All clinical isolates used in this study were preserved at Chiba University Medical Mycology Research Center (Chiba City, Japan). We obtained 19 isolates from a patient at hospital C who had CPA (Tables 1, 2), 3 isolates (IFM 63768, IFM 63666, and IFM 63772) from another CPA patient at hospital C on different days, and additional isolates from 3 other patients with CPA at different hospitals (IFM 62140 at hospital K, IFM 64258 at hospital I, and IFM 64303 at hospital T). We also obtained 16 azole-susceptible clinical isolates (Table 3) from respiratory samples (sputa, pleural effusion, lung tissues, or bronchoalveolar fluids) or pus from a subcutaneous abscess from patients with aspergillosis in various hospitals in Japan.

For genotyping, we PCR amplified and sequenced 6 loci of ≈400 bp from all of the isolates as described previously (15). We counted the repeat numbers of each region (2A, 2B, 2C, 3A, 3B, 3C, 4A, 4B, and 4C) from the sequence.

We performed antifungal susceptibility testing as described previously (16). In brief, we performed tests in triplicate using amphotericin B (AMPH), itraconazole (ITCZ), voriconazole (VRCZ), and posaconazole (PSCZ) in RPMI 1640 medium (pH 7.0) at 35°C, according to the Clinical and Laboratory Standards Institute reference method for broth microdilution (https://clsi.org/standards/products/microbiology/documents/m38), with partial modifications using the dried plate for antifungal susceptibility testing (Eiken Chemicals, Tokyo, Japan).

To detect mutations in cyp51A, erg6, and hmg1 genes, we performed PCR amplification and sequenced these regions using appropriately designed primers (Technical Appendix Table 1). We performed sequence variant detection by comparison with reference sequences from GenBank and Fungi DB (http://fungidb.org/fungidb/) (GenBank accession no. AF338659 for cyp51A, Fungi DB accession nos. AFUB_020770 for hmg1, and AFUB_099400 for erg6).

We performed whole-genome sequencing by using next-generation methods as described previously (15). In brief, we extracted genome DNA from overnight-cultured mycelia by DNeasy Plant Mini Kit (QIAGEN, Hilden, Germany) or by phenol-chloroform extraction and Nucleobond AXG column (TaKaRa, Kusatsu, Japan) with Nucleobond Buffer Set III (TaKaRa). For analyzing IFM 60814 and IFM 63240, we prepared a fragmented DNA library by using a Nextera DNA sample preparation kit according to the manufacturer’s instructions (Illumina, San Diego, CA). The mean length of the libraries was 376–674 bp. We performed sequencing in a paired-end 2 × 150 bp mode on a MiSeq system (Illumina). For analyzing IFM 63666 and IFM 63768, we prepared a fragmented DNA library using KAPA HyperPlus Library Preparation Kit (Kapa Biosystems, Wilmington, MA) according to the manufacturer’s instructions. We performed sequencing in a paired-end 2 × 300 bp mode on a MiSeq system (Illumina).

To search for single-nucleotide polymorphisms (SNPs) between IFM 60814, the first isolate obtained, and IFM 63240–63243, the third set of isolates, we performed read-mapping and SNP detection using CLC Genomics Workbench (CLC bio, Aarhus, Denmark). In brief, we trimmed the reads from each isolate and mapped them to the Af293 reference genome. By comparing them with the Af293 sequence, we detected sites with SNPs in IFM 60814, IFM 63240, IFM 63241, IFM 63242, and IFM 63243. We also compared the sets of varied sites between IFM 60814 and IFM 63240–63243. We considered the unique sites common in IFM 63240–63243 to be the SNPs between the first isolate and the third set of isolates.

We cleaned the read datasets of IFM 63666 and IFM 63768 using Trimmomatic version 0.33 (17). We trimmed reads to generate de novo assembly of the draft genome of IFM 63666 using Platanus version 1.2. 4 with default parameters (18). We performed annotation of all predicted open reading frames of the draft genome using AUGUSTUS version 2.5.5 (19). For nucleotide differences between IFM 63666 and IFM 63768, we mapped the trimmed reads of IFM 63768 to the draft genome of IFM 63666 using SMALT version 0.7.6 (http://www.sanger.ac.uk/science/tools/smalt-0). We identified SNPs using SAMtools version 0.1.19–44428cd (20) and filtered with >10-fold coverage, >30 mapping quality, and 75% consensus using in-house scripts (21,22). We annotated the functional effect of SNPs with SnpEff version 4.1l (23).

We performed disk diffusion assay using the protocol described by Qiao et al. (24). We spread a 200-µL suspension (1 × 10⁶ conidia/mL) of each of the tested isolates uniformly onto YGA (glucose 2%, yeast extract 0.5%, agar 2%, trace element). We impregnated blank paper disks 8 mm in diameter with 2.5 µg amphotericin B, 100 µg nystatin, or 40 µg lovastatin and placed the disks onto the center of the inoculated agar plates. We incubated the plates at 37°C and measured the diameters of the zones of inhibition 48 hours later. We performed each assay 3 times and calculated the mean diameter.

We conducted total ergosterol extraction using the protocol described by Arthington-Skaggs et al. and Alcazar-Fuolietal (25,26). We cultured A. fumigatus isolates for 20 h in glucose minimal medium. We harvested mycelia by filtration, washed them with sterile water, and then dried and weighed them. We added 3 mL of 25% alcoholic potassium hydroxide solution to dried mycelia and mixed for 1 min, then incubated the mixture in an 80°C water bath for 1 h. After incubation, we added water and 3 mL of pentane and mixed for 3 min. We transferred the upper pentane layer to a clean glass tube for evaporation. We dissolved the dried samples in 1 mL of methanol and filtered it through a 0.22-µm pore size membrane filter. We analyzed ergosterol content using the Shimadzu LC-20A system (Shimadzu, Kyoto, Japan) with COSMOSIL 5C¹⁸-MS-II column (4.6 mm ID × 100 mm; Nacalai Tesque, Kyoto, Japan). We established a flow rate of 1 mL/min of acetonitrile to water (95:5 vol/vol). We used peak areas and heights recorded at a 254-nm wavelength in an RF-20Axs (Shimadzu) for quantification, expressing total ergosterol concentration as µg ergosterol per mg fungal dry weight. We repeated each experiment >3 times.

To construct transformants expressing erg6, hmg1, or cyp51A with the mutation, we cloned the alleles using the shuttle vector pPTR I (Takara Bio, Otsu, Japan) and GeneArt Seamless Cloning and Assembly Kit (Invitrogen, Tokyo, Japan). Primers used for the cloning are listed in Technical Appendix Table 2. We used genome DNAs of IFM 60814 (for wild-type [WT] gene) and IFM 63240 (for mutated allele) as templates to clone the alleles. The resultant plasmids were pPTRI-erg6^WT, pPTRI-hmg1^WT, pPTRI-cyp51A^WT, pPTRI-erg6^A350T, pPTRI-hmg1^S269F, and pPTRI-cyp51A^G448S, which we used for transformation of isolate IFM 60814. After selection with pyrithiamine, we verified the sequences of each gene of the candidate transformants by Sanger sequencing.

We recovered a total of 19 A. fumigatus isolates from 1 patient on 9 testing dates during September 2011–May 2016 (Tables 1, 2). Microsatellite analysis showed almost identical short tandem repeats across the isolates, derived from a genetically clonal background. The patient started treatment with VRCZ after the first isolation of A. fumigatus in September 2011. Whereas the first isolate (IFM 60814) was susceptible to antifungal drugs including ITCZ, VRCZ, and PSCZ, later isolates showed resistance to the azoles (MICs: ITCZ, 4 to >8; VRCZ, 8 to >8; PSCZ, 4 to >8). By sequencing the cyp51A gene, we found amino acid substitution G448S IFM 62916 (second isolation date), IFM63248 (fourth isolation date), IFM63595 (sixth isolation date), IFM63712 (seventh isolation date), IFM63713 (seventh isolation date), and IFM64139 (eighth isolation date). These isolates were most likely resistant to VRCZ as a result of the G448S mutation, which has been reported to confer VRCZ resistance in A. fumigatus. However, the multiazole resistance mechanism in the isolates with no G448S mutation remained unknown. The growth of isolate IFM 62916 was markedly impaired with glucose minimal medium and potato dextrose agar, whereas the isolates from the third isolation date also showed a moderately delayed growth compared with those of the first strain (Technical Appendix Figure 1).

To gain insight into the multiazole resistance mechanism in the non-cyp51A strains, we sequenced the genomes of the isolates from the first (IFM 60814) and third (IFM 63240–63243) testing dates by next-generation sequencing. Compared with IFM 60814, IFM 63240–63243 commonly showed 7 nonsynonymous mutations, including 6 SNPs (4 amino acid substitutions and 2 nonsense mutations, which generate termination codon) and a 1-bp deletion resulting in a frame shift of the protein (Technical Appendix Table 2). The 6 genes with mutations included erg6, encoding sterol 24-C-methyltransferase and hmg1, encoding hydroxymethylglutaryl-CoA (HMG-CoA) reductase. The mutation of erg6 (A350T) resided in a Sterol_MT_C domain (PF08498) in the C-terminus, whereas the mutation of hmg1 (S269F) was located at the beginning of the sterol-sensing domain (PF12349) (Technical Appendix Figure 2). Because both of these genes are functionally related to the ergosterol biosynthesis pathway, we assumed that these mutations affect ergosterol biosynthesis and the consequential azole resistance in the strains.

We examined the rest of the isolates from the patient for mutations in the hmg1 and erg6 genes. Despite the presence of A350T, we identified E49K in the erg6 gene in several isolates, whereas Hmg1 S269F existed in all isolates tested (Tables 1, 2). These results indicate that all of the isolates showing multiazole resistance had mutations in both hmg1 and erg6 genes, regardless of the mutation in cyp51A gene.

We investigated the phenotypes associated with the ergosterol biosynthesis pathway in IFM 63240, a representative of the isolates from the third testing day. Disk diffusion assay showed that this isolate was more sensitive to polyene drugs, amphotericin B, and nystatin, suggesting that ergosterol production might be differentially regulated (Figure 1). The ergosterol content in the cells, measured by HPLC, showed significant increase (p = 0.0145) in the third set of isolates compared with the first isolate (Figure 2, panel A). At the same time, we found no differences between the first and third isolates regarding sensitivity to terbinafine, which interferes with the early stage of ergosterol biosynthesis, and fenpropimorph, which inhibits ergosterol biosynthesis (data not shown). Of note, the isolate from the third set showed increased sensitivity to lovastatin, an inhibitor of HMG-CoA reductase (Figure 1).

Considering the different level of ergosterol biosynthesis in the third set of isolates, we wanted to determine whether the mutation in erg6 and/or hmg1 genes was responsible for the phenotypes (i.e., sensitivity to polyenes and resistance to azoles). To address this question, we constructed the isolates carrying these mutant alleles as well as the corresponding wild-type gene by introducing erg6^A350T or hmg1^S269F into the first strain (hereafter designated as 1st+erg6^A350T and 1st+hmg1^S269F). These mutant alleles were ectopically expressed in the transformed strains. We also constructed the strain carrying cyp51A^G448S (1st+cyp51A^G448S) for verification of the role of the mutation in azole resistance as previously reported (27). The MICs for azoles were determined in these strains (Table 4), showing that introducing erg6^A350T or hmg1^S269F did not affect susceptibility to azoles, whereas the cyp51A^G448S conferred resistance to VRCZ.

We studied whether the other non-cyp51A azole-resistant isolates possess mutations in erg6 or hmg1. Over the past decade, we have collected clinical azole-resistant A. fumigatus strains in Japan: 4 non-cyp51A azole-resistant isolates (IFM 62140, IFM 64258, IFM 64303, and IFM 63768) showing a VRCZ MIC of >8 (Table 5). Each isolate was from a patient who had a history of treatment with VRCZ before isolation. Although IFM 64303 has no mutations in erg6 and hmg1 genes, the other 3 isolates possessed different mutations in erg6 and/or hmg1. IFM 62140 had erg6^A75S and hmg1^F261_del, IFM 63768 had hmg1^S269Y, and IFM 64258 had erg6^V206A and hmg1^F390Y, which were different from the pattern observed in IFM 63240. These results suggested a potential link between azole resistance and the mutations in erg6 and hmg1 genes. Notably, IFM 63768 and IFM 63240 had a mutation in Hmg1 at S269 in which serine was substituted by different amino acids (i.e., Hmg1^S269Y and Hmg1^S269F). To investigate the possibility that such mutations are just frequently occurring polymorphisms, we sequenced erg6 and hmg1 in 16 azole-susceptible isolates. We did not find mutation in erg6, although 1 of the 16 isolates, IFM 60901, had a hmg1^H564Y mutation (Table 3). We did not find the mutations hmg1^S269F/Y, hmg1^F261_del, and hmg1^F390Y in the azole-susceptible isolates. These results suggest that the mutations in erg6 and hmg1 found in the azole-resistant isolates are not mere polymorphisms in A. fumigatus but can be implicated in azole resistance.

In our collection, an azole-susceptible isolate (IFM 63666) was isolated 2 months before isolation of azole-resistant IFM 63768 (hmg1^S269Y) from the same patient. The 2 isolates shared the same short tandem repeats, indicating a clonal lineage. We then sequenced the genomes of IFM 63768 and IFM 63666. Comparing the genome sequences, we discovered only 1 nonsynonymous mutation other than hmg1^S269Y in IFM 63768, which was I395M in Afu4g04290 encoding a putative histone H4 deacetylase protein. Although we could not see the effect of the mutation in Afu4g04290, this result supported the possibility that Hmg1^S269Y is the determinant for azole resistance in this isolate. We compared ergosterol levels in the cells between IFM63768 and IFM 63666 and found an increased amount in IFM 63768 with hmg1^S269Y (Figure 2, panel B). We isolated IFM 63772 from the same patient in addition to these isolates, and it showed moderate resistance to VRCZ, resistance to PSCZ, and no mutations in erg6 and hmg1 (Table 5). The amount of ergosterol in IFM 63772 was comparable to that of IFM 63666 (Figure 2, panel B).