Development of Saccharomyces cerevisiae as a Model Pathogen: A System for the Genetic Identification of Gene Products Required for Survival in the Mammalian Host Environment

Corresponding author: John H. McCusker, Department of Microbiology, 3020, Duke University Medical Ctr., Durham, NC 27710., mccus001{at}mc.duke.edu (E-mail)

Communicating editor: A. P. MITCHELL

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION LITERATURE CITED

Saccharomyces cerevisiae, a close relative of the pathogenic Candida species, is an emerging opportunistic pathogen. An isogenic series of S. cerevisiae strains, derived from a human clinical isolate, were used to examine the role of evolutionarily conserved pathways in fungal survival in a mouse host. As is the case for the corresponding Candida albicans and Cryptococcus neoformans mutants, S. cerevisiae purine and pyrimidine auxotrophs were severely deficient in survival, consistent with there being evolutionary conservation of survival traits. Resistance to the antifungal drug 5-fluorocytosine was not deleterious and appeared to be slightly advantageous in vivo. Of mutants in three amino acid biosynthetic pathways, only leu2 mutants were severely deficient in vivo. Unlike the glyoxylate cycle, respiration was very important for survival; however, the mitochondrial genome made a respiration-independent contribution to survival. Mutants deficient in pseudohyphal formation were tested in vivo; flo11 mutants were phenotypically neutral while flo8, tec1, and flo8 tec1 mutants were slightly deficient. Because of its ease of genetic manipulation and the immense S. cerevisiae database, which includes the best annotated eukaryotic genome sequence, S. cerevisiae is a superb model system for the identification of gene products important for fungal survival in the mammalian host environment.

SACCHAROMYCES cerevisiae is closely related to the pathogenic Candida species (BOWMAN et al. 1992 ; LOTT et al. 1993 ), which are clinically the most commonly observed of the pathogenic fungi (BECK-SAGUE and JARVIS 1993 ). It is therefore not surprising that S. cerevisiae is observed clinically at a variety of different body sites and in different patient types and is an emerging opportunistic pathogen (HAZEN 1995 ; MURPHY and KAVANAGH 1999 ). In contrast to laboratory and other nonclinical S. cerevisiae strains, clinically derived strains of S. cerevisiae have characteristics that resemble those found in more commonly observed pathogenic fungi, such as profuse pseudohyphal formation and improved growth at high temperature (MCCUSKER et al. 1994A , MCCUSKER et al. 1994B ). Most significantly, experimental infections showed that clinical isolates and clinically derived strains of S. cerevisiae proliferate and persist in immunocompetent, outbred (CD-1) mice (CLEMONS et al. 1994 ) and kill complement factor five-deficient mice (BYRON et al. 1995 ). This previous characterization laid the initial foundation for using S. cerevisiae as a model to identify conserved gene products critical for fungal survival in the host environment.

The work described here further develops S. cerevisiae as a model system to identify highly conserved signal transduction and metabolic pathways critical for fungal survival in the host environment. Since the importance of many gene products depends upon the environment, one of the fundamental steps in genetic analysis is the identification of gene products that are important for survival in specific environments. This is particularly true for the difficult issue of survival of pathogens in the host environment. Genetic analysis can identify which of these pathways are necessary for in vivo survival and thereby greatly improve our understanding of fungal pathogenesis.

To test specific hypotheses regarding survival in the host environment, dominant drug-resistance markers (DDRM; WACH et al. 1994 ; GOLDSTEIN and MCCUSKER 1999 ) were used to delete specific genes in a clinically derived S. cerevisiae strain, thereby ablating different metabolic processes. Using different DDRMs, differentially marked strains were pooled into single inocula to compare their relative ability to survive in the mouse host. These infection pools always included a diploid strain in which one copy of the HO locus, a gene that is not expressed in diploids (HERSKOWITZ et al. 1992 ), was replaced with a DDRM, thereby allowing in vivo survival competition experiments between wild-type (HO/ho::DDRM) and mutant strains. Accurate determinations of survival phenotypes were possible because wild-type and mutant strains, which could be readily distinguished, were in competition and therefore simultaneously exposed to the same environment.

As an initial test of the system and the generality of fungal gene products required for survival in the host environment, S. cerevisiae ade2 and ura3 mutants were found to be deficient in survival, as is the case for the equivalent Candida albicans and Cryptococcus neoformans mutants (KIRSCH and WHITNEY 1991 ; VARMA et al. 1992 ; PERFECT et al. 1993 ). Cytosine deaminase deficiency, which confers resistance to the antifungal agent 5-fluorocytosine, was not deleterious and appeared to be slightly advantageous in vivo. The histidine and tryptophan biosynthetic pathways made only a slight contribution to survival while the leucine biosynthetic pathway was important for survival in mice. Different aspects of energy metabolism were examined for their contribution to survival in the host environment. Isocitrate lyase (icl1) mutants were only slightly deficient in survival. In contrast, respiratory-deficient mutants (cox15, ⁰) were severely deficient. Finally, the contribution of dimorphism to survival was assessed; flocculin-deficient flo11 mutants were phenotypically neutral while the transcription factor-deficient flo8, tec1, and flo8 tec1 mutants, which are required for expression of FLO11, were slightly deficient in survival in vivo.

The ease of genetically manipulating S. cerevisiae, relative to the more commonly observed pathogenic fungi, along with the immense database and highly annotated genome sequence of this species, makes S. cerevisiae a powerful, genetically tractable model system to identify pathways required for fungal survival in the mammalian host environment.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION LITERATURE CITED

Strains:
All S. cerevisiae strains used in these experiments are listed in Table 1 and were isogenic with YJM145, a homothallic, clinically derived strain. YJM145, and strains isogenic with YJM145, have been extensively characterized genetically (MCCUSKER et al. 1994A ), phenotypically (MCCUSKER et al. 1994B ), molecularly (WINZELER et al. 1998 , WINZELER et al. 1999 ), and in experimental infections (CLEMONS et al. 1994 ; BYRON et al. 1995 ). MATa/MAT diploid strains were used in all experimental infection experiments because, unlike diploid YJM145 background strains, haploid YJM145 background strains exhibited severe clumping, which precluded the accurate determination of colony-forming units (A. L. GOLDSTEIN and J. H. MCCUSKER, unpublished data); in addition, S. cerevisiae diploids, but not haploids, were observed clinically (MCCUSKER et al. 1994A ; CLEMONS et al. 1997 ; MCCULLOUGH et al. 1998 ). Unless otherwise indicated, all strains were HO/HO.

Media:
Yeast extract-peptone-dextrose (YEPD) medium (SHERMAN et al. 1974 ) was used for nonselective growth of yeast strains. To select for strains containing dominant drug-resistance cassettes, YEPD was supplemented with either 100 µg/ml nourseothricin, 200–400 µg/ml hygromycin B, or 200–300 µg/ml G418, resistance to which is conferred by the natMX, hphMX, and kanMX dominant drug-resistance cassettes (WACH et al. 1994 ; GOLDSTEIN and MCCUSKER 1999 ), respectively. The 5-fluorocytosine resistance phenotype of fcy1 mutants was tested on YEPD medium containing 10^-2 M 5-fluorocytosine, which was added to autoclaved medium. The ability to respire was determined by the ability to grow on yeast extract-peptone-ethanol-glycerol medium, which contains the nonfermentable carbon sources ethanol and glycerol, each at 2% final concentrations; ethanol was added after autoclaving. Auxotrophic requirements were determined by growth on synthetic defined media (SHERMAN et al. 1974 ). Pseudohyphal formation was tested on synthetic low ammonium dextrose (SLAD) medium (GIMENO et al. 1992 ). Solid media contained 2% agar.

Gene disruptions and strain construction:
All gene replacements were performed by PCR-mediated gene replacement (WACH et al. 1994 ) with plasmids pFA6kanMX4 (WACH et al. 1994 ), pAG25 (natMX4), and pAG32 (hphMX4; GOLDSTEIN and MCCUSKER 1999 ) as the sources of the gene knock-out cassettes. Primer sequences, with their specific applications, are given in Table 2. To generate targeted gene disruptions, dominant drug-resistance cassettes were PCR amplified in 50-µl reactions containing 50 mM KCl, 20 mM Tris-HCl (pH 8.4), 1.5 mM MgCl₂, 0.05% Tween 20, 0.2 mM dNTPs, 100 µg/ml acetylated bovine serum albumin, 0.5 µM sense primer, 0.5 µM antisense primer, 10 ng DNA template, and 5 units Taq DNA polymerase (GIBCO BRL, Grand Island, NY). Because of the high GC content of the nat1 open reading frame (ORF), PCR amplification reactions of the natMX cassette were supplemented with 5% DMSO. Amplification of the cassettes was initiated with a 1-min 94° denaturation, followed by 30 amplification cycles of 94° 1 min, 55° 1 min, 72° 3 min, and terminated with a 20-min 72° extension. Colony PCR conditions (NIEDENTHAL et al. 1996 ) were used to amplify genomic sequences, including DDRMs already integrated into the yeast genome. Twenty-microliter reactions contained 50 mM KCl, 20 mM Tris-HCl (pH 8.4), 2.5 mM MgCl₂, 0.2 mM dNTPs, and 1 unit Taq DNA polymerase. Cells from a 2- to 4-day-old colony were added to the reaction before PCR amplification was initiated with a 5-min 94° denaturation; this was followed by 40 amplification cycles of 94° 25 sec, 55° 30 sec, 72° 1 min/kb, and terminated with a 72° extension for 7 min.

Between 500 ng and 2 µg of PCR product was used to transform S. cerevisiae strains using either the lithium acetate/dithiothreitol electroporation (THOMPSON et al. 1998 ) or lithium acetate (GIETZ et al. 1995 ) methods. Before plating transformants onto selective media, the cells were grown for 4 to 6 hr in 1 M sorbitol at 30° on a rotator to allow for expression of the transformed drug-resistance marker. To distinguish between abortive transformation events and true gene disruption-integration events, the initial transformation plates were replica plated to fresh antibiotic plates when colonies were 2 mm in diameter. Homologous integration of dominant drug-resistance cassettes was verified by colony PCR (NIEDENTHAL et al. 1996 ) as described above.

Deletion mutations marked by dominant drug-resistance markers were made by transforming a diploid YJM145 background strain to generate deletion heterozygotes. To generate deletion homozygotes for testing in experimental infections, deletion heterozygotes were sporulated at 30°, tetrads were dissected, and the phenotypes of both the dominant drug-resistance marker and the gene deletion mutation were determined. Every deletion homozygote used in this study was isolated from deletion heterozygotes, the tetrads of which showed absolute 2:2 cosegregation of the dominant drug-resistance marker and the deletion mutation.

Experimental infections:
Systemic infections were established in 4- to 5-week-old male CD-1 mice (outbred, immune competent; Charles River Laboratory, Wilmington, MA) as described previously (CLEMONS et al. 1994 ) by injection of the lateral tail vein with 2 x 10⁷ yeast cells suspended in phosphate buffered saline (PBS; pH 7.4); mice were provided food and water ad libitum throughout the course of the experiments. In most instances, the inoculum was composed of a pool of three isogenic YJM145 background strains, each differentially marked with dominant drug-resistance cassettes. Pools of three strains were established by first growing each strain separately in YEPD at 30° to mid-log phase and determining the concentration of cells by hemocytometer count. Approximately equal numbers of cells of each strain were then combined in PBS to produce a total cell concentration of 2 x 10⁸ cells/ml. Viable cell counts were also determined for the inocula by plating serial dilutions onto antibiotic-containing plates.

To recover yeast cells from infected mice, the mice were euthanized in CO₂ and brain tissue was dissected from each animal; each brain was homogenized in 5 ml PBS containing 100 µg/ml ampicillin and streptomycin, for 20 sec with a Tekmar T25 homogenizer with an 8-mm probe at 19,000 rpm. Each brain homogenate was centrifuged for 10 min at 700 x g, the supernatant was discarded, and the pellet was resuspended in 1–2 ml PBS containing 100 µg/ml ampicillin and streptomycin. For each strain with which a mouse was infected, 100 µl of supernatant, or appropriate dilutions of the supernatant, was plated onto three of the appropriate antibiotic-containing plates, and the colonies on the plates were counted. Because at least the wild-type control in every competition was present on average at >10⁵ colony-forming units/brain at 14 days postinfection, dilutions were made such that each of the three plates usually contained 200 or more colonies. The number of colonies of the three plates was added and this number was used to determine the ratio of each mutant to the wild-type control within the homogenate from each animal.

Competitive indices:
The ability of mutant strains to compete with wild type was quantified by competitive indices, a well-established method in bacterial (e.g., CHIANG and MEKALANOS 1998 ) and fungal (BROWN et al. 2000 ) pathogenesis studies. For each mutant, competitive indices at a given time postinfection (i.e., t = x), defined as (mutant/wild type)_t=x/(mutant/wild type)_t=0, were calculated separately for each of five infected mouse brains; these five values were then averaged to give the competitive indices shown in Table 3 and Table 4. As determined previously for experimental infections in CD-1 mice, relative virulence was defined as the relative ability of S. cerevisiae strains to survive in the brain (CLEMONS et al. 1994 ).

To determine in vitro competitive indices, YAG40 (HO/ho::hphMX4) and two mutant strains (orf::natMX4/orf::natMX4 and orf::kanMX4/orf::kanMX4) were grown separately to 1.0 OD₆₀₀ and then pooled in equal amounts before adding 10⁴ cells of the pool to 1 liter of YEPD. The pooled cells were then cultured at 37° with shaking until the OD₆₀₀ of the culture was 1 x 10⁸ cells/ml at which point 0.5 ml of each culture was passaged to 1 liter of fresh YEPD. In total, each pool was passaged through 3 liters of YEPD. After each culture reached 1 x 10⁸ cells/ml, a sample was diluted and plated to YEPD plates containing either hygromycin B, G418, or nourseothricin to determine the composition of the population. After the colonies were counted, competitive indices were calculated as described above. Population doublings were calculated using the formula (final cell concentration)/(initial cell concentration) 2ⁿ, where n is approximately equal to the number of population doublings.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION LITERATURE CITED

Experimental design:
Survival competition experiments were performed by pooling wild-type and mutant strains into a single inoculum. In most instances, the inoculum contained three diploid strains, each containing a unique DDRM integrated into the yeast genome. Because each mouse was infected with a population, each mouse constituted a separate, self-contained experiment where two mutants were assayed for survival relative to a wild-type control.

In every case, one of the three diploid strains was heterozygous at the HO locus where one of the HO ORFs was replaced with a DDRM—HO/ho::DDRM. The decision to use HO/ho::DDRM strains as wild-type controls was based on three factors. First, HO is expressed only in haploids (HERSKOWITZ et al. 1992 ) and all experimental infections in this work were carried out with diploids. Second, mutations at the HO locus have been shown to be phenotypically neutral in vitro (BAGANZ et al. 1997 ). Finally, HO/ho S. cerevisiae strains have been previously tested in experimental infections and were fully virulent (CLEMONS et al. 1994 ).

Since the DDRMs had no effect on the survival of S. cerevisiae in mice, when compared to each other or to an unmarked isogenic strain (see below), the HO/ho::DDRM strains served as the wild-type control in the inocula. The inocula also contained two mutants, each mutant being homozygous for a deletion of the same gene but each mutant strain containing a different DDRM; this redundancy served as an internal control for consistency. For each experiment, systemic infections were initiated by lateral tail-vein injection of 15 4- to 5-week-old male CD-1 mice with a total of 2 x 10⁷ cells, which was previously determined to be the optimum inoculum dose for S. cerevisiae infections in CD-1 mice (CLEMONS et al. 1994 ). After 4 hr, during which time the animals cleared most of the yeast cells from their blood streams, 5 mice were sacrificed; of the remaining mice, 5 were euthanized at 7 days and 5 at 14 days postinfection. Brains were removed from euthanized mice and the brain tissue was homogenized. The homogenates, or dilutions of the homogenates, were plated onto different antibiotic plates, which were incubated at 30° for 2 to 3 days until colonies formed. By counting the colonies on the different antibiotic plates, the composition of the population was determined from each mouse and expressed as competitive indices. The average number of colony-forming units per brain, at least for the wild-type control at the 4-hr, 7-day, and 14-day postinfection timepoints, was similar to those previously reported (CLEMONS et al. 1994 ), being between 1 x 10⁵ and 3 x 10⁵ colony-forming units per brain. Because of the high number of colony-forming units of the wild-type control, it was possible to readily determine both slight and severe competitive defects in mutants.

Neutrality of dominant drug-resistance markers in vivo:
The first experiments established that the hygromycin B, G418, and nourseothricin antibiotic resistance cassettes were neutral for survival in mice, as they were in in vitro competition experiments in liquid media (BAGANZ et al. 1997 ; GOLDSTEIN and MCCUSKER 1999 ). Infection was carried out with a pool of three strains, YJH3 (HO/ho::natMX4), YJH7 (HO/ho::kanMX4), and YJH10 (HO/ho::hphMX4). Between 4 hr and 14 days postinfection, the percentage of each strain within the population remained constant (for YJH3, from 33.0 ± 1.8% to 31.8 ± 3.5%; for YJH7, from 32.1 ± 1.3% to 32.1 ± 2.9%; for YJH10, from 35.0 ± 1.3% to 36.1 ± 3.9%), showing that none of the DDRMs conferred a survival advantage or disadvantage to S. cerevisiae in mice. In two additional repetitions of this experiment, the largest change seen between 4 hr and 14 days postinfection was an increase from 35.0 ± 2.0% to 39.7 ± 4.0% for YJH7.

To get an idea for the noise in the system, the data from the in vivo competitions between YJH3, YJH7, and YJH10 were used to calculate competitive indices where, in successive calculations, one of the three strains was viewed as wild type and the other two strains as mutant. The lowest competitive index calculated from these wildtype strain competitions was 0.75 and the highest was 1.38. Therefore, competitive indices of 0.75 and 1.38 were used as thresholds. Mutants with average competitive indices >0.75 but <1.38 were characterized as phenotypically equivalent to wild type. Mutants with average competitive indices greater than 1.38 were characterized as being advantageous in vivo. Mutants with average competitive indices at 14 days postinfection of <0.75 but >0.2 were characterized as being slightly deficient while those with competitive indices of <0.2 were characterized as being severely deficient.

Pool complexity:
Given that the DDRMs were phenotypically neutral in vivo and that pools of at least three strains could be tested, the next step was to determine whether larger pools could be tested. As the size or complexity of an inoculum pool increases, so does the chance that, due to stochastic effects, insufficient numbers of any one strain can be recovered from infected tissue, resulting in the misidentification of gene products required for survival. To determine how many S. cerevisiae mutants could be combined in an inoculum pool, a competition experiment was performed where three strains were combined in different ratios to mimic different pool sizes; the bulk of the initial inoculum population consisted of an unmarked wild-type strain (YAG129) and the remainder of the population consisted of YAG40 (HO/ho::hphMX4) in an 1:6 ratio or 17.1% of the population and YAG38 (HO/ho::kanMX4) in an 1:13 ratio or 7.5% of the population. Between 4 hr and 14 days postinfection, there was little change in the percentages of YAG129 (75.4 ± 5.1% to 75.6 ± 2.9%) and YAG40 (17.1 ± 3.6% to 20.0 ± 2.8%) within the population. Therefore, a pool of at least six strains can be tested for survival in CD-1 mice over a 14-day time course. The decrease of YAG38 from 7.5 ± 1.5% at 4 hr postinfection to 4.5 ± 0.7% of the population at 14 days postinfection suggested that 13 strains may be approaching the pool size limit, at least for this mouse model system at a 14-day timepoint. However, even in a pool of 13 mutants, minor fluctuations such as these would not prohibit the identification of mutants with severe in vivo survival defects.

Mating does not occur in vivo:
The ability to pool strains within a single inoculum required that mating must not occur in vivo, because mating would result in the complementation of mutations in different strains. All the strains used in these experimental infections were MATa/MAT diploids and therefore unable to mate. However, these strains might, at least in principle, sporulate in vivo to produce mating-competent haploid spores. Since the pool included HO/ho::kanMX4 and HO/ho::natMX4 strains, ho::kanMX4/ho::natMX4 diploids would be generated if mating occurred in vivo. To determine if any mating occurred in vivo, the homogenates of the neutrality experiments, which contained 300–1000 colony-forming units when plated onto YEPD + G418 or YEPD + nourseothricin, were also plated onto YEPD + G418 + nourseothricin. No colonies grew on the YEPD + G418 + nourseothricin plates, which strongly argues that mating did not occur in vivo.

The role of nucleotide biosynthesis in survival in vivo:
Both the purine and pyrimidine biosynthetic pathways are required for pathogenesis in the phylogenetically diverse species of C. albicans (KIRSCH and WHITNEY 1991 ) and C. neoformans (VARMA et al. 1992 ; PERFECT et al. 1993 ). Given the evolutionary conservation of many pathways between different fungi, one hypothesis is that gene products necessary for survival of one pathogenic fungus in most cases should also be necessary for survival of other species of fungi. Therefore, as a test of this hypothesis, ade2, ade4, and ura3 S. cerevisiae mutants were constructed and evaluated for survival in the CD-1 mouse model system.

The ura3 strains YAG216c and YAG218a were compared to the prototrophic strain YAG40 for survival. As shown in Table 3, there was a significant loss of the ura3 strains as rapidly as 4 hr postinoculation. After 14 days, the ura3 mutants were severely depleted in the population. To prove that the reduced survival of the ura3 mutants was due to loss of the URA3 gene, the URA3 gene was transformed into ura3 strain YAG216c to generate the Ura⁺ strain YAG228 (URA3/ura3::natMX4). As shown in Table 4, the reintroduction of the URA3 gene restored survival, confirming the requirement of Ura3p for survival.

Since de novo biosynthesis of pyrimidines was important for survival, the pyrimidine salvage pathway might be important for fungal survival as well. To test this hypothesis, fcy1 strains deficient in cytosine deaminase, a part of the pyrimidine salvage pathway that converts cytosine to uracil, were constructed and tested for survival. As shown in Table 3, the fcy1 mutation was not deleterious and appeared to confer a slight selective advantage in vivo.

Table 3 shows a similar assay for the requirement of purine biosynthesis in vivo using the ade2 strains YAG210a and YAG208a vs. YAG40 and the ade4 strains YAG269a and YAG271a vs. YAG40. In contrast to the ura3 mutants, the ade2 and ade4 mutants were not dramatically reduced from the in vivo population 4 hr postinoculation. However, after 7 days the ade2 and ade4 mutants were both severely depleted. To establish that the survival defect of the ade2 mutants was due to the ade2 mutation, the ADE2 gene was transformed into the ade2 strain YAG208a to generate YAG226 (ADE2/ade2::kanMX4). As shown in Table 4, the wild-type ADE2 gene restored survival to wild-type levels.

To compare in vivo with in vitro competition results, the ade2 strains YAG210a and YAG208a and the ade4 strains YAG269a and YAG271a were competed in vitro against the wild-type strain YAG40. After 23 population doublings in YEPD at 37°, the competitive indices for the ade2 strains were 0.004 for YAG210a and 0.001 for YAG208a; after 34 population doublings, both ade2 strains were undetectable with competitive indices of <0.0008. In contrast, after 23, 34, and 45 population doublings, the in vitro competitive indices for the ade4 strains were 0.54, 0.53, and 0.25 for YAG269a and 0.35, 0.28, and 0.15 for YAG271a, respectively.

The role of amino acid biosynthesis in survival in vivo:
Virtually all of the amino acids are present at substantial concentrations in serum (see CRISPENS 1975 for amino acid concentrations in mouse serum). While tissue-specific levels of amino acids may differ from serum levels, these amino acids are presumably available to mammalian cells and may also be available to pathogens. One hypothesis is that these amino acids will allow auxotrophic fungal mutants to survive. The alternative hypothesis is that, due to inadequate transport capacity or low tissue-specific levels of one or more amino acids, amino acid biosynthetic pathways will be critical for fungal survival.

The leu2 strains YAG185b and YAG183b were compared to prototrophic YAG40 for their ability to survive in mice. As shown in Table 3, the leu2 strains were severely depleted by 14 days postinfection. In contrast, as shown in Table 3, his3 mutants were only slightly deficient in vivo and trp1 mutants were approximately equivalent to the prototrophic control. One possible explanation for this result is that the prototrophic strain in each pool was able to cross-feed the auxotrophs. To determine if the prototrophic YAG40 strain in the mixed infection somehow supplied exogenous histidine or tryptophan to the auxotrophic mutants, three groups of 10 mice were infected, separately, with 2 x 10⁷ cells of each auxotrophic strain and a prototrophic control. Five mice from each group were sacrificed 4 hr postinfection and the remaining 5 were sacrificed 14 days later. Brain homogenate (100 µl) from each animal was plated in triplicate onto antibiotic plates and the colonies counted after incubation at 30°. When infected in the absence of wild-type cells, the survival of his3 and trp1 mutants was almost equivalent to wild type (data not shown). Therefore, cross-feeding between prototrophic and mutant strains in vivo was not responsible for the survival of the his3 and trp1 strains. In contrast to leu2 strains, the phenotype of his3 and trp1 strains suggested that the concentrations of histidine and tryptophan, coupled with the histidine and tryptophan transport capacity, were sufficient for survival in vivo.

The role of the glyoxylate cycle in survival in vivo:
Carbon source utilization to generate energy will be important for survival of pathogens in vivo. One aspect of carbon source utilization and energy metabolism is the glyoxylate cycle. The glyoxylate cycle is not required for utilizing sugars or C3 compounds, such as glycerol, as carbon sources. However, the glyoxylate cycle is required for utilizing C2 compounds, such as ethanol and acetate (FERNANDEZ et al. 1992 ; SCHOLER and SCHULLER 1993 ), and fatty acids, such as oleate (MCCAMMON 1996 ), as carbon sources. Isocitrate lyase (ICL1) is one of the key enzymes in the glyoxylate cycle (FERNANDEZ et al. 1992 ; SCHOLER and SCHULLER 1993 ; MCCAMMON 1996 ). To determine the role of the glyoxylate cycle in survival, the icl1 strains YAG323a and YAG331c were competed against YAG40. As shown in Table 3, the icl1 mutants were slightly deficient in survival compared to wild type at 14 days postinfection; a 21-day postinfection timepoint showed a similar phenotype (data not shown).

In CD-1 mice, S. cerevisiae proliferated and showed extended survival only in the brain (CLEMONS et al. 1994 ). In contrast, in DBA/2N mice, S. cerevisiae strain YJM128, the parent of YJM145 (MCCUSKER et al. 1994A ), was able to survive at least 30 days in spleen, liver, and kidney, and 14 days in lung (BYRON et al. 1995 ). To determine if icl1 mutants had an attenuated ability to survive in other mouse organs, and to assess survival at longer timepoints, survival of the icl1 mutants YAG323a and YAG331c was compared to YAG40 in complement factor 5-deficient DBA/2N mice. Relative to wild type, the icl1 mutants showed equivalent survival in liver, lung, and kidney at 14 days, in spleen at 21 days, and in brain at 28 days (data not shown).

The role of respiration and the mitochondrial genome in survival in vivo:
Respiration is another key aspect of energy metabolism. Unlike most of the pathogenic fungi, S. cerevisiae is petite positive; that is, S. cerevisiae can lose its mitochondrial genome and remain viable (reviewed in WHITTAKER 1979 ). Therefore, in S. cerevisiae it is possible to test two hypotheses: (1) respiration is critical for survival and (2) the mitochondrial genome makes respiration-independent contributions to survival in vivo.

COX15 is a nuclearly encoded gene, the gene product of which is required for cytochrome c oxidase assembly (GLERUM et al. 1997 ); cox15 mutants retain their mitochondrial genome but are unable to respire. The cox15 strains YAG152a and YAG154a were compared to YAG38 for their ability to survive. As shown in Table 3, the cox15 mutants were significantly depleted in vivo. To determine that the reduced survival of the cox15 strains was due to loss of the COX15 gene, the COX15 gene was transformed into YAG152a to generate YAG230 (COX15/cox15::natMX4). As shown in Table 4, the wildtype COX15 gene restored the ability to survive to the cox15 strain.

MIP1 is the nuclearly encoded structural gene for the catalytic subunit of the mitochondrial-specific DNA polymerase. In the absence of Mip1p, yeast cells are unable to respire because they have lost their mitochondrial genome and become ⁰ (GENGA et al. 1986 ; FOURY 1989 ). The mip1 ⁰ strains YAG180b and YAG182a were compared to YAG40 for the ability to survive in mice. As shown in Table 3, the mip1 ⁰ strains were rapidly and severely depleted in mice. Unlike all of the other mutations in this study, mip1 causes loss of the mitochondrial genome, a defect that cannot be complemented by reintroduction of MIP1. Therefore, YAG158, an isogenic MIP1 strain that was made ⁰ with ethidium bromide (SHERMAN et al. 1974 ), was tested for its ability to survive in mice; like the mip1 ⁰ strains, this MIP1 ⁰ strain was rapidly and severely depleted in mice (Table 3). Because MIP1 ⁰ and mip1 ⁰ strains had the same phenotype in mice, the effect on survival in vivo was independent of Mip1p and was due to the absence of the mitochondrial genome.

The cox15 and mip1 results were consistent with the hypothesis that respiration was required for survival. However, although both the cox15 and mip1 mutants were depleted in vivo, the two mutants behaved differently. While the cox15 mutants at 7 days postinfection had a competitive disadvantage relative to wild type of 12-fold, the mip1 mutants had a competitive disadvantage of 130-fold. These results suggest that the mitochondrial genome contributes more to fitness and to survival in vivo than the ability to respire.

The role of dimorphism in survival in vivo:
Although there are exceptions, such as C. glabrata, which does not undergo dimorphism in vivo (FIDEL et al. 1999 ), dimorphism is considered to play an important role in the pathogenesis of most of the pathogenic fungi infecting humans. In S. cerevisiae, the transcription factors Flo8p and Tec1p, which are parts of the cAMP and mitogen-activated protein kinase (MAPK) signaling pathways, respectively, are both required for expression of FLO11 (PAN and HEITMAN 1999 ; RUPP et al. 1999 ), a flocullin required for S. cerevisiae dimorphism (pseudohyphal formation; LAMBRECHTS et al. 1996 ; LO and DRANGINIS 1998 ). The YJM145 genetic background has previously been shown to be able to produce abundant pseudohyphae in vitro (MCCUSKER et al. 1994B ). To assess the effect of mutations on pseudohyphal formation in vitro, flo11, flo8, tec1, and flo8 tec1 mutations were introduced into the YJM145 background; these mutants were streaked onto SLAD plates and scored for pseudohyphal formation after 1 week. While the tec1 mutants were able to form pseudohyphae almost as abundantly as wild type, the flo11, flo8, and flo8 tec1 mutants formed few, if any, pseudohyphae (data not shown).To determine the contribution of dimorphism to the survival of S. cerevisiae in vivo, flo11, flo8, tec1, and flo8 tec1 mutants were competed against wild-type strains in vivo. As shown in Table 3, the flo11 mutant strains were phenotypically equivalent to wild type while the flo8 and tec1 mutant strains were slightly deficient in survival. The flo8 and tec1 mutant strains were somewhat more fit in vivo than the flo8 tec1 mutant strain, suggesting a possible additive phenotypic effect of loss of both Flo8p and Tec1p.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION LITERATURE CITED

Many conserved signal transduction and metabolic pathways are critical for fungal pathogenicity (KIRSCH and WHITNEY 1991 ; MITCHELL 1998 ; LENGELER et al. 2000 ). These conserved signal transduction and metabolic pathways are best understood in S. cerevisiae in large part because this microorganism, in contrast to the pathogenic Candida species, is well characterized genetically and can be readily manipulated. S. cerevisiae is closely related to the pathogenic Candida species (BOWMAN et al. 1992 ; LOTT et al. 1993 ) and is an emerging opportunistic pathogen (HAZEN 1995 ; MURPHY and KAVANAGH 1999 ). In addition, mouse model systems of systemic infection have been developed for S. cerevisiae (CLEMONS et al. 1994 ; BYRON et al. 1995 ). For these reasons, S. cerevisiae can serve as a genetically facile model for the identification of fungal survival factors.

Auxotrophy and survival in vivo:
As a first test of the S. cerevisiae model system, auxotrophs were examined for their ability to survive in the mouse model system. In addition to testing the generality of fungal survival factors, auxotrophic mutants allow several key aspects of fungal biology to be assessed in vivo. For example, the phenotype of an auxotroph is a function of (i) the concentration of the required nutrient combined with the uptake capacity for that nutrient in vivo, (ii) the phenotypic effect of limitation for the required nutrient in the in vivo environment, and, possibly, (iii) the specific block in the pathway.

The survival defect of C. albicans and C. neoformans purine and pyrimidine auxotrophs (KIRSCH and WHITNEY 1991 ; VARMA et al. 1992 ; PERFECT et al. 1993 ) also occurs in S. cerevisiae purine and pyrimidine auxotrophs, which supports generality of factors required for fungal survival. The in vivo phenotypes of the S. cerevisiae ade2 and ade4 mutants suggested that, at least for purine auxotrophy, the specific block in the pathway was not relevant to survival and that purine limitation, whether due to a low concentration of purines or low purine transport capacity in vivo, must severely affect survival in a mammalian host.

The ade2 and ade4 mutants allowed an interesting test of the hypothesis that in vitro phenotypes can predict in vivo phenotypes. S. cerevisiae ade4 mutants are deficient in the first step in purine biosynthesis. ade2 mutants, which are blocked in a late step in the purine biosynthetic pathway (JONES and FINK 1982 ), accumulate red pigment and have an in vitro growth defect (SMITH et al. 1995 ; UGOLINI and BRUSCHI 1996 ). A mutational block earlier in the purine biosynthetic pathway abolished both the in vitro growth defect and red pigment formation of ade2 mutants; consequently, red pigment formation has been suggested to cause the growth defect (UGOLINI and BRUSCHI 1996 ). Since the red pigment appears to contain glutathione (CHAUDHURI et al. 1997 ), the in vitro growth defect might be due to glutathione depletion. Since glutathione is involved in protection from reactive oxygen species, a key aspect of host defenses, one might expect ade2 mutants to be more severely deficient in vivo than ade4 mutants.

In vitro competitions (this work), performed at 37° to mimic body temperature but under otherwise highly favorable environmental conditions (YEPD has a high concentration of glucose and acidic pH and is nutrient replete), showed that while ade4 mutants were slightly deficient in vitro, ade2 mutants were severely deficient in vitro. Given this in vitro differential between ade2 and ade4 mutants, one might have predicted that ade4 mutants would have a considerably less severe defect in vivo than ade2 mutants. In fact, at least at the timepoints tested, ade2 and ade4 mutants were phenotypically indistinguishable in vivo. These results demonstrate that phenotypes are environment specific and that extrapolation from the in vitro environment to the in vivo environment can be misleading and problematic.

Three different amino acid auxotrophs were examined for their ability to survive in vivo: his3, leu2, and trp1. Histidine, leucine, and tryptophan are all present at concentrations in mouse serum (CRISPENS 1975 ) similar to those used in many recipes for synthetic medium (SHERMAN et al. 1974 ). The S. cerevisiae his3 and trp1 mutants were only slightly deficient in their ability to survive. Similarly, a C. albicans his1 mutant has been shown to be unaffected in its virulence (ALONSO-MONGE et al. 1999 ), which, like the results with the ade2 and ura3 mutants, speaks to the generality of fungal survival factors. In contrast, in spite of a substantial concentration of leucine in mouse serum (22 µg/ml; CRISPENS 1975 ), leu2 mutants were severely depleted in vivo. These results suggested that (i) specific amino acids may exist at lower concentrations in tissue(s) compared with serum and/or (ii) significant differences may exist in the transport capacity for different amino acids in vivo. These in vivo competition results were exactly the opposite of what would be predicted on the basis of in vitro competition experiments, where leu2 mutants have no discernible growth defect (PALERMO et al. 1997 ) and his3 and trp1 mutants both have substantial growth defects (SMITH et al. 1995 ; BAGANZ et al. 1998 ). Clearly, in vitro phenotypes are not accurate predictors of in vivo phenotypes.

The survival defect of S. cerevisiae leu2 mutants may be different from that of a C. albicans leu2 mutant, which was approximately as virulent as the prototrophic parent (KIRSCH and WHITNEY 1991 ). However, this difference between leu2 mutants in the two species may be due to the different assays used—the ability of S. cerevisiae leu2 mutants to survive in a nonimmunodeficient host vs. the ability of a C. albicans leu2 mutant to kill an immunosuppressed host. Alternatively, C. albicans, which secretes proteases, may be able to satisfy a leucine auxotrophic requirement through peptides, or C. albicans may have better leucine transport capacity in vivo than S. cerevisiae. Future experiments will be required to distinguish between these different possibilities.

The characterization of auxotrophs for their ability to survive in the in vivo environment addresses an important issue, namely, the ecology of infection. In addition, the in vivo phenotypes of mutants can identify and evaluate potential antifungal drug targets; from this work and other analysis, the histidine and tryptophan pathways would not be good antifungal targets while the purine, pyrimidine, and, possibly, leucine pathways would be good antifungal drug targets. Finally, the identification of auxotrophs with survival defects allows these genes to be used as reporters for in vivo promoter activity for future in vivo expression technology analysis (MAHAN et al. 1993 ; SLAUCH et al. 1994 ).

Drug resistance and fitness in vivo:
Antimicrobial drug resistance is a serious clinical problem for both bacteria (ANDERSSON and LEVIN 1999 ) and fungi (VANDEN BOSSCHE et al. 1994 ; WHITE et al. 1998 ; VERMES et al. 2000 ). Antimicrobial drug resistance is also an interesting area of research for evolutionary and population biology because the clear benefit to the pathogen of becoming resistant to a drug or antibiotic to which it is exposed in vivo is frequently counterbalanced by a deleterious effect on fitness of the mutation or gene that confers the resistance phenotype (reviewed in ANDERSSON and LEVIN 1999 ). Since the effects of resistance mutations on fitness depends upon the environment (BJORKMAN et al. 1998 , BJORKMAN et al. 1999 , BJORKMAN et al. 2000 ; ANDERSSON and LEVIN 1999 ), it is critical to examine the phenotype of drug-resistance mutations in vivo.

Determining the effects on fitness of drug-resistance mutations requires isogenic strains and knowing the exact nature of the resistance mutations. These requirements are difficult to achieve in most pathogenic fungi but are relatively straightforward in S. cerevisiae. Therefore, as a test of the system, fcy1 mutants were tested for their ability to survive. Cytosine deaminase, which is encoded by FCY1, is part of the pyrimidine salvage pathway that converts cytosine to uracil. Cytosine deaminase is also required for the activity of the antifungal drug 5-fluorocytosine which, in order to be toxic, must be converted to 5-fluorouracil. Given the importance of de novo biosynthesis of pyrimidines for survival, the pyrimidine salvage pathway might also be important for fungal survival, which would be consistent with the frequent deleterious effects of drug-resistance mutations. However, the fcy1 mutants were not deleterious and, indeed, appeared to be slightly advantageous in vivo; this result is consistent with the frequent clinical occurrence of 5-fluorocytosine resistance in pathogenic yeasts (VANDEN BOSSCHE et al. 1994 ; WHITE et al. 1998 ; VERMES et al. 2000 ). In the future, the phenotype of S. cerevisiae mutants resistant to other clinically used antifungal drugs, such as azoles, can be examined to determine their effects on fitness in vivo. The same resistance mutations can then be introduced into other pathogenic fungi, such as C. albicans, which will allow comparison between different species.

The glyoxylate cycle and survival in vivo:
The role of the glyoxylate cycle in vivo was of particular interest because of its requirement for utilization of fatty acids, such as oleate (MCCAMMON 1996 ), as carbon sources. It has been shown that Mycobacterium tuberculosis isocitrate lyase mutants were defective in their ability to persist in mice (MCKINNEY et al. 2000 ); the persistence defect of M. tuberculosis isocitrate lyase mutants is thought to be due to an inability to utilize fatty acids as carbon sources when the bacteria are in the lipid-rich environment of macrophages. Interestingly, the ICL1 gene of C. neoformans is induced in vivo (J. PERFECT, personal communication). In addition, the ICL1 gene of both S. cerevisiae and C. albicans has recently been shown to be substantially induced upon exposure to macrophages in vitro and a C. albicans icl1/icl1 mutant showed a substantial reduction in virulence (LORENZ and FINK 2001 ). The fact that icl1 mutants were slightly deficient in vivo suggests that the glyoxylate cycle and, presumably, fatty acid utilization make a minor contribution to S. cerevisiae fitness in vivo.

Respiration, the mitochondrial genome, and survival in vivo:
There are three possible explanations for the critical importance of respiration for survival in the host. First, and most obvious, there is the role of respiration in energy metabolism. Given a concentration of glucose in mouse serum of 0.07% (CRISPENS 1975 ), efficient energy metabolism is likely to be critical for survival. Second, an intact respiratory chain is required for resistance to reactive oxygen species (GRANT et al. 1997 ; PUNGARTNIK et al. 1999 ) that are part of the host defense system. Finally, there is considerable evidence for communication from the mitochondrion to the nucleus (PARIKH et al. 1987 ; LIAO and BUTOW 1993 ; POYTON and MCEWEN 1996 ; HALLSTROM and MOYE-ROWLEY 2000 ). Using S. cerevisiae, it will be possible to assess the contribution of these different processes to survival.

The severe survival defect of ⁰ strains, relative to cox15 strains, suggested that the mitochondrial genome makes respiration-independent contributions to survival. It is clear that, in addition to respiration, a variety of functions take place within the mitochondrion, such as steps in amino acid, fatty acid, and heme biosynthesis (FLAVELL 1971 ; TZAGOLOFF and DIECKMANN 1990 ; PON and SCHATZ 1991 ); it seems likely that the absence of the mitochondrial genome perturbs one or more of these functions. The mitochondrial genome is also necessary for proper mitochondrial morphology; while inhibition of respiration and cytochrome c oxidase deficiency have no effect on mitochondrial morphology (CHURCH and POYTON 1998 ), mitochondrial morphology is defective in ⁰ strains (STEVENS 1981 ). Therefore, mitochondrially encoded gene products may be necessary for proper mitochondrial morphology, which may in turn be required for optimal function of mitochondrially localized, nuclearly encoded gene products.

The severe survival defect of ⁰ strains was particularly interesting because many S. cerevisiae strains grown at high temperatures (37°) show an elevated loss of the mitochondrial genome and become ⁰ (SHERMAN 1959 ; SIMOES-MENDES et al. 1978 ). Given the severe survival defect of S. cerevisiae ⁰ strains, mitochondrial genome stability at the high temperatures experienced in vivo must be an important survival factor that may, in turn, be intrinsically related to the ability of clinically derived S. cerevisiae strains to grow at supra-optimal temperatures, as determined in vitro (MCCUSKER et al. 1994A , MCCUSKER et al. 1994B ). Mitochondrial genome stability at temperatures experienced in vivo is likely to be even more critical for other pathogenic fungi that are petite negative. The identification of S. cerevisiae genes important for high-temperature growth and mitochondrial genome maintenance should aid the identification of the corresponding genes and survival factors in other pathogenic fungi.

In the context of mitochondrial genome stability, 5-fluorocytosine damages the S. cerevisiae mitochondrial genome and is a potent petite-inducing (⁰/^-) agent (OLIVER and WILLIAMSON 1976A , OLIVER and WILLIAMSON 1976B ). Considering the effect of 5-fluorocytosine on the S. cerevisiae mitochondrial genome and the severe effect that loss of the mitochondrial genome has on S. cerevisiae survival, the clinical effect of 5-fluorocytosine on other pathogenic yeasts may be related to mitochondrial genome damage.

Dimorphism and survival in vivo:
Dimorphism is thought to be an important virulence factor for most of the pathogenic fungi infecting humans; it is clear that C. albicans mutants with defects in dimorphism are either avirulent or have attenuated virulence (LEBERER et al. 1997 ; LO et al. 1997 ; STOLDT et al. 1997 ; SCHWEIZER et al. 2000 ). While there are differences between the two species, many aspects regarding the control and expression of dimorphism are conserved between C. albicans and S. cerevisiae (reviewed in ERNST 2000 ; LENGELER et al. 2000 ).

The reversible dimorphic switch between yeast and pseudohyphal forms has been extensively characterized in S. cerevisiae (reviewed in LENGELER et al. 2000 ; GANCEDO 2001 ). In S. cerevisiae, a MAPK-regulated pathway, mediated by the transcription factor Tec1p, and a cAMP-regulated pathway, mediated by the transcription factor Flo8p, are both required to activate transcription of FLO11 (PAN and HEITMAN 1999 ; RUPP et al. 1999 ), encoding the flocculin Flo11p (LAMBRECHTS et al. 1996 ; LO and DRANGINIS 1998 ). To determine their effect on dimorphism in vitro in the clinically derived YJM145 genetic background, flo11, flo8, tec1, and flo8 tec1 mutants were constructed. Compared to wild type, the tec1 mutants were only slightly deficient in their ability to form pseudohyphae. However, the flo11, flo8, and flo8 tec1 mutants formed few, if any, pseudohyphae in vitro.

The flo8 and tec1 mutant strains were slightly deficient relative to wild type in vivo. Since the flo8 tec1 mutant strain was less fit than either of the single mutant strains, this result would suggest that the absence of both Flo8p and Tec1p has an additive effect on phenotype. Interestingly, the flo11 mutant strains were phenotypically indistinguishable from wild type in in vivo competitions. Therefore, the slight phenotypes of the flo8, tec1, and flo8 tec1 mutants cannot be due to lack of transcription of FLO11 and must instead be due to effects on other genes. In addition, given the complete absence of Flo11p in the flo11 strains, the in vivo phenotype of flo11 strains would argue that either production of Flo11p confers no selective advantage in vivo or that FLO11 is not expressed in vivo.

Of 30 S. cerevisiae clinical isolates, nonclinical strains, and laboratory strains tested for survival in vivo, the S288c genetic background, which is deficient in pseudohyphal formation due to a naturally occurring flo8 mutation (LIU et al. 1996 ), is the least able to survive in mice (MCCUSKER et al. 1994A ). However, the in vivo phenotype of flo8 mutants in the YJM145 background (this work) suggests that the flo8 mutation in the S288c background makes only a minor, quantitative contribution to the inability of the S288c background to survive in vivo. We will have to look beyond FLO8, and possibly beyond dimorphism altogether, to find the genetic basis for the large in vivo phenotypic differences between the YJM145 and S288c genetic backgrounds.

Given the importance of dimorphism for C. albicans pathogenicity, why do the S. cerevisiae flo11, flo8, tec1, and flo8 tec1 dimorphism-deficient mutants have little or no phenotype in vivo? One answer to this question is that some virulence or survival factors may be species specific. As an example of a species-specific virulence trait, capsule formation, which is critical for the virulence of C. neoformans, is not observed in other pathogenic fungi infecting humans (CASADEVALL and PERFECT 1998 ). In this sense, dimorphism may be critically important in vivo in some species, such as C. albicans (LEBERER et al. 1997 ; LO et al. 1997 ; STOLDT et al. 1997 ; SCHWEIZER et al. 2000 ), while playing little or no role in vivo in other pathogenic fungi, such as S. cerevisiae (this work) and C. glabrata (FIDEL et al. 1999 ).

Of all the pathogenic fungi, S. cerevisiae is most closely related to C. glabrata (LOTT et al. 1993 ). C. glabrata is one commonly observed pathogenic fungus where dimorphism is apparently not a virulence factor; although recently shown to be capable of undergoing dimorphism in vitro (CSANK and HAYNES 2000 ), C. glabrata does not form pseudohyphae in vivo (FIDEL et al. 1999 ). Similarly, S. cerevisiae pseudohyphae have not been observed in vivo (CLEMONS et al. 1994 ). The ability to undergo dimorphism in the in vivo environment may be one of the characteristics that distinguishes the relatively avirulent C. glabrata and S. cerevisiae from the considerably more virulent C. albicans. In contrast to C. albicans, signaling pathways in S. cerevisiae and C. glabrata may be unable to activate the dimorphism pathway in vivo. In the future, it will be possible to test the hypothesis that the S. cerevisiae signaling pathway(s) fail to turn on FLO11 in vivo. It will also be possible to test the hypothesis that increasing FLO11 expression in vivo will improve S. cerevisiae survival in vivo and/or make S. cerevisiae infections mimic C. albicans infections in becoming more lethal.

	CONCLUSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION LITERATURE CITED

Given the evolutionary conservation of basic signal transduction and metabolic pathways in fungi and the results described here, in most cases, what is true for S. cerevisiae will also be true for the more commonly observed pathogenic fungi. However, even when examining conserved functions and pathways, there will be some cases where S. cerevisiae mutants and mutants in another species have different in vivo phenotypes. Such cases may be directly relevant to the differences in pathogenesis between S. cerevisiae, an emerging opportunistic pathogen, and closely related and more virulent species, such as C. albicans. Interspecific differences in mutant phenotypes may also be relevant to the evolution of fungal pathogenesis. It is now possible to use S. cerevisiae, the most extensively studied eukaryotic microorganism, to directly examine conserved pathways for their role in survival in vivo, which will improve our understanding of fungal pathogenesis.

	ACKNOWLEDGMENTS

The authors thank J. Harrell for some strain construction and P. Hartzog and B. Nicholson for some PCR primer design. The authors also thank G. Cox, J. Heitman, T. Mitchell, and J. Perfect for reading the manuscript and J. Perfect for communication of unpublished results concerning the in vivo expression of the C. neoformans ICL1 gene. A. Goldstein was supported by National Institutes of Health (NIH) training grants T32-CA09111 and T32-AI07392. This work was also supported by NIH grants PO1-AI44975 and RO1-GM58476 and funds from Duke University Medical Center.

Manuscript received May 21, 2001; Accepted for publication July 19, 2001.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION LITERATURE CITED

ALONSO-MONGE, R., F. N