All IRE1 inhibitors were prepared in dimethyl sulfoxide (DMSO) and stored at -80°C until use. 50 mM stock of IRE1 inhibitor III, 4µ8C (#412512 - MilliporeSigma, MA, USA), 50 mM Sunitinib (HY-10255A – MedChemExpress, NJ, USA), 10 mM Z4P (HY-153773 – MedChemExpress, NJ, USA), 10 mM STF-083010 (HY-15845 – MedChemExpress, NJ, USA), 10 mM KIRA6 (HY-19708 – MedChemExpress, NJ, USA) and 10 mM HY-114368 (HY-15845 – MedChemExpress, NJ, USA) were used. DL-Dithiothreitol (#0281 DTT- VWR, PA, USA) was prepared at 1 M concentration in sterile PBS.

The FK1 and FK2 A. fumigatus isolates are from fungal keratitis patients at the University of California at San Francisco-Proctor Foundation. An Af293 derivative strain that constitutively expresses the mcherry protein (Af293::PgpdA-mcherry-hph) was used throughout this study (Kamath et al., 2023; Lightfoot et al., 2024). Strains were maintained on glucose minimal media (GMM) containing 1% glucose, Clutterbuck salts, and Hutner’s trace elements, 10 mM ammonium tartrate (GMM) at pH 6.5 (Cove, 1966). To test the effect of the IRE1 inhibitors on the different Aspergillus strains, conidia were inoculated into GMM broth to a final density of 1x 10⁶/mL against a concentration gradient of drug (0 - 480 µM) and incubated at 35°C. After 72 h, the minimum inhibitory concentration (MIC) was recorded by visually inspecting the plate and looking for the growth of the fungi under the microscope. To assess the fungicidal effect of the IRE1 inhibitors, a broth microdilution assay was set up as described above and incubated at 35°C for 72 h. 100 µL aliquots from wells in which germination was inhibited were then spread across yeast extract, peptone, dextrose (YPD) plates and incubated at 35°C for 24 h prior to the enumeration of colony-forming units (CFU). To assess biofilm metabolic activity, the XTT metabolic assay was performed as described previously (Pierce et al., 2008). XTT buffer was prepared at a stock of 0.5 mg/mL and menadione at 10 mM in 100% acetone (final conc. 1 µM). For this assay, 10⁶ conidia/mL were inoculated in GMM in 96-well plate and incubated at 35°C for 24h to form hyphae followed by treatment with 4µ8C (0 – 240 µM) for 2 h. Next, 100 µL of an XTT-menadione mixture was added to each test well and incubated further for 2 h at 35°C. Supernatants from each well (75 µL) were assayed at 490 nm.

Fungal cultures were grown overnight at 35°C in GMM in 6-well plates were treated with a concentration gradient of 4µ8C (0 – 120 µM) for 2 h followed by treatment with 10 mM DTT or vehicle for 2 h. RNA was then extracted from these cultures using RNeasy Mini kit (#74106 Qiagen, Germany) followed by DNase treatment using the DNase I kit (Millipore Sigma, Massachusetts, USA). The Nanodrop 2000 (Thermo Fisher Scientific, Massachusetts, USA) was used to assess the quantity and quality of the RNA. Subsequently, the RNA was standardized for cDNA conversion utilizing the ProtoScript II First Strand cDNA Synthesis Kit (New England Biolabs, Massachusetts, USA) following the provided protocol. To visualize hacA splicing, the coding sequencing (spanning the predicted 20 bp unconventional intron) was amplified and resolved by electrophoresis (3% agarose gel at 40 V for 10 h) as previously described (Kamath et al., 2023). For qRT-PCR analysis, Luna Universal qPCR master mix (SYBR green; NEB, USA) was used on the QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific, Massachusetts, USA), and the analysis was conducted with QuantStudio Design and Analysis Software v1.5.2. Fold-expression changes were computed using the 2-ΔΔCt method, followed by analysis using MS-Excel. The primers used to amplify hacA, bipA and pdiA in this assay are the same as previously published (Guirao-Abad et al., 2020; Kamath et al., 2023).

Inoculum: 5x10⁶ conidia were incubated in 20 ml YPD at 35°C for approximately 4 h until mostly swollen. Biomass was collected, washed 4x with PBS, and resuspended in 500 μL PBS, adjusting volumes to normalize strains to 0.8 at OD 360 nm. Infections: 6-8-week-old C57B6/6J (Jackson Laboratory) received intraperitoneal immunosuppression with 100 mg/kg Depo-Medrol (Zoetis, USA) on day -1. On day 0, mice were anesthetized using 100 mg/kg ketamine and 6.6 mg/kg xylazine by IP injection, and their right eyes abraded to a 1 mm diameter using Algerbrush II. Fungal inocula (5 μL) were applied to ulcerated eyes and removed with a Kimwipe after 20 minutes. The mice also received Buprenorphine SR (1 mg/kg) which was administered subcutaneously. Contralateral eyes remained uninfected per ARVO guidelines. For controls, we used algerbrushed sham-infected and untouched eyes. Micron IV slit-lamp (Phoenix Research Labs Inc., CA, USA) was used to monitor mice daily post-inoculation (p.i.) up to 72 h. This procedure was performed while the mice were anesthetized with isoflurane. Anterior segment spectral-domain optical coherent tomography (OCT) measured corneal thickness at 48 and 72 h p.i. using a 4x4 mm image using 12 mm telecentric lens from Leica Microsystems (IL, USA). The reference arm was set to 885 according to the manufacturer’s calibration instruction. Image analysis was conducted using the InVivoVue diver software. Corneal thickness quantification involved measuring an 11x11 spider plot covering the entire eye, and an average of 13 readings were obtained. Fluorescein imaging: Corneal fluorescein staining was performed to test the rate of healing of the abraded epithelium post treatment with DMSO or 4µ8C. AK-Fluor 10% fluorescein solution (Akorn, IL, USA) diluted 1:100 was applied to the corneas in isoflurane-anesthetized mice using a cotton swab and excess was wiped off and imaged using a slit-lamp Micron (Phoenix Research Labs Inc., CA, USA). Histopathological analysis: Control eyes were harvested after 72 h and fixed with 10% neutral buffered formalin for 24 h followed by 70% ethanol until they were further processed. 5 μm thick sections of the eyes were stained with Hematoxylin and eosin (H&E) to evaluate host inflammatory response after treatment with DMSO or 4µ8C. Fungal burden determination: Corneas were dissected in a sterile manner, homogenized using a collagenase buffer at a concentration of 2 mg/ml, and then 100 μl aliquots were plated in triplicate on inhibitory mold agar (IMA) plates. The number of colony-forming units (CFU) per cornea was assessed after 24 h of incubation at 35°C. Clinical scoring: Micron images were randomized and scored (0 to 4 scale) for clinical pathology by reviewers not involved with the experiment based on previously established criteria, including surface regularity, area of opacification, and density of opacification (Zhao et al., 2009). Topical Treatments: For the topical treatments of sham or Af293-inoculated corneas, animals were anesthetized using isoflurane and a 5 µL drop of the compound (or DMSO as a vehicle control) was applied to the ocular surface. The animals were kept under anesthesia for 10 minutes with the drop in place before being returned to their cage. The treatment schedule varied slightly across experiments (see the corresponding figure legends), but generally consisted of one treatment on the day of inoculation (4 h p.i.), three treatments (4 h apart) at 1 and 2 days p.i., and one treatment at 3 days p.i., just prior to imaging and removal of the eyes for histological or fungal burden analysis. In the case of the fluorescein experiments, the animals received three treatments 3 days p.i.

All statistical analyses were performed using GraphPad Prism 10 version 10.2.0. The specific tests used for each experiment are described in the corresponding figure legends.

We first predicted that Ire1 inhibitors would display in vitro antifungal activity against Af293 and potentially other A. fumigatus strains in which IreA influences fungal growth and or viability. We began by screening the activity of six commercially available Ire1 inhibitors – including four kinase domain inhibitors (Sunitinib, Z4P, KIRA6, KIRA8) and two endonuclease domain inhibitors (4μ8C, STF-083010) ( Supplementary Figure 1 ) – in a broth microdilution assay. Four A. fumigatus strains were tested in the screen, including the well-studied Af293 and CEA10 strains (both pulmonary isolates) and two uncharacterized corneal isolates from fungal keratitis patients, designated here as FK1 and FK2 (Bertuzzi et al., 2021). No inhibitory activity was observed with any of the four kinase inhibitors even at the highest tested concentration ( Supplementary Figure 2 ), suggesting that these compounds either do not adequately accumulate within the fungal cell, they do not have high affinity for the A. fumigatus kinase domain, or the domain itself is not essential for A. fumigatus growth in these isolates. By contrast, both endonuclease inhibitors fully blocked conidial germination, with 4μ8C and STF-083010 exhibiting comparable MICs against Af293 and FK1, and 4μ8C displaying overall lower MICs against CEA10 and FK2 ( Figure 1A ). To gain some insight into whether their inhibitory action was fungistatic or fungicidal, the microtiter assay was repeated with Af293 and, following 72 h incubation in the presence of drug, aliquots from wells in which germination was inhibited (120-480 µM) were spread onto drug-free YPD plates and incubated for an additional 24 h ( Figure 1B ). Whereas cultures inhibited by STF-083010 all yielded fungal colonies upon sub-culture, all plates corresponding to 4µ8C inhibition remained sterile. These results suggest that 4µ8C is either more potently fungicidal than STF-083010 or that its repressive action, perhaps as a reflection of target binding affinity, is stronger ( Figure 1B ). To determine if 4µ8C can similarly affect fungal hyphae (i.e. the tissue invasive form), an XTT reduction assay was performed on mycelia/biofilms cultured overnight in static culture and subsequently treated with the drug for 2 h. As shown in Figure 1C , the 120 and 240 µM treatments resulted in a markedly reduced or undetectable metabolic activity, respectively, altogether indicating that a similar concentration of drug is inhibitory and potentially fungicidal against both the conidia and hyphae of A. fumigatus. Since 4µ8C provided the best antifungal profile, both in terms of its MIC across the strains and its possible fungicidal action, we decided to prioritize that compound for downstream characterization.

In biochemical assays, 4µ8C forms a Schiff base with distinct lysine residues in both the Ire1 kinase and endonuclease domains, but only inhibits the activity of the latter in treated cells (Cross et al., 2012). This lysine is conserved in the A. fumigatus IreA endocnuclease domain, as is the influence of the drug on hacA splicing in the AfS28/D141 background (Guirao-Abad et al., 2020). We reasoned that if the antifungal action of 4µ8C against Af293 were attributable to an ‘on-target’ effect, then there would be a correspondence between the fungicidal MIC (ranging from 60 to 120 µM) and the inhibition of IreA function. To explore this, fungal mycelia/biofilms were developed in GMM static culture overnight, subsequently pretreated with DMSO (vehicle) or various concentrations of 4µ8C for 2 h, and finally stimulated with 10 mM DTT for an additional 2 h. Samples treated with neither 4µ8C nor DTT served as a control. The splicing status of hacA was then assessed by RT-PCR and gel electrophoresis as previously described (Kamath et al., 2023). As expected, treatment with DTT alone resulted in an observable accumulation of the spliced/induced hacA product (hacA ⁱ), relative to untreated controls, and the inclusion of 30 µM 4µ8C did not have an impact on this response. By contrast, the hacA ⁱ splice form was completely absent in both the 60 and 120 µM 4µ8C treated samples, thus establishing a putative connection between IreA function and fungal cell metabolism or viability ( Figure 2A ; Supplementary Figure 3 ). Sequencing of these end-point PCR products revealed that the canonical (spliceosomal) intron at the 5’ end of the hacA mRNA was absent in all samples, whereas the 20 bp IreA-targeted (non-canonical) intron was absent only in the DTT-treated fungus in the absence of 4µ8C ( Figures 2C, D ; Supplementary Figure 3 ). Finally, in addition to a block in hacA splicing, we further observed that 60 µM blocked the DTT-mediated induction hacA message as well as the steady-state mRNA of two known HacA-regulated transcription factors, BipA and PdiA (Guirao-Abad et al., 2020) ( Figure 2B ). Taken together, these results support a model in which 4µ8C blocks UPR signaling in A. fumigatus, and by extension fungal metabolism and growth, through a specific action on IreA endonuclease activity, rather than a broad inhibition of mRNA processing by the spliceosome.

We were next interested in exploring the in vivo antifungal activity of 4µ8C in a murine model of FK, but first had to consider the concentration to test. Ophthalmic antifungal formulations tend to be highly concentrated to facilitate sufficient accumulation of the drug within the dense collagen matrix of the cornea, particularly in the face of rapid dilution and drainage of the drug at the ocular surface. For example, the in vitro MIC for natamycin against most Aspergillus isolates ranges between 2-32 µg/mL but is administered at 500-25,000X this concentration in the form of hourly 50 mg/mL (75 mM) drops. Similarly, voriconazole drops are often formulated at 1 mg/mL (28 mM), where the in vitro MIC for most A. fumigatus isolates is 0.25-4 µg/mL (Sun et al., 2014). Thus, while testing the highest concentration of 4µ8C that its solubility permits (132.24 mM) would be desirable, we had to further consider that the impact of the drug on corneal tissue homeostasis has not been explored. We therefore began by testing the ocular toxicity profiles of two relatively conservative 4µ8C concentrations, 0.5 and 2.5 mM, which correspond to approximately 5X and 40X the in vitro antifungal MIC, respectively.

The two drug concentrations were evaluated in sham model of FK previously developed by our group, which involves the generation of an epithelial ulcer ahead of fungal inoculation, or in this case, ulceration without the addition of fungus (Sun et al., 2016; Kamath et al., 2023). Each treatment consisted of a 5 µL drop of 4µ8C or vehicle (DMSO) applied to the corneal surface up to three times daily per the schedule described on the corresponding figure legends and methods. Corneas were tracked daily by slit-lamp imaging to assess corneal clarity and ocular surface regularity, as well as by optical coherence tomography (OCT) to measure corneal thickness and assess gross morphology. Corneas were then resected at 72 h post-ulceration for histological analysis with H&E staining. Relative to the vehicle control group, neither 4µ8C concentration had an observable impact on corneal clarity or corneal thickness, suggesting the drug did not lead to acute inflammation or edema ( Figures 3A, B ; Supplementary Figures 4A, B ). Interestingly, histology revealed that whereas the epithelium of vehicle-treated corneas had reformed, those treated with either concentration of 4µ8C still appeared ulcerated ( Figure 3C ; Supplementary Figure 4C ). To evaluate this further, we replicated the experiment and followed re-epithelialization (wound healing) longitudinally with fluorescein staining. Briefly, an intact corneal epithelium precludes the entry of the stain into the underlying tissue, thus giving a negative signal on fluorescence slit-lamp imaging; by contrast, ulcerated eyes give rise to diffuse yellow-green staining of the stromal layer. As opposed to the vehicle-treated eyes, which fully precluded stromal staining within 24 h post-ulceration, the 4µ8C-treated group remained positive for the duration of the 72 h treatment. We observed, however, that the corneas were fully healed within 3 days of stopping the 4µ8C treatment, suggesting that the drug has only a transient impact on epithelial cell proliferation ( Figure 3D ; Supplementary Figure 4D ). Taken together, we concluded that the tested treatment regimens of 4µ8C were minimally toxic to the murine cornea and suitable for evaluation in a treatment model of FK.

To test the impact of 4µ8C on fungal growth and disease outcomes during FK, we inoculated ulcerated corneas with swollen conidia of A. fumigatus Af293 (day 0) and treated on the same schedule as described above – once on the day of ulceration/inoculation, three times at 24 and 48 h p.i, and once at 72 h p.i. Corneal disease metrics were tracked longitudinally by slit-lamp and OCT as before, but included fungal burden assessment at 72 h by homogenizing and plating corneas for CFU enumeration. We began with the 0.5 mM 4µ8C preparation, where the lack of corneal toxicity on sham/uninfected corneas observed in the previous safety trial was recapitulated. Furthermore, treatment at this concentration resulted in an approximately 50% reduction in fungal load at 3 days p.i. relative to vehicle-treated FK corneas ( Supplementary Figure 5A ). This reduction in fungal load did not, however, correspond to improved clinical disease scores or reduced corneal thickness at any of the evaluated time points ( Supplementary Figures 5B–D ), suggesting this treatment regimen is insufficient to block corneal inflammation and tissue damage associated with FK pathology. We next tested the higher (2.5 mM) concentration of 4µ8C in the same manner and observed a >90% reduction in fungal burden at 72 h p.i. that corresponded to an absence of corneal surface disease or cornea swelling at all evaluated time points ( Figure 4 ). These results suggested that the higher 4µ8C dosage led to a rapid corneal sterilization that blocked the establishment of infection and downstream pathology. The relative influence of 0.5 and 2.5 mM 4µ8C on fungal load and disease development was recapitulated across multiple experiments with each concentration.