Identification of Novel α1,3-Galactosyltransferase and Elimination of α-Galactose-containing Glycans by Disruption of Multiple α-Galactosyltransferase Genes in Schizosaccharomyces pombe*

Background: We searched for unidentified α1,3-galactosyltransferases in Schizosaccharomyces pombe and identified three novel genes (otg1+–otg3+). Results: The Otg proteins were found to be glycosyltransferases with the ability to form α1,3-linkages between Gal and α-Man residues. Conclusion: Complete elimination of α-galactosylation in S. pombe was achieved by multiple deletions of newly found otg+ genes. Significance: This is the first study to functionally identify and characterize novel α1,3-galactosyltransferases in S. pombe. The oligosaccharides from fission yeast Schizosaccharomyces pombe contain large amounts of d-galactose (Gal) in addition to d-mannose (Man), in contrast to the budding yeast Saccharomyces cerevisiae. Detailed structural analysis has revealed that the Gal residues are attached to the N- and O-linked oligosaccharides via α1,2- or α1,3-linkages. Previously we constructed and characterized a septuple α-galactosyltransferase disruptant (7GalTΔ) anticipating a complete lack of α-Gal residues. However, the 7GalTΔ strain still contained oligosaccharides consisting of α1,3-linked Gal residues, indicating the presence of at least one more additional unidentified α1,3-galactosyltransferase. In this study we searched for unidentified putative glycosyltransferases in the S. pombe genome sequence and identified three novel genes, named otg1+–otg3+ (α one, three-galactosyltransferase), that belong to glycosyltransferase gene family 8 in the Carbohydrate Active EnZymes (CAZY) database. Gal-recognizing lectin blotting and HPLC analyses of pyridylaminated oligosaccharides after deletion of these three additional genes from 7GalTΔ strain demonstrated that the resultant disruptant missing 10 α-galactosyltransferase genes, 10GalTΔ, exhibited a complete loss of galactosylation. In an in vitro galactosylation assay, the otg2+ gene product had Gal transfer activity toward a pyridylaminated Man9GlcNAc2 oligosaccharide and pyridylaminated Manα1,2-Manα1,2-Man oligosaccharide. In addition, the otg3+ gene product exhibited Gal transfer activity toward the pyridylaminated Man9GlcNAc2 oligosaccharide. Generation of an α1,3-linkage was confirmed by HPLC analysis, α-galactosidase digestion analysis, 1H NMR spectroscopy, and LC-MS/MS analysis. These results indicate that Otg2p and Otg3p are involved in α1,3-galactosylation of S. pombe oligosaccharides.

The fission yeast Schizosaccharomyces pombe is a promising host for expression of therapeutic glycoproteins because it shares greater similarity to higher animals than the budding yeast Saccharomyces cerevisiae in regard to splicing mechanisms, cell division control, transcription-initiation mechanisms, and post-translational modifications (1,2). To date our group has developed S. pombe protein-production systems that have been useful for producing many types of heterologous proteins from various organisms including humans (3). However, for the production of therapeutic glycoproteins intended for human use, yeasts including S. pombe are currently less useful due to their inability to modify proteins with human-compatible glycan structures. To overcome this problem, glycoengineering has been attempted in several yeast species (4). Recently, the secretory production of erythropoietin containing sialylated biantennary glycan was demonstrated in the methylotrophic yeast Pichia pastoris by deletion of yeast-specific glycosyltransferases and introduction of many other genes responsible for 1) processing of high mannose-type glycans to a precursor of complex-type glycan, 2) conversion of the complex type precursor glycan to complex-type glycan, 3) human-like sugar nucleotide synthesis, and 4) their transport from the cytosol to the Golgi lumen (5). With the increasing importance of yeast species other than P. pastoris as alternative hosts for production of therapeutic glycoproteins, glycoengineering targeting human-compatible glycans has received significant attention in such species.
To this end we have been analyzing the precise glycan structures in S. pombe glycosylation mutants to allow use of this organism as an alternative glycoprotein-producing host (15)(16)(17)(18)(19)(20)(21)(22)(23). Although we have succeeded in completely eliminating both the outer chain structures and ␣-linked Gal residues by construction of a gms1⌬och1⌬ deletion mutant, the gms1 ϩ gene encoding a Golgi-localized UDP-Gal transporter is required for ␤1,4-galactosylation processes that produce biantennary complex-type oligosaccharides in a subsequent humanization step in S. pombe. Recently, we constructed a septuple ␣-galactosyltransferase-related gene disruptant (7GalT⌬) as an alternative approach for eliminating ␣-galactosylation and analyzed resultant glycan structures (23). Although the 7GalT⌬ strain was expected to have lost ␣-galactosylation, glycan structural analysis revealed that it still had ␣1,3-linked Gal residues, indicating the presence of unidentified ␣1,3-galactosyltransferase(s). Therefore, to construct the desired S. pombe ␣-galactosylation null mutant, identification of the ␣1,3-galactosyltransferase(s) is required.
Here we describe the identification and characterization of the novel ␣1,3-galactosyltransferase genes. By searching for putative glycosyltransferases in the S. pombe genome, we found three novel uncharacterized genes, otg1 ϩ -otg3 ϩ (␣ one, threegalactosyltransferase). Disruption of these three genes in the 7GalT⌬ 2 mutant resulted in a strain in which 10 ␣-galactosyltransferase genes had been deleted (10GalT⌬) and caused a complete loss of ␣-galactosylation in its glycan. Moreover, in vitro enzymatic assay revealed that Otg2p had ␣-Gal transfer activity toward both N-and O-linked glycans, whereas Otg3p had ␣-Gal transfer activity toward N-glycans.

EXPERIMENTAL PROCEDURES
Strains, Media, and Genetic Methods-The fission yeast strains used in this study are listed in Table 1. Standard rich medium (YES (yeast extract/supplement) and YPD (yeast extract/peptone/dextrose)) and synthetic minimal medium for growing S. pombe were used as described (24). S. pombe cells were transformed by the lithium acetate method (25,26). Standard genetic methods have been described (27). Escherichia coli XL1-Blue (Stratagene) was used for all cloning procedures.
Gene Disruptions-In a previous study the 7GalT⌬otg1⌬ strain (8GalT⌬) had already been constructed because otg1 ϩ gene is located between gmh1 ϩ and gmh2 ϩ genes, and these genes were disrupted simultaneously (23). For construction of the 10GalT⌬ strain, the S. pombe ura4 ϩ marker used for disruptions in the 8GalT⌬ stain was recovered by FOA treatment (28). Because otg2 ϩ and otg3 ϩ are located adjacent to one other and are about 3 kbp in length, they were disrupted simultaneously. The plasmid used to disrupt otg2 ϩ and otg3 ϩ was constructed as follows. A 3.8-kbp fragment carrying the otg2 ϩ and otg3 ϩ genes was amplified from chromosomal DNA by PCR using the Ex Taq DNA polymerase (Takara Co. Ltd., Kyoto, Japan) and the following oligonucleotides: sense (5Ј-GGTTCTTCGCCTGCTAATGTTGCTGTCGG-3Ј) and antisense (5Ј-CAGCACGAAGATAGGGTAAGTCCTCTTCGG-3Ј). The 3.8-kbp fragments were recovered and ligated into pGEM-T Easy vector. The HindIII-XhoI sites within the cloned otg2 ϩ -otg3 fragment were digested, and a 1.6-kbp ura4 ϩ -containing fragment was inserted.
The otg1⌬otg2⌬otg3⌬ triple disruptant strain was constructed by replacing the internal otg1 ϩ gene and otg2 ϩ -otg3 ϩ fragments with the ura4 ϩ marker. For the otg1 ϩ locus, a DNA fragment carrying the otg1 ϩ gene was amplified from chromosomal DNA by PCR using Ex Taq DNA polymerase, and the following oligonucleotides were used: sense (5Ј-AAGCTGTT-TCTGTTGGTTCAACTAATTTGC-3Ј) and antisense (5Ј-TAGCTCTTCATTTCGCTGGATGTTGGGTTG-3Ј). A 3.0kbp fragment was recovered and ligated into a pGEM-T Easy vector (Promega). Replacement of an internal otg1 ϩ gene fragment with ura4 ϩ marker was conducted by the in-fusion method (29). Preparation of a linearized pGEM-T Easy vector carrying only flanking regions of otg1 ϩ was performed by inverse PCR using PrimeSTAR HS DNA polymerase (Takara) and the following oligonucleotide primers, which contained overlapping 15-bp extensions homologous to regions on the ura4 ϩ marker fragment used to replace the otg1 ϩ ORF: sense (5Ј-CTATAGTGTCACCTACTCATTAAAGGAGGAATAC-C-3Ј) and antisense (5Ј-GTGAGTCGTATTACAGATAAGA-GATTAATTCAACG-3Ј). A DNA fragment carrying ura4 ϩ marker fragment was amplified from a pGEM-T Easy vector carrying the ura4 ϩ marker by PCR using PrimeSTAR HS DNA polymerase and the following oligonucleotide primers: sense (5Ј-TGTAATACGACTCACTATAGGGCG-3Ј) and antisense (5Ј-TAGGTGACACTATAGAATACTCAAGC-3Ј). The two DNA fragments were then joined by an In-Fusion enzyme reaction (Clontech). The 9GalT⌬ strain in which only Otg1p is present was constructed from the 10GalT⌬ strain by using of integrating plasmid, pJK148-otg1 ϩ , as described (23). The 10GalT⌬och1⌬ strain was constructed from the 10GalT⌬ strain by replacing an internal och1 ϩ fragment with the ura4 ϩ marker as described (30).
Microscopy-Cells were grown to an A 600 of 1.0 in YES medium and were observed with an Olympus BX-60 fluorescence microscope (Olympus, Tokyo). Images were captured with a Cool SNAP CCD camera using MetaMorph (Roper Scientific) and were saved as Adobe Photoshop files on an Intel ® Macintosh computer.
Western Blot Analysis-Western blot analysis was performed essentially as described (22) with a slight modification. Briefly, to express human transferrin (hTF), transformants harboring pTL2OSTFN-CF-4XL were cultured in 5 ml of YPD (yeast extract/peptone/dextrose) medium at 30°C. Culture supernatants containing hTF were precipitated with acetone (50%, v/v), dissolved in 100 l of SDS-PAGE sample buffer, and boiled for 5 min. A total of 20 l of sample solution was loaded onto a 7% acrylamide gel and electroblotted onto a polyvinylidene difluoride membrane (Millipore). Immunodetection analysis was conducted on the blotted membrane with SNAP i.d. (Millipore) using mouse monoclonal anti-FLAG M2 antibody (Sigma) as the first antibody at a dilution of 1:500 and horseradish peroxidase-conjugated sheep anti-mouse IgG antiserum (GE Healthcare) as the second antibody at a dilution of 1:10,000. FLAG epitope-specific signals were visualized by chemiluminescence (Luminata Forte Western HRP substrate; Millipore).
Preparation of Pyridylaminated (PA) Glycans from Cell Surface Glycoproteins-Cells were cultivated in YES medium at 30°C and harvested in early stationary phase. Cell surface glycoproteins were extracted by autoclaving at 121°C for 90 min in citrate buffer (20 mM citrate-NaOH, pH 7.0) followed by precipitation with methanol. The precipitates were dissolved in hot water, dialyzed, and lyophilized. Oligosaccharides were liberated from galactomannoproteins by hydrazinolysis followed by N-acetylation. The reducing ends of the liberated oligosaccharides were pyridylaminated as described (32,33).
Glycan Structural Analysis-Separation and structural identification of the PA-oligosaccharides were carried out using size-fractionation or reversed-phase high performance liquid chromatography as described (23).
Protein Assay-The amount of protein was determined using BCA protein assay kit (Thermo Scientific, Rockford, IL) according to the manufacturer's instructions with bovine serum albumin as a standard.
Preparation of Solubilized Microsomal Proteins-Solubilized enzymes from various strains were prepared as described (39). Briefly, cells were grown in minimal medium at 30°C, harvested in early stationary phase, washed with ice-cold water, and suspended in 50 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 0.25 M sucrose, and Complete EDTA-free protease inhibitor mixture (Roche Applied Science). Cells were lysed with glass beads by vortexing for 30 s and then holding on ice for 30 s for a total of 15 repeated cycles. Glass beads, unbroken cells, and large cell debris were removed by centrifugation at 1000 ϫ g at 4°C. The resultant supernatant was centrifuged at 100,000 ϫ g for 1 h at 4°C. The membrane pellet was resuspended in 50 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 0.25 M sucrose, and 2% Triton X-100 to a protein concentration of 4 mg/ml and incubated on ice for 30 min. The resultant mixture was then used as the source of solubilized enzyme for the galactosyltransferase assay.

␣1,3-Galactosyltransferases in S. pombe
Galactosyltransferase Assay-The galactosyltransferase assay was carried out as described (39). The reaction mixture in a total volume of 30 l contained 60 g of solubilized enzyme, 100 mM HEPES-NaOH, pH 7.0, 0.1 M sucrose, 1 mM MnCl 2 , 1 mM UDP-Gal, 10 pmol PA-oligosaccharides, and was incubated at 37°C for 12 h. The reaction was terminated by boiling for 5 min. Reaction products were analyzed by size-fractionation HPLC.
NMR Spectroscopy-Sample was exchanged several times in D 2 O followed by lyophilization. Finally, the lyophilized sample was dissolved in 35 l of D 2 O. The 1 H NMR spectrum was recorded on a Varian Inova-600 NMR spectrometer equipped with NANO probe (Agilent Technologies, Inc., Wilmington, DE).
Liquid Chromatography-Tandem Mass Spectrometry-LC-MS/MS was performed with an Agilent Technologies 1200 series instrument (Agilent Technologies, Santa Clara, CA) equipped with HCT plus software (Bruker Daltonics, Bremen, Germany) as described (49).

RESULTS
Identification of a Family of Putative S. pombe ␣1,3-Galactosyltransferase Genes-We previously constructed a septuple ␣-galactosyltransferase mutant and showed that ␣1,3-linked Gal-containing oligosaccharides were still present, indicating an unidentified residual ␣1,3-galactosyltransferase activity (23). To identify undiscovered ␣1,3-galactosyltransferase gene(s), we searched for glycosyltransferases in the S. pombe genome sequence. Among the predicted glycosyltransferases, three genes (SPAC5H10.12c, SPBC4C3.08, and SPBC4C3.09, designated otg1 ϩ , otg2 ϩ , and otg3 ϩ for ␣ one, three-galactosyltransferase, respectively) with unknown functions were identified and exhibited 31-48% identity with each other at the amino acid level (Fig. 1A). The proteins encoded by the otg genes belong to glycosyltransferase family 8 (GT8) of the Carbohydrate-Active EnZymes (CAZy) database (35) and possess a DXD motif that is involved in divalent cation binding necessary for sugar-nucleotide binding. The Otg proteins also have the typical type II architecture of Golgi glycosyltransferases with the prediction of a small N-terminal cytoplasmic domain, a transmembrane domain, and a C-terminal catalytic domain. To predict the function of the genes, we conducted a BLASTp search that revealed greater amino acid sequence similarity of Otg proteins to many GT8 proteins, which included Arabidopsis myo-inositol ␣1,3-galactosyltransferase (galactinol synthase) and E. coli lipopolysaccharide ␣1,3-galactosyltransferase. Therefore, we speculated that these genes might be responsible for ␣1,3-galactosylation in S. pombe. A phylogenetic analysis of these GT8 proteins indicated that the Otg proteins comprised a novel branch relative to GT8 proteins with known activities (Fig. 1B). This branch was relatively close to a branch consisting of pathogenic fungal putative GT8 proteins that may share similar enzymatic functions. These observations are consistent with the possibility that the Otg proteins are ␣1,3-galactosyltransferases. To explore this directly, we disrupted the three genes from the septuple mutant to construct a strain lacking 10 presumptive ␣-galactosyltransferase genes (10GalT⌬).
Phenotypic Characterization of the 10GalT⌬ Mutant-Light microscopic observation using Nomarski optics showed that wild-type cells displayed a normal cylindrical morphology. The morphology of 10GalT⌬ cells was an unusual in that the center swelled abnormally or was rounded, similar to gms1⌬ and the 7GalT⌬ cells (Fig. 2A). The peanut agglutinin PNA, a lectin that recognizes the Gal residues, reacted with wild-type and 7GalT⌬ cells, whereas it did not react with 10GalT⌬ cells or with the gms1⌬ cells, which served as a negative control (Fig. 2B). The 7GalT⌬ cells stained less intensely than wild-type cells, presumably due to the remaining ␣1,3-linked Gal residues. In addition, the gms1⌬, the 7GalT⌬, and the 10GalT⌬ cells were temperature-and hygromycin B-sensitive (Fig. 2C), which are characteristic phenotypes of glycosylation mutants (15,36). Interestingly, 7GalT⌬ and 10GalT⌬ cells grew slower than gms1⌬ cells on YES agar plates. The doubling times for the wild type and gms1⌬, the 7GalT⌬, and 10GalT⌬ strains were 2.70, 4.03, 4.48, and 4.52 h, respectively.
Intracellular Localization of Otg Proteins-To determine the localization of the Otg proteins, we expressed C-terminal GFP fusion Otg in wild-type cells and observed fluorescence by fluorescence microscopy. Cells expressing Otg1-GFP, Otg2-GFP, and Otg3-GFP exhibited punctate fluorescent patterns, which are characteristic of Golgi localization and which overlapped with signals for the Golgi marker protein Gms1-RFP (Fig. 3) (17). However, some Otg-GFP signals did not appear to colocalize with Gms1-RFP. This might reflect differential intracellular dynamics of these proteins for Golgi retention as described in the case of Mnn9p or Vrg4p and Och1p (37,38,50). These results indicate that all three Otg-GFP fusion proteins localize to the Golgi apparatus.
hTF, which is a secreted glycoprotein of ϳ80 kDa with 20 disulfide bonds and two N-glycosylation sites, was used as a heterologous protein with N-linked glycans. Western blot analysis of recombinant hTF secreted into the culture medium of wild-type cells showed a smeared band with a higher molecular ␣1,3-Galactosyltransferases in S. pombe NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46 JOURNAL OF BIOLOGICAL CHEMISTRY 38869 mass, ϳ100 -120 kDa, whereas that from gms1⌬ and och1⌬ cells migrated faster (Fig. 5A), presumably due to these glycosylation defects. The mobility of recombinant hTF from 10GalT⌬ cells was almost the same as that from gms1⌬ cells (Fig. 5A), indicating that the size of N-linked glycan on hTF from gms1⌬ and 10GalT⌬ cells was almost identical. The smaller bands below 80 kDa seen in gms1⌬, och1⌬, and 10GalT⌬ cells were degradation products as described in the case of the recombinant human growth hormone (52). Silver staining after SDS-PAGE analysis was conducted as a loading control (Fig. 5B).

␣1,3-Galactosyltransferases in S. pombe
glycan structure by size-fractionation HPLC. HPLC profile of O-linked glycan structures in both of the 9GalT⌬ and 10GalT⌬ strains were quite similar (data not shown). Furthermore the proteins from the 9GalT⌬ strains did not react with PNA (data not shown). These data suggest that Otg1p does not have galactosyltransferase activity. Furthermore, all Otg proteins were inactive toward Man-PA, Man␣1,2Man-PA, and Gal␣1,2Man-PA as acceptor substrates (data not shown). In addition, the linkage formed by Otg2p was investigated. The enzymatic product (fraction a in Fig. 6B) obtained from a reaction with Man␣1,2Man␣1,2Man-PA was collected and subjected to ␣-galactosidase digestion assays (data not shown). After ␣-ga-lactosidase digestion, the elution position of fraction a shifted toward that of Man␣-Man␣1,2-Man-PA, indicating that the Gal residue was attached by an ␣-linkage. Further structural determination of the enzymatic product was carried out by 1 H NMR spectroscopy (Fig. 6C) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis (Fig. 6D). Three anomeric signals at 5.16, 5.22, and 5.23 were detected in 1 H NMR spectrum. The 1 H NMR spectrum of the enzymatic product was not identical to that of Man␣1,2-(Gal␣1,3-)Man␣1,2-Man as described in the previous report (7), although we pre-FIGURE 2. Cell morphology, lectin blot, and temperature and hygromycin B sensitivity. A, wild-type, gms1⌬, 7GalT⌬, and 10GalT⌬ cells were cultured in YES medium and observed by Nomarski optics. B, a dot-blot lectin staining assay was performed with HRP-PNA. One microgram of protein was blotted onto a Hybond-N nylon membrane, probed by HRP-PNA, and visualized by using 3,3Ј-diaminobenzidine and H 2 O 2 . C, wild-type, gms1⌬, 7GalT⌬, and 10GalT⌬ cells were grown on YES plates at 37 or 30°C in the absence or presence of 25 g/ml hygromycin B for 3 days.   (19). In the schematic structures shown, the vertical and the diagonal bars between the letters indicate ␣1,2and ␣1,3-linkages, respectively. Cellopentaose was added as an internal standard. G, galactose; M, mannose; PA, pyridylamino. ␣1,3-Galactosyltransferases in S. pombe NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46 dicted that the Otg2p enzymatic product would be Man␣1,2-(Gal␣1,3-)Man␣1,2-Man. The signal at 5.16 could be assigned to a 3-substituted ␣1,2-linked Man residue with a downfield shift due to an ␣-Gal attachment (20). This appeared to be quite unlike the 2-substituted ␣1,2-linked Man residue, whose anomeric signal was reported to be around 5.26 (7). The other two anomeric signals (␦ 5.22 and 5.23) are thought to correspond to 2-substituted ␣1,2-linked Man and the terminal ␣1,3-linked Gal residues, whose signals were reported to be detected around 5.26 and 5.28 (7), respectively. Although we attempted to determine these signal assignments by two-dimensional correlation spectroscopy analysis, no signals were found to correlate with the C2-H resonances (data not shown), likely due to limited amounts of the sample. The negative mode MS spectra indicated a precursor ion of m/z 743.7 corresponding to the calculated mass of Hex 4 -PA (m/z ϭ 743.3) (data not shown). In LC-MS/MS analysis, the presence of the 0,3 X 2 -H 2 O/Y 3 ion at m/z 472.1 indicated a 2-or 3-substituted penultimate ␣1,2linked Man. Furthermore, the E2 ion at m/z 305.0 allowed exclusion of a 1,2-substitution on the penultimate ␣1,2-linked Man. These results demonstrated that the terminal ␣-Gal residue was attached to the penultimate ␣1,2-linked Man by 1,3linkage. From these results, the structure of the Otg2p product was deduced to be Gal␣1,3-Man␣1,2-Man␣1,2-Man␣1,2-PA. Collectively these results indicate that Otg2p acts on N-and O-linked glycans and that Otg3p acts on N-linked glycan (Fig. 7).

DISCUSSION
The glycans of S. pombe are known to contain large amounts of Gal residues attached to N-and O-linked glycans via ␣1,2and ␣1,3-linkages (6). Although ␣1,2-galactosyltransferases have been identified and partially characterized (23, 39 -41), before this study, none of the ␣1,3-galactosyltransferases had yet been identified or characterized. We first sought for genes responsible for ␣1,3-galactosylation in the S. pombe genome database to delete them from the 7GalT⌬ strain that still contained ␣1,3-Gal residues in its glycans. Elimination of ␣1,3-Gal residues is required to produce human-type complex-type glycan in S. pombe. The genome-wide search detected only three genes (otg1 ϩ , otg2 ϩ , and otg3 ϩ ) of unknown function. These genes were originally annotated as N-acetylglucosaminyltransferases based on weak sequence similarity to ScGNT1 (34). However, amino acid sequence alignment revealed that ScGnt1p had an extra unshared sequence and a relatively large number of amino acid residues that were not conserved in the SpOtg proteins (Fig. 1A). Furthermore, ScGnt1p were found to be phylogenetically distinct from the Otg proteins (Fig. 1B). Therefore, we assumed that the SpOtg proteins might not have N-acetylglucosaminyltransferase activity but other glycosyltransferase activities instead. The otg genes were then deleted from wild-type and from the 7GalT⌬ strain to generate otg1⌬otg2⌬otg3⌬ and 10GalT⌬, respectively, and glycan structures from the latter strains were analyzed (Figs. 4 and 5). The observed loss of ␣1,3-Gal residues suggested that at least one of the Otg proteins was involved in ␣1,3-galactosylation. The chromatogram of glycan structures from the otg1⌬otg2⌬otg3⌬ mutant shows that ␣1,3-Gal-containing tetrasaccharides were completely eliminated. Although the residual peak could be an ␣1,2-Gal-containing glycan such as Gal␣1,2Man␣1,2Man␣1,2Man-PA, the precise structure has not been determined.
Mammalian-type ␣1,3-galactosyltransferases, which are involved in anti-Gal epitope and ABO blood group B synthesis and belong to the GT6 family have been well characterized (14). However, the mammalian-type ␣1,3-galactosyltransferases and SpOtg proteins share limited amino acid sequence similarity and comprise a distinct clade (Fig. 1B), suggesting that the SpOtg proteins comprise a new family. To determine whether the novel S. pombe Otg proteins have homologs in other organisms, BLASTp searches were conducted against public genome databases. Interestingly, homologous proteins were found to be putative GT8 proteins and mostly found among nematode-, plant-, or human-pathogenic fungal species (e.g. Arthrobotrys oligospora, Leptosphaeria maculans, Aspergillus fumigatus, Cryptococcus neoformans) (Fig. 1B and , suggesting that these species may synthesize glycan containing ␣1,3-Gal residues. However, to our knowledge, no reports have documented the presence of such residues in these species. It would be intriguing if such residues were related to pathogenicity. We have been investigating mechanisms of galactosylation in S. pombe, which consists of UDP-Gal biosynthesis in the cytosol (46), its transport into the lumen of the Golgi apparatus (16), and the transfer of Gal residues to glycans by ␣-galactosyltransferases (23, 39 -41). From our previous results, Gma12p, Gmh2p, and Gmh6p were shown to be involved in galactosylation of both of N-and O-linked glycans, and Gmh3p was shown to be involved in galactosylation of N-linked glycan. From our present in vitro enzymatic analysis, Otg2p and Otg3p have the galactosyltransferase activity toward M9-PA whereas Otg2p acts on Man␣1,2-Man␣1,2-Man-PA (Fig. 6). Analysis of the products of the enzymatic reactions using the combination of

␣1,3-Galactosyltransferases in S. pombe
reversed-phase HPLC, ␣-galactosidase digestion, 1 H NMR analysis, and LC-MS/MS analysis (Fig. 6) indicated that Otg2p was able to transfer ␣1,3-Gal to Man␣1,2-Man␣1,2-Man-PA to generate Gal␣1,3-Man␣1,2-Man␣1,2-Man-PA (Fig. 6C). Glycans with a lone ␣1,3-Gal terminus have never been detected in S. pombe O-linked glycan. This might be because ␣1,2-galactosyltransferases such as Gma12p or putative ␣1,2-mannosyltransferases (presumably Omh proteins) immediately transfer the ␣1,2-Gal or Man residues onto this type of glycan after transfer of ␣1,3-Gal residues. Although we tested Man-PA, Man␣1,2Man-PA, and Gal␣1,2Man-PA as a potential substrate, no Otg proteins were able to transfer ␣1,3-Gal to these PA-O-linked glycans, possibly due to structural differences between the PA-glycan and native glycan. For example, in these PA-O-linked glycans, the mannopyranose ring at the reducing ends are open due to pyridylamination, which might prevent the Otg enzymes from recognizing them and transferring the ␣1,3-Gal moiety to them. Among the Otg proteins, only Otg1p has never shown any enzymatic activities. Therefore, there are still possibilities that Otg1p has other additional glycosyltransferase activities other than ␣1,3-galactosyltransferase activity (e.g. N-acetylglucosaminyltransferase activity). We are currently attempting to overexpress and purify soluble forms of the Otg proteins to better characterize their enzymatic activities. It has been reported that ␣-galactosylation in S. pombe is required for sexual and nonsexual flocculation (18,47,48). The mating process was significantly affected in the gms1⌬ cells during nutritional starvation because cells were incapable of sexual flocculation (18). Recently the gsf2 ϩ gene, encoding a flocculin that binds to Gal residues located on cell surface glycans, was identified (48). Nonsexual flocculation and filamentous invasive growth was tightly controlled by gsf2 ϩ expression. However, whether either of ␣1,2-Gal or ␣1,3-Gal residues is mainly involved in these physiological events remains to be elusive. To address this issue, further work will be required to elucidate the precise mechanism of ␣-galactosylation including the enzymatic characterization of ␣1,2and ␣1,3-galactosyltransferases.
In conclusion, this is the first study to functionally identify and characterize novel ␣1,3-galactosyltransferases in S. pombe. The amino acid sequences of the Otg proteins (GT8) are quite different from those of mammalian ␣1,3-galactosyltransferases (GT6). We have shown for the first time that the Otg proteins are glycosyltransferases with the ability to form ␣1,3-linkages between Gal and ␣-Man residues. More detailed biochemical studies using the single and all combinations of double otg⌬ mutants and purified ␣-galactosyltransferases will provide a more complete understanding of galactosylation pathways in S. pombe and how to manage them for glycoengineering therapeutics and industrial purposes.