Genomic Epidemiology Identifies Azole Resistance Due to TR34/L98H in European Aspergillus fumigatus Causing COVID-19-Associated Pulmonary Aspergillosis

Aspergillus fumigatus has been found to coinfect patients with severe SARS-CoV-2 virus infection, leading to COVID-19-associated pulmonary aspergillosis (CAPA). The CAPA all-cause mortality rate is approximately 50% and may be complicated by azole resistance. Genomic epidemiology can help shed light on the genetics of A. fumigatus causing CAPA, including the prevalence of resistance-associated alleles. We present a population genomic analysis of 21 CAPA isolates from four European countries with these isolates compared against 240 non-CAPA A. fumigatus isolates from a wider population. Bioinformatic analysis and antifungal susceptibility testing were performed to quantify resistance and identify possible genetically encoded azole-resistant mechanisms. The phylogenetic analysis of the 21 CAPA isolates showed that they were representative of the wider A. fumigatus population with no obvious clustering. The prevalence of phenotypic azole resistance in CAPA was 14.3% (n = 3/21); all three CAPA isolates contained a known resistance-associated cyp51A polymorphism. The relatively high prevalence of azole resistance alleles that we document poses a probable threat to treatment success rates, warranting the enhanced surveillance of A. fumigatus genotypes in these patients. Furthermore, potential changes to antifungal first-line treatment guidelines may be needed to improve patient outcomes when CAPA is suspected.


Introduction
Aspergillus fumigatus is the predominant causative agent of invasive aspergillosis (IA) [1,2], with an estimated annual prevalence of > 250,000 cases globally and a mortality rate of 30 to 95% [3,4].Normally, only those who are immunocompromised or immunosuppressed are highly susceptible to developing IA [2]; however, there is a widening group of patients who are at risk of IA, including those with severe influenza or COVID- 19.
In 2019-2020, the novel virus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), causing COVID-19 disease, spread across the globe.Opportunistic pathogens have been widely reported as causing secondary infections in COVID-19 patients with lung damage and a notable proportion of these are fungal coinfections (12.6%) [5].A large proportion of fungal coinfections in COVID-19 cases are caused by Aspergillus species, including A. fumigatus, giving rise to COVID-19-associated pulmonary aspergillosis (CAPA) [6,7].The European Confederation for Medical Mycology (ECMM) and the International Society for Human and Animal Mycology (ISHAM) developed a consensus definition for CAPA at the end of 2020, stating that CAPA is a form of IA that is "in temporal proximity to a preceding SARS-CoV-2 infection" [7].The global prevalence of CAPA is reported to be between 3.8 to 40%, with 15.1% of ICU-admitted COVID-19 patients fulfilling the ECMM definition of CAPA [6][7][8][9][10].Furthermore, CAPA is characterised by low survival, with mortality ranging from 44 to 75% [10][11][12].The MYCOVID cohort study showed that COVID-19 patients who received intensive care treatment and positive sputum cultures for A. fumigatus but could not be classified as CAPA had higher mortality (45.8%) than those who had negative A. fumigatus cultures (32.1%) [10].However, this was lower than the group who had CAPA (61.8%) [10].Therefore, so-called colonisation may not be as harmless as previously thought.
In recent years, A. fumigatus resistance to the azole antifungal drug class has emerged, with its prevalence rapidly increasing, and recently has been declared a public health issue [13,14].In the environment, population genomic studies have determined that depending on the country, between 2.2 to 20% of isolates are azole-resistant, and this proportion is as high as 95.2% in Vietnam [15][16][17][18][19][20].A cause for concern is the increase in the proportion of azole-resistant A. fumigatus (ARAf ) identified in patients, as this is associated with treatment failure, increased mortality rates, and a doubling of health care costs [13,14].In CAPA, the prevalence of azole resistance has not yet been established, as the sample size of the studies has been too small.Two CAPA genomic epidemiology studies have reported no resistant isolates [21,22] while one German study reported 22.2% (n = 6/27) [23] as resistant.Using the Clinical and Laboratory Standards Institute (CLSI) method, a second German study (n = 4) did not find any resistant isolates [24].Finally, a Portuguese study reported a high prevalence of ARAf of 45.5% (n = 5/11 [25]. Genomic epidemiological methods have played a key role in delineating the genetic basis of ARAf [13,26,27].The primary locus, with a high-frequency of non-synonymous SNPs (nsSNPs), known to be involved in azole resistance is the cyp51A gene encoding the 14α-sterol demethylase enzyme [13,26].Common amino acid substitution hotspots are G54, L98H, G138, M220, and G448 (Table A1).Point mutations occur in isolates cultured from patients who have had long-term exposure to azole therapy [26].Additionally, several tandem repeats (TRs) in the gene promoter region of cyp51A lead to the overexpression of the gene and are commonly associated with point mutations in the cyp51A gene (Table A1).The genotypes of A. fumigatus that contain TR-mediated resistance are more common in environmental isolates and isolates obtained from azole-naïve patients.However, there is increasing awareness that resistance mechanisms in ARAf are complex and involve multiple genes (Table A1) [16,[26][27][28][29][30][31][32].In studies that identified ARAf in CAPA isolates, Kirchoff et al. [23] discovered only one CAPA isolate that contained TR 34 /L98H and five polymorphisms of ARAf via a non-cyp51A mechanism.
Whilst the development of a clinical definition for CAPA has aided clinicians in its diagnosis and ensuring treatment is commenced in a timely manner to improve patient outcomes, important questions remain as to the genetic characteristics and identity of A. fumigatus causing CAPA and the prevalence of antifungal resistance [13,[21][22][23][24]33].In A. fumigatus, genomic epidemiological methods have begun to unravel the genetics of environmental and clinical antifungal resistance and the potential mechanisms of dispersion [1,13,16,32].To date, there are only five genomic epidemiologic studies of CAPA using multiple methods [21][22][23][24][25].The results of these analyses were inconclusive on the genetic and epidemiological relatedness of A. fumigatus causing CAPA.The current study is the largest transnational genomic epidemiological investigation to date of CAPA isolates to determine where the genotypes of CAPA isolates group in the wider A. fumigatus population.Secondly, this study aimed to identify the mechanisms and determine the frequency of azole-resistant polymorphisms in CAPA isolates.

CAPA Definition
The definition of CAPA was based on the 2020 ECMM/ISHAM consensus criteria [7].In summary, patients must have a positive SARS-CoV-2 polymerase chain reaction (PCR), require intensive care for COVID-19 and have signs of invasive pulmonary aspergillosis (IPA) infection.Ideally, IPA is confirmed through the histopathological or direct microscopic detection of fungal hyphae obtained by lung biopsy, therefore, showing signs of tissue invasion and/or damage [7,34].However, samples from bronchoalveolar lavage, bronchial aspirate, and tracheal aspirate were used as alternatives, as lung biopsies could rarely be performed in CAPA patients [7].A positive SARS-CoV-2 PCR test must occur between two weeks prior to hospital admission and up to 96 h after ICU admission.CAPA may be further categorised depending on the sensitivity of the diagnostic method used.That is, CAPA can be either proven, probable, or possible [7,11].A fourth category was added in cases where A. fumigatus was isolated from the sputum of a patient with COVID-19 receiving intensive care, but there were no clear signs of invasive disease [10].Furthermore, there was no follow-up bronchoscopy to determine if there was an invasive disease.Therefore, for the first analysis, the patients were categorised as non-CAPA and only colonised with A. fumigatus ('colonising').In the second analysis, the colonising isolates were categorised as clinical non-CAPA isolates.

Fungal Isolates
Twenty-one CAPA isolates from four European countries between 2020-2021 were included (Table 1).In this study, all CAPA isolates were recovered as per the 'standard of care' based on the ECMM criteria [7].Four CAPA isolates (CAPA-A-D) were from four separate cases that originated from two Cologne hospitals in Germany [24,35].Six CAPA isolates were from London, UK (C422-C425, C611, C612).A further two isolates originated from two hospitals in the Netherlands (C403 and C408).A further nine isolates were from two hospitals in Dublin, Ireland (C434-C441, C444).Isolate C444 has previously been described in a case report [36].
In this study, the genetic relatedness of CAPA isolates was compared to (1) A. fumigatus isolates from non-CAPA IA cases, and isolates colonising non-CAPA patients; and (2) non-CAPA clinical and environmental isolates from the wider A. fumigatus population.The first analysis involved comparing CAPA isolates to two other groups.The first group comprised twelve A. fumigatus isolates from eleven patients with different types of IA (C120, C137-C140, C143, C307, C323, C360, C372, C376, C442).This included four isolates from a patient with necrotising aspergillosis, one trauma patient, two allergic bronchopulmonary aspergillosis (ABPA) patients, two with ABPA and asthma, and one with chronic pulmonary aspergillosis (CPA), and two with IPA (Table A2).The selection criteria for these IA isolates were based upon the European Organization for Research and Treatment of Cancer and the Mycoses Study Group definition for invasive disease [34], and were from the UK and Ireland.The second group comprised eight 'colonising' isolates: seven from three Dutch Hospitals (C402, C404-C407, C409, C410) and one from a patient in Ireland (C443) [32,37,38].The second analysis involved comparing 21 CAPA, 167 non-CAPA clinical, and 73 environmental A. fumigatus isolates.In summary, 218 UK and Irish A. fumigatus isolates that had been analysed by Rhodes et al. [32] and 23 additional A. fumigatus isolates (C307, C323, C360, C372, C376, C402, C404-C407, C409, C410, C426-C433, C442, C443) were used (Table A2).All 261 isolates can be found in the Microreact project [37] at https: //microreact.org/project/mPbPTWS2jTvvdNGmuDVyia-capa(accessed on 4 November 2023).

Antifungal Susceptibility Testing
Susceptibility to azole antifungal agents was carried out on all 21 of the CAPA isolates within this study.Isolates were first screened using the low-cost tebuconazole screening test (Tebucheck) to determine which isolates were resistant and then VIPcheck TM was used to give more information on which medical azoles the isolates were resistant to [39].Fifteen of the CAPA isolates were prepared as per the protocol set out in Brackin et al. [39].Briefly, isolates were inoculated in 25 cm 3 tissue culture flasks containing Sabouraud dextrose agar and culture for a minimum of 2 days at 45 • C. Conidia were collected in sterilised 0.05% (v/v) Tween-80 (Calbiochem ® , Sigma-Aldrich, Gillingham, UK using filtration through sterilised glass-wool.A final standardised spore suspension was achieved with a cell density reading at 600 nm between 0.09 and 0.12 using sterilised 0.05% (v/v) Tween-80 with a standard spectrophotometer [39].Twenty microlitres of final standardised spore suspensions were inoculated into Tebucheck multi-well and VIPcheck TM (Mediaproducts BV, Groningen, NL) plates.The Tebucheck multi-well plate consists of four wells comprising of drug-free (growth control well), 6 mg/L, 8 mg/L, and 16 mg/L of tebuconazole [39].All Tebucheck multi-well plates were made in-house as per the protocol set out in Brackin et al. [39].The VIPcheck TM multi-well plate consists of wells with 4 mg/L itraconazole, 2 mg/L voriconazole, 0.5 mg/L posaconazole, and a drug-free growth control well [40].Due to a lower concentration of spore suspension used, cultured plates were left at 45 • C to ensure there was complete growth in the control wells (48 h).Each strain was compared to a non-CAPA clinical isolate C154, which was EUCAST susceptible and verified by both Tebucheck and VIPcheck TM .
The scoring of Tebucheck plates is as follows: a well that had no growth (0 to <10% of the well) was scored 0; if there was partial growth (10 to 50% growth) the well was scored 0.5; and if there was full growth (50 to 100%) then the well was scored 1 [39].Any isolates with a score greater than 1 were considered resistant to tebuconazole [39].
The scoring of VIPcheck TM can be found in Buil et al. [40].Briefly, after 24 h, if there was uninhibited growth in any well which contained an azole, then the well would be scored 2 and the isolate would be resistant to that azole.If there was minimal growth, then this would be scored 1 and the isolate would therefore be partly resistant.If there was no growth in the azole-containing wells, then the well would be scored a 0 and the isolate would be considered susceptible to the azole.

Genomic DNA Preparation and Whole Genome Sequencing
Whole-genome libraries of the A. fumigatus isolates were prepared and sequenced by different research teams and at different times; thus, different platforms were used.Extraction of genomic DNA from the Irish and UK CAPA isolates (n = 15) and some additional Irish non-CAPA isolates (clinical, n = 2, and environmental, n = 8) was carried out at Imperial College London.Briefly, gDNA was extracted using the MasterPure TM Complete DNA and RNA Purification Kit (Lucigen, Middleton, USA), including an additional bead-beating step using a FastPrep-24 TM .Extracted gDNA was purified using a DNeasy Blood and Tissue Kit (Qiagen, Germany), with the concentration measured using a Qubit fluorometer and dsDNA Broad Range Assay kit (Invitrogen, ThermoFisher Scientific, Karnataka, India).DNA purity was assessed using NanoDrop TM spectrophotometry.Purified gDNA was stored at −20 • C prior to the construction of gDNA libraries, normalisation, and indexing (Earlham Institute, Norwich, UK).Libraries were run on a NovaSeq 6000 SP v1.5 flow cell (Illumina, Cambridge, UK) to generate 150 bp paired-end reads.

Bioinformatics Analysis
Raw Illumina whole-genome sequence (WGS) reads were quality checked using FastQC v0.11.9 (Brabham institute) and subsequently aligned to the Af293 reference genome using the Burrow-Wheeler Aligner alignment tool with maximal exact methods algorithm [44,45].The quality of the alignment was improved and converted to sorted BAM format using sequence alignment/map (SAM) tools v1.15.To minimise inaccurate identification of single-nucleotide polymorphisms (SNPs) due to PCR duplication error, duplicated reads were marked using Picard v2.18.7.SNPs were identified using 'Haplotype Caller' from the Genome Analysis Toolkit v4.0.To ensure high confidence calls, SNPs had to achieve at least 1 parameter, QD < 2.0, fisher strand > 60.0, mapping quality < 40.0, mapping quality rank sum test < −12.5, read positive rank sum test < −8.0, and SOR > 4.0.The above expressions have been rigorously tested and benchmarked [32].SNPs were mapped to genes using vcf-annotator v0.5 (Broad Institute).

A. fumigatus and MAT Identification
Previously, 223 UK and Irish A. fumigatus isolates have been confirmed to be members of the A. fumigatus species complex using molecular methods [32].In this study, the CAPA isolates were confirmed as A. fumigatus for congruency with Af293 using the nucleotide sequence of the calmodulin (CaM) gene from Af293, utilising the basic local alignment searching tool (BLAST, v2.13.0) from the National Centre of Biotechnology Information (NCBI) [46,47].A CAPA isolate was identified as A. fumigatus if the top BLAST hit had a percentage identification of >99% [48].

Phylogenetic and Spatial Analyses
Whole-genome SNP data were converted to the presence/absence of an SNP with respect to the reference.SNPs identified as low confidence during variant filtration were assigned as missing.In the first analysis, as there were fewer than 50 taxa, maximum likelihood (ML) phylogenies were constructed using the gamma model (GTRGAMMA) of rate heterogeneity and rapid bootstraps over 1000 replicates in RAxML v8.2.9 (Stematakis, 2006 RAxML-VI-HPC) [50].In the second analysis, with more than 50 taxa, maximum likelihood (ML) phylogenies were constructed using the CAT approximation (GTRCAT) of rate heterogeneity and rapid bootstraps over 1000 replicates in RAxML v8.2.9 (Stematakis, 2006 RAxML-VI-HPC) [50].Phylogenies were annotated and visualised in ggtree and ggtreeExtra in R (v4.2.2).
Genetic similarity and population allocation were investigated via principal component analysis (PCA) and discriminant analysis of principal components (DAPC) based on whole genome SNP data in R (v4.2.2).DAPC was performed using adegenet [51].To determine the number of principal components (PCs) to retain, the a-score method in the adegenet package was used.As the loadings of the PCs themselves are generally uninformative, to gain insight into how each isolate contributes to the cluster's composition, compoplots (adegenet) were generated.

Statistical Analyses
Descriptive analysis of data was conducted in R v4.2.2 and Microsoft Excel.Chi (χ 2 )test was performed to address the association between CAPA and azole-drug resistance.Significant value was considered for p-values < 0.05.Yates's continuity correction was used to calculate χ 2 in cases when values for a variable were less than 20.

WGS of 21 CAPA Isolates
The reference-guided methods were used to analyse the genetic diversity of twentyone CAPA isolates from four European countries (Table 1).All the isolates were confirmed to be A. fumigatus, sharing 100% similarity with the calmodulin (CaM) gene of the Af293 genome.All the sequenced genomes mapped > 96.2% (mean 97.8%) to the reference genome Af293 (Table 1), with a mean coverage of 151× (128-167×) for the German and Dutch CAPA isolates, and 42.1× (27.6-56.2×)for the UK and Irish isolates, reflecting the different sequencing methods and platforms used.The normalised whole-genome depth of coverage confirmed an absence of aneuploidy events in the CAPA isolates.However, deletions and duplications were observed; all the isolates observed had an approximately 260 kilo-base pairs (Kbp) deletion in Chromosome IV (Figure A1).This phenomenon has been observed in previous WGS studies, and the region of the rRNA repeat cluster may be included in this deletion [44,52].Deletions were also found in Chromosome I (region of > 200 Kb), VII (region of > 300 kb), and VIII (approx.60 Kb region), in 7 (24.1%),14 (48.3%), and 8 (27.5%) isolates, respectively (Figure A1).These regions have not been identified as having any known importance in the context of azole resistance.Multiple peaks were observed in all the chromosomes from each isolate, which relate to the presence of copy number variations (CNVs).
An even proportion of mating idiomorphs in the CAPA isolates was observed, with 52% (n = 11/21) containing the MAT1-1 gene and 48% (n = 10/21) containing the MAT1-2 gene.This is consistent with what is expected in the wild, and is suggestive of a sexually recombining population [49,53].Combining the CAPA isolates with the UK and Ireland A. fumigatus isolates, there was a bias towards the MAT1-2 idiomorph (60.9%), which was statistically significant (χ 2 -test, p-value = 0.014, degrees of freedom, d.f.= 2).This means that the population may be more likely to reproduce asexually [32,49,53].Although, there was no significant difference between source types (Table A3).
On average, the CAPA isolates differed from each other by 26,706 SNPs.There was only one CAPA isolate pair (C439-C440) that was highly similar (< 2671 SNPs); however, these were obtained from different patients.When the CAPA isolates were compared to the IA and colonised isolates, they differed by 30,986 SNPs.

Azole Resistance within CAPA Isolates Primarily Centred on Known Polymorphisms within cyp51A
Three CAPA (C438, C441, C444) isolates obtained from different locations in Ireland contained azole-resistant polymorphisms (TR 34 /L98H) (Table 2, Table A4, and Table A5).No known drug-resistant polymorphisms were identified in the Dutch, UK, or German CAPA isolates.C154 was used as a control, as it had previously been shown to be EUCAST susceptible [32] and in this study, this confirmed by Tebucheck and VIPcheck TM .VIPcheck TM score is 2 for uninhibited growth, 1 for minimal growth in well, and 0 for no growth.The wells contain either ITR 4mg/L, POS 0.5 mg/L, or VOR 2mg/L [40].
Susceptibility testing was performed to confirm the isolates containing azole-resistant polymorphisms have raised MICs to azole drugs.Of the CAPA isolates, three had scores of greater than 1 in VIPcheck TM and/or Tebucheck (C438, C441 and C444) (Table 2).Thus, three of the 21 CAPA isolates were ARAf.The remaining CAPA isolates scored 1 in VIPcheck TM and/or Tebucheck, and, therefore, are susceptible phenotypically.These results were verified independently using EUCAST and the CLSI broth microdilution methods (Tables A4 and A5) [24,32,36].

Phylogenetic and Spatial Analysis Shows CAPA Isolates Are Highly Related to Non-CAPA Clinical and Environmental A. fumigatus
To test the phylogenetic relatedness of A. fumigatus cultured from CAPA patients with the existing collection of A. fumigatus from IA and colonised patients, the phylogenetic analysis of twenty-one CAPA, twelve non-CAPA IA isolates, and eight non-CAPA colonising isolates from four European countries was conducted (Table A2).This analysis revealed two broadly divergent clades, which are referred to as clade A and clade B, as previously reported (Figures 1 and A2) [32,54].Of the CAPA isolates, the majority (n = 15, 71.4%) were in clade B, and the remainder were in clade A (n = 6, 28.6%) (Figure 1 and Table A3).All CAPA ARAf isolates (C438, C441, C444) were situated in clade A. All the Dutch CAPA isolates were from clade B, whereas only 55.6% of the Irish CAPA isolates were from clade B. The UK and German CAPA isolates had 84.3% and 75% of isolates in clade B, respectively.Of the IA isolates (C120, C137-C140, C143, C307, C323, C360, C372, C376, C442), eight (66.6%) were in clade A and four (33.3%) were in clade B. The IA cohort contained ten ARAf isolates, which all contained cyp51A polymorphisms.The eight IA isolates in clade A were ARAf and two (C120 and C372) of the four IA isolates in clade B were ARAf isolates (Figure 1 and Appendix Tables A4 and A5).Finally, seven of the eight colonising isolates comprised clade B and all the isolates where not ARAf.
Multivariate methods were used to identify and describe the genetically related clusters.The principal component analysis (four principal components, PCs) identified that the CAPA isolates come from the same population as the IA isolates (Figure 2A).Interestingly, the colonising isolates formed a subsection of the CAPA isolates (Figure 2A).The discriminant analysis of principal components (DAPC) and composition plot (five PCs) confirmed this observation (Figure 2B,C).All the isolate genotypes contained a mixture of CAPA, IA, and coloniser, with twenty-one of the isolates containing >20% CAPA genome (Figure 2B,C).Despite this, the isolates labelled as IA or coloniser are not genetically distinct from the CAPA isolates.Multivariate methods were used to identify and describe the genetically related clusters.The principal component analysis (four principal components, PCs) identified that the CAPA isolates come from the same population as the IA isolates (Figure 2A).Interestingly, the colonising isolates formed a subsection of the CAPA isolates (Figure 2A).The discriminant analysis of principal components (DAPC) and composition plot (five PCs) confirmed this observation (Figure 2B,C).All the isolate genotypes contained a mixture of CAPA, IA, and coloniser, with twenty-one of the isolates containing >20% CAPA genome (Figure 2B,C).Despite this, the isolates labelled as IA or coloniser are not genetically distinct from the CAPA isolates.To gain a more in-depth understanding of the genetic relatedness of the CAPA isolates and whether they represent a distinct genetic identity or are drawn from the wider population of A. fumigatus, the phylogenetic relationship of the CAPA isolates with the wider A. fumigatus population was investigated.A phylogenetic analysis was conducted on 240 clinical and environmental isolates and twenty-one CAPA isolates (Figure A4).Again, two broadly divergent clades were observed: clades A and B (Figures 3, A3 and A4).
A multivariate analysis of the underlying genomic structure of the isolates confirmed that the genomes of the CAPA isolates could not be distinguished from the non-CAPA clinical and the environmental A. fumigatus isolates originating from the UK and Ireland (Figure 4).Composition plot highlights that all the isolates' genotype is composed of genetic material identified as CAPA, IA, or colonising.Four isolates contain 70% control compared to CAPA, whereas there are 21 isolates the genome membership of which is mostly CAPA (>80%).
To gain a more in-depth understanding of the genetic relatedness of the CAPA isolates and whether they represent a distinct genetic identity or are drawn from the wider population of A. fumigatus, the phylogenetic relationship of the CAPA isolates with the wider A. fumigatus population was investigated.A phylogenetic analysis was conducted on 240 clinical and environmental isolates and twenty-one CAPA isolates (Figure S4).Again, two broadly divergent clades were observed: clades A and B (Figures 3, S3 and S4).
A multivariate analysis of the underlying genomic structure of the isolates confirmed that the genomes of the CAPA isolates could not be distinguished from the non-CAPA clinical and the environmental A. fumigatus isolates originating from the UK and Ireland (Figure 4).

Discussion
The COVID-19 pandemic has created a large and growing global cohort of patients at risk of developing A. fumigatus coinfections.Additionally, CAPA has higher mortality

Discussion
The COVID-19 pandemic has created a large and growing global cohort of patients at risk of developing A. fumigatus coinfections.Additionally, CAPA has higher mortality than COVID-19 itself (16 to 25% excess mortality rate) [7].This, in part, has been due to the challenges of determining whether COVID-19 patients have IA [7,10].Therefore, delays in commencing prompt antifungal treatment may have led to higher mortality rates [55].Furthermore, WGS could address these issues by providing information on potential common sources, the genetic relatedness of the CAPA isolates with the wider A. fumigatus population, and the presence of ARAf, informing clinicians about the most effective way to treat CAPA patients [1,13].In this study, WGS was carried out on twentyone CAPA isolates from four European countries to explore the genomic epidemiology of A. fumigatus causing CAPA.
The present genomic analysis of CAPA isolates yielded three major findings.First, the CAPA isolates comprised a diverse range of A. fumigatus genotypes.Secondly, the CAPA genomes are an even mix of A. fumigatus genotypes found in previous studies in either the clinical environment or the environment.Finally, our study provides evidence of ARAf in at-risk patient cohorts.However, an important caveat is that the only ARAf found were from one country, likely due to the overall small sample size of A. fumigatus studied.
To gain insight into CAPA's genetic and epidemiological relatedness, twenty-one CAPA isolates were compared to the different types of A. fumigatus samples including clinical non-CAPA and environmental isolates.This analysis provided evidence that the genotypes of the A. fumigatus isolates from the CAPA patients are genetically diverse, with two broadly divergent clades: A and B. A proportion of the CAPA isolates clustered in clade B (n = 15, 71.4%).In the sample of the IA isolates, 64% were in clade A. The phenomenon of the two-clade structure of A. fumigatus species was first identified in a large global genetic epidemiology study of over 4000 A. fumigatus isolates and has since been replicated in different A. fumigatus populations [32,54,56].
Even though the genotypes of the CAPA isolates in this study were biased towards clade B, the isolates were genetically diverse.The genetic relatedness between the CAPA isolates is on average 26,706 SNPs.The genetic relatedness between the isolates was higher when compared across the twelve IA and eight colonising isolates (average 30,986 SNPs) and the wider A. fumigatus population (25,448 SNPs), comparable to previous studies [32,56].Only one pair of isolates was highly related (<2671 SNPs; C439-C440) and originated from the same country, but were from different patients, suggesting the patients were exposed to the same environmental source.However, no epidemiological information was available to investigate this possibility.
Secondly, using hypothesis-free population genetic methods, we identified that the CAPA isolates were a mixture of A. fumigatus genotypes.The genotypes from the 21 CAPA isolates showed genetic overlapping with the A. fumigatus genotypes isolated from the patients with IA, and those obtained from other patients and the environment.In the five previous genomic epidemiological studies of CAPA, the genotypes from the CAPA isolates were only compared with non-CAPA clinical isolates or COVID-19 patients who were colonised with A. fumigatus and had no active IA [21][22][23][24][25].The results of Steenwyk et al. suggest that overall, the CAPA genotypes are drawn from the wider clinical population of A. fumigatus isolates, despite observing some genetic clustering [24].This has been replicated in a larger transnational genomic study [21].A larger multicentre study using microsatellite methods to genotype isolates identified that the genotypes obtained were genetically distinct from one another [23].This finding was replicated in one Spanish and one Portuguese study using tandem repeats within the exons of surface proteins and erg coding genes (TRESPERG), and microsatellite methods, respectively, to compare isolates obtained from CAPA patients and the environment [22,25].The authors identified that the CAPA isolates were from a diverse genetic pool [22,23,25].However, compared to WGS, TRESPERG and microsatellite methods are limited in their power to provide an in-depth analysis of the genetic structure and diversity within a population [57,58], due to only utilising a small proportion of the whole genome [58].Therefore, this study is the first to conduct an in-depth analysis to delineate the genetic diversity of A. fumigatus in CAPA cases.Furthermore, to show that the genotypes of A. fumigatus obtained from patients with CAPA are a genetically diverse mixture of clinical and environmental A. fumigatus genotypes, WGS is best placed to investigate [27,57].
COVID-19 patients who develop CAPA may have acquired A. fumigatus isolates from the environment.This is supported by the evidence that the genotypes of the CAPA isolates in this study are a mixture of genomes representative of both clinical and environmental A. fumigatus isolates.Previously, studies have identified that clinical non-CAPA A. fumigatus isolates are genetically similar to those sourced from the environment [22,25,32,56].Additionally, ARAf has been isolated from several azole-naïve IA and CAPA cases [36,59,60] suggesting their resistance was pre-acquired from environmental (community and hospital) inocula.It has been hypothesised that patients who develop IA are first colonised with A. fumigatus isolates from airborne conidia in their environment.Thus, CAPA patients may acquire and become colonised with A. fumigatus on exposure to airborne A. fumigatus conidia in either the community or from the hospital environment.These conidia, under favourable conditions (immunosuppression from SARs-CoV-2 and drugs used to treat COVID-19, such as dexamethasone and tazoluzimab), develop hyphae and subsequently invade the surrounding tissue, leading to the onset of CAPA.If the patient is being exposed or treated with triazoles, it is at this stage that resistant inocula will remain in the patient's airway.
The present study used bioinformatic tools and drug susceptibility tests to identify three ARAf out of 21 CAPA isolates containing the predominant cyp51A polymorphism TR 34 /L98H and that were phenotypically resistant to at least one azole.Furthermore, all three were pan-azole-resistant [36].No further CAPA isolates had known azole-resistant polymorphisms and were susceptible to azoles phenotypically.A limitation of the current study is the screening tests to infer if an isolate is potentially susceptible or resistant.Thus, a robust and standardised approach (e.g., EUCAST broth microdilution method) is required to give the accurate and valid phenotyping of the isolates [26,39,41].In a large Spanish tertiary hospital of 28 CAPA patients, the ARAf prevalence was zero [22].This is despite ARAf being previously isolated from patients throughout Spain [61].Similarly, a recently published small transnational study (n = 11) and a small German study (n = 4) did not identify any ARAf in the sample of CAPA isolates [21,24].In a German multicentre study, the prevalence of ARAf isolates was found to be 22.2% (n = 6/27) [23].Only one of the six CAPA isolates was identified to have known cyp51A polymorphisms conferring resistance [23].In the Netherlands, a screening programme of CAPA patients identified one patient in a cohort of twenty-two to have an ARAf strain (4.5%) [62].
In the wider patient population, the prevalence of ARAf has been increasing in the last 30 years from 0.43% to 2.2% in London [63].Different countries have recently reported varying levels: Germany at 3.5%, Denmark at 6.1%, Spain at 7.4%, Netherlands at 11.7%, and Japan at 12.7% [17,20,61,64,65].In the UK, the prevalence of ARAf in the air was recently found to range from 3 to 9%, depending on the season [16,19].Thus, the consideration of ARAf surveillance in CAPA and wider patient populations who may be exposed to Aspergillus conidia in the hospital air could be recommended owing to this near-ubiquitous exposure.However, this is not a cause for changing patient treatments.
There are some limitations to this study.During the COVID-19 pandemic, there were challenges in diagnosing CAPA.Therefore, relatively few patients were diagnosed with CAPA, so the sample size of this study was small.The small sample size of the CAPA isolates prevents the drawing of strong conclusions.Thus, future studies should include larger sample sizes.Secondly, it would have been useful to have clinical indicators and outcomes for the CAPA isolates to analyse any associations between the genotype and clinical phenotypes.This would help with understanding CAPA and improve patient outcomes.Finally, ARAf in CAPA was only identified in isolates from one country, owing to a small sample size.Thus, conclusions about the ARAf prevalence should be taken with caution.Ensuing studies with larger sample sizes from multiple countries across the globe would be better suited to determine the overall prevalence of ARAf in CAPA.

Conclusions
The present study is the largest transnational epidemiological analysis of the genomic relationship between CAPA isolates and the wider A. fumigatus population.We demonstrate that the genomes of the CAPA isolates comprised a diverse mixture of genotypes from the wider A. fumigatus population.The CAPA genomes are similar to those found in other clinical and environmental A. fumigatus, and are not composed of a unique sub-population.Finally, ARAf is identified in this at-risk group of patients.
Future directions should focus on developing user-friendly and easily accessible surveillance methodologies for assaying A. fumigatus for both clinical and environmental settings.Surveillance programmes should include methods of monitoring antifungal resistance, owing to the observed prevalence of resistance to frontline clinical azoles (e.g., itraconazole, voriconazole, isavuconazole, and posaconazole).Furthermore, exposure assessments could be used to aid clinical decisions on the most effective course of antifungal treatment.Finally, a review of local antifungal treatment guidelines in response to the increasing prevalence of ARAf needs to be further considered.

Appendix A Appendix
Table A1.Polymorphisms reported in the literature that are associated with azole resistance in A. fumigatus, were used to search for azole resistance in the CAPA isolates in this study.

Figure 1 .
Figure 1.Phylogenetic tree of isolates from CAPA, IA, and colonising patients.Unrooted maximum likelihood (ML) phylogenetic tree over 1000 replicates performed on WGS SNP data, showing the following: inner track 1, the source of the isolate (CAPA, colonising or IA); middle track 2, if the isolate contains cyp51A polymorphism; and outer track 3, the clade in which the isolate is located.

Figure 1 .
Figure 1.Phylogenetic tree of isolates from CAPA, IA, and colonising patients.Unrooted maximum likelihood (ML) phylogenetic tree over 1000 replicates performed on WGS SNP data, showing the following: inner track 1, the source of the isolate (CAPA, colonising or IA); middle track 2, if the isolate contains cyp51A polymorphism; and outer track 3, the clade in which the isolate is located.

Figure 2 .
Figure 2. Multivariate analysis of isolates from CAPA, IA, and colonising patients.(A) Scatterplot of the principal component analysis (PCA) A. fumigatus genotypes using the first four principal components (PCs) illustrating the genetic identity of CAPA and IA control isolates.(B) Density plot of the discriminant PCA (DAPC), broadly identifying two distinct clusters, CAPA and control.(C)Composition plot highlights that all the isolates' genotype is composed of genetic material identified as CAPA, IA, or colonising.Four isolates contain 70% control compared to CAPA, whereas there are 21 isolates the genome membership of which is mostly CAPA (>80%).

Figure 3 .
Figure 3. Phylogenetic tree of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Unrooted ML phylogenetic tree with over 1000 replicates performed on WGS SNP data, showing the following: outer track 1, if the isolate contains cyp51A polymorphism; middle track 2, the clade in which the isolate is located; and inner track 3, the source of the isolate (CAPA, clinical, or environmental).

Figure 3 .
Figure 3. Phylogenetic tree of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Unrooted ML phylogenetic tree with over 1000 replicates performed on WGS SNP data, showing the following: outer track 1, if the isolate contains cyp51A polymorphism; middle track 2, the clade in which the isolate is located; and inner track 3, the source of the isolate (CAPA, clinical, or environmental).

Figure 3 .
Figure 3. Phylogenetic tree of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Unrooted ML phylogenetic tree with over 1000 replicates performed on WGS SNP data, showing the following: outer track 1, if the isolate contains cyp51A polymorphism; middle track 2, the clade in which the isolate is located; and inner track 3, the source of the isolate (CAPA, clinical, or environmental).

Figure 4 .
Figure 4. Multivariate analysis of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Multivariate analysis, (A) scatterplot of the PCA A. fumigatus genotypes using the first five PCs illustrating the genetic identity of CAPA isolates, and clinical and environmental isolates from Ireland and the UK.(B) Scatterplot of the DPCA identifying 3 types form separate clusters, with CAPA and clinical with some overlap.(C) A composition plot comparing the genetic composition of each isolate.The plot highlights that each isolate is a mixture of genotypes identified as CAPA, environment, and clinical non-CAPA.

Figure 4 .
Figure 4. Multivariate analysis of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Multivariate analysis, (A) scatterplot of the PCA A. fumigatus genotypes using the first five PCs illustrating the genetic identity of CAPA isolates, and clinical and environmental

Figure A1 .Figure A1 .
Figure A1.Normalised depth of coverage by locus indicative of duplications and deletion events.(a) C440 Chromosome 1.The deleted region is also found in C408 and C439-C441.(b) C440 chromosome 4. Two regions, a 500 kbp deletion identified in all isolates, and a 1300 kbp deletion Figure A1.Normalised depth of coverage by locus indicative of duplications and deletion events.(a) C440 Chromosome 1.The deleted region is also found in C408 and C439-C441.(b) C440 chromosome 4. Two regions, a 500 kbp deletion identified in all isolates, and a 1300 kbp deletion around the centromere of Chromosome 4. (c) C440 Chromosome 7. Two deletions in Chromosome 7, relating to the centromere and a region of 1600 kbp found in CAPA-A, CAPA-B, C403, 408, C436 and C439-C441, 444.(d) C444 Chromosome 8. Small deletion of 1400 kbp found in Chromosome 8 in isolates CAPA-B, CAPA-C, C424, C436, C441, 444.

Figure A2 .Figure A2 .
Figure A2.Phylogenetic tree of CAPA isolates, A. fumigatus IA control isolates from London, UK, and isolates from colonising COVID-19 patients (Netherlands and Ireland).Rooted ML phylogenetic tree with bootstraps support over 1000 replicates performed on WGS SNP data.Branch length represents the average number of SNPs.* 0.04 = 3242 SNPs.

Figure A3 .
Figure A3.Phylogenetic tree of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Rooted ML phylogenetic tree, showing: right track 1, if the isolate contains cyp51A polymorphism; middle track 2, the clade the isolate is located; and left track 3, the source of the isolate (CAPA, clinical, or environmental).Branch length represents the average number of SNPs.* 0.1 = 33,839 SNPs.

Figure A4 .
Figure A4.Phylogenetic tree of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Rooted ML phylogenetic tree with bootstraps support over 1000 replicates performed on WGS SNP data.Branch length represents the average number of SNPs.*0.1 = 33,839 SNPs.

Figure A4 .
Figure A4.Phylogenetic tree of CAPA and A. fumigatus clinical and environmental isolates from Ireland and the UK.Rooted ML phylogenetic tree with bootstraps support over 1000 replicates performed on WGS SNP data.Branch length represents the average number of SNPs.* 0.1 = 33,839 SNPs.

Table 1 .
Characteristics of CAPA isolates of A. fumigatus used in this study and details of alignments.

Table 2 .
Azole drug susceptibility of CAPA isolates grown in minimal media and the associated candidate polymorphisms.

Table A2 .
Characteristics of additional clinical and environmental isolates of A. fumigatus used in this study to provide context for CAPA isolates, and details of WGS alignments.

Table A3 .
Relative frequencies of clade assignment, mating type idiomorph, and cyp51A polymorphisms for CAPA, clinical, and environmental A. fumigatus isolates in this study.

Table A4 .
Azole drug susceptibility of CAPA, clinical, and environmental A. fumigatus isolates grown in minimal media and the associated candidate polymorphisms.

Table A5 .
Azole drug susceptibility of A. fumigatus isolates used in study grown in minimal media and the associated polymorphisms.