Reverse Genetics in Predicts ARF Cycling Is Essential for Drug Resistance and Virulence

Summary

Candida albicans, the major fungal pathogen of humans, causes life-threatening infections in immunocompromised individuals. Due to limited available therapy options, this can frequently lead to therapy failure and emergence of drug resistance. To improve current treatment strategies, we have combined comprehensive chemical-genomic screening in Saccharomyces cerevisiae and validation in C. albicans with the goal of identifying compounds that can couple with the fungistatic drug fluconazole to make it fungicidal. Among the genes identified in the yeast screen, we found that only AGE3, which codes for an ADP-ribosylation factor GTPase activating effector protein, abrogates fluconazole tolerance in C. albicans. The age3 mutant was more sensitive to other sterols and cell wall inhibitors, including caspofungin. The deletion of AGE3 in drug resistant clinical isolates and in constitutively active calcineurin signaling mutants restored fluconazole sensitivity. We confirmed chemically the AGE3-dependent drug sensitivity by showing a potent fungicidal synergy between fluconazole and brefeldin A (an inhibitor of the guanine nucleotide exchange factor for ADP ribosylation factors) in wild type C. albicans as well as in drug resistant clinical isolates. Addition of calcineurin inhibitors to the fluconazole/brefeldin A combination only initially improved pathogen killing. Brefeldin A synergized with different drugs in non-albicans Candida species as well as Aspergillus fumigatus. Microarray studies showed that core transcriptional responses to two different drug classes are not significantly altered in age3 mutants. The therapeutic potential of inhibiting ARF activities was demonstrated by in vivo studies that showed age3 mutants are avirulent in wild type mice, attenuated in virulence in immunocompromised mice and that fluconazole treatment was significantly more efficacious when ARF signaling was genetically compromised. This work describes a new, widely conserved, broad-spectrum mechanism involved in fungal drug resistance and virulence and offers a potential route for single or improved combination therapies.

Introduction

Invasive fungal infections pose a serious health risk to hospitalized patients worldwide. Particularly affected are immunocompromised individuals with cancer or AIDS, people undergoing organ and hematopoietic stem cell transplantation (HSCT), and those receiving immunosuppressive therapy or implantable prosthetic devices [1],[2]. The growing population of these at-risk groups is reflected in an increase in invasive fungal infection over the last three decades [3]. Annual treatment costs for fungal therapies reach $2.6 billion in the US alone [4]. Despite available therapy options mortality rates approaching 30–50% (Candida species) and 30–80% (Aspergillus species) remain high [5],[6].

Candida and Aspergillus species together account for ∼70% of all invasive fungal infections, with Candida albicans and Aspergillus fumigatus predominating [7],[8],[9]. Currently, three classes of antifungal drugs are suitable for treatment of systemic infections caused by these fungi: polyenes (most notably amphotericin B) and azoles (e.g. fluconazole, FCZ) have been applied for decades, while the echinocandins (e.g. caspofungin, CF) represent a new class of antifungal that has entered treatment regimes over the past 10 years [10],[11]. While these therapy options can be effective, they also exhibit several shortcomings. First, current antifungals target a very limited number of biological processes. The majority of available drugs target ergosterol (polyenes) or inhibit lanosterol 14α-demethylase (azoles), resulting in the accumulation of toxic sterol intermediates that disrupt membrane integrity and lead to membrane stress. Because ergosterol, the major sterol in fungal cell membranes, is analogous to the mammalian lipid cholesterol, this strategy, particularly when amphotericin B is applied, can be problematic due to host toxicity [10]. Another complication of current antifungal strategies is that available drugs each possess a different spectrum of antifungal activities. For instance, azoles are typically fungistatic against pathogenic yeasts such as Candida species, but fungicidal against molds (Aspergillus species). CF, on the other hand, is fungicidal against yeasts and fungistatic against molds [12]. Finally, and most importantly, the small number of treatment options available has resulted in widespread drug resistance in pathogenic species. For each of the three major classes of antifungals (polyenes, azoles, echinocandins) isolation of drug-resistant clinical strains has been reported [11],[12],[13]; azole-resistant Candida, in particular, is now common among isolates from HIV-positive patients [14]. Developing new antifungal strategies, therefore, remains a pressing need.

One approach to satisfy this need is through combination antifungal therapy, where two (or more) agents combined are significantly more efficacious compared to either agent alone. This approach has recently been validated in a randomized, placebo-controlled trial, where approved antifungals were combined with immune regulatory agents [15]. Results from this study suggested that combining ergosterol inhibitors with a recombinant human monoclonal antibody against heat-shock 90 protein (HSP90) showed increased therapeutic benefits compared to monotherapy against Candida infection. Although the precise mechanisms involved remain elusive [16], extensive experiments have further established the benefits of such combinatorial approaches. For instance, a potent synergy resulted when inhibitors of HSP90 (geldanamycin, radicicol) or inhibitors of HSP90's key client protein, calcineurin (cyclosporin A (CsA), FK506) were combined either with azoles or echinocandins against C. albicans [10],[16],[17],[18],[19],[20],[21]. Similarly, pharmacological compromise of HSP90/calcineurin-signaling enhanced the efficaciousness of echinocandin treatment against A. fumigatus in vitro as well as in insect and mouse infection models [16],[17]. Although these examples clearly demonstrate the potential for combination antifungal therapy, human host toxicity associated with inhibition of HSP90 or suppression of the human immune system by CsA/FK506 currently precludes the use of such inhibitors in the clinic [16],[22]. While a non-immunosuppressive FK506 analogue (L-685, 818) has been identified, proprietary restrictions have currently prevented further testing [22]. Therefore, identification of new antifungal targets for optimal fungal killing remains a priority.

One of the challenges of finding new antifungal targets in C. albicans is the lack of sophisticated screening technologies often employed with, for example, Saccharomyces cerevisiae. Various large-scale chemical-genomic drug screening methods are now well established in S. cerevisiae, and have been effective for elucidating drug targets or revealing insights into the modes of action of bioactive compounds [23],[24],[25],[26],[27]. Similar approaches have only recently been applied directly to fungal pathogens [28]. Using S. cerevisiae as a model, we previously performed chemical-genomics to systematically analyze the genetic requirements to survive FCZ treatment [29]. In that work, we identified 22 genes that become essential for S. cerevisiae survival in the presence of FCZ.

Here, we expanded that work with the aim of identifying synergistic drug interactions that render FCZ fungicidal in C. albicans. To this end, we validated the S. cerevisiae FCZ-cidal gene set [29] in C. albicans. From 22 predicted genes, we found that only one gene, AGE3, mediated FCZ tolerance in the pathogen. We further show that both genetic and pharmacological compromise of ARF (ADP ribosylation factor) activities, a process that depends on Age3p, creates sensitivity to all three classes of antifungals used in clinics (polyenes, azoles, echinocandins), overrides clinical drug resistance and the calcineurin pathway, synergizes with fungistatic drugs against the two major pathogenic fungal species (C. albicans and A. fumigatus) and modifies fungal virulence in two established mouse models of candidiasis. Given that drug treatment in mice was significantly more efficacious when ARF activity was genetically compromised, this demonstrates that targeting ARF signaling has potential for antifungal therapies.

Results

Age3p mediates azole tolerance and sensitivity to cell wall inhibitors in C. albicans

To identify genes that become essential for survival in the presence of FCZ, we previously screened the non-essential S. cerevisiae knock-out collection (about 4900 strains) and identified 22 mutants that showed a robust FCZ-cidal phenotype [29]. BLAST searches identified C. albicans homologs for 21 of those S. cerevisiae genes. YDR532c appears to be the only gene that lacks a clear C. albicans homolog (Table S1). Recreating knockout or transposon insertion mutations of the 21 candidate genes in C. albicans, we found that four mutants (bem2, sac6, srb8 and ssn3) showed FCZ sensitivity comparable to WT (Table S2, Figure S1). Twelve C. albicans mutants (57%) showed increased FCZ sensitivity, but all of these mutants could still resume growth when incubated for extended time in the presence of FCZ. Four genes (GCN5, NGG1, ERG11 and NUP84) were linked to a slightly resistant FCZ phenotype. Only one mutant, age3 (ORF19.3683), showed the FCZ sensitive phenotype predicted from the yeast screen. We therefore focused further investigation on AGE3.

We validated the FCZ sensitivity of age3 cells by three different assays. When tested in a minimal inhibitory concentration (MIC) assay, the age3 mutant initially showed similar drug sensitivity as WT and revertant strains at 24 hours. Since FCZ on its own is fungistatic, however, WT cells demonstrated robust growth above the initial MIC point after prolonged incubation (72 hours), a feature referred to as tolerance [19]. In contrast, age3 cells did not resume growth above the 24 hours MIC point, indicating that age3 mutants lost tolerance to FCZ (Figure 1A, Table S2). These results were confirmed visually by growth on solid rich media in the presence of FCZ (Figure 1B). We further characterized the FCZ sensitivity of age3 cells by time-kill curves. Under FCZ treatment, the number of viable age3 cells slightly decreased over time, while growth of WT cells in the presence of FCZ continued (Figure 1C).

We then tested the age3 mutant against a variety of antifungals to gauge the specificity of the mutation. We included various compounds, including second-generation azoles (voriconazole), non-azole ergosterol inhibitors (terbinafine) and other membrane-targeting drugs (amphotericin B). We found that age3 cells showed a generalized increased sensitivity to these compounds (Figure 1D). Among all cell-membrane drugs tested, the azoles caused by far the most significant enhancement in sensitivity in the age3 mutant.

In order to determine the effect of deleting AGE3 on the integrity of the cell wall, we tested the age3 mutant for sensitivity to a variety of cell wall perturbing agents and other agents whose effect have been linked to altered cell wall and glycosylation. The age3 mutant was slightly more sensitive to the β-1,3 glucan synthase inhibitor CF. Similarly, age3 cells were slightly more sensitive to calcofluor white, a phenotype that is usually associated with altered chitin structures along the cell wall [30]. More remarkably, age3 cells were extremely sensitive to hygromycin B, a phenotype usually seen in glycosylation mutants [31],[32] (Figure 1D, 1E). No change in sensitivity was observed in the presence of other agents such as caffeine, cycloheximide, menadione, nocodazole, rapamycin, 5-FC and wortmannin (data not shown). Together, these data suggest that while AGE3 plays a major role during membrane stress in C. albicans, its influence on the integrity of the cell wall remains somewhat less clear (see discussion).

Deleting AGE3 overrides clinical drug resistance and the calcineurin pathway

Among the most commonly encountered resistance mechanisms in drug treated clinical C. albicans isolates are over-expression of drug pumps or alterations in sterol biosynthesis [13],[33],[34]. To test whether such common mechanisms of drug resistance are still effective in the absence of AGE3, we deleted AGE3 in two FCZ-resistant clinical strains. The strain F5 carries a mutation in the transcription factor MRR1, which leads to constitutive over-expression of drug pumps. The strain S2 carries a mutation in the transcription factor UPC2, which causes up-regulation of ergosterol biosynthesis genes [35],[36]. Figure 2A shows that deleting AGE3 in strains F5 and S2 restored FCZ sensitivity even below WT levels, suggesting that loss of AGE3 abrogates FCZ-resistance in these clinical isolates.

Because age3 mutants lost tolerance to FCZ and because the calcineurin pathway is known to mediate FCZ-tolerance in WT as well as drug resistant clinical isolates [10],[19], we tested whether constitutive calcineurin signaling could reverse the AGE3-dependent FCZ sensitivity. The calcineurin pathway can be activated by addition of extracellular CaCl₂ [19]. C. albicans WT became resistant to FCZ after only 24 hours of growth in the presence of extracellular CaCl₂, while this rescue was not observed in the absence of AGE3 (Figure 2B). To further verify this observation, we deleted AGE3 in a constitutively active calcineurin signaling mutant (DSY2146) that is resistant to FCZ even in the absence of extracellular CaCl₂ [19]. Deleting AGE3 in this constitutively active calcineurin mutant restored FCZ sensitivity both in the absence and in the presence of extracellular CaCl₂ (Figure 2B). These results suggest that constitutive calcineurin signaling does not rescue the age3-dependent FCZ sensitivity. The results also support an argument that AGE3 and calcineurin-dependent processes could be linked (see discussion).

Pharmacological compromise of ARF cycling converts FCZ into a fungicidal drug in C. albicans

Given AGE3's role in sensitivity to various drugs and its implication in clinically relevant processes such as drug resistance, targeting either AGE3 or its biological process seemed a plausible avenue for combination therapies to render FCZ fungicidal. The S. cerevisiae homolog of C. albicans AGE3 is GCS1, which encodes an ARF GAP (ADP-Ribosylation Factor GTPase Activating Protein) [37]. ARFs are small G-proteins of the Ras GTPase superfamily that cycle between an active GTP-bound and an inactive GDP-bound state. ARF guanine nucleotide cycling, and hence function, is regulated by GAPs and GEFs (guanine nucleotide exchange factors) [38]. ARFs are involved in a variety of processes including vesicle trafficking (Golgi-to-ER retrograde vesicle trafficking, trans Golgi network-endosomal transport, transport from the Golgi to the membrane) and actin cytoskeleton organization [39],[40],[41],[42]. Brefeldin A (BFA), a metabolite from the fungus Penicillium decumbens, is a noncompetitive inhibitor of ARF activity. Protein crystal structures showed that BFA binds to a ternary complex of ARF-GDP-GEF, thus stabilizing this otherwise transient protein-protein interaction [43],[44],[45],[46],[47],[48]. Given that ARF cycling is a well-established target of BFA, we reasoned that BFA might be an ideal drug to synergize with FCZ in WT C. albicans by chemically mimicking the age3-dependent FCZ sensitivity. As illustrated by time-kill curves in Figure 3A, combining FCZ with BFA resulted in a potent fungicidal synergy in WT C. albicans, while either drug alone had only minor effects on cell growth. To compare the BFA/FCZ synergy to a well-established fungicidal synergy, we repeated the time-kill curves with CsA in combination with FCZ. CsA/FCZ co-treatment resulted in a similarly strong synthetic phenotype after 24 hours. This observation corroborates previous findings that calcineurin inhibition plus FCZ results in a potent fungicidal combination as assayed at 24 hours of drug treatment [19]. A triple drug combination of FCZ/BFA/CsA was significantly more efficacious than either FCZ/CsA or FCZ/BFA alone at 24 hours. However, when monitored for more than 24 hours, conditions that have not previously been reported [19], cells treated with FCZ/CsA could recover and resume growth, while cells treated with FCZ/BFA or FCZ/BFA/CsA could not resume growth above the detection limit (≈10 cells/ml). At 72 hours, drug combinations of either FCZ/BFA or FCZ/BFA/CsA appeared equally efficacious with no evidence of growth, while cells treated with FCZ/CsA continued to proliferate. Similar results were obtained when another calcineurin inhibitor, FK506, was combined with FCZ. These results suggest that while the combination of calcineurin inhibitors and an azole is initially efficient in pathogen killing, over prolonged drug treatment, combining ARF inhibitors with azoles is more efficacious.

A dose-matrix titration assay measuring growth of treated cells confirmed the synergy between FCZ/BFA (Figure 3B, Table S3). We also tested whether the FCZ/BFA synergy is still effective in C. albicans drug resistant clinical isolates and found that, although considerably higher concentrations of FCZ were needed, there was synergy in isolates F5 and S2 as well as in isolate 5674, which carries a gain-of-function mutation in TAC1, a transcriptional activator of CDR drug efflux pump genes [49]. Thus, genetic compromise of ARF cycling by deleting AGE3 abrogated FCZ tolerance, while pharmacological compromise of ARF cycling by adding BFA converted the fungistatic drug FCZ into a fungicidal agent in WT and FCZ-resistant C. albicans clinical isolates. Together, this suggests that the process of ARF cycling becomes essential during cell membrane stress in this pathogen.

Combining ARF inhibition with other drugs across fungal species

Importantly, genetic or pharmacological compromise of ARF cycling did not appear to significantly affect the cells' initial response to FCZ (Figure 1A, 3B, Table S2). Instead, ARF cycling inhibition seems to act on tolerance. To test whether the effect of BFA on tolerance is observed in combination with other azole drugs, we tested miconazole (MICO) and ketoconazole (KETO) in combination with BFA, because MICO and KETO alone did show a tolerance effect (Figure 4A). When tested in a dose-matrix titration format, BFA synergized with MICO and KETO against WT C. albicans, independently of which media were used (Figure 4A, Figure S2).

Although CF is generally considered fungicidal in C. albicans, we tested whether BFA would synergize with CF. Recent reports showed that C. albicans can start to grow at supra MIC concentrations of CF, an outcome referred to as the paradoxical or trailing growth effect [50],[51]. BFA did synergize with CF against WT C. albicans in a dose matrix titration assay at supra MIC concentrations of CF (Figure 4B). Together, these results indicate that genetic and pharmacological compromise of ARF cycling influences not only cell membrane stress, but also cell wall stress.

To examine whether BFA's inhibition of azole tolerance is conserved across other pathogenic Candida species, we tested BFA/drug interactions in C. tropicalis, C. parapsilosis, C. glabrata and C. krusei. Together, these species account for ∼30–40% of all Candida isolates causing invasive infections; furthermore, C. glabrata and C. krusei are notoriously difficult to treat with FCZ [7],[8],[12]. C. tropicalis, C. parapsilosis and C. glabrata showed tolerance in the presence of FCZ, MICO and KETO (Figure 5A); these azoles synergized with BFA in dose-matrix titration assays with these fungi. FCZ also synergized with BFA in the non-pathogenic yeast S. cerevisiae (data not shown). On the other hand, C. krusei did not show an obvious tolerance effect (i.e. >2 fold variations between 24 hours and 72 hours MIC readings), which could explain why BFA did not synergize with azoles in this pathogen (Figure 5B).

To further investigate ARF cycling inhibition as a mechanism of abrogating drug tolerance in non-Candida human fungal pathogens, we asked whether BFA would synergize with FCZ or CF against A. fumigatus. Consistent with the idea that ARF cycling inhibition acts primarily on tolerance, a disc diffusion assay in A. fumigatus demonstrated that BFA synergized with CF, a drug that is generally fungistatic in A. fumigatus (Figure 6). CF treatment created an inhibition zone against A. fumigatus where cells could still grow within that zone due to the fungistatic nature of CF. In contrast, when BFA was combined with CF, not only was the size of that inhibition zone increased, but growth within that zone was also remarkably reduced. A MIC assay in defined synthetic media independently confirmed that CF synergized with BFA against A. fumigatus (Table S3). On the other hand, we found that BFA did not synergize with FCZ, a drug that is considered fungicidal in A. fumigatus (data not shown) [12]. In general, pharmacological compromise of ARF cycling in A. fumigatus was not as efficacious as in Candida species, possibly because BFA on its own had a more pronounced impact on growth of A. fumigatus compared to Candida species (Figure 6, Table S3). Taken together, these results show that ARF cycling inhibition can couple with different fungistatic drugs in many pathogenic fungi to generate potentially fungicidal activity.

Core transcriptional responses to FCZ and CF treatment are not significantly affected in the absence of AGE3

Analysis of transcriptional regulation has frequently been used to elucidate which cellular processes are linked to drug sensitivity or drug resistance in C. albicans [35],[52],[53]. We therefore performed microarray studies to test how transcriptional regulation is altered in age3 mutants. We first compared WT to age3 cells in the absence of drugs and found that 23 genes were differentially regulated when AGE3 was absent (Table S4). Among those 23 genes were five GPI-anchored cell wall proteins (ECM331, PGA13, CRH11, SAP9, PGA26) and two genes with phospholipase activity (FGR22 and PI-PLC). Because none of these genes have been linked to FCZ tolerance, it is currently unclear how they might influence the age3-dependent drug phenotypes.

We next analyzed the transcriptional response to FCZ. A typical transcriptional signature to FCZ is upregulation of ergosterol genes, presumably to compensate for depletion of these membrane lipids [54],[55]. This core response to FCZ was not changed in the absence of AGE3. All ergosterol genes that became significantly upregulated in FCZ-treated WT cells were similarly upregulated in FCZ-treated age3 cells (Table S5). Besides this core response to FCZ, clustering analysis of genes that were significantly regulated (>2 fold expression change and p-value <0.05) further confirmed that the overall transcriptional response to FCZ was very similar in WT and age3 cells (Figure S3A, Table S6).

A similar lack of AGE3-dependent transcriptional consequences could be observed in microarray experiments when the cell wall inhibitor CF was used. Clustering analysis revealed that CF-treated WT cells and CF-treated age3 cells showed an almost identical transcriptional response, with a significant overlap (p-value 4.8×10⁻¹⁹⁴) of differentially regulated genes (Figure S3B, Tables S7, S8). This overlap of 168 genes showed further statistically significant similarity (p-value 1.8×10⁻⁷) when compared to core C. albicans CF-responsive genes previously identified in two independent studies [52],[56]. Therefore, the core transcriptional response to CF in C. albicans seems not to depend on AGE3. In summary, these microarray experiments suggest that deleting AGE3 does not cause major transcriptional changes in the presence of two different drug classes and further indicates that post-transcriptional processes might play a more dominant role in terms of ARF cycling-dependent drug phenotypes.

The age3 mutant is avirulent in WT mice

To evaluate whether AGE3 is a good drug target in vivo, we first injected age3 mutant cells into WT B6 mice, and found that age3 cells are avirulent in this mouse model. While 100% of mice infected with WT C. albicans became moribund within 11 days post-infection, none of the age3 mutant strain infected mice became moribund during the same time frame (Figure 7A). To test whether mice infected with age3 cells had cleared the infection, we sacrificed half of the mutant group on day 11 to analyze kidney fungal burden (Figure 7B). On average, kidney fungal load was significantly reduced (p-value<0.01, Mann Whitney test) in the age3 group compared to moribund WT-infected mice. We continued with the other half of age3 mutant-infected mice until the end of the experiment (day 21), but again found that none of the age3-infected mice became moribund. Comparing fungal load from these mice showed that three mice had cleared the infection, while two mice had a fungal load that was comparable to WT-infected mice. On average, however, there were significantly fewer age3 cells recovered from the host compared to moribund mice infected with WT C. albicans (p-value <0.02, Figure 7B).

One key virulence factor in C. albicans pathogenicity is hyphal formation [57]. Because Arnold Bito and coworkers (Lettner T., Zeidler U., Gimona M., Breitenbach M., Bito A., manuscript submitted, personal communication from A. Bito) have observed some hyphal formation defects on solid media as well as defects in invasive growth in age3 mutants, we collected kidneys for histological examination on day 11 from both WT-infected and age3-infected mice. No obvious difference was found between kidney sections recovered from WT or from age3-infected mice (Figure 7C). In all kidney sections examined, age3 cells could be observed as elongated hyphal structures. Therefore, it remains unclear why age3-infected mice with a high fungal burden did not show any clinical signs, but might indicate that additional virulence factors are affected in age3 mutants.

Genetic inhibition of ARF cycling results in attenuated virulence in an immunocompromised mouse model of disseminated candidiasis and FCZ treatment is significantly more efficacious when ARF activity is genetically compromised

To help evaluate the therapeutic potential of FCZ treatment of age3 mutant-infected mice, we required a mouse model where age3 cells retain at least partial virulence. Cells of the C. albicans strains WT, age3 mutant and age3 revertant were therefore injected in an immunocompromised C5-deficient A/J mouse model [58]. In this very sensitive animal system, C. albicans WT and revertant-infected A/J mice became rapidly moribund after 20 to 24 hours, while age3 mutant-infected A/J mice survived significantly longer, with a median survival of two days (p-value<0.02, Log-rank test) (Figure 8A). However, comparing fungal burden indicated that mice infected with C. albicans age3 cells accumulated a significantly higher fungal load when moribund (p-value <0.02), with hyphal formation still observed in the age3 mutant-infected mice (Figure 8B, data not shown).

To test whether genetic compromise of ARF activity holds therapeutic potential, we repeated this experiment and injected WT, age3 and the revertant strains in A/J mice to compare survival lengths after two days of FCZ therapy. FCZ administration extended survival of all three groups (p-value <0.001), but was significantly more efficacious in the mutant-infected mice, extending median survival time more than 10-fold compared to 7.5-fold median survival time extension for both the WT and revertant groups (p-value <0.01 for posthoc comparison between age3 mutant group with and without FCZ, versus p-value >0.05 posthoc comparison between WT with and without FCZ or p-value >0.05 for posthoc comparison between revertant with and without FCZ, Dunnett's Multiple Comparison) (Figure 8C). In summary, we conclude from these in vivo experiments that genetic compromise of ARF cycling results in avirulence in WT mice, significant attenuation in virulence in an immunocompromised mouse model and further suggests that FCZ treatment in A/J mice is more efficacious when ARF cycling is genetically compromised.

Discussion

Identifying new drug targets is an important step in the challenging task of developing new antifungal therapies, which are urgently needed due to the emergence of drug resistance to every class of antifungals currently in clinical use [11],[59]. Using a comprehensive reverse-genetics screen, we identified Age3p and the process of ARF cycling as potential new drug targets. We further established a widely conserved, potent fungicidal chemical synergy between the ARF cycling inhibitor BFA and several fungistatic drugs, and, through two murine infection models, validated the potential of ARF cycling as an antifungal target.

To render FCZ fungicidal through combination with other drugs, we applied a large-scale chemical-genomic approach in S. cerevisiae and found 22 FCZ-cidal genes (Table S1) [29]. The S. cerevisiae screen proved to have predictive power in that 12 of the 22 genes (57%) identified in S. cerevisiae were validated in C. albicans with an increased FCZ sensitivity (Figure S1, Table S2). Among those genes, only CDR1 and ADA2 have previously been linked to FCZ sensitivity in C. albicans [52],[60]. On the other hand, our screen linked four genes to FCZ resistance in the pathogen (NUP84, ERG11, GCN5 and NGG1). One of those genes, ERG11, has previously been linked to FCZ resistance in C. albicans [61]. Of note, Sanglard and coworkers have shown that the C. albicans erg11 mutant was sensitive to a variety of drugs, including BFA [61]. Two other C. albicans genes that have been linked to FCZ resistance in our screen (GCN5 and NGG1) are part of the SAGA, a conserved transcriptional co-activator [62]. Our data therefore expand the known events of transcriptional rewiring between S. cerevisiae and C. albicans in regards of drug resistance [63],[64], as another subunit of the SAGA complex, Ada2p, has been linked to increased drug sensitivity in our screen as well as in two previous studies [52],[60]. The observation that different subunits of the same transcriptional co-activator complex regulate the opposing phenotypes of drug sensitivity or drug resistance illustrates the extreme adaptability and flexibility of transcriptional regulatory networks in fungi.

Given that diverse essential processes exhibit significantly different regulation in S. cerevisiae and C. albicans [65],[66],[67], reorganization of transcriptional regulation might in general account for C. albicans mutants that showed different drug phenotypes compared to the yeast prediction. Despite these considerations, budding yeast genetics remains a convenient and powerful approach to predict phenotypes in pathogenic fungal species, at least until similar sophisticated screening techniques become available in the pathogen itself.

An alternative speculation why the success rate of our screen was not higher is that the gene encoding ERG11, the target of FCZ in both species, is essential only in S. cerevisiae [29],[61]. Thus, there may be redundant mechanisms for sterol biosynthesis in C. albicans, a suggestion that is further supported by the observation that another ergosterol biosynthesis gene, ERG24, is essential in S. cerevisiae, but not in C. albicans [68].

Chemical compromise of ARF cycling appeared far more potent than genetic compromise. When age3 mutants were treated with FCZ, the number of viable cells was reduced in time-kill curve experiments by less than one log value, whereas combining FCZ with BFA reduced the number of viable WT cells more than one log value over the same time (Figure 1C, Figure 3A). Similarly, chemical compromise appeared more potent than genetic compromise in terms of synergy with CF. The age3 mutant was only 2-fold more sensitive to CF than WT. On the other hand, combining BFA with CF resulted in a 32-fold reduction in sensitivity to CF in WT cells (Figure 1D and 1E, Figure 4B). These differences likely reflect that, while chemical interference with ARF function inhibits potentially all ARF GEFs, genetic inactivation was restricted to one ARF GAP (AGE3). Thus, it remains possible that some aspects of ARF signaling continue to function in the absence of AGE3 under drug conditions, a suggestion that is supported by work in yeast that demonstrated that several ARF GAPs provide redundant functions [39],[40].

The finding that both genetic and pharmacological blockage of ARF function result in increased azole sensitivity suggests that, among the multiple cellular roles described for the yeast homologue of AGE3, defects in proper ARF cycling and, therefore, defects in intracellular vesicle trafficking are responsible for drug phenotypes. One hypothesis to explain how incorrect vesicle trafficking could result in age3-dependent drug sensitivity is mislocalization of drug pumps. However, CDR1 and MDR1, two major drug pumps in C. albicans appear not be involved, as Arnold Bito and coworkers (Lettner T., Zeidler U., Gimona M., Breitenbach M., Bito A., manuscript submitted, personal communication from A. Bito) observed that CDR1 and MDR1 pumps are correctly localized to the plasma membrane in age3 mutants. They further established that CDR1 drug pump activity was not affected in the absence of AGE3.

Another plausible explanation that could account for the azole sensitivity of age3 mutants is defects in the biosynthesis of ergosterol or problems in transporting this membrane lipid to the cell membrane. Our microarray experiments provided evidence that the target pathway of azoles is not affected transcriptionally in the presence or the absence of FCZ (Figure S3A, Tables S5 and S6). We further found that the amount of ergosterol is similar in the plasma membrane of age3 mutants compared to WT (our unpublished data). These findings together with epistasis experiments showing that deletion of AGE3 restored azole sensitivity in different clinical isolates, suggest that the azole sensitivity of age3 cells is unlikely to depend on established mechanisms.

Finally, in support of the vesicle transport hypothesis is the observation that while age3 cells showed slightly increased sensitivity to different cell wall perturbing agents (CF and calcofluor white), the most impressive effect besides azoles was observed when age3 cells were treated with hygromycin B (Figure 1D). Hygromycin B sensitivity is usually linked to glycosylation defects [31],[32]. Therefore, it remains possible that some glycosylated proteins, including GPI-anchored proteins that normally reside in the cell wall, are not properly localized in age3 cells. How precisely defects in vesicle trafficking influence the observed drug phenotypes and whether drug sensitivity is caused by a general aspect of the secretory pathway or of a particular cell membrane or wall protein remains to be determined.

Drug resistance and virulence are two important biological aspects of pathogenic fungal species. While different fungal drug resistance mechanisms are now well understood [69],[70], various virulence-related attributes have been described that help Candida to cause infections [57],[59],[71]. Whereas genes critically involved either in drug resistance or virulence are attractive drug targets [59], an undeniably better option is targeting genes that are involved in both processes. The broad-range sensitivity to azoles and an echinocandin together with in vivo data showing that age3 mutants are avirulent in WT and exhibit significantly attenuated virulence even in an immunocompromised mouse model, indicates that Age3p and the process of ARF cycling is one such option. We further explored the therapeutic potential of ARF cycling inhibition by demonstrating that, in A/J mice, FCZ treatment was significantly more efficacious when ARF activity was genetically compromised.

One of our major problems was to reproduce the potent in vitro synergy of FCZ/BFA in animal models as we observed that a FCZ/BFA combination failed to rescue A/J mice infected with WT C. albicans (data not shown). One reason why the in vitro synergy failed to translate to in vivo conditions could be that BFA has low bioavailability characteristics [72] and efforts to chemically improve these unfavorable properties have not been successful so far [73],[74]. The ability of BFA to induce apoptosis in cancer cells has stimulated an interest for developing BFA as an anti-cancer therapeutic agent [75],[76],[77],[78],[79],[80]. Increasing evidence also shows that a variety of small G protein signaling pathways of the Ras superfamily, like RHO, RAS and ARF, have been linked to tumorigenesis [81],[82],[83]. Thus, despite being evolutionarily conserved, targeting ARF activities could be beneficial not only for antifungal, but also for cancer therapies. Clearly, one future challenge lies in finding fungal-specific ARF activity inhibitors that retain favorable bioavailability in vivo, a challenge that can now be approached cost- and time-effectively through virtual screening with both mammalian and yeast crystal structures of different ARF cycling proteins at hand [43],[44],[45],[46],[47],[48],[84],[85].

The age3 mutant lost tolerance to FCZ (Figure 1). This phenomenon of losing tolerance under membrane stress is not unique to ARF cycling. Previous work established that when HSP90 or calcineurin is genetically or pharmacologically inhibited in different fungal species, cells show a similar loss of tolerance and are not able to survive membrane stress [17],[18],[22],[86]. Additional phenotypes that calcineurin inhibition shares with ARF inhibition include synergism with azoles and echinocandins as well as reduced virulence in the bloodstream of the host [17],[19],[87]. These overlapping phenotypes, together with our epistasis experiments that showed that deleting AGE3 overrides constitutive calcineurin signaling, provide some arguments that ARF cycling and HSP90/calcineurin-dependent processes could be coupled. Although overlapping functions cannot be excluded, several lines of evidence support the current model that ARF and HSP90/calcineurin inhibition constitute two distinct mechanisms contributing to drug tolerance. First, a triple combination of azole/calcineurin/ARF inhibition is more efficacious than either pairwise combination at 24 hours. Second, consistent with previous observations that long term azole resistance evolves towards HSP90/calcineurin-independency [20], cells treated with calcineurin inhibitors plus FCZ recover after prolonged incubation, while BFA/FCZ treated cells do not. Finally, if ARF cycling is in fact coupled to HSP90/calcineurin signaling, then we might expect age3 mutants to copy other HSP90/calcineurin phenotypes besides the demonstrated reduced virulence and drug sensitivity. One explanation, for instance, why calcineurin mutants are avirulent is that they do not survive in the presence of FBS or Ca²⁺ ions present in host serum [19],[88]. We found, however, that age3 mutants can survive FBS and Ca²⁺ stresses (data not shown).

More work is therefore required to elucidate how ARF G protein signaling relates to HSP90/calcineurin and other signaling pathways that share ARF-related phenotypes. With four ARF proteins, six ARF GAPs and seven ARF GEFs identified in yeast so far [89], this provides a rich resource for further investigations into which aspects of vesicle transport and ARF G protein signaling are responsible for two important aspects of the biology of pathogenic fungal species.

Materials and Methods

Strains, plasmids, primers and culture conditions

All strains, primers and plasmids used in this study are described in supplementary Table S9, S10 and S11, respectively. C. albicans mutants were constructed either with the UAU1-transposon insertion strategy [90] or by deleting the coding sequence of genes (Table S9). For insertion and deletion mutant construction, at least two mutants independently derived from two distinct heterozygous mutants were analyzed in each case. With regards to the AEG3 gene (ORF19.3683), we propose to use AGE3 as standard name for the C. albicans homolog of S. cerevisiae's GCS1, because another gene (ORF19.5059) has already been named ‘GCS1’ in the Candida literature [91]. For the AGE3 deletion mutant, long 100-mer primers flanking up- and downstream sequences, respectively, of the coding sequence of AGE3 were used to amplify marker cassettes pFA-HIS1 and pFA-ARG4 [92]. Transformation was carried out according to standard protocols [93] and selected on synthetic media (2% dextrose, 6.7% yeast nitrogen base without amino acids, 2% agar) containing the necessary auxotrophic supplements. Correct marker integration was PCR-verified as described [94]. As genetic manipulation in C. albicans can frequently lead to aneuploidy [95],[96], we verified the absence of any chromosomal rearrangements by Comparative Genome Hybridization (CGH) and found that deletion mutants of AGE3 were aneuploidy free (Figure S4). For constructing the AGE3-revertant strain, the SAT1 flipper cassette was used [97]. Briefly, the AGE3 coding sequence including upstream and a short downstream flanking sequence was amplified with primers oEE242/oEE243, KpnI/XhoI-digested and cloned into the KpnI/XhoI-digested pSFS2A plasmid, resulting in plasmid pCaEE25, which was sequenced. A long downstream flanking sequence of AGE3 was amplified with primers oEE237/oEE238, NotI/SacII-digested and then cloned into the NotI/SacII-digested plasmid pCaEE25, resulting in plasmid pCaEE27. Following KpnI/SacII double digestion of plasmid pCaEE27, this digestion was transformed directly into the age3 deletion strain. Selection was done on YPD plates containing 200 µg/ml nourseothricin, as described [67] and PCR-verified. Counterselection of the nourseothricin marker was done as described [67].

The nourseothricin marker cassette [97] was also used to delete AGE3 in FCZ-resistant strains F5, S2 and DSY2146. Briefly, a long upstream coding sequence of AGE3 was first amplified from pEE27 with primers oEE235/oEE236, Kpn1/Xho1 digested and then cloned into the Kpn1/Xho1 digested plasmid pEE27, therefore replacing the coding sequence of AGE3. The resulting plasmid was designated pEE43.

Candida strains were routinely cultured at 30°C in either rich YPD media (1% yeast extract, 2% peptone, 2% dextrose, supplemented with 50 µg/ml uridine) or RPMI-MOPS media (RPMI-1640, SIGMA, supplemented with 0.3 g/l L-glutamine, 50 mg/ml uridine, 2% glucose, pH adjusted with MOPS buffer to 7.0). Media plates were supplemented with 2% agar. A. fumigatus was cultured in half-strength YPD media (not enriched with uridine) or RPMI-MOPS media (not enriched with uridine). Media plates were supplemented with 1.5% agar.

Antifungal susceptibility testing

Because drug susceptibility results did not differ significantly between WT C. albicans strains SC5314 (isolate) and SN95 (a standard laboratory auxotrophic mutant used to construct deletion mutants) [98] (Table S2), WT usually refers to SN95 unless indicated otherwise. Drug stock solutions were prepared using ethanol/10% Tween 20 as solvent for FK506 (5 mg/ml), cyclosporin A (25 mg/ml), DMSO for fluconazole (300 mg/ml), miconazole (100 mg/ml), ketoconazole (16.6 mg/ml), itraconazole (10 mg/ml), amphotericin B (20 mg/ml), calcofluor white (50 mg/ml), brefeldin A (20 mM), hygromycin B (50 mg/ml), and water for caspofungin (10 mg/ml). All drugs were obtained from Sigma, except caspofungin (Merck), fluconazole and itraconazole (both from SpectrumChemical, Mfg Corp, USA). Once in solution, drugs were stored at −20°C. Initial antifungal sensitivity testing with all C. albicans FCZ-cidal candidates was done using a modified version of the CLSI (formerly NCCLS) procedure [99]. Briefly, 50 µl of drugs at two-fold the final concentration was serially diluted in flat-bottom 96-well tissue culture plates (Corning Inc., NY, USA) and combined with 50 µl of overnight Candida cultures adjusted to 1×10⁴ cells/ml. MIC plates were incubated at 30°C without shaking and optical densities were read at indicated time points with a Tecan Safire plate reader. The MIC was determined by the first well with a growth reduction of >95% in terms of OD₆₀₀ values in the presence of a compound compared to untreated control cells. Before the 72 hours OD₆₀₀ readings, plates were carefully shaken, so that a representative aliquot of 2 µl of each well could be spotted on fresh YPD recovery plates to assess the extent to which cells recover from the drug treatments. Recovery plates were incubated at 30°C between 24 hours and 48 hours before being photographed. Drug susceptibilities of robust hits that resulted from this initial MIC testing were then confirmed on solid media plates containing FCZ, disc diffusion or time-kill curve assays, as described [19].

Dose-matrix titration assays were used to evaluate drug synergies. Briefly, dose-matrix titration assays were done as described for the MIC assays, except that the final volume was 150 µl; 50 µl of three-fold the final drug concentration of drug A was dispensed in two-fold serial dilution steps across seven columns of the plate, 50 µl of three-fold the final drug concentration of drug B was dispensed in two-fold serial dilution steps down seven rows of the plate; 50 µl of overnight Candida cultures adjusted to 1.5×10⁴ cells/ml was dispensed in all drug containing wells plus one control well containing no drugs. Synergy of a compound pair was quantified with respect to the Loewe additivity model [100] via the fractional inhibitory concentration index (FIC index) = (MIC_{A in combo}/MIC_{A alone}) + (MIC_{B in combo}/MIC_{B alone}), where “MIC_{A in combo}” is the MIC of drug A in combination, “MIC_{A alone}” is the MIC of drug A alone, “MIC_{B in combo}” is the MIC of drug B in combination and “MIC_{B alone}” is the MIC of drug B alone, respectively. Compound pairs were classified as synergistic if its FIC index is ≤0.5, the standard threshold [100],[101]. For calculation purposes of the FIC index, MIC values of >1, >2, >8, >16, >64, >128, >2048, >4096 were assumed to be 2, 4, 16, 32, 128, 256, 4096, 8192, respectively. MIC and dose-matrix titration results were visualized with TreeView version 1.6 (http://rana.lbl.gov/EisenSoftware.htm).

A. fumigatus disc diffusion assays were done as described [16], with the following modifications. Conidiation was induced on YPD plates incubated at 37°C for seven days. Conidia were then washed off the plates and suspended in PBS+0.1% Tween media before spreading 1×10⁵ conidia on appropriate plates. Discs containing 6.4 µg caspofungin were applied and the plates were incubated at 35°C for 48 hours. A. fumigatus MIC assays were done exactly as described [102]. All MIC and dose-matrix titration assays with Candida and Aspergillus were independently performed on at least two different occasions.

Virulence studies

Virulence testing of C. albicans was done as previously described [58]. Briefly, 8- to 12-week old C57BL/6J or A/J mice (Jackson Laboratories, Bar Harbor, ME) were inoculated via the tail vein with 200 µl of a suspension containing 3×10⁵ C. albicans in PBS. Mice were closely monitored over a period of maximally 21 days for clinical signs of disease such as lethargy, ruffled fur, or hunched back. Mice showing extreme lethargy were considered moribund and were euthanized. All experimental procedures involving animals were approved by the Biotechnology Research Institute Animal Care Committee, which operates under the guidelines of the Canadian Council of Animal Care. Statistical analysis of survival curves as well as fungal load was done with GraphPad Prism version 5.0b. For kidney sections, extracted organs were fixed in 10% formaldehyde (Sigma) and processed at the Histology Core Facility at McGill (http://cancercentre.mcgill.ca/) by staining thin slices of tissue sections with Grocot–Gomori methenamine silver to visualize fungal cells.

Microarray and CGH studies

Comparative Genome Hybridization (CGH) analysis was done as previously described [49] with the following modifications. Genomic DNA was extracted from a C. albicans culture grown to saturation with the Qiagen Genomic DNA Extraction kit according to manufacturer's instructions. DNA hybridization was done with the Advalytix SlideBooster for 16 hours at 42°C according to manufacturer's instructions.

For the fluconazole microarray experiment, overnight cultures of C. albicans cells were diluted to OD₆₀₀ of 0.05 in fresh YPD media, grown to early logarithmic phase (OD₆₀₀ 0.8) and split into two 50 ml cultures in 250 ml Erlenmayer flasks. One culture was fluconazole treated (600 mg/ml), while the other group received an equal amount of DMSO as control. Cultures were grown for one hour, spun down, the supernatant was removed and the cell pellet was nitrogen-flash frozen and stored at −80°C until further use. RNA was extracted according to the hot-phenol protocol [103]. For the caspofungin microarray experiment, logarithmically growing cells were treated with 125 ng/ml caspofungin for one hour, as previously described [52]. RNA was extracted with the RNase easy kit (Qiagen), as described [67]. Probe labeling, hybridization and slide washing was done as described [104], except that the SlideBooster was used for hybridization. At least three biological replicates including dye swaps were used for each condition on double spotted ORF microarrays (6,394 intragenic 70-mer oligonucleotide probes) [104]. Scanning was done with a ScanArray Lite microarray scanner (Perkin Elmer). QuantArray was used to quantify fluorescence intensities and data analysis was carried out using Genespring v.7.3 (Agilent Technologies). To compare overlap of different gene lists as well as analyzing Gene Ontology enrichment, p-values were calculated using the hypergeometric distribution as described in the GO Term Finder Tool web site (http://www.candidagenome.org/cgi-bin/GO/goTermFinder). Gene lists can be found in Tables S4, S5, S6, S7 and S8.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7

Attachment 8

Attachment 9

Attachment 10

Attachment 11

Attachment 12

Attachment 13

Attachment 14

Attachment 15