinant human ␤1,4-galactosyltransferase (17), and the Gal␤1,3GlcNAc-␤-pNP standard was purchased from Sigma. The HPLC system consisted of a Jour Research X-ACT degasser, two Waters 510 pumps, a Waters 717 autoinjector, a Jasco C0965 Column Heater (30°C), and a Waters 490 programmable multi-wavelength detector (214, 262, and 350 nm). The system was controlled using Waters Millennium (Version 2.15.2). UDP-[ 14 C]Gal and Hybond-N nylon membranes were from Amersham Corp. Sep-Pak C 18 cartridges were obtained from Waters, and GFA glass-microfiber filters were from Whatman. [ 32 P]CTP was purchased from Hartmann Analytics (Braunschweig, Germany).
Genomic Cloning of ␤3GalT Genes-The mouse genomic library was plated onto Luria-Bertani agar, and plaques were lifted onto Duralose-UV filters. The filters were hybridized to [ 32 P]CTP-labeled DNA probes designed after the EST sequences that were similar to a human galactosyltransferase gene (10). The genomic inserts from the phage clones were subcloned as NotI-NotI fragments into the NotI site of the pBluescript SKII vector, and regions that hybridized to the EST fragments were further subcloned into pBluescript and sequenced.
Sequencing-Sequencing was done using the AmpliTaq FS Dye Terminator Cycle Sequencing kit with the vectors T3 and T7 matching primers flanking the cloned inserts in pBluescript. Sequenced products were purified with CentriSep columns and run on an ABI 310 Genetic Analyzer. Primer walking was used afterward for sequence determination throughout the complete coding region of every ␤3GalT gene. The sequences of the coding and complementary strands were determined and edited using the Sequencher program (Gene Codes, Ann Arbor, MI).
Cloning of Recombinant Baculoviruses and Expression in Sf 9 Cells-Recombinant baculoviruses were generated using the transposon-mediated insertion system developed by Luckow et al. (18) as purchased from Life Technologies, Inc. The mouse ␤3GalT-I, -II, and -III genes were amplified by PCR from the genomic clones using 5Ј-primers containing an EcoRI site (␤3GalT-I) or a BamHI site (␤3GalT-II and -III) just upstream of the ATG codon. The 3Ј-primers for the three ␤3GalT genes contained a XbaI site following the stop codon. The conditions for PCR were 20 cycles at 95°C for 45 s, 55°C for 30 s, and 72°C for 60 s using the Vent polymerase and 50 ng of plasmid DNA as template. The resulting PCR fragments were subcloned into the baculovirus donor vector FastBac1 opened at the EcoRI and XbaI sites (␤3GalT-I) or at the BamHI and XbaI sites (␤3GalT-II and -III). Following the instructions of the manufacturer, the FastBac1-␤3GalT plasmids were transformed into DH10Bac cells for transposition into the bacmid. The integration of the ␤3GalT genes in the bacmids was confirmed by PCR using the 5Ј-primers specific for each ␤3GalT-I gene and the m13r primer (5Ј-AACAGCTATGACCATGATTACG-3Ј), which binds to a bacmid region outside of the transposition element. Sf 9 cells (10 6 ) were transfected with 1 g of recombinant bacmid DNA using Cellfectin (Life Technologies, Inc.) and incubated for 3 days at 27°C. The culture medium was saved as the primary baculovirus stock. The primary stocks were amplified three times to reach titers of ϳ10 9 plaque-forming units/ml. Sf 9 FIG. 1. Schematic alignment of EST sequences with the murine ␤3GalT genes. The EST fragments similar to the ␤3GalT genes were aligned with the corresponding ␤3GalT genes. The protein-coding regions of the three murine ␤3GalT genes are given as black rectangles. National Center for Biotechnology Information sequence identification numbers are provided to the left of the lines representing the EST fragments. The vertical dotted lines indicate 500-bp increments.

TABLE I
Cloning of mouse genomic fragments related to a human ␤-galactosyltransferase gene ESTs that were similar to a human ␤-galactosyltransferase gene (␤xGalT) were retrieved using a tblastn search of the ␣EST division of the GenBank™/EMBL Data Bank. The EST sequences were aligned, and those that overlapped were assembled to form larger ORFs. ESTs representative of each ORF as well as the human ␤xGalT gene were used as DNA probes to screen a mouse genomic DNA library. Seven mouse genomic fragments were isolated and grouped according to their specificity for the DNA probes. ORF2 and ORF3 were found to be different parts of the same gene, as genomic fragments 5 and 14 hybridized to both DNA probes H70574 and R15977. cells were infected at a multiplicity of 10 and further incubated at 27°C for 72 h.
Galactosyltransferase Activity Assays-Baculovirus-infected Sf 9 cells were washed with phosphate-buffered saline and lyzed in 2% Triton X-100 for 15 min on ice. Nuclei were removed from the lysates by centrifugation at 500 ϫ g. 10 l of the lysates were assayed for 30 min at 37°C in 50-l reactions of 50 mM cacodylate buffer, pH 6.6, 10 mM MnCl 2 , 0.5 mM UDP-Gal, 1% Triton X-100, 10% Me 2 SO, and varying acceptors at different concentrations (see Table II). 10 5 cpm of UDP-[ 14 C]Gal (164 pmol) were added to standard assays, whereas 2.5 ϫ 10 5 cpm of UDP-[ 14 C]Gal (410 pmol) were added when kinetic parameters were determined. The reactions were stopped by dilution with 500 l of H 2 O and applied onto a Sep-Pak C 18 cartridge. The cartridge was washed with 15 ml of H 2 O and eluted with 5 ml of methanol. The samples containing asialo-ovine submaxillary mucin as acceptor were stopped by adding 1 ml of cold 15% trichloroacetic acid and 5% phosphotungstic acid solution. The precipitates were collected on GFA glassmicrofiber filters, which were washed with 5 ml of ice-cold ethanol and dried at 80°C for 30 min. The amount of [ 14 C]Gal in the methanol eluates and on the GFA filters was measured in a liquid scintillation counter (Rackbeta, Pharmacia Biotech Inc.).
Analysis of Linkage Specificity of ␤3GalT Enzymes-Sep-Pak C 18 eluates of ␤3GalT-I, -II, and -III assays using unlabeled UDP-Gal with GlcNAc-␤-pNP as acceptor were dried by centrifugal evaporation and redissolved in 100 l of H 2 O. A 20-l aliquot of this material was diluted with acetonitrile to 100 l and subjected to normal phase (NP) HPLC. NP-HPLC was carried out using the 50 mM ammonium formate pH 4.4 buffer system on a GlycoSep-N column (Oxford GlycoSciences, Abingdon, United Kingdom) over 180 min as described previously (19). The pNP-labeled disaccharides were isolated from the sample mixture by comparison to standard UDP, UDP-Gal, GlcNAc-␤-pNP, Gal␤1, 3GlcNAc-␤-pNP, and Gal␤1,4GlcNAc-␤-pNP. The fractions containing the disaccharides were pooled and analyzed by reverse phase HPLC (20) to separate Gal␤1,3GlcNAc-␤-pNP and Gal␤1,4GlcNAc-␤-pNP. Briefly, the first step of the gradient applied to a GlycoSep-R column (Oxford GlycoSciences) consisted of isocratic elution with 75% Solvent A (25% 50 mM ammonium formate, pH 4.4, in acetonitrile) and 25% Solvent B (50 mM ammonium formate, pH 4.4) over 10 min at a flow rate of 1.0 ml/min. This was followed by a linear gradient of 75 to 25% Solvent A over 20 min, followed by a 5-min isocratic step under these conditions. The column was re-equilibrated in the initial solvent composition 20 min prior to the next injection.
Exoglycosidase Digestion Conditions-Streptococcus pneumoniae ␤-galactosidase (Oxford GlycoSciences) digestions were performed at 37°C for 16 -24 h in 100 mM sodium acetate, pH 6.0, at 27 milliunits/ml. At this concentration, the enzyme cleaves only the Gal␤1,4HexNAc linkage (21). An aliquot of the aqueous sample was mixed with acetonitrile at a ratio of 20:80 and applied to the NP-HPLC column.
RNA Preparation and Northern Blotting-Total RNA from 8-weekold ICR mouse tissues was isolated with guanidinium isothiocyanate, followed by centrifugation on cesium chloride cushions (22). 5 g of total RNA from each tissue were separated on formaldehyde-agarose gels. RNA was either stained in ethidium bromide (0.5 g/ml in 0.1 M ammonium acetate) or transferred to Hybond-N membranes by capillary elution. Filters were hybridized overnight at 42°C with [ 32 P]CTP-labeled probes corresponding to the full-length mouse ␤3GalT-I, -II, and -III genes. The filters were washed with 0.1 ϫ SSC and 0.1% SDS up to 60°C and exposed for 3 days at Ϫ70°C using intensifying screens.

Identification and Isolation of Mouse ␤3GalT Genes-The
␣EST division of the GenBank™/EMBL Data Bank was searched for sequences similar to a human ␤-galactosyltransferase protein (␤xGalT) (10) that was regarded as a possible ␤3GalT. The query sequence was compared with the data base using the tblastn algorithm (Version 1.4.9) (23). ESTs bearing the National Center for Biotechnology Information sequence identification numbers R13867, D81474, H14861, R61672, R13064, R15977, T31289, H70585, H70574, Z43147, and Z45582 were identified as being significantly similar to the query ␤xGalT sequence (Table I). It is noteworthy that we have not detected any EST that is identical to the probed ␤xGalT, indicating that this gene is not represented yet in the EST data base. The retrieved EST sequences were aligned with the query sequence, and partially overlapping ESTs were assembled to build larger ORFs (Table I). DNA fragments corresponding to the ␤xGalT gene and to EST probes D81474, H70574, and R15977 were amplified by PCR from a pool of human T-cell cDNAs. The resulting fragments were used as DNA probes in screening a mouse genomic DNA library. We could isolate seven clones, which represented three distinct genomic regions. The three fragments 6, 10, and 13 encompassed the same genomic region and hybridized to EST probe D81474. Both genomic fragments 5 and 14 hybridized to EST probes H70574 and R15977, indicating that the assembled ORF2 and ORF3 were parts of the same gene. Genomic fragments 2 and 15 included the murine ␤xGalT gene (Table I).
The genomic regions hybridizing to the probes were sequenced, and the deduced genes were tentatively named ␤3GalT-I, -II, and -III. The ␤3GalT genes were aligned with the originally identified ESTs (Table I) and additional related ESTs that were retrieved from the EST data base with the blastn algorithm (Fig. 1). This comparison confirmed ORF2 and ORF3 as being parts of the ␤3GalT-III gene. Also, it appeared that the sequence of the human ␤3GalT-III gene could be entirely determined by joining contiguous EST fragments. No EST corresponding to the ␤3GalT-I gene could be identified in the ␣EST data base using the blastn, blastx, and tblastx programs.
Primary Structure and Genomic Organization of Murine ␤3GalT Genes-The murine ␤3GalT-I, -II, and -III genes code for proteins of 326 amino acids (Fig. 2), 422 amino acids (Fig. 3), and 331 amino acids (Fig. 4), respectively. The ␤3GalT-II and FIG. 3. Nucleotide and deduced amino acid sequences of the murine ␤3GalT-II gene. The first in-frame ATG codon at 64 bp is given as the translation initiation codon and is boldface and underlined. The second in-frame ATG codon at 103 bp is boldface. The stop codon at 1330 bp is boldface and underlined. Single-letter notation is used for the amino acids. The amino acids corresponding to the putative transmembrane domain are boldface and underlined.
-III genes contain two in-frame ATG codons spaced by 12 and 11 amino acids, respectively. It is unclear whether the first, second, or both ATG codons are used as the start codon. It appears, however, that in both genes, the proximal ATG codons have stronger contexts for translation initiation (24). A Kyte-Doolittle hydropathy analysis detected single membrane-spanning domains of 19, 19, and 17 amino acids for the ␤3GalT-I, -II, and -III proteins, respectively (Figs. 2-4). The transmembrane domains are located close to the N terminus of each protein, which is typical of the type II transmembrane glycosyltransferase protein family. The protein-coding regions of ␤3GalT-I, -II, and -III are contained in a single exon. However, more proximal untranslated exons might exist, as is the case for other glycosyltransferase genes (25,26). This assumption is supported by the fact that no TATA box sequence was detected in front of the ␤3GalT genes. Also, we noticed that the cDNAs of the three ␤3GalT genes contain ϳ500 bp of untranslated sequence that differs from the genomic sequence 5Ј of the ATG codons (data not shown).
Homology between ␤3GalT Proteins-The three mouse ␤3GalT protein sequences were compared using the ClustalW alignment program (27) (Fig. 5). The greatest identity was observed in the catalytic domains of the proteins. In this region, ␤3GalT-I was 51% identical to ␤3GalT-II and 36% identical to ␤3GalT-III, whereas ␤3GalT-II was 35% identical to ␤3GalT-III. The ␤3GalT enzymes shared no sequence similarity in their N-terminal part corresponding to the cytoplasmic, transmembrane, and stem regions. We noticed that the larger size of ␤3GalT-II reflects a longer stem region, whereas the size of the catalytic domains is conserved among the three ␤3GalT en-zymes. The positions of six cysteine residues in the catalytic domains of the three ␤3GalT proteins were conserved (Fig. 5). This supports the formation of up to three disulfide bridges that would confer a similar structure to the three ␤3GalT enzymes. Such related conservations of cysteine residues have been previously observed in the sialyltransferase protein family (28). Several conserved motifs emerged from the alignment of the ␤3GalT protein sequences (Fig. 5). Notably, we did not detect these same motifs in ␤1,4and ␣1,3-galactosyltransferase sequences, indicating that there are no consensual galactosyl motifs in vertebrate galactosyltransferases. However, a variation of the ␤3GalT motif EDVYVG was recognized in the bacterial galactosyltransferases Lex1 (29), LgtB (30), and LgtE (31) (Fig. 6). This conservation was also detected by others and described as a putative galactosyltransferase motif (32).
Tissue Expression of ␤3GalT Genes-The expression pattern of the three ␤3GalT genes was investigated in adult mouse tissues. We have found the ␤3GalT genes to be predominantly expressed in brain tissue, whereas lower transcript levels were also detected in other tissues (Fig. 7). Beside brain tissue, the 7-kilobase ␤3GalT-I mRNA was detected at low levels in all tissues examined, indicating a possible constitutive expression. Two additional smaller transcripts were also observed in colonic tissue. The ␤3GalT-II transcript was 3 kilobases long and was present in brain and heart tissue and could barely be detected in ovaries, colon, and lymph nodes. The expression of ␤3GalT-III yielded a 2.7-kilobase transcript that was restricted to brain, testes, ovaries, and uterus. In addition, the three ␤3GalT genes were expressed in lymphoid cells as detected by reverse transcriptase-PCR (data not shown). Expression of ␤3GalT Genes in Sf 9 Cells-Sf 9 insect cells were infected with recombinant baculoviruses expressing the three ␤3GalT genes under the control of the polyhedrin promoter. The lysates of Sf 9 cells harvested at 72 h post-infection were assayed for galactosyltransferase activity using various acceptor saccharides (Table II). The highest activity was detected with ␤-conjugated GlcNAc acceptors like GlcNAc-␤-pNP and GlcNAc-␤-benzyl (Bzl). By contrast, the ␣-conjugated Glc-NAc acceptors yielded a 10 -20-fold lower activity compared with the ␤-conjugates. No significant galactosyltransferase activity was measurable toward GalNAc acceptors. This finding was corroborated by assays performed in the presence of the high molecular weight acceptor asialo-ovine submaxillary mucin, which carries multiple O-linked GalNAc monosaccharides (16). Also, the three ␤3GalT enzymes failed to transfer GlcNAc or GalNAc to the acceptors GlcNAc-␤-Bzl, Gal-␤-Bzl, and Gal-NAc-␤-Bzl, showing that the proteins are solely galactosyltransferases (data not shown). The activity of the three ␤3GalT proteins was strictly dependent on the divalent cation Mn 2ϩ (data not shown). Also, the proteins were more active in the presence of 1% Triton X-100 (data not shown). The enzymatic properties of the three ␤3GalT proteins were determined for the donor UDP-Gal and the acceptor GlcNAc-␤-pNP (Table III). When compared with ␤3GalT-I, we found that ␤3GalT-II and ␤3GalT-III exhibited 3-and 2-fold higher K m values for Glc-NAc-␤-pNP, respectively. The opposite was observed with the K m for the donor UDP-Gal, for which ␤3GalT-I showed 4-and 2-fold greater K m values than ␤3GalT-II and ␤3GalT-III, respectively.
Analysis of Linkage Specificity of ␤3GalT-The products of the galactosyltransferase activity assays of the lysates from baculovirus-infected Sf 9 cells expressing ␤3GalT-I, -II, and -III and wild-type baculoviruses were resolved by NP-HPLC. Fig.  8A (panel i) shows the profile for the products of the ␤3GalT-I assay. The elution positions of UDP, UDP-Gal, GlcNAc-␤-pNP, and Gal␤1,3/4GlcNAc-␤-pNP (peak g) were assigned by comparison with known standards (Fig. 8A, panel ii). The HPLC fractions of the ␤3GalT-I, -II, and -III reaction products containing the disaccharide were pooled, evaporated to dryness, redissolved in water, and applied to the reverse phase HPLC column (Fig. 8B, panels ii-iiii), which resolves the linkagespecific isomers Gal␤1,3GlcNAc and Gal␤1,4GlcNAc. The Gal␤GlcNAc linkage was shown to be exclusively ␤1,3. It was assigned by comparing the elution position of the pNP-labeled disaccharides with those of standard Gal␤1,4GlcNAc-␤-pNP (Fig. 8B, panel i) and standard Gal␤1,3GlcNAc-␤-pNP (panel v) and by co-injection with the authentic material (data not shown). The assignment was consistent with the finding that the Gal␤1,3GlcNAc-␤-pNP product was not susceptible to S. pneumoniae galactosidase, in contrast to Gal␤1,4GlcNAc-␤-pNP (data not shown). DISCUSSION Based on the retrieval of ESTs sharing similarity with a ␤-galactosyltransferase gene sequence, we have identified three novel murine ␤3GalT genes. The identification of multiple genes encoding a ␤3GalT activity is consistent with recent similar findings made for other glycosyltransferase families (33). The presence of several isozymes can be interpreted as a way to maintain a greater potential to glycosylate different related acceptor structures, thereby increasing the versatility of glycosyltransferases in shaping the glycoconjugate repertoire.
The three ␤3GalT proteins are heterologous in size, with ␤3GalT-I and ␤3GalT-III representing to date the smallest mammalian glycosyltransferases after the GalNAc ␣2,6-sialyltransferase-III (5). As observed for many glycosyltransferase genes, two in-frame ATG codons are present in the ␤3GalT-II and ␤3GalT-III genes. The two ATG codons are located 5Ј of the region encoding the transmembrane domain, and an alternate initiation of translation would unlikely affect the localization and activity of the two proteins. The three ␤3GalT proteins represent a distinct family from the ␤1,4and ␣1,3-galactosyltransferase proteins, as no sequence similarity can be detected between ␤3GalT enzymes and the other galactosyltransferases. This difference is emphasized by comparing the genomic organization of ␤3GalT genes with that of ␤1,4and ␣1,3-galactosyltransferase genes, which contain multiple exons coding for protein sequence (14,15). These distinctions suggest that the family of ␤3GalT genes separated early in evolution from the other galactosyltransferases and possibly expanded by gene duplication. It remains open as to whether the ␤3GalT gene family comprises additional related members. The search for ESTs similar to the three murine ␤3GalT sequences supports the notion of a gene family consisting of 7-10 members.
The alignment of the three ␤3GalT protein sequences highlighted various conserved regions that were all located in the postulated catalytic domain. These ␤3GalT putative motifs were absent in other vertebrate galactosyltransferase protein FIG. 7. Differential expression of the three ␤3GalT genes in murine tissues. Northern blot analysis was performed using 5 g of total RNA from each tissue sample, as is visible on the ethidium bromide staining (bottom panel). The blots were probed with full-length ␤3GalT gene probes as detailed under "Experimental Procedures."   sequences. However, a short ␤3GalT motif was identified in bacterial galactosyltransferases, suggesting that these amino acids are essential for galactosyltransferase activity. Some of the ␤3GalT motifs were also present in the Drosophila proteins Fringe and Brainiac as outlined by Yuan et al. (32). Fringe (34) and Brainiac (35) have been described as secreted signaling molecules, although their mechanism of action has not been elucidated yet. The similarity between Fringe, Brainiac, and ␤3GalT enzymes suggests that the two Drosophila proteins may represent additional members of the ␤3GalT family.
The acceptor specificity of the three ␤3GalT enzymes was restricted to ␤-conjugated GlcNAc. However, the high K m values measured for the acceptor GlcNAc-␤-pNP indicate that the physiological acceptors are likely more specific oligomeric structures. The ␤3GalT enzymes enable the formation of type 1 carbohydrate chains, which are widely distributed on glycolipids and glycoproteins throughout tissues. Being the acceptor of the Lewis enzyme, the type 1 disaccharide Gal␤1,3GlcNAc is related to the occurrence of Le a and Le b antigens (36,37). Although mainly expressed in brain tissue, the three ␤3GalT genes are differentially expressed in mouse tissues, which suggests that these three ␤3GalT enzymes may participate in the formation of Le a and Le b antigens in mouse tissues. The identification of three ␤3GalT genes highlights the structural diversity encountered in the galactosyltransferase gene family. The difference in sequence and genomic organization between ␤3GalT and other vertebrate galactosyltransferase genes suggests a distinct evolution. The structural information gained from the cloning of the three ␤3GalT genes provides a way to identify additional members of this branch of the galactosyltransferase family.