To identify novel compounds with antifungal activity, high-throughput screening was performed using the Novartis compound archive and S. cerevisiae
cells with a miniaturized 1,536-well plate assay. The assay monitored proliferation using reduction of resazurin as a surrogate marker for metabolic activity (25
). Cycloheximide (100 μM) was used as a positive control, and a final concentration of 1% DMSO was used as a negative control. Typically, robust screening results with Z
′ values of >0.7 were obtained (39
). Overall, 1,101,408 compounds were tested in the primary screen, and approximately 0.9% of the tested compounds showed >90% inhibition of cell proliferation.
Initially, compounds showing >90% inhibition of cell proliferation in the 16-h assay were designated hits (). These hits were further prioritized based on their effects on mammalian cell viability and their predicted physicochemical properties. Hits were initially validated by dose-response characterization, and compound identity and purity were confirmed by LC-MS. The screening campaign identified a number of known antifungal compounds (which were removed from further characterization), suggesting that novel compounds identified in this screen might serve as useful antifungal compounds for further development. This eventually led to a set of 5,000 high-priority compounds that were then characterized further.
Fig 1 High-throughput screening for inhibition of S. cerevisiae proliferation. (A) Scatter plot, with each point representing the inhibitory activity of one individual compound tested during the screening campaign. Each plate contained 24 no-inhibition controls (more ...) Hit characterization and expansion.
Compound 1 () was identified from the primary screen and characterized further by determining the MIC against a panel of pathogenic fungi, using the CLSI antifungal susceptibility testing protocol (). Compound 1 was shown to inhibit the proliferation of S. cerevisiae and pathogenic Candida spp., including Candida albicans and Candida krusei, after incubation for 24 h at 30°C. No MIC was observed for Aspergillus fumigatus or Rhizopus oryzae when tested up to concentrations of 128 μg/ml (395.9 μM) for 48 h. The inhibitory activity of compound 1 is static but not cytocidal, as shown in Fig. S1 in the supplemental material. Thus, compound 1 does have antifungal activity against certain fungal species, demonstrating the utility of the original screen using S. cerevisiae. Further closely related chemical structures (which were not present in the original screening collection of compounds) were selected from the complete Novartis compound collection and tested for antifungal activity ( and ). Compounds 2 and 3 showed slightly more potent antifungal activity; compound 4 did not show activity against any of the species tested, despite its similarity to the other compounds. This clear structure-activity relationship suggests a specific binding target for the active substances.
Fig 2 HIP HOP profiling results. (A) Structures and S. cerevisiae (S.c.) IC30s of substances tested by HIP HOP profiling. (B to D) HIP HOP profiling of compound 1 and derivatives at the indicated concentrations. The gray squares represent strains with deletions (more ...)
To determine the selectivity of these compounds against fungi, the compounds were tested in mammalian cells using the WST-1 cytotoxicity assay against the K562 (CRL-1573), HEK293 (CCL-243), and HEPG2 (HB-8065) cell lines (). Compounds 1, 2, and 3 showed antiproliferative activity with the K562 and HEK293 cell lines (but at concentrations greater than their activity against fungal cells), while compound 4, which lacks antifungal activity, had only weak cytotoxic activity with the K562 and HEK293 cell lines.
Concentrations (IC50) of compounds tested in the WST-1 cytotoxic assay against the K562, HEK293, and HEPG2 cell lines
Target hypothesis. Saccharomyces
HIP and HOP were performed with compounds 1 to 4 of the compound cluster. HIP HOP is a gene dosage-dependent method that assesses the effects of compounds against potential targets encoded by the S. cerevisiae
). We profiled all active compounds at their IC30
in a HIP HOP assay that was performed essentially as described previously (16
). The inactive compound 4 was tested at 200 μM, based on its maximal solubility in DMSO.
The HIP results ( to ) suggest that the active substances, compounds 1, 2, and 3, act by targeting Erg11p, because the heterozygote erg11/ERG11
strain is one of the most sensitive strains in the profile, as is the heterozygote set6/SET6
strain, a key indicator of ergosterol modulation. Erg11p is the established target for azoles and encodes lanosterol 14-alpha-demethylase, which catalyzes an essential step in ergosterol biosynthesis (1
encodes a protein of unknown function, but the heterozygous deletion strain has been previously linked to ergosterol modulation (11
). The observation that the HIP profiles of all active compounds are similar and overlap the HIP profile of voriconazole, an established azole-containing Erg11p inhibitor, strongly suggests that it is the target of these compounds (19
) (). A HOP profile (homozygous profile) is based upon a genome-wide set of homozygous deletions of the nonessential genes in S. cerevisiae
and often highlights parallel or compensatory pathways. The HOP profile of compound 1 identifies five genes with significant sensitivity linked to Erg11p function: DAP1
encodes an established Erg11p-regulating protein, and GCS1
, and YPT31
encode membrane-trafficking components already associated with Erg11p inhibition by previous HIP HOP profiling experiments and genetic studies (11
). Comparison of the HOP profiles for compounds 1, 2, and 3 again revealed significant correlations with the profiles for other established Erg11p inhibitors, such as voriconazole and other azole-containing antifungals ( to and ). Compound 4 was tested at 200 μM but failed to identify any significant hits, in agreement with the observation that it is inactive in the fungal growth assays ().
Compound potencies and similarities of the obtained HOP profilesa
Since Erg11p was identified genetically as the possible target for compounds 1, 2, and 3, we decided to determine if we could measure altered concentrations of the substrate and/or product of Erg11p. To test this hypothesis, we quantified the relative concentrations of ergosterol and lanosterol by a targeted metabolomic-profiling approach following compound exposure for 16 h. In DMSO-treated S. cerevisiae control samples, very low levels of lanosterol were detected (). However, upon treatment with the known Erg11p inhibitor voriconazole, a strong increase in the relative concentration of lanosterol was observed and, thus, a buildup of the Erg11p substrate. Compounds 1, 2, 3, and 4 were analyzed using this metabolomic-profiling protocol to test their effects on lanosterol concentrations. Testing of compound 2, containing a pyridyl moiety, and compound 3, containing a pyrimidyl moiety, showed similar effects on the relative lanosterol concentration, as did voriconazole ( and ). In contrast, treating samples with compound 4, which contains a phenyl moiety but is otherwise identical to compounds 2 and 3, did not result in altered lanosterol concentrations compared to DMSO-treated control samples. The lack of an observable effect on the metabolomic profile caused by compound 4 is consistent with all the data shown above, suggesting that the compound is inactive on the tested fungal species despite high structure similarity to compounds 2 and 3. These results showing a buildup of lanosterol by the active compounds supports inhibition of Erg11p, the target proposed by the genetic-profiling methods, as the mechanism of action of these compounds.
Fig 3 Inhibition of Erg11p in S. cerevisiae cells led to accumulation of lanosterol, the substrate of Erg11p. (A) Chromatograms of ergosterol mass transition 379.3→69.2 (blue line) and lanosterol mass transition 409.4→191.2 (red line) after (more ...)
Statistical analysis of changes in lanosterol concentration compared to the control
Human CYP inhibition.
Besides targeting Erg11p, azoles also have an inhibitory effect on the human orthologue sterol 14-demethylase. The crystal structures of cytochrome P450 enzymes, including proteins in complex with azoles, have been determined (PDB codes 3JUS
, and 3L4D
). The crystal structures reveal that azoles are oriented toward the heme iron, interacting by free electron pairs of the nitrogen of the imidazole moiety.
Since the evidence so far suggests that the compounds in this study interact with Erg11p, we assessed whether they could also inhibit CYP3A4 in vitro despite the absence of a structural azole moiety. Compounds 1, 2, and 3 were indeed potent inhibitors of this enzyme, with IC50s of less than 0.5 μM. Compound 4 exhibited an IC50 of greater than 20 μM. This observation further strengthens the hypothesis that the mode of action of these compounds is similar to those of azole-containing compounds.
Macromolecular modeling supports a conserved azole binding site in humans and S. cerevisiae.
To further explore the experimental observation that compounds 1, 2, and 3 target cytochrome P450 enzymes in a manner similar to azoles, in silico docking was performed with compound 3. Compound 3 was selected because of its structural similarity to compounds 2 and 4 (differing only at the pyridine, pyrimidine, or phenyl moiety) and because it was the most potent compound overall in this study. Using the docking package GOLD, compound 3 was docked into the azole binding site (). The docking experiment provided five solutions with significant docking scores (). Interestingly, the docking solution with the highest GOLD score oriented one nitrogen in the pyrimidine ring toward the heme iron in a manner similar to that of azoles, allowing conjugation of the free electron pair with the heme iron. As the nitrogen of the pyrimidine ring moves away from the heme iron, disrupting this interaction, the GOLD score decreases. As a control, compound 4, which is structurally similar to compound 3 except that it lacks nitrogen in the phenyl moiety, was docked at the same site. None of the docking poses generated by GOLD oriented the phenyl ring toward the heme iron due to lack of a nitrogen atom. Thus, it appears that heme iron coordination with nitrogen in the phenyl moiety may be important for the activity of these compounds. This is in agreement with the lack of activity of compound 4 in the fungal proliferation assays.
Fig 4 In silico docking of Erg11p inhibitors into human CYP51 (PDB code 3LD6). Ketoconazole is represented by light yellow, compound 1 by purple, compound 2 by gray, and the heme cofactor by black. (A) Solution of the in silico docking approach with compound (more ...)
Scoring for the different poses of the compound 3 molecule represented as solutions
We decided to model the 3-dimensional structure of S. cerevisiae Erg11p based on the available crystal structures. Human CYP51 was selected as the template to model S. cerevisiae Erg11p. Based on the sequence alignment, there is 30% identity and 39% similarity between S. cerevisiae Erg11p and human CYP51 (see Fig. S2 in the supplemental material). As shown in Fig. S2 and S3C in the supplemental material, there are a number of conserved residues between CYP51 and the homology model, mostly around the heme-binding pocket, emphasizing the role of heme and its importance for interaction with ligands.
The data presented so far suggest that compounds 1, 2, and 3 target the S. cerevisiae Erg11p enzyme in a manner similar to that of azole-containing compounds. Point mutations in the target, and especially in the binding pocket, can affect binding of a chemical inhibitor and yield resistance. To test the hypothesis that these compounds act on Erg11p in a manner similar to that of azoles in a relevant fungal pathogen, we tested the activities of these compounds against two previously identified azole-resistant C. albicans strains. Each strain carries multiple point mutations in ERG11 () and was used to test susceptibility to Erg11p inhibition by elucidating the MICs for voriconazole and compound 1. The two strains with azole resistance-conferring point mutations showed a 30-fold and a 16-fold increase of the MICs against compound 1 and voriconazole, respectively (). With the exception of E266, all resistance-conferring sites are conserved in S. cerevisiae. Based on our homology model, the corresponding C. albicans azole resistance substitutions (F126L and Y132H) in S. cerevisiae (F134L and Y140H) are located in the azole binding pocket (see Fig. S3C in the supplemental material). In particular, C. albicans Y132 has been shown to be involved in direct interaction with ketoconazole. The observed cross-resistance with compound 1 suggests that in C. albicans, Erg11p is the likely efficacy target of compounds 1, 2, and 3 and that the compounds bind in the same binding pocket supported by the in silico docking approach demonstrated above.