Hosting Infection: Experimental Models to Assay Candida Virulence

Although normally commensals in humans, Candida albicans, Candida tropicalis, Candida parapsilosis, Candida glabrata, and Candida krusei are capable of causing opportunistic infections in individuals with altered physiological and/or immunological responses. These fungal species are linked with a variety of infections, including oral, vaginal, gastrointestinal, and systemic infections, with C. albicans the major cause of infection. To assess the ability of different Candida species and strains to cause infection and disease requires the use of experimental infection models. This paper discusses the mucosal and systemic models of infection available to assay Candida virulence and gives examples of some of the knowledge that has been gained to date from these models.


Candida and Man
1.1. Carriage of Candida Species. In healthy individuals Candida species are harmless members of the normal gastrointestinal (GI), oral, and vaginal microbial flora. It is assumed that everyone carries Candida in their GI tract (reviewed in [1]), with C. albicans the species most frequently identified in faecal sampling, representing 40-70% of isolates [2][3][4]. Other isolates are usually identified as C. parapsilosis, C. glabrata, C. tropicalis, or C. krusei [2][3][4].
In comparison to GI carriage, oral carriage is observed in only ∼40% of healthy individuals, with considerable variation found between studies (reviewed in [1]). Higher carriage levels are generally associated with diabetes, cancer, HIV, or denture use (reviewed in [1]). Again, the majority of isolates (∼80%) are identified as C. albicans, with C. glabrata or C. parapsilosis making up the remainder [5][6][7][8][9].
Vaginal carriage occurs in an even smaller proportion of the healthy population, with only ∼20% of healthy women found to have vaginal Candida carriage [10][11][12][13]. C. albicans is again the most commonly identified species, with C. glabrata the only other species usually found [10,12,[14][15][16][17]. Therefore, C. albicans is the major species found as a commensal in healthy individuals, with four other species, C. tropicalis, C. parapsilosis, C. glabrata, and C. krusei, also found.

1.2.
Candida and Disease. Candida species, however, have an alternative lifestyle, causing opportunistic infection in hosts with altered physiological or immune response. The infections caused by Candida species range from self-limiting, superficial mucosal lesions (commonly referred to as thrush), chronic and/or recurrent mucosal, skin, and nail infections, through to life-threatening invasive or disseminated infection [1,[18][19][20][21].
In humans, the most common infections caused by Candida species are superficial infections of the mucosa, skin, and nails [20][21][22][23][24]. Pseudomembranous oral thrush is common in babies and in the elderly, but is also found in HIVpositive individuals and cancer patients (reviewed in [1,25]). Denture stomatitis is also a significant infection, occurring in approximately 60% of denture wearers [26,27]. In oral candidiasis most infections are caused by C. albicans (58%),

Experimental Models of Candida Infection
Experimental infection models allow disease development to be followed from the moment that fungal cells are introduced into the host. To be a good model, a model should be reproducible, relatively easy to set up, and should reproduce the major clinical symptoms seen in the human disease. It is also an added advantage if the model is cost effective. Models which satisfy these conditions allow further in-depth investigation of Candida virulence to be carried out and, subsequently, allow inferences about Candida virulence in human disease to be made.
Although a great deal of preliminary research on virulence can be carried out by laboratory experiment, infection modelling requires the involvement of a host organism. It is only in a whole organism that the complex host-fungus interactions that determine whether or not disease will occur can be investigated. Although larger animals have been used to study Candida infections, for example, macaques [57,58], piglets [59], rabbits [60][61][62], and guinea pigs [63,64], the majority of Candida virulence studies use rodent infection models. This is due to economic factors, ease of handling, and the availability of genetically modified mouse strains, which allow human genetic conditions to be mimicked.
In this paper, experimental animal models that have been developed for Candida virulence assays are discussed. It should be noted that the majority of models focus on C. albicans as this is the major species associated with human Candida infections.

Mucosal Infection Models.
To model Candida oral and vaginal infections, mucosal models have been developed mainly in rats and mice. The procedures used in rats and mice are generally similar. However, the larger animal has the added advantage that denture-associated fungal biofilms formation can also be studied in a host [65]. Establishment of infection at mucosal sites generally requires treatment with immunosuppressive agents, oestrogen, or antibiotics prior to infection, or the use of germ-free animals [66][67][68]. However, the nude (Foxn1 nu ) mouse model of oral infection allows infection to be established without any immunosuppression or other pretreatment [69]. Greater detail can be found in more extensive reviews of these infection models [67,68,70,71].
In order to assess virulence in mice using the oral infection model, mice are routinely pretreated with corticosteroids and Candida cells are administered into the oral cavity of anaesthetised animals either by applying a Candidasoaked cotton bud under the tongue or by applying the inoculum directly onto the teeth, gums, and oral cavity [67,70,72]. Virulence in this model is usually determined by fungal organ burden and histopathology.
Both rat and mouse models have been used to compare the virulence of C. albicans mutant strains and also clinical isolates [73][74][75][76][77]. Using these models, C. albicans mutant strains which are unable to switch between the yeast and hyphal growth forms were found to be unable to cause oral infection, demonstrating a requirement for yeast-hypha switching in oral infection [75]. In addition, protein kinase Ck2 was also shown to be required for oropharyngeal C. albicans infections [77].
Mouse and rat models have also been developed to assay Candida virulence in vaginal infection. In these models the rodents are maintained in oestrus in order to maintain colonisation and infection, which probably mimics pregnancy-associated candidiasis [78][79][80][81]. In rats, this generally involves surgery to remove the ovaries, with subsequent administration of oestrogen [81]. Recently, however, a new rat model has been developed, similar to the mouse model, where oestrus is maintained merely through administration of oestrogen [82], which will increase the ease of setting up the infection model. Immunosuppression of the host can also prolong colonisation by Candida species [83]. These models allow us to examine single vaginitis episodes; however, a satisfactory model of recurrent, chronic vaginitis is not yet available.
The virulence of C. albicans clinical isolates has been compared in rodent vaginitis models, demonstrating that isolates have varying capacities to cause disease [84,85]. This model has also been used to assess virulence of genetically modified C. albicans mutants [85][86][87].
In addition to assessing C. albicans virulence, this model can be used to examine virulence of other Candida species. As C. glabrata is also associated with human vaginal infection, researchers have used the rat vaginitis model to evaluate the virulence of a C. glabrata petite mutant, discovering than the mutant was more virulent that the parental strain [88].
In addition, C. parapsilosis isolates have also been assessed for their ability to cause vaginal infection in the rat model [80]. In this study only a single isolate, recently obtained from a woman with active vaginal infection, was capable of initiating infection [80].
A major development in Candida virulence testing at mucosal surfaces occurred recently with the development of a concurrent oral and vaginal infection model by Rahman et al. [72]. This mouse model allows both oral and vaginal infections to be initiated in the same host, greatly reducing the numbers of animals required for these virulence assays. A comparison of the virulence of three different C. albicans isolates in this model clearly demonstrated that C. albicans isolates were not equally virulent, with obvious differences in their ability to initiate mucosal infections [72].

Invasive Infection Models.
Mouse models of invasive fungal infection have been the most popular methods to assess Candida virulence up until the present day, although assays have also been carried out in rabbits, guinea pigs, and rats also used in some studies. There are two major models of Candida invasive infection, the intravenous (IV) challenge model and the gastrointestinal (GI) colonisation with subsequent dissemination model. These models were recently reviewed [89].

Intravenous Challenge Model.
The mouse IV challenge model has been used to study Candida virulence since the 1960s and is both well characterised and reproducible [90][91][92]. Candida cells are injected directly into the lateral tail vein, bypassing any requirement of the fungus to cross epithelial and endothelial barriers to gain entry into the bloodstream. In this mouse model, which is similar to human invasive infection occurring with catheter involvement, fungal cells are found in all organs, but disease progresses only in the kidneys and brain, which depends upon inoculum level and mouse strain [91][92][93]. Sepsis develops as invasive disease progresses, which eventually leads to the death of the mouse [92,94,95].
Using immunocompetent mice, the IV challenge model has been used to compare the virulence of different Candida species [97-99, 107, 111-114]. C. albicans is clearly the most virulent species [97,98,111,112,114], followed closely by C. tropicalis [97,98,111,112,114]. In contrast, C. krusei and C. parapsilosis were unable to kill the infected animals, even at high inoculum levels, and fungi were eventually cleared from the host [98,111,114].
In immunosuppressed mice, C. tropicalis showed greater virulence, with disease progressing in the kidneys, rather than infection being controlled which occurs in immunocompetent mice [96,98,99,107,115]. C. parapsilosis and C. krusei remained unable to initiate progressive infections, even with addition of immunosuppressive treatments [98,107], although administration of a very high inoculum potentially allows some C. parapsilosis isolates to initiate disease [108,110].
Within each Candida species, clinical isolates were found to show considerable virulence differences in the IV challenge model. This was true for C. albicans [97,107,116,117], C. tropicalis [97,99,112,115,118], and C. parapsilosis [108,119], with some isolates unable to initiate invasive infections. This raises questions as to whether virulence results found for a single strain or isolate are representative of the entire species. This could be of particular importance for C. albicans studies where the vast majority of gene disruption studies have been carried out in a single strain, SC5314, background.
Numerous studies have evaluated C. tropicalis clinical isolate virulence differences; however, there are very few studies published on the virulence of genetically modified C. tropicalis strains. One study which has been published was able to demonstrate that a secreted acid protease was required for full virulence of C. tropicalis in immunocompetent mice [120]. In contrast to C. tropicalis, vast numbers of studies have been published on the virulence of C. albicans mutants, with over 200 genes identified as contributing to the C. albicans virulence in this model (reviewed in [89]).
C. glabrata behaves very differently from the other Candida species in the mouse model of invasive infection. Although C. glabrata is maintained, or tolerated, at high levels in the kidneys of immunocompetent mice, the mice did not die and there was little inflammation associated with the fungal cells [113,114]. Immunosuppression appears to increase virulence of C. glabrata in terms of higher fungal organ burdens, but mouse survival is only increased in some C. glabrata infections [100,[103][104][105][106]. However, because immunosuppression may allow invasive disease to develop in C. glabrata-infected mice, these treatments have been added to an infection model used in some studies to compare the virulence of genetically modified C. glabrata, with fungal burdens used as the virulence estimate [88,101,102,105]. The immunosuppressed mouse infection model has demonstrated the importance of hypertonic stress responses, the cell wall integrity pathway, and nitrogen starvation responses in C. glabrata virulence [103,104,106]. In addition, this model has identified a petite mutant, strains expressing hyperactive alleles of the transcription factor gene PDR1 and the ace2 null mutant as being more virulent than their parent strains [88,105,121]. However, it should be noted that the hypervirulent phenotype of the C. glabrata ace2 null was completely lost in immunocompetent mice [122]. In other virulence experiments in immunocompetent mice, where virulence was determined from fungal organ burdens at day 7 after infection, researchers were able to demonstrate that the cell wall integrity pathway [123,124] and oxidative stress response [125], as well as the transcription factor Pdr1p and some of the genes that it regulates [101,121], contribute to C. glabrata virulence.

Gastrointestinal Colonisation and Dissemination Model.
Gastrointestinal models can either be set up in neonatal or adult mice. Intragastric infection of neonatal mice leads to persistent colonisation, without any requirement for pretreatment of the mice. However, to obtain colonisation of adult mice, the natural mouse gastrointestinal flora must first be removed by treatment with broad spectrum antibiotics. Adult mice can either be infected by gavage (intragastrically) or orally via their chow or drinking water. Subsequent treatment of Candida colonised mice with immunosuppressants and/or drugs which damage the gut wall allow fungal dissemination to occur (reviewed in [70,126]).
In the gastrointestinal models fungal colonisation is highest in the stomach, caecum, and small intestine [107,[127][128][129], reflecting some of the clinical findings seen in human invasive infection. During the model, persistent colonisation is routinely monitored by noninvasive faecal fungal counts, and after dissemination Candida cells can be cultured from the liver, kidneys, and spleen [128][129][130]. However, differences may be seen between mouse strains [131].
This murine model is believed to be a more accurate reflection of the events occurring in the human patient, with broad spectrum antibiotics allowing fungal overgrowth and later invasive therapies causing mucosal damage. Mucosal damage then allows Candida to enter the bloodstream and disseminate to the internal organs. In the mouse, similar to human patients, there is increased animal-to-animal variation compared to the intravenous challenge model, requiring higher numbers of animals per group to obtain statistically significant results [128][129][130].
Comparison of Candida species virulence in this model demonstrated that C. parapsilosis had lower virulence compared to C. albicans and C. tropicalis, as there was little evidence of dissemination from the gut [107,132]. However, C. parapsilosis was successful in establishing persistent colonisation of the GI tract [107]. In separate studies, C. tropicalis appeared to be more virulent than C. albicans in the gastrointestinal model, with greater dissemination to the internal organs [133,134] and higher mortality rates [97,134]. However, given the levels of variation observed in other models for the virulence of strains of different Candida species, further isolates will require to be assayed before a definitive conclusion on the relative virulence of the two species can be made.
To date, only a limited number of C. albicans mutant strains have been tested in the gastrointestinal colonisation and dissemination infection model, with only 6 mutants identified so far as contributing to virulence [89,135]. However, this model has demonstrated that a constitutively filamentous C. albicans mutant was unable to disseminate, suggesting that the ability to switch between morphological forms may be more important for dissemination [136].
C. glabrata also behaved differently from the other four major Candida species in this model, being unable to colonise the oesophageal tissue in the neonatal mouse gastrointestinal colonisation and dissemination model [137]. Again, there was little host inflammatory response to C. glabrata [137], suggesting that C. glabrata virulence mechanisms may be quite different from those of the other species studied.

Beyond the Genome: Challenges of Candida Virulence Testing in the Postgenomic Era
The genome sequences of C. albicans, C. glabrata, C. tropicalis, and C. parapsilosis are now available [138,139], encouraging the creation of large-scale mutant libraries. The challenge comes, however, when these large libraries are to be screened for genes involved in fungal virulence, with logistical, financial, and ethical issues to be considered. In library screening programmes carried out to date different virulence testing strategies have been taken. Noble et al. [140] used signature-tagged mutagenesis to allow pools of mutants to be assayed in small numbers of animals, significantly reducing the animal numbers required for testing. By contrast, in order to screen a library of 177 C. albicans strains for altered virulence, Becker et al. [141] assayed each strain in 15 mice. From these two examples it is clear that traditional testing methods can lead to large numbers of mice being required to assay virulence. However, researchers have recently begun to address the issues of virulence testing large numbers of Candida strains by developing a range of minihosts, which are mainly based on invertebrate hosts.
Minihosts may not initially appear relevant to the human disease, but these hosts do possess an innate immune system and this is known to be critical in the development of Candida infections [142]. However, many of the minihosts do not possess an adaptive immune system, which may limit their usefulness. In addition, the majority of invertebrate models have the disadvantage that they must be kept at temperatures below normal human body temperature, with the exception of Galleria which can be incubated at 37 • C. Potentially, incubation at lower temperatures may induce physiological changes in the fungus, affecting host-fungus interactions during disease development.

Wax Moth and Silk Worm Larval
Models. The first minihost model developed for Candida virulence testing was International Journal of Microbiology 5 the Galleria mellonella (wax moth) larval model [143]. In this model fungi are injected into larvae, via a proleg, and survival is monitored over a short time period. The model is relatively cheap and has the added advantage that large numbers of larvae can be infected with each mutant strain, increasing the statistical power of the assay. The Galleria model has been successfully used to model C. albicans virulence, with results roughly similar to those found in mouse infection models [143][144][145][146]. A similar model has also been developed using the silk worm (Bombyx mori) [147,148]. Both C. albicans and C. tropicalis are capable of killing silk worm larvae within two days [148], and C. albicans virulence differences were shown to correlate with results previously found in a mouse model [147].

Drosophila melanogaster.
The fruit fly, Drosophila melanogaster, has also been used to assay Candida virulence [149][150][151][152]. The susceptibility of wild-type D. melanogaster continues to be debated; however, both Toll-and Spätzle-deficient fruit flies are susceptible to infection by Candida species when fungi are injected directly into the thorax [149][150][151]. Again, D. melanogaster models also have the advantage that large numbers of flies (>30 flies) can be infected with each Candida strain, increasing the statistical power of the assay.
In fruit flies, C. albicans was shown to be more virulent than other Candida species, confirming the results found in mammalian models (see above; [149]). In addition, virulence results for C. albicans clinical isolates and mutants were broadly similar to those found in the mouse systemic model [149][150][151]. However, differences do occur. In the fruit fly, CO 2 sensing is important for virulence, but this was not the case in the mouse IV challenge model [153]. This model has already been successfully used to screen a C. albicans mutant library, identifying Cas5, a transcription factor involved in cell wall integrity, as being required for full virulence [154].
In addition to the systemic D. melanogaster infection model, a new gastrointestinal infection model has also been developed recently, which should provide new options for virulence screening in a gastrointestinal model [152].

Caenorhabditis elegans.
In addition to fly and larval models, the nematode Caenorhabditis elegans has also been evaluated as an infection model for assaying Candida virulence [155]. This model is particularly suited to highthroughput screening, as the Candida cells are fed to the nematodes in their food and assays are carried out in multiwell plates. This model has also been used successfully to screen a C. albicans transcription factor mutant library, allowing identification of transcription factor genes involved in hypha formation [155].

A Vertebrate Minihost: Zebrafish (Danio rerio).
Zebrafish are the first vertebrate minihost model developed for virulence testing of Candida. This organism has the added advantage of having both innate and adaptive immune systems [156], and methods are also available to allow fish gene expression to be manipulated to mimic human genetic conditions [157].
The first virulence assay developed in zebrafish involved intraperitoneal injection of C. albicans into 7-month-old zebrafish [158]. In this model, similar to mouse models, progressive infection depends upon dose and is associated with increased proinflammatory gene expression. This model also allows increased group sizes, with group sizes of 20 fish being used to date. Using this model, researchers demonstrated that a clinical isolate with reduced virulence in a mouse model also showed reduced virulence in this model [158]. In addition, a C. albicans mutant (efg1/cph1) known to have attenuated virulence due to filamentation defects also had reduced virulence in this model [159,160]. Of greater interest was the finding that, although these mutants were unable to form filaments in vitro, they clearly formed filaments when growing within fish. This model also allows interactions between zebrafish immune cells and Candida cells to be imaged, which will be made even easier in the future with the development of the new transparent adult (casper) zebrafish [161].
A second zebrafish infection model has also been described, where each fish larva (36 h after fertilization) is infected directly into the hindbrain ventricle with approximately 10 fungal cells [162]. In this model the C. albicans efg1/cph1 mutant again demonstrated attenuated virulence, similar to results found in the mouse IV challenge model [162].
There are, however, disadvantages to the zebrafish infection models. One of the major drawbacks of this model, in common with the majority of other minihosts, is that the fish need to be kept at 28-29 • C, which does not allow accurate mimicking of human infection.

Assaying Virulence in Experimental Models: Final Considerations
There are some important points to remember when evaluating Candida virulence in experimental infections. The first concerns the Candida species of interest. Although C. albicans, C. tropicalis, C. glabrata, and C. krusei are all associated with human carriage and infection, they are not natural mouse commensals or pathogens [163]. As such, there may be different interactions occurring between the fungus and the two different host species. This is of particular relevance when considering C. glabrata and its inability to initiate disseminated infection in the IV challenge models, especially when we know that C. glabrata can cause lethal infections in severely ill humans [164]. The second point to consider is that, although the immune systems of mice and men are similar, there are differences that could affect how the host and fungus interact [165][166][167][168]. Of particular relevance to Candida infections are differences in proportions of neutrophils and lymphocytes in the blood, complement receptor expression, and T-cell differentiation, to name but a few (reviewed in [168]). In addition, different mouse strains show differing susceptibility to infection, which could potentially alter virulence results [93,[169][170][171][172].
The third point to consider is which model should be used to evaluate Candida virulence. Some C. albicans isolates exhibit virulence differences depending upon the infection model being used [72,134,173]. A good example is the C. albicans genome sequenced strain SC5314. In the IV challenge model, SC5314 is one of the most virulent C. albicans isolates, causing lethal infection in a relatively short time [92,116]; however, in a vaginal infection model, SC5314 is a very poor coloniser of the vaginal mucosa [72]. In addition, a nongerminative C. albicans strain [173] and a ura3 minus C. albicans strain [174], both of which were attenuated in systemic infection models [173,[175][176][177], successfully established mucosal infections [173,174].
Only careful consideration of the above points will allow the Candida researcher to select the appropriate experimental Candida infection model to answer a particular research question. These models remain essential for increasing our understanding of fungal pathogenesis since both fungal attributes and host responses are known to contribute to the development of clinical disease.