Aspergillus fumigatus is a ubiquitous saprotroph and human-pathogenic fungus that is life-threatening to the immunocompromised. Triazole-resistant A. fumigatus was found in patients without prior treatment with azoles, leading researchers to conclude that resistance had developed in agricultural environments where azoles are used against plant pathogens. Previous studies have documented azole-resistant A. fumigatus across agricultural environments, but few have looked at retail plant products. Our objectives were to determine if azole-resistant A. fumigatus is prevalent in retail plant products produced in the United States (U.S.), as well as to identify the resistance mechanism(s) and population genetic structure of these isolates. Five hundred twenty-five isolates were collected from retail plant products and screened for azole resistance. Twenty-four isolates collected from compost, soil, flower bulbs, and raw peanuts were pan-azole resistant. Resistant isolates had the TR₃₄/L98H, TR₄₆/Y121F/T289A, G448S, and H147Y cyp51A alleles, all known to underly pan-azole resistance, as well as  WT alleles, suggesting that non-cyp51A-mechanisms contribute to pan-azole resistance in some isolates. Minimum spanning networks showed two lineages containing isolates with TR alleles or the F46Y/M172V/E427K allele, and discriminant analysis of principle components (DAPC) identified three primary clusters. This is consistent with previous studies detecting three clades of A. fumigatus and identifying pan-azole-resistant isolates with TR alleles in a single clade. We found pan-azole resistance in U.S. retail plant products, particularly compost and flower bulbs, which indicates the risk of exposure to these products for susceptible populations and that highly resistant isolates are likely distributed worldwide on these products.

Collection and isolation of A. fumigatus from retail plant products

Eight product types were collected from retail stores from September 2019 through April 2021 and surveyed for azole-resistant A. fumigatus. Grapes, apples, almonds, pecans, peanuts, compost, soil, and flower bulbs (tulip, daffodil, Gladiolus, daylily, Dahlia, Canna, Liatris, Caladium, lily of the valley, Clematis, Iris ensata, and magnum elephant ears) were collected from eight retail grocery stores and nine garden centers in the Athens, Georgia, area. The date of collection for fruits and nuts depended on their fresh market season in the U.S. Each product was collected by purchasing a pre-packaged bag. A wide variety of brands, product sizes, and retail stores were sampled. An effort was made to collect products that originated in the U.S; however, some packages of compost and flower bulbs originated from Canada, Costa Rica, and the Netherlands, respectively.

Grapes were first sampled using amended sampling methods (39, 42) where parts of individual grapes including the skin and stems were swabbed or plated directly onto Sabouraud dextrose agar (SDA) amended with 50 µg/ml rose bengal dye, 50 µg/ml chloramphenicol, and 5 µg/ml gentamicin (aSDA) (5). However, this resulted in low yields of A. fumigatus and prevented the entire sample from being tested efficiently, therefore, the following method was devised that was able to be used on grapes, peanuts, pecans, almonds, apples, and flower bulbs not packed in soil. The contents of the entire bag of the purchased product ranging from approximately 0.5 to 2 kg were placed into a sterile 30  38 cm polypropylene bag and rinsed thoroughly with 50 ml sterile 0.05% Tween-20 by massaging the bag to distribute the liquid over the plant products. All grapes contained stems and some peanuts and pecans were unshelled. All almonds were shelled. The liquid was collected into a 50 ml conical tube by cutting one corner of the bag. Almonds and pecans contained large quantities of broken shells, skin, and dust so the liquid was strained through sterile cheesecloth during collection. The tube was centrifuged at 3000 g for 5 min (49) and the supernatant poured off. After observing that pellets from some products contained less A. fumigatus on the surface (grapes, pecans, and almonds), these samples were resuspended in 1 ml of 0.05% Tween-20, and 100 µl aliquots of the entire solution were plated and spread onto 5-15 Petri plates of aSDA so that we could obtain multiple isolates of A. fumigatus. Pellets from products observed with more abundant A. fumigatus (peanuts and flower bulbs not packed in soil), were resuspended in 2.5 ml 0.05% Tween-20 solution and distributed in 100 µl aliquots among 10 plates of aSDA, 10 plates of aSDA amended with 3 µg/ml itraconazole (ITC, Thermo Sci Acros Organics, New Jersey, USA), and 10 plates of aSDA amended with 3 µg/ml tebuconazole (TEB, TCI America, Oregon, USA).

Flower bulbs packed in soil contained very abundant amounts of A. fumigatus so they were processed differently. Bulbs were removed from the soil and placed into a separate bag, shaking off as much soil as possible first. The bulbs were rinsed with 0.05% Tween-20 as described above. The liquid collected into a 50 ml conical tube was immediately vortexed at maximum speed, and 3 ml were transferred to another tube, avoiding any soil that may have been collected in the liquid. The liquid was not centrifuged, unlike with other samples, because the amount of A. fumigatus present was so great that the sample did not need to be concentrated. One and a half milliliters of the suspension were spread plated in 100 µl aliquots onto 5 plates of aSDA, 5 plates of aSDA amended with 3 µg/ml ITC, and 5 plates of aSDA amended with 3 µg/ml TEB. The remaining liquid in the tube was diluted 1:2 with fresh 0.05% Tween-20, and 1.5 ml in 100 µl aliquots was spread plated amongst 5 plates of aSDA, 5 plates of aSDA amended with 3 µg/ml ITC, and 5 plates of aSDA amended with 3 µg/ml TEB. The additional dilution was plated in case the initial sample plating was not dilute enough to visualize individual colonies of A. fumigatus.

Compost and soil were sampled using an amended sampling method (5, 32). Briefly, 2 or 4 g of the material was suspended in 0.1 M sodium pyrophosphate. Only 2 g of soil and compost with lower densities were collected in order to accommodate the size limit of the 50 mL conical tubes, otherwise, 4 g were collected. The suspensions were vortexed for 30 s and allowed to settle for 1 min. Two and a half milliliters of the supernatant was plated as described above for peanuts and flower bulbs not packed in soil.

All agar plates were incubated at 45°C for 2 to 4 days. Individual colonies that were confirmed as A. fumigatus based on morphology were quadrant-streaked onto SDA to obtain a single-spore culture. All colonies were preliminarily screened for azole resistance by quadrant-streaking onto SDA with 3 µg/ml of TEB and SDA with 3 µg/ml of ITC alongside known resistant and susceptible control isolates (5). Plates were incubated at 37°C for 2 days, after which resistance to TEB and ITC was preliminarily scored as resistant, intermediate, or susceptible based on a visual assessment of growth. Isolates scored as resistant had proficient growth in at least two quadrants of an aSDA plate; intermediate had spotty growth in one quadrant of an aSDA plate; and susceptible did not grow at all. Azole sensitivity was then quantified as described in the section below. For long-term storage, a single-spore colony was selected from SDA and streaked onto complete media (50) using a sterile cotton swab. For each isolate, conidia from complete media slants incubated for 2 days at 37°C were harvested, suspended in 15% glycerol in cryotubes, frozen in liquid nitrogen, and stored at -80°C.

Azole-resistance phenotyping

One-hundred-eleven of the isolates preliminarily scored as resistant and 19 of the isolates preliminarily scored as susceptible or intermediate in the screening were selected for azole-resistance phenotyping via minimum inhibitory concentration (MIC) assays to the fungicide TEB, and clinical antifungals ITC, voriconazole (VOR; Thermo Sci Acros Organics, New Jersey, USA), and posaconazole (POS; Apexbio Technology, Texas, USA). Isolates representing a variety of products and samples with elevated growth on azole-amended SDA were selected and assessed using the Clinical Laboratory Standard Institute broth microdilution method (51). Not all isolates were assayed at this stage since many had similar phenotypes from the same sample, such as the same bag of compost or flower bulbs. Some susceptible isolates were included for comparison. Briefly, conidia were harvested from 4-day-old complete media slants using 3 ml sterile 0.05% Tween-20. The spore suspensions were adjusted to 0.09 to 0.13 OD at 530 nm using a spectrophotometer, and 20 µl of suspension was added to 11 ml RPMI 1640 liquid medium (Thermo Sci Gibco, California, USA). The solution was distributed in 100 µl aliquots among 96-wells in microtiter plates containing two-fold serial dilutions of antifungals with the final concentrations ranging from 0.015625 µg/ml to 16 µg/ml. The plates were incubated at 37°C for 48 hr. The MIC (minimum inhibitory concentration) of each isolate to each antifungal was determined visually by selecting the first well that had no fungal growth; the corresponding concentration of the antifungal in that well was the MIC. The accuracy of the MIC assays was checked using both susceptible and resistant A. fumigatus control isolates with known MIC values for TEB, ITC, VOR, and POS. The EUCAST breakpoints defined in February 2020 (52) were used to classify isolates as sensitive (S) or resistant (R): ITC S ≤ 1 µg/ml > R, VOR S ≤ 1 µg/ml > R, and POS S ≤ 0.125 µg/ml and 0.25 µg/ml > R. MIC values of 2 µg/ml for ITC and VOR and 0.25 µg/ml for POS are classified as areas of technical uncertainty (ATU) meaning that treatment with these antifungals may be used for isolates with this resistance breakpoint under certain situations, but for this study they were considered resistant (52). A breakpoint cutoff for TEB was defined as > 2 µg/ml according to previous studies (5). Antifungal resistance phenotypes were further classified into four categories: azole-susceptible, TEB-resistant, pan-azole-resistant, and azole-resistant. Isolates with no azole resistance were classified azole-susceptible. Isolates with a TEB-resistant phenotype but no resistance to any medical azole were classified TEB-resistant. Isolates that were either resistant to only one medical azole or one medical azole and TEB were classified as azole-resistant. Finally, isolates with resistance to more than one medical azole were classified as pan-azole-resistant.

DNA extraction

Hyphae of 102 isolates, including 80 isolates screened by MIC assays and 22 susceptible isolates representing a variety of the sampled products, were grown from spores in liquid complete medium for 16 to 20 hr at 30°C in a 1 g orbital shaker (5). Tissue was gathered by filtering through a 40 µm cell strainer and squeezing the residual liquid from the tissue using a sterile cotton swab. Approximately 100 to 200 mg of tissue was collected and set aside in 2-ml tubes. DNA extractions were performed according to the QIAGEN DNeasy Plant Mini Kit protocol (QIAGEN, Maryland, USA), with a few amendments (53). Briefly, buffer AP1 was warmed to 65°C for at least 10 min prior to use. Four hundred microliters buffer AP1 and 4 µl RNase A were added to each tube with fungal tissue and vortexed for at least 2 min until all tissue was suspended. Each sample was incubated at 65°C for 10 min, vortexing for 10 seconds three times throughout the incubation. One hundred thirty microliters of buffer P3 were added to each sample, vortexed, and then incubated at -20°C for 5 min. The samples were centrifuged for 5 min at 18,407 g to pellet the remaining solids. The supernatant was pipetted into a QIAshredder mini spin column and centrifuged for 2 min at 18,407 g. Five hundred microliters of the flow-through fraction were transferred to a new sterile 2-ml lock-lid tube and 750 µl buffer AW1 (1.5 volume) was added to the flow-through and immediately mixed by pipetting. Six hundred fifty microliters of this solution were pipetted into a DNeasy mini spin column and centrifuged for 1 min at 18,407 g. The flow-through was discarded, and the previous step was repeated with the remaining solution. The DNeasy spin column was transferred into a fresh 2 ml tube, 500 µl buffer AW2 was added, and the column was centrifuged for 1 min at 18,407 g. The flow-through was discarded, and the previous step was repeated except the centrifugation lasted for 2 min rather than 1 min. The DNeasy spin column was transferred to a sterile 1.5 ml tube, and 50 µl buffer AE was added to the membrane. The columns were incubated at room temperature for 5 min, then centrifuged at 18,407 g for 1 min. This step was repeated once, and the DNA was stored at 4°C. The DNA concentration was quantified using NanoDrop One (Thermo Sci, New Jersey, USA).

cyp51A sequencing

Seventy-nine isolates, including 37 sensitive and 42 resistant isolates based on MIC assays, were selected for Sanger sequencing of the promoter and coding regions of cyp51A. These isolates were selected in order to obtain representatives of isolates with varying phenotypes from all sampled plant-based retail products. The selection of isolates was based primarily on elevated MIC values, but some susceptible isolates were included as well for comparison. Isolates with low TEB MIC values and isolates with similar MIC values from the same sample were not always included. PCR was performed using a mix of 12.5 µl OneTaq 2 Master Mix, 6.5 µl RNA-free sterile ddH2O, 2 µl each of previously designed forward primer 5´-CGGGCTGGAGATACTATGGCT-3´ and reverse primer 5´-GTATAATACACCTATTCCGATCACACC-3´ (5, 54). PCR cycling conditions were as follows: 98°C for 2 min followed by 30 cycles of 98°C for 15 sec, 62°C for 15 sec, and 72°C for 2.5 min, followed by a final extension at 72°C for 5 min (5). Sanger sequencing was performed by Genewiz (Genewiz by Azenta Life Sciences, Massachusetts, USA) using 4 primers: 5´-GCATTCTGAAACACGTGCGTAG-3´, 5´-GTCTCCTCGAAATGGTGCCG-3´, 5´-CGTTCCAAACTCACGGCTGA-3´, and 5´-GCGACGAACACTTCCCCAAT-3´ (5). Sequence alignment was performed using Geneious v2019.2 (Biomatters, Auckland, NZ). Briefly, all sequences were trimmed to remove low-quality base pairs with a base quality score <20 from the beginning and end of each sequence. The sequences were aligned for each isolate and the consensus sequence was visually assessed. The promoter regions were aligned and compared with A1163 genomic sequence v43 from Ensembl (55). The coding sequences were translated and aligned to the A. fumigatus A1163 Cyp51A protein (GenBank accession EDP50065).

STRAf genotyping

Single tandem repeats of A. fumigatus (STRAf) are microsatellite markers commonly used to assess genetic diversity and population genetic structure among isolates from different environmental origins, such as clinical or agricultural environments (56-59). Nine previously-developed STRAf markers (STRAf2A, 2B, 2C, 3A, 3B, 3C, 4A, 4B, and 4C) were used to genotype 95 isolates collected in this study: 72 of which we had collected cyp51A sequence data and 23 others that were included to investigate isolates from all of the sampled plant-based retail products (57). Multiplex PCR was performed using a modified protocol for the Type-it Microsatellite PCR kit (Qiagen). Briefly, each of the three multiplex reactions (3 loci per multiplex) contained 5 µl 2 Type-it Master Mix, 1 µl 10 primer mix (2 µM of each of the 6 multiplex primers), 1 µl DNA template, and RNAse-free water. Thermal cycling conditions were as follows: 95°C for 5 min followed by 28 cycles of 95°C for 30 sec, 57°C for 90 sec, 72°C for 30 sec, and a final elongation of 60°C for 30 min. Amplification of several PCR products from each multiplex was confirmed by electrophoresis on a 1% agarose gel with 1× TBE buffer. The PCR products were diluted 1:15 and then sent to the Cornell Institute of Bioinformatics (Ithaca, New York, USA) for addition of the internal size standard Genescan-500 Liz and HiDi-formamide, followed by fragment analysis on an Applied Biosystems 3730x1 96-capillary DNA analyzer. The data were analyzed using the Microsatellite plugin in Geneious v.6 (Biomatters, Auckland, NZ) to identify the nine loci and the amplicon length (or allele) in each sample.

Population genetic analyses

Multilocus genotypes used in the analyses were based on the STRAf data for the 95 isolates from this study and 80 clinical and environmental isolates from the U.S. from a previous study (5). There were 28 isolates from clinical settings and environmental isolates came from agricultural compost (4) and soil with plant debris where apple (2), watermelon (7), strawberry (4), pecan (13), peanut (16), and grape (5) were growing. The laboratory reference used in the previous study (5), Af293, was included as a control for comparison. Isolates from a variety of substrates were chosen to obtain a representative sample, but substrates associated with the retail products sampled in this study were prioritized (e.g., grape, apple, peanut soil, and debris).

To estimate the genetic relatedness among isolates, minimum spanning networks using Bruvo’s genetic distance model (60) and Nei’s 1978 distance (61, 62) were constructed with the Poppr package in R (63). Bruvo’s genetic distance is often useful for analyses based on microsatellite markers (60) and assumes a stepwise mutation model that may not be entirely accurate for an organism like A. fumigatus which is genotypically diverse and known to sexually reproduce (64). Moreover, each STRAf locus contains 11 to 37 alleles that vary in repeat number, so they may not be evolving in a stepwise manner (57). Therefore, we used Nei’s genetic distance, as well, to incorporate an infinite alleles model. Population genetic structure was analyzed using discriminate analysis of principal components (DAPC) in R (65) to identify if populations clustered based on environmental setting, substrate of origin, cyp51A genotype, geographic sampling location, or another factor. Clusters were determined using K-means clustering of principal components.