It was hypothesized that the natural products eugenol (EUG), β-escin, and curcumin (CUR) may act synergistically in combination. This was based on their related, reported mechanisms of action: in causing lipid peroxidation and disruption of cell membrane integrity (EUG), pore formation within the cell membrane (β-escin), and interactions with ABC drug transporters and ERG3 gene downregulation, leading to decreased membrane ergosterol and permeability (CUR) (Morton and Main, 2013; Moghadamtousi et al., 2014; Marchese et al., 2017). Checkerboard assays measuring growth of the model yeast Saccharomyces cerevisiae were used to assess synergy. These showed that pairwise combinations of EUG, β-escin, or CUR did not present any synergistic interaction: all three of the fractional inhibitory concentration (FIC) index values for these pairs of NPs were > 0.5 (Figure 1). In fact, both the EUG + β-escin and CUR + β-escin combinations exhibited antagonism (with FIC indices ≥ 2.5). The interaction between EUG and CUR was deemed additive (FIC index, 1.5). These observations were reproduced in two common laboratory strains of S. cerevisiae, W303 (Figure 1) and BY4743 (Supplementary Figure 1). The results did not support the starting hypothesis.

An organism’s microenvironment can be an important determinant of drug–drug interaction, with previous work reporting a shift from antagonism to synergism for certain antibiotic combinations when bacteria were incubated with glycerol or ethanol instead of glucose (Cokol et al., 2018). To test for similar effects on the stability of the combinatorial interactions among EUG, β-escin, and CUR, these were compared during yeast growth in different carbon sources. [These experiments focused on strain W303, as strain BY4743 has defects in the regulation of respiratory genes (Hon et al., 2005), affecting growth on respired substrates like glycerol and ethanol.] Growth of S. cerevisiae W303 was only slightly slower on the respiratory substrates than on glucose, which is fermentable (Figure 2A). Moreover, the antagonistic and additive combinatorial effects observed in glucose (Figure 1), were reproduced in the other carbon sources (Figure 2B and Supplementary Figure 2). The results indicate, for these combinations, that the carbon source and fermentative or respiratory metabolism were not important determinants of these NP–NP interactions.

To widen the net for enabling the discovery of novel NP antifungal synergies, the Puretitre NP compound library¹ was screened in combinations with the NPs of interest EUG, CUR, and β-escin. The Puretitre library was selected as the 200 NP compounds it comprises mostly have described use in traditional medicine, possessing both high bioactivity and relatively low toxicity so enhancing translational potential. An additional screen of the library in combination with the NP berberine (BER) is described, as BER had been identified as a hit compound in an initial screen of the library with EUG (described later), and there are previous reports of antimicrobial BER actions (Xu et al., 2009; Dhamgaye et al., 2014; Liu et al., 2015). For the combinatorial screening tests, the four selected NPs were supplied at concentrations that were very near sub-inhibitory (∼10% reduction of growth yield compared to no-drug control, see Supplementary Figure 3), to maximize sensitivity-of-detection of novel synergistic interactions. The library drugs were supplied at a largely sub-inhibitory concentration (100 μM) suggested by the manufacturer and consistent with similar previous work (Vallières et al., 2020b). The screens were performed in S. cerevisiae to maximize chances of identifying broad-spectrum antifungal hit combinations, as pathogens commonly have higher intrinsic resistance so offering lower sensitivity of detection of potential synergies. In total, screening of 800 different NP combinations in S. cerevisiae revealed 34 candidate interactions of interest (Figure 3 and Supplementary Table 1): these potential synergies (highlighted in color, Figure 3) were identified according to calculated effect strength > 50 (Vallières et al., 2018, 2020a), from the comparison of growth in the presence of each library compound either with or without the second NP (see Section “Materials and Methods”). The screens with EUG revealed six potentially synergistic interactions and with CUR, 11 hits, while no interactions of interest were found with β-escin (Figure 3). As mentioned earlier, one of the strongest interactions with EUG was with library-compound BER (detailed later) and a subsequent screen of the library in combination with BER suggested 17 potential synergies.

To corroborate synergistic interactions from the most promising screen combinations, checkerboard analysis was performed initially in S. cerevisiae. The three combinations with the greatest effect strengths from each of the three screens that indicated synergies were selected (Figure 4A). Synergy was corroborated for all nine of these combinations by checkerboard analysis (Supplementary Figure 4). The most promising checkerboard result for each of eugenol, curcumin, and berberine is displayed in Figure 4B, with the corresponding library-compound structures in Figure 4C. The other tested combinations gave FIC index values between 0.25 and ≤0.5 (Supplementary Figure 4) but were not pursued further in this study. For eugenol, the EUG + BER combination reduced the MICs for both compounds by up to 8-fold (Figure 4B). Checkerboard analysis for CUR with sclareol also indicated strong synergy, with MICs reduced by up to 4- and 8-fold, respectively, while the combination of BER with pterostilbene reduced these agents’ MICs by up to 16- and 8-fold, respectively.

To assess the wider efficacies of the three selected combinations, they were tested against fungi that are either important human pathogens, including drug-resistant isolates (Candida albicans, C. glabrata, A. fumigatus); phytopathogens (B. cinerea, Z. tritici); or a food spoilage organism (Zygosaccharomyces bailli). The combinatorial effects determined from checkerboard analysis with this wider range of fungi revealed synergy in all cases, with FIC index values ranging between 0.19 and 0.5, the lowest values suggesting the strongest synergy was with EUG + BER (Table 1). Synergies were also retained in drug- (azole-) resistant isolates of C. albicans and A. fumigatus. In addition, the EUG + BER combination retained synergistic activity against biofilms of C. albicans, showing a significantly greater observed effect of the combination than would be expected from an additive interaction of the observed individual-NP effects (Supplementary Figure 5A). This evidence supported our initial use of NP combinatorial screening (with S. cerevisiae) to find novel NP synergies that had broad antifungal spectra.

As the BER + EUG combination gave the strongest overall synergy it was chosen for further investigation of the mechanism of synergistic action. We hypothesized that the synergy between these agents was centered on mitochondria as a target, as previous studies have indicated that each agent causes mitochondrial perturbation [and associated reactive oxygen species (ROS) production] as a mechanism of cell inhibition (Dhamgaye et al., 2014; Alves et al., 2017). Initially, a rho⁰ mutant was tested to assess whether the loss of mitochondrial DNA (mtDNA) could rescue the synergy between BER and EUG. Synergy was retained in the rho⁰ mutant but was weakened with a FIC index of 0.47 compared with 0.25 in the parental wild type (p = 0.028) (Supplementary Figure 6). As the result supported some role for mitochondria in the synergy, we focused next on specific mitochondrial functions. Mitochondrial membrane depolarization by both EUG and BER, individually, has been reported previously (da Silva et al., 2016; Alves et al., 2017). Furthermore, polarity is partly retained in the absence of mtDNA (Appleby et al., 1999; Vowinckel et al., 2021), consistent with the partial-retention of synergy in the rho⁰ mutant shown above. Therefore, we probed mitochondrial membrane depolarization (MMD), using flow cytometric analysis of decreased retention of rhodamine 123 fluorescence by cells, at doses of EUG + BER combinations giving approximately 10, 25, and 50% growth inhibition (Figure 5A). Microscopic observation confirmed cell staining with rhodamine-123 and suggested decreased fluorescence in cells treated with EUG + BER (Figure 5B). The flow cytometric data (presented for the intermediate dosage condition in Figure 5C) corroborated that either NP alone produced decreases in rhodamine 123 stainings. Moreover, the effect was markedly stronger when the NPs were combined. To indicate whether combinatorial effects on retention of the reporter were additive or synergistic, the data were normalized to median fluorescence as percentages of the no-drug controls (see Section “Materials and Methods”). This enabled comparison of the outcome to be “expected” if the combinatorial effect was additive—calculated by multiplying the two fluorescence-medians (% vs. control) obtained for each NP individually—with the observed outcome for the combination (Figure 5D). In all combination concentrations tested (Figure 5A), the observed effects were significantly greater than those which would be expected from additive interactions, suggesting that EUG + BER caused synergistic MMD (Figure 5D). A similar experiment was performed with C. albicans, and showed synergistic MMD also for this yeast pathogen (at a combination concentration giving ∼40% growth inhibition) (Supplementary Figures 5B–D). As effects were evident at combination doses including those near sub-inhibitory to growth, it implied that synergistic MMD was not a non-specific consequence of synergistic growth inhibition but consistent with a mechanistic role for MMD in the antifungal synergy between EUG and BER.

To further elucidate the mechanism of EUG + BER synergy, we focused on the reported effects of ROS in the individual actions of both these agents (Khan et al., 2011; Dhamgaye et al., 2014). To do this we chose to make use of antioxidant molecules and yeast gene deletants with well-characterized properties, rather than fluorescent ROS probes which can be less specific (Winterbourn, 2014). First, the effects of added antioxidants on the synergy were evaluated. Both glutathione and L-ascorbic acid significantly increased the FIC index value, beyond the threshold (≤0.5) that normally encompasses synergy (Figure 6; checkerboard data are shown in Supplementary Figure 7). A similar effect was also evident for the mitochondria-targeted antioxidant, mitoquinol, where suppression of synergy appeared to be slightly stronger than for glutathione and L-ascorbic acid. The results supported the hypothesis that ROS are important for establishing the synergy between EUG + BER.

In keeping with the earlier depolarization results for mitochondrial membrane function, and the indication that (mitochondrial) ROS production is important for the synergy, next we focused specifically on the involvement of mitochondrial ROS in the EUG + BER interaction. Deletion mutants defective for three mitochondrial antioxidant functions were tested: Δsod2 cells lacking the mitochondrial superoxide dismutase that scavenges mitochondrial superoxide; Δogg1 cells lacking a DNA glycosylase that protects the mitochondrial genome from oxidative damage; and Δccp1 cells lacking the mitochondrial cytochrome C peroxidase. The Δogg1 and Δccp1 deletants exhibited FIC indices with EUG + BER that were not significantly different from that of the wild type (Figure 7). In contrast, synergy was accentuated in cells defective for Sod2p, reflected by an FIC index in the mutant that was ∼50% of that in the wild type (Figure 7). This evidence for partial suppression of the synergy by Sod2p (in wild-type cells) further suggests that ROS, more specifically, mitochondrial superoxide is important for the mechanism of synergistic eugenol–berberine action.

The principal yeast strain backgrounds were S. cerevisiae W303 (MATa/MATα leu2-3,112 trp1-1 can1-100 ura3-1 ade2-1 his3-11,15 [phi+]) and BY4743 (MATa/MATα his3-1/his3-1 leu2-0/leu2-0 met15-0/MET15 LYS2/lys2-0 ura3-0/ura3-0). A rho⁰ mutant was derived from S. cerevisiae BY4743 by sub-culturing three times on YPD agar supplemented with 40 μg mL^–1 ethidium bromide. It was confirmed that the resultant mutant was defective for respiratory growth by a lack of growth on YPG agar (recipe as for YPD agar, below, but with glycerol replacing glucose). The sod2Δ, ogg1Δ, and ccp1Δ homozygous diploid deletants (obtained from Euroscarf, Germany), were in the BY4743 background. Other yeast species used in this study included Z. bailli strain NCYC1766, Candida glabrata BG2, Candida albicans SC5314, and an azole-resistant isolate, C. albicans J942148, kindly provided by Carol Munro and Donna MacCallum (University of Aberdeen, United Kingdom). Yeasts were maintained and grown at 30°C (S. cerevisiae) or 37°C (C. albicans and C. glabrata) in YPD broth [2% peptone (Oxoid, Basingstoke, United Kingdom), 1% yeast extract (Oxoid), 2% D-glucose]. For experimental purposes, the yeasts were streaked onto YPD agar [recipe as for YPD broth but with the inclusion of 1.5% agar] from −80°C glycerol stocks and cultured for at least 48 h before single colonies were picked for sub-culture to the broth as described below.

Filamentous fungi used in the study were A. fumigatus CBS 144.89 and an azole-resistant isolate A. fumigatus 3216 (kindly provided by Matthias Brock, University of Nottingham, United Kingdom), Z. tritici K4418 (kindly provided by Syngenta, United Kingdom) and B. cinerea SAR109940. The filamentous fungi were routinely maintained and grown either on Aspergillus complete medium (ACM) at 37°C for A. fumigatus, or Potato Dextrose Agar (PDA, Oxoid) or Potato Dextrose Broth (PDB, Sigma) at room temperature for Z. tritici and B. cinerea (Vallières et al., 2018). Strains of A. fumigatus from −80°C glycerol stocks were grown on PDA slopes for 72 h at 37°C before spores were then harvested for use in experiments. Z. tritici and B. cinerea were cultured for 7 days from −80°C glycerol stocks prior to harvesting pycnidia for experimental use. Where necessary, media were solidified with 1.5% agar (Sigma, United Kingdom).

Eugenol, β-escin, curcumin, berberine hydrochloride, sclareol, capsaicin, parthenolide, ellagic acid, glutathione, L-ascorbic acid were from Sigma–Aldrich (United Kingdom); osthole and pterostilbene were from Stratech (United Kingdom) and mitoquinol from Cayman Chemical Company (United Kingdom); all other NPs were components of the Puretitre natural compound library from Caithness Biotechnologies (United Kingdom). All of the above except eugenol (70% ethanol), glutathione, and L-ascorbic acid (dH₂O) were dissolved in dimethyl sulfoxide (DMSO, Sigma United Kingdom) and added to growth media from the following stock solutions prepared in those solvents: eugenol, 500 mM; glutathione, 375 mM, L-ascorbic acid, 500 mM; osthole, 200 mM; pterostilbene, 200 mM; β-escin, 50 mM; curcumin, 50 mM; berberine hydrochloride, 200 mM; sclareol, 130 mM; capsaicin, 200 mM; parthenolide, 130 mM; ellagic acid, 33.3 mM, mitoquinol, 2.94 mM.

All culturing and preparation for checkerboard assays adhered to EUCAST guidelines, except for the use of YPD broth, ACM, or PDB instead of RPMI as medium (Arendrup et al., 2012). Briefly, for yeasts, overnight cultures in YPD broth, 120 rev. min^–1, 30°C or 37°C (see above), derived from single colonies, were diluted in the morning to OD₆₀₀ 0.5 then grown in YPD for an additional 4 h followed by dilution to OD₆₀₀ 0.1 (S. cerevisiae) or 0.01 (Candida spp.), before use as experimental cell suspensions in assays. For filamentous fungi, spores were inoculated from PDA plates into ACM broth at 10⁵ spores mL^–1 (A. fumigatus) or into PDB at 10⁴ spores mL^–1 (Z. tritici and B. cinerea). Aliquots (50 μL) of these cell or spore suspensions were transferred to flat-bottom 96-well microtiter plates (Greiner Bio-One; Stonehouse, United Kingdom) with compounds added to specified final concentrations by two-fold serial dilution. The inoculated plates were incubated statically for 24 h at 30°C for S. cerevisiae and Z. bailli, 37°C for Candida spp., 48 h at 37°C for A. fumigatus and 7-days at room temperature for Z. tritici and B. cinerea. Subsequently, OD₆₀₀ was determined with a BioTek EL800 microplate spectrophotometer. Fractional inhibitory concentration (FIC) indices were used to assess potential synergy from the checkerboard results, calculated as: [(compound 1 MIC in combination)/(compound 1 MIC alone)] + [(compound 2 MIC in combination)/(compound 2 MIC alone)] (Hsieh et al., 1993). For continuous growth measurements in the presence of NPs, broth cultures of S. cerevisiae were cultivated in 96-well microplates within a BioTek Powerwave XS microplate spectrophotometer (Vallières et al., 2018), in YP (2% peptone, 1% yeast extract) broth supplemented with either 2% D-glucose, glycerol, or ethanol. The growth was monitored from a starting OD₆₀₀ ∼0.1 with continuous shaking at 30°C for 24 h, with OD₆₀₀ readings taken every 30 min.

Biofilm metabolic activity was measured using the XTT (tetrazolium salt, 2,3-bis[2-methyloxy-4nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide) (Sigma, United Kingdom) reduction assay and performed as described previously (Vallières et al., 2020b). Briefly, overnight C. albicans cultures were diluted to OD₆₀₀ ∼0.01 in RPMI 1640 medium and 100 μL aliquots transferred to 96-well microtiter plates (Greiner Bio-One) and placed at 37°C. After 2 h non-adherent cells were removed by washing with PBS and plates were incubated at 37°C for 24 h in fresh medium. Biofilms were then washed again with PBS and eugenol and berberine added at specified concentrations or omitted for controls. Cultures were incubated for a further 24 h, then the biofilm was washed and the XTT reaction was performed using 210 μg mL^–1 XTT and 4.2 μM menadione. Biofilm metabolic activity was measured after 3 h incubation at 490 nm using a BioTek El800 microplate spectrophotometer. The assay was performed in biological triplicate. “Expected values” for combinations were calculated by multiplication of the % biofilm activity determinations for the corresponding individual-compound effects; these were compared with experimental values obtained for the combinations, with statistical analysis by paired t-test.

For the initial determination of sub-inhibitory concentrations (SIC) of selected test compounds (for subsequent use in the high throughput screen), experimental cell suspensions of S. cerevisiae W303 in YPD were prepared from overnight cultures as described above. Aliquots (50 μL at OD₆₀₀ 0.2) were mixed with 50 μL of YPD containing the relevant NP, from 2× solutions of the test compounds. In all conditions including solvent-matched controls, solvent concentrations were <1% of the final assay volume. Subsequent growth was measured by OD₆₀₀ determination with a BioTek EL800 microplate spectrophotometer, after 24 h static incubation at 30°C.

For high-throughput screens, the four test compounds at their SIC (750 μM eugenol, 12.5 μM β-escin, 350 μM berberine, 50 μM curcumin) were assayed in pairwise combinations against the Puretitre natural compound library (Caithness Biotechnologies, United Kingdom), comprising 200 chemicals at 10 mM, dissolved in DMSO. For the screens, aliquots (1 μL) of each library compound were combined with 49 μL YPD and added to 96-well microtiter plates (Greiner Bio-One). Aliquots (50 μL) of yeast cell suspension (prepared as described above) containing one of the four test compounds (added at double the final desired SIC concentration, see above) were added to the 50 μL library-compound preparations in the microtiter plates. This gave final concentrations of 100 μM of each library compound in 100 μL total per well. Solvent-matched controls at 0.35% DMSO or 0.3% ethanol (70%) were used for control assays without added compounds. Subsequent growth was measured according to OD₆₀₀ determinations with a BioTek EL800 microplate spectrophotometer after 12 h and 24 h static incubation at 30°C. OD₆₀₀ from growth with added compounds was expressed as a percentage of control growth without the compounds. Effect strength [(% growth with library compound) - (% growth with library compound + test compound)] was calculated for each combination; screen “hits” were considered as those combinations showing an effect strength > 50, as described previously (Vallières et al., 2020b). Screens were performed in duplicate.

Depolarization of the mitochondrial membrane in yeast cells was determined according to rhodamine 123 dye retention using a method adapted from previous reports (da Silva et al., 2016; Alves et al., 2017). After 24 h of exposure to berberine and/or eugenol in checkerboard format, yeast cell suspensions were removed from 96-well plates, spun down in 1.5 mL microcentrifuge tubes at 1,900 × g for 3.5 min, then washed with and resuspended in 250 μL phosphate-buffered saline (PBS) before supplementation with a final concentration of 20 μg mL^–1 rhodamine 123 (Sigma, United Kingdom) and incubation for 30 min at 30°C in the dark. After incubation, cells were washed once with PBS and resuspended in 500 μL PBS then transferred to 5 mL falcon tubes [Becton Dickinson (BD), United Kingdom]. A BD FACS Canto A flow cytometer (blue filter; excitation at 488 nm, emission at 530 nm) was used to determine the fluorescence intensity of cells. A total of 20,000 cells were evaluated per sample and each condition was assessed in technical triplicate for each independent experiment (n = 4). Cellular debris was gated out from the analysis using Kaluza software. Median fluorescence intensity (MFI) values for cells treated with NP compounds were transformed to percentages relative to the MFI for minus-compound control cells. “Expected values” for % MFI and statistical comparison with experimental values were calculated as described above for biofilm inhibition.

Visualization of S. cerevisiae cells was performed after 24 h treatment with the EUG + BER combination, rhodamine-123 staining, and washing as described above. Cells were resuspended in 50 μL PBS before mounting. A GXML3201LED fluorescence microscope equipped with a GX-CAM controlled by GXCapture software (GX microscopes, Stansfield, United Kingdom) was used to collect images via the FITC filter (excitation, 495 nm; emission, 519 nm) using a ×40 objective lens.