A Human IRE1 Inhibitor Blocks the Unfolded Protein Response in the Pathogenic Fungus Aspergillus fumigatus and Suggests Noncanonical Functions within the Pathway

The unfolded protein response (UPR) is a signaling pathway that maintains endoplasmic reticulum (ER) homeostasis, with functions that overlap virulence mechanisms in the human-pathogenic mold Aspergillus fumigatus. The canonical pathway centers on HacA, its master transcriptional regulator. Translation of this protein requires the removal of an unconventional intron from the cytoplasmic mRNA of the hacA gene, which is achieved by an RNase domain located in the ER-transmembrane stress sensor IreA. Here, we show that targeting this RNase activity with a small-molecule inhibitor effectively blocked UPR activation, resulting in effects that mirror the consequences of genetic deletion of the RNase domain. However, these phenotypes were surprisingly narrow in scope relative to those associated with a complete deletion of the hacA gene. These findings expand the understanding of UPR signaling in this species by supporting the existence of noncanonical functions for the unspliced hacA mRNA in ER stress response.

several species that infect plants, animals, or humans have uniquely adapted this stress response pathway to support the expression of pathogenicity traits, such as antifungal drug resistance, nutrient acquisition, host-temperature adaptation, cell wall or membrane homeostasis, and effector protein secretion (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12). The pathway detects and responds to the accumulation of unfolded proteins, which is a situation that arises whenever the demand for increased output by the secretory pathway exceeds the protein folding capacity of ER-resident chaperones and folding enzymes (13). When unfolded proteins reach a critical threshold in the ER lumen, they activate an ER transmembrane protein known in humans as IRE1 (IreA in Aspergillus fumigatus [ Fig. 1]). IRE1/IreA directs the most evolutionarily conserved branch of the UPR, and the basic paradigm of signaling is conserved across most fungal species, including A. fumigatus (14).
All species homologs of the IRE1 protein contain an ER luminal stress-sensing domain and a cytosolic effector region containing kinase and endoribonuclease (RNase) domains (15,16). Interactions between the sensor and unfolded proteins trigger oligomerization in the ER membrane, resulting in trans-autophosphorylation and activation of the RNase domain. The RNase then removes an unconventional intron from a cytoplasmic mRNA that is known in A. fumigatus as hacA u (uninduced), creating a frameshift in the activated form of the mRNA, hacA i (induced), that is an obligatory step The UPR is activated by ER stress, which occurs when the demand for secretion exceeds the protein folding capacity of the ER or when the cell encounters stimuli that disrupt ER homeostasis. The resulting accumulation of unfolded proteins activates the cytosolic RNase domain of the ER transmembrane stress sensor IreA, which removes an intron from the uninduced mRNA hacA u , converting it into the induced mRNA hacA i , which is translated into the bZIP transcription factor HacA that coordinates a transcriptional program to augment ER folding capacity. The small-molecule inhibitor 48C blocks the activity of the IreA RNase domain required for hacA mRNA processing. (B) Alignment of the RNase domain of human IRE1 with the corresponding region in A. fumigatus, C. albicans, and C. neoformans. The arrow indicates the conserved lysine residue targeted by 48C. Black shading indicates identity to the column consensus and gray indicates similarity.
for the translation of the bZIP transcription factor HacA (1,3,5) (Fig. 1A). This master transcriptional regulator of the canonical UPR pathway helps to buffer physiological fluctuations in ER stress by orchestrating a transcriptional program that augments the folding of proteins trafficking through the secretory pathway (3,13,17).
Selective inhibitors of the kinase or RNase domain of human IRE1 have been developed as tools to experimentally dissect the contribution of the UPR to cell physiology and human disease, particularly with respect to cancer (18)(19)(20)(21)(22)(23)(24). Among these compounds, salicylaldehyde-based inhibitors have shown high efficacy by directly blocking the RNase domain (25). The compounds 48C and STF-083010 are two well-described examples of this class (26)(27)(28)(29), but their effects on a fungal pathogen are still unexplored. Here, we demonstrate that 48C (30) is an inhibitor of the canonical UPR pathway in A. fumigatus, effectively blocking the accumulation of hacA i mRNA and downstream target gene induction. We found that treatment with 48C failed to recapitulate the entire collection of phenotypes associated with a ΔhacA deletion mutant but was similar to phenotypes displayed by an IreA RNase domain mutant that is unable to process hacA u mRNA into hacA i . These data demonstrate the feasibility of pharmacological disruption of the canonical UPR pathway in a fungal pathogen and support emerging evidence that both the unspliced and spliced hacA mRNAs have functions in ER stress responses.

RESULTS
The human IRE1 RNase inhibitor 48C blocks the canonical UPR of A. fumigatus. We previously showed that the canonical UPR pathway of A. fumigatus involves a linear order of molecular events triggered by the activation of IreA in the presence of ER stress (3). Upon activation, the endoribonuclease domain of IreA splices an intron from the uninduced cytoplasmic mRNA hacA u , converting it to its induced form, hacA i , which is subsequently translated into the encoded transcription factor HacA (Fig. 1A). Subsequent studies revealed that deleting the ireA gene, or the ortholog in other fungal species (5,31), is more deleterious than deleting the gene encoding the downstream transcription factor, suggesting that additional branches of the pathway emanating from IreA contribute to the ER stress response independently of HacA (Fig. 1A). To further examine the possibility of more complexity in the pathway, we focused in this study on the RNase domain, incorporating both small-molecule and genetic inhibition approaches.
The synthetic coumarin derivative 8-formyl-7-hydroxy-4-methylcoumarin (abbreviated as 48C) was initially identified in a high-throughput screen for selective inhibitors of the human IRE1 RNase (30). The molecule forms a Schiff base with a lysine residue in the active site of the RNase domain, which prevents IRE1 from splicing its target mRNA (XBP1 u in humans). Since the homologous lysine is conserved in A. fumigatus IreA (Fig. 1B), we hypothesized that 48C could be used to modulate the processing of hacA u mRNA into its activated form, hacA i (Fig. 1A). To test this, we used quantitative reverse transcription-PCR (RT-qPCR) analysis to monitor the induction of hacA i levels during acute ER stress. Dithiothreitol (DTT) is the most widely used approach to experimentally induce ER stress because it reduces the disulfide bridges that stabilize many proteins in the secretory pathway, resulting in a rapid increase in the level of unfolded proteins that activate the IRE1/IreA sensor (32). Consistent with previous reports (33), DTT treatment of A. fumigatus hyphae induced accumulation of hacA i mRNA, reflecting activation of the IreA RNase domain ( Fig. 2A). Since the hacA gene promoter is a target of the HacA transcription factor (34), this surge in hacA i mRNA levels in the presence of DTT is due to a positive feedback loop that involves three linked events: first, initial activation of the IreA RNase converts basally expressed hacA u mRNA into hacA i ; second, translation of hacA i mRNA into the HacA transcription factor increases transcription from the hacA gene promoter itself, thereby replenishing the pool of hacA u mRNA; and third, splicing of hacA u into hacA i continues as long as the IreA RNase remains activated (5,35). We found that incorporation of 48C into the medium prevented DTT from triggering an increase in hacA i mRNA abundance, dem-onstrating that the canonical UPR is not responsive to an acute ER stress stimulus in the presence of this compound. The lowest concentration for blocking the UPR under these conditions was determined to be 10 mg/liter (see Fig. S1 in the supplemental material), so this concentration was used for all subsequent experiments in this study.
We have previously reported that low levels of hacA i are also detectable in the absence of added ER stress, representing a basal UPR that functions to buffer fluctuations in protein folding that arise during normal vegetative growth (5). We found that treatment with 48C also reduced the steady-state levels of hacA i mRNA under these conditions (Fig. 2B), consistent with the ability of the compound to impair basal IreA activation. In contrast, expression of the sod1 gene, which does not depend on the UPR for its expression (36), was unaffected by 48C treatment (Fig. 2B), suggesting that downregulation of hacA i mRNA levels by 48C is not due to a general downregulation of transcription. Moreover, since the effects of DTT on cell physiology are not limited to ER protein folding (37,38), we also compared the effects of 48C on the expression of Afu8g05720, representing a gene that we have previously shown to be induced by DTT independently of the UPR (5). As shown in Fig. 2C, 48C did not block the induction of this gene by DTT, illustrating specificity of 48C treatment for the inhibition of UPR target genes.
Since 48C effectively blocks hacA i induction during acute ER stress ( Fig. 2A), the compound should also prevent the upregulation of UPR target genes that are under the control of the HacA transcription factor. To test this, we compared the expression levels of three genes that are established UPR targets across multiple species: the Hsp70 chaperone gene bipA, the protein disulfide isomerase gene pdiA, and the oxidoreductase gene eroA (34,39). As expected, all three of these UPR markers were strongly induced by DTT in A. fumigatus (Fig. 2D). However, no induction was detected in the presence of 48C, consistent with the ability of this compound to prevent the accumulation of hacA i mRNA (Fig. 2D). The compound also blocked the ability of DTT to induce the expression of srcA and pmrA (Fig. S2A), encoding ER/Golgi P-type Ca 2ϩ ATPases that we have recently identified as targets of the UPR in A. fumigatus (33). Interestingly, we found that a ΔsrcA/ΔpmrA double deletion mutant was hypersensitive to the effects of 48C relative to its parental strain (Fig. S2B). Since the absence of these Ca 2ϩ ATPases exacerbates ER stress (33), the increased sensitivity of this mutant to 48C is likely to be due to the ability of the compound to prevent the canonical UPR from mounting an adaptive response to low Ca 2ϩ levels in the secretory pathway caused by the absence of these Ca 2ϩ pumps. We conclude that 48C is able to enter the fungus and reach the IreA sensor, thereby inhibiting the RNase activity that is required for hacA u mRNA processing and downstream activation of UPR target genes.
48C impairs growth on a polymeric protein substrate. Mutants of A. fumigatus that harbor deletions of the hacA gene grow relatively well on medium containing simple carbon and nitrogen sources (3). Similarly, we found that the growth rate of the A. fumigatus parental strain KU70 on solid minimal medium supplemented with 48C was only slightly reduced, comparable in magnitude to the mild growth inhibition displayed by a KU70-derived mutant lacking the hacA gene (Fig. 3A). In liquid culture, both KU70 and the unmodified wild-type strain CEA10 showed only a 10% reduction in metabolic activity in the presence of 48C (Fig. 3B), making it unlikely that the inhibitory effect of 48C on UPR gene expression is due to broad inhibition of fungal metabolism. In addition, although a slight reduction in spore germination rate was observed in the presence of 48C, the total biomass accumulation after 24 h of incubation in liquid culture was indistinguishable in the presence or absence of the compound (Fig. S3). We conclude that pharmacological inhibition of the IreA RNase domain has minimal growth-impairing effects on A. fumigatus, consistent with the relatively normal growth characteristics of the ΔhacA mutant in the absence of stress.
The ability of 48C to block the induction of hacA i mRNA and downstream UPR target gene activation under ER stress predicted that treatment with this compound would have similar effects as deleting the hacA gene under conditions that require UPR activity. Filamentous fungi that lack the hacA gene grow poorly on polymeric substrates because their secretory pathways are unable to meet the demand for hydrolytic enzyme secretion (3,9,40,41). To determine how 48C would impact utilization of a complex protein substrate, conidia were inoculated into a solution of type I collagen as the only source of carbon and nitrogen. Although both KU70 and CEA10 isolates grew well on this medium, the inclusion of 48C strongly impaired growth (Fig. 3C). The ΔhacA mutant also grew poorly on this substrate, indicating that loss of the canonical UPR, either by genetic deletion of hacA or by treatment with 48C, impairs the ability to grow on a protein substrate that requires enzymatic breakdown by secreted hydrolases prior to absorption. 48C increases sensitivity to the antimicrobial agents carvacrol and hygromycin. The inability of A. fumigatus to activate the canonical UPR in the presence of 48C suggests that the fungus would be hypersensitive to a drug that causes ER stress. To test this, the effects of 48C on growth were compared in the presence of carvacrol, a plant-derived compound that disrupts ER homeostasis in Candida albicans and triggers UPR intervention (42). As shown in Fig. 4A, carvacrol induced accumulation of the hacA i mRNA in the CEA10 strain, confirming that this compound has similar adverse effects on ER homeostasis in A. fumigatus as it does in C. albicans. In addition, the ΔhacA mutant was hypersensitive to carvacrol relative to its parental strain KU70, indicating a role for the canonical UPR in protecting against carvacrol-induced ER stress (Fig. 4B). We found that treatment with 48C prevented carvacrol-induced hacA i accumulation (Fig. 4A), which was associated with increased susceptibility to carvacrol in both the KU70 and CEA10 backgrounds (Fig. 4B). An isobolographic analysis (43) in liquid culture confirmed that the combination of 48C and carvacrol treatment had synergistic toxicity toward the fungus (Fig. 4C). We conclude that preventing the canonical UPR from responding to carvacrol-induced ER stress, either by chemical treatment with 48C or by genetic deletion of hacA, is toxic to A. fumigatus.
Other plant-derived compounds that are structurally related to carvacrol have been shown to enhance the toxicity of drugs that target translation fidelity, such as aminoglycosides (44). Aminoglycosides are a therapeutically important class of antimicrobials that reduce the discrimination between cognate and near-cognate tRNAs, causing the incorporation of incorrect amino acids that promotes the accumulation of toxic protein aggregates (45). As shown in Fig. 4A, the aminoglycoside hygromycin B (HygB) induced hacA i in A. fumigatus, which was blocked by treatment with 48C (Fig. 4A), suggesting a role for the UPR in countering protein folding stress associated with HygB-induced translational errors. In support of this, we found that the ΔhacA mutant was hypersensitive to HygB and that treatment of either KU70 or CEA10 strains with 48C increased sensitivity to this compound (Fig. 4B). An alternative explanation for this increased HygB susceptibility stems from the observation that yeast glycosylation mutants are hypersensitive to this aminoglycoside, a phenomenon that is attributed to one or more HygB resistance proteins that require glycosylation for optimal activity (46). Our previous demonstration that the ΔhacA mutant is deficient in the expression of genes involved in N-glycan biosynthesis (5) would be consistent with a glycosylation defect that enhances HygB sensitivity. This predicts that UPR inhibition by 48C would also reduce the expression of glycosylation-related genes. To test this, we compared the expression levels of glycosylation genes that were previously shown to be downregulated in the ΔhacA mutant (5): msdS (encoding an ␣-mannosidase) (47), the DTTinducible gene Afu5g08970 (WBP1 in Saccharomyces cerevisiae, an oligosaccharyl transferase gene), and Afu5g03500 (ROT2 in S. cerevisiae, a glucosidase I gene). Similarly to ΔhacA (Fig. S4), all three genes were downregulated in A. fumigatus after treatment with 48C as demonstrated by RT-qPCR analysis (Fig. 4D). As expected, the induction of WBP1 and ROT1 upon ER stress with 1 mM DTT was blocked in the presence of the inhibitor (Fig. 4D). We conclude that the canonical UPR is an integral part of the cellular response to HygB in A. fumigatus and that preventing UPR activation, either by treatment with 48C or by genetic deletion of hacA, enhances the toxicity of this aminoglycoside against the fungus.
48C reveals unexpected complexity in the canonical UPR pathway. The ΔhacA mutant of A. fumigatus exhibits increased sensitivity to both the triazole and echinocandin classes of antifungals, as well as to the ER stress agents DTT and tunicamycin (3). Although blocking hacA i mRNA induction by 48C treatment increased carvacrol and hygromycin B sensitivity, we were surprised to find that the compound had minimal effects on the sensitivity of the fungus to DTT, tunicamycin, or itraconazole (Fig. 5A). A fundamental difference between the ΔhacA mutant and chemical inhibition of hacA u processing is that the ΔhacA mutant lacks both hacA u and hacA i mRNAs due to the absence of the hacA gene, whereas the 48C-treated organism still expresses hacA u (Fig. 2B). This raises the possibility that residual hacA i levels in cultures treated with 48C are sufficient to provide some protection against ER stress. We therefore examined a mutant that harbors an inactivating deletion in the RNase domain of IreA but still retains the hacA gene (ireA ΔRNase ). As observed with 48C treatment, no hacA i mRNA could be detected by RT-qPCR in the ireA ΔRNase mutant, even in the presence of a strong ER stress stimulus, confirming that this strain is incapable of mounting a canonical UPR response (Fig. 5B). The levels of hacA u were notably elevated in the ireA ΔRNase mutant (Fig. 5B), consistent with a failure to process basally expressed hacA u mRNA into hacA i . Interestingly, and similar to what we observed with 48C treatment, the ireA ΔRNase mutant showed no hypersensitivity to DTT, tunicamycin, or itraconazole (Fig. 5A).
We have previously shown that the ΔhacA mutant has a lower MIC to caspofungin using an MIC test strip and that the normally fungistatic effects of caspofungin toward A. fumigatus became fungicidal in the absence of hacA (Fig. 5C) (3). In contrast, caspofungin remained fungistatic to the ireA ΔRNase mutant and the 48C-treated parental strain, with no decrease in the MIC (Fig. 5C). In addition, we found that neither the ireA ΔRNase mutant nor treatment with 48C increased thermal stress sensitivity as previously reported for the ΔhacA mutant (Fig. 5D) (3). However, the ireA ΔRNase mutant was hypersensitive to carvacrol and grew poorly on a collagen substrate (Fig. 5E and F), similar to the effects of 48C on the control strain ( Fig. 3C and Fig. 4B). Together, these findings indicate that blocking hacA i accumulation under conditions of ER stress by either a chemical or genetic approach is less detrimental to the fungus than the complete absence of the hacA gene, suggesting that there is additional complexity in the pathway that involves functions for the unspliced hacA u mRNA that are distinct from the spliced mRNA.

DISCUSSION
Selective inhibitors of the kinase and/or RNase domains of human IRE1 are under development for the purpose of creating toxic levels of unfolded proteins in human tumor cells, particularly when used in combination with drugs that exacerbate ER stress (18,19,48,49). These compounds are well tolerated in animal models and have been associated with favorable therapeutic outcomes (21)(22)(23)(24)50). However, their effects on a fungal pathogen have not been reported. Here, we demonstrate that the human IRE1 RNase inhibitor 48C impaired the ability of A. fumigatus to maintain basal levels of hacA i mRNA under vegetative growth conditions, in addition to preventing the accumulation of hacA i under conditions of acute ER stress induced by treatment with DTT or the natural product carvacrol. This correlated with reduced expression levels of known UPR target genes, reflecting inhibition of the canonical UPR. STF-083010 is another inhibitor of the human IRE1 RNase and is structurally related to 48C (21). We found that this compound also impaired hacA i mRNA induction and sensitized the fungus to carvacrol-induced ER stress (see Fig. S5 in the supplemental material), providing additional support for the notion that disabling the fungal UPR is achievable by small-molecule inhibition. However, in contrast to 48C, STF-083010 was only partially effective, so it was not pursued further in this study. Our data on 48C provide the first validation of small-molecule inhibition of the fungal UPR in A. fumigatus, which could provide a valuable tool for future studies into how this stress response pathway integrates with other cellular circuitry to coordinate adaptive responses in fungi. It is also worth mentioning that UPR activity impacts one of the bottlenecks that limit the secretion of proteins by filamentous fungi used in the biotechnology industry (51), suggesting that 48C could have broader utility as a way to understand how the UPR affects industrial secretion processes.
It is well known that filamentous fungi possess a high capacity to secrete large quantities of hydrolytic enzymes that allow them to break down complex substrates (52). Minimal medium that is optimized for the growth of A. fumigatus contains simple forms of carbon and nitrogen that are readily assimilated by the fungus. Treatment with 48C had only a minor inhibitory effect on growth in this medium, similar to the mild growth impairment displayed by a ΔhacA mutant under the same conditions (Fig. 3). In contrast, either deletion of hacA or treatment with 48C severely impaired the ability of A. fumigatus to grow on a collagen substrate. This is consistent with the importance of the UPR in supporting the accurate folding of secreted hydrolytic enzymes that are necessary to break down complex biopolymers (Fig. 3C).
Carvacrol is an essential oil from the oregano plant with in vitro antifungal activity against a variety of fungal species (42,(53)(54)(55)(56) and has been shown to have potential in vivo application as a prophylactic agent for the prevention of avian aspergillosis (57). Although the precise molecular target of carvacrol is unknown, its ability to disrupt ER morphology and induce a UPR transcriptional signature in C. albicans clearly indicates that it causes acute ER stress (42). We found that carvacrol also induced the UPR in A. fumigatus, which could be blocked by treatment with 48C (Fig. 4A). The two compounds showed synergistic toxicity against A. fumigatus, suggesting that 48C prevents the accumulation of hacA i mRNA needed to counteract the toxic effects of carvacrol on ER homeostasis ( Fig. 4B and C). We also found that 48C increased the susceptibility of C. albicans to carvacrol, suggesting a conserved mechanism of carvacrol between species (Fig. S6). Interestingly, our findings also revealed enhanced toxicity of hygromycin B against A. fumigatus in the presence of 48C. This is likely to be due, in part, to the ability of 48C to prevent the canonical UPR from adequately responding to the ER stress caused by hygromycin-induced errors in protein translation (Fig. 4A). In addition, the adverse effects of 48C on the expression of glycosylationrelated genes (Fig. 4D) would be expected to impair glycosylation-assisted protein folding, as well as impair glycosylation-dependent hygromycin B resistance mechanisms that have been well described from studies on yeast glycosylation mutants (46). We conclude that 48C-induced hypersensitivity to hygromycin B is likely to be multifactorial but mediated in part by UPR signaling through the canonical HacA idirected pathway.
Despite the existence of two overlapping reading frames in all species homologs of the mRNA encoding the UPR transcription factor, it is widely assumed that the spliced mRNA is the most relevant to ER homeostasis because it translates a bZIP transcription factor involved in UPR target gene expression. Interestingly, a phylogenetic analysis of vertebrate homologs of this mRNA has indicated that the evolution of the two overlapping reading frames argues for functionality of the unspliced transcript protein (58). Recent studies in human cells have shown that one function of the protein specified by the unspliced transcript is to target the nascent protein-mRNA-ribosome complex to the ER membrane via the signal recognition particle (SRP) pathway (59)(60)(61)(62). If the sole purpose of the protein encoded by the unspliced mRNA in A. fumigatus is bringing the hacA u mRNA in the proximity of the IreA RNase in order to optimize splicing into hacA i , one would predict that preventing hacA i accumulation by treatment with 48C or by IreA RNase domain mutation would have the same effects as deleting the hacA gene (Fig. 1A). However, our data show that the effects of deleting hacA are broader in scope relative to chemical or genetic RNase inhibition (Fig. 5), suggesting that there is unexplained functionality of the unspliced hacA u mRNA that goes beyond serving as the precursor for hacA i . In Cryptococcus neoformans, the majority of the phenotypes associated with ire1 deletion could be complemented with the spliced transcription factor gene but not by the unspliced version (10). However, in that study complementation was performed in the complete absence of Ire1; in our study Ire1 was still present but rendered incapable of inducing the canonical pathway due to chemical or genetic inhibition of the Ire1 RNase. Ire1 is also present in the ΔhacA mutant of A.
fumigatus, but the phenotype of that mutant is more severe than the phenotype of canonical UPR inhibition using 48c treatment or RNase mutation, suggesting that beneficial effects of hacA u are apparent when the Ire1 sensor is present. This raises the possibility that the unspliced protein augments noncanonical functions of Ire1, which would be consistent with evidence that it is targeted to the ER membrane (59)(60)(61)(62). Unique functions for the unspliced protein have also been proposed in human cells (63,64), as well as in Aspergillus oryzae, where mutants that constitutively express only the spliced form exhibit a transcriptional profile that is different from that of a hacA deletion strain (65). The existence of additional noncanonical branches in the UPR that contribute to ER homeostasis is also consistent with studies showing that the UPR is capable of customizing the regulation of the target gene expression contingent upon the nature and scope of the stress that a cell encounters, rather than acting as a simple "on-off" switch (66,67). Experiments to explore the mechanisms involved in A. fumigatus are under way and are expected to further unravel the complexity of the UPR in the biology and virulence of this clinically important fungal pathogen.

MATERIALS AND METHODS
Reagents. Aliquots of 48C (EMD Millipore; 412512) at 25 mg/ml were prepared in dimethyl sulfoxide (DMSO) and stored at Ϫ20°C until use. Dithiothreitol (DTT; Thermo Scientific; R0862) was dissolved in water at 1 M prior to use. Tunicamycin (Cayman Chemical; 11445) and itraconazole (Sigma; 16657) were dissolved in DMSO at 10 mg/liter and stored until use at Ϫ20°C or Ϫ80°C, respectively. Hygromycin B (RPI; H75020) was dissolved in water at 100 mg/ml and stored at Ϫ80°C prior to use. A 6.4 M stock solution of carvacrol (Sigma-Aldrich; W224511) was diluted to 0.64 M (96.74 g/liter) in DMSO prior to immediate use.
Strains and culture conditions. The A. fumigatus strains used in this study are summarized in Table S1  . Colony diameters were determined by spotting 5 ϫ 10 3 conidia onto the center of a 100-mm plate containing AMM ϩ 0.8% (wt/vol) agarose (UltraPure agarose; Invitrogen) with or without 48C, and colony diameters were measured after 6 days. For analysis of stress sensitivity, serial dilutions of conidia (10 5 to 10 1 conidia in a 5-l volume) were spotted onto AMM plates supplemented with 0.8% agarose and the compounds of interest. Plates were incubated at 37°C and photographed after 48 h. For analysis of germination rates, a total of 1 ϫ 10 3 spores were inoculated into liquid AMM containing the indicated concentrations of 48C and incubated at 37°C. Controls with DMSO were run in parallel. The number of germinated conidia was then quantified microscopically (90 to 183 cells counted per condition). For analysis of mycelial biomass, liquid AMM was inoculated with CEA10 conidia at a concentration of 1 ϫ 10 6 conidia/ml and incubated for 24 h at 37°C. The mycelium was then dried and weighed. Growth on a collagen substrate was performed by diluting a 10ϫ stock solution of Aspergillus minimal salts to 1ϫ with a 5-mg/liter solution of type l collagen as the sole source of carbon and nitrogen (Sigma; C3867; 5 mg/liter), and inoculating 100 conidia into the surface of the gel in a 96-well plate. Plates were photographed after 72 h at 37°C. Control cultures in liquid AMM were prepared in parallel.
RNA extraction and RT-qPCR. Cultures were inoculated with 1 ϫ 10 6 conidia/ml in 50 ml of AMM and incubated at 37°C and 200 rpm. After 16 h of incubation, cultures were treated for 1 h with the indicated compounds prior to RNA isolation. RNA was isolated using the RNAzol RT column kit according to the manufacturer's instructions (Molecular Research Center, Inc.). The digestion of genomic DNA and synthesis of cDNA were performed using RNase-free DNase (Roche) and iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad). The RT-qPCR was performed in a StepOne real-time PCR system (Applied Biosystems) using 1 g of template per well and iTaq Universal SYBR Green Supermix (Bio-Rad). The primers used in the reaction mix are summarized in Table S2, and their final concentration was 500 nM, with the exception of 200 nM for the housekeeping gene (18S rRNA). Primer efficacy was evaluated using a standard curve. The specificity of the hacA u and hacA i reverse primers for distinguishing between the unspliced and spliced versions of the hacA mRNA was confirmed by RT-qPCR using the TaqMan probe and primers listed in Table S2. The final concentration for probes and primers was 250 nM and 500 nM, respectively (including the housekeeping gene).The reaction was performed with the Ex Taq master mix (TaKaRa), using the same cycle parameters described above for RT-qPCR. Since expression results obtained with SYBR green were similar to those obtained with a TaqMan probe, the SYBR green method was employed for the results presented in this study, using the same primer concentrations used for TaqMan. For hacA u and hacA i mRNA detection, the cycle conditions were 20 s at 95°C, 40 cycles of 3 s at 95°C, and 20 s at 66°C. For detection of other genes, the cycle conditions were 20 s at 95°C, 40 cycles of 3 s at 95°C, and 30 s at 60°C. The melting curve was monitored to verify the specificity of the amplification reaction. The 18S rRNA was used as a housekeeping gene. Fold change in mRNA levels was determined in comparison to untreated cultures.

Analysis of metabolic activity and antifungal susceptibility.
For analysis of metabolic activity, conidia were inoculated at a concentration of 1 ϫ 10 6 conidia/ml in 150 l of liquid AMM in a 96-well plate and incubated for 16 h at 37°C. The plates were washed three times, and the indicated concentrations of 48C were added to the plates before incubating for an additional hour at 37°C. The final concentration of DMSO was 0.5% (vol/vol) in each well, including the growth controls. The cultures were then washed three times, and the medium was replaced with AMM containing 0.02 mg/ml of the oxidation-reduction metabolic indicator resazurin (14322; Cayman Chemicals). After incubating for 1 h at 37°C, the fluorescence was measured (excitation, 535 nm; emission, 590 nm) in a microplate reader (Synergy H1; BioTek). For analysis of synergy between carvacrol and 48C, conidia were inoculated into liquid AMM containing 0.02 mg/ml of resazurin at a concentration of 2.5 ϫ 10 4 conidia/ml. Serial 2-fold dilutions of 48C were then dispensed from column 2 in the 96-well plate to column 9. Additionally, 2-fold dilutions of carvacrol were dispensed from row A to H into each well containing 48C. Columns 1 and 11 were used as growth and sterility controls, respectively. Columns 10 and 12 were used to display the individual MICs of carvacrol and 48C. The plates were incubated at 37°C for 24 h, after which the fluorescence was measured as described above. The MIC was defined as the lowest concentration required to inhibit 90% of fungal metabolic activity after 24 h of incubation. An isobologram representation was used to determine if the combination of 48C and carvacrol had synergistic activity. The diagonal line in Fig. 4C connects the individual MICs for each compound (shown by the solid circles on the axis lines), and the remaining circles represent the MICs for different combinations of the two drugs. Synergistic, additive, and antagonistic effects are represented by solid circles that fall below the diagonal line, on top of the diagonal line, or above the diagonal line, respectively. The fractional inhibitory concentration (FIC) index was determined as (MIC carvacrol in combination with 48C/MIC carvacrol alone) ϩ (MIC 48C in combination with carvacrol/MIC 48C alone). The interaction between the drugs was defined as synergistic (FIC Ͻ 0.5), indifferent (0.5 Ͻ FIC Յ 4.0), or antagonistic (FIC Ͼ 4.0). Antifungal susceptibility using the MIC test strip (MTS) method was determined following the instructions of the manufacturer. A concentration of 10 6 conidia/ml was spread with a sterile cotton swab onto a plate of RPMI 1640 agar with 0.164 M morpholinepropanesulfonic acid (MOPS) and L-glutamine (pH 7.0). RPMI plates containing 48C were run in parallel. After the surface of the medium appeared dry, an MIC test strip containing caspofungin (Liofilchem; 92154) was applied. Plates were incubated for 48 h at 37°C and photographed.
Statistical analysis. Statistical analysis was performed using GraphPad Prism version 8 for Windows (GraphPad Software, San Diego, CA, USA). Statistically significant differences were determined using one-way analyses of variance (ANOVA) with Dunnett's or Tukey's multiple-comparison tests.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.