Shining a light on the impact of antifungals on Aspergillus fumigatus subcellular dynamics through fluorescence imaging

Oligonucleotides and plasmids used in this study are listed in Table S1. To generate conidia for experiments, strains were grown on solid Aspergillus minimal medium (AMM) (33) at 37°C for 5 days in vented tissue culture flask (Corning) and harvested through Miracloth (Millipore) in PBS + 0.01% Tween-20. Phenotyping on solid medium was carried out by point inoculation of 1 × 10⁴ of each strain in a total volume of 5 µL PBS + 0.01% Tween-20 on solid AMM, pH 5. Plates were incubated for 48 h at 37°C. Liquid growth curves were carried out by inoculating 5 µL of a 1 × 10⁵ spores/mL solution of each strain in 200 µL liquid AMM, pH 7. Microdilution plates were incubated for 48 h at 37°C, and optical density at 600nm (OD₆₀₀) was measured every 10 min on a BioTek Synergy2 plate reader. Growth rates were calculated as the slope of the linear part of the growth curve (12–18 h). These were compared by a one-way ANOVA (n = 3) with Šidák multiple comparisons. PgpdA-FP-expressing strains were imaged in liquid AMM, pH 7. Germlings were grown by inoculating 1 × 10⁴ spores in 200 µL in a µ-Slide 8-well high glass bottom chamber (Ibidi) or a microfluidic chamber made in-house (described in section 2.3) at 30°C for 16 h. PxylP-FP-expressing strains were imaged in liquid RPMI-1640 supplemented with 1 mg/L amphotericin B (Sigma-Aldrich), 0.5 mg/L voriconazole (Sigma-Aldrich), 0.016 mg/L olorofim (Concept Life Sciences) or 0.6 mg/L manogepix (Selleck Chemicals), with 1% xylose to induce gene expression.

FP knock-in cassettes were generated using fusion PCR (34) (Fig. S1a), or by linearization of plasmids where the PxylP promoter and FP were already sequential on a vector (see Table S1 and supplemental methods). The construction of plasmids and the tetrachrome strain is described in detail in the supplemental methods. Transformations were either performed by homologous recombination following Zhao et al. (35) or by CRISPR-Cas9-mediated transformation based upon (33). The method used to generate each strain is found in Table S2. In summary, A. fumigatus A1160P+ (36) was grown overnight at 37°C in a shake flask culture (130 rpm, 37°C). A protoplasting solution was made by dissolving 1 g/10 mL Vinotaste Pro (Novozymes) in 0.6 M KCl + 100 mM citric acid and added to the shake flask culture for 3 h at 130 rpm 37°C. Protoplasts were filtered through Miracloth to remove hyphae and debris, followed by three washes in 0.6 M KCl. Protoplasts were resuspended in 0.6 M KCl + 50 mM CaCl₂. 25 µL of PEG6000 solution, and ~500 ng knock-in cassette was added to the protoplast suspension. If CRISPR-Cas9-mediated transformation was used, ribonucleoproteins (RNPs) were assembled by incubating crRNA with tracrRNA (Integrated DNA Technologies, Inc.) in duplex buffer and further incubated with purified SpCas9 (Integrated DNA Technologies, Inc.). RNPs were added at this stage. Protoplast suspensions were incubated for 50 min on ice. 600 µL of PEG6000 solution was added followed by incubation for 20 min at room temperature. Suspensions were plated onto AMM + 1 M sucrose-containing 100 mM citrate buffer (pH 5). Selection procedures using counter-selectable markers fcyB, fcyA, and uprt were conducted as described previously for A. fumigatus (14, 37, 38). In the case of cntA-based counter-selection, plates containing 50 µg/mL 5-fluorouridine (5FUR) were supplemented with 50 µg/mL of clorgyline (CLG). Strain descriptions and sources for each FP used in each cassette are found in Tables S2 and S3, respectively.

The microfluidic device utilized in this study is fabricated through a process involving photolithography and dry etching to create a negative template. Polydimethylsiloxane (PDMS) is then cast using a 1:5 ratio based on the layout defined by the negative. Inlets and outlets are created by punching holes into the PDMS. Following this, the PDMS and KOH-cleaned glass substrate are subjected to plasma treatment to enhance bonding. The glass substrate and the channel side of the PDMS are then brought together, forming a chemically bonded structure. To reinforce the bonding, the assembled device is placed glass side down on a hotplate at 90°C for 10 min. After cooling, the channels are cleared by injecting a 1:1,000 Tween-20 dilution for 30 min, followed by inoculation with cell suspensions. The spores are allowed to incubate for an hour before applying a flow rate of 0.8 mL/h through the channels. Germlings were grown as in section 2.1. To confirm the predicted vacuolar localization of Vam3 along the hyphae, ΔfcyA::GFP S65T-Vam3^PxylP cultures were stained with 2 µM of CellTracker Blue CMAC dye and further incubated for about 45–60 min as previously described for A. fumigatus (17). Fluorescence microscopy images were captured with a fully motorized Leica SP8x laser scanning confocal microscope equipped with a 40×/0.85NA HCX PL APO dry objective or a 63×/1.4NA HC PL APO CS2 oil objective. Imaging was performed at 37°C. All images were captured in 8-bit at 1,040 × 1,040 pixels. Blue FPs were excited using the 405 nm diode laser at 10%. All other FPs were excited using a white light laser at 20%. Excitation wavelengths were chosen to match the excitation maxima of the FPs, apart from mTagBFP2 and mTurquoise2 where the 405 nm diode laser was used. The fluorescence signal was captured in a 20 nm bandwidth spanning the maximum emission. For relative brightness measurements (Fig. 1a), the 40× objective with a pinhole of 1 AIRY unit, a scan speed of 400 Hz, and a line average of 4 was used. The gain was set to 300%. For the representative image (Fig. 1b), the 63× objective with 3× zoom, a pinhole of 1 AIRY unit, a scan speed of 400 Hz, and a line average of 8 was used. The gain was set to 150%. FPs were captured using HyD detectors. For the tetrachrome strain, mTagBFP2^per was excited at 405 nm with a diode laser. The remaining FPs were excited with at white light laser at the following wavelengths GFP S65T^vac at 490 nm captured, Katushka2S^mit at 670 nm, and mKO2^mem at 588 nm.

Thirty individual germlings from each strain were assessed for FP relative fluorescent signal. The mean grey value of 30 2 × 2 µm sections deemed in focus by viewing the brightfield image was quantified. Fluorescence intensities are displayed in arbitrary units. For each strain, the same settings (magnification, scan speed, laser power, emission range, gain) were used on the background strain, A1160P+, to capture autofluorescence at each excitation/emission. This value was subtracted to give the practical brightness value. Relative fluorescent intensity plots were processed using GraphPad Prism 9.3.1. The brightness of mTagBFP2 and mTurquoise2 were compared using an unpaired t-test. The brightness of all other FPs were compared using a Kruskal–Wallis test with Dunn’s multiple comparisons.

Fluorescent strains were individually cultured on Sabouraud dextrose (SAB) agar for 72 h at 37°C. Spores were harvested with spore buffer for imaging flow cytometry processing. For each strain, 5,000 events were recorded with two technical replicates. All images and data were collected using an ImageStreamX Mark II Imaging Flow Cytometer (Merck). Total events were filtered to exclude speed beads, clumps of spores and out-of-focus spores. Fluorescence data were collected using IDEAS software. Excess speed bead images acquired were further excluded through gating. Gating to distinguish between spores and beads in scatter channel was determined by locating the area of spores using previously used fluorophore tdTomato in channel 3 (560–595) and applying the region containing all spores onto the brightfield channel. Median relative fluorescence intensity values were collected for each fluorophore across six channels: (Channel 1 – excitation 488 nm (emission 480−560 nm), Channel 2 – excitation 488/561 nm (emission 560−595 nm), Channel 3 – excitation 488/561 nm (emission 595−640 nm), Channel 4 – excitation 488/561 nm (emission 640−745 nm), Channel 5 – excitation 405 nm (emission 430−505 nm), Channel 6 – excitation 642 nm (emission 640−745 nm) . All fluorescence values were compared relatively to the non-fluorescent strain A1160P+. Fluorescence density plots were created in R using the ggplot2 package. The interacting effects of practical brightness and theoretical brightness were calculated using simple linear regression.

Data were acquired using ImageJ. Distance between neighboring peroxisomes was measured by auto-thresholding the image and using the nearest neighbor plugin (author Yuxiong Mao). Vacuole area was measured using the “analyze particles” function with a minimum size of five pixel units and a circularity of 0.20–1.00. Mitochondrial fragmentation was measured using the “analyze particles” function with a minimum size of 10 pixel units. Plots and statistical analyses were processed using GraphPad Prism 9.3.1. Significance was determined using a Kruskal–Wallis test with Dunn’s corrects for vacuole area, peroxisome internal, and mitochondrial fragmentation. To model the effect of exposure time of DSMO on mitochondria and vacuole area a mixed effects linear model was fitted to the data using the lmer package in R. The model fit the mitochondria area or vacuole area to the interacting effects of time and DSMO treatment, with replication modelled as a random effect. The model allowed the intercept and the slope to differ between treatments and replicates which resulted in a lower Akaike’s Information Criterion. Normality and linearity of the residuals satisfied the assumptions of the model.

Fluorescence imaging in A. fumigatus has relied mostly on a limited number of fluorophores, with the avGFP derivatives; even though GFP, eGFP, and GFP S65T are most commonly used (39), however, the relatively low brightness of these FPs makes visualizing proteins that are expressed at low levels difficult. Paired with stability and oligomerization issues (40, 41), this makes the current FP palette sub-optimal in A. fumigatus. However, some next-generation FPs are documented to be brighter, have more compatible spectra for multicolor imaging, and tend to be monomeric, overcoming many of the previous caveats for using FPs in A. fumigatus. The use of such FPs opens up novel applications such as single-cell experimentation. To assess next-generation FPs in A. fumigatus, we gathered a collection of fluorophores that have been used in C. neoformans, N. crassa and several fluorophores that have shown excellent performance in mammalian cells: mTagBFP2 and mTurquoise2 (blue); GFP S65T, mGreenLantern, the photoswitchable mEOS3.1, and mNeonGreen (green); EYFP and Citrine (yellow); mKO2, tdTomato, and TagRFP-T, and mRuby3 (orange); mApple, mScarlet-I3, mScarlet-3, and mRFP1 (red); and Katushka2S and mMaroon1 (far-red). These fluorophores span the visible spectrum, have different excitation optima, and have different oligomerization states (Table S3).

To enable systematic testing of the relative brightness of each FP, we generated strains expressing each FP driven by the A. nidulans gpdA promoter from the fcyB locus (Fig. S1a) (14, 42) as a single-copy integration (Fig. S1b). To assess the practical brightness of this set of FPs in A. fumigatus germlings, we imaged each strain and measured the relative fluorescence intensity using confocal microscopy (n = 30) (Fig. 1a). To compare FPs, the peak excitation wavelength was used and a 20 nm bandwidth spanning the maximum theoretical emission wavelength. However, it should be noted that excitation at 405 nm was used for mTurquoise2 and mTagBFP2 due to laser constraints and were therefore analyzed separately. The mean practical brightness ± SD for each FP is shown in Table 1. In the blue channel, the brightest blue protein was mTagBFP2, which was 22% brighter than mTurquoise2 (t = 4.999, df = 58, P < 0.0001). The brightest green protein was mNeonGreen, which was 146% brighter than GFP S65T (P < 0.0001). The yellow fluorescing protein Citrine was 159%, which was brighter than GFP S65T (P < 0.0001) and 70% brighter than EYFP (P < 0.0001). The brightest orange FP, mKO2, was 117% brighter than GFP S65T (P < 0.0001) and 283% brighter than the least bright orange FP, TagRFP-T (P < 0.0001). Aggregates could be observed in mKO2-expressing hyphae, apparent as bright areas ~0.5 µm in diameter (Fig. 1b), suggesting that this protein is poorly tolerated when overexpressed in the cytoplasm under constitutive expression in A. fumigatus, similar to previously reported results in S. cerevisiae (28). mApple was the brightest red FP, which was 108% brighter than GFP S65T (P < 0.0001) and 2,638% brighter than the least bright red FP, mRFP1 (P < 0.0001). While mRFP1 was considered the least bright fluorophore in all tested strains, we could still observe fluorophore under the microscope upon increasing the exposure time and gain. Katushka2S was the brightest overall FP in hyphae. It was 203% brighter than GFP S65T (P < 0.0001) and 705% brighter than the other far-red FP tested here, mMaroon1 (P < 0.0001).

A moderate but significant positive correlation was found between the mean practical brightness of each FP compared to the theoretical brightness from FPbase [Pearson’s correlation Theoretical: Practical T₁₇ = 2.7877, P = 0.01263, r (17) = 0.314] (Fig. 1c). FPs such as mRuby3 or mEOS3.1 (in the un-photoconverted, green state) are brighter in theory—when calculated as the product of extinction coefficient and quantum yield (43)— compared to how they perform in A. fumigatus, whereas Citrine and KatushkaS2 appear brighter in vivo. As PgpdA is reported to induce high expression (44, 45), we assessed the potential growth defects of strains expressing FPs. PgpdA-driven overexpression caused no growth defects in solid media (Fig. 1d) or had a significant effect on the growth rate in liquid media (Fig. S1c).

As many experimental setups in A. fumigatus rely on the infectious propagules, conidia, we assessed fluorescent signals of all the proteins when expressed in conidia. Conidia produced by all fluorescent constructs were distinguishable from non-fluorescent wild-type spores in at least one channel (Fig. S2). The brightest FP in the 430–505 nm channel was mTurquoise2. The brightest in the 480–560 nm channel was mEOS3.1. In the 560–595 nm channel, mKO2 was the brightest FP. Only a single FP showed peak emission within the 595–640 nm channel (mScarlet-3) and in the 640–745 nm channel (Katushka2S).

We have previously described a strategy to integrate three DNA sequences of interest into the A. fumigatus genome by exploiting the pyrimidine salvage pathway (14). In A. nidulans, CntA (NCBI accession number: XP_663097, A. fumigatus homologue AFUB_001570) has been described as the main importer of 5FUR (46). As the activity of 5FUR can be enhanced by the addition of the broad-spectrum inhibitor of fungal efflux pumps CLG (47, 48), we used a combination of 50 µg/mL of 5FUR with 50 µg/mL CLG to improve selection for cntA-disrupted colonies (Fig. S3a). This way, we were able to use cntA together with the endogenous counter-selectable markers fcyB, fcyA and uprt in a sequential manner (Fig. S3b and c). Replacement of all four loci did not affect A. fumigatus growth morphology on a plate assay (Fig. S3d). In addition, the replacement of all four loci did not change the MIC/MEC to voriconazole, amphotericin B, olorofim, and manogepix compared to the wild-type strain. Also, addition of xylose to the culture medium did not change the MIC/MEC to these antifungals (Fig. S4f). Constructs that were integrated at these genomic loci were used to visualize the mitochondria using the CitA₄₀ sequence (49), vacuoles by tagging the Aspergillus oryzae Vam3 homologue (50), the peroxisomes using PTS1-consensus tripeptide SKL (51), and the cell membrane tagging the A. nidulans homologue UapC (52). This single strain, referred to as the tetrachrome strain, allows the visualization of mitochondria (CitA₄₀-Katushka2S), vacuoles (GFP S65T-Vam3), peroxisomes (mTagBFP2-SKL) and cell membrane (UapC-mKO2) simultaneously. We further assessed the potential effect of ectopic expression of GFP S65T tagged Vam3 by simultaneously staining with CMAC, which showed colocalization to the vacuoles as normal (Fig. S3f). Signal from FP expression under the control of the xylose inducible promoter was quantifiable after 1 h and signal increased during the observation period (Fig. S3g). Growth of the tetrachrome strain did not differ from the wild-type strain in different concentrations of xylose (Fig. S3h).

To investigate subcellular dynamics in the tetrachrome strain over a period of 2 h (Fig. 2a), we assessed the four tagged organelles simultaneously using automated analysis (Fig. 2b, individual channels in Fig. S4). The distance between peroxisomes (peroxisome interval) was measured as a proxy for peroxisome abundance to account for growing hyphae. This showed that within the first hour, the peroxisome interval became smaller (0 h median 2.002 µm, 1 h median 1.583 µm) indicating that as hyphae age, peroxisomes become more abundant (P = 0.0059). Within the first 2 h, we observed that mitochondrial fragments became larger (0 h median 4.197 µm, 2 h median 8.622 µm, P = 0.04), which was likely due to cellular growth. The vacuole area remained unchanged.

Next, we investigated the effect of antifungals on the tetrachrome strain. We exposed hyphae grown for 18 h to 2× MIC amphotericin B and monitored the effects on organelles. Amphotericin B exposure significantly reduced the peroxisome interval over 1 h (0 h median 1.615 µm, 1 h median 1.108 µm, P < 0.0001) and 2 h (2 h median 1.276 µm, P < 0.0001). This indicates more peroxisomes per cellular volume. Compared to the wild-type strain, the peroxisomes differed after 1 and 2 h (both P < 0.0001). We observed the mitochondrial fragment size decreased over the course of 2 h (0 h median 5.206 µm, 2 h median 2.473 µm, P = 0.02), but this was not statistically different to the no-drug control (P = 0.87). The vacuole area increased over the same period (0 h median 0.618 µm, 1 h median 1.479 µm, P = 0.0014), which was significantly different to the no-drug control (P = 0.0006). Hyphae treated with amphotericin B for 2 h were visually deflated (Fig. 2a), suggestive of hyphal elements dying from exposure to the drug and organelle degradation.

We further investigated the effect of 2× MIC voriconazole on cellular morphology. An increase in vacuole area was observed (0 h median 0.72 µm, 2 h median 2.54 µm, P = 0.0004) after 1 h, which seemed to plateau at 2 h, seemingly distinct from the vacuole behavior seen for amphotericin B indicating an adaptive response to voriconazole includes compartmentalization of the drug, a toxic by-product resulting from drug action or the recycling of proteins. Additionally, a statistically significant difference to the no-drug control was observed for peroxisome interval after 1 (P < 0.0001) and 2 h (P = 0.0019), and for mitochondrial fragmentation at both time-points (1 h P = 0.01, 2 h P < 0.0001). Upon 2× MIC olorofim exposure, no statistical difference was observed in the peroxisome interval and mitochondrial fragment area. Vacuole area increased after 2 h upon olorofim exposure (0 h median 0.68 µm, 2 h median 2.79 µm, P = 0.0002). Increased vacuole size has previously been documented in response to olorofim exposure (17). Our data reveal a direct correlation between the increase in vacuole size and mitochondrial fragmentation following exposure to voriconazole and amphotericin B. It is not clear whether the apparent fragmentation is caused by a direct effect of the compounds on mitochondrial structure, or an indirect effect of the vacuoles spatially restricting mitochondrial structure. We also observed the colocalization of UapC-mKO2 with the vacuoles during voriconazole treatment. This colocalization also occurs with olorofim and manogepix treatment. This translocation of UapC from the plasma membrane to the vacuolar compartment has been observed in A. nidulans upon ammonium exposure (53). It is not clear what exactly is causing the translocation of UapC to the vacuoles.

Lastly, we investigated the effect of 2× MEC manogepix exposure on subcellular morphology. Surprisingly, no statistical difference in peroxisome interval, mitochondrial fragment area, and overall vacuole area was found. However, we saw a clear change in the distribution of vacuolar size. In the absence of drug, vacuoles exhibited a Gaussian distribution; however, upon the application of manogepix distribution of vacuole size was the binary response (large > 6 µm² or small < 0.3 µm², Fig. 2b). To investigate the response to manogepix in more detail, we looked at individual hyphae over 4.5 h at 15 min intervals (Fig. 3). The addition of drug vehicle, in this case DMSO, did not significantly increase or decrease vacuole or mitochondria fragment area over time (Fig. S5b and c). During manogepix exposure, mitochondrial fragments within individual hyphae became more numerous, and the mean mitochondrial fragment size decreased. At 5 min post-exposure, the mitochondria appeared tubular. There were 60 mitochondrial fragments, and the mean fragment area was 25.09 µm² (±7.32 SEM). Fragmentation of the mitochondrial network was apparent by 20 min and most obvious by 35 min. At 35 min, the mean mitochondrial fragment size was 7.686 µm² (±1.59 SEM), and the number of fragments increased to 147. After 35 min, the number of mitochondrial fragments plateaued (Fig. 3b). Over time, vacuoles appear to fuse. At 5 min manogepix exposure, there were 75 vacuoles, and the average area was 14.97 µm² (±1.56 SEM) (Fig. 3c). At 2 h, vacuole size increased to an average of 46.15 µm² (±6.91 SEM) and the number of vacuoles decreased to 27 (Fig. 3c). After 3 h, vacuole area appears to decrease (Fig. 3a, lower hyphae at 3.5 h). Interestingly, loss of signal is detected in these hyphae in the mitochondria (Katushka2s) channel by 3.5 h. We observed disruptions in the vacuoles, in which they suddenly burst and condensed, potentially releasing their contents (Fig. 3d).