Molecular Cloning and Characterization of Human GnT-IX, a Novel β1,6-N-Acetylglucosaminyltransferase That Is Specifically Expressed in the Brain*

A novel β1,6-N-acetylglucosaminyltransferase (β1, 6GnT) cDNA was identified by a BLAST search using the amino acid sequence of human GnT-V as a query. The full-length sequence was determined by a combination of 5′-rapid amplification of cDNA end analysis and a further data base search. The open reading frame encodes a 792 amino acid protein with a type II membrane protein structure typical of glycosyltransferases. The entire sequence identity to human GnT-V is 42%. When pyridylaminated (PA) agalacto biantennary N-linked oligosaccharide was used as an acceptor substrate, the recombinant enzyme generated a novel product other than the expected GnT-V product, (GlcNAcβ1,2-Manα1,3-)[GlcNAcβ1,2-(GlcNAcβ1,6-)Manα1,6-]Manβ1,4-GlcNAcβ1,4-GlcNAc-PA. This new product was identified as [GlcNAcβ1,2-(GlcNAcβ1,6-)Manα1,3-][Glc-NAcβ1,2-(GlcNAcβ1,6-)Manα1,6-]Manβ1,4-GlcNAcβ1,4-GlcNAc-PA by mass spectrometry and 1H NMR. Namely, the new GnT (designated as GnT-IX) has β1,6GnT activity not only to the α1,6-linked mannose arm but also to the α1,3-linked mannose arm of N-glycan, forming a unique structure that has not been reported to date. Northern blot analysis showed that the GnT-IX gene is exclusively expressed in the brain, whereas the GnT-V gene is expressed ubiquitously. These results suggest that GnT-IX is responsible for the synthesis of a unique oligosaccharide structure in the brain.

The ␤1,6-branched tri-and tetraantennary glycans are preferentially elongated via the action of ␤1,3GnT and ␤1,4-galactosyltransferase to produce a poly-N-acetyllactosamine structure (2,3). In addition, the N-glycans in GnT V-deficient cells are severely depleted of poly-N-acetyllactosamine but not Oglycans (4,5). Poly-N-acetyllactosamine chains are known to serve as ligands for cell adhesion molecules such as selectins and galectins (6 -8). It has been suggested that changes in poly-N-acetyllactosamine content on sympathetic nerve cells affect their migration (9,10). Furthermore, N-acetyllactosamine promotes the neural outgrowth of primary olfactory neurons (11). These collective findings suggest that GnT-V activity in the brain may play an important role in the process of nerve differentiation via the interaction of poly-N-acetyllactosamine chains and their receptors. However, the GnT-V gene (Mgat5)-knockout mice that apparently lack both detectable GnT-V activity and Phaseolus vulgaris leukoagglutinin reactivity in the brain, appear to be neurologically normal (12). Since a detailed analysis of the carbohydrate chains of these glycoproteins has not been examined, the issue of whether poly-N-acetyllactosamine structure is, in fact, altered in Mgat5 Ϫ/Ϫ brain remains to be elucidated.
When the GnT-V gene was cloned, no homologous genes were evident, suggesting that GnT-V has a unique evolutionary origin (13,14). The GnT-V ortholog gene (gly-2) was recently identified in Caenorhabditis elegans, and there is no homologous gene in this organism (15). However, the recent, rapid accumulation of EST and genomic data enabled us to identify a novel human ␤1,6GnT gene homologous to GnT-V. This new GnT catalyzes the synthesis of a unique N-linked oligosaccharide structure and was designated as GnT-IX, since two additional GnT activities (GnT-VII and -VIII) were demonstrated in CHO mutant cells, although their genes have not been identified (16). Interestingly, the GnT-IX gene is exclusively ex-* This work was supported by the New Energy and Industrial Technology Development Organization as a part of Biotechnology Foundation Research Program for Health Maintenance and Improvement and by a Grant-in-aid for Scientific Research (S) 13854010 from the Japan Society for the promotion of Science and by the 21st Century COE Program by the Ministry of Education, Science, Culture, Sports and Technology in Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM /EBI Data Bank with accession number(s) AB109185.

EXPERIMENTAL PROCEDURES
Materials-A fluorescence-labeled acceptor substrate GnGn-bi-PA, GlcNAc␤1,2-Man␣1,3-(GlcNAc␤1,2-Man␣1,6-)Man␤1,4-GlcNAc␤1,4-GlcNAc-PA was prepared as reported previously (17,18). Gn 3 -triЈ-PA, GlcNAc␤1,2-Man␣1,3-[GlcNAc␤1,2-(GlcNAc␤1,6-)Man␣1,6-]Man␤1,4-GlcNAc␤1,4-GlcNAc-PA was prepared by a large scale GnT-V reaction using GnGn-bi-PA as a substrate followed by the isolation of the reaction product using TSKgel ODS-80TM column (7.8 ϫ 300 mm; TOSOH) as described under "GnT Assay." cDNA Cloning of GnT-IX-Using the amino acid sequence of human GnT-V (14) as a query for a BLAST search at the DNA Data Bank of Japan homology search system, we found a fragment sequence, human hypothetical protein T50606, that exhibited a significant similarity to GnT-V. Using the amino acid sequence of T50606 for an EST data base search, seven clones (AL524151, BF569340, BE785253, BE789945, R87580, BF033743, and R87617) were identified that contained partial sequences corresponding to T50606. One g of human brain total RNA (OriGene Technologies) was reverse-transcribed and then amplified by PCR with a pair of primers designed from the consensus sequence of the The potential N-glycosylation sites are circled. B, alignment of the amino acid sequences of human GnT-IX with human GnT-V (14). The asterisks below the sequences indicate identical residues. seven EST clones: S1 (5Ј-AGGGACCAGAAGCAGAT-3Ј, corresponding to nucleotides 841-857 in Fig. 1) and AS1 (5Ј-GAACCTGCTGGT-GCTTT-3Ј, corresponding to nucleotides 2435-2451). The PCR product was subcloned into pT7Blue vector (Novagen) and then sequenced by the dideoxy chain termination method using DNA sequencers (Applied Biosystems model 377 and 310). The 5Ј-end of the GnT-IX was determined by the combination of rapid amplification of cDNA ends (RACE) using a 5Ј-RACE system kit (Invitrogen). Briefly, 5 g of human brain total RNA was reverse-transcribed with the GnT-IX gene-specific antisense primer GSP1 (5Ј-TACTCTGCAGCTCCTT-3Ј, corresponding to nucleotides 1012-1027), and the first strand cDNA was tailed at the 3Ј-end by the terminal deoxynucleotidyltransferase with dCTP followed by PCR with 5Ј-RACE abridged anchor primer (5Ј-GGCCACGCGTC-GACTAGTACGGGIIGGGIIGGGIIG-3Ј) and GSP1. The PCR product was size-selected (approximately 0.5-1 kb) and then subjected to nested PCR with abridged universal amplification primer (5Ј-GGC-CACGCGTCGACTAGTAC-3Ј) and GSP2 (5Ј-CCAGGACATAGAGTG-CAGTC-3Ј, corresponding to nucleotides 963-982). The sequence of the PCR product was used as a query for a further EST data base search. After aligning the sequences of the EST clones (BE276066, BE391943, BE730360, and BG478724) containing the putative start codon, the open reading frame was amplified by PCR with a pair of primers: 5Ј-CTGCTCGCACCAACAAGT-3Ј, corresponding to nucleotides Ϫ25 to Ϫ8 and AS1. The PCR product was subcloned and sequenced as described above.
Construction of a Vector Encoding His 6 -tagged Human GnT-IX-To prepare a C-terminal His 6 -tagged GnT-IX construct, the full-length cDNA and a Myc-His 6 tag sequence were ligated to pcDNA3.1/Zeo(ϩ) (Invitrogen) with KpnI/EcoRV sites. To prepare a soluble N-terminal His 6 -tagged version of the construct, GnT-IX cDNA was amplified by PCR with primers 5Ј-GCCGAATTCACCATCCGCACAG-3Ј and 5Ј-CCGTCTAGATCACAGACAGCCCTGGC-3Ј and then subcloned into pcDNA3.1/Zeo with Ig signal and His 6 tag sequences. were transfected with the linearized plasmid using the SuperFect transfection reagent (Qiagen) according to the standard protocol used for stable transfection and selected for clones that stably express the gene, based on resistance to Zeocin (Invitrogen) followed by detection of the expression of the His 6 -tagged protein by Western blot analysis with an anti-His tag antibody (Tetra His Antibody; Qiagen). The cloned cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin, and 100 g/ml Zeocin. The cells expressing full-length GnT-IX were washed and pelleted, and the microsomal fraction was prepared as described previously (19). For the expression of the soluble form of GnT-IX, the cells were cultured for 3 days in the same medium with 1% serum and antibiotics without Zeocin. The soluble His 6 -tagged enzyme was partially purified by passing it through a chelating Sepharose Fast Flow (Amersham Biosciences) column according to the manufacturer's instruction. For expression of GnT-V as a positive control for the transferase assay, COS-1 cells were transfected with the pSVK3-hGnT-V vector as described previously (19). The microsomal fraction of the transfected cells was prepared, as described above.

Expression of Human GnT-IX in
Capture and Immunoprecipitation of His 6 -tagged Enzyme-For the capture of the tagged enzyme, the solubilized microsomal fraction was incubated with Ni-NTA Superflow beads (Qiagen) in 20 mM imidazole, 1% Triton X-100, phosphate-buffered saline at 4°C overnight, and the beads were then washed four times with 20 mM imidazole, 0.5% Triton X-100, phosphate-buffered saline. Immunoprecipitation was performed using an anti-His tag antibody. Briefly, the solubilized microsomal fraction was incubated with 5-10 g/ml of the antibody for 1 h and then with Protein G-Sepharose 4 Fast Flow (Amersham Biosciences) for 1 h. The immune complex was washed four times with 0.5% Triton X-100, phosphate-buffered saline. The resulting enzyme captured by Ni-NTA beads or immunoprecipitated was subjected to Western blot analysis and a GnT assay.
Western Blot Analysis-The captured His 6 -tagged proteins were separated by SDS-PAGE (10% Laemmli's gel) and transferred to a nitrocellulose membrane (Schleicher & Schuell). The membrane was probed with an anti-His tag antibody and then incubated with horseradish peroxidase-conjugated sheep anti-mouse IgG. The reactive proteins were visualized by means of an ECL kit (Amersham Biosciences).
GnT Assay-GnT activity was assayed using pyridylaminated acceptor substrates under conditions described previously (20) with minor modifications. The sample was incubated at 37°C for 4 h with 20 M pyridylaminated acceptor substrate (GnGn-bi-PA or Gn 3 -triЈ-PA) and 40 mM UDP-GlcNAc in 125 mM MES (pH 6.25) or MOPS (pH 7.5) containing 200 mM GlcNAc, 0.5% Triton X-100, and 10 mM EDTA. The reaction was terminated by boiling for 3 min and then centrifuged at 15,000 rpm for 5 min. The resulting supernatant was injected to a TSKgel ODS-80TM column (4.6 ϫ 250 mm; TOSOH) equipped with a Shimadzu LC-VP HPLC system. Product was eluted with 0.1 M ammonium acetate (pH 4.0) with or without 0.05% n-butyl alcohol at a flow rate of 1.0 ml/min and monitored with a fluorescence spectrophotometer (excitation, 320 nm; emission, 400 nm).
MALDI-TOF MS Analysis-MALDI-TOF MS was performed with a Perseptive Biosystems Voyager RP-DE instrument. The mass spectra were acquired in the reflectron mode under a 20-kV accelerating voltage with positive detection. 2,5-dihydroxybenzoic acid (10 mg/ml) was used as the matrix.
NMR Analysis-For preparation of the NMR sample, a large scale reaction was carried out with 500 M GnGn-bi-PA as a substrate in the presence of 0.5 M glycine. After a 3-day incubation at 37°C, the reaction product was isolated using a TSKgel ODS-80TM column (7.8 ϫ 300 mm; TOSOH) as described above. The separated sugar chains were lyophilized, dissolved in 99.9% D 2 O, and then lyophilized from D 2 O twice and dissolved in this solvent. Proton NMR measurements were carried out with a Varian Unity-400 spectrometer at 400 MHz at 30°C. Chemical shifts are expressed as ppm relative to an external standard of 3-(trimethylsilyl)propionic acid-d 4 .
Northern Blot and Dot Blot Analyses-Human 12-lane multiple tissue Northern blot and human multiple tissue expression array (Clontech) were used for hybridization according to the manufacturer's protocol. Probes were prepared from the cDNA fragments from a PvuII digest (corresponding to nucleotides 1564 -2286) of pSVK3-hGnT-IX, pSVK3 vector (Amersham Biosciences) containing human GnT-IX cDNA, and an EcoRV digest (corresponding to nucleotides 406 -1306) of pSVK3-hGnT-V and then labeled with [␣-32 P]dCTP using a Megaprime DNA labeling system (Amersham Biosciences).

RESULTS
By performing a BLAST search with the amino acid sequence of human GnT-V as a query, we were able to find a human hypothetical protein that has a significant similarity to the catalytic domain of GnT-V (14), but lacking most of the deduced N-terminal region. Using the amino acid sequence of the hypothetical protein in an EST data base search, we found seven clones containing partial sequences corresponding to the hypothetical protein. From the consensus sequence of the seven EST clones, we designed oligonucleotide primers for reverse transcription-PCR using total RNA from various human tissues. Since the PCR products were strongly detected in the brain and the clone of the hypothetical protein was derived from adult brain, we used total RNA from human brain for the subsequent cDNA cloning. The 5Ј-end of the cDNA was determined by a combination of RACE and a further data base search. The resulting full-length cDNA (termed GnT-IX) and deduced amino acid sequences are shown in Fig. 1A. The deduced protein is a type II membrane protein composed of 792 amino acid residues with a calculated molecular mass of 89,531 Da and has eight potential N-glycosylation sites. It appears to consist of an N-terminal cytoplasmic tail, a transmembrane domain, a stem region, and a C-terminal catalytic domain, and the entire sequence identity to human GnT-V is 42%. Comparing the amino acid sequences of human GnT-V and GnT-IX with C. elegans ortholog gly-2 (15) using BLAST 2 sequences (21), the sequences of human GnT-V and GnT-IX exhibit 36 and 32% identities, respectively, to the GLY-2 sequence. The human GnT-V gene (Mgat5) is mapped to chromosome 2q21 (14), whereas the GnT-IX gene is mapped to 17q25.
Initially, we attempted to detect GnT activity using a whole cell lysate or a microsomal membrane fraction prepared from GnT-IX-transfected cells. Although we tested several cell lines, endogenous GnT-V activity was detected, and no other GnT activity was found under the conditions used in the tests. To concentrate the recombinant enzyme and to eliminate endogenous GnT-V activity, full-length human GnT-IX containing a His 6 tag was expressed in Neuro-2a cells, captured by Ni-NTA beads, or immunoprecipitated with an anti-His tag antibody and Protein G beads, and the beads were then directly used for a subsequent GnT assay. Western blot analysis of the captured enzyme showed a 110-kDa band and a smear band at a higher molecular weight (Fig. 2). The captured tagged protein was directly subjected to GnT assay under conditions typical for GnT-V activity with GnGn-bi-PA as an acceptor substrate. When the reaction mixture was subjected to HPLC, the major product (designated as P1) of GnT-IX was eluted (peak 1 in Fig.  3A) with the same retention time as the GnT-V control (Fig.  3C). In addition, two more peaks (peaks 2 and 3) (designated as P2 and P3) that have never been seen in parental cells and GnT-V control cells were observed. These peaks were collected and subjected to MALDI-TOF MS analysis (Fig. 4). The spectrum of P1 showed an m/z value for (M ϩ Na) ϩ of 1621.7 (Fig.  4B), corresponding to that of Gn 3 -triЈ-PA (refer to Fig. 6 for structure), 1621.3 (Fig. 4A). This result and the retention time on HPLC suggest that the structure of P1 could be same as the product of GnT-V. The spectrum of P3 (Fig. 4D) also showed nearly the same m/z value (1620.6) as Gn 3 -triЈ-PA. On the other   Table I. Ac indicates a signal due to ammonium acetate used as elution buffer for HPLC. hand, the m/z value of P2 was 1824.0 (Fig. 4C), indicating that two GlcNAc residues had been transferred to the acceptor substrate. Furthermore, when Gn 3 -triЈ-PA was used as a substrate, a product peak (peak 4 in Fig. 3D) appeared with the same retention time as peak 2 in Fig. 3A. Thus, GnT-IX transfers GlcNAc not only to the core ␣1,6-mannose arm via a ␤1,6-linkage, typical of GnT-V activity, but also to another position and/or via a distinct linkage.
Using recombinant soluble GnT-IX, the pH optimum of the GnT activity of GnT-IX was examined in a reaction mixture composed of 200 mM each buffer, 40 mM UDP-GlcNAc, 200 mM GlcNAc, 10 mM EDTA, 20 M GnGn-bi-PA. The optimal pH was found to be between 7.5 and 8.5 in MOPS or HEPES buffer, whereas that of GnT-V was around pH 6.5 (data not shown). The effects of divalent cations on GnT activity were also examined in the same reaction mixture with 10 mM each of metal chlorides or EDTA. Metal ions such as Mg 2ϩ , Ca 2ϩ , and Mn 2ϩ had no effect on the reaction. GnT activity is strongly inhibited by Ni 2ϩ , Cu 2ϩ , and Zn 2ϩ . EDTA had no effect on the activity as for GnT-V and other ␤1,6GnTs. The activity of GnT-IX was enhanced by the presence of glycine. In the presence of 0.5 M glycine at pH 7.5, the yield of enzymatic products were in-creased about 30% after a 2-h incubation (data not shown). Therefore, we carried out a large scale reaction under this condition to prepare samples for NMR analyses. After a 3-day incubation, most of the substrate was converted to the enzymatic products: 45, 39, and 7% of biantennary substrate was converted to P1, P2, and P3, respectively. Therefore, P1 and P2 were collected and subjected to NMR analyses.
NMR analyses were carried out to confirm of the structures of the enzymatic products P1 and P2. Proton NMR spectra of the substrate and the two products are shown in Fig. 5. The chemical shift values of their anomeric proton signals and methyl proton signals are summarized in Table I. The values for P1 were identical with those of Gn 3 -triЈ-PA (22). In the spectrum of the P2, a methyl proton signal of an additional GlcNAc appeared at 2.044 ppm. This chemical shift value suggests that the GlcNAc binds to the -CH 2 OH group at the 6-position. The anomeric proton signal of Man 4 (assigned at the Gn 4 -tetraЈ-PA structure in Fig. 6) was observed at 5.086 ppm and showed a shift to higher field by 0.50 ppm relative to that of Gn 3 -triЈ-PA (spectra B and C in Fig. 5). When GlcNAc was attached to -CH 2 OH at the 6-position of Man 4Ј, the anomeric proton signal of Man 4Ј shifted to higher field by 0.49 ppm (spectra A and B in Fig. 5). These results indicate that the additional GlcNAc of the P2 is linked to -CH 2 OH at the 6-position of Man 4 of the Gn 3 -triЈ-PA (P1). The anomeric proton of the additional GlcNAc (GlcNAc 7Љ) gave a doublet signal with a coupling constant of 8.0 Hz at 4.561 ppm. This coupling constant value shows that the linkage of the GlcNAc is a ␤-form.
Northern blot analysis showed that the GnT-IX transcript of ϳ5 kb is exclusively expressed in the brain, whereas the GnT-V transcript of ϳ9 kb is expressed in various tissues (Fig. 7). Dot blot analysis also showed that GnT-IX is specifically expressed in all areas of the adult brain as well as the fetal brain (Fig. 8). DISCUSSION Thus far, no homologous gene of GnT-V in the genome in any mammalian species has been reported. Recently, a functional GnT-V ortholog was found in C. elegans, and the gene was the sole GnT-V homolog present in its genome (15). Here we report a novel homologous gene of human GnT-V, designated as GnT-IX, which possesses novel ␤1,6GnT activity.
The domain architecture and the positions of the cysteine residues of mammalian and nematode GnT-Vs are well conserved in GnT-IX. In addition, Leu 189 in the stem region of human GnT-V, corresponding to the position at a L188R point mutation in Lec4A Chinese hamster ovary mutant cells (23), is also conserved in GnT-IX (Leu 201 ). It has previously been reported that the leucine residue is responsible for the Golgi localization of hamster and nematode GnT-Vs (15,23). Therefore, GnT-IX is structurally related to GnT-V and may also be localized in the Golgi apparatus. Compared with GnT-V, GnT-IX has an extra 33 amino acid residues inserted in the C-terminal catalytic domain (Fig. 1B). Since two potential Nglycosylation sites exist in this short sequence, it may be exposed to the molecular surface. Whether this sequence is involved in the difference of substrate specificity between GnT-IX and GnT-V remains to be investigated.
Recombinant human GnT-IX actually exhibited GnT-V activity, catalyzing the transfer of GlcNAc to the 6-OH position of the ␣1,6-linked mannose arm of GnGn-bi-PA and forming Gn 3 -triЈ-PA (Fig. 6). This activity of GnT-IX was considerably lower than that of GnT-V when a similar amount of expression was observed as judged by reactivity to an anti-His tag antibody (data not shown). Interestingly, GnT-IX also acted on the ␣1,3linked mannose arm in the acceptor substrate. Therefore, GnT-IX is able to catalyze the transfer of GlcNAc to the 6-OH position of mannose in the sequence GlcNAc␤1,2-Man␣1, which is present in both ␣-linked mannose arms of the N-glycan, indicating that GnT-IX is different and distinct from GnT-V. GnT-IX preferentially transferred GlcNAc to the ␣1,6-linked mannose arm rather than the ␣1,3-linked mannose arm (peaks 1 and 3 in Fig. 3A), suggesting that the 6-OH position of ␣1,6-linked mannose is more accessible to GnT-IX than that of the ␣1,3-linked mannose.
The expression of the GnT-IX gene was found to be exclusively in the brain, whereas the GnT-V gene is ubiquitously expressed. Similar results were also observed in fetal tissues, suggesting that GnT-IX may play important roles in fetal and adult brains. In addition, although Mgat5 Ϫ/Ϫ mice lacked most detectable GnT-V enzyme activity and P. vulgaris leukoagglutinin reactivity, brain development and function in these mice appeared to be normal (12). Since the poly-N-acetyllactosamine extension is dependent on ␤1,6-branch formation by GnT-V, ␤1,6-branch formation is thought to be important in the process of nerve differentiation via the interaction of poly-N-acetyllactosamine chains and their receptors (4,5). GnT-IX may compensate for the deficiency in Mgat5 Ϫ/Ϫ brain with regard to neuronal development and functions. Fig. 6 summarizes the predicted synthetic pathway of the core structure in N-glycan catalyzed by GnT-IX. GnT-IX catalyzes the synthesis of two unique N-linked oligosaccharide structures, namely the triantennary oligosaccharide possessing a ␤1,6-branched GlcNAc on the ␣1,3-linked mannose arm (Gn 3 -triЉ-PA, peak 3) and the tetraantennary oligosaccharide containing two ␤1,6-branched GlcNAc residues in the core mannose (Gn 4 -tetraЈ-PA, peak 2). These structures have never been reported in any mammalian tissues. For further progress, it will be necessary to determine whether these structures actually exist in the brain.