A Gene of the β3-Glycosyltransferase Family Encodes N-Acetylglucosaminyltransferase II Function in Trypanosoma brucei*

The bloodstream form of the human pathogen Trypanosoma brucei expresses oligomannose, paucimannose, and complex N-linked glycans, including some exceptionally large poly-N-acetyllactosamine-containing structures. Despite the presence of complex N-glycans in this organism, no homologues of the canonical N-acetylglucosaminyltransferase I or II genes can be found in the T. brucei genome. These genes encode the activities that initiate the elaboration of the Manα1–3 and Manα1–6 arms, respectively, of the conserved trimannosyl-N-acetylchitobiosyl core of N-linked glycans. Previously, we identified a highly divergent T. brucei N-acetylglucosaminyltransferase I (TbGnTI) among a set of putative T. brucei glycosyltransferase genes belonging to the β3-glycosyltransferase superfamily (Damerow, M., Rodrigues, J. A., Wu, D., Güther, M. L., Mehlert, A., and Ferguson, M. A. (2014) J. Biol. Chem. 289, 9328–9339). Here, we demonstrate that TbGT15, another member of the same β3-glycosyltransferase family, encodes an equally divergent N-acetylglucosaminyltransferase II (TbGnTII) activity. In contrast to multicellular organisms, where GnTII activity is essential, TbGnTII null mutants of T. brucei grow in culture and are still infectious to animals. Characterization of the large poly-N-acetyllactosamine containing N-glycans of the TbGnTII null mutants by methylation linkage analysis suggests that, in wild-type parasites, the Manα1–6 arm of the conserved trimannosyl core may carry predominantly linear poly-N-acetyllactosamine chains, whereas the Manα1–3 arm may carry predominantly branched poly-N-acetyllactosamine chains. These results provide further detail on the structure and biosynthesis of complex N-glycans in an important human pathogen and provide a second example of the adaptation by trypanosomes of β3-glycosyltransferase family members to catalyze β1–2 glycosidic linkages.

The African trypanosomes are protozoan parasites that cause human African sleeping sickness and Nagana in cattle. The parasite undergoes a complex life cycle between the mammalian host and the blood-feeding tsetse fly vector (Glossina sp.). Throughout this life cycle, Trypanosoma brucei is coated by glycosylphosphatidylinositol (GPI) 8 -anchored proteins. The bloodstream form of the parasite in the mammalian host is covered by a coat of 5 ϫ 10 6 variant surface glycoprotein (VSGs) homodimers and evades the immune system by replacing one VSG coat by another, in a process known as antigenic variation (1)(2)(3)(4). The VSG GPI anchors contain side chains of 0 -6 Gal residues, depending on the VSG variant (5)(6)(7) and between 1 and 3 N-linked glycans. The latter can be of oligomannose, paucimannose, or complex types (6,8,9). T. brucei expresses numerous other GPI-anchored and transmembrane glycoproteins at the cell surface, in the flagellar pocket, and in the intracellular endosomal/lysosomal system, some of which are life cycle stage-specific or display life cycle stage-specific glycosylation differences. For example, the transmembrane invariant surface glycoproteins ISG65 and ISG75 (10) and the GPI-anchored flagellar pocket ESAG6/ESAG7 heterodimeric transferrin receptors (11)(12)(13) are specific to the bloodstream life cycle stage, whereas the major lysosomal glycoprotein p67 is common to bloodstream and procyclic stages but contains complex N-glycans only in the bloodstream stage (14). This control of stage-specific glycosylation resides primarily at the level of oligosaccharyltransferase expression (15). Thus, in the bloodstream form of T. brucei both the TbSTT3A and TbSTT3B genes are expressed, and it appears that TbSTT3A co-translationally scans for glycosylation sequons in relatively acidic local environments, transferring exclusively Man 5 GlcNAc 2 that is destined to be processed to paucimannose or complex N-glycans, whereas TbSTT3B post-translationally modifies any remaining sequons with Man 9 GlcNAc 2 that is destined to be processed no further than Man 5 GlcNAc 2 in the conventional oligomannose series. Conversion from the oligomannose series to the complex series by the conventional mammalian-type route cannot occur because the parasite lacks a Golgi ␣-mannosidase II gene (16). In the procyclic form of T. brucei, the expression of TbSTT3A is repressed at both the mRNA level (15) and protein level (17), favoring the transfer of Man 9 GlcNAc 2 and the predominant expression of the conventional Man 5 GlcNAc 2-Man 9 GlcNAc 2 oligomannose series (18).
The survival strategies of protozoan parasites frequently involve the participation of glycoconjugates. T. brucei expresses many glycoproteins containing Gal and GlcNAc, including glycoproteins with novel bloodstream form-specific giant poly-Nacetyllactosamine (poly-LacNAc) containing N-linked glycans (19). The creation of UDP-glucose 4Ј-epimerase (TbGalE) conditional null mutants showed that this gene, and hence UDP-Gal, is essential for the survival of the parasite in both the bloodstream and procyclic form life stages (20 -22). Similarly, the creation of UDP-GlcNAc pyrophosphorylase (TbUAP) and glucosamine 6-phosphate N-acetyltransferase (TbGNA) conditional null mutants has shown that UDP-GlcNAc is essential for bloodstream form of T. brucei (23,24). From these experiments, it is possible to conclude that one or more of the UDP-Gal-and UDP-GlcNAc-dependent glycosylation pathways are essential to the parasite. This has provided the impetus to identify and characterize the UDP-Gal-and UDP-GlcNAcdependent glycosyltransferase (GT) genes in the parasite. We previously reported a family of 21 genes with predicted amino acid sequences consistent with being UDP-sugar-dependent GTs. All 21 putative T. brucei GT amino acid sequences are similar to those of the mammalian ␤3GT family (25). The mammalian ␤3GT family includes Gal, Glc, glucuronic acid, GlcNAc, and GalNAc ␤-3 transferases, and its members contain N-terminal transmembrane domains followed by three conserved motifs as follows: (I/L)RXXWG, (F/Y)(V/L/M)XXX-DXD, and (E/D)D(A/V)(Y/F)XGX(C/S). The comparable motifs in the T. brucei genes are slightly different, WG, Y(I,V,F)XKXDDD, and ED(A/V/I/L/M)(M/L)X(G/A), but nevertheless, they identify the parasite genes as belonging to the ␤3GT superfamily (26). One of these genes (TbGT8) encodes a ␤1-3 GlcNAc transferase and another (TbGT3) a ␤1-3 Gal transferase that modifies the complex GPI anchor side chains of the procyclins (the major surface glycoproteins of the procyclic life cycle stage) (26 -28). However, we recently reported that another gene (TbGT11) encodes a ␤1-2 GlcNAc transferase that performs a similar role to members of the N-acetylglucosaminyltransferase I family, in that it transfers GlcNAc in ␤1-2 linkage to the 3-arm of Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc (1).
Here, we report that another T. brucei ␤3GT superfamily gene member (TbGT15) encodes another ␤1-2 GlcNAc transferase that was already localized to the Golgi apparatus (29,30). It performs a similar role to members of the N-acetylglucosaminyltransferase II family in that it transfers GlcNAc in ␤1-2 link-age to the 6-arm of Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc, emphasizing the highly divergent nature of the trypanosome genes involved in structurally conserved aspects of complex N-glycan biosynthesis.

Experimental Procedures
Cultivation of Trypanosomes-T. brucei brucei strain 427 bloodstream form parasites, expressing VSG variant 221 and transformed to stably express T7 polymerase and the tetracycline repressor protein under G418 antibiotic selection (31), were used in this study. This genetic background will be referred to as wild-type (WT). Cells were cultivated in HMI-9 medium containing 2.5 g/ml G418 at 37°C in a 5% CO 2 incubator as described previously (31).
DNA and RNA Isolation and Manipulation-Plasmid DNA was purified from Escherichia coli (␣-select chemically competent cells, Bioline, London, UK) using Qiagen Miniprep or Maxiprep kits, as appropriate. Gel extraction and reaction clean up was performed using QIAquick kits (Qiagen). Custom oligonucleotides were obtained from Eurofins MWG Operon or the Dundee University oligonucleotide facility. T. brucei genomic DNA was isolated from ϳ2 ϫ 10 8 bloodstream form cells using DNAzol (Helena Biosciences, UK) by using standard methods. T. brucei mRNA was extracted from 1 ϫ 10 7 cells using RNeasy RNA extraction kit (Qiagen).
Generation of Gene Replacement Constructs-The 517-bp 5Ј and 454-bp 3Ј UTR sequences next to the Tb427.7.300 ORF were PCR-amplified from genomic DNA using Pfu DNA polymerase with primers 5Ј-cgttGTCGACagtatccgcaaaatgcgact-3Ј and 5Ј-gtttaaacttacggaccgtcaagctttatttttctttccctacgcac-3Ј and 5Ј-gacggtccgtaagtttaaacggatccaaagcggaataaaaataaatc-3Ј and 5Ј-ataagtaaGCGGCCGCagatgtcgcgcaagaaaaac-3Ј as forward and reverse primers, respectively. The two PCR products were used together in a further PCR to yield a product containing the 5Ј-UTR linked to the 3Ј-UTR by a short HindIII (underlined), PmeI (italics), and BamHI (underlined) cloning site and NotI and SalI restriction sites at each end (capital letters). The product was cloned between the NotI and SalI sites of the pGEM-5Zf(ϩ) vector (Promega).
The hygromycin phosphotransferase (HYG) and puromycin acetyltransferase (PAC) drug-resistance genes were then introduced into the targeting vector via the HindIII and BamHI cloning sites. For re-expression of Tb427.7.300, the ORF was PCR-amplified from genomic DNA with the primer pair 5Јagaaagcttatggtgtggagtgggcataaa-3Ј and 5Ј-ttcagatcttcatgtgcacgaggcgtgcca-3Ј and cloned into pLEW100-Phleo (31).
For overexpression of full-length TbGT15 with a C-terminal 3ϫ HA epitope tag, a plasmid was generated based on the trypanosome expression vector pLEW82 (31). TbGT15 ORF was amplified from T. brucei genomic DNA and the primers 5Ј-gactaagcttatggtgtggagtgggcataaatac-3Ј and 5Ј-gactttaattaatgcgtaatcagggacgtcataaggatatgcgtaatcagggacgtcataaggatacgctc-ccgcTGTGCACGAGGCGTGCCATC-3Ј containing a HindIII and PacI restriction site (underlined), respectively. The sequence encoding for two HA tags (italics) followed by a sequence encoding an Ala-Gly-Ala linker was attached as a 5Ј-overhang of the reverse primer. The PCR product was subcloned into pLEW82-GPIdeAc-HA (32) via HindIII and PacI restriction sites under replacement of the GPIdeAc insert, but retention of the sequence encoding for one HA tag, resulting in the plasmid pLEW82-TbGT15-HA 3 . The identity of all constructs was confirmed by sequencing.
Transformation of Bloodstream Form T. brucei-Constructs for gene replacement and ectopic expression were purified, digested with NotI to linearize, precipitated, washed with 70% ethanol, and re-dissolved in sterile water. The linearized DNA was electroporated into T. brucei bloodstream form cells (strain 427, variant 221) that were stably transformed to express T7 RNA polymerase and the tetracycline repressor protein under G418 selection. Cell culture and transformation were carried out as described previously (31)(32)(33).
Southern Blotting-Aliquots of genomic DNA isolated from 100 ml of bloodstream form T. brucei cultures (ϳ2 ϫ 10 8 cells) were digested with EcoRI, resolved on a 0.8% agarose gel and transferred onto a Hybond-N positively charged membrane (GE Healthcare, UK). Highly sensitive DNA probes labeled with digoxigenin-dUTP were generated using the PCR digoxigenin probe synthesis kit (Roche Applied Science) according to the manufacturer's recommendations and hybridized overnight at 42°C. Detection was performed using alkaline phosphataseconjugated anti-digoxigenin Fab fragments and the chemiluminescent substrate CSPD (Roche Applied Science).
Mouse Infectivity Studies-Wild-type and TbGT15 null mutant bloodstream form trypanosomes were grown in HMI-9T media, washed in media without antibiotics, and resuspended at 5 ϫ 10 6 cells/ml. Groups of five female BALB/c mice were used for each cell line, and 0.1 ml of the suspension above was injected intraperitoneally per animal. Infections were assessed 3 days post-infection by tail bleeding and cell counting using a Neubauer chamber in a phase contrast microscope.
Small Scale sVSG Isolation-Soluble form VSG (sVSG) was isolated from 100 ml of cultures containing ϳ2 ϫ 10 8 bloodstream form T. brucei by a modification of the method of Cross and co-workers (34, 35) as described previously (36). Briefly, cells were chilled on ice, centrifuged at 2500 ϫ g for 10 min, and washed in an isotonic buffer. The pellet was resuspended in 300 l of lysis buffer (10 mM NaH 2 PO 4 buffer, pH 8.0, containing 0.1 mM tosyllysine chloromethyl ketone hydrochloride (TLCK), 1 g/ml leupeptin, and 1 g/ml aprotinin) and incubated for 5 min at 37°C. The sample was centrifuged at 14,000 ϫ g for 5 min, and the supernatant was applied to a 200-l DE52 anion exchange column pre-equilibrated in 10 mM sodium phosphate buffer, pH 8.0. Elution was performed with 0.8 ml of 10 mM sodium phosphate buffer, pH 8.0, and the eluate was concentrated and diafiltered with water on a YM-10 spin concentrator (Microcon). The final sample of 50 -100 g of sVSG221 was recovered in a volume of 100 l of water.
ES-MS Analysis of Intact sVSG-50 g of aliquots of sVSG preparations were diluted to ϳ0.05 g/l in 50% methanol, 1% formic acid and analyzed by positive ion ES-MS on a Q-Tof 6520 instrument (Agilent). Data were collected, averaged, and processed using the maximum entropy algorithm of the Mass-Hunter software (Agilent).
Lectin Blotting of Cell Extracts-To analyze N-glycosylation of T. brucei bloodstream form cells, ϳ2 ϫ 10 8 cells were first depleted of VSG by hypotonic lysis (34,35). For Western blot analysis, residual cell ghosts were solubilized in SDS sample buffer containing 8 M urea, boiled with DTT, separated by SDS-PAGE (ϳ1 ϫ 10 7 cell eq/lane) on NuPAGE bis-Tris 4 -12% gradient acrylamide gels (Invitrogen) and transferred to a nitrocellulose membrane (Invitrogen). Ponceau S staining confirmed equal loading and transfer. Glycoproteins were probed with 1.7 g/ml biotin-conjugated ricin (RCA-120, Vector Laboratories, UK) in PBS before or after pre-incubation with 10 mg/ml D-galactose and 10 mg/ml ␣-lactose to confirm specific ricin binding. Detection was performed using IRDye 680LTconjugated streptavidin and the LI-COR Odyssey infrared imaging system (LICOR Biosciences, Lincoln, NE).
For mass spectrometric analysis of the reaction product, the assay was performed using 5 mM non-radioactive UDP-GlcNAc. Samples were analyzed by LC-MS using a HILIC column (Tosoh TSKgel Amide column, 1 mm ϫ 10 cm) and a gradient of 80 to 5% acetonitrile in 0.1% formic acid at a flow rate of 50 l/min using a TSQ Quantiva triple-quadrupole mass spectrometer (Thermo Fisher Scientific). For methylation linkage analysis of the product, glycans were converted to constituent monosaccharides in the form of partially methylated alditol acetates and analyzed by GC-MS as described above.
Scanning Electron Microscopy-To analyze bloodstream form cells by scanning electron microscopy, cells were fixed in HMI-9 medium with 2.5% glutaraldehyde. They were further processed and examined in a Philips XL30 ESEM operating at an accelerating voltage of 15 kV by the Centre for High Resolution Imaging and Processing (CHIPS) at the University of Dundee.

Results
Analysis of the TbGT15 Gene Product-We previously characterized the biological function of three members of a family of putative UDP-sugar-dependent GTs (1,26,27). In this study, Tb927.7.300 was selected for functional analysis. The gene encodes for a 367-amino acid protein with a theoretical molecular mass of 43.1 kDa. Stable isotope labeling with amino acids in cell culture-based quantitative proteomic data demonstrated that the protein expression level is 15 times higher in bloodstream form parasites compared with procyclic form parasites (40).
The T. brucei strain that was used in this study (Lister strain 427) differs from the one that was used for the reference genome sequencing project (TREU927). However, an alignment of Tb927.7.300 and its homologue Tb427.7.300 revealed a very high similarity with only three single nucleotide polymorphisms, none of them resulting in amino acid changes. The strain 427 gene and protein product will be referred to here as TbGT15 and TbGT15, respectively.
The protein sequence contains several hallmarks of Golgi apparatus glycosyltransferases. First, a membrane protein topology prediction program based on a hidden Markov model (41) designates TbGT15 as a type II transmembrane protein. In addition, the sequence contains a DXD motif, which is generally involved in catalytic activity of known GTs (42) as well as a dibasic motif, which functions as an endoplasmic reticulum exit signal (43). Indeed, a subcellular Golgi localization of TbGT15 was confirmed previously (29,30).
Creation of Bloodstream Form TbGT15 Null and Conditional Null Mutants-As TbGT15 is predominantly expressed in bloodstream form parasites (40), we decided to investigate the protein function by creating null and conditional null mutants in this life cycle stage. BLAST search of the T. brucei genome suggested that TbGT15 is present as a single copy per haploid genome. Both alleles were sequentially replaced by homologous recombination using PAC and HYG drug resistance cassettes as summarized in Fig. 1A. After selection on the respective antibiotics, the generation of a TbGT15 null mutant (⌬TbGT15:: PAC/⌬TbGT15::HYG) was confirmed by Southern blot using probes for the TbGT15 ORF and 3Ј UTR (Fig. 1B). To allow for a tetracycline-inducible re-expression of the gene, an ectopic copy of TbGT15 was introduced into the rRNA locus of the null mutant background using the pLEW100 vector (31). Clones were selected on phleomycin, and the creation of this conditional null mutant (⌬TbGT15 Ti /⌬TbGT15::PAC/⌬TbGT15:: HYG) was confirmed by RT-PCR (Fig. 1C).
No morphological differences between the WT and TbGT15 null mutant parasites could be ascertained by light microscopy or by scanning electron microscopy ( Fig. 2A). Compared with WT cells, the TbGT15 null mutant parasites exhibited slightly slower growth kinetics in vitro, and this mild growth phenotype was partially reversed in TbGT15 conditional null cells grown under permissive conditions (Fig. 2B). In addition, no difference in its ability to infect mice could be detected for the TbGT15 null mutant (Fig. 2C). From this we can conclude that TbGT15 is a non-essential gene in T. brucei bloodstream form cells.
To assess differences in the glycosylation phenotype, intact sVSGs from WT and TbGT15 null mutant parasites were analyzed by ES-MS in positive-ion mode (Fig. 3). VSG molecules containing a total of four or five GlcNAc residues were present at similar levels in both genotypes, but glycoforms with six GlcNAc residues were completely absent in the TbGT15 null mutant (see arrows in Fig. 3B and Table 1). Bearing in mind that four GlcNAc residues are necessary for the composition of the two N-glycan N-acetylchitobiose core structures, the lack of VSG glycoforms containing six GlcNAc residues strongly indi-  1, 3, and 5) and TbGT15 null mutant cells (lanes 2, 4, and 6). The blot was probed with a TbGT15 ORF probe (left panel) and a TbGT15 3ЈUTR probe (middle panel) and shows the replacement of both alleles with drug resistance cassettes. Equal loading was verified by ethidium bromide staining (right panel). C, ethidium bromide-stained agarose gel of reverse transcription-PCR products from RNA extracted from WT cell, TbGT15 null, and conditional null mutants.  cates a deficiency in biantennary complex N-glycans at Asn-296. From this we can conclude that the mutant cells are unable to express complex N-glycans and that TbGT15 is involved in their biosynthesis.
In Vitro Functional Activity Assay-To verify that TbGT15 encodes a glycosyltransferase that is directly involved in the biosynthesis of hybrid or complex N-glycans, we performed an in vitro assay for enzymatic activity as described previously (1). Briefly, full-length TbGT15 fused to a C-terminal 3ϫ HA epitope tag was expressed in T. brucei bloodstream form cells and immunoprecipitated using anti-HA magnetic beads. The protein was incubated with Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc as acceptor substrate and tritiumlabeled UDP-[ 3 H]GlcNAc as donor substrate. Following desalting and removal of excess UDP-[ 3 H]GlcNAc donor by mixed-bed ion exchange, aliquots were separated by thin layer chromatography (TLC) and analyzed by fluorography. Although a control immunoprecipitation with lysate from WT cells did not result in any [ 3 H]GlcNAc incorporation (Fig. 4A,  lane 2), the sample containing TbGT15-HA 3 showed a strong signal of tritium-labeled reaction product (Fig. 4A, lane 1). This demonstrates that TbGT15 is able to transfer GlcNAc to biantennary Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc core structures. It is noteworthy that the shortened substrate Man␣1-6(Man␣1-3)Man lacking the chitobiose core was not used as an UDP-GlcNAc acceptor by TbG15 (Fig. 4A, lane 3).
To reveal whether GlcNAc is transferred to the Man␣1-3 or Man␣1-6 arm of Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc, the reaction product was treated with ␣1-2,3-mannosidase. As seen in the subsequent TLC/autofluorography analysis, mannosidase treatment caused an increased mobility of the reaction product (compare R f values in Fig. 4B). This suggests that the 3-Man arm was not modified by TbGT15, leaving it susceptible to exoglycosidase cleavage, and allows us to conclude that the transferred GlcNAc residue is attached to the 6-Man arm of the Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc core structure. The anomeric configuration of the newly formed linkage was determined by ␤-Nacetylglucosaminidase digestion of the reaction product. In the following TLC/autofluorography analysis, the band of tritium-labeled GlcNAcMan 3 GlcNAc 2 disappeared, although the amount of free [ 3 H]GlcNAc increased, demonstrating a ␤-configuration (Fig. 4C).
For further characterization of the reaction product, the assay was performed using non-radioactive UDP-GlcNAc. First, HILIC-MS was performed to identify the HexNAc 3 Hex 3 reaction product (Fig. 5, A and B). A subsequent methylation linkage analysis on the reaction sample by GC-MS demonstrated the presence of 1,2,5-tri-O-acetyl-(1-deutero)-3,4,6-tri-O-methyl-mannitol, originating from 2-O-substitued mannose, which reveals that TbGT15 transfers GlcNAc in a 1-2 linkage to one of the non-reducing terminal mannose residues of the trimannosyl core (Fig. 5, C and D).
N-Glycosylation Phenotype of Bloodstream Form TbGT15 Mutant Parasites-To investigate the effect of TbGT15 on the glycosylation of other proteins than VSG, total glycoproteins were extracted with SDS/urea from VSG-depleted trypanosome ghosts and analyzed by lectin blotting. As reported previously for WT T. brucei, ricin (RCA-120), a lectin that predominantly binds to non-reducing terminal ␤-galactose residues, showed strong binding to a series of glycoproteins running between 60 and 150 kDa (Fig. 6, 1st lane). Ricin binding to glycoproteins extracted from the TbGT15 null mutants was slightly reduced, and the apparent molecular mass of all signals was marginally smaller compared with WT signals (Fig. 6, 2nd  lane). Albeit subtle, these changes in blotting pattern suggest an alteration in the synthesis of the large poly-LacNAc-containing glycans of the high molecular weight invariant glycoproteins (19).
To draw structural conclusions, we decided to analyze the ricin-binding glycoprotein fraction by methylation linkage as described previously (19). Briefly, WT and TbGT15 null mutant cells were isolated, depleted of VSGs, solubilized in SDS/urea, and glycoproteins were purified by ricin affinity chromatography. N-Linked glycans were released by peptide:N-glycosidase F and further fractionated by Bio-Gel P-4 gel filtration, resulting in two main fractions as follows: one containing the relatively small mannose-rich N-glycans, and a Gal/GlcNAc-rich high molecular mass fraction eluting at the void volume of the Bio-Gel P-4 column (the total poly-LacNAc fraction). Aliquots of these fractions were subjected to neutral monosaccharide composition analysis by GC-MS. The molar ratios of Gal/Man  Table 1). Significant differences in the sVSG glycosylation patterns are indicated by arrows in B.
in the total poly-LacNAc fraction for WT (19) and TbGT15 null mutant parasites were found to be similar, 14.5:1 and 12.4:1, respectively.
Subsequent GC-MS methylation linkage analysis of the total poly-LacNAc fraction revealed structural similarities but also some quantitative differences (Table 2). Thus, both WT and

Isobaric glycoforms of sVSG221 identified by ES-MS
The molecular masses of different glycoforms of sVSG221 were calculated according to the indicated compositions (the theoretical mass of the assigned VSG composition is shown in parentheses). The relative abundances of those glycoforms observed in Fig. 3 for sVSG preparations from sVSG of WT cells and TbGT15 null mutant cells are indicated by Ϫ, trace, ϩ, ϩϩ, and ϩϩϩ scores. NA means not applicable.

Mass (in Da) WT/TbGT15 null mutant (theoretical) Protein a GlcN-Ino-cP b EtNP GlcNAc
Man ؉Gal WT Traces Ϫ a Protein mass is based on the amino acid sequences of the VSG221 precursor (accession no. P26332) minus residues 1-27 (signal peptide) and 460 -476 (GPI attachment signal peptide) and allows for four disulfide bonds (mass ϭ 46,284 Da). b Components specific to the GPI anchor and common to all glycoforms: GlcN-Ino-cP, glucosamine-␣1-6-myo-inositol-1,2 cyclic phosphate; EtNP, ethanolamine phosphate. c Maximum entropy deconvolved spectra are only semi-quantitative; an indication of relative abundance of the isobaric glycoforms is given based on peak height.  repeats. However, although the numbers of terminal Gal residues were similar, there was a significant decrease in 6-Osubstituted Gal and 3-O-substituted Gal residues and a concomitant increase in 3,6-di-O-substituted-Gal residues in the TbGT15 null mutant glycans. Because TbGT15 has a GnTII-type activity that initiates elaboration of the Man␣1-6 arm of Man␣1-6(Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc core, these data are consistent with a model where linear poly-LacNAc chains predominate on the Man␣1-6 arm, and the Man␣1-3 arm is occupied by both linear and branched poly-LacNAc units (Fig. 7). The increase in the number of 3,6di-O-substituted Gal residues in the TbGT15 mutant glycans further suggests that deletion of the entire (predominantly linear) poly-LacNAc chain attached to the Man␣1-6 arm is, to some extent, compensated by further elaboration of the (linear and branched) poly-LacNAc chain attached to the Man␣1-6 arm.

Discussion
We have cloned the gene that encodes UDP-GlcNAc:␣1-6-D-mannoside-␤1-2-N-acetylglucosaminyltransferase II in T. brucei, an enzyme that catalyzes an essential step on the route to complex N-glycans. In human patients with carbohydrate-deficient glycoprotein syndrome type II, congenital deficiency in GnTII activity is associated with severe psychomotor retardation and other multisystemic abnormalities (45,46). In a mouse model with a homozygous null mutation in the gene encoding GnTII (Mgat2 Ϫ/Ϫ ), 99% of newborns die during the first postnatal week (47). These developmental defects highlight the importance of complex N-glycans in intercellular communication and signaling in multicellular organisms.
The significance of complex N-glycans in the unicellular protozoan T. brucei is less well understood. In the bloodstream form, the parasite expresses both conventional biantennary complex N-glycans and unique highly extended and branched poly-N-acetyllactosamine-containing complex N-glycans (8,19,48,49). However, the T. brucei genome contains no obvious homologues of the canonical GnTI and GnTII genes that encode the ␤1-2GlcNAc transferases usually responsible for the initiation of complex N-glycans. In a previous study, we identified and characterized TbGnTI, the enzyme responsible for the transfer ␤1-2GlcNAc to the Man␣1-3 arm of N-glycan core structures (1). TbGnTI showed unusual activity in that it acts on biantennary Man 3 GlcNAc 2 instead of triantennary Man 5 GlcNAc 2 , the preferred acceptor substrate for vertebrate GnTI activities (50). Remarkably, the TbGnTI gene is highly divergent from the canonical GnTI gene family and, despite the fact that TbGnTI catalyzes a ␤1-2 linkage, it belongs to the so-called ␤3-glycosyltransferase superfamily (1,26).
In this study, a reverse-genetics approach in T. brucei bloodstream form cells indicated that the deletion of TbGT15 (another trypanosome ␤3-glycosyltransferase superfamily member) is accompanied by the absence of complex N-glycans, as well as alterations in the biosynthesis of the giant poly-Lac-NAc-containing glycans. Using a direct enzymatic assay and comprehensive product analysis, we could show that purified TbGT15 catalyzes the conversion of Man␣1-6(Man␣1-3) Man␤1-4GlcNAc␤1-4GlcNAc to GlcNAc␤1-2Man␣1-6 (Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc. We have therefore renamed TbGT15 to TbGnTII. The conversion of Man 3 GlcNAc 2 demonstrates that TbGnTII works indepen-

TABLE 2 Quantitative GC-MS methylation linkage analysis of the total poly-LacNAc fraction
The total poly-LacNAc fraction was permethylated, hydrolyzed, reduced, and acetylated for GC-MS analysis. The resulting PMAA derivatives were identified by retention time and electron impact mass spectra. Quantification was accomplished by integration of the total ion current using molar relative response factors deduced empirically from authentic standards, as described under "Experimental Procedures." dently from prior TbGnTI action, which is in contrast to canonical GnTIIs that only use substrates after modification by GnTI, i.e. Man␣1-6(GlcNAc␤1-2Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc (51,52). This unusual acceptor specificity of the trypanosome enzyme was already indicated by previous data, which show the presence of "pseudohybrid" N-glycans in the absence of TbGnTI (1) and highlights the divergent nature of TbGnTII. A phylogram based on a multiple sequence alignment of TbGT15 and GnTII enzymes of other species is shown in Fig. 8. The GnTIIs of multicellular organisms belong the CAZy (carbohydrate-active enzymes) GT family 13 (39) and are all closely related, with nearly 90% identity between the human and mouse enzymes. In contrast, TbGnTII is a member of the CAZy GT31 family (26) and shares only 9% identity with the human sequence at the amino acid level. Interestingly, although human GnTI and GnTII proteins have only a low level of amino acid sequence homology between them (22%), the TbGnTI and TbGnTII enzymes share 42% identity. This is consistent with the closer functional similarity of the trypanosome enzymes, both of which work on the same acceptor substrate (Man 3 GlcNAc 2 , although only the latter requires the N-acetylchitobiose core), whereas the canonical GnTI and GnTII enzymes work on triantennary Man 5 GlcNAc 2 and Man␣1-6(GlcNAc␤1-2Man␣1-3)Man␤1-4GlcNAc␤1-4GlcNAc, respectively. Methylation linkage analysis of the poly-LacNAc N-glycans of the TbGT15 null mutant showed a reduction in 6-O-substituted Gal and 3-O-substituted Gal but an increase in 3,6-Osubstituted Gal. This allows us to augment our model of the parasites' complex N-glycans and propose that the Man␣1-6 arm is normally occupied by predominantly linear poly-Lac-NAc repeats and the Man␣1-3 arm by branched as well as linear poly-LacNAc repeats (Fig. 7).

PMAA derivative
T. brucei has an unusual dual N-glycosylation mechanism with two paralogous oligosaccharyltransferases, TbSTT3A and TbSTT3B, that transfer biantennary Man 5 GlcNAc 2 and triantennary Man 9 GlcNAc 2 , respectively, in a site-specific manner (15,16). Because of the absence of Golgi ␣-mannosidase II in the parasite, triantennary structures cannot be processed to complex N-glycans, rendering biantennary Man 5 GlcNAc 2 transferred by STT3A the only route to paucimannose and complex structures. Furthermore, the inability of TbGnTI to act on triantennary Man 5 GlcNAc 2 (1) also means that biantennary Man 5 GlcNAc 2 transferred bySTT3A is the only possible route to pseudohybrid N-glycans (i.e. those with only one arm of the trimannosyl-core modified by GlcNAc Ϯ additional sug-ars). RNAi knockdown of TbSTT3A showed that cells are viable in culture but not in mice (15). Interestingly, the deletion of TbGnTI (TbGT11) has no effect on in vitro growth rate, and the infectivity to mice was indistinguishable from wild type (1). This suggests that the presence of pseudohybrid N-glycans with glycan extensions to the 6-arm alone are sufficient to compensate for the loss of complex N-glycans. Here, the in vitro and in vivo viability of the TbGnTII null mutant shows that the reverse is true, in that the presence of hybrid structures with extensions to the 3-arm alone compensates for the loss of complex N-glycans. However, despite extensive attempts, a double knock-out lacking both TbGnTI and TbGnTII genes could not be generated in our hands, suggesting that extension of one or other of the arms of the N-glycan trimannosyl-core is essential for the growth and infectivity of bloodstream form of T. brucei.