Triazole resistance in Aspergillus fumigatus is recognized as a cause of therapy failure in patients suffering from Aspergillus diseases [1] . Azole-resistance can occur primarily, when azole-resistant spores present in environmental air are inhaled, or secondary in a patient on long-term antifungal therapy. Aspergillus susceptibility testing in routine laboratory practice is therefore warranted. However, its implementation is cumbersome due to the considerable workload and cost. Another problem microbiologist are facing is the fact that at least 50% of clinical isolates are due to contamination or colonization [2] . The probability that a positive A. fumigatus culture represents a case of invasive aspergillosis (IA) was only 22% in a Spanish university hospital [3] . Susceptibility testing by broth microdilution has a slow turn-around-time (48 h after a pure sporulating culture became available, so at least 72 h after sampling the patient). The recognition of azole-resistance is therefore often a late finding in the management of the individual patient, which is especially unfortunate in the setting of IA. As a result, systematic Aspergillus susceptibility testing is mainly executed in specialized centers for patient care or for surveillance reasons. Screening techniques to detect azole-resistance rapidly, with minimal effort and cost, are highly sought. Currently described options include the subculture of Aspergillus isolates on selective, azole-containing, screening agars [4] or molecular strategies [5] [6] [7] . Subculturing isolates on screening agars achieve a time gain of 24 hours and are less labour intensive compared to conventional broth microdilution. These agars are now commercially available. Molecular techniques have the advantage that resistance detection can theoretically be performed directly on culture-negative samples and is fast, but this is labour intensive and expensive: Batching the samples will be necessary to be feasible, which will also creates longer turn-around times. Real-time PCR approaches will also miss new emerging mutations, or mechanisms not involving the CYP51A gene.

Matrix-assisted laser desorption time-of-flight mass spectrometry (MALDI-TOF MS) has rapidly gained ground in clinical laboratories as a routine method for microbial species identification. The main advantages of this approach are the simplicity, low cost and speed of analysis (identification in minutes) [8] . MALDI-TOF MS separates the proteome of a microorganism on their mass-charge ratio, disclosing a characteristic spectrum. Species identification is obtained by matching this spectrum to a library of reference spectra. A generated spectrum never “matches” with absolute identity; the software expresses the degree of similarity.

MALDI-TOF MS has a potential use in the subtyping of strains [9] or in microbial resistance detection [10] , when distinctive and conserved differences in, respectively, the spectra of the subspecies or in the susceptible and resistant strains can be detected. This is mainly described for beta-lactamase detection in gram-negative bacteria and methicillin resistance in Staphylococcus aureus.

The aim of this study was to provide a proof-of-concept that mass spectrometry can be used to differentiate susceptible from resistant A. fumigatus strains, the trypsin digested proteome of three azole-resistant A. fumigatus strains and of three susceptible A. fumigatus strains, were analyzed in detail via UPLC-MS/MS analysis. This allows quantifying and identifying peptides/proteins specific for resistance or susceptibility.

Fungal Isolates for UPLC-MS/MS analysis―A large Aspergillus culture collection is at our disposal at the National Reference Center for Mycosis, University Hospitals Leuven.

UPLC-MS/MS analysis―Three azole-resistant A. fumigatus strains (1 with CYP51A genotype TR46/Y121F/T289A, 2 TR34/L98H) and three azole-susceptible A. fumigatus strains, randomly chosen from the culture collection, were each subcultured in triplicate on diluted Sabauroud slants, incubated at 37˚C for 48 h and each subculture was extracted independently. Proteins were extracted in acetonitrile (ACN) 50%, formic acid (FA) 35%, as described by Bruker Daltonics (Bremen, Germany) and dried in a vacuum operator until dry. The resulting protein extracts (n = 18) were dissolved in 40 µl 2 M urea, 50 mM ammonium bicarbonate and reduced with 0.020 M dithiotreitol for 15 min and subsequently alkylated with 0.050 M iodoacetamide for 30 min in the dark. Then the sample was digested with 0.01 µg trypsin (Sigma Aldrich) overnight at 37˚C. The digestion was stopped by adding trifluoroacetic acid to a final concentration of 0.5%. Peptides were purified with Pierce C18 Spin Columns (Thermo Scientific), according to the manufacturer, vacuum dried and dissolved in 10 µl of ACN 5%, FA 0.1%. UPLC-MS/MS analysis was performed on a Q Exactive Orbitrap mass spectrometer (Thermo Scientific). Five microliter from each sample was injected and separated on an Ultimate 3000 UPLC system (Dionex, Thermo Scientific). The samples were separated using as buffer A water 99.9%, FA 0.1% and B ACN 80%, water 20%, FA 0.1%, using an EasySpray C18 column (Thermo Scientific) with a gradient of 4% to 10% B (6 min) followed by 10% - 35% B (25 minutes), 35% - 65% B (5 min) and a final elution and re-equilibration step at 95% and 5% B respectively. The flow-rate was set at 300 µL/min. The Q Exactive was operated in positive ion mode (nanospray voltage 1.5 kV, source temperature 250˚C). The instrument was operated in data-dependent acquisition (DDA) mode with a survey MS scan at a resolution of 70,000 for the mass range of m/z 400 - 1600 for precursor ions, followed by MS/MS scans of the top 10 most intense peaks with +2, +3 and +4 charged ions above a threshold ion count of 16,000 at 35,000 resolution using normalized collision energy (NCE) of 25 eV with an isolation window of 3.0 m/z, an apex trigger 5 - 15 sec and a dynamic exclusion of 10 s. All data were acquired with Xcalibur 2.2 software (Thermo Scientific).

Protein identification―The LC-MS raw data were imported to Progenesis Nonlinear software (version 4.1) and peaks were detected on all aligned runs. An mgf file was generated via Progenesis and searched using Mascot (version 2.2.04) in a first round against our in-house database containing all the uniprot sequences of Neosartorya fumigata (containing 20,414 accessions) and additionally against the whole fungal database of Swissprot taxonomy fungi (containing 16,473 accessions). Parameters were set at: tryptic digestion, one miscleavage allowed, 10 ppm precursor mass tolerance and 0.02 Da for fragment ion tolerance with a fixed modification of cysteine carbamidomethylation and a variable modification of methionine oxidation. Subsequently files were imported in Scaffold (version 3) combining the Mascot search with Xtandem. Proteins were considered as identified when they met the criteria: min 95% protein, min 1 peptide 95%. FDR at those criteria was calculated as 0.1% at protein level and 0.4% peptide level.

Protein annotation―The fasta files of all the identified proteins (min 95% protein, min 1 peptide 95%) were exported from Scaffold and were subsequently annotated via Blast2go Version 2.7.0 (http://www.blast2go.com/b2ghome). Data containing an Interpro annotation were exported and introduced in cytoscape (version 3.0.2) to visualize related proteins.

Peptide/protein quantification―As indicated above the LC-MS raw data were imported to Progenesis Nonlinear software and normalized. Peptides were considered as significantly different between the resistant and susceptible condition, when ANOVA p < 0.05. Protein abundance was calculated via progenesis by considering only the peptides with no conflicts. Proteins were considered as significantly different when ANOVA p < 0.05.

Blind clustering of the proteomes―Protein abundances of the ANOVA significant proteins were exported from Progenesis and imported into Statistica 8 (Nine sigma) to perform a Pincipal Component Analysis (PCA) (NonLinear Iterative Partial Least Squares NIPALS algorithm). Scores were exported and visualized using Microsoft Excel.

UPLC-MS/MS analysis―A total of 7899 tryptic peptides were detected, of which 22.7% (1792 peptides) had significantly different abundances (p < 0.050) between the resistant and susceptible strains. Only 2082/7899 (26.4%) peptides were identified, belonging to 553 proteins when matched against all species in swissprot. A blind clustering of the most important proteins using Principle Component Analysis (PCA) shows both sample groups can be separated (Figure 1). Principle component 1 (PC1) explains 44% of the observed variability and PC2 12%. The proteins with confident identification (defined as a confidence score ≥ 40.0) which count at least one peptide with significantly differing abundance between the susceptible and resistant strains (112 proteins) are listed in Supplemental Table S1. Among these proteins, 16% (18/112) are ribosomal proteins, 14% (16/112) are involved in stress response or oxidation-reduction, 12% (13/112) in carbohydrate metabolic processes-including four alpha-1,2- mannosidases. Another 5% (6/112) are specifically involved in glucan metabolism. Proteins with uncharacterized function represent 19.6% (22/112). A Pubmed literature search and Aspergillus Genome Database search (AspGD, http://www.aspergillusgenome.org/) was performed for every protein with (a)

significantly different abundant peptide (s) between susceptible and resistant strains (or its orthologs), to evaluate for a known role in virulence, host response, diagnostic properties or antifungal susceptibility. For 29 proteins (26%), relevant information was obtained (Supplemental Table S1); twelve interesting proteins are highlighted in Table 1. No peptides of lanosterol-5α-demethylase, the target protein of azole therapy (encoded by CYP51A), were identified from azole-resistant or-susceptible strains and the known differences in this protein are therefore no contributing factor in the proteomic differences observed here.

To the best of our knowledge, this is the first study evaluating proteomic differences between triazole susceptible and resistant A. fumigatus isolates based on UPLC-MS/MS analysis. Our approach of comparative proteome analysis provided proof-of-concept that significant proteomic differences exist. These differences were larger than expected, which indicates that susceptible and resistant A. fumigatus probably accumulated mutations over time. However, only a limited fraction of the differentiating peptides could be identified, which demonstrates the constraints of the current databases. Significant abundancy of a protein in one condition can mean that this protein is indeed less abundant in the other condition, but can also mean that certain peptides of this protein bear mutations/polymorphisms and are therefore not identified in the second condition (independent of their abundancy). Among the proteins which have at least one peptide with significantly different abundance between susceptible and resistant strains, about one out of four proteins (or its orthologs) are known to be relevant in azole resistance, virulence or host response (Supplemental Table S1).

^$Phylome DB database identification [15] . *Condition (susceptible (S) or resistant (R)) with significant abundance of at least one peptide (p < 0.05). The p-values express the minimal level of significance for abundance at the peptide level.

This illustrates the power of comparative proteome analysis to identify interesting targets for research into antimicrobial resistance. Among the differing components, several mitochondrial proteins were detected, involved in stress response (e.g. antigenic mitochondrial protein HSP60, a mitochondrial superoxide dismutase) and also cofilin, which is suggested to play a role in the regulation of mitochondrial function and stress responses and is linked to multi-drug resistance [26] . These data support the hypothesis that mitochondrial activity effects triazole tolerance [27] [28] . Secondly, several conidial proteins (e.g. RodA, RodB, FleA, Arb2, Con-10) and cell-wall modifying enzymes (e.g. glucanases Exg9, EgIC) were also found with significantly different abundances between susceptible and resistant strains. Overall, the identification of many conidial proteins is to be expected as proteomic studies were performed on sporulating strains. A different sporulation rate between resistant or susceptible strains could be an explanation for these different abundances, but could not be objectified visually. A third interesting observation is that many ribosomal proteins are present among the differentiating proteins, which are considered highly conserved intraspecies. This could indicate that the proteome differences reflect a common genomic background of the strains which evolved to azole-resistance. MALDI-TOF MS instruments in clinical laboratories detect proteins in the range of 2000 - 14,000 m/A, which is known to correspond largely with the ribosomal protein fraction.

In conclusion, we proved the presence of substantial proteomic differences between azole-susceptible and azole-resistant A. fumigatus strains. We believe that our data provide interesting new options for research into A. fumigatus resistance.

JM reports grants and personal fees from Pfizer, grants and personal fees from MSD, personal fees from Astellas, personal fees from Gilead, outside the submitted work. KL reports grants and personal fees from Pfizer, grants and personal fees from Gilead Sciences, grants and personal fees from Merck, outside the submitted work. EV, SC, OK and MS have nothing to disclose.

Vermeulen, E., Carpentier, S., Kniemeyer, O., Sillen, M., Maertens, J. and Lagrou, K. (2018) Proteomic Differences between Azole-Susceptible and -Resistant Aspergillus fumigatus Strains. Advances in Microbiology, 8, 77-99. https://doi.org/10.4236/aim.2018.81007

Vermeulen, E., Carpentier, S., Kniemeyer, O., Sillen, M., Maertens, J. and Lagrou, K. (2018) Proteomic Differences between Azole-Susceptible and -Resistant Aspergillus fumigatus Strains. Advances in Microbiology, 8, 77-99. https://doi.org/10.4236/aim.2018.81007