Dihydrofolate Reductase Is a Valid Target for Antifungal Development in the Human Pathogen Candida albicans

The folate biosynthetic pathway is a promising and understudied source for novel antifungals. Even dihydrofolate reductase (DHFR), a well-characterized and historically important drug target, has not been conclusively validated as an antifungal target. Here, we demonstrate that repression of DHFR inhibits growth of Candida albicans, a major human fungal pathogen. Methotrexate, an antifolate, also inhibits growth but through pH-dependent activity. In addition, we show that C. albicans has a limited ability to take up or utilize exogenous folates as only the addition of high concentrations of folinic acid restored growth in the presence of methotrexate. Finally, we show that repression of DHFR in a mouse model of infection was sufficient to eliminate host mortality. Our work conclusively establishes DHFR as a valid antifungal target in C. albicans.

ciated with severe patient toxicity that limits its use. The most recently approved antifungal drugs are the echinocandins, which target cell wall biosynthesis. Drugs in this class are principally active against Candida species and therefore have a narrow spectrum of activity and are available only in intravenous formulations, further limiting their clinical utility (5). A larger concern is the modest efficacy of all three classes of antifungal drugs as this is likely a major determinant of the excessively high mortality rates in patients with IFIs, as well as the high rates of recurrent mucosal infections. For example, approximately one-third of patients with disseminated Candida infections are nonresponsive to treatment with fluconazole (6-10), voriconazole (11), or the recently approved isavuconazole (12), even among those infected by isolates deemed susceptible according to current clinical breakpoints. Favorable response rates of just 52% to 73% have been reported for patients with disseminated candidiasis who are treated with an echinocandin (12)(13)(14)(15)(16) and of 62% for those treated with amphotericin B (14,16). Of grave concern is the recent isolation of Candida auris from patients across the globe that is resistant to all three classes of antifungal drugs (17). As such, there is an urgent need for new antifungal drugs with improved therapeutic efficacy, patient safety, and spectrum of activity.
In its reduced form, tetrahydrofolate (THF) is an essential coenzyme for a number of cellular enzymes, serving as a carrier for the transfer of one-carbon (1C) units as well as their interconversion between various oxidation states (18). THF is required for the synthesis of dTMP, purines, and methionine, as well as a multitude of other important metabolites. Mammals are unable to synthesize folate (FOL) as they lack several of the necessary enzymes and must acquire it through dietary intake. A specialized transport system is required for cellular uptake of folic acid, which is then converted into its active form, THF, by dihydrofolate reductase (DHFR) (19). In contrast, the majority of prokaryotes and microbial eukaryotes lack the transport systems found in mammals and must therefore synthesize folic acid de novo.
The folate biosynthetic pathway has been targeted with enormous success in the development of antineoplastic, antibacterial, and antiprotozoal drugs. Anticancer drugs include folate analogs such as methotrexate (MTX), a potent inhibitor of DHFR (20,21). Other notable antifolates include trimethoprim, which inhibits bacterial DHFR, and a collection of sulfa drugs, including sulfamethoxazole, that inhibit dihydropteroate synthase (DHPS) from some bacterial as well as protozoan parasites (22). Antiprotozoal drugs that target this pathway have been especially important for the treatment of malaria and include the DHFR inhibitors pyrimethamine, proguanil (PRO), and chlorproguanil (23)(24)(25), as well as the DHPS inhibitor dapsone (26). Curiously, a combination of trimethoprim and sulfamethoxazole can also provide an effective treatment for Pneumocystis pneumonia caused by the atypical fungus Pneumocystis jirovecii (27). Although these drugs were not specifically developed to target the folate biosynthetic enzymes of Pneumocystis, they provide clinical evidence that targeted inhibition of the folate (FOL) pathway can provide an effective intervention strategy to treat invasive mycoses. However, the conventional antifolate drugs developed for bacterial or protozoan DHPS or DHFR, or even human DHFR, have little or no antifungal activity upon the major human fungal pathogens (28)(29)(30), either because of divergence of the fungal enzyme structures or permeability issues that prevent antifolates from entering fungal cells (28). Furthermore, until now, no published study has unequivocally determined if the core components of the FOL pathway are required by infectious fungi to cause disease within a mammalian host and are therefore valid and potentially efficacious targets for antifungal development. The objective of this study was to determine if DHFR is essential for Candida albicans, one of the most prevalent human fungal pathogens, to survive and cause disease within its mammalian host. a strain expressing the tetracycline-responsive transactivator protein and replaced with the ARG4 selection marker (32). The promoter of the second allele was then replaced with one of three tetracycline-repressible promoters, P TETO97 , P TETO98 , or P TETO99 , which support different basal levels of transcription (33). DFR1 transcript abundance in strains of each genotype was then compared to that of a wild-type control (DFR1/DFR1) by quantitative reverse transcription-PCR (qRT-PCR) (Fig. 1). In the absence of doxycycline, the P TET97 -DFR1 strain produced levels of the DFR1 transcript similar to those of the wild-type control, while the P TET98 -DFR1 and P TET99 -DFR1 strains overproduced the DFR1 transcript by ϳ4and ϳ20-fold, respectively. However, in the presence of 5 g/ml doxycycline, the DFR1 transcript abundance was dramatically reduced in strains of all three genotypes to below that of the wild-type control (Fig. 1). We noted that the P TET97 -DFR1 strain grew slowly compared to growth of the wild type, even in the absence of doxycycline, suggesting that Dfr1p production is insufficient, while both the P TET98 -DFR1 and P TET99 -DFR1 strains grew at rates comparable to the wild-type rate (see Fig. S1 in the supplemental material). We, therefore, selected the P TET98 -DFR1 strain for further study.
Next, we determined if DFR1 expression is required for C. albicans growth in vitro. When inoculated into rich yeast extract-peptone-dextrose (YPD) medium supplemented with 5 g/ml doxycycline, the P TET98 -DFR1 strain grew to the same cell density as in the absence of doxycycline, and to the same density as the wild-type control ( Fig. 2A). However, when the strain was sequentially passaged in the presence of C. albicans strains of the indicated genotypes were grown to exponential phase in YPD medium in the presence or absence of 5 g/ml doxycycline (DOX). Total RNA was extracted, and DFR1 transcript abundance determined by qRT-PCR. DFR1 transcript abundance was normalized to that of the ACT1 transcript and then expressed relative to the DFR1 transcript abundance in the wild-type control (SC5314) in the absence of doxycycline. These data are the average of three biological and technical replicates. *, P Ͻ 0.0001 (for results compared to those with the parent strain under the same condition using an unpaired Student's t test). P 97 , P TET97 -DFR1 strain; P 98 , P TET98 -DFR1 strain; P 99 , P TET99 -DFR1 strain.

FIG 2
Dfr1p is essential for Candida albicans growth in vitro. A dfr1⌬/P TETO98 -DFR1 strain was sequentially passaged in YPD broth (A), YNB broth (B), or fetal bovine serum (C), each supplemented with 5 g/ml doxycycline (DOX). Following inoculation at approximately 5 ϫ 10 5 cells/ml, each culture was incubated at 30°C for 24 h, and growth was measured after 24 h as the OD 600 value. Relative growth was then expressed as a percentage of the untreated control, i.e., without doxycycline. The values presented here are representative of two independent experiments. doxycycline, growth was dramatically reduced by the third passage. This suggests that DFR1 is essential for C. albicans growth, even in nutritionally rich growth medium containing high concentrations of folate as well as metabolites that depend upon folate for their biosynthesis. The prolonged lag between doxycycline exposure and growth suppression could indicate that cellular stores of THF permit continued growth in the absence of Dfr1p until THF is depleted or that the protein has a long half-life. To distinguish between these possibilities, we attempted to directly inhibit Dfr1p using methotrexate. Previous reports have indicated that MTX lacks antifungal activity against C. albicans (28)(29)(30), despite potent inhibition of purified fungal DHFR in biochemical assays (34). However, we recently reported that MTX has relatively potent and on-target activity upon whole C. albicans cells in unbuffered yeast nitrogen base (YNB) medium (35). We therefore examined how MTX affected the growth of a wild-type C. albicans strain in YNB medium in dose-response assays. Significant growth inhibition was observed in this medium, with a MIC 50 of approximately 0.78 M (data not shown), without the need for passaging. In addition, passaging of the P TET98 -DFR1 strain with doxycycline in YNB medium resulted in detectable growth inhibition without the need for passaging (Fig. 2B). These results suggest that the continued growth of the P TET98 -DFR1 strain upon exposure to doxycycline is likely not due to the utilization of an intracellular store of THF. Instead, these results are consistent with the Dfr1p enzyme having a long half-life, resulting in a long lag before doxycycline-mediated suppression arrests fungal growth.
Candida albicans has a limited capacity to utilize exogenous sources of folate. Folate biosynthesis-deficient mutants of the yeast Saccharomyces cerevisiae are unable to grow in standard lab medium (36,37). However, growth can be restored by supplementing the medium with extremely high concentrations (Ն250 M) of folinic acid (5-formyltetrahydrofolate) (38). To determine if exogenous sources of folate are sufficient to support C. albicans growth in the absence of DHFR activity, we seeded strain SC5314 into unbuffered YNB medium supplemented with either folic acid, dihydrofolate (DHF), THF, or 5-methyltetrahydrofolate (5MTHF) and compared growth in the presence of different concentrations of MTX. Notably, standard YNB medium contains both folic acid (4.5 nM) and para-aminobenzoic acid (PABA), a precursor required for folate biosynthesis. Therefore, MTX sensitivity was also tested in folate-and PABA-free YNB medium. The MIC 50 of MTX in standard YNB medium (0.78 M) was similar to that in the folate-and PABA-free medium, indicating that the folate present in standard YNB medium is not sufficient to affect sensitivity. Additional supplements of folic acid were completely unable to rescue C. albicans growth at any soluble concentration (up to 25 M), and, in fact, the MIC 50 of MTX dropped ϳ10-fold to 0.078 M (Fig. 3). Supplementing the medium with either 1 M DHF, THF, or 5MTHF  (39)(40)(41) had no effect on the MTX-mediated growth inhibition of C. albicans. However, the addition of 250 M folinic acid to YNB was able to partially restore growth in the presence of MTX and elevated the MIC 50 to ϳ5 M (Fig. 3). Nevertheless, even in the presence of this excess of folinic acid, MTX inhibited C. albicans growth in a dose-dependent manner (Fig. 4), indicating that the need for Dfr1p activity is not completely bypassed. Accordingly, we conclude that C. albicans is unable to take up and utilize sufficient quantities of exogenous folates to sustain growth in standard laboratory medium.
A combination of dTMP, adenine, histidine, and methionine, each of which requires folate for its biosynthesis, is sufficient to partially restore S. cerevisiae growth following treatment with antifolates (36). However, these supplements were not sufficient to restore C. albicans growth in the presence of MTX (Fig. 3). This suggests that either the provision of exogenous sources of folate-dependent metabolites is not adequate to bypass the need for de novo folate biosynthesis in C. albicans or that the nutritional requirements of this species in the absence of folate are more complex than those of S. cerevisiae. Finally, we examined if C. albicans was able to scavenge sufficient folates or folate-dependent metabolites from blood serum to negate the need for de novo biosynthesis. The P TET98 -DFR1 strain was passaged for 3 days in 100% fetal bovine serum (FBS) with or without doxycycline. In the first passage, doxycycline inhibited growth by ϳ60%, with increasing growth inhibition seen in subsequent subcultures (Fig. 2C). Thus, C. albicans is unable to acquire sufficient nutrients from blood serum to bypass the need for de novo folate biosynthesis. DFR1 is essential for Candida albicans virulence in a mouse model of disseminated infection. Next, we determined if DFR1 expression is essential for C. albicans virulence in a mammalian host using a mouse model of disseminated candidiasis. Female BALB/c mice were split into two treatment groups: group 1 was treated with doxycycline in a gel food formulation from 72 h before infection and for the duration of the experiment; group 2 was mock treated using unsupplemented gel food. The mice were then infected with ϳ5 ϫ 10 5 yeast cells of the P TET98 -DFR1 strain via the lateral tail vein, and their health was monitored for 12 days. Animals exhibiting signif- icant signs of distress were euthanized, and survival rates were compared. All mice in the mock-treated group succumbed to the infection by day 7 (Fig. 5A), confirming the virulence of the P TET98 -DFR1 strain under nonrepressing conditions. In stark contrast, all doxycycline-treated mice survived for the duration of the experiment. On day 12, surviving animals were euthanized, kidneys were extracted and homogenized, and levels of fungal colonization were determined as CFU counts. This revealed that 5 of the 8 surviving mice had undetectable levels of fungal colonization, essentially clearing the infection, with the remaining animals having extremely low CFU levels (data not shown). To determine if suppression of Dfr1p activity is sufficient to resolve an established infection, a second experiment was performed in which mice were infected with the P TET98 -DFR1 strain, and doxycycline treatment was initiated 24-h postinfection. Again, all mice survived the duration of the experiment (to day 12 postinfection), with initial symptoms resolving within 6 days of doxycycline treatment. Half of these mice had undetectable levels of fungal colonization in their kidneys at the end of the experiment (data not shown). These data confirm that Dfr1p is essential for C. albicans to survive within and cause disseminated disease in a mammalian host. Furthermore, C. albicans is unable to scavenge sufficient sources of folate to survive within mammalian tissue, and de novo biosynthesis is required. Indeed, the antifungal efficacy achieved upon repressing Dfr1p expression appears to be similar to that attained upon suppression of Erg11p expression. Accordingly, DHFR and likely other FOL pathway enzymes can provide potentially efficacious targets for antifungal development.
The pH-dependent antifungal activity of methotrexate is likely due to differential uptake. While MTX has been shown to inhibit fungal DHFR in cell-free biochemical assays using purified enzyme, it has also been reported to lack antifungal activity (28)(29)(30). Given that these previous studies used RPMI 1640 pH 7 medium, we sought to account for the antifungal activity we observed in unbuffered YNB medium. We first confirmed that MTX lacked antifungal activity in RPMI 1640 pH 7 medium (MIC 50 Ͼ 50 M) (Fig. 6). Next, to determine if the antifungal activity of MTX is pH dependent, dose-response experiments were conducted with strain SC5314 in YNB medium buffered at pH 5 or pH 7. MTX inhibited C. albicans growth in YNB pH 5 medium with a MIC 50 of ϳ0.78 M (Fig. 6) but had no effect in YNB pH 7 medium (MIC 50 Ͼ50 M), indicating pH-dependent antifungal activity. Similar results were obtained in YNB medium lacking folate and PABA, buffered to either pH 5 or pH 7 (Fig. S2). In contrast, the P TET98 -DFR1 strain was unable to grow in regular YNB medium, in folate-and PABA-free YNB medium at either pH in the presence of doxycycline (Fig. 7A), or in YNB medium supplemented with different folates (Fig. 7B), indicating that Dfr1p expression and, by inference, de novo folate biosynthesis are essential at either pH. These data are consistent with the previous findings of Navarro-Martinez and colleagues that demon- Groups of BALB/c mice (n ϭ 6) were inoculated with ϳ5 ϫ 10 5 CFU of either the dfr1⌬/P TET98 -DFR1 strain or the erg11⌬/P TETO -ERG11 strain via the lateral tail vein injections. The Erg11p strain was included to compare the antifungal efficacy of targeting Dfr1p with a previously validated target. The mice were monitored three times daily for 12 days, and those showing signs of distress were humanely euthanized. Mice were treated with 2 mg/g doxycycline (DOX) or mock treated with vehicle alone provided in a gel food formulation from 72 h prior to infection and throughout the duration of the experiment. The survival of each group was compared using a log rank test (*, P Ͻ 0.0003). strated that MTX does not accumulate within C. albicans yeast cells at neutral pH (28) and may therefore account for the pH-dependent antifungal activity.
We therefore examined if the antifungal activity of MTX was affected by the most important drug efflux mechanisms in C. albicans. The susceptibility of a C. albicans cdr1⌬/⌬ mutant, lacking an ATP-dependent ABC family transporter, as well as that of a tac1⌬/⌬ mutant, lacking a zinc cluster transcription factor that activates the expression of the Cdr1p and Cdr2p drug efflux pumps (42), is not significantly different from isogenic or wild-type control strains in YNB at pH 5 (MIC 50 of 0.39 M) or pH 7 (MIC 50 Ͼ 6.25 M) (Fig. 8A and B). Similarly, we found that the susceptibility of an mdr1⌬/⌬ mutant, lacking a major facilitator superfamily transporter, which is driven by the proton motive force at the plasma membrane and which has been previously implicated in methotrexate efflux (43,44), was not significantly different from MDR1 ϩ control strains in YNB medium at pH 5 or 7 (Fig. 8C). The susceptibility of strains engineered to overexpress either the Cdr1p or Mdr1p efflux pumps (42) was also FIG 6 Methotrexate has pH-dependent antifungal activity on Candida albicans. Approximately 1 ϫ 10 4 cells/ml of a wild-type C. albicans strain was seeded into a 96-well plate with increasing concentrations of MTX in YNB pH 5, YNB pH 7, and RPMI pH 7 media. After 48 h, the OD 600 was measured and expressed as a percentage of the value for the DMSO-treated control. The results presented here are the averages of two replicates and are representative of two independent experiments.

FIG 7
Dfr1p is essential for Candida albicans growth in both acidic and neutral conditions. The dfr1⌬/P TETO98 -DFR1 strain was passaged in YNB medium at pH 5 or 7 as well as the folate-and PABA-free YNB medium at the same pH (A) or in YNB medium at pH 5 or 7 supplemented with different forms of folate (B) with approximately 5 ϫ 10 5 cells/ml from the previous passage in the presence of 5 g/ml doxycycline. The culture was incubated at 30°C in a rotating incubator, and after 24 h the OD 600 was measured. The growth was calculated as a percentage of that of the untreated control, i.e., without doxycycline. The values presented here are representative of two independent experiments.
unaffected. This lends further support for the hypothesis that the pH-dependent antifungal activity of MTX is a result of differential cellular uptake or permeability.
To determine if MTX's pH-dependent activity was specific to C. albicans, we tested its antifungal activity upon other fungal species in YNB medium at pH 5 and pH 7. As in C. albicans, MTX showed modest and pH-dependent growth inhibition on Candida tropicalis and Candida parapsilosis. In contrast, Candida glabrata was insensitive at either pH (Fig. 9).
Established DHFR inhibitors are unable to inhibit Dfr1p within Candida albicans cells. Finally, we tested a panel of additional antifolates to examine the relationship between their antifungal potencies and capacities to inhibit C. albicans DHFR. The enzyme was purified using a 6ϫHis tag (45,46), and activity was detected by measuring the conversion of NADPH to NADP ϩ at an optical density at 340 nm (OD 340 ) (34). Dose-response experiments confirmed MTX to be a potent inhibitor of C. albicans DHFR, while trimethoprim (TMP) had no activity (Table 1), as previously reported (47). While proguanil (PRO) also lacked activity (50% inhibitory concentration [IC 50 ] Ͼ 10 M), both pyrimethamine (PYR) and pemetrexed (PTX) inhibited C. albicans DHFR activity. Consistent with the biochemical data, neither TMP nor PRO was able to inhibit C. albicans growth in YNB medium buffered to either pH 5 or 7 (MIC Ͼ25 M). PYR also failed to inhibit C. albicans growth at either pH (Fig. S3), suggesting that the compound is unable to access DHFR in whole cells. Similar to MTX, PTX possessed antifungal activity at pH 5 but not at pH 7 in YNB medium, presumably reflecting pH-dependent cell permeability.

DISCUSSION
While folate biosynthesis has been successfully targeted to develop antibacterial, antiprotozoal, and antineoplastic therapies, there have been relatively few efforts to develop antifungals that target this pathway. Although a combination of trimethoprim and sulfamethoxazole is used clinically to treat Pneumocystis pneumonia and Paracoccidioides infections (27,48), many of the most important fungal pathogens are largely insensitive to conventional antifolate drugs (28,49,50). A handful of studies have attempted to produce derivatives of conventional antifolates with enhanced antifungal potency but have failed to yield derivatives with the requisite properties of a viable antifungal drug. For example, an entire series of sulfone compounds completely lack antifungal activity against whole C. albicans cells, despite potent inhibition of fungal The pH-dependent antifungal activity of methotrexate is species specific. Strains of C. albicans (A), C. glabrata (B), C. tropicalis (C), and C. parapsilosis (D) were seeded into YNB medium buffered at pH 5 or pH 7 at approximately 1 ϫ 10 4 cells/ml in the presence of a range of MTX concentrations. After 48 h at 30°C, growth was quantified as the OD 600 and expressed as a percentage of the value for the untreated controls. The values presented here are the average of two replicates and are representative of two experiments. The susceptibility to MTX at the highest concentration at each pH was compared to that of the other pH values using a two-tailed Student's t test (*, P Ͻ 0.01; **, P Ͻ 0.0001). DHPS in vitro (49). Similarly, a large series of diaminopyrimidines, which inhibit bacterial and human DHFR, have little activity against whole C. albicans cells despite potent inhibition of fungal DHFR in cell-free assays (49). DHFR inhibitors based on either quinazoline (51) or pteridine (52) ring scaffolds have potent but pH-dependent antifungal activity in vitro, which likely explains their lack of efficacy in an animal model of infection (52). One factor complicating the interpretation of these findings is that these studies often used different culture conditions or media. We therefore considered that the lack of success could be accounted for by infectious species, such as C. albicans, acquiring exogenous sources of folate from certain growth media or under specific conditions in vitro and/or from mammalian tissue in vivo. This would bypass the need for de novo folate biosynthesis and therefore render antifolate drugs ineffective as antifungals. However, our results using the doxycycline-repressible DFR1 strain revealed that DHFR and, by inference, de novo production of THF are essential for C. albicans growth in a variety of culture conditions, including within blood serum as well as in the mammalian host. Aside from extremely high concentrations of folinic acid, medium supplements of various forms of folate, including the DHFR product THF, were not able to restore growth upon loss of DHFR expression. Thus, DHFR is a valid and potentially efficacious target for antifungal development. These conclusions are further supported by the fact that methotrexate has relatively potent antifungal activity against C. albicans in the lower-pH range. In contrast, doxycycline-mediated suppression of DHFR activity in the P TET98 -DFR1 strain occurred less rapidly, requiring multiple passages to arrest fungal growth completely. This is likely a consequence of an indirect mechanism, acting through suppression of DFR1 transcription, rather than direct inhibition of the fungal enzyme's activity. Nonetheless, the symptoms observed within mice infected with the P TET98 -DFR1 strain resolved fairly quickly upon doxycycline treatment, indicating that the fungus is highly sensitive to perturbation of THF production in vivo. We conclude that an antifungal drug acting through inhibition of DHFR has the potential to provide a highly efficacious antifungal therapy against C. albicans. However, given that Dfr1p is present in humans and fairly well conserved between mammals and fungi, selectivity is likely to be an issue. Interestingly, Anderson and colleagues (18,53) described trimethoprim derivatives with an extended central linker with dramatically enhanced potency against C. glabrata DHFR. However, despite exhibiting selective inhibition of the fungal enzyme over the human enzyme in cell-free biochemical assays, toxicity to mammalian cells remained an issue (54). Accordingly, assuming that inhibition of earlier steps of the folate (FOL) biosynthetic pathway results in similar antifungal efficacy, it may make more sense to target the enzymes absent from the mammalian host.
Folate synthesis starts with the conversion of GTP to dihydroneopterin triphosphate by GTP cyclohydrolase I (Fol2p), which is subsequently condensed with para-aminobenzoic acid (PABA) to form dihydropteroate by the sequential action of dihydroneopterin aldolase (DHNA), hydroxymethyldihydropterin pyrophosphokinase (HPPK), and dihydropteroate synthase (DHPS) (55). In yeast, the last three enzyme activities are provided by a single trifunctional protein, Fol1p, while in filamentous fungi HPPK and DHPS form a bifunctional enzyme (38). Dihydrofolate synthase (DHFS; Fol3p) adds a single L-glutamate residue to produce DHF, which is subsequently reduced by DHFR (Dfr1p) to produce THF that can accept 1C substituents. Notably, mammals completely lack four of the core enzymes required to produce folate, DHNA, HPPK, DHPS (Fol1p), and DHFS (Fol3p). In the noninfectious yeast Saccharomyces cerevisiae, the FOL1 and FOL3 genes are both essential for viability under normal culture conditions (38,56). A recent large-scale study using transposon mutagenesis also predicted FOL1 to be essential in C. albicans (57), and an unpublished study has indicated that the bifunctional FOL1 ortholog is essential in Aspergillus fumigatus (58). Nonetheless, these enzymes are surprisingly poorly studied in infectious fungi or, indeed, in S. cerevisiae.
The fact that the doxycycline-repressible DFR1 strain is unable to grow in any growth medium tested (including those supplemented with THF) under repressing conditions is consistent with conventional antifolates lacking activity upon the target enzyme in whole cells rather than with an inherent capacity of C. albicans to bypass the need for folate biosynthesis. Given that both sulfa-based DHPS inhibitors and various DHFR inhibitors have demonstrated relatively potent inhibition of the respective fungal enzymes in cell-free assays (34,47,49,52,53), the most likely challenge in deriving antifungals that target folate biosynthesis is achieving sufficient intracellular accumulation. In the case of mammalian cells, DHFR inhibitors such as MTX, which are structurally related to folic acid itself, enter primarily through the endogenous folate transport system (59). While the lack of a known folate transport system, in theory, renders fungi vulnerable to antifolates, it simultaneously raises challenges in promoting the uptake of folate analogs. We consider two possible explanations for the pHdependent antifungal activity of MTX. First, fungi could potentially encode a pHdependent folate transport system that enables MTX uptake in more acidic conditions. However, we do not favor this explanation as such a system would be expected to facilitate the uptake of folates in the medium itself simultaneously and thus bypass the need for de novo synthesis. Second, and perhaps more likely, MTX uptake in fungi could occur solely through passive diffusion across the plasma membrane, with the pH dependence simply reflecting the ionization state of MTX. The pKa of MTX is 4.7 (60); thus, deprotonation of the associated carboxylic acid group of MTX at higher pH would confer a negative charge that may render it membrane impermeable. Either way, we found that the best-characterized drug efflux mechanisms in C. albicans, the protondriven Mdr1p and ATP-driven Cdr1p and Cdr2p transporters, had little impact on the sensitivity of strain SC5314 to MTX and therefore do not explain the lack of antifungal activity at neutral pH. Thus, existing drugs that inhibit DHFR cannot provide effective antifungals for two main reasons. First, the fungal enzyme is insensitive to these drugs, presumably due to evolutionary divergence from the DHFR enzyme of their intended target species. For example, while trimethoprim apparently possesses some antifungal activity against Pneumocystis jirovecii, given its clinical efficacy, C. albicans DHFR is insensitive (47). Second, many conventional DHFR inhibitors are unable to access the enzyme within the intact fungal cell, presumably due to limiting membrane permeability. This likely accounts for the lack of antifungal activity of pyrimethamine and the pH-dependent activity of MTX and pemetrexed.
C. glabrata was insensitive to MTX at any pH, which indicates that either the Dfr1p enzyme is structurally distinct or that this species has a different acquisition or permeability of folates. Previous studies have shown that the structure of Dfr1p is conserved between C. glabrata and C. albicans (CgDfr1p and CaDfr1p, respectively) (61). However, this same study established that while an inhibitor could inhibit both CgDfr1p and CaDfr1p, this enzyme inhibition did not always correlate with fungal growth inhibition, nor did the inhibition correlate between species (62). Therefore, we conclude that the differential in MTX pH-dependent activity may be a result of differential uptake of antifolates.
In summary, our results demonstrate that the de novo synthesis of THF is absolutely required for C. albicans to cause disease within the mammalian host and that DHFR is a valid and potentially efficacious target for antifungal development. However, given the structural similarity of fungal and mammalian DHFR and the difficulties in producing DHFR inhibitors that are active upon whole fungal cells, we propose that efforts to exploit the FOL pathway for antifungal development should focus upon the biosynthetic enzymes that have not yet been the subject of significant investigation and which are completely absent from mammals. In addition, this may enhance the chances of discovering novel antifolate scaffolds that have activity upon whole fungal cells. lysed with 1 mg/ml chicken white lysozyme (Sigma) for 60 min at 4°C. Then, Triton X-100 and NaCl were added to 0.1% and 0.5 M, respectively. The lysate was then centrifuged, and the supernatant was collected. The supernatant was added to nickel resin beads (Sigma) and incubated at 4°C for 30 min with rotation. The buffer was removed, and the beads were washed twice with binding buffer. Then the beads were suspended in elution buffer (6 g/liter Tris base, pH 7.5, 8.8 g/liter NaCl, 1 mM DTT, 34 g/liter imidazole, 5% glycerol) and incubated at 4°C for 30 min. The elution was removed and then run on SDS-PAGE gels (Sigma) to confirm the presence of CaDfr1p. SDS-PAGE gels were run according to the manufacturer's protocol.
Dfr1p enzyme assay. The enzymatic activity of C. albicans Dfr1p was determined using a dihydrofolate reductase assay (Sigma). The manufacturer's protocol was adapted to the fungal enzyme by scaling down the volume to 100-l reaction mixtures with 88 l of buffer, 6 l of a 1 mM NADPH stock, 5 l of a 1 mM DHF stock, and 1 l of 3-mg/ml purified enzyme. The loss of NADPH was quantified as the decrease in absorption at the OD 340 over a 30-min time frame. The IC 50 was calculated by converting the change in the OD 340 value per minute into the amount of THF produced per minute and then compared to the level in an untreated control.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.