Identification of a New Family of Genes Involved in β-1,2-Mannosylation of Glycans in Pichia pastoris and Candida albicans*

Structural studies of cell wall components of the pathogenic yeast Candida albicans have demonstrated the presence of β-1,2-linked oligomannosides in phosphopeptidomannan and phospholipomannan. During C. albicans infection, β-1,2-oligomannosides play an important role in host/pathogen interactions by acting as adhesins and by interfering with the host immune response. Despite the importance of β-1,2-oligomannosides, the genes responsible for their synthesis have not been identified. The main reason is that the reference species Saccharomyces cerevisiae does not synthesize β-linked mannoses. On the other hand, the presence of β-1,2-oligomannosides has been reported in the cell wall of the more genetically tractable C. albicans relative, P. pastoris. Here we present the identification, cloning, and characterization of a novel family of fungal genes involved in β-mannose transfer. Employing in silico analysis, we identified a family of four related new genes in P. pastoris and subsequently nine homologs in C. albicans. Biochemical, immunological, and structural analyses following deletion of four genes in P. pastoris and deletion of four genes acting specifically on C. albicans mannan demonstrated the involvement of these new genes in β-1,2-oligomannoside synthesis. Phenotypic characterization of the strains deleted in β-mannosyltransferase genes (BMTs) allowed us to describe the stepwise activity of Bmtps and acceptor specificity. For C. albicans, despite structural similarities between mannan and phospholipomannan, phospholipomannan β-mannosylation was not affected by any of the CaBMT1–4 deletions. Surprisingly, depletion in mannan major β-1,2-oligomannoside epitopes had little impact on cell wall surface β-1,2-oligomannoside antigenic expression.

Structural studies of cell wall components of the pathogenic yeast Candida albicans have demonstrated the presence of ␤-1,2-linked oligomannosides in phosphopeptidomannan and phospholipomannan. During C. albicans infection, ␤-1,2-oligomannosides play an important role in host/pathogen interactions by acting as adhesins and by interfering with the host immune response. Despite the importance of ␤-1,2-oligomannosides, the genes responsible for their synthesis have not been identified. The main reason is that the reference species Saccharomyces cerevisiae does not synthesize ␤-linked mannoses. On the other hand, the presence of ␤-1,2-oligomannosides has been reported in the cell wall of the more genetically tractable C. albicans relative, P. pastoris. Here we present the identification, cloning, and characterization of a novel family of fungal genes involved in ␤-mannose transfer. Employing in silico analysis, we identified a family of four related new genes in P. pastoris and subsequently nine homologs in C. albicans. Biochemical, immunological, and structural analyses following deletion of four genes in P. pastoris and deletion of four genes acting specifically on C. albicans mannan demonstrated the involvement of these new genes in ␤-1,2-oligomannoside synthesis. Phenotypic characterization of the strains deleted in ␤-mannosyltransferase genes (BMTs) allowed us to describe the stepwise activity of Bmtps and acceptor specificity. For C. albicans, despite structural similarities between mannan and phospholipomannan, phospholipomannan ␤-mannosylation was not affected by any of the CaBMT1-4 deletions. Surprisingly, depletion in mannan major ␤-1,2-oligomannoside epitopes had little impact on cell wall surface ␤-1,2-oligomannoside antigenic expression.
A number of yeast species have adapted to colonize human tissue, and the adverse effects caused by these pathogenic yeasts have gained importance over recent decades. In particular, the endosaprophytic yeast Candida albicans is able to invade human tissues in immunosuppressed patients leading to frequent nosocomial systemic infections with a high mortality rate (1). Current research has shown that successful pathogenic adaptation requires the ability to utilize elaborate sensing and regulation pathways, impacting expression of a wide range of virulence factors (2). Among these virulence attributes critical for survival under changing environmental conditions is the yeast cell wall containing large amounts of carbohydrates and carbohydrates covalently linked to a noncarbohydrate moiety classified as glycoconjugates, either glycoproteins or glycolipids. In yeasts and other eukaryotes, glycoproteins typically contain N-and/or O-linked glycans (3,4). Glycans serve diverse and important functions, such as intracellular trafficking of glycoproteins or glycolipids, protein folding, anchoring of macromolecules to the cell membrane, signal transduction, cell/cell interactions, as well as determination of glycoprotein half-life (5). In pathogenic microbes, glycans trigger host innate and/or adaptive immunity responses.
Despite similarities in the early steps of processing, the mature structure of glycans differs substantially between yeasts and mammals. For example, mammalian N-linked glycans are typically of the complex type, whereas typical fungal N-glycans are categorized as high mannose (3). Depending on the species, fungal high mannose glycans contain distinctive modifications, such as the addition of mannosyl phosphate (6) and ␤-linked mannose. Suzuki and co-workers (7) was the first to demonstrate and confirm that C. albicans, in contrast to Saccharomyces cerevisiae, harbors ␤-1,2-linked oligomannosides (␤-Mans) 3 . The expression pattern of ␤-Mans in C. albicans is quite complex. ␤-Mans have been found to be associated with the acid-labile and acid-stable part of the cell wall phosphopeptidomannan (PPM) (8) where they correspond, respectively, to antigenic factors 5 and 6 ( Fig. 1B). They are also associated with phospholipomannan (PLM), a surface cell wall glycolipid derived from the mannose-inositol-phosphoceramide biosynthetic pathway (9) (Fig. 1C). Although not yet confirmed by chemical/ structural studies, association of ␤-Mans with several cell wall non-mannan mannoproteins has been suggested repeatedly by the use of anti-␤-Man-specific antibodies (10,11). The mammalian immune system specifically reacts with synthetic ␤-Mans that display a unique spatial conformation (12). ␤-Mans have been established to be potent antigens for the adaptive immune response and to elicit specific infection-protective antibodies (13). In contrast to ␣-linked oligomannosides, which interact with C-type lectins, ␤-Mans specifically bind to galectin-3 (14,15). PLM from C. albicans displays exclusively ␤-Mans (9) and has been shown to stimulate a variety of host signaling pathways (16,17). Finally, it has been demonstrated that oral administration of synthetic ␤-Mans inhibited C. albicans colonization of the intestinal track in newborn mice, whereas synthetic ␣-Man did not (18). This experimental evidence strongly suggests that ␤-Mans contribute in a very specific manner to the virulence of C. albicans.
To further analyze the importance of ␤-Man residues in C. albicans virulence, the construction of mutants partially or fully depleted of these residues is necessary. However, the genes encoding C. albicans ␤-mannosyltransferases (BMTs) have not been identified. Furthermore, S. cerevisiae lacks ␤-Man residues and thus is a poor model organism to elucidate mechanisms of ␤-Man transfer. To date, attempts to decipher the role of ␤-Mans in virulence and cell wall structure of C. albicans have relied on the deletion of genes homologous to S. cerevisiae genes involved in PLM and PPM biosynthesis pathways upstream of actual ␤-mannose transfer (10, 19 -21).
The presence of ␤-Mans is not restricted to C. albicans and has been demonstrated chemically in bacteria (22), Leishmania (23,24), as well as in nonpathogenic yeasts such as Pichia pastoris (25).
Here we describe the identification and characterization of a novel family of genes involved in the transfer of ␤-Mans in P. pastoris and the pathogenic yeast C. albicans. Based on the available genomic sequence of P. pastoris and C. albicans, we identified four novel genes in P. pastoris and nine previously uncharacterized family members in C. albicans. Interestingly, homologs of these genes are not present in S. cerevisiae. A series of deletion strains of P. pastoris and C. albicans revealed different immunochemical phenotypes allowing us to gain insights into the function of these genes, including the impact of deletion on the global cell wall surface expression of ␤-Man epitopes. Moreover, systematic biochemical and structural analyses of these mutants have allowed us to define their role in different aspects of glycosylation, specifically core mannosylation of N-linked glycans in P. pastoris and ␤-mannose transfer onto mannan in P. pastoris and C. albicans.

EXPERIMENTAL PROCEDURES
Strains and Culture Conditions-Escherichia coli strains TOP10 or DH5␣ were used for recombinant DNA work. The yeast strains used in this study are listed in Table 1. Protein expression in P. pastoris cells was carried out as reported previously (26). C. albicans was grown in YPD-Arg-His-Urd medium (1% yeast extract, 2% peptone, 2% dextrose, 20 mg/liter arginine, 20 mg/liter histidine, 20 mg/liter uridine) at 37°C for 16 h. All procedures for manipulating DNA were performed as described previously (27).
Deletion of BMT Genes-The PpBMT deletion alleles were generated by the PCR overlap method (28 -30). In the first PCR 5Ј-and 3Ј-flanking regions of the PpBMT genes (500 -1000 bp in size) and the NAT or G418 resistance marker (31,32) were amplified. Then, in a second reaction all three first round templates were used to generate an overlap product that contained all three fragments as a single linear allele. The final PCR product was then directly employed for transformation. Transformants were selected on YPD medium containing 200 g/ml of G418 or 100 g/ml of nourseothricin. In each case the proper integration of the mutant allele was confirmed by PCR. Each C. albicans gene, listed in Table 2, was deleted sequentially from strain BWP17 by PCR-based gene targeting (33). Two plasmids were used to release two selectable markers, CaARG4 or CaHIS1, by NotI digest. Each marker was amplified by PCR using primers, including the first and the last 100 bp of genespecific sequences. The generated disruption cassettes were used to transform BWP17 by the lithium acetate method as described previously (34). First, the selectable marker ARG4 was used to generate independent heterozygous strains on synthetic dextrose plates supplemented with 20 mg/liter histidine and 20 mg/liter uridine. Correct insertion of the marker was verified by PCR. A second round of transformation was performed to delete the second allele with the HIS1 marker by the same method. Homologous integration was verified by PCR.

Reintroduction of CaBMTs into the Null Strains-Each
CaBMT open reading frame with ϳ500-bp upstream and downstream nucleotides was amplified by PCR using AccuPrime Pfx DNA polymerase (Invitrogen). The amplified fragment was cloned into the pCRII-TOPO vector (Invitrogen). Then the CaBMT gene was excised by digestion with SacI and NotI and ligated into SacI-and NotI-digested CIp10 (35) to obtain the reintegration vector. The Cabmt⌬ null strains were transformed with StuI-or NcoI-digested reintegration vector. Transformants were screened by PCR to check the reintroduction of the CaBMT gene at the RPS10 locus.
Whole Cell Protein Extraction and Western Analyses-P. pastoris cells were grown for 1 day in YPD medium at room temperature, and whole cell proteins were extracted using Y-PER (yeast protein extraction reagent, Pierce). Extracts were adjusted to the same protein concentration and analyzed by SDS-PAGE on 5-20% acrylamide gels. Membranes were probed with monoclonal antibody (mAb) 5B2 diluted 1:2000 (11).
Reporter Protein Purification and Release of N-Linked Glycans-The Kringle 3 domain was used as a model protein and was purified using His 6 tag as reported previously (26). The glycans were released and separated from glycoproteins by modification of a method reported previously (36). After the proteins were reduced and carboxymethylated, and the membranes blocked, the wells were washed three times with water. The protein was deglycosylated by the addition of 30 l of 10 mM NH 4 HCO 3 , pH 8.3, containing 1 milliunit of N-glycanase (Glyko). After 16 h at 37°C, the solution containing the glycans was removed by centrifugation and evaporated to dryness.
Complementation of P. pastoris OCH1 Gene-A 2.86-kb portion of the P. pastoris OCH1 locus was amplified by PCR using primers PB198 5Ј-GAAGGATCCTGATATAGACCTGCGA-CACCATC-3Ј and PB199 5Ј-ATTCGGATCCTGTCAATGG-GAAGAGATGTCTTGTGCACA-3Ј. The resulting amplified fragment was subcloned into the P. pastoris/E. coli shuttle vec-tor, pRCD252, constructed in our laboratory, which contains the HPH gene obtained from plasmid pAG32 (31), encoding for hygromycin resistance and the P. pastoris TRP2 gene (integration locus). The resulting plasmid, pPB294, was linearized with a unique XbaI site located in the TRP2 gene and transformed into P. pastoris by electroporation. Transformants were selected on YPD medium containing 300 mg/ml hygromycin.
Phenotypic Characterization of C. albicans Strains-To determine the growth rate, strains were grown in YPD-Arg-His-Urd medium at 37°C for 9 h. Every 2 h, absorbance at 620 nm was measured. Hyphae formation was induced in 5% horse serum at 37°C for 3 h. Chlamydospore formation was analyzed in medium containing 17 g/liter corn meal agar (Difco), 0.33% Tween 80, and 2% agar. Cultures were grown at 28 or 37°C for 2-5 days. To determine the sensitivity to chemical agents, 5 l of serial 1:10 dilutions of an overnight culture were spotted onto YPD-Arg-His-Urd agar plates containing 25 or 50 mM CaCl 2 ; 10, 20, or 50 g/ml calcofluor white; 1 or 1.5 M NaCl; 0.001, 0.01, or 0.05% SDS. Plates were incubated at 37°C for 1 day. For sensitivity assays to antifungal agents, including amphotericin B, 5-fluorocytosine, itraconazole, fluconazole, voriconazole, and caspofungin, strains were grown in Sabouraud's dextrose broth at 37°C for 24 h. Two hundred microliters of 1:10 dilution in Shadomy medium were inoculated onto antifungigram plates and incubated at 30°C for 24 h. The growth was determined by measuring the absorbance at 620 nm.
Phosphopeptidomannan Extraction and Western Analyses-PPM from cells grown in YPD-Arg-His-Urd medium was extracted as described previously (37). Briefly, cell pellets were suspended in 0.02 M citrate buffer and autoclaved at 125°C for 90 min. Suspensions were harvested, and Fehling's solution was added to the supernatant to precipitate PPM. The PPM was then washed in methanol:acetic acid (8:1) and dried in a Speed-Vac concentrator. Sugar concentrations were estimated by the sulfuric phenol colorimetric method (38). PPM was analyzed by SDS-PAGE as described above. Membranes were probed with mAbs 5B2 diluted 1:1000 (11), B6.1 diluted 1:1000 (13), or B9E diluted 1:750 (39).
Extraction of Sphingolipids-Cells grown at 37°C in YPD-Arg-His-Urd medium were broken with a French press (Aminco) at 20,000 psi, dialyzed, and lyophilized. Extraction was carried out as described previously (9), using chloroform: methanol:water extractions, butanol:water partitions, and chromatography on phenyl-Sepharose.
Analysis of the ␤-Mans Released from PPM and PLM-Previously extracted PPM or PLM were hydrolyzed in 0.1 N HCl for 1 h at 100°C to release their ␤-Man-containing acidlabile moiety and neutralized with 0.2 N NaOH. Hydrolysates were then dried directly for PLM fractions or after an ultrafiltration step on Centricon YM-30 filters (Amicon) for PPM fractions and tagged with 0.15 M 8-aminonaphthalene-1,3,6trisulfonate (ANTS) and 1 M sodium cyanoborohydride for 16 h at 37°C as described previously (40). The dried samples were resuspended in glycerol:water (1:4) buffer. Electrophoresis of ANTS-labeled oligomannosides was performed on 25 or 35% (w/v) acrylamide separating gels, depending on the degree of polymerization of the oligomannosides, and a 5% acrylamide stacking gel at 600 V. Electrophoresis buffers were the same as for SDS-PAGE except that they did not contain SDS. Acid-hydrolyzed dextran and oligosaccharides were also tagged with ANTS and used as carbohydrate standards. Gels were dried, and images were acquired with the Gel Doc 2000 image analysis apparatus from Bio-Rad equipped with a 365 nm UV transilluminator.
NMR Analyses-For a better selectivity of mutant analyses, NMR experiments were performed separately on acid-resistant domains and on total mixtures of acid-labile oligomannosides isolated from PPM from mutant and wild-type strains. After mild acid hydrolysis, acid-labile and acid-resistant domains of mannans were separated by gel filtration and analyzed separately. All anomeric 1 H NMR signals were then readily attributed to ␣and ␤-linked mannose residues according to published data (37,41 Experiments were recorded at 300 K for acid-labile fractions and 318 K for acid-stable fractions. Chemical shifts were expressed in ppm downfield from the signal of the methyl group of acetone ( 1 H resonating at 2.225 ppm and 13 C at 31.55 ppm for all fractions). For one-dimensional spectra, a spectral width of 4006 Hz was collected as 16,000 complex data points. Sine-shifted bell was used prior to Fourier transformation. Spectra were base-line corrected with a fourth order polynomial function.
Two-dimensional 1 H-13 C heteronuclear (HSQC) spectra were recorded without sample spinning, and data were acquired in the phase-sensitive mode using the time-proportional phase increment method.
ROESY experiment was acquired using 400-ms mixing time and acquired in States mode according to Bax and Davis (42). All parameters (90 hard pulses, soft pulses, delays, and pulse powers) were optimized for each experiment.
Flow Cytometry-Surface expression of ␤-Mans was determined as described previously (43). Briefly, cells grown on YPD-Arg-His-Urd medium at 37°C for 16 h were washed with phosphate-buffered saline and then incubated with factor 5 and 6 sera (Iatron), diluted 1:200. After washing, cells were incubated with specific secondary fluorescein isothiocyanate-labeled antibody, diluted 1:100. Then cells were fixed with paraformaldehyde and analyzed by fluorescence-activated cell sorter.
Data acquisition was performed using an EPICS XLMCL4 (Beckman Coulter, High Wycombe, UK) equipped with an argon ion laser with an excitation power of 15 milliwatts at 488 nm. Fluorescence intensity of 5000 cells was analyzed using WINMDI software (available on line) and represented on a logarithmic scale.
Identification of a Family of Putative P. pastoris ␤-Mannosyltransferase Genes-We speculated that the protein(s) involved in ␤-linked mannose transfer might share sequence similarity or some specific structural motifs with other mannosyltransferases. Therefore, we performed a BLAST search of a P. pastoris partial genomic sequence (Integrated Genomics, Inc.) for genes similar to the known fungal Golgi-residing mannosyland mannosylphosphate transferases. Using the C-terminal part of S. cerevisiae Mnn4 protein containing lysine-glutamic acid rich repeats (KKKKEEEE) (44) as a probe, a sequence with similar lysine-glutamic acid-rich repeats was identified. A detailed analysis of the corresponding contiguous DNA sequence revealed the presence of an open reading frame (ORF) encoding a hypothetical protein of 644 amino acids with a putative N-terminal transmembrane domain, suggesting a possible Golgi localization. In subsequent searches we identified, in the genome of P. pastoris, three more ORFs encoding proteins that share significant sequence similarity (Fig. 3). Interestingly, two of these ORFs are located adjacent to each other. Based on the results of the subsequent experiments, we named these new genes PpBMT1, PpBMT2, PpBMT3, and PpBMT4 (for P. pastoris ␤-mannosyl transfer 1-4).

Analysis of ␤-1,2-Oligomannoside Expression in Ppbmt⌬
Mutants-To characterize the function of these novel genes, PpBMT deletion strains were generated in P. pastoris BK3-1 and YAS137 mutants (lacking outer chain, see Fig. 1A) using fusion PCR-generated mutant alleles (Table 1). Disruptions were confirmed by PCR (data not shown). Both host strains express the Kringle 3 domain of human plasminogen (K3) used as an N-glycosylated reporter protein (26).
Similar to the YAS137 parental strain (Fig. 2A, lane 3), Western blot analyses of whole cell extracts from Ppbmt1-3⌬ mutants generated in this background ( Fig. 2A, lanes 4 -6) revealed that they did not express ␤-Man epitopes recognizable by mAb 5B2. However, strikingly, these epitopes were detected in the Ppbmt4⌬ mutant ( Fig. 2A, lane 7) suggesting that ␤-Man epitopes are accessible following PpBMT4 deletion. Similar results were obtained for deletion mutants constructed in the BK3-1 strain (data not shown). Whole cell analysis was further substantiated by Western blot analyses of the secreted K3 protein (data not shown).
To further characterize the core glycan structure in Ppbmt⌬ mutants, N-glycans were released from secreted Kringle 3 protein by treatment with peptide:N-glycosidase (26) and analyzed by MALDI-TOF mass spectrometry with or without preliminary mannosidase digestion (Fig. 4). As shown previously by Davidson et al. (30)with BK3-1, in reference strain YAS137 (och1⌬ background), core glycans displayed extensive and heterogeneous mannosylation (Fig. 4A). ␤-Mannosylation of a fraction of these glycans blocked the action of ␣-1,2 mannosidase (peak with 10 mannoses; Fig. 4B). By contrast, disruption of the PpBMT2 gene strongly reduced the degree of mannosylation of the core glycans (Fig. 4C) and eliminated glycans resistant to ␣-1,2 mannosidase treatment (loss of peak with 10 mannoses in Fig. 4D) and yielded the Man 5 GlcNAc 2 structure as expected for core N-glycans with terminal ␣-1,2 mannose extensions. In Ppbmt4⌬ strain, an intermediate mannosylation of the core was observed (Fig. 4E), and after ␣-1,2 mannosidase treatment, recalcitrant peaks were still present, but their sizes were smaller than those of the parental strain (Fig. 4F). No  ␤-Mannosylation of Mannan in C. albicans reduction in size or distribution of the recalcitrant peaks was observed in the Ppbmt1⌬ and Ppbmt3⌬ strains (data not shown). These observations suggest that PpBmt2p initiates ␤-mannosylation of core N-linked glycans and that PpBmt4p is responsible for addition of a hexose to the extended ␤-Man chain (Fig. 1A), which obscures the mAb 5B2 epitope, as suggested by the results on Fig. 2A, lane 7.
To test the effect of PpBMT deletions on ␤-Man epitopes on the outer chain, Ppbmt1-4⌬ was transformed with the wildtype P. pastoris OCH1 gene. Western blot analyses of whole cell extracts with mAb 5B2 (Fig. 2B) revealed that Ppbmt1⌬ did not express ␤-Man epitopes (lane 2), in contrast to the parental control strain (Fig. 2B, lane 1). This suggests that PpBmt1p is involved in ␤-mannosylation of the outer chain (Fig. 1A). A slight reduction in the signal was observed in the Ppbmt2⌬ strain (Fig. 2B, lane 3), whereas PpBMT3 deletion had no discernible effect on mAb 5B2 reactivity (lane 4). Interestingly, PpBMT4 deletion resulted in a subtle increase in staining, consistent with results obtained in the och1⌬ mutant background.
The putative functions of PpBmt1-4p, as deduced from this phenotypic analysis, are schematized in Fig. 1A.
Identification of a Related Family of BMT Genes in C. albicans-To determine whether these four new P. pastoris genes have homologs in other organisms, we performed BLAST searches against public genome data bases. Homologs were found in the fungal species C. albicans, Candida glabrata, Debaryomyces hansenii, Saccharomyces castellii, Saccharomyces kluyveri, Aspergillus fumigatus, and Aspergillus terreus. This suggests that these BMT genes form a new family of fungusspecific genes. By homology with the PpBmt2p sequence, nine proteins with unknown function were revealed in C. albicans, a fungal species known to display large amounts of ␤-Mans on various glycoconjugates (8,11). Table 2 summarizes some characteristics of the newly identified putative proteins. The CaBmtp predicted proteins display strong homology to each other (Fig. 5). However, none of these sequences contains the conserved glycosyltransferase domain (pfam Gly_transf_sug; PF04488).
To determine the function of these C. albicans proteins, alleles of each corresponding gene were disrupted using a PCRbased method described previously (33). This generated a set of strains carrying disruptions of single or both alleles of each gene in the parental strain BWP17 (Table 1).
Preliminary characterization of the nine deletion strains revealed phenotypic alterations of PPM, PLM, and mannoprotein ␤-mannosylation (data not shown). We first focused our analyses on four deletion strains selected for displaying phenotypic alterations of PPM ␤-mannosylation, as far as the distribution of ␤-Man within this molecule is clearly known (Fig. 1B) and accessible to structural analyses. We named the corresponding genes CaBMT1-4.

Deletion of CaBMTs Had No Effect on Growth and Morphogenesis in Vitro-When the growth of each
CaBMT deletion strain was monitored in rich medium, no differences were observed. Similarly, all strains had the same ability to produce hyphae in serum or chlamydospores on RAT medium. None of the strains displayed increased sensitivity to CaCl 2 , calcofluor white, NaCl, or SDS. No differences in growth were observed on media containing increasing concentrations of antifungal agents. Results from Alcian blue binding assays demonstrated no effect on glycan phosphorylation in mutant strains (data not shown).

CaBMT1-4 Deletions Affect ␤-Mannosylation of Either the Acid-stable or the Acid-labile Fractions of Phosphopeptidomannan-PPMs
were extracted from selected strains and first analyzed by Western blot. Staining with mAb B9E, specific  (No enzymatic treatment panel). In addition, a portion of the released glycans was digested with ␣-1,2-mannosidase prior to MALDI-TOF MS analysis (␣-1,2 mannosidase-treated panel). Numbers next to the peaks correspond to the number of mannose residues present in N-glycans, calculated on the basis of the molecular mass of the peak (e.g. the peak designated with the number 8 corresponds to a Man 8 GlcNAc 2 N-glycan). A and B, YAS137 (control strain); C and D, PBP130 (Ppbmt2⌬); E and F, PBP135 (Ppbmt4⌬). for ␤-Mans at the nonreducing end of ␣-1,2 chains of the acidstable fraction of C. albicans serotype A strains (antigenic factor 6, see Fig. 1B), revealed a complete loss of reactivity for Cabmt1⌬ and Cabmt3⌬ strains (Fig. 6A, lanes 1 and 3). Reactivity to mAbs 5B2 and B6.1, specific for ␤-1,2-mannotriose, was reduced (Fig. 6, B and C) but was completely abolished (data not shown) after acid hydrolysis, which removes the acidlabile fraction of PPM (see Fig. 1B). This suggests that CaBMT1 and CaBMT3 deletions affect ␤-mannosylation of the acid-stable part of PPM. In contrast, PPMs from Cabmt2⌬ and Cabmt4⌬ strains displayed wild-type reactivity with mAb B9E (Fig. 6A, lanes 2 and 4) but a strongly reduced reactivity with mAb 5B2 suggesting that ␤-mannosylation of the acid-labile part of PPM is affected in these mutants. CaBmt2p, CaBmt3p, and CaBmt4p Are Involved in ␤-Mannosylation of the PPM Acid-labile Fraction-The oligomannosides released from PPMs by mild acid hydrolysis were analyzed by fluorophore-assisted carbohydrate electrophoresis (FACE) (Fig. 7). In comparison with the parental strain (lane T), deletion of CaBMT1 had no effect on ␤-mannosylation of the PPM acid-labile fraction (Fig. 7, lane 1). In contrast, FACE analysis of the acid-labile fraction from Cabmt2⌬, Cabmt3⌬, and Cabmt4⌬ (Fig. 7, lanes 2-4) revealed the accumulation of a single mannose, a mannobiose, and a mannotriose, respectively, suggesting that these three CaBMTs are successively involved in ␤-mannosylation of the PPM acid-labile fraction. Cabmt3⌬ also displayed a slight expression of ␤-Man with a degree of polymerization Ͼ2, which may arise from redundant activity of other CaBmtps (Fig. 7, lane 3).   These results were further confirmed by NMR analysis of the acid-labile fractions from all strains composed of different sets of free oligosaccharides with a reducing mannose residue. As shown in Fig. 8D, the 1 H NMR spectrum of the acid-labile fraction from the serotype A wild-type strain displayed a complex pattern of H-1 ␤-Man signals between ␦ 4.75 and 5.10, as well as several signals around ␦ 5.30 attributed to the reducing H-1 ␣-Man residues of oligomannosides. The ␣-anomer of free mannose residues is also easily identified at ␦ 5.17. These parameters are in agreement with the presence of a heterogeneous family of acid-labile ␤-Mans as described previously in a C. albicans serotype A strain (37).
Analysis of Cabmt2⌬ (Fig. 8A) showed a much simpler 1 H NMR spectrum dominated by two signals at ␦ 5.173 ( 3 J 1,2 1.7 Hz) and 4.890 ( 3 J 1,2 Ͻ 0.8 Hz), which were tentatively attributed to the H-1 signals of ␣and ␤-anomers of free mannose in accordance with published data. This was confirmed by observing their respective spin systems on COSY and ROESY spectra (data not shown). No signal from substituted ␣-Man residues or internal ␤-Mans was identified, indicating the total absence of oligosaccharides in the acid-labile fraction of Cabmt2⌬ mannan. These results confirmed that acid hydrolysis of Cabmt2⌬ PPM released single mannose residues exclusively.
In addition to free reducing ␣and ␤-Man H-1 signals as observed in 1 H NMR spectra of Cabmt2⌬, the Cabmt3⌬ spectra exhibited two intense anomeric signals at ␦ 5.29 (b␣) and 4.78 (a␣) (Fig. 8B), which correlated with their respective H-2 at ␦ 4.13 and 4.05 on a COSY spectrum (data not shown). The strong downfield shifts of b␣ H-1 and H-2 signals at 5.29 and 4.13 ppm compared with free ␣-Man indicated that this residue of ␣-Man is substituted in the C-2 position. The H-1 signal of ␤-Man residue (a␣) of the ␣-form of mannobiosyl was identified based on the H1/H5 internal nuclear Overhauser effect observed from the signal at ␦ 4.78 (data not shown). Furthermore, ROESY analysis showed a cross-peak between the a␣ H-1 signal at ␦ 4.78 ppm and the H-2 signal at ␦ 4.13, confirming that the ␤-Man residue was linked to the C-2 position of the ␣-Man residue (b␣). The H-1 of ␣-Man (b␤) and of the ␤-Man (a␤) residues of the ␤ form of mannobiosyl was also identified as minor signals at ␦ 4.99 and ␦ 4.82, respectively. Thus, NMR data confirmed that the acid-labile fraction of Cabmt3⌬ mannan contains a mixture of free mannose and Man(␤1-2)Man disaccharide.
CaBmt1p and CaBmt3p Are Involved in ␤-Mannosylation of the PPM Acid-stable Fraction-One-dimensional 1 H NMR analysis of the acid-stable fraction of mannan from the serotype A wild-type strain showed a complex pattern of H-1 protons of ␣and ␤-linked Man residues (Fig. 9A). Nondecoupled 13 C-1 H HSQC heteronuclear analysis (data not shown) allowed us to distinguish ␤-Man from ␣-Man residues because of their low 1 J CH constant around 157 ppm, compared with 171 ppm for ␣-Man. As shown on the 13 C-1 H HSQC heteronuclear spectrum (Fig. 9E), four signals labeled a-d were attributed to ␤-Man H-1 depending on their location within the ␤-mannan chain. It has been reported previously that serotype B strains lack ␤-mannosylation on the acid-stable fraction of mannans (7). In accordance with these data, the NMR spectrum of a serotype B strain did not show any signal originating from ␤-Man residues when the same analysis was performed (Fig. 9B).
NMR analysis of the acid-stable fraction of Cabmt1⌬ (Fig.  9C) showed a total absence of signals attributed to ␤-Man residues, identical to the serotype B strain (Fig. 9B). In contrast, NMR analysis of the acid-stable fraction of Cabmt3⌬ (Fig. 9D) showed a single ␤-Man H-1 signal at 4.79 ppm attributed to Man(␤1-2)Man(␣1-terminal) motif. Presence of a single signal out of four was confirmed by nondecoupled and decoupled 13 C-1 H HSQC analysis (data not shown). These data directly confirmed that the Cabmt1⌬ strain does not contain any FIGURE 8. NMR analysis of acid-labile domain. NMR structural analysis of purified acid-labile oligomannosides after mild acid hydrolysis from wild-type strain and mutants. One-dimensional 1 H NMR spectrum of oligomannoside fraction from the following: A, AL86 (Cabmt2⌬); B, AL90 (Cabmt3⌬); C, AL84 (Cabmt4⌬); and D, BWP17 (parental strain). E, 1 H-ROESY spectrum of oligomannoside fraction from AL84 (Cabmt4⌬). The detailed structure of ␤-Man chains shows from which part of the molecules the signals labeled in the various spectra originated. Each anomer a-h is observed as two distinct signals according to the ␣ or ␤ forms of the reducing mannose residue of the oligomannosylated chain.

␤-Mannosylation of Mannan in C. albicans
␤-Man residues on the acid-resistant domain of mannan, whereas Cabmt3⌬ retains a single ␤-linked mannose residue.

Control of Bmtp Activity by Construction of Revertant
Strains-Although all of the deletions had an impact of ␤-mannosylation, it was necessary to confirm specific activity by complementation of one gene copy. Wild copies of CaBMT genes were reintegrated into the RPS10 locus of the corresponding Cabmt⌬ null strains. Complementation was confirmed by PCR (data not shown). Phenotypic analyses either by Western blot for Cabmt1⌬/BMT1 and Cabmt3⌬/BMT3 (Fig. 6, lanes 1Ј and  3Ј) or by FACE for Cabmt2⌬/BMT2, Cabmt3⌬/BMT3, and Cabmt4⌬/BMT4 (Fig. 7, lanes 2Ј, 3Ј, and 4Ј) showed complete restoration of phenotype assigning definitely a function for these representative members of the CaBmtp family.
CaBMT1-4 Deletions Do Not Affect PLM ␤-Mannosylation-Considering homology of some ␤-Man acceptor sites (namely -Man-P-Man) between PPM and PLM (Fig. 1, B and C), we investigated whether deletion of CaBMT1-4 had an impact on PLM. No effect on PLM ␤-mannosylation was observed in Western blot analysis of whole cell extracts (Fig. 10A). These results were confirmed by the more analytical FACE method performed on oligomannosides released from PLM in acidic conditions. Altogether these data demonstrate substrate specificity for CaBmt1-4p between PPM and PLM.
Single Deletions Have a Slight Effect on ␤-1,2-Oligomannoside Surface Expression-Flow cytometry assays were performed using factor 5 serum recognizing homopolymers of ␤-Mans in the acid-labile fraction of C. albicans and factor 6 serum recognizing ␤-Man residues at the nonreducing end of the ␣-1,2 lateral chain specific to the PPM acid-stable fraction of C. albicans serotype A (Fig. 1B). Surface expression of ␤-Mans of deletion strains detected with polyclonal antibodies was similar to that of the parental strain (Fig. 11) except for Cabmt1⌬, which displayed a significant reduction in antigenic factor 6 expression (Fig. 11A), and for Cabmt3⌬ whose reactivity slightly increased with serum factor 5 (Fig. 11C). In contrast to results obtained with purified PPM, deletions did not result in dramatic changes suggesting that, besides PLM which was not affected, other cell wall molecules contribute to compensate cell wall expression of ␤-Mans.

DISCUSSION
The significant increase in opportunistic fungal infections observed over the last 2 decades represents an important med-

␤-Mannosylation of Mannan in C. albicans
ical challenge. In particular, C. albicans is one of the most frequent causes of nosocomial bloodstream infections and is associated with high mortality rates (45). Despite increased investments in antifungal therapies, only limited progress has been achieved in controlling nosocomial candidiasis (1). C. albicans exists as a harmless, commensal micro-organism in the majority of immunocompetent people. However, the increase in number of patients with compromised immunity has contributed to the observed rise in opportunistic fungal infections. Several traits, including differential expression of adhesins (46 -48) and lytic enzymes (49), morphological changes (50), stress response (51), and changes in the glyoxylate cycle (52) have been linked to pathogenicity of C. albicans. However, the way in which C. albicans coordinates these factors in its continuous adaptation to host conditions is still poorly understood.
Among factors contributing to C. albicans virulence, which still pose unresolved questions, are ␤-Mans. ␤-Mans contribute to at least two important pathogenic mechanisms as follows: adhesion to the host cells (14) and modulation of the immune response (53). At the cell wall level, ␤-Mans are known to be associated with PPM (7,8) and PLM (9) as well as several studies which suggest their association with cell wall mannoproteins. Failure to identify genes involved in the biosynthesis of ␤-Mans has hindered the implementation of genetic approaches to study their function, i.e. due to their notable absence in most model fungal species.
In this study, we describe the identification and characterization of a new family of genes responsible for the synthesis of ␤-Mans in the following two yeast species: P. pastoris and C. albicans. First, employing glycan analyses by MALDI-TOF mass spectrometry and Western blot analyses of glycoproteins with mAb 5B2, raised against C. albicans cell wall associated ␤-Mans, we were able to confirm the presence of ␤-Mans in P. pastoris. The ability of P. pastoris cells to modify glycoproteins with immunogenic ␤-Mans could be a potential obstacle in the successful implementation of this organism as a production platform for therapeutic glycoproteins (54). To overcome this problem, we attempted to identify the genes responsible for the transfer of ␤-1,2-mannose in P. pastoris. Making the assumption that putative ␤-1,2-mannosyltransferases might share some structural motifs or domains with other yeast Golgi glycosyltransferases, we employed BLAST to systematically search for novel genes with such sequence similarities. Golgi mannosyltransferases are typically type II membrane proteins that add ␣-1,2-, ␣-1,3-, and ␣-1,6-linked mannoses to protein-and lipidlinked glycans (55). In several yeasts, including S. cerevisiae, C. albicans, and P. pastoris, glycans also contain mannosyl phosphate, and the proteins responsible for mannosyl phosphate transfer are also type II membrane proteins such as Mnn4p (44). A BLAST search using the S. cerevisiae Mnn4p tail as a probe resulted in the identification of a new gene encoding a putative type II membrane protein with C-terminal repeats of lysines and glutamic acids but with no homology to known genes. Subsequently, three similar genes were identified suggesting the presence of a novel gene family with four members in P. pastoris. The results of subsequent BLAST searches against public data bases revealed that the new family of genes is present only in a limited number of fungal species. None of the commonly studied fungal model organisms, including S. cerevisiae and Aspergillus nidulans, harbored homologs of these genes. However, we were able to identify nine homologs in C. albicans consistent with the prominent expression of ␤-Man residues by C. albicans, as established by serological classification (56) and subsequently by biochemical and cytological studies (8,9,11,57).
Analyses of a library of deletion strains revealed that these novel genes are indeed involved in the transfer of ␤-Man residues in P. pastoris. Western blot, mannosidase digests, and MALDI-TOF MS analyses were performed to characterize the N-glycans produced by the mutant strains. The analyses were performed in strains with wild-type glycans and in an och1⌬ mutant background with only core N-glycans present. All the results point to the conclusion that this new family of genes is responsible for the addition of ␤-Man to N-glycans. In the wild-type strain background, the deletion of PpBMT1 eliminated the interaction between whole cell extracts and mAb 5B2, suggesting that the PpBmt1 protein is necessary for the synthesis of a structure recognized by this mAb. PpBMT2 deletion slightly reduced the interaction with mAb 5B2. No changes in ␤-Man were observed in the PpBMT3 deletion strain. Interestingly, deletion of PpBMT4 in the wild-type background enhanced the signal on Western blot probed with mAb 5B2. This suggests that ␤-mannosylation catalyzed by PpBmt4p allows further steps that mask ␤-Man epitopes recognized by mAb 5B2.
In the och1⌬ strains, deletion of PpBMT2 eliminates neutral glycans resistant to ␣-mannosidase treatment as shown by MALDI-TOF MS analysis. In contrast to the wild-type strain, no effect of a PpBMT1 deletion was observed in the och1⌬ strain background. This might indicate that PpBmt1p is responsible for ␤-mannose transfer only on outer chain N-glycans. Again, we were unable to detect any changes in N-glycans associated with the PpBMT3 deletion. The deletion of PpBMT4 in och1⌬ strains seemed to expose the ␤-Man structures recognized by mAb 5B2 and even more dramatically in this case because of the complete lack of signal in the och1⌬ mutant parental strain.
However, MALDI-TOF analysis of the N-glycans from the PpBMT4 deletion strain showed a size reduction of glycans resistant to ␣-mannosidase treatment. This result, in agreement with Western blot analysis of whole cell extracts, confirms that PpBmt4p extends glycans containing ␤-Man with hexose. Future NMR studies of glycans from wild-type P. pastoris as well as Ppbmt⌬ mutants will be necessary to determine the structure of ␤-Man containing glycans and the exact function of PpBMT genes, particularly PpBMT4, in their synthesis.
Similarly, we investigated whether the nine homologs identified in the genome of C. albicans were also involved in the transfer of ␤-Man residues. First, we created a library of C. albicans BMT (CaBMT) deletion strains. In vitro, under the conditions tested, none of the deletion strains displayed alterations in either growth phenotype or sensitivity to chemical or antifungal compounds. However, all of the strains displayed an altered pattern of ␤-Man epitopes in PPM. Based on previously established chemical structures of PPMs, we designed and conducted a series of experiments involving purification of these molecules. Western blot of PPMs with anti-␤-Man antibodies of different specificity, FACE analysis of oligomannosides, and confirmation of their structure by NMR led to the following conclusions, summarized in Fig. 1B. CaBmt1p and CaBmt2p are responsible for the addition of the first ␤-mannose on PPM acid-stable and acid-labile fractions, respectively, and therefore act on ␣-linked mannose residues as acceptors. CaBmt3p and CaBmt4p are active on ␤-mannoses as acceptors, because they are involved in the elongation of ␤-Man chains of the PPM acid-labile fraction. Analysis of the CaBMT3 deletion also suggested its involvement in the addition of the second ␤-mannose to the PPM acid-stable domain. None of these deletions had an impact on PLM ␤-mannosylation demonstrating the substrate specificity of Bmtps. Finally, flow cytometry analysis revealed that modifications of PPM ␤-mannosylation had a limited effect on ␤-Man cell wall surface expression.
A number of studies have implicated a role for mannan in C. albicans biology and pathogenicity. The importance of mannan in virulence has been demonstrated by inactivation of genes involved in transfer of ␣-mannosyls on O-glycans (58) or the N-glycan outer chain (19). In parallel, experimental studies either in vitro (17,53) or in vivo (13,18), as well as clinical studies involving purified (14,53) or synthetic ␤-mannosides (18), or purified molecules selectively containing these residues (16,17), have shown that ␤-Mans also contribute to C. albicans virulence. Moreover, a separate study has revealed that the spatial conformation of ␤-Man (12) differs from the more ubiquitous ␣-Mans, most of which are shared with S. cerevisiae. These differences affect ␤-Man recognition by the adaptative (13) and innate immune system (15,59). Furthermore, an interesting observation is that members of the BMT gene family are present in only a limited number of known fungal species, but among these is the prominent fungal pathogen A. fumigatus (60). Despite these findings, recently published papers have contradicted this notion and led to the conclusion that ␤-Man may actually have little effect on C. albicans virulence (20,21). However, the strategy followed in these papers did not specifically target ␤-Man, but rather was based on the inactivation of C. albicans genes, homologous to S. cerevisiae genes, which prevented downstream association of ␤-Man to PPM (20,21).
The results presented here demonstrate a multiplicity and specificity of Bmtps for their substrates, as well as a relatively limited impact of individual deletions on overall ␤-Man expression. It is striking that deletion of CaBMT2 leading to the complete absence of PPM acid-labile fraction (identified as antigenic factor 5 (61)) had no impact on direct agglutination by the same factor. Even the more analytic flow cytometry method did not find any difference between the mutant and parental strain using factor 5. This implies that the complexity of the process of ␤-mannosylation in C. albicans has been underestimated.
This study was designed as a principal study to define the functions of a new gene family. As such it focused on C. albicans PPM as a reference molecule whose structure elucidation has required many studies. Besides definition of Bmt1-4p functions, it was shown that disruption of these BMTs acting on PPM had no impact on PLM ␤-mannosylation. Further studies will focus on this using CaBMTs.
Considering this specificity of Bmtps in early biosynthetic steps, pleiotropic effects of OCH1 deletion are probably also of importance to be considered. The unknown mechanisms of N-mannan core elaboration in the C. albicans och1⌬ mutant noticed by Bates et al. (19) present striking homologies with what was observed on P. pastoris och1⌬ regarding ␤-Mans.
However, for mature mannoglycoconjugates, considering previous immunological studies on ␤-Man expression in the C. albicans cell wall, and the present results obtained by flow cytometry, the hidden part of the iceberg is probably represented by the numerous cell wall mannoproteins harboring altogether massive amounts of ␤-Man epitopes. As far as the function of a cell wall molecule and its recognition by the host are dependent on glycosylation, it can be anticipated that a role for ␤-Mans in virulence may yet be revealed.
With this is mind, the identification of genes responsible for ␤-mannose transfer creates the opportunity to directly study the ␤-Man biosynthetic pathway in C. albicans and its relation to pathogenicity. More generally detailed analysis of bmt⌬ mutants in different living organisms should allow a more complete understanding of the extent and function of this posttranslational modification and how to manage it for engineering therapeutics and industrial purposes.