Functional Profiling of Human Fungal Pathogen Genomes

Abstract

Fungal infections are challenging to diagnose and often difficult to treat, with only a handful of drug classes existing. Understanding the molecular mechanisms by which pathogenic fungi cause human disease is imperative. Here, we discuss how the development and use of genome-scale genetic resources, such as whole-genome knockout collections, can address this unmet need. Using work in Saccharomcyes cerevisiae as a guide, studies of Cryptococcus neoformans and Candida albicans have shown how the challenges of large-scale gene deletion can be overcome, and how such collections can be effectively used to obtain insights into mechanisms of pathogenesis. We conclude that, with concerted efforts, full genome-wide functional analysis of human fungal pathogen genomes is within reach.

Genome-scale genetic resources (e.g., whole-genome knockout collections) can provide a detailed understanding of fungal pathogenesis. Studies in Cryptococcus neoformans and Candida albicans are the most advanced.

The goal of this article is to describe the promise and challenges of systematic and comprehensive functional approaches to decipher the mechanisms of pathogenesis by human fungal pathogens. To set the stage, we begin with a description of the multifaceted roles of fungi in human affairs before introducing the two broad classes of fungi capable of infecting humans.

FUNGI IN HUMAN WELFARE

The fungal kingdom is arguably the most diverse and successful of the eukaryotic kingdoms. With more than 100,000 described and five million inferred fungal species inhabiting most niches in the biosphere, only bacteria match the diversity of fungi (O’Brien et al. 2005). Fungi are typically saprophytic organisms that extract energy by decomposing organic matter and, thus, play an essential role within the environment: recycling organic matter. Fungi also have many important industrially useful properties. Besides the familiar edible mushroom and brewer’s and baker’s yeast, other fungi produce a variety of bioactive molecules, many of which are well-known drugs, such as penicillin. A major threat to humans from fungi is the destruction of crops. Collectively, plant pathogenic fungi cause 10%–40% of crop loss worldwide (Fisher et al. 2012).

HUMAN FUNGAL PATHOGENS

Although many fungi are plant pathogens, relatively few (about 200 species) can infect animals or humans (Richardson and David 2003). It has been suggested that this is because most fungi cannot grow at mammalian body temperature (indeed, it has been suggested to be why mammals have been successful). Another likely reason is that the innate and adaptive immune systems of mammals provide an effective defense against infection by most fungi. A fungal infection, or mycosis, can be categorized as cutaneous, subcutaneous, or systemic. Cutaneous mycoses are quite common, for example, tinea (“ringworm”) infections caused by Microsporum, Trichophyton, and Epidermophyton fungi, but these are generally easily resolved. Systemic infections are a major clinical problem and challenging to diagnose and treat. They are caused either by fungi that are primary pathogens (e.g., Histoplasma, Blastomyces, and Coccidioides species [spp.]), which infect immunocompetent hosts, or opportunistic pathogens that require some degree of host compromise for infection (e.g., Aspergillus, Candida, Cryptococcus, Murcor spp.).

Presumably, the ability of these pathogens to cause disease involves the requirement for a specific set of properties, not shared with most fungi. Work over the last few decades has shed some light on the underlying molecular mechanisms. There are several simple requirements: proliferation at body temperature of 37°C, adaptation to hostile environments limited in essential nutrients, and specialization to overcome otherwise effective immune barriers, such as recognition and phagocytosis, and killing by macrophages (Levitz and DiBenedetto 1989). Fungi appear to have found a range of solutions for these problems, emphasizing the need for functional studies of a range of pathogens.

THE SYSTEMATIC PERSPECTIVE

Considering the question of how to most effectively approach the biology of the human pathogenic fungi, it is clear that comprehensive genetic and genomic resources represent an efficient set of tools that would enable the field to develop a detailed molecular understanding of fungal mechanisms of infection and drug sensitivities. In support of this perspective, we will briefly highlight the success of such approaches in baker’s yeast, as well as the initial applications of systematic genome-scale genetic studies in human fungal pathogens. Of the many important human fungal pathogens, the application of the approaches described above is most advanced in Cryptococcus neoformans and Candida albicans and, thus, these will be the focus here. However, the approaches are broadly applicable once DNA transformation and homologous replacement strategies are developed in a given organism. The prospect for the latter is rapidly improving because of the development of CRISPR-based genome editing technology (Mali et al. 2013).

Saccharomyces cerevisiae AS A PROTOTYPE

Our perspective is informed by the recent history of genomics, which was driven by studies of Saccharomyces cerevisiae. We review the highlights below.

From Genome to Public Resources

S. cerevisiae was the first eukaryotic organism whose genome was sequenced and annotated (Goffeau et al. 1996). The available genomic data paved the way for the undertaking of the construction of a deletion library for every nonessential gene, as well as a diploid heterozygous for essential genes (Giaever et al. 2002). Critical for the success of this effort was the fact that homologous recombination is unusually efficient in S. cerevisiae, with 50 bp being sufficient for targeting homology (Bahler et al. 1998; Longtine et al. 1998). This allows cloning-free, polymerase chain reaction (PCR)–based construction of targeting vectors. Moreover, transformation using lithium chloride and polyethylene glycol has been highly optimized, as has the ability to rapidly screen colonies for homologous replacement events (Fink and Guthrie 2002). In addition to the well-known gene-deletion library, a number of useful ordered whole-genome libraries have been created. Among these is a library that allows for the overexpression and affinity purification of every yeast protein, which has allowed for efficient activity-based biochemical screens (Martzen et al. 1999), as well as a comprehensive library of genes tagged with green fluorescent protein, which allowed for the determination of protein localization at an unprecedented scale (Ghaemmaghami et al. 2003)

The Efficiency of the Systematic Approach

The ease of genetic manipulations, which include sexual crosses, combined with the ability to systematically assay the behavior of stains lacking every nonessential gene in a controlled, parallel manner under any desired laboratory condition or in combination with other mutations (synthetic genetic array), rapidly enhanced the understanding of the biology of S. cerevisiae (Botstein and Fink 2011). Two additional advances enhanced the utility of the deletion collection. First, robotics for moving arrays of colonies between agar plates was developed. Second, strains and methods were developed that allowed for the selection of the products of a sexual cross in a high-throughput manner (Tong and Boone 2006). The most impressive example of the application of these tools was the generation and analysis of 5.4 million double mutants (Costanzo et al. 2010). Such double mutant analysis (also termed genetic interaction mapping) has been shown to be very useful for linking genes to functional categories or physiological processes. Deletions of single genes often show no easily detectible phenotypes because other genes can perform similar functions inside the cell and can, therefore, “cover” for the absence of the deleted gene. This is the notion of redundant function and, at one extreme, only the double mutant displays a phenotype, producing a so-called synthetic genetic interaction. In reality, nearly all single-gene deletions produce a measurable phenotype (Breslow et al. 2008), and it is in cases in which the double mutant produces more than a multiplicative defect that it is considered a synthetic interaction.

With the availability of the ordered genetic arrays of deletion mutants, the rapid screening for genes whose absence causes sensitivity or resistance to biological compounds of interest (chemogenomics) has become possible (Wuster and Madan Babu 2008). Such studies have been useful for determining drug mode-of-action and gene annotation (as gene deletions with similar phenotypic signature tend to specify factors that function in the same pathway). Because the individual deletion strains of the yeast deletion collection each carry a unique DNA bar code, experiments in which the entire deletion collection is pooled and subject to chemical challenge have also been performed. Custom bar code microarrays or, more recently, bar code sequencing, have been used to assess the fitness of individual strains in the pool after treatment. Such systematic chemical-genetic screens are now routinely used (Lopez et al. 2008; Wuster and Madan Babu 2008).

In addition to allowing for the systematic generation of deletions, the availability of the annotated genomic sequence for budding yeast led to the generation of the first whole-genome DNA microarrays, which allowed for the assessment of messenger RNA accumulation for all genes at the same time (expression profiling) (DeRisi et al. 1997). This ability to assay the expression of every gene at once transformed how questions are framed and investigated (Botstein and Fink 2011). By determining what sets of genes are induced or repressed by a given condition, one can quickly formulate hypotheses as to how the condition affects the physiology of the cell. Although microarray technology is being superseded by RNA-seq, expression profiling is a mainstay in genetic analysis. Likewise, systematic whole-genome chromatin immunoprecipitation of all annotated DNA-binding regulatory proteins has impacted studies of the structure, function, and evolution of transcription circuits. The latter is only beginning to be applied to fungal pathogens, whereas transcriptomics is widespread in its application attributable to the ease of isolating RNA from cells.

The Functionally Annotated Genome

As a result of such large-scale studies, most S. cerevisiae genes have been assigned a function or role in a particular physiological process. The assignments produced by the combination of systematic approaches range in specificity, from the fine-grained and mechanistic to high-level and diffuse. Nonetheless, the current state of genome annotation vastly exceeds what is available in other systems (Dolinski and Botstein 2005). An important aspect of this recent history is the availability of an effective annotation database curated by PhD-level scientists (Saccharomyces genome database, www.yeastgenome.org). In addition to allowing for rapid analysis, the availability of these systematic resources also allows for comprehensive analysis such that entire pathways can be defined. Importantly, a scientific culture of sharing and dissemination of resources has also accelerated progress.

Can We Systematize the Analysis of Human Fungal Pathogens?

S. cerevisiae is not a human pathogen, and it is known that fungal pathogens harbor mechanisms and pathways required for virulence that either do not exist in model yeasts or have been substantially repurposed during their evolution. Thus, the ability to dissect biology of infection and drug responses in these organisms in an efficient and comprehensive manner requires the development of genetic resources and, equally important, the effective systematic analysis of such resources. However, accomplishing this task in a human fungal pathogen is far more challenging than in S. cerevisiae for a variety of technical reasons discussed below. Despite these challenges, substantial beginnings of such efforts have been accomplished in two human fungal pathogens, C. neoformans and C. albicans. Although random insertional mutagenesis methods are effective tools for identifying mutants, they have disadvantages for the application of systematic approaches. Thus, we will focus on what can be accomplished with ordered collections. For Cryptococcus and Candida, we will describe the types of technologies and approaches that allowed the challenges to be addressed successfully, and insights gained from the systematic mindset (Table 1).

Table 1.

Summary of genomic resources available for C. neoformans and C. albicans and some of the insights they have provided

Organism	Genomic resource	Screens	Major findings
C. neoformans	1201 gene-deletion library	In vivo infectivity (Liu et al. 2008)	Identified 197 genes with increased or decreased virulence
		Phagocytosis resistance (Liu et al. 2008)	Identified critical role for GAT201 in preventing engulfment by macrophages
		Melanin production (Liu et al. 2008)	Identified 38 genes with defective melanin production; many of the genes also had virulence defects
		Growth at 37°C (Liu et al. 2008)	Identified 92 genes important for proliferation at body temperature
		Capsule production (Liu et al. 2007, 2008)	Identified CPL1, GAT201, PBX1, and PBX2 as regulating capsule production or assembly
		Growth under hypoxia (Chang et al. 2007; Chun et al. 2007)	SREBP pathway controls growth under hypoxia
C. albicans	∼700 homozygous gene-deletion library	In vivo infectivity (Noble et al. 2010)	Identified 115 mutants defective in infectivity; 46 of those had no defects in morphogenesis
		Growth at 37°C (Noble et al. 2010)	Identified 68 mutants defective in proliferation
		Colony morphology (Noble et al. 2010)	Identified 133 mutants defective in morphogenesis; 66 of those had no defect in infectivity
	143 homozygous transcription factor deletion library	Tested 55 conditions (Homann et al. 2009)	Most mutants affect growth under at least one condition
		Surface adherence (Finkel et al. 2012)	Identified the transcription factor SNF5 as a key regulator of adhesion
	GRACE collection (∼1100 conditional mutants)	Growth requirement (Roemer et al. 2003)	Identified 567 essential genes

See text for details. GRACE, gene replacement and conditional expression; SREBP, sterol regulatory element-binding protein.

C. neoformans: SYSTEMATIC GENETICS IN A HAPLOID FUNGUS

C. neoformans Biology, Life Cycle, and Genome

C. neoformans is a basidiomycetous fungus that is an opportunistic human pathogen, infecting, predominantly, individuals with a compromised immune system. C. neoformans is the leading cause of fungal meningitis and causes more than 600,000 deaths per year in sub-Saharan Africa alone. Cryptoccous is found growing in the environment, typically on decaying wood matter or bird guano (Lin and Heitman 2006). Inhalation of desiccated yeast is likely the major route of infection. Cryptocccus species that infect humans belong to four different serotypes (A–D): Serotype A is termed C. neoformans var. grubii, serotype B and C are called Cryptocccus gatii, and serotype D is called C. neoformans var. neoformans. The different serotypes of C. neoformans display difference in phenotypes and genotypes (Lin et al. 2005; Lee et al. 2009). Serotypes A and, to a lesser extent, D account for most clinical isolates from immunocompromised patients, whereas the C. neformans var. gatii serotype B/C isolates tend to be isolated from more immunocompetent individuals (Lin and Heitman 2006). Virulence factors of C. neoformans include the ability to grow at 37°C and produce melanin and a polysaccharide capsule (Idnurm et al. 2005; Lin and Heitman 2006).

Because of its medical importance, clinical C. neoformans isolates belonging to serotype A and D have been sequenced and their genomes have been annotated. Draft sequence of a serotype B isolate is also available (Broad Institute Serotype B Genome Project, www.broadinstitute.org/annotation/genome/cryptococcus_neoformans_b). C. neoformans var. neoformans has a 19-Mb genome containing 14 chromosomes encoding 6962 predicted proteins (Loftus et al. 2005). The genome of C. neoformans var. grubii is very similar in size. C. neoformans has a complete sexual cycle with two mating types, thus allowing for standard genetic crosses (Hull and Heitman 2002; Lin and Heitman 2006; Idnurm 2010). A current deficit in C. neoformans genomics is that there is no publicly funded annotation database akin to the Saccharomyces genome database or the Candida genome database. However, Broad Institute maintains the sequence database for the H99 serotype A reference strain genome (Broad Institute Serotype A Genome Project, www.broadinstitute.org/annotation/genome/cryptococcus_neoformans).

Gene-Deletion Methods C. neoformans

C. neoformans can be transformed by electroporation or bombardment by DNA-coated particles (biolistic transformation) (Toffaletti et al. 1993). The latter is the mainstay of work in the most common clinical serotype, serotype A (C. neoformans var. grubii). Early studies showed that substantial homology was required for homologous gene replacement, with replacement frequencies increasing steadily until homology reached ∼1 kb on either side of a selectable marker (Davidson et al. 2000; Nelson et al. 2003). Cloning-free fusion PCR methods were found to be suitable for the addition of long homology arms for gene targeting, which aided in the rapid synthesis of targeting constructs. These advances, together with the development of methods for rapid screening of colonies (which was nontrivial for an encapsulated organism) and optimization of transformation parameters, made it feasible to construct large numbers of gene deletions in C. neoformans (Chun and Madhani 2010). In addition, the development of multiple dominant drug selectable marker systems was also critical. This molecular toolbox, together with a complete sexual cycle, make C. neoformans a particularly experimentally tractable pathogen, perhaps the most tractable eukaryotic pathogen of humans.

Systematic Genetic Analysis of Virulence in C. neoformans

To systematically identify virulence factors in Cryptococcus, our laboratory previously constructed a collection of approximately 1200 deletion mutants and screened them in vitro and in vivo for virulence phenotypes (Liu et al. 2008). Each deletion harbored one of 48 DNA oligonucletoide bar codes (Gerik et al. 2005) that enabled quantitative detection of each mutant in a pooled sample. Mutants were pooled in groups of 48 and used to infect mice with an intranasal inhalation model that is thought to mimic the natural route of infection. After allowing for infection to be established, mice were killed, and C. neoformans cells were isolated from the lungs. The relative representation of each of the 48 mutants could be assessed by quantitative PCR using 48 bar-code-specific primer pairs on genomic DNA prepared from cells obtained from the mouse versus those derived from a portion of the input pool. Mutants that were significantly reduced in abundance were reconstructed to verify linkage between genotype and phenotype. Importantly, nearly all mutants that were reconstructed maintained their original phenotypes, indicating that unlinked mutations rarely caused phenotypes in the deletion strains. Moreover, all mutants that displayed a defect in a pooled setting also displayed the anticipated impact on virulence in a single infection experiment. Many of the identified mutants were defective in the production of the well-characterized virulence factors: capsule, melanin or growth at 37°C. However, as many as 40 virulence mutants without clear defects in the well-established virulence markers were also isolated (Liu et al. 2008). These are outstanding candidates for factors that participate in novel virulence pathways. These studies validated the use of large-scale signature-tagged mutagenesis screens for C. neoformans using comprehensive collections of mutants. Two additional screens of this gene-deletion collection (described below) revealed new aspects of cryptococcal disease.

A Major Role for Capsule-Independent Phagocytosis Inhibition in Cryptococcal Disease

A great advantage of a deletion collection is that the same collection can be screened for any phenotype of interest. One such study asked the question of which genes mediate the ability of C. neoformans to resist phagocytosis. C. neoformans can be phagocytosed by activated macrophages in vitro (Levitz and DiBenedetto 1989; Liu et al. 2008). The expectation would be that genes required for the synthesis of the C. neoformans capsule would be the critical factors. Remarkably, screening of the 1200 deletion library on RAW264.7 macrophage cells revealed that a deletion of the gene coding for the GATA family transcription factor Gat201 had caused an increase in phagocytosis that was far greater than that caused by lack of capsule (Liu et al. 2008). Additional genetic studies showed that this phenotype did not require the major capsular polysaccharide (Liu et al. 2008). In subsequent work, the first whole-genome chromatin immunoprecipitation studies of C. neoformans were reported that identified the direct targets of Gat201 (Chun et al. 2011). Systematic deletion of the Gat201 target genes identified two genes, Gat204, encoding another GATA family transcription factor, and Blp1, a member of a family of C. neoformans–specific proteins that each harbors a putative amino-terminal signal sequence, which played a key role in mediating phagocytosis inhibition and pathogen fitness during infection. Further evidence indicated that this was independent of the production of polysaccharide capsule, underlining the conclusion that phagocytosis avoidance and virulence in Cryptococcus requires mechanisms that have yet to be identified (Chun et al. 2011).

A Role for the SREBP Pathway in Virulence and the Discovery of Novel Virulence Factors

Another independent screen of the deletion collection asked the question of whether C. neoformans experienced hypoxia during infection and whether this was important for survival in the host environment. This screen revealed two pathways required for hypoxic adaptation. Both played a role in C. neoformans fitness, implying that the ability to respond to hypoxia is critical for infection (Chun et al. 2007). One of the two pathways encodes a fungal analog of the sterol-sensing SREBP (sterol regulatory element-binding protein) system of mammals. In the mammalian system, SREBP is processed from a membrane-bound precursor in response to sterol depletion in the membrane, yielding an active transcription factor that translocates into the nucleus to activate cholesterol biosynthesis genes. This system had been shown to function as hypoxia sensor in Schizosaccharomyces pombe because of the requirement for molecular oxygen in fungal ergosterol biogenesis (Hughes et al. 2005). A similar mechanism was shown to operate in C. neoformans (Chun et al. 2007). Directed studies of the SREBP pathway came to the same conclusions (Chang et al. 2007).

Although the polysaccharide capsule has been the subject of genetic screen for decades, the systematic approach identified a number of novel players in capsule production. Two of these genes, PBX1 and PBX2, were initially identified on the basis of their dry colony phenotype, which hinted at a capsule defect (Liu et al. 2007). These proteins encode paired β-helix proteins. Many proteins with this structure play a role in carbohydrate biochemistry. Indeed, the two mutants, although capable of producing capsule monomers, were defective in properly assembling the capsule (Liu et al. 2007). The biochemical function of Pbx1 and Pbx2 is at present unclear, but likely plays a key role in capsule assembly, a process that is poorly understood.

The examples described above show the utility and efficiency of screening a large deletion collection, and emphasize the additional utility of construction and analysis of a comprehensive collection.

Cross-Species Genetic Interaction Mapping: Leveraging S. cerevisiae to Study Pathogen Proteins

Because many C. neoformans genes that appear relevant for virulence lack assigned biological function, the genetic tools available for Saccharomyces have been used to uncover the function of C. neoformans genes (Brown and Madhani 2013). The cross-species genetic interaction approach consists of introducing genes of interest from Cryptococcus into the S. cerevisiae deletion collection and measuring their genetic interactions, assessed as growth phenotypes, with any of the S. cerevisiae gene deletions. Based on the plethora of existing data sets of genetic interactions of the identified S. cerevisae genes, one can predict the function of the tested C. neformans gene. Six C. neoformans genes with unknown function, but important for virulence, were initially tested. One of the tested genes, LIV7, had a genetic interaction profile similar to that of S. cerevisae mutants defective in Golgi transport, indicating that Liv7 may be involved in vesicle trafficking, a hypothesis that was tested and supported by further evidence (Brown and Madhani 2013). Predictions about the function of the remaining five tested C. neoformans genes could also be made (endosome vesicle fusion, proteolysis, cell cycle progression, etc.), and preliminary experiments indicate these predictions to be correct (Brown and Madhani 2013).

Lessons Learned

Given its genetic tractability, Cryptococcus represents an obvious test bed for the application of genomics. The concerted investment into methods development by a small number of laboratories was critical in producing the conditions conducive to the construction of the first large-scale genetic resource and its use in studies of pathogenesis. Ongoing efforts are aimed at constructing a whole-genome deletion collection, which we anticipate to have a significant impact on studies of this important human pathogen.

C. albicans: SYSTEMATIC GENETIC ANALYSIS IN A DIPLOID COMMENSAL PATHOGEN WITHOUT A COMPLETE SEXUAL CYCLE

C. albicans Biology, Life Cycle, and Genome

C. albicans, an ascomyceteous fungus, is a commensal of humans found in the gut, skin, and mucous membranes. A common cause of mucosal infections, C. albicans can also cause serious disseminated infection in immunocompromised individuals. According to the Centers for Disease Control, C albicans is the fourth most common cause for hospital acquired bloodstream infections. In response to environmental stimuli, such as pH, hypoxia, and starvation, C. albicans can switch from yeast to hyphal, invasive growth (Whiteway and Bachewich 2007). The switch between yeast and hyphal forms is crucial for pathogenesis (Calderone and Fonzi 2001). The potential sexual cycle of Candida has been the subject of much interest. Although haploid strains can be selected in the laboratory (Hickman et al. 2013), naturally occurring strains are diploid or close to diploid. Candida is thought not to undergo meiosis, a specialized cell cycle during which diploid cells produce haploid gametes, as it lacks many of the key genes that regulate meiosis that are conserved from Saccharomyces to humans. Nonetheless, C. albicans encodes a mating-type locus encoding DNA-binding regulators of cell type (Hull and Johnson 1999). Derivatives capable of mating can be constructed (Hull et al. 2000). These can be used to derive tetraploids, which then lose chromosomes, presumably via errors in chromosome segregation, to produce diploids (Bennett and Johnson 2005).

The C. albicans genome encodes more than 6200 genes, the majority of which (more than 4500 genes) are considered uncharacterized according to the C. albicans genome database (www.candidagenome.org). Aneuploidy is common, as is heterozygosity (Selmecki et al. 2006). These factors and the diploid nature of C. albicans represent enormous challenges for functional analysis. Nonetheless, steady advances in genetic and genomic tool production have led to the development and use of several large-scale genetic resources.

Gene-Deletion Methods in C. albicans: Technical Challenges and Solutions

Because most laboratory C. albicans stains are diploid, introducing recessive mutations in the genome requires two successive transformation steps to modify both copies of the locus of interest (Fonzi and Irwin 1993). Initially, the strategy applied to introduce mutations used ura3 deficient mutants and replaced the gene of interest with a URA3 cassette (Fonzi and Irwin 1993). Flanking of the URA3 gene with repeats allowed for the recycling of the marker using the negative selective agent 5-fluororotic acid. The sequential deletion of the two alleles was challenging for two reasons. First, the marker had to be recycled, and, second, there was no way to select for homologous recombination at the unaltered (vs. deleted) allele. A clever genetic trick was developed by Mitchell and colleagues that used an ARG4 gene flanked by overlapping fragments of the URA3 gene (the UAU1 marker) (Nobile and Mitchell 2009). This construct allows for a double selection, thus, greatly increasing the likelihood for cells lacking both wild-type alleles. This was used to generate the first-ordered homozygous mutant libraries for C. albicans and discovery of the first regulators of biofilm formation (see below). Biofilms are thought to be important clinically as a significant number of Candida infections are associated with indwelling catheters. This early application of an unbiased approach is what led to the discovery of what we now know to be a complex regulatory circuitry that controls biofilms (Nobile et al. 2012). Although clearly powerful, the UAU1 method suffers from the complication that it relies on secondary genetic events that are difficult to control and can result in loss-of-heterozygosity for large regions (Nobile and Mitchell 2009).

Furthermore, URA3 has turned out to not be the ideal marker for studies of C. albicans. ura3 null mutants are defective in both morphogenetic switching an pathogenesis (Lay et al. 1998; Cheng et al. 2003; Brand et al. 2004). Moreover, strains complemented with the wild-type URA3 gene can suffer from position effects that impact the expression of URA3 and the associated phenotypes it controls (Brand et al. 2004). These findings raised the possibility that studies investigating the virulence of C. albicans strains in the ura3 genetic background may have missed important aspects of C. albicans virulence. These findings and other challenges spurred the development of strains better suited to genetic investigations of pathogenesis mechanisms. These strains used the arg4, leu2, and his1 loci, which have no detectible virulence or commensalism phenotypes (Noble and Johnson 2005). Moreover, new heterologous complementing auxotrophic markers and cloning-free methods for the generation of homologous targeting constructs containing several hundred base pairs of targeting homolog were also developed. Finally, the ability to select for the first deletion allele using one marker, and the second using a distinct marker, was used. Together, these advances allowed for the rapid generation of deletions. These strains and approaches are widely used in the field. Dominant drug selectable markers and conditional promoter-based systems have also been developed, and the C. albicans toolbox continues to expand rapidly.

Systematic Analysis of a Homozygous Knockout Collection Reveals Multiple Insights

In the largest such academic effort to date, Noble et al. (2010) generated more than 3000 knockout strains covering nearly 700 genes. Each homozygous deletion was constructed twice independently to enhance the probability that a given genotype and phenotype pair would be linked. The mutants were systematically assayed for growth at 37°C, colony and cell morphology, and fitness in a pooled infection of mice. Prior studies had shown that nearly all mutants defective in virulence that had been described in the field also had defects in the morphology switch from yeast to hyphae (Calderone and Fonzi 2001). However, unbiased screens of the new library showed that many mutants defective in virulence had no detectable defects in morphogenetic switching or growth (Noble et al. 2010). Thus, the first insight from this unbiased approach was that there is evidently much more to C. albicans pathogenesis than morphogenetic switching.

Further analysis of mutants defective in infection, but not in growth or morphogenetic switching, led to studies that revealed the importance of glucosylceramide in infection (Noble et al. 2010). The investigators note that related yeasts, like S. cerevisiae and S. pombe, have lost the glucosylceramide synthesis genes, indicating that this glycolipid may be directly relevant for survival in the host. Additional studies of a transcription factor required for virulence, but not growth or morphology, led to discovery of a new iron regulatory network whose activation is critical for virulence and repression is important for gut commensalism (Noble et al. 2010). Genetic screens of this library in the context of gastrointestinal tract led to the identification of a specialized cell type required for C. albicans commensalism in the gut (Pande et al. 2013). We emphasize that these three insights came directly out of the systematic, unbiased, forward genetic approach enabled by the development of new genetic tools and resources for C. albicans.

Transcription Factor Knockout Collection: A Systematic Analysis

In a related study, the 143 deletion mutants of nonessential C. albicans transcription factors were constructed (Homann et al. 2009). Initial characterization of the mutants revealed most mutants affected growth under at least one condition. In addition, about a quarter of the mutants affected sensitivity to commonly used antifungal drugs, and a quarter of the mutants affected colony morphology (Homann et al. 2009). Other groups used the transcription factor library to investigate the adherence to surfaces by growing Candida cells (biofilm formation) (Finkel et al. 2012). As mentioned above, biofilms are thought to allow C. albicans cells to colonize the surfaces of medical hardware and withstand drug treatment and immune responses (Lynch and Robertson 2008). The study identified multiple factors regulating biofilms but the chromatin-remodeling factor Snf5 had one of the most pronounced phenotypes. The investigators also showed that the effects of Snf5 are largely through regulating another transcription factor, Ace2 (Finkel et al. 2012).

Collections Constructed in Industry: Insights and Challenges

To recognize possible drug targets, several large-scale studies have attempted to identify new essential or haploinsufficient genes in Candida via the systematic construction of mutants. A company constructed a library of more than 1100 mutants, which carry a deletion of one copy of a gene and the second copy is placed under a conditional, doxycycline-repressible promoter (Roemer et al. 2003). The methodology used was termed GRACE (gene replacement and conditional expression), and has the benefit of allowing the investigation of essential genes, as one copy can be kept expressed for the purposes of maintaining the strain. Using this resource, the 567 essential genes were identified and prioritized as potential drug targets. One interesting finding was the discordance of essentiality among homologous genes in C. albicans, S. cerevisiae, and S. pombe. Only 61% of the tested C. albicans genes that are homologous to S. cerevisiae essential genes showed significant reduction in growth. This resource was not made available for academic research until recently. Current use requires negotiation of a materials transfer agreement with Merck Sharp & Dohme (Whitehouse Station, NJ) for academic use. The same is true of a larger heterozygous mutant collection. The latter was successfully used to profile the effects of 35 small molecules and identify their targets (Xu et al. 2007).

Lessons Learned

As an important human fungal pathogen, C. albicans remains the focus of research for molecular mycologists worldwide. As a diploid pathogen without a complete sexual cycle, the technical challenges it poses are substantial. Nonetheless, the patient development of methods for systematic gene deletions has shown that the construction of comprehensive resources is feasible, even in an organism without traditional genetics. Moreover, insights obtained to date from unbiased functional approaches indicate that whole-genome resources, if created and made available without restriction, would have an enormous impact. Importantly, an extremely well-curated genome database (www.candidagenome.org) is in place and widely used in the field. Together, with advances in genetic tools, the field as a whole is poised to make the next move.

CONCLUDING REMARKS

This work makes it evident that the construction of whole-genome deletion collections, as well as other resources (tag libraries, overexpression libraries, etc.), could have a transformative impact on our ability to rapidly and efficiently develop insights into the mechanism of pathogenesis of human fungal pathogens. Although such efforts require substantial investment of time and resources and are not without substantial challenges, there seems little doubt that the impact on the field would be strong and sustained.