Fluconazole-COX Inhibitor Hybrids: A Dual-Acting Class of Antifungal Azoles

When used in combination with azole antifungal drugs, cyclooxygenase (COX) inhibitors such as ibuprofen improve antifungal efficacy. We report the conjugation of a chiral antifungal azole pharmacophore to COX inhibitors and the evaluation of activity of 24 hybrids. Hybrids derived from ibuprofen and flurbiprofen were considerably more potent than fluconazole and comparable to voriconazole against a panel of Candida species. The potencies of hybrids composed of an S-configured azole pharmacophore were higher than those with an R-configured pharmacophore. Tolerance, defined as the ability of a subpopulation of cells to grow in the presence of the drug, to the hybrids was lower than to fluconazole and voriconazole. The hybrids were active against a mutant lacking CYP51, the target of azole drugs, indicating that these agents act via a dual mode of action. This study established that azole-COX inhibitor hybrids are a novel class of potent antifungals with clinical potential.


■ INTRODUCTION
Although humans and yeast have been evolving along different paths over a period of about a billion years, there is still a significant resemblance between the genomes of human and both friendly and pathogenic yeast. 1−5 Approximately onethird of the genes found in the human genome have counterparts in the genomes of yeast; amino acid sequences of the human proteome overlap by more than 30% with those of the yeast proteome. 6 Moreover, when 414 human genes were inserted into yeast cells one at a time, approximately 50% of them were found to be functional and facilitated the survival of the yeast cells. 7 It is, therefore, no wonder that, compared to the relative abundance of unique drug targets in bacteria, few such targets are suitable for selective inhibition of essential cellular processes in pathogenic fungi.
Prevention and treatment of fungal infections currently relies on a relatively limited number of antifungal drugs in only four major drug classes: azoles, echinocandins, allylamines, and polyenes. 8−10 The incidence of fungal infections has risen sharply in recent decades due to growing numbers of immunosuppressed persons and higher prevalence of drugresistant pathogenic fungi. 11,12 Global epidemics are increasingly being caused by drug-resistant (and multidrug-resistant) fungal pathogens, including Aspergillus fumigatus, Candida glabrata, Cryptococcus neoformans, 5,13−16 and, more recently, Candida auris, a pathogen with the potential for extensive multidrug resistance. 17−20 Notably, infections with drugresistant fungi are associated with mortality rates in the range of 50%, granting them high priority for new drug development. 21−24 An increasingly favored approach to rapidly overcome the shortage in fungal drug targets and drug classes is to enhance the efficacy of existing antifungal drugs through combination therapies. 25 To date, several FDA-approved drugs have been reported to synergize with antifungal drugs, including inhibitors of Hsp90, calcineurin, TOR, and PKC pathways, and drug efflux inhibitors. 5,26−29 Several clinically used nonsteroidal anti-inflammatory drugs that act by inhibiting cyclooxygenase (COX) enzymes, including ibuprofen, aspirin, and indomethacin, have been shown to possess moderate antifungal activity; the mechanism is unknown. 30−33 When used in combination with the most commonly used antifungal azole drug fluconazole (FLC, Scheme 1A), COX inhibitors significantly improve antifungal efficacy in vitro. 34,35 The antifungal efficacy of this type of combination was validated in animal models. 36,37 For example, ibuprofen was shown to effectively synergize with FLC against azole-resistant C. albicans. 34,38 A similar effect was observed for a combination of FLC and FK506, a 23-membered-ring macrolide immunosuppressant, that also acts as a broadspectrum inhibitor of pleiotropic drug resistance ATP-binding cassette transporters. 26,39,40 FLC-resistant isolates revert to FLC susceptible after incubation with ibuprofen yet retain high levels of expression of CDR1 and CDR2 efflux pumps. 41 It was shown that ibuprofen can alter the expression of the genes encoding the efflux pumps and that it may also act directly as an efflux pump blocker. 42,43 The arachidonic acid pathway has been associated with the yeast-to-hyphae morphogenesis in several species of Candida, 44,45 the most commonly diagnosed pathogens causing fungal-born infectious diseases in humans. 46,47 In mammals, nonsteroidal anti-inflammatory drugs such as COX inhibitors reduce the formation of prostaglandins generated via the arachidonic acid pathway. 48 Prostaglandins are involved in the morphogenesis and pathogenicity of yeast and mediate the host inflammatory response. 32,49 Prostaglandin E2 (PGE 2 ) regulates growth and colonization and promotes the formation of biofilms of several Candida species. 50,51 Several studies have shown that reduced PGE 2 production limits the virulence of pathogenic fungi, suggesting that the use of inhibitors of the arachidonic acid pathway could improve outcomes of fungal infections. 36,44,45 Physicians are reluctant to prescribe COX inhibitors to patients with infections due to their anti-inflammatory effects as these agents reduce the ability of the innate immune system to combat the pathogen. The efficacy of combination treatments heavily relies on the pharmacokinetic and pharmacodynamic properties of each of the drugs in the combination. 52 Moreover, COX-inhibiting drugs are known to induce gastrointestinal irritation. COX-1 is mainly responsible for mucus formation in the gastrointestinal tract and its inhibition is therefore blamed for inducing irritation. 53−56 These effects have been attributed to the carboxylic acid functionality that is common to all classical COX-inhibiting nonsteroidal anti-inflammatory drugs. 57,58 Ester and amide derivatives of these drugs maintain COX inhibition but cause less gastrointestinal problems, suggesting that the carboxylic acid group present in these drugs may not be required for COX inhibition. 59,60 Based on these observations, in this study, we sought to incorporate the antifungal properties of COX inhibitors with those of antifungal azoles by conjugating the amine-functionalized pharmacophore of FLC to different COX inhibitors via their carboxylic acid to form hybrid drugs. We report here on the synthesis and in vitro efficacies of dualacting antifungals composed of the pharmacophore of FLC and a collection of clinically used COX inhibitors.
inhibitor, we prepared racemic mixture 1a, the azide-functionalized pharmacophore of the first and second-generation antifungal azole drugs FLC and voriconazole (VOR) (Scheme 1A) as we previously reported. 61 Enantiomerically pure 1a-(S) and 1a-(R) were readily obtained by HPLC using a preparative amylose-based chiral resolution column (Scheme 1B). The azide-functionalized pharmacophores 1a-(S) and 1a-(R) were then subjected to catalytic hydrogenation to afford the corresponding amine-functionalized derivatives 1b-(R) and 1b-(S), respectively (Scheme 1B). The absolute configurations of the two amine-functionalized enantiomers of the azole pharmacophore were assigned by solving the X-ray structures of crystals of the two enantiomerically pure N-tosyl derivatives of the amine-functionalized derivatives 1b-(R) and 1b-(S), which readily crystalized from acetonitrile (Scheme 1C).
We generated 24 hybrids by forming an amide bond between the primary amine of the azole pharmacophore and the carboxylic acid of the COX inhibitor following the strategies described in Scheme 2A. Four of the COX inhibitors, ibuprofen, flurbiprofen, naproxen, and ketoprofen, contain a chiral center and were used for the generation of all four diastereomers of each hybrid (1−4, 5−8, 9−12, and 13−16, respectively, Scheme 2B). The achiral COX inhibitors niflumic acid, diflunisal, salicylic acid and diclofenac were used in the synthesis of enantiomeric azole pairs (17−24, respectively, Scheme 2B).
Hybrids 1−4 and 9−12 were prepared by coupling of an enantiomerically pure COX inhibitor to racemate 1b (Scheme 2A). Hybrids 5,6,13, and 14 were prepared by coupling the enantiomerically pure amine-functionalized azole pharmacophore 1b-(S) to a racemate of the COX inhibitors. Hybrids 17, 19, 21, and 23 were prepared by coupling the enantiomerically pure amine-functionalized azole pharmacophore 1b-(S) to achiral COX inhibitors. The same strategy was applied for the preparation of hybrids 7,8,15,16,18,20,22, and 24 from the enantiomerically pure amine-functionalized azole pharmacophore 1b-(R) (Scheme 2A). The purities of the 24 hybrids were determined by chiral semi-preparative HPLC column and confirmed to be ≥95% (Table S1, Figures S2−S25). The structures of the hybrids synthesized were verified using 1 H, 13 C, and 19 62 The panel included strains of C. albicans, C. glabrata, C. parapsilosis, C. tropicalis, C. guilliermondii, C. dubliniensis, and C. auris (for strains information see Table S2 in the Supporting Information). To evaluate the antifungal activity, we determined minimal inhibitory concentration 80% (MIC 80 ) values, which were defined as the lowest drug concentrations with turbidity (measured at OD 600 ) less than or equal to that of specific 1:5 dilutions of the growth control. As controls we tested FLC and VOR. MIC 80 values of the 24 hybrids and of the control azole drugs against the 16 Candida strains tested are summarized in Figure 1 and in Tables S3−S6 in the Supporting Information.
Of the 24 hybrids, three stood out as the most potent agents with the lowest MIC 80 values against all of the azolesusceptible strains in the panel: ibuprofen-based hybrids 1 and 2 and flurbiprofen-based hybrid 5 ( Figure 1). Of these three hybrids, ibuprofen-based azole 1 had the most potent activity against the majority of the azole-susceptible strains in the panel; this hybrid was up to two orders of magnitude more potent than FLC and was as potent as VOR.
In search of structure−activity relationships, we next analyzed the results of the antifungal activity tests in the context of the chiral center or centers of the hybrids. Our analysis revealed a clear connection between the absolute configuration of the chiral center at the benzylic carbon of the azole pharmacophore segment in both the diastereomeric tetrads and enantiomeric pairs. In all cases, the antifungal activity of hybrids with an S-configured benzylic carbon of the azole pharmacophore segment had higher potency than the corresponding hybrids with the R-configured center. Selected examples of two tetrads (ibuprofen-based 1−4 and flurbiprofen-based 5−8) and two enantiomeric pairs (niflumic acidbased 17 and 18 and diflunisal-based 19 and 20), which demonstrate the superior activity of the S-vs R-configured benzylic carbon against the azole pharmacophore are presented in Figure 2.
No general correlation could be made between antifungal potency and the chiral center of the COX inhibitor segments of the diastereomeric tetrads; rather, the results depended on the specific COX inhibitor. For example, hybrid 3 composed of R-configured ibuprofen was more potent than the corresponding S-configured ibuprofen hybrid 4. In the flurbiprofen tetrad, however, the S-configured flurbiprofen hybrid 5 was more potent than the corresponding R-configured ibuprofen hybrid 6 ( Figure 2). Of note, the chiral center of the azole pharmacophore markedly affected the antifungal activity of the hybrids, and the modest contribution of the chiral center of the COX inhibitor supports the hypothesis that the main target of these dual-acting antifungals is CYP51, the target of the azole class of antifungals. The investigation of the antifungal activity indicated that hybrids prepared by conjugation of the carboxylic acid of COX inhibitors to the amine-functionalized pharmacophore of the azole drug FLC can have markedly MIC 80 values were determined using the broth microdilution method over a concentration range of 0.003−64 μg/ mL. Orange circles represent C. albicans strains, yellow circles represent C. glabrata strains, and green circles represent C. parapsilosis, C. tropicalis, C. dubliniensis, and C. auris. Cells were grown in YPAD medium at 30°C (For C. auris strains 37°C) for 24 h. Each concentration was tested in triplicate, and the results were confirmed in at least two independent experiments. improved antifungal activity compared to that of FLC and comparable to that of the potent second-generation azole VOR.
Candida Tolerance to Azole-COX Inhibitor Hybrids Is Lower Than That to FLC and VOR. The majority of treatment failures for patients with invasive candidiasis are caused by apparently susceptible isolates. 63 For example, during a clinical trial on the treatment of invasive candidiasis, the drug anidulafungin, which belongs to the echinocandin class of antifungal drugs that act by inhibiting cell-wall formation, 64,65 was significantly superior to FLC, although the vast majority of isolates were susceptible to both drugs. 66 Apparently susceptible isolates resist antifungal drugs by exhibiting tolerance, defined as the ability of a subpopulation of cells to grow slowly at supra-MIC concentrations. Activation of tolerance mechanisms depends on stress response pathways. 67 Tolerance is, therefore, mechanistically distinct from resistance that relies upon mechanisms that are constantly under alert and do not require activation by stress response signals. Since the subpopulation exhibiting antifungal tolerance is usually characterized by slow growth, it becomes visually detectible after at least 48 h of growth in the presence of the drug, whereas resistance is generally evident after 24 h. 67 The level of tolerance varies between isolates presumably due to genetic differences, and even within a single genetic isolate, tolerance responses of individual cells may differ significantly. 67 Tolerance is thus the result of physiological or epigenetic differences rather than genetic variation. Clinical isolates that cause persistent infections and that fail to respond to a single course of FLC have higher intrinsic tolerance levels than those isolates that cause nonpersistent infections that are cleared with a single FLC course. 67 This suggests that measurement of tolerance may provide useful prognostic information and there is a need for development of drugs that are unaffected by tolerance. To investigate how tolerance is affected by the azole-COX inhibitor hybrids, we compared hybrids 1 and 5 to FLC and VOR in a disk diffusion assay. Tolerance was evaluated by comparing the zone of inhibition after 24 h to that after 48 h.
The assay was carried out on three representative strains: C. albicans, C. parapsilosis, and C. tropicalis (Figure 3).
After 48 h of incubation with FLC or VOR disks, the zones of inhibition that had appeared after 24 h of incubation in plates seeded with C. albicans SN152 or with C. parapsilosis ATCC 22019 were covered by drug tolerant colonies; the drug  tolerant subpopulation was smaller for C. tropicalis 660. All three tested strains displayed reduced tolerance to both hybrids 1 and 5 compared to the tolerance to FLC and VOR with the most pronounced effect observed in C. tropicalis 660 plates (Figure 3). No correlation could be made between MIC 80 values and the level of tolerance. For example, the MIC 80 values of 5, and VOR against C. parapsilosis 22019 were 0.5 μg/mL, and 0.015 μg/mL, respectively (Table S6), whereas the observed tolerance of this strain to hybrid 5 was lower than that to VOR (Figure 3). Since VOR acts predominantly by inhibition CYP51, this suggests that the observed reduced tolerance to the azole-COX inhibitor hybrids is not exclusively due to inhibition of CYP51 and that the antifungal effect of their COX inhibitor segment is likely responsible for the reduction in tolerance to these agents.
Dual-Acting Azole-COX Inhibitor Hybrids Act Predominantly by Inhibiting Ergosterol Biosynthesis. It is well established that clinically used azole antifungals including FLC and VOR act primarily by preventing ergosterol biosynthesis via inhibition of CYP51. 68,69 We asked if fungal growth inhibition by the dual-acting hybrids requires the presence of the ERG11 gene that encodes CYP51. 70 The antifungal activities of hybrids 1 and 5 and of FLC and VOR were determined against an erg3ΔΔ/erg11ΔΔ mutant C. albicans strain and against C. albicans SN152 from which this double knockout strain was derived (Table S2). The erg3ΔΔ/ erg11ΔΔ mutant is viable despite lacking CYP51, which is essential for aerobic growth unless ERG3, which encodes a C-5 sterol desaturase, is inactive. 71 Yeast growth was followed at OD 600 over 48 h in 96-well plates containing serial double dilutions of the tested hybrids. The results are summarized in Figure 4.
As expected, when CYP51 is not present, no significant effect on the growth of the double knockout mutant was observed for the entire range of concentrations of FLC. Modest reduction in growth was observed in wells treated with VOR at 64 μg/mL, the highest concentration tested, presumably due to nonspecific effects of the drug at this high concentration. In contrast, a clear dose-dependent reduction in growth was evident in wells containing hybrids 1 or 5. Dose-dependent growth reduction was also observed in the presence of free ibuprofen and flurbiprofen, from which hybrids 1 and 5, respectively, were derived. This supports that the CYP51-independent antifungal effect of the azole-COX inhibitor hybrids 1 and 5 results from their COX inhibitor segments. Of note, the MIC 80 values of hybrids 1 and 5 against the erg3ΔΔ/erg11ΔΔ mutant C. albicans strain were 64 μg/mL while FLC and VOR were inactive (Table S4). The MIC 80 values of these hybrids against C. albicans SN152, the parent strain of the erg3ΔΔ/erg11ΔΔ mutant were 0.003 μg/mL and 0.007 μg/mL, respectively (Table S4). The high MIC 80 values against the erg3ΔΔ/erg11ΔΔ mutant relative to those against the parent strain support our hypothesis that the contribution to the antifungal activity of the COX-inhibiting segment in these dual-acting agents is modest compared to that of the inhibition of CYP51.

■ CONCLUSIONS
It was previously established that nonsteroidal COX-inhibiting anti-inflammatory drugs and azole antifungals synergize to improve antifungal potency. Combination therapies can be affected by differences in pharmacological properties and by side effects of the drugs in the combination. With the goal of overcoming such potential limitations for combinations of azole antifungals and COX inhibitors, we synthesized a novel type of antifungals by linking an azole pharmacophore with a COX inhibitor to form a hybrid drug molecule. These hybrids were prepared by conjugation of a chiral azole pharmacophore to a collection of chiral and achiral COX inhibitors to form 24 chiral hybrids.
The antifungal activity profiles of the hybrids were tested against a diverse panel of Candida representing seven of the most encountered species of this common fungal pathogen and compared to the activities of the clinically used azole drugs FLC and VOR. The antifungal activities of several hybrids were superior to that of FLC. Two hybrids, ibuprofen-based 1 and flurbiprofen-based 5, stood out due to potency significantly higher than FLC and comparable to VOR. Structure−activity relationship analysis revealed that all hybrids with an S-configured azole pharmacophore were more potent antifungals than the corresponding hybrids with an R-configured azole pharmacophore. No such generalization could be made for the chiral COX inhibitors. In all hybrids with a chiral COX inhibitor, the contribution of the chiral center of the azole pharmacophore to the antifungal activity of the hybrids was markedly higher compared to that of the chiral center of the COX inhibitor.
Importantly, analysis of tolerance, defined as the ability of a subpopulation of cells to grow in the presence of the drug, revealed that yeast cultures were less likely to be tolerant in the presence of the hybrids 1 and 5 than in the presence of FLC and VOR. Clinical isolates with high tolerance are associated with persistent infections, suggesting that lower levels of tolerance to a drug may reduce the chances of the persistence and/or reoccurrence of the infection.
Mechanistic investigation revealed that unlike the clinically used FLC and VOR that target CYP51 as their main mode of action, hybrids 1 and 5 retained activity against an erg3ΔΔ/ erg11ΔΔ mutant C. albicans strain, which lacks CYP51. This activity was significantly lower, however, than the activity of these hybrids against the parent C. albicans strain from which the mutant lacking the target was derived. This indicates that the antifungal activity of these dual-acting hybrids results mainly from the inhibition of CYP51 yet, unlike FLC and VOR, the hybrids also act via a second mode of action contributed by the COX-inhibiting segment.
This study offers guidelines for development of potent antifungal agents that incorporate the antifungal activities of azole antifungals and COX inhibitors in hybrid molecules.
These new antifungals display potent antifungal activity and, importantly, reduced levels of tolerance. The dual-acting hybrids described here offer attractive leads for further clinical development.

■ EXPERIMENTAL SECTION
Chemistry. General Methods and Instrumentation. 1 H-NMR spectra (including one-dimensional total correlation spectroscopy (1D-TOCSY)) were recorded on BrukerAvance 400 or 500 MHz spectrometers, and chemical shifts (reported in ppm) were calibrated to CD 3 OD (δ = 3.31). 13 C-NMR spectra were recorded on BrukerAvance 400 or 500 MHz spectrometers at 100 or 125 MHz, respectively. 19 F-NMR spectra were recorded on BrukerAvance 400 or 500 MHz spectrometers at 375 or 470 MHz, respectively. Multiplicities are reported using the following abbreviations: s = singlet, d = doublet, t = triplet, q = quartet, dd = doublet of doublets, m = multiplet. Coupling constants (J) are given in Hz. Highresolution electrospray ionization (HRESI) mass spectra were measured on a Waters Synapt instrument. Chemical reactions were monitored by thin-layer chromatography (TLC) (Merck, Silica gel 60 F 254 ). Visualization was achieved using a cerium molybdate stain (5 g (NH 4 ) 2 Ce(NO 3 ) 6 , 120 g (NH 4 ) 6 Mo 7 O 24 ·4H 2 O, 80 mL H 2 SO 4 , 720 mL H 2 O) or with UV lamp. All chemicals, unless otherwise stated, were obtained from commercial sources. Reaction products were purified using Geduran Si 60 chromatography (Merck). The preparative reverse-phase high-pressure liquid chromatography (RP-HPLC) system used was an ECOM system equipped with a 5-μm, C-18 Phenomenex Luna Axia column (250 mm × 21.2 mm). The mobile phase was acetonitrile in H 2 O, and the gradient was from 10 to 90% acetonitrile. The flow rate was 20 mL/min. Chiral semipreparative high-pressure liquid chromatography (HPLC) used was performed on an ECOM system equipped with a 5-μm i-Amylose-3 Phenomenex Lux column (250 mm × 10 mm). The flow rate was 5 mL/min. Crystallographic Data. Deposition Numbers 2116277 and 2166299 contain the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre.
Azole-Flurbiprofen Hybrid 7 (110 mg, 90%). HRESI-MS m/z calculated for C 26     Journal of Medicinal Chemistry pubs.acs.org/jmc Article were confirmed by two independent sets of experiments. FLC and VOR were used as control drugs. Disk Diffusion Assay. Antifungal activities of select hybrids against C. albicans SN152, C. parapsilosis ATCC 22019, and C. tropicalis 660 were confirmed by the disk diffusion assay. Strains were streaked from frozen culture onto YPAD agar and incubated for 24 h at 30°C. Two or three colonies were placed into 1 mL of PBS solution, and OD 600 was determined with a TECAN Infinite. OD 600 was adjusted to 0.02 for C. albicans SN152 and to 0.025 for C. parapsilosis ATCC 22019 and C. tropicalis 660 by dilution with PBS. Aliquots of 200 μL of the diluted cultures of each strain were plated onto 15-mL casitone agar plates and spread using sterile beads (3 mm, Fisher Scientific). After the plates dried, a single disk (6-mm diameter, Becton Dickinson) with 25 μg of the hybrid being tested was placed in the center of each plate. Plates were then incubated at 30°C and photographed under the same imaging conditions after 24 and 48 h. FLC and VOR were used as control drugs.
Growth Curve Analyses. Growth curves were determined using the double-dilution method in 96-well plates. Starter cultures were streaked from glycerol stock onto YPAD agar plates and grown for 24 h at 30°C. Colonies were suspended in 1 mL of PBS and diluted to 1 × 10 −3 OD 600 and then diluted 1:100 into fresh medium. Hybrids dissolved in DMSO were added to YPAD broth (32 μL of stock solution in 1218 μL of YPAD broth), and serial double dilutions of hybrids in YPAD were prepared in flat-bottomed 96-well microplates (Corning) to enable testing of concentrations ranging from 64 to 1 μg/mL. Control wells with yeast cells but no-drug (100% growth) and blank wells containing only YPAD (0% growth) were prepared. An equal volume (100 μL) of yeast suspensions in YPAD broth was added to each well with the exceptions of the blank wells. Growth was determined at 30°C by measuring the OD 600 using a plate reader (Infinite M200 PRO, Tecan) every 40 min over 48 h. Each concentration was tested in triplicate, and results were confirmed by two independent sets of experiments. FLC and VOR were used as control drugs.