Identification of Antifungal Compounds Active against Candida albicans Using an Improved High-Throughput Caenorhabditis elegans Assay

Candida albicans, the most common human pathogenic fungus, can establish a persistent lethal infection in the intestine of the microscopic nematode Caenorhabditis elegans. The C. elegans–C. albicans infection model was previously adapted to screen for antifungal compounds. Modifications to this screen have been made to facilitate a high-throughput assay including co-inoculation of nematodes with C. albicans and instrumentation allowing precise dispensing of worms into assay wells, eliminating two labor-intensive steps. This high-throughput method was utilized to screen a library of 3,228 compounds represented by 1,948 bioactive compounds and 1,280 small molecules derived via diversity-oriented synthesis. Nineteen compounds were identified that conferred an increase in C. elegans survival, including most known antifungal compounds within the chemical library. In addition to seven clinically used antifungal compounds, twelve compounds were identified which are not primarily used as antifungal agents, including three immunosuppressive drugs. This assay also allowed the assessment of the relative minimal inhibitory concentration, the effective concentration in vivo, and the toxicity of the compound in a single assay.


Introduction
The soil dwelling nematode Caenorhabditis elegans has become a model host to study mammalian virulence of both bacterial and fungal pathogens [1,2], including Candida albicans, the most commonly isolated fungal pathogen from blood cultures [3]. Importantly, key components of Candida pathogenesis in mammals, such as biofilm and filament formation, are also involved in nematode killing [2]. Because of these features, assays have been devised that involve killing of C. elegans by fungal pathogens which can be used to screen fungal mutants for virulence factors and libraries of compounds for antifungal activity [4][5][6].
Use of a whole animal assay during antifungal screens provides several advantages when compared to more commonly used inhibitory compound screens that are preformed in vitro. First, the immunomodulatory affect of the compound on the host immune system can be assessed in the presence of a pathogen. Additionally, effective compounds that function by inhibiting pathogen virulence factors can be identified, as some of these factors are only expressed in a host. Another important advantage of using a model host during screening is that it allows simultaneous assessment of toxicity. Frequently, compounds toxic to fungi also have undesirable effects on the mammalian host. This toxicity/ efficacy ''bottleneck'' delays antifungal drug discovery.
C. elegans whole-animal screens in particular have several advantages over other mammalian screening models. The C. elegans model does not raise as many ethical concerns as the use of vertebrate models. Moreover, nematodes are easy to manipulate in the laboratory because they are small enough to fit into standard 96-and 384-well microtiter plates and they have short and simple reproductive cycles. C. elegans worms are also relatively inexpensive to propagate in large numbers and are genetically tractable [7][8][9]. The two main objectives of this paper are to first develop an automated, high-throughput, C. elegans assay that can be used to screen chemical compounds in vivo, and second use this new assay to identify chemicals with antifungal activity.

Methods
Media and strains C. albicans strain DAY185 [10], which was used in all assays, was grown overnight to late-log phase with shaking at 30uC in yeast extract peptone dextrose (YPD) (Difco) medium containing the antibiotics kanamycin (45 mg/ml), ampicillin (100 mg/ml), and streptomycin (100 mg/ml). The cells were centrifuged, washed with PBS and re-suspended at a concentration of 2.5610 5

Z9-factor values
Positive and negative control data were used to calculate the Z9-factor as a measure of overall assay quality [14]. Z9 = 12((3s c+ +3s c2 )/(|m c+ 2m c2 |)), where m c+ and m c2 are the averages of test sample signals for both the positive and negative controls respectively, and s c+ and s c2 are the variations in signal measurements for the positive and negative controls respectively. Dimethyl sulfoxide (DMSO, 2%) in the screen medium served as the negative control, and amphotericin B and ketoconazole in screen medium were used individually as positive antifungal controls. Concentrations of the antifungal agents used were effective against the C. albicans DAY185 strain used for the assay. Two columns (16 wells) in each 96-well plates were used for each concentration of the antifungal. Z9-factor values were calculated for each of the antifungal agents.

C. elegans-C. albicans assay
The pre-infection assay was performed as previously described [6]. In brief, worms grown on NGM plates were washed with M9 buffer and placed on 48 h-old C. albicans lawns (on BHI agar plates) for 2 h. The worms were washed off the plates using screen medium, transferred to two separate beakers, and re-suspended at a density of 1-2 worm/ml in screen medium. Twenty ml of the suspension of pre-infected worms were added to wells of 96-well non-binding half area plates (Corning) containing 80 ml of screen medium. The details concerning the addition of compounds to the wells are stated below in the following text. Plates were sealed with Breathe-Easy TM membranes (Diversified Biotech), and incubated at 25uC for 96 h. The survival rates of both exposed and unexposed worms were analyzed using the STATA 6 software.
The procedures for the co-inoculation assay were similar to those of the pre-infection assay above with the following modifications [15]: 70 ml of screen medium was dispensed into each well of a 96-well non-binding half area plate (Corning) using a MultiDrop Combi Plate Filler (Thermo Scientific) equipped with a standard dispensing cassette (Thermo Scientific). When compounds were used, approximately 450 nl of each compound was pin-transferred from compound plates into individual wells (containing 70 ml of screen medium) of 96-well assay plates simultaneously with a 96-pin head (CyBi H -Well, CyBio). Worms were grown and washed as in the pre-infection assay, except that they were not previously infected with C. albicans. A Union Biometrica COPAS-BIOSORT TM worm sorter was used to dispense automatically 15+/21 worms into each well, followed by addition of 10 ml of a C. albicans suspension at a density of 2.5610 5 cells/ml in PBS with a Multidrop Combi Plate Filler.

Acquisition and analysis of data for co-inoculation assay
At the end the incubation period, the plates were imaged using the Molecular Devices Discovery-1 microscope (MDS Analytical Technologies) running MetaExpress software by capturing transmitted light images. Images of the entire well were captured using a 26 low magnification objective lens. Prior to analysis, the images were converted from 16 bit TIF files to 8 bit JPEG files using the CellProfiler TM image analysis software [16,17]. The resulting images were visually analyzed for in vitro fungal growth, followed by visually scoring of live and dead worms based on nematode shape, as live worms appear sinusoidal and dead worms are rod shaped.
The determination of the lowest concentration of the selected compounds showing in vitro antifungal activity was accomplished by following the steps detailed in the co-inoculation assay, using two-fold serial dilutions of the test compounds. The wells were assessed by visually monitoring the turbidity for concentrations exhibiting in vitro inhibition of C. albicans growth. A quantitative scoring system was developed to measure and rank the activity of hits. For this, compounds of interest were tested in a series of twofold dilutions, and the half maximum effective concentrations (EC 50 ) which conferred 50% survival of the worms was determined.

Results/Discussion
Optimization of the C. albicans-C. elegans assay for highthroughput use Previous work has demonstrated that C. elegans can serve as a host for Candida spp. and an infection is established in the worm intestine when nematodes are fed on a lawn of C. albicans on agar medium [6]. Using this pathosystem, a C. elegans-C. albicans ''curing'' assay was developed using liquid media in standard 96well plates which allowed simultaneous assessment of both the compound's antifungal activity and the compound's cellular toxicity. This relatively low-throughput assay entailed preinfection of the nematodes, by placing them on a C. albicans lawn on BHI agar plates prior to the assay, followed by manually pipetting the infected worms into microtiter plates containing the compounds to be screened.
One of the goals of the current work was to increase the throughput of the chemical screening assay by utilizing a commercial instrument to distribute a precise number of infected worms in each microtiter well. However, the use of the instrument to dispense Candida-infected nematodes was not practical, as the fungus formed avid biofilms on the surface of the instrument such that it could not be efficiently decontaminated after each use. To circumvent this problem, direct C. albicans infection of the worms by co-inoculation in assay wells containing the liquid medium was assessed. The survival of worms pre-exposed to C. albicans lawns on agar before transferring to the assay medium was compared to worm survival when the nematode and C. albicans were coinoculated in the assay medium. The rate of worm survival was comparable in the co-inoculation and pre-infection assays ( Figure 1), demonstrating that prior exposure of the nematodes to C. albicans on solid media is not necessary for killing of the nematodes. The killing of worms may commence sooner when worms are pre-infected for two hours before transferring to assay wells, but by four days there was no significant difference between the two treatments, as all of the worms were dead in both instances. Additionally, there was no difference in the frequency and amount of filamention by C. albicans in dead worms in either assay (data not shown).
Several concentrations of fungal inoculum (10 2 to 10 5 cells/well) were examined, and a concentration 2.5610 3 cells/well was optimal for the 96 hour assay period, which resulted in a killing rate comparable to the pre-infection assay (data not shown).
Higher concentrations were equally effective in killing, but resulted in wells with high turbidity that required a washing step prior to imaging in order to score the assay results. Fifteen worms per well of the 96-well half area plate was optimal for the assay, as wells with more than 15 worms per well caused the worms to overlap in the confined area making it more difficult to score whether the nematodes were alive or dead based on shape.
In the C. elegans-C. albicans antifungal discovery assay detailed in Breger et al (2007) [6], preinfection of the worms requires not only additional time to allow ingestion and subsequent colonization of the nematode intestinal tract, but also additional resources such as premade plates with lawns of C. albicans for the infection and materials to recollect the worms post-inoculation. This improved assay also cuts time by dispensing a precise number of worms by instrumentation instead of manual pipetting. As worms are manually pipetted they begin to settle at the bottom of the tube where they are held, creating a ''gradient'' of various worm densities where the top portion of the liquid contains a lower density of worms when compared to the bottom of the tube. This gradient leads to various amounts for worms being pipetted into each well, and thus creates situations where the assay for a compound may need to be repeated due to either too few or too many nematodes within each well. Therefore this improved assay not only saves several hours when setting up the assay, but may save several weeks as some compounds would not have to be repeated by additional screening.

Evaluation of the co-inoculation assay
In order to evaluate the reproducibility and robustness of the coinfection assay the Z9-factor was determined. This calculation takes into account signals from both positive and background samples and is used to assess the probability that a sample scored as positive in a screen is genuine and not due to background noise [14]. For this evaluation, the percentage of worm survival was used in calculations after the following treatments: dimethyl sulfoxide (DMSO) was used as a negative control (background), and amphotericin B (2.5 mg/ml) and ketoconazole (16.0 mg/ml), (which conferred 93-100% worm survival) were used as the positive controls. In three independent experiments during assay development, the Z9 factor values were reproducible for each of the compound concentrations, ranging from 0.69 to 0.72, with an average value of 0.71. Similar to the pre-infection assay, wells that contained no antifungal, compounds without antifungal efficacy, or antifungal compounds at a concentration below the effective dose result in rod-shaped dead worms with fungal filaments protruding from their cuticles (Figures 2A-2D). Interestingly, growth of C. albicans within the liquid medium was not necessary for nematode death, as filaments of C. albicans were seen extruding from some dead nematodes despite no in vitro growth, suggesting the compound inhibited the fungus in vitro but not in vivo. This observation was made in control assay wells, as treatment with 1 mg/ml of amphotericin B exhibited antifungal activity in vitro, but not in vivo ( Figure 2C) as well as compounds identified in the screen ( Figure 2D). However, in wells with effective antifungal agents at inhibitory concentrations, the worms maintain their sinusoidal shape, and the media remained free of fungal growth. This observation was made in both control assay wells with amphotericin B and well containing screened antifungal compounds exhibiting both in vitro and in vivo activities ( Figure 2E and 2F).
During the preliminary screen of the compound library, wells that displayed a reduction or no in vitro fungal growth, however had little or no nematode survival, were counted and potential antifungal compounds. Upon confirmatory experiments, the compounds were able to confer nematode survival at higher concentrations than initially tested in the preliminary screening process. Therefore, all identified compounds conferred in vitro and in vivo antifungal activity. There were no compounds identified in the screen which conferred C. elegans survival in the presence of in vitro fungal growth, an observation which would suggest that the compound may have 1) an immunomodulatory effect on the nematode immune response, and/or 2) properties which inhibit pathogen specific virulence factor(s). Despite not finding any compounds which inhibit C. albicans in vivo but not in vitro, we feel it is possible to do so using this assay as the concentration of fungal cells initially inoculated to the wells still enables the visualization of the nematodes at the conclusion of the screen.
Of note is that in previous studies the scoring of live/dead and worms with and without filaments was accomplished by direct microscopic observation of movement and morphology of the nematodes. In these assays, digital images from the Discovery-1 microscope were captured and visually inspected for the analysis of assay results. This method improved the accuracy of analysis and evaluation of the assay, and moved the process closer to an automated ''screening by imaging'' system. Recently, SYTOX Orange live/dead staining has been utilized to facilitate highthroughput scoring of C. elegans survival [18], although the same approach may not necessarily be feasible for C. albicans due to background resulting from staining of live hyphal filaments extending from the dead nematode. Alternatively, imaging programs that can differentiate between nematode shapes can also possibly be developed to facilitate scoring of nematode survival at the conclusion of the screen [16,17,19].

Identification of antifungal compounds resulting from the screen
After establishing the robustness and repeatability of the assay, a screen of 3,228 compounds was conducted representing compounds not screened in previous studies. This study, in combination with the previous screen by Breger et al (2007), brings the total number of screened compounds to 4,494. This collection was represented by 1,948 bioactive compounds consisting of FDA approved drugs and other compounds from the Microscorce Spectrum collection (Microsource Discovery Systems, Inc.) and the Prestwick compound library (Prestwick Chemicals, Inc.) and 1,280 small molecules derived via diversity- oriented synthesis [20,21] from the Broad Institute of Harvard and MIT (Cambridge, MA). As shown in Table 1, the screening library represented a diverse collection of bioactive compounds composed of multiple chemical classes containing a broad array of functionality. In total, 19 compounds were identified in the screen (Figure 3). More importantly, the C. elegans screen was able to detect most of the well established antifungal agents (Table 1 and  Table 2).
Of the ten antimycotic compounds within the chemical library, all but three (flutrimazole, thiabendazole, and griseofulvin) were identified in the preliminary screen. All three of these compounds are used in topical treatment of fungal infections, and may have limited solubility in water, complicating their identification using the C. elegans-C. albicans ''curing'' assay. Furthermore, griseofulvin has no antifungal activity against Candida spp., as the fungus does not uptake the compound into the fungal cell, and therefore it would not be expected to be identified [22]. The compound thiabendazole is used agriculturally as a fungicide and has been used clinically in treatment of Aspergillus flavus [23,24], however it is also nematocidial [25]. Several other antinematode compounds were represented in the chemical library including the clinically relevant agent albendazole and the avermectin, abamectin. These compounds exemplify three scenarios where a potential antifungal agent would be overlooked in the C. elegans-C. albicans antifungal discovery assay; 1) if the compound has limited solubility in water, 2) if the assayed compound has antifungal activity, as well as nematicidal activity, and 3) if the compound has activity against most fungi, although the fungus used in the screen (C. albicans) maybe unaffected by the compound.
A number of known immuno-modulating agents were found to have antifungal activity in the C. elegans killing assay. This group includes FK-506, ascomycin, and cyclosporin A, all of which have previously been reported to have antifungal activity not only for C. albicans, but also other fungal pathogens such as Cryptococcus neoformans, Aspergillus fumigatus, and members of the zygomycetes [26][27][28][29][30][31][32]. These immunosupressives demonstrated weak antifungal activity when compared to clinically relevant antifungal agents, however they were able to confer ,50% nematode survival at the highest concentration tested ( Figure 4A). The target of these therapeutic drugs, calcineurin, is essential for virulence of C. albicans and is involved in a diverse array of other pathogenicity factors, including the ability to survive in serum and resistance to antifungal azole compounds [27][28][29]. The antifungal activity of these immunosuppressives, coupled with their ability to impede azole resistance, create (at least in vitro) a synergistic effect when used with clinically relevant antifungal agents [33,34]. This synergism has been observed for many fungal pathogens, including azole resistant C. albicans isolates [31][32][33]35]. Importantly, immunotherapy with calcineurin inhibitors does not select for C. albicans strains which are resistant to these compounds [36].
Among the hits was the potential promising compound dequalinium chloride which is a potent anti-tumor and protein kinase C (PKC) inhibitor ( Figure 4B). In Saccharomyces cerevisiae, the Rho 1-PKC pathway (the cell integrity pathway) functions in cell wall biosynthesis and mating [5,37,38]; cell wall integrity is involved in C. elegans killing by fungi [5]. Another compound, triadimefon, also conferred a similar degree of nematode survival ( Figure 4B). The fungicide triadimefon has no affect on DNA, RNA, or protein synthesis, and instead derives it antifungal activity by inhibiting ergosterol biosynthesis [39]. The potent H + V-ATPase inhibitor concanamycin A was also identified, although it had been previously described as having antifungal properties [40]. This compound demonstrated the highest antifungal activity outside of the clinically used antifungal agents in the screen ( Table 2 and Figure 4B). However, concanamycin A is highly toxic, limiting its use clinically. Mice injected with 1 mg per kilogram body mass interperitoneally die [40], however no toxicity was observed at the highest concentration tested (1.4 mg/mL).

Determination of MIC and EC 50 of compounds
Dose response curves were determined to calculate the minimal inhibitory concentration (MIC) and the effective dose that resulted in 50% survival of the nematodes (EC 50 ) for 18 compounds identified in the screen that prevented growth of C. albicans in vitro ( Table 2). The EC 50 for most of the compounds was usually twice the MIC concentration, with a few exceptions. In this assay the Table 1. Compounds exhibiting antifungal activity identified in the C. elegans-C. albicans infection assay.

Compound
Action/Use Reference(s)

Triadimefon
Affects gibberellin and sterol biosynthesis [39] Dequalinium chloride Potent anti-tumor activity and protein kinase C alpha (PKC) inhibitor [58] Valinomycin Apoptosis inducer by disrupting mitochondrial membrane potential; affects morphology of C. albicans [59][60][61] Phenylmercuric acetate Used as fungicide, herbicide, slimicide and bacteriocide; metabolized to diphenylmercury [62] Malachite green carbinol base Malachite green is used to treat parasites, fungal, and bacterial infections in fish and fish eggs; metabolized to the carbinol form and then to the toxic leucomalachite green form. [42,63] Concanamycin A Potent and specific inhibitor of vacuolar-ATPase [40] Ellipticine Potent anti-tumor agent, acts by DNA intercalation and/or inhibition of topoisomerase II [64] Mepartricin Used to treat chronic pelvic pain syndrome, reduces estrogen plasmatic levels and their concentration in the prostrate. Antifungal activity against Candida infections.
[ 65,66] Suloctidil Used as a peripheral vasodilator, formerly in treating peripheral and cerebral vascular disorders; causes hepatotoxicity. [67] doi:10.1371/journal.pone.0007025.t001 MIC values may fluctuate from previously published reports, possibly due to differences in the liquid media used for each of the assays and in particular the time at which the cultures are monitored (two days in the established Clinical and Laboratory Standards Institute (CLSI) protocol for MIC determination versus five days in this assay). However, the standard antifungal used for optimization of the assay amphothericin B had an MIC value of 1.0 mg/mL, which is within the published range of 0.5-2.0 mg/mL [41]. Previously, three compounds were identified which prolonged nematode survival when infected with C. albicans [6]. Two of these compounds, caffeic acid phenethyl ester (CAPE) and enoxacin, also displayed antifungal activity when carried over into a murine model. However, of these two compounds only CAPE was able to inhibit C. albicans growth in vitro. The MIC of CAPE was 64 mg/ mL, well above the 4 mg/mL concentration which was required to observe an increase in C. elegans survival [6], suggesting the compound maybe involved in a role besides inhibition of fungal growth, and thus conferring the observed increase in C. elegans survival. A similar trend was observed in a C. elegans-Enterococcus faecalis antibacterial drug discovery screen, where six classes of compounds displayed effective concentrations lower than the MIC [18]. Therefore, comparison to the MIC and the EC 50 in the C. elegans assay may be used to identify compounds that have higher efficacy during infection than in vitro, as such compounds may affect virulence factors or modulate host immunity. Compounds that may fall into this class are often not identified in a traditional antifungal compound screen as the screen may rely solely on inhibiting microbial growth. Although the concentration of the compounds inside the nematode is probably lower than in the media, the ratio between the MIC and the EC 50 provides a way to identify compounds that most likely have a higher efficacy in vivo than in vitro.

Evaluation of compound toxicity
This screen also allowed simultaneous assessment of the compound's toxicity, as a decrease in nematode survival at higher compound concentrations despite no evidence of fungal growth in vitro or filamentation in vivo was an indication that the compound was toxic to the worm. In addition to the MIC and EC 50 values, the dose-response experiments also detected toxic compounds such as malachite green carbinol base which is the carbinol form of malachite green, usually metabolized to the toxic leucomalachite green form [42]. In the initial portion of the curve, there was no worm survival up to a concentration of 0.95 mg/ml, and there were visible C. albicans filaments on dead worms. At concentrations greater than 0.95 mg/mL up to the EC 50 concentration (7.62 mg/ mL), worm survival increased, however at concentrations higher than the EC 50 there was no survival of any worms ( Figure 5). This was probably due to the toxic effect of the compound since in vitro growth of C. albicans in the wells was absent, and in vivo activity of the compound within the worms did not permit formation of C. albicans filaments penetrating from the cuticle into the surrounding medium.

Concluding remarks
A high-throughput whole animal assay for the identification of compounds with antifungal efficacy has been developed using C. elegans as a heterologous host. The major developments include the use of the worm-dispensing instrument, the use of a standardized fungal inoculum, and the advance of using a co-inoculation procedure of Candida and C. elegans to the wells, thereby eliminating the labor-intensive infection. This facile in vivo model can now be used in high-throughput assays to evaluate large libraries of chemical compounds and avoid some of the main obstacles in current antifungal discovery, including host toxicity screening, costs, and time. The overall excess cost attributable to candidemia is estimated at $1 billion per year [43], while the average cost of treating a single episode of candidemia is $34,123-44,536 (1997 U.S. dollars) [44]. Further complicating issues in treatment of candidemia is the emergence of drug resistant isolates. While most attention has been focused on antibacterial drug resistance, a similar phenomenon has occurred with fungi due to long term use of antifungal agents, particularly in patients who require these medications for the duration of their lives (eg. HIV positive patients, organ transplant recipients). Drug resistant C. albicans isolates have been isolated from patients which are resistant to the most recent class of antifungal compounds developed, the echinocandins [45], further highlighting the need for antifungal drug development [46]. The discovery of new antifungal agents may be a daunting but certainly crucial task.