Multiple Functions of Pmt1p-mediated ProteinO-Mannosylation in the Fungal Pathogen Candida albicans *

Protein mannosylation by Pmt proteins initiatesO-glycosylation in fungi. We have identified thePMT1 gene and analyzed the function of Pmt1p in the fungal human pathogen Candida albicans. Mutants defective inPMT1 alleles lacked Pmt in vitro enzymatic activity, showed reduced growth rates, and tended to form cellular aggregates. In addition, multiple specific deficiencies not known inSaccharomyces cerevisiae (including defective hyphal morphogenesis; supersensitivity to the antifungal agents hygromycin B, G418, clotrimazole, and calcofluor white; and reduced adherence to Caco-2 epithelial cells) were observed in pmt1 mutants.PMT1 deficiency also led to faster electrophoretic mobility of the Als1p cell wall protein and to elevated extracellular activities of chitinase. Homozygous pmt1 mutants were avirulent in a mouse model of systemic infection, while heterozygousPMT1/pmt1 strains showed reduced virulence. The results indicate that protein O-mannosylation by Pmt proteins occurs in different fungal species, where PMT1deficiency can lead to defects in multiple cellular functions.


Protein mannosylation by Pmt proteins initiates Oglycosylation in fungi.
We have identified the PMT1 gene and analyzed the function of Pmt1p in the fungal human pathogen Candida albicans. Mutants defective in PMT1 alleles lacked Pmt in vitro enzymatic activity, showed reduced growth rates, and tended to form cellular aggregates. In addition, multiple specific deficiencies not known in Saccharomyces cerevisiae (including defective hyphal morphogenesis; supersensitivity to the antifungal agents hygromycin B, G418, clotrimazole, and calcofluor white; and reduced adherence to Caco-2 epithelial cells) were observed in pmt1 mutants. PMT1 deficiency also led to faster electrophoretic mobility of the Als1p cell wall protein and to elevated extracellular activities of chitinase. Homozygous pmt1 mutants were avirulent in a mouse model of systemic infection, while heterozygous PMT1/pmt1 strains showed reduced virulence. The results indicate that protein O-mannosylation by Pmt proteins occurs in different fungal species, where PMT1 deficiency can lead to defects in multiple cellular functions.
Candida albicans is the leading cause of superficial and systemic fungal infections of humans. Virulence of C. albicans is a composite of several characteristics, including adaptation to body temperature, adhesion, and penetration of epithelial cells as well as its ability to undergo morphogenetic changes, especially between a yeast and a filamentous growth form (1,2). Conceivably, defects in protein glycosylation may have profound effects on the physical stability and function of the fungal cellular surface and may also affect the secretion efficiency and the activity of secreted soluble proteins.
Present evidence suggests that O-glycosylation proceeds differently in fungal and in higher eukaryotic cells. The first O-glycosylation step in the yeast Saccharomyces cerevisiae, which is catalyzed by protein mannosyltransferases (Pmt) 1 (3), occurs in the endoplasmic reticulum and consists of the co-translational transfer of mannose from Dol-P-Man to serine or threonine residues; further mannosylation extension reactions occur in the Golgi (reviewed in Refs. 4 and 5). In C. albicans, the majority of O-glycosyl chains consist of two or three mannoses, but they can extend up to about seven mannoses in ␣-glycosidic linkages (6). Dol-P-Man-dependent O-glycosylation of secreted proteins has been observed in other yeast species and filamentous fungi (7)(8)(9). In contrast to fungi, most O-glycosylation reactions in higher eukaryotic cells including human cells proceed in the Golgi and commence with the attachment of N-acetylgalactosamine to proteins (10). However, some mammalian proteins also appear to be O-mannosylated (11,12).
A total of seven homologous PMT genes have been identified in the genome of S. cerevisiae (13)(14)(15)(16)(17). The PMT1 and PMT2 gene products appear to form a heterodimer that is necessary for enzymatic activity in vitro and in vivo (18). Deletion of PMT1 does not affect viability but leads to cells that tend to aggregate, especially in a pmt2 genetic background (16,17). Inactivation of both PMT1 and PMT2 causes defects in growth and resistance to killer toxin K1 (16), while some combinations of PMT triple mutants are inviable (19). Recently, a genomic sequence of Schizosaccharomyces pombe has been reported, which encodes a Pmt homologue (GenBank TM accession number Z99126). In Drosophila melanogaster, a gene encoding a protein with high homology to the yeast Pmt1 proteins has been described (20). Furthermore, several "expressed sequence tags" deposited at GenBank TM are derived from genes encoding putative Pmt proteins of human, mouse, rat, Caenorhabditis elegans, and rice cells.
In this report, we describe the C. albicans homologue of the S. cerevisiae PMT1 gene and characterize its function. PMT1 deletion in C. albicans leads to a lack of in vitro Pmt activity and affects two secreted proteins, Als1p and chitinase. In addition, more drastic in vivo phenotypes compared with S. cerevisiae are observed, including a complete block of morphogenesis in some inducing conditions, an increased sensitivity to some antifungal compounds, and a decreased adherence to host cells. Furthermore, PMT1 function is shown to be essential for the virulence of C. albicans in a mouse model of infection.

MATERIALS AND METHODS
Strains and Growth Conditions-PMT alleles of C. albicans and S. cerevisiae are designated CaPMT and ScPMT respectively, whenever necessary for their distinction. Strains and plasmids are listed in Table  I. C. albicans strain CAI4 (21) was used for transformations and gene disruptions. Strains were grown in YPD or SD medium (22), which for Ura Ϫ strains was supplemented with 20 g/ml uridine. YPGal medium is identical to YPD except that it contains 2% galactose and 0.2% glucose as carbon sources. Growth was monitored by measuring the optical density at 600 nm of cultures (A 600 ) using a Novaspec II photometer (Amersham Pharmacia Biotech); cells were dispersed by sonification (1 min) in a bath sonifier (Bandelin Sonorex TK52) before measurements. Transformations were performed using the spheroplast * 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 GenBank TM /EBI Data Bank with accession number(s) AF000232 (for CaPMT1).
Cloning and Sequencing of PMT1-A partial genomic DNA library of C. albicans CAF3-1 (21) was constructed by digesting DNA with Hin-dIII and insertion of 5-6-kb fragments into YEplac195 (23). E. coli DH5␣ transformants carrying the genomic library were analyzed by colony hybridization using a 2.5-kb EcoRV-XhoI fragment containing the S. cerevisiae PMT1 gene (17) as a probe. Plasmid pCT1 was found to strongly hybridize to ScPMT1 and to contain a 5.5-kb C. albicans insert. Subfragments of the pCT1 insert were ligated into pUC19 and sequenced from both ends using forward and reverse primers or by using insert-specific oligonucleotides. The 5.5-kb HindIII CaPMT1 fragment was inserted into the HindIII site of pRC18 (24) to generate replicating C. albicans vectors (resulting in plasmids pCT29 and pCT30 with inverse insert orientation); alternatively, the CaPMT1 fragment was inserted into the HindIII site of YEplac112 or YCplac22 (23) to construct a high copy (pCT31) or a low copy vector (pCT32) for transformation of S. cerevisiae. To place CaPMT1 under transcriptional control of a defined S. cerevisiae promoter, the CaPMT1 coding region was amplified by PCR, using the primers pcP1-9Bam (5Ј-TTAGGATC-CCATTCAATATGGCAA-3Ј) and pcP1-10Hind (5Ј-TTAAAGCTTA-AGTTTTCATCTACAAC-3Ј), by which a BamHI site (italic type) is placed 9 bp upstream of the ATG start codon (underlined), and a HindIII site (italic type) is placed 17 bp downstream of the stop codon. The BamHI-HindIII CaPMT1 fragment was placed downstream of the GAL1 promoter (the EcoRI-BamHI fragment of pBM150 (25)), and the promoter fusion was inserted, as an EcoRI-HindIII fragment, between the EcoRI and HindIII sites of YCplac22 or YEplac112 (23) to generate plasmids pCH1 or pCH2. pCT28 contains the 7.1-kb BamHI-XhoI genomic ScPMT1 fragment (17) as an SacI-SphI insert in YCplac22.
Gene Disruptions-In preparation for disruption of PMT1, a pUC19 derivative (pCT5) containing 2 kb of the PMT1 promoter region and 2 kb of the coding region on a HindIII-HincII fragment was used as template for "divergent" PCR using the oligonucleotides A (5Ј-TTAGTC-GACATTGAATGGGAAACT) and B (5Ј-TTAAGATCTTTAGAGCCG-GATTGC). Oligonucleotide A corresponds to nucleotides Ϫ1-15 relative to the ATG in the PMT1 sequence and also adds a SalI site to the PCR product, while oligonucleotide B corresponds to nucleotides 975-990 and adds a BglII site. The 6.7-kb PCR fragment was cut with SalI and BglII and ligated to the 4-kb SalI-BglII "URA blaster" fragment of plasmid p5921 (26). From the resulting plasmid, pCT23, a 6.7-kb Asp718 -SpeI was isolated that was used to transform strain CAI4 to prototrophy. Correct insertion of the URA blaster into one of the two PMT1 alleles was verified by Southern blots on DNA of transformants, which was cut with HindIII and probed with a 0.7-kb XbaI fragment derived from the PMT1 promoter region or a 1.3-kb NheI-BamHI fragment from p5921 carrying hisG. A pmt1⌬::hisG-URA3-hisG/PMT1 strain was plated out on media containing 0.02% 5-fluoroorotic acid. Spontaneous 5-fluoroorotic acid-resistant strains were analyzed for loss of the URA3 sequence by Southern blotting. One of several identified strains, CAP1-31, of the genotype pmt1⌬::hisG/PMT1 was used for a second round of gene disruption using the original URA blaster fragment. Several transformants had the genotype pmt1⌬::hisG/pmt1⌬::hisG-URA3-hisG, and strain CAP1-312 was chosen to identify strains by 5-fluoroorotic acid resistance (21). Strain CAP1-3121 is a representative of mutant strains of the genotype pmt1⌬::hisG/pmt1⌬::hisG. The CaPMT1 gene was reintroduced in CAP1-3121 by transformation with plasmid pCT29 or pCT30.
To disrupt ScPMT1, we transformed S. cerevisiae strain B76 (24) with an XbaI-SacI fragment containing pmt1⌬::URA3, as described (13,17). Correct replacement of the genomic ScPMT1 gene was verified by Southern blottings (13). The resulting strain, B76⌬pmt1::URA3, was diploidized by transformation with YEplac112-HO, 2 , which contains the HO gene (27). One diploid transformant was grown repeatedly under nonselective conditions in YPD, and a plasmid-free derivative, B76d⌬pmt1, was identified. To induce pseudohyphal growth, B76d⌬pmt1 was transformed with YCplac22RAS2 Val19 , 3 and cells were grown on low nitrogen SLAHD plates as described (28). As a control, the diploid strain B76/4 was analyzed (24). Assay of Pmt1p Enzyme Activity-Cells were grown in 50 ml of YPD or SD to an A 578 of 0.5-1.0. 40 OD units of cells were harvested and washed once with 10 ml of ice-cold TM buffer (50 mM Tris-HCl, pH 7.5, 0.3 mM MgCl 2 ). The cells were resuspended in 100 l of TM buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, 1 g/ml pepstatin, 50 g/ml N-tosyl-L-phenylalanine chloromethyl ketone, 0.25 mM N ␣ -p-tosyl-L-lysine chloromethyl ketone, 20 g/ml antipain, 1 mM benzamidine) and transferred to a 1.5-ml microcentrifuge tube. An equal volume of glass beads was added, and the cells were lysed by vortexing eight times in intervals of 1 min with a 1-min incubation on ice. The lysate was collected into a fresh microcentrifuge tube. Cell debris was pelleted by centrifugation (5 s/4°C), and the supernatant was transferred to a new tube. Membranes were collected by centrifugation at 48,000 ϫ g for 30 min at 4°C and were then resuspended in 100 l of 50 mM bicine, pH 7.7, 1 mM EDTA, 33% glycerol, and protease inhibitors. Microsomal membranes were stored in liquid nitrogen.
The in vitro peptide assay for Pmt1p activity was performed according to Weston et al. (29). The assay was optimized using Dol-P-[ 14 C] mannose (approximately 30,000 cpm; specific activity of 305 Ci/mol) and the acceptor peptide Ac-YATAV-NH 2 (final concentration of 4 mM). Up to 25 l of C. albicans microsomal membranes (protein concentration of 3 mg/ml) were added. Assay mixtures were incubated at 25°C for 20 min.
Adherence to Epithelial Cells-Human colon carcinoma cells (Caco-2) (30, 31) were grown to confluency in Dulbecco's modified Eagle's medium containing 20% fetal calf serum, 0.45% glucose, 292 g/ml glutamine, 1 mM sodium pyruvate, 1% nonessential amino acids, 50 g/ml streptomycin, and 50 g/ml penicillin G (Life Technologies, Inc.) at 37°C (5% CO 2 ). Monolayers established in culture dishes (well diameter of 60 mm) were used 2-4 days after confluency for adhesion studies. Adhesion was determined according to Rotrosen et al. (32) and Fratti et al. (33). Briefly, monolayers were washed twice with 2 ml of phosphatebuffered saline, followed by the addition of approximately 200 C. albicans cells in 1 ml of phosphate-buffered saline. Cells were incubated at 37°C for 45 min in an atmosphere of air containing 5% CO 2 . Following the incubation, monolayers were washed once with 5 ml of phosphatebuffered saline to remove nonadhering cells; the monolayer in each well was then covered by 2 ml of YPD agar (1% agar). Yeast colonies appearing after 48 h of growth at 30°C were counted. As a control, fungal cells were plated directly on YPD.
Animal Experiments-Virulence studies were performed as described, using 8-week-old, male CFW-1 mice (Halan-Winkelmann, Paderborn, Germany) (34). Because of the aggregation of Capmt1 mutants, all strains were sonicated for 5-10 min in a bath sonifier (Bandelin Sonorex TK52) before injection; as determined by plating efficiency before and after sonication, viability was not affected by this treatment. Survival curves were calculated according to the Kaplan-Meier method using the Prism (TM) program (GraphPad Software Inc., San Diego) and compared using the log-rank test. A p value Ͻ0.05 was considered significant. To quantify colony-forming C. albicans units in kidneys, mice were sacrificed 48 h after injection, and kidneys were homogenized in 5 ml of phosphate-buffered saline, serially diluted, and plated.
RNA blots were performed as described using the 2.2-kb XbaI fragment of pCT5 carrying PMT1 or the 1.5-kb SalI-ClaI ACT1 fragment (24) as probes.
For immunodetection of Als1p, cells were grown in SD medium to an A 600 ϭ 3-5 at 30°C. Cells were harvested by centrifugation, resuspended in 10% trichloroacetic acid, 10 mM Tris-HCl (pH 8.0), 25 mM ammonium acetate, 1 mM EDTA and disrupted using glass beads (35). After centrifugation at 13,000 rpm for 30 min (4°C), the supernatant was discarded, and the pellet was solubilized in a small volume of 8 M urea, 100 mM sodium phosphate, 10 mM Tris-HCl (pH 7.0); an equal volume of SDS-PAGE sample buffer (2ϫ) was added, and samples were heated for 5 min at 65°C Proteins in the solubilzed extracts were separated by SDS-PAGE (5% acrylamide) and transferred to Immobilon P (Millipore Corp.) filters. Als1p was visualized by standard procedures after reaction with anti-Als1 antiserum (1:10,000), followed by antirabbit IgG coupled to horseradish peroxidase and staining by the Su-perSignal ULTRA substrate (Pierce). Monoclonal antibody 1B12 was kindly supplied by R. Sentandreu (36); polyclonal antibodies against the gp115 protein, the Pma1 protein, the Cdr1 protein and the Int1 protein were gifts by L. Popolo, A. Goffeau, D. Sanglard, and M. Hostetter, respectively (37)(38)(39)(40).
Chitinase was assayed essentially as described by McCreath et al.

RESULTS
Sequence of CaPMT1-Each of the seven S. cerevisiae PMT genes was used to detect homologous genes in C. albicans genomic DNA by low stringency Southern hybridization. With genomic DNA cut by different restriction enzymes, only two patterns of hybridization were obtained. PMT1 and PMT5 probes yielded one pattern (Fig. 1A), while use of the PMT2, PMT3, and PMT6 genes resulted in an alternative pattern (Fig.  1B), and PMT4 and PMT7 did not show any hybridization. Since only a single hybridizing band was obtained in each case, these results suggested that only two PMT genes, designated PMT1 and PMT2, exist in C. albicans, although the possibility of the presence of less homologous genes could not be excluded. Previously, we had pointed out the sequence-relatedness of the S. cerevisiae PMT1 and PMT5 genes as well as the PMT2, PMT3, and PMT6 genes (13).
Based on the Southern results, we constructed a partial C. albicans genomic bank and cloned CaPMT1 by hybridization to the S. cerevisiae PMT1 gene. A positive clone contained a 5.5-kb fragment, which was sequenced (GenBank TM /EMBO accession number AF000232). The DNA sequence revealed an open reading frame with the potential to encode a protein of 877 amino acids (Fig. 2). Because of N-and C-terminal extensions, CaPmt1p is longer than any other Pmt proteins. The deduced Pmt1 protein has 49% identical residues compared with Pmt1p of S. cerevisiae (17). Identities to the ScPmt2-7 proteins were lower, amounting to 28, 27, 25, 42, 23, and 19% identical residues, respectively. As determined by the TMpred program, 4 CaPmt1p is a likely integral membrane protein containing 10 possible transmembrane regions. Asparagine residues at positions 83, 195, 395, 400, and 721 are potential sites for N-linked glycosylation.
CaPMT1 Complements the S. cerevisiae pmt1 Mutation-To test if CaPmt1p functions in S. cerevisiae, we constructed a low copy centromeric and a high copy S. cerevisiae vector containing CaPMT1 (plasmids pCT32 and pCT31, respectively). In addition, the CaPMT1 coding region was placed under transcriptional control of the GAL1 promoter in low and high copy expression vectors (pCH1 and pCH2, respectively). All plasmids were transformed into the S. cerevisiae pmt1 mutant B76⌬pmt1, and transformants were grown in YPGal (to induce the GAL1 promoter). Chitinase in the culture medium was adsorbed to chitin, separated by SDS-PAGE, and silver-stained (42). As expected, chitinase of the B76⌬pmt1 strain showed a faster electrophoretic mobility compared with the parental wild-type strain B76 (Fig. 3), which is due to a partial defect in O-glycosylation (17,42). In transformants carrying either the S. cerevisiae or the C. albicans PMT1 genes on plasmids, the wild-type migration of chitinase was restored. A complementation of the pmt1 defect was obtained by CaPMT1 present on low or high copy vectors and either containing the authentic promoter or the strong S. cerevisiae GAL1 promoter.
Thus, these results indicate that CaPMT1 is efficiently expressed in S. cerevisiae and that CaPmt1p is able to fully substitute the functions of ScPmt1p in S. cerevisiae.
pmt1 Strains Are Deficient in in Vitro Pmt Activity-The complementation of the S. cerevisiae pmt1 defect by CaPMT1 already had suggested that CaPMT1 encodes a functional Pmt protein. To provide further evidence for this assumption, we tested membrane proteins of the parental and disruptant strains for their ability to O-mannosylate a peptide substrate using Dol-P-Man as the activated sugar (29,43).
Membrane proteins of the wild-type strain SC5314 showed high levels of mannosylation activity, while the activity in the homozygous pmt1 mutant (CAP1-312) was low in comparison, amounting to 1-4% (Table II). Reintroduction of CaPMT1 into the pmt1 background restored in vitro O-mannosylation activity. The pmt1/PMT1 heterozygous strain showed an activity intermediate between the activities in the pmt1 mutant and wild-type cells, indicating an effect of gene dosage on enzymatic activity. Thus, the results of these in vitro enzymatic tests and the in vivo complementation (see above) strongly suggest that CaPMT1 encodes a functional Pmt protein.
CaPmt1p Affects Growth and Secreted Proteins-Strains carrying deleted PMT1 alleles were viable and did not show gross cytological defects. However, in contrast to the wild-type strain SC5314, the heterozygous strain CAP1-3, and the retransformed mutant CAP1-3121(pCT30), cells of the homozygous mutant CAP1-312, lacking both functional PMT1 alleles, formed aggregates (Fig. 5), which at culture densities A 600 Ͻ 1 could be separated into single cells by mild ultrasonic treatment using a bath sonifier. Once disaggregated, however, pmt1 cells remained separate and did not reaggregate, suggesting that aggregates of pmt1 strains arise because of a defect in cell separation after cell division rather than by autoaggregation of individual cells. Growth of the homozygous disruptant strain CAP1-312 was reduced compared with the other strains. For strains SC5314, CAP1-3, and CAP1-312, doubling times in SD or YPD medium were 100 or 70 min, 100 or 70 min, and 150 or 90 min, respectively. This phenotype differs from pmt1 mutants of S. cerevisiae, which are not compromised for growth and do not form aggregates in several genetic backgrounds (e.g. B76⌬pmt1). Simultaneous deletion of both PMT1 and PMT2 are required to generate a comparable aggregation phenotype in S. cerevisiae (16,19) as in the C. albicans pmt1 strains.
The Als1 protein of C. albicans has high homology to ␣-agglutinin of S. cerevisiae and is predicted to be O-and N-glyco-   3 and 4). The presence of a plasmid carrying CaPMT1 in the pmt1 background restored the wild-type pattern of Als1p migration (lane 5). These results are consistent with the assumption that Als1p is a partially O-glycosylated protein in C. albicans; conceivably, glycosylation by Pmt1p leads to the slower migrating protein species, while the absence of Pmt1p in cells reduces O-glycosylation, so that only the unglycosylated (or underglycosylated) Als1 protein is detected. Because the size of Als1p differs among strains (44) and its size in SC5314 and its derivatives is not known, it cannot be decided if the lower band of pmt1/pmt1 strains is underglycosylated or completely lacking O-glycosyl chains. Chitinase of S. cerevisiae is highly O-glycosylated, and homologues in C. albicans have been identified (41). To explore if the pmt1 mutation affects the production and/or activity of C. albicans chitinase, we tested its enzymatic activity in the medium and in association with cells (Table III). Approximately equal amounts of chitinase were found in the medium and cell-associated in the wild type (SC5314), in the heterozygous strains CAP1-3 and CAP1-5, and in the PMT1-reconstituted strains (CAP1-3121(pCT29); CAP1-3121(pCT30)). In contrast, the homozygous pmt1 mutant contained significantly less cellassociated and correspondingly more chitinase activity in the medium. Possibly, C. albicans chitinase is O-glycosylated, and underglycosylation favors its release and/or leads to an increased chitinase enzymatic activity in the medium; alternatively, reduced O-glycosylation of cell wall components may impair retention of chitinase in the cell wall. C. albicans chitinase cannot be adsorbed to chitin (41), and specific antibodies are not available, which at present does not allow a more detailed analysis of this phenotype.
We used immunoblots to determine if other described cell surface components are modified by CaPmt1p (data not shown). The monoclonal antibody 1B12 was described to recognize major cell wall proteins (36). In our immunoblots, monoclonal antibody 1B12 showed five proteins with molecular masses between 130 and 230 kDa; none of these proteins migrated differently in the Capmt1 mutant. An antibody against the S. cerevisiae gp115 cell wall protein, which also reacts with the C. albicans PHR1 gene product (40), recognized an identical 120-kDa protein in immunoblots of both the wild-type and Capmt1 mutant strains. An antibody directed against the S. cerevisiae Pma1 protein, which reacts with the corresponding C. albicans enzyme (38), revealed an identical 100-kDa protein band in immunoblots of PMT1 and pmt1 strains. Also, an antibody directed against the Cdr1p multidrug ABC transporter (39) revealed an identical 170-kDa protein in both PMT1 and pmt1 strains; an antibody directed against the Int1 protein, which is involved in adhesion and filamentous growth of C. albicans (37), revealed an identical 220-kDa protein in the PMT1 and pmt1 strains. In summary, these data indicate that only a subfraction of the cell surface components of C. albicans is extensively modified by the CaPmt1 enzyme. However, the possibility of minor modifications by Pmt1p, e.g. O-glycosylation at a single site not leading to migration differences in SDS-PAGE, cannot be excluded for any of the tested proteins.
Both CaPMT1 Alleles Are Required for Hyphal Morphogenesis-Hyphal development in C. albicans is induced on certain solid media after several days of growth (46) or within minutes by positive stimuli such as serum or GlcNAc (2). As expected, lateral hyphae began to emerge from colonies of the wild-type SC5314 and the retransformed mutant CAP1-3121(pCT30) after prolonged growth on "spider" plates (Fig. 7, A and D). In contrast, mutant CAP1-312 was unable to form hyphae (Fig.  7B), and the heterozygous strain CAP1-3 showed an intermediate phenotype (Fig. 7C). Only Ura ϩ strains were compared in this experiment, since Ura Ϫ strains are known not to undergo morphogenesis in the absence of uridine (47).
In contrast to these results, all strains including pmt1 mutants were able to form hyphae at similar rates on plates containing 5-20% serum or in liquid medium containing 5% serum or 2.5 mM GlcNAc (data not shown). Likewise, pseudohy-  6. The Als1 cell wall protein is affected by a pmt1 muta-tion. Strains were grown in SD medium to an A 600 ϭ 3-5, and proteins in cell extracts were separated by SDS-PAGE (5%). Blots were developed with anti-Als1 antibody (1:10,000), followed by anti-rabbit IgG coupled to horseradish peroxidase.  phal formation induced in YP medium at 37°C proceeded similarly in strain CAP1-312 as in the wild-type strain SC5314. These results suggest that the molecular machinery of hyphal formation is not affected in pmt1 mutants but that a component of a signaling pathway that is operative on spider medium is defective. Because a similar phenotype (no hyphal formation on spider medium, but normal hyphal formation on serum) was obtained for strains defective in components of a conserved mitogen-activated protein kinase pathway (34,48), we assume that this pathway is defective in pmt1 mutants. Because of these results, we examined if the PMT1 gene in S. cerevisiae is also necessary for pseudohyphal growth of this organism. To test this possibility, we disrupted the ScPMT1 gene in a haploid strain (B76), which was subsequently diploidized by transformation with a plasmid containing the HO gene (27). After the loss of the HO plasmid, we had generated strain B76d⌬pmt1 (pmt1⌬::URA3/pmt1⌬::URA3). This strain and strain B76/4 as a control (24) were each transformed with plasmid YCplac22RAS2 Val19 containing the dominant RAS2 Val19 allele (49), and transformants were tested on low nitrogen media as described by Gimeno et al. (28). All tested transformants of strains B76/4 and B76d⌬pmt1 had an identical phenotype, in that vigorous pseudohyphal growth was observed.
Thus, the ability to undergo filamentous morphogenesis is significantly affected by deletion of the Pmt1p-encoding gene in C. albicans, while lack of Pmt1p in S. cerevisiae has no effect. To test if PMT1 expression is up-regulated by hyphal formation, we induced hyphal formation of strain CAI4 containing control plasmid pRC2312 by 2.5 mM GlcNAc. At various times during hyphal induction, the percentages of cells carrying hyphae were determined, and total RNA of cells was prepared and analyzed by Northern blotting for the PMT1 transcript; ACT1 and rRNA were used as control RNAs (50) (data not shown). Although PMT1 transcripts showed a slight increase relative to rRNA during hyphal formation, a similar increase was also observed for the ACT1 transcript. Therefore, PMT1 expression is not tightly correlated to hyphal morphogenesis, and PMT1 is transcribed at relative constant levels during growth and morphogenesis, similar to the ACT1 housekeeping gene (50).

Pmt1
Mutants Are Supersensitive to Some Antifungals-Multiple pmt mutations are known to cause increased antifungal sensitivities in S. cerevisiae (19). Therefore, we tested susceptibility to several antifungal compounds of the C. albicans wild-type and the constructed heterozygous and homozygous pmt1 mutant strains (Fig. 8).
Hygromycin B at 100 g/ml completely blocked growth of the homozygous CAP1-312 mutant on solid media, while the heterozygous strain CAP1-3 (PMT1/pmt1) was affected partially at this concentration, leading to smaller colonies. An effect of heterozygosity was confirmed in liquid YPD at 150 g/ml hygromycin B, where the wild-type strain (SC5314), the heterozygous strain (CAP1-3), and the homozygous strain (CAP1-312) had doubling times of 130, 150, and Ͼ360 min, respectively. Another aminoglycoside antibiotic, G418, mainly prevented growth of the homozygous strain at 1.2 mg/ml G418, while the heterozygous strain was not affected significantly. Similarly, clotrimazole at 2 g/ml was found to significantly block growth of the homozygous disruptant strain while affecting the heterozygous and the wild-type strain less strongly. S. cerevisiae strains defective in cell wall biosynthesis are known to be supersensitive to calcofluor (51). Disruption of both PMT1 alleles was found to cause reduced plating efficiency on media containing calcofluor white (Fig. 8) and led to an inability to grow on YPD media containing low levels (0.06%) of the deter- Strains were grown on YPD-media at 30°C containing substances at the indicated concentrations; hygromycin B was added at 100 g/ml, and G418 was added at 1.2 mg/ml. Sensitivity to calcofluor white was tested by spotting decreasing cell counts (from left to right: 50,000, 5,000, and 500 cells) on YPD plates containing 0 or 10 g/ml calcofluor white. gent SDS; in contrast, the heterozygous strain and the wildtype strain were not supersensitive (data not shown). No differences in susceptibilities of all strains were observed for nystatine (1-15 g/ml), for amphotericin B (0.5-2 g/ml), for a combination of methotrexate and sulfanilamide (0 -150 g and 5 mg/ml, respectively (52)), for fluconazole (5 g/ml), for fluphenazine (50 g/ml), and for caffeine at up to 20 mM (19).
Thus, deletion of PMT1 causes a significant increase in the susceptibility of C. albicans to various antifungal compounds and compounds perturbing the cell wall. Supersensitivity for hygromycin B is observed even with the heterozygous CaPMT1/Capmt1-strain, indicating a gene dosage effect. These results contrast with the findings in S. cerevisiae, where deletions of several PMT genes are required to alter antifungal susceptibilities.
CaPmt1p Is Required for Adherence to Epithelial Cells-Conceivably, membrane and cell wall proteins that are Oglycosylated by CaPmt1p could be essential for C. albicans virulence, by mediating adherence to and migration across epithelial cell layers. To test this possibility, we assayed adherence of C. albicans strains to monolayers of Caco-2 epithelial cells (30,31) (Table IV). In this assay, fungal cells, which had been briefly dispersed by a mild ultrasonification, were placed on the epithelial monolayer for 45 min, after which nonadhering cells were removed by washing; numbers of tightly adhering cells were determined by growth in a YPD agar overlay.
The results demonstrate that the homozygous pmt1 mutant is significantly impaired in adhesion to the epithelial cells, while the heterozygous strain is not affected. Reintroduction of CaPMT1 into the pmt1 mutant background restores adhesion. Thus, a lack of Pmt1 activity has an important effect on a trait considered essential for virulence of C. albicans.
Both CaPMT1 Alleles Are Required for Virulence of C. albicans-To determine directly, if CaPMT1 is involved in pathogenicity of C. albicans, we infected mice (n ϭ 15) intravenously with wild-type and mutant strains, and the survival of animals was tested. Because of the aggregation behavior of C. albicans pmt1 mutants, strains were briefly sonicated prior to injection to ensure that animals were infected with equal cell numbers. Determination of plating efficiencies before and after sonification indicated that this treatment did not affect cellular viability.
Infection with the wild-type strain SC5314 resulted in rapid death (mean survival of 4 days), while deletion of both PMT1 copies (strain CAP1-312) led to complete loss of virulence (Fig.  9). These characteristics correlated with kidney colonization by C. albicans strains, which amounted to 1.5 ϫ 10 6 and 0 colonyforming units/kidney in the wild type and pmt1 mutant, respectively. Interestingly, although the heterozygous CAP1-3 strain almost showed wild-type kidney colonization (9.5 ϫ 10 6 colonyforming units), its survival was increased significantly (mean survival of 7 days; p Ͻ 0.0001). Thus, deletion of PMT1 alleles, in a dosage-dependent manner, decreases the virulence of C. albicans. The lack of kidney colonization suggests that strains lacking both PMT1 alleles are defective in systemic spread or that they are quickly eliminated by defense mechanisms, e.g. by phagocytosis.

DISCUSSION
The PMT1 gene of C. albicans is the first described PMT gene in a species other than S. cerevisiae whose function has been characterized. Other ScPMT homologues of unknown function are known in fungi. Recently, a genomic sequence of S. pombe has been reported, which encodes a putative Pmt protein (GenBank TM accession number Z99126). Also, a PMT1 homologue appears to exist in the yeast Kluyveromycis lactis. 5 Because we also obtained evidence for a homologue of the ScPMT2 gene in C. albicans, it appears that protein O-mannosylation by Pmt-type proteins is a reaction occurring generally in other fungi; this process may require two or more Pmt proteins. The sequence of the C. albicans PMT1 gene predicts a protein with a hydrophilic center and membrane-spanning termini, in agreement with the structures of the S. cerevisiae Pmt1 protein (17). Because we were unable to detect C. albicans homologues of the S. cerevisiae PMT3-7 genes, it is possible that C. albicans only contains two PMT genes. Conceivably, the two alleles of PMT1 and PMT2 provide sufficient genetic redundancy in the diploid yeast C. albicans, a function that is secured by multiple PMT loci in S. cerevisiae, which has a haploid or a diploid genome. The existence of only two PMT genes in C. albicans could also explain the drastic phenotypes of PMT1 deletion in C. albicans, as discussed below.
Although individual deletions of the S. cerevisiae PMT1 and PMT2 genes do not cause significant growth phenotypes, their simultaneous mutation significantly retards growth and causes cells to aggregate strongly (16). This phenotype is obtained in C. albicans by merely deleting both PMT1 alleles. Similarly, while S. cerevisiae pmt1 mutants do not show increased sensitivity toward hygromycin B, C. albicans pmt1 mutants are supersensitive, as is observed for S. cerevisiae pmt1 pmt2 double mutants. 6 Unlike defects in outer chain addition to Nglycosyl chains in S. cerevisiae, which also lead to increased hygromycin B sensitivity (53), the pmt mutations did not cause resistance to vanadate. Surprisingly, deletion of a single PMT1 allele in C. albicans sufficed to increase hygromycin B sensitivity, providing evidence that the dosage of PMT1 expression determines the wild-type phenotype. Haplo-insufficiency has been observed for other genes in C. albicans (48). 7 Supersensi-  9. Virulence of C. albicans strains. Survival of mice (n ϭ 15) infected with 10 6 C. albicans cells was determined. Strains SC5314 (ƒ; wild-type), CAP1-3 (Ⅺ; PMT1/pmt1) and CAP1-312 (⌬; pmt1/pmt1) were compared. tivity for calcofluor white is another example of the strong consequence of PMT1 deletion in C. albicans, which compares to the phenotype of S. cerevisiae pmt1 pmt2 double mutants (19). The calcofluor white phenotype suggests that O-glycosylation of unknown proteins by Pmt1p stabilizes the C. albicans cell wall. In our study, we identified the Als1 protein (44), which recently has been recognized as a potential adhesion factor for C. albicans (45), as the first likely O-glycosylated protein in the cell wall of C. albicans. Recent results suggest that PMT1 activity is required to incorporate several proteins into the cell wall of S. cerevisiae (54). O-Glycosylated proteins may also be involved in an unknown system of resistance to various agents, such as aminoglycoside antibiotics, e.g. by promoting their export (39). However, a defect in Pmt1 function does not cause a general or an equal increase in the susceptibility toward damaging agents, since the sensitivity toward clotrimazole is increased only slightly and is absent for nystatin, amphotericin B, and methotrexate/sulfanilamide. Despite the multiple drastic phenotypes of the pmt1 mutation in C. albicans, a significant effect on osmotic sensitivity, as has been observed for multiple pmt mutations in S. cerevisiae (19), was not detected. Thus, although Pmt1 fulfills similar basic fungions in both fungal species, the resulting phenotypes differ significantly.
Depending on environmental conditions, C. albicans and S. cerevisiae are able to either grow as a budding yeast or in a multicellular filamentous form (reviewed in Ref. 2). Dimorphism of C. albicans is considered an important virulence trait. Hyphal formation of this pathogen can be induced either on certain depauperated solid media (spider media), a process requiring components of a conserved mitogen-activated protein kinase cascade (46); alternatively, positive stimuli, such as serum, are able to induce morphogenesis. Deletion of PMT1 completely abolished the ability of C. albicans to form hyphae on spider media. Surprisingly, even the heterozygous pmt1/ PMT1 strain had a severe defect in morphogenesis, confirming the gene dosage effect observed for hygromycin B sensitivity and virulence in the animal model. Because serum still stimulated hyphae formation, we assume that not morphogenesis per se but rather sufficient levels of a component required for a signaling component operative on spider media requires Oglycosylation by Pmt1 for its function. A similar phenotype (block of hyphae formation on spider medium and normal hyphae formation on serum) has been reported for mutants defective in elements of a conserved mitogen-activated protein kinase pathway (34,48). Therefore, it is possible that an unidentified O-glycosylated protein of this pathway, e.g. a membrane protein involved in signal transduction, is defective in pmt1 mutants. In contrast to the results obtained in C. albicans, pmt1 deletion in S. cerevisiae did not affect its ability to form pseudohyphae on low nitrogen media. Thus, even with regard to morphogenesis, PMT1 has a more important function in C. albicans compared with S. cerevisiae.
Despite the relatively minor growth defect in most media, pmt1 mutants were completely avirulent in the mouse model of systemic infection. Even the pmt1/PMT1 heterozygous strain showed a significant decline in virulence, although this strain grew identically to the wild-type strain and could be distinguished only by increased sensitivity to certain agents and by reduced morphogenesis (see above). Thus, even a 2-fold drop in Pmt1 activity (assuming equal contribution of each allele to overall PMT1 expression) interferes with pathogenicity. It remains to be determined experimentally which of the described various phenotypes of pmt1 mutants is responsible for attenuation of virulence. Conceivably, the alteration of cell surface properties and the morphogenesis defect of pmt1 strains cause defects in adhesion and penetration of epithelial and endothelial cell layers. We obtained experimental support for this possibility by demonstrating that pmt1 mutants adhere less strongly than wild-type cells to a Caco-2 epithelial cell layer.
Thus, it appears that the ability to penetrate endothelial and epithelial layers, which is the precondition to infect organs (1), does not occur with pmt1 mutants. In agreement with this hypothesis, kidneys of infected animals were free of C. albicans pmt1 cells. Recently, the Als1 protein has been described as a likely adhesion factor for C. albicans (45). It appears possible that reduced O-glycosylation of this protein in pmt1 mutants, which is suggested here, reduces the adhesive properties of Als1p and thus affects virulence. Alternatively, the aggregation phenotype of pmt1 mutants could contribute to lower fungal dissemination in the infected animal. A third possibility is an increased host activity, e.g. phagocytosis, against pmt1 strains.
In conclusion, the characterization of the C. albicans PMT1 gene not only demonstrates that Pmt proteins exist and function in O-glycosylation in fungal species other than S. cerevisiae; the results also show that CaPmt1p assumes novel functions that are not observed (in hyphal morphogenesis and antifungal resistance) or are not considered relevant (in adherence and virulence) in S. cerevisiae. O-Glycosylation reactions mediated by Pmt proteins appear to be possible targets for antifungal agents for the following reasons. (i) The bulk of O-glycosylation occurs by different mechanisms in mammalian and fungal cells. (ii) Even a 2-fold reduction in PMT1 gene dosage causes defects in morphogenesis, as well as a significant drop in virulence. (iii) Reduced Pmt1p activity leads to an increased sensitivity to antifungal agents, thereby potentially amplifying any response to antifungals and offering the prospect for a combination therapy. (iv) Cells lacking Pmt1p activity show reduced adherence to host cells and reduced colonization of organs; this characteristic is in contrast to other avirulent C. albicans strains, which still show kidney colonization (55,56). On the other hand, rare O-mannosylation of some human proteins has been reported (11,12), and expressed sequence tags deposited at GenBank TM contain sequences encoding possible Pmt homologs derived from human (accession numbers AA670164, AA425494, AA496399, AA099791, N42494, and N35574), mouse (accession numbers AA274738, AA277592, AA274738, and Z31210), rat (accession number H33465), C. elegans (accession numbers C41029, C47075, C42623, C66447, and C69714), and rice (accession number C73075) cells. It is possible (although this remains to be established) that the corresponding genes perform essential functions in the respective cell types. In D. melanogaster, a gene encoding a Pmt homologue appears to be required for muscle development (20).