Mismatch Repair of DNA Replication Errors Contributes to Microevolution in the Pathogenic Fungus Cryptococcus neoformans

ABSTRACT The ability to adapt to a changing environment provides a selective advantage to microorganisms. In the case of many pathogens, a large change in their environment occurs when they move from a natural setting to a setting within a human host and then during the course of disease development to various locations within that host. Two clinical isolates of the human fungal pathogen Cryptococcus neoformans were identified from a collection of environmental and clinical strains that exhibited a mutator phenotype, which is a phenotype which provides the ability to change rapidly due to the accumulation of DNA mutations at high frequency. Whole-genome analysis of these strains revealed mutations in MSH2 of the mismatch repair pathway, and complementation confirmed that these mutations are responsible for the mutator phenotype. Comparison of mutation frequencies in deletion strains of eight mismatch repair pathway genes in C. neoformans showed that the loss of three of them, MSH2, MLH1, and PMS1, results in an increase in mutation rates. Increased mutation rates enable rapid microevolution to occur in these strains, generating phenotypic variations in traits associated with the ability to grow in vivo, in addition to allowing rapid generation of resistance to antifungal agents. Mutation of PMS1 reduced virulence, whereas mutation of MSH2 or MLH1 had no effect on the level of virulence. These findings thus support the hypothesis that this pathogenic fungus can take advantage of a mutator phenotype in order to cause disease but that it can do so only in specific pathways that lead to a mutator trait without a significant tradeoff in fitness.

A n essential prerequisite for all groups of microbial pathogens is the ability to change rapidly during infection to adapt to the hostile environment of the host. This adaption enables pathogens to avoid recognition or destruction by the host immune system, can result in latency and long-term pathogen persistence, and expedites the emergence of resistance to antimicrobial agents. One mechanism thought to account for the ability to change is the process of microevolution, represented by a change in allelic frequencies that occurs within a population over a short period of time. Microevolution during the course of infection can select for a small proportion of the microbe population that has gained novel traits that facilitate the ability to cause disease. These traits are generated by mutations in the microbe's DNA sequence arising from changes at the chromosomal level, from errors in DNA replication, or from environmental damage. Mutations can be preexisting or can be rapidly acquired in response to the host environment in a process termed "adaptive evolution." Microevolution in fungal pathogens has previously been attributed to changes at the chromosomal level, with evidence of gain, loss, or recombination between chromosomes leading to resistance to antifungal agents (1)(2)(3)(4)(5). Upon exposure to the antifungal agent fluconazole, Candida albicans and Cryptococcus neoformans undergo a process termed heteroresistance in which rapid and yet reversible resistance is conferred by the development of one or more aneuploidies or large-scale chromosomal rearrangements (reviewed in reference 6). The most commonly occurring aneuploidy in C. albicans is an isochromosome of the left arm of chromosome 5, i(5L), which contains TAC1 and ERG11, genes encoding the transcriptional activator of drug efflux genes and the target of fluconazole (7,8). In C. neoformans, increasing concentrations of fluconazole result in duplication of chromosome 1, containing ERG11 and the azole transporter AFR1, in all resistant strains, followed by successive duplication of chromosomes 4, 10, and 14 (9,10). Cells return to normal ploidy when fluconazole is removed, presumably due to reduced fitness as evidenced by reduced proliferation and virulence in vivo (10). Interestingly, recent studies have also indicated that single-base-pair mutations may play an important role in the microevolution of pathogenic fungi during infection (11,12). Genomic sequencing of C. neoformans isolates from a relapsed patient before and after antifungal treatment identified a chromosomal rearrangement and also an additional base pair mutation in the AVC1 gene that controls several virulence phenotypes (11). In addition, high-frequency switching of C. neoformans between yeast cells to a pseudohyphal cellular morphology is governed by single-base-pair mutations in genes of the regulation of Ace2p activity and cellular morphogenesis (RAM) pathway (12). Importantly, mutation of MSH2, MLH1, and PMS1 (three C. neoformans genes from the mismatch repair [MMR] pathway predicted to be required to repair damage to single bases arising from errors in DNA replication) resulted in increased proliferation in a lung assay of cryptococcosis (13). This suggests that single-base mutations arising from errors in DNA replication can generate novel traits that facilitate the ability to cause disease and therefore provide an avenue for the pathogen to undergo microevolution during infection.
Errors generated during DNA replication are corrected by the action of two sequential repair systems, namely, the DNA polymerase 3=-5= exonuclease activity system and the mismatch repair (MMR) system. Possession of sequential repair mechanisms enables a multiplicative effect, increasing the fidelity of DNA replication by a factor of 10 4 to 10 5 . In bacteria, replication errors are corrected by the 3=-5= editing exonuclease activity of the subunit of DNA polymerase III and the mutHLS mismatch repair system (reviewed in reference 14). A MutS homodimer binds the mismatch and is detected by a MutL homodimer, which initiates removal by MutH nicking of the nascent strand. This system is structurally and functionally conserved in eukaryotes. In Saccharomyces cerevisiae, for instance, the Pol2 () and Pol3 (Ѩ) DNA polymerases, which primarily replicate the leading and lagging strands, respectively, also have 3=-5= exonuclease activity. Subsequent repair is then performed by an analogous MMR pathway. The S. cerevisiae genome encodes six MutS homologues (MSH1 to MSH6) and four MutL homologues (MLH1 to MLH3 and PMS1); however, only MSH2, MSH3, MSH6, MLH1, and PMS1 are required for mismatch repair of nuclear DNA, with MSH3 and MSH6 playing partially redundant roles (reviewed in reference 15). MSH1 is required for repair of mitochondrial DNA, whereas MSH4, MSH5, MLH2, and MLH3 are involved in the meiotic recombination processes (15). In eukaryotes, the MutS homodimer found in prokaryotes is replaced with heterodimers referred to as MutS␣ (Msh2 and Msh6) and MutS␤ (Msh2 and Msh3), which are redundant with respect to repairing small indels but are specialized with respect to removal of specific base-base mismatches and large insertion-deletion loops, respectively. MutS␣/␤ is tethered to the proliferating cell nuclear antigen (PCNA), a DNA clamp that acts as a processivity factor for Pol3 (DNA polymerase Ѩ), and, after mismatch recognition, a MutL heterodimer referred to as MutL␣ (comprising Mlh1 and Pms1 [Mlh1-Pms1]) interacts to form a ternary complex that leads to nicks on the new strand and subsequent strand removal and resynthesis. The S. cerevisiae MSH2, MLH1, and PMS1 MMR mutants exhibit a mutator phenotype in which the rate of spontaneous mutation is elevated (15).
This study investigated the contribution of single-base-pair mutations arising from errors in DNA replication to microevolution in a causative agent of cryptococcosis, C. neoformans. Cryptococcosis, caused by C. neoformans or C. gattii, is a significant fungal disease worldwide, with high incidence and morbidity. An estimated 624,000 deaths were once attributed to cryptococcosis each year, a rate higher that attributed to tuberculosis (16). The direction that this research took is presented as follows. Analysis of a collection of environmental and clinical C. neoformans isolates revealed that a number of isolates possessed a mutator phenotype and the ability to change rapidly due to the accumulation of a high frequency of DNA mutations. Whole-genome sequence analysis of two of these mutator strains revealed mutations in MSH2 of the mismatch repair pathway. Deletion of the MMR components encoded by MSH2, MLH1, and PMS1 in C. neoformans resulted in an elevated mutation rate due to an increase in mutations in homopolymeric tracts and transitions, specifically, G to A, suggesting that MMR in C. neoformans repairs errors arising from both replication and oxidative damage. We show that the elevated mutation rate and the types of mutations generated in the MMR mutants are not detrimental to growth and result in only minor sensitivity to stress while enabling rapid microevolution to occur. This microevolution introduces phenotypic variations in traits associated with the ability to grow in vivo, in addition to allowing the rapid generation of antifungal resistance. Interestingly, the elevated mutation rate of the MMR mutants was not detrimental to C. neoformans in vivo, with no decrease in colonization or virulence observed in msh2Δ and mlh1Δ mutants in a murine infection model. This suggests that C. neoformans isolates possessing mutations in MMR components have an opportunity for enhanced adaptive evolution during infection.

RESULTS
Cryptococcus neoformans clinical isolates exhibit a mutator phenotype. Pathogens can retain a larger-than-expected proportion of mutators, i.e., cells with a greaterthan-normal mutation frequency, in their populations (17)(18)(19). In bacteria, mutators found in nature carry defects in a component of the MMR (20). In addition, 55% of clinical isolates of the fungus Candida glabrata contain a mutation in MSH2 (21). To assess if this phenomenon also occurs in C. neoformans, a collection of 11 clinical and 11 environmental strains of C. neoformans (var. grubii, serotype A) from the study described in reference 22 was examined for their mutation rates. This collection of strains is currently the best comparator set, in that the virulence of the strains has been tested, showing a correlation between virulence and clinical origin (22). Mutation rates in these strains were qualitatively assessed by observing the generation of spontaneously resistant 5-fluoroorotic acid (5-FOA) colonies (Materials and Methods). 5-FOA inhibits the growth of wild-type strains as 5-FOA is converted by orotine-5´-monophosphate decarboxylase to 5-fluorouracil (5-FU), which is toxic. Using a high-throughput rapid assay for the emergence of resistance, no differences in mutations rates were seen in comparisons of the 11 environmental isolates to each other or to a laboratory wild-type strain (KN99␣). In contrast, 2 of the 11 clinical isolates, C23 and C45, had an increase in the number of spontaneously resistant 5-FOA colonies indicative of an increased mutation rate (Fig. 1A). C23 originated from bronchoalveolar lavage fluid of a lung transplant patient and C45 from the cerebrospinal fluid of an asthma patient (22). The increased mutation rate was quantitatively assessed using fluctuation analysis (Materials and Methods). This fluctuation analysis involves the inoculation of 20 parallel cultures (in nonselective medium) with a small number of cells (1 ϫ 10 5 cells) that are grown to saturation followed by plating onto selective media to obtain the number of mutants (Materials and Methods). The Lea-Coulson method of the median (23) was used with FALCOR software to estimate the number of mutations, from the observed values of mutants, across the 20 independent parallel cultures. The C23 and C45 clinical isolates showed mutation frequencies approximately 200-fold higher than those seen with the standard KN99␣ wild type (Fig. 1B).
To determine the molecular basis of the increased mutation rate, the genomes of the C23 and C45 isolates were analyzed by next-generation sequencing (Materials and Methods). Reads were aligned to the C. neoformans H99 genome sequence and single nucleotide polymorphisms (SNPs) and sequence variations identified (Materials and Methods). A total of 427 SNPs and sequence variations were identified in the C23 genome sequence. We hypothesized that an increased mutation rate would likely be a consequence of a mutation in a DNA repair gene, so those variants located within the coding region of the 57 genes predicted to play a role in DNA repair in C. neoformans were identified. MSH2, which is 1 of the 57 DNA repair genes in the C23 genome sequence, contained a sequence variation. MSH2 encodes a MutS protein required for mismatch repair of nuclear DNA. The genome of isolate C23 contained a single-base deletion at ϩ1689 which results in deletion of amino acids 379 to 397 and 3 missense mutations (V378N, S398L, and K399E). This region of the 965-amino-acid-long predicted protein (position 378 to position 399) contains 7 amino acids (R380, Q381, L388, P392, D393, R396, and K399) which are highly conserved from S. cerevisiae to humans. Furthermore, the mutational profile of the C23 genome resembles that of a S. cerevisiae MSH2 mutant (24,25). The majority of variations in C23 are SNPs arising from transition (42.9%) and transversion (22.9%) mutations and single-nucleotide deletions or insertions occurring within homopolymeric tracts (22.5%).
In contrast to that of isolate C23, the genome of isolate C45 showed an extremely high level of genetic heterogeneity compared to the genome of H99, with a predicted 269,177 SNPs and genome variations. A total of 32 SNPs were present within MSH2 in isolate C45, with 21 within the coding sequence. Most of these SNPs resulted in silent mutations; however, 6 (V167A, T182I, D238V, L436P, S885F, and N863D) resulted in missense mutations. None of the missense mutations were in highly conserved amino acids; however, the T182I substitution is located in an equivalent position that, when mutated in S. cerevisiae (L183P), resulted in a 220-fold induction of mutation rate due to a lack of structural integrity (24).
To test if the increased mutation rate in the C23 and C45 clinical isolates was due to mutations in MSH2, a wild-type copy of MSH2 was introduced into C23 and C45 at an ectopic location (Materials and Methods). Qualitative assessment of the mutation rate using the number of spontaneously arising 5-FOA colonies was used to assess complementation (Materials and Methods). The introduction of a wild-type copy of MSH2 complemented the increased mutation rate phenotype of both the C23 and C45 isolates, suggesting that the mutations in MSH2 identified by next-generation DNA sequencing were responsible for the enhanced mutation rate in these clinical isolates (Fig. 1B).
Deletion of MMR components results in an elevated mutation rate. In order to further investigate the contribution of single-base-pair mutations arising from errors in mismatch repair to microevolution, three strains with deletions in components of the mismatch repair pathway (the msh2Δ, mlh1Δ, and pms1Δ mutants) were obtained from the Cryptococcus neoformans H99 deletion collection (13). The levels of proliferation of the three mutants within lungs of mice had previously been shown to be increased in competition assays (13). The strains were backcrossed twice to the KN99a MATa congenic parent to remove unwanted melanin and mating differences known to exist in their background (13). Complemented strains were generated by reintroduction of the wild-type gene at an ectopic location using Agrobacterium tumefaciens-mediated transformation and resistance to G-418. Double mutants (msh2Δ mlh1Δ, msh2Δ pms1Δ, and mlh1Δ pms1Δ strains) were generated by crossing.
Mutation rates were qualitatively assessed by observing the generation of spontaneously resistant 5-fluoroorotic acid (5-FOA) and 5-fluorouracil (5-FU) colonies in the mutant strains. 5-FU inhibits the growth of wild-type strains by inhibiting the activity of thymidylate synthetase, which affects pyrimidine synthesis and leads to an imbalance of intracellular dNTP pools. Compared to the wild-type and complemented strains, the msh2Δ, mlh1Δ, and pms1Δ mutants displayed an increase in levels of spontaneously resistant 5-FOA and 5-FU colonies suggestive of a higher mutation rate ( Fig. 2A). The increased mutation rate was quantitatively assessed using fluctuation analysis, and the Lea-Coulson method of the median with FALCOR software was used to estimate the number of mutations from the observed values of mutants across the 20 independent parallel cultures on 5-FOA (

MMR mutants have an increased proportion of transitions and mutations in homopolymeric tracts.
Mutations of the URA5 and FUR1 genes in C. neoformans are the most common cause of resistance to 5-FOA and 5-FU, respectively (12,26). In order to determine the types of mutations generated in the absence of mismatch repair, URA5 and FUR1 were sequenced from 20 5-FOA-resistant and 20 5-FU-resistant strains isolated from the wild-type and msh2Δ, mlh1Δ, and pms1Δ mutant strains derived from the independent parallel cultures used in the fluctuation analysis. URA5 from wild-type 5-FOA-resistant strains contained a variety of mutations, including small insertions or deletions, transversions, and transitions (Fig. 3A). The msh2Δ, mlh1Δ, and pms1Δ mutants displayed a shift in the mutation profile of URA5 ( Fig. 3A and B). In comparison to the wild-type results, URA5 from msh2Δ, mlh1Δ, and pms1Δ 5-FOA-resistant strains showed a reduction in the number of insertion, deletion, and transversion mutations and an increase in the number of transition mutations (from 40% in the wild type to 75%, 100%, and 85% in the mutants, respectively) (Fig. 3B). In addition, the types of transition mutations generated in the msh2Δ, mlh1Δ, and pms1Δ strains differed from the wild-type results. In the wild-type strain, 100% of the transition mutations generated were T-to-C transitions. In contrast, the msh2Δ and mlh1Δ mutants generated predominately G-to-A transitions (Fig. 3C). This altered the positions of mutations to both coding and noncoding regions, as G-to-A transitions frequently occurred at intron splice sites and sites of lariat formation (Fig. 3A). In contrast to the wild-type strain and the msh2Δ and mlh1Δ mutants, the pms1Δ mutant produced all types of transitions (Fig. 3C) 7 at ϩ81 to ϩ87, (A) 6 at ϩ460 to ϩ465, and (T) 14 at ϩ941 to ϩ954 (Fig. 3D). The FUR1 gene sequenced from 5-FU-resistant strains derived from the wild-type strain contained a variety of mutations, including large and small deletions, transitions, and deletions in homopolymeric tracts (Fig. 3D). Similarly to the observation for URA5, the msh2Δ, mlh1Δ, and pms1Δ mutants displayed a shift in the mutation profile of FUR1 compared to that of FUR1 of the wild type (Fig. 3E). In comparison to the wild-type results, FUR1 from msh2Δ, mlh1Δ, and pms1Δ 5-FU-resistant strains possessed a reduction in the numbers of large deletions and an increase in the numbers of deletions and insertions in homopolymeric tracts (from 55% in the wild-type strain to 90%, 95%, and 95% in the mutants, respectively) (Fig. 3E). The numbers of transition mutations in FUR1 between the strains remained equivalent, suggesting that mutations in homopolymeric tracts occur more frequently than transitions. In addition, the type of mutations generated in homopolymeric tracts in the msh2Δ, mlh1Δ, and pms1Δ strains differed from that seen with the wild type. In the wild type, 100% of the mutations generated in homopolymeric tracts were deletions in the (A) 6 homopolymer. In contrast, the msh2Δ, mlh1Δ, and pms1Δ mutants possessed a variety of types of mutations in homopolymeric tracts in addition to deletions in the (A) 6 homopolymer, including deletions in the (C) 7 and (T) 14 homopolymers and an insertion in the (C) 7 homopolymer (Fig. 3F). Also, unlike the wild type, in which 100% of the mutations in homopolymeric tracts involved only one tract, FUR1 from the 5-FU-resistant msh2Δ, mlh1Δ, and pms1Δ strains possessed mutations in multiple homopolymeric tracts (22%, 79%, and 5%, respectively) (Fig. 3F).
Deletion of MMR components results in only minor sensitivity to DNA-damaging agents and oxidative stress. To explore further the role of MMR components in DNA repair, the msh2Δ, mlh1Δ, and pms1Δ mutants were assessed for phenotypes associated with sensitivity to or tolerance of DNA-damaging agents. Strains were cultured on a selection of DNA-damaging chemicals, including those associated with oxidative stress, or were treated with UV light (UV) (Fig. 4). The msh2Δ, msh2Δ pms1Δ, and msh2Δ mlh1Δ mutants showed a minor sensitivity to the following DNA-damaging agents (which affected DNA replication): methyl methanesulfonate (MMS), ethidium bromide (EtBr), and UV (Fig. 4). MMS methylates DNA to stall replication forks, EtBr affects replication by deforming the DNA by intercalation, and UV causes pyrimidine dimers to prevent DNA replication. The msh2Δ mutant also showed minor sensitivity to the following chemicals (which generate reactive oxygen species [ROS]): menadione, paraquat dichloride hydrate (paraquat), tert-butyl hydroperoxide (tBOOH), and hydrogen peroxide (H 2 O 2 ) (although this was not observed in any of the double mutants) (Fig. 4). In contrast, the pms1Δ and mlh1Δ mutants showed increased tolerance of paraquat (Fig. 4). ROS predominately introduces base or sugar damage leading to singlestranded break formation. Minor or no phenotypes are expected on ROS as oxidative damage is mainly repaired by the base excision repair (BER) pathway, although MMR has been shown to play a role in repairing oxidation of guanine to 8-oxo-7,8dihydroguanine (8-oxoG) (27).
Disruption of the mismatch repair pathway leads to rapid microevolution. In order to investigate whether the increased mutation rate observed in the MMR mutants results in an increased rate of microevolution and phenotypic change, the wild-type strain and the msh2Δ, mlh1Δ, and pms1Δ mutants were passaged 3 times independently in culture for approximately 600 generations with population bottlenecks every 90 generations (Materials and Methods). Passaging through population bottlenecks allows nonlethal mutations to accumulate as if they were neutral by mitigating the effect of selection. The resultant strains were compared to the original nonpassaged mutants and assessed for phenotypes that are associated with the ability to grow in vivo, including growth at 37°C, resistance to oxidative stress (paraquat and H 2 O 2 ), alterations in the cell wall (Congo red and calcofluor white stress), and melanization. No phenotypic differences were observed between the original wild-type strain and the 3 passaged wild-type strains at 37°C or on H 2 O 2 , paraquat, calcofluor white, or Congo red (Fig. 5A). In contrast, the phenotypes of the passaged msh2Δ and mlh1Δ mutants differed dramatically under these conditions from the original mutants and from each other, suggesting that the rate of microevolution was occurring more rapidly in these strains than in wild type (Fig. 5A). The msh2Δ strain showed more phenotypic variability than the mlh1Δ strain (Fig. 5A). The passaged pms1Δ strains were phenotypically the same as the original mutant and the wild type (Fig. 5A). Likewise, no differences between the original and passaged wild-type strains in melanization were observed (Fig. 5B). In contrast, the melanization results differed dramatically in the passaged msh2Δ strains compared to the original mutant, with each strain exhibiting a decrease in melanization, an almost complete absence of melanization, and an increase in melanization (Fig. 5B). The passaged mlh1Δ and pms1Δ strains showed differences in melanization compared to the original mutants also, though the differences were not as extreme as those seen with the msh2Δ strains (Fig. 5B).
Deletion of MMR components leads to rapid resistance to antifungal agents. To investigate the clinical relevance of disrupting mismatch repair in C. neoformans and the consequences of rapid microevolution, the fluconazole MIC was determined for the mutant strains using Etests (Materials and Methods). The msh2Δ and pms1Δ mutants had a MIC equivalent to that seen with the wild-type and complemented strains (Fig. 6A). The mlh1Δ mutant possessed much higher resistance to fluconazole than the wild-type, msh2Δ and pms1Δ mutant, and mlh1Δ MLH1 ϩ complemented strains (Fig. 6A). This increase in fluconazole resistance was also observed in the msh2Δ mlh1Δ double mutant but not in the pms1Δ mlh1Δ mutant (Fig. 6A). All of the MMR mutants showed an increase in the number of spontaneously arising fluconazole-resistant colonies in the zone of clearing (Fig. 6B).
To quantify the increase in the number of spontaneously arising fluconazoleresistant colonies, the colony frequency on 72 g/ml fluconazole (16ϫ MIC) was calculated as described in reference 21. The msh2Δ, mlh1Δ, and pms1Δ mutants all showed a statistically significant increase in the frequency of fluconazole-resistant  . One set of plates was exposed to UV light (120 J m Ϫ2 ). Plates were incubated at 28°C for 2 days. colonies (Fig. 6C). The mlh1Δ mutant displayed a higher number of spontaneously arising fluconazole-resistant colonies than the msh2Δ and pms1Δ mutants that was probably a reflection of the higher MIC ( Fig. 6C and D). Unlike the wild-type results, colonies of the msh2Δ, mlh1Δ, and pms1Δ mutants differed dramatically in size, suggesting that newly acquired fluconazole resistance was occurring frequently throughout the time period of the experiment (Fig. 6D).
The number of colonies arising that were spontaneously resistant to an antifungal unrelated to the azoles, amphotericin B, was also assessed. The colony frequency was calculated on 4.8 g/ml amphotericin B (32ϫ MIC) using the previously determined MIC for KN99␣ (21,28). Similarly to the rapid emergence of resistance to fluconazole, the msh2Δ, mlh1Δ, and pms1Δ mutants all showed a statistically significant increase in the frequency of amphotericin B-resistant colonies compared to the wild type, and the mutant colonies also differed in size from those seen with the wild-type and complemented strains (Fig. 6E and F). The mlh1Δ mutant also displayed a higher number of spontaneously arising amphotericin B-resistant colonies than the msh2Δ and pms1Δ mutants (Fig. 6E).
To assess if the single mutations were resulting in multidrug-resistant colonies, 20 fluconazole-resistant and 20 amphotericin B-resistant colonies isolated from the wild-type strain or the msh2Δ mutant were grown on 72 g/ml fluconazole and 4.8 g/ml amphotericin B. Fluconazole-resistant strains derived from either wildtype or msh2Δ strains were not resistant to amphotericin B and vice versa (data not shown).
To determine if the increased mutation rate had any bearing on clinical treatments, the fluconazole MIC was also determined for the C23 and C45 clinical isolates strains using Etests. The C23 isolate possessed a much higher resistance to fluconazole (32 Ϯ 0.0 g/ml) than the wild-type strain (4.5 Ϯ 0.8 g/ml) (Fig. 6B) and also displayed an increase in the number of spontaneously arising fluconazole-resistant colonies in the zone of clearing (Fig. 7A). In contrast, the C45 isolate possessed a MIC lower than that of the wild-type strain (3.0 Ϯ 1.0 g/ml) and very few spontaneously arising fluconazole-resistant colonies in the zone of clearance (Fig. 7A). To quantify the increase in the number of spontaneously arising fluconazole-resistant colonies, the colony frequency was calculated on 72 g/ml fluconazole (16ϫ MIC) as described in reference 21. The C23 isolate showed a statistically significant increase in the frequency of spontaneously arising fluconazole-resistant colonies (Fig. 7B).
Deletion of PMS1, but not MSH2 or MLH1, results in a reduction in virulence. In a large-scale analysis of virulence, 1,201 defined deletion mutants were bar-coded with 1 of 48 unique signature tag DNA sequences, pooled in groups of 48, and inoculated into the lungs of mice, thus allowing many mutants to be tested in a single-animal infection (13). The msh2Δ, mlh1Δ, and pms1Δ mutants showed increased proliferation in this lung assay of cryptococcosis (13). However, as increased proliferation could be attributed to increased adaptation to the host, to a selective advantage in a competitive environment, or to a population bottleneck, virulence of the MMR mutants was also assessed in a conventional murine inhalation virulence assay. In this assay, the virulence, or fungal load in the lung or brain, of the msh2Δ and mlh1Δ mutants did not significantly differ from that seen with the wild-type or complemented strains (Fig. 8). In contrast, the pms1Δ mutant showed a reduction in virulence and fungal load in the lung, compared to the wild type and the pms1Δ PMS1 ϩ complemented strain (Fig. 8).

A B
WT

DISCUSSION
Transmission of an organism's genetic material through mitosis is extremely accurate due to the activity of a number of systems that detect the introduction of incorrect nucleotides during DNA replication. However, mutations that arise through errors in replication from environmental stresses are biologically important since they provide variations upon which selection can act. Those which confer fitness defects are removed by natural selection, whereas others provide an opportunity for adaptive evolution to occur. One mechanism regulating the control of mutagenesis is the mismatch repair pathway. Mutation of genes encoding components of this pathway can provide an increased mutation rate to enhance the possibility of generating phenotypic traits that could increase survival of a microbial lineage.
This study has shown that mutations in MMR components in the human fungal pathogen C. neoformans result in an increase of over 200-fold in the mutation rate compared to the wild-type rate. This corresponds to approximately 1 mutation per genome per generation in contrast to 1 mutation every few hundred generations for the wild type. The msh2Δ mutation rate is equivalent to that observed in msh2Δ mutants in S. cerevisiae (24). The indistinguishable mutation rates of the msh2Δ, mlh1Δ, and pms1Δ C. neoformans mutants, individually or as double gene deletion strains, indicates that, as in S. cerevisiae, the majority of mismatches repaired by MutS␣/␤ require a Mlh1-Pms1 MutL␣ heterodimer. The C. neoformans MMR mutants also displayed a shift in the mutational profile compared to the wild type. The mutants showed an excess of transition mutations, specifically, G-to-A transitions in the msh2Δ and mlh1Δ mutants, and an increase in single-nucleotide deletions and insertions at homopolymeric tracts. The higher mutation rates compared to those seen with 5-FOAresistant strains determined by the isolation of 5-FU-resistant strains likely reflects the fact that mutations at homopolymeric tracts represent the primary mutational event in MMR mutants. This hypothesis is also supported by the decreased transition rate observed in the fur1 mutants compared to the ura5 mutants, where they were the predominant mutation. A similar shift in mutational profile is also observed in S. cerevisiae msh2Δ mutants, in which an equivalent shift to transitions is observed and a large number of single nucleotide variations (Ϫ1 to ϩ1) occur within homopolymeric tracts (24,25). In S. cerevisiae, single nucleotide variations are also detected within dinucleotide/trinucleotide tracts (microsatellites) (24,25). These types of mutations could not be detected with the assays used in this C. neoformans study. This is one of the limitations of using genetic reporter assays to measure mutation rates and to analyze the mutational spectrum. Rates of phenotypically silent mutations are not measured, as evidenced by the frequent deletion of nucleotides in the (T) 14 homopolymer of FUR1 in the 5-FU-resistant msh2Δ, mlh1Δ, and pms1Δ isolates, in combination with mutations in the (C) 7 and (A) 6 homopolymers, as the (T) 14 homopolymer falls within an intron and is therefore silent (Fig. 3D). Despite the increased mutation rate and the shift in the mutational profile, the MMR mutants did not show a decrease in viability in vitro or for the msh2Δ and mlh1Δ mutants in a mouse model of disease. The majority of mutations are likely to occur in noncoding DNA or to be silent, and it is expected that only a minute fraction would be deleterious. Overall, these results suggest that the role of Msh2, Mlh1, and Pms1 in nuclear mismatch repair is conserved in C. neoformans. The types of mutations generated suggest that Msh2, Mlh1, and Pms1 correct errors in nuclear DNA arising both from errors in replication and from oxidative damage. G-to-A transitions, followed by G-to-T transversions, are the most commonly observed mutations arising from oxidative damage to DNA but can also be generated through misincorporation of dUTP instead of dTTP during replication or deamination of cytosine (29). The resulting uracil in DNA is typically removed by the homologue of the Ung1 uracil DNA N-glycosylase of the base excision repair (BER) pathway (30). G-to-A transitions also occur by the deamination of methylated cytosine to thymine. However, none of the G-to-A transitions observed in the MMR mutants were located at CpG sites, the typical sites of cytosine methylation, suggesting that the increase in this transition type is not due to a failure to repair damaged methylated cytosine. A lack of repair of guanine oxidized to 8-oxo-7,8-dihydroguanine (8-oxoG) can also result in an increased number of G-to-A transitions (depending on the sequence context) but normally results in an abundance of G-to-T transversions, which was not observed in the MMR mutants (27,29). Therefore, it is not clear why there was an abundance of G-to-A transitions in the MMR mutants. DNA slippage events and misincorporation of nucleotides during the synthesis of the leading and lagging strands by the DNA polymerases (Pol2) and Ѩ (Pol3), respectively, give rise to single nucleotide variations within homopolymeric tracts and transition and transversion mutations (31,32). However, the majority of mutations are generated by errors during lagging-strand synthesis by DNA polymerase Ѩ (Pol3), which shows inefficient proofreading and, consequently, lower fidelity than that seen with DNA polymerase (Pol2) (31). MMR is more efficient at repairing lagging-strand errors generated by DNA polymerase Ѩ (Pol3) as the MutS␣/␤ complex is directed to the 5= ends of Okazaki fragments and is tethered to the proliferating cell nuclear antigen (PCNA), a DNA clamp that acts as a processivity factor for DNA polymerase Ѩ (Pol3) (31)(32)(33). It is interesting that the mutational profile and phenotype of the pms1Δ mutant differ from those of the msh2Δ and mlh1Δ mutants, and this may relate to the reason that the pms1Δ mutant strain shows attenuation of virulence whereas the other two mutants do not. Although the pms1Δ mutant displays a shift in the mutational profile to more transition mutations and mutations in homopolymeric tracts, like the msh2Δ and mlh1Δ mutants, the types of transition mutations generated differ. This suggests that Pms1 is playing a role in addition to that of the MutL␣ heterodimer in nuclear mismatch repair, possibly through interaction with the PCNA. The range of transitions generated in the pms1Δ mutant may result in more deleterious mutations than in msh2Δ and mlh1Δ mutants, as a greater range of codons can be mutated and as C-to-T transitions also result in premature stops, unlike G-to-A transitions. In addition, little phenotypic variation was observed in the pms1Δ strains passaged for 600 generations, suggesting that most of the mutations that occurred were lethal and were removed from the population by natural selection. Thus, the reduction in virulence observed in the pms1Δ mutant likely arises due to decreasing survival over time. In contrast, passaging the msh2Δ and mlh1Δ mutants through population bottlenecks allowed nonlethal mutations to accumulate as if they were neutral by mitigating the effect of selection.
Mutations resulted in measurable phenotypic change, suggesting that those arising in the msh2Δ and mlh1Δ mutants provide a powerful mode of rapid microevolution in C. neoformans. We observed no difference in proliferation or virulence in the msh2Δ and mlh1Δ mutants in a standard murine inhalation virulence assay. This is in contrast to the original analysis of these mutants, which showed they had increased proliferation in vivo (13). In that experiment, however, 48 pooled mutants were simultaneously inoculated into mice to allow large-scale screening and a less virulent genetic background was used. It is possible that an increase in proliferation would be observed only in a strain with lower virulence than that seen in the KN99 background, when the mutants are in a competitive environment, or when a small number of cells are inoculated, effectively creating a population bottleneck. Similarly to C. neoformans, equivalent levels of mice colonization and infection were also observed between the wild type and the msh2Δ mutant in a standard C. glabrata virulence assay (21). Rather than having a direct impact on the virulence composition of isolates, MMR mutations are more likely to be a significant factor in allowing outbreaks to occur via the generation of hyper-virulent strains from less-virulent ancestors and by allowing recurrence of cryptococcal meningitis through the generation of antifungal resistance.
Microevolution by single-base-pair mutations may potentially play a pivotal role in the development of infectious outbreaks from less-virulent ancestors. One example comes from the outbreak of cryptococcosis that began in the Pacific Northwest of the United States and western Canada in the late 1990s. One of the three clonal subpopulations causing the outbreak consisted of two distinct groups: a group of less-virulent strains containing a frameshift mutation in MSH2 and a group of highly virulent outbreak strains in which the MSH2 mutation is hypothesized to have reverted (34). This suggests that a transient mutator phenotype could have contributed to microevolution and adaptation of virulence in this cryptococcosis outbreak. A transient mutator provides a scenario by which mutators can benefit temporarily from elevated mutation frequencies for adaptation and yet can reduce the risk of accumulating deleterious mutations.
The mutator phenotype of the C. neoformans MMR mutants also promoted rapid development of antifungal drug resistance, as observed by the frequently arising fluconazole-and amphotericin B-resistant colonies. Patients treated for cryptococcal meningitis often relapse due to gain of resistance to antifungal drugs (35). Mutations in MSH2 provide the pathogen with an opportunity for the rapid evolution of resistance to antifungal drugs and could be a cause of patient relapse. Recently, in a study investigating the causes of recurrence of meningitis due to C. neoformans, one isolate with a mutation in the MSH2 gene was identified (35). Deletions of MSH2 in C. glabrata also result in high levels of spontaneous antifungal resistance in vitro, in addition to an increase in the occurrence of resistance to caspofungin arising during in vivo growth (21). MSH2 mutations are present in a higher percentage in fluconazole-resistant clinical isolates than in susceptible strains in C. glabrata (21). In addition, a comparison of paired or triplet isolates from the same patient showed that the MSH2 mutation predated the emergence of fluconazole resistance (21). Likewise, deletion of MSH2 and PMS1 in C. albicans also resulted in resistant colonies arising in the inhibition ellipse of fluconazole Etests, although the rate was not as high as that observed in C. neoformans or C. glabrata (as expected due to the diploid state of C. albicans) (36). Antifungal resistance in C. neoformans is often associated with changes in karyotype (reviewed in reference 6). Fluconazole heteroresistance occurs in C. neoformans via successive duplication of chromosomes 1, 4, 10, and 14 on increasing concentrations of fluconazole. These disomic chromosomes allow the rapid emergence of fluconazole resistance; however, the disomic chromosomes are lost when fluconazole is removed, presumably due to reduced fitness as evidenced by reduced proliferation and virulence (10). The emergence of antifungal resistance via single-base-pair mutations in MMR mutants may provide C. neoformans with another option that results in stable, long-term fluconazole resistance without associated fitness defects.
This study showed that a number of C. neoformans clinical isolates carry mutations in MMR components that give rise to a hypermutation phenotype, high levels of fluconazole resistance, and the ability to rapidly generate spontaneous fluconazoleresistant subisolates. In other microbial populations, retaining a proportion of mutator cells in the population is thought to provide a selective advantage in vivo. Fifty-five percent of C. glabrata clinical isolates recovered from patients are msh2 mutants (other components of the MMR pathway were not sequenced) (21). In addition, the bacterial pathogens Staphylococcus aureus, Eschericha coli, and Salmonella enterica retain a larger-than-expected proportion of mutator cells in their populations with defects in components of MMR, and 20% of Pseudomonas aeruginosa isolates from cystic fibrosis patients are mutators compared to 0% of environmental isolates (17)(18)(19)(20). This suggests that this adaptive mechanism may be conserved in a broad variety of pathogens. Whether C. neoformans clinical populations contain a high proportion of MMR mutants has yet to be determined and requires investigation of larger sample sizes for valid analysis of the proportion of clinical versus environmental isolates that carry an MMR mutation. What is clear is that mutations in MMR components provide C. neoformans Testing for sensitivity to or tolerance of DNA-damaging agents. Strains were cultured overnight in YPD medium, 10-fold serially diluted, and plated onto YPD medium with or without the following stress agents: 0.04% methyl methanesulfonate (MMS) (Sigma-Aldrich; catalog no. 129925), 0.06% ethidium bromide solution (EtBr) (Sigma-Aldrich; catalog no. E1510), 0.25 mM menadione (Sigma-Aldrich; catalog no. M5625), 0.25 mM paraquat dichloride hydrate (paraquat) (Sigma-Aldrich; catalog no. 36541), 0.4 mM Luperox TBH70X tert-butyl hydroperoxide solution (tBOOH) (Sigma-Aldrich; catalog no. 458139), and 5 mM hydrogen peroxide solution (H 2 O 2 ) (Sigma-Aldrich; catalog no. 216763). One set of plates was exposed to UV light (120 J m Ϫ2 ). Plates were incubated at 28°C for 2 days.
Phenotypic characterization of microevolving strains. The wild-type strain and the msh2Δ, mlh1Δ, and pms1Δ mutants were streaked for single colonies on YPD medium or YPD medium-nourseothricin (100 g/ml) plates and incubated at 28°C. Single colonies were restreaked every 7 days for 2 months (approximately 600 generations). Three independent strains were generated for the wild-type strain and each mutant (Table S2). Strains were then cultured overnight using YPD medium, 10-fold serially diluted, and plated onto YPD medium with or without the following stress agents: 0.25 mM paraquat dichloride hydrate (paraquat) (Sigma-Aldrich; catalog no. 36541), 5 mM hydrogen peroxide solution (H 2 O 2 ) (Sigma-Aldrich; catalog no. 216763), 1.5 mg/ml calcofluor white, and 5 mg/ml Congo red. Plates were incubated at 28°C for 2 days. One set of plates was incubated at 37°C. Strains were also plated on L-DOPA (L-3,4-dihydroxyphenylalanine) medium to assess melanization.
Antifungal resistance. Fluconazole Etests were performed on yeast nitrogen base agar plates (pH 7.0) per the manufacturer's instructions and the method described in reference 40. To assess the frequency of colony formation on fluconazole or amphotericin B, 1 ϫ 10 5 cells of each strain from an overnight YPD culture were used to inoculate 3 separate YPD cultures. After growth for 48 h in a roller drum at room temperature, 1 ϫ 10 6 cells were plated onto YPD medium plus 72 g/ml fluconazole (16ϫ MIC) or 4.8 g/ml amphotericin B (32ϫ MIC) and the plates were incubated at 30°C for 7 days. Frequencies were calculated as the number of colonies on the drug plate divided by the total count of CFU plated. Frequency averages and standard errors of the mean were calculated using Prism 4.0c. Two-tailed Student's t tests were performed to determine statistical significance.
Virulence studies in vivo and fungal burden in infected organs. Yeast strains were grown at 30°C overnight, and cultures were washed twice with phosphate-buffered saline (PBS) and resuspended at a final concentration of 2 ϫ 10 5 cells/ml. Groups of female A/Jcr mice (Jackson Laboratory, Bar Harbor, ME) were infected intranasally with 1 ϫ 10 4 yeast cells of each strain. Over the course of experiments, animals that appeared moribund or in pain were sacrificed by CO 2 inhalation. Survival data from the murine experiments were statistically analyzed between paired groups by using the log rank test in Prism program 4.0 (GraphPad Software, Inc., San Diego, CA). P values of Ͻ0.01 were considered significant.
To compare fungal burdens, infected lungs and brains at the endpoint of the experiment were isolated and homogenized in PBS buffer using a homogenizer. Resuspensions were diluted, and 100 l of each dilution was spread on YPD medium. Fungal colonies were counted after 3 days of incubation at 30°C. All statistical analyses were undertaken using the two-tailed Student's t test. P values of Ͻ0.05 were considered statistically significant.
Ethics statement. This study was performed according to the guidelines of NIH and Institutional Animal Care and Use Committee (IACUC). The animal models and procedures used have been approved by the IACUC at Rutgers University (protocol 15041D0518).