An Adhesin of the Yeast Pathogen Candida glabrata Mediating Adherence to Human Epithelial Cells

See allHide authors and affiliations

Science  23 Jul 1999:
Vol. 285, Issue 5427, pp. 578-582
DOI: 10.1126/science.285.5427.578


Candida glabrata is an important fungal pathogen of humans that is responsible for about 15 percent of mucosal and systemic candidiasis. Candida glabrata adhered avidly to human epithelial cells in culture. By means of a genetic approach and a strategy allowing parallel screening of mutants, it was possible to clone a lectin from a Candida species. Deletion of this adhesin reduced adherence of C. glabrata to human epithelial cells by 95 percent. The adhesin, encoded by the EPA1 gene, is likely a glucan–cross-linked cell-wall protein and binds to host-cell carbohydrate, specifically recognizing asialo-lactosyl–containing carbohydrates.

Candida species are responsible for more than 8% of all hospital-acquired infections (1); the two most frequently encountered species areC. glabrata and C. albicans (2).Candida albicans is asexual and diploid, which complicates genetic analysis in this organism because both copies of a gene must be knocked out to uncover a recessive phenotype. Analysis of virulence inC. albicans has, therefore, been limited largely to reverse genetic approaches in which both copies of individual cloned genes are deleted and the resulting phenotype is assessed. Candida glabrata, although asexual, is haploid (3), which facilitates genetic analysis. In C. glabrata, it is possible to generate random mutants and screen for phenotypes of interest (4). Here, we demonstrate that this forward genetic approach can be used to analyze the host-pathogen interaction inC. glabrata and use this approach to identify an adhesin mediating adherence of C. glabrata to host epithelial cells.

The adherence of Candida to host cells has been the subject of intense investigation, and in the case of C. albicans, the yeast expresses a number of adhesins capable of interacting with a variety of ligands, including proteins [reviewed in (5)] and carbohydrates (5–8). Recently, it has also been shown that Hwp1p, a hypha-specific protein, is a substrate for mammalian transglutaminases and mediates covalent attachment ofC. albicans to human buccal epithelial cells (9).

We found that C. glabrata adheres strongly to human epithelial cells in culture. In our assay (10), with a multiplicity of infection (MOI) of 1:1, between 10 and 20% of added yeast adheres to a monolayer of the human laryngeal carcinoma cell line HEp2 compared with 0.1% of added yeast forSaccharomyces cerevisiae (11). Scanning electron micrographs of C. glabrata bound to the surface of the monolayer show a marked and intimate interaction between the epithelial cell filopodia and the yeast cell (Fig. 1) (12). In transmission electron micrographs (13), a similar tight association is seen between the surface of the yeast cell and the surface of the epithelial cell, suggesting that the host ligand is broadly distributed on these tissue culture cells. This interaction is dependent on Ca2+ because adherent yeast can be removed with EGTA or with EGTA titrated with Mg2+ but not with EGTA titrated with Ca2+ (13).

Figure 1

Scanning electron micrograph (12) ofC. glabrata adhering to cultured HEp2 cells. Scale bar, 1 μM.

To identify the yeast gene mediating the interaction of C. glabrata with epithelial cells, we undertook a mutant screen. We implemented a number of genetic tools to facilitate this analysis. First, we used a ura3 deletion strain congenic with a virulent clinical isolate (4). Second, we used a variation of signature-tagged mutagenesis (14), a strategy permitting parallel screening of multiple mutants in a single complex pool of mutants. In this strategy, each mutant in a pool of mutants also carries a unique sequence tag flanked by constant polymerase chain reaction (PCR) priming sites that permit the amplification of all the tags in a pool in a single PCR amplification. The fate of individual mutants is mirrored by the fate of their cognate oligonucleotide tags and is followed by means of hybridization to a membrane on which all the tags have been arrayed. We generated 96 strains of C. glabrata by integrating 96 different sequence tags in the already disrupted URA3 locus (15). In control experiments, differences of twofold to threefold in representation were easily detected (13). Finally, we used a method of insertional mutagenesis that was based on our observation that C. glabrata had a very efficient system of nonhomologous integration. These insertions were randomly distributed at the genome level and gave rise to auxotrophs at a frequency of 0.25% (4).

To generate a mutant library, we transformed each of the 96 strains with linearized YipLac211 [a plasmid carrying the S. cerevisiae URA3 gene (16)] and isolated 100 Ura+ transformants for each of the 96 tagged strains. This collection of 9600 mutants was assembled into 100 pools of 96 in which each pool consisted of a complete set of the 96 tagged strains.

Pools of mutants were allowed to adhere to duplicate monolayers of human cultured epithelial cells (HEp2 cells) in a standard adherence assay (10). The adherent yeast and the input pool were recovered and grown on yeast extract, peptone, and dextrose (YPD) plates overnight. Genomic DNA was prepared from the yeast in the input and two adherent output pools, and the sequence tags were amplified by PCR. The amplified tags were labeled with [α-32P]deoxycytidine triphosphate and were used to hybridize to triplicate membranes on which all 96 tags were immobilized. The filters were analyzed with a phosphorimager, and nonadherent mutants were detected by the presence of the cognate tag in the input pool and its absence in the output pools (Fig. 2A). We screened 50 pools comprising 4800 mutants and isolated 31 mutants that were altered in their adherence to HEp2 cells. Among these mutants, five were hyperadherent (13) and 16 were totally nonadherent (Fig. 2B and Fig. 2A, strain g6). For 10 mutants, adherence was reduced but not eliminated (13).

Figure 2

(A) Signature tag representation in input and recovered adherent output populations of C. glabrata mutant pool 15. Triplicate filters with identical sets of 96 immobilized signature tags. Each filter was probed with labeled PCR product derived from amplification of the tags present in a population of yeast cells. Filter 1, input population of cells after overnight growth in YPD. Filter 2, population adherent to HEp2 cells. Filter 3, population adherent to HEp2 cells in duplicate adherence assay. (B) Adherence of C. glabrata mutants. Individual mutants were reconstructed in parent strain and tested in the standard adherence assay (10)Candida glabrata strain BG2 is the wild-type parent of all of the mutants. The assays were carried out in triplicate.

To identify the loci disrupted in the nonadherent mutants, we cloned the DNA flanking the Yiplac211 insertions. Genomic DNA for each mutant strain was digested with Eco R1 (which does not digest in the Yiplac211 plasmid sequence), circularized with ligase, and transformed intoEscherichia coli. This resulted in recovery of a plasmid consisting of Yiplac211 and the genomic DNA flanking the original insertion site. When we reintroduced the Yiplac211 plasmid into the original site of insertion in the parental strain (17), the regenerated mutants were also nonadherent, demonstrating that the insertion was responsible for the mutant phenotype. Sequence analysis of the DNA flanking the insertion site for 16 nonadherent mutants revealed that 14 of the 16 insertion mutants were in the same 1.4-kb locus (Fig. 3A). These insertions were in the noncoding region upstream of a large open reading frame (ORF), reflecting the fact that illegitimate recombinants in C. glabrata are directed almost exclusively to noncoding regions (4). The sequence of the gene downstream of these 14 insertions revealed an ORF encoding a 1034–amino acid protein (18). We have named this gene EPA1, for Epithelial Adhesin 1. Deletion of the coding region of EPA1rendered the yeast nonadherent (Table 1). Restoration of the EPA1 gene, either on a plasmid or by integration at the natural EPA1 locus, restored adherence, showing that the nonadherent phenotype of the epa1-null strain is due to loss of EPA1 (Table 1).

Figure 3

(A) Physical map of 14 insertions at the EPA1 locus. Insertion sites are marked with vertical hatch marks. The ORF is shown as a hatched box. A “T” marks a consensus TATA box. bp, base pairs. (B) Structure of theEPA1 protein. Filled black boxes are hydrophobic stretches. The open box is highly enriched for Ser (S) and Thr (T); the hatched box has slight homology to FLO1 from S. cerevisiae (34, 35). aa, amino acid. (C) The direct repeat region of EPA1. (D) Alignment of amino acid sequence of EPA1 andFLO1. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Table 1

EPA1 mediates adherence of yeast cells to epithelial cells. Candida glabrata strain: BG2 (4). Candida glabrata epa1Δ: BG176 and BG178, two independent deletions of the EPA1 coding region.EPA1 restored: BG184 and BG186, independent restorations of the EPA1 genomic locus in deletion strains BG176 and BG178 with a two-step replacement strategy (31).Saccharomyces cerevisiae: strain BY4741 (MATaura3Δ0 leu2Δ0 his3Δ0, LYS2 met15Δ0).Saccharomyces cerevisiae + pEPA1: BY4741 transformed with pEPA1 (36). Data shown are for adherence to the CHO-derived epithelial cell line Lec2. Essentially identical results were obtained for adherence to HEp2 cells. All strains were tested in triplicate.

View this table:

Expression of EPA1 in S. cerevisiae permittedS. cerevisiae to adhere as well as C. glabrata to epithelial cells (Table 1). Thus, EPA1 was sufficient to mediate adherence in the context of the S. cerevisiae cell. In its domain structure (Fig. 3B), this protein is a member of a large family of cell-wall proteins in yeast. Amino acids 6 to 24 (12 of 19 hydrophobic) are a consensus signal sequence, and the hydrophobic amino acids 1020 to 1033 at the COOH-terminus are a consensus hydrophobic signal for addition of a glycosyl-phosphatidylinositol (GPI) anchor. Asn1011, which is followed by two small amino acids Ala1012 and Ile1013, is the likely addition point for the GPI anchor. The COOH-terminal two-thirds of the protein (approximately amino acids 331 to 944) is highly enriched for serine and threonine (43.8%); between amino acids 580 and 700 are three direct repeats of the 40–amino acid motif shown in Fig. 3C. The NH2-terminal domain has slight homology with the S. cerevisiae flocculin FLO1, a Ca2+-dependent lectin (Fig. 3D).

As deletion of the EPA1 gene reduced adherence by 95%, we conclude that overall adherence of wild-type C. glabrata in our assay is largely a reflection of EPA1-mediated binding. We characterized the host ligand requirements for EPA1binding. As mentioned above, adherence of C. glabrata to epithelial cells required Ca2+. Binding of C. glabrata to epithelial cells was totally inhibited by 10 mM galactose or lactose but not by 10 mM sialyl-lactose or 100 mM of a number of other sugars or by solutions (1 mg/ml) of various glycoconjugates (Table 2). The concentration at which lactose or N-acetyl lactosamine (LacNAc) inhibited 50% of C. glabrata binding to epithelial cells was 1.25 to 1.5 mM, consistent with LacNAc being closely related to the natural ligand. Adherence of S. cerevisiae cells expressing EPA1 was also dependent on Ca2+ and was 50% inhibited by 1.5 mM lactose (13), suggesting thatEPA1 function in S. cerevisiae closely mirrors its function in C. glabrata.

Table 2

Inhibition of adherence of C. glabrata to HEp2 cells by saccharides. The concentration of glyconjugate at which adherence is 50% of adherence in the absence of glycoconjugate is given. NANA, N-acetyl neuraminic acid (sialic acid); GlcNAc, N-acetyl glucosamine; GalNAc,N-acetyl galactosamine.

View this table:

These data suggested either that EPA1, likeFLO1, is a Ca2+-dependent lectin or thatEPA1 encodes a glycoprotein whose carbohydrate modifications are recognized by a host lectin. To distinguish between these possibilities, we determined whether adherence was affected by treatment with sodium periodate (NaI), which cleaves sugar rings (19). Adherence was eliminated by pretreatment of the HEp2 cell with NaI, whereas treatment of the yeast cell with NaI did not affect adherence (13). This is consistent withEPA1 being a lectin and the host ligand being carbohydrate. To further characterize host requirements for binding byEPA1, we analyzed binding (20) of C. glabrata to chinese hamster ovary (CHO) epithelial cells and to a number of glycosylation-deficient CHO derivative lines (21).Candida glabrata was as adherent to CHO cells or to the glycosaminoglycan (GAG)–deficient line pgsA745 as to HEp2 cells. By contrast, binding to the sialic acid–deficient line Lec2 was increased by fivefold to 10-fold (Fig. 4). In Lec2 cells, the terminal glycosylation of complex N-linked carbohydrates is N-acetyl lactosamine (Galβ1-4 GlcNAc). In cells with reduced levels of surface Gal or GlcNAc (Lec1 and Lec8), binding of C. glabrata was reduced to background levels, equivalent to binding by theepa1-null strain.

Figure 4

Adherence (in percent) of Candidaspecies to cultured CHO cells. Adherence (10, 20) of C. glabrata strains BG2 (EPA1) and BG178 (epa1Δ) to CHO cells and derivative glycosylation-deficient CHO cells. The strains are CHO (ATCC CRL-1781); 745 GAG deficient: pgsA745(ATCC CRL-2242); Lec2 sialic acid deficient (ATCC CRL-1736); Lec1 GlcNAc deficient (ATCC CRL-1735); and Lec8 galactose deficient (ATCC CRL-1737). All experiments were performed in triplicate.

We compared the virulence of the EPA1 andepa1Δ strains in two murine models of mucosal infection (22). In vaginal or gastrointestinal (GI) tract infections (23) with an inoculum of between 106 and 108 C. glabrata, we could find no difference between EPA1 and epa1Δ strains in initial colonization or subsequent persistence (13). In infections of animals with a mixture of oligo-tagged EPA1 andepa1Δ strains, there was no detectable difference in the tag representation (and by extension the yeast strains) either immediately after colonization (days 1 to 2) or in persistently infected animals (days 9 to 15) (13).

Our data show that EPA1 is required for efficient in vitro adherence to human epithelial cells in culture. Elimination of its function reduced in vitro adherence by 20-fold.EPA1-mediated adherence was inhibited by lactose andN-acetyl-lactosamine, but not by sialyl-lactose. Furthermore, adherence was markedly increased by reduction of surface sialylation, consistent with recognition of an asialo-lactosyl-glycoconjugate. The ligand specificity forEPA1 is related to the specificity of a number of differentCandida and fungal lectins. Specifically, Krivan's group has shown that phylogenetically diverse fungi express a lectin specific for lactosyl ceramide (Galβ1-4Glcβ1-1Ceramide) (7). Irvin's group has shown that stationary phaseC. albicans expresses a lectin activity that preferentially binds to asialo-GM1 (gangliotetraosylceramide: Galβ13GalNAcβ1-4Galβ1-4Glcβ1-1Ceramide) (6,24). Cameron and Douglas have shown that stationary phase C. albicans expresses a lectin activity that likely binds to the type 2 H blood group antigen (Fucα1-2)Galβ1-4GlcNAc (25). These three carbohydrate ligands share common features with the ligand for Epa1p, having as their core component Galβ1-4 GlcNAc or Galβ1-4 Glc, both of which are good ligands for Epa1p. This potential conservation in ligand specificity may point to homologous adhesins in C. albicans and other pathogenic fungi. Alternatively, it may reflect convergent evolution of lectin activities specific for conserved glycoconjugates on the mammalian epithelial cell surface.

A number of adhesins have been described in C. albicans that contribute to overall adherence [reviewed in (5)]. Four genes mediating adherence of C. albicans to unknown ligands on human cultured cells have been cloned (9, 26), although their relative contribution to adherence is unknown. A similar redundancy of adhesins probably exists in C. glabrata and might explain why theEPA1-null mutant shows no phenotype in vivo especially given that this adhesin is responsible for 95% of in vitro adherence. Alternatively, a high-affinity ligand present on HEp2 cells and CHO cells may not be present on the tissues colonized in the models we examined. A more complete description of the total adhesin complement of C. glabrata, as well as C. albicans, will be a first step in understanding the role of adherence in host colonization and persistence.

  • * Present address: Johns Hopkins Medical Institute, Department of Molecular Biology and Genetics, 725 N. Wolfe Street, Baltimore, MD 21205, USA.

  • To whom correspondence should be addressed. E-mail: bcormack{at}


View Abstract

Navigate This Article